INTRODUCTION

The homodimeric hemoglobin (HbI)¹ from the blood clam Scapharca inaequivalvis offers a simple model system for studying communication between two chemically identical subunits. Analysis of the mechanistic details of this intersubunit communication is being pursued to provide insights into the regulation of protein function.

Scapharca HbI binds oxygen cooperatively, with a p50 of 7.8 torr and a Hill coefficient (n) of 1.5 at 20 °C, neither of which changes as pH varies from 5.5 to 9.0 (1). Although the individual subunits of HbI have the same myoglobin fold as mammalian hemoglobins, the quaternary assemblage is radically different (2). Upon oxygen binding by HbI, only small tertiary changes are seen at the subunit interface in contrast to the relatively large quaternary changes observed with mammalian hemoglobins (3). Analysis of structures of this hemoglobin at 1.6 Å for the deoxygenated molecule, 1.4 Å for the CO-liganded form, and 1.7 Å for the oxygenated form has provided a framework for understanding the role of individual side chains and interfacial water in mediating cooperativity (2, 4).

Phenylalanine 97 has a central role in the proposed structural mechanism of cooperativity (2, 4). The side chain of Phe⁹⁷ undergoes the largest ligand-linked conformational change in HbI. In the deoxy state, it is tightly packed in the heme pocket, whereas upon ligand binding it is displaced from the heme pocket into the subunit interface. The lowered oxygen affinity of the deoxy state appears to result primarily from the packing of Phe⁹⁷ in the heme pocket, where it restricts movement of the iron atom into the heme plane and lengthens a hydrogen bond involving the proximal histidine. Upon ligation, movement of Phe⁹⁷ into the subunit interface is coupled with disruption of a well ordered interfacial water cluster and movement of the heme groups deeper within each subunit. These effects within a subunit alter the nature of interactions between subunits, which presumably encourages movement of the Phe⁹⁷ side chain (second subunit) into the interface, allowing the second subunit to attain a high oxygen affinity state prior to ligand binding (2). In view of its significant role in the proposed cooperative mechanism of HbI, Phe⁹⁷ is clearly indicated as a target for mutation. Phe⁹⁷, four residues from the proximal histidine at position 101, is analogous to the fourth residue of the F helix in mammalian hemoglobins, which is a leucine in all known vertebrate hemoglobins and most invertebrate hemoglobins. The smaller size of the leucine side chain in these hemoglobins allows it to remain packed in the heme pocket in both liganded and unliganded states.

In this paper, we present functional and high resolution crystallographic studies of the Leu⁹⁷ HbI (F97L) mutant. This mutation results in increased oxygen affinity and diminished cooperativity that appear correlated with observed ligand-linked structural transitions. Our results confirm the crucial role of residue Phe⁹⁷ in modulating oxygen binding by HbI, while demonstrating the persistence of residual cooperativity and suggesting that parallel pathways exist for information transfer between subunits.

MATERIALS AND METHODS

Escherichia coli strain TG1 was used as the host strain for propagating recombinant bacteriophage M13mp18 vector and for in vitro site-directed mutagenesis. E. coli strain W3110lacIq L8 was the host used for overexpression of recombinant HbI (5).

In vitro oligonucleotide-mediated, site-directed mutagenesis was used to generate the F97L mutant. The initially constructed gene for HbI bearing 5 and 3 termini complementary to KpnI and SacI sites was cloned into the polylinker region of a M13mp18 bacteriophage vector (New England Biolabs, Beverly, MA), which had been cut previously with the same two restriction endonucleases. Following several rounds of plaque purification using TG1 cells as hosts, recombinant single-stranded DNA was isolated (6). Subsequently, a 19-mer oligonucleotide (University of Massachusetts Medical Center DNA synthesis facility) complementary to the region of the gene coding for residue Phe⁹⁷ was used in a site-directed mutagenesis protocol (Sculptor in vitro mutagenesis system, Amersham Life Science Inc.) to mutate residue 97 to leucine. Following three rounds of plaque purification, single-stranded DNA was isolated and the changes in DNA sequence were confirmed by dideoxy sequencing using the Sequenase kit (U. S. Biochemical Corp.). Double-stranded DNA of the replicative form of the vector was then prepared and subjected to AccI and SacI restriction cleavage to isolate a segment of the HbI gene that contains the leucine 97 mutation. This fragment was then subcloned into a similarly digested wild-type HbI gene borne on a pCS26 expression vector that, as reported earlier, has allowed for significant overexpression of recombinant, wild-type HbI protein (5).

Recombinant F97L was overexpressed in E. coli and purified as described for wild type (5).

Spectroscopic measurements were carried out at 20 °C with a Cary 3 spectrophotometer at a protein concentration of about 6 × 10⁵ M (heme) in 0.1 M phosphate buffer at pH 7.0. Oxygen equilibria were measured using a tonometer, according to Rossi-Fanelli and Antonini (7), with a reducing system to minimize protein oxidation (8).

Circular dichroism spectra were measured at 20 °C in 0.1 M phosphate buffer at pH 7.0 in a Jasco J710 spectropolarimeter equipped with a Jasco J700 processor in the range 380-650 nm. The deoxygenated derivative was obtained by addition to the oxygenated protein of a small volume of a 1.0 M sodium dithionite solution prepared in deoxygenated buffer.

Kinetic measurements using flash photolysis were performed using a YAG laser giving a 9-ns flash of 25 mJ at 532 nm. The beam was telescoped to give a parallel beam 2 mm in diameter colinear with the observing beam supplied from a 75-W xenon arc passing through a blue filter and imaged at the sample within the photolysis beam. The observing beam continued through a Spex 250-mm monochromator to a photomultiplier and amplifier (Teledyne-Philbrick 1321). The output from the amplifier was digitized by a DAS-50 12-bit A/D converter (Metrabyte) and transferred to an IBM personal computer. The combination had an overall response time of 2 µs.

Stopped flow measurements were performed with the apparatus described by Gibson and Milnes (9) in the slightly modified version of the Durrum Corp. (Palo Alto, CA). Data were digitized and recorded as described for flash photolysis using a tungsten lamp as light source. The dead time of the stopped flow experiment (2-cm path) was 2.4 ms. For the low oxygen experiments the apparatus was filled with a dilute solution of sodium dithionite and allowed to stand overnight. It was then washed out repeatedly with buffer bubbled with pure nitrogen before introducing the working solutions. Low oxygen concentrations were obtained by mixing air-equilibrated buffer with nitrogen-bubbled buffer in a syringe. For the oxygen-pulse experiments a strong (5 mM) solution of deoxygenated hemoglobin was diluted into nitrogen-bubbled buffer and solid dithionite was added (10).

A kinetic equivalent of the two-state model was then developed to simulate the progress of oxygen saturation in oxygen-pulse and other kinetic experiments using a two-step Runge-Kutta process (11). To make comparisons with experimental data, a set of oxygen-pulse experiments with a series of initial concentrations of oxygen was performed and corresponding simulations were run. The early stages of both the experiments and the simulations were then discarded and the remaining time courses, which do not depart from a single exponential, were fitted using a nonlinear least squares program. In addition to the kinetic parameters for the two-state model it is necessary to supply values for the removal of oxygen from solution by dithionite. This was done empirically with attention directed to reproducing the peak fractional saturation observed with the experiment.

Deoxy and carbonmonoxy F97L crystals were grown as described in Ref. 2, except that both crystal forms were grown from solutions of 40-50 mg/ml hemoglobin. Microseeds obtained by crushing corresponding wild-type HbI crystals were used to nucleate mutant HbI crystal growth. The morphologies of both crystal forms were similar to those of wild-type recombinant HbI.

X-ray diffraction data were collected from crystals at room temperature on a R-AXIS IIC imaging plate system mounted on a Rigaku RU 200 rotating anode generator (Molecular Structure Corp., The Woodlands, TX). Data were processed using software supplied by the manufacturer.

The F97L-CO crystals were essentially isomorphous to native crystals, showing the symmetry of monoclinic space group C2, and cell constants a = 93.25 Å, b = 43.98 Å, c = 83.50 Å, and  = 122.03°. A total of 116 frames was collected from two crystals with an oscillation range of 1.5 to 1.8° for each frame. The overall R_merge for 40,595 unique reflections with I/ > 1 was 5.52% (R_merge = I  <I>/ I, where I is the measured intensity of each reflection and <I> is the mean value for the corresponding unique reflection). This data set represents 93.5% of the predicted unique reflections to Bragg spacings greater than 1.6 Å and 88.8% to 1.5 Å.

Deoxygenated F97L HbI crystals showed the symmetry of orthorhombic space group C222₁, with cell constants of a = 91.99 Å, b = 44.27 Å, and c = 143.83 Å, essentially isomorphous to native deoxy crystals. The deoxy F97L HbI data set represents diffraction data from a single crystal that was collected in two different orientations. Initially the large c axis (143.8 Å) was aligned horizontally (perpendicular to the spindle axis). The close spacings between reflections along the c* axis necessitated a small oscillation range (0.9°) in this orientation for the 34 oscillation frames collected. In a second set of frames, the crystal was reoriented with its c axis vertical (parallel to the spindle axis). This allowed for the use of a larger oscillation range of 2.5°. A total of 30 oscillation frames was recorded from this crystal orientation. The overall R_merge for 27,590 unique reflections derived from 95,839 independent observations with I/ > 1 was 7.33%. These represent 84.1% of the predicted unique data to Bragg spacings greater than 1.7 Å.

The starting models for both CO-liganded and deoxy Leu⁹⁷ HbI structures were the native HbI structures (2) in which residue Phe⁹⁷ was mutated to leucine using the macromolecular modeling program "O" (12). Also, the water molecules associated with these structures were deleted and alternate conformers of protein side chains eliminated from the Protein Data Bank coordinate files (entry codes 3SDH and 4SDH).

These models were then refined using the XPLOR package (13). Prior to the commencement of refinement, 10% of randomly selected reflections from each data set were designated as test reflections for use with the Free-R cross-validation method (14). Two cycles of simulated annealing were performed, with the first cycle incorporating noncrystallographic symmetry restraints upon the model. The protocol for simulated annealing that was used called for "heating" the molecule to 1000 K followed by 50 steps of 0.5-fs molecular dynamics. After each set of 50 steps, the temperature was lowered by 25 K and 50 new dynamics steps were performed. This procedure was continued until a temperature of 300 K was reached and was followed by 60 cycles of conjugate gradient minimization combined with B-factor refinement.

Addition of water molecules to the models was initiated at this stage using the programs PEAKS and LOCATE written by Wayne Hendrickson (Columbia University, New York). Peaks of electron density greater than 3.5  in F_o  F_c maps and 0.8  in 2F_o  F_c maps that satisfied hydrogen-bonding criteria were identified as locations for water molecules. This procedure was reiterated, resulting in the addition of five and four shells of water to the CO-liganded and deoxy structures, respectively. Each model was subjected to 120 cycles of conjugate gradient minimization combined with B-factor refinement using XPLOR following the addition of each shell of water molecules. Water molecules with refined B-factors greater than 50 Å² were deleted.

The model was then examined in conjunction with 2F_o  F_c and F_o  F_c maps to locate poorly placed protein side chains and missing water molecules, which were then built. This was followed by 120 cycles of conjugate gradient energy minimization combined with B-factor refinement.

RESULTS

Mutation of phenylalanine 97 to leucine (F97L) in Scapharca HbI resulted in increased ligand affinity and diminished cooperativity as observed by both equilibrium and kinetic measurements. The p50 value for oxygen binding, based on equilibrium measurements, decreased from 7.8 to 1.0 torr, and the Hill coefficient decreased from 1.5 to 1.16 ± 0.05. The statistical analysis of all the data sets (n = 12) indicated that the Hill coefficient was significantly higher than 1.0, such that cooperativity, although drastically reduced, was not abolished.

A summary of kinetic experiments on ligand binding is presented in Table I. The increased oxygen affinity observed in equilibrium measurements was primarily due to a decrease in the oxygen dissociation rate, particularly at low O₂ saturation. Oxygen-pulse and CO-replacement experiments (15) showed a greater than 10-fold decrease in the dissociation rate at low oxygen saturation compared with HbI. In contrast to HbI, there was little or no change in the oxygen-dissociation rate for F97L as a function of varying levels of oxygen saturation (Fig. 1). Additionally, measurements of CO binding to F97L by both stopped flow and flash photolysis showed a clear increase in combination rates compared with wild type.

Table I. Kinetic ligand binding parameters to Scapharca HbI and F97L Parameters are intrinsic (i.e. no statistical factors). Subscripts define the binding step based on a consecutive two-step scheme following Adair (22). A prime designates an on rate, and no prime indicates an off rate. Parameters are intrinsic (i.e. no statistical factors). Subscripts define the binding step based on a consecutive two-step scheme following Adair (22). A prime designates an on rate, and no prime indicates an off rate. Oxygen on rates Oxygen off rates CO on rates k1 k2 k1 k2 l1 l2 µM1 s1 s1 µM1 s1 F97L NDa 26 40b 45b 0.5 2.3 HbI 11c 16c 490c 50d 0.09d 0.2d a ND, not determined. b Oxygen pulse measurements (15). c From Ref. 23. d From Ref. 24.

1

1

The absorption spectra of the oxygenated and CO-liganded derivatives of F97L were almost identical to those reported for native HbI (1). In contrast, significant differences were observed for the deoxygenated derivative. In the Soret region, the band line shape was wider and more symmetric and the absorption maximum was shifted to lower wavelengths by 4 nm. In the visible region, the unusual shoulder present at 590 nm (Q_o band) in wild-type HbI was absent (data not shown).

The circular dichroism spectra of deoxy F97L (Fig. 2) confirmed the peak shift observed in the absorption spectra and displayed additional interesting features: (i) the overall molar ellipticity values were decreased by 20-25% with respect to HbI, a value that is beyond experimental error; (ii) the ellipticity of the Q_o band was reduced; and (iii) the splitting of the Soret band (B_o band), which gives rise to the characteristic negative peak at 419 nm in HbI, was greatly diminished in F97L.

Circular dichroism spectra of deoxygenated HbI () and F97L (- - - -) in the Soret (a) and visible (b) regions.

Upon substitution of Phe⁹⁷ with the smaller leucine side chain, the strong Q_o band at 590 nm, which is diagnostic of proximal strain in wild-type HbI (1, 16), was reduced. In addition, the asymmetry of the energy and polarization of the electronic transition moments of the components of the B_o band, which gives rise to splitting of the Soret band, was diminished in the mutant in which symmetry was restored. The overall decrease in molar ellipticity in F97L supports the coupled oscillator theory of Hsu and Woody (17), since the removal of an aromatic ring from within 4 Å of the heme plane is expected to decrease the rotational strength of the coupled -* transitions of aromatic residues and heme, thus decreasing the intensity of the Cotton effect.

The models for deoxygenated and CO-liganded F97L have been refined against the x-ray data to conventional R-factors of 18.3 and 18.6%, respectively, with excellent stereochemistry (Table II). An electron density map of the heme region of one subunit is shown in Fig. 3. Both the unliganded and liganded structures of F97L were found to be strikingly similar to the corresponding structures of wild-type HbI, with significant differences confined to the disposition of residue 97 side chain, F-helix geometry, and the interfacial water structure. Evidence of similarities in the overall structure is presented in Fig. 4.

Table II. Refinement statistics on Scapharca F97L CO Deoxy Resolution limits (Å) 10.0-1.5 10.0-1.7 No. of reflections used in refinement (F > 1 ) 37,557 24,848   R-factora 18.6% 18.3% No. of test reflections for Rfree (F > 1 ) 3038 2742   Rfree 23.1% 24.4% No. of nonhydrogen atoms   Hemoglobin 2316 2312   Solvent 222 174 Root mean square deviation from ideality   Bond lengths (Å) 0.010 0.011   Bond angles (°) 2.2 2.2   Dihedrals (°) 18.5 18.2   Impropers (°) 1.5 1.5 Average B-factors (Å2)   Main chain 16.7 20.4   Side chain 19.8 23.4   Heme atoms 16.0 17.7   CO ligands 14.4 -   Solvent atoms 37.2 36.6 a R =   Fo  Fc /Fo, where Fo is the observed structure factor and Fc is that calculated from the model.

Simulated annealing omit F_o  F_c map of F97L-CO.

Stereo diagram showing an -carbon trace of HbI (solid lines) and F97L (dashed lines) for the deoxy state (top) and the CO-liganded state (bottom).

As shown in Tables III and IV, the overall B-factors were slightly higher in deoxy F97L than in deoxy wild type and were nearly the same for the CO-liganded structures. In contrast to the similarity in overall B-factors, the core interface water molecules showed significantly higher B-factors in F97L than in wild type. This indicates that these water molecules, which have been implicated in cooperative function in wild-type HbI (18), are less well ordered in F97L.

Table III. Comparison of average B-factors (Å2) between deoxy HbI and deoxy F97L structures Deoxy HbI Deoxy F97L Main chain atoms 18.0 20.4 Side chain atoms 19.4 23.4 Heme atoms 15.1 17.7 Water atoms 34.2 36.6 Core interface water molecules (17) 19.6 29.5

Table IV. Comparison of average B-factors (Å2) between HbI-CO and F97L-CO structures HbI-CO F97L-CO Main chain atoms 16.6 16.7 Side chain atoms 18.2 19.2 Heme atoms 14.3 16.0 CO atoms 13.9 14.4 Water atoms 35.2 37.2 Core interface water molecules 24.7 (14) 34.5 (18)

In the deoxy F97L structure, the disposition of the side chain of residue 97 was very similar to that in the deoxy wild-type structure, where it is packed in the heme pocket (Figs. 4 (top) and 5). Unlike the phenylalanine side chain, which is packed tightly between the heme group and the proximal histidine (residue 101), the smaller Leu⁹⁷ side chain appeared to be accommodated without strain. This was reflected in differences in heme conformation, position of the heme iron, and subtle changes in F-helix geometry compared with deoxy wild-type HbI, as discussed further below.

Upon ligand (CO or O₂) binding to wild-type HbI, the Phe⁹⁷ side chain is displaced into the subunit interface, since its tight pocket is reduced when the iron moves into the heme plane (2). In contrast, the leucine side chain remained in the heme pocket in F97L upon ligand binding (Fig. 4, bottom).

In deoxy wild-type HbI, there is a sharp bend in the F-helix (residues 87-103) (Figs. 4 (top) and 5) that was postulated to result from wedging of the Phe⁹⁷ side chain in the heme pocket (2). Unexpectedly, this bend, although attenuated, persisted in F97L, where the smaller leucine side chain was readily accommodated in the heme pocket. Subtle differences in F-helix geometry were present including a movement of the backbone toward the heme group, most markedly seen involving the C- of residue Leu⁹⁷ (0.5 Å), but also involving Lys⁹⁶ (0.3 Å) and Glu⁹⁵ (0.3 Å), as shown in Fig. 5. This change presumably resulted from loss of the tight packing of Phe⁹⁷ between the F-helix backbone and the heme group. Additionally, the distances between the carbonyl oxygens of residues 96 and 97 and the amide nitrogens of residues 100 and 101, respectively (4.5 and 3.9 Å), were slightly less than the corresponding distances in the wild-type deoxy structure. These effects reflect a subtle straightening of the sharp bend observed in the F-helix of deoxy wild-type HbI. Also, the hydrogen bond between the main chain carbonyl oxygen of residue 97 and the N of the proximal histidine, which is hypothesized to play a role in determining oxygen affinity of the heme iron (2, 19, 20), was optimized in F97L (Tables V and VI).

Table V. Comparison of heme region geometry between deoxy HbI and deoxy F97L Deoxy HbI Deoxy F97L Subunit I Subunit II Subunit I Subunit II Iron to heme plane distance (Å) 0.50 0.52 0.40 0.46 Iron to Npa plane distance (Å) 0.36 0.38 0.22 0.26 Mean iron to Np bond distance (Å) 2.04 2.04 1.98 1.99 Iron-N2 (His101) distance (Å) 2.19 2.18 2.24 2.19 H101 N1-O Phe97 distance (Å) 3.09 3.08 2.85 2.85 a Np denotes the 4 pyrrole nitrogen atoms of the heme group.

Table VI. Comparison of heme region geometry between HbI-CO and F97L-CO HbI HbI-CO F97L-CO Subunit I Subunit II Subunit I Subunit II Iron to heme plane distance (Å) 0.04 0.04 0.09 0.03 Iron to Npa plane distance (Å) 0.01 0.00 0.03 0.02 Mean iron to Np bond distance (Å) 2.01 2.03 1.98 1.97 Iron-N2 (His101) distance (Å) 2.09 2.14 2.18 2.23 H101 N1-O Phe97 distance (Å) 2.87 2.87 2.82 2.86 CO ligand:iron-C distance (Å) 1.83 1.83 1.87 1.88 Iron-C-O angle (°) 155.2 162.8 173.7 174.3 a Np denotes the 4 pyrrole nitrogen atoms of the heme group.

Coupled with ligand binding, a movement of the heme group deeper into each subunit and away from the dimer interface is observed with wild-type and F97L HbI. This movement was previously suggested to result from extrusion of the tightly packed Phe⁹⁷ side chain from the heme pocket (2). Our structural results argue against such a determinative role for Phe⁹⁷. Thus, other interactions between the heme group and protein must be sufficient for this heme movement. In this regard, interactions of protein side chains with the domed deoxy heme as well as intra- and intersubunit interactions with the heme propionates likely contribute to this effect.

There exists an elaborate network of hydrogen-bonded water molecules in the subunit interface of wild-type HbI. These water molecules have been proposed to play an important role in the cooperative mechanism by providing stabilizing hydrogen-bond interactions to the alternate allosteric states of HbI (2, 18).

In the deoxy wild-type structure, of the 219 total water molecules, 17 are located at the central core of the dimer interface. These water molecules are symmetrically distributed about a noncrystallographic dyad (2-fold), even though this restraint was not imposed in model building or refinement, and are very well ordered, with an average B-factor of 19.6 Å². Of these, the lowest B-factors are seen for a cluster of 4 water molecules that are located at the bottom of the interfacial crevice and that are hydrogen-bonded to E-helix residues that remain largely unaltered by ligand binding.

In comparison, 174 water molecules were independently built into the deoxy F97L structure during the course of model refinement. As shown in Fig. 6 (top), the interfacial water architecture for this structure was found to be virtually identical to that of wild type, with 17 water molecules that were structurally equivalent (<0.60 Å) to the waters described above. These water molecules, however, were significantly less well ordered than in wild type, with an average B-factor of 29.5 Å². The deepest cluster of 4 water molecules maintained the lowest B-factors, whereas those closer to the bulk solvent were increasingly disordered. The stability of water molecules correlated with the extent to which they were hydrogen-bonded to protein groups as opposed to other water molecules. Water molecules whose sole hydrogen-bonding partners were other water molecules exhibited the highest B-factors. For instance, the 5 water molecules that hydrogen-bond only with other water molecules showed an average B-factor of 32.7 Å² in F97L compared with 19.1 Å² in HbI. The changes observed for the deoxy F97L structure with regard to water molecules therefore suggest that the subtle changes in the F-helix conformation resulted in a marked loosening of the interfacial water network.

In contrast to the deoxy form, striking differences in the interfacial water architecture were observed between the CO structures of F97L and wild-type HbI. Upon ligand binding to HbI, the displacement of each Phe⁹⁷ side chain into the subunit interface results in the direct expulsion of 3 water molecules (Fig. 6, middle). The liganded wild-type interface thus has fewer and less well ordered water molecules than the deoxy interface (Tables III and IV). As shown in Fig. 6 (bottom), loss of the Phe⁹⁷ side chain resulted in several discrete changes in the interfacial water distribution for the F97L-CO structure. First, 4 of the 6 water molecules that are normally lost due to displacement of the Phe⁹⁷ side chain were now preserved, since the Leu⁹⁷ side chain remained in the heme pocket. The other water molecules, including those participating in additional aspects of ligand-linked transitions, however, resembled wild-type HbI-CO water molecules. Second, the interfacial water molecules in the F97L-CO structure were significantly more disordered, with average B-factors of 34.5 Å² compared with 24.7 Å² for wild-type HbI-CO. The interfacial water architecture for F97L-CO thus appeared to resemble a hybrid version of the deoxy and liganded HbI water distribution. Another difference is that the distribution of water molecules in the mutant CO interface was asymmetric, in contrast to the wild-type structures. This asymmetry was seen to persist in simulated annealing omit maps.

DISCUSSION

HbI represents a simple allosteric system that is useful for the investigation of intersubunit communication. The experiments described here were undertaken to test the hypothesis that Phe⁹⁷ is critical for cooperativity. Based on a comparison of high resolution crystal structures of liganded and unliganded HbI, Phe⁹⁷ was suggested to play a key role in both regulation of oxygen affinity and transfer of information between subunits (2). By virtue of its tight packing in the heme pocket of deoxy HbI, Phe⁹⁷ was thought to lower oxygen affinity by (i) acting as a wedge, restricting movement of the heme iron into the heme plane and (ii) lengthening and weakening a hydrogen bond between the N of proximal histidine and the main chain carbonyl oxygen of Phe⁹⁷. Furthermore, it was suggested that its movement into the interface upon ligand binding would enhance the oxygen affinity of the second subunit through two pathways: (i) by a direct effect of its side chain in the interface and (ii) by triggering a movement of the heme group deeper into its subunit that would alter the hydrogen bonding of the propionates at the interface. As a result, one would predict that mutation of Phe⁹⁷ to Leu would result in a substantial increase in oxygen affinity and a loss of cooperativity.

F97L exhibited an 8-fold increase in oxygen affinity over that of wild type. From the kinetic results, it appeared that the increased oxygen affinity results primarily from an increased affinity of the deoxy state relative to the deoxy state of wild-type HbI. This was reflected in the large decrease in oxygen dissociation rates for F97L, particularly at low oxygen saturation, and a modest increase in ligand combination rates, especially for CO (Fig. 1, Table I). An altered deoxy state was also suggested by spectral results and was clearly shown in the crystallographic analysis. The crystal structures demonstrated that the smaller leucine side chain is accommodated without strain in the heme pocket of both deoxy and F97L-CO. This permitted the heme iron unrestricted movement into the heme plane (Fig. 5, Tables V and VI). Additionally, the hydrogen bond between the main chain carbonyl oxygen of residue 97 and N of the proximal histidine, which has been proposed to increase oxygen affinity, was nearly optimal in both the deoxy and CO structures. Thus, our analysis of the F97L mutant strongly supports the proposal that Phe⁹⁷ plays the central role in determining oxygen affinity.

Cooperativity in F97L was largely decreased due to a loss of the ligand-linked movement of the side chain of residue 97 into the subunit interface. Kinetic measurements showed no distinct change in the oxygen dissociation rate for F97L as a function of varying oxygen saturation, in sharp contrast to wild-type HbI. These observations are consistent with diminished cooperativity resulting from the loss of displacement of Phe⁹⁷ into the interface. Perhaps surprisingly, cooperativity was not fully lost in this mutant. This appears to be associated with our structural finding that the ligand-linked movement of the heme groups continued to occur in this mutant, refuting the earlier hypothesis that this movement was linked to the disposition of the Phe⁹⁷ side chain. Thus, other interactions with the heme are responsible for this movement, which results in communication via an alternative pathway that presumably supplements the primary Phe⁹⁷-based pathway in wild-type HbI.

Royer et al. (18) have recently compared the ligand binding parameters of wild-type HbI and the mutant T72V, which exhibits very high oxygen affinity (p50 = 0.2 torr) in combination with high cooperativity (n = 1.7). The functional behavior of these proteins was analyzed by defining the properties of ideal T and R states of HbI within the context of a two-state model (21). These model properties can never be fully expressed in the experimentally accessible ligand binding behavior of the protein but can be simulated to interpret the experimental data for the liganded and unliganded proteins. That analysis suggested that HbI and T72V have very similar T and R states but that large functional distinctions result from a very different balance between T and R states, as reflected in the allosteric constant L, the ratio between T and R states of the unliganded protein. A similar treatment of data for F97L that assumes no fundamental change in the affinities of the T and R states and that alters the value of L alone does not work. The results from oxygen-pulse and CO-binding experiments agree in indicating more R than T state for singly liganded F97L than predicted by such a simulation as well as greatly diminished cooperativity. The experimental results for oxygen binding data with F97L, however, may be well approximated by large changes in the value of both L and c (c is defined as the ratio between K_T and K_R for unliganded protein, where K_T = T_on/T_off and K_R = R_on/R_off). Fig. 1 shows the agreement between experimental results and calculated values based on reducing L by 25-fold and decreasing T_off by 12-fold relative to wild type. This interpretation is in agreement with our evidence from the deoxy crystal structure, spectroscopy, and kinetic data that the T state of F97L is significantly different from that of wild type.

Previous work has shown that the well ordered water molecules in the subunit interface are crucial for stabilization of the deoxy state of HbI and suggested that these water molecules act as allosteric mediators in the cooperative mechanism (18). In this regard, analysis of the water network in the interface of F97L HbI reveals that quite subtle changes in the deoxy conformation result in a destabilization of the interface water cluster. This suggests that the deoxy protein conformation of wild-type HbI is finely tuned to maintain its very well ordered interface water cluster.