Cytochrome c autocatalyzed carbonylation in the presence of hydrogen peroxide and cardiolipins

Cytochrome c (cyt c) is a small hemoprotein involved in electron shuttling in the mitochondrial respiratory chain and is now also recognized as an important mediator of apoptotic cell death. Its role in inducing programmed cell death is closely associated with the formation of a complex with the mitochondrion-specific phospholipid cardiolipin (CL), leading to a gain of peroxidase activity. However, the molecular mechanisms behind this gain and eventual cyt c autoinactivation via its release from mitochondrial membranes remain largely unknown. Here, we examined the kinetics of the H2O2-mediated peroxidase activity of cyt c both in the presence and absence of tetraoleoyl cardiolipin (TOCL)- and tetralinoleoyl cardiolipin (TLCL)-containing liposomes to evaluate the role of cyt c–CL complex formation in the induction and stimulation of cyt c peroxidase activity. Moreover, we examined peroxide-mediated cyt c heme degradation to gain insights into the mechanisms by which cyt c self-limits its peroxidase activity. Bottom-up proteomics revealed >50 oxidative modifications on cyt c upon peroxide reduction. Of note, one of these by-products was the Tyr-based “cofactor” trihydroxyphenylalanine quinone (TPQ) capable of inducing deamination of Lys ϵ-amino groups and formation of the carbonylated product aminoadipic semialdehyde. In view of these results, we propose that autoinduced carbonylation, and thus removal of a positive charge in Lys, abrogates binding of cyt c to negatively charged CL. The proposed mechanism may be responsible for release of cyt c from mitochondrial membranes and ensuing inactivation of its peroxidase activity.

upon apoptosis onset, reaching submillimolar levels (14). In vitro experiments with liposomes confirmed the permeabilization of CL-containing lipid membranes via the described processes (15,16). Thus, the formation of cyt c-CL complexes triggers a functional shift of the protein function from electron transport to peroxidase activity (9), which, via CL oxidation, contributes to the intrinsic apoptotic pathway (5,8,17).
As it was highlighted by Kagan et al. (9), the fatty acid composition of CL species influences the binding and subsequent unfolding of cyt c. Furthermore, CL acts not only as a passive template for the induction of peroxidase activity in cyt c but actively provides substrates for its enzymatic activity. CL-induced peroxidase activity of cyt c leads to lipid hydroperoxide (LOOH) formation in CL, which further promotes its enzymatic activity as well as the subsequent release of cyt c from the mitochondria (7,18). However, it remains to be clarified whether pre-existing lipid hydroperoxides in CL may be sufficient to initiate the proposed autocatalytic cycle or whether H 2 O 2 is still mandatory for the cyt c peroxidase-derived permeabilization of the mitochondrial membrane upon apoptosis induction (19).
Several studies report an increased degradation of cyt c upon complex formation with CL (20), most likely due to H 2 O 2 -mediated heme degradation and subsequent iron release (8,21,22). Cyt c oxidative modifications were also observed in the absence of CL after incubating the protein with H 2 O 2 (21,23). Thereby, both cyt c-derived peroxidase activity and Fenton chemistry were described as possible oxidative pathways, suggesting that peroxidase activity of cyt c and a subsequent heme degradation may be interlinked (24). It remains unclear whether CL-dependent degradation of the protein is a result of its peroxidase activity and subsequent heme degradation by released radical by-products.
Here, we examined the kinetics of the H 2 O 2 -mediated peroxidase activity of cyt c both in the absence and presence of CL-containing liposomes to evaluate the role of cyt c-CL complex formation in the induction and promotion of its enzymatic activity. Thereby, the role of the CL fatty acyl chains was also considered, thus expanding comparable recent studies (25). Subsequent heme degradation in cyt c was evaluated to gain insights into the self-limiting mechanisms occurring during the cyt c-based peroxidase activity, providing a basis for free radical chemistry. Furthermore, over 50 oxidative modifications of cyt c resulting from the enzymatic activity of cyt c-CL complexes were studied using a bottom-up proteomics approach, and a new mechanism of autocatalyzed cyt c carbonylation via Tyr-derived cofactor leading to the reduction of protein positive charge and dissociation of cyt c-CL complex was proposed.

H 2 O 2 -and cardiolipin-dependent peroxidase activity of cytochrome c
We first performed kinetic measurements to investigate cyt c-based peroxidase activity in the sole presence of H 2 O 2 . The averaged kinetic curves (Fig. 1A) clearly show a time-dependent 2,2Ј-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid (ABTS) radical formation upon addition of increasing H 2 O 2 concentra-tions (straight lines). Three different kinetic phases were observable, namely a lag phase with low enzymatic activity, a subsequent steady-state phase of constant ABTS oxidation followed by a complete cease of the peroxidase activity. Thereby, at higher H 2 O 2 concentrations (bolder lines), shorter lag phases, higher steady-state enzyme activities, and a quicker deactivation of the latter were observed, apparently also leading to lower ABTS radical yields. In the absence of H 2 O 2 (dashed line) or cyt c (not shown), no ABTS oxidation took place. Thus, and 500 M) a shorter lag phase, a higher linear ABTS ϩ⅐ formation rate, and a quicker deactivation of the cyt c-based peroxidase activity were observed. In B, the linear increase of the cyt c-derived enzymatic activity with H 2 O 2 concentration is shown. In C, the coinciding exponential decrease in the relative efficiency of the cyt c-derived peroxidase activity is illustrated. In A, averaged kinetic curves from three independent experiments are shown; in B and C, the corresponding mean (points) and S.D. (error bars) values are given. H 2 O 2 activates and promotes cyt c-based peroxidase activity but subsequently also leads to its autocatalytic limitation.
A replot of the steady-state ABTS radical formation rate against the applied amount of H 2 O 2 (Fig. 1B) showed a linear concentration dependence, which indicates pseudo-first order reaction conditions and translates to a k obs value of 6.0 Ϯ 0.5 M Ϫ1 s Ϫ1 (Fig. 2B). By comparing the obtained ABTS radical yield with the applied H 2 O 2 concentration, the relative efficiency of the cyt c-based peroxidase activity was also calculated, again assuming formation of two ABTS radicals per peroxidase cycle. However, even at 25 M H 2 O 2 , only the formation of 20.13 M ABTS radicals was observed, corresponding to a relative yield of 40.3 Ϯ 3.5%. With increasing hydrogen peroxide concentrations, this value considerably decreased (Fig. 1C). Thus, a strong H 2 O 2 -mediated promotion of cyt c-based peroxidase activity translates to lower relative enzymatic efficiencies due to quick enzyme inactivation.
We next tested whether the presence of CL influences the H 2 O 2 -dependent peroxidase activity of cyt c by applying large unilamellar vesicle (LUV) liposomes with different lipid compositions. The kinetic measurements ( Fig. 2A) clearly showed that both in the absence (black) and in the presence of tetraoleoyl cardiolipin (TOCL)-containing liposomes (blue) no cyt c-based peroxidase activity was observed in the absence of H 2 O 2 (dashed lines), but TOCL strongly promoted the cyt c enzymatic activity in the presence of H 2 O 2 (straight lines). A shorter lag phase and an about 5-fold higher subsequent H 2 O 2dependent steady-state ABTS radical formation rate ( Fig. 2B; 26.7 Ϯ 1.6 M Ϫ s Ϫ1 ) were observed as compared with the lipidfree samples. Thus, TOCL promotes the H 2 O 2 -mediated cyt c-based peroxidase activity, which also translates to an about 2.6-fold higher relative product yield ( Fig. 2A; 53 M ABTS radicals).
Upon application of tetralinoleoyl cardiolipin (TLCL)-containing liposomes (red), cyt c-based peroxidase activity was observed even in the absence of H 2 O 2 ( Fig. 2A, dashed line), yielding about 20 M ABTS radicals after 1 h. Moreover, the three kinetic phases described above were clearly visible. In the additional presence of H 2 O 2 ( Fig. 2A, straight line) a shorter lag phase, an unchanged subsequent steady-state ABTS radical formation rate, and a quicker cease of peroxidase activity were observed, yielding similar overall product yields (22 M after 40 min). As TLCL does not influence the H 2 O 2 concentration-dependent cyt c-based peroxidase activity (Fig. S1), a comparable rate was determined ( Fig. 2B; 4.3 Ϯ 0.6 M Ϫ1 s Ϫ1 ) as compared with the lipid-free measurements. Thus, TLCL is able to replace H 2 O 2 in terms of activation and promotion of cyt c-based peroxidase activity but shows no cumulative effects in the additional presence of H 2 O 2 .
Control measurements with 1-palmitoyl-2-linoleoyl-sn-glycero-3-phosphocholine (PLPC)-containing liposomes (green) showed no effect of this lipid on the cyt c-based peroxidase activity either in the absence or presence of H 2 O 2 . This holds for the kinetics of ABTS radical formation ( Fig. 2A) as well as for the calculated H 2 O 2 -dependent steady-state peroxidase activity of cyt c ( Fig. 2B; 6.0 Ϯ 1.0 M Ϫ1 s Ϫ1 ). Thus, the reported effects of TOCL and TLCL on cyt c-based peroxidase activity are cardiolipin-specific.
Control measurements were also performed at continuous H 2 O 2 production by applying the glucose oxidase (GO)glucose system. Again, the ABTS-based measurements showed sigmoidal kinetics (Fig. S2). Both in the absence of lipid (black) and in the presence of TOCL-containing liposomes (blue), higher constant H 2 O 2 production rates translated to shorter lag phases, higher steady-state enzyme activities, and a quicker cease of cyt c-based peroxidase activity. TOCL strongly promoted those effects. Accordingly, in the presence of the lipid, about 3.5-fold higher relative enzymatic efficiencies of cyt c were observed as calculated from the amount of ABTS radicals formed after 180 min. However, both in the absence and in the presence of TOCL-containing liposomes, the relative product yield again decreased at higher H 2 O 2 production rates.

H 2 O 2 -and cardiolipin-derived degradation of cytochrome c
To further prove the connection between cyt c-based peroxidase activity and its subsequent inactivation, we monitored was added, and the formation of ABTS ϩ⅐ from 1 mM ABTS was followed for up to 3 h at 37°C by monitoring the absorbance increase at 734 nm. The kinetic curves in A show that PLPC (green) has no effect on the cyt c-based peroxidase activity either in the presence (straight line) or in the absence (dashed line) of H 2 O 2 . TOCL (blue) strongly promotes the H 2 O 2 -derived cyt c-based peroxidase activity. TLCL (red) also supports the H 2 O 2 -derived cyt c peroxidase activity promotion but also activates this enzymatic activity on its own. In B, the H 2 O 2 -derived increase in the peroxidase activity of cyt c is shown for the lipid-free system (black) as well as for liposomes with the named lipids. In A, averaged kinetic curves from three independent experiments are shown; in B, the corresponding mean (columns) and S.D. (error bars) values are given.

Autocatalyzed cytochrome c inactivation
spectral changes of cyt c upon its incubation with H 2 O 2 and/or cardiolipin. The cyt c spectrum ( Fig. 3A) clearly showed that in the absence of H 2 O 2 (dashed lines) almost no Soret band loss was observed within 3 h of incubation (gray to black), indicating no substantial heme destruction. However, increasing amounts of H 2 O 2 (bolder straight lines) led to a concentration-dependent decrease both of the Soret band at 409 nm and the lower ␣ and ␤ bands beyond 510 nm, indicating heme bleaching. By continuously monitoring the Soret band loss at 409 nm (  (Fig. S3, circles), thus providing additional proof for the H 2 O 2 -derived heme destruction as a cause for the autocatalytic inactivation of cyt c-based peroxidase activity. Thereby, the free iron concentration was determined by repeated sampling and application of the ferrozine assay. The corresponding calibration curve is given as Fig. S4.
Comparable measurements with TOCL-containing liposomes (Fig. 4A, blue) also showed no Soret band loss in the absence of H 2 O 2 (dashed line). However, in its presence (straight line), considerably faster but only partial heme degradation was observed. Again, a linear correlation between the k obs values (linear part of the kinetics) and the H 2 O 2 concentration was determined (Fig. 4B), yielding a k 2 value of 1.58 Ϯ 0.03 M Ϫ1 s Ϫ1 (Fig. 4C). Thus, the promoting effect of TOCL on H 2 O 2 -mediated cyt c-based peroxidase activity (Fig. 2) translates to an about 3 times faster heme degradation. PLPC (green) did not influence the H 2 O 2 -derived heme degradation (Fig. 4, A-C) as the lipid has no effect on cyt c-based peroxidase activity (Fig. 2, A and B).
The results in the presence of TLCL-containing liposomes (red) were also comparable with the data from the peroxidase activity measurements. Even in the absence of H 2 O 2 (Fig. 4A, dashed line), a short lag phase was followed by an exponential loss of the Soret band intensity, indicating heme destruction in about 77% of the cyt c proteins. In the additional presence of H 2 O 2 (straight line), an instant, even faster, and almost complete loss of the Soret band was observed, corresponding to the quick deactivation of the cyt c-based peroxidase activity observed under these conditions (Fig. 2). From the replot of the linear part of the Soret band loss against the H 2 O 2 concentration, a k 2 value of 2.48 Ϯ 0.21 M Ϫ1 s Ϫ1 was calculated, indicating an almost 5 times faster H 2 O 2 -derived heme degradation in the presence of TLCL. Thereby, the linear relationship is shifted to higher k obs values, illustrating the heme-degrading effect of this lipid in the absence of H 2 O 2 .

Effect of cardiolipin-derived lipid hydroperoxides
The results clearly show that TOCL promotes both cyt c-based peroxidase activity (Fig. 2) and subsequent heme degradation ( Fig. 4) only in the presence of H 2 O 2 , whereas TLCL promotes the named processes even in the absence of H 2 O 2 . This difference may result from the activation of cyt c-based peroxidase activity by pre-existing LOOHs present in TLCL preparations. It is important to note that TLCL used in the study was derived from a bovine heart preparation and thus contained some amount of endogenously or artificially produced LOOH species (Fig. S5). TLCL-derived lipid peroxides can lead to the formation of further LOOHs and thus support both peroxidase activity and subsequent heme loss of cyt c in an autocatalytic fashion. To prove this hypothesis, TLCL was preincubated with triphenylphosphine (PPh 3 ) to reduce the LOOH amount before preparing liposomes, performing peroxidase activity measurements, and studying heme degradation.

Autocatalyzed cytochrome c inactivation
even more so in the presence of H 2 O 2 (straight line) as compared with the lipid-free control (black). In contrast, application of reduced TLCL (violet) with or without H 2 O 2 supplementation resulted in lower peroxidase activity. Furthermore, higher ABTS radical yields were obtained as compared with nonreduced TLCL, indicating lower levels of heme degradation upon application of the lipid with a reduced LOOH content. These results prove a substantial contribution of LOOH to the self-limiting cyt c-based peroxidase activity.
By monitoring the Soret band loss for nonreduced TLCL (Fig. 5B, red), a complete heme loss was observed within the 3-h measuring time, both in the absence (dashed) and in the presence (straight) of hydrogen peroxide. Upon application of reduced TLCL (violet), heme degradation was partially inhibited, indicating a role of lipid hydroperoxides in peroxidasederived heme destruction in cyt c. However, still an almost complete heme loss was obtained at the end of the experiment. Most likely the peroxidase activity-mediated autocatalytic reformation of lipid hydroperoxides in the presence of molecular oxygen keeps the named enzymatic activity running, which, in turn, causes heme degradation and iron release. Indeed, measuring cyt c Soret band loss in the presence of TLCL under hypoxic conditions, a significantly lower rate of heme degradation was observed (Fig. S7), indicating the involvement of oxygen-mediated reactions.

H 2 O 2 -and cardiolipin-derived modification of cytochrome c
Previous results clearly demonstrated inactivation of cyt c during the peroxidase cycle in the presence of H 2 O 2 alone or in combination with cardiolipin. Importantly, TLCL even in the absence of H 2 O 2 was capable not only to activate cyt c peroxidase activity but also to induce its inactivation accompanied by heme destruction and subsequent iron release. Cyt c inactivation was proposed to be derived from oxidative degradation of the protein by reactive products released during the peroxidase cycle. To map oxidative post-translational modifications (oxPTMs) of cyt c (100 M) incubated alone or in the presence of hydrogen peroxide (1 mM) and/or TLCL (10 mM), we used a gel-based bottom-up proteomics approach to monitor over 50 modifications on 10 amino acid residues (Table S2). It is important to note that LC-MS/MS-based identification of the modification sites performed in this study provided only qualitative results and did not include any quantitative values.
Direct oxidation of Met, Trp, Pro, His, Phe, and Tyr amino acid residues was detected in all studied conditions (Table 1 and  Table S3). Both Met residues of cyt c (Met-65 and -80) were detected in the form of methionine sulfoxide. Interestingly, despite the high frequency of peptides carrying Met sulfoxides (especially for Met-65), further oxidation of Met residues to sulfones was not detected. Alternatively, both Cys residues (Cys-14 and -17) were found to be heavily oxidized to sulfonic acid but only in the samples incubated with H 2 O 2 (60 min), TLCL (30 and 60 min), and a combination of both (30 and 60 min). No intermediate oxidation states of Cys residues (sulfenic or sulfinic acids) were identified. The single Trp residue in position 59 was oxidized to hydroxy-Trp/oxindolylalanine (mass increment of 16 atomic mass units relative to unmodified Trp) and dihydroxy-Trp/N-formylkynurenine (32 atomic mass units) in all samples, and no further Trp oxPTMs were detected. Of four Tyr residues, three (Tyr-48, -67, and -97) were modified to dihydroxyphenylalanine (DOPA; Tyr-48, -67, and -97), DOPA quinone (DQ; Tyr-48 and -67), and trihydroxyphenylalanine (TOPA; Tyr-97). Tyr-74, however, was not detected in any of those three oxidation states. Single oxidation (ϩ16 atomic mass units on a proline residue can correspond either to hydroxylation or carbonylation to glutamic semialdehyde. Protein carbonylation, a well-known oxPTM formed usually via metal-catalyzed oxidation of Pro, Thr, Arg, and Lys residues, was reported previously for H 2 O 2 -treated cyt c (25). However, exact confirmation of carbonylated residues using tandem MS has not been provided so far to the best of our knowledge.
Here, we demonstrated oxidative carbonylation on at least 14 cyt c amino acid residues. Among carbonylated amino acid residue, the highest number of modified positions was assigned to Lys. Of 18 Lys residues, seven (Lys-39, -53, -55, -72, -73, -79, and -99) were oxidized to aminoadipic semialdehyde. Seven of eight available Thr residues were oxidized to 2-amino-3-ketobutyric acid (Thr-19, -28, -40, -49, -58, -63, and -78 but not Thr-102). As mentioned before, for Pro residues, both hydroxylation and carbonylation to glutamic semialdehyde result in the same mass increment of 16 atomic mass units and thus are indistin-guishable without carbonyl-specific derivatization. Overall, oxidation of Pro with a mass increment of 16 atomic mass units was detected for all four available Pro residues, including Pro-71 (only in TLCL-treated samples). Interestingly, neither of the two Arg residues (positions 38 and 91) have been shown to be carbonylated to glutamic semialdehyde.

Autogenerated trihydroxyphenylalanine quinone (TPQ)mediated Lys carbonylation
Tyr residues in cyt c are believed to play an important role not only in cyt c peroxidase catalytic activity but also in peroxidase cycle-induced enzyme inactivation. The proposed formation of Tyr radicals and the presence of di-Tyr cross-links in H 2 O 2treated cyt c were confirmed in several studies using specific di-Tyr fluorescence (7,23,26). Based on the mapping of oxidative modifications described above, it was rather surprising that Tyr-74 was not detected in either of the three considered oxidative states (DOPA, DQ, and TOPA). Thus, the possibility of di-Tyr cross-link formation was further evaluated.
Cross-linked peptides are particularly challenging for identification using bottom-up proteomics profiling. Indeed, almost no MS/MS-based identifications of sites of di-Tyr formation

Table 1 Summary of cyt c modifications identified by gel-based bottom-up LC-MS/MS experiments in the absence, or presence of H 2 O 2 , TLCL or combination of both
Cyt c (100 M) was incubated in ammonium bicarbonate buffer alone, in the presence of H 2 O 2 (1mM), TLCL (10 mM) or combination of both for 30-and 60-min. Proteins were separate by SDS-PAGE, digested with trypsin and analyzed by LC-MS/MS. PTMs were identified from obtained tandem mass spectra using search engine-assisted database search.

Autocatalyzed cytochrome c inactivation
are available. Recently, we developed a new MS-based approach to localize cross-linked Tyr residues in in vitro oxidized human serum albumin using mass increments of in silico predicted Tyr-containing tryptic peptides as variable modifications to be used in the search engine-based identification of di-Tyrlinked sequences (27). This strategy was proven to be successful for cyt c samples as well. Indeed, cross-linked peptides representing di-Tyr formation between Tyr-67 and Tyr-74 were identified in all experimental conditions, whereas peptides cross-linked via Tyr-48 (homodimer) and Tyr-48 and Tyr-74 were present only in the samples incubated with TLCL. These results demonstrated significant involvement of the residue at position 74 in Tyr radical-mediated protein cross-linking. That both Tyr modifications and extensive Lys carbonylation were present even in untreated cyt c samples was rather surprising and suggested a possibility of mechanistic correlation between those two modification types in cyt c.
Taking into account the presence of several oxidation states for cyt c Tyr residues accompanied by a high number of carbonylated Lys residues, we speculated that Lys deamination to carbonyl-containing aminoadipic semialdehyde via in situ produced TPQ "cofactor," similar to the enzymatic deamination of primary amines described for copper amine oxidase (28), might occur in cyt c under oxidative environment accompanied by the release of a free iron (Fig. S8). To confirm this hypothesis, mass increments corresponding to Tyr oxidation products leading to TPQ formation as well as their amino derivatives were calculated and used as variable modifications for the database search in our experiments. Those included already-described Tyr oxPTMs such as di-Tyr cross-links, DOPA, TOPA, and DQ as well as new modifications, including TPQ, aminoquinol, and iminoquinol. Indeed, analysis of proteomics data resulted in identification of all proposed Tyr PTMs in cyt c samples (Fig. 6). Of course, MS-based experiments do not provide full structural assignment of modified residues. However, considering the specific elemental composition of modified residues (e.g. C 9 H 14 N 2 O 4 for aminoquinol as a residue) and derived mass increments relative to Tyr residue (e.g. 33.0215 atomic mass units) in combination with the high resolution and mass accuracy of the Orbitrap mass analyzer, our data provide strong support for the proposed structures.
Thus, considering previously described (see "H 2 O 2 -and cardiolipin-derived modification of cytochrome c") and newly considered PTMs, we confirmed modification of Tyr-48 to di-Tyr (homo-and heterodimer with Tyr-74, both in TLCL-con-

Autocatalyzed cytochrome c inactivation
H 2 O 2 ), and aminoquinol (H 2 O 2 , TLCL, and combination of both). Overall, each Tyr residue showed certain specificity in terms of oxPTMs as well as sample treatment types with a general tendency for TPQ and iminoquinol to be present in H 2 O 2containing samples, independent of the presence of TLCL, with Tyr-48 and -74 showing the highest coverage of oxPTM along the whole proposed pathway (Fig. 6).

Cytochrome c modifications by electrophilic lipid peroxidation products
Finally, we investigated cyt c adducts between nucleophilic amino acid residues (Lys, Arg, His, and Cys) and electrophilic lipid peroxidation products, including low-molecular-weight aldehydes such as acrolein, glyoxal, and methylglyoxal as well as well-known ␣,␤-unsaturated aldehydes, hydroxynonenals (HNEs), and oxononenals (ONEs) formed by a ␤-scission of -6 polyunsaturated fatty acid at the C9 position. As expected, HNE and ONE Michael adducts were identified only in the samples coincubated with TLCL. In total, four modification sites were identified, including Lys-5 (ONE), His-33 (ONE and HNE), Lys-39 (ONE), and Lys-99 (ONE).
Adducts with glyoxal and acrolein were confirmed for His-26, Lys-27, His-33, Lys-39, Lys-53, Lys-72, Lys-86, Lys-88, Arg-91, and Lys-100. The majority of the modifications were identified in TLCL-treated samples; however, carboxymethyl-Lys was identified at positions 39, 53, and 88 even in the sample not exposed to TLCL during the experiments. Taking into account that cyt c used in the study was derived from a bovine heart preparation, these modifications might be either present in the original sample or derived from copurified lipids after oxidant exposure.

Comparison of cytochrome c modifications in the presence of H 2 O 2 and/or TLCL
Comparing oxidation sites between different samples, it can be seen that most of the oxPTMs were detectable already in untreated samples (control 0-, 30-, and 60-min incubation) (Fig. 7A). Once more, it is important to note that LC-MS/MSbased identification of modification sites performed in this study provided only qualitative results and did not include any quantitative values.
Interestingly, residues in the middle of the cyt c sequence (Fig. 7B) were oxidized in most of the samples independently of the treatment conditions (Thr-28, Pro-30, His-33, Phe-36, Lys-39, Thr-40, Pro-44, Tyr-48, Thr-49, Lys-53, Thr-58, Trp-59, Met-65, Tyr-67, Pro-76, and Met-80). Oxidation of Cys-14 and -17 to sulfonic acid occurred only in the samples treated with H 2 O 2 , TLCL, and a combination of both, indicating significant conformational changes in the cyt c heme-coordinating pocket in conditions promoting peroxidase activity and protein inactivation (Fig. 7, A-D). Additionally, specific oxidation induced in the presence of H 2 O 2 and TLCL included carbonylation of Thr-19 and Lys-79 as well as a carboxymethyl adduct on Lys-88. Furthermore, several site-specific oxidative PTMs were detected only in the samples treated with TLCL alone or in the presence of H 2 O 2 (His-26; cross-links between Tyr-48 and Tyr-48 and Tyr-48 and Tyr-74; Pro-71; and Lys-99; Fig. 7, C and  D). Interestingly, the presence of H 2 O 2 in TLCL-treated sam-ples did not result in any additional modification, indicating the absence of a cumulative effect between lipid and hydrogen peroxide exposure.

Cyt c carbonylation reduced its ability to bind CL
To demonstrate that cyt c carbonylation attenuates its binding to CL, the extent of cyt c carbonylation upon H 2 O 2 treatment was relatively quantified using cyt c labeling with a carbonyl-reactive fluorescent dye, 7-(diethylamino)-coumarin-3-carbohydrazide (CHH) (see Fig. 9 and Fig. S11A). Cyt c incubation with hydrogen peroxide for 30 and 60 min resulted in more than a 3-fold increase of total protein carbonyls. Using reverse-mode native PAGE, we demonstrated significant reduction in cyt c positive net charge upon oxidation, proportional to the increase of cyt c carbonylation signal ( Fig. 9 and Fig. S11B).
Furthermore, using a competitive binding assay utilizing nonyl acridine orange (NAO), we determined the decrease in binding of CL and cyt c upon oxidation (Table 2). Thus, the CL-cyt c binding constant (K b ) for control protein sample as well as cyt c incubated for 30 and 60 min in the absence of H 2 O 2 was estimated as 15 ϫ 10 7 M Ϫ1 . Incubation with hydrogen peroxide for 30 and 60 min led to a significant decrease of K b values (3-fold), providing a clear correlation among cyt c carbonylation, associated decrease of protein positive charge, and its ability to bind to CL.

Discussion
Recently, cyt c has received a lot of scientific attention due its role in apoptosis associated with cyt c release in cytoplasm and caspase activation. Cyt c forms a complex with the mitochondrion-specific lipid cardiolipin, which, upon oxidation, leads to the detachment of the protein from the inner mitochondrial membrane, permeabilization of the outer membrane, and subsequent release of cyt c to the cytoplasm (29). During this pathway, cyt c gains a peroxidase function, utilizing H 2 O 2 and lipidderived peroxides as substrates. However, the exact mechanism of transition of cyt c from low-to high-peroxidase state remains elusive. To address, at least partially, this question, we used a combination of biochemical methods with bottom-up proteomics for identification of cyt c oxPTMs. Recently, a role of H 2 O 2 -mediated oxPTMs in the promotion of the five-coordinated heme state of cyt c during precatalytic lag phase was proposed (30). Indeed, higher H 2 O 2 concentrations resulted in a shorter lag phase, consistent with the idea of faster oxidation rates of cyt c in the presence of high oxidant concentrations. TOCL promoted cyt c peroxidase activity, leading to the shorter lag phase and faster enzyme inactivation. Electrostatic interactions between positively charged residues of cyt c and negatively charged phosphate groups of CL are essential to cyt c-CL complex formation and believed to be a primary driving force in respect to hydrophobic interaction and hydrogen bonding. Recently, it was demonstrated that several sites on cyt c are engaged in CL-cyt c interactions governed by electrostatic forces (31). It was proposed that sites A and L, although located at the opposite end of the protein, can be seen as a joint binding site containing hydrophobic and positively charged residues. Binding via such a noncontiguous site results in open-

Autocatalyzed cytochrome c inactivation
ing of the heme pocket via membrane pulling and thus promotes cyt c peroxidase activity.
Interestingly, TLCL alone in the absence of H 2 O 2 also resulted in activation of peroxidase activity, which we showed to be dependent on the presence of LOOH in TLCL preparations. Apparently, the combined effect of structural remodeling of cyt c via CL-cyt c complex formation and the presence of initially low levels of organic peroxides was sufficient to initiate the peroxidase cycle. We propose that high levels of TLCLmediated peroxidase activity, despite a low level of initial LOOH (2.2 M) as an initial substrate, can be explained by continuous LOOH production from native TLCL via free radicals generated as by-products of the peroxidase cycle (Fig. 8,  red route). In situ LOOH production results in a continuous substrate supply, thus fueling the peroxidase cycle, producing new radicals and creating a closed loop responsible for the fast enzyme inactivation. Indeed, upon removal of LOOH, peroxidase activity of cyt c-TLCL closely resembled the TOCL curve. Interestingly, we did not observe a cumulative effect between nonreduced TLCL and H 2 O 2 on reaction kinetics, and TLCL alone was capable to stimulate fast cyt c inactivation and degradation.
The majority of the studies on the role of the cyt c-CL complex and its peroxidase activity have been focused on CL oxidative modifications associated with the changes in mitochondrial membrane properties leading to its permeabilization (7), leaving the significance of cyt c oxPTMs for complex formation, peroxidase activity initiation, and cyt c dissociation from mem- In all studied conditions, we observed a high number of oxidative modifications covering the middle part of the cyt c sequence. However, heme-region amino acids were not significantly modified unless severe treatment was applied. Tyr residues and their PTMs are usually found to be relevant in peroxidase activity of classical enzymes as well as cyt c peroxidase (25). Roles of Tyr residues in the peroxidase cycle are usually associated with the formation of Tyr radicals via one-electron reduction of Compound I to Compound II. Tyr radical formation is often followed by its recombination to the corresponding dityrosine cross-links (23); however, the exact localization of PTM sites has never been provided. Here, we detected three different types of di-Tyr cross-links in cyt c involving Tyr-48, -67, and -74. The Tyr-67 hydrogen bond network is essential for heme coordination, and tyrosine may also serve as an electron donor upon Compound I to Compound II one-electron reduction (Fig. 8, green route). Furthermore, our results indicate significant involvement of Tyr-74 in this process because only di-Tyr hetero-but not homodimers between Tyr-67 and Tyr-74 were identified. Interestingly, involvement of Tyr-48, another surface-exposed residue, in Tyr radical-mediated di-Tyr crosslinks was evident only in TLCL-containing samples. It is interesting to note that cyt c Tyr residues are known to be involved in the regulation of apoptosis via other types of regulatory PTMs, namely phosphorylation and nitration. Thus, Tyr-48 phosphorylation was shown to possess antiapoptotic properties by impairing apoptosis-activating factor 1-mediated activation of caspase-9 (32). Although the direct role of Tyr oxidation in regulation of apoptotic signaling is not fully understood, Tyr-48 present in oxidized form would be unavailable for phosphorylation, thus limiting the antiapoptotic function of this modification via PTM cross-talk.
One of the most abundant oxidative modifications of cyt c was carbonylation. Cyt c carbonylation upon incubation with H 2 O 2 was already demonstrated by OxyBlot (23). Recently, H 2 O 2 -induced carbonylation of Lys-72 and -73 during initial lag phase of cyt c peroxidase catalysis was proposed as a mechanism of precatalyst activation (23). Despite detecting carbonylated cyt c proteoforms, the authors, however, did not provide any confirmation of proposed modification sites. Here, we detected 14 carbonylation sites, including seven Lys and seven Thr residues. Such a large number of carbonylation sites made us wonder whether this modification is mechanistically correlated with cyt c peroxidase activity and subsequent radicals formation. Keeping in mind the high involvement of Tyr residues and Tyr-derived PTMs in the peroxidase cycle as well as known reaction mechanisms of Lys carbonylation (33), analogy with the copper-containing enzyme lysyl oxidase, responsible for Lys deamination to aminoadipic semialdehyde in extracellular matrix proteins, first came to mind. However, instead of lysyl oxidase lysine tyrosyl quinone cofactor, we proposed the formation of TPQ, previously described for a copper amine oxidase catalyzing deamination of primary amines (23). Applying these reaction mechanisms to Tyr residues in cyt c, we monitored formation of corresponding TPQ, aminoquinol, and iminoquinol PTMs on both surface-exposed Tyr-48 and -74.
Detection of amino and imino group-containing derivatives of cyt c Tyr residues allows us to propose that autogenerated TPQ "cofactors" in cyt c can mediate Lys deamination to the observed aminoadipic semialdehydes on Lys-39, -53, -55, -72, -73, and -99. Interestingly, of 18 Lys residues, only seven positions were identified in the form of aminoadipic semialdehyde. Furthermore, carbonylated Lys residues are mainly located in site A and contribute to cyt-CL complex formation via electrostatic interactions.
Cyt c is rich in Lys residues, which represent 17% of its sequence versus 6% average Lys abundance in mammalian proteomes. Many Lys residues in cyt c participate in its interaction with lipid membranes and protein-binding partners (30). Our and previously published data indicate high susceptibility of cyt c Lys residues to a variety of modifications, each of which can potentially change the protein net charge and thus interaction surfaces for the binding partners. Indeed, coincubation of cyt c with HNE resulted in a shift of the protein pI from 9.7 to 9.25 (34). Furthermore, using site-directed mutagenesis, it was demonstrated that substitution of Lys-72 and -73 with Asn cancels CL-cyt c binding and CL-induced peroxidase activity, whereas Lys to Arg mutations at the same sites permit cyt c-CL interactions (35). We demonstrated that cyt c carbonylation was associated with the loss of protein positive charge (Fig. 9) and resulted in significantly lower binding constants with CL-containing liposomes, again confirming the importance of Lys residues in establishment of electrostatic interaction between cyt c and cardiolipin. Thus, it is tempting to speculate that similar to the well-studied histone PTM code maintained via methylation of Lys-rich protein tails, cyt c Lys PTMs, formed via autogenerated TPQ-mediated deamination, lead to the elimination of the positive charge and might be seen as a regulatory mechanism responsible for the switch of cyt c interaction partners (e.g. electron transport chain proteins, anionic lipids, or membrane detachment) and thus regulating its moonlighting nature or functional switch.  (n ϭ 3). b Comparisons between the control group (unoxidized cyt c at 0 min) and each other group were performed using the t test. *, significant differences with p Ͻ 0.05 versus control group; ns, not significant.

Liposome preparation
LUVs were prepared by extrusion (22,39). Briefly, singlelipid stock solutions (10 mg/ml in chloroform) were mixed and evaporated in a vacuum centrifuge (Vacuum Concentrator Centrifugal Evaporator R 10.22, Jouan, Fisher Scientific GmbH), and the obtained lipid film was dissolved in phosphate buffer (PB; 10 mM, pH 7.4) (for UV-visible spectroscopy) or ammonium bicarbonate buffer (ABC; 50 mM, pH 7.8) (for gel electrophoresis and MS) by thorough vortexing. The resulting multilamellar vesicles were extruded 19 times through a hydrophilic polycarbonate filter (Merck Millipore, Merck Chemicals GmbH; 25-mm diameter, 0.1-m pore size) in a mini extruder (Avanti Polar Lipids, Inc.). LUV size distribution was determined by nanoparticle tracking analysis using a NanoSight LM20 Nanoparticle Analysis System (NanoSight, Amesbury, UK) equipped with a 650 nm laser. For each sample, three measurements (60 s each) at 23.3°C were performed using a concentration of 10 7 -10 9 particles/ml. The lipid composition and size of the liposomes applied in the experiments are given in Table S1 where the most common lipid compositions are shown in italics.

Autocatalyzed cytochrome c inactivation
was determined spectroscopically (Tecan Infinite 200 Pro, Männedorf, Switzerland) at 830 nm by comparison with a calibration curve prepared with 6.4 -64 nmol of NaH 2 PO 4 .

Lipid hydroperoxide reduction
For selected experiments, LOOHs in the LUVs were reduced by incubating the chloroform stock solution of the lipid mixture (DOPC/TLCL 7:3) with PPh 3 (2 mM chloroform) for 1 h at 30°C in the dark (42). Chloroform was evaporated, and liposomes were prepared as described above. The amount of lipid hydroperoxides before and after PPh 3 reduction was determined by monitoring the enzymatic activity of HRP (0.1 M) in the presence of ABTS (1 mM) in 1 mM liposomes and comparing the obtained values with the ABTS oxidation rates observed in the presence of H 2 O 2 .

Peroxidase activity measurements
If not stated otherwise, samples for the peroxidase measurements were prepared in 96-well plates containing cyt c (5 M), liposomes (500 M lipid concentration), and ABTS (1 mM) in PB. After preincubation for 5 min at 37°C, H 2 O 2 (up to 500 M) was added using the injector device of the plate reader (Tecan Infinite 200 Pro, Männedorf, Switzerland), and the formation of ABTS ϩ ⅐ was followed by absorbance measurement at 734 nm (⑀ 734 ϭ 1.5 ϫ 10 4 M Ϫ1 s Ϫ1 ) (43) at 37°C for up to 3 h. From the linear range of absorbance increase curves, the initial ABTS oxidation rate was determined (5).
Selected experiments were also performed in the presence of GO and glucose as a continuous H 2 O 2 -producing system (21,44). Briefly, samples were prepared as described above in the additional presence of GO (up to 4.70 milliunits) and preincubated for 5 min at 37°C. Glucose (1 mM) was added using the injector device of the plate reader. H 2 O 2 production rates were adjusted by applying different amounts of GO (0.24 -4.70 milliunits). A calibration curve obtained from control measurements with HRP (0.1 M), ABTS (1 mM), glucose (1 mM), and different GO concentrations was consulted.

Heme degradation measurements
Cyt c (5-30 M) was incubated with an up to 100-fold excess of H 2 O 2 either in the absence or presence of a 100-fold lipid excess in PB (10 mM, pH 7.4). Spectral changes between 380 and 580 nm were followed at 37°C for 3 h. From the absorbance loss at 408 nm (Soret peak) (21,36), k obs values were calculated by plotting the linear part of the Soret band loss kinetics against the applied H 2 O 2 concentration. From the slope of these plots, k 2 values were calculated by considering the applied cyt c concentration.
Selected experiments were also performed to correlate (21) the Soret band loss with the release of iron from cyt c by applying the ferrozine assay (21,45) with slight modifications (46,47). Briefly, cyt c (30 M) was preincubated in PB (10 mM, pH 7.4) at 37°C. H 2 O 2 (3 mM) was added by using the injector device of the plate reader. The Soret band loss was followed at 408 nm for up to 3 h. Ascorbic acid (150 mM) was added to selected wells at fixed time points via the injector device of the reader to stop the reaction and reduce iron to the ferrous state. After completion of the absorbance measurements, parts of the samples were transferred either to a well containing ferrozine (2.5 mM; monosodium salt hydrate of 3-(2-pyridyl)-5,6-diphenyl-1,2,4-triazine-p,pЈ-disulfonic acid in 8% ammonium acetate solution) or to a well containing PB. The absorbance was measured at 562 nm (46) using the second well for zero compensation. The free iron concentration in the samples was determined by comparing the obtained net absorbance with a calibration curve generated with ferrozine and FeCl 3 under the same experimental conditions.
To monitor cyt c heme degradation at normoxic and hypoxic conditions, TLCL liposomes (1.5 mM in 10 mM PB) were prepared as stated above and incubated with cyt c (15 M) in PB in a Hellma Precision cell Suprasil cuvette. For normoxic conditions, the decrease of the Soret band (408 nm) absorbance was monitored over 90 min at 37°C. For hypoxic measurements, PB was sonicated (10 min at room temperature), and PB and cyt c stock solutions were degassed by vacuum application in an Eppendorf concentrator (10 min at room temperature) and purged with nitrogen (10 min at room temperature) before measurements of Soret band absorbance carried out in a nitrogen-flushed air-tight screw cap Hellma Precision cell Suprasil cuvette.

SDS-PAGE and in-gel digestion
Cyt c (100 M) was incubated in the absence or presence of TLCL-containing liposomes (10 mM) and H 2 O 2 (1 mM) in ABC at 37°C for 30 or 60 min. Samples were extracted using a tertbutyl methyl ether (MTBE) protocol (48). Briefly, 350 l of methanol and 1250 l of MTBE were added followed by 10 s of vortexing after each step. The mixture was left on a rotating shaker for 1 h at 4°C. To induce phase separation, 313 l of deionized water was added. Samples were left on a rotary shaker for an additional 10 min and centrifuged for 10 min at 1000 ϫ g. The upper phase was removed, and the lower aqueous phase together with protein precipitate was dried in the vacuum centrifuge for protein analysis. Dried samples were dissolved in Laemmli sample buffer (62.5 mM Tris, 2.5% (w/v) SDS, 0.2% (w/v) bromphenol blue, 5% ␤-mercaptoethanol, 10% (w/v) glycerol, pH 6.8), separated by SDS-PAGE (18% T, 1 mm), and stained with Coomassie Brilliant Blue G-250. Visible protein lanes were cut out, destained with acetonitrile (50% (v/v) in 50 mM NH 4 HCO 3 ), centrifuged (1 h at 37°C at 750 rpm), dehydrated with 100% acetonitrile, and dried by vacuum concentration. Proteins were digested with trypsin (375 ng in 3 mM ABC; 4 h at 37°C at 550 rpm), and peptides were extracted using consecutive incubations with 100, 50, and 100% acetonitrile (15-min sonication for each step). Combined extracts were vacuum-concentrated and stored at Ϫ20°C. Before MS analysis, peptides were dissolved in 10 l of 60% aqueous acetonitrile containing 0.5% formic acid and further diluted 1:20 with 3% aqueous acetonitrile.

LC-MS/MS of tryptic peptides
A nanoACQUITY UPLC (Waters GmbH) was coupled online to an LTQ Orbitrap XL ETD mass spectrometer equipped with a nano-ESI source (Thermo Fischer Scientific GmbH, Bremen, Germany). Eluent A was aqueous formic acid (0.1%, v/v), and eluent B was formic acid (0.1%, v/v) in acetoni-

Autocatalyzed cytochrome c inactivation
trile. Samples (10 l) were loaded onto the trap column (nano-ACQUITY Symmetry C 18 ; internal diameter, 180 m; length, 20 mm; particle diameter, 5 m) at a flow rate of 10 l/min. Peptides were separated on a BEH 130 column (C 18 phase; internal diameter, 75 m; length, 100 mm; particle diameter, 1.7 m) with a flow rate of 0.4 l/min using two step gradients from 3 to 30% eluent B over 18 min and then to 85% eluent B over 1 min. After an equilibration time of 12 min, samples were injected every 33 min.
The transfer capillary temperature was set to 200°C, and the tube lens voltage was set to 120 V. An ion spray voltage of 1.5 kV was applied to a PicoTip online nano-ESI emitter (New Objective, Berlin, Germany). The precursor ion survey scans were acquired in the Orbitrap (resolution of 60,000 at m/z 400) for an m/z range from 400 to 2000. Collision-induced dissociation tandem mass spectra (isolation width, 2; activation Q, 0.25; normalized collision energy, 35%; activation time, 30 ms) were recorded in the linear ion trap by data-dependent acquisition for the top six most abundant ions in each survey scan with dynamic exclusion for 60 s using Xcalibur software (version 2.0.7; Thermo Fisher Scientific, San Jose, CA). Peptides were identified using Sequest search engine (Proteome Discoverer 1.4, Thermo Scientific) against the UniProt database, allowing up to two missed cleavages and a mass tolerance of 10 ppm for precursor ions and 0.8 Da for product ions. The list of variable modifications used for the database search is provided in Table  S2. Results were filtered for rank 1 high-confidence peptides and score versus charge states corresponding to Xcorr/z 2.0/2, 2.25/3, 2.5/4, and 2.75/5.

Shotgun MS analysis of lipid hydroperoxides
Dried lipid-containing MTBE extracts were dissolved in ESI solution (methanol/chloroform 2:1 (v/v) containing 5 mM ammonium formate) and analyzed by direct infusion (15 l) using the robotic nanoflow ion source TriVersa NanoMate (Advion BioSciences, Ithaca, NY) and nanoelectrospray chips (1.4-kV ionization voltage, 0.4 p.s.i. nitrogen backpressure) coupled to an LTQ Orbitrap XL ETD mass spectrometer (Thermo Fischer Scientific) operated in negative ion mode. Transfer capillary temperature was 200°C, and the tube lens voltage was set to Ϫ120 V. Mass spectra were acquired with a target mass resolution of 100,000 at m/z 400 in the Orbitrap mass analyzer. Tandem mass spectra (MS/MS) were acquired in the linear ion trap (LTQ) using an isolation width of 1.5 Da and normalized collision energy of 35%. Data were analyzed manually using the XCalibur software (version 2.0.7).

Cyt c carbonylation analysis using CHH derivatization
Cyt c (100 M) in phosphate buffer (50 mM, pH 7.0) was incubated with or without H 2 O 2 (1 mM; 37°C) for 0, 30, and 60 min. Reactions were stopped by ultrafiltration (Vivaspin TM 2 ultrafiltration devices with a 5000-molecular-weight-cutoff polyethersulfone filter, Sartorius AG, Hannover, Germany), and protein concentration in recovered solutions was determined by Bradford assay (49). Cyt c (0.4 M) was incubated with CHH (150 M; 2 h at 37°C) and separated by SDS-PAGE (12% T) (50). Gels were washed twice with water (10 min) and visualized on a ChemiDoc TM MP (Bio-Rad Laboratories GmbH) using Image Lab TM software and a DyLight 488 channel filter for blue epi-illumination. Images were quantified using ImageJ (51).
Reduction of cyt c positive charge upon oxidation was studied using native PAGE (7.5% T) separation performed in reverse electrophoresis mode (11). Resulting gels were stained with Coomassie, and images were acquired on a ChemiDoc MP and quantified using ImageJ.

Cyt c-CL binding assay
Large unilamellar liposomes of POPC (5 mM) or POPC/TLC mixture (2.5 mM each) were prepared in phosphate buffer (50 mM, pH 7) containing 100 M diethylenetriaminepentaacetic acid by extrusion as described in under "Liposome preparation." Liposomes (1 mM total lipid) were incubated with cyt c (20 M; 30 min at room temperature), afterward NAO (4 mM; 1:4 ratio lipid/NAO) was added, and mixtures were incubated for 1 h (room temperature). 10 l of each sample was separated by native PAGE (7.5% T) in reverse electrophoresis mode. The fluorescence of NAO was monitored at 520 nm using a Chemi-Doc MP. Binding constants were calculated according to Belikova et al. (11).