Horseradish peroxidase mutants that autocatalytically modify their prosthetic heme group: insights into mammalian peroxidase heme-protein covalent bonds.

The mammalian peroxidases, including myeloperoxidase and lactoperoxidase, bind their prosthetic heme covalently through ester bonds to two of the heme methyl groups. These bonds are autocatalytically formed. No other peroxidase is known to form such bonds. To determine whether features other than an appropriately placed carboxylic acid residue are important for covalent heme binding, we have introduced aspartate and/or glutamic acid residues into horseradish peroxidase, a plant enzyme that exhibits essentially no sequence identity with the mammalian peroxidases. Based on superposition of the horseradish peroxidase and myeloperoxidase structures, the mutated residues were Leu(37), Phe(41), Gly(69), and Ser(73). The F41E mutant was isolated with no covalently bound heme, but the heme was completely covalently bound upon incubation with H(2)O(2). As predicted, the modified heme released from the protein was 3-hydroxymethylheme. The S73E mutant did not covalently bind its heme but oxidized it to the 8-hydroxymethyl derivative. The hydroxyl group in this modified heme derived from the medium. The other mutations gave unstable proteins. The rate of compound I formation for the F41E mutant was 100 times faster after covalent bond formation, but the reduction of compound I to compound II was similar with and without the covalent bond. The results clearly establish that an appropriately situated carboxylic acid group is sufficient for covalent heme attachment, strengthen the proposed mechanism, and suggest that covalent heme attachment in the mammalian peroxidases relates to peroxidase biology or stability rather than to intrinsic catalytic properties.

Mammalian peroxidases share a unique feature, the formation of two or three covalent bonds to their prosthetic heme 1 group, not found in the peroxidases of other organisms. The two common bonds in all mammalian peroxidases are ester links between a conserved glutamate and aspartate and the heme 1-methyl and 5-methyl substituents, respectively. In MPO a third bond is forged between a methionine and the ␤ carbon of the 2-vinyl substituent (1). Mutagenesis studies with LPO have demonstrated that at least one of the two ester bonds is required for catalytic activity (2). Interestingly, a single covalent link between the heme and protein is also found in most members of the CYP4A, -4B, and -4F classes of cytochrome P450 enzymes (3)(4)(5)(6)(7). In contrast to LPO, however, mutagenesis studies show that the heme-protein covalent bond is not essential for the catalytic activity of at least some of these P450 enzymes (8,6). A long known example of heme covalent binding is provided by cytochrome c, in which a cysteine residue is covalently bound to each of the two porphyrin vinyl groups. Although many hypotheses have been formulated concerning the functional advantages of the links in cytochrome c, including bending of the heme, increasing the heme affinity (9), and increasing the stability of the methionyl-Fe(II) coordination (10), the question has yet to be definitively answered.
Despite mutagenesis studies of MPO and LPO (for a review, see Ref. 11), both the mechanism by which the heme covalent bonds are formed in the mammalian peroxidases and their functional advantage remain unclear. One reason for the slow progress in this area is the relative complexity and relatively low yield of the expression systems required to obtain recombinant mammalian peroxidases, as well as the sensitivity and intransigence of these peroxidases to mutations. In addition, the expression systems only yield partially or completely processed enzymes. It has therefore not been possible to study the bond-forming reaction commencing with an enzyme from which the bond is largely absent. Nevertheless, it has been unambiguously demonstrated that the ester covalent links are formed via an autocatalytic mechanism (12). It has also been shown in MPO that only a carboxylate can trigger covalent tethering of the heme to the protein (13), and in LPO that the length of the carboxylate side chain is critical for the same bond to be formed (2).
It is not known, however, whether there are requirements other than the presence of an appropriate carboxyl side chain for ester bond formation. Despite their common function and the presence of some structural homologies, mammalian peroxidases share very little sequence identity with peroxidases from other organisms (14,15). Is the uniqueness of covalent heme attachment in the mammalian peroxidases simply a result of the presence of appropriately positioned active site carboxylic side chains, for which there is no counterpart in other peroxidases, or is the participation of other residues required? To address this question, we have introduced a carboxylate into the active site of HRP, one of the best studied peroxidases and one that does not have such a group. This monomeric 44-kDa protein is obtained in good expression yields (16,17), is quite stable, and shares virtually no sequence identity with MPO. Furthermore, its crystal structure has been determined (18), making it possible to rationally select sites for introduction of the carboxylic acid side chain.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were purchased from New England Biolabs. Hemin, H 2 O 2 , ABTS, guaiacol, Pronase, bis-tris propane, trifluoroacetic acid, and tissue culture grade water were purchased from Sigma. All other chemicals, including buffer components and HPLC solvents, were from Fisher Scientific. Native HRP was purchased from Roche. The monohydroxylated hemin derivatives used as HPLC standards were isolated from recombinant HRP (19) and LPO (2) according to the published procedures. H 2 18 O (95% 18 O) was obtained from Cambridge Isotope Laboratories (Andover, MA). High-Five TM cells were grown in supplemented Express Five SFM TM medium, both purchased from Invitrogen. The Sf9 cells (Invitrogen) were grown in Excell 420 TM medium purchased from JRH Biosciences (Lenexa, KS). Penicillin/ streptomycin and Fungizone were purchased from the UCSF Cell Culture Facility. UV-visible absorption measurements were undertaken on a Hewlett Packard 8452A diode array detector. Analytical and preparative HPLC separations were performed on a Hewlett Packard 1090 or 1090-II apparatus equipped with a diode array detector using a Vydac protein-C4 5-m 150 ϫ 46-mm column or a MetaChem Kromasil C4 5-m 100 ϫ 46-mm column, both fitted with the appropriate guard column. LC/MS analyses were carried out either on a Perseptive Biosystems Mariner Biospectrometry work station coupled to an Applied Biosystems HPLC apparatus (solvent delivery system 1408, programmable absorption detector 785A) and connected to a 10-cm Applied Biosystems Aquabore C18 7-m 100 ϫ 1-mm column, or a Waters Micromass ZQ coupled to a Waters Alliance HPLC system (separation module 2695, dual wavelength absorbance detector 2487) connected to an Xterra C18 3.5-m 21 ϫ 50-mm column. Stopped flow kinetic measurements were made on a HighTech TM device, equipped with an SF-61 DX2 double mixing apparatus, a CU-61 central unit, a xenon lamp powered by a PS-678 power supply, a M300 monochromator, and an SSU-60 control unit. The mixing chambers and the syringes were thermostated at 25°C, and the data were analyzed with Kaleidagraph 3.5. Modeling was carried out with Swiss-PDB Viewer 3.7.
Cloning and Expression-The sequence GGA TCC CGG GCC CAC CAC CAC CAC CAC CAT, encoding a BamHI site and Arg-Ala-(His) 6 , was added directly to the 5Ј end of the HRP gene, thus modifying the previously employed construct (17), and the sequence TAA TAA GGA TCC was added at its 3Ј end. The resulting construct was subcloned into the BamHI site of pUC19. 35-mer primers were designed for the mutagenesis with the nucleotide substitutions in the center of the sequences; only the regions with changed codons are shown (actual substitutions underlined, numbering starting at the ATG codon of HRP): L37D, 105 TCG ATC GAC, introducing a silent PvuI site; L37E, 105 TCG ATC GAA, introducing a silent PvuI site; F41E, 120 CAC GAG CAT; G69E, 204 TTC GAG AAC; S73E, 216 AAC GAG GCC. Mutagenesis was undertaken with the QuikChange TM kit (Stratagene, La Jolla, CA). After transformation in XL-1 Blue TM , DNA was isolated from positive clones with the Promega Wizard Plus SV Miniprep TM kit, and the mutated genes isolated from their pUC19 plasmids through a BamHI digestion and purified from agarose gels with the QiaEX II TM kit (Qiagen, Valencia, CA). Ligation was accomplished at 4°C into a pAcGP67B plasmid (Pharmingen, BD Biosciences) preliminarily digested by BamHI, dephosphorylated, and purified with the Promega Wizard DNA clean-up TM kit. The plasmids were transformed into DH5␣ TM , and their sequences were verified. Medium sized cultures allowed for the isolation of larger amounts of endotoxin free plasmids though the use of the Purefection TM purification kit (Promega) following the instructions from the manufacturer. Transfection in Sf9 cells was done with the Baculogold transfection kit (Pharmingen, BD Biosciences). A plaque assay on the transfection supernatants allowed for the isolation of recombinant virus that was subsequently amplified to a titer of ϳ1 ϫ 10 8 plaque-forming units/ml using standard baculovirus protocols (20).
Expression and Purification-High-Five TM cells at a density of 2 ϫ 10 6 cells/ml were supplemented with hemin (final concentration 6 M), 1ϫ penicillin/streptomycin, and 1ϫ Fungizone before being infected at a multiplicity of infection of ϳ5. At 3.5 days after infection, the cultures were harvested and centrifuged at 4°C and 10,000 ϫ g for 1 h. The resulting supernatant was then filtered successively through 1.2-, 0.64-, and 0.44-m regenerated cellulose filters to remove remaining cell debris. The clarified liquid was then concentrated ϳ12-fold by ultrafiltration over a spiral-wound regenerated cellulose membrane (molecular mass cut-off ϭ 10 kDa, Millipore) connected to an Amicon TCF10 peristaltic pump equipped with a 2l RA2000S reservoir. The resulting solution was then made 0.4 M with solid NaCl and 20 mM with KH 2 PO 4 / K 2 HPO 4 at pH ϭ 8, and ultracentrifuged for 30 min at 4°C and 100,000 ϫ g. The supernatant was loaded onto a Ni 2ϩ column (nickelnitrilotriacetic acid-agarose, Qiagen) pre-equilibrated with 20 mM KH 2 PO 4 /K 2 HPO 4 , pH ϭ 8, containing 0.4 M NaCl; the resin was then washed with ϳ10 volumes, then with ϳ5 volumes of the same buffer containing imidazole (20 mM imidazole for the wild type, and 5 mM for the mutants). Elution was achieved by increasing the imidazole concentration to 200 mM for the wild type and 100 mM for the mutants and switching to pH ϭ 6. The colored fractions were pooled and dialyzed against 20 mM KH 2 PO 4 /K 2 HPO 4 , pH ϭ 8, buffer for at least 14 h with at least three bath changes. The dialyzed fractions were then concentrated ϳ12-fold by ultrafiltration over an Amicon YM10 membrane, and passed through a pre-equilibrated QFF-Sepharose (Amersham Biosciences) column. The colored fractions were concentrated again over an Amicon Ultra-4 TM centrifugal filter device (10 kDa) to 1-2 ml.
Spectrophotometric Measurements-ABTS oxidation was measured in 50 mM acetate buffer, pH ϭ 4. 5 HPO 4 , pH ϭ 6.7, was added H 2 O 2 (6 mM) in 2 equivalent portions every 5 min. The solutions were further incubated for another 10 min before being subjected to analysis.
HPLC Analysis-Separations were achieved using water containing 0.1% trifluoroacetic acid (solvent A) and acetonitrile containing 0.085% trifluoroacetic acid (solvent B) at a flow rate of 1 ml/min. For the analysis of undigested samples, the Vydac column was eluted with a 3-min isocratic wash of 33% solvent B, followed by a 0.5%/min linear gradient to 95% of B. For isolation of the 3-monohydroxyheme derivative, the Vydac column was eluted with a 3-min isocratic wash of 15% solvent B, followed by a 0.9%/min linear gradient to 43% of B (see above). For analysis of the 8-monohydroxyheme derivative, the Kromasil column was eluted with a 3-min isocratic wash of 15% solvent B, followed by a 0.9%/min linear gradient to 43% of B.
Characterization of the F41E Mutant Porphyrin-The F41E mutant was treated with 6 ϫ 2 equivalents of H 2 O 2 in 100 mM bis-tris propane buffer, pH ϭ 8.2, over a period of 35 min. Disodium EDTA, pH ϭ 8, was added to a final concentration of 2 mM, and the resulting mixture was incubated for 15 min at room temperature before being transferred to a siliconized glass vessel (2). A freshly made 2.5 mg/ml solution of Pronase was then added, and the digestion was allowed to proceed for 36 h at 37°C. The digest was then injected into an HPLC using the same eluant (water/acetonitrile with trifluoroacetic acid) as above with the following gradient: 0 -3 min at 15% acetonitrile, and then a linear gradient of 1%/min. The porphyrin peaks were repeatedly collected in a siliconized glass tube and concentrated under a stream of N 2 with gentle heating. Mass spectrometric analyses were undertaken on the Mariner instrument with the following settings: positive ion mode, spray potential 2500 V, nozzle potential 125 V and 20 scans/min, inlet flow rate 5 l/min.
Characterization of the S73E Mutant Porphyrin-The S73E mutant was treated with 7 ϫ 2 equivalents of H 2 O 2 as described above. The sample was made up to 0.5% trifluoroacetic acid and 33% acetonitrile before being subjected to HPLC separation. The peak eluting at 6.4 min in each HPLC run was collected in a siliconized glass tube and concentrated under a stream of N 2 with gentle heating. LC/MS analysis was undertaken with a Waters Alliance chromatographic system coupled with a ZQ mass analyzer. The solvents used were water and methanol containing 0.1% formic acid; a linear gradient of 2.5%/min starting at 30% of methanol was started after a 2-min wash. O. An S73E mutant sample was concentrated 10-fold over a Millipore Ultrafree centrifugal filter device (Biomax, 10 kDa) and rediluted 10 times with the phosphate buffer in H 2 18 O. The solution was allowed to stand overnight at 4°C. Three new cycles of concentration/dilution as described above were then undertaken, and the resulting solution was treated with 6 ϫ 2 equivalents of H 2 O 2 as previously described. After reaction, the solution was concentrated again 10-fold and re-diluted 2-fold in water/methanol (9/1). Two volumes of 100 mM disodium EDTA, pH ϭ 8, were then added, and the mixture was allowed to stand for 5 min. An identical volume of 4:6 (v/v) water/methanol was finally added, resulting in the formation of a dense precipitate. The supernatant was harvested after centrifugation, diluted with one volume of 0.5% formic acid in water, and directly analyzed by LC/MS. This procedure is required to allow a thorough denaturation of HRP that the methanol used in the LC/MS does not achieve. The solvents, gradient, and MS settings used were the same as described in the previous paragraph.
Pre-steady State Kinetic Measurements-The experiments were done at 25°C in 50 mM KH 2 PO 4 /K 2 HPO 4 , pH ϭ 6.7. All the H 2 O 2 and guaiacol solutions were made daily and kept on ice until used. The integration time was set at 2 ms in the diode array detector mode. For all measurements, at least five different concentrations were used and each measurement was undertaken at least twice. For the measurement of compound I velocity constants, the apparatus was set in single mixing mode and a fixed concentration of enzyme (ϳ4 M prior mixing) was mixed with varying concentrations of H 2 O 2 . Only initial time points were used for the calculations. For concentrations below a 5-fold excess, the apparent velocity constants were deduced from the slope of the variation of the OD at the compound I/compound II isosbestic point versus time. For concentrations above a 5-fold excess, a single exponential decrease model in the form of a ϩ b ϫ [exp(Ϫk ϫ t)] was used to calculate the apparent velocity constant from the variation of the OD at the compound I/compound II isosbestic point versus time (396 nm for the commercial enzyme, 398 nm for the wild type, 397 nm for both mutants before H 2 O 2 treatment, and 400 nm after treatment). For the measurement of compound II velocity constants, the apparatus was set in double mixing mode; a fixed concentration of enzyme (ϳ8 M prior mixing) was mixed with 1 equivalent of H 2 O 2 . The aging time was set at 300 ms except for the F41E mutant before H 2 O 2 treatment, for which a 5-s aging time was employed. The resulting compound I solution was then mixed with varying concentrations of guaiacol and the variation of the OD at the compound II/resting state isosbestic point (410 nm for the commercial enzyme and the wild type, 412 nm for the S73E mutant, 405 nm for F41E mutant before and 417 nm after ] was used to model the reduction of compound I, and only the first time points were used because the excess of guaiacol was less than 10-fold.

Structure Superposition and Choice of Mutations-MPO
is the reference enzyme for the mammalian peroxidases because its crystal structure has been determined (1,22). No other mammalian peroxidase structure is available. The MPO structure makes it possible to define the precise location and arrangement of each of the three bonds that link the protein to the heme. To determine which residues in HRP might be suitable candidates for the introduction of active site carboxylate side chains, the x-ray structure of HRP (18) was superimposed on that of MPO. The only constraint applied was conservation of the average heme plane. As noted elsewhere for MPO/prostaglandin H synthase and yeast cytochrome c peroxidase/lignin peroxidase (14), the superposition revealed striking similarities in the secondary structures immediately surrounding the iron porphyrin ( Fig. 1, a and b). Helix H2 (residues 84 -97) of MPO and helix B (residues 31-44) of HRP, which bear the corresponding distal histidines, coincided almost perfectly. Likewise, MPO helix H8 (residues 326 -338) superimposed very well on HRP helix F and the residues immediately following it (161-171); both segments bear the proximal histidines. On the same side, the first segment of helix H12 of MPO (417-423) was found in a similar position to helix H of HRP (244 -250), even if the length and orientation of the two helices differed as the distance from the heme increased. It is to be noted that these two segments include Asn 421 of MPO and Asp 247 of HRP, respectively, which confer a partial imidazolate character to their respective proximal histidine. At the approximate position of the MPO 234 -243 segment bearing two of the three heme covalent bonds, we found in HRP a similar poorly defined loop (residues 67-78), a region that is among the most variable in class III peroxidases (23). Finally, and to a lesser degree, two additional helical segments further away from the heme were found in comparable positions: part of helix H6 of MPO (287-300) and helix D of HRP (97-112), and part of helix H17 of MPO (493-504) and helix J of HRP (273-283). The HRP segments identified by this superposition of structures are almost the same as the ones invoked as being the remnants of an ancient gene duplication (24). As previously noted (14), the heme in HRP was ''flipped'' by a 180°rotation about the ␤-meso/␦-meso axis relative to that in MPO. The result of this was an exchange of pyrrole rings A and B of MPO with pyrrole rings D and C of HRP. The proximal His 95 and His 42 of MPO and HRP, respectively, as well as the distal His 336 and His 170 , and Asn 421 and Asp 247 superimpose almost perfectly (Fig. 1c). Surprisingly, Arg 239 of MPO is found at a position roughly equivalent to that of Asn 70 of HRP, a residue that is known to interact directly with His 42 to increase the basicity of the latter. Arg 38 of HRP was found at the same spatial location as Gln 91 of MPO, one of the key residues in the active site hydrogen bond network of MPO (1). The chain of hydrogen bonds leading from the distal histidines is quite different in the two; in MPO, His 95 interacts via a water molecule with Asp 237 and His 250 , whereas in HRP, His 42 forms a hydrogen bond with Asn 70 , which in turn interacts with Glu 64 (25). Finally, the location of the distal calcium binding sites was strikingly similar in the two enzymes (data not shown).
We surveyed the residues in HRP closest to the methyl heme substituents in the superimposed MPO structure, reasoning that a covalent ester link might be formed between any of them if it were correctly located relative to a protein carboxylate group. We limited our search to the distal cavity, as the covalent heme-protein links in the mammalian peroxidases appear to be located on this side of the heme. The closest residue lying under the 1-methyl is Gly 69 , a residue that has been implicated as part of the substrate-binding site (26). When mutation of this residue into a glutamate was simulated, the calculated configurations with the acidic side chain pointing toward the heme gave a best distance of ϳ5 Å between the closest oxygen and the 1-methyl carbon atom (Fig. 2c). Interestingly, the carboxylate group of this simulated G69E mutant was also found within ϳ5 Å of the 8-methyl group. Because a glutamate gave a quite long calculated distance, an aspartate substitution was not modeled. Next, the closest and most ideally placed residue relative to the 3-methyl was Phe 41 . This residue has been shown to impede access of reducing substrates to the compound I ferryl oxygen, thus suppressing the oxygenase activity of HRP (19,27). Computer modeling of a Phe 3 Glu mutation this time gave much shorter distances, with the distance between the ␥-oxygen and 3-carbon being as short as 2.5 Å (Fig. 2b). Taking into account the pyrrole ring ''swapping'' between HRP and MPO mentioned earlier, position 41 turned out to be the strict spatial equivalent of position 94 in MPO. The only reasonable candidate for a bond with the 5-methyl was Leu 37 . This residue has not been assigned any function in the HRP active site, but directed evolution experiments have shown that its mutation into an isoleucine improves enzymatic activity (28). Modeling of a glutamate or an aspartate (an isoleucine steric analog) at this site gave the nearest approach distances of 1.5 and 2.1 Å, respectively (Fig. 2a). Both mutations were therefore selected. Finally, the closest and best oriented residue with respect to the 8-methyl group is Ser 73 . This serine may help to stabilize one of the heme propionates in both soybean seed coat peroxidase and HRP (23) and also interacts with Arg 38 (29). However, an isoleucine is found at this position in cytochrome c peroxidase, ascorbate peroxidase, Arthromyces ramosus peroxidase, and lignin peroxidase (18). Again, because of the pyrrole ring nonequivalence between MPO and HRP, this position corresponds to that of Glu 242 in MPO. Simulating the replacement of Ser 73 by a glutamate gave distances to the methyl group of approximately 5 Å (Fig. 2d). Hence, five mutations were chosen for this study: L37D, L37E, F41E, G69E, and S73E.
DNA Construction and Expression of Proteins-An HRP sequence with a tethered 5Ј His 6 tag coding sequence was introduced into pUC19 for subcloning purposes, and the desired mutations were introduced by site-directed mutagenesis (see ''Experimental Procedures''). The desired constructs were then inserted into a baculovirus transfer vector, which was used to co-infect Sf9 cells to produce recombinant viral particles. After plaque purification, successive virus amplification steps allowed for the production of a viral suspension of ϳ1.10 8 plaqueforming units/ml.
The mutants and the wild-type enzyme were first expressed in High-Five cells at an analytical scale to determine the ki- netics of the expression. In the case of the L37D and G69E mutants, no significant ABTS oxidizing activity could be detected in the cell supernatant even after 5 days following infection. The L37E, F41E, and S73E mutants, however, showed a gradual increase in ABTS activity in the cell supernatant that peaked at day 4 after infection. These screening results showed that the L37D and G69E mutants, in contrast to the other three, probably yielded inactive or unstable protein. Medium scale expressions confirmed that the L37D and G69E constructs gave no isolable protein. The results of medium scale expression of the L37E, F41E, and S73E constructs are presented in Table I. An SDS-PAGE analysis of the proteins is also presented in Fig. 3. The wild-type protein exhibited good expression yields, and the protein had an RZ value and a specific activity for ABTS oxidation slightly higher than those of commercial HRP. In the case of the L37E mutant, the yields were low, the heme content of the protein was only half of that of the wild type, and the position of one of the charge transfer bands could not determined. Furthermore, a pure preparation of the protein could not be obtained (see Fig. 3), and even limited freeze/thaw cycles triggered extensive protein denaturation. Work on this mutant was therefore discontinued. The F41E mutant gave goods expression yields, but the ABTS specific activity was 10 times lower than that of the wild type. The Soret band was also slightly red-shifted, sharper, and shoulderless, and the newly introduced carboxylate had a strong effect on the molar absorption coefficient, as previously reported for other mutations at this position (16, 30,31). Finally, the properties of the S73E mutant were quite similar to those of the wild type, except for the specific activity, which decreased by a factor of 3.
HPLC Analysis-In previous studies this laboratory has shown that HPLC readily distinguishes between covalently and noncovalently bound heme in hemoproteins (2,3,5,8,16). We therefore subjected the wild type, both the F41E and S73E mutants, and commercial HRP to HPLC analysis. The resulting chromatograms (Fig. 4) show that, even though the recombinant enzymes elute somewhat earlier than the commercial enzyme (ϳ20 min as opposed to ϳ21 min, respectively), they appear to be homogeneous and elute as single well defined peaks. In none of the proteins was a significant 400 nm absorbance associated with the protein peaks, clearly establishing the absence of covalently bound heme. The only peak with 400 nm absorbance was at 11.4 min, and this peak was shown by co-elution experiments to be unmodified heme.
Treatment with H 2 O 2 -In the case of recombinant LPO, exposure of the freshly isolated protein to a few equivalents of H 2 O 2 results in covalent attachment of the heme to the protein (2,16). This has led to the proposal that covalent bond formation is the result of an autocatalytic process (11,16). A similar autocatalytic process has been identified in the CYP4 family of cytochrome P450 enzymes that also covalently bind their heme group (5). To test whether the HRP mutants undergo heme-protein cross-linking under oxidative conditions, the enzymes were treated with 6 ϫ 2 equivalents of H 2 O 2 . Analysis of the resulting proteins by SDS-PAGE (Fig. 5) shows that reaction with H 2 O 2 did not alter their electrophoretic mobility and did not cause significant degradation of the protein.
The catalytic properties of the commercial and wild-type enzymes were not significantly modified by H 2 O 2 treatment, but the specific activity of the F41E mutant increased 10-fold, whereas the S73E mutant lost approximately 50% of its activity (Table II). The UV-visible spectra revealed no significant changes in the commercial or wild-type enzymes, but the F41E and S73E mutants lost 40 and 20%, respectively, of their total heme content (data not shown). Furthermore, the Soret band of the F41E mutant shifted from 406 to 409 nm as a result of the exposure to H 2 O 2 , whereas that of the S73E mutant shifted from 404 to 406 nm (data not shown).
Finally, the H 2 O 2 -treated enzymes were analyzed by HPLC to determine whether covalent bond formation had occurred. Control experiments showed that neither the commercial nor wild-type recombinant enzyme underwent any significant cross-linking between the heme and protein (Fig. 6, a and b). For the F41E mutant, the heme absorbance was now found to almost completely co-elute with the protein peak (Fig. 6c), indicating that a covalent bond had been formed between the heme and the protein. In contrast, no significant 400 nm absorption was found to co-elute with the protein peak of the S73E mutant but a new peak absorbing at 400 nm was found at 6.5 min (Fig. 6d, arrow). This peak eluted before free heme and accounted for ϳ10% of the total 400 nm absorption. Interestingly, H 2 O 2 treatment caused a splitting of the protein peak for both the wild type and the F41E mutant. The origin of this phenomenon remains unclear, and we decided to treat both samples as if they were homogeneous. In the case of the wild type, because there was no apparent change in the enzyme properties upon treatment, this was probably justified. In the case of the F41E mutant, it appears from the chromatogram in Fig. 7c that only the second peak is associated with 400 nm absorption. The first peak therefore represents heme-free, inactive enzyme, as virtually no free heme is detected. Allowance  was made for this by calculating the enzyme concentration from the absorption at 400 nm. Analysis of the Porphyrins in the F41E Mutant-To examine the type and number of bonds between the enzyme and the porphyrin, the F41E mutant was subjected to complete proteolysis. Preliminary experiments showed that it was necessary to add EDTA to partially denature the protein to achieve efficient digestion of the HRP mutant by Pronase. The digestion medium was injected into the HPLC (Fig. 7a), and the elution time of the new heme-like peak compared with those of authentic 1-hydroxy, 5-hydroxy, and 8-hydroxymethyl heme (2,19). None of these standards co-eluted with the porphyrin derivative isolated here (data not shown). Nevertheless, LC/MS of the new product (Fig. 7b), gave a mass of 632.2 amu, as expected for a monohydroxylated heme. Its absorption spectrum, HPLC properties, mass spectrum, and nonidentity with the three available hydroxymethylheme standards identifies it as the fourth possible hydroxymethyl heme isomer, i.e. 3-hydroxymethyl heme. Thus, the enzyme and the porphyrin in the F41E mutant are linked via an ester bond to the 3-methyl, as predicted by our original modeling studies.
Analysis of the Porphyrins in the S73E Mutant-Even though a heme-protein covalent bond was not formed with this mutant, a more polar species absorbing strongly at 400 nm was formed in its reaction with H 2 O 2 (Fig. 6d). The compound eluting at 6.7 min in Fig. 6d was isolated from a large incubation of the S73E mutant with H 2 O 2 (Fig. 8a), and, upon coinjection with 1-hydroxy-, 3-hydroxy-, and 5-hydroxyheme was shown to co-elute with the 8-hydroxy derivative (data not shown). LC/MS gave a mass of 632.1 amu for the new product (Fig. 8b), consistent with its identification as 8-hydroxymethylheme. Interestingly, although the predicted formation of a covalent bond to the 8-methyl group was not observed, the carboxylate in the mutant nevertheless triggered a partial hydroxylation of that position.
Formation of the 8-hydroxymethylheme derivative but not of a heme-protein covalent bond could arise from competition of a water molecule with the protein carboxylate for the postulated 8-methylene carbocation intermediate in the bond-forming reaction. To test this hypothesis, the experiment was repeated with an enzyme that had been extensively pre-equilibrated in buffer made with 18 O-labeled water. The 8-hydroxymethylheme peak at 6.7 min was obtained as usual on treatment of the enzyme in the same 18 O-labeled buffer with H 2 O 2 . The compound in this instance gave a mass of 634 amu, establishing that the new oxygen atom derived from the bulk medium (Fig. 9a). As an internal control, the unmodified heme present in the same sample was analyzed by ESI mass spectrometry and was found to give the expected mass of 616 amu, confirming that the increased mass of the new product was not simply because of exchange of the carboxylate groups of the heme with the medium (Fig. 9b).
Influence of the Number of H 2 O 2 Equivalents-It has been shown in recombinant LPO that the extent of bond formation is directly related to the amount of H 2 O 2 that is added (2). To explore the relationship of heme modification to H 2 O 2 equivalents in the HRP mutants, the area of the 400 nm peak coeluting with the protein was measured by HPLC for the F41E  mutant, and the area of the 8-hydroxymethylheme peak at 6.7 min for the S73E mutant. As shown in Fig. 10, covalent binding of the heme to the protein can be brought nearly to completion for the F41E mutant, but the proportion of hydroxylated heme obtained with the S73E mutant does not rise above 9 -10%. The majority of the heme in this latter mutant remains unmodified. As indicated by the original spectroscopic studies (Table II), there is some net heme loss with both mutants and this loss increases as the H 2 O 2 concentration increases. We attribute this phenomenon to bleaching, i.e. destruction, of the heme because of increased sensitivity of the mutant proteins to H 2 O 2 in the absence of added reductant.
Influence on the Elementary Rate Constants-The heme moiety of peroxidases oscillates between three different oxidation states as it traverses the catalytic sequence. The order of the rates of formation of the different states in HRP is usually resting state to compound I (k 1 ) Ͼ compound I to compound II (k 2 ) Ͼ compound II to resting state (k 3 ) (15). Even though compound I and compound II of HRP have similar redox potentials, compound II is reduced more slowly than compound I by single electron donors, presumably as a result of the longer distance the electron must travel to quench the hypervalent iron of compound II rather than the porphyrin radical cation of compound I. We have therefore examined the influence of the mutations on the individual rate constants before and after H 2 O 2 treatment (Table III). The values obtained for k 1 before and after H 2 O 2 treatment for both the wild-type and native protein are in good agreement with literature values (30). Introduction of a glutamate in the F41E mutant decreases k 1 by 2 orders of magnitude before H 2 O 2 treatment, but after exposure to 6 ϫ 2 equivalents of H 2 O 2 the k 1 value returns within the same range as that of the wild-type enzyme. The value of k 1 for the S73E mutant was decreased by a factor of 3 relative to the wild-type enzyme and was not significantly altered by H 2 O 2 treatment.
The k 2 values for guaiacol oxidation are within the same order of magnitude as the published value for the wild-type enzyme (25,32), although the value for the commercial enzyme measured here was 10 -30% lower than that in the literature. The values recorded for the wild-type enzyme before and after treatment are consistent with the ones obtained for the commercial enzyme, and show no significant influence of the treatment with peroxide. In contrast, the F41E mutant gave significantly higher values before (50% higher than the wild type) and after (160% higher) treatment with peroxide. Unexpectedly, in the case of the F41E mutant after treatment with H 2 O 2 , no red shift was observed upon mixing compound I with varying concentrations of guaiacol. Instead, we simply observed a gradual increase of the absorbance of the peak, giving the appearance of direct reconversion to the resting state. The use of a slower reducing substrate, p-aminobenzoic acid, did not alter this behavior. We therefore undertook the calculations assuming that the compound II of this enzyme has a spectrum indistinguishable from that of the resting state under the conditions employed. The S73E mutant for its part gave values approximately 3 times lower than the wild type, and this was not significantly changed by the oxidative treatment. The instability of compound II precludes the measurements of k 3 by stopped flow kinetics under the conditions used, so we tried to obtain k 3 values through steady state kinetic measurements as described elsewhere (25,32,33). Unfortunately, in our hands the model did not permit the calculation of reliable values by this method. Further work will therefore be necessary to obtain these values by the pH jump technique (34,35). DISCUSSION Superposition of the structure of MPO with that of HRP revealed significant similarities in their three-dimensional structures in the immediate vicinity of the heme despite the fact that they belong to different superfamilies and share little sequence identity. Five helices or helical segments and one loop were found at spatially equivalent positions. More interestingly, both catalytic histidines and the proximal carboxyl side chain responsible for increasing the proximal histidine basicity were found at almost identical spatial positions. The MPO catalytic arginine was located at a position where HRP has a key auxiliary residue hydrogen-bonded to the distal histidine, and the HRP arginine at a site that corresponds to a key As far as the heme is concerned, the general plane of the tetrapyrrolic assembly is conserved, but not the positions of the heme substituents. This conservation of active site structure in MPO and HRP validates our use of superposition in identifying potential HRP residues to mutate in efforts to introduce a heme-protein covalent bond.
Four positions were thus identified, which, if mutated to a carboxyl-bearing amino acid, might result in covalent bonding of the heme to the protein: Leu 37 , Phe 41 , Gly 69 , and Ser 73 . From these, only Phe 41 and Ser 73 yielded mutants that could be isolated and studied, as the mutations at the other two sites did not yield detectable proteins. The failed mutants show, as might be expected, that the introduction of a negatively charged carboxylic acid residue into the active site can cause serious structural perturbations. The introduction of a carboxylate into the active site of HRP has been achieved earlier for other purposes, notably the substitution of His 42 by a glutamate in efforts to prepare a chloroperoxidase mimic (29,32,36) or to disrupt the active site hydrogen-bonding network associated with a bound benzhydroxamic acid molecule (37). Asn 70 , which interacts with His 42 , has also been successfully replaced by a glutamate (17,38,39). The H42E mutation decreased the catalytic activity 300-fold (32) and the N70D mutation 2-6-fold (17). Finally, a glutamate has been introduced at position 143, in the postulated aromatic reductant binding site, leading to an enzyme with reduced activity (40).
Mutagenesis of Ser 73 has not been reported, but Phe 41 has been replaced by a variety of residues: glycine (41), alanine (16, 31,41,42), valine (30,31,41), leucine (41,43), threonine (43), tyrosine (41), histidine (40,42), and tryptophan (31,41). Indeed, Phe 41 has emerged as a key residue in HRP. In all the sequence-related plant peroxidases (24), an aromatic side chain is found at the corresponding position: a tryptophan in cytochrome c peroxidase and ascorbate peroxidase and a phenylalanine in the others. Phe 41 restricts access of reducing substrates to the oxoferryl moiety and thus limits the peroxygenase activity of the enzyme (19,27). We examined the ability of our mutants to epoxidize styrene as a test for peroxygenase activity (43), but in none of the variants, before of after H 2 O 2 treatment, could we detect any styrene oxide formation (data not shown). Replacement of Phe 41 by a glutamate therefore does not increase access to the ferryl oxygen. This aromatic position governs the resting state spin state in both cytochrome c peroxidase and A. ramosus peroxidase and the stability of compound I in cytochrome c peroxidase. A red shift of the HRP Soret band and an increase in its molar absorption coefficient have been observed when Phe 41 is mutated to a Val or Ala (30,31,37). This shift has been attributed to an increase in 6-coordinate high spin character at the expense of the 5-coordinate high spin character of the wild-type enzyme. Our observations for the F41E mutant are consistent with the literature; introducing a glutamate at position 41 sharpens the Soret band, shifts it by 3 nm, and increases its molar absorption coefficient. Introducing a glutamate at position 73, further away from both the heme and the iron, only causes minor changes in the UV-visible spectrum.
Not surprisingly, the mutations alter the catalytic properties of the enzyme. Under steady state conditions, the mutant enzymes as isolated from the insect cell cultures exhibited a moderately reduced ABTS oxidizing activity (5 and 2 times lower for F41E and S73E, respectively), but the influence on the guaiacol oxidation activity was stronger (Table II, before H 2 O 2 ). This difference in sensitivity between ABTS and guaiacol has been attributed to the existence of additional limiting association/dissociation steps in the case of ABTS (30). However, we found a much stronger decrease of the guaiacol oxidation rates, by a factor of 100 rather than 8, which was paralleled by a similar reduction of the rate of compound I formation (see below). In contrast, the mutation at position 73 only modestly reduced the specific activity of the enzyme despite the fact that Ser 73 has been postulated to interact with Arg 38 and one of the heme propionates (29). Analysis of the individual catalytic steps is informative. In the case of the F41E mutant, a decrease by a factor of 100 of the second order rate constant for the formation of compound I is observed, whereas in the case of the S73E mutant the decrease is only 2-3-fold. After preincubation with H 2 O 2 , the velocity constant for oxidation of F41E to compound I increases greatly, reaching a value in the same range as that for the wild type. In contrast, the value for the S73E mutant is barely changed. The rates for the single electron reduction of compound I by guaiacol are less affected by the mutations, because the measured variations among all the recombinant enzymes do not exceed a factor of 3. These results tend to show that the major change brought about by the substitutions is a decrease of the ease of reaction with H 2 O 2 , whereas the subsequent reduction of the oxidized states is relatively unaffected. In the case of the F41E mutant, one could envision either an ionic or steric interaction between the distal histidine at position 42 and the glutamate at position 41. In the case of the H 2 O 2 pretreated F41E mutant, we were unable to detect the characteristic red shift upon conversion of compound I to compound II. The two obvious possibilities are that either no compound II forms, or it forms but has spectral properties that cannot be differentiated from those of the resting state. The first possibility seems unlikely, however, as it implies simultaneous reduction of the two oxidizing equivalents of compound I. These reducing equivalents would have to come from two different sources, either two substrate molecules or one substrate molecule and the protein itself. Because in the absence of added reductant, compound I of the H 2 O 2 -treated F41E mutant is stable (data not shown), the second hypothesis requires the improbable assumption that reaction with guaiacol (or p-aminobenzoic acid) triggers simultaneous oxidation of the protein. We have therefore proceeded on the assumption that compound II of the H 2 O 2 -treated F41E mutant has spectral characteristics indistinguishable from those of the resting state. In the case of the S73E mutant, the glutamate side chain is further away from the iron center and its perturbations of the catalytic cycle are understandably less pronounced.
The most important result of this work, however, is the finding that both mutants differ from the wild-type and native enzymes in their reactions with H 2 O 2 . In the absence of any other factor, the F41E mutant underwent nearly quantitative covalent linking of its heme 3-methyl group to the protein, presumably via Glu 41 , and the S73E mutant partially hydroxylated the 8-methyl substituent of its prosthetic heme group. Thus, as in the mammalian peroxidases, heme covalent binding occurs via an autocatalytic mechanism (2,11,12). Both of these modifications most likely arise via a common mechanism (12). After formation of compound I, a carbocation is formed either by transient oxidation of the carboxylate and subsequent hydrogen abstraction from the methyl, or via direct deprotonation of the methyl acidified by the porphyrin radical cation (11). The resulting cation is then quenched by a nearby nucleophile, either the carboxylate anion or a nearby molecule of water (Fig.  11). In these mechanisms, it is the very presence of the carboxylate that allows the transfer of oxidizing equivalents from the iron and the -system to the methyl substituent. In the F41E mutant, for which modeling predicted a distance of ϳ2.5 Å between the glutamate oxygen and the 3-methyl carbon, quenching gives the ester. In the S73E mutant, where the distance to the 8-methyl carbon was predicted to be ϳ5 Å, the carboxylate is unable to compete with an active site water molecule in trapping of the cation, so that hydroxylation of the heme rather than covalent binding to the protein is observed.
In accord with this postulated mechanism for the formation of 8-hydroxymethylheme, the hydroxyl oxygen is shown by incubation in 18 O-labeled water to derive from the solvent. The same type of competition between a carboxylate and a water molecule was postulated in the cytochrome P450 system (3,5) and was later confirmed for CYP4B1 (6). It is to be noted that the heterogeneity of heme species in the S73E mutant after H 2 O 2 treatment does not appear to affect its catalytic properties. This is consistent with the previous finding that HRP with an 8-hydroxymethylheme group has the same catalytic properties of the enzyme with a conventional heme group (44). Fi- FIG. 11. Radical mechanism, starting with the compound I obtained by reaction with H 2 O 2 , proposed to explain autocatalytic covalent binding of the glutamic acid to the 3-methyl group in the F41E HRP mutant. The structure of the 8-hydroxyheme formed with the S73E mutant because of effective competition of water for trapping of the carbocation analogous to that shown for the F41E mutant is shown on the bottom far right. nally, we undertook the same type of H 2 O 2 treatment in the presence of guaiacol, reasoning that if an oxidized form of the enzyme is indeed needed for the iron porphyrin to be modified, then a competition between porphyrin and guaiacol oxidation should take place. Indeed, we found that 20 equivalents of guaiacol were enough to completely inhibit covalent linking of the heme to the protein during the H 2 O 2 pretreatment (data not shown). This finding emphasizes the close similarities between the bond-forming process (porphyrin oxidation) and the normal catalytic cycle (exogenous substrate oxidation).
Perhaps the most important conclusion of this study is that all that is needed for the formation of a protein-heme covalent link is the proper juxtaposition of a carboxylate side chain and a heme methyl group within the hemoprotein active site. Previous data suggested this conclusion, but in those instances the carboxyl residue was introduced into an active site that was very closely related in sequence to that of the reference protein, for example, in the CYP4F5 G330E mutant (5). In the case of MPO and HRP, there is no sequence identity in the active site region even though both proteins employ similar iron ligands and catalytic residues to activate hydrogen peroxide. Formation of the covalent ester link thus reflects the simple chemistry of the active site and does not require the committed participation of additional protein residues. Interestingly, the only bond detected in our experiments was that between the 3-methyl and Glu 41 , even though, for example, the 4-vinyl is also close to the carboxylic acid side chain of Glu 41 . The process is thus chemoselective, even though the intrinsic chemistry involved is quite powerful. In sum, perhaps surprisingly, a properly mutated HRP is an excellent model for study of the ester bond formation in mammalian peroxidases because (i) the reaction goes to completion, (ii) the study can be undertaken on an enzyme initially totally devoid of such a bond, (iii) it is autocatalytic, (iv) it is highly chemoselective, and (v) HRP is more stable and easier to express than mammalian peroxidases. Finally, it should be noted that when a tryptophan is introduced in position 41, the resulting enzyme has very low activity. However, a 5-10-min incubation with H 2 O 2 activates the enzyme (31). The authors did not attempt to characterize the chemical modifications responsible for this activation process, but their observation is consistent with our results showing that oxidation of an oxidizable side chain can occur at position 41.
The implications of covalent bond formation on the properties of HRP remain to be clarified. The UV-visible spectrum, the ABTS and guaiacol oxidation efficiencies, and the second order rate constants for the elemental reactions of the catalytic cycle resemble those of wild-type HRP. No clear advantage or disadvantage is conveyed by the presence of the new ester bond. This suggests that the reasons for the ester links to the heme in the mammalian peroxidases are likely to be related to stability, specificity, or peroxidase biology rather than to intrinsic catalytic properties.