Roles of Arg-97 and Phe-113 in Regulation of Distal Ligand Binding to Heme in the Sensor Domain of Ec DOS Protein

The direct oxygen sensor protein isolated from Escherichia coli (Ec DOS) is a heme-based signal transducer protein responsible for phosphodiesterase (PDE) activity. Binding of O2, CO, or NO to a reduced heme significantly enhances the PDE activity toward 3′,5′-cyclic diguanylic acid. We report stationary and time-resolved resonance Raman spectra of the wild-type and several mutants (Glu-93 → Ile, Met-95 → Ala, Arg-97 → Ile, Arg-97 → Ala, Arg-97 → Glu, Phe-113 → Leu, and Phe-113 → Thr) of the heme-containing PAS domain of Ec DOS. For the CO- and NO-bound forms, both the hydrogen-bonded and non-hydrogen-bonded conformations were found, and in the former Arg-97 forms a hydrogen bond with the heme-bound external ligand. The resonance Raman results revealed significant interactions of Arg-97 and Phe-113 with a ligand bound to the sixth coordination site of the heme and profound structural changes in the heme propionates upon dissociation of CO. Mutation of Phe-113 perturbed the PDE activities, and the mutation of Arg-97 and Phe-113 significantly influenced the transient binding of Met-95 to the heme upon photodissociation of CO. This suggests that the electrostatic interaction of Arg-97 and steric interaction of Phe-113 are crucial for regulating the competitive recombination of Met-95 and CO to the heme. On the basis of these results, we propose a model for the role of the heme propionates in communicating the heme structural changes to the protein moiety.

Heme-based sensors are a class of enzymes that regulates the enzymatic activities and DNA binding in response to the presence of diatomic gas molecules CO, NO, or O 2 (1)(2)(3)(4)(5)(6). The direct oxygen sensor from Escherichia coli (Ec DOS) 4 is a heme-based signal transducer protein responsible for phosphodiesterase (PDE) activity (7,8). The Ec DOS is composed of a C-terminal PDE catalytic domain and the N-terminal hemecontaining domain. The latter is a prototypical PAS domain, which is a ubiquitous protein sensory domain found in all kingdoms (9), and has a conserved ␣/␤ folds consisting of ϳ147 residues (10,11). The Ec DOS protein has a PDE activity specific to 3Ј,5Ј-cyclic diguanylic acid (12,13), and the binding of O 2 , CO, or NO to the reduced heme significantly enhances the PDE activity (14,15). Thus, Ec DOS is a novel gas-sensor enzyme that has unprecedented ability to be activated by different gas molecules. Because the CO and NO concentrations in the cells are very low (nanomolar), it is likely that Ec DOS is predominantly an oxygen sensor enzyme whereby catalysis is regulated in response to the micromolar O 2 level (16). It is believed that the structural changes in the heme vicinity caused by ligand binding to the heme provide the initial event in the gas sensing, followed by intramolecular signal transduction from the heme to the functional domain, and thus regulating the PDE activity. Fig. 1 illustrates the crystal structure of the truncated heme sensor domain (Ec DOSH) of Ec DOS protein (10). In the reduced form (Fig. 1A), Met-95 and His-77 are the heme axial ligands in the distal and proximal sides, respectively, and Met-95 forms a hydrogen bond with heme 7-propionate. Upon O 2 binding to the reduced heme, Met-95 is replaced by O 2 and heme 7-propionate forms a hydrogen bond with Arg-97 instead of Met-95. This stabilizes the heme-coordinated O 2 (Fig. 1B). Thus, the replacement of a distal axial ligand from Met-95 to O 2 perturbs the heme 7-propionate hydrogen bonding network, resulting in large conformational changes in the FG loop (10). On the other hand, the role of other distal residues (Fig. 1) and the heme 7-propionate hydrogen bonding network are not clear in the CO-or NO-sensing mechanism.
To understand the gas-sensing mechanism of Ec DOS, it is essential to determine the changes in the heme and surrounding structures caused by distal ligand binding/dissociation. Although the crystal structure provides information about the O 2 -free (reduced) and O 2 -bound forms (10), there is not enough information available about how the structural changes are transmitted from the heme to the PDE catalytic domain. A useful way to answer this question for heme proteins is to use time-resolved studies, in which the ligand dissociation reaction is initiated by photodissociation of heme-bound external ligand.
In the present study, we have performed stationary and timeresolved resonance Raman (TR 3 ) investigations of wild-type (WT) Ec DOSH and several variants (E93I, M95A, R97I, R97A, R97E, F113L, and F113T) and examined the enzymatic activities for the full-length Ec DOS. We found the simultaneous presence of hydrogen-bonded and non-hydrogen-bonded forms for CO and NO adducts, and Arg-97 forms a hydrogen bond with the heme-bound external ligand. We also point out the importance of Arg-97, Phe-113, and heme propionates in the regulation of ligand binding at the distal side.

EXPERIMENTAL PROCEDURES
Sample Preparation-Cloning of the full-length Ec DOS and the heme-containing sensor domain Ec DOSH, the construction of expression plasmids, and the purification procedure of the WT and mutant proteins were performed essentially as described previously (8,15). Site-directed mutagenesis was performed by using a PCR-based approach as implemented in the QuikChange TM kit (Stratagene). The purities of Ec DOS samples were confirmed to be Ͼ95% homogeneous by SDS-PAGE.
For RR measurements, Ec DOSH was further purified by gel filtration through Superdex 75 (26/60 cm) pre-equilibrated with 50 mM Tris-HCl buffer at pH 7.5. The concentration of protein was adjusted to 100 M in 50 mM Tris-HCl buffer (pH 7.5). The oxidized Ec DOSH was prepared by adding an excess amount of potassium ferricyanide to the purified protein and afterward potassium ferricyanide was removed by gel filtration with Sephadex G-25. Reduced Ec DOSH was prepared by adding a minimum amount of sodium dithionite solution (final concentration, 0.5 mM) into the protein solution under nitrogen atmosphere. The NO-adduct of Ec DOSH was prepared by incubating the dithionite-reduced Ec DOSH with NO-saturated buffer.
For TR 3 experiments, ϳ100 l of 200 M oxidized protein (50 mM Tris-HCl buffer at pH 7.5), was transferred to an airtight spinning cell, into which CO gas was incorporated to a pressure of 1 atm (after evacuation of the internal pressure to 0.01 mmHg). The pump/fill procedure was repeated at least three times, and finally a small amount of degassed dithionite solution (final concentration, 0.5 mM) was added for the reduction of the oxidized protein.
Resonance Raman Measurements-RR spectra of Ec DOSH proteins were obtained with a single polychromator (SPEX750M, Jobin Yvon) equipped with a liquid nitrogencooled charge-coupled device detector (Spec10:400BLN, Roper Scientific). The excitation wavelength employed was 413.1 nm from a krypton ion laser (BeamLok 2060, Spectra Physics). The laser power at the sample point was adjusted to 1 milliwatt for the oxidized, reduced, and NO-bound forms and to 0.1 milliwatt for the CO-bound form to prevent photodissociation. Raman shifts were calibrated with indene, acetone, and an aqueous solution of ferrocyanide. The accuracy of the peak positions of well defined Raman bands was Ϯ1 cm Ϫ1 .
Time-resolved Resonance Raman Measurements-Nanosecond TR 3 experiments were performed as described previously (17). TR 3 spectra were obtained with two 10-ns pulsed lasers operating at 30 Hz. The probe light (435.7 nm) was generated by the H 2 -first anti Stokes Raman shift of the second harmonic of an Nd:YAG laser (Quanta-Ray, LAB-130), and its power was made as low as possible (ϳ80 J) to avoid photodissociation by the probe pulse. The pump light (532.0 nm) was the second harmonic of an Nd:YAG laser (Quanta-Ray, LAB-130), and its power was adjusted to 5 mJ/pulse at the sample point. The delay time (⌬t d ) of the probe pulse from the pump pulse was controlled through independent firing of two Q-switch lasers by a pulse generator (Stanford Research, DG-535) for ⌬t d values between Ϫ0.5 s and 1 ms and was monitored on an oscilloscope (Iwasaki, DS-4242) by photodiode detection. Scattered light at right angle was collected and focused by two synthetic quartz lenses onto an entrance slit of a single spectrograph (Spex, 500M) equipped with a non-blazed holographic grating and a liquid nitrogen-cooled charge-coupled device detector (Princeton Instruments, CDD-1100PB). Raman shifts were calibrated using indene, and the accuracy of frequencies are Ϯ1 cm Ϫ1 for well defined peaks.
Phosphodiesterase Assay-PDE activities of the full-length Ec DOS were determined using a colorimetric assay for free phosphate as described previously (15). The measurements were performed under anaerobic conditions (O 2 concentration Ͻ 10 ppm) in a glove box. Phe-113 mutants were fully reduced by adding 10 mM dithiothreitol, and the formation of the reduced form was confirmed by absorption spectroscopy. The CO-and O 2 -bound forms were prepared by diluting the sample in the gas-saturated, reduced PDE solution containing 10 mM dithiothreitol. The NO-bound form was prepared by adding 50 -100 M 6-(2-hydroxy-1-methyl-2-nitrosohydrazine)-N-methyl-1hexanamine to the sample of the reduced form. The PDE reaction was performed at 25°C in a mixture with 71.4 mM Tris-HCl (pH 8.0), 71.4 mM NaCl, 7.14 mM MgCl 2 , 2.14 units/l calf intestine alkaline phosphatase, 14.3 mM dithiothreitol, and 0.29 M Ec DOS and initiated by the addition of 0.3 volume of 0.33 mM 3Ј,5Ј-cyclic diguanylic acid (purchased from BIOLOG, Bremen, Germany). After the desired time, the reaction was terminated by the addition of an equal volume of 1 M HCl. The solution was then centrifuged for 5 min at 15,000 ϫ g to remove denatured proteins as precipitates. The supernatant (100 l) was mixed with 200 l of BIOMOL Green, and the mixture was incubated at 25°C for 30 min. The change in absorbance at 630 nm was measured. The initial rates of the reactions are averages of at least three time course experiments.

RESULTS
The crystal structure of the O 2 -bound form of Ec DOSH showed that Arg-97 forms hydrogen bonds with the heme-coordinated O 2 (Fig. 1B), while the interactions of the hemebound CO or NO with the surrounding residues are not known. To investigate these interactions, we prepared several mutants of the heme distal side residues (E93I, M95A, R97I, and F113L) and measured their RR spectra. It is well established that the heme of WT Ec DOSH adopts a six-coordinate low spin (6c-ls) state in the oxidized and reduced forms (18 -20). The RR spectra of the reduced form for the mutants are similar to that of WT except for M95A mutant, in which the mutation of the distal axial ligand (Met-95 in Fig. 1A), produces a 5-coordinate high spin (5c-hs) structure (supplemental Fig. S1). In addition, the RR spectra of the oxidized form for all the mutants are similar to that of WT (supplemental Fig. S1). The spectral similarity of M95A to WT was expected, because hydroxy ion (or water molecule) is coordinated to the heme axial site in the oxidized form as revealed by the crystal structure (11). These results indicate that the mutations cause no significant change on the coordination-and spin-states of the oxidized heme of the mutants.
Binding of Different Ligands-Because of the sensitivity of the Fe-CO and the C-O stretching frequencies to the heme environment, the CO adducts of heme proteins provide a valuable probe of the heme pocket. The CO-isotope dependence in RR spectra of CO-bound form of WT Ec DOSH is illustrated in Fig.  2. Spectra a and b in Panel A represent the spectra of 12 C 16 O and 13 C 18 O, respectively, in the 300 -650 cm Ϫ1 region, and spectra d and e in panel B represent the spectra of 12 C 16 O and 13 C 18 O, respectively, in the region 1800 -2100 cm Ϫ1 . Spectra c and f display the isotope difference spectra (c ϭ a and b, and f ϭ d and e). It is obvious from the difference spectra that the 488, 1924, and 1970 cm Ϫ1 bands of 12 C 16 O are shifted to 477, 1837, and 1879 cm Ϫ1 , respectively, after isotopic labeling. Accordingly, the 488 cm Ϫ1 band is assigned to the Fe-CO stretching and the 1924 and 1970 cm Ϫ1 bands to C-O stretching.
On the basis of the empirical CO -environment correlation obtained with myoglobin (Mb), the CO frequencies observed at 1910 -1930, 1940 -1950, and 1960 -1970 cm Ϫ1 correspond to CO with strong hydrogen bonding interaction, moderate hydrogen bonding interaction, and neutral surrounding, respectively (21). Accordingly, we assign the major (1970 cm Ϫ1 ) and minor bands (1924 cm Ϫ1 ) to non-hydrogen bonded and strongly hydrogen bonded C-O modes of the heme Fe-CO, respectively. The detection of the major band has been reported previously for the Ec DOSH (18 -20), but the 1924 cm Ϫ1 band was not reported before. This is presumably due to overlap of the isotope-shifted ( 13 C 16 O) component of the 1970 cm Ϫ1 band and the 1924 cm Ϫ1 component of 12 C 16 O in the isotope-difference spectrum (18 -20). The use of 13 C 18 O isotope allowed us to clarify the presence of the strongly hydrogen-bonded conformer (1924 cm Ϫ1 ).
The low (A) and high frequency regions (B) of RR spectra of the CO complexes for several mutants are shown in Fig. 3 and summarized in Table 1. The mutation of Glu-93 showed no significant effect on the stretching vibration modes. On the other hand, the major C-O mode is downshifted by 3 cm Ϫ1 in M95A spectrum (c) compared with WT spectrum (a). This implies that the M95A perturbs the electrostatic field near the iron-bound CO ligand. The Fe-CO and CO stretching frequencies of F113L (d) are largely perturbed (Table 1), implying that Phe-113 is also affected by CO and thus functioning in recognizing the axial ligand CO. Furthermore, the mutation of  Arg-97 (e) leads to the disappearance of the strongly hydrogenbonded conformer. We note that the broad Raman band near 1924 cm Ϫ1 in the spectrum of R97I (e) is assigned to a heme non-fundamental mode but not to CO stretching. This is because it exhibited no significant change upon isotopic labeling as seen in supplemental Fig. S2.
Thus, the RR results indicate that the strongly hydrogen bonding conformer at ϳ1924 cm Ϫ1 is observed for WT, E93I, M95A, and F113L proteins and is absent only for the R97I mutant. This means that Arg-97 forms a hydrogen bond with the bound CO in the hydrogen bonding conformer.
We also examined NO-isotope dependence in RR spectra of NO-bound form of WT and several mutants (Table 1 and see also supplemental Fig. S3). The 557 cm Ϫ1 band is assigned to the Fe-NO stretching, and the 1580 and 1635 cm Ϫ1 bands to N-O stretching. The 1635 cm Ϫ1 component of the NO-bound form means that NO is placed in a hydrophobic environment without a hydrogen bond, but the other N-O stretching fre-quency (1580 cm Ϫ1 ) is characteristic of NO having a negative charge and being stabilized by strong hydrogen bonding (22)(23)(24). Because the strongly hydrogen bonding conformer is observed for WT, E93I, M95A, and F113L and is absent only for R97I, it is reasonable to suggest that Arg-97 forms a hydrogen bond with the heme-bound NO in the hydrogen bonding conformer.
Time-resolved Resonance Raman Investigations of WT Ec DOSH-TR 3 spectroscopy has been successfully used to characterize photodissociated heme-CO complexes in various heme proteins (19,(25)(26)(27)(28). It provides information about dynamic structures in the heme and its pocket. In the present study, TR 3 is used to follow dissociation of CO from Ec DOSH heme. The TR 3 spectra for WT Ec DOSH obtained from the pump/probe experiments are displayed for the high (panel A) and low (panel B) frequency regions in Fig. 4, and they are compared with those of the equilibrium CO (a) and reduced (l) forms. In the spectrum of the equilibrium-reduced form (l), the oxidation state marker ( 4 ) appears at 1361 cm Ϫ1 , characteristic frequency of the heme in the reduced state (18 -20). The 2 band is observed at 1581 cm Ϫ1 and the coresensitive band ( 3 ) at 1494 cm Ϫ1 , characteristic of 6c-ls heme ( Table 2). This indicates that heme adopts the 6c-ls state in the reduced form, where Met-95 is the heme axial ligand in the distal side (Fig. 1A).
Upon binding of CO to the reduced heme, Met-95 is displaced by CO and 4 , 3 , and 2 bands are shifted to 1370, 1499, and 1581 cm Ϫ1 , respectively (Fig. 4A, trace a) (19). The spectrum for ⌬t d ϭ Ϫ0.5 s (Fig. 4, trace b) is similar to that observed without the pump beam (Fig. 4, trace a). This indicates that the recombination of CO to photodissociated species is completed in one turn of the spinning cell. In addition, the fact that the Fe-CO band at 488 cm Ϫ1 almost disappeared at ⌬t d ϭ 20 ns and thenceforth (Fig. 4B, traces c-k) demonstrates that photodissociation is achieved by the pump beam.
In the transient species at 20 ns (Fig. 4A, trace c), 4 , 3 , and 2 bands appeared at 1353, 1470, and 1559 cm Ϫ1 , respectively ( Table 2). These bands are similar to those observed for the equilibrium-reduced form of Mb in which heme adopts the 5c-hs structure (27). Accordingly, the heme of Ec DOSH adopts the 5c-hs structure immediately after CO photolysis. The formation of 5c-hs heme after CO photolysis is consistent with previous studies (19,29,30). However, at 50 s (spectrum h) after CO photolysis, new 4 , 3 , and 2 bands appeared at 1361, 1494, and 1581 cm Ϫ1 , respectively, indicating the generation of 6c-ls heme. The component of 5c-hs heme almost disappeared at 100 s (spectrum i), and then 6c-ls heme is dominant. In addition, spectra i-k are similar to that of the equilibrium-reduced form (l). Therefore, the conversion of the 5c-hs to the 6c-ls heme after CO photolysis is attributed to binding of Met-95 to the heme. The binding of Met-95 to the heme at ϳ100 s is consistent with the flash photolysis results in which Met-95 binding occurs in competition with bimolecular CO binding, and subsequent replacement of Met-95 by CO in the millisecond time scale (30).
Furthermore, the spectrum of the equilibrium-reduced form of WT (Fig. 4B, l) displays Raman bands at 343, 380, and 413 cm Ϫ1 , which can be assigned to 8 , propionate bending  ␦(C ␤ C c C d ), and vinyl bending modes ␦(C ␤ C a C b ), respectively. These band positions are the same as those previously reported (19) and are shifted to 343, 377/384, and 409/427 cm Ϫ1 upon CO-binding (a) ( Table 2). It is evident from Table 2 that the propionate bending mode, ␦(C ␤ C c C d ), is shifted to ϳ363, ϳ375, and ϳ380 cm Ϫ1 at 0.02 (c), 1.0 (e), and 100 s (i), respectively, after photolysis (see also supplemental Fig. S4). The frequency shifts suggest that the interactions of the propionate side chains with surrounding residues at these delay times are different. On the other hand, the vinyl bending mode is shifted to 408 and 413 (broad) cm Ϫ1 within 0.02 (c) and 100 s (i) delay, respectively. This result implies that significant alteration in the interactions of the vinyl groups with the surrounding residues occurs in 0.02-100 s after CO photolysis. Thus, the environment around the heme peripheral groups was changed following CO dissociation. The TR 3 results of M95A mutant (Table 3 and supplemental Fig. S5) demonstrate that the photodissociated transient spectra returned to that of the starting CO-bound form directly without passing through the reduced form. Thus, it confirmed that binding of Met-95 to heme is the origin of the large spectral changes observed for WT at 100 s after CO photolysis. Specifically, in the absence of Met-95, CO begins to rebind to the heme near 100 s as indicated by the increase in intensity of the Fe-CO stretching mode at 488 cm Ϫ1 . In addition, in 1 ms after photolysis, the spectrum is similar to that observed for the equilibrium CO-bound form.
Time-resolved Resonance Raman Investigations of Arg-97 Mutants-Met-95 is heme-coordinated in the reduced form, and the side chain of Arg-97 is oriented toward the surface of the protein (Fig. 1A). Whereas, in the O 2 -bound form, Met-95 is pointing toward the protein surface and Arg-97 reorients inward to the heme distal pocket to provide stabilization to heme-coordinated O 2 (Fig. 1B). Thus, it is possible that the side-chain rotation of Arg-97 in the opposite direction upon ligand dissociation plays a role in regulating the Met-95 binding to the heme upon photodissociation of CO. To examine this idea, in particular the role of the side chain of residue 97, we have prepared R97I, R97A, and R97E variants, and performed their TR 3 measurements. In the case of the R97I mutant ( Fig. 5 and Table 3), spectra c-g (panel A) show 5c-hs heme, for which the 4 , 3 , and 2 bands appeared at 1351, 1471, and 1559 cm Ϫ1 , respectively, similar to those observed for WT. In addition to these, new 4 , 3 , and 2 bands appeared at 1369, 1497, and 1580 cm Ϫ1 , respectively, at 50 s (spectrum h) after CO photolysis.   These frequencies are similar to those observed for the equilibrium CO-bound form (a), implying generation of 6c-ls heme. Furthermore, the Fe-CO band appeared at 50 s, and its intensity gradually increased till 1 ms (panel B, spectra h-k). Thus, the conversion from the 5c-hs to 6c-ls state after CO photolysis is attributed to rebinding of CO to the heme, although a small fraction of 5c-hs heme remained till 1 ms as indicated by the bands at 1351, 1471, and 1559 cm Ϫ1 . Therefore, CO, but not Met-95, rebinds to the heme after the photolysis of CO in the R97I mutant. These results demonstrate that the side chain of Arg-97 is important in regulating the motion of Met-95 in protein dynamics.
On the other hand, the mutation of Arg-97 by Glu (R97E) produces significantly different dynamics ( Fig. 6 and Table 3) than those observed for WT (Fig. 4). Specifically, instead of 5c-hs heme, both 5c-hs (major; 1355, 1472, and 1557 cm Ϫ1 ) and 6c-ls (minor; 1362, 1494, and 1581 cm Ϫ1 ) species were simultaneously observed at a 20 ns delay (panel A, spectrum c). The transient spectra contain both 5c-hs and 6c-ls species till 1 ms. This means that the binding of Met-95 to the heme is not complete till 1 ms (panel A, c-k), whereas Met-95 binding occurs within 100 s in WT (Fig. 4).
Furthermore, we performed similar experiments for R97A (Table 3 and see also supplemental Fig. S6) and the results suggested that the binding of Met-95 and the rebinding of CO to the heme after CO photolysis are faster in R97A variant than those occurring in WT. Thus, the mutation of Arg-97 by a residue with different side chains significantly perturbs the movement of Met-95 in Ec DOSH.

Time-resolved Resonance Raman Investigations of Phe-113 Mutants-
To further investigate the role of Phe-113, especially in the dynamics of Ec DOSH protein, we have performed TR 3 experiments for F113L mutant, and the results are displayed in Fig. 7 and Table 3. At 20 ns   JULY 4, 2008 • VOLUME 283 • NUMBER 27 after photolysis of CO, the 5c-hs heme is formed as indicated by the 4 , 3 , and 2 bands at 1353, 1471, and 1559 cm Ϫ1 , respectively. The 5c-hs heme species was dominant till 10 s, but thenceforth new 4 , 3 , and 2 bands appeared at 1361, 1497, and 1581 cm Ϫ1 , respectively, meaning that 6c-ls heme is generated. These results are similar to those observed for WT (Fig. 4). Accordingly, these bands arise from the Met-95-bound heme. The transient 4 band is broad, and its band fitting analysis suggested the presence of three 4 bands at 1353, 1361, and 1372 cm Ϫ1 at 100 s delay (i). The 1353 and 1361 cm Ϫ1 bands correspond to the 5c-hs and 6c-ls (Met-95 binding) heme, respectively. Because the Fe-CO band appeared at 100 s and its intensity gradually increased till 1 ms (panel B, spectra i-k), we suggest that the 4 band at 1372 cm Ϫ1 arises from the 6c-ls heme in which CO is the sixth ligand of heme. The transient spectrum at 200-s delay (panel A, spectrum j) also showed the simultaneous presence of 5c-hs and the Met-95-and CO-bound forms. The intensity of the 4 band at 1372 cm Ϫ1 corresponds to the CO-bound form and is more increased at 1 ms (spectra k) delay, whereas that of the 5c-hs state (1353 cm Ϫ1 ) almost disappeared. Thus, Met-95 and CO bind to the 5c-hs heme generated upon photolysis of CO. Similar results were observed for the F113T mutant (supplemental Fig. S7). Thus, the mutation of Phe-113 significantly perturbs the Ec DOSH protein dynamics.

Time-resolved Resonance Raman Study of Ec DOS
Phosphodiesterase Activity-We have shown recently the role of distal residues in catalysis (15). For instance, Met-95 coordination to the reduced heme is critical for locking the system and that global structural change around Met-95 caused by the binding of the external ligands releases the catalytic lock and activates catalysis. In the present study, we examined the importance of Phe-113 for the catalytic reaction. We prepared the fulllength mutants for Phe-113 (F113L and F113T) and determined their PDE activities under anaerobic conditions. The results are summarized in Table 4 (see also supplemental Fig. S8). It is evident from Table 4 that the binding of O 2 , CO, or NO to the reduced heme of WT enhances the activity by 6-to 7-fold (15), but the enhancement depends little on a ligand species. The ligand-free reduced form of F113L and F113T exhibited the activities higher and lower than that of WT, respectively. On the other hand, the ligand-bound forms of Phe-113 mutants exhibited similar activities but higher than those of WT irrespective of a ligand species. Thus, our data clearly demonstrate that the mutation of Phe-113 perturbs the PDE activities. This implies that the side chain of Phe-113 plays an important role in determining the conformation of the heme pocket related to that of the catalytic site toward 3Ј,5Ј-cyclic diguanylic acid.

Interactions between Heme-bound Ligand and Distal
Residues-In the present study, we have investigated the interactions of heme-bound CO and NO with the surrounding residues by monitoring the RR spectra for several mutants (E93I, M95A, R97I, and F113L). These mutants were designed to perturb the electrostatic field near the iron-bound gaseous ligand and also allow us to investigate the communication pathway between distal residues of the protein and heme. We found the formation of both hydrogen-bonded and non-hydrogenbonded conformations in the CO-and NO-bound forms for WT, E93I, M95A, and F113L proteins. On the basis of the similarity of the CO-and NO-bound spectra of E93I to those of

TABLE 4 Comparison of the PDE activities for WT and Phe-113 mutants
The errors were estimated from three measurements for variants but those for WT were taken from Ref. 15, and the PDE values were estimated in a different way as discussed in Ref. 15. WT, we conclude that no communication pathway exists between Glu-93 and heme-bound CO or NO. On the other hand, the mutation of Met-95 alters only C-O but not N-O frequency. The Fe-O2 frequency (560 cm Ϫ1 ) (19) is not affected by mutation of Met-95 (supplemental Fig. S9). These results may be related to the binding geometry of different gaseous ligands, where CO binds to the heme iron linearly, whereas a bent geometry is known for both O 2 -and NO-bound forms (31). These results also suggest that the distal residues resulting from the M95A mutation weakly interact with the bound ligand species and perturb the ligand field only a little. Surprisingly, the PDE activity of the ligand-free form of M95A mutant was similar to those of ligand-bound forms irrespective of the ligand species (15) and similar to that of the ligand-bound forms of other mutants (Table 4). Accordingly, it is deduced that the protein conformation of heme pocket (FG loop, including the locations of Arg-97 and Phe-113), which is generated when Met-95 is displaced from the distal coordination site of heme iron, is critical to make the PDE catalytic domain of Ec DOS active. In other words, Met-95 coordination to heme locks the protein conformation in an enzymatically inactive form (15), and such conformation is somewhat relaxed in F113L.

Ec DOS Fe(II) Fe(II)-O 2 Fe(II)-CO Fe(II)-NO
The present RR results indicate that Arg-97 forms a hydrogen bond with heme-bound CO or NO in the hydrogen-bonded conformation. In addition, the absorption spectra of O 2 -bound form of Arg-97 mutants of Ec DOSH could not be detected due to rapid auto-oxidation and/or a low affinity for O 2 . Similar results were observed for the Arg-97 mutants of the full-length Ec DOS (15). Thus, it is likely that the hydrogen bonds from Arg-97 play a critical role for protecting the heme from oxidation by O 2 and thus stabilizing the heme-bound O 2 , but this interaction may not be essential for CO-or NO-bound forms. Recently, Tanaka et al. reported the effects of Arg-97 mutation on the PDE activity (15). Interestingly, the PDE activities of Arg-97 mutants depend on the side chain of the 97th residue. For instance, the PDE activities of the CO-and NO-bound forms of R97A are similar to those of WT, whereas those of R97I and R97E are increased (by 24 -28%) and decreased (by 29 -47%), respectively, compared with those of WT (15). Because the geometry and electric polarization of heme-bound ligand depend on the ligand species, the observed facts indicate the importance of interactions between Arg-97 and hemebound ligand.
The F113L mutant significantly perturbs the Fe-CO , Fe-NO , Fe-O2 , CO , and NO vibrations. A Phe residue has been known to stabilize polar ligands in several natural and variant Mbs, including elephant Mb (32)(33)(34). Similarly, Phe-113 would be critical for stabilizing the heme-bound ligands in the heme distal pocket of Ec DOSH protein. In the reduced form, the mainchain peptide oxygen of Gly-94 of the rigidified FG loop forms hydrogen bonds with the main-chain oxygen of Phe-113 and nitrogen of Leu-115 from H ␤ sheet via a water molecule (10,11). This hydrogen bond network is disrupted upon O 2 binding (10). In addition, the crystal structures of Ec DOSH reveal that there is noticeable movement of the H ␤ sheet upon O 2 binding. Specifically, the root mean square deviation of C ␣ atoms in the H ␤ sheet (residues 108 -120) with respect to the reduced form was 0.41 Å for the O 2 -bound form. The root mean square devi-ation was calculated with Swiss-PdbViewer (35). These observations are compatible with our previous UV resonance Raman results in which Trp-110 from the H ␤ sheet undergoes environmental changes upon O 2 binding (36). Furthermore, the PDE activities of ligand-bound forms of Phe-113 mutants were enhanced by 25-55% compared with those of WT (Table 4), demonstrating that the interactions of Phe-113 with bound ligand perturb the protein conformation to regulate the function of Ec DOS. We also propose that the interactions of Phe-113 with bound ligand would be involved in communicating the heme structural changes to the H ␤ sheet upon ligand binding.
Activation via Heme Peripheral Groups-The propionate bending modes, ␦(C ␤ C c C d ), of the equilibrium-reduced and CO-bound Ec DOSH were observed at 380 and 377/384 cm Ϫ1 , respectively. The ␦(C ␤ C c C d ) mode of the O 2 -bound Ec DOSH was also observed at 384 cm Ϫ1 (not shown). The frequency of ␦(C ␤ C c C d ) is indicative of hydrogen bonding between the heme propionates and the surrounding residues (37). In Mb, the heme 7-propionate constitutes a hydrogen bond with His-97 and Ser-92, and the ␦(C ␤ C c C d ) mode appears at 376 cm Ϫ1 . The disruption of this hydrogen bond by mutations leads to downshifts of the band by ϳ10 cm Ϫ1 (38). In the O 2 -bound form of Ec DOSH, the 7-propionate makes strong hydrogen bonds with Arg-97, whereas, in the reduced form, it makes weak hydrogen bonds with the backbone NH of Met-95 and two water molecules (10). This structural change is compatible with the higher frequency of the ␦(C ␤ C c C d ) mode (384 cm Ϫ1 ) in the O 2 -or CO-bound forms.
A schematic model for the structural change occurring upon ligand photodissociation is illustrated in Fig. 8. After CO photolysis of the CO-bound WT Ec DOSH, the 5c-hs species is formed in which no new prominent bands are observed in the 250 -450 cm Ϫ1 region in the picosecond TR 3 spectra for Ec DOSH (19). This indicates that the heme peripheral modes of the transient species are identical to those of the CO-bound form in this timescale. After a 20 ns delay, however, a weak ␦(C ␤ C c C d ) band appeared at 363 cm Ϫ1 . Because the lower ␦(C ␤ C c C d ) frequency accounts for a weak (or no) hydrogen bond of heme 7-propionate, we suggest that the hydrogen bond between 7-propionate and Arg-97 is cleaved within 20 ns of photolysis (Fig. 8).
The 363 cm Ϫ1 band in the WT spectrum is upshifted to 375 and 380 cm Ϫ1 within 1.0 and 100 s, respectively. Because the ␦(C ␤ C c C d ) frequency of 380 cm Ϫ1 at 100-s delay was observed for the equilibrium-reduced form (Fig. 4B, spectrum l) in which Met-95 is hydrogen-bonded with 7-propionate, the 380 cm Ϫ1 band would arise from the hydrogen-bonded species, and its formation must be completed by 100 s after CO-photolysis. Then, what is the origin of the 375 cm Ϫ1 band observed at 1 s? The crystal structure showed that the 6-propionate constitutes hydrogen bonds with Asn-84 and Gly-94 in the reduced form, but it forms a hydrogen bond only with Asn-84 in the O 2 -bound form, for which Gly-94 from FG loop moves far away from the 6-propionate (10). Thus, we speculate that the formation of a hydrogen bond between Gly-94 and heme 6-propionate is the origin of the 375 cm Ϫ1 band observed at 1 s (Fig. 8). Therefore, we propose the following model upon photolysis of CO from heme: the hydrogen bond networks from both heme propi-

Time-resolved Resonance Raman Study of Ec DOS
onates experience changes, leading to large conformational changes of protein moiety. Specifically, Gly-94 and Met-95 move inward to the heme distal pocket, while Arg-97 orients toward the protein surface as a result of the changes in the propionate hydrogen bonding interactions upon CO photolysis, communicating the heme structural changes to the FG loop. In addition, the inward movement of Gly-94 would reform its hydrogen bonding network with Phe-113 and Leu-115 via a water molecule (10,11), communicating the heme structural changes to the H ␤ sheet as we proposed earlier.
Recently, we have shown that the heme structural changes upon ligand binding in Mb are communicated to the globin moiety through heme propionates (39). In addition, the heme propionates play a crucial role in signaling mechanism of FixL (4) and HemAT-Bs (40) proteins upon ligand binding/dissociation. Similarly, the present results for Ec DOSH strongly suggest that the structural changes of heme upon ligand binding are communicated from the peripheral propionates to the FG loop.
Furthermore, the vinyl bending modes, ␦(C ␤ C a C b ), of the equilibrium-reduced and CO-bound Ec DOSH were observed at 413 and 409/427 cm Ϫ1 , respectively. In the transient states, the ␦(C ␤ C a C b ) bands appeared at 408 and 413 cm Ϫ1 at 0.02 and 100 s, respectively. This indicates the presence of different intermediates in which the interactions of vinyl side chains of heme with surrounding residues are altered upon CO-photo-dissociation. Thus, the cleavage of the hydrogen bond between 7-propionate and Arg-97 at 20 ns and the formation of a new hydrogen bond between 7-propionate and Met-95 at 100 s are accompanied by structural changes around heme vinyl groups. These results are in agreement with our previous UV resonance Raman study in which the heme structural changes are communicated through the 2-vinyl group to Trp-53, an important residue for function of Ec DOS (36).
Distal Residues Regulate Met-95 Binding to the Heme-Liebl and co-workers (30) reported the CO-rebinding kinetics upon its photodissociation from Ec DOSH. They showed that Met-95 binding to the heme occurs in 100 s in competition with bimolecular CO recombination, and subsequent replacement of Met-95 by CO occurs at ϳ8 ms (30). In contrast, in other heme proteins with a hexa-coordinate heme, such as CooA (41) and neuroglobin (42), the binding of an internal ligand does not occur prior to CO-recombination. To understand why Met-95 binding occurs before the recombination of CO in Ec DOSH protein, it is essential to investigate the role of the distal residues such as Arg-97 and Phe-113 (Fig. 1) in the regulation of protein dynamics.
Arg-97 is located on the protein surface in the ligand-free reduced form, but it reorients inward to the heme distal pocket in the O 2 -bound form to provide stabilization to heme-coordinated O 2 (Fig. 1). Thus, it is possible that the rotation of Arg-97 in opposite directions upon ligand dissociation plays a role in regulating the Met-95 binding to the heme. To examine the role of Arg-97, in particular its electrostatic interactions, we mutated Arg-97 by different side chains (Ala, Ile, and Glu) and performed TR 3 experiments for these mutants.
The results indicate that the mutation of Arg-97 significantly perturbs the Met-95 binding to the heme upon photodissociation of CO. Specifically, in the R97I mutant, in which Ile-97 does not provide electrostatic interactions with its surroundings but has a bulky side chain, Met-95 does not bind to the heme but CO rebinds. This must be related to the PDE activities of R97I, which are higher than those observed for WT by 24 -28% (15). On the other hand, Met-95 binding is very slow in the R97E mutant compared with that observed for WT, implying that the negative charge in side chains of Glu-97 in R97E probably raises the potential barrier for rebinding of CO. The destabilization of the CO-bound form in R97E is compatible with its PDE activity, which is lower than that of WT by 29 -47% (15). In Mb, the distal His-64 forms a hydrogen bond with heme-bound O 2 similar to Arg-97 in Ec DOSH (43). In addition, the mutation of His-64 in Mb by apolar residues causes a marked decrease in oxygen affinity and also perturbs the kinetics of CO rebinding (17,31), indicating the importance of the electrostatic interactions between distal His-64 and heme-bound O 2 . Similarly, the electrostatic interactions of Arg-97 with the heme-bound ligand of Ec DOSH would be crucial for regulating the binding of sixth ligand of heme.
Furthermore, we have mutated Phe-113 by polar (Thr) and apolar (Leu) residues to elucidate the role of Phe-113 in the protein dynamics. The mutation of Phe-113 by either Thr or Leu produces similar results but significantly different from those observed for WT. For instance, two conformations are generated upon CO-photolysis and in transient states, Met-95 and CO competitively bind to the heme. The results suggest that the steric but not polar interactions of Phe-113 are substantial for Met-95 binding to the heme prior to CO recombination.

CONCLUSIONS
We have investigated the interactions between the hemebound ligand and some surrounding residues for different ligand species. The RR results showed that Arg-97 and Phe-113 are the major contributors for these interactions, whereas the TR 3 results indicate that the heme propionates experience large structural changes upon photolysis of CO. On the basis of these results, we propose a model for the role of the heme propionates in communicating the structural changes from the heme to the protein moiety. We also point out that the electrostatic interaction of Arg-97 and the steric interaction of Phe-113 are crucial for controlling Met-95 binding to the heme prior to CO recombination.