The Birth of the Hydrogenase

Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1856

The birth of the hydrogenase

Studying the mechanism of [FeFe] hydrogenase maturation

BRIGITTA NÉMETH

ACTA UNIVERSITATIS UPSALIENSIS ISSN 1651-6214 ISBN 978-91-513-0754-1 UPPSALA urn:nbn:se:uu:diva-393261 2019 Dissertation presented at Uppsala University to be publicly examined in Å4101, Ångströmlaboratoriet, Lägerhyddsvägen 1, Uppsala, Wednesday, 6 November 2019 at 09:15 for the degree of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner: Doctor Mohamed Atta (Laboratory of Chemistry and Biology of Metals, iRTSV CEA Grenoble).

Abstract Németh, B. 2019. The birth of the hydrogenase. Studying the mechanism of [FeFe] hydrogenase maturation. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1856. 78 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-0754-1.

The [FeFe] hydrogenases are ancient metalloenzymes that catalyse the reversible interconversion between protons, electrons and molecular hydrogen. Despite the large structural variability within the [FeFe] hydrogenase family, the active site, the so called “H-cluster” is present in every representative. The H-cluster is composed by a four cysteine coordinated [4Fe4S] cluster, ligated via a shared cysteine to a biologically unique [2Fe] subsite decorated with CO and CN ligands and an azadithiolate bridging ligand. The biosynthesis of the [2Fe] subsite requires a maturation machinery, composed of at least three maturase enzymes, denoted HydG, HydE, and HydF. HydE and HydG are members of the radical SAM enzyme family, and are responsible for the construction of a pre-catalyst on HydF. This pre-catalyst is finally transferred from HydF to HydA, where it becomes part of the H-cluster. Recently, a pioneer study combined synthetic chemistry and biochemistry in order to create semi-synthetic HydF proteins. Synthetic mimics of the [2Fe] subsite were introduced to HydF, and this resulting semi-synthetic HydF was used to activate the unmatured hydrogenase (apo- HydA). This technique ushered in a new era in [FeFe] hydrogenase research. This thesis work is devoted to a deeper understanding of H-cluster formation and [FeFe] hydrogenase maturation, and this process is studied using standard molecular biological and biochemical techniques, and EPR, FTIR, XAS and GEMMA spectroscopic techniques combined with this new type of chemistry mentioned above. EPR spectroscopy was employed to verify the construction of a semi-synthetic [FeFe] hydrogenase inside living cells. The addition of a synthetic complex to cell cultures expressing apo-HydA resulted in a rhombic EPR signal, attributable to an Hox-like species. Moreover, the assembly mechanism of the H-cluster was probed in vitro using XAS, EPR, and FTIR spectroscopy. We verified with all three techniques that the Hox-CO state is formed on a time-scale of seconds, and this state slowly turns into the catalytically active Hox via release of a CO ligand. Furthermore, a semi-synthetic form of the HydF protein from Clostridium acetobutylicum was prepared and characterized in order to prove that such semi-synthetic forms of HydF are biologically relevant. Finally,GEMMA measurements were performed to elucidate the quaternary structure of the HydF-HydA interaction, revealing that dimeric HydF is interacting with a monomeric HydA. However, mutant HydF proteins were prepared, lacking the dimerization (as well as its GTPase) domain, and these severely truncated forms of HydF was found to still retain the capacity to both harbor the pre-catalyst as well as transferring it to apo-HydA. These observations highlight the multi-functionality of HydF, where different domains are critical in different steps of the maturation, that is the dimerization and GTPase domain are rather involved in pre-catalyst assembly rather than its transfer to apo-HydA.

Keywords: HydF, HydA, [FeFe] hydrogenase maturation, semi-synthetic enzymes

Brigitta Németh, Department of Chemistry - Ångström, Box 523, Uppsala University, SE-75120 Uppsala, Sweden.

ISSN 1651-6214 ISBN 978-91-513-0754-1 urn:nbn:se:uu:diva-393261 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-393261)

Man muss sich entscheiden was man tut oder lässt Welche Maske man trägt und wohin man gehört Man kann fliehn oder leiden nur eines steht fest: Der einfache weg ist immer verkehrt.

Ki nagyvízre indul, annak dönteni kell: melyik hullám a jó, melyik sodorja el. Legyen bármilyen vonzó egy új állomás. Az egyszerű út, az naivitás.

Michael Kunze and Müller Péter Sziámi

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I. Mészáros L.S., Németh B., Esmieu C., Ceccaldi P., Berggren G. In Vivo EPR Characterization of Semi-Synthetic [FeFe] Hydrogenases. Angewandte Chemie International Edition, 2018. 57(10): p. 2596- 2599. II. Németh, B., Senger, M., Redman, H.J., Stripp, S. T., Haumann, M., Berggren G. H-cluster assembly in [FeFe]-hydrogenase tracked by electron paramagnetic, infrared, and X-ray absorption spectroscopy (Manuscript) III. Németh, B., Esmieu C., Redman H.J., Berggren G. Monitoring H- cluster assembly using a semi-synthetic HydF protein. Dalton Trans, 2019. 48(18): p. 5978-5986. IV. Németh, B., Land, H., Magnuson, A., Hofer A., Berggren G. Study- ing [FeFe] hydrogenase maturation and the nature of the HydF- HydA interaction (Manuscript)

Reprints were made with permission from the respective publishers.

Author contributions

Paper I – Performed the preparation of the purified protein samples (large scale heterologous expression, purification, reconstitution, maturation, sample characterization, biochemical assays) and low temperature CW-EPR measurement of the [2Fe]adt loaded form. Active participation in the data evaluation. Contributed to the writing.

Paper II - Performed the preparation of the protein samples (large scale heterologous expression, purification, reconstitution, maturation, sample characterization, biochemical assays, XAS sample preparation and EPR sample preparation) and low temperature CW-EPR measurement of the [2Fe]adt loaded form. Data evaluation, wrote the first draft of the manuscript, contributed to the further writing.

Paper III – Performed the preparation of the protein samples (large scale heterologous expression, purification, reconstitution, maturation, sample characterization, biochemical assays, FTIR sample preparation, EPR sample preparation) low temperature CW-EPR measurements, FTIR measurements, UV- Vis measurements, hemoglobin assays, hydrogen evolution assays, data evaluation, wrote the first draft of the manuscript, contributed to the later versions.

Paper IV - Performed the molecular biology (mutagenesis, cloning, test expression) and the protein sample preparation (large scale heterologous expression, purification, reconstitution, maturation, sample characterization, biochemical assays, FTIR sample preparation, EPR sample preparation, GEMMA sample preparation) low temperature CW-EPR measurements, FTIR measurements, UV-Vis measurements, GEMMA measurements, hydrogen evolution assays, data evaluation, wrote the first draft of the manuscript, contributed to the later versions.

Contents

1. Introduction ...... 11 1.1 [FeFe] hydrogenases ...... 11 1.2. The catalytic cycle of the [FeFe] hydrogenases ...... 13 1.3. The [FeFe] hydrogenase maturation ...... 15 1.3.1. HydG ...... 16 1.3.2. HydE ...... 17 1.3.3. HydF ...... 18 1.4. The artificial maturation ...... 20 1.4.1. Artificially matured HydA proteins ...... 20 1.4.2. Artificially matured HydF proteins ...... 21 2. Aims ...... 22 3. In vivo activation of the [FeFe] hydrogenase with biomimetic complexes (Paper I) ...... 23 3.1. Motivation and strategy ...... 23 3.2. Sample preparation ...... 24 3.3. In vivo activation of the HydA1 ...... 24 3.3.1. The identification of the EPR signal ...... 24 3.3.2. The time dependence of the activation ...... 25 3.3.2. The power saturation behavior in vivo ...... 27 3.5 Summary and Discussion ...... 29 4. Monitoring the H-cluster formation using a synthetic binuclear iron complex (Paper II) ...... 30 4.1. Motivation and strategy ...... 30 4.2. Sample preparation ...... 31 4.3. X-ray absorption spectroscopy at the Fe K-edge ...... 33 4.3.1. The Fe XANES data analysis ...... 33 4.3.1. The Fe EXAFS data analysis ...... 35 4.4. EPR spectroscopy ...... 37 4.4. FTIR spectroscopy ...... 39 4.5 Summary and discussion ...... 41 5. Monitoring the H-cluster formation using a semi-synthetic HydF protein (Paper III) ...... 42 5.1. Motivation and strategy ...... 42

5.2. Sample preparation ...... 43 5.3. Basic spectroscopic characterization ...... 44 5.4. CO release ...... 46 5.5. Redox chemistry ...... 48 5.6. Summary and discussion ...... 50 6. Quaternary structural study of the [FeFe] hydrogenase maturation (Paper IV) ...... 52 6.1. Motivation and strategy ...... 52 6.2. The quaternary structure of HydF ...... 52 6.2. The GEMMA vs. SEC ...... 54 6.3. Spectroscopic characterization ...... 54 6.4 The effect of the metal content on the quaternary structure ...... 56 6.5. The quaternary structure of the HydF-HydA interaction ...... 57 6.6. Summary and discussion ...... 58 7. The miniaturized HydF proteins (Paper IV) ...... 59 7.1. Motivation and strategy ...... 59 7.2. Sample preparation ...... 59 7.3. The spectroscopic characterization ...... 60 7.4. HydA activation by the miniaturized HydF proteins ...... 63 7.5 Summary and discussion ...... 64 Conclusions and outlook ...... 66 Sammanfattning på svenska ...... 68 Acknowledgements ...... 70 References ...... 74

Abbreviations

adt 2− [2Fe] [Fe2(adt)(CO)4(CN)2] pdt 2− [2Fe] [Fe2(pdt)(CO)4(CN)2] [4Fe4S] Four iron four sulfur cluster

adt Azadithiolate, -SCH2NHCH2S- apo-HydA1 Unmatured HydA with only the [4Fe4S] cluster apo-HydF Unmatured HydF with only the [4Fe4S] cluster ATR-FTIR Attenuated Total Reflection Fourier Transform Infrared Spectroscopy CaHydF HydF from Clostridium acetobutylicum CN Cyanide CO Carbonmonoxide CpI HydA from Clostridium pasteruianum HydA1 HydA from Chlamydomonas reinhardtii CW-EPR Continuous Wave Electron Paramagnetic Resonance E. coli Escherichia coli EDTA ethylenediaminetetraacetic acid, common metal chelator EPR Electron Paramagnetic Resonance EXAFS Extended X-Ray Absorption Fine Structure FTIR Fourier Transform Infrared Spectroscopy GDP Guanosine diphosphate GEMMA Gas-phase Electrophoretic Molecular Mobility Analysis GTP Guanosine triphosphate HEPES 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid holo-HydA Matured, H-cluster carrying hydrogenase holo-HydF Matured, pre-catalyst carrying HydF HydA [FeFe] hydrogenase HydE Radical SAM enzyme, part of the maturation machinery HydF Scaffold protein, part of the maturation machinery HydG Radical SAM enzyme, part of the maturation machinery LB Luria-Bertani medium M9 Minimal medium MBP Maltose Binding Protein

MeHydA HydA from Megasphera elsdenii pdt Propanedithiolate, -SCH2CH2CH2S- SAM S-adenosyl-L-methionine SEC Size Exclusion Chromatography TmHydF HydF from Thermotoga maritima TRIS 2-Amino-2-(hydroxymethyl)propane-1,3-diol UV-Vis Ultraviolet–visible spectroscopy XANES X-ray Absorption near edge structure XAS X-ray Absorption Spectroscopy ΔD TmHydF mutant, lacking in the dimerization domain ΔGD TmHydF mutant, lacking in the dimerization and GTPase domain, fused N-terminally with MBP

1. Introduction

Enzymes are fascinating creatures. Enzymes are the so called “biocatalysts”, the catalysts of Nature. Enzymes are not able to change the thermodynamic nature of the reactions but they are able to accelerate reactions. It means that enzymes cannot turn a thermodynamically uphill reaction into a downhill reaction but they can make things happen faster which would not happen spon- taneously. For instance, at room temperature a stressed PhD student is rather playing with video games than writing her thesis. The latter process has a higher activation barrier. In this case the supervisor acts as a catalyst and ac- celerates the reaction towards the writing direction. Understanding how an enzyme works is crucial in order to be able to take full advantage of their incredible abilities. Why do we care? Because the products of the catalysis are important for us and it is difficult to make them without catalysts, no matter if the product of interest is fixed CO2 from the atmosphere as isopropanol produced by living cells, a doctoral thesis produced by a PhD student and a supervisor on the topic of biofuels or hydrogen produced by a molecular proton reduction catalyst - all these three are goals and results of artificial photosynthesis research in academia. These reactions always require reactants (or what the biochemist calls “substrates”), which the catalyst can turn into product. To the naked eye, proton reduction is a trivial reaction, since the reactants are “just” protons and electrons. While the proton reduction reaction seems simple on paper, it requires a well-designed catalyst. This thesis is dedicated to con- tribute to a deeper understanding of the construction of the best molecular proton reduction catalyst in Nature. This construction process is referred to as the maturation of [FeFe] hydrogenases.

1.1 [FeFe] hydrogenases Nature has developed one of the best molecular proton reduction catalysts, the [FeFe] hydrogenases (HydAs). These enzymes are able to catalyze the reversible interconversion of protons and electrons to molecular hydrogen with high turnover numbers and turnover frequencies, and they work at ambient temperatures, neutral pH and atmospheric pressure, which makes them excellent candidates for solar fuel production.[1] The [FeFe] hydrogenase family has a large diversity with regards to function as well as structure, that is number of subunits, domains and cofactors. [2] What is common in every family member

11 is the so called “H-domain”, which contains the active site. The simplest monomeric [FeFe] hydrogenases – as exemplified by the HydA1 enzyme from the green algae Chlamydomonas reinhardtii (HydA1) – feature only this domain. The H-domain has a “channel” like structure, which is most likely involved in the maturation process. [3] This channel appears to be open in the crystal structure of the unmatured, apo-enzyme and it appears to be closed in the crystal structure of the matured, holo-enzyme. [4] [5] This channel leads directly to the active site of the [FeFe] hydrogenase, the H-cluster (Figure 1). The H- cluster is composed by a canonical 4 cysteine coordinated [4Fe4S] cluster coupled to a unique [2Fe] subsite decorated with carbonyl and cyanide ligands and an azadithiolate bridging ligand. Furthermore, one of the cysteines also provides a bridging thiolate ligand, linking the [4Fe4S] cluster and the [2Fe] subsite together. [5-7]

Figure 1. – Left: [FeFe] hydrogenase from Clostridium pasteurianum featuring the H-domain as well as the auxiliary FeS cluster containing F domain; Right: the H- cluster. Color coding: Gold: sulfur, brown: iron, blue: nitrogen, grey: carbon, red: oxygen (PBD: 4XDC) [8]

It is on this, in biology completely unique, [2Fe] subsite that H2 bond formation and bond breaking happens. Many people tend to fall into the illusion of Nature and assume that as the catalysis occurs only on the [2Fe] subsite, the rest of the protein and the rest of the H-cluster is not involved. The beauty of the smart design of Nature shows up when we have a closer look at the functions required for the proton reduction reaction. In a nutshell, to design a good proton reduction catalyst, we need to think about gas channels, proton relays and electron relays. If we look closer, above the [2Fe] subsite we can see that in the hydrogenase enzyme the [4Fe4S] cluster of the H-cluster functions as an electron relay. In most hydrogenases this electron relay is even more extended, involving numerous FeS clusters, which are not part of the H- cluster. The amino acids of the protein backbone as well as the nitrogen bridgehead of the [2Fe] subsite are acting as proton relays. The gas channels are also provided by the protein backbone, ensuring a good flow of hydrogen

12 gas, and providing a barrier for potentially damaging gases like carbon-monoxide or oxygen. [9, 10]

1.2. The catalytic cycle of the [FeFe] hydrogenases The catalytic cycle of the [FeFe] hydrogenases is a puzzle. As a matter of fact, there are several models and ideas about the catalytic cycle of the [FeFe] hydrogenases. These models are not necessary excluding each other, but cer- tainly not all of them are in complete agreement about what the catalytically relevant species are. These species are specific redox states and catalytic intermediates of the H-cluster. Each and every intermediate state can be expected to have its own spectroscopic fingerprint. However, certain redox states can be short lived or difficult to populate, so there is a high possibility that there are still unknown, catalytically relevant redox states. This thesis will only discuss the redox states, that are the most relevant to know in order to understand this thesis, based mainly on the so called “Mülheim model” of the [FeFe] hydrogenase mechanism, proposed by Lubitz and co-workers (Figure 2). However, it is very important to emphasize that other mechanistic models also exist, like the “Berlin model”, proposed by Stripp, Haumann and co- workers. [11]

The resting state of the [FeFe] hydrogenase is Hox, where the oxidation state of the [4Fe4S] cluster is 2+ and the [2Fe] subsite’s oxidation state is FeIFeII. Since the [4Fe4S] cluster is diamagnetic and the [2Fe] subsite is paramagnetic, this species gives rise to a rhombic S=1/2 EPR signal. One electron reduction leads one step forward to the Hred (sometimes called Hred’) state. In this state, both the [4Fe4S] cluster and the [2Fe] subsite are paramagnetic. Since the [4Fe4S] cluster and the [2Fe] subsite are strongly coupled, the two paramagnetic centers cancel each other, leading to a S=0 spin state, which results in loss of the EPR signal. The protonation of the bridgehead creates an electron hole on the proximal Fe of the [2Fe] subsite – proximal as closer to the [4Fe4S] cluster of the H-cluster – where the electron is tunneling from the [4Fe4S] + cluster. The resulting state is called HredH .

Figure 2. The author’s sketch of the catalytic cycle of the [FeFe] hydrogenases adapted from [12]. The diamond is representing the [4Fe4S] cluster of the H-cluster, and the numbers are referring to its oxidation state. The square bracket symbolizes the open coordination site of the [2Fe] subsite. The states are following the order: + + + Hox, Hred, HredH , HsredH , Hhyd, HhydH , Hox(H2).

In this case the oxidation state of the [4Fe4S] cluster is 2+, which is diamagnetic and the subsite’s oxidation state is FeIFeI, which is also diamagnetic. Thus, species is also EPR silent. Further reduction by one more electron results in the “super reduced” (Hsred) state. The [4Fe4S] cluster with oxidation state 1+ gives a paramagnetic center, while the oxidation state of the [2Fe] subsite remains as FeIFeI. Therefore, this redox state is also EPR active, and it gives rise to a S=1/2 rhombic EPR signal. From a practical point of view, these are the redox species that can be easily populated. However, very recently several research groups were able to populate and capture less stable redox species like the hydride state, where the [4Fe4S] cluster is reduced, the oxidation state of the [2Fe] subsite is FeIIFeII and an terminal hydride can be observed. [13] Furthermore, there are catalytically inactive redox species, like the CO inhibited states (Hox-CO, Hred-CO) where the redox states are identical to their non-inhibited forms, but they coordinate a fourth CO ligand. Hence,

14 Hox-CO is paramagnetic, but in contrast to the rhombic EPR signal of Hox, the Hox-CO state gives rise to an axial EPR signal.

1.3. The [FeFe] hydrogenase maturation If we want to understand the [FeFe] hydrogenase construction process, we need to think in precursors, reactants and building blocks that can be found or made in the organism or in its surroundings. The only solvent of the living cells is water, not acetonitrile or hexane. Thus the intermediates of the catalyst synthesis process has to be stable in water. To construct a functional hydrogenase, first of all we need a protein backbone. The protein backbone and the [4Fe4S] cluster are household building blocks of the hydrogenase enzyme; every living cell is able to perform the synthesis of these. The protein backbone will protect the catalytic cofactor and it will provide gas channels and proton channels as well as binding sites for electron relays to the active site. Furthermore, four conserved cysteines - which are located close to the active site - will coordinate the [4Fe4S] cluster. The [4Fe4S] cluster synthesis requires low valent iron, which is also not too challenging since iron is an earth abundant metal and the cytoplasmic environment of the bacterial cell is relatively reducing. The source of the sulfur in the [4Fe4S] cluster is acid labile sulfide which can be produced by the cleavage of cysteine to alanine and sulfide by the Isc enzyme family, but other FeS cluster biosynthesis enzymes have been identified as well, like the Nif and Suf enzymes families. [14] Since the FeS clusters are common cofactors, the rather unusual part of the H-cluster synthesis is of course the [2Fe] subsite. In living cells, a whole maturation machinery is dedicated to the synthesis of this cofactor. This machinery consists of at least three maturase enzymes, HydG, HydE and HydF (Figure 3). [15]

Figure 3. The [FeFe] hydrogenase maturation scheme based on [1, 8, 16-19]. The protein structures are corresponding to HydG (yellow, PBD: 4WCX), HydE (blue, PBD: 3CIW), HydF (grey, PBD: 5KH0), HydA (pink, PBD: 5LA3). The black lines represent the sequence of the maturation (discussed in detail in later subchapters), the red dashed lines indicate the metal-cofactors highlighted by the magnifying glasses. The bubbles represent hydrogen gas produced by the activated [FeFe] hydrogenase.

1.3.1. HydG HydG belongs to the radical SAM enzyme super family. [20] Radical SAM enzymes are known for their three cysteine coordinated [4Fe4S] cluster that plays a key role in the generation of the 5′-deoxyadenosyl radical. In general, the resulting radical can be used to catalyze a diverse set of reactions. In the case of HydG, this radical will be used later on for the generation of CO and CN from tyrosine (Figure 4). [19] The CO and CN production are catalytically different events and the locations of the reactions are also different. [21, 22] The final product of HydG is a mononuclear iron complex with one CN and two CO ligands, coordinated by a cysteine. [17, 22] [23, 24] [25]

Figure 4. The maturase enzymes: HydG [17] (yellow, PBD: 4WCX), HydE [16] (blue, PBD: 3CIW) and HydF [18] (grey, PBD: 5KH0). The figure was prepared by the author based on published crystal structures (protein structures), and spectroscopic data. [7]

1.3.2. HydE HydE also belongs to the radical SAM enzyme super family and is able to coordinate two [4Fe4S] clusters. Mutagenesis studies have shown that one of the clusters is involved in the radical generation, but the mutation of the other cluster has not shown a phenotype. Hence its role in the catalysis is unclear, and it seems this second cluster is not essential. [16] HydE is most likely involved in the azadithiolate bridging ligand synthesis, but the substrates, the products and the exact reaction mechanism are still unknown (Figure 4). [26, 27]

17 1.3.3. HydF HydF acts like a scaffold protein, carrying the pre-catalyst, which is synthe- tized by HydG and HydE, which will be transferred to the [FeFe] hydrogenase (HydA) to become part of the H-cluster. [28]

Figure 5. HydF timeline based on [7, 12, 15, 29-33] articles. The timeline represents the main findings of the HydF literature that inspired this thesis.

1.3.3.1. The structure of HydF On a tertiary structural level, HydF has three distinct domains: Domain I is the GTPase domain, Domain II is the dimerization domain, and Domain III is the iron-sulfur cluster domain. The dimerization domain exhibits a very hydrophobic surface which leads to dimerization in solution. However, in both crystal structures, tetramers were observed. [18, 34] Due to the fact that HydF is able to form both dimers and tetramers, the “active” quaternary structure of HydF is unclear.

The GTPase domain exhibits GTPase activity in isolation – isolated from the other two domains – both under aerobic and anaerobic conditions. [35] Certain HydF proteins’ GTPase activity can be inhibited by monovalent cations like Na+. [32] The GTPase domain is involved in the interaction with the other maturase enzymes. The interactions between the scaffold protein and the other two maturase enzymes are so strong that the dissociation requires high energy, which is gained from GTP hydrolysis. [36]

18 The iron-sulfur cluster domain contains 3 conserved cysteines, which are able to coordinate a redox-active [4Fe4S] cluster. The cluster can be reconstituted in vitro, and its presence has no effect on the GTP hydrolysis. [29] The fourth ligand of the [4Fe4S] cluster is a glutamate or an aspartate. [18] However, this latter ligand is labile, and it can be exchanged to a water or an imidazole lig- and. [37, 38]

1.3.3.2. The pre-catalyst on HydF HydF was biochemically and spectroscopically characterized first as apo- HydF, obtained from heterologous overexpression in the absence of HydE and HydG. [29] The in vitro reconstituted [4Fe4S] cluster containing form of this protein is referred to as “apo-HydF”, meanwhile the form, which was expressed together with HydG and HydE carries not just a [4Fe4S] cluster, but also a so called “pre-catalyst”. This form is referred to as “holo-HydF”. The nature of the pre-catalyst has remained a subject of speculation. The first insights into this species was published in 2010. Almost at the same time, two studies showed that in the Fourier Transformed Infrared Spectroscopy (FTIR) spectrum of holo-HydF, iron-carbonyl and iron-cyanide vibrations can be identified. [31, 32] Soon thereafter an X-ray absorption spectroscopy (XAS) study showed that the pre-catalyst probably is a binuclear iron complex. [33] In combination the FTIR and the XAS studies proposed that the pre-catalyst on HydF is probably a binuclear iron complex, similar to the [2Fe] subsite of the H-cluster (Figure 6). The presence of the [4Fe4S] cluster and the binuclear iron complex with cyanide and carbonyl ligands raises the question whether the whole complex is being transferred to HydA, or only the [2Fe] subsite. Mulder and co-workers demonstrated that a pre-formed [4Fe4S] cluster is needed on HydA in order to get an active hydrogenase, which means that only the pre-catalyst is transferred. [39]

Figure 6. The proposed structure of the pre-catalyst on HydF based on [7, 31-33] Other models also exist, including a model with bridging carbonyl ligand between the two irons of the pre-catalyst [31]and a model which is suggesting that the cyanide ligand bridging the [4Fe4S] cluster and the pre-catalyst is binding to the [4Fe4S] cluster via the nitrogen of the cyanide ligand not the carbon. [40]remove bracket oxidation state

19 1.4. The artificial maturation Ever since the crystal structure of the [FeFe] hydrogenase from Clostridium pasteurianum was published [5], numerous synthetic chemists have been inspired to prepare synthetic organometallic complexes that mimick the [2Fe] subsite of the H-cluster. [41, 42] These biomimetic complexes are mostly poor proton reduction catalysts, but it turned out that they can be introduced into the [FeFe] hydrogenase enzyme. One specific variant, with 4 carbonyl, 2 cyanide, and a briding azadithiolate ligand ([2Fe]adt) is able to form an H-cluster which is biochemically and spectroscopically indistinguishable from the naturally constructed H-cluster. Other variants with different ligands, different metals and different bridgeheads provide a great tool to explore the function of the different parts of this organometallic active site via “chemical mutagenesis”. [12, 43-45] This has led to a new chapter in [FeFe] hydrogenase research. The semi-synthetic or semi-artificial approach to [FeFe] hydrogenase maturation means that a synthetic complex can be introduced into a biologically produced protein. In our research group these semi-synthetic proteins are not considered as artificial enzymes, because the H-cluster itself is still resem- bling the naturally constructed H-cluster with minor modifications in some cases. Thus, also according to the author’s personal opinion, these semi-synthetic forms should rather be considered as organometallic mutants than artificial enzymes.

1.4.1. Artificially matured HydA proteins [FeFe] hydrogenase (HydA) enzymes from different organisms have been used to create large quantities of active HydA proteins with the artificial maturation method. HydA1 from the microalgae Chlamydomonas reinhardtii, was the first artificially matured HydA ([2Fe]adt-HydA1) together with the clostridial CpI from Clostridium pasteurianum.[7, 43] These HydA proteins were known as typical models for [FeFe] hydrogenase studies, because of the well-established sample preparation. The comparative spectroscopic and biochemical analysis of the artificial forms of these HydA proteins has led to the conclusion that the artificial maturation results in a semi-synthetic HydA protein indistinguishable from the naturally matured form. It is noteworthy that these synthetic binuclear iron complexes are catalytically inactive in solution, but become remarkably active catalysts upon incorporation into the enzyme. This underscores the importance of the protein environment and the presence of the [4Fe4S] cluster for efficient catalysis.

The HydA protein from Megasphaera elsdenii has been characterized in apo form [46] and in semi-synthetic holo-form [47] but the naturally matured form has not been characterized yet. In contrast to HydA1 and CpI, this semi-synthetic MeHydA was a new representative from a previously uncharacterized

20 subclass of the monomeric [FeFe] hydrogenases. This has inspired researchers to explore other unknown hydrogenases with the artificial maturation technique. It is important to mention, that recently a heterodimeric hydrogenase was also activated and characterized, as well as a sensory hydrogenase. [48, 49]

1.4.2. Artificially matured HydF proteins Since 2013, the preparation of large quantities of high quality [FeFe] hydrogenases can be a relatively inexpensive process with the “semi-synthetic” or “semi-artificial” approach, compared to the standard “biological” approach. These synthetic binuclear iron complexes can also be introduced into apo- HydF, creating a semi-artificial or semi-synthetic form of holo-HydF. [7, 18] Such semi-synthetic holo-HydF proteins have been spectroscopically characterized and compared to biologically matured holo-HydF proteins. The results were supporting the notion that the semi-synthetic holo-HydF proteins are similar to the biologically matured ones. Moreover, we know, that if we use this semi-synthetic holo-HydF protein to activate an apo-HydA protein, the resulting holo-HydA is biochemically and spectroscopically indistinguishable from the naturally matured ones. [7] However, it is very important to point out that the biological references were from different organisms than the HydF proteins that were used to prepare the semi-synthetic HydF proteins., the lack of biological controls from the same origin made the biological relevance of these semi-synthetic forms questionable.

21 2. Aims

There are many potential research topics/questions related to [FeFe] hydrogenases that can be addressed. The main motivation of this work was to con- tribute to [FeFe] hydrogenase research by studying how this fascinating catalyst is constructed. In this thesis, I review the results of my research done in projects devoted to the understanding of H-cluster formation, also known as the birth of the hydrogenase.

On a societal level we of course also have to ask the question what can we use hydrogenases for? Biofuel production? Solar fuel production? In order to be able to take full advantage of them, we need to know how they are assembled so we can introduce the optimized machinery into suitable host organisms, which we can use in biotechnological applications.

More specifically, the aims of this research were:

- To prove that the semi-synthetic forms of holo-HydF are biologically relevant model systems.

- To identify the minimal composition of the HydF protein needed to bind the pre-catalyst and to activate HydA.

- Examine the quaternary structure of the HydF-HydA interaction.

- To get a molecular level insight into the mechanism H-cluster assembly.

- Develop a method for following the maturation of [FeFe] hydrogenases spectroscopically in vivo.

22 3. In vivo activation of the [FeFe] hydrogenase with biomimetic complexes (Paper I)

3.1. Motivation and strategy It has recently been shown that the [FeFe] hydrogenase from Chlamydomonas reinhardtii (HydA1) can be heterologously expressed in apo form in both E. coli and cyanobacteria, and activated in vivo via the addition of the synthetic [2Fe]adt complex in complete absence of the three maturase enzymes. [50, 51] Successful enzyme activation was shown via hydrogen production rates, which were significantly above the hydrogen production of the negative controls in both cases. However, there was a complete lack of spectroscopic evidence that the hydrogenase is synthetically activated in the living cells.

The motivation was to develop a model system that can be used to monitor the [FeFe] hydrogenase in “living” cells. If possible, this could open a whole new chapter in the literature. On one hand, new redox states (both catalytically relevant and not relevant) could be identified. This would clarify the in vivo relevance of certain inactive states. On the other hand, imagine, how many putative [FeFe] hydrogenase has been identified over two decades, but still only a few of them have been purified and spectroscopically characterized. [2] To study and characterize unknown hydrogenase enzymes they need to be isolated in soluble form, which can be challenging in E. coli even in apo-form, and to co-express with the maturase enzymes is even more challenging. The idea would be to express unknown [FeFe] hydrogenases in E. coli and use the artificial maturation technique (hydrogen evolution assays and spectroscopy) as a screening method.

The strategy with the primary model system was very simple. We choose E. coli as a host organism to express HydA1, because activity assays have shown that we can activate large quantities of HydA1 with the [2Fe]adt complex. The two most commonly used spectroscopic techniques to characterize [FeFe] hydrogenases are FTIR and low temperature CW-EPR. This is because FTIR can detect the diatomic ligand vibrations, while low-temperature CW-EPR can detect the unpaired electrons. As discussed briefly in Chapter 1, certain redox states are EPR silent. Consequently, with low-temperature CW-EPR we are

23 practically half blind. because of the lack of EPR active catalytic intermediates that can be easily populated. Nevertheless, the first spectroscopic tool was low temperature CW-EPR above FTIR because EPR has a much higher sen- sitivity than standard transmission FTIR. Moreover, the sample preparation is straight-forward and the samples can be stored on liquid nitrogen for long term. In order to avoid the loss of the signal due to the catalytic activity, the choice of biomimetic complex was the [2Fe]pdt complex. The [2Fe]pdt complex has a propane-dithiolate bridging ligand (pdt) instead of an azadithiolate bridging ligand (adt). This small structural difference leads to a very interesting phenomenon. In the case of [2Fe]pdt complex matured HydA1 ([2Fe]pdt- HydA1), the H-cluster is either locked in the Hox or in the Hred state because the carbon bridgehead cannot be protonated in contrast to the nitrogen bridgehead. [52, 53] This Hox or sometimes called “Hox-like” state gives rise to a rhombic EPR signal.

3.2. Sample preparation The sample preparation needed to be improved compared to the first attempts, where HydA1 was expressed in Luria-Bertani medium (LB medium). LB is a complex, undefined medium, full of transition metals like Fe, Cu, Mn. Our model organism, E. coli tends to accumulate Mn ions, which show up in the EPR spectra of the HydA1 expressing E. coli cells as a “six line” Mn signal that correspond to the free Mn ions in E. coli. In order to eliminate the man- ganese lines from the EPR spectra, we changed the medium from the undefined LB to M9 minimal medium. The cells growing in the M9 minimal medium showed practically no “six line” Mn signal (Figure 7, a).

3.3. In vivo activation of the HydA1

3.3.1. The identification of the EPR signal The over expression of apo-HydA1 in M9 minimal medium by E. coli BL21 cells was followed by harvest and transfer of the cells to EPR tubes. We compared the EPR spectra of the HydA1 overexpressing E. coli cells to the cells that were lacking the plasmid with the HydA1 encoding gene. We observed minor differences, but we could not capture any EPR signal that would clearly correspond to the reduced [4Fe4S] cluster of the non-activated apo-HydA1. Although, in the g=1.9 region we observed signals which may correspond to the reduced [4Fe4S] cluster proteins of E. coli.

Figure 7. The low temperature CW-EPR spectra of the apo-HydA1 overexpressing E. coli cells (a), the [2Fe]pdt complex treated E. coli cells expressing apo-HydA1 (b), the difference spectrum of the treated and non-treated cells (c) Spectrum (a) subtracted from spectrum (b), (d) and purified HydA1 treated with [2Fe]pdt complex. Dashed lines represent simulated spectra.

In order to form [2Fe]pdt-HydA1, the apo-HydA1 expressing E. coli cells were treated with [2Fe]pdt complex. The resulting cell paste was analyzed by low temperature CW-EPR spectroscopy. The [2Fe]pdt complex treated cells exhibited a rhombic EPR signal where the g-values were identical to the g-values of the purified form of the [2Fe]pdt-HydA1 enzyme (Figure 7). [53] This signal was not visible in apo-HydA1 expressing E. coli cells before addition of [2Fe]pdt, nor in the [2Fe]pdt complex treated E. coli cells, which were not expressing apo-HydA1. These observations underscore the successful formation of a semi-synthetic H-cluster under in vivo conditions.

3.3.2. The time dependence of the activation Spin quantification studies combined with standard in vitro enzymatic assays showed that we were able to activate 25% of the total hydrogenase population

25 of the cells. This result combined with the time dependence study is suggesting that the activation process was not completed after one hour (Figure 8). To see the rate of incorporation, we incubated the cells together with the [2Fe]pdt cofactor for various times between 0 and 12 hours.

Figure 8. Low temperature CW-EPR spectra of the time dependence of the In vivo activation of HydA1 with the [2Fe]pdt complex. In the inset red circles are indicating the intensity of the peak at g=2.100 and while red circles are indicating the relative intensity at g=2.009

In the time dependence data set we can see that the rhombic Hox-like signal is rising rapidly from the first time point (zero hours) up to 2 hours and then it reaches a plateau. This suggests that the activation is practically done after 2 hours. It raises the question why the activation is so slow compared to a standard in vitro activation on purified protein that can be carried out on minute time scale – which will be demonstrated in the next chapters – rather than hour time scale. [54] To answer this question, we need to take into consideration that the [2Fe]pdt complex has to pass the capsule, the cell wall and the plasma membrane of the E. coli cell. An interesting fact is that in order to activate the HydA1 population within the cell, we need to use 100 times more from the [2Fe]pdt or [2Fe]adt complex in the in vivo activation assay – where the biological barriers are present – than in the in vitro activation assay – where the

26 complex can easily get into contact with the hydrogenase enzyme because of the lack of the biological barriers. [50] It has been suggested during scientific discussions - based on these results - that the uptake of the complex happens with active transport, but in my opinion this data suggests that the uptake of the complex happens either with a passive transport (like facilitated; carrier or channel mediated diffusion) or a “hijacked” active transport. My reasoning behind this is the fact that living bacteria cells cannot have a specialized trans- porter for a synthetic binuclear iron complex such as our [2Fe]pdt or [2Fe]adt complex.

A carrier or channel mediated passive transport would require a higher concentration of complex outside than inside, so in principle a channel or carrier can recognize the complex as something “useful” and let it in. The other option would be a “hijacked” active transport. This would mean that a symporter (like a Na-glucose symporter or something similar) would misrecognize the biomimetic [2Fe] complexes, and the transport would then be glucose dependent. It is an interesting fact that we indeed need to use glucose for the in vivo activation. [50] In principle, a patch-clamp method could be used to identify the nature of the channel or carrier in order to clarify this question.

3.3.2. The power saturation behavior in vivo We carried out power saturation studies (Table 1) to examine the behavior of the [2Fe]pdt H-cluster in a more natural environment. The idea behind this was the observation that all the spectroscopic characterization of HydA1 was so far carried out at neutral pH (pH 7-8) in artificial (Tris or HEPES) buffers with well-defined salts and salt concentrations. Our concern was that the protein and the H-cluster could behave differently in their “natural” environment, where the buffering compounds are intracellular metabolites, which are low molecular weight organic acids (e.g. citric acid), in short: in a more physio- logical (albeit not native) environment. However, the power saturation studies on the purified protein control and the [2Fe]pdt treated whole cells showed that the relaxation properties of the H-cluster is not affected by the cellular environment.

Table1. The power saturation studies of the in vivo activated HydA1 compared to the purified protein control.

27 This is attributable to the fact that the H-cluster in HydA1 is buried and most probably the H-cluster is not affected by the solvent environment. Strongly coupled magnetic centers can cancel each other – for example the [4Fe4S] cluster and the [2Fe] subsite in the H-cluster in the case of the Hred state are both paramagnetic, but because of the physical proximity and coupling the species becomes EPR silent. Meanwhile, weakly coupled species like the [4Fe4S] clusters in the F domain [49] of the sensory hydrogenase from Ther- motoga maritima or the electron transport chain [4Fe4S] clusters in the small subunit of the uptake NiFe hydrogenase from Nostoc punctiforme [55] can be EPR active, and can take more microwave power than species with only one magnetic center. As a matter of fact, the protein shell is protecting the H-cluster from any other magnetic interaction, not just with other activated HydA1 proteins, but also the other magnetic centers coming from the housekeeping proteins of E. coli. This could explain why we have not observed differences in power saturation between the in vivo and in vitro examined samples.

In parallel to power saturation behavior, EPR is sensitive to structural changes as well. Note for example the difference between the Hox and Hox-CO signal, where the oxidation states are identical ([4Fe4S]2+ and FeIFeII on the [2Fe] subsite in both cases) and the difference is an extra carbonyl ligand on the [2Fe] subsite. While Hox gives rise to a rhombic signal, Hox-CO gives rise to an axial EPR signal. [56] Despite the fact that EPR is sensitive also for small structural changes, we have not observed any difference between the in vivo and in vitro examined samples.

28 3.5 Summary and Discussion

In conclusion this study established a method to follow the [FeFe] hydrogenase maturation spectroscopically in vivo. We have not observed any difference between the isolated and the in vivo captured [FeFe] hydrogenase samples. Still, this finding underscores the possibility of in vivo spectroscopic studies of [FeFe] hydrogenases. This can be a new powerful tool in the future to follow the catalytic cycle, maybe even to identify new redox species in vivo, which could not be identified or populated in vitro. Also, this method could be used to examine new [FeFe] hydrogenases which are challenging to isolate. Although, it is very important to point out that the primary characterization of a new, unknown hydrogenase in vivo must be followed by proper isolation and spectroscopic characterization, but this method is a great tool for screening. The mechanism of the cofactor uptake remains unclear, however we must take it into consideration that E. coli is very efficient from a survival and eco- logical niche point of view. E. coli can adapt easily, it can live under various conditions (deprivations, temperature differences, presence or absence of oxygen, various carbon sources) and it can be easily transformed with foreign DNA. This biochemical flexibility can lead to the recognition that E. coli is not a “picky eater”, rather just a microbiological “horder” – which includes the uptake of everything that is in the surroundings and accumulate it, regardless of the presence or the absence of metabolic relevance. Another illustrative example of this is the uptake of a synthetic lanthanoid complex, reported in collaboration with our group. [57] The synthetic lanthanoid complex is cer- tainly not even close to something found in nature. This supports the idea that there should be a generalized “foraging” pathway in E. coli.

29 4. Monitoring the H-cluster formation using a synthetic binuclear iron complex (Paper II)

4.1. Motivation and strategy As introduced in Chapter 1, the H-domain has a channel-like structure, which appears to be open in the crystal structure of the unmatured, apo-enzyme, and it appears to be closed in the matured, holo-enzyme.[4, 5] Recent studies have shown that this channel is directly involved in the pre-catalyst integration process. [3] Our motivation was to provide a deeper insight into this cofactor integration and more specifically the H-cluster formation process.

During the natural H-cluster formation there are several steps that must happen. These steps are: HydF docking to HydA, release of the pre-catalyst from HydF, pre-catalyst transfer to the active site through the channel, pre-catalyst fusion with the [4Fe4S] cluster of HydA. In addition, protein film electrochemistry studies have shown that the reaction has a potential dependence. [58] Under reducing conditions, the fusion between the [4Fe4S] cluster and the pre-catalyst does not occur, rather the pre-catalyst becomes immobilized in the channel. This intermediate was named “intermediate P”, but has so far not been spectroscopically detected. Likewise, there is a complete lack of information about the intermediates occurring between the fusion and the formation of the catalytically active H-cluster.

In this chapter the focus is on the last steps of the maturation events (Figure 9), so for the sake of simplicity we used the synthetic [2Fe]adt pre-catalyst instead of holo-HydF. As will be discussed in the next chapter, the [2Fe]adt pre- catalyst in solution is, if not identical, at least strikingly similar, to the pre- catalyst on both the naturally matured holo-HydF and the semi-synthetic holo- HydF. [54] Thus this study is arguably biologically relevant.

We probed the reaction using X-ray Absorption Spectroscopy (XAS), FTIR spectroscopy and EPR spectroscopy. XAS is a very powerful spectroscopic technique, which gives local information about the distances to the nearest neighboring atoms, redox states and coordination geometry, of the element under study. More specifically, the XAS spectrum has two main parts, the

30 XANES and the EXAFS. The XANES (X-ray Absorption Near Edge Struc- ture) has two main features, these are the pre-edge and the edge. The pre-edge can give us information about the coordination geometry, while the edge can provide information about oxidation states. The EXAFS (Extended X-ray Ab- sorption Fine Structure) is informative about the number of bonds and bond lengths. The EPR spectroscopic fingerprints of the catalytic species are unique, but several redox states, including reduced species are EPR silent. FTIR can give us information about the diatomic ligand vibrations, which re- port on, for example, the electron density of the Fe atoms. These techniques complement each other, and together can provide a better insight into the H- cluster formation reaction.

4.2. Sample preparation Freeze-quench measurements were carried out in the case of XAS and EPR, while the FTIR measurements were carried out in real time.

The XAS and EPR samples were prepared in a very similar way. In the case of the XAS samples, the unmatured, [4Fe4S] cluster coordinating form of HydA (apo-HydA1) and [2Fe]adt solutions were mixed in 1:1 volume ratio. The mixing was ensured by some gentle pipetting before the sample was in- jected into XAS sample holders and flash frozen immediately or after a certain incubation time. In the case of the EPR samples the mixing happened inside the EPR tube and this let us gain a higher time resolution. With the XAS samples, this kind of mixing was nearly impossible to carry out, as the small volume of the XAS cell, and the high concentration of the protein was leading to foaming, which is problematic because the bubbles are interfering with the XAS measurement. In both cases, the captured time points had biological replicates (two biological replicates in two series of samples in the case of the XAS and three biological replicate series with three technical repeats in each series in the case of EPR) and the samples were stored in liquid nitrogen until further use.

31 Figure 9. A schematic representation of the proposed mechanism of the last steps of the H-cluster formation based on [54] and [58].

32 4.3. X-ray absorption spectroscopy at the Fe K-edge The XAS and the data analysis was performed by Dr. Michael Haumann and Dr. Stefan Mebs, at the Free University of Berlin in Germany, the samples were prepared at Uppsala University by the author.

4.3.1. The Fe XANES data analysis As discussed above, the XANES data can provide information about oxidation states and coordination geometry. The XANES spectra (Figure 10) of apo- HydA1 and the [2Fe]adt complex were clearly different from each other, but both the apo-HydA1 and the [2Fe]adt complex spectra matched previous studies. [59-61] The fact that the spectra of the starting materials were in good agreement with previous datasets was a verification of the quality. The sharp differences in K-edge energy between the apo-HydA1 and [2Fe]adt is attributable to the different oxidation states of their iron atoms, the oxidation state of the irons in the [2Fe]adt complex are FeIFeI, while in the [4Fe4S] cluster it is FeII-FeIII-FeII-FeIII when the cluster is in it’s “oxidized” [4Fe4S]2+ state. The first time point (30 sec) of the apo-HydA1 and [2Fe]adt complex mixtures was clearly different from the digital sum of the apo-HydA1 and [2Fe]adt cofactor spectra. Assuming that the electronic structure of intermediate P is highly similar to the [2Fe]adt complex, we can exclude the possibility that the first spectrum was representative of intermediate P. This data suggested that already 30 second was enough for the [2Fe]adt cofactor to enter into the channel and to form a new species. In short, the XANES spectra revealed an oxidation of one of the six irons. This could be attributable to Hox or potentially Hox-CO state.

33 Figure 10. Fe-XANES data. (A) Top: spectra of apo-HydA1 protein and [2Fe]adt complex solutions, the stoichiometric sum (4 x apo-HydA1 + 2x [2Fe]adt) / 6, and the apo- HydA1 + [2Fe]adt mixture after a mean incubation period of 32 s. Bottom: spectra of apo-HydA1 / [2Fe]adt mixtures after indicated mean incubation periods. The insets show the pre-edge feature (left panel, asterisks) and the edge half-height (right panel, dashed lines) in magnification. (B) K-edge energies (at 50 % XANES amplitude) and pre-edge amplitudes (at ~7113.5 eV). Black open squares and triangles denote data from two series of samples. (A) colored solid circles denote data for respective mean spectra. The lines in the main and inset panels show kinetic simulations (For details, see Paper2).

34 4.3.1. The Fe EXAFS data analysis The structural changes were addressed by Fe K-edge EXAFS data analysis (Figure 11). In the EXAFS part of the spectrum we can identify a sum of many different oscillations. These oscillations are not coming from one single species, hence in order to identify the species which are giving rise to the unique oscillation, we need to simulate the oscillation of each bond until we get a good fit. The best fit tells us which are the bonds and bond lengths we have in the sample. In the apo-HydA1 spectrum the Fe-S bond lengths and the Fe-Fe bond distances were in good agreement with previous studies. Similarly, the EXAFS spectrum of the [2Fe]adt solution shows the expected number of ligands and iron-iron bond length. In conclusion we can say that both starting components were intact. [8, 61-63]

In the apo-HydA1 and [2Fe]adt complex mixtures, differences compared to the digital sum of the starting materials spectra were identified. Based on the Fe- S bond lengths and the Fe-Fe distances, the Hox state can be detected, which is in good agreement with previous EXAFS studies and crystal structures. The main mean of the Fe-Fe distances and Fe-S bonds remained unchanged for longer mixing periods than 75 seconds, while the Fe-Fe distance corresponding to the Fe-Fe distance in the [2Fe] subsite became shorter. Furthermore, an additional Fe-C(N/O) bond was formed up to 75 seconds, which was followed by a decrease in the number of Fe-C(N/O) bonds for longer mixing periods. This additional Fe-C(N/O) bond is attributed to the bridging CO ligand of the Hox-CO state. Moreover, the later decrease in the number of Fe-C(N/O) bonds can be attributed to the release of the fourth CO ligand resulting in the Hox state after extended incubation time. [64, 65]

35 Figure 11. EXAFS analysis. (A) Fourier-transforms (FTs) of EXAFS oscillations in the inset of apo-HydA1 protein and [2Fe]adt complex solutions, the stoichiometric sum, and the apo-HydA1 + [2Fe]adt mixtures (mean spectrum over data for all incubation periods. Inset: black lines, experimental data; colored lines, simulations with parameters in. (B) FTs in the left panel of EXAFS oscillations in the right panel for indicated mean apo-HydA1/adt mixing periods. (For details see Paper2)

36 4.4. EPR spectroscopy The low temperature CW-EPR spectroscopy and the data analysis were performed by the author at Uppsala University in Sweden.

The XAS showed that upon mixing of apo-HydA1 and the [2Fe]adt complex, the Hox-CO state can be detected, which turns into Hox after extended incubation time. Both Hox and Hox-CO are paramagnetic S=1/2 species which can be easily detected with EPR. In the low temperature CW-EPR spectrum of oxidized holo-HydA1 samples a mixture of two features, an axial and a rhombic EPR signal can be identified. For demonstration purposes, the author used one of her experimental datasets to deconvolute the Hox/Hox-CO mixtures (Figure 12). The axial signal (g-values g║ = 2.052 and g┴ = 2.007) is attributable to the Hox-CO state, meanwhile the rhombic signal (g-values gx=2.10, gy=2.037, gz=1.996) is corresponding to Hox state. [56]

Figure 12. Simulation of Hox and Hox-CO states based on low temperature CW-EPR spectrum of an Hox/Hox-CO mixture. The author used the published g-values and her experimental spectrum for the simulation. The simulation was carried out in MATLAB programing environment with EasySpin expansion.

37 Fig.13. Low temperature CW-EPR spectra of the captured time points, bottom to top: 5, 10, 15, 30, 60, 100, 200 and 300 seconds. The spectra were recorded with 1 mW microwavepower, 15 Gauss modulation amplitude, at 10 K. The microwave frequency was 9.38 GHz. Each time point has three technical replicates shown as individual spectra, two scans were accumulated and averaged.

38 The EPR spectra were recorded at 10 K, 1 mW microwave power on 24 samples, which represent eight time points with three technical repeats (Figure 13). In isolation, both apo-HydA1 and the [2Fe]adt complex were in their expected, EPR silent states. In contrast, the mixtures gave rise to identifiable EPR signals in every spectrum of the time series. In the first time points (5-10 seconds) we were able to identify only the axial EPR signal, corresponding to the Hox-CO state. In later time points (15-300 seconds) we can also observe the rise of part of the rhombic EPR signal, attributable to the Hox state. The Hox-CO signal shows a quick rise up to 100 seconds, then it reaches a plateau followed by a weak decrease. Meanwhile, the Hox signal starts to be detectable from 15 seconds and it becomes more and more clear and dominant with in- creased incubation times. Furthermore, a rise of a reduced [4Fe4S] cluster signal can be identified after 60 seconds (this is indicated with an asterisk in Fig- ure 13). The parallel g-value is not visible but the perpendicular g-value (g=1.91) is suggesting that this signal is coming from the reduced [4Fe4S] cluster of HydA1. [39]

The results of the EPR spectroscopy confirms the oxidation of the [2Fe]adt pre- catalyst upon H-cluster formation. Furthermore, the delayed appearance of Hox supports the initial formation of Hox-CO followed by the transition into Hox. The continuous increase of both Hox-CO and Hox could be explained in two ways. It is possible, that the first form is an EPR silent species, like Hred-CO which slowly converts into Hox-CO. The other option would be slow cofactor insertion.

4.4. FTIR spectroscopy The ATR-FTIR spectroscopy and the data analysis was performed by Dr. Sven Stripp and Dr. Moritz Senger at the Free Univeristy of Berlin in Germany.

In contrast to the XAS and low temperature CW-EPR experiments, the ATR- FTIR spectroscopy was performed at room temperature under nitrogen atmosphere. Furthermore, in the XAS and low temperature CW-EPR experiments the representative “captured” timepoints were individual samples with biological replicates, in contrast to the ATR-FTIR spectroscopy, where the reaction was followed continuously. The ATR-FTIR experiment was performed on a semi-dry film of apo-HydA1.

39 The reaction was initiated by the addition of the [2Fe]adt complex to the thin film, so this time point should be considered as t0. In the plotted datapoints of the ATR-FTIR dataset we can observe an immediate rise of Hox-CO, followed by the rise of Hox approximately 100-200 seconds later. In addition, two re- + duces species, Hred and HredH also grow into the spectrum upon extended incubation time. The latter two are EPR silent species, but they can be readily identified with FTIR from their characteristic wavenumbers. The FTIR data is in good agreement with the low-temperature CW-EPR data in the sense of the initial formation of Hox-CO followed by a slow transition into Hox (Figure 14).

Hox-CO Hox Hred´ Hred

-3 0 s 0 200

400

600 -3 800

1000

1200 absorbance x 10 x absorbance -6 1400

2000 1900 1800 frequency / cm-1

Hox-CO 4 Hox -3 Hred Hred´ 3

2 [2Fe]adt

1 absorbance x 10 x absorbance 0

0 400 800 1200 1600 time / s

Figure 14. Bottom: The absorbances corresponding to Hox and Hox-CO states plotted as a function of time. Top: the raw data set. Please note that in the figure Hred refers + to HredH and Hred’ refers to Hred.

In the ATR-FTIR spectra we were not able to identify the vibrations which would correspond to Hred-CO state. This finding argues against the possibility of a reduced species, not detected by EPR or XAS, accumulating in the early stages of the reaction.

40 4.5 Summary and discussion In this study we employed three spectroscopic techniques namely XAS, low temperature CW-EPR and ATR-FTIR, in order to gain a better insight into the final steps of the H-cluster formation.

As introduced in Chapter 1 and in the beginning of this chapter, a channel-like structure in the protein backbone of the [FeFe] hydrogenase is most likely directly involved in the maturation of the [FeFe] hydrogenases. The existence of an intermediate P has been proposed, but so far it has not been identified spectroscopically, however this intermediate is stable under reducing conditions according to electrochemical studies. [58]

Despite the different experimental conditions – room temperature vs. cryo- genic temperatures, solution vs. film – these three different spectroscopic techniques are reaching towards the same conclusions. In combination with the aforementioned electrochemistry study our new results allow us to outline a reaction sequence of H-cluster formation. This sequence involves at least three intermediates: 1) The formation of a so-far spectroscopically unidentified intermediate, 2) formation of the Hox-CO state, and 3) Hox formation.

41 5. Monitoring the H-cluster formation using a semi-synthetic HydF protein (Paper III)

5.1. Motivation and strategy In 2013, a new era began in the history and literature of the [FeFe] hydrogenases. A biomimetic complex ([2Fe]adt) was introduced into an in vitro enzymatically reconstituted HydF protein (Figure 15). [7] This “semi-synthetic” HydF protein from Thermotoga maritima, was characterized by CW-EPR, pulsed-EPR, FTIR, UV-Vis spectroscopic techniques and biochemical assays such as iron and sulfur quantification. These semi-synthetic HydF proteins were also tested for HydA activation. The HydA protein - which was activated by this semi synthetic form of HydF – is biochemically and spectroscopically indistinguishable from the naturally matured ones.

The FTIR spectrum this semi-synthetic HydF exhibits peaks, attributable to metal-carbonyl and metal-cyanide vibrations. The FTIR characteristics – the peaks corresponding to Fe-CO and Fe-CN vibrations – were very similar to the previously published FTIR spectra of biologically matured CaHydF (HydF protein from Clostridium acetobutylicum) proteins [31, 32]. Still, the lack of the biological control – a biologically matured TmHydF protein – left the authenticity, or biological relevance, of this form of HydF unanswered.

Figure 15. The biomimetic approach in a nutshell. A- the [2Fe]adt biomimetic complex, B-the [4Fe4S] cluster on HydF. X can indicate a glutamate, aspartate or a his- tidine aminoacid ligand or it can be exchanged to imidazole or water as fourth lig- and [18, 37, 38, 66]C- the proposed structure of the pre-catalyst on HydF based on spectroscopy studies. [7]

Our motivation was to prove the biological relevance of these semi-synthetic HydF proteins by using the synthetic the binuclear iron complex [2Fe]adt to create a semi-synthetic holo-HydF protein, which we can directly compare to

42 previously characterized, naturally matured holo-HydF proteins. There are other HydF proteins with possible biological backgrounds to compare, [67, 68] but our choice was to use HydF from Clostridium acetobutylicum as model system, since it is the most common biological model system for biochemical spectroscopic studies of HydF. [28, 31-33, 36, 40, 66] As a representative example of HydA, our choice was the HydA1 protein from the micro algae Chla- mydomonas reinhardtii, which is a biochemically, electrochemically and spectroscopically well-characterized [FeFe] hydrogenase, both in biological and semi-synthetic forms. [44, 56, 69-71] The expression and purification of HydA1 was carried out following previously published protocols. [39]

5.2. Sample preparation The HydF protein encoding gene from Clostridium acetobutylicum (CaHydF) construct was not codon optimized for E. coli. Hence in order to achieve sufficient expression levels, the Rosetta(II) E. coli strain was employed, as this strain carries a rare codons encoding plasmid. Large scale (6-24 L) expression resulted in sufficient amounts (8 mg purified protein per OD600 per liter cul- ture) of soluble dimeric CaHydF protein. In our hands, the [4Fe4S] cluster incorporation was very poor; the protein was often purified with minimal or no iron content at all. The residual iron was removed via chelation by EDTA in the precense of sodium dithionite (Na-DT). The chelation reaction was stopped by separating the resulting metal free proteins from the chelating rea- gents using size exclusion chromatography (SEC). The CaHydF protein tended to form higher oligomeric forms which eluted in the void volume of the SEC column. This oligomerization was suppressed by slight changes in the expression and purification protocol following literature protocols. [72]

The size exclusion chromatography was followed by in vitro enzymatic reconstitution. The “as prepared” proteins more or less remained in their EPR silent state. However, a small fraction of reduced [4Fe4S] clusters gave rise to an axial EPR signal.

The reconstituted HydF protein was loaded with [2Fe]adt complex in order to form the semi-synthetic holo-HydF protein. This procedure followed the literature protocols for other, previously characterized and published semi-synthetic HydF proteins. [7, 18] The incorporation of the [2Fe]adt complex was observable with UV-Vis spectroscopy by the appearance of a new peak at 350 nm attributable to the metal-ligand charge transfer band in the [2Fe]adt complex (Figure 16). The iron quantification upon maturation also showed an increase in iron content as compared to the [4Fe4S] cluster form.

43 1

0,9

0,8

0,7

0,6

0,5

0,4 Absorbance

0,3

0,2

0,1

0 250 350 450 550 650 750 850 Wavelength (nm)

Figure 16. Monitoring holo-HydF formation by UV/Vis spectroscopy. The solid lines represent experimental data, while the dashed line represent a linear combination of two experimental spectra. 130 µM [4Fe4S]2+-CaHydF (solid black), 130 µM [2Fe]adt-CaHydF (solid blue line); 130 µM [2Fe]adt (solid red); A linear combination of 130 µM [2Fe]adt and 130 µM [4Fe4S]2+-CaHydF spectrum (dashed blue). All spectra were recorded in 100 mM Tris-HCl pH 8.0, 300 mM KCl. The spectra were recorded with a l=0.2 cm cuvette.

5.3. Basic spectroscopic characterization The previously characterized, biologically matured forms of CaHydF and the semi-synthetic forms of HydF [7, 18] were all exhibiting metal-carbonyl and metal-cyanide vibrations in the FTIR spectrum. However the wavenumbers corresponding to these vibrations were not in perfect agreement with each other. [7, 18, 31, 32] The reason for the differences could be structural differences between the model proteins from different organisms or redox differences due to different sample preparation conditions. Having a semi-synthetic holo-HydF protein, we were in the position to investigate the origin of these differences (Figure 17).

44 Consequently, as prepared, sodium-dithionite (Na-DT) reduced and thionine- acetate oxidized proteins were examined by FTIR and CW-EPR spectroscopy. All these three forms of the semi-synthetic CaHydF exhibited four peaks attributable to metal-carbonyl and two peaks attributable to metal-cyanide vibrations. Minor changes (±5) in wavenumbers (cm-1) can be observed in the FTIR spectra. Such differences are considered “small” changes. In general one electron reduction of the [2Fe] subsite usually results in a 10-20 wavenumber shift in the FTIR spectra of [FeFe] hydrogenases, and in model complexes these numbers are even larger. [47, 73]

Figure 17. Low temperature CW-EPR and FTIR spectroscopic characterization of artificially matured CaHydF, holo-CaHydF. Top left: EPR spectrum recorded on as-prepared holo-CaHydF (0.2 mM); bottom left FTIR spectrum recorded on as- prepared holo-CaHydF (2.5 mM); top middle EPR spectrum recorded on Na-DT reduced holo-CaHydF (0.2 mM, 4 mM NaDT); bottom middle FTIR spectrum recorded on NaDT reduced holo-CaHydF (2.5 mM, 50 mM NaDT); top right EPR spectrum recorded on thionine oxidized holo-CaHydF (0.25 mM, 2.5 mM thioinine); bottom right FTIR spectrum recorded on thionine oxidized holo-CaHydF (2.5 mM, 25 mM thioinine). All samples were prepared in 25 mM Tris-HCl pH 8.0 50 mM KCl. EPR spectra were recorded at a microwave power of 1 mW, at 10 K, microwave frequency: 9.28 GHz.

As mentioned in the previous chapter, the oxidation state of the irons in the [2Fe]adt complex is FeI-FeI. The EPR spectrum shows no signal in the “as prepared” and in the “thionine oxidized” forms which implies that both the [4Fe4S] cluster and the pre-catalyst stayed in their initial EPR silent state, attributable to the [4Fe4S]2+ state in the case of the cluster and the FeI-FeI state

45 in the case of the [2Fe]adt pre-catalyst. The EPR spectrum of the reduced sample exhibits a typical axial signal attributable to the reduced [4Fe4S] cluster of CaHydF. However, the EPR signal of the reduced holo-HydF protein did not show notable differences in g-values, linewidth or lineshape compared to the apo form. The small changes in the wavenumbers and the lack of new features in the EPR spectra upon oxidation and reduction leads to the proposal that the pre-catalyst is weakly coupled to the [4Fe4S] cluster.

All three FTIR spectra exhibited 4 Fe-CO peaks and 2 Fe-CN peaks, suggesting that the pre-catalyst on HydF has 4 carbonyl and 2 cyanide ligands just like the synthetic [2Fe]adt complex. Upon oxidation and reduction of holo- HydF, mild shifts in the wavenumbers was observable. Based on the comparison of our new data set to previous studies on the biologically matured Ca- HydF, it was clear that the semi-synthetic form of HydF is authentic and biologically relevant, and the differences in wavenumbers in the previously reported CaHydF FTIR spectra was most likely due to the differences in oxidation states of the [4Fe4S] cluster. It was also clear that we are not able to reduce or oxidize the pre-catalyst on HydF, only the [4Fe4S] cluster is able to accept or donate electrons.

In a functional, catalytically active [FeFe] hydrogenase the [2Fe] subsite features three carbonyl ligands [5], in contrast to HydF, which carries a pre-catalyst with four carbonyls instead.

5.4. CO release To further support the FTIR spectroscopic data which suggest that the pre- catalyst on HydF carries four carbonyl ligands, a hemoglobin based CO release detection was performed. Assuming that the pre-catalyst is a four carbonyl species and that the fourth CO is released, in principle, the hemoglobin assay would allow us to detect the CO released by HydA upon H-cluster formation. The hemoglobin assay was performed based on previously published protocols [43, 74] with slight modifications. In our assay, it was not possible to do kinetic analysis due to the hemoglobin mediated decomposition of the pre-catalyst on HydF. Thus, HydF and HydA were incubated for various times, in various ratios at room temperature and in darkness. The incubation was followed by the addition of deoxyhemoglobin, and the mixture was gently mixed and incubated for 10-15 seconds, after which time UV-Vis spectra were recorded. The released CO was quantified by a home-written Excel script based on a two unknown equation system using the extinction coefficients of deoxyhemoglobin and carboxyhemoglobin at 419 and 430 nms. The extinction coefficients were calculated from experimental UV-Vis spectra and compared to published values which were in good agreement. [24]

Figure 18. The activation of apo-HydA1 using holo-CaHydF probed by enzymatic assays and CO release. From left to right: H2 production assay, CO release- HydA1 titration curve, Time dependence of CO release, CO release as a function of the relative ratio between apo-HydA1 and holo-CaHydF.

In previous studies [7, 28, 31] it was already shown that a large excess of holo- HydF (at least 10 fold molar excess relative to apo-HydA) is required in order to reach the maximal hydrogen evolution activity. The hydrogen evolution activity is directly proportional to the amount of active hydrogenase in solution, thus the maximum activity is corresponding to the complete activation of the hydrogenase population. This need for a large excess of holo-HydF is in contrast to the activation with only the biomimetic complex which can happen in 2:1 molar ratio of [2Fe]adt complex and HydA1. [43] It is very important to emphasize the difference between the hemoglobin assay and the hydrogen evolution assay. From the hemoglobin assay we get information about the H- cluster formation from a structural point of view since the release of the CO is corresponding to a formed H-cluster, in contrast to the hydrogen evolution assay, where the detected hydrogen is the product of a working H-cluster.

In our HydA maturation coupled hydrogen evolution assay we also needed to use an at least 10 fold molar excess of [2Fe]adt loaded forms of TmHydF or CaHydF (Figure 18). Similarly, the maximum CO release was also observed with at least 10 fold molar excess of [2Fe]adt loaded TmHydF or CaHydF proteins. Interestingly, during the method optimization, it was observed that the length of the incubation times in a 5-50 minute interval did not affect significantly the amount of released CO, neither with TmHydF nor with CaHydF in case we used 5:1 molar excess ratio of HydF to HydA. This 5:1 molar excess of HydF to HydA resulted in ~65% yield in CO release, where we considered 100% to correspond to the CO release we observed with 10 and 15 fold molar excess of HydF to HydA. To further investigate the effect of the length of the incubation time, we used a 1:1 ratio of apo-HydA and holo-HydF in a 5-50 minutes time interval. With this ratio, we have not observed the maximal amount of released CO in contrast to the experiments when we used a large excess of HydF, regardless of the incubation period. A reverse equivalent experiment was also performed, where we used more HydA than HydF. In this

47 case, when HydA was in 15 fold molar excess compared to HydF, the same amount of CO was released like in the case when HydF was used in 15 molar excess to HydA. Thus, complete cofactor transfer occurred from holo-HydF to apo-HydA, when one of the proteins was present in excess.

5.5. Redox chemistry As demonstrated and discussed in the previous chapter, oxidized species can be identified (Hox/Hox-CO) upon H-cluster formation, but we also observed the reduction of the [4Fe4S] cluster apo-HydA1 and the formation of reduced H-cluster species in the absence of external electron acceptor. Our hypothesis was that the [4Fe4S] cluster on HydF could serve as electron acceptor during the natural maturation process. In order to identify both the donor and the acceptor of the reaction, we used both FTIR and EPR spectroscopy. We took advantage of the phenomena that we observed complete pre-catalyst transfer when HydA was in a larger molar excess to HydF. This means that in our FTIR and EPR experiments HydA was in molar excess (6.5:1 molar excess of HydA to HydF) in order to force the precatalyst from HydF to HydA. We decided to use FTIR spectroscopy first. The reason for this was that with FTIR spectroscopy there are no “silent” H-cluster states.

In our model where we suspect that the [4Fe4S] cluster of HydF is the electron acceptor and the [2Fe] subsite of the H-cluster is the electron donor, we should identify at least three paramagnetic species by EPR, which would be a rhombic Hox signal and the axial Hox-CO signal from the H-cluster in HydA1, and an axial signal from the reduced [4Fe4S] cluster from the HydF side. Further- more we would expect peaks corresponding to Hox and Hox-CO species in the FTIR spectrum.

In the FTIR spectrum of the HydA-HydF mixture we can identify much sharper peaks compared to the FTIR spectrum of holo-HydF (Figure 20). This is the first sign that the maturation was successful. The carbonyl and cyanide peaks of HydF are broader than the carbonyl and cyanide peaks of HydA. This is mainly attributable to the fact that the carbonyl and cyanide ligands of the pre-catalyst on HydF have much more vibrational freedom on HydF, due to the fact that the pre-catalyst on HydF is localized on the surface in contrast to HydA, where the [2Fe] subsite is buried within the protein.

The carbonyl and cyanide vibrations of the H-cluster in the FTIR spectrum are in good agreement with previously published wavenumbers of the Hox and Hox-CO state. This was a clear evidence that the electron donor of the oxidation event is the [2Fe] subsite or the pre-catalyst. [56]

Figure 19. H-cluster assembly monitored by FTIR and low temperature CW-EPR spectroscopy. (A) FTIR spectrum recorded [2Fe]adt-CaHydF (B) FTIR spectrum adt recorded on a mixture of apo-HydA1 and [2Fe] -CaHydF. Color coding: Hox, red; Hox-CO, purple. (C) low temperature CW-EPR spectra. Black line: apoHydA1 (200 µM) incubated with [2Fe]adt-CaHydF (30 µM) for 60 min; Red line: [2Fe]adt- CaHydF (30 µM); Purple line: apo-HydA1 (200 µM). (Inset): Black line: apo-HydA1 (200 µM) treated with [2Fe]adt-CaHydF (30 µM); Green line: a linear combination of [4Fe4S]+-CaHydF (16 µM) and apo-HydA1 (200 µM). All EPR samples were prepared in 100 mM Tris-HCl pH 7.0 300 mM KCl, and recorded at a microwave power of 1 mW, at 10 K, microwave frequency: 9.28 GH

The FTIR experiment was followed by a low temperature CW-EPR experiment in order to identify the acceptor site of the redox event. In the low temperature CW-EPR experiment we used indentical conditions to the FTIR experiment.

In the low temperature CW-EPR spectrum of the HydF-HydA mixture we can identify four species. These species are the rhombic Hox signal, the axial Hox- CO signal and two additional axial EPR signals, attributable to reduced [4Fe4S] clusters. The Hox-CO signal is clearly dominating compared to the the Hox signal, which is significantly smaller.

To investigate the reduced [4Fe4S] cluster signals, we have done linear combinations of experimental data of reduced [4Fe4S] cluster signals of pure apo-HydA1 and holo-HydF. By looking into the reference spectra of the pure proteins, the feature at g=1.89 was indentified in the spectrum of the reduced unmatured HydA, while the g=1.87 feature belong to the reduced holo-HydF. We also examined the starting material without any additional reductant.

We found that the starting HydA material was partially reduced, and this cluster population remained reduced upon maturation. Meanwhile, the HydF starting material was practically EPR silent. However, the digital linear combination of 16 µM reduced [2Fe]adt-CaHydF experimental spectrum with the HydA starting material resulted in a perfect fit of the final spectrum recorded after mixing the two species (Figure 19).

5.6. Summary and discussion In this chapter we addressed the need for a biochemical and spectroscopic characterization of a semi-synthetic CaHydF protein and its comparison to biologically matured CaHydF proteins, in order to clarify the biological relevance of the semi-synthetic HydF proteins.

Our semi-synthetic HydF protein allows us to reproduce the FTIR spectra previously reported for biologically matured HydF proteins. This finding was suggesting that the semi-synthetic form is identical to the biological ones. This proved the biological relevance of the semi-synthetic HydF proteins.

The release of a CO ligand was identified as part of the HydF mediated maturation. This CO release was observed previously with a maturation using only the synthetic [2Fe]adt cofactor. [43] However, the phenomena we identified with the CO release assay is fascinating. If HydA is in molar excess then the pre-catalyst transfer is as efficient as in the case when HydF is in

50 excess. The fact that we need large excess from HydF or HydA in order to achive maximal pre-catalyst transfer is raising the question: is it possible that there is another maturase enzyme? A chaperon protein which helps HydA to dock to HydF?

Our FTIR and EPR data support the hypothesis that an oxidation event of the [2Fe] subsite happens during H-cluster assembly. Additionally, the observed reduction of CaHydF suggests that this protein acts as an electron acceptor in the process.

51 6. Quaternary structural study of the [FeFe] hydrogenase maturation (Paper IV)

6.1. Motivation and strategy A bioinorganic chemist prefers to focus on cofactor level, meanwhile a biochemist prefers to look into the same reaction on protein level. In the case of the HydF-HydA interaction, we aimed to look into this on both levels.

Our efforts to capture the intermediate HydF-HydA complex with glutaralde- hyde crosslinking, native PAGE and size exclusion chromatography failed. This implied to us, that this protein-protein interaction is transient. To clarify this, we needed a technique which is more sensitive for transient protein-protein interactions than the SEC or PAGE. Our choice was the so called “GEMMA” (Gasphase Electromobility Macromolecule Analysis) technique which is similar to a mass spectrometry method where the proteins are transferred into a volatile buffer like ammonium acetate. The GEMMA technique requires minute quantities of purified proteins and we can see the results within minutes. Despite these great features of the technique, not every protein can be studied via the GEMMA technique.

In this study we used TmHydF (HydF from Thermotoga maritima) and HydA1 (HydA1 from Chlamydomonas reinhardtii) as model proteins.

6.2. The quaternary structure of HydF The quaternary structure of HydF has been a question for a long time. The first biochemical and biophysical characterization study claimed that HydF acts as a monomer in solution. [29] However, since the apo-HydF crystal structure was solved [34], it was clear that HydF cannot be a monomer.

52 As mentioned earlier, the metal free HydF crystal structure has shown that HydF consist of 3 distinct domains (Figure 20), namely the GTPase domain, the dimerization domain and the FeS cluster binding domain. The dimerization domain itself is extremely hydrophobic, thus ensuring that the two HydF monomers forms a stable HydF dimer unit. This hydrophobic interaction is so strong that so far no one has reported isolated monomeric HydF.

Figure 20. The structure of HydF. In the ribbon structure the color coding is: blue – GTPase domain, white – dimerization domain, red – iron-sulfur cluster binding domain. In the surface model blue indicates the more hydrophilic residues and brown the rather hydrophobic residues in the HydF protein. The structures were prepared in Discovery Studio using TmeHydF (pbd: 5kh0) as template.[18] The monomeric unit was created via the manual separation of the monomers in the dimeric units in silico in Discovery studio in order to expose the hydrophobic surface of the dimerization domain.

However, it has been shown previously that even though the majority of the protein is in a dimeric state, there is a small fraction in tetrameric state. In the case of TmHydF, this population corresponds to the first peak of the size exclusion chromatography chromatograms, while the dimeric population consti- tutes the second peak. The nature of the tetrameric state is unclear. On the one hand, the conserved cysteines of the FeS cluster binding domain are able to form disulfide bridges with each other within a HydF monomer and with another cysteine on another HydF monomer. The distance between the two HydF monomers (~40 Å) in the dimeric unit is probably too large for the cysteines to form disulfide bridges within the dimeric unit, hence the cysteines are forming disulfide bridges with the cysteines of another unit. This results in the so called “dimer of dimers”. [34] However, this oligomerization is expected to be prevented by the [4Fe4S] cluster coordination. In the crystal structure of the [4Fe4S] form of the HydF protein (apo-HydF) another form of tetramer was present. This tetramer did not have any obvious contact surfaces, hence it was potentially the result of crystal packing effects. [18]

53 6.2. The GEMMA vs. SEC The metal free HydA1 protein also shows two peaks in the SEC chromatogram just like the metal free TmHydF. The resulting fractions TmHydF and HydA1 of the SEC were collected and used for GEMMA measurements (Figure 21). In GEMMA spectra of metal free TmHydF, the first peak of the SEC mainly consists of tetrameric proteins, while the second peak shows both dimeric and tetrameric forms, which implies that the tetrameric form was formed again after size exclusion chromatography. This behavior implies that this form of the tetramer and the dimer form are not in dynamic equilibrium. Arguably, a dynamic equilibrium would result in the same dimer-tetramer ratio also in the GEMMA spectrum of the first peak, due to the fact that the tetramers would fall apart to dimers in order to keep the ratios. The GEMMA spectrum of the second peak of the HydA1 SEC chromatogram shows that the population is mostly in monomeric form.

Figure 21. Comparison of the SEC and the GEMMA techniques. Panel A: SEC, panel B: GEMMA.

6.3. Spectroscopic characterization The fact that our metal-free proteins did show up in the GEMMA spectrum was promising. This encouraged us to examine the effect of the metal content on the quaternary structure. However, the fact that ammonium acetate was required as buffer concerned us. We performed a basic comparative spectroscopic analysis both in Tris-HCl based buffer, which was the standard buffer in our spectroscopy experiments, and ammonium acetate.

54 Our aim was to see whether the cofactors keep their structural integrity in ammonium acetate, and in this specific case we were interested in both the [4Fe4S] cluster and the pre-catalyst loaded form of HydF (holo-HydF).

In the UV-Vis spectrum (Figure 22, panel A) the black line indicates the reconstituted HydF in Tris-HCl and the dark grey one in ammonium acetate. The reader is now probably assuming that the author forgot to plot the second spectrum, but this is not the case. In the case of the [4Fe4S] proteins, the spectrum in ammonium acetate and Tris-HCl buffers are identical, or rather indistinguishable when they are normalized for absorption at 280 nm.

Figure 22. The spectroscopic characterization of the HydF protein in ammonium acetate. A) UV-Vis spectra: black line apo-TmHydF 50 mM Tris-HCl 150 mM NaCl pH 8.0, grey line apo-TmHydF in 100 mM ammonium acetate, B) Low temperature CW- EPR spectra: top: 200 µl 200 µM apo-TmHydF in 100 mM ammonium acetate, bottom: 200 µl 200 µM apo-TmHydF in 50 mM Tris-HCl 150 mM NaCl pH 8.0. The EPR spectra were recorded at 10 K with 1 mW microwave power. C) FTIR spectra: top: 2 mM holo-TmHydF in 100 mM ammonium acetate, bottom 2 mM holo-TmHydF in 50 mM Tris-HCl 150 mM NaCl pH 8.0.

In short, the low-temperature CW-EPR spectra were recorded at 10 K with 1 mW power on Na-DT reduced [4Fe4S] cluster containing HydF (apo-HydF) proteins using Tris-HCl or ammonium acetate as buffer. In the low temperature CW-EPR spectra there are minor changes in the line shape which can be due to nitrogen coordination by ammonia, but the principal g-values are identical in both buffers and the intensities are relatively similar. In addition, there was no sign of major [4Fe4S] cluster decomposition or change in redox activity in ammonium acetate buffer compared to Tris-HCl. In the FTIR spectrum, the number of peaks, corresponding to Fe-CO and Fe-CN vibrations of the pre-catalyst, as well as their shape and positions, are identical in both buffers. This is an indication that the number of the ligand has not changed, meaning

55 that the pre-catalyst kept its structural integrity. The comparative spectroscopic analysis verified that both the FeS cluster containing apo and the pre- catalyst containing holo-HydF form were qualified for GEMMA analysis.

6.4 The effect of the metal content on the quaternary structure In order to examine the effect of the metal content on the quaternary structure, the metal free proteins were collected after the SEC, and enzymatically reconstituted in Tris-HCl buffer and also loaded with [2Fe]adt complex to generate apo-HydF and holo-HydF. The last step of the sample preparation was the buffer exchange to 100 mM ammonium acetate buffer. The proteins were flash frozen in liquid nitrogen and kept at -80°C until measurement. Prior to measurement the proteins were diluted into anaerobic 20 mM ammonium acetate buffer.

Figure 23 The GEMMA spectra of HydF and HydA as a function of cofactor content. Panel A) bottom: metal free HydA, middle: apo-HydA, top: holo-HydA. Panel B) bottom: metal free HydF, middle apo-HydF, top: holo-HydF.

The quaternary structure of HydA was not perturbed by metal loading. The initial population distribution between monomeric and dimeric form stayed more or less the same in the metal-free sample, the apo-HydA and in the holo- HydA sample. This can be explained by the fact that the [4Fe4S] cluster coordination environment is very buried inside the protein, so the disulfide bridge formation between HydA monomers is practically impossible (Figure 23). In the case of HydF, certain aliquots of the same metal free sample batch exhibited higher oligomeric states. Upon [4Fe4S] cluster coordination and pre-catalyst loading these oligomeric states were decreasingly present which can be explained by the lack of disulfide bridge formation because of [4Fe4S] cluster coordination. In contrast to the [4Fe4S] cluster of HydA, the [4Fe4S]

56 cluster on HydF is on the surface of protein, hence the intramolecular disulfide bridge formation is possible.

6.5. The quaternary structure of the HydF-HydA interaction In order to see what the quaternary structure of the HydF-HydA interaction is, we examined all different combinations of HydA and HydF proteins based on metal loading. In every GEMMA spectrum of this dataset we can identify the three peaks, corresponding to the different quaternary structural forms of HydF and HydA.

Figure 24. The GEMMA spectra of the HydF-HydA mixtures. The text next to the spectra indicates the cofactor content. Column A indicates metal free HydA combined with holo-HydF (top), apo-HydF (middle) and metal free HydF (bottom). Column B indicates apo-HydA combined with holo-HydF (top), apo-HydF (middle) and metal free HydF (bottom). The asterisk indicates the new 150 kDa peak.

A peak at 50 kDa which is corresponding to monomeric HydA, a peak at 100 kDa corresponding to a mixture of dimeric HydF and a small fraction of dimeric HydA, and we can also identify a 200 kDa peak, corresponding to the tetrameric form of HydF. More importantly, the HydF-HydA mixtures revealed a new peak at 150 kDa (Figure 24). This peak is more intense in the case of metal loaded species and practically negligible in the case of the metal free proteins. A possible, but unlikely, explanation would be that a HydA dimer interacts with a HydF monomer based on the molecular masses of the monomeric HydF (45 kDa) and HydA (49 kDa). Instead, the 150 kDa peak is most probably corresponding to a HydF dimer interacting with a HydA monomer. This finding is raising the question: why would a HydA monomer need a HydF dimer for the activation? [75]

57 6.6. Summary and discussion In order to identify the quaternary structure of the HydF-HydA protein complex we used the GEMMA technique. We characterized the metal loaded forms of HydF in ammonium acetate with EPR, UV-Vis and FTIR spectroscopy. Our results show that both the [4Fe4S] cluster and the pre-catalyst is able to keep their structural integrity in ammonium acetate. We performed GEMMA experiments to highlight the effect of the metal content on the quaternary structure. We found that the quaternary structure of HydA is not affected by the metal content, in contrast to HydF where the higher molecular mass species were less present in the metal loaded samples. We suggested that the reason for this is probably disulfide bridge formation. This would be in agreement with crystal structure studies where HydF was found in the crystals in tetrameric form; a dimer of dimer linked together by disulfide bridges. [34] We identified a 150 kDa peak, which is most likely corresponding to a dimeric HydF interacting with a monomeric HydA. This finding is in agreement with previously published HydA activation solution assays, where the HydF dimer fraction was isolated from the tetramer fraction. Both fractions were used for HydA activation. In this assay, only the dimeric fraction activation showed hydrogen evolution activity. [75]

In the author’s opinion, the dimeric HydF interacting with a monomeric HydA appears to be unusual. Why would a monomeric HydA require a dimeric HydF for the activation? Why is HydF a dimer? What is the role of the dimerization domain? What are the roles of the other domains in the interaction with HydA?

58 7. The miniaturized HydF proteins (Paper IV)

7.1. Motivation and strategy In order to elucidate the role of the domains of HydF in the interaction with HydA, and to show the main role of the [4Fe4S] cluster on HydF, we created two mutants. The first mutant (ΔD) is lacking the dimerization domain, hence it is supposed to be a monomeric HydF protein, with only the GTPase domain and the FeS cluster domain. The second mutant (ΔGD) is lacking in both the dimerization domain and the GTPase domain. Previous studies have shown that the GTP hydrolysis has no effect on the HydA activation. [32] Instead, the GTP hydrolysis is critical upstream, in the interaction with HydE and HydG. The GTPase domain in this case acts as a molecular switch, and the GTP hydrolysis is used to dissociate from HydE and HydG, however it has not been shown before if the GTPase domain itself is involved in the interaction with HydA. [36, 76] The importance and the role of the dimerization domain has not been investigated yet.

It is very important to clarify that these HydF variants most likely would not be able to interact with HydE or HydG, and this emphasizes the importance of our study about the authenticity of the semi-synthetic HydF protein. In Chapter 5, we demonstrated that the semi-synthetic HydF proteins are biologically relevant forms. This gives us a very powerful tool to examine the role of the different domains of HydF, as we can now avoid the biological assembly machinery.

7.2. Sample preparation In this experiment series we used TmHydF as full-length control. Our initial idea was to prepare, characterize and use the FeS cluster domain of TmHydF solely. We found in our preliminary experiments that the FeS cluster domain of TmHydF is not soluble or stable, and it was nearly impossible to work with. In order to stabilize it, it was cloned in frame with a maltose binding protein (MBP) solubility tag. This form (ΔGD) was soluble and stable. We have not had these issues with the ΔD mutant which is lacking only the dimerization domain. Both HydF variants were heterologously expressed in Rosetta (II) E. coli cells and aerobically purified.

59 However, while the proteins intended to be in vitro reconstituted, it was visible during the protein purification process that most likely both mutants are able to coordinate a [4Fe4S] cluster, based on the beautiful brown color of the purified protein, which is typical for proteins with [4Fe4S] clusters (Figure 25).

In order to remove the potentially oxidized iron atoms from the protein sample, the proteins were treated with EDTA chelator in the presence of Na-DT under strict anaerobic conditions, followed by size exclusion chromatography and in vitro enzymatic reconstitution following literature protocols. [38, 77]

Figure 25. The purification process of the ΔGD protein. First picture: sonicated cell lysate, second picture: the resulting supernatant of the ultracentrifugation, third picture, left: column loaded with the ultracentrifugation supernatant, right: an empty column for comparison, fourth picture: the concentrated purified ΔGD. The harvest, work-up, and the purification was performed at 4°C, in 100 mM Tris-HCl pH=8.0, 300 mM KCl, 25 mM MgCl2, 10% glycerol containing buffer, under aerobic conditions.

7.3. The spectroscopic characterization The reconstituted proteins were characterized by UV-Vis and low- temperature CW-EPR spectroscopy (Figure 25).

In the UV-Vis spectrum of both variants we can identify the broad features of a typical [4Fe4S] cluster at 400 nm. It can be misleading that the UV-Vis spectrum of the two HydF variant is not identical to each other or to the UV- Vis spectrum of the oxidized WT-TmHydF. This deviation is due to the difference in absorption at 280 nm, which is because the different HydF variants have different extinction coefficients, which leads to a difference in 280 nm and 400 nm ratios. The broad UV-Vis feature of the oxidized HydF variants at 400 nm is disappearing upon chemical reduction with Na-DT. This could mean that the cluster is decomposing during reduction, but the low tempera-

60 ture CW-EPR spectrum proves that the loss of absorbance is due to the formation of the reduced [4Fe4S] clusters. In the low-temperature CW EPR spectra of the HydF variants we can identify the axial EPR signal corresponding to the reduced [4Fe4S] clusters. These EPR signals have identical g-values compared to the low temperature CW-EPR signal of the reduced [4Fe4S] cluster of the WT-TmHydF protein. These results suggests that even though we removed quite significant parts of the parental protein – 25% in the case of ΔD and 75% in the case of ΔGD – we have not significantly perturbed the coordination environment of the [4Fe4S] cluster.

In the next step, the truncated proteins as well as the WT-TmHydF protein, was used to generate holo-HydF. The HydF variants were incubated with 10 fold molar excess of [2Fe]adt complex under dimmed light conditions and 16°C. The reaction was stopped by the separation of the resulting holo-HydF proteins from the excess [2Fe]adt complex via desalting column. The efficiency of the [2Fe]adt loading of the mutants was already visible with UV/Vis spectroscopy due to the increase in absorbance at 350 nm. In the low temperature CW-EPR spectrum of the mutant proteins we have seen that the loading does not affect the EPR signal of the reduced [4Fe4S] clusters. This implies that the pre-catalyst and the [4Fe4S] cluster are weakly coupled. We have seen this exact behavior both with WT-TmHydF and with our CaHydF protein. [54] In order to get unquestionable evidence about the presence of the pre-catalyst on the miniaturized HydF proteins we used FTIR spectroscopy to look into the Fe-CO and Fe-CN vibrations of the pre-catalyst. The Fe-CO and the Fe-CN bands can be clearly identified in both HydF variants, however the FTIR spectra of the miniaturized HydF proteins are not identical to the FTIR spectrum of the pre-catalyst on the parental protein, nor are they identical to the [2Fe]adt complex in solution.

61 Figure 25. Spectroscopic characterization of the truncated HydF proteins. (Panel A-B) UV-Visible spectra of ΔD (A) and ΔGD (B). The panels show the reconstituted ([4Fe-4S]2+) form of the HydF variants (black lines); the [2Fe]adt cofactor loaded ([2Fe]adt-[4Fe-4S]2+) forms (red lines) and the Na-DT reduced ([4Fe-4S]+) forms (dashed lines). The samples were prepared in a buffer containing 100 mM Tris-HCl and 300 mM KCl buffer with a protein concentration of 100 μM (ΔD) and 50 µM (ΔG D). In the case of ΔD and ΔGD, 20 and 10 molar equivalents (0.5 mM) of Na- DT was used as reductants, respectively. (Panel C-D) Low temperature CW-EPR spectrum of the reduced forms of ΔD (C) and ΔDG (D). The panels show both reconstituted [4Fe4S] cluster (black lines) and [2Fe]adt loaded forms (red lines). The EPR spectra were recorded at 10 K, 1 mW microwave power, 10 Gauss modulation amplitude, and 100 kHz modulation frequency. Frequency 9.28 GHz. (Panel E): FTIR spectra of the [2Fe]adt loaded forms. Top spectrum shows HydFΔDG, and the bottom spectrum HydFΔD. The FTIR spectra were recorded at room temperature,

62 and the samples were prepared in a solution containing 100 mM Tris-HCl pH 8.0, 300 mM KCl.

7.4. HydA activation by the miniaturized HydF proteins

The [2Fe]adt loaded forms of the truncated HydF proteins were tested for HydA activation. The truncated [2Fe]adt cofactor loaded forms were mixed with apo- HydA1 and the samples were incubated for 15 minutes, which was followed by reduced methyl viologen addition to initiate the hydrogen production. In this assay the hydrogen evolution activity is proportional to the amount of apo- HydA1 which was activated by the holo-HydF variants.

As discussed in Chapter 5, the results of the hydrogen evolution assay have showed that at least a 10 fold molar excess of holo-HydF protein is required to fully activate apo-HydA1. In this experiment (Figure 26) we demonstrated that the truncated HydF proteins are also able to activate apo-HydA1, although with different efficiencies as compared to the WT holo-HydF protein.

Figure 26. Influence of the different domains of HydF on the transfer of the pre-catalyst to apo-HydA1 (8 nM). Apo-HydA was titrated with 8 – 120 nM of holo-HydF 8- 80 nM holo-HydFΔD; and 8-80 nM holo-HydFΔDG. 15 fold molar excess of [2Fe]adt complex was used as positive control. The maturation reactions were perfomed in 100 mM K-phosphate buffer (pH 6.8), and H2 evolution initiated via addition of dithionite and MV2+ after 15 minutes. The two bars are representing the technical replicates.

63 7.5 Summary and discussion The author prepared 2 HydF mutants, lacking the dimerization domain (ΔD) and lacking both the GTP-ase domain and the dimerization domain (ΔGD).

The spectroscopic characterization of both mutants highlighted that the truncated proteins are able to coordinate a redox-active [4Fe4S] clusters. These [4Fe4S] clusters have EPR and UVVis properties strikingly similar to the [4Fe4S] cluster of the WT-TmHydF.

The results of the FTIR spectroscopy on the holo-HydF variants showed that both HydF variants can coordinate the synthetic [2Fe]adt complex.

The results of the ΔGD mutant are highlighting the importance and the role of the FeS cluster domain in the coordination of the [2Fe]adt complex. This mutant could be used for resolve the structure of the pre-catalyst on HydF. Due to flexibility issues of the GTPase domain the crystallization of HydF was challenging already with the apo protein, and it was impossible to crystallize the holo-HydF protein. [18, 34]

Due to the fact that the ΔD mutant was able to activate apo-HydA1 the role of the dimerization domain remains unclear in the authors oppinion. In Chapter 6 it was demonstrated that a dimeric HydF interacts with a monomeric HydA. Although it is questionable, why? The pre-catalyst synthesis is considered as a biochemically expensive process due to the GTP hydrolysis in order to dissociate from HydE and HydG. [36, 76] Taken into consideration the quaternary structure of HydG and HydE – HydG is a dimer [17] and HydE is a monomer [16] – and the fact that the product (so called “synthon”) of HydG is a mononuclear iron complex with two CO and a CN ligand [25]. Would it be possible that a dimeric HydG interacting with a dimeric HydF and synthon transfer efficiency would be 1 synthon per HydF monomer (Figure 27)? Fol- lowed by synthon transfer and GTP hydrolysis, HydE would dock in – because HydF cannot interact with HydE and HydG simultaneously [36] – and form the bridging ligand between the two mononuclear iron complex?

64 Figure 27. Proposed HydF maturation mechanism. The proteins structures are corresponding to HydG (yellow, PBD: 4WCX), HydE (blue, PBD: 3CIW), HydF (grey, PBD: 5KH0), HydA (pink, PBD: 5LA3). A similar mechanism was proposed by [78] and [25]. The author was taking it into consideration that HydG is a dimer [17] while HydE is a monomer [16]. According to the model, a HydG dimer docks onto a HydF dimer and placing one-one synthon on both [4Fe4S] clusters of the dimeric HydF unit. Upon GTP hydrolysis HydG is dissociating from HydF and a monomeric HydE can dock into the cavity between the two HydF monomers. HydE is creating the azadithiolate bridging ligand, then dissociating upon GTP hydrolysis. The result is one pre-catalyst per dimeric HydF unit.

65 Conclusions and outlook

The aim of this thesis work was to provide a deeper understanding of H-cluster formation and [FeFe] hydrogenase maturation.

As demonstrated and discussed in Chapter 3, in vivo EPR spectroscopy can be a new tool in [FeFe] hydrogenase research. The in vivo artificial maturation, and EPR spectroscopy combined with biochemical assays could be a tool for screening for in silico identified but spectroscopically and biochemically uncharacterized hydrogenases. This method could potentially be a high-through- put screening for promising hydrogenase candidates. The positive hits can be purified and characterized with regards to high oxygen tolerance, heat stabil- ity, organic solvent tolerance, which are requirements for industrial uses of the hydrogenases. These new hydrogenases could be used in combination with solar cells for sun light driven hydrogen production. Work done in our group, and by other labs, have shown that it is possible to combine hydrogenases with semi-conductors for light driven hydrogen production. [79] [80] Furthermore, the [FeFe] hydrogenase and the maturation machinery can be introduced into suitable biotechnological host organisms for biological H2 production or H2 driven CO2 fixation. [81]

As demonstrated and discussed in Chapter 4, our spectroscopic investigations have allowed the formulation of a mechanism of H-cluster assembly. We were not able to capture spectroscopically the electrochemically identified intermediate P. Future efforts should focus on rapid freeze-quench EPR [82] and stopped-flow rapid scan FTIR [73] techniques on ms time scale in order to capture intermediate P.

In Chapter 5 the biological relevance of the semi-synthetic HydF proteins was discussed and demonstrated. This provides a new tool in order to map the interaction surfaces between HydF and HydA by creating HydF mutants lacking certain secondary or tertiary structural motifs or even just aminoacid muta- tions. There is a high chance that these HydF mutants would not be able to interact with HydE and HydG, but our preliminary results (demonstrated and discussed in Chapter 7) with two truncated HydF proteins show that artificial maturation can be a tool for the preparation of miniaturized HydF production.

66 Based on the data discussed in Chapter 6 and 7, in combination with literature data, [78] the author proposed a revised mechanism for how the pre-catalyst is assembled on HydF. This mechanism could be verified with GEMMA measurements on HydF/HydG and HydF/HydE mixtures. Furthermore, in principle mononuclear Fe complexes with CO and CN ligands could be potentially introduced to HydF followed by the addition of a suitable bridging ligand. This maturation system with the mononuclear Fe complexes could provide a model system for the deeper understanding of HydE chemistry.

67 Sammanfattning på svenska

Samhället konsumerar stora mängder icke-förnybara resurser som potentiellt bidrar till global uppvärmning. Denna konsumption kan därav till slut leda till en samhällskollaps, och därför är investering i miljövänliga och förnybara energikällor viktigt. En miljövänlig och polentiellt förnybar energibärare är vätgas, som vid förbränning inte avger koldioxid.

Väte kan produceras via elektrolys, en ej nödvändigtivs miljövänlig process om vi tar i beaktning koldioxidutsläppet som är nödvändigt vid metoden. Al- ternativt kan vätgas produceras som biobränsle av olika typer av mikroorganismer. Dessa mikroorganismer finns i exempelvis sjöar, jorden och haven. Kärnan i dessa mikroorgansimers produktion av vätgas är hydrogenas, ett enzym, som kan göra vätgas av protoner och elektroner. Produktionen av detta enzym är dock komplicerad, och inne i celler är den strikt kontrollerad. Vi kan därför inte bara ta mikroorganismen och koppla dem till en vätgastank för att skapa biovätgas. För att fullt ut utnyttja organismen måste vi förstå enzymet och hur det skapas.

En viktig beståndsdel i dessa enzymer är ett ”hjärta” som finns i deras protein. Det finns många olika typer av hydrogenas-enzymer, med många typer av ”organ”, men gemensamt för dem alla är strukturen på detta hjärta, som i sin tur består av mindre beståndsdelar. Vi kan undersöka dessa med hjälp av olika ”förstoringsglas”.

Denna avhandling är tänkt att bidra till att fördjupa förståelsen för hur dessa hjärtan är uppbyggda. Vårt förstoringsglas är i huvudsak spektroskopin som kan berätta för oss hur hjärtat ser ut. Andra forskningsgrupper har andra för- storingsklass, exempelvis proteinfilmelektrokemi, kan ge information om hur snabbt hjärtat slår.

Förenklat så är hjärtat byggt upp av järnjoner med inslag av koloxid och cy- anid. Som bekant är de senare två giftiga för människor, men även för mikro- organismerna själva. Trots detta finns de alltså i enzymets hjärta. Konstrukt- ionen av hjärtat kräver ett maskineri som kan sköta dess uppbyggnad på ett kontrollerat sätt. Det är intressant att detta hjärta är biologiskt unikt och finns inte i de flesta levande organismer.

68 Kontrollen vid uppbyggnaden är mycket viktig för mikroorganismernas över- levnad, men det gör enzymen svår a att isolera så att vi kan få den i rent kon- centrat. Våra förstoringsglas är känsliga nog för att undersöka hjärtat, men det kräver att vi har stora mängder rent koncetrat av enzymen.

Isolering av enzymen har nyligen blivit enklare med hjälp av en viktig ny tek- nik. En metafor för att förklara den nya tekniken vore att det är som en organ- transplanation. I detta specifika fall kan vi föreställa oss att vi kan tillverka artificiella hjärtan som vi kan sätta in i enzymet. Dessa hjärtan kan inte slå utanför enzymen, men innanför slår de på exakt samma vis som det biologiska, och ser ut på precis samma sätt.

Vi har fokuserat på den första tiden när dessa hjärtan placeras i enzymet och börjar slå. Vi har följt processen med vårt förstoringsglas på olika sätt i ett försök att förstå hur hjärtat är uppbyggt och integrerat i enzymet. Detta arbete har bidragit till en bättre förståelse för denna komplexa process, som i sin tur har lett oss ett steg närmare en effektiv biologisk produktion av vätgas.

69 Acknowledgements

I would like to express my gratitude for many people who contributed to my thesis and to my life in Sweden during my PhD years.

First of all, I would like to say thanks for my collaborators, including Michael Haumann, Stefan Mebs, Sven Stripp, Moritz Senger, Anders Hofer, Haining Tian, Lei Tian and Katka Hola. It was a pleasure to work together with you.

I would like to express my gratitude towards the Swedish Research Council, to the Liljewalchs travel scholarship and to the Royal Swedish Academy for financial support during my PhD.

Thanks to my former supervisors, professors, friends and coworkers who sup- ported me until and even after I came to Sweden. Kornél, András, Duzs Ági, Gábor, Zsuzsi, Csaba, Jolika, Roland, Kis Ági and Attila. Thank you.

I would like to say thank to my co-supervisors, Ann Magnuson and Stenbjörn Styring. Stenbjörn, thank you for your strict and demanding attitude, scientific curiosity and enthusiasm towards everyone’s scientific work in the photosynthesis/BBC group. It was very motivating to show something to you every Monday morning. Ann, thank you for all the mental and scientific support I received from you during these years.

I would like to say big thank for my dear Charlène. (I am kind of crying right now when I am writing these sentences about you :) ) Thank you for being always kind, always supporting. Thank you for our morning coffees and week- end shopping and afternoon complaining sessions. Thank you for being a great friend. I love you Charlène.

For Timofey. Thank you for being with me during the worst times. Thank you for your support. Thank you for our lunches. Thank you for all the great Rus- sian-Hungarian talks. Once we will rule the world again. :) Thank you for your friendship.

To the spinach ladies of the photosynthesis group, Andrea and Nigar (my Mamochka). I had a great time with you in the office, I think we were a great

70 team. I miss you so much. Thanks for all the life, English, FTIR and EPR lessons from you. Thank you for all the great time what we spent together in the office, in Churchill Arms, at the bicycle parking lot and at my backyard.

To the EPR spectroscopists and seniors of the photosynthesis group, Ping, Fikret, Patrícia, Ann (again), Felix and Johannes. Thank you for all the teaching and constructive criticism about my work and my English. I think I improved both.

For Joan B. Broderick and Eric M. Shepard. The visit to your lab was really motivating me to write my thesis and leave “Gustav’s nest” because there is a whole new world outside.

For Claudia. I think both of us learned a lot from each other. Thank youuuuuuu Claudia!

For Alina and Michael. Thank you for being great office mates. I liked a lot to be in the same office with you two. :)

I would like to say thanks for the people, who are like the sunshine on the dark corridor during the winter: Karin, Salauat, and Marco. Thank you for your laughing and insanely positive attitude.

For Starla. Thank you for all the girl talks and your amazing attitude, your radiant positive energy, and your strength as a woman. Thank you for all the time we spent together, and for the time what you spent to organize gender equality seminars with Sigrid.

For Sigrid. Thank you for being curios. Thank you for not giving up on me, and keep being kind. Thank you for appreciating my work and my experience. Thank you for being always kind and smiling.

Astrid. Thank you for being always kind, always positive, never judging. Thank you for our journaling sessions and Pokemon discussions.

Belinda, thanks a lot to be an alfa B (you know the last letters). I really en- joyed our alfa B discussions.

For my twin sister from another mother. Yes, I am talking about you, Nidhi. I wish we could have spent more time together. I will miss you a lot.

For Arwind. Thank you for being cute and always-always curious.

For Mariia. Thank you for your hugs, even when I have not deserved them.

71 For Sergii. Thank you for our discussions during work, about work, instead of work.

For Petko. I know that you can smile, and I am very happy that you smiling to me time to time.

Thanks Henrik Land to be always calm even when I was about to explode from anger. Thank you for your support, smile and positive attitude.

Kelly. Being always cute. Always smiling. Except when we work on week- ends, because that is absolutely not a good thing.

Emily. I have no idea why you like me. I do not like myself. I suppose you are one of those people who can see the beauty behind the beast. Thank you for believe that- I think you believed, because I am not sure that it is exist.

Brian. I think, I am a funny person. I have a lot of nice moments in general, but when we have seen each other on 4th September (2019) and we started to singing to each other simultaneously, I think that was one of the greatest moments of this decade.

I would like express my gratitude for the rest of Hus 7, including the physchem guys (and just to be gender fair, also for the physchem gals), the SMC people, the cyano group and the rest of the BBC group. You created a great work environment.

I would like to say thanks to my family (this is very inclusive, some friends are considered as family), for all the support what I received in the last 29 years. I couldn’t have done it without you. I love you a lot. And because I know that most of you do not speak English, I will write it down in Hungarian as well.

Szeretnék köszönetet mondani a családomnak (Anyu, Apu, Hajnika, Kriszti, Misi, Beni, Rajmi, Dávid, Márk, Zoli, Laci, Laci, Maxi, Zsani, Szilvi, Anna, Rózsi, Eta, Jucus, Joci, Dani, Gábor, Ági, Szasza, Mónika, János, Helga, Melinda) az összes támogatásért amit az elmúlt 29 évben kap- tam. Nélkületek sehol se lennék.

I would like to say thanks for my “sambo”. We met in the worst time probably, but you turned it into a very happy time. Every single day, when I had a bad day at work, I came to home and you were always there for me. I have no words for this, more than I love you and thanks for being a great sambo. I love you Erik Almhagen.

72 And last, but not least, for my dear Gustav. I know that you have been won- dering why your name isn’t on the top of this section, right? I think both of us knows it perfectly well that always-always the last name in the acknowledgements contributed the most for the thesis in one way or another. Plus, you like to be last author now days… so thanks a lot, for everything. Thank you for being a great supervisor. When you think about me, keep it in mind that when you make pancakes, the first pancake is always a disaster.

Figure last – The origin of the thesis.

73 References

1. Lubitz, W., et al., Hydrogenases. Chem Rev, 2014. 114(8): p. 4081-148. 2. Paulette M. Vignais, B.B., Jacques Meyer, Classification and phylogeny of hydrogenases. FEMS Microbiology reviews, 2001. 25: p. 455^501. 3. Lampret, O., et al., The final steps of [FeFe]-hydrogenase maturation. Proc Natl Acad Sci U S A, 2019. 116(32): p. 15802-15810. 4. Mulder, D.W., et al., Stepwise [FeFe]-hydrogenase H-cluster assembly revealed in the structure of HydA(DeltaEFG). Nature, 2010. 465(7295): p. 248-51. 5. Peters, J.W., et al., X-ray Crystal Structure of the Fe-Only Hydrogenase (CpI) from Clostridium pasteurianum to 1.8 Angstrom Resolution. Science, 1998. 282(5395): p. 1853-1858. 6. Silakov, A., et al., 14N HYSCORE investigation of the H-cluster of [FeFe] hydrogenase: evidence for a nitrogen in the dithiol bridge. Physical Chemistry Chemical Physics, 2009. 11(31): p. 6592-6599. 7. Berggren, G., et al., Biomimetic assembly and activation of [FeFe]- hydrogenases. Nature, 2013. 499(7456): p. 66-69. 8. Esselborn, J., et al., A structural view of synthetic cofactor integration into [FeFe]-hydrogenases. Chem Sci, 2016. 7(2): p. 959-968. 9. Morra, S., et al., Atypical effect of temperature tuning on the insertion of the catalytic iron-sulfur center in a recombinant [FeFe]-hydrogenase. Protein Sci, 2015. 24(12): p. 2090-4. 10. Teixeira, V.H., A.M. Baptista, and C.M. Soares, Pathways of H2 toward the active site of [NiFe]-hydrogenase. Biophys J, 2006. 91(6): p. 2035-45. 11. Senger, M., et al., Protonation/reduction dynamics at the [4Fe-4S] cluster of the hydrogen-forming cofactor in [FeFe]-hydrogenases. Phys Chem Chem Phys, 2018. 20(5): p. 3128-3140. 12. Birrell, J.A., et al., Semisynthetic Hydrogenases Propel Biological Energy Research into a New Era. Joule, 2017. 1(1): p. 61-76. 13. Rumpel, S., et al., Direct Detection of the Terminal Hydride Intermediate in [FeFe] Hydrogenase by NMR Spectroscopy. J Am Chem Soc, 2018. 140(11): p. 3863-3866. 14. Bandyopadhyay, S., K. Chandramouli, and M.K. Johnson, Iron-sulfur cluster biosynthesis. Biochem Soc Trans, 2008. 36(Pt 6): p. 1112-9. 15. King, P.W., et al., Functional studies of [FeFe] hydrogenase maturation in an Escherichia coli biosynthetic system. J Bacteriol, 2006. 188(6): p. 2163- 72. 16. Nicolet, Y., et al., X-ray structure of the [FeFe]-hydrogenase maturase HydE from Thermotoga maritima. J Biol Chem, 2008. 283(27): p. 18861-72.

74 17. Dinis, P., et al., X-ray crystallographic and EPR spectroscopic analysis of HydG, a maturase in [FeFe]-hydrogenase H-cluster assembly. Proc Natl Acad Sci U S A, 2015. 112(5): p. 1362-7. 18. Caserta, G., et al., Structural and functional characterization of the hydrogenase-maturation HydF protein. Nat Chem Biol, 2017. 13(7): p. 779- 784. 19. Broderick, J.B., et al., Radical S-adenosylmethionine enzymes. Chem Rev, 2014. 114(8): p. 4229-317. 20. Rubach, J.K., et al., Biochemical characterization of the HydE and HydG iron-only hydrogenase maturation enzymes from Thermatoga maritima. FEBS Lett, 2005. 579(22): p. 5055-60. 21. Pagnier, A., et al., CO and CN- syntheses by [FeFe]-hydrogenase maturase HydG are catalytically differentiated events. Proc Natl Acad Sci U S A, 2016. 113(1): p. 104-9. 22. Nicolet, Y., et al., Crystal structure of HydG from Carboxydothermus hydrogenoformans: a trifunctional [FeFe]-hydrogenase maturase. Chembiochem, 2015. 16(3): p. 397-402. 23. Pilet, E., et al., The role of the maturase HydG in [FeFe]-hydrogenase active site synthesis and assembly. FEBS Lett, 2009. 583(3): p. 506-11. 24. Shepard, E.M., et al., [FeFe]-Hydrogenase Maturation: HydG-Catalyzed Synthesis of Carbon Monoxide. Journal of the American Chemical Society, 2010. 132(27): p. 9247-9249. 25. Suess, D.L., et al., Cysteine as a ligand platform in the biosynthesis of the FeFe hydrogenase H cluster. Proc Natl Acad Sci U S A, 2015. 112(37): p. 11455-60. 26. Rohac, R., et al., Carbon-sulfur bond-forming reaction catalysed by the radical SAM enzyme HydE. Nat Chem, 2016. 8(5): p. 491-500. 27. Betz, J.N., et al., [FeFe]-hydrogenase maturation: insights into the role HydE plays in dithiomethylamine biosynthesis. Biochemistry, 2015. 54(9): p. 1807-18. 28. McGlynn, S.E., et al., HydF as a scaffold protein in [FeFe] hydrogenase H- cluster biosynthesis. FEBS Lett, 2008. 582(15): p. 2183-7. 29. Brazzolotto, X., et al., The [Fe-Fe]-hydrogenase maturation protein HydF from Thermotoga maritima is a GTPase with an iron-sulfur cluster. J Biol Chem, 2006. 281(2): p. 769-74. 30. McGlynn, S.E., et al., In vitro activation of [FeFe] hydrogenase: new insights into hydrogenase maturation. J Biol Inorg Chem, 2007. 12(4): p. 443-7. 31. Czech, I., et al., The [FeFe]-hydrogenase maturase HydF from Clostridium acetobutylicum contains a CO and CN- ligated iron cofactor. FEBS Lett, 2010. 584(3): p. 638-42. 32. Shepard, E.M., et al., Synthesis of the 2Fe subcluster of the [FeFe]- hydrogenase H cluster on the HydF scaffold. Proc Natl Acad Sci U S A, 2010. 107(23): p. 10448-53. 33. Czech, I., et al., The [FeFe]-hydrogenase maturation protein HydF contains a H-cluster like [4Fe4S]-2Fe site. FEBS Lett, 2011. 585(1): p. 225-30. 34. Cendron, L., et al., Crystal structure of HydF scaffold protein provides insights into [FeFe]-hydrogenase maturation. J Biol Chem, 2011. 286(51): p. 43944-50.

75 35. Maso, L., et al., A conformational study of the GTPase domain of [FeFe]- hydrogenase maturation protein HydF by PELDOR spectroscopy. Applied Magnetic Resonance, 2015. 46(4): p. 465-479. 36. Vallese, F., et al., Biochemical analysis of the interactions between the proteins involved in the [FeFe]-hydrogenase maturation process. J Biol Chem, 2012. 287(43): p. 36544-55. 37. Albertini, M., et al., Probing the Solvent Accessibility of the [4Fe-4S] Cluster of the Hydrogenase Maturation Protein HydF from Thermotoga neapolitana by HYSCORE and 3p-ESEEM. J Phys Chem B, 2015. 119(43): p. 13680-9. 38. Berggren, G., et al., An EPR/HYSCORE, Mossbauer, and resonance Raman study of the hydrogenase maturation enzyme HydF: a model for N- coordination to [4Fe-4S] clusters. J Biol Inorg Chem, 2014. 19(1): p. 75-84. 39. Mulder, D.W., et al., Activation of HydA(DeltaEFG) requires a preformed [4Fe-4S] cluster. Biochemistry, 2009. 48(26): p. 6240-8. 40. Scott, A.G., et al., Compositional and structural insights into the nature of the H-cluster precursor on HydF. Dalton Trans, 2018. 47(28): p. 9521-9535. 41. Pullen, S., S. Roy, and S. Ott, [FeFe] Hydrogenase active site model chemistry in a UiO-66 metal-organic framework. Chem Commun (Camb), 2017. 53(37): p. 5227-5230. 42. T. R. Simmons, G.B., Marine Bacchi, Marc Fontecave, Vincent Artero., Mimicking hydrogenases: From biomimetics to artificial enzymes. Coordination Chemistry Reviews, 2014: p. 270-271. 43. Esselborn, J., et al., Spontaneous activation of [FeFe]-hydrogenases by an inorganic [2Fe] active site mimic. Nat Chem Biol, 2013. 9(10): p. 607-609. 44. Siebel, J.F., et al., Hybrid [FeFe]-Hydrogenases with Modified Active Sites Show Remarkable Residual Enzymatic Activity. Biochemistry, 2015. 54(7): p. 1474-1483. 45. Esmieu, C., P. Raleiras, and G. Berggren, From protein engineering to artificial enzymes – biological and biomimetic approaches towards sustainable hydrogen production. Sustainable Energy & Fuels, 2018. 2(4): p. 724-750. 46. Atta, M., Characterization of the gene encoding the [Fe]-hydrogenase from Megasphaera elsdenii. Biochimica et Biophysica Acta (BBA) - Bioenergetics, 1999. Volume 1476(Issue 2): p. 368-371. 47. Caserta, G., et al., Chemical assembly of multiple metal cofactors: The heterologously expressed multidomain [FeFe]-hydrogenase from Megasphaera elsdenii. Biochim Biophys Acta, 2016. 1857(11): p. 1734- 1740. 48. Birrell, J.A., et al., Artificial Maturation of the Highly Active Heterodimeric [FeFe] Hydrogenase fromDesulfovibrio desulfuricansATCC 7757. Israel Journal of Chemistry, 2016. 56(9-10): p. 852-863. 49. Chongdar, N., et al., Unique Spectroscopic Properties of the H-Cluster in a Putative Sensory [FeFe] Hydrogenase. J Am Chem Soc, 2018. 140(3): p. 1057-1068. 50. Khanna, N., et al., In vivo activation of an [FeFe] hydrogenase using synthetic cofactors. Energy & Environmental Science, 2017. 10(7): p. 1563- 1567.

76 51. Wegelius, A., et al., Generation of a functional, semisynthetic [FeFe]- hydrogenase in a photosynthetic microorganism. Energy Environ Sci, 2018. 11(11): p. 3163-3167. 52. Sommer, C., et al., Proton Coupled Electronic Rearrangement within the H- Cluster as an Essential Step in the Catalytic Cycle of [FeFe] Hydrogenases. J Am Chem Soc, 2017. 139(4): p. 1440-1443. 53. Adamska-Venkatesh, A., et al., New redox states observed in [FeFe] hydrogenases reveal redox coupling within the H-cluster. J Am Chem Soc, 2014. 136(32): p. 11339-46. 54. Nemeth, B., et al., Monitoring H-cluster assembly using a semi-synthetic HydF protein. Dalton Trans, 2019. 48(18): p. 5978-5986. 55. Raleiras, P., et al., Turning around the electron flow in an uptake hydrogenase. EPR spectroscopy and in vivo activity of a designed mutant in HupSL from Nostoc punctiforme. Energy & Environmental Science, 2016. 9(2): p. 581-594. 56. Adamska, A., et al., Identification and characterization of the "super- reduced" state of the H-cluster in [FeFe] hydrogenase: a new building block for the catalytic cycle? Angew Chem Int Ed Engl, 2012. 51(46): p. 11458- 62. 57. Kovacs, D., et al., Photophysics of Coumarin and Carbostyril-Sensitized Luminescent Lanthanide Complexes: Implications for Complex Design in Multiplex Detection. J Am Chem Soc, 2017. 139(16): p. 5756-5767. 58. Megarity, C.F., et al., Electrochemical Investigations of the Mechanism of Assembly of the Active-Site H-Cluster of [FeFe]-Hydrogenases. J Am Chem Soc, 2016. 138(46): p. 15227-15233. 59. Kositzki, R., et al., Electronic and molecular structure relations in diiron compounds mimicking the [FeFe]-hydrogenase active site studied by X-ray spectroscopy and quantum chemistry. Dalton Trans, 2017. 46(37): p. 12544- 12557. 60. Noth, J., et al., Lyophilization protects [FeFe]-hydrogenases against O2- induced H-cluster degradation. Sci Rep, 2015. 5: p. 13978. 61. Lambertz, C., et al., Electronic and molecular structures of the active-site H- cluster in [FeFe]-hydrogenase determined by site-selective X-ray spectroscopy and quantum chemical calculations. Chem. Sci., 2014. 5(3): p. 1187-1203. 62. Lambertz, C., et al., O2 reactions at the six-iron active site (H-cluster) in [FeFe]-hydrogenase. J Biol Chem, 2011. 286(47): p. 40614-23. 63. Swanson, K.D., et al., [FeFe]-hydrogenase oxygen inactivation is initiated at the H cluster 2Fe subcluster. J Am Chem Soc, 2015. 137(5): p. 1809-16. 64. Senger, M., et al., Stepwise isotope editing of [FeFe]-hydrogenases exposes cofactor dynamics. Proc Natl Acad Sci U S A, 2016. 113(30): p. 8454-9. 65. Mebs, S., et al., Hydrogen and oxygen trapping at the H-cluster of [FeFe]- hydrogenase revealed by site-selective spectroscopy and QM/MM calculations. Biochimica et Biophysica Acta (BBA) - Bioenergetics, 2018. 1859(1): p. 28-41. 66. Albertini, M., et al., Characterization of the [FeFe]-Hydrogenase Maturation Protein HydF by EPR Techniques: Insights into the Catalytic Mechanism. Topics in Catalysis, 2015. 58(12-13): p. 708-718.

77 67. Kuchenreuther, J.M., et al., High-yield expression of heterologous [FeFe] hydrogenases in Escherichia coli. PLoS One, 2010. 5(11): p. e15491. 68. Kuchenreuther, J.M., et al., Cell-free H-cluster synthesis and [FeFe] hydrogenase activation: all five CO and CN(-) ligands derive from tyrosine. PLoS One, 2011. 6(5): p. e20346. 69. Happe, T., Isolation, characterization and N-terminal amino acid sequence of hydrogenase from the green alga Chlamydomonas reinhardtii. Eurpean Journal of Biochemistry 1993. 214: p. 475-481. 70. Silakov, A., et al., Spectroelectrochemical characterization of the active site of the [FeFe] hydrogenase HydA1 from Chlamydomonas reinhardtii. Biochemistry, 2009. 48(33): p. 7780-6. 71. Kamp, C., et al., Isolation and first EPR characterization of the [FeFe]- hydrogenases from green algae. Biochim Biophys Acta, 2008. 1777(5): p. 410-6. 72. Shepard, E.M., et al., A Redox Active [2Fe-2S] Cluster on the Hydrogenase Maturase HydF. Biochemistry, 2016. 55(25): p. 3514-27. 73. Wang, S., et al., Structural and Kinetic Studies of Intermediates of a Biomimetic Diiron Proton-Reduction Catalyst. Inorg Chem, 2018. 57(2): p. 768-776. 74. Shepard, E.M., et al., [FeFe]-hydrogenase maturation: HydG-catalyzed synthesis of carbon monoxide. J Am Chem Soc, 2010. 132(27): p. 9247-9. 75. Shepard, E.M., et al., Electron Spin Relaxation and Biochemical Characterization of the Hydrogenase Maturase HydF: Insights into [2Fe- 2S] and [4Fe-4S] Cluster Communication and Hydrogenase Activation. Biochemistry, 2017. 56(25): p. 3234-3247. 76. Galazzo, L., et al., Identifying conformational changes with site-directed spin labeling reveals that the GTPase domain of HydF is a molecular switch. Sci Rep, 2017. 7(1): p. 1714. 77. Papini, C., et al., Bioinspired Artificial [FeFe]-Hydrogenase with a Synthetic H-Cluster. ACS Catalysis, 2019. 9(5): p. 4495-4501. 78. Bortolus, M., et al., Overview of the Maturation Machinery of the H-Cluster of [FeFe]-Hydrogenases with a Focus on HydF. Int J Mol Sci, 2018. 19(10). 79. Kornienko, N., et al., Interfacing nature's catalytic machinery with synthetic materials for semi-artificial photosynthesis. Nat Nanotechnol, 2018. 13(10): p. 890-899. 80. Tian, L., et al., Hydrogen evolution by a photoelectrochemical cell based on a Cu2O-ZnO-[FeFe] hydrogenase electrode. Journal of Photochemistry and Photobiology A: Chemistry, 2018. 366: p. 27-33. 81. Chong Liu, B.C.C., Marika Ziesack, Pamela A. Silver, Daniel G. Nocera, Water splitting–biosynthetic system with CO2 reduction efficiencies exceeding photosynthesis. Science 2016. 352(6290): p. 1210-1213. 82. Byer, A.S., et al., Paradigm Shift for Radical S-Adenosyl-l-methionine Reactions: The Organometallic Intermediate Omega Is Central to Catalysis. J Am Chem Soc, 2018. 140(28): p. 8634-8638.

Acta Universitatis Upsaliensis Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1856 Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science and Technology, Uppsala University, is usually a summary of a number of papers. A few copies of the complete dissertation are kept at major Swedish research libraries, while the summary alone is distributed internationally through the series Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology. (Prior to January, 2005, the series was published under the title “Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology”.)

ACTA UNIVERSITATIS UPSALIENSIS Distribution: publications.uu.se UPPSALA urn:nbn:se:uu:diva-393261 2019