Additive Effects of β Chain Mutations in Low Oxygen Affinity Hemoglobin βF41Y,K66T*

In order to decrease significantly the oxygen affinity of human hemoglobin, we have associated the mutation βF41Y with another point mutation also known to decrease the oxygen affinity of Hb. We have synthesized a recombinant Hb (rHb) with two mutations in the β chains: rHb βF41Y,K66T. In the absence of 2,3-diphosphoglycerate, additive effects of the mutations are evident, since the doubly mutated Hb exhibits a larger decrease in oxygen affinity than for the individual single mutations. In the presence of 2,3-diphosphoglycerate, the second mutation did not significantly increase the P 50 value relative to the single mutations. However, the kinetics of CO binding still indicate combined effects on the allosteric equilibrium, as evidenced by more of the slow bimolecular phase characteristic of binding to the deoxy conformation. Dimer-tetramer equilibrium studies indicate an increase in stability of the mutants relative to rHb A; the double mutant rHb βF41Y,K66T at pH 7.5 showed a K 4,2 value of 0.26 μm. Despite the lower oxygen affinity, the single mutant βF41Y and double mutant βF41Y,K66T show only a moderate increase of 20% in the autoxidation rate. These mutations are thus of interest in developing a Hb-based blood substitute.

In order to decrease significantly the oxygen affinity of human hemoglobin, we have associated the mutation ␤F41Y with another point mutation also known to decrease the oxygen affinity of Hb. We have synthesized a recombinant Hb (rHb) with two mutations in the ␤ chains: rHb ␤F41Y,K66T. In the absence of 2,3-diphosphoglycerate, additive effects of the mutations are evident, since the doubly mutated Hb exhibits a larger decrease in oxygen affinity than for the individual single mutations. In the presence of 2,3-diphosphoglycerate, the second mutation did not significantly increase the P 50 value relative to the single mutations. However, the kinetics of CO binding still indicate combined effects on the allosteric equilibrium, as evidenced by more of the slow bimolecular phase characteristic of binding to the deoxy conformation.
Dimer-tetramer equilibrium studies indicate an increase in stability of the mutants relative to rHb A; the double mutant rHb ␤F41Y,K66T at pH 7.5 showed a K 4,2 value of 0.26 M. Despite the lower oxygen affinity, the single mutant ␤F41Y and double mutant ␤F41Y,K66T show only a moderate increase of 20% in the autoxidation rate. These mutations are thus of interest in developing a Hb-based blood substitute.
The search for human hemoglobin (Hb) variants exhibiting a low oxygen affinity without requiring 2,3-diphosphoglycerate (2,3-DPG) 1 is of interest in the view of producing a blood substitute. With this objective, we have previously synthesized the recombinant Hb (rHb) ␤F41Y using the genetic engineering approach (1). The naturally occurring mutated Hb ␤F41Y, first described by Burkert et al. (2), is known as Hb Mequon. The mutation ␤F41Y occurs in an important region of the subunit interface, which undergoes large rearrangements in the transition between the deoxy (T state) and the liganded (R state) conformations (3). We have shown that the recombinant Hb ␤F41Y exhibits a lower oxygen affinity than Hb A, with a well preserved cooperativity of oxygen binding and without increasing the rate of autoxidation. The decreased oxygen affinity of rHb ␤F41Y is attributed mainly to an increase in the allosteric constant, L 0 ϭ T 0 /R 0 , because the oxygen equilibrium dissociation constants for the T and R states, K T and K R respectively, were not modified (1).
Based on the crystallographic structure of Hb A (4), it appears that there could be an additional hydrogen bond in the deoxy conformation between a tyrosyl residue at the ␤41(C7) site and the carbonyl of the ␤97(FG4) His residue within the same ␤ chain. Such an interaction, coupled with the native interchain hydrogen bond between the Tyr-␣ 2 42(C7) and Asp-␤ 1 99(G1) residues, would help stabilize the deoxy state of rHb ␤F41Y.
In addition to the amino acids involved in the ␣ 1 ␤ 2 interface, other key residues are important in the cooperative ligand binding to human Hb. Specifically, the amino acids implicated in the heme contact have a crucial role. The study of naturally occurring human mutants has also confirmed the importance of these regions. Indeed, over 600 natural variants of Hb have been described (5), providing information about the role of certain residues in the structural changes between the deoxy and oxy conformations. Among these natural mutants, 54 displayed a decreased oxygen affinity, with 7 and 47 for the ␣ and ␤ subunits, respectively. In particular, Hb Chico (␤K66T) with a mutation close to the heme group exhibits a low oxygen affinity and a slight instability (6). In deoxy Hb A, the residue Lys-␤66 (E10) forms a salt bridge with the carboxyl group of one propionic acid of the ␤-chain heme; this contact does not exist in the liganded form (4). X-ray analysis of Hb Chico has shown that Thr-␤66 may form a hydrogen bond with His-␤63 via a bridging water molecule. This introduces additional steric hindrance to ligand binding to the T state (7).
With a view to study the combined effects of the mutations F41Y and K66T in the same ␤ subunit, we have produced an artificial human Hb ␤F41Y,K66T. We report here the functional properties of the double mutant compared with those of native Hb A and the singly mutated Hbs.

MATERIALS AND METHODS
The Lys-␤66 3 Thr mutation was introduced by site-directed mutagenesis into the ␤-globin cDNA containing the code for the mutation Phe-␤41 3 Tyr. The doubly mutated ␤ globin was produced as a fusion protein in Escherichia coli using the expression vector pATprTet-cII-FX-␤Gb. After purification and cleavage of the fusion protein by digestion with bovine activated coagulation factor X, the ␣ 2 ␤ 2 tetramer was reconstituted in the presence of cyanohemin and native carbonmonoxy ␣ subunits (8,9). The structure of the mutated chain was checked by reversed-phase high performance liquid chromatography of tryptic digest and amino acid analysis of the two mutated peptides. The purity of the rHb was controlled by isoelectric focusing electrophoresis with a pH gradient ranging from 6.0 to 8.0. The heat stability was tested by incubating rHb ␤F41Y,K66T and Hb A (100 M on a heme basis) at 65°C in 10 mM phosphate buffer, pH 7.0 (10).
The tetramer-dimer equilibrium was studied by gel filtration on a Superose 12 HR 10/30 column (Amersham Pharmacia Biotech, Uppsala, Sweden) as described by Manning et al. (11). All experiments were performed at 25°C, in 150 mM Tris acetate buffer, pH 7.5. For concentrations of Hb ranging from 2 to 500 M on a heme basis, 10-l aliquots were applied and eluted at a flow rate of 0.4 ml/min. The absorbance of the eluent was measured at 415 and 280 nm. Diaspirin cross-linked * This research was supported by the Institut National de la Santé et de la Recherche Médicale, the Direction de la Recherche et de la Technologie (contract 92/177), and the Faculté de Médecine Paris Sud. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
(DLC) Hb was used as a control for undissociable tetrameric Hb, and the peak position of the dimer was determined using the dimeric natural mutant Hb Rothschild (12).
The rate of oxidation for liganded Hb samples was measured by absorption spectrophotometry (SLM-Aminco DW2000) at 37°C for samples under 1 atm of oxygen or under air (13). Hb solutions were 40 M in heme in 20 mM potassium phosphate at pH 7.0.
Oxygen equilibrium curves were recorded with a continuous method using the Hemox Analyzer system (TCS, Huntington Valley, PA) (14). The amount of metHb calculated from the visible absorption spectra was found to be less than 5% at the end of the recordings.
Fluorescence studies were performed on 10 M Hb solutions (on a heme basis) in 10 mM phosphate buffer, pH 7.0, using an SLM 8000 spectrofluorometer. Emission spectra were recorded for both the buffer and the liganded Hb samples.
Kinetics of CO or oxygen recombination were obtained after flash photolysis using 10-ns YAG laser pulses (Quantel, France) providing 160 mJ at 532 nm. Samples were in 1-mm cuvettes, with observation at 436 nm (15). Measurements were made at 25°C, 50 mM bis-Tris at pH 7.0, 100 mM NaCl, for samples equilibrated under air or 1 or 0.1 atm of CO. Kinetics were recorded at different laser energies to probe the Hb tetramer at different ligand saturation levels.
The CO dissociation rate was determined from the kinetics of replacement of CO by NO, with measurements of full spectra versus time using a diode array spectrophotometer (HP 8453).
Kinetics of O 2 replacement by CO were measured with a stopped-flow apparatus (Biologic, France) with detection at 420 nm. Experimental conditions were 50 mM bis-Tris at pH 7.0, 100 mM NaCl, at 25°C. The dead time of this apparatus with a cuvette of 1-cm optical path length is 2 ms. Hb samples equilibrated under 1 atm of oxygen were mixed with a solution equilibrated under 1 atm of CO containing the oxygen scavenger sodium dithionite. The final concentrations after mixing were 5 M in heme, 5 mM sodium dithionite, 0.6 mM O 2 , and 0.5 mM CO.

RESULTS
Properties of rHb ␤F41Y,K66T-Analysis of the purified rHb ␤F41Y,K66T by isoelectric focusing showed that it migrated as a single band (pI ϭ 6.4), with a more cathodic position relative to Hb A (pI ϭ 6.98). Fluorescence studies did not show significant differences between the doubly mutated rHb and native Hb A; the highly quenched emission indicates a correctly reconstituted (folded) rHb. The UV and visible absorption spectra of rHb ␤F41Y,K66T in carboxylated and oxygenated forms were identical to those of native Hb A. Notably, the ratios of absorbance intensity between the Soret band and UV peak at 280 nm were normal, 4.87 and 3.5 for the mutated HbCO and HbO 2 , respectively. The oxy form of rHb ␤F41Y,K66T exhibited the same fraction denaturation as Hb A after a 20-min incubation at 65°C.
Tetramer-Dimer Equilibrium-Hb A, rHb A, rHb ␤F41Y, and rHb ␤F41Y,K66T in the liganded form were eluted as a single peak whose position varied between tetrameric and dimeric forms when the Hb concentrations were in the range of the dissociation constant K 4,2 value. The K 4,2 value for rHb A was 2-fold higher than that for natural Hb A, in agreement with studies by Fronticelli et al. (16); at high protein concentrations (favoring Hb tetramers), the functional properties of Hb A and rHb A are similar as described previously (17,18). The K 4,2 values for rHb ␤F41Y and rHb ␤F41Y,K66T were 3and 6-fold decreased, respectively, compared with that for control rHb A (Table I).
Autoxidation-At 37°C, the oxidation rate of rHb ␤F41Y,K66T under 1 atm of oxygen was increased by 20% compared with that of native Hb A. Under air, the rHb ␤F41Y,K66T, which is less oxygen-saturated than Hb A, exhibited a 2-fold increase in oxidation rate relative to Hb A.
Oxygen Equilibrium Curves- Fig. 1 shows the experimental Hill plots obtained for Hb A, rHb ␤F41Y, and rHb ␤F41Y,K66T in the absence (Fig. 1A) or in the presence of chloride anions (Fig. 1B). Table II   ␤F41Y,K66T in the presence and in the absence of chloride anions.
In the absence of chloride, the oxygen affinity of the rHb ␤F41Y,K66T was decreased 2.5-fold relative to that of control Hb A and by about 1.5-fold relative to that of rHb ␤F41Y. The shift in log P 50 was 0.34 for Hb Chico, 0.18 for rHb ␤F41Y, and 0.42 for the double mutant rHb ␤F41Y,K66T, thus showing a partial additivity of the two mutations (⌬log P 50 ϭ 0.52 for a maximal additivity effect). The Hill coefficient at half-saturation (n 50 ), an index of oxygen binding cooperativity, was more decreased for rHb ␤F41Y,K66T than for rHb ␤F41Y.
In the presence of chloride, the double mutant rHb ␤F41Y,K66T still exhibited an oxygen affinity lower than Hb A, Hb Chico, and rHb ␤F41Y. Nevertheless, the ⌬log P 50 for rHb ␤F41Y,K66T was equal to 0.45 (0.32 in the presence of 2,3-DPG). This indicates that the effects of the two mutations were no longer additive in the presence of effectors. The values of ⌬log P 50 corresponding to the oxygen-linked heterotropic allosteric effectors showed that the chloride, 2,3-DPG and alkaline Bohr effects were in the normal range of values (Table III).
From the experimental curves (Fig. 1), the allosteric parameters were fitted to the equation of the two-state allosteric model (19) to obtain the oxygen dissociation constants K T and K R for the T and R states, respectively (Table IV).
In all experimental conditions, the equilibrium curves could be simulated with values of K T and K R for rHb ␤F41Y,K66T and rHb ␤F41Y, similar to those for Hb A, indicating that the mechanism of the low oxygen affinity was mainly due to the change in the allosteric equilibrium. The allosteric parameter L was higher for rHb ␤F41Y,K66T than for Hb A and rHb ␤F41Y. The switchover point indicates that the allosteric transition T 3 R for rHb ␤F41Y,K66T occurs at a higher oxygen saturation level than for Hb A and rHb ␤F41Y. The calculated amount of rHb ␤F41Y,K66T in the T state for tetramers with three ligands was considerably increased: 85% versus 38 and 16% for rHb ␤F41Y and Hb A, respectively. This large increase in the fraction of T state is due to the presence of the mutation ␤K66T.
Kinetic Studies- Fig. 2 shows the recombination traces of CO after photodissociation. For rHb ␤F41Y,K66T as for Hb A, the traces were biphasic as expected for tetrameric Hb. At low CO photodissociation levels (5%), the CO recombination kinetics of Hb A were fast and monophasic because in these conditions the majority of the photodissociated tetramers are triliganded and remain in the R state. In the same conditions, the CO recombination kinetics of rHb ␤F41Y,K66T still exhibited some slow phase. These results show, as for the oxygen equilibrium curves, that the allosteric equilibrium of partially liganded species of rHb ␤F41Y,K66T are displaced toward the T state relative to Hb A.
The oxygen association and dissociation rates for Hb A and rHb ␤F41Y,K66T were similar (Table V). These results are The allosteric parameters (L, K R , and K T ) were obtained after fitting the experimental curves to the equation of the two-state allosteric model (19) by using a nonlinear least-squares procedure. K R and K T (mm Hg) are the oxygen dissociation constants for the R and T states, respectively; L is the allosteric constant (T 0 /R 0 ); c ϭ K R /K T ; the switchover point i s was calculated as -log L/log c; % T3 is the amount of triply liganded T state species calculated as (Lc 3 )/(1 ϩ Lc 3 ). The standard error per point was typically 0.003-0.006.  consistent with the equilibrium data that show that the value of K R for rHb ␤F41Y,K66T was similar to that of Hb A. However, the CO kinetics suggest a change in the R state properties as well as the shift in allosteric equilibrium.

DISCUSSION
In order to obtain a modified Hb that could be used as a Hb-based artificial oxygen carrier, we have used two different strategies to decrease the oxygen affinity of human Hb. We first introduced a mutation in the allosteric interface (␤F41Y) to shift the equilibrium toward the T state by inducing an additional hydrogen bond in the T state conformation, without perturbing the R state conformation. To further decrease the oxygen affinity, we have then investigated the association of this mutation with a second substitution known to have a similar effect, but due to another mode of action.
In normal human Hb, the Phe-␤41 belongs to a cluster of three highly conserved phenylalanine residues, ␤41(C7), ␤42(CD1), and ␤45(CD4), that participate in the formation of the hydrophobic heme pocket; in Hb A, the Phe-␤41 has contacts with the heme moiety and is critical to the structural integrity and function of the Hb molecule. Decreased oxygen affinity was observed for the naturally mutated Hb Denver Phe-␤41 3 Ser (20) and Hb Bruxelles deletion of Phe-␤41 (21). Hb Denver is an unstable Hb variant; the smaller serine residue may impede movement during the allosteric transition of the FG corner of the ␣ subunits along the C helix of the subunits (20). The mutated Hb corresponding to naturally occurring Hb Mequon (2) was synthesized after site-directed mutagenesis to study the functional properties of the pure form (1). The working hypothesis of the ␤F41Y mutation was to induce an additional hydrogen bond in the T state conformation without perturbing the R state conformation. If the ␤41 tyrosyl residue forms a new hydrogen bond to the carbonyl of the ␤97 (FG4) histidine residue within the same ␤ chain, then the deoxy conformation may be stabilized. While the oxidation rate and heat stability are similar to those of Hb A, this Hb exhibited a 2-fold decrease in oxygen affinity compared with that of Hb A, due to a shift in the allosteric equilibrium.
When two effects have an independent mode of action, the combined effects may be additive. In general, a partial additivity is observed, as shown for the free energy of dimer-tetramer equilibrium for Hb mutants (22). We thus associated the ␤F41Y mutation with a substitution of a residue on the distal side of the heme (␤K66T) to directly act on the heme environment. The abnormal Hb Chico, Lys-␤66 3 Thr, displays a significantly decreased O 2 affinity (6, 7). The ␤66 residue is not involved in the subunit interface and is not expected to have a direct effect on the subunit dissociation. x-ray analysis of the deoxy conformation of Hb A shows that Lys-␤66 makes an ionic bond (salt bridge) between its ⑀ amino group and the carboxyl group of propionate-7 of the heme (4). The disruption of the ionic bond provided by Lys was first suspected to be the cause of the altered O 2 binding. Other single amino acid substitutions have been investigated to date by site-directed mutagenic protein engineering, Lys-␤66 3 Ser and Lys-␤66 3 Arg, that lead also to decreased O 2 oxygen affinity (23), while the natural mutant Hb I-Toulouse, Lys-␤66 3 Glu (24,25), and the engineered Hb Lys-␤66 3 Gly (26) have been reported to have normal oxygen binding function. These substitutions also eliminate the ionic bond; thus, the rupture of the bridge involving the Lys-␤66 would not be the cause of the low oxygen affinity of Hb Chico. The fact that the monomeric as well as the tetrameric ␤ chains of Hb Chico also show a decreased oxygen affinity suggests a tertiary effect (7).
Hb tetramers dissociate reversibly into ␣␤ dimers, involving the breaking of certain bonds at the ␣ 1 ␤ 2 and ␣ 2 ␤ 1 interfaces. The tetramer dissociation constant (K 4,2 ϭ [dimer] 2 /[tetramer]) is on the order of 1 M for liganded Hb A, while the deoxy (T state) tetramers are much more stable (27). The literature  values vary greatly for this parameter. This is in part due to the influence on the solvent conditions; e.g. K 4,2 is about 0.2 M at low ionic force but increases to 1 M at 100 mM NaCl. Protein folding may also play an important role, since the values reported for rHb are often twice that of native Hb A. For mutant Hbs, the value of the tetramer dissociation constant may be a useful probe of the stability of the protein (11,22). In addition to an altered oxygen affinity, Hb Chico Lys-␤66 3 Thr shows an increase in the fraction dimer and in the autoxidation rate (6,7). However, we observed a decrease in the K 4,2 value for the association of both mutations F41Y and K66T. The K 4,2 value for liganded rHb ␤F41Y is decreased 3-fold compared with that of rHb A (Table I), and an additional decrease of a factor of 2 was observed for the double mutant.
Quantification of the additivity requires a choice of parameters. The P 50 value is one obvious choice for its simplicity and reliability of measurement; however, it depends on several microscopic parameters that determine the intrinsic affinity and allosteric equilibrium. An increase due to the allosteric transition could be compensated by a decrease in K T or an increase in the fraction dimers. In the absence of external effectors, the rHb ␤F41Y,K66T exhibits a lower oxygen affinity than for either single mutation; the effects of the mutations are additive. For Hb A, the distance between the residues ␤41 and ␤66 is 15.4 and 17 Å for the T state and R state, respectively. The two residues have no direct contact, as illustrated in Fig. 3, and in principle do not have a correlated participation in their effects on the oxygen affinity. The partial additivity of their effects on the intrinsic oxygen affinity of rHb ␤F41Y,K66T is therefore probably due to two independent mechanisms. One can also consider the combined effects of the mutations with the effectors such as 2,3-DPG or inositol hexaphosphate; with 2,3-DPG or inositol hexaphosphate, the additivity is less pronounced (Fig. 2). This does not imply that the two effects are no longer additive but rather that the change in P 50 may no longer have the same magnitude. The flash photolysis kinetics are sometimes more sensitive to changes in the allosteric equilibrium for a late switchover point, e.g. when the transition from T to R occurs only after binding the third ligand. Low photodissociation levels can isolate the reaction for binding the fourth ligand.
The ligand binding properties of this new rHb demonstrate that several mutations may be introduced to obtain combined effects on the oxygen affinity. This would eliminate the need for an external effector, such as 2,3-DPG, as required for a Hbbased blood substitute. However, lower oxygen affinities are often accompanied by an increased autoxidation rate. Unless these parameters can be decoupled, one must accept a compromise of the best oxygen affinity and lowest autoxidation rate (28). The present results show that Hb can be genetically engineered to regulate the oxygen affinity. By associating sev-eral smaller changes, the perturbations in the protein stability can be minimized. This appears to be the case for the present study of two mutations in the Hb ␤ chains, one inducing a shift in the allosteric equilibrium (␤41 mutation) and the other inducing a decrease in the intrinsic oxygen affinity (␤66 mutation). There is no conflict in accommodating both changes, and an overall additive effect is observed.