INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Electron paramagnetic resonance (EPR) spectroscopy reveals the presence of a free radical in frozen samples of blood (Chernov et al., 1994; Pulatova et al., 1989). This is similar to many other biological tissues. The free radicals in most tissues originate from the mitochondrial electron transport chain (Burgova et al., 1989; Vithayathil et al., 1965). However, there are few mitochondria in blood, so the presence of free radicals there needs a different explanation. We have shown instead that the free radical in blood originates from the reaction of methemoglobin (metHb) with hydrogen peroxide (Svistunenko et al., 1997b). The formation of this species in normal blood, in concentrations ranging from 0.5 to 2 µM, can be used to estimate the intensity of hemoglobin autoxidation in vivo (Svistunenko et al., 1997a,b). However, the chemical nature of the radicals in blood has not been unequivocally established because the EPR spectrum of the blood free radicals is an essentially featureless singlet, making the spectral assignment difficult.

When hemoglobin (Hb) or myoglobin (Mb) in the met form (i.e., in the Fe^III heme state) reacts with H₂O₂ or with other peroxides, the heme is oxidized to the (oxo)ferryl state Fe^IV==O. This one electron oxidation takes place along with H₂O₂ reduction to water, the latter being a two-electron process. The second electron, participating in H₂O₂ reduction, comes from the protein's globin, leaving it in the free radical state (Gibson and Ingram, 1956; Gibson et al., 1958; Kelso King et al., 1967; Kelso King and Winfield, 1963):

(1)

Numerous studies have shown that the optically detectable Fe^IIIFe^IV transition is accompanied by a radical formation demonstrable by EPR spectroscopy. However, the nature and concentration of the radicals formed seem to be strongly dependent on the heme protein being studied. We have shown recently that both peroxyl (ROO) and nonperoxyl radicals are formed in horse heart (HH) metMb, sperm whale (SW) metMb, and human metHb (metHbA) under H₂O₂ treatment and that the yield of these radicals is very different in the different proteins (Svistunenko, 2001).

The peroxyl radical in the HH metMb/H₂O₂ and SW metMb/H₂O₂ systems has been proven to originate from a tryptophan residue (DeGray et al., 1997; Gunther et al., 1995), identified as Trp14 in the case of SW Mb (DeGray et al., 1997). The location of the nonperoxyl radicals is less certain. In low-temperature EPR spectra this radical appears as a singlet (identical to the singlet in blood). The assignment of the singlet to a particular molecular structure is difficult because of the absence of any hyperfine structure or g-factor anisotropy. Comparisons with radicals generated from purified amino acids were equivocal, suggesting that phenylalanine, histidine, or tyrosine could all contribute to the observed spectrum (Kelso King et al., 1967).

Room-temperature studies have the ability to reveal a better resolved structure in EPR spectra. Surprisingly, the reactions of HH metMb and SW metMb with ethyl hydroperoxide (EtOOH) result in two different EPR spectra in the liquid phase: a five-component spectrum in SW metMb and a seven-component signal dominating the spectrum of HH metMb (Miki et al., 1989). The five-component spectrum was assigned to a tyrosyl phenoxyl radical in SW Mb, possibly Tyr151 (Miki et al., 1989). A similar five-component spectrum in leghemoglobin (Lb) treated with peroxides was tentatively assigned to the Tyr132 radical (Davies and Puppo, 1992).

In the case of Hb, the first direct observation of the radical in the metHb/H₂O₂ liquid-phase system was by Shiga and Imaizumi (1975). Although unable to identify the radical precisely because of lack of resolution in the hyperfine structure, the authors pointed out that it must involve a protein in a "slowly tumbling" situation. McArthur and Davies (1993) assigned a similar, also poorly resolved, spectrum in the metHb/peroxide system to a tyrosine-derived phenoxyl radical, stressing that the signal "bears considerable resemblance to those detected with a number of different types of myoglobin and leghemoglobin". However, as indicated above, the room-temperature spectra of the radicals are different in HH and SW myoglobins, making such comparisons ambiguous. In fact, when looked at in detail, the room-temperature EPR spectrum of the metHbA radical is similar to the five-component signal seen in the SW metMb and definitely different from the seven-component signal detected in HH metMb.

This paper will show for the first time a direct comparison between the EPR spectra seen at low temperature and room temperature after peroxide addition to Hb and Mb. This reveals that the nonperoxyl radical detectable as a singlet in the low-temperature EPR spectra is definitely caused by a tyrosine (Tyr) radical. The same species results in a five-component liquid-phase spectrum when studied at room temperature. The seven-component signal detectable in HH Mb arises from a different species.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Protein purification

Human Hb samples were purified from a healthy donor's blood according to Antonini and Brunori (1971). The HH Mb (M-1882) and SW Mb type II (N.M.-0380; this product is no longer commercially available) were from Sigma Chemical Co. (St. Louis MO). Mutagenesis of the synthetic SW Mb gene (Springer and Sligar, 1987) was performed with the U.S.E. Mutagenesis Kit (Amersham Pharmacia, Uppsala, Sweden). The following oligonucleotide was used to mutate Tyr103 to Phe: 5'- CGCTTCAGAGATA-AATTCCAGGAATTTGATCGGG -3'; the nucleotide in bold indicates the mutation site. Screening of the mutants was carried out by restriction analysis, because the oligonucleotide deletes a unique EcoRI site (underlined character), and confirmed by sequencing the entire gene. Wild-type and mutant SW Mbs were expressed in Escherichia coli and purified to homogeneity as previously described (Cutruzzola et al., 1991). The expression levels of the Tyr103Phe mutant were comparable to that of the wild-type protein; no specific effect of the mutation on the protein yield was therefore observed as reported in (Wilks and Ortiz de Montellano, 1992). Because the starting gene used by these authors is different from the one used in this work (and the precise sequence was not published), a possible explanation may be found in a slightly different codon used to build the two wild-type genes and to design the Tyr103Phe mutagenic triplet.

H₂O₂ was purchased from Sigma-Aldrich Co. (Poole, UK). The protein samples were fully oxidized by addition of 2 mM ammonium persulfate and then passed down a Sephadex G-25 column. Oxidized stock solutions were diluted to 100 or 80 µM in 35 mM potassium phosphate buffer at different pH values. All buffers contained 20 µM DTPA (diethylenetriaminepentaacetic acid) as a free iron-chelating agent.

A Hewlett Packard 8453 diode array spectrophotometer was used for optical absorbance measurements. Final concentrations of protein preparations were measured using the ferrous CO complexes as standards. Aliquots of the oxidized proteins were reduced by addition of 10 mM dithionite and then bubbled with CO for 30 s. The following extinction coefficients for the CO-heme forms were used to calculate the heme concentrations (Antonini and Brunori, 1971): ₅₄₀ = 15.4 mM¹cm¹ for HH Mb-CO, ₅₄₂ = 14.0 mM¹ cm¹ for SW Mb-CO, and ₅₄₀ = 13.4 mM¹ cm¹ for HbA-CO.

The final concentrations of the proteins in the met form were measured independently by EPR spectroscopy using the metHb and metMb standards, the concentrations of which were determined as reported earlier (Svistunenko et al., 2000). All quantitative EPR measurements were performed under nonsaturating conditions for all paramagnetic species (a microwave power of 0.05 mW and T = 25 K). The optical and EPR spectroscopy methods showed consistent results in protein determination with differences of less than 10%.

Control experiments with photosynthetic radicals

Arabidopsis thaliana (L.) variety Columbia seeds were germinated in general-purpose compost and grown at 24°C and 70% humidity at a light intensity of 180 µmol m² s¹ with a photoperiod of 16 h and 8 h of darkness in every 24-h cycle. Plants were grown until their rosettes were fully expanded, ~6 days before appearance of the flowering stalk. Plants were light-stressed at 2000 µmol m² s¹ for 1.5 h, and individual leaves were then detached at the petiole, inserted into EPR tubes, and frozen in liquid nitrogen until required.

Reaction of heme proteins with peroxide and EPR sampling

Low-temperature EPR experiments

Room-temperature EPR experiments

EPR measurements and spectra processing

All EPR spectra were measured on a Bruker EMX EPR spectrometer (X-band) at a modulation frequency of 100 kHz. Accurate g-values were obtained using the built-in microwave frequency counter and a 2,2-diphenyl-1-picrylhydrazyl powder standard, the g-value for which is g = 2.0037 ± 0.0002 (Weil et al., 1994). Other instrumental settings are in the figures and have the following abbreviations: microwave frequency  (GHz), microwave power p (mW), modulation amplitude A_m (G), spectra sweep rate  (G/s), time constant  (ms), and number of scans per spectrum NS.

A spherical high quality Bruker resonator SP9703 and an Oxford Instruments liquid helium system were used to measure EPR spectra at low temperature. Unless stated otherwise, the EPR spectra were measured at 10 K. The EPR spectra of the blank samples (frozen water) were subtracted from the EPR spectra of the protein samples to eliminate the base line caused by the resonator's walls, quartz insert, or quartz EPR tube. The g = 2 component of the high spin met heme signal has been subtracted from the EPR spectra of the protein radicals. The first integrals of such corrected spectra converged over the integration range of 200 G in contrast to the spectra before subtraction of the g = 2 heme component, for which the integrals diverged.

Determination of the absolute concentration of the free radicals was performed using a set of Cu²⁺ concentration standards and double integration of the EPR pure lineshapes. The techniques of spectral subtraction with a variable coefficient (Svistunenko et al., 1996, 2000) were used to determine EPR signal intensities. Copper standards and protein samples were measured under nonsaturating conditions.

A 4103TM Bruker resonator was used in the room-temperature experiments. The EPR measurements were started 4-7 s after mixing, and the spectra were run consecutively in an automated regime without time delay between spectra. EPR spectra simulations were performed with WINEPR SimFonia 1.25 (Bruker Spectrospin Ltd., Coventry, UK). The EPR spectra of the SW metMb (Miki et al., 1989) and soybean metLb (Davies and Puppo, 1992) radicals were scanned and digitized with UN-SCAN-IT v.5.0 (Silk Scientific Corp., Orem, Utah).

Three-dimensional protein structure viewing

Swiss-PdbViewer v.3.5 was used to analyze the three-dimensional (3-D) structure of the proteins. Hydrogen atoms cannot be seen in the structures obtained from x-ray crystallography data, so the rotation angle  in Tyr residues (see Fig. 1) cannot be measured directly. Instead, the tyrosine residue in question was oriented as shown in Fig. 1 to represent view A with CR bond being below and not above the phenoxyl ring. We printed out the picture and measured with a protractor the angle  formed by the CR bond and the perpendicular to the phenoxyl ring (from the perpendicular drawn downwards, clockwise to the bond). The angle  was than calculated as  + 60^o.

View larger version (15K): [in this window] [in a new window]   FIGURE 1   Definition of : the angle of the phenoxyl group rotation around C1-C bond in tyrosine.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

The type and the quantity of the radicals formed in metHb and metMb after addition of H₂O₂ is strongly dependent on the protein used. Fig. 2 shows the low-temperature EPR spectra of metHbA, SW metMb, and HH metMb before and 30 s after H₂O₂ addition. The anisotropic EPR signal with g = 6 and g = 2 from the high-spin met form of heme (Fe^III) (Blumberg et al., 1968) decreases on H₂O₂ addition, indicating oxidation of the heme iron to the diamagnetic ferryl state Fe^IV==O. At the same time, a free radical is formed. Although the concentrations of the reactants and the pH values of the medium were the same in all three cases, the free radical EPR signal is different, by size and lineshape, in the three proteins. Two different EPR signals are present in all cases, the proportions of which vary between the proteins (Svistunenko, 2001). One of these two signals is the anisotropic spectrum of a peroxyl radical, with a parallel component at g = 2.037; this species is present at the highest concentration in the HH metMb spectrum. The other signal is an ~19-G wide singlet, present at highest concentration in the metHb spectrum. SW Mb differs only slightly in primary sequence from HH Mb so that the 3-D structures of the proteins are practically superimposable. This difference, however, results in a markedly different pattern of free radical formation. SW metMb responds to H₂O₂ like metHbA, not HH metMb, in that the singlet, and not the peroxyl radical signal, dominates the EPR spectrum (Fig. 2). There is also a third kind of EPR signal that is present only in the HH metMb/H₂O₂ systems. This is a multicomponent signal seen in the low-temperature spectra at pH 6 and is probably due to the same species that is responsible for the seven-component signal in the liquid-phase spectra of the HH metMb/peroxide system (Miki et al., 1989).

View larger version (27K): [in this window] [in a new window]   FIGURE 2   The EPR spectra of human metHb, SW metMb, and HH metMb before and 30 s after addition of H2O2. The 100 µM protein (by heme in the case of Hb) and 100 µM H2O2 were used in all three cases. The reaction was conducted at room temperature at pH 6. The EPR spectra were measured at 10 K. Other instrumental parameters were  = 9.472 GHz, p = 0.8 mW, Am = 4 G,  = 41 ms,  = 14.9 G/s, and NS = 1.

Fig. 3 shows the pure lineshapes of the singlet in the three systems, obtained by the procedure of spectral subtraction with variable coefficients. The high level of noise in the HH Mb spectrum is an indication that the radical responsible for the singlet is formed at a low concentration in this protein under H₂O₂ treatment. The lineshape of the three singlets detected in metHbA, SW metMb, and HH metMb after addition of H₂O₂, are similar. Each is centered at g = 2.005 and has a peak-to-trough width of ~19 G. The microwave power saturation behavior of the signals is also similar. It is therefore likely that the free radicals responsible for the singlet EPR signal in the three proteins are located on analogous amino acid residues. Fig. 4 shows the kinetics of the radical formation and decay in metHbA, SW metMb, and HH metMb after addition of H₂O₂ at pH 7.6. The relative concentrations of the singlet (metHb > SW metMb > HH metMb) are maintained over the whole time course. We have reported before that the low, nonstoichiometric absolute concentration of the radicals often observed in such systems (Kelso King and Winfield, 1963; Kelso King et al., 1967) is a result of a pseudo-steady state, where the rates of radical formation and decay are both H₂O₂ concentration dependent (Svistunenko et al., 1997b). The lineshape of the singlet is essentially unchanged over the pH range 6-8 (illustrated for metHb in Fig. 5).

View larger version (31K): [in this window] [in a new window]   FIGURE 3   The singlet EPR signals detected in the spectra of metHbA, SW metMb, and HH metMb (100 µM by heme) mixed with equimolar amounts of H2O2. The signals shown are results of spectral subtraction procedures, designed individually in each case to attain the best (purest) singlet possible, after the original spectra were detected at optimal conditions for the subtraction procedure (see Appendix). The reaction was conducted at room temperature and pH 8 (metHbA) and pH 7.6 (SW metMb and HH metMb). The original spectra were measured at 10 K,  = 9.468 GHz, Am = 2 G,  = 41 ms, and  = 2.38 G/s. Other instrumental conditions were p = 0.8 mW and NS = 1 (metHbA) or p = 3.18 mW and NS = 2 (SW metMb and HH metMb). The g = 2 component from the high-spin heme signal has been subtracted from all the spectra before applying the procedure of subtraction with variable coefficients. The intensity of all signals was normalized to the same peak-to-trough value. The inset shows the microwave power saturation curve for the singlet in metHbA.

View larger version (16K): [in this window] [in a new window]   FIGURE 4   The kinetic dependences of the radicals with the singlet EPR signal in the three proteins after addition of H2O2. The proteins' concentrations (heme concentration in the case of Hb) and the concentration of H2O2 were 100 µM, pH 7.6. The radical concentrations for HH metMb were multiplied by 5 for illustrative purpose. The random error involved in the concentration measurements was 2-3%. The systematic errors associated with each series were estimated as ±0.3 µM for metHbA, ±0.4 µM for SW metMb, and ±0.15 µM for HH metMb. The overall systematic error associated with the concentration axis was estimated as 20%.

View larger version (21K): [in this window] [in a new window]   FIGURE 5   The singlet EPR signals detected at low temperature in the 100 µM metHbA/100 µM H2O2 systems at three different pH values. The spectra were obtained by using the procedure of spectral subtraction with variable coefficients as described in the Appendix. The original spectra were measured at 10 K,  = 9.471 GHz, Am = 2 G,  = 41 ms,  = 2.38 G/s, p = 0.8 mW, and NS = 1. The g = 2 component from the high-spin heme signal was subtracted from all spectra before applying the procedure of spectral subtraction.

When the reaction of metHbA with H₂O₂ was studied at room temperature, a poorly resolved EPR signal was detected (Fig. 6). We have shown before, both at 10 K and at room temperature (Svistunenko et al., 1996, 1997b), that the free radical in the metHb/H₂O₂ system very rapidly reaches a steady state, the duration of which lengthens with increasing H₂O₂ molar excess. However, the steady-state radical concentration decreases with increasing molar ratio H₂O₂/heme (Svistunenko et al., 1996, 1997b). An experimentally optimal 10:1 ratio of H₂O₂ to heme was therefore used to observe the free radical for a longer duration. The EPR feature observed is best described as an anisotropic quintet; this signal is similar to that reported previously for this system (McArthur and Davies, 1993; Shiga and Imaizumi, 1975; Svistunenko et al., 1996). Fig. 6 shows the kinetic dependences of the signal at three different pH values. The signal decreases more rapidly at higher pH. The time taken to attain a steady state is twice as long at pH 6 as at pH 7 or 8, although the steady-state concentration is not strongly pH dependent. As is the case for the low-temperature singlet (Fig. 5), the lineshape of the five-component signal is similar in the pH range 6-8 (Fig. 6 inset). The lineshape does not change in the course of the reaction.

View larger version (29K): [in this window] [in a new window]   FIGURE 6   Kinetic dependences of the EPR signals detected at room temperature in 0.71 mM (heme) metHbA mixed with 7.1 mM H2O2 (targeted final concentrations) at three different pH values (aerated condition). The EPR spectra were measured in an automated consecutive manner at the following instrumental conditions:  = 9.798 GHz, p = 1.27 mW, Am = 2 G,  = 10.2 ms,  = 9.54 G/s (sweep time 10.5 sec), NS = 1, and T = 293 K. The spectra for the first time point of each dependence, normalized to equal amplitude, are shown in the inset. The data points correspond to the time of half-scan of each spectrum when the clock was started at the moment of mixing.

The lineshape of the singlet detected in the frozen samples of the metHbA/H₂O₂ system (Fig. 5) is not completely featureless, suggesting an underlying hyperfine structure. We suggest that this singlet is caused by the same radicals as those responsible for the better-resolved liquid-phase spectrum (Fig. 6 inset) as the hyperfine structure seen at room temperature becomes less resolved when the sample is frozen. To check this we overlaid the spectra of the metHbA/H₂O₂ system measured at the same field width at room temperature and at 10 K (Fig. 7, A and B). Because microwave frequency differs between low-temperature and room-temperature EPR experiments, the resonance magnetic field for the same paramagnetic species is also different. The magnetic field axis is therefore omitted in Fig. 7, the signals being overlaid on the basis of their common g-factor. The hyperfine features discernible on the singlet fit well the five components of the liquid-phase spectrum (Fig. 7, A and B).

View larger version (35K): [in this window] [in a new window]   FIGURE 7   The EPR signals of the tyrosyl radical. (A) Spectrum of a frozen sample of 100 µM metHbA reacting with 100 µM H2O2 at pH 8. The sample was frozen 29 s after mixing, and the spectrum was measured at 10 K. Other instrumental parameters:  = 9.471 GHz, p = 0.013 mW, Am = 2 G,  = 41 ms,  = 2.38 G/s, and NS = 10. (B) Spectrum measured at room temperature of metHbA reacting with H2O2 at pH 8 (the concentrations and instrumental parameters as in Fig. 6). A single scan was started 5 s after mixing and lasted for 10 s. (C) Difference spectrum ({whole leaves stressed by intense light for 1 h}  {whole unstressed leaves}); both stressed and control leaf spectra were measured at 10 K. Other instrumental conditions were  = 9.473 GHz, p = 0.050 mW, Am = 1 G,  = 82 ms, v = 0.42 G/s, and NS = 1. The spectra were superimposed on the basis of the indicated g-factor, 2.0058. The noisy traces are the experimentally obtained spectra; the smooth traces overlaid with the experimental spectra A and B are the computer simulations with the parameters given in Table 1. The gridlines are drawn at a 10-G interval.

The liquid-phase spectrum can be simulated using the hyperfine splitting constants for a tyrosyl radical published previously (Fig. 7 B and Table 1). To attain the asymmetrical lineshape characteristic of the spectrum (Fig. 7 B), a g-factor anisotropy has to be introduced (Table 1). The 10 K spectrum of the singlet (Fig. 7 A) can then be simulated using a very similar set of parameters, merely requiring a slight increase of all hyperfine splitting constants and a slight decrease in the g-factor anisotropy (Table 1). The third spectrum in Fig. 7 C is shown for comparative purposes. It is the EPR signal of the radical produced in a green plant (A. thaliana) under high light stress. This spectrum has been well characterized as being due to a tyrosyl radical in photosystem II (Barry and Babcock, 1987; Barry et al., 1990; Warncke et al., 1994). The spectra in Fig. 7, B and C, are identical in lineshape.

Fig. 8 shows the EPR spectra of the wild-type recombinant SW metMb and of the Tyr103Phe variant, both proteins frozen 30 s after addition of an equimolar amount of H₂O₂ (spectra a). Both spectra a contain peroxyl radical signals, the features of which can be removed as described in the Appendix. The lineshape of the Trp peroxyl radical is slightly different in the wild type and the Tyr103Phe mutant, indicating that the symmetry of this radical in the mutant is more axial type than in the wild type. However, the integrated intensity of both signals, and hence concentration of the peroxyl radical, is similar in the two proteins (spectra b). In contrast, replacement of Tyr103 with a phenylalanine results in essentially a complete loss of the singlet EPR signal (spectra c).

View larger version (14K): [in this window] [in a new window]   FIGURE 8   The EPR spectra of the wild-type recombinant SW metMb and the Tyr103Phe mutant. (a) Protein samples (80 µM) frozen 0.5 min after addition of H2O2 (80 µM) at pH 7; (b) Peroxyl radical signals pure lineshapes obtained as described in the Appendix; (c) Difference spectra c = a  b. The g = 2 component of the high-spin metMb has been subtracted from the experimental spectra, so that spectra a comprises only the free radical signals. The g-factors are indicated with arrows. Instrumental parameters were  = 9.4681 GHz, p = 3.18 mW, Am = 3 G,  = 164 ms,  = 0.89 G/s, NS = 1, and T = 10 K.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

In this study we have directly compared, for the first time, room- and low-temperature EPR spectra of the nonperoxyl radical formed after peroxide addition to ferric hemoglobin and myoglobin. We demonstrate that the singlet observed at low temperature is due to a tyrosine radical.

The first evidence that a tyrosine residue was involved in free radical formation after the reaction of heme proteins with peroxides was provided in Miki et al. (1989). Tetranitromethane treatment was used to remove only one Tyr from SW Mb, which treatment did not result in the loss of any tyrosines in HH Mb. The changes in the EPR spectra led the authors to suggest that the radical in SW metMb (with the five-component EPR signal) originated from Tyr151, the only one of the three SW Mb tyrosines not present in HH Mb (Miki et al., 1989).

Another five-component liquid-phase EPR signal, similar to that in the SW Mb/EtOOH system (Miki et al., 1989), was detected in soybean metleghemoglobin (metLb) immediately after mixing with H₂O₂ or other peroxides (Davies and Puppo, 1992). Lb, a monomeric oxygen-transporting heme protein in root nodules of legumes, is similar in sequence and structure to mammalian Mbs. However, there is no Tyr151 in Lb. Instead, there are three tyrosines at positions 25, 30, and 133 (Hargrove et al., 1997) (the last residue was identified as 132 in earlier studies (Fuchsman, 1985)). Davies and Puppo (1992) attributed the Lb radical to a tyrosine (possibly Tyr132/133) phenoxyl radical.

A five-component EPR signal is also detectable in the liquid phase in metHbA/peroxide systems. This signal, first reported for the metHb/H₂O₂ system in 1975 (Shiga and Imaizumi, 1975), was later shown for metHb reacting with other peroxides, such as EtOOH, 2-butanone peroxide, iodosylbenzene, or periodate (McArthur and Davies, 1993). When metHb was acetylated (to block tyrosine residues), the signal was not observed. The responsible species was therefore hypothesized to be a tyrosine phenoxyl radical (McArthur and Davies, 1993).

All these findings indirectly point to a possible link between the five-component spectrum and globin tyrosine residues. However, simulations of this signal in the past have not corresponded in detail to the lineshape of the experimental spectrum (McArthur and Davies, 1993). In this paper we show that the five-component signal in of metHbA is identical to the EPR spectrum of whole leaf (Fig. 7, B and C). The latter is known to originate from a tyrosyl radical of photosystem II (Barry and Babcock, 1987; Barry et al., 1990; Warncke et al., 1994). Furthermore, the five-component liquid-phase EPR signal in the metHbA/H₂O₂ systems can be simulated using the hyperfine splitting constants established for a tyrosyl radical elsewhere (Fig. 7 B and Table 1). The simulated and experimental spectra overlay precisely. Therefore, we have no doubt that the five-component EPR signal detected in the liquid phase of the metHb/peroxide system is due to a tyrosine radical.

Could it be that the other two five-component signals in SW Mb (Miki et al., 1989) and Lb (Davies and Puppo, 1992) are also caused by tyrosine radicals? These two signals have different lineshapes, each of which is also different from the Hb quintet. However, the overall widths of all three signals are similar. The SW metMb and metLb quintets can be simulated by hyperfine splitting constants that are characteristics of a tyrosine radical (Fig. 9). The major difference in the simulation parameters for the three globin radicals (Fig. 9 and Table 1) is the splitting constants on the methylene protons. Different splitting on the protons in different proteins can be explained by differences in the dihedral angles formed by the C₁-C-H₁ plane (or C₁-C-H₂) and the ring plane (Fig. 1).

View larger version (15K): [in this window] [in a new window]   FIGURE 9   The EPR spectra assigned in the literature to Tyr radicals and our simulations of the spectra. (A) SW metMb/EtOOH (Miki et al., 1989) (noisy trace); this spectrum was simulated (smooth trace) for a Tyr radical (Fig. 1) with g = 2.006, g = 2.004, a = 5.2 G, a = 0 G, a = 8 G, a = 6.5 G, and H = 3.7 G. (B) Soybean metLb/H2O2 (Davies and Puppo, 1992) (noisy trace); the parameters of the simulated spectrum were as follows: g = 2.0046, g = 2.0042, a = 6.4 G, a = 1.5 G, a = 12.5 G, a = 2.2 G, and H = 5.5 G (smooth trace). Lorentzian lineshape was used in both simulations. Both A and B correspond to the 80-G field width increasing from left to right.

As the phenoxyl ring rotates around the C₁C bond, the spin density of the -orbital system on the 1s orbitals of the methylene hydrogens (-protons) changes according to a cos² pattern, where  is the rotation angle (Stone and Maki, 1962). If  is defined as shown in Fig. 1, then  could be found from the following:

(2)

where a and a are experimentally measured hyperfine splitting constants on the two protons,  is spin density on C₁ and B" is an empirical constant that defines  proton coupling constant B = B' + B"cos² (B' is negligible).

Thus, the hyperfine splitting constants on the two methylene protons follow the same cos² pattern but are phase-shifted with respect to each other by 120°, which is the angle between the CH bonds.

System 3 was solved for the three sets of a and a presented in this paper for the Tyr radicals in the liquid phase in HbA, SW Mb, and soybean Lb (Table 1 and Fig. 9). Generally, Eq. 2 has four solutions on the  range 90° to +90°, two of which must be discarded as having a definitely wrong value for the second variable, B". The other two solutions are identical and indistinguishable as far as the EPR method is concerned. The sum of these two values of  is strictly the angle between the two CH bonds, i.e., 120° (Table 2).

View this table: [in this window] [in a new window]   TABLE 2   Rotation angle  in the Tyr radical as determined from the hyperfine splitting constants on the methylene protons (Table 1 and Fig. 9) and as found for all tyrosines present in the proteins from the known 3-D structure

Thus, the values of the methylene protons' hyperfine splitting constants, obtainable from the EPR spectrum of a Tyr radical, provide information about the dihedral angles in the radical and therefore indicate which Tyr residue in the protein could be the site of the radical. Sometimes the constants cannot be found precisely from an experimental EPR spectrum, for example, when one of the constants is smaller that the line width. This may limit the usefulness of this approach.

Angles  found from Eq. 2 were compared with those determined from the 3-D structure (Table 2). It has been previously suggested that Tyr133 in Lb was the host of the free radical because it is the closest Tyr to the heme (Davies and Puppo, 1992). Our analysis confirms that the dihedral angle in Tyr133 (52° as found from the protein's 3-D structure) is the closest to the angle derived from the radical's spectral features (46.7° as found from Eq. 2), and therefore Tyr133 is the most likely site responsible for the radical seen in Lb (Table 2). In SW Mb, Tyr151 seems the most likely candidate, because  = 87° in this residue is closest to  = 61.7° found as an Eq. 2 solution for the SW Mb EPR spectrum (Table 2). The situation is less clear in HbA, partly because the angles calculated from the x-ray structure vary for the same tyrosines in identical chains (A and C chains,  subunits; B and D,  subunits). However, Tyr42, Tyr35, and Tyr130 seem the most likely candidates (Table 2). As we will discuss later simple mutation of these residues is not likely to yield unequivocal data about where the radical resides in the wild-type protein. Therefore, more accurate information on the hyperfine splitting constants deducible from the EPR spectra and on the exact bond angles in the protein structure are equally important for final assignment of the radical to a specific residue.

The low-temperature EPR of globin radicals has also been extensively studied. How can the low-temperature EPR signals be related to those observed at room temperature? The EPR spectra of the HH metMb/H₂O₂ system measured at a low temperature show mainly a peroxyl radical. This cannot be formed from a tyrosine radical and was demonstrated to originate from a tryptophan residue (DeGray et al., 1997; Gunther et al., 1995). Similar peroxyl radicals are also present in the frozen metHbA/H₂O₂ system. However, the major feature in these spectra is a singlet EPR signal. The SW metMb/H₂O₂ system is in an intermediate situation with the peroxyl radical and singlet signals both contributing strongly to the spectrum (Fig. 2).

When looked at in detail, the low-temperature singlet in metHbA has an underlying hyperfine structure (Fig. 7 A). When overlaid with the liquid-phase spectrum, it becomes clear that the features on the lineshape of the singlet correspond well to the components of the liquid-phase quintet. The singlet could be simulated with parameters only slightly different from those used to simulate the liquid-phase tyrosine radical spectrum (Fig. 7 and Table 1). We conclude therefore that the singlet EPR signal observed at low temperature in the metHbA/H₂O₂ system is caused by a tyrosine radical. Room-temperature EPR has not been able to detect directly free radicals in blood. However, a signal is seen at low temperature. This signal is identical to that produced after peroxide addition to metHb in vitro (Svistunenko et al., 1997a,b). Therefore, we now can state unequivocally that the identity of the free radical seen in frozen blood is a hemoglobin-bound tyrosine radical.

Because the parameters and properties of the singlets are similar in the three globins studied (Figs. 2 and 3), it is likely that tyrosyl radicals are responsible in each case. When the three proteins are compared, a correlation can be seen between the intensities of the liquid-phase quintet (definitively assigned to tyrosine) and the 10 K singlet. The strongest singlet and the strongest quintet are both seen in HbA. The singlet is smaller in SW Mb and much smaller in HH metMb (Fig. 4). This is in accordance with the result that the liquid-phase quintet was detected in the SW metMb/peroxide system but was not seen in the HH metMb under similar conditions (Miki et al., 1989). We therefore suggest that the singlet EPR signals observed at low temperature in SW metMb and HH metMb under H₂O₂ treatment (Fig. 3) are also caused by tyrosyl radicals.

All possible Tyr-to-Phe mutants of recombinant SW metMb were studied in their reaction with H₂O₂ (Wilks and Ortiz de Montellano, 1992). However, in this study, there was a significant contribution from a peroxyl radical signal in all spectra. Not subtracting the contribution of this species from the overall spectra made comparisons between the Tyr mutants difficult. We have undertaken a similar study of the Tyr103Phe mutant. We show that, after subtraction of the signal from the (tryptophan) peroxyl radical the Tyr103Phe mutant has essentially no singlet signal: changing Tyr103 to Phe has completely removed the tyrosine radical. Yet our spectral analysis (Table 2) suggests that Tyr103 is not in the correct conformation to explain the hyperfine constants of the Tyr radical observed in the liquid phase:  in Tyr103 is 134° as can be found from the 3-D structure, whereas the EPR spectrum (quintet in Fig. 9 A) suggests that  = 58.3° or  = 61.7° (the two solutions of Eq. 2 with a = 8.0 G and a = 6.5 G as parameters). Instead, Tyr151 ( = 87°) is a more likely candidate. Interestingly, the Tyr151Phe mutant also seems to show a diminished singlet signal, just like the Tyr103Phe mutant (Wilks and Ortiz de Montellano, 1992). Tyr103 is the closest to the heme. We suggest the radical forms first on Tyr103; a rapid electron transfer between this primary site and Tyr151 then occurs. Tyr151 is the more stable radical detectable in the steady state. Mutations in either Tyr103 or Tyr151 can therefore block the formation of a significant Tyr free radical signal in SW Mb. The fact that both Tyr103 and Tyr151 free radicals can be spin trapped in SW metMb reacting with H₂O₂ (Gunther et al., 1998) is consistent with such a mechanism.