Activation mechanism of the CO sensor CooA. Mutational and resonance Raman spectroscopic studies.

CooA is a CO-dependent heme protein transcription factor of the bacterium Rhodospirillum rubrum. CO binding to its heme causes CooA to bind DNA and activate expression of genes for CO metabolism. To understand the nature of CO activation, several CooA mutational variants have been studied by resonance Raman spectroscopy, in vivo activity measurements, and DNA binding assays. Analysis of the Fe-C and C-O stretching Raman spectroscopy bands permits the conclusion that when CO displaces the Pro2 heme ligand, the protein forms a hydrophobic pocket in which the C-helix residues Gly117, Leu116, and Ile113 are close to the bound CO. The displaced Pro2 terminus is expelled from this pocket, unless the pH is raised above the pKa, in which case the terminus remains in H-bond contact. The pKa for this transition is 8.6, two units below that of aqueous proline, reflecting the hydrophobic nature of the pocket. The proximal Fe-His bond in Fe[II]CooA is as strong as it is in myoglobin but is weakened by CO binding, an effect attributable to loss of an H-bond from the proximal His77 ligand to the adjacent Asn42 side chain. A structural model is proposed for the position of the CO-bound heme in the active form of CooA, which has implications for the mechanism of CO activation.

CooA is a member of the emerging class of heme sensor proteins that modulate biological activity in response to fluctuations in the concentration of the gaseous molecules CO, NO, or O 2 (1,2). Binding of these diatomic ligands to the heme induces the biological response. The mechanisms whereby the protein responds to the binding event are of great current interest.
CooA is isolated from the bacterium Rhodospirillum rubrum, which grows anaerobically on CO as sole energy source (3). CooA regulates the expression of genes, termed coo for CO oxidation, whose products are associated with CO metabolism. The protein is a homodimer (222 amino acids per monomer) containing two b-type hemes that reversibly bind CO. The CO binding enables the protein to bind a specific DNA sequence and induce transcription of the coo genes. CooA contains DNAbinding and signal domains analogous to those found in the catabolite gene activator protein CAP 1 (or CRP) (1, 2) (Fig. 1). The heme is bound within the signal domain, where it is ligated by a histidine (His 77 ) side chain (although this ligand is replaced by the Cys 75 side chain when the Fe[II] is oxidized to Fe[III] (4)). The sixth Fe [II] ligand is unprecedented; it is the N-terminal proline residue (Pro 2 ) from the opposite chain, which is displaced by CO binding (5).
The structure of CO-bound CooA is unknown, but on the basis of the CAP structure (6) it is reasonable to assume that the DNA-binding domains are brought into proper position for DNA binding by movement of the C-helix, which runs the length of the protein and passes close to the heme (1,2,7). It has been shown that a repositioning of the C helices at the dimer interface is a major communication pathway between the hemes and the DNA-binding domains (8), but the basis for this repositioning upon CO binding remains unclear. To elucidate the nature of the CO-bound form of CooA and the mechanism of its CO activation, we have prepared a series of CooA variants by site-directed mutagenesis and examined in vivo activity, specific DNA binding, and the Fe-CO and C-O stretching resonance Raman (RR) bands. The residues chosen for mutagenesis are in the region of the heme groups, as defined by the Fe[II]CooA structure. Fig. 2 gives a close up view of this region.
The data permit us to identify the approximate position of the heme-CO in CooA and propose a specific model for how that positioning affects CooA activation and leads to the C-helix repositioning. An important aspect of the active form involves the breakage of a critical interaction that stabilizes the inactive structure: an H-bond from the His 77 proximal ligand to the side chain of Asn 42 .
These structural inferences also provide a basis for understanding the cooperative binding and complex kinetics of CooA. 2 Finally, we propose a structural model for the heme movement that is consistent with the data and will serve as a working hypothesis for future work.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and in Vivo Assays-The construction of strains overexpressing WT CooA and CooA variants in an Escherichia coli background having a ␤-galactosidase reporter system in the chromosome was described previously (8), and in vivo activities were quanti-tated using the standard protocol (10). All the site-directed and regionrandomized CooA mutations were constructed in a pEXT20-based expression plasmid, which provides tight control of CooA expression (11).
CooA Purification-The purification of WT CooA and most CooA variants was performed with our standard method as described previously (12). Some CooA variants such as C75S, L116K, and M124R CooA were purified with modified methods, which are also described elsewhere (12)(13)(14). The purity of WT CooA and CooA variants was estimated to be Ͼ95% based on SDS-PAGE (in the case of L116K CooA, ϳ90%). The heme content of CooA preparations was estimated using the extinction coefficient of Fe[III] WT CooA (3) or by a modified reduced pyridine-hemochromogen method (3), and protein concentration was measured using the BCA assay (Pierce).
In Vitro DNA Binding Assays-In vitro DNA binding assays of WT CooA and CooA variants were performed using the method described elsewhere (11). As a fluorescence probe, a 26-bp target DNA containing PcooF was labeled with Texas Red on one end of the duplex and used at the concentration of 6.4 nM. Salmon sperm DNA was used as the nonspecific DNA competitor. Dissociation constants (K d ) were calculated by fitting of the binding data to a nonlinear equation with correction of the fluorescence quenching as described elsewhere (15).
Sample Preparation-Purified CooA was diluted in buffer (25 mM Mops/0.1 M NaCl, pH 7.4) to a heme concentration of ϳ20 M. Other pH values were investigated using 0.1 M Tris, pH 7-9, 0.1 M glycine, pH 9.5-10, and ϳ0.1 M NaOH, pH [11][12]. CooA samples were purged with N 2 for 20 min and then with CO for ϳ3 min, followed by reduction of the Fe[III] to the Fe[II] form with sodium dithionite (final concentration ϳ63 mM). Reduction and CO binding were monitored by changes in the absorption spectra.
RR Spectroscopy-RR spectra were obtained with excitation wavelengths of 406.7 and 568.1 nm from a Kr ϩ laser (2080-RS; Spectra Physics) and 441.7 nm from a He-Cd (Liconix) laser in a 270°backscattering sample geometry. Photodissociation of the bound CO was minimized by using low laser power (ϳ5 milliwatts at the sample) and by focusing it with a cylindrical lens onto a spinning sample. The scattered light was collected and focused onto a single spectrograph (Chromex) equipped with a notch filter (Kaiser Optical) and a CCD detector (Roper Scientific) operating at 77 K. Spectra were calibrated with dimethyl formamide and dimethyl sulfoxide-d 6 .

RESULTS
The goal of this study has been to understand the nature of the heme environment of CO-bound CooA and use this information to better understand the mechanism of CooA activation. RR and activity data are gathered and compared in Table  I. We note at the outset that spectroscopic probes, which report on local structural features, are not expected to track the degree of activity, which measures the biochemical competency of the global protein conformation. It is the interplay between the two sets of data that inform our thinking about activation mechanism.
The results of this study are organized in the following manner. 1) There are actually two different forms of CooA-CO, with different C-O stretching frequencies, as revealed by reanalysis of RR spectra. This observation has an effect on the interpretation of RR data in this manuscript and resolves a discrepancy in RR data reported previously. 2) Backbonding analysis of the RR data reveals that the Fe-His proximal bond is weakened in the major form of CooA-CO. Mutational data point to breaking of the His 77 -Asn 42 H-bond as the cause of this weakening and is consistent with a significant role for this process in CooA activation. 3) The bound CO is certainly in close proximity to Ile 113 , Leu 116 , and Gly 117 , which has implications for the movement of the heme and the protein upon CO binding. This proximity was revealed by the RR analysis of mutational variants substituted at these positions that show striking perturbations and is generally consistent with functional characterization of these variants. 4) RR analysis of pH effects of WT CooA, and of variants containing deletions at the N terminus that includes Pro 2 , suggest that CO displacement of Pro 2 leads to its expulsion from the hydrophobic CO binding pocket, because of its positive charge when protonated. The previously reported depression of the Fe[II]CooA 11 heme RR band is found, via the effects of variants, to stem from the strong ligand field of the Pro 2 ligand.
Structurally Distinct Populations of CO-bound CooA-We have previously reported RR spectra of CooA-CO showing a Fe-CO stretching band (FeC) at 487 cm Ϫ1 and an unusually high CO stretching band (CO) at 1982 cm Ϫ1 (16). In contrast, Uchida et al. (17) reported CO ϭ 1969 cm Ϫ1 for their CooA-CO preparation although they reported the same FeC value as ours, 487 cm Ϫ1 . To resolve this discrepancy we reexamined the CO and FeC regions of the RR spectrum at high amplification ( Fig. 3), using 13 CO shifts to identify the bands definitively. FeC was measured at 487 cm Ϫ1 , as before, and we also reproduced the 1982-cm Ϫ1 CO band. However the newly measured CO band has a pronounced shoulder at 1962 cm Ϫ1 . Upon 13 CO substitution, this 1962-cm Ϫ1 band shifted appropriately, indicating a second CO band.
The 487-cm Ϫ1 FeC band is a single band. Deconvolution from the 465-cm Ϫ1 band of excess dithionite (Fig. 3) yields a bandwidth, 14.5 cm Ϫ1 , which is less than that of the CO adduct of myoglobin (data not shown), 18 cm Ϫ1 , ruling out multiple FeC contributions in CooA-CO. Therefore, we conclude that the molecules having CO ϭ 1982 and 1962 cm Ϫ1 both have the same FeC ϭ 487 cm Ϫ1 .
The CO reported by Uchida et al. (17), 1969 cm Ϫ1 , is closer to the lower of the two values we observe, but their CooA-CO preparation also had a fraction with CO ϭ 1979 cm Ϫ1 ; this is the frequency reported in picosecond FTIR experiments (18) for CooA-CO molecules undergoing photolysis. Thus, CooA-CO has two populations of molecules, having distinct CO values (1962 and 1982 cm Ϫ1 in our hands), but the same FeC value, 487 cm Ϫ1 . The relative size of the populations appears to be preparation-dependent. In our preparation the 1962-cm Ϫ1 fraction is small; the species is labeled CooA-COЈ. We suggest that the species represented by 1982 cm Ϫ1 is the active form and that CooA-COЈ is a form that has not rearranged to the active, DNA binding structure. There may be an equilibrium between active and inactive forms, or else a fraction of the molecules are stuck in the inactive configuration because of an energy barrier of some type.
Conformational Change of CooA upon CO Binding Weakens the Fe-His 77 Bond by Breaking the His 77 -Asn 42 H-bond-The anomalous CO of the major form of CooA-CO, 1982 cm Ϫ1 , is produced by a weakened Fe-His 77 bond. This can be seen in a plot of FeC against CO (Fig. 4). Because of backbonding, FeC and CO are negatively correlated in heme proteins (19,20), as illustrated for a set of Mb variants in Fig. 4 (solid line). Changes in the polarity of the binding pocket augment or diminish backbonding and shift the points along the line (19,20). In addition, the correlation is displaced upward when the Fe-proximal ligand bond is weakened, because of diminished bonding competition between the proximal ligand and the CO (19,21). The upper dashed line in Fig. 4 represents data for Fe[II] porphyrin CO adducts without any trans ligand (21).
When the CooA-CO values are plotted on the graph, the minority form, CooA-COЈ, is seen to fall near Mb variants with hydrophobic pockets, in which the distal histidine is replaced by hydrophobic residues (cluster of points marked ϪH64 in Fig.  4). However, the distinctly higher CO of the dominant form, CooA-CO, places it above the Mb line by 7 cm Ϫ1 , implying a weakened bond from the Fe to the proximal ligand, His 77 in CooA (see Fig. 2).  This displacement from the Mb backbonding line was missed in our earlier correlation (16), because a more varied group of heme proteins was used, having greater scatter of the points. The high CO was instead attributed to negative polarity in the CO binding pocket (16). However, examination of the pocket region, defined by the mutagenesis results to be described below, fails to reveal any candidate residue with a negative charge or even a lone pair of electrons that could provide a negative dipole. In principle, peptide dipoles (22) could interact with the bound CO, but this is unlikely, because the residues that form the binding pocket are part of a helix.
The inference of a weakened Fe-His 77 bond in CooA-CO is strongly supported by the picosecond RR study by Uchida et al. (23), which identified the Fe-His stretching band at 211 cm Ϫ1 in the immediate photoproduct of CooA-CO. This band, a direct measure of the Fe-His bond strength, is enhanced in five coordinate Fe[II] hemes, and is found at 220 cm Ϫ1 in deoxyMb (24). The distinctly lower value found by Uchida et al. (23) shows that the Fe-His 77 bond is weaker than in Mb for the immediate photoproduct, consistent with a weaker bond in the CO adduct itself. The minority form, CooA-COЈ, which lies on the Mb backbonding line (Fig. 4), does not have a weakened Fe-His 77 bond and is expected to have an Mb-like photoproduct; however, its Fe-His would escape detection because of the low population.
How strong is the Fe-His 77 bond in Fe[II]CooA itself? Because Fe[II]CooA is six-coordinate, the Fe-His vibration is not enhanced (25), and its frequency cannot be measured. However, we took advantage of a significant five-coordinate fraction in the G117I variant of Fe[II]CooA (7), in which the bulky Ile side chain causes partial dissociation of the Pro 2 ligand (see below). The five-coordinate heme Soret absorption band is redshifted from the six-coordinate band, allowing us to use laser wavelength tuning (442 nm) to enhance the five-coordinate RR signals (Fig. 5). Selection of the five-coordinate species is confirmed by the position of the strong 4 band at 1355 cm Ϫ1 , the standard five-coordinate frequency (25). The Fe-His band is seen at 220 cm Ϫ1 , the same frequency as in deoxyMb. Thus the Fe-His bond is as strong in G117I variant of Fe[II]CooA as in Mb. We infer that the Fe-His bond is similarly strong in WT Fe[II]CooA (there is no reason that steric displacement of Pro 2 should make the proximal Fe-His connection stronger) and that it weakens significantly in the major form of CooA-CO, as well as its immediate photoproduct. At later times, the photoproduct Fe-His position should increase to 220 cm Ϫ1 as the protein relaxes to the Fe[II]CooA structure, but this process cannot be followed in photolysis experiments because of rapid geminate rebinding of CO (18,23).
These results establish that CO binding weakens the Fe-His 77 bond. We therefore asked what could be the mechanism responsible for this property of CooA. It has been known that H-bond donation increases the anionic character of the imidazole ring and strengthens the Fe-His bond (26). For example, the Fe-His frequency is much higher, 240 cm Ϫ1 , in five-coordinate Fe[II] forms of the peroxidases that have a negative Asp side chain in position to accept the proximal His H-bond (24). In Mb the proximal His is H-bonded to a neutral OH group of a Ser residue, as well as to a backbone carbonyl (Fig. 4, inset).
In the Fe[II]CooA structure (Fig. 2), the H-bond acceptor is the amide side chain carbonyl of Asn 42 (Fig. 2) (21). The cluster of Mb points in the middle of the diagram (ϪH64) is from variants in which the distal histidine is replaced with hydrophobic residues. The point for the minor WT CooA-CO (X) lies in the cluster, whereas the major WT CooA-CO and the CO-bound form of L116K CooA variant, as well as the high pH form of WT CooA-CO, lie between the two lines, indicating that the Fe-His bond is weaker in CooA-CO than in MbCO.
With regard to the minor CooA-COЈ form, we note that the low frequency shoulder on the N42A CO band is attenuated relative to the wild-type 1962-cm Ϫ1 shoulder. Curve resolution gives a higher frequency, 1966 cm Ϫ1 for the N42A shoulder, suggesting a weakened H-bond in the N42A CooA-COЈ structure.
Although Asn 42 interacts with His 77 in the Fe[II] form and not the CO-bound form, the nature of the residue is of some importance to activation. N42A, N42D, and N42K CooA all displayed poorer DNA affinity in the presence of CO (and poorer ␤-galactosidase activity) than did WT CooA (Table I), suggesting some additional role for this residue in the active form. The basis for that effect is unknown. Note that the activity assays that measure DNA affinity are somewhat easier to interpret than the in vivo assays reporting ␤-galactosidase activity, because the former only measures DNA binding, whereas the latter measures interaction with RNA polymerase and has other complications (8).
These results indicate that an element of the activation mechanism is the perturbation of the His-Fe bond. The data FIG. 5. RR spectrum of G117I CooA variant in the Fe[II] form, obtained with 442-nm excitation to enhance the five-coordinate high spin fraction, whose Soret band is at ϳ440 nm (7). The 1355-cm Ϫ1 4 band position is diagnostic for five-coordinate high spin heme (23), and a strong Fe-His stretching band is seen at 220 cm Ϫ1 . The spectrum closely resembles that of deoxyMb (21). and arguments below strongly suggest that the His-Fe bond is perturbed by displacement of the heme upon CO binding, which is critical for CooA activation.
Hydrophobic C-helix Residues, but Not Pro 2 , Form the CO Pocket of CooA That Is Necessary for Activation-The distal heme pocket has an important role in ligand discrimination in many heme proteins (27). CooA activation is specific to CO, and an understanding of the CO binding pocket should be valuable in understanding both specificity and the activation in response to CO. We have already shown that hydrophobic residues at positions 113 and 116 are critical for normal CO activation (28) and that Gly 117 is also necessary for this response (7). However, the positioning of these residues with respect to the bound CO was not investigated. We therefore measured the RR spectra of a variety of CooA variants that were either novel in terms of activity and might therefore be altered in their sensing of the bound CO (Table I) Table I, neither replacement of the Pro 2 terminus by tyrosine (P2Y CooA) nor deletion of the two penultimate residues (⌬P3R4 CooA) showed noticeable effect on the FeC or CO band. This implies that Pro 2 in WT CooA does not form the CO pocket and is consistent with the previous observation that Pro 2 is not essential for activation of CooA by CO (11). We note, however, that the DNA binding affinities of these two Pro 2 variants are 5-10-fold poorer than WT CooA, suggesting a minor role for the displaced Pro 2 in the active form of CooA. Properties of the Pro 2 variants are further considered below.
Because Pro 2 is not near the bound CO, then the structure of Fe[II]CooA suggests that the most probable candidates for CO pocket residues are on the C-helix, with Ile 113 , Leu 116 , and Gly 117 being particularly attractive (1,28). The RR analysis of CooA variants altered at these positions establishes significant influences on FeC and CO consistent with a positioning of these residues close to the bound CO (see Fig. 7 and Table I). When these residues are replaced by H-bond donors, histidine and lysine FeC shifts up, and CO shifts down from the position in WT CooA, as expected for a positive polar interaction. 3 The effect is particularly marked for the Gly 117 position. G117H CooA displayed an 11-cm Ϫ1 upshift of FeC and an 18-cm Ϫ1 downshift of CO. These shifts are opposite in sign and comparable in magnitude to those observed when the distal histidine of Mb is replaced by hydrophobic residues (19,20). G117S CooA also showed a sizeable, though more modest effect on the RR spectrum, shifting FeC up by 7 cm Ϫ1 and CO down 12 cm Ϫ1 . These results indicate that the newly introduced His 117 or Ser 117 (albeit to a smaller extent) lie sufficiently close to the CO to exert a positive potential, via H-bond donation. However, the introduced His 117 in G117H CooA is not well ordered and exerts a heterogeneous effect on the CO as evidenced by broad CO and FeC bands (Fig. 7). Consistent with the polarity interpretation, introduction of steric hindrance at 117 position (G117I CooA) had very little effect on FeC and CO (Table I), because the CO environment remains non-polar. The importance of Gly 117 to CooA activation has already been described (7) and is reflected in the dramatically reduced ac-tivity of Gly 117 variants seen in Table I. Because Gly 117 is critical for activity, and its substitution perturbs the RR spectra, it would be tempting to conclude that its role is to contact the bound CO. Although this hypothesis cannot be discounted, it is also possible that Gly 117 is an important determinant of the C-helix structure per se and simply happens to be near the bound CO.
At position 113, the effect of mutation on RR spectra is smaller, as FeC shifts up only 3 cm Ϫ1 and CO down 11 cm Ϫ1 in I113H CooA (see Fig. 7 and Table I). Interestingly, essentially the same effect was seen in I113A and I113D CooA. A possible interpretation is that in these cases water molecules may be brought into contact with the CO, filling the cavity that results from substituting the smaller Ala residue, or attracted to the negatively charged Asp side chain. These results suggest that Ile 113 is close to the bound CO but not as close as Gly 117 . In the case of I113Y CooA, the RR spectrum gives a slight indication of negative polarity, shifting FeC down 2 cm Ϫ1 and CO up 3 cm Ϫ1 . This effect could result from proximity to the electron cloud of the Tyr ring to the CO. Substitutions at position 113 typically have modest effects on activity, as reported previously (28) and shown in Table I. The exception is I113D, and we have already shown that hydrophilic residues at positions 113 and 116 severely perturb activation (28). It is our working hypothesis that the effect is because the hydrophobic pocket is important for eliciting the proper conformational change in the C-helices upon displacement of the Pro 2 ligand by CO.
At position 116, Lys and Phe substitutions were tested. Other possible H-bond donors were not examined, because they 3 An alternative interpretation is that these donors ligate the iron in place of Pro 2 and that CO displaces His 77 instead. However, there is no reason to suppose that distal replacements would labilize the Fe-His 77 bond. When His 77 itself is replaced by Tyr, the FeC/CO values (Table  I) are significantly different, and there is a five-coordinate CO-bound fraction, implying displacement of both ligands (16). have been shown to accumulate heme-containing CooA very poorly, presumably because of disruption of the hydrophobic pocket (28). L116K CooA variant was chosen, because it showed very high CO-independent DNA binding activity (comparable with that of wild-type CooA in the presence of CO) and diminished CO-dependent DNA binding activities (see Ref. 14 and Table I). As shown in Fig. 7, the introduction of Lys at position 116 exerted an almost identical effect on the RR spectrum as did G117H, shifting FeC by 13 cm Ϫ1 up and CO by 18 cm Ϫ1 down (Table I). This result is consistent with the hypothesis that the L116K substitution, like G117H, introduces positive polarity into what is otherwise a non-polar environment around the CO. This CooA variant also showed broad CO and FeC bands. CooA-CO activity is impaired by the L116K substitution but is still considerable, indicating a modest perturbation of the active structure. The high activity of L116K CooA in the absence of CO has already been analyzed, and the result strongly implies that the substituted Lys residue serves as a ligand in both the Fe[III] and Fe[II] forms. We have suggested that this ligation leads to a similar helix repositioning as that seen in WT CooA when bound to CO (14). The absence of significant FeC or CO shifts in L116F is again consistent with maintenance of a hydrophobic environment for the bound CO.
Importantly, in the G117H and L116K CooA variants, the FeC/CO point shifts up and to the left in the backbonding plot (Fig. 4, only L116K is shown, for clarity) but remains displaced from the Mb line by the same amount as WT CooA-CO. This behavior indicates that the Fe-His 77 bond is unaffected by the distal residue replacement, although backbonding is enhanced by the distal polarity. CO binding weakens the Fe-His 77 bond in these variants, just as in WT CooA-CO.
It should be noted that these observations apply only to the major population of CooA-CO molecules as described at the beginning of the "Results." It is impossible to tell how the substitutions affect the minor CooA-COЈ fraction because of its weak signal and the breadth of the perturbed bands. However, a non-polar binding pocket is also indicated for the minor WT CooA-CO by its position in the FeC/CO correlation (Fig. 4), close to those of Mb variants with non-polar pockets.
In summary, the RR data establish that Gly 117 , Leu 116 , and Ile 113 are all in or near the CO binding pocket. However, the data also indicate that substitutions of the Ile 113 and Leu 116 side chains are much more permissive with respect to CO activation of CooA than are those at Gly 117 . A possible interpretation is that the CO directly contacts the Gly 117 and that any steric hindrance destabilizes the DNA binding conformation of the protein. However, it is also possible that Gly 117 induces a critical bend in the C-helix, which is disrupted by substitutions.
We also examined other candidate residues for their effect on RR spectra. These included C-helix variants M124R and W110L and B-helix variants E99L and I95W (the B-helix is the short helix on the "bottom" of CooA as shown in Fig. 1). M124R (13) was chosen because of its significant impact on CO-sensing function of CooA. The others were chosen because of their potential proximity to the heme iron. As shown in Table I, all these variants except M124R had little effect on RR spectra, suggesting that these residues are not near the bound CO and consistent with the pocket being formed primarily by Gly 117 , Leu 116 , and Ile 113 .
The M124R replacement has little effect on the CooA-CO activity (Table I) but produces a double peaked CO band; one component is at the WT CooA position, 1982 cm Ϫ1 , whereas the other is ϳ17 cm Ϫ1 lower. The FeC band is lowered slightly and is broadened (see Table I and Fig. 7). Although Met 124 is fur-ther from the heme iron than are Gly 117 , Leu 116 , and Ile 113 , it is possible that the long Arg side chain reaches the vicinity of the bound CO in a fraction of the CooA-CO molecules, producing the downshifted CO component. Modeling of the introduced arginine in Fe[II]CooA (13) suggested two favored conformations of the side chain, the guanidine cation interacting either with the seven-propionate of the heme or with the backbone carbonyl of Ser 78 . It is possible that in the M124R-CO structure one of the Arg conformers interacts instead with the CO.
The Strongly Basic Pro 2 Ligand Is Expelled from the Hydrophobic CO Pocket in the CO-bound Form of CooA at Physiological pH-Extensive mutational studies have shown that Pro 2 is not critical for the response of CooA to CO (11). Nevertheless, the Pro 2 ligand is a central aspect of cooperativity in CO binding by CooA. 2 Because the global conformational change of CooA upon CO binding is initiated by the displacement of the Pro 2 ligand by CO, it would be informative to trace the fate of the displaced Pro 2 . We have shown above that Pro 2 does not appear to be near the bound CO at normal pH, but when the solution pH was increased above 7, the FeC and CO of WT CooA-CO were affected significantly (Fig. 8). The FeC band at 487 cm Ϫ1 was replaced by one at 497 cm Ϫ1 whereas the CO band at 1982 cm Ϫ1 was replaced by one at 1964 cm Ϫ1 . Although the latter frequency is close to that of CooA-COЈ, the FeC frequency is distinctly higher than in CooA-COЈ. Thus the change at high pH is not simply an increase in the CooA-COЈ fraction, but instead a new species is formed in basic solution.
The high pH FeC and CO values are similar to those displayed by the L116K and G117H variants (see Fig. 7 and Table I). The FeC/CO point remains well above the MbCO backbonding correlation (Fig. 4), indicating little change in the Fe-His 77 bond. We infer that the effect of raising the pH is to bring a H-bond donor into proximity with the bound CO. As in L116K and G117H, the high pH CO and FeC bands are broadened, suggesting heterogeneity in the H-bond donor position.
What might this H-bond donor be? Key evidence comes from the ⌬P3R4 CooA variant, whose FeC and CO bands are unperturbed when the pH is raised, up to 11 (Fig. 8). We conclude that the H-bond donor is the N-terminal Pro 2 , interacting with the bound CO via its secondary amine NH group. When the two penultimate residues, Pro 3 and Arg 4 , are removed, the N terminus is constrained from interacting with the CO.
The Pro 2 interaction in WT CooA requires pH elevation, presumably because the secondary amine becomes protonated when it is displaced from the Fe[II] at pH 7. Because the CO binding pocket is hydrophobic, the protonated N terminus would be expelled from the binding site, because it is positively charged, just as the distal histidine is expelled from the MbCO binding pocket upon protonation at low pH (29,30). At high pH, the displaced Pro 2 , being neutral, can remain in contact with the CO and enhance backbonding, just as the distal histidine does in Mb.
What is the pK a of the displaced Pro 2 ? In aqueous solution the pK a of the Pro amine is 10.5, but it should be lowered in CooA-CO because of the energy cost of expelling Pro 2 from the hydrophobic pocket. In Mb, the pK a of the distal His is 4.3 (29,30), nearly three units lower than the aqueous His pK a . We examined the pH dependence of WT CooA-CO by decomposing the FeC band into 487-and 497-cm Ϫ1 components and plotting their intensities against the solution pH (Fig. 9). The resulting titration curves yield a pK a of 8.6, two units lower than the aqueous pK a of proline. If the assignment of the pK a to Pro 2 is correct, then the energy cost for expelling the distal H-bond donor is about two-thirds as high in CooA as it is in Mb.
We note that the absorption spectrum is not sensitive to the pK a 8.6 process. The Soret band position, which is 422 nm at pH 7, only shifts significantly when the pH is raised to values greater than 10; at pH 12 the band maximum is 419 nm. We found that the RR spectral changes could not be fully reversed for solutions whose pH was raised to 10 or higher, indicating some irreversible change in the protein structure. However the Soret band shift could be reversed, even from pH 12, showing that there is a different physical mechanism for perturbation of the absorption spectrum.
These results indicate that Pro 2 does not lie near the bound CO under physiological conditions, and the role of Pro 2 in the active form of CooA is uncertain. The DNA affinity of the CO-bound forms of P2Y and ⌬P3R4 CooA is 5-to 10-fold poorer than that of WT CooA, implying some small perturbation of the active structure when the N terminus is altered. However the primary roles of Pro 2 appear to be to maintain the inactive form of CooA in the absence of CO (8) and to provide a mechanism for the cooperative binding of CO when it is present. 2 The ligating nitrogen atom of Pro 2 is part of a strongly basic secondary amine and is expected to be a strong donor to the Fe[II]. Evidence for strong donation is seen in the RR spectrum of Fe[II]CooA, which reveals a depressed value, 1532 cm Ϫ1 , for the 11 porphyrin ring stretching vibration (25) (Fig. 10, 568-nm excitation was used to enhance 11 via resonance with the heme Q bands (25)). This mode is known to be sensitive to the strength of the axial ligand field in low spin Fe[II] porphyrin complexes. For example, the band shifts from 1547 to 1533 cm Ϫ1 when methionine is replaced by lysine in Fe[II] cytochrome c (31) and from 1539 to 1527 cm Ϫ1 upon deprotonation of the imidazole complexes of microperoxidase or of bisimidazole Fe[II] protoporphyrin (32,33). The sensitivity to imidazole deprotonation was the basis of our earlier suggestion (16) that His 77 might be deprotonated or strongly H-bonded in Fe[II]CooA. However, His 77 is now revealed by the crystal structure to have a normal H-bond with Asn 42 , and the position of 11 is unaffected when Asn 42 is replaced by the non-H-bonding Ala (Fig. 10).
We conclude that the 11 depression is unrelated to His 77 but instead reflects the strong ligand field of Pro 2 . This interpretation is supported by the observation (Fig. 10) that 11 shifts up, by 4 cm Ϫ1 , in the P2Y and ⌬P3R4 variants. When Pro 2 is replaced by Tyr, the secondary amine is replaced by a primary amine at the terminus, producing a somewhat weaker ligand field. In the ⌬P3R4 variant, the penultimate pair of residues is deleted, shortening the N terminus. The Fe[II] protein remains FIG. 8. pH dependence of the FeC and CO bands for the CO adducts of WT CooA and of the ⌬P3R4 CooA variant. The shift in FeC for WT CooA is also seen in the position of the ϳ1860-cm Ϫ1 band assigned to a combination band with 4 (1371 cm Ϫ1 , the strongest Soret-excited RR band). Another band, at 2049 cm Ϫ1 , is because of a combination of 4 with 7 (677 cm Ϫ1 , another strong fundamental), which is unaffected by pH (and whose intensity indicates the weakness of resonance enhancement for the CO band). The asterisks mark the excess dithionite band.
FIG. 9. pH titration curves of WT CooA-CO. For the analysis, FeC band envelopes of RR spectra were deconvoluted into 487-and 497-cm Ϫ1 components and then the fractional intensity of each band was plotted against the solution pH. The solid curves represent best fits to single proton release and uptake processes having a pK a of 8.6. mainly six-coordinate, but a small fraction of the heme is fivecoordinate, as judged by a shoulder on the 4 porphyrin band at the expected five-coordinate position, 1355 cm Ϫ1 (25). Pro 2 might remain the ligand in ⌬P3R4, but the shortening of the polypeptide chain would strain the Pro 2 -Fe bond, again weakening the field. This interpretation is consistent with the very rapid binding of CO to the ⌬P3R4 variant. 2

DISCUSSION
The objective of this study was to gain insight into the nature of the CO-bound form of CooA and its implications for the conformational change that allows CooA-CO to bind its DNA target. RR signals arising from the heme and its ligands were examined for selected variants, whose activity in vivo and in a DNA binding assay were determined. It is not expected that spectroscopic and activity perturbations should be correlated. On the contrary, it is an important feature of this study that quite different aspects of the CooA molecule are sensed by the spectroscopic and activity probes. Full activity assures that the protein has the correct conformation to bind its DNA target; alterations in many parts of the molecule can destabilize the correct conformation. The RR signals are sensitive to the structure of the heme and its ligands. It is this information that we seek to integrate into a model of how the binding of CO at the heme initiates the conformation change from inactive to active protein.
Two principal findings are central to our model. 1) The Chelix residues Ile 113 , Gly 117 , and Leu 116 become part of the CO binding pocket in CooA-CO. This is revealed by the spectral perturbations observed when their side chains are replaced with H-bond donors. This result implies repositioning of the hemes relative to the C-helices, because all three residues are at substantial distances from the heme iron in the Fe[II]CooA crystal structure (1).
Measured Fe-C ␣ distances are 9.4 (10.2) Å for Leu 116 , 13.0 (7.0) Å for Gly 117 , and 11.2 (9.3) Å for Ile 113 . The first number refers to the residue on the same subunit as the heme, whereas the second number refers to the residue on the opposite subunit. Although not required crystallographically this part of the structure is nearly 2-fold symmetric; the Fe-C ␣ distances from the second heme are within 0.1 Å of those from the first. It is notable that only in the case of Leu 116 does the closer residue reside on the same subunit as the heme. For both Ile 113 and Gly 117 , the residue from the opposite subunit is much closer than is the residue from the same subunit. Thus a simultane-ous rearrangement of both hemes and both C-helices is suggested to bring all three residues close enough to allow H-bond donor replacements to interact with the bound CO. 2) The His 77 -Asn 42 H-bond, which is evident in the Fe[II]CooA crystal structure, appears to be broken upon CO binding. The strongest evidence is that the CO and FeC RR bands are unaffected when Asn 42 is replaced by the non-H-bonding residue Ala (or by any other residue), despite the known sensitivity of these bands to the H-bond status of the proximal histidine ligand in heme-CO adducts (19,20). Moreover, the position of the CO and FeC on the backbonding correlation reveal that the Fe-His bond is weaker in CooA-CO than in MbCO These are the findings that flow from the present study. Together with the Fe[II]CooA crystal structure, they lead to a model for the initial motions leading to activation of CooA. The proposed motions are shown as dashed arrows in Fig. 2.
Upon binding CO, the heme is suggested to slide upward, bringing the FeCO unit closer to the side chain of Leu 116 on the same subunit. As discussed above, the L116K variant is active in the absence of CO, suggesting that a Lys at this position induces the same heme displacement by displacing Pro 2 as the distal ligand. We modeled a Lys side chain in the orientation of the Leu 116 side chain and measured a 4.0-Å separation from the iron to the amine nitrogen. Thus a ϳ2.0-Å displacement of the heme group would be required to form a bond. This displacement provides an explanation for loss of the His 77 -Asn 42 H-bond, because His 77 remains bound to the heme and is displaced with it. It is likely that this heme displacement is also involved in the His 77 /Cys 75 ligand switch that is known to occur The hydrophobic pocket consists of the indicated residues, as well as Val 44 , Ser 78 , Leu 112 , and Ile 113 , which are either obscured in this view or have been omitted to show the heme ligands Cys 75 , His 77 , and Pro 2 . The irregularly shaped orange structure designates one 53-Å 3 cavity, which is proposed to expand to allow the heme to insert further into the hydrophobic pocket in the CooA Fe[III] and Fe[II]-CO states. Structure and cavity analyses were generated using the Swiss Pdb Viewer, version 3.7 program (9) (www.expasy.org/spdbv/); the figure was rendered using POV-Ray, version 3.1g software (www.povray.org/). upon oxidation to Fe[III]CooA (4), because the shifted heme position would accommodate Cys 75 ligation. Similar heme movement and ligand switching has been described for cytochrome cd 1 (35). Although WT Fe[III]CooA is not active, there are a number of mutations that induce activity in the Fe[III] form of the protein (13).
It is further suggested that when the heme is displaced upward, it is approached more closely by the C-helix from the second subunit, whose Gly 117 and Ile 113 residues help form the hydrophobic binding pocket for the CO. A modest motion of the second C-helix would permit H-bond donors at these positions to perturb the CO, whereas a much larger rearrangement of the first C-helix would be required to produce this effect. Thus the overall motion that initiates the conformation change to the active CooA-CO structure is viewed as a clockwise rotation of both hemes and both C-helices, in the plane of Fig. 2.
Some CooA molecules do not undergo this motion. The minority CooA-COЈ fraction represented by the 1962-cm Ϫ1 CO shoulder appears to have an intact His 77 -Asn 42 H-bond (the FeC/CO point falls on the Mb backbonding line), indicating that the heme is undisplaced from its position in the Fe[II]CooA structure. Presumably these molecules are not active in binding DNA. We do not know whether they are in a dynamic equilibrium with the active molecules or else are prevented from undergoing activation by some kind of energy barrier.
What is the driving force for the proposed heme and C-helix displacements? We propose that hydrophobic interactions are responsible (28), perhaps abetted by the electronic reorganization of the heme that results from a acceptor ligand (CO) replacing a donor ligand (the N terminus). Examination of the Fe[II]CooA crystal structure shows the heme to be partially exposed to solvent, whereas it would become more buried by sliding into the cavity (Fig. 11). Hydrophobic interactions would be enhanced by the suggested displacement of the opposite C-helix toward the heme. The hydrophobic character of the CO binding pocket is evidenced by the ϳtwo-unit pK a reduction of the displaced Pro 2 , relative to the expected aqueous pK a . CONCLUSIONS RR spectroscopy, combined with activity measurements, on variants of CooA has provided evidence about the nature of the CO adduct. Three C-helix residues, Ile 113 , Leu 116 , and Gly 117 , are shown to be close to the bound CO, forming a hydrophobic pocket. The displaced Pro 2 ligand is expelled from this pocket, unless the amine is unprotonated; the pK a for this process is 8.6. The RR evidence points strongly to breaking of the proximal His 77 -Asn 42 H-bond when CO binds. On the basis of the Fe[II]CooA crystal structure, these observations lead to the proposal that when CO binds, the heme is displaced into an adjacent cavity and is approached by the C-helix of the opposite subunit. These motions are suggested to trigger the protein conformation change required for DNA binding.