Mitochondrial Peroxiredoxin 3 Is Rapidly Oxidized and Hyperoxidized by Fatty Acid Hydroperoxides

Human peroxiredoxin 3 (HsPrx3) is a thiol-based peroxidase responsible for the reduction of most hydrogen peroxide and peroxynitrite formed in mitochondria. Mitochondrial disfunction can lead to membrane lipoperoxidation, resulting in the formation of lipid-bound fatty acid hydroperoxides (LFA-OOHs) which can be released to become free fatty acid hydroperoxides (fFA-OOHs). Herein, we report that HsPrx3 is oxidized and hyperoxidized by fFA-OOHs including those derived from arachidonic acid and eicosapentaenoic acid peroxidation at position 15 with remarkably high rate constants of oxidation (>3.5 × 107 M−1s−1) and hyperoxidation (~2 × 107 M−1s−1). The endoperoxide-hydroperoxide PGG2, an intermediate in prostanoid synthesis, oxidized HsPrx3 with a similar rate constant, but was less effective in causing hyperoxidation. Biophysical methodologies suggest that HsPrx3 can bind hydrophobic structures. Indeed, molecular dynamic simulations allowed the identification of a hydrophobic patch near the enzyme active site that can allocate the hydroperoxide group of fFA-OOHs in close proximity to the thiolate in the peroxidatic cysteine. Simulations performed using available and herein reported kinetic data indicate that HsPrx3 should be considered a main target for mitochondrial fFA-OOHs. Finally, kinetic simulation analysis support that mitochondrial fFA-OOHs formation fluxes in the range of nM/s are expected to contribute to HsPrx3 hyperoxidation, a modification that has been detected in vivo under physiological and pathological conditions.


Introduction
In most mammalian cells, mitochondria are main sources of oxidizing species [1,2]. Among them, it forms different hydroperoxides such as hydrogen peroxide (H 2 O 2 ), peroxynitrite and either free or lipid-bound fatty acid hydroperoxides ( f FA-OOHs or L FA-OOHs, respectively), many of which participate in cellular redox signaling pathways at low, physiological concentrations [3][4][5]. However, increased steady-state levels of these species can lead to mitochondrial dysfunction, cytotoxicity and several disease states [4,[6][7][8][9][10]. Different peroxidases reduce hydroperoxides in mitochondria [11]. Peroxiredoxin 3 (Prx3) catalyzes the reduction of most of mitochondrial H 2 O 2 [12]. Furthermore, our previous work indicates that Prx3 is an efficient peroxynitrite reductase that represents, together with peroxiredoxin 5 (Prx5), the main target for peroxynitrite in mitochondria [13]. Both H 2 O 2 and peroxynitrite can lead to Prx3 inactivation due to peroxidatic cysteine (Cys P ) hyperoxidation to sulfinic acid (RSO 2 H) or even to sulfonic acid (RSO 3 H), although the enzyme is considered to be less susceptible to this modification than other mammalian typical two cysteine peroxiredoxins (Prxs) such as Prx2 [13,14]. This was ascribed to a slower rate constant of resolution, i.e., change from the fully folded (FF) to locally unfolded (LU) f FA-OOHs but have no-or only a very weak activity with L FA-OOHs [36][37][38]. In some cases, f FA-OOHs not only oxidize but also hyperoxidize Cys P [39,40]. The molecular basis for the oxidizing substrate specificity in Prxs is only starting to be unraveled [41]. Some previous data suggest that Prx3 might be able to catalyze the reduction of f FA-OOHs. For example, cells lacking Prx3 showed an increased susceptibility of the tumor suppressor protein PTEN, a member of the protein tyrosine phosphatase family, to be oxidized by a f FA-OOH [42]. Furthermore, Prx3 got hyperoxidized in cells exposed to AA-derived hydroperoxides [43]. However, since addition of f FA-OOHs to cells can lead to mitochondrial dysfunction and further oxidant production [44], a direct reaction of FA-OOHs and Prx3 remains to be investigated. Although the enzyme is mostly expressed in mitochondria, it can also be located in cell surfaces [45]. For instance, it was detected in membrane-enriched fractions from cancer cell lines exposed to RSL3, an inhibitor of GPx4 that causes ferroptosis. However, Prx3 siRNA failed to demonstrate a protective role against this form of death in those cells [46]. Interestingly, mouse embryonic fibroblast expressing a mutated form of GPx4 (Sec to Cys) showed a compensatory up-regulation of Prx3 [47]. Furthermore, overexpression of Prx3 ameliorates the symptoms of mild but not severe forms of ferroptosis of spinal motor neurons from GPx4 neuron inducible knockout mice [48].
Herein, we investigated the ability of human Prx3 (HsPrx3) to catalyze the reduction of FA-OOHs. In particular, we investigated the reactions of HsPrx3 with f FA-OOHs involved in eicosanoid synthesis, i.e., derived from 15-lipoxygenase-catalyzed peroxidation of AA and EPA (15S-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid (15(S)-HpETE) and 15S-hydroperoxy-5Z,8Z,11Z,13E,17Z-eicosapentaenoic acid (15(S)-HpEPE), respectively) and derived from cyclooxygenase-catalyzed peroxidation of AA (the endoperoxidehydroperoxide prostaglandin G 2 (PGG 2 )), which is a precursor for the 2-series prostaglandin and thromboxane synthesis (structures are shown in Supplementary Figure S1). Our data indicate that the enzyme is very rapidly oxidized as well as hyperoxidized by f FA-OOHs, alerts on the interpretation of kinetic results obtained by different methodologies and could help to explain at least in part Prx3 hyperoxidation detected in mammalian tissues [49,50].

Enzyme Expression and Purification
Recombinant HsPrx3 (wild type, wtHsPrx3 as well as a Cys P to Ser mutated form, C P SHsPrx3) lacking the mitochondrial targeting sequence and containing three additional N-terminal Gly residues were expressed in BL21 Star (DE3) Escherichia coli strain and purified as previously reported [13]. Human thioredoxin 1 (HsTrx1), which acts as an efficient reducing substrate for HsPrx3 in vitro [54], was expressed as in [55]. The plasmid for recombinant expression of Echinococcus granulosus thioredoxin glutathione reductase lacking the glutaredoxin (Grx) domain (EgTR) was obtained from Mariana Bonilla and Lucía Turell, from Institut Pasteur Montevideo and Universidad de la República, Montevideo, Uruguay, respectively, and was expressed and purified as previously [56].

Reduction of HsPrx3
In some experiments, HsPrx3 was reduced immediately before use by incubating the enzyme with DTT (2 mM) for 30 min on ice. Excess reductant was removed by gel filtration using a HiTrap desalting column (Amersham Biosciences, Amersham, UK) and freshly made degassed sodium phosphate buffer 50 mM plus 0.1 mM DTPA, pH 7.8.

Catalytic Activity of HsPrx3
The catalytic activities of HsPrx3 using H 2 O 2 or different f FA-OOHs as oxidizing substrates were measured using a coupled assay following NADPH oxidation at 340 nm (ε 340 nm = 6220 M −1 s −1 ) (150 µM) [57] in the presence of EgTR (40 nM), HsTrx1 (30 µM) and HsPrx3 (0.8 µM) in sodium phosphate buffer 50 mM plus 0.1 mM DTPA at pH 7.8 and 25 • C. The mixture was pre-incubated during a minute and the reaction was started by adding 20 µM of the indicated hydroperoxide, rapidly mixed in a vortex and transferred to a cuvette to start measurements. In order to determine whether the enzyme remained active after the incubation with the hydroperoxide of interest, H 2 O 2 (20 µM) was added 3 min after the initial incubation. Control experiments using different concentrations of EgTR and HsPrx3 were performed to ensure that reduction of HsTrx1 by EgTR was not rate limiting the oxidation of NADPH by the hydroperoxides of interest in the presence of EgTR, HsTrx1 and HsPrx3. EgTR, which as mammalian TR is a selenium-containing protein, was used due to its availability as a highly pure recombinant protein in our lab. Measurements were performed using a Shimadzu UV-Vis spectrophotometer (Shimadzu, Kyoto, Japan) equipped with a Peltier-based temperature control.

Kinetics of HsPrx3 Oxidation and Hyperoxidation by f FA-OOHs
As in other members of the typical two-cysteine Prx family [58][59][60], changes in the redox state of Cys P of HsPrx3 are accompanied by changes in the tryptophan-dependent intrinsic fluorescence intensity of the protein [13]. The oxidation of the thiol group in HsPrx3 Cys P to sulfenic acid causes a rapid decrease in intrinsic fluorescence while resolution causes a slower phase of fluorescence increase. Since oxidation is a bimolecular process, the observed rate constant of the decrease in intrinsic fluorescence depends on the concentration of the oxidant. On the contrary, resolution is an intramolecular process and its rate constant is independent on the oxidant concentration. Hyperoxidation can compete with resolution leading to an acceleration of the second phase of fluorescence change (Supplementary Figure S2). We therefore took advantage of this fluorescence behavior of the protein to measure the kinetics of HsPrx3 Cys P oxidation and hyperoxidation. Reduced wt or C P SHsPrx3 in 50 mM sodium phosphate buffer plus 0.1 mM DTPA pH 7.8 in one syringe was rapidly mixed with the indicated concentrations of the hydroperoxide of interest in the same buffer from a second syringe at 11-12 • C, and the protein intrinsic fluorescence changes were recorded using a SX20 Applied Photophysics stopped-flow spectrofluorimeter (Applied Photophysics, Leatherhead, UK) (λ ex = 295 nm, total emission). Data were fitted by exponential functions to obtain the observed rate constants of each phase (k obs ). The rate constants of the reaction of HsPrx3 Cys P oxidation and hyperoxidation were calculated from the slope of the plot of the k obs of the decrease phase and the increase phase of fluorescence change, respectively, plotted versus hydroperoxide concentration.

SDS PAGE and Western Blot Analysis of HsPrx3 Hyperoxidation
Reduced HsPrx3 (5 µM) was treated with the indicated concentrations of H 2 O 2 or f FA-OOHs in sodium phosphate buffer 50 mM pH 7.8 plus 0.1 mM DTPA. After 15 min, the alkylating agent NEM (20 mM) was added. Samples were resolved in 15% SDS-PAGE electrophoresis under non-reducing (−β-ME) or reducing (+β-ME) conditions and were stained using Coomasie. Alternatively, samples were transferred to a nitrocellulose membrane to detect HsPrx3 Cys P hyperoxidation to sulfinic/sulfonic (Cys P -SO 2/3 H) acid using a commercial rabbit polyclonal antiPrx-Cys P -SO 2/3 H antibody (Abcam, 1b16830, Waltham, MA, USA). Secondary anti-rabbit antibodies were IR-Dye 800CW and data were acquired using an Odyssey imaging system from LI-COR Bioscience (Lincoln, NE, USA).

Mass Spectrometry Analysis of HsPrx3 Modifications by FA-OOHs
Reduced HsPrx3 (10 µM) in 50 mM phosphate buffer plus 0.1 mM DTPA (pH 7.8, 25 • C) without further treatment, exposed to 15(S)-HpEPE (10 or 20 µM) or to H 2 O 2 (20 µM) were alkylated with 2-iodoacetamide (5 mM) for 30 min at 37 • C. The excess alkylating agent was removed by gel filtration using PD SpinTrap G-25 columns (GE Healthcare, Chicago, IL, USA) equilibrated with 20 mM ammonium bicarbonate buffer (pH 7.4, 25 • C). All samples were loaded into a C4 column (214MS5115, Grace Vydac, Hesperia, CA, USA) for HPLC separation. Mobile phase consisted of 0.1% formic acid in nanopure water (solvent A) and 0.1% formic acid in CH 3 CN (solvent B), and elution of the protein was performed with a 10-min linear gradient of solvent B (5-50%) followed by an additional 10 min at 50% solvent B at 100 µL/min. An ESI-triple quadrupole mass spectrometer (QTRAP4500, ABSciex Framingham, MA, USA) was employed for detection. The spectrometer was set in Q1 positive mode in the 500-2000 m/z range with a scan rate of 200 Da/s and Q1 resolution in unit. Parameters used were as follows: Ion Sray voltage, 5000 V; Ion Source Temperature, 300 • C; Declustering Potential, 120 V; Entrance Potential, 10; Curtain Gas, 20 psi; Gas1, 30 psi; Gas2, 20 psi. Data acquisition was performed using Analyst 1.6.2 (AB-Sciex) and PeakView 2.2 (ABSciex) software was used for data analysis and deconvolution of all spectra.

Fatty Acids Binding to HsPrx3
The ability of HsPrx3 to bind fatty acids (FAs) was evaluated by two different experimental methodologies. Firstly, we analyzed the interactions of HsPrx3 with ANS, a fluorescent compound used as a probe of protein surface hydrophobicity in the proximity of cationic groups [61,62]. ANS binding to this protein microenvironment leads to a blue shift and an increase in the probe quantum yield. Furthermore, due to its spectral properties ANS could lead to intrinsic, tryptophan-dependent protein fluorescence resonance energy transfer (FRET) [63]. Fluorescence measurements were performed using a Jasco FP-8500 spectrofluorimeter and 500 µL, 10 mm path length quartz cuvettes. For direct fluorescence measurements, excitation was performed at 375 nm and emission spectra (440-600 nm) were obtained. HsPrx3 (5 µM) was titrated with ANS in sodium phosphate buffer 50 mM pH 7.8 plus 0.1 mM DTPA. Control experiments in the absence of HsPrx3 were also performed. In FRET experiments, HsPrx3 (2 µM, 5 µM and 8 µM) was excited at 295 nm, and emission spectra were recorded between 315 nm and 550 nm. Binding constants were obtained fitting the plot of change in intrinsic fluorescence intensity versus ANS concentration to hyperbolic functions as well as through Scatchard analysis.
Additionally, heat unfolding experiments of HsPrx3 in the absence and presence of AA were performed to analyze the effect of the FA on the thermal stability of the protein. Briefly, the tryptophan-dependent intrinsic fluorescence intensity of HsPrx3 (4 µM) (λ ex = 295 nm, λ em = 335 nm) in the absence or presence of AA (10 or 20 µM) plus an excess of DTT (2 mM) was registered in a Jasco FP-8500 spectrofluorimeter (Jasco, Tokyo, Japan), while an increase in temperature (35-70 • C) with a heating rate of 1.5 • C/min was applied to the sample. Sample buffer was 50 mM phosphate buffer plus 0.1 mM DTPA, pH 7.8 and a 10 mm path length quartz cuvette (500 µL) was used. The apparent melting temperatures (Tm) as midpoint of the denaturation curves were obtained fitting the data to a sigmoidal function.

Computer-Assisted Molecular Modeling of FA Binding to HsPrx3
To perform docking simulations of the binding of 15(S)-HpETE to reduced HsPrx3, a dimeric unit made of chains A and B was extracted from its reported crystal structure PDB id 5jcg and subjected to a short molecular dynamic [64] simulation [64] as previously reported by our group [40,41]. Briefly, the dimer was solvated using a default method, with an octahedral box of 12 Å in radius with TIP3P water molecules [65]. Protein residue parameters correspond to the parm14SB Amber force field [66]. MD was performed using periodic boundary conditions with a 10 Å cutoff and particle mesh Ewald (PME) summation method for treating the electrostatic interactions. The hydrogen bond lengths were kept at their equilibrium distance by using the SHAKE algorithm, while temperature and pressure were kept constant with a Langevin thermostat and barostat, respectively, as implemented in the AMBER program (Amber, University of California, San Diego, CA, USA) [67]. The system was optimized in 1000 steps (10 with steep gradient and the rest with conjugate gradient). Then, it was slowly heated from 0 K to 300 K for 20 ps at constant pressure, with Berendsen thermostat, and pressure was equilibrated at 1 bar for 5 ps. After these two steps, a 10 ns long MD simulation at constant temperature (300 K) and constant volume was performed, followed by a 200 ns long trajectory. The last structure resulting from the MD was used for the molecular docking of 15(S)-HpETE. The structure of the enzyme was considered rigid while 15(S)-HpETE was allowed to be flexible. 500 binding modes were generated and analyzed using standard protocols of AutoDock4.2.6 (The Scripps Research Institute, La Jolla, CA, USA) [68].

Simulations of HsPrx3 Hyperoxidation by Fluxes of Mitochondrial Hydroperoxides in the Presence of Peroxidases
Simulations were performed using COPASI 4.37 program. Rate constants, reactant concentrations and formation rates of H 2 O 2 and f FAOOH are indicated in the text.

HsPrx3 Is Rapidly Inactivated by f FA-OOHs
HsPrx3 catalyzed the reduction of H 2 O 2 (20 µM) as detected by a decrease in NADPH concentration in the coupled assay described in Methods. No consumption of NADPH was observed in the absence of HsPrx3 (not shown). The reaction ended due to oxidizing substrate depletion and not because of enzyme inactivation due to hyperoxidation, since a new addition of H 2 O 2 to the cuvette caused a new phase of NADPH consumption of similar rate. This is consistent with previous data indicating that the rate constant of HsPrx3 hyperoxidation by H 2 O 2 is 1.1 × 10 3 M −1 s −1 versus a rate constant of resolution of 2 s −1 at pH 7.8 and 25 • C [13], which indicates that at 20 µM H 2 O 2 only~1% of the enzyme would be hyperoxidized in the first catalytic cycle, a percentage/cycle that should decrease as the oxidant gets consumed. On the contrary, very slow rates of NADPH consumptions were observed when 15(S)-HpETE or 15(S)-HpEPE were used as substrates in the coupled assay. The addition of H 2 O 2 (20 µM) to the cuvette at the end of the reaction failed to cause a decrease in NADPH concentration, indicating that HsPrx3 was inactive ( Figure 1). The rate of NADPH consumption was modest in the case of PGG 2 (Table 1).

Kinetics of HsPrx3 Oxidation by fFA-OOH
The addition of 15(S)-HpETE ( Figure 2A), 15(S)-HpEPE ( Figure 2B), and PGG2 ( Figure 2C) in excess to reduced HsPrx3 caused a very rapid decrease in the enzyme intrinsic fluorescence (phase of oxidation of the thiolate in CysP to sulfenic acid), followed by a slower phase of fluorescence increase (resolution plus hyperoxidation phase) (Supplementary Figure S2). In the case of 15(S)-HpETE and 15(S)-HpEPE, the first phase was so rapid that it was complete in a few ms even at the lower concentrations of oxidant tested (Figure 2A Figure 2C) in excess to reduced HsPrx3 caused a very rapid decrease in the enzyme intrinsic fluorescence (phase of oxidation of the thiolate in Cys P to sulfenic acid), followed by a slower phase of fluorescence increase (resolution plus hyperoxidation phase) (Supplementary Figure S2). In the case of 15(S)-HpETE and 15(S)-HpEPE, the first phase was so rapid that it was complete in a few ms even at the lower concentrations of oxidant tested (Figure 2A as (1.7 ± 0.1) × 10 7 , (2.6 ± 0.4) × 10 7 and (3.1 ± 0.7) × 10 5 M −1 s −1 , respectively, ( Figure 2D). As controls, the addition of the non-peroxidized fatty acid EPA (3 µM) to reduced HsPrx3 (0.3 µM) caused no change in the protein intrinsic fluorescence. Similarly, the addition of PGG2 to reduced CPSHsPrx3 (0.3 µM) caused no change in protein intrinsic fluorescence, consistent with the requirement of both CysP in HsPrx3 and the hydroperoxide group in the fatty acid for the reaction to occur ( Figure 2E,F).  The time courses of intrinsic fluorescence intensity change were followed (λ ex = 295 nm, total emission). The inset of (C) shows a linear fit of k obs of fluorescence intensity decay plotted versus PGG 2 concentration from which the second order rate constant of HsPrx3 oxidation by PGG 2 was obtained. The rate constant of hyperoxidation of Cys P was determined for the three f FA-OOHs mentioned above, plotting the k obs of the phases of fluorescence intensity recovery against f FA-OOH concentration (D). Both in inset of (C) and in (D) the symbols and error bars represent medians and standard deviations from at least 5 replicates of one of the three independent experiments performed. The effect of EPA (3 µM) on the intrinsic fluorescence of reduced HsPrx3 (0.3 µM) (E) and of PGG 2 (0.3 µM) on C P SHsPrx3 (0.3 µM) (F) are also shown.

Consumption of Hydroperoxides by HsPrx3 under Non-Catalytic Conditions
When H 2 O 2 (50 µM) was mixed with reduced HsPrx3 (20 µM) a fast consumption of 0.8 equivalents of hydroperoxide by HsPrx3 occurred. Similarly, a 1:1 consumption of PGG 2 /HsPrx3 was obtained. However, the stoichiometry of hydroperoxide consumption was higher (1.7:1) for 15(S)-HpEPE. This is consistent with a fast hyperoxidation process in the case of 15(S)-HpEPE that consumes a second equivalent of oxidant ( Figure 3).
obtained. The rate constant of hyperoxidation of CysP was determined for the three fFA-OOHs mentioned above, plotting the kobs of the phases of fluorescence intensity recovery against fFA-OOH concentration (D). Both in inset of (C) and in (D) the symbols and error bars represent medians and standard deviations from at least 5 replicates of one of the three independent experiments performed. The effect of EPA (3 µM) on the intrinsic fluorescence of reduced HsPrx3 (0.3 µM) (E) and of PGG2 (0.3 µM) on CPSHsPrx3 (0.3 µM) (F) are also shown.

Consumption of Hydroperoxides by HsPrx3 under Non-Catalytic Conditions
When H2O2 (50 µM) was mixed with reduced HsPrx3 (20 µM) a fast consumption of 0.8 equivalents of hydroperoxide by HsPrx3 occurred. Similarly, a 1:1 consumption of PGG2/HsPrx3 was obtained. However, the stoichiometry of hydroperoxide consumption was higher (1.7:1) for 15(S)-HpEPE. This is consistent with a fast hyperoxidation process in the case of 15(S)-HpEPE that consumes a second equivalent of oxidant ( Figure 3).   4 and 6). The presence of monomeric form of the enzymes in samples treated with excess hydroperoxides and without β-ME are consistent with the formation of hyperoxidized monomers as confirmed by Western Blot (see below).
When reduced HsPrx3 (5 µM) was exposed to H 2 O 2 , hyperoxidation of Cys P was only marginally detected unless excess oxidant was used. As reported [14], hyperoxidation by H 2 O 2 occurred firstly in one of the subunits of the dimer, while the other was forming an inter-subunit disulfide which is detected as a hyperoxidized dimer, and only at concentrations of H 2 O 2 ≥ 500 µM hyperoxidized monomers, i.e., subunits with both Cys P hyperoxidized, were detected ( Figure 4A). On the contrary, both 15(S)-HpETE and 15(S)-HpEPE caused HsPrx3 Cys P hyperoxidation at lower concentrations, even equimolar, that was detected in both the monomeric and dimeric forms of the protein, in agreement with a fast kinetics of hyperoxidation. PGG 2 -mediated hyperoxidation profile was similar to that of H 2 O 2 , leading to hyperoxidized dimers at concentrations ≤50 µM (not shown).
forming an inter-subunit disulfide which is detected as a hyperoxidized dimer, and only at concentrations of H2O2 ≥ 500 µM hyperoxidized monomers, i.e., subunits with both CysP hyperoxidized, were detected ( Figure 4A). On the contrary, both 15(S)-HpETE and 15(S)-HpEPE caused HsPrx3 CysP hyperoxidation at lower concentrations, even equimolar, that was detected in both the monomeric and dimeric forms of the protein, in agreement with a fast kinetics of hyperoxidation. PGG2-mediated hyperoxidation profile was similar to that of H2O2, leading to hyperoxidized dimers at concentrations ≤50 µM (not shown).

Mass Spectrometry Analysis
In order to confirm the hyperoxidation of HsPrx3 treated with fFA-OOHs, MS analysis was performed. The predicted molecular weight of the reduced HsPrx3 subunit is 21,797 Da (13). Treatment with excess 2-iodoacetamide caused alkylation of reduced HsPrx3 (major peaks of + 113 and +170 which corresponds to the covalent addition of 2 and 3 carbamidomethyl (CAM) groups (+57 each), respectively, Figure 5A and Table 2). The presence of a small fraction of a form of the protein modified by 4 CAM groups in the reduced enzyme that has three Cys residues in its sequence is consistent with the reported side reaction of the alkylating agent with N-terminal as well as side chain amine groups [69]. When reduced HsPrx3 was exposed to H2O2 (1:2 concentration ratio) and then alkylated, the main products detected were the dimeric oxidized disulfide form of the enzyme as well as alkylated derivatives ( Figure 5B and Table 2). However, when reduced HsPrx3 was treated with a 1:1 concentration ratio of 15(S)-HpEPE and then treated with iodoacetamide, the main products had an increase in mass of +89, consistent with CAM addition (+57) plus hyperoxidation (+32) ( Figure 5C and Table 2). When the concentration of 15(S)-HpEPE was further increased (2:1 concentration ratio) the major species detected had a mass increment of +104, consistent with CAM addition plus the addition of 3 oxygen atoms ( Figure 5C and Table 2). This +104 modification was observed in the monomeric and also in the dimeric form of the enzyme exposed to 15(S)-HpEPE ( Table 2). The latter is ascribed to CysP hyperoxidation to sulfonic acid (+48), since similar addition of 2:1 concentration of 15(S)-HpEPE to CPSHsPrx3 yielded a +57 form, consistent with alkylation of the enzyme as the major product ( Figure 5D and Table 2).

Mass Spectrometry Analysis
In order to confirm the hyperoxidation of HsPrx3 treated with f FA-OOHs, MS analysis was performed. The predicted molecular weight of the reduced HsPrx3 subunit is 21,797 Da (13). Treatment with excess 2-iodoacetamide caused alkylation of reduced HsPrx3 (major peaks of + 113 and +170 which corresponds to the covalent addition of 2 and 3 carbamidomethyl (CAM) groups (+57 each), respectively, Figure 5A and Table 2). The presence of a small fraction of a form of the protein modified by 4 CAM groups in the reduced enzyme that has three Cys residues in its sequence is consistent with the reported side reaction of the alkylating agent with N-terminal as well as side chain amine groups [69]. When reduced HsPrx3 was exposed to H 2 O 2 (1:2 concentration ratio) and then alkylated, the main products detected were the dimeric oxidized disulfide form of the enzyme as well as alkylated derivatives ( Figure 5B and Table 2). However, when reduced HsPrx3 was treated with a 1:1 concentration ratio of 15(S)-HpEPE and then treated with iodoacetamide, the main products had an increase in mass of +89, consistent with CAM addition (+57) plus hyperoxidation (+32) ( Figure 5C and Table 2). When the concentration of 15(S)-HpEPE was further increased (2:1 concentration ratio) the major species detected had a mass increment of +104, consistent with CAM addition plus the addition of 3 oxygen atoms ( Figure 5C and Table 2). This +104 modification was observed in the monomeric and also in the dimeric form of the enzyme exposed to 15(S)-HpEPE ( Table 2). The latter is ascribed to Cys P hyperoxidation to sulfonic acid (+48), since similar addition of 2:1 concentration of 15(S)-HpEPE to C P SHsPrx3 yielded a +57 form, consistent with alkylation of the enzyme as the major product ( Figure 5D and Table 2).

Binding of ANS and AA to HsPrx3
The addition of increasing concentrations of ANS to HsPrx3 (5 µM) in sodium phosphate buffer 50 mM, pH 7.8 plus 0.1 mM DTPA, showed an increase in ANS quantum yield (λ ex = 375 nm) and a blue shift when compared with equal concentrations of ANS in the same buffer in the absence of the enzyme ( Figure 6A). Additionally, ANS caused a decrease in tryptophan-dependent HsPrx3 fluorescence emission at 339 nm (λ ex = 295 nm) ( Figure 6A) consistent with FRET, as previously reported for several proteins that bind ANS, including the peroxiredoxin AhpE from Mycobacterium tuberculosis [41]. Relative intrinsic fluorescence change (1 − F/F 0 ) exhibited a hyperbolic behavior when plotted against ANS concentration, resulting in saturation curves from which the value of the dissociation constant (K d ) of 65 ± 4 µM and number of binding sites (n) of 1.34 ± 0.05 were obtained. The inset in Figure 6B shows a Scatchard linealization of the data obtained at 2 µM HsPrx3, which yielded similar parameters. The addition of the saturated fatty acid palmitic acid at concentrations below its CMC (60 µM) [70] did not compete with the binding of ANS to the enzyme (not shown).  CPSHsPrx3-2CAM * the predicted molecular weight of wild type and CPSHsPrx3 reduced subunit are 21,797 Da [13] and 21,781 Da, respectively.  Relative intrinsic fluorescence change (1 − F/F0) exhibited a hyperbolic behavior when plotted against ANS concentration, resulting in saturation curves from which the value of the dissociation constant (Kd) of 65 ± 4 µM and number of binding sites (n) of 1.34 ± 0.05 were obtained. The inset in Figure 6B shows a Scatchard linealization of the data obtained at 2 µM HsPrx3, which yielded similar parameters. The addition of the saturated fatty acid palmitic acid at concentrations below its CMC (60 µM) [70] did not compete with the binding of ANS to the enzyme (not shown). The apparent melting temperature (Tm) of reduced HsPrx3 (4 µM) increased in the presence of AA (10 µM) and did not further increase in the presence of higher AA concentrations (20 µM), with Tm values of 64.6 ± 1 • C for HsPrx3, 68.1 ± 1 • C for HsPrx3 + AA (10 µM) and 67.2 ± 0.3 • C for HsPrx3 + AA (20 µM) (AA critical micelle concentration ranges between 10-60 µM [71,72]). Addition of similar concentrations of FA solvent (ethanol) caused no change in protein Tm value (Supplementary Figure S4). This~4 • C shift is consistent with the ability of HsPrx3 to interact with AA [73].

Modeling 15(S)-HpETE Binding to HsPrx3
The HsPrx3 dimeric unit arising from MD simulations using PDBid 5jcg chains A and B as starting point presents a hydrophobic patch close to Cys P with a solvent accessible surface area value of 256 ± 26 Å 2 ( Figure 7A). This region presents hydrophobic residues from both chains of the dimer, namely: Y39, P40, F43, F45, L142, P143, V167 and P181'. The lowest energy binding modes of 15(S)-HpETE obtained by molecular docking simulations use this hydrophobic region for binding as shown in Figure 7B. Among these conformations, the most populated best ranked cluster presents the hydroperoxide group of 15(S)-HpETE in proximity to the thiolate of Cys P ( Figure 7C). Those were also the structures of overall lower energy, indicating that the hydroperoxide group of the f FA-OOH may also contribute to substrate pose.
(ethanol) caused no change in protein Tm value (Supplementary Figure S4). This ~4 °C shift is consistent with the ability of HsPrx3 to interact with AA [73].

Modeling 15(S)-HpETE Binding to HsPrx3
The HsPrx3 dimeric unit arising from MD simulations using PDBid 5jcg chains A and B as starting point presents a hydrophobic patch close to CysP with a solvent accessible surface area value of 256 ± 26 Å 2 ( Figure 7A). This region presents hydrophobic residues from both chains of the dimer, namely: Y39, P40, F43, F45, L142, P143, V167′ and P181'. The lowest energy binding modes of 15(S)-HpETE obtained by molecular docking simulations use this hydrophobic region for binding as shown in Figure 7B. Among these conformations, the most populated best ranked cluster presents the hydroperoxide group of 15(S)-HpETE in proximity to the thiolate of CysP ( Figure 7C). Those were also the structures of overall lower energy, indicating that the hydroperoxide group of the fFA-OOH may also contribute to substrate pose.

Discussion
HsPrx3 was almost inactive in catalyzing the reduction of 20 µM 15(S)-HpETE and 15(S)-HpEPE by HsTrx1 after 15 s required for appropriate mixing for measurements in a conventional spectrophotometer. Moreover, the enzyme lost its activity towards H 2 O 2 after being treated with those f FA-OOHs (Figure 1 and Table 1). However, the enzyme consumed 1.7 15(S)-HpEPE per reduced HsPrx3 under non-catalytic conditions (Figure 3), indicating that the lack of enzymatic activity shown in Figure 1 could be due to a rapid inactivation caused by hyperoxidation of Cys P . In the case of H 2 O 2 , in which hyperoxidation should not compete with resolution at low oxidant concentrations (see below) the stoichiometry was 0.8 hydroperoxide consumed per HsPrx3. Furthermore, one PGG 2 was consumed per HsPrx3 (Figure 3), in agreement with its lower inactivation by this oxidant under catalytic conditions (Table 1) compared with 15(S)-HpEPE. The fact that reduced HsPrx3 consumed less than one H 2 O 2 per protein reflects a minor fraction of oxidized enzyme (~20%) in our sample after reduction and excess DTT removal (Supplementary Figure S3) due to the presence of adventitial H 2 O 2 in buffer solutions as already reported [74].
The addition of equimolar or higher concentrations of 15(S)-HpETE or 15(S)-HpEPE to HsPrx3 caused hyperoxidation of Cys P as detected by Western Blot analysis (Figure 4). In consonance, MS confirmed hyperoxidation to Cys P sulfinic acid even when added at equimolar concentration with respect to the enzyme. Additionally, in the presence of excess 15(S)-HpEPE, HsPrx3 Cys P was not only oxidized to sulfinic acid but also formed a +48 derivative that was not detected in the C P S mutated form of the enzyme, which is therefore consistent with sulfonic acid formation, and that appears in the monomeric as well as in the dimeric species ( Figure 5). The formation of sulfonic acid in Cys P of HsPrx3 had already been reported in HsPrx3 not only in vitro but also in vivo under physiological conditions, in the mitochondria from old rats [49].
The rate constants of oxidation of HsPrx3 were similar for every f FA-OOHs tested so far (Figure 2), and to those reported for H 2 O 2 -and peroxynitrite-mediated oxidations (Table 3). However, the rate constants of HsPrx3 hyperoxidation differed in several orders of magnitude: 15(S)-HpETE and 15(S)-HpEPE were the fastest and H 2 O 2 was the slowest (Table 3).  These reactivities explain the rapid inactivation of HsPrx3 exposed to 20 µM f FA-OOH but not to 20 µM H 2 O 2 as observed in the coupled catalytic assay (Figure 1). Indeed, k hyperoxidation × [ROOH] is 520 s −1 for 15(S)-HpEPE , 340 s −1 for 15(S)-HpETE, 6.2 s −1 for PGG 2 and 0.022 s −1 for H 2 O 2 . Since hyperoxidation competes with resolution (2 s −1 ) for the sulfenic acid in Cys P of the oxidized enzyme, expected amounts of hyperoxidized enzyme can be calculated as~99% for 15(S)-HpEPE , and 15(S)-HpETE, a fraction that decreased to~70% for PGG 2 and~1% for H 2 O 2 under the conditions employed in Figure 4. Indeed, HsPrx3 hyperoxidation was detected at much lower concentrations of 15(S)-HpEPE , and 15(S)-HpETE than of H 2 O 2 and even at equimolar concentrations hyperoxidation at both Cys P of the dimeric functional unit of HsPrx3 is detected (as hyperoxidized monomers) ( Figure 4). However, there is still a substantial proportion of the enzyme with one Cys P bound to Cys R (as hyperoxidized dimers) indicating that Cys P of HsPrx3 was not~99% hyperoxidized. This probably reflects active site asymmetry in HsPrx3, a dodecameric enzyme under reduced state [75], as already proposed for this and other members of the Prx family [37,58,74]. In this respect, it was previously reported that only half HsPrx3 active sites were susceptible to hyperoxidation even at very high H 2 O 2 concentrations (20 mM), forming hyperoxidized dimers as main products [76]. The slower hyperoxidation rate constant of HsPrx3 by PGG 2 with respect to 15(S)-HpEPE , and 15(S)-HpETE is probably a consequence of its particular structure, since it is an endoperoxide-hydroperoxide, which could affect its interaction with HsPrx3 active site.
Biophysical experiments using ANS showed a dose-dependent quenching of the intrinsic fluorescence of HsPrx3 as well as a blue shift and an increase in ANS quantum yields in the presence of the protein ( Figure 6A-C). These data suggest the presence of a hydrophobic region in HsPrx3 accessible from the solvent and with nearby positive charges [61,62]. However, the Ka value obtained of 0.016 µM −1 is an order of magnitude lower than those reported for fatty acid binding proteins such as bovine serum albumin (0.32 µM −1 ) or even for the 1-Cys Prx from Mycobacterium tuberculosis AhpE (0.16 µM −1 ). Indeed, the Ka value for ANS binding to HsPrx3 is in the range of those attributed to the binding of ANS to molten globule-like states of proteins [77]. Therefore, the existence of a well-defined binding site for hydrophobic compounds such as FA was not unambiguously proven by this technique. The ability of HsPrx3 to bind fatty acids was further analyzed by measuring the effects of AA on the melting temperature of reduced HsPrx3 ( Figure 6D) and a model of interaction is proposed based on the results obtained using docking simulations (Figure 7). Indeed, in an inspection of the HsPrx3 dimer structure and dynamics, we found and characterized a hydrophobic patch in close proximity to the enzyme active site that can allocate f FA-OOHs in such a way that may assist on their reduction by Cys P . It should be taken into account that the binding of f FA-OOH to HsPrx3 is not of high affinity, since the non-oxidable fatty acid palmitic acid did not compete with ANS for the binding to HsPrx3. The effect of low-affinity binding of non-reactive parts of substrates in catalysis was initially described by Jenks and coworkers [78] and is nowadays ascribed at least partially to changes in activation entropy of the reaction [79]. That was indeed the case in the one-cysteine Prx from Mycobacterium tuberculosis AhpE, that reacted much faster with f FA-OOH than with H 2 O 2 at the expense of an important increase in the activation entropy of the Cys P oxidation reaction [41].
From the kinetics of f FA-OOHs reduction by different mitochondrial peroxidases and other potential mitochondrial targets previously and herein reported (Table 4), it emerges that HsPrx3 should be considered a preferential target for these species, particularly of 15(S)-HpETE and 15(S)-HpEPE. This is because of the fast rate constants of reaction combined with the high abundance of the enzyme in many cellular types, with reported concentration values in the 10-120 µM range [13]. In turn, GPx1 is oxidized by FA-OOHs with a similar rate constant (4 × 10 7 M −1 s −1 [28]) but its mitochondrial concentration is lower, so its contribution to f FA-OOH reduction is expected to be minor. Data on the kinetics of oxidation of HsPrx5 by f FA-OOH are lacking so far, but if the high reactivity of the enzyme with other organic hydroperoxides is maintained (~10 7 M −1 s −1 ) [80], then Prx5 would also represent a main target of f FA-OOH since its mitochondrial concentration can attain 16 µM [81]. Of course, these calculations rely on the concentrations of the different peroxidases under reduced state, which can decrease if reduction by the mitochondrial reduction systems rate limits catalysis, or if the fraction of inactive enzymes (which includes but is not limited to hyperoxidized protein) increases. Similar calculations cannot be performed for PGG 2 or similar endoperoxide-hydroperoxides since the kinetic data on its reactions with other mitochondrial targets are still lacking.
Previous data indicated a mitochondrial formation flux of O 2 •− of 0.8 µM/s in endothelial cells under normoglycemia, which can increase almost 10-fold under hyperglycemic conditions [82]. Most of it dismutates to H 2 O 2 and O 2 in a reaction catalyzed by Mnsuperoxide dismutase (MnSOD), a highly efficient enzyme abundant in the organelle. Other potential targets of O 2 •− include • NO, which depending on its steady-state concentrations, can compete for a fraction of O 2 •− and form peroxynitrite [4]. In sake of simplicity, we did not include in our simulations formation fluxes of • NO which largely vary depending on conditions and cell type [4]. Considering mitochondrial formation fluxes of H 2 O 2 in the 0.4-4 µM/s range and rate constants and concentrations indicated in Table 4, we determined expected yields of hyperoxidized Prx3 in the mitochondrial matrix as indicated in Table 5. Formation rates of L FA-OOH in mitochondrial membranes have been previously estimated as 0.1 µM/s [83]. Estimation of formation rates of f FA-OOH in the mitochondrial matrix are lacking, although they have been reported to largely vary under different physiopathological conditions [6]. So, we spanned a three order of magnitude range of f FA-OOH formation rates to investigate the expected effects on HsPrx3 hyperoxidation. Time course simulations that were one hour long were performed in order to avoid other variables such as hyperoxidized HsPrx3 sulfinic acid reduction by sulfiredoxin (which requires the translocation of sulfiredoxin, a cytosolic protein, to the mitochondria [84]) as well as protein turnover [85]. Due to the location of cytochrome c preferentially as-sociated to the outer surface of the inner mitochondrial membrane, reduction of H 2 O 2 (4.6 × 10 1 M −1 s −1 ) or f FA-OOH (k~10 4 M −1 s −1 ) by cytochrome c/cardiolipin complexes [86] were not considered in the simulations related to HsPrx3 hyperoxidation in the mitochondrial matrix shown in Table 5. However, with a calculated t 1/2 value in the ms range (t 1/2 = ln 2/∑ k × [peroxidases], see Table 4), the diffusion of a fraction of f FA-OOH formed in the mitochondrial matrix through the inner mitochondrial membrane to reach the cytochrome c/cardiolipin complex, a process that has been reported to be facilitated by particular proteins, is in principle possible. This would beparticularly important under conditions of increased oxidative stress, where the fraction of oxidized and hyperoxidized HsPrx3 increases [87,88].  [54]. Trx2 plus Grx2 concentration in mitochondria was assumed to sum 10 µM [12] and was fixed under reduced state in the simulations (reduction of the redoxins is not rate limiting peroxide consumption). The concentration of Trx 2 and Grx 2 varies with the cellular type [54]. b The kinetics of the reactions of reduced HsPrx5 with FA-OOHs is unknown so far. In these simulations, we assumed a rate constant similar to those of the enzyme oxidation by the non-natural organic hydroperoxides tert-butyl hydroperoxide and cumene hydroperoxides reported in [80]. The concentration of Prx5 was set as 16 µM as reported in [89]. The susceptibility of HsPrx5 to hyperoxidation was reported to be low, so we did not include it in the simulations [90]. c Mitochondria express GPx1 with a concentration~2 µM [91]. Although GPx4 can also be expressed in the mitochondria, particularly in testis, its expression levels are low in most somatic tissues and therefore it was not included in the simulations [30]. d Glutathione concentration was set at 5 mM and was fixed under reduced state, i.e., in these simulations it is assumed that reduction of glutathione by glutathione reductase/NADPH is faster than glutathione oxidation.
Under low physiological mitochondrial fluxes of H 2 O 2 of~0.4 µM/s, the amount of hyperoxidized HsPrx3 formed in the mitochondrial matrix in one hour is ≤1 nM, unless a concomitant flux of f FA-OOH in the order of 0.001-0.1 µM/s occurs. On the contrary, at higher fluxes of H 2 O 2 formation such as 4 µM/s, hyperoxidized enzyme can reach 10-100 nM levels in the absence of fluxes of f FA-OOH, but still the latter are required for a significant fraction (≥1%) of the enzyme to be hyperoxidized in an hour. For example, for 2% hyperoxidation of the enzyme in an hour, a concomitant flux of 0.1 µM/s of f FA-OOH is needed in mitochondria expressing 100 µM HsPrx3. In mitochondria expressing 20 µM HsPrx3 formation rates of f FA-OOH required for a similar percentage of HsPrx3 inactivation are in the nM/s range (Table 5). Thus, our results indicate that mitochondrial f FA-OOH contributes to the inactivation through hyperoxidation of HsPrx3, a modification that has been detected in cells under both physiological and pathological conditions [49,50,[92][93][94].
Similar data regarding the oxidation and hyperoxidation of the cytosolic Prx1 and Prx2 by f FA-OOH are lacking so far, with the only exception of the initial report by Cha et al. [34]. The peroxidatic active site in the three Prxs is highly conserved. More interestingly, all the residues that make up the hydrophobic patch in HsPrx3 (Figure 7) are conserved in Prx2 (Supplementary Figure S5). Furthermore, the reported rate constants of oxidation and hyperoxidation by H 2 O 2 are similar [13,59,95,96]. Thus, it is tempting to speculate that the rate constants of oxidation and hyperoxidation by FA-OOH would probably be also similar, differences in susceptibility to hyperoxidation arising from differences in rates constants of resolution as in the case of H 2 O 2 . If that is the case, and since Prx2 is the one with the lower resolution rate constant, it would be expected to be the enzyme more prone to hyperoxidation also by f FA-OOH among the three. This is very interesting due to the different cellular compartmentalization of Prx2 (mostly cytosolic) with respect to Prx3 and due to the fact that depending on cells and conditions, f FA-OOH can be preferentially formed in different cell compartments. However, future work is required to test this hypothesis.
Prxs and particularly, typical 2-Cys Prxs participate in redox signaling actions through different mechanisms. In the flood gate mechanism of redox signaling, hyperoxidation and the consequent inactivation of Prxs would allow the oxidation of other, less reactive hydroperoxide targets [97]. In the redox relay mechanism, Prxs act as sensors of hydroperoxides transferring the oxidation to signaling proteins [98,99]. We propose that HsPrx3 oxidation and hyperoxidation by f FA-OOH reported herein contribute to both signaling mechanisms, either by increasing enzyme hyperoxidation yields but also serving as sensors of f FA-OOH in mitochondria.

Conclusions
Herein, we demonstrate that HsPrx3 is rapidly oxidized and hyperoxidized by f FA-OOH, with rate constants that are in the order of ≥3.5 × 10 7 and~10 7 M −1 s −1 for the AAderived 15(S)-HpETE and EPA-derived 15(S)-HpEPE, respectively. Compared with other biologically relevant hydroperoxides such as H 2 O 2 and peroxynitrite, the rate constants of HsPrx3 oxidation by these f FA-OOHs are in the same order of magnitude, but hyperoxidation rate constants are considerably higher ( Table 3). The endoperoxide-hydroperoxide PGG 2 , also derived from AA, oxidized HsPrx3 almost as rapidly but showed a lower rate constant of hyperoxidation. We present a model of interaction of f FA-OOH with HsPrx3 in which the former binds to a hydrophobic patch of the enzyme and positions the hydroperoxide group in close proximity to Cys P that would facilitate the reaction. We propose that, as for other mitochondrial hydroperoxides such as H 2 O 2 and peroxynitrite, HsPrx3 is expected to be a main target for f FA-OOH at least under low oxidative stress conditions in which most of the enzyme is under reduced state. Finally, our results indicate that low fluxes of f FA-OOHs (in the nM/s range) formed in the mitochondrial matrix are expected to increase hyperoxidation yields of HsPrx3 in mitochondria under different conditions. Considering the key roles of Prxs in redox signaling processes we propose HsPrx3 not only as a reductant but also as a sensor of f FA-OOHs formed in mitochondria.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antiox12020408/s1, Figure S1: Molecular representations of fFA-OOH and FA used throughout this work. Figure S2: Intrinsic fluorescence changes in HsPrx3 caused by CysP oxidation, resolution and hyperoxidation. Figure S3: Effect of 15(S)-HpETE versus H 2 O 2 on the distribution of HsPrx3 between monomeric and disulfide-bonded dimeric species. Figure S4: Effect of ethanol on the heat-induced unfolding of HsPrx3. Figure