Photoinduced Electron Transfer in a Radical SAM Enzyme Generates an S-Adenosylmethionine Derived Methyl Radical

Abstract

Radical SAM (RS) enzymes use S-adenosyl-l-methionine (SAM) and a [4Fe–4S] cluster to initiate a broad spectrum of radical transformations throughout all kingdoms of life. We report here that low-temperature photoinduced electron transfer from the [4Fe–4S]¹⁺ cluster to bound SAM in the active site of the hydrogenase maturase RS enzyme, HydG, results in specific homolytic cleavage of the S–CH₃ bond of SAM, rather than the S–C5′ bond as in the enzyme-catalyzed (thermal) HydG reaction. This result is in stark contrast to a recent report in which photoinduced ET in the RS enzyme pyruvate formate-lyase activating enzyme cleaved the S–C5′ bond to generate a 5′-deoxyadenosyl radical, and provides the first direct evidence for homolytic S–CH₃ bond cleavage in a RS enzyme. Photoinduced ET in HydG generates a trapped ^•CH₃ radical, as well as a small population of an organometallic species with an Fe–CH₃ bond, denoted Ω_M. The ^•CH₃ radical is surprisingly found to exhibit rotational diffusion in the HydG active site at temperatures as low as 40 K, and is rapidly quenched: whereas 5′-dAdo^• is stable indefinitely at 77 K, ^•CH₃ quenches with a half-time of ~2 min at this temperature. The rapid quenching and rotational/translational freedom of ^•CH₃ shows that enzymes would be unable to harness this radical as a regio- and stereospecific H atom abstractor during catalysis, in contrast to the exquisite control achieved with the enzymatically generated 5′-dAdo^•.

Graphical Abstract

INTRODUCTION

Adenosylcobalamin (coenzyme B₁₂) enzymes^1–3 and radical S-adenosyl-l-methionine (SAM) (RS) enzymes^4–6 both initiate radical reactions in biology through generation of the reactive 5′-deoxyadenosyl (5′-dAdo^•) radical, as created either by homolysis of the Co–C5′ bond of B₁₂ or by reductive cleavage of the sulfonium S–C5′ bond of SAM. This radical then abstracts an H atom from substrate. The deep parallelism of the two processes has been established by the recent discovery of a primary and mechanistically central organometallic RS intermediate Ω that contains an Fe–C5′ bond, and liberates 5′-dAdo^• through homolysis of this bond in precise analogy to the liberation of 5′-dAdo^• by homolysis of the Co–C5′ bond of B₁₂.^7–9

We recently sought to generate 5′-dAdo^• through photolysis and homolytic Fe–C5′ cleavage of Ω, in analogy to the Co–C5′ photohomolysis of coenzyme B₁₂.^10–13 However, while doing so, we instead discovered that photoinduced electron transfer (ET) from the [4Fe–4S]¹⁺ cluster to bound SAM in pyruvate formate-lyase activating enzyme (PFL-AE) at 12 K directly cleaved the S–C5′ bond of SAM to generate 5′-dAdo^• (Figure 1).¹⁴ In short, the energy of light allowed the system to overcome the intrinsic thermodynamic barrier associated with electron transfer from the [4Fe–4S]⁺ cluster (midpoint potential ~ −450 to −550 mV)^15–20 to SAM (cathodic peak potential ~ −1.4 V).²¹ The key result of this process was that the 5′-dAdo^• radical thus generated was trapped next to the diamagnetic [4Fe–4S]²⁺ cluster in the PFL-AE active site and could be completely characterized by electron paramagnetic resonance (EPR) and electron nuclear double resonance (ENDOR) spectroscopies, thus realizing a goal sought for over half a century.¹⁴ An underlying lesson from this experiment was that the photoinduced ET resulted in reductive cleavage of the same S–C5′ bond of SAM as is cleaved thermally during RS catalysis in PFL-AE.^22,23

Figure 1.

Reductive cleavage of SAM and the structure of HydG. (A) The SAM-cluster interaction in RS enzymes (left), and the three different radical products that can result from reductive S–C bond cleavage in SAM (right). Canonical RS enzymes cleave the S–C5′ bond (red) to generate the 5′-dAdo^• radical intermediate, while the noncanonical RS enzyme Dph2 cleaves the S–C(γ) bond (green) to yield an ACP radical fragment that adds to substrate. (B) Overview of the structure of HydG (PDB 4WCX), with the radical SAM cluster (top) and the auxiliary cluster (bottom) connected by a long active site tunnel (purple mesh).

Essentially all characterized canonical RS enzymes initiate radical catalysis via reductive cleavage of the S–C5′ bond of SAM. This regioselectivity has been proposed to result from the geometry of the SAM-[4Fe–4S] cluster interaction, in which the S–C5′ bond of SAM is trans to the S–Fe interaction, making the unique iron, sulfonium sulfur, and C5′ approximately in the colinear arrangement required for a radical displacement mechanism.²⁴ Configurational interaction between the [4Fe–4S]¹⁺ donor orbital and the S–C5′ σ* acceptor orbital is thought to facilitate specifically S–C5′ bond cleavage.²⁵ The mechanistic formation of Ω during the reductive cleavage of SAM by RS enzymes may help to define this regioselectivity.^8,9 In contrast, a different regioselectivity is observed in the noncanonical RS enzyme Dph2, where the S–C(γ) bond to the methionine side chain is cleaved during catalysis (Figure 1).²⁶ This S–C(γ) regioselectivity, like that in canonical RS enzymes, appears to have a mechanistic origin reflecting the specific geometry of the SAM-cluster interaction, and also involves an organometallic intermediate.²⁷

The observation of S–C5′ cleavage in canonical radical SAM enzymes, and S–C(γ) bond cleavage in Dph2, raises the question of whether enzymatic systems exist that might generate a SAM-derived methyl radical, ^•CH₃, via reductive cleavage of the third S–C bond of SAM (Figure 1).²⁸ Such a possibility was considered by Kampmeier, who proposed that an alternate SAM-cluster interaction geometry would lead to such homolytic S–CH₃ bond cleavage;²⁴ further, a SAM-derived ^•CH₃ radical has been proposed as a possible intermediate during NifB-catalyzed carbon insertion into the iron–molybdenum cofactor of nitrogenase.²⁹ However, in possible opposition to these ideas, while an entire subclass of radical SAM enzymes are known to carry out methylation chemistry using the methyl of SAM, these methylation reactions go via S_N2 rather than radical mechanisms, with subsequent radical chemistry initiated via 5′-dAdo^•.³⁰

HydG is a radical SAM enzyme that catalyzes a key reaction required for maturation of the H-cluster of the [FeFe]-hydrogenase: the synthesis of the CO and CN⁻ ligands of the H-cluster. HydG uses SAM as a cosubstrate, converting it to 5′-deoxyadenosine and methionine while initiating the radical decomposition of tyrosine to produce CO, CN⁻, and p-cresol; the CO and CN⁻ are ultimately incorporated into the active-site H-cluster of the [FeFe]-hydrogenase.^31–35 Two HydG [4Fe–4S] clusters are involved in this process. As is typical in RS enzymes, the [4Fe–4S]_RS cluster binds SAM through the methionyl amino acid and initiates radical chemistry via the reductive cleavage of SAM to provide a 5′-dAdo^• that abstracts H^• from tyrosine to initiate its decomposition.³³ An auxiliary [4Fe–4S]_AUX cluster is involved in subsequent chemistry,^33,36 and can bind a fifth, cysteine-coordinated iron to produce an S = 5/2 [4Fe–4S]_AUX[FeCys] cluster.³⁷ The fifth “dangler” iron of the [4Fe–4S]_AUX[FeCys] cluster has been proposed to serve as a site for formation of an [Fe(CO)₂(CN)] synthon that is transferred to the scaffold protein HydF during hydrogenase maturation.^38–40 The [4Fe–4S]_RS and [4Fe–4S]_AUX clusters are at opposite ends of an active site cavity/tunnel spanning the central TIM barrel of HydG, with the [4Fe–4S]_RS at the N-terminal end and [4Fe–4S]_AUX at the C-terminal end; the distance between the two clusters is approximately 24 Å (Figure 1).^37,41

We previously showed that during catalytic turnover of HydG with SAM and tyrosine, the [4Fe–4S]¹⁺_RS cluster reductively cleaves the S–C5′ bond of SAM homolytically to generate the Ω intermediate, which subsequently liberates 5′-dAdo^• through Fe–C5′ bond homolysis.⁸ The [4Fe–4S]¹⁺_RS cluster of HydG does not reductively cleave SAM in the absence of tyrosine.⁴² We now report that the low-temperature photoinduced electron transfer from the [4Fe–4S]¹⁺_RS cluster to bound SAM in the substrate-free HydG active site results in the specific cleavage of the S–CH₃ bond of SAM, rather than the S–C5′ bond cleaved during catalysis,^31–33,43 to generate primarily a trapped ^•CH₃ radical, with a small concentration of an Ω- like organometallic species with an Fe–CH₃ bond, denoted Ω_M, also observed. The ^•CH₃ radical and Ω_M species and their isotopologues have been characterized by EPR and ENDOR spectroscopies. The ^•CH₃ radical is rapidly quenched even at 77 K and is surprisingly free to tumble in the HydG active site at temperatures as low as 40 K. We discuss the implications of these observations for the evolution of regioselective SAM S–C5′ bond cleavage for the regio- and stereospecific H atom abstraction from substrates in RS enzymes by 5′-dAdo^•.

MATERIALS AND METHODS

All chemicals and other materials used herein were from commercial sources and of the highest purity available. Triton-X100 and imidazole were obtained from Alfa Aesar (Ward Hill, MA). NaCl was obtained from Sigma-Aldrich (St. Louis, MO). HEPES, potassium phosphate, IPTG, PMSF, tryptone, yeast extract, MOPS, DTT, glucose, and streptomycin were obtained from RPI (Mt. Prospect, IL). Ferrous ammonium sulfate was obtained from J.T. Baker Chemical Company (Phillipsburg, New Jersey). MgCl₂, KCl, and glycerol were obtained from EMD (Gibbstown, NJ). Sodium dithionite and sodium sulfide were obtained from Acros Organics (Fair Lawn, NJ). DNase I, RNase A, and lysozyme (hen egg) were obtained from Roche (Indianapolis, IN). Iron(III) chloride and l-cysteine were obtained from Fisher Scientific (Fair Lawn, NJ).

Wild-Type HydG^ΔEF Expression and Purification.

HydG was overexpressed and purified with some modifications to previously published procedures.^31,33,34,44 The Clostridium acetobutylicum hydG gene (CAC1356) was cloned into a pCDFDuet vector to allow for the expression of the N-terminally, 6×-histidine tagged form of the protein;⁴⁴ the vector was transformed into E. coli BL21(DE3) (Stratagene) cells. This expression system of the HydG enzyme in the absence of HydE and HydF is denoted HydG^ΔEF. Fresh agar plates were streaked and single colonies were utilized for 50 mL overnight cultures in media supplemented with 50 μg/mL streptomycin. The following morning, the overnight cultures were used to inoculate 9 L LB that was buffered with 50 mM potassium phosphate, pH 7.5, and additionally contained 10 g/L tryptone, 5 g/L yeast extract, 5 g/L NaCl, 5 g/L glucose, and 50 μg/mL streptomycin. These cultures were then grown aerobically at 37°C with 230 rpm shaking. When OD₆₀₀ values were between 0.5–0.6, 0.09 g/L of ferrous ammonium sulfate (FAS) and IPTG (1 mM final concentration) were added. The cultures continued to grow under aerobic conditions for an additional 2.5 h at 37°C with 230 rpm shaking. At this time, the cultures were cooled to room temperature and supplemented with an additional aliquot of 0.09 g/L FAS. The cultures were then transferred to a 4°C refrigerator and actively sparged with N₂(g) for 15 h. The following morning, the cells were harvested via centrifugation and resulting cell pellets were flash frozen in liquid N₂ and stored at −80°C until lysis and purification.

Cell lysis and protein purification were carried out in a Coy chamber (Grass Lake, MI) that was maintained with a ~97% N₂(g), ~3% H₂(g) atmosphere and housed inside a 4°C walk-in cold room. Frozen cell pellets were transferred to the Coy chamber and then thawed and resuspended in a lysis buffer containing 50 mM Tris pH 8.0, 0.25 M KCl, 5% glycerol, and 10 mM imidazole (Buffer A). Lysis mixtures also contained 20 mM MgCl₂, 1 mM PMSF, 0.8% Triton X-100, 0.4 mg lysozyme per gram of cells, and 0.07 mg DNase and RNase per gram of cells. This mixture was then stirred for 60 min, transferred into gastight centrifuge bottles, and centrifuged for 30 min at 18 000 rpm. Clarified supernatants were then passed over a 5 mL HisTrap Ni²⁺-affinity column (GE Healthcare, Piscataway, NJ) using an Ä_KTA Basic 100 FPLC (GE Healthcare). Purified HydG was eluted from the column using an imidazole gradient with increasing amounts of Buffer B (50 mM Tris, pH 8.0, 0.25 M KCl, 5% glycerol, 500 mM imidazole). Purification fractions were immediately pooled together and buffer exchanged into a 50 mM Tris, pH 8.0, 0.25 M KCl, 5% glycerol buffer using a Sephadex G25 resin column. Eluted fractions were concentrated using Amicon Ultra 10 kDa molecular weight cutoff filter devices (Millipore). Concentrated protein aliquots were then flash frozen in liquid N₂ and stored at −80°C until further use. Protein was quantified via Bradford assays using a bovine serum albumin standard solution (Thermo Scientific). Iron in purified protein samples was quantified via a Varian SpectrAA 220 FS flame atomic absorption spectrometer in comparison to a 0.4–2.0 ppm standard curve created from a 1000 ppm iron AA standard solution (Ricca Chemical Company).

Chemical Reconstitution of HydG.

HydG typically purifies with substoichiometric amounts of Fe loading. For example, the WT HydG^ΔEF enzyme utilized herein contained 3.2 ± 0.1 Fe/protein in its as-purified state. It was therefore necessary to chemically reconstitute protein samples in order to improve [Fe–S] cluster loading. Reconstitutions were carried out with minor modifications to our previously published procedures.^31,33,34 Briefly, HydG enzyme (≤150 μM) was incubated with a 6-fold excess of FeCl₃ and Na₂S in the presence of 5 mM DTT in 50 mM Tris, pH 8.0, 0.25 M KCl, 5% glycerol buffer. The reconstitutions were carried out in scintillation vials for 2.5–3.5 h with gentle stirring in an anaerobic Coy chamber. Following the reconstitution period, mixtures were centrifuged to pellet bulk FeS precipitates and resulting supernatants were passed over a Sephadex G-25 column to remove adventitiously bound iron, sulfide, and DTT. Gel filtered proteins were then concentrated using Amicon Ultra 10 kDa molecular weight cutoff filter devices (Millipore). Aliquots of protein were then flash frozen in liquid N₂ and stored at −80°C until spectroscopic characterization. Bradford analysis and iron quantitation were performed as described above. Following reconstitution, WT HydG^ΔEF harbored 7.7 ± 0.2 Fe atoms/protein.

Preparation of Dangler Loaded HydG.

Reconstituted HydG^ΔEF was loaded with the dangler iron via modification of previously published protocols.^39,45 Dangler loading was accomplished by treating as-reconstituted HydG (150 μM) with 2 mM NaDT, 1.25 equiv of FeCl₃, 10 equiv of SAM, and 7 equiv of l-cysteine. The mixture was incubated at 4°C with stirring in a Coy anaerobic chamber for 45 min. Following the incubation period, the sample was immediately passed over a Sephadex G-25 column, and key fractions were pooled together, and then concentrated using Amicon Ultra 10 kDa molecular weight cutoff filters (Millipore). The resulting protein stocks were then aliquoted, flash frozen in liquid N₂, and stored at −80°C until spectroscopic characterization. Final iron quantitation numbers were 9.4 ± 0.3 Fe/protein.

Sample Preparation.

In an anaerobic Coy chamber, HydG samples for intracavity photolysis and EPR spectroscopic analysis were prepared by mixing enzyme with Na-dithionite to reduce the clusters to the 1+ state, SAM, and buffer (50 mM Tris (pH 8.0), 250 mM KCl, 10% glycerol), to final concentrations of 100 μM protein, 3 mM DT, 5.5 mM SAM, respectively. These samples were then briefly centrifuged, transferred to EPR tubes, flash frozen, and stored in liquid N₂.

EPR and ENDOR Measurements.

These measurements were carried out as described recently.¹⁴ X-band CW EPR spectroscopy was conducted on a Bruker ESP 300 spectrometer equipped with an Oxford Instruments ESR 910, while Q-band CW EPR spectroscopy was conducted on a Bruker EMX spectrometer equipped with an Oxford Instruments Mercury iTC continuous helium flow cryostat. Typical experimental parameters were at 12 and 40 K, 9.38 or 34.0 GHz, and 10 G modulation amplitude. EPR simulations were performed with the EasySpin^5,2,23 program operating in Matlab.⁴⁶ 35 GHz CW and pulse ENDOR spectroscopic data were collected on spectrometers, described previously,^47–49 that are equipped with liquid helium immersion dewars for measurements at 2 K. The CW measurements employed 100 kHz field modulation and dispersion-mode detection under rapid passage conditions. ¹H CW ENDOR spectra employed bandwidth broadening of the RF to 100 kHz to improve signal-to-noise.⁵⁰

RESULTS

Photoinduced ET in the HydG/[4Fe–4S]¹⁺_RS/SAM Complex Generates ^•CH₃.

Irradiation of the SAM-bound [4Fe–4S]⁺ state of substrate-free HydG with 450 nm light at 12 K in the EPR cavity results in rapid loss of the signal from the [4Fe–4S]¹⁺_RS cluster (Figure 2A), while leaving undiminished the signal from the second [4Fe–4S]_AUX cluster. This is accompanied by the appearance of a new free-radical signal (Figure 2B). Figure 2C shows that the time course for the loss of [4Fe–4S]¹⁺_RS signal directly correlates with that of the appearance of the new radical signal. Measurements of the radical signal intensity generated by comparable total flux of light of 400, 450, and 520 nm showed a ratio of intensities of I(400)/I(450)/I(520) ~ 1/0.6/0.1 (Figure S1), in rough correspondence with the wavelength dependence of the absorbance spectrum.

Figure 2.

X-band EPR spectra of (A) ([4Fe–4S]⁺ + SAM) HydG complex before and (B) after photolysis at 12 K with 450 nm LED for 1 h. (C) Time course for decay of ([4Fe–4S]⁺ + SAM) (■) upon photolysis monitored at 3640 G, and increase of ^•CH₃ (●) monitored at 3306 G. The photo reaction appears to be complete at ~90 m, as the EPR spectrum does not change upon further photolysis. The two progress curves are then normalized and fit to stretched-exponential decay, I = exp(−[t/τ]ⁿ), and rise, I = 1 − exp(−[t/τ]ⁿ), functions with the same parameters, τ = 7 ± 1 min and n = 0.52 ± 0.02. In (B), the unchanged signal from [4Fe–4S]_AUX[(Cys)Fe] cluster and that from ^•CH₃ are both identified. EPR conditions: microwave frequency, 9.38 GHz; modulation 10 G; T = 12 K.

As previously,¹⁴ these results indicate that photoinduced electron transfer (ET) from the [4Fe–4S]¹⁺_RS to SAM causes reductive cleavage of SAM to generate the diamagnetic [4Fe–4S]²⁺_RS cluster and a cryotrapped radical. However, this radical is not 5′-dAdo^•, as previously observed upon photolysis of the SAM-bound [4Fe–4S]¹⁺ state of PFL-AE.¹⁴ Rather, the species generated in HydG exhibits a quartet EPR signal, Figure 2B, better resolved at 40 K, Figure 3A, that is quite different from that of 5′-dAdo^•. This signal is precisely simulated by a four-line pattern with 1/3/3/1 intensity ratio as predicted for isotropic hyperfine coupling to three equivalent ¹H with a_iso = 63.6 MHz. This pattern and splitting value are characteristic of the ^•CH₃ radical,⁵¹ thereby identifying the signal as arising from this species. This identification is confirmed by the collapse of the ¹H splittings when CD₃-SAM is used, Figure 3B, and the appearance of an overlapping doublet of quartets, generated by a ¹³C hyperfine coupling, a₃(¹³C) = 80 MHz, when ¹³CH₃–SAM is used, Figure 3C. Photoreduction has cleaved the S–CH₃ bond of SAM, not the S–C5′ bond.

Figure 3.

EPR spectra (black) and simulations (red). (A) Spectrum of rapidly tumbling ^•CH₃ generated with natural abundance SAM, obtained by subtraction of the properly scaled signal from the persistent signal (see text). (B) ^•CD₃ generated with methyl-d₃-SAM. (C) ^•13CH₃ generated with ¹³C-methyl-SAM. Conditions: T = 40 K, modulation amplitude, 5 G; see Figure 4 for spectra with lower modulation. EasySpin simulation parameters: function, chili; g = [2.0015, 2.0015, 2.0055,], (g_iso = 2.003). For spectra (A) and (C), A(¹Ha,b,c) = −[86, 44, 62] MHz (a_iso(¹Ha,b,c) = −63.3 MHz), (α,β,γ) = (120, 0, 0) for 1Ha, (α,β,γ) = (−120, 0, 0) for 1Hb, (α,β,γ) = (0, 0, 0) for 1Hc; for spectrum (C), A(¹³C) = [−10, −10, 260] MHz. τ_C = 3.2 ns; uniform Gaussian line widths 21 MHz.

The quartet signal from the ^•CH₃ radical rapidly decays at 77 K, with a half-time of a few minutes, and progressively more rapidly at higher temperatures (Figure S2). The decay of this signal on annealing reveals a weak underlying minority signal, which persists up to 150 K and has been subtracted from the spectra shown in Figure 3. The signal that persists after annealing, Figure 4, is not that of a “free” methyl radical. It is narrower (Figure S3), does not show resolved ¹H hyperfine splittings (Figures 4, S3), and can be simulated with an envelope that arises from an anisotropic g-tensor, g = [2.021, 2.0047, 1.987] (Figure 4A). This anisotropy is distinctly less than that of the organometallic intermediate Ω.⁷ However, when this species is generated with ¹³CH₃–SAM the signal broadens significantly; the broadening can be simulated by the addition of a ¹³C hyperfine coupling, a_iso(¹³C) ~ 23 MHz (Figure 4B). This observation is confirmed by ¹³C ENDOR (CW, 35 GHz, 2 K): despite the low population of this species, we were able to obtain a ¹³C ENDOR signal that corresponds to a weakly anisotropic coupling with a_iso(¹³C) ≈ 19 MHz, Figure 4C. Although this a_iso(¹³C) is about twice the coupling to the ¹³C5-dAdo^• bound to an Fe of Ω,⁷ both couplings are much smaller than the coupling expected for an isolated, immobilized •13CH₃ radical, which also would be highly anisotropic, rather than nearly isotropic, and with much larger couplings, A(¹³C) ≈ [−10, −10, 260] MHz.⁷ The most plausible interpretation of these observations is that during cleavage of the S–CH₃ bond of SAM a small fraction of the ^•CH₃ reacts with the RS cluster, forming an Fe–CH₃ bond to generate an organometallic species, Ω_M, analogous to Ω. However, for completeness we note an alternative formulation, in which the ^•CH₃ reacts with the RS cluster to form an S–CH₃ bond.²⁹

Figure 4.

X-band EPR spectra. (A) Ω_M (¹²CH₃); (B) Ω_M (¹³CH₃); (C) 35 GHz ¹³C CW ENDOR Spectra of Ω_M (¹³CH₃). Inverted triangle, ¹³C Larmor frequency; “goalpost”, half the hyperfine coupling, a_iso/2. EPR conditions: T = 40 K; modulation amplitude, 5 G; microwave frequency, 9.38 GHz. ENDOR conditions: T = 2 K; modulation amplitude, 4 G; scan speed, 3 MHz/s; scan direction, forward; microwave frequency, 34.8 GHz. EasySpin EPR simulations: function, pepper; g = [2.021, 2.0047,1.987]; Gaussian line widths 30 MHz; for spectrum (B), A(¹³C) = [20, 30, 20] MHz; H_strain = [60, 30, 60] MHz.

Mobility of ^•CH₃ in HydG.

As discussed below, the extent of the motional freedom of ^•CH₃ in the active-site during catalysis at near-ambient temperatures is important in understanding the properties of RS enzymes. As this freedom will be far greater under those conditions than at cryogenic temperature, we have examined the low-temperature results as giving insights into and a lower limit to the rotational and translational mobility during catalysis.

The statistical 1/3/3/1 intensity pattern of the ^•CH₃ radical at 40 K and above (Figures 3A and 5) implies that the radical is undergoing rapid rotational diffusion within the active site even down to these low temperatures. Although •CH₃ is roughly disk-shaped and presumably the rotational diffusion is to some extent anisotropic, the spectra of ^•12CH₃ and ^•13CH₃ at 40 K can actually be jointly described by the simplification of treating reorientation as a fully three-dimensional (classical; isotropic) rotational diffusion. As seen in Figure 3, both these 40 K spectra are satisfactorily reproduced with a rotational correlation time of τ_c ≈ 3 ns, at least 10-fold longer (slower rotation) than would be expected at ambient temperature, although nonetheless in the “fast-motion” regime. The lines in the ·¹³CH₃ spectrum are broader because of less complete averaging of the ¹³C hyperfine anisotropy, whose magnitude is ~6.4 times greater ((A_∥ − A_⊥) ~ 270 MHz) than that of an α-proton (|A_max − A_min| ~ 42 MHz). Some additional remarks about 3D vs 1D ^•CH₃ rotation are presented in the SI (SI text, Figure S4).

Figure 5.

Temperature dependence of the ^•CH₃ signal. Conditions: modulation amplitude, 2G to avoid modulation broadening in the higher-T spectra; power adjusted at each temperature to avoid saturation; because of this adjustment, the intensities are simply normalized to equal height.

The rapid tumbling of the ^•CH₃ radical seen here at ~40 K and above has commonly been observed at liquid nitrogen temperatures and above, as illustrated by a few examples: ^•CH₃ in frozen solutions of solvents ranging from nonpolar to aqueous;⁵² at 77 K generated by X-irradiation of CH₃I;⁵³ at 210 K trapped in zeolite cages;⁵⁴ and as transiently generated through photolysis of MeCob in aqueous solution at 284 K.⁵⁵ In this last case, the lines of the quartet at 280 K are extremely narrow, thus demonstrating the expected enhanced rotational mobility near ambient temperatures.

As the sample is cooled to ~30 K and below, the quartet seen above 30 K does not simply broaden, as would be expected if τ_c for tumbling smoothly increased, nor does the spectrum convert to that characteristic of a low-temperature quantum rotor (for example, refs 56, 57). Instead, the spectrum appears to progressively convert to a broadened quartet that can be assigned to a more slowly tumbling ^•CH₃. Such a transition would appear to reflect progressive conversion of the site into a conformation exhibiting greater hindrance to tumbling, with conversion nearly complete by ~4 K, Figure 5. For completeness, in the SI, we make further comments about ^•CH₃ rotation as exhibited here.

The ^•CH₃ radical signal rapidly decays at 77 K (Figure S2), in clear contrast to the behavior of 5′-dAdo^• formed photochemically in PFL-AE; the 5′-dAdo^• radical is indefinitely stable at 77 K and only decays at temperatures above ~100 K.¹⁴ This decay is not accompanied by the appearance of any new paramagnetic species, other than Ω_M as discussed above. We considered the possibility that the rapid ^•CH₃ decay might reflect the existence of a “tunnel” between the RS and AUX [4Fe–4S] clusters of HydG (Figure 1B). This offered the possibility that ^•CH₃ formed at the [4Fe–4S]_RS cluster might diffuse through the tunnel to, and react with, the [4Fe–4S]_AUX[(Cys)Fe] cluster, and that this reaction might cause the observed decay of the ^•CH₃ signal. We tested this possibility by measuring the intensity of the [4Fe–4S]_AUX[(Cys)Fe] cluster signals, both the S = 1/2 signal seen in Figure 2, and the S = 5/2 signals that appear at low fields,³⁷ after 77 K annealing had eliminated the ^•CH₃ signal: the intensity of the [4Fe–4S]_AUX[(Cys)Fe] cluster signals had not changed from those seen promptly after photolysis, ruling out this possibility, Figure S5. We thus conclude that ^•CH₃ is intrinsically more reactive than 5′-dAdo^•. This is consistent with its greater C–H bond-dissociation free energy, but is more likely associated with its ability to undergo rapid translational diffusion “in search” of a reaction partner. Except when ^•CH₃ is strongly confined by a rigid, covalently bonded environment, for example, when isolated in a zeolite cage⁵⁴ or adsorbed on a silica surface,⁵⁸ it diffuses quite freely and in consequence quenches rapidly, as seen here in HydG.

DISCUSSION

We report here that cryogenic photoinduced electron transfer from the [4Fe–4S]¹⁺_RS cluster to bound SAM in the active site of HydG results in the cleavage of the S–CH₃ bond of SAM, rather than the S–C5′ bond as is cleaved in enzymatic catalysis by nearly all RS enzymes,⁶ including HydG,^34,42 or in photolysis of SAM-bound [4Fe–4S]¹⁺ PFL-AE.¹⁴ The S–CH₃ bond cleavage reported here is also distinct from the S–C(γ) bond cleaved in the noncanonical RS enzyme Dph2.²⁶ The radical product of HydG/SAM photolysis is, remarkably, a ^•CH₃ free radical trapped in the HydG active site. While a methyl radical has been suggested as a possible intermediate during catalysis by the radical SAM enzyme NifB,²⁹ which inserts the methyl carbon of SAM as the central carbide of the iron–molybdenum cofactor of nitrogenase, no prior evidence for reductive homolytic cleavage of the S–CH₃ bond of SAM in a RS enzyme has been reported. Further, while a methyl radical mechanism has been proposed for methyl-coenzyme M reductase,⁵⁹ no direct evidence for a methyl radical has been reported for this or other enzymatic systems.

The ^•CH₃ cleavage product of HydG/[4Fe–4S]⁺/SAM photolysis is highly reactive and rotationally and translationally mobile, and clearly could not reliably carry out regioselective H atom abstraction from a substrate. We have been unable to carry out photolysis of HydG/[4Fe–4S]⁺ in the presence of both SAM and tyrosine due to the rapid reaction of the ternary complex, analogous to the limitations on photolysis of the PFL-AE/[4Fe–4S]⁺/SAM/PFL complex.¹⁴ However, decades of study of ^•CH₃ in the widest range of environments and temperatures have clearly shown that because of its small size and weak interactions with its surroundings,^{52–55 •}CH₃ could not be effectively harnessed. As indicated above, the many published studies of ^•CH₃ show that the small ^•CH₃ radical would not only be intrinsically unstable under turnover conditions, but would also undergo rapid rotation within the active site.^52–55 Clearly, even in the presence of substrate tyrosine, ^•CH₃ would be highly mobile and not subject to enzymatic control: it could not be constrained to adopt the precise positioning required for selective H atom abstraction.⁶⁰

The present study thus provides a powerful reason for nature’s “choice” of radical initiator in RS enzymes: despite the fact that ^•CH₃ would be a somewhat more potent H atom abstractor than 5′-dAdo^• based on their relative C–H bond energies,⁶¹ enzymatic RS reactions require precise control of H atom abstraction from substrate.⁶ In canonical RS enzymes, the large 5′-dAdo^• species fits this requirement, as it can be efficiently directed to its target through nonbonded and H-bonding contacts with active-site residues. Indeed, as we have shown,⁶⁰ once the 5′-dAdo^• free radical is liberated it is “never free”, but is tightly guided through an atom- and regio-/stereospecific reaction with substrate. The enzyme avoids the insurmountable problems associated with formation of ^•CH₃ by using active-site control to ensure the reductive cleavage of S–C5′ to form the relatively easy to control 5′-dAdo^• radical, and not cleavage of S–C(Me) to form ^•CH₃.^24,60 The 3-amino,3-carboxypropyl (ACP) radical formed in the noncanonical RS enzyme Dph2 by S–C(γ) bond cleavage is anchored by amino and carboxylate coordination to the unique iron of the [4Fe–4S] cluster, which enhances control but also limits its ability to react with the wide range of substrates dealt with in the full RS superfamily.

The tight regioselectivity of the enzymatic reductive cleavage of SAM observed across the diverse RS superfamily raises the question: why is the S–CH₃ bond cleaved, with full regiospecificity, in the current study, whereas photoinduced ET of the PFL-AE/[4Fe–4S]¹⁺/SAM complex results in specific S–C5′ bond cleavage? The results for these two systems indeed suggest that photoexcitation of the [4Fe–4S]¹⁺_RS/SAM complex in RS enzymes yields an excited state that is capable of transferring an electron to cleave any of the three S–C bonds of SAM. Although the factors that determine the regioselectivity of S–C bond cleavage during photoinduced ET of [4Fe–4S]¹⁺_RS/SAM complexes are not currently understood, it appears that in contrast to the enzymatically catalyzed reactions, this selectivity does not require major differences in the geometry of the SAM-cluster interaction, as this geometry is very similar in PFL-AE and HydG (Figure S6).^37,62 In addition, while the S–C bond energies in SAM are not currently known, it appears that differences in these energies are not responsible for the S–C bond cleavage regioselectivity during photoinduced ET, as we observed 100% regioselectivity for S–C(Me) bond cleavage in the current study, yet complete regioselectivity for S–C5′ bond cleavage in our recent study of PFL-AE.¹⁴

It is of interest to note that the direct product of catalytic cleavage of both the S–C5′ and S–C(γ) carbons in RS enzymes and the noncanonical RS enzyme Dph2, respectively, is Ω or an Ω-like organometallic intermediate, yet the photoinduced S–C5′ cleavage in PFL-AE generates 5′-dAdo^• without forming Ω,¹⁴ and the photoinduced cleavage of S–CH₃ in HydG yields only a small amount of the Ω_M species in addition to ^•CH₃. These variations in cleavage product demonstrate just how subtle are the active-site interactions that determine the radical created by SAM cleavage. Of central importance to the present considerations, the generation of Ω during catalytic cleavage of SAM S–C5′ and at least partial generation of Ω_M during photoinitiated cleavage of S–CH₃ provides a strong connection between the catalytic and photoinitiated processes.

In summary, active-site control of SAM positioning dictates the cleavage of the S–C5′ bond of SAM to form 5′-dAdo^• during catalysis by RS enzymes,²⁴ and control of 5′-dAdo^• positioning guides the stereospecific reactions of this agent with the broad range of substrate targets within the RS superfamily.⁶⁰ Photoinitiated electron transfer to SAM in the substrate-free [4Fe–4S]¹⁺ HydG active site instead induces cleavage of SAM to exclusively generate ^•CH₃, some of which forms the organometallic species Ω_M with its Fe–CH₃ bond.

The photoinduced ET reported here for the HydG [4Fe–4S]¹⁺/SAM complex, as well as that in our recent study of PFL-AE,¹⁴ are among the first examples of photochemistry involving iron–sulfur clusters.^63–65 The remarkable specificity and efficiency of the photoinduced ET in both these enzymes leads us to wonder whether nature might have taken advantage of such processes. For example, might there be radical SAM enzymes that are directly activated by light to initiate radical reactions? Could such enzymes in this ancient superfamily have been involved in photochemical reactions during the early evolution of life? It is interesting to note that adenosylcobalamin, which exhibits remarkable parallels to the RS [4Fe–4S]/SAM system,⁹ has recently been found to take advantage of its inherent photochemistry to function as a light sensor.⁶⁶ Although to date no evidence for a light-activated RS enzyme has been reported, our ability to use light to quantitatively generate organic radicals in enzyme active sites could very well lead to novel approaches to initiating radical reactions for synthetic or other applications.

The rapid quenching and rotational/translational freedom of ^•CH₃ shows that the HydG active site would be unable to effectively harness this radical as a regio- and stereospecific H atom abstractor during catalysis, thus highlighting the importance of enzymatic control of both the formation and the reactions of 5′-dAdo^•.