Radical SAM Enzyme Spore Photoproduct Lyase: Properties of the Ω Organometallic Intermediate and Identification of Stable Protein Radicals Formed during Substrate-Free Turnover

Abstract

Spore photoproduct lyase is a radical S-adenosyl-l-methionine (SAM) enzyme with the unusual property that addition of SAM to the [4Fe-4S]¹⁺ enzyme absent substrate results in rapid electron transfer to SAM with accompanying homolytic S─C5′ bond cleavage. Herein, we demonstrate that this unusual reaction forms the organometallic intermediate Ω in which the unique Fe atom of the [4Fe-4S] cluster is bound to C5′ of the 5′-deoxyadenosyl radical (5′-dAdo^•). During catalysis, homolytic cleavage of the Fe─C5′ bond liberates 5′-dAdo^• for reaction with substrate, but here, we use Ω formation without substrate to determine the thermal stability of Ω. The reaction of Geobacillus thermodenitrificans SPL (GtSPL) with SAM forms Ω within ~15 ms after mixing. By monitoring the decay of Ω through rapid freeze–quench trapping at progressively longer times we find an ambient temperature decay time of the Ω Fe─C5′ bond of τ ≈ 5–6 s, likely shortened by enzymatic activation as is the case with the Co─C5′ bond of B₁₂. We have further used hand quenching at times up to 10 min, and thus with multiple SAM turnovers, to probe the fate of the 5′-dAdo^• radical liberated by Ω. In the absence of substrate, Ω undergoes low-probability conversion to a stable protein radical. The WT enzyme with valine at residue 172 accumulates a Val^•; mutation of Val172 to isoleucine or cysteine results in accumulation of an Ile^• or Cys^• radical, respectively. The structures of the radical in WT, V172I, and V172C variants have been established by detailed EPR/DFT analyses.

Graphical Abstract

INTRODUCTION

The diverse radical S-adenosyl-l-methionine (radical SAM, RS) enzyme superfamily contains hundreds of thousands of distinct sequences found throughout all kingdoms of life.^1-3 More than 80 distinct radical SAM reactions have been identified, catalyzing key steps in critical processes including DNA repair, protein and tRNA modification, and synthesis of vitamins, antibiotics, protein cofactors, and complex metal clusters. These RS enzymes all contain a site-differentiated [4Fe-4S] cluster that binds SAM through a distinctive methionyl amino/carboxylate chelation of the unique iron of the [4Fe-4S] cluster.^4-6

The reduced [4Fe-4S]⁺ cluster of RS enzymes reductively cleaves the S─C5′ bond of the bound SAM to generate a 5′-deoxyadenosyl radical (5′-dAdo^•), the species that initiates catalysis by abstracting a specific hydrogen atom from the substrate.⁷ Recently, however, we showed that the reductive cleavage of SAM during enzymatic catalysis does not simply liberate 5′-dAdo^•.⁸ Rather, rapid freeze–quench spectroscopic studies of a collection of RS enzymes that span the range of reaction types catalyzed by the superfamily have shown that in all cases SAM cleavage generates a paramagnetic organometallic intermediate denoted Ω (g_∥ = 2.035, g_⊥ = 2.0) in which the unique Fe atom of the [4Fe-4S] cluster forms a bond with C5′ of 5′-dAdo^•.^8,9 Subsequent homolytic Fe─C5′ bond cleavage liberates 5′-dAdo^•,¹⁰ the same reactive species liberated by cleavage of the Co─C5′ bond of adenosylcobalamin (coenzyme B₁₂), for reaction with substrate. The elusive 5′-dAdo^• radical species long implicated in both RS and B₁₂ reactions was finally observed and characterized only recently, a result of the novel discovery of photoinduced [4Fe-4S]⁺ → SAM electron transfer within the [4Fe-4S]/SAM complex.¹¹

One of the RS enzymes that has been observed to form Ω is spore photoproduct lyase (SPL), a radical SAM enzyme that catalyzes the repair of spore photoproduct (SP), a methylene-bridged intra-DNA strand thymine dimer, by direct reversal to generate the native thymine–thymine sequence, Figure 1.^12-14 Upon binding of SAM and SP substrate in the active site,¹⁵ DNA repair is initiated by reductive cleavage of SAM to generate the 5′-dAdo^• radical intermediate, which abstracts a hydrogen atom from C6 of spore photoproduct.^16-18 This substrate radical then undergoes a radical-mediated beta-scission, and the resulting product radical is quenched by addition of an H atom to form the repaired thymine–thymine sequence.¹⁹ It has been proposed that a radical-relay pathway involving a tyrosine and a cysteine residue in the active site is involved in the quenching of the product radical.^20-22

Figure 1.

SPL is a DNA repair enzyme that binds to DNA (left) to repair spore photoproduct (SP), a methylene-bridged thymine dimer, with direct reversion to the repaired TT sequence (right).

In the majority of RS enzymes, addition of SAM to the [4Fe-4S]¹⁺ cluster in the absence of substrate causes a perturbation in the [4Fe-4S]¹⁺ cluster EPR signal with the g values shifting as the result of formation of a [4Fe-4S]¹⁺/SAM complex.¹ Although this complex is generally stable for tens of minutes to hours, it is ultimately subject to nonproductive reductive cleavage of SAM with formation of the EPR-silent [4Fe-4S]²⁺ cluster, methionine, and 5′-deoxyadenosine (5′-dAdoH). If SAM and substrate are added simultaneously to the [4Fe-4S] ¹⁺ state of the enzymes, the organometallic intermediate Ω can be observed on a milliseconds time scale with attendant loss of cluster signal.^8,9 In contrast, in SPL addition of SAM to the reduced enzyme in the absence of substrate results in rapid decay of the [4Fe-4S]¹⁺ cluster EPR signal without its conversion to a signal characteristic of the [4Fe-4S]¹⁺/SAM complex,^18,23 indicating that the cluster has transferred its electron, presumably to SAM with accompanying homolytic S─C5′ bond cleavage.

Herein, we employed this unusual property of SPL to add to our understanding of the mechanism of SAM-dependent radical generation by RS enzymes through studies of the formation of Ω, its thermal stability, and its reactivity. We examined the reaction of Geobacillus thermodenitrificans SPL (GtSPL) with SAM in the absence of SP substrate using a wide range of freeze-quench times to examine the progress of the reaction. We find that in the substrate-free enzyme the [4Fe-4S]¹⁺ cluster is oxidized on a milliseconds time scale upon addition of SAM with generation of Ω. This has allowed us to determine the thermal stability of this Ω at ambient temperature by rapid freeze trapping at progressively longer times, up to 10 s. This in turn allows us to compare the stability of the Ω Fe─C5′ bond with the Fe─C bond of a model compound²⁴ as well as that of the B₁₂ Co─C5′ bond. We have further used long-time hand quenching with quench times up to 10 min, and thus with multiple SAM turnovers, to probe the fate of the 5′-dAdo^• radical liberated by Ω. In this process we use insights from the X-ray structure of GtSPL²⁵ to guide site-directed mutagenesis that, together with EPR and DFT studies of longer time hand-quenched samples, identify the low-yield protein radicals formed in GtSPL.

MATERIALS AND METHODS

Materials.

3′,4′,5′,5″-d₄-Adenosine 5′-triphosphate salt solution was purchased from Cambridge Isotope Laboratories, Inc. Adenosine 5′-triphosphate salt and l-methionine were purchased from Sigma. Unlabeled and labeled SAMs were prepared as previously reported.²⁶

Cloning, Protein Expression, and Purification.

The gene coding for the C140A variant of spore photoproduct lyase from G. thermodenitrificans was synthesized and cloned into a pET-14b vector between the NdeI and the BamHI cloning sites by GenScript to express a N-terminal His₆-tagged SPL fusion protein. Sequence and primer information is provided in the Supporting Information. WT, Y98F, C140G, C140A/V172C, and C140A/V172I variants were obtained by QuickChange PCR mutagenesis using Phusion polymerase (New England Biolabs). The correctness of the cloned DNA sequences was confirmed by sequencing of the entire gene.

SPL preparations were performed as described previously²⁵ with minor modifications. Briefly, the recombinant spl genes from G. thermodenitrificans were expressed in BL21-(ΔIscR) E. coli cells. After inoculation from overnight starter cultures, cell culture growths in phosphate-buffered LB media (9 L in 2.8 L Fernbach flasks) were supplemented with kanamycin (30 μg/mL) and ampicillin (100 μg/mL) antibiotics and maintained at 37 °C with 200 rpm shaking. At an OD₆₀₀ of 0.8, the cultures were supplemented with ferrous ammonium sulfate (0.3 mM) and the protein expression was induced by the addition of 1 mM IPTG, followed by overnight incubation at 18 °C with 200 rpm shaking. The cell pellets (~35 g) were harvested the next day and stored at −80 °C.

To isolate SPL with minimal cluster loss, all lysis and purification steps were performed in an anaerobic Coy vinyl glovebox. At a ratio of ~2 mL of buffer per 1 g of cell pellet, lysis was performed at 4 °C via sonication (15 s pulses, 59 s rest, 60% amplitude, 5 min total pulse time) in buffer A (100 mM Tris pH 8, 300 mM NaCl, 5% glycerol, 20 mM imidazole) in the presence of 1% Triton X-100 (w/v), MgCl₂ (10 mM), PMSF (1 mM), lysozyme (0.5 mg/cell g), and DNase I and RNase A (<1 mg/cell g). After centrifugation (38 000g, 60 min, 4 °C), the SPL-clarified lysates were purified via FPLC on two 1 mL inline HisTrap columns with the following step gradient: (1) 0% Buffer B (100 mM Tris pH 8, 300 mM NaCl, 5% glycerol, 500 mM imidizole) wash for 10 column volumes (CV); (2) 100% Buffer B to elute SPL. After brief centrifugation, the enzyme solutions were gel filtered with Buffer C (100 mM Tris pH 8, 300 mM NaCl, 5% glycerol) (Sephadex G-25 resin column, 75 mL) and reconstituted to increase [4Fe-4S] cluster content; these iron–sulfur cluster reconstitutions were performed at room temperature as described elsewhere.²⁶ Briefly, purified SPL (200 μM) was incubated with DTT (5 mM) for 5 min in Buffer C before ferrous ammonium sulfate (1.2 mM final concentration, from a 10 mM stock) and sodium sulfide (1.2 mM final concentration, from a 10 mM stock) were added in small aliquots (10 μL) over the course of 1 h. After a 3 h incubation at 21 °C, the reconstitution solutions were centrifuged, desalted (Sephadex G-25 resin column, 75 mL) with Buffer C, and concentrated to <60 μM (Amicon 30 kDa MWCO centrifugation filters). The final irons/protein were determined to be 3.7 ± 0.1 for WT, 3.3 ± 0.2 for Y98F, 3.8 ± 0.2 for C140G, 3.3 ± 0.1 for C140A, 3.0 ± 0.2 for C140A/V172C, and 3.2 ± 0.2 for C140A/V172I.

The R-spore photoproduct was prepared as previously described.^17,18,27,28

Sample Preparation.

Hand-Quenched Sample Preparation.

For the 5 min hand-quench experiment, 60 μM SPL samples were reduced with 2 mM sodium dithionite for 5 min, within an MBraun glovebox (O₂ ≤ 1 ppm), before adding 1 mM SAM. The samples were then briefly centrifuged, transferred to EPR tubes (Wilmad LabGlass, 4 mm OD, NJ, USA), capped with rubber septa, and then flash frozen after 5 min in liquid nitrogen outside the MBraun chamber.

For 10 s to 1 min hand-quench experiments, 500 μM SPL samples were reduced with 3 mM sodium dithionite for 5 min within a Coy anaerobic chamber. The samples were then briefly centrifuged and transferred to Q-band tubes. Five millimolar SAM is quickly injected into the tube with a syringe, mixed for 10 s or 1 min, and then flash frozen in an isopentane bath cooled with liquid nitrogen inside the Coy chamber.

RFQ Sample Preparation.

For all RFQ experiments, the same sample preparation was carried out in an anaerobic Coy chamber. The 500 μM SPL sample is reduced with 3 mM sodium dithionite for at least 5 min in 50 mM HEPES pH 7.5, 280 mM NaCl buffer. A SAM mixture containing 5 mM SAM and 3 mM sodium dithionite in the former buffer was prepared separately. This SAM mixture can also contain 500 μM R-SP.

RFQ Experiments.

Rapid freeze–quench experiments were performed with a System 100 apparatus from Update Instrument. Both sample and substrate loops connected to syringes were washed several times with 100 mM dithionite solution and then dried with N₂ gas before loading samples. All samples were loaded in the anaerobic chamber. SPL is dithionite reduced to generate the catalytically relevant [4Fe-4S]⁺ state of the cluster and was loaded into one loop while SAM was loaded into the other. The protein concentrations used were to achieve a ratio of SPL:SAM of 1:10 after mixing. To ensure all protein exited the loop into the mixing chamber and onto the copper wheels, 100 mM dithionite buffer (in water) was placed on each side of the protein. A pocket of N₂ gas was introduced in the loops between the dithionite buffer and the protein. Dithionite was used to react with any oxygen to help protect the protein since the RFQ instrument was not housed in an anaerobic chamber. The mixture was quenched by spraying onto two rotating copper wheels cooled to liquid nitrogen temperatures as previously described^29,30 after 15–500 ms mixing times. The frozen powder was collected in a funnel and packed into precision Q-band tubes (2.5 mm OD) for EPR analysis.

EPR Measurements.

X-Band CW EPR spectroscopy was performed at Montana State University using a Bruker EMX spectrometer fitted with a ColdEdge (Sumitomo Cryogenics) 10 K waveguide in-cavity cryogen-free system with an Oxford Mercury iTC controller unit and helium Stinger recirculating unit (Sumitomo Cryogenics, ColdEdge Technologies, Allentown, PA). Helium gas flow was maintained at 100 psi. Typical spectral parameters were as follows: 2 mW microwave power, 100 kHz modulation frequency, and 6 G modulation amplitude. Spectra were averaged over 10 scans. X-band CW EPR spectroscopic measurements conducted at Northwestern University used a Bruker ESP 300 spectrometer equipped with an Oxford Instruments ESR 910. Typical experimental parameters were at 12 and 40 K, 9.38 GHz, and 10 G modulation amplitude. EPR simulations were performed with the EasySpin^5.2.23 program operated within Matlab.³¹

DFT Calculations.

All density functional theory (DFT) computations were carried out with ORCA 4.0.1.³² The initial geometry for the valyl radical was generated by extracting valine 172 coordinates from the crystal structure of SPL (PDB 4FHD, 2.0 Å resolution).²⁵ Hydrogens were added to make the molecule a neutral valine with the hydrogen from the beta carbon absent in order to generate the radical. The isoleucyl radical was created using the above-described valine and adding an additional methyl group to the gamma carbon of V172. For the two radicals, geometry and single-point calculations utilized the spin-unrestricted B3LYP/G Hybrid functional^33-35 and the Ahlrichs’ valence triple-ξ with a polarization function basis set.³⁶ The molecular orbitals were visualized as Gaussian cubes in Pymol with an isosurface density of 0.08 au. Hyperfine and g tensors were calculated by the coupled–perturbed self-consistent field (SCF) approach as implemented in ORCA 4.0 using the B3LYP hybrid functional and EPR-III basis³⁷ in combination with the accurate spin–orbit coupling operator [RI-SOMF(1X)].³⁸ The EPR calculations were carried out in BP86 in parallel for comparison.

RESULTS

Thermal Stability of Ω.

When reduced GtSPL containing the [4Fe-4S]¹⁺ cluster is mixed with SAM and freeze trapped at 15 ms after mixing, surprisingly the cluster signal (g = [2.03, 1.93, 1.89]) is completely gone and the Ω intermediate has already fully formed, as shown for the C140A/V172I variant in Figure 2; this indicates that the cluster has transferred its electron to SAM with attendant S─C5′ homolytic bond cleavage. Thus, not only is SAM rapidly reductively cleaved in GtSPL without the presence of bound substrate but also Ω forms without substrate as well.

Figure 2.

Time course of Ω decay during increasing freeze–quench delay after mixing GtSPL (C140A/V172I) + SAM in the absence of substrate. X-Band EPR conditions: microwave frequency, 9.37 GHz; modulation amplitude, 5 G; T = 40 K.

The stability of the Fe─C5′ bond of Ω at ambient temperature was assessed by monitoring the Ω signal while varying the freeze–quench delay after mixing, Figure 2. The decay of Ω embodied in this figure can be described by an exponential with decay time τ ≈ 5–6 s (Figure S1).

To further probe the stability of Ω, a sample freeze trapped at 15 ms in the absence of substrate was annealed for 1 min at each of progressively higher temperatures from 150 to 250 K (Figure S2). Annealing at 150 K for 1 min sharpens the Ω spectrum without loss of signal, presumably the result of slight structural relaxation, but 1 min annealing steps at 170 and 190 K begin to cause signal loss, and with a 1 min step at 210 K roughly 3/4 of Ω has been lost. The intermediate completely disappears at 250 K without appearance of detectable radical product, which indicates that the majority of Ω is lost in uncharacterized reactions.

Radical Formed during Prolonged Turnover of GtSPL + SAM.

While following the loss of Ω at longer quench times we observed that when the reaction time became long enough to allow multiple SAM turnovers, a small amount of a radical product began to appear. To identify this product, reduced GtSPL containing the [4Fe-4S]¹⁺ cluster was mixed with SAM at ambient temperature, incubated for up to 10 min, and then freeze trapped by hand, generating samples that have undergone multiple turnovers. Surprisingly, an EPR signal grew in, Figure 3, which clearly arises neither from the [4Fe-4S]¹⁺ cluster nor from the organometallic Ω intermediate. Instead, the enzyme displays a well-defined, strong EPR signal with essentially isotropic g ≈ 2.0 (measured at X-band) and rich proton hyperfine splittings, characteristic of a hydrocarbon radical, Figure 3. This spectrum is not that of homolytically cleaved 5′-dAdo^• itself, as this new signal is broader and with a more complex hyperfine structure (Figure 3).¹¹ Neither is this the spectrum of homolytically cleaved ^•CH₃, which in an RS active site is freely rotating and exhibits a simple 1:3:3:1 quartet created by hyperfine couplings to three equivalent H (Figure 3).³⁹ Finally, it is not associated with the amino-carboxypropyl (ACP) radical liberated by homolytic cleavage of the S─Cγ bond of SAM, as the spectrum is unchanged by use of d₄-SAM deuterated on the β, γ methylene carbons of the Met component of SAM (vide infra). Therefore, we postulated that during the extended substrate-free enzymatic turnover before hand quench, reductive cleavage of SAM indeed generates Ω, which in turn liberates 5′-dAdo^• for H-atom abstraction. In the absence of substrate, the reactive 5′-dAdo^• can undergo a side reaction with some small probability per turnover and abstraction of a hydrogen from a nearby protein residue to form a relatively stable protein-based radical.

Figure 3.

X-Band EPR spectra of WT GtSPL freeze–quenched after 10 min incubation with SAM to allow for accumulation of the unknown radical. Also shown are spectra for 5′-dAdo^• generated through photolyzing (PFL-AE + SAM),¹¹ and ^•CH₃ generated through photolyzing (HydG + SAM).³⁹ EPR conditions: microwave frequency, 9.37 GHz; modulation amplitude, 5 G; T = 40 K.

To identify the residue harboring the stable radical, we examined the active site vicinity of substrate-free GtSPL (PDB 4FHD)²⁵ for residues that would be close to the SAM C5′ upon generation of the 5′-dAdo^•. As can be seen in Figure 4, the substrate-free structure of SPL reveals that the closest amino acid to SAM C5′ is Val 172 with its Cγ at 4.3 Å from C5′. The next-closest residue is Tyr 98 with the ortho-C of the phenol side chain at 4.7 Å. Also nearby is Ala 140 (7.6 Å), which is a Cys residue in the WT GtSPL. Previous work has implicated both Cys 140 and Tyr 98 as key residues involved in a hydrogen-atom transfer pathway during repair of SP to an undamaged TT sequence.^20,21,40 On the basis of this analysis of proximity in the active site as well as evidence for biological function of the enzyme, we used site-directed mutagenesis to probe the possibility that the radical formed during prolonged reaction of GtSPL with SAM resides at one of these residues.

Figure 4.

(Left) Residues in the vicinity of the 5′-C of bound SAM (dark gray) in the active site of GtSPL C140A (4FHE); key residues discussed in the text are shown in teal with distances of closest approach to the 5′C of SAM shown. (Right) Active site of SPL in the substrate-bound state (4FHF) shows the 3.7 Å distance to the site of H-atom abstraction on substrate (purple).

The radical observed after prolonged (~10 min) reaction of WT GtSPL with SAM is unchanged in the C140A, C140G, and Y98F variants of GtSPL (Figure 5), indicating that this radical signal arises neither from a tyrosyl radical at residue 98 nor from a cysteinyl radical at position 140. Observation of a radical at either of these positions would have been consistent with the proposed mechanistic roles for these residues;^21,22 however, the radical that accumulates as described herein is clearly at neither of these amino acid residues.

Figure 5.

X-Band EPR spectra of the indicated GtSPL variants (black), hand-quenched after 10 min reaction with excess SAM and reductant. Each spectrum is labeled with the radical identified via simulations (red) using parameters given in Table 2. EPR conditions: microwave frequency, 9.37 GHz; modulation amplitude, 5 G; T = 40 K.

We next turned our attention to Val172, the residue closest to C5′ of SAM, as the potential site of the stable protein radical resulting from “nonproductive” SAM cleavage in GtSPL. As shown in Figure 4, a methyl of the valine side chain is only 4.3 Å from the SAM C5′ in the substrate-free SPL structure; this distance is similar to the distance of substrate carbons (or nitrogen) to C5′ in radical SAM enzymes. For example, structurally characterized RS enzymes with substrate bound in the active site show that the sites of H-atom abstraction are 3.7–4.1 Å from SAM 5′C (Table 1, Figure 4).^25,41-46 Thus, the close proximity of Val172 in the substrate-free GtSPL structure makes it a reasonable “alternative substrate” for turnover under substrate-free conditions.

Table 1.

Distances from SAM C5′ to the Site of H-Atom Abstraction on Substrate in RS Enzymes

RSE	PDB code	distance (Å)	ref
SPL	4FHF	3.7	25
BioB	1R30	3.7	41
NosL	4R33	3.7	42
LAM	2A5H	3.8	43
PFL-AE	3CB8	3.9	44
QueE	4NJI	4.0	45
anSME	4K38	4.1	46

To test this idea, we subjected the double mutants C140A/V172I and C140A/V172C to the same extended-turnover experiments (the mutation at C140 increases the solubility of the enzyme without altering the function or spectroscopic properties, Figure 5). Each mutation changes the signal of the radical accumulated over multiple SAM turnovers. Incubating SAM with reduced C140A/V172I generates a hydrocarbon-radical EPR spectrum with fewer peaks and smaller overall hyperfine breadth than that of WT or C140A enzyme (Figure 5). Incubation of SAM and mutant C140A/V172C produces yet a different paramagnetic species, one whose g anisotropy (g_∥ = 2.094, g_⊥ ≈ 1.99) is indicative of a cysteine thiyl radical (Cys^•, Figure 5).⁴⁷

Together, these results demonstrate that residue 172 is the site of the stable protein radical observed in WT enzyme under extended substrate-free turnover conditions, and in the next subsection, we describe the analysis that leads to the simulation of its spectrum as well as the spectra generated when V172 is replaced by I or C (Figure 5). We propose that in substrate-free GtSPL variants, the 5′-dAdo^• liberated by thermal Fe─C5′ homolysis of the Ω intermediate undergoes a low-probability side reaction in which it abstracts an H atom from the side chain of residue 172 to form a stable radical, Val^•, Ile^•, or Cys^•, depending on the variant of the enzyme (Scheme 1). Double integration of the Ω and amino acid radical EPR signals indicates that approximately 5% of Ω goes on to form a radical at residue 172 with the majority being quenched by other pathways. This ~5% efficiency is further supported by evaluating the reported kinetics of GtSPL turnover²² relative to the intensity of the residue 172 radical EPR signal.

Scheme 1.

Reaction Sequence Forming the Val172 Radical

Structures of Val172^•, Ile172^•, and Cys172^•.

Val172^•.

Assignment of the radical accumulated in WT GtSPL and variants having valine at position 172 as a valine side chain radical is confirmed, and its structure is established by simulation of its spectrum, Figure 5. The Cγ of Val172 sits closest to the C5′ of SAM, but the tertiary Cβ of the valine side chain is the most susceptible to H-atom abstraction due to its weaker C─H bond and due to the fact that it would give rise to a more stable tertiary radical. We therefore initially evaluated the more stable Cβ radical as the first candidate for the origin of the observed signals of a planar 2pπ radical on the sp²-hybridized tertiary carbon of Cβ (Scheme 2). Such a radical would exhibit proton hyperfine couplings to the six H atoms of the two side-chain methyls, C_γ1─H₃ and C_γ2─H₃, and the Cα─H proton, all of which are “β” to the half-occupied 2pπ orbital on Cβ. β-Proton hyperfine couplings are essentially isotropic with a dependence on the spin density in the carbon radical 2pπ orbital, denoted ρ_π, that obeys the semiempirical relationship

$$a_{{}_{\text{iso}}}\approx {\rho }_{{}_{\pi }}B\langle {\cos }^2\phi \rangle \text{MHz}$$ (1)

where φ is the dihedral angle between the 2pπ orbital and the C─H bond of the β-proton, while the constant B represents the transmission of spin to the β-¹H through hyperconjugation; canonical values are ρ_π ≈ 0.7 and ρ_πB ≈ 140 MHz.⁴⁸

Scheme 2.

To analyze the observed radical spectra we turned to DFT to generate an energy-minimized conformation and estimate for the hyperfine couplings. As a first approximation, we examined the neutral Cβ 2pπ radical in vacuo, namely, uninfluenced by interactions with the surrounding active-site protein scaffold. Starting with the conformation of the crystal structure (PDB 4FHD) and abstracting an H atom from Cβ, the computations yielded a 2pπ radical on Cβ and a planar propyl group with a dihedral angle between the 2pπ orbital and the C_α─H bond of φ = 62°. For the C_α─H coupling the calculation gave a_iso(H) = 48 MHz and a 2pπ spin density of ρ_π ≈ 0.7, thus validating eq 1. However, simulations using this C_α─H coupling showed a discrepancy with experiment that was simply eliminated by slightly reducing the dihedral angle from φ = 62° to φ = 55° (Figure S3). According to eq 1, this increases the C_α─H coupling to a_iso(C_α─H) = 69 MHz. The protons on both Cγ had the same orientation relative to the 2pπ radical orbital, and the calculation (B3LYP) gave the same three distinct ¹H hyperfine couplings for the two methyls a_iso(H_a,H_b,H_c) = 127, 65, 14 MHz.(Table S1) Simulations of the EPR spectrum assuming rotating Cγ methyls and thus equivalent methyl-¹H with a_iso ≈ ρ_πB⟨cos²φ⟩MHz = ρ_πB/2 MHz were slightly better than those assuming the γ-methyls were static with inequivalent ¹H couplings. The resulting simulations using the dihedral angle φ = 55° and corresponding C_α─H coupling (Table 2) match the experimental spectrum within error, Figure 5.

Table 2.

EPR Simulation Parameters for Val172^•, Ile172^•, and Cys172^• from GtSPL

		hyperfine coupling aiso (MHz)
radical	g
Val172•	2.003, 2.003, 2.002	Cα(H) 69	Cγ1(Ha,Hb,Hc) 65	Cγ2(Ha,Hb,Hc) 65
Ile172•	2.005, 2.005, 2.005	Cα(H) 5	Cγ1(Ha,Hb) 63, 10	Cγ2(Ha,Hb,Hc) 52
Cys172•	2.094, 1.998, 1.991		Cγ(Ha,Hb) 90, 50

Ile172^•.

We also used DFT to compute β-proton hyperfine coupling of the radical located on C_β of Ile172^• (Table S1). This EPR spectrum is well simulated assuming C_γ2H₃ is rotating, giving isotropic couplings from three equivalent ¹H but with inequivalent couplings from the locked C_γ1 (H_a,H_b). The in vacuo DFT gives a C_β-C_α torsion angle for the Ile172^• radical that is the same as that for the starting point of V172^•, but simulations of the Ile172^• radical spectrum require that the C_α─H coupling is near zero, which according to eq 1 implies a dihedral angle between the 2pπ radical orbital and the C_α─H bond of φ ≈ 90° (Figure 6, Table 2, and Figure S4). We postulate that the altered conformation with φ ≈ 90° for the isoleucine variant versus φ ≈ 55° determined here for the radical on the native valine residue occurs because steric clashes between the “elongated” Ile residue and other residues within the protein scaffold cause a twist around the C_α─C_β bond; this analysis is confirmed by modeling these radicals in the GtSPL active site (Figure S5). Figure 6 visualizes the DFT models of these two radical residues with φ for Ile172^• as manually adjusted to match experiment and also displays the C_β 2pπ SOMO.

Figure 6.

Computational models of V172^• and I172^•; models pictured here have the C_β─C_α torsion angles manually adjusted to agree with the experimental results (Val172^•, φ ≈ 55°; Ile172^•, φ ≈ 90°) as described in the text. Dashed circle denotes the rotation of the methyl group of V172^• and I172^•. 2pπ radical orbitals are visualized in Pymol with an isodensity of 0.08 au.

Cys172^•.

Finally, the large g anisotropy of the thiyl radical is attributed to sulfur spin–orbit coupling and p–d orbital mixing; this radical has been well investigated with DFT computations by Neese and co-workers.⁴⁷ The poorly resolved hyperfine structure in the EPR spectrum of the C172 thiyl radical in the V172C variant is assigned to its germinate pair of β-protons, ─C_γ(H_a, H_b)─S^•. Our best simulation is achieved with g = [2.094 1.998 1.991] and a_iso(H_a, H_b) = (50, 90) MHz.

DISCUSSION

Radical SAM enzymes use a [4Fe-4S]⁺/SAM complex in the active site to initiate diverse radical reactions via the reductive cleavage of SAM. In the majority of radical SAM enzymes studied to date, when substrate is absent the complex is relatively stable and only undergoes slow “nonproductive” reductive cleavage of SAM; this stability has allowed formation and detailed structural and functional characterization of the [4Fe-4S]⁺/SAM complex by EPR and ENDOR spectroscopies.^4,5,7 SPL, however, behaves somewhat differently, with the EPR signal of the [4Fe-4S]⁺ cluster rapidly vanishing upon addition of SAM, even in the absence of substrate.¹⁸

Here, we asked the question: if SAM rapidly undergoes reductive cleavage by SPL in the absence of substrate, does it do so via Ω and would we be able to observe Ω thus formed in the absence of substrate? Our results clearly show that in GtSPL, nonproductive reductive cleavage of SAM in fact generates Ω, and we used this unique behavior to answer questions about the properties and behavior of the organometallic intermediate Ω. We find that in the absence of substrate, Ω in frozen solution is unstable above 170 K and that at ambient temperature it has a decay time of 5–6 s at RT.

The degree of lability of the Fe─C5′ bond of Ω is in sharp contrast with that of the analogous Co─C5′ bond of B₁₂, which is indefinitely stable until enzymatically activated for bond cleavage. However, this decay time is surprisingly long given that catalytic homolytic Fe─C5′ bond cleavage in Ω to liberate 5′-dAdo^• for reaction with substrate must be more rapid than this to account for known RS turnover rates. This finding thus suggests that the catalytic Fe─C5′ bond cleavage in the presence of substrate, like that of B₁₂, is enzymatically activated. Indeed, the Fe─C5′ bond of the [4Fe-4S]³⁺─dAdo Ω intermediate is far more labile than that of a [4Fe-4S]²⁺–ethyl model complex that was purified and crystallized at ambient temperature.²⁴ A recent report has shown that Fe─C bond homolysis in synthetic [4Fe-4S]–alkyl complexes is promoted by ligand addition, suggesting an effect of coordination number on Fe─C bond lability that is consistent with the higher reactivity of Ω,⁴⁹ both with and without substrate. Another factor that could impact the lability of the Fe─C bond in Ω relative to that in synthetic models is the oxidation state difference. The role of the enzymatic activation of Ω will be illuminated by studies with site-directed mutants and undoubtedly by further model studies and quantum-chemical computations.

While carrying out these experiments, we found that when quenching on the seconds time-scale where Ω had almost completely decayed, we detected formation of a relatively low-intensity product radical corresponding to a small fraction of the original Ω species. This indicates that most of the Ω intermediate reacts nonspecifically in the absence of substrate, likely producing species that are rapidly quenched and therefore not observed by EPR. This low-intensity species could be accumulated to high levels if SPL was allowed to undergo multiple SAM turnovers over the course of minutes, and this allowed us to use a combination of isotopic labeling of SAM and site-directed mutagenesis of SPL to unequivocally identify the site of the stable radical as Val172, which sits in the active site only 4.3 Å from C5′ of SAM.

While it is surprising to see a previously unreported stable valine radical accumulate in SPL under substrate-free turnover conditions, the proximity of this residue to the bound SAM in the active site (Figure 4) is consistent with its ability to serve as an alternative site of H-atom abstraction. The ability of a valine side chain to form a relatively stable tertiary radical explains its accumulation over multiple turnovers. The Cγ of Val172 is nearly as close to the SAM C5′ as are the substrate sites of H-atom abstraction in RS enzymes (Table 1). However, as can be seen in the right panel of Figure 4, Val172 sits in a different position than the substrate C6, from which an H atom is abstracted during reaction with substrate and therefore is presumably not in the direct “line of fire” of the C5′ radical of 5′-dAdo^•. This spatial positioning likely contributes to the low probability for H-atom abstraction from Val172. Our results furthermore show that when other residues are substituted into position 172 they can also undergo this low-probability H-atom abstraction to accumulate a stable protein-based amino acid radical. When Val172 is replaced with isoleucine, a stable isoleucine radical accumulates under substrate-free turnover conditions; when it is replaced with a cysteine, a stable cysteinyl radical accumulates under these conditions.

The isoleucine radical in the V172I variant is particularly interesting as its EPR signal is essentially identical to that of a radical previously reported by Heidinger et al. to accumulate in GtSPL during prolonged incubation of the reduced enzyme with SAM (Figure S6).⁵⁰ It was claimed that this previously observed radical species was a 5′-dAdo^• radical,⁵⁰ but no SAM isotopologs were used to support this assignment, and the EPR spectrum disagrees with the unequivocally identified 5′-dAdo^• reported recently by our groups¹¹ as well as by Britt and co-workers,⁵¹ and the radical was recognized as being misassigned.¹¹ Given the similar experimental approaches employed here for generating that radical and its similar EPR spectrum to that of the isoleucine residue radical in our V172I variant, we conclude that Heidinger et al. were observing the accumulation of an isoleucine radical in their SPL.

In conclusion, the RS enzyme SPL has the unusual property of undergoing relatively rapid reductive cleavage of SAM in the absence of substrate. On examination of this reaction in GtSPL, we have shown that SAM cleavage generates the organometallic intermediate Ω, whose Fe─C5′ bond homolytically cleaves to release the 5′-dAdo^• radical with a decay time of τ ≈ 5–6 s at room temperature, which is surprisingly long, yet represents far greater lability than found in a model complex²⁴ and even more dramatically greater than that of the Co─C5′ bond of B₁₂.⁵² In the absence of substrate this radical can undergo a low-probability conversion to a stable protein radical at residue 172 in the active site with the structures of the radical in WT, V172I, and V172C variants having been established by detailed EPR/DFT analyses.

Whether the radical observed at residue 172 is functionally involved in the previously proposed radical relay in GtSPL awaits further studies. Our results, however, do establish that this RS enzyme can act on a nearby amino acid residue side chain when the native substrate is not available. Such a reaction would presumably be detrimental in the absence of substrate, and in fact, SPL is unusual in the RS superfamily for its propensity to undergo rapid reductive cleavage of SAM in the absence of substrate; for most other RS enzymes studied to date, SAM can be added to the reduced [4Fe-4S]⁺ state to form a relatively stable complex, which turns over appreciably only when substrate is added. This distinction suggests that while many RS enzymes’ catalytic turnover is triggered by substrate binding, SPL turnover may instead be gated by addition of an electron only after substrate is bound. Such “reductive gating” would make sense for an enzyme such as SPL, which must bind its substrate DNA to search for the SPL lesion, and only when it finds the SP would reductive cleavage of SAM be beneficial.