Proton Nuclear Magnetic Resonance Study of the Electronic and Molecular Structure of the Heme Crevice in Horseradish Peroxidase*

High field proton nuclear magnetic resonance spec- troscopy was used to investigate the electronic and molecular structure of the ferric heme in the resting state of horseradish peroxidase. Deuterium labeling of selected positions of hemin and deuterohemin which were subsequently reconstituted into apo-horseradish peroxidase yielded hyperfine shift patterns for the prosthetic group which are consistent with a ferric porphyrin exhibiting appreciable S = 3/2 character in a quantum mixed spin state. All resolved resonances with significant hyperfine shifts can be accounted for by the porphyrin and a proximal histidyl imidazole, although a sixth ligand from the protein cannot be definitely eliminated. The extremely slow exchange rate with bulk water of the proximal histidyl imidazole exchangeable proton and the absence of deviations from Curie behavior for the porphyrin vinyl and propionic acid proton hyperfine shifts indicate a buried heme crevice which is more rigid than in metmyoglobin. The observation of significant deviations from Curie behavior of the proximal histidyl imidazole exchangeable proton in horseradish peroxidase but not in metmyoglobins is suggested to arise from strong hydrogen bonding between the coordinated imidazole and some unspecified protein acceptor residue in the former protein. Many details of the structure of the heme pocket of the resting state of horseradish peroxidase spite by a wide Unambiguously identified

Many details of the structure of the heme pocket of the resting state of horseradish peroxidase remain unresolved in spite of extensive investigations by a wide variety of physicochemical techniques (1). Unambiguously identified are the protoporphyrin prosthetic group and the trivalent oxidation state of the central iron (1). A coordinated histidyl imidazole appears reasonably well established on the basis of photooxidation (2), pH titration (3), ESR (4), and NMR (5) experiments.
Although the ferric ion is generally considered to be essentially high spin, the magnetic moment of 5.23 p~ for free horseradish peroxidase (6-8) is below that expected for a pure high spin protein ( i e . 5.92 p~) .
Initially, this reduced magnetic moment and the complex ESR spectra were interpreted as resulting from chemical or thermal mixtures of high and low spin forms (6-9). More recently, detailed ESR studies have led to the proposal (IO, 11) that horseradish peroxidase, at least in the neutral pH region, is a chemical mixture of high spin (hs) and a quantum-mixed spin (qrns) species, where the latter is a quantum-mechanical admixture of the S = 5/2 and * This work was supported by Grants HL-16087, HL-22252, and GM-26226 from the National Institutes of Health and by the University of California at Davis NMR Facility. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. S = 3/2 states. A similar qms state has been characterized previously for ferricytochrome c' (12). Changes in ESR spectra and magnetic susceptibility with pH and substrate binding are thus interpreted as resulting from a change in the population of the hs uersus qms states (10,11).
The question of the occupation of the sixth iron site is also incompletely resolved. An early postulate (13) was that horseradish peroxidase possessed a "closed" crevice where both coordination sites were occupied by protein side chain groups, as opposed to an "open" crevice hemoprotein such as metmyoglobin, which has a proximal histidine and an easily displaced water (14). An anionic amino acid side chain as sixth ligand has been proposed (15)(16)(17), based on the invariable uptake by the protein of 1 proton upon binding ligands such as cyanide or fluoride ions, and tyrosine has been suggested, although not all plant peroxidases possess tyrosine (18). More prevalent has been the view (15,16,19) that horseradish peroxidase, like metmyoglobin (15), has 1 coordinated water molecule at the sixth site, although little evidence in support of this hypothesis has been offered.
Several recent spectroscopic studies, however, argue strongly against a coordinated water, since the bulk water protons experience minimal enhanced relaxation (20) and "0 hyperfine splitting was found to be absent in the ESR spectra of H2170-doped samples (21), but these studies provide no information on whether the site is occupied by another ligand. The most recent comparison of the resonance Raman spectrum of horseradish peroxidase (22,23) with those of high spin five-and six-coordinate ferric model compounds strongly favors the five-coordinate structure.
Nuclear magnetic resonance is in many ways particularly well suited for elucidating the detailed structure of the active site of paramagnetic hemoproteins because of the strong coupling between the iron unpaired spins and the nuclei of the appended ligands such as the porphyrin, proximal histidine, and any sixth ligand (24,25). This scalar electron-nuclear coupling (26) gives rise to large hyperfine shifts which permit resolution of many of the resonances for ligated groups in the presence of several thousand unresolved peaks from the polypeptide chain (24,25). Considerable work on model compounds (24,(27)(28)(29) and more thoroughly characterized hemoproteins (30) has demonstrated that the hyperfine shift patterns for nonequivalent protons of the prosthetic group are highly characteristic of a particular oxidation-spin-ligation state of the iron. T o date, however, this information has not been applied systematically to probe the active site of a less well characterized hemoprotein.
Studies of horseradish peroxidase-induced paramagnetic relaxation of hydrogen. donor 1:1 complexes with horseradish peroxidase have indicated that the noncovalently bound substrates are 6 to 10 A from the iron (10, 31,32). Effects reported previously (33) of substrate binding on the presumed heme methyl resonances indicate that primarily one methyl peak is Previous proton NMR studies of horseradish peroxidase include a detailed study which emphasized only narrow lines which arise from the amino acid residues (36). Those studies emphasizing the broad, hyperfine shifted resonances (33, 37, 38) have been hampered by a combination of the inability to detect, resolve, and/or assign the resonances which could provide direct information on the heme pocket stereochemistry, although it has been shown that some resolved resonances are sensitive to pH (37, 38) and substrate-induced (33) structural changes. On the premise that the detailed analysis of the NMR spectra in terms of structure and function of horseradish peroxidase can only be meaningfully initiated after assigning unambiguously as many of the heme and coordinated ligand resonances as possible, we have undertaken a systematic investigation of horseradish peroxidase employing hemin ( Fig.  1A) which has been isotope-labeled a t critical positions (29, 30,39). Less direct but equally effective methods have already been employed (5) to locate and assign the exchangeable proximal histidyl imidazole (Fig. 1B) N I H in both horseradish peroxidase and MetMbHnO, as well as deoxymyoglobin (40).

EXPERIMENTAL PROCEDURES
Protein Preparation-Horseradish peroxidase, type VI, was purchased as a lyophilized, salt-free powder from Sigma; the protein is predominantly isozyme C (41). The protein was purified by column chromatography on carboxymethylcellulose, 5 mM in acetate, pH 4.4, and eluted with 0.05 M NaCl in the equilibrating buffer (41). Peroxidase activity of horseradish peroxidase was measured by following spectrophotometrically at 420 nm the oxidation of pyrogallol to purpurogallin in 0.1 M phosphate buffer, pH 6.0, at 20°C in the presence of HZ02 (42), and found to be 290 units/mg of horseradish peroxidase, where 1 unit forms 1.0 mg of purpurogallin from pyrogallol in 20 S. Analytical polyacrylamide gel electrophoresis was carried out a t room temperature in Tris/glycine buffer, pH 9.3, using 7.5% (w/v) acrylamide. Electrophoresis was performed at 1.5 mA/tube, and the gels were stained with Coomassie B d i a n t Blue. Both the native and ' The abbreviations used are: MetMbH20, Met-aquomyglobin; deutero-HRP, deuterohemin-reconstituted horseradish peroxidase; DSS, 2,2-dimethyl-2-silapentane-5-sulfonate; IPA, indole propionic acid; hs, high spin; qms, quantum mixed spin.

B
deuterohemin-reconstituted horseradish peroxidase showed only one band under these conditions. The proton NMR spectrum of the cyanide complex of horseradish peroxidase was identical to that reported by Williams et al. (36) for the isolated and purified isozyme C. The deuterated hemins and deuterohemin were prepared and characterized as described in detail elsewhere (29,30,39). The position of deuteration is indicated by a bracketed prefix, referring to the standard numbering scheme depicted in Fig. LA. Apo-horseradish peroxidase (43) was prepared by the method of Yonetani (44) and was reconstituted with modified hemins and purified by the method of DiNello (43, 45). Solutions of protein (1 to 3 mM) were prepared in 99.8% 'H20 and the pH (uncorrected, hence referred to as 'pH') was adjusted using 0.2 M 2HC1 or 0.2 M NaO'H. The 'pH' was measured with a Beckman model 3550 pH meter equipped with an Ingold microcombination electrode.
The residual water signal was suppressed by a 25-ms presaturation pulse. Signal-to-noise was improved by exponential apodization which introduced a negligible 20-Hz line broadening; the resolved hyperfine shifted resonances generally have natural linewidths of 300 to 600 Hz in horseradish peroxidase. In order to optimize resolution, the probe temperature was maintained a t 35°C except for variable temperature experiments. Peak positions were initially referenced to the residual water line, which in turn was calibrated against internal 2,2-dimethyl-2-silapentane-5-sulfonate (DSS). Chemical shifts are reported in parts per million a t 35"C, referenced to DSS, with downfield shifts taken as positive.

RESULTS
The portions of the 360 'MHz proton NMR spectra of horseradish peroxidase which contain the resolved hyperfine shifted resonances below 12 and above 0 ppm from DSS are illustrated in Fig. 2 a t several temperatures. The narrow peaks in the 0-to 12-ppm region, arising from amino acid residues, have been treated previously (36). The prominent downfield peaks with areas of three protons are resolved at all temperatures, as found earlier (33, 38). Many of the single proton resonances both upfield and downfield have not been reported previously and are well resolved only a t elevated temperatures. The peaks are designated a to p in the downfield region and a' to k' in the upfield portion, with a subscript to indicate the number of protons contributing to the resonance. Three downfield single proton resonances, n to p (not shown), appear as shoulders on the diamagnetic envelope. The temperature dependence of all resolved resonances is illustrated in Fig. 3 in the form of a Curie plot (26). All peaks move in a manner consistent with the respective proton(s) experiencing some hyperfine shift. The apparent intercepts at T' = 0 are given in parentheses on the left. All assigned resonances except one have intercepts in the diamagnetic region, 0 to 10 ppm from DSS, indicating that the Curie law is obeyed reasonably well. Peak a, the previously assigned exchangeable proximal histi-  The samples in Traces B and C had been in 2H20 for over 2 weeks so that the exchangeable Peak a, was lost.  Fig. 5, A and B, identifying two of the expected four single proton peaks, i' and j'.
The downfield portion of the 35°C proton NMR spectrum of native horseradish peroxidase is compared with that of deutero-HRP in Fig. 6, A and B    Hence, the 2,4-H peaks could not be located in the temperature range 25-55°C.
Peaks b to e have already been shown to be sensitive to pH (38), exhibiting inflection points with a pK -5.9. The single proton peaks, when exhibiting a pH effect, yielded the same inflection point. The 35°C shifts in the high pH region, pH 7.0, and the changes in shifts upon lowering the pH to 5.2, are listed in Table I. Similarly, the effect of adding the protondonor substrate, indole propionic acid (IPA), has been reported earlier (33) for b to e. In Table I, we include the change in shifts for all resolved resonances upon addition of a 30-fold excess of IPA.

DISCUSSION
Heme Resonance Assignments-The assignment of resonances relies wholly on deuteration. However, whether such deuteration results in the absence of resolved resonances in the NMR spectrum depends on the functional group resonating outside the intense diamagnetic envelope 0 to 12 ppm, and on exhibiting a linewidth which permits its resolution. Considerable work with ferric models (24,(27)(28)(29) in different spin states has shown unequivocally that, while the resonance positions may differ in a characteristic manner with spin state, the relative linewidths for all peaks are completely predictable and given by the inverse 6th power of their distance from the iron, as expected for the dominant dipolar relaxation mechanism (47). Comparison of the relative heme peak linewidths in MetMbHZO, where essentially all heme peaks are resolved (30), to those of models (24, [27][28][29] has shown that the relative linewidths in the protein are predictable based on their relative r-6 values, where r is the iron-proton distance. The assignment of three methyls (1, 3, 5) in horseradish peroxidase is direct, although the 8-CH:, assignment must also be considered unambiguous. The four methyls are located in the same region as found in both high and intermediate spin models. Four of the expected six vinyl peaks are located, f, g, i', and j', with two remaining unresolved under the diamagnetic envelope, as also found in MetMbHsO (30). Although individual vinyl group peaks are not yet determinable, we tentatively assign f to 2-vinyl H, and g to 4-vinyl H, based on the fact that 1-CH3 exhibits a much larger hyperfine shift than 3-CH:3, and methyl shifts tend to reflect the asymmetric spin distribution for an individual pyrrole (39). The two vinyl H/l's, i' and j', have comparable widths and hence both are either Ho(cis) or Hll(trans). Their width and the fact that Ho(trans) resonates upfield of H,,(cis) in both models (29) and MetMbHPO (30) suggest that they arise from positions of Hp(trans). Although the meso positions are deuterated 260% (30), no clear difference is detected between the traces of [meso-'H4]HRP and native horseradish peroxidase (Fig. 4, A and D). Since the meso-H peaks are generally -5 times broader than the methyls (29,30), which are already -500 Hz wide, the expected -2.5-KHz linewidths for the four meso-H's probably precludes their resolution even at elevated temperatures.
Since [1,3-(C2H&]HRP also has the propionic acid 6,7-H/,'s 275% deuterated (29,30,39), and no single proton peaks exhibit decreased intensity in Fig. 4B, we conclude that all four 6,7-Hp's are unresolved under the diamagnetic envelope. Thus, only the four propionic acid 6,7-H,'s remain unlabeled. Synthetic routes to the desired deuterium-labeled derivatives are currently under investigation. However, these peaks are generally found in the region 30 to 60 ppm from DSS in all ferric models (27)(28)(29), so that of the six remaining resolved downfield single proton resonances, h to m, four are very likely 6,7-H,,'s. On the other hand, since all other heme resonances are assigned, at least two of these six resonances, h to m, must arise from a coordinated protein residue (see below).
In the case of deutero-HRP, the methyls are again assigned unambiguously in Fig. 6. The different methyl shift order, i.e. 8,5,3,1, reading upfield, from that of horseradish peroxidase, 5,1,8,3, can be attributed to the fact that deutero-HRP has the heme reversed by a 180" rotation about the a-y-meso axis (46) from that in horseradish peroxidase, as unequivocally demonstrated for HRP-CN and deutero-HRP-CN.' The minor peaks designated x in Fig. 6 result from the "native" deuterohemin orientation which is present to an extent of 5 to 10%. With the removal of the vinyl H,'s, only seven single proton peaks, a and h to m, are observed in the region 140 to 15 ppm from DSS, although the peaks are not as well resolved as for horseradish peroxidase. Peak a is exchangeable and hence must also arise from the proximal histidyl imidazole N I H (5). The upfield portion of the NMR traces (Fig. 5, A and C ) differ mainly in the absence of the vinyl H,l's in the latter trace, with no additional peaks observable to -100 ppm. Moreover, the trace of [2,4-(ZH)2]deutero-HRP fails to show any difference from that of deutero-HRP, in spite of the fact that the 2,4-H peaks have linewidths only -30% larger than vinyl H,'s or methyls in all models (29) and MetMbHZO (30). Two possible explanations suggest themselves. One is that the 2,4-H shifts are so large (>200 ppm) that the scalar contribution to the paramagnetic relaxation broadens the peaks beyond detection (47,48); such shifts are unprecedented (27) ' G . N. La Mar, J. S. de Ropp, K. M. Smith, and K. C. Lanpy, manuscript submitted for publication. and are considered extremely unlikely. The much more likely location for the 2,4-H peak is under the diamagnetic envelope, 0 to 10 ppm from DSS. Pyrrole-H shifts ( 2 8 ) precisely in this region are characteristic of ferric model compounds which have been demonstrated (49) to exhibit a ground state consisting of quantum mechanical mixtures of S = 5/2 and S = 3/2.
Protein Resonance Assignments-In a ferric hemoprotein which is expected to exhibit at least dominant high spin character, directly coordinated functional groups invariably exhibit downfield contact shifts for most resonances due to the importance of u spin delocalization (27)(28)(29). At least two of the six peaks, h to m, as well as Peak a, must arise from coordinated amino acid residues. These two protons, most likely a methylene pair, could arise from the known coordinated histidine or a sixth ligand. While a sixth nitrogenous ligand would yield low spin iron(III), it has been shown in mutant hemoglobins that the presence of a sixth phenoxide or carboxyl ligand maintains essentially high spin character (50)(51)(52). The only methylene protons of aspartic acid would be much too broad to detect due to their proximity to the iron (48). Models have shown (53) that phenoxide ligands yield readily resolvable m-H peaks at 100 ppm, whose absence in the horseradish peroxidase traces argues strongly against tyrosine as the sixth ligand. Only glutamic acid as sixth ligand has a pair of methylene protons likely to yield resolvable lines with significant hyperfine shifts.
We have already shown (5) that Peak a must arise from the exchangeable N I H of the proximal histidyl imidazole, as also found in MetMbHnO, which is observable in 'Hz0 solution for horseradish peroxidase several days after dissolution of the protein due to the extremely slow exchange rate with bulk 'H20. The 2,4 protons of the known coordinated imidazole (Fig. 1B) would be too broad (>3 KHz) to detect (54). However, the 5-CH2 or P-CH2 should yield two single proton resonances with linewidths very similar to those of the propionic acid H,,'s due to their nearly identical distance from the iron (54). We conclude that the two additional downfield peaks of h to m could arise from a coordinated glutamic acid, but more probably arise from the histidyl imidazole P-CHz. Thus, NMR data fail to provide any direct evidence for a sixth ligand.
The remaining resolved hyperfine shifted resonances appear upfield and all but Peak h exhibit weak temperature dependences and intercepts at T I = 0 sufficiently close to DSS to suggest significant ring-current contributions to the diamagnetic shifts in many cases. These shifts can arise from the small magnetic anisotropy (26) which causes upfield shifts at axial positions, as observed in high spin or nearly high spin ferric complexes (28,55), and are likely to arise from distal residues which are close to the iron but not coordinated (30). Similar peaks are observed (30) in MetMbHnO, where all heme resonances are located and where a sixth ligand from the protein is known to be absent. More definitive assignments must await a single crystal x-ray structural analysis.
Electronic Structure of the Zron-Although the four heme methyls exhibit shifts very similar to those of high spin MetMbHsO (30) and other high spin ferric models (27)(28)(29)40,55), the pyrrole-H shifts differ dramatically, appearing 60 to 80 ppm downfield in an essentially high spin ferric system, but most likely resonating in the region 0 to 10 ppm downfield from DSS in horseradish peroxidase. Comparison of purely high spin models (27,53,55) with the recently characterized models (28) which are quantum mechanical mixtures of S = 5/2 and S = 3 / 2 states (12, 49) shows that, while the methyl shifts are similar, the pyrrole-H shifts are characteristically much further upfield due to the depopulation ofd,+z in the S = 3/2 component. Thus, the heme hyperfine shift patterns in deutero-HRP are consistent with an important contribution from a quantum mixed spin state of S = 5/2 and S = 3/2 to the solution electronic structure of both horseradish peroxidase and deutero-HRP, as was also proposed based on low temperature solid state ESR data (10,11). This difference in spin states invalidates any further comparison of the hyperfine shifts with those of MetMbH20 (30) aimed a t elucidating the coordination geometry. However, the very nature of the quantum mixed spin state argues against the presence of a sixth ligand of any kind, since only in five-coordinated models with relatively weak axial ligands has it been possible to detect this mixed spin state (28,49).
Since all assigned heme resonances follow the Curie law reasonably well (Fig. 2) and exhibit intercepts within 3 to 5 ppm of their expected diamagnetic positions, it is not likely that the relative populations of the hs and qms states differ significantly in the temperature range 25-55°C. Although the methyl shifts are similar in the high spin (29,53,55) and quantum mixed spin states (28), it is very unlikely that all shifts are the same in both states.
This observation of near Curie behavior of the heme resonances takes on added interest when one observes that the only peak which exhibits a substantial deviation from the Curie law is Peak a, the histidyl imidazole NIH. Moreover, although this functional group has a similar shift in MetMbHsO at 25°C (5), it exhibits strict Curie behavior in that protein, as also shown in Fig. 2. This strong non-Curie behavior unique to the NIH signal indicates that either the iron-imidazole or NI-H bonding changes with temperature.
We favor changes in the NI-H bonding, inasmuch as it is known that hydrogen bonding between NIH and a peptide acceptor residue occurs to some extent in all characterized hemoproteins, and has been suggested as an important control mechanism of heme reactivity (56). Although resolution of Peak a precludes complete deprotonation, strong proton donation to some unspecified acceptor would render the axial ligand more capable of stabilizing both the ferric state of the resting enzyme and the iron(1V) states of Compounds I and 11. Since proton donation would decrease the NIH hyperfine shift, the observed deviation from Curie behavior implies stronger proton donation at lower temperatures.
Heme Side Chain Mobility-The hyperfine coupling constants of protons on functional groups such as vinyl and propionic acid side chains are known to depend on the rotational position of the group relative to the heme plane (29,30,57,58). In proteins with only modest constraints on the precise orientation of these groups (14), their oscillatory mobility is manifested clearly in large deviation from Curie law behavior due to the change in average orientation with temperature. Thus, in MetMbHsO (30) and the monomeric insect Met-HbCN's (58) as well as in most models (27,29,39), the H,,'s of both the vinyl groups and propionic acid chains deviate considerably from the predicted T I behavior, which is adhered to by the heme methyl shifts.
In the present case, all heme resonances, including the assigned vinyl H,'s and the probable propionic acid H,'s, obey the Curie law to the same degree as do the methyl shifts, indicating that the orientations of all of these side chains are "frozen" and held immobile in the accessible temperature range by the protein-heme interactions. Thus, we conclude that not only is the heme pocket relatively buried, but that it is also much more rigid or inflexible than in MetMbH20.
Substrate Bindzng-Earlier proposals as to the mode of substrate binding have involved hydrogen bonding to either the proximal histidyl imidazole N,H (32) or one of the propionic acids (33). The latter proposal was based on the pref-erential IPA-induced shift of a methyl thought to be 5-CH, or 8-CH3. Our identification of the imidazole NIH signal (15) and its insensitivity to IPA binding (Table I) eliminates the fist proposal. The unambiguous assignment of the methyl peak most sensitive to IPA binding to 1-CHa (Table I) appears to argue against binding near a propionic acid, in view of our earlier findings (34) that the intercalation of mercuric triiodide into the heme pocket of MetMbHsO resulted in differential heme methyl shift changes proportional to their proximity to the intercalating site (35). However, while the differential methyl shifts with IPA binding support a site near pyrrole I (Fig. lA), several of the Peaks j to m are also sensitive to IPA binding, a t least some of which should arise from propionic acid H,,'s. Thus, the participation of the propionic acid chains in substrate binding cannot be eliminated. Planned '"C labeling of the propionic acid carboxyl groups may shed more light on this matter.