Open Dissertation Rajakovichlj.Pdf

The Pennsylvania State University

The Graduate School

Eberly College of Science

EXPLORING THE FUNCTIONAL AND MECHANISTIC DIVERSITY OF

DIIRON OXIDASES AND OXYGENASES

A Dissertation in

Biochemistry, Microbiology, and Molecular Biology

Lauren J. Rajakovich

 2017 Lauren J. Rajakovich

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2017

The dissertation of Lauren J. Rajakovich was reviewed and approved* by the following:

J. Martin Bollinger, Jr. Professor of Chemistry Professor of Biochemistry and Molecular Biology Dissertation Co-Advisor Committee Co-Chair

Carsten Krebs Professor of Chemistry Professor of Biochemistry and Molecular Biology Dissertation Co-Advisor Committee Co-Chair

Squire J. Booker Howard Hughes Medical Investigator Professor of Chemistry Professor of Biochemistry and Molecular Biology

Amie K. Boal Professor of Chemistry Professor of Biochemistry and Molecular Biology

Christopher House Professor of Geosciences

Scott Selleck Professor of Biochemistry and Molecular Biology Head of the Department of Biochemistry, Microbiology, and Molecular Biology

*Signatures are on file in the Graduate School

iii

ABSTRACT

Approximately half of all enzymes in Nature utilize a metal to perform their biological function.

Many of the metalloenzymes that harbor transition metals activate dioxygen to catalyze a diverse array of oxidation reactions to functionalize unreactive sites in biomolecules. These enzymatic transformations are often inaccessible to synthetic chemists, and consequently, understanding the naturally-evolved mechanisms by which these enzymes enact such challenging reactions will enable the development of new biocatalysts for industrial and therapeutic applications. One common strategy employed in Nature is the coupling of two transition metals, typically iron, to carry out oxidation and oxygenation reactions. Decades of research on this class of non-heme diiron enzymes has focused on three founding members, which has revealed unifying mechanistic features that enable them to enact one- and two-electron oxidation reactions. Chapter 1 summarizes the principles for dioxygen activation employing this bioinorganic scaffold that emerged from this foundational work. Chapter 1 also introduces more recent discoveries of novel non-heme diiron enzymes, facilitated by advancements in genome sequencing and bioinformatics, that have expanded the scope of chemical transformations and mechanistic strategies possible within this extensive metalloenzyme class. My dissertation research focused on three of these newly discovered diiron enzymes that invoke alternative mechanistic strategies to carry out non- canonical transformations. Chapter 2 covers the hydrocarbon-producing cyanobacterial diiron enzyme, ADO, which catalyzes a C-C bond cleavage reaction to effectively remove the chemical functional group from its fatty aldehyde substrate, producing linear hydrocarbons. My work demonstrates that ADO operates by a free-radical mechanism to enact this redox-neutral transformation, and that a cyanobacterial ferredoxin (PetF)/ferredoxin reductase/NADPH reducing system can act as an efficient reducing partner, a requisite for ADO catalysis. These mechanistic studies enabled the identification of inherent vulnerabilities that limit enzymatic

iv efficacy, thereby highlighting direct targets for bioengineering and optimization of biofuel processes deploying this catalyst. Chapter 3 describes a project designed to biochemically characterize a diiron enzyme belonging to a new functional class. This work culminated in the discovery of a novel microbial phosphonate degradation pathway, consisting of two iron- dependent oxygenases with previously misannotated functional assignments. Finally, Chapter 4 describes progress on studies of another potential biofuel catalyst, an iron-dependent enzyme,

UndA. This work provides evidence that UndA employs a diiron cofactor to convert fatty acids into terminal alkenes. This finding corrects its original cofactor assignment, rationalizes its ability to perform this transformation, and suggests a feasible catalytic mechanism. Collectively, these mechanistic studies contribute unique insight into the emerging reactivities of diiron enzymes that diversify their potential biotechnological applications.

TABLE OF CONTENTS

List of Figures ...... vii

List of Tables ...... xviii

Acknowledgements ...... xix

Chapter 1 Non-heme diiron oxygenases and oxidases ...... 1

1.1 Introduction ...... 2 1.2 Canonical ferritin-like diiron-carboxylate oxidases and oxygenases (FDCOOs) ...... 4 1.3 Novel outcomes and variations on the FDCOO mechanistic theme ...... 12 1.3.1 Electrophilic diferric-peroxide intermediates as oxidants...... 13 1.3.2 A nucleophilic dioxygen moiety in aldehyde deformylating oxygenase...... 24 1.3.3 Use of transition metals other than iron by ferritin-like proteins...... 29 1.4 O2-activating diiron cofactors within non-ferrtin-like protein architectures ...... 39 1.4.1 Integral membrane diiron oxidases/oxygenases...... 40 1.4.2 Deoxyhypusine hydroxylase...... 44 1.4.3 Diiron β-hydroxylases...... 47 1.4.4 HD-domain mixed-valent diiron oxygenases...... 50 1.5 Outlook...... 56 1.6 References ...... 57

Chapter 2 A cyanobacterial hydrocarbon production pathway employing a non-heme diiron oxygenase ...... 75

2.1 Exploring physiological reducing partners for ADO catalysis ...... 82 2.1.1 Selection of the cyanobacterial [2Fe-2S] ferredoxin, PetF, as a physiologically-relevant reducing partner ...... 84 2.1.2 Employment of PetF as a reducing system to support ADO catalysis under single turnover conditions ...... 87 III/III 2.1.3 Reduction of the Fe2 ADO by PetF, monitored by stopped-flow absorption spectroscopy (SF-Abs) ...... 88 III/III 2.1.4 Rapid reduction of the ADO Fe2 -PHA intermediate by PetF, monitored by SF-Abs ...... 90 2.1.5 Rapid reduction of the ADO Fe2(III/III)-PHA intermediate by PetF, monitored by rapid freeze-quench (RFQ) Mössbauer spectroscopy ...... 94 2.1.6 Sequential delivery of two electrons from PetF to ADO in a tightly coupled fashion ...... 98 2.2 Interrogating the postulated free-radical mechanism of ADO catalysis ...... 101 2.2.1 Accumulation of radicals upon reduction of the ADO Fe2(III/III)-PHA intermediate ...... 102 2.2.2 Association of the transient EPR signal with a substrate-derived alkylperoxyl radical ...... 106 2.2.3 Dependence of peroxyl radical accumulation on the concentration of O2 ...... 110 2.2.4 Diminished alkane yields due to unproductive alkyl radical quenching by O2 ...... 111

2.2.5 Detection of a protein-based sulfinyl radical using short-chain aldehyde substrates ...... 114 2.2.6 Exploring potential hydrogen atom donors for quenching of the alkyl radical intermediate ...... 122 2.3 Materials and Methods ...... 125 2.4 Acknowledgements ...... 136 2.5 References ...... 136

Chapter 3 A new microbial pathway for organophosphonate degradation catalyzed by two previously misannotated non-heme-iron enzymes ...... 142

3.1 Introduction ...... 143 3.2 Discovery of the TMAEP hydroxylation activity of TmpA...... 147 3.3 TmpA demonstrates substrate specificity for TMAEP...... 151 3.4 Structural basis for TmpA substrate specificity...... 157 II/III 3.5 TmpB harbors a diiron cofactor that is stable in the mixed-valent Fe2 state...... 160 3.6 The TmpA hydroxylation product serves as the TmpB substrate...... 165 3.7 Structural characterization of TmpB ...... 166 3.7 Discussion ...... 169 3.8 Materials and Methods ...... 178 3.9 Acknowledgements ...... 187 3.10 References ...... 187

Chapter 4 Fatty acid decarboxylation to terminal alkene by the diiron oxidase, UndA ...... 195

4.1 Experimental testing of the diiron cofactor hypothesis ...... 199 4.1.1 Structural homology to known diiron proteins ...... 200 4.1.2 UndA metal incorporation by various preparations ...... 202 4.1.3 Interrogating nuclearity and oxidation states of UndA cofactor by Mössbauer and EPR spectroscopies ...... 203 4.1.4 Interrogating nuclearity by X-ray absorption (XAS) spectroscopy...... 205 4.1.5 Interrogating nuclearity by x-ray crystallography ...... 207 4.2 Establishing the active cofactor and operant mechanism for alkene production ...... 209 4.2.1 Substrate triggering effect leads to rapid reaction of reduced UndA with O2...... 209 4.2.2 Rapid accumulation of differic species upon reaction of reduced UndA with O2...... 212 4.2.3 Accumulation of a transient tyrosyl radical...... 217 4.3 Exploring potential mechanisms for UndA-catalyzed fatty acid oxidative decarboxylation ...... 219 4.4 Materials & Methods ...... 224 4.5 Acknowledgements ...... 229 4.6 References ...... 230 Appendix A Chapter 2 Supporting Information ...... 234 Appendix B Chapter 3 Supporting Information ...... 261

vii

LIST OF FIGURES

Figure 1-1. Ferritin-like diiron-carboxylate oxygenase/oxidase structural architecture. Helical architecture of the superfamily namesake, ferritin (PDB accession code 4IWK), is shown on the left in green with the positions of the metal coordinating residues colored in yellow. Diiron active sites of the reduced (diferrous) forms of the three archetypal members are shown on the right – Escherichia coli (Ec) I-a RNR-β (purple; PDB accession code 1PIY), Δ9D (blue; PDB accession code 1AFR), and sMMOH (yellow; PDB accession code 1FYZ)...... 5

Figure 1-3. Possible (hydro)peroxide binding geometries to dinuclear metal clusters...... 13

Figure 1-4. Electrophilic aromatic substitution and single electron transfer mechanisms for arene oxidation invoking a diferric-hydroperoxide intermediate...... 16

Figure 1-5. Aureothin and chloramphenicol abbreviated biosynthetic pathways, highlighting conversions by the diiron N-oxygenases (blue) and β-hydroxylase (red).... 17

Figure 1-6. Proposed nucleophilic attack and single electron transfer mechanisms for reaction of arylamine substrates with electrophilic diferric-(hydro)peroxide intermediates of the N-oxygenases, AurF and CmlI...... 21

Figure 1-7. Divergent pathways suggested for the overall six-electron oxidation sequences of the N-oxygenases, AurF (107) and CmlI (106)...... 24

Figure 1-8. Free-radical mechanism proposed for ADO catalysis, invoking reduction of the ADO diferric-PHA intermediate to yield the alkane product...... 29

Figure 1-9. Active cofactors and mechanistic schemes for the O2 activation pathways of class I RNRs...... 31

Figure 1-10. Potential mechanisms for formation of the Tyr-Val crosslink observed in the R2lox crystal structure employing a heterdinuclear Mn/Fe cofactor...... 36

Figure 1-11. Structural architecture of integral membrane diiron enzymes, stearoyl-CoA desaturase (orange, PDB accession code 4YMK) and fatty acid α-hydroxylase (yellow, PDB, accession code 4ZR0) and depictions of their histidine-rich active sites. The bound acyl-CoA substrate of stearoyl-CoA desaturase is shown as cyan sticks, and the zinc ions are shown as gray spheres. The lines estimate the membrane boundaries and the protein transmembrane domains...... 42

viii

modification of the eIF5A protein lysine residue. (Right, bottom) Zoom in view of the diiron active site displaying the bound peroxide moiety (red)...... 46

Figure 1-13. Metallo-β-lactamase structural fold of CmlA (PDB accession code 4JO0), composed of an N-terminal (purple) domain and a C-terminal catalytic (gray) domain harboring the diiron active site...... 50

Figure 1-15. Reaction mechanism for dioxygen activation by the MIOX mixed-valent II/III Fe2 cofactor and viable mechanistic possibilities for oxygen incorporation into the reaction product. The hydroxyl groups on the substrate at carbon positions 2-5 have been omitted for clarity...... 54

Figure 2-1. Cyanobacterial metabolic pathway for production of long-chain hydrocarbons. n = 11, 13, and 15...... 76

Figure 2-2. Scheme depicting the free-radical mechanism proposed for ADO catalysis...... 81

Figure 2-3. In vitro coupled Np AAR and Np ADO activity assays (t = 1 h at 37 °C) carried out with all seven putative Syn. 6803 Fds (encoded by the genes in the figure) together with Syn. 6803 FNR and NADPH, as well as the spinach Fd/FNR/N system (blue). Heptadecane and octadecanal products were detected by GC-MS. Assay conditions: 10 µM AAR, 20 µM ADO, 0.2 mM 1-[13C]-stearoyl-ACP, 4 mM NADPH, 7.8 μM Fd and 7.8 µM FNR...... 85

Figure 2-4. Characterization of the PetF metallocofactor by UV-visible absorption spectroscopy. (A) Reference absorption spectra of [2Fe-2S]1+ (black trace) and [2Fe- 2S]2+ (red trace) forms of PetF were obtained from a single-mixing SF-Abs experiment, in which sodium dithionite-reduced PetF (0.10 mM) was mixed with an equal volume of either O2-free or O2-saturated 50 mM sodium HEPES, pH 7.5 buffer (~1.8 mM O2), respectively, for 250 s at 5 °C. (B) SF-Abs kinetic trace monitoring oxidation of the [2Fe-2S]1+ PetF at 423 nm (blue trace) in the single- mixing experiment described in A. The black dashed line is the fit described in the text...... 86

Figure 2-5. X-Band CW EPR spectrum of the reduced [2Fe-2S]1+ cluster generated by reduction of the as-isolated PetF (0.28 mM) with sodium dithionite (10 mM) for 30 min in the absence of O2. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.3 mT...... 87

Figure 2-6. Formate production in single turnover ADO assays utilizing different III/III reducing systems. The accumulated Fe2 -PHA intermediate (Figure 2-24) was reacted for ~ 1 s with varying molar ratios of reductant: either chemically reduced PetF (blue) or NADPH (black) in the presence of oxidized PetF and oxidized FNR

in equimolar ratios with ADO (0.050 mM final). Data are averages of at least three replicate assays...... 88

III/III Figure 2-7. Kinetics of PetF oxidation by the as-isolated Fe2 -ADO. (A) SF-Abs kinetic traces monitoring PetF oxidation at 423 nm in single-mix experiments in which a solution of O2-free, chemically reduced PetF ([final] = 0.050 mM) was III/III mixed with an equal volume of an O2-free solution of as-isolated Fe2 -ADO at varying concentrations to give the final ratios indicated in the figure legend. The data were best fit by two exponential functions with the second observed rate constant (kobs2) being concentration dependent. (B) Plot of kobs2 as a function of ADO concentration, from which a second-order rate constant of 8.1 mM-1∙s-1 was obtained. .. 90

Figure 2-8. Sequential-mixing SF-Abs experiments monitoring PetF oxidation upon III/III reaction with the Fe2 -PHA intermediate. (A) Time-dependent absorption spectra III/III after mixing the accumulated Fe2 -PHA intermediate with two equivalents of III/III chemically reduced PetF (0.050 mM final). The Fe2 -PHA intermediate was II/II generated by mixing the Fe2 -ADO•octanal complex with O2-saturated buffer and incubating for 15 s. The “0” ms reference spectrum (black) was constructed mathematically by addition of the individual spectra of the chemically reduced PetF III/III (grey dotted line) and the Fe2 -PHA intermediate (light blue dotted line). The III/III spectrum of the Fe2 -PHA intermediate was generated by mixing a solution II/II containing 0.050 mM Fe2 -ADO and 10 mM octanal with an equal volume of O2- saturated 50 mM sodium HEPES, pH 7.5. The spectrum of the PetF [2Fe-2S]1+ was obtained by mixing chemically reduced PetF (0.10 mM) with an equal volume of O2-free 50 mM sodium HEPES, pH 7.5. (B) Kinetic traces monitoring PetF [2Fe- 2S]1+ oxidation by the increase in absorption at 423 nm in experiments in which the ADO concentration was varied at a constant concentration of chemically reduced PetF (0.050 mM final). Dashed lines are the fits described in the text...... 91

Figure 2-9. Comparison of the kinetics of oxidation of two equivalents of PetF (λ = 423 nm) or one equivalent of dithionite-reduced MeOPMS (λ = 388 nm)(17) upon reaction III/III with the Fe2 -PHA intermediate. The intermediate was accumulated by mixing a II/II MeO solution of 0.1 or 0.3 mM Fe2 -ADO (PetF or PMS experiment, respectively) and 10 mM decanal with an equal volume of O2-saturated buffer and incubating for 15 s before the second mixing event with the respective reductant...... 93

III/III Figure 2-10. 4.2-K/53-mT FQ-Mössbauer spectra monitoring the reaction of the Fe2 - PHA intermediate with two equivalents of reduced PetF. (A) Spectrum of the III/III 57 II/II Fe2 -PHA intermediate accumulated after a solution of the Fe2 -ADO•decanal complex and Cld was reacted with NaClO2 for 30 s. The previously published III/III experimental reference spectrum of the Fe2 -PHA intermediate(17) is overlaid in red with the constituent quadrupole doublets designated with black and grey II/II brackets. The experimental reference spectrum of substrate-free Fe2 -ADO is III/III overlaid in blue. (B) Spectrum of a sample quenched after reaction of the Fe2 - PHA intermediate with two equivalents of chemically reduced, unlabeled (>95% 56Fe) PetF for 0.010 s. The dashed line denotes the high-energy line of the doublet III/III with ΔEQ = 1.2 mm/s (black bracket) arising from the Fe2 -PHA intermediate. (C) B−A difference spectrum (black bars), overlaid with 55% of the experimental III/III reference spectrum of the Fe2 -PHA intermediate (red line).(17) (D) Experimental

III/III III/III reference spectrum of the Fe2 clusters formed upon reaction of the Fe2 -PHA intermediate with two equivalents of reduced PetF for 0.010 s. The spectrum was III/III generated by addition of 55 % of the experimental reference spectrum of the Fe2 - PHA intermediate to the difference spectrum C. (E) Experimental reference III/III III/III spectrum of the Fe2 clusters generated upon reaction of the Fe2 -PHA intermediate with two equivalents of reduced PetF for 0.22 s (black bars; see Figure III/III 2-11 for generation of the spectrum). The derived spectrum of the Fe2 species formed after the 0.010 s reaction with PetF is overlaid for comparison (green line) to III/III depict the changes in relative intensity of the doublets of the two Fe2 forms (δ  0.5 mm/s with ΔEQ  0.8 mm/s and 1.5 mm/s, purple and orange brackets, respectively)...... 96

III/III Figure 2-11. 4.2-K/53-mT FQ-Mössbauer spectra monitoring reduction of the Fe2 - PHA intermediate with 0.5 equivalents of reduced PetF with respect to ADO. (A) Spectrum of a sample that was freeze-quenched 30 s after a solution of 2 mM 57 II/II Fe2 -ADO, 16.6 mM decanal and 10 μM Cld was mixed with 10 mM NaClO2. III/III The Fe2 -PHA intermediate accumulates to 65% of the total Fe and the remaining II/II 35% is of the unreacted Fe2 -ADO. (B) Overlay of the spectrum in A (blue line) III/III with the spectrum of a sample freeze-quenched after the Fe2 -PHA intermediate, accumulated as in A, was reacted with 0.5 molar equivalent (with respect to ADO) chemically reduced, unlabeled PetF for 0.22 s (black bars). (C) Overlay of the B–A difference spectrum (black bars) with the negative of the experimental reference III/III spectrum of the Fe2 -PHA intermediate(17) (red line) plotted at an intensity corresponding to 25 % of the total area of the experimental spectrum in B. (D) Spectrum after addition of 25 % of the experimental reference spectrum of the III/III Fe2 -PHA intermediate to the difference spectrum C. The dashed line denotes the position of the high-energy line of the doublet with ΔEQ = 0.49 mm/s from the III/III spectrum of the Fe2 -PHA intermediate...... 100

III/III Figure 2-12. CW EPR spectra of FQ samples obtained after reduction of the Fe2 -PHA intermediate by (A) one equivalent of reduced MeOPMS or (B) two equivalents of III/III chemically reduced PetF with respect to ADO. The Fe2 -PHA intermediate was II/II accumulated in an initial reaction of the Fe2 -ADO•decanal complex with the Cld/NaClO2 system (Figure 2-24) and was then mixed with the dithionite-reduced MeOPMS or PetF. The dashed lines are positioned at the low field feature (g = 2.037) of the transient radical signal. The arrows indicate the signal features of the [2Fe- 2S]1+ PetF cluster. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.481 GHz, modulation amplitude = 0.4 mT...... 104

Figure 2-13. X-Band CW EPR spectra of samples quenched at 0.020 s in sequential- II/II mixing FQ experiments. In the first mix, 0.60 mM Fe2 -ADO with (red, green) or without (blue) 16.6 mM decanal substrate was reacted for 30 s with O2-saturated (~1.8 mM O2 at 5 °C) 50 mM sodium HEPES, pH 7.5, buffer to allow for formation III/III of the Fe2 -PHA intermediate. In the second mix, the intermediate was reacted with either equimolar reduced MeOPMS (red, blue) or buffer (green) for 0.020 s prior to freezing in liquid 2-methylbutane. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.380 GHz, modulation amplitude = 0.3 mT...... 105

Figure 2-14. Radical yield as a function of equivalents of PetF with respect to ADO. CW III/III EPR spectra of samples quenched 0.010 s after mixing of the Fe2 -PHA intermediate ([ADO]final = 0.23 mM) with varying concentrations of chemically reduced PetF (0.13, 0.23, 0.45, 0.90 mM final concentration). Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT. (inset) Yield of the radical species as a function of equivalents of PetF:ADO...... 105

Figure 2-15. X-band CW EPR spectra of the substrate-derived peroxyl radical species III/III (chemical structure shown in inset). (A) The Fe2 -PHA intermediate was formed 1 2 in the presence of either the 2,2-[ H]2-decanal (black) or 2,2-[ H]2-decanal (red) substrate and was reacted with equimolar reduced MeOPMS for 0.010 s. The spectrum 1 of the 2,2-[ H]2-decanal sample was multiplied by a factor of 1.7 for better comparison to account for the difference in [ADO] (1.49 v. 0.90 mM). Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT. The spectra generated using (B) 2,2- 1 2 [ H]2-decanal and (C) 2,2-[ H]2-decanal were simulated (blue lines) by using the same set of parameters and hyperfine couplings obtained from the simulation of the HYSCORE spectra. The lineshape of the spectra was reproduced by considering an isotropic Voigtian lineshape with peak-to-peak linewidths of 0.17 and 0.24 mT, for the Lorentzian and Gaussian components, respectively, and an anisotropic broadening of [30 22 28] MHz (H-Strain) or [30 18 28] MHz for the spectra 1 2 generated using the 2,2-[ H]2-decanal and the 2,2-[ H]2-decanal, respectively. The only adjustable parameters in the refinement were the g-values. They varied by ± 0.002, which is well within the experimental error of the measurements...... 108

Figure 2-16. X-Band HYSCORE spectra recorded on FQ samples quenched after reaction III/III of the maximally accumulated Fe2 -PHA intermediate with equimolar reduced MeO 2 PMS for 0.02 s in double-mixing experiments using the 2,2-[ H]2-decanal (left 2 column) or the 2,2-[ H]2-3-thiadecanal analog (right column). The numerically simulated spectra are overlaid with grey contours, and the simulation parameters are given in Appendix A (Table S1). Experimental conditions: microwave frequency = 9.388 GHz, τ = 200 ns, π/2 = 8 ns, T = 30 K...... 109

Figure 2-17. Peroxyl radical yield with varying concentrations of O2. CW EPR spectra of III/III FQ samples generated by reaction of the Fe2 -PHA intermediate with one III/III equivalent of reduced PetF for 0.010 s. The Fe2 -PHA intermediate was generated II/II by mixing either the Fe2 -ADO•octanal complex (1.49 mM) containing 10 μM Cld II/II with an equal volume of NaClO2 (black line, 10 mM O2) or the Fe2 -ADO•octanal complex (1.49 mM) with O2-saturated buffer (blue line, ~1.8 mM O2) and incubating for 50 s (Figure 2-24). For clarity, the signal of unreacted [2Fe-2S]1+ PetF (15 %) was subtracted from these spectra. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT...... 111

Figure 2-18. Formate (squares) and alkane (circles) product formation in stoichiometric III/III ADO reactions with varying concentrations of O2. The Fe2 -PHA intermediate II/II was generated by mixing the Fe2 -ADO•decanal complex (0.250 mM ADO) – in the presence of oxidized FNR and PetF (0.250 mM each) – with (A) O2-saturated

xii

buffer (1.8 mM O2 at 5 °C) and incubating for 10 s or (B) air-saturated buffer (~ 0.28 mM O2 at 25 °C) and incubating for 20 s (Figure 2-24). The intermediate was then mixed with varying concentrations of NADPH (a two-electron donor) and quenched after ~ 1 s...... 113

Figure 2-19. X-band CW EPR spectra of RFQ time-course samples using octanal as the II/II ADO substrate. The WT Fe2 -ADO∙octanal complex was reacted in a first mix III/III with the Cld/NaClO2 system to generate the Fe2 -PHA intermediate, which was reacted in a second mix with two equiv of chemically reduced PetF. Experimental conditions: temperature = 60 K, microwave power = 2 mW, microwave frequency = 9.481 GHz, modulation amplitude =0.4 mT...... 115

III/III Figure 2-21. Formation of a Cys71-SO• upon reduction of the Fe2 -PHA intermediate. First derivative of the 2-pulse echo-detected, field-swept Q-band EPR spectrum of the WT ADO 4 s FQ sample. The experimental spectrum (black) was simulated (red) with parameters described in the text. The small discrepancy between the simulation and the experimental spectrum is due to the contribution of the MeOPMS semiquinone radical signal at g ~ 2 and presumably relaxation effects not present in the CW EPR spectrum. Experimental conditions: temperature = 20 K, microwave frequency = 34.08 GHz, shot repetition time (SRT) = 0.5 ms, π/2 = 12 ns, τ = 328 ns...... 118

Figure 2-22. Geometry optimized structures of the putative sulfinyl radical (SO•) harbored on Cys71 of ADO; the adjacent amino acids (Ala70 and Gly72) have also been included in the model. The orientation of the S-O moiety found in typical sulfinyl radicals (A) and the S-O orientation in the sulfinyl radical on Cys71 that supports significant spin density on both of the cysteine β protons (B), as experimentally observed. The plot of the total spin density is shown as mesh contours with a cut-off of V/ r3...... 120

Figure 2-23. X-band CW EPR spectra of the Cys71-SO• in WT Np ADO and NpADO variants. (A) Spectra of FQ-EPR samples quenched after a 0.010 s reaction of the III/III MeO II/II Fe2 -PHA intermediate with equimolar reduced PMS. The Fe2 - ADO•octanal complex ([ADO]f = 0.225 mM WT, C71A, C107A and Y123F, 0.150 mM Y18F and 0.125 mM Y22F) was mixed in equal volume with the Cld/NaClO2 III/III system for 30 s to generate the Fe2 -PHA intermediate. The Y18F and Y22F spectra were multiplied by factors of 1.8 and 1.5, respectively, for ease of comparison...... 121

xiii

formation with air-saturated buffer and the enhanced rate of photolytic decay under illumination by the intense polychromatic light. (C) Effect of substrate chain-length: II/II A solution containing 0.9 mM Fe2 -ADO, 10 μM Cld and 10 mM decanal (red) or octanal (purple) was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the final solution. Data were acquired using monochromatic light and a PMT detector. (D) Effect of high oxygen concentrations: Black trace: A solution of II/II 1.49 mM Fe2 -ADO with 16.6 mM octanal was mixed with an equal volume of O2- II/II saturated buffer (1.8 mM O2). Blue trace: A solution of 1.49 mM Fe2 -ADO with 20 mM octanal and 10 μM Cld was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the reaction. Data were acquired with monochromatic light and a PMT detector. In the experiment with O2-saturated buffer, ~ 80 % -1 -1 intermediate was generated (A450 = 0.7, ε450 = 1,200 mM ∙cm ), in good agreement with the Mössbauer analysis of a parallel sample...... 133

Figure 3-1. Assessment of the γbb hydroxylation activities of TmpA and Ps BBOX. LC- MS chromatograms monitoring γbb (146 m/z, black) and L-carnitine (162 m/z, blue) after a 4 h incubation of 0.01 mM TmpA or PsBBOX, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM sodium L-ascorbate, 1 mM 2OG and 1 mM γbb. The expected conversion is depicted at the top...... 147

Figure 3-2. Sequence similarity network (SSN) illustrating sequence divergence of BBOX- and TmpA-like proteins. The clusters of bacterial protein sequences shown here are derived from the IPR003819 SSN (see Supporting Information, Figures S1- 3). The nodes represent protein sequences with > 90% identity. Edges between the nodes represent pairwise alignment scores of < 10-83 (corresponding to ~ 37% sequence identity). The gene operons in the yellow and green boxes correspond to the genomic context of proteins represented by nodes colored yellow and green, respectively. The large gold circle and dark green diamond represent Lc TmpA and Ps BBOX, respectively. Orange nodes represent predicted fusion proteins with both TmpA- and TmpB-like domains. Diamond shaped nodes represent sequences with an N-terminal Zn(II)-binding motif. Other designations are: LysR, LysR-type transcription regulator; AraC, AraC-type transcription regulator; CDH, carnitine dehydrogenase; HCT, 4-hydroxybenzoyl-CoA thioesterase; HCD, 3- hydroxybutyryl-CoA dehydrogenase; ACT, Acetyl-CoA acetyltransferase/thiolase; ABC, ABC-type glycine/betaine transporter...... 149

Figure 3-3. Testing the activities of Fe/2OG oxygenases toward the TMAEP compound. The expected chemical transformation is shown at the top. (Left) 31P-NMR spectra of the same reactions as in panel A detecting TMAEP (16.8 ppm) and OH-TMAEP (14.0 ppm). (Right) LC-MS chromatograms monitoring TMAEP (168 m/z, black) and OH-TMAEP (184 m/z, red) after a 4-h incubation of 0.01 mM TmpA, PhnY or PsBBOX with 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L-ascorbate, 3 mM 2OG and 2 mM TMAEP...... 151

Figure 3-4. Ultraviolet-visible absorption data showing binding of TMAEP to II TmpA•Fe •2OG and triggering of O2 activation. (A) Difference absorption spectra associated with binding of 2OG (5 mM) to the TmpA•FeII complex [0.6 mM TmpA, 0.5 mM (NH4)2Fe(SO4)2] in the absence of substrate (black) and presence of 5 mM TMAEP (red). (B) Stopped-flow absorption spectra acquired after mixing at 5 °C of

xiv

a solution of 1.2 mM TmpA, 1 mM (NH4)2Fe(SO4)2, 10 mM 2OG and 10 mM TMAEP with an equal volume of air-saturated 50 mM sodium HEPES buffer, pH 7.5 (~ 0.4 mM O2 at 5 °C). The inset shows the absorption spectra at indicated time points after subtracting the spectrum of the anoxic TmpA•FeII•2OG•TMAEP complex. (C) Kinetic traces at 318 nm (blue dots) and 519 nm (red dots), which are attributed to primarily the decay of the ferryl intermediate and the decay and reformation of the TmpA•Fe(II)•2OG•TMAEP reactant complex, respectively. The solid black lines are regression fits to the data, as described in the Methods Section. .... 153

Figure 3-5. Evaluating activity of TmpA toward TMAEP analogues with varying degrees of N-methylation. (A) Absorption difference spectra caused by binding of 2OG (5 II mM) to the TmpA•Fe complex [0.6 mM TmpA, 0.5 mM (NH4)2Fe(SO4)2] in the absence of a substrate (black) and in the presence of 5 mM TMAEP (red), DMAEP (blue) or 2-AEP (green). (B) Kinetic traces at 519 nm after mixing of the solutions described in panel A with air-saturated 50 mM sodium HEPES buffer, pH 7.5 (~ 0.2 mM O2 final at 5 °C). The solid black lines are non-linear regression fits to the data described in the Experimental Section. (C) 31P-NMR spectra of reaction samples after a 4 h incubation of 10 μM TmpA, 20 μM (NH4)2Fe(SO4)2, 200 μM ascorbate, 3 mM 2OG and 2 mM TMAEP (red), DMAEP (blue), or 2-AEP (green). Substrate standards are shown in gray. [TMAEP = 16.8 ppm  14.0 ppm; DMAEP = 17.7 ppm  14.3 ppm; 2-AEP = 18.9 ppm  15.0 ppm] ...... 155

Figure 3-6. Assays testing TmpA activity toward TMAEP in competition with alternative substrates, DMAEP and PC. All reactions were performed at 3 °C and contained 0.02 mM TmpA, 0.03 mM (NH4)2Fe(SO4)2, 0.4 mM ascorbate, 6 mM 2OG and 2 mM of each substrate. Each panel shows product formation over time, monitored by both LC-MS and 31P-NMR. (A) Control reactions containing either TMAEP (black) or DMAEP (blue). (B) Competition reactions containing both TMAEP and DMAEP. (C) Control reactions containing either TMAEP (black) or PC (red). Open red circles correspond to product detected from a reaction containing PC and 2 mM OH- TMAEP. (D) Competition reactions containing both TMAEP and PC. Products from this reaction were monitored by 31P-NMR only because PC and OH-TMAEP have the same m/z and similar retention times...... 157

Figure 3-7. Structural comparison of Lc TmpA and Hs BBOX. (A) Homodimer quaternary structure of TmpA with chain A in dark green and chain B in light green. Fe(II) ions are shown as brown spheres. (B) Homodimer quaternary structure of Hs BBOX (PDB accession code 4C8R) with chain A in dark gray and chain B in light gray. The Zn(II) ion is shown as a purple sphere. (C) Active site of the substrate- bound structure of TmpA. (D) Active site of the substrate-bound structure of Hs BBOX (PDB accession code 3O2G). The co-substrates [2OG, N-oxalylglycine (NOG)], amino acids side chains, and substrates [TMAEP (yellow), γbb (blue)] are shown in stick format. Electrostatic interactions are designated by black dashed lines and the black arrows denote the target carbon position for hydroxylation...... 160

anaerobic chamber. Experimental spectra are shown as black vertical bars. Overall simulations are shown as red lines. The two doublets constituting the simulation of the Fe2(II/II) spectrum are shown as purple and cyan lines with parameters described in the main text. The sub-spectra simulations of spectrum C corresponding to the differic species (orange) and the ferrous (blue) and ferric (green) sites of the Fe2(II/III) species are shown as solid lines, respectively, with parameters described in the main text. The sub-spectrum of the Fe2(II/II) species (22%) was subtracted from the raw experimental spectrum (Appendix B, Figure S19) to give the spectrum shown in panel C for simplicity...... 163

Figure 3-9. X-band EPR spectra of O2-free 0.25 mM TmpB incubated anaerobically with 10 mM sodium L-ascorbate without substrate (black) or in the presence of 10 mM (R)-OH-TMAEP (red) or (R)-OH-AEP (blue). Experimental conditions: temperature = 10 K, microwave power = 0.2 mW, microwave frequency = 9.479 GHz, modulation amplitude = 1 mT...... 164

Figure 3-10. Testing activity of the HD-MVDOs against aminophosphonates with and without N-methylation. Aerobic reactions containing 0.01 mM TmpB or PhnZ, 0.2 mM L-ascorbate, 2 mM substrate [(A) (R)-OH-TMAEP or (B) (R)-OH-AEP] were incubated for 4-h at 3 °C. (Left) 31P-NMR spectra monitoring disappearance of the substrates, (R)-OH-TMAEP (14.0 ppm) or (R)-OH-AEP (15.0 ppm), and production of phosphate (0 ppm). [*contaminant from the PhnZ protein preparation] (Right) LC-MS chromatograms detecting the substrates, (R)-OH-TMAEP (184 m/z) or (R)- OH-AEP (142 m/z), and products, glycine betaine (118 m/z) or glycine (76 m/z)...... 166

Figure 3-11. X-ray crystal structure of TmpB (A) Cartoon depiction of the α-helical tertiary structure characteristic of HD-domain proteins. Iron ions are shown as brown spheres and OH-TMAEP (yellow) is shown in stick format. A dashed line represents unmodeled regions of the structure. (B) Schematic of the first coordination sphere of the diiron cluster. (C) TmpB active site with (R)-OH-TMAEP bound. Protein residues and (R)-OH-TMAEP (yellow) are shown as sticks. Water molecules are shown as red spheres and iron ions as orange spheres. Blue mesh depicts the Fo – Fc omit map contoured at 3σ generated after ligand deletion and subsequent refinement...... 169

Figure 4-1. Chemical reaction catalyzed by the enzyme UndA, where n = 7, 9, or 11...... 196

Figure 4-2. Published mechanism(8) proposed for oxidative decarboxylation employing a mononuclear iron cofactor...... 197

Figure 4-3. General mechanistic scheme possible for oxidative decarboxylation employing a diiron cofactor...... 199

Figure 4-4. (A) Superposition of the known diiron protein, CADD, structure (teal, PDB accession code 1RCW) with that of UndA (purple, PDB accession code 4WWJ) using the PDBeFOLD server. (B) Zoom in view of the two active sites overlaid, depicting the residue ligands for the CADD diiron site and the analogous residues in the UndA primary structure...... 201

xvi

Figure 4-5. (A) Alignment of the ligand residues of selected FDCOOs (bold letters) with the putative ligand residues of UndA (bold letters). (B) Superposition of the core helices of the sMMOH protein structure (blue, PDB accession code 1MYH) with that of UndA (peach, PDB accession code 4WWJ) using the PDBeFOLD server...... 202

Figure 4-6. (Top) 4.2-K/53-mT Mössbauer spectrua of two different preparations (black line and black vertical bars) of aerobically isolated UndA protein. (Bottom) 4.2-K/6- T and 8-T Mössbauer spectra of the aerobically isolated UndA protein. The simulation of 6 T spectrum is shown as a solid red line...... 204

Figure 4-7. CW X-band EPR spectrum of the as-isolated UndA protein reacted with sub- stoichiometric sodium dithionite. Experimental conditions: temperature = 9.5 K, modulation amplitude = 1 mT, microwave power = xx, microwave frequency = 9.480 GHz ...... 205

Figure 4-8. Fe-K-edge EXAFS data (left) and Fourier transform (right) for samples of the aerobically isolated UndA protein. Fit (red) parameters are provided in Table 4-1...... 206

Figure 4-9. Fe-K-edge EXAFS data (left) and Fourier transform (right) for samples of the aerobically isolated UndA protein reduced with sodium dithionite. Fit (red) parameters are provided in Table 4-2...... 207

Figure 4-10. Anomalous density maps (black mesh displayed at a contour of 5σ) demonstrating the presence of two iron sites in the active site of UndA. Fe binding site 2, observed in the original structure, is on the right and site 1 is on the left...... 208

Figure 4-11. Reaction of reduced UndA in the absence or presence of substrate with dioxygen, monitored by SF-Abs spectroscopy. Absorption spectra monitoring the reaction of a solution of apo-UndA protein (0.3 mM), reconstituted with 1.8 molar equivalents of FeII (0.54 mM), with 1 mM LA (A) or LA omitted (B), mixed with an equal volume of O2-saturated buffer (~1.8 mM O2). Panels C and D depict the time- dependent traces of specified wavelenths. *The 412 nm trace was obtained by a dropline analysis of the spectrum, which entails subtraction of the average of the 408 and 416 nm traces from the raw 412 nm trace ...... 211

Figure 4-12. Amplitude change of chromophores as a function of Fe:UndA. Apo-UndA protein (0.3 mM) was reconstituted with varying molar equivalents of Fe(II), preincubated with 1 mM LA substrate and mixed with an equal volume of O2- saturated buffer (~1.8 mM O2 at 5 °C). Amplitudes were obtained from kinetic fits of time-dependent traces monitoring each specified wavelength. *The amplitudes of the 412 nm traces were obtained from the 412 nm traces by a dropline analysis of the spectrum, which entails subtraction of the average of the 408 and 416 nm traces from the raw 412 nm trace...... 212

Figure 4-13. Reaction of reduced UndA with O2 monitored by rFQ-Mössbauer spectroscopy. 4.2-K/53-mT spectra of samples obtained from mixing UndA protein (1.2 mM) reduced with sodium dithionite and preincubated with 10 mM LA substrate with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after various reaction incubation times...... 214

xvii

Figure 4-14. 4.2-K/8-T Mössbauer spectra of samples obtained from mixing UndA protein (1.2 mM) reduced with sodium dithionite and preincubated with 10 mM LA substrate with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after reactions of either 10 ms or 5 s. The 10 ms experimental reference spectrum was obtained by subtraction of the 40 % contribution of the unreacted species represented in the anaerobic control. The red line represents the simulation considering two components shown as green and blue lines...... 215

Figure 4-15. Time-dependent traces monitoring the 550 nm spectral feature in reactions in which the concentration of O2 was varied. Apo-UndA protein (0.3 mM) was reconstituted with 1.8 molar equivalents of FeII (0.54 mM) , preincubated with 1 mM LA substrate and mixed with an equal volume of oxygenated buffer with varying concentrations of O2 (5 °C) denoted in the figure...... 216

Figure 4-16. (Left) CW X-band EPR spectra of FQ samples obtained from reactions of apo-UndA (1.2 mM), reconstituted with 1.8 molar equivalents of FeII (2.2 mM), preincubated with 3.6 mM LA and mixed with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after various reaction times. Experimental conditions: temperature = 20 K, modulation amplitude = 4 G, microwave frequency = 9.625 GHz, microwave power = 20 μW. (Right) Overlay of the time-dependent trace of the 412 nm feature, obtained by the dropline analysis, with the % spin/UndA, determined by spin quantitation relative to a Cu2+-EDTA standard...... 217

Figure 4-17. CW X-band EPR spectra of rFQ samples obtained from reactions of apo- UndA (1.2 mM), reconstituted with 1.8 molar equivalents of 56- or 57-FeII (2.2 mM), preincubated with 3.6 mM LA and mixed with an equal volume of O2- saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after a reaction time of 70 ms. Experimental conditions: temperature = 20 K (70 K), modulation amplitude = 4 G, microwave frequency = 9.625 GHz, microwave power = 20 μW...... 218

Figure 4-18. Possible mechanisms for substrate decarboxylation following formation of the C3-substrate radical...... 222

xviii

LIST OF TABLES

Table 1-1. Spectroscopic parameters of observed diferric-peroxide species in proteins covered in this review...... 11

Table 2-1. Comparison of the yield of the Cys71-centered radical, as determined by FQ- EPR experiments (t = 4 s), and the normalized (with respect to WT ADO) activities (alkane yield) of ADO variants obtained from multiple turnover assays. The yield of the Cys71-SO• is given as a percentage with respect to the total ADO cofactor concentration...... 117

Table 3-1. Target ion m/z ratios detected in positive mode...... 182

Table 3-2. Target ion m/z ratios detected in positive mode...... 182

Table 4-1. EXAFS fitting parameters for the aerobically isolated UndA protein...... 206

Table 4-2. EXAFS fitting parameters for the aerobically isolated UndA protein reduced with sodium dithionite...... 207

Table 4-3. Data collection and refinement statistics for the x-ray structures of UndA┴...... 228

xix

ACKNOWLEDGEMENTS

First, thank you to my co-advisors, Marty and Carsten for providing the perfect balance of independence and guidance that allowed me to thrive. You have set me up for success by letting me participate in grant writing, manuscript preparation, project development, and mentoring. You have supported and encouraged me to go after challenges I might not have yet felt ready for. And although I may not have always heeded your advice, the confidence that you have in me will always stay with me.

Thank you to my third co-advisor, Maria, for being my constant guide, role model, and the most incredible mentor. Our connection as mentor/mentee, as scientists, as co-workers, and as friends was instant. You made my first few years in the lab amazing. I can only hope that one day

I will be as brilliant, kind, and thoughtful as you are. I hope to always make you proud.

Thank you to the BK lab during my time and the whole Penn State bioinorganic community for fostering a scientific environment that is supportive, collaborative, intellectual, and collegial. I am so fortunate to have been a part of such an amazing team. Hopefully our paths will continue to cross.

Thank you to my family for their constant love and support on my journey toward a scientific career. Thank you to my friends for being so understanding throughout this process and for always being there to listen, give advice, or to make life fun. Finally, I would like to thank my fiancé, Michael. Forgetting the miles put on our cars and all of the times I showed up late, you were always there to make every long day, late night, and bad day better. I am so glad I got to share this part of my life with you.

This material is based upon work supported by the National Science Foundation under

Award No. MCB-1122079, MCB-1330784, and CHE-1610676. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the National Science Foundation.

Chapter 1

Non-heme diiron oxygenases and oxidases

Enzymes that harbor dinuclear transition-metal cofactors with histidine and carboxylate ligands enable important biochemical processes in all three domains of life. In many cases, the cofactors promote redox reactions involving dioxygen. A functional paradigm for these enzymes emerged from early studies on the β (formerly R2) subunit of class I-a ribonucleotide reductase, soluble methane monooxygenase, and stearoyl-acyl carrier protein Δ9-desaturase. It comprises eight key features. (1) A diiron cluster held within (2) a conserved, α-helical "ferritin-like"

II/II structural domain activates (3) O2 from the (4) diferrous (Fe2 ) oxidation state. (5) A µ-peroxo-

III/III IV/IV III/IV Fe2 intermediate is produced, and it (6) converts to a high-valent (Fe2 or Fe2 ) complex, which effects (7) a one- or two-electron oxidation reaction that (8) cleaves at least one strong

C–H or O–H bond by abstraction of the hydrogen atom (H•). More recent studies have expanded this original paradigm in each of its eight features. (1) Ferritin-like proteins have been shown to use manganese-iron and dimanganese cofactors, and (2) O2-activating diiron clusters have been found in different protein architectures. (3) In some cases, a partially reduced dioxygen species, rather than O2 itself, reacts with the dimetal cofactor. (4) A new structural and functional

II/III subclass, in which O2 activation by a histidine/carboxylate-coordinated mixed-valent (Fe2 )

III/III cofactor produces (5) a Fe2 -superoxide intermediate that (6) directly abstracts H• to initiate (7) a four-electron oxidation, has been described. In other enzymes functioning in the canonical

II/II Fe2 manifold, (6) peroxide-level complexes themselves, rather than high-valent successors, initiate transformations, and novel (7) zero- and four-electron-oxidations that cleave (8) strong C–

C and N–H bonds, respectively, have been described. Here, we summarize these recent studies,

2 highlighting the specific structural and mechanistic variations that enable the newly recognized outcomes and thereby markedly expand the known repertoire of the functional class.

1.1 Introduction

Molecular oxygen is a powerful oxidant thermodynamically, but its triplet (S = 1) electron-spin ground state imposes a high kinetic barrier for direct reaction with singlet organic matter (1). Upon partial reduction, however, more reactive species – superoxide, peroxide, and hydroxyl radical – are produced. In biology, the controlled, partial reduction of O2 by an organic or transition metal enzyme cofactor is a versatile strategy to harness the potency of reduced oxygen species for useful chemical transformations. For metalloenzymes, the metal-oxygen adducts produced can be potent nucleophiles, electrophiles, or hydrogen-atom (H•) abstractors.

Collectively, these intermediates allow for transformation of even the most chemically inert centers of most organic compounds present in the biosphere.

In reductive activation of O2, the first step, reduction to superoxide, is by far the least thermochemically favorable (1). Enzymes with organic cofactors (e.g., flavins) can kinetically couple the initial uphill step with a second, favorable electron-transfer or radical-coupling step that generates a more stable (yet still potent) peroxide-level intermediate (2). In enzymes that

n activate O2 with transition metals, the right combination of metal ion (M ) and ligands can make addition of O2 more favorable, so that subsequent steps can proceed efficiently from an accumulating (formally) Mn+1–superoxide complex (1). Heme-dependent oxygenases exemplify

III –• this strategy, forming a common, stable Fe -O2 adduct known as "compound 0" that undergoes rapid reduction by outer-sphere electron transfer (3).

An alternative strategy for overcoming the unfavorable initial reduction step is approximation of two redox centers to enable concerted or coupled two-electron reduction of O2

3 to peroxide-level complexes. Multiple classes of O2-activating metalloenzymes conform to this general strategy, but perhaps the clearest embodiment is a large, versatile class of oxygenases and oxidases that employ dinuclear iron cofactors with histidine and carboxylate (glutamate/aspartate) protein ligands (4, 5). Early studies on three structurally related members of this class – the hydroxylase component of soluble methane monooxygenase (sMMOH) (6, 7), the β (formerly

R2) subunit of class I-a ribonucleotide reductase (RNR- β) (8, 9), and stearoyl-acyl carrier protein

Δ9-desaturase (Δ9D) (10, 11) – as well as inorganic models of these enzymes (12-15) revealed a

II/II largely conserved mechanism involving oxidative addition of O2 to the diferrous (Fe2 ) cofactor

III/III to form a peroxide-bridged diferric (Fe2 ) complex, homolytic or reductive cleavage of the

IV/IV III/IV peroxide bond to form a high-valent (Fe2 or Fe2 , respectively) complex, and H• abstraction (from carbon or oxygen) by this complex to initiate a simple one- or two-electron oxidation reaction. The past decade has seen a marked expansion of this functional paradigm to include dimetal cofactors comprising other transition metals (namely, manganese), reactions with reduced oxygen species rather than O2 itself, use of similarly coordinated mixed-valent diiron(II/III) clusters for O2 activation, direct reaction of mid-valent metal-superoxide and - peroxide intermediates with substrates, and more complex outcomes ranging from zero- to four- electron oxidations that target an expanded range of strong chemical bonds (e.g., C–C and N–H).

In this article, we first briefly review the early paradigm-establishing studies before turning to a more detailed analysis of the recent work that has dramatically broadened the known chemical repertoire of reactions between nonheme dinuclear transition metal cofactors and dioxygen species.

4 1.2 Canonical ferritin-like diiron-carboxylate oxidases and oxygenases (FDCOOs)

The three most extensively studied members of the functional class, sMMOH, RNR-β, and Δ9D, belong to a single structural superfamily known as the "ferritin-like" proteins (4, 5). The core architecture of this large group of soluble, cytosolic proteins is a four-helix bundle (Figure

1-1), in which each α-helix contributes at least one metal ligand. The conserved set of two histidine (H) and four-carboxylate [aspartate (D) or glutamate (E)] ligands comes from two recognizable E/DxnE/DxxH sequence motifs, which, together with the predicted helical fold, can reliably identify members of the superfamily (4). In general, each iron(II) site of the reduced form of the cofactor has a distorted square-pyramidal coordination geometry consisting of one histidine residue, one terminal carboxylate residue (coordinated in monodentate or bidentate fashion), and two bridging carboxylates, with at least one in the μ-1,3 bridging mode (Figure 1-1). The flexibility of this coordination sphere, in particular changes during the reaction cycle between bridging and terminal or mono- and bidendate carboxylate coordination modes (termed

“carboxylate shifts" by Lippard and co-workers (16)), would appear to have important functional implications for the geometry and kinetics of O2 addition and for control of the manner in which the initial adducts react (16-18).

II/II Each of these three proteins binds and activates O2 from its diferrous (Fe2 ) form. In each case, O2 ultimately undergoes a four-electron reduction to the oxidation state of two water

III/III molecules, as the cofactor is oxidized by two-electrons to its stable diferric (Fe2 ) "product" state. The source of the two additional electrons required to balance the oxidative half-cycle differs in the three systems. In the case of sMMOH, the hydroxylation of methane to produce methanol effectively extracts both balancing electrons. In this transformation, an O-atom from O2 is effectively inserted into one of the C–H bonds of the substrate (19). In Δ9D, a distinct two- electron oxidation of the substrate, stearic acid bound as thioester to the phosphopantetheine cofactor of acyl carrier protein (ACP), provides both balancing electrons. The fatty acyl chain is

6 dehydrogenated in a reaction that cleaves C–H bonds of adjacent carbons and inserts a cis double bond. In the case of RNR-β, an electron and proton are removed from the phenol side chain of a tyrosine residue, and a stable tyrosyl radical, required for RNR-β to cooperate with its partner α subunit in nucleotide reduction, is introduced. In vitro, the second balancing electron can be provided by one of several exogenous reductants (e.g., ascorbate, thiols, FeII) (20-23). A ferredoxin protein, YfaE, has been implicated in provision of this "extra" electron in vivo during activation of the I-a RNR-β from Ec (24).

The conversion of the cofactor to its diferric form in each oxidative half-cycle mandates that it be reduced back to the diferrous state for subsequent turnovers. sMMO has a dedicated reductase component (sMMOR) with flavin and iron-sulfur cofactors to conduct electrons from

NADH to the cofactor in the hydroxylase component, sMMOH (25, 26). A "coupling" component

(sMMOB) dynamically binds to and dissociates from MMOH during the full cycle to coordinate the oxidative and reductive half-cycles (27). A ferredoxin with a [2Fe-2S] cofactor can reduce the diferric cluster in Δ9D (28, 29). The RNR-β reaction is a post-translational auto-activation process and need not (necessarily) occur multiple times. However, the tyrosyl-radical product of the reaction is subject to adventitious (and possibly regulatory) reduction in vivo and can, accordingly, be regenerated following in situ reduction of the diferric cluster back to the O2-

II/II reactive Fe2 state. The aforementioned ferredoxin, YfaE, has been implicated in this reduction/regeneration (maintenance) process (24).

Foundational mechanistic studies over the last thirty years on these three founding members of the functional class elucidated a number of unifying principles. This work has been exhaustively reviewed (6-11, 30-32) and is summarized here only briefly, to provide proper context for understanding the novel chemistry of the more recently recognized members. All three reactions begin with oxidative addition of O2 to the diferrous cofactors. For sMMOH, this addition is markedly accelerated ("triggered") by complexation with sMMOB (27, 33). In Δ9D,

7 binding of the stearoyl-ACP substrate triggers O2 addition (34). O2 addition in the RNR-β reaction is rapid in the absence of additional factors. Accumulation of a common peroxodiiron(III/III) complex, implicated by spectroscopic data as having a cis-µ-1,2-peroxide bridge, has been observed in each system under appropriate conditions (Figure 1-2) (33-40). The complex was termed Hperoxo or P in studies on sMMOH, where it was first detected (33, 41).

General hallmarks of the structural type, summarized in Table 1-1, are: (i) Mössbauer quadrupole doublet features with isomer shift values that are unusually high (δ ~ 0.6 mm/s) for high-spin FeIII complexes with oxygen and nitrogen ligands; (ii) broad, moderately intense, long-wavelength

-1 -1 absorption bands (λmax ~ 700 nm, ε700 ~ 2,000 M cm ), and (iii) O–O stretching modes at ~ 880 cm-1 observable by resonance Raman spectroscopy with excitation into the ~ 700-nm absorption band (Table 1-1) (33, 34, 37-40). The similarity of these spectroscopic parameters to those of relevant inorganic complexes lends credence to the structural assignment (14, 42, 43).

By contrast to the RNR-β and Δ9D reactions, in which the µ-1,2-peroxides form rapidly and with no apparent accumulation of prior intermediates, kinetic evidence for one precursor to P

(termed O) and spectroscopic evidence for a second (termed P*) were reported for the reaction of

II/II sMMOH as initiated by the mixing of the Fe2 -sMMOH•sMMOB complex with O2 (44-47).

Both precursors were shown to have their cofactors still in the reduced (diferrous) state, leaving

II/III their chemical identities mysterious. The absence of specific evidence for a Fe2 -superoxide- level complex in any of the reactions seems consistent with the notion of concerted (or tightly kinetically coupled) two-electron reduction of O2 upon its addition in a bridging mode, but the characteristics of P* make it a potential intermediate in a more complex, sequential O2-addition pathway. Solomon and co-workers have argued that a perpendicular disposition of the relevant

FeII-based frontier molecular orbitals (FMOs) favors concerted addition (48, 49). The binding

9 events that trigger the O2 reactions (sMMOB to sMMOH, stearoyl-ACP to Δ D) cause adjustments within the first coordination spheres (50) that could bring about this optimal FMO

8 alignment (48). Conversely, coordination geometries yielding sub-optimal FMO orientations could disfavor rapid formation of the canonical cis-µ-1,2-peroxo adduct and perhaps allow structurally distinct complexes with different reactivities to form. Control of the geometry of the initial adduct could thus be important for directing reactions down divergent pathways to the manifold of outcomes represented within the enzyme class.

Although they are thought to be structurally homologous, the peroxide complexes in the three systems have lifetimes that vary by more than 106-fold. In wild-type RNR-β from

Escherichia coli (Ec), the complex decays at > 300 s-1 (5 °C) and thus accumulates minimally and very early (< 3 ms) in the reaction (35). Its fleeting nature initially challenged efforts to establish that it is "on pathway." Substitution of the unique Fe1 aspartate ligand with glutamate was found to stabilize the complex markedly, while still permitting stable tyrosyl radical production (36, 37), and a spectroscopically similar species was found to decay less rapidly in the reaction of wild- type RNR-β from Mus musculus (mouse) (39), making it more feasible to establish the precursor- product relationships. These results bolstered the conclusion of the initial study that it is an on- pathway intermediate in the reaction of wild-type Ec RNR-β (35). The analogous peroxide complex in sMMOH decays with a rate constant of 0.45 s-1 (4 °C) (33). In Δ9D, the complex is remarkably stable, decaying at 6.7 × 10-5 s-1 (6 °C) (40). Importantly, this very slow decay does not result in desaturation of the acyl-ACP substrate. Indeed, the peroxide complex was detected in

9 Δ D only upon reaction of the chemically reduced (dithionite treated) enzyme with O2 in the presence of stearoyl-ACP, whereas only the ferredoxin-reduced enzyme is competent for turnover

(51). The fact that no chemically competent intermediates have been characterized to date in the

Δ9D reaction leaves significant gaps in our understanding of its chemical mechanism, even in comparison to more recently discovered systems.

The mechanistic pathways for decay of the peroxide complexes are also different. In

IV/IV sMMOH, internal redox cleaves the peroxide O–O bond and generates the Fe2 complex

9 designated Q (Figure 1-2, pathway I) (52). Observation of a substrate deuterium-kinetic isotope effect (2H-KIE) on the decay of intermediate Q implicated this species as the H•-abstracting complex (53). Computational work supports this conclusion and provides a consensus view that

IV/III the resultant substrate radical couples with a formally hydroxyl radical harbored on the Fe2 complex to yield the methanol product (54-57). Extensive efforts to define the structure of Q by spectroscopic methods have not yet brought about consensus, but the original proposal by Que,

IV/IV Lipscomb, Münck and co-workers of a di-(µ-oxo)-Fe2 "diamond core" is currently considered most likely (58) and is consistent with the results of recent continuous-flow resonance Raman- spectroscopic experiments (59). In Δ9D, the unproductive decay pathway appears to involve electron transfer between cofactors in paired monomer units of the homodimer, similar to the inter-subunit ET step proposed earlier in a study of the Ec RNR-β activation reaction (60).

Despite the lack of experimental evidence for intermediate states on the productive pathway, a mechanism similar to that for sMMOH can be envisioned for Δ9D, in which a peroxo complex decays by internal redox cleavage of the O–O bond to yield a Q-like state. This complex could

IV/III abstract the first aliphatic H• from the substrate to yield a Fe2 state and a substrate radical.

Then, diverging from the mechanism of sMMOH, this resultant complex could remove a second

H• from the adjacent carbon position on the substrate, followed by radical coupling to form the olefin product (Figure 1-2). For Ec RNR-β, evidence suggests that redox-neutral peroxide bond cleavage does not readily occur. Instead, the peroxide complex, or an extremely reactive and non- accumulating peroxide-level successor, undergoes reductive O–O bond cleavage (Figure 1-2, pathway II). A near-surface tryptophan (Trp) residue donates the electron, and the resultant Trp+• is then reduced from solution to complete the shuttling of the "extra" electron to the cofactor (23,

61). Evidence was presented that a distinct intermediate state, which lacks the long-wavelength absorption and high Mössbauer isomer shift of the canonical µ-1,2-peroxide, accepts the electron from the tryptophan in this reaction (62). This state has been suggested to result from protonation

10 of the bridging peroxide, a step expected to activate it for reductive O–O-bond cleavage (63). The

III/IV Fe2 complex resulting from this reduction, the first intermediate to be detected in the functional class, has been called X (20). It can oxidize the conserved tyrosine by one electron via

H• abstraction to its enzymatically functional tyrosyl radical form (Figure 1-2). As in the case of sMMOH Q, the structure of X in RNR-β has not been resolved. To summarize the controversy as it currently stands: (1) EXAFS (extended X-ray absorption fine-structure) measurements interpreted via DFT (density function theory) calculations yielded an Fe–Fe distance consistent

III/IV with an di-(µ-oxo)-Fe2 core (64); (2) (M)CD [(magnetic) circular dichroism] and computational analysis by the Solomon group suggested that a diamond core could have only one

µ-oxo ligand and, consequently, that the other bridging O-atom would necessarily have to be protonated (i.e., a µ-hydroxo ligand) (65), but 1,2H-ENDOR (electron-nuclear double resonance) experiments have consistently failed to detect the hydron of the bridging hydroxo ligand required by the Solomon model (66); and (3) 17O-ENDOR results revealed only one solvent-equivalent single-atom oxygen bridge (67). We believe that, on balance, the evidence slightly favors a di-(µ-

III/IV oxo)-Fe2 core structure, but this assignment remains uncertain.

Despite the prevailing differences in the natural reaction pathways, studies in which alternative outcomes have been elicited have underscored the overlapping inherent reactivities of the individual cofactors and their adducts with O2. For example, some fatty acid desaturases retain latent hydroxylase activity (68, 69), and others have been manipulated to effect hydroxylation reactions either by amino acid substitutions or by use of strategically deuterated substrates (70, 71). Similarly, RNR-β proteins that self-hydroxylate a phenylalanine residue

(F208) or a tyrosine in its place have been reported (72-74). Alternative outcomes on non-native substrates have been demonstrated for sMMOH, and in some cases (e.g., in oxidation of ethers) evidence suggests direct reaction with the peroxide complex rather than Q (75). The versatility and subtlety of control of this reaction manifold has motivated efforts to deploy the ferritin-like

11 structural scaffold as a template for de novo design of artificial metalloproteins, designated Due

Ferri (DF) proteins (76-78). These minimalistic constructs are capable of accommodating a diiron cluster and have been engineered to activate O2 and oxidize substrates, including phenols and arylamines (79-81). This work has served to highlight the subtle but important influences of differences in the protein architecture on cofactor reactivity and reaction outcome.

Figure 1-2. Canonical mechanisms for oxidation reactions catalyzed by FDCOOs. Pathway I shows those proposed for desaturation (e.g., Δ9D) and hydroxylation (e.g., sMMOH). Pathway II corresponds to assembly of the I-a RNR-β active cofactor. Table 1-1. Spectroscopic parameters of observed diferric-peroxide species in proteins covered in this review. EXAFS resonance Diferric- UV/visible Isomer Quadrupole Fe-Fe Raman peroxide absorption shift, splitting, References distance frequencies species bands (nm) δ (mm/s) ΔE (mm/s) Q (Å) (cm-1) sMMOH 625-650 0.66 1.51 (33) Δ9D 700 0.68, 0.52 1.9, 1.06 ν(O-O) 898 (40) Ec RNR 2.5 ν(O-O) 868 700 0.63 1.73 (36, 37, 39) β-subunit (D84E) (D84E) ToMOH 420* 0.55 0.67 - (82, 83) AurF 500 0.54, 0.61 -0.66, 0.35 - (84) CmlI 500 0.62, 0.54 -0.23, -0.68 ν(O-O) 791 (85) ADO 450 0.48, 0.55 0.49, 1.23 - (86) 1.16, 0.88 0.55, 0.58 DOHH 630 (1.05, 0.99) 3.44 ν(O-O) 855 (87) (0.49, 0.63)┴ ┴ *observed only in the presence of all component proteins ┴parameters for non-nested pair of doublets

12 1.3 Novel outcomes and variations on the FDCOO mechanistic theme

The identification of C/O–H-bond-cleaving high-valent complexes in sMMOH and RNR-

β confirmed the anticipated mechanistic analogy of these two FDCOO systems to the longer- known heme-utilizing cytochrome P450s, which use ferryl/ligand-radical states known as compound I for aliphatic C–H activation (3, 88). However, also mirroring the remarkable mechanistic diversity of the heme proteins (3, 88), alternative manifolds of reactivity that appear not to involve high-valent complexes soon began and still continue to emerge. In one set of enzymes, peroxide-level complexes appear to react formally as electrophiles with substrate nucleophiles, whereas, in another case, the Fe-bound O–O unit reacts, in effect, as the nucleophile with the electrophilic carbonyl group of a bound substrate. In most cases, the nature of the bond- forming steps, polar or radicaloid, remains to be established. In considering the overarching question of how proteins in this class control and direct their reactions to the divergent set of outcomes, two extreme cases may be envisaged. On the one hand, a structurally conserved initial

II/II Fe2 -O2 adduct could be directed by structural/dynamical differences among the individual proteins down different pathways to the distinct outcomes. Alternatively, structural differences among the scaffolds might instead select at the outset for different initial adduct geometries with divergent inherent reactivities. The peroxide intermediates described thus far are all thought to have the cis-μ-1,2 coordination mode, and they all most likely serve as precursors to the high- valent intermediates necessary to abstract hydrogen atoms strongly bonded to sp3-hybridized carbons or heteroatoms. However, a number of alternative O2-addition modes, including end-on,

μ-1,1, μ-η1:η2, and μ-η2:η2, are also possible with a dimetal cofactor (Figure 1-3). These distinct adducts are expected to exhibit different reactivity, as has been seen with dicopper enzymes and inorganic models (14, 89, 90). Moreover, the potential for protonation of the peroxide unit further expands the range of possible intermediate structures. Indeed, for the aromatic hydroxylases and

13 the N-oxygenases in the FDCOO family (discussed in the next section), it has been suggested that the diferric-peroxide intermediates differ from the sMMOH, RNR-β, and Δ9D complexes in peroxide coordination mode, protonation state, or both, and that these differences activate the complexes for direct substrate oxidations not requiring prior conversion to high-valent complexes.

Figure 1-3. Possible (hydro)peroxide binding geometries to dinuclear metal clusters.

1.3.1 Electrophilic diferric-peroxide intermediates as oxidants.

1.3.1.1 Aromatic hydroxylases.

Toluene/o-xylene monooxygenase (ToMO), toluene-4-monooxygenase (T4MO), and phenol monooxygenase (PMO) are bacterial multicomponent monooxygenases (BMMs) in the same functional subgroup as sMMO (91, 92). Their hydroxylase components (denoted by adding "H") are, like sMMOH, ferritin-like proteins. By contrast to the aliphatic substrates oxidized by sMMOH, the aromatic targets for these BMMs are probably not amenable to hydroxylation mechanisms involving H• abstraction by high-valent complexes. Studies on ToMOH and

T4MOH have suggested that, instead, the initial peroxide adducts undergo nucleophilic attack by the π-systems of the aromatic substrates. The first experimental evidence for the intermediacy of

III/III a Fe2 -peroxide complex was obtained by interrogation of the reaction the Ile100  Trp

14 variant of ToMOH (82). The reduced protein reacted with dioxygen to produce an intermediate characterized by a Mössbauer quadrupole doublet with parameters (δ = 0.52 mm/s, ΔEQ = 0.62

III/IV mm/s) characteristic of high-spin ferric ions. In the variant protein, it decayed to a stable Fe2 complex with a radical on the engineered Trp100 (82, 93), a pathway with obvious analogy to that occurring in Ec class I-a RNR-β. By contrast, the indistinguishable (by Mössbauer) complex

III/III in the reaction of the wild-type enzyme was later shown to convert directly to the stable Fe2 product form (83). The Mössbauer characteristics of this species, coupled with chemical considerations and the detection of hydrogen peroxide upon its non-productive decay in the

III/III absence of substrate (83), led to the conclusion that it is a Fe2 -peroxide complex. Decay of the intermediate was enhanced 40-fold by the presence of arene substrates, and no evidence for accumulation of a high-valent intermediate was obtained, consistent with the hypothesis the substrate reacts directly with the peroxide complex (83).

The initial reports on the ToMOH peroxide complex noted a lack of visible absorption features (83), but a more recent study associated it with a weak feature centered at 420 nm (94).

The authors suggested that the difference arose from the presence of the ToMOD protein in the experiments of the second study. They proposed that binding of ToMOD impacts the orientation of a particular glutamate ligand, as suggested by differences in the structures of T4MOH alone and in complex with T4MOD (29), and that the conformation of this Glu impacts the electronic structure and absorption spectrum of the peroxide complex by enforcing altered hydrogen- bonding (vide infra). Irrespective of these subtleties, the spectroscopic features of the ToMOH peroxide complex are obviously distinct from those of the aforementioned μ-1,2-peroxo complexes in sMMOH, RNR-β, and Δ9D, implying a different structure.

A computational study concluded that the small quadrupole splitting of the ToMOH peroxide complex (ΔEQ = 0.62 mm/s) could be rationalized only by assumption of a protonated

III/III peroxide moiety and suggested a µ-1,1-(2-hydroperoxo)-Fe2 structure (95). Subsequent studies

15 implicated a conserved threonine residue (Thr201), positioned above the diiron cluster, in shuttling a proton to the peroxide and stabilizing the protonated adduct by hydrogen bonding (95-

98). This proposal is reminiscent of the prevailing view of the cytochrome P450 pathway, in which a similarly disposed threonine is thought to shuttle protons to the dioxygen unit to enable its heterolysis to produce compound I (99-101). Support for the role of this threonine in ToMOH was provided by a subsequent study on the Thr201  Ser variant. Reaction of its reduced form with O2 generates, in addition to the usual ToMOH peroxide intermediate, a small quantity of a

III/III second Fe2 -peroxide species with the long-wavelength (675 nm) absorption band and high

Mössbauer isomer shift characteristic of the canonical μ-1,2-peroxide complexes (Table 1-1)

(96). Computational analysis suggested that the engineered Ser201, lacking the constraining C4 methyl of the native Thr, might adopt two side-chain conformations leading to two different complexes, with protonated µ-1,1-hydroperoxo and unprotonated µ-1,2-peroxo cores (95).

IV/IV QM/MM calculations suggested that conversion of either species to an Fe2 Q-like state would be endothermic (95), again consistent with the direct reaction of substrate with the peroxide-level complex rather than a high-valent successor.

Experimental insight into the nature of this direct reaction was provided by examination of linear-free-energy relationships (LFER) with T4MO and ToMO substrates bearing ring substituents (83, 102). Hammett analyses gave the large negative values of ρ expected for a mechanism in which positive charge develops on the ring in the transition state. The observations were interpreted in favor of nucleophilic attack of the substrate π-system upon the oxidized diiron intermediate (i.e., electrophilic aromatic substitution, EAS) as shown in Figure 1-4. Because the protonated adduct would be more electrophilic than its unprotonated counterpart, this mechanistic view dovetails nicely with the proposed structure of the intermediate. However, mechanisms initiated by single-electron transfer (SET) can be envisaged (Figure 1-4) and would also develop positive charge in the transition state; SET and EAS mechanisms often cannot be distinguished

16 solely on the basis of such LFER analysis. Nevertheless, the view emerging from these insightful studies is of a peroxide complex, electrophilically activated by protonation and stabilized by hydrogen bonding against progression to a high-valent state, promoting C–O bond formation in one (EAS) or two (SET) steps by electron donation from the π-system of the substrate.

Figure 1-4. Electrophilic aromatic substitution and single electron transfer mechanisms for arene oxidation invoking a diferric-hydroperoxide intermediate.

1.3.1.2 Arylamine N-oxygenases.

Nitroaryl fragments are found in several natural products with antibacterial, antifungal, and antitumor activities (103-105). Aureothin (Aur) and chloramphenicol (Cml) are among the antibiotics in this class (Figure 1-5). The nitroaryl groups in these compounds are produced by the action of a pair of structurally similar ferritin-like diiron N-oxygenases, denoted AurF and

CmlI, upon arylamine precursors (103, 105). AurF oxidizes 4-aminobenzoate (Ar-NH2) to 4- nitrobenzoate (Ar-NO2) on the pathway to aureothin (Figure 1-5) (103). CmlI converts the

17 chloramphenicol arylamine precursor (CAM-NH2) to the actual antibiotic with its nitroaryl moiety (CAM-NO2) (Figure 1-5) (105, 106). By contrast to the two-electron oxidations most common for FDCOOs, these N-oxygenases effect overall six-electron oxidations of their respective N-aryl substrates (106, 107). Evidence suggests that the reactive intermediates for both

AurF and CmlI are, as in the arene hydroxylases, peroxide-level complexes (84, 85).

Figure 1-5. Aureothin and chloramphenicol abbreviated biosynthetic pathways, highlighting conversions by the diiron N-oxygenases (blue) and β-hydroxylase (red).

1.3.1.2.1 The diferric-peroxide complexes of the N-oxygenases

The nature of the AurF cofactor was initially obscured by conflicting results. Various preparations of the protein were found to have either primarily iron (108) or primarily manganese

(109), but in each case with a lesser quantity of the other metal. The discrepant metal content and the nearly contemporaneous elucidation of an RNR-β that employs a heterodinuclear Mn/Fe cluster for radical initiation (see below) (110) led to speculation that AurF might employ an analogous mixed-metal cofactor (111). Ultimately, the Zhao group established that the functional

18 cofactor is, in fact, a diiron cluster and solved its three-dimensional structure (112). The protein was seen to have a second His ligand to Fe1, for an overall 3-His/4-carboxylate coordination sphere. This extra His residue is also conserved in CmlI (113) and may be a crucial means by

II/II which the aryl-N-oxygenases control both the structure of the initial Fe2 -O2 adduct and the pathway/outcome of the reaction.

II/II A study published soon after demonstrated rapid reaction of reduced (Fe2 ) AurF with dioxygen to produce a long-lived (t1/2 ~ 7 min at 25 °C) complex with a broad absorption shoulder at 500 nm (84). Upon exposure to the arylamine substrate (Ar-NH2), the complex decayed rapidly (on the ms time scale) with rates proportional to substrate concentration (84).

The low-temperature/weak-field Mössbauer spectrum of the complex demonstrated a pair of quadrupole doublets with parameters (δ1 = 0.54 mm/s, |ΔEQ,1| = 0.66 mm/s; δ2 = 0.61 mm/s,

|ΔEQ,2| = 0.35 mm/s) characteristic of high-spin ferric ions, indicating oxidation of the cluster by two electrons, presumably concomitant with two-electron reduction of dioxygen to peroxide (84).

The hypso- and hypo-chromicity of the AurF complex relative to the canonical μ-1,2-peroxo-

III/III Fe2 intermediates and its distinct Mössbauer parameters (Table 1-1) imply that it has a different structure. The site-differentiation in the Mössbauer spectrum could be rationalized by an asymmetric peroxide coordination mode or distinct coordination environments of the two iron ions imposed by the additional histidine ligand at site 1. The similarity of the (small) values of

ΔEQ to the value for the ToMOH peroxide complex (0.67 mm/s) suggests that the AurF complex might also be protonated, and thus, electrophilic. NRVS coupled with DFT calculations on the complex support the assignment of a protonated peroxo species and its role as an electrophile in arylamine oxidation (114). Of the structures examined computationally, the most energetically favorable binding mode that can also reproduce the experimental data is a μ-1,2-hydroxoperoxo

(114). However, the authors do posit that this form could rearrange in the presence of substrate to

19 a μ-1,1-hydroperoxo species to undergo reaction with the amine (Figure 1-6). This latter structure is consistent with QM/MM modeling for the AurF peroxo structure (115).

III/III Examination of the chloramphenicol N-oxygenase, CmlI, revealed that similar a Fe2 -

II/II peroxide complex with an absorption band at 500 nm formed upon reaction of the reduced Fe2 enzyme with O2 and decayed rapidly in the presence of substrate (85). The 4.2-K/weak-field

Mössbauer spectrum of the complex, two quadrupole doublets with (small) quadrupole splittings comparable to those of the AurF intermediate (85), is consistent with the expectation that the two enzymes would employ similar oxygenating intermediates. The greater stability of the CmlI species in the absence of substrate (t1/2 ~ 3 h at 4 °C) enabled further structural studies by resonance Raman spectroscopy (85). The presence of a peroxide ligand was verified by detection of an O-isotope-sensitive vibrational mode [ν(16O-16O)] at 791 cm-1 (85). The frequency of this mode is considerably less than that for the canonical μ-1,2-peroxo complexes (Table 1-1), consistent with a different coordination geometry, protonation state, or both. Soaking of crystals

III/III of the µ-oxo-Fe2 resting state of CmlI with H2O2 did yield a structure with density modeled as a putative µ-1,2 peroxide, but the authors speculated that this species is not the relevant intermediate state since no substrate conversion was observed upon incubation of CmlI with H2O2

(113). Instead, based on the distinct vibrational frequency observed in comparison with mixed- metal (Fe/Cu) inorganic model complexes with alterative peroxo-binding modes, the authors

1 2 III/III posited a μ-η ,η -peroxo-Fe2 structure for the active complex (116). Recently, further structural characterization of the peroxo complex by resonance Raman and x-ray absorption

III/III spectroscopies led to a revision of the structure proposal to a μ-oxo-μ-1,1-peroxo-Fe2 species

(117). The presence of a μ-oxo ligand was supported by the detection of a short distance, light atom scatterer in EXAFS measurements and an O-isotope-sensitive vibrational mode [νs(Fe–O–

Fe)] at 485 cm-1 (117). Two other light atom scatterers were detected at distances of 1.98 Å and

2.82 Å that were attributed to the bridging O-atom and the non-coordinated O-atom of the

2 peroxo, respectively (117). The basis of the latter assignment was the low σ value of the scatterer, interpreted to result from spacial fixation of the atom symmetrically between the two Fe ions. The low vibrational frequency for the peroxo was reproduced and was further shown to be insensitive to D2O solvent (117), suggesting that the peroxo moiety in this complex is deprotonated, in contrast to the proposal for the AurF complex, and only becomes protonated in the presence of substrate. The authors rationalized that a peroxo structure that is deprotonated could maintain an ambiphilic character to act as an electrophile in the first step and as a nucleophile in subsequent oxidation steps (vide infra).

Based on the postulated electrophilic character of the AurF and CmlI intermediates, a mechanism for arylamine oxidation invoking nucleophilic attack by the lone pair of the substrate amine on the peroxo (Figure 1-6), similar to the substrate π-system attack on the ToMOH intermediate (84), can be formulated. Alternatively, a mechanism can also be envisioned that initiates by single electron transfer from the arylamine to the electrophilic peroxide (Figure 1-6).

Reductive cleavage of the O-O bond would yield an "open-core" (as opposed to cyclic "diamond-

III/IV core") Fe2 intermediate. Subsequent proton transfer and recombination of the nitrogen-

III/III centered radical with the hydroxo ligand would yield the hydroxylated product and µ-oxo-Fe2 cofactor state. Although no evidence for a high-valent complex has been reported in either AurF or CmlI, such a mechanism cannot be excluded based on the present experimental evidence.

1.3.1.2.2 Six-electron oxidation via 2+4 or 4+2 pathway?

By analogy to the two-electron oxidations known for FDCOOs, the initial published model proposed three consecutive two-electron oxidations for the overall oxidation of the amine to the nitro moiety (108). According to this model, the expected immediate product of the reaction of the diferric-peroxide with the arylamine substrate would be the hydroxylamine

III/III derivative (Ar-NHOH), with concomitant production of a µ-oxo-Fe2 cofactor form. Indeed,

III/III the latter rapidly formed upon exposure of the AurF Fe2 -peroxo to Ar-NH2. It was noted, however, that treatment of the AurF peroxide complex with sub-stoichiometric (< 0.3 equiv) Ar-

NH2 resulted in production of primarily Ar-NO2 (84), hinting that the intermediate might be capable of effecting each of the (presumptively three, two-electron) oxidation steps in the sequence.

A surprise was encountered when this inference was tested by direct examination of the reaction of the 4-hydroxylaminobenzoate (Ar-NHOH) intermediate with the AurF peroxide

22 complex. Rather than effecting a further two-electron oxidation to the nitroso compound (Ar-NO)

III/III as it converted to the µ-oxo-Fe2 cluster, the intermediate reacted to regenerate the fully

II/II reduced (Fe2 ) form of the cofactor, implying a four-electron oxidation of Ar-NHOH to Ar-NO2 to complete the overall "2+4 pathway" (Figure 1-7) (107). This implication was verified by direct

III/III analysis of reaction products, and it was shown that the Ar-NHOH + Fe2 -peroxide  Ar-NO2

II/II + Fe2 four-electron oxidation is catalytic (~ 30 turnovers) in the presence of excess O2 (107). In

III/III addition, evidence was presented that the anticipated µ-oxo-Fe2 /Ar-NO intermediate state fails to accumulate to levels detectable by Mössbauer spectroscopy (< 5% of total iron) (107). This study mandated a revision of the overall pathway from a requirement for three dioxygen and six electron equivalents to two dioxygen and two electron equivalents per six-electron cycle.

III/III In an analogous fashion, the Fe2 -peroxide complex in CmlI was shown to be competent for all steps in the Ar-NH2  Ar-NO2 six-electron oxidation (105, 106). Again, evidence suggested a two-electron and a four-electron oxidation, consistent with conservation of

– the overall reaction stoichiometry, Ar-NH2 + 2 O2 + 2 e + 2 H+  Ar-NO2 + 2 H2O, for the two

N-oxygenases. However, the opposite order to that suggested for AurF was proposed for the CmlI six-electron oxidation (106), whereby a four-electron oxidation of the amine to the nitroso intermediate (CAM-NO) is followed by a two-electron oxidation to the CAM-NO2 product in a

"4+2 pathway" (Figure 1-7). In the 4+2 pathway, the Ar-NHOH intermediate product would not

III/III dissociate and the µ-oxo-Fe2 cofactor form would not be reduced by an exogenous source, so

II/II that the two would react to generate the Ar-NO/Fe2 state. This complex would most likely react with O2 to reform the peroxide complex, which would effect the final Ar-NO  Ar-NO2 oxidation, but could also release the Ar-NO intermediate into solution prior to addition of the

III/III second O2 equivalent to the reduced cofactor to reform the Fe2 -peroxide (106). A kinetic dissection of the steps in the CmlI reaction demonstrated that the reaction from the CAM-

III/III II/II NH2/F2 -peroxo state to the CAM-NO/Fe2 state is rapid, disfavoring dissociation of the

23 CAM-NHOH intermediate (118). Additionally, reduction of the resting, oxidized form of CmlI by exogenous CAM-NHOH was significantly slower than that observed for reduction by CAM-

II/II NHOH generated in situ (118). Furthermore, whereas reaction of the CAM-NO/Fe2 complex with O2 is rapid, reaction of CAM-NO with pre-formed peroxo does not occur (118), such that

NO dissociation precludes processive oxidation to the final CAM-NO2. Collectively, the CmlI reaction kinetics favor a pathway that proceeds without dissociation of intermediate products.

Similar experiments had been performed with AurF to examine the possibility of the “4+2 pathway”. As observed for CmlI, the resting, oxidized form of AurF was slow to react with exogenously provided Ar-NHOH (107), and thus, deemed unlikely to compete with either Ar-

NHOH release or cofactor reduction by an auxiliary reductant. However, unlike CmlI, the

II/II reaction of AurF peroxo with Ar-NH2 does not regenerate the Fe2 form rapidly (107). In this case, the kinetics of dissociation or cofactor reduction via an exogenous reductant would likely predominate in the “2+4” pathway.

Therefore, despite catalyzing similar reactions, the present evidence indicates that AurF and CmlI operate through different pathways. It is possible that intrinsic differences between the substrates (e.g., different reduction potentials) could cause the enzymes to follow different pathways, but it is more likely that (1) both sequences are possible for both enzymes and (2) the predominant pathway is determined by the details of the partition among the possible fates of the

III/III Ar-NHOH/µ-oxo-Fe2 state produced in the initial oxidation step. The “4+2 pathway” would conform more closely to the modus operandi of the canonical diiron oxygenases, because the full reaction cycle would then comprise a single oxidative half-cycle (albeit one comprising two separate O2-dependent steps in an overall six-electron oxidation) followed by a simple reductive half-cycle, whereas the 2+4 pathway would have the oxidative sequence interrupted after the initial two-electron oxidation by the cofactor-reduction event, release of the Ar-NHOH intermediate to react with a second equivalent of peroxide complex, or both. In this regard,

24 identification of the native reducing system would seem relevant to an understanding of the physiological mechanism for each enzyme.

Figure 1-7. Divergent pathways suggested for the overall six-electron oxidation sequences of the N-oxygenases, AurF (107) and CmlI (106).

1.3.2 A nucleophilic dioxygen moiety in aldehyde deformylating oxygenase.

Although the reactions discussed thus far reflect the electron deficiency and

III/III electrophilicity of (hydro)peroxo-Fe2 complexes, chemical and enzymatic precedents establish that M–O–O– and R–O–O– units (where M denotes a transition metal and R an organic functional group) can also be nucleophilic. For example, the C4a-hydroperoxide intermediates of the flavin- dependent enzymes para-hydroxybenozate hydroxylase and phenol hydroxylase are attacked by the electron-rich π-systems of their aromatic substrates, resulting in hydroxylation (as in

ToMOH) (119-122), whereas the flavin C4a-peroxide intermediates of cyclohexanone monooxygenase and bacterial luciferase attack carbonyl groups of their substrates, leading to oxygen insertions (121, 123, 124). The ambivalent reactivity of peroxide complexes is also illustrated by the presumptively nucleophilic intermediates in certain cytochrome P450 systems.

25 For example, in progesterone 17α-hydroxylase-17,20-lyase, it is thought that a peroxo-FeIII complex attacks the substrate carbonyl to initiate C–C-bond cleavage (125).

The first example of an apparently nucleophilic O–O unit in the reaction of a ferritin-like diiron enzyme was provided by cyanobacterial aldehyde-deformylating oxygenase (ADO), an enzyme capable of producing linear hydrocarbons equivalent to commercial diesel and jet fuels.

Production of alkanes and alkenes from fatty acids occurs naturally in organisms from all three domains of life, but this capability in cyanobacteria attracted particular attention because they can produce the fatty acid precursors photosynthetically and could, therefore, potentially support a bioprocess from CO2 and sunlight to renewable hydrocarbon fuels (126). A two-enzyme hydrocarbon production pathway, comprising an acyl-acyl carrier protein reductase and ADO, was identified through a comparative genomics analysis of a group of productive cyanobacterial species and one species found to lack a functional pathway (126). The enzymes were shown to produce linear saturated and monounsaturated hydrocarbons of chain length 13, 15, and 17 from the corresponding saturated/monounsaturated C14, C16, and C18 substrates (126). The second enzyme, ADO, is a ferritin-like diiron protein that catalyzes a reaction – conversion of a Cn linear fatty aldehyde to the corresponding Cn-1 hydrocarbon and formate (127) – that appears at first glance to be hydrolytic but actually requires dioxygen and a reducing system (126), in conformity with other FDCOO systems. The ADO reaction is, in fact, an oxygenation reaction: one O-atom from the O2 co-substrate is incorporated into the formate product (128, 129). The absence of any precedent for such a "cryptically redox" outcome contributed to some initial controversy over the nature of the reaction. The biotechnology company LS9 proposed in their patent application (130) and initial literature report (126) that the C1-derived co-product is carbon monoxide and designated the enzyme aldehyde decarbonylase, whereas an ensuing patent application from the company Joule Unlimited (131) claimed that C1 is converted to CO2 and called it alkanal decarboxylative monooxygenase. Later, after the Penn State group had established that formate is

26 the co-product and O2 a co-substrate (127, 128), two separate reports claimed O2-independent aldehyde cleavage by ADO (132, 133), a claim that was later debunked (129). Once firmly established, the unusual outcome forecast a mechanism very different from those of the aforementioned FDCOOs that effect overt oxidations.

Stopped-flow absorption and freeze-quench Mössbauer spectroscopic studies

II/II demonstrated that, analogously to the initial steps of the aforementioned FDCOOs, the Fe2 -

ADO•aldehyde complex reacts with dioxygen to generate a transient peroxide-level intermediate

(86). However, its spectroscopic features are distinct from those of the other complexes (Table 1-

1), consistent with a different structure and reactivity. The complex has an absorption band at 450 nm (86), even more blue-shifted than the bands of the peroxide complexes in the N-oxygenases.

Its Mössbauer spectrum consists of two partially resolved quadrupole doublets with parameters diagnostic of high-spin FeIII ions (86). The difference in the Mössbauer parameters suggests that the two FeIII sites have different coordination geometries, despite possessing similar residue ligand spheres. This site differentiation, along with the observations that accumulation of the intermediate requires the presence of the aldehyde substrate and cognate linear primary alcohols do not functionally substitute (86), led to the proposal that the intermediate is an asymmetric

III/III Fe2 -peroxyhemiacetal (diferric-PHA) complex resulting from coupling of a partially reduced dioxygen unit with the substrate carbonyl (Figure 1-8), analogous to the mechanisms of the aforementioned flavin and cytochrome P450 systems that process aldehyde substrates.

Formation of the putative diferric-PHA intermediate is slow relative to the O2-addition steps in other FDCOOs (k ~ 1 mM-1s-1 at 5 °C) (86), raising the possibility of a pathway distinct

II/II from the usual concerted/coupled two-electron oxidative addition of O2 to the Fe2 cluster in a bridging mode. Indeed, if certain assumptions are invoked, results from the reactions with alcohol analogues suggest that formation of a bridged peroxide species might be precluded in ADO.

Under the assumptions that (1) alcohols bind in the same manner as aldehydes and properly

27 condition the cofactor for O2 addition, as is suggested by x-ray crystal structures (134, 135), and

(2) the 450-nm-absorbing species that forms in the presence of aldehydes is, as proposed, a diferric-PHA complex, the alcohol analogue should promote formation of the precursor to the diferric-PHA complex, which would be blocked from progressing forward by the absence of the electrophilic carbonyl. According to the published mechanism, this precursor is a bridged peroxo-

III/III Fe2 complex, which would be expected to absorb in the visible regime. The failure of the alcohols to trigger development of any transient chromophore during the slow formation of the µ-

III/III oxo-Fe2 product state (86) could be taken as evidence that a bridged peroxide complex cannot readily form. However, the multiple assumptions invoked in this analysis require experimental validation before firm conclusions can be drawn regarding the pathway for formation of the intermediate.

The redox-neutral cleavage of the aldehyde coupled to complete reduction of O2 necessitates that a total of four electrons be provided by the obligatory reducing system (126).

Two electrons are provided in the usual manner in the reductive half-cycle that converts the µ-

III/III II/II oxo-Fe2 product state of the cofactor back to the O2 reactive Fe2 form. Uniquely in ADO, two additional electrons must also be delivered during cleavage of the substrate, in what corresponds to the O2-dependent oxidative half-cycle of most FDCOOs, to accomplish the redox- neutral conversion. Correlating with this unique pathway and outcome, the diferric-PHA complex is quite stable (t1/2 ~ 7 min at 5 °C) in the absence of a reducing system (86), but reacts rapidly with a number of chemical and biological reductants to trigger formation of the products (86, 128,

129, 136). Diferric-PHA reduction by one cyanobacterial ferredoxin, PetF, was found to be

-1 particularly rapid (kobs > 500 s ) and to give a yield of the two products approaching the theoretical value of 0.5 per reduced ferredoxin (136), whereas reduction by chemical donors

(phenazines) or heterologous ferredoxins was found to be less rapid and to result in diminished product yields (86, 127, 128).

28 Although reduction of the diferric/peroxide-level intermediate could theoretically yield

III/IV an Fe2 intermediate similar to X in RNR-β, no evidence for accumulation of such a species was found (86, 136). This situation was rationalized according to a mechanism (Figure 1-8) in which transfer of a single electron reductively cleaves the O–O bond of the diferric-PHA complex to yield a gem-diolyl radical, which rapidly undergoes β-scission of the C1–C2 bond to generate formate and a radical residing on the formerly C2 carbon. Acquisition of H• from the protein by the radical would complete formation of the second product but leave the enzyme oxidized by one electron. The remaining oxidizing equivalent would need to be quenched by the reducing system. Because this oxidized form of the enzyme apparently does not accumulate sufficiently for direct detection, the proximal source of the H• is still somewhat in doubt (vide infra).

Support for a radical-type C1-C2 cleavage step was provided by experiments employing

"radical-clock" substrate analogs with α-oxiranyl or cyclopropyl groups (137, 138). These substrates were anticipated to undergo ring opening upon formation of an exocyclic carbon- centered radical. Observation of ring-opened products and covalent modification of the enzyme were consistent with the predictions of the radical mechanism. More direct evidence for a C2- derived radical was then obtained with authentic aldehyde substrates by detection of an off- pathway peroxyl radical species (ROO•), produced upon trapping of the primary product radical by a second equivalent of O2 (136).

Several precedents suggested that ADO might have an H• donating amino acid to quench its C2 radical intermediate. The most likely donor would be a Tyr or Cys residue. Indeed, a Cys- derived radical was detected in the reaction under certain conditions (high [O2] and with short- chain substrates) (136). However, its accumulation only as a successor to the off-pathway peroxyl radical and its far distance from the active site ruled out the possibility that the Cys radical might form by quenching of the C2 radical in the productive pathway. Moreover, the susceptibility of the C2 radical to O2 trapping implies that it has a significant lifetime. Thus, an H• donor, if

29 present, must be inefficient. Substitution of other candidates failed to yield the anticipated phenotype of enhanced susceptibility to trapping of the radical by O2 and accumulation of more peroxyl radical (136), suggesting that the cofactor itself might provide the H•. Results of solvent deuterium inventory experiments were also interpreted to favor this possibility (139). The susceptibility of the immediate precursor to the hydrocarbon product to trapping by O2 provides a rationale for previously unexplained observations of diminished yields in bioreactors run at high

O2 flux (140). This modest proficiency in handling its radical intermediates might also present a target for improvement of ADO for biotechnological purposes.

Figure 1-8. Free-radical mechanism proposed for ADO catalysis, invoking reduction of the ADO diferric-PHA intermediate to yield the alkane product.

1.3.3 Use of transition metals other than iron by ferritin-like proteins.

In biology, MnII can also occupy histidine- and carboxylate-rich sites such as those harboring iron in the aforementioned FDCOOs. Indeed, an exciting set of discoveries of the last

30 decade revealed that manganese is actually the functional metal in some ferritin-like proteins previously thought to use iron. These studies have shown how fundamental differences between the metals have been leveraged to cope with challenges faced by organisms in different environmental niches.

Oxidation of MnII to MnIII is less favorable than the corresponding oxidation of iron

(141). In the absence of strongly donating or redox-active ligands, MnII is therefore incompetent for the first uphill reduction step in O2 activation. However, it also confers greater

"bioavailability" for aerobic life by disfavoring adventitious oxidation of MnII to insoluble oxides, as readily occurs for FeII. Conversely, high-valent states (e.g., MIV) are generally more stable for

Mn than for Fe in the relevant coordination environments (141). This property makes Mn ideal for multi-electron processes, such as the photochemically driven four-electron oxidation of water to O2 by photosystem II (142-144). It is also exploited in certain ferritin-like proteins, specifically class I RNR-βs (vide infra).

The possibility for metal substitution to diversify the biochemical repertoire of the ferritin-like scaffold was forecast decades ago by studies on the dimanganese catalases, or

“pseudocatalases,” which disproportionate hydrogen peroxide to O2 and water (145). In this

II/II system, reaction of the Mn2 cluster with a partially reduced oxygen species, H2O2, obviates the

II III/III very unfavorable first reduction of O2 by Mn . The Mn2 form, produced as H2O2 is reduced to water, subsequently oxidizes H2O2 to O2 (145), essentially the reverse of the first step in O2

III/III II/II activation by the FDCOOs. Here, the potential of the Fe2  Fe2 reduction would probably be insufficient to permit oxidation of H2O2, providing a chemical imperative for the use of manganese. The lessons from this system informed subsequent discovery and dissection of the manganese-dependent class I-b and I-c RNRs.

Figure 1-9. Active cofactors and mechanistic schemes for the O2 activation pathways of class I RNRs.

1.3.4.1 Heterodinuclear manganese/iron cofactors.

III/III As discussed above, a class I-a RNR-β harbors a µ-oxo-Fe2 cluster and adjacent tyrosyl radical, which serves as the essential initiator of nucleotide reduction (Figure 1-9). In the early 2000s, McClarty and co-workers showed that the RNR-β from Chlamydia trachomatis (Ct) has a phenylalanine residue at the position aligning with the usual radical tyrosine (146). Genetic data implied that the protein is, nevertheless, active in vivo, and it was shown to support modest levels of nucleotide reduction with its partner α subunit in vitro (146). In a subsequent study, several other examples of apparent RNR-βs with this Y  F replacement were identified (147), and the number of such sequences has since grown. It was initially proposed that the unusual

III/IV RNR-βs could use an Fe2 cluster, analogous to the tyrosyl-radical-generating intermediate X in activation of Ec RNR-β (but stabilized in some way by complexation of the α and β subunits), directly for radical initiation (147-149). This novel modus operandi would obviate the tyrosyl- radical intermediary. Subsequent, we established that the functional cofactor is actually a stable

32 MnIV/FeIII cluster, in which the high-valent manganese functionally replaces the tyrosyl radical as the key initiator (110, 150, 151). This work also showed that the diiron form of the protein lacks measurable activity (110) and that the meager activity observed in the original studies (148, 149) had probably resulted from the presence of manganese in preparations thought to contain only iron. The fact that the active MnIV/FeIII cofactor does not exhibit an X-band EPR signal due to its integer-spin (S = 1) ground state (150) undoubtedly impede the identification of the active cofactor.

The MnIV/FeIII cofactor of Ct RNR-β assembles by reaction of the MnII/FeII-β complex

IV IV with O2 (Figure 1-9). An unprecedented Mn /Fe accumulates nearly to stoichiometric levels under appropriate reaction conditions (152). Spectroscopic and computational analysis suggested

IV IV IV that it has a di-(µ-oxo)-Mn /Fe core with a terminal hydroxide ligand on the Mn ion (153). As in the Ec RNR-β reaction, the electron delivered to the buried cluster as the intermediate decays to the acting resting state is relayed by transient oxidation of a near-surface amino acid sidechain, in this case a tyrosine (222) that has no role in the nucleotide reduction reaction (154). The Trp51 residue corresponding to the electron shuttle in the Ec class I-a RNR-β (Trp48) is also required for proper function of this relay system in Ct RNR-β (154). Reduction of the FeIV ion of the intermediate to the FeIII state of the stable cofactor is most likely accompanied by protonation of one of the oxo bridges (153, 155).

An x-ray crystallographic study by Högbom and co-workers concluded that the MnIV ion of the cofactor resides in site 1 (the site closest to the location of the tyrosyl radical in the class I-a

β proteins) (156), as speculated in the first report of the cofactor. A subsequent careful correlation of metal occupancies with catalytic activity suggested that both 1MnIV/2FeIII and 1FeIII/2MnIV

(where the superscript denotes the site location of the metal) can assemble but the former complex is markedly more active in catalysis and may be the only active form (157). A follow-up study showed that, in the absence of other assembly factors, the RNR-β protein itself has the

33 capacity to direct the proper metal site occupancy under appropriate conditions (anoxic equilibrium with equal concentrations of the two divalent metals at the proper metal:protein stoichiometry) (158) and posited that this capability may derive primarily from the selectivity of site 2 for FeII over MnII.

II/II II/II The reactivities of the homodinuclear (Fe2 and Mn2 ) complexes of Ct RNR-β provided stark illustrations of the general inorganic chemical principles underlying evolutionary

II/II selection of the functional metal(s) in different class I RNR systems. The Fe2 -β complex was shown to react with O2 (154), via an intermediate with the spectral hallmarks of the canonical µ-

III/III II II 1,2-peroxo-Fe2 complex (159), ~10-fold more rapidly than does the Mn /Fe complex, which proceeds without obvious accumulation of a peroxide-level complex under the conditions examined (152). These observations are consistent with the notion of concerted or coupled two- electron reduction of O2, which is likely to be possible only with both sites occupied by the more

II II/II easily oxidized Fe . Conversely, the Mn2 -β complex is completely inert toward O2 (110). At

II least one of the two sites must be occupied by Fe for a reaction with O2 to occur, presumably

III –• because only then is partial reduction to a more reactive complex (e.g., Fe -O2 ) sufficiently favorable to engage the less readily oxidized MnII ion. Following O–O-bond cleavage, however, the presence of Mn then confers greater stability to the high-valent states. For example, whereas the active relay of the extra electron to the MnIV/FeIV intermediate by the Trp51-Tyr222 dyad was found to proceed without accumulation of a Tyr or Trp radical state, both radicals were seen to

II/II accumulate upon decay of the peroxide intermediate in the reaction of the Fe2 complex (154).

Analogously, whereas the MnIV/FeIII form of the protein is stable for hours and during thousands

III/IV of enzymatic turnovers, the inactive Fe2 complex decays in seconds to minutes (110).

Biologically, selection of one Fe and one Mn ion appears precisely to balance the capacity for O2 activation with the potential for stabilization of the resultant oxidized metal complex. These

34 lessons from the class I-c Ct RNR system were instructive in the elucidation of the Mn2- dependent class I-b RNRs by Stubbe and co-workers just a few years later (see below).

The MnIV/FeIV complex in Ct RNR-β is, effectively, the heterodinuclear homolog of the

IV/IV Fe2 complex, Q, in sMMOH. The high homolytic bond dissociation enthalpy (BDE) of methane (105 kcal/mol) (160) makes its oxidation by Q arguably the most difficult transformation mediated by a member of the FDCOO enzyme class. In light of the facts (1) that the MnIV/FeIV intermediate forms without accumulating precursors and is then remarkably stable and (2) that no

IV/IV Fe2 complex has been found to accumulate in the reaction of any FDCOO other than sMMO, it seems apparent that the heterodinuclear high-valent complex is, as expected, more stable than its diiron counterpart. It is therefore interesting to consider the extent to which this relative stability might limit the functional capabilities of the MnIV/FeIV unit for difficult transformations involving cleavage of strong C-H bonds. ii. R2-like ligand-binding oxidase

To date, there is only one additional ferritin-like protein claimed to have a functional

Mn/Fe redox cofactor. The "R2-like ligand-binding oxidase" (R2lox) protein of Mycobacterium tuberculosis (Mt) is highly expressed and has been implicated as a virulence factor (161). It is structurally very similar to RNR-β subunits, including conservation of tyrosine at the position

(175) aligning with the radical-harboring residues of the class I-a and I-b RNR-β proteins (161).

However, R2lox lacks a subunit-interfacial tyrosine that is conserved in all known, active RNR-

βs, wherein it mediates inter-subunit radical translocation during turnover (161). Accordingly, it was found not to support the RNR activity of Mt RNR-α (161). As isolated following over-

III III expression in E. coli, R2lox was found to harbor a Mn /Fe cluster, with Mn is site 1, as in Ct

RNR-β (161-163). The metal ions are bridged by three ligands – a hydroxide and two carboxylates. One of the carboxylates is provided by a protein glutamate residue. The other comes from a bound, exogenous fatty acid. The binding mode of the exogenous ligand led to the

35 suggestion that the protein could be a fatty acid hydroxylase. Indeed, hydroxylated fatty acids were isolated from denatured Geobacillus kaustophilus R2lox purified after heterologous expression in E. coli (163). Although this observation provides circumstantial evidence, the assignment of R2lox as a fatty acid oxygenase remains to be definitively established.

The R2lox protein can direct assembly of the reduced 1MnII/2FeII form of its cofactor from the apo protein and divalent metal ions with remarkable selectivity, analogously to the early steps in assembly of the Ct RNR-β cofactor (163, 164). The reduced heterodinuclear cofactor can then react with O2, resulting in a novel protein crosslink of the phenolic oxygen of Tyr162 to the

C3 of valine 71 (161). This two-electron oxidation reaction necessitates cleavage both a strong tertiary aliphatic C-H bond (BDE ~ 95 kcal/mol (160)) and a phenolic O-H bond (BDE ~ 87 kcal/mol (160)). It is reasonable to anticipate that the reaction may proceed via a MnIV/FeIV complex analogous to the activation intermediate in Ct RNR-β, which would be capable of aliphatic C–H bond activation, but would be the first example of this chemistry for a heterodinuclear cofactor. A mechanism (Figure 1-10) was proposed on the basis of computation analysis (162), involving (1) abstraction of H• from the tyrosine by the MnIV/FeIV intermediate,

(2) abstraction of H• from the valine residue by the resultant Tyr•, (3) electron transfer from the

Val C3 radical to the MnIV/FeIII complex to yield a C3 carbocation and MnIII/FeIII cluster, and (4) polar coupling of the Tyr phenol(ate) with the Val C3 carbocation. However, based on the BDE for the target positions, abstraction of the aliphatic valine H• by the Tyr• would be thermodynamically unfavorable. Alternative mechanisms (Figure 1-10) could proceed by H• atom abstraction from the valine by the MnIV/FeIV in the first step, followed by either (1) electron transfer to the metal center, generating the valine carbocation that could undergo polar chemistry with the deprotonated Tyr, or (2) H• abstraction from the Tyr residue by the resultant MnIV/FeIII complex followed by radical coupling of the two residue radical species to yield the crosslink

II/II species. The Fe2 complex of R2lox, which can form upon incubation of the apo protein with

36 excess FeII, appears also to be active for cross-link formation (163). Thus, a chemical rationale for

Nature's selection of the heterodinuclear cofactor is currently lacking and awaits more definitive insight into the protein's biochemical function.

Figure 1-10. Potential mechanisms for formation of the Tyr-Val crosslink observed in the R2lox crystal structure employing a heterdinuclear Mn/Fe cofactor.

1.3.4.2 A dimanganese cofactor in class I-b RNRs.

In 2010, the Stubbe group resolved a decades-old controversy concerning the nature of the functional cofactor in class I-b RNRs, and the work remains a fascinating example of Nature's expansion of the biological utility of a metalloprotein scaffold by metal substitution. The I-b

RNR-βs are structurally similar to the I-a and I-c proteins, possessing the same ferritin-like architecture, six metal-ligand residues, and pair of redox-active tyrosines (active site and subunit interfacial) (165). Multiple publications over the preceding four decades by Auling, Follman and their co-workers had provided evidence for the involvement of manganese in the archetypal member of the subclass from Corynebacterium (formerly Brevibacterium) ammoniagenes (Ca)

(166). These papers did not provide a clear explanation of how the enzyme might function with manganese, because they detected the reduced (+II) metal in preparations of Ca RNR-β and did

37 not identify an oxidant that might be capable of serving as radical initiator. Moreover, Sjöberg,

Nordlund and co-workers had shown clearly that the protein can, in vitro, assemble a µ-oxo-

III/III Fe2 /Tyr• cofactor equivalent to that found in the Ec class I-a β protein and support nucleotide reduction in that state (albeit at rates ~ 10-fold less than that of the Ec class I-a system) (167).

Stubbe and co-workers synthesized these observations, insights from the dimanganese catalase and Mn/Fe-dependent Ct RNR-β, general principles emerging from a growing literature on the microbial physiology of iron and manganese uptake and utilization (particularly in pathogens), and genomic data from bacteria harboring class I-b RNRs to rationalize the involvement of manganese.

A number of opportunistic pathogens, including E. coli, possess both I-a and I-b RNRs.

In E. coli, the genes encoding the I-b subunits are transcribed only under conditions of iron limitation and oxidative stress, suggesting that β might not use iron in vivo (168, 169). Indeed, other bacteria, including Ca, Bacillus cereus, Bacillus subtilis, Bacillus anthracis,

Mycobacterium tuberculosis, Staphylococcus aureus, and Streptococcus pyogenes, harbor only a

I-b RNR (170-172), and several of these species require manganese for growth (173, 174). The biological data thus strongly implicated manganese as the active metal. The question became how the crucial oxidant needed to initiate turnover could be introduced, given the incompetence of

II Mn for the initial uphill step in O2 activation. The crucial clue came from genomic context. In addition to differences in primary structure, the class I-b RNRs are distinguishable from their I-a and I-c counterparts by the presence of a gene coding for a flavoprotein, NrdI, in the RNR operon

(175-177). Prior work had implicated this protein in provision of the balancing electron during activation of the Ec I-a RNR-β and had shown the S. pyogenes orthologue to be essential for the activity of the class I-b RNR (177). The Stubbe group showed that the role of NrdI is to use its

II/II flavoprotein cofactor to reduce O2 to a form that can react with the Mn2 cluster to introduce the

III/III Tyr• and adjacent Mn2 cluster (176). Later studies by the Stubbe and Sjöberg groups

38 established that the reactive O2-derived species is superoxide (175) and that other class I-b RNR

III/III systems follow the same modus operandi (178, 179). The Mn2 -Tyr• forms of two class I-b

RNR-β orthologues were even isolated from their native hosts (178, 179). A structure of the stable complex between the Ec class I-b RNR-β and NrdI proteins solved by the Rosenzweig and

–• Stubbe groups suggested that a channel could electrostatically conduct O2 produced at the NrdI flavin to the dimanganese cluster in RNR-β (180). Mechanistic analysis later implicated a

III/IV Mn2 complex in Tyr• production (Figure 1-9) (175). Spectroscopic studies showed that,

III/III III unlike the Fe2 clusters formed in FDCOOs, the Mn ions in the product cluster of Ca RNR-β are ferromagnetically coupled to give an STotal = 4 electron-spin ground state (178), rationalizing the markedly different lineshape and temperature dependence of the EPR signal of the Tyr•

III/III compared to those formed adjacent to antiferromagnetically coupled Fe2 clusters.

The increased reduction potential of MnIII relative to FeIII explains both the need for prior

II reduction of O2 by NrdI to initiate Tyr• production and the observation of Mn in the prior biochemical studies on the Ca RNR by Auling, Follman and co-workers. It is apparent that the

III/III III/III Mn2 cluster is less stable toward reduction than the µ-oxo-Fe2 clusters in the class I-a systems. This greater reactivity of the product cluster could be both a virtue and a vice: although

III/III the Mn2 -Tyr• systems may be less robust toward adventitious reduction, they should also be more readily reactivated by in situ cluster reduction and subsequent reaction with NrdI-furnished

–• O2 . However, the clearest rationale for the utilization of manganese instead of iron in the I-b systems relates to the relative bioavailability of the metals. Under conditions of oxidative stress, pathways for iron uptake as well as assembly and stability of iron enzyme cofactors are compromised (181). Moreover, the immune systems of host organisms for pathogenic microbes have evolved measures to deprive the invaders of iron, limiting their virulence (182-184).

Apparently, use of manganese instead of iron represents a countermeasure by the pathogens to maintain essential pathways (173, 185), including nucleotide reduction. The class I RNR story

39 recalls earlier work on Ec superoxide dismutases (SODs). During growth under either hyperbaric

–• O2 or air in the presence of redox cyclers, elevated levels of O2 result in expression of an alternative SOD that uses Mn as a cofactor instead of the Fe used by the constitutively expressed enzyme (186, 187). It has been shown that the Fe-dependent enzyme is still produced during superoxide stress but cannot be properly activated by iron insertion (186, 187). Thus, by multiple mechanisms and in multiple enzyme systems, the challenge of low iron bioavailability has imposed selective pressure for use of manganese in its place. For the case of the class I-b RNRs, the use of the O2-inert metal necessitated the involvement of an activator protein (NrdI) to furnish a more reactive oxidant to push the metal to the higher-valent states required for cofactor assembly. Whether other class I RNR systems might exist that use Mn without such an activator remains to be explored.

1.4 O2-activating diiron cofactors within non-ferrtin-like protein architectures

The ferritin-like fold appears to be Nature's most privileged scaffold for O2-activating dimetal clusters, but discoveries of the last decade have revealed several examples of unusual adaptations to other structural domains that enable them to accommodate diiron cofactors with unique properties. In several cases, although the protein architectures differ, the mechanistic logic and reaction types mediated appear to largely conform to the FDCOO paradigm. However, in one notable case, a completely different pathway both enables unusual outcomes and imposes limitations on the range of transformations that can be effected.

40 1.4.1 Integral membrane diiron oxidases/oxygenases.

Many of the soluble, cytosolic FDCOOs have membrane-bound counterparts that catalyze analogous transformations. These integral-membrane enzymes lack significant sequence similarity to one another and to soluble FDCOOs. However, they have in common a set of eight conserved histidine residues, arranged in three “histidine box” (two HxxHH and one HxxxxH) motifs (188). Mutagenesis studies of various members of this family have shown that these histidines are essential for activity (189-191). Evidence that the enzymes use diiron cofactors was first shown for alkene ω-hydroxylase (AlkB) by Mössbauer-spectroscopic experiments (188). The reduced protein exhibited a quadrupole doublet spectrum with parameters characteristic of FeII ions in O/N coordination environments, and reaction with O2 in the presence of substrate gave a

III/III different spectrum indicative of an antiferromagnetically coupled Fe2 product state (188). The

II/II diminished value of the isomer shift for the Fe2 form (δ ~ 1.1 mm/s compared to typical values of 1.2-1.3 mm/s for FDCOOs) was interpreted as evidence for a more N-rich coordination sphere, consistent with the hypothesis of a ligand set composed primarily or exclusively of the conserved histidines. Thus, it was proposed in this initial study that the presence of the histidine box sequence motifs might be sufficient to identify other members of this family of integral- membrane diiron enzymes (188).

The past two years have seen a breakthrough in the study of these integral-membrane enzymes in the form of x-ray crystal structures of two mammalian (human and mouse) stearoyl-

CoA desaturases in their substrate complexes and a yeast fatty acid α-hydroxylase (192-194). All three structures were solved with ZnII bound at the cofactor site, limiting insight into details of cofactor geometry. The enzymes have a conserved “mushroom-like” topology composed primarily of α-helices. A domain of four nearly parallel transmembrane helices, representing the stem of the mushroom, anchors the protein in the membrane and connects to an entirely cytosolic

41 helical domain (representing the cap) that harbors the cofactor site (Figure 1-11). As anticipated, the dimetal site is composed entirely of histidine residues (Figure 1-11). The electron density

2- – maps did not show evidence for any bridging solvent-equivalent (O , HO , or H2O) ligand, consistent with the presence of divalent zinc (192-194). In the stearoyl-CoA desaturase, one metal site was seen to have five histidine ligands whereas the other had four, with a water molecule completing the first sphere (192, 193). By contrast, each site of the fatty acid α-hydroxylase cofactor had five histidine ligands arranged in square-pyramidal geometry (194). The structures thus identified additional His ligands beyond those predicted from the conserved eight-histidine motif. Intriguingly, the metal-metal distances observed in all structures (> 6 Å) are much greater than those in the soluble diiron enzymes. The presence of ZnII might be partly responsible for this difference. If so, it would call into question the relevance to the active diiron state of structural details in the vicinity of the Zn2 center. It should be noted, however, that the target carbons of the bound substrate (C9 and C10) are positioned between the ZnII ions in a manner that would seem appropriate for desaturation between them. This fact suggests that the use of computational models derived from these structures as starting points for interrogation of reaction coordinates is reasonable.

In addition to fatty acid desaturases and hydroxylases, cyanobacterial ADO also has integral-membrane cognates in plants (195, 196). On the basis of claims that they produce CO as co-product (197-199), the plant enzymes have been designated (as the soluble cyanobacterial enzymes initially and erroneously were) aldehyde decarbonylases (ADs) (195, 196). ADs possess the conserved histidine box motifs and substitution of those His residues leads to loss of activity

43 (195), suggesting the possibility of a diiron cofactor. There is no obvious pathway by which an

ADO-like reaction employing a diiron cofactor could yield CO. Rather, it seems more likely that the integral membrane enzymes also produce formate.

More recently, integral-membrane enzymes that produce hydrocarbons with terminal olefin and acetylene functionality have been discovered in bacterial species (200-202). Their sequences suggest that these enzymes have membrane topologies similar to those of the fatty acid desaturases and hydroxylases, in addition to the conserved histidine box motifs for coordination of a dimetal (presumably, diiron) cofactor. UndB from Pseudomonas was shown to catalyze oxidative decarboxylation of Cn (n = 10, 12, 14) fatty acids to produce Cn-1 1-alkenes and CO2

(200). JamB from a marine cyanobacterium and TtuB from a γ-proteobacterium are reportedly bifunctional enzymes that first oxidatively decarboxylate a Cn fatty acid substrate to the Cn-1 terminal olefin and then further desaturate the olefin to the acetylenic product (201, 202). Soluble bifunctional desaturase/acetylenases that install internal olefin and acetylene groups are known

(203), yet examples of soluble enzymes that install terminal alkyne functionalities remain to be discovered.

The growing number of recognized O2-activating, integral-membrane diiron proteins has important implications for human health and the environment. In higher eukaryotes, fatty acid desaturases have been implicated in obesity, cancer, diabetes and other metabolic diseases (204-

212). Mutations in the fatty acid 2-hydroxylase result in serious disorders of the central nervous system (213-216). These enzymes may thus be targets for drug therapies. In the realm of biotechnology, production of alkanes and alkenes by enzymes such as AD and UndB could underpin bioprocesses to renewable, fungible fuels (217). The recent advances in isolation of these integral membrane diiron enzymes (218, 219) will undoubtedly facilitate deeper investigation into their structures and reaction mechanisms, and the resultant more thorough

44 understanding would undoubtedly enable efforts to target them therapeutically and biotechnologically.

1.4.2 Deoxyhypusine hydroxylase.

The eukaryotic translation initiation factor 5A (eIF5A) is the only protein known to contain the amino acid hypusine (87, 220). A post-translational modification of a specific ribosomally incorporated Lys residue installs hypusine into eIF5A. Deoxyhypusine synthase, an

NAD-dependent alkyl transferase, appends a 4-aminobutyl group derived from spermidine onto the ε-amine of the Lys residue to produce deoxyhypusine (Figure 1-12). Deoxyhypusine hydroxylase (DOHH), an O2-activating nonheme diiron enzyme, then hydroxylates carbon 2 of the appendage with R stereochemistry to complete the modification (Figure 1-12) (220), which is essential for the activity of the eIF5A protein and thus the viability of eukaryotic cells. Sequence and structural analyses revealed that DOHH has a HEAT-repeat, α-hairpin architecture (Figure 1-

12) that positions four histidine and two glutamate residues to coordinate a diiron cofactor (220,

III/III 221). In the published structure, which is of a remarkably stable Fe2 -peroxide complex (see below), similar N- and C-terminal domains each provide one Glu and two His ligands to the diiron site; no carboxylate bridges are present in this state of the cofactor (Figure 1-12) (221).

The unit thus resembles a pair of juxtaposed mononuclear sites having the now-classic

"His2carboxylate facial triad" coordination geometry (222-224), with the distal His of each site being contributed by the opposite domain. Viewed in this light, the structural architecture nicely underscores the general principle of concerted or coupled two-electron reduction of O2 to peroxide repeatedly highlighted above.

The most astounding feature of DOHH is the unprecedented stability of the peroxide

II/II intermediate produced upon addition of O2 to the Fe2 cofactor. In fact, it is so stable that the

45 recombinant DOHH is purified from overproducing E. coli cells in this state (87). The complex has a distinctive blue color and absorption bands at 330 and 650 nm (87), similar to those of μ-

III/III 1,2-peroxo-Fe2 intermediates discussed above (Table 1-1). Mössbauer spectroscopy was used to confirm the +III oxidation state of the iron ions, and analysis by resonance Raman spectroscopy confirmed the presence of the peroxide ligand (87). The DOHH peroxo-diferric intermediate is stable for several days at room temperature, but its decay is accelerated by incubation with its substrate, the eIF5A protein with deoxyhypusine at Lys50 (87). The resultant hydroxylation reaction produces the mature eIF5A protein. Remarkably, even the substrate- triggered decay of the complex proceeds on the timescale of hours under the in vitro conditions examined (87).

The remarkable stability of the DOHH peroxide complex enabled its crystallization and structural characterization. The geometric structure of the peroxide moiety proposed on the basis of spectroscopic data (87) was confirmed by the crystal structure (221). The core was modeled as a μ-1,2-peroxide complex with an additional µ-(hydr)oxo ligand (Figure 1-12) (221). The complex thus has a gross geometry essentially equivalent to that proposed for the peroxide complexes in sMMOH, RNR and fatty acid desaturases, begging the question of why it is so stable. The coordination sphere of the DOHH cofactor, which has two more histidines and two fewer carboxylates than those of the FDCOOs, is less negatively charged, perhaps stabilizing against a redox-neutral O–O-bond cleavage to a Q-like complex. Additionally, the structure reveals that the peroxide moiety is protected from solvent in a hydrophobic pocket, which is expected to preclude its protonation to yield a more electrophilic complex (221).

Figure 1-12. (Left) Bottom-up view of the DOHH (PDB accession code 4D50) HEAT-repeat structural fold with almost identical N-terminal (blue) and C-terminal (pink) domains and the diiron active site coordinated at the interface of the domains. (Right, top) Reaction pathway to instal the hypusine amino acid post-translational modification of the eIF5A protein lysine residue. (Right, bottom) Zoom in view of the diiron active site displaying the bound peroxide moiety (red).

It is extremely unlikely that the peroxide-level complex itself would be competent to initiate the aliphatic C–H-bond cleavage required for the observed hydroxylation outcome. How might the intermediate convert to a more reactive state? EXAFS experiments in the absence and presence of substrate suggested that both iron sites undergo conversion from 6- to 5-coordinate upon substrate binding (225). With the protein architecture holding the glutamate ligands already in terminal coordination mode, one well-precedented type of ligand dissociation, a carboxylate shift, is precluded. Instead, it was attributed to loss of a water ligand and shift of the μ-hydroxo ligand to become terminal mode. It was further suggested that the peroxide might then adopt a new geometry (e.g., μ-η1,η2) that would activate the O-O bond for homolytic cleavage to generate a Q-like oxidant capable of abstracting H• from C2 of the substrate. The reaction would then be

47 completed by HO• transfer to the C2 radical. In this scenario, the coordinative saturation of both

FeIII sites in the µ-peroxo intermediate state in the absence of the substrate would represent an adaptative strategy to mask its reactivity and protect the enzyme from auto-oxidation.

Although neither a high-valent species nor a second peroxide-level species has been detected in DOHH, structural data on the enzyme suggest that it might have some features distinguishing it from the FDCOOs. Analysis by X-ray absorption spectroscopy revealed that the

II/II III/III Fe-Fe distance is nearly constant through the sequence of Fe2 reactant, peroxo-Fe2

III/III intermediate, and Fe2 product states (225), by contrast to the significant changes of the Fe-Fe distance indicated by analogous data on the FDCOOs. The rigidity of the metal center induced by the unique first-coordination sphere and structural architecture might disfavor formation of a di-

IV/IV (μ-oxo)-Fe2 core, which would require contraction along the Fe–Fe vector (58, 226). On this basis, it was suggested that the presumptive H•-abstracting high-valent intermediate in DOHH might adopt an open core, as has been found in certain model complexes (227).

1.4.3 Diiron β-hydroxylases.

Multiple enzyme classes share the αβ/βα sandwich architecture often referred to as the metallo-β-lactamase or metallohydrolase fold (228). Examples include the purple acid phosphatases, metallo-β-lactamases, the glyoxalases II, and the flavodiiron proteins (228-232).

As the designation of the protein fold would imply, most of the well-studied members of this structural superfamily use their mononuclear or dinuclear metal cofactor as a Lewis acid to activate water for a hydrolysis reaction. Purple acid phosphatases use dinuclear clusters with one iron ion and a varied second metal to hydrolyze phosphoester or amide bonds (229, 233, 234); metallo-β-lactamases use mono- or dinuclear zinc centers for hydrolysis of the amide bond of the

48 strained β-lactam ring (228, 230); and the glyoxalases deploy heterodinuclear Zn/Fe cofactors for thioester hydrolysis (228, 230).

In two recently recognized cases, the structural architecture has undergone adaptations to use diiron clusters in reactions with O2. The flavodiiron proteins (FDPs), found in anaerobic organisms, catalyze complete reduction of O2 and in some cases NO as a means of detoxification

(235). Between their diiron and flavin cofactors, these proteins can harbor four reducing equivalents and could thereby efficiently reductively disarm any of the three stable oxidation

–• states of the dioxygen unit (O2, O2 , and H2O2). FDPs have variable domain structures but most typically have a flavodoxin-like N-terminal domain and αβ/βα C-terminal domain that harbors the diiron cofactor (230, 231, 236). Each iron of the cluster is coordinated by two histidines and one carboxylate (Glu/Asp) residue. Recent reviews by Gomes and Kurtz effectively summarize the important evolutionary adaptations distinguishing the FDPs from the hydrolases (230) and the functions and mechanisms of the FDPs (235).

By contrast to the non-productive, reductive decomposition of O2 by the FDPs, a diiron enzyme with a metallo-β-lactamase structural fold, CmlA, was recently shown to catalyze β- hydroxylation of L-4-aminophenylalanine tethered to the non-ribosomal peptide synthase

(NRPS), CmlP (4APA-S-CmlP), on the biosynthetic pathway to chloramphenicol (Figure 1-5)

(237). DNA sequence analyses imply that CmlA is representative of a larger family of

(presumptively diiron) NRPS tailoring β-hydroxylases (237). X-ray crystal structures of CmlA in the reduced and oxidized states revealed practically identical geometries of the dimetal cofactor

(238, 239), confirmed by Mössbauer and EPR spectroscopic studies to be diiron (237). The cofactor is asymmetrically coordinated, with Fe1 held by two histidines and a bidentate glutamate and Fe2 by a single histidine ligand, a monodentate aspartate, and either two solvent molecules

(reduced form) or an acetate (oxidized form) (Figure 1-13). An Asp carboxylate and a hydroxide or oxide ligand bridge the two Fe ions, the former in an unusual μ-1,1 mode. This carboxylate

49 bridging mode, seen previously only in the dinuclear cofactors of two hydrolases (purple acid phosphatase and glyoxylase II (229, 240)), contrasts with the usual µ-1,3 and μ-η1:η2 bridging modes uniformly observed in well-studied FDCOOs and other dimetal proteins (4, 16). A second notable feature of the CmlA cofactor revealed by the structure of the reduced state is the coordinative saturation of Fe1, which would appear to preclude addition of O2 in the expected

II/II bridging mode. However, X-ray absorption spectroscopic studies on the Fe2 state revealed a transition upon binding of the 4APA-S-CmlP substrate (241). The data were interpreted in terms of a change at one FeII site from six- to five-coordinate and the loss of one light-atom scatterer, proposed to be the carbon of a bidentate carboxylate, upon substrate binding (241). The authors posited a carboxylate shift of the site 1 Glu from bidentate to monodentate to open a coordination site for O2 binding. This model would rationalize the observation in transient-state kinetic experiments of pronounced substrate triggering in CmlA (241). Although no high-valent intermediate was detected in this study, it was reasonably speculated that the reaction might

IV/IV proceed via an Fe2 complex equivalent to Q utilized in the sMMO reaction. However, given that the β C–H bond cleaved by CmlA is partially activated by the aniline moiety (BDE ~ 85 kcal/mol compared to 105 kcal/mol for methane (160)), other mechanisms are also feasible.

1.4.4 HD-domain mixed-valent diiron oxygenases.

Myo-inositol (1,2,3,5/4,6-cyclohexan-hexa-ol) oxygenase (MIOX) catalyzes oxidative cleavage of the cyclohexane ring of its substrate, which is a component of cell-signaling second- messenger molecules, to produce D-glucuronate (242-247). Structural and mechanistic studies on

MIOX revealed that it uses an entirely novel cofactor and pathway for this four-electron oxidation

(248). The core of MIOX is a protein scaffold known as the HD-domain, recognized in the late

1990s by Aravind and Koonin as recurring in what was then a relatively small number of

51 phosphohydrolase enzymes (249). The first such enzymes to be characterized were found to use mononuclear divalent metal cofactors coordinated by the histidine and aspartate residues of the conserved HxaHDxbD sequence motif characterizing the domain (249). Studies in the mid-2000s showed that MIOX instead harbors a substrate- and O2-activating diiron cofactor (248, 250-252).

By contrast to all non-heme-diiron enzymes that had been studied prior to that time, MIOX

II/III activates O2 with the mixed-valent Fe2 oxidation state (251). Correlating with its use of the middle redox state of the cofactor, MIOX markedly stabilizes the mixed-valent cluster, supporting accumulation of as much as 70% in equilibrium with the reduced and oxidized states

(252). For comparison, the mixed-valent form of the cofactors in FDCOOs are generally unstable with respect to disproportionation and therefore accumulate to much lesser fractions (253-255).

Perturbations of the electronic and geometric structure of the mixed-valent state upon substrate binding, reflected in the EPR and Mössbauer spectra, were attributed to substrate binding directly to the metal cluster (252). Indeed, the crystal structure of MIOX revealed that the C1 and C6 hydroxyl groups of myo-inositol directly coordinate Fe2 (Figure 1-14), suggested to be the FeIII ion of the valence-localized cluster (250). Analysis by MCD spectroscopy suggested that binding of myo-inositol causes dissociation of a ligand (suggested to be water) from the opposite,

II presumptively Fe site, vacating a position for O2 to add (256).

Figure 1-14. HD-domain structural architecture of MIOX (PDB accession code 2HUO) and the diiron active sites of MIOX (blue) and PhnZ (green, PDB accession code 4MLN) with their respective substrates bound. Mechanistic analysis by transient-state kinetic and spectroscopic experiments revealed functional features of MIOX that further distinguish it from the systems discussed to this point.

II/III III/III Addition of O2 to the Fe2 cofactor was anticipated to generate a superoxo-Fe2 complex, which could theoretically react by multiple possible pathways, including via an H• abstracting high-valent intermediate. However, substitution of the myo-inositol hydrogens with deuteria was shown to promote accumulation of a photolabile intermediate state, termed G, with S = 1/2 ground state and EPR properties distinct from those of the reactant and product complexes (257).

A substrate 2H-KIE of > 8 on decay of this complex was demonstrated, marking it as the intermediate that initiates the reaction by abstracting H• from C1 (Figure 1-15) (257). The spectrum and observed nuclear hyperfine coupling upon incorporation of 57Fe into the cofactor

III/III (257) could be rationalized by a superoxo-Fe2 formulation with terminal superoxide (258).

Importantly, the kinetics of formation of the complex upon mixing with O2 and its decay to the subsequent intermediate, H (which is probably the D-glucuronate product complex), proved that

53 the O2-addition step is reversible, providing strong evidence in favor of an intact O–O-bond in the

H•-abstracting complex (257). Although superoxide-level complexes had previously been implicated as the H•-abstracting intermediates in the catalytic cycles of other iron and copper enzymes (259-261), the studies on MIOX provided the first direct experimental demonstration.

More recently, an H•-abstracting FeIII-superoxide complex was detected in isopenicillin N synthase (IPNS) (262). In the two cases for which C–H cleavage by a mid-valent metal- superoxide complex has been demonstrated (MIOX and IPNS) and the other systems for which it has been proposed, the target substrate C-H bonds are all somewhat activated by the presence of a heteroatom or sp2-hybridized carbon bonded to the target carbon. It has been suggested that such activation is required for H• abstraction by the modestly potent (in comparison to high-valent metal-oxo complexes) superoxide-level oxidants (260, 261). In MIOX, the direct coordination of the C1-hydroxyl group to the FeIII site, presumably leading to its deprotonation, would further activate C1 for H• abstraction. This step would then result in an intermediate state with a C1

III/III radical and hydroperoxo-Fe2 complex (Figure 1-15). The original work suggested that this state would decay via coupling of the C1 radical with either a hydroxyl or hydroperoxyl radical by its attack on either of the two peroxide oxygens (257), but the crystal structure of MIOX strongly disfavored the possibility of hydroperoxyl-radical transfer (250). In the still-viable possibility of hydroxyl-radical transfer from the FeIII–OOH to the C1 radical (Figure 1-15, red), the peroxide bond would be cleaved and an FeIV=O species would result. The ferryl species could, in direct analogy to the key step recently demonstrated in the reaction of 2-hydroxyethyl-1- phosphonate dioxygenase (a mononuclear iron enzyme) (263), abstract H• from the newly- installed C1 hydroxyl, producing a gem-diolyl radical that would undergo radical scission of the

III/III C1–C6 bond. This step would leave a C6 ketyl radical, which would reduce the Fe2 cofactor back to the mixed-valent state. However, emerging precedents from other systems (262) and computational studies (258) suggest an alternative pathway (Figure 1-15, black) involving inner-

54 sphere electron transfer from the substrate ketyl radical to the coordinated FeIII iron to yield a

II/III myo-inosose-1/hydroperoxo-Fe2 intermediate. Subsequent nucleophilic addition of the peroxide moiety to the C1 carbonyl would then initiate cleavage of the C1-C6 bond. This alternative pathway would not involve any high-valent states, thus providing a rationale for the fact that no such species has been detected in the MIOX reaction. The MIOX substrate, myo- inositol, is quite constraining in terms of approaches that might be used to resolve the operant mechanism. The emerging family of HD-domain mixed-valent diiron oxygenases (HD-MVDOs; see below) will undoubtedly provide cases for which the substrate will provide more flexibility and opportunity for deeper insight.

II/III Figure 1-15. Reaction mechanism for dioxygen activation by the MIOX mixed-valent Fe2 cofactor and viable mechanistic possibilities for oxygen incorporation into the reaction product. The hydroxyl groups on the substrate at carbon positions 2-5 have been omitted for clarity.

For several years after the elucidation of the MIOX structure and mechanism, it remained a "one-off" example of an HD-domain diiron oxygenase; bioinformatic analyses of known

55 sequences initially failed to identify any likely functional homologs. The prevalence of many additional potential MVDOs within this superfamily became apparent only after the 2012 discovery of a second example, (R)-1-hydroxyl-2-aminoethylphosphonate oxygenase (PhnZ)

(264). This enzyme promotes the second of two steps in a bacterial pathway that degrades the abundant organophosphonate, 2-aminoethylphosphonate (2-AEP) (264). The first enzyme, PhnY, is an iron- and 2-(oxo)glutarate-dependent dioxygenase that hydroxylates 2-AEP with R stereochemistry (264). The PhnY product serves as the substrate for the HD-MVDO, PhnZ.

Utilizing dioxygen as a co-substrate, PhnZ cleavages the C-P bond of OH-AEP to yield phosphate and glycine (264), incorporating one oxygen atom from O2 into the latter product

(265). Spectroscopic studies confirmed the prediction made on the basis of its similarity to MIOX that PhnZ would harbor a diiron cofactor and function in the mixed-valent oxidation state (265).

Structural comparison of MIOX and PhnZ revealed that both proteins present four conserved histidine and two conserved aspartate residues for coordination of their diiron cofactors

(Figure 1-14) (250, 265, 266), an architecture with clear differences from the 2-His/4-carboxylate coordination afforded by FDCOOs. The Fe1 site is created by the residues in the canonical HD- domain sequence motif, HxaHDxbD. One of these aspartate residues bridges to a second iron site that is completed by the two additional conserved histidines. This extended HD-motif,

HxaHDxbHxcHxdD, was used as a sequence marker to identify additional members of the HD- domain superfamily that might be capable of harboring dimetal clusters, with the expectation that a subset of these sequences might represent novel HD-MVDOs (265). A search using PhnZ as a query sequence returned thousands of new sequences with this extended motif, some of which are known to be or reasonably assigned as hydrolases, but many seem more likely to be MVDOs.

Comparison of these newly identified proteins revealed that previous attempts using MIOX as a query had failed because MIOX possesses an abnormally long "spacer" between the adjacent HD residues and the next conserved His ligand (Figure 1-14, red loop). One candidate, which we

56 designated TmpB, had originally been annotated as a phosphohydrolase but is actually a novel

HD-MVDO. TmpB catalyzes an O2-dependent oxidative C-P bond cleavage reaction analogous to that of PhnZ, but participates in a pathway specific for degradation of the N-trimethyl derivative of 2-AEP – 2-trimethylaminoethylphosphonate.

Despite the success of this bioinformatic analysis in identifying TmpB as a new HD-MVDO, the identification of proteins with a dinuclear sequence motif does not automatically correlate with oxygenase functionality. HD-domain proteins with dinuclear – even diiron – cofactors have been validated as phosphohydrolases, or at least have structures with bound phosphoester ligands to suggest this reactivity (265, 267-270). Therefore, more defined biochemical, spectroscopic, and structural descriptors are necessary to delineate boundaries that will allow for predictive functional assignment within this structural superfamily. With less than a handful of established examples, testing for additional HD-MVDOs is likely to dramatically expand the known repertoire of oxidative transformations of this emerging class of oxygenases.

1.5 Outlook

The relatively narrow paradigm emerging from early studies on what we now think of as canonical FDCOOs provided a framework for understanding the potent oxidative chemistry of which these enzyme are capable. The ongoing discoveries of novel ferritin-like dimetal proteins, as well as entirely new structural and functional classes of O2-activating diiron enzymes, have revealed an unanticipated diversity and versatility of this class of metalloenzymes. The most recently uncovered enzymes include a tRNA monooxygenase (271), epoxidases (272-274), an aging-associated enzyme that catalyzes 5-demethyoxyubiquinone hydroxylation (275-277), and an iron-sulfur cluster repair protein (278, 279). Further structural and mechanistic studies on the diiron enzymes covered in this review and novel representatives that continue to emerge apace

57 will undoubtedly further expand our understanding of what is possible within the realm of O2 activation at protein-bound dinuclear transition metal clusters.

1.6 References

1. Bertini I, G. H., Lippard SJ, Valentine JS (1994) Bioinorganic Chemistry, University Science Books, Mill Valley, CA. 2. Joosten, V., and van Berkel, W. J. H. (2007) Flavoenzymes. Curr Opin Chem Biol 11, 195-202. 3. Sono, M., Roach, M. P., Coulter, E. D., and Dawson, J. H. (1996) Heme-Containing Oxygenases. Chem Rev 96, 2841-2888. 4. Nordlund, P., and Eklund, H. (1995) Di-iron-carboxylate proteins. Curr Opin Struct Biol 5, 758-766. 5. Andrews, S. C. (2010) The Ferritin-like superfamily: Evolution of the biological iron storeman from a rubrerythrin-like ancestor. Biochim Biophys Acta 1800, 691-705. 6. Merkx, M., Kopp, D. A., Sazinsky, M. H., Blazyk, J. L., Muller, J., and Lippard, S. J. (2001) Dioxygen activation and methane hydroxylation by soluble methane monooxygenase: A tale of two irons and three proteins. Angew Chem Int Ed 40, 2782- 2807. 7. Wallar, B. J., and Lipscomb, J. D. (1996) Dioxygen Activation by Enzymes Containing Binuclear Non-Heme Iron Clusters. Chem Rev 96, 2625-2658. 8. Stubbe, J. (2003) Di-iron-tyrosyl radical ribonucleotide reductases. Curr Opin Chem Biol 7, 183-188. 9. Stubbe, J., Nocera, D. G., Yee, C. S., and Chang, M. C. (2003) Radical initiation in the class I ribonucleotide reductase: long-range proton-coupled electron transfer? Chem Rev 103, 2167-2201. 10. Fox, B. G., Lyle, K. S., and Rogge, C. E. (2004) Reactions of the diiron enzyme stearoyl- acyl carrier protein desaturase. Acc Chem Res 37, 421-429. 11. Shanklin, J., Guy, J. E., Mishra, G., and Lindqvist, Y. (2009) Desaturases: emerging models for understanding functional diversification of diiron-containing enzymes. J Biol Chem 284, 18559-18563. 12. Que, L., Jr., and Dong, Y. H. (1996) Modeling the oxygen activation chemistry of methane monooxygenase and ribonucleotide reductase. Acc. Chem. Res. 29, 190-196. 13. Herold, S., and Lippard, S. J. (1997) Carboxylate-bridged diiron(II)-complexes: Synthesis, characterization, and O-2-reactivity of models for the reduced diiron centers in methane monooxygenase and ribonucleotide reductase. J. Am. Chem. Soc. 119, 145-156. 14. Tshuva, E. Y., and Lippard, S. J. (2004) Synthetic models for non-heme carboxylate- bridged diiron metalloproteins: strategies and tactics. Chem Rev 104, 987-1012. 15. Friedle, S., Reisner, E., and Lippard, S. J. (2010) Current challenges of modeling diiron enzyme active sites for dioxygen activation by biomimetic synthetic complexes. Chem Soc Rev 39, 2768-2779. 16. Rardin, R. L., Tolman, W. B., and Lippard, S. J. (1991) Monodentate Carboxylate Complexes and the Carboxylate Shift - Implications for Polymetalloprotein Structure and Function. New J Chem 15, 417-430.

58 17. Rosenzweig, A. C., Frederick, C. A., Lippard, S. J., and Nordlund, P. (1993) Crystal structure of a bacterial non-haem iron hydroxylase that catalyses the biological oxidation of methane. Nature 366, 537-543. 18. Voegtli, W. C., Sommerhalter, M., Saleh, L., Baldwin, J., Bollinger, J. M., Jr., and Rosenzweig, A. C. (2003) Variable coordination geometries at the diiron(II) active site of ribonucleotide reductase R2. J Am Chem Soc 125, 15822-15830. 19. Higgins, I. J., and Quayle, J. R. (1970) OXYGENATION OF METHANE BY METHANE-GROWN PSEUDOMONAS-METHANICA AND METHANOMONAS- METHANOOXIDANS. Biochem. J. 118, 201-&. 20. Bollinger, J. M., Jr., Edmondson, D. E., Huynh, B. H., Filley, J., Norton, J. R., and Stubbe, J. (1991) Mechanism of assembly of the tyrosyl radical-dinuclear iron cluster cofactor of ribonucleotide reductase. Science 253, 292-298. 21. Bollinger, J. M., Jr., Tong, W. H., Ravi, N., Huynh, B. H., Edmondson, D. E., and Stubbe, J. (1994) Mechanism of Assembly of the Tyrosyl Radical-Diiron(Iii) Cofactor of Escherichia coli Ribonucleotide Reductase .2. Kinetics of the Excess Fe2+ Reaction by Optical, EPR, and Mössbauer Spectroscopies. J Am Chem Soc 116, 8015-8023. 22. Bollinger, J. M., Jr., Tong, W. H., Ravi, N., Huynh, B. H., Edmondson, D. E., and Stubbe, J. (1994) Mechanism of Assembly of the Tyrosyl Radical-Diiron(Iii) Cofactor of Escherichia coli Ribonucleotide Reductase .2. Kinetics of the Excess Fe2+ Reaction by Optical, EPR, and Mössbauer Spectroscopies. J Am Chem Soc 116, 8024-8032. 23. Baldwin, J., Krebs, C., Ley, B. A., Edmondson, D. E., Huynh, B. H., and Bollinger, J. M., Jr. (2000) Mechanism of rapid electron transfer during oxygen activation in the R2 subunit of Escherichia coli ribonucleotide reductase. 1. Evidence for a transient tryptophan radical. J Am Chem Soc 122, 12195-12206. 24. Wu, C. H., Jiang, W., Krebs, C., and Stubbe, J. (2007) YfaE, a ferredoxin involved in diferric-tyrosyl radical maintenance in Escherichia coli ribonucleotide reductase. Biochemistry 46, 11577-11588. 25. Blazyk, J. L., Gassner, G. T., and Lippard, S. J. (2005) Intermolecular electron-transfer reactions in soluble methane monooxygenase: a role for hysteresis in protein function. J Am Chem Soc 127, 17364-17376. 26. Müller, J., Lugovskoy, A. A., Wagner, G., and Lippard, S. J. (2002) NMR structure of the [2Fe-2S] ferredoxin domain from soluble methane monooxygenase reductase and interaction with its hydroxylase. Biochemistry 41, 42-51. 27. Liu, Y., Nesheim, J. C., Lee, S. K., and Lipscomb, J. D. (1995) Gating effects of component B on oxygen activation by the methane monooxygenase hydroxylase component. J Biol Chem 270, 24662-24665. 28. Sobrado, P., Lyle, K. S., Kaul, S. P., Turco, M. M., Arabshahi, I., Marwah, A., and Fox, B. G. (2006) Identification of the binding region of the [2Fe-2S] ferredoxin in stearoyl- acyl carrier protein desaturase: insight into the catalytic complex and mechanism of action. Biochemistry 45, 4848-4858. 29. Acheson, J. F., Bailey, L. J., Elsen, N. L., and Fox, B. G. (2014) Structural basis for biomolecular recognition in overlapping binding sites in a diiron enzyme system. Nat Commun 5, 5009. 30. Lippard, S. J. (2005) Hydroxylation of C-H bonds at carboxylate-bridged diiron centres. Philos Trans A Math Phys Eng Sci 363, 861-877; 1035-1040. 31. Wang, W., Liang, A. D., and Lippard, S. J. (2015) Coupling Oxygen Consumption with Hydrocarbon Oxidation in Bacterial Multicomponent Monooxygenases. Acc Chem Res 48, 2632-2639.

59 32. Murray, L. J., and Lippard, S. J. (2007) Substrate trafficking and dioxygen activation in bacterial multicomponent monooxygenases. Acc Chem Res 40, 466-474. 33. Liu, K. E., Valentine, A. M., Wang, D., Huynh, B. H., Edmondson, D. E., Salifoglou, A., and Lippard, S. J. (1995) Kinetic and Spectroscopic Characterization of Intermediates and Component Interactions in Reactions of Methane Monooxygenase from Methylococcus capsulatus (Bath). J Am Chem Soc 177, 10174-10185. 34. Broadwater, J. A., Ai, J., Loehr, T. M., Sanders-Loehr, J., and Fox, B. G. (1998) Peroxodiferric intermediate of stearoyl-acyl carrier protein delta 9 desaturase: oxidase reactivity during single turnover and implications for the mechanism of desaturation. Biochemistry 37, 14664-14671. 35. Tong, W. H., Chen, S., Lloyd, S. G., Edmondson, D. E., Huynh, B. H., and Stubbe, J. (1996) Mechanism of assembly of the diferric cluster-tyrosyl radical cofactor of Escherichia coli ribonucleotide reductase from the diferrous form of the R2 subunit. J Am Chem Soc 118, 2107-2108. 36. Bollinger, J. M., Jr., Krebs, C., Vicol, A., Chen, S. X., Ley, B. A., Edmondson, D. E., and Huynh, B. H. (1998) Engineering the diiron site of Escherichia coli ribonucleotide reductase protein R2 to accumulate an intermediate similar to H-peroxo, the putative peroxodiiron(III) complex from the methane monooxygenase catalytic cycle. J Am Chem Soc 120, 1094-1095. 37. Moënne-Loccoz, P., Baldwin, J., Ley, B. A., Loehr, T. M., and Bollinger, J. M., Jr.

(1998) O2 activation by non-heme diiron proteins: identification of a symmetric -1,2- peroxide in a mutant of ribonucleotide reductase. Biochemistry 37, 14659-14663. 38. Skulan, A. J., Brunold, T. C., Baldwin, J., Saleh, L., Bollinger, J. M., Jr., and Solomon, E. I. (2004) Nature of the peroxo intermediate of the W48F/D84E ribonucleotide reductase variant: implications for O2 activation by binuclear non-heme iron enzymes. J Am Chem Soc 126, 8842-8855. 39. Yun, D., García-Serres, R., Chicalese, B. M., An, Y. H., Huynh, B. H., and Bollinger, J. M., Jr. (2007) (-1,2-peroxo)diiron(III/III) complex as a precursor to the diiron(III/IV) intermediate X in the assembly of the iron-radical cofactor of ribonucleotide reductase from mouse. Biochemistry 46, 1925-1932. 40. Broadwater, J. A., Achim, C., Münck, E., and Fox, B. G. (1999) Mossbauer studies of the formation and reactivity of a quasi-stable peroxo intermediate of stearoyl-acyl carrier protein Delta 9-desaturase. Biochemistry 38, 12197-12204. 41. Liu, K. E., Wang, D. L., Huynh, B. H., Edmondson, D. E., Salifoglou, A., and Lippard, S. J. (1994) SPECTROSCOPIC DETECTION OF INTERMEDIATES IN THE REACTION OF DIOXYGEN WITH THE REDUCED METHANE MONOOXYGENASE HYDROXYLASE FROM METHYLOCOCCUS-CAPSULATUS (BATH). J. Am. Chem. Soc. 116, 7465-7466. 42. Kim, K., and Lippard, S. J. (1996) Structure and Mossbauer spectrum of a (mu-1,2- peroxo)bis(mu-carboxylato)diiron(III) model for the peroxo intermediate in the methane monooxygenase hydroxylase reaction cycle. J. Am. Chem. Soc. 118, 4914-4915. 43. Brunold, T. C., Tamura, N., Kitajima, N., Moro-oka, Y., and Solomon, E. I. (1998) Spectroscopic study of Fe(2)(O(2))(OBz)(2){HB(pz ')3}(2) : Nature of the mu-1,2 peroxide-Fe(III) bond and its possible relevance to O(2) activation by non-heme iron enzymes. J. Am. Chem. Soc. 120, 5674-5690. 44. Lee, S. Y., and Lipscomb, J. D. (1999) Oxygen activation catalyzed by methane monooxygenase hydroxylase component: Proton delivery during the O-O bond cleavage steps. Biochemistry 38, 4423-4432.

60 45. Brazeau, B. J., and Lipscomb, J. D. (2000) Kinetics and activation thermodynamics of methane monooxygenase compound Q formation and reaction with substrates. Biochemistry 39, 13503-13515. 46. Tinberg, C. E., and Lippard, S. J. (2009) Revisiting the Mechanism of Dioxygen Activation in Soluble Methane Monooxygenase from M. capsulatus (Bath): Evidence for a Multi-Step, Proton-Dependent Reaction Pathway. Biochemistry 48, 12145-12158. 47. Banerjee, R., Meier, K. K., Münck, E., and Lipscomb, J. D. (2013) Intermediate P* from soluble methane monooxygenase contains a diferrous cluster. Biochemistry 52, 4331- 4342. 48. Solomon, E. I., and Park, K. (2016) Structure/function correlations over binuclear non- heme iron active sites. J Biol Inorg Chem 21, 575-588. 49. Solomon, E. I., Brunold, T. C., Davis, M. I., Kemsley, J. N., Lee, S. K., Lehnert, N., Neese, F., Skulan, A. J., Yang, Y. S., and Zhou, J. (2000) Geometric and electronic structure/function correlations in non-heme iron enzymes. Chem Rev 100, 235-349. 50. Fox, B. G., Liu, Y., Dege, J. E., and Lipscomb, J. D. (1991) Complex-Formation between the Protein-Components of Methane Monooxygenase from Methylosinus trichosporium Ob3b - Identification of Sites of Component Interaction. J Biol Chem 266, 540-550. 51. Haas, J. A., and Fox, B. G. (1999) Role of hydrophobic partitioning in substrate selectivity and turnover of the Ricinus communis stearoyl acyl carrier protein triangle(9) desaturase. Biochemistry 38, 12833-12840. 52. Lee, S. K., Fox, B. G., Froland, W. A., Lipscomb, J. D., and Münck, E. (1993) A Transient Intermediate of the Methane Monooxygenase Catalytic Cycle Containing an Fe(IV)Fe(IV) Cluster. J Am Chem Soc 115, 6450-6451. 53. Nesheim, J. C., and Lipscomb, J. D. (1996) Large kinetic isotope effects in methane oxidation catalyzed by methane monooxygenase: evidence for C-H bond cleavage in a reaction cycle intermediate. Biochemistry 35, 10240-10247. 54. Siegbahn, P. E. M., Crabtree, R. H., and Nordlund, P. (1998) Mechanism of methane monooxygenase - a structural and quantum chemical perspective. J Biol Inorg Chem 3, 314-317. 55. Siegbahn, P. E. M. (1999) Theoretical model studies of the iron dimer complex of MMO and RNR. Inorg Chem 38, 2880-2889. 56. Basch, H., Mogi, K., Musaev, D. G., and Morokuma, K. (1999) Mechanism of the methane -> methanol conversion reaction catalyzed by methane monooxygenase: A density functional study. J Am Chem Soc 121, 7249-7256. 57. Gherman, B. F., Baik, M. H., Lippard, S. J., and Friesner, R. A. (2004) Dioxygen activation in methane monooxygenase: A theoretical study. J Am Chem Soc 126, 2978- 2990. 58. Shu, L., Nesheim, J. C., Kauffmann, K., Münck, E., Lipscomb, J. D., and Que, L., Jr. (1997) An Fe2IVO2 diamond core structure for the key intermediate Q of methane monooxygenase. Science 275, 515-518. 59. Banerjee, R., Proshlyakov, Y., Lipscomb, J. D., and Proshlyakov, D. A. (2015) Structure of the key species in the enzymatic oxidation of methane to methanol. Nature 518, 431- 434. 60. Elgren, T. E., Lynch, J. B., Juarez Garcia, C., Münck, E., Sjöberg, B.-M., and Que, L., Jr. (1991) Electron-Transfer Associated with Oxygen Activation in the B2 Protein of Ribonucleotide Reductase from Escherichia coli. J Biol Chem 266, 19265-19268. 61. Krebs, C., Chen, S. X., Baldwin, J., Ley, B. A., Patel, U., Edmondson, D. E., Huynh, B. H., and Bollinger, J. M., Jr. (2000) Mechanism of rapid electron transfer during oxygen activation in the R2 subunit of Escherichia coli ribonucleotide reductase. 2. Evidence for

61 and consequences of blocked electron transfer in the W48F variant. J Am Chem Soc 122, 12207-12219. 62. Saleh, L., Krebs, C., Ley, B. A., Naik, S., Huynh, B. H., and Bollinger, J. M., Jr. (2004) Use of a Chemical Trigger for Electron Transfer to Characterize a Precursor to Cluster X in Assembly of the Iron-Radical Cofactor of Escherichia coli Ribonucleotide Reductase. Biochemistry 43, 5953-5964. 63. Cranswick, M. A., Meier, K. K., Shan, X., Stubna, A., Kaizer, J., Mehn, M. P., Münck, E., and Que, L., Jr. (2012) Protonation of a peroxodiiron(III) complex and conversion to a diiron(III/IV) intermediate: implications for proton-assisted O-O bond cleavage in nonheme diiron enzymes. Inorg Chem 51, 10417-10426. 64. Dassama, L. M. K., Silakov, A., Krest, C. M., Calixto, J. C., Krebs, C., Bollinger, J. M., Jr., and Green, M. T. (2013) A 2.8 A Fe-Fe separation in the Fe2(III/IV) intermediate, X, from Escherichia coli ribonucleotide reductase. J Am Chem Soc 135, 16758-16761. 65. Mitic, N., Clay, M. D., Saleh, L., Bollinger, J. M., Jr., and Solomon, E. I. (2007) Spectroscopic and electronic structure studies of intermediate X in ribonucleotide reductase R2 and two variants: a description of the FeIV-oxo bond in the FeIII-O-FeIV dimer. J Am Chem Soc 129, 9049-9065. 66. Shanmugam, M., Doan, P. E., Lees, N. S., Stubbe, J., and Hoffman, B. M. (2009) Identification of protonated oxygenic ligands of ribonucleotide reductase intermediate X. J Am Chem Soc 131, 3370-3376. 67. Willems, J. P., Lee, H. I., Burdi, D., Doan, P. E., Stubbe, J., and Hoffman, B. M. (1997) Identification of the protonated oxygenic ligands of ribonucleotide reductase intermediate X by Q-band H-1,H-2 CW and pulsed ENDOR. J Am Chem Soc 119, 9816-9824. 68. Broun, P., Boddupalli, S., and Somerville, C. (1998) A bifunctional oleate 12- hydroxylase: desaturase from Lesquerella fendleri. Plant J 13, 201-210. 69. Beckmann, C., Rattke, J., Sperling, P., Heinz, E., and Boland, W. (2003) Stereochemistry of a bifunctional dihydroceramide delta 4-desaturase/hydroxylase from Candida albicans; a key enzyme of sphingolipid metabolism. Org Biomol Chem 1, 2448-2454. 70. Broadwater, J. A., Whittle, E., and Shanklin, J. (2002) Desaturation and hydroxylation. Residues 148 and 324 of Arabidopsis FAD2, in addition to substrate chain length, exert a major influence in partitioning of catalytic specificity. J Biol Chem 277, 15613-15620. 71. Behrouzian, B., Savile, C. K., Dawson, B., Buist, P. H., and Shanklin, J. (2002) Exploring the hydroxylation-dehydrogenation connection: Novel catalytic activity of castor stearoyl-ACP 9 desaturase. J Am Chem Soc 124, 3277-3283. 72. Logan, D. T., deMaré, F., Persson, B. O., Slaby, A., Sjöberg, B.-M., and Nordlund, P. (1998) Crystal structures of two self-hydroxylating ribonucleotide reductase protein R2 mutants: structural basis for the oxygen-insertion step of hydroxylation reactions catalyzed by diiron proteins. Biochemistry 37, 10798-10807. 73. Ormö, M., deMaré, F., Regnström, K., Aberg, A., Sahlin, M., Ling, J., Loehr, T. M., Sanders-Loehr, J., and Sjöberg, B.-M. (1992) Engineering of the iron site in ribonucleotide reductase to a self-hydroxylating monooxygenase. J Biol Chem 267, 8711- 8714. 74. Baldwin, J., Voegtli, W. C., Khidekel, N., Moënne-Loccoz, P., Krebs, C., Pereira, A. S., Ley, B. A., Huynh, B. H., Loehr, T. M., Riggs-Gelasco, P. J., Rosenzweig, A. C., and Bollinger, J. M., Jr. (2001) Rational reprogramming of the R2 subunit of Escherichia coli ribonucleotide reductase into a self-hydroxylating monooxygenase. J Am Chem Soc 123, 7017-7030.

62 75. Tinberg, C. E., and Lippard, S. J. (2010) Oxidation reactions performed by soluble methane monooxygenase hydroxylase intermediates H(peroxo) and Q proceed by distinct mechanisms. Biochemistry 49, 7902-7912. 76. Chino, M., Maglio, O., Nastri, F., Pavone, V., DeGrado, W. F., and Lombardi, A. (2015) Artificial Diiron Enzymes with a De Novo Designed Four-Helix Bundle Structure. Eur J Inorg Chem 2015, 3371-3390. 77. Nastri, F., Chino, M., Maglio, O., Bhagi-Damodaran, A., Lu, Y., and Lombardi, A. (2016) Design and engineering of artificial oxygen-activating metalloenzymes. Chem Soc Rev 45, 5020-5054. 78. Yu, F., Cangelosi, V. M., Zastrow, M. L., Tegoni, M., Plegaria, J. S., Tebo, A. G., Mocny, C. S., Ruckthong, L., Qayyum, H., and Pecoraro, V. L. (2014) Protein design: toward functional metalloenzymes. Chem Rev 114, 3495-3578. 79. Reig, A. J., Pires, M. M., Snyder, R. A., Wu, Y., Jo, H., Kulp, D. W., Butch, S. E., Calhoun, J. R., Szyperski, T., Solomon, E. I., and DeGrado, W. F. (2012) Alteration of the oxygen-dependent reactivity of de novo Due Ferri proteins. Nat Chem 4, 900-906. 80. Faiella, M., Andreozzi, C., de Rosales, R. T., Pavone, V., Maglio, O., Nastri, F., DeGrado, W. F., and Lombardi, A. (2009) An artificial di-iron oxo-protein with phenol oxidase activity. Nat Chem Biol 5, 882-884. 81. Snyder, R. A., Betzu, J., Butch, S. E., Reig, A. J., DeGrado, W. F., and Solomon, E. I. (2015) Systematic Perturbations of Binuclear Non-heme Iron Sites: Structure and Dioxygen Reactivity of de Novo Due Ferri Proteins. Biochemistry 54, 4637-4651. 82. Murray, L. J., Garcia-Serres, R., Naik, S., Huynh, B. H., and Lippard, S. J. (2006) Dioxygen activation at non-heme diiron centers: characterization of intermediates in a mutant form of toluene/o-xylene monooxygenase hydroxylase. J Am Chem Soc 128, 7458-7459. 83. Murray, L. J., Naik, S. G., Ortillo, D. O., Garcia-Serres, R., Lee, J. K., Huynh, B. H., and Lippard, S. J. (2007) Characterization of the arene-oxidizing intermediate in ToMOH as a diiron(III) species. J Am Chem Soc 129, 14500-14510. 84. Korboukh, V. K., Li, N., Barr, E. W., Bollinger, J. M., Jr., and Krebs, C. (2009) A long- lived, substrate-hydroxylating peroxodiiron(III/III) intermediate in the amine oxygenase, AurF, from Streptomyces thioluteus. J Am Chem Soc 131, 13608-13609. 85. Makris, T. M., Vu, V. V., Meier, K. K., Komor, A. J., Rivard, B. S., Münck, E., Que, L., Jr., and Lipscomb, J. D. (2015) An unusual peroxo intermediate of the arylamine oxygenase of the chloramphenicol biosynthetic pathway. J Am Chem Soc 137, 1608- 1617. 86. Pandelia, M.-E., Li, N., Nørgaard, H., Warui, D. M., Rajakovich, L. J., Chang, W.-c., Booker, S. J., Krebs, C., and Bollinger, J. M., Jr. (2013) Substrate-triggered addition of dioxygen to the diferrous cofactor of aldehyde-deformylating oxygenase to form a diferric-peroxide intermediate. J Am Chem Soc 135, 15801-15812. 87. Vu, V. V., Emerson, J. P., Martinho, M., Kim, Y. S., Münck, E., Park, M. H., and Que, L., Jr. (2009) Human deoxyhypusine hydroxylase, an enzyme involved in regulating cell growth, activates O2 with a nonheme diiron center. Proc Natl Acad Sci USA 106, 14814- 14819. 88. Denisov, I. G., Makris, T. M., Sligar, S. G., and Schlichting, I. (2005) Structure and chemistry of cytochrome P450. Chem Rev 105, 2253-2277. 89. Lewis, E. A., and Tolman, W. B. (2004) Reactivity of dioxygen-copper systems. Chem Rev 104, 1047-1076. 90. Mirica, L. M., Ottenwaelder, X., and Stack, T. D. P. (2004) Structure and spectroscopy of copper-dioxygen complexes. Chem Rev 104, 1013-1045.

63 91. Cafaro, V., Izzo, V., Scognamiglio, R., Notomista, E., Capasso, P., Casbarra, A., Pucci, P., and Di Donato, A. (2004) Phenol Hydroxylase and Toluene/o-Xylene Monooxygenase from Pseudomonas stutzeri OX1: Interplay between Two Enzymes. Appl Environ Microbiol 70, 2211-2219. 92. Tinberg, C. E., Song, W. J., Izzo, V., and Lippard, S. J. (2011) Multiple roles of component proteins in bacterial multicomponent monooxygenases: phenol hydroxylase and toluene/o-xylene monooxygenase from Pseudomonas sp. OX1. Biochemistry 50, 1788-1798. 93. Murray, L. J., Garcia-Serres, R., McCormick, M. S., Davydov, R., Naik, S. G., Kim, S. H., Hoffman, B. M., Huynh, B. H., and Lippard, S. J. (2007) Dioxygen activation at non- heme diiron centers: oxidation of a proximal residue in the I100W variant of toluene/o- xylene monooxygenase hydroxylase. Biochemistry 46, 14795-14809. 94. Liang, A. D., and Lippard, S. J. (2015) Single Turnover Reveals Oxygenated Intermediates in Toluene/o-Xylene Monooxygenase in the Presence of the Native Redox Partners. J Am Chem Soc 137, 10520-10523. 95. Bochevarov, A. D., Li, J., Song, W. J., Friesner, R. A., and Lippard, S. J. (2011) Insights into the different dioxygen activation pathways of methane and toluene monooxygenase hydroxylases. J Am Chem Soc 133, 7384-7397. 96. Song, W. J., Behan, R. K., Naik, S. G., Huynh, B. H., and Lippard, S. J. (2009) Characterization of a peroxodiiron(III) intermediate in the T201S variant of toluene/o- xylene monooxygenase hydroxylase from Pseudomonas sp. OX1. J Am Chem Soc 131, 6074-6075. 97. Song, W. J., McCormick, M. S., Behan, R. K., Sazinsky, M. H., Jiang, W., Lin, J., Krebs, C., and Lippard, S. J. (2010) Active site threonine facilitates proton transfer during dioxygen activation at the diiron center of toluene/o-xylene monooxygenase hydroxylase. J Am Chem Soc 132, 13582-13585. 98. Song, W. J., Gucinski, G., Sazinsky, M. H., and Lippard, S. J. (2011) Tracking a defined route for O(2) migration in a dioxygen-activating diiron enzyme. Proc Natl Acad Sci USA 108, 14795-14800. 99. Raag, R., Martinis, S. A., Sligar, S. G., and Poulos, T. L. (1991) Crystal-Structure of the Cytochrome-P-450cam Active-Site Mutant Thr252ala. Biochemistry 30, 11420-11429. 100. Shimizu, T., Murakami, Y., and Hatano, M. (1994) Glu(318) and Thr(319) Mutations of Cytochrome-P450 1a2 Remarkably Enhance Homolytic O-O Cleavage of Alkyl Hydroperoxides - an Optical-Absorption Spectral Study. J Biol Chem 269, 13296-13304. 101. Blobaum, A. L., Kent, U. M., and Hollenberg, P. F. (2004) A role for threonine 303 in proton delivery to the active site of cytochrome P450 2E1. Drug Metab Rev 36, 271-271. 102. Mitchell, K. H., Studts, J. M., and Fox, B. G. (2002) Combined participation of hydroxylase active site residues and effector protein binding in a para to ortho modulation of toluene 4-monooxygenase regiospecificity. Biochemistry 41, 3176-3188. 103. He, J., and Hertweck, C. (2004) Biosynthetic origin of the rare nitroaryl moiety of the polyketide antibiotic aureothin: involvement of an unprecedented N-oxygenase. J Am Chem Soc 126, 3694-3695. 104. Lee, J., and Zhao, H. (2006) Mechanistic studies on the conversion of arylamines into arylnitro compounds by aminopyrrolnitrin oxygenase: identification of intermediates and kinetic studies. Angew Chem Int Ed 45, 622-625. 105. Lu, H., Chanco, E., and Zhao, H. (2012) CmlI is an N-oxygenase in the biosynthesis of chloramphenicol. Tetrahedron 68, 7651-7654.

64 106. Komor, A. J., Rivard, B. S., Fan, R., Guo, Y., Que, L., Jr., and Lipscomb, J. D. (2016) Mechanism for Six-Electron Aryl-N-Oxygenation by the Non-Heme Diiron Enzyme CmlI. J Am Chem Soc 138, 7411-7421. 107. Li, N., Korboukh, V. K., Krebs, C., and Bollinger, J. M., Jr. (2010) Four-electron oxidation of p-hydroxylaminobenzoate to p-nitrobenzoate by a peroxodiferric complex in AurF from Streptomyces thioluteus. Proc Natl Acad Sci USA 107, 15722-15727. 108. Simurdiak, M., Lee, J., and Zhao, H. (2006) A new class of arylamine oxygenases: evidence that p-aminobenzoate N-oxygenase (AurF) is a di-iron enzyme and further mechanistic studies. Chembiochem 7, 1169-1172. 109. Zocher, G., Winkler, R., Hertweck, C., and Schulz, G. E. (2007) Structure and action of the N-oxygenase AurF from Streptomyces thioluteus. J Mol Biol 373, 65-74. 110. Jiang, W., Yun, D., Saleh, L., Barr, E. W., Xing, G., Hoffart, L. M., Maslak, M. A., Krebs, C., and Bollinger, J. M., Jr. (2007) A manganese(IV)/iron(III) cofactor in Chlamydia trachomatis ribonucleotide reductase. Science 316, 1188-1191. 111. Krebs, C., Matthews, M. L., Jiang, W., and Bollinger, J. M., Jr. (2007) AurF from Streptomyces thioluteus and a possible new family of manganese/iron oxygenases. Biochemistry 46, 10413-10418. 112. Choi, Y. S., Zhang, H., Brunzelle, J. S., Nair, S. K., and Zhao, H. (2008) In vitro reconstitution and crystal structure of p-aminobenzoate N-oxygenase (AurF) involved in aureothin biosynthesis. Proc Natl Acad Sci USA 105, 6858-6863. 113. Knoot, C. J., Kovaleva, E. G., and Lipscomb, J. D. (2016) Crystal structure of CmlI, the arylamine oxygenase from the chloramphenicol biosynthetic pathway. J Biol Inorg Chem 21, 589-603. 114. Park, K., Li, N., Kwak, Y., Srnec, M., Bell, C. B., Liu, L. V., Wong, S. D., Yoda, Y., Kitao, S., Seto, M., Hu, M., Zhao, J., Krebs, C., Bollinger, J. M., Jr., and Solomon, E. I. (2017) Peroxide Activation for Electrophilic Reactivity by the Binuclear Non-heme Iron Enzyme AurF. J Am Chem Soc 139, 7062-7070. 115. Wang, C., and Chen, H. (2017) Convergent Theoretical Prediction of Reactive Oxidant Structures in Diiron Arylamine Oxygenases AurF and CmlI: Peroxo or Hydroperoxo? J Am Chem Soc 139, 13038-13046. 116. Chavez, F. A., Ho, R. Y. N., Pink, M., Young, V. G., Kryatov, S. V., Rybak-Akimova, E. V., Andres, H., Münck, E., Que, L., Jr., and Tolman, W. B. (2002) Unusual peroxo intermediates in the reaction of dioxygen with carboxylate-bridged diiron(II,II) paddlewheel complexes. Angew Chem Int Ed 41, 149-152. 117. Jasniewski, A. J., Komor, A. J., Lipscomb, J. D., and Que, L., Jr. (2017) Unprecedented (mu-1,1-Peroxo)diferric Structure for the Ambiphilic Orange Peroxo Intermediate of the Nonheme N-Oxygenase CmlI. J Am Chem Soc 139, 10472-10485. 118. Komor, A. J., Rivard, B. S., Fan, R., Guo, Y., Que, L., Jr., and Lipscomb, J. D. (2017) CmlI N-Oxygenase Catalyzes the Final Three Steps in Chloramphenicol Biosynthesis without Dissociation of Intermediates. Biochemistry 56, 4940-4950. 119. Entsch, B., and van Berkel, W. J. (1995) Structure and mechanism of para- hydroxybenzoate hydroxylase. FASEB J 9, 476-483. 120. Cole, L. J., Entsch, B., Ortiz-Maldonado, M., and Ballou, D. P. (2005) Properties of p- hydroxybenzoate hydroxylase when stabilized in its open conformation. Biochemistry 44, 14807-14817. 121. Massey, V. (1994) Activation of molecular oxygen by flavins and flavoproteins. J Biol Chem 269, 22459-22462.

65 122. Maeda-Yorita, K., and Massey, V. (1993) On the reaction mechanism of phenol hydroxylase. New information obtained by correlation of fluorescence and absorbance stopped flow studies. J Biol Chem 268, 4134-4144. 123. Sheng, D., Ballou, D. P., and Massey, V. (2001) Mechanistic studies of cyclohexanone monooxygenase: chemical properties of intermediates involved in catalysis. Biochemistry 40, 11156-11167. 124. Francisco, W. A., Abu-Soud, H. M., DelMonte, A. J., Singleton, D. A., Baldwin, T. O., and Raushel, F. M. (1998) Deuterium kinetic isotope effects and the mechanism of the bacterial luciferase reaction. Biochemistry 37, 2596-2606. 125. Wertz, D. L., and Valentine, J. S. (2000) Nucleophilicity of Iron-Peroxo Porphyrin Complexes, In Metal-Oxo and Metal-Peroxo Species in Catalytic Oxidations (Meunier, B., Ed.), pp 37-60, Springer Berlin Heidelberg, Berlin, Heidelberg. 126. Schirmer, A., Rude, M. A., Li, X., Popova, E., and del Cardayre, S. B. (2010) Microbial biosynthesis of alkanes. Science 329, 559-562. 127. Warui, D. M., Li, N., Nørgaard, H., Krebs, C., Bollinger, J. M., Jr., and Booker, S. J. (2011) Detection of formate, rather than carbon monoxide, as the stoichiometric coproduct in conversion of fatty aldehydes to alkanes by a cyanobacterial aldehyde decarbonylase. J Am Chem Soc 133, 3316-3319. 128. Li, N., Nørgaard, H., Warui, D. M., Booker, S. J., Krebs, C., and Bollinger, J. M., Jr. (2011) Conversion of fatty aldehydes to alka(e)nes and formate by a cyanobacterial aldehyde decarbonylase: cryptic redox by an unusual dimetal oxygenase. J Am Chem Soc 133, 6158-6161. 129. Li, N., Chang, W.-c., Warui, D. M., Booker, S. J., Krebs, C., and Bollinger, J. M., Jr. (2012) Evidence for only oxygenative cleavage of aldehydes to alk(a/e)nes and formate by cyanobacterial aldehyde decarbonylases. Biochemistry 51, 7908-7916. 130. Schirmer, A., Rude M.A., Brubaker S. (2009) Methods and compositions for producing hydrocarbons, (LS9, Ed.), United States of America. 131. Reppas, N. B., and Ridley, C. P. (2010) Methods of compositions for the recominant biosynthesis of n-alkanes, (Inc., J. U., Ed.), United States of America. 132. Das, D., Eser, B. E., Han, J., Sciore, A., and Marsh, E. N. G. (2011) Oxygen-independent decarbonylation of aldehydes by cyanobacterial aldehyde decarbonylase: a new reaction of diiron enzymes. Angew Chem Int Ed 50, 7148-7152. 133. Eser, B. E., Das, D., Han, J., Jones, P. R., and Marsh, E. N. G. (2011) Oxygen- independent alkane formation by non-heme iron-dependent cyanobacterial aldehyde decarbonylase: investigation of kinetics and requirement for an external electron donor. Biochemistry 50, 10743-10750. 134. Jia, C., Li, M., Li, J., Zhang, J., Zhang, H., Cao, P., Pan, X., Lu, X., and Chang, W. (2015) Structural insights into the catalytic mechanism of aldehyde-deformylating oxygenases. Protein Cell 6, 55-67. 135. Park, A. K., Kim, I. S., Jeon, B. W., Roh, S. J., Ryu, M. Y., Baek, H. R., Jo, S. W., Kim, Y. S., Park, H., Lee, J. H., Yoon, H. S., and Kim, H. W. (2016) Crystal structures of aldehyde deformylating oxygenase from Limnothrix sp KNUA012 and Oscillatoria sp KNUA011. Biochem Biophys Res Commun 477, 395-400. 136. Rajakovich, L. J., Nørgaard, H., Warui, D. M., Chang, W.-c., Li, N., Booker, S. J., Krebs, C., Bollinger, J. M., Jr., and Pandelia, M.-E. (2015) Rapid Reduction of the Diferric- Peroxyhemiacetal Intermediate in Aldehyde-Deformylating Oxygenase by a Cyanobacterial Ferredoxin: Evidence for a Free-Radical Mechanism. J Am Chem Soc 137, 11695-11709.

66 137. Paul, B., Das, D., Ellington, B., and Marsh, E. N. G. (2013) Probing the mechanism of cyanobacterial aldehyde decarbonylase using a cyclopropyl aldehyde. J Am Chem Soc 135, 5234-5237. 138. Das, D., Ellington, B., Paul, B., and Marsh, E. N. G. (2014) Mechanistic insights from reaction of alpha-oxiranyl-aldehydes with cyanobacterial aldehyde deformylating oxygenase. ACS Chem Biol 9, 570-577. 139. Waugh, M. W., and Marsh, E. N. G. (2014) Solvent isotope effects on alkane formation by cyanobacterial aldehyde deformylating oxygenase and their mechanistic implications. Biochemistry 53, 5537-5543. 140. Kallio, P., Pasztor, A., Thiel, K., Akhtar, M. K., and Jones, P. R. (2014) An engineered pathway for the biosynthesis of renewable propane. Nat Commun 5, 4731. 141. Law, N. A., Caudle, T., Pecoraro, V. L. (1998) Manganese redox enzymes and model systems: propertes, structure, and reactivity. Adv Inorg Chem 46, 305-440. 142. Pace, R. J., Stranger, R., and Petrie, S. (2012) Why nature chose Mn for the water oxidase in Photosystem II. Dalton Trans 41, 7179-7189. 143. Cox, N., Pantazis, D. A., Neese, F., and Lubitz, W. (2013) Biological Water Oxidation. Acc. Chem. Res. 46, 1588-1596. 144. Perez-Navarro, M., Neese, F., Lubitz, W., Pantazis, D. A., and Cox, N. (2016) Recent developments in biological water oxidation. Curr. Opin. Chem. Biol. 31, 113-119. 145. Whittaker, J. W. (2012) Non-heme manganese catalase--the 'other' catalase. Arch Biochem Biophys 525, 111-120. 146. Roshick, C., Iliffe-Lee, E. R., and McClarty, G. (2000) Cloning and characterization of ribonucleotide reductase from Chlamydia trachomatis. J Biol Chem 275, 38111-38119. 147. Högbom, M., Stenmark, P., Voevodskaya, N., McClarty, G., Gräslund, A., and Nordlund, P. (2004) The radical site in chlamydial ribonucleotide reductase defines a new R2 subclass. Science 305, 245-248. 148. Voevodskaya, N., Lendzian, F., and Gräslund, A. (2005) A stable FeIII-FeIV replacement of tyrosyl radical in a class I ribonucleotide reductase. Biochem Biophys Res Commun 330, 1213-1216. 149. Voevodskaya, N., Narvaez, A. J., Domkin, V., Torrents, E., Thelander, L., and Gräslund, A. (2006) Chlamydial ribonucleotide reductase: tyrosyl radical function in catalysis replaced by the FeIII-FeIV cluster. Proc Natl Acad Sci USA 103, 9850-9854. 150. Jiang, W., Bollinger, J. M., Jr., and Krebs, C. (2007) The active form of Chlamydia trachomatis ribonucleotide reductase R2 protein contains a heterodinuclear Mn(IV)/Fe(III) cluster with S = 1 ground state. J Am Chem Soc 129, 7504-7505. 151. Jiang, W., Yun, D., Saleh, L., Bollinger, J. M., Jr., and Krebs, C. (2008) Formation and function of the Manganese(IV)/Iron(III) cofactor in Chlamydia trachomatis ribonucleotide reductase. Biochemistry 47, 13736-13744. 152. Jiang, W., Hoffart, L. M., Krebs, C., and Bollinger, J. M., Jr. (2007) A manganese(IV)/iron(IV) intermediate in assembly of the manganese(IV)/iron(III) cofactor of Chlamydia trachomatis ribonucleotide reductase. Biochemistry 46, 8709- 8716. 153. Martinie, R. J., Blaesi, E. J., Krebs, C., Bollinger, J. M., Jr., Silakov, A., and Pollock, C. J. (2017) Evidence for a Di-mu-oxo Diamond Core in the Mn(IV)/Fe(IV) Activation Intermediate of Ribonucleotide Reductase from Chlamydia trachomatis. J Am Chem Soc 139, 1950-1957. 154. Jiang, W., Saleh, L., Barr, E. W., Xie, J., Gardner, M. M., Krebs, C., and Bollinger, J. M., Jr. (2008) Branched activation- and catalysis-specific pathways for electron relay to the

67 manganese/iron cofactor in ribonucleotide reductase from Chlamydia trachomatis. Biochemistry 47, 8477-8484. 155. Younker, J. M., Krest, C. M., Jiang, W., Krebs, C., Bollinger, J. M., Jr., and Green, M. T. (2008) Structural analysis of the Mn(IV)/Fe(III) cofactor of Chlamydia trachomatis ribonucleotide reductase by extended X-ray absorption fine structure spectroscopy and density functional theory calculations. J Am Chem Soc 130, 15022-15027. 156. Andersson, C. S., Ohrström, M., Popovic-Bijelic, A., Gräslund, A., Stenmark, P., and Högbom, M. (2012) The Manganese Ion of the Heterodinuclear Mn/Fe Cofactor in Chlamydia trachomatis Ribonucleotide Reductase R2c Is Located at Metal Position 1. J Am Chem Soc 134, 123-125. 157. Dassama, L. M. K., Boal, A. K., Krebs, C., Rosenzweig, A. C., and Bollinger, J. M., Jr. (2012) Evidence that the beta subunit of Chlamydia trachomatis ribonucleotide reductase is active with the manganese ion of its manganese(IV)/iron(III) cofactor in site 1. J Am Chem Soc 134, 2520-2523. 158. Dassama, L. M. K., Krebs, C., Bollinger, J. M., Jr., Rosenzweig, A. C., and Boal, A. K. (2013) Structural basis for assembly of the Mn(IV)/Fe(III) cofactor in the class Ic ribonucleotide reductase from Chlamydia trachomatis. Biochemistry 52, 6424-6436. 159. Saleh, L., Krebs, C., Ley, B. A., Naik, S., Huynh, B. H., and Bollinger, J. M., Jr. (2004) Use of a chemical trigger for electron transfer to characterize a precursor to cluster X in assembly of the iron-radical cofactor of Escherichia coli ribonucleotide reductase. Biochemistry 43, 5953-5964. 160. Luo, Y. R. (2007) Comprehensive Handbook of Chemical Bond Energies, CRC Press, Boca Raton, FL. 161. Andersson, C. S., and Högbom, M. (2009) A Mycobacterium tuberculosis ligand-binding Mn/Fe protein reveals a new cofactor in a remodeled R2-protein scaffold. Proc Natl Acad Sci USA 106, 5633-5638. 162. Griese, J. J., Roos, K., Cox, N., Shafaat, H. S., Branca, R. M., Lehtio, J., Gräslund, A., Lubitz, W., Siegbahn, P. E. M., and Högbom, M. (2013) Direct observation of structurally encoded metal discrimination and ether bond formation in a heterodinuclear metalloprotein. Proc Natl Acad Sci USA 110, 17189-17194. 163. Griese, J. J., Kositzki, R., Schrapers, P., Branca, R. M., Nordstrom, A., Lehtio, J., Haumann, M., and Högbom, M. (2015) Structural Basis for Oxygen Activation at a Heterodinuclear Manganese/Iron Cofactor. J Biol Chem 290, 25254-25272. 164. Kutin, Y., Srinivas, V., Fritz, M., Kositzki, R., Shafaat, H. S., Birrell, J., Bill, E., Haumann, M., Lubitz, W., Högbom, M., Griese, J. J., and Cox, N. (2016) Divergent assembly mechanisms of the manganese/iron cofactors in R2lox and R2c proteins. J Inorg Biochem 162, 164-177. 165. Stubbe, J., and Cotruvo, J. A., Jr. (2011) Control of metallation and active cofactor assembly in the class Ia and Ib ribonucleotide reductases: diiron or dimanganese? Curr Opin Chem Biol 15, 284-290. 166. Willing, A., Follmann, H., and Auling, G. (1988) Ribonucleotide Reductase of Brevibacterium-Ammoniagenes Is a Manganese Enzyme. European Journal of Biochemistry 170, 603-611. 167. Huque, Y., Fieschi, F., Torrents, E., Gibert, I., Eliasson, R., Reichard, P., Sahlin, M., and Sjöberg, B.-M. (2000) The active form of the R2F protein of class Ib ribonucleotide reductase from Corynebacterium ammoniagenes is a diferric protein. J Biol Chem 275, 25365-25371.

68 168. Monje-Casas, F., Jurado, J., Prieto-Alamo, M. J., Holmgren, A., and Pueyo, C. (2001) Expression analysis of the nrdHIEF operon from Escherichia coli. Conditions that trigger the transcript level in vivo. J Biol Chem 276, 18031-18037. 169. McHugh, J. P., Rodriguez-Quinones, F., Abdul-Tehrani, H., Svistunenko, D. A., Poole, R. K., Cooper, C. E., and Andrews, S. C. (2003) Global iron-dependent gene regulation in Escherichia coli. A new mechanism for iron homeostasis. J Biol Chem 278, 29478- 29486. 170. Hartig, E., Hartmann, A., Schatzle, M., Albertini, A. M., and Jahn, D. (2006) The Bacillus subtilis nrdEF genes, encoding a class Ib ribonucleotide reductase, are essential for aerobic and anaerobic growth. Appl Environ Microbiol 72, 5260-5265. 171. Lundin, D., Torrents, E., Poole, A. M., and Sjöberg, B.-M. (2009) RNRdb, a curated database of the universal enzyme family ribonucleotide reductase, reveals a high level of misannotation in sequences deposited to Genbank. BMC Genomics 10, 589. 172. Fieschi, F., Torrents, E., Toulokhonova, L., Jordan, A., Hellman, U., Barbe, J., Gibert, I., Karlsson, M., and Sjöberg, B.-M. (1998) The manganese-containing ribonucleotide reductase of Corynebacterium ammoniagenes is a class Ib enzyme. J Biol Chem 273, 4329-4337. 173. Papp-Wallace, K. M., and Maguire, M. E. (2006) Manganese transport and the role of manganese in virulence. Annu Rev Microbiol 60, 187-209. 174. Oka, T., Udagawa, K., and Kinoshita, S. (1968) Unbalanced growth death due to depletion of Mn2+ in Brevibacterium ammoniagenes. J Bacteriol 96, 1760-1767. 175. Cotruvo, J. A., Jr., Stich, T. A., Britt, R. D., and Stubbe, J. (2013) Mechanism of assembly of the dimanganese-tyrosyl radical cofactor of class Ib ribonucleotide reductase: enzymatic generation of superoxide is required for tyrosine oxidation via a Mn(III)Mn(IV) intermediate. J Am Chem Soc 135, 4027-4039. 176. Cotruvo, J. A., Jr., and Stubbe, J. (2008) NrdI, a flavodoxin involved in maintenance of the diferric-tyrosyl radical cofactor in Escherichia coli class Ib ribonucleotide reductase. Proc Natl Acad Sci USA 105, 14383-14388. 177. Roca, I., Torrents, E., Sahlin, M., Gibert, I., and Sjöberg, B.-M. (2008) NrdI essentiality for class Ib ribonucleotide reduction in Streptococcus pyogenes. J Bacteriol 190, 4849- 4858. 178. Cox, N., Ogata, H., Stolle, P., Reijerse, E., Auling, G., and Lubitz, W. (2010) A tyrosyl- dimanganese coupled spin system is the native metalloradical cofactor of the R2F subunit of the ribonucleotide reductase of Corynebacterium ammoniagenes. J Am Chem Soc 132, 11197-11213. 179. Zhang, Y., and Stubbe, J. (2011) Bacillus subtilis class Ib ribonucleotide reductase is a dimanganese(III)-tyrosyl radical enzyme. Biochemistry 50, 5615-5623. 180. Boal, A. K., Cotruvo, J. A., Jr., Stubbe, J., and Rosenzweig, A. C. (2010) Structural basis for activation of class Ib ribonucleotide reductase. Science 329, 1526-1530. 181. Hantke, K. (2001) Iron and metal regulation in bacteria. Curr Opin Microbiol 4, 172-177. 182. Weinberg, E. D. (1975) Nutritional Immunity - Hosts Attempt to Withhold Iron from Microbial Invaders. J Am Med Assoc 231, 39-41. 183. Weinberg, E. D. (1975) Metal Starvation of Pathogens by Hosts. Bioscience 25, 314-318. 184. Hood, M. I., and Skaar, E. P. (2012) Nutritional immunity: transition metals at the pathogen-host interface. Nat Rev Microbiol 10, 525-537. 185. Juttukonda, L. J., and Skaar, E. P. (2015) Manganese homeostasis and utilization in pathogenic bacteria. Mol Microbiol 97, 216-228. 186. Miller, A. F. (2004) Superoxide dismutases: active sites that save, but a protein that kills. Curr Opin Chem Biol 8, 162-168.

69 187. Miller, A. F. (2012) Superoxide dismutases: Ancient enzymes and new insights. FEBS Lett 586, 585-595. 188. Shanklin, J., Achim, C., Schmidt, H., Fox, B. G., and Münck, E. (1997) Mössbauer studies of alkane omega-hydroxylase: evidence for a diiron cluster in an integral- membrane enzyme. Proc Natl Acad Sci USA 94, 2981-2986. 189. Shanklin, J., Whittle, E., and Fox, B. G. (1994) Eight Histidine Residues Are Catalytically Essential in a Membrane-Associated Iron Enzyme, Stearoyl-CoA Desaturase, and Are Conserved in Alkane Hydroxylase and Xylene Monooxygenase. Biochemistry 33, 12787-12794. 190. Taton, M., Husselstein, T., Benveniste, P., and Rahier, A. (2000) Role of highly conserved residues in the reaction catalyzed by recombinant Delta7-sterol-C5(6)- desaturase studied by site-directed mutagenesis. Biochemistry 39, 701-711. 191. Shanklin, J., and Whittle, E. (2003) Evidence linking thePseudomonas oleovoransalkane ω-hydroxylase, an integral membrane diiron enzyme, and the fatty acid desaturase family. FEBS Lett 545, 188-192. 192. Bai, Y., McCoy, J. G., Levin, E. J., Sobrado, P., Rajashankar, K. R., Fox, B. G., and Zhou, M. (2015) X-ray structure of a mammalian stearoyl-CoA desaturase. Nature 524, 252-256. 193. Wang, H., Klein, M. G., Zou, H., Lane, W., Snell, G., Levin, I., Li, K., and Sang, B. C. (2015) Crystal structure of human stearoyl-coenzyme A desaturase in complex with substrate. Nat Struct Mol Biol 22, 581-585. 194. Zhu, G., Koszelak-Rosenblum, M., Connelly, S. M., Dumont, M. E., and Malkowski, M. G. (2015) The Crystal Structure of an Integral Membrane Fatty Acid alpha-Hydroxylase. J Biol Chem 290, 29820-29833. 195. Bernard, A., Domergue, F., Pascal, S., Jetter, R., Renne, C., Faure, J. D., Haslam, R. P., Napier, J. A., Lessire, R., and Joubes, J. (2012) Reconstitution of plant alkane biosynthesis in yeast demonstrates that Arabidopsis ECERIFERUM1 and ECERIFERUM3 are core components of a very-long-chain alkane synthesis complex. Plant Cell 24, 3106-3118. 196. Aarts, M. G., Keijzer, C. J., Stiekema, W. J., and Pereira, A. (1995) Molecular characterization of the CER1 gene of arabidopsis involved in epicuticular wax biosynthesis and pollen fertility. Plant Cell 7, 2115-2127. 197. Cheesbrough, T. M., and Kolattukudy, P. E. (1984) Alkane biosynthesis by decarbonylation of aldehydes catalyzed by a particulate preparation from Pisum sativum. Proc Natl Acad Sci USA 81, 6613-6617. 198. Dennis, M. W., and Kolattukudy, P. E. (1991) Alkane biosynthesis by decarbonylation of aldehyde catalyzed by a microsomal preparation from Botryococcus braunii. Arch Biochem Biophys 287, 268-275. 199. Schneider-Belhaddad, F., and Kolattukudy, P. E. (2000) Solubilization, partial purification, and characterization of a fatty aldehyde decarbonylase from a higher plant, Pisum sativum. Arch Biochem Biophys 377, 341-349. 200. Rui, Z., Harris, N. C., Zhu, X., Huang, W., and Zhang, W. (2015) Discovery of a Family of Desaturase-Like Enzymes for 1-Alkene Biosynthesis. ACS Catalysis 5, 7091-7094. 201. Zhu, X., Su, M., Manickam, K., and Zhang, W. (2015) Bacterial Genome Mining of Enzymatic Tools for Alkyne Biosynthesis. ACS Chem Biol 10, 2785-2793. 202. Zhu, X., Liu, J., and Zhang, W. (2015) De novo biosynthesis of terminal alkyne-labeled natural products. Nat Chem Biol 11, 115-120. 203. Lee, M., Lenman, M., Banas, A., Bafor, M., Singh, S., Schweizer, M., Nilsson, R., Liljenberg, C., Dahlqvist, A., Gummeson, P. O., Sjodahl, S., Green, A., and Stymne, S.

70 (1998) Identification of non-heme diiron proteins that catalyze triple bond and epoxy group formation. Science 280, 915-918. 204. Dobrzyn, A., and Ntambi, J. M. (2005) The role of stearoyl-CoA desaturase in the control of metabolism. Prostaglandins Leukot Essent Fatty Acids 73, 35-41. 205. Dobrzyn, A., and Ntambi, J. M. (2005) Stearoyl-CoA desaturase as a new drug target for obesity treatment. Obes Rev 6, 169-174. 206. Hulver, M. W., Berggren, J. R., Carper, M. J., Miyazaki, M., Ntambi, J. M., Hoffman, E. P., Thyfault, J. P., Stevens, R., Dohm, G. L., Houmard, J. A., and Muoio, D. M. (2005) Elevated stearoyl-CoA desaturase-1 expression in skeletal muscle contributes to abnormal fatty acid partitioning in obese humans. Cell Metab 2, 251-261. 207. Stefan, N., Peter, A., Cegan, A., Staiger, H., Machann, J., Schick, F., Claussen, C. D., Fritsche, A., Haring, H. U., and Schleicher, E. (2008) Low hepatic stearoyl-CoA desaturase 1 activity is associated with fatty liver and insulin resistance in obese humans. Diabetologia 51, 648-656. 208. Brown, J. M., and Rudel, L. L. (2010) Stearoyl-coenzyme A desaturase 1 inhibition and the metabolic syndrome: considerations for future drug discovery. Curr Opin Lipidol 21, 192-197. 209. Attie, A. D. (2002) Relationship between stearoyl-CoA desaturase activity and plasma triglycerides in human and mouse hypertriglyceridemia. J Lipid Res 43, 1899-1907. 210. Morgan-Lappe, S. E., Tucker, L. A., Huang, X., Zhang, Q., Sarthy, A. V., Zakula, D., Vernetti, L., Schurdak, M., Wang, J., and Fesik, S. W. (2007) Identification of Ras- related nuclear protein, targeting protein for xenopus kinesin-like protein 2, and stearoyl- CoA desaturase 1 as promising cancer targets from an RNAi-based screen. Cancer Res 67, 4390-4398. 211. Powell, D. A. (2014) An overview of patented small molecule stearoyl coenzyme-A desaturase inhibitors (2009 - 2013). Expert Opin Ther Pat 24, 155-175. 212. Igal, R. A. (2011) Roles of StearoylCoA Desaturase-1 in the Regulation of Cancer Cell Growth, Survival and Tumorigenesis. Cancers 3, 2462-2477. 213. Dick, K. J., Eckhardt, M., Paisan-Ruiz, C., Alshehhi, A. A., Proukakis, C., Sibtain, N. A., Maier, H., Sharifi, R., Patton, M. A., Bashir, W., Koul, R., Raeburn, S., Gieselmann, V., Houlden, H., and Crosby, A. H. (2010) Mutation of FA2H underlies a complicated form of hereditary spastic paraplegia (SPG35). Hum Mutat 31, E1251-1260. 214. Edvardson, S., Hama, H., Shaag, A., Gomori, J. M., Berger, I., Soffer, D., Korman, S. H., Taustein, I., Saada, A., and Elpeleg, O. (2008) Mutations in the fatty acid 2-hydroxylase gene are associated with leukodystrophy with spastic paraparesis and dystonia. Am J Hum Genet 83, 643-648. 215. Pierson, T. M., Simeonov, D. R., Sincan, M., Adams, D. A., Markello, T., Golas, G., Fuentes-Fajardo, K., Hansen, N. F., Cherukuri, P. F., Cruz, P., Mullikin, J. C., Blackstone, C., Tifft, C., Boerkoel, C. F., Gahl, W. A., and Program, N. C. S. (2012) Exome sequencing and SNP analysis detect novel compound heterozygosity in fatty acid hydroxylase-associated neurodegeneration. Eur J Hum Genet 20, 476-479. 216. Blackstone, C. (2012) Cellular pathways of hereditary spastic paraplegia. Annu Rev Neurosci 35, 25-47. 217. Hallenbeck, P. (2012) Microbial Technologies in Advanced Biofuels Production, Springer, New York, NY. 218. Chen, H., Gu, Z., Zhang, H., Wang, M., Chen, W., Lowther, W. T., and Chen, Y. Q. (2013) Expression and purification of integral membrane fatty acid desaturases. PLoS One 8, e58139.

71 219. Goren, M. A., and Fox, B. G. (2008) Wheat germ cell-free translation, purification, and assembly of a functional human stearoyl-CoA desaturase complex. Protein Expr Purif 62, 171-178. 220. Park, J. H., Aravind, L., Wolff, E. C., Kaevel, J., Kim, Y. S., and Park, M. H. (2006) Molecular cloning, expression, and structural prediction of deoxyhypusine hydroxylase: a HEAT-repeat-containing metalloenzyme. Proc Natl Acad Sci USA 103, 51-56. 221. Han, Z., Sakai, N., Bottger, L. H., Klinke, S., Hauber, J., Trautwein, A. X., and Hilgenfeld, R. (2015) Crystal Structure of the Peroxo-diiron(III) Intermediate of Deoxyhypusine Hydroxylase, an Oxygenase Involved in Hypusination. Structure 23, 882-892. 222. Koehntop, K. D., Emerson, J. P., and Que, L., Jr. (2005) The 2-His-1-carboxylate facial triad: a versatile platform for dioxygen activation by mononuclear non-heme iron(II) enzymes. J. Biol. Inorg. Chem. 10, 87-93. 223. Hegg, E. L., and Que, L., Jr. (1997) The 2-His-1-carboxylate facial triad - An emerging structural motif in mononuclear non-heme iron(II) enzymes. Eur. J. Biochem. 250, 625- 629. 224. Hausinger, R. P. (2015) 2-Oxoglutarate-Dependent Oxygenases, The Royal Society of Chemistry. 225. Jasniewski, A. J., Engstrom, L. M., Vu, V. V., Park, M. H., and Que, L., Jr. (2016) X-ray absorption spectroscopic characterization of the diferric-peroxo intermediate of human deoxyhypusine hydroxylase in the presence of its substrate eIF5a. J Biol Inorg Chem 21, 605-618. 226. Martinho, M., Xue, G., Fiedler, A. T., Que, L., Jr., Bominaar, E. L., and Münck, E. (2009) Mössbauer and DFT study of the ferromagnetically coupled diiron(IV) precursor to a complex with an Fe(IV)(2)O(2) diamond core. J Am Chem Soc 131, 5823-5830. 227. Xue, G., De Hont, R., Münck, E., and Que, L., Jr. (2010) Million-fold activation of the [Fe(2)(micro-O)(2)] diamond core for C-H bond cleavage. Nat Chem 2, 400-405. 228. Daiyasu, H., Osaka, K., Ishino, Y., and Toh, H. (2001) Expansion of the zinc metallohydrolase family of the beta-lactamase fold. FEBS Lett 503, 1-6. 229. Schenk, G., Mitić, N., Hanson, G. R., and Comba, P. (2013) Purple acid phosphatase: A journey into the function and mechanism of a colorful enzyme. Coord Chem Rev 257, 473-482. 230. Gomes, C. M., Frazao, C., Xavier, A. V., Legall, J., and Teixeira, M. (2002) Functional control of the binuclear metal site in the metallo-beta-lactamase-like fold by subtle amino acid replacements. Protein Sci 11, 707-712. 231. Vicente, J. B., Carrondo, M. A., Teixeira, M., and Frazao, C. (2008) Structural studies on flavodiiron proteins. Globins and Other Nitric Oxide-Reactive Proteins, Part B 437, 3- 19. 232. Wilcox, D. E. (1996) Binuclear Metallohydrolases. Chem Rev 96, 2435-2458. 233. Guddat, L. W., McAlpine, A. S., Hume, D., Hamilton, S., de Jersey, J., and Martin, J. L. (1999) Crystal structure of mammalian purple acid phosphatase. Structure 7, 757-767. 234. Schenk, G., Elliott, T. W., Leung, E., Carrington, L. E., Mitic, N., Gahan, L. R., and Guddat, L. W. (2008) Crystal structures of a purple acid phosphatase, representing different steps of this enzyme's catalytic cycle. BMC Struct Biol 8, 6. 235. Kurtz, D. M., Jr. (2007) Flavo-diiron enzymes: nitric oxide or dioxygen reductases? Dalton Trans, 4115. 236. Vicente, J. B., Justino, M. C., Goncalves, V. L., Saraiva, L. M., and Teixeira, M. (2008) Biochemical, spectroscopic, and thermodynamic properties of flavodiiron proteins. Globins and Other Nitric Oxide-Reactive Proteins, Part B 437, 21-45.

72 237. Makris, T. M., Chakrabarti, M., Münck, E., and Lipscomb, J. D. (2010) A family of diiron monooxygenases catalyzing amino acid beta-hydroxylation in antibiotic biosynthesis. Proc Natl Acad Sci USA 107, 15391-15396. 238. Makris, T. M., Knoot, C. J., Wilmot, C. M., and Lipscomb, J. D. (2013) Structure of a dinuclear iron cluster-containing beta-hydroxylase active in antibiotic biosynthesis. Biochemistry 52, 6662-6671. 239. Vu, V. V., Makris, T. M., Lipscomb, J. D., and Que, L., Jr. (2011) Active-site structure of a beta-hydroxylase in antibiotic biosynthesis. J Am Chem Soc 133, 6938-6941. 240. Cameron, A. D., Ridderstrom, M., Olin, B., and Mannervik, B. (1999) Crystal structure of human glyoxalase II and its complex with a glutathione thiolester substrate analogue. Structure 7, 1067-1078. 241. Jasniewski, A. J., Knoot, C. J., Lipscomb, J. D., and Que, L., Jr. (2016) A Carboxylate Shift Regulates Dioxygen Activation by the Diiron Nonheme beta-Hydroxylase CmlA upon Binding of a Substrate-Loaded Nonribosomal Peptide Synthetase. Biochemistry. 242. Charalampous, F. C. (1959) Biochemical studies on inositol. V. Purification and properties of the enzyme that cleaves inositol to D-glucuronic acid. J Biol Chem 234, 220-227. 243. Charalampous, F. C. (1960) Biochemical studies on inositol. VI. Mechanism of cleavage of inositol to D-glucuronic acid. J Biol Chem 235, 1286-1291. 244. Charalampous, F. C., and Abrahams, P. (1957) Biochemical studies on inositol. I. Isolation of myo-inositol from yeast and its quantitative enzymatic estimation. J Biol Chem 225, 575-583. 245. Charalampous, F. C., and Lyras, C. (1957) Biochemical studies on inositol. IV. Conversion of inositol to glucuronic acid by rat kidney extracts. J Biol Chem 228, 1-13. 246. Reddy, C. C., Pierzchala, P. A., and Hamilton, G. A. (1981) myo-Inositol oxygenase from hog kidney. II. Catalytic properties of the homogeneous enzyme. J Biol Chem 256, 8519-8524. 247. Reddy, C. C., Swan, J. S., and Hamilton, G. A. (1981) myo-Inositol oxygenase from hog kidney. I. Purification and characterization of the oxygenase and of an enzyme complex containing the oxygenase and D-glucuronate reductase. J Biol Chem 256, 8510-8518. 248. Bollinger, J. M., Jr., Diao, Y., Matthews, M. L., Xing, G., and Krebs, C. (2009) myo- Inositol oxygenase: a radical new pathway for O2 and C-H activation at a nonheme diiron cluster. Dalton Trans, 905-914. 249. Aravind, L., and Koonin, E. V. (1998) The HD domain defines a new superfamily of metal-dependent phosphohydrolases. Trends Biochem Sci 23, 469-472. 250. Brown, P. M., Caradoc-Davies, T. T., Dickson, J. M., Cooper, G. J., Loomes, K. M., and Baker, E. N. (2006) Crystal structure of a substrate complex of myo-inositol oxygenase, a di-iron oxygenase with a key role in inositol metabolism. Proc Natl Acad Sci USA 103, 15032-15037. 251. Xing, G., Barr, E. W., Diao, Y., Hoffart, L. M., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) Oxygen activation by a mixed-valent, diiron(II/III) cluster in the glycol cleavage reaction catalyzed by myo-inositol oxygenase. Biochemistry 45, 5402-5412. 252. Xing, G., Hoffart, L. M., Diao, Y., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) A coupled dinuclear iron cluster that is perturbed by substrate binding in myo-inositol oxygenase. Biochemistry 45, 5393-5401. 253. Fox, B. G., Surerus, K. K., Münck, E., and Lipscomb, J. D. (1988) Evidence for a Mu- Oxo-Bridged Binuclear Iron Cluster in the Hydroxylase Component of Methane

73 Monooxygenase - Mossbauer and Electron-Paramagnetic-Res Studies. J Biol Chem 263, 10553-10556. 254. Fox, B. G., Hendrich, M. P., Surerus, K. K., Andersson, K. K., Froland, W. A., Lipscomb, J. D., and Münck, E. (1993) Mossbauer, Epr, and Endor Studies of the Hydroxylase and Reductase Components of Methane Monooxygenase from Methylosinus-Trichosporium Ob3b. J Am Chem Soc 115, 3688-3701. 255. Woodland, M. P., Patil, D. S., Cammack, R., and Dalton, H. (1986) Electron-Spin- Resonance Studies of Protein-a of the Soluble Methane Monooxygenase from Methylococcus-Capsulatus (Bath). Biochim Biophys Acta 873, 237-242. 256. Snyder, R. A., Bell, C. B., 3rd, Diao, Y., Krebs, C., Bollinger, J. M., Jr., and Solomon, E. I. (2013) Circular dichroism, magnetic circular dichroism, and variable temperature variable field magnetic circular dichroism studies of biferrous and mixed-valent myo- inositol oxygenase: insights into substrate activation of O2 reactivity. J Am Chem Soc 135, 15851-15863. 257. Xing, G., Diao, Y., Hoffart, L. M., Barr, E. W., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) Evidence for C-H cleavage by an iron- superoxide complex in the glycol cleavage reaction catalyzed by myo-inositol oxygenase. Proc Natl Acad Sci USA 103, 6130-6135. 258. Hirao, H., and Morokuma, K. (2009) Insights into the (superoxo)Fe(III)Fe(III) intermediate and reaction mechanism of myo-inositol oxygenase: DFT and ONIOM(DFT:MM) study. J Am Chem Soc 131, 17206-17214. 259. Shan, X., and Que, L., Jr. (2005) Intermediates in the oxygenation of a nonheme diiron(II) complex, including the first evidence for a bound superoxo species. Proc Natl Acad Sci USA 102, 5340-5345. 260. Bollinger, J. M., Jr., and Krebs, C. (2007) Enzymatic C-H activation by metal-superoxo intermediates. Curr Opin Chem Biol 11, 151-158. 261. van der Donk, W. A., Krebs, C., and Bollinger, J. M., Jr. (2010) Substrate activation by iron superoxo intermediates. Curr Opin Struct Biol 20, 673-683. 262. Tamanaha, E., Zhang, B., Guo, Y., Chang, W.-c., Barr, E. W., Xing, G., St Clair, J., Ye, S., Neese, F., Bollinger, J. M., Jr., and Krebs, C. (2016) Spectroscopic Evidence for the Two C-H-Cleaving Intermediates of Aspergillus nidulans Isopenicillin N Synthase. J Am Chem Soc 138, 8862-8874. 263. Peck, S. C., Wang, C., Dassama, L. M. K., Zhang, B., Guo, Y. S., Rajakovich, L. J., Bollinger, J. M., Jr., Krebs, C., and van der Donk, W. A. (2017) O-H Activation by an Unexpected Ferryl Intermediate during Catalysis by 2-Hydroxyethylphosphonate Dioxygenase. J Am Chem Soc 139, 2045-2052. 264. McSorley, F. R., Wyatt, P. B., Martinez, A., DeLong, E. F., Hove-Jensen, B., and Zechel, D. L. (2012) PhnY and PhnZ comprise a new oxidative pathway for enzymatic cleavage of a carbon-phosphorus bond. J Am Chem Soc 134, 8364-8367. 265. Wörsdörfer, B., Lingaraju, M., Yennawar, N. H., Boal, A. K., Krebs, C., Bollinger, J. M., Jr., and Pandelia, M.-E. (2013) Organophosphonate-degrading PhnZ reveals an emerging family of HD domain mixed-valent diiron oxygenases. Proc Natl Acad Sci USA 110, 18874-18879. 266. van Staalduinen, L. M., McSorley, F. R., Schiessl, K., Seguin, J., Wyatt, P. B., Hammerschmidt, F., Zechel, D. L., and Jia, Z. (2014) Crystal structure of PhnZ in complex with substrate reveals a di-iron oxygenase mechanism for catabolism of organophosphonates. Proc Natl Acad Sci USA 111, 5171-5176. 267. Bellini, D., Caly, D. L., McCarthy, Y., Bumann, M., An, S. Q., Dow, J. M., Ryan, R. P., and Walsh, M. A. (2014) Crystal structure of an HD-GYP domain cyclic-di-GMP

74 phosphodiesterase reveals an enzyme with a novel trinuclear catalytic iron centre. Mol Microbiol 91, 26-38. 268. Lovering, A. L., Capeness, M. J., Lambert, C., Hobley, L., and Sockett, R. E. (2011) The structure of an unconventional HD-GYP protein from Bdellovibrio reveals the roles of conserved residues in this class of cyclic-di-GMP phosphodiesterases. MBio 2, e00163- 00111. 269. Benda, C., Ebert, J., Scheltema, R. A., Schiller, H. B., Baumgartner, M., Bonneau, F., Mann, M., and Conti, E. (2014) Structural model of a CRISPR RNA-silencing complex reveals the RNA-target cleavage activity in Cmr4. Mol Cell 56, 43-54. 270. Jung, T. Y., An, Y., Park, K. H., Lee, M. H., Oh, B. H., and Woo, E. (2015) Crystal structure of the Csm1 subunit of the Csm complex and its single-stranded DNA-specific nuclease activity. Structure 23, 782-790. 271. Subedi, B. P., Corder, A. L., Zhang, S., Foss, F. W., Jr., and Pierce, B. S. (2015) Steady- state kinetics and spectroscopic characterization of enzyme-tRNA interactions for the non-heme diiron tRNA-monooxygenase, MiaE. Biochemistry 54, 363-376. 272. Liao, R.-Z., and Siegbahn, P. E. M. (2015) Mechanism and selectivity of the dinuclear iron benzoyl-coenzyme A epoxidase BoxB. Chem Sci 6, 2754-2764. 273. Rather, L. J., Weinert, T., Demmer, U., Bill, E., Ismail, W., Fuchs, G., and Ermler, U. (2011) Structure and mechanism of the diiron benzoyl-coenzyme A epoxidase BoxB. J Biol Chem 286, 29241-29248. 274. Teufel, R., Friedrich, T., and Fuchs, G. (2012) An oxygenase that forms and deoxygenates toxic epoxide. Nature 483, 359-362. 275. Behan, R. K., and Lippard, S. J. (2010) The aging-associated enzyme CLK-1 is a member of the carboxylate-bridged diiron family of proteins. Biochemistry 49, 9679-9681. 276. Lu, T. T., Lee, S. J., Apfel, U. P., and Lippard, S. J. (2013) Aging-associated enzyme human clock-1: substrate-mediated reduction of the diiron center for 5- demethoxyubiquinone hydroxylation. Biochemistry 52, 2236-2244. 277. Stenmark, P., Grunler, J., Mattsson, J., Sindelar, P. J., Nordlund, P., and Berthold, D. A. (2001) A new member of the family of di-iron carboxylate proteins. Coq7 (clk-1), a membrane-bound hydroxylase involved in ubiquinone biosynthesis. J Biol Chem 276, 33297-33300. 278. Lo, F. C., Hsieh, C. C., Maestre-Reyna, M., Chen, C. Y., Ko, T. P., Horng, Y. C., Lai, Y. C., Chiang, Y. W., Chou, C. M., Chiang, C. H., Huang, W. N., Lin, Y. H., Bohle, D. S., and Liaw, W. F. (2016) Crystal Structure Analysis of the Repair of Iron Centers Protein YtfE and Its Interaction with NO. Chemistry 22, 9768-9776. 279. Todorovic, S., Justino, M. C., Wellenreuther, G., Hildebrandt, P., Murgida, D. H., Meyer- Klaucke, W., and Saraiva, L. M. (2008) Iron-sulfur repair YtfE protein from Escherichia coli: structural characterization of the di-iron center. J Biol Inorg Chem 13, 765-770.

75 Chapter 2

A cyanobacterial hydrocarbon production pathway employing a non-heme diiron oxygenase

The earth's finite reservoirs of hydrocarbon fuels and the ever-increasing worldwide energy demand necessitate the development of novel renewable energy sources. An attractive solution to this dilemma is industrial fuel production via biological hosts. It has long been known that insects, plants, and fungi naturally produce hydrocarbons.(1, 2) Many insect pheromones are hydrocarbon molecules derived from fatty acid precursors.(3-5) The decarboxylation of fatty acids is mediated by a P450 enzyme, producing CO2 as a by-product,(5) which would be an environmental cost in large-scale industrial applications. In plants, alkanes constitute up to 70 % of cuticular wax total mass, serving to protect the plants from drought, insects, pathogens, and other environmental hazards.(6) The alkanes in plants were recognized decades ago to originate from abundant cellular fatty acids, but it was not until 2012 that the genes cer3 and cer1 from

Arabidopsis thaliana, encoding for the enzymes fatty acyl-ACP reductase (CER3) and aldehyde

“decarbonylase” (CER1), were confirmed to constitute a pathway for metabolism of very long chain fatty acids (Cn = 28-34) to Cn-1 alkanes.(6) Unfortunately, these two proteins are membrane- bound, which poses challenges for studying them biochemically and for bioengineering optimization.

In 2010, prior to the discovery of the plant enzymes responsible for this pathway, an analogous pathway was identified in cyanobacteria for converting long-chain fatty acids to alkanes (Figure 2-1).(2) This discovery was enticing because these microbes are both photosynthetic and genetically alterable, rendering them potential vehicles for inexpensive bioprocesses to generate renewable diesel fuels. Additionally, the two enzymes involved in this biological process are soluble, cytosolic proteins, making them more amenable to structural and

76 mechanistic studies that could facilitate engineering of new biocatalysts. Indeed, biotechnology companies are actively pursuing processes employing these cyanobacterial enzyme components, but quickly have encountered limitations imposed by the modest catalytic efficiencies of the enzymes. Thus, we became motivated to dissect the enzymes’ mechanisms in order to understand and overcome the inherent limitations hindering large-scale hydrocarbon production.

Figure 2-1. Cyanobacterial metabolic pathway for production of long-chain hydrocarbons. n = 11, 13, and 15.

The gene products responsible for this metabolic pathway in cyanobacteria were ascertained through a comparative genomics study.(2) The genomes of ten cyanobacterial strains that produce long-chain alkanes were compared with that of a single non-producing organism,

Synechococcus sp. PCC7002. By identifying genes common amongst the former group and noting their absence in the non-producer, two open reading frames were identified as likely candidates to constitute this pathway and were subsequently confirmed as essential and sufficient for alkane production.(2) The first enzyme of the pathway, acyl-acyl carrier protein reductase

(AAR), utilizes the abundant cellular pool of long-chain saturated and unsaturated Cn-fatty acids

(n = 16, 18, 20), covalently attached via a thioester linkage to an acyl carrier protein (ACP), as

77 substrates (Figure 2-1).(2, 7) AAR catalyzes an NADPH-dependent reduction of the acyl-ACP substrate to produce the corresponding free Cn-fatty aldehyde.(2, 7) AAR forms a stable heterodimeric complex with the second enzyme, effectively channeling the fatty aldehyde intermediate product for its further conversion to a Cn-1-alk(a/e)ne.(7) The second enzyme was originally presumed to generate carbon monoxide (CO) as the co-product in a formally hydrolytic reaction to make alkanes, and was consequently denoted as an aldehyde decarbonylase

(“AD”).(2)

However, there were many inconsistent observations that cast doubt on this initial assignment of the second enzymatic conversion in the pathway. First, this assertion was primarily founded upon analogy to the plant system that was believed to produce CO in its final reaction,(8-

10) despite a lack of biochemical characterization of the enzyme responsible for alkane production. Second, a crystal structure of Prochlorococcus marinus “AD” had been fortuitously reported by the Joint Center for Structural Genomics, prior to its identification and biochemical characterization. This structure, in addition to primary sequence similarity, revealed that the enzyme shares high homology with members of the ferritin-like superfamily of diiron carboxylate-bridged oxidases and oxygenases. Indeed, the crystallographic data demonstrated the presence of two iron ions bound in its active site.(2) Finally, the initial study reported that this enzyme requires a reductant to achieve multiple turnovers.(2) These last two observations would a priori suggest a redox reaction, and thus, were difficult to rationalize with the assertion that

“AD” catalyzes a hydrolytic reaction to produce CO as the by-product in the alkane-forming reaction.

Seminal studies rapidly following the discovery of these two enzymes aimed to define the chemical reaction of “AD” in order to reconcile these incongruities. At the start, two additional hypotheses were put forth for the conversion of the aldehyde substrate to an alkane that could rationalize the need for a reducing system and a redox-active cofactor. Both proposals invoked

78 the consumption of O2 as a co-substrate in the reaction, where one possibility postulated that CO2 would be generated as the by-product,(11) while the other alternative would result in formate

(HCOO-).(12) In vitro biochemical activity experiments showed that formate is produced in a 1:1 stoichiometry with the alkane product, firmly establishing the overall chemical transformation catalyzed by “AD”.(13) Following this work, yet another proposal was put forth positing that

“AD” could produce formate via O2-independent hydrolysis of its fatty aldehyde substrate to alkane, thereby incorporating an oxygen atom from water into the formate co-product.(14, 15)

However, it was definitively determined that “AD” strictly requires O2 as a co-substrate for

18 turnover by employing isotopically-labelled dioxygen ( O2) and experimental observation of one oxygen atom from O2 incorporated into the formate co-product(12, 16). Consequently, “AD” was renamed aldehyde-deformylating oxygenase (ADO) to properly reflect its enzymatic conversion.

These studies revealed the redox nature of ADO catalysis, wherein molecular oxygen is consumed as a co-substrate and reduced by four electrons (formally to the oxidation states of two water molecules), despite the redox-neutral conversion of the aldehyde substrate to a saturated hydrocarbon. This conclusion perfectly rationalized the initial observations reported by Schirmer et al. that, (i) ADO is a member of the FDCOO superfamily, harboring a non-heme diiron cofactor capable of redox chemistry, and (ii) ADO requires reducing equivalents for catalysis. By analogy to other FDCOOs, the diiron cofactor of ADO was presumed to initiate activation of O2, which would enable the C-C bond cleavage of the fatty aldehyde substrate. However, to achieve this formally redox-neutral chemical conversion, the ADO mechanism must diverge drastically from those of traditional FDCOOs that enact oxidation reactions. As covered in Chapter 1, canonical FDCOOs couple the four-electron reduction of O2 with the two-electron oxidation of their substrate and the overall two-electron oxidation of the diiron cofactor. An auxiliary reductant acts to reduce the final diferric state by two electrons after substrate conversion is completed. In contrast, the ADO aldehyde substrate does not become oxidized and thus, ADO

79 requires a total of four electrons from a reducing partner to complete each catalytic cycle, implying that two electrons must be delivered during substrate conversion. Mechanistic studies were then designed and carried out with the goal of elucidating the points of divergence from the established mechanistic steps of FDCOOs that enable ADO to effect its unique chemical transformation.

First, the nature of the enzyme metallocofactor and its role in catalysis were investigated by spectroscopic techniques to confirm the involvement of a non-heme diiron cofactor in ADO catalysis as was suggested by its crystal structure. The Mössbauer spectrum of the as-isolated enzyme showed evidence for an antiferromagnetically(AF)-coupled diferric cluster with a diamagnetic ground state.(17) Chemical reduction of the diferric cluster yielded a diferrous state that would slowly reoxidize to the diferric form upon exposure to O2.(17) In the presence of the aldehyde substrate, however, the diferrous ADO form reacted more rapidly with O2 to produce a product μ-oxo-diferric species that is distinct from the as-isolated form of the cofactor.(17)

During this reaction with substrate, a transient diferric species accumulated that displays distinct absorption and Mössbauer spectroscopic features from those of both the as-isolated and final product states.(17) Substrate analogs (e.g., alcohols, fatty acids, and alkanes) did not stimulate reaction with O2, nor did they result in formation of the transient diferric species.(17) This triggering effect of substrate binding on O2 activation has been observed for other diiron enzymes, as well as metalloenzymes of different classes such as the mononuclear iron and 2- oxoglutarate dioxygenases,(18) and supported the assertion that the ADO diiron center is the relevant cofactor in oxygenative catalysis. Decay of the transient species that accumulates in the reaction with O2 was directly correlated with product formation in a kinetically competent fashion,(17) indicating that this species is a catalytic intermediate and thereby established the essential, catalytic role of the diiron cofactor.

80 Characterization of this intermediate by Mössbauer spectroscopy determined it to be an

AF-coupled diferric species,(17) arising from two-electron reduction of the O2 co-substrate to a peroxide moiety. While this O2-activation step mirrors the standard first step of known FDCOOs, the spectroscopic features of this ADO diferric-peroxide-like species are distinct from those of other FDCOOS that are proposed to have μ-1,2- and μ-1,1-(hydro)peroxo geometric structures as discussed in Chapter 1 (Table 1-1). Its absorption maximum is blue-shifted (λmax = 450 nm) and its Mössbauer spectrum contains two site differentiated quadrupole doublets, suggestive of distinct coordination environments for each of the high-spin ferric ions in the cluster.(17) We postulate, supported by recent unpublished EXAFS and rR data, that the ADO intermediate has a diferric-peroxyhemiacetal (PHA) structure. This species would result from nucleophilic attack by the peroxide moiety on the substrate carbonyl, analogous to the mechanisms of nucleophilic flavin-C4a-(hydro)peroxides and P450s, as discussed in Chapter 1.[refs] Formation of a covalent adduct could explain its slow kinetics of formation and the extended lifetime of this intermediate in the absence of reductant. Additionally, the observed stability of this intermediate could be adventitious for the enzyme, which requires exogenous electrons for catalysis to proceed.

Based on these results, a mechanistic proposal was set forth for ADO catalysis (Figure 2-

II/II 2). With substrate bound at site 2 as shown in crystal structures, the Fe2 cofactor (I) is primed for reaction with O2. Terminal binding of dioxygen at site 1 generates an O2 adduct (II) that would be in rapid equilibrium with the unbound form of the ADO•aldehyde complex (I). This initial step is analogous to that proposed for sMMO, for which the diferrous-O2 adduct state has been termed intermediate P*. Single electron transfer from the ferrous iron at site 1 to the dioxygen moiety generates a non-observable (i.e., kinetically-silent) ferric-superoxide species

II/III (III). Rapid transfer of a second electron from the mixed-valent Fe2 cluster results in a

III/III terminal peroxo-Fe2 state (IV). Attack of the peroxide moiety on the aldehyde carbonyl forms

III/III the stable Fe2 -peroxyhemiacetal (PHA) intermediate (V) observed experimentally. Reductive

81 cleavage of the O-O bond in V generates a gem-diolyl substrate radical (VI) that undergoes β- scission of the C1-C2 bond, forming a C2-alkyl radical intermediate (R•) and the C1-derived co- product, formate (VII). Abstraction of a hydrogen atom (H•) by the alkyl radical from either an amino acid donor in the active site or a coordinated solvent ligand would form the product

III/IV hydrocarbon along with either an amino acid radical or an Fe2 form of the cofactor (VIII).

Subsequent transfer of a second electron from the reducing system (and probably a proton) and

III/III alkane release would result in the formate-bound Fe2 state (IX). Formate dissociation, two-

III/III II/II electron reduction of the product µ-oxo-Fe2 cofactor back to the Fe2 state, and binding of another aldehyde molecule would prepare the enzyme for the next turnover (I).

Figure 2-2. Scheme depicting the free-radical mechanism proposed for ADO catalysis.

Studies interrogating the O2 activation steps with the native aldehyde substrates and substrate analogues provide strong evidence for states I, II, and V. Experiments examining the reduction of state V employing chemical reductants have demonstrated alkane and formate product formation, concomitant with formation of a μ-oxo-diferric product state (state IX).

However, using chemical reductants, the kinetics of electron transfer events to states V and IX

82 are slow, and the product yields are lower than theoretically expected.(17) Finally, there is no direct evidence for any of the intermediate states postulated to follow reduction of state V. This latter half of the proposed mechanism, described as a free-radical mechanism, invokes two mechanistically challenging aspects: (i) delivery of four e– during the catalytic cycle (two during

II/II substrate conversion and two to regenerate the O2-reactive Fe2 cofactor), and (ii) quenching of the alkane precursor, an alkyl radical intermediate. These features would be potential vulnerabilities of the enzyme that might challenge enzymatic efficiency. This chapter describes experimental validation of these steps in catalysis and their overall effect on enzymatic hydrocarbon production.

2.1 Exploring physiological reducing partners for ADO catalysis

The reducing partner for ADO catalysis plays an integral role in substrate conversion itself, in addition to its role in regenerating the O2-reactive diferrous state of the cofactor at the end of the substrate conversion half-cycle. Informative in vitro studies examining the reduction events have employed a small molecule mediator, specifically phenazine methosulfate or 1- methoxy-5-methylphenazinium methosulfate (MeOPMS), chemically reduced by NADH or sodium dithionite, to study the ADO reaction.(12, 14, 17) The phenazine system was shown to support multiple turnovers of the enzyme, but only afforded half the theoretical yield of products with respect to reducing equivalents.(17) This inefficiency was attributed to uncoupled or improperly timed electron delivery to the diiron site by the two-electron reductant. While these artificial chemical reductants have been invaluable tools in elucidating the nature of the ADO reaction and dissecting its mechanism,(12, 14, 17) they are not viable candidates to serve as the reductant for

ADO in a bioprocess producing renewable hydrocarbon fuels. They are kinetically inefficient, do not support robust, stable activity over many turnovers,(14, 16) and would present cost and

83 toxicity issues if added to cultures.(19) A biological electron donor would seem to be an essential component of any such bioprocess.

A priori, the physiological reductant could be either a small biomolecule or a cytosolic redox protein. The genes encoding ADO and AAR are frequently co-transcribed, but there is no other gene associated with this operon that could encode for a protein to serve as a specific electron donor to ADO.(20) Yet, there is substantial precedent for [2Fe-2S] ferredoxins (Fds) serving this function for other FDCOOs. Soluble methane monooxygenase (sMMO) employs an iron-sulfur, flavin-containing reductase (sMMOR) that uses NADH to reduce FAD and transfers the two electrons, one at a time, via a plant-type [2Fe-2S] cluster to the non-heme diiron cofactor in the hydroxylase component (sMMOH).(21, 22) Other bacterial multicomponent monooxygenases (e.g., toluene/o-xylene monooxygenase) have reductase components analogous to sMMOR to service their non-heme diiron hydroxylase components.(23) In the Ec class Ia

III/III ribonucleotide reductase (RNR), the [2Fe-2S] Fd, YfaE, is required to reduce the µ-oxo-Fe2

II/II cluster in the inactive "met" form of the β subunit to the O2-reactive Fe2 state to support regeneration of the catalytically essential tyrosyl radical in a process referred to as the

“maintenance pathway”.(24) In stearoyl-ACP desaturase (Δ9D), activity depends on an abundant, general plant-type [2Fe-2S] Fd.(25, 26)

These precedents suggest that the in vivo reductant for ADO might also be a ferredoxin.

In previous studies, the heterologous spinach ferredoxin/ferredoxin-NADP+ reductase system with NADPH (Fd/FNR/N) system was found to support very few ADO turnovers in vitro.(13, 27)

However, improved activity was demonstrated for Synechococcus elongatus PCC7942 (Se) ADO with its cognate Fd/FNR/N system and in the context of a Se ADO-Fd-FNR fusion protein.(27,

28) The latter studies suggested the possibility that this system might be a physiologically relevant ADO reductant, but no in-depth analysis of electron delivery was reported.

Consequently, details of how ADO and its reducing partner ensure the orderly transfer of a total

84 of four electrons at multiple points in the catalytic cycle have not yet been elucidated, a significant gap in our understanding of the ADO mechanism.

2.1.1 Selection of the cyanobacterial [2Fe-2S] ferredoxin, PetF, as a physiologically-relevant reducing partner

The genome of the cyanobacterium Synechocystis sp. PCC6803 (Syn. 6803) contains seven genes encoding putative Fds ([2Fe-2S]- or [4Fe-4S]-containing) and one encoding an FNR

(FAD-containing).(29) All seven Syn. 6803 Fds were heterologously expressed, purified and assessed in vitro with the recombinantly-produced Syn. 6803 FNR for the capacity to support multiple turnovers by Nostoc punctiforme (Np) ADO. Coupled AAR/ADO reactions were carried out to monitor conversion of the stearoyl-ACP substrate to the octadecanal intermediate and finally to the heptadecane product. After a standard incubation of 1 h, the highest heptadecane yield was observed in the presence of the [2Fe-2S] cluster-containing Fd protein encoded by the

Syn. 6803 ssl0020 gene, also known as PetF or FdI (Figure 2-3, purple). When PetF was employed, octadecanal was not detected, indicating that all of the aldehyde intermediate produced by AAR was subsequently converted by ADO to heptadecane. Conversely, reactions performed with the spinach Fd/FNR system led to lesser heptadecane yields (Figure 2-3, blue) and accumulation of the aldehyde intermediate, in agreement with previous reports by us(13) and others(27) that this system supports only modest ADO activity.

PetF is the most abundant of the Syn. 6803 Fds and is essential for viability, owing to its role as an electron acceptor in photosynthesis.(29, 30) PetF was reported to successfully support in vivo propane production in an engineered microbial platform, employing Prochlorococcus marinus AAR/ADO in Ec as a host.(31, 32) In addition, its closest ortholog in Se (75% sequence identity) was shown to afford faster ADO turnover in vitro than the phenazine system.(27) The cellular abundance of PetF, together with the results demonstrating its suitability as an electron donor to ADO, suggest that it might be the physiological donor and provide the rationale for the use of this one-electron reducing system to probe details of the ADO mechanism that have not been accessible with chemical two-electron donors.

Syn. 6803 PetF was characterized by spectroscopic techniques in order to establish baseline parameters to assess its electron donating abilities in reaction with ADO. First, evidence for an FeS cluster following protein expression and isolation was sought. Based on sequence,

PetF is expected to harbor a 4-cysteine ligated [2Fe-2S] cluster, which would have characteristic absorption features at 330, 423, and 465 nm in the oxidized [2Fe-2S]2+ cluster form. As-isolated

86 PetF demonstrates these expected chromophores, which decrease in intensity upon chemical reduction to the [2Fe-2S]1+ cluster form by sodium dithionite (Figure 2-4A). Re-oxidation of the reduced PetF can be achieved by exposure to molecular oxygen and can be monitored by the increase of the 423 nm feature over time. Fitting a single exponential function to the data gives

-1 parameters of ΔA423 = 0.18 and kobs = 0.014 s for cofactor oxidation (Figure 2-4B).

Figure 2-4. Characterization of the PetF metallocofactor by UV-visible absorption spectroscopy. (A) Reference absorption spectra of [2Fe-2S]1+ (black trace) and [2Fe-2S]2+ (red trace) forms of PetF were obtained from a single-mixing SF-Abs experiment, in which sodium dithionite-reduced PetF (0.10 mM) was mixed with an equal volume of either O2-free or O2-saturated 50 mM sodium HEPES, pH 7.5 buffer (~1.8 mM O2), respectively, for 250 s at 5 °C. (B) SF-Abs kinetic trace monitoring oxidation of the [2Fe-2S]1+ PetF at 423 nm (blue trace) in the single-mixing experiment described in A. The black dashed line is the fit described in the text.

The reduced [2Fe-2S]1+ cluster has an S = ½ total spin state and thus, is amenable to EPR spectroscopy. The spectrum for the reduced PetF protein demonstrates a rhombic signal with principal g-values of 2.05, 1.95, and 1.88, indicative of a metal-based paramagnetic species and characteristic of [2Fe-2S]1+ clusters (Figure 2-5). Spin quantitation relative to a Cu2+-EDTA standard determined the presence of ~0.6 spins/PetF. Considering that excess sodium dithionite was used to prepare the reduced state, the sub-stoichiometric spin per protein likely corresponds to incomplete incorporation of a cluster into all protein molecules rather than incomplete reduction of the clusters.

2.1.2 Employment of PetF as a reducing system to support ADO catalysis under single turnover conditions

The relevant and accessible redox couple for the majority of [2Fe-2S] clusters that participate in biological electron transfer reactions is the one-electron transition from the [2Fe-

1+ 2+ 2S] to the [2Fe-2S] state. Since two electrons are required for productive reduction of the

III/III Fe2 -PHA intermediate to formate and alkane, the theoretical reaction stoichiometry required would be 2 PetF proteins (or 2 electrons) for every 1 molecule of formate or alkane. To determine the experimental stoichiometry, the C1-derived product, formate, was quantified in manual chemically quenched samples prepared by treatment of the pre-formed intermediate with varying molar ratios of a reductant. With the chemically reduced PetF as reductant, a stoichiometry of

0.58 equivalents of formate per two equivalents of PetF (two electrons) was obtained (Figure 2-6, blue). This value deviates from the theoretical stoichiometry, demonstrating an uncoupling of electron delivery and product formation, as previously observed with the phenazine reductant.(17) In contrast, when FNR was included in equimolar quantity with PetF, ADO, and

88 the two-electron donor NADPH as the electron source, the experimental stoichiometry increased to 0.96 formate per 2 electrons (Figure 2-6, black), very close to the theoretical value. Thus, the

III/III reduction of the Fe2 -PHA intermediate via the PetF/FNR/N system is tightly coupled to product formation. In the latter case, the two electrons from NADPH are transferred to the FAD cofactor of FNR, and then shuttled sequentially to the PetF [2Fe-2S] cluster, which finally delivers the electrons to the diiron site of ADO one at a time. The enhanced catalytic performance of ADO with the PetF/FNR/N system may result from well-timed electron delivery and avoidance of reductive quenching of reactive intermediates. More controlled delivery of electrons to the ADO diiron site might also help prevent reduction of O2 to H2O2, reported to be an inhibitor of ADO.(27, 33)

Figure 2-6. Formate production in single turnover ADO assays utilizing different reducing III/III systems. The accumulated Fe2 -PHA intermediate (Figure 2-24) was reacted for ~ 1 s with varying molar ratios of reductant: either chemically reduced PetF (blue) or NADPH (black) in the presence of oxidized PetF and oxidized FNR in equimolar ratios with ADO (0.050 mM final). Data are averages of at least three replicate assays.

III/III 2.1.3 Reduction of the Fe2 ADO by PetF, monitored by stopped-flow absorption spectroscopy (SF-Abs)

The physiological reducing partner to ADO must provide electrons for product formation, as described in the previous section, but also provide two electrons to reduce the product state of

89 the cofactor back to its O2-reactive form for another turnover, acting in the traditional role of the

III/III reducing system for FDCOOs. Reaction of reduced PetF with the Fe2 ADO was examined by

SF-Abs spectroscopy, monitoring the oxidation of PetF (Figure 2-7), which has more intense chromophores (λ = 423 nm) than the weak ligand to metal charge transfer bands of the μ-oxo-

III/III III/III Fe2 ADO cofactor. Complete oxidation of PetF was observed upon reaction with Fe2 -ADO at all ratios of PetF:ADO. The observed rate constant for PetF oxidation demonstrated a dependence on the ratio of ADO:PetF, giving a second order rate constant of 8.1 mM-1∙s-1 using a linear fita for the bimolecular reaction with ADO. The range of observed rate constants (at 5 °C)

III/III is similar to those reported for reduction of the Fe2 forms of other FDCOOs by their [2Fe-2S]

-1 -1 9 -1 Fd partners (RNR-β2 kobs1 ~ 4-5 s and kobs2 ~ 1-2 s at 37 °C; Δ D kobs = 3.4 s at 25 °C;

-1 -1 sMMOH with the sMMOR Fd domain kobs1 = 1 s and kobs2 = 0.2 s at 4 °C).(22, 24, 25)

Considering that the fastest reported multiple-turnover rate for ADO with the PetF/FNR/N reducing system is 0.007 s-1 (at 37 °C),(27) reduction of the diferric state(s) by PetF cannot be rate-limiting. The observed rate constant for reduction of diferric ADO by the reduced phenazine

III/III is almost 100 times less.(17) PetF is thus the most efficient electron donor to Fe2 -ADO reported to date.

a The data could also be reasonably fit with a hyperbola, giving an apparent second order rate -1 -1 constant, kchem/Kd, of 1.3 mM ∙s . However, the experimental data do not reach sufficient saturation to confidently report a rate for the kchem.

III/III Figure 2-7. Kinetics of PetF oxidation by the as-isolated Fe2 -ADO. (A) SF-Abs kinetic traces monitoring PetF oxidation at 423 nm in single-mix experiments in which a solution of O2-free, chemically reduced PetF ([final] = 0.050 mM) was mixed with an equal volume of an O2-free III/III solution of as-isolated Fe2 -ADO at varying concentrations to give the final ratios indicated in the figure legend. The data were best fit by two exponential functions with the second observed rate constant (kobs2) being concentration dependent. (B) Plot of kobs2 as a function of ADO concentration, from which a second-order rate constant of 8.1 mM-1∙s-1 was obtained.

III/III 2.1.4 Rapid reduction of the ADO Fe2 -PHA intermediate by PetF, monitored by SF-Abs

III/III The kinetics of the reaction of the Fe2 -PHA intermediate with reduced PetF were

II/II monitored by sequential-mixing SF-Abs spectroscopy (Figure 2-8). In the first mix, the Fe2 -

ADO•aldehyde complex was reacted with O2-saturated buffer to permit maximal accumulation of

III/III the Fe2 -PHA intermediate (Figure 2-24),(17) which was then reacted in a second mix with chemically reduced PetF. Time-dependent absorption spectra obtained after the second mix reflect, primarily, oxidation of the [2Fe-2S]1+ form of PetF to the [2Fe-2S]2+ form (Figure 2-8A).

The ADO concentration was varied at a constant PetF concentration in order to delineate the

1+ III/III kinetics of oxidation of the [2Fe-2S] cluster in PetF by the Fe2 -PHA intermediate and the

III/III product Fe2 cluster forms. For all ratios of PetF:ADO examined, the same overall change in absorbance at 423 nm was observed (Figure 2-8B). This constant value of ΔA423 (0.18) is consistent with complete PetF oxidation (Figure 2-4). However, as expected, the kinetics of oxidation depend on whether the ADO intermediate is present at sufficiently high concentration

III/III to oxidize the PetF entirely itself (at low PetF:ADO) or, alternatively, the product Fe2 cluster and/or additional turnovers must also contribute to PetF oxidation (at high PetF:ADO).

Figure 2-8. Sequential-mixing SF-Abs experiments monitoring PetF oxidation upon reaction with III/III the Fe2 -PHA intermediate. (A) Time-dependent absorption spectra after mixing the III/III accumulated Fe2 -PHA intermediate with two equivalents of chemically reduced PetF (0.050 III/III II/II mM final). The Fe2 -PHA intermediate was generated by mixing the Fe2 -ADO•octanal complex with O2-saturated buffer and incubating for 15 s. The “0” ms reference spectrum (black) was constructed mathematically by addition of the individual spectra of the chemically reduced III/III PetF (grey dotted line) and the Fe2 -PHA intermediate (light blue dotted line). The spectrum of III/III II/II the Fe2 -PHA intermediate was generated by mixing a solution containing 0.050 mM Fe2 - ADO and 10 mM octanal with an equal volume of O2-saturated 50 mM sodium HEPES, pH 7.5. The spectrum of the PetF [2Fe-2S]1+ was obtained by mixing chemically reduced PetF (0.10 mM) with an equal volume of O2-free 50 mM sodium HEPES, pH 7.5. (B) Kinetic traces monitoring PetF [2Fe-2S]1+ oxidation by the increase in absorption at 423 nm in experiments in which the ADO concentration was varied at a constant concentration of chemically reduced PetF (0.050 mM final). Dashed lines are the fits described in the text.

2.1.4.1 Experiments with PetF:ADO ≤ 2.

In the experiments with a ratio of either 0.5:1 or 1:1 PetF:ADO (Figure 2-8B, blue and pink traces), the kinetic traces had a non-zero initial absorbance, demonstrating that the majority of the PetF undergoes oxidation within ~ 1 ms (the dead-time of the measurements). This oxidation is illustrated by comparison of the PetF absorption spectrum at 1 ms (Figure 2-8A, purple line) to the mathematically generated spectrum for time zero (Figure 2-8A, black line).

Due to this limitation in data acquisition, the kinetics of PetF oxidation in these reactions could

92 not be reliably analyzed. Nevertheless, a lower limit of > 500 s-1 could be set for the apparent first-order rate constant of this rapid oxidation. A slow exponential decay phase was also observed, most discernible in the trace from the experiment with 0.5 equivalents of PetF:ADO

-1 (Figure 2-8B, blue trace). The observed decay rate constant, kobs = 0.017 s , correlates well with

III/III that reported for the unproductive decay of the Fe2 -PHA intermediate (λmax = 450 nm) to a µ-

III/III oxo-Fe2 species (λmax = 350 nm) under the intense, polychromatic incident light employed in this experiment.(17)

In the experiment with a ratio of 2:1 PetF:ADO (Figure 2-8B, green trace), the kinetic trace was best fit by a sum of two exponential functions with observed rate constants of 460 s-1

-1 and 42 s and amplitudes of ΔA423 = 0.121 and 0.051, respectively. These rate constants are significantly greater than those associated with the slow oxidation of PetF observed in control

III/III experiments in which ADO was either used in its as-isolated μ-oxo-Fe2 form (Figure 2-7) or omitted entirely (Figure 2-4). These results establish that the rapid oxidation of PetF with

III/III III/III PetF:ADO ≤ 2 is mediated by the Fe2 -PHA intermediate. In fact, reduction of the Fe2 -PHA

-1 intermediate by PetF is remarkably rapid (kobs > 400 s at 5 °C) and is more than 40 times faster

MeO -1 than reduction by PMS (kobs = 9 s at 5 °C)(17) under these experimental conditions (Figure

2-9).

Figure 2-9. Comparison of the kinetics of oxidation of two equivalents of PetF (λ = 423 nm) or MeO III/III one equivalent of dithionite-reduced PMS (λ = 388 nm)(17) upon reaction with the Fe2 - PHA intermediate. The intermediate was accumulated by mixing a solution of 0.1 or 0.3 mM II/II MeO Fe2 -ADO (PetF or PMS experiment, respectively) and 10 mM decanal with an equal volume of O2-saturated buffer and incubating for 15 s before the second mixing event with the respective reductant.

2.1.4.2 Experiments with PetF:ADO > 2.

With four equivalents of PetF:ADO (Figure 2-8B, orange trace), the kinetic trace could be fit by an equation for three parallel exponential processes. Rapid oxidation of approximately

-1 -1 half of the PetF (ΔA423 = 0.085) occurred with observed rate constants of 440 s and 56 s . These rate constants are comparable to those observed with a ratio of 2:1 PetF:ADO, indicating that this

III/III fast phase represents oxidation of two equivalents of PetF by the Fe2 -PHA intermediate. The remaining half of the PetF (ΔA423 = 0.0937) was oxidized with a markedly diminished observed rate constant, 0.856 s-1. This observation was interpreted to correspond to reduction of the product

III/III II/II μ-oxo-Fe2 cofactor to the Fe2 form, consuming two more equivalents of PetF. Control SF-

III/III Abs experiments in which as-isolated Fe2 -ADO was reacted with reduced PetF gave a slightly slower rate constant of 0.81 s-1 (Figure 2-7). This discrepancy is not unexpected since the as- isolated and product diferric states are distinct by Mössbauer-spectroscopic analysis, the latter of which appears to be more rapidly/easily reduced by PetF. When the PetF:ADO ratio was further

94 increased to 6:1 (Figure 2-8B, grey), similar fast and slow kinetic phases were observed (kobs1 =

-1 -1 230 s and kobs2 = 0.443 s ). The complete oxidation of PetF observed in this experiment with excess O2 and aldehyde substrate is explained by the completion of one full catalytic cycle, which consumes four equivalents of PetF, followed by a second half-turnover, in which the remaining

PetF is oxidized.

2.1.5 Rapid reduction of the ADO Fe2(III/III)-PHA intermediate by PetF, monitored by rapid freeze-quench (RFQ) Mössbauer spectroscopy

III/III RFQ-Mössbauer experiments were performed to monitor reduction of the Fe2 -PHA intermediate by PetF and identify any intermediates that may accumulate during this reaction. A

III/III control sample of the Fe2 -PHA intermediate was generated in a single-mixing experiment by

57 II/II reaction of the Fe2 -ADO•decanal complex for 30 s with the O2-generating chlorite dismutase/chlorite (Cld/NaClO2) enzymatic system according to the kinetics from a corresponding SF-Abs experiment (Figure 2-24).(17, 34, 35) The Mössbauer spectrum of this sample (Figure 2-10A, black bars) is dominated by the pair of partially resolved quadrupole

III/III doublets associated with the site-differentiated Fe2 -PHA intermediate. The parameters for the individual sites are δ1 = 0.48 mm/s, ΔEQ1 = 0.49 mm/s (grey bracket) and δ2 = 0.55 mm/s, ΔEQ2 =

1.23 mm/s (black bracket).(17) The spectrum of the intermediate accounts for 80% of the total iron absorption area, as determined by using the previously published(17) experimental reference spectrum (red line). The remaining absorption area arises from a quadrupole doublet with parameters of δ ~ 1.3 mm/s and ΔEQ ~ 3.0 mm/s (blue line), attributable to the ~ 20% of

II/II substrate-free Fe2 -ADO in the sample.(17)

III/III In a subsequent experiment, the Fe2 -PHA intermediate generated in the same manner was then reacted with two equivalents of chemically reduced, unlabeled (>95% 56Fe) PetF for

95 0.010 s. The spectrum of this sample (Figure 2-10B) does not exhibit the well-resolved features

III/III associated with the Fe2 -PHA intermediate, suggesting that the intermediate has mostly reacted

III/III with the reduced PetF. The accumulation of a new species upon reaction of the Fe2 -PHA intermediate with reduced PetF for 0.010 s was demonstrated in the following way. Spectrum A was subtracted from spectrum B to yield a difference spectrum (Figure 2-10C, black bars), in

III/III which the lines pointing upward represent the features of the Fe2 -PHA intermediate that

III/III decay. Adding back 55 % of the experimental reference spectrum of the Fe2 -PHA intermediate (Figure 2-10C, red line) to this difference spectrum cancels the features of the

III/III decayed Fe2 -PHA intermediate and yields the spectrum of the new species that form (Figure

2-10D). The significant overlap of the various sub-spectra makes the fraction of intensity from

III/III the Fe2 -PHA intermediate to add back rather uncertain (55 ± ~ 8 %); however, varying its contribution within this limited 8% range does not drastically change the resultant reference spectrum of the new species (Appendix A, Figure S6). The spectrum shown in Figure 2-10D can be rationalized as the superposition of two (or more) broad and overlapping quadrupole doublets

III/III with parameters typical of AF-coupled Fe2 clusters.

III/III Figure 2-10. 4.2-K/53-mT FQ-Mössbauer spectra monitoring the reaction of the Fe2 -PHA III/III intermediate with two equivalents of reduced PetF. (A) Spectrum of the Fe2 -PHA 57 II/II intermediate accumulated after a solution of the Fe2 -ADO•decanal complex and Cld was reacted with NaClO2 for 30 s. The previously published experimental reference spectrum of the III/III Fe2 -PHA intermediate(17) is overlaid in red with the constituent quadrupole doublets designated with black and grey brackets. The experimental reference spectrum of substrate-free II/II III/III Fe2 -ADO is overlaid in blue. (B) Spectrum of a sample quenched after reaction of the Fe2 - PHA intermediate with two equivalents of chemically reduced, unlabeled (>95% 56Fe) PetF for 0.010 s. The dashed line denotes the high-energy line of the doublet with ΔEQ = 1.2 mm/s (black III/III bracket) arising from the Fe2 -PHA intermediate. (C) B−A difference spectrum (black bars), III/III overlaid with 55% of the experimental reference spectrum of the Fe2 -PHA intermediate (red III/III line).(17) (D) Experimental reference spectrum of the Fe2 clusters formed upon reaction of the III/III Fe2 -PHA intermediate with two equivalents of reduced PetF for 0.010 s. The spectrum was III/III generated by addition of 55 % of the experimental reference spectrum of the Fe2 -PHA III/III intermediate to the difference spectrum C. (E) Experimental reference spectrum of the Fe2 III/III clusters generated upon reaction of the Fe2 -PHA intermediate with two equivalents of reduced PetF for 0.22 s (black bars; see Figure 2-11 for generation of the spectrum). The derived III/III spectrum of the Fe2 species formed after the 0.010 s reaction with PetF is overlaid for comparison (green line) to depict the changes in relative intensity of the doublets of the two III/III Fe2 forms (δ  0.5 mm/s with ΔEQ  0.8 mm/s and 1.5 mm/s, purple and orange brackets, respectively).

III/III The Mössbauer spectrum of a sample obtained after reaction of the Fe2 -PHA intermediate with reduced PetF for a longer time (t = 0.22 s) was analyzed in an analogous

III/III fashion (Appendix A, Figure S7), and the derived reference spectrum of the Fe2 clusters that form is shown in Figure 2-10E (black bars). This spectrum was generated under the assumption

III/III of complete reduction of the Fe2 -PHA intermediate. Direct comparison to the reference

III/III spectrum of the Fe2 clusters obtained after 0.010 s reaction time (Figure 2-10E, green line)

III/III reveals that the relative amounts of the (at least) two different Fe2 clusters change between

III/III 0.010 and 0.22 s. The relative quantity of Fe2 cluster(s) characterized by smaller quadrupole

III/III splitting (ΔEQ  0.8 mm/s, purple bracket) decreases, whereas the quantity of the Fe2 cluster(s) with larger quadrupole splitting (ΔEQ  1.5 mm/s, orange bracket) increases. Though

III/III the spectral resolution is insufficient to confidently assign these Fe2 forms to specific intermediates in the ADO reaction, their spectroscopic features are similar to those reported for

III/III the purported formate-bound Fe2 cluster (state IX in Figure 2-2),(17) consistent with the

III/III productive reduction of the Fe2 -PHA intermediate by PetF in less than 0.22 s.

The conclusion drawn from the SF-Abs spectroscopic experiments is that PetF donates

III/III electrons more rapidly than the phenazine reductant to the Fe2 -PHA intermediate, which is corroborated by the RFQ-Mössbauer experiments. The Mössbauer spectrum of a sample after a

III/III 0.010 s reaction of the Fe2 -PHA intermediate with PetF is similar to that of a sample after a

0.56 s reaction of the intermediate with MeOPMS;(17) both spectra reveal disappearance of the

III/III 57 Fe2 -PHA intermediate to a similar extent. In the PetF reaction, 55 % of the total Fe in the

III/III III/III sample has undergone conversion from the Fe2 -PHA intermediate to new Fe2 species in

III/III 0.010 s, and, by 0.22 s, the intermediate has completely decayed. At least two distinct Fe2 species formed in the reduction reaction: (a) one species that was not previously observed in the analogous experiment with MeOPMS as reductant, but is the major component at the shortest reaction time point upon reduction by PetF, and (b) a second one that accumulates markedly at

98 later reaction times and has parameters similar to those of the putative formate-bound complex

(state IX in Figure 2-2) previously detected in the reaction with MeOPMS.(17) The first species presumably represents a state that accumulates only in the PetF reaction as a result of its greater efficiency and perhaps its nature as an obligatory single-electron donor. Although structural

III/III assignment of this Fe2 -containing species is beyond the scope of the present work, our results

III/III nevertheless demonstrate that reductive decay of the Fe2 -PHA intermediate is accompanied by formation of only diferric species, strongly supporting the mechanism in Figure 2-2.

2.1.6 Sequential delivery of two electrons from PetF to ADO in a tightly coupled fashion

In the SF-Abs experiment performed with a ratio of 0.5:1 PetF:ADO, the kinetic trace at

423 nm (absorption increase reflecting PetF [2Fe-2S]1+ oxidation) exhibited a decay phase after ~

0.5 s. Because PetF cannot be re-reduced under these experimental conditions, the observed decay phase must be associated with another chromophore absorbing at this wavelength. Indeed,

III/III the observed rate constant is essentially identical to that of the unproductive decay of the Fe2 -

-1 -1 PHA intermediate, which has an absorption maximum at 450 nm (ε450 = 1,200 M •cm ).(17)

This observation suggests that when PetF is sub-stoichiometric with respect to ADO, two

III/III electrons are rapidly transferred to one ADO molecule, such that a portion of the Fe2 -PHA is not reduced at all. Thus, it was hypothesized that the reaction kinetics of reductive C1-C2 bond cleavage might favor rapid transfer of a second electron to the same ADO molecule after the

III/III initial reduction of the Fe2 -PHA intermediate.

This hypothesis was tested in a RFQ-Mössbauer experiment in which a sample

III/III containing ~ 65 % of the Fe2 -PHA intermediate (Figure 2-11A) was mixed with 0.5 equivalents of reduced PetF and allowed to react for 0.22 s (Figure 2-11B, black bars). This reaction time was chosen to ensure complete PetF oxidation, as judged from the corresponding

99 SF-Abs experiment (Figure 2-9, blue trace). The spectrum of the 0.22 s FQ sample (Figure 2-

11B, black bars), when compared to that of the control sample prior to reaction with PetF (blue line), shows an evident decrease in the area of the resolved quadrupole doublets arising from the

III/III III/III Fe2 -PHA intermediate. Quantification of the fraction of Fe2 -PHA intermediate consumed during the reaction with limiting PetF would distinguish between a one-electron reduction of each reacted ADO intermediate (i.e., decay of 50 % of total Fe in the form of the intermediate) or a two-electron reduction (i.e., decay of 25 % of total Fe in the form of the intermediate). To quantify this fraction, spectrum A was subtracted from spectrum B (Figure 2-11C, black bars).

III/III The experimental reference spectrum of the Fe2 -PHA intermediate was then added back to this difference spectrum in intensities ranging from 20-50 % (Appendix A, Figure S8). The spectrum generated by addition of 25 % (Figure 2-11D) is essentially identical to that of Figure

2-10E. Moreover, the derived spectra with varying intensities added back are consistent with the analysis of Figure 2-10 only when the intensity is ≤ 25 % (Appendix A, Figure S8). Therefore,

III/III the fraction of the Fe2 -PHA intermediate that reductively decays upon reaction with limiting

PetF is consistent with the rapid shuttling (sequentially) of two electrons to a single diiron site.

100

III/III Figure 2-11. 4.2-K/53-mT FQ-Mössbauer spectra monitoring reduction of the Fe2 -PHA intermediate with 0.5 equivalents of reduced PetF with respect to ADO. (A) Spectrum of a 57 II/II sample that was freeze-quenched 30 s after a solution of 2 mM Fe2 -ADO, 16.6 mM decanal III/III and 10 μM Cld was mixed with 10 mM NaClO2. The Fe2 -PHA intermediate accumulates to II/II 65% of the total Fe and the remaining 35% is of the unreacted Fe2 -ADO. (B) Overlay of the III/III spectrum in A (blue line) with the spectrum of a sample freeze-quenched after the Fe2 -PHA intermediate, accumulated as in A, was reacted with 0.5 molar equivalent (with respect to ADO) chemically reduced, unlabeled PetF for 0.22 s (black bars). (C) Overlay of the B–A difference III/III spectrum (black bars) with the negative of the experimental reference spectrum of the Fe2 - PHA intermediate(17) (red line) plotted at an intensity corresponding to 25 % of the total area of the experimental spectrum in B. (D) Spectrum after addition of 25 % of the experimental III/III reference spectrum of the Fe2 -PHA intermediate to the difference spectrum C. The dashed line denotes the position of the high-energy line of the doublet with ΔEQ = 0.49 mm/s from the III/III spectrum of the Fe2 -PHA intermediate.

Additional support for this hypothesis can be obtained from analysis of the stoichiometric

III/III reactions of the Fe2 -PHA intermediate with PetF/FNR/N. If each ADO molecule were to be reduced by just one electron, formate should be produced in a 1:1 ratio with respect to reducing equivalents because formate is generated after only a single electron transfer. However, in these experiments varying the reducing equivalents present (Figure 2-6), the stoichiometry of formate:electrons observed was ~ 0.5:1, as opposed to the 1:1 ratio. The observed stoichiometry

III/III thus corroborates the conclusion that half of the Fe2 -PHA intermediate reacts to consume two

101 reducing equivalents. The rapid, tightly coupled delivery of two electrons (sequentially) to the

III/III Fe2 -PHA intermediate demonstrates that PetF can function as an effective redox partner for

ADO, optimally timing electron transfer events for efficient substrate conversion.

2.2 Interrogating the postulated free-radical mechanism of ADO catalysis

The components central to the proposed free-radical mechanism are (i) a nucleophilic diferric-peroxide as the oxidizing intermediate, (ii) product formation that is dependent upon external reductants, and (iii) a C2-alkyl radical substrate intermediate that precedes the alkane product. Strong evidence for the first two points was described in previously published work and in the work described in section 2.1. The results of the RFQ-Mössbauer experiments probing

III/III reduction of the Fe2 -PHA intermediate by the obligatory single-electron donor, PetF, demonstrated that only diferric intermediate states accumulate during substrate conversion, in agreement with our previous studies with the phenazine system.(17) We obtained no spectroscopic evidence for accumulation of a diferrous or a mixed-valent species upon reduction

III/III of the Fe2 -PHA intermediate by PetF, arguing against an alternative mechanism involving such intermediates that was proposed by another group.(36) Likewise, no evidence was obtained for formation of high-valent intermediates (i.e., analogous to “X” in RNR or “Q” in MMOH) during the reaction, weighing against a proposed mechanism involving heterolytic O-O and C1-

C2 bond cleavage.(37) Such species could still possibly form on the ADO reaction pathway but fail to accumulate to detectable levels because of unfavorable kinetics. However, our observation that only diferric species accumulate during catalysis provides strong support for the free-radical mechanism.(12, 17)

As for the final component, indirect evidence for formation of the proposed C2-alkyl radical (R•) was provided in two separate studies employing radical-trapping substrate

102 analogs.(38, 39) An octandecanal cyclopropyl substrate analog was converted to 1-octadecene in an ADO-dependent fashion with oxygen and a reducing system. Production of 1-octadecene necessitates ring-opening of a cyclopropylcarbinyl radical that would be generated upon β- scission of the substrate C1-C2 bond. However, in these experiments, the ADO activity was irreversibly inhibited after one turnover due to covalent modification of the enzyme with the substrate analog radical. Though this study suggests formation of a substrate C2-alkyl radical with a relatively moderate lifetime (deduced to be > 10 µs), these conclusions are inferred from non-catalytic conditions. Experiments using an oxiranyl ring substrate analog also demonstrated support for the C1-C2 bond cleavage event with formation of a C2-radical species with a comparable lifetime. Nonetheless, a kinetically and chemically competent radical intermediate relevant in the ADO mechanism remains to be identified under turnover conditions with aldehyde substrates.

Considering that PetF is an obligatory one-electron donor compared to the two-electron phenazine reductant and facilitates faster electron transfer, we anticipated that one or more intermediate with a half-integer electronic spin state might accumulate between the two sequential electron transfers in the proposed ADO mechanism. Therefore, RFQ-EPR experiments were conducted with the aim of trapping and characterizing any intermediates following

III/III reduction of the Fe2 -PHA intermediate, remaining vigilant for any metal-based EPR-active species that would support alternative mechanistic paths.

2.2.1 Accumulation of radicals upon reduction of the ADO Fe2(III/III)-PHA intermediate

Sequential-mixing RFQ experiments were carried out with both the MeOPMS and PetF reducing systems to test for accumulation of EPR-active intermediates during turnover. The reaction times for the RFQ-EPR experiments were selected on the basis of the results of the SF-

103

III/III Abs experiments in which the Fe2 -PHA intermediate was reduced either with equimolar reduced MeOPMS or with two equivalents of chemically reduced PetF (Figure 2-9).(17) Samples quenched at fast reaction times with either reductant demonstrated a nearly axial EPR signal with principal g-values 2.037, 2.009 and 2.003 (Figure 2-12) that was detectable at temperatures ≤ 70

K2. This radical species did not form if the reducing system was omitted in the second mixing event or if the aldehyde substrate was omitted in the first (Figure 2-13).

With MeOPMS as the reductant, the S = 1/2 signal accumulated to only < 0.05 ADO equivalents at the fastest reaction time of 0.010 s. It largely decayed by a reaction time of 0.50 s, as indicated by the greatly diminished intensity of its low-field component at g = 2.037 (Figure

2-12, dashed line). Its formation and decay thus occur on the same time scale as MeOPMS oxidation (Figure 2-9). With two equivalents of PetF as the reductant, the same EPR signal developed, but to a greater extent. For example, at 0.010 s, ~ 0.2 equivalents of the radical species with respect to ADO accumulated (Figure 2-12). The yield of the radical was strongly dependent on the stoichiometry of reducing equivalents (Figure 2-14); its accumulation increased as the

PetF:ADO ratio increased to 2, but no further enhancement was seen at greater ratios. The increased accumulation of this transient radical in the reaction with PetF in comparison to the

MeOPMS reaction likely reflects its faster formation with the more potent reductant (Figure 2-9), which is almost completely oxidized after 0.034 s (as revealed by disappearance of the EPR

1+ signal of the PetF [2Fe-2S] cluster).

2 We did not attempt to collect EPR spectra at higher temperatures because the EPR samples were freeze-quenched and could not be safely annealed to higher temperatures.

104

III/III Figure 2-12. CW EPR spectra of FQ samples obtained after reduction of the Fe2 -PHA intermediate by (A) one equivalent of reduced MeOPMS or (B) two equivalents of chemically III/III reduced PetF with respect to ADO. The Fe2 -PHA intermediate was accumulated in an initial II/II reaction of the Fe2 -ADO•decanal complex with the Cld/NaClO2 system (Figure 2-24) and was then mixed with the dithionite-reduced MeOPMS or PetF. The dashed lines are positioned at the low field feature (g = 2.037) of the transient radical signal. The arrows indicate the signal features of the [2Fe-2S]1+ PetF cluster. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.481 GHz, modulation amplitude = 0.4 mT.

105

Figure 2-13. X-Band CW EPR spectra of samples quenched at 0.020 s in sequential-mixing FQ II/II experiments. In the first mix, 0.60 mM Fe2 -ADO with (red, green) or without (blue) 16.6 mM decanal substrate was reacted for 30 s with O2-saturated (~1.8 mM O2 at 5 °C) 50 mM sodium III/III HEPES, pH 7.5, buffer to allow for formation of the Fe2 -PHA intermediate. In the second mix, the intermediate was reacted with either equimolar reduced MeOPMS (red, blue) or buffer (green) for 0.020 s prior to freezing in liquid 2-methylbutane. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.380 GHz, modulation amplitude = 0.3 mT.

Figure 2-14. Radical yield as a function of equivalents of PetF with respect to ADO. CW EPR III/III spectra of samples quenched 0.010 s after mixing of the Fe2 -PHA intermediate ([ADO]final = 0.23 mM) with varying concentrations of chemically reduced PetF (0.13, 0.23, 0.45, 0.90 mM final concentration). Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT. (inset) Yield of the radical species as a function of equivalents of PetF:ADO.

106 The spectra of samples quenched at reaction times ≥ 0.034 s exhibited additional S = 1/2 signals that increased in intensity over time (Figure 2-12). In the MeOPMS reaction, these signals can be assigned to (a) the phenazine semiquinone radical with an isotropic signal at g ~ 2 (Figure

2-13, blue trace), and (b) a second radical species with complex fine structure, identified on the basis of evidence presented in section 2.3 as a protein-based sulfinyl radical. The same sulfinyl radical signal was also present in the spectra of samples from the PetF reaction for 2 s (Figure 2-

12).

2.2.2 Association of the transient EPR signal with a substrate-derived alkylperoxyl radical

The kinetic behavior of the transient radical signal observed in the EPR spectra (gav ~

2.02) and the dependence of its intensity on the reducing system suggest that the associated species might either be, or be derived from, an intermediate in the ADO reaction. The unusual g- anisotropy and lack of resolved hyperfine structure in the signal obtained at X-band frequencies weigh against the possibility that the detected species is the purported C2-alkyl radical (the R• resulting from β-scission of the gem-diolyl radical intermediate; state VII in Figure 2-2).(40-42)

To probe the chemical nature of this radical species, identical FQ-EPR experiments were carried

2 out with an aldehyde substrate specifically deuterium-labeled at the C2-position (2,2-[ H]2- decanal). The EPR signal of a sample generated with this substrate isotopolog exhibited a pronounced narrowing (Figure 2-15) due to the smaller hyperfine coupling of deuterium (A1H/A2H

= gn(1H)/gn(2H) = 6.514). This observation verifies that the radical is coupled to the C2-hydrons and therefore be derived from the substrate. Additional experiments carried out with 57Fe-labeled

ADO and 1-[13C]-labeled octanal did not yield any detectable differences in the spectra

(Appendix A, Figure S11). These observations imply that the radical species is not magnetically coupled to the diiron center and that the C1-C2 bond is cleaved prior to its formation.

107 The g-anisotropy of the substrate-derived radical is reminiscent of signals of peroxyl radicals in proteins and organic systems.(43-49) We posited that the radical could result from the reaction of the proposed R• with dioxygen, yielding a peroxyl radical (ROO•), which would have an electronic structure consistent with the observed EPR signal. RFQ-EPR experiments were thus performed with a 3-thia-decanal analog to test for perturbations to the EPR signal of the radical species. The EPR spectrum of the radical signal was essentially the same, with only marginal changes in its principal g-values (2.034, 2.009, and 2.005) (Appendix A, Figure S12). This insensitivity to the sulfur substitution is expected for a peroxyl radical, because its spin density

2 resides primarily on the oxygen atoms. The EPR signal of a sample generated with the 2,2-[ H]2-

3-thia-decanal analog exhibited a smaller but detectable narrowing, also consistent with coupling of the electron spin to the C2-hydrons (Appendix A, Figure S12).

108

Figure 2-15. X-band CW EPR spectra of the substrate-derived peroxyl radical species (chemical III/III structure shown in inset). (A) The Fe2 -PHA intermediate was formed in the presence of either 1 2 the 2,2-[ H]2-decanal (black) or 2,2-[ H]2-decanal (red) substrate and was reacted with equimolar MeO 1 reduced PMS for 0.010 s. The spectrum of the 2,2-[ H]2-decanal sample was multiplied by a factor of 1.7 for better comparison to account for the difference in [ADO] (1.49 v. 0.90 mM). Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT. The spectra generated using (B) 2,2- 1 2 [ H]2-decanal and (C) 2,2-[ H]2-decanal were simulated (blue lines) by using the same set of parameters and hyperfine couplings obtained from the simulation of the HYSCORE spectra. The lineshape of the spectra was reproduced by considering an isotropic Voigtian lineshape with peak-to-peak linewidths of 0.17 and 0.24 mT, for the Lorentzian and Gaussian components, respectively, and an anisotropic broadening of [30 22 28] MHz (H-Strain) or [30 18 28] MHz for 1 2 the spectra generated using the 2,2-[ H]2-decanal and the 2,2-[ H]2-decanal, respectively. The only adjustable parameters in the refinement were the g-values. They varied by ± 0.002, which is well within the experimental error of the measurements.

To measure the hyperfine couplings of the C2-hydrons more precisely, HYSCORE

2 experiments were performed on RFQ samples prepared with the 2,2-[ H]2-decanal and the 2,2-

2 III/III [ H]2-3-thia-decanal substrates and freeze-quenched after a 0.020 s reaction of the Fe2 -PHA intermediate with reduced MeOPMS. The HYSCORE spectra (Figure 2-16) recorded at the principal positions of the g-tensor (g = 2.037 and g ~ 2.01) showed correlations between nuclear || ┴ frequencies from different electronic manifolds corresponding to the 2H nuclei. The cross peaks centered at the 2H Larmor frequency showed that the 2H hyperfine couplings for the putative 1,1-

2 2 [ H]2-2-thia-C8H16OO• analog radical are smaller than those for the 1,1-[ H]2-C9H16OO• radical

(Figure 2-16).

109

Figure 2-16. X-Band HYSCORE spectra recorded on FQ samples quenched after reaction of the III/III MeO maximally accumulated Fe2 -PHA intermediate with equimolar reduced PMS for 0.02 s in 2 2 double-mixing experiments using the 2,2-[ H]2-decanal (left column) or the 2,2-[ H]2-3- thiadecanal analog (right column). The numerically simulated spectra are overlaid with grey contours, and the simulation parameters are given in Appendix A (Table S1). Experimental conditions: microwave frequency = 9.388 GHz, τ = 200 ns, π/2 = 8 ns, T = 30 K.

Because the modest anisotropy of these signals restricts orientation selection in the

HYSCORE experiments, DFT calculations on geometry-optimized models of the substrate- derived radicals were used to guide spectral simulations. The spectra could be reproduced well

2 2 for both radicals, yielding Aiso( H) ~ 2.6 MHz for the 1,1-[ H]2-C9H16OO• considering two

2 equivalent deuterium nuclei. For the 3-thia analog, the Aiso( H) was decreased by almost half, and the two deuterons were slightly inequivalent (Appendix A, Table S1), as anticipated for the more electron-withdrawing sulfur and the larger extent of delocalization. The single-point calculations on geometry-optimized models of these radical moieties reproduced both the g-tensor and the

110 anisotropy of the A-tensor reasonably well (Appendix A, Table S1) and support the proposed structure of a peroxyl radical.

2.2.3 Dependence of peroxyl radical accumulation on the concentration of O2

The substrate-derived peroxyl radical was hypothesized to arise from a side-reaction of the R• with dioxygen, and its yield should therefore depend on the concentration of O2. In all previously described RFQ-EPR experiments, a high [O2] (5 mM after the first mix) was generated

by the Cld/NaClO2 enzymatic system.(34, 35) In a RFQ-EPR experiment with Cld/NaClO2, a ratio of 1:1 PetF:ADO, and octanal as substrate, 0.04 ADO equivalents of the peroxyl radical

III/III accumulated after reaction of the Fe2 -PHA intermediate with the reduced PetF for 0.010 s

(Figure 2-17A, black spectrum). In contrast, with an O2-saturated buffer solution (~0.90 mM O2

III/III after the first mix) sufficient to generate an equal quantity of the Fe2 -PHA intermediate

(Figure 2-24), significantly less peroxyl radical accumulated, amounting to < 0.005 ADO equivalents, as determined by spectral deconvolution analysis (Appendix A, Figure S16). The peroxyl radical was barely detectable in the EPR spectrum (Figure 2-17A), which was dominated by the spectrum of the aforementioned sulfinyl radical. The yield of the peroxyl radical trapped in

RFQ experiments appears to be dependent upon the substrate chain-length, with shorter aldehyde substrates (i.e., C6-C10) supporting greater accumulation and longer aldehydes (i.e., n- octadecanal) supporting minimal accumulation (Appendix A, Figure S14). The dependence of the quantity of peroxyl radical on the O2 concentration supports the hypothesis that it is produced from a reaction of the R• with excess molecular oxygen.

111

Figure 2-17. Peroxyl radical yield with varying concentrations of O2. CW EPR spectra of FQ III/III samples generated by reaction of the Fe2 -PHA intermediate with one equivalent of reduced III/III II/II PetF for 0.010 s. The Fe2 -PHA intermediate was generated by mixing either the Fe2 - ADO•octanal complex (1.49 mM) containing 10 μM Cld with an equal volume of NaClO2 (black II/II line, 10 mM O2) or the Fe2 -ADO•octanal complex (1.49 mM) with O2-saturated buffer (blue line, ~1.8 mM O2) and incubating for 50 s (Figure 2-24). For clarity, the signal of unreacted [2Fe-2S]1+ PetF (15 %) was subtracted from these spectra. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT.

2.2.4 Diminished alkane yields due to unproductive alkyl radical quenching by O2

Peroxyl radicals are well documented in enzymes operating by free-radical mechanisms, resulting from reaction of carbon-centered radicals with molecular oxygen.(50-54) Known ROO• species in these systems have either catalytic or inhibitory roles.(43, 44) In purple lipoxygenases, an ROO• is the on-pathway intermediate in the dioxygenative conversion of linoleic acid to hydroperoxy-octadecadienoic acid.(44, 46, 47) In the Y122F variant of Ec RNR β, but not in the wild-type enzyme, formation of an amino acid-derived peroxyl radical was observed, presumably due to the imperfect containment of free-radical intermediates.(55) Perhaps the best characterized

ROO• adduct is the one observed in pyruvate formate-lyase (PFL), in which the catalytic glycyl radical reacts with O2 to form a peroxyl radical that initiates cleavage of the peptide backbone, ultimately inactivating the enzyme.(43, 56)

112 If the formation of the peroxyl radical in ADO results from the trapping of the R• intermediate by O2 in competition with its productive quenching by H•, then the alkane yield

(relative to formate) should be diminished by this pathway. Indeed, in single turnover assays under O2-saturating conditions, the slope of alkane produced as a function of electrons provided,

0.11 alkane per 2 electrons, was markedly less than that obtained under O2-limiting conditions,

0.76 alkane per 2 electrons (Figure 2-18). The change in O2 concentration did not affect the formate yield, as to be expected because formate is formed immediately after C1-C2 scission, concomitant with formation of the R•. The significant uncoupling between formate and alkane products observed when O2 is in excess is consistent with diversion of some of the R• to peroxyl radical, which is expected to decay to an oxidized product. When O2 is limiting, less R• is unproductively trapped, and, consequently, the alkane:formate ratio approaches the theoretical

1:1 stoichiometry. These results demonstrate that, in the presence of excess O2, an appreciable fraction of the R• intermediate is lost as ROO•, ultimately precluding its productive conversion to alkane. The formation of a substrate-derived peroxyl radical during the productive reaction of

ADO via quenching of an on-pathway intermediate unequivocally establishes that the enzyme operates by a free-radical mechanism.

113

Figure 2-18. Formate (squares) and alkane (circles) product formation in stoichiometric ADO III/III reactions with varying concentrations of O2. The Fe2 -PHA intermediate was generated by II/II mixing the Fe2 -ADO•decanal complex (0.250 mM ADO) – in the presence of oxidized FNR and PetF (0.250 mM each) – with (A) O2-saturated buffer (1.8 mM O2 at 5 °C) and incubating for 10 s or (B) air-saturated buffer (~ 0.28 mM O2 at 25 °C) and incubating for 20 s (Figure 2-24). The intermediate was then mixed with varying concentrations of NADPH (a two-electron donor) and quenched after ~ 1 s.

Detection of the substrate-derived ROO• reveals a limited competency of ADO to contain its free-radical intermediate, R•. Furthermore, although molecular oxygen is absolutely essential for the ADO reaction, it can also limit the yield of the desired alkane product. This observation has, in fact, been reported previously in both in vitro and in vivo studies, but the molecular origins were not identified. In an in vitro study employing the phenazine reducing system, alkane production was greatly enhanced under microaerobic conditions; yields were otherwise very low in the presence of excess O2 (~ 30-fold relative to [ADO]).(39) More recently, in a bioengineered

Ec system overexpressing ADO, the PetF/FNR system, and catalase, production of short-chain alkanes (e.g., propane and heptane) was inhibited by excess O2. In contrast, when the O2 concentration was significantly decreased, alkane production was enhanced.(31) The poor alkane yield was thus independent of the nature of the reducing system (chemical or protein) and of whether the reaction takes place in vitro or in vivo. These observations are in agreement with our experimental results, which clearly demonstrate that the molecular basis for the diminished alkane production is the previously unrecognized unproductive trapping of the R• intermediate by excess O2.

114 2.2.5 Detection of a protein-based sulfinyl radical using short-chain aldehyde substrates

As was alluded to in previous sections, a second EPR-active species was detected in the

RFQ experiments with short-chain (C6-10) substrates. This species persisted long after complete oxidation of the MeOPMS or PetF (> 3 min), in contrast to the rapidly decaying peroxyl radical, and could also be detected at temperatures ≤ 70 K3 (Figure 2-12). In RFQ-EPR experiments with a ratio of 2:1 PetF:ADO and octanal as substrate, this radical signal detectably accumulated by a reaction time of 0.010 s and was still observed at 2 s, with maximal accumulation to 0.04 ADO equivalents (Figure 2-19). The signal of the peroxyl radical developed to only a minor extent in this reaction (Figure 2-19). The octanal substrate afforded faster accumulation of this radical than decanal. Further extending this dependence on chain length, the longer, native substrate, n- octadecanal, afforded no detectable accumulation of this radical species (Appendix A, Figure

S14).

3 We did not attempt to collect EPR spectra at higher temperatures because the EPR samples were freeze-quenched and could not be safely annealed to higher temperatures.

115

Figure 2-19. X-band CW EPR spectra of RFQ time-course samples using octanal as the ADO II/II substrate. The WT Fe2 -ADO∙octanal complex was reacted in a first mix with the Cld/NaClO2 III/III system to generate the Fe2 -PHA intermediate, which was reacted in a second mix with two equiv of chemically reduced PetF. Experimental conditions: temperature = 60 K, microwave power = 2 mW, microwave frequency = 9.481 GHz, modulation amplitude =0.4 mT.

To address the chemical nature of this radical, the aforementioned substrate and cofactor isotopologs were again deployed. However, the EPR signal was not perturbed by using

13 2 57 isotopically-labeled aldehyde substrate (i.e., 1-[ C]-octanal and 2,2-[ H]2-decanal) nor Fe- labeled ADO (Appendix A, Figure S11), indicating that the radical is neither substrate-derived nor magnetically coupled to the ADO diiron cofactor. Its accumulation was dependent upon the presence of a reducing system, but not on the nature of the reducing system (chemical or protein).

Collectively, these characteristics suggested that this second, long-lived radical most likely resides on an amino acid residue of ADO.

However, the anisotropy and fine structure of the second radical signal do not match known signals of amino acid radicals.(57-59) Its spectral features imply that the spin density resides on an atom with appreciable spin-orbit coupling, such as a sulfur or an oxygen atom.(60,

61) Four residues, Tyr18, Tyr22, Cys71 and Tyr123, were identified as candidates to harbor this

116 second radical and were replaced by redox-inert residues (TyrPhe or CysAla) to test for elimination of the associated EPR signal. These residues were selected on the basis of their proximity to the bound substrate (Figure 2-20A) and their strict conservation in cyanobacterial

ADO primary structures. In sequential-mixing RFQ-EPR experiments, the wild-type (WT) or

II/II variant Fe2 -ADO•octanal complex was first reacted with the Cld/NaClO2 system to generate

III/III the Fe2 -PHA intermediate, and the intermediate was then reacted in a second mixing step with equimolar reduced MeOPMS for 4 s. The EPR spectrum of the WT ADO sample exhibited the complex anisotropic S = 1/2 radical signal with an integrated intensity corresponding to 0.08

ADO equivalents. This radical species was formed in all of the ADO variants except C71A

(Figure 2-20B), indicating that the radical must reside on Cys71. The TyrPhe substitutions each resulted in a two-fold decrease in the radical yield, as well as changes in the signal lineshape

(Figure 2-20B). These effects most likely arise from minor perturbations to the environment of the nearby Cys71 residue and radical derived therefrom. In contrast, substitutions of residues located far from Cys71 (e.g., C107A, C117A) resulted in less pronounced changes in the radical signal (Figure 2-20C). All four variant proteins with substitutions that affected accumulation of the Cys71-based radical are less active than wild-type ADO under multiple turnover conditions

(Table 2-1); however, no substitution resulted in complete loss of activity.

117

III/III Figure 2-20. Formation of a Cys71-SO• upon reduction of the Fe2 -PHA intermediate. (A) X- ray crystal structure of Se ADO (PDB accession code: 4RC5), cut away to depict the location of the residues substituted by mutagenesis (Np ADO numbering) relative to the cofactor and bound substrate analog. (B) CW X-band EPR spectra of freeze-quenched samples from the WT and II/II variant ADO reactions in which the Fe2 -ADO•octanal complex (0.9 mM WT, C71A and Y123F, 0.6 mM Y18F and 0.5 mM Y22F) containing 10 μM Cld was mixed with an equal III/III volume of 10 mM NaClO2 and reacted for 30 s to generate the Fe2 -PHA intermediate. Samples were quenched 4 s after a second equal-volume mix of the intermediate with equimolar reduced MeOPMS. For the Y18F and Y22F variants, the spectra were scaled by factors of 1.8 and 1.5, respectively, for purposes of comparison. The spectrum for the C71A variant was scaled by a factor of 3 for better visualization of the low field component of the ROO• signal (g = 2.037, II/II arrow). (C) Samples were prepared as described in B with 0.9 mM Fe2 -ADO (WT, C107A, C117A) Experimental conditions: temperature = 60 K, microwave power = 2 mW, microwave frequency = 9.481 GHz, modulation amplitude =0.4 mT.

Table 2-1. Comparison of the yield of the Cys71-centered radical, as determined by FQ-EPR experiments (t = 4 s), and the normalized (with respect to WT ADO) activities (alkane yield) of ADO variants obtained from multiple turnover assays. The yield of the Cys71-SO• is given as a percentage with respect to the total ADO cofactor concentration.

Normalized ADO variant % radical yield % Activity WT 8.0 100 Y123F 3.9 36 C71A Nd 73 C107A 7.7 92 C117A 6.7 63 Y18F 3.9 57 Y22F 4.3 77

118 The electronic structure of the Cys71-based radical was further investigated spectroscopically and computationally to decipher its chemical nature. To resolve its g-tensor and fine structure, the 2-pulse echo-detected field-swept Q-band EPR absorption spectrum was recorded, the first derivative of which is shown in Figure 2-21. The spectrum was simulated with a slightly anisotropic g-tensor (gx, gy, gz = [2.02, 2.01, 2.00]); the “triplet-like” hyperfine structure

1 was reproduced considering two slightly inequivalent H hyperfine couplings (Aiso1 = 39.7 MHz,

Aiso2 = 44.3 MHz). These g-values and small g-anisotropy of the C71-based radical signal are not consistent with those of known thiyl radicals (S•), but rather, are more typical of (alkyl)-sulfinyl radicals (RSO•).(60, 62, 63) In the former, a large gx value is characteristically observed (gx =

2.10-2.29) as a result of the near degeneracy of the second lone pair orbital.(60) A thiyl radical configuration is thus incompatible with the electronic structure of the Cys71-centered radical. By contrast, the experimentally measured g-tensor of the Cys71-centered radical is essential identical to that of sulfinyl radicals.(43, 45, 56, 63)

119 Furthermore, the isotropic hyperfine coupling constants for the β-hydrogens of the Cys71 residue obtained from the numerical simulation of its radical spectrum are significantly smaller than those of typical thiyl radicals, but are similar in magnitude to the hyperfine coupling constants reported for sulfinyl radicals. In sulfinyl radicals, the isotropic hyperfine coupling constants of the Cβ proton on Hβ(1) is typically negligible, and only the Hβ(2) carries significant spin density, with an average isotropic hyperfine coupling constant of ~ 1.4 mT (~ 40 MHz).(43,

45, 56, 64) However, simulation of the Cys71-based radical spectrum necessitated inclusion of two inequivalent, but similar hyperfine coupling constants, suggesting significant spin density on both of the β-hydrogens, as opposed to only one. The isotropic hyperfine coupling constants of the Cβ protons in sulfur-centered radicals are given by the expression: Aβ(1) =

S SCH 2 S SCH ρ (3p)∙Qβ ∙cos (θ), in which ρ (3p) is the spin density on the sulfur, Qβ is the effective hyperfine coupling constant (86 MHz), and θ is the dihedral angle.(65) Assuming that the Cβ protons are bonded in an sp3 configuration, their dihedral angles differ by 120 degrees, and, thus, there is only one independent angle, θ. This expression implies that changes in the dihedral angle,

Hβ(1)-Cβ-S-O, will impact the hyperfine coupling constants for the two protons, reflecting their spin density.

To assess whether the observed hyperfine coupling constants on the Cβ protons obtained from the numerical simulation of the EPR spectrum could be consistent with hyperfine interactions of a sulfinyl radical with both β-protons, the dihedral angle Hβ(1)-Cβ-S-O was systematically varied between 0 ° and 180 ° in 20 ° increments. Hyperfine coupling constants and g-tensors were obtained from single-point DFT calculations on each of the optimized geometries

(Appendix A, Table S3). The g-tensor principal components are virtually insensitive to the changes in the dihedral angle, whereas the hyperfine coupling constants show a strong dependence on the dihedral angle (Appendix A, Table S3). The theoretical calculations show that, at a 120° torsion angle, the experimentally measured hyperfine coupling constants can be

120 well reproduced, as a result of significant spin density on both β-protons (Figure 2-22), supporting the conclusion that this EPR signal is a sulfinyl radical localized on Cys71. The dihedral angle predicted from the rotamer of C71 in the crystal structure of ADO would be 65 ° and therefore, a rotation about the C-S bond would be needed to achieve this unusual angle. This orientation could be enforced by the hydrophobic substrate-binding pocket and the interaction of the Cys71-SO• with a nearby phenylalanine residue (Appendix A, Figure S18).

Cys71 residue in ADO presumably reacts with the substrate-derived peroxyl radical to generate the sulfinyl radical, considering that its formation is concomitant with peroxyl radical decay. Also consistent with this deduction, the C71A substitution markedly stabilized the peroxyl radical;

III/III after reduction of the Fe2 -PHA intermediate for 0.010 s, much more of the peroxyl radical was present in the C71A variant than in the wild-type protein (Figure 2-23), and it was still present after a 4 s reaction in the variant protein (Figure 2-20).

121

Figure 2-23. X-band CW EPR spectra of the Cys71-SO• in WT Np ADO and NpADO variants. III/III (A) Spectra of FQ-EPR samples quenched after a 0.010 s reaction of the Fe2 -PHA MeO II/II intermediate with equimolar reduced PMS. The Fe2 -ADO•octanal complex ([ADO]f = 0.225 mM WT, C71A, C107A and Y123F, 0.150 mM Y18F and 0.125 mM Y22F) was mixed in III/III equal volume with the Cld/NaClO2 system for 30 s to generate the Fe2 -PHA intermediate. The Y18F and Y22F spectra were multiplied by factors of 1.8 and 1.5, respectively, for ease of comparison.

The extent and the kinetics of Cys71-SO• accumulation are strongly dependent on the chain-length of the aldehyde substrate. Considering that Cys71 is far from the active site and gates the hydrophobic substrate channel,(2, 66, 67) the peroxyl radical must migrate within the active site for this reaction to occur. Short-chain substrates are expected to have greater mobility, explaining the more rapid and greater accumulation of the Cys71-SO• observed. However, the fact that less Cys71-SO• is produced than ROO• precursor accumulated, together with the decay kinetics of ROO•, imply that a fraction of the ROO• decays by one or more than one alternative pathway. At present, the nature of the other decay route(s) and whether the formation of Cys71-

SO• partially or completely inactivates ADO remain unclear.(68)

122 The greater accumulation of off-pathway radical species with short-chain aldehydes (e.g.,

C6-C10) is particularly important because short-chain alkanes have been specifically targeted for their uses as liquefied petroleum gas (e.g., propane and butane) and as “drop-in” gasoline and jet- fuel components (e.g., heptane to undecane).(31, 32, 69) Short-chain aldehydes are also superior substrates for ADO in vitro compared to the natural long-chain substrates, due to their greater solubility.(14, 16, 17) Thus, bioengineered ADO variants have been designed to favor short-chain substrates by sterically occluding the substrate channel, demonstrating higher activity with aldehydes as short as butanal both in vitro and in vivo.(69) In this work, we show that, in the presence of excess O2 and/or short-chain aldehydes, the alkane yields are greatly compromised because ADO lacks an efficient mechanism to productively quench its radical intermediate. Our results establish that ADO employs a free-radical mechanism for conversion of fatty aldehydes to alkanes, and that, although O2 is a co-substrate, it also exposes the Achilles’ heel of this enzyme.

Limiting flux through the unproductive pathways identified here would be a worthwhile goal in efforts to develop more efficient ADO-based bioprocesses.

2.2.6 Exploring potential hydrogen atom donors for quenching of the alkyl radical intermediate

The major conclusion from the work described above is that ADO has limited proficiency in managing its free-radical intermediate. The substrate alkyl radical requires a hydrogen atom equivalent to proceed productively, which we originally posited could be donated by a protein amino acid residue or a solvent molecule in the active site. However, the results demonstrating unproductive diversion of the radical imply that its lifetime is long enough to be intercepted by dioxygen. Therefore, we set out to investigate whether a dedicated hydrogen atom donor exists within the protein architecture or if solvent serves this function.

123 Amino acid residues within or near the active site or substrate-binding channel that could donate hydrogen atoms (i.e., tyrosines and cysteines) were targeted as candidates. We hypothesized that if the putative hydrogen-donating residue were substituted with a residue lacking this capability, we might be able to observe an effect on ADO catalysis. To one extreme, if such a dedicated residue exists, activity would be completely abolished either under single or multiple turnover conditions. More moderate phenotypes could be reflected by altered kinetics, including slower oxidation of the reducing partner, or by diminished alkane product yields.

Finally, we can use the peroxyl radical as a probe of disruption of the hydrogen acquisition step; if the kinetics of hydrogen atom transfer are slower, more of the alkyl radical would partition to the off-pathway reaction with dioxygen to form higher yields of the peroxyl radical.

The ADO protein variants, Y123F, Y40F, Y18F, Y22F, C71A, and C107A, were expressed and purified. All variants were capable of catalyzing multiple reaction turnovers using

MeOPMS as the reducing partner, but were all impaired as compared to the WT enzyme. (ST experiments?) Since none of the variants completely abolished activity, the more discrete steps in catalysis were examined by SF-Abs spectroscopy. In single-mixing experiments, the reduced

II/II III/III Fe2 form was reacted with dioxygen to generate the Fe2 -PHA intermediate. The observed

III/III rates of formation of the Fe2 -PHA absorption feature (λmax = 450 nm) for all variants were comparable to that of the WT enzyme, implying that they are all competent in the first steps of

III/III catalysis. In sequential-mixing experiments, reduction of the Fe2 -PHA intermediate by the phenzine reductant was monitored. The Y18F, Y22F, and C71A variants all demonstrated WT- like behavior, completely oxidizing MeOPMS with similar rates. Oxidation of MeOPMS by the intermediate in Y40F was 1.5-fold slower than WT. The Y123F substitution resulted in ~75 % oxidation of the phenazine provided in a stoichiometric amount and with a 4-fold slower rate. The

Y123F variant was further investigated since it had the most drastic effect on the reduction steps.

We hypothesized that if the hydrogen atom-donating Y123 was missing, the alkyl radical could

124 resort to acquiring a hydrogen atom from the solvent. This partitioning was probed by experiments performed in D2O solvent. In multiple turnover reactions of the Y123F variant in

D2O, a greater incorporation of deuterium into the alkane product was expected than for the WT enzyme. However, the deuterium incorporation results were comparable for the WT and the

Y123F variant.

Finally, the variants were assessed as the postulated hydrogen atom donor, using the yield of the peroxyl radical as a probe. A low yield of peroxyl radical is observed in reactions of ADO with the octanal substrate (~ 4 %), thus we hoped to observe a substantial, observable increase in the yield upon substitution of the correct amino acid. The variant proteins demonstrated comparable yields of the peroxyl radical at the shortest reaction incubation time, with the exception of the C71A variant. This observation was rationalized to result from the elimination of a decay pathway for the peroxyl radical, rather than increased accumulation. This conclusion is substantiated by an increased lifetime of the peroxyl radical, which is still observable after a reaction time of 4 s. Therefore, it was concluded that none of the amino acid residues that are capable of donating a hydrogen atom in close proximity to the active site serve in this dedicated role in catalysis.

Marsh and co-workers explored the possibility that a solvent molecule present in the active, such as a water or hydroxide metal ligand could perform this function, which was supported by proton inventory analysis.(36) Consistent with this conclusion, analysis of the alkane product from multiple turnover reactions of the wild-type ADO performed in D2O demonstrate deuterium incorporation into the hydrocarbon chain. The data collectively suggest that ADO does not possess a dedicated hydrogen atom donor to the alkyl radical during catalysis.

This deficiency could partially contribute to the observed diversion in reactivity of the alkyl radical intermediate that result in low product yields. Bioengineering such a site into the active

125 site cavity of ADO could substantially improve catalytic fidelity, thereby enhancing the efficiency of potential bioreactions deploying this natural biofuel-production catalyst.

2.3 Materials and Methods

Materials. Technical grade (> 85% purity) sodium hydrosulfite (dithionite), β-nicotinamide adenine dinucleotide phosphate, reduced disodium salt hydrate (≥ 97%) (NADPH), 2- nitrophenylhydrazine (97%) (2-NPH), sodium chlorite (NaClO2), spinach ferredoxin, spinach ferredoxin-NADP+ reductase, n-hexadecane and short-chain aldehyde substrates (n-octanal and n- decanal) were purchased from Sigma-Aldrich (St. Louis, USA). 1-(3-Dimethylaminopropyl)-3- ethylcarbodiimide (98%) (EDC) was purchased from Alfa Aesar (Ward Hill, USA). 1-methoxy-

5-methylphenazinium methylsulfate (MeOPMS) was obtained from Acros Organics (New Jersey,

USA). n-octadecanal and n-1-[13C]-octanal were synthesized according to previously described procedures.(2, 13) n-1-[13C]-stearoyl-ACP was prepared according to a previously published procedure.(7) n-1-[13C]-octadecanoic acid and n-1-[13C]-octanoic acid were obtained from

Cambridge Isotope Laboratories, Inc. (Andover, MA). All other chemicals used for protein over- expression and purification were purchased from Sigma-Aldrich (St. Louis, USA), unless stated otherwise.

2 2 Procedures for syntheses of aldehyde substrates. n-1-[ H]-decanal, n-2,2-[ H]2-decanal were prepared according to published procedures.(2, 16) As depicted in Appendix A, Figure S1, BD3 was used to install deuteria onto the C1 carbonyl in the synthesis of n-1-[2H]-decanal. In

2 preparation of n-2,2-[ H]2-decanal, the deuteria were introduced via solvent exchange using

2 CH3O H.

To prepare the 3-thia substrate analog, n-hepta-1-thiol was treated with 1.0 equivalent of

NaH followed by 1.2 equivalent of 1,1-dimethoxy-ethyl-2-tosylate at 0˚C to yield the protected n-

126 3-thia-decanal. Deprotection was carried out by treatment with excess formic acid (~ 10 equivalents) to transform the dimethoxy-protected compound into the aldehyde. For synthesis of

2 2 n-2,2-[ H]2-3-thia-decanal, 1,1-dimethoxy-ethyl-2,2-[ H]2-2-tosylate was used as starting material to introduce deuteria at C2.

Preparation of Np AAR. AAR was prepared as previously described.(7)

Preparation of Np ADO. The plasmid containing the codon-optimized Np ADO gene(13)

(Npun_R1711; accession code YP_001865325) was used to transform Ec BL21 (DE3) competent cells (Invitrogen; Carlsbad, CA). ADO was over-expressed and purified aerobically by procedures similar to those used in previous work,(17) but modified as described below.

Transformed cells were incubated with shaking (250 rpm) at 37 °C in M9 minimal medium with

50 mg/L kanamycin, 0.2% (v/v) glucose, 0.1 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM

(NH4)2Fe(SO4)2·6H2O until an OD600 of 0.6-0.8 was reached. Expression was then induced by addition of 0.25 mM iso-propyl-β-D-1-thiogalactopyranoside (IPTG), and the medium was supplemented with 0.125 mM (NH4)2Fe(SO4)2·6H2O. Cultures were incubated with shaking (250 rpm) at 18 °C for 20-24 h. Cell pellets from centrifugation at 8,000 × g for 15 min were flash frozen in liquid nitrogen and stored at -80 °C. Cells were resuspended in 50 mM potassium phosphate buffer (pH 8.0), 10 mM imidazole and 300 mM NaCl and were lysed by two passages through a French pressure cell at > 1,200 psi. After centrifugation at 22,000 × g for 20 min, the lysate supernatant was loaded onto a Ni2+-nitrilo-tris-acetate (NTA) immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM potassium phosphate buffer (pH 8.0), 10 mM imidazole and 300 mM NaCl. The column was washed with

50 mM potassium phosphate buffer (pH 8.0) containing 40 mM imidazole and 300 mM NaCl.

Protein was then eluted by washing with 50 mM potassium phosphate buffer (pH 8.0), 250 mM imidazole, 100 mM NaCl. Fractions containing the protein were pooled, and the protein was concentrated at 3,500 × g using a 10K MWCO Macrosep® Advance Centrifugal Device (Pall

127 Corporation, Port Washington, New York). In an anoxic chamber (Labmaster, MBraun, Stratham,

NH), the ADO cofactor was reduced by treatment with ≥ 10 mM sodium dithionite for 30 min at

4 °C. Buffer exchange and sodium dithionite removal were achieved using a pre-packed PD-10 desalting column (G-25 Sephadex medium, GE Healthcare) equilibrated with O2-free 50 mM sodium 2-[4-(2-hydroxyethyl)-piperazin-1-yl]ethanesulfonate (HEPES) buffer (pH 7.5), 10% glycerol. Protein aliquots were flash frozen and stored in liquid N2. Protein purity was assessed by SDS-PAGE with Coomassie staining. The concentration of ADO was determined from absorbance at 280 nm by using a molar absorption coefficient of 22,920 M-1•cm-1.(17) Iron content was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) by Mr. Henry Gong at the Pennsylvania State Materials Research Institute. The concentrations of

ADO given are of diiron cluster, calculated as half the measured iron concentration.

Preparation of Syn. 6803 Fds. The Syn. 6803 ssl0020 (petF), ssl2559, slr1828, sll1584, sll1382, ssl3044, and slr0150 genes were each individually codon-optimized for over-expression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by GeneArt (Regensburg, Germany) to give seven Syn. 6803 Fd plasmids. Ec BL21 (DE3) competent cells (Invitrogen, Carlsbad, CA) were doubly transformed with the Fd plasmid and the plasmid pDB1282, containing the Azotobacter vinelandii isc operon in a pARA13 expression vector.(70) Transformed cells with kanamycin and ampicillin resistance were cultured at 37 °C with shaking (250 rpm) in M9 minimal medium with 25 mg/L kanamycin, 150 mg/L ampicillin,

0.2% (v/v) glucose, 0.10 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM

(NH4)2Fe(SO4)2·6H2O until an OD600 of 0.3 was reached. Expression of the isc operon was induced by addition of 2 g/L arabinose and 0.20 mM L-cysteine to cell cultures. Incubation with shaking (250 rpm) at 37 °C was continued until an OD600 of 0.6-0.8 was achieved, at which time

Fd expression was induced by addition of 0.25 mM IPTG. Cultures were also supplemented with

0.125 mM (NH4)2Fe(SO4)2·6H2O at this time. They were then incubated at 18 °C for 24 h with

128 shaking (250 rpm) and then were centrifuged at 8,000 × g for 15 min. Dark brown cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cells were resuspended in 50 mM tris-

(hydroxymethyl)aminomethane (Tris)-HCl (pH 7.5) buffer, 150 mM NaCl, and 10% glycerol.

The cell suspension was lysed by two passages through a French pressure cell at >1,200 psi and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM

Tris-HCl (pH 7.5) buffer, 150 mM NaCl, and 10% glycerol. The column was washed with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl, 30 mM imidazole and 10% glycerol. The protein was then eluted by washing with 50 mM Tris-HCl (pH 7.5) buffer containing 150 mM

NaCl, 250 mM imidazole, and 10% glycerol. Fractions containing the protein were pooled and concentrated at 3,500 × g using a 10K MWCO Macrosep® Advance Centrifugal Device (Pall

Corporation, Port Washington, New York). The concentrated protein was then dialyzed against

100 equivalent volumes of 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol three times for > 6 h each. The protein was frozen in liquid N2 and stored at -80 °C.

Protein purity was assessed by SDS-PAGE with Commassie staining, and protein concentration was determined using a Direct Detect Spectrometer (EMD Millipore, Darmstadt, Germany).

[2Fe-2S] cluster occupancy was determined by EPR spin quantification of a protein sample reduced by treatment with 10 mM sodium dithionite. A frozen solution of Cu2+-EDTA was used as the spin-concentration standard.(71) The PetF concentrations quoted in all experiments are of the [2Fe-2S] cofactor. Chemically-reduced PetF was prepared by anaerobic incubation of the oxidized protein with 10 mM sodium dithionite (> 2-fold excess with respect to [PetF] cofactor) at 4 °C for 30 min. The sodium dithionite was removed and the buffer exchanged by using a PD-

10 desalting column (G-25 Sephadex medium, GE Healthcare) equilibrated with O2-free 50 mM sodium HEPES, pH 7.5.

129 Preparation of Syn. 6803 slr1643 FNR. The Syn. 6803 slr1643 (petH) gene encoding FNR was codon-optimized for over-expression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by GeneArt (Regensburg, Germany). Ec BL21

(DE3) competent cells were transformed with the FNR plasmid and selected for kanamycin resistance. Transformed cells were grown in rich Luria-Bertani medium (LB) with 50 mg/L kanamycin at 37 °C with shaking (250 rpm) until an OD600 of 0.6-0.8 was reached. Protein expression was then induced by addition of 0.25 mM IPTG, and cell cultures were incubated at

30 °C with shaking (250 rpm) for 4 h. Cultures were centrifuged at 8,000 × g for 15 min; cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cell pellets were resuspended in 50 mM Tris-HCl (pH 7.5) buffer containing 150 mM NaCl and 10% glycerol. The suspension was lysed by two passages through a French pressure cell at > 1,200 psi and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol. The column was washed with 50 mM Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 30 mM imidazole and 10% glycerol. The protein was eluted by washing with 50 mM Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 250 mM imidazole and 10% glycerol.

Fractions containing the protein were pooled and concentrated at 3,500 × g using a 10K MWCO

Macrosep® Advance Centrifugal Device (Pall Corporation, Port Washington, New York). The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-HCl (pH

7.5) buffer with 150 mM NaCl and 10% glycerol three times for > 6 h each. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with

Coomassie staining, and protein concentration was determined by using a molar absorption coefficient at 280 nm of 50,210 M-1•cm-1. FNR was reconstituted with equimolar flavin adenine dinucleotide (FAD+) and filtered to remove excess FAD+ using a 30K MWCO Amicon Ultra centrifugal filter (EMD Millipore, Merck KGaA, Darmstadt, Germany). The concentration of

130 bound FAD+ was determined from absorbance at 460 nm by using a molar absorption coefficient of 10,800 M-1•cm-1. The FNR concentrations quoted in all experiments are of the bound FAD+ cofactor.

Preparation of Dechloromonas aromatica Chlorite Dismutase (Cld). Cld was prepared as previously described.(34)

Coupled Activity Assays. Reactions of the AAR and ADO coupled system were performed at 37

°C for 1 h. Assays (0.1 mL) contained 10 µM AAR, 20 µM ADO, 0.20 mM 1-13C-stearoyl-ACP,

4 mM NADPH, 7.8 μM (spinach or Syn. 6803) Fd, and 7.8 µM (spinach or Syn. 6803) FNR.(13)

Multiple turnover assays with the ADO variants (0.20 mL) were performed at 5 °C for 5 min, containing 10 μM ADO, 2.5 mM decanal, 0.2 mM MeOPMS and 10 mM NADH. Reactions were terminated by mixing with an equal volume of a solution of 40 μM hexadecane internal standard in ethyl acetate with vortexing, and were then centrifuged at 16,000 × g for 30 min. The top organic layer containing the aldehyde substrate and alkane product was removed and analyzed on a Shimadzu GCMS-QP5000 gas chromatography mass spectrometer (GC-MS), as previously reported.(13)

Single-Turnover Assays for Product Detection. Single-turnover ADO assays were performed

II/II in the MBraun anoxic chamber. In glass vials, 30 µL of a solution containing 0.25 mM Fe2 -

2 ADO and 12.5 mM 1-[ H]-decanal in 5 mM sodium HEPES (pH 7.5) was mixed with 45 µL O2- saturated (1.8 mM O2 at 5 °C) or air-saturated (0.28 mM O2 at 25 °C) 50 mM sodium HEPES (pH

7.5) buffer by rapid injection via a gas-tight syringe. After being allowed to react for 10 s (O2-

III/III saturated) or 20 s (air-saturated) to accumulate the Fe2 -PHA intermediate (Figure 2-24), this solution was mixed with an equal volume of reductant solution. In experiments employing PetF as the reductant, the reductant solution contained chemically reduced PetF. For reactions

II/II employing the PetF/FNR/N system, the initial Fe2 -ADO•aldehyde solution contained O2-free, oxidized 0.25 mM PetF and 0.25 mM FNR, and the reductant solution contained NADPH. After

131 mixing with the reductant, the 0.15 mL reactions were rapidly terminated (after ~ 1 s) by addition of 0.15 mL of a solution of 40 µM hexadecane internal standard in ethyl acetate, with vigorous vortexing for 30 s. Samples were centrifuged at 16,000 × g for 30 min. The top organic layer containing the alkane product was removed and analyzed on a Shimadzu GCMS-QP5000, as previously reported.(13) Alternatively, the reactions were quenched in 30 μL of a solution of 2-

NPH and EDC for quantification of the formate co-product as its 2-NPH derivative, as previously described.(12, 13, 17) A 1 mM derivatized propionic acid solution (10 μL) was added to each sample as an internal standard. The samples (4 µL injections) were analyzed on a 6410 LC/MS

Agilent QQQ spectrometer using a Hamilton PRP-1 analytical column, as previously described.(12, 17)

Stopped-Flow Absorption Spectroscopy (SF-Abs) and Data Analysis. SF-Abs experiments were carried out at 5 °C in an Applied Photophysics Ltd. (Leatherhead, UK) SX20 stopped-flow spectrophotometer, which was housed in an anoxic chamber (Labmaster, MBraun, Stratham,

USA). The instrument was configured for sequential-mixing, an optical pathlength of 1 cm and data acquisition with white light and the photodiode-array (PDA) detector, unless otherwise

II/II indicated. In the first mix, an O2-free solution containing Fe2 -ADO (0.033, 0.050, 0.10, 0.20,

0.40 mM) and 10 mM of either octanal or decanal substrate was mixed with an equal volume of

O2-saturated (1.8 mM at 5 °C) 50 mM sodium HEPES buffer (pH 7.5). This solution was allowed

III/III to react for 10, 15 or 20 s to accumulate the Fe2 -PHA intermediate, according to the formation kinetics defined in single-mixing experiments employing monochromatic light and a photomultiplier tube (PMT) detector (Figure 2-24). This reaction solution was then mixed with an equal volume of a solution containing 0.10 mM chemically reduced PetF. Time-dependent absorption spectra (1000 points) were acquired on a logarithmic time base after the second mixing event. The dead-time of the stopped-flow apparatus is slightly more than 1 ms. Control

II/II experiments, in which the Fe2 -ADO•aldehyde complex was replaced in the first mixing event

132 with O2-free buffer, were carried out to define the kinetics of oxidation of PetF by O2. Additional

III/III single-mix experiments were carried out to assess the rate of reduction of the as isolated Fe2 -

ADO by PetF. Details of reaction conditions are provided in the figure legends.

1+ ΔA423-versus-time traces, monitoring PetF [2Fe-2S] oxidation in experiments wherein the ratio PetF:ADO was varied (0.5, 1, 2, 4, 6), were fit by sums of exponential functions (eq. 1 and

2) as described in the main text, corresponding to parallel first-order reactions with different amplitudes (ΔA1, ΔA2, ΔA3) and observed rate constants (k1, k2, k3), where A0 is the initial absorbance.

A A1( ) A2( ) (eq. 1)

A A1( ) A2( ) A3( ) (eq. 2)

133

III/III Figure 2-24. SF-Abs kinetic traces monitoring formation of the Fe2 -PHA intermediate (λmax = 450 nm). All experiments were performed at 5 °C in 50 mM sodium HEPES buffer, pH 7.5. (A) Effect of protein concentration: a solution of 0.1 mM (green), 0.2 mM (pink) or 0.4 mM (blue) II/II Fe2 -ADO with 10 mM octanal was mixed with an equal volume of O2-saturated buffer (1.8 mM O2). Data were acquired with monochromatic light and a PMT detector. (B) Effect of oxygen II/II concentration: a solution of 0.1 mM Fe2 -ADO with 10 mM decanal was mixed with an equal volume of either O2-saturated (1.8 mM O2, black) or air-saturated (~ 0.38 mM O2, red) buffer. Data were acquired with white light and a PDA detector. The apparent discrepancy in amplitudes is attributed to the diminished rate of intermediate formation with air-saturated buffer and the enhanced rate of photolytic decay under illumination by the intense polychromatic light. (C) II/II Effect of substrate chain-length: A solution containing 0.9 mM Fe2 -ADO, 10 μM Cld and 10 mM decanal (red) or octanal (purple) was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the final solution. Data were acquired using monochromatic light and a PMT detector. (D) Effect of high oxygen concentrations: Black trace: A solution of 1.49 mM II/II Fe2 -ADO with 16.6 mM octanal was mixed with an equal volume of O2-saturated buffer (1.8 II/II mM O2). Blue trace: A solution of 1.49 mM Fe2 -ADO with 20 mM octanal and 10 μM Cld was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the reaction. Data were acquired with monochromatic light and a PMT detector. In the experiment with O2-saturated -1 -1 buffer, ~ 80 % intermediate was generated (A450 = 0.7, ε450 = 1,200 mM ∙cm ), in good agreement with the Mössbauer analysis of a parallel sample.

Preparation of Freeze-Quenched Samples for Electron Paramagnetic Resonance (EPR) and

Mössbauer Spectroscopies. General procedures for freeze-quench (FQ) EPR and Mössbauer experiments have been published previously.(72-74) Sequential-mixing FQ-Mössbauer

134 experiments were conducted at 5 °C by using three syringes that contained solutions of: (1) 1.4

57 II/II mM Fe2 -ADO, 16.6 mM decanal, 10 μM Cld; (2) 10 mM NaClO2; and (3) 1.4 mM chemically reduced PetF. The buffer for all three solutions was 50 mM sodium HEPES, pH 7.5,

10% glycerol. The contents of syringes 1 and 2 were mixed in equal volume, and this solution

III/III was allowed to react for 30 s to generate the Fe2 -PHA intermediate (Figure 2-24). The resultant solution was then mixed with an equal volume of the syringe 3 solution, and this solution was allowed to react for 10 ms before being rapidly frozen (freeze-quenched) in isopentane at -150 °C.

Sequential-mixing FQ-EPR experiments were conducted at 5 °C by using three syringes.

In experiments with high O2 concentrations, the syringes contained solutions of: (1) 0.9 mM

II/II Fe2 -ADO, 16.6 mM octanal or decanal, and 10 μM Cld; (2) 10 mM NaClO2; and (3) chemically reduced PetF or 0.45 mM MeOPMS reduced with 0.23 mM sodium dithionite (again, all in 50 mM sodium HEPES, pH 7.5, 10% glycerol). In experiments with low O2 concentrations,

Cld was omitted from syringe 1, and syringe 2 contained O2-saturated (1.8 mM O2) 50 mM sodium HEPES, pH 7.5 buffer instead of NaClO2. In both cases, solutions 1 and 2 were mixed in

III/III equal volume, and the resultant solution was incubated for 30 s to generate the Fe2 -PHA intermediate (Figure 2-24). This solution was then mixed with an equal volume of the syringe 3 solution, and the resultant solution was incubated for varying times before being freeze-quenched in isopentane at -150 °C. Minor deviations from these general experimental procedures are noted in the appropriate text and figure legends.

Mössbauer Spectroscopy. Mössbauer spectra were recorded on spectrometers (described previously(72)) from WEB Research (Edina, MN). The spectrometer used to acquire the weak- field spectra is equipped with a Janis SVT-400 variable-temperature cryostat. The external magnetic field was applied parallel to the γ beam. All isomer shifts quoted are relative to the centroid of the spectrum of α-iron metal at room temperature. Simulations of Mössbauer spectra

135 were carried out with the WMOSS spectral analysis software (www.wmoss.org, WEB Research,

Edina, MN).

Continuous-Wave (CW) EPR Spectroscopy. X-Band (~ 9.5 GHz) EPR spectra were acquired on a Bruker ESP-300 spectrometer equipped with an ER/4102 ST resonator (Bruker), an Oxford

Instruments continuous helium flow cryostat, and an Oxford Instruments temperature controller

(ITC 502). For all experiments, quartz tubes with 3 mm inner and 4 mm outer diameter were used

(QSI). The first-derivative EPR spectra were simulated with the MATLAB-based (Mathworks)

EasySpin simulation software (http://easyspin.org/).(75)

Pulse EPR Spectroscopy. Pulse X-Band (~ 9.5 GHz) and Q-band (~ 34 GHz) field-swept EPR spectra and hyperfine sublevel correlation (HYSCORE) spectra were acquired on a Bruker

ELEXSYS E580 spectrometer equipped with a home-built intermediate frequency (IF) extension

(designed and fabricated by Dr. Alexey Silakov of Penn State University) of the SuperX-FT

X‐band bridge.(76) Field-swept electron-spin-echo-detected EPR spectra were recorded using the two-pulse echo sequence (π/2−τ–π–τ–echo). HYSCORE experiments were recorded with the standard pulse sequence (π/2-τ-π/2-t1-π-t2-π/2) and were measured at the field positions corresponding to the principal orientations of the g-tensor. Simulations of the HYSCORE spectra were carried out with the freely available MATLAB-based software KAZAN viewer

(https://sites.google.com/site/silakovalexey/kazan-viewer/) developed by Dr. Alexey Silakov

(Penn State University) and Prof. Boris Epel (University of Chicago).

Density Functional Theory (DFT) Calculations. Geometry optimizations and single-point energy calculations were carried out with the ORCA software package.(77) The structures of the peroxyl radicals considered were first geometry optimized without any restrictions. The optimized geometries were further used as inputs for the calculations of the EPR parameters (g- tensors and A-tensors). All calculations were performed with the unrestricted Hartree-Fock method employing the B3LYP density functional. The TZVP basis set was used for all atoms.

136 The conductor-like screening model (COSMO) was employed to account for electrostatic screening by the protein. A dielectric constant (ε) of 4 was assumed, consistent with previous studies.(78) Calculations were performed on both the C9H18OO• radical and its 2-thia analog.

Calculations were also performed on the R• precursor for purposes of comparison and to provide additional evidence for the favored assignment of the detected species as a peroxyl radical.

2.4 Acknowledgements

Dr. Hanne Nørgaard and Prof. Maria-Eirini Pandelia performed the initial EPR experiments characterizing the peroxyl radical. Prof. Maria-Eirini Pandelia performed data collection for the

Mössbauer experiments. Prof. Maria-Eirini Pandelia carried out the HYSCORE experiments,

EPR simulations, and DFT calculations. Prof. Wei-chen Chang synthesized the isotopically- labelled substrates. Dr. Ning Li and Dr. Doug Warui performed initial experiments using the cyanobacterial ferredoxins and ferredoxin reductase.

This material is based upon work supported by the National Science Foundation under Award

No. MCB-1122079-. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the National

Science Foundation.

2.5 References

1. Hallenbeck, P. (2012) Microbial Technoogies in Advanced Biofuels Production, Springer, New York, NY. 2. Schirmer, A., Rude, M. A., Li, X., Popova, E., and del Cardayre, S. B. (2010) Microbial biosynthesis of alkanes. Science 329, 559-562. 3. Tillman, J. A., Seybold, S. J., Jurenka, R. A., and Blomquist, G. J. (1999) Insect pheromones--an overview of biosynthesis and endocrine regulation. Insect Biochem Mol Biol 29, 481-514.

137 4. Reed, J. R., Vanderwel, D., Choi, S., Pomonis, J. G., Reitz, R. C., and Blomquist, G. J. (1994) Unusual mechanism of hydrocarbon formation in the housefly: cytochrome P450 converts aldehyde to the sex pheromone component (Z)-9-tricosene and CO2. Proc Natl Acad Sci U S A 91, 10000-10004. 5. Qiu, Y., Tittiger, C., Wicker-Thomas, C., Le Goff, G., Young, S., Wajnberg, E., Fricaux, T., Taquet, N., Blomquist, G. J., and Feyereisen, R. (2012) An insect-specific P450 oxidative decarbonylase for cuticular hydrocarbon biosynthesis. Proc Natl Acad Sci U S A 109, 14858-14863. 6. Bernard, A., Domergue, F., Pascal, S., Jetter, R., Renne, C., Faure, J. D., Haslam, R. P., Napier, J. A., Lessire, R., and Joubes, J. (2012) Reconstitution of plant alkane biosynthesis in yeast demonstrates that Arabidopsis ECERIFERUM1 and ECERIFERUM3 are core components of a very-long-chain alkane synthesis complex. Plant Cell 24, 3106-3118. 7. Warui, D. M., Pandelia, M.-E., Rajakovich, L. J., Krebs, C., Bollinger, J. M., Jr., and Booker, S. J. (2015) Efficient Delivery of Long-Chain Fatty Aldehydes from the Nostoc punctiforme Acyl-Acyl Carrier Protein Reductase to Its Cognate Aldehyde- Deformylating Oxygenase. Biochemistry 54, 1006-1015. 8. Cheesbrough, T. M., and Kolattukudy, P. E. (1984) Alkane biosynthesis by decarbonylation of aldehydes catalyzed by a particulate preparation from Pisum sativum. Proc Natl Acad Sci U S A 81, 6613-6617. 9. Dennis, M. W., and Kolattukudy, P. E. (1991) Alkane biosynthesis by decarbonylation of aldehyde catalyzed by a microsomal preparation from Botryococcus braunii. Arch Biochem Biophys 287, 268-275. 10. Schneider-Belhaddad, F., and Kolattukudy, P. (2000) Solubilization, partial purification, and characterization of a fatty aldehyde decarbonylase from a higher plant, Pisum sativum. Arch Biochem Biophys 377, 341-349. 11. Reppas, N. B., and Ridley, C. P. (2010) Methods of compositions for the recominant biosynthesis of n-alkanes, (Inc., J. U., Ed.), United States of America. 12. Krebs, C., Bollinger, J. M., Jr., and Booker, S. J. (2011) Cyanobacterial alkane biosynthesis further expands the catalytic repertoire of the ferritin-like 'di-iron- carboxylate' proteins. Curr Opin Chem Biol 15, 291-303. 13. Warui, D. M., Li, N., Nørgaard, H., Krebs, C., Bollinger, J. M., Jr., and Booker, S. J. (2011) Detection of formate, rather than carbon monoxide, as the stoichiometric coproduct in conversion of fatty aldehydes to alkanes by a cyanobacterial aldehyde decarbonylase. J Am Chem Soc 133, 3316-3319. 14. Eser, B. E., Das, D., Han, J., Jones, P. R., and Marsh, E. N. G. (2011) Oxygen- independent alkane formation by non-heme iron-dependent cyanobacterial aldehyde decarbonylase: investigation of kinetics and requirement for an external electron donor. Biochemistry 50, 10743-10750. 15. Das, D., Eser, B. E., Han, J., Sciore, A., and Marsh, E. N. (2011) Oxygen-independent decarbonylation of aldehydes by cyanobacterial aldehyde decarbonylase: a new reaction of diiron enzymes. Angew Chem Int Ed Engl 50, 7148-7152. 16. Li, N., Chang, W.-c., Warui, D. M., Booker, S. J., Krebs, C., and Bollinger, J. M., Jr. (2012) Evidence for only oxygenative cleavage of aldehydes to alk(a/e)nes and formate by cyanobacterial aldehyde decarbonylases. Biochemistry 51, 7908-7916. 17. Pandelia, M.-E., Li, N., Nørgaard, H., Warui, D. M., Rajakovich, L. J., Chang, W.-c., Booker, S. J., Krebs, C., and Bollinger, J. M., Jr. (2013) Substrate-Triggered Addition of Dioxygen to the Diferrous Cofactor of Aldehyde-Deformylating Oxygenase to Form a Diferric-Peroxide Intermediate. J Am Chem Soc 135, 15801-15812.

138 18. Matthews, M. L., Krest, C. M., Barr, E. W., Vaillancourt, F. H., Walsh, C. T., Green, M. T., Krebs, C., and Bollinger, J. M., Jr. (2009) Substrate-triggered formation and remarkable stability of the C-H bond-cleaving chloroferryl intermediate in the aliphatic halogenase, SyrB2. Biochemistry 48, 4331-4343. 19. Nanayakkara, N. P. D., and Schrader, K. K. (2011) Evaluation of free radical-generating compounds for toxicity towards the cyanobacterium Planktothrix perornata, which causes musty off-flavour in pond-raised channel catfish (Ictalurus punctatus). Aquac Res 42, 1895-1898. 20. Klahn, S., Baumgartner, D., Pfreundt, U., Voigt, K., Schon, V., Steglich, C., and Hess, W. R. (2014) Alkane Biosynthesis Genes in Cyanobacteria and Their Transcriptional Organization. Front Bioeng Biotechnol 2, 24. 21. Muller, J., Lugovskoy, A. A., Wagner, G., and Lippard, S. J. (2002) NMR structure of the [2Fe-2S] ferredoxin domain from soluble methane monooxygenase reductase and interaction with its hydroxylase. Biochemistry 41, 42-51. 22. Blazyk, J. L., Gassner, G. T., and Lippard, S. J. (2005) Intermolecular electron-transfer reactions in soluble methane monooxygenase: a role for hysteresis in protein function. J Am Chem Soc 127, 17364-17376. 23. Acheson, J. F., Bailey, L. J., Elsen, N. L., and Fox, B. G. (2014) Structural basis for biomolecular recognition in overlapping binding sites in a diiron enzyme system. Nat Commun 5. 24. Wu, C. H., Jiang, W., Krebs, C., and Stubbe, J. (2007) YfaE, a ferredoxin involved in diferric-tyrosyl radical maintenance in Escherichia coli ribonucleotide reductase. Biochemistry 46, 11577-11588. 25. Lyle, K. S., Haas, J. A., and Fox, B. G. (2003) Rapid-mix and chemical quench studies of ferredoxin-reduced stearoyl-acyl carrier protein desaturase. Biochemistry 42, 5857-5866. 26. Sobrado, P., Lyle, K. S., Kaul, S. P., Turco, M. M., Arabshahi, I., Marwah, A., and Fox, B. G. (2006) Identification of the binding region of the [2Fe-2S] ferredoxin in stearoyl- acyl carrier protein desaturase: insight into the catalytic complex and mechanism of action. Biochemistry 45, 4848-4858. 27. Zhang, J., Lu, X., and Li, J. J. (2013) Conversion of fatty aldehydes into alk (a/e)nes by in vitro reconstituted cyanobacterial aldehyde-deformylating oxygenase with the cognate electron transfer system. Biotechnol Biofuels 6, 86. 28. Wang, Q., Huang, X., Zhang, J., Lu, X., Li, S., and Li, J. J. (2014) Engineering self- sufficient aldehyde deformylating oxygenases fused to alternative electron transfer systems for efficient conversion of aldehydes into alkanes. Chem Commun 50, 4299- 4301. 29. Poncelet, M., Cassier-Chauvat, C., Leschelle, X., Bottin, H., and Chauvat, F. (1998) Targeted deletion and mutational analysis of the essential (2Fe-2S) plant-like ferredoxin in Synechocystis PCC6803 by plasmid shuffling. Mol Microbiol 28, 813-821. 30. Mazouni, K., Domain, F., Chauvat, F., and Cassier-Chauvat, C. (2003) Expression and regulation of the crucial plant-like ferredoxin of cyanobacteria. Mol Microbiol 49, 1019- 1029. 31. Kallio, P., Pasztor, A., Thiel, K., Akhtar, M. K., and Jones, P. R. (2014) An engineered pathway for the biosynthesis of renewable propane. Nat Commun 5, 4731. 32. Menon, N., Pasztor, A., Menon, B. R., Kallio, P., Fisher, K., Akhtar, M. K., Leys, D., Jones, P. R., and Scrutton, N. S. (2015) A microbial platform for renewable propane synthesis based on a fermentative butanol pathway. Biotechnol Biofuels 8, 61.

139 33. Andre, C., Kim, S. W., Yu, X. H., and Shanklin, J. (2013) Fusing catalase to an alkane- producing enzyme maintains enzymatic activity by converting the inhibitory byproduct H2O2 to the cosubstrate O2. Proc Natl Acad Sci U S A 110, 3191-3196. 34. Dassama, L. M. K., Yosca, T. H., Conner, D. A., Lee, M. H., Blanc, B., Streit, B. R., Green, M. T., DuBois, J. L., Krebs, C., and Bollinger, J. M., Jr. (2012) O(2)-evolving chlorite dismutase as a tool for studying O(2)-utilizing enzymes. Biochemistry 51, 1607- 1616. 35. Krebs, C., Dassama, L. M. K., Matthews, M. L., Jiang, W., Price, J. C., Korboukh, V., Li, N., and Bollinger, J. M., Jr. (2013) Novel approaches for the accumulation of oxygenated intermediates to multi-millimolar concentrations. Coordin Chem Rev 257, 234-243. 36. Waugh, M. W., and Marsh, E. N. G. (2014) Solvent isotope effects on alkane formation by cyanobacterial aldehyde deformylating oxygenase and their mechanistic implications. Biochemistry 53, 5537-5543. 37. Aukema, K. G., Makris, T. M., Stoian, S. A., Richman, J. E., Münck, E., Lipscomb, J. D., and Wackett, L. P. (2013) Cyanobacterial aldehyde deformylase oxygenation of aldehydes yields n-1 aldehydes and alcohols in addition to alkanes. ACS Catalysis 3, 2228-2238. 38. Paul, B., Das, D., Ellington, B., and Marsh, E. N. G. (2013) Probing the mechanism of cyanobacterial aldehyde decarbonylase using a cyclopropyl aldehyde. J Am Chem Soc 135, 5234-5237. 39. Das, D., Ellington, B., Paul, B., and Marsh, E. N. G. (2014) Mechanistic insights from reaction of alpha-oxiranyl-aldehydes with cyanobacterial aldehyde deformylating oxygenase. ACS Chem Biol 9, 570-577. 40. Kochi, J. K., and Krusic, P. J. (1969) Electron Spin Resonance of Primary Alkyl Radicals . Photolysis of Acyl Peroxides. J Am Chem Soc 91, 3940-&. 41. Ellinger, Y., Subra, R., Levy, B., Millie, P., and Berthier, G. (1975) Ab-Initio Study of Electronic-Structure and Hyperfine Coupling Properties in Simple Hydrocarbon Radicals .2. Short-Range and Long-Range Interactions in Alkyl Free-Radicals. J Chem Phys 62, 10-29. 42. Fessenden, R. W., and Schuler, R. H. (1963) Electron Spin Resonance Studies of Transient Alkyl Radicals. J Chem Phys 39, 2147-&. 43. Reddy, S. G., Wong, K. K., Parast, C. V., Peisach, J., Magliozzo, R. S., and Kozarich, J. W. (1998) Dioxygen inactivation of pyruvate formate-lyase: EPR evidence for the formation of protein-based sulfinyl and peroxyl radicals. Biochemistry 37, 558-563. 44. Nelson, M. J., Cowling, R. A., and Seitz, S. P. (1994) Structural characterization of alkyl and peroxyl radicals in solutions of purple lipoxygenase. Biochemistry 33, 4966-4973. 45. Adrait, A., Ohrstrom, M., Barra, A. L., Thelander, L., and Gräslund, A. (2002) EPR studies on a stable sulfinyl radical observed in the iron-oxygen-reconstituted Y177F/I263C protein R2 double mutant of ribonucleotide reductase from mouse. Biochemistry 41, 6510-6516. 46. Nelson, M. J., Seitz, S. P., and Cowling, R. A. (1990) Enzyme-bound pentadienyl and peroxyl radicals in purple lipoxygenase. Biochemistry 29, 6897-6903. 47. Nelson, M. J., and Cowling, R. A. (1990) Observation of a Peroxyl Radical in Samples of Purple Lipoxygenase. J Am Chem Soc 112, 2820-2821. 48. Becker, D., Yanez, J., Sevilla, M. D., Alonsoamigo, M. G., and Schlick, S. (1987) An Electron-Spin-Resonance Investigation of the Motion of Lipid Peroxyl Radicals. J Phys Chem 91, 492-496.

140 49. Yanez, J., Sevilla, C. L., Becker, D., and Sevilla, M. D. (1987) Low-Temperature Autoxidation in Unsaturated Lipids - an Electron-Spin-Resonance Study. J Phys Chem 91, 487-491. 50. Neta, P., Huie, R. E., and Ross, A. B. (1990) Rate Constants for Reactions of Peroxyl Radicals in Fluid Solutions. J Phys Chem Ref Data 19, 413-513. 51. (1997) Peroxyl Radicals, John Wiley & Sons Ltd, West Sussex, England. 52. Stark, M. S. (2000) Addition of peroxyl radicals to alkenes and the reaction of oxygen with alkyl radicals. J Am Chem Soc 122, 4162-4170. 53. Bolsman, T. A. B. M., and Brouwer, D. M. (1978) Reversibility of the Formation of Peroxyl Radicals from Oxygen and Alkyl Radicals. Recl Trav Chim Pay B 97, 320-320. 54. Gilbert, D. L., and Colton, C. A. (2002) In Reactive Oxygen Species in Biological Systems: An Interdisciplinary Approach, Springer, US. 55. Sahlin, M., Cho, K. B., Potsch, S., Lytton, S. D., Huque, Y., Gunther, M. R., Sjoberg, B. M., Mason, R. P., and Gräslund, A. (2002) Peroxyl adduct radicals formed in the iron/oxygen reconstitution reaction of mutant ribonucleotide reductase R2 proteins from Escherichia coli. J Biol Inorg Chem 7, 74-82. 56. Zhang, W., Wong, K. K., Magliozzo, R. S., and Kozarich, J. W. (2001) Inactivation of pyruvate formate-lyase by dioxygen: defining the mechanistic interplay of glycine 734 and cysteine 419 by rapid freeze-quench EPR. Biochemistry 40, 4123-4130. 57. Stubbe, J., and van der Donk, W. A. (1998) Protein radicals in enzyme catalysis. Chem Rev 98, 705-762. 58. Bleifuss, G., Kolberg, M., Potsch, S., Hofbauer, W., Bittl, R., Lubitz, W., Gräslund, A., Lassmann, G., and Lendzian, F. (2001) Tryptophan and tyrosine radicals in ribonucleotide reductase: A comparative high-field EPR study at 94 GHz. Biochemistry 40, 15362-15368. 59. Svistunenko, D. A., and Cooper, C. E. (2004) A new method of identifying the site of tyrosyl radicals in proteins. Biophys J 87, 582-595. 60. van Gastel, M., Lubitz, W., Lassmann, G., and Neese, F. (2004) Electronic structure of the cysteine thiyl radical: a DFT and correlated ab initio study. J Am Chem Soc 126, 2237-2246. 61. Persson, A. L., Sahlin, M., and Sjoberg, B. M. (1998) Cysteinyl and substrate radical formation in active site mutant E441Q of Escherichia coli class I ribonucleotide reductase. J Biol Chem 273, 31016-31020. 62. Kolberg, M., Bleifuss, G., Gräslund, A., Sjoberg, B. M., Lubitz, W., Lendzian, F., and Lassmann, G. (2002) Protein thiyl radicals directly observed by EPR spectroscopy. Arch Biochem Biophys 403, 141-144. 63. Gilbert, B. C. (1990) In Structure and Reaction Mechanisms in Sulphur-Radical Chemistry Revealed by E.S.R. Spectroscopy in Sulphur-Centered Reactive Intermediates in Chemistry and Biology, Springer, New York, USA. 64. Stegmann, H. B. (1988) 10.1.2 Sulfinyl radicals, In Radicals Centered on Heteroatoms with Z > 7 and Selected Anion Radicals I: 10.1.2 Sulfinyl radicals (Fischer, H., Ed.), Springer Berlin Heidelberg. 65. Gordy, W. (1980) Intrepretation of Nuclean Coupling in Oriented Free Radicals in Theory and Applications of Electron Spin Resonance, John Wiley and Sons, New York. 66. Buer, B. C., Paul, B., Das, D., Stuckey, J. A., and Marsh, E. N. G. (2014) Insights into Substrate and Metal Binding from the Crystal Structure of Cyanobacterial Aldehyde Deformylating Oxygenase with Substrate Bound. ACS Chem Biol.

141 67. Jia, C., Li, M., Li, J., Zhang, J., Zhang, H., Cao, P., Pan, X., Lu, X., and Chang, W. (2015) Structural insights into the catalytic mechanism of aldehyde-deformylating oxygenases. Protein Cell 6, 55-67. 68. Hayashi, Y., Yasugi, F., and Arai, M. (2015) Role of cysteine residues in the structure, stability, and alkane producing activity of cyanobacterial aldehyde deformylating oxygenase. PLoS One 10, e0122217. 69. Khara, B., Menon, N., Levy, C., Mansell, D., Das, D., Marsh, E. N. G., Leys, D., and Scrutton, N. S. (2013) Production of propane and other short-chain alkanes by structure- based engineering of ligand specificity in aldehyde-deformylating oxygenase. ChemBioChem 14, 1204-1208. 70. Johnson, D. C., Unciuleac, M. C., and Dean, D. R. (2006) Controlled expression and functional analysis of iron-sulfur cluster biosynthetic components within Azotobacter vinelandii. J Bacteriol 188, 7551-7561. 71. Hagen, W. R. (2006) EPR spectroscopy as a probe of metal centres in biological systems. Dalton Trans, 4415-4434. 72. Price, J. C., Barr, E. W., Tirupati, B., Bollinger, J. M., Jr., and Krebs, C. (2003) The first direct characterization of a high-valent iron intermediate in the reaction of an alpha- ketoglutarate-dependent dioxygenase: A high-spin Fe(IV) complex in taurine/alpha- ketoglutarate dioxygenase (TauD) from Escherichia coli. Biochemistry 42, 7497-7508. 73. Jiang, W., Hoffart, L. M., Krebs, C., and Bollinger, J. M., Jr. (2007) A manganese(IV)/iron(IV) intermediate in assembly of the manganese(IV)/iron(III) cofactor of Chlamydia trachomatis ribonucleotide reductase. Biochemistry 46, 8709- 8716. 74. Ravi, N., Bollinger, J. M., Jr., Huynh, B. H., Edmondson, D. E., and Stubbe, J. (1994) Mechanism of Assembly of the Tyrosyl Radical-Diiron(III) Cofactor of Escherichia-Coli Ribonucleotide Reductase .1. Mossbauer Characterization of the Diferric Radical Precursor. J Am Chem Soc 116, 8007-8014. 75. Stoll, S., and Schweiger, A. (2006) EasySpin, a comprehensive software package for spectral simulation and analysis in EPR. J Magn Reson 178, 42-55. 76. Reijerse, E., Lendzian, F., Isaacson, R., and Lubitz, W. (2012) A tunable general purpose Q-band resonator for CW and pulse EPR/ENDOR experiments with large sample access and optical excitation. J Magn Reson 214, 237-243. 77. Neese, F. (2012) The ORCA program system. Wiley Interdisciplinary Reviews- Computational Molecular Science 2, 73-78. 78. Gilson, M. K., and Honig, B. (1988) Calculation of the total electrostatic energy of a macromolecular system: solvation energies, binding energies, and conformational analysis. Proteins 4, 7-18.

142

Chapter 3

A new microbial pathway for organophosphonate degradation catalyzed by two previously misannotated non-heme-iron enzymes

Assignment of biochemical functions to hypothetical proteins encoded within the growing number of fully sequenced genomes is challenged by functional diversification seen in many known protein structural superfamilies. This diversification, which is particularly common for metalloenzymes, renders functional annotations founded solely on sequence and domain similarities unreliable and often erroneous. Definitive biochemical characterization to delineate functional sub-groups within these superfamilies will aid in improving bioinformatic approaches to assign function. We describe here the structural and functional characterization of two non- heme iron oxygenases, TmpA and TmpB, which are encoded by a genomically clustered pair of genes found in > 350 bacterial organisms. TmpA and TmpB are functional homologues of the pair of enzymes (PhnY and PhnZ) recently shown to degrade 2-aminoethylphosphonate, but instead act on its naturally occurring, N-trimethylated analogue, 2- trimethylaminoethylphosphonate (TMAEP). The iron(II)- and 2-(oxo)glutarate-dependent oxygenase TmpA was found to hydroxylate TMAEP, despite its universal annotation as a γ- butyrobetaine dioxygenase. The hydroxylated product, (R)-1-hydroxy-2- trimethylaminoethylphosphonate ((R)-OH-TMAEP), then serves as the substrate for the second enzyme TmpB. In contrast to its presumed phosphohydrolytic activity, TmpB is an HD-domain oxygenase that activates O2 using a mixed-valent diiron cofactor to enact oxidative cleavage of the C–P bond in OH-TMAEP to produce glycine betaine and phosphate. The high specificity of both TmpA and TmpB for their N-trimethylated substrates implies dedication of this pathway to degradation of TMAEP, which was not previously known to be subject to microbial consumption.

It adds to the growing number of examples of microbial metabolism of organophosphonates,

143 which enable the assimilation of phosphorus, carbon, and nitrogen in nutrient-limited ecological niches.

3.1 Introduction

Metalloenzyme superfamilies are defined by their highly conserved protein scaffolds that harbor their distinct type of metallocofactor. Remarkably, individual members can promote different reaction types on structurally dissimilar substrates (1-4). This evolutionary repurposing, of which there are now many examples (5-7), limits the utility of domain conservation and sequence similarity for de novo prediction of metalloenzyme function. This problem is not unique to metalloenzymes; an estimated 98% of all gene ontology annotations are algorithmically generated and not subsequently curated, leading to misannotations that are propagated and amplified as the number of fully sequenced genomes rapidly grows (8-10). Given the established roles of metalloenzymes in central life processes (1, 4), misannotation of these proteins in particular could be hindering discovery of novel functions with important biomedical or environmental implications.

The broad functional diversity within a metalloprotein superfamily is exemplified by the iron(II)- and 2-(oxo)glutarate-dependent (Fe/2OG) oxygenases (2), with ~50,000 members classified by in the InterPro database. Members of this superfamily share a β-sandwich, cupin structural fold (11) and contain a largely conserved sequence motif, HxD/Ex(40-160)H, that contributes the so-called “facial triad” of protein ligands that bind the mononuclear non-heme

Fe(II) cofactor characteristic of the enzyme family (12, 13). Fe/2OG enzymes activate dioxygen through coupling to the two-electron oxidative decarboxylation of 2OG to succinate and CO2.

This reaction generates an FeIV-oxo (ferryl) complex (14) which, in almost all known cases, extracts two electrons from the primary substrate that balance the four-electron reduction of O2.

144 Whereas hydroxylation is the most common outcome, halogenation, desaturation, and cyclization reactions also occur according to the same functional logic and within this single, conserved protein architecture (2). Reaction prediction is further complicated by the wide range of substrates for these enzymes, from simple small molecules to complex natural products to large macromolecules, such as proteins and nucleic acids. Even high-resolution structures generally will not reveal the substrate identities, owing to the conformational disorder of substrate- interacting loops often seen in the absence of the primary substrate (11).

Another large, but mostly functionally uncharacterized class of metalloenzymes is the

HD-domain structural superfamily (4), with > 75,000 members in the InterPro database. The namesake domain of the superfamily has a characteristic HxaHDxbD sequence motif, which contributes the residues to coordinate a divalent metal cofactor. At the time this superfamily was first recognized, all functionally characterized members were known to promote hydrolysis of phosphate esters or anhydrides (15-18), leading to the automatic annotation of members emerging thereafter as putative phosphohydrolases (4). However, biochemical studies over the last decade on several HD-domain proteins have expanded the range of known metallocofactor types and catalytic activities associated with the superfamily. Most notably, two examples of HD-domain proteins with oxygenase activities – myo-inositol oxygenase (MIOX) and (R)-1-hydroxy-2- aminoethylphosphonate oxygenase (PhnZ) – have been described (19-21). These enzymes harbor histidine- and carboxylate-coordinated non-heme diiron cofactors, which operate in the mixed- valent Fe2(II/III) oxidation state for O2 activation to enact four-electron oxidation reactions that cleave a C–C or C–P bond of their primary substrates (20, 22). The capacity of an HD-domain protein to bind two metals in the active site can be inferred from the observation of two additional, conserved histidine residues (which serve as metal ligands) in the primary structure between the aspartate residues of the canonical domain (HxaHDxbHxcHxdD) (20, 23, 24). The dinuclear nature of the metallocofactor in a given HD-domain protein can thus be predicted from

145 sequence, but this recognition is of limited value in the goal of assigning function because a subset of proteins harboring dinuclear cofactors use them to promote hydrolytic reactions (25-27).

In other words, the nuclearity of the cofactor does not, by itself, serve to distinguish the two major functional classes currently associated with the superfamily, i.e. hydrolases and oxygenases. Indeed, it is now clear that the structural architecture has adapted to accommodate one, two, or even three metal ions (28) that can be used to promote either simple hydrolytic or complex redox outcomes, invalidating the persistent practice of automatic annotation of newly discovered superfamily members as phosphohydrolases.

The functional diversity within structurally conserved superfamilies, exemplified by these two cases, strongly necessitates supplementation of sequence- and domain-based annotation with experimental and/or bioinformatic information in order to more accurately predict the function

(both the substrate and the transformation catalyzed) of a hypothetical metalloenzyme (29-32). In bacterial systems, genomic context can be a valuable tool for deducing the identities of substrates and/or products of an encoded protein of unknown function (33-36), due to the frequent clustering of genes that encode enzymes of the same metabolic pathway in a transcription- regulated operon. Therefore, one can infer that the product of one enzyme will serve as the substrate for another when the corresponding genes are genomically co-localized. A relevant example is the recently discovered pathway for degradation of the most abundant environmental organophosphonate, 2-aminoethylphosphonate (2-AEP). Two syntenic genes encode the enzymes, PhnY and PhnZ, that enable many marine bacteria to catabolize this compound (21).

PhnY is an Fe/2OG oxygenase that hydroxylates C1 of 2-AEP to generate (R)-1-hydroxy-2- aminoethylphosphonate ((R)-OH-AEP) (21). This product serves as the substrate for PhnZ, an

HD-domain diiron oxygenase, that cleaves the C–P bond of (R)-OH-AEP to yield the final degradation products, phosphate and glycine (20, 21, 23).

146 In this work, we identified a similar bicistronic operon, encoding an Fe/2OG oxygenase and an HD-domain protein, within the genomes of > 350 bacteria. In almost all cases, the Fe/2OG oxygenase, designated TmpA, had been annotated as a γ-butyrobetaine dioxygenase (BBOX).

BBOX enzymes catalyze hydroxylation at the C3 position of γ-butyrobetaine (γbb) to yield L- carnitine in the last step of its biosynthesis in eukaryotes (37) and in the first step of bacterial γbb degradation for carbon and nitrogen assimilation (38, 39). The HD-domain protein, designated

TmpB, was annotated either as a phosphohydrolase or, in a few cases, a peptidase. However, the notable sequence similarity of TmpB to the HD-domain oxygenase PhnZ (~ 30% identity), including conservation of the extended HxaHDxbHxcHxdD dinuclear ligand sequence motif, in addition to its analogous genomic synteny with an Fe/2OG oxygenase led us to suspect that

TmpB might actually be an uncharacterized HD-domain mixed-valent diiron oxygenase (HD-

MVDO). In this study, we reassigned the functions of this pair of proteins encoded in the genome of the marine bacterium Leisingera caerulea (Lc). TmpA and TmpB catalyze a previously unidentified pathway for degradation of 2-trimethylaminoethylphosphonate (TMAEP), a naturally occurring organophosphonate. TmpA hydroxylates TMAEP at the C1 position, thereby providing the substrate for TmpB, which promotes the O2-dependent oxidative cleavage of the C–

P bond to yield glycine betaine and phosphate. Biochemical and spectroscopic evidence demonstrate that TmpA and TmpB are highly specific for their N-methylated substrates and thus are not likely to overlap functionally with the PhnY/Z pair that degrades the corresponding unmethylated compound. The x-ray crystal structures of TmpA and TmpB rationalize this selectivity, while also confirming their remarkable similarities to BBOX and PhnZ, respectively, as anticipated from their sequence similarity that led to their misannotation. The biochemical and structural characterization of these enzymes, together with their genomic context, can be used to refine functional predictions for other uncharacterized Fe/2OG and HD-domain enzymes.

147 3.2 Discovery of the TMAEP hydroxylation activity of TmpA.

The Lc TmpA protein studied herein shares 26% sequence identity with Homo sapiens

(Hs) BBOX and 33% with Pseudomonas sp. AK-1 (Ps) BBOX, the only characterized prokaryotic orthologue (38, 40-42). To evaluate the annotation of TmpA as a BBOX, the enzyme was tested in a multiple-turnover reaction for the ability to hydroxylate γbb. TmpA was found neither to consume γbb nor to produce the hydroxylated product, L-carnitine. By contrast, the positive-control enzyme, Ps BBOX, completely converted γbb to L-carnitine under the same experimental conditions (Figure 3-1).

Figure 3-1. Assessment of the γbb hydroxylation activities of TmpA and Ps BBOX. LC-MS chromatograms monitoring γbb (146 m/z, black) and L-carnitine (162 m/z, blue) after a 4 h incubation of 0.01 mM TmpA or PsBBOX, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM sodium L- ascorbate, 1 mM 2OG and 1 mM γbb. The expected conversion is depicted at the top.

The verified γbb hydroxylation activity of Ps BBOX and complete absence of any such activity for TmpA prompted further scrutiny and comparison of their primary structures. A sequence similarity network (SSN) (43, 44) was constructed (Appendix B, Figure S1) for a subset of the Fe/2OG superfamily that included BBOX sequences and TmpA-like sequences.

This subset also included the characterized enzyme ε-trimethyllysine hydroxylase (TMLH), which catalyzes the hydroxylation of free ε-trimethyllysine as the first step in eukaryotic L-

148 carnitine biosynthesis (37, 45). TMLH, a close homologue of BBOX, was used as a rational benchmark for selection of a minimum pairwise alignment score that generates substrate-specific isofunctional clusters, (i.e., groups of proteins that act on the same substrate). Isolation of the isofunctional TMLH cluster from the BBOX cluster was achieved at a pairwise alignment score of < 10-65 (Appendix B, Figure S2). At this threshold, the BBOX-like and TmpA-like sequences segregate into distinct sub-clusters, with the nodes corresponding to the characterized Ps BBOX and Lc TmpA proteins no longer connected via an edge vector (i.e., they have a pairwise alignment score > 10-65). This analysis is consistent with the experimental observation that members of the two sub-clusters act on different substrates. In a network constructed from only bacterial sequences and clustered using a slightly greater stringency (see Supporting Information;

Appendix B, Figure S3), the clusters containing nodes for Ps BBOX and Lc TmpA become completely isolated with a clustering pattern that reflects genomic context (Figure 3-2). Yellow nodes, which generally cluster with Lc TmpA, represent proteins that are encoded adjacent to a gene for a TmpB-like HD-domain protein and often downstream of a gene encoding a LysR-type transcription regulator (Figure 3-2, orange box). Conversely, genes encoding proteins represented by grey and green nodes lack the neighboring gene to encode for an HD-domain protein. Proteins represented by the green nodes are encoded within genomic operons that possess similar genes to those of the γbb degradation operon characterized in Ps. In addition to the containing the gene that encodes the Ps BBOX enzyme, this operon encode proteins involved in betaine uptake and further degradation of L-carnitine to glycine betaine (38) (Figure 3-2, green box). Thus, the proteins represented by green nodes are concluded to be authentic BBOX enzymes with γ-butyrobetaine hydroxylation activity that enable γbb assimilation. The clear dichotomy between the genomic context of TmpA proteins and of those clustered with Ps BBOX is consistent with their use of different substrates, as experimentally observed for Lc TmpA.

149

Figure 3-2. Sequence similarity network (SSN) illustrating sequence divergence of BBOX- and TmpA-like proteins. The clusters of bacterial protein sequences shown here are derived from the IPR003819 SSN (see Supporting Information, Figures S1-3). The nodes represent protein sequences with > 90% identity. Edges between the nodes represent pairwise alignment scores of < 10-83 (corresponding to ~ 37% sequence identity). The gene operons in the yellow and green boxes correspond to the genomic context of proteins represented by nodes colored yellow and green, respectively. The large gold circle and dark green diamond represent Lc TmpA and Ps BBOX, respectively. Orange nodes represent predicted fusion proteins with both TmpA- and TmpB-like domains. Diamond shaped nodes represent sequences with an N-terminal Zn(II)- binding motif. Other designations are: LysR, LysR-type transcription regulator; AraC, AraC-type transcription regulator; CDH, carnitine dehydrogenase; HCT, 4-hydroxybenzoyl-CoA thioesterase; HCD, 3-hydroxybutyryl-CoA dehydrogenase; ACT, Acetyl-CoA acetyltransferase/thiolase; ABC, ABC-type glycine/betaine transporter.

150 We initially predicted that the substrate for TmpA might be structurally similar to the γbb substrate of BBOX and, therefore, tested various analogues of that molecule. None of the analogues most similar to the γbb structure that possess both the trimethylammonium and carboxylate groups were consumed by TmpA (Appendix B, Figure S6). We noted that the

BBOX active site residues that interact with the quaternary amine of γbb (vide infra) are, in fact, conserved (or conservatively substituted) in TmpA sequences. We therefore considered that the

TmpA substrate would retain this structural element but have a different substituent replacing the carboxylate group. TmpA was found to consume the quaternary ammonium-containing molecule phosphocholine in the presence of its presumptive co-substrates, 2OG and O2 (Appendix B,

Figure S7). The identical genomic synteny of the TmpA/B pair to the PhnY/Z pair suggested the possibility that the former might also process a phosphonate-containing compound like the latter.

Thus, the naturally occurring phosphonate analogue of phosphocholine, 2- trimethylaminoethylphosphonate (TMAEP), was tested and efficient consumption was indeed observed under the same conditions. In the presence of excess 2OG and O2, TMAEP was completely converted to a product with a distinct 31P-NMR chemical shift and an increased m/z of

+16 relative to that of the substrate, suggestive of substrate hydroxylation (Figure 3-3). The structure of the purified product was determined by NMR characterization to be 1-hydroxy-2- trimethylaminoethylphosphonate (OH-TMAEP) (Appendix B, Figure S5). The substrate hydroxylation was dependent on the presence of TmpA, 2OG, and O2 and tight coupling between succinate production and TMAEP consumption was observed, with a stoichiometry close to the theoretical value of 1.0 (Appendix B, Figure S8). On the other hand, under analogous conditions, no consumption of TMAEP was observed upon incubation with PhnY (Figure 3-3), which natively hydroxylates the corresponding primary amine analogue, 2-AEP. Ps BBOX was able to convert only a small fraction (< 10%) of the TMAEP present during the course of a 4-h incubation (Figure 3-3). TmpA activity was screened with a panel of other common natural

151 product and synthetic phosphonates to examine whether it could process a variety of phosphonate molecules. TmpA was unable to consume the vast majority of these compounds, excepting 2-

AEP and 3-aminopropylphosphonate (the latter not known to occur in nature), which only supported minimal activity (Appendix B, Figure S9, S12).

Figure 3-3. Testing the activities of Fe/2OG oxygenases toward the TMAEP compound. The expected chemical transformation is shown at the top. (Left) 31P-NMR spectra of the same reactions as in panel A detecting TMAEP (16.8 ppm) and OH-TMAEP (14.0 ppm). (Right) LC- MS chromatograms monitoring TMAEP (168 m/z, black) and OH-TMAEP (184 m/z, red) after a 4-h incubation of 0.01 mM TmpA, PhnY or PsBBOX with 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L- ascorbate, 3 mM 2OG and 2 mM TMAEP.

3.3 TmpA demonstrates substrate specificity for TMAEP.

The capacities of TMAEP to bind to TmpA and trigger its reaction with O2 were probed by UV-visible absorption spectroscopy to provide a foundation for assessing the enzyme’s specificity for this substrate. The ternary TmpA•FeII•2OG complex exhibited a metal-to-ligand charge transfer (MLCT) band with a peak at 530 nm (Figure 3-4A), as previously observed for other Fe/2OG hydroxylases (46). Reaction of this complex with limiting O2 resulted in only a very sluggish reaction marked by slow development of absorption centered at ~ 350 nm

(Appendix B, Figure S10). This change likely reflects unproductive oxidation of the FeII cofactor

152 to FeIII, as observed for other Fe/2OG enzymes (14, 47). Addition of TMAEP to an anoxic solution of the ternary TmpA•FeII•2OG complex elicited additional structure in the MLCT band as well as small but reproducible hyper- and hypsochromic shifts (Figure 3-4A) (48), reflecting the effect of TMAEP binding on the geometry and electronic structure of the cofactor, i.e. dissociation of the axial water ligand to create a square-pyramidal Fe(II) center (49, 50). Reaction of the resultant quaternary complex with limiting O2 (at 5 °C) resulted in rapid, transient changes reflective of a single enzyme turnover (Figure 3-4B). The 520-nm feature of the reactant complex decays and redevelops (Figure 3-4C), signifying oxidation of the Fe(II) center that a species that lacks this chromophore and reformation of the Fe(II)•2OG complex since 2OG is provided in excess. Using a kinetic model of ABC where B is evaluated as a function of time, the observed rates of decay and reformation were determined to be 51 ± 5 s-1 and 2.7 ± 0.9 s-1, respectively. Additionally, a feature with a maximum at ~ 318 nm forms within the dead time of the instrument (3 ms) and decays with an observed rate constant of 65 ± 3 s-1 (Figure 3-4C), using the same kinetic model. This transient UV absorption most likely reflects accumulation of the ferryl intermediate, as has been concluded from similar observations on other Fe/2OG oxygenases (14, 47, 51).

153

Figure 3-4. Ultraviolet-visible absorption data showing binding of TMAEP to TmpA•FeII•2OG and triggering of O2 activation. (A) Difference absorption spectra associated with binding of 2OG II (5 mM) to the TmpA•Fe complex [0.6 mM TmpA, 0.5 mM (NH4)2Fe(SO4)2] in the absence of substrate (black) and presence of 5 mM TMAEP (red). (B) Stopped-flow absorption spectra acquired after mixing at 5 °C of a solution of 1.2 mM TmpA, 1 mM (NH4)2Fe(SO4)2, 10 mM 2OG and 10 mM TMAEP with an equal volume of air-saturated 50 mM sodium HEPES buffer, pH 7.5 (~ 0.4 mM O2 at 5 °C). The inset shows the absorption spectra at indicated time points after subtracting the spectrum of the anoxic TmpA•FeII•2OG•TMAEP complex. (C) Kinetic traces at 318 nm (blue dots) and 519 nm (red dots), which are attributed to primarily the decay of the ferryl intermediate and the decay and reformation of the TmpA•Fe(II)•2OG•TMAEP reactant complex, respectively. The solid black lines are regression fits to the data, as described in the Methods Section.

Close structural analogues of TMAEP were tested for their ability to bind to TmpA, trigger O2 activation, and serve as substrates in multiple-turnover activity assays. Addition of compounds with a hydroxyl, carboxylate or sulfonate group in place of the phosphonate (choline,

γbb, γbb-3, taurine, γ-aminobutyric acid and β-alanine) did not perturb the absorption spectrum of the TmpA•FeII•2OG ternary complex (Appendix B, Figure S11), suggesting that they fail to bind with affinities comparable to that of TMAEP. These analogues also did not trigger the reaction with O2, demonstrated by their failure to promote rapid loss of the ~ 520-nm feature of the reactant complex and development of the transient absorption feature at ~ 318 nm (Appendix B,

Figure S11). Correspondingly, these substrates were not consumed after a 4-h incubation with

TmpA, FeII, and co-substrates (Appendix B, Figure S12).

Analogues that retain the phosphonate moiety, but have reduced degrees of N- methylation were also examined as substrates for TmpA and found to be poorer substrates for

154 TmpA than TMAEP. Addition of 2-dimethylaminoethylphosphonate (DMAEP) to the ternary complex induced a similar shift in the absorption spectrum as that caused by TMAEP (Figure 3-

5A, blue). Subsequent mixing of the quaternary complex formed in the presence of DMAEP with limiting O2 resulted in the spectral changes characteristic of a single turnover as observed with

TMAEP. However, the observed rate constant for decay of the reactant complex was six times

-1 less (kobs = 9 ± 3 s at 5 °C) (Figure 3-5B, blue) than that observed with TMAEP. Detection of a species with an increased m/z of +16 relative to that of the substrate in LC-MS experiments

(Appendix B, Figure S12) and the presence of a new peak with a distinct 31P-NMR chemical shift (Figure 3-5C) confirmed hydroxylation of the dimethylated analogue. In multiple-turnover

-1 reactions with low concentrations of TmpA, the initial rate with DMAEP (v0 = 0.43 s at 3 °C)

-1 was 1.6-fold less than with TMAEP (v0 = 0.67 s at 3 °C). Such a modest difference in rate compelled us to directly assess the enzyme’s preference for the N-trimethylated versus the N- dimethylated substrate in a competition assay with the two compounds present in equal concentration. TmpA exclusively consumed TMAEP at early incubation times and began to process DMAEP only after TMAEP was almost entirely depleted (Figure 3-6B), consistent with a preference for the trimethylammonium substrate. With the corresponding non-N-methylated compound, 2-AEP, TmpA was found to carry out fewer than 10 turnovers after a 4-h incubation with excess 2OG and O2 (Figure 3-5C and S12). Accordingly, addition of 2-AEP to the reactant complex elicited a less pronounced shift in its absorption spectrum (Figure 3-5A, green) and no detectable triggering of the O2 reaction was observed (Figure 3-5B, green), suggesting that the N- methyl groups are important for tight binding in the active site.

155

Figure 3-5. Evaluating activity of TmpA toward TMAEP analogues with varying degrees of N- methylation. (A) Absorption difference spectra caused by binding of 2OG (5 mM) to the II TmpA•Fe complex [0.6 mM TmpA, 0.5 mM (NH4)2Fe(SO4)2] in the absence of a substrate (black) and in the presence of 5 mM TMAEP (red), DMAEP (blue) or 2-AEP (green). (B) Kinetic traces at 519 nm after mixing of the solutions described in panel A with air-saturated 50 mM sodium HEPES buffer, pH 7.5 (~ 0.2 mM O2 final at 5 °C). The solid black lines are non-linear regression fits to the data described in the Experimental Section. (C) 31P-NMR spectra of reaction samples after a 4 h incubation of 10 μM TmpA, 20 μM (NH4)2Fe(SO4)2, 200 μM ascorbate, 3 mM 2OG and 2 mM TMAEP (red), DMAEP (blue), or 2-AEP (green). Substrate standards are shown in gray. [TMAEP = 16.8 ppm  14.0 ppm; DMAEP = 17.7 ppm  14.3 ppm; 2-AEP = 18.9 ppm  15.0 ppm]

As noted previously, TmpA degrades the phosphoester analogue of TMAEP, phosphocholine (PC). The reaction produces phosphate and glycine betaine aldehyde, the former determined by its 31P-NMR chemical shift and the latter by its m/z in LC-MS analysis (Appendix

B, Figure S7). The observed products presumably result from hydroxylation at the C1 position followed by elimination of phosphate. This reaction is robustly catalytic, such that a 1-h incubation of PC with TmpA in the presence of excess 2OG and O2 resulted in its complete consumption, as observed with TMAEP under identical conditions (Figure 3-6C). However, the

156

-1 initial rate was found to be nine-fold less with PC (v0 = 0.074 s at 3 °C) than with TMAEP.

Additionally, no changes to the optical spectrum were seen upon reaction of the presumptive

II TmpA•Fe •2OG•PC quaternary complex with O2 (Appendix B, Figure S11), suggesting that the initial O2-addition step is slow compared to subsequent events leading to reformation of the reactant complex. In a competition experiment, TMAEP is first consumed to completion, at which point PC is then slowly consumed (Figure 3-6D), clearly demonstrating a preference for the phosphonate compound. Furthermore, the TmpA reaction with PC is inhibited in this competition assay compared with a reaction containing solely PC (Figure 3-6C,D). The inhibition was presumed to be caused by the TMAEP product, OH-TMAEP, which was further confirmed by inclusion of purified OH-TMAEP at a high concentration in a reaction with PC as the substrate (Figure 3-6C).

157

Figure 3-6. Assays testing TmpA activity toward TMAEP in competition with alternative substrates, DMAEP and PC. All reactions were performed at 3 °C and contained 0.02 mM TmpA, 0.03 mM (NH4)2Fe(SO4)2, 0.4 mM ascorbate, 6 mM 2OG and 2 mM of each substrate. Each panel shows product formation over time, monitored by both LC-MS and 31P-NMR. (A) Control reactions containing either TMAEP (black) or DMAEP (blue). (B) Competition reactions containing both TMAEP and DMAEP. (C) Control reactions containing either TMAEP (black) or PC (red). Open red circles correspond to product detected from a reaction containing PC and 2 mM OH-TMAEP. (D) Competition reactions containing both TMAEP and PC. Products from this reaction were monitored by 31P-NMR only because PC and OH-TMAEP have the same m/z and similar retention times.

3.4 Structural basis for TmpA substrate specificity.

The x-ray crystal structure of the TmpA•FeII•2OG complex was solved to 1.73 Å resolution by molecular replacement with the C-terminal catalytic domain of the Hs BBOX structure (PDB accession code 3O2G) (52) as the search model. Structures were also solved with the TMAEP substrate (1.70 Å resolution) and OH-TMAEP product (1.78 Å resolution) bound to the protein. As forecast by the sequence similarity that led to the misannotation of TmpA as a

BBOX, the two proteins have a number of structural features in common. The overall topologies

158 are remarkably similar (Figure 3-7), with a root-mean-squared standard deviation (rmsd) of 2.5 Å for the Cα-atoms over 355 residues (53). The tertiary structure of the TmpA monomer consists of a defined N-terminal domain (residues 1-96), a short linker (97-100), and a C-terminal domain comprised of an eight-stranded β-sandwich, known as the cupin fold, that represents the characteristic catalytic domain of Fe/2OG oxygenases (2, 11, 54). Like Hs BBOX, TmpA crystallizes as a homodimer in a “head-to-tail” configuration with the N-terminal domain of one monomer interfaced with the C-terminal domain of the other monomer (Figure 3-7). The N- terminal domains of TmpA and Hs BBOX are structurally homologous, despite sharing < 20 % sequence identity (Appendix B, Figure S13), and are composed of two three-strand antiparallel

β-sheets with an intervening α/β insertion. However, whereas Hs BBOX harbors a Zn(II) ion in the α/β insertion (52), TmpA lacks the CxCxxC…H ligand sequence motif required for Zn(II) binding, and correspondingly, there is no divalent metal observed in the N-terminal domain

(Appendix B, Figure S13).

In all structures, the active site harbored in the C-terminal domain of TmpA reveals a mononuclear non-heme-iron center coordinated by residues His198, Asp200, and His344 in the expected “facial triad” geometry (Figure 3-7, S14) (12, 13). In chain A, 2OG is a bidentate equatorial ligand to the iron center (Figure 3-7, S14) and is stabilized in the active site by electrostatic interactions with the side chain of conserved residue Arg352. In the absence of primary substrate, the octahedral coordination sphere of the iron is completed by a water ligand axial to His344, and a sulfate ion from the crystallization solution occupies the phosphonate binding site described below (Appendix B, Figure S14). In the substrate-bound structure, the

2OG cosubstrate and TMAEP primary substrate can only be modeled in one of two monomers

(chain A) in the asymmetric unit. In the active site with TMAEP bound, the Fe(II) has lost the axial water ligand, resulting in square-pyramidal coordination geometry with an open position for dioxygen coordination (Figure 3-7, S14). Binding of TMAEP also induces inward collapse of

159 two beta strands in the core fold, β2 and β3, which immediately precede the HxD/E metal binding motif, and ordering of the “lid-loop” region (residues 179-196) to close over and bury the active site (Appendix B, Figure S15). This loop region remains disordered in chain B because another symmetry mate encroaches upon the active site, locking it in an open configuration (Appendix B,

Figure S15). Consequently, chain B only retains partial Fe occupancy and does not show reliable density for 2OG or TMAEP in the active site (Appendix B, Figure S14).

The active site residues that interact with TMAEP form a pocket similar to that observed for γbb in Hs BBOX, particularly in the site for binding of the quaternary amine (Figure 3-7). An electron-rich, hydrophobic aromatic cage is formed by the side chains of Tyr173, Phe177, and

Tyr190, positioned with the face of the aromatic rings toward the trimethylammonium moiety to establish cation-π interactions. The phosphonate dianion interacts electrostatically with the side chain of Arg288, and hydrogen bonds with a water molecule and the side chains of Asn187,

Asn201, Tyr203, and Asn286 (Appendix B, Figure S16). Interestingly, the Arg288 residue hydrogen bonds with the side chain of Asn45 located in the N-terminal domain of the opposite monomer, creating an extended hydrogen bonding network only in the presence of substrate

(Appendix B, Figure S16). The pro-(R)-hydrogen of the C1-carbon of TMAEP is poised above the open axial coordination site of the iron center, with the carbon at a distance of 4.2 Å. The product-bound structure (Appendix B, Figure S14) reveals clear electron density for the stereospecific hydroxylated product, (R)-OH-TMAEP, generated by turnover in crystallo, presumably due to slight oxygen contamination. Consistent with turnover in crystallo, the electron density in the equatorial plane cannot be modelled well with 2OG and most likely represents a mixture of succinate (resulting from turnover) and acetate (from the TMAEP synthesis) (Appendix B, Figure S14).

160

Figure 3-7. Structural comparison of Lc TmpA and Hs BBOX. (A) Homodimer quaternary structure of TmpA with chain A in dark green and chain B in light green. Fe(II) ions are shown as brown spheres. (B) Homodimer quaternary structure of Hs BBOX (PDB accession code 4C8R) with chain A in dark gray and chain B in light gray. The Zn(II) ion is shown as a purple sphere. (C) Active site of the substrate-bound structure of TmpA. (D) Active site of the substrate-bound structure of Hs BBOX (PDB accession code 3O2G). The co-substrates [2OG, N-oxalylglycine (NOG)], amino acids side chains, and substrates [TMAEP (yellow), γbb (blue)] are shown in stick format. Electrostatic interactions are designated by black dashed lines and the black arrows denote the target carbon position for hydroxylation.

II/III 3.5 TmpB harbors a diiron cofactor that is stable in the mixed-valent Fe2 state.

Due to the genomic synteny of TmpA and TmpB, and the precedent of the PhnY/PhnZ pathway, we hypothesized that TmpB would utilize the product of the TmpA reaction. The TmpA product, (R)-OH-TMAEP, is the trimethylammonium analogue of the substrate, (R)-OH-AEP, degraded by PhnZ. Lc TmpB shares 32% sequence identity with PhnZ (from uncultured bacterium HF130_AEPn_1), including conservation of the extended HD-domain sequence motif

161 (Appendix B, Figure S17), suggesting it could be a related diiron oxygenase (20, 21), despite its annotation as a phosphodiesterase. We set out to investigate the possibility that, by analogy to

PhnZ and MIOX, TmpB is another HD-MVDO, typified by their use of a diiron cofactor in the mixed-valent Fe2(II/III) state to activate O2, that catalyzes oxidative cleavage of the C–P bond in the TmpA product, (R)-OH-TMAEP.

Inductively-coupled plasma – atomic emission spectroscopy (ICP-AES) detected ~ 1.0-

1.3 molar equivalents of Fe in the purified TmpB protein and no other trace divalent metals above the limits of detection. The nuclearity of the cofactor and its stable oxidation states were assessed by Mössbauer spectroscopy. TmpB enriched in 57Fe (> 95 %) was incubated with the oxidant potassium ferricyanide, the strong reductant sodium dithionite, or the milder reductant L- ascorbate. The 120-K/0-T spectrum of the sample prepared with potassium ferricyanide (Figure

3-8A) demonstrates a single quadrupole doublet with parameters (isomer shift δ = 0.49 mm/s and quadrupole splitting parameter ΔEQ = 0.84 mm/s) typical of high-spin Fe(III). The presence of a quadrupole doublet in the 4.2-K/53-mT spectrum (Appendix B, Figure S18) indicates that the species has an integer spin electronic ground state, which can be rationalized by antiferromagnetic coupling of two high spin ferric ions (S = 5/2) in a dinuclear cluster to give a total spin of S = 0.4 The sample reduced with sodium dithionite has a slightly heterogeneous spectrum at 120-K/0-mT (Figure 3-8B) that is best fit by two quadrupole doublets with parameters characteristic of high-spin ferrous ions with O/N ligands (δ1 = 1.24 mm/s, ΔEQ1 = 3.25 mm/s and δ2 = 1.24 mm/s, ΔEQ2 = 2.5 mm/s). In contrast, when TmpB is reduced with L- ascorbate, the 4.2 K/53 mT spectrum demonstrates a broad multi-line signal (Appendix B,

Figure S19), indicative of a complex with half-integer spin ground state and similar to that

4 The 4.2 K/53 mT spectrum also shows broad baseline features in addition to the quadrupole doublet, suggesting the presence of a mononuclear high spin ferric (S = 5/2) species. This minority species, which is also detected by EPR spectroscopy with an effective g-value of ~ 4.3, is not attributed to partial occupancy at the diiron site, but rather to adventitiously-bound iron that is observed in the X-ray crystal structure (Appendix B, Figure S18)

162 observed for the Fe2(II/III) state of MIOX (55). When measured at 120-K/0-T, the paramagnetic features collapse into quadrupole doublets (Appendix B, Figure S19). The contribution of diferrous species can be identified by the partially resolved high energy line of its corresponding spectrum and was thus subtracted (22 %) from the experimental spectrum to facilitate its deconvolution (Appendix B, Figure S19). The resultant spectrum (Figure 3-8C) was fit with three quadrupole doublets: one with parameters of a high spin Fe(II) ion (δ1 = 1.19 mm/s, ΔEQ1 =

2.57 mm/s), another with parameters of a high spin Fe(III) ion (δ2 = 0.52 mm/s, ΔEQ2 = 0.96 mm/s) that was constrained in the fitting to have equivalent area to that of the Fe(II) doublet, and a final doublet constrained in the fitting with the parameters obtained from the spectrum of the differic state (13%). The Fe(II) and Fe(III) doublets represent the mixed-valent Fe2(II/III) species present in this sample, which, from the sum of their areas, is estimated to amount to ~ 65 % of the diiron species. Taken together, Mössbauer spectroscopy provides evidence for three distinct cluster oxidation states, Fe2(II/II), Fe2(II/III), and Fe2(III/III), consistent with the notion that

TmpB is an HD-MVDO.

163

Figure 3-8. 120 K/0 T Mössbauer spectra of 2 mM O2-free TmpA prepared in three oxidation states – (top) Fe2(III/III) by incubation with 3 mM potassium ferricyanide, (middle) Fe2(II/II) by incubation with 20 mM sodium dithionite, and (bottom) Fe2(II/III) by incubation with 20 mM sodium L-ascorbate, each for 45 min in an anaerobic chamber. Experimental spectra are shown as black vertical bars. Overall simulations are shown as red lines. The two doublets constituting the simulation of the Fe2(II/II) spectrum are shown as purple and cyan lines with parameters described in the main text. The sub-spectra simulations of spectrum C corresponding to the differic species (orange) and the ferrous (blue) and ferric (green) sites of the Fe2(II/III) species are shown as solid lines, respectively, with parameters described in the main text. The sub-spectrum of the Fe2(II/II) species (22%) was subtracted from the raw experimental spectrum (Appendix B, Figure S19) to give the spectrum shown in panel C for simplicity.

TmpB samples reduced by L-ascorbate, i.e. under conditions favoring accumulation of the Fe2(II/III) form, and in the presence or absence of (R)-OH-TMAEP were also analyzed by

EPR spectroscopy. The spectrum without (R)-OH-TMAEP (Figure 3-9, black) demonstrates a broad axial signal with principal g-values < 2 [1.94, 1.78, 1.78], indicative of an antiferromagnetically coupled S = 1/2 mixed-valent Fe2(II/III) species (20, 55, 56). This signal was quantified to be ~0.6 spin/diiron cluster, which is comparable to the yield determined from

164 Mössbauer analysis on an equivalent sample. Addition of (R)-OH-TMAEP drastically perturbs this signal to become more rhombic with principal g-values of 1.94, 1.82 and 1.65, suggestive of substrate binding to the cofactor (Figure 3-9, red), as has been observed for MIOX and PhnZ in the presence of their native substrates (20, 55). When (R)-OH-AEP (i.e., the non-N-methylated analogue and native PhnZ substrate) is incubated with the L-ascorbate-reduced TmpB, the resultant signal is more heterogenous than that with (R)-OH-TMAEP. The complexity of the spectrum suggests a mixture of signals, presumably representing bound and unbound states and/or multiple conformationally distinct states of (R)-OH-AEP binding (Figure 3-9, blue).

165 3.6 The TmpA hydroxylation product serves as the TmpB substrate.

The spectroscopic data demonstrate that the TmpA product, (R)-OH-TMAEP, binds to the diiron cofactor of TmpB and thus, it was tested as a substrate for TmpB in a multiple turnover reaction. TmpB, prepared by L-ascorbate reduction to accumulate the Fe2(II/III), completely consumes (R)-OH-TMAEP in the presence of excess O2. Substrate consumption is strictly dependent on the presence of dioxygen (Figure 3-10). The products of the reaction were determined to be glycine betaine and phosphate by comparison to LC-MS and NMR standards, respectively (Figure 3-10). PhnZ, which natively uses the non-N-methylated analogue, can also convert (R)-OH-TMAEP to the same products (Figure 3-10). Conversely, TmpB does not convert the PhnZ substrate, (R)-OH-AEP, as readily to the products glycine and phosphate, only achieving five turnovers after a 4-h incubation (Figure 3-10). This result suggested that TmpB possesses some degree of preference for the N-methylated substrates, as shown for TmpA. To further interrogate this trend, TmpB activity with the N,N-dimethylated analogue was tested in a coupled reaction with TmpA and DMAEP as the substrate. In this reaction, the products phosphate and N,N-dimethyl-glycine were detected by 31P-NMR and LC-MS, respectively, demonstrating that TmpB can utilize the hydroxylated DMAEP intermediate product (OH-

DMAEP) as a substrate (Appendix B, Figure S23). However, whereas the OH-TMAEP intermediate product accumulates transiently in an analogous coupled reaction, the OH-DMAEP accumulates throughout the reaction, suggesting slower consumption of OH-DMAEP by TmpB than OH-TMAEP (Appendix B, Figure S23). Thus, TmpB prefers the substrate with the highest degree of N-methylation, as observed with TmpA.

166

Figure 3-10. Testing activity of the HD-MVDOs against aminophosphonates with and without N- methylation. Aerobic reactions containing 0.01 mM TmpB or PhnZ, 0.2 mM L-ascorbate, 2 mM substrate [(A) (R)-OH-TMAEP or (B) (R)-OH-AEP] were incubated for 4-h at 3 °C. (Left) 31P- NMR spectra monitoring disappearance of the substrates, (R)-OH-TMAEP (14.0 ppm) or (R)- OH-AEP (15.0 ppm), and production of phosphate (0 ppm). [*contaminant from the PhnZ protein preparation] (Right) LC-MS chromatograms detecting the substrates, (R)-OH-TMAEP (184 m/z) or (R)-OH-AEP (142 m/z), and products, glycine betaine (118 m/z) or glycine (76 m/z).

3.7 Structural characterization of TmpB

The x-ray crystal structure of TmpB was solved to 1.63 Å resolution by molecular replacement using the PhnZ structure (PDB accession code 4MLM) (20) as a search model. The

(R)-OH-TMAEP substrate is present in two of the monomers in the asymmetric unit (chains C and D), but binding is prevented in the other two monomers (chains A and B), due to lattice contacts that occlude the active site (Appendix B, Figure S20) The x-ray structure solved

167 therefore provides two independent views of the active site in the absence of substrate and two views of the complete enzyme-substrate complex. The TmpB topology is entirely α-helical

(Figure 3-11A), similar to PhnZ (1.55 Å rmsd for the Cα-atoms over 170 residues (53)) (20, 23) and other HD-domain proteins (4). Five core helices contribute the conserved histidine and aspartate residues predicted from the extended HD-domain sequence motif (His40, His64, Asp65,

His86, His109, and Asp166) to coordinate a dinuclear cluster of iron ions at the active site

(Figure 3-11B), confirmed by anomalous dispersion data. The site 1 Fe has an additional ligand, conserved tyrosine residue (Tyr30), as has been observed in the PhnZ structure (23). In the substrate-free monomers, the active site cavity is highly solvent-exposed and the first coordination sphere is completed by a (hydr)oxo bridge and two water molecules at iron site 2. In place of the water ligands in the other two monomers, clear electron density for the TmpA product, (R)-OH-TMAEP, is observed, continuous with the density for the iron in site 2 (Figure

3-11C), indicating its direct coordination congruent with the electronic perturbations observed by

EPR spectroscopy. The (R)-OH-TMAEP molecule binds in a bidentate fashion via the C1- hydroxyl group and one of the oxygen atoms originating from the phosphonate moiety (Figure 3-

11C). The same substrate binding mode has been observed for the PhnZ and MIOX substrates

(20, 23, 24) and creates a 5-membered ring with the iron ion and 4 atoms of the substrate, such that the C–C/P bond to be cleaved straddles the iron at site 2. The C1-hydroxyl group of the substrate hydrogen bonds with the side chain of conserved His68 residue (Appendix B, Figure

S21) within the HDIGH68 sequence motif adjacent to the histidine-aspartate metal ligands. This hydrogen bond and substrate-metal coordination poises the target C1-hydrogen to be abstracted toward the iron at site 1, where the dioxygen is suggested to bind, according to the proposed mechanism for HD-MVDOs (57). In this structure of TmpB, the coordination sphere of the iron at site 1 is saturated due to the extra coordinating Tyr30 residue, which would occlude dioxygen

168 binding; however, it has been postulated that this Tyr ligand in PhnZ is transient during the catalytic cycle (23).

The phosphonate moiety of the substrate makes multiple interactions with the protein via the side chains of the conserved residues Arg163, Lys113, Ser134 and Gln138 (Appendix B,

Figure S21). In contrast, the trimethylammonium functional group of (R)-OH-TMAEP has a poorly define pocket in the TmpB active site due to solvent exposure (Appendix B, Figure S22).

In the structures of PhnZ, a flexible loop mediates an open-to-closed transition that seals off the active site from solvent upon substrate binding (Appendix B, Figure S22) (20, 23). The analogous loop (residues 69-84) in the TmpB substrate-bound monomers remains disordered, resulting in the solvent exposed active site. This observation is attributed to the fact that the substrate was introduced by crystal soak methods and the crystal lattice packing prevents this motion in crystallo (Appendix B, Figure S22). The amino acid sequence of this loop contains several aromatic residues and negatively-charged side chains (Appendix B, Figure S17), which could interact favorably with the positively-charged quaternary amine of (R)-OH-TMAEP.

However, the present crystal structure does not depict an obvious aromatic cage structural motif utilized in TmpA.

169

Figure 3-11. X-ray crystal structure of TmpB (A) Cartoon depiction of the α-helical tertiary structure characteristic of HD-domain proteins. Iron ions are shown as brown spheres and OH- TMAEP (yellow) is shown in stick format. A dashed line represents unmodeled regions of the structure. (B) Schematic of the first coordination sphere of the diiron cluster. (C) TmpB active site with (R)-OH-TMAEP bound. Protein residues and (R)-OH-TMAEP (yellow) are shown as sticks. Water molecules are shown as red spheres and iron ions as orange spheres. Blue mesh depicts the Fo – Fc omit map contoured at 3σ generated after ligand deletion and subsequent refinement.

3.7 Discussion

The genomically co-localized genes encoding TmpA and TmpB proteins have been identified in >350 bacterial species. While automated annotation succeeded in correctly classifying these gene products into the appropriate structural metalloprotein superfamilies, their functional assignments were over-predicted and misannotated. The functional assignment of

TmpA as a γ-butyrobetaine dioxygenase designated both the chemistry (i.e., hydroxylation) and

170 the substrate of the reaction (i.e., γbb). Although the nature of the chemical transformation catalyzed was correctly predicted, the substrate which the enzyme acts upon was not. From all compounds screened, TMAEP was shown to be the most suitable substrate based on three measures – reactivity, binding, and triggering of O2 activation. Whereas Fe/2OG dioxygenases exhibit uncoupling, viz the unproductive conversion of 2OG to succinate in the presence of non- native substrates (or in the absence of any substrate), TmpA demonstrates tight coupling of succinate production with TMAEP consumption. TMAEP binding enhances the observed rate for reaction of the Fe•2OG cofactor with dioxygen compared to that in the absence of substrate or in the presence of structurally similar, but poorer substrates (e.g., DMAEP and 2-AEP). This observed rate constant for the O2 addition step is on par with or faster than those for other

Fe/2OG dioxygenases with their native substrates and leads to the rapid accumulation of the known reactive ferryl intermediate (14, 47). This substrate triggering effect is a hallmark of

Fe/2OG enzymes (14, 47) and is proposed to result from displacement of the axial water ligand to open up a coordination site for O2 binding (58, 59). Indeed, in the TMAEP-bound crystal structure, the axial water ligand is displaced, whereas it remains coordinated when only sulfate is bound in the active site substrate pocket. TmpA with TMAEP bound is so poised for reaction with O2 such that turnover was achieved in crystallo due to trace amounts of O2.

TMAEP binding also induces structural changes beyond the first coordination sphere that likely enable reactivity and confer substrate specificity. The contraction of the core β-barrel in the catalytic domain and refolding of the lid loop over the active site positions substrate-interacting residues into the active site. The conformation of the lid loop is stabilized by interactions with the

N-terminal domain of the opposite monomer, suggesting a role for the N-terminal domain in both dimerization and substrate binding. Interestingly, the secondary and tertiary structure of the

TmpA N-terminal domain is almost identical to that of BBOX, except BBOX harbors a Zn(II) ion in its N-terminal domain. The Zn(II) is coordinated by a histidine residue and three cysteine

171 residues that are easily identified in the sequence by a CxCxxC motif (52). Although the exact function of this domain and the Zn(II) metal in BBOX is unknown, removal of the Zn(II) metal abolishes BBOX activity (60). Lc TmpA does not possess this sequence motif, nor a Zn(II) metal in the crystal structure, yet it maintains its structure, presumably through its extended hydrogen bonding network with the opposing monomer.

The lack of this Zn(II)-binding cysteine motif could be used to differentiate TmpA from

BBOX function from the primary structure alone. Indeed, this motif maps onto the SSN in a similar way to the genomic context information (Figure 3-2), with the exception of the fusion proteins (vide infra). The utility of this distinction in predicting function of uncharacterized proteins is illustrated well in analysis of bacterial genomes that contain multiple genes annotated as BBOXs (e.g., Halomonas species). In the final SSN, the nodes representing the two

Halomonas protein sequences encoded within the same genome are separated into distinct clusters, already suggestive that they perform different functions. The protein that clusters with the known Ps BBOX is encoded within a putative γbb degradation operon and possesses the

Zn(II) binding motif, and thus, can be concluded to function as a γ-butyrobetaine hydroxylase.

Conversely, the protein that clusters with the TmpA characterized herein is encoded in an operon together with a TmpB-like HD domain protein and does not possess the Zn(II) binding sequence motif, and correspondingly, should be annotated with TmpA function.

The occurrence of multiple genes annotated as BBOXs within a genome indicates that

BBOX and TmpA are likely paralogues, resulting from gene duplication and subsequent functional divergence. This relationship and common ancestry is evident by their structural homology and the slight promiscuity of BBOX to utilize the TmpA substrate. Unsurprisingly, their active sites are remarkably similar, particularly in the conserved cage of aromatic residues to establish cation-π interactions with the positively-charged quaternary amine common to the γbb and TMAEP substrates (61). Aromatic cages are a classic structural feature of proteins that bind

172 substrates with quaternary amines, including acetylcholine esterases, phosphocholine-binding antibodies, acetylcholine receptors, epigenetic trimethylysine reader domains, trimethylamine dehydrogenases, among others (62). Cation-π interactions are primarily electrostatic, with secondary contributions from hydrophobic interactions and desolvation, which impose a requirement of a positively-charged functionality for substrate recognition and binding (63, 64).

This structurally-induced discrimination is reflected by the lack of TmpA activity with phosphonates that lack an amine entirely. At physiological pH, the compounds 2-AEP, DMAEP and TMAEP are all expected to be positively-charged, yet TmpA demonstrates further selectivity based on the degree of N-methylation. Similar discrimination is observed for epigenetic methyllysine reader domains (65, 66) and alkylamine dehydrogenases (62). The reader domains that bind substrates with higher degrees of lysine N-methylation possess pockets consisting of 1-4 aromatic residues, similar to that observed in TmpA. Conversely, reader domains that recognize lower degrees of lysine N-methylation substitute one or more of the aromatic residues for residues capable of hydrogen bonding with the extra proton(s) on the amine. Furthermore, the N-methyl groups harbor some of the distributed positive charge from the nitrogen (63), perhaps enhancing the binding interaction and conferring the observed specificity of TmpA for higher degrees of N- methylation.

However, the quaternary amine is not the only discriminatory factor for TmpA substrate recognition and subsequent conversion, as the phosphonate functionality is also essential. It is also the major distinguishing feature between the substrates of TmpA and BBOX. A simple, but critical difference between the TmpA and BBOX active sites is the presence of an arginine residue in TmpA (Arg288) that interacts electrostatically with the phosphonate moiety of

TMAEP. Conversely, an alanine residue exists at this position in HsBBOX. A survey of solved substrate-bound structures of enzymes that utilize phosphonates reveals that there is almost invariantly at least one arginine residue that interacts with the phosphonate functional group (e.g.,

173 phosphonoacetaldehyde dehydrogenase, DhpI, 2-AEP transaminase, PhnZ, and HppE). This list also includes TmpB based on the crystal structure solved in this work, which has a conserved arginine (Arg163) to stabilize the phosphonate moiety of its (R)-OH-TMAEP substrate.

Intriguingly, enzymes that utilize carboxylate-, phosphonate- or sulfonate-containing substrates do not exhibit much promiscuity amongst the three functional groups (48), a selectivity that is also observed for TmpA. The equivalent carboxylate analogue of TMAEP (γbb-3) does not serve as a substrate for TmpA, and conversely, BBOX only demonstrates modest activity with TMAEP.

The preference of TmpA for the phosphorus-containing molecule is demonstrated even in cyrstallo. In the absence of substrate, sulfate, which is present at > 3 M in the crystallization solution, is bound in the active site at the position of the phosphonate moiety. Yet when TMAEP is soaked into the crystal at a final concentration of only ~10 mM, it displaces the sulfate molecule in the active site. The specificity of TmpA for the phosphonate functional group might suggest a priori that it would share structural homology and evolutionary history with the phosphonate hydroxylase, PhnY. Indeed, they enact identical chemical transformations – hydroxylation of the C1 position adjacent to a phosphonate – on molecules that differ only in amine methylation. Yet TmpA shares very little sequence similarity with PhnY. Instead, it is the regioselectivity with respect to the quaternary amine that is common to TmpA and its homologue,

BBOX. Thus, TmpA and PhnY likely evolved in parallel to achieve similar mechanisms to initiate phosphonate degradation that is completed by their HD-domain partners.

The HD-domain partners, TmpB and PhnZ, on the other hand likely share an evolutionary history that is reflected by their unique cofactor, the chemical transformation they enact, and their structural homology. The x-ray crystal structure and Mössbauer and EPR spectroscopic data establish that TmpB harbors a diiron cofactor and, most importantly, the latter two techniques demonstrate that TmpB stabilizes the mixed-valent Fe2(II/III) oxidation state.

Whereas canonical diiron enzymes of the ferritin-like superfamily react with dioxygen from the

174

Fe2(II/II) oxidation state, and typically cannot stabilize the mixed-valent oxidation state or can only do so at a very low yield (67-69), the only two known examples of HD-MVDOs, PhnZ and

MIOX, stabilize and utilize the mixed-valent state as the active cofactor (20, 22). The fact that

TmpB demonstrates this hallmark of HD diiron oxygenases provides strong evidence that TmpB is another example of an HD-MVDO. Furthermore, the O2-dependent four-electron oxidation of the phosphonate substrate catalyzed by TmpB solidifies this assignment. Although the prediction of novel HD-MVDOs cannot solely rely on identification of the extended sequence motif of metal ligands, these proteins could serve as a pool to be mined for novel oxygenase function. Thus, the recent discoveries about members of the HD-domain superfamily, both phosphohydrolases and oxygenases, warrant consideration in function assignment of existing and newly identified HD- domain proteins.

The substrate preference demonstrated by TmpB for the N-methylated substrate mirrors the specificity of TmpA, both exhibiting very poor catalytic efficiency with the analogous primary amine compounds. However, the ability of TmpB to discern the methylation state of the amine is not immediately obvious from its structure in comparison to that of PhnZ. The primary amine of OH-AEP forms a single contact with the PhnZ protein via the side chain of Glu27

(PhnZ numbering), which is also conserved in TmpB. While the OH-AEP primary amine can participate in both hydrogen bonding and electrostatic interactions with the acidic residue, the quaternary amine is restricted to the latter. Although this implies a weaker interaction, the observed activity of PhnZ with OH-TMAEP suggests that it is sufficient for substrate recognition.

In the absence of substrate in PhnZ, this key glutamate residue is facing outside of the active and, upon substrate binding, the loop with this residue undergoes a conformation change to position it in the active site. This movement of the peptide backbone is proposed to induce dissociation of the Fe-ligated tyrosine (PhnZ numbering), thereby opening a coordination site for dioxygen binding, a prerequisite for catalysis. On the other hand, in TmpB, the corresponding glutamate

175 (Glu33) is held in place through a hydrogen bonding network in the active site, even in the absence of substrate. This observation implies that another mechanism must be operant to trigger dissociation of the coordinating Tyr30 ligand in this enzyme. Unfortunately, due to the soaking method by which substrate was introduced and the obstructive crystal contacts, the Tyr ligand is not observed in the speculated dissociated form with substrate bound, impeding our ability to draw strong evidence-based conclusions. However, we speculate that the lid loop region of

TmpB, which diverges greatly in sequence from that of PhnZ, could play a key role in substrate recognition.

The observed substrate promiscuity of PhnZ is not expected to be relevant in Nature, because its partner hydroxylase, PhnY, is unable to convert the N-methylated substrate into the hydroxylated intermediate for PhnZ to utilize. Thus, for both the TmpA/B and the PhnY/Z pair, the specificity for degradation of these closely related compounds is predominantly determined by the Fe/2OG dioxygenase, rather than the HD-MVDO partner. Interestingly, 16 examples have been identified in which a single operon possesses a single HD-domain protein, but multiple

Fe/2OG oxygenases, both a BBOX-like and a PhnY-like (i.e., PhyH-like). While this type of operon could potentially represent a novel degradation pathway, it could also perhaps reflect the specificity of the Fe/2OG dioxygenases and the promiscuity of the HD-domain protein, such that the Fe/2OG dioxygenases utilize different substrates, but the HD-MVDO is capable of degrading either hydroxylated intermediate product. The apparent dedication of the pathway partners – an

Fe/2OG dioxygenase and an HD-MVDO – is also reflected in the conservation and occurrence of their respective genomic operons. In addition to the 300+ examples of operons containing separate genes for TmpA and TmpB, there are 59 examples of fusion genes of TmpA and TmpB.

There are also extant gene fusions of PhnY and PhnZ. It is unknown whether the fusion gene product would be a functional two-domain enzyme, but the occurrence of these gene fusions supports the notion that each pairing is specific and functions together. Furthermore, the frequent

176 presence of a gene encoding a LysR-type transcription regulator located upstream of the TmpA and TmpB genes suggests that they are co-transcribed and is consistent with the biochemical evidence presented herein that their gene products constitute a degradation pathway.

Transcription of the operon encoding the PhnY/PhnZ pair also appears to be under control of a

LysR-type regulator. Transcription regulators of the LysR family (70) are commonly found upstream of other known phosphonate degradation operons and, in some cases, have been shown to be induced by the specific phosphonate compounds that their operons degrade (71-73). These precedents would lead to the prediction that transcription of the appropriate degradation pair,

TmpA/B or PhnY/Z, would be specifically induced by the presence of TMAEP or 2-AEP, respectively.

Phosphonates and their degradation products serve valuable roles for bacteria of certain environmental niches. Phosphonates exist in Nature as components of phosphonolipids, phosphonoproteins, and phosphonoglycans, and are incorporated in the place of their phosphoester analogues presumably for the advantage of a more stable C–P bond instead of the hydrolyzable phosphoester C–O–P bond (74-76). The most common phosphoesters are phosphoethanolamine, phosphoserine and phosphocholine, which each have their corresponding phosphonate counterparts – 2-AEP, phosphonoalanine and TMAEP, respectively. While 2-AEP is the most abundant phosphonate in Nature, TMAEP, as well as DMAEP and N-monomethyl-AEP, have been reported in protozoa and marine invertebrates, such as sea anemones, marine plankton and algae (75, 77-79). Degradation pathways for both 2-AEP and phosphonoalanine have been discovered, ultimately producing phosphate and a source of carbon (80-84). The TmpA/B pathway mirrors the PhnY/Z degradation pathway for 2-AEP in its chemical transformations, but is specific for TMAEP, and is the first example of a mechanism for degradation of this naturally- occurring phosphonate compound.

177 The TmpA/B operon is found primarily in Proteobacteria and Actinobacteria, with considerable representation from marine and N2-fixing plant-symbiotic microbes. The products of this degradation pathway, phosphate and glycine betaine, are valuable nutrients for such bacteria.

Organophosphonates have been established as major sources of phosphorus in marine environments, which are limited in this essential nutrient (80). Phosphorus has been found to enhance plant growth and root nodule formation, where the symbiotic bacteria reside, and most critically to support greater N2-fixation capability (85-89). Many organisms possess multiple phosphonate degradation pathways to extract phosphate, but the requirement for different specialized pathways and their cooperation are poorly understood (80-83). Nonetheless, the ubiquity of mechanisms for the release of phosphate from dissolved organic phosphorus sources highlights their necessity and utility. The second product of the TmpA/B pathway, glycine betaine, is a known osmoprotectant. Osmoregulation by compounds with quaternary amines is common in marine organisms to protect against increased salinity (90). Glycine betaine has been shown to be the most potent osmoprotectant in some N2-fixing bacteria, enhancing growth under conditions of osmotic stress (91, 92). Interestingly, nitrogen fixation is sensitive to osmotic stress

(93). Additionally, glycine betaine can be utilized as a source of carbon and nitrogen upon its demethylation to glycine (94, 95). This source of nutrients could provide a competitive advantage for soil bacteria (95), particularly for colonization of the environmental niche of plant roots.

In conclusion, we have defined the biochemical reactions of the Fe/2OG oxygenase,

TmpA, and the HD-domain protein, TmpB, to constitute a novel degradation pathway of the natural organophosphonate, TMAEP. The combination of genomic context and structure- informed sequence motifs will allow for the proper identification and annotation of the genes associated with this pathway in existing and newly sequenced genomes. More broadly, the biochemical and structural information gleaned about these subsets of proteins could potentially

178 inform bioinformatic analyses aiming to categorize and predict functions within their large superfamilies.

3.8 Materials and Methods

Materials. All chemicals used for protein over-expression and purification were purchased from

Sigma-Aldrich (St. Louis, USA) unless stated otherwise.

Preparation of LcTmpA, LcTmpB, and PsBBOX. Detailed procedures for over-expression and purification of the proteins are provided in the Supporting Information.

Preparation of PhnY and PhnZ. Proteins were prepared as previously described (20).

Construction of Sequence Similarity Networks. The sequence similarity network (SSN) was generated using the Enzyme Function Initiative – Enzyme Similarity Tool (EFI-EST) web-based server (43) and visualized in Cytoscape (96). A network was generated including sequences with the IPR003819 “TauD/TfdA-like” domain having a length of more than 350 amino acids, and sequences (Table S1) returned from a Basic Local Alignment Search Tool (BLAST) of the

National Center for Biotechnology Information (NCBI) and the Joint Genome Institute –

Integrated Microbial Genomes (JGI-IMG) databases with the LcTmpA sequence (E-value threshold < 10-10) (97). A representative network was constructed by binning sequences of > 90% sequence identity into single nodes and selecting an initial pairwise alignment score threshold of

< 10-50 for vertices (Appendix B, Figure S1). Further network analysis was performed by increasing the threshold stringency as described in the Supporting Information (Figures S2 and

S3; Figure 3-2).

Syntheses of 2-dimethylaminoethylphosphonate (DMAEP) and 2- trimethylaminoethylphosphonate (TMAEP). The synthetic protocols for DMAEP and TMAEP were modified from those previously published (Appendix B, Figure S5) (98-101).

179 Triethylphosphite was reacted with excess dibromoethane at 160 °C while refluxing for 6 h. The reaction mixture was separated by reduced pressure distillation to isolate the diethyl 2-

31 bromoethylphosphonate product [ P NMR (CDCl3) δ = 25.6 ppm]. The product was mixed with

1.3 molar equiv. of triethylamine in benzene and refluxed at 85 °C for 2 h to give diethyl

31 vinylphosphonate [ P NMR (CDCl3) δ = 17.3 ppm]. The reaction was filtered, washed and dried.

The product was reacted with 2 molar equiv. of dimethyl amine in ethanol while refluxing at 40

31 °C overnight to give diethyl 2-dimethylaminoethylphosphonate [ P NMR (CDCl3) δ = 30.6 ppm]. This dried product was either directly deprotected or was incubated with 2 molar equiv. of methyliodide in diethyl ether at 25 °C overnight to give diethyl 2-

31 trimethylaminoethylphosphonate [ P NMR (CDCl3) δ = 23.3 ppm]. The solid product was filtered, washed and dried. The product was deprotected by reaction with 5 molar equiv. of

31 TMSBr in dichloromethane at 25 °C overnight to give either DMAEP [ P NMR (CDCl3) δ =

31 17.5 ppm] or TMAEP [ P NMR (CDCl3) δ = 16.5 ppm]. The final products were extracted with a solution of 3 molar equiv. of ammonium acetate, dried, and stored at -20 °C.

Enzymatic synthesis of (R)-1-hydroxy-2-trimethylaminoethylphosphonate (OH-TMAEP).

(R)-OH-TMAEP was generated in a large-scale enzymatic conversion of TMAEP by TmpA. A reaction (100 mL) containing 0.005 mM TmpA, 0.01 mM (NH4)2Fe(SO4)2, 3 mM 2OG and 3 mM

TMAEP was prepared in 25 mM ammonium acetate buffer (pH 7.5) and was incubated at room temperature with constant stirring and flushing of atmospheric-composition air. The product was purified according to the procedure used for preparation of (R)-OH-AEP (20) and was analyzed

13 by NMR (Appendix B, Figure S5): C NMR (D2O) δ = 68.83 (d, J = 12.5 Hz), 64.76 (d, J =

31 1 145.0), and 62.39 ppm; P NMR (D2O) δ = 13.0 ppm; H NMR (D2O) δ = 3.60 (m, 1H), 3.45 (m,

1H), and 3.37 (m, 1H), 1.74 (s, 9H).

Multiple-turnover enzyme assays. End-point reactions to test for activity of Fe/2OG oxygenases with substrates and analogues contained final concentrations of 0.02 mM TmpA (or BBOX or

180

PhnY), 0.03 mM (NH4)2Fe(SO4)2, 0.4 mM sodium L-ascorbate, 3 mM 2OG, 2 mM substrate and

2 0.2 mM 2,2,3,3-[ H]4-succinate (d4-succinate) in 50 mM sodium 2-[4-(2-hydroxyethyl)-piperazin-

1-yl] ethanesulfonate (HEPES) buffer, pH 7.5. End-point reactions to test for activity of HD- domain proteins with substrates and analogues contained final concentrations of 0.01 mM TmpB

(or PhnZ), 0.5 mM sodium L-ascorbate and 2 mM substrate in 50 mM sodium HEPES, pH 7.5.

All reaction components (O2-free) were mixed in an anoxic chamber and brought up to 10% of the final volume (1 mL) by addition of O2-free buffer. The solutions were then removed from the chamber, and the reactions initiated by addition of cold, air-saturated buffer (~ 0.4 mM O2 at 5

°C) up to the final volume. Reactions were incubated at 3 °C, stirred, and flushed with air for 4 h.

Aliquots for LC-MS (0.1 mL) and 31P-NMR analysis (0.48 mL) were mixed with 5 % formic acid and 4 % acetic acid, respectively, to terminate the reactions. Minor deviations from these general procedures are described in the appropriate figure legends.

Quantification of TmpA coupling ratio. Reaction solutions (0.1 mL) were prepared in an anoxic chamber, containing 0.22 mM TmpA, 0.2 mM (NH4)2Fe(SO4)2, 0.4 mM TMAEP, 0.1 mM d4-succinate and varying concentrations of 2OG (0, 0.02, 0.04, 0.06, 0.08, 0.1, 0.12, 0.14, 0.16 mM) in 50 mM sodium HEPES, pH 7.5. Assay solutions were removed from the anoxic chamber and mixed with an equal volume of O2-saturated sodium HEPES buffer, pH 7.5 (~1.8 mM O2 at 5

°C). Reactions were quenched after 10 min by addition of 10 μl formic acid and were filtered through 10K centrifugal devices (Pall Corporation, Westborough, MA). Succinate and TMAEP were quantified by the LC-MS analysis described below.

31P-NMR sample preparation and spectroscopy. For 31P-NMR measurements, sodium dithionite, ethylenediaminetetraacetic acid (EDTA), and D2O were added to the samples to final concentrations of 0.2 mM, 0.2 mM, and 20 % (v/v), respectively. Solution 31P-NMR spectra of the various phosphorous containing compounds were recorded at room temperature on an AVX-

360 or HD-500 Bruker spectrometer. The spectra were recorded with a 1D sequence with power-

181 gated 1H decoupling. Chemical shifts are quoted with respect to a phosphoric acid solution (0 ppm) prepared in the reaction buffer, sodium HEPES buffer, pH 7.5, treated with acetic acid,

EDTA, sodium dithionite, and diluted in D2O. NMR spectra were further processed with the freely available Spinworks (version 1.3.8.1) software (Dr. Kirk Marat, University of Manitoba,

Canada).

High performance liquid chromatography – mass spectrometry (LC-MS). LC-MS analysis was carried out on an Agilent 1200 series LC system coupled to a triple quadrupole mass spectrometer (Agilent 6410 QQQ LC/MS; Agilent Technologies). Succinate was detected by injection of 2 μl of filtered samples (0.2 μm filter) onto an extend-C18 (Agilent) column that had been equilibrated with 99% solvent A (water with 0.1% formic acid) and 1% solvent B

(acetonitrile). Succinate was eluted from the column with an isocratic mobile phase of 99% solvent A and 1% solvent B at a flow rate of 0.4 mL/min for 10 min. Detection of succinate and the internal standard d4-succinate was performed by electrospray ionization mass spectrometry

(ESI-MS) in the negative mode with single ion monitoring at mass-to-charge ratios (m/z) of 117 and 121, respectively. Succinate was quantified by comparing the integrated peak area to that of the known concentration of d4-succinate.

For analysis of quaternary ammonium compounds, samples (2 μl) were injected onto a

SeQuant ZIC-HILIC (3.5 μm, 100 Å, PEEK 150 x 2.1 mm) column (Merck, Darmstadt,

Germany) that had been equilibrated with 2% solvent A (10 mM ammonium acetate with 0.2% formic acid) and 98% solvent B (acetonitrile). The reaction mixture was separated with a flow rate of 0.3 mL/min by applying a linear gradient from 2% - 75% A over 28 min, then returning to

2% A over 25 min and re-equilibrating with 2% A for 7 min. The compounds were detected by

ESI-MS (source parameters: gas temperature = 350 °C, gas flow = 9 L/min, nebulizer pressure =

40 psi, capillary voltage = 4000 V) in the positive mode with single ion monitoring (fragmentor voltage = 135 V, dwell time = 200 s, delta EMV = 200 V) of m/z ratios provided in the table

182 below. Comparison of integrated peak intensities to that of an internal standard (γbb or γbb-3) of known concentration enabled quantification of the analytes.

Table 3-1. Target ion m/z ratios detected in positive mode. compound m/z m/z + 16

TMAEP 168 184 DMAEP 154 170 PC 184 200 γbb 146 162 γbb-3 132 148 choline 104 120 glycine betaine 118 glycine betaine aldehyde 102

For analysis of primary amine compounds, samples (2 μl) were injected onto a SeQuant

ZIC-HILIC (3.5 μm, 100 Å, PEEK 150 x 2.1 mm) column (Merck, Darmstadt, Germany) that had been equilibrated with 5% solvent A (water with 0.1% formic acid) and 95% solvent B

(acetonitrile with 0.1% formic acid). The reaction mixture was separated with a flow rate of 0.3 mL/min by applying a linear gradient from 5% - 70% A over 25 min, then returning to 5% A over

25 min and re-equilibrating with 5% A for 10 min. The compounds were detected by ESI-MS

(source parameters: gas temperature = 350 °C, gas flow = 9 L/min, nebulizer pressure = 40 psi, capillary voltage = 4000 V) in the positive mode with single ion monitoring (fragmentor voltage

= 135 V, dwell time = 200 s, delta EMV = 200 V) of m/z values listed in the table below.

Quantification was as described above using 13C-glycine as an internal standard.

Table 3-2. Target ion m/z ratios detected in positive mode. compound m/z m/z + 16

2-AEP 126 142 Taurine 168 184 GABA 104 120 β-alanine 90 glycine 76

183 Stopped-flow absorption (SF-Abs) measurements and data analysis. SF-Abs experiments were carried out at 5 °C with an Applied Photophysics Ltd. (Leatherhead, UK) SX20 stopped- flow spectrophotometer housed in an anoxic chamber (Labmaster, MBraun, Stratham, USA). The instrument was configured for single-mixing, an optical pathlength of 1 cm, and data acquisition with white polychromatic light and a photodiode-array (PDA) detector. Time-dependent absorption spectra (1000 points) were acquired on a logarithmic time base. Specific reaction conditions are provided in the appropriate figure legends.

The ΔA-versus-time trace at 519 nm (where the TmpA reactant complex maximally absorbs) was fit by the equation describing the change in concentration of intermediate “B” as a function of time according to the model ABC (eq. 1), where A0 is the initial absorbance, ΔA1 is the theoretical amplitude, and k1 and k2 are the observed rate constants. The ΔA-versus-time trace at 318 nm (where the ferryl complex absorbs much more than the reactant complex) was also fit by this equation, but with an additional exponential function representing the contribution from absorption of the reactant complex at this wavelength, where ΔA2 is the amplitude change from this feature and k3 is the observed rate constant (eq. 2).

A A ( ) ( ) (eq. 1) 1

A A ( ) ( ) A2( ) (eq. 2) 1

Crystallization and structure solution by X-ray diffraction analysis. For all structures, crystallographic datasets were collected at the 21ID-F/G and 23ID-B beamlines of the Advanced

Photon Source at Argonne National Lab, and the resulting diffraction images were processed with the software package HKL2000 (102). Refinement and model building were performed with the programs Refmac5 (103) and COOT (104), respectively. A summary of data collection and refinement statistics for all structures can be found in Tables S2 and S3. Ramachandran outlier

184 analysis and other validation procedures were carried out using the Molprobity server (105).

Figures were generated in the PyMOL molecular graphics software package (Schrödinger LLC).

In an anoxic chamber at 25 °C, an O2-free solution containing TmpA (15 mg/mL), 0.35 mM (NH4)2Fe(SO4)2, and 1.75 mM 2OG was mixed in a 2:1 ratio with a precipitant solution containing of 1.0-1.4 M LiSO4 and 1.6-1.8 M (NH4)2SO4 in hanging drop vapor diffusion trials.

TMAEP was soaked into existing crystals by 1:1 dilution of crystal drops with 20 mM TMAEP prepared in the precipitant solution for 4-5 h at 25 °C. Crystals were prepared for data collection by cryoprotection in the precipitant solution supplemented with 19% (v/v) glycerol, mounting on rayon loops, and flash freezing in liquid nitrogen. The product-bound TmpA crystal structure was obtained after adventitious exposure of crystals (described below) to O2 during crystallization, likely due to inefficient degassing of the original screen. These crystals were prepared from a solution of O2-free TmpA (15 mg/mL), 0.35 mM (NH4)2Fe(SO4)2, 1.75 mM 2OG and 1.75 mM

TMAEP mixed in a 1:1 ratio with a precipitant solution from a commercial screen (Qiagen) consisting of 1.1 M sodium malonate (pH 7.0), 0.1 M sodium HEPES (pH 7.0) and 0.5 %

Jeffamine ED-2001 (pH 7.0) in a sitting drop diffusion trial in an anoxic chamber. Crystals were prepared for data collection by cryoprotection in the precipitate solution supplemented with 25% glycerol, mounting on a rayon loop, and flash freezing in liquid nitrogen.

All TmpA structures were solved in the P41212 spacegroup with 2 molecules in the asymmetric unit (ASU) unit cell lengths of ~87, 87, and 220. Phase information for the product- bound dataset was obtained by molecular replacement using the software package PHASER

(106) with a truncated HsBBOX structure (PDB accession code 3O2G, residues 101-388), as the search model. The resulting electron density map was subjected to an autobuilding procedure using the software package ARP/wARP (107), followed by manual model building for all residues. All other TmpA structures were solved by molecular replacement using PHASER

185 (106) with the product-bound TmpA structure as the search model. The TMAEP and OH-

TMAEP ligands were generated using the software package JLigand (108).

The TmpA•Fe•2OG structure contains residues -1-375 in chain A and residues -1-376 in chain B. Single iron ions are observed in both chains A and B, but a 2OG molecule is only observed in chain A. The structure additionally contains 4 sulfate ions and 464 water molecules.

The TmpA•Fe•2OG•TMAEP structure consists of residues -1-376 in chain A, residues -2-

188/196-376 in chain B, 2 iron ions, 1 2OG molecule, 1 TMAEP molecule (TMP), 1 sulfate ion, and 521 water molecules. The TMAEP and 2OG molecules are only observed in chain A. A loop consisting of residues 189-195 is disordered in chain B, whereas it is ordered above the substrate- bound active site in chain A, suggesting some minor degree of substrate binding in chain B causing the loop dynamics. The TmpA•Fe•2OG•OH-TMAEP structure contains residues -1-376 in both chains A and B, 2 iron ions, 1 OH-TMAEP molecule (TMO), and 483 water molecules.

The OH-TMAEP molecule is only observed in chain A and the electron density for the co- substrate is not consistent with succinate bound (Appendix B, Figure S13). This density is also not consistent with acetate molecules, known to be present from the chemical synthesis of

TMAEP, (Appendix B, Figure S13). The density observed is likely a result of a mixture of succinate, acetate, and water ligands that co-exist at this location in chain A and was therefore left un-modeled.

In an anoxic chamber, an O2-free solution of as-isolated TmpB (10 mg/mL) and 5 mM sodium L-ascorbate was mixed in a 1:1 ratio with a precipitant solution containing 0.2 M CaCl2,

0.1 M sodium HEPES (pH 7.5), and 27%-33% polyethylene glycol (PEG) 4000 in hanging drop vapor diffusion trials. (R)-OH-TMAEP was incorporated via soaking by 1:1 dilution of crystal drops with 20 mM OH-TMAEP prepared in the precipitate solution and incubation for 24 h at 25

°C. Crystals were prepared for data collection by cryoprotection in the precipitate solution

186 supplemented with 21% (v/v) glycerol, mounting on rayon loops, and flash freezing in liquid nitrogen for data collection.

The TmpB•Fe•(R)-OH-TMAEP structure was solved in the C2221 spacegroup with 4 molecules in the ASU and unit cell lengths of ~70, 151, and 136. Phase information for the dataset was obtained by molecular replacement using the software package MOLREP (109) with the structure of PhnZ (PDB accession code 4MLM) as the search model. The final model contains residues 8-195 in chain A, residues 8-194 in chain B, residues 7-76/85-195 in chains C and D, 6

Fe(II) ions, 4 Fe(III) ions, 2 (R)-OH-TMAEP (TMO) molecules, and 261 water molecules. Each diiron site found in chains A-D was modeled with one Fe(II) ion and one Fe(III) ion in the cluster, as that species accumulates to > 60% in solution when the diferric state is incubated with

L-ascorbate, identical to the crystallization conditions. (R)-OH-TMAEP molecules are only observed in chains A and B where a disordered loop region (residues 77-84) can be modeled over the active site, which is not observed in chains C and D. Additionally, two iron atoms are observed to bind between adjacent chains. These binding sites were confirmed via the iron anomalous dispersion data and modeled at 50% occupancy. The iron anomalous dispersion data was collected at 7.2 keV and phased using the TmpB coordinates. Anomalous dispersion maps were generated with CAD and FFT of the CCP4 Program Suite (110).

Mössbauer spectroscopy. Mössbauer spectra were recorded on WEB Research (Edina, MN) instruments that have been described previously (14). The spectrometer used to acquire the weak- field spectra is equipped with a Janis SVT-400 variable-temperature cryostat. The external magnetic field was applied parallel to the γ beam. All isomer shifts are quoted relative to the centroid of the spectrum of α-iron metal at room temperature. Mössbauer spectra were simulated using the WMOSS spectral analysis software (www.wmoss.org, WEB Research, Edina, MN).

Continuous-Wave (CW) EPR spectroscopy. EPR spectra at X-Band (~ 9.5 GHz) were acquired on a Bruker ESP-300 spectrometer equipped with an ER/4102 ST resonator (Bruker), an Oxford

187 Instruments continuous helium flow cryostat, and an Oxford Instruments temperature controller

(ITC 502). For all experiments, quartz tubes with 3 mm inner and 4 mm outer diameter were used

(QSI).

3.9 Acknowledgements

Dr. Andrew Mitchell collected the anomalous data and assisted with analysis of x-ray crystal diffraction data and model biulding. Prof. Maria-Eirini Pandelia assisted with initial Mössbauer characterization and Dr. Bo Zhang collected final Mössbauer spectra. Prof. Wei-chen Chang synthesized TMAEP and DMAEP substrates and performed NMR analysis of OH-TMAEP. Prof.

Wilfred van der Donk (University of Illinois, Urbana-Champaign) provided the library of phosphonate compounds. Prof. Christopher Schofield (Oxford University) provided the library of betaine compounds and Ps BBOX plasmid. The authors would like to acknowledge the Shared

Fermentation Facility of The Pennsylvania State University (University Park, PA) for use of the

Microfluidics M-110EH-30 microfluidizer processor, and Mr. Henry Gong at the Pennsylvania

State Materials Research Institute (University Park, PA) for ICP-AES analysis.

This material is based upon work supported by the National Science Foundation under Award

No. MCB-1330784. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the National

Science Foundation.

3.10 References

1. Sono, M., Roach, M. P., Coulter, E. D., and Dawson, J. H. (1996) Heme-Containing Oxygenases. Chem Rev 96, 2841-2888. 2. Hausinger, R. P. (2015) 2-Oxoglutarate-Dependent Oxygenases, The Royal Society of Chemistry.

188 3. Andrews, S. C. (2010) The Ferritin-like superfamily: Evolution of the biological iron storeman from a rubrerythrin-like ancestor. Biochim Biophys Acta 1800, 691-705. 4. Aravind, L., and Koonin, E. V. (1998) The HD domain defines a new superfamily of metal-dependent phosphohydrolases. Trends Biochem Sci 23, 469-472. 5. Baier, F., Copp, J. N., and Tokuriki, N. (2016) Evolution of Enzyme Superfamilies: Comprehensive Exploration of Sequence-Function Relationships. Biochemistry 55, 6375- 6388. 6. Roche, D. B., and Bruls, T. (2015) The enzymatic nature of an anonymous protein sequence cannot reliably be inferred from superfamily level structural information alone. Protein Sci 24, 643-650. 7. Galperin, M. Y., and Koonin, E. V. (2012) Divergence and convergence in enzyme evolution. J Biol Chem 287, 21-28. 8. Gaudet, P., Livstone, M. S., Lewis, S. E., and Thomas, P. D. (2011) Phylogenetic-based propagation of functional annotations within the Gene Ontology consortium. Brief Bioinform 12, 449-462. 9. du Plessis, L., Skunca, N., and Dessimoz, C. (2011) The what, where, how and why of gene ontology--a primer for bioinformaticians. Brief Bioinform 12, 723-735. 10. Skunca, N., Altenhoff, A., and Dessimoz, C. (2012) Quality of computationally inferred gene ontology annotations. PLoS Comput Biol 8, e1002533. 11. Aik, W., McDonough, M. A., Thalhammer, A., Chowdhury, R., and Schofield, C. J. (2012) Role of the jelly-roll fold in substrate binding by 2-oxoglutarate oxygenases. Curr Opin Struct Biol 22, 691-700. 12. Koehntop, K. D., Emerson, J. P., and Que, L., Jr. (2005) The 2-His-1-carboxylate facial triad: a versatile platform for dioxygen activation by mononuclear non-heme iron(II) enzymes. J. Biol. Inorg. Chem. 10, 87-93. 13. Hegg, E. L., and Que, L., Jr. (1997) The 2-His-1-carboxylate facial triad - An emerging structural motif in mononuclear non-heme iron(II) enzymes. Eur. J. Biochem. 250, 625- 629. 14. Price, J. C., Barr, E. W., Tirupati, B., Bollinger, J. M., Jr., and Krebs, C. (2003) The first direct characterization of a high-valent iron intermediate in the reaction of an alpha- ketoglutarate-dependent dioxygenase: A high-spin Fe(IV) complex in taurine/alpha- ketoglutarate dioxygenase (TauD) from Escherichia coli. Biochemistry 42, 7497-7508. 15. Murray, K. D., and Bremer, H. (1996) Control of spoT-dependent ppGpp synthesis and degradation in Escherichia coli. J Mol Biol 259, 41-57. 16. Francis, S. H., Colbran, J. L., McAllister-Lucas, L. M., and Corbin, J. D. (1994) Zinc interactions and conserved motifs of the cGMP-binding cGMP-specific phosphodiesterase suggest that it is a zinc hydrolase. J Biol Chem 269, 22477-22480. 17. Johnson, G. S., Adler, C. R., Collins, J. J., and Court, D. (1979) Role of the spoT gene product and manganese ion in the metabolism of guanosine 5'-diphosphate 3'-diphosphate in Escherichia coli. J Biol Chem 254, 5483-5487. 18. Quirk, S., and Bessman, M. J. (1991) dGTP triphosphohydrolase, a unique enzyme confined to members of the family Enterobacteriaceae. J Bacteriol 173, 6665-6669. 19. Bollinger, J. M., Jr., Diao, Y., Matthews, M. L., Xing, G., and Krebs, C. (2009) myo- Inositol oxygenase: a radical new pathway for O(2) and C-H activation at a nonheme diiron cluster. Dalton Trans, 905-914. 20. Wörsdörfer, B., Lingaraju, M., Yennawar, N. H., Boal, A. K., Krebs, C., Bollinger, J. M., Jr., and Pandelia, M.-E. (2013) Organophosphonate-degrading PhnZ reveals an emerging family of HD domain mixed-valent diiron oxygenases. Proc Natl Acad Sci U S A 110, 18874-18879.

189 21. McSorley, F. R., Wyatt, P. B., Martinez, A., DeLong, E. F., Hove-Jensen, B., and Zechel, D. L. (2012) PhnY and PhnZ comprise a new oxidative pathway for enzymatic cleavage of a carbon-phosphorus bond. J Am Chem Soc 134, 8364-8367. 22. Xing, G., Barr, E. W., Diao, Y., Hoffart, L. M., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) Oxygen activation by a mixed-valent, diiron(II/III) cluster in the glycol cleavage reaction catalyzed by myo-inositol oxygenase. Biochemistry 45, 5402-5412. 23. van Staalduinen, L. M., McSorley, F. R., Schiessl, K., Seguin, J., Wyatt, P. B., Hammerschmidt, F., Zechel, D. L., and Jia, Z. (2014) Crystal structure of PhnZ in complex with substrate reveals a di-iron oxygenase mechanism for catabolism of organophosphonates. Proc Natl Acad Sci U S A 111, 5171-5176. 24. Brown, P. M., Caradoc-Davies, T. T., Dickson, J. M., Cooper, G. J., Loomes, K. M., and Baker, E. N. (2006) Crystal structure of a substrate complex of myo-inositol oxygenase, a di-iron oxygenase with a key role in inositol metabolism. Proc Natl Acad Sci U S A 103, 15032-15037. 25. Lovering, A. L., Capeness, M. J., Lambert, C., Hobley, L., and Sockett, R. E. (2011) The structure of an unconventional HD-GYP protein from Bdellovibrio reveals the roles of conserved residues in this class of cyclic-di-GMP phosphodiesterases. MBio 2, e00163- 00111. 26. Benda, C., Ebert, J., Scheltema, R. A., Schiller, H. B., Baumgartner, M., Bonneau, F., Mann, M., and Conti, E. (2014) Structural model of a CRISPR RNA-silencing complex reveals the RNA-target cleavage activity in Cmr4. Mol Cell 56, 43-54. 27. Jung, T. Y., An, Y., Park, K. H., Lee, M. H., Oh, B. H., and Woo, E. (2015) Crystal structure of the Csm1 subunit of the Csm complex and its single-stranded DNA-specific nuclease activity. Structure 23, 782-790. 28. Bellini, D., Caly, D. L., McCarthy, Y., Bumann, M., An, S. Q., Dow, J. M., Ryan, R. P., and Walsh, M. A. (2014) Crystal structure of an HD-GYP domain cyclic-di-GMP phosphodiesterase reveals an enzyme with a novel trinuclear catalytic iron centre. Mol Microbiol 91, 26-38. 29. Gerlt, J. A., Allen, K. N., Almo, S. C., Armstrong, R. N., Babbitt, P. C., Cronan, J. E., Dunaway-Mariano, D., Imker, H. J., Jacobson, M. P., Minor, W., Poulter, C. D., Raushel, F. M., Sali, A., Shoichet, B. K., and Sweedler, J. V. (2011) The Enzyme Function Initiative. Biochemistry 50, 9950-9962. 30. Peng, J., Uygun, S., Kim, T., Wang, Y., Rhee, S. Y., and Chen, J. (2015) Measuring semantic similarities by combining gene ontology annotations and gene co-function networks. BMC Bioinformatics 16. 31. Sjölander, K., Datta, R. S., Shen, Y., and Shoffner, G. M. (2011) Ortholog identification in the presence of domain architecture rearrangement. Brief Bioinform 12, 413-422. 32. Teng, Z., Guo, M., Liu, X., Dai, Q., Wang, C., and Xuan, P. (2013) Measuring gene functional similarity based on group-wise comparison of GO terms. Bioinformatics 29, 1424-1432. 33. Wolf, Y. I., Rogozin, I. B., Kondrashov, A. S., and Koonin, E. V. (2001) Genome alignment, evolution of prokaryotic genome organization, and prediction of gene function using genomic context. Genome Res 11, 356-372. 34. Korbel, J. O., Jensen, L. J., von Mering, C., and Bork, P. (2004) Analysis of genomic context: prediction of functional associations from conserved bidirectionally transcribed gene pairs. Nat Biotechnol 22, 911-917.

190 35. Rogozin, I. B., Makarova, K. S., Murvai, J., Czabarka, E., Wolf, Y. I., Tatusov, R. L., Szekely, L. A., and Koonin, E. V. (2002) Connected gene neighborhoods in prokaryotic genomes. Nucleic Acids Res 30, 2212-2223. 36. Gabaldon, T., and Huynen, M. A. (2004) Prediction of protein function and pathways in the genome era. Cell Mol Life Sci 61, 930-944. 37. Bremer, J. (1983) Carnitine--metabolism and functions. Physiol Rev 63, 1420-1480. 38. Lindstedt, G., Lindstedt, S., Midtvedt, T., and Tofft, M. (1970) Inducible gamma- butyrobetaine-degrading enzymes in Pseudomonas species AK 1. J Bacteriol 101, 1094- 1095. 39. Rebouche, C. J., and Seim, H. (1998) Carnitine metabolism and its regulation in microorganisms and mammals. Annu Rev Nutr 18, 39-61. 40. Lindstedt, G., Linstedt, S., Midtvedt, T., and Tofft, M. (1967) Hydroxylation of Gamma- Butyrobetaine to Carnitine. Biochem J 103, 1271-1282. 41. Lindstedt, G., Lindstedt, S., Olander, B., and Tofft, M. (1968) Alpha-Ketoglutarate and Hydroxylation of Gammabutyrobetaine. Biochim Biophys Acta 158, 503-505. 42. Lindstedt, G., Lindstedt, S., and Nordin, I. (1977) Purification and Properties of Gamma- Butyrobetaine Hydroxylase from Pseudomonas-Sp-Ak-1. Biochemistry 16, 2181-2188. 43. Gerlt, J. A., Bouvier, J. T., Davidson, D. B., Imker, H. J., Sadkhin, B., Slater, D. R., and Whalen, K. L. (2015) Enzyme Function Initiative-Enzyme Similarity Tool (EFI-EST): A web tool for generating protein sequence similarity networks. BBA-Proteins Proteom 1854, 1019-1037. 44. Atkinson, H. J., Morris, J. H., Ferrin, T. E., and Babbitt, P. C. (2009) Using Sequence Similarity Networks for Visualization of Relationships Across Diverse Protein Superfamilies. Plos One 4, e4345. 45. Kazaks, A., Makrecka-Kuka, M., Kuka, J., Voronkova, T., Akopjana, I., Grinberga, S., Pugovics, O., and Tars, K. (2014) Expression and purification of active, stabilized trimethyllysine hydroxylase. Protein Expr Purif 104, 1-6. 46. Pavel, E. G., Zhou, J., Busby, R. W., Gunsior, M., Townsend, C. A., and Solomon, E. I. (1998) Circular dichroism and magnetic circular dichroism spectroscopic studies of the non-heme ferrous active site in clavaminate synthase and its interaction with alpha- ketoglutarate cosubstrate. J. Am. Chem. Soc. 120, 743-753. 47. Matthews, M. L., Krest, C. M., Barr, E. W., Vaillancourt, F. H., Walsh, C. T., Green, M. T., Krebs, C., and Bollinger, J. M., Jr. (2009) Substrate-triggered formation and remarkable stability of the C-H bond-cleaving chloroferryl intermediate in the aliphatic halogenase, SyrB2. Biochemistry 48, 4331-4343. 48. Ryle, M. J., Padmakumar, R., and Hausinger, R. P. (1999) Stopped-flow kinetic analysis of Escherichia coli taurine/alpha-ketoglutarate dioxygenase: Interactions with alpha- ketoglutarate, taurine, and oxygen. Biochemistry 38, 15278-15286. 49. Solomon, E. I., Brunold, T. C., Davis, M. I., Kemsley, J. N., Lee, S. K., Lehnert, N., Neese, F., Skulan, A. J., Yang, Y. S., and Zhou, J. (2000) Geometric and electronic structure/function correlations in non-heme iron enzymes. Chem Rev 100, 235-349. 50. Decker, A., Clay, M. D., and Solomon, E. I. (2006) Spectroscopy and electronic structures of mono- and binuclear high-valent non-heme iron-oxo systems. J Inorg Biochem 100, 697-706. 51. Krebs, C., Fujimori, D. G., Walsh, C. T., and Bollinger, J. M., Jr. (2007) Non-heme Fe(IV)-oxo intermediates. Accounts Chem Res 40, 484-492. 52. Leung, I. K., Krojer, T. J., Kochan, G. T., Henry, L., von Delft, F., Claridge, T. D., Oppermann, U., McDonough, M. A., and Schofield, C. J. (2010) Structural and mechanistic studies on gamma-butyrobetaine hydroxylase. Chem Biol 17, 1316-1324.

191 53. Krissinel, E., and Henrick, K. (2004) Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Acta Crystallogr D Biol Crystallogr 60, 2256-2268. 54. Dunwell, J. M., Purvis, A., and Khuri, S. (2004) Cupins: the most functionally diverse protein superfamily? Phytochemistry 65, 7-17. 55. Xing, G., Hoffart, L. M., Diao, Y., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) A coupled dinuclear iron cluster that is perturbed by substrate binding in myo-inositol oxygenase. Biochemistry 45, 5393-5401. 56. Doi, K., McCracken, J., Peisach, J., and Aisen, P. (1988) The binding of molybdate to uteroferrin. Hyperfine interactions of the binuclear center with 95Mo, 1H, and 2H. J Biol Chem 263, 5757-5763. 57. Hirao, H., and Morokuma, K. (2009) Insights into the (superoxo)Fe(III)Fe(III) intermediate and reaction mechanism of myo-inositol oxygenase: DFT and ONIOM(DFT:MM) study. J Am Chem Soc 131, 17206-17214. 58. Zhou, J., Kelly, W. L., Bachmann, B. O., Gunsior, M., Townsend, C. A., and Solomon, E. I. (2001) Spectroscopic studies of substrate interactions with clavaminate synthase 2, a multifunctional alpha-KG-dependent non-heme iron enzyme: Correlation with mechanisms and reactivities. J. Am. Chem. Soc. 123, 7388-7398. 59. Neidig, M. L., Brown, C. D., Light, K. M., Fujimori, D. G., Nolan, E. M., Price, J. C., Barr, E. W., Bollinger, J. M., Jr., Krebs, C., Walsh, C. T., and Solomon, E. I. (2007) CD and MCD of CytC3 and taurine dioxygenase: Role of the facial triad in alpha-KG- dependent oxygenases. J. Am. Chem. Soc. 129, 14224-14231. 60. Rydzik, A. M., Brem, J., Struwe, W. B., Kochan, G. T., Benesch, J. L., and Schofield, C. J. (2014) Ejection of structural zinc leads to inhibition of gamma-butyrobetaine hydroxylase. Bioorg Med Chem Lett 24, 4954-4957. 61. Kamps, J. J., Khan, A., Choi, H., Lesniak, R. K., Brem, J., Rydzik, A. M., McDonough, M. A., Schofield, C. J., Claridge, T. D., and Mecinovic, J. (2016) Cation-pi Interactions Contribute to Substrate Recognition in gamma-Butyrobetaine Hydroxylase Catalysis. Chemistry 22, 1270-1276. 62. Ma, J. C., and Dougherty, D. A. (1997) The cation-pi interaction. Chem Rev 97, 1303- 1324. 63. Mecozzi, S., West, A. P., and Dougherty, D. A. (1996) Cation-pi interactions in aromatics of biological and medicinal interest: Electrostatic potential surfaces as a useful qualitative guide. Proc Natl Acad Sci USA 93, 10566-10571. 64. Mecozzi, S., West, A. P., and Dougherty, D. A. (1996) Cation-pi interactions in simple aromatics: Electrostatics provide a predictive tool. J Am Chem Soc 118, 2307-2308. 65. Taverna, S. D., Li, H., Ruthenburg, A. J., Allis, C. D., and Patel, D. J. (2007) How chromatin-binding modules interpret histone modifications: lessons from professional pocket pickers. Nat Struct Mol Biol 14, 1025-1040. 66. Kamps, J. J., Huang, J., Poater, J., Xu, C., Pieters, B. J., Dong, A., Min, J., Sherman, W., Beuming, T., Matthias Bickelhaupt, F., Li, H., and Mecinovic, J. (2015) Chemical basis for the recognition of trimethyllysine by epigenetic reader proteins. Nat Commun 6, 8911. 67. Fox, B. G., Hendrich, M. P., Surerus, K. K., Andersson, K. K., Froland, W. A., Lipscomb, J. D., and Münck, E. (1993) Mössbauer, EPR, and ENDOR Studies of the Hydroxylase and Reductase Components of Methane Monooxygenase from Methylosinus trichosporium Ob3b. J Am Chem Soc 115, 3688-3701. 68. Woodland, M. P., Patil, D. S., Cammack, R., and Dalton, H. (1986) Electron-Spin- Resonance Studies of Protein-a of the Soluble Methane Monooxygenase from Methylococcus capsulatus (Bath). Biochim Biophys Acta 873, 237-242.

192 69. Fox, B. G., Surerus, K. K., Münck, E., and Lipscomb, J. D. (1988) Evidence for a μ-Oxo- Bridged Binuclear Iron Cluster in the Hydroxylase Component of Methane Monooxygenase - Mössbauer and Electron-Paramagnetic-Resonance Studies. J Biol Chem 263, 10553-10556. 70. Maddocks, S. E., and Oyston, P. C. (2008) Structure and function of the LysR-type transcriptional regulator (LTTR) family proteins. Microbiology 154, 3609-3623. 71. Kulakova, A. N., Kulakov, L. A., McGrath, J. W., and Quinn, J. P. (2009) The construction of a whole-cell biosensor for phosphonoacetate, based on the LysR-like transcriptional regulator PhnR from Pseudomonas fluorescens 23F. Microb Biotechnol 2, 234-240. 72. Kulakova, A. N., Kulakov, L. A., Villarreal-Chiu, J. F., Gilbert, J. A., McGrath, J. W., and Quinn, J. P. (2009) Expression of the phosphonoalanine-degradative gene cluster from Variovorax sp Pal2 is induced by growth on phosphonoalanine and phosphonopyruvate. FEMS Microbiol Lett 292, 100-106. 73. Kulakova, A. N., Wisdom, G. B., Kulakov, L. A., and Quinn, J. P. (2003) The purification and characterization of phosphonopyruvate hydrolase, a novel carbon- phosphorus bond cleavage enzyme from Variovorax sp Pal2. J Biol Chem 278, 23426- 23431. 74. Mukhamedova, K. S., and Glushenkova, A. I. (2000) Natural phosphonolipids. Chem Nat Compd 36, 329-341. 75. Kirkpatrick, D. S., and Bishop, S. H. (1973) Phosphonoprotein - Characterization of Aminophosphonic Acid Rich Glycoproteins from Sea Anemones. Biochemistry 12, 2829- 2840. 76. Hilderbrand, R. L., Joseph, J. L., Lubansky, H. J., Henderson, T. O. (1977) Biology of Alkylphosphonic Acids, Naval Health Research Center, San Diego, CA. 77. Kittredge, J. S., Isbell, A. F., and Hughes, R. R. (1967) Isolation and Characterization of N-Methyl Derivatives of 2-Aminoethylphosphonic Acid from Sea Anemone Anthopleura xanthogrammica. Biochemistry 6, 289-295. 78. Kittredge, J. S., and Roberts, E. (1969) A Carbon-Phosphorus Bond in Nature. Science 164, 37-42. 79. Kataoka, H., Sakiyama, N., and Makita, M. (1989) Determination of 2- Aminoethylphosphonic Acid and Its N-Methyl Derivative in Animal-Tissues by Gas- Chromatography with Flame Photometric Detection. Agr Biol Chem 53, 2791-2796. 80. McGrath, J. W., Chin, J. P., and Quinn, J. P. (2013) Organophosphonates revealed: new insights into the microbial metabolism of ancient molecules. Nat Rev Microbiol 11, 412- 419. 81. Quinn, J. P., Kulakova, A. N., Cooley, N. A., and McGrath, J. W. (2007) New ways to break an old bond: the bacterial carbon-phosphorus hydrolases and their role in biogeochemical phosphorus cycling. Environ Microbiol 9, 2392-2400. 82. Villarreal-Chiu, J. F., Quinn, J. P., and McGrath, J. W. (2012) The genes and enzymes of phosphonate metabolism by bacteria, and their distribution in the marine environment. Front Microbiol 3, 19. 83. Peck, S. C., and van der Donk, W. A. (2013) Phosphonate biosynthesis and catabolism: a treasure trove of unusual enzymology. Curr Opin Chem Biol 17, 580-588. 84. Kononova, S. V., and Nesmeyanova, M. A. (2002) Phosphonates and their degradation by microorganisms. Biochemistry-Moscow 67, 184-195. 85. Mus, F., Crook, M. B., Garcia, K., Garcia Costas, A., Geddes, B. A., Kouri, E. D., Paramasivan, P., Ryu, M. H., Oldroyd, G. E., Poole, P. S., Udvardi, M. K., Voigt, C. A.,

193 Ane, J. M., and Peters, J. W. (2016) Symbiotic Nitrogen Fixation and the Challenges to Its Extension to Nonlegumes. Appl Environ Microbiol 82, 3698-3710. 86. Bardin, S., Dan, S., Osteras, M., and Finan, T. M. (1996) A phosphate transport system is required for symbiotic nitrogen fixation by Rhizobium meliloti. J Bacteriol 178, 4540- 4547. 87. Eckardt, N. A. (2005) Insights into plant cellular mechanisms: Of phosphate transporters and arbuscular mycorrhizal infection. Plant Cell 17, 3213-3216. 88. Tang, C., Hinsinger, P., Drevon, J. J., and Jaillard, B. (2001) Phosphorus deficiency impairs early nodule functioning and enhances proton release in roots of Medicago truncatula L. Ann Bot 88, 131-138. 89. Høgh-Jensen, H., Schjoerring, J. K., and Soussana, J. F. (2002) The influence of phosphorus deficiency on growth and nitrogen fixation of white clover plants. Ann Bot 90, 745-753. 90. King, G. M. (1988) Distributions and metabolism of quaternary amines in marine sediments., John Wiley & Sons, New York, NY, USA. 91. Bernard, T., Pocard, J. A., Perroud, B., and Lerudulier, D. (1986) Variations in the Response of Salt-Stressed Rhizobium Strains to Betaines. Arch Microbiol 143, 359-364. 92. Lerudulier, D., and Bernard, T. (1986) Salt Tolerance in Rhizobium - a Possible Role for Betaines. FEMS Microbiol Lett 39, 67-72. 93. Lerudulier, D., Bernard, T., Goas, G., and Hamelin, J. (1984) Osmoregulation in Klebsiella pneumoniae - Enhancement of Anaerobic Growth and Nitrogen-Fixation under Stress by Proline Betaine, Gamma-Butyrobetaine, and Other Related-Compounds. Can J Microbiol 30, 299-305. 94. Boncompagni, E., Osteras, M., Poggi, M. C., and Le Rudulier, D. (1999) Occurrence of choline and glycine betaine uptake and metabolism in the family Rhizobiaceae and their roles in osmoprotection. Appl Environ Microbiol 65, 2072-2077. 95. Goldmann, A., Boivin, C., Fleury, V., Message, B., Lecoeur, L., Maille, M., and Tepfer, D. (1991) Betaine Use by Rhizosphere Bacteria - Genes Essential for Trigonelline, Stachydrine, and Carnitine Catabolism in Rhizobium meliloti Are Located on Psym in the Symbiotic Region. Mol Plant Microbe 4, 571-578. 96. Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D., Amin, N., Schwikowski, B., and Ideker, T. (2003) Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res 13, 2498-2504. 97. Markowitz, V. M., Chen, I. M. A., Palaniappan, K., Chu, K., Szeto, E., Grechkin, Y., Ratner, A., Jacob, B., Huang, J. H., Williams, P., Huntemann, M., Anderson, I., Mavromatis, K., Ivanova, N. N., and Kyrpides, N. C. (2012) IMG: the integrated microbial genomes database and comparative analysis system. Nucleic Acids Res 40, D115-D122. 98. Floch, V., Loisel, S., Guenin, E., Herve, A. C., Clement, J. C., Yaouanc, J. J., des Abbayes, H., and Ferec, C. (2000) Cation substitution in cationic phosphonolipids: A new concept to improve transfection activity and decrease cellular toxicity. J. Med. Chem. 43, 4617-4628. 99. Li, J., Ni, X. F., Ling, J., and Shen, Z. Q. (2013) Syntheses and properties of poly(diethyl vinylphosphonate) initiated by lanthanide tris(borohydride) complexes: Polymerization controllability and mechanism. J. Polym. Sci. Pol. Chem. 51, 2409-2415. 100. Lukac, M., Garajova, M., Mrva, M., Devinsky, F., Ondriska, F., and Kubincova, J. (2014) Novel fluorinated dialkylphosphonatocholines: Synthesis, physicochemical properties and antiprotozoal activities against Acanthamoeba spp. J. Fluor. Chem. 164, 10-17.

194 101. Cichowicz, N. R., and Nagorny, P. (2012) Synthesis of Conjugated Polyenes via Sequential Condensation of Sulfonylphosphonates and Aldehydes. Org. Lett. 14, 1058- 1061. 102. Otwinowski, Z., and Minor, W. (1997) Processing of X-ray diffraction data collected in oscillation mode. Method Enzymol 276, 307-326. 103. Murshudov, G. N., Skubak, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F., and Vagin, A. A. (2011) REFMAC5 for the refinement of macromolecular crystal structures. Acta Crystallogr D Biol Crystallogr 67, 355-367. 104. Emsley, P., Lohkamp, B., Scott, W. G., and Cowtan, K. (2010) Features and development of Coot. Acta Crystallogr D Biol Crystallogr 66, 486-501. 105. Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S., and Richardson, D. C. (2010) MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr D Biol Crystallogr 66, 12-21. 106. McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C., and Read, R. J. (2007) Phaser crystallographic software. J Appl Crystallogr 40, 658-674. 107. Langer, G. G., Hazledine, S., Wiegels, T., Carolan, C., and Lamzin, V. S. (2013) Visual automated macromolecular model building. Acta Crystallogr D Biol Crystallogr 69, 635- 641. 108. Lebedev, A. A., Young, P., Isupov, M. N., Moroz, O. V., Vagin, A. A., and Murshudov, G. N. (2012) JLigand: a graphical tool for the CCP4 template-restraint library. Acta Crystallogr D Biol Crystallogr 68, 431-440. 109. Vagin, A., and Teplyakov, A. (1997) MOLREP: an automated program for molecular replacement. J Appl Crystallogr 30, 1022-1025. 110. Winn, M. D., Ballard, C. C., Cowtan, K. D., Dodson, E. J., Emsley, P., Evans, P. R., Keegan, R. M., Krissinel, E. B., Leslie, A. G., McCoy, A., McNicholas, S. J., Murshudov, G. N., Pannu, N. S., Potterton, E. A., Powell, H. R., Read, R. J., Vagin, A., and Wilson, K. S. (2011) Overview of the CCP4 suite and current developments. Acta Crystallogr D Biol Crystallogr 67, 235-242.

195

Chapter 4

Fatty acid decarboxylation to terminal alkene by the diiron oxidase, UndA

Hydrocarbon fossil fuels are being consumed at a much greater rate than they are being re-deposited by natural processes. The ongoing development of new technologies to extract them can postpone the looming crisis, but the imbalance between consumption and production will eventually lead to their depletion.(1) The prodigious capacity of biological systems, particularly bacteria, to rapidly accumulate biomass from atmospheric gases and solar energy could allow them to be deployed for production of renewable energy. However, whereas the important fuels are mixtures of hydrocarbons lacking functional groups that would negatively impact such crucial properties as volatility and viscosity, biomolecules invariably have such functionality, diminishing their versatility and value as alternative fuels. Several enzymes that oxidatively

"defunctionalize" biomolecules to produce alkanes and alkenes have been reported and are being targeted by the biotechnology industry for use in fuel production.(2) Those that derive linear long-chain saturated hydrocarbons from cellular fatty acids, such as soluble cyanobacterial

ADO(3) and its integral membrane functional homologs,(4) have been discussed at length in

Chapters 1 and 2. Whereas these natural products can be utilized as direct replacements for petrodiesel fuel, medium-chain 1-alkenes are attractive “drop-in” biofuel alternatives because of their lower freezing point compared to diesel fuel and potential high recovery in bioprocesses due to volatility.(5, 6) They are also valuable as precursors to numerous industrial chemicals, including lubricants, polymers, pesticides and detergents, owing to the ability of the terminal olefin to be derivatized.(7)

The enzyme UndA from Pseudomonas spp. (> 1,500 homologs) was discovered in 2014 to directly convert medium-chain linear Cn (n = 10, 12, 14) fatty acids to Cn-1 1-alkenes via an

196 oxidative desaturation/decarboxylation reaction producing CO2 as the C1-derived co-product

(Figure 4-1).(8) In this study, 13 Pseudomonas organisms were confirmed to produce 1-undecene

(C11), the highest titer naturally occurring alkene, employing a sensitive solid-phase microextraction technique coupled with gas chromatography-mass spectrometry (GC/MS). A search for the gene(s) responsible for this biological activity was performed by screening a cosmid library generated from the genome of Pseudomonas fluorescens Pf-5. A single gene was found to confer 1-undecene production in a heterologous, non-producing host (E. coli), designated undA. This assignment was confirmed by its genetic disruption that resulted in abolishment of activity in vivo. Subsequent in vitro isolation and biochemical characterization validated the functional assignment of the gene product, UndA, as the sole enzyme responsible for production of 1-undecene from the C12-fatty acid precursor, lauric acid (LA).

Figure 4-1. Chemical reaction catalyzed by the enzyme UndA, where n = 7, 9, or 11.

Further in vitro studies established that UndA requires O2 and Fe(II) to achieve the transformation of LA to 1-undecene.(8) Additionally, a reducing system was determined necessary to achieve multiple turnovers,(8) presumably to provide the balancing two electrons, coupled with the two electron oxidation of the fatty acid, for the four-electron reduction of the dioxygen co-substrate. The reaction mechanism was probed employing substrate isotopologs,

197 revealing that the α-hydrogens (C2) are retained in the final desaturated product,(8) thereby suggesting that β-hydrogen (C3) atom abstraction is a necessary step to enact the decarboxylation reaction. Finally, this study reported a crystal structure of UndA, revealing an active site with a single iron ion bound, coordinated by two histidine residues and one glutamate.(8) This coordination is atypical of a mononuclear non-heme-iron site, in which the ligands are arranged in a “facial triad”, nor are the residues arranged in the primary structure into the HxD/E motif characteristic of Fe/2OG oxygenases.(9) Primarily based on this crystallographic observation, a surprising reaction mechanism was proposed for UndA catalysis (Figure 4-2).(8) In the first step, the Fe(II) cofactor reacts with O2 to generate an Fe(III)-superoxo intermediate. This species then abstracts the substrate β-hydrogen atom, followed by free-radical decarboxylation to yield the terminal unsaturated product and an Fe(IV)-oxo (ferryl) product species. Reduction of the Fe(IV) complex by two electrons then regenerates the resting Fe(II) state for a subsequent turnover.

Figure 4-2. Published mechanism(8) proposed for oxidative decarboxylation employing a mononuclear iron cofactor.

198 This mechanism would be unprecedented for mononuclear non-heme-iron enzymes, which typically require a high-valent intermediate (i.e., Fe(IV)-oxo), a more potent oxidant than the Fe(III)-superoxo species, for aliphatic hydrogen (H•) abstraction. To our knowledge, H• abstraction by Fe(III)-superoxo intermediates is supported only in cases of partially activated C-H bonds and invariably for co-substrate-independent four-electron-oxidation outcomes, as discussed in Chapter 1.(10-13) In these cases, the ability of the Fe(III)-superoxo complex to abstract the H• from the substrate allows the enzyme to extract the initial two electrons to activate the dioxygen

O-O bond, enabling formation of a high-valent intermediate that carries out the subsequent two- electron oxidation of the substrate. The reaction catalyzed by UndA does not share this redox economy, as the conversion of LA to 1-undecene is a two-electron oxidation.

Conversely, dinuclear enzymes are known to carry out reactions similar to that of UndA and involving abstraction of H• from unactivated carbons (as covered extensively in Chapter

1).(14) Deploying two transition metals obviates the need for additional cosubstrates to generate a high-valent state upon activation of dioxygen that would be capable of abstracting the target β-H- atom of the fatty acid substrate, with an estimated BDE of ~95 kcal/mol.(15) Intriguingly, UndA has high structural similarity to a known as-yet-uncharacterized diiron protein, CADD,(16) and shares a structural architecture reminiscent of the ferritin-like fold of diiron oxidases and oxygenases (vide infra). Furthermore, UndA has a functional homolog, UndB,(17) which is an integral membrane-bound enzyme that is expected to utilize a diiron cofactor (see Chapter 1).

Consequently, we have entertained the idea that UndA might, in fact, harbor a dinuclear iron cofactor. This chapter will cover spectroscopic characterization of a diiron cluster in UndA and progress toward establishing this form as the active cofactor, and toward elucidating the mechanism for oxidative decarboxylation by UndA.

199 4.1 Experimental testing of the diiron cofactor hypothesis

The hypothesis that UndA utilizes a diiron cofactor is attractive from a chemical standpoint because a more plausible mechanism can be readily envisioned for oxidative decarboxylation (Figure 4-3) with substantial precedents from the FDCOO family.(18-20) We

II/II posit reaction of the Fe2 state with O2 in the presence of substrate to generate a μ-peroxo-

III/III Fe2 species, similar to intermediate P in sMMO.(21) Homolysis of the O-O bond would result

IV/IV in a high-valent, compound Q-like, Fe2 species that would abstract H• from C3 to generate a

III/IV substrate radical and an Fe2 state.(21-23) This site of C-H bond cleavage is supported by the initial report that both C2 hydrons are retained in the final product.(8) Substrate decarboxylation from this state via various intermediate or transition states (discussed further below) would yield

III/III the Cn-1 1-alkene, CO2, and a μ-oxo-Fe2 product state. Product release, two-electron reduction of the cofactor and binding of substrate would regenerate the reduced cofactor for the next turnover.

Figure 4-3. General mechanistic scheme possible for oxidative decarboxylation employing a diiron cofactor.

200 4.1.1 Structural homology to known diiron proteins

Generally, UndA lacks obvious sequence similarity to known metalloenzymes (< 25 %).

The closest homologs, based on both sequence and structure, belong to a functionally diverse family that shares a heme oxygenase-like domain (16, 24-26). This α-helical domain topology consists of two structural repeats of three-helix bundles. Although the family of proteins that possess this domain do not typically utilize metal cofactors, one example, a protein of unknown function from Chlamydia trachomatis known as CADD (PDB accession code 1RCW), (16) has been crystallized harboring a diiron cluster.(16) Superposition of UndA with CADD ( Figure 4-

4A) using the DALI server (27) verifies their high structural similarity (2.5 Å rmsd over 199 aa).

A closer look at the two active sites reveals that the single Fe site observed in the original structure of UndA overlays with Fe site 2 of CADD, but excitingly, shows that UndA does, in fact, have additional conserved D/H residues poised to coordinate a metal at site 1 (Figure 4-4B).

In the original report, variants were generated substituting only the three residues observed to coordinate the metal center in the crystal structure. These variant proteins did not demonstrate decarboxylation activity. The structural analysis prompted the generation of variants substituting the other three putative ligand residues (in collaboration with Wenjun Zhang, UC-Berkeley).

Substitution of each of these three residues also abolishes UndA activity.

201

Interestingly, these putative ligand residues are arranged in two E/DxxH motifs typical of ferritin-like diiron proteins, despite their lack of overall sequence similarity (Figure 4-5A). The first coordination sphere in UndA is notably lacking a fourth carboxylate ligand, and is supplemented instead by a third histidine residue. This histidine is analogous to the “extra” His ligand in the ferritin-like enzymes, AurF and CmlI, and is likely to have implications for reactivity of the dinuclear cluster, as suggested in Chapter 1. In fact, the heme oxygenase structural domain common to both UndA and CADD is remarkably similar to the ferritin-like

202 structural fold, illustrated by the superposition of the UndA structure with the core helices of sMMOH (Figure 4-5B).

4.1.2 UndA metal incorporation by various preparations

The UndA protein examined in the original study lacked in vitro activity as-isolated, prompting a screen of various organic and inorganic cofactors for recovery of enzymatic activity upon its addition.(8) Among the transition metals tested, Fe(II) was the only metal capable of

203 reconstituting activity. Whereas iron-dependent proteins typically incorporate and retain some transition metals during expression and isolation, no trace metals were detected by ICP-MS in the

UndA protein isolated from heterologous expression in E. coli in rich LB medium. Considering that protein preparation conditions, including over-expression, purification and reconsititution, can dramatically affect metal incorporation and assembly of active metallocofactors, we attempted to optimize these conditions for iron incorporation. UndA isolated from expression in minimal M9-medium supplemented with iron contains 0.5-1 molar equivalents of Fe/protein.

Upon anaerobic reconstitution with Fe(II) of the reduced protein, subsequent oxidation and removal of excess, adventitiously-bound Fe, the Fe incorporation improved to up to 1.5 Fe per

UndA. Thus, we are able to isolate protein in the Fe-bound form and to improve incorporation by methods of reconsititution to allow for further spectroscopic characterization of the metallocofactor assembled.

4.1.3 Interrogating nuclearity and oxidation states of UndA cofactor by Mössbauer and EPR spectroscopies

The 4.2-K/53-mT Mössbauer spectrum of UndA, aerobically isolated from E. coli grown in M9 medium supplemented with 57Fe, exhibits a quadrupole doublet with an isomer shift (δ) of

0.50 mm/s and quadrupole splitting (ΔEQ) of 0.63 mm/s (Figure 4-6), suggesting the presence of

N/O-coordinated high-spin Fe(III). The fact that the majority Fe species (~80% of total 57Fe absorption) gives rise to a quadrupole doublet in a weak applied field establishes that the UndA cofactor has an integer-spin electronic ground state, suggestive of an antiferromagnetically(AF)- coupled high-spin Fe(III) ions in a diiron cluster, as observed in other diiron enzymes. The 4.2-K

Mössbauer spectra with 6- and 8-T applied magnetic field of the same sample (Figure 4-6) support the assignment of an AF coupled Fe2(III/III) cluster. The 6-T spectrum could be

204 simulated reasonably well considering a diamagnetic (S = 0) electronic ground state; however, the lineshape of the 8-T spectrum cannot be simulated considering a strictly diamagnetic ground state, and is reminiscent of systems with antisymmetric exchange of excited electronic states (28).

Figure 4-6. (Top) 4.2-K/53-mT Mössbauer spectrua of two different preparations (black line and black vertical bars) of aerobically isolated UndA protein. (Bottom) 4.2-K/6-T and 8-T Mössbauer spectra of the aerobically isolated UndA protein. The simulation of 6 T spectrum is shown as a solid red line.

The presence of a diiron cluster in the as-isolated UndA protein is further supported by

EPR spectroscopy. Anaerobic addition of sub-stoichiometric amounts of sodium dithionite (0.5 electron per Fe) to the aerobically isolated UndA protein leads to the development of a signal with principal g-values < 2 (g = 1.96, 1.84, 1.77) (Figure 4-7). This EPR signal is characteristic of an S = 1/2 electronic ground state, resulting from an AF-coupled Fe2(II/III) cluster, observed to accumulate to varying yields in several known diiron proteins (29-31). The observation of a

205 mixed-valent Fe2(II/III) EPR signal supports the results from Mössbauer spectroscopy that indicate the presence of a coupled diiron cluster in the as-isolated UndA protein.

4.1.4 Interrogating nuclearity by X-ray absorption (XAS) spectroscopy

The as-isolated and reduced (with excess sodium dithionite) forms of the UndA protein were prepared for analysis by X-ray absorption spectroscopy, which can provide information about the first coordination sphere of a metal cluster. The data from the Fe-K-edge extended edge x-ray absorption fine structure (EXAFS) region (Figures 4-8 and 4-9) were fit based on the parameters given in Tables 4-1 and 4-2, for the oxidized and reduced preparations respectively.

Critically, the Fourier transform of the EXAFS data of the aerobically isolated UndA sample shows an Fe-Fe scatterer with a distance of 3.21 Å, consistent with two iron metals in close enough proximity to couple in a dinuclear cluster. Alternative fits where the single Fe-Fe scattering vector was replaced were considered; however, neither an Fe-C/N/O scatterer nor two

206 Fe-Fe scatterers produced acceptable fits, being deficient in intensity or too broad for the scatterer observed at (R+Δ) of 2.7, respectively. The data were best fit considering a light atom scatterer

(C/N/O) with a short bond distance (R = 1.86 Å), characteristic of a μ-hydroxo ligand bridging the two iron ions. This observation is consistent with the small magnitude of the quadrupole splitting parameter (0.63 mm/s) for the doublets in the Mössbauer spectrum of an equivalent sample. In the Fourier transform of the EXAFS data obtained for the reduced UndA protein, the peak corresponding to the Fe-Fe scatterer at R = 3.21 Å disappears and the data is best fit considering an Fe-Fe scatterer at a distance of 3.97 Å, as to be expected for an increased Fe-Fe vector upon reduction of a diiron cluster.

Figure 4-8. Fe-K-edge EXAFS data (left) and Fourier transform (right) for samples of the aerobically isolated UndA protein. Fit (red) parameters are provided in Table 4-1.

Table 4-1. EXAFS fitting parameters for the aerobically isolated UndA protein. Coordination Scatterer Distance (Å) σ2 number Fe 3.21 0.00207 0.75 C/N/O 1.86 0.0125 1 C/N/O 2.10 0.00797 4 C/N/O 2.57 0.00229 1

207

Table 4-2. EXAFS fitting parameters for the aerobically isolated UndA protein reduced with sodium dithionite. Coordination Scatterer Distance (Å) σ2 number Fe 3.97 0.00903 0.75 C/N/O 2.22 0.0043 2 C/N/O 2.31 0.00446 2.5 C/N/O 3.17 0.00936 2 C/N/O 3.66 0.00645 3

4.1.5 Interrogating nuclearity by x-ray crystallography

In the reported UndA crystal structure, a region of the helix protecting the active site from solvent is disordered,(8) which, importantly, contains the additional residues proposed to coordinate the extra iron site. We originally hypothesized that this disorder, in combination with crystallization artifacts (sulfate bound near unoccupied metal binding site 1), could have induced lability or prevented proper incorporation of a diiron cluster. In addition, the protein preparation used for crystallization in the original study contained insufficient amounts of Fe (1 equivalent

FeII/protein),(8) and thus, we attempted crystallization of the as-isolated UndA protein, in

208 combination with in crystallo methods of Fe supplementation. Anomalous dispersion data collected at the Fe K-edge on these UndA crystals reveal the presence of two anomalous signals

(Figure 4-10), one at the site occupied in the original structure (the position to the right in Figure

4-10) and a second one (albeit weaker), present in the active site. Anomalous density at both positions is present well above background levels (> 8σ) and at levels for which density originating from sulfur atoms is no longer detectable. Additionally, co-crystallization of the

UndA protein with various concentrations of Fe resulted in increased anomalous signals at both sites as a function of Fe provided, providing support that both signals are due to the presence of iron. The second binding site is in the exact position predicted by comparison to the CADD structural homolog. The increased Fe incorporation, however, did not improve the electron density in the disordered helical region enclosing the active site, preventing confident assignments for the orientations of the proposed ligands of the second iron site.

209 4.2 Establishing the active cofactor and operant mechanism for alkene production

Despite the compelling spectroscopic evidence for a diiron cofactor, our experience with other FDCOOs cautions us to entertain the possibility that the native, active cofactor could be a heterodinuclear Mn/Fe cluster, as we first discovered in class I-c RNRs.(32) Deploying a suite of rapid-mixing, transient-state kinetic and spectroscopic techniques, changes to the UndA cofactor(s) can be monitored, which will allow for correlation of reactivity of the relevant cofactor with product formation. Once the active cofactor has been established, spectroscopic techniques can be deployed to characterize reaction intermediates in order to elucidate the mechanism of UndA oxidative decarboxylation.

4.2.1 Substrate triggering effect leads to rapid reaction of reduced UndA with O2.

Diiron oxidases and oxygenases almost exclusively react with dioxygen from the fully

II/II reduced Fe2 oxidation state, with the exception of the aforementioned HD-domain mixed- valent diiron oxygenases (see Chapter 1).(33, 34) The reaction of O2 with UndA reconstituted with Fe(II) was thus assessed initially employing stopped-flow absorption spectroscopy. The reconstituted UndA protein in the reduced state has no distinguishable chromophores in the visible range (Figure 4-11). Upon exposure to dioxygen, in the absence of the LA substrate, a

-1 broad feature slowly develops (kobs =0.5 s ) with an absorption maximum between 330-350 nm typical of μ-oxo-bridged diferric species (Figure 4-11A). On the other hand, the reduced protein preincubated with LA reacts extremely rapidly, resulting in the formation of multiple transient species during a reaction time of less than 1 s (Figure 4-11B). The rate enhancement for the O2 activation step in the presence of the enzyme substrate is typical for metalloenzymes of both the

210 mononuclear and dinuclear classes.(35, 36) Importantly, this substrate triggering effect usually leads to the accumulation of relevant reaction intermediates.

The first transient chromophore observed in the UndA reaction absorbs at 550 nm (ΔA =

0.2) and forms mostly within the deadtime of the measurement (< 1 ms), as observed by the difference with the spectrum of the reduced UndA protein prior to reaction (Figure 4-11B, cyan) and the spectrum 3 ms. Consequently, determination of an observed rate of formation is not possible under these conditions. The disappearance of this feature does not follow an expected exponential decay likely due to its convolution with other chromophores. A distinct second feature (λ ~ 355 nm) also rapidly forms with an observed rate constant of > 200 s-1 and decays

-1 -1 with a rate constant of 30 s (Figure 4-11B, 11D). The decay of this species (kobs = 30 s ) is followed by formation of a sharp signal at 412 nm, a well-established characteristic of tyrosyl radicals (Figure 4-11B). Fitting of the time-dependent trace for the 412 nm dropline, gives observed rate constants of 50 s-1 and 2.4 s-1 for its formation and decay, respectively. Finally, a broad feature absorbing around 330 nm (Figure 4-11B), develops with an observed rate constant of 1.2 s-1 corresponding to an amplitude change of 0.4.

211

Figure 4-11. Reaction of reduced UndA in the absence or presence of substrate with dioxygen, monitored by SF-Abs spectroscopy. Absorption spectra monitoring the reaction of a solution of apo-UndA protein (0.3 mM), reconstituted with 1.8 molar equivalents of FeII (0.54 mM), with 1 mM LA (A) or LA omitted (B), mixed with an equal volume of O2-saturated buffer (~1.8 mM O2). Panels C and D depict the time-dependent traces of specified wavelenths. *The 412 nm trace was obtained by a dropline analysis of the spectrum, which entails subtraction of the average of the 408 and 416 nm traces from the raw 412 nm trace

A priori these species could be associated with any major or minor fraction of the UndA protein, depending on the magnitude of their extinction coefficients. To probe whether these features could be arising from diiron species, the amplitude of these chromophores was evaluated as a function of Fe concentration. In the case that they are correlated with a mononuclear iron cofactor, their amplitudes would be expected to maximize at a 1:1 molar ratio of Fe:UndA. By contrast, if these species originate from a diiron cofactor, the change in absorption would be maximal at a 2:1 molar ratio. Indeed, the amplitudes of all three transient chromophores maximize with two molar equivalents of Fe per UndA protein (Figure 4-12), indicating that these

212 transient species all originate from reaction of O2 with the UndA protein harboring a diiron cofactor. The iron titration curves for each species demonstrate sigmoidal trends. The observed lag at low ratios of Fe:UndA suggests low cooperativity for the affinities of the two binding sites.

The somewhat shallow transition of the curves would imply one tighter binding sight and one with a lower affinity.

Figure 4-12. Amplitude change of chromophores as a function of Fe:UndA. Apo-UndA protein (0.3 mM) was reconstituted with varying molar equivalents of Fe(II), preincubated with 1 mM LA substrate and mixed with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C). Amplitudes were obtained from kinetic fits of time-dependent traces monitoring each specified wavelength. *The amplitudes of the 412 nm traces were obtained from the 412 nm traces by a dropline analysis of the spectrum, which entails subtraction of the average of the 408 and 416 nm traces from the raw 412 nm trace.

4.2.2 Rapid accumulation of differic species upon reaction of reduced UndA with O2.

While the absorption maxima of the transient chromophores observed in SF-Abs experiments can be used as general fingerprints of the corresponding chemical structures, more information about the state of the cofactor of these species can be gleaned from rapid freeze quench (rFQ) Mössbauer spectroscopy. Time points expected to maximally accumulate each transient species were selected from the kinetics of the reaction of reduced UndA with O2 monitored by SF-Abs spectroscopy. The Mössbauer spectrum of the fully reduced state of UndA

213 prior to reaction with O2 demonstrates quadrupole doublets characteristic of high spin Fe(II) with

O/N ligands (isomer shift, δ1 = 1.3 mm/s, quadrupole splitting, ΔEQ1 = 2.85 mm/s; δ2 = 1.3 mm/s,

ΔEQ2 = 3.45 mm/s) (Figure 4-13). The broadness of the spectrum is indicative of substantial heterogeneity in the sample, which could originate from a mixture of substrate-free and substrate- bound states, or could represent both mononuclear and dinuclear ferrous species, as these are not distinguishable at low temperature and weak applied field. After a reaction of the reduced UndA with O2 for only 10 ms, 60 % of the absorption has converted into a new quadrupole doublet with parameters consistent with high spin ferric ions (δ1 = 0.65 mm/s, ΔEQ1 = 0.93 mm/s; δ2 = 0.46 mm/s, ΔEQ2 = 0.89 mm/s) (Figure 4-13). The observation of a quadrupole doublet, as opposed to a magnetically-split spectrum, indicates that this species has an integer-spin electronic ground state, which would not be expected for a mononuclear Fe(III) (S = 5/2), but rather originates from antiferromagnetic coupling of two Fe(III) in a diiron cluster. The 4.2-K/8-T spectrum of this sample (Figure 4-14) verifies a diamagnetic ground state (S = 0). The 53-mT and 8-T spectra can both be simulated using the same parameters. The rapid accumulation of a diferric species confirms that UndA can assemble a dinuclear iron cluster that reacts readily with dioxygen, strongly suggesting that this could be the active form of the cofactor.

214

Figure 4-13. Reaction of reduced UndA with O2 monitored by rFQ-Mössbauer spectroscopy. 4.2- K/53-mT spectra of samples obtained from mixing UndA protein (1.2 mM) reduced with sodium dithionite and preincubated with 10 mM LA substrate with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after various reaction incubation times.

The spectra acquired from samples at later reaction times would at first glance appear to be the same as the spectrum of the 10 ms sample. But upon closer examination, the doublet corresponding to the diferric species in the 10 ms sample evolves over time to a new quadrupole doublet with similar, but distinguishable parameters in the low field spectra. This evolution can be traced in the low field spectra by following the disappearance of the shoulder of the high energy line (Figure 4-13, dotted blue line) and the shift of the doublet low energy line minimum to lower energy at later time points (dotted black line). However, the difference between the species in the 10 ms sample and that in the 5 s sample is most obvious by direct comparison of their spectra obtained with a high applied field, 8-T (Figure 4-14). The low field spectrum of the

215

5 s sample was fit with parameters (doublets in 53 mT spectrum: δ1 = 0.49 mm/s, ΔEQ1 = 0.68 mm/s, δ1 = 0.51 mm/s, ΔEQ1 = 1.18 mm/s) and could represent a final diferric product state of the cofactor in the absence of a reductant. This deduction is consistent with the formation of a broad chromophore absorbing around 330 nm, characteristic of the ligand-to-metal-charge-transfer band of μ-(hydr)oxo-diferric species. The evidence for decay of the first diferric species to a distinct diferric state in less than 5 s correlates to the kinetics of the visible feature absorbing at 550 nm observed in the corresponding SF-Abs experiments.

II/II Considering that most diiron enzymes that activate dioxygen from the Fe2 oxidation state form a diferric-peroxide intermediate state as the first step in the mechanism (as discussed in

Chapter 1), the initial diferric species that forms in the UndA reaction could similarly result from two electron reduction of O2 to a peroxide moiety, oxidizing each ferrous ion by one electron to

216 the ferric oxidation state. Furthermore, the chromophore accumulating on a similar time-scale to this UndA species absorbs at 550 nm, which is consistent with those of known diferric-peroxide species that absorb in the visible region, ranging from 420-700 nm (Table 1-1). The tentative assignment of this transient species in UndA as a diferric-peroxide complex was further tested by examining the relationship of its kinetics of formation on the concentration of O2. As expected for

II/II III/III the bimolecular reaction of Fe2 -UndA with O2 to form a Fe2 -peroxide complex, the observed rate constant for formation of the 550 nm chromophore is dependent on the concentration of O2 in the reaction. Due to the extremely rapid kinetics of formation, this relationship could only be qualitatively assessed under these specific conditions. This observation further supports the assignment of the first diferric species as a peroxo complex.

217 4.2.3 Accumulation of a transient tyrosyl radical.

As previously alluded to, the sharp absorption feature at 412 nm strongly resembles the absorption spectra associated with tyrosyl radicals.(37) Further evidence for this hypothesis was sought employing FQ-EPR spectroscopy, following the kinetics of the 412 nm feature. An EPR signal was detected at low temperature and higher temperatures (up to 70 K) with principal g- values ~ 2 (2.014, 2.005, 1.997) and discernable hyperfine structure (Figure 4-16). These features are all characteristic of an organic-based radical with magnetic hyperfine interactions with nearby protons, consistent with a tyrosyl radical. Additionally, the signal forms and decays with kinetics correlating to those of the 412 nm feature in absorption spectroscopy (Figure 4-16).

However, the lineshape and width of the spectrum are not entirely identical to those observed in other notable enzyme systems (e.g., class I-a RNR-β).(37) We considered that the

218 wider breadth of the spectrum could be due to coupling of the radical with the diiron center.

However, enrichment of the protein with 57Fe, with a nuclear spin of I = ½, does not induce additional broadening of the major signal as was observed in class I-a Ec RNR-β,(38, 39) nor does increasing the temperature result in appreciable narrowing, suggesting that the Tyr• is not coupled to the iron center (Figure 4-17).(40) A survey of EPR signals of tyrosine radicals reported in the literature does show a wide range of lineshapes,(41) reflecting varying degrees of hyperfine interactions with the Cβ-protons. The hyperfine coupling constants and the g-factor components are strongly dependent on the Cβ-Cγ rotation of the phenoxyl ring and the spin density on the Cγ, respectively.(41) While the EPR signal of the UndA radical resembles other

Tyr• previously reported (e.g., ovine prostaglandin H synthase(42)), we are presently preparing

2 the protein with 3,3- H2-tryosine incorporated, which should result in narrowing of the signal linewidth (due to the 6-fold smaller gyromagnetic moment of 2H with respect to 1H), and thereby confirm that this signal originates from a tyrosine radical.

Figure 4-17. CW X-band EPR spectra of rFQ samples obtained from reactions of apo-UndA (1.2 mM), reconstituted with 1.8 molar equivalents of 56- or 57-FeII (2.2 mM), preincubated with 3.6 mM LA and mixed with an equal volume of O2-saturated buffer (~1.8 mM O2 at 5 °C) and freeze quenched after a reaction time of 70 ms. Experimental conditions: temperature = 20 K (70 K), modulation amplitude = 4 G, microwave frequency = 9.625 GHz, microwave power = 20 μW.

219 The UndA protein has 11 tyrosine residues in its primary structure. Candidate residues that could be harboring the observed radical were prioritized by proximity to the active site, or viable translocation pathways via other aromatic residues (e.g., Trp residues), and conservation.

Substitution of these tyrosine residues for redox-inert phenylalanines is expected to result in abolishment of the 412 nm chromophore and the EPR active signal. However, all four variant proteins tested thus far – Y197F, Y104F, Y41F, Y57F – retain the characteristic 412 nm absorption feature without perturbations to its formation kinetics. We are currently exploring the possibilities that (1) the Tyr• is located on another tyrosine residue not yet tested, (2) multiple tyrosyl radicals exist either simultaneously or in parallel alternative pathways, or (3) in vivo self- hydroxylation of the phenylalanine residue regenerates the wild-type protein that is then isolated and tested in vitro.

4.3 Exploring potential mechanisms for UndA-catalyzed fatty acid oxidative decarboxylation

The fast accumulation of these transient species hints at their relevance in productive

UndA turnover, but this assumption can only be confirmed by directly correlating product formation with these changes to the diiron cluster. First, the stoichiometry of products with respect to Fe will be determined in single-turnover reactions with the expectation that theoretical ratios for product:Fe would be 1:1 for a mononuclear cofactor or 1:2 for a dinuclear cofactor. In the latter case, the activity as a function of Fe should reflect the trend observed for the amplitude of the purported diiron transient species observed by SF-Abs (Figure 4-12). Then to determine if any of the observed transient species are kinetically competent, the rate of product(s) will be determined employing rapid-mixing chemical quench experiments. This time-resolved

220 experiment will be essential for narrowing down the number of possible mechanisms that can be envisioned.

Assuming that the diiron form is the active state of UndA and that all transient species observed are potential intermediates in productive turnover, multiple mechanistic variations are

II/II possible. All potential mechanisms start with activation of O2 by the Fe2 cofactor in the

III/III presence of substrate to form a Fe2 -peroxide complex (as shown in Figure 4-3). This first step is supported by the SF-Abs experiments demonstrating substrate triggered O2 activation to rapidly form a species that absorbs at 550 nm, determined by Mössbauer spectroscopy to form concomitantly with a diferric complex. As in the mechanism of sMMOH, internal redox cleavage

IV/IV of the peroxide O-O bond could then give an intermediate Q-like, Fe2 complex (Figure 4-3).

This complex would certainly have the oxidizing potential to abstract a hydrogen-atom (H•) from

III/IV the Cβ-position of the substrate, resulting in a substrate radical and an Fe2 complex (Figure 4-

3). Little experimental evidence has been observed for these high-valent states proposed. An intermediate Q-like species would display a characteristic quadrupole doublet in the Mössbauer spectrum with a diagnostically small isomer shift (δ ≤ 0.2).(21) In the spectrum of the sample obtained at the earliest time point (10 ms), some fraction of this doublet (< 10 %) could be present, but is indiscernible due to the multiple overlapping doublets present in the spectrum.

Employing a deuterium-labelled substrate would be expected to extend the lifetime of the C-H cleaving species and thus enhance its accumulation; however, no such evidence has been observed by absorption spectroscopy. Some evidence may be extant in the available data

III/IV collected for the Fe2 complex. The EPR spectra at early time points shows evidence for a small minority species, hidden underneath the majority signal of the tyrosyl radical, that appears to be isotropic with a g-tensor ~ 2. This minor EPR signal in the spectra enriched in 57Fe appears to split into an unresolved multiplet due to hyperfine coupling to two 57Fe with possible A/g- anisotropy, suggestive that it originates from a metal-based species. The kinetics of this signal

221 appear to qualitatively correlate with the kinetics of the fast forming chromophore that absorbs maximally ~ 355 nm. Both of these spectroscopic features are reminiscent of intermediate X, a

III/IV Fe2 complex, in class I-a RNR-β.(37) However, its minimal accumulation (<0.05 spins/protein) as determined from the EPR spectrum precludes its detection by Mössbauer spectroscopy. Alternatively, the splitting due to 57Fe hyperfine coupling could be observed for a radical species coupled to the metal cluster, as observed for the tryptophan radical cation intermediate in class I-a Ec RNR-β.(38, 39)

Proceeding from the state consisting of an FeIII/IV complex and a substrate radical, two major productive pathways can be envisioned, either invoking the experimentally observed tyrosyl radical or omitting it with the assumption that it is formed in an off-pathway reaction. In the latter case, decarboxylation of the substrate radical could proceed via three distinct intermediate states to the olefin product. The first (and favored) possibility (Figure 4-18, pathway

I-a) is that the C3-substrate radical initiates β-scission of the C1-C2 bond, generating the terminal

•- III/IV alkene and a CO2 radical anion (CO2 ), which would quickly reduce the Fe2 complex via

III/III electron transfer to the product Fe2 state. A second alternative (Figure 4-18, pathway I-b)

III/IV would be electron transfer from the substrate radical to the Fe2 complex, forming a carbocation at C3 that would promote decarboxylation. Third (Figure 4-18, pathway I-c), the substrate radical could couple with a hydroxyl radical from the cofactor, as in the rebound mechanism of sMMOH, to form a hydroxylated substrate intermediate in which the hydroxyl group could be activated as a leaving group by remaining coordinated to the iron cofactor upon substrate decarboxylation. On the other hand, a tyrosyl radical could be generated from H•

III/IV abstraction from the phenolic group by the Fe2 complex, as in the activation pathway of RNR.

This step could either come before (Figure 4-18, pathway II-b) or after (Figure 4-18, pathway II- a) substrate decarboxylation, such that the electron deficient Tyr• would accept the electron

222 donated either from the substrate radical to make the carbocation with concomitant

•- decarboxylation (II-b) or from the CO2 (II-a).

Figure 4-18. Possible mechanisms for substrate decarboxylation following formation of the C3- substrate radical.

The latter possibilities employing a tyrosyl radical to complete the oxidative half-cycle of the UndA mechanism are attractive to consider because they offer a potential mechanism to avoid a dead end hydroxylation outcome. If indeed UndA proceeds through a mechanism that reflects that of sMMOH, how does it promote substrate decarboxylation instead of hydroxylation?

Reduction of the metal cofactor via oxidation of a (nearby) tyrosine residue would mitigate the potential for the cofactor to participate in a hydroxyl-radical group rebound mechanism, thereby decreasing the partition of the substrate radical through alternative pathways and increasing the flux through the desired alkene-forming pathway. This concept that the protein structure directs reactivity of the cofactor has been proposed, and in some cases substantiated, for a number of other classes of metalloenzymes that promote alternative reactions that diverge from the canonical outcome for the family of enzymes. The class of iron- and 2-(oxo)glutarate-dependent dioxygenases perfectly exemplify this principle, manipulating the first and second coordination

223 sphere residues to direct outcomes such as halogenation, desaturation, and cyclization, in contrast to the default hydroxylation reaction.(9) In perfect analogy to the case of UndA, a cytochrome

P450 enzyme, OleTJE, is also known to catalyze oxidative decarboxylation of fatty acids to terminal alkenes.(43-45) Whereas P450s most commonly perform hydroxylation reactions,(46,

47) OleTJE has evolved to suppress this outcome even though it employs the same metallocofactor and oxidizing reaction intermediates. Studies on the reaction mechanism of this enzyme show altered kinetics (i.e., an extended lifetime) for the intermediate species, compound

II, that typically participates in the hydroxyl group rebound mechanism of hydroxylases.(45) It remains unclear how the structure of OleTJE might enable this diversion, but it is possible that a similar strategy could be extended to the function of UndA. For both OleTJE and UndA, a complete understanding of the mechanisms that enable control of reactivity and the relation to structure will be critical for their deployment as biocatalysts in renewable fuel and industrial applications.

224

4.4 Materials & Methods

Materials. Technical grade (> 85% purity) sodium hydrosulfite (dithionite), lauric acid, d12-lauric acid, and tergitol NP-40 were purchased from Sigma-Aldrich (St. Louis, USA). All other chemicals used for protein over-expression and purification were purchased from Sigma-Aldrich

(St. Louis, USA), unless stated otherwise.

Preparation of UndA. The plasmid containing the Pseudomonas fluorescens Pf-5 UndA gene(8) was used to transform Ec BL21 (DE3) competent cells (Invitrogen; Carlsbad, CA). Transformed cells with kanamycin resistance were cultured at 37 °C with shaking (250 rpm) in either rich

Luria Broth (LB) medium with 25 mg/L kanamycin or M9 minimal medium with 25 mg/L kanamycin, 0.2% (v/v) glucose, 0.10 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM

(NH4)2Fe(SO4)2·6H2O until an OD600 of 0.6-0.8 was achieved, at which time UndA expression was induced by addition of 0.25 mM IPTG. M9 medium cultures were also supplemented with

0.125 mM (NH4)2Fe(SO4)2·6H2O at this time. They were then incubated at 18 °C for 24 h with shaking (250 rpm) and then were centrifuged at 8,000 × g for 15 min. Cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cells were resuspended in 50 mM tris-

(hydroxymethyl)aminomethane (Tris)-HCl (pH 8.5) buffer, 300 mM NaCl, and 10% glycerol.

The suspension was lysed by passage through a Microfluidics M-110EH-30 microfluidizer processor at 20,000 psi for 10 min and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 8.5) buffer, 300 mM NaCl, and 10% glycerol.

The column was washed with 50 mM Tris-HCl (pH 8.5) buffer with 300 mM NaCl, 50 mM imidazole and 10% glycerol. The protein was then eluted by washing with 50 mM Tris-HCl (pH

8.5) buffer containing 300 mM NaCl, 250 mM imidazole, and 10% glycerol. Fractions containing

225 the protein were pooled and concentrated at 3,500 × g using a 10K MWCO Macrosep® Advance

Centrifugal Device (Pall Corporation, Port Washington, New York). The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-HCl (pH 8.5) buffer with 300 mM NaCl and 10% glycerol three times for > 4 h each. Protein for crystallography was further purified by size exclusion chromatography using a Sephadex S-200 column (120 mL resin) equilibrated with 50 mM Tris (pH 8.5) buffer with 300 mM NaCl and 10 % glycerol. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with

Commassie staining, and protein concentration was determined by using a molar absorption coefficient of 70,980 M-1•cm-1 at 280 nm. The iron content of the protein over-expressed in LB medium was determined by the ferrozine assay to be < 0.08 Fe/protein and considered to be

“apo”. The iron content of the protein over-expressed in Fe-supplemented M9 minimal medium was determined by the ferrozine assay to range from 0.5-1 Fe/protein. The presence of other metals was assessed by inductively coupled plasma – atomic emission spectroscopy (ICP-AES), which detected less than 0.01 ppm of Mn, Zn and Cu. Chemically-reduced UndA was prepared by anaerobic incubation of the oxidized protein with 10 mM sodium dithionite at 4 °C for 30 min.

The sodium dithionite was removed and the buffer exchanged by using a PD-10 desalting column

(G-25 Sephadex medium, GE Healthcare) equilibrated with 50 mM Tris (pH 8.5) buffer with 300 mM NaCl and 10 % glycerol.

Mössbauer Spectroscopy. Mössbauer spectra were recorded on spectrometers (described previously(48)) from WEB Research (Edina, MN). The spectrometer used to acquire the weak- field spectra is equipped with a Janis SVT-400 variable-temperature cryostat. The spectrometer used to acquire the weak-field spectra is equipped with a Janis 8TMOSS-OM-12SVT variable- temperature cryostat. The external magnetic field was applied parallel to the γ beam. All isomer shifts quoted are relative to the centroid of the spectrum of α-iron metal at room temperature.

226 Simulations of Mössbauer spectra were carried out with the WMOSS spectral analysis software

(www.wmoss.org, WEB Research, Edina, MN).

Continuous-Wave (CW) EPR Spectroscopy. X-Band (~ 9.5 GHz) EPR spectra were acquired on a Bruker ESP-300 spectrometer equipped with an ER/4102 ST resonator (Bruker), an Oxford

Instruments continuous helium flow cryostat, and an Oxford Instruments temperature controller

(ITC 502). For all experiments, quartz tubes with 3 mm inner and 4 mm outer diameter were used

(QSI). The first-derivative EPR spectra were simulated with the MATLAB-based (Mathworks)

EasySpin simulation software (http://easyspin.org/).(49)

X-ray Absorption Spectroscopy (XAS). Samples for XAS measurements were prepared for aerobically isolated UndA with and with sodium dithionite treatment, with a final total iron concentration of 1 mM. X-ray absorption spectra were collected at beamline 7-3 at the Stanford

Synchrotron Radiation Lightsource, operating under ring conditions of 500 mA and 3 GeV. A

Si(220) monochromator (φ = 90°) was used for energy selection of the incident beam, and harmonic rejection was achieved through detuning the monochromator by 50 % and 97.5 % for the ferrous and ferric samples, respectively. The energy of the beam was calibrated using the first inflection point of an Fe foil located upstream of the sample. The sample was placed at 45° with respect to the incident beam, and the iron Kα fluorescence was collected via a 30 element germanium detector. Beam intensity was monitored using an N2-filled ion chamber placed before the sample. The sample temperature was maintained at 10 K using a Displex closed circle cryostat.

Data processing was done using the EXAFSPAK software package(50). The pre-edge background was subtracted using a Gaussian function, and three-segment splines (of orders 2, 3, and 3) were used to subtract the background from the EXAFS region (in PySpline(51)). The

EXAFS were then fit using OPT. Scattering paths for EXAFS fitting were generated using FEFF

9.0(52) using the crystallographic coordinates described in the next section, modified to include

227 an additional iron center 3.4 Å from the crystallographically determined iron. Paths generated from an additional model where the Fe-Fe distance was lengthened to 4.0 Å did not significantly improve the fit of the ferrous sample. During EXAFS fitting, the distance, Debye-Waller factors, and the E0 parameter were allowed to float, while the coordination numbers were systematically varied.

X-ray Crystallography. Crystallographic datasets were collected at the 21ID-F/G and 23ID-B beamlines of the Advanced Photon Source at Argonne National Lab and the resulting diffraction images were processed with the software package HKL2000. Refinement and model building were performed with the programs Refmac5 and COOT, respectively. A summary of data collection and refinement statistics can be found in Table 4-3. Ramachandran outlier analysis and other validation procedures were carried out using the Molprobity server. Figures were generated using the PyMOL molecular graphics software package (Schrödinger LLC).

In an anaerobic chamber at 25 °C, O2-free UndA (15 mg/mL) was cocrystallized with varying amounts of (NH4)2Fe(SO4)2 and 1.75 mM LA. The sample was mixed in a 2:1 ratio with a precipitant solution consisting of 0.1 M MES, pH 6.0-6.5, 1.8 M M (NH4)2SO4 in hanging drop vapor diffusion trials. Additional iron was incorporated via soaking of existing crystals for 4-5 h at 25 °C by 1:1 dilution of crystal drops with a 20 mM solution of (NH4)2Fe(SO4)2 prepared in the crystallization solution. Crystals were prepared for data collection by mounting on rayon loops after cryoprotection in well solution supplemented with 20% (v/v) glycerol, followed by flash freezing in liquid nitrogen. UndA crystallizes in the P2221 space group with two molecules in the asymmetric unit. Phase information for the product-bound data set was obtained by molecular replacement using PHASER with the published UndA structure (PDB accession code 4WWJ), as the search model.

228 Table 4-3. Data collection and refinement statistics for the x-ray structures of UndA┴. UndA Anomalous Data collection Wavelength 0.97857 Å 1.72200 Å

Space group P2221 P2221 Cell Dimensions a, b, c (Å) 68.009, 74.148, 67.962, 74.020, 142.563 142.455 α, β, γ (°) 90, 90, 90 90, 90, 90 Resolution (Å) 50-1.86 (1.89- 80-2.61 (2.66- 1.86) 2.61)

Rmerge 0.058 (0.469) 0.076 (0.163) < I/σ > 29.1 (3.0) 22.5 (12.1)

CC1/2 0.942 Completeness (%) 98.9 (100.0) 100.0 (99.9) Redundancy 5.8 (6.1) 6.3 (5.3)

Refinement Resolution (Å) 1.86 No. reflections 61998

Rwork/Rfree 0.2011/0.2349

No. atoms 4357 Protein 4138 Ligand/ion 32 Water 187 B-factors Protein 27.9 Ion/Ligand 29.0 Water 25.1 r.m.s deviations Bond lengths (Å) 0.0205 Bond angles (°) 1.9023 ┴A single crystal was used to solve each structure. *Highest resolution shell is shown in parenthesis.

USA). The instrument was configured for single-mixing, an optical pathlength of 1 cm, and data acquisition with white light and the photodiode-array (PDA) detector. An O2-free solution containing apo-UndA (0.3 mM), 0.54 mM (NH4)2Fe(SO4)2·6H2O with preincubation or omission

229 of 1 mM LA was mixed with an equal volume of O2-saturated (1.8 mM at 5 °C) 50 mM Tris-HCl buffer (pH 8.5) with 300 mM NaCl and 10 % glycerol. Time-dependent absorption spectra (1000 points) were acquired on a logarithmic time base after a dead-time of the stopped-flow apparatus

(~ 1 ms). Changes of reaction conditions are provided in the appropriate figure legends.

ΔA-versus-time traces were fit by sums of exponential functions as described in the main text, corresponding to parallel first-order reactions with different amplitudes (ΔA1, ΔA2, ΔA3) and observed rate constants (k1, k2, k3), where A0 is the initial absorbance.

A1( ) A2( ) (eq. 1)

Preparation of Freeze-Quenched Samples for Electron Paramagnetic Resonance (EPR) and

Mössbauer Spectroscopies. General procedures for freeze-quench (FQ) EPR and Mössbauer experiments have been published previously.(48, 53, 54) Single-mixing FQ-Mössbauer and EPR experiments were conducted at 5 °C in which the contents an O2-free solution containing chemically-reduced UndA (1.2 mM), preincubated with 10 mM LA was mixed with an equal volume of O2-saturated (1.8 mM at 5 °C) 50 mM Tris-HCl buffer (pH 8.5) with 300 mM NaCl and 10 % glycerol. The resultant solution allowed to react for varying incubation times before being rapidly frozen (freeze-quenched) in isopentane at -150 °C. Minor deviations from these general experimental procedures are noted in the appropriate text and figure legends.

4.5 Acknowledgements

Dr. Bo Zhang carried out the Mössbauer data collection and analysis. SF-Abs and FQ experiements were performed together with Dr. Bo Zhang. Dr. Beth Blaesi carried out the

EXAFS data collection and analysis. Prof. Wenjun Zhang (University of California, Berkeley) provided the UndA plasmid and performed the mutagenesis studies. Devon van Cura generated the tyrosine variant proteins and assisted with SF-Abs experiments on those variants.

230 This material is based upon work supported by the National Science Foundation under

Award No. CHE-1610676. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the

National Science Foundation.

4.6 References

1. Shafiee, S., and Topal, E. (2009) When will fossil fuel reserves be diminished?, Energy Policy 37, 181-189. 2. Wang, J., and Zhu, K. (2017) Microbial production of alka(e)ne biofuels, Curr Opin Biotechnol 50, 11-18. 3. Schirmer, A., Rude, M. A., Li, X., Popova, E., and del Cardayre, S. B. (2010) Microbial biosynthesis of alkanes, Science 329, 559-562. 4. Bernard, A., Domergue, F., Pascal, S., Jetter, R., Renne, C., Faure, J. D., Haslam, R. P., Napier, J. A., Lessire, R., and Joubes, J. (2012) Reconstitution of plant alkane biosynthesis in yeast demonstrates that Arabidopsis ECERIFERUM1 and ECERIFERUM3 are core components of a very-long-chain alkane synthesis complex, Plant Cell 24, 3106-3118. 5. Lennen, R. M., and Pfleger, B. F. (2013) Microbial production of fatty acid-derived fuels and chemicals, Current Opinion in Biotechnology 24, 1044-1053. 6. Peralta-Yahya, P. P., Zhang, F. Z., del Cardayre, S. B., and Keasling, J. D. (2012) Microbial engineering for the production of advanced biofuels, Nature 488, 320-328. 7. Ray, S., Rao, P. V. C., and Choudary, N. V. (2012) Poly-alpha-olefin-based synthetic lubricants: a short review on various synthetic routes, Lubrication Science 24, 23-44. 8. Rui, Z., Li, X., Zhu, X., Liu, J., Domigan, B., Barr, I., Cate, J. H., and Zhang, W. (2014) Microbial biosynthesis of medium-chain 1-alkenes by a nonheme iron oxidase, Proc Natl Acad Sci U S A 111, 18237-18242. 9. (2015) 2-Oxoglutarate-Dependent Oxygenases, The Royal Society of Chemistry. 10. van der Donk, W. A., Krebs, C., and Bollinger, J. M., Jr. (2010) Substrate activation by iron superoxo intermediates, Curr Opin Struct Biol 20, 673-683. 11. Bollinger, J. M., Jr., and Krebs, C. (2007) Enzymatic C-H activation by metal-superoxo intermediates, Curr Opin Chem Biol 11, 151-158. 12. Xing, G., Diao, Y., Hoffart, L. M., Barr, E. W., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) Evidence for C-H cleavage by an iron- superoxide complex in the glycol cleavage reaction catalyzed by myo-inositol oxygenase, Proc Natl Acad Sci U S A 103, 6130-6135. 13. Tamanaha, E., Zhang, B., Guo, Y., Chang, W.-c., Barr, E. W., Xing, G., St Clair, J., Ye, S., Neese, F., Bollinger, J. M., Jr., and Krebs, C. (2016) Spectroscopic Evidence for the Two C-H-Cleaving Intermediates of Aspergillus nidulans Isopenicillin N Synthase, J Am Chem Soc 138, 8862-8874. 14. Lippard, S. J. (2005) Hydroxylation of C-H bonds at carboxylate-bridged diiron centres, Philos Trans A Math Phys Eng Sci 363, 861-877.

231 15. Luo, Y. R. (2007) Comprehensive Handbook of Chemical Bond Energies, CRC Press, Boca Raton, FL. 16. Schwarzenbacher, R., Stenner-Liewen, F., Liewen, H., Robinson, H., Yuan, H., Bossy- Wetzel, E., Reed, J. C., and Liddington, R. C. (2004) Structure of the Chlamydia protein CADD reveals a redox enzyme that modulates host cell apoptosis, J Biol Chem 279, 29320-29324. 17. Rui, Z., Harris, N. C., Zhu, X., Huang, W., and Zhang, W. (2015) Discovery of a Family of Desaturase-Like Enzymes for 1-Alkene Biosynthesis, ACS Catalysis 5, 7091-7094. 18. Merkx, M., Kopp, D. A., Sazinsky, M. H., Blazyk, J. L., Müller, J., and Lippard, S. J. (2001) Dioxygen Activation and Methane Hydroxylation by Soluble Methane Monooxygenase: A Tale of Two Irons and Three Proteins, Angew Chem Int Ed Engl 40, 2782-2807. 19. Fox, B. G., Lyle, K. S., and Rogge, C. E. (2004) Reactions of the diiron enzyme stearoyl- acyl carrier protein desaturase, Acc Chem Res 37, 421-429. 20. Shanklin, J., Guy, J. E., Mishra, G., and Lindqvist, Y. (2009) Desaturases: emerging models for understanding functional diversification of diiron-containing enzymes, J Biol Chem 284, 18559-18563. 21. Liu, K. E., Valentine, A. M., Wang, D., Huynh, B. H., Edmondson, D. E., Salifoglou, A., and Lippard, S. J. (1995) Kinetic and Spectroscopic Characterization of Intermediates and Component Interactions in Reactions of Methane Monooxygenase from Methylococcus capsulatus (Bath), J Am Chem Soc 177, 10174-10185. 22. Shu, L., Nesheim, J. C., Kauffmann, K., Münck, E., Lipscomb, J. D., and Que, L., Jr. (1997) An Fe2IVO2 diamond core structure for the key intermediate Q of methane monooxygenase, Science 275, 515-518. 23. Nesheim, J. C., and Lipscomb, J. D. (1996) Large kinetic isotope effects in methane oxidation catalyzed by methane monooxygenase: evidence for C-H bond cleavage in a reaction cycle intermediate, Biochemistry 35, 10240-10247. 24. Toms, A. V., Haas, A. L., Park, J. H., Begley, T. P., and Ealick, S. E. (2005) Structural characterization of the regulatory proteins TenA and TenI from Bacillus subtilis and identification of TenA as a thiaminase II, Biochemistry 44, 2319-2329. 25. Lad, L., Friedman, J., Li, H. Y., Bhaskar, B., de Montellano, P. R. O., and Poulos, T. L. (2004) Crystal structure of human heme oxygenase-1 in a complex with biliverdin, Biochemistry 43, 3793-3801. 26. Magnusson, O. T., Toyama, H., Saeki, M., Rojas, A., Reed, J. C., Liddington, R. C., Klinman, J. P., and Schwarzenbacher, R. (2004) Quinone biogenesis: Structure and mechanism of PqqC, the final catalyst in the production of pyrroloquinoline quinone, Proc Natl Acad Sci USA 101, 7913-7918. 27. Holm, L., and Rosenstrom, P. (2010) Dali server: conservation mapping in 3D, Nucleic Acids Res 38, W545-549. 28. Kauffmann, K. E., Popescu, C. V., Dong, Y. H., Lipscomb, J. D., Que, L., and Münck, E. (1998) Mössbauer evidence for antisymmetric exchange in a diferric synthetic complex and diferric methane monooxygenase, J Am Chem Soc 120, 8739-8746. 29. Fox, B. G., Hendrich, M. P., Surerus, K. K., Andersson, K. K., Froland, W. A., Lipscomb, J. D., and Münck, E. (1993) Mössbauer, Epr, and Endor Studies of the Hydroxylase and Reductase Components of Methane Monooxygenase from Methylosinus-Trichosporium Ob3b, J Am Chem Soc 115, 3688-3701. 30. Woodland, M. P., Patil, D. S., Cammack, R., and Dalton, H. (1986) Electron-Spin- Resonance Studies of Protein-a of the Soluble Methane Monooxygenase from Methylococcus-Capsulatus (Bath), Biochim Biophys Acta 873, 237-242.

232 31. Fox, B. G., and Lipscomb, J. D. (1988) Purification of a High Specific Activity Methane Monooxygenase Hydroxylase Component from a Type-II Methanotroph, Biochem Bioph Res Co 154, 165-170. 32. Jiang, W., Yun, D., Saleh, L., Barr, E. W., Xing, G., Hoffart, L. M., Maslak, M. A., Krebs, C., and Bollinger, J. M., Jr. (2007) A manganese(IV)/iron(III) cofactor in Chlamydia trachomatis ribonucleotide reductase, Science 316, 1188-1191. 33. Xing, G., Barr, E. W., Diao, Y., Hoffart, L. M., Prabhu, K. S., Arner, R. J., Reddy, C. C., Krebs, C., and Bollinger, J. M., Jr. (2006) Oxygen activation by a mixed-valent, diiron(II/III) cluster in the glycol cleavage reaction catalyzed by myo-inositol oxygenase, Biochemistry 45, 5402-5412. 34. Wörsdörfer, B., Lingaraju, M., Yennawar, N. H., Boal, A. K., Krebs, C., Bollinger, J. M., Jr., and Pandelia, M.-E. (2013) Organophosphonate-degrading PhnZ reveals an emerging family of HD domain mixed-valent diiron oxygenases, Proc Natl Acad Sci U S A 110, 18874-18879. 35. Broadwater, J. A., Ai, J., Loehr, T. M., Sanders-Loehr, J., and Fox, B. G. (1998) Peroxodiferric intermediate of stearoyl-acyl carrier protein delta 9 desaturase: oxidase reactivity during single turnover and implications for the mechanism of desaturation, Biochemistry 37, 14664-14671. 36. Matthews, M. L., Krest, C. M., Barr, E. W., Vaillancourt, F. H., Walsh, C. T., Green, M. T., Krebs, C., and Bollinger, J. M., Jr. (2009) Substrate-triggered formation and remarkable stability of the C-H bond-cleaving chloroferryl intermediate in the aliphatic halogenase, SyrB2, Biochemistry 48, 4331-4343. 37. Bollinger, J. M., Jr., Edmondson, D. E., Huynh, B. H., Filley, J., Norton, J. R., and Stubbe, J. (1991) Mechanism of assembly of the tyrosyl radical-dinuclear iron cluster cofactor of ribonucleotide reductase, Science 253, 292-298. 38. Baldwin, J., Krebs, C., Ley, B. A., Edmondson, D. E., Huynh, B. H., and Bollinger, J. M., Jr. (2000) Mechanism of rapid electron transfer during oxygen activation in the R2 subunit of Escherichia coli ribonucleotide reductase. 1. Evidence for a transient tryptophan radical, J Am Chem Soc 122, 12195-12206. 39. Krebs, C., Chen, S. X., Baldwin, J., Ley, B. A., Patel, U., Edmondson, D. E., Huynh, B. H., and Bollinger, J. M., Jr. (2000) Mechanism of rapid electron transfer during oxygen activation in the R2 subunit of Escherichia coli ribonucleotide reductase. 2. Evidence for and consequences of blocked electron transfer in the W48F variant, J Am Chem Soc 122, 12207-12219. 40. Chang, W. C., Guo, Y., Wang, C., Butch, S. E., Rosenzweig, A. C., Boal, A. K., Krebs, C., and Bollinger, J. M., Jr. (2014) Mechanism of the C5 stereoinversion reaction in the biosynthesis of carbapenem antibiotics, Science 343, 1140-1144. 41. Svistunenko, D. A., and Cooper, C. E. (2004) A new method of identifying the site of tyrosyl radicals in proteins, Biophys J 87, 582-595. 42. Dorlet, P., Seibold, S. A., Babcock, G. T., Gerfen, G. J., Smith, W. L., Tsai, A. L., and Un, S. (2002) High-field EPR study of tyrosyl radicals in prostaglandin H(2) synthase-1, Biochemistry 41, 6107-6114. 43. Rude, M. A., Baron, T. S., Brubaker, S., Alibhai, M., Del Cardayre, S. B., and Schirmer, A. (2011) Terminal olefin (1-alkene) biosynthesis by a novel P450 fatty acid decarboxylase from Jeotgalicoccus species, Appl Environ Microbiol 77, 1718-1727. 44. Belcher, J., McLean, K. J., Matthews, S., Woodward, L. S., Fisher, K., Rigby, S. E., Nelson, D. R., Potts, D., Baynham, M. T., Parker, D. A., Leys, D., and Munro, A. W. (2014) Structure and biochemical properties of the alkene producing cytochrome P450

233 OleTJE (CYP152L1) from the Jeotgalicoccus sp. 8456 bacterium, J Biol Chem 289, 6535-6550. 45. Grant, J. L., Mitchell, M. E., and Makris, T. M. (2016) Catalytic strategy for carbon- carbon bond scission by the cytochrome P450 OleT, Proc Natl Acad Sci U S A 113, 10049-10054. 46. Sono, M., Roach, M. P., Coulter, E. D., and Dawson, J. H. (1996) Heme-Containing Oxygenases, Chem Rev 96, 2841-2888. 47. Denisov, I. G., Makris, T. M., Sligar, S. G., and Schlichting, I. (2005) Structure and chemistry of cytochrome P450, Chem Rev 105, 2253-2277. 48. Price, J. C., Barr, E. W., Tirupati, B., Bollinger, J. M., Jr., and Krebs, C. (2003) The first direct characterization of a high-valent iron intermediate in the reaction of an alpha- ketoglutarate-dependent dioxygenase: A high-spin Fe(IV) complex in taurine/alpha- ketoglutarate dioxygenase (TauD) from Escherichia coli, Biochemistry 42, 7497-7508. 49. Stoll, S., and Schweiger, A. (2006) EasySpin, a comprehensive software package for spectral simulation and analysis in EPR, J Magn Reson 178, 42-55. 50. George, G. N. EXAFSPAK, SSRL, SLAC, Stanford University, Stanford, CA. 51. Tenderholt, A., Hedman, B., and Hodgson, K. O. (2007) PySpline: A modern, cross- platform program for the processing of raw averaged XAS edge and EXAFS data, X-Ray Absorption Fine Structure-Xafs13 882, 105-+. 52. Rehr, J. J., Kas, J. J., Vila, F. D., Prange, M. P., and Jorissen, K. (2010) Parameter-free calculations of X-ray spectra with FEFF9, Phys Chem Chem Phys 12, 5503-5513. 53. Ravi, N., Bollinger, J. M., Jr., Huynh, B. H., Edmondson, D. E., and Stubbe, J. (1994) Mechanism of Assembly of the Tyrosyl Radical-Diiron(Iii) Cofactor of Escherichia-Coli Ribonucleotide Reductase 1. Mössbauer Characterization of the Diferric Radical Precursor, J Am Chem Soc 116, 8007-8014. 54. Jiang, W., Hoffart, L. M., Krebs, C., and Bollinger, J. M., Jr. (2007) A manganese(IV)/iron(IV) intermediate in assembly of the manganese(IV)/iron(III) cofactor of Chlamydia trachomatis ribonucleotide reductase, Biochemistry 46, 8709- 8716.

234 Appendix A

Chapter 2 Supporting Information

2 2 Procedures for syntheses of aldehyde substrates. n-1-[ H]-decanal, n-2,2-[ H]2-decanal were

1,2 prepared according to published procedures. As depicted in Scheme S1, BD3 was used to install

2 2 deuteria onto the C1 carbonyl in the synthesis of n-1-[ H]-decanal. In preparation of n-2,2-[ H]2-

2 decanal, the deuteria were introduced via solvent exchange using CH3O H.

To prepare the 3-thia substrate analogue, n-hepta-1-thiol was treated with 1.0 equivalent of NaH followed by 1.2 equivalent of 1,1-dimethoxy-ethyl-2-tosylate at 0˚C to yield the protected n-3-thiadecanal. Deprotection was carried out by treatment with excess formic acid (~ 10 equivalents) to transform the dimethoxy-protected compound into the aldehyde. For synthesis of

2 2 n-2,2-[ H]2-3-thiadecanal, 1,1-dimethoxy-ethyl-2,2-[ H]2-2-tosylate was used as starting material to introduce deuteria at C2.

2 2 Scheme S1. Summary of procedures for syntheses of n-1-[ H]-decanal, n-2,2-[ H]2-decanal, n-3- 2 thiadecanal and n-2,2-[ H]2-3-thiadecanal. Preparation of Np AAR. AAR was prepared as previously described.3

235 Preparation of Np ADO. The plasmid containing the codon-optimized Np ADO gene4

(Npun_R1711; accession code YP_001865325) was used to transform Ec BL21 (DE3) competent cells (Invitrogen; Carlsbad, CA). ADO was overexpressed and purified aerobically by procedures similar to those used in previous work,5 but modified as described below. Transformed cells were incubated with shaking (250 rpm) at 37 °C in M9 minimal medium with 50 mg/L kanamycin,

0.2% (v/v) glucose, 0.1 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM (NH4)2Fe(SO4)2·6H2O until an OD600 of 0.6-0.8 was reached. Expression was then induced by addition of 0.25 mM iso- propyl-β-D-1-thiogalactopyranoside (IPTG), and the medium was supplemented with 0.125 mM

(NH4)2Fe(SO4)2·6H2O. Cultures were incubated with shaking (250 rpm) at 18 °C for 20-24 h. Cell pellets from centrifugation at 8 000g for 15 min were flash frozen in liquid nitrogen and stored at

-80 °C. Cells were resuspended in 50 mM potassium phosphate buffer (pH 8.0), 10 mM imidazole and 300 mM NaCl and were lysed by two passages through a French pressure cell at >

1 200 psi. After centrifugation at 22 000g for 20 min, the lysate supernatant was loaded onto a

Ni2+-nitrilo-tris-acetate (NTA) immobilized affinity chromatography column (~100 mL resin per

500 mL lysate) equilibrated with 50 mM potassium phosphate buffer (pH 8.0), 10 mM imidazole and 300 mM NaCl. The column was washed with 50 mM potassium phosphate buffer (pH 8.0) containing 40 mM imidazole and 300 mM NaCl. Protein was then eluted by washing with 50 mM potassium phosphate buffer (pH 8.0), 250 mM imidazole, 100 mM NaCl. Fractions containing the protein were pooled, and the protein was concentrated at 3 500g using a 10K MWCO Macrosep®

Advance Centrifugal Device (Pall Corporation, Port Washington, NY). In an anoxic chamber

(Labmaster, MBraun, Stratham, NH), the ADO cofactor was reduced by treatment with ≥ 10 mM sodium dithionite for 30 min at 4 °C. Buffer exchange and sodium dithionite removal were achieved using a pre-packed PD-10 desalting column (G-25 Sephadex medium, GE Healthcare) equilibrated with O2-free 50 mM sodium 2-[4-(2-hydroxyethyl)-piperazin-1-yl]ethanesulfonate

(HEPES) buffer (pH 7.5), 10% glycerol. Protein aliquots were flash frozen and stored in liquid

236

N2. Protein purity was assessed by SDS-PAGE with Coomassie staining, and protein concentration was determined using a molar absorption coefficient at 280 nm of 22 920 M-1∙cm-1.5

Iron content was determined by inductively coupled plasma atomic emission spectroscopy (ICP-

AES) by Mr. Henry Gong at The Pennsylvania State Materials Research Institute. The concentrations of ADO quoted in all experiments are of diiron cluster, calculated as half the measured iron concentration.

Preparation of Syn. 6803 Fds. The Syn. 6803 ssl0020 (petF), ssl2559, slr1828, sll1584, sll1382, ssl3044, and slr0150 genes were each individually codon-optimized for overexpression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by GeneArt (Regensburg, Germany) to give seven Syn. 6803 Fd plasmids. Ec BL21 (DE3) competent cells (Invitrogen, Carlsbad, CA) were doubly transformed with the Fd plasmid and the plasmid pDB1282, containing the Azotobacter vinelandii isc operon in a pARA13 expression vector.6 Transformed cells with kanamycin and ampicillin resistance were cultured at 37 °C with shaking (250 rpm) in M9 minimal medium with 25 mg/L kanamycin, 150 mg/L ampicillin, 0.2%

(v/v) glucose, 0.10 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM (NH4)2Fe(SO4)2·6H2O until an OD600 of 0.3 was reached. Expression of the isc operon was induced by addition of 2 g/L arabinose and 0.20 mM L-cysteine to cell cultures. Incubation with shaking (250 rpm) at 37 °C was continued until an OD600 of 0.6-0.8 was achieved, at which time Fd expression was induced by addition of 0.25 mM IPTG. Cultures were also supplemented with 0.125 mM

(NH4)2Fe(SO4)2·6H2O at this time. They were then incubated at 18 °C for 24 h with shaking (250 rpm) and then were centrifuged at 8 000g for 15 min. Dark brown cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cells were resuspended in 50 mM tris-

(hydroxymethyl)aminomethane (Tris)-HCl (pH 7.5) buffer, 150 mM NaCl, and 10% glycerol.

The cell suspension was lysed by two passages through a French pressure cell at >1 200 psi and centrifuged at 22 000g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized

237 affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM

NaCl, 250 mM imidazole, and 10% glycerol. Fractions containing the protein were pooled and concentrated at 3 500g using a 10K MWCO Macrosep® Advance Centrifugal Device (Pall

Corporation, Port Washington, NY). The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol three times for > 6 h each. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with Commassie staining, and protein concentration was determined using a Direct Detect spectrometer (EMD Millipore, Darmstadt, Germany). [2Fe-2S] cluster occupancy was determined by EPR spin quantification of a protein sample reduced by treatment with 10 mM sodium dithionite. A frozen solution of Cu2+-EDTA was used as the spin concentration standard.7 The PetF concentrations quoted in all experiments are of the [2Fe-2S] cofactor. Chemically reduced PetF was prepared by O2-free incubation of the oxidized protein with 10 mM sodium dithionite (> 2-fold excess with respect to [PetF] cofactor) at 4 °C for 30 min. The sodium dithionite was removed and the buffer exchanged by using a PD-10 desalting column (G-25 Sephadex medium, GE Healthcare) equilibrated with O2-free 50 mM sodium

HEPES, pH 7.5.

Preparation of Syn. 6803 slr1643 FNR. The Syn. 6803 slr1643 (petH) gene encoding FNR was codon-optimized for over expression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by GeneArt (Regensburg, Germany). Ec BL21

238 expression was then induced by addition of 0.25 mM IPTG, and cell cultures were incubated at

30 °C with shaking (250 rpm) for 4 h. Cultures were centrifuged at 8 000g for 15 min; cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cell pellets were resuspended in 50 mM Tris-

HCl (pH 7.5) buffer containing 150 mM NaCl and 10% glycerol. The suspension was lysed by two passages through a French pressure cell at > 1 200 psi and centrifuged at 22 000g for 20 min.

The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column

(~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol. The column was washed with 50 mM Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 30 mM imidazole and 10% glycerol. The protein was eluted by washing with 50 mM

Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 250 mM imidazole and 10% glycerol. Fractions containing the protein were pooled and concentrated at 3 500g using a 10K MWCO Macrosep®

Advance Centrifugal Device (Pall Corporation, Port Washington, NY). The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol three times for > 6 h each. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with Coomassie staining, and protein concentration was determined by using a molar absorption coefficient at 280 nm of 50 210 M-

1∙cm-1. FNR was reconstituted with equimolar flavin adenine dinucleotide (FAD+) and filtered to remove excess FAD+ using a 30K MWCO Amicon Ultra centrifugal filter (EMD Millipore,

Merck KGaA, Darmstadt, Germany). Bound FAD+ was quantified by using a molar absorption coefficient at 460 nm of 10 800 M-1∙cm-1. The FNR concentrations quoted in all experiments are of the bound FAD+ cofactor.

Preparation of Dechloromonas aromatica Chlorite Dismutase (Cld). Cld was prepared as previously described.8

239 Cartesian coordinates (Å) of the optimized geometries used as input for the single-point energy calculations. Model 1 (C9H18OO•)

C -3.304290 0.423657 -0.083290 C -2.889476 -1.046764 0.078334 H -2.995131 0.815708 -1.061666 H -2.838934 1.051926 0.688208 C -1.371557 -1.263077 -0.027766 H -3.240309 -1.420233 1.052020 H -3.395349 -1.654956 -0.686410 H -1.021261 -0.889006 -1.002558 H -0.865866 -0.652592 0.736724 C -0.947327 -2.731153 0.133845 H -1.453314 -3.340525 -0.631099 C 0.570527 -2.947527 0.028729 H 1.077048 -2.339822 0.794504 C 0.989403 -4.416974 0.190195 C 2.507361 -4.632687 0.087491 H 0.636344 -4.790367 1.163553 H 3.013034 -4.026572 0.853872 H -1.299004 -3.104324 1.108302 H 0.923254 -2.574763 -0.945403 H 0.485363 -5.024741 -0.576644 H 2.861519 -4.264216 -0.886788 H -4.392158 0.544070 -0.002508 C 2.910799 -6.107803 0.254030 H 2.418065 -6.717642 -0.514803 C 4.417843 -6.289062 0.144537 H 2.573935 -6.477848 1.231450 H 4.966318 -5.757987 0.930290 H 4.807490 -6.001498 -0.838351 O 4.704799 -7.721555 0.315699 O 6.002521 -7.979187 0.255862

240

Model 2 (2-[S]-C8H18OO•) C -3.227600 0.439888 -0.340455 C -2.852061 -0.995531 0.058274 H -3.013847 0.623100 -1.402146 H -2.657382 1.173916 0.244786 C -1.363822 -1.313231 -0.156960 H -3.107475 -1.159539 1.115791 H -3.460496 -1.706462 -0.520432 H -1.108755 -1.146947 -1.215202 H -0.755209 -0.602050 0.423114 C -0.980062 -2.748182 0.236904 H -1.589126 -3.458398 -0.343162 C 0.508170 -3.063414 0.017562 H 1.119128 -2.359050 0.602065 C 0.880530 -4.504706 0.406796 C 2.367315 -4.794542 0.171892 H 0.636035 -4.673258 1.464972 H 2.994356 -4.131312 0.778718 H -1.233335 -2.913928 1.295361 H 0.762744 -2.896084 -1.039781 H 0.273158 -5.208832 -0.178765 H 2.624753 -4.661914 -0.883752 H -4.294798 0.634911 -0.174748 H 4.935739 -7.480927 0.804787 H 5.035029 -5.684982 0.652774 O 4.753372 -6.723088 -1.131823 O 6.040819 -6.589844 -1.414568 C 4.515557 -6.590870 0.329000 S 2.751204 -6.537739 0.649208

241 Cartesian coordinates (Å) of the optimized geometry of the sulfinyl radical (SO•) in a tripeptide model of the Ala70-Cys71-Gly72 triad (Np ADO numbering). This structure was used as the input for the single-point energy calculations. The original, non-optimized coordinates were obtained from the crystal structure of Se ADO (PDB accession code: 4RC5). C -3.49625582614082 2.05081029942340 -14.11952281337485 O -2.95866620266036 1.06868109761725 -14.62834699731787 N -3.69382364500963 2.22707141515584 -12.78472435942896 C -3.17343169919328 1.32352162659688 -11.77039724330367 C -4.22238821056885 0.34496900466382 -11.17708823090019 O -3.92044099933675 -0.37387354901434 -10.21815890833368 C -2.49611813850262 2.08741539774942 -10.62151889298729 S -0.94939527635186 2.94005663336615 -11.13594057365249 N -5.42941336824022 0.32267676290544 -11.77122871523500 H -2.23708603768534 1.38656387543501 -9.82342965444937 H -3.14406785398336 2.87131264505256 -10.20862626651170 H -2.42899930261949 0.68361895202023 -12.25835653088118 H -4.11467395108544 3.09816371967482 -12.49126642674058 H -3.88698766502313 2.89292997972284 -14.71828350406941 H -6.12042568330514 -0.33950703224945 -11.45028422545990 O -0.05873219490069 1.84553457102295 -11.74810694243275 H -5.63622796539300 0.89587083085718 -12.57297362492119

Cartesian coordinates (Å) of the optimized geometry of the sulfinyl radical (SO•) in a tripeptide model of the Ala70-Cys71-Gly72 triad (Np ADO numbering). This structure reproduces the observed hyperfine coupling constants of the Cβ protons, considering a Hβ(1)-Cβ-S-O dihedral angle of 122 °. C -3.44888450464006 1.97432700512647 -14.08547303316077 O -3.06648842511528 0.90913469350437 -14.56781479508432 N -3.59695358567152 2.21405450054116 -12.75563343657273 C -3.13808899001597 1.28942421009148 -11.73011806227250 C -4.23019182288836 0.32437136548650 -11.19184868635209 O -3.95608298615895 -0.46379969693626 -10.27958377692405 C -2.51400956443067 2.03147074967609 -10.54500856156268 S -0.81185605227318 2.69602166218487 -10.88401762984867 N -5.43791601937155 0.39323332824965 -11.77790709476728 H -2.41775762482445 1.35649694455233 -9.68967198833732 H -3.12727594175437 2.88710450260059 -10.23481803428083 H -2.38945451227811 0.63839584984866 -12.19621769154491 H -3.87123449171419 3.14709827763502 -12.48117980133327 H -3.72967233338986 2.84267971018971 -14.70782036451384 H -6.15651498200209 -0.25795202875311 -11.49799377623266 O -0.45646786927022 2.31910069035941 -12.33175120581728 H -5.61828431420117 1.02465446564304 -12.54139597139473

242 Table S1. Comparison of experimentally determined and computationally predicted electronic structure parameters of the substrate-derived peroxyl radical, obtained from the simulation of the EPR and HYSCORE spectra and from single-point calculations on geometry-optimized models 2 2 of the 1,1-[ H]2-C9H18OO• and the 1,1-[ H]2-2-[S]-C8H18OO•, respectively.

g1 g2 g3 A1, MHz A2, MHz A3, MHz 2 1,1-[ H]2-C9H16OO• experiment 2.037 2.010 2.004 3.1/1.5 1.5/3.1 3.3/3.3 theory 2.028 2.009 2.002 3.0/3.2 3.2/3.4 5.3/5.4 2 1,1-[ H]2-2-[S]-C9H16OO• experiment 2.035 2.009 2.004 1.8/1.4 1.3/0.7 1.3/0.7 theory 2.027 2.009 2.002 5.6/4.4 3.6/2.2 3.5/2.0

Table S2: Comparison of the yield of the Cys71-SO• radical, as determined by FQ-EPR experiments (t = 4 s), and the normalized (with respect to WT ADO) activities (alkane yield) of ADO variants obtained from multiple turnover assays. The yield of the Cys71-SO• is given as a percentage with respect to the total ADO cofactor concentration. Normalized ADO variant % Cys71-SO• % Activity WT 8.0 100 Y123F 3.9 36 C71A nd 73 C107A 7.7 92 C117A 6.7 63 Y18F 3.9 57 Y22F 4.3 77

243 Table S3. Calculated A- and g-tensors for the Cys71-SO• obtained from single-point energy calculations on geometry-optimized structures, in which the Hβ(1)-Cβ-S-O dihedral angle was varied between 0 and 180 °. Finer steps around the angle value of 120 ° that reproduced best the observed hyperfine couplings were also undertaken, and the calculated parameters are presented for 118 ° and 122 °.

Hβ(1)-Cβ-S-O

g1 g2 g3 Aβ(1)iso (MHz) Aβ(2)iso (MHz) dihedral angle (°)

0 2.002 2.011 2.021 -0.62 30.2

20 2.002 2.012 2.022 4.23 16.6

40 2.002 2.012 2.022 17.86 4.33

60 2.002 2.012 2.022 33.5 -1.3

80 2.002 2.012 2.022 45.1 2.9

100 2.002 2.012 2.021 48.3 17.1

118 2.002 2.012 2.021 39.4 34.8

120 2.002 2.012 2.021 39.1 35.3

122 2.002 2.012 2.021 37.8 36.6

140 2.002 2.012 2.021 21.1 46.3

160 2.002 2.012 2.021 4.2 42.9

180 2.002 2.011 2.022 -1.26 31.6

244

Figure S1. CW EPR spectrum of reduced [2Fe-2S]1+-containing PetF (0.28 mM) generated by reduction of the as-isolated PetF with sodium dithionite (10 mM) for 30 min in the absence of O2. The signal of the reduced PetF has principal g-values of 2.05, 1.95, and 1.88. Spin quantitation relative to a 256 μM Cu2+-EDTA standard yielded 162 μM of [2Fe-2S]1+ cluster (~ 58 % of [PetF]). Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz frequency, modulation amplitude = 0.3 mT.

245

III/III Figure S2. SF-Abs kinetic traces monitoring formation of the Fe2 -PHA intermediate (λmax = 450 nm). All experiments were performed at 5 °C in 50 mM sodium HEPES buffer, pH 7.5. (A) Effect of protein concentration: a solution of 0.1 mM (green), 0.2 mM (pink) or 0.4 mM (blue) II/II Fe2 -ADO with 10 mM octanal was mixed with an equal volume of O2-saturated buffer (1.8 mM O2). Data were acquired with monochromatic light and a PMT detector. (B) Effect of oxygen II/II concentration: a solution of 0.1 mM Fe2 -ADO with 10 mM decanal was mixed with an equal volume of either O2-saturated (1.8 mM O2, black) or air-saturated (~ 0.38 mM O2, red) buffer. Data were acquired with white light and a PDA detector. The apparent discrepancy in amplitudes is attributed to the diminished rate of intermediate formation with air-saturated buffer and the enhanced rate of photolytic decay under illumination by the intense polychromatic light. (C) II/II Effect of substrate chain-length: A solution containing 0.9 mM Fe2 -ADO, 10 μM Cld and 10 mM decanal (red) or octanal (purple) was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the final solution. Data were acquired using monochromatic light and a PMT detector. (D) Effect of high oxygen concentrations: Black trace: A solution of 1.49 mM II/II Fe2 -ADO with 16.6 mM octanal was mixed with an equal volume of O2-saturated buffer (1.8 II/II mM O2). Blue trace: A solution of 1.49 mM Fe2 -ADO with 20 mM octanal and 10 μM Cld was mixed with an equal volume of 10 mM NaClO2 to generate ~ 5 mM O2 in the reaction. Data were acquired with monochromatic light and a PMT detector. In the experiment with O2-saturated -1 -1 buffer, ~ 80 % intermediate was generated (A450 = 0.7, ε450 = 1 200 mM ∙cm ), in good agreement with the Mössbauer analysis of a parallel sample.

246

Figure S3. In vitro coupled Np AAR and Np ADO activity assays (t = 1 h at 37 °C) carried out with all seven putative Syn. 6803 Fds (encoded by the genes in the figure) together with Syn. 6803 FNR and NADPH, as well as the spinach Fd/FNR/N system (blue). Heptadecane and octadecanal products were detected by GC-MS. Assay conditions: 10 µM AAR, 20 µM ADO, 0.2 mM 1-[13C]-stearoyl-ACP, 4 mM NADPH, 7.8 μM Fd and 7.8 µM FNR.

Figure S4. SF-Abs experiments performed at 5 °C in 50 mM sodium HEPES, pH 7.5 (A) Reference absorption spectra of [2Fe-2S]1+ (black trace) and [2Fe-2S]2+ (red trace) PetF were obtained from a single-mixing experiment, in which chemically reduced PetF (0.10 mM) was mixed with an equal volume of either O2-free or O2-saturated buffer (1.8 mM) for 250 s, respectively. (B) SF-Abs kinetic trace monitoring oxidation of the [2Fe-2S]1+ PetF at 423 nm (blue trace) in the single-mixing experiment described in A. Fitting a single exponential function -1 to the data (black dashed line) gives ΔA423 = 0.180(3) and kobs = 0.014(1) s . (C) Individual absorption spectra used to mathematically construct the “0” ms reference spectrum (black trace), III/III which is additionally shown in Figure 2A in the main text. The spectrum of the Fe2 -PHA intermediate (green dotted trace) was generated by mixing of a solution containing 0.050 mM II/II Fe2 -ADO and 10 mM octanal with an equal volume of O2-saturated buffer. The spectrum of the [2Fe-2S]1+ PetF (orange dotted trace) was obtained from mixing chemically reduced PetF (0.10 mM) with an equal volume of O2-free buffer.

247

III/III Figure S5. Kinetics of PetF oxidation by the as-isolated Fe2 -ADO. (A) SF-Abs kinetic traces monitoring PetF oxidation at 423 nm in single-mix experiments in which a solution of O2-free, chemically reduced PetF ([final] = 0.050 mM) was mixed with an equal volume of an O2-free III/III solution of varying concentrations of as-isolated Fe2 -ADO to give the final ratios indicated in the figure legend. The data were best fit by two exponential functions with the second observed rate constant (kobs2) being concentration dependent. (B) Plot of the kobs2 as a function of ADO -1 -1 concentration, from which a second-order rate constant of 8.1 mM ∙s was obtained.

III/III Figure S6. 4.2-K/53-mT Mössbauer spectra of the Fe2 species generated upon reaction of the III/III Fe2 -PHA intermediate for 0.010 s with two equiv of chemically reduced PetF. The spectra were generated by adding back varying amounts of the experimental reference spectrum of the III/III Fe2 -PHA intermediate [47 % (blue line), 55 % (black line), 63 % (red line)] to the difference spectrum shown in Figure 3C of the main manuscript. The spectra were scaled to have equal intensity for ease of comparison. The resulting spectra are qualitatively similar, consisting of at least two quadrupole doublets with parameters of δ1 ≈ 0.50 mm/s, ΔEQ1 ≈ 0.75 mm/s and δ2 ≈ III/III 0.55 mm/s, ΔEQ2 ≈ 1.5 mm/s. Two additional doublets of equal intensity, arising from the Fe2 - PHA intermediate, were included in fits to better determine the maximum quantity of intermediate to have undergone reduction. The relative contributions of all four quadrupole doublets were allowed to vary. This analysis suggested that the derived spectrum obtained after adding back 63 % of the intermediate contains maximally 5 % of the intermediate, whereas in the other two cases considered, the contributions of the intermediate were sufficiently removed to make the remainder indiscernible. The above analysis places an upper limit of 58 % (of the initial ~ 80 %) on the amount of intermediate reduced by the PetF at 0.010 s.

248 Figure S7. Analysis and derivation of the reference Mössbauer III/III spectrum of the Fe2 species generated upon reduction of the III/III Fe2 -PHA intermediate by PetF after a reaction time of 0.22 s. (A) 4.2-K/53-mT FQ-Mössbauer spectrum of a sample 57 II/II generated by reacting the Fe2 -ADO•decanal complex with III/III Cld/NaClO2 for 30 s. The Fe2 -PHA intermediate accumulated to ~ 75% of the total iron, and the remaining II/II fraction (~ 25 %) is in the substrate-free Fe2 form. (B) III/III Spectrum of a FQ sample obtained after reaction of the Fe2 - PHA intermediate, generated as in A, with two equiv of II/II chemically reduced PetF for 0.22 s. The Fe2 -ADO in this sample (~ 39 %) is comprised of ~ 27% of the substrate-free II/II II/II Fe2 form and ~ 12 % of the substrate-bound Fe2 form. (C) B−A difference spectrum (black bars), overlaid with 61% of the III/III experimental reference spectrum of the Fe2 -PHA intermediate (red line). (D) Spectrum obtained after addition of III/III 61% of the experimental reference spectrum of the Fe2 - PHA intermediate to the difference spectrum, C. The II/II discrepancy (~12 %) in the total amount of the remaining Fe2 species in the 0.22 s reaction sample compared to that in the control sample corresponds to the contribution from the II/II unreacted, substrate-bound Fe2 -ADO (overlaid in blue). This difference suggests that a fraction of the substrate-bound, reduced ADO did not react with O2 in the first mixing event. III/III Considering that the Fe2 -PHA intermediate has thus accumulated to a lesser extent (i.e., ~ 61 %) than in the analogous control sample (spectrum A), it is likely that the Cld used in these experiments may have partially inactivated during the course of the experiment (6-8 h). (E) Experimental III/III reference spectrum of the Fe2 species generated after III/III reduction of the Fe2 -PHA intermediate for 0.22 s. The spectrum was obtained after subtraction of the contribution II/II from unreacted substrate-bound Fe2 form (12 %) from spectrum D. The spectrum can be analyzed as two quadrupole doublets with parameters δ1 ≈ 0.45 mm/s, ΔEQ1 ≈ 0.8 mm/s (purple line) and δ2 ≈ 0.55 mm/s, ΔEQ2 ≈ 1.6 mm/s (orange line). The total simulation is shown as the dark red line. It should be noted that these parameters are given to provide the range of quadrupole and isomer shift parameters of the diferric clusters. Because these features are relatively broad, unreolved quadrupole doublets, the Mössbauer parameters are somewhat correlated with the linewidth parameters, making the errors in both relatively large.

249 Figure S8. Detailed procedure by which the quantity of the III/III III/III Fe2 species formed upon reduction of the Fe2 -PHA intermediate by 0.5 equiv of chemically reduced PetF for 0.22 s was assessed by FQ-Mössbauer spectroscopy. The difference spectra (blue lines, Figure 4C) were generated as described in III/III the main text by subtraction of the spectrum of the Fe2 -PHA intermediate control sample (Figure 4A) from the spectrum of the 0.22 s reaction sample (Figure 4B). The reference spectra of III/III the product Fe2 species (red lines) were then constructed after addition of varying percentages (20 – 50 %) of the III/III experimental reference spectrum of the Fe2 -PHA intermediate (black lines) to the difference spectra (blue lines). III/III Addition of contributions from the Fe2 -PHA intermediate in III/III excess of 25% leads to unreasonable spectra for the new Fe2 III/III species. The features of the Fe2 -PHA intermediate dominate these resultant spectra (red lines), as inferred by inspection of the high energy line of the doublet with ΔEQ = 0.49 mm/s of the intermediate (black lines), indicating that these are overestimations of the fraction of intermediate that decayed.

250

Figure S9. CW X-band EPR spectra of the substrate-derived peroxyl radical species. (A) The III/III 1 Fe2 -PHA intermediate was formed in the presence of 2,2-[ H]2-decanal (black spectrum) or 2 MeO (B) 2,2-[ H]2-decanal (red spectrum) substrate and was reacted with equimolar reduced PMS for 0.010 s. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT. The spectra generated using 1 2 the 2,2-[ H]2- and the 2,2-[ H]2-decanal were simulated (blue lines) by using the same set of parameters and hyperfine couplings obtained from the simulation of the HYSCORE spectra. The lineshape of the spectra was reproduced by considering an isotropic Voigtian lineshape with peak-to-peak linewidths of 0.17 and 0.24 mT, for the Lorentzian and Gaussian components, respectively, and an anisotropic broadening of [30 22 28] MHz (H-Strain) or [30 18 28] MHz for 1 2 the spectra generated using the 2,2-[ H]2-decanal and the 2,2-[ H]2-decanal, respectively. The only adjustable parameters in the refinement were the g-values. They varied by ± 0.002, which is well within experimental error of the measurements.

251

Figure S10. CW EPR spectra of samples quenched at 0.020 s in sequential-mixing FQ II/II experiments. In the first mix, 0.60 mM Fe2 -ADO with (red, green) or without (blue) 16.6 mM decanal substrate was reacted for 30 s with O2-saturated (1.8 mM O2 at 5 °C) 50 mM sodium III/III HEPES, pH 7.5, buffer to allow for formation of the Fe2 -PHA intermediate. In the second mix, the intermediate was reacted with either equimolar reduced MeOPMS (red, blue) or buffer (green) for 0.020 s prior to freezing in liquid 2-methylbutane. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.380 GHz, modulation amplitude = 0.3 mT.

252

Figure S11. CW X-band EPR spectra of samples prepared by employing substrate isotopologues and/or 56Fe/57Fe-labeled ADO. The substrate-based peroxyl radical (A) or the Cys71-centered sulfinyl radical (B, C) radicals were targeted for maximal accumulation. (A) Either 57Fe-labeled 56 II/II (1.4 mM) or Fe-unlabeled (0.90 mM) Fe2 -ADO with 10 mM decanal and 10 μM Cld was III/III reacted in a first mix with 10 mM NaClO2 to generate the Fe2 -PHA intermediate, which was then mixed with two equiv of chemically reduced PetF and quenched in liquid 2-methylbutane after being allowed to react for 0.010 s. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.481 GHz, modulation amplitude = 0.4 mT. The 56Fe- sample spectrum was scaled by a factor of 1.56 for comparison. (B) Either 57Fe-labeled or 56Fe- II/II 13 12 unlabeled Fe2 -ADO with 1-[ C]- or 1-[ C]-octanal, respectively, was reacted with O2- III/III saturated buffer (1.8 mM) to generate the Fe2 -intermediate, which was then mixed with equimolar MeOPMS and quenched in 2-methylbutane after being allowed to react for 0.054 s. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave 56 II/II frequency 9.381 GHz, modulation amplitude = 0.3 mT. (C) 1.49 mM Fe-unlabeled Fe2 -ADO 2 1 with 14.3 mM 2,2-[ H]2-decanal or 2,2-[ H]2-octanal was reacted with O2-saturated buffer (1.8 III/III MeO mM) to generate the Fe2 -intermediate, which was then mixed with equimolar PMS and 2 quenched in 2-methylbutane after being allowed to react for 3 min (2,2-[ H]2-decanal) or 0.2 s 1 1 (2,2-[ H]2-octanal). The spectrum of the sample generated with 2,2-[ H]2-octanal was scaled by 0.5 for ease of comparison. Experimental conditions: temperature = 28 K, microwave power = 0.2 mW, microwave frequency 9.483 GHz, modulation amplitude = 0.4 mT.

253

Figure S12. CW EPR spectra of the substrate-derived peroxyl radical obtained using 3- II/II 1 thiadecanal substrate analogues. A solution of Fe2 -ADO containing either the 2,2-[ H]2-3- 2 thiadecanal (blue) or the 2,2-[ H]2-3-thiadecanal (black) was reacted in a first mix with O2- III/III saturated (1.8 mM) 50 mM sodium HEPES, pH 7.5 to generate the Fe2 -PHA intermediate. III/III MeO The Fe2 -PHA intermediate was reacted in a second mix with equimolar reduced PMS for 0.020 s prior to freezing in 2-methylbutane. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.381 GHz, modulation amplitude = 0.3 mT.

254

Figure S13. X-Band HYSCORE spectra recorded on FQ samples quenched after reaction of the III/III MeO maximally accumulated Fe2 -PHA intermediate with equimolar reduced PMS for 0.02 s in 2 2 double-mixing experiments using the 2,2-[ H]2-decanal (left column) or the 2,2-[ H]2-3- thiadecanal analog (right column). The numerically simulated spectra are overlaid with grey contours, and the simulation parameters are given in Table S1. Experimental conditions: microwave frequency = 9.388 GHz, τ = 200 ns, π/2 = 8 ns, T = 30 K.

255

Figure S14. CW X-band EPR spectrum of the peroxyl radical obtained using octadecanal as the II/II substrate. In a first mix, 0.60 mM Fe2 -ADO containing 1.5 mM octadecanal and 10 μM Cld III/III was reacted with an equal volume of 10 mM NaClO2 for 30 s to generate the Fe2 -PHA intermediate, which was then mixed with an equal volume of 0.60 mM chemically reduced PetF and allowed to react for 0.010 s before being freeze-quenched. Experimental conditions: temperature = 20 K, microwave power = 0.2 mW, microwave frequency = 9.481 GHz, modulation amplitude = 0.4 mT.

256

Figure S15. CW X-band FQ-EPR spectra of the Cys71-SO• in WT ADO and ADO variants. (A) II/II FQ-EPR time-course using octanal as the substrate. The WT Fe2 -ADO∙octanal complex was III/III reacted in a first mix with the Cld/NaClO2 system to generate the Fe2 -PHA intermediate, which was reacted in a second mix with two equiv of chemically reduced PetF. (B) A solution of II/II 0.9 mM Fe2 -ADO (WT, C107A, C117A) with 16.6 mM octanal and 10 μM Cld was mixed III/III with an equal volume of 10 mM NaClO2 for 30 s to generate the Fe2 -PHA intermediate. This solution was then mixed with an equal volume of equimolar MeOPMS and samples were quenched after 4 s. Experimental conditions: temperature = 60 K (detection of the Cys71-SO• signal at temperatures > 70 K was not attempted because the EPR samples were freeze-quenched and could not be safely annealed to higher temperatures), microwave power = 2 mW, microwave frequency = 9.481 GHz, modulation amplitude =0.4 mT.

257

Figure S16. CW X-band EPR spectra of the Cys71-SO• in WT and ADO variants. (A) Spectra of III/III FQ-EPR samples quenched after a 0.010 s reaction of the Fe2 -PHA intermediate with MeO II/II equimolar reduced PMS. The Fe2 -ADO•octanal complex ([ADO]f = 0.225 mM WT, C71A, C107A and Y123F, 0.150 mM Y18F and 0.125 mM Y22F) was mixed in equal volume with the III/III Cld/NaClO2 system for 30 s to generate the Fe2 -PHA intermediate. The Y18F and Y22F spectra were multiplied by factors of 1.8 and 1.5, respectively, for ease of comparison. (B) Spectral deconvolution depicting the individual contributions from the peroxyl radical, the MeOPMS semiquinone (SQ) radical and the sulfinyl radical signals. The spectrum of a sample III/III MeO obtained from a 0.010 s reaction of the WT Fe2 -PHA intermediate with PMS was fitted as a linear combination of the experimental reference spectra of the substrate-derived peroxyl radical, the MeOPMS SQ radical and the Cys71-SO• radical (“4 s FQ sample” spectrum), giving fractional contributions for the three radicals of 0.248, 0.497 and 0.255, respectively, of the total spin concentration. Experimental conditions: temperature = 60 K, microwave power = 2 mW, microwave frequency = 9.480 GHz, modulation amplitude = 0.4 mT.

258

Figure S17. Geometry optimized structures of the sulfinyl radical (SO•) harbored on Cys71 of ADO; the adjacent amino acids (Ala70 and Gly72) have also been included in the model. The orientation of the S-O moiety found in typical sulfinyl radicals (A) and the S-O orientation in the sulfinyl radical on Cys71 that supports significant spin density on both of the cysteine β protons (B), as experimentally observed. The plot of the total spin density is shown as mesh contours with a cut-off of V/ r3.

According to Gordy et al., the isotropic hyperfine coupling constants of the Cβ protons in sulfur- S SCH 2 S centered radicals are given by the expression: Aβ(1) = ρ (3p)∙Qβ ∙cos (θ), in which ρ (3p) is the SCH spin density on the sulfur, Qβ is the effective hyperfine coupling constant (86 MHz), and θ is the dihedral angle.9 Assuming that the Cβ protons are bonded in an sp3 configuration, the dihedral angles differ by 120 degrees, and, thus, there is only one independent angle, θ. In sulfinyl radicals, the isotropic hyperfine coupling constants of the Cβ proton on Hβ(1) is typically negligible, and only the Hβ(2) carries significant spin density, with an average isotropic hyperfine coupling constant of ~ 1.4 mT (~ 40 MHz). This hyperfine structure is evident in the EPR spectra of reported sulfinyl radicals in solution and in proteins, in which only one proton hyperfine coupling dominates the rhombic EPR spectrum with g = [2.02, 2.01, 2.0].10-12 In the spectrum of the Cys71-based radical in ADO, the g-values measured coincide with those of a sulfinyl radical 10-15 (RSO•) and not with those of a thiyl (RS•) radical. In the latter, a large gx value is characteristically observed (gx = 2.10-2.29) as a result of the near degeneracy of the second lone pair orbital.13 In addition, the isotropic hyperfine coupling constants for such radicals are significantly larger than those obtained from the numerical simulation of the spectrum of the Cys71-based radical. A thiyl radical configuration is thus incompatible with the electronic structure of the Cys71-centered radical. By contrast, the experimentally measured g-tensor of the Cys71-centered radical is essential identical to that of a sulfinyl radical.10-12,15 Considering that this radical is generated concomitantly with the decay of the substrate-derived peroxyl radical, a molecular structure of the type Cys-SO• is likely for the Cys71-centered radical signal and is in agreement with previous observations in other proteins.10-12 Therefore, to assess whether the observed hyperfine coupling constants on the Cβ protons obtained from the numerical simulation of the EPR spectrum are consistent with this assignment, the dihedral angle Hβ(1)-Cβ-S-O was systematically varied between 0 and 180 degrees in 20 degree increments (see Table S3; in the crystal structure, it is 65 °). Each of the structures was geometry optimized. Hyperfine coupling constants and g-tensors were obtained from single-point calculations on each of the optimized geometries. The g-tensor principal components are virtually insensitive to the changes in the

259 dihedral angle, whereas the hyperfine coupling constants show a strong dependence on the dihedral angle (Table S3).

Figure S18. The tertiary structure of Se ADO (PDB accession code: 4RC5) is shown in light blue in ribbon representation. The Ala70-Cys71-Gly72 triad (Np ADO numbering), in which the cysteine has been modeled as Cys71-SO•, is shown in stick format (blue) and was obtained by an unrestricted geometry optimization. In this orientation, only one of the Cβ protons has significant spin density. The Cys71-SO• geometry that reproduces the experimental data [Hβ(1)-Cβ-S-O dihedral angle = 120 °] is shown in stick format and colored by heteroatom identity. In this orientation, the SO• is ‘shielded’ by a hydrophobic environment, in particular by a phenylalanine residue (Phe132, Se ADO numbering).

REFERENCES (1) Schirmer, A.; Rude, M. A.; Li, X.; Popova, E.; del Cardayre, S. B. Science 2010, 329, 559. (2) Li, N.; Chang, W.-c.; Warui, D. M.; Booker, S. J.; Krebs, C.; Bollinger, J. M., Jr. Biochemistry 2012, 51, 7908. (3) Warui, D. M.; Pandelia, M.-E.; Rajakovich, L. J.; Krebs, C.; Bollinger, J. M., Jr.; Booker, S. J. Biochemistry 2015, 54, 1006. (4) Warui, D. M.; Li, N.; Nørgaard, H.; Krebs, C.; Bollinger, J. M., Jr.; Booker, S. J. J Am Chem Soc 2011, 133, 3316. (5) Pandelia, M.-E.; Li, N.; Nørgaard, H.; Warui, D. M.; Rajakovich, L. J.; Chang, W.-c.; Booker, S. J.; Krebs, C.; Bollinger, J. M., Jr. J Am Chem Soc 2013, 135, 15801. (6) Johnson, D. C.; Unciuleac, M. C.; Dean, D. R. J Bacteriol 2006, 188, 7551. (7) Hagen, W. R. Dalton transactions 2006, 4415. (8) Dassama, L. M. K.; Yosca, T. H.; Conner, D. A.; Lee, M. H.; Blanc, B.; Streit, B. R.; Green, M. T.; DuBois, J. L.; Krebs, C.; Bollinger, J. M., Jr. Biochemistry 2012, 51, 1607. (9) Gordy, W. Intrepretation of Nuclean Coupling in Oriented Free Radicals in Theory and Applications of Electron Spin Resonance; John Wiley and Sons: New York, 1980. (10) Reddy, S. G.; Wong, K. K.; Parast, C. V.; Peisach, J.; Magliozzo, R. S.; Kozarich, J. W. Biochemistry 1998, 37, 558. (11) Zhang, W.; Wong, K. K.; Magliozzo, R. S.; Kozarich, J. W. Biochemistry 2001, 40, 4123. (12) Adrait, A.; Öhrström, M.; Barra, A. L.; Thelander, L.; Gräslund, A. Biochemistry 2002, 41, 6510. (13) van Gastel, M.; Lubitz, W.; Lassmann, G.; Neese, F. Journal of the American Chemical Society 2004, 126, 2237.

260 (14) Kolberg, M.; Bleifuss, G.; Gräslund, A.; Sjoberg, B.-M.; Lubitz, W.; Lendzian, F.; Lassmann, G. Arch Biochem Biophys 2002, 403, 141. (15) Gilbert, B. C. In Structure and Reaction Mechanisms in Sulphur-Radical Chemistry Revealed by E.S.R. Spectroscopy in Sulphur-Centered Reactive Intermediates in Chemistry and Biology; Springer: New York, USA, 1990; Vol. 197.

261 Appendix B

Chapter 3 Supporting Information

Preparation of Lc TmpA. The Lc gene CAER_RS0122985 encoding TmpA

(WP_027237574.1) was codon-optimized for over-expression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by

GenScript (Piscataway, NJ). Ec BL21 (DE3) competent cells (Invitrogen; Carlsbad, CA) were transformed with the TmpA plasmid and selected for kanamycin resistance.

Transformed cells were grown in rich Luria-Bertani medium (LB) with 50 mg/L kanamycin at 37 °C with shaking (250 rpm) until an OD600 of 0.6-0.8 was reached.

Protein expression was then induced by addition of 0.25 mM IPTG, and cell cultures were incubated at 18 °C with shaking (250 rpm) for 16-20 hours. Cultures were centrifuged at 8,000 × g for 15 min; cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cell pellets were resuspended in 50 mM Tris-HCl (pH 7.5) buffer containing

150 mM NaCl and 10% glycerol. The suspension was lysed by passage through a

Microfluidics M-110EH-30 microfluidizer processor at 20,000 psi for 10 min and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol.

The column was washed with 50 mM Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 30 mM imidazole and 10% glycerol. The protein was eluted by washing with 50 mM Tris-HCl

(pH 7.5) buffer, 150 mM NaCl, 250 mM imidazole and 10% glycerol. Fractions containing the protein were pooled and concentrated at 3,500 × g using a 30K MWCO

262 Macrosep® Advance Centrifugal Device (Pall Corporation, Port Washington, NY). The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-

HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol for > 6 h in the presence of 5 mM EDTA, pH 8 and 10 mM dithiothreitol (DTT) and twice for > 6 h without EDTA and DTT. Protein for crystallography was further purified by size exclusion chromatography using a Sephadex S-200 column (120 mL resin) equilibrated with 50 mM HEPES (pH 7.5) buffer with 150 mM NaCl. Fractions containing the protein were pooled and concentrated at 3,500 × g using a 30K MWCO Macrosep® Advance

Centrifugal Device. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with Coomassie staining, and protein concentration was determined by using a molar absorption coefficient of 49,500 M-1•cm-1 at 280 nm.

The iron content was determined by the ferrozine assay to be < 0.08 Fe/protein and considered to be “apo”.

Preparation of Lc TmpB. The Lc gene CAER_RS0122980 encoding TmpB

(WP_027237573.1) was codon-optimized for over-expression in Ec, synthesized, and inserted into the NdeI and XhoI restriction sites of expression vector pET-28a(+) by

GenScript. Ec BL21 (DE3) competent cells were transformed with the TmpB plasmid and selected for kanamycin resistance. Transformed cells were cultured at 37 °C with shaking (250 rpm) in M9 minimal medium with 50 mg/L kanamycin, 150 mg/L ampicillin, 0.2% (v/v) glucose, 0.10 mM CaCl2, 200 mM MgSO4·7H2O, and 0.125 mM

(NH4)2Fe(SO4)2·6H2O until an OD600 of 0.6-0.8 was reached. Protein expression was then induced by addition of 0.1 mM IPTG and cultures were also supplemented with 0.125 mM (NH4)2Fe(SO4)2·6H2O at this time. They were then incubated at 18 °C for 24 h with

263 shaking (250 rpm) and then were centrifuged at 8,000 × g for 15 min. Cell pellets were flash frozen in liquid N2 and stored at -80 °C. Cell pellets were resuspended in 50 mM

Tris-HCl (pH 7.5) buffer containing 150 mM NaCl and 10% glycerol. The suspension was lysed by passage through a Microfluidics M-110EH-30 microfluidizer processor at

20,000 psi for 10 min and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM

NaCl and 10% glycerol. The column was washed with 50 mM Tris-HCl (pH 7.5) buffer,

150 mM NaCl, 30 mM imidazole and 10% glycerol. The protein was eluted by washing with 50 mM Tris-HCl (pH 7.5) buffer, 150 mM NaCl, 250 mM imidazole and 10% glycerol. Fractions containing the protein were pooled and concentrated at 3,500 × g using a 10K MWCO Macrosep® Advance Centrifugal Device. The concentrated protein was then dialyzed against 100 equivalent volumes of 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl and 10% glycerol three times for > 6 h. Protein for crystallography was further purified by size exclusion chromatography using a 120 mL S-200 Sephadex column (GE) equilibrated with 50 mM HEPES (pH 7.5) buffer with 150 mM NaCl.

Fractions containing the protein were pooled and concentrated at 3,500 × g using a 10K

® MWCO Macrosep Advance Centrifugal Device. The protein was frozen in liquid N2 and stored at -80 °C. Protein purity was assessed by SDS-PAGE with Coomassie staining, and protein concentration was determined by using a molar absorption coefficient of

14,650 M-1•cm-1 at 280 nm. The iron content was determined to be 0.9-1.3 Fe/protein by ferrozine assay and was confirmed by inductively coupled plasma – atomic emission spectroscopy (ICP-AES), which detected less than 0.01 ppm of Mn, Zn and Cu. The

264 concentrations of TmpB quoted in all experiments represent the concentration of diiron cluster, calculated as half the measured iron concentration.

Preparation of PsBBOX. Ec BL21 (DE3) competent cells were transformed with the plasmid containing the PsBBOX gene in a pCOLDi vector and selected for ampicillin resistance. Transformed cells were grown in rich Luria-Bertani medium (LB) with 150 mg/L ampicillin at 37 °C with shaking (250 rpm) until an OD600 of 0.6-0.8 was reached.

150 mM NaCl and 10% glycerol. The suspension was lysed by passage through a French press cell twice at >12,000 psi and centrifuged at 22,000 × g for 20 min. The supernatant was loaded onto a Ni2+-NTA immobilized affinity chromatography column (~100 mL resin per 500 mL lysate) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM

NaCl and 10% glycerol. The column was washed with 50 mM Tris-HCl (pH 7.5) buffer,

Tris-HCl (pH 7.5) buffer with 150 mM NaCl. The protein was further purified by size exclusion chromatography using a Sephadex S-200 column (120 mL resin) equilibrated with 50 mM Tris-HCl (pH 7.5) buffer with 150 mM NaCl. Fractions containing the

265 protein were pooled and concentrated at 3,500 × g using a 30K MWCO Macrosep®

Advance Centrifugal Device. The protein was pelleted in liquid N2 and stored at -80 °C.

Protein purity was assessed by SDS-PAGE with Coomassie staining, and protein concentration was determined by using a molar absorption coefficient of 55,190 M-1•cm-1 at 280 nm. Iron content was determined to be < 0.02 Fe/protein by ferrozine assay and to be < 0.01 by inductively coupled plasma atomic emission spectroscopy (ICP-AES). Zinc content was determined to be 0.89 Zn/protein by ICP-AES.

266

Table S2. Data collection and refinement statistics for the X-ray structures1 of TmpA in the absence of substrate/product and in the substrate-bound and product-bound complex states. TmpA•Fe(II)• TmpA•Fe(II)• TmpA•Fe(II)•

2OG 2OG•TMP 2OG•TMO Data Collection Wavelength 0.97857 Å 0.97857 Å 0.97857 Å Space Group P41212 P41212 P41212 Cell Dimensions 87.051, 87.051, 86.901, 86.901, 87.564, 87.564, a, b, c (Å) 220.696 220.613 221.119 α, β, γ (°) 90, 90, 90 90, 90, 90 90, 90, 90 50.00-1.73 (1.76- 50.00-1.70 (1.73- 50.00-1.78 (1.81- Resolution (Å) 1.73) 1.70) 1.78) Rmerge 0.065 (0.871) 0.052 (0.633) 0.083 (0.467) < I/s > 26.4 (2.2) 34.9 (3.1) 25.4 (3.9) CC1/2 0.776 0.851 0.914 Completeness (%) 98.8 (99.9) 99.6 (100.0) 99.8 (99.7) Redundancy 8.2 (7.9) 8.1 (7.2) 8.3 (7.7)

Refinement Resolution (Å) 1.73 1.70 1.78 No. Reflections 83209 88905 79060 Rwork / Rfree 0.194/0.214 0.200/0.220 0.197/0.217 No. atoms Protein 5880 585 2 5890 Ion/Ligand 32 27 13 Water 464 521 483 B-Factors Protein 21.4 15.6 21.5 Ion/Ligand 34.4 11.2 13.9 Water 24.9 21.1 26.3 r.m.s. deviations Bond lengths (Å) 0.006 0.006 0.006 Bond angles (°) 1.17 1.20 1.20 Ramachandran Outliers (%) 2 (0.27) 0 (0.00) 1 (0.13) Molprobity Score (%-tile) 0.71 (100) 0.79 (100) 0.65 (100) 1A single crystal was used to solve each structure. *Highest resolution shell is shown in parenthesis

267

Table S3. Data collection and refinement statistics for the X-ray structure of TmpB. TmpB•Fe(II)•TMO TmpB•Fe(II)•TMO (7.2 keV) Data Collection Wavelength 0.97857 Å 1.72200 Å Space Group C2221 C2221 Cell Dimensions 70.110, 151.322, 80.760, 66.653, a, b, c (Å) 135.833 62.889 α, β, γ (°) 90, 90, 90 90, 90, 90 50.00-1.73 (1.76- 50.00-2.54 (2.58- Resolution (Å) 1.73) 2.54) Rmerge 0.067 (0.957) 0.056 (0.190) < I/s > 25.9 (2.1) 13.8 (6.7) CC1/2 0.874 0.967 Completeness (%) 99.9 (100.00) 78.3 (80.6) Redundancy 7.4 (7.3) 3.3 (3.2)

Refinement Resolution (Å) 1.72 No. Reflections 71844 Rwork / Rfree 0.224/0.251 No. atoms Protein 5882 Ion/Ligand 32 Water 261 B-Factors Protein 23.9 Ion/Ligand 20.8 Water 26.8 r.m.s. deviations Bond lengths (Å) 0.006 Bond angles (°) 1.07 Ramachandran Outliers (%) 1 (0.14) Molprobity Score (%-tile) 0.75 (100) *Highest resolution shell is shown in parenthesis.

268

Figure S1. (Top) SSN constructed from sequences with the IPR003819 domain (length > 350 amino acids) and sequences resulting from BLAST searches of the NCBI and IMG databases using LcTmpA as the sequence query. Nodes represent individual sequences or multiple sequences with > 90% identity. Edges between nodes represent pairwise alignment scores with an E-value < 1 × 10-50. The encircled cluster contains the nodes for Hs BBOX (blue), Hs TMLH (purple), Ps BBOX (green) and Lc TmpA (yellow) and thus was isolated for further analysis. (Bottom) Zoom-in view of the “BBOX-TMLH” cluster circled in the top panel.

269

Figure S2. (Left) “BBOX-TMLH” cluster upon increase of the E-value threshold for edges to < 1 × 10-65. At this threshold, the node corresponding to Hs TMLH (purple) separates into an isofunctional cluster. (Right) Zoom-in view of the “BBOX-only” cluster at the same threshold, showing sub-clustering based on domain [eukaryotic Hs BBOX (blue) vs. prokaryotic Ps BBOX (green)] and genomic context [Ps BBOX (green) vs. Lc TmpA (yellow)]. Green nodes represent sequences found within a γbb degradation operon, including PsBBOX. Yellow nodes represent sequences encoded adjacent to a gene encoding a TmpB-like HD-domain protein, including LcTmpA.

270

Figure S3. (Left) The “BBOX” cluster upon increase of the E-value threshold for edges to < 1 × 10-71. At this threshold, the eukaryotic clusters (blue) have become isolated from the prokaryotic clusters. (Right) The bacterial clusters only were selected for further analysis of genomic context. The putative isofunctional clusters containing the TmpA- like nodes (yellow, as described in Figure S2) and BBOX-like nodes (green, as described in Figure S2) have become more separated, with very few nodes (only 3) connecting the two major clusters. These clusters separate completely at a threshold E-value < 1 × 10-83, which roughly corresponds to sequence identity > 38% (main text, Figure 2).

Figure S4. Synthetic scheme to prepare TMAEP and DMAEP. Complete procedure is described in the Experimental Section.

271

Figure S5. NMR (1H, 13C, 31P) characterization of purified OH-TMAEP prepared by enzymatic synthesis from TMAEP with TmpA.

272

Figure S6. γbb analogues tested for activity with TmpA and PsBBOX. Succinate production was monitored by LC-MS from a reaction of 0.01 mM TmpA or Ps BBOX, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L-ascorbate, 1 mM 2OG and 1 mM analogue. Uncoupled succinate production in the absence of any substrate was determined to be 17 ± 2 turnovers (TON) after a 4 hour reaction. While PsBBOX is promiscuous with substrate analogues, as previously reported (1), TmpA succinate production is not stimulated by any of these substrates. Additionally, LC-MS analysis does not show any new peaks with m/z + 16 relative to those of the substrates that would correspond to hydroxylated products.

273

Figure S7. Assay testing activity of TmpA with phosphocholine (PC). (A) 31P-NMR spectra of a reaction of 0.01 mM TmpA, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L-ascorbate, 3 mM 2OG and 2 mM PC after a 4-h incubation with air-flushing. (PC = -0.5 ppm; phosphate = 0 ppm) (B) LC-MS chromatograms of the same samples as in panel A, monitoring PC (184 m/z) and glycine betaine aldehyde (102 m/z, red).

Figure S8. Reaction requirements for TMAEP hydroxylation by TmpA. (A) 31P-NMR spectra of a complete reaction of 0.01 mM TmpA, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L- ascorbate, 3 mM 2OG and 2 mM TMAEP after a 4-h incubation with air-flushing and parallel reactions omitting 2OG or TmpA, or performed anaerobically. (TMAEP = 16.8 ppm; OH-TMAEP = 14.0 ppm) (B) LC-MS single-ion-monitoring (184 m/z) chromatograms corresponding to the same samples as in panel A, monitoring OH- TMAEP product formation. (C) Determination of the “coupling ratio”, defined as the ratio of TMAEP consumed to succinate produced in reactions with limiting 2OG; the slope of the plot is 1.0 ± 0.1.

274

Figure S9. Phosphonate analogues tested for activity with TmpA. Succinate production was monitored by LC-MS from a reaction of 0.01 mM TmpA, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM L-ascorbate, 1 mM 2OG and 1 mM analogue. Uncoupled succinate production in the absence of any substrate was determined to be 17 ± 2 turnovers (TON) after a 4 hour reaction. Succinate production by TmpA does not appear to be stimulated in the presence of any analogues lacking an amino functionality. Additionally, LC-MS analysis does not show any new peaks with m/z + 16 relative to those of the substrates that would correspond to hydroxylation products.

275

Figure S10. SF-Abs experiment monitoring reaction of TmpA with O2 in the absence of substrate. (A) Absorption spectra at select time points upon mixing a solution of 2 mM TmpA, 1.9 mM (NH4)2Fe(SO4)2 and 10 mM 2OG with an equal volume of O2-saturated 50 mM HEPES buffer, pH 7.5 (1.8 mM O2 at 5 °C). (C) Time-dependent traces from the same reaction, monitoring 318 nm (blue) and 519 nm (red).

276

Figure S11. Absorption spectra determining substrate binding to TmpA and triggering of O2 activation. (A and B) Absorption spectra of the ternary complex, TmpA•Fe(II)•2OG, in the absence of substrate (black) and in the presence of various substrate analogues. Time-dependent traces of 318 nm (C and D) or 519 nm (E and F) upon mixing of a solution of 2 mM TmpA, 1.9 mM (NH4)2Fe(SO4)2, 10 mM 2OG and 10 mM substrate analogue with an equal volume of O2-saturated 50 mM HEPES buffer, pH 7.5 (1.8 mM O2 at 5 °C).

277

Figure S12. LC-MS chromatograms of samples after a 4 h reaction of 0.01 mM TmpA, 0.02 mM (NH4)2Fe(SO4)2, 0.2 mM ascorbate, 1 mM 2OG, and 1 mM substrate analogue. TmpA was able to hydroxylate DMAEP and 2-AEP as described in the main text and the non-natural product 3-aminopropylphosphonate (3-APP), achieving 15 turnovers after the 4 h reaction. No substrate consumption, nor hydroxylated product (M+16 m/z), was observed for the remaining analogues.

278

Figure S13. (Top) Sequence alignment for the N-terminal domain regions of Lc TmpA (numbering), Hs BBOX and Ps BBOX with the Zn(II)-binding motif in BBOX sequences highlighted in red. (Bottom) Overlay of the TmpA (dark green) and HsBBOX (grey) N- terminal domains. The Zn(II) metal is shown as a blue sphere and the coordinating residues in the HsBBOX structure, Cys38, Cys40, Cys43 and His82, are shown as sticks. The C-terminal domain of the opposite, interfacing TmpA monomer is shown in light green with the Fe(II) metal as a brown sphere and the TMAEP substrate in stick format.

279

Figure S14. Active site models for the TmpA structures. For all panels, the Fe(II) is shown as a sphere and the coordinating residues His198, Asp200, and His344 are shown as green sticks. The Fo-Fc difference (green and red mesh) and omit maps (blue mesh) are shown at ± 3.0σ and the 2Fo-Fc map (gray mesh) is shown at 1.5σ. (A) Chain A of the TmpA ternary complex structure. The Fo-Fc omit map is shown for Fe(II), 2OG, a water ligand, and a sulfate molecule. (B) Chain A of the TmpA quaternary complex structure. The Fo-Fc omit map is shown for Fe(II), 2OG, and TMAEP (yellow). (C) Chain A of the TmpA product-bound complex structure. The Fo-Fc omit map is shown for Fe(II), OH- TMAEP (yellow), and the Fe(II) equatorial non-proteinatious ligands. Three models were considered to account for the density for the non-proteinatious ligands: (G) two coordinating water molecules and an acetate molecule resulting from the TMAEP synthesis procedure, (H) 2OG from the crystallization condition, or (I) succinate resulting from turnover in crystallo. None of these options alone accurately model the observed density, suggesting that there is a mixture of states, and thus, this density was left unmodeled in the final structure. The negative density for 2OG suggests that 2OG likely does not contribute substantially to this mixture, consistent with the observation of substrate hydroxylation. (D) Chain B of the TmpA ternary complex structure. The Fo-Fc omit is shown for Fe(II), water ligands, and a sulfate molecule. The Fe(II) occupancy is estimated to be < 80 %. (E) Chain B of the TmpA quaternary complex structure. The Fo- Fc omit map is shown for Fe(II), water ligands, and a sulfate molecule. The Fe(II) occupancy is estimated to be < 65 %. (F) Chain B of the TmpA product-bound complex structure. The Fo-Fc omit map is shown for Fe(II), water ligands, and the partially occupied substrate or product mixture. The Fe(II) occupancy is estimated to be < 40 %.

280

Figure S15. (Top) Overlay of molecules A (light green) and B (dark green) in the asymmetric unit of the TmpA crystal soaked with TMAEP. In molecule B, the residues 189-196 constituting the lid-loop region are disordered and not modeled (indicated by a dashed line), leaving the active site exposed. Whereas, in molecule A, TMAEP is bound (yellow sticks), the core contracts, and the lid-loop becomes ordered to close over the active site. (Bottom) Depiction of the TmpA molecule B (dark green cartoon) and its symmetry mate in the asymmetric unit, which occludes the active site.

281

Figure S16. Substrate contacts between the phosphonate moiety of TMAEP (yellow) and the TmpA protein (green). Residues from the C-terminal domain of monomer A are depicted in light green and the residues from the N-terminal domain of monomber B are shown in dark green. The Fe(II) metal is shown as a brown sphere.

282

Figure S17. Pairwise sequence alignment of PhnZ from uncultured bacterium HF130_AEPn_1 and LcTmpB with the conserved histidine and aspartate residues of the extended HD-domain sequence motif highlighted in red.

283

Figure S18. Spectroscopic evidence for adventitiously-bound Fe in TmpB. (A) 4.2 K/53 mT Mössbauer spectrum of 2 mM O2-free TmpB in the Fe2(III/III) state prepared by incubation with 3 mM potassium ferricyanide for 45 min in an anaerobic chamber. (Inset) X-band EPR spectrum of a sample of as-isolated, oxidized TmpB (0.25 mM). Experimental conditions: temperature = 10 K, microwave power = 0.2 mW, microwave frequency = 9.479 GHz, modulation amplitude = 1 mT. (C) X-ray crystal structure of TmpB, showing the interface of monomers C and D in the asymmetric unit. Orange mesh depicts the anomalous Fourier density contoured at 5σ. The occupancy of these metals was adjusted to xx in the final model. Their observation in the crystal structure supports the conclusion that the high-spin Fe(III) signals observed in the Mössbauer and EPR spectra are due to adventitiously-bound iron, not mononuclear Fe(III) in the active site.

284

A 0.5%

B 0.5%

C 0.2%

-10 -8 -6 -4 -2 0 2 4 6 8 10 Velocity (mm/s) Figure S19. Mössbauer spectra of a sample prepared to contain significant amount of TmpB in the Fe2(II/III) state. The sample was generated by incubation of a solution of 2 mM O2-free TmpB in the Fe2(III/III) state with 20 mM sodium L-ascorbate for 45 min under anaerobic conditions. The 4.2K/53mT spectrum with the external magnetic field oriented parallel to the γ beam is shown as black vertical bars in left panel, A. The spectral contributions from Fe2(II/II) (22% of total intensity) and Fe2(III/III) (13% of total intensity) are shown as turquoise and orange lines, respectively. Removal of these contributions yields the experimental reference spectrum of TmpB Fe2(II/III) state in a magnetic field of 53 mT applied parallel (left panel, B, black vertical bars). The corresponding reference spectrum of the Fe2(II/III) cluster with the external field oriented perpendicular to the γ beam (left panel, B, blue line) was prepared in an identical fashion. The resulting parallel-minus-perpendicular difference spectrum (left panel, C) of TmpB Fe2(II/III) state exhibits the field-orientation dependence expected for a cluster with S = 1/2 ground state. The 120-K/zero-field spectrum of this sample (right panel, black bars) overlaid with 26% of the experimental Fe2(II/II) spectrum collected under identical conditions (orange line), which was then subtracted to give the resultant spectrum in the main text Figure 8.

285

Figure S20. Symmetry mate (gray mesh surface representation with cartoon) occupies the active site of chain B (purple) in the TmpB structure, preventing substrate binding. Iron ions are shown as brown spheres and ligands are shown in stick format.

Figure S21. Interactions of (R)-OH-TMAEP (yellow) in the active site of TmpB. Iron ions are shown as brown spheres. Dashed lines represent hydrogen bonding or electrostatic interactions.

286

Figure S22. (Left) PhnZ structure (PDB accession code 4MLN), highlighting the lid loop (dark blue) that seals off the active site from solvent exposure in the presence of substrate (sticks). (Right) TmpB structure showing the unmodeled region (dashed line) of the analogous lid loop (dark purple) and the contacts between the loop with the symmetry mate (gray mesh surface depiction) that prevent its motion.

Figure S23. Coupled reactions of TmpA and TmpB with (A) TMAEP or (B) DMAEP. Reactions were incubated for 4 h at 3 °C and contained 0.02 mM TmpA, 0.03 mM (NH4)2Fe(SO4)2, 0.4 mM L-ascorbate, 6 mM 2OG, and either 2 mM TMAEP or DMAEP. Substrates (black), hydroxylated intermediate products (blue) and glycine betaine (A, red) or dimethyl glycine (B, red) were monitored by LC-MS.

1. Rydzik, A. M., Leung, I. K. H., Kochan, G. T., Loik, N. D., Henry, L., McDonough, M. A., Claridge, T. D. W., and Schofield, C. J. (2014) Comparison of the substrate selectivity and biochemical properties of human and bacterial gamma-butyrobetaine hydroxylase. Organic & Biomolecular Chemistry 12, 6354-6358.

287

VITA Lauren J. Rajakovich

EDUCATION Pennsylvania State University, University Park, PA 2012-2017 Ph.D. candidate, Department of Biochemistry, Microbiology and Molecular Biology; December 2017 Dissertation Title: Exploring the functional and mechanistic diversity of diiron oxidases and oxygenases Advisors: Profs. J. Martin Bollinger, Jr. and Carsten Krebs Wake Forest University, Winston-Salem, NC 2007-2011 Bachelor of Science cum laude, Department of Chemistry Concentration in Biochemistry; Minors in Music & Biology Honors Thesis Title: Biosynthesis of 4-thiouridine in Bacillus subtilis Advisor: Prof. Patricia Dos Santos

HONORS & AWARDS Poster Award – Enzymes, Coenzymes & Metabolic Pathways Gordon Research Conference 2016 Richard L. and Norma L. McCarl Graduate Scholarship Endowment in Biochemistry, Microbiology and Molecular Biology – Pennsylvania State University 2015 Paul M. Althouse Memorial Outstanding Teaching Assistant Award – Pennsylvania State University 2014 Homer F. Braddock, Nellie H. and Oscar L. Roberts Fellowship – Pennsylvania State University 2012 Honors in Chemistry – Wake Forest University 2011 The John W. Nowell Award in Undergraduate Chemistry – Wake Forest University 2011

PROFESSIONAL EXPERIENCE Elected Chair, Bioinorganic Chemistry Gordon Research Seminar 2017 Elected Vice Chair, Bioinorganic Chemistry Gordon Research Seminar 2016 Workshop Instructor, Bioinorganic Workshop, Pennsylvania State University 2014 & 2016 Teaching Assistant, Pennsylvania State University 2012 Research Technician, Wake Forest University 2011-2012 Chemistry Tutor, Wake Forest University 2009-2011 President of Women in Science, Wake Forest University 2007-2011

PUBLICATIONS (1) Peck SC, Wang C, Dassama LM, Zhang B, Guo Y, Rajakovich LJ, Bollinger JM Jr., Krebs C, van der Donk WA. O-H Activation by and Unexpected Ferryl Intermediate during Catalysis by 2- Hydroxyethylphosphonate Dioxygenase. J. Am. Chem. Soc. 2017, doi: 10.1021/jacs.6b12147. (2) Rajakovich, L.J., Nørgaard, H., Warui, D.M., Chang, W.-c., Li, N., Booker, S.J., Krebs, C., Bollinger, J.M., Jr., Pandelia, M.-E. Rapid reduction of the diferric-peroxyhemiacetal intermediate of aldehyde- deformylating oxygenase by a cyanobacterial ferredoxin: evidence for a free-radical mechanism. J. Am. Chem. Soc. 2015, 137(36): 11695-709. (3) Warui, D.M., Pandelia, M.-E., Rajakovich, L.J., Krebs, C., Bollinger, J.M., Jr., Booker, S.J. Efficient delivery of long-chain fatty aldehydes from the Nostoc punctiforme acyl-acyl carrier protein reductase to its cognate aldehyde deformylating oxygenase. Biochemistry 2015, 54(4): 1006-15. (4) Pandelia, M.-E., Li, N., Nørgaard, H., Warui, D.M., Rajakovich, L.J., Chang, W.-c., Booker, S.J., Krebs, C., Bollinger, J.M., Jr. Substrate-triggered addition of dioxygen to the diferrous cofactor of aldehyde-deformylating oxygenase to form a diferric-peroxide intermediate. J. Am. Chem. Soc. 2013, 135(42): 15801-12. (5) Rajakovich, L.J., Tomlinson, J., Dos Santos, P.C. Functional analysis of Bacillus subtilis genes involved in the biosynthesis of 4-thiouridine in tRNA. J. Bacteriol. 2012, 194(18): 4933-40.