Mechanism of Reduction of an Aminyl Radical Intermediate in the Radical SAM GTP 3′,8-Cyclase MoaA

Abstract

The diversity of the reactions catalyzed by radical S-adenosyl-l-methionine (SAM) enzymes is achieved at least in part through the variety of mechanisms to quench their radical intermediates. In the SPASM-twitch family, the largest family of radical SAM enzymes, the radical quenching step is thought to involve an electron transfer to or from an auxiliary 4Fe-4S cluster in or adjacent to the active site. However, experimental demonstration of such functions remains limited. As a representative member of this family, MoaA has one radical SAM cluster ([4Fe-4S]_RS) and one auxiliary cluster ([4Fe-4S]_AUX), and catalyzes a unique 3′,8-cyclization of GTP into 3′,8-cyclo-7,8-dihydro-GTP (3′,8-cH₂GTP) in the molybdenum cofactor (Moco) biosynthesis. Here, we report a mechanistic investigation of the radical quenching step in MoaA, a chemically challenging reduction of 3′,8-cyclo-GTP-N7 aminyl radical. We first determined the reduction potentials of [4Fe-4S]_RS and [4Fe-4S]_AUX as −510 mV and −455 mV, respectively, using a combination of protein film voltammogram (PFV) and electron paramagnetic resonance (EPR) spectroscopy. Subsequent Q-band EPR characterization of 5′-deoxyadenosine C4′ radical (5′-dA-C4′●) trapped in the active site revealed isotropic exchange interaction (~260 MHz) between 5′-dA-C4′● and [4Fe-4S]_AUX¹⁺, suggesting that [4Fe-4S]_AUX is in the reduced (1+) state during the catalysis. Together with density functional theory (DFT) calculation, we propose that the aminyl radical reduction proceeds through a proton-coupled electron transfer (PCET), where [4Fe-4S]_AUX serves as an electron donor and R17 residue acts as a proton donor. These results provide detailed mechanistic insights into the radical quenching step of radical SAM enzyme catalysis.

Graphical Abstract

INTRODUCTION

Radical SAM enzymes catalyze chemically challenging biological reactions by free radical-mediated mechanisms initiated by a highly reactive primary radical, 5′-deoxyadenosine radical (5′-dA●), which is transiently generated by reductive cleavage of S-adenosyl-l-methionine (SAM, 1) using a reduced [4Fe-4S]¹⁺ cluster.¹ While the use of 5′-dA● may be compared to adenosylcobalamin-dependent enzymes, one of the unique aspects of radical SAM enzymes is the diversity in the mechanism of radical quenching. Adenosylcobalamin-dependent enzymes catalyze reductive quenching of radical intermediates using 5′-dA to catalyze net redox neutral reactions. In contrast, the radical quenching step in radical SAM enzymes can be either reductive or oxidative. This flexibility allows the evolution of various radical chemistry catalyzed by these enzymes. However, despite being a critical step, the radical quenching mechanisms in most radical SAM enzymes remain ambiguous and lack experimental evidence.²

Radical SAM enzymes in the SPASM-Twitch family, the largest family in the radical SAM superfamily, are characterized by an extended C-terminal domain harboring one (Twitch subfamily) or two (SPASM subfamily) auxiliary [4Fe-4S] clusters ([4Fe-4S]_AUX).^3,4 Although [4Fe-4S]_AUX has frequently been proposed to have redox functions,^5–10 the experimental demonstration has been difficult. The major challenge is connecting the redox functions of these [4Fe-4S]_AUX to the catalytic mechanisms. So far, the best-characterized systems are SPASM subfamily enzymes responsible for modifying ribosomally synthesized peptide natural products (RiPPs). Among such enzymes, SCIFF maturase and MftC were reported to have two [4Fe-4S]_AUX clusters with ~50 and ~100 mV more negative reduction potentials than that of [4Fe-4S]_RS.^11,12 However, these potentials have not yet been correlated to the function of [4Fe-4S]_AUX as an electron acceptor. More recently, a study on another SPASM subfamily member SuiB provided experimental evidence for the redox function of auxiliary cluster.¹³ SuiB catalyzes a C–C bond formation reaction between Lys2 and Trp6 on its substrate SuiA. In this study, the cross-linked Lys-Trp radical intermediate was trapped and shown to reduce the auxiliary cluster in the frozen solution.

Little is known about the catalytic functions of [4Fe-4S]_AUX in other radical SAM enzymes, especially those in the Twitch subfamily or those with the proposed [4Fe-4S]_AUX function as an electron donor. Here, we focused on the redox function of the [4Fe-4S]_AUX in MoaA, one of the founding members of the Twitch subfamily. MoaA catalyzes the first step of the Moco biosynthesis by transforming GTP (2) into a unique cyclic nucleotide intermediate, 3′,8-cH₂GTP (3).⁵ This transformation was proposed to proceed through an H-3′ abstraction from GTP by 5′-dA●. Subsequent addition of C3′ radical on GTP (GTP-C3′●, 4) to C-8 of guanine generates a 3′,8-cyclo-GTP-N7 aminyl radical (3′,8-cH₂GTP-N7●, 5, Figure 1). The 3′,8-cH₂GTP-N7● was then proposed to be reductively quenched by an electron transfer from reduced [4Fe-4S]_AUX and a proton transfer from an unknown donor to form 3′,8-cH₂GTP. Previously, we performed detailed characterization of MoaA catalysis and found a shunt pathway that accumulates a 5′-deoxyadenosyl-C4′ radical¹⁴ (5′-dA-C4′●, 6, Figure 1). Detailed characterization of this shunt pathway and the normal pathway allowed determination of the rate constant of the rate-limiting step of the transformation of GTP-C3′● to 3′,8-cH₂GTP, which is likely the C3′ radical addition to C8. The results were consistent with MoaA to accelerate the addition of GTP-C3′● to C-8 of guanine by 6–9 orders of magnitude through conformational restraint of GTP and transition state stabilization by the positively charged R17 residue. However, the mechanism by which 3′,8-cH₂GTP-N7● is reduced to 3′,8-cH₂GTP remained ambiguous.

Figure 1.

Proposed mechanism of MoaA catalysis.

The reduction of 3′,8-cH₂GTP-N7● is expected to be chemically challenging. In DNA, 5′,8-cyclization of 2′-deoxyguanosine proceeds through oxidation of 5′,8-cH₂-guanosine-N7● and specifically yields 5′,8-cycloguanosine.¹⁵ Therefore, the reduction of 3′,8-cH₂GTP-N7● into 3′,8-cH₂GTP in MoaA requires a specific mechanism. Here, we report the first mechanistic characterization of the MoaA-catalyzed 3′,8-cH₂GTP-N7● reduction using a combination of PFV, X- and Q-band EPR spectroscopy, and DFT calculation. Together, the observations suggest that the 3′,8-cH₂GTP-N7● reduction proceeds with a proton-coupled electron transfer (PCET) mechanism with [4Fe-4S]_AUX as an electron donor and R17 as a proton donor (Figure 1). These characterizations suggest that the presence of the electron donor with finely tuned reduction potential and an appropriate proton donor enables MoaA to catalyze the otherwise chemically challenging reduction of 3′,8-cH₂GTP-N7● into 3′,8-cH₂GTP. Therefore, these results further extend our understanding of enzyme-catalyzed radical reactions in general.

RESULTS

PFV Characterization of 4Fe-4S Clusters.

To assess the redox function of the [4Fe-4S]_AUX in MoaA, we characterized wt-MoaA with PFV. Cyclic voltammetry at a pyrolytic graphite edge (PGE) electrode revealed an electrochemical current centered at −460 mV vs SHE (Figure S1A). Modification of the PGE electrodes with multiwalled carbon nanotubes (MWCNTs) previously used for radical SAM enzymes improved the magnitude of the signal (Figure S1B). To confirm the potentials observed in cyclic voltammetry, we characterized wt-MoaA with square wave voltammetry and observed a similarly broad feature at −450 mV (Figure S1C). These analyses revealed only one electrochemical current peak, albeit with a peak width at half height larger than anticipated for a single one-electron redox process. Thus, we also characterized ΔRS-MoaA, a MoaA variant with the three Cys ligands of the N-terminal [4Fe-4S]_RS cluster replaced with Ser (see the subsequent section for the biochemical characterization of ΔRS-MoaA). The cyclic and square wave voltammetry of ΔRS-MoaA all showed a single electrochemical current peak narrower than those observed with wt-MoaA (Figure S1D–S1F). These observations are consistent with the presence of two closely spaced reduction potentials in wt-MoaA with two 4Fe-4S clusters and the presence of a single redox center in ΔRS-MoaA due to the removal of [4Fe-4S]_RS.

The electrochemical data were assigned to be [4Fe-4S]^2+/1+ redox couples. The wt-MoaA cyclic voltammogram was best simulated with two independent one-electron transfers at −455 mV and −510 mV vs SHE (Figure 2A, E_mid in Table 1). In contrast, the ΔRS-MoaA voltammogram was better represented by a singular electron transfer at −485 mV (Figure 2B, E_mid in Table 1). The full-width at half-height (fwhm) was larger for wt-MoaA compared to ΔRS-MoaA (140 vs 97 mV), which also supports that wt-MoaA has two redox-active cofactors with close reduction potentials (−455 mV and −510 mV) while ΔRS-MoaA has only one with −485 mV. Notably, the data collected at MWCNT-modified PGE electrodes was too broad to fit compared to bare PGE, suggesting that interfacial electron transfer kinetics were fundamentally slow. We also determined the reduction potentials of wt-MoaA at pH 5.5–9.5 (Figure S2), which suggested the absence of amino acid residues close to either of the two clusters with pK_a in the tested range. This observation is consistent with the reported structures of wt-MoaA, in which no amino acid residues within 5 Å from each cluster would have pK_a between 5.5–9.5.

Figure 2.

Cyclic voltammograms of wt- (A) and ΔRS-MoaA (B). The experimental voltammograms (dotted lines) were fit to two independent one-electron transfers (red) with the sum of these fits (blue) for wt-MoaA or one-electron transfer (red) for ΔRS-MoaA.

Table 1.

Reduction Potentials of wt- and ΔRS-MoaA Based on the Fittings Shown in Figure 2

	wt-MoaA (mV vs SHE)		ΔRS-MoaA (mV vs SHE)
E ox a	−484	−427	−469
E red b	−537	−482	−502
E mid c	−510	−455	−485
fwhm	140		97

^aE_ox represents the cathodic peak potential of the simulated voltammogram.

^bE_red is the anodic peak potential of the simulated voltammogram.

^cE_mid (midpoint potential) is the mean of E_ox and E_red.

The presence of GTP or SAM minimally affected the wt-MoaA voltammograms (Figure S3A–C, Table S1), while in ΔRS-MoaA, a shoulder was observed in the presence of GTP, suggesting a small shift of the reduction potential by GTP binding (Figures S3D–F, Table S1). Quantitation of this shift is challenging due to the small extent and the apparent incompleteness of the shift. As described below, our EPR characterization of MoaA suggests that the GTP binding to MoaA causes a protein conformational change and a slight shift in the reduction potential of [4Fe-4S]_AUX. Depending on how the protein is adsorbed on the electrode surface, such protein conformational change could be partially inhibited, and therefore the reduction potential shift may also be partially inhibited. Accordingly, our PFV observations qualitatively suggest that the redox potential of [4Fe-4S]_AUX shifts to the positive side upon GTP binding (Table S1).

X-Band EPR Characterization of 4Fe-4S Clusters.

PFV showed two reduction potentials very close to each other (−510 mV and −455 mV). Still, it was not possible to unambiguously assign the two potentials to each of the two clusters based on a comparison with ΔRS-MoaA (−485 mV). Therefore, wt- and ΔRS-MoaA were further characterized by EPR; 150 μM of wt- or ΔRS-MoaA were prereduced with 3 mM sodium dithionite (SDT) for 1 h at 25 °C, mixed with buffer, SAM, GTP, or both SAM and GTP to a final concentration of 100 μM enzyme, 2 mM SDT, and 1 mM substrates, and incubated for another 2 min at 25 °C. The samples were then freeze-quenched in an isopentane slush bath and characterized by X-band EPR at 15 K. Wt-MoaA harbors 1.84 ± 0.02 eq of 4Fe-4S clusters based on Fe quantitation, suggesting >90% overall cluster loading. The wt-MoaA EPR spectra exhibited two axial signals (Figure 3A). In the absence of either substrate, the two species exist at stoichiometric amounts, and the sum of these species was 1.19 ± 0.05 eq per MoaA, suggesting a partial reduction of the 4Fe-4S clusters (Table 2). SAM did not significantly alter the total amount of the reduced clusters or the ratio of the two axial signals. GTP increased the total amounts of the reduced cluster to ~1.3 eq, and significantly shifted the ratio of the two signal intensities (Table 2). The effect of GTP on the total spin concentration (~0.1 eq increase) was small but reproducible between samples with different preparations of MoaA. Under the turnover conditions, the total amount of signals and the ratio of the two species were very similar to that with GTP alone.

Figure 3.

EPR spectra of wt- (A) and ΔRS-MoaA (B) at 15 K. Prereduced wt- or ΔRS-MoaA was incubated with buffer only, SAM, GTP, or both SAM and GTP for 2 min and manually frozen in an isopentane slush bath. The narrower signal was designated as species 1, and the wider signal as species 2. The g values and the ratio of the two signals are listed in Table 2. The feature at g = 2.0 in wt-MoaA with SAM and GTP is from 5′-dA-C4′●.

Table 2.

Simulated g Values and Ratio of the Two Axial Signals from Reduced 4Fe-4S Clusters of wt- and ΔRS-MoaA^a

		species 1		species 2
enzyme	substrate	g ⊥	g//	g⊥	g//	species 1 (%)	total [4Fe-4S]+ (eq of MoaA)b
wt-MoaA		1.900	2.028	1.879	2.060	49.7 ± 2.8	1.19 ± 0.05
	SAM	1.901	2.028	1.879	2.059	50.9 ± 2.9	1.22 ± 0.02
	GTP	1.895	2.030	1.882	2.061	22.9 ± 2.8	1.32 ± 0.06
	GTP+SAM	1.895	2.030	1.882	2.061	22.8 ± 2.8	1.30 ± 0.05
ΔRS-MoaA		1.895	2.029	1.875	2.060	49.1 ± 0.7	0.87 ± 0.005
	SAM	1.895	2.028	1.878	2.059	45.1 ± 1.0	0.85 ± 0.06
	GTP	1.891	2.020	1.881	2.061	14.3 ± 0.5	0.96 ± 0.04
	GTP+SAM	1.890	2.020	1.880	2.062	17.1 ± 2.0	0.96 ± 0.05

^aValues are averages of three replicates, and the errors represent standard deviation.

^bTotal amounts of reduced [4Fe-4S] clusters are quantified by EPR based on 1 mM Cu standard.

ΔRS-MoaA had 0.90 ± 0.04 eq of 4Fe-4S cluster loaded based on Fe quantitation, consistent with almost complete loading of [4Fe-4S]_AUX and no [4Fe-4S]_RS. Since ΔRS-MoaA exhibited no detectable catalytic activity, the binding of SAM and GTP to ΔRS-MoaA was characterized by anaerobic isothermal titration calorimetry (ITC) (Figure S4, Table S2). The K_d of SAM with ΔRS-MoaA was higher than the upper limit of detection for ITC analysis (90 μM), consistent with the absence of [4Fe-4S]_RS. The K_d (3.2 ± 0.8 μM) and ΔH (−15.2 ± 1.5 kcal/mol) of GTP binding to ΔRS-MoaA are comparable to those for wt-MoaA¹⁶ (3.5 ± 0.8 μM and −19.9 ± 2.0 kcal/mol). These characterizations suggest that the structure and the GTP binding capability of ΔRS-MoaA are not significantly perturbed in ΔRS-MoaA compared to wt-MoaA.

Our EPR characterization of ΔRS-MoaA at 15 K revealed the presence of two axial signals very similar to those found in wt-MoaA (Figure 3B, Table 2). Almost stoichiometric amounts of the two species were observed without substrates or with SAM, whereas the wider species became predominant in the presence of GTP. GTP also increased the total amount of [4Fe-4S]¹⁺ EPR signals in ΔRS-MoaA to a degree similar to that in wt-MoaA (~0.1 eq, Table 2). 5′-dA-C4′● was not observed even in the presence of SAM and GTP (compare the bottom spectra in Figure 3A and 3B), consistent with the absence of catalytic activity of this mutant. These observations in ΔRS-MoaA suggest that [4Fe-4S]_AUX exhibits two axial signals. Although the origin of the difference in these two species is not clear, we assigned them as two conformations of [4Fe-4S]_AUX whose ratio is altered by the conformational change of MoaA induced by GTP binding.

The overall line shapes of 4Fe-4S cluster EPR signals and their response to GTP were very similar between wt- and ΔRS-MoaA (Figure 3). The presence of GTP increased the total amount of EPR signals in both wt- and ΔRS-MoaA to a very similar degree (0.1 equiv), while SAM did not significantly affect the shapes or the amounts of EPR signals. The spin quantitation revealed 0.85–0.96 eq of [4Fe-4S]¹⁺ in ΔRS-MoaA, compared to 1.2–1.3 eq in wt-MoaA. The difference in the spin quantity was reproducible (0.35 ± 0.03 equiv) under all four conditions and therefore likely represents the amount of reduced [4Fe-4S]_RS. The [4Fe-4S]_RS signal is not apparent from the line shape, likely due to its low abundance and the significant overlap with [4Fe-4S]_AUX. Thus, these observations suggest that in the presence of SDT, [4Fe-4S]_AUX can be almost fully reduced (85–96%), whereas [4Fe-4S]_RS is reduced only partially (35%), suggesting that [4Fe-4S]_AUX has more positive reduction potential than [4Fe-4S]_RS. Therefore, we assign the reduction potential of −510 mV to [4Fe-4S]_RS and −455 mV to [4Fe-4S]_AUX.

To test our reduction potential assignment, we calculated the reduction potential difference of the two clusters based on the difference in the amount of reduced clusters in wt- and ΔRS-MoaA using the following equation

$$E_2-E_1=\frac{\text{RT}}{\text{nF}}\ln \frac{{[\text{Red}]}_2{[\text{Ox}]}_1}{{[\text{Red}]}_1{[\text{Ox}]}_2}$$

where E₁ and E₂ are the reduction potentials of the two clusters, R is the ideal gas constant, T is the temperature in K, n is the number of electrons, F is Faraday constant, [Red]₁ and [Red]₂ are the concentrations of the reduced state of the two clusters, and [Ox]₁ and [Ox]₂ are the concentrations of the oxidized form. On the basis of the quantitation that 0.35 ± 0.03 eq [4Fe-4S]_RS and 0.86 ± 0.03 eq [4Fe-4S]_AUX are reduced in the absence of GTP, the reduction potential difference between [4Fe-4S]_RS and [4Fe-4S]_AUX is calculated as ~61 mV. This value is consistent with the reduction potential difference observed by PFV (55 mV). Similarly, the GTP-induced increase of the population of the reduced [4Fe-4S]_AUX from 0.86 eq to 0.96 eq corresponds to a ~ 34 mV positive shift of the reduction potential, consistent with the values measured by square wave voltammograms (10–25 mV, Table S1). Therefore, the PFV and EPR data are consistent with each other and provide strong support for the measured reduction potentials and their assignments to the two clusters.

Q-Band EPR Characterization of the Spin–Spin Interaction between 5′-dA-C4′● and [4Fe-4S]¹⁺.

To understand the catalytic function of [4Fe-4S]_AUX, it is critical to understand its redox state during catalytic turnover. To this end, we used 5′-dA-C4′● that accumulates in the active site during wt-MoaA catalysis. 5′-dA-C4′● is formed directly from an on-pathway GTP-C3′● intermediate through H-4′ atom abstraction from 5′-dA in a manner competitive to and kinetically comparable to the addition of GTP C3′● to C8 (Figure 1).¹⁴ Since 5′-dA-C4′● is <15 Å from [4Fe-4S]_AUX if [4Fe-4S]_AUX is in the reduced paramagnetic state, there should be spin–spin interactions between [4Fe-4S]_AUX¹⁺ and 5′-dA-C4′● detectable by EPR. In fact, 5′-dA-C4′● showed a fast relaxation property, indicating the presence of another paramagnetic center nearby.¹⁴ Therefore, 5′-dA-C4′● could be used to probe the reduction state of [4Fe-4S]_AUX.

At X-band, the 5′-dA-C4′● EPR signal significantly overlaps with the [4Fe-4S]¹⁺ signals (Figure 3A, bottom spectrum), preventing an accurate assessment of spin–spin interaction from the EPR line shape. Therefore, we resorted to temperature-dependent EPR measurements at Q-band frequency, allowing for a better spectroscopic separation due to the large g-anisotropy of the 4Fe-4S signals. As expected, at 40 K, Q-band EPR signal of 5′-dA-C4′● (Figure 4A, top trace) closely resembles that of high-temperature 5′-dA-C4′● X-band EPR signal. At temperatures below 40 K, the 5′-dA-C4′● signal noticeably broadened (Figure 4A). Q-band EPR spectra measured below 20 K showed a distinct splitting coinciding with the appearance of signal from a 4Fe-4S cluster (Figure 4A, Figure S5). We were able to simulate the splitting (Figure 4B) using a purely isotropic spin–spin interaction and spin Hamiltonian parameters of 5′-dA-C4′● reported previously¹⁴ and the 4Fe-4S cluster species 2 (Table 2). Using the ${\mathcal{H}}_{\text{ex}}=J_{\text{ex}}{S}_1{S}_2$ formalism for the exchange spin–spin interaction, we have estimated the interaction constant to be J_ex≅263 MHz (0.0088 cm⁻¹). Simulations with a predominantly anisotropic dipolar spin–spin interaction did not result in reasonable fits. Therefore, while we cannot exclude the presence of a dipolar interaction, we conclude that it is a relatively minor contribution to the overall spin–spin interaction between the radical and the 4Fe-4S cluster.

Figure 4.

Characterization of 5′-dA-C4′● by Q-band EPR. (A) Temperature dependence of EPR spectra of 5′-dA-C4′● in wt-MoaA. (B) Experimental (blue trace) and simulated (red trace) EPR spectra of 5′-dA-C4′● observed in wt-MoaA with GTP and SAM at 40 and 10 K.

For the exchange interaction partner of 5′-dA-C4′●, the following considerations eliminate the possibility of [4Fe-4S]_rs¹⁺ and the observations are most consistent with [4Fe-4S]_AUX¹⁺. First of all, the formation of 5′-dA-C4′● requires reductive cleavage of SAM in which [4Fe-4S]_RS is oxidized to the 2+ state. Thus, when 5′-dA-C4′● is formed, [4Fe-4S]_RS is in the oxidized diamagnetic 2+ form. We also considered a case where [4Fe-4S]_RS was rereduced by SDT after accumulation of 5′-dA-C4′●. However, such rereduction of [4Fe-4S]_RS is unprecedented in any radical SAM enzymes and unlikely considering that 5′-dA-C4′● is formed in the time scale of the on-pathway C3′–C8 cyclization. Furthermore, EPR and PFV data described above revealed that [4Fe-4S]_RS can only be partially reduced by SDT without any detectable shift of the reduction potential of [4Fe-4S]_RS by SAM or GTP binding. Thus, if [4Fe-4S]_RS is the interaction partner, we should observe a mixture of exchange coupled and uncoupled signals. Consequently, the homogeneity of the exchange-coupled 5′-dA-C4′● signal is inconsistent with [4Fe-4S]_RS being the exchange interaction partner.

The exchange interaction between 5′-dA-C4′● and [4Fe-4S]_AUX was unexpected because, in the crystal structure, C4′ of SAM is too far removed from [4Fe-4S]_AUX (>13 Å) for a direct exchange interaction that requires an orbital overlap. To understand this observation, we considered two possible mechanisms: a dynamic relocation of 5′-dA in the active site and superexchange interaction through the guanine base. In radical SAM enzymes, 5′-dA● has been shown to migrate 0.5–1 Å from the position in the resting state to the position of H atom abstraction.^1,17 To investigate such motion in MoaA, we modeled the potential locations of 5′-dA during the catalysis. In this analysis, we used a MoaA structural model created by overlaying the crystal structures of MoaA in complex with SAM¹⁸ and MoaA in complex with GTP.¹⁹ While no structure is currently available for MoaA with both GTP and SAM, this model is reasonable for this purpose as the C5′ of SAM is facing toward C3′ of GTP and the distance between C5′ of SAM and C3′ of GTP (5.4 Å, Figure 5A) is close to those found in crystal structures of other radical SAM enzymes (3.8–4.6 Å).^1,20 To model the position of 5′-dA● for H atom abstraction, we fixed the position of 6-NH₂ and N1 of adenine of SAM as they form H-bonds with backbone amide of Met197 and moved the rest of the molecule so that the 5′-methyl group of 5′-dA is closer to H3′ of GTP without causing a clash with the other parts of GTP or protein (Figure 5B). In this putative H-3′ abstraction conformation, the C4′ of 5′-dA is still 11–12 Å away from the [4Fe-4S]_AUX. Since this distance is still too far for the direct exchange interaction, we also considered further relocation of 5′-dA-C4′●. In the crystal structure, Leu279 positions between SAM and [4Fe-4S]_AUX and likely prevents the migration of 5′-dA-C4′● closer than ~10 Å. Therefore, while the migration of 5′-dA-C4′● could potentially shorten the distance to [4Fe-4S]_AUX, it is unlikely that the distance will become short enough for the direct orbital interaction between 5′-dA-C4′● and [4Fe-4S]_AUX¹⁺.

Figure 5.

Putative 5′-dA relocation during MoaA catalysis. (A) Model structure of MoaA active site in complex with SAM and GTP created by an overlay of the structures of MoaA in complex with SAM (PDB ID 1TV8) and complex with GTP (PDB ID 2FB3). (B) Putative position of 5′-dA● for the abstraction of H3′ of GTP. The position of 5′-dA in the MoaA●SAM complex (before the cleavage) is shown in the cyan line.

The active site structure is consistent with a superexchange interaction through the guanine base of GTP. Long-range superexchange interaction may occur when two paramagnetic centers are spaced by a nonmagnetic molecule. In solids, an exchange coupling of ~0.01 cm⁻¹ can occur for centers spaced 6–11 Å apart.²¹ In proteins, superexchange over center-to-center distances of 7–11 Å has been reported through superexchange pathways with conjugated aromatic rings (e.g., FMN²² and His^23,24). In the modeled structure of MoaA in complex with GTP and SAM, the distance between C4′ of SAM and N9 of GTP is 7.5 Å (Figure 5A). This distance is significantly shorter in the putative H atom abstraction position (~5.3 Å, Figure 5B). Further migration of 5′-dA-C4′● toward the guanine base may be possible considering the space available above the ribose ring of GTP. Such superexchange interaction would also explain the observation of mostly isotropic spin–spin interaction because the through-space C4′-Fe distance of 10–12 Å would result in a relatively small anisotropic dipolar spin–spin coupling of about 30 MHz, which is well within our upper limit for anisotropic spin–spin interaction constant of 50 MHz. Therefore, the observed interaction between 5′-dA-C4′● and [4Fe-4S]_AUX¹⁺ can be explained by the dynamic relocation of 5′-dA and the superexchange through the guanine base of GTP. Together, these observations and analyses support that [4Fe-4S]_AUX is in its reduced state during catalysis and electronically coupled to the guanine base of GTP.

DFT Calculation of Aminyl Radical Reduction Potential.

To study the relevance of [4Fe-4S]_AUX to the radical quenching, we investigated the mechanism of the aminyl radical reduction step by DFT calculation and the experimentally determined reduction potential of [4Fe-4S]_AUX. The aminyl radical reduction to 3′,8-cH₂GTP must accompany transfers of an electron and a proton. In the crystal structures, R17 is the only deprotonatable amino acid residue within 4 Å from N7 of GTP. Also, previous DFT calculation of the aminyl radical intermediate in the active site suggested that the distance between a proton on R17 and N7 of the aminyl radical is 2.2 Å,¹⁴ consistent with a proton transfer. Therefore, we calculated the aminyl radical reduction with R17 as a proton donor by three mechanisms; an electron transfer followed by a proton transfer (ET-PT), a proton transfer followed by an electron transfer (PT-ET), and a proton-coupled electron transfer (PCET) (Figure 6, Figure S6).

Figure 6.

Stepwise (ET-PT or PT-ET) or concerted (PCET) mechanism for reductive quenching of 3′,8-cyclo-GTP-N7● aminyl radical, where R17 serves as a proton donor and [4Fe-4S]_AUX as an electron donor. Redox potentials or ΔG were calculated based on DFT calculation.

In our previous study, we performed DFT calculation of the transformation of GTP-C3′● to 3′,8-cyclo-GTP-N7● using a simplified active site model constituted of GTP-C3′● and three methylguanidinium ions as mimics of three catalytically essential arginine residues (R17, R266, and R268).¹⁴ Thus, in this study, we used the product state (3′,8-cyclo-GTP-N7●) of the previous calculation as the initial state and calculated the reduction of 3′,8-cyclo-GTP-N7● into 3′,8-cH₂GTP by the three mechanisms. In the PT-ET mechanism, there is a significant mismatch in pK_a of Arginine (13.8²⁵) and N7 of GTP (2.4–3.5), resulting in a significantly uphill PT with ΔG of 13.8 kcal/mol. The E_a for this PT was calculated to be 24.4 kcal/mol. Our previous kinetic characterization of MoaA revealed the rate constant for the transformation of GTP-C3′● into 3′,8-cH₂GTP as 2.7 ± 0.7 s⁻¹ and the activation energy as 12–16 kcal/mol.¹⁴ Therefore, the calculated activation energy of the PT-ET mechanism is inconsistent with the experimentally determined rate of MoaA catalysis. In the ET-PT stepwise mechanism, we observed a reduction potential of −1040 mV, which is significantly more negative than the experimentally determined reduction potential of [4Fe-4S]_AUX (−455 mV). The PCET mechanism was predicted to proceed with the reduction potential of −430 mV, which is slightly more positive than the reduction potential of [4Fe-4S]_AUX (−455 mV). These results suggest that the PCET mechanism is the kinetically favored mechanism of the aminyl radical reduction.

DISCUSSION

Elucidation of the catalytic function of auxiliary clusters in radical SAM enzymes in the SPASM-twitch family remains difficult. While reduction potentials of several other members of the SPASM-twitch family have been reported,^11,12,26,27 ambiguity remains about their functions in catalysis due to the lack of understanding in the reduction potentials of reaction intermediates and the redox state of the auxiliary clusters during catalytic turnover. In this study, we used a combination of PFV, EPR, and DFT calculation to provide experimental evidence that the reduced 1+ state auxiliary cluster of MoaA reduces the aminyl radical intermediate to form the final product.

Characterization of MoaA by PFV and EPR revealed the reduction potentials of −455 and −510 mV vs SHE for the [4Fe-4S]_AUX and [4Fe-4S]_RS, respectively. The observed potentials were not affected by SAM and only minimally affected by GTP binding. In the presence of SDT, [4Fe-4S]_AUX was mostly reduced, whereas the [4Fe-4S]_RS cluster was only partially reduced. The more negative reduction potential of [4Fe-4S]_RS relative to [4Fe-4S]_AUX ensures that [4Fe-4S]_AUX is reduced when [4Fe-4S]_RS is reduced. Reduction of [4Fe-4S]_RS without reducing [4Fe-4S]_AUX would cause an abortive cleavage of SAM or unwanted reaction outcome, such as oxidative quenching of the aminyl radical. Thus, the observed reduction potentials are likely a mechanism for MoaA to trigger the SAM cleavage only in the presence of the catalytically relevant redox state (1+) of [4Fe-4S]_RS. This is a sharp contrast to some of the characterized SPAMS subfamily members that use [4Fe-4S]_AUX as the electron acceptor. For example, in MftC and SCIFF maturese,^11,12 the reduction potentials of [4Fe-4S]_RS (−460 or −490 mV) is more positive than those for [4Fe-4S]_AUXI(−550 or −540 mV) and [4Fe-4S]_AUXII (−500 or −585 mV), allowing the reduction of [4Fe-4S]_RS without reducing [4Fe-4S]_AUX so that it could serve as an electron acceptor during the radical quenching step. A more recent example from SuiB¹³ suggested that the binding of its peptide substrate, SuiA, apparently shifts the reduction potentials of [4Fe-4S]_RS and/or [4Fe-4S]_AUXI, allowing a selective reduction of [4Fe-4S]_RS. Therefore, these observations suggest that the relative redox potentials of 4Fe-4S clusters are finely tuned based on the redox functions of [4Fe-4S]_AUX during the radical quenching step.

Our Q-band EPR characterization of 5′-dA-C4′● revealed its exchange interaction with [4Fe-4S]_AUX¹⁺ with the coupling constant of J_ex ≅ 263 MHz (0.0088 cm⁻¹). This exchange coupling constant was unexpectedly large considering the long-distance (>13 Å) between the C4′ of SAM and [4Fe-4S]_AUX in the crystal structure. The MoaA active site structure analysis suggested that the observed exchange interaction is likely mediated by the guanine base of GTP, suggesting the electronic coupling of guanine and [4Fe-4S]_AUX¹⁺. This electronic interaction between GTP and [4Fe-4S]_AUX is also consistent with the small (+10 ~ + 25 mV) but detectable positive shift of the reduction potential of [4Fe-4S]_AUX upon GTP binding. In synthetic Fe(II)²⁸ and Ni(II)²⁹ polypyridine complexes, electronic coupling of the redox noninnocent polypyridine ligand to Fe(II) or Ni(II) has been reported to cause positive shifts of the reduction potentials of the polypyridine ligand. Thus, although the extent of the electronic coupling between GTP and [4Fe-4S]_AUX is currently unclear, it is possible that such electronic coupling positively shifts the reduction potential of the aminyl radical and serves as an additional mechanism to facilitate the aminyl radical reduction.

The study also suggested the function of R17 as a proton donor for the aminyl radical reduction. Our earlier study had shown that the cationic charge of R17 is essential to catalyze the GTP-C3′● addition to C8 by stabilizing the transition state.¹⁴ Therefore, while amino acid residues with lower pK_a would be preferable as a proton donor for the aminyl radical reduction, R17 cannot be replaced by an acidic residue. Instead, the presence of the strong reductant, [4Fe-4S]_AUX and the coupling of ET and PT allow the use of R17 as a proton donor. The dual role of R17 in the MoaA catalysis explains why this residue is strictly conserved among MoaA in all kingdoms of life and why a mutation in this residue causes the Moco deficiency disease in humans.³⁰

CONCLUSIONS

Although radical SAM enzymes have been emerging rapidly, radical quenching mechanisms remain largely unexplored. Our comprehensive characterization of the auxiliary cluster of a Twitch-family member, MoaA, revealed the reduction potential and the redox state of [4Fe-4S]_AUX during the catalytic cycle and strongly suggested its catalytic role as an electron donor for the quenching of product radical. Our Q-band EPR results also provided the first evidence for an electronic coupling between GTP and [4Fe-4S]_AUX, which could be used to alter the reduction potential of the aminyl radical intermediate. These results provide important insights into the mechanism of post-H atom abstraction steps in radical SAM enzymes in general.

MATERIALS AND METHODS

General.

Sodium dithionite (SDT) was purchased from Sigma-Aldrich. β-Mercaptoethanol (βME) was from Calbiochem. Dithiothreitol (DTT) was from Amresco. Guanosine 5′-triphosphate (GTP) was from Chem-Impex. G-25 Sephadex resin was from GE Healthcare. Ni-NTA agarose resin was from Qiagen. SAM was enzymatically synthesized from l-methionine and ATP using the same protocol as described before.¹⁴ Escherichia coli DH5α and BL21(DE3) competent cells were from Invitrogen. N-terminally His₆-tagged Staphylococcus aureus wt-MoaA or a C24S/C28S/C31S MoaA triple variant (ΔRS-MoaA) in pET15b was expressed, purified, and characterized using the same protocol as described before.¹⁴ Evaluation of statistical significance was carried out using GraphPad Prism 7. All anaerobic experiments were carried out in an MBRAUN glovebox maintained at 10 ± 2 °C with an O₂ concentration <0.1 ppm. All anaerobic buffers were degassed on a Schlenk line and equilibrated in the glovebox overnight. All plastic devices were evacuated in the antechamber of the glovebox overnight before use.

Determination of Reduction Potentials of MoaA.

Electrochemical experiments were performed with a PGSTAT 12 potentiostat (EcoChemie) under anaerobic conditions in an MBRAUN Labmaster glovebox. A three-electrode configuration was used where the reference electrode was a standard calomel electrode, and the counter electrode was a platinum wire. The temperature of the cell was controlled by a water jacket. All potentials are reported relative to a standard hydrogen electrode (SHE). Baseline scans were measured using a pyrolytic graphite edge (PGE) electrode that had either been treated with alumina polishing or pretreated by polishing with sandpaper followed by incubating overnight with 20 μL of 3 mg of multiwalled carbon nanotubes (MWCNT) (Sigma, 10 nm ±1 nm × 4.5 nm ±0.5 nm × 3–6 μm) dissolved in 1 mL of dimethylformamide (DMF) by sonication for 15 min.³¹ Whether MWCNT-modified or not, electrodes were then placed in the electrochemical cell containing a mixed buffer solution (10 mM MES, CHES, HEPES, CAPS, and TAPS), 200 mM NaCl, and pH 7.5 that was kept at a temperature of 4 °C. A 4 μL aliquot of wt- or the ΔRS-MoaA variant (>500 μM) was deposited directly on the electrode and allowed to sit for ~4 min at room temperature, after which the electrode was immersed in the electrochemical cell, and cyclic or square wave voltammetry was measured. Nonturnover electrochemical signals were analyzed by correcting the non-Faradaic component of the current from the raw data using the QSoaS package.³²

Voltammograms were analyzed to assess how many electrons were transferred based on peak current (I_p) and full-width at half-height (fwhm) based on the eqs 1 and 2

$$I_{\text{p}}=\frac{{\text{n}}^2{\text{F}}^2\text{A}\Gamma \text{v}}{4\text{RT}}$$ (1)

$$\text{FWHM}=3.53\frac{\text{RT}}{\text{nF}}$$ (2)

where n is the number of electrons transferred, F is the Faraday constant, Γ is the electroactive surface coverage, A is the electrode surface area, ν is the scan rate, R is the universal gas constant, and T is temperature. Envelope signals were fit to either one or two redox-active species using the Nernst equation with the scan rate (ν) and temperature (T) set and fitting for the peak potential (E_p), the number of electrons involved in the transfer (n), and the electroactive coverage (γ) for each species. The electroactive coverage was set equal to one another for both species. The best-fitting parameters were determined using orthogonal distance regression (ODRPACK) within the QSoaS program.

To assess the dependence of reduction potential on pH, a Pourbaix diagram was performed on the wt-MoaA. Reduction potentials were determined by square wave voltammetry measured with a frequency of 15 Hz and amplitude of 50 mV. The pH of the buffer was adjusted by the addition of NaOH or HCl to cover a range of pH 5.5–9.5. The protein film was allowed to equilibrate at the new pH for ~5 min before performing electrochemical experiments.

To study the effect of SAM or GTP binding on the midpoint potential of the clusters, SAM (1 mM) or GTP (1 mM) or both were briefly incubated with wt-MoaA or ΔRS-MoaA before the quick deposition of the solution directly on the PGE electrode modified with MWCNT. The scans were initiated after 1 min of exposure. Potentials determined for the enzymes in the presence of SAM or GTP were determined from square wave voltammograms measured with a frequency of 10 or 15 Hz.

X-Band EPR Experiments.

MoaA (wt- and ΔRS-MoaA) samples were prepared under strictly anaerobic conditions (<0.1 ppm of O₂). MoaA (150 μM) was first prereduced by 3 mM sodium dithionite (SDT) in 100 mM Tris·HCl pH 7.6, 0.3 M NaCl, 10% glycerol with 5 mM DTT at 25 °C for 60 min. To initiate the reaction, the prereduced MoaA was mixed with SAM or GTP to the final concentrations of 100 μM MoaA, 2 mM SDT, 1 mM SAM and 1 mM GTP. The samples were transferred to EPR tubes and freeze-quenched manually by submerging in an isopentane slush bath (~110 K) in the glovebox. The samples were then taken out of the glovebox and kept in liquid nitrogen for X-band continuous wavelength EPR characterization.

The EPR spectra for the 4Fe-4S clusters were measured at 15 K in an EMXplus 9.5/2.7 EPR spectrometer equipped with an In-Cavity Cryo-Free VT system (Bruker Biospin Corporation). EPR parameters were a microwave frequency of 9.36–9.39 GHz, a power of 1.589 mW, modulation amplitude of 10 G, modulation frequencies of 100 kHz, time constants of 0.01 ms, and a scan time of 100 s. Quantitation of the [4Fe-4S]¹⁺ signals was carried out using a 1 mM Cu(II) sample as standard. The Cu(II) standard was prepared from a ~100 mM CuSO₄ stock solution in H₂O with the Cu(II) concentration determined by light absorbance at 810 nm (ε_{810 nm} = 11.8 M⁻¹ · cm⁻¹). The stock solution was diluted in 2 M NaClO₄, 10 mM HCl, 20% glycerol to prepare 1 mM Cu(II) standard. EPR spectra of Cu(II) standard were determined under the following conditions: a microwave frequency of 9.36–9.39 GHz, a power of 0.02 mW, modulation amplitude of 4 G, modulation frequencies of 60 kHz, time constants of 0.04 ms, and a scan time of 30 s.

The baseline correction was performed in Bruker Xenon software. The signal intensity was determined by calculating the double integral of the first derivative EPR spectra. The normalized intensity was obtained from the double integrals (DI) by the eq 3

$$\text{normalized intensity }(\text{DI}/\text{Nc})=\frac{\text{observed intensity(DI) }}{\text{M}\cdot \text{g}\cdot \text{V}\cdot \sqrt{\text{P}}}$$ (3)

where M represents modulation amplitude, P is the microwave power, g is the average g value of the radical, and V is a volume factor to correct for the difference in the inner diameter of EPR tube. The normalized intensity was compared to that of the standard to get the exact spin concentration of the sample. All EPR spectral simulations were performed using the EasySpin software.³³

Anaerobic ITC.

Using a GE MicroCal VP-ITC instrument under the flow of argon, anaerobic ITC was performed on wt-MoaA and the ΔRS-MoaA variant. A solution containing GTP or SAM (200 μM in 0.5 mL) or assay buffer (25 mM Tris·HCl pH 7.6, 150 mM NaCl, 5% glycerol with 2.5 mM DTT) was titrated into 1.5 mL of 20 μM MoaA in the same buffer over 29 injections at 25 °C, 307 rpm. The resulting ITC data were analyzed by Origin 7.

Q-Band EPR Experiments.

MoaA samples were prepared under strictly anaerobic conditions (<0.1 ppm of O₂). MoaA (450 μM) was first prereduced by 1.5 mM sodium dithionite (SDT) in 50 mM HEPES pH 7.6, 150 mM NaCl, 10% glycerol with 5 mM DTT at 25 °C for 60 min. To initiate the reaction, the prereduced MoaA was then mixed with SAM or GTP in the same buffer. The final concentration of each component was 300 μM MoaA, 1 mM SDT, 1 mM SAM and 1 mM GTP. The samples were transferred into Q-band EPR tubes and freeze-quenched manually by submerging in liquid nitrogen immediately after taking out of the glovebox.

Q-band EPR experiments were carried out with Bruker E500 spectrometer equipped with Bruker ER 5106 QT-W resonator and Oxford cryogenics CF935 liquid He flow cryostat. All spectra were acquired with the following settings: microwave Frequency, 33.98 GHz; modulation amplitude, 5 G; modulation frequency, 100 kHz; conversion time, 40 ms; MW power 1.2 mW (≤30 K) or 0.38 mW (>30 K). Data processing was performed in Matlab using Kazan Viewer software. All Q-band EPR spectral simulations were performed using the “pepper” routine from EasySpin software.³³

DFT Calculation.

DFT calculation was performed using the previously reported simplified model of the MoaA active site.¹⁴ This model includes 3′,8-cyclo-GTP-N7● or 3′,8-cyclo-GTP–N7-H^+• with methyl-guanidinium ions as mimics of three active site Arg residues (R17, R266, and R268). The heavy atoms in the truncated Arg residues, the backbone atoms of the triphosphate group (Pγ, O3β, Pβ, O3α, Pα), and the three atoms of the purine group (O6, N1, N2) were frozen.¹⁴ All calculations were performed using the Gaussian 16 program³⁴ by applying hybrid functional Becke,³⁵ 3-parameter, Lee–Yang-Parr³⁶ with the D3 version of Grimme dispersion with Becke-Johnson damping³⁷ (B3LYP-GD3BJ) and the 6–311G(d) basis set.³⁸ The vibrational frequencies for all structures were calculated at 298.15 K and 1 atm pressure. On the basis of the optimized structures in the gas phase, the solvation free energies were calculated by the SMD solvent model with M06–2X functional and 6–31G(d) basis set. The Gibbs free energies in a solvent were obtained by adding the gasphase Gibbs free energies and solvation free energies.