Article pubs.acs.org/JACS
Cite This: J. Am. Chem. Soc. 2018, 140, 3798−3808
The Catalytic Reaction of Nitroxides with Peroxyl Radicals and Its Relevance to Their Cytoprotective Properties Markus Griesser, Ron Shah, Antonius T. Van Kessel, Omkar Zilka, Evan A. Haidasz, and Derek A. Pratt* Department of Chemistry and Biomolecular Sciences, University of Ottawa, Ottawa, Ontario K1N 6N5, Canada S Supporting Information *
ABSTRACT: Sterically-hindered nitroxides such as 2,2,6,6tetramethylpiperidin-N-oxyl (TEMPO) have long been ascribed antioxidant activity that is thought to underlie their chemopreventive and anti-aging properties. However, the most commonly invoked reactions in this contextcombination with an alkyl radical to give a redox inactive alkoxyamine or catalysis of superoxide dismutationare unlikely to be relevant under (most) physiological conditions. Herein, we characterize the kinetics and mechanisms of the reactions of TEMPO, as well as an N-arylnitroxide and an N,N-diarylnitroxide, with alkylperoxyl radicals, the propagating species in lipid peroxidation. In each of aqueous solution and lipid bilayers, they are found to be significantly more reactive than Vitamin E, Nature’s premier radical-trapping antioxidant (RTA). Inhibited autoxidations of THF in aqueous buffers reveal that nitroxides reduce peroxyl radicals by electron transfer with rate constants (k ≈ 106 to >107 M−1 s−1) that correlate with the standard potentials of the nitroxides (E° ≈ 0.75−0.95 V vs NHE) and that this activity is catalytic in nitroxide. Regeneration of the nitroxide occurs by a two-step process involving hydride transfer from the substrate to the nitroxide-derived oxoammonium ion followed by H-atom transfer from the resultant hydroxylamine to a peroxyl radical. This reactivity extends from aqueous solution to phosphatidylcholine liposomes, where added NADPH can be used as a hydride donor to promote nitroxide recycling, as well as to cell culture, where the nitroxides are shown to be potent inhibitors of lipid peroxidation-associated cell death (ferroptosis). These insights have enabled the identification of the most potent nitroxide RTA and anti-ferroptotic agent yet described: phenoxazine-N-oxyl.
■
INTRODUCTION Nitroxides have demonstrated exciting biological activities linked to their ability to mitigate the effects of oxidative stress. Their protective effects have been documented in a staggering array of cell and animal models of disease, including neurodegeneration,1 Alzheimer’s,2 Huntington’s,3 Parkinson’s,4 multiple sclerosis,5 diabetes,6 ischemia-reperfusion injury,7 obesity, and hyperlipidemia.8 Furthermore, nitroxide administration has been associated with a significant increase in lifespan in a variety of organisms, including fruit flies,9 freshwater annelids,10 and mice.11 Very recently, mitochondria-targeted nitroxides have been reported to be potent inhibitors of ferroptosis,12 a form of regulated necrosis associated with the accumulation of lipid hydroperoxides,13−15 perhaps hinting that the mechanism that underlies their cytoprotective effects involves the inhibition of lipid peroxidation. The ability of nitroxides to mitigate oxidative stress is often ascribed to their reactivity as superoxide dismutase (SOD) mimics. This reactivity, established largely by the efforts of Samuni, Krishna, Goldstein, and co-workers,16−19 is believed to result from the sequence of reactions illustrated in Scheme 1A for the archetypal nitroxide, 2,2,6,6-tetramethylpiperidine-N© 2018 American Chemical Society
oxyl (TEMPO). Although there is no doubt that nitroxides undergo this chemistry, it is unclear if this mechanism is physiologically relevant because cells generally have a comparable or higher level of SOD enzymes:20 Cu/Zn-SOD to protect the cytoplasm 21 and Mn-SOD to protect mitochondria.22 Another mechanism often invoked to account for the biological activity of nitroxides involves the scavenging of alkyl radicals, a reaction that is purportedly the basis for many newly developed nitroxide-containing redox probes.23−25 Despite being the quintessential reaction of nitroxides, this reaction is too slow (k ≈ 1−3 × 108 M−1 s−1)26,27 to compete with the formation of peroxyl radicals from the combination of alkyl radicals and O2 (k ≈ 3 × 109 M−1 s−1),28 a competition made more unlikely because O2 is generally present in much higher concentration than nitroxide (mM vs μM to nM). Although nitroxides have long been demonstrated to be unreactive to peroxyl radicals in organic solution,29 Goldstein and Samuni provided compelling evidence using pulse radiolysis that nitroxides react readily with peroxyl radicals in Received: January 25, 2018 Published: February 16, 2018 3798
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society
water (i.e., k = 3−10 × 107 M−1 s−1 for TEMPO and a variety of alkylperoxyl radicals).30 The mechanism was believed to be an electron transfer, analogous to the first step of the SOD mimicking mechanism shown in Scheme 1A, with concomitant formation of the alkylperoxide anion and the oxoammonium ion derived from TEMPO (2,2,6,6-tetramethylpiperidinium-Noxide, hereafter TEMPO+). In collaboration with the Valgimigli group, we reported that nitroxides undergo acid-promoted reactions with peroxyl radicals31 and, more recently, that the nitroxide can be recycled from the resultant oxoammonium ion by reduction with a substrate-derived radical (Scheme 1B).32 Thus, nitroxides are catalytic radical-trapping antioxidants (RTAs), and the substrate-derived radicals are themselves the stoichiometric reductants. Evidence for this mechanism was obtained at ambient temperatures in organic solution to which acids were added31,32 as well as at elevated temperatures in hydrocarbons where acids are formed in situ.32 We wondered whether this mechanism could contribute to the antioxidant activity of nitroxides in aqueous solution and, more importantly, in lipid bilayers where lipid peroxyl radicals propagate the peroxidation of lipids. The results of our investigations are reported here.
Scheme 1. Catalytic Trapping of Peroxyl Radicals by Nitroxides: (A) TEMPO Catalytically Reduces Hydroperoxyl Radicals in Aqueous Solution (See Ref 19), Generally Referred to as Superoxide Dismutation, and (B) in the Presence of Acid (See Ref 22), TEMPO Catalytically Reduces Alkylperoxyl Radicals in Organic Solution
■
RESULTS
I. Nitroxides Catalytically Trap Peroxyl Radicals in Aqueous THF. TEMPO was initially investigated for its ability to inhibit the free radical chain oxidation (autoxidation) of THF. THF is a convenient substrate for determination of both the kinetics and stoichiometry of peroxyl radical reactions as it is both miscible with aqueous buffers and sufficiently reactive to
Figure 1. Co-autoxidation of THF (3.1 M) and STY-BODIPY (10 μM) initiated by AAPH (1 mM) in either acetate or phosphate buffer (100 mM) of pH 4 (A), 7 (B), and 10 (C) at 37 °C (black) inhibited by 2 μM TEMPO (red), TEMPO+BF4− (green), or Trolox (blue). Reaction progress was monitored by loss of STY-BODIPY (D) absorbance at 562 nm (ε = 132,261 M−1 cm−1) which enables the determination of rate constants (kinh) and reaction stoichiometries (n) for reactions of inhibitors with chain-carrying peroxyl radicals (E). Transient concentrations of TEMPO monitored in parallel by EPR at pH 4 (black), 7 (red), and 10 (blue) (F). 3799
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society participate in a chain reaction with well-defined kinetics.33 Reactions were initiated with the water-soluble azo initiator AAPH and monitored by the consumption of STY-BODIPY, a highly absorbing and oxidizable co-substrate (Figure 1D).34 Representative results obtained at pH 4, 7, and 10 are shown in Figures 1A−C, respectively. Plotted alongside the data are the results of THF/STY-BODIPY co-autoxidations inhibited by Trolox, a water-soluble analogue of α-tocopherol (α-TOH), the most biologically active form of Vitamin E, and an important benchmark of RTA activity.35 Visual comparison of the reaction progress data reveals that TEMPO is both significantly more reactive to peroxyl radicals (lower initial rate) and has a greater radical-trapping capacity (longer inhibited period) than Trolox. The rate constants derived from the initial rates of the inhibited autoxidations (eq 1, Figure 1E) indicate that the reactivity of TEMPO toward peroxyl radicals is essentially pH-independent, viz. kinh = (2.8 ± 0.4) × 106, (4.6 ± 0.2) × 106, and (3.7 ± 0.2) × 106 M−1 s−1 at pH 4, 7, and 10, respectively. These data can be compared with those obtained for Trolox (kinh = 2.1 × 105, 4.1 × 105, and 8.7 × 105 M−1 s−1, respectively), which increase ever so slightly with pH due to the increasing contribution of a sequential proton-loss electron-transfer mechanism.33 The rate constants determined for the reactions of TEMPO are roughly 1 order of magnitude lower than those determined by Goldstein et al. by pulse radiolysis, consistent with the depressed polarity of the current medium.30 The benefits of carrying out inhibited autoxidations in lieu of pulse radiolysis experiments to assess the reactivity of the nitroxides to peroxyl radicals are not limited to the “cleanliness” of the reaction,36 but include the ability to directly determine the stoichiometry of peroxyl radical trapping, which is derived from the length of the inhibited period (τinh) as in eq 2 (Figure 1E). Although this is easily done for Trolox, yielding n ∼ 2 at each pH, this is possible for TEMPO only at pH 4 because the autoxidations at pH 7 and 10 remain retarded well beyond the initial inhibited period. The duration of the inhibited period at pH 4 indicates that TEMPO traps roughly 3-fold more radicals than Trolox (i.e., n ∼ 6). Thus, as in organic solution, TEMPO appears to catalytically trap radicals in aqueous solution, but in the absence of acid. THF/STY-BODIPY co-autoxidations were also carried out in the presence of TEMPO+ (added as the tetrafluoroborate salt). The reaction progress curves are also included in Figure 1. Interestingly, the data are practically superimposable on those obtained with TEMPO, suggesting that TEMPO+ can be reduced in situ to yield TEMPO. To corroborate this result, corresponding experiments were carried out in the cavity of an EPR spectrometer. The results are shown in Figure 1D, from which it is clear that the temporal evolution of the TEMPO signal at each pH is consistent with the extent to which each of the autoxidations are inhibited. That is, TEMPO is consumed relatively rapidly at pH 4, but is consumed progressively more slowly with increasing pH. Since the autoxidation remains inhibited and TEMPO is not consumed, it must be reformed in situ. Moreover, because no external reductant is supplied, the substrate must serve as the stoichiometric reductant. II. The Substrate is the Stoichiometric Reductant in Nitroxide Inhibited Autoxidations of THF. During efforts to monitor the foregoing THF/STY-BODIPY co-autoxidations by EPR, it became apparent that TEMPO+ was not stable under the reaction conditions, yielding significant amounts of TEMPO at t = 0, even in the absence of STY-BODIPY and
AAPH. Since THF and phosphate buffer were the only other components of the reaction mixture, their interaction with TEMPO+ was investigated by monitoring the loss of its absorbance at 300 nm (where TEMPO+ and TEMPO have a maximum difference in their spectra, see Figure 2A inset). Representative results are shown as a function of THF concentration in Figure 2A from which the rate constant for the reaction was determined to be (1.4 ± 0.3) × 10−4 M−1 s−1.
Figure 2. TEMPO+ (1 mM) in phosphate buffer of pH 7 reacts with THF as monitored by loss of its absorbance at 300 nm (37 °C). Inset: Difference in absorbance of TEMPO+ and TEMPO (A). CBS-QB3calculated transition state structure for the reaction of TEMPO+ with THF (B).
The observation that TEMPO+ can oxidize THF was somewhat surprising because there are few examples of ether oxidation by oxoammonium ions in the literature and they are limited to activated (benzylic) ethers. Alkyl ethers are generally described as unreactive.37−39 The oxidation of alcohols by oxoammonium ions has been the subject of considerable study40 and is believed to occur by formal hydride transfer from the Cα-H to the oxygen atom of the oxoammonium cation.41 The results of CBS-QB342 calculations (including a continuum solvation model) suggest that a similar mechanism is at play for the reaction of THF and TEMPO+, as illustrated by the transition state structure shown in Figure 2B. The calculated free energy barrier for the reaction of ΔG‡ = 26.2 kcal mol−1 predicts a second order rate constant of 5.3 × 10−5 M−1 s−1 upon application of transition state theory, which is in excellent agreement with the experimental data. Additional support for this mechanism was obtained by carrying out kinetic experiments in d8-THF, which yielded a kinetic isotope effect of kH/kD = 3.5 on the oxidation of THF by TEMPO+ (see Supporting Information for the data). To assess the relevance of each of the TEMPO+/THF and the TEMPO+/R• reactions on the turnover of TEMPO in the 3800
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society
Figure 3. Simulation of the reaction progress curves for the TEMPO-inhibited autoxidation of THF according to variations in the (abbreviated) kinetic model defined by eqs 3−8 (dotted lines), eqs 3−10 (dashed lines), and eqs 3−13 (solid lines) (A). Numerical fits of experimental traces of THF/STY-BODIPY coautoxidations inhibited by TEMPO at pH 7 and 10 as well as the best possible fits without including eq 11 (B). Experimental data at pH 4 inhibited by TEMPO (red) as well as TEMPOH (green) in contrast to the theoretical fit without eq 11 (blue, dashed) (C).
TEMPO + R•/THF• → TEMPO‐R/TEMPO‐THF
autoxidation of THF, we constructed a kinetic model of the autoxidation using the simulation software COPASI.43 An abbreviated scheme (omitting the reactions of STY-BODIPY for clarity) and the corresponding rate constants are given below, and the results of the simulation are visualized in Figure 3. The complete scheme and associated rate constants can be found in the Supporting Information. AAPH → 2 R• •
(14)
Simulations of inhibited autoxidations carried out without inclusion of the TEMPO+/THF reaction (eq 11) or subsequent reactions of TEMPOH (eqs 12 and 13) in the kinetic model displayed neither the protracted inhibited period nor the subsequent retardation of the autoxidation observed experimentally (Figure 3A). This suggests that the steady-state concentration of alkyl radicals that propagate the autoxidation is simply too low to sustain enough TEMPO to effectively inhibit the reaction. Thus, once TEMPO is initially consumed (corresponding to n = 1), the autoxidation proceeds at the uninhibited rate. Inclusion of the TEMPO+/THF reaction in the kinetic model, as well as the subsequent reaction of TEMPOH with peroxyl radicals (eq 12, reported kinh = 2.4 × 106 M−1 s−1),31 affords simulated data that is strikingly consistent with experiment. The known comproportionation of TEMPO+ and TEMPOH to yield TEMPO was also included (eq 13), but given the known rate constant of k = 50 M−1 s−1,44 this is not expected to be a major pathway as it requires simultaneous accumulation of both species, whereas TEMPOH will react readily with ROO•. The loss of TEMPO via its combination with alkyl radicals (eq 14) limits the turnover of the cycle; omission of this reaction in the kinetic scheme leads to perpetual retardation of the autoxidation. The data sets obtained at pH 7 and 10 could be readily fit to this kinetic model, but the data sets at pH 4 could not. Attempts to do so yielded rate constants that exceeded diffusion for eq 14 (k ≈ 2 × 1011 M−1 s−1), suggesting that at least one other off-cycle reaction limits radical-trapping activity. Since the pKa of TEMPOH2+ has been reported to be
(3) •
R + O2 → ROO
(4)
•
•
ROO + THF → ROOH + THF
(5)
THF• + O2 → THF‐OO•
(6)
kp
THF‐OO• + THF → THF‐OOH + THF• kt
2 ROO• /THF‐OO• → non‐radical products
(7) (8)
k inh
ROO• /THF‐OO• + TEMPO ⎯→ ⎯ ROO− /THF‐OO− + TEMPO+ k reg1
TEMPO+ + R•/THF• ⎯⎯⎯→ TEMPO + R+/THF+ k reg2
TEMPO+ + THF ⎯⎯⎯→ TEMPOH + THF+
(9) (10) (11)
k inh2
ROO• /THF‐OO• + TEMPOH ⎯⎯⎯→ ROOH/THF‐OOH + TEMPO
(12)
TEMPO+ + TEMPOH → 2 TEMPO
(13) 3801
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society
Figure 4. Investigated nitroxides (A). Co-autoxidation of THF (3.1 M) and STY-BODIPY (10 μM) initiated by AAPH (1 mM) in phosphate buffer (100 mM) of pH 7 at 37 °C (black) and inhibited by 2 μM TEMPO, TBPNO, DHQNO, and Trolox (B) monitored at 562 nm (ε = 132,261 M−1 cm−1). Cyclic voltammograms of TEMPO, TBPNO, and DHQNO recorded in acetonitrile (E° vs NHE) (C).
(recall for TEMPO+ it was 5.3 × 10−5 M−1 s−1). Although some of the predicted rate acceleration must be based on the differing electronics, the lesser steric hindrance surrounding the key N− O moiety of TBPNO+ as compared to that of TEMPO+ must also contribute. The relative stability of the monoaryloxoammonium ion (hereafter DHQNO+) implies that its reaction with THF will be slower than that of TEMPO+, but because of the reduced steric hindrance, it is predicted to be 6-fold faster (3.2 × 10−4 M−1 s−1). The poor turnover of the monoaryl compound may therefore be attributed to competing reactions. The (highly reproducible) “kink” in the initial inhibited region of the TBPNO-inhibited autoxidation (see inset, Figure 4B) deserves further comment. Since the conversion of TBPNO+ to its hydroxylamine (TBPNOH) via reaction with THF is predicted to be relatively facile, we wondered if the inhibition improves with time because TBPNOH is a better peroxyl radical-trapping agent than TBPNO. Hence, TBPNOH was synthesized and the same inhibited autoxidations were carried out. Indeed, the data obtained were essentially superimposable on those shown in Figure 4B after the initial decay in STY-BODIPY between 0 and ∼500 s (kinh = (4.7 ± 0.5) × 106 M−1 s−1, see Supporting Information). IV. Nitroxides Inhibit the Autoxidation of Phosphatidylcholine Lipid Bilayers. The reactivity of nitroxides as catalytic RTAs in aqueous solution, but not organic solution, prompted our investigation of the same series of compounds in biphasic media more representative of the biological contexts in which this reactivity may be relevant. Unilamellar liposomes of egg phosphatidylcholine (Egg-PC) supplemented with STYBODIPY as the signal carrier were used for this purpose, such that the kinetics and stoichiometry of radical-trapping could be determined as above for the THF/STY-BODIPY coautoxidations.48,49 Representative results are included in Figure 5A. The reaction progress curves reveal that each of the nitroxides are reactive RTAs in lipid bilayers and that all are significantly more reactive than α-TOH (from the initial rates, kinh = (3.2 ± 0.2) × 104, (3.3 ± 0.4) × 104, and (5.3 ± 0.3) × 104 M−1 s−1 for TEMPO, TBPNO, and DHQNO, respectively, versus (4.7 ± 0.4) × 103 M−1 s−1 for α-TOH). In this medium, the RTA activity of the nitroxides is not obviously catalytic. In fact, TEMPO appears to trap only a single peroxyl radical, and after the initial inhibited period of ∼1000 s, the reaction progress is essentially that of the uninhibited reaction. On the other hand, the autoxidations inhibited by TBPNO and
∼7.5,44 we surmised that at pH 4 TEMPOH is largely protonated, and a much poorer H-atom donor. In fact, the O− H BDE of TEMPOH2+ is calculated to be 116 kcal mol−1, 46 kcal mol−1 higher than in TEMPOH (CBS-QB3). Consistent with the foregoing, when autoxidations were carried out at pH 4 in the presence of TEMPOH, little inhibition was observed (see Figure 3C). III. Peroxyl Radical-Trapping Antioxidant Activity Extends to Other Relevant Nitroxides. Since arylnitroxides are intermediates in the catalytic inhibition of hydrocarbon autoxidation by diarylamine RTAs45−47 and may contribute to the anti-ferroptotic activity of liproxstatin (and other monoarylamines),48 we expanded the foregoing studies to include a model diarylnitroxide (bis(4-tert-butylphenyl)nitroxide, TBPNO) and a monoarylnitroxide (the dihydroquinoline-Noxyl, DHQNO) derived from a liproxstatin model compound (shown in Figure 4A).48 The results of THF/STY-BODIPY coautoxidations in phosphate buffer of pH 7 inhibited by these compounds are shown in Figure 4B (corresponding results obtained at pH 4 and 10 are included in the Supporting Information). Introduction of the aryl substituents substantially changes the reactivity of the nitroxide. Although they remain more reactive than Trolox, the profiles differ from both TEMPO and each other. The reactivity of the monoaryl nitroxide DHQNO to peroxyl radicals is initially indistinguishable from TEMPO (kinh = (5.4 ± 0.4) × 106 M−1 s−1), whereas the diarylnitroxide TBPNO is noticeably less reactive (kinh = (1.3 ± 0.4) × 106 M−1 s−1), but only initially so (more on this later). However, although DHQNO traps fewer radicals than TEMPO (similarly to Trolox, for which n ∼ 2), TBPNO traps more. To rationalize these observations, and based on the evidence, which suggests that the (initial) reaction between the nitroxide and peroxyl radical is an electron transfer, the standard potentials of the three nitroxides were determined by cyclic voltammetry. The results are shown in Figure 4C, where it is clear that the trend in the potentials matches kinh; DHQNO is very slightly more oxidizable than TEMPO (E° = 0.85 and 0.88 V vs NHE, respectively), and both are more oxidizable than TBPNO (E° = 0.99 V). The corresponding relative instability of the diaryloxoammonium ion (hereafter TBPNO+) suggests that it will be the most easily reduced by the substrate, providing a rationale for its greater catalytic activity. Indeed, the CBS-QB3-calculated rate constant for the reaction of TBPNO+ with THF is 1.6 M−1 s−1 3802
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society
Figure 5. Co-autoxidations of egg phosphatidylcholine lipids (1 mM) and STY-BODIPY (8 μM, monitored at 565 nm, ε = 123,676 M−1 cm−1) suspended in phosphate-buffered saline (10 mM) at pH 7.4 initiated by MeOAMVN (0.2 mM) at 37 °C and inhibited by 2 μM of nitroxides (A). Co-autoxidations under identical conditions except initiated by 0.68 mM MeOAMVN and using PBD-BODIPY (10 μM, monitored at 584 nm, ε = 118,548 M−1 cm−1) as indicator. In addition, the experiments were performed in an EPR cavity under identical conditions to directly monitor the nitroxide concentrations. The inhibitory effect of 8 μM nitroxide, 8 μM NADPH, as well as their combination is depicted for TEMPO (B), lipoTEMPO (C), TBPNO (D), and DHQNO (E).
concentration then rebounds at ∼1500 s (see Figure 5B), peaking at around half the initial concentration at ∼3500 s and then falling to zero, all while the autoxidation remains retarded. These results can be rationalized on the basis of the greater hydrophilicity of TEMPO+ relative to DHQNO+ and TBPNO+. Thus, upon oxidation of TEMPO with a peroxyl radical, TEMPO+ partitions to the aqueous phase where it is readily reduced to TEMPOH by NADPH. TEMPOH and TEMPO+ can comproportionate to yield TEMPO that accumulates in the aqueous phase, gradually partitioning back to the lipid where it can inhibit the autoxidation. In contrast, DHQNO+ and TBPNO+ remain largely in the lipid and/or interfacial region where, once reduced by NADPH, the hydroxylamines readily undergo reactions with peroxyl radicals. The nitroxides never accumulate because they are formed in the lipid bilayer, or at least at the interface, and react readily with peroxyl radicals as they are formed in the lipid. To provide some additional insight on this point, we synthesized a TEMPO derivative of greater lipophilicity (4amino-TEMPO acylated with dodecanoic acid, hereafter lipoTEMPO) and examined its ability to inhibit lipid peroxidation in the same liposomes. Although this nitroxide is inherently less reactive than TEMPO (due to both inductive electron withdrawal by the amide group and reduced mobility in the bilayer due to the C12 side chain, see Supporting Information for more details), it is still an effective inhibitor (Figure 5E). Moreover, although it appears to act synergistically with NADPH, the nitroxide is not initially consumed and then reformed as is TEMPO, presumably due to its anchoring in the lipid. Instead, the concentration of the nitroxide is largely maintained, whereas the autoxidation is inhibited. Since the nitroxide itself is less reactive than the corresponding
DHQNO are retarded beyond the initial inhibited period, but only slightly. Given the lack of THF or other potential hydride donors to reduce the product oxoammonium ions, it is possible that the reductant is a lipid-derived alkyl radical as we first surmised. If this were the case, TEMPO, being the most polar of the three nitroxides, may not turn over because TEMPO+ more readily partitions to the aqueous phase away from lipid radicals. The foregoing experiments were also carried out in the presence of NADPH to determine if a hydride donor can facilitate catalytic RTA activity in liposomes as in aqueous THF. The results are shown in Figure 5B−D, where superimposed on the reaction progress determined by PBD-BODIPY50 coautoxidation, the concentrations of the nitroxides (determined by EPR spectroscopy) are also indicated. In the presence of 1 equiv of NADPH, the inhibition period corresponding to the reaction of the nitroxide with chain-carrying peroxyl radicals at least doubled compared to its absence. This result implies that NADPH acts synergistically with the nitroxide, presumably by reduction of the corresponding oxoammonium ion,51 because NADPH alone only retards the oxidation.52 The changes in concentration of the nitroxide over the course of the autoxidations indicate very different dynamics between TEMPO and either TBPNO or DHQNO. In the absence of NADPH, the rate of nitroxide loss (and corresponding formation of oxoammonium ion) is similar for all compounds with the small differences being consistent with the relative reactivity of each with peroxyl radicals (kinh). When NADPH is present, similar curves are observed for DHQNO and TBPNO wherein the rate of loss of nitroxide is slightly lower. However, the TEMPO profile is strikingly different. Although the initial loss of TEMPO is similar, the nitroxide 3803
DOI: 10.1021/jacs.8b00998 J. Am. Chem. Soc. 2018, 140, 3798−3808
Article
Journal of the American Chemical Society
dismutation is unlikely because most cells express superoxide dismutase enzyme(s) that are greater in both concentration and specific activity than the nitroxides. If one considers these facts, it is likely that another reactivity is key. Insight into an alternative mechanism of cytoprotection comes from Goldstein’s oft-overlooked work that shows a direct reaction between nitroxides and peroxyl radicals,30,56 which are the propagating species in lipid peroxidation. Our results are in good agreement with Goldstein’s. Although the absolute rate constants we have determined from inhibited autoxidations are ∼1 order of magnitude lower (4.6 × 106 vs 3−5 × 107 M−1 s−1), this seems reasonable given that our reactions contain a significant amount of THF, which will slow electron transfer from the nitroxides to peroxyls to form the ionic oxoammonium and peroxide products. Most importantly, our results clearly show that the oxoammonium ion can be recycled by THF or other hydride donors in solution (see data obtained for hydride transfer with dioxane, dimethoxyethane, and co-autoxidations with 4-ethylbenzoic acid in the Supporting Information). These results enable the proposal of a mechanistic scheme accounting for the catalytic radical-trapping antioxidant activity of nitroxides under ambient conditions in the absence of added acid, which is illustrated with TEMPO in Scheme 2. TEMPO
hydroxylamine, its concentration does not drop until all the NADPH, and therefore hydroxylamine, is depleted. V. Nitroxides Can Be Potent Inhibitors of Ferroptosis. To ascertain how the RTA activity of nitroxides in liposomes may translate to cells, their potencies as inhibitors of ferroptosis were determined. Ferroptosis was induced in Pfa1 mouse embryonic fibroblasts53 by inhibition of the lipid hydroperoxide detoxifying enzyme glutathione peroxidase 4 (Gpx4) with (1S,3R)-RSL3.13,54,55 The results are shown in Figure 6 alongside corresponding data obtained with ferrostatin-1 (Fer-1) and liproxstatin-1 (Lip-1), the archetype ferroptosis inhibitors.
Scheme 2. Catalytic Cycle of Autoxidation Inhibition by TEMPO Figure 6. Dose−response curves obtained from RSL3 (100 nM) induced ferroptosis in Pfa1 cells inhibited by TEMPO (red ●), lipoTEMPO (black ■), TBPNO (blue ▲), DHQNO (green ⧫), Fer-1 (purple ▼) and Lip-1 (yellow ◀).
The reactivity trends of the nitroxides observed in the liposome autoxidations translate well to the potency of the compounds in cell culture. That is, the cytoprotection afforded by nitroxide supplementation follows the same general trend wherein TBPNO ∼ DHQNO > lipo-TEMPO > TEMPO. Although TEMPO was a rather poor inhibitor of ferroptosis (EC50 = 5300 nM), simple attachment of a lipophilic side chain resulted in much greater potency (EC50 = 241 nM). The results are consistent with our hypothesis that TEMPO (and/or TEMPO+) partitions significantly to the aqueous phase, reducing its efficacy at inhibiting lipid peroxidation occurring in the lipid-rich regions of the cell. The monoaryl and diarylnitroxides are both more lipophilic than TEMPO and have preferential reactivity for both the initial electron transfer and the subsequent regeneration steps, leading to EC50 values of 99 and 80 nM, respectively, only ∼2-fold higher than those of Fer-1 (45 nM) and Lip-1 (35 nM).
undergoes electron transfer to a peroxyl radical, inhibiting the propagation of autoxidation and generating TEMPO+. The rate of regeneration of TEMPO via reduction of TEMPO+ with an alkyl radical is generally too slow to contribute, therefore requiring the presence of a hydride donor to convert TEMPO+ to TEMPOH. TEMPOH can trap another peroxyl radical via H-atom transfer, thereby regenerating TEMPO. The catalytic cycle is limited by the deleterious (in this context, at least) combination of TEMPO with an alkyl radical, which forms a stable alkoxylamine. Moreover, turnover is limited at low pH due to the protonation of TEMPOH, whose conjugate acid is not an H-atom donor. This becomes important at pH
