Insights into Hemoglobin Assembly through in Vivo Mutagenesis of α-Hemoglobin Stabilizing Protein*

Background: The α-hemoglobin-stabilizing protein (AHSP) facilitates hemoglobin assembly. Results: AHSP mutations that enhance binding affinity for α-globin or slow its rate of autooxidation in vitro do not affect normal or stressed erythropoiesis in mice. Conclusion: AHSP exhibits robust molecular chaperone activity in vivo even when its biochemical interactions with reduced α-globin are perturbed. Significance: Our findings support new models for AHSP activities in vivo. α-Hemoglobin stabilizing protein (AHSP) is believed to facilitate adult Hemoglobin A assembly and protect against toxic free α-globin subunits. Recombinant AHSP binds multiple forms of free α-globin to stabilize their structures and inhibit precipitation. However, AHSP also stimulates autooxidation of αO2 subunit and its rapid conversion to a partially unfolded bishistidyl hemichrome structure. To investigate these biochemical properties, we altered the evolutionarily conserved AHSP proline 30 in recombinantly expressed proteins and introduced identical mutations into the endogenous murine Ahsp gene. In vitro, the P30W AHSP variant bound oxygenated α chains with 30-fold increased affinity. Both P30W and P30A mutant proteins also caused decreased rates of αO2 autooxidation as compared with wild-type AHSP. Despite these abnormalities, mice harboring P30A or P30W Ahsp mutations exhibited no detectable defects in erythropoiesis at steady state or during induced stresses. Further biochemical studies revealed that the AHSP P30A and P30W substitutions had minimal effects on AHSP interactions with ferric α subunits. Together, our findings indicate that the ability of AHSP to stabilize nascent α chain folding intermediates prior to hemin reduction and incorporation into adult Hemoglobin A is physiologically more important than AHSP interactions with ferrous αO2 subunits.

␣-Hemoglobin stabilizing protein (AHSP) is believed to facilitate adult Hemoglobin A assembly and protect against toxic free ␣-globin subunits. Recombinant AHSP binds multiple forms of free ␣-globin to stabilize their structures and inhibit precipitation. However, AHSP also stimulates autooxidation of ␣O 2 subunit and its rapid conversion to a partially unfolded bishistidyl hemichrome structure. To investigate these biochemical properties, we altered the evolutionarily conserved AHSP proline 30 in recombinantly expressed proteins and introduced identical mutations into the endogenous murine Ahsp gene. In vitro, the P30W AHSP variant bound oxygenated ␣ chains with 30-fold increased affinity. Both P30W and P30A mutant proteins also caused decreased rates of ␣O 2 autooxidation as compared with wildtype AHSP. Despite these abnormalities, mice harboring P30A or P30W Ahsp mutations exhibited no detectable defects in erythropoiesis at steady state or during induced stresses. Further biochemical studies revealed that the AHSP P30A and P30W substitutions had minimal effects on AHSP interactions with ferric ␣ subunits. Together, our findings indicate that the ability of AHSP to stabilize nascent ␣ chain folding intermediates prior to hemin reduction and incorporation into adult Hemoglobin A is physiologically more important than AHSP interactions with ferrous ␣O 2 subunits.
Recent studies have helped to define the biochemical and structural features of AHSP/␣ interactions (15, 18 -23). AHSP is comprised of three antiparallel ␣ helices with 11 unstructured amino acids at the carboxyl terminus (21)(22)(23). AHSP helix 1, helix 2, and the interconnecting loop interact with helices G and H of ␣-globin at a surface that overlaps with the interface for ␤-globin binding. However, ␤ subunits bind ␣ sub-* This work was supported, in whole or in part, by National Institutes of Health units with higher affinity and displace ␣ from AHSP complexes in solution and in vivo (9,10,14,24,25). These observations indicate that AHSP can act as a molecular chaperone to bind nascent ␣-globin and stabilize its folding prior to incorporation into HbA during normal erythropoiesis (12,14,24,25).
An evolutionarily conserved proline residue within loop 1 of AHSP regulates binding to ferrous ␣O 2 and its subsequent autooxidation (18). In the structure of the ASHP⅐met-␣ complex, the cis conformation at the AHSP Asp 29 -Pro 30 peptide bond projects loop 1 toward ␣-globin helix G where AHSP Pro 30 promotes backbone hydrogen bonding interactions. Pro 30 amino acid substitutions in AHSP convert the 29-30 peptide bond to a completely trans conformation, which generally enhances the affinity of AHSP for ferrous ␣CO and ␣O 2 and reduces the rates of autooxidation and subsequent conversion to the bishistidyl form (18).
Because of these large effects on AHSP affinity for reduced ␣ subunits and on the redox and coordination properties of bound ␣ subunits, we chose to examine the in vivo phenotypes of P30A and P30W variants of murine AHSP (18). Although these mutations significantly altered AHSP/␣O 2 interactions in vitro, they surprisingly produced no detectable consequences in mice. These observations led us to re-examine ␣ subunit binding, autooxidation, and hemin loss from the mutant AHSP complexes in vitro. Our combined in vitro and in vivo findings refine and revise current models for AHSP function during erythropoiesis and emphasize the physiological importance of high affinity met-␣ binding to AHSP.

EXPERIMENTAL PROCEDURES
Rates of ␣ Subunit Binding and Release from Pro 30 Variants-Detailed descriptions of measurements of ␣ subunit binding to and release from AHSP variants are given in the accompanying paper by Mollan et al. (42). Time courses for association of reduced and oxidized ␣ subunits with AHSP were measured in an Applied Photophysics PiStar kinetic stopped-flow spectrophotometer (Leatherhead, Surrey, UK), monitoring the intrinsic fluorescence of AHSP. The binding of holo-␣ subunits quenches the fluorescence of AHSP, and its release to ␤ subunits restores the fluorescence emission. A detailed discussion of these signal changes and analysis of the observed time courses are also given in Mollan et al. (42).
Autooxidation Rates of AHSP⅐␣O 2 Complexes-Autooxidation reactions were initiated by diluting a cold concentrated solution of ␣O 2 into an equimolar solution of AHSP in 0.1 M sodium phosphate, pH 7.0 at 37°C and then recording optical absorbance visible spectra. Spectra were recorded using a Cary50Bio instrument (Varian, Inc., Palo Alto, CA). First-order rate constants for autooxidation to the hemichrome met-␣ were obtained by fitting the absorbance change time courses to a single exponential expression using the Microsoft Excel Solver plug-in (Microsoft Corp., Redmond, WA).
Rate of Hemin Loss from Oxidized ␣ Subunits-Hemin dissociation time courses were measured by mixing the ferric form of ␣ or ␣⅐AHSP with the apoglobin form of H64Y/V68F sperm whale myoglobin (Mb) as described previously for assaying hemin dissociation from metHb and metMb (43)(44)(45)(46). Hemin loss from free ferric ␣ generates ␣-apoglobin, which immediately precipitates at 37°C, causing a dramatic increase in solution turbidity. Increasing solution turbidity interferes with monitoring the absorbance time courses at 600 nm for hemin transfer from ␣-globin to the apoMb reagent. To obtain analyzable time courses, the temperature was lowered to 10°C, and 0.6 M sucrose was added to stabilize the apoprotein and inhibit precipitation (46). In addition, the reaction pH was lowered to 5.5 to speed up the reaction at the low temperature, which also helps prevent interference from apo-␣-globin precipitation, particularly for free met-␣ subunits.
Hargrove et al. (44 -46) have shown that hemin loss rates increase markedly at low pH as the proximal histidine becomes protonated. However, the relative effects of mutagenesis and oligomer dissociation for both Hbs and Mbs remain the same as that at higher, more physiological pH values. For the free met-␣Hb and met-␣⅐AHSP complexes, the hemin loss reaction mixtures contained 6 M met-␣, 40 M H64Y/V68F apoMb, 30 M potassium ferricyanide to ensure oxidation, and 6 M either WT, P30A, or P30W AHSP. Spectra were recorded every minute for ϳ12 h, and the increase at 600 nm is due to formation of the "green" H64Y/V68F holo-metMb complex as the ␣ subunits lose hemin. Data were plotted as the normalized absorbance change at 600 nm versus time.
Animals-Ahsp P30W-and P30A-targeting constructs were generated by modification of a previously described Ahsp-targeting construct (supplemental Fig. 1) (10). Site-directed mutagenesis was used to replace the CCT codon for proline (Pro) with either GCT for alanine (P30A) or TGG for tryptophan (P30W). The targeting plasmids were linearized with SacI and electroporated into 129SV embryonic stem cells. Clones with stable integration of the targeting vector were selected with G418 and ganciclovir and further screened for homologous recombinants by Southern blotting and karyotyping. Selected clones were injected into C57BL/6 blastocysts to produce chimeric mice. These were bred with C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME) to produce Ahsp P30A/ϩ or Ahsp P30W/ϩ heterozygotes. Neomycin selection cassettes were removed by interbreeding with E2A-Cre mice (47). Mice were genotyped using the primer set AHSP-KI-F (5Ј-AGA AAC GGG GGA TGC AAA TCC AAA ACA AAG-3Ј) and AHSP-KI-R (5Ј-CAA GTT CAC CCA GTC ATG AAC CAC AAT CAC-3Ј), generating a 190-base pair product for the wild-type allele and a 236-base pair product for the mutant allele. All mice used for experiments were backcrossed onto a C57BL/6 background for five to seven generations. Ahsp Ϫ/Ϫ mice were generated as described previously (10). ␤-Thalassemic (␤-globin Th3/ϩ ) mice were kindly provided by Oliver Smithies (University of North Carolina, Chapel Hill, NC) (48). The Ahsp allele is genetically linked to the ␤-globin (Hbb) locus (11); Ahsp/␤-globin Th3/ϩ double mutant mice were generated as described previously (11). Animal protocols were approved by the Institutional Animal Care and Use Committee of The Children's Hospital of Philadelphia (Philadelphia, PA).
Hematological Analysis-Mice were analyzed at 2 and 8 months of age. Blood was collected by submandibular bleeding, anticoagulated with EDTA, and analyzed on a Hemavet HV950FS analyzer (Drew Scientific, Dallas, TX). Reticulocyte counts were done using Retic-COUNT reagent according to the manufacturer's protocol using a FACSCalibur flow cytometer (BD Biosciences). Flow cytometry data were acquired using equipment maintained by the Flow Cytometry Core Laboratory of The Children's Hospital of Philadelphia Research Institute.
AHSP Expression Analysis-Erythrocytes were lysed with 40 volumes of water with 1 mM dithiothreitol and 1:500 protease inhibitor mixture (Sigma-Aldrich). Bone marrow erythroid precursors were purified from mice of the indicated genotypes by immunomagnetic positive selection using the EasySep PE Selection Kit (StemCell Technologies, Vancouver, British Columbia, Canada) with Ter119-PE antibody (Biolegend, San Diego, CA) according to the manufacturer's instructions (51). Protein lysates from purified erythroblasts were prepared using radioimmune precipitation assay buffer with 1 mM DTT and 1:500 protease inhibitor mixture (Sigma-Aldrich). Protein concentrations were determined using a Pierce bicinchoninic acid assay kit (Thermo Scientific, Rockford, IL), and 10 g of protein were resolved by 15% SDS-PAGE. After transfer to 0.2-m PVDF membranes (Whatman), AHSP was detected using a rat polyclonal antibody raised against purified mouse AHSP protein and HRP-conjugated goat anti-rat secondary antibody. HRP-conjugated anti-␤-actin antibody was used as a loading control. Western blots were developed using SuperSignal West Pico chemiluminescent reagent. Western blotting secondary antibodies, markers, and reagents were obtained from Thermo Scientific.
Isoelectric Focusing of 35 S-Labeled Reticulocyte Extracts to Detect Globins and AHSP⅐␣Hb Complexes-10 l of freshly collected erythrocytes and reticulocytes (normalized to reticulocyte percentage and hematocrit) were incubated in methionine/cysteine-free DMEM (Invitrogen), 2 mM L-glutamine, 10% dialyzed FBS, and 0.1 mM non-essential amino acids at 37°C for 1 h. 50 Ci of 35 S-labeled methionine and cysteine (Perkin-Elmer Life Sciences) were added, and the cells were labeled for 10 min at 37°C. The cells were washed with PBS and exposed to carbon monoxide to stabilize ferrous heme groups. The cells were lysed in 5 volumes of water, and the lysate was clarified by centrifugation. Lysates were resolved by isoelectric focusing (pH 6 -8) on precast agarose gels from the Hemoglobin Test kit on an LKB 2117 Multiphor apparatus at 1500 V for 50 min (PerkinElmer Life Sciences). For autoradiography, gels were fixed in 10% trichloroacetic acid for 10 min, washed in deionized water, and dried. Images were acquired using a phosphor screen and Storm 865 scanner (GE Healthcare). Bands were quantified using National Institutes of Health ImageJ software. For analyzing band globin and AHSP content, isoelectric focusing gel bands were excised, boiled in 2ϫ Laemmli sample buffer (4% SDS, 20% glycerol, 10% 2-mercaptoethanol, 0.004% bromphenol blue, 0.125 M Tris HCl, pH 6.8; Sigma-Aldrich), and proteins were resolved by 15% SDS-PAGE. Gels were either dried for autoradiography or used for mouse AHSP Western blotting as described above.
Statistical Methods-Statistical analysis was performed using GraphPad Prism 4.0 software (GraphPad Software, La Jolla, CA). All group comparisons were done using one-way analysis of variance.
Sequence Alignments-AHSP protein sequences were obtained using the Basic Local Alignment Search Tool (BLAST) at the NCBI using the human AHSP sequence (NCBI accession number NM_016633.2) as the query, and using the Ensembl genome browser. After manual inspection for sequence quality, 31 sequences were trimmed to include only helices 1 and 2 and aligned using the ClustalW algorithm with Strap multiple sequence alignment software (52). 30 Mutations Alter ␣ Chain Binding, Autooxidation, and Hemin Loss-The evolutionarily conserved proline at position 30 in loop 1 of AHSP contacts helix G of ␣-globin at the edge of the protein-binding interface (Fig. 1). To examine the function of this amino acid, Gell et al. (18) made a series of mutations that convert AHSP Pro 30 to Gly, Ala, Val, Phe, and Trp and screened the resultant protein interactions with ␣-globin. Except for the P30G mutant, all AHSP variants showed enhanced affinity (lower K D values) for ferrous ␣CO or ␣O 2 as measured by isothermal titration calorimetry. All of these mutants also showed 3-4-fold decreased rates of autooxidation and concomitant hemichrome formation (18).

AHSP Pro
In concert with our mutagenesis studies of the endogenous Ahsp gene, the in vitro effects of P30A and P30W AHSP substitutions were reevaluated for recombinantly produced proteins. Direct measurements of the association and dissociation rate constants for the binding of these AHSP variants to native ␣CO and met-␣ subunits were made using stopped-flow fluorescence methods as described in the accompanying paper by Mollan et al. (42) and as described previously (25). Direct comparisons of the speeds of ␣(Fe 2ϩ )CO and met(Fe 3ϩ )␣ binding to WT, P30A, and P30W AHSP are shown in Fig. 2A. Remarkably, the rates of binding of reduced and oxidized ␣ subunits to AHSP were the same and were unaffected by mutation of Pro 30 to either Ala or Trp. The only major difference is that WT AHSP showed a secondary slow phase when binding reduced ␣ subunits. This slow phase for WT AHSP appears to be due to a cis to trans isomerization of Pro 30 (see accompanying paper (42)).
In contrast, the rate of dissociation of ␣ subunits from the AHSP complexes in the presence of excess ␤ subunits depended markedly on both the oxidation state of the ␣-hemes and in the case of reduced ␣CO on the amino acid at position 30. As shown in Fig. 2B, reduced ␣CO was released from WT and P30A AHSP with half-times of roughly 2-4 s, whereas the half-time for dissociation from P30W AHSP was roughly 25-fold larger. In contrast, WT and both Pro 30 AHSP mutants released oxidized ␣ subunits very slowly with half-lives on the order of 200 -400 s, similar to that for the release of reduced ␣ subunits from the P30W mutant (Fig. 2B).
As a result of these differences in dissociation rate constants, the equilibrium dissociation constant (K D ) for ␣ subunit binding depended strongly on the heme oxidation state. When the heme was reduced, the P30W AHSP mutant showed a ϳ30-fold higher affinity for ␣ subunits than WT and P30A AHSP, which showed K D values of roughly 20 nM. However, when the ␣ subunits were oxidized, WT AHSP and the two Pro 30 mutants showed the same high affinity with K D values equal to 0.2-0.6 nM. Similar relative K D differences for the AHSP variants were observed by isothermal titration calorimetry, although the isothermal titration calorimetry values were all 5-10-fold larger than those calculated from the association and dissociation rate constants. The cause of this discrepancy is discussed in Mollan et al. (42). The remarkably high affinity of AHSP for ferric ␣ subunits and the lack of significant differences between the Pro 30 mutants were discovered as a result of our efforts to understand the results of the in vivo mutagenesis studies.
Autooxidation of WT, P30A, and P30W AHSP⅐␣O 2 complexes was also re-examined. As shown in Fig. 3, the rates of oxidation were 3-4-fold smaller than that for the wild-type complex as was observed by Gell et al. (18), but even these rates were still 2-3-fold greater than that for free ␣O 2 ( Table 1). As estimated by increases in turbidity, the overall resistances of the AHSP⅐␣O 2 complexes to precipitation were roughly the same for the wild type, P30A, and P30W AHSP (data not shown) even though the latter mutant⅐␣O 2 complexes autooxidized more slowly. This observation and the lack of phenotype observed for mice expressing the mutant Ahsp genes (described in subse-quent sections) led us to examine the rates of hemin loss from the oxidized hemichrome met-␣⅐AHSP complexes.
Time courses for hemin loss were obtained by mixing AHSP⅐met-␣ complexes with excess H64Y/V68F apoMb as a hemin-scavenging reagent (Fig. 4). The ultrastable apoMb mutant turns "green" as indicated by the appearance of a large absorbance peak at 600 nm when it accepts hemin from a donor (44 -46). Unexpectedly, the rates of hemin loss from the P30A and P30W AHSP⅐met-␣ complexes were 2-7-fold greater than that from the wild-type AHSP⅐met-␣ complex ( Fig. 4 and Table  1). Unfortunately, the apo-␣-globin product rapidly precipitates at 37°C at pH 7, particularly in the absence of AHSP, and the resultant turbidity increases make quantitative analyses of hemin dissociation time courses impossible. However, the relative order of the initial rates of the absorbance increases at 600 nm for the wild-type and two variant AHSP⅐met-␣ complexes under these more physiological conditions was the same as the order of the rate constants determined quantitatively at low temperature and pH (not shown).
As shown in Fig. 4B, binding to wild-type AHSP enhanced the rate of hemin loss from met-␣ by a factor of 2. Binding to the Pro 30 AHSP mutants accelerated hemin dissociation even further. These increases in k Ϫhemin seem to suggest increased exposure of the heme pocket to water in the AHSP complex. Liong et al. (53) have shown that exposing the proximal histidine to water causes dramatic (ϳ1000-fold) increases in the rate of hemin loss from metMb variants. Thus, even partial unfolding of the heme pocket in the AHSP⅐met-␣ complexes to form the hemichrome would be expected to increase the rate of hemin loss markedly.
One key result in Table 1 is that the net resistance of the variant AHSP⅐␣O 2 complexes to apoglobin formation, unfolding, and precipitation may be the same for all three variants. In these displacement reactions, the fluorescence of AHSP increases as ␣ subunits dissociate from the complex, and the normalized increases were calculated as 1 Ϫ (F t ϪF ϱ )/(F 0 ϪF ϱ ). A detailed description of the reactions and analyses to obtain association (kЈ AHSP ) and dissociation (k AHSP ) rate constants is given in the accompanying paper (42). Finally, the fluorescence changes for the P30W AHSP mutant were roughly 2 times greater than those for either wild-type or P30A AHSP due to the "extra" Trp side chain. FIGURE 3. Autooxidation of AHSP⅐␣O 2 complexes. A, absorbance spectra of 20 M wild-type AHSP⅐␣O 2 complexes over time during the autooxidation process. B, time courses for free ␣O 2 subunits and AHSP⅐␣-globin complexes at 577 nm during the autooxidation process where the normalized absorbance changes were computed as (A t Ϫ A ϱ )/(A 0 ϪA ϱ ) where A t , A ϱ , and A 0 equal the absorbance at time t, infinite time, and time 0, respectively. AHSP⅐␣O 2 concentrations were fixed at 20 M in 100 mM sodium phosphate buffer, pH 7.0 at 37°C. Scans were taken every minute for ϳ24 h. Autooxidation of ␣O 2 subunits in the absence of AHSP was accompanied by an increase in solution turbidity due to precipitation at this temperature.
Even though the Pro 30 mutant AHSP⅐␣O 2 complexes autooxidized more slowly, the resultant AHSP⅐met-␣ complexes lost hemin more rapidly so that net rates of apo-␣-globin appearance are probably similar for the Pro 30 mutants and wild-type AHSP.
Generation of AHSP Pro 30 Substitutions in Mice-We created Ahsp P30A and Ahsp P30W missense mutations in the endogenous murine gene (supplemental Fig. 1A). We reasoned that comparing hematologic parameters in these mutant strains would allow us to distinguish between effects caused by altered AHSP affinity for ␣O 2 and reduced rates of autooxidation and bishistidyl conversion. Mutant animals were born at normal Mendelian ratios, similar to what we observed previously for Ahsp Ϫ/Ϫ mice (10,11). Erythroblast AHSP protein levels were normal in the mutant animals (supplemental Fig. 1B).
Analysis of AHSP⅐␣-Globin Complexes in Vivo-Endogenous AHSP⅐␣-globin complexes are more easily detected under conditions of ␣-globin excess (12,48). Thus, we crossed the Ahsp P30W and Ahsp P30A mutant mice onto a ␤-thalassemia intermedia background (␤-globin Th3/ϩ ) (48) and examined the formation of mutant AHSP⅐␣-globin complexes by isoelectric focusing of radiolabeled reticulocyte extracts (12). We detected several globin-containing bands that could potentially be an AHSP⅐␣-globin complex (supplemental Fig. 2A). We used Western blotting of proteins isolated from these bands to iden-tify the band that contained both ␣-globin and AHSP (supplemental Fig. 2B), allowing us to quantify the relative amount of labeled ␣-globin in the complex in subsequent experiments. Accumulation of the AHSP⅐␣-globin complex was significantly increased in Ahsp P30W/P30W ␤-globin Th3/ϩ mice consistent with biochemical studies showing that the mutant protein had increased affinity for ␣-globin (Fig. 5). AHSP⅐␣-globin complexes were present at wild-type levels in Ahsp P30A/P30A ␤-globin Th3/ϩ mice. Thus, it was possible to increase AHSP⅐␣-globin complex formation in vivo through targeted mutagenesis of Ahsp gene as predicted from the higher affinity of P30W AHSP for reduced ␣O 2 or ␣CO in vitro ( Fig. 5B and Table 1). However, our assay did not allow us to measure the fraction of the AHSP⅐␣-globin complexes, which were in the bishistidyl hemichrome form in vivo.
Hematological Features of Ahsp P30A and Ahsp P30W Knock-in Mice-We first compared baseline erythropoiesis in AHSP Pro 30 -substituted and control mice (Table 2A and supplemental Table 1). Ahsp Ϫ/Ϫ mice exhibited reticulocytosis and elevated red cell distribution width consistent with mild hemolysis. These parameters were entirely normal in Ahsp P30A and Ahsp P30W mutant mice (heterozygotes and homozygotes) at ages 2 ( Table 2) and 8 months (not shown). Ahsp P30A and Ahsp P30W mouse erythrocytes exhibited normal morphology, whereas Ahsp Ϫ/Ϫ erythrocytes exhibited shape abnormalities and eosinophilic inclusions (Refs. 10 and 11 and data not shown). Of note, Ahsp mutant mice used in this study were backcrossed onto a C57BL/6 genetic background. In this pure strain, loss of Ahsp caused minimal anemia in contrast to Ahsp Ϫ/Ϫ mice with a mixed 129SV/C57BL/6 genetic background (10,11).

TABLE 1 In vitro properties of AHSP P30A and P30W variants
The bimolecular rate constants (kЈ AHSP ) for ␣CO binding to and the first-order dissociation rate constants for ␣CO release (k AHSP ) from wild-type AHSP and Pro 30 mutants were measured at pH 7.0 at 25°C in rapid mixing experiments (see Fig. 2 and Mollan et al. (42)). The equilibrium dissociation constants for ␣CO or met-␣ binding to AHSP under these conditions were calculated from k AHSP /kЈ AHSP . The first-order rates of autooxidation (k autox ) of ␣O 2 free in solution and bound to the AHSP variants were measured as described in Fig. 3. The first-order rates of hemin dissociation (k Ϫhemin ) were measured as described in Fig. 4 (18). These measurements are consistently higher than the values calculated from the rate constants. This discrepancy is discussed in the accompanying paper by Mollan et al. (42).
Compared with control animals, the hematocrits of Ahsp Ϫ/Ϫ mice declined more rapidly after treatment with 5-FU, consistent with a reduced circulating lifespan of mutant erythrocytes (11). During subsequent recovery, there was increased reticulocytosis in the Ahsp Ϫ/Ϫ mice compared with controls. In contrast, Ahsp P30W/P30W and Ahsp P30A/P30A mice showed normal kinetics of erythrocyte decline and reticulocyte recovery after treatment with 5-FU. Together, these studies indicate that hemoglobin assembly and function are not impaired significantly by AHSP P30A or P30W substitution either at baseline or under conditions of erythropoietic stress despite substantial differences in ␣-globin binding affinity, induced rates of autooxidation, and hemin loss measured in vitro. Effects of AHSP Pro 30 Substitutions on ␤-Thalassemia Intermedia-Loss of AHSP exacerbates ␤-thalassemia intermedia (genotype ␤-globin Th3/ϩ ) in mice (11). We postulated previously that in ␤-thalassemia AHSP may reduce the toxicity of excess free ␣-globin by converting it to a bishistidyl form, which exhibits a reduced rate of ROS generation (26). To test this model, we intercrossed ␤-globin Th3/ϩ mice with animals harboring AHSP Pro 30 substitutions. We reasoned that the AHSP P30A and AHSP P30W substitutions might have a greater impact on erythropoiesis in ␤-thalassemia, where ␣-globin is present in excess. The reduced rates of autooxidation of ␣O 2 and concomitant conversion to the hemichrome form when bound to the Pro 30 AHSP mutants might aggravate ␤-thalassemia by increasing ROS generation in a manner similar to what was observed upon Ahsp gene ablation (11). Alternatively, the increased affinity of the AHSP mutants for ferrous ␣-globin (Table 1 and Fig. 5) might selectively enhance the ability of AHSP to protect against excess ␣-globin degradation and precipitation in ␤-thalassemia, particularly for the P30W mutant, which bound ␣O 2 about 30-fold more tightly. AHSP proteins were expressed at similar levels in erythroid precursors of Ahsp ϩ/ϩ , Ahsp P30W/P30W , and Ahsp P30A/P30A mice with concomitant ␤-thalassemia (supplemental Fig. 3). Erythroid indices were indistinguishable between thalassemic mice expressing wild-type AHSP or either Pro 30 mutation at 2 and 8 months (Table 2 and supplemental Tables 2 and 3). Moreover, the levels of erythroid membrane-associated ␣-globin precipitates were similar in ␤-thalassemic mice expressing wild-type AHSP and Pro 30 mutations in contrast to those lacking AHSP where ␣-globin precipitation was increased (Fig. 8 and Ref. 11). Thus, the P30W and P30A substitutions have no detectable positive or negative effects on the ability of AHSP to protect against accumulation of toxic free ␣-globin in ␤-thalassemia.

DISCUSSION
Prior studies indicate two potential functions for AHSP during normal and pathologic erythropoiesis (8,26,63). First, AHSP acts as a molecular chaperone to bind and stabilize nascent ␣-globin folding intermediates prior to reaction with ␤-globin and assembly into HbA (Fig. 9). This function is gov-   MARCH 30, 2012 • VOLUME 287 • NUMBER 14 erned by the rates of holo-and apo-␣-globin binding to and release from AHSP (18,25,42,64). Thus, AHSP/␣-globin interactions during HbA synthesis must be optimized to promote ␣-globin stability yet not interfere significantly with release of ␣-globin for binding to ␤-globin during HbA assembly.

In Vivo Analysis of AHSP Mutants
Second, AHSP forms stable complexes with excess free ␣-globin under conditions of ␤-globin deficiency to reduce the toxicity associated with ␣-globin oxidation, hemin loss, and precipitation. Mitigation of excess ␣-globin toxicity was postulated to occur in part through AHSP-mediated conversion of bound ␣ subunits to a bishistidyl, hemichrome state, which limits ROS production. Biochemical studies support this model, although it has been difficult to test in vivo. Pro 30 regulates both the affinity of AHSP for reduced ␣-globin and the rate of conversion of bound ␣O 2 to the ferric bishistidyl met-␣ form (i.e. autooxidation). To examine the importance of the latter properties in vivo, Pro 30 missense mutations (P30A and P30W) were introduced into endogenous murine Ahsp loci. As shown in Table 1, these mutations significantly altered the biochemical properties of AHSP in vitro. However, in contrast to the complete loss of AHSP, the P30A and P30W substitutions produced no detectable in vivo effects on hemoglobin synthesis or stability at base line, during erythropoietic stress, or under conditions of ␣-globin excess (␤-thalassemia) in mice.
This lack of phenotype for the Ahsp mutants requires the re-evaluation of current models for the physiological functions of AHSP during erythropoiesis and led us to re-examine the hypothesis that a major function of AHSP is to stabilize excess ␣O 2 by oxidizing it to a hemichrome species to prevent ROS generation. More generally, our work also highlights the importance of using in vivo approaches to test hypotheses generated by in vitro biochemical experiments, which in this case suggested that there should be marked physiological differences between the P30A and P30W variants and wild-type AHSP.
An alternative and perhaps complementary function of AHSP is to act as a molecular chaperone for ␣-globin during HbA assembly. This role is supported by kinetic studies showing that both ferrous and ferric holo-␣ bind ϳ20 times more rapidly to AHSP than to ferrous ␤ subunits (14,25,42). However, ␤ subunits have a 10,000-fold higher affinity for ␣-globin than AHSP due to an extremely small ␣1␤1 dissociation rate constant, which is on the order of 10 Ϫ5 s Ϫ1 (25,65). In comparison, the rate of ␣ dissociation from AHSP is 10 2 -10 4 larger for all the AHSP variants listed in Table 1. These kinetic results suggest that newly synthesized ␣-globin either with or without hemin binds initially to AHSP and subsequently is released for interaction with ␤-globin to form HbA (Fig. 9). At high AHSP and ␤-globin concentrations, the rate, but not the extent, of HbA assembly could be limited by the rate of release of ␣-globin from AHSP ( Fig. 9) (25).
The AHSP concentration in late stage erythroblasts is ϳ0.1 mM (10). The concentration of HbA subunits is ϳ10 -20 mM, although the concentrations of free, non-tetrameric ␣ and ␤ subunits are unknown (9, 10). One question is whether the ratio of AHSP to the pool of newly synthesized free ␤ chains is sufficiently low for optimal speed of HbA assembly and how easily this process can be perturbed under physiological conditions. For example, increased AHSP concentration and/or enhanced affinity for ␣-globin could interfere with the rate of HbA tetramer formation. However, the ϳ30-fold increased affinity of AHSP P30W for ␣O 2 (Table 1) and the increased levels of the complex in reticulocytes (Fig.  5B) did not appear to impair the net amount of HbA synthesis in mouse erythrocytes. Moreover, 10-fold overexpression of AHSP does not reduce steady-state HbA concentrations in erythroid cells even during ␤-globin deficiency (66). Thus, experimental manipulation of both AHSP levels and its affinity for ␣ subunit binding demonstrated that the ratio of AHSP to newly synthesized ␤-globin chains is not restrictive for HbA synthesis in mice.
All of these results suggest that within erythroblasts, the function of AHSP as a molecular chaperone for HbA assembly retains considerable plasticity to tolerate variations in ␣-globin interaction rates and affinities. The limitation appears to be the ratio of the rates of subunit denaturation versus formation of the initial stable oligomeric states, i.e. ␣⅐AHSP, ␤ 4 , and ␣␤ complexes (Fig. 9). Once these homo-and heterocomplexes are formed, the subunits are stable enough to quantitatively reassemble into the holo-HbA tetramer, which is an effectively irreversible process.  The formation, spectral properties, and structure of AHSP⅐bishistidyl met-␣ complexes have been examined extensively in vitro (18,21,22,28). The physiological function of this complex in vivo was examined in the current study by introducing endogenous Ahsp mutations that altered its rate of ␣O 2 autooxidation. AHSP amino acid substitutions that reduced this rate by 3-4-fold did not alter hemoglobin formation during normal erythropoiesis. Moreover, these mutations did not worsen ␤-thalassemia as might be expected if AHSP neutralizes the toxicity of free ␣O 2 by inducing formation of a stable and ROS-inhibiting bishistidyl structure. One explanation is that the lower autooxidation rates of the mutant AHSP⅐␣O 2 complexes are still sufficient to generate a stable bishistidyl complex that prevents ROS formation and ␣-globin precipitation in vivo.
It seems more likely that the toxicity of excess ␣Hb is due to hemin loss, apoglobin denaturation, and Heinz body formation, which is a major source of ROS in red cells. This explanation is supported by the lack of effect of the Pro 30 mutations on erythropoiesis in vivo. Although the Pro 30 mutant AHSP⅐␣O 2 complexes autooxidized more slowly than the wild-type complex ( Fig. 3 and Table 1), the mutant AHSP⅐met-␣ complexes lost hemin much more rapidly (Fig. 4). Thus, the resistance of the initial ferrous ␣O 2 complex to apo-␣-globin formation is roughly the same for wild-type AHSP and the two variants, which could account for a lack of phenotype in the Ahsp mutant mice.
Finally, our findings and those presented in the accompanying paper (42) are also relevant to an unsolved problem concerning HbA synthesis. The redox state of heme during its insertion into apoglobins is not known. We suggest that during normal erythroid HbA assembly ferric heme is incorporated into AHSP-bound apo-␣-globin, rapidly forming a hemichrome holo-␣ species with the heme pocket partially collapsed (Fig. 9). Subsequently, this intermediate is reduced by endogenous methemoglobin reductases to generate ferrous ␣-globin, which is released rapidly from AHSP for binding to ␤-globin.
Several lines of evidence support this model. Free ferrous heme (Fe 2ϩ ) autooxidizes rapidly in milliseconds to hemin (Fe 3ϩ ) even at low levels of O 2 . Gell et al. (18) and others (64) have shown that AHSP binds and stabilizes apo-␣-globin in vitro. Labeling studies in reticulocytes suggest that hemoglobin is initially synthesized with the heme in a ferric state (67, FIGURE 9. Model for AHSP/␣-globin interactions and hemoglobin assembly during erythropoiesis. After translation, newly formed apo-␣-globin subunits will precipitate rapidly unless they bind to their partner ␤ subunit, AHSP, or free heme, which is in the ferric state (hemin). Both apo-and holo-␤-globin subunits can self-assemble to form tetramers (HbH). Even in Ahsp Ϫ/Ϫ mice, almost normal concentrations of HbA are present in mature red cells, demonstrating that the simple assembly pathway indicated by the lower rung of reactions occurs readily. However, AHSP can facilitate assembly by stabilizing apo-and/or holo-␣globin folding intermediates and allow rapid displacement of AHSP by ␤-globin (solid pale blue arrow). There is a buildup of AHSP⅐␣-globin complexes in ␤-thalassemia in addition to ␣-globin precipitates. In excess, ␤-globin can self-chaperone by forming ␤ 4 tetramers (HbH), as is observed in ␣-thalassemia (pale blue dashed arrow). 68). As shown in Fig. 2B and Table 1, AHSP bound ferric met-␣-globin with a 100-fold higher affinity than that for various reduced forms of ␣ chains (18,42). Culbertson and Olson (69) have shown that hemichrome species similar to that of the met-␣⅐AHSP complex are key folding intermediates in the formation of holo-Hbs and holo-Mbs. Hemichrome ␣-globin bound to AHSP can be reduced anaerobically in vitro to generate fully folded, pentacoordinate deoxy-␣ with no evidence of a hemochrome species (bishistidyl coordination of ferrous iron) (28,42,69). In vitro, bishistidyl hemoglobins have faster reduction kinetics than pentacoordinate forms (70), suggesting that the ␣-globin hemichrome intermediate state induced by AHSP may facilitate reduction of the ferric heme iron in vivo.
Our new in vitro data also support this model of AHSP chaperone activity involving stabilization of an ␣ ferric, hemichrome folding intermediate. The AHSP P30A and P30W mutations did not significantly affect the uptake and release of met-␣globin (see Fig. 2 and Table 1). Thus, although ␣O 2 autooxidation is impaired, the Pro 30 mutations may have no detectable effects in vivo because the key function of AHSP is to prevent the precipitation of apo-␣-globin conformers, facilitate uptake of hemin, stabilize the holo-met-␣ folding intermediate, and prevent release for binding to ␤-globin until the ␣-globin is reduced. These biochemical chaperone properties of AHSP have been exploited to enhance ␣-globin synthesis in bacterial expression systems (20) and to quantify the pool of free ␣ chains in ␤-thalassemia (19). Although in vitro experiments have been valuable for establishing models of hemoglobin assembly and the role of AHSP, here we show that in vivo studies are required to test and refine these models. Further mutagenesis and in vivo studies will improve our understanding of AHSP/␣-globin interactions and enhance the use of AHSP as a diagnostic and biotechnological tool.