The role of Tyr34 in proton coupled electron transfer and product inhibition of manganese superoxide dismutase

About 25% of known enzymes perform redox reactions to promote cellular functions¹. These enzymes, known as oxidoreductases, use proton-coupled electron transfers (PCETs)^1,2,3. PCETs are required for a wide range of biological functions, including energy generation and DNA synthesis, and oxidoreductase dysfunction leads to many disease states^4,5,6,7,8. While PCETs are a fundamental biochemical reaction, how they are facilitated by enzymes is not well understood due to difficulties in unraveling the precise proton transfer and electron transfer steps^2,9,10. Some enzymes use multiple proton transfers per electron transfer step or vice-versa^11,12,13. Defining how PCET reactions occur in oxidoreductases benefits the understanding of redox-mediated disease states and furthers therapeutic interventions^{14,15,16,17,18}.

Some oxidoreductases use their PCETs to regulate the concentrations of reactive oxygen species (ROS) in cells. These PCETs are crucial as fluctuations of ROS levels stimulate mitophagy and programmed cell death, and excessive levels lead to the damage of DNA, proteins, and lipids¹⁹. Impairment of oxidoreductase function promotes several diseases, including cardiovascular and neurological conditions and cancer progression^20,21. In particular, the oxidoreductase human manganese superoxide dismutase (MnSOD) is responsible for eliminating superoxide (\({{{{\rm{O}}}}}_{2}^{\bullet -}\)) in the mitochondrial matrix. MnSOD dysfunction contributes to a broad range of diseases including psoriasis, inflammatory bowel disease, multiple sclerosis, cardiovascular disease²², and breast and prostate cancer^23,24,25,26. Overall, the PCETs of MnSOD contribute to health and longevity^27,28. \({{{{\rm{O}}}}}_{2}^{\bullet -}\) is endogenously produced as a byproduct of the mitochondrial electron transport chain, and MnSOD is the only means to reduce \({{{{\rm{O}}}}}_{2}^{\bullet -}\) in the mitochondrial matrix^29,30.

Within the mitochondrial matrix, MnSOD uses PCETs to decrease \({{{{\rm{O}}}}}_{2}^{\bullet -}\) concentrations rapidly and efficiently (k_cat/K_m > ~ 10⁹M^-1 s^-1). In the first half-reaction, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) is oxidized to molecular oxygen (O₂) with a Mn³⁺ ion (k₁), and in the second half-reaction, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) is reduced to hydrogen peroxide (H₂O₂) with a Mn²⁺ ion (k₂). The two half-reactions regenerate the Mn³⁺ ion²⁸.

For PCET catalysis, MnSOD uses a hydrogen bond network (dashed blues lines, Fig. 1a right panel) to couple proton transfers with changes in Mn oxidation state^28,31,32. The Mn ion is covalently bound to His26, His74, His163, Asp159, and a solvent molecule called WAT1. From WAT1, the hydrogen bond network extends to Asp159, Gln143, Tyr34, WAT2, His30, and Tyr166 from the adjacent subunit. To enter the active site, the substrate and solvent pass through a gateway formed by Tyr34 and His30. The distance between Tyr34(O^H) and His30(N^δ1) is 5.4 Å. Hydrophobic residues Trp123, Trp161, and Phe66 hold the hydrogen bonding atoms of Asp159, WAT1, Gln143, and Tyr34 close together³³. In our previous work, we investigated the protonation states of the hydrogen bond network in relation to the oxidation state of the Mn ion¹³.

a The active sites of tetrameric MnSOD are found in positively charged cavities made by two subunits. Solvent and substrate must pass through gateway residues His30 and Tyr34 to interact with the catalytic Mn ion. The distance between Tyr34(O^H) and His30(N^δ1) is 5.4 Å. Blue dashed lines indicate hydrogen bonds. The inset indicates the chain identity derived crystallographically, where the asymmetric unit is composed of an AB dimer, and the CD dimer is generated through symmetry to form the native tetrameric assembly. b Room temperature neutron structures revealed changes in protonation among WAT1 and Gln143 and alterations in the hydrogen bond network of WAT1-Gln143-Trp123. Dotted lines indicate hydrogen bonds ≥ 1.8 Å while hashed lines indicate SSHBs that are hydrogen bonds < 1.8 Å. c A shared proton, indicated with rounded dots, was observed between His30 and Tyr166. This proton helps modulate the protonation state of N^δ1(His30) when the Mn ion changes oxidation states. d Tyr34 is deprotonated in Mn³⁺SOD and protonated in Mn²⁺SOD. There is no water within hydrogen bond distance on Tyr34 in Mn²⁺SOD. e Superposition of wildtype MnSOD (yellow, PDB ID 5VF9) and Tyr34Phe MnSOD (red, PDB ID 9BWR) active sites with a root-mean-square deviation of 0.07 Å among C^α atoms. Panel a was created from MnSOD X-ray structure (PDB ID 5VF9)⁷⁰, and panels (b–d) were created from the neutron structures of Mn³⁺SOD (PDB ID 7KKS) and Mn²⁺SOD (PDB ID 7KKW)¹³. All hydrogen positions were experimentally determined except for solvent molecules in panel a that were randomly generated to accentuate the solvent in the active site funnel. All distances are in Å.

Full size image

Using neutron crystallography, we previously observed how changing the Mn oxidation state shifts the pK_a of active site amino acids and leads to several unusual protonation states¹³. Neutron diffraction is advantageous for studying PCET mechanisms because the scattering of deuterium is on par with carbon, nitrogen, and oxygen, and neutrons do not photoreduce metal ions, unlike X-rays^34,35. The neutron structures of Mn³⁺SOD and Mn²⁺SOD revealed three important observations on how PCET catalysis is facilitated. First, during the Mn³⁺ to Mn²⁺ redox transition, an unconventional proton transfer occurs between Gln143 and metal-bound ⁻OH(WAT1), leading to an unusual Gln143 amide anion and H₂O (Fig. 1b). The amide anion is stabilized by two short-strong hydrogen bonds (SSHBs) with WAT1 and Trp123. SSHBs stabilize catalytic steps and enhance kinetic rates (hashed lines, Fig. 1b)^36,37,38. Second, the N^δ1 of His30 is protonated only in the Mn²⁺ oxidation state (Fig. 1c) and a low-barrier hydrogen bond (LBHB) is formed between Tyr166 and His30 in the Mn²⁺ state. A LBHB is a type of SSHB where the heteroatoms transiently share a proton, and the hydrogen bond distances are nearly equivalent (1.2–1.3 Å)³⁹. Third, Tyr34 is deprotonated in the Mn³⁺ oxidation state and becomes protonated in the Mn²⁺ state (Fig. 1d). We concluded that Tyr34 and His30 are probably the two proton donors needed during the Mn²⁺ to Mn³⁺ redox transition to form H₂O₂ from the protonation of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) (k₂)^28,32,40,41. Overall, the previous MnSOD neutron structures indicate that multiple proton transfers are coupled to electron transfer events and that the active site metal leads to residues with unconventional pK_as essential for PCET catalysis.

How \({{{{\rm{O}}}}}_{2}^{\bullet -}\) interacts with the active site for catalysis is unclear^28,31,32. \({{{{\rm{O}}}}}_{2}^{\bullet -}\) may bind the Mn ion directly for an inner-sphere electron transfer or between His30 and Tyr34 for a long-range outer-sphere electron transfer with the Mn ion^28,31,32,41. The inner sphere binding site for \({{{{\rm{O}}}}}_{2}^{\bullet -}\) is unknown and could form a six or five coordinate active site binding. The sixth site is predicted to bind opposite Asp159⁴². Five coordinate binding would displace Asp159 or WAT1⁴³. Quantum chemistry computational studies have postulated that the first half-reaction (k₁, Mn³⁺ → Mn²⁺) proceeds through an inner-sphere electron transfer while the second half-reaction (k₂, Mn²⁺ → Mn³⁺) proceeds through an outer-sphere electron transfer⁴¹. The outer-sphere mechanism for the second half-reaction is a promising hypothesis because His30 and Tyr34 have been shown to lose protons during the Mn²⁺ → Mn³⁺ redox transition (Fig. 1c, d)¹³. His30 and Tyr34 could protonate \({{{{\rm{O}}}}}_{2}^{\bullet -}\) in concert with long-range electron transfer to form the H₂O₂ product. While there is no structural evidence for how \({{{{\rm{O}}}}}_{2}^{\bullet -}\) interacts with the active site for electron transfer, modeling, and quantum chemical calculations support the second half-reaction proceeding through an outer-sphere mechanism⁴¹.

Human MnSOD is inhibited by its product, H₂O₂, to regulate the output of mitochondrial H₂O₂. Physiologically, MnSOD product inhibition is thought to be related to H₂O₂ acting as a second messenger^31,32,44. Mitochondrially-derived H₂O₂ plays a role in apoptosis^45,46, mitochondrial biogenesis⁴⁴, and protein localization and activity⁴⁷. Furthermore, mitochondrial H₂O₂ has been shown to modulate the activity of PTEN (phosphatase and tensin homolog), a protein tyrosine phosphatase that contributes to cancer and neurological disease^48,49. Mitochondrial H₂O₂ may also be scavenged by peroxiredoxins (PRDXs) and glutathione peroxidases (GPXs). These enzymes are dependent on nicotinamide adenine dinucleotide phosphate (NADPH) for function^50,51,52. MnSOD product inhibition regulates mitochondrial H₂O₂ levels independent of metabolic status. However, little is known about how MnSOD mechanistically achieves product inhibition.

Product inhibition occurs when the ratio of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) to MnSOD is high^31,32. At concentrations of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) that are lower than enzyme, catalysis proceeds rapidly through the first-order reactions k₁ and k₂. At concentrations of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) that are much greater than enzyme, catalysis proceeds through the reversible inhibition reactions k₃ (first-order) and k₄ (zero-order). The k₃ reaction forms a product-inhibited complex (indicated by square brackets) that is unreactive to \({{{{\rm{O}}}}}_{2}^{\bullet -}\) and only disassociates through the zero-order k₄ reaction that dictates the lifetime of the complex. Several kinetic models suggest that inhibition is initiated from the reaction of a Mn²⁺ ion with \({{{{\rm{O}}}}}_{2}^{\bullet -}\) in the absence of protonation to form [Mn³⁺-O₂²⁻] or [Mn³⁺-⁻OOH] (k₃)^31,32. The inhibited complex is then relieved by at least one protonation to yield Mn³⁺ and H₂O₂ (k₄). Since k₂ and k₃ both use Mn²⁺ and \({{{{\rm{O}}}}}_{2}^{\bullet -}\) as reactants, they are competing reactions, and their ratios determine the propensity for MnSOD to become product-inhibited. Human MnSOD has equal k₂ and k₃ (Supplementary Table 1), which means at high \({{{{\rm{O}}}}}_{2}^{\bullet -}\) to enzyme ratios, ~50% of Mn²⁺ reactions with \({{{{\rm{O}}}}}_{2}^{\bullet -}\) form the inhibited complex^53,54. The factors that govern k₃ and k₄ are unclear, although Tyr34 appears to contribute to the inhibition process^55,56.

or

Tyr34 is a conserved residue in the active site of MnSOD^55,57. Physiologically, the residue becomes nitrated in several human neurodegenerative diseases, leading to inactivated MnSOD⁵⁸. During the PCET mechanism, Tyr34 is thought to be one of the sources of protons needed to produce H₂O₂ during the Mn²⁺ to Mn³⁺ redox transition^28,32,57. Our previous neutron structures show that Tyr34 is poised to donate a proton in Mn²⁺SOD (Fig. 1d)¹³. Interestingly, the Tyr34Phe MnSOD variant is unable to proceed through the fast Mn²⁺ to Mn³⁺ redox transition (k₂), and catalysis proceeds exclusively through the product-inhibited pathway k₃ (see rate constants listed in Supplementary Table 1). When k₂ <<k₃ then ~99% of Mn²⁺ reactions with \({{{{\rm{O}}}}}_{2}^{\bullet -}\) form the inhibited complex. There is also higher retention of the inhibited complex, with half the disassociation rate compared to wildtype (see k₄, Supplementary Table 1). Since the active site structure of the Tyr34Phe variant is practically identical to wildtype (Fig. 1e)⁵⁷, the perturbed kinetics and enrichment of product inhibition of the Tyr34Phe variant are probably due to the loss of the hydroxyl group for proton transfer. Interestingly, the Trp161Phe variant, which does not eliminate a hydroxyl group, also primarily uses the product inhibited pathway. In fact, Trp161Phe has similar kinetics to the Tyr34Phe variant (Supplementary Table 1)³³. This suggests that the contribution of Tyr34 toward catalysis is not solely proton transfer. To study the effects that govern the reactions k₁-k₄ of MnSOD, we performed a series of experiments with the Tyr34Phe variant.

Several studies have reported that mixing high H₂O₂ concentrations with Mn³⁺SOD leads to the formation of the product-inhibited complex that decays to Mn²⁺SOD^33,59,60,61. Mutagenesis studies found the process to be correlated with the forward reactions k₂-k₄. These puzzling results may be explained by excessive concentrations of H₂O₂ initiating a backward reaction with Mn³⁺SOD to produce \({{{{\rm{O}}}}}_{2}^{\bullet -}\), and then a subsequent forward reaction of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) with MnSOD to form the inhibited complex^60,61. Importantly, this would provide an avenue to structurally analyze product inhibition without relying on volatile \({{{{\rm{O}}}}}_{2}^{\bullet -}\) solutions⁶². We combined H₂O₂ soaking with the Tyr34Phe variant to study the product-inhibited complex and gain new information on the MnSOD PCET mechanism and an enhanced understanding of product inhibition.

To help understand the role of the hydroxyl group of Tyr34 in MnSOD activity we used the Tyr34Phe MnSOD variant in combination with neutron crystallography, X-ray absorption spectroscopy (XAS), and quantum mechanical (QM) chemistry calculations. With neutron crystallography, we captured the product-bound, reduced, and oxidized states of Tyr34Phe MnSOD free from radiation-induced perturbations. For each state, we used XAS to determine the oxidation state and coordination number of the manganese with each treatment, and to probe the electron orbitals of the Mn ion. Solution XAS data was useful in guiding the interpretation of neutron crystal structures that sometimes have asymmetric active site structures between subunits due to the crystal lattice. Then, we used QM calculations to quantitatively interrogate the Mn ion orbitals that determine redox activity. Lastly, the Tyr34Phe data were compared with wildtype¹³ and Trp161Phe⁶³ to piece together a MnSOD mechanism that describes the reactions k₁-k₄. With respect to oxidoreductases in general, our work presents a thorough characterization of how a single tyrosine modulates PCET catalysis. Our main finding is that Tyr34 plays a part in every MnSOD kinetic step: proton donor/acceptor, orienting nearby residues for rapid proton transfer, limiting the product inhibited complex, and keeping the lifetime of the inhibited state short. Thus, Tyr34 in MnSOD is a central regulator of H₂O₂ levels in the mitochondria and signaling.

X-ray absorption spectroscopy (XAS) characterization of Tyr34Phe MnSOD

Tyr34Phe MnSOD was selected for this study to help understand the role of Tyr34’s hydroxyl group on the MnSOD PCET and product inhibition mechanisms. Solution XAS data is useful in guiding the interpretation of neutron crystal structures. In Mn K-edge XAS, the X-ray absorption near edge structure (XANES) contains information on the oxidation and coordination states of the Mn, while the extended X-ray absorption fine structure (EXAFS) region provides information on Mn-ligand bond distances. The Tyr34Phe MnSOD variant is advantageous in this study because it accumulates and retains the product-inhibited complex, and from that, we can infer information about the PCET mechanism (see Supplementary Table 1 for rate constants). First, we measured the XANES spectral signature to characterize four states of Tyr34Phe MnSOD: oxidized, reduced, H₂O₂ soaked, and \({{{{\rm{O}}}}}_{2}^{\bullet -}\) soaked (see Supplementary Fig. 2 for all Tyr34Phe, Trp161Phe and WT MnSOD HERFD-XANES spectra, and structural comparisons).

The energy of the rising edge of the XANES region increases with higher oxidation states, while the intensity of the pre-edge contains information about the coordination of the Mn ion (Fig. 2a). The pre-edge corresponds to 1 s to 3 d orbital transitions, with gains in intensity from 4p character mixing into the 3 d orbitals from symmetry distortions and/or loss of inversion symmetry^64,65,66. Based on electron paramagnetic resonance (EPR) data, the Mn ion is expected to be high-spin in both Mn³⁺ and Mn²⁺ states⁶⁷. For oxidized Tyr34Phe MnSOD, the rising edge is observed at higher energy than the reduced data, indicating a difference in the Mn ion valency. Given that these samples were mixed with stoichiometric excesses of redox reagents, we interpret oxidized Tyr34Phe MnSOD as Mn³⁺ and reduced Tyr34Phe MnSOD as Mn²⁺. For both oxidized and reduced Tyr34Phe MnSOD, the pre-edge intensity (i.e., the area under the curve, inset Fig. 2a) corresponds to a distorted five-coordinate trigonal bipyramidal complex that has been observed in crystal structures^57,68. A strong pre-edge intensity requires electric dipole allowed transitions that can only occur due to the presence of Mn 4p character mixed into the 3 d orbitals. Such a mixing requires the absence a local inversion center around the Mn ion as is the case for a 5-coordinated complex. For a 6-coordinated complex, a sufficient amount of 4p character could only be reached by one short Mn-ligand bond distance, as it is observed for example for some Fe-O complexes⁶⁶. However, neither our neutron scattering, XAS or DFT data suggest such a geometry.

a The HERFD-XANES region of Tyr34Phe MnSOD treated with potassium dichromate to isolate the Mn³⁺SOD resting state, sodium dithionite to isolate the Mn²⁺SOD resting state, and either superoxide or hydrogen peroxide to isolate the product-inhibited state. b Overlay of the Fourier transform of Mn K-edge EXAFS data [k²χ(k)] from superoxide- and peroxide-soaked Tyr34Phe MnSOD with the EXAFs spectrum in k-space shown in the inset. Due to the scattering phase shift, the distance found by the Fourier Transformation (R) is ~0.5 Å shorter than the actual distance and a 0.5 Å correction (α) was implemented. c, d Superoxide- and peroxide-soaked Tyr34Phe MnSOD with the EXAFs spectrum. Colored lines represent experimental data, while the black line is simulated EXAFS spectra fit to the experimental data.

Full size image

To isolate the product-inhibited complex with XAS, we introduced \({{{{\rm{O}}}}}_{2}^{\bullet -}\) or H₂O₂ to Tyr34Phe MnSOD. \({{{{\rm{O}}}}}_{2}^{\bullet -}\) was used to isolate the complex from the forward reaction and H₂O₂ was used to attain the complex with the back reaction^60,61. For both approaches, the inhibited complex has been reported to form within 44 ms of mixing with Tyr34Phe MnSOD⁶⁰. Indeed, the approaches yield nearly identical XANES spectra and reveal that the same electronic structure is formed (Fig. 2a). The superoxide and peroxide-soaked spectra are like Tyr34Phe Mn²⁺SOD, indicating the presence of a divalent Mn ion. However, these spectra have unique features relative to Tyr34Phe Mn²⁺SOD, with lower intensities at ~6550-6555 eV and higher intensities at ~6565-6575 eV. Similarly, the pre-edge intensities are like Tyr34Phe Mn²⁺SOD to indicate that the resulting complex from either mixing \({{{{\rm{O}}}}}_{2}^{\bullet -}\) or H₂O₂ is five-coordinate (Fig. 2a). Altogether, by using the Tyr34Phe MnSOD variant that enriches for the product-inhibited complex, we show that an electronically distinct five-coordinate Mn²⁺ complex forms from either exposure to \({{{{\rm{O}}}}}_{2}^{\bullet -}\) or H₂O₂.

Next, we investigated the EXAFS region of the \({{{{\rm{O}}}}}_{2}^{\bullet -}\) and peroxide-soaked Tyr34Phe MnSOD K-edge spectra. The Fourier transform of the EXAFS data, χ(k), yields χ(R) that provides information on the atomic radial distribution (bond lengths) around the absorbing Mn ion (Fig. 2b–d). Overall, the k-space (inset) and R-space spectra are highly similar for \({{{{\rm{O}}}}}_{2}^{\bullet -}\) and H₂O₂ soaked Tyr34Phe MnSOD and yield the same Mn bond distance solutions (Table 1). For both samples, the first shell of coordination observed at ~2.1 Å is best fit by three N atoms at 2.15 Å and two O atoms at 2.11 Å to indicate a five-coordinate complex (Supplementary Table 2 and 3). These account for N^ε2 atoms of His26, His74, and His163, O^δ2 atom of Asp159, and another O atom (denoted O¹). As the next closest scatterer is a single atom found at 2.44 Å (denoted O²), we interpret O¹ and O² as belonging to a dioxygen species. The largest contributors to the second peak seen at ~3.0 Å are seven carbon atoms that correspond to the C^δ2 and C^ε1 of the three histidines and the C^γ of the aspartate residue. Overall, the EXAFS data show that when \({{{{\rm{O}}}}}_{2}^{\bullet -}\) or H₂O₂ are given to Tyr34Phe MnSOD, a five-coordinate Mn complex that includes a dioxygen species is formed.

Our XAS data indicate that H₂O₂ has both a reducing effect on MnSOD and can form the inhibited complex. The reducing effect of MnSOD by H₂O₂ has been reported previously^59,60. Thermodynamically, the process is unfavorable as the reduction potential of wildtype Mn³⁺SOD to Mn²⁺SOD is 0.4 V, and the oxidation potential of H₂O₂ to \({{{{\rm{O}}}}}_{2}^{\bullet -}\) is -0.85 V⁶⁹. However, a 100-fold excess of H₂O₂ was used, probably driving a backward reaction^60,61. Mutagenesis studies have indicated that forming the inhibited complex through peroxide-soaking depends on reactions k₂-k₄ and suggests that the \({{{{\rm{O}}}}}_{2}^{\bullet -}\) generated from the backward reaction then reacts with MnSOD in the forward direction⁶⁰. In these past experiments, detection of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) was lost during the dead time of the instrument (1.4 ms). As H₂O₂ is more stable than \({{{{\rm{O}}}}}_{2}^{\bullet -}\) in aqueous solution, we used H₂O₂ for the structural studies on Tyr34Phe MnSOD.

Neutron structure of the Tyr34Phe product-inhibited complex

To visualize the product-inhibited complex and the corresponding active site protonation states, we solved a cryocooled neutron structure of perdeuterated Tyr34Phe MnSOD soaked with deuterium peroxide (D₂O₂) at 2.30 Å resolution. For the neutron structures, proton positions can be assigned at 2.5 Å or better. We first solved the all-atom structure of the entire enzyme, excluding the active site. Then, with these phases, we carefully interrogated the active site nuclear density maps for the active site coordination and protonation states.

For chain A, the omit |F_o | - |F_c| nuclear scattering-length density next to the Mn ion was oblong at 3.0 σ (Fig. 3a). We interpreted the density as a dioxygen species with a single proton (denoted as LIG for ligand) that has taken the place of the WAT1 position observed in the resting states (Fig. 1b). The proton of LIG, D², points toward Trp161 and appears to be interacting with the pi system (Fig. 3a). The O²(LIG) and D^ε21(Gln143) atoms are close in proximity (2.2 Å apart) though LIG and Gln143 are not in optimal geometry for hydrogen bonding (Figs. 3a, b and Supplementary Fig. 1a). LIG refined well at full occupancy and led to Mn bond distances that closely resemble those measured from the EXAFS spectra (Table 1). At physiological temperatures, optical absorption spectra in the literature also suggest a displacement of WAT1 and binding of a dioxygen species, which agrees with our cryocooled diffraction data⁴³. For chain B, the omit |F_o | - |F_c| is instead a spherical shape at 3.0 σ (Supplementary Fig. 1b), indicating a ^-OD molecule that has been previously observed in wildtype neutron structures¹³. Contrasting structures between the two subunits of the asymmetric unit are often observed for the MnSOD P6₁22 crystal form due to differences in solvent accessibility^13,70. In chain B, the six-coordinate nuclear density near the Mn ion and ⁻OD molecule did not correspond with the five-coordinate XAS data, making it less useful in terms of defining the PCET mechanism (Supplementary Fig. 1b). Regardless, from the nuclear density at chain A, a singly protonated dioxygen species replaces WAT1 upon D₂O₂ soaking and forms a complex with bond distances like those found from EXAFS spectra (Table 1).

a D₂O₂-soaked Tyr34Phe MnSOD for chain A with a singly-protonated dioxygen species denoted LIG. The inset highlights the elongated |F_o | - |F_c| difference density of LIG. b D₂O₂-soaked Tyr34Phe MnSOD for chain A rotated 90° along the horizontal axis relative to panel (a). c Tyr34Phe Mn²⁺SOD for chain A. Notably, this neutron structure is the lowest resolution, 2.5 Å of the three in Supplementary Table 6. d Active site overlay of wildtype Mn²⁺SOD (PDB 7KKW, solved at 2.3 Å resolution) and Tyr34Phe Mn²⁺SOD (this study, solved at 2.5 Å resolution). e Tyr34Phe Mn³⁺SOD for chain A. f Active site overlay of wildtype Mn³⁺SOD (PDB 7KKS, solved at 2.3 Å resolution) and Tyr34Phe Mn³⁺SOD (this study, solved at 2.3 Å resolution). Green, orange, and magenta omit |F_o | - |F_c| difference neutron scattering length density of protons displayed at 2.5σ, 3.0σ, and 3.5σ. respectively. Light blue 2|F_o | - |F_c| density is displayed at 1.0σ. Distances are in Å. Dashed lines indicate hydrogen bonds ≥ 1.8 Å, and hashed lines indicate SSHBs that are hydrogen bonds < 1.8 Å. Chain B active sites are shown in Supplementary Fig. 1.

Full size image

How the product-inhibited complex is formed as well as its oxidation state and coordination, have been historically controversial^{31,32,61,71,72,73}. The generally accepted mechanism is that inhibition is initiated from \({{{{\rm{O}}}}}_{2}^{\bullet -}\) binding Mn²⁺ to form a Mn³⁺-O₂^2- complex (k₃, Supplementary Table 1)^{28,31,32,74,75}. Our XAS data indicates a five-coordinate Mn²⁺ as the identity of the complex^59,60. Indeed, the neutron structure of Tyr34Phe MnSOD soaked with D₂O₂ yields a five-coordinated active site with a singly-protonated dioxygen species. The Mn bond distances of the solved neutron structure with Mn²⁺ and a hydroperoxyl anion⁻,O₂H, resemble those from the EXAFS data. Also, the density function theory (DFT) calculations using Mn²⁺ and ⁻O₂H most closely match the neutron structure (Table 1). Overall, these data and calculations indicate that the inhibited complex is a five-coordinate Mn²⁺ complex where the WAT1 position has been replaced with ⁻O₂H.

Tyr34 orients the Gln143-WAT1 SSHB of Mn²⁺SOD and limits product inhibition

We next wondered whether enrichment of the product-inhibited complex in the Tyr34Phe variant was because of structural perturbations near the Mn ion or due to the absence of the Tyr34 hydroxyl group for proton transfer. To this end, we solved neutron structures of reduced and oxidized Tyr34Phe MnSOD at 2.5 and 2.3 Å resolution, respectively. For reduced Tyr34Phe MnSOD, the Mn bond distances are similar to five-coordinate wildtype Mn²⁺SOD (Supplementary Table 4). Furthermore, the protonation states resemble those of wildtype Mn²⁺SOD, where WAT1 is of the D₂O form while Gln143 is deprotonated to the amide anion (Fig. 3c). Deprotonated amino acids are identified when attempts to model and refine a proton result in negative |F_o | - |F_c| difference neutron scattering length density, and all the other protons of the amino acid can be placed. Interestingly, the lack of the hydroxyl group in Tyr34Phe Mn²⁺SOD perturbs the orientation of Gln143 and lengthens the WAT1-Gln143 hydrogen bond (Fig. 3d). The WAT1-Gln143 SSHB of the wildtype enzyme is critical for the back-and-forth proton transfers needed for redox cycling of the Mn ion¹³, and the Tyr34Phe Mn²⁺SOD neutron structure suggests one role Tyr34 plays in catalysis is correctly positioning Gln143 for rapid PCET catalysis. Indeed, the Mn²⁺ to Mn³⁺ redox transition is greatly diminished for Tyr34Phe MnSOD (k₂, Supplementary Table 1)^31,32. Another consequence of the perturbation of the WAT1-Gln143 interaction is the potentially easier displacement of WAT1 by a dioxygen species to form the product-inhibited complex. Overall, the reduced Tyr34Phe MnSOD neutron structure suggests that Tyr34 orients Gln143 for a tight hydrogen bonding interaction with WAT1.

For oxidized Tyr34Phe MnSOD, the Mn ion is like wildtype with a bound ⁻OD molecule (Fig. 3e) with similar covalent bond distances (Supplementary Table 4), and identical hydrogen bond distances for O(WAT1)-D^ε21(Gln143) and D^ε1(Trp123)-O^ε1(Gln143) (Fig. 3f). As PCET mechanisms depend on the distribution of electrostatic vectors to propagate proton and electron transfers, the missing anionic deprotonated Tyr34 in this variant probably perturbs enzyme kinetics. Such electronic effects are indicated by the lower kinetic rates of the Mn³⁺ to Mn²⁺ redox transition for the Tyr34Phe variant (k₁, Supplementary Table 1). Overall, the interactions between WAT1, Gln143, and Trp123 are the same in Tyr34Phe and wildtype Mn³⁺SOD making this variant. Because these active sites are practically identical the use of Tyr34Phe MnSOD and its handy kinetic data to study the product inhibited state is valid and can help interpret the wildtype mechanism.

The neutron structures of reduced and oxidized Tyr34Phe Mn²⁺SOD suggest the roles of Tyr34 in catalysis are (1) orient Gln143 for efficient interaction and proton transfer with WAT1 during the Mn²⁺ to Mn³⁺ PCET reaction, (2) limit formation of the product-inhibited complex by strengthening the WAT1-Gln143 hydrogen bond, and (3) contribute subtle electronic effects as a phenolate anion during the Mn³⁺ to Mn²⁺ reaction. These interpretations are supported by the kinetic rates of Tyr34Phe MnSOD^31,32. For Tyr34Phe MnSOD, the fast Mn²⁺ to Mn³⁺ redox reaction is dramatically reduced (k₂, Supplementary Table 1), formation of the product-inhibited complex is enriched (k₃ >> k₂, Supplementary Table 1), and the Mn³⁺ to Mn²⁺ redox reaction is cut in third (k₁, Supplementary Table 1). A unifying theme among Tyr34Phe MnSOD and other variants studied kinetically is that the precise orientation of Gln143 is critical for catalysis. Mutating residues directly adjacent to Gln143, such as Trp161, Trp123, and Tyr34, lead to similar kinetic consequences, where the Mn²⁺ → Mn³⁺ half reaction is greatly diminished, product inhibition is enriched, and the Mn³⁺ → Mn²⁺ half reaction is slower compared to wildtype^31,32. These redundant effects of the Trp161Phe, Trp123Phe, and Tyr34Phe variants suggest that a crucial role of residues neighboring Gln143 is to enforce a SSHB between WAT1 and Gln143. In the case of Tyr34, our neutron structures indicate that Tyr34 is responsible for positioning Gln143 for a SSHB with WAT1 during the Mn²⁺ resting state and that the strength of the bond between Gln143 and WAT1 correlates with the extent of product inhibition.

Retention of the product-inhibited complex is dependent on the Gln143 position

As the XAS and neutron crystallographic data of Tyr34Phe MnSOD demonstrated the formation of a five-coordinated Mn²⁺SOD with a singly-protonated dioxygen species replacing the WAT1 position, we wondered if the formation and retention of the inhibited complex could be distinguished with other variants of MnSOD. In the context of Tyr34Phe MnSOD, we also sought to define whether the enrichment of product inhibition was due to the loss of the ionizable Tyr group only or also due to the perturbation of the Gln143-WAT1 SSHB interaction. For example, the Trp161Phe MnSOD variant alters a residue directly adjacent to Gln143 and has enriched product-inhibition kinetics like those of Tyr34Phe MnSOD (Supplementary Table 1)⁵⁹. For wildtype, Tyr34Phe, and Trp161Phe, we performed XANES in the HERFD mode of detection, which allows a large improvement in energy resolution and sensitivity compared to conventional XANES^76,77,78. Here, we focused on comparing the reduced form of the variants to those of peroxide-soaked counterparts. As all these complexes are expected to be five-coordinated d⁵ with spin S = 5/2, we sought to distinguish fine details of the spectra offered by the HERFD mode of detection (detailed evaluation of these data can be found in the Supplementary Information pages 10-13 and Supplementary Fig. 3). Overall, reduced and peroxide-soaked Tyr34Phe MnSOD spectra had distinct features between each other, and the general shape trends were reproducible with conventional XANES (Fig. 2a) and HERFD-XANES (Supplementary Fig. 3a). Similar trends were observed between reduced and peroxide-soaked forms of Tyr34Phe and Trp161Phe MnSOD (Supplementary Fig. 3ab). Since the wildtype enzyme exhibits physiological product inhibition levels (k₃ = k₂), compared to the Tyr34Phe and Trp161Phe variants that are enriched for product inhibition (k₃ >> k₂ and lower k₄ compared to wildtype, Supplementary Table 1), a potential explanation for the less pronounced differences for the wildtype pair is that the peroxide-soaked data may reflect a mixture of species rather than a fully isolated product-bound complex (Supplementary Fig. 3c). The XERFD-XANES data for peroxide soaked Y34F, W161F and wild-type are similar and follow these trends (Supplementary Fig. 3d). These observations indicate that retention of the inhibited complex is not exclusively dependent on a proton transfer from Tyr34 but instead on other factors of the active site. We compared our D₂O₂-soaked Tyr34Phe MnSOD neutron structure with that of our previous D₂O₂-soaked Trp161Phe MnSOD neutron structure (Supplementary Fig. 3e, PDB ID 8VHW⁶³ and concluded that the LIG orientation in Tyr34Phe MnSOD is probably close to what would occur physiologically. When the reduced structures are compared, both variant structures have a lengthened WAT1-Gln143 interaction compared to wildtype (Supplementary Fig. 3f, PDB ID 8VHY)⁶³. Altogether, these comparisons show the importance of the WAT1-Gln143 interaction for catalysis and hydrogen bonding correlates with longer retention of the inhibited complex.

The electronic configuration of the Mn ion

For enzymes that use metal centers to catalyze redox reactions, the arrangement of the metal 3 d orbitals determines how electrons are exchanged and how substrates orient for catalysis⁷⁹. A detailed analysis of the electronic configuration of the Mn ion is presented in the Supplementary information and is summarized here. For MnSOD, the metal ion is in a distorted C_3v symmetry environment with either 4 or 5 occupied electrons in the α-manifold, depending on the oxidation state of the metal⁶⁷. For Mn³⁺ with S = 2, the e_π (xz/yz) and e_σ (xy/x²-y²) α orbitals are occupied, while Mn²⁺ with S = 5/2 also has the z² α orbital occupied (Supplementary Fig. 4a). The z² α orbital exchanges electrons during redox reactions as it is the lowest unoccupied molecular orbital (LUMO) for Mn³⁺ and the highest occupied molecular orbital (HOMO) for Mn²⁺. This means that mono- or dioxygen species are most likely to bind the Mn ion along the z-axis for reactivity (Supplementary Fig. 4b). For further insight into the MnSOD metal 3 d orbitals, we compared the K-pre-edge spectra found through time-dependent DFT (TD-DFT) simulations with spectra measured from HERFD-XANES (detailed evaluation of these data can be found in the Supplementary Information pages 16-18 and Supplementary Table 5 and Fig. 4). We found that the experimental pre-edge spectra are similar between wildtype and Tyr34Phe Mn³⁺SOD, indicating that the Tyr34Phe variant does not significantly alter the Mn³⁺ ion orbital configuration (Supplementary Figs. 4c, d). Also, the H₂O₂-soaked Trp161Phe and Tyr34Phe complexes have similar experimental spectra though the calculated spectra suggest differences in orbital transitions (Supplementary Figs. 4g, h).

a, b Neutron structures of D₂O₂-soaked Tyr34Phe MnSOD (2.3 Å resolution). c, d Neutron structures of Tyr34Phe Mn²⁺SOD (2.5 Å resolution). e, f Neutron structures of Tyr34Phe Mn³⁺SOD (2.3 resolution). Blue, green, and orange omit |F_o| - |F_c| difference neutron scattering length density of protons displayed at 2.0 σ, 2.5σ, and 3.0σ, respectively. Light blue 2|F_o | - |F_c| density is displayed at 1.0σ. Distances are in Å. Dashed lines indicate hydrogen bonds ≥ 1.8 Å and hashed lines indicate SSHBs that are hydrogen bonds < 1.8 Å. Chains are colored according to the inset in Fig. 4a.

Full size image

Tyr34 contributes to the pK_as of Tyr166/His30 and influences the proton shuffle across the subunit interface

Second-sphere residues His30 and Tyr166 have unusual pK_as in wildtype MnSOD and change protonation states (Fig. 1c)¹³. To identify whether the Tyr34 mutation to Phe affects the protonations of His30/Tyr166, we investigated the nuclear density maps of Tyr34Phe MnSOD in D₂O₂-soaked, reduced, and oxidized forms. For chain A of D₂O₂-soaked MnSOD that has a dioxygen species bound (Fig. 3a, b), the omit |F_o | - |F_c| density indicates that Tyr166 is in the neutral, protonated form while His30 is singly protonated on the N^δ1 atom (Fig. 4a). Interestingly, for chain B, Tyr166 is instead deprotonated and His30 is now protonated on N^ε2 (Fig. 4b). The two residues form a 1.7 Å SSHB due to the negative charge of ionized Tyr166. Different chains of the same MnSOD neutron structure have been observed before to differ in protonation states, with the most explicable cause being differences in solvent accessibility¹³. However, the protonation configuration observed in chain B is unique and has not been seen before (Fig. 4b). It is not known if this proton configuration occurs in wildtype enzyme as neutron data for D₂O₂-soaked wildtype has yet to be collected. But we can compare reduced and oxidized Tyr34Phe MnSOD to wildtype¹³.

For reduced Tyr34Phe MnSOD, both chains have the same protonation states (Fig. 4c, d). Tyr166 is protonated and neutral while His30 is singly-protonated on the N^δ1 atom. The omit density for D^η(Tyr166) is elongated and spans to N^ε2(His30) (Fig. 4d). Elongated density has been observed between the two residues before in reduced wildtype MnSOD¹³. This may be indicative of a proton transfer occurring between Tyr166 and His30, in line with the different protonation states observed in the D₂O₂-soaked structure (Fig. 4a, b).

The protonation states of oxidized Tyr34Phe (Fig. 4e, f) are the same as reduced Tyr34Phe (Fig. 4c, d), which is in contrast to wildtype MnSOD where the His30 protonation alters between oxidized and reduced states (Fig. 1c). This difference is probably due to the Phe mutation at position 34. In wildtype, Tyr34 is ionized when the Mn ion is oxidized (Fig. 1d). The negative charge on Tyr34 may exert electronic effects that alter the pK_a of His30. Thus, Tyr34 may help modulate the pK_a of nearby residues so efficient proton transfers occur for PCET catalysis.

Our three neutron structures of Tyr34Phe MnSOD highlight the importance of residues Tyr166, His30, and Tyr34 in maintaining the proton pool of the active site. Two protons are needed to protonate O₂²⁻ to H₂O₂ during the fast Mn²⁺ → Mn³⁺ reaction (k₂, Supplementary Table 1). The D₂O₂-soaked structure unambiguously indicates that His30 and Tyr166 shuffle protons, with proton transfers between O^η(Tyr166) and N^ε2(His30) coinciding with changes in the N^δ1(His30) protonation state (Fig. 4a, d). In wildtype MnSOD, the proton between O^η(Tyr166) and N^ε2(His30) was instead observed to be shared, perhaps because the wildtype structures were collected at room temperature (Fig. 1c, d). Another distinction between Tyr34Phe and wildtype MnSOD is that in the oxidized forms, the N^δ1(His30) have different protonation states. For Tyr34Phe, N^δ1(His30) is protonated (Fig. 4e, f) while in wildtype N^δ1(His30) is deprotonated (Fig. 1c). The pK_as of active site residues are a result of multiple effects, including residue composition and ionization states. An ionized Tyr34 coincides with a deprotonated N^δ1(His30) in wildtype MnSOD and suggests that Tyr34 contributes to modulating the pK_a of nearby residues for oxidized MnSOD. Overall, the Tyr34Phe MnSOD neutron structures help elucidate the role of Tyr166, His30, and Tyr34 in MnSOD catalysis.

Summary and comparison of Tyr34Phe active site configurations

Comparing the Tyr34Phe MnSOD neutron structures with our previous wildtype and Trp161Phe neutron structures helps our understanding of the role of Tyr34 in catalysis. For oxidized wildtype and Tyr34Phe MnSOD, the WAT1-Gln143 protonation states and hydrogen bond interaction are similar (Fig. 5a, b). However, in reduced Tyr34Phe, His30 is observed to have different protonation states compared to wildtype, and the proton between His30 and Tyr166 is not shared. For the reduced counterparts, the hydrogen bonding of Gln143 with WAT1 and Trp123 is weaker in Tyr34Phe MnSOD (Fig. 5c, d). The Gln143 hydrogen bonds in wildtype are SSHBs (<1.8 Å), while those of Tyr34Phe are typical hydrogen bonds (>1.8 Å). Reduced wildtype is observed to have a shared proton between His30 and Tyr166 while Tyr34Phe does not. However, chain B of reduced Tyr34Phe MnSOD is observed to have elongated nuclear density between His30 and Tyr166 (Fig. 4d), which indicates that a proton may be shared. For peroxide-bound Trp161Phe and Tyr34Phe, HO₂^- binds the Mn ion in different orientations (Fig. 5e, f). Furthermore, in Trp161Phe, a H₂O₂ molecule is bound with SSHBs between His30 and an anionic Tyr34. The lack of H₂O₂ at this site in Tyr34Phe may be due to the absence of the hydroxyl group. Overall, Tyr34 plays a role in several aspects of the active site, and several enzymatic details are revealed from this collection of neutron structures.

a Wildtype oxidized resting state from chain B of PDB ID 7KKS¹³. b Tyr34Phe oxidized resting state from chain A (present study, Figs. 3e, 4e). c Wildtype reduced resting state from chain B of PDB ID 7KKW¹³. d Tyr34Phe reduced resting state from chain A (present study, Figs. 3c, 4c). e The product-inhibited complex of Trp161Phe, consisting of a HO₂^- molecule bound to a reduced active site. From chain B of PDB ID 8VHW⁶³. f The product-inhibited complex of Tyr34Phe, consisting of a HO₂^- molecule bound to a reduced active site. From chain A of the present study (Figs. 3a, b, 4a). Dashed lines represent normal hydrogen bonds (> 1.8 Å), wide-dashed represent SSHBs (hydrogen bonds < 1.8 Å), and round-dotted lines represent a shared proton. The portrayal of hydrogen bond lengths in 2D are not representative of those seen experimentally in 3D.

Full size image

First, Tyr34 helps control the charge at the active site and provides an environment conducive to charge-dependent proton and electron transfers. For example, Tyr34 and His30 both lose protons when the active site is oxidized and gain protons when the active site is reduced (Figs. 5a, 5c, 1c, d). As a result, we previously hypothesized that Tyr34 and His30 protonate O₂^2- during the fast Mn²⁺ → Mn³⁺ reaction (k₂, Supplementary Table 1)¹³. In the Tyr34Phe variant, the His30 protonation state remains consistent between resting states, which implicates Tyr34 in influencing the pK_a of His30 through electronic effects (Figs. 5b, d, 4c-f). These charge effects of Tyr34 may be a partial contributor to the deficient catalysis observed for the Tyr34Phe variant (Supplementary Table 1).

Second, Tyr34 plays a significant role in the fast Mn²⁺ → Mn³⁺ reaction as suggested by the very low k₂ for Tyr34Phe (Supplementary Table 1). Comparison of the resting Mn²⁺SOD neutron structures of wildtype and Tyr34Phe indicates that Tyr34 enforces a strong WAT1-Gln143 interaction (Figs. 5c, d, 3d). This interaction is important for PCET, where the WAT1 → Gln143 proton transfer coincides with the Mn²⁺ → Mn³⁺ redox reaction (Fig. 1b)¹³. Another potential reason for diminished k₂ catalysis for the Tyr34Phe variant is the loss of an ionizable group that could protonate O₂^2- or HO₂^-. However, the Trp161Phe variant also has a nearly extinguished k₂ and maintains the number of ionizable groups (Supplementary Table 1). The shared feature for these two variants is the weakened WAT1-Gln143 interaction (Fig. 4f). Since our previous structures indicate that Tyr34 loses a proton during the Mn²⁺ → Mn³⁺ reaction (Fig. 1d), it is possible that the WAT1 → Gln143 proton transfer precludes a Tyr34 → O₂²⁻ proton transfer. Overall, our structures indicate Tyr34 contributes to orienting WAT1 and Gln143 for a proton transfer event during the Mn²⁺ → Mn³⁺ redox reaction.

Third, the formation of the product-inhibited complex is dependent on Tyr34. The complex is characterized by an HO₂⁻ molecule replacing the WAT1 position in the Mn²⁺ oxidation state (Fig. 5e, f, Supplementary Fig. 3e). As such, the capacity to form the inhibited complex may correlate with the ease an HO₂^- molecule can displace WAT1. The Tyr34Phe and Trp161Phe variants have a higher propensity to accumulate the inhibited complex, and both have a weakened WAT1-Gln143 interaction in the Mn²⁺ oxidation state that would permit easier displacement of WAT1 (Fig. 4f). This suggests that the Tyr34-Gln143-WAT1 hydrogen bond network suppresses product inhibition.

Lastly, retention of the product-inhibited complex correlates with the strength of hydrogen bonding between HO₂⁻ and Gln143. The inhibited complexes captured in the Tyr34Phe and Trp161Phe variants reveal two different HO₂⁻ binding orientations and hydrogen bond interactions (Fig. 5e, f, Supplementary Fig. 3e). The hydrogen bonding between HO₂⁻ and Gln143 is stronger in Trp161Phe compared to Tyr34Phe, and this stronger interaction correlates with a slower disassociation of the inhibited complex for Trp161Phe (k₄, Supplementary Table 1). As protonation of HO₂⁻ to H₂O₂ is required for relief of the inhibited complex, the hydrogen bonding of Gln143 with HO₂⁻ may compete with a H₂O → HO₂⁻ proton transfer to form H₂O₂. Comparison of the product-inhibited complex of Tyr34Phe and Trp161Phe provides clues into the relief of the inhibited complex.

Suggested mechanism

From our collection of neutron structures and XAS data, we have constructed a mechanism for the fast reaction pathways k₁ and k₂ and the reversible product inhibition reaction pathways k₃ and k₄ (Fig. 6). Note that our data describe the inhibited complex as Mn²⁺-containing (Figs. 2a, Supplementary Fig.3a–c), in contrast with several mechanistic models that assign a Mn³⁺ oxidation state to the complex^{31,32,33,59,60}. Furthermore, in the absence of data that describe how \({{{{\rm{O}}}}}_{2}^{\bullet -}\) interacts with an oxidized active site, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) reacting with the Mn³⁺ ion is represented only by Mn³⁺ gaining an electron (Fig. 6). For a reduced active site, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) likely binds between His30 and Tyr34 for PCET to form H₂O₂^13,41. Overall, the mechanism delineates how MnSOD’s reversible inhibition pathway may branch from the fast reaction pathway.

The associating and dissociating molecules are indicated by the reaction on the left side of the figure for each panel. a, b The fast Mn³⁺ → Mn²⁺ half-reaction corresponding to k₁. Note that the electrons are contributed to by superoxide binding. The ___location of this binding site is still unknown, so this is indicated with a thick arrow. c, d The fast Mn²⁺ → Mn³⁺ half-reaction corresponding to k₂. e, f Formation of the inhibited complex that is dependent on the presence of H₂O₂ during the reduction of the Mn³⁺ ion. In the absence of H₂O₂, the enzyme proceeds through the fast Mn³⁺ → Mn²⁺ half-reaction (k₁). g, h Relief of the inhibited complex. Our data describe the inhibited complex as Mn²⁺-containing in contrast to other published mechanistic models^{31,32,33,59,60}. In the absence of data that indicate how \({{{{\rm{O}}}}}_{2}^{\bullet -}\) interacts with an oxidized active site, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) reacting with the Mn³⁺ ion is represented only by Mn³⁺ gaining an electron. For a reduced active site, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) likely binds between His30 and Tyr34 for PCET to form H₂O₂^13,41. Dashed lines represent normal hydrogen bonds (>1.8 Å), wide-dashed represent SSHBs (hydrogen bonds < 1.8 Å), and round-dotted lines represent a shared proton. The portrayal of hydrogen bond lengths in 2D are not representative of those seen experimentally in 3D.

Full size image

For the k₁ reaction that represents the fast Mn³⁺→ Mn²⁺ half-reaction, reduction of the Mn ion occurs alongside three proton transfers, Gln143 → WAT1 and protonation of His30 and Tyr34 by solvent (Fig. 6a). In the reduced state, anionic Gln143 is stabilized by SSHBs with Trp123, WAT1, and Tyr34 (Fig. 6b). For the k₂ reaction that represents the fast Mn²⁺ → Mn³⁺ half-reaction, a WAT1 → Gln143 proton transfer occurs alongside protonation and reduction of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) (Fig. 6c). Here, \({{{{\rm{O}}}}}_{2}^{\bullet -}\) participates in a long-range electron transfer and is protonated by His30 and Tyr34 to form H₂O₂ (Fig. 6d). If Mn³⁺ gains an electron while the resultant H₂O₂ is still present, H₂O₂ donates a proton to His30 and subsequently replaces WAT1 to form the product inhibited complex (k₃, Fig. 6e, f). The inhibited complex is characterized by the inability to perform the WAT1 → Gln143 proton transfer and prohibits fast catalysis (Fig. 6f). Relief of the inhibited complex is achieved by protonation of HO₂⁻ by a solvent molecule (k₄, Fig. 6g). H₂O₂ departs the active site and is replaced by WAT1 to form the Mn²⁺ resting state (Fig. 6h). From relief of inhibition, fast catalysis may proceed again through k₂ (Fig. 6c, d).

The key determinant of whether product inhibition is engaged is the His30 proton donor during Mn³⁺ reduction. If H₂O₂ is the proton donor, the inhibited complex is formed (k_3, Fig. 6e). If a solvent molecule is the proton donor, the enzyme proceeds through fast catalysis (k₁, Fig. 6a). This explains why high concentrations of \({{{{\rm{O}}}}}_{2}^{\bullet -}\) lead to product inhibition, as \({{{{\rm{O}}}}}_{2}^{\bullet -}\) may enter the active site before H₂O₂ departs. Tyr34 may also serve as a potential proton acceptor for H₂O₂ since Tyr34 also gains a proton during reduction to Mn²⁺. However, Tyr34 is not necessary to form the inhibited complex, as indicated by our Tyr34Phe neutron structure (Fig. 3a, b). Our work provides insights into the PCET mechanism of MnSOD, especially in the context of product inhibition.

Altogether, our investigation reveals how a single tyrosine residue has a profound effect on PCET catalysis. Tyr34 plays a part in every MnSOD kinetic step from its roles of (1) acting as a proton donor/acceptor, (2) orienting nearby molecules Gln143 and WAT1 for proton transfer, (3) limiting the formation of the inhibited complex, and (4) shortening the lifetime of the inhibited complex. These roles of Tyr34 place the residue as a central regulator of H₂O₂ output in the mitochondria. H₂O₂ produced from MnSOD has several cellular effects, including stimulating apoptotic signaling pathways^45,46, coordinating protein localization and activity⁴⁷, and mitochondrial biogenesis⁴⁴. Inactivation of Tyr34 by nitration is observed in neurological disease and further highlights the physiological role of Tyr34⁵⁸. In total, the work provides insight into how a PCET enzyme facilitates catalysis and how an oxidoreductase molecularly facilitates mitochondrial function.

Perdeuterated expression and purification

For deuterated protein expression of MnSOD, the pCOLADuet-1 expression vector harboring full-length cDNA of MnSOD was transformed into Escherichia coli BL21(DE3) cells. Transformed cells were grown in D₂O minimal media within a bioreactor vessel using D₈-glycerol as the carbon source⁸⁰. Induction was performed with 1 mM isopropyl β-D-thiogalactopyranoside, 8 mM MnCl₂, and fed D₈-glycerol until an OD₆₀₀ of 15.0. Expression was performed at 37 °C for optimal Mn metal incorporation⁸¹. Harvested cell pastes were stored at -80 °C until purification. For protein purification (with hydrogenated reagents), cells were resuspended in a solution of 5 mM MnCl₂ and 5 mM 3-(N-morpholino)propanesulfonic acid (MOPS), pH 7.8. Clarified lysate was incubated at 55 °C to precipitate contaminant proteins that were subsequently removed by centrifugation. Next, soluble protein was diluted with an equal volume of 50 mM 2-(N-morpholino)ethanesulfonic acid (MES) pH 5.5, yielding a final concentration of 25 mM. Measurement of pH verified a value of 5.5 after dilution. Protein was applied onto a carboxymethyl sepharose fast flow column (GE Healthcare) and eluted with a sodium chloride gradient that contained 50 mM MES pH 6.5.

Crystallization

Perdeuterated Tyr34Phe MnSOD crystals were grown in a microgravity environment aboard the International Space Station (ISS)⁸². Crystals growth was achieved in Granada Crystallization Boxes (GCBs, Triana) through capillary counterdiffusion using fused quartz capillary tubes (VitroCom) that had inner diameters of 2.0 mm and outer diameters of 2.4 mm⁸³. 25 mg ml^-1 protein-filled capillaries were plugged with 40 mm of 2% agarose (w/w) and inserted into GCBs filled with precipitating agent composed of 4 M potassium phosphate, pH 7.8. The pH of the phosphate buffer was achieved through 91:9 ratios of K₂HPO₄:KH₂PO₄. The GCBs were delivered to the ISS by SpX-17 as part of the Perfect Crystals NASA payload and returned to earth 1 month later on SpX-18. The crystals within GCBs were observed to be resilient against travel damage and were placed within carry-on baggage during further aircraft travels to the UNMC Structural Biology Core Facility and ORNL. Perdeuterated Tyr34Phe MnSOD crystals were 0.3-0.6 mm³ in size, and further details of microgravity crystallization have been published⁸². For X-ray diffraction, crystals were grown in 1.8 M potassium phosphate, pH 7.8 by hanging-drop vapor diffusion. Protein (23 mg ml^-1) and reservoir solution were mixed at a 1:1 ratio to give a 4.0 µL drop. Crystals for X-ray diffraction were less than 0.1 mm³ in size and were fully grown after 14 d.

Crystal manipulations

For deuterium exchange, microgravity-grown crystals were first placed in 1 mL of hydrogenated 4 M potassium phosphate pH 7.8. Deuterium was introduced with 0.1 mL incremental additions every 2 min of 4 M deuterated potassium phosphate (K₂DPO₄:KD₂PO₄) pD 7.8 (calculated by adding 0.4 to the measured pH reading) for a total of five times and a net volume addition of 0.5 mL. After 10 min, 0.5 mL of the solution was removed leading to a 1 mL solution consisting of 33% deuterium. The process was repeated enough times to gradually increase the deuterium content to ~100%. The 4 M deuterated potassium phosphate also served as the cryoprotectant for the cryocooling process. Further details of the process were published⁸⁴.

For redox manipulation, the deuterated potassium phosphate solutions were supplemented with either 6.4 mM potassium permanganate (KMnO₄) to achieve the Mn³⁺ oxidation state or 300 mM sodium dithionite (Na₂S₂O₄) to achieve the Mn²⁺ state. Crystals were either sealed in capillaries or 9-well glass plates to maintain the desired oxidation state. Redox reagents were not used for the Tyr34Phe MnSOD structure soaked with D₂O₂. The crystal was in the as is resting state, which is >90% oxidized. The dioxygen-bound complex was achieved by supplementing the cryoprotectant that the crystal was immersed in with D₂O₂ at a final concentration of 1% w/v (~0.28 M) and soaking for 1 min before cryocooling. Flash-cooling was performed with an Oxford diffraction cryostream⁸⁵. Further details of ligand cryotrapping for neutron crystallography were published⁸⁴.

Crystallographic data collection

Time-of-flight, wavelength-resolved neutron Laue diffraction data were collected from perdeuterated crystals using the MaNDi instrument^86,87 at the Oak Ridge National Laboratory Spallation Neutron Source with wavelengths between 2 to 4 Å. Sample sizes ranged from 0.3 to 0.6 mm³ and data were collected to 2.30 Å resolution for oxidized and D₂O₂-soaked structures, while the reduced structure was collected to 2.5 Å resolution (Supplementary Table 6). Crystals were held in stationary positions during diffraction, and successive diffraction frames were collected along rotations of the Φ axis. X-ray diffraction data were collected using a Rigaku FR-E SuperBright home source or the Stanford Synchrotron Radiation Lightsource (SSRL) beamline 14–1 (Supplementary Table 6).

Crystallographic data processing and refinement

Neutron data were integrated using the MANTID software package^88,89,90 and wavelength-normalized and scaled with LAUENORM from the Daresbury Laue Software Suite⁹¹. X-ray diffraction data were processed using HKL-3000⁹². Refinements of both neutron and X-ray models were completed separately with PHENIX.REFINE from the PHENIX suite⁹³. The refinements were intentionally performed separately due to the known perturbations that X-rays have on the solvent structure, metal redox state, and metal coordination^34,94. The X-ray model was first refined against its corresponding data set and subsequently used as the starting model for neutron refinement. Torsional backbone angle restraints were derived from the X-ray model and applied to neutron refinement using a geometric target function with PHENIX.REFINE⁹³. Mn-ligand restraints for neutron refinement were derived from DFT calculations rather than the X-ray model to remove any influence of photoreduction. The neutron refinement was performed by modeling the D atoms of the active site last to limit phase bias. Initial rounds of refinement to fit protein structure included only non-exchangeable D atoms, which have stereochemical predictable positions. Afterward, H/D atoms were modeled onto the position of each amide proton, and occupancy was refined. In general, the asymmetric units of the neutron crystal structures had a deuterium content of ~85% for the amide backbone, and areas with low deuterium exchange (<50 %) coincided with the presence of hydrogen bonds forming a secondary structure. Next, exchangeable proton positions of residues outside the active site (e.g., hydroxyl group of serine/tyrosine) were manually inspected for obvious positive omit |F_o | - |F_c| neutron scattering length density at a contour of 2.5σ or greater and modeled as a fully occupied deuterium. If the density was not obvious, and there was no chemically sensible reason for the residue to be deprotonated (which is the case for residues outside the active site), the proton position was H/D occupancy refined. D₂O molecules outside the active site were then modeled and adjusted according to the nuclear density. Last, D atoms of the active site were modeled manually. At the active site, a residue is considered deprotonated when (1) attempts to model and refine a proton result in negative |F_o| - |F_c| difference neutron scattering length density, (2) all the other protons of the residue can be placed, and (3) the heavy atom that is deprotonated acts as a hydrogen-bond acceptor. As chemically ideal covalent bond distances of D atoms were ensured during model building and refinement, small deviations from D atom positions and omit |F_o | - |F_c| neutron scattering length density centers were expected from the data resolution (2.3–2.5 Å). To test if the active site ligands for chain A and chain B were truly different we performed a refinement forcing chain B to be identical to chain A. This resulted in difference density that supported our original structural interpretation of the D₂O₂-soaked structure (Supplementary Figs. 1a and 1b). Chain B is different than chain A as shown.

X-ray absorption spectroscopy measurements

A solution of 3 mM MnSOD (~70 mg ml^-1) in 25 mM potassium phosphate pH 7.8 was treated with 8 mM potassium dichromate to isolate the Mn³⁺SOD resting state, 200 mM sodium dithionite to isolate the Mn²⁺SOD resting state, and either 20 mM superoxide or 280 mM (1% w/v) hydrogen peroxide for the product-inhibited state. For EXAFS spectra the sample included 30% ethylene glycol to avoid water crystals. Samples were mounted in 150 µl mylar cells. Superoxide stock solutions were generated by mixing 1.3 M potassium superoxide (KO₂, Sigma Aldridge) in a mixture of dry DMSO (Thermo Scientific) and 0.30 M di-benzo-18-crown-6-ether (Thermo Scientific) following published protocols^{59,60,62,95,96}. The concentration of superoxide in the DMSO/18-crown solution was measured by a UV-Vis spectrophotometer at a wavelength of 250 nm with a molar extinction coefficient of 2686 M^-1 cm^{-1 62,97}. Superoxide solutions were stored at 4 °C under a layer of argon in glass vials sealed with rubber septum.

Mn K-edge HERFD-XANES spectra were recorded at beamline 15-2 of the Stanford Synchrotron Radiation Lightsource (SSRL), while Mn K-edge EXAFS spectra were collected at beamline 7-3. At both beamlines, data were collected at 10 K using a liquid He cryostat, and the incident energy was tuned to the first derivative of an internal Mn foil at 6539 eV. X-ray irradiation was carefully monitored so that two subsequent scans of the same spot did not have photoreduction differences, and different spots along samples were scanned. When appropriate, aluminum foil was inserted into the beam path to attenuate the incident flux. For HERFD-XANES measurements, a Johann-type hard X-ray spectrometer with six Ge(333) analyzer crystals was used with a liquid-nitrogen cooled Si(311) double crystal monochromator, and energy was calibrated to a glitch with measurement of Mn Foil. For EXAFS, measurements were recorded with a 30-element Ge solid-state detector, and a Si(220) monochromator at Φ = 90° was used.

X-ray absorption spectroscopy data analysis

EXAFS data reduction, averaging, and refinement were carried out with the LARCH software package⁹⁸. Refinement of the k²χ(k) EXAFS data used phases and amplitudes obtained from, FEFF⁹⁹. For each fit, optimization of the radial distribution around the absorbing Mn ion (r) and the Debye-Waller factor (σ²) was performed. The goodness-of-it was evaluated by χ² values, reduced χ² values and R-factors.

Computational methods

All DFT calculations were performed with the ORCA quantum chemistry package version 5.0 using the B3LYP functional, the def2-TZVP basis set for all atoms, and the CPCM solvation model^{100,101,102,103}. For geometry optimizations, the full active site (i.e., residues shown in the right panel of Fig. 1a) was included from the neutron structure where the peptide backbone was truncated and C^α was fixed. All atoms for residues Trp161, Phe66, and Tyr166 were fixed with the exception of hydroxyl Tyr166 proton to mimic the packing found in the native enzyme. The Mn ion used the high-spin quintet and sextet states for trivalent and divalent systems, respectively, per experimental observations⁶⁷. A dense integration grid and tight convergence were enforced. For CPCM calculations, the dielectric constant was 80.4. The atomic radii were 2.04 Å for carbon, 1.86 Å for nitrogen, 1.32 Å for hydrogen, 1.82 Å for oxygen, and 2.40 Å for manganese. The solvent probe radius was 1.3 Å.

For TD-DFT calculations, the Mn ion was instead assigned the core property basis set, CP(PPP)^104,105. The geometry-optimized model was truncated to only the Mn ion and its immediate ligands. Including all active site residues for TD-DFT did not significantly alter the simulated spectra. Computed Mn K pre-edge data were plotted using a Gaussian broadening of 1 eV, and a 32.3 eV energy correction was applied in line with previous studies^64,106.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.