Stationary and Time-resolved Resonance Raman Spectra of His77 and Met95 Mutants of the Isolated Heme Domain of a Direct Oxygen Sensor from Escherichia coli *

The heme environments of Met95and His77 mutants of the isolated heme-bound PAS domain (Escherichia coli DOS PAS) of a direct oxygen sensing protein from E. coli (E. coli DOS) were investigated with resonance Raman (RR) spectroscopy and compared with the wild type (WT) enzyme. The RR spectra of both the reduced and oxidized WT enzyme were characteristic of six-coordinate low spin heme complexes from pH 4 to 10. The time-resolved RR spectra of the photodissociated CO-WT complex had an iron-His stretching band (νFe-His) at 214 cm−1, and the νFe-CO versus νCO plot of CO-WT E. coli DOS PAS fell on the line of His-coordinated heme proteins. The photodissociated CO-H77A mutant complex did not yield the νFe-His band but gave a νFe-Im band in the presence of imidazole. The RR spectrum of the oxidized M95A mutant was that of a six-coordinate low spin complex (i.e. the same as that of the WT enzyme), whereas the reduced mutant appeared to contain a five-coordinate heme complex. Taken together, we suggest that the heme of the reduced WT enzyme is coordinated by His77 and Met95, and that Met95 is displaced by CO and O2. Presumably, the protein conformational change that occurs upon exchange of an unknown ligand for Met95 following heme reduction may lead to activation of the phosphodiesterase domain of E. coli DOS.

Heme-containing signal-transducing proteins (1)(2)(3) respond to diatomic molecules, which act as physiological, environmental messengers. This has attracted the attention of biophysical chemists. The O 2 sensing proteins so far identified include FixL (an oxygen-sensing kinase of Rhizobia meliloti) (1,4), HemAT (an oxygen sensor heme protein discovered from Bacillus subtilis (HemAT-Bs) and Halobacterium salinarium (HemAT-Hs)) (5, 6), PDEA1 (7), and putatively a heme protein from E. coli (designated Escherichia coli DOS) (8). There is only one CO sensor protein known (CooA, a CO-binding transcriptional regulation factor from Rhodospirillum rubrum) (9,10) and one NO sensor (soluble guanylate cyclase) (11,12). In each case, binding of an external ligand to the heme located in an N-terminal sensory domain transmits a signal to the functional C-terminal domain (either enzymatic or DNA binding). We are curious to know how these proteins recognize a specific diatomic molecule to generate the appropriate physiological response and what kind of structural changes occur to transmit the signal from the sensory domain to the functional domain.
The sensory domain of FixL belongs to the large family of signal-transducing PAS domain 1 proteins, whereas those of HemAT, CooA, and soluble guanylate cyclase do not. The PAS domain proteins found in eukarya, archaea, and bacteria contain a partly conserved tertiary structure despite their limited sequence homology (Ͻ15%) and dissimilar cofactors (13). Although structures of three PAS proteins including the human voltage sensor (HERG) (14), the rhizobial oxygen sensor (FixL) (15,16), and bacterial light sensor (PYP) (17) have been solved, interactions between the sensory domain and the functional domain are not clearly understood. Namely, hydrophobic interactions seem important to regulate the K ϩ channel of HERG, whereas polar interactions in the EF loop of the PAS domain seem to be essential to PYP. In the case of FixL, either a protein conformational change associated with the location of the heme iron (in-plane/out-of-plane) (15) or a ligand-protein interaction on the distal side of the heme pocket (16) appears to play a substantial role in regulating the activity of the functional domain.
E. coli DOS was found in E. coli by the Gilles-Gonzalez group who predicted, on the basis of sequence homology to the PAS domain of FixL, that it is an O 2 sensor enzyme (8). The same group later found an O 2 -sensitive phosphodiesterase (PDE A1) 2 in Acetobacter xylinum (designated AxPDEA1) (7). The N-terminal 140 residues of AxPDEA1 contain the heme binding PAS motif, whereas PDE activity is present in the C-terminal region. Importantly, the E. coli DOS protein is highly homologous (50%) to the C-terminal region of AxPDEA1. In our previous paper, we found that E. coli DOS is a PDE and that the activity is sensitive to the heme redox state rather than O 2 binding (18).
The study demonstrated that the enzyme is inhibited by oxidation of the heme iron and on the binding of external ligands such as CO and NO.
Resonance Raman (RR) spectroscopy is a powerful tool for elucidating the structural characteristics of heme domains, by providing detailed information about the coordination structure of the heme and the protein environment of the bound ligand (19 -21). Tomita et al. (22) recently reported comparative RR analysis of various heme-bound PAS domains including FixL, AxPDEA1, and E. coli DOS. They found unusual characteristics in the heme environment of E. coli DOS compared with other PAS domains. In the present paper, we examine visible stationary and time-resolved resonance Raman spectra of the wild type form of the isolated heme-bound PAS domain of E. coli DOS (E. coli DOS PAS) (WT) and the His 77 and Met 95 mutants in order to further elucidate the structurefunction relationships of the heme domain. From the RR data, it appears that, in the ferrous complexes, one of the axial ligands to E. coli DOS PAS is His 77 and the other is Met 95 .

EXPERIMENTAL PROCEDURES
Materials-Cloning of E. coli DOS, construction of the expression plasmids and purification of the wild type and mutant proteins was performed essentially as described in our previous paper (18). The His 77 mutant proteins were expressed and purified in the presence of imidazole (10 mM Resonance Raman Measurements-Continuous wave Raman scattering was performed by exciting at 421 nm with a blue diode laser (Hitachi Metals, model ICD-430), at 413 nm with a Kr ϩ ion laser (Spectra Physics, model 2016) or 441.6 nm with a helium/cadmium laser (Kinmonn Electric, model CD4805R). The excitation light was focused on the sample contained in a variable speed spinning cell. The laser power at the sample point was typically 3-4 mW but was made 0.1-0.2 mW for CO-E. coli DOS PAS to minimize photodissociation of bound CO. The scattered light along a right angle from the incident radiation was dispersed by a 100-cm single polychromator (Ritsu Oyo Kogaku, model DG-1000) equipped with a cooled CCD detector (Princeton Instruments, model CCD-1100). Raman shifts were calibrated using indene, carbontetrachloride, dimethylformamide, and cyclohexane. The accuracy of frequencies are Ϯ1 cm Ϫ1 for well defined peaks.
Picosecond time-resolved resonance Raman (TR 3 ) spectra were measured using a homemade pump/probe system, details of which have been described elsewhere (23,24). Briefly, the probe beam at 442 nm (0.2 J/pulse) was the first Stokes stimulated Raman line of methane gas derived from a homemade Raman shifter, whereas the pump beam at 540 nm (12 J/pulse) was generated by optical parametric generation and amplification. Both pulses were obtained from the second harmonic of the 784-nm output of a Ti-sapphire laser operated at 1 kHz. Raman scattered light was detected with a liquid nitrogen-cooled CCD detector (Princeton Instruments, model CCD-1100PB), which was attached to a single spectrograph (Chromex, model 500IM-CM). Nanosecond TR 3 spectra were measured with a single-color arrangement using 10-ns pulses at 427 nm (ϳ0.3 mJ/pulse at a sample) operated at 100 Hz. The light pulse was obtained from an XeCl excimer laser-pumped dye laser (Lambda Physik, model EMG 103MC/FL2002). In this experiment, the 308-nm output of the XeCl laser was converted to 427 nm with stilbene 420. The Raman scattered light was detected with the same detection system as used for continuous wave RR experiments. Fig. 1 shows the unpolarized RR spectra of reduced (a) and oxidized (b) E. coli DOS PAS excited at 421 nm in the 180 -620 cm Ϫ1 region (A) and the parallel and perpendicular components of the spectra in the 1300 -1700 cm Ϫ1 region (B). The bands of the reduced form at 1361, 1493, 1580, and 1609 cm Ϫ1 are assigned to 4, 3, 2 , and 10 of porphyrin in-plane vibrations, respectively (19). The 4 , 3 , and 2 band positions are the same as those previously reported (22), whereas the 10 band was located at 1625 cm Ϫ1 in Tomita et al. (22). The 4, 3 , and 10 bands, which are used as the oxidation and coordination/spin state markers, have frequencies similar to those of cytochrome . The RR spectra in B are represented as polarized spectra in which the electric vector of scattered light is parallel (//) or perpendicular (Ќ) to that of the incident light. The ordinate scales of the spectra in the 1380 -1700 cm Ϫ1 region are 5 times expanded relative to those in the 1300 -1380 cm Ϫ1 region. The solvent was 50 mM sodium phosphate buffer at pH 7.7, and the excitation wavelength was 421 nm with a power of 4 mW at the sample. b 5 (25) and CooA (26,27), indicating that E. coli DOS PAS adopts a six-coordinate low spin (6c-ls) state. The 4, 3, 2 and 10 bands of the oxidized form are observed at 1372, 1505, 1577, and 1641 cm Ϫ1 , respectively, similar positions to those previously reported by Tomita et al. (22) and typical 6c-ls type. The Raman bands assignable to the vinyl side chains of the heme are seen at 1620, 1433, and 412 cm Ϫ1 for the oxidized form, similar to those of the reduced form at 1620, 1432, and 410 cm Ϫ1 . This indicates that the structure of the vinyl side chains is similar in the oxidized and reduced forms. Although these spectra were observed for E. coli DOS PAS at pH 7.7, RR spectra were hardly changed between pH 4.4 and 10.0. This means that the heme coordination is not altered in this pH range. Fig. 2 compares the RR spectra of CO-E. coli DOS PAS (b) and O 2 -E. coli DOS PAS (c) excited at 421 nm with those of reduced (a) and oxidized (d) forms excited at the same wavelength. The 4, 3 , and 2 bands of CO-E. coli DOS PAS are identified at 1370, 1496, and 1581 cm Ϫ1 , which are distinctly different from those of reduced E. coli DOS PAS but are close to those of general CO-bound 6c-ls hemes (28). The 4, 3, 2 , and 10 bands of O 2 -E. coli DOS PAS are observed at 1375, 1505, 1580, and 1640 cm Ϫ1 , which are rather close to those of the oxidized form but also close to those of oxy-R. melilori FixL ( 4 ϭ 1376, 3 ϭ 1502, 2 ϭ 1577, and 10 ϭ 1636 cm Ϫ1 ) (29). The bands in this region of the CO-and O 2 -E. coli DOS PAS spectra have not been reported previously.

RESULTS
To confirm that the RR spectra of the lower frequency region arise from the ligand bound forms, their dependences on isotopically labeled ligands were examined. Fig. 3 shows the RR spectra of 16 . It is apparent from the isotope difference spectrum that the ϳ559 cm Ϫ1 band of 16 Simulation of the difference spectrum with Gaussian band shape functions enabled us to determine the precise band positions in the raw spectra, which were 561 cm Ϫ1 for 16 (22).
The CO-isotope dependence of the CO-E. coli DOS PAS RR spectra is illustrated in Fig. 4, where spectra a and b in panel A represent the spectra of 12 C 16 O-E. coli DOS PAS and 13 C 18 O-E. coli DOS PAS in the 300 -800 cm Ϫ1 region, respectively; spectra d and e in panel B represent the spectra of 12 C 16 O-E. coli DOS PAS and 13 C 16 O-E. coli DOS PAS in the 1500 -2050 cm Ϫ1 region, respectively; and spectra c and f display the isotope difference spectra (c ϭ a Ϫ b, and f ϭ d Ϫ e). It is obvious from the difference spectra that the 486 and 1973 cm Ϫ1 bands of 12 C 16 O-E. coli DOS PAS are shifted to 472 and 1927 cm Ϫ1 after isotopic labeling. Accordingly, the former and latter are assigned to the Fe-CO stretching ( Fe-CO ) and C-O stretching ( CO ) modes, respectively. These isotope-dependent spectral changes were not obtained by Tomita et al. (22). The RR spectra of CO-E. coli DOS PAS did not change with pH over the range 4.4 -10.0.
Usually, CO bound to histidine-coordinated heme is photodissociable, resulting in five-coordinate high spin (5c-hs) heme. To see this state, the laser power was increased during the continuous wave Raman experiments with CO-E. coli DOS PAS. This resulted in a frequency shift of 4 from 1370 to 1361 cm Ϫ1 (results not shown). Nevertheless, the iron-histidine stretching band, which is observable in the 200 -250 cm Ϫ1 region only for the ferrous 5c-hs form (20), was not observed. Furthermore, the 3 band was observed at 1493 cm Ϫ1 , similar to that in spectrum a in Fig. 2. This means that the heme is still ferrous 6c-ls but has formed a different complex. In other words, the CO has photodissociated, but an endogenous ligand has bound rapidly in its place. The transit time for a molecule to go across the laser beam under the particular cell spinning conditions used is estimated to be a few hundred microseconds; coordination of an internal residue to the axial site of heme must therefore be completed in the nanosecond time range. This was investigated by collecting TR 3 spectra for CO-E. coli DOS PAS.
The results from picosecond TR 3 experiments on CO-E. coli DOS PAS are displayed in Fig. 5, where A and B represent the TR 3 spectra in the low (150 -800 cm Ϫ1 ) and high frequency (1250 -1650 cm Ϫ1 ) regions, respectively, and spectra a-d correspond to the pump/probe results with time delays of Ϫ5, 1, 10, and 1000 ps. Spectra e in both panels show the spectrum obtained with the probe beam only and therefore reflect CO-E. coli DOS PAS before photolysis. This spectrum agrees with spectrum b in Fig. 2, demonstrating that the probe beam is weak enough to protect the sample from photolysis. Since about 30% of the CO is photolyzed under these conditions, the observed spectrum contains a contribution from the unphotolyzed species. This contribution was subtracted from the observed spectra by the use of spectrum e. Accordingly, spectra a-d reflect the transient species only. Spectra a in both panels exhibit no band features, indicating that the sample is completely restored to CO-E. coli DOS PAS in one turn of the spinning cell and gives rise to the same spectrum as that observed without the pump light. After 1 ps (b), the 4 band is shifted to 1354 cm Ϫ1 , the same frequency as observed for deoxy-Mb with histidine-coordinated 5c-hs heme. The spectral pattern of spectrum b in Fig. 5B is distinctly different from that shown in Fig. 2, spectrum a, but close to that observed for photolyzed MbCO (30). In the lower frequency region, two bands appeared at 214 and 298 cm Ϫ1 , which were absent in the spectrum of the reduced form. The two bands can be assigned to the iron-histidine stretching ( Fe-His ) and ␥ 7 modes of the ferrous 5c-hs heme, respectively. The intensities of the two bands increased after 10 ps and did not decrease even after 1000 ps. This means that the equilibrium structure of the 5c-hs heme is attained in a short time and that photolyzed CO leaves the heme pocket, unlike in the case of CO-CooA (31). Thus, the picosecond TR 3 experiment has established that the transligand to CO is histidyl imidazole and that the coordination of an internal ligand to the sixth coordination position is not as fast as in the case of CooA. Similar TR 3 experiments were carried out with O 2 -E. coli DOS PAS, but the 4 band was observed at the same frequency as in Fig. 2c, indicating that no photodissociation of O 2 occurs.
It has been suggested from amino acid sequence alignment and mutation experiments (8,18) that Met 95 and His 77 are the heme axial ligands. Therefore, we examined the RR spectra of the Ala and His mutants of these residues. Fig. 6 shows the 421-nm excited RR spectra in the 1300 -1700 cm Ϫ1 region of M95H in the reduced state (a), and of M95A in the reduced (b) and oxidized (c) states. Spectrum a is of a typical ferrous 6c-ls type, suggesting the formation of bishistidine-coordinated heme. The addition of CO yielded the Fe-CO and CO bands at 489 and 1965 cm Ϫ1 , respectively (spectra not shown). Spectrum b is definitely different from spectrum a in Fig. 2 as well as from the spectrum of oxidized M95H. The 2 , 3 , and 4 bands are observed at 1558, 1468, and 1353 cm Ϫ1 , indicating the presence of the ferrous 5c-hs heme. The RR spectrum in the lower frequency region, which was obtained with excitation at 441.6 nm, is displayed in the inset of Fig. 6. Here, a band assignable to the Fe-His mode (20) is observed at 213 cm Ϫ1 . Since E. coli DOS PAS has two His residues (His 77 and His 83 ), it is highly likely that a His-coordinated 5c-hs heme complex is formed in the reduced form of M95A. The addition of CO to it yielded the Fe-CO and CO bands at 487 and 1966 cm Ϫ1 , respectively, which are fairly close to those for M95H. Unexpectedly, the spectrum of the oxidized form of M95A (c) is very close to spectrum d in Fig. 2, indicating the presence of 6c-ls heme (i.e. similar to the WT form). It suggests the coordination of a residue other than Met to the sixth site of the heme ferric complex. We have examined the pH dependence of the RR spectrum of the mutant but noticed no pH-induced spectral changes between pH 6 and 10.
As described in our previous paper, mutation of His 77 destabilized heme binding, whereas mutation of His 83 did not (18). The H77A mutant expressed and prepared in the presence of imidazole (Im), however, did bind heme and retained it even after the removal of Im by gel filtration chromatography like the corresponding Mb mutant (32). We measured RR spectra of the H77A mutant prepared like this, both in the absence and presence of exogenous Im, in which the concentration of Im is adjusted to 500 equivalents of the protein. Fig. 7 shows the 421-nm excited RR spectra of the reduced form of the H77A mutant in the presence (a) and absence (b) of exogenous Im. In the absence of Im, two 3 bands are observed at 1468 and 1492 cm Ϫ1 with nearly equal intensity, indicating the coexistence of the 5c-hs and 6c-ls heme complexes. The addition of a 500-fold excess of Im to the protein solution resulted in weakening of the 5c-hs marker band at 1468 cm Ϫ1 and concomitant intensification of the 6c-ls marker band at 1492 cm Ϫ1 . This indicates that an alternative residue is coordinated to the heme iron in the absence of exogenous Im. To identify it, we examined its CO adduct.
The 421-nm excited RR spectra of the CO adduct of H77A in the absence (a) and presence (b and c) of exogenous Im are shown in Fig. 8, where A and B display the 250 -600 and 1800 -2100 cm Ϫ1 regions of the spectra, respectively. In both A and B, spectra a and b were obtained with 12 C 16 O, spectra c were obtained with 13 ; Fig. 4). This means that the environments around the CO are likely to be similar, because both frequencies are sensitive to residues surrounding the bound CO (33). It appears that exogenous Im is bound instead of His 77 and that CO occupies the same site as in the WT protein. In the absence of exogenous Im, on the other hand, the Fe-CO and CO bands appear at 486 and 1978 cm Ϫ1 , respectively. The Fe-CO and CO frequencies are summarized in Table I and compared with those of related proteins. The CO frequency of the H77A mutant in the absence of Im is highest and shifted in the opposite direction to that of the M95A mutant. To probe the heme coordination of the H77A mutant in the absence of Im, photolysis experiments were carried out with the nanosecond pulse single color observation system. Fig. 9 shows the RR spectra of photolyzed CO-WT E. coli DOS PAS (a), CO-H77A E. coli DOS PAS in the absence (b) and presence (c) of exogenous Im obtained with 427-nm 10-ns pulses. In Fig. 9, A and B display the spectra in the 160 -470 and 1250 -1650 cm Ϫ1 regions, respectively. The appearance of the 4 band at 1354 cm Ϫ1 (B) means that the dominant species present is a photolyzed 5c-hs heme, although there is a small amount of the unphotolyzed species in which the 4 band appears around 1370 cm Ϫ1 as a shoulder. For the WT E. coli DOS PAS, the Fe-His band is observed at 213 cm Ϫ1 , in agreement with the results of the ps-TR 3 spectra (b-d) in Fig. 5. In contrast, the photoproduct of CO-H77A E. coli DOS PAS in the absence of exogenous Im does not yield a band corresponding to the Fe-His mode around 200 -260 cm Ϫ1 (Fig. 9A, b). Except for this point, the general pattern of the RR spectrum of the H77A mutant is close to that of the WT enzyme, suggesting the formation of 5c-hs heme in the photoproduct. If the trans ligand to the iron-bound CO is His, a band corresponding to the Fe-His mode is expected to appear for the photoproduct. Its absence indicates that the CO-H77A E. coli DOS PAS complex does not have His as the axial ligand, supporting the coordination of His 77 to the heme iron in the CO-WT E. coli DOS PAS complex.
In the presence of a 500-fold molar excess of exogenous Im, on the other hand, a weak but reproducible band was detected at 225 cm Ϫ1 for the H77A mutant (c). This feature is very close to that of the H77Y mutant of CooA, for which the Fe-Im stretch- ing band was observed at 222 cm Ϫ1 in the presence of exogenous Im but not in its absence (31). Similar observations are also reported for a cavity mutant of Mb in which the proximal His is replaced by Gly (226 cm Ϫ1 ) (34). Accordingly, this observation suggests that Im is bound to the heme iron of a proportion of the CO-H77A E. coli DOS PAS complex allowing the Fe-Im stretching mode to be weakly observed. In the H77A mutant of E. coli DOS PAS, a nonhistidine residue is coordinated to the heme iron, and CO is bound in the trans position. This structure is retained in the presence of exogenous Im for most of the sample, but for a minor proportion, this residue is replaced by CO and Im is bound to the original site of His 77 (i.e. trans to CO).

DISCUSSION
Analysis of full-length E. coli DOS demonstrated that ferrous, but not ferric, E. coli DOS shows phosphodiesterase activity with cAMP (18). To understand the molecular mechanism of this sensing process, elucidation of structural changes of the heme site caused by redox state changes and binding of exogenous ligands is essential.
Heme Axial Ligand-The present RR experiments demonstrated that both the ferric and ferrous forms of WT E. coli DOS PAS have 6c-ls heme. The spectral insensitivity to pH of the ferric form excludes a possibility of water coordination, because if water is coordinated to an oxidized heme at neutral pH, the transition from high to low spin state, which is usually accompanied by frequency shifts in the 2 , 3 , and 10 bands, would occur upon raising pH (35,36). Therefore, the two axial sites of the heme iron must be occupied by amino acid residues in both the ferrous and ferric states. Since the M95A mutant gave the RR spectrum of a 5c-hs type heme in the reduced state, it is likely that Met95 occupies one of the axial sites in the reduced complex.
It is well known that Fe-CO frequencies have a linear inverse correlation when they are plotted against CO frequencies (37,38). The Fe-CO versus CO plot for CO-E. coli DOS PAS falls on the line of the neutral histidine-coordinated heme proteins as shown in Fig. 10. This fact strongly suggests that the trans ligand of CO in CO-WT E. coli DOS PAS is a neutral histidine. The Fe-CO versus CO plot of CO-M95A E. coli DOS PAS also falls on the line of His-coordinated heme proteins. Presumably, Met 95 is replaced by CO. This conclusion is consistent with the fact that the TR 3 experiments with CO-E. coli DOS PAS demonstrated the presence of the Fe-His band for the CO-photodissociated form. The Fe-His band was also observed at 213 cm Ϫ1 for the stationary state of the reduced M95A mutant but was  absent for the photoproduct of CO-H77A E. coli DOS PAS. Consequently, it is highly plausible that His 77 is the other axial ligand in the reduced form. The Fe-CO versus CO plot for the H77A mutant also falls on the straight line of the His-coordinated heme proteins, although its location is different from that of CO-WT E. coli DOS PAS as shown in Fig. 10. This may mean that some unknown residue other than histidine is bound to the trans site of CO in CO-H77A E. coli DOS PAS. If Met 95 remains bound in the reduced state and can yield the CO adduct, all of the results are satisfactorily interpreted, although so far there is no evidence for Met-Fe(II)-CO heme. Alternatively, some nitrogenous ligand may be bound instead of His 77 . Anyway, the bond length of the Fe(II)-Met 95 or Fe(II)-nitrogenous ligand is not strong, and accordingly, when exogenous Im is present with the H77A mutant, Im displaces the residue, binds to the heme iron, and yields the Im-Fe(II)-CO adduct in the presence of CO, in addition to the Met (or nitrogenous base)-Fe(II)-CO adduct.
In the oxidized form, however, the RR spectrum of the M95A mutant was very close to that of WT E. coli DOS PAS. It remained unchanged over a pH range of 6 -10. Also, Met 95 mutants retained PDE activities. 3 Therefore, it is deduced that some residue other than Met 95 is coordinated to the axial site of Fe(III). If so, the axial ligand could be replaced upon changing the redox state of the heme iron. Biochemical experiments (18) demonstrated that E. coli DOS is enzymatically active in the reduced state but inactive in the oxidized state and in the O 2 -bound form. It is also noted that binding of CO and NO inhibits the enzymatic activity (18). Although the name of this protein is an "oxygen sensor," the protein might be a redoxcontrolled enzyme. A similar kind of axial ligand replacement occurs upon changing the redox state of the heme iron in CooA (39).
Iron-Histidine Bond-The iron-His stretching mode was observed at 214 cm Ϫ1 for CO-photodissociated E. coli DOS PAS. The Fe-His frequency reflects the nature of the protein. It is well known for deoxy-Hb that geometrical distortion of the axial His due to strain exerted by the protein causes a shift of this mode toward lower frequencies and that the magnitude of the frequency shift correlates with the oxygen affinity (40). The strain is relieved from the heme when the proximal His is replaced with Gly or Tyr, which allows an exogenous Im to be bound to the heme iron. Its Fe-Im stretching frequency is free from the strain exerted by the protein (31,34,41). The Fe-His frequency of CO-photodissociated WT E. coli DOS PAS is fairly low compared with other His-coordinated heme proteins, but since the Fe-His frequency of the photodissociated transient form may not be the same as that in the equilibrium state, we compared the transient Fe-His frequency of the native enzyme with the transient Fe-Im frequency of the cavity mutant for both photodissociated species (i.e. ⌬ (His 3 Im) ϭ Fe-Im Ϫ Fe-His ). The larger the ⌬ (His 3 Im) value is, the stronger the strain in the Fe-His bond. The value of ⌬ (His 3 Im) for H77A E. coli DOS PAS is 11 cm Ϫ1 , which is larger than that of Mb (⌬ (His 3 Im) ϭ 6 cm Ϫ1 ) (34) but close to that of other sensor protein such as CooA (⌬ (His 3 Im) ϭ 11 cm Ϫ1 ) (31). This means that some strain is exerted on the heme by the protein moiety via the iron-His bond in E. coli DOS PAS, which might be a common property of signal-transducing proteins.
The other feature of the Fe-His frequency is that it reflects the basicity of the bound histidyl imidazole. When a strong hydrogen bond is formed between proximal His and the surrounding protein, the Fe-His band shifts toward higher frequency (42). In fact, the Fe-His frequency of Coprinus cinereus peroxidase is shifted to a higher frequency (245 cm Ϫ1 ) due to a strong hydrogen bond formed with Asp 245 , but when this hydrogen bond is disrupted by the substitution of Asp 245 with Asn, the Fe-His frequency shifts down to 204 cm Ϫ1 (43). Accordingly, the relatively low frequency of Fe-His for E. coli DOS PAS suggests that the axial His (His 77 ) does not form a strong hydrogen bond to surrounding residues.
Environments around Bound CO-It is known from the RR and IR studies of CO-mutant Mb complexes that CO serves as a sensitive probe of the heme distal pocket, because the Fe-CO and CO frequencies are mainly determined by back donation from CO to Fe(II) and thus by the electrostatic field generated by the surrounding residues rather than the Fe-C-O geometry (44). The Fe-CO and CO frequencies of Mb complexes, which were identified at 486 and 1973 cm Ϫ1 , respectively, suggest that the bound CO lies in an electronically neutral (or slightly negative) environment. The Fe-C-O bending RR band was not recognized in Fig. 4, suggesting that the Fe-C-O adopts a nearly linear and upright structure. The RR spectra of CO-E. coli DOS PAS exhibited no pH dependence between pH 4 and 10. This means that protonation/deprotonation does not occur in the distal side of the heme pocket between pH 4 and 10. These facts imply that the distal side of the heme pocket is fairly hydrophobic. During preparation of this paper, Tomita et al. (22) published similar results for the wild type E. coli DOS PAS, in support of this.
The results of the ps-TR 3 experiments provide evidence of further characteristic features of the heme pocket. It was shown in Fig. 5 that recombination of photodissociated CO to the heme was negligible within 1000 ps following photolysis in E. coli DOS PAS. This is in sharp contrast with the case of CooA, for which most of the photodissociated CO was recombined with the heme with a time constant of 70 ps (31). The rapid recombination of CO is thought to reflect a small cagelike distal heme pocket. Kinetic studies of CO recombination for various CO-Mb mutants indicated that replacement of Leu 29 and Val 68 on the heme distal side with bulky residues such as Trp or Phe resulted in faster recombination of CO (45,46). Thus, it appears that the bound CO in the CO-E. coli DOS PAS complex is not as crowded as in CooA, although CO binding is accompanied by displacement of an internal axial ligand (Pro2 in CooA and most probably Met 95 in E. coli DOS PAS) from the axial coordination site of heme.
Environments around O 2 -In the RR spectrum of oxy-E. coli DOS PAS (Fig. 3) the Fe-O 2 stretching mode ( Fe-OO ) was located at 561 cm Ϫ1 , which is considerably lower than that of Mb. For Mb, it was reported that the Fe-OO frequency is insensitive to hydrogen bonding between the terminal oxygen atom and distal His 64 (47). The considerably low Fe-OO frequency is noted for Mycobacterium tuberculosis Hb ( Fe-OO ϭ 564 cm Ϫ1 ) (48) and the Tyr(B10)3 Leu mutant of Chlamydomonas Hb ( Fe-OO ϭ 561 cm Ϫ1 ) (49). Contrary to the Mb case, the Fe-OO frequencies for these systems are sensitive to mutation of the nearby residues in the distal heme pocket as shown in Table I. This sensitivity is attributed to the formation of a hydrogen bond between the proximal O atom of bound O 2 and the distal residue (Tyr in M. tuberculosis Hb; Gln and Tyr in Chlamydomonas Hb), which directly constrains and weakens the Fe-O 2 bond, resulting in the lowering of the Fe-OO frequency to ϳ560 cm Ϫ1 (48,49). Accordingly, we deduce that the existence of hydrogen bonding between the proximal oxygen atom and amino acid residues is responsible for the low frequency of  (50)). An extremely slow O 2 dissociation rate is also observed for the Hbs of Chlamydomonas (k off ϭ 0.014 s Ϫ1 ) (51) and Ascaris summ (k off ϭ 0.004 s Ϫ1 ) (52), in which strong hydrogen bonds between the proximal oxygen atom of bound O 2 and surrounding residues has been noted.
Flexibility of Protein Structures in the Heme Pocket-The RR spectra confirmed that both axial positions of the heme iron are occupied by amino acid residues in WT E. coli DOS PAS, and therefore an exogenous ligand must displace one in order to bind to the heme iron, as Gilles-Gozalez and co-workers have pointed out (8). The present study suggested that His 77 and Met 95 are coordinated to the reduced heme iron and that Met 95 is the residue replaced by exogenous ligands such as CO and O 2 . The reduced M95A mutant gave the RR spectrum of a 5c-hs type heme, but the reduced M95H mutant yielded RR marker bands at the frequencies of a 6c-ls heme; 2 ϭ 1579, and 3 ϭ 1491 cm Ϫ1 . This means that His 95 is coordinated to the Fe(II) heme. However, the addition of CO to it yielded the Fe-CO and CO bands at 489 and 1965 cm Ϫ1 , respectively, which are the same as those of the M95A form (487 and 1966 cm Ϫ1 ). Probably, His 95 is displaced from the sixth coordination site by CO, although bis-His coordinated hemes like cytochrome b 5 do not easily form CO adducts. The same frequencies suggest that the displaced His does not have a positive charge; otherwise, the polar NH group of its imidazole ring would increase the Fe-CO frequency in a distance-dependent manner (53). This leads us to speculate that binding of CO to the heme iron causes a large movement of the axial residue, but the protein is somewhat flexible, and the size of the structural change may depend on the ligand species. While the Fe-bound CO is surrounded by hydrophobic residues, the proximal oxygen atom of iron-bound O 2 forms a strong hydrogen bond with nearby residues.
It might be useful to refer to the crystal structure of the heme domain of the O 2 -sensing B. japonicum FixL enzyme (15,55) for deducing structural changes to E. coli DOS PAS. In B. japonicum FixL, Arg 220 moves toward O 2 to form a hydrogen bond with it as it binds while, simultaneously, the side chain of Ile 215 , which is present in the proximity of the 5c-hs heme iron, is forced to move out to provide a room for the guanidinyl group of Arg 220 . By analogy, some hydrophobic residues, which are located near the heme, might be forced to move, causing charged residues to occupy the room and provide hydrogen bonds for the proximal oxygen atom of bound O 2 in E. coli DOS PAS. Arg 97 of E. coli DOS PAS, which corresponds to Arg 220 of B. japonicum FixL, is a candidate for the hydrogen-bonding residue.
Structural changes of the protein near the ligand binding site may be critical in the signal-transducing proteins. In CooA, the replacement of an internal axial ligand, Pro2, by CO is a trigger for DNA binding (56,57). For R. meliloti FixL, Mukai et al. (36) pointed out that the steric repulsion between the side chain of Ile 209 (and/or Ile 210 ) on the F/G loop, which is close to the sixth coordination site of the heme, and O 2 binding to iron causes a conformational change inhibiting the kinase activity. The structural characteristics of the heme pocket of E. coli DOS PAS revealed in this experiment are as follows: 1) redox-dependent axial ligand exchange occurs, 2) there may be a large movement in the position of the sixth ligand (most probably Met 95 ) upon binding of exogenous ligands; and 3) there are structural differences between the O 2 -and CO-bound forms. All of these characteristics suggest structural flexibility of the heme pocket, which allows us to speculate that the conformational change on the heme distal side upon ligand binding is critical for the regulation of enzymatic activity in the catalytic domain.