Characterization by ENDOR Spectroscopy of the Iron-Alkyl Bond in a Synthetic Counterpart of Organometallic Intermediates in Radical SAM Enzymes

Abstract

Members of the radical S-adenosyl-l-methionine (SAM) enzyme superfamily initiate a broad spectrum of radical transformations through reductive cleavage of SAM by a [4Fe−4S]¹⁺ cluster it coordinates, to generate the reactive 5’-deoxyadenosyl radical (5’-dAdo•). However, 5’-dAdo• is not directly liberated for reaction, and instead binds to the unique Fe of the cluster to create the catalytically competent S=½ organometallic intermediate Ω. An alternative mode of reductive SAM cleavage, especially seen photochemically, instead liberates CH₃•, which forms the analogous S=½ organometallic intermediate with an Fe-CH₃ bond, Ω_M. The presence of a covalent Fe-C bond in both structures was established by the ENDOR observation of ¹³C and ¹H hyperfine couplings to the alkyl groups that show isotropic components indicative of Fe-C bond covalency. The synthetic [Fe₄S₄]³⁺–CH₃ cluster, M-CH₃, is a crystallographically characterized analogue to Ω_M that exhibits the same [Fe₄S₄]³⁺ cluster-state as Ω and Ω_M, and thus analysis of its spectroscopic properties—and comparison with those of Ω and Ω_M—can be grounded in its crystal structure. We report cryogenic (2 K) EPR and ¹³C/^1/2H ENDOR measurements on isotopically-labelled M-CH₃. At low temperature, the complex exhibits EPR spectra from two distinct conformers/subpopulations. ENDOR shows that at 2 K one contains a static methyl, but in the other the methyl undergoes rapid tunneling/hopping rotation about the Fe-CH₃ bond. This generates an averaged hyperfine coupling tensor whose analysis requires an extended treatment of rotational averaging. The methyl-group ¹³C/^1/2H hyperfine couplings are compared with the corresponding values for Ω and Ω_M.

Graphical Abstract

Introduction

Radical S-adenosyl-l-methionine (RS) enzymes, a superfamily with over 5×10⁵ members spanning all kingdoms of life, initiate a broad spectrum of radical transformations through use of S-adenosyl-l-methionine (SAM) and an [Fe₄S₄] cluster to liberate the reactive 5’-deoxyadenosyl radical (5’-dAdo•) for H-atom abstraction.^1,2 SAM first binds to the [Fe₄S₄]¹⁺ cluster through its methionyl amino-acid moiety,^3–6 and for decades it was believed that reductive cleavage of the SAM sulfonium by electron transfer from the cluster directly liberates the 5’d-Ado• radical for reaction with substrate.^7–10 The RS mechanism underwent a paradigm shift when it was discovered that across the superfamily, 5’-dAdo• formed by cleavage of the S-C5’ bond of SAM first binds to the unique Fe of the cluster to create the catalytically competent organometallic intermediate Ω (Fig 1); 5’-dAdo• is only liberated for reaction with substrate upon homolytic cleavage of the Fe-C5’ bond of Ω.^11–14 Subsequent studies showed an alternative mode of reductive SAM cleavage, especially seen photochemically, in which reductive homolytic cleavage of the SAM methyl-sulfonium bond generates a CH₃• that reacts to form an analogous intermediate, Ω_M, in which the unique Fe forms an organometallic Fe-CH₃ bond, Fig 1.¹⁵ Both intermediates were established by EPR and ENDOR spectroscopies to contain an [Fe₄S₄]³⁺−alkyl cluster. The S = ½ spin of each was shown to reside on an [Fe₄S₄]³⁺ cluster by its axial g-tensor with average g-value, g_iso > 2, and Ω furthermore was shown to exhibit the large ⁵⁷Fe hyperfine couplings associated with such a cluster.¹² The presence of a covalent Fe-C bond in each was established by the observation of ¹³C and ¹H hyperfine couplings to the alkyl that show isotropic components indicative of Fe-C covalency.^11,12

Figure 1.

Proposed structures of Ω and Ω_M (A) and crystallographically determined structure of M-CH₃ (B).

Synthetic [Fe₄S₄]–alkyl clusters have been prepared to better understand the properties of these enzymatic intermediates.^16–18 Of the synthetic [Fe₄S₄]–alkyl clusters reported to date, cluster M-CH₃ (Fig. 1B), a structurally-characterized analogue to Ω_M (Fig 1), is the only one with the same core charge state as that in Ω and Ω_M: [Fe₄S₄]³⁺.¹⁸ Like Ω and Ω_M, it adopts an S = ½ ground state with average g-value, g_iso, > 2.¹⁸ Importantly, M-CH₃ has been crystallographically characterized, and thus its composition, connectivity, and bond lengths are known without ambiguity. As such, analysis of its spectroscopic properties—and comparison with those of Ω and Ω_M—can be grounded in its crystal structure. We report here cryogenic EPR and ¹³C/^1/2H ENDOR measurements on isotopically labelled M-CH₃ and compare them with the corresponding measurements on Ω and Ω_M.

Materials & Methods:

Sample preparation:

The previously reported cluster, [(L^B(NIm)₃)Fe₄S₄CH₃][B(C₆F₅)₄] (M-CH₃), was prepared as described.¹⁸ The isotopologues M-¹³CH₃ and M-C(²H)₃ were prepared in an identical manner except with the methyl groups sourced from (¹³CH₃)₂Mg and (C(²H)₃)₂Mg, respectively, instead of natural-abundance (CH₃)₂Mg. Samples were prepared in an N₂-filled glove box as 1 mM solutions, loaded into custom quartz Q-band tubes, frozen in LN₂, and kept under LN₂ until spectroscopic characterization.

EPR and ENDOR measurements:

All CW (continuous wave) X-band EPR measurements were performed on a Bruker ESP-300 spectrometer with a liquid helium flow Oxford Instruments ESR-900 cryostat. The 35 GHz CW EPR and ENDOR spectra were recorded on a modified Varian E-110 spectrometer equipped with a helium immersion dewar.¹⁹ As the X- and Q-band spectra are collected in different spectrometers at different frequencies and in different detection modes – unsaturated absorption mode at X-band with derivative display through field modulation, rapid-passage dispersion mode with absorption-display at Q-band – the lineshapes and g-values determined by EPR simulations differ very slightly (third significant figure in g-values) for the two types of spectra. This has no impact on the collection of 2D field-frequency patterns of ENDOR spectra collected across the Q-band EPR envelope, and in the text g-values are only given to two significant figures. ¹³C ENDOR used swept CW ENDOR. The ¹H CW ENDOR was collected using the field modulation detected stochastic ENDOR sequence for better resolution of the features. Pulsed ENDOR measurements were collected at ~2 K, where electron-spin relaxation is slow in both M-CH₃ conformers, on a spectrometer described previously, with SpinCore PulseBlaster ESR_PRO 400 MHz digital word generator and Agilent Technologies Acquiris DP235 500 MS/s digitizer using SpecMan4EPR software.²⁰ For ²H pulsed ENDOR a Mims pulse sequence [π/2-τ-π/2-T_RF-π/2] was employed in which τ is the microwave pulse duration and T_RF denotes the time interval in which the RF is applied. For ¹H pulsed ENDOR a Davies pulse sequence [π-T_RF-π/2-τ-π] was employed. Both EPR and ENDOR simulations were carried out with EasySpin, using the pepper and salt functions, respectively.²¹

Hyperfine Sign Determination.

To obtain the signs of the measured hyperfine couplings (more precisely, the sign of A/g_N – where g_N is the nuclear g-factor and is positive for ¹H and ¹³C studied here) the pulsed-ENDOR-saturation and recovery (PESTRE) method was used at 35 GHz as described in detail previously²² and discussed below.

Results

EPR of M-CH₃:

Figure 2 shows the temperature dependence of EPR spectra of M-CH₃. At 40 K, M-CH₃ appears to give a somewhat poorly defined signal with an axial g-tensor characteristic of an [Fe₄S₄]³⁺ cluster, g_|| ≈ 2.09 > g_┴ ≈ 2.04 > 2, g_iso = (g_|| + 2g_┴)/3 > 2.²³ As the temperature is lowered, the poor resolution at 40 K improves, and by 12 K, the spectrum can be decomposed into well-defined contributions from two cluster conformers with comparable populations, one with g_|| ≈ 2.09 that persists at higher temperatures, denoted M1-CH₃, and a more-anisotropic signal with g_|| ≈ 2.11 from the second conformer, denoted M2-CH₃. Figure S1 shows that the properties of the two conformers vary only slightly with solvent, but that the relative populations vary significantly. Rapid increases in spin-relaxation with temperature restrict ENDOR measurements to cryogenic temperatures (and thus also to frozen solution). Whereas two states are observed by EPR of the frozen-solution, the fluid-solution ¹H NMR spectra of M-CH₃ recorded at T > 193 K show only a single set of resonances.¹⁸ This difference indicates that in fluid solution the two conformers rapidly interconvert on the NMR timescale.

Figure 2.

CW X-band EPR spectra of M-CH₃ in 3:1 DFB/toluene. Conditions: temperature 12 K, 30 K and 40 K, 9.375 GHz microwave frequency, 10 G modulation amplitude, 320 ms time constant, 200 μW microwave power. Simulation parameters: (a) g = [2.088, 2.052, 2.042] (red), g = [2.113, 2.05, 2.022] (blue), sum (dashed).

¹³C ENDOR:

The signature feature of M-CH₃ is its organometallic Fe-CH₃ bond, and isotopic labeling of the methyl group has enabled the use of ¹³C ENDOR to characterize this bond. The 2D field-frequency pattern of ¹³C ENDOR spectra collected across the EPR envelope of M-¹³CH₃ is displayed in Fig. 3. It shows a single ¹³C doublet with a near-isotropic splitting of magnitude, A = 5.5 MHz. As this signal is the summation of responses from the two conformers in comparable amounts, the observation of a single doublet at all fields indicates that the Fe-C bonding is essentially identical in both conformers. The presence of this isotropic ¹³C hyperfine interaction establishes the covalency of the Fe-C bond, as also inferred from previous ⁵⁷Fe Mossbauer spectroscopic analysis.¹⁸ The slight changes in apparent lineshape across the field can be attributed to an extremely small anisotropic component to the hyperfine couplings that manifest themselves itself differently in the signals from the two conformers, in part because of the difference in their g-values.

Figure 3.

Q-band CW ¹³C ENDOR spectra (right) of M-(¹³C)H₃ (black), natural abundance M-CH₃ (red), and absorption-display EPR spectrum of M-CH₃ (left) and 2-component simulations with parameters as in Fig 2. Spectra centered at ¹³C Larmor frequency. Feature centered at −5 MHz is ¹⁴N signal from the scorpionate ligand. ENDOR Conditions: 2K, 34.934 GHz microwave frequency for natural abundance sample, 35.01 GHz microwave frequency for ¹³C sample, 3.2 G modulation amplitude, 1.2 μW MW power, forward swept, 32 ms time constant, scan speed 3 MHz/second, 10–100 scans for each. Intensity of natural abundance spectra were scaled to the corresponding ¹³C spectra intensity by scaling the scorpionate ¹⁴N signal.

The sign of the ¹³C hyperfine coupling is critical for connecting hyperfine values to bonding descriptions, but is not given by conventional ENDOR measurements. To determine the hyperfine sign we thus conducted Pulsed ENDOR SaTuration-REcovery, PESTRE measurements, Fig 4.²² This protocol employs a multi-sequence of Davies ENDOR electron spin-echo pulse sequences. The first set of sequences is applied without applied RF and establishes a ‘baseline (BSL)’ steady-state electron spin-echo response; each sequence in the second set incorporates an RF pulse at a chosen frequency (ENDOR); the response in the third set exhibits a dynamic reference level (DRL) whose offset from BSL is induced by relaxation effects that depend on the sign of the hyperfine coupling and that decays to BSL.²² The signs of the deviation of the DRL from the BSL (denoted DRL-δ) at the ν₊ and ν₋ partners of a hyperfine doublet independently establish the sign of the hyperfine coupling.²²

Figure 4.

Absolute hyperfine sign determination for methyl ¹³C coupling at g = 2.05. (Top) Baseline-corrected Q-band Davies pulsed ¹³C ENDOR. (Bottom) ¹³C PESTRE trace collected from the ν+ branch of the ¹³C coupling. ENDOR conditions: 34.753 GHz microwave frequency, π = 120 ns, τ = 600 ns, t_rf = 35 μs, rf tail = 5 μs, repetition time = 50 ms, spectral resolution 256 points, temperature 2K, PESTRE collected at rf frequency 15.8 MHz.

The ¹³C ENDOR response from M-C(²H)₃ exhibits the well-resolved ν₋/ν₊ doublet of Fig 3, and both branches have been probed with PESTRE. The weaker ν₋ peak does not give a clear PESTRE signature for sign assignment, but the clean observation of DRL-δ < 0 when probing ν₊ (Fig 4) unambiguously establishes that the ¹³C hyperfine coupling is positive: the essentially isotropic ¹³C hyperfine coupling thus has the positive value a_iso(¹³C) = +5.5 MHz. This coupling is caused by positive spin density on the methyl ¹³C, which arises because the spin-down unpaired electrons on the alkylated Fe center polarize the doubly-occupied Fe–C σ-bond, inducing positive spin density at C (Scheme S1).

The conclusion that the methylated Fe center exhibits negative spin density is consistent with previous NMR analysis of M-CH₃.¹⁸ The surprising absence of a distinct anisotropy to the ¹³C coupling implies that the through-space dipole-dipole interaction with the down-spin electrons on the unique Fe, as modified by smaller contributions from the other Fe’s, is almost exactly offset by an oppositely signed anisotropic interaction with the positive (spin-up) local spin on C.

^1,2H ENDOR analysis of the M1-C(^1,2H)₃ Conformer:

In parallel, we carried out ^1,2H ENDOR studies of M-CH₃ and its isotopologue, M-C(²H)₃, at 2 K. Figure 5A shows the richly featured ¹H ENDOR spectrum of M-C(H)₃ taken near the peak of the absorption-display EPR spectrum (g = 2.065), whose breadth corresponds to a maximum ¹H hyperfine coupling of A ~ 15 MHz. The ¹H signals specifically associated with the Fe-bound C(¹H)₃ are absent in the spectrum of M-C(²H)₃, and are better visualized by subtraction of the two spectra, Fig 5B. These spectra are in turn complemented by the ²H Mims ENDOR signals introduced by deuterium substitution in M-C(²H)₃, which match well with the methyl ¹H signals upon scaling the ²H frequencies by the ratio of proton and deuteron nuclear g-factors, g_N(H)/g_N(D) = 6.518, given that the features of the ²H spectra are broadened by unresolved ²H quadrupole coupling (see SI), with the appearance of broadening enhanced by the presentation on the scaled frequency axis. This is illustrated in Fig 5C for g = 2.065, while Fig S2 shows that the full 2D field-frequency pattern of ²H Mims spectra of M-C(²H)₃ mirrors the pattern of (¹H)₃ spectra of M-C(¹H)₃, Figs 6, S3.

Figure 5.

(A) Q-band stochastic ¹H ENDOR at field corresponding to g=2.05 of M-CH₃ (black) and M-CD₃ (red), scaled to equalize the intensity of non-exchangeable proton couplings. (B) Spectrum upon subtraction of M-CD₃ from M-CH₃ partitioned into M1-CH₃ (blue) and M2-CH₃ (gold) contributions. Simulation of M1-CH₃ (blue, dashed) uses a single axial hyperfine tensor. M2-CH₃ contribution obtained by subtracting M1 simulation from CH₃-CD₃ difference spectrum. M2-CH₃ simulation discussed in text and described in SI. (C) Q-band Mims ²H ENDOR spectra scaled to ¹H frequency by a factor g_N(¹H)/g_N(²H) = 6.514 at field corresponding to g = 2.05. (D) ¹H PESTRE trace collected at the ν+ branch (red) and ν- branch (black) of the ¹H coupling at g = 2.07. ¹H stochastic ENDOR conditions: 2 K, ~35.1 GHz MW frequency, 3.2 G modulation amplitude, 1.2 μW MW power, 2 s rf on, 2 s rf off, 0.5 s wait time, 150–1250 scans each. ²H Mims ENDOR conditions: 2 K, microwave frequency 34.74 GHz, π = 100 ns, τ= 300 ns, t_rf= 60 μs, rf tail = 10 μs, repetition time 50 ms, spectral resolution 256 points. ¹H PESTRE conditions: 34.753 GHz microwave frequency, π = 120 ns, τ = 600 ns, t_rf = 35 μs, rf tail = 5 μs, repetition time = 50 ms, spectral resolution 256 points, temperature 2 K, PESTRE collected at rf frequency 47.67 and 54.55 MHz. M1-CH₃ ENDOR simulation parameters: g = [2.091, 2.054, 2.042], where g₁ = g_|| corresponds to the z-direction in a conventional coordinate frame; A = [−12.7 −6.45 −6.45], and [α, β, γ] = [0, 55, 0].

Figure 6.

Subtraction of Q-band stochastic ¹H ENDOR spectra of M-CH₃ and M-CD₃ to give the methyl proton spectrum (black), simulation of a single axial hyperfine tensor for M1-CH₃ (red), and result of subtracting simulation from CH₃-CD₃ difference spectrum (blue) (right), along with the absorption-display EPR spectrum of M-CH₃ (left) with simulations as in Fig 3. ¹H stochastic ENDOR conditions: 2 K, ~35.1 GHz MW frequency, 3.2 G modulation amplitude, 1.2 μW MW power, 2 s rf on, 2 s rf off, 0.5 s wait time, 150–1250 scans each. ENDOR simulation parameters: g = [2.091, 2.054, 2.042], where g₁ = g_|| corresponds to the z-direction in a conventional coordinate frame; A = − [12.7 6.45 6.45], and [α, β, γ] = [0, 55, 0]; the absolute sign is determined as described in text.

Figure 2 shows that the EPR signal of M-CH₃ is a sum of signals from two conformers, and consequently the ¹H ENDOR spectra of M-CH₃ are a superposition of ENDOR responses from the two at fields where the EPR spectra overlap (Figs 5B, 6, S3). M2-CH₃ alone contributes to the EPR and thus to both ¹H (Figs 6, S3) and ²H ENDOR spectra at the high- and low-field edges of the EPR spectrum because of the greater M2-CH₃ g-anisotropy.

Simulations of the cryogenic (2 K) 2D field-frequency pattern of orientation-dependent ¹H ENDOR spectra collected across the M1-CH₃ portion of the EPR envelope, Figs 6, S3, show that the three methyl protons of the M1-CH₃ conformer exhibit a well-defined joint, averaged ¹H response characteristic of the frozen-solution of a center with axial g-tensor and a single hyperfine-coupled proton with non-coaxial hyperfine tensor that is itself precisely axial.^28,29 As discussed further below, this behavior at 2K is most simply understood as arising from tunneling and/or hopping of the methyl protons through rotations of 2π/3 among the energetic minima of a three-fold symmetric potential surface that rotationally averages the hyperfine tensor. This, behavior of M1-CH₃ is analogous to the tunneling/hopping averaging for H₂ in non-classical Fe- and Co-H₂ complexes,^30,31 and is captured in the cartoon of Scheme 1.

Scheme 1:

Representation of Fe-CH₃ behavior in the rotationally averaged M1 and static M2 conformers.

The complete averaged ¹H hyperfine tensor derived above from simulations is A1(¹H)^av = –[12.7, 6.45, 6.45] MHz, with Euler angles defining the hyperfine-tensor orientation relative to the g-tensor frame of [α, β, γ] = [0, 55, 0]. The relative signs of the components are fixed by the simulations of the field dependence; the negative absolute sign of the components is obtained by use of the multi-sequence PESTRE protocol. This is shown in Figs 5D for a measurement at g = 2.065, with the RF set to a frequency near the maximum C(¹H)₃ ENDOR intensity of M1-CH₃; PESTRE measurements at additional fields corresponding to the other principal g-values of M1-CH₃, give the same negative sign for the interaction, Fig S4, while PESTRE measurements described below show that the ¹H couplings for M2-CH₃ likewise are negative. The observed, averaged M1-CH₃ ¹H hyperfine tensor, A1(¹H)^av, thus can be decomposed into a negative isotropic coupling, a_iso^av= −8.5 MHz, and a negative axial anisotropic tensor T^av = [T^av_||, T^av_┴, T^av_┴] = −[4.2, −2.1, −2.1] MHz. The negative sign of the A(¹H)^av tensor of M1-CH₃, in particular that of the isotropic coupling, is in agreement with that implied by the negative paramagnetic shift of the methyl protons previously observed in VT-NMR.¹⁸ To understand the negative sign we note that the Fe-CH₃ protons are ‘β’ to Fe, and as a result the isotropic contribution to the hyperfine coupling, a_iso^av = −8.5 MHz, can be attributed to spin transfer from the ‘down-spin’ on Fe to the protons through hyperconjugation and/or through-bond spin polarization from Fe through C to the coupled ¹H.

¹H ENDOR of Conformation M2-CH₃:

For fields at which the EPR spectrum of M2-CH₃ overlaps that of M1-CH₃, subtraction of the ¹H ENDOR simulation of the single proton representing the rotationally averaged methyl protons of M1-CH₃ from the observed spectrum gives the ¹H ENDOR response of M2-CH₃, as illustrated in Fig 5B. In addition, as noted, the M2-CH₃ EPR signal, and thus its ENDOR signals, persist to both lower and higher field because of its greater g-anisotropy. Figs 6 and S5 show the resulting M2-CH₃ ¹H ENDOR spectra collected across the EPR envelope of this conformer. These signals are of low intensity and nearly featureless across the EPR envelope, with a maximum coupling that is somewhat larger than the M1-CH₃ maximum, and that remains roughly constant across the field at a value, A ~ −15 MHz, Figs 6, S5. The negative sign of A(¹H) for this conformer is likewise established by PESTRE measurements at multiple fields and frequencies, Figs 4, S4, S6.

The broad, low-intensity ¹H ENDOR responses of M2-CH₃ indicate that this methyl is essentially static, Scheme 1. For a static M2-CH₃, the three methyl protons necessarily have differently oriented hyperfine tensors, and may well have different hyperfine components as the result of a hyperconjugative interaction with the Fe anchor and/or dipolar interactions with the other cluster Fe. Thus, they resonate over a different range of frequencies at each g-value of observation. Their signals neither track nor reinforce each other at fields across the EPR envelope, resulting in an essentially featureless signal whose total breadth varies little. For illustration, Fig S5, shows the 2D pattern of M2-CH₃ spectra is well reproduced by summation of three tetrahedrally-oriented methyl-group ¹H hyperfine tensors with similar isotropic couplings (a_iso = −8 ↔ −10 MHz) and anisotropic coupling tensor components (average value of unique unique component, T2_||^av ≅ −5 MHz). The proton signals from conformer M2-CH₃ are substantially less intense than those of M1-CH₃, despite the ~50/50 ratio of conformer populations in the simulation of the 12 K EPR spectrum (Fig 1), because of the M1-CH₃ rotational averaging. The averaged M1 ¹H signal arises from three protons that resonate at the same frequency with every orientation of the Fe-C center relative to the external field (same hyperfihne tensor components and orientation), and thus their contributions coherently add, unlike those of the M2 protons with their different tensor orientations.

Finally, why is methyl rotation stopped for conformer M2, but not M1? We speculate that in a frozen solution, the two conformers exhibit differential solvent access to the unique Fe site. Specifically, small differences thus introduced in the Fe–N, N–C(imidazole), and/or N–C(tolyl) torsion angles of the M2 conformer could yield large secondary structure changes in the scorpionate ligand that hinder the methyl rotation, or bring solvent molecules near the methyl that have this effect. Presumably the effects that quench M2 methyl rotation in the solid state at 2 K also slightly distort the cluster geometry, thereby changing the spin coupling among the Fe ions and/or individual site g-tensors sufficiently to slightly modify the observed g-values. An underlying solvent influence is supported by the observation that varying the solvent composition changes the relative occupancy of the two conformers (Fig S1). However, as we show below, the properties of the Fe-CH₃ bond are the same in the two conformers.

Rotational Averaging of the M1-C(¹H)₃ Anisotropic Hyperfine Interactions:

As presented above, the averaged orientation-selective, ¹H ENDOR response from the three CH₃ protons of the M1-CH₃ conformer at 2 K is characteristic^28,29 of the frozen solution of a center with axial g-tensor and hyperfine coupling to a single proton whose hyperfine tensor is itself precisely axial, but non-coaxial with the g-tensor; it is not the sum of distinct and differing individual signals from each of the three, as seen for M2-CH₃. As visualized in Scheme 1 (left), at 2 K the M1-CH₃ methyl-group undergoes rapid rotation about the Fe-C bond through tunneling/hopping on its three-fold symmetric ground-state potential-energy surface, thereby averaging the M1-CH₃ methyl-group proton hyperfine tensors in behavior analogous to the tunneling/hopping of H₂ in a non-classical trigonally-symmetric Fe-H₂ complex.³¹ The extensive literature discussing the dynamics of methyl-group rotation includes numerous considerations of how such rotation can affect the ¹H ENDOR spectra of the methyl protons,^{24,25,27,32,33} However, here we focus on the properties of the M-CH₃ complex and its organometallic Fe-C bond, and thus address the consequences of the rotational averaging on the ¹H hyperfine tensor of the M1 conformation. In most prior studies the methyl group under consideration is part of an organic radical whose EPR spectrum is without resolved g-anisotropy, while the hyperfine coupling is large and largely isotropic. To our knowledge, it has not been explicitly recognized that rotation does not merely average the coupling constants of the three methyl protons.³⁴ When the methyl group is part of a paramagnetic center that exhibits well-resolved g- and hyperfine-anisotropy, an additional level of averaging is required to generate orientation-resolved ENDOR spectra in which the three methyl ¹H nuclei jointly give a 2D field-frequency pattern of ENDOR spectra that appears as though it were from a single ¹H with an averaged, anisotropic hyperfine tensor.

The three protons of a static methyl likely have hyperfine tensors with different component values, but must have differently oriented (non-colinear) anisotropic interaction tensors, and this is necessarily true for the H-atoms of the M1-CH₃ methyl. Thus, the observation of a single joint, orientation-selective ¹H ENDOR pattern requires not only that the three ¹H nuclei of the M1-methyl exhibit hyperfine tensors with the same components, whether intrinsically or through rotational averaging, but beyond that it requires that rotational averaging of the single-site tensors themselves gives them a common averaged orientation. As discussed next, and in more detail in SI, this rotational averaging reorients the observed, averaged, joint hyperfine tensor, redefines its ‘symmetry’ by making it precisely axial, and modifies the component values of the anisotropic interaction from those of the single-site tensors.

To understand the rotational averaging of a methyl-proton hyperfine tensor, consider Fig 7A, which shows one of the three CH₃ protons of an Fe-CH₃ fragment with each proton having an intrinsic traceless anisotropic hyperfine tensor of arbitrary rhombicity and with its unique T₃ axis tipped within its Fe-C-H plane by an arbitrary angle δ with respect to the Fe-C rotation (z) axis. We may write the tensor components for this ¹H as,

$$T=\left[T_1,T_2,T_3\equiv T_{\left|\right|}\right]\equiv T\left[-\left(1-\eta \right),-\left(1+\eta \right),2\right],$$ (1)

with T as a scaling factor that sets the strength and overall sign of the interaction, and where η defines the rhombicity, with η = 0 corresponding to a perfectly axial tensor. As illustrated in Fig 7A, if the anisotropic term were dominated by through-space interactions between spin on the Fe and the ¹H nucleus, its unique hyperfine axis, T₃, would point closely along the Fe-H vector, with tip angle δ_T ~ 22°; if it were dominated by interaction with local spin on carbon, the unique axis would point along the C-H bond, with δ_L ~ 70°.

Figure 7.

(A): Depiction of methyl Fe-C-H structure with angle δ_T the angle between the Fe-C bond/ rotation-axis and the dashed line, which represents the unique axis of through-space ¹H dipolar coupling to spin on Fe, and the angle δ_L between the Fe-C bond and the C-H bond, along which lies the unique axis for ¹H anisotropic coupling to spin on C. (B): Plot of the axial anisotropic hyperfine parameters, T^av_||, T^av_┴, obtained upon assumption of rapid rotation of the methyl about Fe-C (Eqs S3), for convenience normalized to a reference scaling factor (eq 1), T = +1 (positive spin density on Fe), and plotted as function of δ for several values of rhombicity parameter, η. For negative T (negative Fe spin density), and a corresponding reference factor T = −1, the plot y axis is inverted.

As shown in the SI, when the methyl group rotates about the Fe–C axis rapidly enough for the three protons to yield a single, joint, averaged ¹H hyperfine tensor in the 2D field-frequency pattern of orientation-selective ENDOR spectra collected at fields across the EPR envelope, then this rapid rotation must have averaged the components of the site tensors to yield a single average traceless anisotropic hyperfine tensor with a common orientation, A1^av, and this averaging does not depend on the nature of the motion – being the same for all types of averaging in which the potential surface for methyl rotation does not favor a particular methyl rotamer. The anisotropic contribution to this averaged tensor, T1^av (eqs S3), is precisely axial, and is reoriented so that its unique component, T1^av_||, lies along the Fe-C bond, the rotation axis, as shown in Scheme 1. Thus, the Euler angles determined from the simulation of the 2D pattern of M1-CH₃ ¹H ENDOR spectra, which describe the orientation of the axial hyperfine-tensor A1^av relative to the g-frame, φ = 0, θ = 55° for X, in fact describe the orientation of the Fe-C bond relative to the g-frame. As illustrated in Fig 8, this places g_|| nearly parallel to the compression axis of the M-CH₃ [Fe₄S₄]³⁺ cluster,¹⁸ analogous to the orientation of g_|| seen in single-crystal EPR studies of [Fe₄S₄]³⁺ clusters.^35,36

Figure 8:

The ENDOR-derived relative orientations of g_|| (red arrow) and A1^av_|| (blue arrow), and the resulting orientation of g_|| relative to the compression axis of the M-CH₃ [Fe₄S₄]³⁺ cluster, which as described in ref 18, is normal (dashed red line) to the face for which the Fe-Fe-CH₃ angle is 138°. Thermal ellipsoids (30%); orange (Fe), yellow (S), blue (N), gray (C),white (H). Ligand is trimmed; anion and solvent removed.

The rotational averaging further causes the values of the observed averaged tensor components to become functions of the tip-angle δ of the site-tensors relative to the Fe-C bond (Fig 7, eqs S3). As illustrated in Fig 7B, the value of the observed T1^av_|| drops with increasing δ, crosses through zero at an angle δ > 45° that depends on η, then changes sign (eqs S3). The components of the averaged ¹H anisotropic hyperfine tensor determined for M1-CH₃ were given above, T1^av_|| = −4.2 MHz, T1^av_┴ = +2.1 MHz. The measured negative value of the unique component, T1^av_||, is consistent in its negative sign with assignment of the scaling factor, T (eq 1), to a through-space dipolar interaction of ‘down’ spin on the methyl-bound Fe center with the methyl ¹H (Fig 7A). Given the geometry of the Fe-CH₃ moiety, if the intrinsic (non-averaged) anisotropic interaction T (eq 1) is assigned to a such an interaction, the intrinsic T_|| would be tipped by an angle of δ_T ~ 22° (Fig 7, left), and thus its contribution to T1^av_|| would retain the negative sign of the scale-parameter, T (Fig 7, right).

This assignment of the anisotropic hyperfine coupling to the through-space interaction with Fe, rather than to interaction with local spin on C, is supported by the magnitude of T1^av_||, which is consistent with that expected for dipolar coupling to Fe when it is noted that eqs S3 then predict an observed averaged value, T1^av_||/2T ~ 0.8 (see Fig 7B). As discussed below, the combination of this decrease in the measured M1 anisotropic coupling caused by rotational averaging, plus the overall hyperfine averaging for the rotating M1 methyl, leads to the slightly smaller breadth of the M1 methyl ¹H ENDOR pattern compared to that for the M2 methyl.

Discussion and Conclusions

Complex M-CH₃, with its [Fe₄S₄]³⁺-CH₃ organometallic bond, provides a crystallographically characterized realization of the ENDOR-derived structures of the alkyl-ligated [Fe₄S₄]³⁺ states of radical SAM enzymes: the Ω_M state, whose Fe-CH₃ bond precisely parallels that displayed by M-CH₃ (Fig 1), as well as paralleling the Fe-alkyl bond of Ω itself, the catalytically-central enzymatic intermediate in which it is the 5’-dAdo moiety that forms an Fe-C5’ bond.¹² The Mossbauer and NMR studies of M-CH₃ established the unusual electronic-structure effects imparted to the cluster by a strong-field alkyl ligand, most especially the Fe³⁺ valence localization of the alkylated iron, and thus lay the foundation for understanding the overall electronic structures of [Fe₄S₄]³⁺−alkyl radical SAM intermediates.¹⁸ This report has built on this foundation by probing the Fe-C bond in detail through ¹H and ¹³C ENDOR measurements that can be compared directly to those in Ω_M, and more generally to those of Ω itself.

As a foundation for analysis, we first note that all freeze-trapped radical-SAM enzymatic [Fe₄S₄]³⁺−alkyl intermediates exhibit the doublet cluster ground state with g_iso > 2, characteristic of an [Fe₄S₄]³⁺ cluster.^11,12,23 Thus, the finding that M-CH₃, with its [Fe₄S₄]³⁺cluster and crystallographically characterized structure, also exhibits g_iso > 2, confirms the EPR-derived characterization of the cluster of the transient enzymatic intermediates. Secondly, it is of interest that M-CH₃ exhibits two [Fe₄S₄]³⁺−alkyl conformers with similar properties of the Fe-C bond, but slightly different g-values, for work in progress indicates that the same is true for the enzymatic intermediates. The methyl group of the M1-CH₃ complex rotates rapidly at 2 K, and we have clarified how this rotation creates the observed ¹H hyperfine tensor of a single proton associated with the rotationally averaged methyl protons, Scheme 1. Such rotation: redefines the symmetry of a single-site hyperfine tensor, making the joint, averaged tensor precisely axial; it reorients this axial tensor so that the unique axis is parallel to the Fe-C bond; and it modifies the component values from those of the site tensor as shown in Fig 7B (see eqs S3).

Considering the properties of the Fe-alkyl bond itself, methyl-group interactions with the ‘down’-spin on the unique Fe cause both conformers of M-CH₃ to exhibit a small positive and essentially isotropic ¹³C coupling constant, indicative of positive (‘up’) spin density induced on carbon. The magnitude of the M-CH₃ coupling, a_iso(¹³C) = ~ +6 MHz, is quite comparable to that observed for Ω, |a_iso(¹³C)| ~ 9 MHz,¹² while both couplings are somewhat smaller than |a_iso(¹³C)| ~ 18 MHz for Ω_M;¹⁵ notably, the coupling is more than an order of magnitude smaller than that purported for Ω in a recent DFT study.³⁷ In contrast, the ‘down’ spin on the unique Fe induces a negative isotropic ¹H hyperfine coupling to the methyl protons of both M-CH₃ conformers, and the magnitude of their a_iso(¹H) also compare well with the ¹H coupling in Ω, |a_iso(¹H)| = 7 MHz, as follows.¹¹ The rotationally averaged methyl hyperfine tensor of M1-CH₃ has an isotropic coupling of a_iso(¹H) = −8.5 MHz and anisotropic coupling with unique value T1^av_|| = −4.2 MHz, , while the 2D ENDOR pattern of M2-CH₃ (Fig S5) is satisfactorily reproduced by the sum of signals from the three tetrahedrally-oriented protons of a static methyl group (Scheme 1) with ¹H tensors that have similar isotropic couplings, a_iso2(¹H) ~ −(8 ↔ 10) MHz, and an average value for the unique component of the anisotropic coupling tensor of, T2^av_|| = −5 MHz. This latter corresponds to T1^av_|| after taking into account the decrease in T1^av_|| associated with rotational-averaging according eqs S3 (Fig 7B). The resulting equivalence of the intrinsic ¹H hyperfine tensors in the two conformers shows that the effects that slightly alter the g-values of the M-CH₃ conformers and modulate the dynamics of methyl group rotation nonetheless do not alter the actual Fe-CH₃ bonding.

¹³C and ¹H ENDOR/PESTRE hyperfine-sign measurements of the enzymatic Ω and Ω_M intermediates will test whether they exhibit the same isotropic coupling signs as that of M-CH₃, as well as the corresponding magnitudes; if so, this will confirm that the alkyls in the enzymatic intermediates are likewise bound to a valence-localized, spin-down Fe³⁺, with implications for the cluster electronic structure previously discussed.¹⁸ Such comparisons of the properties of Ω and Ω_M with those of M1-CH₃, in combination with additional quantum computations and synthetic modeling, will address the extent to which the higher coordination number of the unique Fe in the Ω’s, with the methionyl amino-acid moiety of SAM providing additional ligand(s), affects the electronic structure of the alkylated Fe site, and thus the properties of the Fe-alkyl bond. Moreover, the hyperfine coupling values reported herein—the first measured for a crystallographically characterized [Fe₄S₄]³⁺–alkyl cluster—will help benchmark computational predictions for the hyperfine coupling in the enzymatic [Fe₄S₄]³⁺–alkyl intermediates Ω and Ω_M.