The Pennsylvania State University

The Graduate School

Department of Chemical Engineering

DEVELOPMENT OF BIOLOGICAL PLATFORM FOR THE

AUTOTROPHIC PRODUCTION OF BIOFUELS

A Dissertation in

Chemical Engineering

by

Nymul Khan

 2015 Nymul Khan

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

May 2015

The dissertation of Nymul Khan was reviewed and approved* by the following:

Wayne Curtis Professor, Chemical Engineering Dissertation Advisor Chair of Committee

Esther Gomez Assistant Professor, Chemical Engineering

Manish Kumar Assistant Professor, Chemical Engineering

John Regan Professor, Environmental Engineering

Phillip E. Savage Walter L. Robb Family Endowed Chair Head of the Department of Chemical Engineering

*Signatures are on file in the Graduate School

iii

ABSTRACT

The research described herein is aimed at developing an advanced biofuel platform that has the potential to surpass the natural rate of solar energy capture and CO2 fixation. The underlying concept is to use the electricity from a renewable source, such as wind or solar, to capture CO2 via a biological agent, such as a microbe, into liquid fuels that can be used for the transportation sector. In addition to being renewable, the higher rate of energy capture by photovoltaic cells than natural photosynthesis is expected to facilitate higher rate of liquid fuel production than traditional biofuel processes. The envisioned platform is part of ARPA-E’s (Advanced Research Projects Agency - Energy) Electrofuels initiative which aims at supplementing the country’s petroleum based fuel production with renewable liquid fuels that can integrate easily with the existing refining and distribution infrastructure (http://arpa- e.energy.gov/ProgramsProjects/Electrofuels.aspx). The Electrofuels initiative aimed to develop liquid biofuels that avoid the issues encountered in the current generation of biofuels: (1) the reliance of biomass-derived technologies on the inefficient process of photosynthesis, (2) the relatively energy- and resource-intensive nature of agronomic processes, and (3) the occupation of large areas of arable land for feedstock production. The process proceeds by the capture of solar energy into electrical energy via photovoltaic cells, using the generated electricity to split water into molecular hydrogen (H2) and oxygen

(O2), and feeding these gases, along with carbon dioxide (CO2) emitted from point sources such as a biomass or coal-fired power plant, to a microbial bioprocessing platform. The proposed microbial bioprocessing platform leverages a chemolithoautotrophic microorganism (Rhodobacter capsulatus or

Ralstonia eutropha) naturally able to utilize these gases as growth substrates, and genetically modified to produce a triterpene hydrocarbon fuel molecule (C30+ botryococcenes) native to the alga Botryococcus braunii. In addition to the genetic modification and bioreactor performance studies of these organisms for the production of botryococcene or squalene, the research examined the potential economic feasibility of

iv the proposed platform through the use of bioreactor, microbial energetic models and experimentally measured growth yield and maintenance coefficients.

In order to carry out an economic analysis, a process model was created in Aspen with the bioreactor at the center. This is presented in Chapter 2. The model looked at the effects of growth yield and maintenance coefficients of R. capsulatus and R. eutropha, reactor residence time, gas-liquid mass- transfer coefficients, gas composition and specific fuel productivity on the volumetric productivity and fuel yield on H2. It was found that the organism with the lowest maintenance coefficient performed better under very low growth rates evaluated in the model (based on residence time through the reactor) performed the best. The optimum parameter values were then used to determine the capital and operating costs for a 5000 bbl-fuel/day plant and the final fuel cost based on the Levelized Cost of Electricity

(LCOE). It was found that under the assumptions used in this analysis and crude oil prices, the LCOE required for economic feasibility must be less than 2¢/kWh. While not feasible under current market prices and costs, this work identifies key variables impacting process cost and discusses potential alternative paths toward economic feasibility. This was the best case scenario of the two organisms evaluated, and an optimally suited organism with high growth yield and low maintenance coefficient should obviously improve the economics. This economic constraint will improve with the rise of fossil fuel prices, which should occur if the environmentally detrimental effects of their use are factored into the price, through higher taxation, for example.

A review of the current status of metabolic engineering of chemolithoautotrophs is carried out in order to identify the challenges and likely routes to overcome them. This is presented in Chapter 3 of this dissertation. The initial metabolic engineering and bioreactor studies was carried out using a number of gene-constructs on R. capsulatus and R. eutropha. The gene-constructs consisted of Plac promoter followed by the triterpene synthase genes (SS or BS) and other upstream genes. In R. capsulatus, by genetically supplementing the methylerithrotol phosphate (MEP) pathway and supplementing the growth with glucose, it was found that the triterpene synthase enzymes were substrate-limited i.e. depended on

v the carbon-flux to them. A comparison of the production of triterpenes were done in the different growth modes that R. capsulatus was capable of growing – aerobic heterotrophic, anaerobic photoheterotrophic and aerobic chemoautotrophic. Small-scale testing (<50 ml) under typical (un-supplemented) growth conditions showed that the per-cell triterpene production levels were surprisingly similar in all the different growth modes (around 5 mg/gDW). However, the results were much improved when tested in controlled fed-batch bioreactors, capable of reaching significantly higher cell densities. In the heterotrophic case, production was found to increase up to 40 mg/L (~11 mg/gDW), unfortunately inhibited by some sort of toxic effect at OD660 around 12. Autotrophic growth on H2, O2 and CO2, on the other hand, showed no such effect and growth occurred well up to an OD660 of 17 (corresponding to about

7 gDW/L), limited only by the mass-transfer of the gases and triterpene productivity increased continuously to greater than 100 mg/L (16 mg/gDW) in the batch mode. Continuous autotrophic operation further increased the specific titer to 23 mg/gDW, reaching a steady state. The specific productivity was found to be around 0.5 mg/gDW-hr. This demonstrated that autotrophic productivity could likely be improved much further by increasing the available mass-transfer of the reactor. These efforts are presented in Chapter 4 of this dissertation.

vi

TABLE OF CONTENTS

List of figures ...... ix

List of tables ...... xv

List of symbols and abbreviations ...... xvi

Acknowledgements ...... xvii

Chapter 1 Background and outline...... 1

I. Motivation ...... 1 II. My research ...... 2 III. Rationale for Selection of the Microbial Bioprocessing Platform ...... 6 IV. Metabolic engineering strategy ...... 8 V. Dissertation outline ...... 10

Chapter 2 A process economic assessment of hydrocarbon biofuels production using chemoautotrophic organisms ...... 11

I. Preface ...... 11 II. Specific contributions ...... 11 III. Introduction ...... 12 IV. Model development ...... 13 2.IV.I. Proposed electrofuels production process ...... 13 2.IV.II. Bioreactor modeling ...... 14 V. Results and discussion ...... 22 2.V.I. Sensitivity analysis: effect of reactor residence time (τ) ...... 22 2.V.II. Sensitivity analysis: effect of specific productivity (Rfuel) ...... 24 2.V.III. Sensitivity analysis: effect of gas-liquid mass transfer coefficient (kLa)..... 27 2.V.IV. Process economic analysis of capital and operating costs...... 29 2.V.V. Alternative perspective ...... 33 2.V.VI. Targets based on the analysis ...... 34 VI. Conclusions ...... 36

Chapter 3 Literature review: Metabolic engineering in chemolithoautotrophic hosts for the production of fuels and chemicals ...... 37

I. Preface ...... 37 II. Specific contributions ...... 38 III. Introduction ...... 38 IV. Background ...... 41

vii

3.IV.I. Carbon Fixation in Chemolithoautotrophs...... 43 3.IV.II. Hydrogen Utilization by Chemolithoautotrophs ...... 47 3.IV.III. Microbial Electrosynthesis via Direct Electron Feeding ...... 49 V. Metabolic Engineering of Natural Chemolithoautotrophic Organisms ...... 50 3.V.I. Status and Constraints for Metabolic Engineering in Chemolithoautotrophs 50 3.V.II. Metabolic Engineering in Aerobic Chemolithoautotrophs ...... 53 3.V.III. Metabolic Engineering in Acetogens ...... 55 3.V.IV. Metabolic Engineering in Other Chemolithotrophs ...... 58 VI. Using Metabolic Engineering to Improve / Creating New Chemolithoautotrophs ... 59 3.VI.I. Engineering Carbon Capture ...... 59 3.VI.II. Engineering improved H2-utilization ...... 62 3.VI.III. Metabolic Engineering of Alternative Electron Delivery Pathways ...... 64 VII. Metabolic Engineering Toolbox for Existing Autotrophs ...... 65 3.VII.I. Genetic Transformation ...... 65 3.VII.II. Plasmids, Promoters and Vectors of Particular Utility for Chemolithoautotrophs ...... 68 VIII. Status of productivities and scale-up considerations for metabolic engineering .... 71 IX. Conclusion ...... 76

Chapter 4 Triterpene hydrocarbon production engineered into a metabolically versatile host – R. capsulatus ...... 78

I. Preface ...... 78 II. Specific contributions ...... 78 III. Introduction ...... 79 IV. Results ...... 83 4.IV.I. Achieving Substrate-Limited Triterpene Biosynthetic Constructs ...... 83 4.IV.II. Productivity Enhancement through Bioreactor Operational Strategies ...... 88 V. Discussion ...... 93

Chapter 5 Ongoing and future work and conclusion ...... 96

I. Triterpene production in R. eutropha...... 96 5.I.I. Pathway enhancements ...... 100 II. Future work ...... 103 III. Conclusion ...... 107

Appendix A Bioreactor material balance equations for section process economic analysis (section 2.III.II) ...... 108

Appendix B A comparison of experimental measurements and thermodynamic predictions of true yield () ...... 111

Calculation of true yield by TEEM ...... 112 Modifications applied to the original TEEM to incorporate autotrophic growth, reversed electron transport, and BPF synthesis for R. eutropha and R. capsulatus ...... 115 Comparison of original TEEM and Modified Electron Balance methods of predicting ...... 118

viii

Appendix C Measurement of growth yield and maintenance coefficient ...... 121

Growth yield and maintenance coefficient ...... 121 Theory ...... 122 Experimental procedure and results ...... 124

Appendix D Materials and methods...... 129

Bacterial Strains and Growth Conditions ...... 129 Small Scale Phototrophic and Autotrophic Growth Conditions for R. capsulatus Strains130 Heterotrophic Bioreactor Growth Condition ...... 131 Autotrophic Bioreactor Growth Conditions ...... 132 Cloning Procedures ...... 132 Intergeneric Mating Between E. coli S17-1 and R. capsulatus SB1003 ...... 136 Hydrocarbon Analysis ...... 137

Appendix E Background calculations for Figure 4-5...... 139

IV. Aerobic heterotrophic growth on glucose ...... 139 V. Anaerobic photoheterotrophic growth on malate ...... 140 VI. Autotrophic growth on H2 ...... 141 VII. Carbon balance ...... 141 Aerobic heterotrophic growth on glucose ...... 141 Anaerobic photoheterotrophic growth on malate ...... 142 Autotrophic growth on H2 ...... 142

Appendix F Other miscellaneous work ...... 143

VIII. Small scale multiplexed autotrophic screening device...... 143 IX. 100-L plastic bag trickle-bed reactor...... 145 X. Secretion of hydrocarbon by R. capsulatus ...... 146

References ...... 148

ix

List of figures

Figure 1-1. System load with and without large (16 GW) PV system on two spring days (reproduced from Denholm & Margolis 2006)...... 4

Figure 1-2. Conceptual depiction of an Electrofuels process. Energy from the sun is captured in the form of renewable electricity (depicted as solar photovoltaics, PV) and is used to split water. The resulting O2 and H2 is combined with CO2 and fed into a bioreactor where R. capsulatus consumes these gases in stoichiometric requirement. R. capsulatus is genetically engineered to produce a C30+ triterpene hydrocarbon which can be recovered as an extracellular phase-separated oil...... 5

Figure 1-3. Metabolic engineering strategy for the production of tritepenes in R. capsulatus. 9

Figure 2-1. Process flow diagram of the envisioned Electrofuels process. CO2, H2, O2, fresh nutrient and recycled cells enter the reactor and cells+oil leave the reactor. The hydrophobic top oil layer is separated in oil-water separator and pass through fuel filtration and processed to final product. The remaining cells+water+residual oil from oil-water separator enters a clarifier where the denser cell layer is split into two fractions, one recycled back to the reactor and one sent to sludge processing. The top, less dense overflow from this clarifier enters a second clarifier where cell debris are separated from the water, which is processed through waste-water processing to be recycled to the reactor with fresh nutrients and makeup water...... 14

Figure 2-2. Framework for capturing microbial growth energetics, cellular maintenance and fuel production. (A) Incorporating the experimentally measured biological yield and maintenance coefficients into model reaction/stoichiometry equations suitable for reactor design. (B) Concise version of metabolic pathway for calculating the minimum required Gibbs’ free energy for fuel synthesis in a bacteria...... 17

Figure 2-3. Variation of volumetric productivity, (A), fuel yield on H2, /2 (B) and cell density, (C) as a function of residence time () through the reactor for R. capsulatus (●) and R. eutropha (o). Simulation parameters: true growth yield (2) of R. capsulatus = 2.66, R. eutropha = 7.68 gDW/mol-H2, maintenance coefficients (2) of R. capsulatus = -3 -3 2.01x10 , R. eutropha = 6.8x10 mol-H2/gDW.hr, kLa = 330/hr, specific productivity = 0.5 g fuel/g biomass.hr, cell recycle efficiency = 95%...... 24

Figure 2-4. (A) Variation of electricity requirement for H2 generation (●) and volumetric productivity (o) with specific fuel productivity representing an increasing fraction of energy being converted into fuel instead of biomass (Simulation parameters: residence time = 15 hr; other parameters same as in Figure 2-3). (B) Variation in reactor volume as a function of kLa showing the reduction in culture process volume that is enabled as the mass transfer rate is improved ( ). The range of electricity requirements that would be required to achieve the specified kLa in different bioreactor configuration is shown as the shaded area between the dashed lines. The high and the low value of the electricity requirement at each kLa corresponds to the highest and the lowest P/V able to produce the specified kLa; the range of P/V for a variety of bioreactor configurations are presented in Figure 2-5...... 26

x

Figure 2-5. Gas-liquid mass transfer coefficient (kLa) vs. power/volume (P/V) for various reactor types. Surface aerator 1 = wastewater treatment plant demonstration run, Chattanooga, TN (Cosby & Gay, 2003); Surface aerator 2 = wastewater treatment plant demonstration run, Yuba City, CA (Lewis & Gay, 2003); Stirred 1 = stirred tank reactors correlation for electrolytes, (Van’t Riet, 1979), Vs = 0.005 m/s; Stirred 2 = stirred tank reactor correlation for non-electrolytes (Van’t Riet, 1979), Vs = 0.005 m/s; Stirred 3 = stirred tank reactors correlation for electrolytes (Linek et al., 1987), Vs = 0.005 m/s; TBR 1 = trickle-bed reactor correlation set 1 (Reactor parameters: Roininen et al., 2009, kLa correlation: Goto & Smith, 1975, pressure-drop correlation: Larachi et al., 1991); TBR 2 = trickle-bed reactor correlation set 1 (Reactor parameters: Roininen et al., 2009, kLa correlation and pressure- drop correlation: Reiss, 1967); Microbubble – various = experimental data from various investigators (Bredwell & Worden, 1998; Hensirisak et al., 2002). This compilation of literature values focuses on large-scale commercial units (except for microbubble technology)...... 28

Figure 2-6. (A) Capital costs of various process components for a 5000 bbl/day fuel production plant. Results are grouped into high (black bards) and low (dashed bars) range capital costs. (B) A breakdown of process costs (the dominant cost of electricity generation is not presented). Top (grey) group represents the major operating costs which make up 36% of total; middle (hashed) group represents the major non-photovoltaic capital costs which make up 46% of total; bottom (black) group estimates the major non-PV fixed costs corresponding to 17% of the total. (C) A sensitivity analysis of the final fuel cost based on LCOE of various generation methods reported in two different sources: REN21, 2013 (black); IRENA, 2012 (dashed). CSP = Concentrate Solar Power, PT = Parabolic Trough, ST = Solar Tower. The horizontal line indicates year 2020 crude oil price estimate (Gruenspecht, 2012). Off-grid hydropower values were not available in IRENA, 2012...... 31

Figure 3-1. Analogy between direct photosynthetic growth and indirect photosynthetic growth based on the feeding of electrons as reducing power to reduce CO2...... 39

Figure 3-2. An examination of the CO2 reduction reactions from the perspective of the location of the delivery of the reducing power to either the reactor, or to the inside of the cells as a major determinant of the application of these for biotechnological purposes. Light cannot penetrate dense cultures (Beer-Lambert Law), where gasses face limitations of solubility in the liquid phase (Henry’s Law). Electrons are particularly challenging to deliver (Ohm’s Law) either to a reactor electrode or into the cell although this can be facilitated by the reduction of a ‘carrier’...... 40

Figure 3-3. A summary of autotrophic growth modes with the common theme of reducing CO2 to biomass (carbohydrate) using various forms of reducing power. These different autotrophic growth modes are often associated with other classification names that are listed. These common name classifications are useful, but also lead to problems typical of generalizations.42

Figure 3-4. Schematic illustrating the diversity of the cbb operons including paralogs of the various accessory proteins in addition to the large and small subunit (cbbL/S). Form I Rubisco is denoted by cbbLS and Form II by cbbM, both of which are found in R. capsulatus and R. sphaeroides (Panel A&B). R. eutropha also contains two CBB operons, one on the chromosome (Panel C), and one on the megaplasmid pHG1 with a defective transcriptional activator (cbbR*, Panel D). The iron-oxidizing bacteria Acidithiobacillus ferooxidans

xi

contain four different CBB operons with two encoding Form I Rubisco subunits. Other genes are: cbbF=SBPase/FBPase, cbbP=PRK, cbbT=TKT; cbbA=FBA/SBA, cbbE=PPE, cbbG=GAPDH, cbbK=PGK, cbbZ=phosphoglycolate phosphatase), cbbBXYQ=unknown. The presence of these functional operons in plasmids provided the historical basis of elucidating autotrophic genes including CO2 fixation and hydrogen use, now adapted to metabolic engineering of these phenotypes in chemolithoautotrophs...... 44

Figure 3-5. The various carbon fixing pathways...... 46

Figure 3-6. Schematic representation of carbon and energy metabolism, regulation and gene expression in a typical aerobic chemolithoautotroph...... 48

Figure 3-7. (A) Maximum reported productivities in grams of carbon fixed in products per liter (gC-fixed/L) in autotrophic systems. Native pathways: PHB by R. eutropha (I), acetate by M. thermoacetica (II), ethanol by C. ljungdahlii (III); engineered pathways: botryococcene by R. capsulatus (IV), methyl ketone by R. eutropha (V), butyrate by C. ljungdahlii (VI). All are timecourses adapted from literature reports presented in Table 3 and reflect growth on CO2 only (not chemoautotrophic hosts utilizing organic carbon source). (B) Maximum theoretical yield on H2 for the native molecules...... 75

Figure 4-1. R. capsulatus is a metabolically diverse organism that can utilize and interconvert the energy provided from the sun in a variety of different ways. Metabolism within the dashed line representing Rhodobacter cell are those reported in this study: aerobic chemoautrophic growth consumes H2 and O2 while fixing carbon, with these gases being provided by photovoltaic-driven electrolysis in this scenario (or other natural sources). Aerobic heterotrophic growth is the typical consumption of sugars and associated aerobic respiration, anaeobic photoheterotrophic growth consumes energy poor organic acids with photosystem- II mediated ATP generation using light energy. Depiction of the evolution (green arrows) or uptake (red arrows) of metabolic gases emphasizes the bioreactor requirements for gas- exchange as a major constraint for process design and operation. The lower part of the figure depicts the metabolic engineering strategy within the Rhodobacter cell to achieve the production of the high energy terpenes botryococcene or squalene. The operon includes the enzyme that commits carbon flux to the MEP pathway: (1-Deoxy-D-xylulose 5-phosphate synthase, dxs) and the isopentenyl diphosphate (IPP) isomerase (idi) to enhance the ratio of IPP to dimethylallyl pyrophosphate (DMAPP). This DNA construct also includes the gene for farnesyl diphosphate synthase (fps) which utilizes one DMAPP and two IPP to produce C15 farnesyl pyrophosphate, the substrate for the terpene synthases (botryococcene synthase, BS; squalene synthase, SS)...... 80

Figure 4-2. Engineering of triterpene synthases with the native or engineered MEP pathways of E. coli DH5α. Constructs were engineered as a single polycistron under the control of the PLac promoter with triterpene synthase only (BS, botryococcene synthase; or SS, squalene synthase); in combination with a prenyltransferase (fps); or with one plasmid-borne copy of the engineered MEP pathway (dxs-idi-fps) or two plasmid-borne copies. Five biological replicates of each strain were grown aerobically for 24 hours in 2xYT-1% glycerol and then extracted with 1:1 acetone and hexane to determine triterpene content...... 84

Figure 4-3. R. capsulatus triterpene metabolic engineering. (a) Constructs containing botryococcene synthase (BS, upper panel) or squalene synthase (SS, lower panel) with

xii

increasing enhancements in metabolic flux by including the avian farnesyl pyrophospate synthase (fps), and putative rate-limiting enzymes in the MEP pathway. (b) Accumulation of triterpenes in R. capsulatus grown in standard growth medium (Std. medium) and that supplemented with an additional carbohydrate source, 80 mM glucose (+Glu suppl.) for the adjacent constructs: botryococcene for the upper panel, and squalene for the lower panel. The level of enhancement beyond standard growth medium alone is indicated by the extended bars (open bars). Experimental values were determined from five replicate cultures; error bars depict standard error of the mean...... 85

Figure 4-4. Shake flask autotrophic time-course of R. capsulatus harboring pBBR:Plac:BS-dxs-idi- fps and pBBR:Plac:BS only (Top panel) and pBBR:Plac:SS-dxs-idi-fps and pBBR:Plac:SS only (Bottom panel)...... 86

Figure 4-5. Exploring alternative trophisms for hydrocarbon production in Rhodobacter. (a) Triterpene specific productivity of R. capsulatus with pBBR:BS-dxs-idi-fps (blue-left panel), and pBBR: SS-dxs-idi-fps (red-right panel). (b) Reaction for hydrocarbon (HC) production for given tropism. (c) Trophims are presented quantitatively in terms of thermodynamic energy required to assimilate a mole of carbon into pyruvate as well as the energy required to produce an ATP normalized to a molar carbon basis. The carbon balance is shown to illustrate the relative magnitude of carbon produced by aerobic and photo-heterotrophic growth as compared to autotrophic CO2 assimilation...... 88

Figure 4-6. Heterotrophic bioreactor growth of R. capsulatus pBBR:Plac:BS-dxs-idi-fps. R. capsulatus genetically engineered with the above expression vector was grown in a 5 L BioFlo with glucose provided as the sole initial carbon and energy source, followed by fed- batch supplementation based on feedback control of dissolved oxygen (DO). Ammonia was also added in a fed-batch manner for pH control. (a) Growth monitored as OD660 (black circles) and botryococcene accumulation (blue squares) respectively. (b) The DO (red squares) and cumulative glucose supplementation (black line), where glucose fed batch was initiated at ~40 hr, and the approximate times for incremental increases in O2 supplementation (light blue arrows) are noted. Cultures stopped growing and accumulating hydrocarbon at about 70 hr; respiration slowed considerably in stationary phase as indicated by the sharp rise in DO which could not be recovered with additional glucose feed...... 89

Figure 4-7. Growth tests with heterotrophic bioreactor culture and culture supernatant. (a and b) Replicate samples from the bioreactor inoculated into RCVB (-malate) minimal media supplemented with 10 g/L glucose. (c) Sample from the bioreactor inoculated into YCC complex medium supplemented with 10 g/L glucose. (d) Fresh culture of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into bioreactor supernatant supplemented with concentrated YCC medium nutrients (to make the final concentration similar to YCC). (e and f) Replicate fresh cultures of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into bioreactor supernatant supplemented with concentrated RCVB (-malate) medium nutrients (to make the final concentration similar to RCVB). (g) Fresh culture of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into RCVB (-mal) medium supplemented with 27 g/L glucose. (h) Fresh culture of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into RCVB (-mal) medium supplemented with 10 g/L glucose...... 90

Figure 4-8. Performance of R. capsulatus pBBR:PLac:BS-dxs-idi-fps in an autotrophic bioreactor. R. capsulatus expressing this plasmid was grown with gas phase feeding of H2, O2 and CO2

xiii

(for growth) and liquid phase feeding of ammonia (for pH control) in batch operation for the first 110 hours and under continuous operation after that. Continuous flow was established with 12.5 g/L flow of the CA medium, which corresponds to 10% of µmax of R. capsulatus. (a) The inlet gas composition profile. During the batch growth mode, the compositions were varied to accommodate the growth demand, based on outlet gas composition measurement. Gas consumption did not vary greatly during continuous flow and the inlet composition was essentially kept constant. (b) The OD660 and botryococcene profile during batch and continuous operation. Cells grew exponentially to OD 5 and linearly after that reaching a maximum of 17. Botryoccocene accumulation increased proportionally as the cells and achieved a maximum volumetric accumulation of 110 mg L-1 botryococcene at the end of batch growth. Under continuous operation, the cells reached a steady state within about 130 hrs of starting flow, reaching an OD of ~7 and maintaining a relatively constant level 60 mg L-1 botryococcene. (c) The specific botryococcene productivity profiles. The specific botryococcene levels increased steadily during batch growth to about 17 mg/gDW. Batch specific productivities were around 0.5 mg/gDW-hr. Specific botryococcene continued to increase during the continuous flow and reached 23 mg/gDW. However, specific productivity decreased somewhat and reached a steady level of about 0.3 mg/gDW-hr. 92

Figure 5-1. Triterpene production by R. eutropha transformed with pRK:Plac:SS-fps. The three different colored bars indicate three different clones...... 97

Figure 5-2. Heterotrophic reactor growth of R. eutropha pRK:Plac:SS-fps. (A) Defined medium fed-batch with fructose. (B) LB medium fed-batch with fructose...... 98

Figure 5-3. Comparison between two backgrounds of R. eutropha: PHB- and wild-type in terms of triterpene production, transformed with pRK:Plac-SS-FPS. The cultures are grown in LB medium supplemented with indicated levels of fructose...... 99

Figure 5-4. Bioreactor growth of R. eutropha wild-type transformed with pRK:Plac-SS-FPS. 100

Figure 5-5. Triterpene production by R. eutropha PHB- pRK:Plac:SS-fps + pBBR:Plac:MevT- MBIS (dual plasmid)...... 103

Figure 5-6. Triterpene production by R. eutropha H16 wild-type transformed with pBBR:Plac:BS- dxs-idi-fps...... 101

Figure 5-7. Desired mega-plasmid containing all the MVA pathway genes, the triterpene synthase and fps...... 104

Figure A1. Bioreactor flow streams and mass-balance envelops. 1. Cell separator and recycle, 2. bioreactor and cell separator, 3. bioreactor only...... 108

Figure C1: Experimental setup for the measurement of growth yield and maintenance coefficient.125

Figure C2. Media reservoir weight throughout the experiment. The first derivative of y gives the instantaneous media flow rate, while the second derivative gives the deceleration rate. Media flow rates varies from 10.03 g/hr at the initial steady state to 6.92 g/hr at the final steady state, with a deceleration rate of -0.0974 g/hr2...... 127

xiv

Figure C3. Change of culture density with media flow rate. At the initial steady state the culture density remains around 1 g/L, increasing with decrease of flow rate to about 1.4 g/L during the final steady state...... 127

Figure C4. Plot of µ/Y vs. µ for two D-stat experiments covering two different ranges of dilution rates. Inverse of the slope gives the growth yield and the intercept gives the maintenance coefficients...... 128

Figure F1. (Top) A multiplexed autotrophic growth device design for continuous monitoring of growth using LED/phododiode assemblies at the base of the culture. Gas inlet to multiplexed cultures is achieved with a very small sapphire orifice manifold. (Bottom) Examples of online optical density collected from the screening device while growing different strains of R. capsulatus...... 144

Figure F2. A 100-L low cost plastic bag trickle bed that has been assembled to explore regions of high mass transfer efficacy operation for autotrophic growth...... 146

Figure F3. Hydrocarbon secretion in R. capsulatus autotrophic bioreactor cultures. Triterpene levels in cells, media and total culture during late lag and stationary phase of two independent batch cultures...... 147

xv

List of tables

Table 2-1. Parameter values used in final scaled-up design. 29

Table 3-1. Catalytic properties of various hydrogenases (compiled from BRENDA). 65

Table 3-2. Comparison of productivities of different compounds in various autotrophic and non-

autotrophic hosts. 75

Table D1. Strains and plasmids used in this study. 130

Table D2. Nucleotide sequences of the wildtype and R. capsulatus codon-optimized genes used. 133

xvi

List of symbols and abbreviations a, b, c, d stoichiometric coefficients in cell growth equation bbl US fluid barrel -1 concentration of fuel in the bioreactor, mol·L -1 liquid phase concentration of i-th gas component, mol·L 2 -1 diffusion coefficient of i-th gas component, m ·s -1 -1 rate of consumption of i-th gas component, mol·L ·h rate of substrate utilization for growth, fuel synthesis, maintenance requirements and total -1 -1 respectively, mol·L ·h EROI Energy Return on Investment F rate of liquid feed into reactor, L·h-1 gDW grams dry weight -1 -1 gas transfer rate of i-th gas component, mol·L ·h -1 Henry’s law coefficient of i-th gas component, atm·L·mol IGCC Integrated Gasification Combined Cycle -1 gas-liquid mass transfer coefficient of the i-th component, h LCOE Levelized Cost of Electricity -1 -1 maintenance coefficient of cells on H2 and O2 respectively mol·gDW ·h , NGCC Natural Gas Combined Cycle PC Pulverized Coal -1 -1 volumetric fuel productivity, g-fuel·L ·h total pressure, atm -1 -1 specific fuel productivity, g-fuel·gDW ·h -1 -1 rate of growth of cells, gDW·L ·h V reactor volume, L X cell density, gDW·L-1 -1 yield of fuel on H2, g-fuel· (mol-H2) / gas phase mole fraction of i-th component -1 true growth yield of cells on H2, gDW·mol ε cell recycle efficiency -1 µ, µmax specific growth rate and maximum specific growth rate of cells, h τ residence time through reactor, V/F, h-1

xvii

Acknowledgements

I would like to thank Prof. Wayne Curtis for his excellent support and mentoring throughout my

PhD. Prof. Joseph Chappell at University of Kentucky provided valuable insights and direction for the project. My colleagues Dr. S. Eric Nybo at University of Kentucky, Dr. Alex Rajangam and Ryan

Johnson at Penn State University were directly involved in the project and provided great assistance in major aspects of the project. My PhD cohorts, classmates and friends Sergio Florez and Dr. Trevor

Zuroff, provided both intellectual and material help. I am immensely grateful to the numerous excellent undergraduate students who voluntarily put their time and effort for the project - Stephanie Tran, Bill

Muzika, Andrew Barmasse, John Myers, Erik Wolcott and Justin Yoo. Microbial energetic model modifications developed by Amalie Tuerk (Masters from Curtis lab, currently doing her PhD at

University of Delaware) were used in the process economic analysis. She also provided major inputs and insights to help write the manuscript related to this. Ben Woolston (honors student from Curtis lab, currently doing his PhD at Massachusetts Institute of Technology) provided valuable inputs for the writing of the literature review on metabolic engineering. Dr. Steven Singer at Lawrence Berkeley

National Lab kindly provided two strains of R. eutropha. Financial assistance for this project was provided by ARPA-E electrofuels award DE-AR0000092.

Chapter 1

Background and outline

I. Motivation

Energy crisis and environmental concerns for CO2 emission have been continuing challenges for mankind in the 21st century. Fossil fuels have provided a readily accessible and energy-dense fuel source to pave the way for science and technology advancements. But their supply is limited, and their use contribute towards the increase greenhouse gas emissions, adversely affecting global climate. Thus sustainable and renewable sources of biofuels are needed.

Global carbon cycle is maintained by photosynthetic plants and microorganisms sequestering CO2 into biomass using Sun’s energy, providing sources of food and energy for life on earth. In the process of their consumption by humans and animals, the CO2 is released back into the atmosphere. In the last hundred years or so, the exponential growth of human civilization and technologies, and thus the use of fossil fuels, have significantly outpaced natural rate of carbon fixation. Maintenance of economic growth has relied on using millions of years of natural accumulation in fossil fuel reserves within a very short period of time. Further implication of this is the increase of atmospheric CO2 by greater than 10%, which is believed to be causing global adversities (Melillo, Jerry M., Terese (T.C.) Richmond, and Gary W. Yohe 2014). Estimates indicate that the current rate of consumption may be sustained for a few more centuries (IEA

2013), but more catastrophic consequences of rising CO2 levels are prompting research towards reduced greenhouse gas (GHG) emission and carbon capture and storage (CCS) technologies.

These technologies, however, offset the real problem of renewable use of carbon. In addition,

2 depleting reserves and requirement of advanced technologies for unlocking more difficult underground resources mean that prices of fossil fuels will continue to rise in the long term. Only recently have these problems attracted serious attention and concerns from general public and policy-makers and as such, numerous approaches are being funded to develop transformative technologies to address these problems.

Notable efforts in this direction include bioethanol and biodiesel, which are first generation biofuels, i.e. produced directly from food crops (for example corn in US, sugarcane in

Brazil etc.). Corn-ethanol, currently the most widely produced biofuel in the United States, still made up less than 5% of transportation fuels in 2011 (www.eia.gov). Furthermore, there is ongoing controversy related to corn-ethanol contributing to higher food prices. Second generation, non-food biofuels, such as cellulosic ethanol are only recently starting to be commercialized but are limited by natural biomass supply. Third generation algae-biofuels are under active research and development. Even with increasing degree of productivity and sustainability, these technologies still rely on the natural rate of photosynthesis and carbon capture, and this needs to be surpassed if renewable liquid fuels are to be reasonably inexpensive.

II. My research

My research aimed at developing a more advanced biofuel platform that could have the potential to surpass the natural rate of solar energy capture and CO2 fixation. The underlying concept is to use the electricity from a renewable source, such as wind or solar, to capture CO2 via a biological agent, such as a microbe, into liquid fuels that can be used for the transportation sector. The envisioned platform is part of ARPA-E’s (Advanced Research Projects Agency -

Energy) Electrofuels initiative which aims at supplementing the country’s petroleum based fuel

3 production with renewable liquid fuels that can integrate easily with the existing refining and distribution infrastructure (http://arpa-e.energy.gov/ProgramsProjects/Electrofuels.aspx).

In addition to being renewable, the higher rate of energy capture by photovoltaic cells than natural photosynthesis is expected to facilitate higher rate of liquid fuel production than traditional biofuel processes. A conservative estimate of efficiency of today’s photovoltaic cells is around 10% (Peng et al. 2013). Using a maximum biomass yield of 7.68 gDW/mol-H2 for R. eutropha (measured in this work), biomass heat content of 5.411 kcal/gAFDW (Ho & Payne

1979) and electrolyzer energy requirement of 53.4 kWh/kg-H2 (Ivy 2004), we can estimate that the overall efficiency of a system converting CO2 into biomass using H2 should be around 4.3% in terms of energy captured in biomass. Improved efficiency of photovoltaic or other solar energy capture technologies (Bartels et al. 2010) in future should improve this efficiency further.

Compared to this, the best case of sugarcane production (based on a yield of 75 tons/hectare/yr) has an efficiency of 0.38% in terms of total energy captured in biomass (Rosa 2005). Even with a theoretical maximum yield of 280 tons/hectare/yr, the efficiency could only get to 1.4%.

Due to the inherent fluctuations of renewable energy sources, the future integration of renewables into the power grid remains a challenge. Because of operational issues such as non- coincident peaks, non-dispatchability and the stability of the power supplied due to their inherent fluctuating nature (Figure 1-1), there are very legitimate concerns about reliably operating an electrical grid that derives a large fraction of its power renewable sources. Also, solar photovoltaics produce low-voltage DC electricity, which incurs substantial loss in efficiency when converting and stepping up to high voltage AC and subsequently stepping down to low voltage DC for most of its use. Much of the renewable electricity generation is also quite distributed and there is significant cost to create transmission and distribution infrastructure for this electricity.

4

Figure 1-1. System load with and without large (16 GW) PV system on two spring days (reproduced from Denholm & Margolis 2006).

The transportation sector is the largest consumer of liquid fossil fuels (13.44 million bbl/day in 2011, www.eia.gov/forecasts/aeo/MT_liquidfuels.cfm) that cannot be readily replaced by electricity. It is also the second largest source of CO2 emission (1.9 billion metric tons in 2008, http://www.eia.gov/oiaf/1605/ggrpt/carbon.html) after power plants. Thus, storing the intermittent and variable energy of renewable electricity into the chemical bonds of a fuel molecule may be a viable alternative to other forms of storage technologies. Also, when comparing liquid fuel products to other energy storage forms, it is apparent that a liquid hydrocarbon fuel is a much more efficient energy carrier by weight and volume than thermal, mechanical, battery and hydrogen energy storage.

The Electrofuels initiative aimed to develop liquid biofuels that avoid the issues encountered in the current generation of biofuels: (1) the reliance of biomass-derived technologies on the inefficient process of photosynthesis, (2) the relatively energy- and resource- intensive nature of agronomic processes, and (3) the occupation of large areas of arable land for feedstock production. To address these issues, the Electrofuels initiative funded research efforts that sought to develop biological processes that would convert distributed, off-grid, renewable electricity into alternative liquid fuels. The logic of this approach rests in its ability to provide a

5 reliable energy source for the transportation sector by storing transiently-available electrical energy in a chemical bond (

Figure 1-2). In addition, the Electrofuels approach is synergistic with advances in photovoltaic cells.

SUN Hydrocarbons Metabolic CO2 Engineering Bioreactor CO Engineering RuBP Botryococcene P.V. 2 3GP Platform O Squalene 2 Rhodobacter Organism e- MVA Engineering IPP NADP H2 MEP H20 Process Modeling NADPH & Economics H2

Figure 1-2. Conceptual depiction of an Electrofuels process. Energy from the sun is captured in the form of renewable electricity (depicted as solar photovoltaics, PV) and is used to split water. The resulting O2 and H2 is combined with CO2 and fed into a bioreactor where R. capsulatus consumes these gases in stoichiometric requirement. R. capsulatus is genetically engineered to produce a C30+ triterpene hydrocarbon which can be recovered as an extracellular phase-separated oil.

Under the Electrofuels initiative, a range of approaches to this challenge were funded

(summary found in (A. Tuerk 2011)). The concept for our approach is depicted schematically in

Figure 1-2 and involves collecting and transporting electrons to a centralized bioreactor for biological capture of the reducing power in the chemical bonds of a hydrocarbon fuel. This proceeds by: (1) the capture of solar energy into electrical energy via photovoltaic cells (with demonstrated laboratory efficiencies upwards of 40%, a seven-fold improvement on photosynthesis), (2) the use of the generated electricity to split water into molecular hydrogen

(H2) and oxygen (O2), and (3) feeding these gases, along with carbon dioxide (CO2) emitted from point sources such as a biomass or coal-fired power plant, to a microbial bioprocessing platform.

6 Our proposed microbial bioprocessing platform leverages a chemolithoautotrophic microorganism (R. capsulatus or R. eutropha) naturally able to utilize these gases as growth substrates, and genetically modified to produce a triterpene hydrocarbon fuel molecule (C30+ botryococcenes) native to the alga Botryococcus braunii. The details and rationale for these choices are discussed below.

III. Rationale for Selection of the Microbial Bioprocessing Platform

Biological approaches for producing alternative fuels are benefitting from advances in molecular biology. These developments have increased the range of fuel molecule targets that can be synthesized by living organisms (Kim et al. 2006; Farmer & Liao 2001), as well as expanded the selection of alternative hosts for production through the development of new genetic engineering tools (Steinbrenner & Sandmann 2006). However, current approaches most frequently use E. coli or yeast as the microbial host with carbohydrate (i.e. fixed-carbon) based substrates. Our approach to the Electrofuels initiative leverages developments in alternative hosts capable of autotrophic growth on H2 as well as developments in fuel targets.

1.III.I.I Microbial Host Selection

For my work, I studied two different chemolithoautotrophs (microbes growing on H2, O2 and CO2) based on specific differences in their physiology, metabolism and energetic yields – R. capsulatus and R. eutropha. I have primarily looked at R. capsulatus, a purple non-sulfur facultative phototroph, as a candidate because of its capability of diverse metabolic modes

(including autotrophic), ability to utilize a range of growth substrates, and innate high level production of carotenoids (Hunter et al. 2009) indicating that pathways for isoprenoid

7 biosynthesis are active in this organism. On the other hand, the chemoautotrophic growth mode was developed as an alternative approach to bioprocessing using R. eutropha (Tanaka et al.

1995). R. eutropha is a well-studied chemoautotroph, whose use was initially motivated by its robust growth and the production of the industrially relevant biopolymer poly-hydroxy butyrate

(PHB). High-density growth (40 – 90 gDW/L) of R. eutropha on gaseous substrates was achieved in high gas-liquid-mass-transfer bioreactors with controlled addition of inorganic nutrients, demonstrating the technical feasibility of such a process. As will be discussed later, R. capsulatus is less efficient in utilizing H2 for growth, but has lower maintenance requirements than R. eutropha, differences which provide economic tradeoffs in an Electrofuels process. The range of yield and maintenance parameters between R. capsulatus and R. eutropha is likely representative of most chemoautotrophs and we were interested in quantifying how these contrasting energetic needs affect the overall economics of the process.

1.III.I.II Hydrocarbon Fuel Molecule Selection

One of the largest barriers faced by alternative energy technologies is poor economics relative to fossil fuels. Having been the most widely developed biofuel to date, ethanol has achieved the most success competing economically with gasoline. However, ethanol’s relative competitiveness results in part from governmental subsidies and mandates that ensure lower cost of ethanol feedstocks and higher selling prices (Solomon et al. 2007). Estimates of the Energy

Return on Investment (EROI) of corn-derived ethanol currently range from a mere 0.7 to 1.65, while those estimated for cellulosic ethanol range from 4.40 to 6.61 (Hammerschlag 2006; de

Castro et al. 2014).

We have chosen the hydrocarbon botryococcene as the product of the Electrofuels process as a means of addressing issue of energy-intensive industrial processing, which is a major

8 drawback to ethanol production. In choosing this fuel molecule target, the project leveraged our experience in working with the alga Botryococcus braunii (Khatri et al. 2014) and the associated discovery of the enzymes responsible for the synthesis of C30 squalene and botryococcene (Okada et al. 2004). C30+ triterpenes include methylated botryococcenes and squalenes, and are produced by B. braunii race B, a colonial green algae that accumulates these hydrocarbon oils up to 30 percent of its dry weight composition (Metzger & Largeau 2005). Ancestors of this alga have been implicated in producing up to 1.4 percent of the total hydrocarbon in Maar-type oil shales based on the unique carbon footprint of C34 botryococcene oil (Derenne et al. 1997). They have properties similar to crude oil (Hillen et al. 1982), a higher energy density than ethanol (38.1 vs.

23.5 kJ/L) (A. Tuerk 2011) and can be easily separated from an aqueous phase by decantation

(Khatri et al. 2014). Given these advantages, it was anticipated that the production of botryococcenes could achieve a higher EROI than ethanol.

IV. Metabolic engineering strategy

The synthesis of triterpenes occur via the terpenoid synthesis pathways. There are two distinct natural terpenoid pathways – the mevalonic acid (MVA) patway and the methylerithritol phosphate (MEP) pathway. The metabolic engineering strategy relied on introducing one or both of these pathways into the autotrophic host in addition to the specific triterpene synthase – squalene or botryococcene synthase (SS or BS). The strategy is schematically depicted in Figure

1-3.

9

This is the preferred means for engineering high levels of FPP biosynthesis into Rhodobacter

Operon 1 Operon 2 Operon 4

Pro ERG10 HMGS tHMGR Pro ERG12 ERG8 MVD1 Idi1 FPS Pro SS/BS TMT

Acetyl-CoA Acetoacetyl-CoA HMG-CoA Mevalonate Mev-P Mev-PP IPP DMAPP

Mevalonate (Mev) pathway (eukaryotic) 2x TRITERPENES

FPP

Methyl erythritol phosphate (MEP) pathway (prokaryotic) 2x METHYLATED Pyruvate TRITERPENES + DXP X MEP CDP-ME CDP-ME2P ME-2,4cPP HMB4PP IPP DMAPP G3P

Knockout mutation to be introduced This pathway exists and into the DXP Pro DXS Idi2 FPS Operon 3 operates in Rhodobacter reductoisomerase natively – but is subject locus to endogenous regulatory mechanisms controlling carbon flux down the pathway. This is the alternative approach to engineering high FPP levels in Rhodobacter. But this is not the preferred approach because it relies on too many of the “native” Rhodobacter genes, which will be subject to endogenous regulatory mechanisms.

Figure 1-3. Metabolic engineering strategy for the production of tritepenes in R. capsulatus.

MEP pathway is naturally found in many bacteria including the ones relevant for this work – E. coli, R. capsulatus and R. eutropha. Therefore, the natural metabolic flux of these bacteria can be channeled into the target triterpene molecule by introducing only the key limiting enzymes of this pathway. This is designated as Operon 3 in Figure 1-3. On the other hand, all the enzymes of the MVA pathway must be installed in order for this pathway to be functional in the hosts of our choice. This involves eight enzymes in total and therefore the pathway is broken down into two synthetic operons – Operon 1, converting acetyl-coA to MVA and Operon 2, converting MVA into farnesyl pyrophosphate (FPP). Farnesyl pyrophosphate synthase (FPS) is technically not part of either of these pathways but it is also found to be a rate-limiting enzyme and therefore are included in both of these pathways. More details can be found in Chapter 4.

10 V. Dissertation outline

The project can be divided into three major aspects: 1. Modeling and economic assessment of the electrofuels process, 2. Engineering of the organisms and pathways for the production of biofuel, 3. Bioreactor design and scale-up to improve productivity. The economic assessment provides the potential for economic feasibility of the electrofuels process under some given scenarios as well as identifies the areas for improvement that will have the highest impact.

The electrofuels scenario is quite different from the existing biofuel (e.g. corn or cellulosic ethanol) production processes. No model for this process thus exists for the prediction of the performance and cost of the fuel from this process. In Chapter 2, I describe a framework for integrating the different aspects of the electrofuels process as well as develop bioreactor and organism model that are used for carrying out a process economic assessment. This is published as an article in Bioresource technology (Khan et al. 2014).

Chapter 3 is a literature review of the current status of metabolic engineering of chemolithautotrophic hosts. It briefly covers chemolithoautotrophic metabolism in terms of electron capture and carbon fixation and more in-depth the recent advancements in genetic engineering of these organisms to produce fuels and chemicals, performance of the native and engineered systems and finally the major current and future challenges. It is submitted as an invited review in the journal Metabolic engineering. Chapter 4 describes our efforts to genetically engineer R. capsulatus to produce C30 triterpene hydrocarbons and studying its performance under different conditions including autotrophic.

Chapter 2

A process economic assessment of hydrocarbon biofuels production using chemoautotrophic organisms

I. Preface

In this chapter, the economic analysis of an ARPA-e Electrofuels process is presented, utilizing metabolically engineered R. capsulatus or R. eutropha to produce the C30+ hydrocarbon fuel, botryococcene, from hydrogen, carbon dioxide, and oxygen. This is published as an article in the journal Bioresource Technology. The analysis is based on an Aspen plus® bioreactor model taking into account experimentally determined R. capsulatus and R. eutrohpha growth and maintenance requirements, reactor residence time, correlations for gas-liquid mass-transfer coefficient, gas composition, and specific cellular fuel productivity. Based on reactor simulation results encompassing technically relevant parameter ranges, the capital and operating costs of the process were estimated for 5000 bbl-fuel/day plant and used to predict fuel cost. Under the assumptions used in this analysis and crude oil prices, the Levelized Cost of Electricity (LCOE) required for economic feasibility must be less than 2¢/kWh. While not feasible under current market prices and costs, this work identifies key variables impacting process cost and discusses potential alternative paths toward economic feasibility.

II. Specific contributions

I acted as first author, wrote major portion of the text and facilitated submission of the manuscript to Bioresource Technology. I carried out the measurement of the growth yield and

12 maintenance coefficients for R. capsulatus and R. eutropha, refined the microbial energetic model for fuel production, developed the bioreactor model in Aspen plus® to integrate the experimentally measured parameters with the energetic model of fuel production to predict the operating cost of the process. Amalie Tuerk assisted in refinement of writing and energetic analysis. John Myers facilitated the capital cost estimation of the process and had done considerable writing associated with his thesis. Erik Wolcott helped with the compilation of the gas-liquid mass-transfer coefficients for the various reactor systems.

III. Introduction

This exercise of using process economic analyses to research priorities is an important aspect of the ARPA-E program. An initial analysis of the economic feasibility of the microbial process, based on microbial energetic theory, can be found in an extensive thesis (A. Tuerk

2011). A preliminary process economic model can also be found in (Myers 2013). In this part of the work, I expand upon these previous analyses by constructing a more detailed bioreactor model in Aspen Plus® (described in section 2.IV.II). Furthermore, I examined specific scenarios based on reactor residence time (τ), specific fuel productivity () and gas-liquid mass transfer coefficient (kLa) to predict their effect on the overall process volumetric fuel productivity () and fuel cost ($/bbl-fuel). The ultimate goal of this analysis was to identify limiting parameters for the process and ranges of these variables needed to achieve economic feasibility.

13 IV. Model development

2.IV.I. Proposed electrofuels production process

The process flow diagram of the proposed Electrofuels production process is depicted in

Figure 2-1. Production of the C30+ hydrocarbon fuel by the host organism occurs in the bioreactor,

which is fed the gaseous substrates H2, O2 and CO2. The relative amount of these gases fed is equal to the ratio that supports growth and fuel production by the bacteria. The determination of this stoichiometric ratio is discussed in further detail under Section 2.IV.II.IV. Other nutrients required for cell growth, such as salts and vitamins, are fed to the bioreactor via a liquid stream.

The bioreactor is operated continuously and the bioreactor effluent contains biomass, botryococcene fuel product, and spent culture medium. An underlying assumption critical to this process configuration is that the botryococcene fuel is excreted and deposited as a hydrophobic layer on top of the culture, which can be separated by decantation (oil-water separator). The decanted hydrocarbon layer then flows from the oil-water separator to a fuel filtration system where the residual aqueous phase is removed, and the fuel molecules can then exit the process as the crude hydrocarbon product. We did not include subsequent processing, such as hydrocracking, within the scope of this economic analysis, since the point of comparison is crude oil, which would require similar downstream processing (Hillen et al. 1982). The aqueous phase that exits the oil-water separator flows to a cell clarifier. This concentrates the biomass phase before returning a portion of the biomass to the bioreactor, and the remaining biomass to sludge processing to avoid accumulation of dead cells and cellular waste (perfusion mode operation).

The spent media passes through a waste treatment process (a second clarifier or filter) to remove cellular debris and waste material, and is recycled as fresh medium after supplementing with salts and other nutrients.

14

Figure 2-1. Process flow diagram of the envisioned Electrofuels process. CO2, H2, O2, fresh nutrient and recycled cells enter the reactor and cells+oil leave the reactor. The hydrophobic top oil layer is separated in oil-water separator and pass through fuel filtration and processed to final product. The remaining cells+water+residual oil from oil-water separator enters a clarifier where the denser cell layer is split into two fractions, one recycled back to the reactor and one sent to sludge processing. The top, less dense overflow from this clarifier enters a second clarifier where cell debris are separated from the water, which is processed through waste-water processing to be recycled to the reactor with fresh nutrients and makeup water.

2.IV.II. Bioreactor modeling

The biological processes taking place in the bioreactor (cell growth, maintenance and fuel molecule synthesis) determine the rate of substrate consumption (H2, CO2, O2), and therefore the gas-liquid mass-transfer (kLa) requirements, which are the most critical components of the bioreactor operating cost. The bioreactor volumetric fuel productivity (dependent on the specific fuel productivity, residence time and kLa) and the scale of the process (bbl-fuel/day) determine the required size of all associated process equipment, and thus the capital and fixed costs. The

15 capital investment at that plant size can then be amortized at a nominal internal rate of return and combined with the operating costs to arrive at an estimated cost for crude fuel production ($/bbl- fuel). A detailed model of the bioreactor processes associated with hydrocarbon fuel production, which includes cell growth and maintenance, fuel synthesis, and gas transport, is therefore a critical prerequisite to accurately estimating the capital and operating costs. The assumptions related to the bioreactor and development of the associated material balance equations are described in Appendix A.

In aerobic H2-utilizing chemoautotrophic metabolism (referred to as ‘autotrophic’ metabolism from here on), electrons are transferred from H2 to O2 through a series of steps which generate the energy needed for various cellular processes, including CO2 fixation for growth and fuel synthesis. Thermodynamic models (Mccarty 1971; McCarty 2007; A. Tuerk 2011) can be used to calculate these requirements (examples of these methods for calculating the true yield are demonstrated in Appendix B). The microbial substrate requirements ultimately reduce to stoichiometric equations that can be incorporated in a bioreactor model. To integrate the bioreactor model with the full model of the Electrofuels process, we used Aspen plus® software.

We selected Aspen plus® as it can solve complicated material and energy balance equations with rigorous physical property determinations, and we selected the RCSTR reactor model to incorporate the microbial energetic equations described below in sections 2.IV.II.I through

2.IV.II.IV.

2.IV.II.I Substrate requirement for cell growth

Figure 2-2A depicts the conceptual framework for modeling cell growth and maintenance processes. In this context, growth processes refer to the use of energy and carbon substrates for the production of new cellular biomass, whereas cellular maintenance is the use of energy

16 substrates for non-growth processes such as motility, ion pumping, futile cycling etc. (van

Bodegom 2007; Pirt 1965). For autotrophic metabolism, the process of cell growth (excluding energy demands for maintenance) can be represented by a stoichiometric relationship, the

‘growth equation’ as shown below in Equation 2-1.

CO + aH + bNH + cO → CellCNHO() + dHO + bH (2‐1)

In this relationship, the inorganic carbon source (CO2), electron donor (H2), electron

acceptor (O2), and nitrogen source (ammonium, NH ) are consumed, resulting in the formation of

+ cell biomass, water (H2O), and hydrogen ions (H ). Consequently, the rate of consumption of

each of the gases due to growth processes , where i represents H2, CO2, or O2) is

[] related to the rate of cell growth by Equation 2-2,

[] r = = = = (2‐2)

The indices for the elements N, H and O in the cell biomass term of Equation 2-1 [,

2 + 3 − 2 and 2 + 2 − , respectively] were obtained from literature reports of the empirical elemental formula for cell biomass for each R. capsulatus and R. eutropha (Hoekema,

R. D. R. D. Douma, et al. 2006; Ishizaki & Tanaka 1990), resulting in three of the four equations needed to calculate the four coefficients a, b, c and d. The fourth equation needed to complete the set is from the true growth yield (substrate requirement for growth that could occur in the absence

of cellular maintenance requirements) of the bacteria on the energy substrate H2 () and is related to the coefficient a by Equation 3,

Y = (2‐3)

17

where is the formula weight of one ‘mole’ of cell according to the empirical

formula for cell biomass. Because is the quantity of H2 needed for growth-only processes

(and does not include energy requirements for maintenance), this parameter can be obtained through the use of microbial energetic models or experimental measurements (Pirt 1965).

Energetics of bacterial Energetics of fuel synthesis growth A 20ATP B CO2 10NADH

Growth yield Maintenance 9ATP RuBP (YT) coeffcient (m) 6NADPH Glyceraldehyde-3- phosphate (G3P)

2ATP 2NADH H2 + ½O2  ½H2O Maintenance equation 3- phosphoglycolate ΔGfuel (PGA) 2ATP + + 2NADH CO2 + aH2 + bNH4 + cO2 Cell + dH2O + bH Growth equation Acetyl CoA 2NADPH

Mevalonate 22ATP 11NADPH

34CO2 + (29+f)H2 + f/2O2  C34H58 + fH2O Fuel synthesis equation Botroycoccene

Figure 2-2. Framework for capturing microbial growth energetics, cellular maintenance and fuel production. (A) Incorporating the experimentally measured biological yield and maintenance coefficients into model reaction/stoichiometry equations suitable for reactor design. (B) Concise version of metabolic pathway for calculating the minimum required Gibbs’ free energy for fuel synthesis in a bacteria.

18

2.IV.II.II Substrate requirement for cellular maintenance

Additional consumption of H2 and O2 occurs in order to generate energy for cellular maintenance. Because cellular maintenance requirements are defined as independent of energy requirements for growth, the simplest way to conceive of these requirements is to consider them

on a per-time, per mass of cell basis, and thus the maintenance coefficients ( and ) are a measure of the rate of substrate required to satisfy the maintenance needs of one unit mass of cells. Cells are also known to utilize the already formed cellular material under energy-starved condition. This is termed endogenous respiration and is not specifically taken into account in this model. It is assumed that in active, albeit suboptimal, lithotrophic growth with abundance of H2, this quantity would be relatively small. The rate of consumption of H2 and O2 due to cellular maintenance is therefore proportional to the total biomass concentration in the reactor, X, as shown in Equation 4.

m X = = m X (2‐4)

is related to by a factor of two because of the stoichiometry of water formation, which is the reaction for energy generation in autotrophic metabolism since electrons from H2 are

transferred to O2. We initially presented values of and based on literature reports and used them in conjunction with microbial energetic theory to predict net growth and fuel yields under variety of conditions (A. Tuerk 2011). To confirm these estimates towards improving the

accuracy of this economic analysis, we have recently measured the values of and under

19 identical conditions for both R. capsulatus and R. eutropha using an adaptation of the traditional methods using continuous chemostat cultures under autotrophic conditions (Pirt 1965).

2.IV.II.III Substrate requirement for fuel synthesis

To incorporate botryococcene fuel synthesis into the bioreactor model, we first defined a parameter, , the specific fuel productivity. The rate of fuel synthesis is therefore given by

Equation 5.

= RX (2‐5)

The botryococcene fuel molecule, C34H58, is energetically expensive to synthesize.

Therefore, additional H2 and O2 must be consumed to provide energy for fuel synthesis, beyond the requirements for cell growth. We designate the parameter to account for the additional H2 which must be oxidized in order to provide this energy; this is above and beyond the hydrogen incorporated into the botryococcene molecules themselves. The overall fuel synthesis reaction is summarized in Equation 6. This treatment is analogous to cell growth and maintenance described in sections 2.IV.II.I and 2.IV.II.II.

34CO + (29 + f)H + O → C H + fH O (2‐6)

The minimum energetic requirement for botryococcene synthesis from CO2 via the

Calvin-Benson-Bassham (CBB) and mevalonic acid (MVA) pathways can be calculated by summing the total Gibbs energy required for generating the cofactors (mainly ATP, NAD(P)H and NADH) in the metabolic steps. A simplified version of the relevant metabolic pathways is shown in Figure 2-2B. The cofactor requirement for going from mevalonate to botryococcene has been previously elucidated by (Okada et al. 2004). The total Gibbs energy required for

20 synthesizing one mol of botryococcene from CO2 was converted to the amount of H2 (f in

Equation 6) that must be oxidized to provide this energy. The rate of consumption of H2, O2 and

CO2 for fuel synthesis can then be calculated through Equation 2-7,

RX = = = (2‐7)

The fuel mass yield on H2 ( ) is defined as the quantity , where is the / formula weight of botryococcene.

2.IV.II.IV Mass transfer limitations for gas transport and optimization

The total rate of consumption of each of the gases is the sum of the contributions from cell growth, cellular maintenance, and fuel production (Equation 8),

= + + (2‐8)

Based on this premise, the volumetric productivity of botryococcene is ultimately constrained by the mass transfer rate of the limiting gas substrate into liquid. At steady state conditions in the bioreactor, where liquid-phase gas concentrations are constant, the uptake rate

of the i-th gas substrate by the organism is equal to the gas transfer rate of this gas component (). For gaseous component i, these rates can be calculated from the gas phase fractional composition (), the total gas pressure (), the gas-liquid mass transfer coefficient

(,), Henry’s law coefficient () and the liquid phase concentration (), as shown in

Equation 9:

= GTR = ka − C (2-9)

21

Thus, this equation illustrates that the , and therefore the total possible gas uptake by the organisms, are limited by the that is possible in a particular reactor type.

To simultaneously satisfy the organism’s H2, O2, and CO2 demands, there is only one set of at which all the gases are transported at a stoichiometrically balanced rate. At any other gas composition, only the limiting component will transport at the maximum rate ([] = 0), while the other components are transported at a sub-maximal rate ([] > 0). Therefore, the bioreactor must be operated at the gas composition that results in maximum transport rate of each of the gases to achieve the highest volumetric productivity. In practice, this could be achieved by controlling the gas phase composition via feedback control loops. In our simulations, we used the optimization functionality in Aspen plus® to converge on the optimal gas composition. Because the bioreactor

also affects the , as described above, we constrained the optimization by limiting

to values correlated for bioreactor configurations. Fixing then fixes and

, as these are related to by the diffusion coefficient correction to mass transfer

(Equation 10),

/ ka = ka (2‐10)

An inter-conversion based on diffusion coefficient to 2/3 power is used. This is because

correlations in literature are typically reported for as a function of operating conditions

(rpm, gas rate etc.) and mass transfer coefficients have an inherent dependence on gas diffusion in the associated boundary layer, which needed to be captured in the analysis.

22 V. Results and discussion

Continuous operation provides for the high productivity required for fuel production.

Extracellular botryococcene accumulation allows for the recycle of biomass for a higher operating concentration and higher process volumetric productivity. Since achieving economic feasibility is anticipated to be challenging, these operational enhancements are assumed as baseline requirements.

2.V.I. Sensitivity analysis: effect of reactor residence time (τ)

In a perfusion bioreactor, the residence time through the reactor system ( = V/F) and cell retention efficiency (ε) determine the required cell proliferation rate (µ) and operational cell concentration (X). The associated equations relating these quantities are developed in Appendix

A. These subsequently affect the rate at which fuel will be produced for a given specific fuel productivity (). The economics of this process are highly dependent on the volumetric fuel productivity (; this sets the required reactor size and therefore the capital cost) and the fuel

yield on H2, (/; the major operating cost). Therefore, we performed an initial set of simulations to evaluate the effect of reactor residence time on these two variables (Figure 2-3).

The steady state cell density in the reactor is also presented to assist in interpretation (Figure

2-3C) and to establish the operational cell concentration that must be achievable using the host organism. For these simulations experimentally determined values for microbial energetic

parameters were used (unpublished data): true growth yield () and maintenance coefficient

-3 () of both R. capsulatus (2.66 gDW/mol-H2 and 2.01x10 mol-H2/gDW·hr, respectively) and

-3 R. eutropha (7.68 gDW/mol-H2 and 6.8x10 mol-H2/gDW·hr, respectively). Figure 2-3 presents

23 the results for the chosen process variables: = 330/hr, = 0.5 g-fuel/gDW·hr and ε =

95%.

The results indicate that X, and / all increase with residence time, asymptotically approaching maximum values. This behavior results from the opposing effects of X and : in a

continuous flow bioreactor, = growth rate of the bacteria () = (given < ).

Therefore, as increases, specific growth rate decreases. Since the volumetric rate of substrate availability is fixed by gas mass transfer (kLa), the reduction in specific growth rate and reduced gas consumption per cell permit an increased steady state cell concentration with an associated greater fuel production per unit volume. Based on the assumption that metabolism can be directed to produce fuel instead of cell mass results in operating at lower growth rates. At lower growth rates less of the available H2 must be partitioned to cell growth, enabling more of the H2 to be

used towards synthesizing fuel, and enabling a higher /. This is reflected in Figure 2-3C.

The preceding discussion also explains the important observation that at higher residence times (and lower growth rates) the organism with lower maintenance requirement (R. capsulatus, in this case) is predicted to perform better despite its less efficient utilization of H2 for growth.

This is because at a lower growth rate, a smaller fraction of the available energy substrate is used for maintenance, leaving more H2 available for fuel synthesis. The assumption that mass transfer limitations are the dominant constraint on bioreactor operation is central to this argument.

Because it is desirable to operate at higher fuel yield and volumetric productivity, and of

R. capsulatus and = 15 hr were chosen for further simulations.

24

1 A .hr)

3 0.8 Rhodobacter capsulatus 0.6

Volumetric Volumetric Ralstonia eutropha productivity (kg fuel/m (kg 0.4 3.0 B ) 2 2 H H on mol 2.0 Yield Yield (gfuel/ 1.0

10 C

density 8 biomass/L) Cell Cell

gDW 6 ( 4 0 5 10 15 Residence time (hr) Figure 2-3. Variation of volumetric productivity, (A), fuel yield on H2, / (B) and cell density, (C) as a function of residence time () through the reactor for R. capsulatus (●) and R. eutropha (o). Simulation parameters: true growth yield () of R. capsulatus = 2.66, R. eutropha = 7.68 -3 - gDW/mol·H2, maintenance coefficients () of R. capsulatus = 2.01x10 , R. eutropha = 6.8x10 3 mol-H2/gDW·hr, kLa = 330/hr, specific productivity = 0.5 g fuel/g biomass.hr, cell recycle efficiency = 95%.

2.V.II. Sensitivity analysis: effect of specific productivity (Rfuel)

Specific fuel productivity is an intrinsic biological property of a genetically engineered production strain, and affects the volumetric productivity as well as the fuel yield on H2. The range of specific fuel productivities examined in this analysis spanned from 0.1 to 2 g- fuel/gDW.hr (Figure 2-4A). This is a wide range covering conservative to very optimistic specific

25 productivity targets. As a reference, the ethanol specific productivities of E. coli growing on various fixed-carbon substrates reported in the literature range from about 0.3 to 0.8 g·gDW-1·hr-1

(estimated from Ohta et al, 1991; Hildebrand et al, 2013). The resulting fuel yields on H2 were converted to electricity required (kWh/bbl fuel), based on an average electrolytic efficiency of

80% for H2 generation (Myers 2013). This is the major component of the operating cost of the process. In the range of low specific productivities, the electricity requirement for H2 generation initially decreases sharply as specific productivity increases. This is attributed to the fact that when specific productivity is low, the majority of the energy is expended in the growth of cells rather than ending up in the fuel molecule, so small improvements in specific productivity give large electricity cost savings. However, the electricity requirement remains essentially unchanged above specific productivities of 1 g-fuel·gDW-1·hr-1. This asymptotic minimum is the sum of the energy fixed in the fuel molecule and the minimum loss that is associated with cell growth and maintenance.

Specific productivity is the target associated with genetic engineering and development of the host organism. Obviously, higher is better, but the more important question is: at what range of specific productivity will the process achieve economic viability? Initial efforts to introduce the botryococcene synthesis pathway via metabolic engineering approaches have successfully produced the botryococcene molecule (Khan et al. 2013); however, the observed production levels are more than an order of magnitude lower than the low-range values used in this simulation. For the current, experimentally observed productivity levels, it was calculated that the electricity cost would be about 175 times that of the 0.1 g-fuel·gDW-1·hr-1 specific productivity case. The genetic engineering efforts are well justified because a small increase in specific productivity (which is possible at the early stages of development) translate to large savings in electricity cost. However, after a significant productivity has been attained, and electricity requirements level off, other aspects of the process become increasingly important for

26 achieving economic feasibility. For our subsequent analysis, the moderately optimistic level of

0.5 g-fuel·gDW-1·hr-1 was used.

Figure 2-4. (A) Variation of electricity requirement for H2 generation (●) and volumetric productivity (o) with specific fuel productivity representing an increasing fraction of energy being converted into fuel instead of biomass (Simulation parameters: residence time = 15 hr; other parameters same as in Figure 2-3). (B) Variation in reactor volume as a function of kLa showing the reduction in culture process volume that is enabled as the mass transfer rate is improved (). The range of electricity requirements that would be required to achieve the specified kLa in different bioreactor configuration is shown as the shaded area between the dashed lines. The high and the low value of the electricity requirement at each kLa corresponds to the highest and the lowest P/V able to produce the specified kLa; the range of P/V for a variety of bioreactor configurations are presented in Figure 2-5.

27

2.V.III. Sensitivity analysis: effect of gas-liquid mass transfer coefficient (kLa)

The mass transfer rate affects the volumetric productivity and therefore determines the reactor volume required at a given production capacity, as discussed in section 2.IV.II.III.

Increasing comes at the expense of increased operating costs for mass transfer, but this facilitates a decrease in the capital cost due to a reduction in the required reactor volume. Figure

2-4B shows the decrease in reactor volume and increase in energy cost (kWh/bbl-fuel) for an

-1 increase in kLa from 330 to 1000 hr for a range of bioreactor configurations. To serve as a guide in determining relevant reactor costs and realistic kLa values, the ranges of kLa used in Figure

2-4B and the associated power-per-unit-volume (P/V) are based on values for various gas-liquid contacting reactors reported in the literature. The compilation is presented in Figure 2-5. In performing the literature review, we focused on systems that can be operated at the large scale required for this process. The key is to choose a bioreactor where high mass transfer can be sustained at relatively low power levels. In this respect, trickle-bed reactors were found to perform the best at commercial scales. While microbubble reactors have similar performance, they have only been demonstrated at the laboratory scale. We have therefore used trickle-bed reactor costs in our final analysis. Slight biofilm formation on reactor wall is expected and observed experimentally, but operational strategies such as pulsatile flow can be implemented to keep this to a minimal and is not expected to be a problem.

Comparing the energy costs for kLa enhancements in Figure 2-4B to the costs of energy for H2 production in Figure 2-4A, energy required for mass transfer is only 1.5 to 4.5% of the minimum energy required for H2 generation (~4000 kWh/bbl-fuel). This result emphasizes the dominance of H2 generation requirements in overall energy requirements. Optimizing kLa, therefore, will only become important in the later stages of plant design when more accurate

28 equipment costs are available. Therefore in determining the total capital and operating cost of the

-1 process, we selected a kLa of 1000 hr .

Surface aerator 1 Surface aerator 2 10000 Stirred 1 Stirred 2 Stirred 3 TBR 1 TBR 2 Microbubble - various 1000

100 a (1/hr) L k 10

1 10 100 1000 10000 P/V (W/m3)

Figure 2-5. Gas-liquid mass transfer coefficient (kLa) vs. power/volume (P/V) for various reactor types. Surface aerator 1 = wastewater treatment plant demonstration run, Chattanooga, TN (Cosby & Gay, 2003); Surface aerator 2 = wastewater treatment plant demonstration run, Yuba City, CA (Lewis & Gay, 2003); Stirred 1 = stirred tank reactors correlation for electrolytes, (Van’t Riet, 1979), Vs = 0.005 m/s; Stirred 2 = stirred tank reactor correlation for non- electrolytes (Van’t Riet, 1979), Vs = 0.005 m/s; Stirred 3 = stirred tank reactors correlation for electrolytes (Linek et al., 1987), Vs = 0.005 m/s; TBR 1 = trickle-bed reactor correlation set 1 (Reactor parameters: Roininen et al., 2009, kLa correlation: Goto & Smith, 1975, pressure-drop correlation: Larachi et al., 1991); TBR 2 = trickle-bed reactor correlation set 1 (Reactor parameters: Roininen et al., 2009, kLa correlation and pressure-drop correlation: Reiss, 1967); Microbubble – various = experimental data from various investigators (Bredwell & Worden, 1998; Hensirisak et al., 2002). This compilation of literature values focuses on large-scale commercial units (except for microbubble technology).

29 2.V.IV. Process economic analysis of capital and operating costs

The preceding analysis, based on specific productivity and mass-transfer requirements, helps define the realistic range to conduct a more detailed analysis of capital and operating costs.

Table 2-1 shows the values selected for parameters used in this section of the work, all of which are based on the calculations described above. The scaled-up process design is based on a daily production rate of 5,000 barrels of oil per day (795,000 L/d). As a reference point, in 2012 the largest ethanol production facility in the United States had a nameplate capacity of slightly over 27,000 bpd, while the average facility produced around 4,500 bpd (Nebraska 2014).

Table 2-1. Parameter values used in final scaled-up design.

Parameter Value

Plant size 5000 bbl/day

Specific productivity 0.5 gfuel/gDW.hr

Oxygen mass-transfer coefficient (kLa) 1000/hr

Cell recycle 95%

The results of the capital cost analysis are presented in Figure 2-6A; as can be seen, the costs of the process components varied over a range of three orders of magnitude. We grouped the equipment into categories of lower and higher cost, and note that the capital cost of photovoltaics included in H2 generation costs corresponds to the highest capital expense. The capital cost of the platform (excluding H2 production) is $24.44/bbl-fuel, assuming an after-tax rate of return of 15% (B). The fixed cost component was much smaller than the capital cost component at $8.81/bbl-fuel. The operating cost of the process is highly dependent on the cost of electricity and the LCOE of electricity is quite variable for the various types of renewable energy

30 technologies. However, the cost of nutrient feed-water, wastewater, cooling water, and CO2 are relatively insensitive to electricity cost, and amount to ~$18.9/bbl-fuel. Although we expected

CO2 cost to be minimal due to it being a waste product, surprisingly, this cost was found to be the most significant cost after electricity. In this analysis we assume that CO2 from a power plant or a similar point source will be purified, compressed, and transported to this facility for use. The magnitude of this cost varies substantially with the types of point source, carbon capture technology employed, and the means of transport.

The average of the cost of capture and compression of CO2 for IGCC, NGCC and PC power plants is reported to be $31/tonne-CO2 (Metz et al. 2005). Transportation costs for CO2 via pipeline or tanker transport varies from 1 to 5 USD/tonne-CO2/250 km (Metz et al. 2005). Using an average cost of $3/tonne-CO2/250 km and an average distance of 500 km between the CO2 point source and the Electrofuels plant, we predict CO2 transportation costs of $6/tonne-CO2.

Thus, using the total predicted CO2 cost of $37/tonne-CO2 results in a platform cost of $17/bbl- fuel. As shown in Figure 2-6B, the overall cost of the Electrofuels production platform (excluding the cost of H2 and the much smaller bioreactor operational requirements for mass transfer) is therefore estimated to be around $54.3/bbl-fuel, a price well below the target sales price of

$127/bbl-fuel (based on the DOE’s estimate of the 2020 price of oil (Gruenspecht 2012)).

31

A Electrolyzers Figure 2-6. (A) Capital costs of Reactors various process components for Reactor Packing Heat Exchangers a 5000 bbl/day fuel production Pumps plant. Results are grouped into Feed and Product Vessels 0 20 40 60 80 100 120 high (black bards) and low (dashed bars) range capital Cell Waste Clarifier Cell Clarifier costs. (B) A breakdown of Values in Millions Fuel Filtration process costs (the dominant cost of 2010 Dollars Oil Water Separation of electricity generation is not 0.0 0.2 0.4 0.6 0.8 1.0 presented). Top (grey) group represents the major operating costs which make up 36% of CO2 B Feed Water total; middle (hashed) group Waste Water Cooling Water represents the major non- Electrolyzers Reactors photovoltaic capital costs which Reactor Packing make up 46% of total; bottom Heat Exchangers $18.9 Pumps (black) group estimates the Feed and Product Vessels Cell Waste Clarifier major non-PV fixed costs Cell Clarifier $8.8 corresponding to 17% of the Fuel Filtration $24.4 Oil Water Separator total. (C) A sensitivity analysis Property Tax of the final fuel cost based on SG&A Rep & maint. Capital Costs (Minus electricity generation) LCOE of various generation Overhead Operating Costs (Minus electricity) Insurance methods reported in two Fixed Costs Labor different sources: REN21, 2013 0 3 6 9 12 15 18 (black); IRENA, 2012 (dashed). Cost per bbl ($/bbl) CSP = Concentrate Solar Power, PT = Parabolic Trough, ST = Solar Tower. The horizontal line C 2500 indicates year 2020 crude oil REN21 price estimate (Gruenspecht, 2000 2012). Off-grid hydropower IRENA values were not available in IRENA, 2012. 1500

1000 Fuel cost (US$/bbl) Fuel cost 500

0 wind Wind CSP (ST) CSP (PT) CSP Solar PV Solar (onshore) (offshore) (Off-grid) Hydropower

32 The challenge is therefore to determine the true effect of renewable electricity costs on this technology, which is complicated by the range of various technologies available and the large variability of their costs in North America and around the world. The International Renewable

Energy Report (IRENA 2012) shows that the worldwide LCOE of various renewable energy technologies varies from as low of 4 cents/kWh for onshore wind power to a high of >30 cents/kWh for large-array solar PV. For North America the averages range from 4 cents/kWh for hydropower to 50 cents/kWh for solar PV (REN21 2013; IRENA 2012). The range of final fuel costs, based on the potential range of renewable electricity costs, is shown in Figure 2-6C. As can be seen, the range of final fuel costs spans a very wide range, which reflects the wide variation found in the renewable electricity sources. Furthermore, the lowest projected cost (187.2-

748.8/bbl-fuel; based on onshore wind electricity) is still about two to six times higher than the

2020 crude oil price estimate ($127/bbl-fuel) (Gruenspecht 2012). The highest costs are generally for solar photovoltaics, which are about twenty times that of this baseline crude oil price.

To place the cost of these electricity generation technologies in perspective, we are presenting a brief comparison based on two traditional, fossil fuel based technologies for H2 generation: coal gasification and natural gas reforming. In a recent study, (Bartels et al. 2010) reported hydrogen production costs of 0.36-1.83 $/kg and 2.48-3.17 $/kg from coal and natural gas respectively. This produces ranges of 68.8-205.4 and 265.8-330 $/bbl-fuel respectively based on the Electrofuels scenario. Although much lower in cost than some of the renewable technologies for H2 generation, both coal gasification and natural gas reforming generate a much higher ratio of CO2 to H2 than can be consumed by the process. From our analysis, we found that the maximum ratio of consumption of CO2 to H2 in the electrofuels process is 0.2 mol-CO2/mol-

H2. On the other hand, the best of coal gasification and natural gas reforming technologies would generate 1.08 and 0.42 mol-CO2/mol-H2. According to these numbers, at least 81% and 52% of

33 the CO2 generated by these technologies respectively would be either released to the atmosphere or have to captured by some other process.

It is obvious that at the current crude oil prices and renewable electricity costs, the

Electrofuels process is not economically favorable. There are several opportunities that were not taken into account in this analysis that could provide for somewhat improved economics. These include cost reductions for carbon credit, use of low-cost plastic fermentation systems (T. Hsiao et al. 1999), use of the waste biomass from the process for heating, etc. Furthermore, some of the

LCOE costs used in the analysis include capital and operating costs for transmission and distribution of the electricity to the grid, which is not necessary if a direct electrolysis process is adopted. Nonetheless, the overall analysis provides quantitative evidence that even with improvements and meeting cellular productivity metrics, the proposed process does not achieve economic feasibility in terms of producing a fuel product. It is not entirely unexpected that an electrofuel would have difficulty competing with crude oil at the current price, as the energy content of crude oil is a result of millions of years of accumulation.

2.V.V. Alternative perspective

A potential advantage of the hypothetical electrofuels process that is not captured in this analysis is the potential for using this microbial platform to make alternative, high-value products by leveraging the high specificity of synthesis that is an inherent capability of biological systems.

Through the use of genetic engineering, the autotrophic host could be designed to produce biochemical products with very high degree of specificity. For example, in parallel to the genetic engineering efforts to produce botryococcene, we have also demonstrated squalene production in

R. capsulatus (unpublished data). The current bulk price of squalene is $15-20/kg, which translates to about $2027-2702/bbl-fuel. Other isoprenoids, such as lycopene, valencene, beta-

34 carotene, etc., are also high value products whose production are demonstrated in non-native hosts (Alper et al. 2005; Beekwilder et al. 2014a; Verwaal et al. 2007). Wax ester (a health product commonly sold as Jojoba oil) is a fatty ester for which we have also demonstrated production in R. capsulatus (unpublished data). At ~$68/L, the retail price of Jojoba oil translates to a price of $8064/bbl-fuel. These examples demonstrate the potential value of alternative products able to be generated by a biological production platform using a sustainable, renewable substrate. This alternative paradigm for Electrofuels commercialization could facilitate the simultaneous production of multiple high value products (different reactors employing different genetically modified strains) using remote sources of electricity/energy. This could be an alternative use of the electrofuels process during the technology development phase until reductions in the cost of renewable electricity and/or the rise in the price of crude oil make

Electrofuels production economically feasible.

2.V.VI. Targets based on the analysis

One of the primary goals of this analysis was to first establish the effect of the key

process parameters – and of the bacteria, τ, and kLa – on the cost of fuel produced by an Electrofuels process, and subsequently define the target productivity metrics for potential feasibility. The range of values for and kLa used in this analysis were selected based on

realistic engineering considerations of what is actually feasible, while and are measured physiological values. The value of the residence time parameter (τ) is subsequently constrained by these parameters.

It was found that the cost of H2 generation (i.e. the cost of electricity) accounts for >90% of the current cost of fuel from the Electrofuels process. The next largest costs are the capital investment required for other non-H2 process equipment and the operating costs for providing

35 mass transfer and supplying CO2. Therefore, steps that decrease H2 consumption would

drastically improve process economics. Critical steps include using a host organism with high

- and low , and improving the specific fuel productivity to achieve a target >0.3 g-fuel·gDW

1 -1 .hr . The sharpest change in H2 cost occurs at specific productivities below 0.5, where H2 costs escalate rapidly. This means that incremental changes in productivity in this range have a large impact on the process economics. Also, since the cost of kLa is low relative to the H2 generation cost, and because increases in mass transfer have the potential to drastically decrease the required reactor volume (and thus reduce capital cost), it is advantageous to use a bioreactor that can achieve large kLa, even at the expense of increased energy costs for achieving a target kLa.

The other major goal of this analysis was to assess the overall economic feasibility of an

Electrofuels process. It was found that above a specific fuel productivity of 0.3, the process can be feasible at an LCOE <2¢/kWh at the current crude oil prices; this LCOE is about 2-20 times lower than the cost of electricity by current renewable electricity technologies. Given that the economic feasibility is the primary concern, producing fuels via the Electrofuels process is not a viable option at the moment. However, the various technologies considered in this analysis

(renewable electricity, autotrophic host, engineering and energetics of hydrocarbon pathways) are in early developmental phase and have the potential to drastically change the economics in future.

Alternative higher-value biochemical targets such as lubricants may be pursued during this phase of development to act as economic motivator for near-term profitability, so that the technologies can be developed for application when the rising crude oil prices (due to the depletion of reserves) allow acceptable economics for this process.

36 VI. Conclusions

Cost of electricity (for H2 generation) was found to have the largest impact on process economics. Given the assumptions made and current crude oil prices, the process is expected to be feasible at an LCOE ≤2¢/kWh. Performance parameters examined in this analysis included both organism-specific physiological parameters, and parameters for process operational modes.

A critical requirement for economic feasibility is an autotrophic host with a high yield on hydrogen, low metabolic maintenance energy requirements, and able to meet a specific fuel productivity metric of >0.3 g-fuel·gDW-1·hr-1. Gas-liquid mass transfer capabilities of the bioreactor were not found to be critical.

37 Chapter 3

Literature review: Metabolic engineering in chemolithoautotrophic hosts for the production of fuels and chemicals

I. Preface

This chapter presents a comprehensive review of the literature from the perspective of the metabolic engineering of chemolithoautotrophic hosts and written in this form as an invited review for the journal Metabolic Engineering. The use of chemolithoautotrophic hosts as production platforms has been renewed by the prospect of metabolically engineered commodity chemicals and fuels. Efforts such as the ARPA-E electrofuels program highlight both the potential and obstacles that chemolithoautotrophic biosynthetic platforms provide. This review surveys the numerous advances that have been made in chemolithoautotrophic metabolic engineering with a focus on hydrogen oxidizing bacteria such as the model chemolithoautotrophic organism (Ralstonia), the purple photosynthetic bacteria (Rhodobacter), and anaerobic acetogens.

Two alternative strategies of microbial chassis development are considered: (1) introducing or enhancing autotrophic capabilities (carbon fixation, hydrogen utilization) in model heterotrophic organisms, or (2) improving tools for pathway engineering (transformation methods, promoters, vectors etc.) in native autotrophic organisms. Unique characteristics of autotrophic growth as they relate to bioreactor design and process development are also discussed in the context of challenges and opportunities for genetic manipulation of organisms as production platforms.

38 II. Specific contributions

This review was a collective effort among Dr. Wayne Curtis, Dr. S. Eric Nybo, Ben

Woolston and myself. Dr. Curtis provided an overall guidance as well as critical input for the writing of this review. He also helped write the introduction and background sections. Dr. Nybo provided useful material and helped with the writing of sections 3.IV.I.I, 3.V.II and 3.VI.I.I. Ben

Woolston provided critical input as well as wrote the entirety of sections 3.V.III and 3.VII.I.II.

Although not first author, the vast majority of writing, figures and technical assembly of references was conducted by myself including the logistics of submission. A large fraction of the highly biological focus of initial writing by Eric nybo was displaced in various re-organizations, however, it was decided that his overall contribution to the Electrofuels effort and academic carrier focus warranted first authorship.

III. Introduction

The vast majority of biological activity on our planet derives its energy from the sun.

Heterotrophic organisms rely on the primary productivity of photosynthetic organisms to provide reduced carbon. This photosynthetic activity is divided into the process of photon capture to release high energy electrons (light reactions) and the channeling of those electrons to their use in the reduction of carbon dioxide (dark reactions). Electrons can be made available by a variety of sources (wind, hydroelectric, etc.) where the most analogous to photosynthesis is the production of electricity from photovoltaics. In Figure 3-1, this analogy is used to illustrate the concept of

‘replacing’ the light reactions of biological photosynthesis with photovoltaics and utilizing those electrons to reduce CO2. The path traversed by these electrons defines broad classes of microorganisms. Since free electrons are rapidly reacted, the reducing power is often shuttled by

39 other compounds. The prevalence of dihydrogen (H2) from numerous biological or geochemical reactions has provided motivation for biological utilization of this form of reducing power to fix

CO2, and can be viewed simplistically as a form of autotrophic growth where H2 is used as a substitute for light, and the dark reactions of CO2 fixation proceed as they do in photosynthetic organisms. Not surprisingly, nature has found numerous other pathways to fix CO2 (Berg 2011); however, the basic concept of capturing available reducing power into central metabolism remains the basis of autotrophic growth.

Figure 3-1. Analogy between direct photosynthetic growth and indirect photosynthetic growth based on the feeding of electrons as reducing power to reduce CO2. DEF: direct electron feeding, Fd: ferredoxin.

Included in the ‘dark reactions’ of Figure 3-1 is also the process of splitting water, which can be achieved external to the cell by electrolysis, or with the assistance of electron carriers that

40 might be reduced either inside or outside the cell. As will be discussed in more detail below, some organisms can directly accept the reducing power from an electrode (Nevin et al. 2010), and the nature of configuring bioreactors to accommodate this broad range of electron delivery becomes an integral component of an eventual process to use the reducing power within metabolically engineered organisms to produce biochemicals (conceptually depicted in Figure

3-2). The scope of metabolic engineering of autotrophs therefore spans from improving the ability of an organism to accept electrons into metabolism and fix CO2 to the introduction of biosynthetic pathways into this diverse class of organisms that already possesses chemolithoautotrophic metabolism. Since the metabolic engineering of facile heterotrophic organisms such as E. coli or yeast is highly advanced, it is tempting to pursue introducing autotrophic metabolism into these model chassis. This remains an exciting possibility, particularly if the host chassis is an extremophile or industrially robust organism, however this is currently limited by the tremendous complexity of both the associated enzyme systems and their regulation.

- I = V/ X-H Electron e Ohm’s law acceptors

Photosynthetic hv Beer‐Lambert law apparatus

H , CO , H2-ases, 2 2 carboxylating CO, O2 Henry’s law enzymes

Reactor Boundary Cell Boundary

Figure 3-2. An examination of the CO2 reduction reactions from the perspective of the mode of transport of the reducing power to either the reactor, or to the inside of the cells as a major

41 determinant of the application of these for biotechnological purposes. Light cannot penetrate dense cultures (Beer-Lambert Law), where gas transfer rates (GTR) face limitations of solubility in the liquid phase (Henry’s Law). Electrons are particularly challenging to deliver (Ohm’s Law) either to a reactor electrode or into the cell although this can be facilitated by the reduction of a ‘carrier’ (X-H).

IV. Background

By definition, autotrophic growth encompasses any organism that can ‘feed itself’ from non-organic carbon sources – namely CO2. Since there are microorganisms that can utilize essentially any reduced chemical as a component of their metabolism, it is useful to briefly review this classification to define the scope of this review. Aerobic photoautotrophs utilize light to extract hydrogen from water (Figure 3-3), and are not the focus of this review. Many organisms can utilize H2 including the anaerobic methanogens and acetogens, as well as aerobic hydrogen oxidizing bacteria that are often referred to as Knallgas bacteria (Schwartz et al. 2013).

The methanogens are excluded from this review because their ubiquitous presence in association with anoxic breakdown of organic material (sediments, wastewaster, animal gut) require its own specific review (Demirel & Scherer 2008; Thauer et al. 2008). Acetate production from CO2 and

H2 is of interest in this review in part because acetate can readily be converted to alternative biochemicals.

The aerobic hydrogen-oxidizing bacteria are of particular interest for this review because of their demonstrated potential as production platforms. The model organism R. eutropha (used here in preference to the new classification of Cupriavidus necator) has been the subject of study for over 100 years (Kaserer 1906). Its ability to accumulate biomass and poly-3-hydroxybutyrate

(PHB) to very high levels both heterotrophically (Ryu et al. 1997) and autotrophically (Tanaka et al. 1995) demonstrated its ability to be competitive with E. coli as a production platform.

Considerable efforts have been made to conduct scale-up and optimization studies. R. capsulatus

42 is also a noteworthy chemolithoautotrophic (M. Madigan & Gest 1979) candidate due to its study as a model organism and diverse growth modes (Hunter et al. 2009). Obligate anaerobic acetogens, like Clostridium ljungdahlii, are attractive hosts because of their ability to use synthesis gas (syngas) or CO2 and H2 for growth. C. ljungdahlii naturally produces acetate, ethanol, butanol, and 2,3-butanediol; companies such as INEOS, Coskata and LanzaTech are actively commercializing it for production of variety of chemicals from syngas (Köpke, Mihalcea,

Bromley, et al. 2011). A more in-depth examination of some of the characteristics of these model, sequenced organisms provides some additional perspectives on considerations for metabolic engineering of biochemicals production in these organisms. The autotrophs which can utilize

+ 2+ reduced inorganic compounds (NH4 , H2S, Fe ) are also of notable interest because of either utilization of an industrial waste stream or the potential of the substrate to undergo cyclic reduction at an anode as the basis of delivering the reducing power. Finally, the potential for direct electron feeding from a biocathode (sometimes referred to as microbial electrosynthesis) represents the ultimate example of delivery of reducing power from a renewable electrical generation plant to store that reducing power in chemical bonds of a microorganism (Lovley &

Nevin 2013).

hv Photoautotrophs CO2 CH2O H2O

H2 Knallgas bacteria CO2 CH2O

CO Acetogens CO2 CH2O H2

Nitrifying, Sulfur, Iron XH Oxidizing bacteria CO2 CH2O

e- Biocathode CO CH O 2 H O 2 2

Figure 3-3. A summary of autotrophic growth modes with the common theme of reducing CO2 to biomass (carbohydrate) using various forms of reducing power. These different autotrophic

43 growth modes are often associated with other classification names that are listed. These common name classifications are useful, but also lead to problems typical of generalizations.

3.IV.I. Carbon Fixation in Chemolithoautotrophs

3.IV.I.I Calvin-Benson-Bassham (CBB) Pathway and Rubisco

The predominant mechanism for CO2 capture in aerobic chemolithoautotrophs is via the photosynthetic dark reactions of the Calvin-Benson-Bassham (CBB) pathway (Figure 3-4). This pathway is present in many of the model autotrophic organisms including R. eutropha,

Rhodobacter, and some species of Nitrosomonas, and Acidithiobacillus. The first committed step of the CBB pathway is a carboxylation reaction involving CO2 and ribulose-1,5-bisphosphate

(RuBP) by ribulose bisphosphate carboxylase/oxygenase (Rubisco), which many autotrophs compartmentalize into organelles called carboxysomes. Regeneration of RuBP involves enzymes of central metabolism, a sedoheptulose bisphosphatase (SBPase), that converts a 7-carbon sedoheptulose-1,7-bisphosphate to sedoheptulose-7-P, and phosphoribulokinase (PRK), that converts ribulose-5-P to RuBP. Enzymes Rubisco, SBPase and PRK are considered to be unique to the CBB pathway. Rubisco is responsible for most of the CO2 captured on the planet and is the

CO2-fixing enzyme in the majority of aerobic chemolithoautotrophs. The extensive studies of energetic and catalytic efficiency of Rubiscos are of particular interest to metabolic engineering.

Three forms of Rubisco exist, which differ in catalytic efficiency (kcat) and CO2-specificity (SCO2)

–important properties which generally represent trade-offs in functionality. Variants of these forms are found in chemolithoautotrophic bacteria (Badger & Bek 2008). Sometimes more than one variant exist simultaneously in a single bacteria and are generally differentially regulated in response to CO2 concentration (Berg 2011).

44

An affinity for O2 is the biggest disadvantage of the Rubisco enzymes and causes the phenomenon called photorespiration in plants, which is the tendency of Rubisco to oxidize RuBP, rather than carboxylating it. The resulting need to ‘recover’ the oxidized product is energetically very expensive and involves loss of CO2 and consuming the amino acid glycine to give ammonium. Photorespiratory CO2 loss is estimated to be 21% of the net CO2 assimilation in plants (Berg 2011). In bacteria, the product of Rubisco oxidation is metabolized more efficiently than plants, where 2-phosphoglycolate is dephosphorylated to glycolate, which is either secreted from the cell or recycled back to central metabolism (Berg 2011). Both of these mechanisms waste carbon and energy, thus there is room for improvement in chemolithoautotrophic bacteria as well.

Figure 3-4. Schematic illustrating the diversity of the cbb operons including paralogs of the various accessory proteins in addition to the large and small subunit (cbbL/S). Form I Rubisco is denoted by cbbLS and Form II by cbbM, both of which are found in R. capsulatus and R. sphaeroides (Panel A&B). R. eutropha also contains two CBB operons, one on the chromosome (Panel C), and one on the megaplasmid pHG1 with a defective transcriptional activator (cbbR*, Panel D). The iron-oxidizing bacteria Acidithiobacillus ferooxidans contain four different CBB operons with two encoding Form I Rubisco subunits. Other genes are: cbbF=SBPase/FBPase, cbbP=PRK, cbbT=TKT; cbbA=FBA/SBA, cbbE=PPE, cbbG=GAPDH, cbbK=PGK, cbbZ=phosphoglycolate phosphatase), cbbBXYQ=unknown. The presence of these functional operons in plasmids provided the historical basis of elucidating autotrophic genes including CO2 fixation and hydrogen use, now adapted to metabolic engineering of these phenotypes in chemolithoautotrophs.

45

The Rubisco forms have evolved for optimized performance (by a balance of kcat and

SCO2) in their respective niches with varying degrees of CO2 and O2 availability; therefore the target platform organism may not have the appropriate properties for the intended bioreactor environment. Therefore the choice of the Rubisco form for metabolic engineering will be largely dictated by the O2 and CO2 operational conditions. For example, choice for a Form II Rubisco

(having lower SCO2 but higher kcat) can be made in an autotroph if high catalytic rate is desired by increasing CO2 and limiting O2 reactor concentrations. Carbon concentrating mechanisms (CCM) and carboxysomes found in bacteria are a reminder that the bioreactor operating conditions can be significantly modified by the microbial chassis. Rubisco and associated genes needed for functional CO2 fixation are nearly always located in cbb operons (Figure 3-4). These operons reflect that a functional CBB pathway requires at least four genes including a transcriptional regulator (CbbR) while at the same time a tremendous diversity has evolved. These carbon- fixing operons tend to be ‘complete’ by the principle of the ‘selfish operon’ (Lawrence & Roth

1996), where the likelihood of a gene being passed along to subsequent host is enhanced by the clustering of genes in an operon that encodes a complete function. This clustering of genes by function continues to be a significant asset to metabolic engineering efforts that has been accelerated by next generation sequencing and associated reverse genetics.

3.IV.I.II Other Carbon-fixation Pathways

Due to the diverse habitats of chemo-/photoautotrophic bacteria, various other CO2 fixation pathways have evolved. In addition to CBB, at least five other CO2 fixation pathways have been characterized (Figure 3-5): (1) the reductive tricarboxylic acid cycle (rTCA), (2) reductive acetyl-CoA or Wood-Ljungdahl pathway (rAC), (3) 3-hydroxypropionate bicycle

(3HP), (4) 3-hydroxypropionate-4-hydroxybutyrate cycle (HP/HB), and (5) dicarboxylate-4-

46 hydroxybutyrate cycle (DC/HB) (Berg 2011). rTCA cycle has a wide distribution in chemolithoautotrophic and photosynthetic bacteria. It involves running the TCA cycle in reverse.

Wood-Ljungdahl pathway is limited mostly to obligate anaerobic acetogens of Clostridial genera and methanogenic Archaea and fixes CO2 as well as CO via a linear pathway. Distribution of

3HP pathway is very limited to within green non-sulfur bacteria with Chloroflexus aurantiacus being the most heavily studied. The DC/HB and HP/HB are found in archaea - mostly aerobic/microaerobic thermophiles. Although the HB branch is common to both the cycles, the enzymes appear to have evolved independently of each other. With the interest in using CO2 as a substrate for biochemical production, these pathways are being intensely studied including efforts to enhance or even transplant these pathways to create new autotrophic chassis for metabolic engineering.

CBB CO2 3-Phosphoglycerate Ribulose-1,5-2P Ribulose-5-P 1

Glyceraldehyde-3-P

Sedoheptulose-1,7,-2P Sedoheptulose-7-P

Wood-Ljungdahl CO (acetogen) 2 5-Methyl- CO2 Corrinoid Citrate Pyruvate 6 Acetyl-CoA - HCO3 8 Citramalyl-CoA 5,10-Metylene-THF Glyoxylate Acetoacetyl-CoA L-Malyl-CoA Malonyl-CoA Methylmalyl-CoA

Phosphoenol 5,10-Metyenyl-THF cis-Aconitrate pyruvate Propionyl-CoA 3-HPA HCO - 3 8 7 - HCO3 4-Hydroxybutyrate 10-Formyl-THF (S)-Methylmalonyl-CoA

DC/HB Oxaloacetate Malate Succinyl-CoA Formate

HP/HB 3 5 4 rTCA Isocitrate 2-Oxoglutarate CO H CO2 CO2 2 2

Figure 3-5. Representation of various CO2-fixation pathways in relation to each other: CBB (black ≡), Wood-Ljungdahl (red =), Reverse TCA (rTCA; yellow - -), 3-hydroxypropionate (3-

47 HPA; blue —), 3-hydroxypropionate/4-hydroxybutyrate (HP/HB; green ▪▪▪), dicarboxylate/4- hydroxybutyrate (DC/HB; purple — ▪ ▪). The carboxylating enzymes are indicated by numbered circles: 1, ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco); 2, bifunctional carbon monoxide dehydrogenase/acetyl-CoA synthase (CODH/ACS); 3, formate dehydrogenase; 4, 2- oxoglutarate synthase (Fd-dependent); 5, isocitrate dehydrogenase; 6, pyruvate synthase; 7, phosphoenolpyruvate carboxylase; 8, bifunctional acetyl-CoA/propionyl-CoA carboxylase.

3.IV.II. Hydrogen Utilization by Chemolithoautotrophs

Hydrogenases are metal-containing enzymes that are common in bacteria (Vignais &

Billoud 2007). They are classified based on the H2-binding sites as the [FeFe] and [NiFe], with a third rare [Fe]-only type found only in methanogens (Schick et al. 2012). [FeFe] hydrogenases, have simple monomeric forms, consisting of only the catalytic subunit, which in principle would be easier to introduce into a heterologous host. [NiFe] hydrogenases are composed of small and large subunits, and generally have relatively large number of maturation and assembly accessory proteins. Hydrogenases are also subdivided as membrane bound (MBH) uptake or H2-evolving and bidirectional soluble (SH) forms (Vignais & Billoud 2007). The uptake MBH of interest for generating an autotroph is generally attached to cytochrome b complex and ultimately involved in

ATP generation. The SH contains subunits that bind to soluble cofactors such as NAD or NADP.

They are able to catalyze the evolution or consumption of H2, subsequently oxidizing or reducing the cofactors, depending on the physiological conditions (i.e. NAD(P)/NAD(P)H ratio, H2 partial pressure etc.). Some SH also act as H2 sensors, being part of the regulatory mechanism and not H2 activation. Further discussion of the details of hydrogenases is beyond the scope of this review and are thoroughly reviewed elsewhere (Lubitz et al. 2014). A schematic representation of an aerobic chemolithoautotroph in terms of energy generation, utilization, CO2-fixation and regulation is presented in Figure 3-6.

The genes for hydrogenase synthesis, assembly and maturation are usually conveniently organized onto operons which facilitates their heterologous expression (Rousset &

48 Liebgott 2014). In fact, Alcaligenes hydrogenophilus contains the hydrogenase activity (Hox+) on a mega-plasmid that allowed early identification of hydrogenase genes by mobilization into other species which conferred ability to grow on hydrogen (Yagi et al. 1986). The transitions of this work into the current bioinformatic era is discussed further in Section 4.3. These historical studies on the genetic transfer of the Hox+ phenotype, were followed by a period of intense study of hydrogenases for the production of bio-hydrogen (Lubitz et al. 2014; Rousset & Liebgott 2014), providing considerable genetic resources and experience in heterologous hydrogenase expression.

Figure 3-6. Schematic representation of carbon and energy metabolism, regulation and gene expression in a typical aerobic chemoautotroph. Both membrane bound (MBH) and soluble (SH)

49 hydrogenases generate reducing equivalents. MBH also channels electrons to the membrane- bound Electron Transport Chain for the production of ATP. SH facilitates ATP production via the NADH oxidoreductase (Complex I). The energy and reducing power are used by the carbon fixation pathway (typically CBB). The Regulatory Network is a complex interaction among external substrates as well as internal metabolites and cofactors, which eventually controls the transcription of the hydrogenase operons (hox, hup, hyp etc.) and CBB operons. There is direct regulation of these operons as well (e.g. H2).

3.IV.III. Microbial Electrosynthesis via Direct Electron Feeding

Where the ability of microorganisms to donate electrons to the anode of a microbial fuel cell (MFC) has been extensively studied, the acceptance of electrons at a biocathode to reduce

+ CO2 falls within the scope of this review. The ability of hydrogenase to reduce NAD mediated by electrode potential was demonstrated over 20 years ago (Cantet et al. 1992). Biohydrogen and biomethane production subsequently dominated the study of electrosynthesis where the typical feedstock included organic-laden wastewater streams. Microbial electrosynthesis (MES) of organic compounds is recognized and recently reviewed (Rabaey & Rozendal 2010; Lovley &

Nevin 2013). While most of microbial electron feeding has involved undefined consortia, (Nevin et al. 2011) recently reported a number of pure culture of acetogens (belonging to Spormusa,

Clostridium and Moorella genera) are capable of accepting electron from a cathode and fix CO2 into acetate (on the order of 0.1-1 mM over several days). Autotrophic electrosynthesis of acetate, methane, and hydrogen gas was recently reported by a microbial consortia (Marshall et al. 2012).

The cathode was populated predominantly by the archaebacteria Methanobacterium spp. and the eubacteria Acetobacterium spp., and peak yields of methane reached 7 mM per day, acetogenesis was over 4 mM per day (28.5 mM in total over 12 days), and hydrogen production was over 11 mM per day. Similarly, a mixed culture of bacteria achieved rates of 129 mL per day of CH4 and

94.72 mg/day of acetic acid, further demonstrating the capability for carbon fixation directly from

50 electrical power (Jiang et al. 2013). However, to date we are not aware of any genetically engineered microbial electrosynthetic system.

V. Metabolic Engineering of Natural Chemolithoautotrophic Organisms

The diversity of strategies for metabolic engineering in chemolithoautotrophs is as diverse as the range of autotrophic hosts. The status of genetic engineering is very organism- dependent with the two major categories of aerobic H2-oxidizing bacteria and anaerobic acetogens. The current focus of metabolic engineering study is largely defined by considerations for the level of basic science study, existing metabolism, genetic engineering tools and ease/danger of culture conditions (using H2 and CO). Model chemolithoautotrophic organisms including R. eutropha, Rhodobacter sp. and C. ljungdahlii, provide a useful basis for discussing key characteristics of autotrophic metabolisms. An assessment of performance under autotrophic conditions is of particular relevance, but is often not tested since these model organisms can grow heterotrophically.

3.V.I. Status and Constraints for Metabolic Engineering in Chemolithoautotrophs

R. eutropha is a model chemolithoautotroph that has very high aerobic autotrophic productivity when grown on H2, and accumulates large quantities of PHB where its knockout in principle could provide extensive flux within central metabolism via acetyl-CoA. Combined with relatively mature genetic engineering (transformation, chromosomal modifications, plasmids), this organism has been the basis of the full array of commodity biochemicals and biofuels. R. capsulatus is also robust hydrogen-oxidizing chemolithoautotroph that benefits from the

51 extensive study of its closely related species R. sphaeroides, which can also be coerced into chemolithoautotrophic growth as well (Paoli & Tabita 1998). The study of its diverse metabolism including anaerobic photoheterotrophy has resulted in extensive understanding of its basic physiology and genetics. Rhodobacter also has moderate capacity for PHB accumulation as well as isoprene metabolism associated with photosynthetic pigments. Due to the anaerobic nature of acetogens, these organisms display natural accumulation of fermentative end-products of industrial interest from CO2, including ethanol, butanol and even acetone and 2,3-propanediol. C. ljungdahlii has emerged as a front-runner for further development as a metabolically engineered biochemical platform in part since the genetic tools are becoming available relative to other candidates. Acetogenic bacteria also have a high autotrophic flux to acetyl-CoA which makes them attractive candidates for autotrophic production of a variety of fuels and chemicals (Hu et al.

2013).

This range of model organisms represents a gradual change of metabolism from highly

O2 tolerant to highly O2 sensitive. This distinction is quite important from the standpoint of metabolic engineering of chemolithoautotrophs due to energetic considerations. Anaerobic organisms are generally more efficient at energy capture because of the inherently smaller available thermodynamic free energy to drive heterologous metabolism (Bar-Even et al. 2012).

For this same reason the anaerobic rate of energy capture is quite constrained particularly for the production of highly reduced molecules (Fast & Papoutsakis 2012).

Acetogenesis from H2/CO2 proceeds via the stoichiometry

4H2 + 2CO2  CH3COOH + 2 H2O (ΔG = -95 kJ/mol acetate)

and is thus one of the most constrained modes of energy conservation known (Drake &

Daniel 2004). Energy is conserved through the Wood-Ljungdahl pathway by a combination of substrate level phosphorylation (production of acetate from acetyl-CoA) and chemiosmosis

(Mayer & Müller 2014). Diverting acetyl-CoA flux away from acetate production therefore puts

52 severe energy limitations on the cell, and stoichiometric models predict that this will drastically limit the yield of the target molecule (Fast & Papoutsakis 2012). This may explain why attempts to produce highly reduced chemicals through the Wood-Ljungdahl pathway to-date have met with limited titers and yields. Unlike genetic systems, which can be gradually improved and refined over time, these thermodynamic considerations are immutable, and may ultimately constrain the commercialization of these microbes.

On the other hand, aerobes are capable of capturing large amount energy from an electron donor because of the ability to use O2 as an electron acceptor, and thus drive high rates of productivity of energy-dense molecules at the expense of energy capture efficiency. In addition to the inherently less efficient energy capture in aerobic chemolithoautotrophs, the presence of O2 poses numerous additional problems (see Section 3.VI.I.II for further discussion). Almost all the key enzymes for CO2-fixation and H2-activation are either permanently or temporarily deactivated by O2 or have undesired side reactions that are energetically wasteful. Natural systems have solved this problem by evolving a balance among these antagonistic effects of O2.

Where light is available in high quantity, higher photoautotrophs use the abundance of ATP to circumvent the wasteful photorespiratory branch of photosynthesis and use carbon concentrating mechanisms to exclude O2 from reaction centers. Bacteria that do not have this option carry out a slightly more efficient (but still wasteful) cycling mechanism to counteract this non-specific reaction of Rubisco. Rubiscos and hydrogenases in nature have also evolved highly ‘optimized’ forms based on the organism’s natural habitats. Catalytically faster Rubiscos are less specific towards CO2. Catalytically faster [FeFe] hydrogenases are highly sensitive to O2 and deactivate permanently, while [NiFe] SH and MBH display gradual tolerance towards O2 and corresponding reduced catalytic activity. The goal of metabolic engineering is to develop organisms to produce chemicals at high rates of productivity and using low energy as both of these contribute towards the cost – biofuel being the most extreme example where both of these conditions have to be

53 simultaneously met. From this standpoint, the engineering needs to capture the best of both worlds – through genetic, metabolic and bioreactor engineering.

3.V.II. Metabolic Engineering in Aerobic Chemolithoautotrophs

The native biosynthetic capacity of R. eutropha to produce PHB at greater than 70% of dry weight, even in presence of carbon monoxide (Volova et al. 2002), has motivated the engineering of this organism for even higher levels of PHB and other biopolymers. Fukui and co- workers engineered non-native poly[(R)-3-hydroxybutyrate-co-3-hydroxypropionate] polyesters in R. eutropha by introducing malonyl-CoA reductase and 3-hydroxypropionyl-CoA synthetase genes from C. aurantiacus (Kichise et al. 1999). Most metabolic engineering strategies of heterologous pathways in R. eutropha have similarly incorporated some form of knock-out of the

PHB synthesis pathway, to divert a portion of that carbon flux. However, in recent work producing methyl ketones from a truncated lipid -oxidation pathway, there was surprisingly no enhancement observed for the ±knockout comparison, and the highest level of chemolithoautotrophic production was only 180 mg/L. Hydrocarbon biosynthesis engineering in

R. eutropha using decarboxylation of fatty acids also used -oxidation mutants, and was executed in a manner to explore promoters, origins of replication and ribosomal binding (Bi et al. 2013).

Where broad host range pBBR and arabinose-inducible PBAD outperformed other typical vectors, an unexpected combination of a high copy number mutant of pCM62 combined with a synthetic

RBS gave 6 mg/L in this generally low expression testing.

Only R. eutropha PHB knockout strains were tested under heterotrophic growth conditions in efforts to produce branched-chain isobutanol and 3-methyl-1-butanol production

(Lu et al. 2012). Batch production was over 300 mg/L and repeated media removal for 50 days achieved a cumulative production totaling 14 g/L. In a subsequent effort from the Sinskey

54 laboratory, the systematic optimization of isopropanol production demonstrated substantial effects for codon usage, and gene dosage and was able to achieve 3.44 g/L in batch cultures with less than 1 gDW/L. This represents one of the highest specific productivities achieved to date for metabolic engineering of R. eutropha (Grousseau et al. 2014). However, this was achieved on fructose, and remains to be demonstrated under chemolithoautotrophic conditions. As part of a demonstration of integrated electrolytic reduction of CO2 to formate (with subsequent uptake and formation of CO2 and NADH), Liao laboratory (Li et al. 2012) executed an autotrophic fermentation (80:10:10, H2, O2, CO2) to produce 1 g/L alcohols (~50:50 isobutanol:3-methyl-1- butanol). This work included a PHB knockout and the ‘Ehrlich pathway’ in conjunction with enhanced α-keto acid flux toward the branched valine/leucine amino acid synthesis pathway. The elevated production is partly assisted by the ability of autotrophic growth to achieve high cell concentrations of ~12 gDW/L (OD600=24).

Purple non-sulfur bacteria, such as R. sphaeroides and R. capsulatus, naturally synthesize large amounts of carotenoids as well as increased lipid metabolism during intracellular membrane production under O2/light limitation (Beekwilder et al. 2014b). Utilizing the endogenous

IPP/DMAPP flux of R. sphaeroides, valencene production (57 mg/L) was achieved by simply expressing Callitropsis nootkatensis valencene synthase (CnVS) which was 40-fold higher than the same construct in Saccharomyces cerevisiae. Further enhancement to 352 mg/L was observed upon introduction of the mevalonate (MVA) pathway from Paracoccus zeaxanthinifaciens, indicating that the CnVS enzyme was substrate limited. We took an alternative approach of enhancing the endogenous R. capsulatus methylerithrotol phosphate (MEP) pathway to produce the C30 triterpenes, squalene and botryococcene, using the triterpene synthases from the colony algae Botryococcus braunii race B (Niehaus et al. 2011). This was accomplished by expressing high activity avian FPS to catalyze conversion of IPP/DMAPP to farnesyl diphosphate (FPP), along with expressing the rate-limiting MEP enzymes 1-deoxy-D-xylulose 5-phosphate synthase

55 (DXS) and isoprenyl diphosphate isomerase (IDI) from pBBR vectors (Khan et al. n.d.). As observed for most other biochemicals tested, autotrophic production was comparable to screening results on complex heterotrophic media, indicating a robust metabolism under fully chemolithoautotrophc growth conditions. Autotrophic bioreactors achieved 110 mg/L when biomass reached 6.6 gDW/L and subsequent continuous operation stabilized at 60 mg/L (at 2.6 gDW/L). This represents a specific productivity of 0.5 mg·gDW-1·h-1 which is comparable to the average alcohol specific productivity of 0.67 mg∙gDW-1∙hr-1 (though this work of Liao displayed a growth-dissociated burst of 3.3 mg∙gDW-1∙hr-1 after 3 days of culture). These results are still an order of magnitude lower than the 93 mg∙gDW-1∙hr-1 isopropanol reported for growth on fructose

(Grousseau et al. 2014). Comparative interpretations of productivity are often difficult to interpret in the absence of details of bioreactor operational strategy, which is addressed in some additional detail in Section VIII of this review.

3.V.III. Metabolic Engineering in Acetogens

Metabolic engineering of acetogens is in its infancy, largely due to the limited genetic tools available for these microbes. In contrast, the native productivity of these organisms for producing their final electron acceptors is well beyond metabolic engineering efforts in aerobic chemolithoautotrophs. Since acetate is the precursor to a variety of important chemical compounds, for example polyvinyl acetates, the ability to generate it autotrophically instead of from petrochemicals continues to be developed, including the use of genetic engineering. For example, the acetogenic capacity of Acetobacteium woodii was recently improved by overexpression of native pathway enzymes. Transformation of plasmids carrying either the phosphotransacetylase (PTA) and acetate kinase (ACK) or the four THF-dependent enzymes in the Wood-Ljungdahl pathway was achieved by methylating these vectors using an E. coli strain

56 expressing the C. ljungdahlii DNA methylation genes prior to electroporation (Straub et al. 2014).

The PTA-ACK strain had significantly higher specific productivity of 0.9 g∙gDW-1∙hr-1 and final acetate titer 50.5 g/L compared to the already impressive ‘wild-type’ production strain performance (0.85 g∙gDW-1∙hr-1 and 44.7 g/L). In addition to acetate, syngas-fermenting organisms natively produce several other compounds, including ethanol, butyrate, butanol, and

2,3-butanediol (Liew et al. 2013; Daniell et al. 2012; Köpke, Mihalcea, Liew, et al. 2011). Three companies: Coskata, INEOS and LanzaTech are commercializing autotrophic ethanol production

(Schiel-Bengelsdorf & Dürre 2012). LanzaTech is also working on 2,3-butanediol production

(Köpke, Mihalcea, Liew, et al. 2011) and although no results are yet available in the literature, numerous patents have been filed in the last year by LanzaTech in this space.

The first example of an engineered acetogen for non-native production involved transplanting the butanol biosynthesis pathway from the heterotrophic acetogen Clostridium acetobutylicum into C. ljungdahlii via the pIMP1 shuttle vector. The recombinant strain produced

0.15 g/L (5 kJ/L) butanol from synthesis gas in mid growth-phase, but this was subsequently consumed by the end of the fermentation (Köpke et al. 2010). More recently Derek Lovley’s group improved the transformation protocol for C. ljungdahlii, reporting high efficiency replicating plasmids, as well as integrative vectors and the ability to generate knockouts (Leang et al. 2013). The genetic toolkit was further expanded by the discovery that the lactose-inducible promoter and glucuronidase reporter on the pAH2 vector originally designed for Clostridium perfringens functioned in C. ljungdahlii. The inducible system was used to drive expression of genes coding for acetone biosynthesis, leading to the successful production of acetone from syngas at a titer of ~0.87 g/L (26.8 kJ/L) (Banerjee et al. 2014), a substantial improvement over 9 mg/L acetone production previously obtained in C. aceticum (Schiel-Bengelsdorf & Dürre 2012).

Butyrate production in C. ljungdahlii was recently enhanced using plasmid-based expression of the 8-gene butyrate synthesis pathway from C. acetobutylicum resulted in 0.74 g/L

57 for H2/CO2 and 1.24 g/L for CO/CO2 (Ueki et al. 2014). By chromosomal integration of the butyrate pathway and shutting down Pta-dependent acetate, AdhE1-dependent ethanol and Ctf- dependent fatty acid synthesis pathways, the production improved to 1.35 g/L for H2/CO2 and

1.48 g/L for CO/CO2. A notable result of this effort is that chromosomal knockouts predicted to eliminate ethanol and acetate could not abolish accumulation, indicating the existence of additional unidentified genes/pathways for these competing molecules. In addition, for the initial plasmid introduction the acetate/ethanol levels were essentially unchanged from wild type at roughly 7 g/L, where the 0.64 g/L enhancement in butyrate was accompanied by a 5.9 g/L decline in these alternative acetate/ethanol fluxes. This indicates that much needs to be learned about the pathways and compensatory energetics of these organisms to effectively utilize them as biochemical production platforms.

Genetic systems have also recently been developed for the promising acetogen Moorella thermoacetica. First, a pre-methylated suicide vector was used to generate a uracil auxotroph by inactivation of the pyrF gene, selected for with 5-FOA. Uracil prototrophy was then restored along with integration of a lactate dehydrogenase gene behind the GAPDH promoter, that allowed production of lactate at 0.61 g/L from fructose (Kita et al. 2012). Later, a kanamycin resistance gene was integrated into the same locus, conferring resistance to 300 g/mL (Iwasaki et al. 2013). A functional antibiotic resistance marker opens up the ability to perform integration into any site within the genome. Earlier this year, Evonik (in collaboration with LanzaTech) announced the production of 2-HIBA, the precursor for Plexiglas, from syngas using an engineered strain (http://corporate.evonik.com/en/media/focus/Pages/bacteria-syngas.aspx).

Although the details of the engineering are not publically available, the work likely takes advantage of a recently discovered CoA-carbonyl mutase enzyme (Rohwerder & Müller 2010).

Recently, a series of studies by Tyurin and co-workers at Syngas Biofuels Energy, Inc. reported engineering several Clostridium sp. strains for the continuous high level production of

58 several heterologous compounds such as acetone, n-butanol, mevalonate etc. (Kiriukhin & Tyurin

2014; Berzin et al. 2013; Kiriukhin & Tyurin 2013; Berzin et al. 2012). By elimination of key processes in the cell, such as spore formation, acetate and acetaldehyde production, they have reported very high specific productivity of these molecules in continuous syngas or CO2/H2 fermentations of recombinant systems. However, strong doubts have been expressed on the validity of these results, and they should be considered with caution (Bengelsdorf et al. 2013).

3.V.IV. Metabolic Engineering in Other Chemolithotrophs

Nitrosomanas europaea is an example of a chemolithoautotroph that utilizes the reducing power of ammonium to reduce CO2. The Scott Banta group received an ARPA-E award to engineer isobutanol into N. europaea (Khunjar et al. 2012); however, this exemplified the challenge of working with non-model organisms. Although preliminary genetic tools were developed by the group and GFP expression and isobutanol production were achieved, the slow growth and excessive energetic penalty for transport and oxidation of ammonia by this bacteria ultimately resulted in this effort being abandoned in favor of an iron-oxidizing chemolithotroph

(personal communication). Utilizing the bioreactor concept of an electrochemical reactor to repeatedly reduce an electron carrier, this effort has transitioned to ferric-ferrous reduction at a cathode as basis of autotrophic CO2 reduction by Acidithiobacillus ferrooxidans (Li et al. 2014).

Although this work is still quite new with initial work demonstrating introduction of isobutyrate production (Banta & West 2014); the work will continue as an NSF-funded project with preliminary demonstration of autotrophic heptadecane hydrocarbon production.

59 VI. Using Metabolic Engineering to Improve / Creating New Chemolithoautotrophs

While genetic engineering in existing autotrophic systems has shown promise and demonstrated a wide range of compounds produced directly from CO2, the limitations of tools development combined with the likelihood of desirable characteristics for an industrial platform from a non-autotrophic host has motivated the alternative approach of improving or transplanting pathways for carbon and reducing power capture. Besides avoiding homologous metabolic regulation, heterologous expression often provides insights into scientific details obscured in the native host.

3.VI.I. Engineering Carbon Capture

3.VI.I.I Engineering CBB pathway

The limitations of rate and CO2/O2 specificity of Rubisco have made this the target for improving primary productivity in agriculture by extensive research efforts that are chronicled in many reviews (Mueller-Cajar & Whitney 2008). Although these efforts have largely ‘failed’ as stated in a recent evaluation “ … although the challenge of making a ‘better Rubisco’ has exceeded the grasp of career of many scientists” (Whitney et al. 2011) this must be qualified by understanding that the context of the goal of agricultural productivity is very different from the metabolic engineering goal that is the basis of this review. There is both tremendous natural diversity associated with evolution ranging from aerobic to anaerobic conditions, as well as presence / absence of carbon concentrating mechanisms. Screening assays have improved both kcat, affinity and specificity where tradeoffs have largely precluded improved overall performance. R. capsulatus has been used as a host to screen for improved Rubisco by creating a

60 large subunit deletion mutant with impaired photoautotrophic growth, that was then used in complementation studies with a mutated cyanobacterial Rubisco (Smith & Tabita 2003). The defined conditions of a bioreactor provide a clearly defined objective to exploit the extensive research on Rubisco. For example, an improved affinity for CO2 can allow a higher H2 partial pressure with associated improved gas transport rates. Similarly, there is an opportunity to manipulate wasteful RuBP oxidation through bioreactor operation and control, particularly under the continuous steady-state operation that is pragmatically the only means of achieving economic feasibility for an autotrophic process (Khan et al. 2014). This emphasizes the importance of integration of metabolic engineering efforts with bioprocess design. It seems highly likely that an improved Rubisco would be component of an autotrophic production platform.

In addition to Rubisco, other components of the CBB pathway can also be limiting. It was observed that overexpression of SBPase from the cyanobacteria Synechococcus and from

Arabidopsis thaliana in tobacco (Miyagawa et al. 2001; Lefebvre et al. 2005) and from

Chlamydomonas reinhardtii in Dunaliella bardawil (Fang et al. 2012) improved photosynthetic

CO2 fixation by 20-100%. A gene found on bacterial cbb operons, cbbZ, encodes for phosphoglycolate phosphatase (PGP) that dephosphorylates phospohglycolate (formed due to the oxidation of RuBP; analogous to photorespiration in plants). This allows glycolate to recycle back to CBB cycle via the D-glycerate pathway while also relieving phosphoglycolate inhibition of triosephosphate isomerase (Berg 2011). By defining a production platform organism and operating condition, the relevance of such potentially rate-limiting steps within the CBB pathway becomes a much clearer optimization objective.

61 3.VI.I.II Engineering Other Carbon Capture Pathways

Engineering the CBB pathway may not be the most logical choice from biotechnological perspective where kinetic and energetic efficiency are critical criteria. The CBB pathway in algae and plants evolved under conditions where the supply of ATP and NADPH were not typically limiting. In aerobic systems, robustness of CBB enzymes to O2 appears to be a dominant criterion, where co-evolution of carbon concentrating mechanisms further protected the kinetic shortcomings of this pathway. The five other known carbon fixation pathways have varying degrees of kinetic and energetic efficiencies and O2 tolerance (Bar-Even et al. 2012; Fast &

Papoutsakis 2012) and computational analysis suggests that superior pathways may be possible based on several quantitative criteria (such as pathway specific activity, ATP and NADPH cost, number and compatibility of the enzymes with existing network etc.) These existing alternative pathways have also been subject to the various natural selective pressures for their respective niche (Berg 2011), which may require modifications for optimal biotechnological application.

Reconstituting a functional heterologous carbon fixation pathway in a non-autotrophic organism will be the next major step towards creating a synthetic chemolithoautotrophic production platform. The 3-HPA bicycle (from Chloroflexus aurantiacus) has been largely installed into into E. coli by Pam Silver laboratory (Mattozzi et al. 2013). Although each component of the pathway was demonstrated to be functional, the complete pathway failed to achieve autotrophic growth. By introduction of a portion of the HP/HB pathway from

Metallosphaera sedula into the hyperthermophile Pyrococcus furiosus, partial chemolithoautotrophy was conferred to this archaeon to produce 3-hydroxypropionic acid, which represents a major advance toward a process-rationalized host rather than a convenient genetic model (Keller et al. 2013). As an example of altering carbon fluxes through expression of heterologous CO2 fixation pathways, an “autotrophic bypass” was engineered into yeast by

62 introducing Rubisco and PRK from the chemolithoautotroph Thiobacillus denitrificans

(Guadalupe-Medina et al. 2013). The associated CO2 fixation created an electron sink to decrease glycerol production by 90% and increase ethanol production by 10%. These recent results, combined with the broadened emphasis toward using biological metabolism of CO2 as a component of reducing GHG emissions is poised to utilize metabolic engineering to create more carbon efficient biochemical production processes.

3.VI.II. Engineering improved H2-utilization

In contrast to the challenge of transplanting CO2 fixation, conferring the ability to utilize

H2 is comparatively easy, with the initial successes preceding modern genomics as a result of the transmission of plasmids containing complete functional hydrogen utilization operons (Umeda et al. 1986). This does not mean that hydrogenase function is by any means simple, as it involves metal incorporation and maturation accessory proteins that vary considerably by hydrogenase type and organism (Lubitz et al. 2014). The role of metabolic engineering is parsing through the available hydrogenase diversity, including the focus on biohydrogen production, to identify the appropriate platform candidate.

In chemolithoautotrophic bacteria, oxidation of molecular H2 provides the reducing equivalents and ATP needed for CO2-fixation. The mechanism for energy conservation in MBH are generally less efficient than SH. SH can provide the reducing equivalents directly to the redox carriers such as NAD(P) or ferredoxin (Fd) and ATP is generated via proton-motive force. On the other hand, although ATP generation is facilitated more directly by MBH through membrane linked electron transport chain, less efficient reverse electron transport is generally required to generate reducing equivalents (Vignais & Billoud 2007). A practical example of the relative effect of SH and MBH is R. eutropha, where doubling time increased significantly for SH

63 knockout mutants (12 h vs. 3.6 h for wild-type) while only slightly for MBH mutants, indicating that reductant supply is growth-limiting. Furthermore, transfer of R. eutropha SH genes into

Pseudomonas facilis (containing MBH) reduced its doubling time from 12 to 9 h (Friedrich &

Schwartz 1993). This is likely also reflected in apparent H2 yield of R. eutropha of 6.4 gDW/mol-

H2 (Bongers 1970) versus only 2.7 for R. capsulatus (Siegel & Ollis 1984). Energetic efficiency is also affected the specific redox carrier used. For example, Fd is more efficient than others because of its lower reduction potential (Bar-Even et al. 2012).

The catalytic properties of hydrogenases are of relatively minor importance compared to carboxylating enzymes where some approach ‘kinetic perfection’ (Berg et al. 2002), being limited by the diffusion rates of H2 (see Error! Reference source not found. for rates). On the other hand O2 tolerance is an important factor, [NiFe] MBH are the most O2 tolerant, followed respectively by [NiFe] SH and [FeFe] hydrogenases, with the catalytic properties generally reflecting an evolution that varies inversely with their oxygen tolerance. Therefore, choices of hydrogenases and cofactors are ultimately dictated by the consideration of whether the chemolithoautotrophic platform is being designed for aerobic or anaerobic conditions. Although

[FeFe] hydrogenases are easier to genetically engineer (due to relatively small number of maturation factors) and are also generally faster and more efficient, both the hydrogenases and the cofactor Fd are very sensitive to O2. However, if bioreactor operating conditions permit maintaining low dissolved O2 (e.g. to avoid explosion hazard), there may be added advantage to using [FeFe] versus [NiFe] hydrogenases.

Much of the work related to engineering hydrogenases in heterologous hosts was motivated by the production of hydrogen (Rousset & Liebgott 2014) rather than the creation of better autotrophic organisms. None-the-less, the tools, and in many cases, the outcomes of these efforts are equally useful for improvement to organisms as autotrophic production platforms. In fact, at bioreactor operating conditions where partial pressure of H2 is anticipated to be above 0.5

64 atm (Bongers 1970), it is likely that most hydrogenases will act as uptake hydrogenases, because at typical ratios of cellular redox pairs (NAD(P)H/NAD(P) or Fdred/Fdox) the equilibrium partial pressure for H2 uptake appears to be quite low (<1000 Pa) (Veit et al. 2008). Therefore, studies seeking to increase the efficiency of H2 production via improved O2 tolerance (Rousset &

Liebgott 2014) or channeling substrate and reductant flow (Kontur et al. 2012) may directly translate to enhanced H2 uptake rates.

Table 3-1. Catalytic properties of various hydrogenases (compiled from BRENDA). Organism Type of Substrate Redox Km Kcat (1/s) Km/kcat

-1 -1 H2ase partner (mM) (M s )

6 R. eutropha [NiFe] H2 0.037 187.1±108.4 5.06x10 soluble hydrogenase NAD+ 0.293 126±24 4.3x105 NADH 0.048 196.5±36 4.11x106

R. opacus H2 0.063 - - NAD+ 0.137 - -

- 9 C. [FeFe] H2 Fd 3.3x10 133 4.03x10 acetobutylicum soluble 5 hydrogenase

- 9 H2 Flavodoxin 8.8x10 483 5.49x10 5

3.VI.III. Metabolic Engineering of Alternative Electron Delivery Pathways

Although there is no demonstration of an engineered organism directly accepting electrons from a cathode to produce non-native chemicals, components of this system have been demonstrated separately, namely electron delivery to autotrophic organisms (Nevin et al. 2011) that have been genetically modified to produce various compounds (Ueki et al. 2014). However,

65 the volumetric rates of product formation are extremely slow (compared to the performance of the same organisms in other systems), likely due to the very small densities (depending on the surface area of the electrode only) and/or the rate of electron delivery to the cells. The mechanism for electron delivery to the cell is not yet fully understood. Genetic engineering has been used to improve the electron transfer in microbial fuel cells (MFC) via overexpression of electron carriers or synthesis of novel electron carriers (Sydow et al. 2014), an approach which may be applicable to microbial electrosynthesis. A summary of the status, challenges and potential for improving the electrosynthesis platform is available (Rabaey & Rozendal 2010). The choice and tradeoffs between aerobic versus anaerobic is reiterated in terms of productivity and yield. The field is very new, however, with much to be learned about the potential of microbial electrosynthesis for the autotrophic production of fuels and chemicals.

VII. Metabolic Engineering Toolbox for Existing Autotrophs

For those trained in molecular techniques using only facile organisms such as E. coli, transitioning to work with most autotrophs requires patience, persistence and attention to detail as there are fewer tools and methods are less forgiving and reproducible. As details were largely omitted in the review of metabolic engineering advances (Section 3), the critical importance of the metabolic engineering toolbox warrants some additional discussion.

3.VII.I. Genetic Transformation

Transformation by mating and counter-selection against the conjugal donor is still a dominant method utilized for transformation of both Ralstonia and Rhodobacter. Relative

66 to heat shock or electroporation in E. coli, this method increases the time required for a given transformation by five or more. Transformation procedures also often require a period of developing mastery, which adds additional delay in achieving results that would otherwise seem rather mundane. Although recent successes provide methods to achieve metabolic engineering in chemolithoautotrophs, these tools are still under active development, where new methods are slow to be adopted. It is therefore useful to compile some of the recent advances for manipulating chemolithoautotrophic organisms – particularly for acetogens that have only very recently yielded to reliable transformation.

3.VII.I.I Genetic Transformation of Aerobic Chemolithoautotrophs

Relatively routine techniques are available for transformation of Ralstonia and

Rhodobacter. Both species are amenable to conjugal plasmid transformation, electroporation and homologous chromosomal transformation. One of the seminal developments in engineering a wide range of chemolithoautotrophs was the establishment of intergeneric conjugation from specialized E. coli donor strains in a bi- or tri-parental mating (Ditta et al. 1985). These systems take advantage of broad host range vectors that feature an origin of transfer (oriT), which can be mobilized from E. coli strains harboring the RK2 mobilization machinery, either integrated into the chromosome (i.e. E. coli S17-1, diparental mating) or harbored on the pRK2013 plasmid,

(triparental mating). The recipient strain preferably exhibits a drug resistance to counter-select the

E. coli host, such as rifampicin resistance for R. capsulatus SB1003 or gentamicin for R. eutropha. This adds considerable additional constraints on vector design. Homologous recombination in gram-negative bacteria has been available for over 20 years utilizing the efficient sacB counter-selection (Quandt & Hynes 1993). These methods provide both for deletion, gene modification and insertion chromosomal insertion (Jaschke et al. 2011) as noted

67 for the most recent high-level expression systems based on the T7 polymerase vectors described in the next section.

One of the ongoing challenges concerning genetic engineering of chemolithoautotrophs is the recalcitrance of many host strains to electroporation of foreign DNA, which if successful would circumvent the need for time-consuming conjugation protocols. Improved vectors that utilize electroporation continue to be developed for R. eutrophoa (Solaiman et al. 2010).

Restriction systems are believed to be responsible for recalcitrance by rapidly degrading foreign

DNA. Recently, the rsh1 endonuclease gene of R. sphaeroides was knocked out, which subsequently made the strain amenable to electroporation (Jun et al. 2013), a method which is described in more detail below for acetogens.

3.VII.I.II Genetic Transformation of Acetogens

The primary limiting factor for acetogen metabolic engineering is the lack of robust genetic tools. Transformation in these organisms is frequently hindered by extensive restriction modification systems (Leang et al. 2013; Straub et al. 2014; Kita et al. 2012; Tsukahara et al.

2014). There are a number of strategies to combat this. In a strategy named Plasmid Artificial

Modification Systems (PAMS), the transformation vector is purified from an E. coli host heterologously expressing the native methyltransferases, increasing transformation efficiency

(Yasui et al. 2009; Suzuki & Yasui 2011). The methylation domains of these enzymes are readily identified bioinformatically, thus the strategy can be implemented based solely on the sequenced genome of the target organism. Alternatively, the specific methylation pattern, or ‘methylome’ of the target organism can be established via Single Molecule Real Time (SMRT) sequencing, and then avoiding these sites in the design of transformation vectors (Murray et al. 2012; Flusberg et al. 2010). Even when DNA digestion has been abated, the plasmid must either successfully

68 replicate in the host, or integrate into the genome. Toward the first aim, a modular Clostridia plasmid system, with multiple origins of replication and antibiotic resistance markers has been generated by the Minton lab (Heap et al. 2009) that facilitates rapid screening of potential vectors.

Integration via the universal homologous recombination mechanism occurs with low efficiency in

Clostridia (Al-Hinai et al. 2012), hypothesized to be partially due to the absence of the resolvase enzyme in many of these species (Rocha et al. 2005). This scheme as well as generally implementing knock-outs and knock-ins is limited by the availability and effectiveness of negatively selectable markers. A summary of the breadth of transformation methods currently being used for chemolithotrophs is provided as Supplemental Table S2.

In extensive work with homologous recombination in C. ljundalhii, the Lovley lab has noted an inability to get any of the three negative selections to work (e.g. pyrF, galK, mazF)

(Ueki et al. 2014) thereby preventing marker reuse of the limited selectable markers using the

Cre-LOX excistion procedures (Marx & Lidstrom 2002). Despite the noted limitations, transformation systems now exist for the three model acetogens: Moorella thermoacetica,

Clostriidum ljungdahlii, and Acetobacter woodii, facilitating both metabolic engineering and studies to expand and improve the genetic systems.

3.VII.II. Plasmids, Promoters and Vectors of Particular Utility for Chemolithoautotrophs

The most heavily utilized broad host range plasmids for engineering of gram-negative bacteria are the IncP-based pRK plasmids (Ditta et al. 1985) and the pBBR1MCS vectors

(Kovach et al. 1995). Most IncP-based vectors rely on light-sensitive tetracycline selection, although pRKD418 employs a tetrahydrofolate reductase gene for trimethoprim resistance

(Mather et al. 1995). The largest drawback of most plasmid-based expression in a production

69 environment is the need for antibiotic supplementation for constant selection pressure. Recently, the narrow-host range plasmid, pIND4, was maintained in R. sphaeroides without selection pressure (Ind et al. 2009). This vector incorporates several attractive features, including an IPTG- inducible lacIq cassette and the repressible hybrid PA1/A0/A3 promoter and stable MG160 replicon

(Inui et al. 2003). This plasmid has also been used to build a BioBricks® toolbox for R. sphaeroides with expression levels (of DsRed fluorescent protein) comparable to an E. coli Plac system (Tikh et al. 2014). In our hands, pIND4 displayed poor stability in R. capsulatus; we observed that R. capsulatus transformants appeared to express the kanamycin marker poorly and replacement of KanR with the neomycin phosphotransferase of pBBR1MCS-2 doubled the kanamycin resistance phenotype to a level of 20-25 μg/mL suggesting that this plasmid might be improved for the more robust chemolithoautotroph R. capsulatus which would very useful for dual plasmid testing of operons.

A hoxA-mediated ‘plasmid addiction’ system was recently developed specifically for R. eutropha lithoautotrophic growth (Lütte et al. 2012). The chromosomal gene for the HoxA hydrogenase operon transcriptional regulator is essential for utilization of hydrogen. The transgenes of interest are provided on the pLO11 plasmid which complements the hoxA- mutation and renders the plasmid indispensable, thereby eliminating the need for a selectable marker under chemolithoautotrophic conditions. A similar obligatory plasmid system based on the KDPG aldolase (eda) system (Voss & Steinbüchel 2006) is not applicable to chemolithoautotrophic growth because the enzyme, KDPG aldolase, is part of the Entner-Douderoff pathway which is only active during heterotrophic growth – although this demonstrates the generality of the approach. The mobilization efficiency and stability of four different plasmid systems (based on incompatibility groups IncP, IncQ, IncW and pBBR) capable of replicating in R. eutropha were recently assessed (Gruber et al. 2014). They showed that RP4 and RSF1010 based mobilization sequences are respectively about 50000 and 5000 times more efficient than pBBR1 mobilization

70 sequence. Notably, the inclusion of a 2.3 kb par region of RP4 onto other plasmids nearly completely prevented plasmid loss for up to 96 hours in the absence of antibiotic selection. This par region apparently has broad utility for plasmid retention as it previously conferred a high degree of stability to E. coli plasmids in absence of antibiotic selection for up to 200 generations

(Gerlitz et al. 1990). It is now believed that par region encodes for proteins that stabilize plasmid retention based on site-specific recombination that can resolve plasmid multimers. A similar new high-stability plasmid for R. eutropha has been generated by cloning oriV28 and parABS28 regions from pMOL plasmid of Ralstonia metallidurans into an E. coli cloning vector which also permitted transformation by electroporation (Sato et al. 2013). The relative copy number for the four plasmids having common mob and par regions was assessed and the pBBR based vector was found to be about 2 and 4 times higher than those of RSF1010 and pSa respectively.

Early promoter engineering efforts in R. eutropha developed native promoters to drive expression of acetoin, polyhydroxyalkanoate, and pyruvate biosynthetic genes (Delamarre & Batt

2006). While most of these fusions with an artificial polyhydroyalkanoate operon succeeded in producing PHA, the overall yields were less than 10%, reflecting an innate limitation of this method. Orthogonal promoters for R. eutropha gene expression under chemolithoautotrophic conditions have the potential to bypass the host’s innate genetic regulation machinery, where many of the transcription initiation signals were adapted from E. coli. In comparison of heterologous promoters it was found that the T5 phage-derived promoter Pj5 enhanced expression in R. eutropha by 4 to 5-fold relative to Ptac and Plac (Gruber et al. 2014). The PcbbL promoter which is induced under autotrophic conditions (Kusian et al. 1995), was effective in increasing flux to the amino acid oligomer cyanophycin (Lütte et al. 2012).

The T7 RNA polymerase (T7 RNAP) system that is used to drive high level expression in

E. coli vectors has been adopted to R. eutropha and R. capsulatus, although most metabolic engineering efforts may not require ultra-high protein expression. In R. eutropha the T7 RNAP

71 gene was placed under the control of phaP promoter of the PHA biopolymer operon and the oph gene for organophosphohydrolase (OPH) target protein under the control of T7 promoter

(Barnard et al. 2004). This approach expressed OPH at 6% to 18% of total soluble protein. A similar T7 RNAP expression system has been developed for R. capsulatus resulting in upwards of 80 mg/L of yellow fluorescent protein (Katzke et al. 2010).

The anhydrotetracycline-inducible promoter system based on high-affinity binding to the tetracycline repressor/operator (TetR/O; Lutz and Bujard 1997), has recently been adapted for tunable gene expression in R. eutropha by screening for variable expression of TetR (Li & Liao

2014). An order of magnitude dynamic range of tunable expression was achieved. Although this is two orders of magnitude smaller range than the analogous E. coli system, it was sufficient to alleviate toxicity associated with an intermediate accumulation in the isobutanol metabolic engineering in R. eutropha. These recent advances in genetic toolbox development should accelerate progress in metabolic engineering which is just beginning to tap into the potential of chemolithoautotrophic organism as illustrated in the next section.

VIII. Status of productivities and scale-up considerations for metabolic engineering

Design of metabolic engineering strategies are affected by scale-up considerations, where there are additional challenges for autotrophs as compared to traditional organisms. Downstream factors such as product separation, genetic stability, induction of gene expression and growth/non-growth associated product formation, all affect the choices for metabolic engineering approach. The choice of platform organism is also quite critical and is coupled with the energetic efficiency (growth yield and maintenance coefficients) (Khan et al. 2014), aforementioned scale- up considerations and various other factors such as product toxicity. Many of these issues relate

72 specifically to the target product and are beyond the scope of this review. In this section we will briefly discuss the performance of autotrophs during bioreactor culturing and the relevance to metabolic engineering.

Since R. eutropha was known to produce large amounts of PHB, early developments of autotrophic bioreactor design to achieve high density and titer focused on this organism. A limited number of recent studies addressed scale-up of engineered lines as well (see Error!

Reference source not found.). The volumetric and specific productivities reflect the performance of the bioreactor and the cells respectively. This compilation represents the best productivities reported in the literature for autotrophic hosts for the conditions noted and includes native production of metabolites as a comparative benchmark. E. coli hosting the PHB pathway is also included for comparison.

The production of native compounds by autotrophs represents the high capacity of these organisms for producing biochemicals from CO2. The volumetric productivity reflects not only the biosynthetic capacity, but the permissiveness of the organism to be cultured at high density. It is interesting to note that for autotrophic culture conditions the continuous provision of carbon from CO2 is analogous to heterotrophic fed-batch and the maximum biomass cell densities represent non-carbon limitations on growth. It is noteworthy that the highly productive PHB (Ahn et al. 2000) and ethanol (Dumsday et al. 1999) could be largely replicated in E. coli for which the most advanced metabolic engineering tools exist. By comparison, metabolic engineering efforts are generally two to three orders of magnitude lower in volumetric and specific productivity in autotrophs. This is true even in the high density (~25 gDW/L) autotrophic productivity of R. eutropha for cyanophycin (Lütte et al. 2012). This situation is also generally not improved for the metabolically engineered strains even for heterotrophic conditions, suggesting that the current ability to manipulate fluxes in autotrophs is still tapping into only a fraction of the potential of these organisms. This is expected, since in contrast to heterotrophic metabolism,

73 chemolithoautotrophic metabolism is much more complex due to sensitive coupling of energy generation and utilization. Thus more global perspectives are needed while designing metabolic strategies. Strategies that integrate the target product formation tightly with the existing cellular machinery for energy generation and regulation, while eliminating non-essential competing pathways are much more likely to succeed.

In an effort to place these comparisons on a more common basis, the incorporation of carbon into natural and metabolically engineered metabolites is presented in Figure 6. The time courses represent best case data presented in the literature with the case of PHB (Tanaka et al.

1995) outpacing other chemolithoautotrophs by a large margin (and comparable to the best results of even E. coli). Acetate and ethanol production by M. thermoacetica and C. ljungdahlii are also quite impressive as compared to engineered metabolites (Hu et al. 2013). The best case for metabolically engineered butyrate production is still very low. Figure 3-7 also presents a final reminder of the fundamental differences between aerobic and anaerobic chemolithoautotrophic organisms as production platforms, where much more hydrogen is required to fuel the rapid production rates – resulting in a much lower theoretical energetic yield. However, at this time the capability of metabolic engineering in autotrophs is so far from realizing these yield differences that making such arguments for excluding one chemolithoautotrophic platform over the other is not warranted.

74 Table 3-2. Comparison of productivities of different compounds in various autotrophic and non- autotrophic hosts. Productivities have been calculated using data reported in the cited works.

PS, avg Density, PV, avg Mode of (max) biomass Substrate Organism Compound (max) Reference operation [g/gDW (product) [g/L-h] -h] [g/L]

Autotrophic hosts – engineered systems

Methyl 0.0017 0.0019 (Müller et H /O /CO R. eutropha Batch 1* 2 2 2 ketones (0.0036) (0.0046) al, 2013) (Lütte et al, H /O /CO R. eutropha Cyanophycin Fed-batch 0.0097 0.00041 25.4* 2 2 2 2012) R. Botryococcene Continuou 0.001 0.0003 (Khan et al, H2/O2/CO2 7 (0.11) capsulatus (C30H50) s (0.0023) (0.0006) 2015) C. (Ueki et al, CO/CO Butyrate Batch 0.011 0.028 0.8* 2 ljungdahlii 2014) 0.56 0.28 (Straub et al, H /CO A. woodii Acetate Fed-batch 2 (51) 2 2 (1.2) (0.9) 2014) 0.24 0.0026 (Barnard et Glucose R. eutropha OPH Fed-batch 113 (0.31) (0.0034) al, 2004) (Grousseau Fructose R. eutropha Isopropanol Batch 0.076 0.093 1.25* et al, 2014) R. Valencene (Beekwilder Yeast extract Batch 0.0049 - - sphaeroides (C15H24) et al, 2014) Fructose/Yea C. (Banerjee et Acetone Batch 0.016 0.019 0.885 st extract ljungdahlii al, 2014)

Autotrophic hosts – non-engineered systems

0.05 91.3 (Tanaka et H /O /CO R. eutropha PHB Fed-batch 1.55 (5) 2 2 2 (0.18) (61.9) al, 1995) M. 0.14 (Hu et al, CO/CO thermoaceti Acetate Fed-batch 0.55 3.92 2 (0.4) 2013) ca C. Continuou (Phillips et H /CO/CO Ethanol 0.322 0.184 4 2 2 ljungdahlii s al, 1993) 3.14 0.06 (Ryu et al, Glucose R. eutropha PHB Fed-batch 281 () (9.6) (0.14) 1997) Alcaligenes 4.94 0.38 (Wang & Sucrose PHB Fed-batch 111.7 () latus (5.13) (0.9) Lee, 1997)

75

Model, non-autotrophic systems

E. coli 194.1 (Choi et al, Glucose (LB) PHB Fed-batch 4.63 0.088 (engineered) (141.6) 1998) E. coli Continuou (Dumsday et Glucose (LB) Ethanol 0.57 0.18 3.2* (engineered) s al, 1999) S. cerevisiae (Bayrock & Continuou Glucose (non- Ethanol 12.7 1.41 9 (106) Ingledew, s engineered) 2005)

PV = volumetric productivity, PS = specific productivity. Non-engineered systems indicate production of native compounds by hosts that have not been otherwise genetically engineered. Engineered systems indicate production of heterologous compounds by hosts that have been genetically engineered. Non-autotrophic, engineered systems indicate production of heterologous compounds by non-autotrophic hosts that have been genetically engineered. * Estimated using a conversion factor of 0.5 g/L/OD

A B Native Engineered Substrate: H , CO +/- O productivities productivities 2 2 2 40 1 0.5 2

I 0.4 VI 0.8 30 III 0.6 0.3 fixed/L 20 - fixed/L - C fixed/mol H C fixed/mol

II gC 0.4 0.2 - gC

10 0.1 mol

0.2 PHB Ethanol Butanol Acetate Ethanol Butanol

IV Acetate V 0 0 0 Anaerobic Aerobic 0 200 400 6000 100 200 (Wood-Ljungdahl pathway) 1 (CBB pathway) Time (hr)

Figure 3-7. (A) Maximum reported productivities in grams of carbon fixed in products per liter (gC-fixed/L) in autotrophic systems. Native pathways: (I) PHB by R. eutropha, (Tanaka et al, 1995); (II) acetate by M. thermoacetica, (Hu et al, 2013); (III) ethanol by C. ljungdahlii, (Phillips et al, 1993). Engineered pathways: (IV) botryococcene by R. capsulatus, (Khan et al., 2015), (V) methyl ketone by R. eutropha (Müller et al, 2013); (VI) butyrate by C. ljungdahlii, (Ueki et al, 2014). All are autotrophic timecourses adapted from literature reports presented in Table 3. (B) Maximum theoretical yield (using values reported by Fast & Papoutsakis, 2012) on H2 of acetate, ethanol, PHB and butanol using aerobic CBB and anaerobic Wood-Ljungdahl (acetogen) pathways.

76 IX. Conclusion

The status of metabolic engineering of chemolithoautotrophs is very dependent on the specific organism where the genetic tools range from reasonably well-developed to non- existent. The results of metabolic engineering efforts can be generally characterized as falling short of tapping into the native fluxes of these organisms by several orders of magnitude. On one hand the genetic tools developed in model organisms have provided for rapid advances of methods, where opportunities to take advantage of unique characteristics of chemolithoautotrophic behavior such as the Hox-addiction plasmid have been developed, but have limited use since the majority of metabolic engineering testing is carried out under the convenient heterotrophic conditions. At this early stage, metabolic engineering strategies are understandably driven by proof-of-concept studies. But the next level of strategies and strain development must be designed with autotrophic operating conditions and scale-up in mind and performances tested in autotrophic conditions.

While transplanting H2-utilization to more facile model organisms is straight-forward, the goal of conferring either CO2-fixation or enhanced direct electron uptake has not yet been reported. Creation of an ‘optimal’ chemolithoautotrophic production platform should include considerations for bioreactor operating conditions where the enzymes chosen can reflect achieving high gas transport rates. Oxygen concentrations are of particular importance because of this and appears to be one of the dominant evolutionary drivers for the key chemolithoautotrophic enzymes. The efficiency advantage of anaerobes appears to come with an inherent constraint on rates of metabolism, the impact of which is also dependent upon operations strategy where continuous high density culture provides the highest productivity at minimum cost.

Chemolithoautotrophic systems have the potential to utilize an inexpensive carbon source as well as interface to thermochemical deconstruction of renewable biomass with associated

77 environmental benefits. This opportunity will continue to drive commercialization efforts and overcome the scientific and technical challenges presented by these useful metabolic pathways.

Chapter 4

Triterpene hydrocarbon production engineered into a metabolically versatile host – R. capsulatus

I. Preface

This chapter presents genetic engineering of R. capsulatus to produce squalene and botryococcene. Genes of triterpene hydrocarbon biosynthesis – squalene and botryococcene synthase, were obtained from the ancient algae B. braunii. R. capsulatus was used as the host because of its existing mechanisms for terpenoid synthesis and use, as well as its capability to grow in diverse trophic modes and substrates – utilizing carbohydrate, sunlight or hydrogen (with

CO2 fixation) as alternative energy feedstocks. Engineering an enhanced MEP pathway was also used to augment triterpene accumulation. The hetero-, photo- and autotrophic growth modes were tested in small-scale (<50 mL) cultures. Fed-batch heterotrophic and continuous autotrophic bioreactor studies were performed in order to test the performance of the strains under scaled-up

(1 and 5 L) heterotrophic and autotrophic conditions. Productivities were found to be improved under the controlled bioreactor conditions, autotrophic conditions in fact being the best. This work is accepted for publication in the journal Biotechnology and bioengineering.

II. Specific contributions

All the growth studies – including small-scale screening in the different trophic modes and reactor operations – batch and continuous, hydrocarbon extraction and analyses were done by myself. William Muzika helped set up some of the growth and reactor capabilities as well as help

79 with sampling. I have also facilitated the writing, figures, technical assembly of references, logistics of submission and revision of the manuscript. Dr. S. Eric Nybo generated the gene constructs used in this work. Dr. Alex Rajangam and Stephanie Tran helped with the conjugation of the constructs into R. capsulatus.

III. Introduction

Feedstock flexibility is being targeted to improve the economic feasibility and sustainability of the production of biofuels and chemicals. Corn-ethanol and the ongoing transitioning to cellulosic-ethanol or biobutanol are examples of allowing plants to capture photonic energy in chemical bonds, and subsequently selectively releasing that energy to produce a more convenient liquid transportation fuel. Similarly, the production of biofuels by algae reflects an even more direct use of the sun that relies on the storage of triacylglycerol lipids by these photosynthetic organisms. These oils can then be converted to biodiesel with minimum processing requirements. The use of these plant-derived energy sources, is inherently competitive with their alternative use as food and adds an additional dimension to the interaction of energy and food production costs. A more desirable alternative would be a host organism capable of chemical production from a variety of energy and carbon sources. To demonstrate this, we have targeted the production of two C30 hydrocarbons, potential fuel and chemical – botryococcene and squalene, in R. capsulatus – a host capable of diverse metabolism and robust growth.

Nature has devised extensive options to tap into the energy of the sun, with equally diverse carbon balances. This presents alternative approaches for biofuels production, where

Figure 4-1 illustrates the metabolic scenarios addressed in the presented work. Heterotrophic growth is the familiar consumption of carbohydrates, where aerobic growth takes advantage of metabolic efficiencies to produce CO2 and water. When oxygen is constrained, anaerobic

80 heterotrophic metabolism can result in the production of alcohols and organic acids and dihydrogen gas. Photoheterotrophic metabolism utilizes a carbon source such as malic or succinic acid from the TCA cycle and achieves tremendous carbon use efficiency by generating most of its energy by ATP production from anaerobic photosynthesis. The final R. capsulatus trophism illustrated in Figure 4-1 is aerobic chemolithotrophic metabolism, where the photonic energy generates hydrogen (via electrolytic splitting of water), and this energy can then be used to fix carbon dioxide. In nature the CO2 and H2 can be derived from anaerobic metabolism.

Figure 4-1. R. capsulatus is a metabolically diverse organism that can utilize and interconvert the energy provided from the sun in a variety of different ways. Metabolism within the dashed line representing Rhodobacter cell are those reported in this study: aerobic chemoautrophic growth consumes H2 and O2 while fixing carbon, with these gases being provided by photovoltaic-driven electrolysis in this scenario (or other natural sources). Aerobic heterotrophic growth is the typical consumption of sugars and associated aerobic respiration, anaeobic photoheterotrophic growth consumes energy poor organic acids with photosystem-II mediated ATP generation using light

81 energy. Depiction of the evolution (green arrows) or uptake (red arrows) of metabolic gases emphasizes the bioreactor requirements for gas-exchange as a major constraint for process design and operation. The lower part of the figure depicts the metabolic engineering strategy within the Rhodobacter cell to achieve the production of the high energy terpenes botryococcene or squalene. The operon includes the enzyme that commits carbon flux to the MEP pathway: (1- Deoxy-D-xylulose 5-phosphate synthase, dxs) and the isopentenyl diphosphate (IPP) isomerase (idi) to enhance the ratio of IPP to dimethylallyl pyrophosphate (DMAPP). This DNA construct also includes the gene for farnesyl diphosphate synthase (fps) which utilizes one DMAPP and two IPP to produce C15 farnesyl pyrophosphate, the substrate for the terpene synthases (botryococcene synthase, BS; squalene synthase, SS).

The choice of the biofuel product molecule is the critically important complement to the opportunities presented by feedstock flexibility. Our work focuses on the triterpene hydrocarbons

(C30+), squalene and botryococcene, because of their higher energy density than ethanol (A. Tuerk

2011) and ability to be processed nearly identically to that of crude oil (Hillen et al. 1982). Most notable from a process design perspective is that the hydrocarbon fuels are sufficiently hydrophobic to phase-separate, thereby eliminating the energy-intensive distillation step that constrains ethanol production processes. The specific choice of the fuel molecule botryococcene from the colonial algae Botryococcus braunii race B was purposeful because this is one of a very few biological derived molecules known to directly contribute up to 1% of all oil shales (Derenne et al. 1997; Glikson et al. 1989). In its native algae host, botryococcene oils are unique as compared to other photosynthetic algae associated with biofuels production (Metzger & Largeau

2005). These oils are not for energy storage; their apparent use for floatation results in their production to exceeding high levels > 30% of the carbon budget that is independent of the growth rate (Khatri et al. 2014). As a result, they accumulate at all stages of growth, and not only during storage conditions that characterize most algae biofuels scenarios. We previously isolated the genes for the biosynthesis of botryococcene and characterized their unusual biosynthetic mechanisms (Wu et al. 2012; Niehaus et al. 2011). This understanding has led to the generation of the functional chimeric botyrococcene synthase (BS) gene used in this work. The more direct biosynthetic route to squalene, another linear triterpene analog to botryococcene, was also

82 compared using the squalene synthase (SS) gene from the same B. braunii host as illustrated in

Figure 4-1.

The work presented here combines the concept of feedstock flexibility with the production of advanced hydrocarbon biofuels. The multi-trophic utilization of feedstocks for hydrocarbon biofuels production introduces constraints and opportunities for bioprocess engineering. Within the production bioreactor, gas exchange is a dominant consideration due to the low solubility of gases in aqueous solution and the energy required to transport these gases across the interface. The familiar aerobic heterotrophic reactor has oxygen uptake and CO2 evolution, which is the opposite of the oxygenic photosynthetic algae. The less familiar anaerobic photoheterotrophic condition has minimal gas transport considerations due to only minor CO2 evolution, but has a large photon flux requirement. Aerobic chemolithoautotrophic metabolism fixes CO2, while also taking up H2 and O2, and therefore has the greatest requirements for gas transport of the presented options. At the same time, since gas transport is unidirectional (into the culture), chemolithoautotrophic growth can be performed with 100% conversion of the gas phase (Bongers 1970), and thereby enables the most efficient utilization of the feedstocks presented. Our choice of the purple photosynthetic bacteria R. capsulatus was thereby planned to give great latitude in examining diverse modes of feedstock metabolism in the same organism for triterpene production.

In this study, we generated genetic constructs to produce C30 squalene and botryococcene in R. capsulatus that are shown to have sufficient activity to be limited by substrate availability.

We then characterize the productivity for these triterpene oils for the three trophic modes of aerobic heterotrophic, anaerobic photoheterotrophic and aerobic chemoautotrophic metabolism.

Finally, we demonstrate the ability to further amplify productivity by operating under fed-batch and continuous production conditions in heterotrophic and autotrophic stirred tank bioreactors.

83 IV. Results

4.IV.I. Achieving Substrate-Limited Triterpene Biosynthetic Constructs

To probe the productivity of R. capsulatus in various trophic modes, codon- optimized genetic constructs were assembled based on the triterpene target and known rate- limiting steps in biosynthesis of terpenes. Our recent work on botryococcene biosynthesis revealed a remarkable biochemical evolution where the triterpene botryococcene is synthesized by two separate enzyme homologs (Niehaus et al. 2011). This led to the generation of a chimeric fusion of these two sequentially acting enzymes, which provided the functional botryococcene synthase (BS) used in this work. For comparative purposes, the B. braunii squalene synthase (SS) is included which utilizes the same precursor as BS – farnesyl diphosphate (FPP) (Figure 4-1).

The avian FPP synthase (fps) gene was included in these constructs to ensure the robust production of the common FPP substrate (Wu et al. 2006). The genes encoding for the microbial equivalent of 1-deoxy-D-xylulose 5-phosphate synthase (dxs) and isoprenyl diphosphate isomerase (idi) were also included because of their known enhancement of carbon flux in the

MEP pathway and to improve the stoichiometry for FPP formation (Miller et al. 2000).

The initial construction and functional testing of these constructs was carried out in E. coli, which achieved successively higher titers when the dxs, idi and fps genes were added to the constructs (Figure 4-2). Their performance in R. capsulatus for both squalene and botryococcene in heterotrophic mode is shown in Figure 4-3, confirming a requirement for fps, and the expected improved performance of the native squalene enzyme. However, the addition of fps was not sufficient to pull large carbon flux out of central metabolism. This limitation was relieved when the rate-limiting enzymes from the MEP pathway - dxs and idi- were added.

Productivity increased by 20 to 100 fold over fps-only constructs. Further increases in

84 productivity of 1.5 to 2 fold were observed when the medium (YCC) was supplemented with 80 mM glucose (Figure 4-3b). This demonstrated that the introduced triterpene synthase activities were sufficient to be limited by general metabolic rates within central metabolism. The constructs containing the triterpene synthase and the enhanced MEP pathway components

(pBBR:Plac::BS/SS-dxs-idi-fps) were further utilized to probe for major differences in available flux from central metabolism under different trophic modes of growth.

Figure 4-2. Engineering of triterpene synthases with the native or engineered MEP pathways of E. coli DH5α. Constructs were engineered as a single polycistron under the control of the PLac promoter with triterpene synthase only (BS, botryococcene synthase; or SS, squalene synthase); in combination with a prenyltransferase (fps); or with one plasmid-borne copy of the engineered MEP pathway (dxs-idi-fps) or two plasmid-borne copies. Five biological replicates of each strain were grown aerobically for 24 hours in 2xYT-1% glycerol and then extracted with 1:1 acetone and hexane to determine triterpene content.

85

a b Std. +Glu medium suppl. PLac BS Botryococcene

Squalene

PLac BS FPS

PLac BS DXS IDI FPS

PLac SS

PLac SS FPS

PLac SS DXS IDI FPS

0 5 10 15 mg Triterpnes/g DW

Figure 4-3. R. capsulatus triterpene metabolic engineering. (a) Constructs containing botryococcene synthase (BS, upper panel) or squalene synthase (SS, lower panel) with increasing enhancements in metabolic flux by including the avian farnesyl pyrophospate synthase (fps), and putative rate-limiting enzymes in the MEP pathway. (b) Accumulation of triterpenes in R. capsulatus grown in standard growth medium (Std. medium) and that supplemented with an additional carbohydrate source, 80 mM glucose (+Glu suppl.) for the adjacent constructs: botryococcene for the upper panel, and squalene for the lower panel. The level of enhancement beyond standard growth medium alone is indicated by the extended bars (open bars). Experimental values were determined from five replicate cultures; error bars depict standard error of the mean.

The specific productivity at stationary phase for the different trophic growth modes were compared in small scale (<50 mL) cultures (Figure 4-5a). Photoheterotrophic growth mode improved the transition to autotrophic growth, through induction of the Calvin-Benson-Bassham pathway for carbon fixation (McKinlay & Harwood 2010), and therefore sequentially preceded autotrophic growth. The production levels paralleled growth over time (Figure 4-4) and were

86 found to be quite similar for all three trophic growth modes for either triterpene product, with squalene levels being two-fold higher than the botryococcene levels. Squalene productivity was greater than botryococcene probably because of the enzymatic efficiencies of the enzymes.

Conversion of FPP to squalene is catalyzed by a single active site in the enzyme, where the conversion of FPP to PSPP to squalene occurs without the substrate leaving the active site. In contrast, BS chimeric enzyme requires two catalytic sites, one converting FPP to PSPP, the second, PSPP to botryococcene, which would entail the release of PSPP from the active site of one enzyme and its re-binding at the second active site. These mechanisms provide an explanation for the different rates observed at similar substrate levels. Since there is a corresponding reduced accumulation of botryococcene, a degree of feedback regulation known to occur for the endogenous MEP pathway (Banerjee & Sharkey 2014) appears to be functional in conjunction with the heterologous pathway enhancements.

4 BS+Op3 )

1 BS - 2 mg L mg yococcene (

Botr 0

SS+Op3 SS )

1 4 - (mg L (mg

Squalene 2

0 0 20 40 60 80 100 120 140 Time (hours)

Figure 4-4. Shake flask autotrophic time-course of R. capsulatus harboring pBBR:Plac:BS-dxs-idi- fps and pBBR:Plac:BS only (Top panel) and pBBR:Plac:SS-dxs-idi-fps and pBBR:Plac:SS only (Bottom panel).

87 Independent of the precise rate-limiting mechanisms, obtaining similar productivities in the different trophic modes was unexpected, because the three trophic conditions represent very different physiological conditions, each having different energetic requirements. The differences for typical growth conditions in these diverse trophic modes had less impact than simple glucose supplementation. This result contrasts with the dramatically different metabolic modes (Figure

4-5b) that have been quantified in the “trophism snapshots” compiled in Figure 4-5c. The trophisms were quantified in terms of their carbon balance, with net release of about 50% of the carbon to CO2 in aerobic heterotrophic mode, and net capture of 100% of the CO2 in the autotrophic mode. By comparison, about 80% of the carbon from malate ends up in the biomass for photoheterotrophic metabolism, since the ATP generation is facilitated largely by light and not from malate catabolism. The contrasting energetic features of these trophic modes are depicted in the energy ‘source’ and ‘sink’ values. ‘Source’ was calculated as the efficiency with which input energy is captured as ATP. These inputs were sugar, light and hydrogen for heterotrophic, anaerobic photoheterotrophic and aerobic autotrophic modes of growth respectively. To place these numbers on a common carbon basis with ‘sink’ energy requirements, a conversion factor of 2.28 mol-ATP/mol-biomass-carbon was used (A. Tuerk 2011). The basis of these calculations is presented in Appendix E. The most important features of these calculations is that the sugar, malate and H2 are all products of photonic energy produced externally to the production host for those alternative trophic production modes (Figure 4-1). The comparable results for these dramatically different metabolisms and their associated fluxes of carbon and reducing power suggest a strong homeostatic regulation of the precursor pools of the central metabolism that is largely independent of the trophic growth mode.

88

CO2 Energy a b c Balance (kJ/mol C)

Botryococcene g-CO2 (g-C biomass) Source Sink Squalene -100% 0% +100% 0 250 500 -100 0 100

Aerobic (CH O) + O  HC + CO + H O (CH2O)n Heterotrophic 2 n 2 2 2

photons malate Anaerobic Photo- (RCOOH)  HC + CO heterotrophic 2

H2 CO2 Aerobic Chemo- CO + H + O  HC + H O autotrophic 2 2 2 2

0 5 0 5 10 Reductant Assimilation (e- → ATP) (C → pyruvate) mg triterpene / gDW

Figure 4-5. Exploring alternative trophisms for hydrocarbon production in Rhodobacter. (a) Triterpene specific productivity of R. capsulatus with pBBR:BS-dxs-idi-fps (blue-left panel), and pBBR: SS-dxs-idi-fps (red-right panel). (b) Reaction for hydrocarbon (HC) production for given tropism. (c) Trophims are presented quantitatively in terms of thermodynamic energy required to assimilate a mole of carbon into pyruvate as well as the energy required to produce an ATP normalized to a molar carbon basis. The carbon balance is shown to illustrate the relative magnitude of carbon produced by aerobic and photo-heterotrophic growth as compared to autotrophic CO2 assimilation.

4.IV.II. Productivity Enhancement through Bioreactor Operational Strategies

To determine if the growth-associated specific productivity could be maintained while increasing culture density through fed-batch addition of carbon and nitrogen, a typical feeding strategy of concentrated glucose feed (540 g/L) based on dissolved oxygen (DO) feedback control was implemented (Gleiser & Bauer 1981). NH4OH (8 N) feed was also provided for pH control while providing nitrogen (Figure 4-6). The heterotrophic bioreactor culture reached an OD660 of 12.7 and accumulated a total of about 40 mg/L botryococcene with an average specific titer (over the entire run) of 11.7±3.0 mg/gDW. This represents a 2-fold improvement over small scale YCC cultures with glucose supplementation, which indicates controlled pH and higher aeration significantly improved the substrate utilization of the cells to

89 generate a greater relative flux in the MEP pathway. However, we expected to reach a much higher density, but repeatedly encountered abrupt growth cessation at relatively low culture densities of less than 5 gDW/L, which are over an order-of-magnitude lower than E. coli cultures that routinely reach densities upwards of 100 gDW/L (Lee 1996). We suspect that this may be a result of some form of metabolic inhibition or a quorum sensing behavior that Rhodobacter is known to display (Schaefer et al. 2002; Curtis 2011). To test further, aliquots of media from the reactor stationary phase were tested for continued growth in subsequent batch flask culture

(Figure 4-7). Cultures inoculated into fresh media grew robustly, whereas cultures provided with concentrated nutrients from the stationary phase bioreactor supernatant displayed minimal growth. More work would be required to determine if this is typical homoserine lactone quorum sensing behavior or a different metabolic inhibition.

45 16

12 30 ) 660 8 15 Growth (OD Growth

4 (mg/L) Triterpene a 0 0

100 Gas phase O2 increased b 50 2 g/L) ( 50 25 Cumulative Dissolved O Dissolved glucose fed fed glucose (% of air saturation) (% of air 0 0 0 50 100 Time (hr)

Figure 4-6. Heterotrophic bioreactor growth of R. capsulatus pBBR:Plac:BS-dxs-idi-fps. R. capsulatus genetically engineered with the above expression vector was grown in a 5 L BioFlo

90 with glucose provided as the sole initial carbon and energy source, followed by fed-batch supplementation based on feedback control of dissolved oxygen (DO). Ammonia was also added in a fed-batch manner for pH control. (a) Growth monitored as OD660 (black circles) and botryococcene accumulation (blue squares) respectively. (b) The DO (red squares) and cumulative glucose supplementation (black line), where glucose fed batch was initiated at ~40 hr, and the approximate times for incremental increases in O2 supplementation (light blue arrows) are noted. Cultures stopped growing and accumulating hydrocarbon at about 70 hr; respiration slowed considerably in stationary phase as indicated by the sharp rise in DO which could not be recovered with additional glucose feed.

14

a b 12 c d 10 e f 8 g h

OD660 OD660 6

4

2

0 0 50 100 150 Time (hr)

Figure 4-7. Growth tests with heterotrophic bioreactor culture and culture supernatant. (a and b) Replicate samples from the bioreactor inoculated into RCVB (-malate) minimal media supplemented with 10 g/L glucose. (c) Sample from the bioreactor inoculated into YCC complex medium supplemented with 10 g/L glucose. (d) Fresh culture of R. capsulatus pBBR:Plac::BS- dxs-idi-fps inoculated into bioreactor supernatant supplemented with concentrated YCC medium nutrients (to make the final concentration similar to YCC). (e and f) Replicate fresh cultures of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into bioreactor supernatant supplemented with concentrated RCVB (-malate) medium nutrients (to make the final concentration similar to RCVB). (g) Fresh culture of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into RCVB (- mal) medium supplemented with 27 g/L glucose. (h) Fresh culture of R. capsulatus pBBR:Plac::BS-dxs-idi-fps inoculated into RCVB (-mal) medium supplemented with 10 g/L glucose. These tests were performed in order to confirm if something accumulating in the

91 heterotrophic bioreactor prevented further growth of R. capsulatus. Freshly growing cultures of R. capsulatus with the said construct inoculated into reactor supernatant supplemented with concentrated media components failed to grow, while reactor cultures inoculated into fresh media (both complex and defined) displayed proper growth. Freshly growing cultures inoculated into fresh media with high and low glucose concentrations emulating the reactor media are provided as control and displayed proper growth.

Sustainable biofuel production is dependent on CO2 fixation and we were interested in evaluating Rhodobacter’s capacity to produce triterpene under chemoautotrophic growth. To sufficiently characterize the metabolic capacity of the R. capsulatus platform for triterpene production, we scaled up to a 1 L autotrophic bioreactor. R. capsulatus pBBR:Plac::BS- dxs-idi-fps was grown in batch and continuous flow operation with gas phase H2, O2, CO2 feeding and NH4OH based pH control (Figure 4-8). It was found to be much more robust in terms of growth and triterpene accumulation in the scaled-up autotrophic reactor compared to heterotrophic. The inlet gas composition was manually adjusted via precise mass flow controllers to promote the high gas transport rate into the liquid (Figure 4-8a). The culture reached an OD660 of about 17 (7 gDW/L), comparable to that attained for heterotrophic growth, yet accumulated a total of about 110 mg/L botryococcene (Figure 4-8b), more than double that observed with heterotrophic reactor runs. Interestingly, in contrast to the heterotrophic bioreactor, there was a steady increase in specific triterpene level during the course of batch growth up to about 16.7 mg/gDW (Figure 4-8c). The specific productivity of botryococcene calculated during batch growth ranged between 0.17 and 0.5 mg/gDW-hr. Because steady state operation unambiguously quantifies the specific productivity, we switched the reactor to continuous operation by flowing

12.5 g/hr CA medium (inorganic salts and vitamins) into the reactor and removing culture at the same rate. This corresponded to a dilution rate of 9.62x10-3 hr-1, roughly 10% of the maximum

-1 growth rate of R. capsulatus (µmax = 0.112 hr ). The culture density came to steady-state around

2.75 gDW/L and the specific triterpene level was found to increase further and became relatively constant at about 23 mg/gDW. The inlet gas composition was kept constant during continuous

92 operation (Figure 4-8a) to allow the culture to reach steady state. Constant outlet gas composition was obtained when the culture reached a steady state (data not shown). The steady-state triterpene specific productivity was found to be 0.32 mg/gDW-hr, thereby achieving significant improvements based on all productivity measures as compared to heterotrophic small scale and reactor cultures.

a 1 0.4 H2 0.8

0.3 2 0.6 O2 0.2 CO 2 0.4 H

0.1 and 0.2 CO2 2

0 0 O Concentration Concentration

20 100 b ) 80

660 15 (mg / L) (mg OD ( 60 10

40

Growth 5 20

BATCH CONTINUOUS Botryococcene 0 0 c

0.5 20 ) 0.4 15 ) 0.3 c titer c DW

10 g gDW∙hr 0.2

Productivity 5 (mg/ (mg/ 0.1 Titer Specifi 0 0 Specific Productivity Specific 0 50 100 150 200 Time (hr)

Figure 4-8. Performance of R. capsulatus pBBR:PLac:BS-dxs-idi-fps in an autotrophic bioreactor. R. capsulatus expressing this plasmid was grown with gas phase feeding of H2, O2 and CO2 (for growth) and liquid phase feeding of ammonia (for pH control) in batch operation for the first 110 hours and under continuous operation after that. Continuous flow was established with 12.5 g/L flow of the CA medium, which corresponds to 10% of µmax of R. capsulatus. (a) The inlet gas composition profile. During the batch growth mode, the compositions were varied to accommodate the growth demand, based on outlet gas composition measurement. Gas consumption did not vary greatly during continuous flow and the inlet composition was

93 essentially kept constant. (b) The OD660 and botryococcene profile during batch and continuous operation. Cells grew exponentially to OD 5 and linearly after that reaching a maximum of 17. Botryoccocene accumulation increased proportionally as the cells and achieved a maximum volumetric accumulation of 110 mg L-1 botryococcene at the end of batch growth. Under continuous operation, the cells reached a steady state within about 130 hrs of starting flow, reaching an OD of ~7 and maintaining a relatively constant level 60 mg L-1 botryococcene. (c) The specific botryococcene productivity profiles. The specific botryococcene levels increased steadily during batch growth to about 17 mg/gDW. Batch specific productivities were around 0.5 mg/gDW-hr. Specific botryococcene continued to increase during the continuous flow and reached 23 mg/gDW. However, specific productivity decreased somewhat and reached a steady level of about 0.3 mg/gDW-hr.

V. Discussion

Although biofuels like ethanol and other small chain alcohols are readily obtained from genetically modified microbial hosts supplied with photosynthetic derived sugars and cellulosic feedstocks, questions remain about the sustainability and efficiency of such processes, including carbon balance and net energy generated. These observations have led to efforts aimed at coupling the production of higher energy fuel molecules to a variety of solar and geochemical energy sources. Utilizing the considerable knowledge that has become available about the bacterial MEP pathway and the biosynthesis of high-energy triterpenes, hydrocarbon biosynthetic pathways were introduced into R. capsulatus such that alterations in substrate availability within central metabolism through glucose feeding were reflected in enhanced triterpene accumulation. In choosing R. capsulatus there was a hope that the large differences in carotenoid metabolism in these different growth modes – particularly photoheterotrophic – that might provide a more favorable trophism for terpenoid flux. However, the observation that the levels of triterpene production were similar for fundamentally different modes of growth suggested a persistent degree of ‘homeostasis’ existed where the pools for metabolites of central metabolism were comparable between the different trophic modes of growth and no one mode was more advantageous relative to another for hydrocarbon production. It is possible that some of

94 the inherent potential of the alternative trophisms is obscured by endogenous regulation of the

MEP pathway, and work is planned to follow up by testing a heterologous MVA (mevalonate) pathway that should lack such regulation. This strategy has been successful for the production of various isoprenoids in E. coil (Kim et al. 2006; Martin et al. 2003).

Productivity can be enhanced considerably through bioreactor operational strategies which increase cell density while maintaining a relatively fixed triterpene content.

However, this is largely limited to an order of magnitude enhancement and compliments improvements in metabolic engineering strategies. A comparison of this work to other efforts at terpene hydrocarbon metabolic engineering is very informative as compiled in Table 1, most of which are based on complex carbon source or sugars. The maximum specific productivity in this work of 0.5 mg/gDW/hr is on par with several lycopene (C40) tetraterpene efforts, however, the distinct advantage of high-throughput screening for colored compounds has allowed for selecting production levels of both lycopene and carotene to about 3 mg/gDW/hr. By comparison, the production rates of ethanol are on the order of 500 mg/gDW/hr (estimated from (Ohta et al.

1991)), which illustrates the challenge of competing with a pathway that is a natural obligatory final electron acceptor. What is most surprising, is a comparison to the native Botryococcus braunii algae, where recent continuous culture work has provided very accurate specific productivity levels at 4 mg/gDW/hr (Khatri et al. 2014) which represents photosynthetic conditions from an organism that is notoriously slow growing (doubling times of 4+ days). This terpeneoid flux has only been exceeded by the recent work using a modified MVA pathway to produce C5 isoprene at 9 mg/gDW/hr in E. coli fed glucose and beef extract (Yang et al. 2012).

Combined with the comparative achievements of modest production levels, suggests that there are critical aspects of metabolic engineering of hydrocarbons that remain to be discovered. It is also interesting that the native algae biochemistry produces nearly exclusively tetra-methyl- botryococcene. This suggests that there is subcellular management of carbon flux that will likely

95 require intricate recapitulation in heterologous hosts to achieve high stable hydrocarbon production. Although organisms with alternative trophic modalities are more difficult to work with experimentally, the results presented here demonstrate a comparable potential for biochemicals production that do not rely on sugar as a carbon source.

96 Chapter 5

Ongoing and future work and conclusion

I. Triterpene production in R. eutropha

We have prioritized R. capsulatus for our initial studies of triterpene production using enhanced MEP pathway because of its already existing enzymes for terpenoid synthesis and use

(in light-harvesting machinery) as well the ability to grow in diverse trophic modes and substrates. This was described in Chapter 4. R. eutropha is another chemolithoautotroph that is very widely studied due to its robust growth and natural biopolymer (PHB) production. R. capsulatus and R. eutropha are, however, very different in their metabolism. The former is more widely classified as a facultative anaerobic phototroph, while the latter as a classic “knall-gas” bacteria. The most recent studies are the exploration of triterpene production in R. eutropha in order to carry out a comparison between the two bacteria in terms of their ability to produce C30 hydrocarbons. These studies are not yet completed at the time of submission of the dissertation but is expected that they will be soon after and generate sufficient data for a publication.

5.I.I. R. eutropha PHB-

For the initial set of studies I chose a strain of R. eutropha that is knocked-out in its ability to produce PHB (R. eutropha PHB-). PHB is a storage compound for these bacteria. Once a nutrient limitation other than carbon (nitrogen or phosphorous for example) is encountered, R. eutropha channels its carbon flux into PHB, for use later as a carbon source as other nutrients become available. In removing its ability to make PHB during nutrient limitation, it is expected that there will be additional flux to channel towards our target triterpene production.

97 First, to observe the triterpene production without any enhancement in the pathway, R. eutropha was transformed with pRK:Plac:SS-fps. This is a construct containing gene for farnesyl pyrophosphate synthase FPS (converting IPP and DMAPP into FPP) and the triterpene synthase

SS (converting FPP into squalene) (see Figure 1-3 for more details). With only these genes, there was small amount of triterpene production (Figure 5-5). These levels are very similar to those observed with R. capsulatus using the same construct but in a different vector backbone (i.e. pBBR:Plac:SS-fps). Supplementation of fructose clearly improved production, as a result of both increased biomass and increased substrate availability (shown later).

3

2.5

2

1.5

lene (mg/L)lene 1 Squa 0.5

0 LB LB+Fru 5g/L Medium

Figure 5-1. Triterpene production by R. eutropha transformed with pRK:Plac:SS-fps. The three different colored bars indicate three different clones.

To study its performance under controlled conditions, the best performing line (based on

Figure 5-1) was grown in a bioreactor with heterotrophic carbon sources (Figure 5-2). Figure

5-2A shows growth on defined medium and Figure 5-2B shows growth on LB medium, both supplemented with fructose (fed-batch). Their growth performance was not very good, that is they did not reach the high densities normally reached according to reports for growth of the wild-type strain of R. eutropha (Ryu et al. 1997). In both the current bioreactor efforts growth ceased midway through the run and it is not clear at this point what caused this. It is very likely

98 because of the PHB knockout that limited its ability to regulate internal redox. It was observed that pH changes had very strong effect on growth. Any time acid or base was added to stabilize pH, growth would cease (apparent from the subsequent rise in DO).

12 pH 100 A OD600 Squalene 90 10 DO 80 70 8 60

ne (mg/L)ne 6 50 40 DO (%) Squale

, 4 30 20

OD600 2 , 10 pH 0 0 0 20 40 60 80 Time (hr)

25 100 B 90 20 80 70 15 pH 60 OD600 Squalene 50 Squalene Squalene (mg/L) 10 DO 40 (%) DO pH, pH,

, 30 5 20 OD600 10 0 0 0 20 40 60 80 100 Time (hr)

Figure 5-2. Heterotrophic reactor growth of R. eutropha pRK:Plac:SS-fps. (A) Defined medium fed-batch with fructose. (B) LB medium fed-batch with fructose.

99 5.I.II. R. eutropha wild-type

In comparison to the PHB- lines the wild-type without any knockout, performed much better, both in terms of total and per-cell productivity. Figure 5-3 is a comparison between the two backgrounds grown in LB supplemented with different amounts of fructose. On a per-cell basis the wildtype performs about two-times better than the PHB-. The wildtype line also performs much better than the PHB- during bioreactor growth as well – readily reaching OD600 greater than 120 (corresponding to 50 gDW/L) and good triterpene production Figure 5-4. This is further proof that the PHB knockout line is highly constrained in its ability to grow well. Further work is needed to elucidate the exact nature of the issues with this line but it is quite likely because of its limited ability to control internal redox due to the knockout.

Re wildtype pRK 8.0 Re PHB- pRK 8.0

6.0 6.0

4.0 4.0

2.0 2.0 Squalene (mg/L) Squalene 0.0 (mg/L) Squalene 0.0 5 10 15 20 25 5 10 15 20 25 3.0 3.0

2.0 2.0

1.0 1.0 ualene (mg/gDW)

Sq 0.0 0.0 5 10 15 20 25 5 10 15 20 25 Squalene Squalene (mg/gDW) Fructose added (g/L) Fructose added (g/L) (+LB) (+LB)

Figure 5-3. Comparison between two backgrounds of R. eutropha: PHB- and wild-type in terms of triterpene production, transformed with pRK:Plac-SS-FPS. The cultures are grown in LB medium supplemented with indicated levels of fructose.

100

140 100

120 80

100 60 80 OD600 Squalene 40 Squalene (mg/L)Squalene 60 DO (% air sat) DO (%)

pH, pH, 20 , , 40

OD600 20 0

0 -20 0 20 40 60 80 100 Time (hr)

Figure 5-4. Bioreactor growth of R. eutropha wild-type transformed with pRK:Plac-SS-FPS.

5.I.III. Pathway enhancements

5.I.III.I MEP pathway

In parallel to the above studies, I have also chosen to use the wild-type of R. eutropha to see the effect of the enhanced MEP pathway. Based on our results for the different constructs with R. capsulatus, pBBR-Plac:SS-dxs-idi-fps was chosen for transforming R. eutropha for triterpene production. Figure 5-5 shows triterpene production by four ex-conjugants picked from the plate and grown on LB medium supplemented with 20 g/L glucose and antibiotic (500 µg/ml kanamycin). Specific production levels are somewhat less than those seen for R. capsulatus with the same construct and glucose supplementation. However, comparison based on complex

101 medium is slightly ambiguous and bioreactor studies using defined medium as well as autotrophic growth will clarify any obvious difference between Ralstonia and Rhodobacter.

Re wildtype pBBR-Plac-SS-dxs-idi-fps Re wildtype pBBR-Plac-SS-dxs-idi-fps 10.0 6.0 Medium: LB+Glucose (20 g/L) Medium: LB+Glucose (20 g/L) 8.0 4.0 6.0

4.0 2.0

Squalene (mg/L) Squalene 2.0 Squalene (mg/gDW) Squalene 0.0 0.0 1 2 3 4 5 1 2 3 4 5 Colony Colony

Figure 5-5. Triterpene production by R. eutropha wildtype transformed with pBBR:Plac:SS-dxs- idi-fps, grown on LB medium with 20 g/L glucose – total (left) and specific (right) titer levels.

5.I.III.II MVA pathway

The other pathway supplementation is via the MVA pathway. So I studied supplementation of the basic triterpene synthesis pathway (SS-fps) with an enhancement of the flux to IPP/DMAPP (substrates for FPS) using the MVA pathway genes. One way of achieving this was to transform the lines discussed above (i.e. R. eutropha wildtype or PHB- pRK:Plac:SS- fps) with the MVA pathway genes (see Figure 1-3). However, there have been considerable difficulty in cloning the codon-optimized versions of Operons 1 and 2 (R. capsulatus and R. eutropha codon-usage are approximately similar). Considerable effort was expended to obtain codon-optimized genes where postdoc Alex Rajangam delegated the assembly to Justin Yoo.

Amplification of the long operons with high GC content presented substantial problem and PCR polymerases and conditions needed to be optimized. Repetitive sequences in the operons interfered with primer design and caused delays and needed to be troubleshot. Working with wrong DNA for a period of about two months also contributed to a lot of wasted efforts.

102 Therefore, while that work was in progress, we have attempted to proceed with a construct developed by Greg Stephanopoulos lab (pBbA5c:Plac:MevT-MBIS; available from Addgene), that is codon-optimized for E. coli. This is essentially the same as Operons 1 and 2 driven by a single promoter Plac. The construct is then cloned out from the pBbA5c backbone into the pBBR vector for expression in R. eutropha and R. capsulatus, generating pBBR:Plac:MevT-MBIS.

This vector was transformed into the pRK line discussed above. Figure 5-6 shows the results of growth screening of five clones on LB medium with and without fructose. Results are slightly higher on average than those with only SS-fps but almost on the same level as the best performing line above. This was surprising as it was expected that the addition of the MVA pathway would provide a substantial flux to the triterpene synthesis. This could be the result of the very different codon-usage of E. coli vs. R. eutropha. However, we cannot definitively answer this question until the codon-optimized version of Operons 1 and 2 are successfully cloned.

3.50 3.00 2.50 g/L) 2.00 1.50 1.00 Squalene (m Squalene 0.50 0.00 LB LB+Fru 5g/L Medium

103

Figure 5-6. Triterpene production by R. eutropha PHB- pRK:Plac:SS-fps + pBBR:Plac:MevT- MBIS (dual plasmid).

II. Future work

There are immense opportunities to take this platform to the next level in terms of metabolic engineering, protein engineering, improvements in host organism and bioreactor. Much of this is described in Chapter 3. The task that can be finished in the near future without the involvement of much more additional resources is the study of the comparison of the MEP and

MVA pathway in R. capsulatus and R. eutropha, in heterotrophic as well as autotrophic. The goal of this would be to select the best combination of host and constructs to move forward to the next level of improvements. Some potential work that will require much more additional resources are described below.

5.II.I.I Generation of a mega-plasmid

A dual plasmid system is problematic in that it takes much longer for the conjugations due to the added step of conjugating the first plasmid, selecting and screening for hydrocarbon production before the next plasmid can be conjugated. Once the bacterial lines are obtained, it must be propagated under two to three antibiotic selections, which tend to negatively affect growth (and therefore triterpene production). Extent of plasmid loss is also more in this case. In case of R. capsulatus, we have generally seen fair degree of plasmid loss with the pBBR based constructs, and although measurements are not done for dual-plasmid systems, it can only be expected to be worse. Furthermore, due to the possibly different copy numbers of the two plasmids, the relative gene dosage is also variable. It is quite difficult to obtain accurate copy

104 number information, particularly because it varies with the growth stages. These problems prompted us to attempt to construct a single plasmid with all the genes of interest of the MVA pathway as well as the triterpene synthase and FPS (Figure 5-7). This will yield a proper comparison between the enhanced MEP and MVA pathways. However, the genetic engineering of this ‘mega-plasmid’ has proven particularly difficult to obtain as a result of these long operons, not only because of length (~3-5 kb) but also because of high GC content (>65%) and repetitive sequences in the intergenic regions. Restriction digest based cloning is also not possible because of the lack of compatible sites. These problems have been partially overcome by the use of several different polymerases and optimization of PCR conditions, but there are still some way to go before this construct can be entirely constructed and tested for functionality.

pBBR-Plac:Op1-Plac:Op2-Plac-SS-FPS e 17032 bp

Figure 5-7. Desired mega-plasmid containing all the MVA pathway genes, the triterpene synthase and fps.

105 5.II.I.II Promoter

The promoter used in the majority of our work, Plac, is a common, moderately strong and constitutive promoter in R. capsulatus and R. eutropha. But there are much stronger inducible promoters available that can increase gene expression and potentially improve the triterpene level. Some of these include Ptac, PBAD, Ptet etc. Much of the recent advancements regarding promoters is described in Section 3.VII.II. However, the highest level of expression of all the genes is probably not optimal for the desired pathway engineering. Much more important would be the relative level of expression of the genes and their respective enzyme activity within the engineered pathway. Furthermore, the gene expression should be inducible in such a way that it is regulated with respect to the growth, energy or carbon metabolism by the organism. Some such important host-specific inducible systems in R. capsulatus that respond to environmental factors are puc, puf, cbb, hydrogenase operon promoters. The first two respond to light levels and oxygen

(increasing expression with low levels of both), cbb promoters respond to CO2 and O2 levels and hydrogenase operon promoters respond to H2. All of these promoters also affected by the cellular redox state and global regulatory network. Similarly complicated gene regulatory systems exist for R. eutropha as well. As a result, it is not a straight-forward choice as to which environmental cue will be most suitable to use. For example, oxygen tension can be used based on a bioreactor operating condition that is also suitable from a safety standpoint. However, it is not directly related to growth so may not be so well-suited in that respect.

In the same way, fine tuning the relative expression level of the genes that result in maximal production is not a straight-forward objective, as a change in the expression level of one will require a change in all the others. A screening strategy could be designed in such a way that each of the genes (or at least each operon) has its own inducible promoter and/or operator.

Common promoters such as Plac, PBAD, Pfru, Ptet etc. could be used where each responds to a

106 different chemical substance over a substantial range. Growth tests could be performed where different levels of the inducer substance are used and transcript levels of the genes for the best triterpene producing line could be selected. Unless a high-throughput promoter modification method such as MAGE can be combined with rapid product screening (e.g. lycopene (Alper et al.,

2005)) optimization of productivity is a daunting task. In addition, since the proximity of the genes from the promoter also affects its expression level, the genes could be shuffled around within a particular operon using a strategy that maximizes the product of that operon.

5.II.I.III Other potential areas of improvement

Further factors controlling gene expression such as plasmid stability, gene dosage etc. must also be improved. Plasmid stability has been a significant problem in our studies. A strategy to make the plasmid indispensable to the cell - for example having a key gene in the hydrogenase or cbb operon on the plasmid while knocking out the corresponding copy in the genome is a useful strategy that has been used successfully. Gene expression is not the only factor determining the product formation through an engineered pathway. Strategies such as knocking out major non-essential pathways (such as PHB in R.eutropha) to divert the carbon flux will prove to be important at a later stage. Protein engineering, such as creating tethers or scaffolds for the proteins, is also a major area that can vastly improve product formation by channeling the substrate or preventing the loss of the substrate to other competing pathways in the cell. Much of this is discussed in detail in Chapter 3 and so is not repeated here.

107 III. Conclusion

In my PhD work, I tried to lay the foundation of the different aspects of the specific

Electrofuels paradigm we worked with, starting from analyzing the process economic feasibility and identifying major areas that need to be improved. I worked with two different classes of bacteria that are promising in different ways, although R. capsulatus is prioritized more than the

R. eutropha to begin with. Their comparison in terms of growth yield and maintenance parameters was completed but a proper comparison in terms of triterpene production still needs to be captured. I believe this work would be the basis of next level of engineering to improve biofuel production in autotrophic hosts.

108

Appendix A

Bioreactor material balance equations for section process economic analysis (Section 2.V.I)

In performing simulations the reactor is modeled as a continuous-flow perfusion with cell recycle. Since product is in the media steady operation can be characterized by the mean residence time of the liquid, however the amount of biomass present is determined from cell recycle.

2 1

XR XR Sep R F 3 F X = 0

XR, V XR F+R

Figure A1. Bioreactor flow streams and mass-balance envelops. 1. Cell separator and recycle, 2. bioreactor and cell separator, 3. bioreactor only.

Symbols:

F = Feed flow rate (L/h) R = Recycle flow rate (L/h) V = Volume of reactor (L) XR = Cell density in the reactor (gDW/L) Rfuel = specific productivity of the fuel (g-fuel/gDW-h) = Ratio of cell density of recycle stream to reactor outlet stream

109 = Ratio of cell density of the separator outlet stream to reactor outlet stream

The cell recycle efficiency is defined as,

= = (A1) () ()

From a biomass balance around the separator (1), the recycle ratio is,

= (A2)

From the overall biomass balance (2) at steady state,

= − = (A3)

where is the growth rate of the bacteria in the reactor.

= = (A4)

Thus, in a perfusion type system such as this, the actual growth rate of the organism is decreased as the cell recycle efficiency increases. Lower growth rates decreases the metabolic demand for growth and therefore increases the impact of maintenance coefficient.

Mass balance of the dissolved H2 in the liquid phase of the reactor (3),

[ ( ) ] = − + + + − − = ( ) (A5) /

where / is the true growth yield of biomass on H2, is the maintenance

coefficient of biomass on H2, is the mass transfer coefficient of H2. , and are

110 the liquid phase concentration, gas phase partial pressure and Henry’s law coefficient of H2 respectively.

In a simplified form assuming = 0 (limiting reactant),

[ − ( + ) + ] + = (A6) /

Similar equations can be written for O2 and CO2. These equations combined with the growth, cell maintenance and fuel synthesis equations discussed in the main body of the text for all three gases and biomass can be solved to get the stream values at a given gas composition. The optimization routine in Aspen can solve this set of equations within the gas composition range of each gas and the constraints of kLa to find the stoichiometric mix.

Volumetric fuel productivity is defined as,

= (A7)

Fuel yield on H2 is defined as

/ = (A8)

111 Appendix B

A comparison of experimental measurements and thermodynamic predictions of true yield ()

A modified form of the Thermodynamic Electron Equivalents Model (TEEM), developed and presented by McCarty (McCarty 2007; McCarty 1971; Rittman & McCarty 2001), was used as the basic framework for predicting the true yields attainable for cell growth and botryococcene product fuel (BPF) production by R. capsulatus and R. eutropha as a function of process conditions (Tuerk, 2011). Abbreviated here as the modified Electron Balance (EB) method, this approach incorporates knowledge about the carbon fixation and energy-capturing processes of

Rba. capsulatus and R. eutropha into the previously formulated TEEM model. By accounting for these differences, we attempted to improve the estimate of the metabolic efficiency factor (ε) for

each organism, and thus improving the estimates of true yield on hydrogen () for each organism, with the ultimate aim of improving projections of actual cell growth and botryococcene product fuel (BPF) yield across a range of process conditions.

Extensive details can be found in (A. L. Tuerk 2011); here we outline a simplified

version of the calculations producing values. This includes a summary of the original formulation of TEEM (McCarty 1971; McCarty 2007) and the subsequent modifications implemented in the modified EB method (Tuerk, 2011). Finally, the values of the cellular efficiency model parameter ε that most closely matched experimentally determined yield and maintenance values are highlighted.

112 Calculation of true yield by TEEM

For simplicity and demonstration purposes, we are here assuming an empirical biomass formula of C5H7O2N for both R. capsulatus and R. eutropha. Gibbs free energy values for half reactions on a per-electron equivalent (eeq) basis were obtained from (Rittman & McCarty 2001).

First, the electron donor half reaction (red) is constructed:

: + → , ∆ = . / (B1)

along with the electron acceptor half reaction (rea):

: + + → , ∆ = −. / (B2)

These are combined to form the overall energy-generating reaction at standard biological conditions:

= − : + → , ∆ = ∆ − ∆ = −. / (B3)

Adjustments to the Gibbs free energy of this reaction should be made for deviations of culture conditions (temperature, pH, activity of products/reactants) relative to standard biological conditions; for simplicity here, we are only invoke energy values under standard biological conditions.

The cell synthesis half reaction (rsynth) is constructed based on the empirical biomass formula:

: + + + → + (B4)

113 Unlike the half reactions for the electron donor and the electron acceptor, the

Gibbs free energy associated with the cell synthesis half reaction is not explicitly known.

Under TEEM, it is assumed that the energy required for cell synthesis comes from the energy necessary to reduce the carbon source to the level of a metabolic intermediate

(Acetyl-CoA), using electrons from the electron donor. This is predicted to be an exergonic reaction at standard biological conditions):

= − : + + → + + (B5)

∆ = ∆ − ∆ = . − . = −. /

Additionally, ATP generation is another substantial component of the energy demands for cell synthesis. This is envisioned as a second step in the cell synthesis process, where ATP is used to synthesize cellular components from the intermediate

Acetyl-CoA:

. × / × ∆ = = = . / (B6)

The value of / (cost of ATP synthesis per eeq cell) was obtained from Rittman and McCarty, 2001, is the formula weight of biomass according to the empirical biomass formula, and is the number of electron eeq required to produce one mol of cell from CO2 (obtained from the cell synthesis half reaction).

The overall Gibbs free energy of cell synthesis (∆) is assumed to be the sum of the energy required for these two steps (using values adjusted for non-standard

114 conditions); to account for the irreversibility of cellular metabolism, the energetic of the individual steps must be adjusted for cellular inefficiency (ε). Thus, exergonic reactions are penalized so that less energy is available to the cell than thermodynamics predict, and endergonic reactions are penalized so that more energy is required:

∆ ∆ = ∆ + ∆ < (B7)

Finally, an electron balance on the electrons from the electron donor substrate partitions the fraction of electrons from the ED into those required to be used for energy

generation ( ), and those able to be used for cell synthesis ( ). These fractions can be calculated from the relative energetic of cell synthesis and energy generation as follows:

= (B8)

= (B9)

∆ = − (B10) ∆

The values of and are employed to construct the overall stoichiometric equation for growth:

= + (B11)

From the overall stoichiometric equation for growth, true yield can be calculated from the stoichiometric coefficients () for electron donor and cell biomass. For this particular case,

115

× = = (B12)

Modifications applied to the original TEEM to incorporate autotrophic growth, reversed electron transport, and BPF synthesis for R. eutropha and R. capsulatus

A number of modifications were made to the original formulation of TEEM to account for chemolithoautotrophic metabolism and botryococcene hydrocarbon synthesis, as TEEM was primarily was developed for heterotrophic organisms. Accounting for autotrophic growth and reversed electron transport was accomplished by altering the method for calculating the Gibbs free energy of cellular synthesis (∆), although the overall stoichiometry of is unchanged. The modifications are summarized in Figure S2.1 below, with equations S2.13 and

S2.14 providing the details of this altered computation; these can be compared to their equivalent in the original method, equation S7.

116

Figure B1: A step-by-step comparison of the calculation by the original TEEM method, and the modified version developed specifically for Electrofuels process calculations. (Figure reproduced from Tuerk, 2011).

for R. eutropha (without Reversed Electron Transport): (B13)

= + + +

utilization of e- from e- from ATP for cell Step: e- from H to NADH NADH to acetyl-coA 2 and BPF acetyl-coA cell and BPF synth.

117

for R. capsulatus (with Reversed Electron Transport): (B14)

= + + + +

e- from H e- from utilization of 2 e- from e- from to ubiquinone ATP for cell Step: NADH to acetyl-coA ubiquinone pool to and BPF acetyl-coA cell and BPF pool NADH synth.

The factor is +1 for endergonic reactions, and -1 for exergonic reactions. The use of

NADH as an intermediate electron carrier of reducing energy for CO2 fixation is based in physiology, as this is the true electron donor for carbon fixation via the Calvin-Benson-Bassham pathway. In reversed electron transport (RET), NAD+ is not reduced directly by the electron donor, as it is for R. eutropha (no RET). Instead, electrons from the ED must first reduce

+ ubiquinone (E0’=+113 mV (Thauer et al. 1977); ∆ = -10.9 kJ/eeq), and then NAD is reduced by expending the proton motive force to force electrons from ubiquinone uphill (E0’ = -320 mV

(Thauer et al. 1977); ∆ = 30.9 kJ/eeq). In addition, the baseline value of / was increased to 5.36 kJ/g cell (28.5 kJ/eeq) for calculations in the Electrofuels scenario, based on data provided in (Kelly 1990), to account for the observation that autotrophy is energetically more demanding than heterotrophy.

To account for BPF production, two additional modifications were applied. First, the selected ratio of BPF synthesis to cell synthesis was used to adjust the empirical cell biomass formula. The effect of this was to increase the degree of reduction of the cell, requiring more electrons from the electron donor to enable synthesis. Secondly, the additional energy required for synthesizing BPF in addition to cellular components was accounted for by increasing the

118

value of / in accordance with the ratio of BPF to cell, based on an experimentally determined value for the heat of combustion of botryococcene. For details, see Tuerk, 2011.

A summary of the Gibbs free energy values employed for calculating are provided in

Table B1 below; for further details of the calculations, see (A. L. Tuerk 2011). For simplicity, the values in this table are based on calculations at standard conditions, and have not been corrected for predicted process conditions. For illustration, we used the empirical biomass formula

(C5H7O2N) reported above, without the production of BPF.

Table B1: Calculation steps using modified TEEM

∆ of electron transport steps R. capsulatus R. eutropha

(with RET) (no RET)

∆ = ∆ − ∆ −50.8 −8.97 (∆ − ∆) 41.8 ∆ = 0.25 ∆ = ∆ (33.3 − 30.9) = 0.12 / 0.25 0.25 × 20 − ∆ ≡stoichiometric coefficient of CO2 in biomass (cell + fuel) half reaction ≈ 0 / ∆ ∆/ 5.36 = 28.5

Comparison of original TEEM and Modified Electron Balance methods of predicting

Using the methods and values described above, was calculated for a range of cellular efficiency values at standard biological conditions. The results are shown in Tables B2 (TEEM) and B3 (Modified Electron Balance) below. It should be noted that for simplifying these

119 calculations, a single baseline elemental composition of R. capsulatus was used for both R. capsulatus and R. eutropha; the use of empirical biomass formulas specific for each species yields slightly different values.

Table B2: as a function of ε, calculated using original TEEM method ε ∆ A 0.20 92.00 3.879 0.205 0.795 2.32 0.25 72.79 2.455 0.289 0.711 3.27 0.30 59.84 1.682 0.373 0.627 4.21 0.35 50.45 1.216 0.451 0.549 5.10 0.37 47.38 1.080 0.481 0.519 5.43 0.40 43.31 0.913 0.523 0.477 5.91 0.45 37.65 0.705 0.586 0.414 6.63 0.50 33.03 0.557 0.642 0.358 7.26 0.55 29.17 0.447 0.691 0.309 7.81 0.60 25.88 0.364 0.733 0.267 8.29 0.65 23.03 0.299 0.770 0.230 8.70 0.70 20.52 0.247 0.802 0.198 9.06

Table B3: as a function of ε, calculated using the modified Electron Balance method R. capsulatus R. eutropha ∆ ε A ε ∆ A 0.20 341.9 14.42 0.065 0.935 0.73 0.20 141.3 5.96 0.144 0.856 1.62 0.25 269.0 9.07 0.099 0.901 1.12 0.25 112.2 3.79 0.209 0.791 2.36 0.30 219.5 6.17 0.139 0.861 1.58 0.30 92.7 2.61 0.277 0.723 3.13 0.35 183.4 4.42 0.185 0.815 2.09 0.35 78.6 1.89 0.345 0.655 3.90 0.37 171.5 3.91 0.204 0.796 2.30 0.37 74.0 1.69 0.372 0.628 4.21 0.40 155.7 3.28 0.233 0.767 2.64 0.40 68.0 1.43 0.411 0.589 4.65 0.45 133.6 2.50 0.285 0.715 3.22 0.45 59.6 1.12 0.473 0.527 5.34 0.50 115.4 1.95 0.339 0.661 3.83 0.50 52.8 0.89 0.529 0.471 5.98 0.55 100.1 1.53 0.395 0.605 4.46 0.55 47.1 0.72 0.581 0.419 6.56 0.60 86.9 1.22 0.450 0.550 5.09 0.60 42.3 0.59 0.627 0.373 7.09 0.65 75.3 0.98 0.506 0.494 5.72 0.65 38.2 0.50 0.669 0.331 7.56 0.70 65.0 0.78 0.561 0.439 6.34 0.70 34.6 0.42 0.706 0.294 7.98

At a ‘typical’ cellular efficiency of 0.37 (McCarty, 2007), the values calculated by the modified EB method for R. capsulatus and R. eutropha were 2.30 and 4.21 gDW/mol-H2 respectively, compared to the value of 5.43 gDW/mol-H2 by the original TEEM method.

120 Comparing these values to experimentally determined values (Table B4 below), it is clear that the original TEEM method substantially overestimates true yield for R. capsulatus, which is expected since the energetically demanding process of reversed electron transport is not explicitly accounted for by the original TEEM method. Accounting for reversed electron transport via the modified EB method predicts true yield values closer to experimentally determined values; an efficiency factor of 0.40-0.45 would be a more appropriate parameter choice for R. capsulatus in this model. For R. eutropha, the choice of ε=0.37 underestimates true yields using both methods

(compared to experimentally determined values in this work). This suggests that R. eutropha may have a higher inherent cellular efficiency than the ‘typical’ value, on the order of 0.65.

Table B4: Comparison of the true yield and maintenance coefficient values from various sources R. capsulatus R. eutropha ∙ ℎ References ∙ ℎ References 2.86 - (Siefert & Pfennig 4.1 0.001 (Siegel & Ollis 1984) 1979) 4.5 0.025 6.4 0.007 (Bongers 1970) 2.66 0.002 Experimental (this 7.68 0.0068 Experimental (this work) work) 2.30 - Calculated by 4.21 - Calculated by modified EB modified EB (ε=0.37) (ε=0.37) 5.43 - Calculated by 5.43 - Calculated by original TEEM original TEEM (ε=0.37) (ε=0.37)

The power of this type of modeling is the ability to use experimentally determined values to project outcomes for future research directions - namely the ability to predict BPF yields based on what is known about cell growth and behavior in the absence of fuel production.

121 Appendix C

Measurement of growth yield and maintenance coefficient

Growth and maintenance processes of an organism are the major sinks for the energy substrate, H2 in this case. Since the cost of H2 is also the most significant fraction of the fuel cost from the electrofuels process, it was hypothesized that the organism with higher growth yield and lower maintenance coefficient will ultimately result in lower fuel cost. For these reasons, the growth yield and maintenance coefficient of R. capsulatus and R. eutropha were determined under similar autotrophic conditions to provide an unbiased comparison between the two hosts.

The process model described in Chapter 2 of this thesis utilizes these parameters to predict the relative effect of the two parameters on the fuel cost. In this appendix, I describe the general background and procedure adopted for their determination.

Growth yield and maintenance coefficient

For a microbe, growth yield is defined as the energy needed for biomass generation and cell division only and not for the processes related to general maintenance of the cell such as homeostasis, motility, futile cycling etc. (Pirt 1965). The growth yield and maintenance coefficient of R. capsulatus on H2, are not reported in the literature. In fact, values of these parameters are generally very scarce for any organism under autotrophic conditions. Only those of R. eutropha can be found on multiple instances in the literature. Siegel & Ollis (1984) noted that the maintenance coefficient of R eutropha is a strong function of dissolved O2 concentration.

As observed by other researchers (Howells 1982), O2 can be both inhibitory and limiting for the growth of a bacteria at different concentrations and is strongly strain dependent. Under continuous culture conditions, the steady state in between inhibitory and limiting concentrations

122 is unstable and tends to move either to washout at higher concentration or towards O2 limitation

(Siegel & Ollis 1984). In general, high O2 concentration is problematic for most chemoautotrophs because of the deactivation (permanent or temporary) of the hydrogenase enzymes (this is described in more detail in Section 3.IV.II). Therefore, lower O2 concentration is expected to improve the values of these parameters. O2 concentration is a very important parameter from not only this perspective, but also from a safety standpoint, because a mixture of H2 and O2 is generally explosive at an O2 concentration greater than about 6.9%. These factors ultimately largely dictate a low (and/or limiting) O2 concentration in the reactor. I have thus used O2-limited conditions while measuring the growth yield and maintenance coefficients.

Theory

In principle, maintenance coefficient and growth yield can be calculated if the yield on the substrate is known as a function of the growth rate. (Pirt 1965) derived the following equation

as the inter-relationship among specific growth rate () and yield () and true yield () and maintenance coefficient ().

= + (6)

and are both experimentally measurable parameters. Therefore plotting against

a straight line can be obtained whose slope and intercept would be equal to the and respectively. This has been traditionally used in the determination of true yield and maintenance coefficient. However, in a batch system the growth rate varies continuously, therefore an accurate measure of instantaneous growth rate and yield are not possible. At the same time, the culture goes through physiological changes as it progresses through the different growth stages and therefore the yields obtained this way may not refer to the same condition. So, true yield and

123 maintenance are measured in continuous systems, in which the growth rate of an organism can be fixed at a particular value by fixing the dilution rate (inverse of residence time in a reactor).

Cell balance on a continuous reactor:

= − (7)

where = volume of reator, = growth rate, = flow rate, and = inlet and outlet cell concentration respectively. With 0 inlet cell concentration at steady state,

= (8)

From the definitions of specific growth rate, μ = and dilution rate, = , we have μ

= in a steady state continuous flow reactor.

Multiple steady states are established by varying the dilution rate and yield values are measured at these steady states. The corresponding yield and dilution rates are plotted on a 1/ vs. 1/ graph to get the true yield and maintenance coefficient. However, this method is very time consuming and produces only a limited number of data points. Increasingly, these parameters are being measured in deceleration-stat or acceleration-stat (D-stat or A-stat) (Barbosa et al. 2005) in contrast to a chemostat. Instead of going from one dilution rate to the next in a step and waiting for a steady state to be established, the dilution rate is varied smoothly from one to the next. If the rate is slow enough for the cells to adapt physiologically, the culture goes through a succession of pseudo steady-states and the yield and dilution rate within this slowly changing dynamic period is used to calculate the maintenance and true yield. The D-stat method is followed in my determination of the yield and maintenance coefficient.

124 Experimental procedure and results

The experimental setup and schematic of the autotrophic bioreactor is shown below

(Figure C1). H2, O2 and CO2 gases are mixed via three mass flow controllers (SLA 5850, Brooks

Instruments, CA; http://www.brooksinstrument.com/downloads/Product%20Documentation/Thermal%20Mass%20

Flow%20Meters%20Controllers%20Digital%20Gas/Data%20Sheets/ds-tmf-sla5800-mfc- eng.pdf) and delivered to the reactor after sparging through a humidifying column. The outlet gas passes through a drying column, then through a mass flow meter (M-50SCCM, Alicat Scientific; http://www.alicat.com/products/mass-flow-meters-and-controllers/mass-flow-meters/) that measures the flow rate and a process GC (8610, SRI Instrument; http://www.srigc.com/2005catalog/cat38-39.htm) that measures the composition. The composition data is collected once every hour. The reactor used in this work is a 1 L reactor

(Celligen, New Brunswick). The liquid flow rate into and out of the reactor is controlled via a highly precise peristaltic pump (120 U, Watson-Marlow), whose rotation can be controlled to within 0.01 rpm by a 0-5 V analog input. This is very critical in maintaining an accurate and consistent flow rate and smoothly varying it for D-stat. The actual flow rate of the inlet media is calculated by periodically recording the weight of the media bottle which sits on a balance and fitting appropriate regression lines through the associated dynamics. The culture is recirculated via a second peristaltic pump for online optical density measurement. Samples are taken periodically from the outlet for offline OD and dry weight measurements. The media used for growth is based on a previous work in the literature (M. T. Madigan & Gest 1979) where the autotrophic growth of R. capsulatus was first demonstrated. A pH probe continuously measures the pH of the culture. As discussed before, the culture is allowed to come to a steady state at a particular dilution rate and then the dilution rate is decreased slowly. The deceleration rate is

125 based on previous literature work on R capsulatus (Hoekema, R. D. Douma, et al. 2006) and experience with running the experiment multiple times with various difficulties over a period of more than six months.

Figure C1: Experimental setup for the measurement of growth yield and maintenance coefficient.

Figure C2: Watson-Marlow 120 U three-channel precision peristaltic pump used for media delivery to and removal from the reactor. A 15-pin D-sub connector was used to interface with computer to change the flow rates and log the actual rotor speed.

126 Another critical aspect of the experiment is the measurement of gas flow rate and composition, which is much more difficult to do accurately than liquid phase because of strong temperature and pressure effects and leaks. Mass flow meters, although capable of measuring flow rate of a single gas very accurately, are not very accurate for a mixture of gases and at low flow rates. A soap-bubble meter (1-10-100 ml meter, Bubble-O-Meter, Dublin, OH; http://www.bubble-o-meter.com/bom.php?curPos=BOM1-10-100ml) was therefore placed at the end of the line to calibrate the outlet flow based on manual measurements. This was checked against the flow rate based on composition correction of the MFM and they compared very well.

Furthermore, a state-of-art GC was used to analyze these gases continuously in a single pass for composition measurement.

Figures C2 through C4 shows representative data from an experiment for the measurement of growth yield and maintenance coefficient of R. capsulatus. Similar methods were used for R. eutropha as well (data not shown). Figure C3 is the total media consumption with time at various stages of the experiment. The instantaneous flow rates are calculated by the first derivative of the fitting equations. As can be seen, this parameter could be maintained very accurately using the precision peristaltic pump. The deceleration rates used in these experiments were based conservatively on a previous work of (Hoekema, R. D. Douma, et al. 2006) where they measured the yield and maintenance of R. capsulatus in the photoheterotrophic growth mode, producing H2. Plot of culture dry weight against time (Figure C4) shows that the expected variation with media flow rate deceleration. The apparent yields calculated in two such D-stat experiments for R. capsulatus is plotted in Figure C5, with µ/Y vs. µ. Inverse of the slope then gives the growth yield and intercept the maintenance coefficient. The values obtained in these experiments are shown in Table C1.

127

1900 Initial steady state 1700 Deceleration y = 6.9181x - 1068.2 1500 Final steady state

1300 y = -0.0487x2 + 41.799x - 7312.4 1100

Media consumed (g) Media consumed y = 10.026x - 2131.5 900

700

500 300 320 340 360 380 400 420 Time (hr)

Figure C3. Media reservoir weight throughout the experiment. The first derivative of y gives the instantaneous media flow rate, while the second derivative gives the deceleration rate. Media flow rates varies from 10.03 g/hr at the initial steady state to 6.92 g/hr at the final steady state, with a deceleration rate of -0.0974 g/hr2.

1.5 10.5

10 1.4 9.5 1.3 9 1.2 Biomass dry weight 8.5

1.1 Media flow rate 8 7.5 1 Flow rate (g/hr) rate Flow Dry weight (g/L) weight Dry 7 0.9 6.5

0.8 6 300 320 340 360 380 400 Time (hr)

Figure C4. Change of culture density with media flow rate. At the initial steady state the culture density remains around 1 g/L, increasing with decrease of flow rate to about 1.4 g/L during the final steady state.

128

30 6 H2 (Expt. 1) O2 (Expt. 1) H2 (Expt. 2) O2 (Expt. 2) 25 5 CO2 (Expt. 1) CO2 (Expt. 2)

20 y = 372.99x + 2.6521 4 g.hr)

15 3 y = 377.81x + 1.3736 y = 43.457x + 0.4721 10 2 μ/Y x 1000 (mol/ x 1000 μ/Y

y = 42.378x + 0.1957 5 1 y = 76.632x + 0.9908 y = 95.132x + 0.5096 0 0 0 0.02 0.04 0.06 0.08 μ (1/hr)

Figure C5. Plot of µ/Y vs. µ for two D-stat experiments covering two different ranges of dilution rates. Inverse of the slope gives the growth yield and the intercept gives the maintenance coefficients.

Table C1: Growth yield and maintenance coefficients of R. capsulatus and R. eutropha.

Growth yield (gDW/mol-H2) Maintenance coefficient

(mol-H2/gDW.hr)

R. capsulatus 2.66 2.01x10-3

R. eutropha 7.68 6.8x10-3

These results indicate that in the non-growing conditions, R. eutropha requires more than

3x more enerfy for maintenance. In the ideal production system, growth would be minimized since it represents loss of substrate to biomass rather than conversion to product. Therfore despite the fact that biomass yield of R. eutropha is twice that of R. capsulatus it may not be the best platform. It comes down to the efficiency with which the reducing power can be delivered to the heterologously introduced pathways.

129 Appendix D

Materials and methods

Bacterial Strains and Growth Conditions

E. coli DH5α (Invitrogen) was used as the host for routine cloning experiments and for E. coli triterpene engineering. E. coli DH5α was grown on LB agar or in LB broth at

37°C for cloning and routine maintenance. For triterpene engineering, E. coli DH5α was transformed with the plasmid constructs using standard heat shock protocols (Schmidt et al.

1999). R. capsulatus SB1003 (ATCC BAA-309) was used as a second prokaryotic host for triterpene engineering experiments. For heterotrophic growth and maintenance, R. capsulatus was grown in YCC liquid medium (5 g/L yeast extract, 6 g/L casamino acids) and on YCC agar plates at 34oC. When phosphate buffered saline was required, phosphates were added to a final concentration of 6.60 mM K2HPO4, 3.30 mM KH2PO4 just before inoculation. Glucose supplementation was accomplished by, adding a sterilized 1M glucose solution to a final concentration of 20-100 mM just before inoculation. The bacterial cell lines are described in

Table D1. When appropriate, E. coli strains were grown with 50 µg mL-1 kanamycin and 100 µg mL-1 ampicillin. R. capsulatus strains were grown in the presence of 20 µg mL-1 kanamycin and

10 µg mL-1 rifampicin.

For heterotrophic production of triterpenes, both E. coli and R. capsulatus were picked as a single colony from a plate and grown in 1 mL of LB or YCC respectively and were incubated for 20 hours at 37°C or 35°C. This allowed for normalization of growth between biological replicates the same growth-stage before performing the time course and terminal endpoint fermentations. Terminal endpoints for heterotrophic and photoheterotrophic samples were taken after 48 hours for five replicate cultures.

130

Table D1. Strains and plasmids used in this study.

Strain Characteristics and Relevance References

E. coli DH5α Host for routine cloning and operon expression Invitrogen™ E. coli S17-1 Host for mobilization of broad host range plasmids into R. capsulatus; recA pro hsdR RP4-2-Tc::Mu- Km::Tn7 R. capsulatus SB1003 Rhodobacter expression host for triterpene production. Plasmid Characteristics References pBlueScript II SK(+) Plasmid cloning vector; pMB1 replicon, lacZ, ApR Agilent pGEM-T Easy Vector Plasmid cloning vector for TA-cloning of PCR Promega products; pMB1 replicon, lacZ, ApR pBBR1MCS-2 Broad Host Range E. coli expression vector; pBBR R replicon, Km , oriT, PLac promoter pBBR:PLac:SS SS from Botryococcus braunii race B with the C- This study terminal membrane spanning domain deleted; pBBR R replicon, Km , oriT, PLac promoter pBBR:PLac:SS+fps SS and fps from chicken expressed as a single operon This study R under PLac promotion; pBBR replicon, Km , oriT, PLac promoter pBBR:PLac: SS, dxs, idi, and fps expressed as a single operon under This study R SS+dxs+idi+fps PLac promotion; pBBR replicon, Km , oriT, PLac promoter pBBR:PLac:BS A gene fusion of SSL-1 and SSL-3 botryococcene This study synthase enzymes from Botryococcus braunii race B R under PLac promotion; pBBR replicon, Km , oriT, PLac promoter pBBR:PLac:BS+fps BS gene fusion and fps expressed as a single operon This study R under PLac promotion; pBBR replicon, Km , oriT, PLac promoter pBBR:PLac:BS+dxs+ BS gene fusion and dxs, idi, and fps expressed as a This study idi+fps single operon under PLac promotion; pBBR replicon, R Km , oriT, PLac promoter

Small Scale Phototrophic and Autotrophic Growth Conditions for R. capsulatus Strains

To induce photoheterotrophic growth, R. capsulatus SB1003 wildtype strain and its derivatives were picked as single colonies directly from YCC plates conjugation plates

131 (supplemented with antibiotics when appropriate) and inoculated into 13 mm screw cap culture tubes filled completely with RCVB medium (M. Madigan & Gest 1979). This photoheterotrophic growth phase was found to be necessary to efficiently transition R. capsulatus

SB1003 from chemoheterotrophic growth to autotrophic growth. The cells were grown anaerobically under light to saturation and inoculated to a starting OD660 of about 0.1 (Beckman

DU520 spectrophotometer) into CA media (M. Madigan & Gest 1979) with antibiotics (M.

Madigan & Gest 1979). After the lag period, 2 mL samples are removed every 24 hrs to measure

OD660 and analyze for product.

The input gases for autotrophic growth (H2, CO2, and O2) were mixed via three mass flow controllers (SLA5850, Brooks Instruments, Hatfield, PA) to a set composition of 80:10:10. The autotrophic growth setup consists of 125 mL baffled Erlenmeyer flasks, each with a gas inlet and outlet, and a sampling port (5” 16G needle extending to the bottom of flask with 3 mL syringe).

The flasks are connected in series for gas delivery and are shaken at 200 rpm (G2 shaker, New

Brunswick). The entire setup is contained in an incubated chamber kept at 34°C.

Heterotrophic Bioreactor Growth Condition

Cells were grown in a 5 L bioreactor (BioFlo, New Brunswick, Edison, NJ) with a two

Rushton impellers in glucose fed-batch mode. The working liquid volume of the reactor was 4 L.

The pH of the culture was maintained at 7 by feeding 8 N NH4OH solution. The O2 fraction in the gas feed and the agitation was increased manually initially to maintain dissolved oxygen (DO) around 20% of air saturation until an RPM of 250 and dissolved O2 fraction of 15%. Temperature was maintained at 34 oC. The reactor was inoculated from a 100 mL shake-flask culture grown aerobic heterotrophically in RCVB(-M) media+80 mM glucose+25 µg/mL kanamycin to a starting OD660 of around 0.1 in the reactor of the same media (M. Madigan & Gest 1979). 2-3 mL

132 sample was removed from the reactor at regular interval for the measurement of OD660 and hydrocarbon content.

Autotrophic Bioreactor Growth Conditions

Cells were grown in a 1 L bioreactor (BioFlo, New Brunswick, Edison, NJ) with a cell- lift impeller in batch and continuous growth mode. The working liquid volume of the reactor was

900 mL. The inlet gas stream was mixed using three high-precision mass flow controllers

(SLA5850, Brooks Instruments, Hatfield, PA) and the outlet gas flow rate and compositions were measured using a mass flow meter (M-50SCCM-D, Alicat Scientific, Tucson, AZ) and two gas flow sensors for H2 (BCP-H2, BlueSens, Herten, Germany) and O2/CO2 respectively (BlueInOne,

BlueSens, Herten, Germany). The pH of the culture was maintained at 7 by feeding 25% NH4OH solution. The RPM and temperature were maintained at 1000 and 34 oC. The inlet gas flow rate and compositions were varied during batch growth but kept constant during continuous growth mode (discussed in the results section). The reactor was inoculated from an autotrophically grown flask culture (described above) to a starting OD660 of 0.1 in CA media with antibiotic (Imam et al.

2013). After about 15 hrs from inoculation, 2 mL sample was removed from the reactor at a regular interval for the measurement of OD660 and hydrocarbon.

Cloning Procedures

Nucleotide sequences for the native and codon optimized versions of genes used in this report are included in Table D2. The Rhodobacter codon optimized genes were synthesized by

Genscript. These genes were cloned into the pUC57 cloning vector and were the source of template for PCR amplification. All gene combinations were amplified from pUC57 templates by

133 PCR according to standard procedures (Schmidt et al. 1999). Genes were amplified and spliced by overlap extension (SOE) into polycistronic cassettes, which were then cloned into pGEM-T- easy vector for sequencing (Promega). The outermost primer set for amplifying the fully spliced region was modified to have a unique 5’- BamHI restriction site and a 3’-XbaI restriction site.

The cassettes for SS, BS, SS-fps, BS-fps, SS-dxs-idi-fps, and BS-dxs-idi-fps were cut with unique

BamHI and XbaI restriction sites and spliced into the same sites of pBBR1MCS-2 under the control of the PLac promoter, which is a strong constitutive promoter in R. capsulatus due to the lack of the lacI repressor.

Table D2. Nucleotide sequences of the wildtype and R. capsulatus codon-optimized genes used.

Wildtype squalene synthase (SS) nucleotide sequence ATGGGGATGCTTCGCTGGGGAGTGGAGTCTTTGCAGAATCCAGATGAATTAATCC CGGTCTTGAGGATGATTTATGCTGATAAGTTTGGAAAGATCAAGCCAAAGGACGA AGACCGGGGCTTCTGCTATGAAATTTTAAACCTTGTTTCAAGAAGTTTTGCAATCG TCATCCAACAGCTCCCTGCACAGCTGAGGGACCCAGTCTGCATATTTTACCTTGTA CTACGCGCCCTGGACACAGTCGAAGATGATATGAAAATTGCAGCAACCACCAAGA TTCCCTTGCTGCGTGACTTTTATGAGAAAATTTCTGACAGGTCATTCCGCATGACG GCCGGAGATCAAAAAGACTACATCAGGCTGTTGGATCAGTACCCCAAAGTGACAA GCGTTTTCTTGAAATTGACCCCCCGTGAACAAGAGATAATTGCAGACATTACAAAG CGGATGGGGAATGGAATGGCTGACTTCGTGCATAAGGGTGTTCCCGACACAGTGG GGGACTACGACCTTTACTGCCACTATGTTGCTGGGGTGGTGGGTCTCGGGCTTTCC CAGTTGTTCGTTGCGAGTGGACTACAGTCACCCTCTTTGACCCGCAGTGAAGACCT TTCCAATCACATGGGCCTCTTCCTTCAGAAGACCAACATCATCCGCGACTACTTTG AGGACATCAATGAGCTGCCTGCCCCCCGGATGTTCTGGCCCAGAGAGATCTGGGG CAAGTATGCGAACAACCTCGCTGAGTTCAAAGACCCGGCCAACAAGGCGGCTGCA ATGTGCTGCCTCAACGAGATGGTCACAGATGCATTGAGGCACGCGGTGTACTGCCT GCAGTACATGTCCATGATTGAGGATCCGCAGATCTTCAACTTCTGTGCCATCCCTC AGACCATGGCCTTCGGCACCCTGTCTTTGTGTTACAACAACTACACTATCTTCACA GGGCCCAAAGCGGCTGTGAAGCTGCGTAGGGGCACCACTGCCAAGCTGATGTACA CCTCTAACAATATGTTTGCGATGTACCGTCATTTCCTCAACTTCGCAGAGAAGCTG GAAGTCAGATGCAACACCGAGACCAGCGAGGATCCCAGCGTGACCACCACTCTGG AACACCTGCATAAGATCAAAGCTGCCTGCAAGGCTGGGCTGGCACGCACAAAAGA TGACACCTTTGACGAATTGAGGAGCTAA Wildtype botryococcene synthase fusion, SSL-1 and SSL-3, (BS) nucleotide sequence ATGACTATGCACCAAGACCACGGAGTCATGAAAGACCTTGTCAAGCATCCAAATG AATTTCCATACTTGCTCCAACTAGCTGCAACAACGTACGGCTCACCGGCTGCACCG ATCCCCAAGGAACCGGACCGAGCTTTCTGCTACAATACTCTTCACACCGTTTCGAA GGGGTTCCCCAGATTTGTTATGAGACTTCCGCAGGAACTCCAAGATCCGATATGCA TATTCTACCTCCTGTTGCGAGCACTAGACACGGTGGAGGATGATATGAACCTCAAA

134

AGTGAGACGAAGATTTCACTCCTACGCGTTTTCCATGAACACTGTTCAGACAGGAA CTGGAGTATGAAAAGTGATTATGGCATATATGCAGATCTGATGGAAAGATTCCCC CTGGTCGTATCCGTCTTAGAGAAGCTCCCTCCCGCCACACAGCAGACTTTCAGGGA GAATGTCAAATACATGGGCAATGGCATGGCAGATTTTATTGATAAGCAGATCCTG ACAGTGGATGAGTACGACCTCTACTGCCACTATGTGGCCGGCAGTTGCGGCATTGC TGTCACCAAGGTCATTGTGCAGTTCAACCTTGCCACGCCTGAAGCTGACTCCTACG ACTTTTCCAACAGTCTGGGCCTCTTGCTTCAGAAGGCCAACATCATCACTGACTAC AATGAAGACATCAATGAAGAGCCCAGGCCCAGGATGTTCTGGCCCCAGGAGATTT GGGGGAAGTACGCGGAGAAGTTGGCTGACTTCAATGAACCCGAAAATATTGATAC AGCCGTGAAGTGCTTGAACCACATGGTCACAGATGCAATGCGGCACATTGAGCCT TCCCTCAAAGGCATGGTTTATTTCACAGACAAGACAGTCTTTCGGGCGCTCGCTCT TCTGCTGGTCACAGCCTTTGGCCATTTGTCCACTTTGTACAACAACCCCAATGTCTT TAAAGAGAAAGTGAGACAGCGGAAGGGAAGGATTGCACGGCTGGTCATGTCATCC AGGAATGTACCAGGCCTCTTCCGTACATGCCTCAAACTCGCAAACAACTTCGAGTC CAGGTGCAAGCAAGAGACGGCAAATGATCCCACTGTGGCCATGACTATCAAGCGC TTGCAATCTATTCAAGCTACATGCAGAGATGGCCTGGCCAAGTATGACACACCCTC TGGGCTGAAATCTTTCTGCGCAGCCCCAACTCCCACCAAGGGTGGTTCTGGTGGTG GTTCTGGTGGTGGTTCTGGTATGAAACTTCGGGAAGTCTTGCAGCACCCGGGTGAG ATTATCCCTCTCCTGCAAATGATGGTCATGGCCTACCGCAGGAAGAGGAAGCCTCA AGATCCCAATTTGGCCTGGTGCTGGGAGACGCTGATTAAAGTTTCGAGAAGTTACG TTCTAGTCATTCAGCAGCTTCCTGAAGTACTTCAGGACCCTATCTGCGTCAACTATC TTGTTCTTCGAGGCTTGGACACACTGCAGGATGACATGGCAATTCCCGCAGAGAA GCGGGTTCCACTCCTCCTCGACTACTACAACCATATTGGAGACATAACTTGGAAGC CGCCTTGCGGATATGGGCAGTATGTGGAGCTGATTGAGGAGTATCCAAGGGTGAC AAAAGAGTTCTTGAAACTCAACAAGCAAGATCAGCAGTTTATCACGGACATGTGC ATGCGGCTGGGAGCGGAGATGACAGTATTTCTCAAGAGGGACGTGTTGACAGTTC CTGACTTGGATCTGTATGCCTTCACTAATAACGGGCCAGTTGCTATCTGCCTGACC AAGTTATGGGTGGACAGAAAGTTTGCAGACCCAAAGCTTCTGGACCGGGAGGACC TATCGGGCCACATGGCCATGTTCTTGGGCAAGATTAACGTCATCCGCGACATCAAG GAGGATGTCTTGGAGGATCCTCCTCGCATCTGGTGGCCGAAGGAGATCTGGGGAA AGTACCTCAAGGACCTGAGGGACATCATCAAGCCTGAGTATCAAAAGGAAGCGCT GGCCTGTCTCAATGACATCCTCACAGATGCACTGCGCCATATCGAGCCCTGCCTTC AGTACATGGAGATGGTTTGGGACGAGGGCGTTTTTAAGTTCTGCGCCGTGCCAGA GCTCATGTCCTTGGCTACCATCTCGGTGTGTTACAACAATCCGAAGGTCTTCACAG GTGTTGTCAAGATGAGGAGGGGCGAAACAGCAAAGCTGTTTCTGAGCGTAACAAA TATGCCAGCTCTGTACAAGAGTTTTTCAGCCATTGCTGAAGAAATGGAGGCCAAGT GTGTGAGGGAGGATCCCAACTTTGCACTCACAGTCAAGCGGCTTCAGGATGTCCA GGCGTTATGCAAGGCAGGCCTAGCAAAATCAAATGGAAAGGTTTCAGCTAAGGGT GCTTAG Rhodobacter codon optimized SS GATGTCGACTAGGAGGAATATAAAATGGGCATGCTGCGGTGGGGCGTCGAGAGCC TCCAGAACCCGGACGAGCTGATCCCCGTGCTGCGGATGATCTATGCGGACAAGTT CGGCAAGATCAAGCCGAAGGACGAGGACCGCGGCTTCTGCTACGAGATCCTGAAC CTCGTCTCGCGGAGCTTCGCGATCGTGATCCAGCAGCTGCCGGCCCAGCTCCGCGA CCCCGTCTGCATCTTCTATCTGGTGCTCCGGGCGCTGGACACGGTCGAGGACGACA TGAAGATCGCCGCGACCACGAAGATCCCGCTGCTCCGCGACTTCTACGAGAAGAT CTCGGACCGCAGCTTCCGGATGACCGCCGGCGACCAGAAGGACTATATCCGCCTG CTCGACCAGTACCCGAAGGTGACGTCGGTCTTCCTGAAGCTCACCCCCCGCGAGCA GGAGATCATCGCGGACATCACGAAGCGGATGGGCAACGGCATGGCCGACTTCGTG

135

CACAAGGGCGTCCCGGACACCGTGGGCGACTACGACCTGTATTGCCATTACGTCG CGGGCGTGGTCGGCCTGGGCCTCTCGCAGCTGTTCGTGGCCAGCGGCCTGCAGTCG CCCAGCCTCACGCGCTCGGAGGACCTGAGCAACCACATGGGCCTGTTCCTCCAGA AGACCAACATCATCCGGGACTATTTCGAGGACATCAACGAGCTGCCGGCGCCCCG CATGTTCTGGCCGCGGGAGATCTGGGGCAAGTACGCGAACAACCTGGCCGAGTTC AAGGACCCCGCCAACAAGGCCGCGGCCATGTGCTGCCTGAACGAGATGGTCACGG ACGCGCTGCGCCATGCCGTGTATTGCCTCCAGTACATGTCGATGATCGAGGACCCG CAGATCTTCAACTTCTGCGCGATCCCCCAGACGATGGCCTTCGGCACCCTGTCGCT CTGCTATAACAACTACACGATCTTCACCGGCCCGAAGGCGGCCGTGAAGCTGCGT CGCGGCACCACGGCGAAGCTCATGTATACCTCGAACAACATGTTCGCGATGTACC GCCACTTCCTGAACTTCGCCGAGAAGCTCGAGGTCCGGTGCAACACGGAGACCTC GGAGGACCCCAGCGTGACCACGACCCTGGAGCACCTCCATAAGATCAAGGCCGCG TGCAAGGCGGGCCTGGCGCGGACCAAGGACGACACGTTCGACGAGCTGCGGAGCT GAGATGTCGACT

Rhodobacter codon optimized BS GATGTCGACTAGGAGGAATATAAAATGACCATGCACCAGGACCACGGCGTGATGA AGGACCTCGTCAAGCACCCCAACGAGTTCCCGTATCTGCTCCAGCTCGCGGCCACC ACCTATGGCAGCCCGGCCGCGCCGATCCCCAAGGAGCCCGACCGCGCCTTCTGCT ACAACACCCTGCACACGGTGTCGAAGGGCTTCCCGCGCTTCGTCATGCGGCTGCCG CAGGAGCTCCAGGACCCCATCTGCATCTTCTATCTGCTCCTGCGCGCGCTGGACAC CGTGGAGGACGACATGAACCTCAAGAGCGAGACGAAGATCAGCCTCCTGCGCGTC TTCCACGAGCATTGCTCGGACCGGAACTGGTCGATGAAGAGCGACTACGGCATCT ATGCCGACCTGATGGAGCGCTTCCCCCTCGTGGTCTCGGTGCTGGAGAAGCTCCCG CCCGCCACCCAGCAGACGTTCCGGGAGAACGTCAAGTACATGGGCAACGGCATGG CGGACTTCATCGACAAGCAGATCCTGACCGTGGACGAGTACGACCTCTATTGCCAT TACGTGGCCGGCAGCTGCGGCATCGCGGTCACCAAGGTGATCGTCCAGTTCAACCT GGCCACGCCGGAGGCGGACTCGTATGACTTCTCGAACAGCCTCGGCCTCCTGCTCC AGAAGGCCAACATCATCACGGACTACAACGAGGACATCAACGAGGAGCCGCGCC CCCGGATGTTCTGGCCGCAGGAAATCTGGGGCAAGTATGCCGAGAAGCTGGCGGA CTTCAACGAGCCCGAGAACATCGACACCGCCGTGAAGTGCCTGAACCACATGGTC ACGGACGCGATGCGCCATATCGAGCCGTCGCTCAAGGGCATGGTGTATTTCACCG ACAAGACGGTCTTCCGGGCCCTGGCGCTGCTCCTGGTGACCGCGTTCGGCCACCTG TCGACGCTCTACAACAACCCGAACGTGTTCAAGGAGAAGGTCCGCCAGCGGAAGG GCCGCATCGCCCGGCTGGTGATGTCGAGCCGCAACGTCCCCGGCCTCTTCCGGACC TGCCTGAAGCTCGCGAACAACTTCGAGAGCCGCTGCAAGCAGGAGACGGCCAACG ACCCGACCGTGGCGATGACGATCAAGCGCCTCCAGTCGATCCAGGCCACCTGCCG GGACGGCCTGGCGAAGTATGACACGCCGTCGGGCCTCAAGAGCTTCTGCGCCGCG CCGACCCCCACGAAGGGCGGCTCGGGCGGCGGCAGCGGCGGCGGCTCGGGCATG AAGCTGCGCGAGGTGCTCCAGCACCCGGGCGAGATCATCCCCCTCCTGCAGATGA TGGTCATGGCCTACCGCCGGAAGCGCAAGCCGCAGGACCCCAACCTGGCGTGGTG CTGGGAGACCCTCATCAAGGTGTCGCGGAGCTATGTGCTGGTCATCCAGCAGCTGC CGGAGGTCCTGCAGGACCCCATCTGCGTGAACTACCTGGTCCTCCGCGGCCTGGAC ACCCTCCAGGACGACATGGCCATCCCGGCGGAGAAGCGGGTGCCCCTCCTGCTCG ACTACTATAACCATATCGGCGACATCACGTGGAAGCCGCCCTGCGGCTATGGCCA GTACGTGGAGCTCATCGAGGAGTATCCCCGCGTCACCAAGGAGTTCCTGAAGCTC AACAAGCAGGACCAGCAGTTCATCACGGACATGTGCATGCGCCTGGGCGCGGAGA TGACCGTGTTCCTGAAGCGGGACGTGCTCACGGTCCCGGACCTGGACCTCTACGCC TTCACCAACAACGGCCCCGTGGCGATCTGCCTGACGAAGCTCTGGGTCGACCGCA

136

AGTTCGCCGACCCGAAGCTGCTCGACCGGGAGGACCTGTCGGGCCACATGGCGAT GTTCCTCGGCAAGATCAACGTGATCCGCGACATCAAGGAGGACGTCCTCGAGGAC CCGCCCCGGATCTGGTGGCCGAAAGAAATCTGGGGTAAGTACCTGAAGGACCTCC GCGACATCATCAAGCCCGAGTACCAGAAGGAGGCCCTGGCGTGCCTCAACGACAT CCTGACCGACGCCCTCCGCCATATCGAGCCGTGCCTGCAGTATATGGAGATGGTGT GGGACGAGGGCGTCTTCAAGTTCTGCGCCGTGCCGGAGCTGATGAGCCTCGCGAC CATCTCGGTGTGCTACAACAACCCCAAGGTCTTCACGGGCGTGGTCAAGATGCGCC GGGGCGAGACCGCCAAGCTGTTCCTCTCGGTGACGAACATGCCGGCGCTGTACAA GTCGTTCAGCGCCATCGCGGAGGAGATGGAGGCCAAGTGCGTGCGCGAGGACCCC AACTTCGCGCTGACCGTGAAGCGGCTCCAGGACGTCCAGGCGCTGTGCAAGGCCG GCCTCGCGAAGAGCAACGGCAAGGTGTCGGCCAAGGGCGCGTGAGCTGTCGACT

Rhodobacter codon optimized fps GCTGTCGACTAGGAGGAATATAAAATGCAGCCGCACCACCACCATAAGGAGGGC CGGATGCACAAGTTCACGGGCGTCAACGCCAAGTTCCAGCAGCCCGCGCTCCGCA ACCTGTCGCCCGTGGTGGTGGAGCGCGAGCGGGAGGAGTTCGTCGGCTTCTTCCCG CAGATCGTGCGCGACCTCACCGAGGACGGCATCGGCCACCCGGAGGTGGGCGACG CCGTGGCCCGGCTGAAGGAAGTGCTCCAGTACAACGCCCCGGGCGGAAAGTGCAA CCGCGGCCTGACCGTGGTGGCGGCCTATCGGGAGCTCTCGGGCCCGGGCCAGAAG GACGCGGAGAGCCTGCGCTGCGCGCTCGCCGTCGGCTGGTGCATCGAGCTGTTCC AGGCGTTCTTCCTCGTGGCCGACGACATCATGGACCAGTCGCTGACCCGTCGCGGC CAGCTCTGCTGGTACAAGAAGGAAGGCGTCGGCCTGGACGCGATCAACGACAGCT TCCTGCTCGAGTCGAGCGTCTACCGCGTGCTCAAGAAGTATTGCCGCCAGCGGCCG TACTATGTGCACCTGCTCGAGCTGTTCCTCCAGACCGCGTATCAGACGGAGCTGGG CCAGATGCTGGACCTCATCACGGCCCCGGTCTCGAAGGTGGACCTCTCGCATTTCA GCGAGGAGCGCTACAAGGCGATCGTGAAGTATAAGACGGCCTTCTACTCGTTCTA TCTGCCCGTGGCGGCCGCCATGTACATGGTGGGCATCGACAGCAAGGAAGAGCAC GAGAACGCGAAGGCCATCCTGCTCGAGATGGGCGAGTACTTCCAGATCCAGGACG ACTATCTGGACTGCTTCGGCGACCCCGCCCTCACGGGCAAGGTCGGCACGGACAT CCAGGACAACAAGTGCTCGTGGCTGGTCGTGCAGTGCCTGCAGCGCGTGACCCCC GAGCAGCGGCAGCTGCTCGAGGACAACTACGGCCGCAAGGAGCCCGAGAAGGTC GCGAAGGTGAAGGAGCTCTATGAGGCGGTGGGCATGCGGGCCGCCTTCCAGCAGT ACGAGGAGTCGAGCTATCGTCGCCTGCAGGAGCTCATCGAGAAGCATTCGAACCG GCTGCCCAAGGAGATCTTCCTCGGCCTCGCCCAGAAGATCTACAAGCGCCAGAAG TGAGATGTCGACT

Intergeneric Mating Between E. coli S17-1 and R. capsulatus SB1003

E. coli S17-1 (ATCC 47055) was used as a host for intergeneric conjugation of pBBR1MCS-2 derivatives into R. capsulatus SB1003 (Simon et al. 1983). In brief, E. coli S17-1 strains harboring the pBBR1MCS-2 derivative constructs was grown from a single colony in 1

137 mL of lytic broth (LB) (supplemented with 50 µg mL-1 kanamycin) to early log phase. In parallel, R. capsulatus SB1003 was grown in 1 mL of YCC (supplemented with 10 µg mL-1 rifampicin) for 24 hours before the conjugal mating experiment. The cells were spun down and washed twice with 1 mL of sterile LB, then the E. coli S17-1 donor and R. capsulatus recipient cells were each resuspended in 100 µL of sterile LB and mixed. The entire mating mixture was spot plated (10 µL) onto dried YCC agar plates and allowed to mate for 8-12 hours. Matings were scraped with a toothpick, resuspended in 200 µL of sterile LB, and serially diluted onto

YCC plates (supplemented with 10 µg mL-1 rifampicin and 20 µg mL-1 kanamycin) and incubated for 48 to 72 hours before exconjugants appeared. Exconjugants were picked with a sterile toothpick and inoculated in 1 mL of YCC supplemented with antibiotics for 20 hours so that all the strains were in early to middle stationary phase. These cultures were used to inoculate the shake flask and culture tube experiments.

Hydrocarbon Analysis

Bacterial culture and acetone were mixed in 1:1 volumetric ratio, either in a 4 mL or 2 mL HPLC vial and vortexed for 30 seconds to lyse the cells and to extract the triterpenes. 0.5 mL isooctane was then added to the mixture and vortexed for another 30 seconds to extract the triterpenes. The culture-acetone-isooctane mixture is then centrifuged for 5 minutes at >3000g.

Cell debris pellets at the bottom and liquid phases separate into acetone-water (bottom) and acetone-iso-octane (top) layers. 100 to 300 µL of the top layer is collected into glass inserts for injection (1 µl) into GC-FID for analysis. The injection port is equipped with an inlet liner having silica wool (deactivated) to trap the non-volatile components.

138 GC-FID analysis was conducted on an Agilent 7820A GC. Squalene standard was analyzed in concentrations ranging from 1 ng to 100 ng to generate standard curves.

139 Appendix E

Background calculations for Figure 4-5.

The calculations below provide the basis for carbon and energy balance depicted in

Figure 3. The three different trophic modes described in this work encompass very distinct carbon and energy sources. Aerobic heterotrophic uses glucose as both energy and carbon source to generate biomass and release CO2. Anaerobic photoheterotrophic growth uses light as the energy source to generate ATP which is used to assimilate the more oxidized carbon source malate into biomass. There is a small net release of CO2. Chemoautotrophic growth uses H2 as the energy source to fix CO2 and therefore causes a net uptake of CO2.

To place the source of energy on a common basis with assimilation, the energy utilized per ATP is calculated, then a conversion factor of 10.5 mol ATP consumed per gram-dry-weight and a biomass composition of 50% by weight carbon yields 2.28 mol-ATP/mol-C assimilated into biomass.

The energy sink is evaluated as simple heat of reaction of the carbon source to pyruvate, which is the most immediate common metabolite of central metabolism.

Heats of combustion are obtained from (Domalski 1972).

IV. Aerobic heterotrophic growth on glucose

Source – the energy content of the energy source per mol ATP

The oxidative phosphorylation of 1 mole of glucose via the electron transport chain generates 30 to 38 moles of ATP.

The heat of combustion of glucose = 2800.35 kJ/mol (Domalski 1972).

140 Therefore, energy required for 1 mol of ATP = 73.68 to 93.34 kJ/mol-ATP (Average =

83.51 kJ/mol-ATP)

Using a value of 10.5 mol-ATP/gDW (A. Tuerk 2011) (=2.28 mol-ATP/mol-C-fixed) we have 190.4 kJ/mol-C-fixed for aerobic heterotrophic growth on glucose.

Sink – energy released/consumed per mole of C fixed in pyruvate

→ 2 + 2

∆Hr = 2*(-587.02) – (-1271) = 96.96 kJ/mol-Glucose = 16.16 kJ/mol-C-pyruvate

V. Anaerobic photoheterotrophic growth on malate

Source – the energy content of the energy source per mol ATP

2 photons yield 1 mol of ATP (Berg et al. 2002)

.× .×. Energy of 1 mol of 800 nm photon (hc/λ) = = 149.8 kJ/mol- × photon.

Energy required for 1 mol ATP = 299.6 kJ/mol-ATP

Using a value of 10.5 mol-ATP/gDW(A. Tuerk 2011) (=2.28 mol-ATP/mol-C-fixed) we have 683.1 kJ/mol-C-fixed for anaerobic photoheterotrophic growth on malate.

Sink – energy released/consumed per mole of C fixed in pyruvate

→ +

∆Hr = (-587.02 – 393.51) – (-1103.66) = 123.13 kJ/mol-pyruvate = 41.04 kJ/mol-C- pyruvate.

141

VI. Autotrophic growth on H2

Source – the energy content of the energy source per mol ATP

Hydrogenases can generate 1 NADH/NADPH molecule upon oxidation of 1 H2 (Lubitz et al. 2014). The number of ATPs generated by NADH and NADPH range from 2.5 to 3.5

(Lodish et al. 2000; Van Walraven et al. 1996). For simplicity we are assuming it roughly equal to 3 i.e. 3ATP/H2.

Heat of combustion of H2 = 286 kJ/mol-H2

Energy consumed per ATP = 95.33 kJ/mol-ATP

Using a value of 10.5 mol-ATP/gDW(A. Tuerk 2011) (=2.28 mol-ATP/mol-biomass-C) we have 217.4 kJ/mol-C-fixed for autotrophic growth on H2.

Sink – energy released/consumed per mole of C fixed in pyruvate

3 3 + 2 → + 2

∆Hr = (-587.02 – 0) – (-393.51*3) = 593.5 kJ/mol-Pyr = 197.8 kJ/mol-C

VII. Carbon balance

Aerobic heterotrophic growth on glucose

The carbon yield on glucose (g-C-biomass/g-C-glucose) varies from 0.4 to 0.6 (McCarty

2007). Assuming an average of 0.5 g-C-biomass/g-C-glucose, the rest must leave as CO2.

Therefore, C-CO2-released/C-consumed = 0.5 or 50%.

142 Anaerobic photoheterotrophic growth on malate

Under photoheterotrophic growth we have measured the yield of biomass on malate to be

0.84±0.045 gC-biomass/gC-malate. Therefore, C-CO2-released/C-consumed = 0.26 or 26%.

Autotrophic growth on H2

Since there is no other sink for CO2 consumed than end up in biomass, essentially 100% of the CO2 consumed is captured into biomass. That is, C-CO2-released/C-consumed = -100%.

143 Appendix F

Other miscellaneous work

This section includes works that had been started but discontinued due change of priorities and lack/limitation of resources.

VIII. Small scale multiplexed autotrophic screening device.

It was desirable to have the ability to grow many small autotrophic cultures at once to allow for screening of multiple genetic constructs. A volume of one mL was sufficient for chemical hydrocarbon screening, therefore the design was implemented in test tubes with a working culture volume of less than 5 mL. We developed a robust optical density monitoring system based on an LED and a photodiode that was placed in a solid rack in which the test-tubes sat (Figure F1A). CO2, H2 and O2 gasses were precisely mixed using a bank of mass flow controllers, and this gas was bubbled into the test tubes through a stainless steel tubing. To minimize problems of daisy-chaining of gas delivery, a means of manifolding gas delivery was devised based on pressure drop through precise ruby orifices (RB#22104, Bird Precision,

Waltham, MA). Since the pressure drop in the orifices was larger than in the device, the source pressure (and flow through each orifice) could be made reasonably accurate at a minimum cost compared other means of multiplexed gas delivery. To improve the accuracy of OD measurements, the gas delivery could be stopped briefly with a solenoid just prior to making measurements of the LED-photodiode voltage. All of this was interfaced through a multi-channel analog input card with monitoring and control provided by custom programming in LabView software. A multiplexed growth assessment is shown in Figure F1B illustrating the ability to

144 monitor the kinetics of growth online for growth under autotrophic conditions. Ryan Johnson,

Reed Taylor and Bill Muzika helped extensively with this effort.

Figure F1. (Top) A multiplexed autotrophic growth device design for continuous monitoring of growth using LED/phododiode assemblies at the base of the culture. Gas inlet to multiplexed cultures is achieved with a very small sapphire orifice manifold. (Bottom) Examples of online optical density collected from the screening device while growing different strains of R. capsulatus.

145 IX. 100-L plastic bag trickle-bed reactor.

To demonstrate the low capital and operating cost at a higher scale, we constructed a 100-

L capacity plastic trickle-bed reactor. Our literature review on large-scale reactors showed that trickle-bed reactors have the lowest power-per-volume requirements for gas-liquid mass transfer coefficient (kLa) and can achieve substantially high kLa at large scales (Figure 2-5). Combined with the low capital and operating cost, scalability and low explosion hazard, plastic trickle-beds appear to be ideal for such a process. Dr. Wayne Curtis already has a patent for such a reactor system (T. Y. Hsiao et al. 1999; Curtis 2004). Therefore, in anticipation of much higher levels of production with current genetic engineering efforts, we constructed a 100 L plastic bag trickle- bed reactor (working volume of 10 L; Figure F2) for demonstration purposes. It is made almost entirely of plastic materials, which not only reduces the explosion hazard, but is extremely low cost by several orders of magnitude compared to traditional stirred tank bioreactors. It is built within an explosion bay in Fenske laboratory and all the necessary safety features in place, in order to prevent damages in case of explosion. We have been able to significantly improve the kLa of this reactor by altering the packing, spray nozzles etc.

146

Figure F2. A 100-L low cost plastic bag trickle bed that has been assembled to explore regions of high mass transfer efficacy operation for autotrophic growth.

X. Secretion of hydrocarbon by R. capsulatus

We also had the interesting observation that when R. capsulatus was grown to these densities (> 2 gDW/L) we could detect accumulation of botryococcene in the aqueous phase of the culture (Figure F3). This was observed in repeated experiments. It is, however, deemed highly unlikely that a hydrophobic molecule such as botryococcene would simply be excreted by the cells and remain in the aqueous phase. Our hypothesis in this case was that botryococcene is somehow incorporated into the carotenoid-based pigment-protein complex of Rhodobacter and released as particles in the aqueous phase. Part of the reason for this idea is that at relatively higher density the culture was highly pigmented and the hydrocarbon extract (from the regular extraction method used throughout the project) became increasingly pigmented with increasing cell density. We could also found a previous report in the literature where the incorporation of a non-native hydrocarbon species, β-carotene, was shown to be incorporated in the light harvesting

147 complex of R. sphaeroides. However, we did not pursue this observation because of other priorities.

Figure F3. Hydrocarbon secretion in R. capsulatus autotrophic bioreactor cultures. Triterpene levels in cells, media and total culture during late lag and stationary phase of two independent batch cultures. Cultures were spun down and supernatant and pellet were separately analyzed. Samples of the whole culture was also analyzed separately. The blue columns indicate the amount of triterpene found in the supernatant of the cultures while the pink portion represents the triterpene in the culture pellet. The red columns indicate the total triterpene. Error bars indicate deviation of two replicates.

148 References

Ahn, W.S., Park, S.J. & Lee, S.Y., 2000. Production of Poly(3-hydroxybutyrate) by fed-batch culture of recombinant Escherichia coli with a highly concentrated whey solution. Applied and environmental microbiology, 66(8), pp.3624–7. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=92194&tool=pmcentrez&render type=abstract [Accessed July 13, 2014].

Al-Hinai, M. a, Fast, A.G. & Papoutsakis, E.T., 2012. Novel system for efficient isolation of clostridium double-crossover allelic exchange mutants enabling markerless chromosomal gene deletions and DNA integration. Applied and environmental microbiology, 78(22), pp.8112–21.

Alper, H., Miyaoku, K. & Stephanopoulos, G., 2005. Construction of lycopene-overproducing E. coli strains by combining systematic and combinatorial gene knockout targets. Nature biotechnology, 23(5), pp.612–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15821729 [Accessed February 1, 2014].

Badger, M.R. & Bek, E.J., 2008. Multiple Rubisco forms in proteobacteria: their functional significance in relation to CO2 acquisition by the CBB cycle. Journal of experimental botany, 59(7), pp.1525–41. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18245799 [Accessed July 9, 2014].

Banerjee, a & Sharkey, T.D., 2014. Methylerythritol 4-phosphate (MEP) pathway metabolic regulation. Natural product reports, 31(8), pp.1043–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24921065 [Accessed July 18, 2014].

Banerjee, A. et al., 2014. A Lactose-Inducible System for Metabolic Engineering of Clostridium ljungdahlii.,

Banta, S. & West, A., 2014. Electrofuel production using genetically engineered Acidithiobacillus ferrooxidans. In Electrochemical Society Meeting Abstracts. The Society for Solid-state and Electrochemical Science and Technology. Available at: http://ma.ecsdl.org/content/MA2014-02/50/2290.abstract.

Barbosa, M.J. et al., 2005. Optimization of biomass, vitamins, and carotenoid yield on light energy in a flat-panel reactor using the A-stat technique. Biotechnology and bioengineering, 89(2), pp.233–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15593095.

Bar-Even, A., Noor, E. & Milo, R., 2012. A survey of carbon fixation pathways through a quantitative lens. Journal of experimental botany, 63(6), pp.2325–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22200662 [Accessed July 31, 2014].

Barnard, G.C. et al., 2004. High level recombinant protein expression in Ralstonia eutropha using T7 RNA polymerase based amplification. Protein expression and purification, 38(2), pp.264–71. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15555942 [Accessed July 15, 2014].

149 Bartels, J.R., Pate, M.B. & Olson, N.K., 2010. An economic survey of hydrogen production from conventional and alternative energy sources. International Journal of Hydrogen Energy, 35(16), pp.8371–8384. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0360319910007317 [Accessed August 1, 2014].

Beekwilder, J. et al., 2014a. Valencene synthase from the heartwood of Nootka cypress (Callitropsis nootkatensis) for biotechnological production of valencene. Plant biotechnology journal, 12(2), pp.174–82. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24112147 [Accessed June 3, 2014].

Beekwilder, J. et al., 2014b. Valencene synthase from the heartwood of Nootka cypress (Callitropsis nootkatensis) for biotechnological production of valencene. Plant biotechnology journal, 12, pp.174–182. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24112147 [Accessed January 20, 2014].

Bengelsdorf, F.R., Straub, M. & Dürre, P., 2013. Bacterial synthesis gas (syngas) fermentation. Environmental technology, 34(13-16), pp.1639–51. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24350425 [Accessed September 5, 2014].

Berg, I. a, 2011. Ecological aspects of the distribution of different autotrophic CO2 fixation pathways. Applied and environmental microbiology, 77(6), pp.1925–36. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3067309&tool=pmcentrez&ren dertype=abstract [Accessed May 25, 2013].

Berg, J.M., Tymoczko, J.L. & Stryer, L., 2002. Biochemistry 5th ed., New York, NY: W. H. Freeman. Available at: http://www.ncbi.nlm.nih.gov/books/NBK22519/.

Berzin, V., Kiriukhin, M. & Tyurin, M., 2012. Selective production of acetone during continuous synthesis gas fermentation by engineered biocatalyst Clostridium sp. MAceT113. Letters in applied microbiology, 55(2), pp.149–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22642684 [Accessed July 9, 2014].

Berzin, V., Tyurin, M. & Kiriukhin, M., 2013. Selective n-butanol production by Clostridium sp. MTButOH1365 during continuous synthesis gas fermentation due to expression of synthetic thiolase, 3-hydroxy butyryl-CoA dehydrogenase, crotonase, butyryl-CoA dehydrogenase, butyraldehyde dehydrogenase, and. Applied biochemistry and biotechnology, 169(3), pp.950–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23292245 [Accessed July 9, 2014].

Bi, C. et al., 2013. Development of a broad-host synthetic biology toolbox for Ralstonia eutropha and its application to engineering hydrocarbon biofuel production. Microbial cell factories, 12, p.107. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3831590&tool=pmcentrez&ren dertype=abstract.

Van Bodegom, P., 2007. Microbial maintenance: a critical review on its quantification. Microbial ecology, 53(4), pp.513–23. Available at:

150 http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1915598&tool=pmcentrez&ren dertype=abstract [Accessed April 1, 2012].

Bongers, L., 1970. Energy generation and utilization in hydrogen bacteria. Journal of bacteriology, 104(1), pp.145–51. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=248194&tool=pmcentrez&rende rtype=abstract.

Cantet, J., Bergel, A. & Comtat, M., 1992. Kinetics of the catalysis by the Alcaligenes eutrophus H16 Kinetics of the catalysis by the Alcaligenes eutrophus H16 hydrogenase of the electrochemical reduction of NAD+. Journal of molecular catalysis, 73(3), pp.371–380. Available at: http://www.sciencedirect.com/science/journal/03045102/73.

De Castro, C. et al., 2014. A top-down approach to assess physical and ecological limits of biofuels. Energy, 64, pp.506–512. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0360544213009018 [Accessed May 14, 2014].

Curtis, B., 2011. Development of Rhodobacter for the Production of Functional Membrane Proteins. The Pennsylvania State University.

Curtis, W.R., 2004. Growing cells in a reservoir formed of a flexible sterile plastic liner. Available at: http://www.google.com/patents/US6709862.

Daniell, J., Köpke, M. & Simpson, S., 2012. Commercial Biomass Syngas Fermentation, Available at: http://www.mdpi.com/1996-1073/5/12/5372/ [Accessed August 19, 2013].

Delamarre, S.C. & Batt, C. a, 2006. Comparative study of promoters for the production of polyhydroxyalkanoates in recombinant strains of Wautersia eutropha. Applied microbiology and biotechnology, 71(5), pp.668–79. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16362422 [Accessed July 16, 2014].

Demirel, B. & Scherer, P., 2008. The roles of acetotrophic and hydrogenotrophic methanogens during anaerobic conversion of biomass to methane: a review. Reviews in Environmental Science and Bio/Technology, 7(2), pp.173–190. Available at: http://link.springer.com/10.1007/s11157-008-9131-1 [Accessed July 11, 2014].

Denholm, P. & Margolis, R., 2006. Very Large-Scale Deployment of Grid-Connected Solar Photovoltaics in the United States : Challenges and Opportunities Preprint. , (April).

Derenne, S. et al., 1997. Chemical structure of the organic matter in a Pliocene maar-type shale: Implicated Botryococcus race strains and formation pathways. Geochimica et Cosmochimica Acta, 61(9), pp.1879–1889. Available at: http://www.sciencedirect.com/science/article/pii/S0016703797000422 [Accessed June 1, 2014].

Ditta, G. et al., 1985. Plasmids related to the broad host range vector, pRK290, useful for gene cloning and for monitoring gene expression. Plasmid, 13(2), pp.149–153. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2987994.

151 Domalski, E.S., 1972. Selected Values of Heats of Combustion and Heats of Formation of Organic Compounds Containing the Elements C, H, N, O, P, and S. Journal of Physical and Chemical Reference Data, 1(2), p.221. Available at: http://scitation.aip.org/content/aip/journal/jpcrd/1/2/10.1063/1.3253099 [Accessed October 20, 2014].

Drake, H.L. & Daniel, S.L., 2004. Physiology of the thermophilic acetogen Moorella thermoacetica. Research in microbiology, 155(10), pp.869–83.

Dumsday, G.J. et al., 1999. Comparative stability of ethanol production by Escherichia coli KO11 in batch and chemostat culture. Journal of Industrial Microbiology and Biotechnology, 23(1), pp.701–708. Available at: http://link.springer.com/10.1038/sj.jim.2900690 [Accessed January 9, 2015].

Fang, L. et al., 2012. Expression of the Chlamydomonas reinhardtii sedoheptulose-1,7- bisphosphatase in Dunaliella bardawil leads to enhanced photosynthesis and increased glycerol production. Plant biotechnology journal, 10(9), pp.1129–35. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22998361 [Accessed October 13, 2014].

Farmer, W.R. & Liao, J.C., 2001. Precursor balancing for metabolic engineering of lycopene production in Escherichia coli. Biotechnology progress, 17(1), pp.57–61. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11170480.

Fast, A.G. & Papoutsakis, E.T., 2012. Stoichiometric and energetic analyses of non- photosynthetic CO2-fixation pathways to support synthetic biology strategies for production of fuels and chemicals. Current Opinion in Chemical Engineering, 1(4), pp.380–395. Available at: http://linkinghub.elsevier.com/retrieve/pii/S2211339812000457 [Accessed August 17, 2014].

Flusberg, B. a et al., 2010. Direct detection of DNA methylation during single-molecule, real- time sequencing. Nature methods, 7(6), pp.461–5.

Friedrich, B. & Schwartz, E., 1993. Molecular biology of hydrogen utilization in aerobic chemolithotrophs. Annual review of microbiology, 47, pp.351–83. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8257102 [Accessed January 7, 2015].

Gerlitz, M., Hrabak, O. & Schwab, H., 1990. Partitioning of Broad-Host-Range Plasmid RP4 Is a Complex System Involving Site-Specific Recombinationt. , 172(11), pp.6194–6203.

Gleiser, I. & Bauer, S., 1981. Growth of E. coli W to high cell concentration by oxygen level linked control of carbon source concentration. Biotechnology and Bioengineering, XXIII(1981), pp.1015–1021. Available at: http://onlinelibrary.wiley.com/doi/10.1002/bit.260230509/abstract [Accessed March 26, 2014].

Glikson, M., Lindsay, K. & Saxby, J., 1989. Botryococcus—A planktonic green alga, the source of petroleum through the ages: Transmission electron microscopical studies of oil shales and petroleum source rocks. Organic Geochemistry, 14(6), pp.595–608. Available at:

152 http://www.sciencedirect.com/science/article/pii/0146638089900399 [Accessed June 2, 2014].

Grousseau, E. et al., 2014. Isopropanol production with engineered Cupriavidus necator as bioproduction platform. Applied microbiology and biotechnology, 98(9), pp.4277–90. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24604499 [Accessed July 13, 2014].

Gruber, S. et al., 2014. Versatile and stable vectors for efficient gene expression in Ralstonia eutropha H16. Journal of biotechnology, pp.1–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24998763 [Accessed July 10, 2014].

Gruenspecht, H., 2012. Annual Energy Outlook 2012, Available at: http://www.eia.gov/forecasts/aeo/pdf/0383%282012%29.pdf [Accessed February 24, 2013].

Guadalupe-Medina, V. et al., 2013. Carbon dioxide fixation by Calvin-Cycle enzymes improves ethanol yield in yeast. Biotechnology for biofuels, 6(1), p.125. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3766054&tool=pmcentrez&ren dertype=abstract [Accessed October 5, 2014].

Hammerschlag, R., 2006. Ethanol’s Energy Return on Investment: A Survey of the Literature 1990−Present. Environmental Science & Technology, 40(6), pp.1744–1750. Available at: http://pubs.acs.org/doi/abs/10.1021/es052024h.

Heap, J.T. et al., 2009. A modular system for Clostridium shuttle plasmids. Journal of microbiological methods, 78(1), pp.79–85.

Hildebrand, A. et al., 2013. Engineering Escherichia coli for improved ethanol production from gluconate. Journal of biotechnology, 168(1), pp.101–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23942377 [Accessed June 10, 2014].

Hillen, L.W. et al., 1982. Hydrocracking of the oils of Botryococcus braunii to transport fuels. Biotechnology and bioengineering, 24(1), pp.193–205. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18546110.

Ho, K.P. & Payne, W.J., 1979. Assimilation efficiency and energy contents of prototrophic bacteria. Biotechnology and Bioengineering, 21(5), pp.787–802. Available at: http://doi.wiley.com/10.1002/bit.260210505.

Hoekema, S., Douma, R.D.R.D., et al., 2006. Controlling light-use by Rhodobacter capsulatus continuous cultures in a flat-panel photobioreactor. Biotechnology and bioengineering, 95(4), pp.613–626. Available at: http://onlinelibrary.wiley.com/doi/10.1002/bit.20907/pdf.

Hoekema, S., Douma, R.D., et al., 2006. Controlling light-use by Rhodobacter capsulatus continuous cultures in a flat-panel photobioreactor. Biotechnology and bioengineering, 95(4), pp.613–626. Available at: http://onlinelibrary.wiley.com/doi/10.1002/bit.20907/pdf [Accessed July 24, 2011].

153 Howells, E.R., 1982. Single-cell protein and related technology. Chemistry Industry, (15), pp.508–511.

Hsiao, T. et al., 1999. Development of a low capital investment reactor system: application for plant cell suspension culture. Biotechnology progress, 15(1), pp.114–22. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9933521 [Accessed November 15, 2010].

Hsiao, T.Y. et al., 1999. Development of a low capital investment reactor system: application for plant cell suspension culture. Biotechnology progress, 15(1), pp.114–122. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9933521 [Accessed November 15, 2010].

Hu, P., Rismani-Yazdi, H. & Stephanopoulos, G., 2013. Anaerobic CO2 fixation by the acetogenic bacterium Moorella thermoacetica. AIChE Journal, 59(9), pp.3176–3183. Available at: http://doi.wiley.com/10.1002/aic.14127 [Accessed January 10, 2015].

Hunter, C.N. et al., 2009. Advances in Photosynthesis and Respiration. Volume 28: The Purple Phototrophic Bacteria, Dordrecht, The Netherlands: Springer. Available at: http://medcontent.metapress.com/index/A65RM03P4874243N.pdf [Accessed November 12, 2013].

IEA, 2013. Resources to reserves, Available at: http://www.iea.org/publications/freepublications/publication/Resources2013_free.pdf.

Imam, S., Noguera, D.R. & Donohue, T.J., 2013. Global insights into energetic and metabolic networks in Rhodobacter sphaeroides. BMC systems biology, 7, p.89. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3849096&tool=pmcentrez&ren dertype=abstract [Accessed May 27, 2014].

Ind, A.C. et al., 2009. Inducible-expression plasmid for Rhodobacter sphaeroides and Paracoccus denitrificans. Applied and environmental microbiology, 75(20), pp.6613–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19684165.

Inui, M. et al., 2003. Isolation and molecular characterization of pMG160, a mobilizable cryptic plasmid from Rhodobacter blasticus. Applied and environmental microbiology, 69, pp.725– 733.

IRENA, 2012. Renewable Power Generation Costs in 2012 : An Overview, Available at: http://costing.irena.org/media/2769/Overview_Renewable-Power-Generation-Costs-in- 2012.pdf.

Ishizaki, A. & Tanaka, K., 1990. Batch culture of Alcaligenes eutrophus ATCC 17697T using recycled gas closed circuit culture system. Journal of fermentation and bioengineering, 69(3), pp.170–174. Available at: http://linkinghub.elsevier.com/retrieve/pii/0922338X9090041T [Accessed July 6, 2011].

Ivy, J., 2004. Summary of Electrolytic Hydrogen Production: Milestone Completion Report. , (September). Available at:

154 http://www.osti.gov/energycitations/product.biblio.jsp?osti_id=15007167 [Accessed March 31, 2013].

Iwasaki, Y. et al., 2013. Engineering of a functional thermostable kanamycin resistance marker for use in Moorella thermoacetica ATCC39073. FEMS microbiology letters, 343(1), pp.8– 12.

Jaschke, P.R. et al., 2011. Modification of the genome of Rhodobacter sphaeroides and construction of synthetic operons 1st ed., Elsevier Inc. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21601102 [Accessed January 3, 2015].

Jiang, Y. et al., 2013. Bioelectrochemical systems for simultaneously production of methane and acetate from carbon dioxide at relatively high rate. International Journal of Hydrogen Energy, 38, pp.3497–3502.

Jun, D. et al., 2013. Use of new strains of Rhodobacter sphaeroides and a modified simple culture medium to increase yield and facilitate purification of the reaction centre. Photosynthesis Research, pp.1–9.

Kaserer, H., 1906. Die oxydation des Wasserstoffes durch Mikroorganismen. Zent. Bakt. Par. II, 16, p.681.

Katzke, N. et al., 2010. A novel T7 RNA polymerase dependent expression system for high-level protein production in the phototrophic bacterium Rhodobacter capsulatus. Protein expression and purification, 69(2), pp.137–46. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19706327 [Accessed July 6, 2011].

Keller, M.W. et al., 2013. Exploiting microbial hyperthermophilicity to produce an industrial chemical, using hydrogen and carbon dioxide. Proceedings of the National Academy of Sciences of the United States of America, 110(15), pp.5840–5. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3625313&tool=pmcentrez&ren dertype=abstract [Accessed October 18, 2014].

Kelly, D.P., 1990. Energetics of Chemolithotrophs. In T. A. Krulwich, ed. THE BACTERIA: A TREATISE ON STRUCTURE AND FUNCTION Volume XII. London: Academic Press, Ltd., pp. 479–503.

Khan, N. et al., Triterpene hydrocarbon production engineered into a metabolically versatile host – Rhodobacter capsulatus. Biotechnology and bioengineering.

Khan, N.E. et al., 2014. A process economic assessment of hydrocarbon biofuels production using chemoautotrophic organisms. Bioresource technology, 172, pp.201–211. Available at: http://www.sciencedirect.com/science/article/pii/S0960852414012346 [Accessed September 8, 2014].

Khan, N.E. et al., 2013. Progress and economic considerations for the biological production of triterpene biofuels from gases. In AICHE 2013 Annual Meeting. Pittsburgh, PA, pp. 1–7.

155 Available at: https://aiche.confex.com/aiche/2013/webprogram/Paper343452.html [Accessed September 8, 2014].

Khatri, W. et al., 2014. Hydrocarbon production in high density Botryococcus braunii race B continuous culture. Biotechnology and bioengineering, 111(3), pp.493–503. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24122424 [Accessed February 17, 2014].

Khunjar, W.O. et al., 2012. Biomass production from electricity using ammonia as an electron carrier in a reverse microbial fuel cell. PLoS ONE, 7.

Kichise, T. et al., 1999. Biosynthesis of polyhydroxyalkanoates (PHA) by recombinant Ralstonia eutropha and effects of PHA synthase activity on in vivo PHA biosynthesis. In International Journal of Biological Macromolecules. pp. 69–77.

Kim, S.-W. et al., 2006. Over-production of beta-carotene from metabolically engineered Escherichia coli. Biotechnology letters, 28(12), pp.897–904. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16786275 [Accessed June 13, 2014].

Kiriukhin, M. & Tyurin, M., 2013. Expression of amplified synthetic ethanol pathway integrated using Tn7-tool and powered at the expense of eliminated pta, ack, spo0A and spo0J during continuous syngas or CO2 /H2 blend fermentation. Journal of applied microbiology, 114(4), pp.1033–45. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23289641 [Accessed July 9, 2014].

Kiriukhin, M. & Tyurin, M., 2014. Mevalonate production by engineered acetogen biocatalyst during continuous fermentation of syngas or CO/H blend. Bioprocess and biosystems engineering, 37(2), pp.245–60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23775000 [Accessed July 9, 2014].

Kita, A. et al., 2012. Development of genetic transformation and heterologous expression system in carboxydotrophic thermophilic acetogen Moorella thermoacetica. Journal of Bioscience and Bioengineering, xx(xx), pp.1–6.

Kontur, W.S., Noguera, D.R. & Donohue, T.J., 2012. Maximizing reductant flow into microbial H2 production. Current opinion in biotechnology, 23(3), pp.382–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22036711 [Accessed January 7, 2015].

Köpke, M., Mihalcea, C., Liew, F., et al., 2011. 2,3-Butanediol production by acetogenic bacteria, an alternative route to chemical synthesis, using industrial waste gas. Applied and environmental microbiology, 77(15), pp.5467–75. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3147483/.

Köpke, M. et al., 2010. Clostridium ljungdahlii represents a microbial production platform based on syngas. Proceedings of the National Academy of Sciences of the United States of America, 107(29), pp.13087–92. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2919952&tool=pmcentrez&ren dertype=abstract [Accessed November 10, 2013].

156 Köpke, M., Mihalcea, C., Bromley, J.C., et al., 2011. Fermentative production of ethanol from carbon monoxide. Current opinion in biotechnology, 22(3), pp.320–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21353524 [Accessed October 13, 2014].

Kovach, M.E. et al., 1995. Four new derivatives of the broad-host-range cloning vector pBBR1MCS, carrying different antibiotic-resistance cassettes. Gene, 166(1), pp.175–176. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8529885.

Kusian, B. et al., 1995. Characterization of the duplicate ribulose-1,5-bisphosphate carboxylase genes and cbb promoters of Alcaligenes eutrophus. Journal of bacteriology, 177, pp.4442– 4450.

Lawrence, J.G. & Roth, J.R., 1996. Selfish Operons: Horizontal Transfer May Drive the Evolution of Gene Clusters. , (3).

Leang, C. et al., 2013. A genetic system for Clostridium ljungdahlii: a chassis for autotrophic production of biocommodities and a model homoacetogen. Applied and environmental microbiology, 79(4), pp.1102–9.

Lee, S.Y., 1996. High cell-density culture of Escherichia coli. Trends in biotechnology, 14(3), pp.98–105. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8867291.

Lefebvre, S. et al., 2005. Increased Sedoheptulose-1 , 7-Bisphosphatase Activity in Transgenic Tobacco Plants Stimulates Photosynthesis and Growth from an Early Stage in Development 1. , 138(May), pp.451–460.

Li, H. et al., 2012. Integrated electromicrobial conversion of CO2 to higher alcohols. Science, 335(6076), p.1596. Available at: http://www.sciencemag.org/cgi/doi/10.1126/science.1217643 [Accessed March 30, 2012].

Li, H. & Liao, J.C., 2014. A synthetic anhydrotetracycline-controllable gene expression system in Ralstonia eutropha H16. ACS synthetic biology. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24702232.

Li, X. et al., 2014. Addition of citrate to Acidithiobacillus ferrooxidans cultures enables precipitate-free growth at elevated pH and reduces ferric inhibition. Biotechnology and bioengineering, 111(10), pp.1940–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24771134 [Accessed January 5, 2015].

Liew, F.M., Köpke, M. & Simpson, S.D., 2013. Gas fermentation for commercial biofuels production. In Z. Fang, ed. Liquid, Gaseous and Solid Biofuels - Conversion Techniques.

Lodish, H. et al., 2000. Molecular Cell Biology 4th ed., New York, NY: W. H. Freeman. Available at: http://www.ncbi.nlm.nih.gov/books/NBK21528/.

Lovley, D.R. & Nevin, K.P., 2013. Electrobiocommodities: powering microbial production of fuels and commodity chemicals from carbon dioxide with electricity. Current opinion in

157 biotechnology, 24(3), pp.385–90. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23465755 [Accessed December 4, 2014].

Lu, J. et al., 2012. Studies on the production of branched-chain alcohols in engineered Ralstonia eutropha. Applied Microbiology and Biotechnology, 96, pp.283–297.

Lubitz, W. et al., 2014. Hydrogenases. Chemical reviews, 114(8), pp.4081–148. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24655035 [Accessed October 20, 2014].

Lütte, S. et al., 2012. Autotrophic production of stable-isotope-labeled arginine in Ralstonia eutropha strain H16. Applied and environmental microbiology, 78(22), pp.7884–90. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3485953&tool=pmcentrez&ren dertype=abstract [Accessed January 9, 2015].

Lutz, R. & Bujard, H., 1997. Independent and tight regulation of transcriptional units in Escherichia coli via the LacR/O, the TetR/O and AraC/I1-I2 regulatory elements. Nucleic acids research, 25(6), pp.1203–10. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=146584&tool=pmcentrez&rende rtype=abstract [Accessed December 23, 2014].

Madigan, M. & Gest, H., 1979. Growth of the Photosynthetic Bacterium Rhodopseudomonas capsulata Chemoautotrophically in Darkness with H2 as the Energy Source. Journal of bacteriology, 137(1), pp.524–530. Available at: http://jb.asm.org/content/137/1/524.short [Accessed April 5, 2013].

Madigan, M.T. & Gest, H., 1979. Growth of the photosynthetic bacterium Rhodopseudomonas capsulata chemoautotrophically in darkness with H2 as the energy source. Journal of bacteriology, 137(1), pp.524–30. Available at: http://jb.asm.org/cgi/content/abstract/137/1/524 [Accessed November 22, 2011].

Marshall, C.W. et al., 2012. Electrosynthesis of commodity chemicals by an autotrophic microbial community. Applied and environmental microbiology, 78(23), pp.8412–20. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3497389&tool=pmcentrez&ren dertype=abstract [Accessed July 11, 2014].

Martin, V.J.J. et al., 2003. Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nature biotechnology, 21(7), pp.796–802. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12778056.

Marx, C.J. & Lidstrom, M.E., 2002. Broad-host-range cre-lox system for antibiotic marker recycling in gram-negative bacteria. BioTechniques, 33(5), pp.1062–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12449384.

Mather, M.W., McReynolds, L.M. & Yu, C.A., 1995. An enhanced broad-host-range vector for gram-negative bacteria: avoiding tetracycline phototoxicity during the growth of photosynthetic bacteria. Gene, 156, pp.85–88.

158 Mattozzi, M.D. et al., 2013. Expression of the sub-pathways of the Chloroflexus aurantiacus 3- hydroxypropionate carbon fixation bicycle in E. coli: Toward horizontal transfer of autotrophic growth. Metabolic engineering, 16, pp.130–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23376595 [Accessed October 2, 2014].

Mayer, F. & Müller, V., 2014. Adaptations of anaerobic archaea to life under extreme energy limitation. FEMS microbiology reviews, 38(3), pp.449–72.

McCarty, P.L., 1971. Energetics and Bacterial Growth. In S. J. Faust & J. V. Hunter, eds. Organic Compounds in Aquatic Environment. New York, NY: Marcel Dekker, Inc., pp. 495–531.

Mccarty, P.L., 1971. Engetics and Bacterial Growth. In S. J. Faust & J. V. Hunter, eds. Organic Compounds in Aquatic Environment. New York, NY: Marcel Dekker, Inc., pp. 495–531.

McCarty, P.L., 2007. Thermodynamic electron equivalents model for bacterial yield prediction: Modifications and comparative evaluations. Biotechnology and bioengineering, 97(2), pp.377–388. Available at: http://onlinelibrary.wiley.com/doi/10.1002/bit.21250/abstract [Accessed August 13, 2011].

McKinlay, J.B. & Harwood, C.S., 2010. Carbon dioxide fixation as a central redox cofactor recycling mechanism in bacteria. Proceedings of the National Academy of Sciences of the United States of America, 107, pp.11669–11675.

Melillo, Jerry M., Terese (T.C.) Richmond, and Gary W. Yohe, E., 2014. Climate Change Impacts in the United States: The Third National Climate Assessment., Available at: nca2014.globalchange.gov\nThis.

Metz, B. et al., 2005. IPCC special report on carbon dioxide capture and storage B. Metz et al., eds., New York, NY: Cambridge University Press. Available at: https://www.ipcc.ch/pdf/special-reports/srccs/srccs_wholereport.pdf [Accessed May 9, 2014].

Metzger, P. & Largeau, C., 2005. Botryococcus braunii: a rich source for hydrocarbons and related ether lipids. Applied microbiology and biotechnology, 66(5), pp.486–496. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15630516 [Accessed March 13, 2012].

Miller, B., Heuser, T. & Zimmer, W., 2000. Functional involvement of a deoxy-D-xylulose 5- phosphate reductoisomerase gene harboring locus of Synechococcus leopoliensis in isoprenoid biosynthesis. FEBS letters, 481(3), pp.221–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11007968.

Miyagawa, Y., Tamoi, M. & Shigeoka, S., 2001. Overexpression of a cyanobacterial fructose-1,6- /sedoheptulose-1,7-bisphosphatase in tobacco enhances photosynthesis and growth. Nature biotechnology, 19(10), pp.965–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11581664 [Accessed October 13, 2014].

159 Morinaga, Y. et al., 1978. Growth Characteristics and Cell Composition of Alcaligenes eutrophus in Chemostat Cultures. Agricultural and biological chemistry, 42(2), pp.439–444.

Mueller-Cajar, O. & Whitney, S.M., 2008. Directing the evolution of Rubisco and Rubisco activase: first impressions of a new tool for photosynthesis research. Photosynthesis research, 98(1-3), pp.667–75. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2758363&tool=pmcentrez&ren dertype=abstract [Accessed August 28, 2014].

Müller, J. et al., 2013. Engineering of Ralstonia eutropha H16 for autotrophic and heterotrophic production of methyl ketones. Applied and environmental microbiology, 79(14), pp.4433–9. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3697500&tool=pmcentrez&ren dertype=abstract.

Murray, I. a et al., 2012. The methylomes of six bacteria. Nucleic acids research, 40(22), pp.11450–62.

Myers, J., 2013. The scaled-up process design and economic analysis of an electrofuels platform for producing botryococcene as an alterantive transportation fuel. The Pennsylvania State University. Available at: https://honors.libraries.psu.edu/paper/17879/.

Nebraska, 2014. Ethanol Facilities Capacity by State and Plant. Available at: http://www.neo.ne.gov/statshtml/122.htm.

Nevin, K.P. et al., 2011. Electrosynthesis of organic compounds from carbon dioxide is catalyzed by a diversity of acetogenic microorganisms. Applied and environmental microbiology, 77(9), pp.2882–6. Available at: http://aem.asm.org/content/77/9/2882.short [Accessed October 14, 2014].

Nevin, K.P. et al., 2010. Microbial electrosynthesis: feeding microbes electricity to convert carbon dioxide and water to multicarbon extracellular organic compounds. mBio, 1(2), pp.1–4. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2921159&tool=pmcentrez&ren dertype=abstract [Accessed January 5, 2015].

Niehaus, T.D. et al., 2011. Identification of unique mechanisms for triterpene biosynthesis in Botryococcus braunii. Proceedings of the National Academy of Sciences of the United States of America, 108(30), pp.12260–5. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3145686&tool=pmcentrez&ren dertype=abstract.

Ohta, K. et al., 1991. Genetic improvement of Escherichia coli for ethanol production: chromosomal integration of Zymomonas mobilis genes encoding pyruvate decarboxylase and alcohol. Applied and environmental microbiology, 57(4), pp.893–900. Available at: http://aem.asm.org/content/57/4/893.short [Accessed June 3, 2014].

160 Okada, S. et al., 2004. Characterization of botryococcene synthase enzyme activity, a squalene synthase-like activity from the green microalga Botryococcus braunii, Race B. Archives of biochemistry, 422(1), pp.110–118. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0003986103006568 [Accessed July 23, 2011].

Paoli, G.C. & Tabita, F.R., 1998. Aerobic chemolithoautotrophic growth and RubisCO function in Rhodobacter capsulatus and a spontaneous gain of function mutant of Rhodobacter sphaeroides. Archives of microbiology, 170(1), pp.8–17. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9639598 [Accessed November 12, 2010].

Peng, J., Lu, L. & Yang, H., 2013. Review on life cycle assessment of energy payback and greenhouse gas emission of solar photovoltaic systems. Renewable and Sustainable Energy Reviews, 19, pp.255–274. Available at: http://www.sciencedirect.com/science/article/pii/S1364032112006478 [Accessed July 9, 2014].

Phillips, J.R. et al., 1993. Biological production of ethanol from coal synthesis gas. Applied Biochemistry and Biotechnology, 39-40(1), pp.559–571. Available at: http://www.springerlink.com/index/10.1007/BF02919018.

Pirt, S.J., 1965. The maintenance energy of bacteria in growing cultures. Proceedings of the Royal Society of London. Series B, Biological Sciences, 163(991), pp.224–231. Available at: http://www.jstor.org/stable/75569.

Quandt, J. & Hynes, M.F., 1993. Versatile suicide vectors which allow direct selection for gene replacement in Gram-negative bacteria. Gene, 127(1), pp.15–21. Available at: http://linkinghub.elsevier.com/retrieve/pii/0378111993906116.

Rabaey, K. & Rozendal, R.A., 2010. Microbial electrosynthesis - revisiting the electrical route for microbial production. Nature reviews. Microbiology, 8(10), pp.706–16. Available at: http://dx.doi.org/10.1038/nrmicro2422 [Accessed July 10, 2014].

REN21, 2013. Renewables 2013: Global Status Report, Available at: http://www.ren21.net/portals/0/documents/resources/gsr/2013/gsr2013_lowres.pdf [Accessed May 13, 2014].

Rittman, B.E. & McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications, Boston: Mcgraw-Hill.

Rocha, E.P.C., Cornet, E. & Michel, B., 2005. Comparative and evolutionary analysis of the bacterial homologous recombination systems. PLoS genetics, 1(2), p.e15.

Rohwerder, T. & Müller, R.H., 2010. Biosynthesis of 2-hydroxyisobutyric acid (2-HIBA) from renewable carbon. Microbial cell factories, 9, p.13.

Rosa, A.V. da, 2005. Fundamentals of renewable energy processes, Amsterdam: Elsevier Academic Press.

161 Rousset, M. & Liebgott, P., 2014. Engineering Hydrogenases for H2 Production: Bolts and Goals. In D. Zannoni & R. De Philippis, eds. Microbial BioEnergy: Hydrogen Production. Advances in Photosynthesis and Respiration. Dordrecht: Springer Netherlands, pp. 43–77. Available at: http://link.springer.com/10.1007/978-94-017-8554-9 [Accessed October 27, 2014].

Ryu, H. et al., 1997. Production of poly(3-hydroxybutyrate) by high cell density fed-batch culture of Alcaligenes eutrophus with phospate limitation. Biotechnology and …, 5(1), pp.28–32. Available at: http://onlinelibrary.wiley.com/doi/10.1002/(SICI)1097- 0290(19970705)55:1<28::AID-BIT4>3.0.CO;2-Z/abstract [Accessed July 12, 2014].

Sato, S., Fujiki, T. & Matsumoto, K., 2013. Construction of a stable plasmid vector for industrial production of poly(3-hydroxybutyrate-co-3-hydroxyhexanoate) by a recombinant Cupriavidus necator H16 strain. Journal of bioscience and bioengineering, 116(6), pp.677– 81. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23816763 [Accessed May 23, 2014].

Saur, G., 2008. Wind-To-Hydrogen Project: Electrolyzer Capital Cost Study, Available at: http://www.nrel.gov/docs/fy09osti/44103.pdf [Accessed July 6, 2014].

Schaefer, A.L. et al., 2002. Long-Chain Acyl-Homoserine Lactone Quorum-Sensing Regulation of Rhodobacter capsulatus Gene Transfer Agent Production. Journal of Bacteriology, 184(23), pp.6515–6521. Available at: http://jb.asm.org/content/184/23/6515.short [Accessed June 30, 2014].

Schick, M. et al., 2012. Biosynthesis of the iron-guanylylpyridinol cofactor of [Fe]-hydrogenase in methanogenic archaea as elucidated by stable-isotope labeling. Journal of the American Chemical Society, 134(6), pp.3271–80.

Schiel-Bengelsdorf, B. & Dürre, P., 2012. Pathway engineering and synthetic biology using acetogens. FEBS letters, 586(15), pp.2191–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22710156 [Accessed October 19, 2014].

Schmidt, C.O. et al., 1999. Isolation, characterization, and mechanistic studies of (-)-alpha- gurjunene synthase from Solidago canadensis. Archives of biochemistry and biophysics, 364(2), pp.167–77. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10190971 [Accessed June 2, 2014].

Schwartz, E., Fritsch, J. & Friedrich, B., 2013. H2-metabolizing prokaryotes. In E. Rosenberg et al., eds. The Prokaryotes - Prokaryotic Physiology and Biochemistry. Berlin, Heidelberg: Springer Berlin Heidelberg, pp. 119–199. Available at: http://link.springer.com/10.1007/978-3-642-30141-4 [Accessed July 18, 2014].

Siefert, E. & Pfennig, N., 1979. Chemoautotrophic growth of Rhodopseudomonas species with hydrogen and chemotrophic utilization of methanol and formate. Archives of Microbiology, 122(2), pp.177–182. Available at: http://www.springerlink.com/index/V323851512727030.pdf [Accessed June 3, 2011].

162 Siegel, R.S. & Ollis, D.F., 1984. Kinetics of growth of the hydrogen-oxidizing bacterium Alcaligenes eutrophus (ATCC 17707) in chemostat culture. Biotechnology and bioengineering, 26(7), pp.764–70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18553444.

Simon, R., Priefer, U. & Pühler, A., 1983. A broad host range mobilization system for in vivo genetic engineering: transposon mutagenesis in gram negative bacteria. Nature Biotechnology, 1(9), pp.784–791. Available at: http://www.nature.com/nbt/journal/v1/n9/abs/nbt1183-784.html [Accessed June 18, 2013].

Smith, S. a & Tabita, F.R., 2003. Positive and Negative Selection of Mutant Forms of Prokaryotic (Cyanobacterial) Ribulose-1,5-bisphosphate Carboxylase/Oxygenase. Journal of Molecular Biology, 331(3), pp.557–569. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0022283603007861 [Accessed October 4, 2014].

Solaiman, D.K.Y., Swingle, B.M. & Ashby, R.D., 2010. A new shuttle vector for gene expression in biopolymer-producing Ralstonia eutropha. Journal of Microbiological Methods, 82, pp.120–123.

Solomon, B.D., Barnes, J.R. & Halvorsen, K.E., 2007. Grain and Cellulosic Ethanol: History, Economics, and Energy Policy. Biomass and Bioenergy, 31(6), pp.416–425. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0961953407000396 [Accessed March 4, 2013].

Steinbrenner, J. & Sandmann, G., 2006. Transformation of the green alga Haematococcus pluvialis with a phytoene desaturase for accelerated astaxanthin biosynthesis. Applied and environmental microbiology, 72(12), pp.7477–84. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1694260&tool=pmcentrez&ren dertype=abstract [Accessed June 3, 2014].

Straub, M. et al., 2014. Selective enhancement of autotrophic acetate production with genetically modified Acetobacterium woodii. Journal of biotechnology, pp.67–72. Available at: http://www.sciencedirect.com/science/article/pii/S0168165614001084#.

Suzuki, T. & Yasui, K., 2011. Plasmid Artificial Modification: A Novel Method for Efficient DNA Transfer into Bacteria J. A. Williams, ed. Plasmid, 765.

Sydow, A. et al., 2014. Electroactive bacteria--molecular mechanisms and genetic tools. Applied microbiology and biotechnology, 98(20), pp.8481–95. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25139447 [Accessed January 8, 2015].

Tanaka, K. et al., 1995. Production of poly (D-3-hydroxybutyrate) from CO2, H2, and O2 by high cell density autotrophic cultivation of Alcaligenes eutrophus. Biotechnology and bioengineering, 45(3), pp.268–275. Available at: http://onlinelibrary.wiley.com/doi/10.1002/bit.260450312/abstract [Accessed April 14, 2011].

163 Thauer, R.K. et al., 1977. Energy conservation in chemotrophic anaerobic bacteria. Bacteriological reviews, 41(3), p.809. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16350228.

Thauer, R.K. et al., 2008. Methanogenic archaea: ecologically relevant differences in energy conservation. Nature reviews. Microbiology, 6(8), pp.579–91. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18587410 [Accessed December 4, 2014].

Tikh, I.B., Held, M. & Schmidt-Dannert, C., 2014. BioBrick(TM) compatible vector system for protein expression in Rhodobacter sphaeroides. APPLIED MICROBIOLOGY AND BIOTECHNOLOGY, 98(7), pp.3111–3119.

Tsukahara, K. et al., 2014. Genome-guided analysis of transformation efficiency and carbon dioxide assimilation by Moorella thermoacetica Y72. Gene, 535(2), pp.150–5.

Tuerk, A., 2011. An assessment of photosynthetic biofuels and electrofuels technologies under rate-limited conditions. The Pennsylvania State University. Available at: https://etda.libraries.psu.edu/paper/12318/8169.

Tuerk, A.L., 2011. An assessment of photosynthetic biofuels and electrofuels technologies under rate-limited conditions. Pennsylvania State University. Available at: https://etda.libraries.psu.edu/paper/12318/.

Ueki, T. et al., 2014. Converting carbon dioxide to butyrate with an engineered strain of Clostridium ljungdahlii. mBio, 5(5), pp.19–23. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4212834/.

Umeda, F. et al., 1986. Conjugal transfer of hydrogen-oxidizing ability of Alcaligenes hydrogenphilus to Pseudomonas oxalatus. Biochemical and biophysical research communications, 137(1), pp.108–113. Available at: http://www.sciencedirect.com/science/article/pii/0006291X86911824#.

Veit, A. et al., 2008. Constructing and testing the thermodynamic limits of synthetic NAD(P)H:H2 pathways. Microbial biotechnology, 1(5), pp.382–94. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3815245&tool=pmcentrez&ren dertype=abstract [Accessed October 22, 2014].

Verwaal, R. et al., 2007. High-level production of beta-carotene in Saccharomyces cerevisiae by successive transformation with carotenogenic genes from Xanthophyllomyces dendrorhous. Applied and environmental microbiology, 73(13), pp.4342–50. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1932764&tool=pmcentrez&ren dertype=abstract [Accessed June 3, 2014].

Vignais, P.M. & Billoud, B., 2007. Occurrence, classification, and biological function of hydrogenases: an overview. Chemical reviews, 107(10), pp.4206–72. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17927159 [Accessed October 24, 2014].

164 Volova, T.G., Kalacheva, G.S. & Altukhova, O. V., 2002. Autotrophic synthesis of polyhydroxyalkanoates by the bacteria Ralstonia eutropha in the presence of carbon monoxide. Applied Microbiology and Biotechnology, 58, pp.675–678. Available at: http://link.springer.com/article/10.1007%2Fs00253-002-0941-8.

Voss, I. & Steinbüchel, A., 2006. Application of a KDPG-aldolase gene-dependent addiction system for enhanced production of cyanophycin in Ralstonia eutropha strain H16. Metabolic engineering, 8(1), pp.66–78. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16266816 [Accessed July 14, 2014].

Van Walraven, H.S. et al., 1996. The H+/ATP coupling ratio of the ATP synthase from thiol- modulated chloroplasts and two cyanobacterial strains is four. FEBS letters, 379(3), pp.309– 13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8603713 [Accessed October 20, 2014].

Wang, F. & Lee, S.Y., 1997. Poly(3-Hydroxybutyrate) Production with High Productivity and High Polymer Content by a Fed-Batch Culture of Alcaligenes latus under Nitrogen Limitation. Applied and environmental microbiology, 63(9), pp.3703–6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1389255&tool=pmcentrez&ren dertype=abstract.

Whitney, S.M., Houtz, R.L. & Alonso, H., 2011. Advancing our understanding and capacity to engineer nature’s CO2-sequestering enzyme, Rubisco. Plant physiology, 155(1), pp.27–35. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3075749&tool=pmcentrez&ren dertype=abstract [Accessed August 6, 2014].

Wu, S. et al., 2012. Engineering triterpene metabolism in tobacco. Planta, 236(3), pp.867–77. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3810399&tool=pmcentrez&ren dertype=abstract [Accessed June 1, 2014].

Wu, S. et al., 2006. Redirection of cytosolic or plastidic isoprenoid precursors elevates terpene production in plants. Nature biotechnology, 24(11), pp.1441–7. Available at: http://dx.doi.org/10.1038/nbt1251.

Yagi, K. et al., 1986. Isolation of hydrogen-oxidation gene from Alcaligenes hydrogenophilus and its expression in Pseudomonas oxalaticus. Biiochemical and biophysical research communications, (1), pp.114–119. Available at: http://www.sciencedirect.com/science/article/pii/0006291X86911836#.

Yang, J. et al., 2012. Enhancing production of bio-isoprene using hybrid MVA pathway and isoprene synthase in E. coli. PloS one, 7(4), p.e33509. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3338741&tool=pmcentrez&ren dertype=abstract [Accessed May 30, 2014].

Yasui, K. et al., 2009. Improvement of bacterial transformation efficiency using plasmid artificial modification. Nucleic acids research, 37(1), p.e3.

Nymul Khan

Pennsylvania State University, 226 Fenske Lab • University Park, PA, 16802 • (336) 790 8954 [email protected]

Education Bangladesh University of Engineering and Technology Dhaka, Bangladesh Bachelor of Science in Chemical engineering 2001 - 2006 North Carolina A&T State University Greensboro, NC, USA Masters of Science in Chemical engineering 2007-2010 The Pennsylvania State University University Park, PA, USA Ph.D. in Chemical Engineering 2010-2015

Research Publications / Presentations Journal  Nybo, S. E., Khan, N. E., Woolston, B. M., & Curtis, W. R. (2015). Metabolic Engineering in Chemolithoautotrophic Hosts for the Production of Fuels and Chemicals. Metabolic Engineering. (Invited review; co-first author)  Khan, N., Nybo, S. E., Chappell, J., & Curtis, W. R. (2015). Triterpene Hydrocarbon Production Engineered Into a Metabolically Versatile Host–Rhodobacter capsulatus. Biotechnology and bioengineering. DOI: 10.1002/bit.25573.  Khan, N. E., Myers, J. A., Tuerk, A. L., & Curtis, W. R. (2014). A process economic assessment of hydrocarbon biofuels production using chemoautotrophic organisms. Bioresource technology, 172, 201-211.  Adewuyi, Y. G., & Khan, N. E. (2012). Modeling the ultrasonic cavitation‐enhanced removal of nitrogen oxide in a bubble column reactor. AIChE Journal,58(8), 2397-2411.  Khan, N. E., & Adewuyi, Y. G. (2011). A new method of analysis of peroxydisulfate using ion chromatography and its application to the simultaneous determination of peroxydisulfate and other common inorganic ions in a peroxydisulfate matrix. Journal of Chromatography A, 1218(3), 392-397.  Khan, N. E., & Adewuyi, Y. G. (2010). Absorption and oxidation of nitric oxide (NO) by aqueous solutions of sodium persulfate in a bubble column reactor.Industrial & Engineering Chemistry Research, 49(18), 8749- 8760.  Rahman, M. A., & Khan, N. E. (2010). Study of an Evaporation System for Sodium Hydroxide Solution. Journal of Chemical Engineering, 24, 35-36. Conference  Nymul Khan, Alex Rajangam, Eric Nybo, Joe Chappell, Wayne Curtis. Rhodobacter as a platform for autotrophic biofuel production. AIChE annual meeting, Pittsburgh, PA, 2012.  Nymul Khan, Yousuf Adewuyi. Modeling of the sonochemical removal of nitric oxide using Raleigh Plesset bubble dynamics equation, AIChE annual meeting, Nashville, TN, 2009.  Nymul Khan, Yousuf Adewuyi. Removal of nitric oxide using aqueous solution of sodium persulfate in a bubble column reactor. AIChE Annual Meeting, Nashville, TN, 2009.