Biophysical Characterization of the α-Globin Binding Protein α-Hemoglobin Stabilizing Protein*

α-Hemoglobin stabilizing protein (AHSP) is a small (12 kDa) and abundant erythroid-specific protein that binds specifically to free α-(hemo)globin and prevents its precipitation. When present in excess over β-globin, its normal binding partner, α-globin can have severe cytotoxic effects that contribute to important human diseases such as β-thalassemia. Because AHSP might act as a chaperone to prevent the harmful aggregation of α-globin during normal erythroid cell development and in diseases of globin chain imbalance, it is important to characterize the biochemical properties of the AHSP·α-globin complex. Here we provide the first structural information about AHSP and its interaction with α-globin. We find that AHSP is a predominantly α-helical globular protein with a somewhat asymmetric shape. AHSP and α-globin are both monomeric in solution as determined by analytical ultracentrifugation and bind each other to form a complex with 1:1 subunit stoichiometry, as judged by gel filtration and amino acid analysis. We have used isothermal titration calorimetry to show that the interaction is of moderate affinity with an association constant of 1 × 107 m −1 and is thus likely to be biologically significant given the concentration of AHSP (∼0.1 mm) and hemoglobin (∼4 mm) in the late pro-erythroblast.

␣-Hemoglobin stabilizing protein (AHSP) is a small (12 kDa) and abundant erythroid-specific protein that binds specifically to free ␣-(hemo)globin and prevents its precipitation. When present in excess over ␤-globin, its normal binding partner, ␣-globin can have severe cytotoxic effects that contribute to important human diseases such as ␤-thalassemia. Because AHSP might act as a chaperone to prevent the harmful aggregation of ␣-globin during normal erythroid cell development and in diseases of globin chain imbalance, it is important to characterize the biochemical properties of the AHSP⅐␣globin complex. Here we provide the first structural information about AHSP and its interaction with ␣-globin. We find that AHSP is a predominantly ␣-helical globular protein with a somewhat asymmetric shape. AHSP and ␣-globin are both monomeric in solution as determined by analytical ultracentrifugation and bind each other to form a complex with 1:1 subunit stoichiometry, as judged by gel filtration and amino acid analysis. We have used isothermal titration calorimetry to show that the interaction is of moderate affinity with an association constant of 1 ؋ 10 7 M ؊1 and is thus likely to be biologically significant given the concentration of AHSP (ϳ0.1 mM) and hemoglobin (ϳ4 mM) in the late pro-erythroblast.
Mammalian hemoglobin, the red blood cell oxygen transport molecule, is a tetramer of two ␣and two ␤-globin chains. Exquisitely coordinated expression of these ␣and ␤-globin chains is required during erythropoiesis to generate high concentrations of hemoglobin, without production of either chain in excess; any disruption of normal globin gene expression patterns can lead to serious human disease (1,2). One such disease is ␤-thalassemia, a common genetic disorder caused by mutations in one or more of the ␤-globin gene loci that result in reduced ␤-globin production. In addition to the direct effects of reduced ␤-globin synthesis, many of the symptoms of this disorder appear to be consequences of the resulting cytotoxic buildup of free ␣-globin (1, 2). Free ␣-globin is highly unstable and readily precipitates, damaging membrane structures and triggering the apoptotic cell death of erythroid precursors (1). The effects of dysregulated expression of individual globin chains can be severe, and consequently, it has long been thought that additional factors within the cell may assist with the processing of free globin chains and their assembly into mature hemoglobin (3). The identification of any such factors proved elusive for a long time.
Recently however, Kihm et al. (4) identified ␣-hemoglobin stabilizing protein (AHSP), 1 a small 102-residue protein that may act to neutralize the harmful effects of any free ␣-globin generated during either normal erythropoiesis or in circumstances of disease (4). Expression of the AHSP gene is strongly up-regulated by the essential erythroid transcription factor GATA-1 such that AHSP accumulates to relatively high concentrations (ϳ0.1 mM) in late erythroid precursor cells (4,5). AHSP mRNA is present specifically in all hematopoietic tissues of the fetus and adult mouse, consistent with a role in the regulation of hemoglobin production throughout pre-and postnatal life.
AHSP specifically binds to the ␣-chain of hemoglobin, but not to the ␤-chain or to tetrameric hemoglobin, making it an ideal candidate for an ␣-globin-specific chaperone. Consistent with a role for AHSP in regulating coordinated globin expression, gene-targeting studies in mice showed that ablation of AHSP function leads to erythrocyte abnormalities that are also observed in ␤-thalassemia. The staining of erythrocytes from these mice with crystal violet reveals that they contain inclusion bodies of denatured hemoglobins, know as Heinz bodies (4). AHSP also prevents the precipitation of ␣-globin both in vitro and in COS cells, further supporting the idea that AHSP may prevent pathological aggregation and precipitation of ␣-globin in vivo.
Little further is currently known about the action of AHSP or the molecular details of its interaction with ␣-globin. Interestingly, however, the protein was previously identified (and named EDRF, for erythroid differentiation related factor) as a marker for transmissible spongiform encephalopathies (6), although the significance of this finding is currently unclear. Nothing is known about the physical or conformational properties of AHSP. Little information is available from sequence comparisons, because AHSP displays no clear homology to any protein of known structure. Here, we present the first biophysical analysis of AHSP and its specific interaction with ␣-globin. We demonstrate that AHSP is primarily ␣-helical in conformation, probably with an extended C-terminal region, and that it is slightly elongated. We also show that AHSP is monomeric in solution at concentrations of up to 1 mM and that the AHSP⅐␣-globin complex has a 1:1 stoichiometry with an association affinity constant (K A ) of 1.0 ϫ 10 7 M Ϫ1 at 20°C. The formation of this complex does not appear to involve large-scale structural rearrangements of either component, judging from circular dichroism data, although small changes in either one or both components may take place. In addition, we find that the heme group of ␣-globin is not required for AHSP interaction. Finally, the behavior of truncation mutants of AHSP indicates that at least six amino acids from the N terminus and 17 from the C terminus are dispensable for ␣-globin binding activity.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Full-length human AHSP cDNA (GenBank TM accession number AF147435) was sub-cloned into the bacterial expression vector pGEX-2T (Amersham Biosciences). Expression from this plasmid in Escherichia coli BL21 cells yielded fulllength AHSP as a protein fusion with glutathione S-transferase (GST). AHSP expression was performed in shaker flasks overnight at 25°C. Following cell lysis, the fusion protein was captured on a reduced glutathione-agarose column (Amersham Biosciences), and the AHSP polypeptide was subsequently released from the column by treatment with thrombin (Sigma), which acted at the engineered cleavage site encoded by the pGEX-2T vector. AHSP fractions were dialyzed into 10 mM sodium phosphate, pH 7.0, and applied to an UNO-Q anion exchange column (Bio-Rad). Isocratic elution of AHSP in 50 mM NaCl, 20 mM sodium phosphate buffer, pH 7.0, yielded protein of Ͼ95% purity as determined by reversed-phase HPLC (Phenomenex C18) and SDS-PAGE analysis. A final gel filtration step on Superose 12 (Amersham Biosciences) produced a single protein peak corresponding to AHSP. The molecular mass of the purified protein as determined by electrospray mass spectrometry was 11,984 Ϯ 0.5 Da, in agreement with the predicted mass of 11,984 Da (including an additional Gly-Ser dipeptide at the N terminus of the protein, which remains after thrombin cleavage). The final yield was ϳ26 mg of purified AHSP per liter of bacterial culture.
Purified ␣and ␤-globin were obtained from human blood. All purification steps were carried out at 4°C, where possible. Human red blood cells were obtained from whole blood by centrifugation (3000 rpm, 10 min) and washing with 0.9% (w/v) NaCl. Carbon monoxide was bubbled through the cell suspension to form CO-liganded hemoglobin, a stable form more resistant to oxidation-induced precipitation. Hemolysates were generated by addition of five cell-pellet volumes of deionized water on ice, and membrane fractions were removed by centrifugation (7000 rpm, 15 min) after addition of NaCl to a final concentration of 0.9% (w/v). The ␣and ␤-globin chains were separated using the well-established method of reaction with p-hydroxymercuribenzoate (PMB) (7). In this case, 2.4 g of hemoglobin was incubated overnight with 143 mg of PMB, and the globin chains were separated on DEAE-Sepharose following Ikeda-Saito et al. (8). PMB was removed from both chains by overnight incubation with 20 mM dithiothreitol. Mass spectrometry analysis was used to monitor PMB-globin adduct formation and to confirm quantitative removal of PMB from the final globin preparations.
Further purification of ␣-globin was performed by fast-protein liquid chromatography using an Amersham Biosciences Mono-S column. Protein was loaded in 20 mM sodium phosphate buffer (pH 6.6), and ␣-globin eluted in linear gradient of 0 -175 mM NaCl over 6 column volumes. For ␤-globin fractions, protein was applied to a Mono-Q column in 10 mM sodium phosphate buffer (pH 8.5, buffer A) and eluted in a gradient of 0 -100% buffer B (20 mM sodium phosphate, 100 mM NaCl, pH 7.0). Purified ␣and ␤-globin were stored frozen at Ϫ80°C until required. After thawing, purified globins were re-charged with CO, and a final gel filtration step on Superose 12 (Amersham Biosciences) was performed directly prior to use in all experiments (any apo-␣-globin formed during purification was excluded from the column). Concentrations of purified globin chains were estimated from the concentration of the associated heme group, which was calculated by measuring absorption at 390 nm of globin samples unfolded in 6 M guanidine hydrochloride. Measured absorbances were compared with a standard curve generated from purified hemin (Sigma). As a further control, heme concentration was also estimated from iron concentration as determined by flame ionization atomic absorption spectroscopy (using a Varian ApectrAA-20). These two methods gave consistent results.
Aliquots of apo-␣-globin and apo-␤-globin, which lack the heme prosthetic group, were prepared by reversed-phase HPLC. For each aliquot, ϳ5 mg of protein from the DEAE-Sepharose ␣or ␤-globin fraction was loaded onto an Alltima C18 column (Alltech) in 10% acetonitrile, 0.1% trifluoroacetic acid, and a gradient of 10 -90% acetonitrile was applied over 50 min at a flow rate of 1 ml min Ϫ1 . The apo-globin fraction was lyophilized and stored at Ϫ20°C. Apo-globins were re-suspended in 50 mM sodium phosphate buffer (pH 7.4) and filtered through a 0.2 m membrane directly prior to use.
Deletion Mutants of AHSP-N-and C-terminal truncation mutants were generated by PCR and cloned into the pGEX6P1 expression plasmid (Amersham Biosciences), and their identities were verified by DNA sequencing. Expression and purification were carried out as described for the full-length protein, except that AHSP was liberated from the GST portion of the fusion protein using the PreScission protease (Amersham Biosciences). The masses of the deletion proteins were verified by electrospray mass spectrometry to exclude the possibility of further unwanted proteolytic digestion.
Circular Dichroism Spectropolarimetry-Spectra were collected on a Jasco J-720 spectropolarimeter, using a 1-mm path-length cell, and the temperature was controlled using a water-jacketed cell holder. Spectra were collected at 20°C over the wavelength range 184 -260 nm, using a resolution of 0.5 nm and a bandwidth of 1 nm. Final spectra were the sum of three scans accumulated at a speed of 20 nm min Ϫ1 with a response time of 1 s. For the thermal denaturation measurements, the ellipticity at 222 nm was monitored over a temperature range of 10 -88°C, using a resolution of 0.2°C, a bandwidth of 1 nm, and a temperature gradient of 1°C min Ϫ1 . CD spectra were analyzed, essentially by the method of Chen and Yang (9), using the CDPro software package available from lamar.colostate.edu/ϳsreeram/CDPro/main.html (10).
Analytical Size-exclusion Chromatography-Size exclusion chromatography was carried out at either 16 or 4°C using either a Superose 12 column (Amersham Biosciences) with a running buffer containing 150 mM NaCl and 20 mM sodium phosphate, pH 7.0, or a TSK-GEL column (G2000SWXL, MAC-MOD) with a running buffer containing 150 mM NaCl, 50 mM Hepes, pH 7.4, 1 mM EDTA. The columns were calibrated using the molecular mass standards RNase A (13.7 kDa), chymotrypsinogen A (25 kDa), ovalbumin (43 kDa), and albumin (67 kDa), and a standard curve was constructed by plotting log 10 M r against elution volume. Molecular weights for injected samples were estimated by comparison to the standard curve.
Amino Acid Analysis-Amino acid analysis was performed by the Australian Proteome Analysis Facility (Sydney, Australia) to determine the compositions and concentrations of purified AHSP, ␣-globin, and AHSP⅐␣-globin complex. Due to the similar retention times of the AHSP and AHSP⅐␣-globin peaks during gel filtration, complexes for amino acid analysis were collected after injection of mixtures containing excess ␣-globin to avoid contamination with free AHSP. The quantity of each amino acid was expressed as a molar percentage of the total measured amino acid content, and these values were compared with calculated amino acid percentages for theoretical complexes of 1:1, 1:2, and 2:1 ratios of AHSP⅐␣-globin. To determine the goodness of fit, 2 values were calculated as ⌺(quantity expected Ϫ quantity calculated ) 2 . Serine was excluded from our analysis due to systematic error in this measurement.
Isothermal Titration Calorimetry-Calorimetric experiments were carried out using a MicroCal VP-ITC titration calorimeter. Protein solutions were buffer-exchanged into 20 mM sodium phosphate buffer (pH 7.0) containing 150 mM NaCl by gel filtration chromatography. Heats of reaction were determined by multiple injections of 100 M AHSP into 10 M ␣-globin at 20°C, with stirring at 310 rpm and a delay of 250 s between each injection. The reverse experiment (i.e. injections of 100 M ␣-globin into 10 M AHSP) was also performed. Heats of dilution for all experiments were measured from injection of each protein into buffer, and these values were subtracted from the binding data. Titration curves were analyzed using Origin 5.0 software (Micro-Cal) by fitting to various simple models provided in the software.
Analytical Ultracentrifugation-Equilibrium and sedimentation velocity experiments were performed using a Beckman model XL-A analytical ultracentrifuge equipped with an An-60 Ti rotor. Proteins were buffer exchanged into 20 mM sodium phosphate buffer (pH 7.0) containing 150 mM NaCl by gel filtration chromatography, and protein concentrations were determined in the XL-A by reference to a blank containing this buffer alone. For equilibrium measurements, data were collected in six-sector cells as absorbance (280 nm) versus radius scans (0.001-cm increments, 10 averages) at 18,000, 30,000, and 42,000 rpm. Scans were collected at 3-h intervals and compared to determine when the samples reached equilibrium. Analysis of the data was carried out using the NONLIN software (11), and the best model and final parameters were determined by examination of the residuals derived from fits to several models.
For sedimentation velocity experiments, absorbance data were collected in two-channel centerpieces in continuous mode with a step size of 0.003 cm and at time intervals of 900 s, with no signal averaging. Preceding each run, the samples were spun at 3000 rpm for at least 3 h so that they reached thermal equilibrium in the rotor. Data were analyzed using the program Sedfit (12) to simultaneously determine the apparent molecular weight (M r ,app) and sedimentation coefficient for a given buffer and temperature (s T,b ).
In the above analyses, the partial specific volumes (v) of each protein were determined from the amino acid sequences (13) and adjusted for temperature (14) using the program SEDNTERP. 2 The buffer density was taken to be 1.0068 g ml Ϫ1 and the viscosity 0.0105 Poise at 20°C. For other temperatures, buffer density ( T,b ) and viscosity ( T,b ) were calculated using the SEDNTERP program, assuming the bulk of the temperature dependence to be contributed by the majour component, water. Values for s T,b were transformed to standard sedimentation coefficients in water at 20°C (s 20,w ) using the following equation (16), where the density of water at 20°C ( 20,w ) is 0.9982 g ml Ϫ1 and the viscosity ( 20,w ) is 0.01002 poise. For consideration of theoretical spherical molecules, the anhydrous radius (R) was calculated from molecular weight following the empirical relationship r ϭ 6.72 ϫ 10 Ϫ9 M r 1/3 (17), and values for the frictional coefficient (f o ) and s were calculated as described previously (16,18).

AHSP and the AHSP⅐␣-Globin Complex Are Predominantly
␣-Helical-A secondary structure prediction carried out on the amino acid sequence of AHSP revealed that the protein may contain substantial amounts of ␣-helix (Fig. 1A). We therefore sought to confirm this experimentally. A far-UV circular dichroism spectrum of AHSP (Fig. 1B) displays minima at 209 and 222 nm and a maximum at ϳ190 nm, indicating that the protein is folded and adopts a largely ␣-helical conformation. Analysis of the CD spectrum using the CDPro software package suggests that AHSP contains ϳ70% ␣-helix. A thermal denaturation experiment, monitored using the ellipticity at 222 nm, showed that AHSP unfolds reversibly with a T m of 60°C. All of these data are consistent with the protein adopting a well-defined, largely ␣-helical conformation.
Given that free oxy-␣-globin has a propensity to aggregate and precipitate, we investigated the possibility that the protein might be partially unfolded and that the binding of AHSP might stabilize ␣-globin by inducing substantial conformational change. A far-UV CD spectrum (Fig. 1C), deconvoluted using CDPro, reveals that free ␣-globin contains ϳ80% ␣-helix. Notably, ␣-globin within the hemoglobin tetramer is ϳ75% ␣-helix, suggesting that free ␣-globin retains similar levels of secondary structure even in the absence of its ␤-globin partner, a conclusion in line with previous studies (19). In addition, thermal denaturation of ␣-globin, carried out as described above, revealed a single highly cooperative unfolding transition with a T m of 59°C, consistent with a well-defined fold.
Formation of the AHSP⅐␣-globin complex appears to result in a small reduction in total ␣-helical content (Fig. 1C), but there is little evidence for large-scale structural rearrangement, suggesting that ␣-globin stabilization by AHSP occurs through an alternative mechanism. It should be borne in mind, however, that changes in the spatial organization of ␣-helices might not be detected in this experiment.
Both AHSP and ␣-Globin Are Monomeric in Solution-We next sought to define the aggregation states of both AHSP and ␣-globin in solution. As noted previously, oxy-␣-hemoglobin is unstable, largely as a result of oxidation of the iron (II) heme group to iron (III), and subsequent reactions leading to protein precipitation (20,21). In contrast, the CO-liganded form of ␣-globin is much more resistant to this oxidation-induced precipitation but retains the ability to bind AHSP (4). Thus, we have used the CO-liganded form of ␣-globin to study the physical properties of the molecule and its interaction with AHSP.
Gel filtration chromatography of purified AHSP indicated that the protein forms a single species of discrete molecular size, with no evidence of aggregation into high molecular weight forms, even at concentrations up to 1 mM (AHSP is estimated to attain a concentration of 0.1 mM in pro-erythroblasts (4) 1. Secondary structure analysis of human AHSP and ␣-globin. A, sequence of the AHSP protein used in these studies (Nterminal Gly-Ser residues remain from thrombin cleavage of the expressed GST-AHSP fusion protein; these are underlined). A consensus secondary structure prediction (sspred) was generated using the Jpred2 web server (jura.ebi.ac.uk:8888 (34)), confidence scores included. B, far-UV circular dichroism (CD) spectrum of 15 M AHSP in buffer (50 mM NaF, 20 mM sodium phosphate, pH 7.0, 20°C), indicating that the protein contains predominantly ␣-helical secondary structure. C, conformational analysis of the AHSP⅐␣-globin complex. CD spectra of both 10 M AHSP (open circles) and 10 M ␣-globin (filled circles) in 37.5 mM NaCl, 5 mM sodium phosphate (pH 7.0, 20°C) are shown. To detect changes in secondary structure upon binding, these two samples were mixed and the resulting CD spectrum was numerically doubled (solid line) and compared with the theoretical sum of the spectra from the individual proteins (dashed line). The inset shows these two lines scaled to reveal small qualitative signal differences. standards indicated that AHSP migrates in a position consistent with a globular protein of ϳ23 kDa, roughly twice the mass predicted by the primary amino acid sequence. By comparison, ␣-globin ran close to its expected monomeric size of ϳ14 kDa. However, because the elution properties of proteins in size exclusion chromatography are shape-dependent, we chose to characterize the aggregation state of both AHSP and ␣-globin using sedimentation equilibrium experiments. The data in Fig.  2A show clearly that AHSP is monomeric, suggesting that its gel filtration profile may reflect anisotropy in the shape of the protein. A single species fit was observed for all protein concentrations examined, up to 0.1 mM (the higher protein concentrations were monitored at 250 nm; not shown).
To investigate the hydrodynamic properties of AHSP further, we performed a series of sedimentation velocity experi-ments to measure the sedimentation coefficient (s) of AHSP (Fig. 2B). The model that best fitted the experimental data incorporated a non-interacting single species with M r ,app ϭ 11,800 Ϯ 460 Da (n ϭ 14) and a sedimentation coefficient, extrapolated to infinite dilution (s 20,w 0 ), of 1.52 S (Fig. 2B). This result is in close agreement with the mass determined from equilibrium experiments and with a theoretical mass of 11,986 Da for the monomeric protein. Calculation of s for a theoretical sphere of the same mass yielded values in the range 1.86 Ն s Ն 1.58, depending upon the level of hydration (ranging from anhydrous to hydration at 0.41 g of H 2 O per 1 g of protein based on amino acid composition (22)). Hence, our data are consistent with AHSP assuming a globular fold, but, because the rate of sedimentation observed was slower than predicted by a spherical model, we suggest that AHSP may have a slightly elongated shape, consistent with the behavior of the molecule by gel filtration.
A similar analysis of CO-liganded ␣-globin indicated that it is also monomeric in solution (Fig. 3). Some small systematic deviations between the non-interacting monomer model and the experimental data were observed (Fig. 3A, residuals), but introducing either self-association or non-ideality into the model could not reduce these deviations. One possibility is that small amounts of oxidized material form aggregates during the course of the experiment, although the majority of ␣-globin was in the CO-liganded form as judged by spectrophotometric measurements in the region 540 -550 nm (not shown). Given the small magnitude of the residuals, we propose that the monomer model is valid. Sedimentation velocity experiments yielded a value for s 20,w of 1.97 S (Fig. 3B). As before, s was calculated for a theoretical sphere of the same mass and yielded values in the range 2.09 Ն s Ն 1.64 (ranging from anhydrous to hydration at 3.9 g of H 2 O per 1 g of protein). These values suggest that ␣-globin is rather spherical, in accordance with the gel filtration data. Note that the possible effects of the heme prosthetic group were ignored for the purposes of this analysis. Taken together, the results from gel filtration and sedimentation analysis indicate that purified AHSP and ␣-globin are both monomers in solution, but that AHSP is probably more elongated than ␣-globin.
The AHSP⅐␣-Globin Complex has a 1:1 Stoichiometry-To examine the AHSP⅐␣-globin complex in detail, we titrated AHSP into a solution of ␣-globin and monitored the formation of complex by gel filtration chromatography (Fig. 4A). These experiments showed that the interaction between AHSP and ␣-globin is sufficiently tight that, when mixed together, free monomers only occur in measurable quantities when one of the two proteins is in excess, even at protein concentrations below 10 M (suggesting a K A Ͼ 10 6 M Ϫ1 for formation of the complex). Upon addition of 1 molar equivalent of AHSP to a sample of ␣-globin, a single species eluted from the gel filtration column with an apparent molecular size of ϳ30 kDa, suggesting the formation of a heterodimeric AHSP⅐␣-globin complex. To confirm the stoichiometry, gel filtration fractions corresponding to the AHSP⅐␣-globin complex were collected and subject to amino acid analysis. This method does not rely on estimates of protein concentration derived from optical absorbance measurements and, therefore, provides an independent measurement of subunit concentrations in the complex. The experimentally derived amino acid composition was compared with values calculated for theoretical AHSP⅐␣-globin complexes of 1:1, 2:1, and 1:2 subunit stoichiometry. Analysis of the relative levels of amino acids characteristic of either AHSP or ␣-globin indicated a 1:1 subunit stoichiometry for the complex (Fig. 4B). Global analysis of the data by determination of ⌺residuals 2 between observed and calculated amounts of every residue gave values of 0.8, 8.5, and 9.8 for the 1:1, 2:1, and 1:2 models, respectively, confirming the 1:1 model to be the best fit. Samples of AHSP and ␣-globin alone were also analyzed, and the expected amino acid compositions were confirmed. In summary, all the data indicate the AHSP⅐␣-globin complex is a heterodimer.
AHSP and ␣-Globin Form a Moderately Tight Complex-We next used isothermal titration calorimetry to measure thermodynamic parameters for the AHSP⅐␣-globin interaction. Fig. 5 shows that formation of the complex is exothermic, and the data fitted well to a simple 1:1 binding model (Fig. 5, lower  panel) with ⌬H ϭ 8.75 Ϯ 0.07 kcal mol Ϫ1 and K A ϭ 1.0 ϫ 10 7 M Ϫ1 at 20°C. We note that this interaction affinity is greater than that observed for the ␣ 1 ␤ 1 (dimer) 7 ␣ 1 ␤ 1 ⅐␣ 2 ␤ 2 (tetramer) equilibrium state of fully oxygenated hemoglobin. Hence, AHSP binding could have consequences for oxygen transport if it contacted  4. AHSP and ␣-globin form a 1:1 complex. A, Superose-12 gel filtration chromatography of purified AHSP (I) and purified ␣-globin (II) indicate species of discrete molecular size. Addition of increasing molar fractions of AHSP to an ␣-globin solution of constant concentration shows that the stoichiometry of the interaction is 1:1. Three points of this titration are shown (III, IV, and V). The running buffer was 150 mM NaCl, 20 mM sodium phosphate, pH 7.0, and the excluded volume of the column is indicated (V o ). Optical absorbance at 540 nm (dashed lines) was used to specifically monitor the elution profile of ␣-globin and total protein was monitored at 280 nm. B, amino acid analysis of the Superose-12-purified AHSP⅐␣-globin complex showing the six residue types most characteristic of the AHSP or ␣-globin sequences. The experimentally determined composition (filled bars, average from two samples) is compared with the theoretical compositions for 1:1, 2:1, or 1:2 AHSP⅐␣-globin ratios (hatched bars). the residues of ␣-globin that mediate this dimer 7 tetramer association (i.e. those residues at the ␣ 1 ␤ 2 interface). To investigate this possibility, we performed binding competition experiments using AHSP and an equimolar mixture of CO-liganded ␣and ␤-subunits (these display virtually identical dimer 7 tetramer equilibrium to oxygenated subunits; K A ϭ 0.8 ϫ 10 6 M Ϫ1 at 20°C (23)). We observed that ␣-globin became exclusively incorporated into a complex with ␤-globin, with no evidence of a ternary complex with AHSP (Fig. 6B). This result was observed irrespective of the order of addition of components, hence, the pre-formed ␣-globin⅐AHSP complex was not protected from dissociation. Furthermore, addition of recombinant AHSP to purified hemoglobin A had no effect on the Hill coefficient for oxygen binding, indicating that the cooperativity of oxygen binding is unaffected by AHSP (not shown). Together these results suggest that AHSP is unlikely to interact with ␣-globin at the ␣ 1 ␤ 2 interface of hemoglobin.
AHSP Binds ␣-Globin in the Absence of the Heme Prosthetic Group-To begin to probe the structural determinants of the AHSP⅐␣-globin interaction, we first investigated the role of the heme prosthetic group. Removal of the heme group from ␣-globin was achieved by reversed-phase HPLC chromatography at low pH. Apo-␣-globin fractions were lyophilized and then taken up in phosphate buffer at neutral pH. Analysis by CD spectropolarimetry revealed that the resulting apo-␣-globin retains considerable, but much reduced, ␣-helical structure (ϳ30% ␣-helix) in the absence of the heme group. This result is in agreement with previous studies of apo-hemoglobin and its subunits, which demonstrated that removal of heme causes a greater destabilization of ␣-globin secondary structure than of ␤-globin (24,25). A UV-visible wavelength absorption spectrum indicated that contamination of the apo-␣-globin preparation with heme-containing protein was at levels below 1% (ratio of absorption, A 420 /A 280 ϭ 0.04 (26)).
Using gel filtration analysis we observed the formation of a complex between apo-␣-globin and AHSP eluting at a similar volume to the holo-␣-globin⅐AHSP complex, suggesting that the heme group is not required for this interaction (Fig. 6B).
In the absence of AHSP, apo-␣-globin cannot be detected by gel filtration chromatography; it appears to interact with the chromatography system in such a way that no discrete protein peak is observed. Notably, apo-␣-globin has previously been shown to form high sedimentation coefficient aggregates resulting in exclusion from gel filtration systems (24). In addition, higher yields of soluble apo-␣-globin were obtained when lyophilized protein was taken up in buffer already containing AHSP. These observations suggest that the apo-␣-globin may somehow be stabilized by AHSP binding, although CD measurements of apo-␣-globin in the presence or absence of AHSP did not reveal substantial changes in secondary structure. Apo-␤-globin showed no binding to AHSP in our assay (not shown).
The N and C Termini of AHSP Are Not Required for ␣-Globin Binding-To further define structural requirements for the AHSP⅐␣-globin interaction, we generated a series of truncation mutants of AHSP and assayed their ability to bind ␣-globin by gel filtration (Fig. 7). Deletions from the C terminus of up to 17 amino acids (mutant ⌬C17) had no significant effect on ␣-globin binding. In addition the ⌬C17 mutant adopted an ␣-helical conformation similar to the full-length protein as determined by CD (not shown), supporting the hypothesis that this region may not contribute to the folded globular region of AHSP. Deletion of a further 6 residues (mutant ⌬C23) completely abrogated the interaction in this assay. Thus, the C-terminal region that is not required for the formation of a complex with ␣-globin corresponds closely to the unstructured C-terminal tail predicted from sequence analysis (Fig. 1A). Likewise, an N-terminal truncation of 6 residues had no effect on binding, whereas removal of 18 residues abolishing binding (Fig. 7). An intermediate effect was seen for the ⌬N12 mutant: binding was observed but with a substantially reduced affinity. DISCUSSION Given that the precipitation of ␣-globin in vivo carries severe medical consequences, it is reasonable to suppose that an organism might contain proteins to assist in the regulation of correct globin assembly. AHSP is a small globular protein that has recently been shown to bind specifically to ␣-globin and prevent its precipitation in vitro. Moreover, loss of AHSP through gene ablation in mice causes hemoglobin precipitates to accumulate in red blood cells, demonstrating that AHSP is required for normal hemoglobin production in animals. Further investigations into the structure of AHSP and the nature of its interaction with ␣-globin are likely to provide the basis for new therapeutic strategies to inhibit the formation of toxic hemoglobin precipitates in disorders of globin chain imbalance, such as ␤-thalassemia. Here we show that AHSP is a globular ␣-helical protein with a somewhat asymmetric shape and with flexible N and C termini that are not required for ␣-globin binding. AHSP exists as a monomer in solution and binds ␣-globin with only minor changes in the secondary structure content of either protein. Interestingly, the ␣-globin heme group is not required for interaction with AHSP, despite the fact that ␣-globin cannot attain a native fold in the absence of this prosthetic group. Indeed, we find that interaction with AHSP stabilizes apo-␣-globin in solution, suggesting that AHSP may also prevent aggregation of ␣-globin from which the heme group has been lost. AHSP might bind either to the regions of residual structure in apo-␣globin or induce structure in regions that are otherwise less well defined. In this regard, CD analysis of ␣-globin proteolytic fragments has suggested that the apo-␣-globin N terminus is a region that maintains residual structure (27).
The AHSP⅐␣-globin complex forms with a 1:1 subunit stoichiometry and an association constant in the sub-micromolar range (1 ϫ 10 7 M Ϫ1 ). With the observed affinity, the AHSP⅐␣globin interaction is likely to be biologically significant, given the concentrations of AHSP (0.1 mM) and hemoglobin (4 mM) found in red blood cells. In addition there are consequences regarding the site of AHSP binding on ␣-globin. Hemoglobin exists in a dimer 7 tetramer equilibrium, with each ␣-subunit contributing to an intra-dimer ␣ 1 ␤ 1 interface and an interdimer ␣ 1 ␤ 2 interface. The ␣ 1 ␤ 1 interface is essentially unchanged in all ligand states of hemoglobin and constitutes a very high affinity interaction (K A Ͼ Ͼ 10 10 M Ϫ1 ). It is the ␣ 1 ␤ 2 interface that becomes reorganized upon oxygen binding, thereby giving rise to the well-described allosteric effects. For deoxyhemoglobin, the equilibrium constant for tetramer formation through the ␣ 1 ␤ 2 interface is 4.2 ϫ 10 10 M Ϫ1 at 21.5°C (28), well above the binding affinity of AHSP. Upon full oxygenation (or CO binding), this association constant is reduced substantially to ϳ1 ϫ 10 6 M Ϫ1 (23), potentially making the ␣ 1 ␤ 2 interface accessible to AHSP, if it were to bind this region. However, we have shown that ␤-globin out-competes AHSP, even under conditions of CO saturation. These findings suggest that the AHSP binding site might overlap with the high affinity ␣ 1 ␤ 1 interface rather than the ␣ 1 ␤ 2 face. Alternatively, a conformational change in ␣-globin upon binding ␤-globin may be responsible for masking the AHSP interaction site, although we favor FIG. 7. Analysis of ␣-globin binding by deletion mutants of AHSP. A, schematic diagram of deletion mutants of AHSP with reference to their predicted secondary structure and a summary of their ␣-globin binding properties, as determined in B. B, gel filtration data showing the capacity for AHSP deletion mutants to bind to ␣-globin. Recombinant AHSP (10 M) deletion mutants were mixed with purified human ␣-globin (5 M) and run on a TSK-GEL gel filtration column. The retention time for ␣-globin was determined by monitoring absorbance at 540 nm, indicating the level of incorporation into the complex with AHSP. the former possibility; CD measurements have demonstrated that there is no significant change in the conformation of ␣-globin upon binding to ␤-globin (19). It is likely that high resolution structural information will ultimately distinguish these possibilities. In particular, it will be interesting to determine whether the AHSP⅐␣-globin interface structurally mimics the ␣ 1 ␤ 1 interface, especially given that AHSP is, like the globins, a predominantly ␣-helical protein.
Regardless of the site of interaction, our finding that AHSP binding is effectively excluded upon formation of ␣␤ complexes is consistent with the proposed role for AHSP in erythroid cells as an ␣-globin-specific chaperone. According to this model, AHSP sequesters free ␣-globin to limit its toxicity. This interaction would only be required prior to hemoglobin assembly, in the event of hemoglobin breakdown, or under conditions of relative ␣-globin excess. The moderate binding affinity observed for the AHSP⅐globin interaction is sufficient to sequester free ␣-globin but would not interfere with hemoglobin formation.
In some respects, these potential functions for AHSP resemble those of Rbl2p, a yeast protein that binds ␤-tubulin, a component of microtubules (29 -31). Free ␤-tubulin (i.e. ␤-tubulin not associated with ␣-tubulin) disrupts microtubule assembly and function. Rbl2p limits the toxicity of free ␤-tubulin by forming a heterodimeric complex. However, ␤-tubulin interacts more strongly with ␣-tubulin than Rbl2p, favoring microtubule formation when tubulin subunits are balanced. ␤-Tubulin bound to Rbl2p can be released, either for assembly into microtubules or for storage in higher order nontoxic aggregates. The fate of ␣-globin bound to AHSP remains to be determined. In vitro experiments indicate that AHSP-bound ␣-globin can be released for assembly into hemoglobin A tetramers. Other outcomes are also possible. For example, AHSP might enhance the catabolism of excess free ␣-globin through interactions with protein degradation machinery (32), which are known to occur with other chaperone proteins. It is also possible that AHSP plays other roles in erythrocyte maturation, for example, by acting as a monitor of free ␣-globin levels that feeds back to globin gene regulation. These potential functions could be mediated through the non-helical N-and Cterminal domains of AHSP, which we have demonstrated to be dispensable for ␣-globin binding.
Based on our current and previous findings, AHSP appears to represent a novel type of molecular chaperone with distinct characteristics. It most closely resembles members of the small heat shock protein family in its size and ATP independence, although, unlike small heat shock proteins, AHSP does not form higher order oligomeric complexes (33). Also, in contrast to most other chaperones, which are widely expressed and relatively promiscuous with respect to substrate interactions, AHSP appears to be highly tissue-and substrate-specific. Our current studies on the biochemistry of the AHSP⅐␣-globin interaction provides fundamental information that is relevant to understanding its biological function. Future studies to define the structure of the AHSP⅐protein complex should lead to an improved understanding of hemoglobin biology and elucidate novel approaches for minimizing its pathological precipitation in various disease states