Detection of Water Molecules on the Radical Transfer Pathway of Ribonucleotide Reductase by ¹⁷O Electron–Nuclear Double Resonance Spectroscopy

Abstract

The role of water in biological proton-coupled electron transfer (PCET) is emerging as a key for understanding mechanistic details at atomic resolution. Here we demonstrate ¹⁷O high-frequency electron–nuclear double resonance (ENDOR) in conjunction with H₂¹⁷O-labeled protein buffer to establish the presence of ordered water molecules at three radical intermediates in an active enzyme complex, the α₂β₂E. coli ribonucleotide reductase. Our data give unambiguous evidence that all three, individually trapped, intermediates are hyperfine coupled to one water molecule with Tyr-O···¹⁷O distances in the range 2.8–3.1 Å. The availability of this structural information will allow for quantitative models of PCET in this prototype enzyme. The results also provide a spectroscopic signature for water H-bonded to a tyrosyl radical.

Water is no longer known as just the solvent in which biochemical reactions take place but has been recognized as an essential player in these reactions.¹ Of particular interest is water involvement in electron transfer processes,²⁻⁵ its action as a proton wire⁶⁻⁸ or its role in proton-coupled electron transfer (PCET).⁹⁻¹² The identification of internal water in proteins can be achieved by X-ray diffraction.¹³⁻¹⁵ However, the crystallization of transient protein complexes is difficult. One key approach for detection of water in biological systems has been the use of ¹⁷O-enriched water in conjunction with magnetic resonance spectroscopy.¹⁶⁻²¹ Among these methods, electron paramagnetic resonance (EPR) can take advantage of high selectivity, as it detects nuclei only in the ligand sphere (r ≲ 1.5 nm)²² of paramagnetic centers.

EPR-based ¹⁷O hyperfine (hf) spectroscopy has been established for the detection of water binding to transition-metal ions, where the oxygen usually coordinates to the ion and large hyperfine couplings (several MHz) can be observed.²³⁻²⁵ However, the most common water coordination motif to biological radicals occurs via H-bond interactions. The hf coupling to ¹⁷O is diminished in comparison to the metal ion coordination, due to a longer interspin distance. In addition, the small ¹⁷O gyromagnetic ratio (γ_H/γ_¹⁷O ≈ 7.4)²⁶ and high nuclear spin (I = 5/2) have rendered the ¹⁷O hf coupling difficult to resolve.

Here we illustrate that high-frequency (94 and 263 GHz) electron–nuclear double resonance (ENDOR) spectroscopy can detect the ¹⁷O signal of ordered water molecules at an H-bond distance to radical intermediates in E. coli ribonucleotide reductase (RNR). The enzyme uses a long-range (32 Å) radical transfer (RT) to initiate nucleotide reduction (Scheme 1).²⁷ Three tyrosines (Y₃₅₆, Y₇₃₁, and Y₇₃₀) are essential pathway residues, which form transient intermediates in the active complex α₂β₂, consisting of the two homodimeric subunits α₂ and β₂.^13,27 Water has been observed only crystallographically in inactive α₂s without β₂.^13,14,28,29 Using site-selectively inserted tyrosine analogues to trap Y intermediates,³⁰ our previous ¹H/²H ENDOR and DFT studies^10,11,31 revealed H-bonds attributed to water molecules and proposed a role of water in RT. However, all active sites of proteins have exchangeable protons, and thus alternative interpretations to our water proposal were possible. Recently, a cryo-EM structure of α₂β₂ was reported but the resolution was insufficient for water observation.³² Since our original proposals, studies using photo-RNRs and MD simulations implying waters in α₂β₂ have appeared.^33,34 However, water has never been directly detected.

Scheme 1. Current PCET Model of the 32 Å (Y₁₂₂ to C₄₃₉) RT in E. coli RNR^27,32,

Redox-active tyrosines 356, 731, and 730 are shown in cyan, electron transfer steps as red arrows, and proton transfer steps as blue arrows. Water molecules revealed in this study in respective site-selective mutants are shown in boldface.

Therefore, we explored the capability of H₂¹⁷O ENDOR spectroscopy by exchanging the RNR buffer with H₂¹⁷O. α₂β₂-Y₃₅₆^• was generated by a 2,3,5-F₃Y₁₂₂^• mutation in β₂,³⁵ whereas radicals at Y₇₃₁ and Y₇₃₀ were trapped by replacing the respective residue with 3-aminotyrosine (NH₂Y),³⁶ leading to α₂β₂-NH₂Y₇₃₁^• and α₂β₂-NH₂Y₇₃₀^•. The individual variants were mixed with the complementary α₂ or β₂ protein, CDP as a substrate, and ATP as an effector. The reaction was then quenched after a few seconds inside EPR tubes. Details on the sample preparation are given in sections SI1 and SI2.

Figure 1 displays representative 94 GHz ¹⁷O Mims³⁷ ENDOR spectra of the radical intermediates.

Figure 1.

94 GHz ¹⁷O Mims³⁷ ENDOR spectra of (A) the intermediate Y₃₅₆^• and (B) NH₂Y₇₃₁^• and NH₂Y₇₃₀^• at B₀∥g_y in the EPR line (T = 50 K, τ_Mims = 390 ns). Acquisition time: 46 h (Y₃₅₆^•), 40 h (NH₂Y₇₃₁^•), and 18 h (NH₂Y₇₃₀^•). Y₃₅₆^• is from β₂-F₃Y₁₂₂^•/α₂-Y₇₃₀F, which gives the highest radical yield (section SI2). Experimental spectra are shown in gray, with a Savitzky–Golay filter (fourth-order polynomial, 20-point window) shown in color. Simulations used Easyspin³⁸ (section SI1.6) with parameters given in Table 1 and section S3. Solid lines (teal) represent transitions among m_I > 0 manifolds and dashed lines (orange) those among m_I < 0 manifolds. The simulation does not distinguish between dihedral θ = 0°or θ = 180°.

Each spectrum shows a sharp doublet centered on the ¹⁷O Larmor frequency (19.3 MHz at 3.4 T), which can be assigned to the central spin transition (m_I(¹⁷O) = +1/2 → −1/2) of one coupled ¹⁷O nucleus. As ¹⁷O is contained only in the water of the protein buffer, these sharp signals must arise from water molecules coupled to the radicals. Control experiments with only β₂ protein confirmed that the signal is associated to the radicals generated in α₂β₂ (section SI3). The broad resonances at ±2.5 MHz are attributed to other nuclear transitions of the I = 5/2 spin system, broadened by nuclear quadrupole coupling (Figure 1A). Additionally, we note asymmetry of the doublet, which arises from second-order effects of the quadrupole coupling (section SI4). A comparison of the ENDOR spectra at the low (B₀∥g_x) and high-field (B₀∥g_z) edges of the EPR line (section SI5) indicates an almost isotropic hf coupling, with the dipolar contribution dominating the line width of the central doublet. The ¹⁷O ENDOR spectra could be simulated with one ¹⁷O nucleus, from which the asymmetry of the central peaks resulted using full diagonalization of the spin Hamiltonian (Figure 1 and section SI8). Parameters are given in Table 1 and section SI3. The spectra of Y₃₅₆^• and NH₂Y₇₃₁^• additionally contain signals close to the Larmor frequency not reproduced in the simulations, which likely originate from second-sphere water molecules at the subunit interface. Additional broadening is also observed, particularly at NH₂Y₇₃₁^•. It might be caused by conformational distribution of this residue, which was found to have flexibility.^39,40,33

Table 1. Simulation and DFT Parameters for ¹⁷O and ¹H hf Couplings of Water in RNR Intermediates^a.

	Y356• sim/DFTsmall	NH2Y731• sim	NH2Y730• sim/DFTlargeb
Ax (17O)	0.43/0.19	0.70	0.65/0.24
Ay (17O)	0.66/0.59	0.84	0.80/0.6
Az (17O)	0.70/0.65	0.89	0.89/0.6
A(H1)	6.231/7.4	≲2.5b	2.7b/4.2
ρ(17O)c (%)	0.05		0.03
rO···17O (Å)	2.9 ± 0.1	∼3.0	∼3.0

^aExcept as noted, values are in MHz. Simulated quadrupole values for ¹⁷O were {P_x;P_y;P_z} = {−0.02;–0.32;0.34} MHz with e²qQ/h = 6.8 MHz and η = 0.93.⁴¹

^bValues from ²H couplings in refs (11 and 10) using γ_¹H/γ_²H ≈ 6.5.²⁶

^cLoewdin spin density⁴² from DFT. Uncertainties in coupling constants are less than 10% for simulations and up to 20% for DFT.

To rationalize the coupling, we began with a DFT-optimized small model (25 atoms, details in section SI1) of Y₃₅₆^•, as previous ENDOR spectra revealed ¹H couplings consistent with one water at the H-bond distance r_O–H ≈ 1.8 Å.³¹ The ¹⁷O coupling from this model was A_max(¹⁷O) ≈ 1 MHz, slightly exceeding the present experimental value of 0.6 ± 0.05 MHz. To optimize the model, we computed dihedral θ (C₃–C₄–O···H) and distance scans for ¹⁷O couplings, including the quadrupole tensor and the relative energies (section SI6). The DFT equilibrium distance always resulted in r_O–H ≈ 1.8 Å. We found that hf couplings and energies vary significantly with θ, while the quadrupole coupling is less affected (Figure S9A–C). A_xyz values of ≲1 MHz are found for θ in the range ≲±30° (or equivalently 150° ≲ θ ≲ 240°): i.e., close to the ring plane. Water coordination in the ring plane also results in minimal relative energies (Figure S9B). Importantly, predicted spin densities on ¹⁷O are <0.1% but are sufficient for producing a marked ¹⁷O isotropic splitting. The spin density transfer or spin polarization is likely related to the H-bond nature. A distance scan for the optimized dihedral of +20° predicts A_max(¹⁷O) in the range 0.75–0.56 MHz (Figure S10A) for r_O–H ≈ 1.8–2.0 Å. Consideration of the DFT-predicted ¹H couplings (Figure S10B) and comparison with the experimental values³¹ of ∼6.2 (H₁) and ∼1.6 MHz (H₂) indicates that the water is located at r_{Tyr-O···¹⁷O} = 2.9 ± 0.1 Å, corresponding to an r_O–H value of 1.9 ± 0.1 Å. Notably, the DFT-predicted dipolar coupling (T_|| ≈ 0.3 MHz, section SI6) is consistent with the point-dipole model and the aforementioned broadening of the sharp peaks.

Analogous DFT calculations were performed on the isolated amino tyrosyl NH₂Y^•.^10,11,36 We observed a trend for the ¹⁷O hf coupling in the dihedral and distance scans (section SI7) very similar to the Y^• model. The calculation predicts that A_iso(¹⁷O) of NH₂Y^• is slightly larger (10–15%) than that of Y^• at similar Tyr-O···¹⁷O distances and orientations, which could explain the experimental observation. The amino group introduces an asymmetry in the radical, and the energetically most favored water orientation is found at the opposite side of the amino group (Figure S11B). Nevertheless, this small model could not account simultaneously for the ¹⁷O and ¹H couplings observed for these two intermediates (Figure S12). As noted in a previous g_x calculation,¹⁰ the coordination of the water molecule to NH₂Y^•s is influenced by the surrounding second-sphere residues, as these two intermediates are buried in α₂β₂ (Scheme 1).

Having established that at least one water molecule is hf-coupled to each of the three intermediates, we examine their current molecular models in light of this finding. First, we consider the radical site Y₃₅₆^• (Scheme 1). To explain the unprecedented g_x value of Y₃₅₆^• (g_x = 2.0062), we previously proposed that two almost equivalent waters might be simultaneously bonded to Y₃₅₆^•.³¹ While the present results are most consistent with the distance and orientation proposed for one water, the 94 GHz ¹⁷O ENDOR spectra (Figure 1A) cannot resolve a second water. We note that the spectral line shape and ¹⁷O hf coupling in Figure 1A are conserved in other RNR constructs that generate Y₃₅₆^• (section SI8), including the F₃Y₁₂₂^•/E₅₂Q-β₂ double mutant used to solve a recent cryo-EM structure.⁴³

To gain spectral resolution, we recorded ¹⁷O ENDOR spectra of Y₃₅₆^• at 263 GHz/9.4 T (Figure 2).⁴⁴⁻⁴⁶

Figure 2.

Comparison of 94 and 263 GHz Mims ENDOR of Y₃₅₆^• at the three canonical positions in the EPR line. Total acquisition time for 263 GHz (T = 20 K): 18 h (B₀∥g_x), 10 h (B₀∥g_y), and 11 h (B₀∥g_z). Experimental spectra are shown in gray, with a Savitzky–Golay filter (fourth-order polynomial, 10 point window) in color. Simulations of 263 GHz spectra are in black with parameters as for 94 GHz (see Table 1 and Table S4).

The results illustrate that the line width of the central doublet substantially narrows, particularly at B₀∥g_z (Figure 2). Despite the narrowing, a factor of approximately 2 from 94 to 263 GHz, we cannot discern two distinct ¹⁷O contributions. Simulations of the 263 GHz spectra with the same parameters used at 94 GHz reproduce the line narrowing and support the analysis at 94 GHz. The lack of evidence for a second, almost equivalent water H-bonded to Y₃₅₆^• strongly suggests that the two-water model has become very unlikely and alternative explanations for the shifted g_x value of Y₃₅₆^• will have to be examined. The precise location of second-sphere residues might play a role,¹² which will require further experimental and computational investigation.

For the radical intermediates in the subunit α, a previous combined ENDOR/DFT model of NH₂Y₇₃₀^• proposed a water molecule coordinated in plane at a distance r_{NH2Y730-O···¹⁷O} ≈ 3.0 Å.¹⁰ The present results are consistent with this model and provide direct evidence for this postulated water in the enzyme complex α₂β₂-NH₂Y₇₃₀^•. The DFT-predicted hf parameters (DFT_large) for this large model (140 atoms) are reported in Table 1, and the model is displayed in section SI9.

Finally, for α₂β₂-NH₂Y₇₃₁^•, large-scale (215 atoms) DFT calculations previously proposed three models of the trapped intermediate (section SI10). Among these models, only one (model 3, Figure S15) contained a water molecule at an H-bond distance. The DFT-predicted ¹⁷O hf couplings of model 3 (∼2.5 MHz), however, largely exceed the present experimental values (Table S5). However, this DFT model did not include residues from the β subunit, which we now know are close to this residue in the active complex.⁴³ Therefore, the model will require further refinement. Nevertheless, the present results give evidence for a water molecule coordinated almost in the plane of NH₂Y₇₃₁^•.

In conclusion, we have reported the capability of ¹⁷O high-frequency ENDOR to detect water H-bonded to tyrosyl radicals. The spectroscopic approach led to the first detection of ordered water molecules at three trapped radicals proposed to be representative of Y^• intermediates in the PCET of E. coli RNR. These results verify previous hypotheses on the presence and role of water in the RNR mechanism and provide a new starting point for computational studies. Knowledge of this ¹⁷O signature will also be generally useful for many other biological systems, in which tyrosyl radicals are involved.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.1c01359.

• Experimental procedure, radical yield determination, ENDOR spectra of Y₁₂₂^• and F₃Y₁₂₂^•, ENDOR spectra of I = 5/2 nuclei, orientation-selective ENDOR spectra, DFT models of Y^• and NH₂Y^•, ¹⁷O Y₃₅₆^• spectra of different mutants, previous large DFT models of α₂β₂-NH₂Y₇₃₀^• and α₂β₂-NH₂Y₇₃₁^• (PDF)

This work was funded by the Max Planck Society and the DFG priority program SPP1601 to M.B. as well as an NIH grant 29595 to J.S.

The authors declare no competing financial interest.