Quinone methide dimers lacking labile hydrogen atoms are surprisingly excellent radical-trapping antioxidants† †Electronic supplementary information (ESI) available: Additional kinetic and thermodynamic data and analyses, synthesis and characterization data, computed optimized geometries, energies, and Cartesian coordinates. See DOI: 10.1039/d0sc02020f

Quinone method dimers, (bio)synthetic intermediates en route to many naturally products derived from resveratrol, are potent radical-trapping antioxidants, besting the phenols from which they are derived and to which they can be converted.


Introduction
The efficient inhibition of autoxidation continues to be a widelypursued objective, given the indispensability of hydrocarbonbased products (either petroleum-derived or plant-derived) in day-to-day life and their propensity to undergo autoxidation. Nevertheless, the inhibitors which are commonly added to these products have remained essentially the same for decades. 1,2 The most ubiquitous radical-trapping antioxidants (RTAs) are hindered phenols, such as butylated hydroxytoluene (BHT). BHT and related phenols react with the peroxyl radicals (ROOc) that propagate the autoxidation chain reaction. The mechanism is well understood to involve hydrogen atom transfer (HAT) from the phenolic O-H to a chain-carrying peroxyl radical followed by rapid combination of the resultant phenoxyl radical with a second peroxyl radical (Fig. 1A). 2 A vast literature has established that substituents can have a substantial impact on the rate of the initial H-atom transfer, which is characterized by the inhibition rate constant (k inh ). 2,3 Electrondonating groups accelerate the rate as they stabilize the electron-decient phenoxyl radical 4 and the transition state leading to its formation, whereas electron-withdrawing groups have the opposite effect.
As part of ongoing efforts to better understand the antioxidant activity of resveratrol and its oligomers, 5 we found that the dimeric quinone methides which we employed as the key synthetic intermediates en route to resveratrol dimers are signicantly better RTAs than their phenolic precursors or products (Fig. 1B). 6 This was entirely unexpected given that the quinone methide dimers (hereaer QMDs) lack phenolic Hatoms. In our preliminary report, we suggested four possible reaction paths that would account for this observation (Fig. 1C): (1) HAT from the C-H bonds on the carbon atoms which link the quinone methide moieties; (2) tautomerization or hydration of the quinone methide(s) followed by HAT from the resultant phenolic O-H; (3) direct addition of peroxyl radicals to the quinone methide to form persistent phenoxyl radicals, and (4) homolysis of the weak central C-C bond in the QMDs followed by combination of the resultant persistent phenoxyl radicals with peroxyl radicals. Herein, we offer the details of our efforts to elucidate the mechanism.

Synthetic procedures
A set of hindered resveratrol analogues (1a-j) were synthesized wherein the resorcinol ring was replaced with aryl rings bearing substituents of differing electronics (Chart 1). These phenols were dimerized via anodic oxidation in the presence of 2,6lutidine to produce quinone methide dimers (2a-j). 6 Aer basemediated isomerization of one quinone methide, Lewis acid activation and Friedel-Cras cyclization afforded a series of substituted quadrangularin (quad) A analogues (3a-e).

Inhibited autoxidations
The reactivities of the QMDs as RTAs were determined by the classical inhibited autoxidation approach 7 utilizing a co-autoxidation of 1-hexadecene and PBD-BODIPY ( Fig. 2A). 8 Inclusion of the latter enables monitoring of reaction progress via conventional spectrophotometry by loss of its absorbance at l max ¼ 588 nm upon addition of peroxyl radicals to its 1-phenylbutadiene moiety. Rate constants for the reaction of peroxyl radicals with the 10 QMDs (2), their 10 precursor stilbenoid phenols (1), and 5 quadrangularin A analogues derived therefrom (3) (k inh ) and corresponding reaction stoichiometries (n) were determined from the initial rate and inhibited period (t inh ) of the inhibited autoxidations, respectively, according to the expressions in Fig. 2B. Representative results are shown in Fig. 2C-E for series of equivalently-substituted 1, 2 and 3, respectively.
All 10 of the QMDs studied were determined to be excellent RTAs, with k inh $ 4 Â 10 6 M À1 s À1 at 37 C. The striking difference in reactivity between the QMDs and the equivalentlysubstituted precursor stilbenoid phenols and product quadrangularin A analogues is evident simply upon consideration of the raw inhibited autoxidation reaction progress data (compare the prominent inhibited periods in Fig. 2D to the retarded autoxidations in Fig. 2C and E, respectively), which amount to a difference in k inh of >10-fold. This is even more impressive given that the stilbenoid phenols from which the QMDs are derived are already $10-fold more reactive than the archetype hindered phenolic RTAs, such as BHT (k inh ¼ 2 Â 10 4 M À1 s À1 ). 9 Indeed, the reactivity of the QMDs under these conditions is on par with that of a-tocopherol, the most potent form of vitamin E and among the most reactive RTAs . 9 The observed inhibition periods correspond to the trapping of $2 radicals by the QMDs, which is similar to the precursor phenols, but roughly half that observed for the quadrangularin A analogues. The stoichiometries of the phenols are consistent with precedent (cf. Fig. 1A), 9 and since the quadrangularin A analogues contain two phenolic moieties, it follows that they trap twice as many peroxyl radicals. We also carried out inhibited autoxidations at 70 C and found that the superiority of QMDs compared to their precursor and product phenols is maintained (see ESI for the raw data †). However, pushing the temperature to 100 C reveals a noticeable drop in reactivity. The stoichiometry exhibited by the QMDs was also found to attenuate at progressively higher temperatures; see the kinetic data summarized in Table 1.

Mechanistic studies
Without a labile H-atom, the mechanism by which QMDs act as RTAs was not immediately obvious. The lack of substituent effects on the reactivity of the QMDs (r + $ 0 at 37 C, Fig. 2F) is in direct contrast to the sensitivity observed for the precursor phenols (r + ¼ À0.18 at 37 C), which is consistent with known structure-reactivity relationships for phenolic RTAs, in general (vide supra). The reactivities of the quadrangularin A analogues, which are also phenols, are lower than the precursor stilbenoid phenols and essentially independent of their substitution. This difference can be rationalized based upon the reduced conjugation between the substituted phenyl rings and the reactive hindered phenolic moieties in the quadrangularin A analogues relative to the stilbenoid phenols, which are fully conjugated (see ESI for the calculated minimum energy structures †). 10 Upon increasing the temperature, the trends among the two sets of phenols remain consistent (see ESI for the plots †), but those for the QMDs changedemonstrating a slight positive correlation at 70 C which increases slightly along with the diminution of reactivity at 100 C (Fig. 3A).
The rst mechanistic possibility we considered involves Hatom transfer from the benzylic positions adjacent to the methide carbons of the QMD. Although HAT from a benzylic carbon to a peroxyl radical is generally a sluggish reaction (k $ 1 M À1 s À1 ), 11 in this case HAT restores the aromaticity of one of the aryl rings, which suggests it may be signicantly enhanced. To probe this mechanism, an analogue of 2d was synthesized wherein both benzylic H-atoms were replaced with D-atoms (structure shown in Fig. 3B; for synthetic details see Experimental section), and inhibited autoxidations were carried out as above. 12 The results of this experiment (shown in Fig. 3B) revealed no kinetic isotope effect. Corresponding DFT computations on a model reaction (HAT from an analogous monomeric quinone methide to a methylperoxyl radical, shown in Fig. 3C) yield k HAT ¼ 0.3 M À1 s À1 from DG ‡ ¼ 20.9 kcal mol À1 and k H /k D ¼ 6. 13 Clearly, this mechanism fails to account for the reactivity of the QMDs.
The second mechanistic possibility we considered involves in situ tautomerization of the QMD to a stilbenoid phenol that can react with peroxyl radicals (Fig. 1C). The simple fact that the k inh values for the QMDs exceed that of the precursor stilbenoid (B) determination of inhibition rate constants (k inh ) and stoichiometries (n) for reactions of inhibitors with chain-carrying peroxyl radicals from initial rates and inhibition periods (t inh ). Co-autoxidations of 1-hexadecene (2.9 M) and PBD-BODIPY (10 mM) initiated by AIBN (6 mM) in chlorobenzene at 37 C (dashed black traces in (C-E)) and inhibited by 5 mM 1 (C), 1 mM 2 (D), and 5 mM 3 (E) (colour traces); (F) linear free energy relationships for 1, 2, and 3 at 37 C. phenols by more than an order of magnitude strongly suggests that a phenol is not involved in the mechanism; 6 however, we sought corroborating evidence from kinetic solvent effect (KSE 14 ) and solvent kinetic isotope effect (SKIE) experiments (see Fig. 3D). Thus, we carried out co-autoxidations of dioxane and PBD-BODIPY in chlorobenzene (PhCl) and a 2 : 1 PhCl : DMSO mixture. Previous studies 15 conducted under identical conditions have shown a predictable attenuation of k inh for HAT from X-H groups to peroxyl radicals due to a combination of the H-bond accepting capacities of the substrate (1,4-dioxane; b H 2 ¼ 0.41) and co-solvent (DMSO; b H 2 ¼ 0.78). 15 The results are shown in Fig. 3E, which reveal no suppression in the reactivity of the QMD in the presence of Hbond accepting (HBA) solvents. Furthermore, QMD-inhibited co-autoxidations of dioxane and PBD-BODIPY in PhCl to which 1% v/v MeOH or 1% v/v MeOD were added were indistinguishable (Fig. 3F), suggesting no exchangeable protons are involved in the RTA activity. 16 The third mechanistic possibility, addition of a peroxyl radical to one of the methide carbons of the QMD (Fig. 1C), has some precedent. Volodkin found quinone methides to be modest antioxidants (k inh $ 10 3 M À1 s À1 ), 17 but since the rate constants they determined were much lower than those we found for the QMDs, and they investigated a limited number of structurally similar compounds, it is possible that quinone methide reactivity is enhanced when part of a QMD. 18 When we calculated the TS for addition of a peroxyl radical to a model QMD (non-tert-butylated 2d, see Fig. 4A), we found a barrier of DG ‡ ¼ 16.0 kcal mol À1 , which corresponds to k add ¼ 9 Â 10 2 M À1 s À1 upon application of transition state theoryclose to the reported values for simple QMs, but about 4500-fold lower than the k inh values derived from the inhibited autoxidations. 19 Aside from this discrepancy between theory and experiment, the addition mechanism is fully consistent with the lack of substituent effects on the reactivity at ambient temperature and the absence of kinetic solvent effects and either O-H/O-D or C-H/C-D kinetic isotope effects.
To provide insight on the origin of the possibility of an enhancement of quinone methide reactivity when part of a dimer, we calculated the TS for the addition of a peroxyl radical to a simple QM for comparison to the calculated data for the model QMD (Fig. 4A). To our surprise, the reaction was predicted to be much faster (DG ‡ ¼ 13.0 kcal mol À1 , which was used to derive k add ¼ 1 Â 10 5 M À1 s À1 ). The discrepancy between the calculated rate constant for the reaction of 4a and the reported experimental values (e.g. 1.4 Â 10 3 and 5.8 Â 10 2 M À1 s À1 Table 1 Inhibition rate constants (k inh ) and stoichiometries (n) for substituted QMDs (2), stilbenoid phenols (1) and quadrangularin A analogues (3) measured from inhibited co-autoxidations of 1-hexadecene (2.9 M) and PBD-BODIPY (10 mM) initiated by AIBN (6 mM) at 37 C, t BuOO t Bu (87 mM) at 70 C, and dicumyl peroxide (1 mM) at 100 C at 60 C in styrene and cumene, respectively) 17 prompted us to synthesize it to corroborate Volodkin's results using our own approach and under the same conditions that the reactivity of the QMDs had been determined. Inhibited co-autoxidations of PBD-BODIPY in 1-hexadecene and of STY-BODIPY in cumene (both in chlorobenzene) yielded k inh ¼ 8.2 Â 10 4 and 7.2 Â 10 3 M À1 s À1 , respectively, at 37 C (see Fig. 4B for a direct comparison of the reactivity of 4a and 2d, tabulated in Table  2). 20 We believe that these values are slightly higher than those obtained by Volodkin due to differences in the experimental conditions employed. 21 Nevertheless, these results conrm that simple QMs are much less reactive than the QMDs. 22 Given that the observed reactivity of 4a was highly coincident with the corresponding hindered phenols, we considered that the simple QMs undergo hydration in situ to produce hindered phenols that are the active RTAs, and that the QMDs are simply more resistant to hydrationmaking them appear more reactive. Thus, we prepared the hydrated form of 4a and evaluated its reactivity in a 1-hexadecene/PBD-BODIPY co-autoxidation to nd nearly indistinguishable reactivity to that of 4a. To conrm that 4a and hydrated 4a react via different mechanisms, we carried out additional experiments wherein DMSO was added as a co-solvent. While the reactivity of 4a was unchanged, a signicant suppression in the reactivity of the hydrated 4a was observed, consistent with the sequestration of the phenolic Hatom as part of an H-bonded complex with DMSO (Fig. 4C). We were also able to monitor the QM chromophore (l max $ 300 nm) by UV-vis spectrophotometry and found that it disappeared steadily during the inhibited period of the autoxidation (Fig. 6D).
The conrmation of the lacklusterbut authentic -RTA activity of the quinone methide 4a suggests that there is something special about the QMD structure which confers greater reactivity compared to monomeric quinone methides. To provide further insight on this point, we carried out autoxidations inhibited by the tautomerized QMD precursor to quadrangularin A analogue 5a and an isomer thereof (5b)both of which contain one hindered phenol and one quinone methide moiety. Both compounds exhibited kinetics consistent with the phenolic moiety (refer to Table 2 and Fig. 4D), in agreement with the larger k inh of the phenols compared to the quinone methides.
The fourth mechanistic possibility (Fig. 1C) involves trapping of peroxyl radicals by combination with the small amount of persistent phenoxyl radical that is in equilibrium with the QMD (Fig. 5A). 23,24 We had previously shown that QMD 2e does not possess the weakest C-C bond (with a BDE of 17.0 vs. 6.1 kcal mol À1 ), 25 it is sufficiently weak that a relevant concentration of phenoxyl radical exists at ambient temperatures. Based upon the equilibrium constant we previously reported for 2e (K eq ¼ 5.5 Â 10 À10 M at 37 C), 24 the phenoxyl radical concentration at the beginning of an autoxidation inhibited by 1 mM of the QMD is $23 nM. If establishment of this equilibrium is fast relative to propagation of the autoxidation, then This value is in excellent agreement with ndings of Jonsson et al., who determined rate constants ranging from 1 Â 10 8 M À1 s À1 to 5 Â 10 8 M À1 s À1 for the reactions of various stabilized phenoxyl radicals with peroxyl radicals derived from i PrOH in water (the effects of nonviscous solvents are generally negligible on radical combination reactions). 26 Since we have no information on how substituents impact the QMD-phenoxyl radical equilibrium (Fig. 5A), we investigated it by UV-vis spectrophotometry at different temperatures (representative plot shown in Fig. 5B), enabling determination of the C-C BDEs (DH C-C , tabulated in the ESI †) from the resultant Van't Hoff plots (e.g. Fig. 5C). 27 The BDEs of a broader series of QMDs were also computed using DFT and dispersion-corrected DFT (see ESI †), and we nd that the trend in the sensitivity of DH C-C to the substituent is similar to that of the experimental values, with derivatives substituted with electrondonating groups possessing weaker bonds. These data imply that electron-donating substituents should enhance RTA activity, but this is not observed. In fact, the relationship between log k inh and DH C-C (Fig. 5D) shows the opposite trendat elevated temperatures, at least.
Nevertheless, to probe whether this pre-equilibrium is relevant, we synthesized a QMD bearing alkyl substituents in lieu of aryl substituents (6, structure shown in Fig. 5F) in the anticipation that it would have a stronger central C-C bond. Indeed, we could nd no evidence for phenoxyl radicals in the UV/vis spectra of the QMD up to 100 C even though the predicted spectrum of the radical suggests l max ¼ 325 nm (see  Table 2 Inhibition rate constants (k inh ) and stoichiometries (n) of QMs and phenols during inhibited co-autoxidations of cumene (3.6 M) and STY-BODIPY (10 mM) or hexadecene (2.9 M) and PBD-BODIPY initiated by AIBN (6 mM) in chlorobenzene at 37 C alongside calculated addition rate constants (k calc add , gas phase, 37 C) and literature k inh values (k lit inh , cumene or styrene, 60 C) ESI †). Furthermore, we were unable to identify crossover products when equivalent amounts of 2d and 6 were heated together (see ESI for spectral data †). Indeed, only poorly resolved (low intensity) spectra were obtained when concentrated samples of 6 (10 mM) were placed in the cavity of an EPR spectrometer, in contrast to the strong signals observed from samples of 2d (1 mM, see Fig. 5E), which can be readily identied as being fully consistent with the radicals derived therefrom (see ESI †). Nonetheless, integration of the spectra observed from samples of 6 yields K eq ¼ 4.2 Â 10 À13 M at 20 C. This can be directly compared with a value of K eq ¼ 1.7 Â 10 À10 M at 20 C that we obtained with 2d, and assuming DS is similar for C-C bond homolysis in 2d and 6, suggests that the difference in their C-C BDEs is at least 3.5 kcal mol À1 . Most importantly, this compound is a very poor RTA (k inh $ 2 Â 10 3 M À1 s À1 ) 28 compared to the other QMDs (see Fig. 5F) and is even several-fold worse than the simple QM 4a.

Discussion
We recently prepared QMD 2e as the key intermediate in the total synthesis of pallidol and quadrangularin Anatural products resulting from the oxidative dimerization of resveratrol. 6 We were very surprised to nd that 2e and related QMDs are potent RTAs (k inh $ 4 Â 10 6 M À1 s À1 at 37 C); ca. 10-fold more reactive than the phenols from which they are derived. In fact, the QMDs were similarly reactive to a-tocopherolthe most biologically active form of vitamin E and the standard to which all other RTAs are compareddespite lacking labile Hatoms, the key structural feature of canonical RTAs. The preparation of 2e and its subsequent stereoselective conversion to pallidol and quadrangularin A was enabled by its reversible fragmentation to two persistent phenoxyl radicals. 5,24 The foregoing mechanistic investigations suggest that this equilibrium is also responsible for the impressive RTA activity of 2e and related QMDs. Although the persistent phenoxyl radicals are present in small quantities at equilibrium (ca. 1% of the QMD under the conditions which were investigated here), their combination with peroxyl radicals is quite rapid (k inh $ 2 Â 10 8 M À1 s À1 at 37 C). Thus, one molecule of QMD can be expected to trap two peroxyl radicalsconsistent with the inhibition times observed in the QMD-inhibited autoxidations, from which n $ 2 were derived.
It is noteworthy that the resultant peroxidic adducts still contain QM moieties which can, in principle, react with additional peroxyl radicals. Indeed, simple QMs such as 4a react with peroxyl radicals with rate constants similar to those of the hindered phenols from which they are derived (e.g. 4a-H 2 O), k inh $ 8 Â 10 4 M À1 s À1 at 37 C. Since the initial addition of a peroxyl radical forms a phenoxyl that can subsequently combine with a peroxyl radical (as do the phenoxyl radicals in equilibrium with the QMDs), the simple QMs were also found to trap 2 peroxyl radicals (see Fig. 6B). Thus, in principle, QMDs should trap a total of 6 peroxyl radicals, as is shown in the overall mechanistic scheme in Fig. 6A.
The inhibited autoxidation data for the isomerized QMD structures 5a and 5b (Fig. 4D) provide some insight to why the QMDs only appear to trap 2 peroxyl radicals in our experiments. Each of these compounds, which feature one phenolic moiety and one QM moiety react as if only the phenol were present with n $ 2 and k inh ¼ 8.1 Â 10 4 M À1 s À1 and 6.4 Â 10 4 M À1 s À1 , respectively, which are subject to the same substantial KSEs. The lack of retardation of oxidation due to the QM moiety implies that it is signicantly less reactive to peroxyl radicals than a simple monomeric QM. Indeed, when we monitored the QMD-inhibited autoxidations by looking at the whole spectrum rather than simply l max of PBD-BODIPY at 588 nm, we noticed consumption of the characteristic absorbance of the QM moiety at $300 nm only aer the initial inhibited period, which coincided with consumption of the characteristic absorbance of the QMD moiety at $330 nm (Fig. 6C). This slower phase is consistent with the loss at $300 nm in autoxidations of the simple QM 4a (Fig. 6D). In an attempt to quantify the slower reactivity of the more hindered QM moieties in the peroxyl adducts derived from the QMD (or 5a or 5b), we carried out inhibited autoxidations of cumene in PhCl with much higher concentrations of QMD (up to 50 mM). Upon doing so, we could indeed see a retardation of autoxidation beyond the initial inhibited period (Fig. 6E), which yielded k inh $ 4 Â 10 3 M À1 s À1 on par with a k inh for a simple QM measured in cumene/PhCl at 37 C. Ultimately, the most compelling result in support of the dissociation/combination mechanism is the observation that QMD 6, which has a much stronger central C-C bond than 2, is >1000-fold less reactive as an RTA than 2. The reactivity of 6 is even lower than the simple QM 4a, implying some steric effects on the addition of peroxyl radicals. This is consistent with computational predictions 22 as well as the slower reactivity of the QM moieties in the tautomerized QMD precursor to quad A (5a) and isomer thereof (5b) relative to 4a.
The one piece of mechanistic data that is not, at rst glance, fully consistent with the dissociation/combination mechanism is the erosion in RTA activity with increasing temperature. Increasing the temperature shis the QMD/radical equilibrium toward phenoxyl radical formation, which should improve peroxyl radical-trapping, but instead only a slight increase is observed on going from 37 C to 70 C and then a noticeable drop on going from 70 C to 100 C. We believe this simply reects the reversible nature of the phenoxyl-peroxyl radical combination. It is well known that the activities of hindered phenolic RTAs diminish at elevated temperatures and that n values erode from 2 to 1 due to the reversible nature of the phenoxyl-peroxyl radical combination step that follows the initial H-atom transfer from phenol to peroxyl radical. 29 The same phenoxyl-peroxyl combination step features here. The fact that the erosion in reactivity is not as severe for the more electron-poor QMDs (cf. Fig. 5D) presumably reects a stronger C-O bond in the phenoxyl-peroxyl adduct.
The step-wise homolytic substitution mechanism of the QMDs is reminiscent of that by which the dimeric form of the Ciba (now BASF) antioxidant Irganox HP-136 reacts. The HP-136 dimer was determined to have largely solvent-independent RTA kinetics, 30 with k inh ¼ 4.3 Â 10 5 M À1 s À1 and n ¼ 0.8 from styrene autoxidations in chlorobenzene at 30 C. 31 The QMDs are therefore roughly 10-fold more reactive than the HP-136 dimer, presumably due to their weaker central C-C bond: 14-17 kcal mol À1 vs. 23 kcal mol À1 for (HP-136) 2 . Thus, the QMDs add to a short list of molecules which lack labile H-atoms but are nevertheless reactive RTAs. Two of the few other examples of which we are aware are tetrasuldes 32 and trisulde-1-oxides. 33 These compounds undergo concerted (bimolecular) homolytic substitution by peroxyl radicals to produce stable and persistent perthiyl radicals that do not propagate the autoxidation, but combine to give tetrasuldes. Although these substitution reactions are also insensitive to solvent effects, they are much slower at ambient temperatures (k inh $ 10 3 and 10 4 M À1 s À1 for the tetrasuldes and trisulde-1-oxides, respectively, compared to k inh $ 10 6 M À1 s À1 for the QMDs at 37 C), limiting their applications to elevated temperatures.
The lack of solvent effects and high reactivity of the QMDs at ambient temperatures prompted us to carry out one additional set of experiments (see Fig. 7). It has recently come to light that the poor performance of phenolic antioxidants as inhibitors of lipid peroxidation in biological membranes is the result of strong H-bonding interactions between phenols and the phosphate diester moiety of the membrane phospholipids (Fig. 8A). 34 Indeed, the reactivity of a-tocopherol (a-TOH), Nature's premier lipophilic RTA, toward peroxyl radicals is suppressed almost 1000-fold on going from chlorobenzene to a phospholipid bilayer (k inh $ 10 6 vs. 10 3 M À1 s À1 ). As such, we wondered if QMDs would be effective RTAs in this context, and evaluated their ability to suppress (phospho)lipid peroxidation using our recently described FENIX assay. 34a Disappointingly, we found that they were quite poor inhibitors. Alas, the large microviscosity of the phospholipid bilayer 35 presumably suppresses cage escape and combination of the QMD-derived phenoxyl radicals with lipid peroxyl radicals (Fig. 8B). This result further supports a stepwise substitution by peroxyl radicals on the QMDs in contrast to the concerted substitution favoured for the tetrasuldes and trisulde-1-oxides, which do not depend on the viscosity of the medium. Alongside the foregoing experiment, we evaluated the RTA activity of a simple QM (4a) as a control. In fact, the QM turned out to be a much better RTA under these conditions, even besting a-TOH (see    7). 36 The apparent k inh value measured for 4a in the phospholipid bilayer is scarcely different from that obtained in chlorobenzenereinforcing that addition of peroxyl radicals is operative and medium independent (Fig. 8C). As far as we are aware, this is the rst instance of the inhibition of (phospho) lipid peroxidation by a non-canonical (HAT)-like RTA, and is compelling indirect evidence underlining the deleterious role of H-bonding on the reactivity of canonical RTAs.

Conclusions
This work establishes the mechanistic basis which underpins the surprisingly potent RTA activity of quinone methide dimers (QMDs) of the general formula 2. These dimers, which are formed by the oxidative coupling of hindered stilbenoid phenols, have previously served as synthetic precursors to resveratrol oligomer natural products. Despite lacking any conventional RTA motif (e.g. a phenolic O-H bond), these QMDs are not only reactive in that capacity, but best both the phenolic compounds from which they are derived and to which they are converted. Our work corroborates previous data which shows that monomeric quinone methides are modest RTAs at best, and shows that the diaryl QMD scaffold is privileged in that it opens up a new mode of reactivity via its reversible fragmentation. The resultant persistent phenoxyl radicals can combine with peroxyl radicals leaving the second quinone methide moiety to react with additional peroxyl radicals. This mode of reactivity is not inuenced by solvent, in contrast to traditional phenolic RTAs, whose H-atom transfer reactions are slowed by H-bonding interactions. This fact suggests that QMDs may be useful RTAs for applications in non-viscous H-bonding media. Experiments in phospholipid bilayers reveal that the high microviscosity of the lipid phase suppresses the reactivity of QMDs, but that the lacklustre reactivity of simple QMs, which undergo addition, is fully consistent with what is observed in solutionmaking it the only class of RTA reported to date whose solution phase reactivity directly translates to lipid bilayers.

Experimental section
General All chemicals and solvents obtained from commercial suppliers were used as received unless indicated otherwise. 1-Hexadecene and cumene were puried and stored as previously described. 8,32 The synthesis, purication, and characterization of substituted stilbenes (1), QMDs (2), and quad A analogues (3) as well as hybrid phenol-QMs (5 and 6) used in this study are described in previous reports. 5,6 The synthesis, purication, and characterization of 4a, 37 4a-H 2 O, 38 4b 39 and S4 40 were carried out according to literature reports.
4-((1S,2S)-6-(Benzyloxy)-2-(4-(benzyloxy)phenyl)-3-((E)-3,5-ditert-butyl-4-hydroxybenzylidene)-2,3-dihydro-1H-inden-1-yl)-2,6-di-tert-butylphenol (S2). Compound S1 (120 mg, 0.145 mmol) was dried down into a ame-dried round bottom ask charged with a stir bar. The atmosphere was evacuated and replaced with N 2 , and the starting material was dissolved in CH 2 Cl 2 (14 mL, 0.01 M reaction concentration). The solution was cooled to the reaction temperature, and BF 3 $OEt 2 (0.036 mL, 0.29 mmol, 2 equiv.) was added dropwise. The reaction was stirred for 3 hours, at which point it was raised from the ice bath and quenched via the addition of saturated NaHCO 3 . Once the reaction had thawed, it was poured into a separatory funnel, and the layers were separated. The aqueous layer was extracted with additional portions of CH 2 Cl 2 , and the combined organic layers were washed with brine, dried over magnesium sulfate, and concentrated under reduced pressure. The crude product was puried by column chromatography (10% to 50% DCM/ hexanes) to afford compound S2 (100 mg, 83% yield). 1   (2S,3S)-1-((E)-3,5-Di-tert-butyl-4-hydroxybenzylidene)-3-(3,5di-tert-butyl-4-hydroxyphenyl)-2-(4-hydroxyphenyl)-2,3-dihydro-1H-inden-5-ol (3b). Compound S2 (73 mg, 0.088 mmol) was dried down in a ame-dried round bottom ask charged with a stir bar. Pentamethylbenzene (134 mg, 0.88 mmol) was added in a single portion, and the solids were dissolved in dichloromethane (9 mL, 0.01 M reaction concentration) under an inert atmosphere. The reaction mixture was cooled to À78 C, at which point BCl 3 (0.53 mL, 0.53 mmol, 1.0 M in CH 2 Cl 2 ) was added via syringe, turning the reaction mixture deep purple. The reaction was stirred at this temperature for 1.5 h, at which point it was lied from the dry ice bath and quenched with saturated NaHCO 3 . The reaction was stirred vigorously while the ice thawed and until the reaction mixture stopped changing colour. Once the quench was complete (at this point the reaction was a pale-yellow colour), the reaction was poured into a separatory funnel containing DI H 2 O. The layers were separated, and the aqueous layer was extracted with additional CH 2 Cl 2 . The combined organic layers were washed with brine, dried over MgSO 4 and concentrated under reduced pressure. The crude material was puried by ash chromatography using a 0 to 15% acetone in DCM gradient to afford compound 3b (49 mg, 86% yield). 1  2,6-Di-tert-butyl-4-(4-(3,5-di-tert-butyl-4-hydroxycyclohexa-2,5-dien-1-ylidene)-2,3-dimethylbutylidene)cyclohexa-2,5-dien-1-one (6). Phenol S4 (25 mg, 0.100 mmol) was added to a 10 mL reaction ask charged with a stir bar and KPF 6 (74 mg, 0.4 mmol, 4.0 equiv.). The solids were dissolved in MeCN (8 mL) and 2,6-lutidine (12 mL, 0.1 mmol, 1.0 equiv.) was added to the reaction solution. See ESI † for details regarding the electrochemical setup. The reaction was stirred at 750 rpm for 1 h at a constant voltage of 0.8 V. Upon completion of the reaction (as judged by TLC), the electrodes were removed and rinsed into a collection ask with DCM ($40 mL). The contents of the reaction vial were also rinsed into the collection ask. The solvent was removed under reduced pressure, the crude material was resuspended in DCM, and the electrolyte was ltered away with a plug of Celite. The ltrate was then concentrated and puried by column chromatography (1% to 10% EtOAc/ hexanes) to afford compound 6 as a yellow foam (23.8 mg, 95% yield, $7 : 1 dr). 1 (25 mL). The resulting solution was heated at reux for 4 hours aer which it was cooled to room temperature and quenched with aqueous acetic acid and diluted with ether. The organic phase was washed with water and brine, dried over MgSO 4 , ltered and concentrated in vacuo. The crude yellow oil was puried by ash column chromatography using 10% EtOAc in hexanes as the mobile phase to yield the nal product as a light beige oil (0.320 g, 60%). 1  (E)-2,6-Di-tert-butyl-4-(2-(phenyl-d 5 )vinyl-2-d)phenol (1d-d 6 ). p-Toluenesulfonic acid monohydrate (5.7 mg, 0.03 mmol) was added to a solution of 2,6-di-tert-butyl-4-(1-hydroxy-2-(phenyld 5 )ethyl-2,2-d 2 )phenol (200 mg, 0.60 mmol) in benzene (12.0 mL) under nitrogen and the solution was heated to reux (85 C) for 4 hours. Aer it was cooled to room temperature, the solution was diluted with ethyl acetate (50 mL) and washed with water and brine, dried over MgSO 4 , ltered and concentrated in vacuo. The crude yellow oil was puried by ash column chromatography using hexanes as the eluent to yield the nal product as a white solid (0.130 g, 69%). 1  4,4 0 -(2,3-Bis(phenyl-d 5 )butane-1,4-diylidene-2,3-d 2 )bis(2,6di-tert-butylcyclohexa-2,5-dien-1-one) (2d-d 12 ). 23 A solution of stilbene 1d-d 6 (25 mg, 0.080 mmol) in dry THF (0.9 mL) was cooled to 0 C in an ice bath and purged with nitrogen for 5 minutes. KHMDS (0.090 mmol, 1 M in THF, 0.09 mL) was added slowly and the resulting bright yellow solution was stirred for 10 minutes at 0 C. Ferrocenium hexauorophosphate (28 mg, 0.090 mmol) was added in two 14 mg portions within a 15 minute interval. The resulting orange solution was stirred at 0 C for 1 hour aer which it was ltered through a pad of Celite, concentrated in vacuo and then puried by ash column chromatography using a gradient of 95 : 2.5 : 2.5 to 90 : 5:5 hexanes/ EtOAc/DCM to obtain the nal product as a yellow solid (19 mg, 76%). 1  AIBN (6 mM) at 37 C, t BuOO t Bu (87 mM) at 70 C, or dicumyl peroxide (1 mM) at 100 C. A 3.5 mL quartz cuvette was charged with PhCl (0.44 mL) and 1-hexadecene (2.00 mL). The cuvette was preheated to the desired temperature in a thermostated sample holder of a UV-vis spectrophotometer and allowed to equilibrate for approximately 15 min. To the cuvette was added PBD-BODIPY (12.5 mL of a 2.00 mM stock solution in 1,2,4-trichlorobenzene) and initiator (50 mL of a 300 mM stock solution of AIBN in chlorobenzene, 40 mL of neat t BuOO t Bu, or 50 mL of a 50 mM stock solution of dicumyl peroxide in chlorobenzene). The solution was thoroughly mixed prior to monitoring the uninhibited co-autoxidation via the disappearance of PBD-BODIPY at 588 nm (37 C), 587 nm (70 C), or 586 nm (100 C) for 5-10 min to ensure the reaction was proceeding at a constant rate. Finally, the antioxidant under investigation was added (5.0-10.0 mL of a 0.25 or 2.5 mM solution in chlorobenzene), the solution was mixed thoroughly, and the absorbance readings were resumed. The resulting Abs vs. time data were processed as previously reported. 8 The rate of initiation (R i ¼ 1.30 Â 10 À9 M s À1 (37 C), 1.26 Â 10 À9 M s À1 (70 C), 6.58 Â 10 À9 M s À1 (100 C)) and second order rate constant for propagation of the dye (k PBD-BODIPY ¼ 3792 M À1 s À1 (37 C), 7633 M À1 s À1 (70 C), 8283 M À1 s À1 (100 C)) necessary to compute stoichiometric data (n) and inhibition rate constants (k inh ) were determined using 2,2,5,7,8-pentamethylchromanol (PMC) as a standard, which has an established stoichiometry of 2. 42 Similar experiments were conducted at 37 C employing cumene (3.6 M) and STY-BODIPY (10 mM) initiated by AIBN (6 mM) in chlorobenzene. Reaction progress was monitored at 571 nm. The rate of initiation, also determined using PMC, was measured to be R i ¼ 2.28 Â 10 À9 M s À1 and the second order rate constant for propagation had been determined previously (k STY-BODIPY ¼ 141 M À1 s À1 ) (37 C). Inhibited autoxidation experiments involving liposomes were conducted at 37 C employing egg-phosphatidylcholine liposomes (1 mM), STY-BODIPY (10 mM), initiated by DTUN 34a (0.2 mM) in PBS buffer (10 mM). Reaction progress was monitored at 565 nm. The rate of initiation, determined using PMC, was measured to be R i ¼ 2.29 Â 10 À9 M s À1 and the second order rate constant for propagation had been determined previously (k STY-BODIPY ¼ 894 M À1 s À1 ) (37 C).
Determination of kinetic solvent effects (KSEs) in inhibited co-autoxidation experiments. Autoxidations of 1,4-dioxane (2.9 M) and PBD-BODIPY (10 mM) at 37 C were initiated by AIBN (6 mM) in chlorobenzene. A 3.5 mL quartz cuvette was charged with 1,4-dioxane (0.620 mL) and either PhCl (1.790 mL) or PhCl and DMSO (1.180 mL and 0.620 mL, respectively). The proceeding steps are the same as those described above but where only AIBN (50 mL of a 300 mM stock solution in chlorobenzene) is used to initiate the co-autoxidation which was monitored at 588 nm (3 ¼ 123 000 M À1 cm À1 in PhCl, 118 200 M À1 cm À1 in 2 : 1 PhCl/DMSO). The rate of initiation in PhCl (R i ¼ 2.40 Â 10 À9 M s À1 ) and in 2 : 1 PhCl/DMSO were previously standardized by PMC. The second order rate constants for propagation are k PBD-BODIPY ¼ 5310 M À1 s À1 (PhCl) and 5900 M À1 s À1 (2 : 1 PhCl/DMSO). 15 Determination of solvent kinetic isotope effects (SKIEs) in inhibited co-autoxidation experiments. Autoxidations of 1,4dioxane were conducted as described above but where the 3.5 mL quartz cuvette was charged with 1,4-dioxane (0.620 mL), PhCl (1.770 mL), and 0.025 mL of MeOL (L ¼ H or D). The experiments containing MeOL were conducted in competition and each set was determined in triplicate.
Determination of QMD C-C bond dissociation enthalpies (DH). 24 To a 3.5 mL quartz cuvette was added 2.475 mL of 1,2dichlorobenzene and 0.025 mL of a 5 mM stock of QMD in 1,2dichorobenzene. The cuvette was affixed with a rubber septum and the contents were purged with N 2 for 10 min. A scan on the UV-vis spectrophotometer was acquired using a cuvette containing only solvent and used to record baseline-corrected spectra throughout the experiment. The cuvettes containing the QMDs equilibrated to the set temperature for at least 5 min before recording the spectra. The measurements were conducted in triplicate.

EPR spectra
Electron paramagnetic resonance (EPR) spectra were recorded using a Bruker EMXplus (X-band) spectrometer equipped with an ER 4119HS cavity at 20 C. The samples were 0.1-10 mM in benzene and degassed (3 cycles of freeze-pump-thaw) and placed under an atmosphere of N 2 prior to acquisition. The radical concentration was determined using the quantitative EPR package of the Bruker Xenon soware. Spectral simulations were performed using EasySpin. 43 Calculations Calculations were conducted using the B3LYP method 44 (CBSB7 basis set) and CBS-QB3 (complete basis set) method 45 as implemented in the Gaussian 16 suite of programs. 46 Rate constants were calculated using transition state theory at 37 C. Dispersion corrections were applied using Grimme's D3 approach. 47 Hyperne coupling constants were predicted by spin density distributions at the B3LYP/TZVP level of theory. 48

Conflicts of interest
There are no conicts to declare.