Iron Coordination Structures of Oxygen Sensor FixL Characterized by Fe K-edge Extended X-ray Absorption Fine Structure and Resonance Raman Spectroscopy*

FixL is a heme-based O2 sensor protein involved in a two-component system of a symbiotic bacterium. In the present study, the iron coordination structure in the heme domain of Rhizobium meliloti FixLT (RmFixLT, a soluble truncated FixL) was examined using Fe K-edge extended x-ray absorption fine structure (EXAFS) and resonance Raman spectroscopic techniques. In the EXAFS analyses, the interatomic distances and angles of the Fe-ligand bond and the iron displacement from the heme plane were obtained for RmFixLT in the Fe2+, Fe2+O2, Fe2+CO, Fe3+, Fe3+F−, and Fe3+CN− states. An apparent correlation was found between the heme-nitrogen (proximal His-194) distance in the heme domain and the phosphorylation activity of the histidine kinase domain. Comparison of the Fe-CO coordination geometry between RmFixLT and RmFixLH (heme domain of RmFixL), based on the EXAFS and Raman results, has suggested that the kinase domain directly or indirectly influences steric interaction between the iron-bound ligand and the heme pocket. Referring to the crystal structure of the heme domain ofBradyrhizobium japonicum FixL (Gong, W., Hao, B., Mansy, S. S., Gonzalez, G., Gilles-Gonzalez, M. A., and Chan, M. K. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 15177–15182), we discussed details of the iron coordination structure of RmFixLT and RmFixLH in relation to an intramolecular signal transduction mechanism in its O2 sensing.

The soil bacterium Rhizobium meliloti establishes a symbiotic association with alfalfa, which facilitates nitrogen fixation by nitrogenase enzymes in plant root nodules (1)(2)(3). A requirement for nitrogen fixation is a low O 2 tension in the nodules, since the nitrogenase is readily deactivated by O 2 (4). In response to O 2 concentration, the transcription of nitrogen fixation genes (nif and fix) in the root nodules (5) is mediated by two proteins, FixL and FixJ (6); FixL senses the O 2 concentration in the nodules, transmits the information to FixJ via phosphorylation with ATP, and the phosphorylated FixJ promotes the expression of the nitrogen fixation genes as a transcriptional activator (7,8). The FixL/FixJ system is one of the so-called two-component regulatory systems, which are ubiquitous signal transduction devices and respond to environmental stimuli in prokaryotic cells and even in some eukaryotic cells (9 -13).
The O 2 sensor protein FixL, the transmitter molecule of the two-component system (14,15), consists of two different functional domains (16 -18), sensor and histidine kinase domains. Since the O 2 sensor domain of FixL contains a heme as a prosthetic group, FixL is a member of heme-containing proteins (hemoproteins). The iron of the heme domain in FixL is in an equilibrium state between the O 2 -bound (oxy) and the O 2unbound (deoxy) states, in response to the O 2 concentration (19). However, compared with other O 2 -binding hemoproteins, FixL is quite unique, since O 2 ligation to the heme iron is directly coupled with the switching of the kinase activity in itself. The kinase is active when the heme domain is in the deoxy state, whereas it is deactivated when O 2 is bound to the heme iron (6). As a result, the heme-based O 2 sensor FixL can be classified into a new class of hemoproteins (19).
For the case of FixL, the signaling for the O 2 association to/dissociation from the heme iron is transferred to its histidine kinase domain (20). Possibly, structural changes in the vicinity of the heme, caused by transition between the deoxy and oxy states, constitutes an initial event in O 2 sensing, which is then followed by the intramolecular signal transduction from the heme to the histidine kinase site, thus regulating kinase activity. In order to understand better the O 2 -sensing mechanism in FixL, the structural changes of the heme environment induced by O 2 dissociation have been examined using bio-and physicochemical techniques. Recently, resonance Raman (21,22), NMR (23), and ESR (24,25) spectroscopic studies have been applied to the characterization of the heme environmental structure of FixL, and kinetic measurements (19,20,24) have been used to determine its ligand binding properties as well. These studies have shown that the basic structure of the iron coordination is similar to that of hemoproteins such as myoglobin (Mb), 1 but the environment around the iron sixth coordination site is different.
However, these studies have yielded little direct structural information on the heme site of FixL resulting from the association or dissociation of O 2 . The relevance of the heme site coordination structure to the kinase activation is still not clear. Gilles-Gonzalez and co-workers (6,20) have measured the ki-nase activity of FixL in several iron oxidation, ligation, or spin states and found that the kinase domain is inactive in the low spin state of the heme iron, whereas it is active in the high spin state. Based on these observations, they proposed "the spin state" hypothesis, where activation/deactivation of the kinase site could occur through changes in the spin state of the heme iron. Most recently, they reported the crystal structure of the heme domain of Bradyrhizobium japonicum FixL (BjFixLH) in the ferric (Fe 3ϩ ) and ferric-cyanide (Fe 3ϩ CN Ϫ ) forms at 2.4 and 2.7 Å resolution, respectively. Based on structural comparison between these two states, it was suggested that the heme doming/flattening upon the ligand dissociation/association results in rearrangement of the hydrogen bond network between the heme 6,7-propionates and the "regulatory loop" and is possibly relevant to the activation/deactivation of the kinase domain (26).
In the present study, we report Fe K-edge EXAFS (extended x-ray absorption fine structure) of FixL in several iron states, and we characterize its iron coordination structures in active and inactive forms of the kinase. The EXAFS provides structural information that is limited to the metal-ligand coordination site (bond distance and angle) but is more precise than crystallographic data. To support the EXAFS information, we also measured the resonance Raman spectra of the CO adduct of FixL. Based on these spectroscopic results, we examined the change in the iron coordination structure of FixL as this relates to the activation/deactivation mechanism of the kinase domain in FixL.

MATERIALS AND METHODS
Sample Preparation-A soluble truncated FixL (RmFixLT, monomer molecular mass of 43 kDa) and the heme-containing domain of FixL (RmFixLH, molecular mass of 17 kDa) of R. meliloti were overexpressed in Escherichia coli strain JM109. RmFixJ (monomer molecular mass of 23 kDa) was co-expressed with RmFixLT in the same system. Expression plasmid for RmFixLT was generated by subcloning of the BamHI-HindIII fragment of the fixLJ genes (27) with the EcoRI-BamHI adapter into pUC18 at the EcoRI and HindIII sites. The N-terminal amino acid sequence of the soluble FixL is MTMITNSGSV 131 . Rm-FixLH is the heme-binding domain of FixL, which lacks the transmembrane and the kinase domains. We constructed the expression system of RmFixLH in E. coli using the fixL DNA as follows. The XhoI and XbaI sites were filled in with Klenow fragment and ligated to introduce the stop codon (L260D-stop) after removal of the kinase domain. The filled-in RsrII-KpnI DNA fragment encoding the heme domain and FixJ was joined to the filled-in NheI-KpnI fragment of pRSET vector to generate the 6ϫ His-tagged N-terminal sequence, MRGSH 6 GMART 128 .
The purification procedure of RmFixLH was described elsewhere (28). The purity of the protein was checked by SDS-polyacrylamide gel electrophoresis. The ratio of absorbances at 280 and 416 nm (Soret band), A 280 nm /A Soret , in the oxy form was 0.79 for the RmFixLT⅐RmFixJ complex and 0.56 for the RmFixJ-free RmFixLT, showing that the heme is not largely removed from the protein by the gel filtration. Horse heart myoglobin in the lyophilized form (Sigma) was dissolved in 5 mM potassium phosphate buffer (pH 6.0) and then applied to the cation exchange column (CM52; Whatman) equilibrated with the same buffer and developed with the gradient of the 5 mM phosphate (pH 6.0) and 50 mM K 2 HPO 4 .
Dynamic Light Scattering Measurements-Dynamic light scattering of the protein was measured using DynaPro-MS (Protein Solutions). For this measurement, a beam of monochromatic light was directed into the sample to monitor fluctuation of light intensity scattered by the molecule. From analyses of the data, the translational diffusion coefficient, DT (nm 2 /s), of the protein particle in solution was obtained. Assuming Brownian motion, this coefficient was converted to the Hydrodynamic Radius, R H , of the protein particle using the Stokes-Einstein Equation 1, where k b (J/K) represents Boltzman's constants, T (K) the absolute temperature in Kelvin, and (Pa⅐s) the solvent viscosity. The hydrodynamic molecular weight of the protein particle can be estimated from the R H . The sample condition for the dynamic light scattering measurement was 10 M (RmFixLT) 2 (RmFixJ) 2 in 50 mM Tris/HCl (pH 8.5), 10% glycerol, 10% ␤-mercaptoethanol, and 0.2 M NaCl at 20°C. In the measurement of the ferrous deoxy, ferrous O 2 , and ferrous CO complexes, ␤-mercaptoethanol was present in the system. EXAFS Measurements and Analyses-For the Fe K-edge EXAFS measurement, we prepared the ferrous deoxy, ferrous CO, ferrous O 2 (oxy), ferric (met), ferric fluoride (metF Ϫ ), and ferric cyanide (metCN Ϫ ) forms of the (RmFixLT) 2 (RmFixJ) 2 complex. The iron concentration of the samples was less than 1 mM for RmFixLT, because it precipitated when RmFixLT was concentrated over 1 mM. In addition, the ferrous CO complexes of RmFixLH (2 mM) and the O 2 , deoxy, and CO complexes of Mb (8 mM) were also prepared. The samples were frozen in liquid nitrogen in Perspex EXAFS cells with Mylar windows.
The EXAFS measurements were carried out using synchrotron radiation at Station 8.1 of the synchrotron radiation source of Daresbury Laboratory (UK) and at BL12C of Photon Factory of KEK (Japan) (29,30). All measurements were carried out at liquid nitrogen temperature in the fluorescence mode using multi-element germanium detector.
Each EXAFS scan took approximately 1 h. Due to the low concentration of iron in the samples, it took 20 -24 h to obtain good quality EXAFS data for the RmFixLT complexes, by averaging. After calibration of monochromator position to energy, the EXAFS was normalized to a unit iron atom and extracted from the background absorption using the Daresbury Laboratory program EXBACK (31). The EXAFS data were converted into k space using k ϭ (2 m e (E Ϫ E 0 )/h 2 )] 1/2 , where E and E 0 are the energy of the incident x-ray radiation and the absorption edge of the iron atom, respectively, and k is the photoelectron wave vector. The fitting of the data was performed using the non-linear least squares program EXCURV92 (32), which calculates the theoretical EXAFS function using fast curved wave theory and incorporates multiple scattering up to 3rd order (33,34). Phase shifts were calculated using the Hedin-Lundqvist approximation (35). Curve fitting was carried out in k space on raw EXAFS data weighted by k 3 to compensate for the diminishing amplitude of the EXAFS at high k.
Constrained and restrained refinement procedures were used to minimize the number of free parameters in the least squares analysis (36,37). The quality of the simulations was assessed by the least squares fit index and R factor (37). The most important information obtained by the fitting procedures were the radial distances of the inner shell ligands in the iron coordination sphere (Fe-N pyr , Fe-N im , Fe-ligand, and Fe-Ct), where the d(Fe-Ct) represents the distance of the iron from the center of the pyrrole plane (iron displacement). The errors in the structural parameters deduced from EXAFS analysis arise from the data collection (e.g. photon counting statistics or sample inhomogeneity) and the subsequent data analysis (e.g. uncertainties in the EXAFS theory used or the quality of the background subtraction). For the FixL and Mb spectra, estimates of the errors are Ϯ0.02 Å for the first shell distances of the N pyr and the sixth ligand atom and Ϯ0.03 Å for the histidine ligand. The Debye-Waller factors for atoms belonging to the same ligand were found to increase systematically with their distance from the iron atom.
Resonance Raman Spectral Measurements-The resonance Raman spectra were measured with a single spectrophotometer (JASCO NR-1800) equipped with a cooled charge-coupled device (Princeton Instru-ments). The excitation sources were Kr ϩ laser (Coherent) at 406.7 and 413.1 nm. The Raman cell was spinning and kept below 10°C by flushing with cold N 2 gas.

Solution States of FixL and FixL-FixJ Complex
Prior to Fe K-edge EXAFS measurements, the solution states of RmFixLT were checked using the dynamic light scattering technique. Fig. 1 shows the results for RmFixLT in the met (Fe 3ϩ ) state in the presence or absence of RmFixJ, which show the distribution of R H (hydrodynamic radius) of the protein in the solution. This measurement indicated that R H of RmFixLT (RmFixJ-free form; bold line in Fig. 1) is widely distributed and that the average molecular mass estimated from the R H value is about 300 kDa, suggesting that RmFixLT is highly and randomly aggregated in the solution, since the molecular mass of monomeric RmFixLT is 43 kDa. On the other hand, the RmFixLT⅐RmFixJ complex gives the result shown in Fig. 1 (thin line), in which the distribution of the R H is narrow and the average molecular mass is 130 kDa. The same results were obtained for the metF Ϫ , deoxy, oxy (Fe 2ϩ O 2 ), and CO (Fe 2ϩ CO) forms of RmFixLT (data not shown). These results suggest that the RmFixLT⅐RmFixJ complex is monodispersive in the solution and is most probably in the heterotetrameric form, (RmFixLT) 2 (RmFixJ) 2 , since the monomer molecular mass of RmFixJ is 23 kDa.
The (RmFixLT) 2 (RmFixJ) 2 in the met, metF Ϫ , oxy, and CO states are stable at 10°C, since their dynamic light scattering patterns were unchanged for several hours after the purification. The deoxy form was stable in the tetrameric form for 30 min after the fresh preparation under anaerobic condition but gradually aggregated afterward. Therefore, we prepared the EXAFS samples immediately after the preparations.

Fe K-edge EXAFS of RmFixLT
In view of the results of the light scattering measurements, EXAFS studies were carried out on the (RmFixLT) 2 (RmFixJ) 2 complex to investigate the iron coordination structure of Rm-FixLT, since the (RmFixLT) 2 (RmFixJ) 2 complex is apparently more intact than the aggregated RmFixLT. Fig. 2a shows the EXAFS data and their Fourier transforms for RmFixLTs in the oxy, CO, and deoxy forms. Also shown are the corresponding data for horse heart myoglobin. The clear similarity between these spectra for each pair of samples implies a general structural equivalence between their iron coordination sites. From this observation it was anticipated, for example, that the iron atom in deoxy RmFixLT is displaced from the porphyrin plane, as is the case for deoxy Mb (38). The detailed curve-fitting of the deoxy RmFixLT EXAFS confirmed this observation.
We have also examined the Fe K-edge EXAFS for the met, metF Ϫ , and metCN Ϫ complexes (Fig. 2b). The strategy used in  the data analysis was to adopt the relevant model for the iron coordination site for the RmFixLT samples that is known, by crystallography (39), to hold true for Mb. The three-dimensional crystallographic information on Mb was simplified, and an averaged ("idealized") structure of the porphyrin ring, the sixth ligand (CO, O 2 ), and the imidazole ring of the histidine ligand were constructed. The application of restraints guaranteed the structural integrity of these ligands throughout the fitting procedure and also ensured that the refinements remained overdetermined. These features of the data analysis method are described in detail elsewhere (37) for the EXAFS of fetal deoxy Mb. The results obtained by this approach are compiled in Table I, top and middle parts.
Fifth Ligand His-194 Imidazole Coordination-In our EX-AFS analyses of RmFixLT, the best fit of the theoretical curves to the raw data was obtained in all cases when the imidazole ring was placed at the iron axial position in all states, indicating coordination of the histidyl imidazole to the heme iron of RmFixLT as the fifth axial ligand. The result agrees with the previous resonance Raman and NMR spectroscopic results (23,25) and confirms initial structural expectations based on similarity with the EXAFS of Mb. In the mutagenesis study, replacement of His-194 with other amino acid residues lost the heme in RmFixL (16,27), indicating that His-194 is a proximal ligand of RmFixL (proximal His-194). Indeed, the recent crystallographic study of BjFixLH (26) (Table I, bottom part) indicated that the Fe-N(His-194) bond length in deoxy RmFixLT (2.11Å) was the same as that of the Fe-N(His-93) bond in deoxy Mb (2.11Å). On the other hand, the resonance Raman study indicated that the Fe-N(His-194) stretching of deoxy RmFixLT (209 cm Ϫ1 ) is lower than the Fe-N(His-93) stretching frequency of deoxy Mb (220 cm Ϫ1 ) and therefore suggests that the Fe-N(His-194) bond in deoxy RmFixLT is weaker than the Fe-N(His-93) of deoxy Mb (25). One possible explanation of the apparent discrepancy between the EXAFS and the resonance Raman results is distortion of the Fe-imidazole bond in RmFixLT, compared with that in Mb. Bond distortion causes different bonding-antibonding interactions between the imidazole nitrogen p-orbital and the iron d-orbitals, resulting in the differences in the strength of the Fe-N(His) bond between FixL and Mb.
Iron Displacement from Heme Plane-The iron displacement relative to the heme plane, d(Fe-Ct), is also dependent on the iron spin states; in the oxy, CO, and metCN Ϫ forms (low spin state), the iron is located in the heme plane (in-plane configuration), whereas it is displaced by 0.44 Ϯ 0.06 and 0.55 Ϯ 0.06 Å from the heme plane (out-of-plane configuration) in the met and deoxy forms (high spin state), respectively. The results are consistent with the general properties of hemoproteins in that the high spin iron is of considerably larger covalent radius than the low spin iron (40), thus making it difficult to force the high spin iron into the rigid porphyrin plane.
Sixth Ligand Coordination-The EXAFS analysis clearly showed that the iron sixth coordination site in the met form of RmFixLT was vacant (5-coordination), which is in good agreement with previous conclusions based on optical absorption and the resonance Raman measurements (19,22). Support for the 5-coordination of the met iron in RmFixLT was also provided in the iron near-edge spectra (41), in which the magnitude of the pre-edge for RmFixLT was comparably larger than that for six-coordinated metaquo Mb (data not shown).
The Fe-C-n ϭ 163°for the metCN Ϫ form. These values fall into the allowable range of the Fe-ligand bond distances and angles, which were determined for the corresponding complexes of some hemoproteins by crystallographic and EXAFS analyses (42).
Fluoride Coordination to Ferric Iron-We also compared the F Ϫ binding to met RmFixLT with those to other met hemoproteins examined thus far. However, the structural analysis of the metF Ϫ complexes of hemoproteins is rare and controversial. Aime et al. (43) and Bolognesi et al. (44) reported the crystal structures of the metF Ϫ complexes of sperm whale Mb and Aplysia Mb, respectively. For the case of sperm whale Mb, the Fe-F Ϫ and the Fe-N im bond lengths were reported to be 2.23 and 2.47 Å, respectively, and those for Aplysia Mb to be 2.21 and 2.26Å, respectively. In these reports, the d(Fe-N im ) in these Mb complexes are the longest among those for hemoprotein complexes reported thus far, and the value of 2.47 Å is too distant to have a strong interaction. In addition, it is also puzzling that the iron in the metF Ϫ complex of Aplysia Mb has been reported to be located in the heme plane (in-plane configuration), despite the high spin heme iron. On the other hand, Deatherage et al. (45) reported the structural change of metHb induced by the F Ϫ binding based on x-ray difference Fourier technique, which showed that the F Ϫ binding to the met iron as a sixth ligand causes a small movement of the iron toward the porphyrin plane.
Powers et al. (  I Final refined parameters obtained from restrained refinement of EXAFS The abbreviations used are: Fe-Ct, magnitude of iron displacement, distance between iron and the porphyrin plane center; Fe-N pyr , bond distance between iron and porphyrin pyrrol nitrogen atom; Fe-N im , bond distance between iron and proximal His imidazole nitrogen atom; Fe-L 1 , bond distance between iron and sixth ligand atom; Fe-L 1 -L 2 , bond angle between iron and sixth ligand. Values in parentheses are Deby-Waller factors of the ligand atom coordinating iron.

Changes of Iron Coordination Structure of FixL upon Ligand Binding
The present EXAFS study clearly indicates that the iron coordination structure of the O 2 -bound heme in RmFixLT, in which the kinase domain is inactive, is significantly different from that of the deoxy form, which contains the active kinase domain. This structural difference is comparable to the pattern observed for oxy and deoxy Mb (38) and for oxy (R state) and deoxy (T state) hemoglobin (Hb). In the case of Hb, the iron moves up to 0.3 Å in the R-T conversion (46,47). Thus, comparing the EXAFS data (Table I, top part), we found two changes of the iron coordination structures between the oxy and the deoxy states. One is an iron displacement (Fe-Ct) from the heme plane by 0.55 Å, which accompanies an increase in the Fe-N pyr length from 2.01 to 2.07Å. The similarity between deoxy Mb and deoxy RmFixLT EXAFS spectra already suggested this similarity in structure. Another similarity is an increase in the Fe-N im bond length from 2.01 to 2.11 Å. As a consequence, the distance between the heme plane and the nitrogen atom of the proximal His-194 imidazole, d(Ct-N im ) ϭ d(Fe-Ct) ϩ d(Fe-N im ), is increased by 0.63 Å upon the dissociation of O 2 from the ferrous heme iron.
It has been reported that the kinase of FixL is inactive in the metCN Ϫ state of its heme domain, although active in the met and the metF Ϫ states (20). It was also shown, without any raw data, that CO inhibits the kinase activity of ferrous FixL. Therefore, we also examined the d(Ct-N im ) of these complexes of RmFixLT. In the met form, the d(Ct-N im ) is 2.55 Å, compa-rable to that of the kinase-active deoxy form (2.66 Å). The d(Ct-N im ) of the FixL in the metF Ϫ state is 2.31 Å. On the other hand, the d(Ct-N im ) of the metCN Ϫ and the CO complexes are 2.06 and 2.08 Å, respectively, which are similar to that of the kinase-inactive oxy (2.03 Å) forms. In Fig. 3, the kinase activities of the RmFixLT complexes are plotted against the d(Ct-N im ) estimated in the present study. Inspection of this figure clearly shows that a threshold is present in the region of 2.1-2.3 Å of the d(Ct-N im ); the RmFixLT complexes having a d(Ct-N im ) longer than the threshold contains active kinase, whereas the kinase domain of the RmFixLT complex with the shorter d(Ct-N im ) is inactive. The result indicates that the iron displacement relative to the heme plane is apparently correlated to the kinase activation/deactivation.
In hemoproteins, it is a general feature that the iron moves out of the heme plane in conversion from the low to the high spin state, due to the increment of the iron covalent radius. Perutz (46,47) focused on this iron movement induced by the O 2 dissociation from hemoproteins and proposed a well known mechanism in the R-T transition of Hb, i.e. the "trigger" model, where the movement of the proximal histidine (fifth ligand) relative to the heme plane upon the dissociation of O 2 from one Hb subunit acts as a trigger to induce the conformational change of the Hb tertiary and quaternary structure, eventually controlling the O 2 affinity of the other subunits. The present EXAFS results on deoxy and oxy RmFixLT suggest that the proximal trigger model might be applicable to FixL; the movement of the imidazole group of His-194 relative to the heme plane upon the association/dissociation of O 2 is related to the deactivation/activation of its histidine kinase domain.
However, recent crystallographic study of BjFixLH (26) suggested that the Hb-type mechanism is not operative in the O 2 sensor FixL, because of no structural change in the heme proximal side upon the CN Ϫ association to the ferric iron. Instead, another mechanism was proposed, in which rearrangement of the hydrogen bond pattern of the heme 7-propionate with its surroundings, caused by the heme flattening upon the ligand (CN Ϫ ) association, is possibly important in the signal transduction to the kinase domain. The present EXAFS data of RmFixLT appears to be consistent with the proposal based on the crystal structures of BjFixLH, since the heme flattening/doming is closely associated with the iron movement relative to the heme plane. These results support the idea that ligand binding could act as a trigger in the intramolecular signal transition in FixL through the heme flattening/doming.  (19). The activity value of the CO complex has not been reported.

Structural Characterization of Heme Pocket in Sixth Ligand (Distal) Side
In the heme-based O 2 sensor FixL, the O 2 binding to the heme sixth site is a signal that induces kinase deactivation (6). Therefore, the ligand (O 2 , CO, NO, CN Ϫ , and F Ϫ ) binding properties of FixL and property of the autoxidation of its oxy complex have been extensively studied via kinetic and spectroscopic techniques and have been discussed in relation to the structure around the iron sixth coordination side (the distal heme pocket) (6, 7, 19 -24). Through these extensive studies, it has also been suggested that steric and/or electrostatic properties of the distal heme pocket are different for RmFixLT vis à vis RmFixLH (heme domain of FixL), and consequently interaction of the iron-bound ligand with the heme pocket is also different between these forms. In the present study, we attempted to characterize the structure in the distal heme pocket and to reveal the effect of the kinase domain, by virtue of the Fe K-edge EXAFS and the resonance Raman spectroscopy of the CO complexes of RmFixLT, RmFixLH, and Mb. The simulation of the EXAFS of RmFixLT is shown in Fig. 4 together with the data for the CO complexes of RmFixL and Mb, and their resonance Raman spectra are also illustrated in Fig. 5. The main reason for the choice of the CO complex is that the coordination of CO to the ferrous heme iron has been the most comprehensively investigated using many biochemical and physicochemical techniques and has been discussed in detail with relevance to the heme pocket structure and its chemical environments (48,49 and references therein). The CO coordination is a sensitive marker to evaluate characters around the Fe-CO moiety in hemoproteins.
Structural Characterization of FixL in Heme Distal Side by Comparison with Myoglobin-To characterize the heme pocket structure, we compared the CO binding geometry between Rm-FixLT and Mb (see Table I, top and bottom parts), and in particular, we focused on the bond angle of the CO coordination, 208 Fe-C-O, which is different between RmFixLT (157°) and Mb (149°). A relative difference of 8°is significant in terms of the maximum fitting errors, but a note of caution is required for the absolute errors. 2 The Fe-CO coordination in RmFixLT is less bent than that of Mb, although other parameters, including the Fe-CO distance (1.81 Å) and the Fe-N im distance (2.08 -2.09 Å), are basically identical.
Because the linear coordination is favorable for the CO coordination in the model system, the bent CO binding in the proteins can be explained in terms of effects of protein parts in the distal heme pocket. To examine the protein effect on the Fe-CO bond, we also measured the resonance Raman spectrum of the CO-bound form of RmFixLT (Fig. 5), in which the Fe-CO stretching ( Fe-CO ) and the C-O stretching ( CO ) frequency bands were observed at 498 and 1962 cm Ϫ1 , respectively. These bands were shifted to 496 and 1921 cm Ϫ1 upon the 13 CO replacement and to 490 and 1873 cm Ϫ1 upon the 13 C 18 O replacement, respectively. Compared with the spectrum of MbCO ( Fe-CO ϭ 507 cm Ϫ1 , CO (main band) ϭ 1945 cm Ϫ1 ), the Fe-CO locates at lower positions, whereas the CO locates at higher positions, indicating that the Fe-CO bond character in Rm-FixLT is different from that in Mb. The back donation from the iron d orbital to the ligand CO p* orbital is less in RmFixLT than in Mb, decreasing the Fe-CO bond order, while increasing the C-O bond order.
According to the comprehensive studies of vibrational spectroscopies of the CO adduct of hemoproteins (49 -52), two factors that significantly effect CO coordination have been considered as follows: the steric hindrance to the iron-bound CO by the distal residues and electrostatic interactions of the bound CO with a hydrophobic heme pocket. The less bent CO coordination in RmFixLT, compared with that in Mb, increases in the back donation by increment of overlap between the ferrous iron d orbital and the CO p* orbital, resulting in increase in the back donation. Therefore, the Fe-CO bond order must be increased, but this is not the case for RmFixLT and Mb.
On the other hand, since the polar and cationic His-64 (distal His) is present in the distal heme pocket of Mb, the back donation is facilitated to increase in the Fe-CO bond order. Therefore, we can suggest that the Fe-CO environment of Rm-FixLT should be apolar, retarding the back donation and decreasing the Fe-CO bond order (53). In the case of RmFixLT, the hydrophobic effect is predominant to determine its Fe-CO bond character. This suggestion is in good agreement with that reported on the basis of the NMR, ESR, resonance Raman, and kinetic studies (21)(22)(23)(24)(25). In addition the recent crystallographic study showed that its distal heme pocket is constructed by some hydrophobic residues such as Ile-215, Leu-236, and Ile-238 of BjFixLH.
Considering the van der Waals radii of side chains of the distal residues, the heme sixth sites of FixL are packed. Despite such crowded space in the distal heme pocket, some spectroscopic results including EXAFS have suggested the less bent CO coordination in FixL than in Mb. These observations could suggest that steric environment in the distal heme pocket of FixL allows the less bent CO coordination, but highly bent coordination might be unfavorable. Indeed, the CN Ϫ ligand is linearly bound to the ferric iron, as was found in the crystal structure of BjFixLH (26), whereas the triatomic pseudoha-lides, SCN Ϫ and N 3 Ϫ , which generally bind the heme ferric iron in the highly bent fashion (54), do not bind to FixL (21,23). On the other hand, the bulky imidazole can bind to RmFixLT (55). Therefore, the relationship between the ligand binding and the steric effect in RmFixLT still seems controversial. Further studies to elucidate the ligand selection mechanism of FixL should be carried out.
Interaction of Heme Distal Side with Kinase Domain-As shown in Table I, top and bottom parts, the linear CO coordination is more favorable in RmFixLH (171°) than in RmFixLT (157°). The change of the CO coordination from bent to linear increases in the overlap of the porphyrin and CO * orbitals, eventually increasing the back donation from the iron/porphyrin moiety to the CO ligand. Indeed, in the resonance Raman spectra of the CO complex of RmFixLH (Fig. 5), the Fe-CO and C-O frequency bands are located at 503 and 1955 cm Ϫ1 , respectively, which were shifted to 499 and 1912 cm Ϫ1 upon the 13 CO substitution and to 492 and 1868 cm Ϫ1 upon the 13 C 18 O substitution, respectively. The Raman spectral results show that the Fe-CO bond order in RmFixLH is increased, whereas the C-O bond order is decreased, compared with those of RmFixLT (Fe-CO ϭ 498 cm Ϫ1 , C-O ϭ 1955 cm Ϫ1 ). The results are quite consistent with the previous Raman result ( 21), in which the porphyrin -electron density marker, 4 , is shifted to higher frequency (up-shifted) by 2.7 cm Ϫ1 from the ferrous-CO complex of RmFixLT to that of RmFixLH. The up-shifted 4 line in the spectrum of RmFixLH suggests diminished * electron density resulting from the increased back donation from the iron to the ligand CO, in good consistency with the Fe-CO bond of RmFixLH stronger than that of RmFixLT.
According to the suggestion discussed above, the Fe-CO bond is affected by steric hindrance and/or hydrophobicity in the distal heme pocket. Since it is not expected that the hydrophobicity around the iron-bound ligand (CO) is drastically altered from RmFixLT to RmFixLH, the change in the CO coordination geometry is possibly caused by change in steric factors. In RmFixLH, the steric constraint to the iron-bound CO is less than in RmFixLT. The steric interaction between the ironbound ligand (CO) and its surrounding is present, and its magnitude is different between RmFixLT and RmFixLH.
Recently, we have reported that RmFixLH which we purified is stable in a dimeric form like RmFixLT (28). Therefore, the structural difference in the heme distal pocket between Rm-FixLH and RmFixLT is caused by removal of the kinase domain, rather than by the monomer/dimer formation. It is also true that the heme distal pocket indirectly communicates to the kinase domain. These indications allow us to suggest that the ligand binding to the heme iron possibly affects the kinase domain conformation through the interaction, possibly steric interaction, between the iron-bound ligand and the distal heme pocket.
At present, we cannot conclude whether the interaction of the iron-bound ligand with the distal heme pocket is important or not in the intramolecular signal transduction in FixL. Our suggestion is sharply contrasted to that arisen form the crystallographic study of BjFixLH (26), i.e. less steric interaction; the side chain of the distal residues such as Leu-236 and Ile-238 shift slightly to accommodate the bound CN Ϫ , but the main chain in the ␤-strand does not significantly change in its position. However, their BjFixLH in its single crystal is in monomeric state, whereas our RmFixLT is in a heterotetrameric (RmFixLT) 2 (RmFixJ) 2 state; the latter sample is more intact than the former sample because dimeric FixL is biologically significant in vivo (56 -58). In addition, several spectroscopic measurements including EXAFS have proved that the structural characteristics in the heme distal side of FixL are distinguishable in the presence and absence of the kinase domain. These differences in the sample condition might be responsible for the discrepancy in the suggestion on the interaction between the ligand and the heme pocket, which have arisen from the crystallographic and EXAFS studies.
In the present study, we have characterized the iron coordination structure of FixL in atomic level with the Fe K-edge EXAFS and resonance Raman techniques, which has quantitatively shown the iron displacement relative to the heme plane upon the ligand binding, the apparent correlation of the Fe-N(His-194) bond length with the kinase activity of FixL, and the steric interaction of the iron-bound ligand with the heme pocket. The EXAFS and the Raman data are supported by each other and agree with several spectroscopic and kinetic results reported thus far. In understanding the structure-function relationship of FixL, these data should be complementary to those provided by the crystallographic data (26), which is informative on the overall topology and the heme pocket structures of the FixL heme domain. However, there are also some discrepancies, as were stated in this text. In addition to the difference in the sample condition, another reason for these discrepancies, therefore, may be that the resolution of the crystal data of BjFixLH (26) is not sufficiently high for evaluating subtle structural changes that can be assessed by spectroscopic techniques. This suggestion is based on the Protein Data Bank data of BjFixLH (1BV6 and 1BV5), where, for example, temperature factors of some chains are too high and completeness of the data of the metCN Ϫ complex is less than 90%. The crystal data with higher resolution will be discussed in the context of the spectroscopic data to understand the interaction between the heme and the kinase domains in structural terms.