Deciphering the Structural Role of Histidine 83 for Heme Binding in Hemophore HasA*

Heme carrier HasA has a unique type of histidine/tyrosine heme iron ligation in which the iron ion is in a thermally driven two spin states equilibrium. We recently suggested that the H-bonding between Tyr75 and the invariantly conserved residue His83 modulates the strength of the iron-Tyr75 bond. To unravel the role of His83, we characterize the iron ligation and the electronic properties of both wild type and H83A mutant by a variety of spectroscopic techniques. Although His83 in wild type modulates the strength of the Tyr-iron bond, its removal causes detachment of the tyrosine ligand, thus giving rise to a series of pH-dependent equilibria among species with different axial ligation. The five coordinated species detected at physiological pH may represent a possible intermediate of the heme transfer mechanism to the receptor.

Heme carrier HasA has a unique type of histidine/tyrosine heme iron ligation in which the iron ion is in a thermally driven two spin states equilibrium. We recently suggested that the H-bonding between Tyr 75 and the invariantly conserved residue His 83 modulates the strength of the iron-Tyr 75 bond. To unravel the role of His 83 , we characterize the iron ligation and the electronic properties of both wild type and H83A mutant by a variety of spectroscopic techniques. Although His 83 in wild type modulates the strength of the Tyr-iron bond, its removal causes detachment of the tyrosine ligand, thus giving rise to a series of pH-dependent equilibria among species with different axial ligation. The five coordinated species detected at physiological pH may represent a possible intermediate of the heme transfer mechanism to the receptor.
The ability of bacteria to cause diseases depends on many parameters allowing the pathogen to invade the vertebrate host. One of them is its skill to scavenge the host iron ion. Indeed, iron is an essential metal for nearly all living organisms. However, its oxidized iron(III) form is hardly soluble, and its reduced iron(II) form is highly toxic. Therefore, in biological fluids, iron mostly exists as a complex with iron-binding proteins or in heme carrier proteins and is scarcely available. Hence, Gram-negative bacteria have developed multiple iron/heme acquisition systems to survive. Most of them rely on outer membrane receptors either specific for exogenous iron/heme sources present in the biotope or for molecules called siderophores or hemophores. These molecules are synthesized and released by bacteria into the extracellular medium (1). Their function is to capture iron or heme from the host and to return it to the specific receptor.
The first hemophore was discovered in Serratia marcescens (2). Since then, several orthologs have been found in Pseudomo-nas aeruginosa (3), Pseudomonas fluorescens (4), Yersinia pestis (5), Yersinia enterocolytica (Sanger Institute) and Yersinia pseudotuberculosis (6). HasA (heme acquisition system) hemophores form an independent family of highly conserved hemecarrier proteins that are not homologous to any other known proteins. They bind free heme, or extract heme from proteins such as hemoglobin and hemopexin, and deliver it to hemophore-specific outer membrane receptors, termed HasR (1), which in turn release it into the bacteria (2,7).
S. marcescens wild-type HasA (HasA WT ) is a 19-kDa monomeric protein that binds b-heme with a very high affinity (K d ϭ 1.9 ϫ 10 Ϫ11 M) and a 1:1 stoichiometry (8,9). In the heme loaded hemophore holoHasA WT , the iron ion is in the oxidized form with the lowest redox potential (Ϫ550 mV) reported so far for heme-binding proteins (9). The x-ray structure of holo-HasA WT shows an unprecedented heme-binding site (10). Heme is held by two extended loops that connect the ␣ and ␤ faces of the protein and is rather exposed to solvent (186 Å 2 ) (Fig. 1). The heme iron ion is bound by an unusual pair of ligands: a tyrosine (Tyr 75 ) and a histidine (His 32 ) (10). This ligand pair has only been observed in a very few proteins, e.g. some hemoglobins from invertebrates (11)(12)(13), the reduced cytochrome c maturation protein CcmE (14), and the oxidized cytochrome cd 1 nitrite reductase (15). Some abnormal human hemoglobins (16) and mutated heme proteins also present this type of coordination (17). X-ray and NMR data previously showed that, in holoHasA WT, the Tyr 75 oxygen forms a tight hydrogen bond with the N␦1 of a neighboring histidine, His 83 , that increases the nucleophilic character of Tyr 75 and strengthens the Tyr 75 -iron coordination bond (18,19). The role of His 83 in heme binding was highlighted by alanine mutagenesis of the two iron axial ligands and of His 83 , followed by the determination of the K d for heme in the mutant proteins. Consistently, heme affinity in H83A is 265 times smaller (K d ϭ 5.0 ϫ 10 Ϫ9 M) than in HasA WT , whereas the mutation of the axial ligand induces a 400-fold loss of affinity for Y75A and only a 5-fold loss for H32A. Moreover, His 83 has been proposed as an alternative iron ligand in the absence of Tyr 75 or of both His 32 and Tyr 75 (20). Noteworthy, in contrast to His 32 , the Tyr 75 -His 83 pair is conserved in all the hemophores, which suggests that the Tyr 75 -His 83 hydrogen bond plays a central and recurrent role for the activity of these heme carrier proteins.
To decipher the structural role of His 83 in HasA, we characterized the electronic properties and the iron coordination of holoHasA WT and holoH83A in the oxidized form as a function of pH, using a panel of techniques such as NMR, EPR, x-ray crystallography, resonance Raman, and UV-visible spectrosco-pies. Although holoHasA WT possesses a unique coordination as a function of pH, substitution of His 83 with an alanine residue gives rise to pH-dependent coordination forms with different axial ligation. The role of the Tyr 75 -His 83 hydrogen bond for the stabilization of the Tyr 75 -iron axial coordination was unraveled. Disruption of this hydrogen bond at physiological pH results in a five-coordinate species that most probably represents an intermediate of the heme transfer mechanism to the receptor.

EXPERIMENTAL PROCEDURES
Preparation of HasA Proteins-Wild-type hemophore HasA WT from S. marcescens, and mutant H83A were obtained and cloned as previously reported (2,20). Uniformly 15 N-and 15 N/ 13 C-labeled proteins were produced at 303 K in M9 minimal medium containing 15 NH 4 Cl and 13 C-glycerol as the sole nitrogen and carbon sources and were purified as described previously (9,21). The purity of the proteins was checked by SDS-PAGE. The protein concentrations were calculated using the previously determined ⑀ 277 nm values of 19500 M Ϫ1 cm Ϫ1 for HasA WT and 18500 M Ϫ1 for H83A mutant. Cleavage of the last nine residues (180 -188) of HasA proteins was performed using the S. marcescens protease Prt SM to prevent sequential proteolysis of the C-terminal amino acids (22). Unless otherwise specified, the samples were dissolved in 20 mM phosphate buffer. Heme loading was performed as previously described (19).
The cyano-complexes were prepared by adding a 15-fold excess of KCN to the proteins (100 mM KCN stock solution in 20 mM phosphate, pH 7).
Absorption Spectroscopy-The absorption measurements were achieved in a Perkin-Elmer Lambda 2 spectrophotometer at ambient temperature using 1-or 0.2-cm path length cells.
Resonance Raman Spectroscopy-Resonance Raman (rR) 2 spectra were recorded from protein samples that were 25-80 M, contained in 5-mm NMR tubes spinning at ϳ20 Hz. Raman scattering was excited using either 413.1-nm emission from a Krϩ laser (15-20 milliwatt) or 514.1 nm emission from an Ar laser (140 milliwatt). UV-visible absorbance spectroscopy was used to check sample integrity before and after the rR spectra were recorded. The spectra were recorded at ambient temperature as previously described (23). The spectra were calibrated using the Raman bands of toluene and dimethylformamide. The samples of HasA-H83A examined for isotopic sensitivity were prepared from a 0.3 mM stock protein solution. A 9-l aliquot was diluted into 50 mM CAPS buffer at pH 10 NMR Spectroscopy-NMR experiments were performed on protein samples with concentration ranging from 0.8 to 2.5 mM in 20 mM phosphate buffer adjusted at the requested pH. Onedimensional 1 H NMR spectra were acquired with 8192 data points either at 16.4 T with a spectral width of 240 ppm or at 11.7 T with a spectral width of 200 ppm. During the recycle delay of 150 or 30 ms, respectively, water suppression was achieved by presaturation. One-dimensional 15 N NMR spectra were acquired at 9.4 T with a 30-s 90°pulse, a spectral width of 450 ppm, 2048 data points, and a recycle delay of 10 ms. Onedimensional 13 C NMR spectra were acquired at 16.4 T with a 12.4-s 90°pulse, a spectral width of 500 ppm, and 4096 data points. Several experiments were performed with carriers at Ϫ200, 150, 567, and 850 ppm to identify signals over wide spectral regions. Recycle delays were 10 or 200 ms. 13 C T 1 values were measured via the inversion recovery sequence. Several series were performed using 10 and 100 ms as recycle delays with the 13 C-carrier frequency at different positions to properly invert and excite all 13 C resonances. Direct detected 13 C-13 C COSY experiments were acquired at 16.4 T with 2048 ϫ 256 data points, 1024 scans, and a recycle delay of 300 ms for holo-HasA WT and 4096 ϫ 512 data points, 512 scans, and a recycle delay of 250 ms for holoH83A. Spectral widths of 220 ppm in both dimensions with the carriers at 150 ppm for the indirect dimension and at 110 ppm for the direct dimension were used. Unless otherwise specified, NMR spectra were recorded at 303 K.
EPR Spectroscopy-EPR spectra were recorded with a Bruker ESP300E spectrometer fitted to an Oxford Instrument ESR900 helium flow cryostat. The experiments were carried out on protein samples with concentration ranging from 0.1 to 0.5 mM in 50 mM sodium acetate buffer, pH 5.5, 50 mM phosphate buffer, pH 7, and 50 mM Tricine, pH 9. Spin quantitations of the low and high spin heme signals were performed in nonsaturating conditions as previously described (23).
X-ray Crystallography-The crystallization experiments were performed using the hanging drop vapor diffusion method. The drops were prepared by mixing 2 l of the protein solution (4.0 mg/ml in 50 mM phosphate buffer, pH 7, 0.02% X-ray diffraction data were collected at 100 K at the ESRF (Grenoble, France) beamline ID14-EH1 using an ADSC Quantum 4R CCD detector. The wavelength of the synchrotron x-rays was 0.933 Å. The crystal was rotated through 90°with a 1.0°oscillation range/frame. The raw data were processed and merged with MOSFLM (24) and SCALA (25). The crystals of holoH83A (space group P3 2 21) diffracted to a maximum resolution of 2.6 Å. Molecular replacement with holoHasA WT (Protein Data Bank code 1B2V) as model gave a unique solution with a correlation coefficient of 0.72 and a R factor of 35.4%. Refinement was performed using REFMAC5, part of the CCP4 package. The electron density clearly indicated the presence of the heme group but was diffuse and featureless ( Fig. 2). Therefore, the geometry of the heme group was tightly constrained to standard values and refined in the two heme orientations related by 180°rotation along the ␣-␥ meso axis identified in holoHasA WT (19,26). Water molecules were added with CCP4/wARP (27). The stereochemistry of the final structure was evaluated using PROCHECK (28). The refinement statistics and final R factors are listed in Table 1. Fig. 3 shows the absorbance spectra of holoHasA WT and holoH83A. The spectrum of holoHasA WT at pH 5.6 ( Fig. 3A) shows maxima at 406 nm (Soret band), 494, 537, 568, and 618 nm that are consistent with a mixture of two species, one high spin (HS) and one low spin (LS). Indeed, whereas the charge transfer band at 618 nm is characteristic of a HS species, the presence of two Q-bands at 537 and 568 nm indicates the presence of a LS species. The absorbance spectrum did not change upon titration from pH 4.4 to 10.4, showing that the population of the two species is pH-insensitive. It is noteworthy that the addition of KCN to the protein at a 15-fold excess did not produce any spectroscopic changes, either at pH 7.3 or at pH 9.4 (data not shown).

Visible Absorbance Spectra of Holoproteins
In contrast to holoHasA WT , the absorbance spectrum of holoH83A is sensitive to pH. At acidic and neutral pH, the spectra exhibit a Soret maximum at 404 nm, Q-bands at 502 nm and 536 nm, and a CT band at 630 nm (Fig. 3B). This pattern is consistent with the presence of a HS species. When pH increases, bands at 502, 536, and 630 nm decrease, whereas new bands emerge at 490, 543, and 610 nm. The disappearance of the charge transfer band at 630 nm at the advantage of the 610-nm band indicates that the HS species observed at acidic and neutral pH progressively disappears when pH increases for the benefit of a new HS species. Moreover, a LS species appears, as shown by a new band at 576 nm. Thus, at alkaline pH, holoH83A exists as a mixture of a HS and a LS species, the HS species being different from the one observed at acidic pH. Like holoHasA WT , no detectable change was observed after the addition of a 15-fold excess of a KCN solution neither at pH 7.3, where a HS species is dominant, nor at pH 9.4, where two HS species and one LS species are present.

Resonance Raman Spectra of holoHasA WT and holoH83A Mutant
Consistent with previously reported NMR data, the rR spectra of holoHasA WT show a mixture of LS and HS species. The rR data further indicate that the HS component of holoHasA WT is a hexacoordinate species. The 1300 -1700 cm Ϫ1 region of heme rR spectra comprises bands corresponding to porphyrin in-plane vibrational modes. They are diagnostic for oxidation The diffraction data have been refined with the two heme group orientations related by 180°rotation along the ␣-␥ meso axis as identified in holoHasA WT (19,26). The low resolution of the electron density might be due to high flexibility and/or the presence of several conformations. state, coordination number, and spin state of the heme iron atom (44,45). The spectrum of holoHasA WT (Fig. 4A) is indicative of a ferric species with its characteristic 4 band at 1372 cm Ϫ1 . The spin state marker bands indicate an equilibrium mixture of six coordinate high spin (6c HS) ( 2 , 1561 cm Ϫ1 ; 3 , 1476 cm Ϫ1 ; 10 , under the 1624 cm Ϫ1 band) and six coordinate low spin (6c LS) ( 2 , 1580 cm Ϫ1 ; 3 , 1503 cm Ϫ1 ; 10 , 1633 cm Ϫ1 ) states. The ratio of the two 3 and two 10 band intensities does not change between pH 5.2 and 9.2, whereas the LS bands increase in intensity at the expense of their HS counterparts as the temperature is lowered (data not shown). These observations suggest that the mixture of 6c HS and 6c LS hemes arises from either two pH independent axial ligand sets or a single axial ligand set characterized by a pH-insensitive spin state equilibrium.
For hemoproteins, excitation into the tyrosinate-Fe(III) CT band (near 500 nm) can provide resonance enhancement of Raman scattering by vibrations characteristic of bound phenolate. No bands attributable to bound tyrosine vibrations were identified in the 514.5-nm excited rR spectra of HasA WT . However, most rR bands assigned to Fe-Tyr Ϫ stretching modes have been observed in spectra of 5c HS hemes (14, 29 -33). Because the heme in HasA WT is clearly six coordinate, Raman scattering by these (Fe-Tyr Ϫ ) and coordinated tyrosinate modes may be poorly enhanced with green excitation.
Spectra of holoH83A were recorded over the pH range of 5.2 to 10.0 with both B-band (Soret) and Q-band excitation at 413.1 ( Fig. 4) and 514.5 nm (Fig. 5), respectively. Three pH-dependent forms, two HS and one LS, were detected, as suggested by the optical absorbance spectra and show their pH-dependent interconversion.
High frequency spectra recorded at pH 5.2 or 7.2 contain 2 , 3 (Fig.  4A), 10 , 11 , and 19 (Fig. 5) bands that reveal the presence of 5c and 6c HS hemes. They fall within frequency ranges characteristics of coordination number and spin state, as indicated in the figures. The 3 band observed at 1477 cm Ϫ1 is consistent with a 6c HS heme. The second 3 band at 1487 cm Ϫ1 falls at the low end of the frequency range expected for 5c HS proteins (29). The more intense 2 band is consistent with the presence of a 6c LS species, although no 3 band attributable to 6c LS heme can be observed in the Soret-excited spectra recorded at pH 5.2 and 7.2. The evidence of the 6c LS species is corroborated by pairs of 10 and 11 bands that arise from both HS and LS hemes in the Q-band excited spectra (Fig. 5). Both the 1477-and 1487-cm Ϫ1 3 bands corresponding to the 6c and 5c HS species persist at pH 9.0, whereas a new 3 band grows at 1501 cm Ϫ1 characteristic of the 6c LS species. The 6c HS and 6c LS 3 bands increase in intensity with increasing pH relative to the 1487 cm Ϫ1 band are consistent with a transition from 5c to 6c and with an equilibrium between 6c HS and 6c LS hemes at alkaline pH. The relative intensities of the HS and LS pairs of 10 and 11 bands with increasing pH corroborate the presence of this equilibrium (Fig. 5). Spin state equilibria of this nature have been reported for hydroxide complexes of various heme proteins (34 -36). Persistence of bands attributable to 5c HS heme at pH 9 and 10 clearly shows that complete conversion to the hydroxide complex is not achieved under the conditions of the rR experiment.
The low frequency region rR spectra are useful in identifying metal-ligand vibrations and porphyrin deformations, many of which involve motion of peripheral substituents (37,38). HoloH83A (Fig. 4B) exhibits bands in the 480 -570-cm Ϫ1 range, where (Fe-OH) bands of HS and LS heme hydroxides occur. However, because the frequencies of these bands are insensitive to both D 2 O and 18 OH 2 , they are assigned to porphyrin modes (44). The bands between 380 and 425 cm Ϫ1 are due to propionate and vinyl bending modes and do not vary significantly as a function of pH (Fig. 4B). The HasA WT spectrum exhibits two vinyl bending bands, consistent with the two vinyl-protein interactions. The spectra of holoH83A exhibit a single vinyl bending band in acidic and neutral solution and a broadened band in alkaline solution, suggesting that the struc-  FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 ture of the WT heme pocket has been modified. At low pH, a pronounced band is observed at 196 cm Ϫ1 (Fig. 4B). Although this is in a region often associated with modes having axial iron-His stretching character, the only such modes that have been identified in ferric hemes are in those having the bis-His axial ligand set. Because this set of ligands is inaccessible in holoH83A, the 196 cm Ϫ1 band is concluded to arise from a porphyrin mode. Based on its greater intensity at low pH, where HS pentacoordinate iron dominates the heme speciation, the 196 cm Ϫ1 band is attributed to a mode having porphyrin outof-plane character.

Characterization of holoHasA WT and holoH83A Mutant by NMR
1 H NMR Spectroscopy-The analysis of 1 H-15 N heteronuclear single-quantum correlation spectra of holoHasA WT and holoH83A indicate that for both systems the secondary structure and protein folding are maintained over the investigated pH range (supplemental Fig. S1 and Ref. 19).
One-dimensional experiments over large spectral windows were used to monitor the heme coordination sphere. We previously showed that, in HasA WT , HS and a LS species were in fast exchange on the NMR time scale (19). Indeed, the onedimensional 1 H NMR spectrum of the protein shows heme methyl resonances with chemical shift values intermediate  between those expected for a purely HS S ϭ 5/2 heme iron(III) and those for a purely LS S ϭ 1/2 heme iron(III) (Fig. 6A). One major and one minor set of resonances are visible on the spectrum, corresponding to the two possible orientations of the heme along the ␣-␥ meso axis (10,19). The chemical shifts and the nonselective T 1 and T 2 values for the hyperfineshifted signals are in the 1-10-ms range, and the line widths are ϳ400 Hz at 16.4 T (supplemental Table S1). The relaxation times are close to those expected for a HS species (39). Two exchangeable, broad and fast relaxing signals with a 70:30 intensity ratio are observed at 65.8 and 73.2 ppm, respectively (Fig. 6A). Their line widths are ϳ1000 Hz, and their T 1 values are shorter than 2 ms. They typically correspond to imidazole N␦1H of iron coordinated histidines (39 -42) and are attributed to His 32 N␦1H in the two heme orientations. Most of the hyperfine-shifted resonances in the 80 to Ϫ30 ppm region exhibit a pronounced temperature dependence of their chemical shift, with a nonlinear and anti-Curie behavior (supplemental Fig. S2A) for the heme methyls and His 32 N␦1H, consistent with an equilibrium between a LS and a HS form (43)(44)(45). Hyperfine-shifted signals of holoHasA WT do not show significant shifts from pH 4.5 to 10.4, indicating that the equilibrium between the two spin states is not affected by pH. The addition of a 10-fold excess of cyanide failed to influence the position and intensity of any of the resonances and did not show any detectable signal indicative of a cyanide adduct.
The holoH83A 1 H NMR spectrum obtained at acidic pH displays two sets of well resolved peaks in a 3:1 ratio with four distinct heme-methyl resonances between 60 and 80 ppm, typical of HS species (Fig. 6B) (39). They correspond to the two different heme orientations in the heme-binding pocket. A very broad, exchangeable, resonance associated with the high spin species is present at 102.3 ppm, in the spectra at 303 K and pH 5.6, with a T 1 Ͻ 1 ms and a line width of ϳ1000 Hz. It is characteristic of the N␦1H of an axial histidine for a HS species (39,46). Because no other histidine can coordinate the iron in holoH83A, it corresponds to His 32 N␦1H. Longitudinal relaxation times are less than 1 ms for the four heme methyls and ϳ4 -5 ms for all the other downfield shifted signals. As illustrated in supplemental Fig. S2B, hyperfine-shifted resonances exhibit a Curie behavior, indicative of a pure HS state (39). The meso 1 H shifts are characteristically downfield (ϳ40 ppm) for the hexacoordinate HS ferric complexes and upfield (Ϫ20 to Ϫ70 ppm) for the pentacoordinate high spin ferric complexes (39,47). The 1 H NMR spectrum of holoH83A clearly locates meso H-peak at ϳϪ25 ppm. Therefore, at acidic pH values holoH83A is a pentacoordinate HS species. In addition to the two sets of resonances described above, the spectrum of holoH83A at acidic and neutral pH exhibits unresolved overlapping signals between 20 and 40 ppm (Fig. 6, B and C). With increasing pH, the acidic pentacoordinate HS species decreases in intensity, whereas the overlapping signals at 20 -40 ppm increase. Additionally, a minor set of signals (ϳ10%) appears in the region 40 -60 ppm. They are overwhelmed by the acidic HS form and can be unambiguously identified only at pH Ͼ 9 (Fig.  6D). The two alkaline species undergo line narrowing at increasing pH, indicative of an exchange process in the quasislow regime. Line narrowing at higher pH allows the identification of several well resolved peaks in the 20 -40-ppm region with T 1 of ϳ3 ms and line widths of ϳ500 Hz. The temperature dependence of these signals is nonlinear (supplemental Fig.  S2C). All of these findings are consistent with an hydroxidebound form (36,(61)(62)(63). Two broad and exchangeable signals, missing at acidic pH, are observable at alkaline pH (a major one at 64.7 ppm and a minor one at 70.5 ppm at pH 9.9 and 303 K) (Fig. 6D, inset). These resonances, characteristics of the imidazole N␦1H of a heme iron-coordinated histidine (39 -42), correspond to His 32 N␦1H in the two heme orientations. Thus, His 32 is a heme iron ligand in the hydroxide-bound species. The minor alkaline form (10%) is characterized by larger shifts with respect to the hydroxide form, accounting for a predominantly HS species.
No spectra changes were observed upon the addition of a 15-fold buffered KCN to holoH83A at pH 7.3 or 9.4. Thus, whatever the dominant species in solution, holoH83A does not bind cyanide at any pH. 15 N NMR Spectroscopy-holoH83A spectra were recorded at various pH values at 303 K. The histidine resonances were assigned by comparison with the already available assignment of holoHasA WT and holoH83A at pH 5.5 and 7.0, at 317 K (19). The His 32 N␦1 resonance is observed at 194.2 ppm, at pH 7 (Fig.  7). The His 32 N⑀2 signal is undectectable as in holoHasA WT , in agreement with its coordination to the heme iron. The spectrum recorded at pH 9 shows dramatic broadening of all reso-  FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 nances, thus preventing histidine identification at alkaline pH (data not shown). 13 C NMR Spectroscopy-Resonances of the iron-coordinated residues can be observed using 13 C direct detection in one-dimensional experiments (19). The comparison of holoHasA WT and holoH83A spectrum at pH 5.6 is shown in Fig. 8. The holoH83A spectra show eight hyperfineshifted well resolved peaks in the downfield region (Fig. 8B, peaks  A-H). Two additional signals were observed at lower temperature (283 K) (data not shown). On the basis of the relative intensity, five pairs corresponding to the two heme orientations could be identified, a set of major signals labeled A, C, D, F, and H and a set of minor signals. Four pairs have T 1 values shorter than 1 ms, and one pair has T 1 in the 10 -13-ms range ( Table 2). T 1 values reported for holoHasA WT are of the same order of magnitude (19). Analogous with the findings for the 1 H heme resonances, larger hyperfine shifts with respect to the wild-type and Curie behavior are observed in the 13 C spectrum of the mutant. This behavior can be accounted for by the presence of a HS form. The number of hyperfine-shifted resonances is smaller than for holo-HasA WT , consistent with the presence of a single protein ligand. We therefore assigned them to the axial ligand His 32 .

Structural Role of His 83 in HasA
At pH 8.3, a new set of weakly hyperfine-shifted signals is observed in the range ϩ200/Ϫ50 ppm (Fig. 8C,  inset). Their relative intensity with respect to that of the strongly hyperfine-shifted signals is the same as observed in the corresponding 1 H spectrum between the 5c HS and hydroxide-bound forms. Resonances in a similar chemical shift range have been detected for the axial ligands in LS cytochrome c. 3 The paramagnetic tailored 13 C-13 C COSY spectra in the aromatic region at acidic pH of holoHasA WT and holoH83A are essentially superimposable. However, in the mutant, a new cross-peak appears in the region diagnostic for C␤-C␥ correlations of aromatic residues. This peak can only correspond to the Tyr 75 C␤-C␥ correlation because all the other aromatic residues have been already assigned in both spectra, which are similar in this region (Fig. 9, A and B). In the holoH83A spectrum at pH 8.3, the latter correlation is still present. Moreover, an additional peak appears that corresponds to the Tyr 75 C␤-C␥ cor-3 P. Turano, unpublished results.   6 (B), and holoH83A (1.9 mM) at pH 8.3 (C). The expanded scale of trace C permits the identification of 13 C signals (indicated with asterisks) attributed to the alkaline major form. The spectra were recorded at a 175 MHz 13 C Larmor frequency. relation of the low spin form (Fig. 9C). 13 C-13 C COSY crosspeaks are detectable only if their line width is smaller than ϳ300 Hz (48). A coordination bond induces a strong electron spin delocalization from the metal to the ligand affecting both chemical shift and nuclear relaxation (49), thus making the signals of the ligands undetectable in the 13 C-13 C COSY. This means that Tyr 75 is not a heme iron ligand in any of the two forms detected by NMR as a function of pH. In holoHasA WT 13 C-13 C COSY, Tyr 75 C␤-C␥ correlation was not observed (Fig. 9A).
Electron Paramagnetic Resonance Spectroscopy of the Ferric Proteins-At neutral pH, holoHasA WT exhibits a rhombic LS signal with g values at 2.80, 2.20, and 1.71 (Fig. 10A). A weak HS S ϭ 5/2 signal with a slight rhombic character is also visible at g ϭ 6.3 and 5.8. It represents less than 10% of the total spin intensity. The spectrum does not change at pH 5.5 (not shown), but at pH 9.0 some spectral heterogeneities are revealed by the appearance of a second LS component at g ϭ 2.76, 2.20, and 1.76. However, the ratio of the HS to LS signals remains essentially unchanged at this pH (Fig. 10B). Thus, the EPR results indicate that the coordination sphere of the iron ion is not modified between pH 5.5 and 9.0.
To compare the local environment of the heme Fe 3ϩ ion for the protein in solution and in the crystal, EPR experiments were also performed on a sample made of small HasA WT crystals spread in mother liquor at pH 8 and frozen. The spectrum of this polycrystalline sample shows a major LS signal at g ϭ 2.8, 2.2, and 1.71 and a minor HS signal (Fig. 10C). Their close similarity with those given by the protein solution shows that the magnetic properties and the coordination of the heme Fe 3ϩ ion are the same.
The EPR spectrum of holoH83A variant at acidic pH shows a HS signal with axial symmetry (Fig. 10D) at g Ќ ϭ 5.8 and g // ϭ 1.99. At neutral and basic pH (Fig. 10, E and F), a LS signal appears as the major component with g ϭ 2.73, 2.19, and 1.78, and a slightly rhombic HS signal at g ϭ 6.5 and 5.4 is also present. The relative proportion of the HS signal appeared to be preparation-dependent but remains lower than 20%.
X-ray Structure of Ferric holoH83A Mutant-One molecule was present in the asymmetric unit, giving a crystal volume/ protein mass (V M ) of 4.33 Å 3 Da Ϫ1 and a solvent content of 72% by volume (50). The coordinates describing one protein molecule with one heme group per asymmetric unit were refined to a final R factor and R free of 23.6% and 27.3%, respectively. As for the structure of holoHasA WT , the last 14 C-terminal residues are undetectable.
Although holoH83A only diffracted to medium resolution, and the B-factors are substantially larger than those of holo-HasA WT (supplemental Fig. S3), comparison of both structures show that they are very similar (supplemental Fig. S4). Overall root mean square deviation is 0.51 Å. The most significant differences concern the two heme-binding loops, i.e. residues 32-44 and 73-81 that move away from the heme in holoH83A, with an apparent increase in the His 32 N⑀2-Fe and Tyr 75 O-Fe bond lengths. However, the quality of the structure precludes to determine whether Tyr 75 is coordinated to the heme iron in the crystal of holoH83A or there is no more an axial ligand, as established for holoH83A in solution. Indeed, in holoHasA WT , FIGURE 9. Portions of two-dimensional, paramagnetic tailored, 13 C-13 C COSY spectra containing the C␤-C␥ correlations of phenylalanine and tyrosine residues. A, holoHasA WT at pH 5.6. B, holoH83A at pH 5.6. C, holoH83A at pH 8.3. the tyrosinate form of Tyr 75 is stabilized by a tight hydrogen bond between His 83 N␦1 and Tyr 75 O (1.86 Å). In holoH83A, the only potential candidate to form a hydrogen bond with the hydroxyl proton of Tyr 75 is the Ala 83 carbonyl oxygen. Although the distance between the two atoms, ϳ4 Å, does not preclude the formation of a weak hydrogen bond, the tyrosinate character of Tyr 75 would remain poor.

DISCUSSION
The heme iron spin state and the coordination sphere of both holoHasA WT and holoH83A have been defined over a wide range of pH values. All of the spectroscopic techniques indicate that the heme iron coordination in holoHasA WT does not undergo changes in the pH range 4.5-10.4. Resonance Raman shows that the sixth position in the high spin state is neither vacant nor occupied by a water molecule, supporting His 32 and Tyr 75 ligation. NMR data accounts for a thermal high spin-low spin equilibrium in fast exchange on the NMR time scale with a single set of axial ligands, His 32 and Tyr 75 . The equilibrium is related to modulations in the strength of the coordination bond between the iron and Tyr 75 , which in turn depends on the strength of the hydrogen bond between Tyr 75 O and His 83 N␦1 and on the tyrosine protonation state. For holoH83A, the different techniques monitor changes in iron coordination number and spin state as a function of pH that are summarized in Table 3.
In solution at acidic pH the major form of holoH83A is a high spin pentacoordinate species. The fifth ligand is unambiguously identified by NMR as His 32 . Resonance Raman reveals the presence of two additional six coordinated species with HS and LS state. Relative intensities of rR bands do not generally correlate with the relative populations because of differences in their resonance enhancement. NMR spectra at acidic pH show evidence of a minor species, characterized by broad features in the 20 -40-ppm range, accounting for ϳ20% of total signal intensity and attributed to an OH Ϫ bound form. Signal broadening arises from an exchange process, quasi-slow on the NMR time scale, with another protein form, whose intensity is beyond detection at acidic pH and probably accounting for the 6c HS species observed via rR spectroscopy.
At increasing pH, all of the spectroscopic techniques monitor a decrease of the 5c HS species. UV-visible and EPR spectroscopies show the appearance of two new species, one HS and one LS species. They are consistent with the six-coordinated species detected by rRaman. Relative intensity of NMR signals shows that the six-coordinated LS form is the dominant species at alkaline pH values and that the HS spin species is ϳ10%. However, the nature of these species and their spin state need to be commented.
As outlined above, relaxation data and chemical shifts of the dominant alkaline species account for a His 32 /OH Ϫ axial coordination. As described in the literature for a variety of ferric heme proteins, this is not a purely low spin state as also confirmed by the temperature dependence of the chemical shifts (51)(52)(53)(54).
Acid-base transitions taking place in ferric heme proteins with a H 2 O/OH Ϫ character are generally correlated to a change in the protonation status of a close residue in the heme pocket (55). This residue, which is a histidine in most cases, forms a hydrogen bond with the iron bound water (55). The histidine deprotonation causes ionization of the bound water, resulting in a predominantly low spin hydroxide form. In holoH83A, the most likely residue for deprotonation in close proximity to the bound water is Tyr 75 . Thus, we propose that, at increasing pH, the phenolic oxygen atom of Tyr 75 forms a hydrogen bond with the water ligand, thereby stabilizing the hydroxide complex, as summarized in Scheme 1 and Table 3. The modulation of the H-bond strength involving H 2 O/OH Ϫ may broaden the HS and LS Fe-OH modes, such that they are not observed in the low frequency rR spectrum. The pK a of the transition from the fivecoordinated form and the six-coordinated form is ϳ8.4 (Scheme 1). Transitions involving deprotonation of bound water have been reported to occur with a wide range of pK a values (56,57). In the present case, exchange broadening prevents us from an accurate estimate of the pK a for the transition between the 6c HS and the 6c LS form (Scheme 1). However, their relative intensity does not vary at pH Ͼ 8. Estimates of rate constants for the two equilibria depicted in Scheme 1 are ϳ6 ϫ 10 4 s Ϫ1 for the transition between the 6c HS and the 6c LS form and much lower than 5 ϫ 10 4 s Ϫ1 for the transition between five-coordinated to six-coordinated species. Although the various spectroscopic techniques we employed here probe a wide range of time scales, they converge to a consistent view of pHdependent heme speciation in solution.
Accessibility of heme to anionic ligands seems restricted in holoHasA WT and holoH83A, which do not bind cyanide, either at neutral or at basic pH, despite their spin state. Cyanide is usually considered to be a strong ligand for ferric heme. How-SCHEME 1 ever, the nature of the residues in the heme active site heavily affects the binding properties of the cyanide anion. Reduced or even loss of affinity toward HS heme iron (III) has been found in the case of heme hydrophobic environment (58 -61). In holo-HasA WT and holoH83A, the heme is surrounded by residues involved in both hydrophobic and stacking interactions. The only three polar residues in the vicinity of the heme are all on the His 32 side. In holoH83A, the replacement of a histidine with an alanine mutant makes the Tyr 75 side of the heme more accessible but also more hydrophobic. His 83 modulates the stability of hexacoordinate heme in holoHasA WT through hydrogen bonding with the coordinated side chain of Tyr 75 . The purified hemophore, HasA WT , has a much higher intrinsic affinity for heme (K d ϭ 1.9 ϫ 10 11 M) than its purified physiological target, HasR (K d ϭ 0.2 ϫ 10 Ϫ6 M) (23). As previously suggested, relative affinities should invert to accomplish heme transfer from the hemophore to the cell surface receptor (19). Removal of His 83 is not sufficient to cause such an inversion but reduces heme affinity by approximately 2 orders of magnitude. The modulation of the iron-Tyr 75 bond strength caused by the Tyr 75 -His 83 hydrogen bond suggests a certain plasticity of the heme-binding pocket on the Tyr 75 side of the heme that could play crucial roles in the mechanism of heme release. Mutagenesis studies have suggested that the binding of HasA WT to HasR involves two regions that make independent contacts with the receptor: the regions of residues 51-60 and 95-107 (62). Residues 51-53 are at the edge of the heme-binding pocket and very close to tyrosine 75 (C␣ Ser 51 -C␣ Tyr 75 , 5.8 Å). Conformational changes in holo-HasA WT that are induced by binding to HasR could propagate to the heme-binding loop bearing Tyr 75 , thereby triggering the elongation and breaking of the His 83 -Tyr 75 hydrogen bond.
The results presented herein show that disruption of the His 83 -Tyr 75 H-bond results in a 5c heme. Because transfer of heme to HasR requires scission of the Fe-HasA bonds, the present characterization of holoH83A at physiological pH provides a picture of a possible mechanistic intermediate in that heme transfer reaction.