Metabolic Mechanisms and Regulation of Mixed Culture Fermentation Robert Donald Hoelzle

Bachelor of Science (with Distinction) in Chemical and Biomolecular Engineering

The Ohio State University

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2016

School of Chemical Engineering

Advanced Water Management Centre Abstract

Fermentation of waste streams is an emerging method for production of biofuels and commodity chemicals, and mixed-culture fermentation allows for robustness and flexibility in the range of waste streams that can be processed, and in varying product mix from the fermenter. However, mechanistic models of mixed culture fermentation are limited, with mixed culture fermentation generally resulting in acetate, butyrate, and ethanol, with even these production modes are not clearly understood. Results from pure-culture fermentation, as well as those from methanogenic reactors indicate the possibility to produce high-value lactate and propionate if the cellular mechanisms which control the expression of these production modes can be harnessed and understood. Lactate and propionate are key chemicals in production of bioplastics, pesticides, preservatives, and food additives, among many other uses. Their production by mixed cultures has been observed at high substrate concentrations (shock loads), and inconsistently at high pH. They are produced via metabolic pathways which branch off of the primary energy-generation pathways. In general, they are thought to be produced by microbes as a way to redirect excess carbon flow in transient states, and are also thought to be produced as an electron sink. In this work, both of these possible production influences are explored for steady state production of lactate and propionate.

By increasing substrate from limiting to non-limiting quantities, and thereby increasing the carbon load on a fermentation system, lactate production was successfully achieved and maintained during steady state only when substrate was in excess. Substrate consumption at this point was 14.1±0.8 gCOD·L-1, and lactate accounted for 25±1% of the total product COD. Upon return to substrate-limiting conditions, the process returned to butyrate production. This demonstrates the reversibility of this production mode. Taxonomic analysis via Illumina pyrosequencing of the 16S rRNA gene revealed that the microbial community experienced a shift from Megasphaera-dominant during butyrate production to Ethanoligenens-dominant during lactate production. Taxonomy was also regained upon return to substrate-limiting conditions. However, taxonomy alone could not explain the product shift because Ethanoligenens produces little to no lactate in cultured references. A mechanistic description of the product shift therefore required molecular analysis of the metabolic mechanisms, and this was achieved through metaproteomics via SWATH-MS. The evidence suggests that at high carbon loads, toxic methylglyoxal is formed in many microbe groups, including Megasphaera, due to limited processing capacity of the

i dihydroxyacetone-P intermediate of glycolysis. The methylglyoxal is converted into lactate for disposal. This detoxification need is abated upon return to substrate-limited conditions.

The impact of redox stress was examined using cathodic electro-fermentation with soluble mediator chemicals ranging from slightly negative (-119 mV vs SHE) to near that of the intracellular NAD(P) and Fd pools (-394 mV). Propionate production was successfully achieved and maintained during steady state at 17±3% of total product COD (5.68±0.02 gCOD·L-1 substrate consumed) using methyl viologen as a mediator. However, minimal cathodic current was measured in the system. Taxonomic analysis revealed a gradual change from Megasphaera to Pectinatus with mediators of increasingly negative redox potential. This corresponded with a shift from butyrate production to propionate production. Pectinatus is a known propionate producer, and so this describes the product spectrum, but not the switching mechanism. Metaproteomic analysis revealed increased expression of both oxidative stress enzymes and electron mediation in mediated systems compared to non-mediated systems. Oxidative stress response in particular appears to be the causative factor for the change in product spectrum, and this could explain why others have also reported very low current compared to the reduction of the product spectrum. Additionally, expression of oxidative stress response enzymes was inversely related to expression of propionate pathway enzymes (p < 0.01). This mechanism could explain the occasional observation of propionate formation at high pH, which is also inversely related to oxidative stress response. It is hypothesized that in general, propionate formation is repressed by oxidative stress. However, in this system, propionate formation was due to the inhibition of bacteria other than Pectinatus.

In general, it is proposed that mixed culture lactate production is a response to high carbon loads. This includes transient states, such as exponential growth and inter-steady-state periods, where the substrate supply is greater than the biomass processing capacity. By contrast, propionate production is an electron mediation response at low oxidative stress. Transiently increased substrate loads may also induce propionate formation via the same mechanism due to increased NAD reduction from increased substrate processing. However, at steady state this effect appears to diminish in favor of lactate production. This may occur due to electron rebalancing to other reduced products, such as ethanol. These conclusions are not substantively linked, and not related to the simpler (and more intuitive) models previously proposed. In this thesis, while there are still limitations in database information, the metaproteomic approach was critical to identify functional mechanisms in comparison with the more traditional taxonomic approaches.

ii Declaration by Author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my research higher degree candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis.

iii Publications During Candidature

Peer-reviewed papers:

Hoelzle, R. D., Virdis, B. & Batstone, D. J. Regulation mechanisms in mixed and pure culture microbial fermentation. Biotechnol. Bioeng. 111, 2139–2154 (2014). doi:10.1002/bit.25321

Mohd-Zaki, Z., Bastidas-Oyanedel, J., Lu, Y., Hoelzle, R., Pratt, S., Slater, F., & Batstone, D. Influence of pH Regulation Mode in Glucose Fermentation on Product Selection and Process Stability. Microorganisms 4, 2 (2016). doi:10.3390/microorganisms4010002

Conference presentations:

Hoelzle, R. D., Virdis, B. & Batstone, D. J. Regulation Mechanisms in Mixed and Pure Culture Microbial Fermentation. The University of Queensland Postgraduate Engineering Conference. 3rd of June 2013. Brisbane (QLD, Australia) (Oral Presentation)

Hoelzle, R. D., Virdis, B. & Batstone, D. J. Assessing Metabolic Response to Increased Organic Loading Rate in Mixed-Culture Fermentation of Waste Water. Metabolic Engineering X. 15 to 19 June, 2014. Vancouver (BC, Canada). (Poster Presentation)

Hoelzle, R., Puyol, D., Huelsen, T., Virdis, B. & Batstone, D. Sludge Fermentation: On-Site Production of VFAs for PPB Nutrient Recovery. CRC for Water Sensitive Cities. 21 to 22 October, 2014. Melbourne (VIC, Australia). (Poster Presentation)

Puyol, D., Huelsen, T., Barry, E., Hoelzle, R., Lu, K. & Batstone, D. J. The partition-release- recovery concept for domestic wastewater treatment: from theoretical approach to practical conceptualization. XXXV Biennial Meeting of the Spanish Royal Society of Chemistry. 19 to 23 July, 2015. A Coruña (Spain). (Oral Presentation)

iv Publications Included in this Thesis

Hoelzle, R. D., Virdis, B. & Batstone, D. J. Regulation mechanisms in mixed and pure culture microbial fermentation. Biotechnol. Bioeng. 111, 2139–2154 (2014). doi:10.1002/bit.25321

This review paper has been updated and included as part of Chapter 1. The original paper is attached as Appendix A.

Contributor Statement of Contribution

Robert Hoelzle (Candidate) Designed the study (70%)

Wrote the paper (100%)

Bernardino Virdis Designed the study (10%)

Critically reviewed and edited paper (40%)

Damien Batstone Designed the study (20%)

Critically reviewed and edited paper (60%)

v Contributions by Others to the Thesis

This thesis includes the reporting of some important contributions made by other researchers that I have collaborated with throughout the duration of my PhD. These contributions are acknowledged as follows:

• Prof. Damien Batstone, Dr. Bernardino Virdis, and Dr. Daniel Puyol, who worked closely with me as advisors and shared with me their technical knowledge, and contributed in experimental and system design and in data interpretation • Nicola Angel from the Australian Centre for Ecogenomics (ACE, UQ) for performing the 16S rRNA Illumina pyrosequencing • Dr Amanda Nouwens, from the School of Chemical and Molecular Biosciences (SCMB, UQ) for assistance with proteomic procedures and the operation of the mass spectrometer, as well as initial interpretation of proteomic spectra

Statement of Part of the Thesis Submitted to Qualify for the Award of Another Degree

None.

vi Acknowledgements

One constant in any PhD is that the candidate must go through substantial personal and professional growth through the course of their studies, and this can in no way be accomplished alone. I would therefore be remiss to not acknowledge and thank those who have helped me along the way. First and foremost, I must thank my advisory team: Prof Damien Batstone, Dr Dino Virdis, and Dr Dani Puyol. Damien, thank you for giving me this opportunity and for supporting and encouraging me in my research endeavors. Dino, I have truly enjoyed working with you, and I would not have made it very far without all of our hours- long discussions. Dani, you have been a great friend, and I have become a much better scientist because you pushed me to improve and justify my writing and my lab methods.

I would also like to thank Paul Jensen, who took it upon himself to act as a mentor to me, both in working through my PhD, and in my broader professional development. You got me to think about where I want to go with my career, and how to get the most out of my PhD in order to start down that path, not to mention that you helped me line up my next step along the way. I will be forever grateful for your generosity and our friendship.

And to my first mentor, my 7th and 8th grade science teacher who first encouraged me to think critically and analytically about the world and its countless and amazing phenomena. Thank you Ms Hickey for inspiring me in the sciences!

Outside of advisors and mentors, I must also acknowledge and thank those who helped me learn the analytical techniques which were so important to my work. Kenn Lu, thank you for all your help in learning taxonomic analysis. I think you made time to help me with QIIME at least 3 times alone! Also, Christy Grobbler and Amanda Nouwens, thank you for helping me to learn and implement proteomic analysis. I enjoyed this part of my work most of all.

While I was generally familiar with basic statistical techniques when I began my PhD, I must thank my good friend Tegan Cruwys for helping me to understand how to properly implement and interpret principal component analysis. This ended up being a relatively small, but nonetheless vital part of the overall analysis of this thesis, but it created quite a few headaches while trying to grapple with it. You took what was at first for me an enigma of data manipulation, and made it perfectly understandable.

Thank you also to all of my friends at AWMC and in Brisbane. Overall, I’ve had an amazing 4 years and made some incredible life-long friends. A big thank you especially to the

vii Wednesday Morning Breakfast Club, without whom I probably would have lost my mind at some point.

Finally, and most importantly, thank you to my family for all of your love and support. Especially to my wife, Rachel. You were always there to support me through all of the late nights in the lab, the reactor failures, and the frustrations with data analysis, and you were always first to celebrate with me after any success. It takes an incredible person to stick with someone through all of the stress and anxiety of a PhD, and incredible doesn’t even begin to describe you!

viii Keywords bio-production, biochemicals, fermentation, mixed-culture, metabolic regulation, electron mediation, energy metabolism, waste management, metaproteomics, pyrosequencing

Australian and New Zealand Standard Research Classifications (ANZSRC)

100303 Fermentation, 40%

060104 Cell Metabolism, 40%

090703 Environmental Biotechnology, 20%

Fields of Research (FoR) Classification

1003 Industrial Biotechnology, 40%

0601 Biochemistry and Cell Biology, 40%

1002 Environmental Biotechnology, 20%

ix Table of Contents

Abstract ...... i

Declaration by Author ...... iii

Publications During Candidature ...... iv

Publications Included in this Thesis ...... v

Contributions by Others to the Thesis ...... vi

Statement of Part of the Thesis Submitted to Qualify for the Award of Another Degree ..... vi

Acknowledgements ...... vii

Keywords ...... ix

Australian and New Zealand Standard Research Classifications (ANZSRC) ...... ix

Fields of Research (FoR) Classification ...... ix

Table of Contents ...... x

Table of Figures ...... xiv

List of Tables ...... xix

List of Abbreviations ...... xxi

1. Introduction ...... 1-1

1.1. Review of Metabolic Pathways ...... 1-3

Primary Energy Pathways ...... 1-5 Redox Balancing through Electron Mediation ...... 1-5 Pyruvate Oxidation ...... 1-8 Acetyl-CoA Branches ...... 1-10 Chain Elongation ...... 1-11 Lactate Branches ...... 1-13 Dicarboxylic Acid Cycle ...... 1-14 Alternative Pyruvate Branches ...... 1-15 Glycolysis Side-Branches...... 1-16 1.2. Feedback Mechanisms of Electron Mediation ...... 1-18

Repressor Enzymes ...... 1-18 NADH:NAD+ Response to Environmental Factors ...... 1-19 Response to Changes in External Redox State ...... 1-20 1.3. Environmental Influences on Metabolic Regulation ...... 1-20

Regulation with pH ...... 1-20

x Regulation through Substrate Supply ...... 1-21 Regulation of the Redox Environment ...... 1-22 Regulation with Substrate Type ...... 1-25 Regulation with Temperature ...... 1-26 2. Thesis Motivation ...... 2-27

2.1. Research Objectives ...... 2-27

3. Research Methods ...... 3-30

3.1. Media and Culture ...... 3-30

Varied Substrate Concentration Media ...... 3-30 Electro-fermentation Media ...... 3-31 Culture...... 3-32 3.2. Bioreactors ...... 3-33

Varied Substrate Concentration ...... 3-33 Electro-fermentation ...... 3-33 Reactor Sterilization and Preparation ...... 3-33 Biofilm Removal ...... 3-35 3.3. Controls and Process Data ...... 3-36

Flow Rate ...... 3-38 Temperature ...... 3-38 pH ...... 3-39 Set Potential ...... 3-39 3.4. Analytical Methods ...... 3-39

Headspace Gas ...... 3-39 Chemical Analysis ...... 3-40 Biomass ...... 3-40 Chemical Oxygen Demand ...... 3-41 Cyclic Voltammetry ...... 3-41 16S rRNA Amplicon Analysis ...... 3-42 Metaproteomics ...... 3-42 3.5. Calculations ...... 3-43

Flowrates ...... 3-43 Theoretical Yields ...... 3-43 Kinetic Calculations ...... 3-44 Bioinformatics ...... 3-45 Statistics ...... 3-46 4. Function of Metabolic Shift from Butyrate to Lactate Production Triggered by Substrate Concentration ...... 4-48

xi 4.1. Approach ...... 4-48

4.2. Experimental Setup ...... 4-49

Continuous Experiments ...... 4-49 Batch Activity Experiments ...... 4-49 4.3. Results ...... 4-51

Product Analysis ...... 4-51 Culture Kinetics ...... 4-55 Taxanomic Analysis ...... 4-56 Metaproteomic Analysis ...... 4-58 4.4. Discussion ...... 4-60

Defining the Fermentation Types ...... 4-60 Change in Metabolic Enzyme Expression between Low and High Levels .. 4-61 The Methyglyoxal Bypass and the Switch between Butyrate and Lactate Production ...... 4-63 Substrate Uptake and Glycolysis ...... 4-64 Chain Elongation in Level 5a ...... 4-65 Additional Observations from the Metaproteome ...... 4-66 Linking Continuous and Batch Observations...... 4-66 Energy and Electron Mediation ...... 4-68 4.5. Conclusions ...... 4-69

5. Redox Stress from Exogenous Electron Mediators, not Electron Uptake, Causes Shift from Butyrate to Propionate Production in a Poised Potential System ...... 5-70

5.1. Approach ...... 5-71

Experimental Setup ...... 5-71 5.2. Results ...... 5-72

Cyclic Voltometry ...... 5-72 Cathodic Current ...... 5-73 Product Spectrum Analysis ...... 5-73 Taxonomic Analysis ...... 5-75 Metaproteomic Analysis ...... 5-77 5.3. Discussion ...... 5-82

Cathodic Current and Product Formation ...... 5-82 Taxonomy and Proteomics of Butyrate Production ...... 5-82 Taxonomy and Proteomics of Propionate Production ...... 5-85 Fermentation Characteristics of Pectinatus ...... 5-87 Glucose Processing ...... 5-87 Electron Mediation ...... 5-89

xii Oxidative Stress Response ...... 5-91 Effects of Mediators versus Applied Potential on Product Spectrum ...... 5-92 Energy and Electron Mediation ...... 5-95 5.4. Conclusions ...... 5-96

6. Integrated Analysis ...... 6-98

6.1. Characteristics of Fermentation Types ...... 6-100

HBu-type Fermentation ...... 6-100 HLa-type Fermentation ...... 6-101 HPr-type Fermentation ...... 6-103 6.2. Description of Metabolic Mechanisms ...... 6-105

Methylglyoxal Detoxification ...... 6-105 Oxidative Stress Regulation and Electron Mediation ...... 6-106 Pathway Selection ...... 6-107 6.3. Comparison of Taxonomic and Molecular Analyses ...... 6-107

HBu-type Fermentation ...... 6-108 HLa-type Fermentation ...... 6-109 HPr-type Fermentation ...... 6-110 Contrast of Taxonomic and Metaproteomic Approaches ...... 6-110 6.4. Biochemical Market Analysis ...... 6-111

Market Analysis of HLa and HPr-type Fermenation ...... 6-111 Purification Technologies and Viability of Obtained Production Yields ...... 6-112 7. Recommendations for Future Work and Conclusions ...... 7-114

7.1. Future Work ...... 7-114

Expansion of Proteome Libraries ...... 7-114 Metabolic Analysis for Alternative Pathway Switching Mechanisms ...... 7-114 Expansion within Electro-fermentation ...... 7-115 Applications to Fermentation Modelling ...... 7-116 7.2. Conclusions ...... 7-116

Production Modes ...... 7-116 Pathway Switching Mechanisms ...... 7-117 Taxonomic versus Metaproteomic Analysis ...... 7-118 8. References ...... 8-119

Appendix A ...... A-1

xiii Table of Figures

Figure 1: Overview of the glucose-based mixed-culture fermentation metabolism. Each section of the pathway is separated by a colored bubble. From the top and moving clockwise: (I) contains the energy generation pathways; (II) contains the alternate branches from glycolysis; (III) contains alternate pyruvate-derived branches; (IV) contains the acetyl-CoA branches; (V) contains the L-lactate branches; and (VI) contains the dicarboxylic acid cycle. The superscripts R, N, and O on the products indicate whether their redox state relative to -1 glucose is reduced, neutral, or oxidized, respectively, based on their gCOD·molC value. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-2 Figure 2: Primary carbohydrate-based energy pathways for sugar consumption in anaerobic systems, simplified to main branching points and intermediates. EMP glycolysis extends vertically downward from glucose, while ED glycolysis and the pentose-phosphate pathway extend to the right and downward from glucose. The intermediates fructose 6- phosphate, G3P, and pyruvate are underlined to highlight the multiple pathways which cells can use to generate the same intermediates. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-4 Figure 3: Products and pathways branching directly from acetyl-CoA generation. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-8 Figure 4: Reverse β oxidation pathway of chain elongation, depicting elongation of ethanol and acetate to butyrate, followed by elongation to n-hexanoate, and finally to n-octanoate. This same elongation process can also be used to elongate other carboxylic acids by 2 carbons, such as propionate to n-valerate, and n-valerate to n-heptanoate. The stoichiometry for each elongation step is based on the stoichiometry of 1 acetate formation per 6 ethanol oxidations. Acetate must already be present in excess for elongation with ethanol to butyrate. Furthermore, the elongation cycles can take place from any formation of acetyl-CoA under the right environmental conditions71...... 1-12 Figure 5: Products and pathways branching directly from pyruvate reduction to L-lactate. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-14 Figure 6: Products and pathways of the dicarboxylic acid cycle. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-15 Figure 7: Products of the alternative pyruvate-consuming pathways. R-acetoin formation from pyruvate is a side reaction of pyruvate decarboxylase that reacts one pyruvate with

xiv one acetylaldehyde. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis...... 1-16 Figure 8: Products and pathways branching directly from glycolysis. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis. 1-17 Figure 9: Diagram depicting the repression mechanism of Rex enzymes for a generic dehydrogenase gene. NAD+ binds with the Rex enzyme, which changes to an active conformation. The active Rex enzyme then binds to the operon upstream of the dehydrogenase gene, blocking transcription by RNA polymerase...... 1-19 Figure 10: Standard reduction potentials (E°’, mV vs. SHE) of common electron shuttle molecules41,123,145,146 and metabolic half-reactions at pH 7 and an ionic strength of 0.10 M33,40,41...... 1-24 Figure 11: Bioreactor setup for varied substrate concentration experiments...... 3-34 Figure 12: Bioreactor setup for electro-fermentation experiments...... 3-34 Figure 13: Circuit diagram of varied substrate concentration experiment. Multiple load cell signal amplifiers and pump control switches were setup in parallel as required by the experiment. In the electro-fermentation experiments, the float switch was removed and the load cell and amplifier circuit was replaced with lab balances with RS-232 connections directly to the computer. Connected lines are represented by a black dot...... 3-37 Figure 14: Time-course product data of the steady state periods by COD composition. Vertical bars separate the individual operational levels, and the substrate concentration of each operational level is listed below the x-axis...... 4-50 Figure 15: COD balances of each operating level. A) Total effluent COD versus COD of feed. B) Effluent SCOD as measured by HPCL versus effluent SCOD as measured by COD vials...... 4-52 Figure 16: Ratio of measured to calculated biomass COD...... 4-53 Figure 17: Gas composition as determined by both calculation from theoretical yield and measured by GC...... 4-53 Figure 18: Product COD composition normalized by total product composition...... 4-54 Figure 19: Relative abundance of the top 10 most prevalent genera across the 5 sample sets. Normalized OTUs were grouped by genera, and the sum of relative abundance across all samples was used to determine the most prevalent...... 4-56 Figure 20: Biplot of the pyrosequencing results. Operating levels are labelled by groups as L5a, L5b, L10, L15, and L20. Points within each group are labelled sequentially in the order of which the samples were taken as S1, S2, and S3. The red arrows show the order of operation of the reactor...... 4-57

xv Figure 21: Sum of proteins with change in regulation between high (positive) and low (negative) substrate concentrations. Protein regulation shifts calculated from the total catabolic protein spectrum...... 4-58 Figure 22: Heat map of total metabolic enzyme expression. Enzymes are grouped by metabolic pathway or function. L2FC values are based on comparison to the baseline. Values were adjusted up such that the minimum L2FC was equal to 0, and then all enzymes which were not detected at a given level were also set to 0...... 4-59 Figure 23: Mixed culture model metabolism and key metabolic functions. Protein name, abbreviation, and Uniprot accession number are listed in the table, and log2-fold change (L2FC) of each protein is listed alongside its function. Bifidobacterium showed an increased in expression of Ldh from L5b to L10, and then did not show any change in expression from L10 to L20...... 4-62 Figure 24: Redundancy analysis of product formation and the sum of key metabolic processes...... 4-64 Figure 25: CVs of mediators and background fermentation broth...... 5-72 Figure 26: Product spectrum of electro-fermentation experiments A) is the product balance (%gCOD out vs gCOD consumed), and B) is the product relative abundance. Control is level 5b from the varied substrate concentration experiment. The mediated experiments depict the average of 5 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-74 Figure 27: Relative abundance of the top 12 most prevalent genera across all sample sets. Normalized OTUs were pooled by genera, and the sum of relative abundance across all samples was used to determine the most prevalent. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-76 Figure 28: Biplot of the reactor taxonomic groupings as represented by PC1 and PC2. Experimental (closed circuit) reactors are grouped together within each mediator, and separate from the open circuit (OC) reactors...... 5-77 Figure 29: Sum of proteins with change in regulation between experimental (positive) and control (negative) reactors. Protein regulation shifts calculated from the total catabolic protein spectrum...... 5-78 Figure 30: Heat map of total metabolic enzyme expression. Enzymes are grouped by metabolic pathway or function. L2FC values are based on comparison to the baseline.

xvi Values were adjusted up such that the minimum L2FC was equal to 0, and then all enzymes which were not detected at a given level were also set to 0...... 5-79 Figure 31: Expression of enzymes of the butyrate production pathway A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-83 Figure 32: Single factor linear regression of HBu production as a percentage of total product formation versus Clostridium (squares, solid line), Megasphaera (diamonds, dashed line), Bifidobacterium (crosses, long-dashed line), and Ethanoligenens (circles, dotted line). Data includes only through NR, and excludes MV...... 5-84 Figure 33: Expression of enzymes of the DCA cycle of propionate production A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-85 Figure 34: Linear regression of HPr production as a percentage of total product formation versus Pectinatus relative abundance. The solid line (—) incorporates the full data set, while the dashed line (- -) incorporates only up through NR, excluding MV...... 5-86 Figure 35: Expression of enzymes of the glycolysis pathway A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-88 Figure 36: Expression of electron mediation enzymes A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-90 Figure 37: Expression of oxidative stress regulation enzymes A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE...... 5-91

xvii Figure 38: Linear regression of Pectinatus relative abundance versus reduction potential of the mediator compounds. The solid line (—) incorporates the full data set, while the dashed line (- -) incorporates only up through NR, excluding MV...... 5-94 Figure 39: Difference in functional expressions between closed circuit (positive) and open circuit (negative) reactors...... 5-94 Figure 40: Biplot of all samples from varied substrate concentration and electro- fermentation experiments showing taxonomic variance between fermentation types. ....6-98 Figure 41: Product composition of the three identified production types. A) Gas composition by mol%. B) Liquid phase products by mol%. C) Total products by COD%...... 6-99 Figure 42: Transition states of increasing organic load from the varied substrate concentration experiment. A) Initial reactor start-up. B) Transition from 5 to 10 g·L-1. C) Transition from 10 to 15 g·L-1...... 6-102 Figure 43: Comparison between this work (arrow) and results of Horiuchi et al.21 of ATP production during HPr-type fermentation. Operation in this work was at pH 5.50 while that of the literature was as 8.00. Trend line is for ATP production at pH 8.00 only...... 6-104 Figure 44: Detected and inferred metabolic interactions and pathway expression under high carbon load, low oxidative stress, and standard conditions...... 6-108

xviii List of Tables

Table 1: Standard reduction potentials (E°’) in mV (vs. SHE) of primary EMCs involved with anaerobic activity. Unless otherwise noted, data is from is taken from Alberty40...... 1-7 Table 2: Product and cofactor molar yields of the acetyl-CoA branches associated with Pdhc/Pfo (top) and Pfl (bottom) activity with respect to glucose (Glu) and EMP Glycolysis...... 1-10 Table 3: Product and cofactor molar yields of the L-lactate branches with respect to glucose (Glu) and EMP Glycolysis...... 1-13 Table 4: Product and cofactor molar yields of the Dicarboxylic Acid Cycle with respect to glucose (Glu) and EMP Glycolysis...... 1-15 Table 5: Product and cofactor molar yields of the alternative pyruvate branches with respect to glucose (Glu) and EMP Glycolysis...... 1-16 Table 6: Product and cofactor molar yields of the glycolysis side-branches with respect to glucose (Glu) and EMP Glycolysis...... 1-17 Table 7: Product-based pathway balance equations. All units are in moles, and Glu is glucose (C6H12O2), HPy is pyruvic acid, HMa is malic acid, HSu is succinic acid, HAc is acetic acid, EtOH is ethanol, HLa is lactic acid, HPr is propionic acid, HFo is formic acid, - and X is biomass. 1 H2 is interchangeable with 2 e ...... 3-44 Table 8: Average COD composition of each operation level. Value listed as percent gCOD by total product gCOD...... 4-50 Table 9: Yield and specific substrate consumption rate results from batch tests...... 4-55 Table 10: Percent difference within and between groups for each kinetic factor. P-values for each comparison are listed directly below...... 4-55 Table 11: Significant loadings of principal components 1-3. Percent of total variation for each principal component is listed in brackets in the header...... 4-58 Table 12: Specific substrate consumption rates of the steady state levels...... 4-65 Table 13: Production of ATP and sinking of e- per biomass associated with the product -1 spectrum of each level. ATP and e- values are each in mmol·gCODX . The “Total” section describes functionality per product pathway from glucose uptake through end product; the “Glycolysis” section describes functionality within glycolysis distributed by product yield; and the “Post-HPy” section describes functionality per product pathway from pyruvate through end product. The values are heat-mapped level-by-level and across sections to highlight spread of functionality within each level...... 4-68

xix Table 14: Measured reduction potentials of mediator peaks and their allosteric means in abiotic fermentation broth at pH 5.50 and 30°C. Reduction potentials are compared to standard literature values (pH 7 and 25°C)...... 5-72 Table 15: Theoretical maximum electron supply from mediators in feed before interaction with cathode (Mediator) and measured electron supply from the electrode array (Cathode), both in Amps. % Reduced and % Oxidized is a measure of the redox state of the mediator in the feed before entering the vessel and interacting with the cathode...... 5-73 Table 16: Significant loadings of principal components 1-3. Percent of total variation for each principal component is listed in brackets in the header...... 5-76 Table 17: Production of ATP and sinking of e- per biomass associated with the product - -1 spectrum of each mediator. ATP and e values are each in mmol·gCODX . The “Total” section describes functionality per product pathway from glucose uptake through end product; the “Glycolysis” section describes functionality within glycolysis distributed by product yield; and the “Post-HPy” section describes functionality per product pathway from pyruvate through end product. The values are heat-mapped level-by-level and across sections to highlight spread of functionality within each level...... 5-96 Table 18: Product yields of HBu-type fermentation. Biomass is presented as COD yield -1 (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase...... 6-100 Table 19: Product yields of HLa-type fermentation. Biomass is presented as COD yield -1 (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase...... 6-101 Table 20: Product yields of HPr-type fermentation. Biomass is presented as COD yield -1 (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase...... 6-104 Table 21: US theoretical production capacity compared to estimated 2016 global market for acids and acid derivatives...... 6-111

xx List of Abbreviations

AQ Anthraquinone-2,7-disolfonate HPr Propionic Acid ATP Adenosine Triphosphate HPy Pyruvic Acid BA Basal Anaerobic HRT Hydraulic Retention Time CoA Coenzyme A HSu Succinic Acid COD Chemical Oxygen Demand kM Specific Glucose Uptake Rate CSTR Continuous Stirred Tank Reactor KS Half-Velocity Constant CV Cyclic Voltomagraph MV Methyl Viologen Cyt Cytochrome NAD Nicotinamide Adenine Dinucleotide DCA Dicarboxylic Acid (Cycle) NADP Nicotinamide Adenine Dinucleotide Phosphate E' Reduciton Potential NR Neutral Red E°' Standard Reduction Potential OLR Organic Loading Rate ED Entner-Doudoroff (pathway) OTU Operational Taxonomic Unit EMC Electron-Mediation Cofactor(s) PC Principal Component EMP Embden-Meyerhof-Parnas PCA Principal Component Analysis (pathway) EtOH Ethanol Pdhc Pyruvate Dehydrogenase Complex FAD Flavin Adenine Dinucleotide Pfl Pyruvate-Formate Lyase Fd Ferredoxin Pfo Pyruvate-Ferrodoxin Oxidoreductase Fdh Formate Dehydrogenase Q Quinone FID Flame Ionization Detector Rex Redox Repressor (Enzymes) FMN Flavin Mononucleotide RF Riboflavin G3P Glyceraldehyde-3-Phosphate RID Refractive Index Detector GC Gas Chromatograph SCOD Soluble Chemical Oxygen Demand Glu Glucose SHE Standard Hydrogen Electrode HAc Acetic Acid TCD Thermal Conductivity Detector HBu Butyric Acid TCOD Total Chemical Oxygen Demand HFo Formic Acid TGY Tryptone-Glucose-Yeast Extract HHx Hexanoic Acid VFA Volatile Fatty Acid(s) HLa Lactic Acid VWD Variable Wavelength Detector HMa Malic Acid X Biomass HPLC High Pressure Liquid Chromatograph

xxi 1. Introduction*

World population growth, unprecedented rates of urbanization, and consumerism are defining the engineering challenges of the 21st century. Prominent among these challenges is the management of increasing waste streams1. Global climate change and the economic strain of a volatile oil market are also difficult challenges resulting from the change in population dynamics, and have created a necessity to find alternative and sustainable methods of producing not only fuels, but also feedstock chemicals traditionally derived from fossil sources2. New strategies must be developed in order to meet these challenges, and waste stream carbon recycling into biochemicals via fermentation may help alleviate both demands. Fermentation by mixed cultures, or natural consortia of microbes, allows for robustness and flexibility within waste processing to produce a wide variety of biochemicals, but requires control of product outputs through manipulation of in-reactor conditions, rather than limiting available pathways through selection of culture capability3,4.

Fermentation is a key central step in treating organic waste streams in which soluble organic monomers are converted into short-chain fatty acids, alcohols, sugar alcohols, hydrogen, and carbon dioxide4,5. While the definition is not universally agreed upon, fermentation is in general biological catabolism without the use of an external electron acceptor, such as O2, - 2- NO3 , or SO4 , and commonly relies on substrate-coupled electron-transfer reactions for adenosine triphosphate (ATP) generation6. The primary electron sink is the substrate, but excess electrons may be sunk by reducing protons to elemental hydrogen. Oxidation reactions provide the potential energy to generate cellular energy in the form of ATP. This process is known as substrate level phosphorylation. Fermenting cells can also leverage energy from electron transfer in other ways, including through proton, cation, and electron transport chains7.

The lack of an external electron acceptor is crucial to the understanding of the fermentation metabolism. In aerobic respiration, organics are oxidized to CO2 by coupled reduction of intracellular electron-mediation cofactors (EMC). Re-oxidising the EMCs by sinking electrons to terminal electron acceptors removes excess electrons from the cell and provides the driving force for the majority of respirative ATP generation via oxidative phosphorylation.

* This chapter was modified for this thesis from (Appendix A): Hoelzle, R. D., Virdis, B. & Batstone, D. J. Regulation mechanisms in mixed and pure culture microbial fermentation. Biotechnol. Bioeng. 111, 2139–2154 (2014). doi:10.1002/bit.25321 1-1

Because microorganisms in anaerobic systems are not able to utilize external electron acceptors, they must utilize substrate-level phosphorylation to generate the majority of their ATP. Excess electrons are then usually sunk by substrate reduction, primarily by production of short chain fatty acids, alcohols, and sugar-alcohols, or in the form of hydrogen gas if electrons cannot be utilized in substrate-level reduction (Figure 1). As such, fermentation can be regarded as coupled substrate-level oxidation-reduction reactions, with catabolic energy being mainly derived from oxidation reactions (through substrate-level phosphorylation), and the reduction reactions mainly existing to sink electrons. Excess electrons can be sunk to hydrogen, but this can be energetically unfavorable when the hydrogen partial pressure is high8.

Fermentation is widely used (in pure and mixed culture forms) to produce food products (e.g. beverages, fermented dairy products)9–11, renewable fuels (e.g. hydrogen, ethanol)12,13, pharmaceuticals (e.g. antibiotics, diagnostic markers)14,15, and industrial chemicals (e.g. organic acids, biopolymers)16–18. Some pure culture processes, such as bioethanol production from starch by Saccharomyces cerevisiae, rely on fast fermentation times and inoculation with highly active cultures which quickly generate product concentrations that are inhibitory to competing species, and thereby minimize loss of product yield while not implementing expensive sterilization equipment12. However, in

1-2 pharmaceutical production, tight production regulations enforced by agencies like the US Food and Drug Administration and the European Medicines Agency require that secondary (contaminating) organisms be completely eliminated. Even in biochemical production, the uncontrolled high-rate fermentation of sugars by mixed cultures is normally regarded as an undesirable side-reaction caused by microbial contamination19.

Mixed culture fermentation is not widely used in industrial fermentation, largely because factors determining product spectrum of mixed culture systems are still not well understood. Hence, mixed culture product output is difficult to control. While secondary, higher value, products such as propionate and lactate have been observed for mixed cultures20,21, the normal range of products from fermentation on simple sugars is acetate, butyrate, and ethanol4,22–25. Additionally, outside of electron mediation mechanisms, there has been limited investigation of the metabolic mechanisms through which various fermentation pathways are regulated. Even within the area of electron mediation, this has been limited to investigation of the mediator pools26 and mechanistic assumptions in fermentation models27.

It is therefore necessary to understand production modes within mixed culture fermentation at a mechanistic level in order to develop reliable production of biochemical products from waste streams. Given the growth in availability of tools for exploring gene and protein expression in complex cultures through methods of metatranscriptomics and metaproteomics, descriptions of fermentation modes must now include molecular exploration of the metabolic mechanisms. First, it is necessary to understand fermentation at the pathway level, from which mechanistic descriptions can be built upon.

1.1. Review of Metabolic Pathways

Fermentation products are predominantly produced from the metabolic intermediate pyruvate (Figure 1). Pyruvate is the primary intermediate resulting from the initial carbohydrate oxidation step, which is the central ATP-generating pathway in most cells. During respiration, pyruvate is then oxidized through the citric acid cycle to produce ATP while the excess electrons are used to reduce an external electron acceptor to maintain the redox balance. By contrast, in fermentation a variety of pyruvate-reducing pathways have evolved which couple to pyruvate oxidation and close the electron balance. These pathways are reviewed in more detail in this section.

1-3

phosphate, - phosphate,

f all products and intermediates are are and intermediates products all f

main branching points and intermediates. EMP glycolysis extends extends glycolysis EMP intermediates. and points branching main

phosphate pathway extend to the right and downward from glucose. The intermediates fructose 6 fructose intermediates The glucose. from downward and right the to extend pathway phosphate

to simplified systems, in anaerobic consumption sugar for pathways energy based -

are underlined to highlight the multiple pathways which cells can use to generate the same intermediates. The stoichiometry o stoichiometry The intermediates. same the generate to use can cells which pathways multiple the highlight to underlined are

Primary carbohydrate Primary

: 2

Figure Figure pentose the and glycolysis ED while glucose, from downward vertically and pyruvate G3P, glycolysis. and EMP glucose to respect with shown

1-4 Primary Energy Pathways

The product pathways described here are primarily acidogenic, and therefore branch from pyruvate-generating pathways. For carbohydrate catabolism, this generally means Embden- Meyerhof-Parnas (EMP) glycolysis, Entner-Doudoroff (ED) glycolysis, and the pentose- phosphate pathway28. A wide variety of saccharides, including sucrose, lactose, glucose, fructose, and xylose, as well as a variety of sugar alcohols, can be metabolized by one or more of these pathways (Figure 2). They each eventually are metabolized to glyceraldehyde-3-phosphate (G3P). G3P is then oxidized to pyruvate through the second half of EMP glycolysis. Oxidation of 1 G3P to 1 pyruvate generates 2 ATP. This oxidation also liberates 2 electrons per G3P, which reduce EMCs and used to form reduced end products. This use of reduced end products as electron sinks maintains the catabolic electron balance.

While ED glycolysis and the pentose-phosphate pathway both supply G3P for oxidation to pyruvate, they are less efficient at generating ATP than EMP glycolysis. ED glycolysis occurs mainly in aerobic environments and as a pathway to catabolize sugar acids, such as gluconate28. The pentose-phosphate pathway, by contrast, is primarily anabolic, generating precursors for biomass generation29. Therefore, only EMP glycolysis will be considered from this point on as it is the most relevant to the fermentation metabolism.

Redox Balancing through Electron Mediation

In order to understand the fermentation metabolism, it is first necessary to understand the importance of electron mediation within a fermenting culture. Electrons are transferred in metabolic redox reactions by a set of cofactors, which we refer to collectively as electron- mediating cofactors (EMCs). These cofactors vary in functional group, specificity, and reaction mechanism, but all serve the same basic purpose as electron carriers (by conversion between oxidized and reduced forms). EMCs play a major role in both oxidation of substrates and sinking of electrons into reduced end-products. In the anaerobic energy metabolism, the primary EMCs are nicotinamide adenine dinucleotide (NAD), nicotinamide adenine dinucleotide phosphate (NADP), flavin adenine dinucleotide (FAD), and ferredoxin (Fd)6,30. Other EMCs involved in electron transfer in anaerobic environments, though not necessarily in fermentation, include flavin mononucleotide (FMN)31,32 and a wide array of protein complexes involved in transmembrane electron transport, such as cytochromes and terminal oxidases33.

1-5 The structure of the functional groups of NAD and NADP is identical. NADP is primarily associated with anabolic activity34. However, it is also the active EMC in several catabolic reactions, including butanol and isopropanol production35 and methylglyoxal conversion to acetol and lactaldehyde36. Additionally, some catabolic reactions, such as glycerol to 1,3- propanediol37, utilize either NAD or NADP depending on the active enzyme. In these cases, NAD and NADP are represented together as NAD(P). Due to their similarities, we will use NAD(P) unless it is necessary to distinguish between them. NAD(P), FAD, and FMN each utilize an aromatic ring structure to carry electrons. EMCs which use this mechanism are referred to collectively as quinones (Q). Additional information on the structures and mechanisms of activity is available in the literature38,39.

Electrons are transferred between EMCs and metabolic intermediates via hydrogenase enzymes, which function by transferring hydrogen bonds between EMC and the intermediate. They are the primary enzyme responsible for electron removal from the cell and generally act by hydrogenating double bonds in alkene or ketone groups. Some, such as H2:Fd hydrogenase, reduce protons to make molecular hydrogen. They may also work in reverse to transfer electrons from a metabolic intermediate to an EMC by removing hydrogen from an intermediate, often from an -OH group resulting in a ketone, such as with glycerol dehydrogenase37. Hydrogenase enzymes are part of a wider group of enzymes known as oxidoreductases. Oxidoreductases catalyze metabolic redox reactions and do not necessarily involve hydrogenation. These enzymes are responsible both for the generation of reduced EMCs, through steps like glucose oxidation to pyruvate, and for oxidation of the cofactors to facilitate electron removal via the various reduced end products that make up the fermentation product spectrum.

Information on the standard reduction potentials (E°’) of these EMCs under typical pH and ionic strength conditions is provided in Table 1. The effect of the ionic strength on E°’ for each of the EMCs is notable, as is the effect of pH on the hydrogen-binding EMCs, highlighting the impact of these two factors on metabolic thermodynamics. More extensive reduction potential data for these species is available in the literature40,41

The reduction potentials in Table 1 represent the potential at the reaction midpoint. Midpoint potentials, while informative about the thermodynamics of cellular processes involving these EMCs, do not fully represent the reduction potential energy of the intracellular EMC pools. Measurements of the redox states of bacterial EMCs have shown that the Fd and NADP pools are generally maintained at 90% or more reduced42, while the

1-6 Table 1: Standard reduction potentials (E°’) in mV (vs. SHE) of primary EMCs involved with anaerobic activity. Unless otherwise noted, data is from is taken from Alberty40.

I = 0.10 M I = 0.25 M Half-reaction pH 5 pH 7 pH 8 pH 5 pH 7 pH 8 †Fd3+ + e- → Fd2+ -400.9 -400.9 -400.9 -403.0 -403.0 -403.0 NAD+ + H+ + 2e- → NADH -258.9 -318.1 -347.7 -256.9 -316.0 -345.6 NAPD+ + H+ + 2e- → NADPH -263.6 -322.7 -352.3 -257.4 -316.6 -346.2 ‡ + + - FAD + 2H + 2e → FADH2 - -219.0 - - - -

+ - FMN + 2H + 2e → FMNH2 -100.7 -219.0 -278.1 -102.7 -221.0 -280.1 ҂ - Cyt cox + e → Cyt cred +222.3 +222.3 +222.3 +212.1 +212.1 +212.1 † The source of Fd is not specified in Alberty40, however Lundblad and MacDonald41 list ΔE’° of Fd from Clostridium as -413 mV. ‡ ΔE’° data for FAD is not available in Alberty40. The value listed is sourced from Lundblad and MacDonald41. This value was measured at pH 7.14 with unspecified ionic strength. ҂ The source of Cyt c is not specified in Alberty40, however Lundblad and MacDonald41 list ΔE’° of Cyt c552 from Pseudomonas and Cyt c4 from Azotobacter as +300 mV.

NAD pool is maintained at 80% or more oxidized26,42. This shifts the intracellular reduction potential of these EMCs down to around (vs SHE) -500 mV for Fd and -360 mV for NADP, while shifting NAD up to around -280 mV43.

Cells are able to conserve the reduction potential energy of these EMC pools by coupling redox reactions of multiple pools through the processes of electron bifurcation43 and confurcation44. Electron bifurcation utilizes highly exergonic substrate reduction steps to pass electrons from an EMC pool of higher reduction potential to a pool of lower reduction potential (generally Fd). For instance, the reduction of crotonyl-CoA to butyryl-CoA by 1 NADH in butyrate and butanol formation (ΔG°’ = -60 kJ·mol-1, Section 1.1.4) can instead utilize 2 NADH to reduce crotonyl-CoA and 1 Fd through an intermediate flavin group (ΔG°’ = -40 kJ·mol-1)45. This process helps maintain the highly reduced Fd pool while also conserving reduction potential energy. At a required enthalpy of at least -75±5 kJ·mol-1 for 46 2 ATP formation , electron bifurcation can increase ATP yield by up to /7 mol per reduction reaction.

Electron confurcation acts in reverse to bifurcation by utilizing electrons from EMC pools of low reduction potential to drive endergonic reactions, such as lactate oxidation to pyruvate (Section 1.1.6), by passing those electrons to an EMC pool of higher reduction potential through a flavin intermediate44.

1-7 Pyruvate Oxidation

Fermenting cells pair substrate oxidation to electron mediation via a number of different pathways, which primarily branch from pyruvate. Most commonly, a portion of pyruvate is oxidized for further substrate level phosphorylation. This step begins with oxidation to acetyl- CoA, and can be in induced by one of three enzymes: the Pyruvate Dehydrogenase Complex (Pdhc), Pyruvate-Formate Lyase (Pfl), and Pyruvate-Fd Oxidoreductase (Pfo)47– 49. Pdhc is a compound enzyme structure that oxidizes pyruvate through a series of steps, which ultimately liberates CO2 and reduces a bound FAD. The reduced FAD is re-oxidized either by passing its hydrogen bonds to NAD(P) or by producing H2, regenerating Pdhc for oxidation of the next pyruvate. By contrast, Pfl oxidizes pyruvate by direct replacement of the carboxyl group with CoA, generating formate rather than CO2 and H2. At low pH, formate + can be oxidized to CO2 by reduction of NAD through the enzyme formate dehydrogenase (Fdh)50, however generation of NADH in this way is associated with formate oxidation rather than pyruvate oxidation. Pfo acts similarly to Pdhc, directly producing CO2 and reducing an

Figure 3: Products and pathways branching directly from acetyl-CoA generation. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis.

1-8 EMC, though Fd is reduced instead of FAD. Pfo appears to be highly prevalent among strictly anaerobic bacteria, especially in species of Clostridia, as well as some species of algae capable of dark fermentation48.

While there are several enzymes which each facilitate the oxidation of pyruvate to acetyl- CoA, they have different implications for cellular electron mediation, and are expressed under different circumstances. From Figure 3, it is evident that pyruvate oxidation via both

Pdhc and Pdo each accounts for production of 1 CO2 and 1 reduced EMC per pyruvate. By contrast, oxidation via Pfl accounts for 1 formate per pyruvate. Due to this difference in product output, differences in which enzyme is expressed will have implications both on how a cell must sink electrons and on the acidification rate of the cytoplasm and the extracellular medium. Electron balancing and pH management is discussed in further detail below, but first it is necessary to understand under which conditions the various pyruvate-oxidizing enzymes are expressed.

From studies on pure cultures, it appears that enzyme selection for pyruvate oxidation is primarily dependent on redox environment (including O2 content) and pH. Pdhc has been 49,51,52 shown to be active in both aerobic and anaerobic systems . By contrast, O2 is inhibitory to expression of both Pfl53–56 and Pfo57, and therefore they are both primarily active in anaerobic systems. This means that species in a mixed culture capable of producing Pdhc will be more likely to produce acetyl-CoA-derived products in a system that is not completely anaerobic. This may also lend them more resistance to other oxidative stressors (Section 1.2.3) by maintaining the ability to produce ATP through acetate and butyrate generation.

Expression based on pH is a little more complicated and less well-defined. Snoep et al.49 demonstrated using Enterococcus faecalis under anaerobic conditions that Pdhc is most active at low pH (below 6.5) while Pfl is most active at high pH (above 6.5). However, it is uncertain whether this pH-dependent expression occurs universally. Mixed-culture fermentation experiments by Lu et al.22 and Temudo et al.4 show low formate production and high CO2 production at low pH, while at high pH, the reverse was true. However, it is not clear whether these changes in formate and CO2 formation were due to changes in relative simultaneous activity of Pdhc/Pfo and Pfl, the pH-dependent activity of Fdh with only Pfl expression, or some combination of each of these functions. It can therefore not be said for certain whether EMC reduction is directly associated with pyruvate oxidation or formate oxidation at low pH, or with what ratio they are reduced at this stage of the metabolism.

1-9 Acetyl-CoA Branches

The metabolic pathways branching from pyruvate-derived acetyl-CoA (Figure 3) make up the most common pathways active in mixed-culture fermentation58, likely due to their ATP generating capacity via substrate-level phosphorylation. These pathways are able to efficiently regulate energy production and electron removal through selection of a range of end products59. End products derived from acetyl-CoA are acetate, n-butyrate (butyrate), ethanol, n-butanol (butanol), acetone, and isopropanol.

The acetate and butyrate pathways are major sources of ATP generation in fermentative systems, responsible for nearly half of the substrate-level phosphorylation in an EMP glycolysis system60,61. Additionally, butyrate generation serves to sink excess electrons post- pyruvate. Ethanol, another major fermentation product, and butanol are also routes for electron sinking by reduction of acetyl-CoA and butyryl-CoA (an acetyl-CoA derivative), respectively59. Acetone results from a decarboxylation of acetoacetate, which is produced when acetate and butyrate are re-consumed back to acetyl-CoA and butyryl-CoA, respectively. Re-consumption of acetate and butyrate in this method occurs as part of the sporulation process in solventogenic species of Clostridia and results in the production of ethanol and butanol62. Some strains of C. beijerinckii are also able to sink electrons to isopropanol by reducing acetone63.

Depending on the specific enzyme used for pyruvate oxidation, varying strain will be put on the electron mediation system due to production of either formate or CO2 and a reduced EMC. To highlight this, Table 2 lists the difference in oxidized EMC yield (from

Table 2: Product and cofactor molar yields of the acetyl-CoA branches associated with Pdhc/Pfo (top) and Pfl (bottom) activity with respect to glucose (Glu) and EMP Glycolysis.

Yield Acetate Butyrate Acetone Isopropanol Ethanol Butanol Formate Pdhc/Pfo Y (Product·Glu-1) 2 1 1 1 2 1 0 Y (ATP·Glu-1) 4 3 2 2 2 2 0 Y (e-·Glu-1) 8 4 8 6 0 0 0

Pfl Y (Product·Glu-1) 2 1 1 1 2 1 2 Y (ATP·Glu-1) 4 3 2 2 2 2 2 Y (e-·Glu-1) 4 0 4 2 -4 -4 4 Note: The electron (e-) yield is the number of electrons that are released per glucose as a result of the given pathway. Every 2 e- yielded is equivalent to reduction of 1 NAD+ to NADH.

1-10 glucose) associated with each product based on Pdhc/Pfo and Pfl oxidation of pyruvate. As pointed out by Kleerebezem et al.30, it should not be assumed that any given reduced EMC can be used to produce any given reduced end product. Most enzymatically-catalyzed redox reactions have specificity for one particular EMC, as with NADP specificity in 1,3- propanediol production37, and butanol and isopropanol production35. Additionally, production of some reduced products under certain conditions is only thermodynamically possible with an EMC with a low enough E°’. An example is proton reduction to H2 under high H2 partial pressures, which is only possible using reduced Fd30. It is therefore necessary to determine which specific enzymes are active during these oxidation steps, and in what ratios, in order to identify the range and likely stoichiometric ratios of possible products.

Chain Elongation

Some bacteria, such as Clostridum kluyveri and Megasphaera elsdenii, are able to use reverse β oxidation, a pathway analogous to butyrate formation from pyruvate oxidation, to produce the longer carbon chain products, such as n-hexanoate and n-octanoate, by chain elongation of shorter carbon chain intermediates64–68. Carbon chains are elongated by 2 carbons at a time through condensation of an acetyl-CoA with another carbonyl-CoA. The carbonyl group from acetyl-CoA is then fully reduced, and finally the CoA group is transferred to a carboxylic acid with fewer carbons than the new carbonyl-CoA via a CoA transferase. This process is represented in Figure 4 with ethanol and acetate being elongated first to butyrate, then to n-hexanoate, and finally to n-octanoate. However, this process also works analogously for elongation of other carboxylic acids, such as propionate to n-valerate, and n-valerate to n-heptanoate69.

This process appears to occur when high reducing equivalents are available in the form of ethanol or H2, and only when high-energy substrates are not readily available. Lactate reduction to pyruvate (Section 1.1.6) is also able to provide the required reducing equivalents70. As such, it provides a unique way to sink electrons into longer short chain fatty acids. While effective as an electron sink, the reverse β oxidation pathway itself is endergonic and requires co-oxidation of 1 ethanol to acetate for ATP generation for every 5 ethanol elongated to be thermodynamically feasible66. Additionally, this process is slow, and so is subject to competition from acetoclastic methanogens under normal mixed culture fermentation conditions (pH 7, 37°C)64,66,67. It is therefore necessary to inhibit methanogenesis for chain elongation to occur, either by operating at low pH or by use of a methanogenic inhibitor compound66.

1-11

Figure 4: Reverse β oxidation pathway of chain elongation, depicting elongation of ethanol and acetate to butyrate, followed by elongation to n-hexanoate, and finally to n-octanoate. This same elongation process can also be used to elongate other carboxylic acids by 2 carbons, such as propionate to n-valerate, and n-valerate to n-heptanoate. The stoichiometry for each elongation step is based on the stoichiometry of 1 acetate formation per 6 ethanol oxidations. Acetate must already be present in excess for elongation with ethanol to butyrate. Furthermore, the elongation cycles can take place from any formation of acetyl-CoA under the right environmental conditions71.

1-12 Table 3: Product and cofactor molar yields of the L-lactate branches with respect to glucose (Glu) and EMP Glycolysis. Yield Lactate 1,2-Propanediol Propanol Propionate (a)† Propionate (b)‡

-1 Y (Product·Glu ) 2 2 2 2 2 Y (ATP·Glu-1) 2 2 2 4 2 Y (e-·Glu-1) 0 -8 -12 -4 -4 † Propionate (a) refers to propionate production via the 1,2-propanediol pathway. ‡ Propionate (b) refers to propionate production via the acrylate pathway.

Lactate Branches

The ability to produce L-lactate (lactate) is widespread among microorganisms. It is the primary fermentation product for some species of Lactobacillus72,73, but for many organisms it serves as an intermediate which is later re-consumed59,74. Lactate has the same carbon oxidation state as glucose. It is therefore generally ineffective as an electron sink to enable ATP generation through acetate or butyrate production, even though it is produced by direct pyruvate reduction. However, several species are able to metabolize lactate into products that are more reduced than glucose, making them more effective as electron sinks than lactate. The acrylate pathway, utilized by Clostridium propionicum75,76, Megasphaera elsdenii77, and Prevotella ruminicola78, reduces lactate to propionate, sinking 4 electrons per glucose (Table 3). Similarly, Lactobacillus buchneri reduces lactate to 1,2-propanediol79, also eliminating 4 electrons per glucose. 1,2-Propanediol can also be reduced to propanol or converted to propionate for ATP generation by Listeria innocua and species of Salmonella80. Production of propanol and propionate from 1,2-propanediol has not been observed as originating from lactate-produced 1, 2-propanediol, though it is possible that this kind of syntrophic relationship could exist in mixed culture systems. This branch of the metabolism is therefore presented here as proposed possible mixed culture metabolism (Figure 5).

Though lactate is generally not expressed as a primary fermentation product, it has the same carbon oxidation state as glucose, and so is useful as an energy source. When consumed, it is readily metabolized through pyruvate to acetate and butyrate. Some species, such as Propionibacterium acidipropionici, couple ATP generation via lactate oxidation to acetate, with electron sinking via lactate reduction to propionate81. Additionally, lactate can be supplied as a primary substrate to cultures containing lactate reduction pathways, in which case some lactate is metabolized for ATP generation through pyruvate oxidation, while some is reduced. In anaerobic digestion processes, lactate and propionate accumulation are viewed as evidence of inhibition within the methanogenicprocess and are often associated with shock loads of substrate20,82. Lactate and propionate have also been

1-13 Figure 5: Products and pathways branching directly from pyruvate reduction to L-lactate. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis. observed as products of mixed culture fermentation4,22, though the stimuli for their production has yet to be described.

Dicarboxylic Acid Cycle

An alternative method of propionate generation is via the dicarboxylic acid (DCA) cycle, in which fumarate, malate, and succinate are also produced as intermediates83,84 (Figure 6). The DCA cycle acts analogously, but in reverse, to the citric acid cycle of aerobic metabolism, reducing oxaloacetate to succinate through L-malate and fumarate. Propionate is generated when propanoyl-CoA transfers its CoA to succinate, starting the regeneration stage of the cycle. The final step is regeneration of oxaloacetate by transfer of the CoA to pyruvate, generating propanoyl-CoA.

Because this pathway is a cycle with two separate entry points (pyruvate and phosphoenolpyruvate), the ATP and electron mediation effects of propionate and the in- cycle products must be accounted for differently. Propionate generation is the result of transfer of a CoA to pyruvate, while fumarate, malate, and succinate generation are each

1-14 Figure 6: Products and pathways of the dicarboxylic acid cycle. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis. Table 4: Product and cofactor molar yields of the Dicarboxylic Acid Cycle with respect to glucose (Glu) and EMP Glycolysis. Yield Malate Fumarate Succinate Propionate

-1 Y (Product·Glu ) 2 2 2 2 Y (ATP·Glu-1) 0 0 0 2 Y (e-·Glu-1) -2 -2 -4 -4 the result of phosphoenolpyruvate entering the DCA cycle. Therefore, propionate generation from the DCA cycle is associated with the final ATP generation step of glycolysis, while the other products are not (Table 4). By extension, since each propionate produced from the DCA cycle constitutes one turn of the cycle, the oxidized EMC yield of each propionate is equivalent to the full oxidized EMC yield of one turn of the cycle.

Alternative Pyruvate Branches

Other products that can be metabolized from pyruvate include ethanol, acetoin, and 2,3- butanediol (Figure 7). Ethanol derived in this way bypasses acetyl-CoA, using a direct decarboxylation of pyruvate to acetylaldehyde, and then reduction to ethanol. This ethanol pathway is utilized most notably by Zymomonas mobilis and eukaryotes, such as Saccharomyces cerevisae, for industrial ethanol biofuel fermentation12. Meanwhile, acetoin and 2,3-butanediol are produced from decarboxylation of pyruvate to α-acetolactate. Acetoin production generally occurs during the exponential growth phases of bacteria such as Bacillus subtilis and Klebsiella aerogenes as a pH-neutral energy storage molecule which can be re-consumed during stationary phase85. Though acetoin has also been shown to result from a side product of pyruvate decarboxylase in species of Saccharomyces86. 2,3-

1-15 Figure 7: Products of the alternative pyruvate-consuming pathways. R-acetoin formation from pyruvate is a side reaction of pyruvate decarboxylase that reacts one pyruvate with one acetylaldehyde. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis. Table 5: Product and cofactor molar yields of the alternative pyruvate branches with respect to glucose (Glu) and EMP Glycolysis. Yield Ethanol Acetoin 2,3-Butanediol

-1 Y (Product·Glu ) 2 1 1 Y (ATP·Glu-1) 2 2 2 Y (e-·Glu-1) 0 4 2

Butanediol is one of the products produced by acetoin re-consumption in species of Bacillus, Klebsiella, Serratia, and Pseudomonas85,87.

These products do not result in any additional post-glycolysis ATP production, and their electron-sinking capacity is quite straight forward (Table 5). Ethanol’s effect on the electron balance is identical whether or not it is produced in this pathway or through an acetyl-CoA step. Acetoin and 2,3-butanediol, however, each represent a net reduction of EMCs when counting from glucose. As a result, acetoin is not useful to cells as an electron sink, and 2,3- butanediol must be produced alongside other reduced cofactors to maintain the cellular redox balance.

Glycolysis Side-Branches

Two pathways can extend off of glycolysis at dihydroxyacetone-phosphate: the glycerol pathway, which is commonly used as an electron sink; and the methylglyoxal bypass, in which toxic methylglyoxal is converted to other products. Each results in a direction of carbon flow away from pyruvate production (Figure 8). Carbon flux away from pyruvate in this manner may result in production of glycerol, 1,3-propanediol, 1,2-propanediol, and D- lactate.

1-16 Figure 8: Products and pathways branching directly from glycolysis. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP glycolysis. Table 6: Product and cofactor molar yields of the glycolysis side-branches with respect to glucose (Glu) and EMP Glycolysis. Yield Glycerol 1,3-Propanediol 1,2-Propanediol D-Lactate

-1 Y (Product·Glu ) 2 2 2 2 Y (ATP·Glu-1) 0 0 -2 -2 Y (e-·Glu-1) -4 -8 -8 0

The glycerol pathway is active in many species during sugar fermentation, most notably by Bacillus88 and Saccharomyces89,90. ATP is generated in this pathway, but the phosphorylation steps of glycolysis cancel this ATP formation (Table 6). Additionally, this pathway is reversible, and glycerol can be readily consumed by many bacteria for energy production. Since it is a reduced substrate compared to glucose (Figure 1), anaerobic catabolism of glycerol requires increased production of reduced products. Some species, including species of Klebsiella, Clostridia, Citrobacter, Lactobacillus, and Enterobacter, deal with the increased load of reduced EMCs by directly reducing some of the glycerol to 1,3- propanediol91. However, no species are known to be able to ferment sugars through glycerol to 1,3-propanediol due to sugar repression of the 1,3-propanediol pathway92.

The methylglyoxal bypass is a highly conserved pathway utilized in conditions of phosphate limitation and when carbon flow into glycolysis is greater than the processing capacity of the cell’s metabolism93,94. However, methylglyoxal is highly toxic, and so detoxification mechanisms are necessary to avoid cell death. The most common pathway for dealing with methylglyoxal is hydration to D-lactate, which can then either be isomerized to L-lactate via lactate racemase95, or oxidized to pyruvate94. Some species of Clostridia are also able to detoxify methylglyoxal by reducing

1-17 it to 1,2-propanediol96–98. There is no additional ATP generation associated with the methylglyoxal bypass, though conversion to 1,2-propanediol provides an effective electron sink (Table 6).

1.2. Feedback Mechanisms of Electron Mediation

All cells use a variety of molecular signalling pathways to respond to and interact with their environment. These include metabolic regulators, such as the release of hydrolytic enzymes in response to the presence of cellulose, and anabolic regulators, such as the signals involved in regulating the cell cycle. Cells also possess a set of metabolic regulatory enzymes which specifically sense changes in the redox state of the environment and alter the metabolic flow of electrons accordingly. Some detect changes in the ratio of reduced to oxidized EMCs99, while others detect the presence of reduced products100. Another signalling system regulates the transcription of respirative and fermentative genes in facultative anaerobes in response to the presence or absence of an external electron sink101. These signalling mechanisms make up the range of redox-responsive metabolic controls.

Repressor Enzymes

Redox repressor (Rex) enzymes respond to changes in the ratio of NADH:NAD+ by regulating the transcription of genes that code for NADH oxidising enzymes99,102,103. They regulate the oxidation state of the NAD pool and control product formation by binding to operons upstream of genes which code primarily for hydrogenases and cytochromes when the NADH:NAD+ ratio is low, preventing production of enzymes that oxidize NADH (Figure 9). Forms of Rex have been found to be highly conserved across many species of bacteria, including aerobes and anaerobes, and span a wide range of distantly related taxonomic groups104. In function, it has been found that Rex enzymes are either inhibited by NADH99 or activated by NAD+102, though the outcome of down-regulating NADH oxidation is the same.

In addition to regulation based on the NADH:NAD+ ratio, there is recent evidence for redox potential regulating enzymes which respond to concentrations of reduced products and down regulate NADH dehydrogenases100. In this work, Pei et al.100 also describe regulation of the ethanol pathway of Thermoanaerobacter ethanolicus based on NADH:NAD+ ratio. Further evidence for redox potential control based on interactions of NADH:NAD+ ratio andreduced products, specifically butanol, is also available105,106. These complex interactions between the NADH:NAD+ ratio, reduced product concentrations, and expression of cytochromes and NADH oxidizers maintain a stable redox potential in the cell

1-18 Figure 9: Diagram depicting the repression mechanism of Rex enzymes for a generic dehydrogenase gene. NAD+ binds with the Rex enzyme, which changes to an active conformation. The active Rex enzyme then binds to the operon upstream of the dehydrogenase gene, blocking transcription by RNA polymerase. and, similar to ATP pool maintenance, ensure that the cytoplasm maintains an appropriate level of NAD+ for reducing power in the energy metabolism.

NADH:NAD+ Response to Environmental Factors

Previous research on fermentation using Enterococcus faecalis found a correlation between the NADH:NAD+ ratio, culture pH, and relative reduction potential of the substrate107,108. Using combinations of mannitol, glucose, pyruvate, and gluconate, it was found that generally the highest NADH:NAD+ ratios occurred with the most reduced substrate (with the exception of gluconate). Additionally, NADH:NAD+ ratio decreased with decreasing pH. Several studies have also shown increased production of reduced products as a response to fermentation of more reduced substrates such as glycerol109,110, xylose89, and mannitol111.

Taking into consideration the recent research on Rex enzymes and enzymes that respond to the redox state of the products, it can be assumed that the NADH:NAD+ ratio is optimized within the cell for the most thermodynamically efficient breakdown of the substrate, and stabilized at that ratio by the redox repressor enzymes, given the range of both metabolic pathways and switch points of repressor enzymes. This requires extension of the basic assumption of thermodynamic optimum30 to also consider constraints imposed by regulation mechanisms (but still treating the system as an optimization problem).

1-19 Response to Changes in External Redox State

Many facultative anaerobes also rely on redox-sensing enzymes to respond to changes from aerobic to anaerobic conditions. The ArcA-ArcB system of Escherichia coli responds to increases in the reduced forms of membrane-bound quinones to down regulate transcription of respiratory genes and up regulate transcription of fermentative genes101. In the Fnr system, also of Escherichia coli, a transcription factor is bound to several Fe-S clusters 112 which oxidize or reduce depending on the presence of O2 . This is used in turn to regulate many fermentative genes in anaerobic conditions. Bacillus subtilis responds to changes in nitrosative stress via the ResD-ResE system, which responds to changes in highly reactive nitric oxide (NO) concentration by regulating the transcription of various genes involved in aerobic and anaerobic metabolism113, as well as up-regulating flavohemoglobin production - to either oxidize NO to NO3 in aerobic conditions, or reduce NO to N2O in anoxic conditions114.

Clearly, redox sensing enzymes play a vital role in regulation of cellular metabolism and in maintaining a stable redox potential within the cell. In addition to stabilizing the NADH:NAD+ ratio, they respond to fluctuations in the redox state of the quinone pool, concentration of

O2, and presence of highly reactive oxidized nitrogen. This complex set of redox regulators shifts metabolism between aerobic and anaerobic, and maintains a ratio of NADH:NAD+ that is thermodynamically favorable for breakdown of the available substrates by moving electrons between EMCs, H2, and reduced products. It is likely that successful control of the fermentation product spectrum will rest on controlling the activity of the redox repressor enzymes.

1.3. Environmental Influences on Metabolic Regulation

It is currently known that pH4,22, substrate concentration20,25,82, substrate type81,109,110, and temperature5,61,115 act as important controls on both the product spectrum and microbial population diversity. Each interacts with different aspects of the cellular metabolism or system ecology and results in formation of specific ranges of products based on these interactions.

Regulation with pH

No environmental control on the fermentation product spectrum has been more widely researched than pH. The pH affects the extent of dissociation of organic acids produced by the cells, which in turn effects the amount of energy (as ATP) a cell must expend in

1-20 maintenance8. Undissociated organic acids (at low pH) passively diffuse into the cell, thereby requiring an increased rate of ATP-consuming active transport to remove the acid products and maintain intracellular pH. It is possible that this increased energy consumption signals cells to change their primary mode of ATP generation from the acetate pathway to the lower ATP-yielding butyrate pathway at low pH. Butyrate production has a reduced impact on pH per mole glucose due to the stoichiometry of reaction. It therefore reduces active transport energy expenditure. Conversely, at high pH, where passive diffusion of acid products is blocked due to extracellular acid dissociation, cells can rely primarily on acetate production for ATP production while mediating the concurrent production of excess electrons through production of reduced products, such as ethanol.

The metabolic shift from acetate and ethanol at high pH to acetate and butyrate at low pH has been well documented in mixed cultures4,22. This production shift is also the most commonly accepted norm for pH-controlled glucose fermentation in substrate-limited conditions, and has been well-modelled27. However, it has also been reported that mixed cultures will respond to low pH by increasing pH-neutral ethanol and maintaining ATP generation through acetate production24. It is not fully understood what causes mixed cultures to react to low pH by producing butyrate in some cases and ethanol in other cases, though it may be linked to substrate type. Lu et al.22 and Temudo et al.4 each used glucose as their only carbon source (resulting in butyrate at low pH), while Ren et al.24 used a combination of molasses and sucrose (resulting in ethanol at low pH). Similarly, the product spectrum at high pH has also been shown to sometimes include high propionate production in place of ethanol21. This production mode still allows for electron sinking from acetate production. It was suggested in the study that this product shift was the result of a taxonomic shift, though this claim was not evidenced. While the exact mechanisms responsible for pH- influenced product shift are not yet fully understood, the strong influence of pH on the product spectrum is clearly evident. However, current understanding of the effects of pH are limited to substrate-limited conditions. In Chapter 4, low pH fermentation is described in both substrate-limited and non-substrate-limited conditions.

Regulation through Substrate Supply

Substrate concentration is often conflated with hydraulic retention time (HRT) into the lumped factor of organic loading rate (OLR)20,116,117, which describes the rate of supply of the substrate. While this is useful as a descriptive term, and substrate concentration and hydraulic retention time (HRT) both affect the availability of substrate in continuous systems, the two factors affect microbial processes in different ways118. Substrate concentration has

1-21 a direct effect on metabolic expression and energy availability, while HRT controls substrate supply, removal of end products, and applies selective pressure based on growth rate.

When substrate availability is low, cells preference metabolic pathways with high ATP yield, such as acetate and butyrate61. Alternatively, excess substrate availability can cause inhibitory effects either through product inhibition8,100,119 or through build-up of toxic metabolic intermediates94. Research into the effect of substrate concentration on mixed cultures has so far focused on defining the limits of methanogenic activity in anaerobic 20 digestion, as well as on high rate H2 production. Eng et al. described an increase in lactate and propionate levels in response to increasing the substrate concentration, or shock 120 loading, using sucrose in anaerobic digestion processes. Within H2 production, Wu et al. , by varying both substrate concentration and HRT, reported that the H2 production rate is highly dependent on substrate concentration. In this study, H2 production increased by as much as 9-fold with a 4-fold increase in substrate concentration. They also reported increased VFA production with high substrate concentration, though the description of the VFA spectrum was limited.

Voolapalli and Stuckey82 induced shock loads decreasing HRT at constant substrate concentration and, similar to Eng et al.20, reported increased VFA production in response to the shock loading. However, they also did not differentiate between the acid products. In contrast, Agler et al.121 varied HRT while holding OLR constant (varied substrate concentration) and reported increased production rate of butyrate and other unspecified fermentation products with decreased HRT. This effect may result from the washout of methanogens from mixed cultures at low HRT, favoring fermenting organisms5. In general, there has been limited description of the effects of loading (either through HRT or substrate concentration) on the fermentation product spectrum, or how varied loading influences metabolic expression. Chapter 4 of this thesis explores this space through variation of substrate concentration at fixed HRT.

Regulation of the Redox Environment

A number of methods can be used to alter the redox environment of a fermenting system, including interaction with an electrode, O2 content, and content of other chemical additives. Electrode-driven redox control has garnered a great deal of attention in recent years through research into bioelectrosynthesis33,122. Application of a sufficiently negative potential to a fermenting culture via an electrode has been shown to increase the yield of reduced

1-22 products, including increasing propionate yield of Propionibacterium freudenreichii123, and increasing ethanol yield of Clostridium thermocellum and Saccharomyces cerevisiae124.

The assumed mechanism of bioelectrosynthesis is via electron uptake by the cells, and subsequent redirection of electrons to reduced end products. Cells uptake electrons via one of two mechanisms: direct and indirect electron transfer. Direct electron transfer relies on direct interaction of outer membrane redox active components (EMCs and/or bionanowires)125,126. This mechanism has been observed for anode-driven systems127, and has been postulated in cathode-driven systems122,128,129, but thus far there has been no confirmation of this mechanism in fermentation systems. Indirect electron transfer occurs when redox active shuttling molecules transport electrons between microorganisms and the electrode via diffusion through the culturing media. The shuttling molecule, or electron mediator, can be secondary metabolites such as phenazine130 and flavins32, hydrogen or oxygen from hydrolyzed water (produced at electrode surface)33,131, or added synthetic molecules123,124.

Anthraquinone 2,6-disulfonic acid123, cobalt(III) sepulchrate123, methyl viologen132–134, and neutral red124,135,136 have all been successfully used as synthetic electron mediators in cathodic systems. However, the cellular response of reduced product formation is not always consistent with the measured consumption of electrons. Emde and Schink123, for instance, accounted for 83% of the measured current through reduced product formation, and explained the remaining difference with biomass formation. By contrast, Kracke137 recorded a reduced product formation of 4 times greater than the measured current. There have been no mechanistic studies on these systems thus far, and so it is not clear why the cellular response may be greater than the stimulus.

From a control viewpoint, use of selection of a set redox potential via a synthetic mediator appears to be the best method for regulating extracellular redox state in fermentation systems. Since each carrier has a different reduction potential (and stoichiometry), the regulation is based on the different energy levels at which electrons interact with the cells, and by extension, the redox-active cellular functions (Figure 10). As shown by Emde and Schink123, the use of mediators with more negative reduction potentials results in increased yield of reduced products in pure culture fermentation. Cathodes without mediators have also been used in mixed-culture fermentation138. However, no reports of mediated mixed- culture electro-fermentation could be found at this time. Chapter 5 of this thesis investigates the mechanisms for reduced product formation as induced by redox mediator compounds.

1-23

Figure 10: Standard reduction potentials (E°’, mV vs. SHE) of common electron shuttle molecules41,123,145,146 and metabolic half-reactions at pH 7 and an ionic strength of 0.10 M33,40,41.

1-24 139 140,141 Exposure to redox chemicals, such as O2 and methyl viologen are known to activate the oxidative stress response, which alters the routes of electron mediation within the cell101,112,139. The end result of mild oxidative stress is a change in the product spectrum to less reduced products139,140. Similarly, exposure to the various oxidation states of nitrogen142 and sulfur143,144 alter the electron mediation mechanisms. Highly oxidized nitrate, for instance, has been shown to alter Clostridium metabolism from acetate-butyrate-ethanol- producing to primarily acetate-producing142.

Regulation with Substrate Type

As discussed above (Section 1.3.1), the difference in substrate type between the experiments performed by Lu et al.22, Temudo et al.4, and Ren et al.24 may explain differences in production of ethanol vs butyrate at low pH. Different substrates have different and distinct degrees of reduction, and therefore generate different quantities of reduced EMCs during catabolism. Additionally, each substrate enters the fermentation metabolism at a different point (Figure 2), and therefore has a unique set of pathways and yields available to it. Certain substrates also can be inhibitory to particular pathways, such as sugar inhibition of the 1,3-propanediol pathway92.

Work with pure cultures has shown extensively that substrate type has a strong effect on the product spectrum. Lewis and Yang81 demonstrated the effects on product spectrum of Propionibacterium acidipropionici grown on glucose, lactose, and lactate. Propionate and acetate production were highest and succinate production lowest when lactate was the fermentation substrate, while lactose resulted in the highest succinate production. Yu et al.147 demonstrated how the substrates lactose and glucose affect fermentation gene expression in Clostridium acetobutylicum. They showed not only that genes related to lactose fermentation were only expressed in the presence of lactose, but that when provided a mix of glucose and lactose, the lactose fermentation-related genes were not expressed until glucose had been sufficiently consumed. This is evidence that the effect on product spectrum is at least a result of an adjustment in the metabolism. Girbal and Soucaille105 and Snoep et al.107 took this a step further, showing that substrate type affects the NADH:NAD+ ratio of C. acetobutylicum and Enterococcus faecalis, respectively. However, their results were inconclusive as to whether the change in ratio was directly correlated to the relative redox state of the substrate.

In addition to the effect caused by the substrate redox potential, substrate type can also affect the system ecology. Zhang et al.148 showed that the bacterial microbiomes of

1-25 anaerobic digesters are highly correlated to substrate type. One likely explanation for this is that many microorganisms do not possess the necessary catabolic pathways for certain substrates. For example, Ethanoligenens harbinense, and many phylogenetically similar species, cannot consume inositol, glycogen, mannitol, arabinose, or inulin, and only weakly consume starch149. Additionally, it is well known in the bioethanol industry that wild-type strains of Saccharomyces cerevisiae are not able to ferment xylose150,151. The inability of a species to digest a substrate provides a selective pressure against that species in the mixed culture.

Regulation with Temperature

Similar to the ecological effects of substrate type, it is well known that system temperature has a profound effect on the ecology of mixed culture systems, providing selective pressure for psychrophiles below about 20°C, mesophiles around 30-40°C, and thermophiles above about 50°C152. Additionally, temperature directly affects the activation energy of a bioreaction, as well as thermodynamics of equilibrium. As shown in studies on ethanol production with Saccharomyces cerevisiae, total substrate consumption as well as ethanol yield and production rate increase with increasing temperature, while yield of volatile by- products (e.g. acetaldehyde, isobutanol, VFAs) decrease with increasing temperature153,154. Infantes et al.155 also showed increased fermentation rate with increasing temperature with mixed cultures, however the individual product responses (VFAs and H2 only) were more heavily based on pH than on temperature. In general, oxidative reactions (producing hydrogen as reduced product) are favored at high temperature over reductive reactions5.

A secondary thermodynamic effect of temperature is on cellular membrane permeability. Bischof et al.156 showed this effect using a fluorescent dye to track membrane permeability versus temperature using two types of animal cells. Infantes et al.155 also hypothesized that increased membrane permeability was responsible for incomplete substrate consumption at high temperature and low pH. No references could be found that assess temperature- dependent membrane permeability studies on fermenting micro-organisms.

1-26 2. Thesis Motivation

As discussed in Chapter 1, there are substantial knowledge gaps in the product selection mechanisms of mixed culture fermentation. These knowledge gaps must be overcome in order to reliably and reproducibly control biochemical outputs from organic waste streams for a biochemical-from-waste industry. Control modes based on pH, substrate concentration, redox environment, and substrate type show the most promise in terms of dynamic product response, and so are logical schemes with which to start. However, since glucose metabolism in general is the most well-understood, descriptions of metabolic feedback mechanisms should start with glucose feeds, and then expand into alternate and mixed substrates. Therefore, the work in this thesis will focus on the effects of carbon and electron loading (via substrate concentration and redox environment) on the product spectrum within glucose fermentation at controlled pH. There has also generally been limited direct mechanistic analysis in mixed culture systems, with only taxonomic analysis having been applied. This thesis will expand to metaproteomics to assess the metabolic mechanisms which cause the change in product spectrum.

The product spectrum of glucose-fed mixed cultures at low pH has shown to be the most reliably reproducible, with acetate and butyrate being the primary products at this pH in multiple studies4,21,22. A pH of 5.50 in a glucose-fed system was therefore chosen as the baseline from which to compare, with acetate and butyrate as the expected baseline products.

2.1. Research Objectives

I. Describe and characterize production when culture is subjected to high carbon loads and redox stresses.

Previous studies into the effects of fermenting systems taken from substrate- limiting conditions to non-limiting conditions have focused on hydrogen production120,157, and on shock (transient) loading and how the outputs of shock loaded systems alter methanogenesis in anaerobic digestion20,82. Each of these works measured increased VFA production at high substrate concentrations, but they each either did not explore a complete product spectrum82,120,157, did not maintain high loads through steady state20, or varied substrate availability only by varying the HRT157. Additionally, none of these studies used controlled pH.

2-27 Combined, this makes it difficult to describe the mechanisms of product selection at high substrate concentration.

Notably, Ren et al.157 did report increased ethanol production with increased substrate availability, while Eng et al.20 reported increased lactate and propionate production in response to shock loads. These results suggest that cultures may respond to non-limiting organic loads either by either producing electron sink products, or by redirecting carbon flow away from ATP production in general.

Similarly, there is evidence that use of a cathode to supply reducing power in substrate limited conditions will cause both mixed138 and pure123,124 cultures to increase production of electron sink products. Emde and Schink123 demonstrated an increased yield of reduced products with mediator compounds of increasingly negative redox potential, indicating that at least some bacteria can sense an increased electron load, and will manipulate their product spectrum accordingly. However, this same effect has not been demonstrated in a mixed culture.

II. Describe the metabolic mechanisms utilized by mixed cultures for carbon and redox mediation.

It has been shown that many species of microorganisms use a variety of enzymatic signalling pathways to respond to changes in the local redox potential by adjusting the transcription levels of redox-related genes. Additionally, it is known that certain key catabolic steps, such as conversion of pyruvate to acetyl- CoA, can be performed by a number of enzymes depending on the media conditions. In each case, the enzymes that are produced affect the cytoplasmic redox potential. This research aims to show that these processes are directly involved in the changing of fermentation types and to categorize the cytoplasmic redox states that arise as their result.

For high carbon loads, the mechanisms which product to switch to reduced products, such as lactate and propionate, will be investigated. For redox stress, it has been shown that degree of reduction of the product spectrum can be much greater than the stoichiometric changes that would be anticipated by simple electron uptake137, indicating that electron uptake cannot be the only factor influencing the change in product spectrum. Metaproteomic analysis will be used to analyze the switching mechanisms in both scenarios.

2-28 The objective based on this research question is to characterize the ecological and biomolecular responses as they correspond to applied stimuli and product spectrums of the first objective. This will be accomplished through metaproteomic analysis, with focus on the regulation of electron mediation, oxidative stress, and product pathway enzymes.

III. Describe the importance of taxonomy versus functional expression in determining the product spectrum.

The product spectrum generated by mixed cultures has so far been most thoroughly described based on the taxonomy, either in relation to the dominant genus group in the system158, or with a view of the whole taxonomy22,23. This view is based on the assumption that the group which is producing the most biomass must also have the greatest influence on the product spectrum. However, this ignores the metabolic functional range of expression often described as a main benefit of mixed versus pure culture, as well as the ability of cells to consume intermediates and products from other cells for further processing.

An alternative view of mixed culture metabolic processes is that of the mixed culture model metabolism8. While this view incorporates pathways from the entire culture, it ignores individual species at the functional level and instead views product selection through the lens of metabolic pathways. However, this approach assumes that all pathways are equally available at under all conditions and does not consider specific restrictions.

In order to expand on the model metabolism approach used by Rodriguez et al.8, mechanistic information obtained through metaproteomics will be used to describe the regulation of the various product pathways available to the mixed culture. This approach will be integrated with the current methods of taxonomic description and unregulated mixed culture metabolism.

2-29 3. Research Methods

This chapter describes the experimental setup and methods used to pursue the objectives of this thesis. Two primary series of experiments were carried out, one investigating substrate concentration, and the other investigating mediated electro-fermentation. The two series used similar experimental setups, with some key differences highlighted in the following text.

3.1. Media and Culture

Varied Substrate Concentration Media

The feed for the substrate concentration tests was split between substrate, nutrients, and water to enable constant hydraulic retention time (HRT) and nutrient feed rate while varying the glucose concentration. Each feed bottle was prepared in 10L Schott bottles, 9L per batch.

Substrate

For the substrate feed, 200 g·L-1 of glucose was mixed into a Schott bottle. The bottle was then sterilized in at autoclave at 121°C for 30 minutes. To ensure an anaerobic state, the bottle was removed from the autoclave at 75°C and immediately sealed and sparged with 99% argon through a sterile 0.22µm filter for 1 hour. The bottle was then fitted with a flexfoil gas bag (Air-Met, Melbourne, Australia) of 99% argon to maintain atmospheric pressure.

Nutrient Media

A basal anaerobic (BA) nutrient medium, as defined by Bastidas-Oyanedel et al.159, was prepared at 4x concentration, excluding vitamins and Na2S. This solution was then sterilized and sparged in the same fashion as the substrate solution. After the bottle was cooled to below 50°C, concentrated vitamins and Na2S solutions were added through a sterile 0.22 µm filter.

Water

Purified RO water was sterilized and sparged in the same fashion as the substrate and nutrient solutions.

3-30 Activation Media

Tryptone-glucose-yeast extract (TGY) medium was prepared in 100 mL aliquots in 165 mL serum bottles. The bottles were sealed with a butyl rubber septum, heated to 75.0°C, and then the headspace was flushed with 99% argon gas at 100 mL·min-1 for 5 minutes. The bottles were then autoclaved for 15 minutes, and the headspace was immediately flushed with argon again while the still hot, this time through a sterile 0.22 µm filter. After cooling to room temperature, the bottles were flushed one final time with sterile argon for one minute, and then vented to atmospheric pressure using a monometer. Sterile, anaerobic bottles were then stored at 4°C until use.

Inoculation Media

1 Initial reactor inoculation from TGY activation medium used BA medium at /5 of the continuous salt and nutrients concentration with 10 g·L-1 glucose. This ensured high initial growth rate and low saline shock upon inoculation160. For the varied substrate concentration experiment, this inoculation medium was fed into the reactor before inoculation as described in Section 3.2.3.

Batch Activity Media

Fermentation tests in 165 mL serum bottles used 100 mL 1xBA media with 0.5 g·L-1 glucose. Bottles were prepared, deoxygenated, and sterilized by the same method as the activation media. As with preparation of the nutrient media, Na2S and vitamins were added through a sterile 0.22 µm filter after autoclaving. Initial nutrient and substrate preparation was at 1.1x concentration, and bottles were filled to 90 mL to allow for addition of 10 mL of 0.5 M MES buffer through a sterile filter as the final preparation step. MES buffer solution was prepared with sterile, anoxic MilliQ water in an anaerobic chamber, and the pH was adjusted to 5.50 with 4 M NaOH.

Electro-fermentation Media

Media for the electro-fermentation tests was supplied to each reactor from a single feed bottle. Bottles were prepared in similar fashion to the varied substrate concentration nutrient media, using 1xBA media and 5 g·L-1 glucose substrate. A triplicate of feed bottles was prepared simultaneously for each mediator compound. To ensure that each triplicate was as internally consistent as possible, the nutrients and substrate for all three bottles were measured out together, and then split between bottles with vigorous mixing using volumetric flasks. RO water was then added to each bottle to 9L, using volumetric flasks. Concentrated

3-31 mediator solutions were added to a final concentration of 100µM alongside the vitamin and

Na2S solutions. Inoculation medium for electro-fermentation was identical to that of the varied substrate concentration experiment, and did not contain any mediator. This allowed for gradual build-up of mediator concentration in the reactor at the start of continuous feeding.

Culture

A diverse mixed culture was required for these experiments in order to ensure exploration of the interacting effects between functional groups. Granules from the upflow anaerobic sludge blanket digester at Golden Circle Cannery in Brisbane, Australia were chosen as the base culture. This digester is fed on carbohydrate-rich pineapple canning wastewater, and so the culture is well suited for fermentation of glucose.

The culture was collected less than 48 hours before beginning the first experiment, and activated overnight in TGY media at 30°C. To activate, granules were crushed in a plastic bag by hand using a roller in order to break up the granule structure while not affecting culture activity161. The crushed granules were then injected into pre-warmed TGY bottles. To ensure high growth among both sporulated and non-sporulated cells, the culture was activated in a pair of TGY bottles, one of which was heat shocked for 10 minutes at 75°C.

To ensure continuation of the same culture throughout the experiments, broth samples were taken at weekly intervals from the fermenter and stored in 20% glycerol at -80°C in 2 mL aliquots. In the event of reactor failure, the reactor was re-started from the most recent broth sample, activated overnight in TGY media at 30°C. At the end of the varied substrate concentration experiment, a set of 30 samples were taken and stored in this fashion for use in the electro-fermentation experiments.

Reactors were inoculated with 1% v/v activated TGY media in all cases. For start-up of the varied substrate concentration experiment, where both heat-shocked and non heat-shocked activated granules were used, the total inoculum was split evenly by volume between the two activation types. These two separate inocula were injected into the reactor, one immediately following the other. For all inoculations, culture injection was via needle and syringe through the liquid sampling port, with the needle tip submerged to minimize physical stress on the cells.

3-32 3.2. Bioreactors

Varied Substrate Concentration

In order to accomplish the continuous system objectives of this thesis, an automated CSTR bioreactor was constructed (Figure 11). The bioreactor was a tempered glass vessel at 1.2 L working volume, with ports on the top and sides, and was manufactured by the UQ Glassblowing Service. The bioreactor was constantly stirred at 200 rpm with a 40 mm stir bar on a stir plate (Big Squid, IKA, Germany). Gas and liquid samples were taken through blue chlorobutyl septums (Bellco Glass, USA). Feed was combined from three separate vessels (water, 200 g·L-1 glucose solution, and 4x BA minerals and nutrients), each with a drip immediately coming out of the bottle, before passing through a second drip at the top of the bioreactor to prevent back contamination. Biogas was vented through from the headspace to a fume extraction hood, through a gas line of 40 mL total volume. This ensured no back pressure in the vessel, and no oxygen contamination when sampling. The remaining controls and sensors are described in Section 3.3 below.

Electro-fermentation

Due to limitations associated with incorporating electrodes, smaller (340 mL) working volume vessels were used in the electro-fermentation experiments (Figure 12). Thesevessels were run in much the same way as in the varied substrate concentration experiments, except that feed was supplied from a single vessel. This was feasible because of the short time scale of these experiments (2 weeks), where back-contamination was unlikely. Additionally, in these vessels, there was not space for a volume-based switching mechanism to control the effluent, so a drop-pipe set at the desired volume was used instead. Biogas was collected in flexfoil gas bags to ensure no back pressure in the vessel, and no oxygen contamination when sampling.

Reactor Sterilization and Preparation

Two days prior to inoculation, the fermentation vessel was completely assembled, except for an air-tight cap in place of the pH probe. The reactor was sterilized by pumping 100% ethanol in through the effluent lines until full, allowed to sit in contact with stirring for 1 hour, then pumped out through the effluent line while flushing with 99% argon through a sterile 0.22 µm filter. Argon flushing was maintained until inoculation. The ethanol was rinsed out with sterile MQ water by the same method, and then the reactor and lines were allowed to dry overnight while activating the temperature control (Section 3.3.2). Feed and effluent lines

3-33 Figure 11: Bioreactor setup for varied substrate concentration experiments.

Figure 12: Bioreactor setup for electro-fermentation experiments.

3-34 were then connected to their respective bottles under flame, and the reactor was filled to the inoculation composition from the feed bottles.

Preparation of the electro-fermentation vessels varied slightly from that of the varied substrate concentration vessel. Due to the use of single feed bottles, inoculation medium was prepared and sterilized directly in the reactors. To ensure consistency between reactors, the media was prepared in the same method as the electro-fermentation media (Section 3.1.2), filling each reactor to a final volume of 340 mL. Reactors were also sterilized with cathodes inserted and made anaerobic using the same method as for the media bottles. Feed lines, which could not be autoclaved due to the pump connectors, were sterilized, flushed with sterile argon, and connected to the feed bottles and reactor vessel by the same method as the varied substrate concentration reactor.

Sterilization of the vessels and feed lines was essential to ensure continuity of the culture throughout the experiments, especially following biofilm removal (Section 3.2.4) and during electro-fermentation start-up. For all reactors, pH probes and additional electrodes were inserted with continued sterile argon flushing. Immediately before insertion, these were each cleaned with a lint-free wipe wetted with 100% ethanol. Finally, for the electro-fermentation vessels, the effluent drop pipe was adjusted to the level of the media to set the reactor volume at 340 mL.

Biofilm Removal

The varied substrate concentration reactor required regular removal of the biofilm. This was done immediately following each steady state period, as well as pre-steady state around the time product composition began to approach stabilization. After pre-steady state biofilm removal, at least 1 week, but not more than 2, was allowed for the reactor to reach steady state. Electro-fermentation reactors did not require this step due to the short experimentation time. The biofilm was removed to ensure measured results were only representative of free- cell activity. This was necessary because steady state sampling of the biofilm would not have been possible without disturbing the steady state conditions.

In order to remove the biofilm, a temporary holding reactor was constructed, sterilized, and made anaerobic in identical fashion to the experimental reactor, except for lacking an effluent control mechanism and pH and temperature sensors. Once prepared, reactor broth was transferred quickly to the holding reactor through the effluent port by slight pressurization of the headspace with sterile 99% argon. The experimental reactor and components were cleared of biofilm with a bottle brush and 100% ethanol. Sampling

3-35 septums were also replaced at this time. The reactor was then sterilized and rinsed as in Section 3.2.3. After flushing with sterile 99% argon for 1 hour, broth was returned to the experimental reactor by the same method, and pH and effluent controls were re-enabled.

In both experiments, only the planktonic community was assessed, and biofilm activity was assumed to be insignificantly different from the planktonic community. This is assumption is supported by the active biofilm removal strategies, which ensured that the biofilm was never allowed to grow for more than 2 weeks before the start of a steady state assessment period. These were similar biofilm removal strategies to those employed in comparable studies which assumed a planktonic community for the purpose of community assessment23,158. However, it is necessary to note that in both the varied substrate concentration experiment and the electro-fermentation experiment, biofilm growth on the reactor walls and on the probes and electrodes will have played an active role in the observed activity. Additionally, though these biofilms were never allowed to form more than 2 weeks before the start of a steady state period, this does not guarantee that the biofilm was composed similarly to the planktonic community, or that it had similar catabolic function.

3.3. Controls and Process Data

Initially, LabVIEW was used for both process automation and data logging. LabVIEW was run on a Dell Inspiron 3000 Series laptop running Windows 8.1 and interfaced with the bioreactor via an iUSBDAQ (HYtek Automation, Canada). The controls were run on a 1000 second cycle, where feed pumps were turned on for an adjustable number of seconds each cycle and pH checked and adjusted once every 25 seconds. A level switch controlled the effluent pump. Process data, including pH, temperature, outgas flow rate, and feed and effluent bottle levels, was measured once every second, and then averaged and logged every 5 minutes. The entire control and instrument system was backed up with a 1200VA uninterruptible power supply (CyberPower Systems, USA).

LabVIEW eventually proved unstable as a control system when operated in this fashion. As such, the system control functions were re-programmed onto and run via an Arduino (Mega 2560, Arduino, Italy), while data logging functions remained in LabVIEW. This allowed for continuous and uninterrupted control for the duration of the experiments, as well as the ability to run multiple reactors simultaneously. Control system and data logging code is available on at https://github.com/RDhoelzle/Fermenter_Controls. The signal processing and control system is diagrammed in Figure 13.

3-36 232 connections directly directly - 232 connections was removed and the load cell and amplifier circuit was replaced with lab balances with RS with balances lab with replaced was circuit amplifier and cell load the and removed was

experiment. Multiple load cell signal amplifiers and pump control switches were setup in parallel as required by the the by as required in parallel setup were switches control pump and amplifiers signal cell load Multiple experiment. concentration substrate varied fermentation experiments, the float switch float the experiments, fermentation - Circuit diagram of of diagram Circuit

13 : Figure Figure electro the In experiment. dot. a black by represented are lines Connected computer. the to

3-37 Flow Rate

Feed and effluent bottles were positioned on weighing balances, which were interfaced with the iUSBDAQ to track the flowrate via change in mass. Initially, a custom hanging balance loading cell was built. However, these were replaced after the varied substrate concentration experiment with benchtop balances with RS232 interface (EM-30KAM, A&D Weighing, Australia) for better signal stability. The RS232 interface allowed for direct signal processing by the computer in LabVIEW.

The bioreactor was fed via dosing pumps (PR-1 rinse pump, Seko Spa, Italy) controlled as described above. All pumps used in this system were of the same make and model. Feed was separated from the bioreactor just prior to entering one of the top ports by a glass drip.

The initial effluent setup utilized a u-bend tube extending from the top side-port of the bioreactor, and level was controlled by overflow into this tube. The suction created from this overflow method meant that the gas phases of the bioreactor and the effluent bottle had to be linked in order to measure gas production rate. However, since the effluent continued to generate off gas, this was not a suitable method for measuring gas flow rate.

To solve this, an effluent pump was installed, and controlled using a stainless steel float switch mounted in the central top-port of the bioreactor. Effluent was drawn from the bottom side-port, and the effluent line was separated from the bioreactor via in-line check valve. This method allowed for ±10 mL control of the bioreactor volume and separation of the gas phase and effluent.

The electro-fermentation bioreactors were too small to use the same effluent control method, and so instead utilized a drop tube set to the desired level. This unfortunately meant that the gas phase could not be separated from the effluent, and so gas production rate could not be measured.

Temperature

1 Temperature was controlled at 30.0±0.5°C with /4 inch ID PVC tubing, wrapped in serpentine fashion around the bioreactor, which was attached to the output of a water bath heater (TH7100, Ratek, Australia). Temperature was measured using a RTD temperature probe (EI-1022, Electronic Innovations Corp, USA), which was calibrated against an ethanol laboratory thermometer.

3-38 pH

The pH was controlled at 5.50±0.01 with a miniCHEM pH meter and intermediate junction pH sensor (TPS, Australia). The pH meter was interfaced in parallel with both the Arduino and the iUSBDAQ for control and data logging. When below the set point, pH was adjusted with 4 M NaOH by switching on a dosing pump for 50 ms. This dosed an equivalent of 1-3 drops of 4 M NaOH at a time. The number of doses were logged over time and correlated to the flow rate of the NaOH solution in order to adjust the real bioreactor HRT.

The reactor pH was verified every 3-5 days by testing with a freshly calibrated pH meter and probe of the same make. Offsets in pH reading were assumed to occur linearly with time, and the pH reading and setpoint of the reactor was adjusted accordingly. Reference gel in the pH probe was replaced and the meter was re-calibrated with each biofilm cleaning (Section 3.2.4).

Set Potential

For the electro-fermentation experiments, set potential was controlled at -700 mV vs 3 M Ag/AgCl reference electrode (MF-2079, BASi, USA) via a potentiostat (1000C, CH Instruments Inc, USA). The working electrode was a 6 mm graphite rode (4cm working length) with platinum wire counter electrode immersed in PBS solution and separated by a 3cm2 ion-exchange membrane. Current was logged once per second using the potentiostat.

3.4. Analytical Methods

Headspace Gas

Headspace composition was determined by 8 mL gas sampling taken with a 10 mL gas- tight syringe (SGE, Australia). The syringe was first flushed once with 1 mL of headspace gas before taking the sample, with 10s allowed for the syringe pressure to equilibrate to 1atm. Headspace sampling always preceded other sampling to ensure no backflow from the outgas lines would skew the gas sample readings.

Samples were analyzed via a GC-TCD (2014 Series, Shimadzu, Japan), with a HayeSep Q 1 80/100 column (2.4 m length; /8 inch OD, 2 mm ID, Agilent, CA, USA) heated to 45°C. The carrier gas was 99% argon at 28 mL·min-1, and the injection volume was 1 mL controlled by a sample loop with valve. Injection temperature was 75°C and detector temperature was 100°C. By flushing experimental vessels with argon and using this GC method, potential oxygen leaks in the experiments were easily detectible.

3-39 Chemical Analysis

2 mL broth samples were taken regularly and centrifuged at 10,000 x g for 10 min to remove cell mass. Supernatant was extracted and passed through a sterile 0.22 µm filter to remove remaining cell mass. 150 µL of filtered supernatant was then combined in a HPLC vial with 250 µL insert (Agilent, USA) with 50 µL of 0.02 N sulfuric acid to make 0.005 N to ensure pH uniformity across samples. Calibration standards were prepared in the same fashion. Prepared HPLC vials were stored at -20°C until analysis, and the remainder of filtered supernatant was also stored at -20°C in a microcentrifuge tube as backup.

Samples were analyzed via HPLC (1050 Series, Hewlett-Packard, USA) using a 7.7x300 mm Hi-Plex H column with associated guard column and inline filter (Agilent, USA) heated to 40.0°C (TC-50/CH-30, Eppendorf, Germany). The mobile phase was 0.005 N sulfuric acid at 0.800 mL·min-1, which passed through an inline degasser, and the sample injection volume was 30 µL. Two detection modes were used in series to ensure separation of analysis between acids and other compounds. Acids were detected first by VWD (1050, Hewlett-Packard, USA) at 210nm. All products were then detected via RID (1047, Hewlett- Packard, USA) at 30°C and a detection range of 32×10-5 refractive index units full scale, with the tube internal diameter and length both minimized between detectors. Results from acid products detected by both methods were averaged together for the final result.

It was found that a number of minor products, including malate and succinate, interfered with glucose measurements using this method. Glucose was therefore separately measured on the same equipment under a different method, in which MQ water was used as the mobile phase and the column temperature was increased to 65.0°C. All other method specifications were identical. Because the fermentation media was inherently acidic, pH was still fixed with 0.02 N sulfuric acid to a final concentration of 0.005 N to ensure uniformity between samples and calibration standards.

For the electro-fermentation analysis, it was found that the mediator compounds interfered with the RID analysis. To account for this, new calibration standards were prepared for each mediator set containing 100 µM of the appropriate mediator compound.

Biomass

The cell pellet from the chemical analysis sample was washed in 2 mL of 4°C PBS solution and centrifuged again at 10,000 x g for 10 min. The supernatant was discarded and the pellet was resuspended again in 2 mL of 4°C PBS solution. Resuspended biomass was then

3-40 analyzed for total protein content using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Inc, USA) in triplicate according to the manufacturer’s protocol.

Once per week, 10 mL of broth was analyzed for total suspended solids and volatile suspended solids (TSS/VSS) in triplicate according to the standard methods162. This value was correlated to the total protein content to obtain grams dry weight of biomass (VSS). The 8,163 mixed culture fermentation biomass molecular formula of C5H9O2.5N1 was used for carbon and COD balancing.

Chemical Oxygen Demand

Chemical oxygen demand (COD) was calculated from the theoretical COD values of constituent components, represented as CnHaObNc, in which the moles of O2 per mole of component required for oxidation to CO2 is defined by the following equation:

3 = + × 4 2 4 Eq. 1 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑎𝑎 𝑏𝑏 𝑂𝑂2 𝑔𝑔 𝐶𝐶𝐶𝐶𝐶𝐶 � �𝑚𝑚𝑚𝑚𝑚𝑚� �𝑛𝑛 − − 𝑐𝑐� 𝑀𝑀𝑀𝑀 � �𝑚𝑚𝑚𝑚𝑚𝑚� COD balances were verified by both comparison of total products and unconsumed substrate to substrate fed, and comparison of COD of soluble components (SCOD) to measured SCOD. Measured was determine by potassium dichromate colorimetric assay (14555, Merck Millipore, MA, USA) and measured on a Spectroquant® Move 100 colorimeter (Merck Millipore, MA, USA). SCOD measurements used the same samples as those for HPLC measurement, and dilutions for colorimetric measurement were made with MQ water by mass using a 3000±0.1 g analytical balance (A&D, Japan). Similarly, biomass COD composition was verified by difference of SCOD and total COD (TCOD), measured with the same colorimetric assay. All colorimetric measurements were assayed in triplicate.

Cyclic Voltammetry

Cyclic voltammetry (CV) was run for each mediator compound at 1 mM in abiotic fermentation broth, composed of BA media solution with 1 mL·L-1 each of acetic acid, butyric acid, and ethanol. The pH was adjusted to 5.50 with 4 N HCl and measured again at the end to ensure no pH drift. Temperature was held at 30.0°C. CVs were performed in the electrofermentation CSTRs using the potentiostats and electrodes as described above. The scan rate was 50 mV·s-1 and measured 12 sweeps from -1000 to +200 mV vs 3 M Ag/AgCl, logging current at every mV.

3-41 The final 3 forward sweeps were averaged together to reduce the signal noise, as were the final 3 reverse sweeps. The potentials of the forward and reverse peaks were then averaged together to define the standard reduction potentials of each mediator compound.

16S rRNA Amplicon Analysis

Samples for taxonomic analysis were taken 3 times per steady state period. Biomass was taken and prepared similarly to the cell mass density samples. Final resuspension was in 1.0 mL of 1xPBS, and samples were stored at -20°C until analysis. DNA was extracted using the FastDNA™ SPIN Kit for Soil (MP Biomedicals, Santa Ana, California) per the manufacturer’s protocol. The 16S rRNA gene was PCR amplified with iTAG 16S 926F and 1392wR primers164 and Taq 2X Master Mix (New England Biolabs Inc., Ipswich, MA, USA), and DNA extraction and gene amplification were verified by agarose gel electrophoresis. Pyrosequencing was then performed on the raw DNA extracts by the Australian Centre for Ecogenomics (University of Queensland, Brisbane, Australia) via Illumina sequencing, with amplification via the same primers. Pyrosequencing results were processed and interpreted by a combination of QIIME and Ampvis as described in Section 3.5.4 below.

Metaproteomics

Protein extraction solution consisting of 77 mg dithiotheretol (DTT) and 1 tablet of complete Protease Inhibitor Cocktail (Roche, Switzerland) was dissolved into 10 mL of B-PER II Bacterial Protein Extraction Reagent (Thermo Fisher Scientific Inc, USA). Extraction solution was used within 48 hours and stored at 4°C to prevent oxidation of DTT. Samples for metaproteomic analysis were taken 3 times per steady state period. Biomass was prepared similarly to the cell mass density samples. Final resuspension was in 500 µL of protein extraction solution. Proteins were extracted using the method described by Grobbler et al.165.

Extracted protein samples were analyzed via the SWATH-MS modified method of Chang et al.166, which consisted of using a Shimadzu Prominence nanoLC-MS/MS connected to a Triple-ToF 5600 mass spectrometer (AB SCIEX, Framingham, MA, USA). Mobile phases A and B were 1% acetonitrile/0.1% formic acid and 80% acetonitrile/0.1% formic acid respectively. Samples (5 µg, pooled for identification, and 1 ug aliquots for individual samples) were loaded onto the autosampler maintained at 12 °C, desalted online in a 150 µm × 5 mm, 5 µm C18 trap column for 3 min with 100% mobile phase A at a flow rate of 30 µL·min-1. Chromatographic separation was performed using a Vydac Everest C18 column (150 µm ID × 150 mm, 300A, 5 µm particle size; Mandel Scientific, Guelph, ON,

3-42 Canada) with the following gradient: 10% buffer B from 0-3.01 min, 10-60% buffer B from 3.01–48 min, 60–97% buffer B from 48-56 min, 97% buffer B from 56-61 min, 97-10% buffer B from 61–63 min, and 10% buffer B from 63-70 min, at a flow rate of 1 µL·min-1. Eluted peptides were detected by TripleToF 5600 MS (AB SCIEX) coupled with a Nanospray III ion source in positive ion mode using PicoTip Emitter SilicaTips (12 cm, 10 µm diameter, uncoated; New Objective, Woburn, MA, USA). The following MS conditions were used: declustering potential of 80 V, curtain gas of 30 psi, Gas1 of 10 psi and heater interface temperature of 150 °C. For IDA, TOF MS was acquired over 350-1600 m·z-1 (0.5 sec) and MS/MS performed across 100-1600 m·z-1 (0.05 sec) with rolling collision energy. For SWATH analyses, TOF MS was acquired over 350-1600 m·z-1 (0.5 sec) and MS/MS performed across 100-1600 m·z-1 (0.05 sec) with rolling collision energy, using 26 Da windows across the mass range 350-1250 m·z-1. SWATH results were processed and interpreted for expression analysis using a combination of PeakView, ProteinPilot, and MSstats, as described in Section 3.5.4 below.

3.5. Calculations

Flowrates

Flowrates were calculated by linear regression of the full steady state data set as measured by the feed and effluent bottle balances. Density of each feed and effluent fluid was measured at room temperature via a 1000.000±0.400 mL volumetric flask (Duran, Germany) and 3000.0±0.1 g analytical balance (A&D, Japan).

Theoretical Yields

- Theoretical H2, CO2, ATP, and e were calculated based on the molar product yields as described in Chapter 1 with glucose consumption split from product formation by pyruvate (Table 7). Meanwhile, biomass-associated theoretical yields were based on the 8 stoichiometry of Rodriguez et al. , and H2 consumption was calculated as the difference in theoretical COD between glucose and biomass on a carbon mole basis, with 16

-1 gCOD·molH2 . Comparison of theoretical gas composition to measured composition served to verify stoichiometric assumptions.

The change in electron sinking capacity of a particular product spectrum was calculated using the difference in molar yields of individual products. The change in yield was then multiplied by the associated electron sinking capacity per mole of that product pathway. This - was assessed at 2 e per H2 from the stoichiometries listed in Table 7.

3-43 Table 7: Product-based pathway balance equations. All units are in moles, and Glu is glucose (C6H12O2), HPy is pyruvic acid, HMa is malic acid, HSu is succinic acid, HAc is acetic acid, EtOH is ethanol, HLa is lactic acid, HPr is propionic acid, - HFo is formic acid, and X is biomass. 1 H2 is interchangeable with 2 e .

Pathway Balance Equation

Pyruvic Acid (C3H4O3) = 2 + 2 + 2 Eq. 2

𝐺𝐺𝐺𝐺𝐺𝐺 𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻2 𝐴𝐴𝐴𝐴𝐴𝐴 Acetic Acid (C2H4O2) = + + + Eq. 3

𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻2 𝐶𝐶𝐶𝐶2 𝐴𝐴𝐴𝐴𝐴𝐴 Butyric Acid (C4H8O2) 2 = + 2 + Eq. 4

𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻𝐻𝐻𝐻𝐻 𝐶𝐶𝐶𝐶2 𝐴𝐴𝐴𝐴𝐴𝐴 Ethanol (C2H6O) + = + Eq. 5

𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝐶𝐶𝐶𝐶2 Lactic Acid (C3H6O3) + = Eq. 6

𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻2 𝐻𝐻𝐻𝐻𝐻𝐻 Propionic Acid (C3H6O2) + 2 = Eq. 7

𝐻𝐻𝐻𝐻𝐻𝐻 𝐻𝐻2 𝐻𝐻𝐻𝐻𝐻𝐻 Malic Acid (C4H6O5) + 2 + 2 = 2 Eq. 8

𝐺𝐺𝐺𝐺𝐺𝐺 𝐶𝐶𝐶𝐶2 𝐴𝐴𝐴𝐴𝐴𝐴 𝐻𝐻𝐻𝐻𝐻𝐻 Succinic Acid (C4H6O4) + 2 + 2 + 2 = 2 Eq. 9

𝐺𝐺𝐺𝐺𝐺𝐺 𝐻𝐻2 𝐶𝐶𝐶𝐶2 𝐴𝐴𝐴𝐴𝐴𝐴 𝐻𝐻𝐻𝐻𝐻𝐻 Formic Acid (CH2O2) 2 + = Eq. 10

1 𝐻𝐻2 𝐶𝐶𝐶𝐶2 𝐻𝐻𝐻𝐻𝐻𝐻 + + 6.325 = + Biomass (C5H9O2.5N1) 2 Eq. 11 𝐺𝐺𝐺𝐺𝐺𝐺 𝐻𝐻2 𝐴𝐴𝐴𝐴𝐴𝐴 𝑋𝑋 𝐶𝐶𝐶𝐶2

ATP and electron generation were used to create heat maps of energy production and electron mediation. These are split between total (glucose-to-product) yield, glycolysis (glucose-to-pyruvate) yield, and post-pyruvate (pyruvate-to-product) yield. The heat mapping is defined individually within each operational level in order to highlight the split in activity pre- and post-pyruvate processing.

Kinetic Calculations

Product and biomass yields from batch tests were calculated by linear regression from the start of biomass growth to the time of maximum biomass concentration. Yield of lactate, being first produced and then reconsumed, was instead calculated only to the time of maximum lactate concentration. Yields from each individual bottle were then averaged together for the triplicate set.

Specific substrate consumption rate (kM) was calculated by linear regression of the substrate concentration through the time span corresponding to the four points of highest immediate biomass growth rate. This was then divided by the average biomass concentration of this same time span. As with the yield calculations, kM values from each individual bottle were then averaged together for the triplicate set. Comparisons between kinetic values of different

3-44 levels were calculated with reference to the higher level value (substrate concentration, absolute value of mediator E’, or reactor number).

Bioinformatics

16S rRNA Amplicon Pyrosequencing

Operational taxonomic units (OTUs) were picked via cross-reference with the Greengenes database on 18 November, 2015 for varied substrate concentration sample analysis, and on 16 May, 2016 for electro-fermentation analysis, each at 97% similarity using QIIME. OTU sets were rarefied in Ampvis167 using the rarefy_even_depth() function based on the sample with the lowest total OTU count, and using the default random number generation seed of 712. Since 16S rRNA pyrosequencing has limited accuracy at the species level, the rarefied OTU tables were summarized based on genus before further data analysis. Genera relative abundance was calculated as the rarefied and summarized OTU count of each individual genus divided by the rarefication depth.

It must be noted when interpreting 16S rRNA amplicon results that there are inherent errors in this type of analysis resulting from amplification primer biases, non-uniform gene copies across species, and extraction inefficiencies. Therefore, while this thesis frequently refers to the relative abundance of genera, these values are in actuality the relative abundance of the measured OTUs. However, the approximation of OTUs as representative of the relative abundance of genera is considered a reasonable assumption in this system given the following: the extraction procedures were uniform across all experiments; biological triplicate analyses were conducted for each operational level; all identified OTUs were of the kingdom Bacteria.

Metaproteomics

A reference proteome was assembled from all bacteria, archaea, and fungi on the UniProt database168 on 9 November, 2015 for varied substrate concentration sample analysis, and on 6 April, 2016 for electro-fermentation sample analysis. Reference proteomes were used to map ion fragments in ProteinPilot (v5.0, SCIEX, Framingham, MA, USA). Proteins were imported into PeakView (v2.1, SCIEX, Framingham, MA, USA) for analysis of the SWATH data at 5 peptides per protein, 3 transitions per peptide, 99% peptide confidence threshold, 1.0% global false discovery rate threshold, 5.0 min XIC extraction window, and 50ppm XID width. The global false discovery rate was calculated within the ProteinPilot software using the Paragon algorithm, which compares positive protein matches from the reference

3-45 proteome to matches based on a decoy database composed of reversed protein sequences169.

Accession numbers from the PeakView results were converted to genus and protein names using the UniProt Retrieve/ID mapping tool. Species-level names were available, but genus- level identification was used for consistency with the limitations of the pyrosequencing. For each individual ion, a blank dataset was also inserted into the data table with intensity of 25,000 (roughly the total dataset mean), which was used for baseline comparisons. Ion intensity tables were then converted to the “label free LCMS” 10 column format for processing in MSstats170. This 10 column ion table was imported in whole for total culture assessment, and split by genus or by function for more detailed assessment in MSstats.

Raw ion data was processed using the dataProcess() function of MSstats with default settings, which normalize ion intensity data by the equalized medians method. Group comparisons were performed as described for the individual analyses, and adjusted p < 0.05 was used to define significant differences in expressions between groups.

Metabolic maps used only group-to-group comparisons, and did not include baseline comparisons. When comparing against the baseline, the log2-fold changes (L2FC) of each protein set were shifted up such that the reading with the minimum response was set to zero, and then all NA responses were also set to zero. This comparison was used as a basis for discussing the full proteome.

Statistics

Confidence Intervals

All error bars and confidence intervals expressed in the text are 95% confidence intervals based on two-tailed t-tests (5% significance threshold). Where confidence intervals provided in stacked histograms, they represent confidence in added totals from replicate measurements (technical replicates).

Confidence intervals for baseline proteomic comparisons were generated by re-labelling individual proteins by their function (ie butyrate-generating, glycolysis, electron mediation, etc), and then re-normalizing by genus. The full normalized protein set was then reassembled and compared as a whole to the baseline, and confidence intervals were calculated from the standard error and t-crit (based on DOF) given for each set by MSstats.

3-46 Group Comparisons and Regression

Sample groups were compared by ANOVA and Tukey’s range test in R using the aov() followed by the TukeyHSD() functions. Groups were considered significantly different for p < 0.05.

Regression analysis was performed in R using the anova() function on linear models. Individual factors were eliminated from multiple regression analysis one at a time based on the highest p-value until all p-values were < 0.05.

Principal Component Analysis

Principal component analysis (PCA) was performed in Ampvis using the amp_ordinate() function with the rarefied and summarized OTU tables. Genera with principal component (PC) loadings of > 0.50 were considered to be positively (directly) described by that PC, while those with loadings of < -0.50 were considered to be negatively (inversely) described.

Redundancy Analysis

Redundancy analysis (RDA) was was used to assess whether changes in the product spectrum could be explained by variations in total pathway expression (sum of individual protein L2FCs). RCA was performed using the vegan package in R with the rda() function. Data sets were internally normalized before RDA to ensure even weighting.

3-47 4. Function of Metabolic Shift from Butyrate to Lactate Production Triggered by Substrate Concentration

High lactate and propionate formation have previously been noted in transient states and increased substrate loading. These phenomena have predominantly been described under conditions of high substrate concentrations in anaerobic digestion (to methane)20,74. The fact that increased lactate and propionate production occurred with high substrate concentration makes high carbon loading an ideal condition to investigate sustained production of these products. However, these studies either did not bring the system to steady state at high carbon loads20 or did not study the complete product spectrum82, and none have investigated the mechanisms which cause the culture to switch to formation of these products in the first place. Similar work on high substrate concentrations found that hydrogen production and ethanol production are also increased under these conditions157. This suggests that production of lactate and propionate may be an electron mediation response to increased NAD reduction associated with increased glycolysis, or a redirection of carbon flow away from ATP production in general. Whatever the mechanism, analysis of the culture metabolic function under these conditions is necessary for controlled lactate and propionate production from waste streams.

This chapter describes the investigation of lactate formation from mixed culture fermentation on glucose in substrate non-limiting conditions. Specifically under investigation are the mechanisms that initiate the switch from butyrate production in substrate-limiting conditions to lactate production in substrate non-limiting conditions. The functionality of the culture is described at various substrate concentrations via SWATH-MS metaproteomics, as well as kinetic batch tests at each operational level. Additionally, the taxonomy of the culture is determined through 16S rRNA Illumina pyrosequencing. These factors are considered together, along with the product spectra, to describe the changes in metabolic functionality as the reactor enters different operational modes.

4.1. Approach

In this work, a CSTR (Figure 11) was operated under a range of substrate concentrations with fixed HRT using glucose as the substrate, and controlled at pH 5.50. In operational -1 order, the substrate concentrations tested were 5, 10, 15, 20, and 5 gglu·L . At each substrate concentration, the reactor was allowed to reach steady state (<10% change in major products over 1 week). Once at steady state, the reactor was analyzed for product

4-48 composition, microbial community structure, and protein expression through 6 samples over the course of 2-3 weeks. During this time, a set of batch fermentations were run to determine culture kinetics by directly inoculating fermenter broth into serum bottles containing BA media with 0.5 g·L-1 of glucose, and buffered to pH 5.50 with 0.5 MES. Protein expression analysis was then compared to the product composition to describe the response at a metabolic level. This was then compared to the microbial community analysis methods used by Mohd-Zaki et al.158 and Temudo et al.23, which described culture function based solely on known product spectra of observed taxa.

4.2. Experimental Setup

Continuous Experiments

Prior to inoculation, the fermenter was sterilized with 70% ethanol, purged overnight with sterile argon while maintaining the vessel at 60°C to ensure complete ethanol evaporation, filled to 1.0L with 10 g·L-1 glucose and 0.1x BA medium, and brought to 30.0°C. The native pH of this medium was 7.25. Motile cells from the activated culture were injected into the fermenter for an inoculum size of 1.0% v/v (0.5% each from the heat shocked and non-heat shocked seeds), and the pH was adjusted to 7.0-7.2. Continuous feeding was started when the pH reached 5.50, and the level was then maintained at 1.2 L for the duration of the experiment.

The dilution rate was held constant at 1.0 d-1, and the substrate concentration was varied by adjusting the water and glucose feed rates with constant nutrient feed rate. Substrate -1 concentration was increased progressively in 5 gglu·L intervals. New operational levels continued until substrate was no longer consumed in full, after which the system was -1 returned to 5 gglu·L to ensure re-emergence of glucose-limited results. For each substrate concentration, the system was allowed to reach steady state, defined by less than 10% variation in the two major products over the course of a week. The system was then held at steady state for 2-3 weeks. During steady state, the fermenter was sampled regularly for chemical analysis, headspace composition, dissolved gas composition, cell mass density, taxonomic analysis, and metatranscriptomic analysis.

Batch Activity Experiments

In addition to the continuous fermenter, a set of batch fermentations was carried out at each substrate concentration to measure the specific glucose uptake rate (kM) and the yields of biomass and major products. During the steady state period of operation in the continuous

4-49

0.6 0.8 0.8 of each

± ± ±

20.9 18.7 12.8

1.0 1.3 0.8

± ± ± X

12.9 12.8 11.2

the substrate concentration substrate the 0.1 0.2 0.0

± ± ± HFO

0.6 0.5 0.1

0.0 0.1 0.2

± ± ± HSU

0.0 0.1 0.5

0.7 0.5 0.0

± ± ± HMA

2.2 2.2 0.3

5.1 0.5 0.3

± ± ± HHX

1.3 0.8 12.2

1.1 1.5 1.0

± ± ± HBU

7.6 24.1 32.4

3.9 1.0 1.0

± ± ± ETOH

7 7.2 5.2 8.

0.8 0.8 0.2

± ± ± HPR ch operation level. Value listed as percent gCOD by total product gCOD. product by total gCOD as percent listed Value level. ch operation 0.5 1.3 1.5

f the steady state periods by COD composition. Vertical bars separate the individual operational levels, and and levels, operational individual the separate bars Vertical composition. COD by periods state steady the f 1.2 1.2 0.9 - axis.

± ± ± HAC

19.4 25.2 31.0

0.0 0.9 1.2

± ± ± course product data o data product course - HLA

0.0 0.4 Time Average COD composition of ea of composition COD Average

25.5 : 8

14 :

Level 5a Low High Table Table

Figure Figure the x below listed is level operational

4-50 fermenter, a triplicate of batch bottles were inoculated with 5 mL of broth from the fermenter, while a fourth control bottle was injected with 5 mL deoxygenated RO water. At inoculation, and then again every 1-3 hours, 6 mL of broth was sampled for analysis, and the volume was replaced with 6 mL of argon at 1atm using a gas-tight syringe. Samples were analyzed for cell density by absorbance at 560nm, and for chemical analysis as described in the methods. Toward the end of the exponential growth phase, 15 mL of broth was sampled -1 and analyzed in triplicate for TSS/VSS, which was used to convert OD560 to gCOD·L as described in the methods. The experiment was terminated once cell death outpaced cell growth. Product yields and specific substrate consumption rates were calculated as described in Section 3.5.2.

4.3. Results

Product Analysis

Based on the chemical analysis, steady state conditions were successfully achieved at all operation levels after 2-3 weeks of operation, with a clear shift in production from primarily butyrate and acetate (HBu-type) to primarily lactate and acetate (HLa-type) between 10 and 15 g·L-1 (Figure 14). Critically, HBu-type fermentation was regained when the substrate concentration was reduced from 20 g·L-1 back to 5 g·L-1, showing reversibility of the production types. Additionally, the initial 5 g·L-1 period produced hexanoate at 12.2±5.1% of the product COD. This high hexanoate production was not observed in any of the other operational levels, and so this initial state will be considered separate from the remaining steady state periods and referred to as level 5a. The 10 g·L-1 and second 5 g·L-1 states, being similar as acetate-butyrate production states at lower substrate concentration, will be referred to collectively as the “Low” levels, while the 15 g·L-1 and 20 g·L-1 states will be referred to collectively as the “High” levels. The total product spectrum by percentage COD for each collective level is described in Table 8.

The COD balance was first checked versus feed, and compared total COD of all fermentation products (including biomass and H2), as well as unconsumed substrate, to total COD of substrate fed (Figure 15A). Levels 10, 20, and 5b g·L-1 were each within 10% of balance. However, level 5a g·L-1 was slightly underbalanced, while level 15 g·L-1 was slightly overbalanced. To ensure the accuracy of the HPLC measurements, soluble COD (SCOD) was checked against measured SCOD as described in Section 3.4.4 (Figure 15B). Measurements versus theoretical SCOD were not significantly different for any operational level, verifying the HPLC measurements.

4-51

Figure 15: COD balances of each operating level. A) Total effluent COD versus COD of feed. B) Effluent SCOD as measured by HPCL versus effluent SCOD as measured by COD vials.

Biomass theoretical COD values were also verified against measured COD (Figure 16). The ratio of measured to theoretical COD was not significantly different from 1.0 for any level except 15 g·L-1. This lower ratio likely describes the over-balance in total feed-versus- effluent COD of the same level. Biomass COD for level 5a g·L-1 could not be verified because the TCOD verification method was not introduced until 10 g·L-1. Gas composition as determined by GC and by theoretical stoichiometry (Eqs 1-10) is presented in Figure 17. Measured and theoretical values are within 10 mol% in all cases except for at 10 g·L-1. A possible explanation for the discrepancy at level 10 is that there was a small gas leak either through the butyl rubber septum or through a screw cap port during this steady state period.

4-52 A small gas leak would result in H2 escaping disproportionately to CO2, whereas in a large leak, there would be detectible O2 in the headspace. As there was no detectible O2 at any point during the experiment, and the measured H2 appears to have dropped off relative to

CO2, it is assumed that there was in fact a small leak. Furthermore, the measured and

Figure 16: Ratio of measured to calculated biomass COD.

Figure 17: Gas composition as determined by both calculation from theoretical yield and measured by GC.

4-53 100%

90%

80%

70%

60%

50%

40%

30%

20% Total Product Composition COD by Composition Product Total 10%

0% L5a L10 L15 L20 L5b

HLA HAC HPR ETOH HBU HHX Minor Products Biomass H2

Figure 18: Product COD composition normalized by total product composition. theoretical values quickly returned to agreement following the reactor cleaning at the end of -1 the 10 g·L steady state period. For consistency however, the theoretical H2 was used for the remaining analysis.

The close agreement between measured and theoretical gas composition confirms the assumed metabolic stoichiometry. After normalizing the products as gCOD percentage of total product output (Figure 18), products were compared level-by-level by ANOVA. The first 5 g·L-1 level (5a) was found to be significantly different (p < 0.05) than each of the remaining levels for at least two of the major products (acetate, butyrate, lactate, and ethanol) in each case. In addition to high acetate and butyrate production, this level was also defined by high hexanoate production. For the remainder of the levels, hexanoate was produced only as a minor product. Critically, this means that the second 5 g·L-1 level (5b) did not regain hexanoate production, and likely indicates that reactor history has played a role in selection against hexanoate production. This is considered further in the discussion.

No products of levels 15 and 20 g·L-1 were significantly different. These production modes can therefore be considered identical. Within the 5b and 10 g·L-1 substrate concentrations, acetate was the only major product which had significantly different expression, at 9.98% higher in 5b (p = 0.03). However, the carbon yield of acetate from glucose was not significantly different between 5b and 10 g·L-1. Therefore, the major products of 5b and 10 g·L-1 can be also considered identical. It is acknowledged, however, that the minor products

4-54 Table 9: Yield and specific substrate consumption rate results from batch tests.

-1 -1 -1 -1 -1 Yield (gCOD·gCODGlu ) 5 g·L 10 g·L 15 g·L 20 g·L

Lactate 0.29 ± 0.14 0.21 ± 0.05 0.20 ± 0.07 0.13 ± 0.02 Acetate 0.17 ± 0.05 0.39 ± 0.08 0.24 ± 0.09 0.37 ± 0.04 Ethanol 0.07 ± 0.08 0.01 ± 0.08 0.18 ± 0.06 0.21 ± 0.05 Butyrate 0.37 ± 0.08 0.33 ± 0.08 0.01 ± 0.02 0.09 ± 0.02 Biomass 0.17 ± 0.06 0.29 ± 0.07 0.18 ± 0.05 0.19 ± 0.03

-1 kM (gCODGlu·(gCODX·h) ) 0.20 ± 0.03 0.25 ± 0.03 0.28 ± 0.09 0.38 ± 0.12

propionate, formate, and valerate, as well as biomass, are each significantly different between the low substrate concentrations. This may indicate a transition or intermediate state in some minor functions or microbes of the community at 10 g·L-1.

Culture Kinetics

Yields and specific substrate consumption rate (kM) were calculated as described in Sections 3.5.2 and 3.5.3, respectively, and the results are presented in Table 9. It was found that glucose consumption had not reached completion when biomass growth exited exponential phase, and therefore KS was not determined. There was minimal lag phase in any of the bottles, and so yields were calculated by regression starting from inoculation. Results from level 5a could not be used due to over-acidification, which prompted adding pH buffering to

Table 10: Percent difference within and between groups for each kinetic factor. P-values for each comparison are listed directly below.

Comparison Low-High L5-L10 L20-L15 HAc -7.7 -56 -33 > 0.05 0.002 > 0.05

HBu 580 -12 -93 3×10-6 > 0.05 > 0.05

HLa 50 34 64 0.04 > 0.05 > 0.05

EtOH -75 -22 0.67 2×10-5 > 0.05 > 0.05

HPr 210 74 -110 0.04 > 0.05 > 0.05

X 23 -44 -2.8 > 0.05 0.004 > 0.05

kM -32 -18 -28 0.004 > 0.05 0.002

4-55 the method starting with level 10. Because of this, only results from level 5b of the two 5 g·L-1 substrate concentrations are presented.

The calculated kinetic values were compared as described in Section 3.5.3 (Table 10). Yields of butyrate, lactate, and ethanol were found to be not significantly different within the low and high substrate concentrations, but were significantly different between the two groups. Acetate and biomass yields were not significantly different between the low and high groups or within the high group, but were significantly different within the low group. kM, conversely, was not significantly different within the low group, but was significantly different within the high group and between the two groups. Acetate and biomass yields do not appear to follow any specific trend, however lactate yield and kM each appear to correlate with substrate concentration. This was verified by linear regression analysis, and it was found that lactate yield correlates inversely (p = 0.003) while kM correlates positively (p = 2×10-5).

Taxanomic Analysis

Relative abundance of the 10 most prevalent genera (Figure 19) highlights the dominant genera at each substrate concentration. Here, it is evident that 5b and 10 g·L-1 are dominated by Megasphaera, with high (above 10%) abundance of Ethanoligenens, Bifidobacterium, and Pectinatus. By contrast, 15 and 20 g·L-1 are dominated by

100%

90%

80%

70%

60%

50%

40%

Relative Abundance Relative 30%

20%

10%

0% L5a L10 L15 L20 L5b Ethanoligenens Megasphaera Bifidobacterium Pectinatus Clostridiaceae Ruminococcaceae Clostridium Clostridiales Other Unassigned

Figure 19: Relative abundance of the top 10 most prevalent genera across the 5 sample sets. Normalized OTUs were grouped by genera, and the sum of relative abundance across all samples was used to determine the most prevalent.

4-56 Ethanoligenens, with high abundance of Bifidobacterium and a Clostridiaceae family group. Clostridium is also notably increased from the low to the high levels. Megasphaera and Ethanoligenens are co-dominant in level 5a, also with notable abundance of Bifidobacterium and Pectinatus.

Principle component analysis of the pyrosequencing results makes clear the taxonomic migration of the system from level to level (Figure 20). The culture in level 5a is taxonomically distinct from the remaining levels along PC2, and falls between the high and low levels along PC1. There is a clear taxonomic grouping between the low levels, and a second grouping between the high levels. These groupings are primarily described by Megasphaera and Pectinatus in the first case, and Ethanoligenens, Clostridium, and Bifidobacterium in the second case (Table 11). This shift also corresponds to the noted change in product spectrum between low and high substrate concentrations. 5a, lying around a PC2 value of -3.5, is described by a combination of Ethanoligenens, Prevotella, and Bacteroides, as well as distinctly low levels of Bifidobacterium and Clostridium. The fact that level 5b did not regain the production or taxonomy of level 5a indicates that 5a was not reproducible

Figure 20: Biplot of the pyrosequencing results. Operating levels are labelled by groups as L5a, L5b, L10, L15, and L20. Points within each group are labelled sequentially in the order of which the samples were taken as S1, S2, and S3. The red arrows show the order of operation of the reactor.

4-57 Table 11: Significant loadings of principal components 1-3. Percent of total variation for each principal component is listed in brackets in the header.

PC1 [66.6%] PC2 [16.1%] PC3 [5.3%] Correlation Genus Load Genus Load Genus Load Ethanoligenens 3.18 Bifidobacterium 1.09 Bifidobacterium 0.66 Clostridiaceae 1.47 Clostridium 0.99 Bacteroides 0.57 Positive Clostridium 1.09 Clostridiaceae 0.96 Sutterella 0.52 Bifidobacterium 0.76 Enterobacteriaceae 0.72

Coriobacteriaceae -0.61 Coriobacteriaceae -0.73 Enterobacteriaceae -0.61 Veillonellaceae -0.66 Bacteroides -0.83 Megasphaera -0.62 Enterobacteriaceae -0.69 Prevotella -0.94 Negative Clostridiales -0.70 Ethanoligenens -1.24 Lachnospiraceae -1.06 Pectinatus -2.58 Megasphaera -2.93 after start up while level 10 was. This is further evidence of the history effect noted above (Section 4.3.1).

Metaproteomic Analysis

At a false discovery rate of 1%, 3201 distinct peptides and 736 distinct proteins were positively identified, roughly half of which were involved in metabolic function. Full culture analysis shows that the primary genera actively switching expression were Bifidobacterium, Clostridium, Ethanoligenens, and Megasphaera (Figure 21). However, a full culture view of the proteomics is more descriptive of the culture taxonomy than of specific activity due to the inability to account for relative abundance.

Up-Regulated in High 20 Up-Regulated in Low

-10 Dorea Firmicutes Veillonella Pectinatus Hungatella Shewanella Clostridium Treponema Megamonas Thermincola Coprococcus Eubacterium Anaerovibrio Megasphaera Ruminococcus Ethanoligenens -20 Heliobacterium Bifidobacterium Lachnospiraceae Sum ofProteins with Change in Expression Pseudoramibacter

-30

Figure 21: Sum of proteins with change in regulation between high (positive) and low (negative) substrate concentrations. Protein regulation shifts calculated from the total catabolic protein spectrum.

4-58

Figure 22: Heat map of total metabolic enzyme expression. Enzymes are grouped by metabolic pathway or function. L2FC values are based on comparison to the baseline. Values were adjusted up such that the minimum L2FC was equal to 0, and then all enzymes which were not detected at a given level were also set to 0.

4-59 To account for taxonomic relative abundance and generate functional descriptions of the experimental conditions, the raw MS data was split and analyzed by individual genera (Megasphaera, Ethanoligenens, Bifidobacterium, Clostridium, and minor genera), and then regrouped by metabolic function, as described in Section 3.5.4. A heat map of the total metabolic proteome across all levels is depicted in Figure 22.

4.4. Discussion

Defining the Fermentation Types

The product spectrum of the low substrate concentrations, being primarily distinguished by butyrate production, is subsequently defined as HBu-type fermentation, while that of the high substrate concentrations is HLa-type fermentation. These product spectra align well with that reported in the literature. Lu et al.22 and Temudo et al.4, who operated similar systems at low substrate concentration and pH 5.5, also reported primarily butyrate and acetate, with some ethanol. This is also the product spectrum predicted in fermentation models at these operating conditions27. The change to HLa-mode production is in line with results reported by Eng et al.20, where lactate concentrations spiked with shock loads, and acetate production increased relative to butyrate production. Similarly, Voolapalli and Stuckey82 also described an increase in acetate production relative to butyrate. However, as previously stated, neither of these studies investigated sustained, steady state production, and Voolapalli and Stuckey82 did not measure the full VFA product spectrum.

In general, the defining product spectra of the high and low substrate concentrations align well with the known product spectra of their dominant genera59,171,172, though with one major exception. That is that Ethanoligenens, the dominant and most descriptive genus of the high substrate concentrations, is known to produce only very small levels of the defining product of this production mode, and instead is associated with high production of acetate and ethanol149,173. Additionally, as shown in Figure 20, the level 5b community is still migrating away from that of the high levels into the production steady state period. This suggests that the community response to the operating conditions is more important in determining the product spectrum than the specific taxonomy. Clearly, the method used by Mohd-Zaki et al.158 and Temudo et al.23 to explain the product spectrum via the taxonomy is not sufficient, and an exploration of the culture metabolism is required.

Comparison of relative abundances of catabolic enzymes between the high and low groups shows that genera which have the largest changes in protein expression are also those identified as genera of dominant and high abundance by pyrosequencing (Figure 21). This

4-60 analysis is only a comparison of the high and low substrate concentrations, and does not include level 5a, and the change in protein expression for Pectinatus could not be analyzed due to the lack of proteome information in the database. However, the remainder follow the same expression trends noted from pyrosequencing. However, because the relative abundance of each genus will directly affect the relative levels of protein detection, this level of analysis only shows that the change in product spectrum is related to the change in culture. Deeper analysis of genus-level protein regulation is needed to investigate the changes in metabolic regulation through the whole culture.

Change in Metabolic Enzyme Expression between Low and High Levels

Expression results of the high and low levels from the metabolic proteome were compared within each genus to determine which enzymes showed a significant change in expression. These results were plotted to depict the relative up and down-expression of pathways and mediation functions within a model mixed culture metabolism (Figure 23). Here it becomes evident that, in general, Megasphaera accounts for most of the difference in activity at low substrate concentrations, while Bifidobacterium and Ethanoligenens account for most of the difference in activity at high substrate concentrations. This aligns with the relative abundances of the taxonomic analysis. Notably, substrate uptake and the methylglyoxal bypass are each up-expressed in Megasphaera at high substrate concentrations.

Additionally, the trends in pathway expression align with the trends of product expression, though no enzymes relating to formate or ethanol production showed a significant change in expression. Propionate, which is produced via two pathways (Section 1.1), was not produced as a significantly greater proportion of product COD in either high or low substrate concentrations, though both pathways were up-expressed during low substrate concentrations. Finally, glycerol, while not detected in the product spectrum, was observed to be produced by Megasphaera and simultaneously consumed by Clostridium, which also exhibited higher expression of glucose uptake enzymes at low substrate concentrations.

The change from butyrate production at low substrate concentration to lactate production at high substrate concentration was largely attributable to the change in metabolic activity of Megasphaera (Figure 23). At high substrate concentration, Megasphaera down-expressed enzymes involved in butyrate formation, pyruvate oxidation, and electron mediation. Additionally, Megasphaera changed activity in the glycolysis side-branch pathways (Figure 8) from lactate reduction and glycerol production to lactate generation via glyoxylase. Megasphaera is known to consume lactate into propionate, and will preferentially consume

4-61 fold change (L2FC) (L2FC) change - fold ed inedthe table, and log2 rium showed an increased in expression of Ldh from L5b to L10, and then did not show any change in expression from L10 to L10 to from expression in change any show not did and then L10, to L5b Ldh from of expression in increased an showed rium Mixed culture model metabolism and key metabolic functions. Protein name, abbreviation, and Uniprot accession number are list are number accession Uniprot and abbreviation, name, Protein functions. metabolic key and metabolism model culture Mixed

23 :

Figure Bifidobacte function. its alongside listed is protein each of L20.

4-62 lactate through this pathway when available95,174. This shift away from propionate formation and the corresponding increase in glyoxalase expression appears to be the causal factor for the switch to lactate formation at high substrate concentrations.

The Methyglyoxal Bypass and the Switch between Butyrate and Lactate Production

Methylglyoxal is a toxic byproduct of glycolysis which is produced from dihydroxyacetone-P when glucose consumption outpaces phosphate uptake93,94. It is thought that methylglyoxal forms under these conditions as a mechanism to slow down substrate consumption by requiring the cell to spend energy detoxifying methylglyoxal to lactate. This detoxification occurs through the methylglyoxal bypass, and is carried out by glyoxalase.

This mechanism agrees with the observed increase in substrate uptake activity of Megasphaera and corresponding decrease in relative abundance. This also agrees with the observed decrease in activity for electron mediation and production of butyrate and propionate by Megasphaera during expression of the methylglyoxal bypass. However, given that phosphate is supplied in excess, and the rest of the community did not show the same response, it seems that this activity in Megasphaera has more to do with limited phosphate uptake capacity than a limitation of the BA media.

In addition to the metabolic change in Megasphaera, several other key functionalities changed among other genera in the system, which contributed to the expression of the HLa- type fermentation mode. Most importantly, Bifidobacterium began to up-express lactate production starting from 10 g·L-1, and subsequently increased expression of proteins involved in glucose uptake, glycolysis, and electron mediation. No other metabolic enzymes from Bifidobacterium showed any change in expression, which may indicate that lactate production was increased as a means to sink the excess electrons taken up as a result of increased glucose oxidation. Given the slight increase in propionate production measured at 10 g·L-1, which is simultaneous to the increased lactate production from Bifidobacterium, this may be attributable to increased lactate availability for usage in the acrylate pathway of Megasphaera. Though there was no significant change in acrylate pathway activity between 5b and 10 g·L-1. Bifidobacterium maintained high lactate pathway activity into the high substrate concentrations, and increased in relative abundance from 13±4% to 19±5%.

Notably, there was no detection under any condition of lactate dehydrogenase from Ethanoligenens, the dominant genus during HLa-type fermentation. However, and as noted previously, Ethanoligenens is not generally associated with lactate production, preferring to

4-63 Figure 24: Redundancy analysis of product formation and the sum of key metabolic processes. 173 sink electrons instead to ethanol and H2 . This is then consistent with the lack of Ethanoligenens lactate dehydrogenase detection.

Redundancy analysis of normalized product formation versus normalized total community metabolic pathway response shows that lactate production does in fact correlate along RDA1 with increased expression of both the methylglyoxal bypass and glycolysis (Figure 24). Furthermore, the correlation between lactate formation and expression of these two pathways is stronger than the correlation between lactate formation and expression of the pyruvate-to-lactate pathway. Additionally, the inverse relationship along RDA1 between lactate formation and expression of the lactate-consuming acrylate pathway of propionate formation highlights the other key metabolic shift away from lactate consumption. Together, these correlations confirm that the high substrate consumption and the methylglyoxal bypass are the key factors which cause the shift to HLa-type formation.

Substrate Uptake and Glycolysis

In addition to the change in production type going from low to high substrate concentrations, glucose consumption was no longer 100% (Figure 15). Based on the glucose consumption

4-64 Table 12: Specific substrate consumption rates of the steady state levels.

-1 Level kM (gCODglu·(gCODX·h) ) 5a 0.41 ± 0.01 Low 0.36 ± 0.04 High 0.34 ± 0.03 rates of the high substrate concentrations, the maximum glucose consumption capacity was -1 -1 13±1 g·L (14±1 gCOD·L ). The kM, meanwhile, remained consistent throughout the experiment (Table 12), and was comparable to the maximum batch test kM value of -1 0.383±0.117 gCODGlu·(gCODX·h) . However, the kM of level 5a was higher compared to the remaining levels. This may be related to the differing product spectrum featuring hexanoate which was described for this level.

The up-expression at high substrate concentration of substrate uptake and electron mediation functionalities by Bifidobacterium were additionally each consistent with increased glucose processing to lactate. As glucose became available in excess, Bifidobacterium cells were able to uptake more substrate. This resulted in greater glycolysis activity, which in turn required greater electron mediation to process the reduction of the NAD pool. Lactate dehydrogenase was then up-expressed in order to sink electrons generated during pyruvate production.

Chain Elongation in Level 5a

As described above, level 5a was distinct from the other levels in both product spectrum and taxonomy. 5a was co-dominated by Ethanoligenens and Megaspheara (Figure 19), and produced high amounts of hexanoate and significantly less acetate and butyrate than the low levels (Table 8). As discussed in Section 1.1.5, Megasphaera is capable of chain elongation via the reverse β oxidation pathway, which utilizes butyrate pathway enzymes and their analogues to elongate ethanol and acetate to butyrate, hexanoate, and octanoate.

The Megasphaera butyrate pathway enzymes 3-hydroxybutyryl-CoA dehydrogenase, acetyl-CoA acetyltransferase I and II, crotonase, and thioesterase were all most highly expressed in level 5a (Figure 22), despite decreased overall butyrate production at this level. Specific downstream enzymes of reverse β oxidation were not detected in the proteome. However, the UniProt database does not specifically annotate these enzymes, and so this may be due to either lack of annotation, or analog function of the butyrate-producing enzymes. In either case, the high relative abundance of Ethanoligenens combined with

4-65 increased hexanoate and decreased butyrate and acetate suggests that Megasphaera is likely chain-elongating acetate, ethanol, and butyrate to hexanoate in this level.

The chain elongation activity is then lost at level 10 g·L-1, and not regained. According to Spirito et al.68, chain elongation activity occurs when other sources of energy are low. The increase from 5 to 10 g·L-1 of glucose may have been sufficient to supply Megasphaera with the energy to out-compete Ethanoligenens. Later, when the substrate concentration was decreased from 20 back to 5 g·L-1, Megasphaera may have been able to react more quickly than Ethanoligenens to the now substrate-limiting conditions. At this final operational level, Megasphaera no longer required expression of the methylglyoxal bypass, while Ethanoligenens had been enriched in excess-substrate conditions. In any case, the failure of the reactor to regain chain elongation function appears to be due to the operating history of the reactor, which is a known factor affecting reactor performance22.

Additional Observations from the Metaproteome

Two additional changes were observed in overall culture function between operational states. Clostridium was observed to increase glycerol uptake at low substrate concentrations, which explains the lack of glycerol detection in the product spectrum, and may explain the increase in Clostridium glycolysis activity as Clostridium cells replaced carbon sourcing with glucose. Finally, the increase in acetate production was attributable to Ethanoligenens, which up-expressed acetate kinase at each increasing level. Bifidobacterium formate acetyltransferase was detected, but had not significant change in expression. Additionally, there was no detection of ethanol pathway enzymes for any genus.

Linking Continuous and Batch Observations

Batch kinetics were used to better understand culture function in continuous operation. The change in butyrate and ethanol yields between low and high groups aligns with the continuous reactor product spectrum, where butyrate yields are higher with biomass grown at low substrate concentrations and ethanol yields are higher with biomass grown at high substrate concentrations. And within the low group, the biomass yield in the batch tests follows the same trend as that of the continuous reactors within the low groups, with biomass from 5 g·L-1 growing at a 43.7% lower yield than that from 10 g·L-1 (compared to 17.4% lower biomass yield in the continuous reactor). However, the acetate yield is 7.66% lower with biomass from 5 g·L-1 compared to 10 g·L-1. This is the inverse of what was observed in the continuous reactor, where acetate yield was 9.98% higher in 5 g·L-1 compared to 10 g·L-1.

4-66 As described in Section 4.3.2, lactate production in the batch test correlates inversely to the substrate concentration in which the biomass was grown, while kM correlates positively. The inverse correlation between lactate yield and substrate concentration conditions of the biomass is surprising given that the reverse is true in the continuous reactors. This increased yield in batch mode with biomass grown in low substrate concentration conditions (in which the biomass was exposed to very little lactate) supports the assumption that lactate is an transient state product which is reconsumed when advantageous59,74. Under this mechanism, the culture consumes glucose through glycolysis as rapidly as possible for quick ATP generation. Additional ATP is then also produced via pyruvate oxidation, and excess electrons are sunk in a single reduction step to lactate, which can be reconsumed later for further pyruvate oxidation (Section 1.1.6). By contrast, the biomass grown in high substrate concentration conditions was already well-exposed to high concentrations of lactate, and so could readily switch to lactate consumption in an advantageous situation, such as the lower glucose concentration in batch mode.

The direct correlation between kM and substrate concentration indicates that the biomass is consuming substrate at a rate proportional with the increase in substrate concentration. This is a surprising result considering the lack of change in biomass yield between the low and high groups in both batch and continuous fermentation, as well as the stability of the continuous kM and product spectra within each group. This appears to contradict the common assumption that fermenting cultures select product pathways predominantly based on maximising ATP yield for biomass growth8,27, since under that assumption the biomass yield should either increase proportionally with kM, or be significantly higher at high substrate concentrations compared to low substrate concentrations.

The interplay of the expressional changes of the culture overall may offer a clue as to this observed contradiction. This functionality again appears to be mainly attributable to Megasphaera. As noted previously, there is an increase in biomass yield both in the continuous system and the batch tests from level 5b to 10 g·L-1. This initial increase in biomass yield is consistent with the assumption of maximal biomass growth8,27. However, when glucose becomes in excess, the biomass yield again drops to that of level 5b and remains constant through the high substrate concentrations, despite the continuing increase of kM. This is likely partially caused by the build-up of methylglyoxal in Megasphaera at high substrate concentration, inhibiting biomass growth while maintaining increased glucose uptake93,94. However, it is not clear why the other genera in the system do not compensate for the lost Megasphaera biomass yield. One possible explanation is that the biomass

4-67 Table 13: Production of ATP and sinking of e- per biomass associated with the product spectrum of each level. ATP and -1 e- values are each in mmol·gCODX . The “Total” section describes functionality per product pathway from glucose uptake through end product; the “Glycolysis” section describes functionality within glycolysis distributed by product yield; and the “Post-HPy” section describes functionality per product pathway from pyruvate through end product. The values are heat- mapped level-by-level and across sections to highlight spread of functionality within each level.

ATP e- Level HLa HAc HPr EtOH HBu HLa HAc HPr EtOH HBu Total 5a 0 47 0 6 35 0 -94 1 0 -47 5b 0 72 0 6 53 0 -143 0 0 -70 10 1 54 2 3 44 0 -108 3 0 -58 15 22 83 1 7 11 0 -166 2 0 -15 20 25 91 1 10 15 0 -182 2 0 -19

Glycolysis 5a 0 23 0 6 23 0 -47 -1 -12 -47 5b 0 36 0 6 35 0 -72 0 -12 -70 10 1 27 2 3 29 -1 -54 -3 -6 -58 15 22 42 1 7 8 -45 -83 -2 -14 -15 20 25 45 1 10 10 -50 -91 -2 -19 -19

Post-HPy 5a 0 23 0 0 12 0 -47 1 12 0 5b 0 36 0 0 18 0 -72 1 12 0 10 0 27 0 0 15 1 -54 6 6 0 15 0 42 0 0 4 45 -83 5 14 0 20 0 45 0 0 5 50 -91 5 19 0 composition varies significantly between genera. However the observed variance in the ratio of measured gCOD to gVSS varied by only 14.6% through the entire experiment (Figure 16) and shows no correlation to PC1 or PC2. Alternatively, biomass growth rate may already be maximized at the operated HRT. Notably, there is not an observed increase in substrate uptake by the dominant genus at high substrate concentration. However, this does not explain the noted increase in biomass yield at 10 g·L-1.

Energy and Electron Mediation

The product-associated ATP and electron-sink yields were calculated from Equations 2-11, and are displayed as a heat map (Table 13) as described in Section 3.5.2. This heat map provides a useful visual for how energy generation and electron mediation functionalities are segregated between glycolysis and post-pyruvate processing.

Overall, it is observed that acetate production is the primary driver of ATP generation in all cases. This is coupled with butyrate production in both 5a and low substrate concentrations, and with lactate production at high substrate concentrations. The butyrate-associated ATP

4-68 generation is evenly split between glycolysis and post-pyruvate, while the lactate-associated ATP generation is entirely via glycolysis. Additionally, all propionate, and ethanol-associated ATP generation takes place in glycolysis, as these are not ATP-producing pathways post- glycolysis.

Since oxidation is required for ATP generation, these product pathways are also the primary source of electrons. Electron sinking is much more polarized between the glycolysis/post- propionate split, with most electron-sourcing occurring during glycolysis, and electron-sink functionality switching from ethanol at 5a and low substrate concentrations, to ethanol and lactate at high substrate concentrations only after glycolysis (with propionate also used at 10 g·L-1). This discrepancy in electron mediation functionality highlights the active expression of electron-sink pathways as a response to ATP formation (and substrate oxidation) in glycolysis, subject to gene availability.

4.5. Conclusions

Increasing the substrate concentration beyond the glucose-consumption capacity resulted in a change in product expression from HBu-type to HLa-type fermentation. HBu-type was regained upon reducing the substrate concentration back below the glucose-consumption threshold. Observations from proteomic analysis suggest that this change in production type was directly attributable to a combination of methylglyoxal detoxification to lactate in Megasphaera and electron sinking to lactate in Bifidobacterium.

Batch yield and specific substrate consumption rate results were consistent with the observed product yields and genus functionalities of the continuous reactor. The inverse relationship between batch lactate yield and substrate concentration of the biomass appears to be a result of lactate processing by Megasphaera. The relationship between the increasing kM and stagnant biomass yield, which seems to contradict the assumption of maximal biomass growth, appears to result from the methylglyoxal-induced biomass growth inhibition in Megasphaera, though it is not clear why the other genera do not compensate when experiencing reduced competition and increased substrate availability. Ultimately, it seems that the assumption of maximized biomass growth holds true in substrate-limited conditions, but not when substrate is in excess.

4-69 5. Redox Stress from Exogenous Electron Mediators, not Electron Uptake, Causes Shift from Butyrate to Propionate Production in a Poised Potential System

Electro-fermentation attempts to alter the electron load separate from the substrate load or other conditions of the culture through control of the extracellular redox state using a poised potential electrode array. This generally refers only to increasing the electron load relative to the substrate rather than providing an electron sink, and has been studied in pure culture fermentations as a means to increase the production of reduced products123,124,137. Electron transfer can be accomplished through direct electron transfer via bionanowires, or through indirect electron transfer via electrode-generated hydrogen or using a mediating compound of known redox potential (Section 1.3.3). Direct electron transfer has only been observed in anodic systems, and so cannot be used to supply excess electrons in electro- fermentation127. As such, this makes indirect electro-fermentation using mediators of known reduction potential the ideal method for which to study the effects of redox state on the mixed culture product spectrum.

Previous studies using electro-fermentation did report increased production of electron sink products, but either did not control the terminal redox state via mediators138, or studied the effects using a pure culture123,124. However, comparison between the measured electron uptake and the shifted product spectrum indicates that the additional electron uptake alone does not always account for the additional reduced product formation. Additionally, product formation mechanisms in previous studies have been limited to taxonomic descriptions and product spectrum analysis.

This chapter describes the formation of electron sink products in mixed culture fermentation of glucose in substrate-limiting conditions using mediated fermentation in a cathode cell. Metaproteomics techniques are employed in order to describe the mechanism of product formation shift. As with the investigation into substrate-non-limiting conditions, the functionality of the culture is described via SWATH-MS metaproteomics at steady state operation with each mediator compound. The taxonomy is assessed via 16S rRNA Illumina pyrosequencing, and considered alongside the product spectra and the proteomics to describe the metabolic function of the culture under each condition.

5-70 5.1. Approach

Triplicate CSTRs (Figure 12) were operated in parallel for each mediator tested. All three reactors were fed with identical BA media with 5 g·L-1 glucose substrate and 100µM of mediator compound, prepared as described in the methods. Two of the reactors, designated “experimental reactors”, were operated with -700 mV vs 3 M Ag/AgCl set potential on the cathode as described in the methods. The third control reactor was operated in open circuit. For each condition, the final 5 g·L-1 steady state period of the varied substrate concentration experiment was used as a mediator-free control.

Mediator compounds with a range of electrochemical reduction potentials were chosen as such to explore a range of negative redox potentials (within a potential window of 0 to -400 mV vs SHE), in order to interact with a wide range of electron mediation systems within the culture, including FAD, FMN, NAD(P), and Fd33. Useful mediators and concentrations were identified from those used in pure culture studies123,124.

Experimental Setup

Four mediator compounds were chosen for controlling the redox state across a range of negative potentials (vs SHE): anthraquinone-2,7-disulfonate (AQ), riboflavin (RF), neutral red (NR), and methyl viologen (MV). Mediator reduction potentials were measured in abiotic, anaerobic reactor broth at steady state operation conditions using cyclic voltammetry as described in Section 3.4.5.

Biomass was taken from the final 5 g·L-1 operational level of the varied substrate concentration experiment and activated as described in the methods. As with the varied substrate concentration experiment, continuous feeding at a dilution rate of 1 d-1 was initiated once the pH dropped to 5.50. The pH was maintained from this point at 5.50±0.01. A constant set potential was applied to the working electrodes of at least 50 mV less than the reductive peak of the most negative mediator in the two experimental reactors 24 hours after inoculation. This ensured rapid reduction of mediator compound123. The reactors were held in steady state feeding and effluent extraction for the duration of the experiment. After 1 week of start-up time, the reactors were considered to be at steady state production, and were maintained at that state for one additional week.

During start-up, reactors were sampled once daily for product spectrum, biomass concentration, and headspace analysis. During the steady state production period the start- up sampling was continued, except headspace analysis was decreased to every other day

5-71 3.0E-04

2.0E-04

1.0E-04

0.0E+00

-1.0E-04

-2.0E-04 Background Current (A) Current -3.0E-04 AQ RF -4.0E-04 NR -5.0E-04 MV -6.0E-04 -1200 -1000 -800 -600 -400 -200 0 200 400 Potential (mV vs 3 M Ag/AgCl)

Figure 25: CVs of mediators and background fermentation broth. Table 14: Measured reduction potentials of mediator peaks and their allosteric means in abiotic fermentation broth at pH 5.50 and 30°C. Reduction potentials are compared to standard literature values (pH 7 and 25°C).

Oxidative Reductive E' vs Ag/AgCl E' vs SHE Literature E°' Mediator Peak (mV) Peak (mV) (mV) (mV) vs SHE (mV)

AQ -301 -396 -349 -119 -184‡҂ RF -307 -393 -350 -120 -15† NR -439 -577 -508 -278 -325‡ MV -599 -649 -624 -394 -430‡ †Malinauskas et al.175 ‡Emde and Schink123 ҂Literature value for anthraquinone-2,7-disolfonate not available, so E°' of anthraquinone-2,6-disolfonate123 presented in its place. to prevent over-drawing. Additionally, during steady state, samples were taken every second day for pyrosequencing and metaproteomic analysis. Additionally, one 15 mL sample was taken alongside one of the biomass samples to calibrate the VSS concentration to the BCA result.

5.2. Results

Cyclic Voltometry

CVs for the four mediator compounds, as well as the background media, were obtained and are presented in Figure 25. No background peaks were detected in the non-mediated broth, and peaks were successfully obtained for each of the compounds. H2 production is observed to begin at the cathode around -850 mV (vs Ag/AgCl), well below the MV reductive peak

5-72 Table 15: Theoretical maximum electron supply from mediators in feed before interaction with cathode (Mediator) and measured electron supply from the electrode array (Cathode), both in Amps. % Reduced and % Oxidized is a measure of the redox state of the mediator in the feed before entering the vessel and interacting with the cathode.

Mediator Cathode % Reduced % Oxidized AQ -7.4×10-5 ± 6×10-6 -1.87×10-5 ± 3×10-8 75 25 RF -7.3×10-5 ± 6×10-6 -1.38×10-5 ± 2×10-7 81 19 NR -7.4×10-5 ± 1×10-5 -5.12×10-6 ± 1×10-8 93 6.9 MV -7.7×10-5 ± 1×10-6 -6.49×10-5 ± 1×10-7 15 85

(-649 mV). This allowed for operation at -700 mV, or roughly 50 mV below the most negative mediator peak, during continuous mode to ensure complete reduction of all mediator compounds without risk of producing H2.

Reduction potentials (E’) were calculated as described in Section 3.4.5, and are presented in Table 14, along with the literature values for standard reduction potential (E°’) of each mediator compound. Measured E’ of riboflavin was 105 mV less than the literature value, while the remaining mediators were 36-66 mV greater than their literature values. This discrepancy is likely due to measuring at non-standard pH for this experiment.

Cathodic Current

The measured electric current, as well as the theoretical maximum electron supply from mediators in the feed (before interaction with the cathode), are presented in Table 15. For each reactor, cathodic current was less than the theoretical electron supply of the fed mediator. This, combined with the very low order of magnitude of the measured current, suggests that all activity at the cathode was solely for reducing the oxidized fraction of mediator as it was fed to the reactor, and that there was very little cycling of mediators between cells and the cathode. Under this assumption, the relative reduced to oxidized forms of the mediator in the feed were calculated by electron balance with reference to the supplied current from the mediator. While the low set potential of the cathode ensures full reduction in the closed circuit experimental reactors, the mediators fed to the open circuit reactors do not interact with a cathode, and therefore do not get further reduced. Any production differences between the closed and open circuit reactors may then be partially attributable to this difference in redox pressure from the equilibrium redox state of the mediator in the feed.

Product Spectrum Analysis

Product analysis shows a distinct shift in product spectrum between neutral red and methyl violgen (Figure 26). The product spectrum remained HBu-type from non-mediated control

5-73 Figure 26: Product spectrum of electro-fermentation experiments A) is the product balance (%gCOD out vs gCOD consumed), and B) is the product relative abundance. Control is level 5b from the varied substrate concentration experiment. The mediated experiments depict the average of 5 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. through neutral red, and then increased acetate and propionate production at the expense of butyrate with the methyl viologen mediator. This fermentation mode will be called HPr- type. One of the NR experimental reactors failed and is therefore not considered in this analysis.

Products were normalized by percentage of total product COD content (Figure 26B). Level- by-level ANOVA on the normalized products showed that there was no significant difference in product spectra between experimental reactors for each mediator. When compared to the

5-74 control reactor, RF, AQ, and NR each show no significant difference in any major product. This is true of both the experimental reactors and the open circuit reactors. There is therefore no significant shift in product spectrum through NR. The shift to HPr-type occurs only with the MV set, where acetate and propionate each increase and butyrate decreases in both the experimental and open circuit reactors.

RF, AQ, and NR each additionally showed no significant difference between experimental reactors and the open circuit reactor, except for the biomass of AQ which was 19.8% lower than open circuit in reactor 1 and 16.9% lower in reactor 2 (p = 0.006 and 0.04, respectively). For MV, there was no significant difference in product spectrum between experimental reactor 2 and the open circuit reactor. Experimental reactor 1, however, produced 27.2% less propionate than open circuit (p = 0.03). This translated to 23.1% less propionate production in the grouped experimental reactors compared to open circuit (p = 0.02). Because there was no significant difference between reactor 2 and either of the other reactors, the entire MV set will generally be considered identical, though possible reasons for the discrepancy between reactor 1 and open circuit will be discussed in Section 5.3.8.

Taxonomic Analysis

Control, AQ, RF, and NR 16S rRNA amplicons were all dominated by Megasphaera, while the MV system was dominated by Pectinatus (Figure 27). Pectinatus was also a prominent secondary genus in control through NR, while Megasphaera was the secondary genus in MV. Additionally, Clostridium was prominent in the mediated reactors, with decreasing relative abundance from RF to MV. The same trend was seen with Bifidobacterium, though it was also prominent as a secondary genus in control. Finally, Ethanoligenens, which was another prominent secondary genus in control, was a minor genus in the mediated reactors at less than 1% in each except AQ (1.5% experimental, 4.5% open circuit).

Principal component analysis also shows that the phylogeny of MV is separate from the other reactors along PC1 (Figure 28), which is descriptive of the genera Pectinatus and Sutterella versus Megasphaera, Clostridium, Bifidobacterium, and Ethanoligenens (Table 16). The PC1 loadings of Megasphaera to Pectinatus are an order of magnitude larger than all other descriptive genera of PC1 except for that of Clostridium, which is still more than 3x less than Megasphaera and Pectinatus. This indicates that the primary taxonomic variance described by PC1 is almost entirely attributable to changes in these two genera.

There is not much spread between mediator groups along PC2, which is descriptive of the genera Bifidobacterium and Clostridium versus Ethanoligenens, Prevotella, and

5-75 Bacteroides. However, the experimental reactors cover a larger spread than the open circuit reactors within PC2, with the open circuit reactors clustered along the positive side. The AQ experimental reactors, as well as some samples from AQ open circuit and MV and RF experimental reactors, fall within the negative PC2 space. From the relative abundance data, this appears to be less attributable to the descriptive genera of PC2, and instead seems to be primarily attributable to higher abundance of an Enterobacteriaceae family group.

Figure 27: Relative abundance of the top 12 most prevalent genera across all sample sets. Normalized OTUs were pooled by genera, and the sum of relative abundance across all samples was used to determine the most prevalent. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. Table 16: Significant loadings of principal components 1-3. Percent of total variation for each principal component is listed in brackets in the header.

PC1 [39.1%] PC2 [26.7%] PC3 [13.0%] Correlation Genus Load Genus Load Genus Load Pectinatus 6.18 Lactobacillus 1.35 Ethanoligenens 2.58 Sutterella 0.68 Pectinatus 1.32 Clostridiaceae 1.09 Clostridium 1.27 Clostridiales 1.03 Positive Coriobacteriaceae 0.96 Unassigned 1.01 Megasphaera 0.95 Megasphaera 0.73 Lactococcus 0.63 Ruminococcaceae 0.61

Ethanoligenens -0.53 Clostridiales -0.74 Sutterella -0.51 Bifidobacterium -0.59 Ethanoligenens -0.96 Lactococcus -0.55 Clostridium -1.65 Gammaproteobacteria -0.96 Ruminococcus -0.83 Negative Megasphaera -5.37 Enterobacteriaceae -6.21 Enterobacteriaceae -1.50 Lactobacillus -1.69 Clostridium -2.41

5-76 Figure 28: Biplot of the reactor taxonomic groupings as represented by PC1 and PC2. Experimental (closed circuit) reactors are grouped together within each mediator, and separate from the open circuit (OC) reactors. Metaproteomic Analysis

At a global false discovery rate of 1%, 898 distinct peptides and 398 distinct proteins were detected, roughly half of which are metabolic. However, the protein database for the dominant genus in MV (Pectinatus) is limited to two non-metabolic proteins, and therefore the function of this genus must be inferred from a combination of literature sources and correlations to the product spectrum, rather than directly analyzed by proteomics.

Full culture proteomic analysis shows that the primary active genera are Bifidobacterium, Clostridium, Megasphaera, and Propionibacterium (Figure 29). Given the lack of database entries for Pectinatus, this full culture analysis largely agrees with the pyrosequencing results except for Propionibacterium, which makes up less than 1% of the total relative abundance of any reactor. However, due to the inability to account for relative abundance in the analysis, a total-culture view of the proteomics is more descriptive of the culture taxonomy than of specific activity. To account for relative abundance and generate functional descriptions of the experimental conditions, the raw MS data was split and analyzed by individual genera, and then regrouped by metabolic function, as described in 3.5.4. A heat map of the total metabolic proteome across all levels is depicted in Figure 30.

5-77

shifts calculated from the total catabolic protein protein catabolic total the from calculated shifts

Sum of proteins with change in regulation between experimental (positive) and control (negative) reactors. Protein regulation Protein reactors. (negative) control and (positive) experimental between in regulation change with proteins of Sum

29 :

Figure Figure spectrum.

5-78

comparison to the baseline. Values Values baseline. the to comparison

ere not detected at a given level were also set to 0. to set also were level a given at detected not ere

Heat map of total metabolic enzyme expression. Enzymes are grouped by metabolic pathway or function. L2FC values are based on based are values L2FC function. or pathway metabolic by grouped are Enzymes expression. enzyme metabolic total of map Heat

30 :

Figure Figure w which enzymes all then and 0, to equal was L2FC minimum the that such up adjusted were

5-79

Continued

30: Figure

5-80

Continued

30:

Figure

5-81 5.3. Discussion

Cathodic Current and Product Formation

As described in Table 15, the steady state cathodic current for each level was less than the theoretical maximum electron supply from the fed mediators. This implies that there was very little redox state turnover of the mediators between the community and the cathode. Furthermore, the mediator in the feed bottle must have been at some equilibrium between the reduced and oxidized states, and the measured cathodic current must have been primarily for fully reducing the mediator once in the reactor. While these measured currents were roughly 100x smaller per unit volume than those reported in Kracke’s137 continuous study, there was an induced shift in product spectrum from HBu-type to HPr-type going from a mediator with -278 mV to -394 mV reduction potential. This shift accounted for 0.27 extra electrons being sunk per glucose compared to control.

As with Kracke’s137 continuous study, the change in product spectrum did not occur until a mediator with sufficiently low reduction potential was used. Furthermore, in both studies, the community response in generating a more reduced product spectrum was far greater than the measured electron load, indicating some cellular response other than only electron balancing. Potential response mechanisms will be explored later in the discussion through metaprotemeomic analysis, but first the pathway responses of the two production types (HBu and HPr) will be described.

Taxonomy and Proteomics of Butyrate Production

Expression of butyrate pathway enzymes was limited to Clostridium and Megasphaera. From the total culture pathway expression, there is no definitive trend relating to the standard reduction potential of the mediator compounds (Figure 31). Genus-normalized culture activity analysis reveals that the culture butyrate pathway activity in MV consistently had the weakest response, at about half the activity of control, while community had a greater response compared to control for each of the other mediators in both closed and open circuit. The reduced response to MV is consistent with the product spectrum. Additionally, because Pectinatus does not possess the capacity to produce butyrate176, the lack of a Pectinatus proteome library did not affect analysis of this pathway.

The fact that butyrate pathway activity is so low in MV when normalized, but that total pathway expression remains high in MV when not normalized suggests that the decrease in

5-82 Figure 31: Expression of enzymes of the butyrate production pathway A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. butyrate production is more dependent on the taxonomic shift toward non-butyrate- producing Pectinatus than on a change in the expression of the butyrate pathway.

To test this assumption, the effect of the first amplicon principle component (Table 16) on butyrate concentration (% gCOD/gCODtotal products) was tested by regression using a linear model. For the total data set, this model is highly significant with p = 7×10-15, and shows that

5-83 70

60 )

products 50 gCOD / 40

20 HBu Formation (% gCOD (% Formation HBu 10

0 0 10 20 30 40 50 60 70 Relative Abundance of PC1 Taxonomic Indicators (%)

Figure 32: Single factor linear regression of HBu production as a percentage of total product formation versus Clostridium (squares, solid line), Megasphaera (diamonds, dashed line), Bifidobacterium (crosses, long-dashed line), and Ethanoligenens (circles, dotted line). Data includes only through NR, and excludes MV. butyrate production is strongly correlated to the abundance of Megasphaera, Clostridium, Bifidobacterium, and Ethanoligenens. Butyrate production in this system is therefore dependent on the abundance of PC1 genera. Bifidobacterium, however, is not known to produce butyrate172,177, and so the individual PC1 genera were analyzed together by multiple factor linear regression. For the full mediator data set, it was found that Megasphaera and -11 Ethanoligenens were both important factors (pMe = 1×10 , pEt = 0.001). Surprisingly, when testing the more conservative set of control through NR, the only genus that showed any correlation was the weak butyrate producer Ethanoligenens149,173, which was weakly correlated (p = 0.04, Figure 32).

From multiple regression and the PC1 loadings, the butyrate-producer Megasphaera appears to be the primary driver of butyrate production. Multiple regression on the full data set combined with regression on control through NR shows that, despite being only a weak butyrate producer, abundance of Ethanoligenens is also linked to butyrate production, either through coincidence of growing conditions or through some enhancing role. It has been shown previously that butyrate production of Megasphaera is enhanced by the presence of acetate178, which is a primary product of Ethanoligenens. The correlation of Ethanoligenens

5-84 Figure 33: Expression of enzymes of the DCA cycle of propionate production A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. to butyrate production is likely due to this enhancing effect from its acetate production, though this functionality was not observed in the proteomics analysis.

Taxonomy and Proteomics of Propionate Production

Expression of dicarboxylic acid (DCA) cycle enzymes is fairly consistent within the measureable culture as a whole throughout all levels (Figure 33). This holds true for both

5-85 30

) 25

total products 20 gCOD /

5 HPr Formation (% gCOD (% Formation HPr

0 0 10 20 30 40 50 60 70 80 Pectinatus Relative Abundance (%)

Figure 34: Linear regression of HPr production as a percentage of total product formation versus Pectinatus relative abundance. The solid line (—) incorporates the full data set, while the dashed line (- -) incorporates only up through NR, excluding MV. total pathway expression and when adjusted for taxonomic relative abundance. The pathway expression in the mediated, closed-circuit tests does appear to increase with mediator E’, however, linear regression modelling showed no significant correlation. The acrylate pathway, as expressed by Megasphaera95, is considerably more variable (Figure 30), but also does not show any significant expression trend. The expression of propionate- pathways from the measurable culture is therefore neither dependent on the mediator nor the presence of a polarized electrode. Propionate production must instead be reliant only on the action of Pectinatus, which was the dominant genus in HPr-type fermentation and could not be assessed through proteomic analysis due to lack of database entries for this organism (Section 5.2.5).

To test this hypothesis, the effect of Pectinatus relative abundance (%) on propionate concentration (% gCOD/gCODtotal products) was tested via linear regression. For the total data set, this model is highly descriptive (p = 8×10-22, Figure 34). However, because there is high clustering amongst control through NR, and propionate and Pectinatus are already the defining features of the MV set, a more conservative test of the model robustness was conducted by excluding the MV data. Linear regression of this limited data set was still highly correlated with propionate formation (p = 2×10-5). Therefore, propionate formation appears

5-86 to be primarily a linear function of Pectinatus abundance in this system, even though this cannot be directly described in terms of metabolic expression.

Fermentation Characteristics of Pectinatus

It is insufficient to merely show correlation between a genus and a production mode in lieu of proteomic data without also discussing the physiological capacity of that genus. Pectinatus is a strictly anaerobic genus closely related to Megasphaera, which are both ruminal genera known for beer spoilage176,179. Like Megasphaera, Pectinatus species ferment sugars to acetate, propionate, and lactate171,180. However, unlike Megasphaera, Pectinatus species are not able to produce butyrate, valerate, or caproate, and instead produce succinate and acetoin. The lack of butyrate production capability also requires that Pectinatus species produce much more propionate than Megasphaera species to account for the lack of the butyrate electron sinking capacity. This propionate production works out to about 4x the molar ratio of acetate when grown in a basal medium181, as opposed to the 1x or less HPr:HAc molar ratio of Megasphaera species grown in basal medium174,182. The reason for this is twofold. Firstly, lack of butyrate pathway requires that Pectinatus produce all post-glycolysis substrate-level ATP from acetate. This greatly increases the intracellular electron load and requires a high-yield electron acceptor, such as propionate, to balance the electrons.

The fact that Pectinatus species also produce succinate hints that, unlike Megaspheara, they make use of the DCA cycle for propionate production. However, as they also produce lactate, it is unclear whether they also have acrylate pathway functionality. Use of the DCA cycle may explain why lactate production remains quite low as propionate production increases, as it does not make use of a lactate intermediate. Though it is unclear whether lactate production increases under high substrate loads. This preference for electron sinking via propionate production explains why there is such a strong correlation between propionate production and Pectinatus relative abundance in this study. However, it does not explain why Pectinatus was able to achieve such high abundance in the first place. The fact that there was no measureable affect ascribable to the cathodic current suggests that the high Pectinatus growth could be dependent on biological interactions with the mediators.

Glucose Processing

Total culture glycolysis function appears to decline steadily with progressively negative mediator E’ (Figure 35). This was true for both closed and open circuit reactors, and the difference between these sets was less than 10% in all cases. In order to check for any

5-87 Figure 35: Expression of enzymes of the glycolysis pathway A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open- circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. effect of mediator E’ on glycolysis activity, a linear model was tested via regression on the combined closed and open circuit sets. However, this model showed no significant correlation (p > 0.05). Abundance-adjusted glycolysis expression also showed a steady decline from control through MV, though this again was likely linked to lack of the Pectinatus proteome.

Despite the measured decline in glycolysis activity, glucose consumption remained at 98.0% or higher in all cases except for the closed circuit RF reactors, which each had 97.9% glucose consumption. There is a significant inverse correlation between glycolysis activity

5-88 and relative abundance of Pectinatus (p = 0.03), indicating that the total culture glycolysis activity may actually remain unchanged if not for the lack of the Pectinatus proteome.

Electron Mediation

Now that the metabolic functionalities of each of the production modes have been described, the metabolic mechanisms responsible for this change in functionality can be explored. It has been assumed in previous electro-fermentation studies that interaction with the electron mediation system, via surface flavoproteins such as cytochromes, is responsible for the observed changes in product spectrum123,124,137.

Here, the non-normalized enzyme expression related to electron mediation was consistently 6-12 L2FC higher across all mediators compared to control, with the highest expression levels in NR for both the closed and open circuit sets (Figure 36). Expression difference between closed and open circuit reactors was less than 10% for all mediators except MV, which had 13.8% higher expression in closed than open circuit. When normalized for relative abundance, expression remained 4-6 L2FC higher in mediated reactors versus control, except for in MV. In the normalized sets, NR again showed the highest regulation in both the closed and open circuit sets. The drop-off in measured expression at MV was almost certainly a result of lack of proteome information for Pectinatus.

This increase in total expression was largely due to oxidoreductase activities of Clostridium and Propionibacterium, surface-bound electron transfer flavoproteins of Megasphaera, and ferrodoxin from Streptomyces. The increased activities of these systems seems to support the assumption that it is the electron mediation mechanism, especially surface flavoproteins, that is responsible for the change in product spectrum in electro-fermentation systems. However, there was no significant correlation found between expression of the electron mediation system and expression of either the butyrate pathway or the propionate pathway.

While electron mediation activity was increased in mediated reactors compared to control, extremely low current rules out any mediator redox cycling between cell and cathode. The lack of mediator cycling indicates that there was limited interaction with the culture electron transport chain. While similar pure culture studies resulted in an increasing effect on generation of reduced products which correlated with reduction potential of the mediator, the action of the mediator on a mixed culture appears to be more similar to that of Kracke137, in which no notable shift in product spectrum occurred except with a mediator of lower reduction potential than the ferredoxin pool. This is unexpected because a mixed culture should have access to a full range of electron transport pathway mechanisms, which should

5-89 Figure 36: Expression of electron mediation enzymes A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open-circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE. offer some redox interaction with the mediators. It is possible that the reason this does not happen in a mixed culture is due to the wide variety of metabolic pathway availability (also described as culture robustness) combined with interspecies redox signalling which supersedes the activity of the mediator compounds and prevents alteration of sub-ferredoxin EMC pools.

5-90 Oxidative Stress Response

Oxidative stress response was consistently up-regulated in mediated reactors compared to control when measured as non-normalized expression (Figure 37). This was true for both closed circuit (3-6 L2FC higher) and open circuit (0.5-3.5 L2FC higher), and indicates that the presence of mediator compounds imparts an oxidative stress on the community. Expression in closed circuit reactors was 20-35% higher than in open circuit reactors, except for RF, which was less than 1% different. This seems to indicate that the degree of reduction

Figure 37: Expression of oxidative stress regulation enzymes A) total expression, and B) expression normalized by genus relative abundance. The mediated experiments depict the average of 3 samples from the steady state period of open- circuit control reactors on the left of each set, and the experimental reactors (-700 mV vs Ag/AgCl) averaged together on the right. The mediator E’ are given relative to SHE.

5-91 of the mediators influences the extent of the oxidative stress response, where a more reduced state imparts a higher oxidative stress response on the culture than when mediators are present in equilibrium between reduced and oxidized states (Table 15). The lack of significant difference in oxidative stress expression between closed and open circuit for RF may be due to the presence of specific nutrient uptake mechanisms for RF.

When adjusted for genus relative abundance, total oxidative stress expression remains higher in mediated reactors than control, but expression decreases with increasingly negative mediator E’. Low expression in the MV set is again most likely due to lack of proteome data for Pectinatus, while decline from AQ to NR is primarily due to Kl-Sudm and Kl-Thrd in closed circuit and Mg-Glpd in open circuit. Considering the lack of any such trend in the total culture expression, this expression pattern appears to be due to changes in relative abundance of Klebsiella and Megasphaera, as well as the lack of proteome information for Pectinatus (Figure 27).

As described in Section 1.3.3, the oxidative stress response mechanism is highly integrated with electron mediation, and by extension, influences the degree of reduction of the product spectrum. The consistently increased oxidative stress response observed here indicates that oxidative stress is a key mechanism by which external electron mediators influence the product spectrum in mediated electro-fermentation.

While the extent of the oxidative stress response does not appear to be linked to the mediator E’, it does appear linked to the expression of propionate pathway enzymes. A linear model of propionate enzyme expression response to oxidative stress mediator expression was tested via regression, and found that there is a significant inverse correlation between the two (p = 0.009). This may explain why propionate is sometimes observed at high pH in mixed culture systems21, as oxidative stress response has been shown to be inversely related to pH183. Propionate formation therefore appears to be generally supressed by oxidative stress. This result at first appears to contradict the fact that oxidative stress response is generally increased in the culture as a whole in conditions where propionate production has increased. However, these results need to be taken in context with the fact that the expression changes of oxidative stress mediating and propionate producing enzymes of Pectinatus could not be measured at this time.

Effects of Mediators versus Applied Potential on Product Spectrum

The steady state cathodic current was consistently in the 10 µA to 1 mA range, and always below the max theoretical electron supply from fully reduced mediators (Table 15). This

5-92 suggests that redox cycling of mediators between cells and the cathode occurred either at a very slow rate, or not at all. Since the mediators, as fed to the reactors in the media, are expected to be at some equilibrium between oxidized reduced forms, the measured cathodic current must have been primarily the result of the reduction of the mediator from the feed. Preliminary analysis of the voltammograms recorded on the mediators (Figure 25), as well as the current drawn upon initiation of the potentiostat, suggests that the lack of mediator cycling was due to the poor redox interaction between mediators and culture rather than between mediator and the electrode.

The shift from HBu-type to HPr-type fermentation clearly demonstrates that the mediators had some effect on the production capacity of the culture. However, the low measured cathodic current makes it highly unlikely that the effect was due to the increased intracellular electron availability due the presence of the mediators. A more likely explanation is that the change in production capacity was due to the change in culture phylogeny. In fact, it appears that MV provided some competitive advantage to Pectinatus, likely through toxicity to the other groups.

- MV toxicity occurs through spontaneous chemical formation of an O2 radical, which is mitigated by the oxidative stress mechanism141. A cell’s ability to mitigate the oxidative stress is a result of its capacity to produce oxidative stress enzymes184. Increased oxidative stress activity in MV-mediated reactors compared to control is consistent with this mechanism. Furthermore, increased abundance of Pectinatus in MV suggests that Pectinatus has a greater capacity to mitigate oxidative stress than other groups in the community. As described in Section 5.3.7, there is a general increase in community oxidative stress response in all mediated vessels. Linear regression of Pectinatus relative abundance versus mediator reduction potential reveals that Pectinatus abundance is at least in part determined by reduction potential (p = 5×10-10, full set), and this is true even before MV (p = 0.046, AQ to NR) (Figure 38). If Pectinatus does in fact have a high capacity to mitigate oxidative stress, then this correlation suggests that electron mediators with a more negative reduction potential generally impart a higher oxidative stress on a culture. This could be confirmed with proteomic information on Pectinatus.

Notably, the oxidative stress response is the only major metabolic function of this system that is consistently more highly expressed in closed circuit than in open circuit (Figure 39), suggesting that applied potential at the cathode may also play a role in inducing oxidative stress. Furthermore, this difference between closed and open circuit is most pronounced with MV (-394 mV). The decreased oxidative stress response of open circuit MV may explain

5-93 80

Relative Abundance Relative (%) 30

20 Pectinatus

0 0 50 100 150 200 250 300 350 400 450 Mediator Reduction Potential (-mV vs SHE)

Figure 38: Linear regression of Pectinatus relative abundance versus reduction potential of the mediator compounds. The solid line (—) incorporates the full data set, while the dashed line (- -) incorporates only up through NR, excluding MV.

10 HBu Pathway 8 HPr Pathway Glycolysis 6 e- Mediation Oxidative Stress 4

0 0 50 100 150 200 250 300 350 400 450 -2 Difference in Total Culture Response (L2FC) Response Culture Total in Difference -4

-6 Mediator E' (-mV vs SHE)

Figure 39: Difference in functional expressions between closed circuit (positive) and open circuit (negative) reactors. why propionate production is higher in the open circuit reactor than in the closed circuit reactors (Section 5.2.3). The difference between closed and open circuit oxidative stress responses appears to increase with progressively negative mediator potential, however

5-94 regression via a linear model of this response difference versus mediator potential shows no significant correlation.

While successful electro-fermentation has been demonstrated across a range of mediators using a variety of pure cultures124, it has also been shown that biological interaction with thereduced mediator is dependent on the available cellular machinery and specific thermodynamics of each species137. Each cell relies on a variety of proteins and mechanisms within the electron transport chain to maintain balanced redox potential in several interacting pools of EMCs, which range in SRP from ubiquinones in the +50 to +100 mV range to ferredoxins in the -400 to -500 mV range. Furthermore, in a robust mixed culture which presumably possesses a near full representation of available electron transport chain machinery, there are a variety of intercellular signal compounds and metabolic intermediates which may provide the means to maintain redox balance over the entire culture if not in individual cells or species. Successful electron exchange between mediator and a mixed culture may therefore require a mediator reduction potential of less than the ferredoxin pool in order to fully engage the electron transport chain of the culture as a whole.

Energy and Electron Mediation

Calculating the ATP and electron sink yields of the product spectrum (Section 3.5.2) reveals how the energy and electron mediation functionality of the culture changes with mediator (Table 17). ATP generation is linked almost exclusively to acetate and butyrate production from AQ through NR, switching to primarily acetate with some production linked to butyrate and propionate with MV. Through glycolysis, the propionate and butyrate-linked ATP generation is the same, then extra ATP is generated via butyrate formation post-glycolysis.

Surprisingly, the change in electron mediation is more gradual than is the change in energy generation. Acetate and butyrate production are the primary electron sources through all mediators, though this balance shifts more heavily in favor of acetate with MV. Through glycolysis, acetate and butyrate account for similar electron sourcing through NR, after which the butyrate load is halved and picked up by propionate. However, post-glycolysis, the switch in electron-sinking functionality switches from ethanol to propionate between AQ and NR rather than NR and MV. This again is reflective of the increase in Pectinatus relative abundance, and possible increase in oxidative stress response associated with a more electronegative mediator.

5-95 Table 17: Production of ATP and sinking of e- per biomass associated with the product spectrum of each mediator. ATP - -1 and e values are each in mmol·gCODX . The “Total” section describes functionality per product pathway from glucose uptake through end product; the “Glycolysis” section describes functionality within glycolysis distributed by product yield; and the “Post-HPy” section describes functionality per product pathway from pyruvate through end product. The values are heat-mapped level-by-level and across sections to highlight spread of functionality within each level.

ATP e- Level HLA HAC HPR ETOH HBU HLA HAC HPR ETOH HBU

Total C 0 73 0 6 54 0 -145 0 0 -72 AQ 0 50 2 5 37 0 -99 3 0 -49 AQ_OC 0 42 1 3 31 0 -84 2 0 -41 RF 0 57 1 3 35 0 -114 2 0 -46 RF_OC 0 48 1 4 30 0 -95 2 0 -40 NR 0 69 2 3 42 0 -139 5 0 -56 NR_OC 0 57 2 2 35 0 -114 3 0 -47 MV 0 77 9 3 18 0 -153 19 0 -23 MV_OC 0 61 10 2 12 0 -122 20 0 -16

Glycolysis

C 0 36 0 6 36 0 -73 0 -12 -72 AQ 0 25 2 5 25 -1 -50 -3 -10 -49 AQ_OC 0 21 1 3 21 -1 -42 -2 -5 -41 RF 0 29 1 3 23 -1 -57 -2 -7 -46 RF_OC 0 24 1 4 20 0 -48 -2 -8 -40 NR 0 35 2 3 28 0 -69 -5 -6 -56 NR_OC 0 29 2 2 24 0 -57 -3 -5 -47 MV 0 38 9 3 12 -1 -77 -19 -6 -23 MV_OC 0 30 10 2 8 0 -61 -20 -3 -16

Post-HPy C 0 36 0 0 18 0 -73 1 12 0 AQ 0 25 0 0 12 1 -50 6 10 0 AQ_OC 0 21 0 0 10 1 -42 3 5 0 RF 0 29 0 0 12 1 -57 5 7 0 RF_OC 0 24 0 0 10 0 -48 4 8 0 NR 0 35 0 0 14 0 -69 10 6 0 NR_OC 0 29 0 0 12 0 -57 6 5 0 MV 0 38 0 0 6 1 -77 38 6 0 MV_OC 0 30 0 0 4 0 -61 40 3 0

5.4. Conclusions

The pathways for product formation were consistent with the observed product spectrum in both fermentation modes. Pectinatus does not produce butyrate, so this pathway could be fully described. Additionally, propionate is the primary product of Pectinatus, and therefore analysis of this pathway was necessarily limited. Propionate production showed a strong

5-96 correlation to Pectinatus relative abundance, both including the HPr-type data (p = 8×10-22) and when considering only propionate formation in HBu-type fermentation (p = 2×10-5). While this analysis is clearly limited by lack of a Pectinatus proteome, it is clear that the observed changes in propionate formation were linked very closely to the competitiveness of Pectinatus within the system, which in turn appears linked to the oxidative stress imposed by the mediators on the culture.

From the total measureable culture, it was found that expression of propionate pathway enzymes decreases with increased oxidative stress response, and this is reflected within the HPr-type set, where the open circuit reactor demonstrated a lower oxidative stress response than the closed circuit reactors, and also had higher propionate production. This propionate pathway response to oxidative stress may explain why propionate is sometimes observed in mixed culture fermentation at high pH, where there is thought to be a decreased oxidative stress response.

In general, the closed circuit reactors demonstrated a greater oxidative stress response compared to their closed circuit counterparts, and the non-mediated control demonstrated the lowest oxidative stress response. This suggests that electron mediator compounds in general can induce an oxidative stress response on a fermentation system, and that this response is increased by application of a potential via a cathode. The fact that the product spectrum became more reduced despite the lack of current suggests that the primary driving force for this change is the oxidative stress imposed by the reduced mediators, and not electron uptake.

5-97 6. Integrated Analysis

A set of experiments were carried out in order to identify the metabolic mechanisms which cause mixed fermenting cultures to switch fermentation type from butyrate to lactate and propionate production. In these experiments, glucose was fermented continuously and at steady state under conditions which varied carbon and redox load. All other factors, including media composition, temperature, pH, and HRT were maintained constant, while substrate concentration or redox mediator compound were varied. This resulted in three distinct fermentation types: butyrate (HBu), lactate (HLa), and propionate (HPr) (Figure 41). A fourth fermentation type featuring hexanoate production (at the expense of butyrate and acetate) was also observed during initial acclimatization of the varied substrate concentration experiment. Hexanoate production was linked to chain elongation of ethanol and butyrate by Megasphaera, but this production mode was not reproducible (Section 4.4.5). Distinct taxonomies were identified for the different fermentation types (Figure 40). However, the known metabolic capacities of the dominant taxonomic groups did not always align with the product spectrum (Section 4.4.1). Through metaproteomic analysis, key switching mechanisms were identified based on methylglyoxal detoxification, oxidative stress regulation, and electron mediation.

Figure 40: Biplot of all samples from varied substrate concentration and electro-fermentation experiments showing taxonomic variance between fermentation types.

6-98 Figure 41: Product composition of the three identified production types. A) Gas composition by mol%. B) Liquid phase products by mol%. C) Total products by COD%.

6-99 6.1. Characteristics of Fermentation Types

As per the first research objective of this thesis, the characteristics of the three main production modes are here described. Where possible, these modes are compared to similar results from the literature.

HBu-type Fermentation

Low pH (5.50) was maintained throughout all of the experiments, and butyrate production was established as a baseline fermentation type for comparison to both high carbon loading and redox stress. HBu-type was characteristic of substrate-limited fermentation at acidic pH with little other stress on the system, which was consistent with the reported results from similar systems and with pH-based model predictions (Table 18). However, lower butyrate and higher acetate yields were measured here than in the literature average and in the model, which reported as much as 4-10x higher carbon processing to butyrate than to acetate, as opposed to the nearly equivalent yields by molC as measured here (25.7±0.7 22 HBu, 28.2±1.1 HAc molC%). This work used the same source of biomass as Lu et al. and achieved similar butyrate and acetate yields, and so the yield differences are likely attributable to differences in taxonomy. In this work, the primary identified butyrate producer was Megaspheara, whereas in Lu et al.22, and subsequently Mohd-Zaki et al.158, butyrate production was associated with Clostridium. Temudo et al.23 also identified Clostridium as

-1 Table 18: Product yields of HBu-type fermentation. Biomass is presented as COD yield (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase. Product Measured Yield Literature Yield† Model‡ -1 Liquid Phase mmol·molGlu HLa 10 ± 4 200 ± 460 0 HAc 846 ± 34 540 ± 460 130 HPr 40 ± 6 40 ± 20 100 EtOH 113 ± 10 170 ± 340 30 HBu 386 ± 10 430 ± 530 680 HHx 2 ± 1 0 ± 0 0 HVa 18 ± 2 0 ± 0 0 HSu 2.2 ± 0.5 0 ± 0 0 HFo 16 ± 8 0 ± 0 0 Other 3 ± 2 60 ± 90 0 X (mgCOD) 163 ± 9 170 ± 40 166 Gas Phase mol%

H2 53.2 ± 0.7 40 ± 30 47

CO2 46.8 ± 0.7 60 ± 30 53 †Literature data averaged from 20h HRT data from Temudo et al.4 and both operation modes of Lu et al.22 ‡Model data derived from González-Cabaleiro et al.27

6-100 the primary genus associated with butyrate production. The fact that HBu-type fermentation was achieved under similar conditions by two different biomass cultures taken from geographically distant locations is a testament to the robustness of the mixed culture fermentation process.

HLa-type Fermentation

Production changed from HBu-type to HLa-type fermentation when the system was supplied with substrate in excess of its consumption capacity. This product spectrum changed was accompanied by a taxonomic change from Megasphaera-dominated to Ethanoligenens- dominated. The production of lactate, however, was associated with Megasphaera and Bifidobacterium in the Ethanoligenens-dominated condition. HLa-type fermentation was observed at steady state with excess supplied substrate, as well as in transient states when substrate concentration was increased (Figure 42). Steady state product yields are listed in -1 Table 19. The kM was calculated as 340±30 mgCOD·(gCODX·h) during steady state, which did not vary significantly from the kM from HBu-type fermentation (360±40 -1 mgCOD·(gCODX·h) ). The condition which produced hexanoate did have a slightly higher -1 kM at 410±10 mgCOD·(gCODX·h) , but in general this appears to be about the maximum loadable COD for a fermentation system, above which lactate production takes over.

Under this fermentation scenario, acetate remains the primary product by molar yield, though only a slightly higher percentage of total carbon is processed to acetate (37±3%)

-1 Table 19: Product yields of HLa-type fermentation. Biomass is presented as COD yield (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase. Product Measured Yield -1 Liquid Phase mmol·molGlu HLa 614 ± 57 HAc 1100 ± 100 HPr 30 ± 5 EtOH 209 ± 22 HBu 109 ± 16 HHx 7 ± 2 HVa 3.8 ± 0.6 HSu 10 ± 4 HFo 14 ± 4 Other 0 ± 0.00 X (mgCOD) 140 ± 20 Gas Phase mol%

H2 52 ± 2

CO2 48 ± 2 -1 Consumption mgCODGlu·(gCODX·h) Kinetics (kM) 340 ± 30

6-101

Figure 42: Transition states of increasing organic load from the varied substrate concentration experiment. A) Initial reactor start-up. B) Transition from 5 to 10 g·L-1. C) Transition from 10 to 15 g·L-1.

6-102 than to lactate (31±3%). Even though ATP production is coupled to butyrate production and not to lactate production, this product spectrum actually yields greater ATP production than the acetate-butyrate combination of HBu-type fermentation (76±11 and 57±6 -1 mmolATP·gCODX , respectively). Considering the intracellular pH regulation mechanism proposed by Rodriguez et al.8, this is not entirely surprising. Under this mechanism, the acids produced at low pH trend toward an undissociated equilibrium outside of the cell

(depending on pKa). Undissociated acids are then able to freely diffuse back into the cell, where they re-dissociate. This therefore requires increased ATP expenditure to actively remove acid products and maintain the intracellular pH. The higher an acid’s stoichiometric coefficient from glucose (Table 7), the greater the required ATP expenditure to maintain intracellular pH per glucose consumed. Up to 2 lactate molecules are able to be produced from each glucose, as opposed to 1 butyrate molecule, therefore lactate production at low pH should require more ATP expenditure than butyrate production.

In addition to increased lactate and acetate yields, ethanol molar yield roughly doubled from 110±10 to 210±20, while butyrate molar yield nearly quartered from 380±10 to 110±20. All other product changes were small or insignificant.

Lactate production was also observed during start-up and transition phases, accounting for as much as 55% of the liquid phase product by COD (Figure 42). This always occurred when glucose load was increased, but was not always immediate. This delayed response is observable in the transition from 5 to 10 g·L-1 substrate concentration, where biomass yield first increased for nearly a week, and then decreased over the course of 2 weeks as the total COD concentration in the reactor built up until lactate production began at day 43. A similar trend is observable in the transition from 10 to 15 g·L-1, though at this high of substrate concentration, the transition to lactate production began much sooner.

HPr-type Fermentation

The system tested in this work switch from HBu-type to HPr-type fermentation under substrate-limited conditions at acidic pH under the influence of a chemical redox stressor which altered the taxonomy from Megasphaera-dominant to Pectinatus-dominant. This is in contrast to propionate formation in the literature, which has been achieved at basic pH21, though literature taxonomic description of HPr-type fermentation is not available. Under the literature conditions, a greater percentage of the substrate was processed into propionate and ethanol, whereas in this system, the substrate was directed to higher production of butyrate, while acetate yields were indistinguishable (Table 20). This redirection from

6-103 100

90 )

1 80 - X 70 gCOD ·

ATP 60

ATP Production (mmol Production ATP 20

0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Retention Time (d-1)

Figure 43: Comparison between this work (arrow) and results of Horiuchi et al.21 of ATP production during HPr-type fermentation. Operation in this work was at pH 5.50 while that of the literature was as 8.00. Trend line is for ATP production at pH 8.00 only. -1 Table 20: Product yields of HPr-type fermentation. Biomass is presented as COD yield (mgCOD·gCODGlu ), and gas phase products are presented as total composition of the gas phase. Product Measured Yield Literature Yield† -1 Liquid Phase mmol·molGlu HLa 7 ± 7 0 ± 0 HAc 1038 ± 59 1020 ± 90 HPr 283 ± 39 500 ± 50 EtOH 76 ± 13 160 ± 100 HBu 154 ± 39 90 ± 40 HHx 0 ± 0 0 ± 0 HVa 27 ± 10 0 ± 0 HSu 1.2 ± 0.2 0 ± 0 HFo 45 ± 14 0 ± 0 Other 0 ± 0 0 ± 0 X (mgCOD) 154 ± 9 169 ± 40 Gas Phase mol% H2 55.0 ± 0.8 20 ± 4 CO2 45.0 ± 0.8 68 ± 4 †Literature data averaged from pH 8.0 values of Horiuchi et al.21 propionate and ethanol at basic pH to butyrate at acidic pH may again result from the Rodriguez8 intracellular pH mechanism. In consideration of this, a projection of expected ATP production from the literature values HRT toward the conditions of this work predicts that ATP production at a comparable HRT and pH 8.0 should be about 40% than what was calculated at pH 5.5 (Figure 43). This is the opposite of what is expected from the Rodriguez mechanism if it is assumed that a stable ATP pool is maintained (after expenditure of

6-104 maintenance energy) in the same way that a stable NAD pool is maintained26. This could imply that there are additional pH-related maintenance tasks at high pH, such as maintenance of the cellular membrane185, which require even greater ATP generation than does intracellular pH maintenance. Additionally, this discrepancy could be due to only considering substrate-level phosphorylation, while not considering the increased ATP production from high proton gradients available at low pH186.

Similar to lactate, propionate production also notably increased during transition states (Figure 42). However, in contrast to lactate production, propionate production always began immediately, and then subsided through the transition to lactate production. This indicates that at least two separate switching mechanisms are at work as the culture metabolism evolves through the transient fermentation states.

6.2. Description of Metabolic Mechanisms

The primary observed metabolic switching mechanisms from this work are response to methylglyoxal formation, oxidative stress regulation, and electron mediation. Methylglyoxal detoxification and oxidative stress regulation were linked directly to lactate and propionate formation, respectively. Electron mediation was observed as well, but this was a response to general environmental conditions and not able to be directly correlated to any of the three production modes. The actions of these metabolic mechanisms are here described, as well as how they are employed by cells to switch between pathways.

Methylglyoxal Detoxification

Methylglyoxal is formed in cells as an offshoot of glycolysis from dihydroxyacetone-P when substrate consumption outpaces the cell’s ability to process phosphate187. This can result from phosphate limitation, but is more dependent on the substrate consumption rate considering that its production can be activated under high glucose concentrations when phosphate is in excess188.

In this system, the methylglyoxal bypass was upregulated at high substrate concentration after the glucose supply surpassed the glucose consumption rate. Despite activation of this pathway, substrate uptake and glucose processing were increased relative to low substrate concentration. Additionally, this mechanism was only observed in one taxonomic group. However, the net effect of the methylglyoxal mechanism from that one group on the culture as a whole was quite pronounced, with a complete carbon rebalance in the product spectrum. As such, it is apparent that this mechanism is meant to slow down substrate

6-105 consumption and cell growth. Presumably this is due to the inability of other parts of the metabolism to keep pace, namely phosphate uptake and processing. The end result is detoxification of methylglyoxal to lactate, and this is an advantageous product because it can be reconsumed for ATP generation when growing conditions become more favorable.

Oxidative Stress Regulation and Electron Mediation

- Oxidative stress is caused by the presence of strong oxidizers, such as O2 , H2O2, and methylglyoxal, and is mediated by an array of mediators and response systems183,189–191. Some of these mediators, such as thioredoxin, directly react with and reduce the oxidizing compound, and then require an electron mediating compound in order to re-reduce192. This consequently adds an additional strain on the electron mediation system, requiring the pathway fluxes to adjust. Other response systems, such as the FNR-FNA system, actively change the expression of pathways used for electron sinking in response to the detection of oxidizers in order to directly alter these pathway fluxes191. As such, while the oxidative stress regulation systems do not directly respond to changes in electron mediation, they do generate a strong effect on the electron balance of the system, and therefore act as a subset within the electron mediation system as a whole.

The oxidative stress-related proteins detected in this work were all of the variety which directly react with oxidizers. High oxidative stress response was predominantly detected in the mediated electro-fermentation tests, and a higher response was consistently detected in closed circuit reactors than in open circuit reactors (Section 5.3.7). The interaction between the oxidative stress response system and the electron mediation system was also observed through the similar up-expression of electron mediation proteins under those same conditions. The up-expression of the oxidative stress mechanism was independent of the cathodic current in these tests, but was dependent on the extent of mediator reduction in the vessel. It appears then that the change in product spectrum was the result of an oxidative stress response induced by the reduced mediators, and not electron uptake.

Through these tests, it was also established that oxidative stress response and propionate formation are inversely proportional (Section 5.3.7). It is surprising then that the HPr-type fermentation was established in this experimental set, and specifically with use of the mediator which has a well-established toxicity mechanism which acts by forming an - oxidative stress-inducing O2 radical. However, upon comparison to other reported HPr-type fermentations, it is evident that the propionate yield was actually lower than should be expected given the taxonomy (Figure 43).

6-106 So why then is there a correlation between oxidative stress response and propionate pathway activity? This was not directly measured through the proteomics, but a hypothesis can be developed from knowledge of the oxidative stress response system and the assumption that propionate formation is employed by cells primarily as an electron sink mechanism. The oxidative stress response system places a strain on the electron mediation system through diversion of reducing power toward regeneration of response mediators112,139,192. This effectively reduces the need to sink electrons to the substrate, allowing more carbon to be diverted to ATP generation. This effect can be seen in the oxidative stress response of fermenting Lactobacillus, where carbon flow is directed away from lactate and toward a combination of acetate, ethanol, 2,3-butanediol, and acetoin139.

Similarly, Clostridium has been shown to redirect electron sinking from H2 to 1,3-propanediol in the presence of oxidative stress-inducing methyl viologen140. As established in Sections 1.1.6 and 1.1.7, propionate is a highly effective electron sink, and so a decreased need to sink electrons would result in a decreased need to produce propionate. The reason then that propionate formation remains high in the HPr-type fermentation of this work is that Pectinatus must always rely on propionate formation as an electron sink because it is incapable of producing butyrate and lactate. The lack of butyrate production requires that it produce additional acetate for ATP generation. This results in the production of extra reducing equivalents, which must then be sunk to propionate.

Pathway Selection

The effects of these described metabolic mechanisms on the combined mixed culture metabolism are illustrated in (Figure 44). Here, it is shown that glucose uptake rate is the controlling factor on the methylglyoxal bypass, while the oxidative stress response interacts with the DCA cycle of propionate formation. Under the standard low pH and low substrate concentrations, the metabolism reverts to high butyrate generation. Low pH conditions are referred to here as standard due to the tendency toward acid rather than alcohol production in a system without pH control.

6.3. Comparison of Taxonomic and Molecular Analyses

Taxonomic identification has been traditionally used for identification of the primary active groups within a fermenting mixed culture. Comparison to the known properties of these groups in a pure culture setting has generally been used to describe metabolic function23,158. This is a technique that often describes culture function well enough, but it has limitations when the taxonomy does not obviously match the product spectrum. Here, the traditional

6-107 Figure 44: Detected and inferred metabolic interactions and pathway expression under high carbon load, low oxidative stress, and standard conditions. taxonomic approach is compared to the molecular approach for the production modes observed in this work, and an argument will be made for necessity of using both techniques together when investigating culture function.

HBu-type Fermentation

Butyrate production was observed in both experimental sets, and there was a slight taxonomic variation between the two. In the varied substrate concentration study, the

6-108 taxonomy of HBu-type fermentation was defined primarily by Megasphaera, along with Pectinatus and Bifidobacterium as sub-dominant genera. The taxonomy of HBu-type fermentation in the electro-fermentation was similarly dominated by Megasphaera, but with Clostridium as the secondary group. With Megasphaera making up 35-57% of the relative abundance throughout all HBu-type conditions, it makes sense that butyrate production is the defining characteristic of this product spectrum. However, in pure culture studies, species of Megasphaera tend to produce higher yields of butyrate than acetate171. Clostridium is also highly associated with acetate and butyrate formation in pure culture studies59, and depending on the specific species, will produce a host of other products as well59,76,193. Neither Bifidobacterium nor Pectinatus produce butyrate. However, Bifidobacterium does produce predominantly acetate under standard substrate-limited conditions172, and Pectinatus produces acetate along with propionate181.

In general, this taxonomic description agrees quite well with the observed product spectrum, though the relatively high acetate production is not completely resolved. Proteomic analysis confirms Megasphaera and Clostridium as the groups responsible for butyrate production, and identifies Ethanoligenens, Spiroplasma, and Mycobacterium as additional contributors to acetate generation. Meanwhile, the low levels of propionate formation are attributable to Megasphaera through the acrylate pathway, and Acetonema, Dialister, Megamonas, and Anaerovibrio through the DCA cycle. Pectinatus is presumably active in propionate formation as well. However, proteomic assessment of Pectinatus was not possible due to lack of information in the database.

HLa-type Fermentation

In the HLa-type production phase, the taxonomy was heavily dominated by Ethanoligenens at about 45% of the relative abundance, while Bifidobacterium, Clostridium, and Megasphaera collectively accounted for another 40%. While Bifidobacterium, Clostridium, and Megasphaera each have the capacity for lactate production, none are particularly associated with its production in pure culture59,171,172,177. Even more surprising in this regard is Ethanoligenens, which is reported to either produce no149 or extremely little173 lactate, and instead is associated with acetate and especially ethanol production. Ethanol yield did increase in this production mode, as did acetate, but lactate was the defining product.

A purely taxonomic approach for this production mode would likely attribute acetate production to each of these genera, and assume that lactate production was the result of some combined action of the three sub-dominant genera. While the molecular analysis did

6-109 confirm this, the purely taxonomic approach is not able to offer any deeper explanation for why the defining product of this product spectra was not produced by the taxonomic group which makes up nearly half of the identifiable relative abundance. Through metaproteomics, it is clear that lactate production resulted from a combination of increased lactate production by two of the sub-dominant genera, as well as decreased lactate consumption to propionate by Megasphaera.

HPr-type Fermentation

The HPr-type fermentation of this work revealed the current limitations of the molecular approach. Pectinatus accounted for 50-60% of the relative abundance for this production mode, with Megasphaera and Bifidobacterium as the main sub-dominant genera. These three genera alone can adequately describe this product spectrum, with propionate formation primarily attributable to Pectinatus, as well as some additional production from Megasphaera. All three of these genera produce acetate, but the high levels relative to propionate are likely due to the action of Bifidobacterium. Finally, the butyrate formation is attributable to Megasphaera.

The ability to attribute the switch in production mode to the action of Pectinatus was crucial for analysis of this production mode, and in Section 5.3.3, a very strong direct correlation was established between propionate production and Pectinatus relative abundance. While propionate production activity was also detected from other genera, the lack of a Pectinatus proteome meant the full proteome and propionate formation could not be correlated. Crucially, however, a link was detected between propionate pathway activity and oxidative stress response of the remaining culture in Section 5.3.7. The ability to establish this mechanism is solely due to the use of metaproteomics.

Contrast of Taxonomic and Metaproteomic Approaches

Taxonomic and metaproteomic analyses clearly each have their strengths and weaknesses. Analysis of the product spectrum through the lens of the taxonomy can reveal a great deal about presence of key populations within a mixed culture, and taxonomic analysis is well supported by the databases, but it cannot reveal any mechanistic information. Even in systems in which the product spectrum is fully describable by the taxonomy, the action of key mechanisms and minor genera are lost in taxonomic analysis alone. Conversely, metaproteomics enables detailed descriptions of the metabolic mechanisms of the culture and reveals the action of minor players within the system. However, it is limited by the availability of sequenced proteomes. Clearly, there is more descriptive value in using the

6-110 two techniques together wherein the potential functional population can be identified by taxonomy where the metaproteomic database is limited.

6.4. Biochemical Market Analysis

Having successfully achieved production of lactate and propionate, it is worthwhile to investigate possible production scenarios based upon real wastewater production and market sizes. Additionally, the challenges of extracting and purifying mixed fermentation products must also be considered.

Market Analysis of HLa and HPr-type Fermenation

Lactate and propionate are each highly valuable commodity chemicals which are used in a variety of applications, including plastics, preservatives, and pharmaceutical production194,195. In 2011, global production of lactate was estimated at 300,000- 400,000tonne with a total market value of about USD 500M and growing at a rate of 19% per year195. Propionate, by comparison, was reported to have a global production of 400,000tonne and a market value of about USD 1.07B in 2014, with an annual growth rate of 7%194. Due to their end uses, there is large demand in both markets to move away from petroleum-based production and toward production from renewable, biological processes.

In order to get an idea for a region’s potential production capacity, an estimated United States domestic waste water production of 325 L·(p·d)-1 by Smith et al.196 was used in conjunction with the typical waste water COD content of 500 mg·L-1 estimated by McCarty et al.197 and the estimated US population on 1 July, 2016 of 324 from the US Census Bureau198. These values were compared against the measured yields of major products (lactate, acetate, propionate, ethanol, and butyrate) from HLa and HPr-type fermentation in this work, and the estimated production capacities of each product from the two production modes are presented in (Table 21). Using this rough estimate, the organic matter contained in US domestic wastewater alone is enough to out-produce the global markets for each of these chemicals several times over at the yields reported in this thesis. However, these values should be interpreted with the understanding that the glucose fermentation used in Table 21: US theoretical production capacity compared to estimated 2016 global market for acids and acid derivatives.

Fermentation Measured US Production Estimated 2016 Global % of Global Type Yield (COD) Capacity (tonne·yr-1) Market (tonne) Market HLa 0.307 5,540,000 835,000† 664 HPr 0.165 2,100,000 458,000‡ 459 †UK National Non-Food Crops Centre bioeconomy lactic acid fact sheet195 ‡Market Research Store global propionic acid market report194

6-111 this thesis generally represents the best-case scenario for COD conversion and product yields. Additionally, it should be noted that the substrate concentration control scheme used to induce HLa-type fermentation requires either pre-settling of the raw sludge, or the widespread incorporation of household and workplace water-saving technologies in order to increase the organic load.

Purification Technologies and Viability of Obtained Production Yields

A significant hurdle for taking fermentation processes from production of on-site COD generators for water treatment to production of industrially viable biochemicals is economical extraction and purification of the products. Due to the low titres obtained in most fermentation processes, it is often uneconomical to extract fermentation products from raw fermentation broth. To overcome this issue, a large body of research has gone into developing bio-compatible extraction processes.

While in situ extraction processes such as sparging can be utilized in high yield production of highly volatile products199, these are generally not suitable for lower yield and acid products. Ex situ product extraction techniques currently being developed in pure culture biochemical production tend to focus on increasing the solids retention time relative to the hydraulic retention time, either by use of a growth matrix in the reactor200, or with hollow fibre membranes and a cell-recycling stream201,202. These techniques serve two purposes. First, to increase the total product yield and production rate inside the reactor through use of high cell mass densities. And second, to form a product stream which can be treated through some extraction process and then recycled back into the reactor without causing the culture to sporulate203. Liquid-liquid extraction of the cell-free stream using a toxic, but highly insoluble solvent, such as dodecanol204 or oleyl alcohol205, can then be accomplished for recovery of products before recycling partially fermented broth back to the reactor. Membrane pertraction can be used in a similar fashion, though this method does not involve a recycle stream206.

Liquid-liquid extraction appears to be the most viable method for recovering lower-titer acid products, and has been shown to recover butyrate from an HBu-type fermentation from 510 mM in the reactor to 3420 mM at 91% purity in the recovered product205. For comparison, the lactate and propionate produced in these experiments were 41.05 and 8.39 mM, respectively. Even the best-case reported mixed culture HPr-type fermentation only resulted in 24 mM propionate21. However, the in-reactor butyrate production of Wu and Yang205 was accomplished through a combination of cell immobilization and recycle. A typical butyrate

6-112 production in a more comparative system to that operated in this work is 190 mM207,208. Additionally, as identified in this thesis, there are still knowledge gaps in mixed culture fermentation, especially for HPr-type fermentation. Investigation of these knowledge gaps and application of alternative fermentation schemes will undoubtedly push these product titres higher, and possibly to levels which are economically viable for recovery and purification.

6-113 7. Recommendations for Future Work and Conclusions

7.1. Future Work

Expansion of Proteome Libraries

Analysis of the electro-fermentation proteomics in this work was limited by lack of substantial proteome information for the key genus, Pectinatus. This highlights the key current limitation of the metaproteomic method, which can only be resolved by ongoing sequencing of the proteomes of more species. Upon checking the database entries of all identified genera from literature mixed culture fermentation taxonomic analysis23,158, including the experiments within this thesis, it was found that 17% of investigated taxa did not have metabolic proteome information in the UniProt database168. Of the prominent taxa identified, 8% did not have metabolic proteome information. While this indicates that there is metabolic proteome data available for the majority of genera so far identified as major contributors to mixed culture fermentation, it does also show that mixed culture proteomic analysis is going to be limited until the proteome libraries are expanded.

From taxonomic studies of related mixed culture anaerobic process, anaerobic digestion, it was found that only 4% of investigated taxa did not have available metabolic proteome information209–211. This suggests that proteomic investigations of anaerobic digestion systems are less likely to be limited by database availability than are fermentation systems, though this will always be dependent on what contribution a particular taxa has on the system.

The taxonomy of fermentation and anaerobic digestion systems is highly related to that of the human (and other animals) microbiome, specifically that of the intestinal tract212,213. Metabolic analysis of microbiomes through metaproteomcs for medical and agricultural/veterinary applications will therefore be similarly limited by database availability. Throughout these studies, the genera identified as currently lacking complete proteome information are Pectinatus, Hespellia, Pseudobutyrivibrio, Dendrosporobacter, Zymophilus, Schwartzia, and Sporotomaculum.

Metabolic Analysis for Alternative Pathway Switching Mechanisms

Common fermentation process control mechanisms, such as pH and substrate type, have already been thoroughly explored using product analysis and taxonomic approaches4,22,23,25,158. The product spectrum effect of pH in glucose fermentation has been

7-114 especially well described, with mathematical models now able to accurately describe the observed product spectra27. However, modelling descriptions still rely on assumptions about the internal pH regulation mechanisms and electron mediation of the culture, which have not been verified through molecular analysis, and do not include functionality for controls outside of pH. Other controls, such as temperature and hydraulic retention time, have only been lightly explored compared to pH and substrate type. Even substrate type has been mainly limited to glucose, glycerol, xylose, and lactose within mixed cultures, and these only describe carbohydrate functionality.

Therefore, it is recommended that future work in mixed culture fermentation should explore the metabolic mechanisms within these other controls. Continued development in this area should focus on mechanistic verification of current models, as well as expansion to include pathways not based around glycolysis, and culture responses to changing thermal conditions, oxidative stress, retention times, and electro-fermentation.

Expansion within Electro-fermentation

Despite limitations in the proteomic analysis of the electro-fermentation results, pathway- switching mechanisms were still able to be inferred. However, it was evident that the change in fermentation type was not due to any significant mediator cycling, but due to redox stress imposed by the mediators. There was a measureable increase in intracellular electron mediation activity with all mediators, but as there was no cycling with the cathode, the mediation activity appears to have been for maintaining intracellular electron balance in response to the increased oxidative stress.

This is not to say that electro-fermentation is impossible in mixed cultures. Instead, this implies that the primary mechanism for reduction of the product spectrum is not electron uptake. In order to verify this result, a larger set of mediators needs to be explored, both to rule out chemical effects and differences in electron exchange functionality, and to induce redox pressure on the culture at a reduction potential below that of the Fd pool to see if electron uptake is possible under these conditions.

Additionally, this work only explored mediators in the negative reduction potential space in an attempt to introduce excess electrons into the system. There has been little exploration of anoxic carbohydrate metabolism in the presence of an electron mediator with a positive reduction potential. Rodriguez et al.8 proposed a mechanism for pathway switching based on upkeep ATP for maintaining intracellular pH during acetate production, and the fermentation model of González-Cabaleiro et al.27 agrees strongly with this assumption. A

7-115 good experimental test of this assumption would be to perform electro-fermentation at low pH using positive applied potentials (vs SHE). Should this mechanism be correct, there should be an observable metabolic and product spectrum shift away from acetate and toward butyrate or some other more reduced product to maintain the ATP pool, and essentially sink excess carbon. This may even result in high alcohol production as the culture would not need to balance electron sinking with ATP production via butyrate.

As such, it is recommended that exploration into electro-fermentation continue through a wide array of mediators, both in the negative and in the positive reduction potential space. Ideally, there should be an exploration of sets of mediators with quite similar reduction potentials and varying mechanisms of electron transfer to the cell. Additionally, mediators having high toxicities should be avoided in favor of synthetic cobalt compounds, such as those used by Emde and Schink123 and Kracke137. Finally, the range of mediators should extend beyond the reduction potentials of the intracellular mechanisms while avoiding H2 production in cathodic cells and O2 production in anodic cells.

Applications to Fermentation Modelling

The mechanisms described in this thesis suggest that the generalized energy mediation approach of González-Cabaleiro et al.27 are insufficient to describe metabolic shifts outside of the pH-influenced acetate-butyrate-ethanol system, from which the model was built to describe. Following from this, construction of more inclusive fermentation models will need to take into account these fundamental metabolic mechanisms. Modelling approaches will need to consider activation/deactivation of particular pathways based on stimuli such as carbon load, substrate type, and redox environment. This may also require inclusion of key enzyme activity constants, such as constants for methylglyoxal synthase, thioredoxin, and superoxide dismutase.

7.2. Conclusions

Production Modes

Three distinct production modes were detected in this work: HBu-type, HLa-type, and HPr- type. The HBu-type production was established as a baseline from which to measure changes in the metabolism, as well as for comparison to the literature. Steady state HLa- type fermentation had not been previously established for a mixed culture, instead only being observed as a transient state during high substrate loads. In this system sustained

7-116 lactate production was achieved by subjecting the culture to a sustained high substrate load. HBu-type fermentation was then recoverable from this state.

Production of propionate, also of industrial interest, was additionally achieved, though not through the hypothesized mechanism. A mediator with a known toxicity mechanism, which works through oxidative stress, likely caused the taxonomic shift to a known propionate producer. Comparison of the observed HPr-type to that of the literature revealed that this culture relied more heavily on the butyrate pathway for ATP production than did previously observed HPr-type fermentations.

Pathway Switching Mechanisms

The transition from butyrate to lactate formation resulted from methylglyoxal formation at high substrate concentrations. Under this condition, some of the glucose processing through glycolysis begins to form methylglyoxal from dihydroxyacetone-P. Being extremely toxic, the cells dedicate metabolic function to detoxifying the methylglyoxal into lactate while continuing to produce ATP though the pyruvate oxidation pathways.

The metabolic description of the redox stress response in this work was limited by the lack of proteome information of key genus Pectinatus. However, basic functionality was still able to be inferred based on measured activity of the remaining culture, known chemical properties of the mediator, and known functionality of Pectinatus. In addition to this, mechanistic information was derived about the remaining culture through metaproteomics. A consistent increase in oxidative stress response was observed in mediated vessels, particularly when the mediators were in the fully reduced state. This appears to explain why altered product spectra in electro-fermentation systems are often more reduced than is necessary for the amount of measured current.

Additionally, an inverse relationship was observed between propionate pathway expression and oxidative stress response. This response may explain the occasional reporting of propionate formation, rather than the standard ethanol formation, at high pH. The known oxidative stress imposed on the system by the mediator may then explain why this system resulted in a lower propionate yield than reported in the literature.

Increased electron mediation was also detected under increased redox load, and appears to be in direct response to the increased oxidative stress repsonse. From related pure culture work, it appears that this increased electron mediation response acted to balance the redox load through the culture as a whole. Under this scenario, mediator redox cycling

7-117 will not occur in a mixed culture except with mediators of a more negative redox potential than the intracellular Fd pool.

Taxonomic versus Metaproteomic Analysis

Crucially, this work has also demonstrated the capability of studying mixed culture processes through correlation of production data, taxonomic information, and molecular analysis. Furthermore, it demonstrates the capacity to use the different data sets to fill in otherwise missing information. For instance, the taxonomic description of the HLa-type fermentation mode was not sufficient to describe the culture functionality under those conditions. Furthermore, the metabolic mechanism which caused the transition to this fermentation mode was more intricate than expected, and relied not only on carbon redirection away from lactate reduction, but also on detoxification and processing of metabolic by-products.

Despite the importance of the metabolic description, the results of the HPr-type fermentation also revealed the current limits of this method. There is still more mechanistic information to be described for HPr-type fermentation, even though some information was discernible from the sub-dominant taxonomic groups. A general relationship between oxidative stress and propionate production was described from the proteomics analysis, but the mechanism could not be described for the change in taxonomy which caused the shift in product spectrum.

Overall, both methods of community analysis were necessary to describe the underlying mechanisms of product spectrum change. Additionally, correlations between the two enhanced the overall description of the process mechanisms. As sequencing technology and proteome databases continue to improve, the ability to correlate these analyses will become even more intricate and powerful.

7-118 8. References

1. Mundial, B. What a Waste: A Global Review of Solid Waste Management. FEBS J. 281, 4556 (2014). 2. United Nations/Framework Convention on Climate Change. Paris Agreement. 21st Conf. Parties 3 (2015). doi:FCCC/CP/2015/L.9 3. Angenent, L. T., Karim, K., Al-Dahhan, M. H., Wrenn, B. a & Domíguez-Espinosa, R. Production of bioenergy and biochemicals from industrial and agricultural wastewater. Trends Biotechnol. 22, 477–85 (2004). 4. Temudo, M. F., Kleerebezem, R. & van Loosdrecht, M. Influence of the pH on (open) mixed culture fermentation of glucose: a chemostat study. Biotechnol Bioeng 98, 69–79 (2007). 5. Batstone, D. J. et al. Anaerobic Digestion Model No . 1: IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. (IWA Publishing, 2002). 6. El-Mansi, E. M. T., Bryce, C. F. A., Demain, A. L. & Allman, A. R. Fermentation Microbiology and Biotechnology. (CRC Press, 2006). 7. Konings, W. N., Poolman, B. & van Veen, H. W. Solute transport and energy transduction in bacteria. Antonie Van Leeuwenhoek 65, 369–80 (1994). 8. Rodriguez, J., Kleerebezem, R., Lema, J. M. & van Loosdrecht, M. C. Modeling product formation in anaerobic mixed culture fermentations. Biotechnol Bioeng 93, 592–606 (2006). 9. Fonteles, T. V, Costa, M. G. M., de Jesus, A. L. T. & Rodrigues, S. Optimization of the Fermentation of Cantaloupe Juice by Lactobacillus casei NRRL B-442. Food Bioprocess Technol. 5, 2819–2826 (2012). 10. Libkind, D. et al. Microbe domestication and the identification of the wild genetic stock of lager-brewing yeast. Proc. Natl. Acad. Sci. U. S. A. 108, 14539–44 (2011). 11. Satish Kumar, R. et al. Traditional Indian fermented foods: a rich source of lactic acid bacteria. Int. J. Food Sci. Nutr. 64, 415–428 (2013). 12. Lin, Y. & Tanaka, S. Ethanol fermentation from biomass resources: current state and prospects. Appl. Microbiol. Biotechnol. 69, 627–42 (2006). 13. Sreela-or, C., Imai, T., Plangklang, P. & Reungsang, A. Optimization of key factors affecting hydrogen production from food waste by anaerobic mixed cultures. Int. J. Hydrogen Energy 36, 14120–14133 (2011). 14. Damasceno, L. M. et al. An optimized fermentation process for high-level production of a single-chain Fv antibody fragment in Pichia pastoris. Protein Expr. Purif. 37, 18–26 (2004). 15. Singh, A. K. et al. Fermentation and downstream process for high yield production of Plasmodium falciparum recombinant HRP II protein and its application in diagnosis. J. Ind. Microbiol. Biotechnol. 40, 687–95 (2013). 16. Jiang, L. et al. Enhanced butyric acid tolerance and bioproduction by Clostridium tyrobutyricum immobilized in a fibrous bed bioreactor. Biotechnol Bioeng 108, 31–40 (2011). 17. Moon, S. K., Wee, Y. J. & Choi, G. W. A novel lactic acid bacterium for the production of high purity L-lactic acid, Lactobacillus paracasei subsp paracasei CHB2121. J Biosci Bioeng 114, 155–159 (2012). 18. Rehm, B. H. a. Bacterial polymers: biosynthesis, modifications and applications. Nat. Rev. Microbiol. 8, 578–92 (2010). 19. Brar, S. K., Dhillon, G. S. & Soccol, C. R. Biotransformation of Waste Biomass into High Value Biochemicals. (Springer New York, 2014). doi:10.1007/978-1-4614-8005-1 20. Eng, S. C., Fernandes, X. A. & Paskins, A. R. Biochemical effects of administering shock loads of sucrose to a laboratory-scale anaerobic (UASB) effluent treatment plant. Wat Res 20, 789–794 (1986). 21. Horiuchi, J.-I., Shimizu, T., Tada, K., Kanno, T. & Kobayashi, M. Selective production of organic acids in anaerobic acid reactor by pH control. Bioresour Technol 82, 209–213 (2002). 22. Lu, Y., Slater, F. R., Mohd-Zaki, Z., Pratt, S. & Batstone, D. J. Impact of operating history on mixed culture fermentation microbial ecology and product mixture. Water Sci. Technol. 64, 760 (2011). 23. Temudo, M. F., Muyzer, G., Kleerebezem, R. & van Loosdrecht, M. C. Diversity of microbial communities in open mixed culture fermentations: impact of the pH and carbon source. Appl Microbiol Biotechnol 80, 1121–1130 (2008). 24. Ren, N., Wang, B. & Huang, J. C. Ethanol-type fermentation from carbohydrate in high rate acidogenic reactor. Biotechnol. Bioeng. 54, 428–433 (1997). 25. Temudo, M. F., Poldermans, R., Kleerebezem, R. & van Loosdrecht, M. C. Glycerol fermentation by (open) mixed cultures: a chemostat study. Biotechnol Bioeng 100, 1088–1098 (2008). 26. de Kok, S., Meijer, J., van Loosdrecht, M. C. M. & Kleerebezem, R. Impact of dissolved hydrogen partial pressure on mixed culture fermentations. Appl. Microbiol. Biotechnol. 97, 2617–25 (2013).

8-119 27. González-Cabaleiro, R., Lema, J. M. & Rodríguez, J. Metabolic energy-based modelling explains product yielding in anaerobic mixed culture fermentations. PLoS One 10, 1–17 (2015). 28. Peekhaus, N. & Conway, T. What’s for Dinner?: Entner-Doudoroff Metabolism in Escherichia coli. J. Bact 180, 3495–3502 (1998). 29. Kruger, N. J. & von Schaewen, A. The oxidative pentose phosphate pathway: structure and organisation. Curr. Opin. Plant Biol. 6, 236–246 (2003). 30. Kleerebezem, R., Rodriguez, J., Temudo, M. F. & van Loosdrecht, M. C. M. Modeling mixed culture fermentations; the role of different electron carriers. Water Sci Technol 57, 493–497 (2008). 31. von Canstein, H., Ogawa, J., Shimizu, S. & Lloyd, J. R. Secretion of flavins by Shewanella species and their role in extracellular electron transfer. Appl. Environ. Microbiol. 74, 615–23 (2008). 32. Marsili, E. et al. Shewanella secretes flavins that mediate extracellular electron transfer. Proc. Natl. Acad. Sci. U. S. A. 105, 3968–73 (2008). 33. Kracke, F., Vassilev, I. & Krömer, J. O. Microbial electron transport and energy conservation - the foundation for optimizing bioelectrochemical systems. Front. Microbiol. 6, 1–18 (2015). 34. Madigan, M., Martinko, J. & Stahl, D. in Brock Biology of Microorganisms 113–144 (Benjamin Cummings, 2012). 35. Hiu, S. F., Zhu, C., Yan, R. & Chen, J.-S. Butanol-ethanol dehydrogenase and butanol-ethanol-isopropanol dehydrogenase: different alcohol dehydrogenases in two strains of Clostridium beijerinckii (Clostridium butylicum). Appl Env. Microbiol 53, 697–703 (1987). 36. Ko, J. et al. Conversion of methylglyoxal to acetol by Escherichia coli aldo-keto reductases. J. Bacteriol. 187, 5782– 9 (2005). 37. Nakamura, C. E. & Whited, G. M. Metabolic engineering for the microbial production of 1,3-propanediol. Curr. Opin. Biotechnol. 14, 454–459 (2003). 38. Madigan, M. T., Martinko, J. & Stahl, D. Brock Biology of Microorganisms. (Benjamin Cummings, 2012). 39. Nelson, D. L. & Cox, M. M. Lehninger Principles of Biochemistry. (W.H. Freeman, 2012). 40. Alberty, R. A. in Thermodynamics of Biochemical Reactions 155–170 (John Wiley & Sons, Inc, 2003). 41. Lundblad, R. L. & MacDonald, F. M. in Handbook of Biochemistry and Molecular Biology 537–938 (CRC Press, 2010). 42. Bennett, B., Kimball, E. & Gao, M. Absolute metabolite concentrations and implied enzyme active site occupancy in Escherichia coli. Nat. Chem. Biol. 5, 593–599 (2009). 43. Buckel, W. & Thauer, R. K. Energy conservation via electron bifurcating ferredoxin reduction and proton/Na+ translocating ferredoxin oxidation. Biochim. Biophys. Acta - Bioenerg. 1827, 94–113 (2013). 44. Schink, B. Electron confurcation in anaerobic lactate oxidation. Environ. Microbiol. 17, 543 (2015). 45. Herrmann, G., Jayamani, E., Mai, G. & Buckel, W. Energy conservation via electron-transferring flavoprotein in anaerobic bacteria. J. Bacteriol. 190, 784–791 (2008). 46. Thauer, R. K., Jungermann, K., Decker, K. & Pi, P. P. H.--. Energy conservation in chemotrophic anaerobic bacteria. Bacteriol. Rev. 41, 809 (1977). 47. Demuez, M., Cournac, L., Guerrini, O., Soucaille, P. & Girbal, L. Complete activity profile of Clostridium acetobutylicum [FeFe]-hydrogenase and kinetic parameters for endogenous redox partners. FEMS Microbiol. Lett. 275, 113–21 (2007). 48. Noth, J., Krawietz, D., Hemschemeier, A. & Happe, T. Pyruvate:ferredoxin oxidoreductase is coupled to light- independent hydrogen production in Chlamydomonas reinhardtii. J. Biol. Chem. 288, 4368–77 (2013). 49. Snoep, J. L., de Mattos, M. J. T., Postma, P. W. & Neijssel, O. M. Involvement of pyruvate dehydrogenase in product formation in pyruvate-limited anaerobic chemostat cultures of Enterococcus faecalis NCTC 775. Arch Microbiol 154, 50–55 (1990). 50. Ferry, J. G. Formate dehydrogenase. FEMS Microbiol. Rev. 87, 377–382 (1990). 51. Bosma, H. J. et al. Pyruvate Dehydrogenase Complex from Azotobacter vinelandii: Structure, Function, and Inter- Enzyme Catalysis. Ann. N. Y. Acad. Sci. 378, 265–286 (1982). 52. Hansen, R. G. & Henning, U. Regulation of pyruvate dehydrogenase activity in Escherichia coli K12. Biochim. Biophys. Acta - Enzymol. Biol. Oxid. 122, 355–358 (1966). 53. Nnyepi, M. R., Peng, Y. & Broderick, J. B. Inactivation of E. coli pyruvate formate-lyase: role of AdhE and small molecules. Arch. Biochem. Biophys. 459, 1–9 (2007). 54. Leibig, M. et al. Pyruvate formate lyase acts as a formate supplier for metabolic processes during anaerobiosis in Staphylococcus aureus. J. Bacteriol. 193, 952–62 (2011). 55. Abbe, K., Takahashi, S. & Yamada, T. Involvement of Oxygen-Sensitive Pyruvate Formate-Lyase in Mixed-Acid

8-120 Fermentation by Streptococcus mutans Under Strictly Anaerobic Conditions. J Bacteriol 152, 175–182 (1982). 56. Lindmark, D. G., Paolella, P. & Wood, N. P. The Pyruvate Formate-lyase System of Streptococcus faecalis I. Purification and Properties of the Formate-Pyruvate Exchange Enzyme. J. Biol. Chem. 244, 3605–3612 (1969). 57. Uyeda, K. & Rabinowitz, J. C. Pyruvate-Ferredoxin Oxidoreductase: III. Purification and Properties of the Enzyme. J Biol Chem 246, 3111–3119 (1971). 58. Decker, K., Jungermann, K. & Thauer, R. K. Energy production in anaerobic organisms. Angew Chem Intern. Ed. 9, 138–158 (1970). 59. Jones, D. T. & Woods, D. R. Acetone-Butanol Fermentation Revisited. Microbiol. Rev. 50, 484–524 (1986). 60. Jungermann, K., Thauer, R. K., Leimenstoll, G. & Decker, K. Function of reduced pyridine nucleotide-ferredoxin oxidoreductases in saccharolytic Clostridia. Biochim. Biophys. Acta - Bioenerg. 305, 268–280 (1973). 61. Schroder, C., Selig, M. & Schonheit, P. Glucose fermentation to acetate, CO2 and H2 in the anaerobic hyperthermophilic eubacterium Thermotoga maritima: involvement of the Embden-Meyerhof pathway. Arch Microbiol 161, 460–470 (1994). 62. Jones, S. W. et al. The transcriptional program underlying the physiology of clostridial sporulation. Genome Biol 9, R114 (2008). 63. Survase, S. a, Jurgens, G., van Heiningen, A. & Granström, T. Continuous production of isopropanol and butanol using Clostridium beijerinckii DSM 6423. Appl. Microbiol. Biotechnol. 91, 1305–13 (2011). 64. Barker, H. A., Kamen, M. D. & Bornstein, B. T. The synthesis of butyric and caproic acids from ethanol and acetic acid by Clostridium Kluyveri. Proc. Natl. Acad. Sci. U. S. A. 31, 373–381 (1945). 65. Seedorf, H. et al. The genome of Clostridium kluyveri , a strict anaerobe with unique metabolic features. PNAS 105, 2128–2133 (2008). 66. Agler, M. T., Wrenn, B. A., Zinder, S. H. & Angenent, L. T. Waste to bioproduct conversion with undefined mixed cultures: the carboxylate platform. Trends Biotechnol 29, 70–78 (2011). 67. Ganigué, R., Sánchez-Paredes, P., Bañeras, L. & Colprim, J. Low fermentation pH is a trigger to alcohol production, but a killer to chain elongation. Front. Microbiol. 7, 1–11 (2016). 68. Spirito, C. M., Richter, H., Rabaey, K., Stams, A. J. M. & Angenent, L. T. Chain elongation in anaerobic reactor microbiomes to recover resources from waste. Curr. Opin. Biotechnol. 27, 115–122 (2014). 69. Grootscholten, T. I. M., Steinbusch, K. J. J., Hamelers, H. V. M. & Buisman, C. J. N. High rate heptanoate production from propionate and ethanol using chain elongation. Bioresour. Technol. 136, 715–718 (2013). 70. Prabhu, R., Altman, E. & Eitemana, M. A. Lactate and acrylate metabolism by Megasphaera elsdenii under batch and steady-state conditions. Appl. Environ. Microbiol. 78, 8564–8570 (2012). 71. González-Cabaleiro, R., Lema, J. M., Rodríguez, J. & Kleerebezem, R. Linking thermodynamics and kinetics to assess pathway reversibility in anaerobic fermentations. in 13th World Congress on Anaerobic Digestion PPA15 (IWA Publishing, 2013). 72. Petrova, P. & Petrov, K. Direct starch conversion into L-(+)-lactic acid by a novel amylolytic strain of Lactobacillus paracasei B41. Starch-Starke 64, 10–17 (2012). 73. Mercier, P., Yerushalmi, L., Rouleau, D. & Dochain, D. Kinetics of Lactic-acid Fermentation on Glucose and Corn by Lactobacillus-amylophilus. J. Chem. Technol. Biotechnol. 55, 111–121 (1992). 74. Romli, M., Keller, J., Lee, P. L. & Greenfield, P. F. Model prediction and verification of a two-stage high-rate anaerobic wastewater treatment system subjected to shock loads. Process Saf. Environ. Prot. 73, 151–154 (1995). 75. Luo, W. et al. Toxic effects of acrylic acid on C lostridium propionicum and isolation of acrylic acid-tolerant mutants for production of acrylic acid. Eng. Life Sci. 12, 567–573 (2012). 76. Selmer, T., Willanzheimer, A. & Hetzel, M. Propionate CoA-transferase from Clostridium propionicum: Cloning of gene and identification of glutamate 324 at the active site. Eur J Biochem 269, 372–380 (2002). 77. Kim, J., Hetzel, M., Boiangiu, C. D. & Buckel, W. Dehydration of (R)-2-hydroxyacyl-CoA to enoyl-CoA in the fermentation of alpha-amino acids by anaerobic bacteria. FEMS Microbiol. Rev. 28, 455–68 (2004). 78. Caspi, R. The MetaCyc Database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Research (2014). Available at: http://metacyc.org/. 79. Oude Elferink, S. J. W. H. et al. Anaerobic conversion of lactic acid to acetic acid and 1,2-propanediol by Lactobacillus buchneri. Appl. Environ. Microbiol. 67, 125–132 (2001). 80. Xue, J., Murrieta, C. M., Rule, D. C. & Miller, K. W. Exogenous or L-rhamnose-derived 1,2-propanediol is metabolized via a pduD-dependent pathway in Listeria innocua. Appl. Environ. Microbiol. 74, 7073–9 (2008). 81. Lewis, V. P. & Yang, S.-T. Propionic acid fermentation by Propionibacterium acidipropionici: effect of growth substrate. Appl Microbiol Biotechnol 37, 437–442 (1992).

8-121 82. Voolapalli, R. K. & Stuckey, D. C. Hydrogen production in anaerobic reactors during shock loads-influence of formate production and H2 kinetics. Water Res. 35, 1831–41 (2001). 83. Papoutsakis, E. T. & Meyer, C. L. Fermentation Equations for Propionic- Acid Bacteria and Production of Assorted Oxychemicals from Various Sugars. Biotechnol Bioeng 27, 67–80 (1985). 84. Suwannakham, S. & Yang, S. T. Enhanced propionic acid fermentation by Propionibacterium acidipropionici mutant obtained by adaptation in a fibrous-bed bioreactor. Biotechnol Bioeng 91, 325–337 (2005). 85. Xiao, Z. & Xu, P. Acetoin metabolism in bacteria. Crit. Rev. Microbiol. 33, 127–40 (2007). 86. Chen, G. C. & Jordan, F. Brewers’ Yeast Pyruvate Decarboxylase Produces Acetoin from Acetaldehyde: A Novel Tool To Study the Mechanism of Steps Subsequent to Carbon Dioxide Loss. Biochemistry 23, 3576–3582 (1984). 87. Garg, S. K. & Jain, A. Fermentative production of 2,3-butanediol: A review. Bioresour. Technol. 51, 103–109 (1995). 88. Skraly, F. a & Cameron, D. C. Purification and characterization of a Bacillus licheniformis phosphatase specific for D-alpha-glycerophosphate. Arch. Biochem. Biophys. 349, 27–35 (1998). 89. van Dijken, J. P. & Scheffers, W. A. Redox balances in the matabolism of sugars by yeasts. FEMS Microbiol. Rev. 32, 199–224 (1986). 90. Pagliardini, J. et al. The metabolic costs of improving ethanol yield by reducing glycerol formation capacity under anaerobic conditions in Saccharomyces cerevisiae. Microb. Cell Fact. 12, 1–14 (2013). 91. Saxena, R. K., Anand, P., Saran, S. & Isar, J. Microbial production of 1,3-propanediol: Recent developments and emerging opportunities. Biotechnol. Adv. 27, 895–913 (2009). 92. Cameron, D. C., Altaras, N. E., Hoffman, M. L. & Shaw, a J. Metabolic engineering of propanediol pathways. Biotechnol. Prog. 14, 116–25 (1998). 93. Booth, I. R. et al. Bacterial production of methylglyoxal: a survival strategy or death by misadventure? Biochem. Soc. Trans. 31, 1406–8 (2003). 94. Ferguson, G. P., Tötemeyer, S., MacLean, M. J. & Booth, I. R. Methylglyoxal production in bacteria: suicide or survival? Arch. Microbiol. 170, 209–18 (1998). 95. Hino, T. & Kuroda, S. Presence of lactate dehydrogenase and lactate racemase in Megasphaera elsdenii grown on glucose or lactate. Appl. Environ. Microbiol. 59, 255–259 (1993). 96. Huang, K.-X., Rudolph, F. B. & Bennett, G. N. Characterization of Methylglyoxal Synthase from Clostridium acetobutylicum ATCC 824 and Its Use in the Formation of 1,2-Propanediol. Appl. Environ. Microbiol. 65, 3244– 3247 (1999). 97. Liyanage, H., Kashket, S., Young, M. & Kashket, E. R. Clostridium beijerinckii and Clostridium difficile Detoxify Methylglyoxal by a Novel Mechanism Involving Glycerol Dehydrogenase. Appl. Environ. Microbiol. 67, 2004–2010 (2001). 98. Tran-Din, K. & Gottschalk, G. Formation of D(-)-1,2-propanediol and D(-)-lactate from glucose by Clostridium sphenoides under phosphate limitation. Arch Microbiol 142, 87–92 (1985). 99. Brekasis, D. & Paget, M. S. B. A novel sensor of NADH/NAD+ redox poise in Streptomyces coelicolor A3(2). EMBO J. 22, 4856–65 (2003). 100. Pei, J. et al. The mechanism for regulating ethanol fermentation by redox levels in Thermoanaerobacter ethanolicus. Metab. Eng. 13, 186–93 (2011). 101. Georgellis, D., Kwon, O. & Lin, E. C. Quinones as the redox signal for the arc two-component system of bacteria. Science 292, 2314–6 (2001). 102. Gyan, S., Shiohira, Y., Sato, I., Takeuchi, M. & Sato, T. Regulatory loop between redox sensing of the NADH/NAD(+) ratio by Rex (YdiH) and oxidation of NADH by NADH dehydrogenase Ndh in Bacillus subtilis. J. Bacteriol. 188, 7062–71 (2006). 103. Wang, E. et al. Structure and functional properties of the Bacillus subtilis transcriptional repressor Rex. Mol. Microbiol. 69, 466–78 (2008). 104. Ravcheev, D. A. et al. Transcriptional Regulation of Central Carbon and Energy Metabolism in Bacteria by Redox- Responsive Repressor Rex. J Bacteriol 194, 1145–1157 (2012). 105. Girbal, L. & Soucaille, P. Regulation of Clostridium acetobutylicum metabolism as revealed by mixed-substrate steady-state continuous cultures: role of NADH/NAD ratio and ATP pool. J. Bacteriol. 176, 6433–8 (1994). 106. Vasconcelos, I., Girbal, L. & Soucaille, P. Regulation of carbon and electron flow in Clostridium acetobutylicum grown in chemostat culture at neutral pH on mixtures of glucose and glycerol. J Bacteriol 176, 1443–1450 (1994). 107. Snoep, J. L., de Mattos, M. J. T. & Neijssel, O. M. Effect of the energy source on the NADH/NAD ratio and on pyruvate catabolism in anaerobic chemostat cultures of Enterococcus faecalis NCTC 755. FEMS Microbiol. Lett. 81, 63–66 (1991).

8-122 108. Snoep, J. L., de Graef, M. R., de Mattos, M. J. T. & Neijssel, O. M. Effect of culture conditions on the NADH/NAD ratio and total amounts of NAD(H) in chemostat cultures of Enterococcus faecalis NCTC 775. FEMS Microbiol. Lett. 116, 263–268 (1994). 109. Himmi, E. H., Bories, A., Boussaid, A. & Hassani, L. Propionic acid fermentation of glycerol and glucose by Propionibacterium acidipropionici and Propionibacterium freudenreichii ssp. shermanii. Appl Microbiol Biotechnol 53, 435–440 (2000). 110. Saint-Amans, S., Girbal, L., Andrade, J., Ahrens, K. & Soucaille, P. Regulation of Carbon and Electron Flow in Clostridium butyricum VPI 3266 Grown on Glucose-Glycerol Mixtures. J Bacteriol 183, 1748–1754 (2001). 111. Crabbendam, P. M., Neijssel, O. M. & Tempest, D. W. Metabolic and energetic aspects of the growth of Clostridium butyricum on glucose in chemostat culture. Arch Microbiol 142, 375–382 (1985). 112. Kiley, P. J. & Beinert, H. The role of Fe-S proteins in sensing and regulation in bacteria. Curr. Opin. Microbiol. 6, 181–185 (2003). 113. Nakano, M. M. Induction of ResDE-Dependent Gene Expression in Bacillus subtilis in Response to Nitric Oxide and Nitrosative Stress. J Bacteriol 184, 1783–1787 (2002). 114. Rogstam, A., Larsson, J. T., Kjelgaard, P. & von Wachenfeldt, C. Mechanisms of adaptation to nitrosative stress in Bacillus subtilis. J. Bacteriol. 189, 3063–71 (2007). 115. Maeda, T., Yoshimura, T., Shimazu, T., Shirai, Y. & Ogawa, H. I. Enhanced production of lactic acid with reducing excess sludge by lactate fermentation. J Hazard Mater 168, 656–663 (2009). 116. Leitão, R. C., van Haandel, A. C., Zeeman, G. & Lettinga, G. The effects of operational and environmental variations on anaerobic wastewater treatment systems: a review. Bioresour. Technol. 97, 1105–18 (2006). 117. Bertolino, S. M., Rodrigues, I. C., Guerra-Sa, R., Aquino, S. F. & Leao, V. A. Implications of volatile fatty acid profile on the metabolic pathway during continuous sulfate reduction. J Env. Manag. 103, 15–23 (2012). 118. Kuenen, J. G. & Johnson, O. J. in Encyclopedia of Microbiology, Third Edition, Volume 6 (ed. Schaechter, M.) 130–147 (Elsevier Inc., 2009). 119. Ezeji, T. C., Qureshi, N. & Blaschek, H. P. Acetone butanol ethanol (ABE) production from concentrated substrate: reduction in substrate inhibition by fed-batch technique and product inhibition by gas stripping. Appl Microbiol Biotechnol 63, 653–658 (2004). 120. Wu, S.-Y. et al. Fermentative hydrogen production and bacterial community structure in high-rate anaerobic bioreactors containing silicone-immobilized and self-flocculated sludge. Biotechnol. Bioeng. 93, 934–946 (2006). 121. Agler, M. T. et al. Shaping reactor microbiomes to produce the fuel precursor n-butyrate from pretreated cellulosic hydrolysates. Environ. Sci. Technol. 46, 10229–38 (2012). 122. Rabaey, K. & Rozendal, R. a. Microbial electrosynthesis - revisiting the electrical route for microbial production. Nat. Rev. Microbiol. 8, 706–16 (2010). 123. Emde, R. & Schink, B. Enhanced Propionate Formation by Propionibacterium freudenreichii subsp. freudenreichii in a Three-Electrode Amperometric Culture System. Appl Env. Microbiol 56, 2771–6 (1990). 124. Shin, H. S., Zeikus, J. G. & Jain, M. K. Electrically enhanced ethanol fermentation by Clostridium thermocellum and Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol. 58, 476–81 (2002). 125. Gorby, Y. a et al. Electrically conductive bacterial nanowires produced by Shewanella oneidensis strain MR-1 and other microorganisms. Proc. Natl. Acad. Sci. U. S. A. 103, 11358–63 (2006). 126. Reguera, G. et al. Extracellular electron transfer via microbial nanowires. Nature 435, 1098–101 (2005). 127. Bond, D. R. & Lovley, D. R. Electricity Production by Geobacter sulfurreducens Attached to Electrodes. Appl. Environ. Microbiol. 69, 1548–1555 (2003). 128. Gregory, K. B., Bond, D. R. & Lovley, D. R. Graphite electrodes as electron donors for anaerobic respiration. Environ. Microbiol. 6, 596–604 (2004). 129. Lovley, D. R. Powering microbes with electricity: direct electron transfer from electrodes to microbes. Environ. Microbiol. Rep. 3, 27–35 (2011). 130. Rabaey, K., Boon, N., Höfte, M. & Verstraete, W. Microbial phenazine production enhances electron transfer in biofuel cells. Environ. Sci. Technol. 39, 3401–8 (2005). 131. Rabaey, K. et al. Microbial ecology meets electrochemistry: electricity-driven and driving communities. ISME J 1, 9–18 (2007). 132. Kim, T. S. & Kim, B. H. Electron flow shift in Clostridium acetobutylicum fermentation by electrochemically introduced reducing equivalent. Biotechnol. Lett. 10, 123–128 (1988). 133. Peguin, S. & Soucaille, P. Modulation of metabolism of Clostridium acetobutylicum grown in chemostat culture in a three-electrode potentiostatic system with methyl viologen as electron carrier. Biotechnol. Bioeng. 51, 342–8 (1996).

8-123 134. Steinbusch, K. J. J., Hamelers, H. V. M., Schaap, J. D., Kampman, C. & Buisman, C. J. N. Bioelectrochemical ethanol production through mediated acetate reduction by mixed cultures. Environ. Sci. Technol. 44, 513–7 (2010). 135. Park, D. H., Laivenieks, M., Guettler, M. V, Jain, M. K. & Zeikus, J. G. Microbial utilization of electrically reduced neutral red as the sole electron donor for growth and metabolite production. Appl. Environ. Microbiol. 65, 2912–7 (1999). 136. Park, D. H. & Zeikus, J. G. Utilization of electrically reduced neutral red by Actinobacillus succinogenes: physiological function of neutral red in membrane-driven fumarate reduction and energy conservation. J. Bacteriol. 181, 2403–10 (1999). 137. Kracke, F. Understanding extracellular electron transport of industrial microorganisms and optimization for production application. (The University of Queensland, 2016). doi:10.14264/uql.2016.202 138. Dennis, P. G., Harnisch, F., Yeoh, Y. K., Tyson, G. W. & Rabaey, K. Dynamics of cathode-associated microbial communities and metabolite profiles in a glycerol-fed bioelectrochemical system. Appl. Environ. Microbiol. 79, 4008–14 (2013). 139. Miyoshi, A. et al. Oxidative stress in Lactococcus lactis. Genet. Mol. Res. 2, 348–59 (2003). 140. Reimann, A., Biebl, H. & Deckwer, W. D. Influence of iron, phosphate and methyl viologen on glycerol fermentation of Clostridium butyricum. Appl. Microbiol. Biotechnol. 45, 47–50 (1996). 141. Bus, J. S., Aust, S. D. & Gibson, J. E. Superoxide- and singlet oxygen-catalyzed lipid peroxidation as a possible mechanism for paraquat (methyl viologen) toxicity. Biochem. Biophys. Res. Commun. 58, 749–755 (1974). 142. Ishimoto, M., Umeyama, M. & Chiba, S. Alteration of fermentation products from butyrate to acetate by nitrate reduction in Clostridium perfringens. Z. Allg. Mikrobiol. 14, 115–121 (2007). 143. Widdel, F. & Pfennig, N. Studies on Dissimilatory Sulfate-Reducing Bacteria that Decompose Fatty Acids II. Incomplete Oxidation of Propionate by Desulfobulbus propionicus gen. nov., sp. nov. Arch Microbiol 131, 360–365 (1982). 144. Wrighton, K. C. et al. Fermentation, Hydrogen, and Sulfur Metabolism in Multiple Uncultivated Bacterial Phyla. Science (80-. ). 337, 1661–1665 (2012). 145. Chen, K. I., McEwan, A. G. & Bernhardt, P. V. Cobalt hexaamine mediated electrocatalytic voltammetry of dimethyl sulfoxide reductase: Driving force effects on catalysis. J. Biol. Inorg. Chem. 16, 227–234 (2011). 146. Fultz, M. Lou & Durst, R. A. Mediator Compounds for the Electrochemical Study of Biological Redox Systems: A Compilation. Anal. Chim. Acta 140, 1–18 (1982). 147. Yu, Y., Tangney, M., Aass, H. C. & Mitchell, W. J. Analysis of the mechanism and regulation of lactose transport and metabolism in Clostridium acetobutylicum ATCC 824. Appl. Environ. Microbiol. 73, 1842–1850 (2007). 148. Zhang, W., Werner, J. J., Agler, M. T. & Angenent, L. T. Substrate type drives variation in reactor microbiomes of anaerobic digesters. Bioresour. Technol. 151, 397–401 (2014). 149. Xing, D. et al. Ethanoligenens harbinense gen. nov., sp. nov., isolated from molasses wastewater. Int. J. Syst. Evol. Microbiol. 56, 755–760 (2006). 150. Ha, S.-J. et al. Engineered Saccharomyces cerevisiae capable of simultaneous cellobiose and xylose fermentation. Proc Natl Acad Sci U S A 108, 504–509 (2011). 151. Jeffries, T. W. Engineering yeasts for xylose metabolism. Curr. Opin. Biotechnol. 17, 320–6 (2006). 152. Pervin, H. M. et al. Drivers of microbial community composition in mesophilic and thermophilic temperature-phased anaerobic digestion pre-treatment reactors. Water Res. 47, 7098–108 (2013). 153. Bakoyianis, V., Koutinas, A. A., Agelopoulos, K. & Kanellaki, M. Comparative Study of Kissiris, γ-Alumina, and Calcium Alginate as Supports of Cells for Batch and Continuous Wine-Making at Low Temperatures. J. Agric. Food Chem. 45, 4884–4888 (1997). 154. Galanakis, C. M. et al. Effect of pressure and temperature on alcoholic fermentation by Saccharomyces cerevisiae immobilized on γ-alumina pellets. Bioresour. Technol. 114, 492–8 (2012). 155. Infantes, D., González del Campo, a., Villaseñor, J. & Fernández, F. J. Influence of pH, temperature and volatile fatty acids on hydrogen production by acidogenic fermentation. Int. J. Hydrogen Energy 36, 15595–15601 (2011). 156. Bischof, J. C. et al. Dynamics of cell membrane permeability changes at supraphysiological temperatures. Biophys. J. 68, 2608–14 (1995). 157. Ren, N., Li, J., Li, B., Wang, Y. & Liu, S. Biohydrogen production from molasses by anaerobic fermentation with a pilot-scale bioreactor system. Int. J. Hydrogen Energy 31, 2147–2157 (2006). 158. Mohd-Zaki, Z. et al. Influence of pH Regulation Mode in Glucose Fermentation on Product Selection and Process Stability. Microorganisms 4, 2 (2016). 159. Bastidas-Oyanedel, J. R., Mohd-Zaki, Z., Pratt, S., Steyer, J. P. & Batstone, D. J. Development of membrane inlet mass spectrometry for examination of fermentation processes. Talanta 83, 482–492 (2010).

8-124 160. Aston, J. E. & Peyton, B. M. Response of Halomonas campisalis to saline stress: Changes in growth kinetics, compatible solute production and membrane phospholipid fatty acid composition. FEMS Microbiol. Lett. 274, 196– 203 (2007). 161. Juste-Poinapen, N. M. S., Turner, M. S., Rabaey, K., Virdis, B. & Batstone, D. J. Evaluating the potential impact of proton carriers on syntrophic propionate oxidation. Sci. Rep. 5, 18364 (2015). 162. Rice, E. W., Baird, R. B., Eaton, A. D., Clesceri, L. S. & Editors. Standard Methods for the Examination of Water and Wastewater. (American Public Health Association, American Water Works Association, Water Environment Federation, 2012). 163. Heijnen, J. J. in Basic Biotechnology 55–71 (2006). doi:10.1017/CBO9780511802409.005 164. Engelbrektson, A. et al. Experimental factors affecting PCR-based estimates of microbial species richness and evenness. ISME J. 4, 642–647 (2010). 165. Grobbler, C. et al. Use of SWATH mass spectrometry for quantitative proteomic investigation of Shewanella oneidensis MR-1 biofilms grown on graphite cloth electrodes. Syst. Appl. Microbiol. 1–5 (2014). doi:10.1016/j.syapm.2014.11.007 166. Chang, R. Y., Etheridge, N., Nouwens, A. S. & Dodd, P. R. SWATH analysis of the synaptic proteome in alzheimer disease. Neurochem Int 87, 1–12 (2015). 167. Albertsen, M., Karst, S. M., Ziegler, A. S., Kirkegaard, R. H. & Nielsen, P. H. Back to Basics – The Influence of DNA Extraction and Primer Choice on Phylogenetic Analysis of Activated Sludge Communities. PLoS One 10, e0132783 (2015). 168. Bateman, A. et al. UniProt: a hub for protein information. Nucleic Acids Res. 43, D204–D212 (2015). 169. Tang, W. H., Shilov, I. V. & Seymour, S. L. Nonlinear fitting method for determining local false discovery rates from decoy database searches. J. Proteome Res. 7, 3661–3667 (2008). 170. Choi, M. et al. MSstats: an R package for statistical analysis of quantitative mass spectrometry-based proteomic experiments. Bioinformatics 30, 2524–2526 (2014). 171. Marounek, M., Fliegrova, K. & Bartos, S. Metabolism and some characteristics of ruminal strains of Megasphaera elsdenii. Appl. Environ. Microbiol. 55, 1570–1573 (1989). 172. de Vries, W. & Stouthamer, a. H. Fermentation of glucose, lactose, galactose, mannitol, and xylose by bifidobacteria. J. Bacteriol. 96, 472–478 (1968). 173. Castro, J. F., Razmilic, V. & Gerdtzen, Z. P. Genome based metabolic flux analysis of Ethanoligenens harbinense for enhanced hydrogen production. Int. J. Hydrogen Energy 38, 1297–1306 (2013). 174. Hino, T., Shimada, K. & Maruyama, T. Substrate preference in a strain of Megasphaera elsdenii, a ruminal bacterium, and its implications in propionate production and growth competition. Appl. Environ. Microbiol. 60, 1827–1831 (1994). 175. Malinauskas, A., Ruzgas, T. & Gorton, L. Tuning the redox potential of riboflavin by zirconium phosphate in carbon paste electrodes. Bioelectrochemistry Bioenerg. 49, 21–27 (1999). 176. Schleifer, K. H. et al. Taxonomic Study of Anaerobic, Gram-Negative, Rod-Shaped Bacteria from Breweries: Emended Description of Pectinatus cerevisiiphilus and Description of Pectinatus frisingensis sp. nov., Selenomonas lacticifex sp. nov., Zymophilus raffinosivorans gen. nov.,. Int. J. Syst. Bacteriol. 40, 19–27 (1990). 177. Pokusaeva, K., Fitzgerald, G. F. & Van Sinderen, D. Carbohydrate metabolism in Bifidobacteria. Genes Nutr. 6, 285–306 (2011). 178. HINO, T., MIYAZAKI, K. & KURODA, S. Role of extracellular acetate in the fermentation of glucose by a ruminal bacterium, Megasphaera elsdenii. J. Gen. Appl. Microbiol. 37, 121–129 (1991). 179. Satokari, R., Juvonen, R., Mallison, K., Von Wright, A. & Haikara, A. Detection of beer spoilage bacteria Megasphaera and Pectinatus by polymerase chain reaction and colorimetric microplate hybridization. Int. J. Food Microbiol. 45, 119–127 (1998). 180. Haikara, A., Penttila, L., Enari, T. M. & Lounatmaa, K. Microbiological, biochemical, and electron microscopic characterization of a Pectinatus strain. Appl. Environ. Microbiol. 41, 511–517 (1981). 181. Tholozan, J.-L., Membré, J.-M. & Grivet, J.-P. Physiology and development of Pectinatus cerevisiiphilus and Pectinatus frisingensis, two strict anaerobic beer spoilage bacteria. Int. J. Food Microbiol. 35, 29–39 (1997). 182. Kung, L. & Hession, a. O. Preventing in vitro lactate accumulation in ruminal fermentations by inoculation with Megasphaera elsdenii. J. Anim. Sci. 73, 250–256 (1995). 183. Maurer, L. M., Yohannes, E., Bondurant, S. S., Radmacher, M. & Slonczewski, J. L. pH Regulates Genes for Flagellar Motility , Catabolism , and Oxidative Stress in Escherichia coli K-12 pH Regulates Genes for Flagellar Motility , Catabolism , and Oxidative Stress in Escherichia coli K-12 †. J. Bacteriol. 187, 304–319 (2005). 184. Hassan, H. M. & Fridovich, I. Superoxide Radical and the Oxygen Enhancement of the Toxicity of Paraquat in Escherichia coli. J. Biol. Chem. 253, 8143–8148 (1978).

8-125 185. Spilling, K., Brynj??lfsd??ttir, ??sa, Enss, D., Rischer, H. & Svavarsson, H. G. The effect of high pH on structural lipids in diatoms. J. Appl. Phycol. 25, 1435–1439 (2013). 186. Meyer-Rosberg, K., Scott, D. R., Rex, D., Melchers, K. & Sachs, G. The effect of environmental pH on the proton motive force of Helicobacter pylori. Gastroenterology 111, 886–900 (1996). 187. Kalapos, M. P. Methylglyoxal in living organisms: chemistry, biochemistry, toxicology and biological implications. Toxicol. Lett. 110, 145–75 (1999). 188. Russell, J. B. Glucose toxicity in Prevotella ruminicola: Methylglyoxal accumulation and its effect on membrane physiology. Appl. Environ. Microbiol. 59, 2844–2850 (1993). 189. Nanou, K., Roukas, T. & Kotzekidou, P. Role of hydrolytic enzymes and oxidative stress in autolysis and morphology of Blakeslea trispora during ??-carotene production in submerged fermentation. Appl. Microbiol. Biotechnol. 74, 447–453 (2007). 190. O’Sullivan, D. M., McHugh, T. D. & Gillespie, S. H. Analysis of rpoB and pncA mutations in the published literature: an insight into the role of oxidative stress in Mycobacterium tuberculosis evolution? J Antimicrob Chemother 55, 674–679 (2005). 191. Bitoun, J. P., Nguyen, A. H., Fan, Y., Burne, R. a & Wen, Z. T. Transcriptional repressor Rex is involved in regulation of oxidative stress response and biofilm formation by Streptococcus mutans. FEMS Microbiol. Lett. 320, 110–7 (2011). 192. Holmgren, A. Thioredoxin structure and mechanism: conformational changes on oxidation of the active-site sulfhydryls to a disulfide. Structure 3, 239–243 (1995). 193. Girbal, L., Croux, C., Vasconcelos, I. & Soucaille, P. Regulation of metabolic shifts in Clostridium acetobutylicum ATCC 824. FEMS Microbiol. Rev. 17, 287–297 (1995). 194. John, J. Global Propionic Acid Market Set for Explosive Growth, To Reach Around USD 1.55 Billion by 2020, Growing at 7% CAGR. (2016). 195. Chemicals, R. Renewable Chemicals Factsheet Lactic Acid. Renew. Chem. Factsheet 44, 3–4 (2011). 196. Smith, A. L., Skerlos, S. J. & Raskin, L. Psychrophilic anaerobic membrane bioreactor treatment of domestic wastewater. Water Res. 47, 1655–1665 (2013). 197. McCarty, P. L., Bae, J. & Kim, J. Domestic wastewater treatment as a net energy producer - can this be achieved? Environ. Sci. Technol. 45, 7100–6 (2011). 198. Commerce, U. D. of. United States Census Bureao. (2016). Available at: http://www.census.gov/. (Accessed: 1st July 2016) 199. Xue, C. et al. High-titer n-butanol production by clostridium acetobutylicum JB200 in fed-batch fermentation with intermittent gas stripping. Biotechnol. Bioeng. 109, 2746–2756 (2012). 200. Huang, Y. & Yang, S. T. Acetate production from whey lactose using co-immobilized cells of homolactic and homoacetic bacteria in a fibrous-bed bioreactor. Biotechnol. Bioeng. 60, 498–507 (1998). 201. Kwon, S. G., Park, S. W. & Oh, D. K. Increase of xylitol productivity by cell-recycle fermentation of Candida tropicalis using submerged membrane bioreactor. J Biosci Bioeng 101, 13–18 (2006). 202. Chang, H. N., Yoo, I. K. & Kim, B. S. High density cell culture by membrane-based cell recycle. Biotechnol. Adv. 12, 467–487 (1994). 203. Huang, W.-C., Ramey, D. E. & Yang, S.-T. Continuous Production of Butanol by Clostridium acetobutylicum Immobilized in a Fibrous Bed Bioreactor. Appl Biochem Biotechnol 113, 887–898 (2004). 204. Tanaka, S. et al. Membrane-assisted extractive butanol fermentation by Clostridium saccharoperbutylacetonicum N1-4 with 1-dodecanol as the extractant. Bioresour. Technol. 116, 448–452 (2012). 205. Wu, Z. & Yang, S. T. Extractive fermentation for butyric acid production from glucose by Clostridium tyrobutyricum. Biotechnol. Bioeng. 82, 93–102 (2003). 206. Zigová, J., Šturdík, E., Vandák, D. & Schlosser, Š. Butyric acid production by Clostridium butyricum with integrated extraction and pertraction. Process Biochem. 34, 835–843 (1999). 207. Canganella, F. & Wiegel, J. Continuous cultivation of Clostridium thermobutyricum in a rotary fermentor system. J. Ind. Microbiol. Biotechnol. 24, 7–13 (2000). 208. Michel-Savin, D., Marchal, R. & Vandecasteele, J. P. Butyric fermentation: Metabolic behaviour and production performance of Clostridium tyrobutyricum in a continuous culture with cell recycle. Appl. Microbiol. Biotechnol. 34, 172–177 (1990). 209. Narihiro, T. & Sekiguchi, Y. Microbial communities in anaerobic digestion processes for waste and wastewater treatment: a microbiological update. Curr. Opin. Biotechnol. 18, 273–278 (2007). 210. Mladenovska, Z., Dabrowski, S. & Ahring, B. K. Anaerobic digestion of manure and mixture of manure with lipids: Biogas reactor performance and microbial community analysis. Water Sci. Technol. 48, 271–278 (2003).

8-126 211. Lu, Y., Slater, F., Bello-Mendoza, R. & Batstone, D. J. Shearing of biofilms enables selective layer based microbial sampling and analysis. Biotechnol. Bioeng. 110, 2600–2605 (2013). 212. Macfarlane, G. T. & Macfarlane, S. Human colonic microbiota: ecology, physiology and metabolic potential of intestinal bacteria. Scand. J. Gastroenterol. Suppl. 222, 3–9 (1997). 213. Gill, S. R. et al. Metagenomic analysis of the human distal gut microbiome. Science (80-. ). 312, 1355–1359 (2006).

8-127 Appendix A

Regulation Mechanisms in Mixed and Pure Culture Microbial Fermentation

Robert D Hoelzle1, Bernardino Virdis1,2, Damien J Batstone*1

Email address: [email protected], [email protected], [email protected]

1Advanced Water Management Centre, The University of Queensland, Brisbane, QLD 4072, AUSTRALIA 2Centre for Microbial Electrosynthesis, The University of Queensland, Brisbane, QLD 4072, AUSTRALIA

*Correspondence should be addressed to: Damien J. Batstone, Advanced Water Management Centre, The University of Queensland, St. Lucia, QLD 4072, Australia Phone: +61 7 3346 9051; Fax: +61 7 3365 4726; E-mail: [email protected]

A-1 ABSTRACT

Mixed-culture fermentation is a key central process to enable next generation biofuels and biocommodity production due to economic and process advantages over application of pure cultures. However, a key limitation to the application of mixed-culture fermentation is predicting culture product response, related to metabolic regulation mechanisms. This is also a limitation in pure culture bacterial fermentation. This review evaluates recent literature in both pure and mixed culture studies with a focus on understanding how regulation and signalling mechanisms interact with metabolic routes and activity. In particular, we focus on how microorganisms balance electron sinking while maximising catabolic energy generation. Analysis of these mechanisms and their effect on metabolism dynamics is absent in current models of mixed-culture fermentation. This limits process prediction and control, which in turn limits industrial application of mixed-culture fermentation. A key mechanism appears to be the role of internal electron mediating cofactors, and related regulatory signalling. This may determine direction of electrons towards either hydrogen or reduced organics as end-products and may form the basis for future mechanistic models.

Key words

Mixed microbial population; fermentation; electron mediation; fermentation products; waste treatment; enzyme expression

A-2 1 INTRODUCTION

Anaerobic digestion is a stepwise conversion process in which complex organic material, generally from industrial and/or municipal waste streams, is converted into biomass and biogas utilising a mixed consortia of microbes. Fermentation is a key central step in anaerobic digestion in which soluble organic monomers are converted into short-chain fatty acids, alcohols, sugar alcohols, hydrogen, and carbon dioxide (Batstone et al., 2002; Temudo et al., 2007). The generation of these products during fermentation offers an opportunity to industrially produce biochemicals useful as feedstock chemicals to replace fossil feeds (Angenent et al., 2004; Temudo et al., 2007). Fermentation is biological - catabolism without the use of an external electron acceptor, such as O2, NO3 , or CO2, and commonly relies on substrate-coupled electron-transfer reactions for adenosine triphosphate (ATP) generation (El-Mansi et al., 2006). The primary electron sink is the substrate, but excess electrons may be sunk by reducing protons to elemental hydrogen. Oxidation reactions provide the potential energy to generate cellular energy in the form of ATP. This process is known as substrate-level phosphorylation. Fermenting cells can also leverage energy from electron transfer in other ways, including through proton, cation, and

A-3 electron transport chains (Konings et al., 1994).

The lack of an external electron acceptor is crucial to the understanding of the fermentation metabolism. In aerobic respiration, the feedstock is oxidised to CO2 by coupled reduction of intracellular electron-mediation cofactors (EMC). Re-oxidising the EMCs by sinking electrons to terminal electron acceptors removes excess electrons from the cell and provides the driving force for the majority of respirative ATP generation via oxidative phosphorylation.

Because microorganisms in anaerobic systems are not able to utilise external electron acceptors, they must utilize substrate-level phosphorylation to generate the majority of their ATP. Excess electrons are then usually sunk by substrate reduction, primarily by production of short-chain fatty acids, alcohols, and sugar-alcohols, or in the form of hydrogen gas if electrons cannot be utilised in substrate-level reduction (Figure 1). As such, fermentation can be regarded as coupled substrate-level oxidation-reduction reactions, with catabolic energy being mainly derived from oxidation reactions (through substrate-level phosphorylation), and the reduction reactions mainly existing to sink electrons. Excess electrons can be sunk to hydrogen, but this can be energetically unfavorable when the hydrogen partial pressure is high (Rodriguez et al., 2006).

Fermentation is widely used (in pure and mixed culture forms) to produce food products (e.g. beverages, fermented dairy products) (Fonteles et al., 2012; Libkind et al., 2011; Satish Kumar et al., 2013), renewable fuels (e.g. hydrogen, ethanol) (Lin and Tanaka, 2006; Sreela-or et al., 2011), pharmaceuticals (e.g. antibiotics, diagnostic markers) (Damasceno et al., 2004; Singh et al., 2013), and industrial chemicals (e.g. organic acids, biopolymers) (Jiang et al., 2011; Moon et al., 2012; Rehm, 2010). Some pure culture processes, such as bioethanol production from starch by Saccharomyces cerevisiae, rely on fast fermentation times and inoculation with highly active cultures which quickly generate product concentrations that are inhibitory to competing species, and thereby minimise loss of product yield while not implementing expensive sterilization equipment (Lin and Tanaka, 2006). However, in pharmaceutical production, tight production regulations enforced by agencies like the US Food and Drug Administration and the European Medicines Agency require that secondary (contaminating) organisms be completely eliminated. Even in biochemical production, the uncontrolled high-rate fermentation of sugars by mixed cultures is normally regarded as an undesirable side- reaction caused by microbial contamination (Brar et al., 2014). This is largely because factors determining product spectrum of mixed culture systems are still not well

A-4 understood, and are hence difficult to control. While secondary products such as propionate and lactate have been observed for mixed-cultures (Eng et al., 1986; Horiuchi et al., 2002), mixed-culture fermentation under controlled conditions thus far results only in acetate, butyrate, and ethanol production from glucose (Lu et al., 2011; Temudo et al., 2007; Temudo et al., 2008a). As a result, current mixed-culture fermentation models are only able to accurately predict product spectrums at low substrate concentrations and under a limited pH range (Kleerebezem et al., 2008; Rodriguez et al., 2006). In this review, we aim to provide a broader analysis of mixed-culture fermentation from a molecular standpoint, first through a discussion of the metabolic pathways available, and second through an exploration of relevant metabolic control systems. On the basis of this, we provide several methods that could be utilised to exploit these metabolic control systems to gain control over a wider section of the fermentation product spectrum.

2 REVIEW OF METABOLIC PATHWAYS

Fermentation products predominantly are produced from the metabolic intermediate pyruvate (Figure 1). Pyruvate is the primary intermediate resulting from the initial carbohydrate oxidation step, which is the central ATP-generating pathway in most cells. During respiration, pyruvate is then oxidised through the citric acid cycle to produce ATP while the excess electrons are used to reduce an external electron acceptor to maintain the redox balance. By contrast, in fermentation a variety of pyruvate-reducing pathways have evolved to counter pyruvate oxidation and close the electron balance. These pathways are reviewed in more detail in this section.

2.1 Primary Energy Pathways

The product pathways described in this work are primarily acidogenic, and therefore branch from pyruvate-generating pathways. For carbohydrate catabolism, this generally means Embden-Meyerhof-Parnas (EMP) glycolysis, Entner-Doudoroff (ED) glycolysis, and the pentose-phosphate pathway (Peekhaus and Conway, 1998). A wide variety of saccharides, including sucrose, lactose, glucose, fructose, and xylose, as well as a variety of sugar alcohols, can be metabolised by one or more of these pathways (Figure 2). They each eventually are metabolised to glyceraldehyde-3-phosphate (G3P). G3P is then oxidised to pyruvate through the second half of EMP glycolysis. Oxidation of 1 G3P to 1 pyruvate generates 2 ATP. This oxidation also liberates 2 electrons per G3P, which reduce EMCs and used to form reduced end products. This use of reduced end products as electron sinks maintains the catabolic electron balance.

A-5

Glycolysis Glycolysis

mediates. The stoichiometry of all

end to the right and downward from glucose. The intermediates

Phosphate Phosphate Pathway ext

based energy pathways for sugar consumption in anaerobic systems, simplified to main branching points and intermediates. EMP

tes are shown with respect to glucose and EMP Glycolysis. and respect with glucose EMP toshown are tes

Phosphate, G3P, and Pyruvate are underlined to highlight the multiple pathways which cells can use to generate the same inter

Primary Primary carbohydrate : :

Figure extends vertically downward from glucose, while ED Glycolysis Fructose 6 and the Pentose productsintermediaand

A-6 While ED glycolysis and the pentose-phosphate pathway both supply G3P for oxidation to pyruvate, they are less efficient at generating ATP than EMP glycolysis. ED glycolysis occurs mainly in aerobic environments and as a pathway to catabolize sugar acids, such as gluconate (Peekhaus and Conway, 1998). The pentose-phosphate pathway, by contrast, is primarily anabolic, generating precursors for biomass generation (Kruger and von Schaewen, 2003). Therefore, only EMP glycolysis will be considered from this point on as it is the most relevant to the fermentation metabolism.

2.2 Electron-Mediating Cofactors

Electrons are transferred in metabolic redox reactions by a set of cofactors, which we refer to collectively as EMCs. These cofactors vary in functional group, specificity, and reaction mechanism, but all serve the same basic purpose as electron carriers (by conversion between oxidised and reduced forms). EMCs play a major role in both oxidation of substrates and sinking of electrons into reduced end-products. In the anaerobic energy metabolism, the primary EMCs are nicotinamide adenine dinucleotide (NAD), nicotinamide adenine dinucleotide phosphate (NADP), flavin adenine dinucleotide (FAD), and ferredoxin (Fd) (El-Mansi et al., 2006; Kleerebezem et al., 2008). Other EMCs involved in electron transfer in anaerobic environments, though not necessarily in fermentation, include flavin mononucleotide (FMN) (von Canstein et al., 2008; Marsili et al., 2008) and cytochromes, particularly cytochrome c (Cyt c) (Inoue et al., 2011).

The structure of the functional groups of NAD and NADP is identical. NADP is primarily associated with anabolic activity (Madigan et al., 2012a). However, it is also the active EMC in several catabolic reactions, including butanol and isopropanol production (Hiu et al., 1987) and methylglyoxal conversion to acetol and lactaldehyde (Ko et al., 2005). Additionally, some catabolic reactions, such as glycerol to 1,3-propanediol (Nakamura and Whited, 2003), utilize either NAD or NADP depending on the active enzyme. In these cases, NAD and NADP are represented together as NAD(P). Due to their similarities, we will use NAD(P) unless it is necessary to distinguish between them. NAD(P), FAD, and FMN each utilize an aromatic ring structure to carry electrons. EMCs which use this mechanism are referred to collectively as quinones (Q). Additional information on the structures and mechanisms of activity is available in the literature (Madigan et al., 2012b; Nelson and Cox, 2012).

Information on the standard reduction potentials (ΔE’°) of these EMCs under typical pH and ionic strength conditions is provided in Table 1. The effect of the ionic strength on ΔE’°

A-7 Table 1: Standard reduction potentials (ΔE’°) in volts (versus standard hydrogen electron) of primary EMCs involved with anaerobic activity. Unless otherwise noted, data is from is taken from Alberty (2003). I = 0.10M I = 0.25 M Half-reaction pH 5 pH 7 pH 8 pH 5 pH 7 pH 8 †Fd3+ + e- → Fd2+ -0.4009 -0.4009 -0.4009 -0.4030 -0.4030 -0.4030 NAD+ + H+ + 2e- → NADH -0.2589 -0.3181 -0.3477 -0.2569 -0.3160 -0.3456 NAPD+ + H+ + 2e- → NADPH -0.2636 -0.3227 -0.3523 -0.2574 -0.3166 -0.3462 ‡ + + - FAD + 2H + 2e → FADH2 - -0.219 - - - - + - FMN + 2H + 2e → FMNH2 -0.1007 -0.2190 -0.2781 -0.1027 -0.2210 -0.2801 ҂ - Cyt cox + e → Cyt cred 0.2223 0.2223 0.2223 0.2121 0.2121 0.2121

†The source of Fd is not specified in Alberty (2003), however Lundblad and MacDonald (2010b) list ΔE’° of Fd from Clostridium as -0.413V. ‡ΔE’° data for FAD is not available in Alberty (2003). The value listed is sourced from Lundblad and MacDonald (2010). This value was measured at pH 7.14 with unspecified ionic strength. ҂ The source of Cyt c is not specified in Alberty (2003), however Lundblad and MacDonald (2010b) list ΔE’° of Cyt c552 from Pseudomonas and Cyt c4 from Azotobacter as 0.300V.

for each of the EMCs is notable, as is the effect of pH on the hydrogen-binding EMCs, highlighting the impact of these two factors on metabolic thermodynamics. More extensive reduction potential data for these species is available in the literature (Alberty, 2003; Lundblad and MacDonald, 2010).

Hydrogenase enzymes facilitate the transfer of electrons between EMCs and metabolic intermediates by transferring hydrogen bonds. They are the primary enzyme responsible for electron removal from the cell and generally act by hydrogenating double bonds in alkene or ketone groups. Some, such as H2:Fd hydrogenase, reduce protons to make molecular hydrogen. They may also work in reverse to transfer electrons from a metabolic intermediate to an EMC by removing hydrogen from an intermediate, often from an -OH group resulting in a ketone, such as with glycerol dehydrogenase (Nakamura and Whited, 2003). Hydrogenase enzymes are part of a wider group of enzymes known as oxidoreductases. Oxidoreductases catalyse metabolic redox reactions and do not necessarily involve hydrogenation. These enzymes are responsible both for the generation of reduced EMCs, through steps like glucose oxidation to pyruvate, and for oxidation of the cofactors to facilitate electron removal via the various reduced end products that make up the fermentation product spectrum.

2.3 Acetyl-CoA Branches

The metabolic pathways branching from pyruvate-derived acetyl-CoA (Figure 3) make up the most common pathways active in mixed-culture fermentation (Decker et al., 1970),

A-8 Figure 3: Products and pathways branching directly from acetyl-CoA generation. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP Glycolysis.

likely due to their ATP generating capacity via substrate-level phosphorylation. These pathways are able to efficiently regulate energy production and electron removal through selection of a range of end products (Jones and Woods, 1986). End products derived from acetyl-CoA are acetate, n-butyrate (butyrate), ethanol, n-butanol (butanol), acetone, and isopropanol.

The acetate and butyrate pathways are major sources of ATP generation in fermentative systems, responsible for nearly half of the substrate-level phosphorylation in an EMP glycolysis system (Jungermann et al., 1973; Schroder et al., 1994). Butyrate generation also serves to sink excess electrons. Ethanol, another major fermentation product, and butanol are also routes for electron sinking by reduction of acetyl-CoA and butyryl-CoA (an acetyl-CoA derivative), respectively (Jones and Woods, 1986). Acetone results from a decarboxylation of acetoacetate, which is produced when acetate and butyrate are reconsumed back to acetyl-CoA and butyryl-CoA, respectively. Re-consumption of acetate and butyrate in this method occurs as part of the sporulation process in solventogenic species of Clostridia and results in the production of ethanol and butanol

A-9 (Jones et al., 2008). Some strains of C. beijerinckii are also able to sink electrons to isopropanol by reducing acetone (Survase et al., 2011).

In the initial step of this set of pathways, pyruvate is oxidised to acetyl-CoA by one of three enzymes: the Pyruvate Dehydrogenase Complex (Pdhc), Pyruvate-Formate Lyase (Pfl), and Pyruvate-Fd Oxidoreductase (Pfo) (Demuez et al., 2007; Noth et al., 2013; Snoep et al., 1990). Pdhc is a compound enzyme structure that oxidises pyruvate through a series of steps which ultimately liberates CO2 and reduces a bound FAD. The reduced FAD is re- oxidised either by passing its hydrogen bonds to NAD(P) or by producing H2, regenerating Pdhc for oxidation of the next pyruvate. By contrast, Pfl oxidises pyruvate by direct replacement of the carboxyl group with CoA, generating formate rather than CO2 and H2. + At low pH, formate can be oxidised to CO2 by reduction of NAD through the enzyme formate dehydrogenase (Fdh) (Ferry, 1990), however generation of NADH in this way is not associated with pyruvate oxidation. Pfo acts similarly to Pdhc, directly producing CO2 and reducing an EMC, though Fd is reduced instead of FAD. Pfo appears to be highly prevalent among strictly anaerobic bacteria, especially in species of Clostridia, as well as some species of algae capable of dark fermentation (Noth et al., 2013).

Pdhc from Azotobacter (Bosma et al., 1982), Escherichia (Hansen and Henning, 1966), and Enterococcus (Snoep et al., 1990) is active in both aerobic and anaerobic systems. By contrast, Pfl from Escherichia (Nnyepi et al., 2007), Staphylococcus (Leibig et al., 2011), and Streptococcus (Abbe et al., 1982; Lindmark et al., 1969), and Pfo from Clostridium (Uyeda and Rabinowitz, 1971) are inhibited by oxygen and therefore primarily active in anaerobic systems . Additionally, Snoep et al. (1990) demonstrated using Enterococcus faecalis under anaerobic conditions that Pdhc is most active at low pH (below 6.5) while Pfl is most active at high pH (above 6.5). However, it is uncertain whether this pH-dependent expression occurs universally. Mixed-culture fermentation experiments by Temudo et al. (2007) and Lu et al. (2011) show low formate production and high CO2 production at low pH; at high pH, the reverse was true. However, it is not clear whether these changes in formate and CO2 formation were due to changes in relative activity of Pdhc/Pfo and Pfl, or due to pH-dependent activity of Fdh. It can therefore not be said for certain whether reduction of Fd, FAD, NAD, and NADP is directly associated with pyruvate oxidation or formate oxidation, or with what ratio they are reduced at this stage of the metabolism.

Table 2 highlights the difference in oxidised EMC yield (from glucose) associated with each product based on Pdhc/Pfo and Pfl oxidation of pyruvate. As pointed out by

A-10 Table 2: roduct and cofactor molar yields of the acetyl-CoA branches associated with Pdhc/Pfo (top) and Pfl (bottom) activity with respect to glucose (Glu) and EMP Glycolysis.

Yield Acetate Butyrate Acetone Isopropanol Ethanol Butanol Formate Pdhc/Pfo Y (Product/Glu) 2 1 1 1 2 1 0 Y (ATP/Glu) 4 3 2 2 2 2 0 Y (e-/Glu) -8 -4 -8 -6 0 0 0

Pfl Y (Product/Glu) 2 1 1 1 2 1 2 Y (ATP/Glu) 4 3 2 2 2 2 2 Y (e-/Glu) -4 0 -4 -2 4 4 -4 Note: The electron (e-) yield is the number of electrons that are sunk into the product per glucose. Every 2 e- yielded is equivalent to oxidation of 1 NADH to NAD+, which are necessary to regenerate to continue glucose oxidation. Refer to Table 1 for the e- yields of other primary EMCs.

Kleerebezem et al. (2008), it should not be assumed that any given reduced EMC can be used to produce any given reduced end product. Most enzymatically catalyzed redox reactions have specificity for one particular EMC, as with NADP specificity in 1,3- propanediol production (Nakamura and Whited, 2003), and butanol and isopropanol production (Hiu et al., 1987). Additionally, production of some reduced products under certain conditions is only thermodynamically possible with an EMC with a low enough

ΔE’°. An example is proton reduction to H2 under high H2 partial pressures, which is only possible using reduced Fd (Kleerebezem et al., 2008). It is therefore necessary to determine which specific enzymes are active during these oxidation steps, and in what ratios, in order to identify the range and likely stoichiometric ratios of possible products.

2.4 L-Lactate Branches

The ability to produce L-lactate (lactate) is widespread among microorganisms. It is the primary fermentation product for some species of Lactobacillus, but for many organisms it serves as an intermediate which is later re-consumed (Jones and Woods, 1986). Lactate has the same carbon oxidation state as glucose. It is therefore generally ineffective as an electron sink, even though it is produced by direct pyruvate reduction. However, several species are able to metabolise lactate into products that are more reduced than glucose, making them more effective as electron sinks than lactate. The acrylate pathway, utilised by Clostridium propionicum (Luo et al., 2012; Selmer et al., 2002), Megasphaera elsdenii (Kim et al., 2004), and Prevotella ruminicola (Caspi, 2014), reduces lactate to propionate, sinking 4 electrons per glucose (Table 3). Similarly, Lactobacillus buchneri reduces lactate to 1,2-propanediol (Oude Elferink et al., 2001), also eliminating 4 electrons per glucose. 1,2-Propanediol can also be reduced to propanol or converted to propionate for ATP

A-11

Figure 4: Products and pathways branching directly from pyruvate reduction to L-lactate. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP Glycolysis. generation by Listeria innocua and species of Salmonella (Xue et al., 2008). Production of propanol and propionate from 1,2-propanediol has not been observed as originating from lactate-produced 1,2-propanediol, though it is possible that this kind of syntrophic

Table 3: Product and cofactor molar yields of the L-lactate branches with respect to glucose (Glu) and EMP Glycolysis. Yield Lactate 1,2-Propanediol Propanol Propionate (a)† Propionate (b)‡ Y (Product/Glu) 2 2 2 2 2 Y (ATP/Glu) 2 2 2 4 2 Y (e-/Glu) 0 8 12 4 4 † Propionate (a) refers to propionate production via the 1,2-propanediol pathway. ‡Propionate (b) refers to propionate production via the acrylate pathway.

A-12 relationship could exist in mixed culture systems. This branch of the metabolism is therefore presented here as proposed possible mixed culture metabolism (Figure 4).

As noted above, lactate is generally not expressed as a primary fermentation product except in species of Lactobacillus. Since it has the same carbon oxidation state as glucose, it is useful as an energy source and is readily metabolised through pyruvate to acetate and butyrate. Some species, such as Propionibacterium acidipropionici, couple ATP generation via lactate oxidation to acetate with electron sinking via lactate reduction to propionate (Lewis and Yang, 1992). In anaerobic digestion processes, lactate and propionate accumulation are viewed as evidence of inhibition within the methanogenic process and are often associated with shock loads of substrate (Eng et al., 1986; Voolapalli and Stuckey, 2001). Lactate and propionate have also been observed in the mixed-culture fermentation product spectrum (Lu et al., 2011; Temudo et al., 2007), though the stimuli for their production has yet to be accurately described or modelled. If lactate is supplied as a primary substrate, it can either be reduced by the pathways in Figure 4 or oxidised to pyruvate. When it is reduced, electrons must be sourced from molecular hydrogen (Junicke et al., 2013).

2.5 Dicarboxylic Acid Cycle

Species of Propionibacteria are able to produce propionate and succinate through the dicarboxylic acid (DCA) cycle (Papoutsakis and Meyer, 1985; Suwannakham and Yang, 2005). Succinate is used in many polyesters and resins, and is also used as a pH buffer in the food industry (Zeikus et al., 1999). The DCA cycle acts analogously in reverse to the citric acid cycle of aerobic metabolism, reducing oxaloacetate to succinate through L- malate and fumarate (Figure 5). Propionate is generated when propanoyl-CoA transfers its CoA to succinate, starting the regeneration stage of the cycle. The final step is regeneration of oxaloacetate by transfer of the CoA to pyruvate, generating propanoyl- CoA.

Propionate generation is the result of transfer of a CoA to pyruvate, while succinate generation is the result of phosphoenolpyruvate entering the DCA cycle. Therefore,

Table 4: Product and cofactor molar yields of the Dicarboxylic Acid Cycle with respect to glucose (Glu) and EMP Glycolysis. Yield Succinate Propionate Y (Product/Glu) 2 2 Y (ATP/Glu) 0 2 Y (e-/Glu) 4 4

A-13

Figure 5: Products and pathways of the Dicarboxylic Acid Cycle. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP Glycolysis. propionate generation from the DCA cycle is associated with the final ATP generation step of glycolysis, while succinate is not (Table 4). By extension, since each propionate produced from the DCA cycle constitutes one turn of the cycle, the oxidised EMC yield of each propionate is that of one turn of the cycle.

2.6 Alternative Pyruvate Branches

Other products that can be metabolised from pyruvate include ethanol, R-acetoin (acetoin), and 2R,3R-butanediol (Figure 6). Ethanol derived in this way bypasses acetyl- CoA, using a direct decarboxylation of pyruvate to acetylaldehyde, and then reduction to ethanol. This ethanol pathway is utilised most notably by Zymomonas mobilis and eukaryotes, such as Saccharomyces cerevisae, for industrial ethanol biofuel fermentation (Lin and Tanaka, 2006). Acetoin, being one of the molecules which make the characteristic butter flavour, is widely used in the food industry for flavouring and fragrance (Crout et al., 1986; Pavia, 2007). It is produced primarily from decarboxylation of α-acetolactate during the exponential growth phases of bacteria such as Bacillus subtilis and Klebsiella aerogenes as a pH-neutral energy storage molecule which can be re-consumed during stationary phase (Xiao and Xu, 2007). Acetoin is also produced as a side product of pyruvate decarboxylase in species of Saccharomyces (Chen and Jordan, 1984). 2,3-Butanediol is one of the products produced by acetoin re-consumption in species of Bacillus, Klebsiella, Serratia, and

A-14

Figure 6: Products of the alternative pyruvate-consuming pathways. R-acetoin formation from pyruvate is a side reaction of pyruvate decarboxylase that reacts one pyruvate with one acetylaldehyde. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP Glycolysis. Table 5: Product and cofactor molar yields of the alternative pyruvate branches with respect to glucose (Glu) and EMP Glycolysis. Yield Ethanol R-Acetoin 2R,3R-Butanediol Y (Product/Glu) 2 1 1 Y (ATP/Glu) 2 2 2 Y (e-/Glu) 0 -4 -2 Pseudomonas (Garg and Jain, 1995; Xiao and Xu, 2007) and has commercial value as an antifreeze, fuel additive, and in rubber production (Garg and Jain, 1995).

Aside from ethanol, which results in identical yields in ATP and oxidised EMCs whether or not it is produced through an acetyl-CoA step, the products of this pathway group each represent a net loss in oxidised EMCs (Table 5). As a result, acetoin is not useful to cells as an electron sink, and 2,3-butanediol must be produced alongside other reduced cofactors to maintain the cellular redox balance.

2.7 Glycolysis Side-Branches

Electrons can be sunk by directing carbon flow away from pyruvate production (Figure 7). Carbon flow of this kind is generally associated with the glycerol branch of the energy pathways (Figure 2) and the methylglyoxal bypass (Booth et al., 2003). Products produced by carbon flux away from pyruvate are glycerol, 1,3-propanediol, 1,2-propanediol, and D-lactate. Their respective yields are described in Table 6.

Glycerol is produced by many species during sugar fermentation, most notably by Saccharomyces (van Dijken and Scheffers, 1986; Pagliardini et al., 2013) and Bacillus

A-15

Figure 7: Products and pathways branching directly from glycolysis. The stoichiometry of all products and intermediates are shown with respect to glucose and EMP Glycolysis. Table 6: Product and cofactor molar yields of the glycolysis side-branches with respect to glucose (Glu) and EMP Glycolysis. Yield Glycerol 1,3-Propanediol 1,2-Propanediol D-Lactate Y (Product/Glu) 2 2 2 2 Y (ATP/Glu) 0 0 -2 -2 Y (e-/Glu) 4 8 8 0

(Skraly and Cameron, 1998). Glycerol is readily consumed by many bacteria for energy production. Since it is a reduced substrate compared to glucose (Figure 1), anaerobic catabolism of glycerol requires increased production of reduced products. Some species, including species of Klebsiella, Clostridia, Citrobacter, Lactobacillus, and Enterobacter, deal with the increased load of reduced EMCs by directly reducing some of the glycerol to 1,3-propanediol (Saxena et al., 2009). However, no species are known to be able to ferment sugars to 1,3-propanediol due to sugar repression of the 1,3-propanediol pathway (Cameron et al., 1998).

The methylglyoxal bypass is a highly conserved pathway utilised in conditions of phosphate limitation and when carbon flow into glycolysis is greater than the processing capacity of the cell’s metabolism (Booth et al., 2003; Ferguson et al., 1998). However, methylglyoxal is highly toxic, and so detoxification mechanisms are necessary to avoid cell death. The most common pathway for dealing with methylglyoxal is hydration to D-lactate, which can then either be removed from the cell or oxidised to pyruvate (Ferguson et al., 1998). Some species of Clostridia are also able to detoxify methylglyoxal by reducing it to 1,2-propanediol (Huang et al., 1999; Liyanage et al., 2001; Tran-Din and Gottschalk, 1985).

A-16 3 REDOX-RESPONSIVE METABOLIC CONTROLS

All cells use a variety of molecular signalling pathways to respond to and interact with their environment. These include metabolic regulators, such as the release of hydrolytic enzymes in response to the presence of cellulose, and anabolic regulators, such as the signals involved in regulating the cell cycle. New to the field of signalling mechanisms is a set of metabolic regulatory enzymes which specifically sense changes in the redox state of the environment and alter the metabolic flow of electrons accordingly. Some detect changes in the ratio of reduced to oxidised EMCs (Brekasis and Paget, 2003), while others detect the presence of reduced products (Pei et al., 2011). Another signalling system regulates the transcription of respirative and fermentative genes in facultative anaerobes in response to the presence or absence of an external electron sink (Georgellis et al., 2001). These signalling mechanisms make up the range of redox-responsive metabolic controls.

3.1 Repressor Enzymes

Recently, a set of regulatory enzymes where discovered that are able to respond to changes in the ratio of NADH:NAD+ by regulating the transcription of genes that code for NADH-oxidising enzymes (Brekasis and Paget, 2003; Gyan et al., 2006; Wang et al., 2008). These Rex (redox repressor) enzymes bind to operons upstream of genes coding primarily for hydrogenases and cytochromes when the NADH:NAD+ ratio is low, preventing production of enzymes that oxidise NADH (Figure 8). Forms of Rex have been found to be highly conserved across many species of bacteria, including aerobes and anaerobes, and span a wide range of distantly related taxonomic groups (Ravcheev et al., 2012). In function it has been found that Rex enzymes are either inhibited by NADH (Brekasis and Paget, 2003) or activated by NAD+ (Gyan et al., 2006), though the outcome of down- regulating NADH oxidation is the same.

In addition to regulation based on the NADH:NAD+ ratio, there is new evidence for redox potential regulating enzymes which respond to concentrations of reduced products and down regulate NADH dehydrogenases (Pei et al., 2011). In this work, Pei et al. (2011) also describe regulation of the ethanol pathway of Thermoanaerobacter ethanolicus based on NADH:NAD+ ratio. Further evidence for redox potential control based on interactions of NADH:NAD+ ratio and reduced products, specifically butanol, is also available (Girbal and Soucaille, 1994; Vasconcelos et al., 1994). These complex interactions between the NADH:NAD+ ratio, reduced product concentrations, and

A-17

Figure 8: Diagram depicting the repression mechanism of Rex enzymes for a generic dehydrogenase gene. NAD+ binds with the Rex enzyme, which changes to an active conformation. The active Rex enzyme then binds to the operon upstream of the dehydrogenase gene, blocking transcription by RNA polymerase. expression of cytochromes and NADH oxidisers maintain a stable redox potential in the cell and, similar to ATP pool maintenance, ensure that the cytoplasm maintains an appropriate level of NAD+ for reducing power in the energy metabolism.

3.2 NADH:NAD+ Response to Environmental Factors

Previous research on fermentation using Enterococcus faecalis found a correlation between the NADH:NAD+ ratio, culture pH, and relative reduction potential of the substrate (Snoep et al., 1991; Snoep et al., 1994). Using combinations of mannitol, glucose, pyruvate, and gluconate, it was found that generally the highest NADH:NAD+ ratios occurred with the most reduced substrate (with the exception of gluconate) (Snoep et al., 1991; Snoep et al., 1994). Additionally, NADH:NAD+ ratio decreased with decreasing pH (Snoep et al., 1991). Several studies have also shown increased production of reduced products as a response to fermentation of more reduced substrates such as glycerol (Himmi et al., 2000; Saint-amans et al., 2001), xylose (van Dijken and Scheffers, 1986), and mannitol (Crabbendam et al., 1985).

A-18 of the basic assumption of thermodynamic optimum (Kleerebezem et al., 2008) to also consider constraints imposed by regulation mechanisms (but still treating the system as an optimisation problem).

3.3 Response to Changes in External Redox State

Many facultative anaerobes also rely on redox-sensing enzymes to respond to changes from aerobic to anaerobic conditions. The ArcA-ArcB system of Escherichia coli responds to increases in the reduced forms of membrane-bound quinones to down regulate transcription of respiratory genes and up regulate transcription of fermentative genes (Georgellis et al., 2001). In the Fnr system, also of Escherichia coli, a transcription factor is bound to several Fe-S clusters which oxidise or reduce depending on the presence of O2 (Kiley and Beinert, 2003). This is used in turn to regulate many fermentative genes in anaerobic conditions. Bacillus subtilis responds to changes in nitrosative stress via the ResD-ResE system, which responds to changes in highly reactive nitric oxide (NO) concentration by regulating the transcription of various genes involved in aerobic and anaerobic metabolism (Nakano, 2002), as well as up-regulating flavohemoglobin - production to either oxidise NO to NO3 in aerobic conditions, or reduce NO to N2O in anoxic conditions (Rogstam et al., 2007).

Clearly, redox sensing enzymes play a vital role in regulation of cellular metabolism and in maintaining a stable redox potential within the cell. In addition to stabilizating the NADH:NAD+ ratio, they respond to fluctuations in the redox state of the quinone pool, concentration of O2, and presence of highly reactive oxidised nitrogen. This complex set of redox regulators shifts metabolism between aerobic and anaerobic, and maintains a ratio of NADH:NAD+ that is thermodynamically favorable for breakdown of the available substrates by moving electrons between EMCs, H2, and reduced products. It is likely that successful control of the fermentation product spectrum will rest on controlling the activity of the redox repressor enzymes.

4 REGULATION OF METABOLISM

It is currently known that pH (Lu et al., 2011; Temudo et al., 2007), organic loading rate (Eng et al., 1986; Temudo et al., 2008b; Voolapalli and Stuckey, 2001), substrate type (Himmi et al., 2000; Lewis and Yang, 1992; Saint-amans et al., 2001), and temperature (Batstone et al., 2002; Maeda et al., 2009; Schroder et al., 1994) are important controls on both the product spectrum and microbial population diversity. Each interacts with different

A-19 aspects of the cellular metabolism or system ecology and results in formation of specific ranges of products based on these interactions.

4.1 Regulation with pH

The pH of the media affects the extent of dissociation of organic acids produced by the cells. Rodriguez et al. (2006) hypothesized that the extent of dissociation affects the amount of ATP a cell must expend in maintenance. Undissociated organic acids (forming at low pH) passively diffuse into the cell, thereby requiring an increased rate of ATP- consuming active transport to remove the acid products. It is possible that this increased energy usage signals cells to change their primary mode of ATP generation from the acetate pathway to the lower ATP-yielding butyrate pathway at low pH. Butyrate production has a reduced impact on pH per mole glucose due to the stoichiometry of reaction. It therefore reduces active transport energy expenditure.

The metabolic shift from acetate to butyrate has been well documented in mixed cultures (Lu et al., 2011; Temudo et al., 2007). However, it has also been reported that mixed cultures will respond to low pH by increasing pH-neutral ethanol and maintaining ATP generation through acetate production (Ren et al., 1997). It is not understood what causes mixed cultures to react to low pH by producing butyrate in some cases and ethanol in other cases, though it may be linked to substrate type. Lu et al. (2011) and Temudo et al. (2007) each used glucose as their only carbon source, while Ren et al. (1997) used a combination of molasses and sucrose. Additionally, Horiuchi et al. (2002) reported increased propionate production in mixed culture glucose fermentation at pH 8.0, a result which has not been replicated.

4.2 Regulation with Organic Loading Rate and Hydraulic Retention Time

Organic loading rate (OLR) and hydraulic retention time (HRT) both affect the availability of substrate in continuous systems. OLR has not generally been considered an important primary variable in fermentation because it is a net effect of the hydraulic retention time (HRT) and input substrate concentration. However, OLR is the principal lumped variable determining maximum cellular flux (i.e., available feed). We therefore argue that OLR is a highly relevant controlling factor for determining fermentation metabolics.

The effect of OLR on mixed cultures has so far been studied primarily in the context of defining the limits of methanogenic activity in anaerobic digestion. Eng et al. (1986) described an increase in lactate and propionate levels during increased OLR (shock loads)

A-20 of sucrose in anaerobic digestion processes in relation to changes in H2 and methane production, possibly due to inhibition in degradation of these products. Voolapalli & Stuckey (2001) also reported increased VFA production in shock loads of sucrose and glucose, but did not differentiate between the acid products.

OLR has not been widely assessed separately from HRT. For example, Voolapalli and Stuckey (2001) increased OLR by decreasing HRT (at constant substrate concentration) and found results as noted above. In contrast, Agler et al. (2012) varied HRT while holding OLR constant (varied substrate concentration) and reported increased production rate of butyrate and other unspecified fermentation products with decreased HRT. This indicates that HRT and OLR can be decoupled, and it should be investigated whether varying OLR with constant HRT has a separate influence. In general, limited work has been done on loading (either through HRT or OLR) and there is ample opportunity for expanded research on this topic.

4.3 Regulation with Substrate Type

As discussed above (Section 4.1), the difference in substrate type between the experiments performed by Lu et al. (2011), Temudo et al. (2007), and Ren et al. (1997) may explain differences in production of ethanol vs butyrate at low pH. Different substrates have different and distinct degrees of reduction, and therefore generate different quantities of reduced EMCs during catabolism. Additionally, each substrate enters the fermentation metabolism at a different point (Figure 2), and therefore has a unique set of pathways and yields available to it. Certain substrates also can be inhibitory to particular pathways, such as sugar inhibition of the 1,3-propanediol pathway (Cameron et al., 1998).

Work with pure cultures has shown extensively that substrate type has a strong effect on the product spectrum. Lewis & Yang (1992) demonstrated the effects on product spectrum of Propionibacterium acidipropionici grown on glucose, lactose, and lactate. Propionate and acetate production were highest and succinate production lowest when lactate was the fermentation substrate, while lactose resulted in the highest succinate production. Yu et al. (2007) demonstrated how the substrates lactose and glucose affect fermentation gene expression in Clostridium acetobutylicum. They showed not only that genes related to lactose fermentation were only expressed in the presence of lactose, but that when provided a mix of glucose and lactose, the lactose fermentation-related genes were not expressed until glucose had been sufficiently consumed. This is evidence that the effect on product spectrum is at least a result of an adjustment in the metabolism. Girbal &

A-21 Soucaille (1994) and Snoep et al. (1991) took this a step further, showing that substrate type affects the NADH:NAD+ ratio of C. acetobutylicum and Enterococcus faecalis, respectively. However, their results were inconclusive as to whether the change in ratio was directly correlated to the relative redox state of the substrate.

In addition to the effect caused by the substrate redox potential, substrate type can also affect the system ecology. Zhang et al. (2014) showed that the bacterial microbiomes of anaerobic digesters are highly correlated to substrate type. One likely explanation for this is that many microorganisms do not possess the necessary catabolic pathways for certain substrates. For example, it is well known in the bioethanol industry that wild-type strains of Saccharomyces cerevisiae are not able to ferment xylose (Ha et al., 2011; Jeffries, 2006). The inability of a species to digest a substrate provides a selective pressure against that species in the mixed culture.

4.4 Regulation with Temperature

Similar to the ecological effects of substrate type, it is well known that system temperature has a profound effect on the ecology of mixed culture systems, providing selective pressure for psychrophiles below about 20°C, mesophiles around 30-40°C, and thermophiles above about 50°C (Pervin et al., 2013). Additionally, temperature directly affects the activation energy of a bioreaction, as well as thermodynamics of equilibrium. As shown in studies on ethanol production with Saccharomyces cerevisiae, total substrate consumption as well as ethanol yield and production rate increase with increasing temperature, while yield of volatile by-products (e.g. acetaldehyde, isobutanol, VFAs) decreased with increasing temperature (Bakoyianis et al., 1997; Galanakis et al., 2012). Infantes et al. (2011) also showed increased fermentation rate with increasing temperature with mixed cultures, however the individual product responses (VFAs and H2 only) were more heavily based on pH than on temperature. In general, oxidative reactions (producing hydrogen as reduced product) are favored at high temperature over reductive reactions (Batstone et al., 2002).

A secondary thermodynamic effect of temperature is on cellular membrane permeability. Bischof et al. (1995) showed this effect using a fluorescent dye to track membrane permeability versus temperature using two types of animal cells. Infantes et al. (2011) also hypothesized that increased membrane permeability was responsible for incomplete substrate consumption at high temperature and low pH. No references could be found that

A-22 assess temperature-dependent membrane permeability studies on fermenting microorganisms.

4.5 Regulation with Electro-fermentation

Application of a sufficiently negative potential to a fermenting culture via an electrode has been shown to increase the yield of reduced products, including increasing propionate yield of Propionibacterium freudenreichii (Emde and Schink, 1990), and increasing ethanol yield of Clostridium thermocellum and Saccharomyces cerevisiae (Shin et al., 2002). Cells uptake electrons in an electro-fermentation system by one of two mechanisms: direct and indirect electron transfer. Direct electron transfer relies on direct interaction of outer membrane redox active components (EMCs and/or nanowires) (Gorby et al., 2006; Reguera et al., 2005). This mechanism has been observed for anode-driven systems (Bond and Lovley, 2003), and has been postulated in cathode-driven systems (Gregory et al., 2004; Lovley, 2011; Rabaey and Rozendal, 2010), but we could not find any reported use of this mechanism in fermentation systems at the time of this review.

Figure 9: Standard reduction potentials (ΔE’°) in volts (versus standard hydrogen electrode) of common electron shuttle molecules (Bishop, 1972) and metabolic half-reactions at pH 7 and an ionic strength of 0.10 M (Alberty, 2003).

A-23 Indirect electron transfer occurs when redox active shuttling molecules transport electrons between microorganisms and the electrode via diffusion in the culturing media. The shuttling molecule, or electron mediator, can be secondary metabolites such as phenazine (Rabaey et al., 2005) and flavins (Marsili et al., 2008), hydrogen from hydrolysed water (produced at cathode surface) (Rabaey et al., 2007), or added synthetic molecules. Anthraquinone 2,6-disulfonic acid (Emde and Schink, 1990), cobalt(III) sepulchrate (Emde and Schink, 1990), methyl viologen (Kim and Kim, 1988; Peguin and Soucaille, 1996; Steinbusch et al., 2010), and neutral red (Park et al., 1999; Park and Zeikus, 1999; Shin et al., 2002) have all been successfully used as synthetic electron mediators in cathodic systems (see Figure 9 for standard reduction potentials). From a control viewpoint, use of synthetic mediators appears to be the best method for regulating extracellular electron transfer in fermentation systems due to the ability to select for a desired reduction potential. Since each carrier has a different reduction potential (and stoichiometry), the regulation would be based of the different energy levels at which electrons would be entering the cells. As shown by Emde and Schink (1990), the use of mediators with more negative reduction potentials results in increased yield of reduced products in pure culture fermentation. Cathodes have also been used in mixed-culture fermentation (Dennis et al., 2013). However, no reports were found on the effect of using mediators of different reduction potentials in mixed-culture fermentation at the time of this review. Nevertheless, using mediators with increasingly negative reduction potentials is expected to make reductive metabolic pathways more thermodynamically favorable (Figure 9).

5 MODELLING REGULATION

In order to gain control over fermentation product output, it is necessary to develop a mathematical model that can accurately predict product formation in a variety of fermentation conditions. Mathematical modelling of fermentation began as a process step in the modelling of mixed-culture anaerobic digestion. Mosey (1983) assumed that the + system was fundamentally regulated by NAD :NADH ratio via the partial pressure of H2.

This model set butyrate and propionate production as a thermodynamic response to H2 partial pressure while maximizing ATP production through acetate. Costello et al. (1991) expanded upon the Mosey model by introducing product and pH inhibition terms for each group of bacteria. Keller et al. (1993) then introduced shock loading to account for the increased production of VFAs at high organic loading rates. VFAs were split generically into even-carbon and odd-carbon groups by Ruzicka (1996), and the directed production of one group versus the other was assumed to be controlled by the partial pressure of H2.

A-24 Current mixed-culture fermentation models use a static mixed culture metabolism approach that assumes a stable substrate to product stoichiometry and set enzyme expressions to predict a product spectrum. For instance, the model developed by Rodriguez et al. (2006) and expanded upon by Kleerebezem et al. (2008) uses a combination of balancing redox potential and maximising ATP generation to solve for a set product spectrum at given input conditions of pH, partial pressure of H2, and substrate concentration. The Rodriguez et al. (2006) model was expanded to include redox considerations with FAD and Fd in addition to NAD by Kleerebezem et al. (2008), but remains inadequate in predicting acetate, butyrate, and ethanol levels at low loading rates, as well as predicting lactate, propionate and other reduced product levels under any input conditions.

The Rodriguez et al. (2006) and Kleerebezem et al. (2008) models each related the ratio of reduced to oxidised EMCs to the partial pressure of H2, in the same way that Mosey

(1983) and Ruzicka (1996) assumed direct thermodynamic influence of H2 partial pressure. It was recently shown by de Kok et al. (2013) that this assumption is invalid and that H2 partial pressure has little or no influence on EMC redox state. This finding is supported by the discovery of redox-responsive metabolic controls (Section 3) and suggests that a dynamic stoichiometry approach is needed which accounts for the changes in enzyme expression resulting from various environmental stimuli. One approach to addressing this would be to amend the current thermodynamic models with additional constraints that also include metabolic controls, possibly as dedicated state variables. This would also allow incorporation of (for example) electro-fermentation and possibly allow better manipulation and direct control of the process.

6 CONCLUSIONS

The wide range of fermentation products that arise from catabolic electron balancing present an opportunity to develop a versatile biochemical-from-waste production industry. However, more research is needed to understand how stimuli such as pH, substrate type, OLR, and HRT effect catabolic (especially oxidoreductive) enzyme expression and, by extension, the product spectrum. While many of these factors are well understood in certain pure-culture systems, in general, there is a lack of understanding of the relationship between product formation and, particularly, its relationship to the internal ratio of reduced to oxidized EMCs. The following are especially important: the conditions that cause increased generation of reduced EMCs; and the factors which control whether H2 or organic acids are used as an electron sink. There is broad scope for targeted experiments

A-25 in mixed culture to develop more competent and mechanistic predictive models. This has the potential to substantially expand the field of application for mixed culture fermentation, including in novel processes such as electrofermentation.

Acknowledgements

This project was supported under Australian Research Council’s Discovery Projects funding scheme (project number DP0985000) and Dr. Damien Batstone is the recipient of an ARC Research Fellowship.

A-26 REFERENCES

Abbe K, Takahashi S, Yamada T. 1982. Involvement of Oxygen-Sensitive Pyruvate Formate-Lyase in Mixed-Acid Fermentation by Streptococcus mutans Under Strictly Anaerobic Conditions. J Bacteriol 152:175–182. Agler MT, Werner JJ, Iten LB, Dekker A, Cotta M a, Dien BS, Angenent LT. 2012. Shaping reactor microbiomes to produce the fuel precursor n-butyrate from pretreated cellulosic hydrolysates. Environ. Sci. Technol. 46:10229–38. Alberty RA. 2003. Oxidation-Reduction Reactions. In: . Thermodyn. Biochem. React. Hoboken, NJ: John Wiley & Sons, Inc, pp. 155–170. Angenent LT, Karim K, Al-Dahhan MH, Wrenn B a, Domíguez-Espinosa R. 2004. Production of bioenergy and biochemicals from industrial and agricultural wastewater. Trends Biotechnol. 22:477–85. Bakoyianis V, Koutinas AA, Agelopoulos K, Kanellaki M. 1997. Comparative Study of Kissiris, γ-Alumina, and Calcium Alginate as Supports of Cells for Batch and Continuous Wine-Making at Low Temperatures. J. Agric. Food Chem. 45:4884–4888. Batstone DJ, Keller J, Angelidaki I, Kalyuzhnyi S V, Pavlostathis SG, Rozzi A, Sanders WTM, Siegrist H, Vavilin VA. 2002. Anaerobic Digestion Model No . 1: IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. London: IWA Publishing. Bischof JC, Padanilam J, Holmes WH, Ezzell RM, Lee RC, Tompkins RG, Yarmush ML, Toner M. 1995. Dynamics of cell membrane permeability changes at supraphysiological temperatures. Biophys. J. 68:2608–14. Bishop E. 1972. Indicators. New York: Oxford. Bond DR, Lovley DR. 2003. Electricity Production by Geobacter sulfurreducens Attached to Electrodes. Appl. Environ. Microbiol. 69:1548–1555. Booth IR, Ferguson GP, Miller S, Li C, Gunasekera B, Kinghorn S. 2003. Bacterial production of methylglyoxal: a survival strategy or death by misadventure? Biochem. Soc. Trans. 31:1406–8. Bosma HJ, de Graaf-Hess AC, de Kok A, Veeger C, Visser AJWG, Voordouw G. 1982. Pyruvate Dehydrogenase Complex from Azotobacter vinelandii: Structure, Function, and Inter-Enzyme Catalysis. Ann. N. Y. Acad. Sci. 378:265–286. Brar SK, Dhillon GS, Soccol CR. 2014. Biotransformation of Waste Biomass into High Value Biochemicals. Ed. Satinder Kaur Brar, Gurpreet Singh Dhillon, Carlos Ricardo Soccol. New York, NY: Springer New York. Brekasis D, Paget MSB. 2003. A novel sensor of NADH/NAD+ redox poise in Streptomyces coelicolor A3(2). EMBO J. 22:4856–65. Cameron DC, Altaras NE, Hoffman ML, Shaw a J. 1998. Metabolic engineering of propanediol pathways. Biotechnol. Prog. 14:116–25. Von Canstein H, Ogawa J, Shimizu S, Lloyd JR. 2008. Secretion of flavins by Shewanella species and their role in extracellular electron transfer. Appl. Environ. Microbiol. 74:615–23. Caspi R. 2014. The MetaCyc Database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res. http://metacyc.org/. Chen GC, Jordan F. 1984. Brewers’ Yeast Pyruvate Decarboxylase Produces Acetoin from Acetaldehyde: A Novel Tool To Study the Mechanism of Steps Subsequent to Carbon Dioxide Loss. Biochemistry 23:3576–3582. Costello DJ, Greenfield PF, Lee PL. 1991. Dynamic modelling of a single-stage high-rate anaerobic reactor - I. Model derivation. Wat Res 25:847–858. Crabbendam PM, Neijssel OM, Tempest DW. 1985. Metabolic and energetic aspects of the growth of Clostridium butyricum on glucose in chemostat culture. Arch Microbiol 142:375–382. Crout DHG, Littlechild J, Morrey SM. 1986. Acetoin metabolism: stereochemistry of the acetoin produced by the pyruvate decarboxylase of wheat germ and by the a-acetolactate decarboxylase of Klebsiella aerogenes. J. Chem. Soc. Perkin Trans. 1:105. Damasceno LM, Pla I, Chang H-J, Cohen L, Ritter G, Old LJ, Batt C a. 2004. An optimized fermentation process for high- level production of a single-chain Fv antibody fragment in Pichia pastoris. Protein Expr. Purif. 37:18–26. Decker K, Jungermann K, Thauer RK. 1970. Energy production in anaerobic organisms. Angew Chem Intern. Ed. 9:138– 158. Demuez M, Cournac L, Guerrini O, Soucaille P, Girbal L. 2007. Complete activity profile of Clostridium acetobutylicum [FeFe]-hydrogenase and kinetic parameters for endogenous redox partners. FEMS Microbiol. Lett. 275:113–21. Dennis PG, Harnisch F, Yeoh YK, Tyson GW, Rabaey K. 2013. Dynamics of cathode-associated microbial communities and metabolite profiles in a glycerol-fed bioelectrochemical system. Appl. Environ. Microbiol. 79:4008–14. Van Dijken JP, Scheffers WA. 1986. Redox balances in the matabolism of sugars by yeasts. FEMS Microbiol. Rev. 32:199–224.

A-27 El-Mansi EMT, Bryce CFA, Demain AL, Allman AR. 2006. Fermentation Microbiology and Biotechnology 2nd ed. Boca Raton, Florida, USA: CRC Press. Emde R, Schink B. 1990. Enhanced Propionate Formation by Propionibacterium freudenreichii subsp. freudenreichii in a Three-Electrode Amperometric Culture System. Appl Env. Microbiol 56:2771–6. Eng SC, Fernandes XA, Paskins AR. 1986. Biochemical effects of administering shock loads of sucrose to a laboratory- scale anaerobic (UASB) effluent treatment plant. Wat Res 20:789–794. Ferguson GP, Tötemeyer S, MacLean MJ, Booth IR. 1998. Methylglyoxal production in bacteria: suicide or survival? Arch. Microbiol. 170:209–18. Ferry JG. 1990. Formate dehydrogenase. FEMS Microbiol. Rev. 87:377–382. Fonteles T V, Costa MGM, de Jesus ALT, Rodrigues S. 2012. Optimization of the Fermentation of Cantaloupe Juice by Lactobacillus casei NRRL B-442. Food Bioprocess Technol. 5:2819–2826. Galanakis CM, Kordulis C, Kanellaki M, Koutinas A a, Bekatorou A, Lycourghiotis A. 2012. Effect of pressure and temperature on alcoholic fermentation by Saccharomyces cerevisiae immobilized on γ-alumina pellets. Bioresour. Technol. 114:492–8. Garg SK, Jain A. 1995. Fermentative production of 2,3-butanediol: A review. Bioresour. Technol. 51:103–109. Georgellis D, Kwon O, Lin EC. 2001. Quinones as the redox signal for the arc two-component system of bacteria. Science 292:2314–6. Girbal L, Soucaille P. 1994. Regulation of Clostridium acetobutylicum metabolism as revealed by mixed-substrate steady-state continuous cultures: role of NADH/NAD ratio and ATP pool. J. Bacteriol. 176:6433–8. Gorby Y a, Yanina S, McLean JS, Rosso KM, Moyles D, Dohnalkova A, Beveridge TJ, Chang IS, Kim BH, Kim KS, Culley DE, Reed SB, Romine MF, Saffarini D a, Hill E a, Shi L, Elias D a, Kennedy DW, Pinchuk G, Watanabe K, Ishii S, Logan B, Nealson KH, Fredrickson JK. 2006. Electrically conductive bacterial nanowires produced by Shewanella oneidensis strain MR-1 and other microorganisms. Proc. Natl. Acad. Sci. U. S. A. 103:11358–63. Gregory KB, Bond DR, Lovley DR. 2004. Graphite electrodes as electron donors for anaerobic respiration. Environ. Microbiol. 6:596–604. Gyan S, Shiohira Y, Sato I, Takeuchi M, Sato T. 2006. Regulatory loop between redox sensing of the NADH/NAD(+) ratio by Rex (YdiH) and oxidation of NADH by NADH dehydrogenase Ndh in Bacillus subtilis. J. Bacteriol. 188:7062– 71. Ha S-J, Galazka JM, Kim SR, Choi J-H, Yang X, Seo J-H, Glass NL, Cate JHD, Jin Y-S. 2011. Engineered Saccharomyces cerevisiae capable of simultaneous cellobiose and xylose fermentation. Proc Natl Acad Sci U S A 108:504–509. Hansen RG, Henning U. 1966. Regulation of pyruvate dehydrogenase activity in Escherichia coli K12. Biochem. Biophys. Acta - Enzymol. Biol. Oxid. 122:355–358. Himmi EH, Bories A, Boussaid A, Hassani L. 2000. Propionic acid fermentation of glycerol and glucose by Propionibacterium acidipropionici and Propionibacterium freudenreichii ssp. shermanii. Appl Microbiol Biotechnol 53:435–440. Hiu SF, Zhu C, Yan R, Chen J-S. 1987. Butanol-ethanol dehydrogenase and butanol-ethanol-isopropanol dehydrogenase: different alcohol dehydrogenases in two strains of Clostridium beijerinckii (Clostridium butylicum). Appl Env. Microbiol 53:697–703. Horiuchi J-I, Shimizu T, Tada K, Kanno T, Kobayashi M. 2002. Selective production of organic acids in anaerobic acid reactor by pH control. Bioresour Technol 82:209–213. Huang K-X, Rudolph FB, Bennett GN. 1999. Characterization of Methylglyoxal Synthase from Clostridium acetobutylicum ATCC 824 and Its Use in the Formation of 1,2-Propanediol. Appl. Environ. Microbiol. 65:3244–3247. Infantes D, González del Campo a., Villaseñor J, Fernández FJ. 2011. Influence of pH, temperature and volatile fatty acids on hydrogen production by acidogenic fermentation. Int. J. Hydrogen Energy 36:15595–15601. Inoue K, Leang C, Franks AE, Woodard TL, Nevin KP, Lovley DR. 2011. Specific localization of the c-type cytochrome OmcZ at the anode surface in current-producing biofilms of Geobacter sulfurreducens. Environ. Microbiol. Rep. 3:211–7. Jeffries TW. 2006. Engineering yeasts for xylose metabolism. Curr. Opin. Biotechnol. 17:320–6. Jiang L, Wang J, Liang S, Cai J, Xu Z, Cen P, Yang S, Li S. 2011. Enhanced butyric acid tolerance and bioproduction by Clostridium tyrobutyricum immobilized in a fibrous bed bioreactor. Biotechnol Bioeng 108:31–40. Jones DT, Woods DR. 1986. Acetone-Butanol Fermentation Revisited. Microbiol. Rev. 50:484–524. Jones SW, Paredes CJ, Tracy B, Cheng N, Sillers R, Senger RS, Papoutsakis ET. 2008. The transcriptional program underlying the physiology of clostridial sporulation. Genome Biol 9:R114.

A-28 Jungermann K, Thauer RK, Leimenstoll G, Decker K. 1973. Function of reduced pyridine nucleotide-ferredoxin oxidoreductases in saccharolytic Clostridia. Biochim. Biophys. Acta - Bioenerg. 305:268–280. Junicke H, Kleerebezem R, van Loosdrecht MCM. 2013. Kinetic and thermodynamic approach for the quantification of interspecies hydrogen transfer in non-defined microbial communities. In: . 13th World Congr. Anaerob. Dig. Santiago de Compostela, Spain: IWA Publishing, p. PPA16. Keller J, Romli M, Lee PL, Greenfield PF. 1993. Dynamic model simulation and verification of a two-stage high-rate anaerobic treatment process with recycle. Water Sci. Technol. 28:197–207. Kiley PJ, Beinert H. 2003. The role of Fe-S proteins in sensing and regulation in bacteria. Curr. Opin. Microbiol. 6:181– 185. Kim J, Hetzel M, Boiangiu CD, Buckel W. 2004. Dehydration of (R)-2-hydroxyacyl-CoA to enoyl-CoA in the fermentation of alpha-amino acids by anaerobic bacteria. FEMS Microbiol. Rev. 28:455–68. Kim TS, Kim BH. 1988. Electron flow shift in Clostridium acetobutylicum fermentation by electrochemically introduced reducing equivalent. Biotechnol. Lett. 10:123–128. Kleerebezem R, Rodriguez J, Temudo MF, van Loosdrecht MCM. 2008. Modeling mixed culture fermentations; the role of different electron carriers. Water Sci Technol 57:493–497. Ko J, Kim I, Yoo S, Min B, Kim K, Park C. 2005. Conversion of methylglyoxal to acetol by Escherichia coli aldo-keto reductases. J. Bacteriol. 187:5782–9. De Kok S, Meijer J, van Loosdrecht MCM, Kleerebezem R. 2013. Impact of dissolved hydrogen partial pressure on mixed culture fermentations. Appl. Microbiol. Biotechnol. 97:2617–25. Konings WN, Poolman B, van Veen HW. 1994. Solute transport and energy transduction in bacteria. Antonie Van Leeuwenhoek 65:369–80. Kruger NJ, von Schaewen A. 2003. The oxidative pentose phosphate pathway: structure and organisation. Curr. Opin. Plant Biol. 6:236–246. Leibig M, Liebeke M, Mader D, Lalk M, Peschel A, Götz F. 2011. Pyruvate formate lyase acts as a formate supplier for metabolic processes during anaerobiosis in Staphylococcus aureus. J. Bacteriol. 193:952–62. Lewis VP, Yang S-T. 1992. Propionic acid fermentation by Propionibacterium acidipropionici: effect of growth substrate. Appl Microbiol Biotechnol 37:437–442. Libkind D, Hittinger CT, Valério E, Gonçalves C, Dover J, Johnston M, Gonçalves P, Sampaio JP. 2011. Microbe domestication and the identification of the wild genetic stock of lager-brewing yeast. Proc. Natl. Acad. Sci. U. S. A. 108:14539–44. Lin Y, Tanaka S. 2006. Ethanol fermentation from biomass resources: current state and prospects. Appl. Microbiol. Biotechnol. 69:627–42. Lindmark DG, Paolella P, Wood NP. 1969. The Pyruvate Formate-lyase System of Streptococcus faecalis I. Purification and Properties of the Formate-Pyruvate Exchange Enzyme. J. Biol. Chem. 244:3605–3612. Liyanage H, Kashket S, Young M, Kashket ER. 2001. Clostridium beijerinckii and Clostridium difficile Detoxify Methylglyoxal by a Novel Mechanism Involving Glycerol Dehydrogenase. Appl. Environ. Microbiol. 67:2004–2010. Lovley DR. 2011. Powering microbes with electricity: direct electron transfer from electrodes to microbes. Environ. Microbiol. Rep. 3:27–35. Lu Y, Slater FR, Mohd-Zaki Z, Pratt S, Batstone DJ. 2011. Impact of operating history on mixed culture fermentation microbial ecology and product mixture. Water Sci. Technol. 64:760. Lundblad RL, MacDonald FM. 2010. Physical and Chemical Data. In: . Handb. Biochem. Mol. Biol. 4th ed. New York: CRC Press, pp. 537–938. Luo W, Cai J, Zhu L, Zhu X, Huang L, Xu Z, Cen P. 2012. Toxic effects of acrylic acid on C lostridium propionicum and isolation of acrylic acid-tolerant mutants for production of acrylic acid. Eng. Life Sci. 12:567–573. Madigan M, Martinko J, Stahl D. 2012a. Nutrition, Culture, and Metabolism of Microorganisms. In: . Brock Biol. Microorg. 13th ed. San Francisco: Benjamin Cummings, pp. 113–144. Madigan MT, Martinko J, Stahl D. 2012b. Brock Biology of Microorganisms 13th ed. San Francisco: Benjamin Cummings. Maeda T, Yoshimura T, Shimazu T, Shirai Y, Ogawa HI. 2009. Enhanced production of lactic acid with reducing excess sludge by lactate fermentation. J Hazard Mater 168:656–663. Marsili E, Baron DB, Shikhare ID, Coursolle D, Gralnick J a, Bond DR. 2008. Shewanella secretes flavins that mediate extracellular electron transfer. Proc. Natl. Acad. Sci. U. S. A. 105:3968–73. Moon SK, Wee YJ, Choi GW. 2012. A novel lactic acid bacterium for the production of high purity L-lactic acid, Lactobacillus paracasei subsp paracasei CHB2121. J Biosci Bioeng 114:155–159.

A-29 Mosey FE. 1983. Mathematical modelling of the anaerobic digestion process regulatory mechanisms for the formation of short-chain volatile acids from glucose. Wat Sci Tech 15:209–232. Nakamura CE, Whited GM. 2003. Metabolic engineering for the microbial production of 1,3-propanediol. Curr. Opin. Biotechnol. 14:454–459. Nakano MM. 2002. Induction of ResDE-Dependent Gene Expression in Bacillus subtilis in Response to Nitric Oxide and Nitrosative Stress. J Bacteriol 184:1783–1787. Nelson DL, Cox MM. 2012. Lehninger Principles of Biochemistry 6th ed. New York: W.H. Freeman. Nnyepi MR, Peng Y, Broderick JB. 2007. Inactivation of E. coli pyruvate formate-lyase: role of AdhE and small molecules. Arch. Biochem. Biophys. 459:1–9. Noth J, Krawietz D, Hemschemeier A, Happe T. 2013. Pyruvate:ferredoxin oxidoreductase is coupled to light- independent hydrogen production in Chlamydomonas reinhardtii. J. Biol. Chem. 288:4368–77. Oude Elferink SJWH, Krooneman J, Gottschal JC, Spoelstra SF, Faber F, Driehuis F. 2001. Anaerobic conversion of lactic acid to acetic acid and 1,2-propanediol by Lactobacillus buchneri. Appl. Environ. Microbiol. 67:125–132. Pagliardini J, Hubmann G, Alfenore S, Nevoigt E, Bideaux C, Guillouet SE. 2013. The metabolic costs of improving ethanol yield by reducing glycerol formation capacity under anaerobic conditions in Saccharomyces cerevisiae. Microb. Cell Fact. 12:1–14. Papoutsakis ET, Meyer CL. 1985. Fermentation Equations for Propionic- Acid Bacteria and Production of Assorted Oxychemicals from Various Sugars. Biotechnol Bioeng 27:67–80. Park DH, Laivenieks M, Guettler M V, Jain MK, Zeikus JG. 1999. Microbial utilization of electrically reduced neutral red as the sole electron donor for growth and metabolite production. Appl. Environ. Microbiol. 65:2912–7. Park DH, Zeikus JG. 1999. Utilization of electrically reduced neutral red by Actinobacillus succinogenes: physiological function of neutral red in membrane-driven fumarate reduction and energy conservation. J. Bacteriol. 181:2403–10. Pavia DL. 2007. Introduction to Organic Laboratory Techniques: A Microscale Approach, 4th ed. Belmont, CA, USA: Thomson Brooks/Cole. Peekhaus N, Conway T. 1998. What’s for Dinner?: Entner-Doudoroff Metabolism in Escherichia coli. J. Bact 180:3495– 3502. 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. Biotechnol. Bioeng. 51:342–8. Pei J, Zhou Q, Jing Q, Li L, Dai C, Li H, Wiegel J, Shao W. 2011. The mechanism for regulating ethanol fermentation by redox levels in Thermoanaerobacter ethanolicus. Metab. Eng. 13:186–93. Pervin HM, Dennis PG, Lim HJ, Tyson GW, Batstone DJ, Bond PL. 2013. Drivers of microbial community composition in mesophilic and thermophilic temperature-phased anaerobic digestion pre-treatment reactors. Water Res. 47:7098– 108. Rabaey K, Rodriguez J, Blackall LL, Keller J, Gross P, Batstone D, Verstraete W, Nealson KH. 2007. Microbial ecology meets electrochemistry: electricity-driven and driving communities. ISME J 1:9–18. Rabaey K, Boon N, Höfte M, Verstraete W. 2005. Microbial phenazine production enhances electron transfer in biofuel cells. Environ. Sci. Technol. 39:3401–8. Rabaey K, Rozendal R a. 2010. Microbial electrosynthesis - revisiting the electrical route for microbial production. Nat. Rev. Microbiol. 8:706–16. Ravcheev DA, Li XQ, Latif H, Zengler K, Leyn SA, Korostelev YD, Kazakov AE, Novichkov PS, Osterman AL, Rodionov D a. 2012. Transcriptional Regulation of Central Carbon and Energy Metabolism in Bacteria by Redox-Responsive Repressor Rex. J Bacteriol 194:1145–1157. Reguera G, McCarthy KD, Mehta T, Nicoll JS, Tuominen MT, Lovley DR. 2005. Extracellular electron transfer via microbial nanowires. Nature 435:1098–101. Rehm BH a. 2010. Bacterial polymers: biosynthesis, modifications and applications. Nat. Rev. Microbiol. 8:578–92. Ren N, Wang B, Huang J-C. 1997. Ethanol-type fermentation from carbohydrate in high rate acidogenic reactor. Biotechnol Bioeng 54:428–433. Rodriguez J, Kleerebezem R, Lema JM, van Loosdrecht MC. 2006. Modeling product formation in anaerobic mixed culture fermentations. Biotechnol Bioeng 93:592–606. Rogstam A, Larsson JT, Kjelgaard P, von Wachenfeldt C. 2007. Mechanisms of adaptation to nitrosative stress in Bacillus subtilis. J. Bacteriol. 189:3063–71. Ruzicka M. 1996. An extension of the Mosey model. Wat Res 30:2440–2446. Saint-amans S, Girbal L, Andrade J, Ahrens K, Soucaille P. 2001. Regulation of Carbon and Electron Flow in Clostridium butyricum VPI 3266 Grown on Glucose-Glycerol Mixtures. J Bacteriol 183:1748–1754.

A-30 Satish Kumar R, Kanmani P, Yuvaraj N, Paari K a, Pattukumar V, Arul V. 2013. Traditional Indian fermented foods: a rich source of lactic acid bacteria. Int. J. Food Sci. Nutr. 64:415–428. Saxena RK, Anand P, Saran S, Isar J. 2009. Microbial production of 1,3-propanediol: Recent developments and emerging opportunities. Biotechnol. Adv. 27:895–913. Schroder C, Selig M, Schonheit P. 1994. Glucose fermentation to acetate, CO2 and H2 in the anaerobic hyperthermophilic eubacterium Thermotoga maritima: involvement of the Embden-Meyerhof pathway. Arch Microbiol 161:460–470. Selmer T, Willanzheimer A, Hetzel M. 2002. Propionate CoA-transferase from Clostridium propionicum: Cloning of gene and identification of glutamate 324 at the active site. Eur J Biochem 269:372–380. Shin HS, Zeikus JG, Jain MK. 2002. Electrically enhanced ethanol fermentation by Clostridium thermocellum and Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol. 58:476–81. Singh AK, Athmaram TN, Shrivastava S, Merwyn S, Agarwal GS, Gopalan N. 2013. Fermentation and downstream process for high yield production of Plasmodium falciparum recombinant HRP II protein and its application in diagnosis. J. Ind. Microbiol. Biotechnol. 40:687–95. Skraly F a, Cameron DC. 1998. Purification and characterization of a Bacillus licheniformis phosphatase specific for D- alpha-glycerophosphate. Arch. Biochem. Biophys. 349:27–35. Snoep JL, de Graef MR, de Mattos MJT, Neijssel OM. 1994. Effect of culture conditions on the NADH/NAD ratio and total amounts of NAD(H) in chemostat cultures of Enterococcus faecalis NCTC 775. FEMS Microbiol. Lett. 116:263– 268. Snoep JL, de Mattos MJT, Neijssel OM. 1991. Effect of the energy source on the NADH/NAD ratio and on pyruvate catabolism in anaerobic chemostat cultures of Enterococcus faecalis NCTC 755. FEMS Microbiol. Lett. 81:63–66. Snoep JL, de Mattos MJT, Postma PW, Neijssel OM. 1990. Involvement of pyruvate dehydrogenase in product formation in pyruvate-limited anaerobic chemostat cultures of Enterococcus faecalis NCTC 775. Arch Microbiol 154:50–55. Sreela-or C, Imai T, Plangklang P, Reungsang A. 2011. Optimization of key factors affecting hydrogen production from food waste by anaerobic mixed cultures. Int. J. Hydrogen Energy 36:14120–14133. Steinbusch KJJ, Hamelers HVM, Schaap JD, Kampman C, Buisman CJN. 2010. Bioelectrochemical ethanol production through mediated acetate reduction by mixed cultures. Environ. Sci. Technol. 44:513–7. Survase S a, Jurgens G, van Heiningen A, Granström T. 2011. Continuous production of isopropanol and butanol using Clostridium beijerinckii DSM 6423. Appl. Microbiol. Biotechnol. 91:1305–13. Suwannakham S, Yang ST. 2005. Enhanced propionic acid fermentation by Propionibacterium acidipropionici mutant obtained by adaptation in a fibrous-bed bioreactor. Biotechnol Bioeng 91:325–337. Szentrimay R, Yeh P, Kuwana T. 1977. Evaluation of Mediator-Titrants for the Indirect Coulometric Titration of Biocomponents. In: Sawyer, DT, editor. Electrochem. Stud. Biol. Syst. 38th ed. Washington, DC: American Chemical Society, pp. 143–169. Temudo MF, Muyzer G, Kleerebezem R, van Loosdrecht MC. 2008a. Diversity of microbial communities in open mixed culture fermentations: impact of the pH and carbon source. Appl Microbiol Biotechnol 80:1121–1130. Temudo MF, Poldermans R, Kleerebezem R, van Loosdrecht MC. 2008b. Glycerol fermentation by (open) mixed cultures: a chemostat study. Biotechnol Bioeng 100:1088–1098. Temudo MF, Kleerebezem R, van Loosdrecht M. 2007. Influence of the pH on (open) mixed culture fermentation of glucose: a chemostat study. Biotechnol Bioeng 98:69–79. Tran-Din K, Gottschalk G. 1985. Formation of D(-)-1,2-propanediol and D(-)-lactate from glucose by Clostridium sphenoides under phosphate limitation. Arch Microbiol 142:87–92. Uyeda K, Rabinowitz JC. 1971. Pyruvate-Ferredoxin Oxidoreductase: III. Purification and Properties of the Enzyme. J Biol Chem 246:3111–3119. Vasconcelos I, Girbal L, Soucaille P. 1994. Regulation of Carbon and Electron Flow in Clostridium acetobutylicum Grown in Chemostat Culture at Neutral pH on Mixtures of Glucose and Glycerol. J Bacteriol 176:1443–1450. Voolapalli RK, Stuckey DC. 2001. Hydrogen production in anaerobic reactors during shock loads-influence of formate production and H2 kinetics. Water Res. 35:1831–41. Wang E, Bauer MC, Rogstam A, Linse S, Logan DT, von Wachenfeldt C. 2008. Structure and functional properties of the Bacillus subtilis transcriptional repressor Rex. Mol. Microbiol. 69:466–78. Xiao Z, Xu P. 2007. Acetoin metabolism in bacteria. Crit. Rev. Microbiol. 33:127–40. Xue J, Murrieta CM, Rule DC, Miller KW. 2008. Exogenous or L-rhamnose-derived 1,2-propanediol is metabolized via a pduD-dependent pathway in Listeria innocua. Appl. Environ. Microbiol. 74:7073–9. Yu Y, Tangney M, Aass HC, Mitchell WJ. 2007. Analysis of the mechanism and regulation of lactose transport and metabolism in Clostridium acetobutylicum ATCC 824. Appl. Environ. Microbiol. 73:1842–1850.

A-31 Zeikus JG, Jain MK, Elankovan P. 1999. Biotechnology of succinic acid production and markets for derived industrial products. Appl. Microbiol. Biotechnol. 51:545–552. Zhang W, Werner JJ, Agler MT, Angenent LT. 2014. Substrate type drives variation in reactor microbiomes of anaerobic digesters. Bioresour. Technol. 151:397–401.

A-32