α-Hemoglobin-stabilizing Protein (AHSP) Perturbs the Proximal Heme Pocket of Oxy-α-hemoglobin and Weakens the Iron-Oxygen Bond

Background: α-Hemoglobin stabilizing protein (AHSP) is a hemoglobin chaperone.

Results: AHSP causes a subtle perturbation of the proximal heme pocket of O₂-α-hemoglobin, lengthening the Fe-O₂ bond and enhancing O₂ dissociation.

Conclusion: Pro-30 in wild-type AHSP promotes αHb autooxidation by introducing strain into the proximal heme pocket.

Significance: αHb·AHSP complexes are intermediates in Hb assembly and achieve αHb detoxification.

Abstract

α-Hemoglobin (αHb)-stabilizing protein (AHSP) is a molecular chaperone that assists hemoglobin assembly. AHSP induces changes in αHb heme coordination, but how these changes are facilitated by interactions at the αHb·AHSP interface is not well understood. To address this question we have used NMR, x-ray absorption spectroscopy, and ligand binding measurements to probe αHb conformational changes induced by AHSP binding. NMR chemical shift analyses of free CO-αHb and CO-αHb·AHSP indicated that the seven helical elements of the native αHb structure are retained and that the heme Fe(II) remains coordinated to the proximal His-87 side chain. However, chemical shift differences revealed alterations of the F, G, and H helices and the heme pocket of CO-αHb bound to AHSP. Comparisons of iron-ligand geometry using extended x-ray absorption fine structure spectroscopy showed that AHSP binding induces a small 0.03 Å lengthening of the Fe-O₂ bond, explaining previous reports that AHSP decreases αHb O₂ affinity roughly 4-fold and promotes autooxidation due primarily to a 3–4-fold increase in the rate of O₂ dissociation. Pro-30 mutations diminished NMR chemical shift changes in the proximal heme pocket, restored normal O₂ dissociation rate and equilibrium constants, and reduced O₂-αHb autooxidation rates. Thus, the contacts mediated by Pro-30 in wild-type AHSP promote αHb autooxidation by introducing strain into the proximal heme pocket. As a chaperone, AHSP facilitates rapid assembly of αHb into Hb when βHb is abundant but diverts αHb to a redox resistant holding state when βHb is limiting.

Introduction

Adult human hemoglobin (Hb)² is a tetramer of two αHb and two βHb subunits. The αHb and βHb chains share a common fold, and each binds a single, iron-containing protoporphyrin IX (heme) molecule in a deep pocket, protected from solvent. The central iron atom in each heme is the site of physiological dioxygen (O₂) binding. α-Hemoglobin stabilizing protein (AHSP) is a molecular chaperone that binds free αHb in a heterodimeric complex (see Fig. 1A) (1–3) and is essential for normal erythropoiesis (4–12).

FIGURE 1.

Structure of the bis-histidyl αHb·AHSP complex. A, shown is a schematic of the αHb·AHSP complex (PDB 3IA3). Loop 1 and Pro-30 (spheres) of AHSP are positioned at the interface with αHb. B, shown are the E-F helices of αHb in the native conformation, as seen in the deoxy Hb complex (PDB 2DN2) (53), with His-87 (His F8) coordinating the heme iron. His-58 (His E7) sits away from the iron site and regulates the binding of diatomic ligands. C, in a complex formed from AHSP and met-αHb (PDB 3IA3), His-87 and His-58 simultaneously coordinate the Fe(III) heme iron (bis-histidyl heme coordination).

One role of AHSP is to protect erythroid cells from oxidative damage arising from the action of the heme iron in free αHb (6). In aqueous solutions containing air-equilibrated buffer, the Fe(II) atoms of heme groups spontaneously oxidize to the ferric (Fe(III)) state, which results in the production of superoxide anions. This process is termed autooxidation. Autooxidation is inhibited by the physicochemical properties of the heme pocket that allow reversible oxygen binding. However, spontaneous oxidation is not blocked completely, and the superoxide anions generated undergo rapid enzymatic and non-enzymatic disproportionation to H₂O₂ and O₂. H₂O₂ is a key mediator of oxidative stress that can react with either reduced or oxidized globins to produce highly reactive ferryl or oxo-ferryl (Fe(IV)) heme groups (13, 14), and subsequent reactions lead to destruction of Hb as well as production of potent pro-oxidants that damage membranes and other cellular components (15–17). Additionally, autooxidation to the ferric state renders Hb incapable of O₂ transport and greatly enhances the rates of hemin loss and globin denaturation.

Compared with mature Hb, isolated αHb and βHb chains produce higher levels of reactive oxygen species (18–20), presumably linked to a loss of structural integrity of the heme pocket. Oxidative damage occurs in erythroid cells of β-thalassaemic mice that contain elevated levels of unpaired αHb (21–24). Kong et al. (6) showed similar increases in reactive oxygen species in AHSP^−/− mice, suggesting that AHSP has a role in limiting reactive oxygen species production during normal erythropoiesis. In a purified system AHSP inhibits the reaction of Fe(III) αHb with H₂O₂ (2, 20), and recently Mollan et al. (20) demonstrated that AHSP binding prevents the production of Fe(IV)O heme and associated protein radicals when Fe(III) αHb is exposed to H₂O₂.

The low reactivity of Fe(III) αHb in the presence of AHSP correlates with a change in αHb heme pocket structure. In native Hb, the αHb heme group is coordinated through a single His-87 side chain, termed the proximal histidine (see Fig. 1B), allowing exogenous ligands in the distal heme pocket to interact with iron. In the ferric αHb·AHSP complex, Fe(III) becomes coordinated by both the proximal and distal histidine side chains (bis-histidyl heme coordination; see Fig. 1C) (1, 2, 11, 25–27). Coordination of low-spin Fe(III) with six strong ligands blocks the binding of exogenous ligands and associated electron transfer reactions involving oxygen species. The switch from native to bis-histidyl heme coordination is accompanied by a change to a non-native αHb tertiary structure (r.m.s deviation of 3.3 Å over 135 C^α atoms), which dissociates from AHSP ∼100-fold slower than Fe(II) αHb (20, 25, 27).

As well as inducing changes in the structure of Fe(III) αHb, AHSP binding causes a >10-fold increase in the autooxidation rate of Fe(II) αHb (11, 20, 25, 28, 29), also resulting in formation of bis-histidyl Fe(III) αHb. To understand how promoting αHb autooxidation is compatible with reduced reactive oxygen species production in erythroid cell, it is important to appreciate that repeated movement between different iron oxidation states can occur for a single heme group, and this redox cycling increases the probability of irreversible changes in the globin and other cellular components. In contrast, autooxidation of αHb to the bis-histidyl αHb structure dramatically lowers the Fe(III)/Fe(II) redox potential (20) and inhibits interactions with exogenous ligands, thus inhibiting redox cycling. However, the structural mechanism by which AHSP accelerates αHb autooxidation and, more generally, how AHSP promotes the protective bis-histidyl structure is not well understood.

Previous work from our group has suggested that AHSP induces strain into the O₂-αHb structure that is relieved upon autooxidation by reorganization of the αHb protein fold (25). Mutation of Pro-30 to Ala (AHSP^P30A) in loop 1 of AHSP (Fig. 1A) leads to an ∼4-fold reduction in αHb·AHSP autooxidation rates, suggesting that Pro-30 contributes to the strained conformation. However, how contacts at the αHb·AHSP interface are transmitted to the αHb heme pocket and the nature of the strained O₂-αHb structure are unknown. Difficulties in characterizing the O₂-αHb·AHSP complex arise from its propensity to spontaneously convert to a bis-histidyl structure. A previous x-ray crystallographic study suggested that the strained O₂-αHb structure involved large-scale changes in the globin polypeptide structure and heme coordination geometry (1).

To explore these issues, we have used NMR and EXAFS to probe the complexes of CO- and O₂-liganded αHb with AHSP in solution. The results show that the tertiary structure of CO/O₂-αHb in these complexes closely resembles that of CO/O₂-αHb in the native HbA tetramer and, therefore, that our previous model of O₂-αHb·AHSP (1) is unlikely to represent the initial AHSP binding event in solution. We show that mutation of Pro-30 in AHSP leads to NMR chemical shift perturbations in the F-helical region of CO-αHb, and the concomitant changes in O₂ and CO binding reactivity help identify how this strain introduced into the proximal portion of the heme pocket promotes autooxidation of O₂-αHb.

EXPERIMENTAL PROCEDURES

Protein Production

The structured region, residues 1–90 of human AHSP, and mutants thereof were expressed and purified as described previously (25). CO-αHb was purified from human blood (3). O₂-αHb was generated by gently blowing oxygen across the surface of CO-αHb under a concentrated 10-watt halogen light source. The procedure was carried out on ice, and the formation of the O₂-αHb was monitored by UV-visible spectroscopy (Shimadzu UV-1601). Isotopically labeled αHb for NMR studies was co-expressed with βHb and methionine aminopeptidase in Escherichia coli from the previously described plasmid pHE7 (30). Expression was carried out in the E. coli strain JM109. One liter of overnight culture in DM-4 minimal media as described by Looker et al. (31) containing d-glucose (5 g liter⁻¹), NH₄Cl (0.7 g liter⁻¹), thiamine (30 mg liter⁻¹), ampicillin (100 mg liter⁻¹), and trace metal solution (32) was inoculated into a final volume of 3.5 liters of DM-4 medium containing thiamine (30 mg liter⁻¹), ampicillin (100 mg liter⁻¹), trace metal solution, yeast extract (50 μl of 10% w/v solution), and antifoam (Sigma) in a 5-liter fermentor vessel (New Brunswick BioFlo). Growth was maintained at 37 °C, pH 6.8, with aeration and monitored in real time based on dissolved O₂ concentration. The culture exhausted the nitrogen supply at an absorbance of ∼1 (600 nm) and was provided with 2 g of ¹⁵NH₄Cl. The culture exhausted the glucose supply at an absorbance of ∼3, and 1.5 g of d-[U(99%)-¹³C]glucose was provided, and the temperature was shifted to 32 °C. The addition of a further 2 g of ¹⁵NH₄Cl, 1.5 g of d-[U(99%)-¹³C]glucose occurred at an absorbance of ∼3.5, and expression was induced with 0.2 mm isopropyl β-d-1-thiogalactopyranoside and 15 μm hemin. Cells were collected by centrifugation after 5 h, and the resulting cell pellets were washed once in 10 mm Tris-HCl, 0.5 mm EDTA, pH 8.0. Cell pellets were resuspended in the same buffer and saturated with CO before lysis by sonication (Ultrasonic Processor 500 W, Sonics & Materials Inc.). After clarification by centrifugation, recombinant Hb was purified as previously described (33). The Hb chains were split by the addition of a 2-fold molar excess (per cysteine) of p-hydroxymercuribenzoate (pMB) and purified as previously described (3).

Size Exclusion Chromatography (SEC) and In-line Light Scattering

SEC chromatography was carried out on a 24-ml Superose 12 column (GE Healthcare) in 0.1 m sodium phosphate, pH 7.0 (21 °C). Protein concentrations were determined based on UV-visible absorbance. The concentrations of AHSP and mutants thereof were obtained after unfolding in 6 m guanidine-HCl using molar extinction coefficients calculated based on the amino acid compositions (34). CO/O₂-αHb concentrations were determined based on absorption peaks arising from π → π* transitions in the porphyrin ring using molar extinction coefficients for the corresponding CO/O₂ forms of Hb (35). The weight-average molecular weight (M_w) of particles in solution was determined directly from measurements of light scattering and protein concentration (36). In-line light scattering measurements were made at 41.5°, 90.0°, and 138.5° with respect to the incident beam using a mini-DAWN with 690 nm laser (Wyatt Technology Corp., Santa Barbara CA), calibrated against toluene. Protein concentrations were determined based on the refractive index measured by an Optilab differential refractometer (Wyatt Technology Corp.) calibrated against NaCl standards. Molecular weight calculations were performed for each 50-μl fraction across the elution peaks using the ASTRA software (Wyatt Technology Corp.) with the assumption of a standard value for the specific refractive index increment with respect to sample concentration (dn/dC) of 0.19 ml g⁻¹. The presence of the heme and pMB groups upon the value of dn/dC was neglected in this case.

NMR Spectroscopy

All NMR spectroscopic measurements were conducted using a Bruker Avance II 600 MHz spectrometer equipped with a triple-resonance TCI cryoprobe (Bruker, Karlsruhe, Germany), then processed using Topspin 2.1 (Bruker), and analyzed using SPARKY 3.1 (37). Samples were formulated as described, with the addition of 20 μm 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) and 5% D₂O. Backbone assignments were made using HNCA, HNCACB, CBCA(CO)NH, and ¹⁵N-edited NOESY according to standard assignment strategies. His side chain resonances were assigned based on ¹⁵N-HSQC, ¹⁵N-NOESY-HSQC, and two-dimensional TOCSY (two-dimensional total correlation spectroscopy)/NOESY spectra. Tautomer/charge state was inferred for His side chains from cross-peak intensities in the ¹⁵N-HSQC (38). Spectra were recorded at sample temperatures of 15 and 25 °C for αHb and αHb·AHSP complexes, respectively, to optimize spectral quality.

EXAFS

Samples for x-ray absorption spectroscopy were prepared in 20 mm sodium phosphate buffer, pH 7.0 (21 °C), with 30% sucrose (0.88 m) as the cryoprotectant. Sucrose was the preferred cryoprotectant, as it caused minimal changes in the UV-visible spectrum. The O₂-αHb·AHSP complex was formed by mixing O₂-αHb with a 1.1 molar excess of AHSP. Final protein concentrations were 6 and 2.5 mm for O₂-αHb and O₂-αHb·AHSP samples, respectively. The solutions were injected into 140-μl Lucite XAS cells (23 × 2 × 3 mm) with 63.5-μm Kapton tape windows. This operation was conducted in a nitrogen-filled glove bag. Immediately after their preparations, the samples were frozen in a liquid-nitrogen/n-hexane slurry to prevent autooxidation. X-ray absorption spectra were recorded at the Stanford Synchrotron Radiation Lightsource (SSRL) on beamline 9-3 (3 GeV, 100–50 mA) using a Si(220) double-crystal monochromator. Beamline 9-3 had an upstream collimating mirror and a downstream sagitally focusing mirror; both were rhodium-coated and also provided harmonic rejection. Data were collected as fluorescence spectra using a 30-element germanium array detector (39). For each sample, 10–21 scans were recorded from 6785 to 8257 eV. The energy was calibrated using an iron foil standard in which the first inflection point was assigned to an energy of 7111.2 eV (40). Data scans were averaged, background-subtracted, and normalized as described previously (41).

EXAFS Data Analysis

The model-fitting calculations were performed by means of the program XFIT (42). The values of the parameters that were varied to optimize the agreement between the observed and calculated EXAFS were the coordinates x, y, and z of all atoms in the model in relation to an arbitrary set of Cartesian axes, the Debye-Waller factors, σ², a scale factor, S₀², and E₀. The value of S₀² was initially restrained to be 0.9 with an estimated standard deviation (e.s.d.) 0.1 for O₂-αHb and 0.92 (e.s.d. 0.02) for O₂-αHb·AHSP (i.e. to typical values for low-spin O₂-myoglobin and O₂-leghemoglobin). Both the observed and calculated EXAFS were Fourier-filtered (42). The k windows used for the EXAFS analyses are shown in the figures. The goodness-of-fit parameter R was calculated as described by Ellis and Freeman (42). The model used in the XFIT analyses is that described previously (41). The sixth coordination site was a dioxygen molecule where the terminal oxygen was arbitrarily constrained to lie in the plane perpendicular to the heme and passing through the meso-Cs. The O¹-O² distance was restrained to be 1.22 Å (e.s.d. 0.01 Å), consistent with that reported by Jameson et al. (43). The constraints and restraints used for the bond lengths and bond angles in the porphyrin ring and the imidazole ligand and for the Debye-Waller factors were as described previously (41). Constraints and restraints and multiple-scattering (MS) paths used in the fitting calculations are given in supplemental Tables.

O₂/CO Binding Measurements

Rate constants for O₂ binding to and dissociation from ferrous αHb·AHSP complexes as well as rate constants for CO binding were measured using methods developed for relatively unstable Fe(II) O₂-Hb complexes as described by Birukou et al. (45). In these measurements, stable CO-αHb complexes are injected into laser photolysis cuvettes containing mixtures of buffers equilibrated with 1 atm O₂ or 1 atm CO. The mixtures in % were O₂:CO = 100:0, 75:25, 50:50, 25:75, and 0:100. The CO-αHb samples were photolysed with a 0.5-μs, 577-nm dye laser pulse. Absorbance time courses for O₂ and CO rebinding to deoxy-αHb were measured on μs time scales at 436 nm, and then time courses for replacement of the transiently bound O₂ by CO were measured on millisecond to seconds time scales at 424 nm (see Fig. 1 in Ref. 45). This method prevents significant autooxidation during the measurements of O₂ binding because the sample remains in the Fe(II)-CO form in between laser photolysis pulses. The dependence of the pseudo first order rates for the fast and slow phases on O₂ and CO were fitted globally to obtain values for the association rate constants for O₂ and CO binding to αHb, k′_O2 and k′_CO, and the rate of O₂ dissociation, k_O2, as described in Birukou et al. (45). Our measurements were made using 100 mm potassium phosphate buffer, pH 7.0, at 20 °C. Similar kinetic measurements of O₂ and CO binding to native and recombinant wild-type Fe(II)-αHb·AHSP complexes were made by Vasseur-Godbillon et al. (28).

RESULTS

A Proline Residue Is Precisely Positioned in Loop 1 of AHSP to Promote Autooxidation of αHb

The Asp-29–Pro-30 peptide bond in loop 1 of AHSP undergoes cis-trans isomerization that is coupled with small changes in the packing of the three α-helices in AHSP (1, 46). Mutation of AHSP Pro-30 to a range of other residue types abolishes cis-trans peptidyl isomerization of AHSP and leads to a consistent 4-fold reduction in the rate of O₂-αHb autooxidation (11, 25). We previously proposed that cis-peptidyl Pro-30 makes physical contact with O₂-αHb that promotes autooxidation (25). However, it is possible that other changes in AHSP conformation related to cis-trans isomerization might explain the role of Pro-30 in αHb autooxidation. To investigate this possibility, we introduced proline mutations into AHSP^P30A at position 28 (AHSP^P30A/N28P), 29 (AHSP^P30A/D29P), 31 (AHSP^P30A/L31P), or 33 (AHSP^P30A/S33P) as shown in Fig. 2A. We argued that proline residues in these positions might still allow AHSP to undergo conformational exchange, but the proline side chains would presumably not preserve contacts with αHb that are made by the native Pro-30. The locations of the introduced proline mutations at the αHb·AHSP interface are shown in Fig. 2B. We did not mutate Val-32 because this side chain packs into the hydrophobic core between the AHSP helices.

FIGURE 2.

Conformational exchange in AHSP is dependent on the position of proline residues in loop 1. A, shown is the sequence of the first two α-helices and intervening loop 1 of wild-type human AHSP and AHSP mutants used in this study. Conserved positions relative to the wild-type sequence are marked with dashes. B, shown is the structure of the bis-histidyl αHb·AHSP interface (white, PDB 3IA3; the Pro-30 side chain is shown in space-fill) with the AHSP^P30A structure overlaid (black, PDB 1Z8U). The positions of side chains mutated to proline in AHSP^P30A are shown in space-fill. C and D, portions of the assigned ¹⁵N-HSQC spectra of AHSP (800 μm) and AHSP^P30A (900 μm). E–H, portions of the ¹⁵N-HSQC spectra of AHSP^P30A/N28P (400 μm), AHSP^P30A/D29P (550 μm), AHSP^P30A/L31P (900 μm), and AHSP^P30A/S33P (530 μm) are shown. All ¹⁵N-HSQC spectra were recorded at 15 °C in 20 mm sodium phosphate buffer, pH 6.9–7.0.

The propensity of the mutant AHSP proteins to undergo cis-trans isomerization was assessed from ¹⁵N-HSQC NMR spectra. A portion of the ¹⁵N-HSQC spectrum of wild-type AHSP is shown in Fig. 2C, with peaks arising from N-H^N groups in the same residue in cis-prolyl and trans-prolyl conformers of AHSP labeled and joined by a dashed line. In the AHSP^P30A spectrum, a single N-H^N peak was observed for each residue, consistent with a single AHSP conformer (Fig. 2D and Table 1) (1, 46). The spectrum of AHSP^P30A/N28P contained an additional set of weaker HSQC signals compared with AHSP^P30A (Fig. 2E, arrows, and Table 1) that was consistent with ∼20% of molecules adopting an alternative conformation, most likely due to isomerization of the Phe-27–Pro-28 peptide bond. The ¹⁵N-HSQC spectra of AHSP^P30A/D29P, AHSP^P30A/L31P, and AHSP^P30A/S33P (Fig. 2, F–H) were all consistent with a single protein conformation, which indicated that prolines at positions 29, 31, and 33 do not result in measurable cis/trans isomerization. Chemical shift analyses were consistent with minor conformational changes in AHSP proximal to loop 1 and indicated that none of the mutations conferred gross structural changes (not shown). These results suggest that only cis-peptide bonds before Pro-30 or Pro-28 result in a significant amount of cis/trans isomerization.

TABLE 1.

αHb-interaction parameters for AHSP mutants

Protein	Kautoox, pH 7.0, 30 °C	Residues	N-HN peaksa
	h−1
AHSP	0.59 ± 0.07	90b	121b
AHSPP30A	0.14 ± 0.01	90	80
AHSPD29A	0.53 ± 0.02	90	134b
AHSPP30A/N28P	0.13 ± 0.02	90	113
AHSPP30A/D29P	0.14 ± 0.03	90	84
AHSPP30A/L31P	0.14 ± 0.05	90	85
AHSPP30A/S33P	0.14 ± 0.03	90	81

^a Spin systems with characteristic chemical shifts for Asn, Gln, Arg, and Trp side chains were excluded.

^b The AHSP construct used comprises residues 1–90 with the C-terminal unstructured residues removed (1). Approximately 50% of residues gave rise to two sets of signals in triple resonance NMR spectra, corresponding to the cis/trans AHSP conformers. Differences in signal overlap in HSQCs results in a total of 121 resolved N-H^N peaks for WT AHSP and 134 resolved peaks for AHSP^D29A.

The effect of AHSP proline mutations on αHb autooxidation is shown in Fig. 3. Autooxidation of O₂-αHb to bis-histidyl αHb was measured by changes in the visible absorption spectrum of αHb (Fig. 3, A and B). The absorbance time courses fit to single exponential expressions and yielded first-order rate constants for autooxidation (k_autoox; Fig. 3C and Table 1). AHSP^P30A showed a 4-fold reduction in the value of the rate constant compared with that of wild-type AHSP, as previously described (11, 25). All introduced-proline mutants had the same low activity as the AHSP^P30A (Fig. 3C and Table 1). A control mutation AHSP^D29A had the same activity as wild-type AHSP. The results indicate that the positioning of Pro-30 in loop 1 is critical for function. In addition, the observation that free AHSP^P30A/N28P showed cis/trans isomers in solution, yet conferred the same slowed rate of O₂-αHb autooxidation as AHSP^P30A, suggests that the propensity of free AHSP to undergo conformational changes does not in itself contribute to O₂-αHb·AHSP autooxidation activity. We have shown previously that stabilization of the cis-peptidyl Asp-29–Pro-30 AHSP conformation does correlate with increased autooxidation activity (25). Together the above data are consistent with the hypothesis that the effects of Pro-30 are mediated through direct contact between loop 1, containing the Pro-30 residue, and αHb.

FIGURE 3.

The position of proline in AHSP loop 1 is important for function. A, conversion of O₂-αHb·AHSP to bis-histidyl αHb·AHSP was monitored by a decrease in intensity of the absorption band at 575.5 nm. B, time courses for conversion to bis-histidyl αHb in the presence of AHSP or selected AHSP mutants are shown, with fits to an exponential function. C, rate constants for O₂-αHb conversion reactions using AHSP and AHSP mutants shown in Fig. 1B are shown (average ± S.D. for 3–9 repeat assays). Mutation of Pro-30 to Ala (AHSP^P30A) reduces the rate constants by a factor of 4. The control mutation (AHSP^D29A) had no significant effect. Introducing Pro at position 28, 29, 31, and 33 failed to rescue any activity.

Free Liganded αHb Monomer Adopts Native-like Secondary Structure

We next used NMR spectroscopy to probe for structural differences between αHb in the native Hb tetramer and free αHb in solution. The aim of these studies was to probe for structural changes in αHb that could explain the more rapid autooxidation observed in vitro and provide a suitable reference point for subsequent NMR studies of αHb·AHSP complexes. We found that sequential backbone assignments could not be obtained for residues 104–126 (corresponding to helices G and H), as also reported previously by Martineau and Craescu (47). The absence of NMR signals corresponding to these residues suggests that they might be in intermediate chemical exchange, perhaps as a result of αHb self-association.

In line with that possibility, we noted a marked decrease in SEC peak elution time with increasing αHb load concentration. This behavior is typical of protein self-association and is consistent with previous studies of αHb (48, 49). We reasoned that treatment of αHb with pMB, which covalently modifies Cys-104 of αHb and inhibits Hb chain homo (48) and hetero interactions (50), would allow more complete NMR data to be obtained. It has previously been shown that pMB treatment has no effect on the O₂ and CO binding parameters of isolated αHb (51). Light scattering was used to monitor the self-association status of CO-αHb before and after pMB treatment. We found that untreated CO-αHb eluted from SEC with a light scattering data profile characteristic of reversible self-association, with a marked increase in M_w toward the center (i.e. the highest concentration) of the elution peak (Fig. 4A, each M_w measurement is indicated with a red +). In contrast, pMB treated CO-αHb (CO-αHb^PMB) eluted with a constant M_w of 16.7 ± 1.3 across the peak (Fig. 4A, black dots), indicating that CO-αHb^PMB is predominantly monomeric (formula weight of 16,053).

FIGURE 4.

Identification of the αHb homodimerization interface. A, self-association of αHb was inhibited by treatment with pMB. SEC traces are shown for CO-αHb (red) or CO-αHb^PMB after reaction with a 4-fold molar excess of pMB (black). Molecular weight from light scattering measurements is plotted across each peak (red/black symbols). B, portions of the ¹⁵N-HSQC for CO-αHb (red) and CO-αHb^PMB (black) were recorded at 15 °C in 20 mm sodium phosphate, pH 6.9. Assignments are shown for new peaks in the CO-αHb^PMB spectrum that are not present in the CO-αHb spectrum. C, shown are a ribbon (αHb) and surface (βHb) representation of the αHb/βHb dimer from CO-Hb (PDB 2DN3). Residues missing in the CO-αHb spectrum (red) or experiencing significant chemical shift difference between CO-αHb and CO-αHb^PMB (orange) map to the tight βHb-interaction face, which is, therefore, also identified as the αHb homodimerization face.

The ¹⁵N-HSQC spectrum of CO-αHb^PMB (Fig. 4B, black) contained numerous additional peaks (labeled) that were not present in the untreated CO-αHb spectrum (Fig. 4B, red) and sequential backbone assignments could now be obtained for 130/134 non-proline residues of CO-αHb^PMB. Comparison of the CO-αHb^PMB and CO-αHb backbone chemical shifts indicated that only the modified Cys-104 and immediately surrounding residues were strongly influenced (not shown), confirming that pMB modification does not substantially alter the tertiary structure of αHb, consistent with functional measures of O₂ and CO binding (51). The αHb self-association interface is clearly identified by the NMR data (Fig. 4C, red/orange) and is the same face that interacts with βHb (Fig. 4C, surface) to form the αHb₁βHb₁ heterodimer.

To obtain information about the secondary structure of monomeric CO-αHb^PMB, we used TALOS+ (52) to derive φ and ψ backbone angle predictions from NMR chemical shift data. Seven helical secondary structure elements could be clearly identified corresponding to the A-C and E-H helices (Fig. 5). The angles derived by TALOS+ were very similar to angles determined from crystal structures of native CO-Hb (PDB 2DN3) (53), with correlation coefficients of 0.82 (φ) and 0.87 (ψ) over 127 residues. One difference was that TALOS+ consistently assigned the C helix region as α-helix rather than the 3₁₀-helix that is identified by the program DSSP (54) based on crystal structure coordinates (PDB 2DN3).

FIGURE 5.

NMR chemical shift-derived ψ angles for αHb in the absence and presence of AHSP. Backbone ψ and φ angle predictions were calculated from H^N, ¹⁵N, ¹³C^α, and ¹³C^β chemical shifts using the program TALOS+ (52). Predictions for CO-αHb^PMB, CO-αHb·AHSP, and CO-αHb·AHSP^P30A are displayed as bars with a height ±1 S.D. The ψ and φ angles from the αHb subunit of a CO-HbA crystal structure (PDB 2DN3) are superimposed (filled circles) with α-helix (black bars) and 3₁₀ helix (white bars) secondary structure elements as assigned by DSSP (54).

To obtain information about the heme pocket, we assigned His side chain resonances for CO-αHb^PMB (Fig. 6 and Table 2). His-58 and His-87 side chain resonances displayed large upfield deviations from random coil chemical shifts consistent with positions above or below the plane of the heme ring. The chemical shifts were very similar to values previously obtained for CO-αHb at pH 5.3–5.7, 36 °C, (55, 56), and the αHb subunit of intact CO-Hb at pH 6.9, 29 °C (30, 57), strongly supporting the conclusion that free αHb adopts a heme pocket structure similar to that in Hb tetramers, with the heme iron bound through N^ϵ2 of His-87. In summary, preparation of monomeric αHb through treatment with pMB allowed us to obtain backbone and His side chain chemical shift data suitable for comparative NMR studies with αHb·AHSP complexes.

FIGURE 6.

Histidine region of the ¹⁵N-HSQC spectrum of CO-αHb^PMB. A, shown is a portion of ¹⁵N-HSQC spectrum of CO-αHb^PMB (0.8 mm) in 20 mm sodium phosphate, pH 7.0, at 15 °C. The heme pocket His-58 and His-87 spin systems contain highly shifted resonances (linked by solid lines). Note that axes are discontinuous to remove large empty regions of the spectrum. B, shown is the central region of the spectrum shown in A with peaks assigned to surface histidine side chains.

TABLE 2.

NMR resonances of His side chains for CO-αHb^PMB and CO-αHb·AHSP

Resonances for CO-αHb^PMB in 10 mm sodium phosphate, pH 7.0, at 288 K and CO-αHb·AHSP in 10 mm sodium phosphate, pH 7.0, at 298 K.

His residue	Nδ1	Nϵ2	Hδ2	Hϵ1	Hδ1
20	191.0	182.7	7.20	8.21
45	229.0	175.1	7.27	7.99
50	201.4	181.2	7.06	8.07
58	250.6	153.8	4.47	8.30
72	188.1	182.8	7.44	8.38
87	163.3	218.2	1.02	1.47	9.40
89	218.0	182.0	6.92	7.74
112	187.3	179.1	6.76	8.38
58 (with AHSP)	250.1	153.8	4.48	8.05
87 (with AHSP)	163.2	218.6	1.10	1.43	9.31

CO-αHb Retains Near-native Secondary Structure and Heme Ligation in Complex with AHSP

To understand how AHSP promotes autooxidation of αHb, we examined chemical shift differences between free CO-αHb^PMB and the CO-αHb·AHSP complex. Widespread changes in the ¹⁵N-HSQC of CO-αHb were observed upon binding of AHSP (Fig. 7A) despite the complex remaining in the CO-liganded form, as confirmed by UV-visible spectroscopy (not shown). A plot of backbone chemical shift differences between CO-αHb^PMB and CO-αHb·AHSP reveals that the B-C corner and the F, G, and H helices of αHb are most affected (Fig. 7B).

FIGURE 7.

AHSP binding perturbs helix F of CO-αHb^PMB. A, shown are ¹⁵N-HSQC spectra of CO-αHb and CO-αHb·AHSP recorded at 298 K in 20 mm sodium phosphate, pH 7.0. B, shown are combined, weighted, chemical shift differences (Δδ_N,HN,Cα) between CO-αHb^PMB (15 °C) and CO-αHb·AHSP (25 °C). C, shown are chemical shift differences between CO-αHb^PMB (15 °C) and CO-αHb·AHSP^P30A (25 °C). D, shown are αHb chemical shift differences between the CO-αHb·AHSP and CO-αHb·AHSP^P30A complexes (both at 25 °C). Horizontal dotted and dashed lines indicate the average Δδ and increments of 1 S.D., respectively. E, shown is a molecular model of the O₂-αHb·AHSP complex generated by substituting O₂-αHb (PDB 2DN1) for the bis-histidyl αHb subunit in the bis-histidyl αHb·AHSP crystal structure (PDB 3IA3). Residues in αHb experiencing a chemical shift difference (Δδ_N,HN,Cα) >0.2 ppm between the AHSP and AHSP^P30A complexes are colored red (residues missing assignments in one or both complexes are black).

The B-C corner, helix G, and helix H comprise the AHSP binding face of αHb (1, 2, 25), and chemical shift changes are expected here as a result of the interface forming (Fig. 7B, gray background shading). In contrast, chemical shift changes in helix F cannot be explained by proximity to the AHSP surface, which implies that they arise from an induced structural change in αHb. However, NMR signals arising from the F helix region had similar line widths to signals from other regions of the αHb subunit, and TALOS+ derived φ and ψ angles indicated that the F helix region remains well ordered with similar secondary structure in the absence and presence of AHSP (Fig. 5).

His-87 side chain resonances remained strongly shifted in the complex with AHSP (Fig. 8A, red), with similar chemical shift values to the free CO-αHb^PMB form (Fig. 8A, black). This provides strong evidence that His-87 remains the heme-ligating residue in CO-αHb·AHSP. In summary, AHSP binding to CO-αHb results in small but significant conformational changes in helix F even though the α-helical character and heme binding function are retained. As such, these findings contrast with a previous x-ray crystallographic study in which we reported that O₂-αHb·AHSP^P30A has a highly unusual heme site geometry wherein the F helix is disordered, and heme is coordinated by His-58 in helix E (1). It is possible that the unusual crystal structure is characteristic of the O₂- and not the CO-αHb·AHSP complex but, as described below, that now seems unlikely at least for the O₂-αHb·AHSP complex in solution.

FIGURE 8.

His-87 is the heme ligand in O₂-αHb·AHSP. Overlay of portion of the ¹⁵N-HSQCs of CO-αHb^PMB (20 mm sodium phosphate, pH 7.0, 15 °C; black) and CO-αHb·AHSP (20 mm sodium phosphate, pH 7.0, 25 °C; red) show that the His-58 and His-87 side chains experience only minor chemical shift perturbations upon AHSP binding.

AHSP Pro-30 Perturbs the Proximal Heme Pocket of αHb

Because mutation of AHSP Pro-30 to Ala or other residue types leads to a 4-fold decrease in the rate of O₂-αHb·AHSP autooxidation, we reasoned that CO-αHb·AHSP and CO-αHb·AHSP^P30A complexes would display structural differences that give insight into the mechanism underlying AHSP-mediated αHb autooxidation. We, therefore, obtained backbone resonance assignments for CO-αHb in complex with AHSP^P30A using standard triple resonance methods. Compared with wild-type AHSP, the AHSP^P30A mutant induced more limited perturbations of CO-αHb (Fig. 7C). A direct comparison of CO-αHb chemical shifts in the CO-αHb·AHSP and CO-αHb·AHSP^P30A complexes shows that the wild-type AHSP causes significantly larger chemical shift perturbations to the N-terminal fragment of αHb helix G and the whole of helix F (Fig. 7D), which suggests that AHSP Pro-30 makes contacts that are transmitted through the αHb structure and lead to perturbations of helix F.

Interestingly, we could not obtain backbone assignments for residues 135–138 of CO-αHb in complex with wild-type AHSP, and signals for these residues appeared to be absent from the NMR spectra. In NMR spectra of CO-αHb·AHSP^P30A, residues 135–138 were assigned and predicted as the regular α-helix by TALOS+ (Fig. 5). Residues 135–138 of native αHb are in a region of helix H that contacts helix F (Fig. 7E), which suggests that loss of NMR signals could reflect a destabilization of helix H in αHb·AHSP related to changes in structure of the F helix. Overall the results here indicate that AHSP Pro-30 makes contacts that perturb helices F, G, and H of αHb, which together constitute the proximal heme pocket of αHb (Fig. 7E).

Comparison of the Iron-Ligand Structures in O₂-αHb and O₂-αHb·AHSP

NMR spectroscopic data indicated that His-87 is retained as the iron binding His residue in CO-αHb·AHSP, but these data do not provide detailed information about the exact geometries of the axial iron ligands. In addition, we could not determine NMR structures for the O₂-αHb·AHSP complexes due to autooxidation during NMR data collection. To probe the electronic structure and coordination environment around the heme iron more directly, we acquired x-ray absorption spectra at the iron K-edge for the physiologically more relevant O₂-αHb and O₂-αHb·AHSP complexes. Spectra were recorded at 10 K to minimize photo-damage and increase the importance of multiple-scattering contributions to the EXAFS (58, 59). Samples were prepared in buffer containing 0.88 m sucrose, which allowed the formation of a homogeneous glass upon freezing (59). This procedure permitted measurements on the unstable O₂-αHb·AHSP complex to be performed over extended time periods without conversion to the bis-histidyl ferric αHb form. The x-ray absorption near edge structure (XANES) spectra of O₂-αHb and O₂-αHb·AHSP (not shown) show no irradiation damage, and the similarity to spectra of the low-spin Fe(III) proteins metleghemoglobin, oxymyoglobin, and oxyleghemoglobin (60) suggest that all have a low-spin Fe(III) center, which is consistent with an Fe(III)⁺-O₂^˙̄ species (61). This classification is also supported by the diagnostic oxidation- and spin-state resonance Raman marker bands (1, 62). The slightly lower edge energy of O₂-αHb·AHSP compared with free O₂-αHb suggests a lengthening of the Fe-O₂ bond in the AHSP complex.

EXAFS data obtained from O₂-αHb and O₂-αHb·AHSP were fitted to models for the iron sites using restrained and constrained multiple-scattering models (58, 63–65). The observed and calculated EXAFS and the corresponding Fourier transforms are shown for O₂-αHb and O₂-αHb·AHSP in Fig. 9, A–D, and the derived molecular structures for O₂-αHb and O₂-αHb·AHSP are shown in Fig. 9, E and F, respectively. The bond length/angle and other fitted parameters obtained are listed in Table 3.

FIGURE 9.

EXAFS analyses of the heme sites of O₂-αHb and O₂-αHb·AHSP. Shown are EXAFS (A) and Fourier-transform amplitude (B) of EXAFS of O₂-αHb. Shown are EXAFS (C) and Fourier-transform amplitude (D) of EXAFS of O₂-αHb·AHSP. Shown are the observed data (solid line), data calculated from the best fit to the model (dashed line), residual data (dotted line), and window used in the Fourier filter (dot-dash). Molecular structures and Fe-ligand bond lengths and Fe-O–O bond angle for O₂-αHb (E) and O₂-αHb·AHSP (F) were obtained from EXAFS data.

TABLE 3.

A comparison of Fe-ligand geometry from EXAFS and X-ray crystallographic analyses

Species	Method	Fe-Ligand distances				Fe-O-O angle (°)	Debye-Waller factors, σ2				Other fit parameters
		Np	Nϵ	O1	O1-O2		Np	Nϵ	O1	O2	E0	S02	R
		Å					Å				eV		%
O2-αHb	Crystala	2.02 (8)	2.07 (8)	1.82 (8)	1.23 (8)	124
	MSb	2.00 (2)	2.04 (2)	1.83 (2)	1.21 (2)	130 (3)	0.002	0.001	0.003	0.004	7125.1 (2)	0.92 (2)	15.9
O2-αHb·AHSP	Crystalc	2.0 (5)	2.1 (5)	2.7 (5)	1.3 (5)	143
	MSd	2.01 (2)	2.08 (2)	1.86 (2)	1.22 (2)	135 (3)	0.001	0.002	0.002	0.008	7125.4 (3)	0.915 (3)e	16.8
	MSf	2.03f	2.12f	2.74f	1.25f	143f	0.002	0.059	0.079	0.080	7131.2	0.932	32.5

^a PDB 2DN1 data at 1.25 Å resolution. Bond length errors are Cruickshank diffraction-data precision indicator (DPI) as calculated by REFMAC multiplied by 2^1/2 to convert from average coordinate error to average bond length error.

^b This work; 12-scan average. Error estimates in parentheses are calculated as described under “Comparison of the Iron-Ligand Structures.”

^c PDB 1Y01 data at 2.8 Å resolution. Bond length errors are 2^1/2 × DPI based on R_free factor.

^d This work; 21-scan average. Error estimates in parentheses are calculated as described under “Comparison of the Iron-Ligand Structures.”

^e Restraint on S₀² was tightened to be 0.92(2) Ų; i.e., similar to O₂-αHb.

^f Iron−ligand bond lengths and Fe-O-O bond angle were constrained to be those in the O₂-αHb·AHSP^A30 structure (PDB 1Y01).

The errors in the iron-ligand bond lengths determined by EXAFS were estimated as the r.m.s. of contributions from both random and systematic errors. The random (statistical) errors due to noise in the data were estimated by Monte-Carlo calculations (42), and the systematic errors were assigned a conservative consensus value, 0.02 Å (44). For O₂-αHb, the calculated statistical errors were 0.002 Å for Fe-N^p, 0.008 Å for Fe-N^ϵ, and 0.005 Å for Fe-O₂. For O₂-αHb·AHSP, Monte-Carlo calculations yielded r.m.s. errors of 0.003 Å for Fe-N^p, 0.012 Å for Fe-N^ϵ, and 0.006 Å for Fe-O₂. These statistical errors were combined with the maximum systematic error to obtain the estimated maximum r.m.s. errors in the reported bond lengths (Table 3). Monte-Carlo calculations yielded an r.m.s. error of 1.7° and 2.3° for the Fe-O-O bond angle in O₂-αHb and O₂-αHb·AHSP, respectively. These errors were combined with the respective r.m.s. errors of 2.1° and 3.2°, resulting from varying the O-O distance by ±0.03 Å (Table 4) to give respective estimated maximum r.m.s. errors of 2.7° and 3.9° for O₂-αHb and O₂-αHb·AHSP. Therefore, the error in this bond angle estimated by two independent methods is ∼3–4°. The resulting Fe-O-O angles determined from restraining the starting angle between 115° and 175° and allowing the fit to go to convergence then removing the restraint were 130° (for 6 minima and 143° for 1 minimum; the latter was a local minimum with a higher R value and a low S_o² value for a heme protein) in O₂-αHb (Table 5) and 135° (for 6 minima and 101° for 1 minimum) in O₂-αHb·AHSP (Table 6). The fit obtained for the Fe-O-O bond angle of 101°, when the angle was initially restrained to be 115° and then this restraint was removed and the fit allowed to go to convergence, is likely to be a false minimum. This is because it is not chemically reasonable and is not consistent with the similarity in the shape of the XANES of O₂-αHb and O₂-αHb·AHSP (not shown), as XANES is sensitive to bond angles. These analyses show that the bond lengths are independent of all of these changes in starting models and constraints/restraints, and the bond angles relax to the non-constrained values shown in Table 3. These detailed analyses of parameter space provide confidence that the EXAFS fits are robust and are not trapped in false minima. Similarly, varying the tilt or rotation of the proximal imidazole did not significantly alter the fitted bond length parameters (not shown).

TABLE 4.

Effect of varying the O-O bond distance restraint on the Fe-O-O angle in human O₂-αHb and O₂-αHb·AHSP

The quantities enclosed by the brackets ({}) are the values of σ_res used by XFIT, which is analogous to e.s.d.

Species	Restraint on O-O bond {σres}	Fe-Ligand distances				Fe-O-O angle	Debye-Waller Factors, σ2			S02	R
		Np	Nϵ	O1	O2		Np	Nϵ	O1
	Å	Å				°		Å2			%
O2-αHb	1.19 {0.01}	2.00	2.04	1.83	1.18	131.4	0.001	0.001	0.003	0.920	14.8
	1.25 {0.01}	2.00	2.04	1.83	1.25	133.5	0.001	0.001	0.003	0.906	15.4
	1.22 {0.03}	2.00	2.04	1.83	1.21	130	0.002	0.001	0.003	0.92	15.9
O2-αHb·AHSP	1.19 {0.01}	2.00	2.08	1.86	1.19	137.6	0.001	0.002	0.002	0.916	16.8
	1.25 {0.01}	2.01	2.08	1.86	1.25	134.0	0.001	0.002	0.002	0.915	16.8
	1.22 {0.03}	2.01	2.08	1.86	1.22	135	0.001	0.002	0.002	0.915	16.8

TABLE 5.

Effect of restraining the Fe-O-O bond angle in human O₂-αHb on the MS XAFS analysis

The quantities enclosed by the brackets ({}) are the values of σ_res used by XFIT, which is analogous to e.s.d.

Restraint on Fe-O-O bond angle {σres}	Fe-Ligand distances				Fe-O-O angle (°)	Debye-Waller factors, σ2				S02	R
	Np	Nϵ	O1	O2		Np	Nϵ	O1	O2
	Å					Å					%
115° {1°}	2.00	2.05	1.83	1.22	115.0	0.001	0.001	0.003	0.080	0.893	15.9
125° {1°}	2.00	2.04	1.83	1.22	126.6	0.001	0.001	0.003	0.004	0.902	15.2
135° {1°}	2.00	2.05	1.83	1.21	134.9	0.001	0.001	0.003	0.009	0.909	15.3
145° {1°}	2.00	2.05	1.83	1.22	144.6	0.001	0.001	0.003	0.004	0.886	15.6
155° {1°}	2.00	2.05	1.83	1.22	155.0	0.001	0.001	0.003	0.080	0.895	15.8
165° {1°}	2.00	2.05	1.83	1.22	165.0	0.001	0.001	0.003	0.080	0.894	15.9
175° {1°}	2.00	2.05	1.83	1.22	175.0	0.001	0.001	0.003	0.080	0.894	15.9
No constraint	2.00	2.05	1.83	1.21	129	0.001	0.001	0.003	0.004	0.919	14.1
115°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0
125°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0
135°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0
145°b	2.00	2.06	1.84	1.22	143.4	0.001	0.001	0.003	0.004	0.884	15.6
155°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0
165°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0
175°b	2.00	2.04	1.83	1.21	129.7	0.002	0.001	0.003	0.004	0.922	15.0

^a The original fit where Fe-O-O is unrestrained is included for comparison.

^b Fits were performed in which the Fe-O-O angles restrained the starting angle between 115° and 175° and then allowed the fit to go to convergence, then removing the restraint.

TABLE 6.

Effect of restraining the Fe-O-O bond angle in human O₂-αHb·AHSP on the MS XAFS analysis

The quantities enclosed by the brackets ({}) are the values of σ_res used by XFIT, which is analogous to e.s.d.

Restraint on Fe-O-O Bond angle {σres}	Fe-Ligand distances				Fe-O-O angle	Debye-Waller factors, σ2				S02	R
	Np	Nϵ	O1	O2		Np	Nϵ	O1	O2
	Å				°	Å2					%
115° {1°}	2.01	2.08	1.86	1.22	115.0	0.001	0.003	0.002	0.080	0.914	17.4
125° {1°}	2.01	2.08	1.86	1.23	126.7	0.001	0.002	0.002	0.010	0.915	17.3
135° {1°}	2.01	2.08	1.86	1.22	135.1	0.001	0.002	0.002	0.008	0.915	16.8
145° {1°}	2.01	2.07	1.86	1.22	144.6	0.001	0.003	0.002	0.029	0.914	17.3
155° {1°}	2.01	2.08	1.86	1.22	155.0	0.001	0.003	0.002	0.080	0.914	17.4
165° {1°}	2.01	2.08	1.86	1.22	165.0	0.001	0.003	0.002	0.080	0.914	17.4
175° {1°}	2.01	2.08	1.86	1.22	175.0	0.001	0.003	0.002	0.080	0.914	17.4
No constraint	2.01	2.08	1.86	1.22	135	0.001	0.002	0.002	0.008	0.915	16.8
115°b	2.01	2.07	1.86	1.22	101	0.001	0.004	0.002	0.004	0.914	14.7
125°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8
135°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8
145°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8
155°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8
165°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8
175°b	2.01	2.08	1.86	1.22	135.4	0.001	0.002	0.002	0.008	0.915	16.8

^a The original fit where Fe-O-O is unrestrained is included for comparison.

^b Fe-O-O angles were initially restrained between 115° and 175° and the fit allowed to go to convergence, then the restrain was removed and the fit allowed to go to convergence.

Overall the iron-site geometry of O₂-αHb and O₂-αHb·AHSP is similar to that derived from high resolution (1.25 Å) crystal structures of O₂-HbA (53). However, the Fe-N_ϵ and Fe-O bond lengths were shorter in O₂-αHb than in O₂-αHb·AHSP (2.04 cf. 2.08 Å and 1.83 cf. 1.86 Å). Although these differences are within the conservative estimates of accuracy given in Table 3 (0.02 Å), the systematic error contributions are expected to be the same for both proteins and, hence, the errors obtained from Monte Carlo calculations are a better indication of the relative errors. For the Fe-O₂ bond these errors were 0.005 and 0.006 Å in O₂-αHb and O₂-αHb·AHSP, respectively, indicating that the 0.03 Å lengthening of Fe-O₂ in O₂-αHb·AHSP is significant. It should be emphasized that the Fe-O₂ bonds are substantially shorter than those of the other heme-ligand bonds, which is why the precision is so high. This is due to the rapid increase in the importance of single and multiple-scattering contributions as the bond length decreases.

The O₂-αHb·AHSP data were also fitted to a model where the iron-ligand bond lengths and Fe-O-O bond angle were constrained to be that in the O₂-αHb·AHSP^P30A structure (1), and E₀, S₀², and the Debye-Waller factors were fitted (Table 3, bottom row). The resulting temperature factors were unacceptably high (0.059 Ų for the proximal imidazole N^ϵ, 0.079 Ų for O¹, and 0.080 Ų for O²) with an R value of 32.5%. Clearly, this model does not fit the solution structure.

The ability of this methodology to determine precise bond length and angle information on axial ligands has been established in studies of porphyrin model complexes and by the determination of precise bonding parameters in the heme environments of met- and deoxymyoglobin (41, 66) before the publication of the high resolution protein crystallographic structures (67, 68). Even taking the small differences between O₂-αHb·AHSP and O₂-αHb into account, it is clear that the initial solution complex of O₂-αHb·AHSP has a “normal” coordination geometry similar to that in free O₂-αHb chains.

AHSP Reduces Reactivity of αHb with O₂ and CO

In 2006, Vasseur-Godbillon et al. (28) reported that AHSP binding reduces the O₂ and CO affinities of αHb roughly 3–5-fold. We repeated these ligand binding experiments with wild type and two Pro-30 mutants of AHSP. Table 7 presents a summary of the rate constants for O₂ and CO association (k′_O2 and k′_CO) to, and O₂ dissociation (k_O2) from, αHb in the presence and absence of AHSP variants. The ratio of k′_O2/k_O2 provides the O₂ association equilibrium or affinity constant (K_O2). As reported by Vasseur-Godbillon et al. (28), AHSP binding to O₂-αHb causes a small decrease in the rate of O₂ binding and a larger 3–4-fold increase in the rate of O₂ dissociation. The net result is an ∼4-fold decrease in O₂ affinity (Table 7). These results support the XAFS data and provide direct evidence that AHSP binding leads to weakening of the Fe-O₂ bond, which in turn leads to a significant increase in the susceptibility of O₂-αHb to autooxidation.

TABLE 7.

AHSP·αHb ligand binding kinetic parameters at pH 7, 20 °C

Protein	k′O2	kO2	k′CO	KA,O2
	μm−1 s−1	s−1	μm−1 s−1
αHb	37 ± 9 (40 ± 2)a (50)b	29 ± 7 (22 ± 6)a (40)b	5.3 ± 0.3 (5.2 ± 0.5)a (6)b	1.3 ± 0.5 (1.8 ± 0.5)a (1.2)b
AHSP·αHbc	23 (35)b	73 (130)b	1.4 (2)b	0.30 (0.35)b
AHSPP30A·αHb	35	47	2.2	0.7
AHSPP30W·αHb	28	25	3.8	1.1

^a Rate parameters were taken from Birukou et al. (45).

^b These values were measured at pH 7.0, 25 °C by Vasseur-Godbillon et al. (28).

^c The values of the kinetic parameters for AHSP·αHb represent the average of two independent determinations.

AHSP binding to αHb also slows the rate of bimolecular CO binding by a factor of ∼3–4-fold, as also reported previously by Vasseur-Godbillon et al. (28) (Table 7). CO binding is limited primarily by the rate of internal bond formation and, therefore, provides a sensitive measure of iron reactivity that is primarily determined by interactions between the heme iron and the proximal histidine. The ∼4-fold decrease in k′_CO and O₂ affinity are consistent with conformational strain introduced on the proximal side of the heme pocket of αHb·AHSP due to interactions with AHSP Pro-30. This interpretation is strongly supported by the effects of Pro-30 mutations on ligand binding. As shown in Table 7, AHSP^P30A and AHSP^P30W mutations diminish the effects of AHSP binding on both O₂ affinity and the rate of CO association, and in the case of AHSP^P30W, the observed ligand binding parameters are similar to those of free αHb. Thus, all our structural, spectroscopic, and functional data suggest that structural changes in the proximal heme pocket are induced by AHSP Pro-30, causing a weakening of the Fe-O₂ bond and an increase in the rate of autooxidation.

DISCUSSION

AHSP promotes autooxidation of O₂-αHb to prevent it from reacting with H₂O₂ under conditions of oxidative stress. The resultant bis-histidyl heme does not react with H₂O₂ and, therefore, cannot generate reactive ferryl species and protein radicals (2, 20). The bis-histidyl αHb·AHSP structure has been well characterized by x-ray crystallographic and spectroscopic methods (1, 2, 25–27), but the formation of this structure from O₂-αHb has been less well understood. The data presented here provide key insights as to how AHSP facilitates formation of the bis-histidyl state by making specific interactions at the αHb·AHSP interface and fit a model, explained below, in which αHb·AHSP complexes can adopt one of two well ordered conformations that are intermediates in Hb assembly and achieve αHb detoxification (Fig. 10).

FIGURE 10.

Model for αHb·AHSP intermediates in Hb assembly. Free αHb (left) has a native-like conformation (gray color). Pathways for incorporation of αHb into Hb are shown as black arrows. Deoxy Fe(II) αHb or O₂-αHb bind AHSP (top row) with a near-native structure and are rapidly combined with available βHb to produce native Hb. Fe(III) αHb binds AHSP in the high affinity bis-histidyl structure (orange, bottom row), inhibiting release of Fe(III) αHb to Hb. Methemoglobin reductase reduces bis-histidyl αHb to the native αHb structure. In Fe(II)/O₂-αHb·AHSP, interaction between AHSP Pro-30 and helix G of αHb (star) introduces strain in the proximal heme pocket (orange shading). Under conditions of limiting βHb or elevated redox stress, autooxidation of Fe(II)/O₂-αHb to the inert bis-histidyl complex occurs (red arrows).

Structure of αHb

By reacting pMB with Cys-104 and blocking αHb self-association, we obtained chemical shift-derived backbone dihedral angles and chemical shift analysis of His side chains, which together show that the conformation of free αHb is very similar to its structure in the native Hb tetramer. This conclusion is consistent with previous findings that heme pocket structure is well preserved in isolated subunits, based on far-UV (69, 70) and near-UV CD (69–72) and homonuclear NMR studies of ferrous and ferric αHb (47, 55, 56, 73, 74). None of these data provide any hints as to why free αHb is more prone to oxidation and subsequent rapid loss of the heme moiety and aggregation. Similar observations have been made in other systems. For example, the monomeric point mutants of the carrier protein transthyretin display identical x-ray crystal structures to the wild-type protein and yet have substantially reduced thermodynamic stabilities (75). In such cases, a native-like conformation still represents the lowest energy state available; however, the free energy well in which this conformer sits is shallower due to a reduction in the number of stabilizing interactions. In the case of αHb, stabilization is provided by multiple subunit interactions in the tetramer.

CO- and O₂-Hb Have Native-like Conformations in Complex with AHSP

A previous x-ray crystallographic study from our group suggested that O₂-αHb underwent large conformational changes upon binding AHSP (1), and it was assumed that these changes were responsible for the accelerated autooxidation. In that study residues Asp-74–Leu-91 of αHb (helix F) appeared disordered (as judged from a lack of interpretable electron density), and the heme-ligating residue was switched from the usual His-87 in helix F to His-58 in helix E. However, NMR line shape and TALOS+ chemical shift data presented here show that, in the solution state, helix F retains an ordered α-helical structure in CO-αHb·AHSP. In addition, analysis of His side chain resonances strongly suggests that His-87 retains the heme binding role. In the crystal structure, the Fe-O₂ bond is unusually long (2.7 Å) (1). Such a long Fe-O₂ bond is inconsistent with visible absorption and XANES and EXAFS spectroscopic data (Table 3), which indicate that the iron coordination geometries of O₂-αHb and O₂-αHb·AHSP are very similar. The EXAFS data do indicate a slight lengthening of the Fe-O₂ bond from 1.83 Å in O₂-αHb to 1.86 Å in O₂-αHb·AHSP, providing structural evidence that AHSP promotes superoxide dissociation and a decrease in O₂ affinity (Table 7).

Several possible explanations exist for the differences between the crystal structure described previously and the solution structure reported here. A detergent molecule located close to the heme group in the crystal might have influenced the structure; however, the detergent makes very few direct contacts with the protein. Perhaps more significantly, the crystals were grown under a standard atmosphere making it likely that partial autooxidation of αHb would occur. In addition, the intense x-ray beams used for protein crystallography can promote reduction of ferric proteins (59). A mixture of heme states arising from these processes would introduce static disorder in the crystal, complicating analysis of the x-ray data. Nevertheless, the crystal structure does provide evidence that helix F of αHb is susceptible to conformational distortion in the presence of AHSP.

AHSP Pro-30 Promotes Changes in the Proximal Heme Pocket of O₂-αHb That Enhance Autooxidation

We have previously shown that mutation of Pro-30 to a range of other residues types leads to a 4-fold reduction in autooxidation of αHb. Here we show that introducing Pro at other positions in loop 1 did not restore native O₂-αHb autooxidation and, although the AHSP^P30A/N28P may undergo cis/trans isomerization, O₂-αHb autooxidation was not enhanced above the rate observed for AHSP^P30A, which is only 2–3-fold greater than that of free O₂-αHb (27). These data emphasize that the position of Pro-30 in AHSP loop 1 is critical to enhancing autooxidation.

AHSP Pro-30 is positioned adjacent to the beginning of helix G in the bis-histidyl αHb·AHSP complex. However, in the absence of crystallographic data, the degree of interaction between Pro-30 and helix G in O₂-αHb·AHSP is unknown. Based on mutation and binding data, which show that Pro-30 destabilizes the O₂-αHb·AHSP interface but not the bis-histidyl αHb·AHSP interface (11, 25, 27), we previously proposed that Pro-30 introduces a steric clash at the O₂-αHb·AHSP interface, which is relieved upon transition to the bis-histidyl αHb·AHSP structure. Here we show that AHSP binding induces NMR chemical shift perturbations in F, G, and H helices of CO-αHb, which constitute the proximal heme pocket, and that mutation of Pro-30 markedly reduces these disturbances. These findings are consistent with a steric interaction between AHSP Pro-30 and αHb helix G that is propagated through packing interactions of the F, G, and H helices (represented as orange shading of these helices in Fig. 10, top row). Bonding interactions between the heme iron and proximal His-87 in helix F and the geometry of the proximal pocket are important factors influencing ligand reactivity of the iron (76, 77). It is, therefore, significant that AHSP decreases O₂ affinity and CO association rates of αHb and that mutation of Pro-30 restores ligand reactivity of the heme iron to near-normal levels, which suggests strongly that Pro-30-induced disturbance of the proximal heme pocket contributes to enhanced O₂-αHb autooxidation and lower ligand reactivity in the presence of wild-type AHSP.

O₂-αHb·AHSP and Bis-histidyl·AHSP Intermediates in Hb Assembly

Our data suggest a model in which AHSP can bind αHb in one of two well ordered conformations depending on the oxidation/ligand state of the heme (Fig. 10). O₂-αHb is bound in a near-native conformation (Fig. 10, top), whereas Fe(III) αHb adopts a non-native globin conformation and bis-histidyl heme coordination (Fig. 10, bottom). We expect that Fe(II) deoxy αHb and CO-αHb will also bind to AHSP with near-native protein conformations on the basis that these forms of αHb all show the same affinity for AHSP as O₂-αHb, which is 100-fold weaker than the interaction of AHSP with the non-native bis-histidyl Fe(III) αHb conformation (20, 25, 27). Notably, deoxy-Fe(II) αHb does not adopt bis-histidyl heme coordination (26) despite the availability of an open iron axial coordination site. These studies emphasize that specific coordination between the distal His-58 and the Fe(III) iron center and the structural changes in the αHb polypeptide chain occur in a concerted fashion during oxidation of the αHb·AHSP complex and do not occur when the iron atom is reduced.

Based on kinetic data, Khandros et al. (11) and Mollan et al. (27) proposed that αHb synthesized in the Fe(III) form in the presence of AHSP (78) proceeds through a bis-histidyl αHb·AHSP folding intermediate before heme reduction and incorporation into Hb (Fig. 10, solid black arrows). Bis-histidyl ferric αHb can in principle be transferred to βHb and subsequently reduced (Fig. 10, dashed arrows); however, the rate of ferric αHb·AHSP dissociation is probably too slow to be physiologically relevant, and reduction almost certainly occurs more rapidly (27). Based on autooxidation rates measured in vitro (11, 20, 25, 28, 29), newly reduced Fe(II)/O₂-αHb should be rapidly incorporated into Hb before autooxidation can occur if the concentration of free βHb is high. However, under conditions of limiting βHb that are known to occur during normal erythropoiesis (79, 80), the likelihood increases for the conversion of Fe(II)/O₂-αHb·AHSP species to bis-histidyl ferric complexes (Fig. 10, red arrows). The more rapid autooxidation of αHb in the presence of AHSP may be advantageous, as it provides a means to trap αHb in a relatively unreactive Fe(III) state that inhibits redox cycling and reactions with reactive oxygen species that could otherwise occur with free αHb. In addition, both free O₂-αHb and met-Hb subunits are more prone to H₂O₂-induced ferryl and protein radical formation and precipitation (20, 81), making it advantageous to remove free O₂-αHb and stabilize it as bis-histyl met-αHb·AHSP under conditions of oxidative stress.

To investigate the biological importance of autooxidation rate, Khandros et al. (11) performed gene knock-in experiments in mice. Mutations of AHSP Pro-30 to Ala or Trp did not disturb erythropoiesis (11), which suggested that a 4-fold decrease in αHb·AHSP autooxidation rate is not sufficient to elicit changes in erythropoiesis in the normal mouse model (11, 27). However, it is important to recognize that O₂-αHb·AHSP^P30A still autooxidizes more quickly than does free O₂-αHb (11, 25, 27) and that other interactions across the αHb·AHSP interface, in addition to those with Pro-30, are important for autooxidation and hemichrome formation.

A transition between two metastable αHb·AHSP structures, as outlined in Fig. 10, avoids persistent disordered states that would favor unregulated reactions with an exposed heme and rapid heme dissociation. In this regard, it is interesting that mutations of Pro-30 to Ala or Trp decrease O₂-αHb autooxidation rates but enhance heme dissociation rates (11). In conclusion, it appears that the AHSP sequence has evolved to achieve a fine balance of distorting αHb sufficiently to promote changes in heme coordination that inhibit reactions with H₂O₂ without inducing an unacceptably high risk of heme loss or other irreversible denaturation events.

A Physiological Role for O₂-αHb·AHSP beyond the Red Blood Cell

Recently Straub et al. (82) reported that αHb and AHSP (but not βHb) are expressed in arterial endothelial cells at sites where the endothelial cells make direct contact with underlying smooth muscle (myoendothelial junctions) and that αHb expression regulates NO signaling from the endothelium to smooth muscle. NO signaling is important for maintaining vascular tone and modulating the response of smooth muscle to vasodilator and vasoconstrictor messengers. Importantly, the effect of heme proteins on NO signaling is dependent on the iron oxidation and ligand state. Fe(III) heme interacts only very weakly with NO, providing no barrier to NO diffusion. In contrast, O₂-heme reacts rapidly with NO to form nitrate (NO₃⁻) and Fe(III) heme, thus quenching NO signaling (83–85). In the endothelium, the reaction of NO with O₂-αHb·AHSP is expected to generate bis-histidyl αHb·AHSP. Therefore, in endothelial cells the balance between O₂-αHb·AHSP and Fe(III) bis-histidyl αHb·AHSP will impose restrictive/permissive states for NO signaling, respectively. In this regard, autooxidation and the interaction of Fe(III) bis-histidyl αHb·AHSP with reductase systems in endothelial cells are critical processes. Understanding these processes is now necessary for a full understanding of NO signaling in the vasculature.