Protein Conformation Changes of HemAT-Bs upon Ligand Binding Probed by Ultraviolet Resonance Raman Spectroscopy*

HemAT from Bacillus subtilis (HemAT-Bs) is a heme-based O2 sensor protein that acts as a signal transducer responsible for aerotaxis. HemAT-Bs discriminates its physiological effector (O2) from other gas molecules (CO and NO), although all of them bind to a heme. To monitor the conformational changes in the protein moiety upon binding of different ligands, we have investigated ultraviolet resonance Raman (UVRR) spectra of the ligand-free and O2-, CO-, and NO-bound forms of full-length HemAT-Bs and several mutants (Y70F, H86A, T95A, and Y133F) and found that Tyr70 in the heme distal side and Tyr133 and Trp132 from the G-helix in the heme proximal side undergo environmental changes upon ligand binding. In addition, the UVRR results confirmed our previous model, which suggested that Thr95 forms a hydrogen bond with heme-bound O2, but Tyr70 does not. It is deduced from this study that hydrogen bonds between Thr95 and heme-bound O2 and between His86 and heme 6-propionate communicate the heme structural changes to the protein moiety upon O2 binding but not upon CO and NO binding. Accordingly, the present UVRR results suggest that O2 binding to heme causes displacement of the G-helix, which would be important for transduction of the conformational changes from the sensor domain to the signaling domain.

Recently, a variety of heme-containing gas sensor proteins have been discovered through gene analysis of various organisms from bacteria to mammals (1)(2)(3)(4)(5)(6). The O 2 -sensing proteins identified so far include FixL (7), phosphodiesterase A 1 from Acetobacter xylinum (8), Escherichia coli direct oxygen sensor (DOS) 3 (9), and HemAT (10). FixL, A. xylinum phosphodiesterase A1, and E. coli DOS belong to the PAS (PER/ ARNT/SIM) family and contain a heme-bound PAS domain as a sensor, but HemAT is globin-like, belonging to the methyl-accepting chemotaxis protein (MCP) family. The effector domain of FixL serves as a protein kinase, which regulates the expression of the nitrogen fixation gene by phosphorylating the FixJ protein (7,11,12). A. xylinum phosphodiesterase A1 catalyzes the hydrolysis of di-cGMP, which is required for the activation of cellulose synthase in cellulose-producing bacteria (8). E. coli DOS also exhibits significant phosphodiesterase activity toward di-cGMP (13).
HemAT is a heme-based signal transducer protein responsible for bacterial aerotaxis (10, 14 -17). HemAT consists of two domains, the sensor and signaling domains, connected by a linker region. The sensor domain has globin folds and contains a heme that acts as the O 2 binding site, whereas the signaling domain interacts with a histidine kinase protein (CheA), a component of the CheA/CheY two-component signal transduction system that regulates the rotation direction of the flagellar motor (18 -20). The x-ray crystallographic analysis of HemAT has been completed only for the reduced ligand-free and oxidized CN-bound forms of the truncated sensor domain, which stays as a homodimer (21). Fig. 1A displays the crystallographic structure of one subunit of the CN-bound form of the truncated sensor domain of Bacillus subtilis HemAT (HemAT-Bs). The CN ion and His 123 are the heme axial ligands in the distal and proximal sides, respectively. The CN ligand forms a hydrogen bond with Tyr 70 (21). The truncated sensor domain of HemAT-Bs contains one Trp and five Tyr residues; Trp 132 and Tyr 133 are contained in the G-helix in the proximal side of heme, but Tyr 70 is present in the distal side, whereas Tyr 13 , Tyr 49 , and Tyr 148 are located far from the heme (Fig. 1A). These aromatic residues will be used as probes to monitor the protein conformational changes upon ligand binding in this ultraviolet resonance Raman (UVRR) study.
One of the important issues concerning the structural biology of gas sensor proteins is clarifying structural changes of the heme and the nearby residues (Tyr 70 , His 86 , and Thr 95 ) (Fig. 1A) upon binding of the signaling molecule. The structural changes in the active site upon binding of different ligands have been studied with different techniques (15,16,(22)(23)(24). For instance, by using visible excited RR spectroscopy in combination with site-directed mutagenesis, it is proposed that His 86 forms a hydrogen bond with a heme 6-propionate upon binding of O 2 to the heme (23). The formation of this hydrogen bond induces conformational changes of the protein by which Thr 95 is moved to a position suitable to form a hydrogen bond with the hemecoordinated O 2 , whereas Tyr 70 does not form a hydrogen bond with the heme-bound O 2 (22,23). In contrast, Zhang et al. (16) reported that mutation of Tyr 70 (Y70F, Y70L, and Y70W) of the truncated sensor domain of HemAT-Bs brought about larger dissociation constants for O 2 than the wild type (WT) and thus proposed that Tyr 70 would form a hydrogen bond to the hemebound O 2 in the WT form. These contradictory results demand directly probing the environmental change of Tyr 70 upon binding of O 2 because this might be substantially related to the discrimination of ligands to be signaled.
Another important issue is concerned with the structural changes in the protein moiety of the G-helix. The recent timeresolved RR study suggested that Tyr 133 in the G-helix (Fig. 1A) forms a hydrogen bond with the proximal axial ligand (His 123 ) upon CO binding (25). This is based on the 2-cm Ϫ1 downshift of the iron-histidine-stretching ( Fe-His ) band in the time-resolved RR spectra of WT hundreds of picoseconds after photodissociation of CO, whereas such a frequency shift was not observed for the Y133F mutant. Furthermore, G-and H-helices of different subunits of a homodimer form an antiparallel four-helix bundle (Fig. 1B) and exhibit appreciable displacement upon CN binding (21). Although this displacement is small, it is perceptible and may be related to the signal transduction (21). On the other hand, it is unknown whether such conformational changes occur upon binding of the effector molecule, O 2 .
To monitor the conformational changes in the protein moiety upon binding of O 2 , CO, and NO to a heme, we applied UVRR spectroscopy to WT and several mutants (Y70F, H86A, T95A, and Y133F) of the full-length HemAT-Bs for the first time. The vibrational spectra of the aromatic side chains such as Tyr and Trp residues can be selectively obtained by choosing an appropriate excitation wavelength in the 220 -250-nm region, and the spectra provide structural information about their conformations, local environments, and hydrogen bonding interactions (26,27). Such kinds of UVRR-specific information have proven to be essential to understanding the structural mechanisms of a variety of heme proteins, including hemoglobin (28), myoglobin (29), CooA (30), and E. coli DOS (31). Thus, UVRR spectra of HemAT-Bs are expected to probe directly the structural and/or environmental changes of Tyr 70 and Tyr 133 . Furthermore, UVRR spectroscopy will provide some information on the G-helix motion through side chain vibrations of Trp and Tyr residues because HemAT-Bs contains Trp 132 as well as Tyr 133 in the G-helix. Finally, we examined the effect of removal of the hydrogen bond between His 86 and heme 6-propionate and between Thr 95 and heme-bound O 2 on the protein conformational changes. As discussed below, the UVRR data will reveal specific features regarding the protein structural changes upon binding of different ligands.

EXPERIMENTAL PROCEDURES
Sample Preparation-In this study, we used the full-length HemAT-Bs with a His 6 tag at the C terminus, which was expressed with an E. coli BL21(DE3) cell system under the control of the T7 promoter in the pET-24(ϩ) vector (Novagen). Site-directed mutagenesis was carried out using a QuikChange site-directed mutagenesis kit (Stratagene). For the expression of HemAT-Bs, the E. coli cells were grown aerobically at 37°C for 4 h in Terrific Broth containing 30 g/ml kanamycin. The expression was induced by the addition of isopropyl ␤-D-thiogalactopyranoside to a final concentration of 1 mM, and then the cultivation was continued at 22°C for 18 h. The cells were harvested by centrifugation at 4,000 ϫ g and were stored at Ϫ80°C until used.
The cells were thawed and resuspended in Buffer A (50 mM Tris-HCl (pH 8.0) containing 15 mM glycine and 1 M NaCl) and then were broken by sonication. The resulting suspension was centrifuged at 100,000 ϫ g for 20 min, and the supernatant was loaded on a Ni 2ϩ -charged HiTrap chelating column (Amersham Biosciences). After the column was washed with Buffer A and then with 50 mM Tris-HCl (pH 8.0), the adsorbed proteins were eluted by 50 mM Tris-HCl (pH 8.0) containing 100 mM imidazole. The fractions containing HemAT-Bs were combined and loaded on a HiTrap Q HP column (Amersham Biosciences). The column was washed with 50 mM Tris-HCl (pH 8.0) containing 100 mM NaCl, and then HemAT-Bs was eluted by increasing the concentration of NaCl in the buffer.
For Raman measurements, the concentration of protein was adjusted to 80 M in 50 mM Tris-HCl (pH 7.5). As an internal intensity standard to generate UVRR difference spectra, 400 mM sodium perchlorate (NaClO 4 ) was contained in the solution for the excitation at 229 nm. In addition, Trp Raman bands were used as a secondary internal standard to normalize the UVRR spectra of different Tyr mutants.
The oxidized HemAT-Bs was prepared by adding an excess amount of potassium ferricyanide to the purified protein, and afterward potassium ferricyanide was removed by gel filtration on a Sephadex G-25 column. A 5-fold excess of sodium dithionite was added into degassed oxidized HemAT-Bs solution to reduce it. To prepare CO-, NO-, and O 2 -bound HemAT-Bs, the reduced HemAT-Bs solution was exposed to CO, NO, and O 2 gas, respectively.
Ultraviolet Resonance Raman Measurements-UVRR measurements were performed using instrumentation described previously in detail (32). The 229-nm excitation light was generated by an intracavity frequency doubling of the 457.9-nm line of an argon ion laser (Coherent, Innova, 300 FReD). The second harmonic in the laser output was separated from the fundamental with a Pellin-Broca prism and focused into a sample solution. An ϳ100-l aliquot of the protein solution was incorporated into a spinning cell with a stirring function (32), the inside of which was replaced with the corresponding gas. Raman-scattered light at a right angle was collected with a UV microscope objective lens, dispersed with a 126-cm single monochromator (Spex 1269) equipped with a 3600groove/mm holographic grating, and detected by an intensified charge-coupled detector (Princeton Instruments, ICCD-1024MG-E/1). We adopted spectral resolutions of 7.8 cm Ϫ1 for spectra. The laser power at the sample point was very low (0.3 milliwatts) to avoid photodissociation of the ligand from the heme. The protein sample was replaced with a fresh one every 10 min, and the total exposure time to get one spectrum was ϳ1 h. The integrity of the sample after exposure to UV laser light was carefully confirmed by comparing the visible absorption spectra obtained before and after the UVRR measurements. If some spectral changes were recognized, the Raman spectrum was discarded. Raman shifts were calibrated with cyclohexane, trichloroethylene, 1,2-dichloroethane, and toluene. Fig. 2 depicts the 229-nm excited raw UVRR spectrum of the O 2 -bound form of full-length WT HemAT-Bs (spectrum a) and the difference spectra, O 2 Ϫ ligand-free, for full-length (spectrum b) and truncated sensor domain proteins (spectrum c). The raw spectrum of full-length WT (spectrum a) is dominated by the bands arising from one Trp (Trp 132 ) residue and six Tyr residues (Tyr 13 , Tyr 49 , Tyr 70 , Tyr 133 , Tyr 148 , and Tyr 184 ), which are labeled with W and Y, respectively, followed by their mode numbers (26). The difference spectra were calculated so that the band of ClO 4 Ϫ (934 cm Ϫ1 ), which was present at the same concentration in all the samples as an internal intensity standard, could become zero. The presence of negative peaks in the difference spectra means that Raman intensities of the Trp and Tyr bands are reduced by oxygen binding.

UVRR Spectral Changes of the Sensor Domain and Fulllength Proteins of HemAT-Bs upon O 2 Binding-
The frequency of the W17 band (ϳ875 cm Ϫ1 ) is known to serve as a marker of hydrogen bonding of the Trp indole ring (33,34). The W17 band of WT (spectrum a) is observed at 876 cm Ϫ1 . This frequency corresponds to a moderate strength of hydrogen bonding (33,34). The crystal structure of the CNbound form of the truncated sensor domain in the homodimer suggests that Trp 132 does not form a hydrogen bond with any nearby residues (21). The W3 mode of Trp, which is known to be sensitive to the absolute value of torsion angle ͉ 2,1 ͉ about the C ␤ -C 3 bond connecting the indole ring to the peptide main chain (35), is observed at 1554 cm Ϫ1 for the full-length WT, which corresponds to a ͉ 2,1 ͉ angle of 102°. The x-ray structure of the truncated sensor domain indicated that Trp 132 in each monomer of the cyanide-bound form has 2,1 values of 111 and 127° (21). These values are larger than that predicted from the W3 frequency. The W17 and W3 vibrations of the truncated sensor domain are also observed at frequencies similar to those of the full-length protein (data not shown). These observations suggest that the hydrogen bonding interaction and side chain conformation of Trp 132 might be slightly different between the O 2 -and CN-bound forms.
Because of the importance of an effector molecule, O 2 , for HemAT-Bs, next we focused on the UVRR spectral changes upon binding of O 2 to the heme. We noted that the absorption spectrum of the O 2 -bound form showed no significant change after exposure to UV light, indicating that the O 2 -bound form was not oxidized under our experimental conditions. The O 2 Ϫ ligand-free difference spectra (Fig. 2,   Intensity comparison between spectra b and c indicates that the intensity change of Y8a is larger for the full-length protein than for the truncated sensor domain (see also supplemental Fig. S1). Because this difference spectrum (Fig. 2, spectrum b) reflects the spectral changes of Tyr residues upon O 2 binding and the truncated sensor domain (residues 10 -178) does not contain Tyr 184 , which is located in the linker region (residues 176 -195) between the sensor and signaling domains, the change in the Y8a band (spectrum b) is partially assigned to Tyr 184 . This suggests that an environmental change of Tyr 184 takes place, in fact, upon binding of O 2 to the heme. However, the overall similarity between spectra b and c means that the structural changes of Trp and Tyr residues in the sensor domain upon O 2 binding are hardly affected by the presence of the linker region.
UVRR Spectral Changes of HemAT-Bs upon Binding of Different Ligands- Fig. 3 compares the UVRR spectral differences among the O 2 -bound (spectrum b), CO-bound (spectrum c), and NO-bound (spectrum d) forms of the full-length WT, for which the 229-nm excited raw UVRR spectrum of the O 2 -bound form is presented again by spectrum a for reference. For clarification, spectra b-d in Fig. 3 are represented by the differences, ligand-bound Ϫ ligand-free reduced forms. Although CO and NO are not effector molecules for HemAT-Bs, we have monitored the spectral changes for Trp and Tyr residues of WT upon CO (Fig. 3, spectrum c) or NO (spectrum d) binding to elucidate whether or not the protein moiety can discriminate between different ligands. The difference spectrum c (CO Ϫ ligand-free) displays negative features for Trp and Tyr bands similar to those observed upon O 2 binding (spectrum b), but the features of the W18 and W16 modes are slightly different between CO and O 2 binding. In addition, spectrum c reveals stronger negative peaks for the Y9a (1174 cm Ϫ1 ), Y8b (1598 cm Ϫ1 ), and Y8a (1617 cm Ϫ1 ) bands than those in spectrum b, implying that the Raman intensity of some Tyr residues is more reduced in the CO than O 2 form. Furthermore, the difference spectrum d (NO Ϫ ligand-free) in Fig. 3 yields negative features for the Trp and Tyr bands similar to those observed upon O 2 binding (spectrum b). However, the features for the W3 and Y8a modes appear different between NO and O 2 . For instance, the W3 feature is stronger than Y8a in spectrum b, but it is reversed in spectrum d. Thus, the UVRR difference spectra indicate that the Raman bands of Trp and Tyr residues change in different ways upon binding of different ligand species (O 2 , CO, or NO). This suggests that the protein structural changes, including those of Trp 132 and some Tyr residues, are specific to a ligand species.
Spectral Changes of Tyr Residues Induced by O 2 Binding-The 229-nm excited UVRR spectra of Tyr mutants are shown in Fig. 4, where the raw spectra of the O 2 form of Y133F (spectrum a) and of Y70F (spectrum b) of full-length HemAT-Bs are displayed, whereas the difference spectra, O 2 Ϫ ligand-free, are also represented for Y133F (spectrum c) and Y70F (spectrum d).
The Trp and Tyr Raman bands in spectra a and b are located at frequencies similar to those of WT (Fig. 3, spectrum a). As shown in Fig. 4, spectrum c, the W18, W16, W7, and W3 bands of the Y133F mutant exhibit prominent negative features, implying the intensity decreases of the Trp bands in the O 2 -bound form are similar between WT (Fig. 3, spectrum b) and Y133F (Fig. 4, spectrum c). However, the intensities of the difference peaks of Y9a and Y8a are weaker for Y133F (Fig. 4, spectrum c) than for WT (Fig. 3, spectrum b). On the other hand, all Tyr negative peaks are abolished by mutation of Tyr 70 (Fig. 4, spectrum d). These results are clearly seen in the double difference spectra (spectra e and f in supplemental Fig. S2), demonstrating similarity between the spectra of WT and Y133F and some dissimilarity between the spectra of WT and Y70F with regard to Tyr bands. This would mean that binding of O 2 to heme induces conformational changes of the protein in both the distal (Tyr 70 ) and proximal (Tyr 133 ) sides of heme, but the former is much larger than the latter.
Spectral Changes of Tyr Residues Induced by CO Binding- Fig. 5 depicts the raw UVRR spectrum of the CO form of full-length WT excited at 229 nm (spectrum a) and the difference spectra, CO Ϫ ligand-free, for WT (spectrum b), Y70F (spectrum c), and Y133F (spectrum d). Spectrum b reveals negative peaks for Y9a, Y7a, Y8b, and Y8a bands, indicating that some Tyr residues undergo spectral changes upon CO binding. These negative peaks in the WT spectrum (spectrum b) became weaker in Y70F (spectrum c) and Y133F The contribution from Tyr 70 appears larger than that from Tyr 133 as shown by the double difference spectra (spectra e and f in supplemental Fig. S3). The larger contribution from Tyr 70 than from Tyr 133 , in addition to the absence of Trp negative peaks, except for the W3 mode, distinguishes between the COand O 2 -bound forms. These results also mean that Tyr 70 and Tyr 133 , in addition to Trp 132 in the sensor region, undergo spectral changes upon CO binding that are different from those seen for O 2 binding. (23) proposed that a hydrogen bond is formed between His 86 and heme 6-propionate only when O 2 is bound to the heme in HemAT-Bs. The formation of this hydrogen bond induces a conformational change of the CE loop and E-helix by which Thr 95 is moved to a position suitable to form a hydrogen bond with the heme-bound O 2 . The conformational changes around the heme upon O 2 binding would be propagated to the protein moiety within the sensor domain and afterward to the signaling domain. If this is the case, these hydrogen bonds must control the conformational changes in the protein moiety. To examine this idea, we performed UVRR measurements of H86A and T95A mutants, in which a hydrogen bond between His 86 and heme 6-propionate or between Thr 95 and heme-bound O 2 would be absent, respectively. The raw spectra of T95A and H86A are quite similar to those of WT (Fig. 3, spectrum a), implying that mutations of Thr 95 and His 86 incorporate no significant effect on the Raman frequencies of Trp and Tyr residues. The difference spectrum of T95A (Fig. 6, spectrum c) is different from that of WT (Fig. 3, spectrum b) regarding the negative features at W18, W17, W16, W7, W3, and Y8a; the difference peak intensities are much weaker in the T95A spectrum than in WT (see also supplemental Fig. S4).

Effects of Mutations of Thr 95 and His 86 on the O 2 -induced Spectral Changes-Yoshimura et al.
We reported previously for the O 2 -bound WT HemAT-Bs that there are three O 2 isotope-sensitive bands ( Fe-O 2 ) at 554, 566, and 572 cm Ϫ1 (22). In the T95A mutant, the band at 554 cm Ϫ1 , which is deduced to arise from the species with a hydrogen bond between the proximal oxygen atom of heme-bound O 2 and a protein residue, and the band at 566 cm Ϫ1 disappeared, and only the band at 572 cm Ϫ1 was observed (22). The frequency of 566 cm Ϫ1 implies the presence of a moderately strong hydrogen bond between the distal oxygen atom of hemebound O 2 and a protein residue similar to that in MbO 2 . The results suggest that the T95A mutant has a single conformation of the distal heme pocket, which would correspond to the open form with Fe-O 2 at 572 cm Ϫ1 (22). Thus, the small negative features in spectrum c (Fig. 6) are produced as a result of intensity reduction of the open form (572 cm Ϫ1 ) compared with that  of ligand-free form. Therefore, the removal of a hydrogen bond between Thr 95 and heme-bound O 2 significantly perturbs the conformation of Trp 132 , Tyr 133 , and Tyr 70 residues and thus may make the information transduction impossible.
On the other hand, the H86A spectrum (Fig. 6, spectrum d) displays negative peaks at the W18, Y1, W17, W16, W7, and W3 bands similar to those for WT (Fig. 3, spectrum b). However, the negative peak of Y8a in the WT spectrum (Fig. 3,  spectrum b) is absent in the H86A spectrum (Fig. 6, spectrum d).
In addition, the Y9a mode appears as a small derivative near 1174 cm Ϫ1 , implying that the Y9a band is shifted to a higher frequency in the O 2 -bound form.
In the H86A mutant, only the Fe-O 2 band at 566 cm Ϫ1 disappeared (23), implying that the H86A mutant yields two conformations corresponding to the open form with Fe-O 2 at 572 cm Ϫ1 and the closed form with Fe-O 2 at 554 cm Ϫ1 . Thus, the negative features in spectrum d (Fig. 6) are probably because of the existence of two conformations with Fe-O 2 at 572 and 557 cm Ϫ1 . Therefore, the removal of a hydrogen bond between His 86 and heme 6-propionate perturbs only the conformation of Tyr 70 and Tyr 133 residues, resulting in the disappearance of the 566 cm Ϫ1 species. Taking this result together with that of T95A, we conclude that the hydrogen bonds between His 86 and heme 6-propionate and between Thr 95 and heme-bound O 2 significantly influence the conformational states of Tyr 70 , Trp 132 , and Tyr 133 .

Protein Conformational Changes upon Binding of Different
Ligands-Full-length HemAT-Bs protein contains a single Trp (Trp 132 ) and six Tyr (Tyr 13 , Tyr 49 , Tyr 70 , Tyr 133 , Tyr 148 , and Tyr 184 ) residues. Trp 132 and Tyr 133 are located in the proximal side of heme but Tyr 70 in the distal side, whereas Tyr 49 , Tyr 148 , and Tyr 184 are far from the heme (Fig. 1A). Only Tyr 184 stays in the linker region between the sensor and the signaling domains. These aromatic residues are used as probes to monitor the protein structure in the present UVRR spectroscopy. The UVRR results demonstrate that Raman intensities of Trp 132 decrease for the O 2 -, CO-, and NO-bound forms in different ways. Generally, the Raman intensity of Trp is sensitive to environmental hydrophobicity and/or hydrogen bonding interactions of the indole side chain (28,36,37). Because a frequency shift is not observed for the hydrogen bond marker band (876 cm Ϫ1 ), a moderate strength hydrogen bonding between Trp 132 and nearby residues or a water molecule is hardly altered by ligand binding. Thus, it is reasonable to conclude that the environment around Trp 132 shifts to more hydrophilic in the O 2 -, CO-, and NO-bound forms.
Generally for proteins, the W7 band appears as a doublet at ϳ1358/1340 cm Ϫ1 in the raw spectrum (see Fig. 2, spectrum a). This doublet arises from Fermi resonance between the N 1 -C 8 stretching fundamental (W7 mode) and the combination of out-of-plane bending vibrations (26). The intensity ratio of the W7 doublet serves as a marker of the hydrophobicity of the environment around the indole ring; the intensity ratio of the W7 doublet decreases as the environment around the Trp residue shifts to more hydrophilic (33). The difference spectrum, O 2 Ϫ ligand-free, reveals a negative peak at 1358 cm Ϫ1 (Fig. 2, spectrum b), indicating the intensity reduction of the 1358-cm Ϫ1 component of the W7 band, and hence the intensity ratio of the W7 doublet, is decreased by O 2 binding. This observation confirms that the environment around Trp 132 shifts to more hydrophilic in the O 2 -bound form.
The UVRR bands of Tyr residues also exhibited an intensity decrease upon binding of ligands in a manner specific to a ligand species, and these changes were assigned to Tyr 70 and Tyr 133 (Figs. 4 and 5). These spectral changes are not compatible with deprotonation or formation of a strong hydrogen bond of the phenol side chain. Our previous RR study with visible excitation (22,23) suggested that Tyr 70 does not form a hydrogen bond with heme-coordinated O 2 . However, Zhang et al. (16) proposed that Tyr 70 would form a hydrogen bond to the heme-bound O 2 on the basis of the observation that mutation of Tyr 70 gave a larger dissociation constant for the hemebound O 2 than that for WT. On the other hand, Yoshimura et al. (25) recently suggested that Tyr 133 forms a hydrogen bond with the proximal ligand (His 123 ) in the CO-bound form. This proposal is based on a 2-cm Ϫ1 downshift of the Fe-His band in time-resolved RR spectra of WT hundreds of picoseconds after the photodissociation of CO. This frequency shift did not occur in that of the Y133F mutant. However, the size of the downshift of Fe-His suggests that a very weak hydrogen bond (38) is present. It was clarified from the model compound studies that the intensity of a Tyr band is influenced by the hydrophobicity of its surroundings (37). On the other hand, the formation or breaking of a hydrogen bond between the hydroxy side chain of Tyr and any surrounding residue is expected to influence significantly both the intensity and frequency of Tyr bands (28,31,39,40). Because UVRR results of WT, Y70F, and Y133F showed no significant frequency shift for any of Tyr Raman bands, all Tyr residues, including Tyr 70 and Tyr 133 , do not undergo a change of hydrogen bonds. Only the intensities of Tyr 70 and Tyr 133 are altered upon ligand binding; we concluded that the environment around Tyr 70 and Tyr 133 shifts to more hydrophilic in the ligand-bound forms without a change in their hydrogen bonding interactions. These results confirmed our previous model (22,23), which suggested that Tyr 70 does not form a hydrogen bond, but Thr 95 does with heme-coordinated O 2 . Presumably, the change of hydrophobicity around Tyr 133 would be associated with the observed Fe-His shift.
Both the full-length and truncated sensor domain proteins of HemAT-Bs revealed similar spectral changes for Trp and Tyr residues upon O 2 binding (Fig. 2). Although the main contributors to Tyr difference peaks arise from Tyr 70 and Tyr 133 , Tyr 184 in the linker region exhibits a small spectral change (Y8a band) upon ligand binding, and the magnitude of its change seems to depend on a ligand species (Fig. 3). This implies that conformational change of the sensor domain upon ligand binding, specific to a ligand species, is communicated to the signaling domain through the linker region. One of the possible explanations for the small spectral change of Tyr 184 is that this residue may be exposed to solvent, and binding of O 2 will cause only small changes in its environment. However, the main changes in UVRR spectra are attributed to environmental changes of Trp 132 , Tyr 70 , and Tyr 133 in the heme neighborhood.
The previous RR results showed that Thr 95 forms a hydrogen bond to the heme-bound O 2 of HemAT-Bs (22,23). In contrast, the RR and Fourier transform infrared results indicated the absence of the hydrogen bonding interaction between the heme-coordinated CO and Thr 95 (23,24). Therefore, it is proposed that the hydrogen bonding interactions of the hemebound O 2 with the surrounding residues such as Thr 95 are crucial for ligand discrimination in HemAT-Bs (23). Such a specific hydrogen bonding between a heme-bound O 2 and a protein residue has been also observed for other O 2 sensor heme proteins such as FixL and E. coli DOS (4,5). The present results in Fig. 3 are compatible with the idea that the heme of HemAT-Bs discriminates between O 2 and CO, resulting in different conformational changes in the protein moiety of HemAT-Bs. In addition, binding of O 2 , CO, or NO to a five-coordinate, highspin, ligand-free (reduced) form changes the spin state of the heme iron of HemAT-Bs (15). Thus, the observed structural changes upon O 2 binding are mainly due to conversion of the high-spin to low-spin state and the formation of hydrogen bonds between O 2 and Thr 95 , whereas those observed upon CO or NO binding are probably due to only the change in the spin state of heme iron.
Insights into Signal Transduction Mechanism of HemAT-Bs-A conformational change specific to O 2 binding would occur in HemAT in the heme neighborhood, and then intramolecular signal transduction would take place from the sensor domain to the signaling domain through the linker region. As a result, the self-kinase activity of CheA is regulated through the HemAT-CheA interaction by ligand binding to heme. This signaling event takes place only with O 2 but not with other gas molecules that bind to the heme. Yoshimura et al. (23) proposed that a hydrogen bond is formed between His 86 and heme 6-propionate only when O 2 is bound to the heme. This proposal seems likely because the formation of a hydrogen bond induces conformational change in the heme distal side by which Thr 95 is displaced to a proper position to form a hydrogen bond with the bound O 2 . In the present study, we have investigated the role of these hydrogen bonds and found that removal of these hydrogen bonds in the H86A and T95A mutants strongly perturbs the conformational changes of Tyr 70 (B-helix), Tyr 133 , and Trp 132 (G-helix) in both the distal and proximal sides of the heme. Recently, we have shown that the heme structural changes upon ligand binding in myoglobin are communicated to the globin through heme propionates (29). This is based on the observation that mutations of the nearby residues or chemical modification of the propionate side chain of heme significantly changes the conformation of Trp residues in the A-helix of the globin, which are far from the heme in myoglobin. Similarly, the present UVRR results suggest that hydrogen bonds between Thr 95 and heme-bound O 2 and between His 86 and heme 6-propionate in HemAT-Bs communicate the structural changes of heme to the protein moiety (B-and G-helices) upon O 2 binding.
Furthermore, the present UVRR results showed for the first time that Trp 132 and Tyr 133 on the G-helix in the proximal side of heme undergo a change in hydrophobicity upon O 2 or CO binding. Recently, Pinakoulaki et al. (41) pointed out the presence of a ligand cavity in the protein from the Fourier transform infrared experiments for the heme-bound CO of HemAT-Bs and suggested that the docked CO is interacting with Tyr 133 . This observation is compatible with different conformational changes of G-helix residues (Trp 132 and Tyr 133 ) upon binding of different ligands. The conformational changes of G-helix residues are quite important for the signal transduction mechanism of HemAT-Bs, as discussed below. The signal transduction mechanism has been proposed for MCPs, which are membrane-bound proteins (42,43). A pair of two antiparallel helices in a MCP dimer forms a transmembrane four-helix bundle that connects the periplasmic sensor domain and the cytoplasmic signaling domain. The binding of the effector molecule to the sensor domain induces a slide and/or rotational movement of this helix bundle, which is a key step in the signal transduction mechanism of MCPs (42,43).
HemAT-Bs is a member of the MCP family. Although HemAT-Bs lacks the transmembrane region, the G-and H-helices of the two subunits of the homodimer form an antiparallel four-helix bundle (Fig. 1B) and exist in the C-terminal region of the sensor domain, and the H-helix is apparently continuous to the extended helical structure of the signaling domain. Given that HemAT-Bs adopts the same mechanism for intramolecular signal transduction as membrane-bound MCPs, the helix bundle consisting of the G-and H-helices will correspond to the transmembrane helix bundle in MCPs. In fact, x-ray analysis of the truncated HemAT-Bs has demonstrated that both G-and H-helices experience displacement upon CN binding (21), although CN is not bound to the reduced form. If similar displacements of the four-helix bundle occur upon binding of O 2 , it would trigger the transduction of the conformational changes from the sensor domain to the signaling domain.
Furthermore, the truncated sensor domain of HemAT-Bs maintains classic globin folds (21). In myoglobin, binding of an external ligand to the heme induces significant displacement of the F-and E-helices (44). In addition, the protein structural changes are propagated from the heme to the C-terminal region of the H-helix upon ligand binding through the structural change in the proximal side of the heme (29). The x-ray structure of the truncated HemAT-Bs has shown that the F-helix suffers displacement upon CN binding (21). In addition, Yoshimura et al. (23) proposed that the E-helix experiences considerable conformational changes upon O 2 binding. On the other hand, further study is necessary to monitor the structural changes in the H-helix of HemAT-Bs upon O 2 binding.
In conclusion, the present UVRR study of the ligand-bound (O 2 , CO, and NO) forms of WT and several mutants of HemAT-Bs has brought new insights into the communication pathway between the heme and the protein moiety. More concretely, the hydrogen bonds between heme-bound O 2 and Thr 95 and between His 86 and heme 6-propionate are considered to communicate the heme structural changes upon O 2 binding to the protein moiety. Furthermore, the UVRR results suggest that the G-helix experiences different amounts of displacement upon ligand binding depending on the ligand species, which would be important for communication of the ligand-specific conformational changes from the sensor domain to the signaling domain.
We have demonstrated that UVRR spectroscopy is a powerful tool to elucidate fine structural changes of protein associated with the binding of ligand to the heme. The method can be applied in a time-resolved mode to monitor dynamic changes in the protein moiety upon photodissociation of the gaseous ligand using the pump-probe (nanoseconds to microsecond) technique. A time-resolved UVRR study of HemAT-Bs will be our next project.