Catalytic Function of Tyrosine Residues in para-Hydroxybenzoate Hydroxylase as Determined by the Study of Site-directed Mutants*

The role of protein residues in activating the sub- strate in the reaction catalyzed by the flavoprotein p-hydroxybenzoate hydroxylase was studied. indicates that Tyr-201 and Tyr-385 form a hydrogen bond network with the 4-OH of p-hydroxybenzoate. Therefore, site directed mutants were constructed, converting each of these tyrosines into phenylalanines. Spectral (visible and fluorescence) properties, reduction potentials, and binding constants are very similar to those of wild type, indicating that there are no major structural changes in the mutants. In the absence of substrate, the mutants and wild type exhibit similar pH-dependent changes in the FAD spectrum. However, the enyme-substrate complex of Tyr-201- Phe lacks an ionization observed in both wild type and Tyr-385 + Phe, which preferentially bind the phenolate form of substrates. Tyr-201 + Phe shows no preference, indicating that Tyr-201 is required to ionize the sub- strate. The mutants less than 6% the activity of at least 100-fold faster than their decay, at least 95% of the enzyme was in the transient state. The spectra of transient intermediates for the WT enzyme have been reported before (9).

The mutants have less than 6% the activity of the wild type enzyme. The effects on catalysis were studied by stopped flow techniques. Reduction of FAD by NADPH is slower by 10-fold in Tyr-201 Phe and 100-fold in Tyr-385 + Phe. When the reduced Tyr-201 + Phe-p-hydroxybenzoate complex reacts with oxygen, a long-lived flavin-C(4a)-hydroperoxide is observed, which slowly eliminates H202 with very little hydroxylation. Thus, the role of Tyr-201 is to activate the substrate by stabilizing the phenolate. Tyr-385 + Phe reacts with oxygen to form 25% oxidized enzyme, and 75% flavin hydroperoxide, which successfully hydroxylates the substrate. This mutant also hydroxylates the product (3,4-dihydroxybenzoate) to form gallic acid. para-Hydroxybenzoate hydroxylase (EC 1.14.13.2) (p-OHbenzoate hydroxylase),' which catalyzes the reaction in Equation l , has become the model for the study of flavoprotein monooxygenases.
NADPH + H' + p-OHB + 0, (1) -+ 3,4-DOHB + NADP' + HZ0 A large body of mechanistic (1-4) and crystallographic information (5-8) has been published for the enzyme isolated from Pseudomonas fluorescens. Several intermediates in both the reductive and oxygenative half-reactions of the catalytic cycle have been identified and characterized (9-14). Analogous intermediates have been found in the catalytic pathway of many other flavoprotein monooxygenases (e.g. [15][16][17][18][19][20]. The gene for p-OH-benzoate hydroxylase (p06A) has been isolated from Pseudomona aeruginosa (by R. H. Olsen, University of Michigan) and studied (21). It was shown that the enzyme from this related source was almost identical in sequence to that from P. fluorescens (22). Of the 394 residues, p-OHbenzoate hydroxylase from P. aeruginosa differs from the enzyme of P. fluorexens by only two amino acids, and these are very conservative changes. This explains the nearly identical kinetic and spectral properties (9) and the isomorphous crystalline state of the enzyme from both sources (23). Thus, the enzyme from either source can be used interchangeably for structural and mechanistic investigations. Very recently, the p06A gene from P. fluorescens has also been isolated and expressed in Escherichia coli (24). Although the protein coding portion of the gene from this related bacterium is very similar, the regulatory sequences are substantially different.* para-Hydroxybenzoate hydroxylase is a particularly good model for testing chemical explanations for structure and catalysis in proteins. It is a homodimer with each 45-kDa subunit independently catalyzing a multisubstrate, irreversible reaction. The most significant advantage for study is the presence of FAD in the active site, which takes part in, and acts as a reporter group for chemical intermediates in the reaction. Thus, it has been possible to characterize individual steps in catalysis (10).
A. H. Westpbal, personnal commununication. and the oxidative half-reaction, in which the reduced enzymep-OHB complex reacts with 0 2 to form the product (Scheme I) (1,25,26). An important biological control feature is seen in the reductive half-reaction. The binding of substrate greatly stimulates (by a factor of lo5) the rate of reduction of the enzyme by NADPH (13, 27,28). At the conclusion of the reduction reaction, the oxidized pyridine nucleotide dissociates from the reduced enzyme to complete the first half of catalysis. Since the presence of substrate has very little effect on either the binding constant for NADPH or the redox potential of the enzyme-bound flavin (25), the effect of substrate must be on conformational properties of the enzyme-NADPH complex. Although the binding constants for substrate to oxidized and reduced forms of the enzyme are nearly identical, the rates for association and dissociation of p-OHB are lO"-fold slower with the reduced enzyme (10). This property may help in assuring that the substrate does not dissociate before oxygen reacts with the reduced flavin.
In the second half of catalysis, the reduced flavin reacts with oxygen (probably through the channel vacated by NADP+), forming a C(4a)-flavin hydroperoxide, which acts as the oxygenating agent (9, 10). The terminal oxygen atom is transferred, forming the product, 3,4-dihydroxybenzoate (3,4-DOHB), and a C(4a)-flavin hydroxide. After the elimination of water from the flavin and the release of product, oxidized enzyme is regenerated.
Complexes of the enzyme with substrate or product have been structurally characterized by x-ray crystallography (6, 29). One striking feature of the structure is the presence of 3 tyrosine residues in intimate hydrogen bonding association with the substrate, p-hydroxybenzoate (p-OHB). The 4-hydroxyl of the substrate (or product) is hydrogen bonded to the hydroxyl of Tyr-201, which is also hydrogen bonded to the hydroxyl of Tyr-385, forming a hydrogen bond network (29), as illustrated in Fig. 1. There are no other protein groups in the substrate-binding site with obvious catalytic potential. Thus, these residues are of primary interest for analysis of protein involvement in catalysis. Previously attempted chemical modification of these residues resulted in problems common to such methods, such as the lack of specificity, and derivatization that eliminated activity (30). This paper reports properties of two site-directed mutants produced from the pobA gene from P. aeruginosa, and expressed in E . coli  defined differences from wild type in both the reductive and oxygenative half-reactions. These differences provide information on the role of the protein in catalysis. A preliminary account of some of this work has been reported (23).

EXPERIMENTAL PROCEDURES
Materials-The following compounds were purchased from commercial sources and purified by recrystallization: p-OHB, 2,4-DOHB, 3,4-DOHB, and 4-hydroxycinnamic acid. The following compounds, of the highest purity available, were used as purchased from Sigma: NADPH, FAD, isopropylthiogalactoside, riboflavin, and ampicillin. Analytical mixtures of oxygen and nitrogen were purchased from Matheson. Dry argon was passed through a column of Ridox (Fisher), and then a solution of reduced methyl viologen to remove traces of oxygen, and through H20 to humidify.
The following restriction endonucleases were purchased from Pharmacia LKB Biotechnology Inc.: SphI, PstI, ClaI, and BamHI. New England Biolabs provided AsuII, and the isoschizomer, Csp451, was from Promega. Other reagents for DNA manipulation were obtained from Boehringer Mannheim and Pharmacia.
A dideoxynucleotide sequencing kit using the Klenow enzyme came from Bresatech, and one using bacteriophage T 7 DNA polymerase came from U. S. Biochemicals. A kit for site-directed mutagenesis by the Kunkel method (31) was obtained from Bio-Rad. Samples of E. coli JM-105 and double-stranded DNA from the bacteriophage M13 mp19 were obtained from Pharmacia.
Plasmids-Enzymes were produced from the genepobA, which was identified originally in a BamHI fragment from a genomic library of P. aeruginosa in the plasmid pR01931 by its ability to complement a mutant lacking functional pobA (Dr. Ronald Olson, University of Michigan). The stock genetic material for this work was the plasmid, pNE101, which contained the protein coding portion of pobA (21), and was purified from E. coli.
The pNE130 expression vector used to produce wild type and mutant enzymes was prepared from the plasmid pNElOO (21). In pNE100, pobA is aligned with the lac promoter in an orientation opposite to that in pNElOl and is out of phase with the lac2 fragment. pNElOO was cut with restriction enzymes ClaI and BanHI. The 4.24kb DNA fragment generated was end-filled and ligated to form pNE130 with the retention of a unique BamHI site. pNE130 contains pobA under control of the lac promoter, amp resistance, and a multicopy origin of replication.
T o provide DNA for mutation, double-stranded DNA from the phage M13 mp19 (32) and pNElOl was cut with restriction enzymes SphI and PstI. The 1.75-kb piece of DNA removed from pNElO1 containing the protein coding portion of pobA was ligated into the M13 polylinker region. This construct was used to produce purified single-stranded phage DNA-containing uracil substitutions, according to the general methods of Kunkel et al. (31).
Mutations were generated from the phage template DNA by the Bio-Rad adaptation of the method of Kunkel et al. (31). Mismatched oligonucleotides (21-mer) were annealed to the single-stranded DNA, where they primed the synthesis of the complementary strand. Phage plaques were isolated from transformed E. coli. Mutants were identified by dideoxynucleotide sequencing (33).
To transfer each mutation to the expression vector, mutated phage was isolated in the double-stranded DNA form, cut with the restriction enzymes AsuII (which uniquely cuts near the middle of pobA) and BamHI (which cuts uniquely past the 3'-end of pobA), and the 1.32-kb fragment was isolated. The same enzymes were used to cut pNE130 (4.24 kb), and the 3.12-kb fragment (containing the remaining fragment ofpo6A) was purified. These two fragments were ligated to form a 4.44-kb plasmid, containing a mutant of pobA. E. coli J M -105 (32) was transformed with the plasmid, and mutants were identified by the unique size of the plasmid. A single isolate of each mutant plasmid was selected for enzyme production in E. coli J M -105.
The sequences of the mutated segments of pobA in the final expression plasmids were verified by the dideoxynucleotide procedure using bacteriophage T 7 DNA polymerase (34). Oligonucleotide primers that generated overlapping sequences from a single sample of purified plasmid DNA were used. Only the expected changes from wild type were found in each case.
Enzyme Production-To produce enzyme in E. coli JM-105, cells were plated on minimal salts-glucose agar with 30 pg ml" ampicillin to select for the expression and episome plasmids and grown at 37 "C.
A few colonies were spread on a plate of Luria-Bertani broth plus 30 pg ml" ampicillin. After 12 h a t 37 "C, a quarter of the plate was transferred to a Fernbach flask containing about 650 ml of 4/3 X Luria-Bertani broth (35) and 80 pg ml" ampicillin. After shaking for 3.5-4.0 h at 37 "C, the contents of the flask were transferred to a fermentor containing 10-11 liters of 4/3 X Luria-Bertani broth and 30 pg ml" ampicillin a t 37 "C. Following a lag of about 1 h, the cells grew in log phase for 2.5-3.0 h. The pH was held a t or below 7.0 by the addition of 6.0 M HCl. At the end of log phase, at a cell density of 6 g of compacted cells/liter, the culture was induced with isopropylthiogalactoside (0.5 mM final) and supplemented with riboflavin (0.1 mM final). Induction caused a linear increase of active, soluble p-OH-benzoate hydroxylase with time. Cells ceased growth a t 7 g of compacted cells/liter and were harvested 5.0 h after induction, before a severe decline in cell density. Compacted cells were frozen a t -70 "C until used.
The enzyme was extracted and purified from compacted cells using the same methods developed for P. aeruginosa (3). Cells contained 1.8-2.0% of cellular protein as p-OH-benzoate hydroxylase. Thus, one fermentor load gave 80-100 mg of purified enzyme. Mutants and wild type behaved the same in purification. Experimental Conditions-Unless otherwise stated, the results were obtained under conditions favorable for the detection of catalytic intermediates (10): 50 mM phosphate buffer (K+ salt) and 10 mM EDTA (Na+ salt), pH 6.5, a t 3 "C.
Extinction Coefficients-Enzyme solutions were prepared in 0.05 M sodium phosphate, pH 7.2, by gel filtration, and spectra recorded at 25 "C before and after the addition of 0.1% sodium dodecyl sulfate. Denaturation of the enzyme was complete within 30 s, without any precipitate being formed, permitting the determination of the extinction coefficient of the enzyme-bound flavin relative to that of free FAD (11.3 mM" cm"). The spectrum of FAD is unaffected by sodium dodecyl sulfate.
Reduction Potentials-The reduction potentials of the purified enzymes were determined at pH 7.0, 25 "C, by a simple spectrophotometric method employing a reducing system of xanthine and xanthine oxidase (36).
Dissociation Constants-Spectral perturbations (absorbance or fluorescence) of the enzyme-bound FAD were followed upon titration with the ligands, p-OHB, 2,4-DOHB, or 3,4-DOHB. Titration results were analyzed graphically (double reciprocal) or by non-linear least squares fitting to obtain the dissociation constants.
Product Analysis-Product formation was demonstrated by analyzing reaction mixtures containing 0.5 mM p-OHB, 20 p~ NADP+, a n NADPH regenerating system (25 mM glucose 6-phosphate and glucose-6-phosphate dehydrogenase), and catalase to recycle oxygen. Reaction mixtures were air-saturated. The reactions were initiated by the addition of p-OH-benzoate hydroxylase. Aliquots were periodically acidified with trifluoroacetic acid, spun, and separated isochratically with a C18-HPLC column, using 20% CH30H, 1% acetic acid. The substrates and products were identified by coinjection with authentic compounds.
The extent to which oxygen consumption resulted in hydroxylated products rather than Hz02 was determined by the analysis of airequilibrated reaction mixtures containing various amounts of substrate and NADPH. Reactions were initiated by the addition of p -OH-benzoate hydroxylase and monitored by absorbance changes a t 340 nm. Aliquots were withdrawn periodically, quenched with trifluoroacetic acid, spun, and chromatographed as above. Both the amounts of substrate consumed and product formed could be measured, and compared to NADPH consumed (Table I).
Stopped Flow Spectrophotometry-Experiments were carried out on a 2-cm path length stopped flow spectrophotometer built by one of us (D. P. B.), and described in detail elsewhere (15). The instrument was particularly suited to anaerobic work at low temperature. The absorbance changes recorded were digitized and stored in a computer. Rate constants were obtained from exponential fits using the programs FIT-87 (developed by C. Batie, University of Michigan) or Program A (developed by Chung-Jen Chiu and Rong Chang, University of Michigan) based on the Marquardt algorithm (37).
Turnouer Numbers-The maximum turnover number for each enzyme was obtained by enzyme-monitored turnover in a stopped flow spectrophotometer as described by Gibson et al. (38).
Kinetic Analysis of Transient States in Catalysis-Reduction of each enzyme by NADPH was studied in the stopped flow spectrophotometer by anaerobically mixing enzyme solutions with solutions of NADPH with or without saturatingp-OHB. Reduction of the enzyme flavin was followed a t 450 nm. The maximum rate constants for reduction were obtained from double-reciprocal plots of the observed rate constant versus NADPH concentrations (39). Transient charge transfer complexes between FAD and pyridine nucleotide were observed at wavelengths greater than 520 nm beyond the absorbance of oxidized enzyme.
The oxygen half-reaction was studied by reacting reduced p-OHbenzoate hydroxylase with OZ-containing buffer. Substrate-saturated or substrate-free enzyme in a glass tonometer was made anaerobic by repeated cycles of evacuation and equilibration with argon and was reduced with a buffered solution of sodium dithionite. After complete reduction of the enzyme, a 25% excess of dithionite was added. Accurate results depended on the maintenance of fully reduced enzyme at the start of each reaction.
The oxygen dependence of the reactions was examined in concentrations ranging from 0.13 to 1.1 mM. Rate constants for the interconversion of intermediates were obtained by analyzing traces collected a t appropriate isosbestic points.
The spectra of intermediates were obtained by observing reactions of reduced enzyme with 0.65 mM oxygen at a large number of wavelengths and reading the absorbance a t 20 ms. Because the observed intermediates were formed at least 100-fold faster than their decay, a t least 95% of the enzyme was in the transient state. The spectra of transient intermediates for the WT enzyme have been reported before (9).

RESULTS
Reaction Catalyzed-The maximum turnover number for W T p-OH-benzoate hydroxylase is 5.7 s" at pH 6.5, 4 "C, as measured by oxygen consumption in enzyme-monitored turnover. Changing either of the tyrosines to phenylalanine reduced this value significantly; the Tyr-201-Phe mutant has a turnover number of 0.42 s-l, and the Tyr-385 + Phe mutant has a turnover number of 0.36 s-'.
Analysis of products from reactions catalyzed by the Tyr-201 -Phe mutant showed that only 5-6% of the NADPH consumed resulted in the formation of 3,4-DOHB, with the remainder of the reducing equivalents nonproductively forming H202. The behavior of the Tyr-385 + Phe mutant was more complex. It was found to hydroxylate p-OHB to form 3,4-DOHB, and additionally, to catalyze the further hydroxylation of 3,4-DOHB to form 3,4,5-trihydroxybenzoate (gallic acid, see Fig. 2). To measure the amount of NADPH leading solely to the formation of 3,4-DOHB, reaction mixtures containing a limiting concentration of p-OHB were analyzed; to prevent the formation of gallic acid, protocatechuate dioxygenase, which catalyzes the cleavage of 3,4-DOHB, was also included. By this method, the Tyr-385 --., Phe enzyme was found to hydroxylate p-OHB with an efficiency of 75%. The efficiency of formation of gallic acid was measured by using 3,4-DOHB as a substrate. In this case, 20% of the NADPH consumed led to gallic acid formation. A dissociation constant of 66 p~ for 3,4-DOHB was independently measured for this mutant enzyme, compared to a value of 230 p~ for WT. Enzyme assays at pH 6.5 and 7.9 showed that p-OHB and 3,4-DOHB cause NADPH to be oxidized at about the same rate, in contrast to the WT enzyme, in which there is a 10fold difference in rates between p-OHB and 3,4-DOHB at pH 8.0 (40). Thus, the Tyr-385 + Phe mutant has lost an interesting feature of the WT, the ability to discriminate the substrate from the chemically more reactive product.
Physical Properties of Proteins-It should be noted that the mutants described here behave exactly the same as the WT enzyme during purification, implying that there are no major changes in the structures of the mutant proteins. Samples of each mutant have been crystallized and are being subjected to x-ray diffraction a n a l y~i s .~ The crystal structure of the WT enzyme (6) shows that there are no immediate contacts between either tyrosine 201 or tyrosine 385 and the isoalloxazine ring of the FAD. Thus, it is unlikely that a change from Tyr to Phe will have any large effect on the properties of the flavin. Indeed, the absorbance spectra of the mutant enzymes were almost identical with that of wild type enzyme, with t450 values of 10.3 f 0.1 mM" cm" for WT and both mutants. Similarly, the fluoresence properties were almost identical to those of wild type (Table I). There were no experimentally distinguishable differences in the redox potentials between the WT and mutants, a further indication that the flavin environment is unchanged. They each exhibited simple two-electron reductions in redox titrations, in the presence or absence of p-OHB; there is no thermodynamic stabilization of the semiquinone state of the :' X-ray analysis of the Tyr-385 -+ Phe mutant complexed with p-OHB shows the structure to be essentially the same as that for wild type (M. S. Lah, private communication).
flavin. There was no indication of other redox groups involved in the reaction, unlike an earlier report on the enzyme from P. fluorescens (41). In that report, it was found that more than two electrons were initially needed to reduce each molecule of the enzyme, due to the presence of a redox active cysteine on the surface of the enzyme. Our enzyme preparations did not have a mixed population of molecules with partly oxidized cysteine.
The mutations under study should disrupt the hydrogen bond network of phenolic groups of the tyrosines and the substrate. It might be expected that substrate binding would be weaker. However, the Kd values measured for two substrates were essentially unchanged for each mutant (Table I).
This result implies that these hydrogen bonds do not provide a net contribution to the stability of the enzyme-substrate complexes.
It has been reported that p-OH-benzoate hydroxylase, both free and in complex with p-OHB, has an ionizable group in the active site that results in marked changes in the absorbance spectrum (12, 42). We found that the pK, observed for the WT enzyme-substrate complex is due to binding of the substrate in the phenolate form. This was shown by titration of the enzyme with p-hydroxybenzoate at various pH values. In agreement with results reported by Shoun et al. (28) for the enzyme from Pseudomonas desmolytica, there is a large increase in absorbance centered around 285 nm due to binding of the substrate. The results of a typical experiment carried out at pH 7.5 are shown in Fig. 3. A double-reciprocal plot of the increase in absorbance at 286 nm versus the concentration of added p-OHB gave a straight line with an intercept equivalent to a AtzR6 of 9000 M" cm" and a Kd of 33 p~. In the same experiment the flavin fluorescence decreased to less than 20% the initial value, with a Kd for p-OHB of 31 p~.
Similar experiments a t different pH values show p-OHB binding in its phenolate form with a pK, of 7.4 (maximum AtzR6 of -15,000 M" cm"; see inset, Fig. 3). Ionization of free p-OHB shifts the X , , , from 250 to 280 nm, with a Acmax around 285 nm of -16,000 M" cm". It should be emphasized that the AtzR6 value of -15,000 M" cm" observed at high pH values cannot be ascribed to the ionization of a tyrosine residue, since such ionizations result in At of only -2,000 M" cm". The perturbation of the pKa of the bound substrate must be due to hydrogen bonding with Tyr-201 (see Scheme 2), since the Tyr-201+ Phe mutant (Fig. 3) shows absolutely no increase in absorbance in the 280-300 nm region, even though binding of p-OHB similar to WT is evident from the perturbation of the flavin absorption spectrum, and a quenching of the flavin fluorescence to 30% its initial value, with a A similar perturbation of the phenolic pK, is observed on binding the inhibitor, p-hydroxycinnamate to WT. Free in solution this compound in its phenolic form has a wavelength maximum at 287 nm, with an extinction coefficient of 19,200 M" cm"; in its phenolate form (pK, 9.1) the X, , , is at 334 nm, with an extinction coefficient of 22,300 M" cm". With this compound it is very easy to distinguish which ionization state of the inhibitor is bound, since the absorbance increase due to binding of the phenolate occurs in a region of minimal absorbance of the flavoprotein. Furthermore, the phenolate form has a bright blue fluorescence, with emission maximal at 440 nm, while the phenolic form is non-fluorescent.
Results of a typical experiment with wild type enzyme at pH 7.5 are shown in Fig. 4. Binding of 4-hydroxycinnamate is accompanied by a large increase in absorbance in the 300-400 nm range, with a difference spectrum maximum at 350 nm. It is also accompanied by development of fluorescence  a Activities refer to the rate of NADPH oxidation or the rate of oxygen consumption. I, Activities at pH 7.9 were measured a t 25 "C, using standard assay conditions (3) and are therefore not maximum Turnover numbers were measured at pH 6.5 and 3 "C by enzyme monitored turnover, and are maximum values Hydroxylation stoichiometries were determined at pH 6.5, 25 "C, and refer to the percentage of NADPH Dissociation constants were determined at pH 6.5, 3 "C. Formation of the phenolate form of the substrate upon binding to wild type p-OH-benzoate hydroxylase, but not upon binding to Tyr-201 -. , Phe. The enzymes in 0.1 M phosphate buffer, pH 7.5,25 "C, were titrated withp-hydroxybenzoate and absorption spectra recorded with a diode array spectrophotometer. Spectra recorded for the same concentrations of p-OHB at pH 7.5 in the absence of enzyme were subtracted from the enzyme spectra t o correct for the free p-OHB spectral contributions. The curves shown are the difference spectra, the corrected enzyme spectra in the presence of different quantities of substrate minus the enzyme spectrum in the absence of substrate. The top set of curves shows the difference spectra a t 14.9, 39.4, and 112 PM p-OHB added to 19.  Fig. 4 and indicate a pK, of 6.9 for enzyme-bound 4-hydroxycinnamate, versus one of 9.1 for the free compound. Similar binding of the phenolate anion occurs with the mutant enzyme, Tyr-385 + Phe, except that in this case the maximum extinction change at high pH values is not as great as with wild type enzyme (Fig. 4, inset). The Kd for binding Tyrosine Mutants ofp-Hydroxybenzoate Hydroxylase to this mutant is not significantly different from that with wild type enzyme, in the range 100-200 p~ from pH 7.1 to 8.5.
In contrast, binding of 4-hydroxycinnamate to Tyr-201 -+ Phe gave no increase in absorbance at 350 nm, either at p H 7.3 or at pH 8.5, nor was there any sign of the blue fluorescence typical of the phenolate. However, contrary to the quenching of the flavin fluorescence found with wild type enzyme, binding to Tyr-201 -+ Phe resulted in an increase in the flavin fluorescence (by 85% at pH 7.3, 46% at p H 8.1, and 14% at p H 8.5). While it is clear that binding of 4-OH cinnamate to this mutant does not result in a lowering of the phenolic pK,, as it does with wild type and Tyr-385 "-f Phe, the strength of binding is approximately the same as with W T ( K d of 130 ~L M at pH 7.3 and 100 p~ at pH 8.1 and 8.6).
The Reductive Half-reaction-When the reductive half-reaction of Tyr-201 -+ Phe was studied under anaerobic conditions, it was found that the rate of reduction changed from approximately 7 X s-' for the substrate-free enzyme to 5.0 s-' for the complex with p-OHB, an acceleration of 7 X lo3 (see Table 11). This compares with the rates for WT: approximately 4 X s-' without p-OHB, and 50 s-' with p-OHB, a rate acceleration of lo5. The Kd determined for NADPH from the Tyr-201-Phe-substrate complex was 580 pM, compared to the WT value of 210 pM. The slower rate of reduction (a factor of 10) is not related to the redox potentials, since they are unchanged from W T (Table I).
Larger perturbations to the reductive half-reaction were observed with Tyr-385 + Phe. The rate of reduction of free enzyme was approximately 8 X s-', and the rate in the presence of p-OHB was 0.5 s-' , 100-fold slower than WT, but still a rate acceleration of 620 due to the binding of substrate. The Kd for NADPH in the ternary complex was 240 p~, almost the same as for WT.
Past studies of p-OH-benzoate hydroxylase show a strong correlation between the rapid reduction of flavin by NADPH and the presence of two transient charge transfer complexes (indicated by weak absorbance between 500-800 nm) on the catalytic pathway (13, 27,43,44). The first charge transfer complex is between oxidized flavin and NADPH; the second is between reduced flavin and NADP+. No charge transfer interactions were detected in the reaction with either of the mutants. It would be expected that the second charge transfer complex would be kinetically invisible due to the slower rate of reduction compared to the (perhaps) unchanged rate of release of NADP+. Conversely, the first charge transfer interaction between oxidized FAD and NADPH should be more prominent, due to favorable rate separations; however, it is not seen. The lack of observable charge transfer complexes may be an indication that these mutants are unable to orient the pyridine nucleotide optimally for reaction with the flavin. Oxidative Half-reaction-This reaction has been described for the WT enzyme in detail (9, lo), and is outlined in Scheme 1. The hydroperoxyflavin and hydroxyflavin intermediates in the reaction of oxygen with the reduced enzyme-p-OHB complex are readily detected in stopped flow experiments. Rate constants for this reaction are presented in Table 11. When the p-OHB complex of NazSz04-reduced Tyr-201-+ Phe was reacted with oxygen, only two consecutive reactions could be detected. The first reaction was an oxygen-dependent second-order reaction leading to the complete formation of a transient chemical species with the characteristics of flavin-C(4a)-hydroperoxide (Fig. 5). The decay of this intermediate to the complex of oxidized enzyme andp-OHB was independent of oxygen concentration and monophasic, with an observed rate constant of 0.76 s-'. Since a small fraction (5-6%) of the NADPH used in turnover results in product formation (see Table I), there must be two competing processes in this reaction phase: oxygen transfer leading to product, and elimination of H20z from the flavin hydroperoxide. The observed rate constant for the disappearance of the hydroperoxide must be the sum of the rate constants for both reaction pathways leaving this branch point; values for k3, 0.04 s-', and ks, 0.72 s-', were calculated. The transfer of oxygen from the hydroperoxide to the substrate leads to the flavin C(4a)-hydroxide, which eliminates water, and product is released. However, this third phase was not detected. If the release of product from the small proportion of enzyme that undergoes actual oxygen transferwas as slow as 1.5 s-', the C(4a)-hydroxyflavin would not be detected.
When the complex between reduced Tyr-385 -+ Phe andp-OHB was reacted with oxygen, two consecutive reactions were observed. The first reaction was an oxygen-dependent second order reaction (see Fig. 6). The spectrum of the transient species formed, calculated assuming two consecutive reactions, has the characteristics of a mixture of flavin-C(4a)hydroperoxide and oxidized flavin. The kinetics a t all wavelengths were the same, indicating that both species were derived simultaneously from the reaction of O2 with the reduced enzyme-substrate complex. The adjusted spectrum of the intermediate (to 100% hydroperoxide) plotted in Fig. 6 was calculated by assuming that the observed spectrum contained 77% of the hydroperoxide and 23% oxidized flavin.  . 6. Spectra of reduced, oxidized, and intermediates of the Tyr-385 + Phe mutant. Conditions were similar to those described in Fig. 5. Enzyme concentration, 13 p~. A transient spectrum was formed by 20 ms (data plotted as circles) and decayed slowly to the oxidized form of the enzyme-substrate complex (upper curue). The spectrum of the C(4a)-hydroperoxide (triangles) was calculated on the assumption that the transient absorbance was a mixture of 77% 4a-hydroperoxide and 23% oxidized enzyme. Inset, observed rate constants for the formation of the transient intermediate as a function of oxygen concentration.
Product analysis shows that 75% of the NADPH consumed by this mutant results in the formation of 3,4-DOHB (see Table I). We conclude that the partial failure of this mutant to form product is a consequence of a change in the initial reaction of reduced flavin with Oz, which now converts only about 75% of the enzyme into transiently stable hydroperoxide. A corollary of this is that once formed, all the hydroperoxide transfers oxygen to make product. It was confirmed that the formation of oxidized flavin in the first phase was not due to free reduced enzyme reacting with O2 ( k = 3 x lo4 M" s" for the latter reaction). This is very similar to W T ( k = 2.8 X lo5 M" s-' with substrate; k = 2.5 X lo4 M-' s-' without substrate).
The WT enzyme exhibits two clearly defined phases after hydroperoxide formation: oxygen transfer (marked by formation of the C(4a)-hydroxide), followed by water elimination and product release to form oxidized enzyme (9,lO). The Tyr-385 + Phe mutant shows only one phase after the formation of the first intermediate. Therefore, the second, oxygen-independent phase in the reaction of Tyr-385 + Phe (with a rate constant of 1.7 s-') must represent primarily one of the above reactions. Simulations show that if the second reaction (water elimination/product release) has a rate constant at least 10-fold greater than the first (oxygen transfer), the two reactions will appear as a single, oxygen-independent phase, primarily representing the oxygen transfer step. Thus, the intermediate observed when oxygen reacts with the reduced Tyr-385 + Phe-substrate complex is the hydroperoxyflavin.
An alternative explanation is that with this mutant, the oxygen-dependent first phase represents the formation of the C(4a)-hydroxide, and the second, oxygen-independent phase represents the elimination/product release step. This would require the rate constant for the oxygen transfer reaction to be much greater than that for the formation of the hydroperoxide. To be consistent with the data, the oxygen transfer reaction would be >300 s-', much faster than that with WT. Also mitigating against this possibility are results obtained in 0.1 M N3-. This anion (among other effects) slows the rate of oxygen transfer to substrate with WT enzyme (a rate change from 47 s-' to 6.5 s-'). With the Tyr-385 + Phe enzyme the oxygen-dependent formation of the first intermediate was essentially unaffected, while the second phase was slowed from 1.7 s" to 0.31 s-', yet no third phase was resolved. Thus the effect of N3-on Tyr-385 + Phe was similar to that on W T (6-7 fold decrease), supporting the conclusion that the observed intermediate is the hydroperoxyflavin, and the oxygen-independent step is due to oxygen transfer. Surprisingly, in the presence of azide, the Oz-dependent first phase results in the formation of 100% of the hydroperoxide.

DISCUSSION
The turnover rate of W T is determined by a combination of several steps ( k l , kB, k4 and the release of NADP+ from reduced enzyme, see Scheme 1). Our results show that the mutations of p-OH-benzoate hydroxylase have two major effects on catalysis. They each slow the reduction of FAD by NADPH and the oxygen transfer from the flavin hydroperoxide top-OHB. For Tyr-385 + Phe, it is these two reactions that have become the rate-determining steps in catalysis and account for the low turnover number (0.36 s-', Table I). The Tyr-385 + Phe mutant has other effects on catalysis as well. For Tyr-201 + Phe, the maximum NADPH oxidation rate (0.42 s-') is accounted for by the shortened pathway (ks in Scheme 1) leading to slow release of HaOa (analogous to a flavoprotein oxidase). These specific changes in the catalytic pathway of the mutants are discussed below.
When p-OHB is present in the active site, Tyr-201 and Tyr-385 are well-removed from the flavin; it is unlikely that they would have much influence on the reactivity of the flavin. Consistent with this, the mutations of the tyrosines to phenylalanines have little effect on the spectral or the redox properties of the flavin. Tyr-201 and Tyr-385 form a hydrogen bond network with p-OHB through the 4-hydroxyl group (see Fig. 1) (6). Thus, it is a reasonable conclusion that the effects of the mutations on the specific rate constants indicate functions for the hydrogen bond network in catalysis.
We propose that the ability to transfer oxygen to the substrate by this enzyme, and possibly by other flavoprotein hydroxylases, is a balance between the lability of the flavin hydroperoxide and its reactivity toward the substrate. The latter is determined by the electrophilicity of the flavin hydroperoxide and the nucleophilicity of the aromatic substrate. Each of these factors can be influenced by pH. The WT enzyme catalyzes the reaction effectively a t high rates over a wide range of pH, but this is readily disrupted by changes in any of the three factors involved. In any specific combination of enzyme and substrate, the half-life of the flavin hydroperoxide should decrease as the pH is raised, due to base-catalyzed elimination. This has already been documented for the analogous flavin hydroxide on the catalytic pathway (12). Therefore, as the pH is raised, the fraction of the hydroperoxide formed that transfers oxygen in the reaction should decrease unless there is a compensating increase in the rate of oxygen transfer.
The decay rate of the flavin hydroperoxide to oxidized flavin and H202 cannot be measured in the WT, due to the small fraction of the enzyme involved. However, upper and lower limits may be estimated. The WT enzyme hydroxylates the substrate with an efficiency >98%; assuming that 2% of the oxygen transfer rate (47 s-') is due to HzOa elimination, an upper limit of 0.9 s-' is obtained. The substrate is required for the stabilization of the flavin hydroperoxide. This intermediate cannot be detected when substrate-free enzyme is reacted with oxygen, unless azide is also present (10). A lower limit to its rate of decay can be estimated by assuming that rapid elimination of H202 occurs upon dissociation of p-OHB from the p-OHB-E-FlHOOH complex and that p-OHB dissociates from the complex at the same rate as from reduced enzyme. The latter has been measured and suggests a lower limit of 0.004 s" (10). Since the Tyr-201 -Phe enzyme catalyzes the formation of 95% H202, it is a useful model for the estimation of the hydroperoxide decay rate. A rate of 0.72 s" was measured for peroxide release (see Table 11), a value within the limits estimated for WT.
Substituted aromatic rings show the following reactivity for electrophilic attack -0-> -NH2 > -OH > -H (45). None of the known flavoprotein hydroxylases are able to attack an unactivated aromatic ring. Thus, the reactivity of the substrate is an important factor in the function of these enzymes. p-OH-benzoate hydroxylase influences this reactivity by modulating the pK, of the 4-hydroxyl of the substrate; the protein shifts the pKa from 9.3 to 7.4 (see Scheme 2). Thus, a substantial fraction of the more reactive phenolate form is present at physiological pH values. This lowering of the substrate pK, is lost in Tyr-201-Phe, with a simultaneous thousandfold decrease in the rate of oxygen transfer. The proposal of a balance between the reactivity of the hydroperoxide/substrate pair and stability of the hydroperoxide can also explain the production of H20z by flavoprotein hydroxylases, even with a substrate in the active site. This hypothesis explains the behavior of p-aminobenzoate with p -OH-benzoate hydroxylase, which is completely hydroxylated at low pH, and is less effectively hydroxylated as the pH is increased (46). p-Aminobenzoate, a nearly perfect analog of p-OHB, cannot ionize on the enzyme to increase its reactivity in compensation for the increased rate of H202 elimination at higher pH. It should be noted that the reaction of the reduced enzyme-p-aminobenzoate complex of Tyr-201 + Phe with oxygen at pH 6.5 leads to formation of intermediates and product completely analogously to wild type.* An example of a flavoprotein that hydroxylates an amineactivated substrate is anthranilate hydroxylase. After hydroxylating anthranilate in the 3-position, it catalyzes the hydrolysis of the imine intermediate to yield 2,3-dihydroxybenzoate (18,47). The amino group in the 2-position of anthranilate is apparently sufficient activation for hydroxylation, analogous t o p-OH-benzoate hydroxylase. The hydrolysis is aided by acid catalysis. Therefore, the enzyme is not likely to be able t o assist the deprotonation of a phenolic group at the 2position, which may be required for the hydroxylation of a phenol. This could explain why salicylate (Z-hydroxybenzoate) is not a substrate for anthranilate hydroxylase (48). Nevertheless, salicylate promotes both rapid reduction of the flavin by NADPH and rapid reaction with oxygen; the latter reaction leads solely to the formation of HzOz and oxidized flavin. In the case of p-OH-benzoate hydroxylase reacting with p-aminobenzoate, hydrolysis of the transient imine is not catalyzed, probably because the hydrogen bond network is set up to stabilize a phenolate.
It should be mentioned that the ionization of the phenol is not a general property of all flavoprotein hydroxylases. Melilotate (16) and phenol (17) hydroxylases bind their substrates as phenols both to the oxidized and to the reduced enzymes. Thus, the activation of substrate by individual hydroxlases may involve a variety of factors.
The Tyr-385 + Phe mutant exhibits an ability to hydroxylate 3,4-DOHB with moderate efficiency. This illustrates a major problem for the biological control of these hydroxylation reactions, the products are more reactive toward hydroxylation than the original substrates. Triphenols are a menace to cells because their spontaneous oxygen reactivity leads to oxidation products that chemically damage proteins and other cellular constituents. p-OH-benzoate hydroxylase is known to bind 3,4-DOHB in a specific orientation (29). The rate of flavin reduction in this complex is lower, no measurable flavin hydroperoxide is formed (lo), and thus no measurable hydroxylation occurs (Fig. 2). How does Tyr-385 + Phe permit hydroxylation of 3,4-DOHB? It must allow 3,4-DOHB to bind in an orientation exposing the 5-position to the flavin e.g. rotated around its C(l)-C(4) axis, so that the 3-OH group can be accommodated due to the removal of the hydroxyl group of Tyr-385. Thus, some or all of the enzyme must form the stabilized hydroperoxide, as it does with p-OHB, to achieve hydroxylation.
The control over the reduction of the flavin is essential to the biological function of this and similar enzymes. Nearly all these enzymes require a ternary complex of phenolic substrate, pyridine nucleotide, and enzyme for efficient reduction of the flavin. Reduction is commonly stimulated by a factor of -lo3 over reduction from the binary complex of enzyme with pyridine nucleotide. With p-OH-benzoate hydroxylase, there is further stimulation of reduction by 100 to severalhundred-fold, depending on the molecule occupying the substrate site. Interestingly, the Tyr-385 + Phe mutant, which no longer has an intact hydrogen bond network, has lost two orders of magnitude of acceleration of the rate of reduction compared to WT. Thus, p-OH-benzoate hydroxylase appears to have a second level of control not present in other flavoprotein hydroxylases, apparently involving the hydrogen bond network. It should be noted that the redox potential of the enzyme flavin does not change significantly on substrate binding, either with wild type or the two mutant enzymes (see Table I).
p-Aminobenzoate stimulates reduction of the wild type enzyme by only about 100-fold? Reduction is thus 2000-fold slower than with p-OHB. Consequently, the enzyme is catalytically slow with p-aminobenzoate, apparently because it does not utilize the H-bond network process. Since p-aminobenzoate is a critical metabolite in folate synthesis, it suggests that a possible biological function for this additional level of control on p-OH-benzoate hydroxylase is the need to prevent the oxidation of p-aminobenzoate in the cell while p-OHB is being used as a source of cellular energy, and p-OH-benzoate hydroxylase is strongly induced.
How can the disruption of the hydrogen bonding network in the mutants result in slower rates of flavin reduction by NADPH (100-fold in the case of Tyr-385 + Phe)? It is known that catalytically sufficient rate enhancements in wild type p -OH-benzoate hydroxylase do not occur unless there is a phenolic group (or an aromatic thiol) capable of ionization at the 4-position of the substrate or effector (43). Thus, there is evidence suggesting that the pKo of the substrate in the hydrogen bond network is important in the control of reduction. However, it is not the only factor involved, since the maximum rate of reduction is constant from pH 5.5 to 9.0 (28). Furthermore, our results show that in the Tyr-201 "-* Phe mutant, where the phenolate anion of the substrate is not preferentially bound, the rate of flavin reduction by NADPH is only a factor of 10 slower than with WT.
It is also known that the control of the reduction rate by the binding of p-OHB is not due to dramatic changes in the Kd for NADPH, since the Kd value hardly changes upon bindingp-OHB to the WT (27). Furthermore, the Tyr-385 + Phe-p-OHB complex shows almost no change from WT in its Kd for NADPH, yet is reduced 100 times slower than WT, while the Kd of the Tyr-201-Phe-p-OHB complex is higher by about a factor of three, and reduction is only slowed by a factor of 10. Apparently, the control is exerted through some ' B. Entsch and D. Ballou, unpublished data.

Tyrosine Mutants of p-H]
subtle change in the orientation of NADPH to the flavin ring. The correct orientation for such hydride transfer is usually signaled by the appearance of charge transfer interactions between the reacting species. These charge transfer interactions are lost in the Tyr-201 and Tyr-385 mutants.
Unfortunately, it has not been possible to obtain a structure of a pyridine nucleotide-p-OH-benzoate hydroxylase complex (49), so no detailed structural information exists. Stereochemical studies have shown that the pyridine ring of NADPH binds on the re side of the flavin (50), which is traversed by protein residues 293 to 300. This important loop probably interacts with both the NADPH and the flavin ring. Its importance may be indicated by the fact that the amino acid sequence of this loop and its extensions (a stretch of 27 amino acids) is entirely preserved in p-OH-benzoate hydroxylase from Acinetobacter calcoaceticas,6 even though the sequence of the corresponding gene is very different (Acinetobacter is -47% G+C while Pseudomonas is 67% G+C). The crystal structure also shows that the backbone carbonyl groups of Pro-293 and Thr-294 are in contact with the hydrogen bond network about the 4-hydroxyl group of the substrate, while the backbone carbonyl group of Tyr-385 is hydrogen bonded to the amine group of Lys-297. Thus, there is scope in the structure for the transmission of small conformational changes from the hydroxyl group of the substrate to the flavin ring and pyridine ring of NADPH, via Pro-293 and Thr-294, or Tyr-201, Tyr-385, and Lys-297.