The Chemical Basis of Thiol Addition to Nitro-conjugated Linoleic Acid, a Protective Cell-signaling Lipid*♦

Nitroalkene fatty acids are formed in vivo and exert protective and anti-inflammatory effects via reversible Michael addition to thiol-containing proteins in key signaling pathways. Nitro-conjugated linoleic acid (NO2-CLA) is preferentially formed, constitutes the most abundant nitrated fatty acid in humans, and contains two carbons that could potentially react with thiols, modulating signaling actions and levels. In this work, we examined the reactions of NO2-CLA with low molecular weight thiols (glutathione, cysteine, homocysteine, cysteinylglycine, and β-mercaptoethanol) and human serum albumin. Reactions followed reversible biphasic kinetics, consistent with the presence of two electrophilic centers in NO2-CLA located on the β- and δ-carbons with respect to the nitro group. The differential reactivity was confirmed by computational modeling of the electronic structure. The rates (kon and koff) and equilibrium constants for both reactions were determined for different thiols. LC-UV-Visible and LC-MS analyses showed that the fast reaction corresponds to β-adduct formation (the kinetic product), while the slow reaction corresponds to the formation of the δ-adduct (the thermodynamic product). The pH dependence of the rate constants, the correlation between intrinsic reactivity and thiol pKa, and the absence of deuterium solvent kinetic isotope effects suggested stepwise mechanisms with thiolate attack on NO2-CLA as rate-controlling step. Computational modeling supported the mechanism and revealed additional features of the transition states, anionic intermediates, and final neutral products. Importantly, the detection of cysteine-δ-adducts in human urine provided evidence for the biological relevance of this reaction. Finally, human serum albumin was found to bind NO2-CLA both non-covalently and to form covalent adducts at Cys-34, suggesting potential modes for systemic distribution. These results provide new insights into the chemical basis of NO2-CLA signaling actions.

Nitroalkene fatty acids are endogenous adaptive signaling mediators formed in vivo upon addition of nitric oxide (NO ⅐ )-or nitrite (NO 2 Ϫ )-derived nitrogen dioxide (NO 2 ⅐ ) to unsaturated fatty acids (1)(2)(3). The presence of the electronwithdrawing nitro group renders the nitroalkene ␤-carbon electron-deficient and thus susceptible to attack by nucleophiles in reversible Michael addition reactions (4). The electrophilicity of nitroalkene fatty acids is critical to the biological actions of these molecules, as demonstrated by the inhibitory effects of increased nitroalkene reductase activity (5). Nitroalkene fatty acids exhibit potent anti-inflammatory and cyto-protective properties and thus are beneficial in many models of disease, including atherosclerosis, restenosis, ischemia reperfusion, renal injury, diabetes, metabolic syndrome, and endotoxemia (6 -13). Furthermore, endogenous formation of nitroalkene fatty acids has been associated with the cardioprotective effects of the Mediterranean diet (14,15). Notably, the potential application of soft electrophiles as pharmacological agents is underscored by the Food and Drug Administration approval of dimethyl fumarate for the treatment of multiple sclerosis (16,17).
Thiols are excellent nucleophiles and are able to react with different electrophiles, including oxidants. This property constitutes the basis of the fundamental roles of particular thiols in signaling, detoxification, and antioxidant response processes. For most chemical and enzymatic reactions, thiol reactivity involves the nucleophilic attack of the thiolate (RS Ϫ ) on the electrophile. The intracellular compartment presents an elevated concentration of reduced thiols. Glutathione (GSH) is the main low molecular weight thiol in the cytosol (2-17 mM), and protein thiols represent ϳ70% of the total reduced intracellular pool (18 -21). In contrast, plasma has much lower total thiol concentrations that, in addition, are predominantly oxidized.
Low molecular weight plasma thiols include cysteine, cysteinylglycine, GSH, homocysteine, and ␥-glutamylcysteine, which together constitute a total of 12-20 M reduced thiol. The most abundant thiol in this compartment is cysteine 34 in human serum albumin (HSA) 5 which is ϳ600 M and 75% reduced (22).
From a kinetic, mechanistic, and structure-reactivity relationship aspect, Michael addition and ␤-elimination reactions between various activated olefins and nucleophiles have received considerable attention in the literature. Depending on the nucleophilic and electrophilic partners as well as on the reaction conditions, two types of mechanisms have been proposed, concerted or stepwise, with the latter involving intermediate carbanions (23)(24)(25)(26).
Conjugated linoleic acid (CLA) is a preferential substrate for biological nitration leading to the formation of 9-and 12-nitrooctadecadienoic acid (9-and 12-NO 2 -CLA) (27). Nitroalkene derivatives of CLA are present in plasma and urine of healthy individuals with and without CLA supplementation and are generated during digestion, metabolic stress, and inflammation (1,2,27). The reaction between thiols and NO 2 -CLA has implications for the modulation of not only signaling actions but also circulating levels and bio-elimination pathways (2,6,28).
Importantly, in the case of NO 2 -CLA, both the ␤and ␦-carbons with respect to the nitro group are, in principle, electrophilic and thus susceptible to reaction with nucleophiles.
Herein, we provide an experimental and computational study of the reaction of NO 2 -CLA, the most abundant endogenous nitroalkene, with biological thiols. We found that two reversible products are formed, the ␤-adducts and the ␦-adducts. The ␤-adducts are formed with faster kinetics, but the ␦-adducts are more stable. The reaction mechanisms are stepwise, with thiolate attack on the nitroalkene being the ratecontrolling step. In addition, HSA is able to bind NO 2 -CLA non-covalently. Our findings contribute to the understanding of the chemical basis of the signaling and pharmacological actions of nitroalkene fatty acids.

NO 2 -CLA Reacts Biphasically and Reversibly with Low
Molecular Weight Thiols-Stopped-flow mixing of a solution consisting of 9-and 12-NO 2 -CLA ( Fig. 1A) with excess GSH led to a reduction in absorbance at 330 nm, consistent with loss of double bond conjugation. The decrease was biexponential and included a fast phase that lasted ϳ0.5 min followed by a second phase that was slower by a factor of ϳ30 (Fig. 1B). No significant changes in absorbance occurred in the absence of thiols. Although the NO 2 -CLA stock contains an ϳ1:1 mixture of the positional isomers 9-and 12-NO 2 -CLA, the existence of two phases could not be explained by a differential reactivity of the isomers, as purified 9-and 12-NO 2 -CLA yielded comparable biphasic kinetics (Fig. 1B, inset). In contrast, the reaction with the monounsaturated derivative nitro-oleic acid (NO 2 -OA), 5  which has only one electrophilic site, was monophasic ( Fig. 1C) (29). When NO 2 -CLA was mixed with the yellow thiol thionitrobenzoate (TNB), which allows us to follow changes in thiol concentration due to its absorbance at 412 nm, the kinetics were also biphasic, confirming that thiol consumption occurred in both phases (Fig. 1D). Regarding the concentration dependence of the reaction between NO 2 -CLA and excess GSH, a linear dependence with a non-zero y axis intercept was observed between the pseudo-first order rate constant (k obs ) of the fast phase and GSH concentration (Fig. 1E), while a hyperbolic dependence with a non-zero intercept was observed for the slow phase (Fig. 1F). Taken together, the kinetic results are consistent with two parallel and reversible processes involving two non-equivalent electrophilic centers in NO 2 -CLA. These centers react with thiols forming two products that we hypothesize are the ␤and the ␦-adducts as shown in to Scheme 1, where k on␤ and k on␦ are second-order rate constants at pH 7.4 for the forward addition reaction of the fast and slow processes, respectively, and k off␤ and k off␦ correspond to first-order rate constants for the reverse elimination reactions. 6 The coupled differential equations derived from Scheme 1 can be solved in matrix form yielding complex biexponential concentration functions. The relation of the exponential constants k obs with the rate constants and concentrations is simplified when one phase is faster than the other (30,31). In this case, the larger exponential constant, k obs(fast) , increases linearly with thiol concentration according to Equation 1 (Fig. 1E).
For GSH reactions, k on␤ ϭ 34 Ϯ 4 M Ϫ1 s Ϫ1 was determined from the slope of the plot, whereas k off␤ ϭ 0.10 Ϯ 0.02 s Ϫ1 was obtained from the y axis intercept. The equilibrium dissociation constant (K eq␤ ) was calculated by dividing k off␤ by k on␤ and its value was (2.8 Ϯ 0.9) ϫ 10 Ϫ3 M (25°C, pH 7.4).
The smaller exponential constant, k obs(slow) , is given by the sum of the forward and reverse effective rate constants, where the former is k on␦ [thiol] multiplied by the fraction of free NO 2 -CLA, and it increases hyperbolically with thiol concentration according to Equation 2 (Fig. 1F).
From fits to this equation, the k on␦ for the reaction with GSH was determined to be 3.5 Ϯ 0.5 M Ϫ1 s Ϫ1 whereas K eq␤ was (3.0 Ϯ 0.2) ϫ 10 Ϫ3 M, in excellent agreement with the value determined from the fast phase. The k off␦ was obtained using data at relatively low thiol concentrations and was (3 Ϯ 1) ϫ 10 Ϫ4 s Ϫ1 . The K eq␦ was (9 Ϯ 4) ϫ 10 Ϫ5 M.
Similar determinations were carried out for cysteine, homocysteine, cysteinylglycine, and ␤-mercaptoethanol (BME), and the rate and equilibrium constants are shown in Table 1. The rate constants were higher for the presumed ␤-adducts in all cases. The equilibrium constants were lower by factors of 20 -65 for the ␦-adducts, reflecting higher stability. It is worth mentioning that at higher cysteine concentrations (30 mM) a third slow phase became evident (k obs ϭ (1.33 Ϯ 0.01) ϫ 10 Ϫ3 s Ϫ1 ), which was not further characterized. No reaction was observed with histidine (Յ40 mM), suggesting that additions of ␣-amino and imidazole groups were not significant, consistent with nitrogenous bases being weaker nucleophiles toward Michael acceptors than thiolates (23,26).
NO 2 -CLA Has Two Non-equivalent Electrophilic Carbon Centers in C ␤ and C ␦ -The electrophilic Fukui f ϩ (r) function, which measures the propensity of NO 2 -CLA to gain electron density in a nucleophilic attack, is shown in Fig. 2, mapped on a total electronic density isosurface calculated at the PCM-DFT level in aqueous solution. Two well defined and distinct soft electrophilic regions (Fig. 2, depicted in blue) involving the C ␤ and C ␦ centers are found in both isomers. Whereas the region around C ␤ appears extended over the molecular surface and connected to the electrophilic nitro group envelope, the region enclosing C ␦ is considerably smaller. These features would make C ␤ more prone to nucleophilic attack by thiols. Atomic softness data calculated from global softness and condensed Fukui f A ϩ ( Table 2) completes the picture enabling a quantitative comparison, showing that C ␤ is significantly softer than C ␦ , establishing the non-equivalency of the electrophilic C ␤ and C ␦ centers and contributing to consolidate the hypothesis of their association with the fast and slow reactions, respectively.
The ␤-Adduct Is the Kinetic Product and the ␦-Adduct Is the Thermodynamic Product-To define the products generated in each phase of the reaction, spectrophotometric, LC-UV-Visible, and LC-MS/MS experiments were performed. First, the time course of the reaction with GSH was followed at three different wavelengths. During the fast phase, the decrease in absorbance at 330 nm correlated with an increase at 250 nm, although no changes were observed at 290 nm. In the slow phase, however, the decrease at 330 nm correlated with an increase at 290 nm (Fig. 3A). These results suggest that the product formed during the slow reaction contains an intact nitroalkene moiety due to thiol addition at the ␦-carbon (29,32) whereas that formed during the fast reaction does not.
LC-UV-Visible analyses were performed using BME as a model thiol due to its slower addition kinetics and better chromatographic separation (33, 34). 9-NO 2 -CLA eluted at a retention time of 15.5 min (peak 1, max 320 nm) (Fig. 3, B and C).
The peak at 15.1 min (peak 1*) corresponds to contaminant 12-NO 2 -CLA (data not shown). In the first 10 s of the reaction with BME, a peak appeared at 8.8 min (peak 2, max 240 nm) (Fig. 3, B, inset, and C) that was still present after 5 min but disappeared after 1 h due to the reverse elimination process and further reactions of NO 2 -CLA, and it was thus assigned to the fast reaction product (␤-addition). After 5 min of reaction, another peak appeared at 8.2 min (peak 3, max 280 nm) (Fig.  3B), became dominant as the reaction progressed, and was assigned to the ␦-addition product. This peak was not symmetric, suggesting the presence of two diastereomers, both with maximum absorbances at 280 nm (Fig. 3C). The secondary peaks at 8.6 (peak 2*) and 7.7 min (peak 3*) were assigned to the products of the fast and slow processes for the reaction between contaminant 12-NO 2 -CLA and BME.
To obtain further mechanistic insight, the reaction between purified 9-NO 2 -CLA and BME was monitored by LC-MS/MS. The disappearance of this product coincides with the formation of two isobaric species (MRM 402/324, retention times: 6.5 and 6.8 min, ␦ 1 -and ␦ 2 -adducts, respectively) that become the dominant products during the slow phase of the reaction and likely represent diastereomers as further discussed below (Figs. 6C and 7B). To confirm the relative rates of elimination from the adducts formed during the different phases of the reaction, aliquots were incubated with excess NEM (100 mM, 30 min) to efficiently trap BME (k ϭ 7 ϫ 10 4 M Ϫ1 s Ϫ1 at pH 7.4 (35)) ( Fig. 4, C and D). Consistent with the measured k off␤ and k off␦ values (Table 1), NEM incubation completely reversed adducts formed at early time points but had no effect on the late products. Similar results were obtained with the purified 12-NO 2 -CLA isomer, with the particularity that only one ␦-addition product could be resolved to such an extent as to allow reliable peak integration (supplemental Fig. S1).
Overall, LC-MS/MS results and UV-Visible analysis were in excellent agreement, and it was concluded that the fast reaction corresponded to the ␤-addition of the thiol and the slow reaction to the ␦-addition (Scheme 1). No evidence for the formation of other products was obtained. The results are consistent with a fast reaction of NO 2 -CLA with the thiol to form initially the ␤-adduct. Because of the reversibility of this reaction, the ␤-adduct undergoes elimination and releases free NO 2 -CLA, which engages in further reactions finally leading to the slow accumulation of the more stable ␦-adduct.
Addition Requires a Thiolate and Elimination Occurs through Two Independent Pathways-The pH dependence of the reaction kinetics of NO 2 -CLA with GSH was studied using threecomponent buffers of constant ionic strength. Plots of k obs versus GSH concentration were constructed at each pH, from which the corresponding apparent k on, app and k off, app values were obtained for both fast and slow processes. For the fast reaction, the k on␤, app increased with pH ( Fig. 5A) according to a single pK a equation (Equation 3),

Fast reaction (␤) Slow reaction (␦)
a A mixture of 9-and 12-NO 2 -CLA was used unless otherwise specified. b Thiol pK a values reported in Ref. 36.

TABLE 2 Atomic softness s A ؉ for the carbon centers in the conjugated nitroalkene moieties embedded in each NO 2 -CLA regioisomer (in au)
Data were determined according to Equation 10. Positive values correspond to atoms more prone to receive electrons, acting as electrophilic centers. The softest and most reactive C ␤ site in each isomer is underlined.

Regioisomer
Carbon center where k on, pH-indep represents the rate constant for the completely ionized thiol. The pK a was 8.78 Ϯ 0.02, in agreement with the reported pK a of GSH (8.94 (36)) and consistent with the requirement for rapid prior ionization of the thiol to thiolate in the forward addition process. The k off␤, app presented an upward bend indicating that the elimination process occurs through two independent pathways ( Fig. 5B) (37). At biologically relevant pH values, elimination appears to be unimolecular or water-assisted. The increase above pH 9 likely reflects the assistance of alternative bases. Similar pH dependences were observed for the slow phase (data not shown), suggesting that comparable mechanisms are operative.
Correlations with Thiol pK a Suggest That the Thiolate Participates in the Rate-controlling Step-Because the thiolate is the reactant in the forward addition process, the k on, app values at pH 7.4 (Table 1) were corrected according to Equation 3 to obtain pH-independent values (k on, pH-indep ), that reflect the intrinsic reactivity of each thiolate toward NO 2 -CLA. For the fast reaction, the log of k on, pH-indep increased with thiol pK a according to Equation 4, where ␤ nuc is the Brønsted nucleophilic coefficient, and C on is a constant. According to the slope and the y axis intercept of the plot, ␤ nuc was 0.64 Ϯ 0.08, and C on was Ϫ2.8 Ϯ 0.7 (Fig. 5C, top trace).The ␤ nuc value of 0.64 indicates that thiolate nucleophilicity correlates with proton basicity and is consistent with thiolate participation in transition state formation. It also suggests a relatively high degree of charge transfer at the transition state level. For comparison, ␤ nuc values of 0.45 and 0.16 were reported for the addition of thiols to acrylonitrile (26) and ␣-nitrostilbene, respectively (23).
For the reverse reaction (pH Ͻ9), inverse correlations between the log k off and thiol pK a were obtained, consistent with Equation 5,  where ␤ lg is the Brønsted leaving group coefficient, and C off is a constant. The values of ␤ lg and C off were determined from the slope and y axis intercept of the plot to be Ϫ0.73 Ϯ 0.12 and 5.3 Ϯ 1.1, respectively (Fig. 5C, bottom trace). The ␤ lg value of Ϫ0.73 indicates that the reactivity of the adduct correlates with the proton acidity of the thiol that is eliminated. This is consistent with a rate-controlling step that involves thiolate departure and partial charge formation, with considerable amount of C-S bond breaking in the transition state, in agreement with a reversal of the mechanism proposed for the forward reaction. Values of ␤ lg of Ϫ0.68 and Ϫ0.54 were reported for the elimination of thiols from ␣-nitrostilbene and acrylonitrile adducts, respectively, with leaving group expulsion argued to be the rate-controlling step in the latter case (23,25). For the slow reaction, similar trends were observed, with ␤ nuc ϭ 0.6 Ϯ 0.2, C on ϭ Ϫ4 Ϯ 2, ␤ lg ϭ Ϫ0.72 Ϯ 0.09, and C off ϭ 2.7 Ϯ 0.8 (Fig. 5D).
Lack of Solvent Kinetic Isotope Effects-The reaction of 20 M NO 2 -CLA with 4 mM GSH exhibited k obs of 0.23 Ϯ 0.03 and (6.4 Ϯ 0.2) ϫ 10 Ϫ3 s Ϫ1 for the fast and slow processes, respectively. When H 2 O was replaced by D 2 O (92%) and the pD was adjusted for similar GSH ionization fractions, k obs values were 0.25 Ϯ 0.04 and (6.1 Ϯ 0.3) ϫ 10 Ϫ3 s Ϫ1 , indicating the absence of significant deuterium solvent kinetic isotope effects. This rules out rate-controlling steps that involve protonation and suggests that concerted addition-elimination processes are not involved. Instead, the reactions likely occur through stepwise mechanisms involving anionic intermediates and thiolate attack as the rate-limiting step, in agreement with the Brønsted correlations shown in  Table S1). No evidence was found here for a concerted four-membered ring TS as reported for the attack of thiols on ␣,␤-unsaturated carbonyls (38). Both TS ␤ /TS ␦ resemble the open structures found for thio-Michael addition on that kind of acceptor (39,40), taking place after fast deprotonation of the thiolate in solution.
Reactions at C ␤ and C ␦ proceed both through entropically disfavored loose reactant complexes (RC ␤ /RC ␦ , Fig. 6, A and B) leading to transition states (TS ␤ /TS ␦ ) with a more asymmetric distribution of charge in the acceptor and associated free-energy barriers of 18.9 and 20.2 kcal/mol, respectively, calculated with respect to the isolated reactants in solution (Table 3). This supports the assignment of ␤-/␦-addition processes, respectively, to the fast/slow phases observed in the experiments. In fact, a difference of 1.3 kcal/mol in activation free-energy translates into a 9-fold increase in the k on rate for ␤-addition with respect to ␦-addition, in excellent agreement with the kinetic experiments (Table 1). Charge transfer of 0.33 and 0.28 atomic units (au) between reacting moieties is respectively found at TS ␤ and TS ␦ (mainly accommodated at the NO 2 group) reflecting the similarity between ␤ nuc parameters for both fast and slow NO 2 -CLA reactions. The R1/R2 substituents at the conju-  Table  1; a, cysteinylglycine; b, cysteine; c, GSH; d, homocysteine; e, BME. . Atoms are colored by element, and a selection of relevant bond lengths, WBI, and NPA atomic/group charges labeled as q X (X represents an atom or group of atoms) featuring geometrical and electronic reorganization along each reaction channel is highlighted in proximity to the structures. Net charge transfer between reactants is evidenced at each TS. C, relevant properties toward I ␤ /I ␦ nitronate protonation: MEP mapped on a total electron density surface of 0.004 au (notice that the underlying structures retain the orientation shown immediately above in A and B for each species), NPA atomic charges, and proton affinities (PA) at each O/C protonation site. All properties calculated at the PCM(IEF)-B97X-D/6 -31ϩG(d,p) level in aqueous solution. MEP coloring scheme spans from Ϫ0.245 (red) to Ϫ0.04 au (cyan-blue). Although electrostatics favor a faster protonation at NO 2 oxygens in both cases, C ␣/␥ targets lead to more stable products, displaying mostly anti stereochemistry due to steric restrictions, as evidenced in red labels. gated nitroalkene moiety (Scheme 1) modulate its intrinsic reactivity toward thiolate attack (23,26). In particular, for nitroalkene fatty acids the alkyl/carboxyl(ate) chains are expected to increase the reactivity toward thio-Michael addition with respect to the model system studied here, without altering the mechanism 7 leading to a more pronounced charge reorganization between partners at the C ␤ /C ␦ adduction TSs, closer to the expectations from the ␤ nuc values derived from experimental data. A similar rationale applies to bond forming/ breaking at each TS, for which Wiberg bond indices (WBIs) of 0.27/0.15 for the nascent S⅐⅐⅐C ␤ /S⅐⅐⅐C ␦ bonds imply advances of 28 or 16% at each reaction involving the model compounds. A lag in S-C ␦ bond formation with respect to the advance in charge transfer from thiolate to nitroalkene moieties (not found in TS ␤ ) may explain the higher barrier found for the slower reaction leading to the ␦-adduct (41).
Concerning the anionic intermediates (I ␤ /I ␦ , Fig. 6), they both would be nitronate species, sharing a strengthened C ␣ -N bond (WBIs: 1.48 and 1.32 au, respectively) and a net charge of Ϫ1 au localized at the NO 2 moiety in their structures. Whereas the second NO 2 -CLA unsaturation remains intact at C ␥ ϭC ␦ in I ␤ , it appears shifted into C ␥ ϭC ␤, in I ␦ enabling a more extended electron delocalization that translates in a more stable ␦-intermediate (Table 3). No evidence of formation of any stable carbanion was found. The NO 2 group progressively captures all the electronic density transferred between reactants up to completion of each of these two parallel adduction processes, resulting in completely formed C-S single bonds (1.85-1.86 Å and WBIs 0.94 au) and loss of planarity at C ␤ /C ␦ . Thus, whereas a thermodynamically disfavored nitronate I ␤ is obtained through a faster and slightly endoergic (but exothermic) process, the more stable I ␦ would be the prevalent outcome of the adduction step in the longer timescales. Reverse free-energy reaction barriers in the range of 15-22 kcal/mol ( Table 3) are indicative of reversible processes leading in both cases to elimination of a thiolate, more facilitated from the ␤-intermediate (once again a result qualitatively in line with the relationship between k off values obtained for the fast/slow processes).
As a required step in reaching neutral products for these processes, protonation of the anionic intermediates has to be placed in the mechanistic scheme. Natural population analysis (NPA) atomic charges at C ␣ /C ␥ /O, the corresponding proton affinities (PA C␣ , PA C␥ , and PA O ), and the molecular electrostatic potential (MEP) mapped on the molecular surface are shown in Fig. 6C as the properties of I ␤ /I ␦ nitronates that determine the kinetics and thermodynamics of proton capture in aqueous solution. Two possibilities arise here (Fig. 7) as follows: O-protonation leading to an aci-Nitro RЈCRϭNO 2 H derivative (aci-Nitro ␤-/␦-adducts) and C ␣ / ␥ -protonation leading to nitroalkane/nitroalkene products (tautomers of the former). A protonation preference for the O-site over the C ␣ -site has been shown for phenylnitromethanes both by experiments under acidic and neutral conditions in aqueous solution and methanol (42)(43)(44) and by computational modeling at the B3LYP/ 6-31ϩG(d,p) level in gas phase (42) or mimicking aqueous solution by including two water molecules in the system (45). The latter study also presented a water-assisted mechanism for conversion of aci-Nitro species toward a nitroalkane C ␣ tautomer, more stable by 8.3 kcal/mol (45). PCM(IEF)-B97X-D/ 6-31ϩG(d,p) modeling in the aqueous solution conducted here on the protonation of I ␤ and I ␦ showed both species to be more prone for O-protonation (aci-Nitro would be thus a kinetic outcome of protonation), but all the C-protonated tautomers were found to be more stable by 5-6 kcal/mol than their aci-Nitro counterparts. Among the final species derived from I ␦ , a small 0.7 kcal/mol difference in stability would favor a C ␥ -protonated nitroalkene over a C ␣ -protonated and unsaturated nitroalkane, with the former being the most stable product achievable by all means.
Regarding stereoisomerism, while protonation at C ␣ must be anti-periplanar to the C ␤ -S bond linking the thiolate moiety that blocks any syn approach to I ␤ (only (R,S) or (S,R) enantiomers can be obtained among ␤-adducts), both syn/anti protonation appears feasible at the more accessible C ␣ position in I ␦ (Fig. 6) yielding either (R,R), (R,S), (S,R), or (S,S) stereoisomers, an outcome that makes it possible to find enantiomers and diastereomers among the final ␦-adducts.
Thus, on the basis of our calculations, a faster O-protonation of I ␤ /I ␦ that would be followed by water-assisted tautomerization into the corresponding more stable counterparts complete the mechanism of reaction leading to the final products. According to modeling, the expected products would be then a ␤-thio-Michael-unsaturated adduct in nitroalkane form (Fig.  7A) in the short timescale and a mix of ␦-thio-Michael adducts, including a prevalent nitroalkene form and two companion diastereomeric unsaturated nitroalkane species (Fig. 7B), in the long timescale.

TABLE 3 Relative energetics of the species participant in ␤/␦-channels using a representative conjugated nitroalkene model
Enthalpy and Gibbs free-energy at 298 K and 1 atm relative to isolated reactants for each species involved in the stepwise mechanism for the thio-Michael addition and further protonation are calculated at the PCM(IEF, water)-B97X-D/6-31ϩG(d,p) level and expressed in kilocalories/mol. Free-energy reaction barrier for the first step and stability of the main Michael adduct(s) corresponding to each of the reaction channels are highlighted in boldface.

Reaction Channel Property
Thiolate adduction step a Protonation step, final products b RC X TS X I X  Fig. 6. b See the corresponding structures in Fig. 7.

NO 2 -CLA-Cysteine ␦-Addition Products Are Found in Human
Urine-Based on experimental and modeling results, endogenous Cys-NO 2 -CLA in human urine is expected to be primarily represented by the more stable ␦-addition products. Isotopically labeled Cys-9-NO 2 -CLA and Cys-12-NO 2 -CLA enriched in ␦-adducts were synthesized, mixed in a 1:1 ratio, and compared with urinary Cys-NO 2 -CLA (Fig. 8). LC-MS/MS analysis is consistent with a predominant presence of the Cys-␦-adducts of 9-and 12-NO 2 -CLA versus the corresponding ␤-products, which presented a different chromatographic profile (supplemental Fig. S2).
NO 2 -CLA Reactions with HSA-Cys-34 in HSA is the most abundant reduced thiol in plasma and is therefore a potential target for reaction with NO 2 -CLA. In addition, HSA has a central role in fatty acid binding and transport; therefore, both covalent and non-covalent interactions with NO 2 -CLA might occur. The absorbance of NO 2 -CLA changes with solvent polarity ( max 330 nm in aqueous solution versus 312 nm in methanol), and thus it is expected to also change upon interaction with the hydrophobic binding sites in HSA. NO 2 -CLA incubation with delipidated thiol-blocked HSA resulted in a blue shift in absorbance (Fig. 9A). Titrations performed by adding aliquots of delipidated thiol-blocked HSA to a fixed amount of NO 2 -CLA allowed us to estimate up to 3.8 Ϯ 0.5 NO 2 -CLA bound per delipidated thiol-blocked HSA (Fig. 9B). When titrations were carried out using lipidated (stearic acid/HSA, 5:1) thiol-blocked HSA, a similar value of 3.2 Ϯ 0.2 NO 2 -CLA per HSA was obtained indicating that NO 2 -CLA was able to displace stearic acid (data not shown). Fitting of the titration data to hyperbolic equations yielded global apparent dissociation constants of (0.9 Ϯ 0.3) ϫ 10 Ϫ6 and (1.2 Ϯ 0.3) ϫ 10 Ϫ6 M for delipidated and lipidated HSA, respectively, consistent with reported values for non-nitroalkene fatty acids (46).
The reactivity of Cys-34 toward NO 2 -CLA could not be evaluated following changes in absorbance at 330 nm because of interference from non-covalent binding. Therefore, the decay in HSA thiol concentration after incubation of reduced and lipidated HSA (50 M) with NO 2 -CLA (300 M) for 1 h was assessed using 4,4Ј-dithiodipyridine (DTDP). The mole of thiol per mol of HSA were 0.47 and 0.25 in the absence and presence of NO 2 -CLA, respectively. Considering the concentrations used and the incubation time, a lower limit of 0.5 M Ϫ1 s Ϫ1 (25°C, pH 7.4) was estimated for the rate constant.

Discussion
NO 2 -CLA reacts reversibly with thiols forming Michael adducts. In agreement with the presence of two electrophilic centers in NO 2 -CLA, two products were detected, the ␤-adducts and the ␦-adducts. The ␤-adducts are formed at faster rates and are thus the kinetic products. In contrast, the ␦-adducts are formed with slower kinetics but present higher stabil- ity and are thus the thermodynamic products. In the case of GSH, the ␦-adduct is 30 times more stable than the ␤-adduct.
From a mechanistic viewpoint, our experimental and computational results are consistent with a stepwise addition process where the rate-controlling step involves the nucleophilic attack of the thiolate on NO 2 -CLA to form an anionic intermediate, followed by proton incorporation, and the reverse process would be involved in elimination. Although nitroalkanes possess relatively high carbon acidity (pK a 10.28 for nitromethane (43,47)), deprotonation rates can be outstandingly slow, with rate constants for proton transfer from nitromethane to OH Ϫ measured at 27.6 M Ϫ1 s Ϫ1 (43,47). In fact, deprotonation has been proposed to control the rate of elimination of other activated alkanes (25). However, in the case of the reaction between NO 2 -CLA and thiols, the Brønsted correlations obtained together with the lack of a solvent deuterium kinetic isotopic effect rule out protonation as the rate-controlling step in the addition process (or deprotonation in the elimination).
The results obtained in this study can be generalized to predict the stability of NO 2 -CLA adducts as a function of thiol pK a . If specific aspects of protein thiol reactivity and steric constraints are dismissed, the K eq at a certain pH can be calculated from dividing k off and k on and combining Equations 3-5 to obtain Equation 6.
For both the ␤and the ␦-adducts, substitution in Equation 6 with the ␤ lg , ␤ nuc , C off , and C on values obtained from Fig. 5 allows the generalization that, at pH 7.4, the stability of the adducts increases as the pK a increases, so that adducts formed with relatively less acidic thiols (higher pK a ) are the more stable at neutral pH (Fig. 10).
Considering the wide variety of thiols present in the biological context and the reversibility of the reactions described herein, it is likely that NO 2 -CLA will exist as part of a dynamic pool alternating between free and bound forms. Our study predicts that under intracellular conditions, NO 2 -CLA will quickly react with GSH and other low and high molecular weight thiols to initially form ␤-adducts. Because of the reversibility of these reactions, the ␤-adducts will undergo elimination, and the free NO 2 -CLA will eventually give rise to the more stable ␦-adducts. Thus, the transient formation of ␤-adducts could contribute to establish a dynamic buffer of NO 2 -CLA. Considering the high concentrations of GSH (2-17 mM) and protein thiols inside cells (10 -50 mM) (18,19), we can predict that if equilibrium is achieved Ͼ99% of the NO 2 -CLA pool will correspond to covalently bound forms (assuming similar reactivity for protein thiols and free cysteine). Furthermore, Ͻ4% of these products will correspond to ␤-adducts and Ͼ96% to ␦-adducts, given the higher stability of the latter. In plasma, considering the lower concentrations of thiols (ϳ450 M), it can be predicted that a significant fraction of the NO 2 -CLA pool will be non-covalently bound to HSA, probably impacting its transport and storage. Nevertheless, it is important to understand that the interaction of NO 2 -CLA with a particular protein thiol will be affected by the environment of the nucleophilic residue.
The adducts formed with GSH can be exported out of the cell through multi-drug resistance protein channels, which is a prominent pathway for nitroalkene fatty acid cell clearance. Once in the extracellular milieu, GSH-NO 2 -CLA adducts are processed by ␥-glutamyl transpeptidases and dipeptidases/ aminopeptidases. The resulting cysteine conjugates are ex-  creted in the urine. A small portion of the cysteine conjugates can be N-acetylated intracellularly forming mercapturic acid derivatives (2,28,48).
The finding of adducts of NO 2 -CLA with cysteine in urine provides evidence for the in vivo reaction of NO 2 -CLA with thiols. Moreover, it is likely that these adducts originated intracellularly by conjugation with GSH followed by cell export, removal of the glutamyl and glycyl moieties, and excretion. The fact that the Cys-NO 2 -CLA adducts found in urine correspond to ␦-addition products is consistent with the higher stability of the ␦versus ␤-adducts of cysteine and GSH.
Overall, we have performed an in-depth study of the interactions of NO 2 -CLA with biological thiols and HSA. As the beneficial health effects of nitroalkene fatty acids continue to be unveiled, our study contributes to the understanding of the chemistry that underlies the protective actions of these endogenously formed potential drug candidates.

Experimental Procedures
General Solutions-All experiments were performed in 0.1 M phosphate buffer at pH 7.4, containing 0.1 mM DTPA unless otherwise specified. Low molecular weight thiol solutions were prepared in nanopure water and used on the same day. TNB was synthesized as described previously (49). Thiol concentrations were determined with DTNB before and after experiments using an absorption coefficient at 412 nm of 14,150 M Ϫ1 cm Ϫ1 (50). Stearic acid solutions (100 mM) were freshly prepared in methanol (80°C, with agitation). DTDP solution (0.25 mM) was prepared in phosphate buffer, 0.1 M, pH 7.0.
HSA Solutions-HSA was delipidated with activated charcoal as described (53). Reduced HSA was prepared by incubation with BME (10 mM, for 30 min at room temperature with agitation) followed by gel filtration on PD-10 columns equilibrated with phosphate buffer (0.1 M, pH 7.4, 0.1 mM DTPA). Thiolblocked HSA was prepared by incubating delipidated HSA with BME followed by addition of NEM (150 mM, for 15 min at room temperature with agitation), followed by gel filtration against phosphate buffer. Lipidated and reduced HSA was prepared by incubating delipidated HSA with 5:1 stearic acid (stearic acid/ HSA) (30 min at room temperature with agitation) followed by addition of BME and subsequent gel filtration. The HSA concentration was determined from the absorbance at 279 nm (⑀ ϭ 0.531 (g/liter) Ϫ1 cm Ϫ1 , 66,438 Da) (46). HSA thiols were measured with DTNB in sodium pyrophosphate buffer (0.1 M, pH 9, 5 min) (54). Reactivity Patterns in 9-/12-NO 2 -CLA as Explored by Electronic Structure Computational Modeling-Aqueous solution structures of 9-/12-NO 2 -CLA were fully optimized and verified by inspection of the Hessian eigenvalues at the PCM(IEF)-B97X-D/6-31ϩG(d,p) level (57-59), including non-electrostatic contributions to solvation as implemented in Gaussian09 revision D.01 (60). Each solute was contained in a molecular shaped cavity constructed with Bondi's radii (61). B97X-D density functional was chosen based on its known superior performance in modeling thio-Michael additions (62). Energies of the highest occupied and lowest unoccupied Kohn-Sham orbitals (⑀ HOMO and ⑀ LUMO , respectively) were used to assess global softness S as the inverse of the global chemical hardness , calculated as follows: The electrophilic Fukui function f ϩ (r), measuring the propensity of NO 2 -CLA to gain electron density in a nucleophilic attack, was also evaluated for each regioisomer within a finite difference approach (63), as shown in Equation 8,

UV-Visible Assessment of Nitroalkene Fatty Acid Reactions with Thiols-NO
where N (r) and Nϩ1 (r) represent the electron density at each point r around the molecule, respectively, obtained for NO 2 -CLA in the anionic state of reference (N) and after gaining one extra electron at the ground-state geometry (N ϩ 1) determined by single-point calculations at the same level of theory. Electrophilic sites were thus inspected by mapping f ϩ (r) on a molecular surface of 0.0004 au isodensity by using Gaussview5 for generating the molecular graphics represented in Fig. 2 (64). To further assess differences among carbon electrophilic sites within each molecule, condensed Fukui function for each atom A (65) was also obtained by applying a Natural Bond Orbital (NBO) calculation and deriving NPA atomic charges q A from the corresponding population analysis (66) as expressed in Equation 9, Electrophilic atomic indices (condensed atomic softness, s A ϩ ) were finally obtained as shown in Equation 10, Computational Modeling of the Detailed Mechanism of Thio-Michael Additions at C ␤ /C ␦ Using Representative Model Species-Geometries of the RC, TS, and the anionic intermediate products (I) for the addition of methane thiolate at the C ␤ /C ␦ positions of a representative conjugated nitroolefin moiety (2-nitrohexa-2,4-diene) were fully optimized and verified at the same level of theory as previously applied for NO 2 -CLA 9-/12-regioisomers. IRC reaction paths (67) correspondingly interconnecting them were generated with the HPC algorithm (68). Thermochemistry was calculated at 298 K and 1 atm under the usual statistical thermodynamics approaches (unscaled harmonic vibrational frequencies and rigid rotors) as implemented in Gaussian09 D.01 (60). NPA atomic charges (66) and WBIs (69) were also extracted from the corresponding electronic structures. The structure and relative stability of the tautomeric products further obtained after protonation of the anionic intermediate outcoming from each thiolate direct addition on C ␤ /C ␦ were also characterized at the same level of theory, assessing the corresponding PA as Gibbs free-energy differences between the protonated and deprotonated form of each species. This approach has been recently validated by Taunton and co-workers (70) in an integrated kinetic and computational characterization of thio-Michael adducts established between activated acrylonitriles and ␤-mercaptoethanol. Molecular graphics representing the 3D structure of the species characterized were prepared with Discovery Studio Visualizer 4.0 (Accelrys). Molecular electrostatic potential was mapped for both nitronate intermediates on a total electronic density surface of 0.004 au, obtaining graphical representations with Gaussview5 (Semichem Inc.) (64).
LC-MS/MS Analysis of the Reaction between NO 2 -CLA and ␤-Mercaptoethanol-9or 12-NO 2 -CLAs (10 M) were reacted with BME (1.76 mM) in phosphate buffer, pH 7.4, at 25°C. Aliquots were removed at different time points, and the reaction was stopped by dilution into 2 volumes of either 10% formic acid or 100 mM NEM. Samples were allowed to react for 30 min, diluted in 15 volumes of methanol, and subjected to LC-MS/MS analysis.
LC-UV-Visible Analysis of the Reaction between NO 2 -CLA and BME-High performance liquid chromatography experiments were performed in an Agilent 1260 Infinity instrument. 9-NO 2 -CLA (100 M) was mixed with BME (1.76 mM) in phosphate buffer at 25°C. At increasing times (10 s, 5 min, and 1 h), aliquots (70 l) were mixed with 50% acetonitrile, 10% formic acid (140 l). Samples (150 l) were injected in a C18 column (Zorbax Eclipse Plus C18, 4.36 ϫ 100 mm; 3.5 m) and resolved using a linear gradient from 35 to 90% solution B in 19 min at a flow rate of 1 ml/min (solution B, acetonitrile, 0.1% acetic acid; solution A, H 2 O, 0.1% acetic acid). The absorbance was registered using a diode array detector.
Determination of Non-covalent NO 2 -CLA Binding to HSA-Increasing concentrations of thiol-blocked delipidated or lipidated HSA (0.2-10 M) were mixed with NO 2 -CLA (10 M) in phosphate buffer, and the UV-Visible spectra were recorded. Control procedures were performed in the absence of NO 2 -CLA.
Reaction between NO 2-CLA and HSA Thiol-Lipidated reduced HSA (50 M) was mixed with NO 2 -CLA (300 M) or an equivalent volume of methanol (control) and incubated for 1 h (25°C, pH 7.4). The samples were gel filtrated to remove excess NO 2 -CLA, and protein concentration was determined using both absorbance measurements at 279 nm and the bicinchoninic acid assay (71). The remaining free thiols were quantified using DTDP (215 M). Protein was first removed via ultrafiltration. Thiopyridone concentration was determined at 324 nm (⑀ ϭ 21,400 M Ϫ1 cm Ϫ1 ) (50) in the ultrafiltrate after baseline subtraction.
Nitroalkene Fatty Acid Extraction from Human Urine-Urine samples were collected from healthy human volunteers and used immediately (2) (University of Pittsburgh IRB PRO07110032). Briefly, C18 SEPAK columns were conditioned with 1 column volume of 100% methanol and equilibrated with 2 volumes of 10% methanol, and 5 ml of urine were loaded. Columns were then washed with 2 volumes of 10% methanol and dried under vacuum, and lipids eluted with 1 volume of methanol. The solvent was evaporated, and the samples were resuspended in 100 l of methanol for LC-MS/MS analysis.
Data Processing-Data were plotted and analyzed using OriginPro 8.0 (OriginLab) or Prism 6 (GraphPad). Unless specified, results are expressed as the means Ϯ S.E. of independent experiments.
Author Contributions-L. T. and D. A. V. prepared the manuscript, contributed to writing, designed, performed, and analyzed experiments. L. L. performed experiments. E. L. C. designed, performed, and analyzed (with the assistance of C. S. in characterizing 12/9-NO 2 -CLA local softness) computational modeling and wrote the corresponding sections of the manuscript. M. N. M. designed and performed LC-UV-Visible experiments. S. R. S. contributed to LC-MS/MS method development. S. R. W. synthesized nitrated fatty acids. B. A. and F. J. S. contributed to the overall concept, experimental design, data interpretation, and manuscript preparation. All authors have given approval to the final version of the manuscript.