Mutants That Alter the Covalent Structure of Catalase Hydroperoxidase II from Escherichia coli*

The three-dimensional structures of two HPII variants, V169C and H392Q, have been determined at resolutions of 1.8 and 2.1 Å, respectively. The V169C variant contains a new type of covalent bond between the sulfur atom of Cys169 and a carbon atom on the imidazole ring of the essential His128. This variant enzyme has only residual catalytic activity and contains heme b. The chain of water molecules visible in the main channel may reflect the organization of the hydrogen peroxide substrates in the active enzyme. Two alternative mechanisms, involving either compound I or free radical intermediates, are presented to explain the formation of the Cys-His covalent bond. The H392Q and H392E variants exhibit 75 and 25% of native catalytic activity, respectively. The Gln392 variant contains only heme b, whereas the Glu392 variant contains a mixture of heme b andcis and trans isomers of heme d, suggesting of a role for this residue in heme conversion. Replacement of either Gln419 and Ser414, both of which interact with the heme, affected the cis:trans ratio of spirolactone heme d. Implications for the heme oxidation mechanism and the His-Tyr bond formation in HPII are considered.

Under certain conditions, compound I can be reduced by a one-electron addition resulting in the formation of compound II (a formal Fe 4ϩ state) which can lead to inactivation of the enzyme. Small subunit catalases (12), utilize NADPH to prevent the accumulation of compound II while the large subunit enzymes, HPII and PVC, do not form compound II, which may explain why they do not require NADPH.
The structure of native HPII, solved and refined at 1.9-Å resolution contains a cis-hydroxychlorin ␥-spirolactone heme d and a novel type of covalent bond joining the C ␤ of the essential Tyr 415 and the N ⑀ of His 392 (4,13). As part of an ongoing study of catalases, HPII mutant variants with changes in Val 169 , Asp 197 , His 392 , His 395 , Ser 414 , and Gln 419 were produced. The HPII variant V169C, designed to investigate the effect of changes in the major channel leading to the active site, was found to contain an unusual covalent bond between the essential histidine (His 128 ) and the substituted residue (Cys 169 ). Variants in other positions were produced to study the mechanism of heme oxidation and Tyr-His bond formation in HPII. In particular, a variant of His 392 was designed to change the residue to its Gln (14) counterpart in PVC which has heme d but not the His-Tyr bond. We report here the enzymatic characterization of 12 mutant variants of HPII and the structure determination of two of them, V169C and H392Q, refined at 1.8-and 2.1-Å resolution, respectively. Implications of the biochemical and structural peculiarities found for these variant enzymes are discussed.

EXPERIMENTAL PROCEDURES
Materials-Standard chemicals and biochemicals were obtained from Sigma. Restriction nucleases, polynucleotide kinase, DNA ligase, and the Klenow fragment of DNA polymerase were obtained from Life Technologies, Inc.
Oligonucleotide-directed Mutagenesis-Oligonucleotides were synthesized on a PCR-Mate synthesizer from Applied Biosystems and are listed in Table I. The restriction nuclease fragments, that were mutagenized following the Kunkel procedure (18), sequenced and subsequently reincorporated into pAMkatE72 to generate the mutagenized katE genes, are also listed. Sequence confirmation of all sequences was by the Sanger method (20) on double-stranded plasmid DNA generated in JM109. Subsequent expression and purification were carried out as described previously (21).
Catalase Assay and Protein Determination-Catalase activity was determined by the method of Rørth and Jensen (22) in a Gilson oxygraph equipped with a Clark electrode. One unit of catalase is defined as the amount that decomposes 1 mol of H 2 O 2 in 1 min in a 60 mM H 2 O 2 solution at pH 7.0 at 37°C. Protein was estimated according to the methods outlined by Layne (23).
Matrix-assisted Laser Desorption/Ionization Mass Spectrometry (MALDI-MS)-250 g of protein was boiled for 3 min in 50 mM potassium phosphate, pH 7.0, and mixed with 7 g of trypsin at room temperature for 1 h. The solution was boiled a second time and treated with an additional 7 g of trypsin for 1 h after which the solution was boiled and lyophilized. The dried sample was dissolved in 2 ml of saturated ␣-cyano-4-hydroxycinnamic acid in acetonitrile, 0.1% trifluoroacetic acid (1:2 by volume). 1 l was applied to the sample probe and air dried for MALDI-MS using the Manitoba reflecting time-of-flight mass spectrometer (24). The instrument was operated in positive-ion linear mode. Desorption/ionization was achieved by a pulsed ultraviolet laser beam (N 2 laser, ϭ 337 nm). The acceleration voltage was 20 kV and the laser power density was approximately 10 6 W/cm 2 . The mass spectrum of each sample was the average of more than 100 shots. For mass calibration, bovine insulin (5733 Da) was used.
Crystallization, Data Collection, and Structure Determination-Crystals of HPII variants V169C and H392Q were obtained in the vicinity of the conditions previously described for the wild type enzyme HPII (3). A diffraction data set for the V169C variant was collected at station X11 in DESY. The crystal was frozen in liquid nitrogen with 30% MPD added to the crystallization buffer. Diffraction data were processed with DENZO and scaled with SCALEPACK (25) (Table II). For the H392Q variant, data were collected using a rotating anode x-ray source and graphite monochromator at 100 K in the same cryobuffer. Data were processed and scaled using MOSFLM (26) and SCALA (27,28), respectively (Table II). The refined model of native HPII at 1.9-Å resolution (4) was used as the starting model for both structures, and this model was improved by automatic refinement alternated with manual rebuilding using programs XPLOR (29), REFMAC (30), and O (31), respectively. The R free criteria was used during all the refinement but no reflections were omitted in the map Fourier synthesis calculations. A distance restraint and the removal of the non-bonded interactions between atoms participating in the new His-Cys bond were used during the final steps of the V169C refinement. Restraints due to the non-crystallographic symmetry were only taken into account in the initial steps of refinement to 2.5-Å resolution. Bulk solvent correction was included and the whole resolution was used in the refinement. The corresponding final crystallographic agreement factors, R and R free , together with other refinement parameters are summarized in Table II. Solvent accessibility and cavities were calculated with VOIDOO (32). For polar atoms the atomic radii used were smaller than the van der Waals value (33). This allows us to take into account the fact that when two atoms form a hydrogen bond, the observed interatomic distance is smaller than the sum of the atomic radii. Atomic coordinates and structure factors have been deposited with Protein Data Bank with entry codes 1qf7 and 1cf9 for the H392Q and V169C variants of HPII, respectively.

Major Channel Variants of HPII-
The side chain of Val 169 is situated immediately above the essential residue His 128 where the major access channel enters the distal heme pocket (Fig.  1A). This valine is fully preserved among all known catalase   . The corresponding electron densities are shown for the three cases. In the V169C variant structure, a chain of well defined solvent molecules reach the distal pocket as compared with the native and H392Q structures where solvent molecules are sequences and it is placed in the narrowest and most hydrophobic section of the channel. Changing the equivalent residue in catalase A from S. cerevisiae to a smaller residue (V111A) resulted in an enzyme with reduced catalatic activity but enhanced peroxidatic activity for larger substrates (34). This was attributed to a disruption in the flow of H 2 O 2 in the enlarged channel and to easier access for the larger peroxidatic substrates. The rationale behind changing Val 169 3 Cys in HPII was that subsequent selective reaction of the sulfhydryl group might allow progressive blockage of the access channel. This objective was precluded by the finding that the purified V169C mutant had only about 0.1% of the native HPII activity, and contained heme b as determined by HPLC and spectral analyses (Table III). In order to define if these dramatic and unexpected changes were a result of steric factors or another property specific to the cysteine residue, the V169A and V169S mutant variants were constructed. These two variant enzymes contained heme d and exhibited about 25% of wild type activity, similar to the change in activities of the equivalent Val mutants in catalase A from S. cerevisiae and significantly higher than the activity of the V169C variant. No o-dianisidine peroxidatic activity was detected. From this, it was concluded that the Cys residue was a determining factor in the reduced catalatic activity and in the lack of conversion of heme b to heme d. How a residue located outside the heme pocket could have such important catalytic effects remained puzzling and structure determination of the V169C variant was considered appropriate.
Crystal Structure of the V169C Variant of HPII-The threedimensional structure of the V169C variant of HPII was determined and refined using synchrotron data at 1.8-Å resolution. The starting model was the structure of native HPII at 1.9-Å resolution (4), from which the substituted residue and neighboring solvent molecules were omitted. The final refined V169C structure, with crystallographic agreement factors of r ϭ 18.1% and R free ϭ 23.7%, provided a clean electron density map which showed a number of significant differences with respect to wild type HPII (Fig. 1, A and B). All of these differences were clustered in the vicinity of either the mutated residue or the heme group.
The most striking feature evident in the final electron density maps of the V169C variant is the presence of a covalent bond between the sulfur atom of Cys 169 and the imidazole ring of His 128 (Fig. 1B and Fig. 2). The shape of the electron density and the refinement behavior strongly suggest that, within the limitations of the resolution available, the sulfur atom from Cys 169 together with all the atoms from the imidazole ring of His 128 are in a common plane. Perfect planarity implies that the atoms in the imidazole ring retain the sp 2 hybridization state which imposes strong constraints on the possible chemical mechanisms that can produce the covalent bond. The modified imidazole ring is rotated about 30 o relative to its position in wild type HPII and its orientation can be unambiguously defined by the proximity of a water molecule within 3.1 Å of the imidazole N ⑀ atom and by the bifurcated hydrogen bond formed by N ␦ with both O ␥ from Ser 167 and main chain carbonyl oxygen from Thr 168 (Fig. 2). This leaves the C ⑀ of the imidazole ring situated just 1.84 Å from the cysteine sulfur atom, a distance consistent with a covalent bond. The rigidity introduced by the new Cys-His covalent bond that would likely raise the pK a of the imidazole ring together with the altered imidazole-heme stacking, should both contribute to the diminished catalytic activity of the V169C variant.
The moderate increase in the heme channel volume that results from the Val to Cys replacement can only explain in part the presence of four extra solvent molecules in the variant enzyme (Figs. 1, B and D, and 2). These solvent molecules form a continuous hydrogen bonded chain that extends the fulllength of the channel from the molecular surface to the distal pocket, ending with the water molecule above the heme iron. The presence of this continuous chain of solvent molecules reaching the active center proves that the low activity of this mutant is not a result of steric hindrance to substrate access. In native HPII, the solvent chain is interrupted in the vicinity of Val-169.
The V169C variant crystal structure confirmed the presence of heme b, rather than heme d as had already been suggested by HPLC analysis, and also confirmed the mass spectroscopic analyses that showed the absence of the Tyr 415 -His 392 covalent bond found in wild type HPII. The absence of the two covalent modifications in the V169C variant is similar to the situation found in the structure of the inactive or weakly active variants, H128A, H128N, and N201H ( Fig. 3) (12). 2 In these structures, Gln 419 is rotated forming a hydrogen bond with the heme propionate side chain and with a water molecule located near Thr 416 . Furthermore, the main chain from residues 414 to 417 is shifted with respect to the native model, and Ile 126 , on the heme distal side, exists in two conformations likely due to the restrictions imposed by its proximity to the propionate chain (Fig. 3).
Mass Analysis of Tryptic Peptides-An independent confirmation of the presence of the His 128 -Cys 169 covalent bond was provided by a mass analysis, using MALDI-MS, of trypsin digests of wild type HPII and of the V169C variant. Complete digestion of HPII by trypsin should generate a mixture of 75 peptide products, which will be further complicated, particu-  larly below 3000 Da, by the presence of partial digest products. His 128 and Val 169 (or Cys 169 ) are situated on separate tryptic peptides (Fig. 4B). The His 128 -containing peptide extends from Ile 126 to Arg 130 and has a predicted mass of 596 Da. The Val 169 -containing peptide extends from Phe 166 to Arg 180 , and replacement of Val with Cys would increase the predicted mass from 1453 to 1457 Da. Covalent linkage of the His 128 -and Cys 169 -containing peptides through the imidazole-sulfur bond would produce a 2051-Da peptide, assuming that 2 hydrogens were lost as part of the covalent bond formation. The mass spectrum of a tryptic digest of wild type HPII (Fig. 4A) confirms the absence of significant peptide peaks in the 2000 to 2100 Da range. By comparison, the mass spectrum of a tryptic digest of the V169C variant (Fig. 4B) shows a prominent peak in this range with a mass of 2051, consistent with the presence of the His 128 -Cys 169 covalent linkage.
Heme Proximal Side Variants of HPII-The conversion of heme b to heme d and of the formation of the His 392 -Tyr 415 covalent bond in HPII have been linked in a concerted mechanism that requires the formation of compound I. The proposed mechanism was supported by the observation that neither modification is found in the catalytically inactive variants of HPII, H128A, H128N, and N201H (13). 2 In PVC, the His-Tyr bond cannot be formed because the residue that would correspond to the histidine is a glutamine. However, PVC does form heme d proving that heme oxidation and His-Tyr bond formation can be unlinked and suggesting the possibility that the two reactions may occur separately in HPII. To test this possibility, the HPII mutant variant, H392Q, was constructed to make HPII reflect the structure of PVC, and the H392E, H392A, and H392D variants were constructed for comparison. All of these variants retained activity ranging from 20 to 75% of wild type activity (Table III). Spectroscopic and HPLC analysis revealed that three of the variants, including the Gln, Asp, and Ala replacements, contained only heme b consistent with a coupling of His-Tyr bond formation and heme modification. Despite the activity of the variants, the characteristic heme b spectra were indistinguishable from the spectra for the inactive mutants, H128A, H128N, and N201H previously reported (21). Surprisingly, the H392E variant was found to have a mixture of heme b and heme d (Fig. 5) indicating that heme conversion could occur, albeit at a slower rate. However, the picture is complicated by the presence of large amounts of the trans isomer of the spirolactone in the H392E variant as compared with the predominantly cis isomer in PVC and HPII, and by the presence of an unidentified heme peak that eluted between the peaks of heme d and b.
Crystal Structure of H392Q-Determination of the crystal structure of the H392Q variant of HPII was undertaken because it differed significantly from wild type HPII in lacking both the His-Tyr bond and heme d despite retaining near wild type activity. The structure, determined and refined following a similar approach to the one described for the V169C variant, gave crystallographic agreement factors r ϭ 14.4% and R free ϭ 21.0% for data at 2.1-Å resolution collected with a conventional rotating anode x-ray source and a graphite monochromator. A number of structural changes with respect to native HPII, including a few rearrangements in the solvent structure and the presence of the heme b group were evident in the electron density (Fig. 1C). Changes directly related to the presence of heme b, such as the double conformation of Ile 126 on the heme distal side and the rotation of Gln 419 on the heme proximal side, are consistent with changes already described for the V169C structure (Fig. 3). Additional changes around the mutated residue, Gln 392 , included the hydrogen bonding of Gln 419 with both the propionate side chain of the heme group and a molecule of water that is also hydrogen bonded with the amide nitrogen atom of Gln 392 (Fig. 3). This water, not present in the native HPII structure, is found in PVC and in the HPII variants, V169C and H392Q. Gln 419 has the same disposition in the structures of all the HPII variants containing heme b regardless of whether residue 392 is Gln or His (Fig. 3).
A water molecule in PVC and in the H392Q variant of HPII occupies the space that corresponds to the imidazole group in the native HPII structure. This water is in direct contact with the C ␤ atom of the essential tyrosine (Tyr 350 in PVC and Tyr 415 in HPII) suggesting a strong polarization of this region. Otherwise, the solvent organization on the heme proximal side in the H392Q variant is very similar to that of active HPII. The solvent structure on the distal side of the heme is similar to wild type HPII with no solvent molecules being present in the final part of the major channel which indicates that neither the FIG. 2. Stereo view (rotated by about 180 o with respect to Fig. 1) of the environment of the covalent bond formed between the S atom of the mutated residue (Cys 169 ) and the carbon atom C ⑀ from the imidazole ring of the essential histidine (His 128 ). Also the 2F o Ϫ F c electron density map is shown, calculated after simulated annealing with His 128 , Cys 169 , and the five waters filling the channel excluded. The electron density suggests that the bonded sulfur and all the atoms in the imidazole ring are situated in a common plane (see the text) with a dihedral angle (Cys ␤169 -Ser 169 -Cys ⑀128 -Asn ␦128 ) of 76 o . The relative orientation of the imidazole ring, defined by both the presence of a water molecule hydrogen bonded to the N ⑀ and the bifurcated hydrogen bond formed between the N ␦ with O ␥ from Ser 167 and O ␥ from Thr 168 departs about 30 o from the orientation found in the active enzyme. The geometry of this Cys-His bond imposes strong restraints on the possible mechanisms of formation. FIG. 3. Stereo views of the heme proximal side in native HPII (A), PVC (B), the variant V169C (C), and the variant H392Q (D). The native HPII structure contains a modified heme d and also presents a covalent bond between the C ␤ of Tyr 415 and N ␦ of His 392 . The structure of the HPII variant H392Q contains an unmodified heme b cofactor but retains a high catalatic activity. The structure of PVC also contains a modified heme d group, although the configuration in the vicinity of the essential tyrosine is very close to the one determined for the HPII variant H392Q. presence of heme b nor the double conformation of Ile 126 affect the solvent organization in the major channel. The presence of only one solvent molecule in the pocket immediately above the heme ( Fig. 3C; corresponding to W0 in native HPII) rather than two found in native HPII could indicate that accessibility to the iron is somewhat diminished in the H392Q structure. Whether this is due to Ile 126 , to the heme b, or to the water being a weaker ligand in the heme b environment is not yet clear.
Involvement of Other Proximal Side Residues-The concerted mechanism that couples heme oxidation to His-Tyr bond formation implicated His 395 and Asp 197 as playing a role in the reaction (13). The H395A, H395Q, D197A, and D197S variants of HPII were constructed to test this hypothesis. These four mutants all retained near wild type levels of activity and contained both heme d and the His-Tyr bond (Table III). The only difference from wild type was the apparently slightly slower heme conversion rate in the case of the H395A variant indicating that His 395 and Asp 197 do not have a significant role in heme conversion or His-Tyr bond formation.
Both Ser 414 and Gln 419 are located in close proximity to the heme. Ser 414 is hydrogen bonded with the hydroxyl group on ring III of the heme d, and its interaction with the incoming hydroxyl group might potentially influence the oxidation reaction. Gln 419 is hydrogen bonded with the carbonyl group of the spirolactone on ring III in HPII and with the propionate side chain in the HPII variants that contain heme b. It seems likely that the association continues throughout the oxidation and cyclization reaction providing the Gln residue with an opportunity to influence the heme modification pathway. The mutant variants S414A, Q419A, and Q419H were constructed to test these hypotheses. The three variants retained 40% or more of wild type activity (Table III) indicating that the residues were not critical to the catalytic function of the enzyme. Spectral and HPLC analyses revealed a distribution of hemes quite different from that of wild type HPII (Fig. 5). For the two Gln 419 mutant variants, there was a larger than normal peak of the trans isomer of heme d (ϳ30% of the total as compared with 10% in the native HPII) indicating that changing the Gln residue affected the direction of lactone cyclization. Even more dramatic was the elution pattern of heme from the S414A variant which contained more trans than cis isomer and some heme b indicating that the Ser 414 residue has important cata- lytic roles both in directing the path of reaction and in facilitating the reaction. DISCUSSION The report of unusual covalent modifications in proteins is becoming increasingly common as accurate information about large structures is becoming available. Examples in redox related systems include the methionine sulfone found in the active site of catalase from P. mirabilis (2), the cysteine-sulfenic acid in the NADH peroxidase from Streptococcus faecalis (35), the oxidized tryptophan residue in lignin peroxidase (36), the modified cysteine in catalase HPII (37), the internal cyclization of the peptide backbone in the green fluorescent protein (38), and the His-Tyr covalent bond in catalase HPII (13). The bond between the C ⑀ of the imidazole ring of His 128 and the sulfur of Cys 169 in the V169C mutant variant of HPII is yet another example of an unexpected and unusual covalent bond located in the core of a protein. The presence of this linkage, with essentially full occupancy, is well supported by both the electron density of the crystal structure and the sizes of fragments in the tryptic digest maps. The strong peptide peak in the mass spectrum indicates that the bond is relatively stable and does not break down readily under trypsin digest or MALDI-MS conditions.
Two possible mechanisms to explain the formation of the imidazole-sulfur bond can be proposed involving either an acidcatalyzed, nucleophile-mediated reaction (39) or a free radical mediated reaction. The protonation of His 128 may facilitate the nucleophilic attack of the thiol group of Cys 169 on the C ⑀ of the imidazole ring (Fig. 6A). This would yield an intermediate with the C ⑀ converted to sp 3 hybridization which could return to sp 2 hybridization through release of a hydride ion to the oxyferryl group of compound I. The same mechanism could be revised slightly such that the protonation of His 128 at the initial stage of compound I formation triggers the nucleophilic attack by the thiol group (Fig. 6B).
Free radical addition of thiols to double bonds can be initiated by peroxides (40) providing several pathways for a free radical mechanism. The initial oxidation of the thiol group of Cys 169 by H 2 O 2 would yield a reactive thiyl radical which could then attack the C ⑀ of His 128 as shown in Fig. 6C. The participation of compound I as the oxidant of the thiol seems less likely given the 7.9 Å distance between the Cys-SH and the heme iron.
The absence of heme modification and of the His-Tyr covalent bond in the V169C mutant variant, both of which require compound I formation, suggest that compound I is not formed in this HPII variant, apparently invalidating mechanisms such as those in Fig. 6, A and B. The solution to this dilemma lies in the likelihood that the formation of the His-Cys bond is a more rapid reaction than either the heme modification or the His-Tyr bond formation. Reaction with the sulfur may occur in the first few catalytic rounds before any other modification occurs resulting in rapid modification of the essential His 128 imidazole ring. Once His 128 is modified, compound I cannot be formed and further modification cannot take place. A second explanation involves the replacement of compound I as primary oxidant by hydrogen peroxide as a hydride acceptor in the nucleophilic mechanism or the free radical mechanism (Fig. 6C).
The concerted mechanism that combines heme oxidation with His-Tyr bond formation in HPII seems to be corroborated FIG. 6. Possible mechanisms for the formation of the covalent bond found in the V169C variant of HPII between the mutated residue Cys 169 and the essential histidine His 128 . Scheme A shows the proposed nucleophilic mechanism based on the acid catalyzed nucleophilic addition of the thiol group of Cys 169 onto the imidazole ring of His 128 . Scheme B shows an alternate mechanism where protonation of His 128 , in the initial step of compound I formation, triggers the nucleophilic attack of the thiol group of Cys 169 on the imidazole ring. Scheme C depicts the proposed free radical mechanism based on the oxidation of the thiol group of Cys 169 to yield a thiyl radical which can attack the imidazolic system. by the properties of the H392Q variant of HPII. This variant was surprising in retaining a high percentage of native activity despite the presence of heme b and the absence of the His-Tyr bond which demonstrated that neither heme d nor the His-Tyr bond is essential for the activity of HPII. The H392E variant which contains some heme d but no His-Tyr bond appears to contradict the assumption of a linkage between the heme modification and His-Tyr bond formation and confirms that an alternate mechanism is possible in HPII. However, the slower rate of heme conversion and the abnormal heme profile as compared with native HPII suggest that the alternate mechanism may be different from that normally operating in the native enzyme. By the same criteria, the mechanism in the H392E variant would appear to be different from that operating in PVC.
Both Gln 419 and Ser 414 have roles in controlling the heme oxidation reaction as is evident in the fact that changing either residue causes changes in the cis:trans ratio of the spirolactone. The most dramatic change results from removal of Ser 414 , consistent with it being hydrogen bonded with the OH on ring III. Interaction with the Ser 414 side chain must stabilize the hydroxyl group on the proximal side of the heme, presumably limiting isomerization to the more stable trans isomer and ensuring formation of the cis isomer. The S414A variant contains the His-Tyr bond even when retaining some of the unmodified heme b group. All the inactive HPII variants analyzed lacked the two covalent modifications (13), which suggest that compound I formation is required for these modifications to take place.