The Hydrogen Peroxide Reactivity of Peptidylglycine Monooxygenase Supports a Cu(II)-Superoxo Catalytic Intermediate*

We have investigated the reaction of peptidylglycine monooxygenase with hydrogen peroxide to determine whether Cu(II)-peroxo is a likely intermediate. When the oxidized enzyme was reacted with the dansyl-YVG substrate and H2O2, the α-hydroxyglycine product was formed. The reaction was catalytic and did not require the presence of additional reductant. When 18O-labeled H2O2 was reacted with peptidylglycine monooxygenase and substrate anaerobically, oxygen in the product was labeled with 18O and must therefore be derived from H2O2. However, when the reaction was carried out with H 162O2 in the presence of 18O2, 60% of the product contained the 18O label. Therefore, the reaction must proceed via an intermediate that can react directly with dioxygen and thus scramble the label. Under strictly anaerobic conditions (in the presence of glucose and glucose oxidase, where no oxygen was released into the medium from nonenzymatic peroxide decomposition), product formation and peroxide consumption were tightly coupled, and the rate of product formation was identical to that measured under aerobic conditions. Peroxide reactivity was eliminated by a mutation at the CuH center, which should not be involved in the peroxide shunt. Our data lend support to recent proposals that Cu(II)-superoxide is the active species.

Peptidylglycine monooxygenase (PHM) 2 catalyzes the hydroxylation of peptidylglycine substrates at the C␣ position, the first step in the amidation of peptides by the bifunctional enzyme peptidylglycine ␣-amidating monooxygenase (1, 2). The enzyme requires two coppers for activity (3) and undergoes redox cycling during catalysis via the intermediacy of both dicopper(II) and dicopper(I) forms (4). Structural (5)(6)(7), spectroscopic (8 -12), and theoretical (13,14) studies have provided a detailed description of the ligand environment of the copper centers, which are bound in separate domains, about 11 Å apart (Fig. 1). One copper (Cu H , also termed Cu A ) is bound to three histidines (His 107 , His 108 , and His 172 ) in domain 1. The other copper (Cu M , also termed Cu B ) is bound to two histidines (His 242 and His 244 ) and a methionine (Met 314 ) in domain 2. X-ray absorption spectroscopy studies have iden-tified large changes in coordination between Cu(II) and Cu(I) states (11). In the oxidized enzyme, Cu H is further ligated by at least one solvent molecule, whereas Cu M coordinates two histidines and two water molecules in the equatorial plane with the methionine in an axial position undetectable by EXAFS (11,12). Reduction causes the water ligands to dissociate and the methionine to move close (2.25 Å) to the Cu M center (9,11). In contrast, crystallographic studies have failed to detect large changes in metal coordination during redox (6).
The detailed mechanism of substrate hydroxylation has been the subject of much debate. It is generally accepted that the enzyme cycles through a reductive phase in which the two copper centers are reduced to Cu(I) and an oxidative phase in which O 2 is activated by binding at one of the copper centers and subsequently hydroxylates the substrate. How this chemistry occurs is still unclear. Crystal structures of substrate-bound forms of PHM have located a di-iodo-YG substrate bound in the vicinity of the Cu M center (2,5,6) (Fig. 1A), whereas a precatalytic complex of PHM with the slow substrate tyrosyl-D-threonine and dioxygen shows O 2 bound at Cu M but rotated away from the Cu-C␣(substrate) vector (7). These structures strongly support the premise that oxygen activation occurs at the Cu M center but provide no information on the chemical identity of the reactive species. If the reactive oxygen species is a Cu M -peroxo or hydroperoxo complex, an electron must be transferred from Cu H to Cu M to complete the 2-electron reduction of O 2 to peroxide, but this itself presents a mechanistic challenge, since the two coppers are separated across an 11-Å solvent-filled cavity, and the shortest through-bond pathway is Ͼ80 Å. To overcome this problem, Prigge et al. (6) identified a potential electron transfer (ET) pathway involving the Cu H ligand His 108 , Gln 170 , a hydrogen-bonded water molecule, and the peptide substrate, which reduced the ET pathway to ϳ20 Å. Invoking a different strategy, Jaron and Blackburn (10) suggested that O 2 might react initially at Cu H and that the superoxide so formed could channel to Cu M providing a carrier for the electron and possibly a proton.
Neither of these mechanisms is consistent with all of the available data. Glutamine 170, a critical residue in the substrate-mediated ET pathway, can be mutated to alanine with no loss of catalytic activity (15), whereas in the related enzyme dopamine ␤-monooxygenase, oxygen reduction and substrate hydroxylation remain tightly coupled even in the case of extremely slow substrates, apparently ruling out superoxide channeling, where some leakage of superoxide into bulk solution would be expected (16). These results led Klinman and co-workers (16 -18) to argue against peroxide as a viable intermediate in both PHM and dopamine ␤-monooxygenase and to propose that the reactive oxygen species is a Cu(II)-superoxo species, which abstracts a hydrogen atom from substrate prior to the electron transfer step.
If a Cu M (II)-peroxo or hydroperoxo species is an intermediate, then it should be possible to generate product by reacting the oxidized enzyme with hydrogen peroxide and substrate as depicted schematically in the oxidative phase of the mechanism shown in Fig. 1B. This "peroxide shunt" has been shown to occur in other oxygenase systems such as cytochrome P450 (19,20), methane monooxygenase (21), and naphthalene 1,2-dioxygenase (22). Peroxide shunt reactions have the characteristic that when labeled peroxide is used as the source of the hydroxylating oxygen atom, the label is quantitatively transferred to the products (22). In this paper, we have investigated the reactivity of hydrogen peroxide with the oxidized form of the PHM catalytic core (residues 42-356, PHMcc) and find that hydroxylated product is indeed formed in the reaction. However, when 18 O-labeled peroxide was used as the source of oxygen, we observed scrambling of the label with atmospheric O 2 . We also observed that peroxide reactivity is eliminated by a mutation at the Cu H center, which should not be involved in the peroxide shunt chemistry. The results have led to the conclusion that Cu(II)peroxo or -hydroperoxo species are probably not involved in the reaction pathway, and an alternative mechanism involving Cu(II)-superoxo species appears more likely.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Unless otherwise stated, chemicals and buffers were obtained from Sigma or Fisher and were used without further purification. 18 O-Labeled hydrogen peroxide (2% (v/v) solution) and oxygen gas were obtained from Icon Isotopes, New Jersey, at 90 and 99 atom %, respectively.
Expression and Purification of PHMcc and Its H242A and H172A Mutants-PHMcc and the H242A and H172A mutants were expressed and purified as described previously (10,11,23) using a Cellmax 100 (Spectrum Laboratories) hollow fiber bioreactor with a 1.1-m 2 cellulose acetate cartridge. In some experiments, wild type PHMcc was expressed using a Biovest Minimax automated cell culture system (Biovest International). Typically, 5 days of bioreactor harvest were pooled. Ammonium sulfate was added to 50% saturation, and the solution was stirred for 1 h. The resulting precipitate was centrifuged at 12,000 ϫ g and redissolved, with gentle shaking, in 10 ml of 50 mM sodium phosphate buffer (pH 7.5), containing 0.001% Triton X-100. The sample was then centrifuged to remove insoluble material, filtered through a 0.45-m sterile filter, and applied to a 26/60 Hiload Superdex 75 prep grade filtration column (Amersham Biosciences) at a flow rate of 2.5 ml/min. The enzyme began elution at 0.6 column volumes and continued to elute over 45 ml. At this stage of purification, SDS-PAGE revealed a purity level of 90 -95%. The enzyme was further purified by anion exchange chromatography on a Biocad 700E perfusion chromatography system (Applied Biosystems), using a 10 ϫ 100-ml Peek column packed with Poros 20-m HQ anion exchange resin. Partially purified PHM from size exclusion chromatography was loaded in 20 mM Tris acetate buffer, pH 8.2, and then washed with 2 column volumes of the buffer. The column was eluted by a 0 -300 mM NaCl gradient in loading buffer, over 10 column volumes. Purified PHM eluted close to 100 mM NaCl. Yields of pure PHMcc for 5 days of harvest ranged from 30 to 50 mg.
Copper Reconstitution-The purified protein in 20 mM sodium phosphate, pH 8.0, was placed in a 50-ml conical centrifuge tube to which 100 mM copper sulfate was added at 1 l/min on ice with gentle stirring until the molar ratio of copper added per protein was 2.5:1. The protein was then concentrated in an Amicon ultrafiltration device (10,000 Da cut-off) from 30 to 1 ml. This was followed by three wash sequences to remove excess and/or adventitiously bound copper. During each wash, 10 ml of 20 mM sodium phosphate buffer, pH 8, containing 25 M Cu(II) (as Cu(SO 4 )) was added to the ultrafiltration device, and the solution was concentrated to 1 ml. A final concentration step adjusted the pure PHMcc to 1 mM. The protein was then flash frozen in cryotubes and stored in aliquots at Ϫ80°C. The final Cu/protein ratio was in the region of 2.0 -2.2:1.
Copper and Protein Concentration-Protein concentration was determined using the A 280 and an extinction coefficient (1 mg/ml) of 0.98 as previously described (10). A 280 measurements were recorded on a Shimadzu UV-265 spectrophotometer at ambient temperature. Copper concentrations were determined using a PerkinElmer Optima 2000 DV inductively coupled plasma optical emission spectrometer.
HPLC Separation and Detection of Product and Substrate-Reversephase HPLC was performed with a Varian Pro Star solvent delivery module equipped with a Varian Pro Star model 410 autosampler (250-l syringe, 100-l sample loop), on a 250 ϫ 4.6-mm Varian Microsorb-MV 100-5 C18 column. Substrate (dansyl-YVG; American Peptide Co.) and product (produced by the PHM-catalyzed reaction) were monitored by their dansyl fluorescence using a Waters 474 scanning fluorescence detector ( Ex ϭ 365 nm, Em ϭ 558 nm). Solvent A was 0.1% trifluoroacetic acid in water, and solvent B was 0.1% trifluoroacetic acid in acetonitrile). Product was separated from substrate via isocratic loading and elution at 25% B in A.
Steady State Kinetic Measurements-The kinetics of the peroxide reaction were determined by following the rate of substrate consumption (or product formation) as a function of time. The reaction was performed in a water-jacketed glass reaction vessel, with stirring, in 100 mM MES buffer, pH 5.5, at 25°C. All reagents except for hydrogen peroxide were added to the following final concentrations: dansyl-YVG (50 -400 M), Cu 2ϩ as copper sulfate (5 M), and PHM (2.5-5 M). After the reagents were allowed to incubate for 2 min, the reaction was initiated by adding H 2 O 2 from a 15 mM stock, to a final concentration of 0.5-4.0 mM. In a typical experiment using 1 mM H 2 O 2 , 330-l aliquots were removed every 30 s, transferred to a 1.5-ml microcentrifuge tube containing 10 l of 10% trifluoroacetic acid, and vortexed for 10 s. Substrate and product were separated by HPLC, and their concentrations were determined using a standard curve built from chromatograms of 10 -200 M samples of dansyl-YVG run under identical conditions. Kinetic constants were extracted from the raw data by fitting to the Michaelis-Menten equation using nonlinear regression in Sigmaplot 8.0.
Measurement of the Dansyl-YVG Dissociation Constant K D -For the oxidized enzyme, an Amicon (5-ml) ultrafiltration device was first preincubated overnight with 300 M Cu(II)-loaded PHM (2.0 copper/protein) in 100 mM MES, pH 5.5, containing 1 mM dansyl-YVG. The protein solution was then washed repeatedly with buffer until the substrate concentration in the filtrate had fallen to a low level as determined by measurement of dansyl fluorescence of the filtrate. This conditioning procedure ensured that all irreversible substrate and/or protein binding sites on the membrane were occupied. Next, dansyl-YVG was titrated into the PHM solution, and aliquots of the filtrate were extracted for analyses as follows. A septum was placed over the ultrafiltration cell, and the cell was pressurized by injecting 0.6 ml of air over the solution. After a small amount of filtrate had been collected, it was returned to the concentrator, and the process was repeated three times. On the fourth pressurization, 10 l of filtrate was saved for analysis of the concentration of "free" substrate by fluorimetry ( Ex ϭ 365 nm; Em ϭ 558 nm). The remaining filtrate was returned to the concentrator along with the next titration aliquot (30 l) of 10 mM substrate for a net volume gain of 20 l. This procedure was repeated until a total substrate concentration of 1 mM was reached.
For the reduced protein, the following modifications to the procedure were used. All solutions were first purged with argon and placed in an anaerobic chamber. Cu(II)-loaded PHM (2.0 copper/protein) in 100 mM MES, pH 5.5, was reduced with a 5-fold excess of buffered ascorbate and then titrated with dansyl-YVG in a conditioned ultrafiltration cell in the anaerobic chamber using an identical protocol.
The data were analyzed by constructing plots of fractional binding, versus free substrate S F (where the subscripts T and F represent total and free concentrations, respectively). These data were fit by nonlinear regression (Sigmaplot 8.0) to the equation, where K D is the dissociation constant for binding of dansyl-YVG to PHM, and n is the number of binding sites. Values of K D and n were refined in the fits. 18 O Incorporation Experiments- 18 O-Labeling reactions were carried out similarly to those for kinetic analysis, except that all reactions were carried out in an anaerobic chamber. 18  All reactions were quenched with trifluoroacetic acid (10% in water) and purified on HPLC prior to mass spectrometry measurements.
Mass Spectrometry-Samples for mass spectrometry were purified on HPLC and then diluted 250 times for infusion with 50:50 acetonitrile/water in 0.1% formic acid. Samples were directly infused into the electrospray ionization source of a LCQ Deca XP Plus (Thermo, San Jose, CA) ion trap mass spectrometer at 3 l/min. Typically, 100 profile scans were acquired (200-ms maximum injection time and three microscans) over a range of m/z 500 -1500. The standard isotope distribution for substrate and product was calculated using MS-Isotope in Protein Prospector 4.0.5 (by P. Baker and K. Clausner; available on the World Wide Web at prospector.ucsf.edu/ucsfhtml4.0/msiso.htm).
Measurement of Peroxide Concentration-Hydrogen peroxide concentrations were determined using a BIOXYTECH H 2 O 2 -560 quantitative peroxide formulation kit (OXIS International Inc.). 25 l of quenched reaction mixture was diluted to 2 ml and mixed thoroughly. 100 l of this dilution was then added to 1 ml of working reagent. The A 560 was recorded for the samples and a series of standards, after incubating for 1 h at ambient temperature.
Peroxide Degradation Monitored by Oxygen Production-Oxygen production was monitored using a Rank Brothers oxygen electrode at 25°C. 100 mM MES (pH 5.5) was added to a stirred cell until a stable base line was achieved. The stirred cell was capped, and H 2 O 2 was then added with a Hamilton syringe, through a small opening in the cap, to a final concentration in the cell of 1 mM. Reagents from the standard reaction including Cu 2ϩ , PHM, and substrate, were added consecutively, by Hamilton syringe, with oxygen levels monitored after each addition.
Reaction Using Glucose and Glucose Oxidase (GO) to Generate H 2 O 2 -Experiments in which a glucose/GO system was used to generate H 2 O 2 were performed in a Rank Brothers oxygen electrode at 25°C. 100 mM MES buffer (pH 5.5) containing 50 mM D-glucose was added to the cell and allowed to equilibrate, and the response was adjusted to 21%. The base line was allowed to stabilize, the original buffer was removed, and fresh buffer at 100 or 21% oxygen saturation was added. The stirred cell was capped, and glucose oxidase was immediately added with a Hamilton syringe to a concentration of 43 g/ml. Once all the oxygen was converted to hydrogen peroxide, substrate and PHM were added with a Hamilton syringe to final concentrations of 200 and 5 M, respectively. Each reaction was performed in a total volume of 2 ml. The reaction was quenched at the desired time, by the addition of 70 L 10% trifluoroacetic acid. The quenched reaction was immediately analyzed for H 2 O 2 , product, and substrate concentrations.
Stoichiometry of Peroxide Consumption to Product Formation Using the Glucose/GO Reaction to Generate Hydrogen Peroxide-These experiments were performed in the same manner as the glucose/GO reactions above, with the exception that substrate was added to 400 M, and only 21% oxygen saturation was used. The initial H 2 O 2 concentration (formed by conversion of the dissolved oxygen to peroxide by glucose/GO) was assayed prior to the addition of substrate and PHM and again after the reaction was quenched with trifluoroacetic acid.
EPR Spectroscopic Quantitation of the Reduction of the Copper Centers in PHMcc by Hydrogen Peroxide-EPR spectra were obtained from PHM samples with [Cu(II)] ϭ 250 M, on a Bruker E500-X-band EPR spectrometer equipped with a SuperX microwave bridge, a superhigh Q cavity, and a nitrogen flow cryostat (Helitran; Advance Research Systems). The following experimental conditions were used: temperature, FIGURE 1. A, active site structure of the oxidized form of PHMcc with the substrate diiodo-YG bound by its C-terminal carboxylate to the guanidinium group of Arg 240 and the phenolate OH of Tyr 318 . The Cu H center (also termed Cu A ) is coordinated by His 107 , His 108 , and His 172 , whereas the Cu M center (also termed Cu B ) is coordinated by His 242 , His 244 , and Met 314 . The coordinates were taken from Protein Data Bank file 3PHM. B, schematic general mechanism for the monooxygenase activity of PHM. The reaction can be divided into a reductive phase in which the dicopper(II) form is reduced to the dicopper(I) form by a reductant (ascorbate) and an oxidative phase in which the dicopper(I) form reacts with O 2 to generate a reduced oxygen intermediate. In the figure, the oxidative phase is depicted as involving a dicopper(II)-peroxo species that corresponds to a "peroxide shunt" reaction in which product can be formed from oxidized enzyme, substrate, and hydrogen peroxide, without the need for reducing equivalents. 80 K; microwave frequency, 9.4 GHz; microwave power, 20 milliwatts; modulation frequency, 100 kHz; modulation amplitude, 10 G. EPR signals were quantified by double integration under nonsaturating conditions and by comparison with 100, 200, and 300 M Cu(II) (EDTA) standards. Titration with hydrogen peroxide was performed by adding 10-l aliquots of a 30% hydrogen peroxide stock solution to a 200-l initial sample volume.

RESULTS
The reaction cycle of PHM (Fig. 1B) involves a reductive phase in which the enzyme is reduced by ascorbate to a dicopper(I) reduced intermediate and an oxidative phase in which the reduced enzyme reacts with peptide substrate and dioxygen to generate the hydroxylated product. As depicted in Fig. 1, many of the proposed mechanisms have suggested a Cu M (II)-peroxide or hydroperoxide as the hydroxylating species, formed from the reaction of PHM-Cu(I) 2 with O 2 and subsequent transfer of one electron from each Cu(I) center to dioxygen. If such an intermediate exists, then it should also be generated directly from the reaction of hydrogen peroxide with the oxidized (dicopper(II)) enzyme in the presence of substrate.
Measurement of Peptidyl-␣-hydroxyglycine Product Using HPLC-To test whether this peroxide shunt chemistry occurred in PHMcc, we measured the reaction of the oxidized enzyme with dansyl-YVG and hydrogen peroxide. Initial concentrations of reagents were 200 M substrate, 1 mM peroxide, and 5 M PHM in 100 mM MES buffer, pH 5.5. Aliquots were sampled at 30-s time intervals, and product was separated from substrate by HPLC, using the fluorescence of the dansyl group for detection (Fig. 2a). Under these conditions, all of the substrate was converted into product within 2-3 min. Further, since the substrate was present in 40-fold excess over the enzyme, at least 40 enzyme turnovers occurred, implying that the reaction was catalytic. Control experiments where peroxide was incubated with substrate and 5 M Cu 2ϩ in the absence of PHM or in the presence of PHM that had been heated at 90°C for 30 min gave no product (data not shown), demonstrating that the reaction was enzymatic. Fig. 2 shows rate data for the peroxide reactivity over a range of H 2 O 2 concentrations (Fig. 2c) or dansyl-YVG concentrations (Fig. 2d). Kinetic parameters extracted from these plots are listed in Table 1. Table 1 also lists kinetic parameters for the ascorbate/O 2 -dependent reaction (Fig.  2b). These data show that although the rate of the peroxide reaction is much slower than the ascorbate/O 2 reaction under the standard reaction conditions of 1 mM H 2 O 2 and 200 M dansyl-YVG, this is due primarily to a large increase in the value of K m for the dansyl-YVG substrate, which is 2 orders of magnitude larger than for the ascorbate reaction. A modest decrease in k cat from 9.2 to 5 s Ϫ1 is observed for the peroxide reaction.
The large increase in K m for the peroxide pathway suggests that the dansyl-YVG binds to a different form of the enzyme than in the ascorbate/O 2 -dependent pathway, consistent with reactivity occurring within the oxidized rather than the reduced form of the enzyme. To gain further insight into this possibility, we determined the dissociation constant for dansyl-YVG to PHM in its Cu(I) and Cu(II) forms. PHM was titrated with aliquots of dansyl-YVG in an ultrafiltration cell at 23°C,   and a small volume of filtrate was extracted for measurement of the free substrate concentration from its fluorescence signal. Plots of fractional formation of enzyme-dansyl-YVG versus free substrate were fit to theoretical curves for substrate binding. The results are shown in Fig. 3. The K D values were 22.5 and 145 M for the Cu(I) and Cu(II) forms of PHM, respectively. The K D for Cu(I)-PHM is in the same range as K m for the ascorbate/O 2 -dependent pathway, consistent with the dansyl-YVG binding to the Cu(I) form as predicted by the general mechanism of Fig.  1. The K D for Cu(II)-PHM is about 7 times larger than the reduced K D and indicates that structural factors in the oxidized enzyme weaken the binding. Thus, the increased K m measured for the peroxide pathway, although larger than K D for Cu(II)-PHM, is more consistent with substrate binding to the Cu(II) than the Cu(I) form of the enzyme. We also examined the possibility that hydrogen peroxide does not interact directly with the dicopper(II) enzyme but instead reduces it to the dicopper(I) form and that product forms by reaction of the dicopper(I) form with dioxygen and substrate. Here peroxide would be fulfilling the same role as ascorbate. As a test for this reductive peroxide pathway, we studied the reaction of PHM with peroxide and substrate under anaerobic conditions. For the anaerobic reactions, stock H 2 O 2 solutions were vacuum-flushed with argon to remove oxygen and diluted with anaerobic buffer. All reactions were performed in an anaerobic chamber. Under these conditions, reaction rates were found to be almost identical to those measured under aerobic conditions (data not shown), which would appear to rule out a simple reductive role for peroxide. However, subsequent experiments showed that products of the PHM catalytic reaction are able to accelerate the spontaneous disproportionation of peroxide to dioxygen and water, such that anaerobiosis could never be achieved by the above procedure. A more sophisticated method of obtaining anaerobic conditions utilizing the glucose/ glucose oxidase reaction needed to be developed to test the effect of anaerobiosis on the PHM peroxide reactivity (see below).
It has been shown (25) that the K m for peptidylsubstrates is unaffected by the nature of the reductant. To test this with the dansyl-YVG substrate, we determined K m,dansyl-YVG using N,NЈ-dimethylphenylenediamine as reductant. The measured K m was 4.1 M as compared with 5.5 M for ascorbate. These data further suggest that peroxide is not merely fulfilling a reductive role, since if this were the case, K m,dansyl-YVG would not be expected to increase.
Isotope Distribution in the Peptidyl-␣-hydroxyglycine Product-Electrospray ionization mass spectrometry was used to determine the oxygen isotope composition in the hydroxylated product. In previous studies, it was shown that when PHM reacts by the ascorbate/O 2 pathway, the oxygen atom incorporated into the ␣-hydroxy group of the ␣-hydroxyglycine product is derived entirely from molecular oxygen (26). If the peroxide pathway represents a peroxide shunt, then the ␣-hydroxy oxygen atom should similarly be derived from peroxide (22). Accordingly, we carried out reactions using H 2 Table 1. anaerobic conditions gave peaks corresponding to ␣-18 OH (m/z 589, 90%) and ␣-16 OH (m/z 587, 10%), as expected for the 90% enrichment of 18 O in the labeled peroxide. This confirmed that the oxygen atom was derived from peroxide rather than from solvent. However, when H 2 18 O 2 reacted under aerobic conditions, only 35% of the product was labeled with 18 O, and the remainder had exchanged with 16 O. Since solvent exchange could be ruled out from the previous experiment, this result implied that the oxygen in the product was derived in part from molecular oxygen. An experiment where H 2 16 O 2 was reacted with substrate in the presence of 99 atom % 18 O 2 yielded both the ␣-16 OH and ␣-18 OH products in the ratio 40:60. These results indicate that, as expected for a monooxygenation reaction, the oxygen atom at the ␣-OH group is derived from peroxide but that a pathway exists for this oxygen to exchange with oxygen from molecular oxygen. Two mechanisms appear plausible for this: (i) reaction of peroxide with the dicopper(II) enzyme may lead to an intermediate that is in rapid equilibrium with a Cu(I)-dioxygen species that can subsequently exchange with atmospheric O 2 , or (ii) like ascorbate, peroxide is acting as a reducing agent and forms the dicopper(I) intermediate, which itself generates product through the dicopper(I)-dioxygen route. As discussed above, the latter scenario would predict that product formation should not occur (or should be dramatically slower) under anaerobic conditions, contrary to observation. However, nonenzymatic peroxide disproportionation could potentially generate sufficient oxygen to drive the reaction even under formally anaerobic conditions. To test this possibility, we measured the ratio of peroxide consumed to product formed.
Stoichiometry of the Peroxide Reaction- Table 3 compares peroxide consumed with peptidyl-␣-hydroxyglycine product formed for a number of different determinations. It is clear that peroxide consumption exceeds product formation by 2-3-fold. Oxygen electrode measurements showed that peroxide was decomposed to molecular oxygen and water in a nonenzymatic chain (Haber-Weiss) reaction,

REACTIONS 1 AND 2
This reaction was accelerated in the presence of the components of the PHM catalytic reaction, suggesting that PHM catalysis generates a species capable of initiating and/or propagating nonenzymatic peroxide decomposition (for further details, see supplemental Fig. S1 and accompanying discussion). The presence of oxygen from the decomposition of peroxide may explain why the rate of the PHM-peroxide reaction does not decrease under "anaerobic" conditions, since O 2 is continuously generated by peroxide decomposition. Also in mixed isotope experiments such as the reaction of PHM with H 2 16 O 2 under 18 O 2 , the product would still contain 16 O even if peroxide acted solely as a reductant due to 16 O 2 generation from nonenzymatic peroxide decomposition.
We therefore sought to monitor product formation under conditions where the nonenzymatic reaction was suppressed or eliminated. We used the reaction of glucose oxidase in the presence of 50 mM glucose and O 2 -saturated buffer to generate a 1.2 mM solution of anaerobic hydrogen peroxide. This reaction was allowed to proceed in a sealed oxygen electrode cell with no head space until all of the oxygen had been consumed (Fig. 5). Then PHM and dansyl-YVG were introduced into the cell, and the rate of peptidyl-␣-(OH)-glycine was determined by HPLC. This system had the advantage that any oxygen produced by peroxide decomposition was rapidly recycled to hydrogen peroxide by the GO reaction without releasing any superoxide into solution (27). Fig. 5 shows oxygen electrode traces corresponding to the GO reaction. The system was able to absorb oxygen entering the cell even when the cap was removed, and the cell was exposed to atmospheric O 2 . With the cell sealed, the peroxide concentration after all oxygen had been consumed was found to be 1.2 mM, and it remained unchanged, provided that no further oxygen was introduced into the system. Fig. 5 shows the rates of product formation from the PHM/GO system compared with appropriate controls. Of great significance, the rate of product formation did not decrease when peroxide was kept strictly anaerobic. Furthermore, peroxide consumption was now strictly coupled to product formation, with the ratio of peroxide consumed to product formed equal to unity (Table 3). This result suggests that peroxide reacts with PHM to generate product by a pathway that does not rely on simple reduction to the dicopper(I) form and reaction of the latter with dissolved oxygen, since the concentration of dissolved O 2 in the PHM/GO system is vanishingly small. Rather, the mechanism must involve formation of a catalytic intermediate from Cu(II) and peroxide within the PHM catalytic cavity. The scrambling of the isotope label observed under aerobic conditions then suggests that a Cu(I)-O 2 species must be a common intermediate.
In order for the peroxide pathway and the ascorbate/O 2 pathways to   3 Stoichiometry of dansyl-YV-(OH)-Gly production to hydrogen peroxide consumed for the standard PHM peroxide reaction (200 M dansyl-YVG ؉ 5 M Cu 2؉ ؉ 5 M PHM ؉ 1 mM hydrogen peroxide, pH 5.5) and for the reaction carried out anaerobically in the presence of 50 mM glucose ؉ 45 g/ml glucose oxidase In the latter reaction, the H 2 O 2 was produced in situ from the reaction of the glucose/glucose oxidase with oxygen-saturated MES buffer and was measured to be 1.2 mM. To test this hypothesis, we measured the ability of peroxide to reduce the copper centers of PHM at pH 5.5 using EPR spectroscopy as shown in Fig. 6. The EPR integrated intensity dropped to 75% of the fully oxidized control. These experiments demonstrate that peroxide can reduce the Cu(II) centers in PHM. The extent of reduction did not increase with increased time of incubation and represented only a fraction of the total copper in the protein. Thus, it is likely that an equilibrium exists between Cu(II)-peroxo and Cu(I)-superoxo and/or Cu(I)-dioxygen species, which must be accounted for in any mechanism for peroxide reactivity. H172A and H242A Mutants-A true peroxide shunt requires no additional electrons to complete the monooxygenase reaction. Thus, if a Cu M (II)-peroxo entity were the reactive species, it could produce peptidyl ␣-hydroxyglycine and water without the need for electron transfer from the other copper center (Cu H ). This would imply that the ascorbate/O 2 pathway might be abrogated in mutants with impaired function at Cu H , whereas the peroxide reactivity was unaffected. Accordingly, we measured the ability of H172A, a Cu H mutant known to retain less than 1% of wild type activity (23), to generate product via the peroxide pathway. The H172A mutant was incubated with 1 mM peroxide and either 400 M or 1 mM dansyl-YVG. The reaction was allowed to proceed for up to 60 min, 20 times longer than the time required by the wild type protein to convert 200 M of substrate completely into product. The results showed that H172A produced no observable product.

Sample
We also tested the peroxide reactivity of Cu M site deletion mutants. H242A has been shown to bind copper only at the Cu H site (11). The H242A mutant was also unable to produce product, indicating that the peroxide reactivity is not centered at Cu H , since it does not proceed when only Cu H is occupied.

DISCUSSION
We have demonstrated that PHM is able to catalyze the hydroxylation of peptidylglycine substrates starting from the oxidized enzyme and using hydrogen peroxide as the only source of oxygen. When peroxide labeled with 18 O was reacted with PHM and substrate in the presence of 16 O 2 , only 35% of the label was incorporated into the product. This scrambling of the label was not due to solvent exchange, since full incorporation of 18 O occurred when both the peroxide and the ambient oxygen were labeled with 18 O or when labeled peroxide was reacted under anaerobic conditions. Hence, the peroxide pathway must generate an FIGURE 5. The PHM/dansyl-YVG/H 2 O 2 reaction carried out in the presence of glucose and glucose oxidase. The glucose/glucose oxidase system was used to generate H 2 O 2 under anaerobic conditions and to eliminate oxygen production via peroxide disproportionation. Reactions were performed in a Rank Brothers oxygen electrode at 25°C. 100 mM MES buffer (pH 5.5) containing 50 mM D-glucose was added to the cell and allowed to equilibrate, and the response was adjusted to 21%. The base line was allowed to stabilize, the original buffer was removed, and fresh buffer at 100 or 21% oxygen saturation was added. The stirred cell was capped, and glucose oxidase was immediately added with a Hamilton syringe to a concentration of 43 g/ml. Once all of the oxygen was converted to peroxide, substrate and PHM were added with a Hamilton syringe to final concentrations of 200 and 5 M, respectively. Top, time course of the PHM-peroxide reaction in the presence of glucose/glucose oxidase. Middle, O 2 -electrode traces showing the response of the O 2 concentration to glucose/glucose oxidase. Arrow a, glucose oxidase added to a 50 mM solution of glucose in a capped O 2 -electrode cell; arrow b, the cell cap was removed and stirring continued, but the solution remained anaerobic; arrow c, the glucose oxidase was inactivated by reaction with 1% trifluoroacetic acid, and the oxygen level returned quickly to the air-saturated value. Bottom, comparison of reaction rates for aerobic (standard reaction) and anaerobic (glucose/GO) reactions. The anaerobic reactions used H 2 O 2 generated from the glucose/GO reaction at either 21 or 100% initial oxygen concentration. The standard aerobic reactions used 1 mM peroxide and were run in the presence of 50 mM glucose, 45 mg/ml GO, or 200 g/ml SOD as controls. Other reagents were 200 M dansyl-YVG, 5 M Cu 2ϩ , 100 mM MES, pH 5.5, and 5 M PHM. intermediate that is capable of exchange with atmospheric dioxygen. We demonstrated that peroxide is decomposed in a nonenzymatic process to dioxygen and that the decomposition is accelerated when PHM is catalyzing substrate hydroxylation, suggesting that PHM catalysis generates a species capable of initiating and/or propagating nonenzymatic H 2 O 2 decomposition. However, we were able to fully suppress the nonenzymatic peroxide decomposition by carrying out the PHM/ peroxide reaction in the presence of glucose and glucose oxidase, which recycled any free oxygen back to peroxide and thus provided a strictly anaerobic environment. Using this system, we were then able to show that the PHM/peroxide reaction was tightly coupled and that its rate did not decrease when oxygen was totally excluded from the bulk solution.
Mixed labeling of the product with O from both peroxide and atmospheric dioxygen could occur if peroxide acted primarily to reduce the Cu(II) centers to Cu(I). In this scenario, it would be fulfilling the same role as ascorbate and hence would produce a dicopper(I) intermediate that would be poised for reaction with dioxygen in solution. Although the Michaelis constant for binding of O 2 to the enzyme-dansyl-YVG complex has not been measured, K m,O 2 for dansyl-Gly-Gly-Ser is 73 M at pH 6 and 37°C (18). Since the PHM reaction is known to be equilibrium-ordered with O 2 binding to the enzyme-peptidylglycine complex (24), the K m,O 2 will vary with substrate, but it is unlikely that K m,O 2 for dansyl-YVG would vary substantially from that for dansyl-GGS. Using this assumption, we would predict that k cat for the reductive pathway should decrease dramatically when the O 2 concentration in the bulk solution decreased to undetectable levels as measured by the O 2 electrode in the glucose/glucose oxidase system. Since we observed no decrease in the rate of the PHM-peroxide reaction, we conclude that peroxide cannot be simply fulfilling the role of reductant and must be generating reactive oxygen species within the PHM active site cavity.
The increase in K m,dansyl-YVG observed in the peroxide reaction and mirrored in the K D for dansyl-YVG binding to the Cu(I) and Cu(II) forms of PHM provides strong corroborating evidence that the peroxide reactivity resides primarily within the oxidized enzyme.
The observation that isotopic molecular oxygen is able to exchange into product generated from hydrogen peroxide and oxidized enzyme implies that an intermediate exists along the reaction pathway that is electronically equivalent to metal-bound dioxygen. The most likely candidate for this intermediate would be a Cu M (I)-dioxygen complex. Thus, for Cu(II)-peroxo to be a viable intermediate, it must be in equilibrium with Cu(I)-dioxygen. This would require that the long range electron transfer from Cu H to Cu M be reversible. As discussed in the Introduction, a number of novel suggestions have been necessary to explain the absence of a direct (through-bond) ET pathway from Cu H to Cu M , including substrate mediation (6), superoxide channeling (10), or oriented solvent (16). Given these constraints on the available ET pathways, we consider reversible electron transfer to be highly improbable, as has also been argued by Klinman and co-workers (16).
Another test of the viability of a Cu(II)-peroxo would be the formation of peptidyl ␣-(OH)-glycine from the reaction of mutants that lacked a functional Cu H center with peroxide and Cu(II)-PHM, since the Cu(II)-peroxo species does not require additional ET for activity. We found that the Cu H mutant H172A was unable to catalyze product formation from peroxide and oxidized enzyme. This mutant has been shown to be less than 1% as active as wild type PHM in the ascorbate/ dioxygen pathway (23), and it has been suggested that the decrease in activity is due to impaired ET from the modified Cu H site. The complete absence of product formation even after 60 min may suggest that H172A abrogates the peroxide pathway in some additional way, perhaps by impairing the ability of peroxide to reduce the Cu H center. As a control, we also tested the Cu M site mutant H242A, which we have previously shown causes loss of copper binding at the M center (11). If this mutant was active in the peroxide reaction, it would suggest that peroxide binding and reactivity could occur at Cu H . This mutant was similarly inactive in the peroxide reaction. These data provide compelling evidence that electron transfer from Cu H to Cu M is still obligatory in the peroxide pathway and hence that a Cu M (II)-peroxo cannot be the catalytic intermediate.
A mechanism for the peroxide reactivity that is in accord with all of our data is shown in Fig. 7. We propose that hydrogen peroxide reacts initially at Cu H (II), reducing the copper atom and forming a species that is electronically equivalent to Cu(I)-superoxide but may also be protonated. The superoxide formed can either react further within the PHM cavity or be lost to the bulk solution. The entity most likely to react with superoxide inside the cavity would be the Cu M (II), which could capture FIGURE 7. Suggested mechanism for the production of hydroxylated product in the PHM/dansyl-YVG/H 2 O 2 reaction. A, the peroxide reacts initially at Cu H (II), reducing the copper atom and forming a species that is electronically equivalent to Cu(I)-superoxide. The superoxide formed can either react further within the PHM cavity or be lost to the bulk solution. B, the superoxide reacts inside the cavity with Cu M (II), which captures the superoxide as a Cu(II)-superoxide species electronically equivalent to or in equilibrium with Cu(I)-dioxygen, providing a mechanism for oxygen exchange with atmospheric molecular oxygen.