Reversible Alkyl-Group Migration between Iron and Sulfur in [Fe₄S₄] Clusters

Abstract

Synthetic [Fe₄S₄] clusters with Fe–R groups (R = alkyl/benzyl) are shown to release organic radicals on an [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺ redox couple, the same that has been proposed for a radical-generating intermediate in the superfamily of radical S-adenosyl-l-methionine (SAM) enzymes. In attempts to trap the immediate precursor to radical generation, a species in which the alkyl group has migrated from Fe to S is instead isolated. This S-alkylated cluster is a structurally faithful model of intermediates proposed in a variety of functionally diverse S transferase enzymes and features an “[Fe₄S₄]⁺-like” core that exists as a physical mixture of S = ¹/₂ and S = ⁷/₂ states. The latter corresponds to an unusual, valence-localized electronic structure as indicated by distortions in its geometric structure and supported by computational analysis. Fe-to-S alkyl group migration is (electro)chemically reversible, and the preference for Fe vs. S alkylation is dictated by the redox state of the cluster. These findings link the organoiron and organosulfur chemistry of Fe–S clusters and are discussed in the context of metalloenzymes that are proposed to make and break Fe–S and/or C–S bonds during catalysis.

Graphical Abstract

Introduction

Members of the radical S-adenosyl-l-methionine (SAM) superfamily of enzymes¹ perform a wide array of chemical reactions that support diverse cellular functions.^2,3 The common intermediate in these processes is the highly reactive 5′-deoxyadenosyl radical (5′-dAdo•), which is generated from reductive homolysis of the S–C(5′) bond in SAM.² The reducing equivalent is supplied by an [Fe₄S₄]⁺ cluster that binds SAM at one Fe site; the remaining three Fe centers are bound to the protein polypeptide via a highly conserved CX₃CX₂C motif. In addition to supplying an electron, the cluster may play a more intimate role in radical generation. Specifically, recent biochemical and spectroscopic studies have led to the characterization of a precursor to the 5′-dAdo• that is proposed to feature an Fe–C(5′) bond (Scheme 1A).⁴ This intermediate, termed ‘Ω,’ has subsequently been recognized across the radical SAM enzyme family,^5,6 and related species have also been proposed in mechanisms of non-canonical radical SAM enzymes that release the 3-amino-3-carboxypropyl radical instead of the 5′-dAdo•.⁷ Both the structural identities⁸ and the roles of these intermediates in catalysis⁹ remain under investigation.

Scheme 1.

An organometallic intermediate proposed in mechanisms of radical SAM enzymes (A) and functional models thereof (B).

The transient nature of intermediate Ω has prevented its comprehensive characterization and limited our understanding of its formation, electronic structure, and reactivity patterns. To circumvent these challenges, our group has studied synthetic [Fe₄S₄]–R (R = benzyl, alkyl) clusters whose properties and reactivity are more amenable to thorough characterization.^10–12 Although several such clusters have now been prepared and characterized, each features a four-coordinate alkylated Fe site^10–12 (unlike Ω, which is thought to be at least five-coordinate); binding a fifth ligand results in a dramatically weakened Fe–C bond that readily undergoes homolysis to release radicals in an analogous fashion to the chemistry that occurs in radical SAM enzymes (Scheme 1B).¹² Thus, as for intermediate Ω, the high reactivity of this five-coordinate intermediate has hampered its characterization. An additional limitation of these prior studies is that the clusters that undergo Fe–C bond homolysis generate radicals on an [Fe₄S₄]²⁺–R/[Fe₄S₄]⁺ redox couple (Scheme 1B; cluster charge states are tabulated in Table S1) whereas intermediate Ω is proposed to release the 5′-dAdo• on an [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺ redox couple.⁴ As such, two related objectives in our research have been to study clusters in which Fe–C bond homolysis occurs on an [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺ couple and to characterize intermediates featuring a higher-coordinate alkylated Fe center.

In this paper, we first demonstrate that organic radicals can be generated on an [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺ couple. Then, in attempts to characterize [Fe₄S₄]³⁺–alkyl clusters in which the alkylated Fe site has a high coordination number (≥ 5), we reveal an unexpected finding: that in the [Fe₄S₄]³⁺ core charge state, the alkyl group migrates from Fe to S in a two-electron reductive elimination reaction. We fully characterize this novel S-alkylated cluster and demonstrate the reversibility of alkyl-group migration. Lastly, we discuss the consequences of these findings in the context of both radical SAM and S transferase enzymes.

Results

Reaction pathways upon oxidizing [Fe₄S₄]–R clusters to the [Fe₄S₄]³⁺ core charge state

To test if organic radicals can be generated from the [Fe₄S₄]³⁺ state, we first attempted two-electron oxidation of two previously reported [Fe₄S₄]⁺ clusters: (IMes)₃Fe₄S₄–CH₂Ph (1–Bn; IMes = 1,3-dimesitylimidazol-2-ylidene) and (IMes)₃Fe₄S₄–(CH₂)₇CH₃ (1–Octyl), both of which had been shown to release organic radicals on an [Fe₄S₄]²⁺–R/[Fe₄S₄]⁺ couple upon addition of pyridine ligands to their one-electron oxidized, [Fe₄S₄]²⁺ forms (Scheme 1B).¹² Indeed, reaction of 1–Bn with 2 equiv [Cp₂Fe][BAr^F₄] (Ar^F = 3,5-bis(trifluoromethyl)phenyl) in tetrahydrofuran (THF) results in the rapid and near-quantitative generation of bibenzyl (96 ± 5% yield) and [1–THF][BAr^F₄]₂ (98 ± 5% yield) within 1 min at room temperature (Scheme 2A; Fig. S10 and S65 for spectroscopic characterization of [1–THF][BAr^F₄]₂ and structural characterization of [1–Et₂O][B(C₆F₅)₄]₂, an analogue of [1–THF][BAr^F₄]₂). Based on these results, we hypothesized that transiently generated [1–Bn][BAr^F₄]₂ binds THF at the apical Fe site, generating a reactive five-coordinate intermediate [1–Bn(THF)][BAr^F₄]₂ that quickly releases benzyl radical (Scheme 2A). Addition of 2 equiv [Cp₂Fe][BAr^F₄] to 1–Octyl generated only traces (~0.5%) of the radical dimerization product, hexadecane. The differences in reaction outcomes between 1–Bn and 1–Octyl may be due to the greater reactivity of 1° radicals over benzyl radicals (resulting in alternative pathways for free radical termination) and/or unproductive reactions that occur for 1–Octyl prior to Fe–C bond homolysis (vide infra). Regardless, these results demonstrate that organic radicals can be liberated from synthetic clusters in the [Fe₄S₄]³⁺ charge state, as has been proposed for radical SAM enzymes.

Scheme 2.

Radical generation on an [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺ redox couple via an intermediate with a five-coordinate Fe center.

We then sought to construct a more stable analogue of the intermediate that precedes Fe–C bond homolysis: [1–R(THF)]²⁺ (Scheme 2A). Our strategy was to combine the ether and alkyl donors into a chelating ligand, and for these purposes, we used the 3-methoxypropyl (MP) group (Scheme 2B). Treatment of (IMes)₃Fe₄S₄–Cl (1–Cl) with 3-methyoxypropyl magnesium bromide gave (IMes)₃Fe₄S₄–MP (1–MP), and oxidation of 1–MP with 1 equiv cobaltocenium ([Cp₂Co][PF₆] or [Cp₂Co][BAr^F₄]) in o-difluorobenzene (o-DFB) generated the [Fe₄S₄]²⁺ cluster [(IMes)₃Fe₄S₄–MP]⁺ ([1–MP]⁺, anion = [PF₆] or [BAr^F₄], respectively) (Fig. 1A). The solid-state structure of [1–MP][BAr^F₄] was determined from single-crystal X-ray diffraction (XRD) analysis (Fig. 1B) and established that the 3-methoxypropyl group binds as a monodentate ligand, with the ether group remaining unbound. The ¹H NMR resonances for the α- and β-CH₂ protons of the MP group in [1–MP][PF₆] are shifted to 72.5 ppm and −3.4 ppm, respectively, which closely correspond to the chemical shifts for alkyl groups bound to [Fe₄S₄]²⁺ clusters previously reported by our group.^10–12 In contrast, the γ-CH₂ protons (at 3.4 ppm) and -CH₃ protons (at 3.3 ppm) of the MP group display minimal paramagnetic shifting, indicating that they are not closely associated with the cluster. Based on these observations, we conclude that the ether group is not bound to the apical Fe site in solution.

Figure 1.

Synthesis and characterization of 3-methoxylpropyl-ligated clusters. (A) Synthetic scheme. Panels (B) and (C) display thermal ellipsoid plots (50%) of [1–MP]⁺ and [2]²⁺, respectively. Hydrogen atoms, solvent molecules, and anions omitted for clarity. Fe (orange), S (yellow), O (red). N (blue), C (gray). (D) A schematic of the [Fe₄S₄] core of [2]²⁺ showing Fe–S distances in Å. Average Fe–S bond lengths for each site shown in bold. Standard uncertainties for bond distances omitted for clarity.

We next explored further oxidation of [1–MP]⁺. Treatment of [1–MP][PF₆] (or [1–MP][BAr^F₄]) with 1 equiv [Cp₂Fe][PF₆] (or [Cp₂Fe][BAr^F₄]) in o-DFB at −35 °C quickly produced a dark-red solution, the color of which persisted upon warming to room temperature. ¹H NMR analysis of the reaction mixture revealed the generation of a species with unusually broad resonances between −4 and 13 ppm (Fig. S6, S8, S9). The new product slowly decomposes at room temperature (t_1/2 ≈ 3 days in o-DFB; Fig. S11–S13) to a complex mixture of species. Nevertheless, in non-coordinating solvents, the degradation rate was sufficiently slow at −35 °C to grow single crystals by layering an o-DFB solution with fluorobenzene. Surprisingly, instead of obtaining an Fe-alkylated [Fe₄S₄]³⁺ cluster, single-crystal XRD analysis (Fig. 1C) revealed that the ether group of the MP ligand is now bound to Fe, and the alkyl ligand has migrated to an adjacent S site, forming the S-alkylated cluster [2][PF₆]₂. In [2]²⁺, the apical Fe is part of a six-membered ring featuring the methyoxypropane thiolate ligand whose μ₃-thiolato group binds to two additional Fe sites. The observed S–C bond (1.837(2) Å) is unremarkable and falls within the typical range for alkane thiolates bridging two or more Fe centers.^13–15

The conversion of transiently generated [1–MP]²⁺ to [2]²⁺ could occur by multiple mechanisms in which alkyl migration proceeds in a stepwise or a concerted fashion (Scheme 3). Given that solvent binding to [1–Bn]²⁺ rapidly releases benzyl radicals (Scheme 2A), it is reasonable to assume that intramolecular coordination of the ether group in [1–MP]²⁺ would generate the higher-coordinate intermediate (A) that undergoes Fe–C bond homolysis to generate a tethered 3-methoxypropyl radical (MP•) (in B). Recombination of the MP• with S instead of Fe would yield [2]²⁺. In this scenario, the net two-electron reductive migration of the MP group from Fe to S occurs in two, one-electron steps. Alternatively, the process could be concerted, occurring either directly from [1–MP]²⁺ or from intermediate A, thereby circumventing intermediate B.

Scheme 3.

Potential mechanisms of alkyl-group migration.

Because [2]²⁺ is structurally unprecedented in the synthetic [Fe₄S₄] cluster literature, and structurally homologous, S-functionalized [Fe₄S₄] clusters have been proposed as intermediates in a number of biological processes (see discussion),^16–22 we undertook a detailed examination of the physical properties of [2]²⁺ and of the reversibility of the alkyl migration step (vide infra).

Characterization of [2]²⁺

Although [2]²⁺ is generated by oxidation of [1–MP]⁺, its Fe–S cluster core is formally more reduced because Fe-to-S alkyl-group migration is a two-electron reductive process. This is reflected in the 80 K Mössbauer spectra of the series 1–MP, [1–MP][PF₆], and [2][PF₆]₂. Although none of the spectra can be simulated with a unique four-site model (see SI page 21 for further simulation details), the clusters’ average isomer shifts (δ_avg) inform on the average valences of the Fe ions. For 1–MP and [1–MP][PF₆], δ_avg = 0.474 and 0.377 mm s⁻¹ (Fig. 2), respectively, and the magnitude of the shift upon one-electron oxidation (a decrease of approximately 0.1 mm s⁻¹) is consistent with a one-electron oxidation averaged over four Fe sites.²³ In the absence of alkyl-group migration, a similar decrease in δ_avg would be expected upon further one-electron oxidation. Instead, the trend is reversed: δ_avg of [2][PF₆]₂ is 0.504 mm/s (Fig. 2), which is higher than that of [1–MP][PF₆] and indeed similar to that of 1–MP. Thus, the Fe–S core of [2]²⁺ can be considered as “[Fe₄S₄]⁺-like”, consistent with its Fe oxidation states of 3×Fe²⁺ and 1×Fe³⁺ centers.

Figure 2.

Zero-field ⁵⁷Fe Mössbauer analysis. (A) 80 K spectra of 1–MP, [1–MP][PF₆] and [2][PF₆]₂ with δ_avg indicated. Experimental data (black circles) with fits shown as colored traces (see SI page 21 for fitting details). (B) Graphical depiction of δ_avg for 1–MP, [1–MP][PF₆] and [2][PF₆]₂.

X-band EPR spectra of frozen solutions and microcrystalline solids of [2][PF₆]₂ show evidence for non-interconverting S = ¹/₂ and S = ⁷/₂ spin systems whose ratios vary depending on the physical properties of the sample (vide infra). The 15 K spectrum of a sample prepared as a frozen solution in 9:1 o-DFB:toluene (Fig. 3A) reveals a dominant signal with g = [2.00, 1.96, 1.89] (g_avg = 1.95; see Fig. S18 for simulation) that is consistent with an S = ¹/₂ spin system. In addition to this signal, a derivative-like feature is present at g_eff = 5.17 (Fig. 3A) as is an absorption-like feature at g_eff = 12.2; the latter is more evident in spectra of solid samples of [2][PF₆]₂ (Fig. 3B) and in spectra of poorly glassed frozen-solution samples (vide infra; Fig. S19). Comparison with computed rhombograms²⁴ reveals that these two low-field features arise from transitions within an S = ⁷/₂ spin system at an E/D ratio of ~0.10 (where D and E are the axial and transverse zero-field splitting parameters, respectively). Thus, the nearly isotropic signal at g_eff = 5.17 arises from transitions within the m_s = |±³/₂〉 Kramers doublet (g_x, g_y, and g_z), while the feature at g_eff = 12.2 corresponds to g_y of the m_s = |±¹/₂〉 Kramers doublet. Transitions within other Kramers doublets of the S = ⁷/₂ spin state are not visible due to their extreme breadth.²⁴ The broadness of the features arising from the S = ⁷/₂ state (Fig. S20), and the low intensity of the signal arising from the m_s = |±¹/₂〉 Kramers doublet in particular, preclude an accurate determination of D in the current work, however the slower growth of the m_s = |±¹/₂〉 doublet with increasing temperature relative to that of the m_s = |±³/₂〉 doublet indicates that D < 0 (Fig. S21). Note that similar EPR signals have been reported for the [Fe₄S₄] cluster in benzoyl Co-A reductase and attributed to an S = ⁷/₂ spin system.²⁵

Figure 3.

X-band (9.37 GHz) EPR characterization of [2][PF₆]₂. (A) Sample prepared as a frozen solution (9:1 o-DFB:toluene); 15 K; 40 μW. (B) Sample prepared as a crushed solid; 15 K; 160 μW. g-values indicated with arrows.

Whereas the EPR signals arising from the S = ⁷/₂ spin state vary with temperature because of the changing population of m_s states, the temperature-normalized spectra of the S = ¹/₂ signal are nearly superimposable (Fig. S19). This demonstrates that the S = ¹/₂ and S = ⁷/₂ spin states are not interconverting on the EPR timescale and that, instead, samples of [2]²⁺ behave as physical mixtures of S = ¹/₂ and S = ⁷/₂ states. Similar effects have been observed for [Fe₄S₄]⁺ clusters that exist as physical mixtures of S = ¹/₂ and S = ³/₂ states.^26–33 In addition, the ratio of the S = ¹/₂ and S = ⁷/₂ signals can vary from sample to sample and exhibits a strong dependence on the solvent composition. Solvent combinations that give poorer glasses or that less efficiently solubilize [2]²⁺ tend to give a greater proportion of the S = ⁷/₂ signal (Fig. S22). These observations suggest that the S = ⁷/₂ state is favored in the solid state (including in aggregates formed upon freezing frozen solutions) and that the S = ¹/₂ state is favored in glassy, frozen solutions (as high as ~0.7 spins per cluster). Consistent with this proposal, the EPR spectrum of a solid sample of [2][PF₆]₂ features a dominant S = ⁷/₂ signal, although both signals are significantly broadened (Fig. 3B). Taken together, these observations and the computational results discussed below indicate: (i) that the S = ⁷/₂ and S = ¹/₂ states are both accessible for [2]²⁺; (ii) that these electronic isomers do not interconvert in the solid state or in frozen solution but rapidly interconvert in fluid solution at ambient temperature; (iii) that the two spin states correspond to a cluster with the same connectivity (namely, that revealed by XRD analysis) rather than to two different structures in solution (e.g., an S-alkylated and an Fe-alkylated cluster).

The S = ¹/₂ state of [2]²⁺ is typical of [Fe₄S₄]⁺ clusters and arises from antiferromagnetic coupling between an S = ⁹/₂ [2Fe^2.5+] pair (derived from parallel alignment of an Fe³⁺ ion and an Fe²⁺ ion that undergo valence delocalization by the double-exchange mechanism) and an S = 4 [2Fe²⁺] pair comprised of the remaining two spin-aligned Fe²⁺ sites (Fig. 4, left).³⁴ The observed g_avg is < 2 as is common for [Fe₄S₄]⁺ clusters.^34,35 In contrast, the S = ⁷/₂ spin state is more exotic. Although an octet state has been theoretically predicted to exist as one of several ground spin states for [Fe₄Q₄]⁺ clusters (Q = S or Se),³⁵ it has so far only been spectroscopically observed in a few cases^25,36–40 (and, relatedly, for the P-cluster of the Mo nitrogenase)⁴¹ and has not yet been observed for a structurally characterized [Fe₄Q₄]⁺ cluster. The electronic-structure basis for the S = ⁷/₂ state is readily apparent from the crystal structure of [2]²⁺. As shown in Fig. 1D, the average Fe–S bond length at the Fe1 site (that which is not coordinated to the alkylated S site) is short: 2.251(2) Å (0.041(6) Å shorter than the average Fe–S bond length in the [Fe₄S₄]⁺ cluster 1–Bn),¹² which we attribute to the localization of Fe³⁺ at this site. The remaining Fe sites are thereby localized Fe²⁺ centers because they are bound by a comparatively poorer μ₃-thiolato ligand. Antiferromagnetic coupling between the Fe³⁺ center (S₁ = ⁵/₂) and the remaining three Fe²⁺ sites (aligned parallel with total spin S₂₃₄ = 2 × 3 = 6; Fig. 4, right) gives S_tot = S₂₃₄ – S₁ = ⁷/₂, as is observed experimentally. This 3:1 spin-coupling pattern is consistent with the theoretical treatment of [Fe₄S₄]⁺ clusters with an S = ⁷/₂ spin state³⁵ and has been proposed to explain the magnetic properties of [Fe₄Se₄]⁺ and [Fe₄S₄]⁰ clusters (the latter with S_tot = S₂₃₄ – S₁ = 6 – 2 = 4).^38,42

Figure 4.

Spin-coupling patterns for the S = ¹/₂ (left) and S = ⁷/₂ (right) states of [2]²⁺, specifically the BS34 and BS1 determinants, respectively. The bolded Fe–Fe axes represent spin aligned sets of Fe ions.

Valence localization in the S = ⁷/₂ state is corroborated by broken-symmetry density functional theory (BS DFT) calculations. We first evaluated the relative energies of various BS determinants for [2]²⁺ by constraining the atomic coordinates to those observed crystallographically (the Mes groups were truncated to hydrogen, and the H-atom positions were optimized at the BP86/ZORA-def2-TZVP level; single-point energy calculations were performed at the TPSSh/DKH-def2-TZVP level; see SI page 48 for additional information). Among the four S = ⁷/₂ BS determinants, the configuration in which Fe1 is in the minority spin (BS1) is the lowest in energy. Population analysis (see SI page 47) reveals that this corresponds to the electronic structure proposed based on the XRD analysis (vide supra): a localized Fe³⁺ site (Fe1) antiferromagnetically coupled to the remaining Fe²⁺ sites. This BS determinant is also 7.4 kcal/mol more favorable than the lowest-energy S = ¹/₂ BS determinant (see Table S5); the energetic preference for the S = ⁷/₂ state using the crystal structure coordinates further supports our conclusion that the crystal structure represents the geometry of the S = ⁷/₂ state.

Because the EPR spectrum of [2]²⁺ reveals both an S = ⁷/₂ and an S = ¹/₂ signal, we were interested to learn if allowing the Fe–S core to relax in silico would result in the two states being closer in energy. We therefore conducted analogous BS-DFT calculations on partially optimized structures (Mes groups were truncated to hydrogen; Fe–Fe–C–N dihedral angles were fixed; geometry optimizations and energy calculations were performed at TPSSh/DKH-def2-TZVP level; see SI page 44–45 for additional information). The optimized structure using the S = ⁷/₂ BS1 determinant is similar to that observed crystallographically, with short Fe–S distances around Fe1 (2.247 Å on average; Fig. S44). In contrast, the six BS determinants with an S = ¹/₂ spin state (BS12, BS13, BS14, BS23, BS24, BS34) all exhibit tetragonally compressed core structures and pairwise spin coupling (Table S6, Fig. S43, S48–S53 and Fig. 4, left), both of which are hallmarks of [Fe₄S₄]⁺ clusters with S = ¹/₂ ground states.^43–46 In these fully optimized structures, the energies of the S = ¹/₂ BS determinants are still higher than that of the S = ⁷/₂ BS1 determinant, but the gap is smaller than those in the unoptimized structures discussed in the preceding paragraph. Indeed, the S = ¹/₂ BS determinant with the lowest energy (BS34, with minority spin on Fe3 and Fe4) is only 2.7 kcal/mol higher than the S = ⁷/₂ BS1 determinant. Due to the modest difference in computed energies, it is reasonable to expect that both spin states could be observed by EPR spectroscopy and that subtle perturbations from the environment (e.g., solvent interactions, glassing effects, etc.) could change the ratios of the two states, resulting in the occurrence of physical spin-mixtures. In contrast, because of the high homogeneity in a single-crystal lattice, only one state is observed crystallographically (specifically, the S = ⁷/₂ spin state).

Assessing the reversibility of S-alkylation

Given that oxidation of [1–MP]⁺ triggers alkyl-group migration from Fe to S, we were interested to learn if the reverse reaction—alkyl-group migration from S to Fe—would occur upon reduction of [2]²⁺ (Fig. 5). Such a finding would build on prior work that has established the cleavage of S–C bonds at Fe–thiolate complexes,^47–51 though in these examples the organic groups were lost as radicals instead of recombining with the complex to form an Fe–C bond. In the present study, we observed that treatment of [2][PF₆]₂ with 1 equiv Cp₂Co in o-DFB at −35 °C for five minutes resulted in a color change to dark-brown. Examination of the solution by ¹H NMR spectroscopy at room temperature showed complete consumption of [2][PF₆]₂ as evidenced by the disappearance of the characteristic peak at −2.3 ppm, and clean conversion to [1–MP][PF₆] (with characteristic α- and β-CH₂ protons of the MP group at 72.5 and −3.4 ppm, respectively; Fig. 5). Thus, [1–MP]⁺ and [2]²⁺ can be cleanly and reversibly interconverted by one-electron redox processes. Alkyl-group migration from Fe to S is therefore chemically reversible, with the preference for Fe vs. S alkylation depending on the redox state: Fe-alkylation is preferred at an [Fe₄S₄]²⁺ core whereas S-alkylation is preferred at an [Fe₄S₄]³⁺ core (to give an “[Fe₄S₄]⁺-like” core).

Figure 5.

Assessment of the chemical reversibility of redox-coupled interconversions of [1–MP]⁺ and [2]²⁺. ¹H NMR spectra of isolated [1–MP][PF₆] (top), isolated [2][BAr^F₄]₂ (middle), and the product from reduction of [2][PF₆]₂ using Cp₂Co (bottom) recorded in d₄-o-dichlorobenzene at room temperature.

To shed further light on these redox-coupled alkyl-group migration processes, we recorded cyclic voltammograms (CVs) of [1–MP][PF₆] and [2][PF₆]₂ (henceforth denoted as [1–MP]⁺ and [2]²⁺ for simplicity). The CV of [1–MP]⁺ in the potential range of −1.0 V to −2.0 V features a reversible event at −1.54 V that corresponds to the [1–MP]⁺/1–MP redox pair (Fig. S35 and S36; all potentials referenced to Cp₂Fe/[Cp₂Fe]⁺).^10–12 More interestingly, sweeping positively reveals an irreversible wave with a peak potential of ca. –0.3 V that arises from oxidation of [1–MP]⁺ (Fig. 6A). The return sweep displays a new event at ca. −0.9 V that becomes more reversible with increasing scan rate (Fig. 6A) and only appears if the CV is first swept positively past the wave at ca. −0.3 V (Fig. S37). Additionally, the same quasi-reversible feature is observed in the CV of [2]²⁺ when swept in the negative direction (Fig. 6B). We therefore assign this redox event, observed in CVs of both [1–MP]⁺ and [2]²⁺, to the [2]⁺/[2]²⁺ redox couple. As such, the irreversible wave at ca. –0.3 V observed in the positive sweep of [1–MP]⁺ corresponds to the oxidation-induced transformation of [1–MP]⁺ to [2]²⁺. Note that the same irreversible wave at ca. –0.3 V is observed in the CV of [2]²⁺ after sweeping negatively through the [2]⁺/[2]²⁺ redox couple because, under these conditions, in-situ generated [2]⁺ is partially converted to [1–MP]⁺. The sweep-rate dependence in the CV of [2]⁺ (Fig. 6B) is consistent with this interpretation; at slower sweep rates, the return wave of the [2]⁺/[2]²⁺ redox couple (at ca –0.8 V) is diminished, and that of the [1–MP]⁺/[2]²⁺ couple (at ca. –0.3 V) is enhanced because a greater proportion of [2]⁺ has rearranged to [1–MP]⁺. Taken together, these data enable the complete assignment of the CVs of [1–MP]⁺ and [2]²⁺. They also demonstrate that the rearrangement upon reduction of [2]²⁺ to [1–MP]⁺ occurs on the CV timescale, whereas the rearrangement upon oxidation of [1–MP]⁺ to [2]²⁺ is sufficiently fast that the [1–MP]⁺/[1–MP]²⁺ couple cannot be observed (at scan rates up to 2 V sec⁻¹; Fig. S38).

Figure 6.

Cyclic voltammetry experiments and analysis. (A) CV plots of [1–MP]⁺. The slowest scans (25 mV/s) are shown in blue and the fastest scans (500 mV s⁻¹) are shown in red. Other scan rates (50, 100, 150, 200, 300 and 400 mV s⁻¹) are shown in gray. The starting scanning directions and positions are labeled with black arrows. (B) CV plots of [2]²⁺ with the same formatting as in (A). (C) Square scheme for the redox-coupled alkyl-group migration. (D) CV plot of [2]²⁺ recorded at 2.5 mV s⁻¹. All CV experiments were referenced to Cp₂Fe/[Cp₂Fe]⁺ and performed in o-DFB with 0.5 M [Bu₄N][PF₆] supporting electrolyte at 2 mM [cluster] at room temperature.

Having assigned the features in the CVs of [1–MP]⁺ and [2]²⁺, we next considered the thermodynamics of these redox-coupled alkyl-group migrations. The equilibrium constants for the two redox-decoupled rearrangements (K_AC, defined as [C]/[A], and K_BD, defined as [D]/[B]; Fig. 6C) are connected to the redox couples, E₁° and E₂°, corresponding to equations (1) and (2) below:

$${\left[1–MP\right]}^{2+}+{\text{e}}^{-}\leftrightarrows {\left[1–MP\right]}^{+}$$ (1)

$${\left[2\right]}^{2+}+{\text{e}}^{-}\leftrightarrows {\left[2\right]}^{+}$$ (2)

Because the redox interconversion of [2]²⁺ and [2]⁺ is quasi-reversible, a value of E₂° ≈ −0.90 V can be estimated from the mid-peak potentials in the CVs recorded at fast scan rates. In contrast, the irreversibility of the oxidation of [1–MP]⁺ precludes a reliable determination of E₁°. However, E₁° can be estimated from analysis of the CV of 1–Bn, which exhibits a quasi-reversible [1–Bn]⁺/[1–Bn]²⁺ couple when recorded in non-coordinating solvent (o-DFB) with a weakly coordinating electrolyte ([ⁿPr₄N][BAr^F₄]; Fig. S39). Based on this analysis, we estimate E₁° for the [1–MP]⁺/[1–MP]²⁺ couple to be −0.26 V.

The thermodynamic potential for the overall equilibrated system,

$$\left({\left[1–MP\right]}^{2+},{\left[2\right]}^{2+}\right)+{\text{e}}^{-}\leftrightarrows \left({\left[1–MP\right]}^{+},{\left[2\right]}^{+}\right)$$ (3)

is defined as E_K° and can be related^52,53 to the standard potentials and the equilibrium constants, K_AC and K_BD, according to the relationships

$$E_1°=E_{\text{K}}°+RT/F\ln \left(K_{\mathrm{BD}}+1\right)$$ (4)

$$E_2°=E_{\text{K}}°-RT/F\ln \left({K_{\mathrm{AC}}}^{-1}+1\right)$$ (5)

when K_AC << 1 and K_BD >> 1 (or, ΔG_AC >> 0 and ΔG_BD << 0, respectively, as is the case in our system). E_K° can in principle be used to derive the equilibrium constants for the redox-decoupled rearrangements, and its value is typically obtained by reading the center position of a reversible redox pair from a fully equilibrated system observed at sufficiently slow sweep rates.^53,54 However, even at very slow sweep rates (e.g., 2.5 mV sec⁻¹ for the CV of [2]²⁺; Fig. 6D), a fully equilibrated system is not achieved as evidenced by the wide separation and slow convergence of the two redox pairs (Fig. S40, S41). Nevertheless, we can still determine the upper and lower limits of the rearrangement free energies, ΔG_BD and ΔG_AC, using the relationship

$$\text{Δ}{G}_{\text{AB}}+\text{Δ}{G}_{\text{BD}}=\text{Δ}{G}_{\text{AC}}+\text{Δ}{G}_{\text{CD}}$$ (6)

At the limit of ΔG_BD = 0,

$$\text{Δ}{G}_{\text{AC}}\le \text{Δ}{G}_{\text{AB}}-\text{Δ}{G}_{\text{CD}}=-nF{E}_1°+nF{E}_2°=15{\text{kcal mol}}^{-1}$$ (7)

where n = 1 (corresponding to the number of electrons transferred) and F is Faraday’s constant.

Likewise, at the limit of ΔG_AC = 0,

$$\text{Δ}{G}_{\text{BD}}\ge \text{Δ}{G}_{\text{CD}}-\text{Δ}{G}_{\text{AB}}=-nF{E}_2°+nF{E}_1°=-15{\text{kcal mol}}^{-1}$$ (8)

Since neither ΔG_BD nor ΔG_AC is near zero, it is likely that both |ΔG_BD| and |ΔG_AC| are substantially lower than 15 kcal mol⁻¹. Regardless, the foregoing CV analysis provides a range for the free-energy change of alkyl-group migration in both the cationic and dication states, as well as a qualitative indication of the rates of these processes.

Trapping intermediates upon oxidation of clusters featuring non-chelating alkyl groups

Alkyl-group migration in the transformation of [1–MP]⁺ to [2]²⁺ is accompanied by binding of the ether group to the formerly alkylated Fe site, and the thermodynamics of this rearrangement are impacted in part by the stabilization afforded by the formation of the six-membered ring in [2]²⁺ (or, conversely, by destabilization if there is significant ring strain). We were therefore interested to learn if S-alkylated clusters can be generated in the absence of a chelating alkyl group.

We began by reexamining two-electron oxidation of clusters 1–R (where R is the non-chelating benzyl or octyl group; vide supra). In particular, we attempted to trap intermediates via freeze-quench techniques and to analyze the products by EPR spectroscopy. Samples generated by freezing mixtures of 1–Bn and 2 equiv [Cp₂Fe][BAr^F₄] after incubating in THF at −78 °C for ~1 min display no EPR signals; apparently, the Fe–C bond is sufficiently weak that no Fe- or S-alkylated species can be observed under these conditions. This is also consistent with the generation of [1–THF][BAr^F₄]₂ (vide supra), which is diamagnetic and therefore EPR-inactive. In contrast, when 1–Octyl samples are prepared under identical conditions, a new S = ¹/₂ EPR signal is observed (0.13 spins per equiv of starting 1–Octyl; Fig. 7C) with a g-tensor (g = [1.99, 1.96, 1.91]) very similar to that of the S = ¹/₂ signal of isolated [2]²⁺ (g = [2.00, 1.96, 1.89]). Accompanying this spin-doublet signal, a small derivative-like signal was also detectable at g_eff = 5.17 (Fig. 7B), a characteristic g_eff value for an S = ⁷/₂ system with E/D near 0.10 (also akin to that observed for [2]²⁺). Similar S = ¹/₂ signals were observed when conducting the same oxidation in 2-MeTHF or mixtures of THF and 2-MeTHF, and small perturbations to the EPR signals were apparent when the composition of the solvent mixture was changed (see SI page 30–31), indicating that solvent is coordinated to the cluster in the trapped intermediate (as depicted in Fig. 7A, top). Taken together, these EPR results demonstrate that the S-alkylated cluster [3]²⁺ forms in the presence of donor solvents (Fig. 7A, top), and we therefore conclude that the generation of S-alkylated Fe–S clusters does not require a chelating alkyl group.

Figure 7.

Fe- and S-alkylated species trapped in freeze-quench EPR studies. (A) Solvent dependence of reaction outcome in the two-electron oxidation of 1–Octyl. EPR spectra of [2]²⁺, [3]²⁺, and [1–Octyl]²⁺ in the low-field (B) and mid-field (C) regions. Spectra recorded at 15 K, 9.37 GHz, and 40 μW (for [2]²⁺ and [3]²⁺) or 10 μW (for [1–Octyl]²⁺).

Alkyl-group migration to S does, however, require a donor ligand. A markedly different S = ¹/₂ EPR signal (with g = [2.18, 2.03, 1.98]) was observed upon two-electron oxidation of 1–Octyl in the non-coordinating solvent o-DFB, corresponding to 0.36 spins per equiv 1–R (Fig. 7C, S30); a similar EPR signal was observed upon two-electron oxidation of 1-Bn in o-DFB (Fig. S34). Notably, the signals feature g_avg > 2, which is characteristic of [Fe₄S₄]³⁺ clusters^55–60 including Fe-alkylated [Fe₄S₄]³⁺ intermediates proposed in biological systems^4,5,7 and synthetic models thereof.¹¹ We therefore assign the species generated in o-DFB to the clusters [1–Octyl]²⁺ and [1–Bn]²⁺, which feature four-coordinate, alkylated/benzylated Fe sites (Fig. 7A, bottom). Thus, in the absence of a donor solvent, no R group migration occurs due to the absence of a suitable ligand for the Fe site, whereas in coordinating solvents like THF, an S-alkylated dicationic cluster forms with THF bound to the apical Fe site ([3]²⁺; Fig. 7A, top).

Discussion

In this work, we have demonstrated the facile and reversible migration of alkyl groups between Fe and S in synthetic [Fe₄S₄] clusters, and we now relate these findings to mechanistic proposals for [Fe₄S₄] enzymes. First, we note that [2]²⁺ is a high-fidelity structural model for several intermediates in S transferase enzymes. For example, in lipoic acid biosynthesis, LipA catalyzes the insertion of two S atoms into unactivated C–H bonds of a protein-bound n-octanoyl chain;^61–65 in this reaction, the two S atoms derive from an [Fe₄S₄] cluster that is cannibalized during catalysis.^17,19,20 The first S–C bond forms when a 2° carbon radical attacks the [Fe₄S₄] cluster to generate an S-alkylated [Fe₄S₄] cluster that loses Fe to form an S-alkylated [Fe₃S₄] cluster (Fig. 8A). Similar S-methylated [Fe₄S₄] and [Fe₃S₄] intermediates have been proposed in reactions performed by alkylthiotransferases^21,22 such as MiaB and RimO as well as the nitrogenase maturase NifB, which is proposed to generate the central carbide in nitrogenase cofactors via an S-methylated intermediate (Fig. 8B).^16,18,66,67 In the latter, the S-bound C atom eventually migrates to Fe, a process that is conceptually related to the alkyl-group migration reactions reported herein. The discovery and characterization of [2]²⁺ therefore supports the plausibility of S-alkylated intermediates in several Fe–S enzymes that serve a variety of cellular functions.

Figure 8.

Proposed enzymatic intermediates that feature S–C bonds at [Fe₄S₄] and [Fe₃S₄] clusters. (A) Mechanism of the first S–C bond-forming step in LipA. (B) Structures of methylated intermediates proposed for methiothiotransferases and in nitrogenase maturation.

The reversibility of S-alkylation is also noteworthy in this context. We observed rapid migration from S to Fe in the reduction of [2]²⁺ to [1–MP]⁺, and, by analogy, it would be reasonable to assume that reduction of enzymatic S-alkylated [Fe₄S₄] intermediates could likewise induce S-to-Fe alkyl-group migration. However, this appears not to take place, and one mechanism for preventing such processes could entail the rapid loss of Fe after S-alkylation of [Fe₄S₄] clusters (or, in some cases, before S-alkylation), as has been proposed in mechanisms for LipA (Fig. 8A), MiaB, and RimO.^17,19,21,22 Thus, Fe loss may not be an incidental feature of these mechanisms, but instead a critical step to prevent alkyl migration to Fe and favor productive organosulfur chemistry.

The characterization here of reasonably stable S-alkylated [Fe₄S₄] clusters raises the question of why SAM-derived organoiron—but not organosulfur—intermediates have been observed in radical SAM enzymes. Indeed, our findings indicate that, in the absence of steric constraints imposed by the polypeptide, S-alkylation may be favored over Fe alkylation in the [Fe₄S₄]³⁺ state. Thus, it is likely that the active site geometry in radical SAM enzymes disfavors S–C bond formation. It is also possible that S-alkylated [Fe₄S₄] intermediates will be characterized in future studies of a radical SAM enzyme (observed, for example, under non-native reaction conditions, using non-native substrates, and/or in a protein mutant with a more open active site that increases the steric accessibility at S), and our work provides spectroscopic signatures that can be used to identify such species.

Lastly, our findings highlight the exceptional weakness of Fe–C bonds in [Fe₄S₄] clusters featuring alkylated Fe sites with high (> 5) coordination numbers. In the [Fe₄S₄]³⁺ state, we were able to trap species featuring a four-coordinate Fe–alkyl site (in the absence of an external ligand) or products of S-alkyl migration (in the presence of an internal or external ligand); in no cases have we observed the likely intermediate in these reactions: an [Fe₄S₄]³⁺ cluster featuring a 5-coordinate, alkylated Fe center. The favorability of Fe-to-S alkyl migration in this redox state speaks to both the stability of the S-alkylated species in the absence of a protein environment and the instability of the Fe-alkylated species. Given this unexpected reaction pathway, new strategies will be required to generate [Fe₄S₄] clusters with Fe-alkylated sites having a coordination number > 4.

Conclusion

We have shown that organic radicals can be generated on the same redox couple as has been proposed for radical SAM enzymes, [Fe₄S₄]³⁺–R/[Fe₄S₄]²⁺, in quantitative yield for the relatively stable benzyl radical and low yield for a reactive 1° radical. In our attempts to trap the key Fe-alkylated intermediate in this process, we made an unexpected discovery: in this small-molecule, synthetic system (i.e., one in which the cluster is not constrained by a protein active site), alkyl groups migrate from Fe to S in a net two-electron reductive elimination. The resulting S-alkylated [Fe₄S₄]⁺-like cluster was characterized, its spectroscopic signatures were elucidated, and its electronic structure was rationalized. We also showed that Fe-to-S migration is reversible upon re-reduction to the [Fe₄S₄]²⁺ state, and therefore that the regioselectivity of cluster alkylation can be controlled by the cluster’s redox state. This work connects the organoiron and organosulfur chemistry of Fe–S clusters, highlights the ease with which Fe–C and S–C bonds can be interchanged at Fe–S clusters, and lends credence to the intermediacy of S-alkylated intermediates in S transferase reactions.