Spin Trapping and Protein Cross-linking of the Lactoperoxidase Protein Radical*

Lactoperoxidase (LPO) reacts with H2O2 to sequentially give two Compound I intermediates: the first with a ferryl (FeIV=O) species and a porphyrin radical cation, and the second with the same ferryl species and a presumed protein radical. However, little actual evidence is available for the protein radical. We report here that LPO reacts with the spin trap 3,5-dibromo-4-nitroso-benzenesulfonic acid to give a 1:1 protein-bound radical adduct. Furthermore, LPO undergoes the H2O2-dependent formation of dimeric and trimeric products. Proteolytic digestion and mass spectrometric analysis indicates that the dimer is held together by a dityrosine link between Tyr-289 in each of two LPO molecules. The dimer retains full catalytic activity and reacts to the same extent with the spin trap, indicating that the spin trap reacts with a radical center other than Tyr-289. The monomeric protein recovered from incubations of LPO with H2O2 is fully active but no longer forms dimers when incubated with H2O2, clear evidence that it has also been structurally modified. Myeloperoxidase, a naturally dimeric protein, and eosinophil peroxidase do not undergo H2O2-dependent oligomerization. Analysis of the interface in the LPO dimers indicates that the same protein surface is involved in LPO dimerization as in the normal formation of myeloperoxidase dimers. Oligomerization of LPO alters its physical properties and may alter its ability to interact with macromolecular substrates.

LPO 1 (EC 1.11.1.7; donor-H 2 O 2 oxidoreductase), which oxidizes thiocyanate (HSCN) to hypothiocyanate (HOSCN) and more highly oxidized species, is a component of the mammalian antimicrobial defense system (1). It is closely related by sequence and function to MPO and EPO and also by sequence to thyroid peroxidase (2,3). LPO is not only of intrinsic interest but is also a useful model for the study of MPO and EPO, neither of which has been successfully expressed in other than mammalian cells (4).
Bovine LPO is a protein of 612 amino acids with a molecular mass of approximately 78,000 Da, approximately 10% of which is carbohydrate (2,5). The prosthetic group of the enzyme is a heme in which the 1-and 5-methyl groups bear hydroxyl groups esterified with the side chains of Glu-275 and Asp-125, respectively (4, 6 -8). The prosthetic heme in myeloperoxidase is bound via two similar ester bonds to the protein (9) but in addition is linked to the protein by a bond between the 2-vinyl group and a methionine residue (10,11). We established earlier that the prosthetic group is bound to the protein in LPO via a self-activating process in which noncovalently bound heme becomes covalently bound on exposure of the heme-protein complex to H 2 O 2 (4). Recent evidence indicates that a similar mechanism is involved in covalent attachment of the heme to the protein in EPO and thyroid peroxidase (12,13).
The peroxidative mechanism for LPO, as is true for all hemoprotein peroxidases, involves the following three-step sequence, where AH is a substrate (14). LPO  Compound II is a fairly well defined intermediate, with an Fe IV ϭO species in which the iron is one oxidation state above its resting ferric state. The nature of Compound I is less well defined. In horseradish peroxidase, the prototypical peroxidase, Compound I consists of a ferryl species and a porphyrin radical cation (15). In cytochrome c peroxidase, another well characterized peroxidase, Compound I consists of the same ferryl species plus a tryptophan-centered protein radical (16 -18). Evidence exists for both types of Compound I species in LPO (19 -22), in that LPO reacts with H 2 O 2 to give a ferryl/ porphyrin radical cation Compound I intermediate that is thought to convert rapidly to a ferryl/protein radical species (23)(24)(25). Reaction of the proposed LPO ferryl/protein radical Compound I with phenol (20) or ferrocyanide (21) involves preferential initial reduction of the ferryl species at low pH and the putative protein radical at high pH (20 -22). Transformation of the ferryl/porphyrin radical cation Compound I to an intermediate that retains the ferryl species but not the porphyrin radical cation clearly suggests that electron transfer from a protein residue quenches the porphyrin radical cation. However, no independent evidence exists for a protein radical beyond the fact that the Compound I species without the porphyrin radical cation retains two oxidation equivalents, as shown by the fact that it oxidizes two molecules of oneelectron-reducing substrates (20 -22).

EXPERIMENTAL PROCEDURES
Materials-All chemicals were of analytical grade and were purchased from Sigma or Roche Molecular Biochemicals. Lyophilized LPO from bovine milk (80 -150 units/mg of protein, A 412 /A 280 ϭ 0.88 -0.95) was also from Sigma. This preparation was shown to be homogeneous by gel filtration chromatography and was used without further purification. The sequencing grade modified porcine trypsin was purchased from Promega (Madison, WI). MPO from human polymorphonuclear leukocytes (180 -220 units/mg of protein) and EPO from human eosinophils (2960 units/mg of protein) were obtained from Calbiochem and U. S. Biochemical Corp., respectively. Prepacked Sephadex G-25 (PD-10) gel filtration cartridges were purchased from Amersham Pharmacia Biotech.
Analytical Methods-Absorption spectra were recorded on a Cary 1E UV-visible spectrometer (Varian, Victoria, Australia). Dityrosine content was measured with a Perkin-Elmer LS 50 B fluorimeter. High pressure liquid chromatography was carried out on a Hewlett Packard 1090 liquid chromatography system (Palo Alto, CA and quenched by adding catalase, as described above. The resulting solution was applied to a 2.6 ϫ 20-cm SP-Sepharose fast flow column (Amersham Pharmacia Biotech) equilibrated with 50 mM potassium phosphate, pH 6.8. The column was washed with 50 mM potassium phosphate buffer, pH 6.8, and then eluted using the stepwise procedure described in the text (0 -750 mM NaCl). The fractions corresponding to monomeric and oligomeric forms (as assessed by SDS-PAGE analysis) were pooled in two lots. Each lot was desalted and concentrated by ultrafiltration (Centricon 30, Amicon; molecular weight cut off ϭ 30,000). The concentrated samples were stored at 4°C for subsequent analysis. Trypsin Digestion of Oligomeric LPO-A 100-l sample of oligomer (300 M) was diluted to 2 ml with 6 M guanidine-HCl in 50 mM Tris-HCl, pH 8, then concentrated down to 100 l by diafiltration (Centricon 30, Amicon). After adding 20 l of 50 mM dithiothreitol, the oligomeric fraction was incubated at 60°C for 60 min. Trypsin was then added to a final protease:protein ratio of 1:20 (w/w), and the mixture was incubated at 37°C for 12 h. The reaction was stopped by injecting the final mixture directly onto a reverse phase HPLC column.
High Pressure Liquid Chromatography of the LPO Proteolytic Digest-Dityrosine-containing peptides were isolated by high pressure liquid chromatography on a Vydac 218TP54, 4.6 mm x 250 mm, C18 column eluted with a linear gradient going from 0.1% trifluoroacetic acid in water to 0.085% trifluoroacetic acid in 50% acetonitrile over 80 min. The eluent was monitored with a UV-visible detector at 216 and 280 nm. Fractions of 0.8 ml were collected at a flow rate of 0.8 ml/min. A 200-l aliquot of each fraction was collected and diluted to 500 l with a solution containing 8 M urea and 200 mM NaOH. After 315-nm excitation, the fluorescence intensities of the resulting mixtures were measured at 410 nm, which is the emission maximum of the Tyr-Tyr bond. Fractions were allowed to stand in the urea-NaOH solution for at least 30 min before reading the fluorescence intensity.
Spin Trapping Experiments-EPR measurements were performed with an ER/200D EPR spectrometer operating at 9.80 GHz with a TM cavity. X-Band first derivative absorption spectra were obtained with the following settings: microwave power, 25 mW; center field, 3480 G; time constant, 100 ms; sweep time, 50 s; modulation, 0.32 mT at a frequency of 100 kHz; and total sweep width, 125 G. Spectra were taken at 18 -22°C. The magnetic field range and center were estimated by comparing the EPR spectrum from an LPO/H 2 O 2 reaction mixture with that of the stable nitroso compound potassium nitrosodisulfonate. The potassium nitrosodisulfonate splitting was taken to be 13.091 G, and the center peak to correspond to a g value of 2.0056 (31). The reactions were initiated by adding H 2 O 2 to the mixture of LPO and the spin trap in 50 mM potassium phosphate buffer, pH 6.8, containing 200 M diethylenetriaminepentaacetic acid to inhibit possible catalysis by trace transition metals. For all EPR experiments, the reactions were allowed to proceed for 3 min. For the experiments shown in Fig. 1B, the LPO preparation subjected to reaction with H 2 O 2 in the presence of DBNBS was loaded onto a PD-10 column and eluted with the phosphate buffer containing diethylenetriaminepentaacetic acid.
Mass Spectrometry-MALDI experiments were carried out on a Voyager DESTR time-of-flight mass spectrometer operated in reflectron mode with delayed extraction or on a Voyager Elite time-of-flight mass spectrometer operated in linear mode with delayed extraction. 3,5-Dimethoxy-4-hydroxy cinnamic acid and ␣-cyano-4-hydroxycinnamic acid were used as the matrices for intact (i.e. nondigested) proteins and the HPLC-purified cross-linked peptide, respectively. Bovine serum albumin was used as the calibration standard in determining the molecular mass of intact proteins. For the first molecular mass screening of the tryptic digest of LPO, two-point external calibration was used based on angiotensin II (m/z 1046.54) and ACTH clip peptide (m/z 2465.2). The monoisotopic molecular mass of the cross-linked peptide was determined more accurately using a two point internal calibration, based on a coeluting LPO proteolytic peptide ( (353-362), MH ϩ at m/z 1124.56) and mellittin (MH ϩ at m/z 2845.76). The post-source decay (PSD) experiment was performed on the Voyager DESTR-TOF spectrometer, lowering the reflectron voltage by 25% for each subsequent frame. PSD data were smoothed to yield average masses.
Preparation and Isolation of Authentic Dityrosine-Authentic dityrosine was prepared by the LPO-catalyzed oxidation of L-tyrosine. A 1 mM L-tyrosine stock solution was prepared in 50 mM potassium phosphate buffer, pH 8.5. LPO (128 nM, final concentration) and H 2 O 2 (25 mM, final concentration) were sequentially added, and the resulting mixture was incubated at 25°C for 5 min. Excess H 2 O 2 was then consumed by incubation with catalase (1.25 units/ml) for 15 min. The resulting solution was injected onto a Vydac 218TP54, 4.6 x 250-mm, C18 reverse phase HPLC column. The column was eluted with a linear gradient going from 0.1% trifluoroacetic acid in water to 0.085% trifluoroacetic acid in 50% acetonitrile over 20 min at a flow rate of 0.8 ml/min. The eluent was monitored with a UV-visible detector at 275 nm. Fractions of 0.8 ml were collected. Fluorescence measurements on the collected fractions (performed as above) demonstrated the presence of only one 410-nm-emitting component following excitation at 315 nm. This component yielded a single ninhydrin-positive spot on silica gel with n-butyl alcohol/acetic/water (4:1:1) as the solvent system. It was identified as authentic dityrosine free of tyrosine impurities by mass spectrometry. Final concentrations were determined by measuring the absorbance at 315 nm using an extinction coefficient of 5.2 mM Ϫ1 cm Ϫ1 at pH 7.5 (32).

Spin Trapping of LPO Protein-derived Radicals by DB-NBS-
The addition of 1.2 eq of H 2 O 2 to LPO in the presence of 18 mM DBNBS produced an EPR spectrum characteristic of a highly immobilized nitroxide radical (Fig. 1A). The DBNBS/ LPO-derived radical adduct remained associated with the protein following gel filtration chromatography through a PD-10 column (Fig. 1B), demonstrating that the DBNBS-trapped radical was protein-bound. When DBNBS was incubated with LPO in the absence of peroxide and then reacted with H 2 O 2 after passing the sample through a PD-10 gel filtration column, the immobilized nitroxide was not observed (Fig. 1C). This rules out formation of the immobilized radical adduct by the ene reaction known to occur with DBNBS (33). Formation of the immobilized radical adduct required the presence of DBNBS, LPO, and H 2 O 2 (not shown). Inclusion of the LPO inhibitor azide (5 eq) prevented the formation of the DBNBS adduct (Fig.  1D), confirming the requirement for active peroxidase in this process. There was no effect when 5 eq of sodium azide was added after the spin-trapped adduct had already been formed. Interestingly, the addition of the two-electron donor substrate thiocyanate (120 eq) also prevented formation of the DBNBS adduct (Fig. 1E). The addition of 120 eq of potassium thiocyanate after the adduct had already been formed did not effect the spectra in any way (not shown). When the DBNBS/LPO-derived radical adduct was submitted to nonspecific proteolysis, an isotropic three-line spectrum was detected with a hyperfine coupling constant of 13.6 G (Fig. 1F). This was similar to that reported for proteolysis of the DBNBS/metmyoglobin, DBNBS/ cytochrome c, and DBNBS/cytochrome c oxidase adducts (34 -36).
Unmodified LPO and the DBNBS/LPO-derived radical adduct were subjected to MALDI mass spectrometric analysis. The unmodified LPO gave an average molecular weight of 77,658 ( Fig. 2A), whereas the calculated molecular mass from the primary protein sequence is 69,460 (2). The mass discrepancy of 8,198 Da is accounted for by the carbohydrate content of the protein, which represents about 10% of the total weight (2,5). The protein after reaction with DBNBS and H 2 O 2 exhibited a molecular ion ([M ϩ H] 1ϩ ) that was 340 Da higher than that of the unmodified protein (Fig. 2B). This increase in the mass of the parent protein corresponds within experimental error to the addition of 1 molecule of DBNBS (M r ϭ 344) to native LPO, suggesting that only one free radical site per protein molecule was trapped by the spin trap. No ions corresponding to unmodified LPO were observed after reaction with DBNBS, indicating that virtually all the LPO in the reaction mixture was converted into the DBNBS/LPO-derived radical adduct.
H 2 O 2 -mediated Cross-linking of LPO-SDS-PAGE analysis of LPO after incubation with H 2 O 2 showed that the monomeric protein (M r ϭ 80,000) was partially converted to dimeric (M r ϭ 184,000) and trimeric (M r ϭ 280,000) products (Fig. 3A). Oligomer formation increased to a maximum as the H 2 O 2 /LPO ratio rises to a value of 4 but did not increase further as the H 2 O 2 /LPO ratio increased further to a value of 8 (not shown). The maximum yields of the LPO dimer and trimer under these conditions were 49 and 5%, respectively. Furthermore, the remaining monomeric protein, after reisolation and purification, remained catalytically active but did not oligomerize when treated with H 2 O 2 . Formation of trimeric products suggests that at least two residues in separate locations on the LPO surface can participate in oligomerization reactions. These cross-linking reactions cannot be attributed to S-S bond formation, as the protein samples were treated with 2-mercaptoethanol before electrophoresis.
Various agents were found to inhibit peroxide-mediated cross-linking. The two-electron donor substrates iodide and thiocyanate and the one-electron donor substrate ferrocyanide prevented polymerization (Fig. 3A). The one-electron donor substrates ascorbate and ABTS and the suicide substrate 3-amino-1,2,4-triazole were equally effective (not shown). Interestingly, DBNBS also prevented polymerization. However, when DBNBS was removed from the system after the reaction by gel filtration chromatography through a PD-10 column, the monomeric protein oligomerized when re-exposed to H 2 O 2 (not shown).
The amino acid sequences of MPO, EPO, and LPO show a remarkable degree of similarity that ranges from 50 to 70% residue identity (37). This led us to investigate whether H 2 O 2 also causes covalent oligomerization of MPO and EPO. Fig. 3B shows the analysis of intact and H 2 O 2 -treated MPO, EPO, and LPO on a 20% SDS-polyacrylamide gel. As known (38,39) Isolation of Monomeric and Oligomeric LPO-In the initial stages of this work, separation and quantification of monomeric and oligomeric forms of LPO was accomplished by using a gel filtration column (Sephacryl S-300). Subsequent experiments showed that the oligomeric form can be obtained in higher yield by cationic ion exchange chromatography (SP-Sepharose fast flow). Typical elution patterns of LPO preparations submitted to gel filtration chromatography before and after incubation with H 2 O 2 are given in Fig. 4A. The intact LPO appeared as one major peak (peak 1). Treatment of the native protein with H 2 O 2 caused the formation of two new peaks of higher molecular weight (peaks 2 and 3). Calibration of the column with proteins of known molecular weight indicated that the earliereluting peak (peak 3) corresponded to a trimeric species of LPO with a molecular weight of about 250 kDa, and peak 2 corresponded to a dimeric species with a molecular weight of about 160,000. The molecular weight of peak 1 was in the range of that reported in the literature for native LPO (76,000 -80,000) (2,5,40). These results corroborate the evidence obtained by SDS-PAGE that in the presence of H 2 O 2 the monomeric protein is partially converted to dimeric and trimeric products. When LPO preparations were submitted to cationic ion exchange chromatography after incubation with H 2 O 2 , two major protein components were obtained as indicated in Fig. 4B. The inset shows the analysis of the peak 1 and peak 2 proteins on an SDS-polyacrylamide gel. As can be seen, peak 1 and peak 2 represent the monomeric and oligomeric forms of LPO, respectively. The two peaks were well resolved, indicating that the oligomerization reaction significantly altered the charge properties of the protein. Comparison of the areas of the two peaks at 280 nm indicates that approximately 50% of the LPO is converted to oligomers after incubation with H 2 O 2 .

Characterization of Monomeric and Oligomeric LPO-The
UV-visible spectra of the purified products were almost identical in the Soret (350 -450 nm) and visible (450 -700 nm) regions of the spectrum to that of the native enzyme (not shown), indicating that the heme environment of the enzyme was minimally altered by oligomerization. The influence of oligomerization on activity was investigated by using ABTS and KSCN as electron donor substrates. Plots of the rate versus the reductant concentration at a fixed H 2 O 2 concentration and the rate versus the H 2 O 2 concentration at a fixed reductant concentration are presented in Fig. 5. Figs. 5, B and C indicate that both the monomeric and oligomeric LPO were inhibited by high H 2 O 2 concentrations with KSCN as the substrate and by high ABTS concentrations. In contrast, the normal Michaelis-Menten kinetic profiles in Figs. 5, A and D show that neither form of the enzyme was inhibited by high concentrations of KSCN or of H 2 O 2 when ABTS was the substrate. Most importantly, the kinetic profiles obtained with the oligomeric form of LPO did not differ significantly from those obtained with the monomeric form, which indicates that oligomerization of LPO does not block access to the heme group or alter the kinetic mechanism of the reaction. Thus, the residue responsible for the oligomerization reaction must be located in a different region of the LPO surface than the entrance of the substrate access channel. Furthermore, the cross-linked residue is not critically involved with the residues responsible for the catalytic action of the enzyme. To determine whether oligomerization alters the ability of LPO to form a protein-centered radical, the spin trapping experiments were repeated with the monomeric and oligomeric fractions of LPO isolated by cationic exchange chromatography and, for control purposes, with the native form of the enzyme. As indicated in Fig. 6, the purified products form the DBNBSprotein adduct with an intensity similar to that seen for the native protein, indicating that the site at which the DBNBS adduct is formed is still accessible after oligomerization. This finding requires that at least two free radical sites exist on the LPO protein surface: the site responsible for the oligomerization reaction and the DBNBS binding site.
Previous investigations have established that the reaction of some hemoproteins with H 2 O 2 results in oligomerization of the protein due to the formation of tyrosine-tyrosine cross-links (41)(42)(43). The dityrosine cross-links can be detected by their characteristic ultraviolet fluorescence (44,45). This led us to determine the fluorescence spectrum of the monomeric and oligomeric forms of LPO (Fig. 7A). As can be seen, the oligomers fluoresce strongly when exposed to 320-nm light, whereas the monomer does not. The excitation and emission maxima of the oligomers (320 and 410 nm, respectively), as well as the pH dependence of the fluorescence, were similar to that of authentic, purified dityrosine (data not shown), providing strong evidence that the oligomer fluorescence was due to the presence of a dityrosine bond.
Identification of the Residues Involved in the Oligomerization Reaction-Proteolytic digestion of the LPO oligomer provides, after a high pressure liquid chromatographic purification step (Fig. 8), a fraction with a fluorescence spectrum identical to that of dityrosine (Fig. 7B). The molecular masses of the peptides in this fraction were determined by MALDI-TOF mass spectrometry. Some predicted LPO peptides were thus identified. The fraction contained a peptide (363-372) (MH ϩ at m/z 1124.68) and a series of disulfide-linked dimers, most likely formed via disulfide shuffling during the digestion, with molecular ions such as m/z 21115.10, 2540.34, and 3551.48, corresponding to peptides (140 -148)S-S(363-372), (571-579)S-S(280 -294), and (571-579)S-S (679 -701), respectively. MALDI molecular weight measurements also revealed a cross-linked peptide with MH ϩ at m/z 3208.54. Based on this mass, a Tyr-Tyr cross-linkage in addition to a disulfide link was suspected for tryptic peptide (280 -294). The sequence of this peptide is Ala 280 -Gly-Phe-Val-Cys-Pro-Thr-Pro-Pro-Tyr 289 -Gln-Ser-Leu-Ala-Arg 294 . MH ϩ for the monomer is at m/z 1606.8, whereas dimer formation via the Tyr-Tyr cross-linkage would yield an MH ϩ at m/z 3210.57. The calculated MH ϩ for a dimer formed via loss of 4 hydrogens is m/z 3208.55, which agrees well with the measured mass.
To confirm the identity of the cross-linked peptide, a PSD analysis was performed by selecting its molecular ion cluster as the precursor and following its metastable decomposition induced by MALDI (Fig. 9). Very few ions were observed, mainly in the low mass region (Ͻm/z 500). These ions, N-terminal (m/z 276 ϭ b 3 , 347 ϭ a 4 , 375 ϭ b 4 ) and C-terminal (m/z 175 ϭ y 1 , 229 ϭ y 2 -NH 3 , 446 ϭ y 4 ) fragments, some immonium ions, and internal fragments were sufficient to confirm that the (280 -294) peptide was involved in the cross-linking reaction (for nomenclature, see Ref. 46). They also suggested that at least four amino acids at the N terminus as well as the C terminus of at least one peptide were not modified. However, the PSD data did not provide further direct information on the chemical nature of the cross-link between the parent peptides.
A reduction experiment was performed by dissolving the peptides in 25 mM ammonium acetate buffer, pH 4.5, and incubating the mixture with 300 mM tris(2-carboxyethyl)phosphine at 53°C for 1 h. All the disulfide-linked peptides present in the fraction were reduced under these conditions, whereas the molecular mass of the suspected cross-linked peptide remained the same, in accord with its postulated dityrosine linkage. The molecular ion suggests that the Cys-Cys bond is also preserved despite the treatment with tris(2-carboxyethyl)phosphine, possibly because the close proximity enforced by the vicinal dityrosine cross-link increases the resistance of the disulfide bond to reduction.

DISCUSSION
The reaction of LPO with H 2 O 2 initially yields a horseradish peroxidase-like Compound I with a ferryl species and a porphyrin radical cation that decays with time to a poorly defined Compound I that retains the ferryl species but not the porphyrin radical cation (19 -22). The second oxidation equivalent in this latter Compound I species presumably resides on the protein, but no actual evidence for this has been available. The present spin trapping studies provide the first direct evidence that a protein radical is indeed formed in the reaction of LPO with H 2 O 2 (Fig. 1). The relevance of the radical and the proteinbound nature of the spin-adduct are confirmed by the demonstration that (a) its formation requires catalytically active LPO and H 2 O 2 , (b) the EPR signal of the spin adduct is anisotropic and remains associated with the protein upon passage of the protein through a PD-10 gel filtration column, and (c) tryptic digestion of the protein adduct yields a peptide with a sharp isotropic EPR signal.
The demonstration that LPO undergoes a radical dimerization reaction in which a tyrosine residue in each of two LPO molecules couples to give a dityrosine cross-link provides solid independent evidence for a protein radical and identifies one of the residues in which the radical density resides. The nature of the cross-link is definitively established by (a) the fact that the dimeric protein (Fig. 7A) and the cross-linked peptide obtained upon digestion of the protein dimer (Fig. 7B) exhibit the characteristic dityrosine fluorescence and (b) mass spectrometric identification of the cross-linked peptide as Ala 280 -Gly-Phe-Val-Cys-Pro-Thr-Pro-Pro-Tyr 289 -Gln-Ser-Leu-Ala-Arg 294 (Fig.  9), with Tyr-289 as the cross-linked residue. The yield of the cross-linked peptide (40 -50%) shows that the radical density is largely accessible through Tyr-289, although other residues share the unpaired electron density, because the formation of an LPO trimer implicates at least one residue in addition to Tyr-289 as a locus of unpaired electron density. Furthermore, the radical trapped by DBNBS is not Tyr-289, because the LPO dimer reacts just as well with DBNBS as the parent LPO monomer when both proteins are exposed to H 2 O 2 (Fig. 6). In addition, although preincubation of LPO with DBNBS and H 2 O 2 prevents dimer formation, chromatographic removal of the spin trap gives LPO that dimerizes when incubated with H 2 O 2 (not shown). Thus, a minimum of two radical sites are present, one of which is centered on Tyr-289. The protection offered by DBNBS against dimerization is probably due to reducing impurities in the spin trap, which is present in very high concentration.
The finding that the monomeric LPO recovered from incubations of LPO with H 2 O 2 yields no dimeric protein when reexposed to H 2 O 2 indicates either that the native protein exists in two forms, only one of which can undergo dimerization, or that a reaction that modifies the LPO structure competes with the dimerization process. The protein modification, whether preexisting or the result of a peroxide-dependent reaction, either prevents transfer of unpaired electron density to Tyr-289 or suppresses subsequent cross-linking of this residue. The unimpaired catalytic activity of the recovered LPO monomer and the fact that it reacts normally with DBNBS to give a spin-adduct (Fig. 6) clearly show that the recovered protein functions more or less normally and is still able to form protein Coordinates for the MPO x-ray structure (10) and the LPO homology model (37) were obtained from the protein data base and from Springer's server, respectively. radicals. Thus, the modification that prevents LPO dimerization reflects a specific effect at Tyr-289 rather than a general disabling of the protein or its ability to form protein radicals.
The relationship between protein radical formation, spinadduct formation with DBNBS, and protein cross-linking is delineated by studies with LPO substrates and inhibitors. Iodide and thiocyanate undergo two-electron oxidations catalyzed by the Compound I with the ferryl/porphyrin radical species (21,22). Inhibition of DBNBS spin-adduct formation and cross-linking by both of these substrates therefore stems from a competition between their oxidation by the initial Compound I and its conversion to the Compound I form with a protein radical. Ferrocyanide is a one-electron donor and may inhibit LPO dimer formation by reducing both Compound I forms to Compound II, Compound II to the resting ferric state, and the protein radical directly to the diamagnetic state. As already noted, DBNBS appears to inhibit cross-linking due to the presence of reducing impurities in the spin trap rather than by trapping the radical(s) involved in dimerization. Cross-linking and spin-adduct formation can thus be inhibited by consuming the first Compound I species before it decays to the second Compound I species, by reducing Compound I to Compound II, and by directly trapping or quenching the protein radicals that are formed.
EPO and MPO do not detectably dimerize when incubated with H 2 O 2 even though they share a high degree of sequence identity with LPO (Fig. 3B). A model of the LPO structure based on a sequence alignment with MPO, using the MPO crystal structure as a template, provides interesting insights into this difference (Fig. 10). MPO normally exists as a dimer in which Cys-153 links the two monomers through a disulfide bond. Cys-153, as this implies, is located at the dimer interface in MPO. Sequence alignments indicate that Cys-284 in LPO corresponds to Cys-153 in MPO. However, Cys-284 in LPO, although it appears from the model to be surface-exposed, is not involved in the formation of a disulfide bond analogous to that of MPO, because LPO is a monomeric enzyme. The LPO homology model places Tyr-289 close to Cys-284 on the protein surface (Fig. 10). The radical dimerization of Tyr-289 thus occurs at the same protein surface in LPO as the cysteine cross-link in MPO. This requires that two LPO protein molecules interact closely through the same surface as the two subunits in the MPO dimer, although dimer formation involving Cys-284 is blocked in some manner. The residue that corresponds to Tyr-289 in MPO is Ile-324, which readily explains why a dityrosine cross-link is not formed in MPO. H 2 O 2 -dependent cross-linking of MPO, if observed, would require interaction of two dimer molecules through a surface other than that involved in dimer formation. The fact that LPO slowly forms oligomeric species indicates that such interactions are possible but are less favored than interaction of the proteins through the surface used in MPO to form the protein dimer.
Evidence exists that the LPO dimer may be formed in vivo and may be physiologically relevant. Early studies showed that LPO catalyzes the cross-linking of proteins such as collagen and that this cross-linking is inhibited in vitro by iodide and thiocyanate (46). Dimerization of LPO itself was not examined, but later analyses of human salivary peroxidase showed that this peroxidase is present in both monomeric (M r ϳ 80,000) and polymeric (M r ϳ 280,000) states (47,48). As salivary peroxidase is closely related to LPO except in carbohydrate content (49), oligomerization of LPO may also occur physiologically. As we show here, dimerization does not alter the intrinsic catalytic activity of the protein toward small substrates but alters the physical properties of the protein and may alter its ability to interact with other proteins. The present study unambiguously demonstrates that a protein radical is formed, but the possible role of this radical in substrate turnover by LPO remains to be definitively explored.