Sickle Cell Hemoglobin with Mutation at αHis-50 Has Improved Solubility*

Background: Deoxy sickle cell hemoglobin (Hb S) tetramers polymerize in solution via lateral and axial contacts among neighbors. Results: αHis-50 → Gln mutation can, whereas αHis-20 → Gln mutation cannot, improve the solubility of Hb S. Conclusion: α2His-50 interacts with the βS1 CD corner, whereas α2His-20 has minimal interaction with βS1Glu-22. Significance: α2His-50 contributes significantly to the polymerization of the Hb S tetramers. The unliganded tetrameric Hb S has axial and lateral contacts with neighbors and can polymerize in solution. Novel recombinants of Hb S with single amino acid substitutions at the putative axial (recombinant Hb (rHb) (βE6V/αH20R) and rHb (βE6V/αH20Q)) or lateral (rHb (βE6V/αH50Q)) or double amino acid substitutions at both the putative axial and lateral (rHb (βE6V/αH20R/αH50Q) and rHb (βE6V/αH20Q/αH50Q)) contact sites were expressed in Escherichia coli and purified for structural and functional studies. The 1H NMR spectra of the CO and deoxy forms of these mutants indicate that substitutions at either αHis-20 or αHis-50 do not change the subunit interfaces or the heme pockets of the proteins. The double mutants show only slight structural alteration in the β-heme pockets. All mutants have similar cooperativity (n50), alkaline Bohr effect, and autoxidation rate as Hb S. The oxygen binding affinity (P50) of the single mutants is comparable with that of Hb S. The double mutants bind oxygen with slightly higher affinity than Hb S under the acidic conditions. In high salt, rHb (βE6V/αH20R) is the only mutant that has a shorter delay time of polymerization and forms polymers more readily than Hb S with a dextran-Csat value of 1.86 ± 0.20 g/dl. Hb S, rHb (βE6V/αH20Q), rHb (βE6V/αH50Q), rHb (βE6V/αH20R/αH50Q), and rHb (βE6V/αH20Q/αH50Q) have dextran-Csat values of 2.95 ± 0.10, 3.04 ± 0.17, 11.78 ± 0.59, 7.11 ± 0.66, and 10.89 ± 0.83 g/dl, respectively. rHb (βE6V/αH20Q/αH50Q) is even more stable than Hb S under elevated temperature (60 °C).

Hb S 2 is a naturally occurring mutant of human adult hemoglobin (Hb A) that has the normally occurring glutamate at the ␤6 position of Hb A replaced with a valine (1). Heterozygous individuals carrying both Hb S and Hb A genes are protected against Plasmodium infection (2). However, homozygous individuals carrying only the Hb S genes develop hemolytic anemia (3,4). The deoxygenated Hb S molecules form long fibers within the red blood cells, making them rigid and distorting them into the shape of a sickle. This leads to the blocking of capillary vessels and the destabilization of the red blood cell membranes and premature destruction of erythrocytes (5).
EM studies proposed that the Hb S fibers contain 14 (6) or 16 (7) individual strands twisted in a ropelike fashion. These strands can be arranged into seven or eight double strands, which are held together by inter-double strand, intra-double strand axial and intra-double strand lateral interactions (8,9). The crystal structure of deoxy-Hb S has been determined (8,10) and refined to 2.05 Å (11). In this paper, we adopt the designations of Padlan and Love (10,12) in denoting the subunits of the tetramers in the double strands. The ␤-subunits that provide acceptor pockets for the ␤Val-6 are designated as ␤ S 1 . The subunits that act as valine donors are assigned as ␤ S 2 . The ␣-subunits are named according to the ␤-subunits with which they form dimers. According to the 2.05 Å structural model of Harrington et al. (11), each double strand is held together by the lateral interactions among ␤ S 1 -subunits of one strand with the ␤ S 2 -and ␣ 2 -subunits of tetramers from the other strand (Fig.  1A). The individual double strand is further stabilized by axial interactions between tetramers translated one unit cell along the vertical axis. The ␤ S 1 -subunit of the lower tetramer makes contacts with the ␤ S 2 -and the ␣ 2 -subunits of the upper tetramer. In addition, ␣ 1 -and ␣ 2 -subunit interactions are also present (Fig. 1B). The model gives a static picture of interacting amino acid residues that stabilize the polymer.
Numerous natural or recombinant Hb S mutants with substitution(s) at the putative contact sites have been selected or generated for studying the mechanism and inhibition of Hb S polymerization. The hydrophobicity and the size of the side chain of the amino acid at ␤ S 2 6 or the ␤ S 2 4 position affects the solubility of the resulting mutants (13)(14)(15). Solubility can be viewed as the concentration of hemoglobin that remains in solution after completion of polymerization. Mutations at the ␤ S 1 acceptor pocket can improve moderately the solubility of Hb S, but the mutants are inherently unstable (16 -18). Himanen et al. replaced the ␤Lys-95 at the acceptor pocket with Ile and disrupted one of the lateral contact sites (19,20). The mutant is 2.6 times more soluble than Hb S under deoxy conditions (19). Sivaram et al. (21) removed one of the axial interactions by replacing the ␣Leu-113 with histidine and improved the Hb S solubility by 1.8 times. Mutants that carry a replacement at one of the axial contact sites and a substitution at the acceptor pocket (␣P114R/␤T87K and ␣H20Q/␤T87Q) have improved solubility but still less than twice that of Hb S (22,23).
The ␣His-20 and the ␣His-50 residues are located on the surface of the Hb S tetramer. The imidazole nitrogen of ␣ 2 His-20 is putatively in close proximity to the carboxyl side chain of ␤ S 1 Glu-22 (11). A ␤E22Q mutant presumably disrupted part of the axial interactions between Hb S tetramers and showed a moderate increase (1.28-fold) in solubility (24). The ␣ 2 His-50 is postulated in lateral hydrogen-bonding distance from the ␤ S 1 Asp-79 and the ␤ S 1 Asn-80 side chains (11). Surprisingly, Hb S mutants with substitutions at this position have not been generated to study its effect on Hb S polymerization. We report here the structural, functional, and polymerization properties of five recombinant Hb S mutants: rHb (␤E6V/ ␣H20R), rHb (␤E6V/␣H20Q), rHb (␤E6V/␣H50Q), rHb (␤E6V/␣H20R/␣H50Q), and rHb (␤E6V/␣H20Q/␣H50Q). We provide evidence that the putative axial interaction between ␣His-20 and ␤ S Glu-22, as suggested by an x-ray crystallographic study, has minimal contribution in stabilizing the Hb S polymers. The solubility of Hb S can be improved markedly by replacing the ␣His-50 with a glutamine. In particular, rHb (␤E6V/␣H50Q) and rHb (␤E6V/␣H20Q/␣H50Q) are nearly 4 times more soluble than Hb S. Furthermore, rHb (␤E6V/ ␣H50Q) is more heat-stable than Hb S. To our knowledge, rHb (␤E6V/␣H50Q) is the most soluble Hb S mutant reported.

Experimental Procedures
Materials-Blood samples were obtained from the local blood bank (a normal human donor) and the National Institutes of Health (an Hb SS human donor) for the isolation of normal Hb A and Hb S, respectively. Restriction and related enzymes used in molecular biology work were purchased from New Eng- land Biolabs. The QuikChange site-directed mutagenesis kit was a product from Stratagene. Chromatographic materials used in hemoglobin purification were obtained from GE Healthcare. Reagent grade chemicals were purchased from Sigma and used without further purification.
The plasmids were transformed into JM109 cells for protein expression. Cells were grown in DM-4 medium (25) at 30°C in a 20-liter fermenter (B. Braun Biotech International, model Biostat C). The induction and purification of rHb were carried out as described previously (26). Hb A and Hb S from whole blood were purified according to established methods in our laboratory (27,28). Briefly, red blood cells were lysed with water and then loaded onto a Sephadex G-25 column equilibrated and eluted with 0.05 M Tris-HCl, pH 7.45, and 0.1 M NaCl. For the Hb S sample, it was further separated from other components by loading onto a Mono S column equilibrated with 10 mM phosphate, pH 6.8, and 0.5 mM EDTA. The column was developed with the 20 mM phosphate, pH 8.3, and 0.5 mM EDTA. Fractions corresponding to Hb S were collected. The Hb samples were then saturated with CO and stored frozen at Ϫ80°C. The molecular weights of the Hb subunits were confirmed by mass spectrometry in an ion trap instrument equipped with an electrospray ionization source. The amount of N-terminal methionine cleavage of the samples was estimated by Edman degradation (26). All rHb samples in this study had the correct molecular weights and less than 5% unprocessed N-terminal methionine.
Oxygen Binding Properties-Oxygen dissociation curves were determined with a Hemox analyzer (TCS Medical Products) according to Shen et al. (26). The oxy-Hb samples at 0.1 mM Hb (in terms of heme) were prepared in 0.1 M sodium phosphate in the presence of catalase and superoxide dismutase to reduce the amount of met-Hb to less than 5% in all samples measured (29). Experiments were conducted at 29°C as a function of pH. The experimental results were fit to the Adair equation using a nonlinear least-squares procedure. The partial pressure (in mm Hg) at 50% oxygenation (P 50 ) and the Hill coefficient (n 50 ) were determined from each curve. The experimental values had an accuracy of Ϯ10% and S.D. values of Ϯ4% between runs.
Dextran-Csat Assay and Polymerization Kinetic Measurements-Dextran-Csat assays were performed according to Tam et al. (28) for microsample handling. The polymerization progress curves were determined by the temperature-induced method of Adachi and Asakura (30) with modifications. Concentrated oxy-Hb sample (5-50 l) was added to 3 ml of 1.8 M phosphate buffer, pH 7.4, and the solution was purged with nitrogen for 15 min. Sodium dithionite was then added to a final concentration of 10 mM, and the mixture was transferred to a sealed cuvette. The concentration of deoxy-Hb in the sample was determined by light absorbance at 555 nm and calculated by using a millimolar extinction coefficient (mM Ϫ1 cm Ϫ1 ) of 12.5. The cuvette was chilled in ice water, and the temperature was monitored closely with a thermocouple. Once the temperature in the cuvette dropped to 2°C, it was placed in a Cary 50 UV-visible spectrophotometer (Varian) with a cuvette holder preheated and maintained at 30°C. Immediately, light absorbance at 700 nm, which reflects the turbidity of the sample, was recorded at 15-s intervals up to 2000 or 4000 s, depending on the polymerization speed of the sample. Routinely, the temperature of the sample increased from 2 to 30°C in 325 s. The delay time of polyermerization (t d ) was determined from an A 700 versus time plot according to Adachi and Asakura (30).
Autoxidation and Thermal Stability of rHb S Mutants-The autoxidation rate (k ox ) of the samples was determined as described (31). The concentration of oxy-Hb at each time point was calculated according to the equation, [Oxy-Hb] ϭ 0.2174A 560 Ϫ 0.0573A 577 Ϫ 0.1616A 630 and expressed as a percentage of the original sample. The logarithm (base 10) of the percent oxy-Hb was plotted against time, and the slope of the plot gives the autoxidation rate. Thermal stability measurements were performed according to the outline of Adachi et al. (16). Briefly, oxygenated samples (50 l, ϳ20 mg/ml) in 0.1 M sodium phosphate, pH 7.0, and 1 mM EDTA were aliquoted in quadruplicates into thin-walled polypropylene tubes and inserted into a GeneAmp PCR system 2700 (Applied Biosystems). The instrument was programmed to run at 60°C for 10 min and then cooled to 25°C in 1 min. The precipitates in the samples were removed by centrifugation, and the absorbance of the supernatants at 577 and 700 nm were recorded after appropriate dilution. The readings at 700 nm were treated as background and subtracted from the 577 nm readings. The background readings were less than 1% of the absorbance at 577 nm for the most diluted samples. Samples in triplicates without the heating step were used as controls, and the amount of precipitated proteins was calculated accordingly and expressed as a percentage.
Structural Study with 1 H NMR Spectroscopy-Proton NMR spectra of Hb A, Hb S, and its five mutants were obtained at 29°C on Bruker Avance DRX-300 or DRX-600 spectrometers. Hb samples in carbonmonoxy form were concentrated to ϳ5% (3.1 mM in terms of heme) and exchanged into 0.1 M sodium phosphate at pH 7.0 in 95% water and 5% deuterium oxide (D 2 O). Hb S and rHb mutants carrying a substitution at the ␣His-50 position (rHb (␤E6V/␣H50Q), rHb (␤E6V/ ␣H20Q/␣H50Q), and rHb (␤E6V/␣H20R/␣H50Q)) were diluted to 2 and 4% solutions, respectively, before converting into the deoxy form for the NMR measurements. A jumpand-return pulse sequence was used to suppress the water signal (32).

Oxygen Binding and Cooperativity Properties of rHbs-The
O 2 binding properties of Hb A, Hb S, and Hb S recombinant mutants with additional single (rHb (␤E6V/␣H20Q), rHb (␤E6V/␣H20R), and rHb (␤E6V/␣H50Q)) or double (rHb (␤E6V/␣H20Q/␣H50Q) and rHb (␤E6V/␣H20R/␣H50Q)) substitutions are summarized in Table 1. The ␣His-20 and the ␣His-50 are located on the surface of the tetramer (33). As expected, substitutions at these positions do not significantly change the O 2 binding properties of the macromolecules. The O 2 binding curves of the Hb S single mutants are essentially indistinguishable from that of Hb A or Hb S, whereas the Hb S double mutants have slightly higher O 2 binding affinity at low pH. The P 50 values determined for the double mutants at acidic pH are ϳ20% lower than those of Hb S. All mutants in this study have Hill coefficients (n 50 ) of 2.5 or higher (Table  1). These results are indicative that these rHbs are cooperative in binding O 2 .
Hemoglobin releases hydrogen ions upon oxygenation, and this alkaline Bohr effect can be expressed mathematically as ⌬H ϩ ϭ Ϫ⌬log P 50 /⌬pH (34). The numbers of protons released per heme for Hb A, Hb S, and the recombinant mutants are listed in Table 1. The rHbs release 0.43-0.48 proton/heme, whereas Hb A and Hb S release 0.46 and 0.48 proton/heme, respectively. Hence, the mutants we generated are effective oxygen carriers.
Dextran-Csat and Polymerization Kinetic Measurements-The solubilities of Hb S and its recombinant mutants in the presence of dextran and low ionic strength buffer were determined according to the microassay of Tam et al. (28) and are listed in Table 2. Harrington et al. (11) estimated that in the double-stranded form of Hb S fibers, the imidazole nitrogen of the ␣ 2 His-20 is 3.55 Å from the side chain oxygen of ␤ S 1 Glu-22 and suggested that they could form axial interaction. By eliminating the positive charge from the ␣His-20, the rHb (␤E6V/ ␣H20Q) mutant has a Csat value of 3.04 Ϯ 0.17 g/dl, which is close to that of Hb S (2.95 Ϯ 0.10 g/dl). By elongating the side chain, the rHb (␤E6V/␣H20R) mutant apparently has improved interaction with neighboring tetramers, and the Csat value decreased to 1.86 Ϯ 0.20 g/dl. These results imply that the interaction between the ␣His-20 and the ␤Glu-22 of Hb S is minimal under dextran-Csat assay conditions at pH 7.5. In contrast, eradicating the positive charge from the ␣His-50 position improves the solubility of the mutated rHb S. The rHb (␤E6V/ ␣H50Q) mutant has a Csat value of 11.78 Ϯ 0.59 g/dl, which is 4 times higher than Hb S. This substitution can also improve the solubility of the rHb S mutated at the ␣His-20 position. The rHb (␤E6V/␣H20R/␣H50Q) mutant has a Csat value of 7.11 Ϯ 0.66 g/dl, which is 3.8-and 2.4-fold higher than that of the rHb (␤E6V/␣H20R) mutant and Hb S, respectively. The rHb (␤E6V/ ␣H20Q/␣H50Q) mutant has a Csat value of 10.89 Ϯ 0.83 g/dl, which is ϳ3.6-fold higher than that of the rHb (␤E6V/␣H20Q) mutant and Hb S. Therefore, the solubility of Hb S can be improved by disrupting the lateral contacts between the ␣ 2 CD and the ␤ S 1 EF corners. The polymerization kinetics of the rHb S mutants was determined in 1.8 M phosphate buffer, pH 7.4, at 30°C. The increase in turbidity of the sample, as measured by light scattering at 700 nm, reflects sample polymerization. The A 700 values at various time points for each protein concentration can be plotted as a function of time to generate a sigmoidal curve. Some representative results for Hb S and its five mutants are presented in Fig.  2. The intercept of the slope at the x axis gave the delay time of polymerization (t d ). The delay time between the beginning of the experiment and the onset of polymerization depends on the initial hemoglobin concentration and increases as the protein concentration decreases. The logarithmic plots of the reciprocal of the delay time (t d ) versus the concentration of proteins are presented in Fig. 3 and summarize the data for multiple mutants at various concentrations for direct comparison. The plots for rHb (␤E6V/␣H20Q), rHb (␤E6V/␣H50Q), rHb (␤E6V/␣H20Q/␣H50Q), and rHb (␤E6V/␣H20R/␣H50Q) fall to the right of that representing Hb S. The results indicate that these mutants have longer polymerization delay time than Hb S. The plot for rHb (␤E6V/␣H20R) is shifted to the left of the plot for Hb S. The result suggests that rHb (␤E6V/␣H20R) forms polymers faster than Hb S. Among these six plots in Fig.  3, only the line representing rHb (␤E6V/␣H20R) has a slope (1.91) significantly steeper than others (1.46 -1.79). The slope of the line (m) is related to the size of the nucleus in forming aggregates (35,36). Hence, the nuclei formed in the rHb (␤E6V/ ␣H20R) solution are larger than for Hb S and other mutants in this study.
Autoxidation and Thermal Stability-It has been suggested that Hb mutants have the potential to be used in gene therapy for sickle cell anemia (22). It is imperative to determinate the autoxidation rate (k ox ) and stability of the putative candidates. The successful candidate should have k ox and stability similar to that of Hb S, if not Hb A. The k ox of Hb A, Hb S, and its mutants were determined and are listed in Table 2. Hb S has a k ox of 0.0029 Ϯ 0.0002 h Ϫ1 , which is 3.4 times higher than that of Hb A (0.00086 Ϯ 0.0002 h Ϫ1 ). The Hb S mutants in this study all have substitution(s) on the protein surface and away from the heme pocket (33). Consequently, these mutants have k ox ranging from 0.0023 to 0.0031 h Ϫ1 , very similar to that of Hb S.
Reportedly, oxy-Hb S was found mechanically (18) and thermally (16,18) less stable than oxy-Hb A. We examined the thermal stability of the oxy form of Hb A, Hb S, and its mutants in a PCR thermal cycler, which has a more precise temperature control than in previous studies. Hb S has 15.7 Ϯ 0.2% of denatured protein after heating at 60°C for 10 min. Hb A has only 5.8 Ϯ 0.1% denatured protein under the same experimental conditions. The results for the Hb S mutants are listed in Table  2. With the exception of rHb (␤E6V/␣H50Q), all mutants have thermal stability equal to or better than Hb S. In particular, rHb (␤E6V/␣H20Q/␣H50Q) has only 12.0 Ϯ 0.3% denatured protein under the same treatment, significantly better than Hb S. Apparently, Hb S mutants with substitution(s) on the protein surface are thermally more stable than Hb S with an additional mutation in the acceptor pocket (16,18).
Structural Properties Investigated with 1 H NMR-The possible conformational changes of the rHb S mutants were investigated with 1 H NMR. ␣His-122 and ␣His-103 are located at the ␣ 1 ␤ 1 interface, and the NH ⑀1 of their side chains give signals at 12.9 and 12.1 ppm, respectively (37)(38)(39). The exchangeable proton resonance spectra of the CO form of the samples show clearly that these two amino acid residues have not been disturbed in the five mutants studied (Fig. 4A).
␤Trp-37 is located at the ␣ 1 ␤ 2 interface, and its NH ⑀1 atom gives a resonance peak at 10.6 ppm (39,40). As demonstrated in Fig. 4A, the resonances between 9.5 and 11.0 ppm are similar among Hb A, Hb S, and the five recombinant mutants. There-fore, substitutions at ␣His-20 and ␣His-50 do not change the quaternary structure of the liganded Hb A at the ␣ 1 ␤ 2 interface.
Resonances from 0 to Ϫ3 ppm belong to the non-exchangeable ring current-shifted protons, and they provide insight into the tertiary structure of the heme pockets. The ␥ 2 -CH 3 groups of ␣Val-62 and ␤Val-67 of the distal heme pocket give resonances at Ϫ1.75 and Ϫ1.82 ppm relative to 2,2-dimethyl-2silapentanesulfonic acid, respectively (41,42). The signals at Ϫ1.75 ppm are similar among all six proteins studied. However, the signal at Ϫ1.82 ppm is shifted slightly upfield for the rHb (␤E6V/␣H20R/␣H50Q) and the rHb (␤E6V/␣H20Q/␣H50Q) mutants, although the mutations are on the ␣-subunit (Fig. 4B). Moreover, the resonances between Ϫ0.7 and Ϫ1.1 ppm for the double mutants differ significantly from those for the other five proteins. But, we have not yet assigned the resonances for these signals and cannot comment on their importance.
The 1 H NMR spectra for the deoxy form of Hb A, Hb S, rHb (␤E6V/␣H50Q), rHb (␤E6V/␣H20R/␣H50Q), and rHb (␤E6V/ ␣H20Q/␣H50Q) are presented in Fig. 5. The deoxy forms of Hb S mutants with a single substitution at ␣His-20 are at least 3-fold less soluble than mutants carrying a substitution at the ␣His-50 position, and they were excluded from the NMR experiments. The hyperfine-shifted proton resonances of the N ␦ H exchangeable proton of ␣His-87 and ␤His-92 in the proximal heme pocket appear at 63 and 76 ppm from 2,2-dimethyl-2-silapentanesulfonic acid, respectively (43,44). The signals for the mutants, Hb S, and Hb A appear at the same positions (Fig.  5A). Fig. 5B presents the spectral region of the deoxy form of the proteins between 11 and 25 ppm downfield from 2,2-dimethyl-2-silapentanesulfonic acid. The porphyrin rings of the ␣and ␤-subunits resonate at 17.0 and 22.6 ppm, respectively (45). The signals at 14.1 and 11.2 ppm have been assigned to the hydrogen bonds between ␣Tyr-42 and ␤Asp-99 and between ␤Trp-37 and ␣Asp-94, respectively. These residues are located in the ␣ 1 ␤ 2 interface, and the resonances are important T-state markers of the deoxy-Hb A (39,46,47). We did not detect any difference in chemical shift for these signals among the five proteins. The signals at 12.9 and 12.1 ppm belong to ␣His-122 and ␣His-103, as in the spectra for the liganded form of the molecules. Again, we cannot detect any shift in the signals for the mutants.
In summary, the ␤-heme pockets of the liganded form of rHb (␤E6V/␣H20R/␣H50Q) and rHb (␤E6V/␣H20Q/␣H50Q) changed slightly when compared with that of Hb A or Hb S.  Besides that, the tertiary and quaternary structures of the four mutants are similar to that of Hb A or Hb S.

Discussion
The oxy forms of Hb A and Hb S have similar solubility. Upon deoxygenation, the solubility of Hb A was decreased by half, whereas that of Hb S diminished by 100 times (48). The primary sequences of Hb S and Hb A differ in a single amino acid. Hb S has a hydrophobic valine instead of a hydrophilic glutamic acid at the ␤6 position. Furthermore, mutants with isoleucine (49) and lysine (13,50) at the ␤6 position have lower and higher solubility, respectively, than Hb S. There-  fore, lowering solubility and forming polymers can be attributed to this single amino acid substitution, a direct result of introducing undesirable interactions among surface residues of neighboring tetramers.
We have generated five Hb S mutants for this study. Three of them carry an additional substitution at the ␣20 or the ␣50 position (rHb (␤E6V/␣H20R), rHb (␤E6V/␣H20Q), and rHb (␤E6V/␣H50Q)), whereas the other two carry mutations at both ␣20 and ␣50 positions (rHb (␤E6V/␣H20R/␣H50Q) and rHb (␤E6V/␣H20Q/␣H50Q)). These Hb S mutants have substituted amino acid(s) on the surface of the tetrameric protein and away from the heme pockets and the ␣ 1 ␤ 1 and the ␣ 1 ␤ 2 interfaces. Hence, these mutants are structurally (Figs. 4 and 5) and functionally ( Fig. 2 and Tables 1 and 2) similar to Hb S. Their autoxidation rate is similar to that of Hb S. For some mutants (rHb (␤E6V/␣H20Q), rHb (␤E6V/␣H20Q), and rHb (␤E6V/␣H20Q/␣H50Q), they are thermally more stable than Hb S. EM studies (6,51) indicate that the double strands within the Hb S fibers have a slight helical twist. The results give an overall conformation of the fiber but cannot pinpoint the interacting residues with certainty. Data from x-ray crystallography have higher resolution, but the double strands found in the Hb S crystals are straight and do not have the helical twist (10 -12).
Harrington et al. (11) refined the crystal structure of deoxy-Hb S to 2.05 Å and then employed the program CON-TACT to identify inter-and intrastrand interacting residues. With this approach, 34 residues from the ␤ 1 /␤ 2 -subunits and 9 residues from the ␣ 1 /␣ 2 -subunits are suggested to be within interacting distances either laterally or axially among Hb S tetramers. These putative interacting residues are listed in Table 3.
According to the model of Harrington et al. (11), the axial contacts can be grouped as follows: (i) the GH corner (␣ 1 Pro-114 and ␣ 1 Ala-115) of the ␣ 1 -subunit of the lower tetramer interacts with the GH corner (␣ 2 Glu-116) and A helix (␣ 2 Lys-16) of the ␣ 2 -subunit of the upper tetramer; (ii) the A helix (␤ S 1 Gly-16, ␤ S 1 Lys-17, and ␤ S 1 Val-18) of the ␤ S 1 -subunit of the lower tetramer interacts with the G helix (␤ S 2 His-116 and ␤ S 2 His-117) and the GH corner (␤ S 2 Phe-118 and ␤ S 2 Lys-120) of the ␤ S 2 -subunit of the upper tetramer; (iii) the G helix (␤ S 1 His-117) and the GH corner (␤ S 1 Phe-118, ␤ S 1 Gly-119, and ␤ S 1 Glu-121) of the ␤ S 1 -subunit of the lower tetramer interacts with the GH corner (␣ 2 Pro-114 and ␣ 2 Ala-115) of the ␣ 2 -subunit of the upper tetramer; and (iv) the B helix (␤ S 1 Glu-22) of the ␤ S 1 -subunit of the lower tetramer interacts with the B helix (␣ 2 His-20) of the ␣ 2 -subunit of the upper tetramer. These interacting residues are presented as a diagram in Fig. 1B. Because the coordinates in the Protein Data Bank file 2HBS cannot pinpoint the relative positions of the interacting residues, we summarized the interacting residues as did Wishner et al. (8).
The lateral interactions contributing to the Hb S polymerization are mainly between ␤ S 1 of one strand and the ␤ S 2 of the neighboring strand. Seven residues from the A helix, including the substituted ␤Val-6, plus two residues from the H helix of ␤ S 2 interact with 13 residues from the E and F helices and the EF and FG corners of the neighboring ␤ S 1 -subunit (Table 3 and Fig. 1A). The interactions between the ␤ S 1 EF corner (␤Asp-79 and ␤Asn-80) and the ␣ 2 CD corner (␣Ser-49 and ␣His-50) constitute the only lateral contact between ␤ S -and ␣-subunits. In particular, the side chains of ␤ S 1 Asp-79 and ␣ 2 His-50 are reportedly separated by only 2.83 Å (11). However, according to the deposited coordinates (Protein Data Bank code 2HBS) of the two Hb S tetramers in the unit cell, the imidazole nitrogen of ␣ 2 His-50 is 6 Å from the side chain oxygen of ␤ S 1 Asp-79. The distance between the side chains of ␣ 2 His-50 and ␤ S 1 Asn-80 is shorter (3.86 Å) but still too far for hydrogen bonding. Therefore, these results generated from docking calculations should be confirmed with biochemical and mutagenesis studies.
Careful scrutiny of the docking results reveals that only five (␣Lys-16, ␣His-20, ␣Pro-114, ␣Ala-115, and ␣Glu-116) and two (␣Ser-49 and ␣His-50) residues on the ␣-subunits contribute to either axial or lateral interactions, respectively. Russu and Ho (52) have suggested the presence of histidyl residues at contact areas between Hb S molecules in pregelation aggregates. Therefore, ␣His-20 and ␣His-50, the only two histidines among the list, were selected for mutation, and its effect on Hb S solubility was studied.
Presumably, ␣ 2 His-20 and ␤ S 1 Glu-22 form an interacting pair in the Hb S fibers. Nagel et al. (53) compared the ␤ and ␥ sequences of the hemoglobin subunit and determined the minimum gelling concentration of a mixture of Hb S and Hb G Coushatta (␤Glu-22 3 Ala). They suggested that ␤ S Glu-22 possibly partially stabilized the deoxy-Hb S polymers. Acharya and Seetharam (54) modified the ␤ S Glu-22 and the ␤ S Glu-43 with glucosamine, and they observed an increase in the solubility of the modified deoxy-Hb S by about 55%. However, the addition of glucosamine not only eliminated the carboxyl side chain of ␤ S Glu-22 but appended also a bulky carbohydrate derivative that might interfere with subunit-subunit interactions. Therefore, whether disrupting the putative ␣ 2 His-20 and ␤ S 1 Glu-22 interaction can improve Hb S solubility is debatable. Hb S carrying mutations at the ␣20 position have been studied (23,24,55). Rhonda et al. (55) measured the solubility of an equal mixture of Hb S and Hb Le Lamentin (Hb S with an ␣His-20 3 Gln mutation) and claimed a 1.6-fold increase in Csat value. Srinivasulu et al. (23) measured the solubility of Hb Le Lamentin with an oxygen affinity method and reported only a 26% increase in Csat. Banerjee et al. (24) used the dextran-Csat method of Bookchin et al. (56) and reported a 28% increase in Csat value for the same mutant.
We used the dextran-Csat method of Tam et al. (28) developed for microsample handling in this study. Our Hb Le Lamentin (rHb (␤E6V/␣H20Q)) has a Csat value similar to that of Hb S without any pronounced improvement (Table 2), although the polymers formed more slowly than for Hb S (Fig.  3). The 3.55-Å distance (11) between the N ⑀2 atom of the ␣ 2 His-20 and the O ⑀2 atom of the ␤ 1 Glu-22 is possibly too long for optimal hydrogen bond formation. However, the structure of Hb in solution is flexible, and its backbone and side chains can exhibit various types of motions (see Ref. 57 for a review). Hence, it is conceivable that a subpopulation of the Hb possibly assumes a conformation with a shorter distance between ␣ 2 His-20 and ␤ 1 Glu-22 for hydrogen bonding. The conservative substitution of Gln for His at the ␣20 position disrupts this interaction and abolished the contribution of ␣ 2 His-20 and ␤ 1 Glu-22 to the polymerization process, resulting in a lengthening in the delay time of polymerization.
To further explore the possible interaction between ␣ 2 His-20 and ␤ S 1 Glu-22, we generated the rHb (␤E6V/␣H20R) mutant. rHb (␤E6V/␣H20R) has a Csat value of 1.86 Ϯ 0.20 g/dl (Table 2), which is only two-thirds of that of Hb S. It has also a shorter delay time of polymerization, and the size of nucleus in forming aggregates is larger than for Hb S (Fig. 3). Hence, the interactions between neighboring Hb S tetramers are intensified by introducing a longer side chain carrying a positive charge at the ␣20 position. Adachi et al. (58) have investigated the polymerization of Hb S in the presence of Hb A 2 variants carrying Val, Glu, and Thr at positions 6, 22, and 87 of the ␦-subunit. They concluded also that it is not necessary to change Glu to Ala at the ␤22 position in preparing anti-sickling hemoglobin. Therefore, we conclude that the interaction between ␣ 2 His-20 and ␤ S 1 Glu-22 contributes minimally to the polymerization of Hb S.
The lateral interactions among residues on the ␤ S 2 -subunit with the neighboring ␤ S 1 EF pocket have been studied extensively. For instance, mutants with an aromatic residue (phenylalanine or tryptophan) at the ␤6 position are more soluble than Hb S (59). Probably, it is more difficult to insert a bulky side chain into the EF pocket. The observation that the rHb (␤E6V/ ␤L88F) mutant, with an aromatic residue in the EF pocket, also has higher solubility substantiates this speculation (16). Other residues on or near the acceptor site have also been mutated to disrupt the lateral interactions in various studies. In general, the mutants have moderately improved solubility (  Putative amino acid residues that participate in lateral and axial contacts in double strand Hb S polymers (11) Interacting residues are highlighted with the same color or connected with doubleheaded arrows. 61), but some of them reportedly suffered from thermal instability (16,18,61).
According to the model proposed by Harrington et al. (11), ␤ S 1 Asp-79 interacts with both ␣ 2 Ser-49 and ␣ 2 His-50, whereas ␤ S 1 Asn-80 interacts only with ␣ 2 His-50. Nagel et al. (53) have investigated the solubility of Hb S in the presence of an Hb A 2 mutant containing a ␤Asn-80 3 Lys mutation and observed a 19% increase in solubility. McCune et al. (62) generated an Hb S mutant with the addition of ␤Glu-22 3 Ala and ␤Asn-80 3 Lys substitutions. This mutant is better than Hb A but less effective than Hb F in inhibiting the polymerization of Hb S. Srinivasulu et al. (63) substituted the ␣Ser-49 on Hb S with Arg and observed a 60% increase in solubility. However, the contribution of ␣ 2 His-50 in stabilizing the Hb S polymer has never been tested.
To address this issue, we generated the rHb (␤E6V/␣H50Q) mutant, which has a Csat value 4 times higher than that of Hb S ( Table 2). The results show clearly that removing the positive charge on ␣His-50 interrupts its interaction with the neighboring ␤ S 1 Asp-79 and/or ␤ S 1 Asn-80. Furthermore, this substitution at the ␣His-50 position also improves the solubility of rHb (␤E6V/␣H20R) and rHb (␤E6V/␣H20Q). The resulting rHb (␤E6V/␣H20Q/␣H50Q) mutant has a dextran-Csat value of 10.89 Ϯ 0.80 g/dl, which is 3.7 times higher than Hb S. It has similar autoxidation rate and better thermal stability than Hb S ( Table 2). The rHb (␤E6V/␣H20R/␣H50Q) mutant has a dextran-Csat value of 7.11 Ϯ 0.66 g/dl (Table 2), which is 3.8 times better than that of the rHb (␤E6V/␣H20R) mutant. Both mutants have a longer delay time of polymerization compared with that of Hb S (Fig. 3). Furthermore, the results also suggest that the effects of mutation at positions ␣20 and ␣50 are independent of each other.
In summary, the results in this report show that (i) the interaction between ␣ 2 His-20 and ␤ S 1 Glu-22 is minimal in stabilizing the Hb S polymer, and (ii) the Hb S polymer can be destabilized by replacing the histidyl residue at the ␣50 position with a Gln. The resulting protein is functionally and structurally similar to Hb S. The mutant carrying the ␣His-50 3 Gln substitution can be a candidate for gene therapy of sickle cell disease.
Author Contributions-M. F. T. and C. H. designed the study and wrote the paper. T. C. S. T. purified all of the samples in this study and performed and analyzed the experiments in Table 2. V. S. collected and analyzed all NMR spectra. N. T. C. performed and analyzed the experiments shown in Table 1