An Artificial [Fe4S4]-Containing Metalloenzyme for the Reduction of CO2 to Hydrocarbons

Iron–sulfur clusters have been reported to catalyze various redox transformations, including the multielectron reduction of CO2 to hydrocarbons. Herein, we report the design and assembly of an artificial [Fe4S4]-containing Fischer–Tropschase relying on the biotin–streptavidin technology. For this purpose, we synthesized a bis-biotinylated [Fe4S4] cofactor with marked aqueous stability and incorporated it in streptavidin. The effect of the second coordination sphere provided by the protein environment was scrutinized by cyclic voltammetry, highlighting the accessibility of the doubly reduced [Fe4S4] cluster. The Fischer–Tropschase activity was improved by chemo-genetic means for the reduction of CO2 to hydrocarbons with up to 14 turnovers.


INTRODUCTION
Iron−sulfur metallocofactors are ubiquitous in nature and involved in some of the earth's most fundamental biological processes. Among them, the cuboidal [Fe 4 S 4 ] cluster is the most common representative and is widely known for its role in mediating electron transfer. However, the catalytic function of [Fe 4 S 4 ] is increasingly recognized. 1−3 While electron transport chains rely on the redox couples [Fe 4 S 4 ] 3+/2+ and [Fe 4 S 4 ] 2+/1+ , the oxidation state [Fe 4 S 4 ] 0 is particularly intriguing for catalytic purposes, as such highly reduced species are capable of activating very inert moieties. For instance, it was reported recently that a radical Sadenosylmethionine (SAM) enzyme of Methanocaldococcus jannaschii catalyzes the coupling of two lipid chains. 4 Thereby, an [Fe 4 S 4 ] 0 cluster activates two sp 3 -carbon centers, ultimately leading to the formation of a C−C bond. Further, Suess and co-workers have investigated the electronic configuration of a synthetic [Fe 4 S 4 ] 0 cluster supported by N-heterocyclic carbene ligands. 5 The binding of a CO ligand to the FeS core induces an intramolecular valence disproportionation, and the CObound Fe site adopts a low-valent Fe 1+ oxidation state. Thereby, the C−O bond exhibits remarkable activation, as evidenced by spectroscopy. However, elucidating the fascinating properties of [Fe 4 S 4 ] 0 clusters is challenged by their pronounced reactivity. Although all-thiolate ligated [Fe 4 S 4 ] 0 clusters have been observed electrochemically since the 1970s, isolating such clusters has not been realized until recently. 6,7 The aforementioned lipid-modifying SAM enzyme, as well as Suess' CO-bound [Fe 4 S 4 ] 0 cluster, relies on a 3:1 sitedifferentiated [Fe 4 S 4 ] cluster, in which the unique iron atom is coordinated to a labile ligand (i.e., histidine/Cl − ) before being replaced by the substrate/CO ligand. The 3:1 site-differ-entiated pattern is found throughout several classes of catalytically active FeS proteins, including isoprenoid synthesis proteins (IspG and IspH), aconitase, (R)-2-hydroxyacyl-CoA dehydratase, and the superfamily of SAM enzymes. 8−11 It has been postulated that these unsaturated forms may be essential for reactivity. 12 The remarkable catalytic properties of FeS clusters are further highlighted by the work of the Ribbe and Hu groups. They reported on the propensity of [Fe 4 S 4 ]-containing metalloproteins to catalyze the reduction of CO and CO 2 to hydrocarbons (alka/enes hereafter). 13−16 Strikingly, they showed that the reaction was also catalyzed by a synthetic [Fe 4 S 4 (SCH 2 CH 2 OH) 4 ] cluster in the presence of strong reducing agents, in either aqueous or organic solvents. 17 These results contrast with most catalytic systems, which rarely lead to the formation of multiple C−C bonds upon reduction of CO 2 . 18−20 Inspired by these results, we speculated that embedding a biotinylated [Fe 4 S 4 ] cluster into a protein environment may enable us to engineer and evolve an artificial metalloenzyme (ArM) for the reduction of CO 2 to alka/enes (Fischer− Tropschase, FTase hereafter). Anchoring a metallocofactor into a protein scaffold provides a well-defined second coordination sphere around the cofactor, thus offering straightforward means of optimizing the catalytic performance by chemical and genetic methods. 21 A scaffold of particular interest is streptavidin (Sav), thanks to its exceptionally high affinity for biotin. 22,23 In the past 20 years, Sav has proven to be a privileged host for incorporating various biotinylated cofactors. The resulting ArMs were chemo-genetically optimized to catalyze various reactions including metathesis, C−H activation, hydroamination, hydrogenation, and hydrogen production. 24−31 Other versatile host proteins that have been used for the generation of ArMs include carbonic anhydrase, hemoproteins, prolyl oligopeptidase, helical bundles, the lactococcal multiresistance regulator, and de novo designed metallopeptides. 32 6,47 We speculated that an incoming substrate might displace one of the thiols of the bidentate ligand. However, as the thiol remains in proximity of the Fe center during and after substrate turnover, the decomposition of the cluster by aquation may be minimized.
The homotetrameric structure of Sav can be described as a dimer-of-dimers, with each dimer consisting of two biotinbinding sites facing each other. Capitalizing on this feature, we hypothesized that it might be possible to coordinate the [Fe 4 S 4 ] cluster with two biotinylated 3,5-bis(mercaptomethyl)benzene ligands�and thus four thiolate donors�to firmly anchor the cofactor within the biotin-binding vestibule. Relying on QM-MM calculations, we selected glycine as a spacer to enforce the coordination of the cofactor to two adjacent biotin-binding sites and thus minimize the crosslinking of Sav to afford oligomers. The modeled structure of [(Biot-gly) 2 Fe 4 S 4 ]·Sav WT is displayed in Figure 1. Computational details are collected in the Supporting Information.

Synthesis of [(Biot-gly) 2 Fe 4 S 4 ].
We set out to synthesize a biotinylated ligand bearing a glycine spacer between the biotin anchor and the 3,5-bis(mercaptomethyl)aniline, Scheme 1. Esterification of the m-dicarboxylate 1 was followed by LiAlH 4 reduction to afford the corresponding diol 2. Next, N-Boc protection of the aniline and nucleophilic substitution of the benzylic alcohol yielded the bis(thioester) 3. After N-Boc deprotection, the glycine spacer was introduced using HATU as a coupling agent. Finally, biotin was introduced using biotin pentafluorophenyl ester to afford the bis(thioester) 4. For purification purposes, the bis(thioester) 4 was converted into the corresponding bis(disulfide) 5, which was isolated in analytically pure form following silica gel chromatography. The addition of excess dithiothreitol yielded the desired ligand (Biot-gly). Having purified and fully characterized the (Biot-gly) by NMR and HRMS, it was reacted with half an equivalent of [(t-BuS) 4 Fe 4 S 4 ] 2− to afford the corresponding bis-biotinylated cluster [(Biotgly) 2 Fe II 2 Fe III 2 S 4 ] 2− and four equivalents of t-BuSH, which were removed in vacuo. 48 The cluster was characterized by UV−vis, HRMS, and paramagnetic 1 H NMR, to confirm the purity of the oxygen-sensitive bis-biotinylated cluster [(Biotgly) 2 Fe 4 S 4 ]; see the Supporting Information for experimental details.

Aqueous Stability of [(Biot-gly) 2 Fe 4 S 4 ].
Typically, [(RS) 4 Fe 4 S 4 ] clusters display limited stability toward water. 49,50 As the thiolate ligands are prone to substitution reactions with coordinating solvents, the addition of excess ligand stabilizes the FeS cores under basic conditions. 51−54 Further, the stability can be improved by using large hydrophilic ligands or adding surfactants. 47,50,55,56 However, in the absence of excess ligand or surfactant, the reported [(RS) 4 Fe 4 S 4 ] clusters are only stable up to a water content of around 40%. We hypothesized that the chelating nature of the dithiolate ligand (Biot-gly) might minimize aquation, thus increasing the stability of [(Biot-gly) 2 Fe 4 S 4 ] in water. Inspired by a publication by Holm and co-workers, we examined the aqueous stability of [(Biot-gly) 2 Fe 4 S 4 ] spectrophotometrically. 50 For comparison, the same experiment was conducted with (NMe 4 ) 2 [(HOCH 2 CH 2 S) 4 Fe 4 S 4 ] (6), which previously had been reported to be stable in partially aqueous solutions. 50 Accordingly, the two clusters were dissolved in mixtures of DMSO and borate buffer (pH 8.2, 0.2 M), and the spectral

Scheme 1. Thirteen-Step Synthesis of the Biotinylated (Biot-gly) and Ligand Exchange with [(t-BuS) 4 Fe 4 S 4 ] 2− to Afford the Corresponding Bis-Biotinylated Cluster [(Biotgly) 2 Fe 4 S 4 ]
changes were monitored over time, Figure 2. As reported, the spectrum of the model cluster [(RS) 4 Fe 4 S 4 ] 6 in DMSO remained unchanged for >18 h. However, in the presence of 40% borate buffer, the spectrum already exhibits changes after 1 h. After 18 h, a significantly elevated baseline and decreased features at 400 nm are observed, indicative of cluster decomposition. 47 At 99% aqueous content, the spectrum is nearly featureless after 15 min. Gratifyingly, [(Biot-gly) 2 Fe 4 S 4 ] proved remarkably stable even in 99% borate buffer: no notable spectral change was apparent for >18 h. To the best of our knowledge, [(Biot-gly) 2 Fe 4 S 4 ] represents the first synthetic [Fe 4 S 4 ] cluster that is stable in an aqueous solution in the absence of excess ligand. 49

Characterization of [(Biot-gly) 2 Fe 4 S 4 ]·Sav.
Next, we set out to investigate the incorporation of [(Biotgly) 2 Fe 4 S 4 ] within Sav. Both circular dichroism spectroscopy (CD) and native mass spectroscopy (HRMS) unambiguously highlight the formation of discrete [(Biot-gly) 2 Fe 4 S 4 ] 2 ·Sav intramolecular assemblies (i.e., two clusters per homotetrameric Sav), rather than the formation of cross-linked, oligomeric [(Biot-gly) 2 Fe 4 S 4 ] n ·Sav m assemblies. As can be appreciated, anchoring of two [(Biot-gly) 2 Fe 4 S 4 ] clusters within homotetrameric Sav WT is unambiguously confirmed by the presence of a peak at 68273.8 m/z in native HRMS experiments ([(Biot-gly) 2 Fe 4 S 4 ] 2 ·Sav WT calculated peak: 68273.6 m/z), Figure 3a. No significant peak at higher m/z was detected, thus supporting the hypothesis that the topology   Figure 4a. 57 The anodic scan after the second reduction reveals a nontrivial behavior: (i) a high current density at −649 mV is observed, and (ii) the anodic peak corresponding to the [Fe 4 S 4 ] 1+/2+ process splits into two peaks, indicative of some structural changes in the [Fe 4 S 4 ] core. However, this does not affect the basic integrity of the cluster as derived from two consecutive CV scans, Figure S10. We hypothesize that a thiolate ligand dissociates from the [Fe 4 S 4 ] core during the second reduction. 58 The redox processes of [(Biot-gly) 2 Fe 4 S 4 ]·Sav are diffusion-controlled. To circumvent this limitation, the FTase was adsorbed on an L-cysteine-modified gold electrode to scrutinize its redox behavior. 59 The corresponding cyclic voltammograms reveal that the [Fe 4 S 4 ] 2+/1+ reduction potential of [(Biot-gly) 2 Fe 4 S 4 ]·Sav shifts by +38 mV to −305 mV upon incorporation into Sav AA, Figure 4b. However, a much larger effect of the protein environment on the [Fe 4 S 4 ] 1+/0 redox couple is observed: the potential shifts by +500 mV to −514 mV, Figure 4b.  (7), whose redox potentials were determined in the presence of excess ligand, Figure 4c. 57 Upon embedding the cofactor [(Biot-gly) 2 Fe 4 S 4 ] into Sav AA, marked differences in the redox behavior are clearly apparent, Figure 4a,b. Inspired by the Fe protein of nitrogenase, we hypothesize that hydrogen bonds between the cluster and close-lying amino acid residues may stabilize the highly reactive [Fe 4 S 4 ] 0 species and shift the potentials of the [Fe 4 S 4 ] 2+/1+ and [Fe 4 S 4 ] 1+/0 redox events closer together. The anodic peak splitting, which was observed during the [Fe 4 S 4 ] 1+/2+ process for the free cofactor after the second reduction, disappeared for [(Biot-gly) 2 Fe 4 S 4 ]·Sav AA. If operative, the ligand dissociation from the [Fe 4 S 4 ] core is potentially minimized by the preorganization of the proteinconfined ligands (Biot-gly), which may favor the rapid thiolate recoordination in the event of ligand (partial) dissociation.
As highlighted in Figure 1, residues 112 and 121 of Sav lie close to the four thiolate ligands of the [Fe 4 S 4 ] core. To investigate the effect of these residues on the redox potential, cyclic voltammograms of four single mutants of [(Biotgly) 2 Fe 4 S 4 ]·Sav were recorded, Figure 4c. The [Fe 4 S 4 ] 2+/1+ redox couple appears to be affected by the S112A mutation (+68 mV), whereas the mutation K121A only had a marginal effect. On the other hand, the potential for the [Fe 4 S 4 ] 1+/0 couple ranges from −637 to −514 mV, whereby the mutations in both positions S112A and K121A resulted in substantial positive shifts. Thus, it appears that increased hydrophobicity around the thiolate ligands of the [Fe 4 S 4 ] 2+ core contributes to 2.6. Fischer−Tropschase Activity. In the presence of CO 2 (1 atm) and Eu(II)-DTPA (E 0 = −1.3 V vs Ag/AgCl at pH 8) 69 as reductant, [(Biot-gly) 2 Fe 4 S 4 ]·Sav AA catalyzes the production of short alkanes and alkenes (C 1 −C 4 ), which were detected by GC-FID and GC-MS. Compared to the free cofactor, the ArM displays improved turnover numbers (TONs), Figure 5a. The addition of equimolar amounts of FeCl 3 , Na 2 S, and (Biot-gly) instead of the assembled [(Biotgly) 2 Fe 4 S 4 ] led to minimal background activity. This strongly suggests that the FTase [(Biot-gly) 2 Fe 4 S 4 ]·Sav AA is the catalyst precursor for the reduction of CO 2 to alka/enes. Nonspecific hydrophobic interactions between [(Biotgly) 2 Fe 4 S 4 ] and BSA or Biot·Sav led to slightly increased FTase activity compared to the free cofactor. However, to achieve maximal TONs, embedding the cofactor in the biotinbinding pocket is essential.
In the absence of CO 2 , some residual FTase activity is detected, presumably due to the reduction of the cosolvent DMF, Figure 5b. Upon relying on 13 CO 3 HNa as a CO 2 source, the detection of 13 C-alka/enes by GC-MS unambiguously confirms that dissolved CO 2 (i.e., HCO 3 − ) is indeed the major C-source of the alka/enes (100% for C 3 and C 4 , 70% for C 2 H 6 , and 30% for C 2 H 4 ), Figure S20.
A general challenge when performing CO 2 reduction in water is the competing production of dihydrogen in the presence of strong reducing agents. 18 The hydrogen evolution during CO 2 reduction with [(Biot-gly) 2 Fe 4 S 4 ] and [(Biotgly) 2 Fe 4 S 4 ]·Sav AA was monitored over time by GC-TCD, Figure 5c. Importantly, the kinetic behavior for both H 2 and alka/enes follows similar trends: the corresponding first-order kinetic fit reveals no catalytic onset, neither for H 2 nor for C n H m production. The consumption of the reducing agent leads to leveling-off of C n H m and H 2 production after 2 h ([(Biot-gly) 2 Fe 4 S 4 ]) and 48 h ([(Biot-gly) 2 Fe 4 S 4 ]·Sav AA). However, the addition of Eu(II)-DTPA restores the FTase activity, Figure S21. This suggests that the nature of the active catalyst is by-and-large maintained beyond the indicated times. Further, [(Biot-gly) 2 Fe 4 S 4 ]·Sav AA reacts much more slowly than the free [(Biot-gly) 2 Fe 4 S 4 ] cofactor, despite overall higher TON. As observed in the cyclic voltammogram, the free cofactor exhibited structural changes (i.e., potential ligand dissociation) on the CV time scale. If such an event is essential for substrate binding, this might explain the increased CO 2 fixation rate (albeit at the cost of a reduced TON). In the presence of Biot·Sav (1 equiv Sav AA and excess biotin added), the FTase activity is very similar to that of the free cofactor. This supports the hypothesis that the catalytically active species is indeed embedded in Sav during catalysis.
2.7. Chemo-genetic Optimization. Next, we turned to the chemo-genetic optimization of the FTase activity. For chemical optimization purposes, we synthesized and charac-  Table S2). As the C n H m and H 2 production levels off, addition of Eu(II)-DTPA restores the FTase activity, Figure S21. The experiments were performed in triplicate with standard deviation displayed.  Figure S1 for stability assessments, S5 for CD titration, S8 for native HRMS, and S12, S17 for cyclic voltammograms). We evaluated the FTase activity of both [(Biot-gly) 2   We thank Yanne Darile Yaimyse Amassoka Bayanga for initial CD titration studies.