Binding of oxygen and carbon monoxide to a heme-regulated phosphodiesterase from Escherichia coli. Kinetics and infrared spectra of the full-length wild-type enzyme, isolated PAS domain, and Met-95 mutants.

The heme-regulated phosphodiesterase, Ec DOS, is a redox sensor that uses the heme in its PAS domain to regulate catalysis. The rate of O(2) association (k(on)) with full-length Ec DOS is extremely slow at 0.0019 microM(-1) s(-1), compared with >9.5 microM(-1) s(-1) for 6-coordinated globin-type hemoproteins, as determined by the stopped-flow method. This rate is dramatically increased (up to 16-fold) in the isolated heme-bound PAS domain. Dissociation constants (K(d)) calculated from the kinetic parameters are 340 and 20 microm for the full-length wild-type enzyme and its isolated PAS domain, respectively. Mutations at Met-95 in the isolated PAS domain, which may be a heme axial ligand in the Fe(II) complex, lead to a further increase in the k(on) value by more than 30-fold, and consequently, a decrease in the K(d) value to less than 1 microM. The k(on) value for CO binding to the full-length wild-type enzyme is also very low (0.00081 microM(-1) s(-1)). The kinetics of CO binding to the isolated PAS domain and its mutants are similar to those observed for O(2). However, the K(d) values for CO are considerably lower than those for O(2).

The phosphodiesterase (PDE) 1 from Escherichia coli, Ec DOS, is composed of an N-terminal heme-bound PAS domain and a C-terminal PDE catalytic domain (1). The basic physicochemical characteristics and function of this enzyme have been partially elucidated by our group and that of Kitagawa et al. (1,2). PDE activity is dependent on the redox state of Ec DOS in that the enzyme is active only when the heme is in the Fe(II) state. Changes in the redox state of the heme bound to the N-terminal PAS domain may induce a subtle conformational change, which intramolecularly transmits signals to the C-terminal PDE domain to initiate and/or regulate catalysis. Ec DOS therefore constitutes a novel class of heme enzymes designated "heme-based sensors" (3)(4)(5). These include proteins such as FixL (6,7), CooA (8,9), sGC (10,11), and Hem-AT (12,13). In these enzymes, association or dissociation of the exogenous axial ligand (O 2 , CO, or NO) from the heme iron leads to protein conformational changes, which in turn transmit signals to other domains to regulate catalysis or binding to DNA. The Ec DOS signal transducing mechanism appears to be unique, since changes in the redox state of the PAS domain, rather than iron coordination chemistry, are responsible for signal transduction (1). However, in other words, signal transduction triggered by ligand binding (CO and NO) is common to Ec DOS and FixL, based on the finding that CO or NO binding abolishes catalysis by Ec DOS (1).
The physicochemical properties of the isolated heme-bound PAS domain of Ec DOS were initially characterized by Gilles-Gonzales and colleagues (14). They reported that Ec DOS is a direct O 2 sensor enzyme that takes advantage of its characteristic O 2 binding affinity to modulate catalysis. The association rate constant (k on ) for O 2 (0.0026 M Ϫ1 s Ϫ1 ), CO (0.0011 M Ϫ1 s Ϫ1 ), and NO (0.002 M Ϫ1 s Ϫ1 ) and dissociation constants (K d ) for O 2 (13 M) and CO (10 M) were similar for the isolated PAS domain (14). To date, the corresponding association kinetics and equilibrium constants have not been reported for the fulllength Ec DOS. The same group characterized a PDE from Acetobacter xylinum (AxPDEA1) and suggested that this enzyme is O 2 -sensitive, based on the finding that O 2 binding suppressed PDE catalysis by 70% in the Fe(II) enzyme (15). AxPDEA1 displayed higher k on and k off values for O 2 (6.6 M Ϫ1 s Ϫ1 and 77 s Ϫ1 , respectively) than the isolated PAS domain of Ec DOS, despite similar K d values (12 M).
The kinetics of exogenous axial ligand binding to the heme protein and associated equilibrium constants provide useful information on the structure and characteristics of the heme distal site and ligand access channel (16 -19). It is important to study O 2 and CO binding, particularly to full-length Ec DOS, to clarify whether the enzyme is a direct O 2 sensor. These analyses would also be useful in elucidating the structure of the heme distal site and ligand access channel and their relation to the signal transduction mechanism. Based on the amino acid sequence alignment and crystal structure of a similar PAS enzyme, FixL (7,20,21), Met-95 is suggested as a heme axial ligand trans to His-77.
In the present study, we report rate and equilibrium constants for O 2 and CO binding to the full-length enzyme in the Fe(II) state, isolated heme-bound PAS domain, and Met-95 * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cloning and expression in E. coli and purification of full-length Ec DOS (amino acids 1-807) and the isolated heme-bound PAS domain (amino acids 1-147) were performed as described previously (1,2). Site-directed cassette mutagenesis was performed using oligonucleotides. The sequence was confirmed by Sanger's method using an automatic sequencer DSQ-2000L (Shimadzu Co.). Ec DOS-PAS gene mutants were expressed in BL21 E. coli cells. There are slight differences in isolation of the isolated heme-bound PAS domain and the full-length enzyme. Briefly, the full-length protein was expressed at 20°C to avoid proteolytic digestion, whereas the isolated heme-bound PAS domain was expressed at 25°C. The yield for purified full-length Ec DOS was about 10% that for the isolated heme-bound PAS domain from the same culture scale. Full-length Ec DOS, isolated PAS domain, and Met-95 mutants were more than 95% homogenous, as evaluated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.
Optical Absorption, Stopped-flow, and Flash Photolysis Spectra-Spectral experiments were performed under aerobic conditions on Shimadzu UV-1650, UV-2500, and Hitachi U-2010 spectrophotometers maintained at 25°C by a temperature controller. Anaerobic spectral experiments were conducted on a Shimadzu UV-160A spectrophotometer in a glove box under a nitrogen atmosphere with an O 2 concentration of less than 50 ppm (22)(23)(24). Following reduction of the heme by sodium dithionite, excess dithionite was removed using a Sephadex G-25 column in the glove box. To ensure that the appropriate temperature of the solution was maintained, the reaction mixture was incubated for 10 min, prior to spectroscopic measurements.
To obtain the O 2 and CO association rates, solutions of Fe(II) Ec DOS proteins (containing nearly 10 M heme) reduced with sodium dithionite in 50 mM Tris-HCl buffer (pH 8.0) were rapidly mixed with 260 -1300 M O 2 or 200 -1000 M CO solutions using a stopped-flow spectrophotometer under anaerobic conditions (Otsuka Electronics model RA-401) at 25°C. Formation of O 2 and CO complexes was monitored at 428 and 423 nm, respectively. To measure the O 2 dissociation rate, O 2 -bound Ec DOS was mixed with CO-saturated buffer in the absence of dithionite using a stopped-flow spectrometer. For the full-length wild-type enzyme, the CO-saturated buffer contained a trace of dithionite. To determine the CO dissociation rate, dithionite-reduced CO-bound Ec DOS protein was mixed with NO-saturated buffer in the presence of a trace of dithionite, using a rapid scan spectrometer (Shimadzu model Multispec 1500). The NO gas solution was prepared using the NO donor, NOR-1. The effects of EDTA on ligand binding kinetics were additionally evaluated. No significant differences were observed in the presence or absence of EDTA. At least three experiments were conducted to obtain each rate constant. Regression analyses were performed and lines representing an optimum correlation coefficient were selected (23). Experimental errors were within 20% for O 2 kinetics and 10% for CO kinetics.
To obtain autoxidation rates, solutions of the Fe(III) Ec DOS proteins (containing ϳ10 M heme) were reduced with sodium dithionite in 50 mM Tris-HCl (pH 8.0) buffer containing 1 mM EDTA. Excess dithionite was removed using a Sephadex G-25 column in the glove box. After removal, the protein solution was mixed with air-saturated buffer to form the Fe(II)-O 2 complex in a cuvette. Changes in UV/visible spectra were monitored with respect to time at 25°C, using a rapid scan spectrometer (Shimadzu model Multi-spec 1500).
Laser flash photolysis experiments were conducted in a 10 ϫ 10-mm cell at 25°C using a 433-nm light of OPO of a Nd:YAG laser (Quanta-Ray, GCR-130) producing an excitation power of 40 mJ with a repeated pulse duration of 1.2 s. Details of this apparatus are described elsewhere (25). Kinetic data were analyzed using Igor Pro software (Wave-Metrics). Experiments were repeated at least three times. Data were reproducible, with errors within 20%.
FT IR Spectra-FT IR spectra were obtained using a Bio-Rad FTS-3000MX spectrometer with 2 cm Ϫ1 resolution, employing an accumulation time of 5 min and a set of 3-mm CaF 2 windows with a 0.05-mm Teflon spacer. Proteins were concentrated to 2-3 mM heme using Ultrafree-MC (Millipore Co., Billerica, MA). Protein samples were reduced with 1 mM sodium dithionite and exposed to a stream of CO gas. Fig. 1) were conducted using a stopped-flow spectrometer. The time profile (inset of Fig. 1A) of O 2 binding to full-length Ec DOS was composed of two phases. The rate constant of the fast O 2 association phase was dependent on the O 2 concentration (Fig. 1B). The k on value (0.0019 M Ϫ1 s Ϫ1 ) obtained with Ec DOS was extremely low, compared with those of other PAS heme proteins (BjFixL, RmFixLH, and AxPDEA1) or 6-coordinated Hb such as Arabidopsis trHb, barley Hb, cytoglobin, neuroglobin, and Synechocystis trHb (27)(28)(29)(30) (Table  I). The slow phase was independent of the O 2 concentration. Because the rate of the slow phase is similar to that of autoxidation (described below), it is possible that the phase reflects autoxidation and/or protein relaxation. The dissociation rate constant (k off ) was calculated as 0.64 s Ϫ1 . Accordingly, the K d value for O 2 obtained from the k on and k off values was calculated as 340 M. This value is comparable with that of BjFixL (140 M), but higher than those of other PAS heme proteins, RmFixLH and AxPDEA1 (12-31 M). Interestingly, this value is much higher than that of O 2 binding to 6-coordinated Hb (0.00056 -0.0029 M) ( Table I).

O 2 Association Kinetics-The O 2 association kinetics experiments (shown in
The heme redox state appears to regulate PDE activity, such that only the enzyme in the Fe(II) state is active (1). Conversely, the substrate may affect the structure of the heme binding site. Accordingly, the effect of the substrate, cAMP, on O 2 binding kinetics was examined. The presence of cAMP in our experiments did not significantly alter the k on value.
The k on value of the isolated heme-bound PAS domain for O 2 was increased 16-fold compared with the wild-type protein (up to 0.031 M Ϫ1 s Ϫ1 ). The k off value was estimated as 0.61 s Ϫ1 (Fig. 2), and the K d value was 20 M. Delgado-Nixon et al. (14) reported that the k on , k off , and K d values for O 2 binding to the isolated PAS domain were 0.0026 M Ϫ1 s Ϫ1 , 0.034 s Ϫ1 , and 13 M, respectively. Interestingly, these k on and k off values are significantly lower than those obtained in the present study.
Met-95 of the PAS domain of Ec DOS is a suggested heme axial ligand (7,20,21). Mutation of this residue increased the k on value to more than 1 M Ϫ1 s Ϫ1 , and thus evaluation by the stopped-flow method was unfeasible. The k off values for M95A, M95L, and M95H mutant proteins for O 2 were determined as 0.73, 0.45, and 0.79 s Ϫ1 , respectively. The corresponding K d values are therefore likely to be less than 1 M.
The laser flash photolysis method was applied to the Fe(II)-O 2 complexes to accurately determine the faster association rates of O 2 with the Met-95 mutants. However, the amount of O 2 photodissociation from Ec DOS proteins was extremely low under our experimental conditions. Consequently, establishing the k on values was not feasible. We additionally attempted to measure K d values by monitoring Soret spectral changes upon the addition of O 2 . However, relatively fast autoxidation hampered the estimation of the exact K d values of Ec DOS proteins.
Autoxidation Rates of Fe(II)-O 2 Complexes-To elucidate the local environment of the Ec DOS heme, we evaluated the autoxidation rates of the Fe(II)-O 2 complexes. The autoxidation time profile of all Fe(II)-O 2 complexes studied comprised only one phase. The rates obtained are summarized in Table II. The rate for the full-length wild-type enzyme was 0.015 min Ϫ1 . Addition of the substrate, cAMP, did not significantly alter the rate. However, the rate in the isolated heme-bound PAS domain was decreased by 2.6-fold to 0.0058 min Ϫ1 . The redox stability of the Fe(II)-O 2 complex of the isolated PAS domain was evident during the purification procedure. Specifically, the characteristic Soret peak of the Fe(II)-O 2 complex was occasionally detected in the lysis solution of E. coli expressing the PAS domain, but scarcely observed in the solution of E. coli expressing the full-length wild-type enzyme. M95A and M95L mutations further decreased the rate to 0.0013-0.0017 min Ϫ1 . Interestingly, the autoxidation rate for the M95H mutant was comparable (0.018 min Ϫ1 ) to that of the full-length wild-type enzyme. Met-95 mutations in full-length Ec DOS also led to a decrease in the autoxidation rate. The effects of the mutations in the full-length enzyme were similar to those in the isolated PAS domain. The autoxidation rate of the M95H full-length mutant was similar to that of the wild-type full-length enzyme, whereas those of M95A and M95L mutants were lower, compared with the wild-type full-length enzyme. All the full-length Met-95 mutant proteins display similar autoxidation rates to those of the corresponding isolated PAS domain.
CO Binding Kinetic Constants-To clarify the ligand discrimination mechanism of the heme of Ec DOS, we obtained k on values for the binding of CO to Fe(II) complexes. The k on value for full-length Ec DOS wild-type enzyme was obtained by the stopped-flow method (Fig. 3). The observed time profile was monophasic, and the rate was dependent on the CO concentration. The k on value was determined as 0.00081 M Ϫ1 s Ϫ1 , whereas the k off rate was calculated as 0.0025 s Ϫ1 (Fig. 4). The K d value was estimated from the k on and k off values as 3.1 M. Addition of cAMP to the full-length wild-type enzyme did not significantly alter these kinetic parameters. The k on value determined for the full-length enzyme was much lower than those reported for other PAS proteins and 6-coordinated globins (Table III).
The k on value for CO increased 9.4-fold (up to 0.0078 M Ϫ1 s Ϫ1 ) in the isolated PAS domain, whereas the k off value increased by only 1.8-fold. The k on value for the isolated PAS domain is similar to that of the PAS protein, BjFixL (Table III).    Table III). Mutation of Met-95 markedly increased the k on value for CO, which made it difficult to determine values using the stoppedflow method. We therefore employed flash photolysis to obtain more accurate rate constant values for CO association. The rates observed for the Met-95 mutants were linearly dependent on the CO concentration (Fig. 5). The k on values for the Met-95 mutant proteins were between 3.4 and 9.3 M Ϫ1 s Ϫ1 . However, the k off value was not markedly altered in the isolated heme-bound PAS domain or Met-95 mutants. The k off values for the Met-95 mutants were between 0.0032 and 0.0049 s Ϫ1 (Table  III). The K d values obtained from the k on and k off values were between 0.00053 and 0.00094 M, suggesting that the Met-95 mutations dramatically decrease the K d value of the isolated PAS domain for CO.
Changes in the Soret absorption spectra of Fe(II) complexes upon the addition of CO were additionally evaluated to determine equilibrium K d values for CO. However, exact K d values were not obtained, because estimation of the exact concentration of free CO in the heme solution proved difficult because of the high CO affinity of the enzyme. A marked increase in the CO affinity was also observed with this method for the isolated PAS domain and Met-95 mutants.
C-O Stretching in FT IR Spectra-FT IR spectroscopy was employed to observe the C-O stretching mode in Fe(II)-CO complexes of the isolated PAS domain (Fig. 6, Table IV). The mode was located at 1970 cm Ϫ1 in the wild-type enzyme. This position is similar to those (1969 -1973 cm Ϫ1 ) of other PAS proteins (except RmFixL) and relatively similar to those of CooA (1969 cm Ϫ1 ) and Paramecium caudatum Hb (1974 cm Ϫ1 ), but higher than those of mammalian myoglobin (1945 cm Ϫ1 ) and human hemoglobin (1951 cm Ϫ1 ). Among the Met-95 mutants, only the M95A mutant enzyme displayed a significant change in band position. This finding is surprising, considering that K d and k on values for the association of CO with the PAS domain were affected by the mutations (Table III). The FT IR spectrum of the CO complex of the full-length wild-type enzyme was similar to that of the isolated PAS domain. 2 Notably, in addition to the band at 1970 cm Ϫ1 , a band at 1925 cm Ϫ1 was consistently observed in the spectra of the wild-type enzyme and Met-95 mutants.

DISCUSSION
An extensive number of kinetic and equilibrium studies on the association/dissociation of O 2 and CO to/from myoglobin (including heme distal mutants) have been performed (36,37,38). From the data, it is clear that first, there is no defined relationship between the Fe-C-O geometry and CO affinity. Thus, steric hindrance does not play a dominant role in regulating the CO affinities of the native proteins. Second, a major kinetic barrier to O 2 and CO binding involves the disruption of hydrogen-bonding interactions between the distal histidine and an adjacent water molecule. Third, the electrostatic field surrounding the ligand access channel and bound ligand is important for O 2 and CO association kinetics, i.e. distal pocket residues regulate the binding of O 2 and CO primarily by electrostatic interactions. Moreover, autoxidation at high O 2 concentrations is caused by direct dissociation of the neutral superoxide radical (HO 2 ) from O 2 myoglobin. This process is accelerated by decreasing the pH. Finally, the stretching frequency of bound CO serves as an empirical voltmeter that can be used to measure the polarity of the distal pocket and predict the extent of electrostatic stabilization of bound O 2 . Ec DOS is a PAS protein in which the heme is surrounded by five antiparallel ␤-sheets on the distal side and one ␣-helix on the proximal side (3, 5-7), compared with globins, which are mainly composed of ␣-helices. Additionally, Ec DOS has a 6-coordinated heme, unlike myoglobin and hemoglobin that contain a 5-coordinated heme. Despite these structural differences, Ec DOS and globins would share several common properties.
Combination by exogenous ligand such as O 2 and CO with a Fe(II) 5-coordinate species in equilibrium with the 6-coordinate species with endogenous ligands is important to consider kinetic mechanisms of the wild-type Ec DOS enzyme. The Gilles-Gonzalez group (20) and our laboratory (1, 2) have established that the heme-axial residue displaced by the exogenous ligand is Met-95. Our laboratory (1, 2) further confirmed that Fe(II) full-length wild-type Ec DOS is 6-coordinated and that Met-95 is the displaced ligand by performing comparisons with FixL. Note that M95A and M95L mutants are in the high-spin state, whereas the M95H mutant is in the low-spin state. Therefore, exogenous ligands bind to a 5-coordinated species generated from the 6-coordinated species in the Ec DOS wild-type and M95H mutants similar to Arabidopsis trHB, barley Hb, cytoglobin, neuroglobin, and Synechocystis trHb (Tables I and III). In contrast, exogenous ligands bind to the 5-coordinated spe-  cies in the M95A and M95L mutants, analogous to Hb ␣-chain, Hb ␤-chain, and myoglobin.
The Gilles-Gonzalez group (14) reported that Ec DOS is an oxygen sensor PDE whereby the heme senses O 2 binding, which triggers a protein conformational change and transfers a signal to the PDE domain to regulate catalysis. This theory is based on the FixL signal transduction mechanism in which O 2 dissociation from the heme in FixL initiates His kinase activity on the catalytic domain (3)(4)(5)(6)(7). The catalytic mechanism is initiated as a result of preferred CO binding over O 2 , i.e. K d values for O 2 and CO are 140 and 9 M, respectively, for FixL. Another PDE A1 protein, AxPDEA1, from A. xylinum, binds the Fe(II) heme as the active form. O 2 moderately suppresses the activity of this enzyme, but does not cause total inhibition (15). Gilles-Gonzalez et al. (14) reported that the k on (0.0026 M Ϫ1 s Ϫ1 ), k off (0.034 s Ϫ1 ), and K d ( (20). In their report, Gonzalez et al. (20) provide an excellent discussion on the mechanism that gives rise to such kinetics and the consequences for oxygen affinity.
The overall k on for O 2 is governed roughly by the rate of diffusion into the protein, the binding site, and the rate of iron-oxygen bond formation (38). The k on value for O 2 binding to the full-length wild-type protein was very low, whereas the K d value calculated from the k on and k off values was very high at 340 M (Table I) (Table I). The binding affinity of O 2 for the Fe(II) heme in the PAS domain is very low compared with those of other 6-coordinated (Arabidopsis trHb, barley Hb, cytoglobin, neuroglobin, and Synechocystis trHb) (27)(28)(29)(30) and 5-coordinated globin-type hemoproteins (29) (Table I). This appears to be a key characteristic of PAS proteins.
The Met-95 mutations in the isolated PAS domain significantly increased the k on value and decreased the K d value. These observations may be explained by the theory that Met-95 is an axial ligand in the Fe(II) heme. Mutation of Met-95 to Ala and Leu will result in 5-coordinated heme, enabling direct binding of exogenous ligands without prior dissociation of an endogenous ligand. It is additionally possible, particularly in the case of the M95H mutant, that the O 2 access channel is widened or the electrostatic character of the O 2 access channel is altered, similar to that observed in myoglobin distal mutants. Liebl et al. (39) reported that the decay phase of geminate recombination of O 2 with the isolated PAS domain is 5.3 ps. This value is not significantly different from those obtained with FixL and myoglobin (39). For myoglobin, the overall k off for O 2 is governed equally by the rate of Fe(II)-O 2 bond disruption and the rate of O 2 dissociation from the protein (38). Similar factors may contribute to the O 2 k off value of the Ec DOS protein. However, isolation of the PAS domain and mutation of Met-95 do not appear to influence these factors.
The autoxidation rate of the full-length wild-type enzyme is 0.015 min Ϫ1 (Table II). Therefore, even if O 2 is bound to the Fe(II) form of full-length Ec DOS, the Fe(II)-O 2 complex is easily oxidized to the Fe(II)-O 2 form. The autoxidation rate of AxPDE A1 is less than 0.001 min Ϫ1 (15), and consequently, the effect of O 2 on catalysis is easily observed. For FixL, dissociation of O 2 from the Fe(II)-O 2 complex initiates His kinase activity, and the effect of O 2 on catalysis is readily detected (6,7,26,43). However, it is not feasible to examine the effect of O 2 on catalysis for Ec DOS, because the rapid autoxidation rate of the full-length enzyme hampers detection of catalytic activity. Isolation of the PAS domain greatly decreased the autoxidation rate, suggesting that a specific steric and/or electrostatic effect, which destabilizes the Fe(II)-O 2 complex in the full-length wild-type enzyme, was removed on isolating the PAS domain, leading to stabilization of the Fe(II)-O 2 complex. Perhaps isolation of the PAS domain decreases the accessibility of the distal pocket to either solvent water molecules or another polar residue(s), which inhibits protonation of the Fe(II)-O 2 complex (38). Mutations at Met-95 to Ala and Leu further stabilized the Fe(II)-O 2 complex and greatly decreased the autoxidation rate of the isolated PAS domain. The M95A mutation in the fulllength enzyme has a similar effect. Because Met-95 axially ligates to the Fe(II) complex, an indirect interaction between the Met-95 side chain and heme-bound O 2 via a water molecule or polar residue may accelerate autoxidation of the wild-type enzyme. Similarly, a direct or indirect interaction between the His side chain and bound O 2 in the M95H mutant may destabilize the Fe(II)-O 2 complex and enhance the autoxidation rate, compared with the other Met-95 mutant proteins.
CO Binding Characteristics-The k on value for CO binding to the full-length wild-type enzyme is lower than those of the other PAS proteins (BjFixL, RmFixLH, and AxPDEA1) and considerably lower than those of the 6-coordinated globins (Table III). The K d value of the full-length wild-type enzyme is comparable with those of BjFixL and RmFixLH, but considerably higher than those of the 6-coordinated globins. Therefore, the k on and K d values of full-length Ec DOS are within the range of those of the PAS proteins. The unusual kinetic and equilibrium binding constants observed for Ec DOS define the character of this enzyme in relation to other PAS proteins. Interestingly, the CO binding affinity is higher in the isolated PAS domain. It is possible that the catalytic domain hinders CO binding by blocking the CO access channel, as observed for O 2 . The k on value for CO binding to the isolated PAS domain was 7-fold higher than that obtained by the Gilles-Gonzalez group (14). Similarly, the K d value for CO of the isolated PAS domain obtained in the present study was 17-fold lower than the reported value. We additionally observed a marked difference in the cyanide binding rate and affinity for the isolated PAS domain (40), compared with the data obtained by Gilles-Gonzalez and colleagues (20).
The laser flash photolysis time profile for the CO complex of the 6-coordinated isolated PAS wild-type enzyme is composed of three phases: 1) a very fast (picosecond order) phase, which is independent of the CO concentration; 2) a CO-dependent fast reaction (about 70% of the total absorption change) with k on ϭ 19 M Ϫ1 s Ϫ1 , and 3) a CO-dependent slow reaction (about 10% of the total absorption change) with k on ϭ 0.071 M Ϫ1 s Ϫ1 . Ligand binding to the 5-coordinated species is observed when flash photolysis is applied to the 6-coordinated species and endogenous ligand is dissociated (27-30, 50, 51). Under specific conditions, the rate of CO binding to the 5-coordinated species is linearly dependent on the concentration of CO. We do not know the association rate of Met-95 dissociated by laser flash, although Vos et al. (39,44) reported that the rebinding rate of dissociated Met-95 of the isolated PAS domain of Ec DOS is of the order of picoseconds. We speculate that the fast CO-dependent reaction is because of a combination of endogenous axial ligand (Met-95) re-association and CO association reactions, whereas the slow CO-dependent reaction is a result of a combination of axial ligand (Met-95) dissociation and CO association (27)(28)(29)(30)41). Further analyses on the fast reactions are currently underway.
The Met-95 mutations markedly increased the CO association rate obtained by the laser flash photolysis method. The Fe(II) complexes of the M95A and M95L mutants are 5-coordinated. Thus, CO must have easy access to the heme, similar to O 2 . The M95H mutation possibly affects the electrostatic character of the CO access channel, rather than the steric character. Again, the k on and K d values of the M95I mutant obtained by the Gilles-Gonzalez group (20) are distinct from those obtained for the Met-95 mutants characterized in this study (20). For myoglobin, the overall k off for CO is determined almost exclusively by the thermal rate of disruption of the Fe-CO bond (38). For Ec DOS proteins, isolation of the PAS domain and Met-95 mutants modified the CO association rate more than the dissociation rate, suggesting that the thermal rate remains unaltered.
Comparison between the O 2 and CO Association Kinetics and Equilibrium Constants-The Gilles-Gonzalez group (14,20,21) reported that Ec DOS is a direct oxygen sensor, although the various kinetic and equilibrium parameters measured in their experiments for the O 2 -bound isolated PAS domain were not significantly different from those of the CO-bound form. However, in the present study, we observed significant differences in the rate and equilibrium constants determined for O 2 and CO binding to the full-length wild-type enzyme (Tables I and  III). In particular, the K d value for O 2 binding to the full-length wild-type enzyme was much higher than that for CO. The considerable difference between the affinity of O 2 and CO (K d (O 2 )/K d (CO) 100 for full-length Ec DOS) was similar to that observed in 25 for myoglobin (39), and 10 for BjFixL (26). The reasons for the discrepancy between our data and results of the Gilles-Gonzalez group (14,20) are currently unclear. However, apparently there are differences in some experimental conditions between the two laboratories. In the present study: (a) all proteins used were highly purified with more than 95% homogeneity (1, 2); (b) all experiments were buffered with 50 mM Tris-HCl (pH 8.0); (c) we purified the wild-type and Met-95 mutant proteins in their Fe(III) low-spin form as determined by optical absorption, resonance Raman (1, 2), and magnetic circular dichroism spectra (52); (d) we strictly ascertained whether spectral changes accompanying the ligand associations and dissociations had clear isosbestic points as shown in Figs. 1-4. a and d were not shown in Refs. 14 and 20. b is in contrast with the reported conditions of the Gilles-Gonzalez group, which used either 0.1 M sodium phosphate buffer (pH 7.0) or 0.1 M Tris-HCl (pH 8.0) in Refs. 14 and 20, respectively. About c, this is in contrast to the reported conditions of the Gilles-Gonzalez group (20), who purified their M95I mutant in the high-spin Fe(III) form. Determination of dissociation constants of gaseous ligands with high affinity (K d ϭ 10 M or less) using "free concentration" (not total concentration added) of ligand in the solution can often lead to incorrect values because of various experimental errors. We thus calculated a more precise K d value from the k on and k off values. Taken together, these differences in experimental conditions may explain the discrepancies between our results and those of the Gilles-Gonzalez group. Under our experimental conditions, significant differences were observed in the kinetics of CO/O 2 binding between the full-length protein and the isolated PAS domain of wild-type Ec DOS.
It should be emphasized that rates of ligand recombination to the 6-coordinate species observed by rapid mixing (stoppedflow) are distinguished from those of recombination with the 5-coordinated form by flash photolysis. The former method is used for very slow ligand binding compared with the rate of endogenous ligand dissociation, which is applicable to the 6-coordinated wild type Ec DOS. On the other hand, the flash photolysis method was applied to the Met-95 mutants, because ligand association rates are fast and occur in the 5-coordinated species (or 6-coordinated M95H mutant) following rapid dissociation of endogenous ligand (30,50,51).
The autoxidation rate of the Fe(II)-O 2 complex of the fulllength wild-type enzyme is faster than that of the isolated PAS domain, suggesting that Ec DOS is not an oxygen sensor enzyme. True biochemical and physicochemical characteristics are more relevant to the in vivo situation, because these values have been measured in our experiments using the full-length enzyme rather than the isolated PAS domain.
FT IR-The stretching frequency of CO bound to Fe(II) heme serves as an empirical voltmeter that may be used to measure the polarity of the distal pocket and predict the extent of electrostatic stabilization of bound O 2 (35,37). The C-O stretching band of the wild-type PAS domain was located at 1970 cm Ϫ1 . This value is higher than that of myoglobin (1945 cm Ϫ1 ) and hemoglobin (1956 cm Ϫ1 ), but similar to the CooA (1969 cm Ϫ1 ) and P. caudatum Hb (1974 cm Ϫ1 ) values. The same stretching frequencies in the resonance Raman spectra of the wild-type, M95A, and M95H mutants are located between 1965 and 1973 cm Ϫ1 (2). CO may interact with a distal His in myoglobin and hemoglobin, but does not bind similarly to distal amino acids in CooA (48) and P. caudatum Hb (49). Therefore, CO is unlikely to interact directly with distal amino acids in Ec DOS. If Met-95 does not interact with CO in the isolated Ec DOS PAS domain, mutation of this residue would not influence the C-O stretching frequency. The small shift in C-O stretching frequency observed in the mutants suggests that this is indeed the case. The FT IR findings are consistent with kinetic results in that the mutants displayed similar CO dissociation rates as the wild-type isolated PAS domain. The k off value reflects the interaction between the coordinated ligand and the distal amino acid. The Met-95 mutations possibly affect the CO access channel rather than the heme distal structure. Only the CO stretching frequency of the M95A mutant was distinct from those of the wild-type enzyme and the other two Met-95 mutants.
A band at 1925 cm Ϫ1 was consistently observed in the wildtype and Met-95 mutant spectra. It is hypothesized that the heme distal site is heterogeneous in character, because CO may interact with an amino acid located at this distal site, such as Arg-97 (20,21,47).
Mechanism of Signal Transduction-It is difficult to infer the signal transduction mechanism of Ec DOS based on the O 2 and CO binding kinetic and equilibrium data obtained in this study. Movement of distal amino acids accompanied by exogenous ligand binding is critical for signal transduction of FixL (5,15,47). Liebl et al. (39,44) suggest that the isolated PAS domain has a very tight heme pocket for exogenous ligands, in view of the ultra-fast (nanosecond to picosecond order) O 2 and CO binding data. The Gilles-Gonzalez group (14, 20, 21) extensively discuss the signal transduction mechanism of Ec DOS in association with Met-95, based on the O 2 and CO binding kinetics and equilibrium data obtained for the wild-type enzyme and M95I mutant of the isolated PAS domain of Ec DOS. Interestingly, the Met-95 mutants of the full-length Ec DOS display PDE activities comparable with those of the wild-type enzyme (52). Structural studies (42,45,46) on signal transducer enzymes with the PAS family domain suggest that the conformational flexibility of this domain is used to communicate ligand binding/activation to downstream transducer proteins as a possible common mechanism. In fact, the amino acid sequence reveals that the Ec DOS PAS domain has a surface salt bridge, which is identified as a signaling pathway that plays a key regulatory role in LOV-mediated signal transduction (46). The role of the isolated heme-bound PAS domain in catalysis of Ec DOS is also interesting to note (53).
Summary-In summary, our study disclosed a number of novel findings. 1) Rate constants for O 2 and CO association (k on ) to the full-length Ec DOS wild-type enzyme were extremely low.
2) The isolated heme-bound PAS domain displayed a markedly increased k on value. 3) Mutation of Met-95, an axial ligand of the Fe(II) heme, to Ala, Leu, and His further increased the k on value of O 2 and CO. 4) The differences between the O 2 and CO binding rates and equilibrium constants were significant. 5) The autoxidation rate of the Ec DOS wildtype enzyme was dramatically decreased in the isolated hemebound PAS domain. 6) Mutation of Met-95 to Ala or Leu further decreased the autoxidation rates in both the full-length wildtype enzyme and the isolated PAS domain. 7) The C-O stretching frequency located at a higher wavelength region at 1970 cm Ϫ1 was not significantly affected by mutation of Met-95, suggesting that CO does not interact ionically with a distal amino acid residue(s). These findings illustrate the characteristic features of the heme binding site of this enzyme. Significant differences between the full-length enzyme and isolated PAS domain were observed with regard to the O 2 and CO rate and equilibrium constants, suggesting close interactions of the N-terminal PAS domain and the C-terminal catalytic domain. Our results support the theory that by sensing the redox state or exogenous ligand binding, a conformational change in the PAS domain of Ec DOS directly regulates the conformation of the catalytic domain, as proposed for other PAS sensor enzymes (7,42). Ligand (CO and NO) binding that triggers signal transduction is common to Ec DOS and FixL, because CO or NO binding terminates catalysis with Ec DOS (1). O 2 adaptation is critically important in many organisms and cell types (43). In anaerobic bacteria such as E. coli, a moderate rise or drop in O 2 concentration may modulate numerous important physiological functions. However, much of the O 2 sensing may be a response to the consequential effects of changes in O 2 tension, such as alterations in redox potential (43). Biologically important redox-related cofactors may alter the O 2 concentration as well as the redox state of cells. Further studies are required to explore the O 2 sensing system of Ec DOS.