Single residue modification of only one dimer within the hemoglobin tetramer reveals autonomous dimer function

Abstract

The mechanism of cooperativity in the human hemoglobin tetramer (a dimer of αβ dimers) has historically been modeled as a simple two-state system in which a low-affinity structural form (T) switches, on ligation, to a high-affinity form (R), yielding a net loss of hydrogen bonds and salt bridges in the dimer–dimer interface. Modifications that weaken these cross-dimer contacts destabilize the quaternary T tetramer, leading to decreased cooperativity and enhanced ligand affinity, as demonstrated in many studies on symmetric double modifications, i.e., a residue site modified in both α- or both β-subunits. In this work, hybrid tetramers have been prepared with only one modified residue, yielding molecules composed of a wild-type dimer and a modified dimer. It is observed that the cooperative free energy of ligation to the modified dimer is perturbed to the same extent whether in the hybrid tetramer or in the doubly modified tetramer. The cooperative free energy of ligation to the wild-type dimer is unperturbed, even in the hybrid tetramer, and despite the overall destabilization of the T tetramer by the modification. This asymmetric response by the two dimers within the same tetramer shows that loss of dimer–dimer contacts is not communicated across the dimer–dimer interface, but is transmitted through the dimer that bears the modified residue. These observations are interpreted in terms of a previously proposed dimer-based model of cooperativity with an additional quaternary (T/R) component.

Many tertiary and quaternary conformational events important for catalysis and cooperativity in biological macromolecules are energetically unfavorable, and must be driven by favorable free energy of ligand binding. In the case of human hemoglobin (Hb), initial binding of O₂ to the heme Fe drives intrinsically unfavorable conformation change within the globin. The associated energetic “penalty” results in lower binding constants for initial ligation. However, penalties for the additional binding steps are progressively reduced, generating Hb's characteristic sigmoidal binding curve. Contributing to this curve are the partially ligated Hb intermediates, composed of two different subunit types (α and β) with four binding sites (α¹, β¹, α², β²). Up to four labile ligands (O₂) are present on each tetramer in multiple configurations of site occupancy within at least two distinct quaternary structures (T and R). The challenge in understanding the mechanism of Hb cooperativity has been to identify and measure the separate energetic penalties for each O₂ binding step, and then to correlate individual penalties with specific structural events.

To approach a resolution of this complex problem, energetic components of the system's ligation intermediates were evaluated with no prior assumptions as to their quaternary structures. Analysis of the thermodynamic linkages between dimer–tetramer assembly and O₂ binding permitted the ligand binding constant of the noncooperative free dimer to be used as the thermodynamic reference species (1, 2). The energetic penalty observed in ligand binding to the tetramer was quantitated as the free energy difference between each tetrameric species having bound O₂ in a specific site configuration and the corresponding set of dimeric reference species, and was represented as ΔG_c, the cooperative free energy penalty (3).

Hb cooperativity is thus comprised of stepwise decreases in ΔG_c penalty with successive O₂ bound, traditionally observed in four macroscopic ligand binding constants (Fig. 1). These binding constants, also known as Adair constants, are composites of microscopic constants for the different configurations of bound O₂ among the four hemesites. The cumulative ΔG_c penalty, 6.3 kcal/mol, likewise distributes among the 10 microstates (Fig. 2). There is significant redundancy within the partially ligated species: those with a single O₂ bound exhibit very similar ΔG_c penalties, as do those having O₂ bound to both dimers. However, the Hb tetramer with one dimer unligated and the other dimer fully ligated exhibits a unique ΔG_c penalty in comparison with the other species. This pattern of ΔG_c distribution was found to be consistent over a range of solution conditions and hemesite substitutions (reviewed in ref. 4), and is the basis for a microstate model of cooperativity in which the Hb tetramer functions as a dimer of cooperative αβ dimers (reviewed in ref. 5).

Figure 1.

Components of the cooperative free energy penalty (ΔG_c) in wild-type Hb. Macroscopic O₂ binding free energies, ΔG_i i = 1, … , 4 (in kcal/mol), for both free dimer (D) and assembled tetramer (Tet) and corresponding tetramer assembly free energies, ΔG_tet, previously determined at pH 7.4 and 21.5°C (5, 35). The ligand binding free energy for the free dimer, ΔG_int, also serves as the intrinsic ligand binding free energy for the tetramer. The difference in tetramer ΔG_i minus free dimer ΔG_int is the cooperative free energy penalty, ΔG_c, paid by the tetramer. As additional O₂ ligands are bound, the stepwise ΔG_c penalty decreases from 2.9 for the initial binding step to −0.6 kcal/mol for the final O₂. The cumulative penalty for complete oxygenation is 6.3 kcal/mol (a 4 × 10⁴-fold enhancement in equilibrium constant). This also equals the difference in assembly ΔG_tet (oxy minus deoxy), as required by path-independence of free energy. Thus, assembly parameters provide an equivalent measure of the structural free energy components to that obtained by direct O₂ binding.

Figure 2.

Experimentally observed penalties for wild-type Hb microstates. Cumulative penalties for the 10 ligation microstates, previously determined for the Zn/FeO₂ hemesite analog system (5). Dark gray bars indicate ligation onto only one dimer within the tetramer.

The energetically unfavorable structure changes that accompany the ΔG_c penalties were originally attributed by Perutz and Kilmartin to the net loss of noncovalent bonds in the dimer–dimer interface (6, 7). Crystal structures of the tetrameric end states showed that the two αβ dimers reorient relative to one another on complete ligation, breaking hydrogen bonds (including salt bridges) within the polar dimer–dimer interface. Consistent with this, the dimer → tetramer assembly free energy, ΔG_tet, was significantly more favorable for the deoxy (T) relative to the oxy (R) Hb (i.e., the less stable tetramer exhibits higher overall binding affinity). Perhaps the most convincing argument that binding the first O₂ results in loss of dimer–dimer interface contacts has been the near-universal observation that removal of any dimer–dimer contact, either hydrogen bond or salt bridge, results in an increase in oxygen affinity and a decrease in cooperativity. Thus, both structural and energetic results have commonly been interpreted in terms of Hb cooperativity resulting from global quaternary T ⇋ R switching. In this view, O₂-driven subunit conformational changes are transmitted from the heme solely to the dimer–dimer interface, with no intradimer cooperativity.

The presence of a direct relationship between the strength of noncovalent bonding interactions across the interdimer interface and a quaternary T ⇋ R equilibrium thus became a fundamental premise in Hb cooperativity which is still widely accepted. In combination with the two-state energetic model (MWC; ref. 8), the Perutz salt bridge model has accounted for a wide range of both structural and functional Hb properties, as long as the symmetry of the Hb tetramer was preserved and as long as the system was studied as the average of its ligation intermediates. Still, elements of the two-state model, which permits global quaternary rearrangement, but no sequential inter-subunit coupling, have frequently been questioned from a wide range of perspectives (9–16). Systematic deviations from the two-state model become most apparent when the Hb system is resolved into its separate, partially ligated microstates, some of which are asymmetrically ligated (17–21). The fact remains that experimental perturbations to the Hb system, when limited to symmetric modifications (solvent conditions, allosteric effectors, globin modifications), cannot, even in principle, yield a clear distinction between a single quaternary (cross-dimer) effect vs. two tertiary (intradimer) effects.

In this study, the change in ΔG_c penalty resulting from single residue modification of just one dimer within the tetramer is measured for all configurations of bound ligand. The modification desArg α141, traditionally used to evaluate cross-dimer functional coupling in Hb, is examined by using the O₂ analog Fe⁺³CN (CNmet). In this Hb, the C-terminal α141Arg is enzymatically removed, eliminating two key cross-dimer salt bridges. Results show that the loss of a single α141Arg residue destabilizes the deoxy tetramer as expected, but only the modified dimer binds ligand with increased affinity: the wild-type dimer within the modified tetramer binds ligand with wild-type affinity. Similar results are observed for Hb Yakima β99Asp → His, a naturally occurring mutation that breaks two cross-dimer hydrogen bonds in the α¹β² interface. These findings are interpreted in terms of a localized intradimeric (or tertiary) response to the modification of interface contacts, rather than a cross-dimer quaternary response. This interpretation is shown to be valid not only for hybrid tetramers, but for the classical doubly modified tetramers as well.

Materials and Methods

Human Hbs A (HbA₀) and the naturally occurring variants S (HbS) (β6 Glu → Val), Bunbury (β94 Asp → Asn), and Yakima (β99 Asp → His) were prepared from whole blood (22). DesArg Hb was prepared by the carboxypeptidase digestion method of Kilmartin (7). CNmet derivatives were prepared by ferricyanide oxidation and the addition of KCN (23). All purified Hbs were exchanged into 0.1 M Tris, 0.1 M NaCl, and 1 mM Na₂ EDTA, at pH 7.40, T = 21.5°C (total chloride concentration = 0.18 M). Oxygen binding was performed on doubly modified desArg Hb by using both the Imai and Gill techniques (24, 25). The 16 partially ligated microstates containing a single residue modification were constructed by mixing the doubly modified parent species with wild-type parents. Their hybrids formed as a result of dimer exchange via the normal equilibrium between free αβ dimer and assembled tetramer. Dimer to tetramer assembly free energy values for each parent Hb were independently determined either by the haptoglobin kinetics method (26) or by analytical gel chromatography (27). ΔG_c penalties were calculated by subtracting the assembly free energy, ΔG₂, of deoxy hybrid from that of the partially ligated hybrid. Each modified hybrid was separated from its two parents by cryogenic isoelectric focusing, from which the relative concentrations of parents and hybrid were evaluated (28). HbS was used instead of normal HbA₀ to further enhance electrophoretic separation of hybrids from parents, as it was previously demonstrated to have energetic properties indistinguishable from those of wild-type Hb. HbS is referred to herein as “wild type” with specific reference to its unperturbed dimer–dimer interface contacts. In the case of the doubly ligated species in which both ligands are located on one dimer, the CNmet analog yields an experimental artifact because of ligand exchange, causing this species to have a ΔG_c penalty similar to singly ligated species (5). A correction has been applied to the data shown here for this species, which has no significant effect on the results or conclusions of this study.

Results

The overall effect of the double desArg modification (Fig. 3) on assembly free energies of the CNmet end-states is a marked 4.2 kcal/mol destabilization of the deoxy tetramer and a small 0.8 kcal/mol stabilization of the fully ligated tetramer, consistent with previous findings for the doubly modified O₂ species (29). The overall ΔG_c penalty for complete ligation is reduced from 6.0 kcal/mol for CNmet ligation of wild-type Hb (HbA₀) to only 1.0 kcal/mol for CNmet desArg Hb. As expected, the ligation end states for singly modified A₀/desArg hybrid have properties intermediate between those of normal Hb and the doubly modified desArg Hb: the deoxy tetramer is destabilized by 2.4 kcal/mol and the oxy tetramer is stabilized by 0.3 kcal/mol, yielding an overall ΔG_c of 3.3 kcal/mol.

Figure 3.

Cross-dimer contacts in deoxyHb. Deoxy αβ dimers, showing α-subunits (teal), β-subunits (dark blue), and hemes (red) (36, 37). The α-carbon atoms of each residue modified in this study are shown in yellow. The C-terminal α141Arg forms salt bridges with both 126Asp and 127Lys on the opposite α-subunit; the β99Asp is hydrogen bonded to α42Tyr and α97Asn on the opposite dimer; and β94Asp is hydrogen bonded to the C-terminal 146His on the same β-subunit (green dashed lines). Remaining contacts relate to the data of Fig. 7 (gray dashed lines). Overall, the cross-dimer interface in the deoxy tetramer is composed of α¹β², α²β¹, and α¹α² polar interactions, each of which are in contact with the central cavity. No β¹β² contacts are found in the deoxy or fully ligated tetramers, and no α¹α² contacts are found in the fully ligated tetramer. Ligands bind to the distal side of each heme, which is facing away from the dimer–dimer interface in this view (crystal structures of Hb do not exhibit a clear path for the entry of O₂). Free dimer structures are taken from the tetramer structure (37) for this illustration.

Partially ligated A₀/desArg hybrids show a clear asymmetric response to CNmet ligation (Fig. 4). Within the singly modified tetramer, the wild-type dimer exhibits the same ΔG_c penalty found in the wild-type tetramer, and the desArg dimer has the same penalty as the double desArg tetramer. Thus, the ΔG_c penalty measured on ligation to one dimer within the tetramer is independent of the presence or absence of a desArg modification on the opposite dimer. This is characteristic of the Hb system despite the fact that removal of the C-terminal Arg weakens the tetramer relative to free dimer, in both the single and double desArg Hb. This response is consistently observed throughout all 16 desArg hybrids (Fig. 5); independent of the state of the opposing dimer, ligation to a modified dimer yields a decreased penalty, whereas ligation to a wild-type dimer exhibits a wild-type penalty.

Figure 4.

Asymmetric effect of single desArg modification at the first ligand binding step. The first CNmet ligand, denoted as X, binds to either α- (shown) or β-subunit in wild-type Hb (gray subunits) with a ΔG_c penalty of 3.0 ± 0.2 kcal/mol at pH 7.4 and 21.5°C (38). When both α141Arg residues are removed (green subunits), effectively eliminating α¹–α² contacts between the two dimers, the penalty is reduced to 1.0 ± 0.2 kcal/mol. The hybrid desArg Hb, in which only one Arg has been removed, is an asymmetric tetramer, and responds to ligation in an asymmetric manner: the wild-type dimer binds CNmet with the wild-type penalty, whereas the desArg dimer binds CNmet with the same penalty exhibited by the doubly modified desArg Hb. The ΔG_c penalties for binding the first CNmet ligand to wild-type Hb and to double desArg Hb are similar to the O₂ binding penalties, 2.9 and 1.1 kcal/mol, respectively.

Figure 5.

Distribution of ΔG_c penalties in desArg Hb microstates. Ligation of CNmet to the 10 ligation states of symmetric Hbs (Top, wild type; Bottom, double desArg) and all 16 combinations of desArg hybrids (Middle) at pH 7.4 and 21.5°C. Tetramers with ligands on only one dimer are boxed. Experimental errors are ± 0.1 to 0.3 kcal/mol, or approximately one-third the height of the tetramer symbol.

Single modification of the β subunit was also examined, using two naturally occurring mutations. Hb Bunbury (β94 Asp → Asn) eliminates an intra-subunit hydrogen bond at the β N terminus, and Hb Yakima (β99 Asp → His) alters two hydrogen bonds in the α¹β² interface (Fig. 3). In contrast, the double mutant of Hb Bunbury exhibits tetramer stability and ligand affinity indistinguishable from that of normal HbA₀ (29), whereas the double mutant of Hb Yakima destabilizes the deoxy tetramer by 4.5 kcal/mol and stabilizes the fully ligated tetramer by 1.9 kcal/mol, comparable to that found with the O₂ ligand (29). The total cumulative ΔG_c penalty is 6.2 kcal/mol for double Bunbury and only 0.1 kcal/mol for double Yakima, compared with 6.0 kcal/mol for wild type. Under conditions of this study, the single β94 modification responds essentially as a silent mutation in the 16 hybrids (Fig. 6), whereas the single β99 mutation generates a distribution of penalties having the same characteristic pattern as that of single desArg.

Figure 6.

Distribution of ΔG_c penalties in β-modified Hb microstates. The 16 hybrids of Hb Bunbury (β94 Asp → Asn) (Upper) and Hb Yakima (β99 Asp → His) (Lower). Experimental conditions are the same and experimental errors are similar to those for desArg Hbs shown in Fig. 5.

The Hb microstate composed of one unligated and one doubly ligated dimer is relatively straightforward to prepare, requiring only the wild type and doubly mutated parents. A survey of assembly free energies over a range of modifications and naturally occurring mutations in the dimer–dimer interface was conducted in which the CNmet ligand was placed on the modified dimer. These results are compared with previously published perturbations of the corresponding hybrid in which the CNmet ligand was placed on the wild-type dimer (30) (Fig. 7). The asymmetric response observed in the single desArg and single Yakima hybrid data are evident in the ΔG_c perturbations of 19 modifications at 16 residue positions throughout the dimer–dimer interface (Fig. 7), indicating that this functional asymmetry is not limited to a particular set of dimer–dimer contacts.

Figure 7.

Dimer–dimer interface mutational perturbation at 16 residue sites. Perturbation of the ΔG_c penalty in fully ligated and α¹β¹ doubly ligated (CNmet) hybrid Hb caused by single-residue modification (green subunit). Residues throughout the dimer–dimer interface (7 in α-subunits, 9 in β-subunits) were modified either chemically or by naturally occurring mutation. The ΔG_c penalty is calculated as the dimer–tetramer assembly free energy of the ligated hybrid tetramer minus the assembly free energy of the unligated hybrid tetramer bearing the same modified residue. The ΔG_c penalty for ligation to the wild-type dimer (black bars) is within experimental error (typically ± 0.2 kcal/mol) of the wild-type value (A₀) for all modified Hbs. The ΔG_c penalty for ligation to a dimer carrying the modified residue varies from 7.8 to 2.5 kcal/mol for fully ligated hybrid Hb (white bars) and from 4.4 to 0.6 kcal/mol for doubly ligated hybrid Hb (gray bars). Modified Hbs are: SM β102 Asn → Tyr; AG β40 Arg → Lys; BU β94 Asp → Asn; TG β124 Pro → Gln; GN α85 Asp → Asn; SD β109 Val → Met; TH β93 Cys → thio-CH₃Cys; A₀ wild type; DH β146 His deleted; WO β97 His → Leu; SC α127 Lys → Thr; NE β93 Cys → NESCys; DL α97 Asn → Lys; CH α92 Arg → Leu; LE α141 Arg → Leu; TA α126 Asp → Asn; KA α40 Lys → Glu; DA α141 Arg deleted; YA β99 Asp → His; RA β99 Asp → Ala. Residues identified as salt bridge contacts in either α- or β-subunits (DH, SC, LE, TA, KA, DA) result in no significantly different pattern of perturbation than other interface contact residues. Data shown as solid bars was previously published (31). Experimental errors are ± 0.1–0.3 kcal/mol.

The use of a nonlabile O₂ analog (such as CNmet) is expected to perturb the system. The extent of perturbation, observed by comparing the measured O₂ binding curve for double desArg Hb with the CNmet binding curve calculated from microstate constants, is not large enough to significantly alter the ligation characteristics of the desArg modification (data not shown). The ligand exchange artifact previously observed for the α¹β¹-doubly ligated species is observable only on long (>24 h) hybridization, which does not apply to the other ligation hybrids. For example, the asymmetric effect is clearly observed on initial ligation, as shown in Fig. 4, in which the ΔG_c penalty for the CNmet analog is the same, within error, as that for O₂, for both wild type and desArg Hbs. Thus, ligation analog perturbations and artifacts are present in these systems, but do not have a significant impact on the ΔG_c values reported here.

Discussion

Hybrid Hbs with only a single desArg α141 or β99 Asp → His modification demonstrate that removal of interdimer contacts weakens the T interface, but is not communicated from heme to heme across that interface. Instead, the perturbation appears to be long-range (30–40 Å) within the dimer itself, in that an α-subunit modification affects the β-subunit heme within the same dimer, and a β-subunit modification affects the α-subunit heme within the same dimer. In the interpretation of the microstate model of Hb cooperativity (also referred to as the “symmetry rule” or “molecular code” model, refs. 3–5), ligation to a single dimer within the tetramer produces a ΔG_c penalty without T → R quaternary switching. On further ligation, when both dimers contain at least one ligand, the combined ΔG_c penalty exceeds the free energy of the T interface, which then rearranges to the less-stable R interface in a concerted quaternary switch (Fig. 8). In this microstate model, ligands bind cooperatively to a single dimer within the tetramer, followed by a single concerted quaternary switch, and subsequent cooperative ligation by the other dimer.

Figure 8.

Microstate-based allosteric model of Hb along the preferred pathway. The energetically preferred pathway through the Hb microstates is that in which one dimer becomes fully ligated before the remaining dimer binds ligands (deoxy subunits are blue, ligated subunits are red). The pathway begins with the deoxy tetramer, in which each unligated dimer is designated D0. Free energy of the initial binding step (to either α- or β-subunit) drives changes in conformation not only in the ligated subunit and surrounding solvent, but throughout the ligated dimer, designated D1. These long-range effects are manifested as an unfavorable free energy penalty, whose origin is the mismatch between the partially ligated dimer conformation and the T interface; when the interface is absent, as in the free αβ dimer, this penalty is never developed, and binding occurs noncooperatively. The second binding step on the same dimer develops a smaller penalty, resulting in positive cooperativity within the dimer (D2). The quaternary T alignment between the two dimers is not significantly altered until the third binding step. On ligation of the second dimer, the unfavorable free energy penalty exceeds the favorable free energy of the T interface contacts, which rearrange to R in a concerted fashion. The partially unligated dimer within R, d1, produces only a small penalty relative to that within T. On binding the final ligand, this penalty is released, designated in both dimers as d2. In the reverse pathway, from the fully ligated to deoxy tetramer, ligand-driven conformational change is no longer a penalty, but instead becomes an energetic “dividend.”

Structural Basis for Intradimer Cooperativity.

The intradimer (or α¹β¹) interface is typically viewed as inert, and not contributing to cooperative ligand binding. This conclusion is based primarily on the lack of significant change in the interface between crystal structures of deoxy and fully ligated Hb. However, cooperativity within the αβ dimer would likely require such a change, but by what mechanism? Two proposals that provide a structural framework for intradimer cooperativity have been made, one involving rigid body movement from α- to β-subunit (31), and another based on movements of helices on the distal side of the heme (32). Recent dynamic structural approaches show hydrogen exchange in histidine residues in the intradimer interface, which differ significantly between native deoxy and oxy Hb (33), as well as communication across the α¹β¹ interface (16), which are consistent with the present findings of dimer functionality.

The Functional Dimer.

The results presented here yield a functional picture of the mechanism of Hb cooperativity that is fundamentally different from that traditionally perceived. Although crystal structures of the tetramer end-states show a molecule that is structurally organized as a dimer of αβ dimers, the structural symmetry of the tetramer suggests that the dimers act in concert, maintaining functional symmetry as well. The most enduring models of cooperative ligand binding introduced in the 1960s [i.e., the two-state concerted model (8) and the sequential model (34)] both viewed Hb function in terms of a tetramer of four similar subunits, rather than as a dimer of cooperative αβ dimers. Symmetrical perturbations of Hb, coupled with the stoichiometric composite ligand binding constants, supported these early proposals. Results of the present study indicate that the assignment of thermodynamic, kinetic, spectroscopic, and static structural properties to quaternary features of the Hb tetramer based on symmetric perturbations averaged over many microstates should be reevaluated. Classically inferred quaternary characteristics of this molecule are likely to be due, in part or in whole, to the coupling effects within dimers, as demonstrated here.