Mass Spectral Analysis of Protein-based Radicals Using DBNBS NONRADICAL ADDUCT FORMATION VERSUS SPIN TRAPPING*

Protein-based radicals generated in the reaction of ferricytochrome c (cyt c ) with H 2 O 2 were investigated by electrospray mass spectrometry (ESI-MS) using 3,5-di-bromo-4-nitrosobenzenesulfonate (DBNBS). Up to four DBNBS-cyt c adducts were observed in the mass spectra. However, by varying the reaction conditions (0–5 molar equivalents of H 2 O 2 and substituting cyt c with its cya- nide adduct which is resistant to peroxidation), noncovalent DBNBS adduct formation was inferred. Nonethe-less, optical difference spectra revealed the presence of a small fraction of covalently trapped DBNBS. To probe the nature of the noncovalent DBNBS adducts, the less basic proteins, metmyoglobin (Mb) and a -lactalbumin, were substituted for cyt c in the cyt c /H 2 O 2 /DBNBS re- action. A maximum of two DBNBS adducts were observed in the mass spectra of the products of the Mb/ H 2 O 2 /DBNBS reactions, whereas no adducts were detected following a -lactalbumin/H 2 O 2 /DBNBS incubation, which is consistent with adduct formation via spin trapping only. Titration with DBNBS at pH 2.0 yielded noncovalent DBNBS-cyt c adducts and induced folding of acid-denatured cyt c , as monitored by ESI-MS and optical spectroscopy, respectively. Thus, the noncovalent DBNBS-cyt c mass adducts observed are Methods— Reactions were carried out by mixing 500 m M protein with 10 m M DBNBS and 0, 1, or 5 molar equivalents of H 2 O 2 in 50 m M ammonium acetate solution (pH 7.5) containing 200 m M DTPA. The reactions were terminated after 10 min by injecting the reactants onto a Vydac C 18 column (4.6 3 300 mm) and separating the protein from salts and low molecular weight reactants by HPLC (HP1090, Hewlett-Packard) using a 10–55% acetonitrile (ACN) gradient in 0.05% triflu- oroacetic acid at 1 ml/min over 20 min. The protein peak was collected, lyophilized, suspended in 1:1 methanol/water with 0.5% acetic acid, and infused at a flow rate of 5 m l/min using a syringe pump (Harvard Apparatus) directly into the electrospray ionization (ESI) source of a Finnigan SSQ7000 mass spectrometer (ThermoFinnigan) for molecular weight determination. The protein was further digested with 1:50 (w/w) trypsin at 50 °C for 4 h. The digests were separated for peptide mass mapping on a Vydac microbore C 18 column (1 3 300 mm) using a 10–55% ACN gradient in 0.05% trifluoroacetic acid at 40 m l/min over 100 min. The spray voltage was set at 4.5 kV, and the capillary tem- perature was maintained at 210 °C, while the sheath and auxiliary gas pressures were 40 and 15 p.s.i., respectively. Full-scan acquisition was performed in profile mode using scan rates of 320–380 atomic mass units/s.Toprobe DBNBS interaction with acid-unfolded cyt c , the protein was incubated for 30 min with increasing amounts of DBNBS in 0.05% trifluoroacetic acid (pH 2.0) before carrying out optical or MS measure- ments. For molecular weight determination, samples were infused into the ESI source of the mass spectrometer by flow injection at 50 m l/min in 75% ACN containing 0.05% trifluoroacetic

Reactive oxygen species, such as H 2 O 2 and superoxide, are generated by all aerobic cells as by-products of a number of metabolic reactions and in response to various stimuli. Oxidative damage can occur when H 2 O 2 reacts with heme proteins, such as ferricytochrome c (cyt c), 1 to form highly reactive oxy-ferryl-heme and transient protein-based radical species (X ⅐ ) that are linked to the initiation of lipid peroxidation (1,2). Detection of X ⅐ in biological systems is often difficult because they are short-lived and highly reactive. Spin traps, which are diamagnetic compounds containing a functional group that reacts with X ⅐ to form a more stable paramagnetic adduct (XST ⅐ ), are frequently used in electron paramagnetic resonance (EPR) investigations (3). Although EPR signals can provide information about a radical center and its environment, the specific sites of radical formation in biomolecules are not identified. Coupling of high performance liquid chromatography (HPLC) and mass spectrometry (MS) has been used to identify spin adducts of various small molecules (4 -6). Our research group has extended the use of LC/MS of spin adducts to proteins to overcome the inherent limitations of EPR. We have found that conversion of the spin adduct XST ⅐ to a stable diamagnetic mass adduct (XMA) via ascorbate reduction permits the assignment of XMA to a specific amino acid residue when spin trapping and peptide mass mapping by on-line LC/MS (ST/LC/MS) are coupled (7)(8)(9)(10). In addition to the increased specificity offered by ST/LC/MS, it possesses enhanced sensitivity over EPR, since considerably smaller quantities (picomole versus nanomole) of sample can be analyzed (11).
3,5-Dibromo-4-nitrosobenzenesulfonate (DBNBS), sometimes referred to as Perkin's trap, was developed to trap carbon-centered radicals. It is stable to temperature and light, and the introduction of the sulfonate group onto the benzene ring has helped overcome problems encountered in the use of lipophilic nitroso spin traps such as 2-methyl-2-nitrosopropane (MNP) (12). However, it is known that DBNBS adducts can be formed through several nonradical reactions. For example, prolonged incubation of DBNBS with unsaturated fatty acids (13) or with free tryptophan (14) causes chemical modifications through nonradical reactions such as the ene reaction between the nitroso group in DBNBS and the double bond in tryptophan.
Barr and co-workers (11,15) recently reported the trapping of a protein-based tyrosyl radical by DBNBS in the reaction of cyt c with H 2 O 2 using EPR. In addition, peaks corresponding to (DBNBS) n -cyt c adducts, with n ϭ 1-4, were observed by MALDI-MS, suggesting that as many as four protein-based X ⅐ species were trapped by DBNBS during the reaction of cyt c with 5 molar equivalents of H 2 O 2 . However, the sites of DBNBS-cyt c adduct formation were not identified. The radicals formed in the reaction of cyt c with H 2 O 2 are of interest, since they could cause mitochondrial membrane damage and play a role in the apoptotic process (16 we detected DBNBS-cyt c mass adducts in the absence of H 2 O 2 , and in the cyanide-ligated protein, even though cyanide binding inhibits heme-mediated peroxidation (11,17).
Cyt c, a highly basic protein (pI ϳ10), unfolds at low pH, resulting in increased absorption and blue-shifting of the Soret maximum from 408 at pH 7.0 to 394 nm in the acid-denatured protein at pH 2.0 (18). The addition of anions converts the unfolded state of cyt c to a conformation resembling a molten globule by reducing the electrostatic repulsion of positive charges on the protein surface via Debye-Hü ckel screening and ion pairing (18,19). We provide data here that support the formation of noncovalent adducts under the MS conditions (pH 2.0); in fact, DBNBS was found to stabilize a molten globule state of acid-unfolded cyt c at pH 2.0 in a manner similar to that reported for 8-anilino-1-naphthalenesulfonate (ANS) (19). Our results reveal that caution must be used in interpretation of the mass spectra of DBNBS-protein adducts, and the different limitations in EPR and MS approaches to the analysis of spin-trapped species are also highlighted.

EXPERIMENTAL PROCEDURES
Materials-Horse (Type VI), cow and tuna (Type XI) heart cytochromes c, horse heart metmyoglobin (Mb), ␣-lactalbumin from bovine milk, trifluoroacetic acid, and DBNBS were purchased from Sigma and used without further purification. H 2 O 2 was purchased from Fisher, while potassium cyanide (KCN) was obtained from BDH Chemicals. Sequencing grade trypsin (Roche Molecular Biochemicals) was used for digestion, and ammonium acetate (JT Baker Chemical Co.) Solutions containing diethylenetriamine-N,N,NЈ,NЉ,NЉ-pentaacetic acid (DTPA, ICN) were prepared using 18-M⍀-cm water obtained from a Barnstead Nanopure system.
Methods-Reactions were carried out by mixing 500 M protein with 10 mM DBNBS and 0, 1, or 5 molar equivalents of H 2 O 2 in 50 mM ammonium acetate solution (pH 7.5) containing 200 M DTPA. The reactions were terminated after 10 min by injecting the reactants onto a Vydac C 18 column (4.6 ϫ 300 mm) and separating the protein from salts and low molecular weight reactants by HPLC (HP1090, Hewlett-Packard) using a 10 -55% acetonitrile (ACN) gradient in 0.05% trifluoroacetic acid at 1 ml/min over 20 min. The protein peak was collected, lyophilized, suspended in 1:1 methanol/water with 0.5% acetic acid, and infused at a flow rate of 5 l/min using a syringe pump (Harvard Apparatus) directly into the electrospray ionization (ESI) source of a Finnigan SSQ7000 mass spectrometer (ThermoFinnigan) for molecular weight determination. The protein was further digested with 1:50 (w/w) trypsin at 50°C for 4 h. The digests were separated for peptide mass mapping on a Vydac microbore C 18 column (1 ϫ 300 mm) using a 10 -55% ACN gradient in 0.05% trifluoroacetic acid at 40 l/min over 100 min. The spray voltage was set at 4.5 kV, and the capillary temperature was maintained at 210°C, while the sheath and auxiliary gas pressures were 40 and 15 p.s.i., respectively. Full-scan acquisition was performed in profile mode using scan rates of 320 -380 atomic mass units/s.
To probe DBNBS interaction with acid-unfolded cyt c, the protein was incubated for 30 min with increasing amounts of DBNBS in 0.05% trifluoroacetic acid (pH 2.0) before carrying out optical or MS measurements. For molecular weight determination, samples were infused into the ESI source of the mass spectrometer by flow injection at 50 l/min in 75% ACN containing 0.05% trifluoroacetic acid. Optical spectra were recorded on a Beckman DU 650 spectrophotometer between 200 and 600 nm, using a scan rate of 2400 nm/min. Difference spectra were generated using Origin 3.0 software (MicroCal).

Analysis of DBNBS Mass Adduct Formation with Ferricytochrome c in the Presence and Absence of H 2 O 2 -
The deconvolved ESI mass spectrum (Fig. 1a) of horse cyt c revealed that (DBNBS) n -cyt c adducts (n ϭ 1-4) were formed in the reaction of cyt c with 5-fold molar excess of H 2 O 2 , as observed previously by MALDI-MS (11,15). A similar pattern of (DBNBS) n -cyt c adduct formation was observed when cyt c was reacted with only 1 molar equivalent of H 2 O 2 (Fig. 1b) despite the fact that H 2 O 2 , a two-electron oxidant, can remove a maximum of two electrons to generate two radical sites on cyt c. The reaction of horse cyt c with 1-5 molar equivalents of H 2 O 2 was repeated several times in the presence of DBNBS, and multiple DBNBS adducts were consistently observed. Moreover, in the absence of H 2 O 2 , up to two DBNBS adducts were detected not only with horse cyt c, but also with bovine and tuna cyts c (Fig. 1, c-e). These data suggest that DBNBS interacts with cyt c via a non-spin-trapping mechanism.
The cyanide ion serves as a high affinity ligand for the ferric state of cyt c by displacing the axial Met 80 ligand (17,20). Hence, cyanocyt c formation, which inhibits heme-catalyzed reactions and prevents cyt c heme degradation by organic hydroperoxides (11,17), was used to probe direct (i.e. non-heme mediated) peroxidation of the polypeptide. Fig. 1f reveals that the addition of 3 mM KCN to the horse cyt c/H 2 O 2 /DBNBS reaction results in the same mass spectrum as that observed in the absence of H 2 O 2 (Fig. 1, f versus c). Therefore, the (DBNBS) n -cyt c adducts from the cyanocyt c/H 2 O 2 /DBNBS reaction are unlikely due to peroxidation, supporting the absence of non-heme-mediated oxidation of cyt c residues by H 2 O 2 . Nonetheless, the mass spectra reveal that the (DBNBS) n -cyt c adduct intensities increase in the presence of H 2 O 2 ( Fig. 1, a and b versus c); hence, spin trapping of X ⅐ by DBNBS is also likely occurring, consistent with the EPR data that indicated trapping of a tyrosyl radical in cyt c (11).
To further probe the nature of the DBNBS adducts detected by MS, optical spectra of the reaction products obtained for horse cyt c in the absence and presence of H 2 O 2 were recorded (Fig. 2). The difference spectrum (Fig. 2, trace 3) of HPLCpurified products from the cyt c/H 2 O 2 /DBNBS (1:5:20) reaction minus those from the cyt c/H 2 O 2 (1:5) reaction reveals loss of DBNBS absorption seen at 288 nm (Fig. 2, trace 2) and growth of a new absorption band at 300 nm. This is consistent with increased conjugation of the chromophore, where the nitroso group of DBNBS traps a radical on the aromatic ring of a tyrosine residue (11). The possibility that the 300 nm absorption was due to H 2 O 2 -induced oxidation of DBNBS was also considered, but incubation of DBNBS with 5 molar equivalents of H 2 O 2 gave rise to a DBNBS species with an absorption spectrum (Fig. 2, trace 1) essentially identical to that of untreated DBNBS (Fig. 2, trace 2). Moreover, the cyt c/DBNBS (1:10) minus cyt c difference spectrum (data not shown) resembles that of DBNBS alone (Fig. 2, trace 2). As an additional control, the cyt c/H 2 O 2 (1:5) minus cyt c difference spectrum (Fig. 2, trace 4) was generated. A relatively flat base line between 230 and 370 nm with negligible UV absorption was observed, suggesting minimal oxidation of aromatic residues by H 2 O 2 . Thus, the new band at 300 nm in the cyt c/H 2 O 2 /DBNBS minus cyt c/H 2 O 2 difference spectrum (Fig. 2, trace 3) is not an artifact due to the subtraction procedure, but can be assigned to a protein-based DBNBS spin adduct. Unfortunately, tryptic digestion of the products of the cyt c/H 2 O 2 /DBNBS reaction yielded only native peptides. It has been observed that ascorbate reduction of MNP-Mb spin adducts yields stable XMAs that can be identified by ST/LC/MS (9,10). However, addition of 5 mM ascorbate to the cyt c/H 2 O 2 /DBNBS reaction did not convert XST ⅐ to XMA, since the mass spectrum (data not shown) was the same as that in Fig. 1a. This is likely due to either low trapping efficiency of DBNBS toward cyt c radicals or instability of the covalently trapped DBNBS-cyt c adducts under the digestion conditions. The radical is bonded directly to the nitroxide in the XST ⅐ formed with DBNBS, which renders reverse or cleavage reactions of the spin-trapped species favorable (3). Interestingly, Kim and co-workers (21) observed by EPR that the decay of 5,5-dimethylpyrroline-N-oxide (DMPO) adducts of Mb and hemoglobin were accelerated by denaturation (urea or guanidine HCl) and proteolysis of the protein moiety. The data in Fig. 2  Many reactions other than spin trapping have been reported for DBNBS. For example, the formation of a nitroxyl free radical was detected by EPR after a 60-min incubation of DBNBS with free tryptophan (14). Chemical modification after 24-h incubation of low density lipoprotein by DBNBS was detected by agarose gel electrophoretic mobility (13). The DBNBS labeling was assigned to the ene reaction between the nitroso group and a double bond, which results in an allylic hydroxylamine (22). However, both reported studies indicated that labeling required prolonged (Ն1 h) exposure to DBNBS. We limited cyt c exposure to DBNBS to 10 min and immediately separated the protein from the low molecular weight reactants by HPLC.
The possibility of "inverted spin trapping" was also considered, which could occur if the cyt c/H 2 O 2 /DBNBS reaction yielded an oxidized form of the spin trap (DBNBS ⅐ ϩ ). Inverted spin trapping of DBNBS ⅐ ϩ by an amino acid residue could lead to a DBNBS-cyt c adduct identical to that formed by "normal" spin trapping of X ⅐ by DBNBS (reviewed in Ref. 23  Procedures") were collected, lyophilized, and suspended in 1:1 methanol/water with 0.5% acetic acid. Samples (ϳ1 mg/ ml) were directly infused into the ESI source of the mass spectrometer at 5 l/ min for molecular weight determination. sumed that the E°DBNBS ⅐ ϩ /DBNBS is comparable with E°M NP ⅐ ϩ /MNP (2.06 V) (23), which is beyond the reach of the high oxidation states of heme proteins under non-forcing conditions. Hence, "inverted spin trapping" is ruled out in the cyt c/H 2 O 2 /DBNBS reaction, since (i) the reaction time was only 10 min and (ii) the Fe IV ϭ O center is rapidly converted to Fe III by endogenous electron transfer to heme (data not shown). Also, the reduction potential of the Fe III /Fe II couple in horse cyt c (E°ϭ 0.25 V) is clearly insufficient to oxidize the nitroso group.
DBNBS could also form sulfonamides with the ⑀-amino groups of lysines, which account for ϳ20% of the residues in cyt c. However, the deconvolved mass spectra (Fig. 1, a-f) reveal ⌬m increments of 360 Ϯ 2 Da for the cyt c adducts, corresponding to ammoniated DBNBS adducts, whereas sulfonamide formation would give rise to ⌬m increments of 326 Da. Furthermore, peptide mass mapping of the products of the cyt c/H 2 O 2 / DBNBS reaction yielded exclusively native horse cyt c peptides (data not shown), indicating that no stable derivatives such as sulfonamides were formed. This agrees with the work of Kalyanaraman and co-workers (13) who reported that DBNBSmodified low density lipoprotein was formed by a lysine-independent process.

Analysis of DBNBS Mass Adduct Formation with Metmyoglobin and ␣-Lactalbumin in the Presence and Absence of H 2 O 2 -To establish whether or not DBNBS adduct formation
with proteins exhibits specificity, Mb and ␣-lactalbumin were selected for further MS investigations. Mb, a heme-containing protein with a pI of 7.0, is not reported to have a high affinity for anionic compounds (29). Spin trapping of X ⅐ generated in the Mb/H 2 O 2 reaction has been demonstrated by both EPR and ST/LC/MS using DBNBS and other spin traps, and X ⅐ has been assigned primarily to Tyr 103 (10, 30 -33). The difference spectrum (data not shown) of HPLC-purified products from the Mb/H 2 O 2 /DBNBS (1:5:20) reaction minus those from the Mb/ H 2 O 2 (1:5) is similar to that seen with cyt c (Fig. 2, trace 3), which is indicative of DBNBS-tyrosine spin adduct formation in both proteins. An estimate of ⑀ 302 for the DBNBS-tyrosine spin adduct (42 mM Ϫ1 cm Ϫ1 ) was obtained by assuming a trapping efficiency of 70% for the tyrosyl radical in Mb, based on the relative intensities of the peaks in the mass spectrum of the Mb/H 2 O 2 /DBNBS (1:5:20) reaction products (Fig. 3a). Using this estimated ⑀ 302 , the yield of DBNBS-cyt c spin adduct formation in the cyt c/H 2 O 2 /DBNBS (1:5:20) reaction is ϳ20%, compared with ϳ10% estimated from EPR measurements (11). Yields of 10 -20% are sufficient to identify modified peptides by mass mapping, but tryptic digests of DBNBS-labeled cyt c contained exclusively native peptides, indicating that DBNBS labeling is not stable to peptide mass mapping, as discussed above.
In Fig. 3a, the deconvolved mass spectrum of the Mb/H 2 O 2 / DBNBS (1:5:20) reaction products shows the formation of (DBNBS) n -Mb adducts with n ϭ 1 and 2. DBNBS ene addition to Mb and/or ion pair formation can be ruled out, since only native globin is detected in the deconvolved mass spectrum in the absence of H 2 O 2 (Fig. 3b). Inverted spin trapping was not reported in the Mb/H 2 O 2 /DBNBS reaction (26), since the oxyferryl heme of Mb (E°Fe IV ϭ O/Fe III ϳ 1 V) cannot oxidize DBNBS. When H 2 O 2 was present at 1 molar equivalent (Fig.  3c), the relative intensity of the (DBNBS) 2 -Mb peak was negligible and that of the (DBNBS)-Mb peak decreased by 40% compared with the 1:1 adduct following oxidation with 5 molar equivalents of H 2 O 2 (Fig. 3a). Likewise, a less intense (DMPO)-Mb peak was observed by ESI-MS upon decreasing the amount of H 2 O 2 from 3 to 1 molar equivalents in the Mb/H 2 O 2 /DMPO reaction (32). It was reported that formation of oxyferryl Mb requires Ͼ1 molar equivalent of H 2 O 2 , since H 2 O 2 is consumed in side reactions at the porphyrin or other locations on the globin (9,34). Nevertheless, the efficiency of (XST ⅐ ) n -Mb adduct formation depends on the number of oxidizing equivalents present in the Mb/H 2 O 2 /ST reaction for both DMPO and DBNBS, in contrast to the cyt c/H 2 O 2 /DBNBS reaction (Fig. 1,  a versus b).
DBNBS adduct formation with ␣-lactalbumin was also investigated here, since it is an acidic protein with a pI ϳ4 and contains basic residues evenly dispersed over its surface. In fact, due to its high negative charge under the spin trapping conditions used (pH 7.5), ␣-lactalbumin should repel the negatively charged DBNBS. Similarly, Matulis and co-workers (35) observed by fluorescence spectroscopy that very little ANS bound to bovine serum albumin (pI 5.8) at pH Ͼ 11. ␣-Lactalbumin also lacks the heme prosthetic group found in Mb and cyt c; thus there should be no reaction with H 2 O 2 to generate X ⅐ unless H 2 O 2 directly oxidizes the polypeptide. As expected, the deconvolved mass spectra of ␣-lactalbumin both in the presence (Fig. 4a) and absence (Fig. 4b) of H 2 O 2 revealed no DBNBS adduct formation, since only native ␣-lactalbumin was detected. This rules out direct peroxidation of the polypeptide, consistent with the results for cyanocyt c (Fig. 2f), and it also rules out ion pair formation. observed by monitoring the Soret band upon incubation of cyt c with ANS or with various strong acids and their neutral salts (18,19). The molten globule state of cyt c, which forms at high salt (500 mM NaCl) and low pH (pH 2.0), is characterized by a red shift from 394 nm (the Soret maximum of acid-denatured cyt c) to 400 nm, which is accompanied by band broadening (18). To elucidate the mechanism by which the (DBNBS) n -cyt c mass adducts (Fig. 1) are formed, acid-unfolded horse cyt c was titrated with DBNBS at pH 2.0, and changes in the Soret and mass spectra were monitored. The Soret maximum red-shifted from 394 to 402 nm following 30-min incubation of cyt c with 25 molar excess DBNBS, consistent with molten globule formation (Fig. 5a). Similar effects were observed with ANS (19), which promotes the refolding of cyt c at low pH by Debye-Hü ckel screening and ion pair formation (18). The mass spectra in Fig.  1, a and b versus c, reveal enhanced binding of DBNBS to the H 2 O 2 -oxidzied cyt c, which is also seen in the greater red shifting of the Soret bands in Fig. 5, b versus a. Of note, the absorbencies of the acid denatured oxidized cyt c (⑀ 394 ϭ 140 mM Ϫ1 cm Ϫ1 ) and its DBNBS-stabilized molten globule (⑀ 402 ϭ 94 mM Ϫ1 cm Ϫ1 ) are less than those of the unoxidized forms (⑀ 394 ϭ 209 mM Ϫ1 cm Ϫ1 ; ⑀ 394 ϭ 120 mM Ϫ1 cm Ϫ1 ).
DBNBS titration of horse cyt c was also monitored by MS (Fig. 6). The (DBNBS) n -cyt c (n ϭ 1-4) peaks in the mass spectra of samples with Ն10-fold excess DBNBS recorded with- out prior HPLC purification are more intense than the corresponding peaks in the HPLC-purified sample (Fig. 6, c-f versus Fig. 1c). Fig. 6 also reveals that the abundance and stoichiometry of the (DBNBS) n -cyt c adducts increased as a function of DBNBS concentration, which is consistent with noncovalent adduct formation. Ali and co-workers (19) observed peaks corresponding to (ANS) n -cyt c adducts (n ϭ 1-7) at pH 2.0 by ESI-MS, but no adduct formation was observed with nile red, a neutral hydrophobic dye, revealing the importance of electrostatic interactions.
Precipitation of cyt c was observed upon addition of a large excess of DBNBS at pH 7.5, which required limiting the amount of DBNBS added to 500 M cyt c to 10 mM. Ion pair formation between the sulfonate group of DBNBS and basic residues of cyt c would reduce the solubility of the protein.
Interestingly, cyt c is known to possess anionic binding sites; specifically, there are two phosphate binding sites, one near Lys 87 with a dissociation constant (K d ) of 200 M and another close to Lys 25 -His 26 -Lys 27 with a K d of Ͼ2 mM (36). However, the DBNBS-cyt c adducts must possess K d values in the low M range, since phosphate-cyt c adducts are not observed by ESI-MS following reversed-phase HPLC purification, whereas DBNBS-cyt c adducts are seen in Fig. 1, c-e. In fact, a K d of ϳ36 M was estimated from a double-reciprocal plot (⌬Abs Ϫ1 versus [DBNBS] Ϫ1 ) of the data in Fig. 5 for both oxidized and nonoxidized horse cyt c, similar to the calculated K d (3-50 M) for the interaction of ANS to cationic polyamino acids at pH 2.0 (35).
Conclusions-DBNBS complexes with cyt c via strong electrostatic interactions at pH 2.0, thereby complicating the analysis of spin trapping by ST/LC/MS. However, noncovalent DB-NBS adduct formation clearly shows specificity for cyt c (Fig. 1, c-e versus Figs. 3b and 4b), indicating that its formation cannot simply be correlated with the total number of lysine residues, since horse cyt c possesses 19, while horse Mb and cow ␣-lactalbumin have 19 and 12, respectively. Lysine residues are highly conserved in cyts c and are clustered predominately around the exposed heme edge, forming anionic binding sites (36). In Mb and ␣-lactalbumin, the lysine residues are more or less distributed evenly over the entire protein surface. Also, the fact that DBNBS and ANS exhibit comparable efficiencies in inducing cyt c folding at low pH, despite the significantly larger hydrophobic moiety of ANS, is consistent with electrostatic interactions being the principal determinant of binding. Ion pair formation between proteins and negatively charged probes such as DBNBS and ANS may be a common occurrence at the low pH values used for ESI-MS analyses in positive ion mode. Therefore, it is essential to carry out the appropriate controls before interpreting MS data involving protein-probe adducts such as the (DBNBS) n -cyt c mass adducts observed in the present study. Finally, our results reveal that compared with Tyr 103 of Mb, the cyt c radicals are not very reactive and/or accessible at pH 7.5.