|Home | About | Journals | Submit | Contact Us | Français|
Protein engineering strategies seek to develop a hemoglobin-based oxygen carrier with optimized functional properties, including (i) an appropriate O2 affinity, (ii) high cooperativity, (iii) limited NO reactivity, and (iv) a diminished rate of autooxidation. The mutations αL29F, αL29W, αV96W and βN108K individually impart some of these traits and in combinations produce hemoglobin molecules with interesting ligand-binding and allosteric properties. Studies of the ligand-binding properties and solution structures of single and multiple mutants have been performed. The aromatic side-chains placed in the distal-heme pocket environment affect the intrinsic ligand-binding properties of the mutated subunit itself, beyond what can be explained by allostery, and these changes are accompanied by local structural perturbations. In contrast, hemoglobins with mutations in the α1β1 and α1β2 interfaces display functional properties of both “R”- and “T”-state tetramers because the equilibrium between them is altered. These mutations are accompanied by global structural perturbations, suggesting an indirect, allostery-driven cause for their effects. Combinations of the distalheme pocket and interfacial mutations exhibit additive effects in both structural and functional properties, contribute to our understanding of allostery, and advance protein-engineering methods for manipulating the O2 binding properties of the hemoglobin molecule.
Human hemoglobin (Hb) serves as a classical model of allostery in proteins, and its study has contributed greatly to understanding the relationship between structure and function in biological molecules. The two-state model of allostery in proteins described by Monod, Wyman and Changeux was based, in part, on structural and functional studies in Hb (1). According to this two-state model, cooperative O2 binding results from a conversion of Hb between high affinity, “R”- and low affinity “T”-states, and allosteric control operates by changing the equilibrium between these two states, measured as the equilibrium constant, L. Comparison of x-ray crystal structures of Hb allowed Perutz to assign R- and T-states to quaternary structures of the Hb tetramer and establish a stereochemical description of allostery that has been widely used in explaining and understanding cooperative O2 binding by Hb (2).
A large amount of work suggests that a simple two-state mechanism does not fully account for cooperativity and allostery in O2 binding. Structural studies have detected conformations of Hb distinct from those originally noted, including R2 (3), RR2 and R3 (4) conformations. NMR studies suggest that the solution structure of HbCO A is a dynamic intermediate between R and R2 conformations (5), with the R structure representing an intermediate form between T and R2 structures (6), consistent with the pathway proposed by Srinivasan and Rose (7). Alternate structures may fit with a two-state model, but only if the structural ensemble sorts into only two functional states. The low-affinity T-state form of Hb appears to comprise at least two different forms with differing affinities (8, 9). Trapping these states in a sol-gel has allowed detection of functional behaviors intermediate between classic R- and T-state traits (10). Numerous high-quality x-ray crystal structures have been reported for deoxy-Hb, but none appear to match the solution structure, which could possibly represent rapidly inter-converting species (11). To account for inadequacies of the original two-state model, several new models of allostery in hemoglobin have been proposed including the global allostery (12, 13), molecular code (14, 15) and tertiary-two state (16) models.
Understanding of allostery in Hb can be advanced by the study of mutant recombinant Hbs (rHbs) possessing interesting new structural or functional properties. Such rHbs are being produced in a rational protein engineering approach to develop a hemoglobin-based oxygen carrier (HBOC). Promising rHbs should possess: (i) appropriate O2 affinity, in order to facilitate O2 delivery in the absence of the intracellular allosteric effector, 2,3-bisphosphoglycerate (2,3-BPG), (ii) resistance to auto-oxidation, in order to increase the functional half-life of the product and inhibit the production of radical oxygen species, and (iii) limited reactivity with nitric oxide (NO) to diminish the ‘hypertensive side effect’ (17). Hbs containing αL29F, αL29W, αV96W, and βN108K substitutions possess many of these desirable qualities.
Hb Presbyterian, containing the βN108K substitution, was first found in three generations of the same family (18). Incorporation of this mutation into rHb conferred: (i) a pronounced chloride effect, (ii) an enhanced Bohr effect, (iii) lowered O2 affinity, and (iv) the capacity to switch to the low-affinity T-state conformation without changing ligation state (19). Acharya and coworkers found that the unusual T-state like character of the liganded form of transgenic (Tg) Hb (β3N108K) is: (i) detectable as an increase in geminate yield of carbon monoxide; (ii) enhanced by the allosteric effector phosphate; and (iii) manifested at both ‘hinge’ and ‘switch’ regions of the α1β2 interface, while the structure of the deoxygenated form is unperturbed compared to Hb A (20). Properties of Hb Presbyterian, including decreased O2 affinity and increased Bohr effect facilitate O2 delivery and led the research group at Somatogen to use the βN108K mutation in their first-generation rHb-based blood substitute product, rHb1.1 (21, 22). This group believed, as do we, that low affinity (P50 between 20 and 30 Torr) enhances O2 delivery, in both in vitro and in vivo (23, 24). In contrast to this view is the belief that higher O2 affinity is required in a HBOC in order to prevent premature O2 delivery and overcompensation in auto-regulation of capillary flow (25).
The αV96W substitution within the α1β2 interface that was created and characterized by our laboratory was found to increase P50 (26). Unlike many mutations within the α1β2 interface that disrupt H-bond or salt-bridge interactions and result in high O2 affinity and low cooperativity, the αV96W substitution lowers O2 affinity while maintaining high cooperativity. We have found that the CO bound form could adopt the T-type quaternary structure without changing ligation state upon addition of the allosteric effector inositol hexaphosphate (IHP) and/or a decrease in temperature (26). X-ray crystal structures have shown that this Hb creates a water-mediated H-bond network between α96W and β101E that appears to stabilize the T-state conformation, and that the R-state structure contains a β101E-β104R interaction, which is normally seen only in T-state structures (27).
The lowered affinity and high cooperativity of βN108K and αV96W substitutions have prompted our laboratory to explore the effect of the double mutation on the structure and function of Hb (19). We have found that rHb (αV96W/βN108K) has a greater tendency to switch to the T-state conformation without changing ligation state than either single mutant, and that the double mutant has a lower O2 affinity than either single mutant. This combination of mutations appears to promote low O2 affinity by stabilizing the T-state of Hb while destabilizing or altering the high-affinity R-state, creating a promising prototype for a blood substitute molecule with a high P50 value. To advance this approach, we added the B10 mutation, αL29F, to the double mutant (28). The L29F substitution in both myoglobin (Mb) and Hb was found previously to decrease the rates of both auto-oxidation and NO dioxygenation (29–32). However, this mutation markedly increases O2 affinity in both Mb and Hb to levels that would preclude O2 delivery in a normal capillary which has PO2 of ~30 to 50 Torr (30, 33–35). Thus, we have combined the αL29F mutation with the allosteric mutants to produce a rHb (αL29F/αV96W/βN108K) triple mutant with lower affinity than Hb A and lower auto-oxidation rates than rHb (αV96W/βN108K) (28).
Here, we present results of flash photolysis, equilibrium binding, 1H-NMR, and rapid-mixing studies on Hbs containing αL29F, αL29W, αV96W, βN108K, αV96W/βN108K, αL29F/αV96W/βN108K, and αL29W/αV96W/βN108K mutations. The distal pocket mutations at the B10 helical position alter the intrinsic ligand binding properties of the α-subunits, whereas the interface mutants affect the allosteric equilibrium, and both types of effects are additive in the multiple mutants.
Recombinant hemoglobins were designed, expressed and purified according to previously described procedures (36, 37). The plasmids co-express methionine aminopeptidase (Met-AP), which removes the N-terminal methionine necessary for bacterial expression and results in production of authentic Hb. An oxidation-reduction step for the rHb molecule is included in the purification procedure in order to convert incorrectly oriented heme groups to the correct orientation (for details, see refs 36, 37).
All experiments were performed in 0.1 M sodium phosphate buffer at pH 7.0 and 20 °C, with the exception of the geminate rebinding studies which were performed at 22 °C. Hb concentrations are reported on per heme or subunit basis.
Association and dissociation rate constants were determined as described previously for Hbs (38–40). Flash-photolysis experiments monitored ligand rebinding as a change in the absorbance at 436 nm following photo-dissociation of ~10% of the bound ligands. Buffers equilibrated with 1 atm O2 (1.25 mM O2) or 1 atm CO (1.0 mM CO) ensured pseudo-first order conditions for the rebinding reaction to 0.1 mM Hb (heme basis). Partial photolysis was employed to produce a comparatively large population of tri-liganded molecules so that absorbance changes would reflect the last step of the ligand binding (i.e., Hb4X3 + X → Hb4X4). Time courses were fit to a two-exponential expression, and the resulting observed rates divided by the ligand concentration to yield association rate constants. Dependence of the observed rates on ligand concentration was confirmed by experiments performed in buffers containing 0.25 mM O2 or 0.10mM CO.
Rate constants for CO dissociation from the rHbs were measured by rapid mixing techniques. rHbs in solutions containing 0.1 mM CO were mixed with buffer containing 2.0 mM (equilibrated with 1 atm) nitric oxide (NO). Free NO was present in excess and has an association rate constant approximately 10-fold greater than k'co for Hb, allowing the CO-rebinding reaction to be ignored. The CO dissociation rate constants were determined for α- and β-subunits directly from the time courses by fitting with a two-exponential expression.
Rate constants for O2 dissociation were also measured by rapid mixing techniques. In this case, 0.1 mM (heme basis) Hb in buffer equilibrated with air was mixed with buffer solutions equilibrated with 1-atm N2, 1-atm CO, and air using a SFM-400 stopped flow (Biologic, France). Proportions of mixing were manipulated to record a number of time courses over a range of [CO]/[O2] values. The time courses were fit to two-exponential expressions to measure fast and slow phases. Because the O2 rebinding reaction cannot be ignored, the rate constant, kRO2, was obtained by fitting the dependence of the observed pseudo first order replacement rate constants, robs, on [CO]/[O2] to the following expression:
The relatively low final Hb concentration of 5 to 10 µM results in some dimer formation and an over-estimation of R-state character for the low affinity mutants since the ability to switch to the T-state relies on an intact α1β2 interface.
Oxygen-dissociation curves were measured using a Hemox Analyzer (TCS Medical Products, Huntington Valley, PA), as described previously (36, 37) with modifications. Hb concentration was 60 µM and 0.3 µM catalase and superoxide dismutase were included to limit oxidation. Oxygen affinity, measured as P50, and cooperativity, measured as the Hill coefficient (n50), were determined from the resulting oxygen equilibrium curves (OECs) with an accuracy estimated at ± 10% based on reproducibility of measurements with Hb A.
1H-NMR spectra were measured for Hbs equilibrated under 1-atm CO gas using a 300-MHz Bruker Avance DRX-300 NMR spectrometer. Hb concentration was 4 mM (heme basis), dissolved in 0.1 M sodium phosphate pH 7.0 with 5% D2O added. The water signal was suppressed by using a jump-and-return pulse sequence (41). Proton chemical shifts are referenced to the methyl proton resonance of 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) indirectly by using the water signal, which occurs 4.85 ppm downfield from DSS at 20 °C, as the internal reference.
Representative time courses for CO and O2 rebinding following partial flash-photolysis are shown in Figure 1. In these experiments, ~10% of the bound ligand was photo-dissociated. As a result, the last step of ligand binding (i.e., Hb4X3 + X → Hb4X4, where X is the ligand O2 or CO) contributes virtually all of the recorded absorbance change. This measurement is model-independent. Considered within the context of the two-state MWC model, the last step of ligand binding normally represents association with the high-affinity R-quaternary state of the Hb tetramer (1). Fitting using two exponentials allows calculation of bimolecular ligand association rate constants k'Rα and k'Rβ for α- and β-chains, respectively (24). However, the set of mutant rHbs being examined contain substitutions which may promote low affinity T-state character in Hb, even in the fully saturated state, and thus, mixtures of k'R and k'T values can occur. This possibility is considered in assigning apparent rate constants for the faster and slower phases (kf and ks, respectively) of ligand rebinding.
Placement of Phe or Trp at position B10 of the α-subunit produced dramatic decreases in the rate of the slow phase of O2 binding in all four mutants containing these substitutions. The scale of these reductions is greater than that seen in rMb (L29W), which produces a 70-fold decrease in k'O2 compared to wild-type, and matches the large decreases in k'O2 seen previously in rHb (αV1 M/βV1M/αL29F) and rHb (αV1M/βV1M/αL29W) (42, 43). The amplitude of the markedly slower phase accounts for roughly 50% of the total observed reaction in each of the mutants containing the single distal mutation. The relatively unperturbed faster phase, stoichiometric amplitude distribution, and agreement with previous results for distal-heme pocket mutations of Hb and Mb lead to straightforward assignments of the slow and fast phases to α- and β-subunits of the Hb, respectively. Rate constants for ligand association were calculated by fitting time courses to two-exponential functions and are reported in Table 1, together with values previously reported for Hb A (44–46). Values of k'RαO2 measured in all four samples containing aromatic side-chains substituted into the distal binding pocket are ~10- to 100-fold smaller than the reported values of k'TαO2, indicating that the effect cannot be explained by allostery alone. The αL29F and αL29W mutations directly alter the intrinsic ligand-binding properties of the α-subunit, and that effect is clearly conserved in the multiple mutants.
Similar experiments were performed using CO as the ligand, as illustrated in Figure 1B. Again, the time courses are markedly biphasic and exhibit a slow phase rate constant, which is 10- to 300-fold lower than the reported T-state values, making the assignment of the mutant B10 α-subunit phase straightforward. This shows that the distal-heme pocket substitutions interfere with ligand association in a ligandindependent fashion. Calculated rate constants are presented in Table 1. Interestingly, the rate constants for the faster phase of CO rebinding in the rHb (αL29F/αV96W/βN108K) and rHb (αL29W/αV96W/βN108K) triple mutants match those reported for k'TβCO rather than k'RβCO suggesting that even with only 10% photolysis, the Hb tetramers are still in the T or low affinity quaternary state. This pattern differs from that seen with the same rHbs using O2 as ligand, or that seen in the single rHb (αL29F) and rHb (αL29W) mutants that do appear to switch to the R-state when three ligands have been bound. Thus, the T-quaternary conformation may occur for the Hb4(CO)3 species during the partial photolysis experiments.
To test this idea, the time courses following full photolysis have been analyzed. In the case of Hb A, full photolysis allows a significant fraction of the unliganded tetramers to switch quaternary conformations during the dye laser pulse, leading to biphasic rebinding time courses with slower apparent rate constants due to the presence of a large fraction of tetramers in the T-state quaternary form. In the case of the triple mutants, the fast and slow rate constants measured for CO binding after full photolysis are nearly identical to the rate constants determined after 10% photolysis. The full photolysis values are 0.003 and 0.11 µM−1s−1 for rHb (αL29F/αV96W/βN108K) and 0.0002 and 0.03 µM−1s−1 for rHb (αL29W/αV96W/βN108K), and can be compared to the partial photolysis values reported in Table 1. The lack of dependence of the measured rate constants on the degree of photolysis and the excellent agreement between the faster phase and reported values of k'TβCO demonstrate that these triple mutant Hbs exhibit functional properties of the T quaternary state, even with three ligands bound [Hb4(CO)3]. Such T-state behavior was not seen in the single αL29 mutants, showing that the combined αV96W and βN108K substitutions cause retention of the T-state in tri-liganded intermediates. The appearance of the T-state functional properties after partial photolysis of the two triple mutants arises from a quaternary switch to the low affinity form in the newly produced Hb4(CO)3 intermediate, which has time to occur due to ultra slow CO rebinding to the B10 mutants.
The βN108K single mutant exhibits a fast O2 rebinding phase after partial photolysis, with a rate constant of 54 µM−1s−1, and a slower phase with a rate constant of 4.3 µM−1s−1. Assignments based on these time courses alone are not straightforward. The apparent rate constant of the faster phase is similar to the average of the two R-state subunit association rate constants. The apparent rate constant for the slower phase, 4.3 µM−1s−1, is similar to that for O2 binding to the T-state subunits and suggests that this mutation by itself can slow down the rate of the last step in ligand binding by creating a population of Hb4(O2)3 tetramers in the T-quaternary conformation. The amplitudes of the fast and slow phases of rebinding support this interpretation. Following partial photolysis of native HbCO A, the fast and slow phases of bimolecular rebinding contribute roughly equally to the total absorbance change. These phases arise from reactions of the α- and β-subunits, which are similar in both: (i) spectral changes between liganded and unliganded states and (ii) geminate recombination yield. The slower phase of the βN108K mutant rebinding time course comprises only ~25% of the total signal, indicating that the phases are not due to roughly equal contributions of α- and β-subunits as was seen in Hb A, rHb (αL29F), and rHb (αL29W). Instead, this slow phase probably reflects the appearance of T-state character in the Hb4(O2)3 species.
The T-state of Hb A has very low (≤ 1%) CO rebinding geminate yield and a high overall quantum yield (~1.0) for complete photo-dissociation (47). Thus, the presence of any T-state conformations will be amplified in partial photolysis experiments since the quantum yield for the R-state hemoglobin is roughly 0.7. Thus, in our view, the partial photolysis results for rHb (βN108K) indicate the presence of a major population of tetramers rebinding at the average value of k'RO2 (i.e., at R-state tetramer rates), mixed with a smaller fraction of molecules reacting at slow T-state like rate constants. This interpretation is also supported by full photolysis experiments in which 100% of the O2 is photo-dissociated and the two phases of rebinding occur at average R- and T-state rates. For rHb (βN108K), kslow and kfast following full photolysis were measured as 6.0 and 45 µM−1s−1, respectively. These values agree well with rate constants of 4.3 and 54 µM−1s−1 measured in the partial photolysis experiments. Both full and partial photolysis results agree well with previously reported average rate constants of approximately 6 and 60 µM−1s−1 for the T- and R-state O2 association to Hb A, respectively, as shown in Table 1.
Rebinding of CO to rHb (βN108K) produces fast and slow phases with apparent rate constants of 3.2 µM−1s−1 and 0.34 µM−1s−1. As with O2 association, the rate constants for these two phases are similar to reported values of the last (R-state like) and first (T-state like) steps of CO binding, respectively. The fraction of fast and slow phases of the total absorbance change were 0.23 and 0.77, respectively, in rHb (βN108K), and not equal as is observed in Hb A. Thus, the effects of mutation on rebinding are consistent for O2 and CO. Apparent rate constants for fast and slow phases presented in Table 1 reflect the R- and T-state behavior and not different subunits.
Similar behavior was observed for the αV96W mutant. O2 rebinding time courses exhibit a large amplitude phase, with an apparent rate constant that matches the average of R-state α- and β-subunit rate constants, and a smaller amplitude slow phase reflecting T-state rebinding. Least-squares fitting of partial photolysis O2 rebinding traces produced an average slow phase rate constant of 10 µM−1s−1. The full photolysis time courses provided apparent rate constants of 10 and 60 µM−1s−. As was also seen with rHb (βN108K), the major difference between full and partial photolysis time-courses is an increase in the relative amplitude of the slow phase with increasing photolysis. CO rebinding to rHb (αV96W) shows time courses similar to those for the O2 rebinding, including a slow T-state like phase with a small amplitude. After full photolysis, the two rate constants for CO binding to rHb (αV96W) were 0.15 and 5.9 µM−1s−1 in excellent agreement with those for the T- and R-state of Hb A.
O2 rebinding to rHb (αV96W/βN108K) also proceeds in two phases with rate constants of 43 and 4.1 µM−1s−1. These values are nearly identical to those measured for rHb (βN108K), and the same R vs. T trend is observed for k'CO. The effects of the αV96W and βN108K mutations are additive on the amplitude of the slower phase. The slow phases for bimolecular 02 rebinding to rHb (αV96W), rHb (βN108K) and rHb (αV96W/βN108K) accounted for 10, 20 and 36 % of the observed absorbance change, respectively, after partial photolysis, with the double mutant, as expected showing the greatest extent of T-state behavior.
Dissociation rate constants were determined by observing displacement of ligands in rapid-mixing mixing experiments. For the O2 dissociation reaction, replacement of O2 from HbO2 by CO was monitored after mixing at a variety of ratios of free ligands. The time courses were fitted to a two-exponential function and the dependence of observed rates (robs) on ligand ratio was used to determine kO2 (Figure 2). Assignment of the resultant rate constants was straightforward for each sample since at least one phase matched rate constants previously reported for the α- and β-subunits, and in these experiments the starting liganded state is Hb4(O2)4, which even in the case of the α1β2 interface mutants normally exhibits R-state like properties.
Placement of Phe at the B10 position within the distal-heme pocket produces dramatic 70- and 40-fold decreases in kO2 for α-subunits within tetrameric rHb (αL29F) and rHb (αL29F/αV96W/βN108K), respectively, compared to α-subunits within Hb A. This finding parallels the effect seen in Mb, where the L29F mutation caused a dramatic ~10-fold reduction in kO2 compared to the wild-type value (30). The effect in Mb was explained as due to electrostatic stabilization of bound O2 by the positive edge of the phenyl-ring multipole. This phenomenon appears to be conserved in α-subunits, where the L29F substitution reduces kO2 more dramatically than in Mb. Placement of Trp at position 29 of α-subunits within tetrameric rHb (αL29W) and rHb (αL29W/αV96W/βN108K) lowered kO2 in α-subunits to a much lesser extent, 10- and 5-fold, respectively, than the αL29F mutation. Again, similar effects of the Leu(B10)Trp mutation are seen for single mutants of Mb and Hb A (24, 42, 48). The faster phase for O2 dissociation from each of the four samples containing aromatic substitutions within the α-subunit distal-heme pocket matched the rate constant measured for the native β-subunit, reflecting R-state type function.
The slower phase of the ligand dissociation reaction for rHb (αV96W), rHb (βN108K), and rHb (αV96W/βN108K) (14, 15, and 16 s−1, respectively), matches the values of 12 and 16 s−1 previously reported for the native α-subunit of Hb A, leading to straightforward assignments. The amplitudes of the slow and fast phases are roughly equal for Hb A and rHb (αV96W). However, the replacement reactions in rHb (βN108K) and rHb (αV96W/βN108K) have a larger, ~65%, fast phase and ~3-fold larger rate constants, indicating that the βN108K interface mutation acts to facilitate O2 dissociation, either by increasing the intrinsic rate of O2 dissociation from the β-subunits and/or by conferring the T-state behavior to the fully oxygenated tetramer. The distal-pocket mutations suppress this effect, suggesting that the elevation in kO2 is not due to an intrinsic change to the β-subunit itself, but instead is due to increased T-state character imparted by βN108K mutation.
Similar rapid-mixing experiments were employed to measure the replacement of bound CO through competition with NO. In this case, the high relative concentration of NO, the high association rate constants of NO, and low dissociation rate constants for NO allow direct observation of kco (49). Time courses were fitted to a two-exponential function using equal amplitudes, and the apparent rate constants uniformly match those of Hb A. This indicates that the effect of the βN108K substitution on ligand dissociation, unlike its effect on association, is ligand-dependent.
After rapid photolysis of bound ligand from the heme group, a portion of the dissociated ligands rebind internally to the iron atom to which it was originally bound (i.e., a geminate pair). The rate and amplitude of this ultrafast, nanosecond geminate rebinding process is dependent on the ability of the surrounding protein structure to capture the ligand in apolar cavities or to eject it into the solvent phase by steric hindrance (50). Normalized CO and O2 geminate rebinding traces are presented in Figure 3 and show clear differences in the fraction geminate rebinding (Fgem). The overall trend for Fgem is rHb (αL29F/αV96W/βN108K) ≈ rHb (αL29W/αV96W/βN108K) < rHb (αV96W/βN108K) < rHb (βN108K) < rHb (αV96W) ≈ Hb A and is summarized in Panel 3C.
The alteration in Fgem for rHb (αV96W), rHb (βN108K) and rHb (αV96W/βN108K) reflects a decrease in iron reactivity and iron-ligand-bond formation by proximal restraints of in plane movement of the iron atom as result of these substitutions at the subunit interfaces. The effects of the two interface mutations are roughly additive in the double mutant. Dramatic decreases in geminate rebinding have been observed in rHb (αV1M/βV1M/αL29F) and rHb (αV1M/βV1M/αL29W) (42), and it was expected that these substitutions in the distal-heme binding pocket would cause the same effect in the triple mutants. However, the observed reductions in CO geminate recombination to < 10% is more severe than was found in the single αL29F and αL29W mutants and reflects the decrease in iron reactivity due to the greater T-state character caused by the interface mutants. The allosteric transition in Hb has a pronounced effect on the CO geminate reaction reducing the population of rebinding from ~40% in the R-quaternary structure to < 1 % for the T-state (47).
The partial pressure of O2 at 50% saturation (P50) and the slope of the Hill plot at 50% saturation, the Hill coefficient (n50), are reported in Table 2. These empirical parameters are independent of any allosteric model and provide insight to the effects of mutations on O2 equilibrium binding parameters.
OECs were measured in 0.1 M sodium phosphate buffer at pH 7.0 at 20 °C, which matches conditions of the kinetic experiments, but differs in temperature, pH or both from previously reported binding curves for some of these rHbs. Wiltrout and coworkers reported P50 values of 15, 10.9, and 40 Torr for Hb A, rHb (αL29F) and rHb (αL29W) at pH values of 7.0, 6.8 and 7.13, respectively at 29 °C (35). Their results match the trend seen in Table 2 and indicate a roughly uniform ~2-fold change in P50 due to the 9 °C difference between experimental conditions. Tsai and Ho report P50 values of 8.0, 12.8, 24.5 and 48.8 for Hb A, rHb (αV96W), rHb (βN108K) and rHb (αV96W/βN108K), respectively, at pH 7.4 and 29 °C (51). This trend of decreasing O2 affinity is reproduced at pH 7.0 and 20 °C. The effects of pH and temperature on P50 appear to compensate each other in these studies, leading to exact matches for the P50 values of Hb A and rHb (αV96W) and to close agreement for the rHb (βN108K). The exact compensation between the effects of the pH and temperature is again seen when comparing our results to those of Jeong and coworkers (28). Their P50 values of 8.0, 4.0, 38 and 22 Torr for Hb A, rHb (αL29F), rHb (αV96W/βN108K) and rHb (αL29F/αV96W/βN108K), respectively, measured at pH 7.4 and 29 °C exactly match our results which were recorded at pH 7.0 and 20 °C.
The equilibrium O2 binding curve for rHb (αL29W/αV96W/βN108K) is difficult to analyze due to: (i) its very low affinity and (ii) its tendency to oxidize during the course of the experiment. Oxidation was estimated by spectral analysis to be < 5% for the other Hbs, but was as high as 9% for rHb (βL29W/βV96W/βN108K).
A representative equilibrium oxygen-binding dataset for rHb (αL29W/αV96W/βN108K) is shown in Figure 4B along with a binding curve for Hb A collected on the same day. The upper asymptote for the rHb (αL29W/αV96W/βN108K) curve is undefined and the lower asymptote deviates from that of Hb A. The discrepancy between Hb A and the mutant Hb at low O2 tensions could be due to oxidation. To account for this oxidation, an asymptotic value for the absorbance of the deoxygenated species was used, which is the average of the deoxy-Hb A lower limit and the lowest mutant deoxy-Hb absorbance value. The upper asymptotic value for the absorbance of fully oxygenated Hb A was used for the mutant protein. These scaling procedures allow estimation of the P50 of rHb (αL29W/αV96W/βN108K) as 120 ± 10 Torr in three trials. In addition to altering the absorbance values, oxidation is expected to increase the apparent O2 affinity of the Hb, suggesting that the O2 affinity of rHb (αL29W/αV96W/βN108K) may be even lower. Regardless of the exact P50 determination, it is clear that these three mutations combine to create a recombinant Hb with dramatically reduced O2 affinity.
Oxygen affinity is critical in the rational design of a HBOC. The single-mutant Hbs affect O2 affinity either directly, by interfering with ligand binding at the iron atom, or indirectly by altering allostery in the tetramer. The current studies of the combined effects of mutations, together with previous reports reveal a relatively simple dependence of P50 in the multiple mutant on the P50 of each single mutant according to the relationship:
The effect of each single mutation is expressed as the quotient of the P50 value of that single mutant rHb, P50mutation, divided by the P50 value of Hb A in the same conditions. The product of these proportionality factors and of the P50 of Hb A, yields the predicted P50 value of a multiply-mutated hemoglobin containing each of the single mutations, predicted P50mutations1+2+3.... A plot of the predicted versus measured P50 of double- and triple-mutant Hbs is shown in Figure 5. This simple linear relationship is expected to be an over-simplification, and there are relatively few rHbs for which both the single and multiple mutants have been measured. However, the correlation is striking and the prediction of the ultra-large P50 value of the rHb (αL29W/αV96W/βN108K) triple mutant is remarkably good. These multiple mutants contain substitutions at different regions of the protein, minimizing the potential for direct interference or overlapping effects of mutation. The agreement between the available data points and the line y = x suggests that this relationship may be useful, as a first approximation, for predicting P50 values of new multiple mutant Hbs based on the effects of the single mutations. Deviation from the predicted value would indicate a more complicated interaction between the single mutations.
1H-NMR spectra were collected for samples in the same buffer conditions as were used for kinetic and equilibrium studies, 0.1 M sodium phosphate buffer at pH 7.0 and 20 °C. Figure 6 shows regions of the HbCO spectra featuring exchangeable and ring-current shifted resonances with selected marker peaks labeled. The α1β1 interface was monitored by NεH resonances of α122His and α103His side-chains, which arise shifted 12.9 and 12.1 ppm from DDS, respectively (52). These two His side-chains span the α1β1 interface to contact β35Y and β131N, respectively and produce markers conserved in both liganded- and deoxy-spectra (19). The α1β2 interface was monitored by the characteristic R-quaternary state marker at 10.7 ppm. This marker arises from the NεH of the β37W side chain and moves from ~10.7 ppm in liganded Hb to ~11.0 ppm in deoxy Hb (53). The β37W side chain is located within the hinge region of the α1β2 interface, and forms a H-bond with α94D following flash photolysis, early in the R→T transition (54). Noble and coworkers systematically replaced over 20 interfacial side chains with Ala and Gly and suggested that the β37W-α140Y interaction is the “keystone” of the network stabilizing the T-state conformation (55). The distal-pocket environment was tracked by the position of the ring-current shifted α62V and β67V methyl resonances at −1.8 ppm (56). The distal valine side-chains approach within 4 Å of the bound O2 according to a recent x-ray crystal structure determination (pdb accession code 2DN1), and is a sensitive probe of the distal-pocket region comprising the O2 binding site (57). A resonance at -1.1 ppm arises from the β141L residue (58). This side-chain is located in the proximal portion of the heme pocket, ~ 4.5 Å from the proximal histidine, and adjacent to the organic phosphate binding site of the ββ cleft. Marker proton resonances of Hb A are labeled in Figure 6.
Spectra of rHb (αL29F) and rHb (αL29W) are unchanged from HbCO A control for markers in the α1β1 and α1β2 interfaces, but show clear shifts in the resonance for α62V, indicating a perturbation to the structure of the α-chain distal-heme pocket environment. These results indicate that the structural consequence of these mutations is local in nature, with the substituted side-chains altering the resonance of the nearby distal valine side-chain, without extending conformational changes to either α1β1 or α1β2 interface, or the β-subunit heme-pocket in the R-state quaternary structure. The large ring-current shift in the mutated α (L29F) subunit suggests, unexpectedly, that the α Phe(B10) side-chain causes the Cγ2 atom of Val62 to move toward the center of the porphyrin ring. Both Hbs containing the αL29F substitution exhibit this upfield shift of for the α62V Cγ2 resonance. The opposite effect is observed in the Hbs containing αL29W substitutions where the α62V Cγ2 resonance shows a downfield shift of roughly 0.5 ppm, indicating that the valine side-chain is pushed away from the center of the porphyrin ring.
These findings match the trend for aromatic substitutions at helical position B10 reported previously (28, 35). Such a direct change to the ligand-binding site is not surprising considering the large sizes of the aromatic side chains and the proximity of the α29L position to both α62V and the bound ligand. The shifts in the resonance of the α62V marker in these single mutants are conserved in rHb (αL29F/αV96W/βN108K) and rHb (αL29W/αV96W/βN108K).
The αV96W substitution causes three subtle changes in the 1H-NMR spectrum of the rHbCO. In the α1β2-interface region, the resonance for the β37W side-chain is shifted 0.1 ppm downfield, indicating an alteration to the local environment of the β37W side-chain within the α1β2 interface. This change is accompanied by small shifts of the β67V and β141L resonances, indicating a change in the heme-pocket of the β-chain and showing that this mutation acts across the (α1β2 interface to alter the heme-pocket structure of the wild-type partner subunit. A small shift of the α62V marker resonance seems to indicate a perturbation of the α-heme pocket environment, as well. Both markers in the α1β1 interface for rHb (αV96W) are unchanged from those of wild-type Hb A.
We have reported a temperature- and allosteric-effector-dependent appearance of a resonance peak at ~14.2 ppm in the 1H-NMR spectrum of rHb (αV96W) (26). This peak originates from the H-bond between α42Y and β99D in the α1β2 interface of the unliganded T-state molecule, and its emergence reflects a shift in the quaternary structure of the Hb molecule (59). The presence of both the 14-ppm T-state marker and the 10.7-ppm R-state marker indicates an intermediate quaternary state. We have further reported that IHP binding causes an increase in the strength of the T-state marker signal and loss of the R-state marker, indicating that rHb (αV96W) has a propensity to adopt the T-state quaternary structure even while fully saturated with CO (26). Thus, the structural perturbations noted in Figure 6 indicate a population, which is not uniformly in the normal R-state quaternary structure, but is tending towards the T-state structure.
The βN108K substitution clearly and dominantly perturbs the α1β1 interface, as seen in the ~0.25 ppm upfield shift of the α103H marker. Figure 6 shows that this feature is conserved in each mutant containing the βN108K mutation. Acharya and coworkers found a similar perturbation in α1β1 interface of Tg-Hb Presbyterian, which is Hb containing the βN108K substitution, and expressed in a transgenic pig (20). Alteration of the α103H marker has been reported for rHb (βN108Q), rHb (βN108R), and rHb (βN108E) indicating a role for the naturally occurring Asn side chain in the structure of the normal R-state α1β1 dimers, and demonstrating the dominant nature of this mutation on the structure of the protein (51). In addition to the shift in the α103H resonance, the βN108K substitution induces a change in the characteristic R-quaternary state marker at 10.7 ppm, which is associated with β37W in the α1β2 interface of HbCO A. In rHb (βN108K), this peak appears to have shifted upfield, indicating a marked change in the environment of β37W in the α1β2 interface, implying communication between the two major dimer interfaces in the tetrameric Hb A.
Thirdly, there is a small, downfield shift of the β141L resonance in the ring-current shifted region for the βN108K substitution, indicating a structural perturbation in the vicinity of the proximal side of the heme group. This structural feature exactly matches that seen in the rHb (αV96W) spectrum for the β141L marker of the β-subunit proximal heme-pocket area. Like the αV96W mutation, the βN108K substitution also increases the population of the T-state conformation at low temperatures and in the presence of the allosteric effector IHP, as indicated by the emergence of the resonance at 14.2 ppm (19, 20). It appears that the structural changes resulting from the βN108K substitution shown in Figure 6 propagate from the site of the mutation in the α1β1 interface to both the α1β2 interface and heme-pocket regions of the liganded molecule, promoting conversion to the T-state structure. Again, this propagation is expected for an “allosteric” mutation.
The rHbs containing both distal-pocket and interface mutations reflect the cumulative structural effects of their constituent single mutations. The two α1β1-interface markers of rHb (αV96W/βN108K), rHb (αL29F/αV96W/βN108K), and rHb (αL29W/αV96W/βN108K) contain structural perturbations seen for the rHb (βN108K). In addition, small but distinct peaks are apparent at ~14 ppm in the spectra of rHb (αV96W/βN108K) and rHb (αL29F/αV96W/β108K) and at ~11.2 ppm in the spectra of rHb (αL29F/αV96W/βN108K) and rHb (αL29W/αV96W/βN108K). An additive effect of the αV96W and βN108K substitutions on the strength of the resonance at 14.2 ppm was noted by Tsai and coworkers, and appears to underlie the appearance of that signal in our spectrum (19). The peak evident at 11.2 ppm arises from the α94D-β37W H-bond and, like the 14-ppm marker, is also in the region of a characteristic T-state quaternary structural marker. Emergence of these resonances indicates the presence of T-state character in the CO saturated tetramer of all three multiple mutants. This conclusion is supported by the waning of the characteristic R-state marker associated with β37W in the three multiple mutants, indicating that the ligand-bound forms of these multiple mutant molecules are not in a normal R-state conformation in 0.1 M phosphate buffer at pH 7.0 and 20 °C. The heme-pocket regions of these three proteins also shows the additive effects of the single mutations. The 0.1-ppm downfield shift of the β141L resonance shared in the spectra of rHb (αV96W) and rHb (βN108K) is conserved in all three multiple mutants. The pattern of migration of the α62V resonance in rHbs containing αL29F and αL29W substitutions is also conserved. The distal pocket marker resonance for αV62 Cγ2 shifts upfield by ~0.25 ppm in both αL29F containing mutants, and downfield by 0.45 and 0.55 ppm for all the αL29W containing mutants.
In general, the mutations αL29F, αL29W, αV96W, and βN108K can be thought of as producing either direct, local structural and functional changes at the ligand-binding site, which alter the intrinsic properties of that subunit, or of producing indirect, global effects in the tetramer that promote changes from the R- to T-like states. The simplest hypothesis is that the B10 mutations affect only the distal-pocket and that the α1β1- and α1β2-interface mutations affect only the equilibrium between the R- and T-quaternary states. The purpose of this work is to examine that hypothesis and verify whether these effects are additive in multiple mutants that could be potential thirdgeneration blood substitute prototypes.
Placement of Phe or Trp at position B10 of the α-subunit produced dramatic 40- and 330-fold decreases, respectively, in the rate constant for bimolecular O2 binding to α-subunits after partial photolysis. In contrast, the rate constants for the wild-type β-subunits are, as expected, relatively unperturbed. These decreases in k'O2 are similar to those observed for the same mutations in sperm whale Mb and are unlikely to be due to changes in the population of the R- and T-states given that T-state rate constants for Hb A are only ~ 5- to 15-fold smaller than those of the R-state. The Phe (B10) substitution also produces a marked decrease in kO2, but not kCO, indicating a specific electrostatic interaction of bound O2 with the positive edges of the phenyl-ring multipole. Similar favorable interactions for the Phe (B10) mutation are seen in recombinant Mb (L29F) (30). In contrast to Phe (B10), the large size of the indole ring in the Trp (B10) mutant markedly hinders the binding of all ligands due to unfavorable steric interactions, and this effect dominates, causing marked decreases in O2 affinity for all Trp (B10) mutants, including both Hb subunits and all Mbs that have been investigated. Thus, there is a balance between favorable electrostatic and unfavorable steric effects for aromatic substitutions at the B10 position. The former results in higher O2 affinity for the αL29F subunit, but the latter results in a much lower O2 affinity for the αL29W subunit. Thus, O2 affinity can be manipulated independently of the O2 association or NO dioxygenation rate constant, both of which decrease due to filling the distal pocket with the large aromatic side chains. The distal-pocket structures of rHb (αL29F) and rHb (αL29W) are clearly perturbed, as observed by changes in the αV62 Cγ2 resonance in the ring-current shifted region (Figure 6). Changes in the extent of the geminate recombination further indicate that these substitutions modify the structure of the active site. Thus, in both rHbs, the intrinsic O2 affinity of the altered subunit is changed markedly, but there is no effect on the unmodified β-subunit or on the R-state quaternary structure of the tetramer. The ability to produce changes in the intrinsic affinity of different subunits has the potential to lower the cooperativity as measured by the Hill coefficient (n50). Significant subunit heterogeneity is expected to reduce the apparent cooperativity as measured by n50 (60, 61). NMR studies have detected subunit heterogeneity of Hb A for the first step of O2 binding in the presence of organic phosphate (62). Table 2 shows that reductions in the n50 values are observed for rHbs containing αL29F and αL29W substitutions, as would be expected if the α-subunit has much higher and much lower O2 affinity, respectively, than its partner β-subunit.
The βN108K substitution in the α1β1 interface produces a small slow phase for both O2 and CO rebinding after partial photolysis (~25% of the total amplitude). The rate constant for this slow phase matches reported values for the first step of the ligand binding to the native Hb A tetramers, which are considered to be the T-state rate constants for O2 and CO binding. This lowered rate constant does not appear to be due to a change in the intrinsic reactivity of the mutated subunit, as was seen for the B10 mutants, but instead appears to be due to a shift toward the T-quaternary state. The relatively large distances, 15 and 21 Å, between the α-carbon of the βN108K amino acid and the β- and α-subunit iron atoms, respectively, seems to preclude a simple direct effect. The agreement between the values of the faster and slower apparent rate constants and those reported for the R- and T-state Hb A suggests that the partially liganded (Hb4X3) form of the mutant rHb is comprised of populations of the R- and T-type conformational states, which are not seen in the Hb A control.
Results of the 1H-NMR measurements and geminate recombination studies show that the βN108K substitution produces spectral changes indicative of a change in the population of the quaternary structures with only small alterations in the heme-pocket compared to the αL29F and αL29W mutations. Thus, as expected, the βN108K α1β1 mutation acts indirectly through global structural changes in the tetramer, to produce less reactive T-like active sites for slower ligand association, more rapid dissociation, and less geminate rebinding. Supporting this interpretation is our NMR evidence that fully liganded rHb (βN108K) and rHb (αV96W) convert to the T-state like structure in response to lowered temperatures and upon addition of the allosteric effector IHP(19, 20, 26).
The αV96W mutation produces changes that are similar to, but less dramatic than those of the βN108K replacement. Bimolecular association of O2 with Hb4(O2)3 occurs in two phases, with the slower phase accounting for only 10% of the observed absorbance change, but at a rate similar to that expected for the T-state hemoglobin. This mutation shares with rHb (βN108K): (i) the ability to switch quaternary forms without changing ligation state; (ii) perturbation of the proximal heme-pocket marker at −1.2 ppm; and (iii) lowered O2 affinity. Unlike rHb (βN108K), rHb (αV96W) shows no changes for the α1β1-interface 1H-NMR markers. Thus, a perturbed α1β1 interface is not necessary for the ability to switch quaternary forms independent of the ligation state. However, the correlation between the shifts in the −1.2-ppm resonance raises the possibility that perturbation of the β141L marker is linked to alterations at the α1β1 interface.
As described above, both rHb (βN108K) and rHb (αV96W) can convert to the T-state in fully liganded tetramers at lowered temperatures and upon addition of the allosteric effector IHP (19, 20, 26). The ability to switch quaternary conformation, even while fully saturated, is increased when these two mutations are combined in rHb (αV96W/βN108K) (51). This propensity is reflected in the 1H-NMR spectra in Figure 6, which show increasing perturbation from the native R-state-like structure according to the trend: αV96W/βN108K > βN108K > αV96W. Table 2 shows that P50 follows the same trend, indicating, as expected, that lowered O2 affinity coincides with increased ability to access the T-state conformation. This trend is evident in the cooperativity of O2 binding, measured as the Hill coefficient (n50). The value of n50 diminishes as the molecule exhibits greater T-state character in the presence of saturating ligand concentrations. Also, in the partial photolysis ligand rebinding measurements, the increases in amplitude of the slow bimolecular T-state phases and the decreases in the extent of geminate rebinding for O2 and CO follow the same trend.
The molecular code model developed by Ackers and coworkers describes O2 binding to Hb in terms of a series of intermediate steps involving intra-dimer cooperativity and a role for the α1β1 interface in influencing the overall O2 binding reaction (14, 15). Our results for rHb (βN108K) show that the perturbation of the α1β1 interface can affect O2 binding by acting upon the allosteric transition, supporting the idea of an important role of this interface in the overall reaction. In addition, the 1H-NMR spectra of all rHbs containing the βN108K substitution display a shift in the resonance at −1.2 ppm, indicating that the liganded structure of these mutant Hbs differs from the normal R-state structure of Hb A. However, the most important result is that the effects of the α1β1- and α1-β2-interface mutants are additive, a result that is important for engineering ligand-binding properties in Hb-based blood substitutes.
The appropriate O2 affinity for an Hb-based O2 carrier (HBOC) remains controversial. Hb A serves as an O2 delivery agent, with its transport capabilities maximal at its P50, which is ~28 Torr in blood. Normal capillaries have PO2 of ~20 to 30 Torr (33, 34), supporting the view that as free O2 diffuses from the capillary, it is quickly replenished from the store of HbO2 within RBCs. A high-affinity rHb (P50 < 10 Torr) will remain saturated at these normal PO2 values. Conversely, a low-affinity rHb, such as rHb (αL29W/αV96W/βN108K) may not fully saturate in the lung and will start to desaturate at higher P02 values similar to those found in arterioles and large arteries, resulting in premature O2 delivery. Thus, for Hb in RBCs, a P50 of 20 to 30 Torr is optimal and most workers feel that the same logic applies to extracellular HBOCs. However, this view has been challenged (62). Extracellular HBOCs have no unstirred layers surrounding them, permeate the cell-free plasma layer lining the vessel walls of arteries and arterioles, and as result, deliver oxygen two to three times more efficiently on an iron basis than RBCs alone (23). To prevent premature and excessive O2 delivery by HBOCs, Cole and coworkers suggest that a P50 < 15 Torr is optimal for HBOC (25). Otherwise the excess O2 delivery will cause overcompensating auto-regulatory responses that will cause vasoconstriction. These workers have argued that the hypertensive side effect of all first generation HBOCs is due to this problem of too much O2 delivery.
However, Doherty and coworkers examined the hypertensive side-effect of a large number of rHb mutants that had been well characterized with respect to O2 binding and NO scavenging, including in-vitro measures of the rate constants for NO dioxygenation. They showed a striking correlation between k'No.ox (from 2 to 70 µM−1s−1) and change in mean arterial pressure (ΔMAP) in 10% top-load, rat model. On the other hand, there was no correlation between P50 and ΔMAP for a set of rHbs with P50 values ranging from 3 to 50 Torr (32, 64). These findings strongly suggest that NO scavenging is the underlying cause of the hypertensive effect.
Our long-term goal is to limit the rate of NO dioxygenation through mutation of the distal-pocket with large amino acid side-chains, while adjusting P50 values to a range of 20 to 30 Torr for optimal O2 transport in a normal capillary. In previous work, Olson, Doyle, Lemon, and others have shown that replacement of Leu (B10) and Val (E11) with Phe and Trp can markedly reduce the rate of NO dioxygenation (29, 32, 64). In this work, we have shown that O2 affinity is profoundly altered by the Phe (B10) and Trp (B10) mutations in the α-subunits, but through adjustments to α1β1 and α1β2 interfaces that alter the allosteric equilibrium, the P50 values of the distal-pocket mutant can be re-adjusted to more favorable values, particularly in the case of the αL29F mutant.
Cooperativity in O2 binding is decreased in the mutant rHbs in the order Hb A > rHb (αL29F) ≈ rHb (αV96W) ≈ rHb (βN108K) > rHb (αL29W) > rHb (αV96W/β108K) > rHb (αL29F/αV96W/βN108K) > rHb (αL29W/αV96W/βN108K). There are two trends within this overall variation. First, in comparing the distal-pocket mutants rHb (αL29F) and rHb (αL29W), the rHb with the greater subunit difference in the O2 affinity, rHb (αL29W), also exhibits the largest decrease in the n50 value. In this mutant, there is an ordered addition of ligand, with O2 binding first to T-state β-subunits and then to the R-state rHb α(L29W) subunits that have roughly the same low affinity due to steric hindrance by the large indole ring adjacent to the bound ligand. Second, the rHbs showing the greatest manifestation of the T-state functional and structural traits, even when liganded, also exhibit low n50 values because little switching to the high-affinity state occurs until after all four ligands have been bound. In principle, high cooperativity is desired in a HBOC in order to provide more efficient O2 delivery over small changes in the PO2 values. Our mutants show that the cooperativity is reduced by either markedly increasing or decreasing the affinity of one subunit versus the other causing ordered addition of O2 with little change in the affinity even with a quaternary transition or by inhibiting the allosteric transition from low- to high-affinity quaternary structure.
In the triple mutant rHb (αL29F/αV96W/βN108K), inhibition of the T to R transition by the two interface mutations compensates for the intrinsic increase in O2 affinity caused by the βL29F substitution. The resultant triple mutant has a moderate overall affinity, low rates of auto-oxidation, and presumably low rates of NO scavenging by the α-subunit. Thus, it serves as a promising prototype HBOC molecule. In contrast, the triple mutant with the αTrp (B10) replacement, rHb (αL29W/αV96W/βN108K), starts with an intrinsically low affinity of the α-subunits and when combined with the interface mutations, results in a molecule that cannot be saturated in air. Perhaps the most remarkable result of this study is that the effects of the single distal-pocket and interface mutations are additive and their individual properties can be used to predict those of the multiple mutants. These correlations are highly encouraging for using rational protein engineering and our library of single-point mutations to design safer, more efficient, and more stable HBOCs.
†This work is supported by research grants from the National Institutes of Health (R01HL-024525 and R01GM-084614 in support of VS, TS, NH and CH, and HL-047020, GM 35649, and Welch Foundation Grant C-0612 in support of JSO), and DHM was supported by a postdoctoral fellowship sponsored by the American Heart Association (0625507U).