key: cord-0823573-kk7rajr5
authors: Senft, Laura; Moore, Jamonica L.; Franke, Alicja; Fisher, Katherine R.; Scheitler, Andreas; Zahl, Achim; Puchta, Ralph; Fehn, Dominik; Ison, Sidney; Sader, Safaa; Ivanović-Burmazović, Ivana; Goldsmith, Christian R.
title: Quinol-containing ligands enable high superoxide dismutase activity by modulating coordination number, charge, oxidation states and stability of manganese complexes throughout redox cycling
date: 2021-07-13
journal: Chemical science
DOI: 10.1039/d1sc02465e
sha: 6f2ead392fb9b3d2d98a648e3d68a0b312584864
doc_id: 823573
cord_uid: kk7rajr5

Reactivity assays previously suggested that two quinol-containing MRI contrast agent sensors for H(2)O(2), [Mn(H2qp1)(MeCN)](2+) and [Mn(H4qp2)Br(2)], could also catalytically degrade superoxide. Subsequently, [Zn(H2qp1)(OTf)](+) was found to use the redox activity of the H2qp1 ligand to catalyze the conversion of O(2)˙(−) to O(2) and H(2)O(2), raising the possibility that the organic ligand, rather than the metal, could serve as the redox partner for O(2)˙(−) in the manganese chemistry. Here, we use stopped-flow kinetics and cryospray-ionization mass spectrometry (CSI-MS) analysis of the direct reactions between the manganese-containing contrast agents and O(2)˙(−) to confirm the activity and elucidate the catalytic mechanism. The obtained data are consistent with the operation of multiple parallel catalytic cycles, with both the quinol groups and manganese cycling through different oxidation states during the reactions with superoxide. The choice of ligand impacts the overall charges of the intermediates and allows us to visualize complementary sets of intermediates within the catalytic cycles using CSI-MS. With the diquinolic H4qp2, we detect Mn(iii)-superoxo intermediates with both reduced and oxidized forms of the ligand, a Mn(iii)-hydroperoxo compound, and what is formally a Mn(iv)-oxo species with the monoquinolate/mono-para-quinone form of H4qp2. With the monoquinolic H2qp1, we observe a Mn(ii)-superoxo ↔ Mn(iii)-peroxo intermediate with the oxidized para-quinone form of the ligand. The observation of these species suggests inner-sphere mechanisms for O(2)˙(−) oxidation and reduction that include both the ligand and manganese as redox partners. The higher positive charges of the complexes with the reduced and oxidized forms of H2qp1 compared to those with related forms of H4qp2 result in higher catalytic activity (k(cat) ∼ 10(8) M(−1) s(−1) at pH 7.4) that rivals those of the most active superoxide dismutase (SOD) mimics. The manganese complex with H2qp1 is markedly more stable in water than other highly active non-porphyrin-based and even some Mn(ii) porphyrin-based SOD mimics.

High concentrations of reactive oxygen species (ROS), such as hydrogen peroxide (H 2 O 2 ) and superoxide (O 2 c À ), are known to oxidatively degrade biomolecules, and their over-accumulation is believed to be involved in the pathology of several severe health conditions, including COVID-19. [1] [2] [3] [4] [5] [6] [7] [8] [9] The potential negative impacts of ROS overproduction on human health have motivated extensive efforts to develop methods to detect and detoxify ROS in biological environments. [10] [11] [12] [13] One attractive antioxidant design strategy is to synthesize small molecules that resemble the enzymes that the body itself uses to regulate ROS concentrations. A small dose of such an antioxidant would alleviate oxidative stress by catalytically degrading ROS. Investigated antioxidants include functional mimics of superoxide dismutases (SODs), which are manganese-, iron-, nickel-, or copper-zinc-containing enzymes that catalyze the degradation of O 2 c À to O 2 and H 2 O 2 (eqn (1)). [14] [15] [16] [17] [18] [19] [20] [21] The enzymatic catalysis relies upon the metal center cycling between two oxidation states; there are no known SODs that instead use an organic redox partner for the disproportionation of O 2 c À .

Previously, our labs developed and characterized three waterand air-stable Mn(II) complexes with redox-active organic ligands (Scheme 1) that respond to H 2 O 2 with changes in their T 1 -weighted relaxivity (r 1 ). [22] [23] [24] Consequently, the Mn(II) complexes can serve as redox-responsive contrast agents for magnetic resonance imaging (MRI) . Notably, the MRI responses result from the oxidation of the organic component, rather than the metal, distinguishing these from redox-responsive probes developed by the Caravan group. [25] [26] [27] [28] [29] The rst sensor, [Mn(Hptp1)(MeCN)] 2+ (1), reacts with H 2 O 2 to form a binuclear Mn(II) species; the reaction with H 2 O 2 oxidatively couples the para-methyl phenol groups from two equiv. of the pre-activated contrast agent. 22 The other two compounds, [Mn(H 2 -qp1)(MeCN)] 2+ (2, MeCN ¼ acetonitrile) and [Mn(H 4 qp2)Br 2 ] (3), feature quinol subunits which are oxidized to para-quinone groups upon their reactions with H 2 O 2 (Scheme 2). 23, 24 In the reduced form of the ligands, the quinols can deprotonate and bind tightly to Mn(II) ions as quinolates; in the oxidized form, conversely, the neutral para-quinones have much weaker metal binding affinities and appear to be displaced by water molecules. 24 The better aquation of the Mn(II) ion is proposed to increase r 1 and thereby enhance MRI contrast.

Curiously, prolonged exposure of 2 and 3 to a large excess of H 2 O 2 converts only 70% of the quinols to para-quinones. 23, 24 The existence of a possible equilibrium mixture of the reduced and oxidized forms of the ligands in the presence of excess ROS led us to speculate that the ligands were cycling between these different oxidation states; this could potentially allow 2 and 3 to catalyze ROS degradation with or without a classical metalcentered redox pathway. We initially screened the antioxidant activity of the three manganese complexes using the xanthine oxidase/hypoxanthine/lucigenin assay. 30, 31 Our results suggested that 1, 2, and 3 all behave as SOD mimics. 23, 24 The SOD activity of 1 was then conrmed using more quantitatively accurate stopped-ow kinetics measurements. 32 Comprehensive study of 1 revealed that in contrast to the majority of nonporphyrin-based manganese-containing SOD mimics, its SOD activity mainly proceeds through an outer-sphere pathway, with additional superoxide and oxygen reductase activity operating through an inner-sphere mechanism involving a Mn(IV)-oxo intermediate and the generation of OHc radicals. 32 As was observed with H 2 O 2 , these reactive species oxidatively couple the phenol groups of separate molecules to dimerize 1. The dimerization reaction has little impact on the SOD activity and essentially protects the catalyst from inadvertently generated ROS. Much like most other non-porphyrin-based Mn SOD mimics, [14] [15] [16] [33] [34] [35] [36] [37] 1 is more active under more acidic and phosphate-free conditions.

More recently, we found that a Zn(II) complex with the H 2 qp1 ligand, [Zn(H 2 qp1)(OTf)](OTf) (4), efficiently catalyzes the degradation of O 2 c À . 38 These results demonstrate that the quinol in the H 2 qp1 ligand is, by itself, a viable redox partner for superoxide dismutation. The free H 2 qp1 ligand, however, cannot act as a SOD mimic, demonstrating that a metal ion is essential for the observed catalysis.

Complexes 2 and 3 contain manganese centers in addition to quinol moieties, and catalytic superoxide degradation could involve either one or both of these as redox partners. It is also possible that the two redox-active components may act synergistically to improve either the rate of the reactions with O 2 c À or the stability of the catalyst; the latter effect was observed with 1, which channels other ROS towards the functionally innocuous dimerization reaction. 32 In order to determine the roles that the metal and ligand have in catalysis, we have investigated the (de) protonation equilibria, water exchange processes, and solution stability of 2 and have used stopped-ow kinetics techniques to conrm and quantitatively assess the abilities of 2 and 3 to act as SOD mimics. We have also identied several catalytically relevant intermediates by cryo-ionization mass spectrometry (CSI-MS). We observe that the ligand inuences the catalysis through (A) its ability to access multiple oxidation states, (B) its charge, and (C) its ability to control the accessibility and rate of ligand exchange at the metal center through steric and conformational effects. Despite the high structural similarity of 1 to 2, we nd that 2 and 3 both react with O 2 c À through more efficient inner-sphere mechanisms, rather than the predominantly outer-sphere pathway observed for 1. 32 Our results have allowed us to identify some of the key factors that modulate SOD catalysis by manganese complexes with this class of redoxactive ligand.

All chemicals and solvents were purchased from Sigma-Aldrich and used as received unless otherwise noted. All deuterated solvents were bought from Cambridge Isotopes. Diethyl ether (ether) and methanol (MeOH) were bought from Fisher. Methylene chloride (CH 2 Cl 2 ) was purchased from Mallinckrodt Baker. N- (2) , and [Mn(H 4 qp2)Br 2 ] (3) were prepared through previously described procedures. 23, 24 Instrumentation All 1 H and 13 C NMR spectra were recorded on a 400 MHz or 600 MHz AV Bruker NMR spectrometer at 293 K. In each spectrum, the reported NMR resonance peak frequencies were referenced to internal standards. 17 O NMR data were collected on a Bruker AVANCE DRX 400WB spectrometer with a superconducting wide-bore magnet operating at a 54.24 MHz resonance frequency and a 9.4 T magnetic induction. A Varian Cary 50 spectrophotometer was used to collect optical data, which were then processed using soware from the WinUV Analysis Suite.

The speciation chemistry of 2 in water was assessed using a METROHM 765 Dosimat with a jacketed, airtight glass titration vessel. A Fisher Scientic Accumet Research AR15 pH meter was used to determine the pH of the sample solutions during the titrations. The electrode was calibrated before each titration using commercially available standard solutions buffered to pH 4.0, 7.0, and 10.0. All samples were purged with argon prior to analysis and subsequently analyzed under an argon atmosphere at 25 C. All solution samples were prepared in solutions of 100 mM KCl in deionized Millipore water. The titrations investigating metal-ligand speciation were run with solutions that contained a 1 : 1 molar mixture of the ligand and MnCl 2 $(4H 2 O). Carbonate-free solutions of 0.10 M KOH and 0.10 M HCl were prepared using argon-saturated deionized Millipore water. These data were analyzed and t to speciation models using the Hyperquad2006 program. 39 17 O NMR measurements of aquation numbers (q) and water exchange processes 17 O NMR spectra were recorded on a Bruker AVANCE DRX 400WB spectrometer equipped with a spectrospin superconducting widebore magnet operating at a resonance frequency of 54.24 MHz at a magnetic induction of 9.4 T. The measurements at atmospheric pressure were performed with a commercial 5 mm Bruker broad band probe thermostated with a Bruker B-VT 3000 variable temperature unit. Relaxation rates were measured for both the paramagnetic solutions and metal-free aqueous buffer solutions. The line widths at half the maximum height of the signal were determined by a deconvolution procedure on the real part of the Fourier transformed spectra with a Lorentzian shape function in the data analysis module of Bruker Topspin 1.3 soware. Pressure-dependent measurements were done with a custom-made thermostated high-pressure probe. The sample was measured in a standard 5 mm NMR tube cut to a length of 50 mm. To enable pressure transmittance to the solution, the NMR tube was closed with a moveable Macor piston. The advantage of this method is that oxygen-sensitive samples can be easily placed in the NMR tube and sealed with the Macor piston under an argon atmosphere. The pressure was applied to the high-pressure probe via a per-uorinated hydrocarbon (hexauoropropylene oxide, Hostinert 175, Hoechst) and measured by a VDO gauge with an accuracy of AE1%. The temperature was adjusted with circulating, thermostated water (Colora thermostat WK 16) to AE0.1 K of the desired value and monitored before each measurement with an internal Pt-resistance thermometer with an accuracy of AE0.2 K. Enriched (10%) 17 O-labeled water (D-Chem Ltd Tel Aviv, Israel) was used for the 17 40 Relaxation rates were measured both for aqueous solutions containing 2 and for metal-free solutions buffered to pH 4.0 and 7.4.

Cryospray-ionization mass spectrometry (CSI-MS) measurements were performed on a UHR-TOF Bruker Daltonik maXis plus, an ESI-quadrupole time-of-ight (qToF) mass spectrometer capable of a resolution of at least 60.000 (FWHM), which was coupled to a Bruker Daltonik Cryospray unit. Detection was in positive ion mode with a source voltage of 3.5 kV and a ow rate of 240 mL per hour. The temperatures of the spray gas (N 2 ) and the dry gas used for solvent removal were maintained at À40 C and À35 C, respectively. The mass spectrometer was calibrated prior to every experiment via direct infusion of an Agilent ESI-TOF low concentration tuning mixture, which provided a m/z range of singly charged peaks up to 2700 Da in both ion modes. For the reactions with O 2 c À , 1 mM solutions of either 2 or 3 in MeCN (1% DMF) were cooled to À40 C and mixed with excess solid KO 2 . The mixture was diluted to roughly 1 Â 10 À5 M and quickly injected into the mass spectrometer. Applied solvents were not extra dry in order to provide a source of protons. The measured data were processed and analyzed with Bruker Data Analysis. The complete CSI-MS spectra recorded for the reactions of 2 and 3 with excess superoxide are shown in Fig. S1 and S2, respectively, in ESI. †

We previously assessed the abilities of 2 and 3 to catalytically degrade superoxide using the xanthine oxidase/hypoxanthine/ lucigenin assay. 23, 24, 31 Each assay was carried out in a total volume of 1 mL containing 50 mM Tris (pH 8.0), 50 mM hypoxanthine, 0.005 U mL À1 xanthine oxidase (Calbiochem), and 5 mM dark adapted lucigenin in the presence of either 0.1 nM or 10 mM of the Mn(II) complex or its vehicle.

The ability of the Mn(II) complexes to catalytically degrade superoxide was tested by a direct method using stopped-ow techniques as described elsewhere. 34 Experiments were carried out using syringes 1, 2, and 3 on a Biologic SFM-400 instrument that was equipped with an Energetiq LDLS ENQ EQ-99-FC laser driven light source and a J&M TIDAS diode array detector (integration time ¼ 0.5 ms, l ¼ 180-724 nm). The source of superoxide was commercially available KO 2 dissolved in dry and non-buffered DMSO ([O 2 c À ] z 1-2 mM). Complexes 2 and 3 were each tested at four different concentrations between 0.9 and 9 mM in aqueous solutions buffered with HEPES or sodium phosphate to either pH 7.4 or pH 8.1. The ionic strength of all solutions was 111 mM. The aqueous solution containing the studied Mn(II) complex was mixed in a 9 : 1 ratio with the superoxide solution in DMSO using a high-density mixer. In each experiment, the concentration of superoxide exceeded that of the metal-containing catalyst by at least ten-fold to ensure catalytic conditions. Millipore water was used for the preparation of the buffer solutions. All of the prepared buffers were treated with Chelex 100 sodium exchange resin for at least 12 h before use in order to remove adventitious metal ions. The data analysis was performed using the BioKine V4.66 soware. Each reported k obs value is the average of at least 10 measurements. The reported k cat values were determined from the slope of the k obs vs.

[SODm] plot. The presence of H 2 O 2 was qualitatively conrmed using Baker Testrips for peroxides.

Cryo UV/vis experiments were performed with the use of low temperature stopped-ow techniques involving a biologic cryo SFM-4000 four-syringe stopped-ow combined with a 150 W xenon lamp and a J&M TIDAS diode array detector (200-724 nm) with an integration time of 0.5 ms. A Huber C-905 cryostat lled with silicon oil was used to cool the samples. Spect, Global Analysis System soware (version 3.0.38 for 32 bit Windows system) was used to analyze the spectral changes and perform a global analysis of kinetic traces at various wavelengths.

EPR spectra were recorded on a JEOL continuous wave spectrometer JES-FA200, equipped with an X-band Gunn diode oscillator bridge, a cylindric mode cavity, and a helium cryostat. The samples were measured in the solid state in quartz glass EPR tubes at 95 K and 7 K. The spectra shown were measured using the following parameters: microwave frequency ¼ 8.959 GHz, modulation width ¼ 1.0, and 0.5 mT, microwave power ¼ 1.0 mW, modulation frequency ¼ 100 kHz, time constant ¼ 0.1 s.

All calculations were done by applying the B3LYP functional [41] [42] [43] in combination with the def2svp basis set. [44] [45] [46] The characterization as minima or transition states was done by computation of vibrational frequencies at the same level. All wavefunctions were tested successfully for stability. [47] [48] [49] Relative energies were corrected for zero-point vibrational energies (ZPE). All calculations were performed using the Gaussian 16 program package. 50

Before one can identify the factors that govern the ability of manganese complexes with polydentate ligands (Scheme 1) to catalytically decompose superoxide, one needs to understand how they behave in water. We have therefore characterized the pH-dependent speciation, stability, and water exchange kinetics of [Mn(H 2 qp1)(MeCN)](OTf) 2 (2) in aqueous solutions using potentiometric and spectrophotometric pH titrations and variable temperature and pressure 17 O NMR spectroscopy. We previously reported analogous studies for [Mn(Hptp1)(MeCN)] 2+ (1) and [Mn(H 4 qp2)Br 2 ] (3). 24, 32 The new data enable us to fully compare the aqueous chemistries of 1, 2, and 3. Further, we determined catalytic rate constants for superoxide dismutation by 2 and 3 using the same techniques that were used for 1, again enabling the three to be compared. 32 Mechanistic insights were gained by characterizing reactive intermediates using lowtemperature UV/vis and EPR spectroscopies and CSI-MS.

Potentiometric and spectrophotometric pH titration data were collected to assess the stability and ligand protonation states of 2 in water. Previously, we found that the H 2 qp1 ligand itself undergoes two ionization events consistent with pK L values of 7.24 and 4.72 which were assigned to the (de)protonation of one of the tertiary amines (H 3 qp1 + ) and one of the pyridine rings (H 4 qp1 2+ ), respectively. 38 We observed precipitation upon raising the pH above 10; consequently, we were unable to measure reliable pK a values for the quinolic O-H groups of H 2 qp1, which likely fully deprotonate at pH values higher than 10. Similar ligand instability above pH 9 was observed for H 4 qp2. 24 By comparison, complex 2 displays ionization events with pK a values of 6.3 and 3.1 (Fig. S3 , S4, † and Table 1 ). The pK a value of 6.3 for the Mn(II) complex is consistent with the (de) protonation of a M(II)-bound phenolate. 26, 51 The assignment of the 6.3 pK a value to the (de)protonation of a metal-bound quinol is supported by spectrophotometric pH titration data. At pH 6.0 and above, there is a strong peak at 305 nm; the energy of this band is consistent with an intraligand electronic transition for a phenolate, rather than a phenol ( Fig. S5 †) . 24, 26 As the pH drops below 6, this band disappears and is replaced by a prominent shoulder at 295 nm, an energy which is more consistent with the conjugate acid phenol. At physiologically relevant pH 7.4, 93% of the complex has a [Mn(Hqp1)] + core, with the other 7% existing as the dicationic [Mn(H 2 qp1)] 2+ (Fig. 1 ). Additional aquation number (q) measurements were performed to fully describe the water speciation (vide infra).

Analogous studies for 1, which features a para-methyl phenol group in place of the quinol, revealed that 1 is slightly more acidic, and the metal-bound phenol has a pK a value of 6.16. 32 Consequently, the conjugate base phenolate should be a slightly weaker base and electron pair donor than the quinolate in 2. To our initial surprise, 2 is much more stable in water than 1 as evidenced by their respective pMn 52 values of 7.25 and 5.40 (Table 1) . 32 A priori, the hydroxyl group para to the coordinating O atom in H 2 qp1 would be anticipated to be a better electron donor than the methyl group in Hptp1, 53 possibly explaining why H 2 qp1 coordinates more tightly to Mn(II) than Hptp1. We recently found that N,N 0 -bis(2,5-dihydroxybenzyl)ethanediamine-N,N 0 -diacetic acid formed a more stable Mn(II) complex than N,N 0 -bis(2-hydroxybenzyl)ethylenediamine-N,N 0 -diacetic acid, providing another example of a quinol-for-phenol substitution improving complex stability. 54 That quinols/quinolates are indeed better ligands than phenols/phenolates is supported by previously obtained crystallographic data. In the structures of 1 and 2 obtained from MeCN, Hptp1 and H 2 qp1 remain protonated and coordinate to the Mn(II) centers in a hexadentate fashion. 22, 23 The Mn-OH(quinol) bond (2 22, 23 demonstrating that quinol binds to Mn(II) more tightly than para-methyl phenol. At pH 7.4, the H 2 qp1 ligand is also found to coordinate Mn(II) more tightly than H 4 qp2. The pMn value of 3 was determined to be 5.36. 24 Although over 99% of both the H 2 qp1 and H 4 qp2 ligands are bound to the Mn(II) above pH 7.0 ( Fig. 1) , 24 H 4 qp2 protonates and dissociates from the metal ion to a greater extent than H 2 qp1 under acidic conditions. Less than 10% of the Mn(II) is ligand-free at pH 3 with 1.0 mM amounts of metal and H 2 qp1; conversely, over 80% of the Mn(II) dissociates from H 4 qp2 under the same conditions. 24 The pK a values of the rst and the second Mn(II)-quinol groups in 3 were found to be 5.82 and 7.14, respectively, resulting in 66% of species with a doubly deprotonated [Mn(H 2 qp2)] core and 33% of species with a singly deprotonated [Mn(H 3 qp2)] + core at pH 7.4. 24 The crystallographic data that we have obtained thus far with the H 2 qp1 and H 4 qp2 ligands suggest that protonated quinols bind less avidly to metal ions than pyridine rings. 23, 24, 38 Whenever these ligands have not been fully coordinated to the metal centers in the solid-state structures, it is always the quinol groups that are not ligated. The stronger metal-binding affinity of H 2 qp1 relative to H 4 qp2 is therefore likely a consequence of replacing one of the H 4 qp2 quinols with a pyridine. The singly deprotonated H 3 qp2 À thereby acts more like a pentadentate ligand at physiological pH; Hqp1 À , conversely, behaves more fully like a hexadentate ligand.

The aquation number (q) and rate of water exchange at 298 K (k 298 ex ) for 2 were determined from variable temperature 17 O NMR data at pH 4.0 and 7.4 using a protocol developed by Gale et al. (Fig. 2) . 40 The aquation of 2 is not as extensive as it is for either 1 or 3 ( Table 2 ). The simplest interpretation of the 0.7 value of q at pH 7.4 for 2 suggests that approximately 70% of the Mn(II) binds a single water molecule, with the other 30% without coordinated water molecules. The overall composition of the major component is therefore proposed to be [Mn(Hqp1)(OH 2 )] + . The pH 7.4 (Table 2) . 24, 32 That the aquation number for 2 is lower than it is for 1 and 3 under similar conditions may suggest that the monoaqua species for 2 is thermodynamically favored to a lesser degree than the analogous species for the Hptp1 and H 4 qp2 ligands. Between pH 7.0 and 7.4, H 2 qp1 appears to be more strongly electron-donating than the other two ligands, as evident from its stronger chelating ability, vide supra. The more electron-rich metal center in 2 may thereby have a weaker affinity for the additional aqua ligand. That less of the Mn(II) in 2 is heptacoordinate suggests that water exchange proceeds through a dissociative mechanism (D), which is supported by the highly positive activation volume of +10.9 (AE0.6) cm 3 mol À1 as well as a positive activation entropy (Fig. S6 , S7, † and Table 2 ). These temperature-and pressuredependent 17 O NMR data conrm that Hqp1 À labilizes the coordinated water molecule. They also explain the co-existence of the seven-coordinate [Mn(Hqp1)(OH 2 )] + and the sixcoordinate [Mn(Hqp1)] + species at pH 7.4, with the latter being the intermediate for the dissociative water exchange mechanism. DFT calculations (B3LYP/def2svp; Fig. 3 ) support the operation of a D mechanism and predict that the energy difference between [Mn(Hqp1)(OH 2 )] + and [Mn(Hqp1)] + is only 2.2 kJ mol À1 , with the seven-coordinate structure being slightly more stable. These calculations also predict a low activation barrier for ligand exchange (8.6 kJ mol À1 , Fig. 3 also contribute to the slower rate of water exchange at the lower pH. The faster rate and a dissociative character of the exchange mechanism under more basic conditions suggest that the coordination of quinolate groups increases the kinetic lability of the Mn(II) center. The higher ligand substitution reactivity of 2 at pH 7.4, may be benecial for catalytic activity involving an inner-sphere mechanism.

The SOD activities of complexes 2 and 3 were initially assessed using an established procedure that uses xanthine oxidase and hypoxanthine to produce O 2 c À and a subsequent reaction with lucigenin to detect it. 23, 24, 30, 31 The concentrations at which half of the sensing reaction was blocked (IC 50 ) were found to be 11.3 nM (2) and 18.2 nM (3). Based on the IC 50 values and the 7.7 Â 10 5 M À1 s À1 rate constant reported for the reaction between O 2 c À and lucigenin, 55 one would estimate k cat values of 3.4 Â 10 8 M À1 s À1 (2) and 2.1 Â 10 8 M À1 s À1 (3) from these data.

Side reactions between the various components of the assays have been hypothesized to inuence the measured IC 50 , and these assays frequently provide inaccurate accounts of antioxidant activity. 17, 34, [56] [57] [58] [59] In order to obtain the best possible assessment of the SOD mimicry of 2 and 3, we performed an extensive kinetic analysis of the reactions between O 2 c À and the complexes in aqueous solutions buffered to pH 7.4 and 8.1 (Table 3) . We recently reported similar studies for complexes 1 and 4, the latter being the Zn(II) complex [Zn(H 2 qp1)(OTf)] + . 32, 38 The H 2 qp1 ligand by itself was previously tested and found to be inactive as a catalyst for the decomposition of O 2 c À at either pH 7.4 or 8.1. 38 The degradation of superoxide is second-order in O 2 c À in the absence of added catalyst, conrming that treatment of the solutions through a Chelex exchange matrix successfully removed all potentially catalytic metal impurities. The decomposition of O 2 c À in solutions containing either 2 or 3, conversely, were found to be rst-order with respect to both O 2 c À and the manganese catalyst. The k cat values for 2 and 3 are both within the ranges seen for manganese-containing SOD mimics, with the most efficient of the complexes having k cat values ranging from 10 7 to 10 8 M À1 s À1 . 15, 18, 19, 30, 32, 33, 35, [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] Complex 2 is especially effective in HEPES buffer, with a catalytic activity (log k cat ¼ 7.99) that rivals those of cationic porphyrin complexes. 18, 19 One particularly effective porphyrinbased SOD mimic is Mn II Br 8 TM-4-PyP 4+ , which has both a highly positive charge and a large catalytic rate constant (log k cat ¼ 8.34, determined by the cytochrome c assay) that approaches the 8.84-9.30 values measured for SODs. [71] [72] [73] However, Mn II Br 8 TM-4-PyP 4+ is unstable at neutral pH and releases free manganese; approximately 30% of the Mn is estimated to be lost. 74 The decomposition of the complex has a substantial impact on the reactivity; if the Mn II Br 8 TM-4-PyP 4+ catalyst were fully intact, its estimated log k cat value would exceed 8.67. 74 The release of neurotoxic Mn(II) ions would severely limit the ability of this and similarly unstable complexes to serve as pharmaceuticals. 71 Complex 2 also compares well to manganese complexes with pentaazamacrocyclic ligands, which comprise the other major class of highly active SOD mimetics. Arguably the best of these mimics is M40401, which has a catalytic rate of approximately 1.5 Â 10 9 M À1 s À1 . 68, 69 As with Mn II Br 8 TM-4-PyP 4+ , M40401 is markedly less stable in water than 2, with a pMn ¼ 4.7 versus the 7.25 value measured for 2. 75 Although complex 2 is about an order of magnitude less active than M40401, its much greater aqueous stability may allow it to engage in longer-lived catalysis.

Much like most other manganese-containing SOD mimics, complex 2 displays lesser activity at pH 8.1 and in phosphate buffer. [14] [15] [16] [32] [33] [34] [35] [36] [37] The inhibition of the SOD activity by phosphate through its competitive binding to the metal center 34 is usually indicative of an inner-sphere pathway for superoxide decomposition. 14 The activity of complex 3, conversely, is less sensitive to the composition of the solution; despite this, the data suggest that 3 likewise reacts with superoxide through inner-sphere pathways. The second quinol in 3 does not enhance the catalysis. To the contrary, 2 is a superior catalyst to 3 in two of the investigated aqueous solutions and is approximately equivalent to 3 in phosphate buffer. The second quinol in 3 partly deprotonates to a quinolate, which reduces the overall charge of the complex in water. Complex 24 The lesser average positive charge on the H 4 qp2 complexes is anticipated to weaken their abilities to attract and bind anions, both slowing the rate of superoxide complexation and decomposition and attenuating the inhibitory effect of phosphate. The lessened positive charge for 3 also accounts for its slower rate of water exchange which unlike 2, appears to proceed through an associative mechanism on the basis of its highly negative entropy of activation. 24 These results and the intermediate species characterized within the catalytic cycles (vide infra) are consistent with both 2 and 3 degrading O 2 c À through inner-sphere pathways.

Despite its strong structural similarity to 2, complex 1 exhibits much lower SOD activity than either 2 or 3 and catalytically degrades O 2 c À through mostly outer-sphere pathways. 32 These differences can be rationalized by considering the Mn(III/ II) reduction potentials. Engagement of the coordinated water molecule in proton coupled electron transfer (PCET) lowers the Mn(III/II) reduction potential (466 mV vs. Ag/AgCl) 32 and allows an outer-sphere reduction of O 2 c À , yielding a Mn(III)-OH species that can then oxidize superoxide to O 2 and regenerate the Mn(II)-OH 2 complex. The calculated structure of the Mn(III) form of 1, [Mn III (ptp1)(OH)] + , features a strong bond between the OH À and the metal center. Although the Mn(III) center in [Mn III (ptp1)(OH)] + is predicted to be pentacoordinate with three neutral N-donors and an anionic O-donor from the ptp1 À ligand and a second anionic O-donor from the hydroxide, its distorted square pyramidal structure, the diamine backbone of the ligand, and the negatively charged phenolate and hydroxide groups block access to the only accessible coordination site for potential sixth ligands. 32 Therefore, our calculations suggest that the Mn(III) form of 1 is inaccessible to O 2 c À and cannot oxidize it through an inner-sphere mechanism. Conversely, the Mn(III/II) reduction potentials of 2 (645 mV vs. Ag/AgCl) 23 and 3 (725 mV vs. Ag/AgCl) 24 are both too high to support efficient reduction of O 2 c À through an outer-sphere catalytic pathway; 2 and 3 are instead proposed to catalyze superoxide degradation through inner-sphere pathways that involve the oxidation and reduction of the quinol/para-quinone groups (vide infra). Complexes 2 and 3 differ from 1 in that the stabilizations of Table 3 Catalytic rate constants, k cat (M À1 s À1 ), for the reactions of 1, 2, and 3 with superoxide 

The reactions between KO 2 and the manganese complexes were probed by cryo-MS, rapid-scan stopped-ow and cryo UV/vis techniques, and EPR spectroscopy. Based on our results, we propose a detailed common mechanism for complexes 2 and 3 (Scheme 4). We were able to identify most of the reactive and volatile intermediates participating in redox catalysis during the reactions between 2 and 3 with excess superoxide (species a-k in Scheme 4) by analyzing these reactions in MeCN at À40 C with ultra-high resolution cryo mass spectrometry. The initial reaction between the starting Mn(II) species with the ligand in a fully reduced state (a) and O 2 c À oxidizes the metal center to Mn(III) (b, Fig. 4) , with concomitant reduction of O 2 c À to H 2 O 2 . For compound 3, the presence of two negative charges from the two quinolates in the coordination sphere stabilizes the +3 oxidation state of the metal. Since an external anion, such as OH À , is not needed to stabilize the complex, the metal center remains accessible to additional equivalents of O 2 c À , unlike the Mn(III) form of 1. 32 Species b can therefore coordinate O 2 c À (c, Fig. S10 in ESI †) and continue to participate in inner-sphere catalytic cycles (red cycle in Scheme 4). This type of reactivity is well-established for many Mn-containing SOD mimics with redox-innocent polydentate ligands. [15] [16] [17] 30, 61 However, the introduction of redox-active, non-innocent ligands enables more versatile reactivity, introducing pathways involving ligand oxidation and reduction (d and e) in addition to traditional metal-based redox catalysis.

The inclusion of redox-active ligands also appears to allow the generation of a higher-valent Mn(IV)-oxo intermediate (i). The observation of this species would imply that the Mn(III)superoxo species formed with 3 (c) can itself undergo heterolytic O-O bond cleavage, with a two-electron oxidation of the hydroquinone ligand to a para-quinone and loss of water. Similar reactivity was recently reported for an iron porphyrin complex with a pendant quinol group; in this work, an Fe(III)superoxo reacts to form a ferryl species (Fe(IV)-oxo) with an oxidized para-quinone moiety. 76 The reaction goes through an Fe(III)-semiquinone (quinoxyl radical) species as an intermediate. 76 This path (upper orange arrow in Scheme 4) might contribute to our detection of a quinoxyl-radical by the UV/vis spectroscopy (vide infra) during the catalytic degradation of O 2 c À by 3.

In the case of 2, the initial oxidation by O 2 c À appears to generate a Mn(II)-quinoxyl radical species (d) through Scheme 4 Proposed mechanism of catalytic superoxide removal by 2 and 3. Depicted are two quinol moieties of H 4 qp2 within the structure of 3 (the rest of the pyridylamine ligand is omitted for clarity). In the case of 2, the leftmost quinol moiety is replaced by a pyridine ring from the ligand. Within the cycle of 3, the intermediates marked with could not be observed by CSI-MS due to their neutral charge. Note that species presented as Mn-O-O À refer to formally coordinated superoxide and those presented as Mn-O À -O À to formally coordinated peroxide.

tautomerization of Mn(III)-quinolate since one quinolate anion is not sufficient to stabilize the formally +3 oxidation state of the metal center. The quinoxyl radical is a stronger acid than quinolate, which facilitates its deprotonation (d, Fig. 5 ). Thus, 2 probably does not enter the metal-only cycle observed for 3 (red cycle), and its SOD activity relies more on the ligand's ability to undergo redox transformation. Although intermediate d was not directly observed for 3, it too may be capable of forming a Mn(II) intermediate with a quinoxyl radical through either a valence tautomerism or direct oxidation of the quinols. We nd this likely since Mn(II)-para-quinone species (e in Fig. 4 and S11 †) are observed with both 2 and 3; the para-quinone results from further oxidation of the quinoxyl radical in species d. 38 Cycling between a metal-coordinated quinolate, quinoxyl radical, and para-quinone (violet cycle) was the sole possible mechanism for the SOD activity of the Zn(II) complex with H 2 qp1 (4). With 2 and 3, however, the ligand-only redox cycle is not likely to be a major contributor to the overall superoxide degradation since the catalysis exhibited by 4 is much slower (k cat ¼ 3.4 Â 10 6 M À1 s À1 for 4 at pH 7.4 in HEPES) 38 than what was detected for 2 and 3. The Mn(II)-quinone species of 2 and 3 (e) can enter alternative metal-based catalytic cycles for superoxide degradation (blue cycle; for CSI-MS spectra of f/g see Fig. S11 †) . We speculate that these pathways are more efficient than those with the reduced ligands for two reasons. First, the oxidized para-quinone forms of the ligands, qp1 and H 2 qp2, are effectively pentadentate ligands, with either N5 or N4O donor sets respectively. The removal of a chelating arm improves the accessibility of the metal center to O 2 c À in both the Mn(II) and Mn(III) states. The N5 coordination provided by qp1 is stronger than that of the N4O coordination by H 2 qp2, maintaining the higher stability of 2 under catalytic conditions (vide infra) and bolstering its SOD activity. Second, the transformation of the quinolate to a neutral para-quinone increases the overall positive charge of the catalyst, which would more readily attract and bind O 2 c À .

An intermediate that is noticeably unobservable during the catalysis is the Mn(III)-para-quinone species j, which is either tricationic (for 2) or dicationic (for 3). We believe that j rapidly reacts with O 2 c À to yield species k (Fig. S12 †) , which then oxidizes O 2 c À to O 2 and regenerates the Mn(II) species e. The heightened positive charges of j species would attract O 2 c À more efficiently and enable its stronger coordination to the metal center due to the stronger coulombic interactions. The inability to detect intermediates j with cryo-MS would be consistent with the heightened positive charge increasing the rate of reactivity to the extent that these species are too reactive to observe by CSI-MS. The ability of 2 to attain a tricationic form would partially explain its high activity, particularly relative to that of 3. Notably, the most active reported manganese-containing SOD mimic M40401 (based on the catalytic rate constants determined by direct stopped-ow measurements) 68,69 features the neutral N5 pentaazamacrocyclic ligand. M40401 likewise has a +2 or +3 overall charge depending on the manganese redox state during the catalytic cycling and predominantly exists as a pentacoordinate species in solution, supporting our observed correlations between catalyst charge, the accessibility of the metal center for superoxide binding, and the rate of superoxide decomposition.

The hydroperoxo species h from the blue cycle in Scheme 4 can either get protonated to release H 2 O 2 or undergo homolytic cleavage to generate a high-valent Mn(IV)-oxo species i (Fig. 6) and a hydroxyl radical. The ability of the H 4 qp2 to provide a dianionic coordination sphere in its reduced form (H 2 qp2 2À ) and a monoanionic one in its two-electron oxidized form (Hqp2 À ) allows this ligand system to better stabilize the highervalent intermediate than H 2 qp1. We can therefore observe i for 3, but not 2, using CSI-MS. We note that Mn(IV)-oxo intermediates could also form directly from the Mn(III)-superoxo species c (vide supra; orange paths). Although i could not be detected under CSI-MS experimental conditions in the reaction with 2, perpendicular EPR measurements suggest that this species is formed transiently with the H 2 qp1 system (vide infra). The endproducts of the reactions between 2 and O 2 c À provide further, but indirect, evidence for the generation of the OHc by-products anticipated from the formation of Mn(IV)-oxo species. Upon prolonged reaction times, we observe Mn(II) complexes with oxygenated ligand backbones and missing quinol/para-quinone groups by CSI-MS (Fig. 7) . We believe that O 2 c À will normally intercept and react with Mn(IV)-oxo and OHc oxidants quickly enough to prevent them from oxidizing the organic ligands. When the catalysis has depleted the O 2 c À , however, this protective benet is lost, resulting in ligand oxidation. The benzylic carbons that connect the quinols/para-quinones to the rest of the ligand are particularly susceptible to oxidation due to their weak C-H bonds. 38, 77 With 3, the long-term decomposition renders the ligand tetradentate. In water, neutral and acyclic N4 coordination environments cannot out-compete water molecules as ligands; consequently, the end-products of the reactions between 3 and O 2 c À consist of a mixture of organic ligand fragments and no distinct Mn(II) or Mn(III) species. With 2, the loss of a quinol group from H 2 qp1 yields a pentadentate ligand with a N5 coordination environment. This decomposition product can still bind Mn(II) tightly enough to allow us to observe discrete Mn(II) complexes with the degraded ligand by CSI-MS (Fig. 7) . The m/z features of the decomposition products are less prominent for 2 than those for 3; this is consistent with the H 2 qp1 system generating fewer Mn(IV)-oxo and OHc species during catalysis. Importantly, the species with the oxygenated H 2 qp1 ligand could still be catalytically active even without its quinol/para-quinone group (vide supra). Lastly, we also detect a Mn(II) species with an oxygenated benzylic carbon of H 2 qp1

where the quinol is still attached to the framework (Fig. 7B) . The quinol is in its reduced form, which suggests that the ligand indeed cycles between its quinol and para-quinone oxidation states during catalysis (purple pathway) in parallel with the metal redox cycling (red and blue cycles).

Although parallel MS experiments using a 18 O-labeled source of superoxide would be invaluable for corroborating the identities of many of the detected intermediates, such studies were impractical due to the nature of the available resources. We were unable to nd a reliable and economic source of 18 Olabeled KO 2 but were previously able to synthesize this reagent. 78 Regrettably, the synthetic procedure contaminates the product with substantial amounts of basic impurities that have proven to be difficult to remove. These basic impurities are not present in commercially available K 16 O 2 . In prior studies, we found that the buffer solutions used under the required CSI-MS experimental conditions cannot sufficiently mitigate the added basicity. As a consequence of the more basic conditions that arise from the use of the 18 O-labeled oxidant, the data are dominated by metal-OH species which yield m/z features that overlap with and obscure the regions corresponding to the target intermediates. 38 Observation of reaction intermediates by low-temperature UV/ vis and EPR When 0.1 mM 2 is rapidly mixed with an excess of superoxide (0.5-1 mM) in MeCN at À40 C, three bands at 422, 448, and 595 nm appear within 3 s (Fig. 8, green line) . Over the next 30 s, the features at 422 and 448 nm begin to diminish while the broad band at 595 nm continues to intensify (Fig. 8, blue line) . The appearance of an isosbestic point at 474 nm indicates that the observed spectral changes can be ascribed to two distinct species that are formed as intermediates in the reaction between 2 and O 2 c À at low temperature. When the reaction is monitored at À40 C for 1200 s, the 595 nm band slowly disappears as the absorbance below 400 nm increases without the formation of distinct peaks (Fig. S13 in ESI †) . The long-term spectra likely correspond to a complex mixture of products (vide supra).

Although absorption bands above 400 nm may be consistent with higher-valent manganese intermediates formed from the oxidation of the metal center in 2 by O 2 c À , the energies and characteristic shapes of the features at 422 and 448 nm resemble those found in the spectrum of the semiquinone radical anion. 79, 80 When the H 2 qp1 ligand by itself is allowed to react with an excess of O 2 c À in MeCN, we observe similar bands by UV/vis (Fig. S14 †) , conrming that the species observed as the rst intermediate in the low-temperature reaction between 2 and O 2 c À is best interpreted as a Mn(II) complex with a semiquinone radical anion ligand. The UV/vis changes observed at low temperature therefore result from ligand-centered, rather than metal-centered, oxidation. The UV/vis data corroborate the identity of species d (Scheme 4 and Fig. 5) .

The formation of the semiquinone radical anion is accompanied by an additional process that produces the species associated with the broad band centered at 595 nm. This secondary reaction is also observed in the reaction between O 2 c À and the free H 2 qp1 ligand; the peak at 595 nm therefore cannot be ascribed to a low-energy charge transfer band between the metal and the ligand. The most probable explanations for the 595 nm absorbance band involve the occurrence of either intermolecular self-association between semiquinone radicals or an electron transfer equilibrium between semiquinone radical electron donors and para-quinone electron acceptors to form corresponding p-dimers. Such p-association charge transfer equilibria of organic radical ions have been the subject of many investigations. 81, 82 They have been shown to be thermochromic systems which display a strong absorption band at $600 nm at lower temperatures. Unfortunately, these intermolecular self-association processes could not be observed with the extremely dilute solutions and low-pressure conditions associated with the CSI-MS experiments.

Mn(III)-peroxo complexes with amino-pyridine ligands have been documented to have UV/vis bands between 550 and 600 nm. [83] [84] [85] We cannot preclude the possibility that these are partly contributing to the 595 nm band observed for 2. To further investigate the potential contribution of multiple species to the 595 nm band, we compared the kinetic traces for the reaction between KO 2 and the H 2 qp1 ligand and that between KO 2 and complex 2 (Fig. S15 †) . Although the kinetic trace obtained for the ligand reaction can be satisfactorily t to a one-exponential function, the kinetic trace for the reaction with complex 2 cannot, suggesting that more than one process is contributing to the increased absorbance at $600 nm.

The subsequent slow disappearance of the 595 nm band over 10 min at À40 C indicates that further oxidation processes occur. Unfortunately, these processes cannot be well resolved from the UV/vis data since (A) the more intense absorption bands of the oxidized ligands dominate the spectra and (B) the high reactivities of higher-valent manganese intermediates preclude their accumulation to concentrations high enough to be clearly observed by UV/vis spectroscopy under the experimental conditions. In the reactions between 2 and O 2 c À , we observe a small but signicant increase in the absorbance at 700 nm that is not duplicated in the reactions between free H 2 qp1 and O 2 c À (Fig. S16 †) . This may be consistent with the presence of a small amount of a higher-valent manganese species, but the data do not unambiguously support this.

Additional evidence of high-valent manganese species in the reaction cycles of 2 was provided by EPR analysis of freezequenched samples. To prepare the samples, an MeCN solution of 2 was rapidly mixed with an excess of O 2 c À at 233 K and then rapidly cooled to 9 K aer 30 s. The X-band EPR spectrum recorded in frozen MeCN at 9 K in the perpendicular mode displays the characteristic features of the high-spin d 3 electronic conguration anticipated for a monomeric Mn(IV) complex, with a strong signal at g $ 2 and weak broad signals at g $ 4 and g $ 1.23 (Fig. S17 †) . The EPR spectra of most Mn(IV) complexes tend to follow one of two patterns, depending on the magnitude of the zero-eld splitting. When the axial parameter D is much larger than the energy provided by the spectrometer (0.31 cm À1 with X-band), a weak signal is found at g $ 2 and a strong signal is observed at g $ 4. When D is much smaller than the excitation energy, the predominant peak instead occurs at g $ 2 with weaker features at g $ 4; this tends to occur in high symmetry environments. [86] [87] [88] [89] [90] [91] [92] With complex 2, the more intense signal at g $ 2 indicates that the Mn(IV) species has a small zero-eld splitting, with 2jDj ( hn.

The spectrum shown in Fig. S17 † was analyzed and tted using EasySpin in MATLAB®, 93 and the values of the obtained parameters are given in Table S3 . † The linewidth (tted using a Gaussian distribution) is relatively large; most probably due to unresolvable hyperne coupling between the manganese nucleus and nitrogen atoms in the surrounding ligand. The magnitude of the D value was determined to be 0.048 cm À1 (note: the sign of D is unknown), hence conrming that 2jDj ( hn. The E/D is approximately equal to zero (0.05), showing that there is negligible rhombicity in the system. These values suggest that the effective electronic environment of the Mn(IV) closely approximates an ideal octahedral conguration. The N5 core of the two electron-oxidized para-quinone form of the ligand (qp1) and the oxo group would provide such a coordination environment.

It should be noted that it is theoretically possible to model the spectrum for 2 as a Mn(II) species. However, although residual Mn(II) species, superoxide, and/or ligand radicals could contribute to the strong signal at g $ 2, our control experiments analyzing the EPR spectra of MeCN solutions of 2, the semiquinone radical ligand, and O 2 c À (Fig. S18 †) demonstrate that none of these three species can account for the feature at g $ 2 shown in Fig. S17 . † The vast majority of the superoxide has been consumed by the 30 s timepoint. In principle, the broad feature at g $ 2 can also be ascribed to another Mn(II) species generated in the catalytic cycle (Scheme 4). This species should display high symmetry, regardless of whether the spectrum is simulated as a Mn(II) or as a Mn(IV) species, since the magnitude of the E/D ratio is small. The only plausible candidate with these characteristics would be a Mn(II) intermediate bearing a quinone ligand (species e shown in Scheme 4). Although we cannot preclude this possibility, we believe that the UV/vis data taken during the catalysis performed by 2 is indicative of a transient high-valent manganese intermediate, and consequently, we assign the feature at g $ 2 to a Mn(IV) species. Although the zero-eld splitting parameters obtained in the present study are in line with those observed for other Mn(IV) centers, 92 the D parameter is signicantly smaller than those reported for other Mn(IV)-oxo adducts. The D values for other reported Mn(IV)-oxo complexes are usually much larger than the microwave frequency, resulting in a more intense EPR signal situated near g $ 4. A rare counterexample of a Mn(IV)-oxo species with a more prominent signal at g $ 2 was described by van Eldik et al. for the oxidation of [Mn II (bpy) 2 Cl 2 ] by H 2 O 2 . 94 Although the authors did not perform explicit calculations to reveal the value of D for this system, limiting direct comparison, the formation of a Mn(IV)-oxo complex was supported by UV/vis spectroscopy. Consequently, the features at g $ 2 observed in the present study could plausibly be assigned to high-valent manganese-oxo intermediates. Very recently, an unprecedented signal at g $ 1.8 was observed during water oxidation reactions involving manganese-containing heterogeneous catalysts; the feature was assigned to a low-spin Mn(IV)-oxo species with S ¼ 1/2. 95 Although the signal at $2 described in the present work could be also generally ascribed to a low-spin Mn(IV)-oxo species, we could not adequately t the data using a S ¼ 1/2 model; a S ¼ 3/2 model provides a superior t.

When UV/vis is used to analyze a rapidly mixed MeCN solution containing 0.1 mM 3 and 0.5-1.0 mM O 2 c À at À40 C, we observe bands similar to those found for analogous reactions with 2: sharp features at 423 and 448 nm and a much broader band centered at 595 nm (Fig. 9 , green and blue lines). Over longer periods of time, however, a new band appears at 520 nm as the 595 nm feature vanishes, resulting in the solution changing color from blue to purple (Fig. 9 , magenta line). Since a 520 nm band was not observed for 2, this suggests that 3 proceeds through an additional intermediate. The 520 nm peak is also observed when the free H 4 qp2 ligand is oxidized by O 2 c À under similar reaction conditions, suggesting that this new feature corresponds to a purely ligand-derived electronic transition (Fig. S19 †) .

As with the other bands, the 520 nm peak observed during oxidation of both 3 and H 4 qp2 cannot correlate to a charge transfer process between the ligand and the metal center but must instead be associated with a redox process on the ligand itself. The purple color of the solution is reminiscent of the charge-transfer complex quinhydrone, which consists of a reduced quinone interacting with an oxidized para-quinone through ring stacking. [96] [97] [98] In the presence of light, an electron from the reduced hydroquinone donor transfers to the oxidized benzoquinone acceptor, accounting for the characteristic purple hue of quinhydrone. 99, 100 This charge-transfer complex can only form when both redox forms of the quinol groupthe two-electron oxidized para-quinone and fully reduced quinolexist in the solution at the same time. That a similar complex is not observed during the oxidation of either H 2 qp1 or 2 suggests that this is an intramolecular, rather than an intermolecular, process that requires the molecule to have two redox-active quinol/para-quinone subunits. Although the existence of this species cannot be clearly resolved from the UV/vis data due to interference from the strong absorbance bands of the semiquinone radicals, the cryo-MS data demonstrate that a Mn(II)quinol-para-quinone intermediate e is produced during the oxidation of 3 by O 2 c À (Scheme 4, the m/z ¼ 538.1404 species in Fig. 4) . In intermediate e, the quinhydrone would form from intramolecular ring stacking interactions between the quinol and para-quinone groups in the coordination sphere.

Since the UV/vis data recorded during the oxidation of 3 by O 2 c À are dominated by the strong absorption bands assigned to semiquinone radical anions (422 and 448 nm), 79,80 quinhydrone (520 nm), [96] [97] [98] and also partially to p-association charge transfer equilibria (the red shied band at 595 nm), 81, 82 it is difficult to identify additional intermediates using these data by themselves (Fig. S20 †) . With this system, the CSI-MS results are quite helpful to understanding the overall mechanism (Scheme 4) and, concerning their complexity, are rare for SOD catalytic cycles. 38 Lastly, we note that the UV/vis data suggest that the H 4 qp2 ligand in 3 undergoes secondary oxidation reactions, specically oxygenation of the methylene positions and cleavage of the quinol groups from the 1,2-ethanediamine backbone, more slowly than the H 2 qp1 ligand in 2. This can be rationalized by the much higher SOD activity of 2, which more quickly decomposes the O 2 c À that could sufficiently scavenge reactive Mn(IV)-oxo and hydroxyl radicals (vide supra).

The reaction between 3 and O 2 c À was also characterized by EPR. In these experiments, MeCN solutions of 3 were rapidly mixed with an excess of O 2 c À at 233 K, and the resulting mixtures were frozen at 9 K aer 30 s and analyzed by X-band EPR (Fig. S21 †) . Similar to the reactivity of complex 2, the vast majority of the superoxide has been consumed by the 30 s time point. When the reaction is quenched at 10 s, however, we observe an additional sharp signal that can be attributed to residual superoxide (Fig. S22 †) . The 9 K spectrum obtained at 30 s (Fig. S21 †) is complicated and suggests the presence of multiple species. The intense signal centered at g $ 2.0 with the partially resolved hyperne splitting from the manganese nucleus (I ¼ 5/2) is most likely a Mn(II) species. In addition, the spectrum exhibits multiple low-eld components below 2000 G and high-eld components above 4000 G. These features probably result from the superposition of signals from one or more distinct S ¼ 3/2 species, with the S value being characteristic of a mononuclear Mn(IV) species with D > hn. [86] [87] [88] [89] [90] [91] [92] Using EasySpin, 93 it was possible to simulate the spectrum with two components: a Mn(IV) species (S ¼ 3/2) and a Mn(II) species (S ¼ 5/2) (Fig. S21 †) . Fitting this spectrum was not trivial due the multiple components and many parameters needed for optimization. As a result, the obtained t, although reasonable, is not as accurate as that for complex 2. Nevertheless, the model still accounts for the main features and reproduces most of the features observed for a highly complicated spectrum. According to our model, the Mn(II) and Mn(IV) species are present in an approximately 1 : 1 ratio (Table S3 †) . The Mn(IV) complex has a D value of magnitude 0.323 cm À1 , which is slightly greater than hv. The E/D ratio for the Mn(IV) species generated from 3 is small suggesting that this species, like the one generated from 2, is slightly axially distorted with negligible rhombicity. An alternative EasySpin model that also provided a reasonable t to the data included two Mn(IV) species: one with 2jDj ( hn and another with a D value greater than hn (Fig. S23 † and Table S3 †). Although there is little difference between the two models used to t 3, we believe that, though possible, it is unlikely for two different reactive Mn(IV) species to coexist together under these conditions. As previously mentioned, the hyperne splitting observed around the central g $ 2 feature suggests that a Mn(II) species is more likely, although we were unable to simulate these features using either model. Irrespective of which model is correct, both give evidence for the existence of a Mn(IV) species. It should be noted that other models were also tested that consisted of one or two Mn(IV) species with 2jDj ( hn, but these did not yield accurate simulations representative of our spectrum for 3. Models involving Mn(II) alone were also tested, but they likewise did not replicate as many of the spectral features.

Although complexes 1 and 2 are structurally similar, with the only difference being the identity of the para substituent in the phenolic portion of the ligand, the two compounds display markedly different activities with O 2 c À . The para-methyl group in 1 provides a Mn(III/II) reduction potential that is low enough to enable outer-sphere electron transfer; whereas, the parahydroxyl group in 2 increases this reduction potential enough to make this pathway energetically non-competitive with innersphere oxidation. The hydroxyl-for-methyl substitution on the phenol also increases the catalytic rate, with 2 displaying activity comparable to those of cationic porphyrin manganese complexes 62, 71 and potent pentaazamacrocyclic Mn SOD mimetics that have entered clinical trials, but with superior aqueous stability to the latter. [101] [102] [103] The inclusion of a second quinol in 3 was found to destabilize the Mn(II) complex in water and reduce both the catalytic activity and its substitution reactivity. Both effects are proposed to result from a lesser overall charge on the manganese complex, which should weaken the ability of the complex to attract both O 2 c À and competitive inhibitors, such as phosphate. Although a typical counter-ion inhibition is not observed, 3 catalytically degrades O 2 c À through an inner-sphere mechanism. The CSI-MS and time resolved UV/vis data provide evidence for a host of reaction intermediates that demonstrate the inner-sphere nature of the catalysis and reveal that both the manganese centers and the ligands are oxidatively active during catalysis. Though involvement of high-valent manganese species is well documented in the catalase-like activity of porphyrin-and non-porphyrin-based enzyme mimetics, 67,104 generation of Mn(IV)-oxo in a direct reaction between a monomeric Mn(II) complex and superoxide within a SOD catalytic cycle had not been documented prior to this report. In the case of a polymeric mixed valence Mn(III)/ Mn(II) complex, Mn(III)-O-Mn(IV) was identied as an intermediate in a reaction with superoxide by EPR spectroscopy. 105 Our explicit observation of Mn(IV)-oxo intermediates generated from mononuclear Mn(II) SOD mimetics by CSI-MS, UV/vis and EPR spectra is unique. Using the SOD activity of the Zn(II) analogue of 2 (4) as an estimate for the ligand-only contribution to the reactivity, we conclude that the metal-based inner-sphere SOD mechanism provides the more efficient pathway for superoxide degradation in Mn-quinol/phenol-based systems. The steric/ conformational features enabling stable pentacoordination upon ligand oxidation and higher overall positive charges of 2 allow it to efficiently bind O 2 c À . The oxidation of the Hqp1 À ligand to neutral qp1 also limits the accumulation of a potentially deleterious Mn(IV)-oxo species, relative to the H 4 qp2 system. These aforementioned factors combine to make 2 arguably the most effective reported Mn SOD mimic when one accounts for its higher stability. The results here demonstrate that the concerted/complementing redox-activity of manganese and organic ligands can be a highly effective strategy for developing small molecule antioxidants.

Most of the data that support the ndings of this study are present in the article and its ESI. The corresponding authors will provide additional data not available in these documents upon request.

Computational Advances in Organic Chemistry

Oxidative Stress and Antioxidant Protection: The Science of Free Radical Biology and Disease

Redox-Active Therapeutics. Oxidative Stress in Applied Basic Research and Clinical Practice

Functions of Quinones in Energy Converting Systems

Charge Transfer Interactions of Biomolecules

The two clinical trials involving M40403 have been suspended (NCT00033956) or terminated (NCT00101621). The studies on the safety, tolerability, and pharmacokinetics of the derivatives GC00034702 and GC00034711 are being performed (NCT03096756 and NCT03099824)

We thank Auburn University, the National Science Foundation Meyer for enabling EPR measurements in his laboratory and Prof. Lena J. Daumann for the support in the analysis and interpretation of EPR data.

The authors declare no competing nancial interest.