Linkage of interactions in sickle hemoglobin fiber assembly: inhibitory effect emanating from mutations in the AB region of the alpha-chain is annulled by a mutation at its EF corner.

The AB and GH regions of the alpha-chain are located in spatial proximity and contain a cluster of intermolecular contact residues of the sickle hemoglobin (HbS) fiber. We have examined the role of dynamics of AB/GH region on HbS polymerization through simultaneous replacement of non-contact Ala(19) and Ala(21) of the AB corner with more flexible Gly or rigid alpha-aminoisobutyric acid (Aib) residues. The polymerization behavior of HbS with Aib substitutions was similar to the native HbS. In contrast, Gly substitutions inhibited HbS polymerization. Molecular dynamics simulation studies of alpha-chains indicated that coordinated motion of AB and GH region residues present in native (Ala) as well as in Aib mutant was disrupted in the Gly mutant. The inhibitory effect due to Gly substitutions was further explored in triple mutants that included mutation of an inter-doublestrand contact (alphaAsn(78) --> His or Gln) at the EF corner. Although the inhibitory effect of Gly substitutions in the triple mutant was unaffected in the presence of alphaGln(78), His at this site almost abrogated its inhibitory potential. The polymerization studies of point mutants (alphaGln(78) --> His) indicated that the inhibitory effect due to Gly substitutions in the triple mutant was synergistically compensated for by the polymerization-enhancing activity of His(78). Similar synergistic coupling, between alphaHis(78) and an intra-double-strand contact point (alpha16) mutation located in the AB region, was also observed. Thus, two conclusions are made: (i) Gly mutations at the AB corner inhibit HbS polymerization by perturbing the dynamics of the AB/GH region, and (ii) perturbations of AB region (through changes in dynamics of the AB/GH region or abolition of a specific fiber contact site) that influence HbS polymerization do so in concert with alpha78 site at the EF corner. The overall results provide insights about the interaction-linkage between distant regions of the HbS tetramer in fiber assembly.

Sickle cell anemia arises because of a point mutation at the sixth position of the ␤-chain (␤Glu 6 3 Val) in the hemoglobin (Hb) 1 molecule (1). Deoxygenated sickle hemoglobin (HbS) polymerizes into long helical fibers that are believed to be responsible for the pathophysiology of the sickle cell disease. The knowledge gleaned so far from structural analysis of HbS crystal (2)(3)(4), solution polymerization studies of natural variants or engineered mutant hemoglobins (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24), and electron microscopic studies (25) have led to a 14-stranded model of the fiber (26,27). These strands appear as seven double strands of the type found in HbS crystals, albeit with a slight twist caused by fiber packing. The models specify several amino acid residues from both ␣and ␤-chains that participate in inter-or intradouble-strand contacts stabilizing the fiber structure. By and large, the fiber models agree with the solution polymerization experiments of mutant hemoglobins in that polymerizationsensitive mutations are usually located in fiber contact regions (27). However, the models present an approximate description of the HbS fiber that are limited to the identity of the contact residues and totally lack information on the hierarchy of interactions or additive or synergistic effect of two or more contact residues (4,26,27). Furthermore, the models project a static view of the contact sites and do not provide any clues about the influence of flexibility or dynamics of a given contact region on fiber assembly. Here we have examined the consequence of flexibility changes at the boundary of the A-and B-helix (AB region) of the ␣-chain on the polymerization of HbS. The linkage of interaction in the polymerization process, between the AB region and a remote site located at the EF corner of the ␣-chain, has also been explored.
The AB region of the ␣-chain contains several contact sites of the fiber (␣16, ␣20, and ␣23) and lies in spatial proximity to a cluster of GH corner residues that are implicated in fiber interactions (residues ␣113-␣116). Besides, AB-GH corners of the ␣-chain engage in extensive inter-tetrameric interactions with AB-GH corners of the ␤-chain, yielding physiologically important intra-double-strand axial contacts of the fiber. Therefore, it is conceivable that even subtle alterations of the conformational dynamics generated through rational amino acid changes at non-contact sites at the AB or GH corners of ␣/␤ chains might influence the HbS polymerization by perturbing the inter-tetrameric interaction interface of the fiber. We have tested this surmise by replacing two Ala residues of the AB corner of the ␣-chain, Ala 19 (AB1) and Ala 21 (B2), with the relatively more flexible Gly residues or less flexible ␣-aminoisobutyric acid (Aib) residues (Fig. 1A).
Aib (also called ␣-methylalanine), an analog of alanine, is a non-protein amino acid found in microbial peptides that form membrane channels. Aib exhibits higher helix-forming ability compared with Ala residues, which has the highest helical propensity among the 20 coded amino acids. Model studies on short peptides have shown that incorporation of one or more Aib residues significantly enhance the helical backbone structure (28,29). The stabilizing effect of Aib emanates from the presence of an extra methyl group (relative to Ala) that restricts the peptide backbone torsional angles (, ) to helical conformation. In fact, Gly 3 Ala substitutions also introduce a methyl group on C␣ atom, and similar effects, albeit to a lesser extent than Aib, are realized due to Ala. Therefore, a hierarchy of conformational constraint is expected to be generated at the AB corner while traversing from Gly to Ala to Aib; the ensuing sequence of residues 18 -22 at the AB corner ( Fig. 1B) with Gly substitution (GGHGG) is likely to be potentially more flexible than the native sequence (GAHAG), which in turn, would be more flexible than the sequence generated by Aib replacements (GAibHAibG).
We have earlier successfully employed a semi-synthetic reaction, namely, V8 protease-mediated coupling of ␣1-30 and ␣31-141 fragments to produce ␣141 (␣-globin), for the construction of mutant ␣-subunits (23,24). This strategy, although restrictive in nature, is especially suited for the incorporation of non-coded Aib residues and, in appropriate situations, could be effectively utilized for the introduction of 20 standard amino acids. Accordingly, we have constructed the aforementioned ␣-globin mutants containing Gly or Aib residues at 19 and 21 positions through the ␣-globin semisynthetic reaction. We have also taken advantage of the fortuitous co-presence of Gly residues at the 19/21 positions and a contact point mutation at position 78 (EF7, Asn 3 His or Gln) in the ␣-chain of the Hanuman langur (Presbytus entellus) hemoglobin (30) to authenticate the role of conformational flexibility of the AB region as well as to explore linkage of interactions, if any, between these two remote regions of the tetramer.

MATERIALS AND METHODS
The ion-exchange resins (CM52 and DE52) were obtained from Whatman. The chemicals used in peptide synthesis were obtained from Nova Biochem. The hemoglobins from homozygous (SS) sickle cell patients and Hanuman langur (HL) were purified from their red cell lysates by DE52 and CM52 chromatography according to standard procedures. Heme-bound individual subunits of hemoglobins were obtained as hydroxymercuribenzoate derivatives following the procedure described by Bucci (31). The heme-free chains were obtained by acidacetone precipitation.
Isolation and Characterization of HL ␣-Chain-Hemoglobin from the HL erythrocyte lysate was subjected to CM52 cation-exchange chromatography, and the elution profile was monitored at 540 nm. The chromatographic profile revealed the presence of two hemoglobin components that were present in the ratio of 45:55. This finding was unexpected in view of the fact that HL is known to contain a single hemoglobin. To locate the "variant" chain(s), the two hemoglobins (referred to as HL1 and HL2 in the order of elution) from CM52 chromatography were subjected to RP-HPLC for separation of their ␣ and ␤ globins. The RP-HPLC profile of the above HL hemoglobin samples (HL1 and HL2) displayed two globin peaks with 1:1 stoichiometry. The globin peaks from respective hemoglobins were collected and subjected to positive ion electro-spray mass spectrometry (ESMS). The experi- mental mass of the first peak in each case (15,896.44 Da and 15,896.39 Da, respectively) was similar and in agreement with the reported sequence (calculated mass, 15,895.24 Da) of langur ␤-globin. In contrast, the mass of the second peak of RP-HPLC, corresponding to the ␣-chain from the two samples, was different. Interestingly, the mass of this chain from HL2, 15,123.03 Da, fitted well to the calculated mass (15,121.36 Da) of the documented sequence of the HL ␣-chain (30) that differs from human at only three sites (Gly 19 , Gly 21 , His 78 ). The experimental mass of the ␣-chain from the HL1 sample was, however, less by about 10 Da (15,113.81). Thus the heterogeneity in the HL hemoglobin was localized in the ␣-chain.
To define the region of sequence change(s), the two ␣-chains of HL were subjected to tryptic peptide mapping (Fig. 2). A comparison of the peptide maps of the two HL chains revealed altered retention times for two peptides namely, T8ϩ9 (residues 61-90) and T9 (residues 62-90), suggesting that the origin of sequence variation in the two chains was the region encompassing residues 61-90. Further, the positions of the three peptides, T4 (17-31), T8ϩ9, and T9 were distinct between the human and the two HL chains. This result is consistent with the fact that the HL ␣-chain sequence is known to differ from human at positions 19, 21, and 78. For locating exactly the site of sequence changes in the variant chain, T8ϩ9 and T9 peptide fragments were isolated by RP-HPLC and subjected to 25 cycles of Edman degradation. The sequence assignment of 22 and 25 residues in T9 and T8ϩ9, respectively, were made unambiguously and were found to be identical to the human sequence, except at the 78th position. Although the 17th and 18th cycles of amino acid sequencing of T9 and T8ϩ9 (corresponding to the 78th position of the ␣-chain) peptides derived from HL1 ␣-chain yielded a Gln residue, the same peptides of HL2 ␣-chain yielded His. The sequencing results were further confirmed by mass measurements. T8ϩ9 fragment obtained from human sequence had a mass of 3124.89 Da, which was expected for the presence of Asn at 78th position (calculated mass, 3125.51). In the case of HL1␣ and HL2␣, the masses of the T8ϩ9 fragments were found to be 3139.22 and 3148.94 Da, respectively. These values fit the Asn 3 Gln or His substitutions in T8ϩ9 fragment and are also in agreement with the mass of the corresponding intact ␣-chains. Thus, the presence of two kinds of ␣-chain in HL hemoglobin was unequivocally established. Although these chains show three sequence changes relative to the human ␣-chain (Gly 19 , Gly 21 , and His 78 / Gln 78 ), they differ from each other only at a single residue at position 78 (His or Gln).
Semi-synthesis of Mutant ␣-Chains-The complementary fragments of ␣-globin, ␣1-30 and ␣31-141, were obtained by subjecting the V8 protease digest of ␣-globin to Sephadex G50 chromatography (24). The sequence comprising ␣1-30 with Gly or Aib residues at positions 19 and 21 was assembled by solid-phase peptide synthesis. The respective purified peptides were ligated with the complementary fragment (␣31-141) derived from human or langur chains by V8 protease-mediated ligation reaction (23,24). Semi-synthetic mutant globins were purified by CM52 chromatography in the presence of 8 M urea as described previously (24). ESMS, tryptic peptide mapping, and amino acid sequencing were employed to establish the chemical identity of the mutant chains.
Generation of Mutant HbS Tetramers-The heme-free mutant globins were reconstituted with ␤ s -chain and heme to tetrameric mutant HbS through the "alloplex" route (32) and according to the protocol described by Sivaram et al. (24). HbS tetramers composed of hemebound HL ␣-chain were prepared by incubating the ␣-chain (0.5 mM) with 1.2-fold molar excess of ␤ s -chain (50 mM Tris-HCl, pH 7.4, containing 1 mM dithiothreitol and 1 mM EDTA) at 4°C for 24 h. The tetramers were obtained in pure form by CM52 chromatography. The purity was further established by analytical anion-exchange chromatography on a Mono Q HR 5/5 column using FPLC (AKTA, Amersham Biosciences).
Circular Dichroism (CD) Spectroscopy-CD spectra of HbS constructs were recorded on a J710 spectropolarimeter using a 1-mm path-length cell. The respective buffer baselines were subtracted from the sample data, and the final spectrum was represented as mean residue ellipticity in degree/cm 2 /dmol units.
Oxygen-affinity Measurements-Oxygen-affinity measurements were done on a Hemox analyzer (TCS Medical Products, New Hope, PA) in Bis-Tris buffer, pH 7.4. The value of partial oxygen pressure at 50% saturation (P 50 ) and Hill coefficient (n H ) was calculated from each dissociation curve.
Measurements of Gelation Concentration (C sat )-The C sat was measured in the presence of 12 g/dl concentration of dextran (70 kDa) as described by Bookchin et al. (33). This quasi-physiological method has been amply demonstrated to produce the same information as that of standard C sat measurements but at a much-reduced hemoglobin concentration. HbS samples were taken in 50 mM phosphate buffer (pH 7.5) and were allowed to polymerize in the presence of sodium dithionite at 37°C. The concentration of HbS in soluble phase (C sat ) was determined after centrifugation of the gelation mixture at room temperature.
Molecular Dynamics (MD) Simulation-MD simulations on the ␣-chain of HbS and the mutants were carried out using the GROMACS suite of programs (34). The initial structure of the native protein was taken from the high-resolution x-ray crystal structure (4) of HbS (Protein Data Bank code 2HBS). The structures of the mutants were generated interactively using Insight II (Accelrys). The initial model structures were placed in a simulation box of size 42.8 ϫ 31.5 ϫ 42.7 Å 3 . The closest distance from any protein atom to the walls of the box was not less than 9 Å. The system was then solvated by adding a bath of simple point charge (35) waters in such a way that the density of the system was as close to 1 as possible. The ␣-chain of HbS including the heme has an equal number of negative and positive charges; hence, there was no need of adding counterions for ensuring charge neutrality. The system was first relaxed by energy minimization using the steepest descent algorithm (1000 steps). Subsequently, the added solvent around the protein was allowed to equilibrate by means of position-restrained MD for 25 ps and energy was minimized again with steepest descents (1000 steps). Constant-volume MD simulations were carried out for 2 ns at 300 K with periodic boundary conditions imposed. The bond lengths and bond angles of the protein were restrained around their equilibrium values.

Rationale and Design of Study-
The hemoglobin molecules in the crystal pack as two strands of HbS molecules. Hemoglobin tetramers in each strand are connected through a network of axial contacts (4). These contacts involve AB and GH regions of both the ␣and ␤-chains that include ␣Lys 16 , ␣His 20 , ␣Pro 114 , ␣Ala 115 , ␣Glu 116 , ␤Gly 16 , ␤Lys 17 , ␤Glu 22 , ␤His 117 , ␤Phe 118 , ␤Gly 119 , and ␤Glu 121 . The presence of a large number of contacts at the interface between the two HbS molecules in each strand of the fiber raises the intriguing possibility that even subtle changes in the conformational dynamics generated because of the perturbation of non-contact residues around AB or GH regions of the chains might influence the polymerization process by interfering with inter-tetrameric interaction interface of the fiber.
We chose sites Ala 19 and Ala 21 at the AB region of the ␣-chain to address the above issue based on the fact that these are not contact points themselves, but their replacements with Gly residues generate a sequence ( 18 GGHGG 22 ) that has the propensity to enhance the conformational flexibility of the AB region. Secondly, the readily available V8 protease-mediated ␣-globin semisynthetic procedure permits us to incorporate non-coded Aib residues that are expected to contrast the effects because of Gly residues and confer conformational rigidity to the AB region.
Molecular Dynamics Simulations of the Mutant ␣-Chains-We carried out MD simulations studies on native (Ala) and mutant (Gly or Aib) ␣-chains to investigate the extent of dynamic perturbations in the mutant chains. The MD simulations for each mutant ␣-chain were checked for stability by monitoring several global properties like the radius of gyration, root mean squared deviation from the initial structure as well as the kinetic and potential energies. These parameters were found to stabilize after about 0.1 ns and, hence, data from 0.1-2 ns were used for all subsequent analyses. The root-mean-squared fluctuations (RMSF) of the C␣ carbon atoms were calculated for the native (␣A 19 -A 21 ) and mutant chains (␣A19G/A21G and ␣A19Aib/A21Aib) and is shown in Fig. 3A. The crystallographic temperature factors for the starting structure are also shown for comparison (Fig. 3B). Structural fluctuations in the mutant ␣-chains are, overall, similar to those of the native chain and follow the same trend as that seen in the experimental structure, suggesting that the introduction of mutations did not cause any large-scale deviation of the dynamical properties of the ␣-chain.
The regions around residues 42 and 93 within the CD and FG corner, respectively, show the most fluctuations and a large difference between the native ␣-chain and its mutants. Both of these regions are particularly close to the highly mobile propi-onate side chains of the heme moiety. The RMSF difference in these regions also varies when different time windows of the simulation were used for calculation. This result indicates that, perhaps because of strong interaction with the highly mobile propionate side chains, the dynamics in these regions becomes much more complicated, and it was not possible to sample fully the total conformational space even during 2 ns of MD simulation, resulting in the observed difference in conformational fluctuation between the native ␣-chain and its mutants. The RMSF values in the AB corner regions rise to a local maximum in all three trajectories (Ala, Gly, Aib), indicating that this region is relatively mobile. Introduction of Ala 3 Gly and Ala 3 Aib mutations do cause small differences in RMSF values in this region. The GH corner region also lies in a local maximum of the RMSF plot. Although the average fluctuation in the GH region is comparable between the three trajectories, a significant difference between the fluctuation profile of the AB and the GH region is observed. This is clearly apparent in Fig. 4, where the frequency distribution of C␣-C␣ distance between pairs of AB and GH corner residues is shown. The distribution of C␣-C␣ distance between Ala 13 and Glu 116 is more variable in the Gly mutant and tends toward larger values. A similar trend is also apparent in the C␣-C␣ distance distribution calculated for the Ala 13 -Leu 113 pair. In the case of the Lys 16 -Glu 116 and Lys 16 -Leu 113 pairs, the effect is less pronounced for the C␣-C␣ distances, but is clearly visible when one considers the side chain atoms NZ of Lys 16 and CD of Glu 116 (data not shown).
We used dynamical cross-correlation maps to identify corre- lated movements of residues from MD simulations (36,37). Fig.  5 shows the dynamical cross-correlation maps calculated from the simulations by averaging across the entire length of the simulations, except for the first 0.1 ns. Averages calculated with smaller time windows (between 50 ps to 0.1 ns) showed essentially similar patterns (data not shown). Positive peaks in the maps indicate that the movements of an atom pair are correlated, and negative peaks indicate that they are anticorrelated. There exist both positive and negative correlations between the AB and GH region residues in the Ala trajectory (native ␣-chain), as indicated by peaks in the appropriate regions. One finds a strong peak of positively correlated fluctuations (correlation coefficient 0.6 -0.8) between the AB and the GH regions in the Ala simulation. However, in the same region, the correlation coefficient reduces to around 0.2 in the Gly simulation. In general, one finds that the motion of the entire stretch of residues including the A-and the B-helices are better correlated with the region spanned by the G-and H-helices in the native ␣-chain, as compared with the Gly mutant. This indicates that consequent to Ala 3 Gly mutation at the AB corner, the movements of the AB and GH regions become more independent. Similar cross-correlation maps calculated for the Aib simulations show that, in this case, the AB and GH corners exhibit a small shift in the opposite direction, with a slight increase in positively correlated movements as compared with the Ala simulation.
Characterizations of HbS Tetramers Containing Gly/Aib-substituted ␣-Chains-Under identical conditions of anion-exchange FPLC (Fig. 6), both the mutant tetramers containing Gly or Aib substitutions (henceforth referred to as ␣ 2 (A19G/ A21G)␤ 2 S and ␣ 2 (A19Aib/A21Aib)␤ 2 S , respectively) exhibited the same retention time as that of the native HbS (␣ 2 (A 19 -A 21 )␤ 2 S ). This chromatographic behavior is in accord with the fact that Gly or Aib substitutions for Ala on the protein surface will not cause any charge alterations in the mutant hemoglobins. The chain separation profile obtained by RP-HPLC of both the mutant HbS was superimposable with the profile of native HbS that was run under identical conditions. The ratios of the area under the two peaks (␣ and ␤) obtained from each mutant HbS was comparable with and similar to that of the native HbS, suggesting the correct stoichiometry of the ␣and ␤ s -chains in the semisynthetic HbS.
The mutant ␣-chains obtained by RP-HPLC runs of each semi-synthetic construct was further analyzed by ESMS and tryptic peptide mapping to rule out any possible chemical modifications during the process of chemo-enzymic assembly of these chains. The experimental mass of ␣(A19G/A21G) and ␣(A19Aib/A21Aib) mutant chains were found to be 15,098.03 and 15,154.76 Da, respectively (Fig. 7). These values were in excellent agreement with their theoretical masses (15,098.32 and 15,154.38 Da, respectively). That the sequence of the polypeptide chain was intact and the sequence alteration was specific to sites 19 and 21 was confirmed by tryptic peptide mapping. The tryptic peptide profiles of the mutant ␣-chains were compared with the ␣-chain that was isolated from native HbS. RP-HPLC profiles of the tryptic digest of the respective mutant globins showed altered retention time (relative to that of the native globin) for only one peptide namely, ␣T4. This result is consistent with the fact that sites 19 and 21 are contained in the tryptic fragment ␣T4 (residues 17-31). The order of elution of ␣T4 peptide (Gly substitutes first followed by Ala and Aib) from the reverse-phase column was also found to be in accord with the hierarchical hydrophobicity of Gly, Ala, and Aib residues. Taken together, the results unequivocally establish the chemical composition and sequence integrity of the ␣-chains in the mutant tetramers.
Structural Characterization of the Gly/Aib Mutant HbS-Circular dichroism spectra of the hemoglobins in the far UV region (200 -250 nm) were recorded to assess the secondary structures of the polypeptide chains in the tetramer. The spectra (Fig. 8A) of both of the mutant tetramers ␣ 2 (A19G/ A21G)␤ 2 S and ␣ 2 (A19Aib/A21Aib)␤ 2 S were similar to those of the native HbS (␣ 2 (A 19 -A 21 )␤ 2 S ), suggesting that Gly or Aib substitutions at the AB corner of the ␣-chains in the tetramer did not cause any global changes in the polypeptide helical fold and that the effects on conformation, if any, were subtle and localized to the AB-corner residues.
We also used CD spectroscopy to probe the heme-environment of the mutant proteins. For this, CD spectra of the proteins (Fig. 8B) were recorded in the Soret region. Both the mutants exhibited a CD spectrum that was very similar to that of the native HbS, indicating the conservation of the native-like folding of the heme pocket in the mutants. Thus, simultaneous replacements of Gly or Aib for Ala at positions 19 and 21 of the ␣-chain do not produce any tertiary structural aberrations in the mutant HbS.
Functional Behavior of Gly/Aib Mutant HbS-The oxygenbinding properties of the mutants were studied to evaluate the functional integrity of the mutant tetramers (Table I). Both of the mutants, ␣ 2 (A19G/A21G)␤ 2 S and ␣ 2 (A19Aib/A21Aib)␤ 2 S , bound oxygen cooperatively, and their Hill coefficients (n max ) were similar (2.4 and 2.5, respectively) or identical to that of the native HbS (2.5). The oxygen affinity of the mutants, as reflected in their respective P 50 value (7.5 or 8 mmHg), was also comparable with native HbS (8.5 mmHg). In both the mutants, 2,3-bisphosphoglycerate (BPG) induced a decrease in oxygen affinity that was comparable with that of native HbS. The high cooperativity of oxygen binding was also maintained in the presence of BPG. The retention of normal oxygen affinity and cooperativity in the mutant HbS suggests the presence of native-like quaternary structure and dimer interactions in the tetramers. This interpretation seems consistent with the results of the CD spectroscopy described earlier.
Influence of the Gly/Aib Residues at the AB Corner on HbS Polymerization-We measured the polymer solubility (C sat ) of the mutant tetramers under near-physiological conditions in the presence of 70-kDa dextran following the protocol of Bookchin et al. (33). The dextran C sat of native HbS control ␣ 2 (A 19 -A 21 )␤ 2 S and the two mutants, ␣ 2 (A19G/A21G)␤ 2 S and ␣ 2 (A19Aib/A21Aib)␤ 2 S , is shown in Fig. 9A. Under identical conditions, the solubility of mutant HbS (C sat ϭ 53 mg/ml) containing Ala 3 Gly substitutions was found to be considerably higher as compared with the native HbS (C sat ϭ 30 mg/ ml). In contrast, HbS mutant with Ala 3 Aib replacements produced a C sat ϭ 28 mg/ml, which was similar to or slightly less than that of the native control. Importantly, the C sat values of both the native and mutant proteins were independent of the initial hemoglobin concentrations in the tested range of 70 -90 mg/ml. Thus, the solubility measurements clearly demonstrate that the HbS tetramers with Ala 3 Gly substitutions at the AB corner of the ␣-chain are endowed with significant polymeriza-tion-inhibitory potential. However, the Aib residues at the same sites behave in much the same way as Ala residues.
Coupling of the Gly Mutations of the AB-corner with ␣78 Mutation (His/Gln) at the EF-corner-To further substantiate the inhibitory effect of the Gly substitutions at the AB corner on HbS polymerization, we undertook the construction of triple-mutant HbS comprising langur ␣-chain and ␤ s -chain. As described under "Materials and Methods," the langur ␣-chain contains an Asn 78 3 His or Gln mutation at the EF-corner, as compared with the human ␣-chain, besides the Gly 19 and Gly 21 residues at the AB-corner. Thus, the studies of polymerization behavior of the above triple-mutant HbS would corroborate the inhibitory effect because of Gly mutations and also reveal the additive or synergistic nature of the interactions between the two distant regions (AB and EF corner) of the tetramer. Be-

FIG. 5. Dynamical cross-correlation maps for simulations of Ala (native)
and Gly mutant ␣-chain. The maps were calculated according to the formula C ij ϭ ͗r i r j ͘ Ϫ ͗r i ͗͘r j ͘, where r k refers to the position vector of the k th atom, and ͗r k ͘ is the mean position vector of the same atom. The angle brackets represent time averages. The averaging was carried out during the simulation period from 100 ps to 2 ns. Values below the main diagonal correspond to the Ala simulation (native), whereas those above the main diagonal correspond to the Gly simulation. The bars besides the axes depict the secondary structure assignments of the ␣-chain. The color scale is shown in the box placed outside the figure at top right. 6. FPLC of mutant HbS. An aliquot of the native HbS (␣ 2 (A 19 -A 21 )␤ 2 S ) and the respective mutant proteins ␣ 2 (A19G/A21G)␤ 2 S or ␣ 2 (A19Aib/A21Aib)␤ 2 S was chromatographed on a Mono Q (HR5/5) column using a linear pH gradient (8.5-7) of Tris-acetate buffer over a period of 20 min at a flow rate of 1 ml/min. sides, the polymerization results of the mutants would provide information on the effects of chemically disparate side chain at the ␣78 site (Asn 78 3 His/Gln) on fiber formation; the inhibitory potential of this site has been previously deciphered (38) from the co-polymerization studies of HbS and HbA Stanleyville II (Asn 78 3 Lys).

FIG.
HbS tetramers composed of ␤ S -chain and respective triplemutant ␣-chains, ␣(A19G/A21G/N78Q) or ␣(A19G/A21G/ N78H), were generated. The tetramers were isolated by CM52 cation-exchange chromatography followed by an analytical FPLC using a Mono Q column, as described previously. The heme stoichiometry in the purified tetramers was ascertained by comparing the 280:540 nm absorbance ratio for these pro-teins with that of the native HbS. These ratios (2.50 and 2.52, respectively) compared well with a value of 2.54 of native HbS, thus confirming the correct stoichiometry of heme in the reconstituted tetramers. RP-HPLC analysis of the respective hemoglobins revealed correct composition of ␣ and ␤ chains. The far-UV CD and Soret region CD spectral characteristics of both the triple mutants were comparable with and similar to those of native HbS (data not shown).
The functional integrity of the triple mutants was assessed by oxygen-affinity measurements. The P 50 for the ␣ 2 (A19G/ A21G/N78H)␤ 2 S and ␣ 2 (A19G/A21G/N78Q)␤ 2 S mutants were found to be 8 and 7 mmHg, respectively. Thus, both HbS mutants exhibited P 50 values that were similar to native HbS (P 50 ϭ 8.5 mmHg). Besides, the mutants also displayed normal cooperativity of oxygen binding. The results suggest that substitution of His or Gln for Asn at the 78th position in the co-presence of Ala 3 Gly replacements at the AB-corner of the ␣-chains do not have any deleterious effect on the function of HbS.
Polymerization Behavior of Triple Mutants-The gelation concentration (C sat ) of the above triple mutants was determined to ascertain the combinatorial influence of sequence changes at the AB and EF corners on the polymerization reaction. For this, C sat of the triple mutants (Fig. 9B) was measured by the dextran-C sat method in a manner identical to that used for the double mutants (Fig. 9A). The ␣ 2 (A19G/A21G/ NA8H)␤ 2 S tetramer yielded a C sat of about 28 mg/ml that was

TABLE I Oxygen binding characteristics of the mutant HbS
Oxygen equilibrium measurements were done at pH 7.4 (100 mM sodium phosphate buffer, 29 Ϯ 0.5°C), in the absence or presence of BPG (2 mM) using a Hemox analyzer. The hemoglobin concentration used was about 0.1 mM on heme basis. P 50 (partial pressure at 50% saturation) of hemoglobin is expressed in mmHg. The P 50 values were reproducible within Ϯ 0.5 mmHg. very close to the C sat of native HbS (C sat ϭ 30 mg/ml). In contrast, the C sat of ␣ 2 (A19G/A21G/N78Q)␤ 2 S tetramer was considerably higher (C sat ϭ 50 mg/ml), as compared with native HbS, but very similar to that of the ␣ 2 (A19G/A21G)␤ 2 S tetramer (C sat ϭ 53 mg/ml).
The C sat measurements of HbS containing only the point mutation of a Gln or His for Asn at ␣78 was determined to assess the individual contribution of these residues in the polymerization of HbS. The C sat of ␣ 2 (N78Q)␤ 2 S mutant was found to be similar (31 mg/ml) to that of the native HbS, suggesting that Gln residue behaves in much the same way as Asn. However, the HbS mutant ␣ 2 (N78H)␤ 2 S containing His at this site yielded a C sat ϭ 22 mg/ml, which was significantly lower than that of native HbS, suggesting that the presence of His residue at the ␣78 site confers polymerization-enhancing propensity to the HbS molecule. The potentiating effect of His at this site has been commented upon in a recent publication (39).
Given that the C sat of ␣ 2 (A19G/A21G/N78Q)␤ 2 S and ␣ 2 (A19G/A21G)␤ 2 S mutants are similar (50 and 53 mg/ml, respectively), and the point mutant, ␣ 2 (N78Q)␤ 2 S , behaves in much the same way as native HbS, it can be concluded that the inhibition of polymerization observed in the triple mutant must emanate from the Gly substitutions at the AB corner. Further, inhibition of HbS polymerization because of Gly substitutions was not noticeable in the His-containing triple mutant, ␣ 2 (A19G/A21G/N78H)␤ 2 S , because of the polymerization-enhancing effect of His 78 . However, the C sat of the triple mutant ␣ 2 (A19G/A21G/N78H)␤ 2 S (28 mg/ml) is far less than that expected from an additive contribution of constituent doublemutant ␣ 2 (A19G/A21G)␤ 2 S (53 mg/ml) and single mutant ␣ 2 (N78H)␤ 2 S (22 mg/ml). Because the C sat of the HbS (30 mg/ ml) increased by 23 mg/ml because of Gly substitutions and was lowered by 8 mg/ml because of His at ␣78, a simple additive effect would have yielded a C sat of about 45 mg/ml (30 mg/ml ϩ (23Ϫ8) mg/ml) for the triple mutant. The shift in C sat of the triple mutant by about 17 mg/ml ((45Ϫ28) mg/ml) clearly suggests that the polymerization-enhancing effects due to the presence of His 78 at the EF corner synergistically compensate for the inhibitory effect emanating from the Gly substitutions at the AB corner.

Coupling of His 78 at the EF Corner with a Mutation of a Known
Contact (␣Lys 16 3 Gln) Located in the Proximity of the AB Corner: Polymerization Behavior of ␣ 2 (K16Q/N78H)␤ 2 S -To see whether the synergistic coupling between His 78 and the AB corner was unique to the Gly substitutions or a general attribute of the AB region, we constructed a double-mutant HbS (␣ 2 (K16Q/N78H)␤ 2 S ) in which His 78 mutation was coupled to mutation of a known intra-double-strand contact site at ␣16. The C sat of the above double mutant (Fig. 10) was found to be 17 mg/ml. Using identical assay conditions, we had previously reported a C sat of about 39 mg/ml for the HbS-carrying point mutation of ␣Lys 16 3 Gln as against 30 mg/ml for native HbS (24). Thus, the presence of His 78 in the double mutant not only annulled the inhibitory effect due to mutation at ␣16 site but also conferred significant polymerization-enhancing propensity to the double-mutant HbS. Interestingly, His 78 shifted the C sat of double-mutant ␣ 2 (K16Q/N78H)␤ 2 S in a similar synergistic fashion as it did in the case of the triple-mutant ␣ 2 (A19G/ A21G/N78H)␤ 2 S . Taken together, the results unequivocally suggest synergistic coupling of interactions between AB region of the ␣-chain and ␣78 located at its EF corner in the polymerization of HbS.

DISCUSSION
HbS fiber formation is a complex process that requires specific contacts and intricate interplay between several residues from both, ␣ and ␤, chains of the tetramer. In the current scenario, the definition of a fiber contact by both theoretical as well as experimental approach is centered on specific amino acid residues. The solution studies consider a given residue as a fiber contact only if the mutation of that site results in an altered polymerization behavior relative to the native HbS (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). On the other hand, the fiber models define a contact residue based on the distance; residues with contact distances of about 5 Å or less are generally considered as interacting partners (26,27). Notwithstanding the approximate concordance and inter-dependence between experimental and modeling studies, the above specific amino acid residue-based description of contact points does not shed any light on the influence of regio-specific conformational dynamics of HbS molecule on the polymerization process, as also the short-or longrange cumulative interaction between the contact sites. Fur-

Cooperative Interactions in the HbS Fiber Assembly
ther, although no studies have so far been undertaken to gain insights into the protein dynamics vis-à -vis fiber formation, definitive experimental data on combinatorial effects due to contact sites are rare. In this report, we have addressed the aforementioned aspects of HbS polymerization by engineering sequence changes in selected regions of the ␣-chain at specific contact sites, as well as at sites that do not fall within the ambit of the traditional definition of contact points but are capable of perturbing the fiber formation by modifying the conformational dynamics of a contact-rich region of the fiber.
In designing the studies to probe the role of dynamics of the AB corner on HbS polymerization, it was pertinent to see that amino acid changes made to modulate the conformational features of this region exerted only local influence and did not perturb the gross secondary structure or overall global folding of the hemoglobin molecule. The choice of residues Ala 19 and Ala 21 , located at the boundary of A-and B-helices of the ␣-chain was appropriate in this regard, because alterations at the beginning or end of helices, as compared with those located in the middle of helices, were less likely to cause such untoward structural changes. Indeed, CD spectroscopy showed that HbS tetramers with either Ala 3 Gly or Ala 3 Aib substitutions exhibited secondary structure and tertiary heme folding that was very similar to the native HbS. Besides, the functional behavior of the mutants, as reflected by the oxygen-affinity measurements of the tetramers, were also similar, suggesting the preservation of native-like quaternary structure and dimer interactions in the mutant proteins. These results confer legitimacy to the solubility data of the HbS mutants carrying Ala to Gly or Aib substitutions at the AB corner of the ␣-chains, and indicate that the effects on HbS polymerization are primarily related to their intrinsic ability to perturb the conformational dynamics of the polypeptide chain.
MD simulations of mutant ␣-chains show that both Gly and Aib mutants perturbed the dynamics of the AB and GH region of the chain. However, distribution of C␣-C␣ distances between pairs of AB-GH (Fig. 4) residues indicated that Gly mutants exhibit greater fluctuations as compared with Aib and Ala (native chain) mutants. This was further corroborated by the fact that dynamical cross-correlation maps of native (Ala) and Aib-substituted ␣-chains showed strong motional coupling of AB and GH region residues, whereas that of Gly-substituted chains displayed significantly less coordinated movement. This can be interpreted to mean that AB-GH regions of native or Aib mutant chains may be more rigid than the Gly mutant chain. The shift in the tail of the distance distribution profiles for the Gly mutant simulation toward larger values (shown in Fig. 4), therefore, appear to have arisen because of the relative motional independence that the GH corner residues gain in the presence of Gly residues at the AB corner. The perturbation of Lys 16 , Leu 113 , and Glu 116 in the Gly mutant is particularly interesting, because mutations of these residues of the GH corner are known to affect the polymerization reaction. Thus, the results of the MD simulation of the ␣-chains indicate that Gly mutations inhibit HbS polymerization by perturbing the intra-chain interactions of the ␣-chain that, in turn, influences the overall intra-double-strand axial interactions in the fiber emanating from AB-GH regions of the ␣/␤ chains.
The ability of a polymerization-enhancing mutation of an established inter-double-strand contact point (␣His 78 ) at the EF corner of the ␣-chain to completely neutralize the inhibitory effect of Gly substitutions at the AB corner is an interesting result that suggest long-range and non-additive coupling of mutational effects in HbS polymerization. This view is strengthened with the observation that the coupling of ␣His 78 at the EF corner to mutation of a specific intra-double-strand fiber contact (␣16) in the AB region occurs in a similar synergistic fashion as it happened with Gly mutants. The fact that ␣Lys 16 makes axial contacts with the GH corner residues (␣114, ␣115), and Gly substitutions at the AB corner exert their influence on polymerization by inducing dynamic changes in the AB-GH region of the ␣-chain, raises the intriguing possibility of interaction-linkage between AB, GH, and EF regions of the ␣-chain in the HbS fiber assembly. Our previous work (24), demonstrating that the inhibitory effect of ␣16 was additive with ␣113 (GH1), is consistent with this idea.
The polymerization studies of the mutant HbS molecules described in this report are important in another context. Transgenic mice expressing the human ␤ s chain do not fully exhibit the sickling phenotype due to inhibition of HbS polymerization by mouse ␣-chains (40). There are 19 mutations in the mouse ␣-chain compared with the human sequence. Of these, only three (␣78, ␣113, and ␣116) are at the known contact points of the fiber. In previous work, the 19 sites of sequence changes were segregated into 8 (located in ␣1-30 segment) and 11 (located in ␣31-141 fragment) through the construction of human/mouse chimeric ␣-chains. Interestingly, along with other sequence changes, the mouse ␣-chain contains Ala 19 3 Gly, Ala 21 3 Gly, and Asn 78 3 Gly substitutions. HbS constructs composed of either chimeric ␣-chains were shown to inhibit the polymerization reaction, albeit to different extents. Further, the inhibitory potential of both the constructs was much lower than the intact mouse ␣-chain. The results lead to the speculation that one or more residues among the eight sites of sequence changes in the ␣1-30 segment must make intermolecular contacts in the HbS fiber and that the interactions of contact residue(s) in the two complementary fragments are synergistic. The results of the present work, demonstrating inhibitory effects due to residues 19 and 21 at the AB corner and interaction-linkage with ␣78 located at the EF corner, provide a rational explanation for the earlier observations. However, it must be reiterated here that ␣19 and ␣21 are not fiber contact sites in the conventional sense, because they are not implicated in any interactions by either solution experiments or model-building studies. It turns out that modification of fiber interactions through perturbation of conformational dynamics of a contact-rich region of the HbS may be an effec- FIG. 10. Synergistic action of AB region and EF region mutations on polymer solubility. The C sat measurements were performed in the presence of dextran at 37°C as described in Fig. 9 legend. The data for HbS carrying a point mutation at ␣16 site (␣Q16) is taken from Sivaram et al. (24). tive mechanism to inhibit HbS polymerization.
In summary, the results show that Ala 19/21 3 Gly mutations of the non-contact residues at the AB corner of the ␣-chain inhibit HbS polymerization by enhancing the dynamics of the AB-GH region that contains numerous intra-double-strand contact residues of the HbS fiber. The synergistic coupling of Ala 19/21 3 Gly or Lys 16 3 Gln mutations of the AB region with a mutation located in the EF corner (Asn 78 3 His) of the ␣-chain suggest some sort of linkage between fiber interactions that emanate from AB, GH, and EF regions of the ␣-chains. Further studies involving NMR and x-ray crystallography of the mutant HbS molecules described here would illuminate the mechanism(s) of short-or long-range communication between specific regions of the tetramer in fiber assembly.