EPR Spectroscopic Studies of [FeFe]-Hydrogenase Maturation

Abstract

Proton reduction and H₂ oxidation are key elementary reactions for solar fuel production. Hydrogenases interconvert H⁺ and H₂ with remarkable efficiency and have therefore received much attention in this context. For [FeFe]-hydrogenases, catalysis occurs at a unique cofactor called the H-cluster. In this article, we discuss ways in which EPR spectroscopy has elucidated aspects of the bioassembly of the H-cluster, with a focus on four case studies: EPR spectroscopic identification of a radical en route to the CO and CN⁻ ligands of the H-cluster, tracing ⁵⁷Fe from the maturase HydG into the H-cluster, characterization of the auxiliary Fe–S cluster in HydG, and isotopic labeling of the CN⁻ ligands of HydA for electronic structure studies of its H_ox state. Advances in cell-free maturation protocols have enabled several of these mechanistic studies, and understanding H-cluster maturation may in turn provide insights leading to improvements in hydrogenase production for biotechnological applications.

1 Introduction

Hydrogenases catalyze the interconversion of H⁺ and H₂ and have attracted considerable attention in the context of solar fuel production [1]. The [FeFe]-hydrogenases are highly active [2] and feature a hexa-iron catalytic cofactor known as the H-cluster (Fig. 1), which consists of a [4Fe–4S] subcluster (“[4Fe–4S]_H”) linked via an L-cysteine (“Cys”) residue to a unique di-iron subcluster (“[2Fe]_H”) [3, 4]. In the last decade, the bioassembly of the H-cluster has emerged as an active topic of research that presents fresh challenges in the coordination chemistry, radical chemistry, and biochemistry of Fe–S enzymes [5]. Unraveling this process is motivated partly by the goal of improving hydrogenase production for biotechnological applications, and partly by its exquisite chemical complexity. The objective of this article is to explicate a few examples in which EPR spectroscopy has provided insight into the mechanism of H-cluster maturation, and to show how the goals of understanding both the mechanism of maturation and the properties of the mature H-cluster are intellectually and practically related. We begin with a brief overview of what is known about H-cluster bioassembly and then focus in greater detail on two investigations of the maturase, HydG, and two investigations of the mature hydrogenase, HydA. More extensive presentations of other biochemical, spectroscopic, and structural studies can be found in recent reviews [1, 5–10].

Fig. 1.

a Ball-and-stick representation of the H-cluster from the X-ray crystal structure of Clostridium pasteurianum HydA (pdb code 3C8Y [68]) with the dithiolate bridging ligand taken as 2-azapropane-1,3-dithiolate. b General scheme of the assembly and installation of the [2Fe]_H subcluster. Color codes: orange Fe; yellow S; gray C; blue N; red O. The four proteinaceous Cys ligands to the [4Fe–4S]_H subcluster have been truncated

A central tenet of HydA maturation is that the H-cluster of HydA is assembled in a stepwise manner: the [4Fe–4S]_H subcluster is generated by the cell’s typical Fe-S cluster machinery and the [2Fe]_H subcluster is synthesized and installed by three maturases: HydE, HydF, and HydG (Fig. 1) [11–13]. As such, active HydA is only obtained when it is coexpressed with HydE, HydF, and HydG; when HydA is not coexpressed with the maturases, the resulting H-cluster consists only of the [4Fe–4S]_H subcluster and is called either “HydA^ΔEFG” or “apo-HydA.” Thus, the primary objective in studying HydA bioassembly is to learn how the maturases synthesize and install the [2Fe]_H subcluster; because Fe–S cluster chemistry is central to these processes, the Fe–S cluster content and function of the maturases will be emphasized herein.

HydE is a member of the radical S-adenosyl-L-methionine (SAM) family of enzymes [7] which employs a [4Fe–4S] cluster to coordinate and reduce SAM by one electron to generate the reactive 5′-deoxyadenosyl radical (“5′-dA·”). EPR and UV/visible spectroscopic studies of Thermatoga maritima (Tm) and Clostridium acetobutylicum (Ca) HydE suggest that it coordinates both a SAM-binding [4Fe–4S] cluster and an auxiliary [4Fe–4S] cluster [14, 15]. However, not all of the Cys residues implicated in auxiliary cluster binding are strictly conserved, and mutants lacking these residues have been shown to promote HydA maturation (as measured by hydrogenase activity) [16]. Several X-ray crystal structures of TmHydE have been reported, and each shows the presence of a SAM-binding [4Fe–4S] cluster; however, the auxiliary cluster binding site contains only between zero and three Fe atoms, depending on crystallization conditions [16–18]. As such, the relevance of an auxiliary cluster to the function of HydE remains unclear. Moreover, neither the substrate nor the product of HydE has been identified, though it has been recently shown that thiols stimulate SAM cleavage by HydE [15]. Thus, although it has been broadly suggested that HydE may be responsible for synthesizing the 2-azapropane-1,3-dithiolate (ADT) bridging ligand in the [2Fe]_H subcluster [5–7, 9, 15, 19], there is not yet any direct evidence in support of this hypothesis, and HydE’s role in H-cluster assembly remains a fertile area for further exploration.

The GTPase, HydF, [20, 21] is also an Fe–S enzyme. The accounts of Fe-S cluster composition of HydF have been considerably variable. When not co-expressed with HydE and HydG, HydF has been reported to show EPR spectroscopic properties consistent with the presence of a [4Fe–4S] cluster [22–25] or of both a [4Fe–4S] cluster and a [2Fe–2S] cluster [20, 26]. When coexpressed with HydE and HydG, HydF displays EPR, FTIR, and XAS spectroscopic features broadly consistent with a [2Fe]_H-like subcluster with an additional [4Fe–4S] cluster [20, 27, 28]. It was also shown that HydF (not coexpressed with HydE and HydG) can bind synthetic models of the [2Fe]_H subcluster in a mode that is similar (or identical) to that of the [2Fe]_H subcluster precursor in HydF that is coexpressed with HydE and HydG (and therefore contains the native [2Fe]_H subcluster precursor) [29]. In spite of the considerable progress that has been made toward elucidating the role of HydF in H-cluster maturation, most of the precise chemical details about the Fe-S cluster conversions that occur on HydF remain up for debate and invite further experimentation.

HydG is a radical SAM enzyme with both a SAM-binding [4Fe–4S] cluster and an auxiliary Fe–S cluster [14, 30–33]. Although the complete reaction promoted by HydG remains unknown, it has been shown that it uses SAM, L-tyrosine (“Tyr”), and Fe to generate an organometallic complex with minimum stoichiometry of [Fe(CO)₂(CN)] that eventually comprises the [2Fe]_H subcluster of HydA [34]. This conclusion was reached primarily by consideration of three sets of experiments which showed: (1) that the CO and CN⁻ ligands of the [Fe(CO)_x(CN)_y] intermediates observed in HydG are derived from Tyr [34]; (2) that the CO and CN⁻ ligands of the [2Fe]_H subcluster in HydA are derived from Tyr [35]; and (3) that Fe from HydG is incorporated into the [2Fe]_H subcluster of HydA [34]. These studies built on prior discoveries that HydG produces CN⁻ [36] and CO [30] from SAM-mediated cleavage of Tyr [37]. Following this brief introduction, we will now further discuss recent insights into the role of HydG in H-cluster maturation obtained from EPR spectroscopy, starting toward the beginning of the HydG reaction: the 5′-dA·-initiated Cα–Cβ of substrate Tyr.

2 Identification of a Radical in the HydG Reaction

It is generally accepted that the HydG reaction is initiated by reductive cleavage of SAM to give the 5′-dA· which abstracts an H atom from substrate Tyr, thereby inducing Tyr Cα–Cβ bond cleavage. Several pathways can be envisioned for this general reaction (Scheme 1), which we will outline here. The first consideration is the site of initial H-atom abstraction. Given that the 5′-dA· abstracts a solvent-exchangeable H-atom [33, 38], the initial site of H-atom abstraction could be either the phenolic H–O atom or the amino H–N atom of Tyr. Many reports and reviews have assumed that the phenolic group is the site of initial H-atom abstraction for both HydG [6, 7, 19, 33, 34, 38, 39] and the closely related Tyr lyase, ThiH [7, 19, 40–42]. The L-tryptophan (“Trp”) lyase, NosL, similarly promotes Cα–Cβ cleavage of its substrate amino acid [43], and a recently reported X-ray crystal structure of Streptomyces actuosus NosL [44] shows the amino group of substrate Trp oriented toward the 5′-C atom of SAM, suggesting that the amino H–N atom is the likely site of initial H-atom abstraction. On this basis, it was proposed that the amino group of Tyr is, by analogy, the site of initial H-atom abstraction for ThiH and HydG [44]. Although neither of the reported X-ray crystal structures of HydG shows bound Tyr, modeling studies of both structures indicate that Tyr could bind with the amino group oriented toward the site of 5′-dAdo· formation [45, 46], consistent with the proposal based on the NosL structure. Regardless, there are no reports of experimental evidence that distinguishes H–N or H–O atom abstraction for HydG, and therefore both possibilities will be considered here.

Scheme 1.

Pathways for radical-initiated Tyr Cα–Cβ bond cleavage. Protons denoted in parentheses are employed for cases in which the site of H-atom abstraction may undergo subsequent protonation. Only H-atom abstraction from a neutral amine group is shown here; similar pathways may be drawn for H-atom abstraction from an ammonium group

Following H–N/O-atom abstraction, the Cα–Cβ bond may be cleaved heterolytically or homolytically (Scheme 1). Homolytic cleavage of the resultant O-centered Tyr radical (1) would give p-quinone methide (3) and a glycyl radical (4); the latter would presumably react with further H⁺ and e⁻ equivalents to give the observed p-cresol byproduct. Likewise, if the N-centered Tyr radical (2) is cleaved heterolytically, the resulting fragments would be p-quinone methide (3) and, following N-protonation, a glycyl radical (4). On the other hand, heterolytic Cα–Cβ cleavage of the O-centered Tyr radical (1) would give the 4-oxidobenzyl radical anion or its protonated form, the 4-hydroxybenzyl radical (collectively referred to herein as the 4(H)OB·, or 5) as well as dehydroglycine (6). Homolytic cleavage of the Cα–Cβ bond of the N-centered Tyr radical (2) results in the same two products: the 4(H)OB· (5) and dehydroglycine (6).

In an attempt to distinguish between these plausible pathways, the discovery and spectroscopic elucidation of radical intermediates were pursued. When the HydG reaction is freeze-quenched early in its time course (<1 min.), an organic radical may be trapped and observed by EPR spectroscopy [33]. Although the hyperfine splitting is poorly resolved in the X-band continuous wave (“CW”) EPR spectrum, the Q-band CW EPR spectrum shows a rich hyperfine coupling pattern (Fig. 2, trace A); for this reason, further studies aimed at elucidating the identity of the radical were performed at Q-band [33]. When the radical is generated with ²H₇-¹⁵N-Tyr, the corresponding EPR spectrum (Fig. 2, trace B) is dramatically narrower and nearly featureless¹; this immediately suggests that the radical is Tyr-derived and therefore, unsurprisingly, cannot be the highly reactive 5′-dA· (which has thus far eluded direct spectroscopic observation in any enzyme). The EPR spectrum of the HydG radical generated using ¹³Cα-Tyr (Fig. 2, trace C) is essentially identical to that generated from natural abundance Tyr; thus, the glycyl radical (3)—which is expected to have significant spin density on Cα—can be eliminated as the identity of the observed intermediate. On the other hand, deuteration of only the ring H atoms using 2,3,5,6-²H₄-Tyr results in a much simpler EPR spectrum (Fig. 2, trace D), which points to significant spin density on the phenol ring of the intermediate. This observation is consistent with the structures of both the O-centered Tyr radical (1) and the 4(H)OB· (5), and is inconsistent with that of the N-centered Tyr radical (2).

Fig. 2.

Q-band CW EPR spectra of the HydG radical generated with various Tyr isotopologues. Corresponding structural proposals for the radical are shown for clarity. Adapted with permission of the authors from [33]

Oxygen-centered Tyr radicals such as 1 are well known to have significant spin density on the O atom and alternating ring C atoms (at the 1, 3, and 5 positions) [47, 48]. In contrast, the 4(H)OB· (5) is calculated to have significant spin density on the Cβ atom and alternating C atoms in the ring (at the 2, 4, and 6 positions) which would result in substantial spin polariziation of the two Hβ atoms as well as the ring 2- and 6-H atoms [33]. Thus, these two candidate structures may be distinguished by generating the radical using 3,5-²H₂-Tyr: if the O-centered Tyr radical (1) is the correct structure, then the resulting EPR spectrum should be markedly different from that generated using natural abundance Tyr. On the other hand, if the resulting EPR spectrum is little changed from the natural abundance Tyr spectrum, then there must be greater spin density on the 2- and 4-positions than on the 3- and 5-positions, and the 4(H)OB· (5) is the correct structure. Because the observed spectrum of the radical generated using 3,5-²H₂-Tyr (Fig. 2, trace E) is essentially identical to that of the radical derived from natural abundance Tyr (Fig. 2, trace A), it may be deduced that the 4(H)OB· (5) is correct. Further support for this assignment came from the EPR spectrum of the radical generated from β,β-²H-Tyr (Fig. 2, trace F), which is dramatically collapsed compared to that of the natural abundance Tyr spectrum and therefore is consistent with a radical that has significant spin density on Cβ.

The observation of the 4(H)OB· (5) allows for several mechanistic pathways to be eliminated. In particular, the two pathways invoking the glycyl radical (3) may be ruled out in favor of the pathways that invoke the 4(H)OB· (5). This finding is particularly relevant to H-cluster maturation because it indirectly suggests that dehydroglycine (6) is an intermediate en route to the CO and CN⁻ ligands of the H-cluster.

3 Tracing ⁵⁷Fe from HydG to HydA Using ENDOR Spectroscopy

The Cα–Cβ bond cleavage of substrate Tyr is only the beginning of the reaction for HydG, which must further cleave dehydroglycine (6) to give CO and CN⁻. The production of CN⁻ by HydG was first demonstrated by acidification of the HydG reaction mixture, followed by CN⁻ derivatization and quantification using fluorescent HPLC analysis [36], and CO production was monitored by in situ addition of deoxymyoglobin which binds free CO to give characteristic UV/visible and IR absorption bands [30]. Subsequent SF-FTIR spectroscopic studies showed the buildup of organometallic [Fe(CO)_x(CN)_y] intermediates, suggesting that the product of the HydG reaction might in fact be an organometallic complex rather than free CO and CN⁻ [34].

This hypothesis was tested using a previously developed in vitro maturation protocol [35, 49, 50]. First, samples of fully matured and uniformly ⁵⁷Fe-labeled Chlamydomonas reinhardtii (Cr) HydA1 (“HydA1^57Fe”) were generated by coexpression with the Shewanella oneidensis (So) maturases (HydE, HydF, and HydG) in growth medium that contained ⁵⁷Fe. In a separate experiment, the maturases were individually expressed, with only HydG grown on ⁵⁷Fe-containing medium (“HydG^57Fe”). Then, natural-abundance apo-HydA1 was matured using HydE, HydF, and HydG^57Fe with no additional Fe, and the resulting “HydA1^57Fe-HydG” was isolated for EPR spectroscopic analysis in the S = ½ H_ox state.

The X-band CW EPR spectra of the H_ox states of the HydA1^57Fe and HydA1^57Fe-HydG samples are broadened in comparison with that of natural-abundance HydA1, indicating some degree of ⁵⁷Fe incorporation [34]. More detailed information was gleaned from the ⁵⁷Fe electron-nuclear double resonance (“ENDOR”) spectra, which were recorded at g₁ = 2.10 where the ENDOR peaks are sharp due to strong orientation selection. The HydA1^57Fe ENDOR spectrum displays several features, consistent with ⁵⁷Fe-labeling of multiple Fe centers in H_ox (Fig. 3, trace A). The HydA1^57Fe-HydG ENDOR spectrum is much simpler (Fig. 3, trace B), containing only a doublet with peak positions at ν_± = ½·A(⁵⁷Fe) ± ν_L as expected for a strongly coupled ⁵⁷Fe nucleus (i.e. A(⁵⁷Fe) >2·ν_L) where A(⁵⁷Fe) and ν_L are the effective hyperfine coupling tensor component and nuclear Larmor frequency, respectively (at a given field position). Thus, the doublet observed in the HydA1^57Fe-HydG spectrum corresponds to A(⁵⁷Fe) = 16.0 MHz and can be ascribed to one of the Fe centers in the [2Fe]_H subcluster (or potentially to both Fe centers if the ⁵⁷Fe hyperfine coupling is the same for each ⁵⁷Fe nucleus at this field position). The A(⁵⁷Fe) = 16.0 MHz doublet in the HydA1^57Fe-HydG spectrum cannot correspond to the [4Fe–4S]_H subcluster because the latter is fully formed in apo-HydA1 and is therefore composed of natural abundance Fe. Consistent with this interpretation is that subtraction of the HydA1^57Fe-HydG spectrum from the HydA1^57Fe spectrum yields a clean doublet with A(⁵⁷Fe) = 10.55 MHz (Fig. 3, trace C) which can be ascribed to the [4Fe–4S]_H cluster. The assignments of the larger and smaller A(⁵⁷Fe) hyperfine tensors to the [2Fe]_H and [4Fe–4S]_H subclusters, respectively, are in line with previous ENDOR and Mössbauer spectroscopic work on the H_ox states of the Desulfovibrio vulgaris and Clostridium pasteurianum [FeFe]-hydrogenases [51–55], though somewhat different from an ENDOR analysis of H_ox from Desulfovibrio desulfuricans which suggested A(⁵⁷Fe) hyperfine tensors of similar magnitude for the two subclusters [56]. Regardless, using ⁵⁷Fe ENDOR spectroscopy largely as an analytic tool, these studies demonstrate that Fe from HydG is incorporated into the [2Fe]_H subcluster of HydA1.

Fig. 3.

Q-band Davies ENDOR spectra of HydA1^57Fe (A), HydA1^57Fe-HydG (B), and the difference spectrum (A – B), each recorded at g₁ = 2.10. Data (black); simulations using hyperfine values from the text (red). Adapted with permissions of the authors from [34]

When viewed in light of the finding that [Fe(CO)_x(CN)_y] intermediates build up in the HydG reaction [34] and that diatomic ligands in these intermediates and in the [2Fe]_H subcluster are derived from Tyr [34, 35], these results suggest that the product of the HydG reaction is an organometallic complex that is eventually incorporated into apo-HydA. The transfer of [Fe(CO)_x(CN)_y] species from HydG to apo-HydA is likely mediated by HydF (though this remains to be directly demonstrated), whereby HydF may serve as an acceptor for the organometallic complex produced by HydG rather than as a [2Fe–2S]-cluster-containing scaffold that is decorated by CO and CN⁻ ligands produced by HydG [20, 25].

4 Characterization of the Auxiliary Cluster in HydG

Although SF-FTIR spectroscopic studies established that [Fe(CO)_x(CN)_y] intermediates build up in HydG, most of the structural details of these intermediates and the mechanisms by which they are formed and released are still to be determined. EPR spectroscopic evidence for the presence of an auxiliary [4Fe–4S] cluster has been reported for TmHydG [14], Clostridium acetobutylicum (Ca) HydG [30, 32], and SoHydG [33]; it has also been suggested that CaHydG may not bind an auxiliary cluster whatsoever [27]. Operating under the assumption that an auxiliary [4Fe–4S] cluster is the site of [Fe(CO)_x(CN)_y] synthon formation, a simple model was proposed in which the non-Cys-ligated Fe of the auxiliary [4Fe–4S] cluster could bind CO and CN⁻ before the synthon is released to its acceptor [34]. In this manner, the auxiliary cluster could cycle between [4Fe]- and [3Fe]-cluster forms during HydG turnover [34, 57].

An important revision to this mechanism was enabled by an X-ray crystal structure of TiHydG [46], which features two monomers per asymmetric unit: one contains the auxiliary cluster in a [4Fe–4S] form while the other contains an unusual [5Fe–5S] form, consisting of a traditional cuboidal [4Fe–4S] cluster linked to a dangler Fe by a bridging μ₂-sulfide. The dangler Fe (present in 73 % occupancy) is further bound to the protein via a highly conserved His ligand positioned trans to the bridging μ₂-sulfide.

The solution EPR spectroscopic features of SoHydG are consistent with the crystallographically observed 5Fe structure of TiHydG [46]. The X-band CW EPR spectrum of dithionite-reduced SoHydG shows two major signals with very distinct properties (Fig. 4, panels A and B, top traces). One signal follows typical temperature- and power-dependence for a [4Fe–4S] cluster and has been shown to coordinate SAM [33] (this has also been studied in CaHydG [30, 32]). The other signal is broad and its major resonances occur at low fields, with g_eff-values of 9.5, 4.7, 4.1, and 3.8. The positions of these resonances are indicative of an S = ⁵/₂ spin system with a rhombicity of E/D = 0.255 as predicted with the aid of a rhombogram [58] or spectral simulation [59]. In general, identification of the spin state of an Fe–S cluster with non-integer spin and S > ½ is quite simple using X-band CW EPR spectroscopy because at X-band frequencies, the electron Zeeman term of the spin Hamiltonian is typically much smaller than the zero-field splitting term [58]. Thus, each populated m_S level gives rise to its own EPR signal with g_eff values that depend, to a first approximation, only on S and E/D. For the S = ⁵/₂ signal in HydG, the resonance at g_eff = 9.5 arises from the m_S = ± ½ spin manifold, while the resonances at g_eff = 4.7, 4.1, and 3.8, arise form the m_S = ± ³/₂ spin manifold. Because the energy difference between these two manifolds is defined as 2D and the relative populations of these spin manifolds are dictated by the Boltzmann distribution, recording and simulating the EPR spectra over a range of temperatures allowed for D to be estimated as +4.5 cm⁻¹.

Fig. 4.

EPR studies of the Fe–S cluster composition of dithionite-reduced SoHydG. a and b X-band CW EPR spectra recorded at 126 μW (a) or 5.00 mW (b). Top traces spectra of samples with no CN⁻ added (black); simulation of the [5Fe] form of the auxiliary cluster with g_eff values indicated (red). Middle traces spectra of samples with K¹³CN added (black). Bottom traces spectra of samples desalted three times in the presence of K¹³CN as described in the text (black) with the g₁ values of the three principle spectral components indicated; simulation of the ¹³CN-bound [4Fe] form of the auxiliary cluster (blue). c Simplified exchange coupling scheme for the auxiliary cluster of HydG and the proposed structure of the auxiliary cluster upon treatment with K¹³CN. d X-band HYSCORE spectrum recorded at g = 2.09 of a sample of SoHydG desalted three times in the presence of K¹³CN. e HYSCORE simulation of the ¹³CN hyperfine interaction (blue contours) with the experimental data from (d) reproduced for clarity (gray contours). Reproduced with permission of the authors from [46]

The S = ⁵/₂ spin system observed by EPR spectroscopy may be understood as resulting from exchange coupling between a reduced, S = ½ [4Fe–4S]⁺ cluster and a high-spin S = 2 dangler Fe²⁺ center. Nonetheless, it was important to demonstrate that this EPR signal does in fact arise from a [5Fe] cluster with the aforementioned coupling scheme, rather than from a [4Fe–4S] cluster with a highly unusual S = ⁵/₂ ground state [60, 61]. To this end, chemical means for removing the dangler Fe to give an S = ½ [4Fe–4S] cluster were pursued [46]. Addition of 20 mM KCN to SoHydG results in partial conversion of the S = ⁵/₂ cluster signal to new S = ½ signals (Fig. 4, panels A and B, middle traces). When SoHydG is subjected to three treatments of first diluting with buffer containing KCN, SAM, and sodium dithionite (DTH), and then concentrating through a centrifugal spin concentration filter, nearly quantitative conversion of the S = ⁵/₂ signals is achieved (Fig. 4, panels A and B, bottom traces). This process effectively desalts the solution while maintaining constant KCN, SAM, and DTH concentrations, and thereby perturbs the equilibrium set up by KCN addition by driving dangler Fe removal. Further evidence for dangler Fe removal upon addition of exogenous KCN was provided by the ¹³C hyperfine tensor for ¹³CN⁻ bound to the [4Fe–4S] cluster species with g = [2.09, 1.94, 1.93]. X-band hyperfine sublevel correlation (“HYSCORE”) spectra recorded at multiple field positions were well-simulated using A(¹³C) = [−5.0, −4.0, 0.9] MHz with Euler angles of [−90°, −40°, 0°] (Fig. 4, panels D and E), parameters that are highly reminiscent of those reported for Pyrococcus furiousus ferrodoxin with ¹³CN bound [62], thereby giving further credence to this structural assignment.

Overall, these structural and spectroscopic studies of the auxiliary cluster of HydG indirectly implicate the dangler Fe as the site of [Fe(CO)_x(CN)_y] synthon formation [46]. Thus, a mechanism involving a [5Fe]↔[4Fe] (or possibly a [5Fe]↔[3Fe]) cluster conversion is likely operative during HydG turnover, the details of which remain to be elucidated. In addition, it is important to note that although a dangler-Fe containing cluster was observed crystallographically for TiHydG and spectroscopically for SoHydG, no auxiliary cluster whatsoever was observed in the X-ray crystal structure of Carboxydothermus hydrogenoformans HydG [45]. Moreover, S = ⁵/₂ spin systems have not been reported in the EPR data for TmHydG [14] and CaHydG [30, 32]. The reasons for why dangler Fe-containing clusters have not been observed in certain HydG studies are not entirely clear. However, given the likely central role of the dangler Fe in both the HydG reaction and in H-cluster maturation, it will be important to resolve any discrepancies regarding the Fe-S cluster composition of HydG.

5 Incorporation of Magnetic Isotopes into the H-Cluster

The final example we’ve chosen to consider illustrates how protocols developed for H-cluster maturation can be applied toward electronic structure studies of the H-cluster. The H-cluster can be poised in multiple chemical states whose structures and relevance to catalysis are central to understanding the mechanism of H⁺/H₂ interconversion [1]. One such state, H_ox, has received a great deal of attention because it is thought to be the state from which H₂ is released (for H⁺ reduction) or to which H₂ binds (for H₂ oxidation). H_ox has an S = ½ ground state, with two limiting electronic structure descriptions involving an $S=0{[4\text{Fe}-4\mathrm{S}]}_{\mathrm{H}}^{2+}$ subcluster exchange-coupled to either a valence-delocalized S = ½ [Fe^1.5+Fe^1.5+]_H subcluster or a valence-localized S = ½ [Fe²⁺Fe¹⁺]_H subcluster. Discriminating between these (as well as other) electronic structure descriptions of H_ox requires extensive computational work that is calibrated to the available structural and spectroscopic data [63–65], most commonly the ⁵⁷Fe hyperfine coupling tensors as determined from ⁵⁷Fe ENDOR and Mössbauer spectroscopic studies (vide supra) [1, 51–55]. In order to expand the number and variety of hyperfine interactions that can be determined for H_ox—and therefore the parameters by which to refine our electronic structure models—a cell-free maturation protocol [35] was employed in order to isotopically label the CN⁻ ligands of the [2Fe]_H subcluster [66]. In principle, this cell-free maturation technology allows for any atom in the [2Fe]_H subcluster to be labeled with a magnetically active nucleus so long as its small-molecule precursor is known.

The first study toward these ends utilized ¹³CN- and C¹⁵N-labeled H_ox samples generated by performing the cell-free maturation with 2-¹³C-Tyr and ¹⁵N-Tyr, respectively [66]; we will focus here on the EPR properties of the ¹³CN-labeled H_ox sample. A strongly coupled ¹³C hyperfine interaction is readily apparent in the X-band CW EPR spectrum owing to the dramatic splitting of the signal at all three principle g values (Fig. 5). Q-band Davies ENDOR spectra recorded at multiple field positions display features in the “strongly coupled” regime with peaks at ν_± = ½·A(¹³C) ± ν_L (Fig. 5). The high-frequency peaks of these doublets are well separated from other features in the ENDOR spectra, and simulation of their orientation-selected shapes and positions allowed for the hyperfine tensor to be accurately determined (A(¹³C) = [30.9, 23.2, 30.2] MHz). A more weakly coupled ¹³C doublet with peaks at ν_± = ν_L ± ½· A(¹³C) was observed using the Mims ENDOR pulse sequence, and simulation of the orientation-selected data yielded a hyperfine tensor of A(¹³C) = [5.22, 5.24, 4.16] MHz. What is most immediately apparent upon comparison of the two ¹³CN hyperfine tensors is that the magnitudes of their isotropic components (a_iso = ⅓•(A₁ + A₂ + A₃)) are dramatically different—28.1 MHz versus 4.87 MHz—and point to a lopsided spin distribution in H_ox wherein the distal Fe (Fe_d) harbors greater spin density than the proximal Fe (Fe_p). This conclusion is in broad agreement with prior computational studies of H_ox [63, 65] and in contrast with ⁵⁷Fe ENDOR studies on H_ox in Desulfovibrio desulfuricans, which suggest a more even spin distribution [56]. External ligand binding (e.g. CO, H₂ and/or H₂O) at Fe_d has been invoked as a factor that could give rise to a more even spin distribution [1, 65]; further work aimed at a precise structural-electronic description of H_ox is clearly warranted. Nonetheless, this CN⁻ labeling study illustrates how atom-specific control over H-cluster maturation can enable further refinement of the electronic structure descriptions of catalytically relevant states. A complementary approach, wherein apo-HydA is treated with synthetic [2Fe]_H subcluster precursors to give natural and unnatural variants of holo-HydA, was recently employed for determining the ¹³C and ¹⁵N hyperfine coupling tensors of the CN⁻ ligands in an unnatural variant of H_ox [67]; the observed hyperfine tensors were very similar to those determined for the natural H-cluster generated by the cell-free maturation as described above [66].

Fig. 5.

EPR studies of ¹³CN-labeled H_ox. Structural assignments and EPR parameters (center). Features ascribed to the ¹³CN ligands bound to the proximal and distal Fe centers are shown in blue and red, respectively. Orientation-selective Q-band Mims ENDOR data (black traces, left) and simulations (blue traces, left); the sharp peaks that are not simulated here arise from an H_ox–CO impurity. X-band CW EPR spectrum (black trace, right, top) and simulation (red trace, right, top). Orientation-selective Q-band Davies ENDOR data (black traces, right, bottom) and simulations (red traces, right, bottom). Adapted with permission of the authors from [66]

6 Conclusion

The examples highlighted in this article illustrate how EPR spectroscopy has played an important role in elucidating several aspects of the mechanism of H-cluster maturation, from the details of Tyr Cα–Cβ bond cleavage to the migration of Fe into the [2Fe]_H cluster of HydA. Advancements in H-cluster maturation protocols have allowed for the execution of otherwise challenging experiments for studying the mature H-cluster, such as incorporating isotopically labeled CN⁻ into H_ox. Conversely, we anticipate that these and future spectroscopic insights into H-cluster maturation may allow for the development of improved maturation protocols with applications for solar fuel production.