Biochemical Fates of α Hemoglobin Bound to α Hemoglobin-stabilizing Protein AHSP*

Alpha hemoglobin-stabilizing protein (AHSP) is an erythroid protein that binds free α hemoglobin (αHb) to maintain its structure and limit its pro-oxidant activity. Prior studies have defined two different αHb·AHSP complexes. Binding of AHSP to Fe(II) αHb induces an unusual configuration in which the F helix of the globin becomes disordered and the heme ring becomes solvent-exposed. Over time, this intermediate oxidizes to form a stable hemichrome in which the proximal (F8) and distal (E7) histidines are coordinated to the heme iron atom. The addition of βHb to either Fe(II) or Fe(III) αHb·AHSP displaces AHSP to generate tetrameric (α2β2) HbA species. The biochemical properties and in vivo significance of the two αHb·AHSP complexes are poorly understood. Here we show that Fe(III) αHb·AHSP forms from auto-oxidation of oxygenated αHb bound to AHSP and that this process is greatly accelerated at physiologic temperature and oxygen pressures. In contrast to free Fe(III) αHb hemichromes, AHSP-bound Fe(III) αHb does not precipitate and can be recycled into functional HbA. This requires enzymatic reduction of AHSP-bound αHb, either prior to or after extraction by β subunits. In contrast, reaction of Fe(II) αHb-AHSP with βHb generates functional HbA directly. Our findings support a model in which AHSP can either stabilize αHb transiently en route to HbA formation during normal erythropoiesis or convert excessive free αHb into a more chemically inert state from which recovery of αHb is possible by redox cycling.

Alpha hemoglobin-stabilizing protein (AHSP), 4 also called erythroid-associated factor or ERAF, is an abundant erythroid protein originally identified in a screen for genes that are regulated by the essential hematopoietic transcription factor GATA-1 (1). Previously, we have shown that AHSP specifically binds and stabilizes the ␣ subunit of hemoglobin and reduces its cytotoxic effects in vivo, at least partly by inhibiting its ability to generate reactive oxygen species (2). Certain characteristics of the AHSP-␣Hb interaction are established (3)(4)(5)(6). 1) AHSP binds ␣Hb via the ␣ 1 ␤ 1 dimer interface. Therefore, AHSP and ␤Hb cannot bind ␣Hb simultaneously. 2) ␤Hb binds ␣Hb more tightly than AHSP. Hence, ␤Hb easily displaces AHSP from ␣Hb. 3) AHSP can induce oxidation of the heme iron of ␣Hb. 4) AHSP binding induces significant structural rearrangements in the globin chain of ␣Hb.
The initial binding of AHSP to free oxy-␣Hb causes disordering of the globin F helix, movement of the heme group, displacement of the F8 histidine, coordination of the Fe(II) heme iron by the distal histidine (E7), and binding of O 2 to the proximal side of the heme group (5). In this unique structure, the oxygen binding site is exposed to solvent, which predisposes it to the spontaneous loss of HO ⅐ 2 and oxidation of the heme iron (auto-oxidation) (7). Indeed, the Fe(II) oxy-␣Hb⅐AHSP complex oxidizes much more rapidly at ambient temperature and oxygen pressure than free Fe(II) oxy-␣Hb alone. Concomitantly, there are further structural changes, and the heme iron becomes coordinated by both the F8 (proximal) and the E7 (distal) histidines (6). This AHSP-bound Fe(III) bis-histidyl or hemichrome structure is more resistant to denaturation and hemin loss than oxidized free ␣Hb. This configuration also inhibits the reaction of Fe(III) ␣Hb with H 2 O 2 and other oxidants, presumably because the hexacoordinate iron is less available for direct reaction with these substrates. Thus, hemichrome formation is one mechanism by which AHSP stabilizes ␣Hb and renders it more chemically inert. Formation of AHSP⅐␣Hb complexes may be particularly important to limit erythrocyte damage in ␤ thalassemia, a hemoglobinopathy in which reduction of ␤ globin synthesis causes excessive free ␣Hb to accumulate (2,8). It is also likely that AHSP acts as a chaperone to stabilize either apo or holo ␣Hb during normal HbA synthesis (1).
Although a number of ␣Hb-AHSP interactions have been described by in vitro biochemical and structural studies, certain processes remain unclear. For example, we have shown that AHSP induces the formation of bis-histidyl ␣Hb over several hours under ambient conditions, but how the rate of this reaction is affected by temperature and oxygen tension is unknown. ␤Hb displaces AHSP from ␣Hb to form ␣ 2 ␤ 2 HbA-like com-plexes, suggesting that AHSP can stabilize ␣ chains transiently prior to incorporation into HbA. However, AHSP binding causes significant structural rearrangements around the ␣Hb heme group in both the Fe(II) oxy-␣Hb⅐AHSP and Fe(III) bishistidyl ␣Hb⅐AHSP complexes. In this current work, we have shown that some of these structural changes are retained when the ␣ subunits are displaced from AHSP by ␤Hb and that fully functional tetrameric HbA is not always formed. Determining whether or not fully active hemoglobin can be generated from ␣Hb⅐AHSP complexes is critical for determining how AHSP functions in vivo. We conducted the present study to address two key questions. 1) How readily is the bis-histidyl Fe(III) ␣Hb⅐AHSP complex formed under physiologic conditions? 2) Can AHSP-bound Fe(II) or bis-histidyl Fe(III) ␣Hb recombine with ␤Hb to form functional HbA? Our findings provide the basis for a model to explain how AHSP may function in erythroid cells to augment Hb assembly and stability.

EXPERIMENTAL PROCEDURES
Interactions of ␣Hb and AHSP-Recombinant human AHSP, AHSP⅐ glutathione S-transferase, ␣Hb, and ␤Hb, were prepared as previously described (2,4). Fe(III) ␣Hb was prepared by incubating oxy-␣Hb with a 4-fold excess of potassium ferricyanide followed by gel filtration chromatography using Sephadex G50 beads. Fe(II) ␣Hb⅐AHSP complex was prepared by incubating 10 M purified ␣Hb with equimolar recombinant AHSP on ice for 30 min in 20 mM sodium phosphate, pH 7.4, 100 mM NaCl, and 10 M diethylenetriaminepentaacetic acid (DTPA). Samples of this complex were incubated at various temperatures in a quartz cuvette and the spectral changes followed over time in a PerkinElmer Life Sciences Lambda 25 UV-visible spectrophotometer to monitor conversion to the Fe(III) bis-histidyl form. Deoxygenated Fe(II) ␣Hb was generated from oxy-␣Hb by three cycles of vacuum and nitrogen purge in a Teflon sealed quartz cuvette followed by adding a molar excess of sodium dithionite. Equimolar AHSP was added using a gas-tight Hamilton syringe to generate deoxy-Fe(II) ␣Hb-AHSP. Fe(III) ␣Hb⅐AHSP complex was prepared by incubating Fe(II) ␣Hb⅐AHSP for several hours and verifying the complete conversion spectrophotometrically. Standard extinction coefficient spectra of Fe(II) and Fe(III) complexes were prepared from a series of concentrationdependent spectra. The conversion rates of Fe(II) to Fe(III) complexes were calculated following linear regression of the experimental spectra to the two extinction coefficient spectra utilizing Sigma plot software (Systat Software) (9).
Formation and Characteristics of Tetrameric Hb from AHSP⅐␣Hb Complexes-To examine the displacement of AHSP from its complex with ␣Hb, equimolar ␤Hb was added and the mixture incubated at room temperature for 5 min. Cellulose acetate electrophoresis was performed using chromatography strips (Helena Laboratories, Beaumont, TX) according to the manufacturer's instructions. The tetramers resulting from displacement of AHSP by ␤Hb were examined for oxygen binding by standard tonometry (10). Briefly, 3 ml of the reaction mixtures were deoxygenated via three cycles of vacuum and nitrogen purge in a 300-ml tonometer. Known volumes of oxygen were added at room temperature and pressure, the samples rolled at 22°C for 10 min, and spectra recorded. Semi-oxidized Hb heterotetramer ␣(FeIII) 2 ␤(FeII) 2 was prepared by mixing Fe(III) ␣Hb with an equimolar ratio of oxy-␤Hb (11). Rates of cyanomet Hb formation were monitored spectrophotometrically after the addition of a 4-fold molar excess of potassium cyanide to various Hb species. Reduction of oxidized Hbs was performed utilizing a ferredoxin reductase system, as previously described (12). Briefly, Hbs were premixed with glucose-6-phosphate, nicotinamide adenine dinucleotide phosphate, ferredoxin-nicotinamide adenine dinucleotide phosphate reductase, ferredoxin with or without superoxide dismutase and catalase (Sigma). Thereafter, the reaction was initiated by adding glucose-6-phosphate dehydrogenase (Sigma) at 37°C under aerobic conditions, and the accompanying spectral changes were recorded.
EPR Measurements-EPR spectra of Fe(III) ␣Hb subunits free in solution, bound to AHSP, and bound to Fe(II) carbonmonoxy (CO) ␤Hb subunits, were recorded with a Brucker EMX EPR spectrometer. The instrument conditions for the EPR measurements were: frequency, 9.60 GHz; power, 10 milliwatts; modulation amplitude, 10.9 G; modulation frequency, 100 kHz; and temperature, 4.5 K. The high spin signal at g ϭ 6 (ϳ1200 G) was quantified by double integration between 800 and 1700 G and comparison with the signal for (␣FeIII) 2 (␤FeII) 2 Hb prepared by mixing newly oxidized ␣Hb with reduced CO ␤Hb subunits at pH 7 (see Fig. 6C). The original ␣Hb samples were prepared at 4°C in EPR tubes as described under "Results" and then frozen, first in a dry ice/methanol bath and then stored in liquid nitrogen.

RESULTS
Through interactions with ␣Hb, AHSP is essential for normal erythropoiesis and Hb homeostasis. AHSP binds several forms of ␣Hb, but the functional and biological implications of these different interactions are unknown. Here we investigate the biochemical properties of AHSP⅐␣Hb complexes, providing potential molecular mechanisms for AHSP actions in vivo. Previously, we have shown that Fe(II) ␣Hb⅐AHSP converts to a Fe(III) bis-histidyl hemichrome complex (5). However, this reaction occurs over several hours at room air and temperature, which brings into question its biological significance, as erythropoiesis occurs at higher temperatures and lower oxygen tension. Therefore, we investigated the effects of these parameters on the rate of Fe(III) complex formation (Fig. 1). The conversion of ␣Hb⅐AHSP from the oxygenated Fe(II) form to the Fe(III) bis-histidyl hemichrome is accompanied by characteristic spectrophotometric changes that allow the kinetics of this reaction to be monitored. The UV-visible spectrum of the Fe(II) complex was followed during incubation at a variety of temperatures ( Fig. 1, A-C). The spectra were fitted to a two-component linear regression model utilizing standard spectra for the Fe(II) and Fe(III) ␣Hb⅐AHSP complexes (Fig. 1D). From these fits, the proportion of oxidized heme was calculated and plotted against time (Fig. 1E). A dramatic increase in the rate of conversion to the Fe(III) complex with increasing temperatures from 22 to 37°C occurred (Fig. 1F). This large temperature dependence was quantitated in terms of an activation energy (E a ) obtained from an Arrhenius plot of lnk autox versus 1/T. The observed value is quite large (121 kJ/mol) and is almost identical to that reported for the auto-oxidation of intact HbA (13), implying that both processes have a similar mechanism. The value of E a for auto-oxidation is markedly higher than that for either O 2 binding to (E a Ϸ 20 -30 kJ/mol) or dissociation from (E a Ϸ 60 -70 kJ/mol) mammalian Hbs and myoglobins (14).
AHSP-bound oxy-␣Hb contains a disorganized F helix on which the proximal histidine F8 resides, increasing exposure of the heme iron to solvent (5). This open heme pocket, with a lack of ligand coordination, is known to be associated with increased . From the curves in B, the rate of oxidation (k autox ) was calculated and is shown versus the complex concentration. These data fit best to a K d of 5 M and a k autox value of 0.077 min. Ϫ1 . The relative k autox value for free ␣Hb is 0.001 min Ϫ1 (4) and is similar to that of oxygenated HbA dimers (15).
rates of auto-oxidation of the heme iron (7). Therefore, we reasoned that the conversion of ␣Hb⅐AHSP from the Fe(II) to the Fe(III) form might occur through auto-oxidation. In this case, the conversion would be oxygen-dependent. We tested this by examining the Fe(II) ␣Hb⅐AHSP complex after deoxygenation by sodium dithionite (Fig. 2). Anaerobic addition of one equivalent of dithionite induced a spectrum indicative of deoxygenated Fe(II) Hb ( Fig. 2A, dashed line). This form of ␣Hb⅐AHSP was stable, with minimal spectral changes after 1 h of incubation at 37°C (Fig. 2A, solid line). In contrast, 100% of fully oxygenated Fe(II) ␣Hb⅐AHSP converted to the Fe(III) species upon similar incubation in room air ( Figs. 1 and 2).
When Fe(II) ␣Hb⅐AHSP complex was diluted from 15 to ϳ1 M complex in air, the observed rate of auto-oxidation decreased markedly (Fig. 2, B and C). This result is the opposite of what is observed when intact oxyhemoglobin is diluted over the same concentration range. Zhang et al. (15) show that the rate of auto-oxidation increased from ϳ0.0001 min Ϫ1 for Hb tetramers at high concentrations to 0.0007 min Ϫ1 for Hb dimers observed at low concentrations. The most straightforward explanation for the decrease in the auto-oxidation rate of oxy-␣Hb⅐AHSP with dilution is that the complex is dissociating into AHSP and free oxy-␣Hb, which has a much smaller autooxidation rate. In the simplest model, the observed rate of autooxidation would be given by, where Y complex is the fraction of oxy-␣Hb bound to AHSP. k autox,␣Hb⅐ASHP and k autox,free␣Hb are the first order rates of autooxidation of oxy-␣Hb bound to AHSP and free in solution, respectively. The value of Y complex at equimolar ␣Hb and AHSP is given by, where P 0 is the total concentration of ␣Hb and K d is the equilibrium dissociation constant for ␣Hb binding to AHSP. The rate of auto-oxidation of free oxy-␣Hb is roughly equal to that of Hb dimers, ϳ0.001 min Ϫ1 (16). The dependence of the observed rates of auto-oxidation on total ␣Hb (P 0 ) were fitted to Equations 1 and 2 (Fig. 2C). The best fit was obtained with K d Ϸ 5 M and a limiting rate at high protein concentration for auto-oxidation of oxy-␣Hb⅐AHSP equal to ϳ0.080 min Ϫ1 . This rate is ϳ80-fold greater than that for auto-oxidation of free ␣Hb and Hb dimers and roughly 800-fold greater than that for tetramers. Very similar auto-oxidation rates were reported by Vasseur-Godbillon et al. (17): 0.0007 min Ϫ1 for ␣Hb, 0.0002 min Ϫ1 for HbA, and .05 min Ϫ1 for ␣Hb-AHSP.
The fitted value of the K d for oxy-␣Hb binding to AHSP is significantly higher than expected based on previous estimates by isothermal titration calorimetry (4). This discrepancy may be the result of using a simple one-step model to analyze both types of measurements, when it is clear that auto-oxidation and large secondary conformational changes occur after ␣Hb binds to AHSP. Regardless of the exact interpretation, it is clear the oxy-␣Hb binding to AHSP causes a dramatic increase in the rate of auto-oxidation. Together, the data in Figs. 1 and 2 support a mechanism in which the formation of Fe(III) bis-histidyl ␣Hb⅐AHSP occurs through ␣Hb auto-oxidation facilitated by association with AHSP. As auto-oxidation is accelerated under partially saturated conditions (18), this reaction is likely to be favored at the relatively low oxygen concentrations that occur normally in vivo.
Previously, we suggested that AHSP may stabilize ␣Hb until free ␤ chains become available for the formation of HbA. In support of this, both the Fe(II) and Fe(III) ␣Hb⅐AHSP complexes form apparently tetrameric Hb species upon ␤Hb addition (6). However, the biochemical properties of these HbA-like species are not well studied. We incubated both Fe(II) and Fe(III) ␣Hb⅐AHSP with a freshly prepared excess of oxy-␤Hb and examined the resultant Hb complexes by cellulose acetate electrophoresis, which separates proteins according to size and charge (Fig. 3). For these experiments, we used AHSP fused to glutathione S-transferase, which permits eventual removal of AHSP and AHSP-bound Hbs by absorption to glutathione beads. Similar to AHSP, AHSP-GST binds free ␣Hb and converts it to a more stable Fe(III) bis-histidyl form (not shown). The addition of a 2-fold excess oxy-␤Hb to the Fe(II) ␣Hb⅐AHSP complex produced four bands including the original complex, free AHSP, free ␤Hb, and a new species with similar electrophoretic mobility to HbA (Fig. 3, lane 7). Reaction of ␤Hb with Fe(III) ␣Hb⅐AHSP produced similar bands, except that the newly formed HbA-like complex (slowest migrating species in Fig. 3, lane 8) had slightly reduced mobility. Passage of the AHSP⅐Hb reaction mixtures over glutathione beads removed most free AHSP and AHSP⅐␣Hb complexes (Fig. 3, lanes 9 and 10). However, the two complexes with similar mobility to HbA were unaltered, indicating that they did not contain AHSP. Together, these data indicate that the newly formed slow migrating bands in Fig. 3, lanes 7-10, represent ␣ 2 ␤ 2 heterotetramers. However, the reduced mobility of the species resulting from reaction of ␤Hb with Fe(III) ␣Hb⅐AHSP suggests that this tetrameric Hb differs from normal HbA. For example, the oxidized ␣Hb heme may retain the bis-histidyl conformation acquired during its interaction with AHSP and thus alter the mobility of the Hb tetramer.
To address these possibilities, we studied the biochemical features of the Hb heterotetramers generated from interaction between different forms of ␣Hb⅐AHSP and ␤Hb. First, we examined the oxygen binding properties of Hb resulting from the interaction of ␤Hb with Fe(II) ␣Hb⅐AHSP (Fig. 4A). This Hb exhibited similar electrophoretic mobility to oxy-HbA (Fig. 3, compare lanes 4, 7, and 9) and also displayed spectral properties identical to HbA at room temperature, pressure, and oxygen tension (Fig. 4A, solid line). Upon deoxygenation via three cycles of vacuum and nitrogen purge, a pure deoxygenated HbA spectrum was observed (Fig. 4A, dotted line). Sequential addition of oxygen resulted in progressive oxygenation with P 50 ϭ 1.85 mm Hg, slightly lower than that of normal HbA (ϳ3 mm Hg). The resultant spectrum showed evidence of heme oxidation, as indicated by the ratio of peak absorbances at 542 and 577 nm. Based on deconvolution across the entire visible spectrum, ϳ30% of the final Hb was oxidized (Fig. 4A, dashed line), presumably via auto-oxidation during reoxygenation. However, the majority of AHSP-bound Fe(II) ␣Hb appeared to be recycled into functional HbA.
Strikingly different results were obtained when similar experiments were performed utilizing Hb heterotetramers derived from the Fe(III) ␣Hb⅐AHSP hemichrome complex. The initial spectrum of this Hb species was consistent with a 50:50 mix of oxygenated Fe(II) ␤Hb and Fe(III) ␣Hb heme moieties (Fig. 4B, solid line). Furthermore, deoxygenation of this Hb produced the expected 50:50 mix of deoxygenated Fe(II) and Fe(III) hemes (Fig. 4B, dotted line). However, sequential addition of oxygen produced minimal oxygenation and resulted predominantly in further heme oxidation and heme loss, as evidenced by the loss of absorbance in the 520 -580 nm range and increased absorbance at 630 nm (Fig. 4B, dashed line). As mentioned previously, heme oxidation is autocatalytic, and thus, it is not surprising that a semi-oxidized Hb tetramer is unstable (18).
To examine further whether Fe(III) ␣Hb retains its bis-histidyl hemichrome state in the Hb heterotetramers, we measured their reactivity toward cyanide, which binds Fe(III) heme iron to produce cyanomet Hb with a distinct spectrophotometric signature (Fig. 5A)    14). Cyanide addition to purified Hb (␣FeIII) 2 (␤FeII) 2 prepared by adding oxy-␤Hb to freshly prepared Fe(III) ␣Hb resulted in rapid formation of cyanomet Hb, with complete conversion occurring within 5 min (Fig. 5, A and C). Conversion to cyanomet Hb appeared to occur at 100% of the hemes, suggesting that within this time period, Fe(II) ␤Hb was also oxidized to the Fe(III) form. Wallace et al. showed that anions such as cyanide and azide markedly accelerate auto-oxidation of native hemoglobin (19), and this marked acceleration also occurs for the partially oxidized Hb tetramers shown in Fig. 5A.
In contrast, when cyanide was added to Hb heterotetramers prepared from oxy-␤Hb and Fe(III) ␣Hb⅐AHSP, the generation of cyanomet Hb was much slower (Fig. 5, B and D). After 40 min of incubation with cyanide, less than 20% of the hemes were converted to the cyanomet form. Therefore, in this complex, the ␣ subunits contain hemichromes that inhibit cyanide binding. Moreover, this hemichrome form also inhibits auto-oxidation of oxy-␤ subunits by cyanide, which we observed in the more native Hb valency hybrids shown in Fig. 5A.
We measured the electron paramagnetic resonance (EPR) spectra of frozen samples of free Fe(III) ␣Hb, ␣Hb bound to AHSP, and ␣Hb reconstituted into Hb by addition of CO ␤Hb to examine the spin states of the ␣Hb Fe(III) atom. EPR is a standard method for assessing coordination of Fe(III) in terms of the spin states of the unpaired metal electrons. Concentrated samples of free Fe(III) ␣Hb or Fe(III) ␣Hb⅐AHSP were prepared and stored on ice. A portion of each was mixed with 2-fold excess Fe(II) CO ␤Hb and incubated at 4°C for 30 min to reconstitute ␣ 2 ␤ 2 HbA tetramers. The CO form of ␤Hb was used to prevent oxidation of these subunits and the appearance of ␤ Fe(III) signals. Then, all samples were frozen for EPR analysis of the different ␣Hb⅐AHSP or ␣ 2 ␤ 2 HbA complexes.
As shown in Fig. 6A, there was virtually no g ϭ 6 high spin signal in the 4.5 K spectrum of the Fe(III) ␣Hb⅐AHSP complex. The low spin signal was complex and indicates multiple hemichrome conformers. Peisach and coworkers reported similar complex EPR spectra for Fe(III) HbA treated with urea or salicylate denaturants (20). Thus, AHSP rapidly converts all Fe(III) ␣Hb to low spin complexes, with most of the EPR components being consistent with a bis-histidyl hemichrome. In contrast, free Fe(III) ␣Hb frozen at about 30 min after oxidation showed a significant amount of high spin signal (about 20%, based on comparison to Fe(III) HbA standards). This signal typically reflects unbound or water bound heme iron. About 80% of the sample formed a low spin hemichrome complex, as judged by the amplitude of the g ϭ 6 peak, the appearance of signals in the g ϭ 4 to g ϭ 2 region (Fig. 6A, solid line) and characteristic broad 540 and 570 nm visible absorbance bands (not shown). These findings are consistent with slow formation of hemichrome-like spectra after the oxidation of free human oxy-␣Hb chains [(20) and data not shown].
When CO ␤Hb was added to free Fe(III) ␣Hb to generate a mixed valence (␣FeIII) 2 (CO ␤FeII) 2 Hb, there was complete conversion to a simple 100% high spin EPR spectrum with a dominating peak at g ϭ 6 (Fig. 6B). In contrast, reaction of Fe(III) ␣Hb⅐AHSP with CO ␤Hb produced a hybrid Hb with incomplete conversion of the Fe(III) ␣Hb subunit to the simple aquomet form found in native Fe(III) HbA (Fig. 6C). Based on electrophoresis and gel filtration, all of the Fe(III) ␣Hb subunits were removed from AHSP (not shown). However, the amplitude of the g ϭ 6 signal and the optical spectra (Figs. 4 and 5) indicated that most (ϳ70%) of the Fe(III) ␣Hb subunits that were previously bound to AHSP still had a low spin hemichrome conformation in the hybrid tetramer. This result suggests that the AHSP-mediated rearrangement of ␣Hb to a bis-histidyl structure is retained upon binding to ␤Hb.
Cytochrome b 5 metHb reductase converts oxidized Hb iron to the ferrous state in erythroid cells (21). We investigated whether a similar in vitro enzymatic system could reduce the Hb tetramer derived from Fe(III) ␣Hb⅐AHSP (Fig. 7). We mixed CO ␤Hb with either purified Fe(III) ␣Hb (Fig. 7A) or Fe(III) ␣Hb⅐AHSP (Fig. 7B) to generate semi-oxidized HbA heterotetramers (␣FeIII) 2 (CO ␤FeII) 2 . We used the more stable CO ␤Hb in these experiments to minimize further breakdown of the tetrameric complex from in vitro ␤Hb auto-oxidation, as observed in Fig. 4B. We incubated the Hb tetramers with a reconstituted ferredoxin reductase system (12) in air and monitored reduction of the ␣ subunit by visible light absorbance. Both semi-oxidized Hbs showed heme loss after treatment with the reductase system, as reflected by dampened absorbance through most of the visible spectrum (Fig. 7, A and B). As with any reduction system, it is possible for the ferredoxin reductase to generate accessory oxidants, which could account for the observed Hb instability. Therefore, we repeated these Hb reduction experiments in the presence of the antioxidant enzymes superoxide dismutase and catalase, both of which are relatively abundant in nor-mal erythroid cells (Fig. 7, C and D). Under these conditions, the HbA tetramers derived from either free Fe(III) ␣Hb or Fe(III) ␣Hb⅐AHSP were both entirely reducible and bound oxygen, such that the resultant spectra matched that of HbA. The reaction was slower for the hemichrome containing AHSP-derived Fe(III) ␣Hb. Interestingly, this reduction system with catalase and superoxide dismutase was also capable of reducing the Fe(III) ␣Hb⅐AHSP complex itself (data not shown). Taken together, our findings suggest that Fe(III) ␣Hb derived from the AHSP complex is relatively stable and retains its hemichrome structure upon replacement of the AHSP with ␤Hb.
However, it appears that the resultant tetrameric Hb is functionally competent and that the hemichrome structure is reversible under appropriate reducing conditions.

DISCUSSION
Gene targeting studies demonstrate that AHSP is essential for normal Hb production. AHSP null mice have a compensated hemolytic anemia, and their erythrocytes exhibit increased reactive oxygen species, globin chain precipitates, and shortened life span (1,2). It is likely that these effects are initiated by destabilized ␣Hb that results from the loss of AHSP. Here we used biochemical approaches to study some of the factors that influence ␣Hb-AHSP interactions. These findings refine and extend our model to explain AHSP actions in vivo (Fig. 8). We propose that newly synthesized oxygenated Fe(II) ␣Hb is transiently bound and stabilized by AHSP during erythropoiesis. This interaction alters the structure of ␣Hb reversibly, such that ␤Hb can readily displace AHSP to generate functional HbA under reducing conditions.
If oxy-␣Hb remains bound to AHSP, the structure of the heme pocket is altered further, rapid auto-oxidation occurs, and additional conformational changes lead to the formation of a Fe(III) bis-histidyl complex. This hemichrome form of ␣Hb is resistant to further oxidation and hemin loss, because the sixth coordinate position of heme iron is occupied and unable to participate in chemical reactions that generate reactive oxygen species. However, the stability of AHSP-bound ␣Hb is achieved at an energetic cost, as the heme must be enzymatically reduced to generate functional HbA. This reduction

AHSP and HbA Assembly
OCTOBER 27, 2006 • VOLUME 281 • NUMBER 43 can occur either before or after displacement by ␤Hb. The former would appear to represent a futile redox cycle in the presence of oxygen because of the rapid rate of auto-oxidation of bound oxy-␣Hb.
In vivo, the balance between Fe(II) ␣Hb⅐AHSP and Fe(III) bis-histidyl ␣Hb⅐AHSP is probably controlled by the redox environment, oxygen saturation, and most importantly, ␤Hb availability. Specifically, when ␤Hb is replete, less bis-histidyl complex is formed because AHSP-bound Fe(II) ␣Hb is rapidly recruited into HbA. Alternatively, ␤Hb deficiency, as occurs in ␤ thalassemia, favors conversion of AHSP-bound ␣Hb to the more inert bis-histidyl form. Hence, AHSP has at least two functions: to stabilize ␣Hb transiently during normal HbA synthesis and to sequester ␣Hb in a more stable form during conditions of ␣Hb excess. In future studies, it will be important to characterize further the biochemical features of HbA derived from ␣Hb⅐AHSP complexes.
Genetic evidence in mice supports our model. Loss of AHSP exacerbates anemia and Hb precipitation in ␤ thalassemia, consistent with a physiologic role for bis-histidyl ␣Hb⅐AHSP in ␤Hb deficiency (2). Under these conditions, impaired ␤Hb production and increased oxidative stress favors the accumulation of free ␣Hb subunits; converting these into the bis-histidyl form could provide a mechanism for more effective and prolonged stabilization. The ultimate fate of bis-histidyl ␣Hb⅐AHSP under these circumstances is unknown. One unproven possibility is that this complex is targeted for degradation by proteolytic systems. Additionally, there could be new functions conferred by the bis-histidyl structure. For example, the bis-histidyl form of neuroglobin is proposed to function as a scavenger of reactive oxygen species (22). Ferric bis-histidyl neuroglobin also binds specifically to GDP-bound G␣ and is thereby believed to modulate GDP/GTP signal transduction (23).
Remarkably, loss of AHSP also worsens the ␣ thalassemia silent carrier state (24). 5 Under these circumstances, ␤Hb is present in relative excess so that levels of free ␣Hb are predicted to be low. Despite this, both ␣Hb and ␤Hb are destabilized by AHSP deficiency. This likely reflects destabilization of ␣Hb prior to interaction with ␤Hb, underscoring a physiologic role for the Fe(II) ␣Hb⅐AHSP complex. Hence, AHSP probably has different roles in Hb synthesis according to globin chain synthetic ratios. It may be possible to examine this hypothesis further by measuring the levels of various AHSP⅐␣Hb complexes in normal and thalassemic strains of mice. Additional roles for AHSP might also exist. For example, AHSP binds apo (non heme-bound) ␣ globin protein in vitro (4) and facilitates its folding and expression in Escherichia coli and in eukaryotic cells (1,17). In this capacity, AHSP could interact with the nascent polypeptide chain to promote its stability, folding, or heme insertion during HbA synthesis.