Assembly of human hemoglobin. Studies with Escherichia coli-expressed α-globin

The α-globin of human hemoglobin was expressed in Escherichia coli and was refolded with heme in the presence and in the absence of native β-chains. The functional and structural properties of the expressed α-chains were assessed in the isolated state and after assembly into a functional hemoglobin tetramer. The recombinant and native hemoglobins were essentially identical on the basis of sensitivity to effectors (Cl− and 2,3-diphosphoglycerate), Bohr effect, CO binding kinetics, dimer-tetramer association constants, circular dichroism spectra of the heme region, and nuclear magnetic resonance of the residues in the α1β1 and α1β2 interfaces. However, the nuclear magnetic resonance revealed subtle differences in the heme region of the expressed α-chain, and the recombinant human normal adult hemoglobin (HbA) exhibited a slightly decreased cooperativity relative to native HbA. These results indicate that subtle conformational changes in the heme pocket can alter hemoglobin cooperativity in the absence of modifications of quaternary interface contacts or protein dynamics. In addition to incorporation into a HbA tetramer, the α-globin refolds and incorporates heme in the absence of the partner β-chain. Although the CO binding kinetics of recombinant α-chains were the same as that of native α-chains, the ellipticity of the Soret circular dichroism spectrum was decreased and CO binding kinetics revealed an additional faster component. These results show that recombinant α-chain assumes alternating conformations in the absence of β-chain and indicate that the isolated α-chain exhibits a higher degree of conformational flexibility than the α-chain incorporated into the hemoglobin tetramer. These findings demonstrate the utility of the expressed α-globin as a tool for elucidating the role of this chain in hemoglobin structure-function relationships.

The basis of cooperative oxygen binding by human hemoglobin is an old problem that is being elucidated with the aid of modern technologies. Of particular importance is the elucida-tion of the properties of ␣and ␤-chains, both in the isolated state and incorporated into the hemoglobin tetramer. Such studies should provide fine details on the linkage between tertiary and quaternary structure that forms the basis of cooperative oxygen binding. The recent development of hemoglobin expression systems offers great promise for facilitating the study of such structure-function relationships. This approach has been particularly useful for investigating the physicochemical properties of hemoglobin and myoglobin and has been employed in the development of hemoglobin variants with properties suitable for clinical use as red cell substitutes (1).
Although Escherichia coli expression systems for HbA 1 have been developed, the product exhibits structural and functional modifications relative to native HbA (2)(3)(4). Alternatively, tetrameric hemoglobins in which only the ␣or ␤-subunits are recombinant (␣rHbA and ␤rHbA) can be obtained by in vitro refolding of the expressed globin. ␤-Globin has thus been expressed (5,6) and refolded in the presence of native ␣-chains to yield recombinant hemoglobin whose conformational and functional properties resemble those of natural HbA (5,6). In contrast, development of an analogous ␣-globin expression system was not as successful (7), although it is highly desirable as a tool for elucidating the tetrameric assembly and cooperativity of HbA.
In this paper we report the construction of a system for the synthesis of high levels of an ␣-globin fusion protein in E. coli. The conformational properties of the isolated chain are presented along with the functional and conformational characterization of reconstituted HbA tetramer following refolding of the expressed ␣-globin with heme in the presence of native ␤-chains. Spectral measurements and oxygen binding of these proteins reveal new insights into the linkage between the conformation of the heme pocket, ligand access channel, and quaternary interactions.

MATERIALS AND METHODS
Plasmid Construction-The ␣-globin expression plasmid pNF␣ is structurally analogous to pJKO5, from which ␤-globin has been routinely produced as a fusion protein, and comprises 83 residues of a flu virus protein, NS1, a Factor X recognition sequence followed by ␤-globin. The construction of pNF␣ is similar to that of pJKO5 (6). Human ␣-globin cDNA was provided by B. G. Forget (Yale University) as a 700-base pair PstI insert in pKT218 (8). In an initial effort to express ␣-globin in E. coli, the 700-base pair PstI fragment was inserted into the PstI site of pKK-233-2 (Pharmacia Biotech Inc.), and the short region between the NcoI sites at the initiation ATG site in the globin gene was subsequently removed by digestion with NcoI followed by religation. The resulting plasmid, pB␣2, in JM101 induced with isopropyl-1-thio-␤-D-galactopyranoside was found to produce very low levels of ␣-globin. Although unsatisfactory as an expression system, the 750-base pair NcoI to XmnI fragment from this plasmid was used as a source of the ␣-globin gene for the construction of pNF␣. pNF␣ was constructed in two steps. The initial construct, pN␣, was formed by inserting the NcoI to XmnI fragment from pB␣2 into pJKO4 (6), from which the ␤-globin gene had been removed by digestion with NcoI and EcoRV. (Note that both EcoRV and XmnI produce blunt ended DNA.) pJKO4 codes for a fusion protein composed of 81 residues of NS1 followed by ␤-globin but does not have an intervening Factor X recognition sequence. pN␣ in E. coli strain AR120 (9), induced with nalidixic acid, produced high levels of the fusion protein NS1-␣-globin (data not shown). A Factor X recognition sequence was inserted between the NS1 segment and ␣-globin by first digesting pN␣ with NcoI and making the DNA blunt ended with mung bean nuclease. Subsequent digestion with BglII removed the NS1 portion of the fused gene. This segment was replaced by the BglII to StuI fragment from pJKO1 (6), which contains the same NS1 portion followed by a Factor X recognition sequence. The resulting plasmid, pNF␣, is structurally identical to pJKO5 except that ␣-globin cDNA replaces the ␤-globin cDNA of pJKO5. This plasmid in strain AR120, induced with nalidixic acid, yielded large amounts of the NS1-FX-␣globin fusion protein as the principal component of the insoluble inclusion bodies.
Expression and Purification of the ␣-Globin-Growth, expression, and purification of the fusion protein followed a previously described protocol (6). Enzymatic cleavage was monitored by reverse phase HPLC using a Vydac C 4 column. The solvents were: A, 20% CH 3 CN, 0.1% trifluoroacetic acid; and B, 60% CH 3 CN, 0.1% trifluoroacetic acid. The gradient was 44% B to 75% B over 120 min. Peaks were identified by eight cycles of Edman sequencing.
Reconstitution of ␣rHbA-Upon cleavage with Factor X, the ␣-globin was recovered as a precipitate, dissolved in a minimum volume of cold 0.1 M NaOH, and diluted to 2 mg/ml (A 280 ϭ 1.0 is taken as 1 mg/ml) with 0.04 M borate buffer, pH 9.0, 0.002 M EDTA. In order to solubilize the aggregates, the solution was deoxygenated by stirring under a constant stream of N 2 followed by addition of dithiothreitol to a final concentration of 0.001 M. After 24 h at 4°C the protein was reconstituted with cyanoheme and native ␤-chains, prepared from carboxy-HbA by reaction with p-mercuribenzoate (10). p-Mercuribenzoate was removed from the isolated ␤-chains as described by Geraci and Li (11). After 1 week the protein was concentrated and the heme converted to the CO derivative by addition of sodium dithionite under a stream of CO. The hemoglobin was purified on an affinity column of immobilized hemoglobin (12,13). The collected fractions were analyzed on paragon gel electrophoresis (Beckman Instruments).
Refolding with Heme of Isolated ␣r-The ␣-globin was purified and reconstituted with cyanoheme as described. After 1 week at 4°C the protein was concentrated and the heme converted to the CO form followed by dialysis against 0.001 M NaCl saturated with CO. The ␣r were purified by preparative isoelectrofocusing (Bio-Rad) according to the manufacturer's instructions. The pH range of the ampholyte was 3-10. Screening of the fractions was carried out on paragon electrophoresis (Beckman). The ampholyte was removed by repetitive washing with water using a Centriprep concentrator (Amicon).
Peptide Analysis-The natural and recombinant ␣-globins were isolated from the respective hemoglobin using reverse phase HPLC and a Vydac C 4 column (14). Prior to tryptic digestion, the isolated ␣-globins were S-pyridylethylated and desalted (15). TPCK-treated trypsin (Sigma) was added to a ratio of 1:50 (w/w) and the solution let stand at room temperature for 20 h. The resulting clear solution was then injected into a 4.6 ϫ 250 mm Vydac C 18 column previously equilibrated with Buffer A (0.1% trifluoroacetic acid). The peptides were eluted using a gradient of 0 -50% Buffer B (0.1% trifluoroacetic acid in 9:1 (v/v) CH 3 CN:H 2 O) over 60 min. The absorbance was monitored at 220 nm. Sequences were determined using a Hewlett-Packard model G1005A protein sequencing system following the procedures standardized by the manufacturer.
Oxygen Equilibrium Curves-Oxygen equilibrium measurements were performed using the thin layer dilution technique of Dolman and Gill (16) with an Aviv 14DS spectrophotometer. In this technique the oxyhemoglobin is deoxygenated by stepwise dilution of the equilibrating gas with constant volumes of N 2 . Formation of methemoglobin as judged by spectral deconvolution of the initial and final spectra was less than 10%. Protein concentration was 1.5-2.0 mM in heme in 0.05 M HEPES buffer at 25°C. The experimental data were fitted to the Adair equation (17) using an iterative procedure incorporating the Marquardt algorithm, where Y is the fractional saturation with O 2 , pO 2 is the partial pressure of oxygen in millimeters of mercury, and ␤ i represents the overall Adair constants related to the intrinsic statistical affinity constant K i of the subsequent steps of oxygenation by ␤ i ϭ ⌸ i K i . The value of the median ligand activity, P m , was determinated using the relation P m ϭ ␤4 Ϫ0.25 (18). CO Binding Kinetics-Flash photolysis was carried out on solutions containing 5 M heme and 50 M CO at 23°C in 0.1 M Bis-Tris (pH 7.0) containing 0.1 M KCl. Approximately 0.5 mg of sodium dithionite was added to reduce any ferric heme to the ferrous state. The instrumentation and experimental details for laser flash photolysis were essentially as described previously (19). A pulse (0.6 s) from a dye laser disrupted the photolabile heme Fe-CO bond, and the recombination of CO with the heme protein was then monitored by following the absorbance change at 436 nm. Data were transmitted to a microcomputer for processing and analysis.
Standard multiexponential analysis of the kinetic data was performed according to, where ⌬A t is the total absorbance change observed at time t, a i is the absorbance change for component i at t ϭ 0, k i is the observed pseudofirst order rate constant for component i, and n is the number of independent components. Least square analysis was performed with RS/1 software (BBN Software Products, Cambridge, MA) on a Dell 450/ME microcomputer. Statistical significance (p Ͻ 0.05) was evaluated using Student's t test assuming equal variances. Circular Dichroism-The CD spectra of the carboxyl derivatives were recorded in 0.05 M phosphate buffer at pH 7.0 using an Aviv CD 60 spectropolarimeter. Before recording the spectra hemoglobin was treated with dithionite under an atmosphere of CO and rapidly filtered through Sephadex G-25 resin. The optical spectra were deconvolved to verify that the heme was present only in the carboxyl form and to determine the exact protein concentration. The CD spectra, which are the average of three scans, were recorded every 0.3 nm using a bandwidth of 0.5 nm and a time constant of 1 s in a cuvette of 0.5 cm path length. Protein concentration was 0.2 mg/ml. NMR Measurements-The NMR experiments were carried out on a VXR-400/54 spectrometer operating at 9.4 tesla. All NMR spectra reported here were obtained at 29°C using the jump-and-return pulse sequence 90°(⌽)--90°(-⌽) (20). The delay was adjusted for each experiment such that maximum excitation was obtained for the spectral region of interest. The relaxation delay between successive transients was 2.5 s. The proton chemical shifts are referenced to 2,2-dimethyl-2silapentane-5-sulfonate. 2 The samples of native HbA and ␣rHbA used for the NMR measurements were in 0.1 M Bis-Tris buffer, pH 6.85, in 90% H 2 O, 10% D 2 O. Hemoglobin concentration was between 7 and 9%, and methemoglobin content was 5% or less. Fig. 1 shows SDS-polyacrylamide gel electrophoresis of the total cell proteins before and after induction as well as the detergent-insoluble fusion protein. This figure illustrates the high expression level of NS1-FX-␣-globin and shows that treatment of the insoluble fraction with detergent eliminates a large part of the contaminant proteins.

Protein Expression, Digestion, and Refolding-
The procedures for purification, cleavage, and reconstitution of the tetrameric hemoglobin were essentially the same as those used for ␤rHbA (6). The pH values of the various reactions were increased due to the higher isoelectric point of the ␣-globin, which renders it less soluble than ␤-globin at pH 8.5. The yield of ␣rHbA was 15-20 mg/liter of cell culture. Cleavage of the fusion protein with Factor Xa was followed by reverse phase HPLC. Fig. 2A shows the elution pattern of the detergent-purified fusion protein, and Fig. 2B shows the elution pattern of the fusion protein following 20 h of digestion with Factor Xa. The major peak corresponds to ␣-globin with the correct amino-terminal end, whereas the smaller peak is a product of overdigestion, which occurs at Arg ␣31 (B12).
Peptide Maps of ␣-Chain from Normal Human HbA and r␣HbA- Fig. 3, A and B, shows the tryptic peptide patterns obtained from normal and recombinant ␣-globins, respectively. The numbers above the peaks correspond to the tryptic fragments of the ␣-chain as listed in Table I. The HPLC profiles of tryptic digestion of the two ␣-globins are identical. Furthermore, five cycles of Edman sequential degradation on intact ␣-globins isolated from natural and recombinant HbA gave comparable yields of the expected 5 residues at the amino terminus. Sequence analysis of the peak eluting at 22 min identified an unresolved mixture of fragments 1 (residues 1-7) and 3 (residues 12-16). The peak eluting at 24 min contains fragment 1ϩ2, a partial cleavage product (residues 1-11). The difference between peaks 12 and 13 eluting at 52 and 54 min is 1 Lys residue located at position ␣61(E10).
Circular Dichroism--The spectrum in the Soret region is sensitive to the interaction of the heme with the surrounding aromatic residues (21). Fig. 4A shows the CD spectra of the carboxyl derivatives of HbA and ␣rHbA. Although these exhibit the same ellipticity at the peaks of their spectra, small differences are evident at lower wavelengths. Fig. 4B shows the CD spectra of native and recombinant ␣-chains. The recombinant ␣-chains have a lower ellipticity and a broader spectrum than the native ␣-chains.
Nuclear Magnetic Resonance- Fig. 5 shows the downfield region of the NMR spectra of ␣rHbA and HbA in the deoxy form. The resonances between 15 and 24 ppm originate from protons in the heme groups and/or from protons in amino acid residues located in the heme pockets. These resonances are shifted downfield by the hyperfine interactions between the corresponding protons and the unpaired electrons of the iron atoms. In deoxy-HbA, the hyperfine shifted resonances at 22.2 and 18.9 ppm have been assigned to the ␤-subunits, and those at 20.3 and 16.8 ppm have been assigned to the ␣-subunits (22)(23)(24). As shown in the figure, in deoxy-␣rHbA the same hyperfine shifted resonances are observed, and their chemical shifts are, within experimental error, the same as in deoxy-HbA. The spectral region from 11 to 15 ppm in Fig. 5 contains several other hyperfine shifted resonances and four exchangeable proton resonances. The hyperfine shifted resonances are significantly broader than the exchangeable proton resonances (i.e. 350 -500 versus 50 -75 Hz). Due to these differences in the line widths and to the spectral overlap, only the exchangeable proton resonances in the spectral region 11-15 ppm in Fig. 5 can be observed accurately. In deoxy-HbA, these four exchangeable proton resonances have been assigned to specific hydrogen bonds in the Hb molecule as follows: the resonance at 14.1 ppm to the hydrogen bond between Tyr ␣42 (C7) and Asp ␤99 (G1) at the ␣ 1 ␤ 2 interface (25,26); the resonance at 13.0 ppm to the hydrogen bond between Asp ␣126 (H9) and Tyr ␤35 (C1) at the ␣ 1 ␤ 1 interface; the resonance 12.2 ppm to the hydrogen bond between His ␣103 (G10) and Asn ␤108 (G10) at the ␣ 1 ␤ 1 interface (26); and the resonance at 11.1 ppm to the hydrogen bond between Asp ␣94 (G1) and Trp ␤37 (C3) at the ␣ 1 ␤ 2 interface (27). 3 As shown in the figure, in deoxy-␣rHbA these four exchangeable proton resonances are very close if not identical to those in deoxy-HbA.
Our NMR results for ␣rHbA in the ligated state are shown in Figs. 6 and 7. Fig. 6 shows the spectral region from 9.5-14.5 ppm. In carboxy-HbA, the resonances at 12.95, 12.1, and 10.2 ppm originate from exchangeable protons and have been assigned as follows: the resonance at 12.95 ppm to the hydrogen bond between Asp ␣126 (H9) and Tyr ␤35 (C1) at the ␣ 1 ␤ 1 interface (26); the resonance at 12.1 ppm to the hydrogen bond between His ␣103 (G10) and Asn ␤108 (G10) at the ␣ 1 ␤ 1 interface (26); and the resonance at 10.23 ppm to the hydrogen bond between Asp ␣94 (G1) and Asn ␤102 (G4) at the ␣ 1 ␤ 2 interface (25). In carboxy-␣rHbA, these exchangeable proton resonances occur at the same spectral positions as those in carboxy-HbA.
The spectra shown in Fig. 6 also contain several resonances from the heme groups. In carboxy-HbA these resonances are (28,29): the resonance at 10.45 ppm (mesoproton ␥ of the heme in ␣-chains); the resonance at 10.1 ppm (mesoproton ␣ of the 3 The assignment of this resonance is currently tentative. For a detailed discussion of the possible origin of this resonance, see Ho (22).  Table I. The ␣-globins were isolated by HPLC, S-pyridylethylated, and digested with TPCK-treated trypsin. Chromatography was performed on a Vydac C 18 column as described under "Materials and Methods."  heme in the ␤-chain); and the resonance at 9.7 ppm (mesoproton ␦ of the hemes in ␣and ␤-chains and mesoproton ␣ of the heme in the ␣-chain). In carboxy-␣rHbA, the latter two heme resonances occur at the same positions as in carboxy-HbA. However, the resonance of the mesoproton ␥ of the heme in ␣-chains (10.45 ppm) is missing from the spectrum. Fig. 7 shows the region from Ϫ2.2 to Ϫ0.4 ppm of the NMR spectra of carboxy-␣rHbA and HbA. This region contains resonances from protons in the heme groups and/or amino acids in the heme pockets that are shifted upfield by the ring-current effect of the heme groups. In carboxy-HbA, the resonance at Ϫ1.77 ppm has been assigned to the ␥ 2 -CH 3 protons of the distal Val residues in the ␣and ␤-subunits, Val ␣62 (E11) and Val ␤67 (E11) (28). As shown in Fig. 7, the position of this resonance in carboxy-␣rHbA is nearly identical to that in carboxy-HbA.
Functional Studies-The binding isotherms of HbA and ␣rHbA were measured using 1.5-2.0 mM heme, a concentration at which dimers and their associated complications are negligible. For these experiments, heme oxydation was below 5%, thus eliminating the need for adding the reducing system. The curves were analyzed according to the Adair equation (17) and yielded the thermodynamic parameters listed in Table II. Although these parameters do not differ significantly within the 66.7% confidence limit, the cooperativity of ␣rHbA is decreased at all oxygen fractional saturations as shown in Fig. 8.
The oxygen binding parameters of HbA and ␣rHbA were measured in the presence of Cl Ϫ and 2,3-diphosphoglycerate.
The data (Table III) indicate that the oxygen affinity and cooperativity of natural HbA and ␣rHbA are similarly affected by these effectors and that these exhibit the same Bohr effect.
CO Recombination Kinetics-The kinetics of CO binding to native HbA and ␣rHbA were measured (Fig. 9) and the data analyzed using a multiexponential model. Each binding curve was well represented by two components whose rates varied by about 30-fold (Table IV). The slower phase originated from CO binding to the tetramer, and the fast phase originated from CO binding to the dimer as described previously (30). Comparison of the rate constants shows that HbA and ␣rHbA exhibit similar rates of CO recombination for both dimers and tetramers.
The kinetic analysis also yields absorbance values that reflect the amounts of tetramer and dimer. These values were used to calculate the dimer-tetramer association constant (K a ) and the accompanying ⌬G value for this transition (Table IV). The kinetics of CO binding to native and recombinant ␣-chains were also measured. Each binding curve was well represented by two components whose parameters are presented in Table V. The major components of the native and recombinant ␣-chains reacted at the same rate (k 2 ϭ 18.2 and 19.7 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively). However, the two samples differed significantly in that the major component corresponded to virtually the entire (97%) native ␣-chain but to only 63% of the recombinant ␣-chain.

DISCUSSION
The long term objective of this work is elucidation of the role of ␣-chains on the structure and function of hemoglobin. Our strategy involved comparison of the expressed and native ␣-chains both as the isolated chain or incorporated into the HbA tetramer. Any observed differences in conformational or functional properties can then be attributed to the ␣-chain. Thus, although the ␣rHbA was almost identical to native HbA, the subtle differences that were observed provide new details on the role of proper ␣-chain folding in maintaining the structure and cooperativity of the hemoglobin molecule. The conformational and functional aspects will be considered separately. Conformational Studies-The NMR spectra in Fig. 5 indicate that the overall conformation of the heme pockets in deoxy-␣rHbA is similar to that in deoxy-HbA. A subtle difference, however, is observed for the relative intensities of the hyperfine shifted resonances at 22.2 and 16.8 ppm in ␣rHbA. In deoxy-HbA, the intensity of the ␣-chain hyperfine shifted resonance at 16.8 ppm is the same as that of the ␤-chain hyperfine shifted resonance at 22.2 ppm (namely, six protons/heme) (31). 4 In contrast, in deoxy-␣rHbA the ␣-chain resonance at 16.8 ppm appears more intense than the ␤-chain resonance at 22.2 ppm.
In order to understand the origin of this difference we have measured the absolute intensity of the hyperfine shifted resonance of ␣rHbA using a reference standard (tris (6,6,7,7,8,8,8-heptafluoro-2,2-dimethyl-3,5-octanedionate) europium) complexed with t-butyl alcohol (31). The results showed that the ␤-chain resonance at 22.2 ppm in ␣rHbA has the same intensity as the corresponding resonance in HbA. In contrast, the intensity of the ␣-chain resonance at 16.8 ppm in ␣rHbA is increased by a factor of approximately 2 relative to that in HbA. The origin of this difference is difficult to ascertain since, at present, it is not known whether the ␣-chain hyperfine shifted resonance at 16.8 ppm originates from protons in the heme groups, from protons in the amino acids in the heme pockets, or from both. Future NMR experiments on HbA molecules containing fully deuterated heme groups may provide an answer to this question.
Similarly, in carboxy-␣rHbA (Fig. 6) the resonance of the mesoproton ␥ of the heme in the ␣-chain (10.45 ppm) is missing from the spectrum, probably shifted to a different spectral position in which it cannot be resolved. This result suggests that in the ligated form the heme environment in the ␣-subunits of ␣rHbA may also differ from that in HbA. Alterations in the heme pockets are also suggested by the ring-current shifted resonances (Fig. 7). All these resonances are broader in ␣rHbA than in native HbA, and in ␣rHbA the resonance at Ϫ0.64 ppm appears to be shifted from its position in HbA. Similar differences in the ring-current shifted resonances have been previously observed for a full recombinant HbA (4) and have been attributed to a different mode of heme insertion. One possibility previously suggested is the existence of two heme orientations that differ by a 180°rotation about the ␣,␥-mesoaxis of the porphyrin (32). Our results for ␣rHbA do not support this suggestion, since a rotation about the ␣,␥-mesoaxis would conserve the environment of the ␥-mesoproton. In contrast, our results show that in ␣rHbA the ␥-mesoproton resonance is shifted from its position in the spectrum of HbA (Fig. 6). Moreover, our results show that the position of the distal (E11)Val residues relative to the heme groups in ligated ␣rHbA is very close to that in HbA. This is consistent with the observed similarity of the CD spectra (Fig. 4A) and indicates that the small difference observed near 410 nm is not due to heme inversion. Further investigations are necessary in order to characterize fully the conformation of the ␣-chain heme pockets in ␣rHbA. Figs. 5 and 6 show that the exchangeable proton resonances of deoxy-␣rHbA are very similar if not identical to those in HbA in both deoxy and carboxyl forms. These findings   indicate that the hydrogen bonds at the ␣ 1 ␤ 2 and ␣ 1 ␤ 1 interfaces that can be monitored by NMR are identical in deoxy-and carboxy-␣rHbA and HbA.
The Soret region of the CD spectrum primarily reflects the interaction of the heme chromophore with the surrounding aromatic residues (21). Thus, it serves as a sensitive probe of correct heme pocket refolding. In the heme pocket, in the absence of amino acid substitutions a decrease in ellipticity has been attributed to the presence of inverted heme (33). The CD spectra of HbA and ␣rHbA have a similar ellipticity (Fig. 4A), suggesting correct heme insertion. However, a small difference in the shape of the spectrum is observed near 410 nm, which suggests the presence of subtle conformational modifications. Although no direct correlation between the NMR and the CD spectra can be made, it is interesting to note that NMR data suggest the correct refolding of the heme pocket, the absence of inverted heme, and the presence of subtle modifications in the heme pocket resonaces.
The CD spectrum of ␣r-chains (Fig. 4B) is broader than that of native ␣-chains and has a decreased ellipticity, suggesting some heme disorder and heterogeneity. The absence of similar heterogeneity in the ␣rHbA suggests that optimal refolding of the ␣-heme pocket occurs upon the tetrameric assembly of the protein.
Functional Studies-The thermodynamic analysis reported in Table II shows that HbA and ␣rHbA exhibit oxygen binding properties that are indistinguishable except for a reduced cooperativity. Since the NMR data indicate identical resonances for residues at the ␣ 1 ␤ 1 and ␣ 1 ␤ 2 interfaces, the reduced cooperativity cannot be attributed to altered quaternary assembly. The free energy of cooperativity, ⌬G c , is the same for HbA and ␣rHbA (Ϫ1.8 Ϯ 0.3 and Ϫ1.6 Ϯ 0.2 kcal/heme). Thus, the decreased value of n max measured for ␣rHbA is probably due to phenomena linked to the intermediate stages of oxygenation. At present the molecular events associated with the intermediates of oxygenation are still elusive, as these partially oxygenated forms are very unstable and their fractional distribution is much smaller than anticipated by a statistical distribution of ligands on tetrameric hemoglobin (34). These results suggest that substitution of the recombinant for the native ␣-globin is sufficient to alter these oxygenation intermediates independently of the quaternary assembly and propose a key role for nonquaternary interactions in the modulation of oxygen cooperativity.
The Bohr effect and the oxygen affinity regulation by allosteric effectors are sensitive probes of tertiary structure and quaternary assembly. In ␣rHbA the Bohr effect and the sensitivity to Cl Ϫ and 2,3-diphosphoglycerate are equivalent to that in native HbA (Table III). This indicates that the Bohr effect groups and the effector binding sites are correctly aligned in the half-recombinant protein. This is probably favored in the half-recombinant hemoglobin by the native partner chains, which help direct the correct refolding and reassembly of the recombinant chains (35).
The CO binding experiments confirm the oxygen equilibrium data ( Fig. 9 and Table IV) and reveal that HbA and ␣rHbA are indistinguishable on the basis of two additional criteria: rate of CO binding and the energetics of dimer-tetramer association. Our observation that the major component of both recombinant and native ␣-chains exhibited the same rate (19.7 and 18.2 ϫ 10 5 M Ϫ1 s Ϫ1 ) as that of the dimers of native HbA (20.5 ϫ 10 5 M Ϫ1 s Ϫ1 ) and ␣rHbA (19.5 ϫ 10 5 M Ϫ1 s Ϫ1 ) is consistent with a similar conformation of isolated ␣-chains, whether monomeric or polymeric, and the ␣-chains within a hemoglobin dimer. The differences observed between the binding kinetics of the expressed and native ␣-chains thus are eliminated during tetramer formation, as the energetics of quaternary assembly more than compensate for the decreased conformational flexibility of ␣r upon its incorporation into the tetramer.
In addition to characterizing HbA, it is also important to determine the properties of the isolated chains. The ability of isolated ␣-globin to refold with heme may perhaps be explained by the results from the CO binding kinetics and CD. The greater heterogeneity in CO binding kinetics of the ␣r-chain suggests a correspondingly greater degree of structural or dynamic heterogeneity relative to that of the native ␣-chain. This result is consistent with the CD measurements, which reflect the time-averaged structure of all ␣-chain components and which show reduced ellipticity and a broader spectrum in the Soret region for the ␣r-chains. Taken together, these approaches suggest the presence of alternating conformations for the isolated ␣r-chain. A dynamic access to a wide conformational space would enhance its prospects for correctly refolding with the heme and ultimately attaining a conformation suitable for pairing with ␤-chain and formation of a functional HbA tetramer. CONCLUSION In conclusion, we have expressed ␣-globin that can refold with heme in the absence of the ␤-chain. The isolated ␣-globin exhibits a wide range of conformational flexibility. Furthermore, this ␣-globin can recombine with the heme and assemble TABLE IV Kinetic parameters, dimer-tetramer equilibrium constants, and dimer-tetramer assembly free energies for HbA and ␣rHbA k 1 and k 2 are the bimolecular association rate constants for CO binding to tetramer and dimer, respectively. a 1 and a 2 are the absorbance parameters for tetramer and dimer, respectively. K a is the dimer-tetramer association equilibrium constant calculated from the a 1 and a 2 values. ⌬G is the dimer-tetramer assembly free energy calculated from the dimer-tetramer association equilibrium constant. Conditions: 0.1 M Bis-Tris and 0.1 M KCl, pH 7.0, 23°C. with ␤-chain to produce a recombinant HbA whose functional and conformational characteristics resemble those of natural HbA. Although subtle modifications in the heme pocket and a decreased cooperativity are observed, the contacts across the ␣ 1 ␤ 1 and ␣ 1 ␤ 2 interfaces are retained. These findings underscore the key role played by the tertiary structure of the heme region, exclusive of the quaternary structure, in modulating the functionality of the HbA molecule. Our results also suggest a key role for nonquaternary interactions in processes involving the relatively unstable oxygenation intermediates. The expressed ␣-globin is thus suitable for construction of mutants to be used in future studies of HbA structure-function relationships.