α-Hemoglobin Stabilizing Protein (AHSP) Markedly Decreases the Redox Potential and Reactivity of α-Subunits of Human HbA with Hydrogen Peroxide*

Background: α-Hemoglobin stabilizing protein (AHSP) modifies the redox properties of bound α-subunits. Results: Isolated hemoglobin subunits exhibit significantly different redox properties compared with HbA. A significant decrease in the reduction potential of α-subunits bound to AHSP results in preferential binding of ferric α. Conclusion: AHSP·α-subunit complexes do not participate in ferric-ferryl heme redox cycling. Significance: AHSP binding to α-subunits inhibits subunit pseudoperoxidase activity. α-Hemoglobin stabilizing protein (AHSP) is a molecular chaperone that binds monomeric α-subunits of human hemoglobin A (HbA) and modulates heme iron oxidation and subunit folding states. Although AHSP·αHb complexes autoxidize more rapidly than HbA, the redox mechanisms appear to be similar. Both metHbA and isolated met-β-subunits undergo further oxidation in the presence of hydrogen peroxide (H2O2) to form ferryl heme species. Surprisingly, much lower levels of H2O2-induced ferryl heme are produced by free met-α-subunits as compared with met-β-subunits, and no ferryl heme is detected in H2O2-treated AHSP·met-α-complex at pH values from 5.0 to 9.0 at 23 °C. Ferryl heme species were similarly not detected in AHSP·met-α Pro-30 mutants known to exhibit different rates of autoxidation and hemin loss. EPR data suggest that protein-based radicals associated with the ferryl oxidation state exist within HbA α- and β-subunits. In contrast, treatment of free α-subunits with H2O2 yields much smaller radical signals, and no radicals are detected when H2O2 is added to AHSP·α-complexes. AHSP binding also dramatically reduces the redox potential of α-subunits, from +40 to −78 mV in 1 m glycine buffer, pH 6.0, at 8 °C, demonstrating independently that AHSP has a much higher affinity for Fe(III) versus Fe(II) α-subunits. Hexacoordination in the AHSP·met-α complex markedly decreases the rate of the initial H2O2 reaction with iron and thus provides α-subunits protection against damaging oxidative reactions.

one for ␣-subunits would accelerate autoxidation. We previously suggested that AHSP stabilizes a hemichrome folding intermediate and prevents its incorporation into HbA until the bound ␣-subunit can be reduced to the ferrous form (17).
Ferrous and ferric forms of HbA and Mb are known to react with H 2 O 2 to produce ferryl heme (Fe(IV)) species (6) as part of a pseuodoperoxidative cycle, where ⅐ Hb refers to a protein-based radical (6). Hamdane et al. (18) reported that reacting ferric ␣-subunits with H 2 O 2 in the absence of AHSP yields the formation of a ferryl heme intermediate with a rate constant of 320 M Ϫ1 s Ϫ1 , whereas Feng et al. (15) reported that AHSP binding to ferric ␣ markedly inhibits the formation of H 2 O 2 -induced ferryl heme species. However, these events have not been well characterized for isolated ␣-subunits in either the absence or presence of AHSP, even though both ferryl heme and protein-based radicals are known to initiate a cascade of lipid oxidation reactions, which are thought to be involved in Hb toxicity in vivo (4).
Both Tyr and Trp are capable of acting as redox intermediates by facilitating electron flow pathways in metalloproteins (19). These residues are thought to function as part of a pathway for electron flow between solvent reducing equivalents and the heme iron. Tyr-42 in ␣-subunits has been shown to play a critical role in reducing ferryl heme through such a mechanism (20). These asymmetric, through protein electron transfer pathways may explain the apparent differences between the pseudo-peroxidase activities of ␣and ␤-subunits observed in intact HbA tetramers (20).
To investigate underlying mechanisms of AHSP action, we studied the formation of ferryl heme intermediates and their protein-based radicals ( ⅐ HbFe(IV) ϭ O) in isolated HbA subunits and ␣-chains in the absence and presence of AHSP. Autoxidation rates, redox potentials, and radical formation after addition of H 2 O 2 were measured and compared for isolated ␣and ␤-subunits and AHSP⅐␣-subunit complexes. Significant differences between isolated ␣-subunits and AHSP⅐␣subunit complexes were observed. Unlike ␤-subunits, isolated ␣-subunits do not form significant amounts of protein radicals after reaction with H 2 O 2 , although a small amount of ferryl heme is detected. In contrast, no ferryl heme species or associated protein radicals are observed when ␣-subunits are bound to AHSP. These findings show that ␣-subunits become resistant to the generation of protein radicals and ferryl heme species when bound to AHSP, and that hexacoordination of the heme iron within AHSP⅐met-␣-subunit complexes appears to be the physiologically preferred species.

EXPERIMENTAL PROCEDURES
Protein Expression, Purification-Recombinant human AHSP was produced using previously published methods (12,17). HbA was obtained from expired units of human blood (Gulf Coast Regional Blood Bank, Houston, TX) and was purified using previously published methods (21). HbA subunits were also isolated using previously published methods (22), as modified (17). This method occasionally resulted in catalase contamination in ␤-subunit samples. Catalase could be removed by size exclusion chromatography using Superdex 200 media (GE Healthcare). In the case of HbA and subunits, previously determined extinction coefficients were used to determine protein concentrations in heme equivalents (23). The AHSP extinction coefficient utilized was 11,460 M Ϫ1 cm Ϫ1 at 280 nm, and was calculated using ExPASy Proteomics Server Protparam. Purified haptoglobin (Hp) solution was a kind gift from Bio Products Laboratory (Hertfordshire, United Kingdom). The isolation and fractionation of this protein from human plasma was done as previously reported (24). Size exclusion HPLC chromatograms of the Hp samples used in this study show the following molecular weight distribution: 60% ␣␤ dimers (Hp 1-1), 21% ␣␤ trimers (Hp 1-2), and 19% larger polymers (Hp 2-2).
Spectrophotometry-Autoxidation studies were performed at 37°C using air equilibrated 0.05 M potassium phosphate, pH 7.0, at 37°C (25,26). Spectra were recorded every 1 min for 4 h, using an integration time of 0.5 s and an interval of 1 nm. AHSP and ␣-subunit concentrations were fixed at 10 M in heme equivalents. Where indicated, 10 g/ml of superoxide dismutase and 200 units/ml of catalase were added to the cuvettes prior to the start of the reactions (25). Co-oxidation of epinephrine was followed under the same buffer conditions by monitoring absorbance at 475 nm using 600 M epinephrine and the same concentrations of superoxide dismutase and catalase (25).
To detect the ferryl heme oxidation state, ferric subunits were first generated by addition of a 10-fold molar excess of potassium ferricyanide to O 2 -bound ferrous proteins, followed by brief incubation and removal of the potassium ferricyanide using Sephadex G-25 chromatography media. These steps were performed as quickly as possible at 4°C because ferric subunits are highly unstable at room temperature. Materials were used within minutes of preparation. Ferryl heme detection studies were completed by manual mixing and stopped-flow spectrophotometry. In our stopped-flow experiments, 30 M protein in heme equivalents was mixed with 3 mM H 2 O 2 (post-mixing concentrations given) at 8°C. In the manual mixing experiments, 60 M protein in heme equivalents was mixed with 90 M H 2 O 2 at 22°C. Both sets of experiments utilized 10 mM potassium phosphate buffer, pH 7.0. In the stopped-flow experiments, higher H 2 O 2 concentrations were chosen to accelerate the time courses and to mitigate sample denaturation stemming from necessarily longer sample preparation times. Lower H 2 O 2 concentrations were used in benchtop spectrophotometer experiments to ensure that the reactions were not too rapid to be measured conventionally. Where indicated, 2 mM sodium sulfide (Na 2 S) was used to derivatize the ferryl heme oxidation state as sulfhemoglobin using established methods (27). Buffers used for variable pH experiments include 200 mM potassium phosphate (pH 7.0 and 8.0 at 23°C) and 200 mM sodium acetate (pH 5.0 and 6.0 at 23°C) (28).
Spectroelectrochemistry-Intrinsic reduction-oxidation potentials were determined under anaerobic condition using a custom-built optically transparent thin layer electrode cell. The path length of this cell is 0.06 cm and [Ru(NH 3 ) 6 ]Cl 3 was used as a cationic mediator (29 -31). Well resolved differences in Soret band absorption for Fe(II) and Fe(III) heme species were used to follow the extent of reduction at different applied potentials E applied , which is defined as follows.
Where E 1 ⁄2 o is the midpoint potential when [Hb red ] ϭ [Hb ox ], R is the universal gas constant, T is the absolute temperature, n is the number of electrons transferred in the redox process for an ideal Nernstian system and is a measure of cooperativity for a non-Nernstian system, and F is the Faraday constant. All potentials are reported with respect to normal hydrogen electrode (31).
In a typical experiment, hemeproteins were placed in 1 M glycine buffer, pH 6.0, at 8°C, and degassed gently on a salt/ice mixture and then purged using argon gas. ␤-Subunit concentrations were fixed at 10 M, and free ␣-subunit and AHSP⅐␣subunit complexes were fixed at 80 M (heme equivalents). The heme to mediator ratio was kept at 1:10 for all reactions. This anaerobic solution was then introduced into an optically transparent thin layer electrode cell fitted with a platinum mesh working electrode, platinum wire auxiliary electrode, and a silver/silver-chloride reference electrode. The potentials were controlled using an EG & G Princeton Applied Research Potentiostat Model 263 (AMETEK, Berwyn, PA) and the Soret bands for oxidized and reduced species were recorded using a Cary 100 spectrophotometer (Varian, Inc., Palo Alto, CA). Heme proteins were oxidized to the Fe(III) state by application of a positive potential, and spectra were recorded after a steady absorbance measurement was obtained at 405 nm for oxidized ␣-subunits and 413 nm for oxidized AHSP⅐␣-subunit complexes. Once a stable absorbance was achieved for the Soret band at an oxidizing potential, ensuring complete conversion of all the heme groups to the Fe(III) state, the potential of the working electrode was made progressively more negative. At each relatively negative potential the analyte solution was equilibrated for 15 min and the absorbance of both oxidized and reduced species were recorded until at the end point all of the protein samples were converted to reduced heme leading to a stable reduced form Soret band (430 nm for both free reduced ␣-subunits and reduced AHSP⅐␣-subunit complex). At the end of the experiment, the electrode potential was reversed in polarity to observe the reversibility of the redox process (31).
EPR-EPR spectra were recorded using a Bruker EMX spectrometer equipped with a liquid helium continuous flow cryostat, connected with a GFS600 transfer line and an ITC503 temperature controller (Oxford Instruments, Abingdon, Oxfordshire, UK). The experimental parameters were: frequency, 9.6 GHz; power, 1.0 milliwatts; modulation amplitude, 1 G; modulation frequency, 100 kHz; and temperature, 10 K. Samples were prepared as follows using 0.05 M potassium phosphate buffer, pH 7.0, at 22°C. In the case of AHSP⅐␣-subunit complexes, stoichiometric amounts of each protein were added to a 1.0-ml volume of buffer prior to oxidation using known molar extinction coefficients (17,23). In the case of Hp⅐Hb complexes, excess HbA was added to a small volume of Hp, the sample was allowed to incubate for 15 min on ice, and excess HbA was removed using a Superose 12 10/300 GL column (GE Healthcare). Samples were oxidized to the ferric state by addition of a 10-fold molar excess K 3 Fe(CN) 6 and incubation for ϳ3 min on ice. Each sample was then passed through a 10-DG (Bio-Rad) desalting column to remove excess K 3 Fe(CN) 6 Table 2.
Reagents and Instrumentation-Unless otherwise specified, all reagents and chemicals were obtained from Thermo Fisher Scientific (Waltham, MA) or Sigma. Chromatographic media, columns, and chromatography equipment were obtained from GE Healthcare Bio-Sciences Corporation (Piscataway, NJ) and Whatman International Ltd. (Maidstone, Kent, United Kingdom). UV-visible absorbance spectroscopy measurements were made using an Agilent 8453 diode-array spectrophotometer (Agilent Technologies, Inc., Santa Clara, CA) or a Cary 100 Spectrophotometer (Varian, Inc., Palo Alto, CA). In both cases, 1-cm path length cells were used. Stopped-flow measurements were made using an Applied Photophysics SX-18 microvolume stopped-flow spectrophotometer (Leatherhead, Surry, UK). The path length was 10 mm and the entrance and exit slit widths were set to 1 mm each to give 4.8-nm spectral band widths. The volume of the cell was 20 l. Shot volumes were between 100 and 200 l, and mixing was performed using equal volumes of reactant solutions.
Data Analysis-Microsoft Excel was used for nonlinear least square data fitting to a single exponent expression to obtain the observed rate constants (Microsoft Corp., Redmond, WA) (32). Fitting routines in Origin were also used to verify the values obtained from Excel (OriginLab Corp., Northampton, MA). Spectroelectrochemical and sulfheme measurements were repeated three times to obtain standard deviations, and autoxidation reactions were repeated four times. Structure images were created using the PyMOL Molecular Graphics System (Schrödinger, LLC, New York).

RESULTS
Autoxidation-HbA oxidation from the ferrous (Fe(II)) to the ferric (Fe(III)) state occurs spontaneously in the presence of air-equilibrated aqueous buffer and produces superoxide anion (O 2 . ) (33,34). The released superoxide anion then undergoes dismutation to H 2 O 2 and O 2 , and the H 2 O 2 produced in this step can react with the protein to produce several other species (6). This spontaneous conversion of ferrous to ferric state in hemeproteins under aerobic conditions and the subsequent redox processes involving the superoxide anion, H 2 O 2 , and ferric Hb are broadly classified as autoxidation. Although AHSP has previously been shown to accelerate the initial rate of ␣⌷ 2 -subunit autoxidation (16), the generation of subsequent ferryl heme and protein radical species (Eqs. 1 and 2) have not been well defined. Initially we examined whether AHSP⅐␣⌷ 2 autoxidation occurs by mechanisms similar to those reported for native tetrameric HbA. AHSP⅐␣O 2 complexes were allowed to autoxidize in the presence and absence of superoxide dismutase (an O 2 . scavenger) and/or catalase (an H 2 O 2 scavenger) in accordance with the methods of Misra and Fridovich (25). The rate of autoxidation was monitored by optical absorbance changes in the visible region. AHSP⅐␣O 2 complexes exhibit peaks at ϳ541 and ϳ576 nm (16,23), whereas AHSP⅐met-␣ complexes exhibit peaks at 535 and 565 nm following oxidation to the ferric state (13,15,16). The latter positions are indicative of a hemichrome or bishistidyl conformation in which the imidazole nitrogen atoms of ␣His-58 and ␣His-87 both coordinate the ferric iron atom and exclude H 2 O and other exogenous ligands (15). Representative time courses produced from monitoring absorbance decays at 576 nm are shown in Fig. 1A.
In agreement with previous studies (16,35), 10 M AHSP⅐␣subunit complexes autoxidize with an apparent rate constant of ϳ2.3 Ϯ 0.2 h Ϫ1 (Fig. 1A). The addition of catalase reduced the apparent rate constant to 1.2 Ϯ 0.2 h Ϫ1 under the same experimental conditions. Superoxide dismutase did not appreciably affect the apparent rate constant of autoxidation, which was measured to be 2.4 Ϯ 0.3 h Ϫ1 in its presence. For comparison, Fig. 1B depicts the initial phases of the autoxidation plots for isolated ␣and ␤-subunits, which show at least 10-fold slower apparent rate constants (Յ0.3 h Ϫ1 ). AHSP binding to ␣-subunits induces conformational changes that facilitate superoxide dissociation and hemichrome formation. Because catalase converts H 2 O 2 into H 2 O and O 2 , these data suggest that the slower rate of autoxidation in the presence of catalase is due to the diminished availability of H 2 O 2 , which is capable of reacting with ferrous and ferric HbA to facilitate more rapid oxidation of the protein. Caughey and co-workers (26) observed a similar decrease in the rate of autoxidation of HbAO 2 due to addition of catalase. The absence of a significant effect of superoxide dismutase agrees with most previous work (25,26) Even in the case of catalase, the effect is very small compared with the dramatic effect of AHSP binding on the autoxidation of ␣O 2 (Fig. 1B).
The absence of an effect in the case of superoxide dismutase, which facilitates the rapid dismutation of O 2 . into H 2 O 2 and O 2 , agrees with most previous work (25), although at least one conflicting report exists (26). However, both effects are very small compared with the effect of AHSP binding to ␣O 2 . We also examined H 2 O 2 and O 2 . production from AHSP⅐␣O 2 complexes during autoxidation. We recorded spectral changes during the co-oxidation of epinephrine to adrenochrome in the same buffer and at the same temperature. We recorded changes in absorbance at 475 nm in the presence and absence of catalase and superoxide dismutase (25). This process has previously been used to assay for H 2 O 2 and O 2 . production during HbA autoxidation (25). Representative data from these studies are shown in Fig. 1C. Both enzymes showed additive effects on the diminution of the observed rate of co-oxidation of epinephrine. Similar effects of catalase and superoxide dismutase are observed for the autoxidation of native HbA (25). Ferryl Heme Species-In addition to the oxidation of ferrous heme to the ferric heme state, iron atoms within HbA can undergo further oxidation to the ferryl heme state in the presence of reactive oxygen species (6,7). This process has been associated with radical formation both in vitro and in vivo (36,37), and significant work on HbA shows that ferryl heme formation occurs following oxidation of both ferrous and ferric protein by H 2 O 2 (6,7,36). Recently, unique tyrosine-mediated inter-subunit electron transfer pathways have been proposed for intact HbA (20). Because autoxidation of oxy-Hb produces O 2 . , which spontaneously dismutates to form H 2 O 2 , we sought to characterize the behavior of both subunits in isolation with respect to H 2 O 2 reactivity. HbA oxidation to the ferryl heme state can be induced in vitro by subjecting the protein to moderate excesses of H 2 O 2 (36). Ferryl heme formation causes the emergence of two broad . production by AHSP⅐␣-subunit complexes. A, representative autooxidation time courses in the absence and presence of catalase or superoxide dismutase. B, time courses for autoxidation of isolated ␣O 2 -and ␤O 2 -subunits and the AHSP⅐␣O 2 complex. C, co-oxidation of epinephrine by AHSP⅐␣O 2 in the presence of catalase or superoxide dismutase. Autooxidation was performed at 37°C using air-equilibrated 0.05 M potassium phosphate, pH 7.0, at 37°C, and 10 M protein in heme equivalents. Where indicated, 10 g/ml of superoxide dismutase (SOD) and 200 units/ml of catalase (CAT) were added prior to data collection. Co-oxidation of epinephrine was followed at 475 nm using 600 M epinephrine and the same buffer and temperature conditions as were used for autoxidation (25). Complete time courses for isolated ␣ and ␤ could not be obtained due to protein precipitation. Data in panel A were normalized to total absorbance signal changes. Open circles, squares, and triangles represent data points for AHSP⅐␣ alone, AHSP⅐␣ with superoxide dismutase, and AHSP⅐␣ with catalase, respectively. Lines in panel A are theoretical fits generated using the single-exponent expression Y t ϭ Y 0 ϩ Y 1 e Ϫk1*t . Observed rate constants from four replicates performed on two different days are given in the text. Data in panels B and C were normalized to correct for slight concentration differences. FEBRUARY 8, 2013 • VOLUME 288 • NUMBER 6 optical absorbance peaks with maxima at ϳ545 and 585 nm (36). We first mixed ferrous oxygenated HbA, ␣, ␤, and AHSP⅐␣-subunit complexes with H 2 O 2 using conditions similar to those reported by Tomoda et al. (38). These reactions were spectrally monitored for up to 2 min in a benchtop optical absorbance spectrophotometer. In agreement with the previous report (38), we found that isolated ferrous ␣O 2 -and ␤O 2subunits form hemichromes and precipitate on this time scale. In contrast, we observed that aquo-or hydroxy-(met) ␣and ␤-subunits form relatively stable, transient ferryl heme complexes, which are similar to those observed for tetrameric met-HbA.

Redox Chemistry of AHSP
As shown in Fig. 2A, mixing 30 M ferric ␤-subunits with 3 mM H 2 O 2 (post-mixing concentrations) results in the emergence of broad absorbance bands at ϳ547 and 585 nm, which are indicative of formation of a Fe(IV) ϭ O complex. The reaction of H 2 O 2 with ferric ␣-subunits under equivalent conditions results in similar spectral transitions, with the emergence of ferryl-hemichrome-like spectra with absorbance peaks at ϳ543 and ϳ582 nm (Fig. 2B). The red lines in Fig. 2 represent spectra obtained for the reaction of intact ferric HbA with H 2 O 2 using identical experimental conditions. Fig. 2C shows spectra for the reaction of AHSP⅐(met)-␣-subunit complexes with H 2 O 2 using identical experimental conditions. We observed no absorbance changes between 3 ms and 20 s post-mixing. These data suggest that AHSP completely inhibits ferryl heme formation on these time scales (Fig. 2C), even in the presence of large excesses of H 2 O 2 .
On longer time scales between 20 s and 2 min, isolated ferric subunits precipitated following exposure to H 2 O 2 , as evidenced by marked increases in solution turbidity (not shown). In agreement with other reports (13, 15), however, we find that AHSP binding to ␣-subunits inhibits this H 2 O 2 -induced precipitation.
We detect no spectral evidence for ferryl heme intermediates under any conditions examined.
Because the absorbance spectra of hemichrome and ferryl heme species are similar (compare red spectra in Fig. 2, A and B, with the black spectrum in C), we used sodium sulfide (Na 2 S) to further characterize the extent of ferryl heme formation in isolated met-subunits and AHSP⅐(met)-␣-subunit complexes. This reagent has previously been shown to react with ferryl heme groups in HbA and myoglobin to produce quantifiable derivatives that give an absorbance peak at ϳ620 nm (27). The iron in the ferryl heme is reduced to the ferrous state and the sulfur is incorporated onto the porphyrin (27). This reaction depends both on the kinetics of ferryl heme formation and its rate of reaction with Na 2 S, and has widely been used to derivatize ferryl Hb in vitro and in biological samples (27,39).
We added 90 M H 2 O 2 to 60 M ferric ␣-subunits, ␤-subunits, HbA, and AHSP⅐␣-subunit complexes, incubated the samples at room temperature for 60 -120 s (see the legend of Fig. 3 for the precise conditions), and recorded spectra before and after this incubation. After recording spectra at the completion of each incubation, 2 mM Na 2 S was rapidly added to each sample, and again spectra were recorded. Representative spectra of the absorbance peaks at ϳ620 nm are shown in Fig. 3,  A, B, and D. These data indicate that H 2 O 2 -treated met-HbA and met-␤ form sulfheme species quite readily (Fig. 3). However, ␣-subunit sulfheme formation is more limited, due in part to a slower rate of ferryl heme formation and a greater rate of precipitation (Fig. 3D). In contrast, the AHSP⅐(met)-␣-subunit complexes do not form any detectable sulfheme species, verifying the spectral measurements in Fig. 2 (see Fig. 3C). Using a previously published extinction coefficient (27), the calculated  .0, at 22°C, 2 mM Na 2 S was added and spectra were immediately recorded. Na 2 S was added 70 s following H 2 O 2 addition in the case of ␣ and AHSP⅐␣, 90 s in the case of ␤, and 120 s in the case of HbA. Various other incubation times were also assayed (not shown). However, simultaneous oxidation and precipitation of ferric subunits creates a limited time in which ferryl heme intermediates can be detected. The peak at ϳ620 nm is due to ferrous sulfhemoglobin formation (27). Each panel contains an overlay of ferric (black), ferryl (red), and sulfheme (blue) Hb spectra. To further investigate our findings, we altered the evolutionarily conserved AHSP proline 30 in recombinant AHSP to generate AHSP(P30A) and AHSP(P30W) mutant proteins (17,40). Although these mutations do not detectably perturb erythropoiesis in mouse models, AHSP(P30W) has been shown to bind oxygenated ␣-subunits with a 30-fold increased affinity in vitro, and both variants cause decreased rates of autoxidation and increased rates of hemin loss relative to wild-type AHSP (17,40). In conducting experiments analogous to those presented in Fig. 3, we found that these variants exhibit the same behavior as wild-type AHSP (data not shown). This suggests that once the hemichrome is formed, the reaction with H 2 O 2 is strongly inhibited through a mechanism that is independent of the specific residue at position 30.
Alkaline conditions have previously been reported to stabilize the ferryl heme oxidation state (28,41). We therefore examined the H 2 O 2 reactions and sulfide addition at various pH values using isolated ferric ␣and ␤-subunits and AHSP⅐␣-subunit complexes, using previously reported experimental conditions (28). We were able to detect large amounts of the ferryl heme state in ␤, small amounts in ␣, but none in the AHSP⅐␣-subunit complex at any pH (Fig. 4).
Midpoint Redox Potential-The redox potentials of isolated HbA subunits have previously been investigated by two groups using potentiometric methods (42,43). Neither group reported the effects of AHSP binding to ␣-subunits on midpoint potential. Our previous measurements showed that AHSP binds met-␣ with a much higher affinity than ␣CO, ␣O 2 , and deoxy-␣ (17,44). This result predicts that AHSP binding should mark-edly decrease the reduction potential of isolated ␣-subunits, provided the binding equilibrium is significantly dependent on the oxidation state of heme.
To verify this prediction, we measured the reduction potential of ␣-subunits in the presence and absence of AHSP under anaerobic conditions using a spectroelectrochemical method. As shown in Fig. 5, the Nernst plots for isolated ␣-subunits and AHSP⅐␣-subunit complexes are linear and show evidence of a simple, noncooperative, single electron reduction process. Based on these data, we calculate an E 1 ⁄2 o of ϩ40 (versus normal hydrogen electrode) for free ␣-subunits, which agrees with the reports of both Abraham and Taylor (42) and Banerjee and Cassoly (43). Binding to AHSP lowers the midpoint potential to Ϫ78 mV. These values are listed in Table 1, along with the reduction potentials we determined for HbA and Hp⅐Hb complexes. Like AHSP, Hp binding also lowers the reduction potential of HbA, but to a lesser extent, from ϩ120 to ϩ54 mV (24). Interestingly, the redox process involving AHSP⅐␣-subunit complexes was completely reversible as observed by reversing the polarity of the working electrode at the end of each spectroelectrochemical experiment. This result indicates that reduction does not dissociate the AHSP⅐␣ complex at the micromolar protein concentrations used in the measurements, as predicted from previous measurements of the K D for binding of the ferric (0.0006 M) and ferrous (0.017 M) forms of ␣ to AHSP (17). E 1 ⁄2 o values measured for isolated ␤-subunits were variable and differed substantially from those reported by other groups (42,43). Although the reason for this is unclear, we found that irreversible ␤-subunit denaturation during our experiments prevented reliable measurement. Also, ␤-subunits readily self-associate into homotetramers (45). Neither our experiments nor those of Abraham and Taylor (42) or Banerjee and Cassoly (43) adequately controlled for this phenomenon.
Radical Formation-We also investigated the effects of AHSP on protein radical formation using low temperature EPR spectroscopy. We prepared ferric ␣-subunits, ␤-subunits, HbA,   FEBRUARY 8, 2013 • VOLUME 288 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 4293 AHSP⅐␣-subunits, and Hb⅐Hp complexes using the methods described under "Experimental Procedures," and recorded spectra before and following the addition of either 0.1 M sodium fluoride or a 1.5-fold molar excesses of H 2 O 2 . Fig. 6A contains spectra recorded in the presence or absence of sodium fluoride, which normally converts aquo-met or hydroxy-met samples into completely high spin forms. All three ferric resting samples showed EPR spectra indicative of a mixture of high and low spin ferric iron. The g values of the low spin ␣-subunits (2.792, 2.248, and 1.687) and AHSP⅐(met)-␣subunit complexes (2.932, 2.247, and 1.737) indicate that both species have a histidine imidazole distal ligand according to modified Truth-Diagram analysis (46), and the g values for the low spin ␤ (2.752, 2.253, and 1.830) indicate a hydroxide ligand (Fig. 6A). The amount of low spin ferric iron in all three samples is substantial and should not be underestimated by directly comparing the apparent amplitudes of the high and low spin heme signals. The extent of low spin character is revealed by comparing the heights of the g ϭ 6 signals of the met samples with those in the EPR spectrum of the corresponding fluoride derivatives. Fluoride binding converts all Fe 3ϩ atoms into the high spin state with the g ʈ component at g ϭ ϳ2 showing a doublet due to the nuclear hyperfine of the fluorine (Fig. 6A) and a maximal intrinsic g ϭ ϳ6 signal, which can be used to estimate the total amount of Fe 3ϩ in each sample by converting it all to high spin. The intensity of the observed g ϭ 6 signals for the untreated samples can then be compared with that for the met⅐fluoride complex to estimate the amount of high spin iron. Thus, the high spin heme signals of ␣and ␤-subunits are ϳ22 and 21% of the total heme iron, respectively, assuming that the height of the derivative signals remain proportional to the concentration of high spin heme detected at g ϭ 6 in each case.

Redox Chemistry of AHSP
Fluoride failed to convert the low spin ferric heme iron in the AHSP⅐(met)-␣-subunit complex to high spin, indicating that fluoride cannot access the iron once AHSP is bound due to its strong stabilization of bis-His coordination. We estimated the amount of high spin heme in AHSP⅐␣-subunit complexes using the amplitude of the EPR g ϭ 6 signal for the fluoride complexes of the ␣and ␤-subunits. As shown in Table 2, the percentage of high spin heme is Յ10% in the AHSP⅐(met)-␣-subunit complex, much less than that of the ␣and ␤-subunits, and in sharp contrast to that of met-HbA, which is as high as 76.4%. This much reduced level of high spin heme in AHSP⅐(met)-␣-subunit complex is clearly due to bis-His coordination. Estimates of the amounts of high and low spin ferric ␣-subunit, ␤-subunit, HbA, AHSP⅐␣-subunit complex, and Hp⅐␣␤ are summarized in Table 2. Fig. 6B shows spectra recorded from samples incubated with H 2 O 2 for 10 s prior to freezing. ␤-Subunits treated with H 2 O 2 show a significant amount of protein-based free radical signal at g ϭ 2.0 under these conditions. The H 2 O 2 -reacted met-␣ sample shows less signal, and the AHSP⅐(met)-␣ sample shows almost no detectable free radical peak g ϭ 2.0.
EPR spectra of the radical species recorded in the 100 gauss range are presented in Fig. 7. The radical generated in the ␣-subunit is centered at g ϭ 2.0038 with a symmetric line shape and an overall width of 20 gauss. The radical signal found for ␤-subunits, although with a similar center at g ϭ 2.0039, is asymmetric with an additional low field component at 2.033 and clear hyperfine structures in the major signal, with an overall width of 24.5 gauss. The half-saturation power is also different, 2.4 versus 10.4 milliwatts for the ␣and ␤-subunits, respectively, indicating the location and structure of the protein radical found in the ␣and ␤-subunits are likely different.  The percentages of radical yields are shown in Table 2, together with similar data obtained for the Hp⅐Hb complex for comparison under equivalent experimental conditions. The reaction of H 2 O 2 with Hp-met-Hb dimers shows a level of radical signal, which is comparable with those observed for isolated met-␤ or met-HbA, consistent with a recent observation (47), and is 50-fold greater than that seen for the AHSP⅐(met)-␣ complex.

DISCUSSION
The reaction between H 2 O 2 and ferrous Hb is a two-electron process that results in the formation of ferryl heme iron (6). If H 2 O 2 reacts with ferric Hb, a radical species is formed in addition to the oxoferryl species (Eqs. 1 and 2) (6). Both the ferryl heme and its associated protein radical can induce a wide variety of oxidative reactions that affect nearby molecules due to their high midpoint redox potentials, ϳ1 V (48). Specific chemical modifications to free Hb have been observed if proteins are challenged with excess H 2 O 2 in vitro. First, the heme vinyl groups may become modified and covalently linked to the protein (48). Second, extensive globin chain cross-links and irreversible modifications of several amino acids, primarily in ␤-subunits, have been observed (49). Studies have also revealed the presence of oxidative changes to the surrounding tissues following exposure to extracellular Hb, both in animals infused with these proteins and when Hb is released from human red cells leading to kidney or brain injury (8,50,51). It is not surprising that several pathways exist in mammals to control these events. Two key globin-binding proteins that appear to inhibit oxidative damage are: (a) AHSP, which provides protection against oxidative damage to ␣-subunits and surrounding proteins during erythropoiesis (12)(13)(14) and (b) Hp and the CD163 receptor on macrophages, which coordinate Hb dimer clearance during hemolysis (50). Our results indicate that AHSP and Hp exert these functions through distinct mechanisms.
Preferential Binding of AHSP to Met-␣-Current evidence suggests that AHSP protects free ␣-subunits from oxidative degradation by preferentially binding to the ferric form and rapidly inducing structural changes that generate a stable hexacoordinated species (12,13,15,17). Our spectroelectrochemical experiments confirm that AHSP binds preferentially to met-␣-subunits. The midpoint reduction potential of free ␣-subunits decreases from ϩ40 to Ϫ78 mV when the subunit is bound to AHSP (Fig. 5B, Table 1). These values can be used to independently estimate the ratio of the equilibrium dissociation constants for AHSP binding to reduced deoxy-␣ and met-␣-subunits (i.e. K D,red /K D,ox ). A complete derivation of the effect of AHSP binding on the reduction potential of ␣-subunits is given in the Supplemental Scheme 1 in Supplemental Derivation. Using this method and the parameters in Table 1, the computed ratio is ϳ130, which is almost identical to the ratio computed from the bimolecular association and unimolecular dissociation rate constants (i.e. 100) reported by Mollan et al. (17).
Interestingly, under comparable experimental conditions, the midpoint potential of HbA was found to decrease from ϩ120 mV in the tetrameric form to ϩ54 mV in the Hp-bound ␣␤-dimeric form (Table 1) (24). Thus, both Hb-binding proteins facilitate oxidation, AHSP by inducing hemichrome formation and Hp by promoting dissociation into more easily oxidized dimers. These two proteins also deal with peroxideinduced radical chemistry in markedly different ways (see below).
Inhibition of H 2 O 2 Reactions by AHSP-Early structural work on AHSP⅐␣-subunit complexes led to the proposal that the bishistidyl conformation strongly inhibits ferric-ferryl redox cycling following exposure to H 2 O 2 (15,52). It was estimated that at least 33% of the heme in ferrous ␣⌷ 2 ⅐AHSP complexes converts to the ferryl heme form following H 2 O 2 exposure, whereas less than 10% of the heme in ferric met-␣⅐AHSP complexes converts to the ferryl heme form following the same exposure (15). A more recent report indicates that H 2 O 2 -induced covalent modification of ␣-subunits can be prevented by AHSP binding, and that two exposed ␣-subunit Tyr residues (Tyr-24 and Tyr-42) are unable to take part in electron transfer and ferryl heme protein radical formation (Fig. 8) (18).
In this work, we focused on ferric-ferryl redox transitions within the AHSP⅐␣ complex. We confirmed that AHSP⅐␣ autoxidizes ϳ10-fold more rapidly than isolated ␣-subunits (13,16), and found that this process both produces and consumes H 2 O 2 . We also found that in the presence of molar excesses of H 2 O 2 , the ferric forms of HbA, ␣, and ␤ can react to form ferryl heme intermediates, which can be detected using Na 2 S. By contrast, AHSP⅐(met)-␣-subunit complexes show no evidence of H 2 O 2 -induced ferryl heme intermediates or species (Figs. 2 and 3). This lack of peroxidase activity is likely a consequence of low accessibility of H 2 O 2 to the heme iron once AHSP is bound and induces bis-His coordination (Fig. 8C). Internal hexacoordination also explains the lack of conversion from low spin to high spin Fe 3ϩ by addition of excess fluoride to the met-␣⅐AHSP complex (Fig. 6A). These findings are confirmed in our experiments using buffers with pH values varying FIGURE 7. Close-up EPR spectra of the radical species before and after H 2 O 2 treatment. Individual spectra are labeled accordingly. EPR conditions were: microwave power, 1 milliwatt; microwave frequency, 9.3 GHz; range 140 G; modulation amplitude, 2 G; temperature, 20 K. Spectra were measured using the same samples as were used to generate the data in Fig. 6. from 5.0 to 8.0 (Fig. 4). Fig. 8D outlines a scheme of AHSP function based on these findings.
Differences between Met-␣, Met-␤, and AHSP⅐(met)-␣-Subunits-Although our data show that isolated ␣-subunits behave markedly differently than isolated ␤-subunits with respect to ferryl heme and protein radical formation, it is clear that more work is needed to determine the underlying mechanistic differences, including more directed measurements of the rates of formation and decay of other intermediates and site-directed mutagenesis studies designed to determine the structural origin of the radical signals in Fig. 7.
For example, the smaller radical and ferryl heme signals for ␣-subunits could be due to stronger water coordination to Fe(III), which inhibits the initial reaction with H 2 O 2 (53). There is evidence that the distal histidine in ␣-subunits stabilizes bound ligands and distal pocket water to a greater extent than in ␤-subunits because of a more rigid conformation and closer proximity of the imidazole side chain to the coordinated O atom (Fig. 8, A and B) (54,55).
In addition, Reeder et al. (20) have shown that ferryl-␣-subunits both autoreduce or react with added reducing agents ϳ10 times more rapidly than ferryl-␤-subunits, and that these differences are due in part to facilitation of electron transfer by the solvent-exposed Tyr-42 side chain at the CD corner of the ␣but not ␤-subunits (Fig. 8). However, the extent of ␣Tyr-42 participation in the peroxidase reaction is not fully defined and studies with isolated mutant subunits remain to be done.
By contrast, the mechanistic cause of the lack of reactivity of the AHSP⅐(met)-␣ complex with H 2 O 2 and formation of ferryl heme species is much clearer. A dramatic conformational change in the CD corner of ␣-globin occurs on binding to AHSP and autoxidation, which results in hemichrome formation with both His-58(E7) and His-87(F8) coordinated to the Fe(III) atom (Fig. 8C). This hexacoordination appears to completely inhibit the reaction of AHSP⅐(met)-␣ with H 2 O 2 to form the initial intermediates observed in free isolated subunits.
Summary and Physiological Relevance-Our study provides three biochemical findings of relevance to erythroid physiology. First, met-␤-subunits are more prone to the H 2 O 2 -induced formation of ferryl heme species and protein-based radicals than met-␣-subunits. This may explain why ␣and ␤-thalassemic erythroid cells exhibit distinct membrane abnormalities and oxidative modifications (56,57). It also suggests that Hbbased O 2 therapeutics should be designed with more attention directed at mitigating oxidative reactions of ␤-subunits. Second, AHSP binding renders met-␣Hb nearly inert to oxidative degradation by H 2 O 2 , with no ferryl heme species or proteinbased radicals detected by either optical absorbance or EPR. Third, AHSP binding dramatically lowers the redox potential of ␣ to a much more negative value, and thermodynamically favors the ferric over ferrous iron.
Although AHSP and Hp both protect against Hb-mediated oxidative damage, the biochemical mechanisms are distinct. AHSP stabilizes nascent ␣-subunits by inducing the formation of a stable hemichrome that inhibits interactions with H 2 O 2 and other ligands. By contrast, Hp-bound ␣␤ dimers retain their pseudoperoxidase activity, although the resulting ferryl heme species appear to be less reactive due in part to the stabilization of a protein-based radical on ␤Tyr-145 (47).