Redox-controlled Ligand Exchange of the Heme in the CO-sensing Transcriptional Activator CooA*

The transcriptional activator CooA fromRhodospirillum rubrum contains a b-type heme that acts as a CO sensor in vivo. CooA is the first example of a transcriptional regulator containing a heme as a prosthetic group and of a hemeprotein in which CO plays a physiological role. In this study, we constructed an in vivo reporter system to measure the transcriptional activator activity of CooA and prepared some CooA mutants in which a mutation was introduced at Cys, His, Met, Lys, or Tyr. Only the mutations of Cys75 and His77affected the electronic absorption spectra of the heme in CooA. The electronic absorption spectra, EPR spectra, and the transcriptional activator activity of the wild-type and mutant CooA proteins indicate that 1) the thiolate derived from Cys75 is the axial ligand in the ferric heme, but it is not coordinated to the CO-bound ferrous heme; 2) Cys75 is protonated or displaced in the ferrous heme; and 3) His77 is the proximal ligand in the CO-bound ferrous heme and probably also in the ferrous heme, but it is not coordinated to the ferric heme. NMR spectra reveal that the conformational change around the heme, which will trigger the activation of CooA by CO, takes place upon the binding of CO to the heme.

The purple, non-sulfur, photosynthetic bacterium Rhodospirillum rubrum can grow on CO as a sole energy source under anaerobic conditions in the presence of CO (1,2). The expression (which is regulated at the transcriptional level) of the proteins coded in the cooFSCTJ and cooMKLXUH operons is induced under these conditions (3)(4)(5). The genes of key enzymes that gain energy for growth on CO such as CO dehydrogenase and hydrogenase are coded in the coo operons (3)(4)(5). The cooA gene product has been reported to be the transcriptional activator for regulation of the expression of the coo operons and to be a member of the CRP 1 /FNR family of transcriptional regulators on the basis of amino acid sequence homology (3)(4)(5).
CRP is a homodimer of a 209-amino acid monomer that is composed of two domains (6 -11). The small carboxyl-terminal domain contains a helix-turn-helix DNA-binding motif, and the large amino-terminal domain is responsible for subunit-subunit contact for dimerization and binds cAMP as the effector (9 -13). The two domains are connected by a hinge region (res-idues 135-139) (9 -11). CooA contains a helix-turn-helix motif as the DNA-binding motif in its C-terminal region (3), which indicates that CooA is a DNA-binding transcriptional regulator protein. We have reported that the amino-terminal region from Met 1 to Met 131 is the heme-binding domain in CooA, which corresponds to the cAMP-binding domain in CRP (14). The homology between CRP and CooA suggests that they share a common mechanism of transcriptional regulation; however, the effectors of these regulatory proteins are completely different. The most interesting feature of the transcriptional regulation with CooA is that CO is required for the expression of the coo operons (2)(3)(4)(5), indicating that CO acts as the effector of CooA.
It has been reported that recombinant CooA can be expressed in Escherichia coli and that it contains a b-type heme as a prosthetic group (15,16). The ferrous heme in CooA can bind CO as an axial ligand (14 -16). DNase I footprint analysis with CooA has shown that the protection of the target sequence on DNA is observed only in the presence of CO under anaerobic conditions (4,5,16), indicating that binding of CO to the heme in CooA causes the specific binding of CooA to the target DNA. These results show that the heme in CooA acts as a CO sensor in vivo and regulates the activity of CooA by the binding of CO. Although CO is widely used as a probe to study the biochemical and biophysical properties of hemeproteins, it does not have any physiological role in these cases. CooA is the first example of a hemeprotein in which CO plays a physiological role.
The function of the heme in CooA, which is the sensor of the effector, is a new one. Two hemeproteins, FixL (17)(18)(19)(20)(21)(22)(23)(24) and soluble guanylate cyclase (25)(26)(27)(28)(29)(30)(31)(32)(33)(34), have been reported as members of the hemeprotein family in which the heme acts as the sensor of the effector. The hemes in FixL and soluble guanylate cyclase are O 2 and NO sensors, respectively, and are responsible for the regulation of the enzymatic activity with the effector (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34). CooA is a new member of the "heme-based sensor proteins." The coordination structure of the heme is an important factor in the regulation of the function of hemeproteins. In this study, we tried to determine the coordination structure of the heme in CooA by measurements of the electronic absorption spectra, EPR spectra, and the transcriptional activator activity of wild-type and some mutant CooA proteins prepared by site-directed mutagenesis. To measure the activity of CooA, we constructed an in vivo reporter system containing a cooF-lacZ operon fusion as the reporter gene. We propose that the heme in CooA exists in a six-coordinate form in ferric, ferrous, and CO-bound ferrous forms and that the exchange of the axial ligand takes place upon the change in the redox state of the heme. By measuring NMR spectra of ferrous and CO-bound ferrous CooA, we found that the conformational change around the heme, which will trigger the activation of CooA by CO, takes place upon the binding of CO to the heme.

EXPERIMENTAL PROCEDURES
Vectors and Strains-R. rubrum (IFO 3986) was obtained from the Institute for Fermentation (Osaka, Japan). Plasmid pRS551, RS74 phage, and E. coli P90C (35) were kindly provided by Professor R. W. Simons (UCLA). Plasmid pKK223-3 and the TA cloning kit were obtained from Amersham Pharmacia Biotech and Invitrogen, respectively. E. coli strains were grown on LB medium. Ampicillin (50 mg/ liter) and/or kanamycin (30 mg/liter) was added to the medium when necessary.
Construction of the Expression System-The expression vector pKK3CO5 was constructed as reported previously (15). To improve the yield of recombinant CooA, a new expression vector harboring two copies of cooA (pKK3CO6) was also constructed as follows. The DNA fragment containing cooA was excised from pCRCO (15) by digestion with EcoRI and was ligated with EcoRI-treated pKK223-3. The clone harboring pKK3CO6 was selected from the cells transformed by the above ligation mixture on LB agar plates containing ampicillin and 1 mM isopropyl-␤-D-thiogalactopyranoside. The desired clone showed a deep orange-red color compared with the clone harboring pKK3CO5.
Expression and Purification of CooA-Recombinant CooA was expressed in E. coli JM109 as reported previously (15). Purification was carried out according a previous method (14,15) with some modification. A Q-Sepharose column was used in the first step instead of a butyl-Sepharose column. The fractions containing CooA eluted from the first column were combined and applied to a chelating Sepharose column charged by zinc ion. The chelating Sepharose column was preequilibrated with 20 mM Na 2 HPO 4 buffer (pH 7.2) containing 1 M NaCl. Adsorbed proteins were eluted by a 0 -0.06 M imidazole linear gradient. The fractions containing CooA were combined and dialyzed against the appropriate buffer. Further purification on a gel filtration column (Sephacryl S-100) was carried out if necessary. The CooA solution after dialysis was concentrated by ultrafiltration with a YM-10 membrane (Amicon, Inc.) for spectroscopic measurements. CooA was dissolved in 50 mM Tris-HCl buffer (pH 8.0) unless otherwise noted.
All of the mutants constructed in this work were expressed and purified as the wild type except for C75A CooA. As the amount of the expressed C75A CooA was less than one-fifth of that of the wild type and this mutant was labile compared with the wild type, C75A CooA could not be purified to homogeneity. Therefore, a crude sample of C75A CooA was used for the spectroscopic measurements.
Determination of the Heme Content-The concentration of CooA was determined by the pyridine ferrohemochrome method (36). A value of 34 mM Ϫ1 cm Ϫ1 (36) at the absorption maximum of the ␣ band for the pyridine ferrohemochrome derived from the protoheme was used to calculate the concentration of CooA.
Site-directed Mutagenesis-Site-directed mutagenesis was carried out using the QuickChange site-directed mutagenesis kit (Stratagene) or the Chameleon double-stranded, site-directed mutagenesis kit (Stratagene). pKK3CO5 was used as the template for mutagenesis.
Formation of Reduced and CO-bound CooA Proteins-Reduced CooA was prepared by adding dithionite under argon atmosphere into CooA solution degassed by a vacuum pump. CO-bound CooA was prepared by introducing CO by a gas-tight syringe into reduced CooA solution.
Construction of the Reporter System-The construction of the reporter system for the measurement of CooA activity in vivo was carried out according to a previously reported procedure (14). A transcriptional fusion of the cooF promoter region to a reporter gene (lacZ gene in this work) was carried out as described below. pCR-COP was constructed by cloning a 337-base pair fragment bearing the cooF promoter region, which consisted of the region of the cooF transcript from positions Ϫ250 to ϩ87 (2, 4) (position ϩ1 represents the transcriptional start point of the cooF transcript (4)), into the pCRII vector with the TA cloning kit. This cooF promoter region, which is denoted as the COP region hereafter, was synthesized by the polymerase chain reaction with the chromosomal DNA of R. rubrum (IFO 3986) as the template. The two synthetic oligonucleotides (5Ј-GATATCCCGCTGATCGTCAAC-3Ј and 5Ј-CGAGACGAAACAAAGACTTCGC-3Ј) purchased from Cruachem, which were designed according to the DNA sequence of the coo region (2), were used as the primers. The DNA fragment containing the COP region was fused to lacZ in pRS551, which is a multicopy vector used in analysis of lacZ gene fusion (35). pRS-COP was constructed by inserting the EcoRI fragment containing the COP region, which was cut from pCR-COP, into the EcoRI site of pRS551. The direction and nucleotide sequence of the COP region in pRS-COP were confirmed by DNA sequence analysis with an Applied Biosystems Model 373A sequencer. This plasmid-borne fusion was transferred to the RS74 phage vector by homologous recombination in E. coli P90C according to a previously reported procedure (35), and the phage lysate containing the recombinant phage was prepared. Lysogens were obtained by infecting E. coli JM109 with this resulting phage lysate and selected on LB agar plates containing kanamycin. A clone bearing a single copy of the recombinant -prophage was selected and used in the following study. The resulting strain was named E. coli COP. Prophage copy number was determined by polymerase chain reaction according to the literature (37).
␤-Galactosidase Assay-␤-Galactosidase levels were determined by hydrolysis of o-nitrophenyl-␤-D-galactopyranoside according to the method of Miller (38). The cell-free extract prepared from the reporter strain was used for the assay. E. coli COP bearing an appropriate CooA expression vector was grown in a 50-ml cultivation flask containing 10 ml of LB medium with 50 g/ml ampicillin, 30 g/ml kanamycin, and 1 mM isopropyl-␤-D-thiogalactopyranoside at 37°C for 8 h. The cultivation was carried out on a rotary shaker at 180 rpm. When the cells were grown with CO, 10 ml of a head-space gas in a flask sealed by a rubber septum was replaced by the same volume of CO with a gas-tight syringe before starting the cultivation. The collected cells after the cultivation were washed and resuspended in 4 and 1 ml of Z-buffer (38), respectively. The resuspended cells were broken by sonication and then centrifuged to prepare the cell-free extract for the measurement of ␤-galactosidase activity. The assay was carried out at 28°C. The protein concentration of the cell-free extract was determined by the absorbance at 280 nm using bovine serum albumin as a standard. The specific activity of ␤-galactosidase is expressed as nanomoles of o-nitrophenyl-␤-D-galactopyranoside hydrolyzed per min/mg of protein.
EPR and NMR Measurements-X-band EPR spectra were measured on a Jeol RE1X or RE3X apparatus. CooA was dissolved in 50 mM Tris-HCl buffer (pH 8.0) for the measurement of EPR spectra. Azidebound metmyoglobin was used as a standard for quantitation of the low-spin heme. Myoglobin from horse heart was obtained from Sigma. The 1 H NMR spectra were measured on a Varian Unity 750 (750 MHz) spectrometer at 25°C. Chemical shifts were referenced to 2,2-dimethyl-2-silapentane 5-sulfonate. For the measurement of NMR spectra, CooA was dissolved in 50 mM KH 2 PO 4 /NaOH buffer (pH 7.6) containing 15% D 2 O.

Electronic Absorption and EPR Spectra of Wild-type CooA-
The electronic absorption spectra of wild-type CooA are shown in Fig. 1. While we reported previously that ferric (oxidized) CooA purified using a butyl-Sepharose column showed the Soret absorption peak at 418.5 nm (15), CooA purified by the modified procedure described in this work revealed the Soret absorption peak at 423.5 nm in the oxidized state, which is consistent with the observations of Shelver et al. (16). Using a butyl-Sepharose column may cause the dissociation of an axial ligand of the heme in CooA to form a five-coordinate heme in the oxidized state to some extent. Therefore, the above-modified method was used to purify CooA in this study. The electronic absorption spectra of CooA in the ferrous (reduced) and CO-bound states were identical in both samples purified using a butyl-Sepharose and a Q-Sepharose column.
Ferric CooA showed the Soret, ␣, ␤, and ␦ bands at 423.5, 570.0, 538.5, and 362.5 nm, respectively. In addition to these bands, a weak charge transfer band was observed at 760 nm, which was tentatively assigned to be a charge transfer band from sulfur to Fe 3ϩ . Ferrous CooA showed the Soret, ␣, and ␤ bands at 424.5, 557.5, and 528.5 nm, respectively, which was typical of six-coordinate low-spin hemeproteins. CO-bound ferrous CooA showed the Soret, ␣, and ␤ bands at 422.0, 568.5, and 540.5 nm, respectively. These spectra were consistent with the observations of Shelver et al. (16) on CooA purified from R. rubrum.
The EPR spectrum of ferric CooA showed a rhombic component with g values of 2.46, 2.26, and 1.90 as shown in Fig. 2a. These values are typical of low-spin hemeproteins, indicating that the heme in ferric CooA is in the six-coordinate low-spin form. The g values of CooA are almost the same as those of substrate-free ferric cytochrome P450 cam (g ϭ 2.45, 2.26, and 1.91) (39) and ferric cytochrome P420 cam (g ϭ 2.45, 2.27, and 1.91) (40), in which a thiolate derived from a cysteine is the axial ligand of the heme.
The quantitation of the spin concentration revealed that the intensity of the signal of ferric CooA was 0.8 spins/CooA monomer. The content of the heme was also determined to be 0.8 mol/CooA monomer by the pyridine ferrohemochrome method. These results show that CooA contains 1 mol of protoheme/mol of CooA monomer, which was also consistent with the observations of Shelver et al. (16). We also confirmed by gel filtration experiments that CooA existed in a dimeric state in the ferric, ferrous, and CO-bound ferrous states as reported by Shelver et al. (16) (data not shown).
Electronic Absorption and EPR Spectra of CooA Mutants-To determine the axial ligands of the heme in CooA, we constructed some mutants by site-directed mutagenesis and measured their electronic absorption spectra and their transcriptional activator activity. Cysteine, histidine, methionine, lysine, and tyrosine were chosen as candidates for the axial ligand. CooA contains five Cys residues at positions 35, 75, 80, 105, and 123; five His residues at positions 28, 77, 133, 146, and 200; and five Met residues except for the first one at positions 73, 76, 108, 124, and 131. We constructed all of the Cys-to-Ala (C35A, C75A, C80A, C105A, and C123A), His-to-Ala (H28A, H77A, H133A, H146A, and H200A), and Met-to-Leu (M73L, M76L, M108L, M124L, and M131L) CooA mutants at these positions. A His-to-Tyr mutant (H77Y CooA) was also constructed. For Lys-to-Ala and Tyr-to-Phe mutants, three (K26A, K30A, and K101A) and two (Y55F and Y67F) CooA mutants were constructed. All of the Lys and Tyr residues located in the heme-binding domain (131 residues from Met 1 to Met 131 (14)) in CooA were chosen as a target for mutagenesis.
The absorption maxima of the ferric, ferrous, and CO-bound ferrous forms of wild-type and mutant CooA proteins are summarized in Table I. Among the above 21 mutants, only C75A, H77A, and H77Y CooA showed different electronic absorption spectra from wild-type CooA. The other 18 mutants showed the same spectra as the wild type in the ferric, ferrous, and CObound ferrous forms.
The electronic absorption spectrum of C75A CooA in the ferric form is shown in Fig. 3. The Soret band was observed at 411.0 nm, which was blue-shifted by 12 nm compared with the wild type, and the clear ␣ and ␤ bands in the Q-band region were not observed in ferric C75A CooA. The EPR spectra of C75A CooA are shown in Fig. 4. Ferric C75A CooA showed EPR signals in the g ϭ 6 region at 4 K due to the high-spin heme, whereas it did not show any signals in the g ϭ 2 region due to the low-spin heme at 4 and 77 K. These features are typical of five-coordinate, high-spin hemeproteins, suggesting that Cys 75 is an axial ligand of the heme in ferric CooA. Ferrous C75A CooA showed the Soret, ␣, and ␤ bands at 423.0, 557.0, and 527.0 nm, respectively (Fig. 3). This spectrum is similar to that of the wild type, suggesting that the heme in ferrous C75A CooA is in the six-coordinate form, as is the wild type. CObound ferrous C75A CooA showed the Soret, ␣, and ␤ bands at 420.5, 569.0, and 532.0 nm, respectively.
The electronic absorption spectra of H77A and H77Y CooA are shown in Figs. 5 and 6, respectively. Ferric H77A and H77Y CooA showed the Soret, ␣, ␤, and ␦ bands at 422.5, 570.0, 538.5, and 362.0 nm and at 423.5, 570.0, 538.5, and 363.5 nm, respectively. The charge transfer band at 760 nm in the ferric form was also observed in both mutants. The EPR spectra of H77A and H77Y CooA are shown in Fig. 2 (b and c,   the electronic absorption and EPR spectra are almost the same as those of the wild type, indicating that the mutation at position 77 does not affect the coordination structure of the ferric heme, i.e. His 77 is not thought to be coordinated to the ferric heme in CooA. Although the electronic absorption spectra of H77A and H77Y CooA in the ferric state were identical, the ferrous forms of these mutants showed different properties. Ferrous H77A CooA showed the Soret, ␣, and ␤ bands at 422.5, 557.5, and 528.0 nm, respectively. This spectrum is similar to that of wild-type CooA, although the Soret band was slightly blueshifted by 2 nm compared with the wild type, and the molar extinction coefficients of the absorption maxima were different from those of wild-type ferrous CooA (see the legends of Figs. 1 and 5). On the other hand, the electronic absorption spectrum of ferrous H77Y CooA showed the Soret band and a single peak in the Q-band region at 424.0 and 558.5 nm, respectively, which resembled the spectrum of deoxymyoglobin (41) and was typical of five-coordinate, high-spin ferrous hemeproteins. Shelver et al. (16) have reported that H77Y CooA cannot be stably reduced by dithionite (although the data are not shown),   The microwave power, modulation frequency, and modulation amplitude were 1 milliwatts, 100 kHz, and 1 millitesla (mT), respectively. The signals at g ϭ 2.01 and 4.27 seem to be due to some organic radical in the sample or the contaminant proteins and an adventitious Fe 3ϩ , respectively. which is inconsistent with our result. A reason for the difference may be that they used crude extracts prepared from E. coli. However, H77Y CooA in crude extracts can be reduced by dithionite in our preparation. The reasons for this difference are not obvious at present.
CO-bound ferrous H77A and H77Y CooA showed the Soret, ␣, and ␤ bands at 419.5, 563.0, and 538.5 nm and at 419.5, 568.0, and 538.5 nm, respectively. Wild-type CooA in the CObound ferrous form showed the Soret, ␣, and ␤ bands at 422.0, 568.5, and 540.5 nm, respectively. The Soret absorption max-ima of the His 77 mutants were blue-shifted by 2 nm compared with the wild type, suggesting that there is some alteration of the surrounding structure of the heme in the CO-bound form when the mutation is introduced at His 77 .
Conformational Change around the Heme upon the Binding of CO-The CO-bound ferrous heme in CooA was formed by the reaction of ferrous CooA with CO as described above. CO should replace one of the axial ligands of the heme in ferrous CooA to form CO-bound ferrous CooA because the heme in ferrous CooA is in the six-coordinate form with two axial ligands. The binding of CO to the heme will remove an axial ligand from the heme and will cause some conformational change around the heme, which will trigger the activation of CooA by CO as described under "Discussion." The conformational change around the heme upon the binding of CO was pursued by 1 H NMR in this work. The 1 H NMR spectra of ferrous and CO-bound ferrous CooA are shown in Fig. 7. Ferrous CooA showed several signals that were shifted by the ring current of the heme in the Ϫ2ϳϪ6 ppm region as shown in Fig. 7a, which were due to protons of the amino acid(s) being the axial ligand(s) of the heme and/or those located nearby above the heme plane. These signals observed in the Ϫ2ϳϪ6 ppm region in ferrous CooA disappeared when CO bound to the heme as shown in Fig. 7b. These results indicate that the amino acid residue(s) giving the signals that were shifted by the ring current of the heme in ferrous CooA move away from the vicinity of the heme to be free from the ring current effect upon the binding of CO. The movement of the amino acid residue(s) as described above may cause the conformational change around the heme.
Wild-type and Mutant CooA Transcriptional Activator Activities-To measure the transcriptional activator activity of CooA in vivo, we constructed the reporter strain E. coli COP containing a single copy of the recombinant -prophage that contains a cooF-lacZ operon fusion. When E. coli COP transformed by the expression vector of wild-type CooA (pKK3CO5/E. coli COP) was grown in the presence of CO, the specific activity of ␤-galactosidase increased with cultivation time and then reached a constant value (Fig. 8, filled circles). On the other hand, ␤-galactosidase activity was scarcely observed when pKK3CO5/E. coli COP was grown in the absence of CO as shown in Fig. 8 (empty circles). These results show that recombinant CooA expressed in E. coli cells can act as the transcriptional activator only in the presence of CO, which is the same situation as that in R. rubrum, and that the transcriptional activator activity of CooA can be evaluated using the in vivo reporter system constructed in this study.
Immediately after the disruption of the cells, the crude extract prepared from E. coli cells expressing wild-type CooA showed the ␣ and ␤ bands at 557.5 and 528.5 nm, respectively, which were the same as those of ferrous CooA (data not shown). This result shows that the expressed CooA in E. coli exists in the ferrous form even when E. coli cells are grown aerobically. As ferrous CooA reacts readily with CO to form CO-bound CooA, CooA in E. coli cells grown in the presence of CO should exist in the CO-bound ferrous form. Therefore, the results shown in Fig. 8 indicate that CO-bound ferrous CooA is active as the transcriptional activator, but ferrous CooA is not.
The activity of the CooA mutants prepared in this study was measured, and the results are summarized in Table II. These mutants can be classified into three groups according to the dependence of the activity on CO as follows: group 1, the dominant-negative mutant that is inactive regardless of the presence and absence of CO; group 2, the dominant-positive mutant that is active regardless of the presence and absence of CO; and group 3, the mutant that is active in the presence of CO and inactive in the absence of CO, as the wild type.
H77A CooA was dominant-negative. C105A, H77Y, and M131L CooA were dominant-positive. The activities of C105A and M131L CooA were up-regulated by ϳ15and 20-fold, respectively, compared with that of the wild type in the presence of CO. Although H77Y CooA was dominant-positive, the activity of this mutant was down-regulated to be about one-fifth of that of the wild type. Other mutants, except for H77A, H77Y, C105A, and M131L CooA, were active only in the presence of CO, as the wild type. Among those mutants in group 3, H133A, C123A, and Y67F CooA showed less than one-half of the activity compared with the wild type, although the dependence of the activity on CO was the same as that of the wild type. The activity of H28A, C35A, M73L, and M76L CooA in the presence of CO was ϳ60% of that of the wild type. Other mutants in group 3 showed activity similar to that of the wild type.
In the course of site-directed mutagenesis experiments to determine the axial ligands of the heme in CooA, we found that some mutants show different activity compared with that of the wild type as described above, although their coordination struc-ture of the heme was identical to that of the wild type. Among these mutants, M131L CooA was previously obtained as a dominant-positive mutant by random mutagenesis (14), and the possible mechanism by which this mutant is active even in the absence of CO is described below. The detailed properties of other mutants showing activity different from that of the wild type remain to be elucidated. DISCUSSION Electronic absorption and EPR spectroscopies revealed that the heme in ferric CooA was in the six-coordinate, low-spin form. In the resonance Raman spectrum of ferric CooA, 2 and 3 bands have been observed at 1580 and 1501 cm Ϫ1 , respectively. 2 These values are consistent with the model that the ferric heme in CooA is in the six-coordinate, low-spin state. Ferric CooA shows a clear ␦ band and a weak charge transfer band at 362.5 and 760 nm, respectively, which are similar to absorptions in P450 cytochromes containing a thiolate as an axial ligand of the heme (42). The EPR spectrum of ferric CooA also resembles that of thiolate-ligated hemeproteins such as cytochromes P450 and P420, i.e. the g values observed in ferric CooA (g ϭ 2.46, 2.26, and 1.90) are almost the same as those of substrate-free ferric cytochrome P450 cam (g ϭ 2.45, 2.26, and 1.91) (39) and ferric cytochrome P420 cam (g ϭ 2.45, 2.27, and 1.91) (40), in which a thiolate derived from a cysteine is the axial ligand of the heme. These results show that a cysteine is an axial ligand of the ferric heme in CooA.
Ferric C75A CooA shows the typical electronic absorption and EPR spectra of five-coordinate, high-spin hemeproteins. C75A CooA showed complex EPR signals in the g ϭ 6 region as shown in Fig. 4, which may be the superposition of the signals due to the heme with a different conformation. As the mutation is introduced at the amino acid acting as the axial ligand in C75A CooA, the conformation of the heme will not be fixed. The spectroscopic properties of wild-type and C75A CooA indicate  that Cys 75 is an axial ligand of the ferric heme in CooA. The electronic absorption spectra of the ferric CooA mutants at His, Met, Cys, Lys, or Tyr, except for Cys 75 , are identical to that of the wild type, indicating that these residues are not the sixth axial ligand of the ferric heme in CooA. The EPR spectra of the His 77 mutants (H77A and H77Y CooA), which are almost the same as that of the wild type, support that His 77 is not coordinated to the ferric heme in CooA. A water molecule is a possible candidate for the sixth ligand of the ferric heme in CooA because ferric CooA shows almost the same EPR spectrum as the substrate-free cytochrome P450 cam , which contains a thiolate and a water as the axial ligands (39,40). However, we cannot determine whether the sixth axial ligand in the ferric heme is a water or some amino acid residue. Ferrous CooA shows the typical electronic absorption spectrum of the six-coordinate, low-spin hemeproteins. In the resonance Raman spectrum of ferrous CooA, 2 and 3 bands have been observed at 1579 and 1491 cm Ϫ1 , respectively (43). These values are similar to those of the ferrous cytochrome b 5 ( 2 ϭ 1583 and 3 ϭ 1493 cm Ϫ1 (44)). These results indicate that the heme in ferrous CooA is in the six-coordinate, low-spin form. Ferrous H77Y CooA showed an electronic absorption spectrum similar to that of deoxymyoglobin that contains a five-coordinate heme. The resonance Raman spectrum of ferrous H77Y CooA shows the split 3 bands at 1470 and 1492 cm Ϫ1 . 2 The frequency of 3 ϭ 1470 and 1492 cm Ϫ1 is similar to that of deoxymyoglobin ( 3 ϭ 1473 cm Ϫ1 (45)) and wild-type CooA ( 3 ϭ 1491 cm Ϫ1 (43)), respectively. Resonance Raman spectroscopy therefore reveals that both the five-coordinate and sixcoordinate hemes exist in ferrous H77Y CooA. 2 On the other hand, ferrous H77A CooA shows an electronic absorption spectrum similar to that of wild-type CooA. The electronic absorption spectra of ferrous H77A and H77Y CooA are typical of six-coordinate and five-coordinate hemeproteins, respectively. These results suggest that His 77 may be the axial ligand of the ferrous heme or may be located near the heme in ferrous CooA.
Cys 75 is coordinated to the ferric heme in CooA as described above. Is it also coordinated to the ferrous heme? Ferrous low-spin, thiolate-ligated heme complexes and hemeprotein such as H450 (a soluble iron protoporphyrin IX-containing protein of unknown function (46)) typically exhibit Soret peaks at ϳ445 nm (46 -49), compared with the 424.5 nm Soret band exhibited by ferrous CooA. It has been reported that the blue shift in the wavelength of the Soret peak of ferrous H450 takes place from near 450 nm to ϳ425 nm upon lowering the pH to 6, which requires that the thiolate (cysteinate) ligand is either protonated or displaced upon lowering the pH (46,48,49). In the case of ferrous CooA, therefore, Cys 75 will be protonated or replaced by another amino acid residue to show the Soret peak at 424.5 nm.
The resonance Raman spectrum of CO-bound CooA has revealed that a histidine is the fifth ligand of the heme in the CO-bound form (43). The stretching modes of Fe-CO and CϭO, (Fe-CO) and (CϭO), have been observed at 487 and 1969 cm Ϫ1 , respectively, in CO-bound CooA (43). Among the His mutants in CooA, only the mutation at His 77 affected the properties of the electronic absorption spectra of CooA. These results indicate that His 77 is the proximal ligand in CO-bound CooA. The properties of H77A and C75A CooA transcriptional activator activity support that His 77 is the proximal ligand of CO-bound CooA, but Cys 75 is not. The binding of CO to the heme activates CooA as the transcriptional activator by the conformational change around the heme and finally in the whole molecule as discussed below. Therefore, if the mutation is introduced in the proximal ligand in CO-bound CooA, it will cause the change in the transcriptional activator activity. C75A CooA shows activity similar to that of the wild type, whereas H77A CooA is inactive even in the presence of CO, which supports the above conclusion.
On the basis of the above discussion, the model of the coordination structure of the heme in CooA we propose is shown in Fig. 9. CooA shows the unique property that the axial ligand is exchanged upon the change in the redox state of the heme iron. Cys 75 is coordinated to the ferric heme, but not to the CObound heme. On the contrary, His 77 is coordinated to the CObound heme, but not to the ferric heme. The similar redoxcontrolled ligand exchange has been reported for cytochrome cd 1 from Thiosphaera pantotropha (51,52). The exchange of the axial ligands is thought to be responsible for the adjustment of the redox potential to regulate the internal electron transfer and for the release of the reaction product (NO) from the d 1 heme (51,52). In the case of CooA, the exchange of the axial ligand upon the change in the redox state of the heme may be concerned with the regulation of the redox potential of the heme.
Recombinant CooA expressed in E. coli shows transcriptional activator activity when the reporter strain is grown with CO, when CooA exists in CO-bound form. It has been reported that CooA exhibits sequence-specific DNA binding and binds DNA only in the presence of CO under anoxic, reducing conditions (4,5,16), which shows that only CO-bound CooA can bind the target DNA to be active as the transcriptional activator. These results indicate that the binding of CO to the heme in CooA is a very important step in activating CooA as the transcriptional activator. CO should replace one of the axial ligands to bind the heme because the ferrous heme in CooA, which can bind CO, is in the six-coordinate form. NMR spectra of ferrous and CO-bound CooA reveal that the release of the axial ligand from the heme induced by the binding of CO causes some conformational change around the heme. This signal of the conformational change around the heme, which is induced by the binding of CO, will finally change the conformation of the FIG. 9. Proposed model of the coordination structure of the heme in CooA. L represents an unidentified ligand. The coordination structure of the ferrous heme is tentative. The possible three models are shown. Cys 75 will be protonated if it is coordinated to the ferrous heme (see "Discussion"). whole molecule to be adapted for the specific binding to the target DNA.
The properties of M131L CooA transcriptional activator activity suggest that the alteration of the relative orientation of the two domains, the heme-binding domain and the DNAbinding domain, will be involved in the activation process of CooA, which is triggered by the binding of CO. As Met 131 is located at the end of the heme-binding domain and adjacent to the hinge region that connects the heme-binding domain and the DNA-binding domain (14), the replacement of Met by Leu at position 131 will cause the change in the relative orientation of the two domains by the change in the steric hindrance of the residue at position 131. M131L CooA is active even in the absence of CO probably because the conformation of M131L CooA in the absence of CO will be the same as that of wild-type CooA in the presence of CO. A similar effect of the mutations around the hinge region has been reported in the case of CRP (53)(54)(55). The activity of M131L CooA is up-regulated compared with that of the wild type, but the reasons for this up-regulation are not obvious at present.
CooA is a new member of the heme-based sensor protein family. So far, two hemeproteins, FixL (17)(18)(19)(20)(21)(22)(23)(24) and soluble guanylate cyclase (25)(26)(27)(28)(29)(30)(31)(32)(33)(34), have been reported to be hemeproteins in which the heme acts as a sensor of an effector molecule. CooA, FixL and soluble guanylate cyclase contain the heme sensing CO, O 2 , and NO, respectively. Although the function of the heme that senses the gaseous effector molecules such as CO, O 2 , and NO by binding them is the same in CooA, FixL and soluble guanylate cyclase, the signal transduction mechanism is different, especially at the initial stage. Whereas FixL and soluble guanylate cyclase contain the five-coordinate heme and the effector binds to the vacant sixth position of the heme (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34), CO replaces one of the axial ligands to bind the heme in CooA. The conformational change triggered by the replacement of the axial ligand by CO is responsible for the activation of CooA, just as that triggered by the cleavage of the Fe-His bond upon binding NO trans to the proximal histidine in the case of soluble guanylate cyclase. The binding of the exogenous ligand is the trigger of the conformational change in the case of CooA and soluble guanylate cyclase. A similar conformational change induced by the binding of the ligand has been reported for some cЈ-type cytochromes, although the physiological relevance of these observation remains unclear because the real functions of cЈ-type cytochromes are unknown (56). However, this suggests that the ligand-induced conformational change will be the common mechanism regulating the function of hemeproteins (57).