A Tryptophan Neutral Radical in the Oxidized State of Versatile Peroxidase from Pleurotus eryngii

Versatile peroxidases are heme enzymes that combine catalytic properties of lignin peroxidases and manganese peroxidases, being able to oxidize Mn2+ as well as phenolic and non-phenolic aromatic compounds in the absence of mediators. The catalytic process (initiated by hydrogen peroxide) is the same as in classical peroxidases, with the involvement of 2 oxidizing equivalents and the formation of the so-called Compound I. This latter state contains an oxoferryl center and an organic cation radical that can be located on either the porphyrin ring or a protein residue. In this study, a radical intermediate in the reaction of versatile peroxidase from the ligninolytic fungus Pleurotus eryngii with H2O2 has been characterized by multifrequency (9.4 and 94 GHz) EPR and assigned to a tryptophan residue. Comparison of experimental data and density functional theory theoretical results strongly suggests the assignment to a tryptophan neutral radical, excluding the assignment to a tryptophan cation radical or a histidine radical. Based on the experimentally determined side chain orientation and comparison with a high resolution crystal structure, the tryptophan neutral radical can be assigned to Trp164 as the site involved in long-range electron transfer for aromatic substrate oxidation.

exhibits manganese-independent activity toward veratryl alcohol and p-dimethoxybenzene. Furthermore, it oxidizes hydroquinones and substituted phenols that are not efficiently oxidized by LiP or MnP in the absence of veratryl alcohol and Mn 2ϩ , respectively. VP is even able to degrade directly high redox potential dyes, which can be eventually oxidized by LiP only in the presence of veratryl alcohol (9,10).
Two genes encoding VP isoenzymes VPL and VPS1, expressed in liquid-and solid-state fermentation cultures, respectively, have been cloned from Pleurotus eryngii (11,12). The deduced amino acid sequences for both isoenzymes were used to build molecular models by homology modeling, taking advantage of sequence identity to P. chrysosporium LiP and MnP and Coprinopsis cinerea (synonym Coprinus cinereus) peroxidase (13). Very recently, the crystal structure of recombinant P. eryngii VP expressed in Escherichia coli and activated in vitro (14) has been determined at 1.33-Å resolution (Protein Data Bank code 2BOQ).
Catalytically, VP would follow the classical heme peroxidase cycle, in which hydrogen peroxide is the final electron acceptor, acting as a 2-electron oxidizing substrate for the resting enzyme, which results in the formation of Compound I (15)(16)(17). Compound I is reduced back by the substrate in a two-step reaction that involves the formation of Compound II (a 1-electron oxidized form) and then the closure of the catalytic cycle to the resting state. It is known that H 2 O 2 attains the distal side of the heme through the access channel present in all peroxidases. However, identification of sites involved in substrate oxidation has been successful only on a few occasions (18 -20).
In ligninolytic peroxidases, only the manganese interaction site in MnP, situated near the internal propionate of the heme, has been confirmed by site-directed mutagenesis and x-ray diffraction of MnP⅐Mn 2ϩ complexes (21)(22)(23). A putative manganese-binding site similar to that of MnP has been identified in P. eryngii VP isoenzymes (12,24). Substitution of one of the residues from the putative manganese interaction site in VPL yields an enzyme with its Mn 2ϩ oxidation ability strongly impaired (25). Such a binding site is not present in LiP according to its lack of activity with this divalent cation.
In the case of LiP, there is no x-ray structure with any of its substrates available, and structure-function studies have been based on techniques that led to the indirect characterization of the substrate interaction sites. The veratryl alcohol molecule was initially modeled at the heme access channel (26), and the residues hypothetically involved are conserved also in the isoenzyme VPS1 of P. eryngii, which has a main heme access channel remarkably similar to that of LiP (24). Apart from the possibility of direct electron transfer to the heme from some substrates, long-range electron transfer (LRET) should be considered to justify the capacity of VP and LiP to oxidize high molecular size substrates that do not fit into the heme access channel.
The first suggested LRET pathway involved the LiP distal histidine and started at His 82 (27). This pathway does not exist in MnP and in the isoenzyme VPL of P. eryngii, but it has been identified in VPS1. Two additional pathways have been proposed in LiP. One starts at the second exposed histidine (LiP His 239 ) to the proximal histidine and is absent in MnP and VPS1 and present in VPL (His 232 ). The second pathway starts from an exposed tryptophan (Trp 171 ) present in LiP and proceeds to the porphyrin ring. Its involvement in veratryl alcohol and non-phenolic lignin model compound oxidation has been confirmed in LiP, and Trp 171 has been revealed to be a redox-active residue (28 -32). Multisequence alignment revealed that the above-mentioned tryptophan residue is conserved in all LiP sequences as well as in Pleurotus VP and is absent in the typical MnP (15). The involvement of a surface tryptophan as a catalytic site in LiP was also inferred in a study on the S168W variant of MnP (33).
Moreover, a putative electron transfer pathway from the exposed tryptophan residue was identified in P. eryngii isoenzymes VPL (Trp 164 ) and VPS1 (Trp 170 ) (12,13,24). It is interesting to point out that VPs, as all other ligninolytic enzymes described until now, have no Tyr residues in their sequences (15).
In this work, we report the results from multifrequency (9.4 GHz, X-band; and 94 GHz, W-band) EPR studies on a freeze-quenched radical intermediate in the reaction of P. eryngii VP with H 2 O 2 . Comparison of the EPR and pulse electron nuclear double resonance (ENDOR) experimental data with theoretical results from density functional theory (DFT), in particular for the hyperfine (hf) tensor values, excludes assignment to a tryptophan cation radical or a histidine radical, but strongly suggests assignment to a tryptophan neutral radical. It has been shown in earlier studies on tryptophan radicals in ribonucleotide reductase variants that evaluation of the hf-tensors of the side chain methylene protons enables determination of the side chain geometry, which can be used for site-specific assignment (34,35). Based on the side chain orientation deduced from the experimental hyperfine data of these protons, the observed tryptophan neutral radical can be assigned to Trp 164 , which is proposed as the surface site involved in LRET in VPL of P. eryngii.

MATERIALS AND METHODS
Chemicals and Protein Preparation-Potassium hydrogen phthalate and H 2 O 2 were purchased from Sigma and used without further purification. Recombinant VP (VP*) was obtained by E. coli expression (14), and the cDNA encoding the mature sequence of P. eryngii allelic variant VPL2 (GenBank TM accession number AF007222) (12) was cloned in the expression vector pFLAG1 and used to transform E. coli W3110. Transformed cells were grown for 3 h in Terrific Broth (36), induced with 1 mM isopropyl ␤-D-thiogalactopyranoside, and grown for an additional 4 h. The apoenzyme accumulated in inclusion bodies was recovered using 8 M urea. Subsequent in vitro folding was performed using 0.15 M urea, 5 mM Ca 2ϩ , 20 M hemin, oxidized glutathione/dithiothreitol (5:1), and 0.1 mg/ml protein at pH 9.5 (14). Active enzyme was purified by Mono Q chromatography using a 0 -0.5 M NaCl gradient in 10 mM sodium tartrate (pH 5.5) supplemented with 1 mM CaCl 2 .
UV-visible and EPR Measurements-UV-visible measurements of 6 M VP* in 0.1 M phthalate buffer (pH 4.5) were performed on a Hewlett-Packard 8453 spectrophotometer at 24°C before and after the addition of 8 eq of H 2 O 2 . EPR solutions were prepared with a final concentration of 0.16 mM enzyme and 1.3 mM H 2 O 2 in the same buffer, and the reac-tion was stopped by rapid immersion of the EPR tube in liquid nitrogen after 10 s. Continuous wave X-band (9.4 GHz) EPR measurements were carried out with a Bruker eleXsys E500 series using the Bruker ER4122 SHQE cavity and an Oxford ESR900 helium continuous flow cryostat. X-band pulse ENDOR experiments were performed on a Bruker E580 pulse EPR/ENDOR spectrometer using a Bruker EN4118X-MD4-W1 resonator equipped with a helium immersion cryostat. Spin quantification was performed by double integration of the experimental EPR radical signal compared with the iron signal.
EPR saturation data were collected by measuring the EPR absorption derivative signal intensity as a function of microwave power at different temperatures. The saturation data were fit to Equation 1, where S is the EPR derivative signal intensity, P is the microwave power, P1 ⁄ 2 is the half-saturation power, and b is the inhomogeneity parameter. A nonlinear least-squares fit to Equation 1 yielded P1 ⁄ 2 and b values for each particular temperature (37)(38)(39).
94 GHz high-field EPR experiments were performed on a Bruker eleXsys E680 spectrometer. Spectra were recorded at 40 K. Low microwave power, typically only a few microwatts, was used to avoid saturation effects. For determination of precise g-tensor components, the microwave frequency was measured using a frequency counter, which was integrated in the spectrometer. The magnetic field was calibrated with a g-tensor standard (lithium in LiF, g ϭ 2.002293 (2)) (40) at two different frequencies (typically, 93.8 and 94.2 GHz). All spectra were recorded in the "persistent mode" of the superconducting magnet, using the room temperature coils for the field sweep to ensure high linearity and stability of the field (maximum sweep width of 800 G). The modulation amplitude was kept at 0.2 millitesla (mT) to avoid modulation broadening. The 94 GHz EPR microwave resonator is a fundamental mode cylindrical cavity that has a high filling factor (compared with a transmission line set up or a Fabry-Perot resonator) and a small sample volume (0.7-mm diameter and ϳ1.5-mm height), thereby assuring high sensitivity and high homogeneity of the field over the active sample volume (Ͼ5 ppm). This enables the measurement of very accurate g-tensor values with an instrumental limit for the absolute g-tensor resolution obtained for the narrow signal of the g-tensor standard of Ϯ0.000005 (lithium in LiF, approximate line width of 0.01 mT).
Simulation of the EPR Spectra-The EPR spectra were analyzed using a software for simulating and fitting EPR spectra for S ϭ 1/2 systems with anisotropic g-and hf-tensors (34). Thereby, the spectra are simulated by computing the resonant field position correct to second order at the given microwave frequency, dependent on the orientation of the g-and hf-tensors with respect to the external magnetic field (41,42). No restriction for the relative orientation of the principal axes of the different tensors is applied (34).
DFT Calculations-DFT calculations were performed with the program package Gaussian 03 (43) at the unrestricted Kohn-Sham level using the hybrid exchange-correlation functional B3LYP. Geometry optimizations were done using a contracted basis set augmented with polarization functions (6 -31G**) (43). Magnetic properties (g-tensor, spin densities, and hyperfine couplings) were calculated on the geometry-optimized structure in a second step using the TZVP basis set, which is included in the program. The resulting output was fed into the program MOLDEN for visualization of the spin density distribution in the molecule (44). Fig. 1 shows the electronic absorption spectrum of P. eryngii VP* (expressed in E. coli and folded in vitro) at pH 4.5 before and after H 2 O 2 addition. The spectrum is characterized by the presence of the Soret band at 409 nm and the two charge transfer bands at 505 and 638 nm. The 600 -650-nm wavelength region is sensitive to the heme pocket environment (45). This band (638 nm), observed only in the high spin state, has been assigned to a charge transfer transition from the porphyrin to the iron. In the inset of Fig. 1, the EPR spectrum of the resting state enzyme confirms the presence of the dominant contribution of the high spin iron species (g Ќ ϭ 6.00 and g ʈ ϭ 2.00).

UV-visible Spectroscopy-
In the UV-visible spectrum ( Fig. 1) recorded 10 s after the addition of 8 eq of H 2 O 2 , the Soret absorption band is broadened, and its intensity is reduced with respect to the native enzyme. The visible region shows a broad absorption with peaks at 583 and 652 nm, suggesting the formation of Compound I (17,46).
The reduced absorption of the Soret region of Compound I compared with native VP* and the characteristic absorption band at 652 nm suggest that this compound contains 2 oxidizing eq over the native enzyme (47). The first oxidizing equivalent is contained in the ferryl state of the iron. The second equivalent can be on the porphyrin ring, as for horseradish peroxidase, or on an amino acid side chain of the protein, as for cytochrome c peroxidase (16,17). Through EPR measurements, it is possible to distinguish the oxidizing equivalent location and to assign the nature of the organic radical formed. Fig. 2, the EPR signal obtained after the addition of 8 eq of H 2 O 2 and rapid cooling is shown. The strong decrease in the EPR signal of the ferric high spin species and the formation of an intense radical signal corresponding to a protein radical species are evident. The yield of the radical signal is ϳ0.25 spin/heme.

GHz EPR Spectroscopy-In
The isotropic EPR spectrum shown in Fig. 2A is characterized by a doublet, centered at g ϭ 2.0027. It seems that the reaction of the heme porphyrin ring and H 2 O 2 proceeds to the formation of an amino acid radical, considering the absence of the characteristic signal of the ferryl porphyrin intermediate at 10 s of freezing time (48,49). In Fig. 2B, the intensity of the radical signal against temperature is plotted. The curve shows a strong temperature dependence especially at low temperature, as reported for cytochrome c (50). ) can be calculated for all different temperatures. At 20 K, the b value obtained from the fitting is equal to 0.9. This is typical for protein radicals with a weak coupling to a paramagnetic ion. The P1 ⁄ 2 value is in good agreement with those   . VP* was measured after a 10-s reaction with 8 eq of H 2 O 2 . The region between 450 and 700 nm has been expanded 3-fold. Inset, X-band EPR spectra of resting P. eryngii VP* (Fe(III) spin state before the addition of H 2 O 2 ). The spectrum was recorded under the following conditions: temperature, 20 K; ϭ 9.385 GHz; modulation amplitude, 1 mT; microwave power, 2 mW; and modulation frequency, 100 kHz.
reported for similar organic radicals that coupled to the ferryl heme iron in other peroxidase systems (51,52).
An expansion of the well resolved 9.4 GHz (X-band) EPR spectrum of the radical is shown in Fig. 4. The spectrum is dominated by a large doublet splitting, showing well resolved subsplittings on each of the two components. The overall shape of the spectrum is very symmetric, which is expected in the case of very small g-tensor anisotropy. The upper trace shows a best simulation and fit of the spectrum based on one nitrogen and two anisotropic ring proton hf-tensors, typical for a tryptophan radical (34,52,53), and two fairly isotropic hf-tensors from the ␤-protons of the side chain. The simulation shows that the 9.5 GHz EPR spectrum is fully consistent with the proton and nitrogen hf-tensor values, typical for a tryptophan neutral radical (Scheme 1). The g-tensor values obtained from the 94 GHz EPR spectrum were used in the simulation shown in Fig. 4 (see below and Table 1). At 9.5 GHz, the EPR simulation is very sensitive for the hf-tensor values, but less sensitive for the g-tensor values in the case of such small g-tensor anisotropy. The obtained hf-tensor values are given in Table 1.
High-field EPR Spectroscopy (94 GHz Frequency)-To unequivocally identify the nature of the observed radical, we performed high-field/ high frequency EPR experiments on samples of P. eryngii VP*, prepared in the same way as those used for recording the 9.4 GHz EPR spectra. Although the hyperfine structure observed in conventional X-band EPR can be rather similar for different types of organic radicals, it has been demonstrated that the principal values (g xx,yy,zz ) of the g-tensor, which are resolved for such radicals only in high-field EPR, can serve as a molecular fingerprint for identification of the type of radical (48,49,(51)(52)(53). In particular, tyrosine and tryptophan radicals are difficult to distinguish in X-band (9.4 GHz) EPR, whereas at 94 GHz EPR, they are clearly distinguished based on the much larger g xx and g yy values observed for tyrosyl radicals exhibiting large spin density on the oxygen, which has a large spin orbit coupling constant, compared with tryptophan radicals with spin density only on carbons and nitrogen, both being nuclei with only small spin orbit coupling constants (51)(52)(53). For a tryptophan radical, the field corresponding to 94 GHz EPR (3.3 tesla) does still not represent the so-called high-field limit, where the g-tensor shifts exceed the hyperfine splitting and all three g-tensor components are separated in the spectrum. This situation was nearly met at 285 GHz EPR in a study of a tryptophan radical in azurin (54). However, the hf-tensor resolution in the spectrum was lost at 285 GHz. Nevertheless, at 94 GHz EPR, the g-tensor anisotropy contributes already significantly to the shape and overall width of the spectrum, and the g-tensor values can be obtained from EPR simulations, provided partially resolved hyperfine splitting is observed at 94 GHz, and all hf-tensor values are known from 9.5 GHz EPR experiments (53). Fig. 5 shows the 94 GHz high-field EPR spectrum from the radical obtained at 40 K. The spectrum is again dominated by a large doublet splitting. However, different from the X-band EPR spectrum, it shows a clear asymmetry; the low-field side shows a sharper rise and is more structured, whereas the high-field side has a more extending wing with weak intensity. This asymmetry in the spectrum is indicative of the presence of small g-tensor anisotropy, which nevertheless significantly affects the 94 GHz EPR spectrum.
The upper trace in Fig. 5 shows a best simulation and fit of the spectrum, from which the g-and hf-tensor values were obtained. 94 GHz EPR simulations were started using the proton and nitrogen hf-tensor values from the X-band EPR simulations. The simulated 94 GHz EPR spectrum is particularly sensitive to the g-tensor values. The spectral positions of the g xx , g yy , and g zz components shown in Fig. 5 indicate the magnitude of the g-tensor anisotropy, which significantly determines the overall width, shape, and splitting of the simulated 94 GHz spectrum. The X-band EPR spectrum of the radical was then simulated using the g-tensor values from the 94 GHz EPR simulations, in this way refining the hf-tensor values.
The final simulations shown in Figs. 4 and 5 were performed for the 9.4 and 94 GHz EPR spectra using the same g-and refined hf-tensor values (Table 1), which fitted both spectra equally well. We want to emphasize that only this multifrequency approach ensures obtaining a unique set of g-and hf-tensor values for the radical. Table 1 shows that the g-and hf-tensor values for the radical in P. eryngii VP* agree very well with values recently obtained for a tryptophan neutral radical in FIGURE 4. High resolution narrow scan 9. 5 GHz EPR spectrum of the radical reported in Fig. 2 together with the best simulation (see "Results"). EPR spectra were recorded under the following conditions: microwave power, 0.5 mW; modulation amplitude, 0.1 mT; and temperature, 40 K. Parameters for the EPR simulation (Sim.) are given in Table 1.
Exp., experimental. SCHEME 1. Molecular structures of the tryptophan (upper) and histidine (lower) neutral radicals. The inner numbers denote the molecular positions, which were chosen for the histidine radical so as to have the same numbering for comparable positions, except for positions 5 (His) and 9 (Trp). The outer numbers are -spin densities derived from the experimental hf-tensors (see "Discussion").
Bjerkandera adusta VP (52), as well as with an earlier study on the Y122F variant of the ribonucleotide reductase from E. coli (34,53).
9.5 GHz ENDOR Spectroscopy-Smaller couplings, e.g. from positions of small or negative spin densities in the indole ring or from a proton hydrogen-bonded to the nitrogen in a tryptophan neutral radical, are not resolved in the EPR spectra. They are resolved, however, by ENDOR spectroscopy. This technique gives basically the NMR spectrum of the respective radical detected via the electron (55). ENDOR lines from protons (hydrogens) occur as pairs in the spectra according to the first-order resonance condition (Equation 2), where H represents the Larmor frequency of the proton (I ϭ 1/2, H ϭ 14.9 MHz at a magnetic field of 345 mT) and A j (j ϭ xx, yy, zz) is the electron-nuclear hyperfine coupling. In frozen solution, the hyperfine coupling is anisotropic, defined by three tensor principal components, A xx , A yy , and A zz (55). According to Equation 2, each proton gives rise to a group of three pairs of spectral features (one pair for each component A xx,yy,zz ) that are spaced symmetrically about H or A j /2, whichever value is larger. Fig. 6 shows a Davies pulse ENDOR spectrum (56) from the radical in P. eryngii VP*. Because of the low radical yield for the freeze-quenched reaction of VP* with H 2 O 2 and unfavorable spin relaxation rates, probably caused by the nearby Fe 4ϩ ϭO heme (S ϭ 1), the ENDOR spectrum shows noise, even after ϳ7 h of accumulation. Therefore, we concentrated on the range of smaller hyperfine couplings, which give larger ENDOR signal intensities and are not resolved in the EPR spectra. Several pairs of lines are clearly detected in Fig. 6 and are spaced symmetrically about the free proton Larmour frequency ( H ϭ 14.9 MHz). The broad line groups at 7-10 and 20 -24 MHz result from the two larger tensor components of protons 5 and 7 of the indole ring (Scheme 1). The lower trace in Fig. 6 shows a simulation of the ENDOR spectrum in the experimental frequency range using the hf-tensor components determined by EPR ( Table 1). The extra intensity on the high frequency side (20 -24 MHz) results from superposition of the low frequency lines (Equation 2) from the large ␤-proton hf-tensor values (H-␤1). The low frequency lines from the smaller ␤-proton hf-tensor values (H-␤2) are expected to be close to zero frequency, where intensities decrease strongly beyond detection. The respective high frequency lines of H-␤1 and H-␤2 are expected at frequencies Ն30 MHz, beyond the spectral range analyzed. Both hf-tensors from these protons are well resolved, however, in the EPR spectra and were analyzed.
Smaller hyperfine couplings (not resolved in the EPR spectra) give rise to the ENDOR line pairs between 11 and 18 MHz. Lines are observed corresponding to hf-tensor components of 0.28, 0.19, and ϳ0.11 mT and a smaller component between 0.07 and 0.04 mT with larger intensity, probably indicating contributions from two protons (0.1 mT obtained from the EPR spectra corresponds to 2.8 MHz a The g-tensor principal values (g xx,yy,zz ) are given with a maximum error of Ϯ0.0001. The g z axis is expected to be out of the tryptophan plane (34,53,62); the g x axis is in the tryptophan plane and forms an angle of ϳ20°with the line connecting the center of the C-6 -C-7 bond with C-2 (Scheme 1). b hf-tensor principal components are given in mT (estimated error of Ϯ0.05 mT). The g-and hf-tensor values fit X-and W-band EPR spectra equally well (Figs. 4 and 5). The remaining deviations are small and could be a result of weak coupling to the Fe 4ϩ ion (S ϭ 1 ground state) of the heme. g-tensor and ring proton hf-tensor values (in mT) are similar to those obtained recently from the tryptophan neutral radical in B. adusta (52) and previously reported for the Y122F variant of E. coli RNR (34, 53) (see center rows). The differences in the hf-tensor values of the side chain protons H-␤1 and H-␤2 do not reflect different spin densities at C-3 (Scheme 1), but rather result from different dihedral angles for different side chain orientations (see "Discussion"). c Negative signs are for ring proton hf-tensor values expected from theory in the case of positive carbon spin density; for the simulations, the x axes of H-5 and H-7 hf-tensors were rotated 30°with respect to the g-tensor x axis (estimated error of Ϯ0.05 mT). d The axis of A zz (N-1) is parallel to the g z axis, as expected for a nitrogen in a planar -system (34,53). e The hf-tensor was assigned to the proton of a hydrogen bond donor to the indole nitrogen of the tryptophan neutral radical (Fig. 7A). f Small hf-tensor components are for H-2 (Scheme 1) and for the proton hydrogen-bonded to N-1 from the ENDOR spectra and their simulations (Fig. 6). Principal axes are not related to g-tensor axes. g ND, not determined. h Theoretical g-and hf-tensor principal values are from DFT for the tryptophan neutral radical, hydrogen-bonded at N-1 to a water molecule (Fig. 7) using Gaussian 3.0 (B3LYP, TZVP basis set) (43). A comparison of calculated g-and hf-tensor values for the tryptophan neutral and cation radicals using, in addition, the EPRII basis set, yielded the same values within ϳ5% for the larger tensors. Side chain orientation 85°out of plane corresponds to dihedral angles of ϳ55°and ϳ65°for H-␤1 and H-␤2 (orientation for Trp 244 ) (Fig. 9). Side chain orientation 21°out of plane corresponds to dihedral angles of Ϫ9°for H-␤1 and 52°for H-␤2 (crystal structure for Trp 164 ) (Figs. 7 and 9 and Table 2). The experimental values are in agreement only with an assignment to Trp 164 (see "Discussion").
obtained from the ENDOR spectra). The best ENDOR simulation ( Fig.  6) was obtained assuming two proton hf-tensors with the principal values given in Table 1 (last two columns). Comparison with the results from the DFT calculations (see below) shows that these couplings can be assigned to the hydrogen at C-2 (Scheme 1) and to a proton from a hydrogen bond to N-1 in the case of a tryptophan neutral radical (Fig. 7, Table 1, columns 7 and 8, and Table 2, row 2). The presence of these two small anisotropic proton hf-tensors strongly supports an assignment to a tryptophan neutral radical, hydrogen-bonded at N-1. In the tryptophan cation radical, a hydrogen forms a real covalent bond to N-1. In this case, an additional large hf-tensor, larger as for position 5, was calculated (data not shown), which is in disagreement with the experimental value (see "Discussion"). DFT Calculations-To corroborate the assignment of experimental hf-and g-tensor components and to further investigate the question of the presence of a tryptophan neutral or cation radical (both forms were observed in DNA photolyase) (57), we performed density functional calculations of spin densities and hf-and g-tensors (43,44). Earlier DFT studies on tryptophan radicals were reported with qualitatively similar results for the hf-tensors (34,58). However, we performed our own DFT calculations with the program package Gaussian 03 (43) at the restricted Kohn-Sham level using the hybrid exchange-correlation functional B3LYP and a TZVP basis set, in particular for the correct side chain geometry obtained from the x-ray structure, to obtain the g-tensor and taking into account hydrogen bonding by introducing water molecules. Inspection of the structure of tryptophans in various proteins shows a large variety of side chain geometries. It seems obvious that the actual side chain orientation is largely imposed by constraints of the protein and not by a small energy minimum of the isolated tryptophan molecule. After performing a general geometry optimization, we performed calculations for the tryptophan neutral radical with two fixed side chain orientations: (i) 85°out of plane, which corresponds approximately to Trp 244 in the crystal structure of P. eryngii VP, and (ii) only 21°out of plane, which corresponds to the side chain orientation of Trp 164 in VP. The best agreement with experimental hf-tensor values was obtained for the tryptophan neutral radical, deprotonated and hydrogen-bonded at N-1. For the above two side chain orientations of this radical, all calculated larger hf-tensor and g-tensor components are listed in Table  1 and compared with the experimental values. Table 2 compares experimental -spin densities derived from the measured hf-tensors with theoretical Mulliken spin densities (43,44), calculated by DFT for a tryptophan cation and neutral radical. The experimental -spin density for N-1, ( 14 N-1), was estimated from the anisotropic 14 N 1 hf-tensor with elements AЈ ii ϭ A ii Ϫ A iso (ii ϭ xx, yy, zz), where A iso ϭ (A xx ϩ A yy ϩ A zz )/3. Assuming axial symmetry (AЈ xx ϭ AЈ yy ), the magnitude of this tensor is determined by the uniaxial constant B (59), where AЈ zz ϭ 2B ( 14 N-1) and AЈ xx ϭ AЈ yy ϭ ϪB ( 14 N-1). Using the 14 N 1 hf-tensor values from Table 1 and the value of B ϭ 47.8 MHz (1.706 mT) given for ( 14 N-1) ϭ 1 (59), an experimental -spin density of ( 14 N-1) ϭ 0.20 was obtained (34,53). For the ␣-protons at ring positions 5 and 7, -spin densities were estimated from the hf-tensor elements of the respective protons (Table 1) using the relation A iso (H ␣ ) ϭ Q (C-5/7), with an established Q value of Ϫ2.48 mT (60). The -spin density at C-3 was obtained from the hf-tensors of the two side chain ␤-protons H-␤1 and H-␤2 as discussed below. Calculated Mulliken spin densities are given in Table 2 for the tryptophan neutral radical, deprotonated and hydrogen-bonded at N-1 to a water molecule for the two side chain conformations (85°and 21°out of plane); for the tryptophan cation radical, with a N-H bond at N-1; and for the respective neutral and cation radicals of histidine. A visualization of the calculated Mulliken spin densities is given in Figs. 7 and 8.

DISCUSSION
The VP Radical Is a Tryptophan Neutral Radical-Identification and assignment of radicals immobilized in proteins are not straightforward, and simulations of conventional 9.4 GHz EPR spectra alone may lead to ambiguities, viz. that more than one set of g-and hf-tensor data may fit the observed EPR spectrum equally well. However, the multifrequency FIGURE 5. 94 GHz EPR spectrum of the protein radical together with the best simulation. EPR spectra were recorded under the following conditions: microwave power, 0.25 microwatts; modulation amplitude, 0.2 mT; and temperature, 40 K. The positions for the three g-tensor principal values are shown to indicate the significant contribution of g-tensor anisotropy in the spectrum. The shape of the 94 GHz EPR spectrum is clearly asymmetric, and its total width is increased compared with that of the spectrum at 9.5 GHz due to the larger shifts of the g-tensor components at the correspondingly higher field (3.3 versus 0.34 teslas). The g-and hf-tensor components used for the simulation (Sim.) are the same as in Fig. 4 (Table 1), except for a somewhat larger intrinsic line width (0.4 versus 0.35 mT for the 9.5 GHz EPR spectrum). Exp., experimental. approach used here avoids this problem. Simulations of the 9.4 GHz EPR spectra yielded an hf-tensor data set that was used for simulations of the 94 GHz high-field EPR spectra, from which accurate g-tensors were obtained. This data set was further refined by simulations of both spectra with these g-tensor values, yielding finally a common hf-and g-tensor set (given in Table 1), which fits the EPR spectra at both frequencies equally well.
The obtained g-tensor values show very low anisotropy. Radicals with spin density on oxygen (like tyrosyls) have significantly larger g-tensor anisotropy, with typical values for tyrosyls of g xx ϭ 2.0092 to 2.0068 (depending on the environment polarity and hydrogen bonding), g yy ϳ 2.0044, and g zz ϳ 2.0022 (48,51,53,61,62). The g-tensor values obtained here for the radical in P. eryngii VP* are in perfect agreement with a tryptophan neutral radical, as reported for the Y122F variant of E. coli ribonucleotide reductase (34,53,62) and recently detected in VP from B. adusta (52). Theoretical g-tensor values (Table 1) were calculated with DFT for the tryptophan neutral radical, hydrogen-bonded at N-1 to a water molecule (Fig. 7), for two different side chain orientations. Considering a reported general ϳ10% overestimation by DFT methods (63,64) of the g-tensor shifts (⌬g jj ) from the free electron value g e , (⌬g jj ϭ g jj Ϫ g e (jj ϭ xx, yy, zz), g e ϭ 2.002319), the calculated g-tensor values are in good agreement with the experimental values. Similar g-tensor values were, however, also calculated for the tryptophan cation radical (Table 2). Therefore, discrimination between a tryptophan neutral or cation radical is not possible based on the g-tensor values alone.
The common hf-tensor data set obtained from the 9.4 and 94 GHz EPR spectra shows the presence of two large rather isotropic proton hf-tensors typical for ␤-protons of the side chain, only two large anisotropic hf-tensors from two ring protons, and one large anisotropic hftensor from an S ϭ 1 ( 14 N) hf-tensor (Table 1). From these data, experimental spin densities for C-3, for two other carbon ring positions, and for N-1 were deduced (Table 2, first column). Comparison with the calculated Mulliken spin densities for tryptophan and histidine neutral and cation radicals ( Table 2) shows that such a spin density distribution is calculated only for the tryptophan neutral radical. According to the calculations, ϳ60% of the spin density is at C-3, followed by 23 and 18% at C-5 and C-7, giving rise to the observed ring proton hf-tensors; N-1 carries 26% spin density. This distribution is very similar to that  MOLDEN (44). The molecular structure was energy-minimized, including the water molecule hydrogen-bonded at N-1. A, tryptophan neutral radical, hydrogen-bonded to a water molecule at N-1 (for numbering, see Scheme 1), with a side chain orientation 21°out of plane, corresponding to the geometry of Trp 164 in the crystal structure (see Fig. 9). B, tryptophan cation radical, with an N-1-H covalent bond. Blue, positive spin densities; red, negative spin densities. Note that the cation radical has reduced spin densities at C-3 (where the side chain is attached) and at N-1 and additional large spin density between these two positions at C-2. The larger spin densities for both molecules are given in Table 2. The experimental results agree only for the neutral radical (A; see "Results").

Trp cation radical b,c
His neutral radical (in plane) b,d a -Spin densities ( ) at C-3, C-5, and C-7 were estimated from the experimental hf-tensor values (see "Results"). Isotropic hyperfine coupling values, A iso ϭ (A xx ϩ A yy ϩ A zz )/3, from both H-␤1 and H-␤2 were used for estimating (C-3) (see "Results"). b DFT calculations were made with the program Gaussian 03 (43) using the B3LYP functional and the TZVP basis set for final calculation of the spin densities and g-and hf-tensors of the respective radical (see "Results"). Given are the Mulliken atomic spin densities ( ) for all molecular positions where Ͼ ͉0.06͉. The orientation of the side chain with respect to the tryptophan head group was changed and found to have no major influence on the spin densities within the indole ring system. However, the hf-tensor values of the side chain ␤-protons strongly depend on the side chain orientation (see Eq. 3 and "Discussion"). The 21°out-of-plane orientation of the side chain corresponds to the geometry of Trp 164 in the crystal structure of P. eryngii VP and is close to the orientation deduced from the experimental values of the ␤-proton hf-tensors. The 85°o ut-of-plane orientation corresponds to Trp 244 and does not agree with the experiments (see Fig. 9 and "Discussion"). The tryptophan neutral radical is deprotonated at N-1. We assume the presence of a hydrogen bond at N-1, which is modeled in the calculation by a water molecule (see "Discussion"). c This is the tryptophan cation radical, protonated at N-1 (Fig. 7B). d This is the histidine neutral radical, deprotonated and hydrogen-bonded at both nitrogens. Side chain geometry is as shown in Fig. 8. Note the different numbering schemes for tryptophan and histidine. C-5 is not comparable in both radicals (Scheme 1). The cation radical is protonated at one nitrogen. Side chain geometry is as shown in Fig. 8. The experimental data do not agree with the spin densities calculated for a histidine neutral or cation radical. In particular, the large spin densities at C-2 (0.33) and C-5 (0.47) lead to very large and anisotropic hf-tensors from the ring protons (between 1.5 and 2.0 mT), which is not in agreement with the experimental values (see "Discussion"). e NA, not applicable. Position 4 in tryptophan is a non-magnetic 12 C atom (Scheme 1).
obtained previously for the tryptophan radical in the Y122F variant of E. coli ribonucleotide reductase (34). The calculations also predict small anisotropic hyperfine couplings for the proton at position 2 and for the water proton, hydrogen-bonded to N-1, in good agreement with the small proton hyperfine splitting observed in the pulse ENDOR spectra ( Fig. 6 and Table 1). The tryptophan neutral radical presumably evolves from a short-lived cation radical that is generated from the proposed Fe 4ϩ ϭO porphyrin cation radical precursor by LRET, leaving the heme iron in the Fe 4ϩ state, but promoting the cation radical state to the tryptophan. The tryptophan cation radical then, in a second step, becomes stabilized by deprotonation at N-1. This proton could be taken in by one of the oxygens of the nearby residue Glu 243 and then become hydrogen-bonded to the remaining lone electron pair of N-1 at the tryptophan. In any case, the second small anisotropic proton hf-tensor, observed by ENDOR, strongly supports a neutral radical hydrogenbonded at N-1. The hf-tensor of this hydrogen-bonded proton is mostly of dipolar nature and is expected to be very similar for a hydrogenbonded water proton or a proton from an amino acid residue as the hydrogen bond donor. For simplicity, the calculations were performed with water as the hydrogen bond donor, whereby the distance and geometry were obtained by energy minimization.
The calculations for the tryptophan cation radical predict significantly smaller spin densities for C-3 and N-1 and are not in agreement with the experimental results. In addition, a large positive spin density is predicted for C-2 and a large hf-tensor is expected for the NH proton ( Fig. 7 and Table 2). The proton bound to the nitrogen and the proton at position 2 would both lead to additional large ring proton hf-tensors, giving rise to visible additional splitting in the EPR spectrum, which is clearly in contradiction with the experimental EPR spectra. In particular, the simulated 9.4 GHz EPR spectra were very sensitive to the multiplicity of the proton hyperfine couplings. Furthermore, the hf-tensor from a proton hydrogen-bonded to N-1, observed by ENDOR, is expected only in the case of the neutral radical. Hence, the obtained experimental set of g-and hf-tensor data and the deduced -spin densities are in perfect agreement with a tryptophan neutral radical and do not support a tryptophan cation radical.
Because the possible involvement of a histidine radical has been discussed in some mechanistic studies (13,65), we also performed calculations for the histidine neutral radical, deprotonated and hydrogenbonded at both nitrogens, and for the cation radical, protonated at N-4 ( Fig. 8 and Table 2). For the neutral radical, large spin densities were calculated for C-2 and in particular for C-5. Significantly smaller spin density compared with the tryptophan neutral radical was calculated for C-3; and surprisingly, small and even negative spin densities were calculated for the two nitrogens. The calculated hf-tensor value for ring proton 5 (data not shown) was at least a factor of 2 larger than the values derived from the experimental spectra. On the other hand, the two calculated rather isotropic hf-tensor values of the side chain ␤-protons were significantly smaller than the experimental values, and the very small calculated 14 N hf-tensor elements were also in disagreement with the experimental EPR spectra. Interestingly, very similar spin densities were calculated for the histidine cation radical, with a proton bonded at N-4. The Mulliken spin densities and the calculated hf-tensor values for both forms of histidine radicals are clearly in disagreement with the experimental values. We therefore ruled out a histidine as site for the radical in P. eryngii VP.
Site-specific Assignment of the Radical to Trp 164 -It had been shown in previous studies (34,53,62) that the hf-tensor values of the side chain ␤-protons of a tryptophan neutral radical, like those of tyrosyl radicals (34,61,66), can vary significantly, even though the spin density at C-3 is very similar in all cases. The magnitude of the isotropic part (A iso ) of the hf-tensor of ␤-protons is a function of the spatial orientation of the side chain and is given by the McConnell relation (67) where BЈ and BЉ are empirical constants (BЈ ϳ 0, BЉ ϭ 5.0 -5.8 mT) and is the dihedral angle between the adjacent -carbon p z axis and the projected C-␤-H-␤ bond (Scheme 1) (34,51,66). The dependence of A iso (H-␤) from the side chain orientation with respect to the tryptophan indole plane is strong because cos 2 can vary between 1 and 0 for each of the two ␤-protons. Using the above range for the BЉ value and considering tetragonal symmetry for the two side chain ␤-protons and the second side chain carbon (i.e. 1 (H-␤1) Ϫ 2 (H-␤2) ϭ 120°) and a spin density of (C-3) ϳ 0.52 (1), dihedral angles of 1 (H-␤1) ϳ ʈ15°ʈ and 2 (H-␤2) ϳ ʈ45°ʈ (estimated error of Ϯ5°) were calculated from the isotropic part (A iso ) of the hf-tensors of the ␤-protons H-␤1 and H-␤2 given in Table 1. This side chain orientation for the tryptophan radical obtained from the experimental data is shown in Fig. 9 together with the orientations for the two VP residues Trp 164 and Trp 244 from the P. eryngii VP* crystal structure very recently solved at atomic resolution.
The side chain (R) orientation for Trp 244 , taken from the x-ray struc-  MOLDEN (44). A, histidine neutral radical, deprotonated and hydrogenbonded to water molecules at both nitrogens (compare with Scheme 1). B, histidine cation radical, hydrogen-bonded at N-␦ (N-4; see Scheme 1). Note that, for the neutral radical, both nitrogens show hydrogen bonds to a water molecule. Blue, positive spin densities; red, negative spin densities. For both radicals, large spin densities were calculated for C-3, giving rise to large hf-tensors of the side ␤-protons (Scheme 1), and for C-2. A particularly large spin density was calculated for the carbon between the two nitrogens (C-5; see Scheme 1), and only small and even negative spin densities were calculated for the two nitrogens. These values are not in agreement with the experimental values (see "Results"). The larger spin densities together with those from the tryptophan radicals are listed in Table 2.
ture (Fig. 9C), is ϳ85°out of the indole plane. The dihedral angles for the ␤-protons are 52°for H-␤1 and Ϫ63°for H-␤2 and, for a tryptophan neutral radical, would lead to hf-tensor elements more than a factor of 2 smaller than those observed in the experiment (Table 1). However, for Trp 164 , these ␤-proton dihedral angles are Ϫ9°and 52° (Fig. 9B), which is in fair agreement with the experimental range (15 Ϯ 5°and 45 Ϯ 5°) when considering that a small reorientation upon radical generation (68) could be possible (see above and Table 1). It is interesting to mention that these two tryptophan residues are homologous to Trp 171 and Trp 251 of P. chrysosporium LiP and have similar side chain orientations, although a third tryptophan residue (Trp 17 ) exists in LiP (whose side chain has the same orientation as LiP Trp 251 ) (26,69). Based on the experimentally determined side chain geometry, we assign the tryptophan neutral radical in P. eryngii to Trp 164 . Furthermore, kinetic data obtained with different single mutants (W164S and H232F) and one double mutant (W164S/P76H) have shown that the LRET pathway starting at Trp 164 is the only pathway that is kinetically competent for the oxidation of (high redox potential) aromatic substrates by VP (70). Closer inspection of the sequences of LiP from P. chrysosporium and VPL from P. eryngii shows that these enzymes have no tyrosines. In fact, the absence of tyrosine residues is a general characteristic of all ligninolytic peroxidases described up to the present (Ͼ30 available sequences) (15). This characteristic seems to be a strategy to prevent competitive generation of tyrosine radicals. In fact, charge transfer between tryptophan and tyrosine and its pH-dependent formation has been established (71) in dipeptides and in a cytochrome c peroxidase variant including a K ϩ -binding site (CcPK2) (72). It has been shown that the solution redox potential for the reaction of the tryptophan neutral radical (TrpT rp ⅐ ϩ H ϩ ) is pH-dependent and is, at ambient pH values, higher than for the respective reaction of a tyrosine neutral radical. On the other hand, the potential required for generation of the cation radical (TrpT rp . ϩ ) is lower for tryptophan and very high for a tyrosine cation radical (73). However, these redox potentials are significantly influenced by the protein environment (72,73). It is very possible that the true catalytic intermediate in VP is the tryptophan cation radical, which becomes (in the absence of substrate), during the time needed for freezing, the protein deprotonated at the indole nitrogen to form the probably more stable observed tryptophan neutral radical. Formation of a relatively stable deprotonated neutral radical was also observed for a tryptophan radical generated by photoinduced electron transfer in azurin (54). Moreover, the existence of exposed tyrosine residues may lead to enzyme inactivation due to intermolecular cross-linking (between the phenoxy radicals formed), as reported during spontaneous decay of cytochrome c peroxidase Compound I (74). Therefore, the absence of tyrosine residues in ligninolytic peroxidases could be a way of preventing enzyme inactivation by dimerization reactions.
Conclusion-Both forms of tryptophan radicals (neutral and cation) are observed in the electron transfer reaction of DNA photolyase (57,75). Tryptophan cation radicals are observed for residues inside the protein, whereas the last radical in the chain, close to the surface of the protein, is deprotonated and becomes a neutral tryptophan (57). Similarly, we observed in P. eryngii a tryptophan neutral radical at position 164 (LiP Trp 171 ), which is located at the protein surface. Our results presented in this study strongly support mechanistic models that involve a tryptophan radical at this position (13,28,32). This tryptophan radical, generated close to the protein surface via LRET from a proposed Fe 4ϩ ϭO porphyrin -cation radical precursor, provides a powerful reactant, which is apparently capable of oxidizing also large molecular mass substrates, which, because of their large size, are not able to reach the active heme site of VP. This ability makes the enzyme particular interesting for biological degradation of large molecules such as industrial dyes and lignin (9,10).
Although our experimental data clearly show that a tryptophan neutral radical is formed in the course of the reaction of VP with H 2 O 2 , the detailed mechanism of the reaction still remains to be resolved (70). For instance, the actual catalytically active form of the radical in the presence of substrate could nevertheless alternatively be a tryptophan cation radical, which, in the absence of substrate, becomes deprotonated after ϳ10 s at the nitrogen to form the more stable neutral radical, which we observed. Additional experiments using fast freeze-quench techniques FIGURE 9. Locations and side chain orientations of tryptophan residues in P. eryngii VP. A, distances from the two tryptophan residues (Trp 164 and Trp 244 ) to heme in the VP* crystal structure determined at 1.33-Å resolution. B and C, orientations of C-␤-H bonds and side chains (edge on view of the C-␤-C-3 bond; see Scheme 1) for the VP* tryptophan radical obtained from the EPR data (Exp.) in comparison with Trp 164 (B) and Trp 244 (C) orientations in the VP* crystal structure. R corresponds to C-␣.
with a time resolution of a few milliseconds, combined with EPR techniques, are envisaged to resolve earlier steps of the catalytic reaction of VP.