The C-helix in CooA rolls upon CO binding to ferrous heme.

CooA is a homodimeric transcriptional activator from Rhodospirillum rubrum containing one heme in each subunit. CO binding to the heme in its sensor domain activates CooA, facilitating the binding to DNA by its DNA-binding domain. The C-helix links the two domains and shapes an interface between the subunits. To probe the nature of CO activation, residues at positions 112-121 on the C-helix were replaced by Asn or Gln and their effects were evaluated by resonance Raman spectroscopy and by the measurements of CO binding affinity. The nu(Fe-CO) stretching Raman line in CO-bound wild-type CooA was up-shifted by 6 cm(-1) in the L116Q, G117N, and L120Q mutants, indicating unequivocally that these residues are close to the bound CO. Residues Leu116 and Leu120 from each subunit form contacts with the corresponding residues in the opposite subunit, enabling hydrophobic interactions in the inactive ferrous form. Thus, in the CO-bound activated form, both C-helices appear to roll to direct these residues toward the heme, forming a hydrophobic pocket for the bound CO. The CO affinity is approximately one order of magnitude higher in the L112Q, I115Q, L116Q, G117N, L120Q, and T121N mutants but reduced in A114N mutant. The variation indicates that these residues are close to the heme in the ferrous and/or CO-bound forms and are responsible for CooA activation. A roll-and-slide mechanism is proposed for CO activation of CooA.

fore, we propose a roll-and-slide model for CO activation of CooA.

MATERIALS AND METHODS
Preparation of CooA Mutants-We have reported a high expression system for CooA, which employs a synthetic gene optimized for expression in Escherichia coli (8). The QuikChange system (Stratagene) was used to introduce mutations into the CooA coding sequence, and the DNA sequence of the plasmid products were confirmed with a Li-Cor Model 4200S2 DNA sequencer. Cell growth and protein purification were performed as described previously (6) with slight modifications. Soluble fractions of the cell extract were applied to a Q-Sepharose FF column (Amersham Biosciences), and the adsorbed proteins were eluted with linear gradient of NaCl in 50 mM Tris-HCl (pH 8.0). The CooA fractions were subsequently purified by Sephacryl S-100 column chromatography (Amersham Biosciences) and concentrated (Amicon Ultra-15, Millipore). Proteins were purified to homogeneity as judged by SDS-PAGE criteria. CooA concentrations were evaluated from the heme content, which was measured by a pyridine hemochrome assay, and hence were based on a monomer unit.
Resonance Raman Spectroscopy-The spectra were recorded using a double monochromator (Jasco R-800) with a slit width of 6 cm Ϫ1 following excitation by a krypton ion laser (406.7-nm line, Coherent I-302). A photomultiplier detector was used (Hamamatsu Photonics, R595), and the frequencies were calibrated with indene. The frequencies reported are accurate within 1 cm Ϫ1 for sharp and discrete Raman lines. A spinning Raman cell was used throughout the measurements to minimize local heating and sample damage. The samples contained 100 M protein in 0.1 M Tris-HCl (pH 8.0). Ferrous proteins were prepared by the addition of sodium dithionite after purging extensively with nitrogen gas. The carbonmonoxy forms were prepared by the addition of sodium dithionite under 1 atm CO. The spectra of the CO forms were obtained with a defocused laser beam.
CO Titration-CO titration of ferrous CooA was performed in 0.1 M Tris-HCl (pH 8.0) at 25°C as described previously (8). A cuvette was filled with a CooA solution, which was purged extensively with nitrogen gas stoppered with a rubber septum, and the protein was reduced by an excess of sodium dithionite solution, which was injected by a syringe. Full reduction of CooA was confirmed by absorption spectroscopy (Beckman DU640). Aliquots of CO saturated in 0.1 M Tris-HCl (pH 8.0) and dithionite were added to solutions containing ϳ2 M protein, and the absorption spectra were recorded. The CO concentrations were calibrated using pig myoglobin. The absorbance changes at the Q-band maxima were traced and analyzed after normalization.

RESULTS
Ferric and Ferrous CooAs-In our mutation strategy, those residues having long and hydrophobic side chains (Leu and Ile) were replaced by Gln, which has long but polar side chain. On the other hand, small residues (Ala, Gly, and Thr) were replaced by the short and polar Asn, which is analogous to Gln. Arg 118 , the sole charged residue in our target set, was replaced by the hydrophobic Leu. The effect of these mutations was evaluated by resonance Raman spectroscopy, which is a powerful tool for revealing the coordination states of heme iron (22,23). In the ferric state, the spectral profiles of these mutants are essentially the same as that of WT CooA (Fig. 2). 2 , 38 , and 3 lines, which are sensitive to the iron coordination state, are observed at 1583, 1549, and 1501 cm Ϫ1 , respectively, indicating that the Cys 75 and Pro 2 ligands in the WT CooA are unperturbed by the mutations. However, in the A114N, I115Q, and A119N mutants, the 38 line appears stronger than the 2 line and it is also slightly up-shifted. In these mutants, a relatively strong line is observed at 268 cm Ϫ1 , which is assignable to the ␦(C ␤ C 1 ) bending vibration of peripheral carbon atom of the heme (24).
In the ferrous CooA mutants, the resonance Raman spectra are again essentially the same as those of the WT protein (Fig.  3). The 2 , 38 , and 3 lines are observed at 1581, 1555, and 1493 cm Ϫ1 , respectively, corresponding to six-coordination with axial ligands contributed by the His 77 and Pro 2 side chains ( Fig. 1). However, in the L116Q, G117N, and L120Q mutants, the 2 and 38 lines are slightly up-shifted and a weak line appears around 1470 cm Ϫ1 ( 3 ). A similar profile was detected in the P2H and H77A mutants of CooA in which one of the Pro 2 -His 77 axial ligands is replaced (8). Thus, some of the C-helix residues affect the axial coordination of the ferric and ferrous hemes in CooA but the effects are slight.
CO Probe-In Fig. 4, resonance Raman spectra of ferrous CO-bound CooA are shown. The (Fe-CO) stretching frequency is sensitive to the polarity of the residues surrounding the bound CO, and it therefore represents an excellent probe of the distal environment (25)(26)(27)(28)(29). The (Fe-CO) stretching line is observed at 490 cm Ϫ1 in WT CooA as reported previously (8,19,30), indicating that the bound CO is in a hydrophobic distal environment. This line is clearly affected by mutations of some of the C-helix residues, and especially noted are the large up-shifts of 6 cm Ϫ1 in the L116Q, G117N, and L120Q mutants. A small but measurable up-shift of 3 cm Ϫ1 in the L112Q and A119N mutants is also noted. In these mutants, hydrophobic residues are replaced by polar Gln or Asn residues. The effect of hydrophobic to polar substitutions on the (Fe-CO) stretching frequency has been reported in myoglobin (31). In ferrous CO-bound myoglobin, the bound CO is close to the distal His 64 and the (Fe-CO) stretching line is observed at ϳ510 cm Ϫ1 (32,33). The line is observed at 490 cm Ϫ1 in H64L and at 492 cm Ϫ1 in H64I mutants that have apolar replacements, but it is restored to 508 cm Ϫ1 in H64Q (31). Thus, it is clear that Leu 116 , Gly 117 , and Leu 120 in WT CooA are close to the bound CO, contributing to a hydrophobic distal heme pocket. Leu 112 and Ala 119 may also locate close to the CO but in a rather remote fashion. The frequency of the (Fe-CO) stretch and the shift values relative to that in WT protein are summarized in Table II.
Along with the frequency shift, the band shape of the (Fe-CO) stretch is slightly affected by the mutations of some of the C-helix residues (Fig. 4). The (Fe-CO) lines are relatively broad in the I113Q, A114N, L116Q, and L120Q mutants, suggesting the fluctuation in the Fe-CO geometry. A minor band seems to exist at ϳ520 cm Ϫ1 in the I115Q mutant, and the presence of multiple Fe-CO conformers is suggested. Similar multiple (Fe-CO) stretching modes were detected in some of the His 77 axial ligand mutants of CooA (8), and hence, the Fe-His bond may partly be affected in the I115Q mutant. The main (Fe-CO) stretching frequency at 489 cm Ϫ1 , however, is almost the same as that in WT CooA; hence, Ile 115 is suggested to be remote from the heme in the CO-bound state.
Because some of the C-helix residues affect the bound CO, the effect of these residues on the CO binding affinity was evaluated. The CO dissociation constants (K d ) for the ferrous-CO bound CooA mutants are given in Table II. The CO affinity of WT CooA (K d ϭ 11 M) is similar to that for FixL (9.0 M) (34) and Dos (10 M) (35), which are heme-based gas sensor proteins. Although the CO binding kinetics is reported to be multi-step (36,37), the equilibrium CO binding is a one-step process (8). Thus, binding was analyzed in terms of a simple equilibrium between the ferrous and CO-bound forms of the protein. The K d values generally decreased in the C-helix mutants, indicating that polar side chains in the mutants facilitate CO binding to the ferrous heme. The logarithm of the ratio of the K d values for the mutant and WT CooA (i.e. ⌬log K d ) is calculated and also summarized in Table II. The CO affinity is less affected in I113Q and R118L, indicating that Ile 113 and Arg 118 are remote from the heme. It should be noted here that the K d value greatly increased in the A114N mutant. Thus, Ala 114 is suggested to play a key role in CO activation of CooA. Because of this low affinity, resonance Raman measurement of CO-bound A114N was very difficult (Fig. 4) and the laser power had to be reduced to 0.3 milliwatts. Even under this condition, the ferrous heme is not fully coordinating CO because saturated CO concentration is 1 mM. This is the main reason for the relatively weak (Fe-CO) stretching line in this mutant. Fig. 5, the positions of C-helix residues are schematically drawn in rectangle and in helical wheel projection models. Because the C-helix shapes a coiledcoil structure in the inactive ferrous form (4), a 3.5 residue/turn scheme was adopted. As summarized in Table I, the C␣ atoms of Ile 113 , Leu 116 , Leu 120 , Cys 123 , and Ile 127 residues on the respective subunits are located close to each other (Ͻ7.0 Å) and these residues are marked with small pink circles (Fig. 5, top). It is clear that these residues align on one face of the C-helix and shape a dimer interface. The C␣ atoms of Leu 112 and Leu 116 (marked with cyan circles) are close to the ferrous heme within the same subunit, whereas those of Ile 113 , Ala 114 , Gly 117 , Arg 118 , and Thr 121 (marked with green circles) are close to the heme within the opposite subunit. Pro 2 , one of the heme axial ligands, is provided by the opposite subunit and is shown in green. In the ferric state, the A114N, I115Q, and A119N mutants affected the relative intensity of the 38 and ␦(C ␤ C 1 ) bending modes (Fig. 2) and, hence, these residues are probably close to the heme periphery in the WT CooA. In the crystal structure of ferrous CooA (4), the side chain of Ile 115 is close to the heme 1-methyl and 2-vinyl groups and Ala 119 is close to the 1-methyl group, although Ala 114 is remote from the heme periphery (Fig. 1). The proximal ligand in the ferric state is the side chain of Cys 75 , which is replaced by that of His 77 in the ferrous state (6,7). The redox-linked ligand exchange in CooA may reposition the heme, providing a close contact of heme periphery with Ala 114 side chain.

Inactive Ferrous Form-In
In the ferrous state, the mutation of residues Leu 116 , Gly 117 , and Leu 120 slightly affected the axial coordination of the heme (Fig. 3). Because His 77 ligand is opposite to the heme plane when viewed from the C-helix, relatively weak coordination by  (21), and this is similar to that observed in our G117N mutant. In these Leu 116 , Gly 117 , and Leu 120 mutants, the K d values for CO dissociation increased by ϳ10-fold (Table II). Although many factors may affect CO affinity, this increase may partly be attributed to the weaker Pro 2 coordination in these mutants. Pro 2 coordination greatly reduces the CO affinity, and this coordination is the way in which CooA regulates its CO-sensing level (8). Leu 116 and Leu 120 locate in the dimer interface in the inactive ferrous state, and the side chain of Leu 120 is too far from the Pro 2 ligand for direct interaction (Fig. 5, top). Thus, the C-helices must roll relative to each other at least in the L120Q mutant to weaken the Fe-Pro 2 bond in the ferrous state. The replacement of these aliphatic residues by the polar Gln should weaken the hydrophobic interaction between the helices. This may promote the rolling of the helices, mimicking the active CO-bound form as described below and facilitating CO binding with high affinity. Active Ferrous-CO Form-As we revealed previously (8), the Pro 2 ligand is replaced by the incoming CO molecule and this replacement triggers CooA activation (11)(12)(13)(14). The residues affecting the (Fe-CO) stretch (Fig. 4) by Ͼ3 cm Ϫ1 are marked with red circles (Fig. 5). Among these residues, the effects of Leu 116 , Gly 117 , and Leu 120 mutation were dominant again; hence, these three residues appear to be located close to the CO molecule. Because Leu 116 and Leu 120 locate at the dimer interface in the ferrous state, the C-helices must roll to direct these residues to the CO molecules. To maximize the interaction between these residues and CO, the hemes must slide along the C-helices. Because Pro 2 is provided from the opposite subunit, CO replacement should heave the Pro 2 anchor in CooA. This should increase the freedom for relative movement of C-helices and hemes. Regardless, the movement of both helices as well as both hemes must be symmetrical because CooA is homodimeric and the recognition sequence of CooA is palindromic (15-17).
The residues that affected the CO affinity more than one order of magnitude are marked by small purple circles in Fig. 5 (bottom). The CO affinity increased predominantly in the L112Q, I115Q, L116Q, Gl117N, L120Q, and T121N mutants. The Leu 116 , Gly 117 , and Leu 120 side chains are close to one another on one face of the C-helix, thus lining the distal heme pocket and greatly affecting the CO affinity. The remaining Leu 112 , Ile 115 , and Thr 121 residues are directed outward from the dimer interface in the ferrous form (Fig. 5, top) and may not interact directly with the bound CO. These residues surround the Leu 116 , Gly 117 , and Leu 120 triad, possibly forming an entrance allowing CO to access the heme.
Among the CooA mutants examined, A114N is the sole mu-  tant, which greatly reduced the CO affinity, although Gly 117 and Thr 121 , which lie on the same C-helix face, enhanced the CO affinity when replaced by Asn. It is interesting to note that Ala 114 is sandwiched by Gly 111 and Gly 117 , which have small side chains. Such small residues may facilitate the helical roll in the CO activation. When the small Ala 114 is substituted by the polar and bulkier Asn, it may inhibit C-helix rolling. This inhibition may stabilize the ferrous inactive form so that CO affinity is reduced.
Roll-and-Slide Model-As discussed above, the C-helices must roll, and here, we propose a roll-and-slide model for the CO activation of CooA (Fig. 5, bottom). In this section, we validate the model based on the experimental evidence. Since A114N showed low CO affinity, we propose that Ala 114 could be the fulcrum of the helix rolling to shape the CO-bound form. This means that the heme must slide to the C-terminal side on the helix, and this view is consistent with the close proximity of the Leu 116 , Gly 117 , and Leu 120 triad to the CO-bound heme. Leu 116 is reported to be crucial for CooA activity, and the L116K mutant is active in both ferrous and CO-bound states (20). Leu 116 is close to the heme in our model (Fig. 5, bottom), enabling the coordination of Lys side chain in L116K.
It is proposed (19) that the C-helix relatively moves upon the CO binding and that the heme also moves to the N-terminal side. In our systematic mutation study, however, we favor the movement of the heme toward the C-terminal portion of the C-helices for the reasons outlined below. In the I113Q mutant, the K d value and (Fe-CO) stretching frequency are less affected than in the other mutants examined (Table II), although Ile 113 is one helical turn up from Leu 116 and Gly 117 and is close to the heme (marked with green circle in Fig. 5, top). In con-trast, Met 124 is reported to affect iron coordination in the activated form (38). This residue is one helical turn down from Leu 120 , supporting the notion of translation of the heme toward the C terminus of the helix.
Three major factors contribute to cofactor retention in heme proteins (39): (i) covalent bonding between the heme iron and axial ligand residues; (ii) apolar interactions between the heme and surrounding hydrophobic residues; and (iii) polar interac-  Table I) are marked with small pink circles. The residues having d(C␣-Fe) and d(C␣-FeЈ) of Ͻ10 Å (boldface in Table I) are marked with cyan and green circles, respectively. Pro 2 is shown in green to indicate that the ligand is provided by the opposite subunit. Bottom, active ferrous-CO form. The residues that showed ⌬(Fe-CO) stretching frequencies greater than 3 cm Ϫ1 (boldface in Table II) by the mutation are marked with red circles. The residues that showed K d values of one order of magnitude greater or smaller than that for WT CooA (boldface in Table II) by the mutation are marked with small purple circles.
tions between the heme propionates and surrounding hydrophilic residues (40). In ferrous CooA, the bonds to the Pro 2 -His 77 ligands may be enough to anchor the heme, although the heme propionate is not surrounded by amino acid side chains. However, in the CO-bound state, Pro 2 is replaced by CO, requiring a compensating mechanism for the heme retention. In our model, the helical roll may allow the positively charged Arg 118 to interact favorably with the negatively charged heme propionate in the opposite subunit. Although Arg 118 interacts with Asp 72 in the ferrous form (4), this residue may partly contribute electrostatically to the stabilization of the slipped heme.
In summary, the C-helix in CooA rolls upon CO binding to the heme as revealed by resonance Raman spectra and CO dissociation constants. We propose a roll-and-slide model for the CO activation mechanism of CooA. This model satisfactorily accounts for our observations as well as those reported earlier. Although further studies on the interactions between the C-helices and the DNA-binding domains are required, our roll-and-slide model may be a good starting point for fully understanding the CO activation mechanism.