The role of the hydrophobic distal heme pocket of CooA in ligand sensing and response.

CooA from Rhodospirillum rubrum is a heme-containing transcriptional activator that becomes activated only upon binding CO. The basis for this specificity has been probed in a CooA variant, termed DeltaP3R4 CooA, lacking two residues adjacent to the Pro(2) heme ligand, which weakens that ligand. DeltaP3R4 CooA can bind imidazole and CN(-), as well as CO, and form a 6-coordinate low spin adduct with each. However, in contrast to the case with CO, imidazole and CN(-) do not stimulate the DNA binding activity of DeltaP3R4 CooA. This result indicates that the CO-specific activation of CooA is not merely the result of creation of a 6-coordinate CooA adduct but that there must be another element to this response. One feature of CooA activation is modest repositioning of the C-helices upon CO binding, so we altered a portion of the C-helix (residues Ile(113) and Leu(116)) located near the heme-bound CO in wild type CooA, and we investigated the effect on CO-specific activation. Surprisingly, the sizes of Ile(113) and/or Leu(116) positions are not critical for CooA activation by CO, disproving a precise interaction between these residues and the CO-bound heme as a basis for the CO activation mechanism and CO ligand specificity. In contrast, hydrophobic residues at these positions contribute to the activation. Some CooA variants altered at these positions in the background of DeltaP3R4 were also found to show low but reproducible activation in response to imidazole binding to the heme. A model for the role of hydrophobicity in CooA activation and specificity is suggested.

Rhodospirillum rubrum is a photosynthetic bacterium that can grow with CO as a sole energy source (1). The response of this organism to CO is regulated by the CooA protein, which binds CO and then activates the transcription of a series of genes encoding the CO oxidation system of R. rubrum (2). CooA contains a b-type heme prosthetic group as do many other gas-sensing proteins such as soluble guanylyl cyclase, FixL, DOS, and HemAT (3)(4)(5)(6). The heme of CooA is 6-coordinate and low spin in all oxidation and ligation states (7), indicating that the incoming CO must displace one of the internal protein ligands. The structure of Fe(II) CooA (without CO and therefore unable to bind DNA) revealed an unprecedented ligation arrangement for a heme protein wherein Pro 2 (N-terminal proline from the opposite subunit) and His 77 serve as the heme axial ligands (8). NMR studies (9) have indicated that Pro 2 is the ligand displaced by CO, and analysis of CooA variants has shown that His 77 , the retained ligand in Fe(II)-CO CooA, is critical for CooA activation by CO. In contrast to the critical role of His 77 , alteration of Pro 2 did not dramatically affect CooA activity in response to CO (10), disproving any important role of the displaced Pro 2 residue in the active form of CooA.
CooA belongs to the same family of transcriptional activators as the cAMP-receptor protein (CRP) 1 (11). Each of these proteins exists in equilibrium between an active form that can bind specific DNA target sequences and an inactive form that cannot. In each case, that equilibrium is shifted toward the active form by binding a small molecule effector, CO for CooA and cAMP for CRP. The structure of the effector-bound form of CRP has been known for some time (12), but the nature of the effector-free form has largely been speculative, so that the exact conformational change caused by effector binding is unknown. Conversely, although the effector-free structure of CooA has been reported, its effector-bound structure remains unsolved. Nevertheless, because CooA binds a DNA sequence reminiscent of that bound by CRP, it is a reasonable hypothesis that the effector-bound forms of the two proteins will be rather similar, so that a comparison of the two known structures is potentially informative. Each protein is a dimer, and each monomer contains two functionally distinct domains (Fig. 1A). The effector-binding domain of each protein senses its respective small molecule, which in turn leads to the repositioning of the DNA-binding domain. In each protein, the dimer interface is composed of a long ␣-helix (designated as C-helix) (8), and comparison of effector-free CooA with effector-bound CRP (8) reveals a repositioning of these two C-helices with respect to each other. Alterations of particular amino acids along these C-helices exert a variety of effects on activity in CRP (13,14) and fumarate and nitrate reductase activator protein (15), another member of this family of transcriptional activators. We have also mutationally repositioned the C-helices of CooA, mimicking the structure of effector-bound CRP, and this resulted in CooA variants with activity in the absence of effector (27). These facts suggest a role of the repositioning of the C-helices for the activation of these proteins in response to their respective effectors, and Fig. 1B compares the C-helix position of inactive CooA with that of active CRP. However, it is not clear how CO binding to the heme of CooA results in such C-helix repositioning.
One of the particularly interesting features of CooA is its specificity for CO as the only small molecule that can activate the protein. Small molecules that are weaker ligands than CO fail to displace the Pro 2 ligand, whereas NO displaces both protein ligands and leads to an inactive protein (16), and O 2 binding oxidizes CooA. It was therefore a reasonable hypothesis that the specificity for CO simply reflected its liganding strength. If correct, this would predict that any ligand that could displace Pro 2 yet allow a 6-coordinate heme would be similarly active. On the other hand, this result would not be seen if CO bound to the heme affected the achievement of the active form in ways other than simply creating a 6-coordinate species with Pro 2 displacement.
In this report, we have examined the issue of the role of CO in CooA activation and the related issue of CO specificity. We reasoned that any direct role of the heme-bound CO would almost certainly be affected by the protein residues nearest to that CO, and we have examined the importance of these through a combination of mutational, functional, and spectroscopic analyses.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and in Vivo Activity Assays-The construction of strains overexpressing wild-type (WT) CooA and CooA variants in an Escherichia coli background having a CooA-dependent ␤-galactosidase reporter system in the chromosome was described previously (17). In vivo activities of WT CooA and CooA variants were measured using the above system and quantitated using the standard protocol (18). All the CooA variants were constructed in a pEXT20-based expression plasmid that provides tight control of cooA expression (10).
Creation and Screening of cooA Mutations-Site-directed mutagenesis involved PCR amplification of cooA with primers designed to incorporate the desired nucleotide changes, as described elsewhere (19). The method used for codon randomization was essentially identical to the method used for site-directed mutagenesis, except that the primers contained randomized codons for the desired positions. Screening of CooA variants involved the analysis of their ability to cause ␤-galactosidase accumulation in colonies on agar plates incubated under different growth conditions as described previously (10). Based on colony color, CooA variants were classified as active, weakly active, and inactive. Selected variants were examined quantitatively by the in vivo ␤-galactosidase assay, after which the cooA genes were sequenced to determine the causative residue changes. Imidazole (25 mM final concentration) was used for the screening of imidazole-activated CooA variants.
Purification of WT CooA and CooA Variants-The purification of WT CooA and the CooA variants (Ͼ95% homogeneity) was performed as described previously (17). The heme content of CooA preparations was estimated using the extinction coefficient of WT CooA (7) or by a modified reduced pyridine-hemochromogen method (7), and protein concentration was measured using the BCA assay (Pierce).
Preparation of Hydroxylapatite Batch-treated CooA Samples-Preparation of hydroxylapatite batch-treated CooA samples was carried out using the procedure described previously (20). By this method, hemecontaining CooA was enriched to ϳ10% of total protein in case of WT CooA. These preparations were used for the preliminary measurement of UV-visible spectra and in vitro DNA binding activities of some CooA variants.
Measurement of Heme-containing CooA Accumulation-20 ml of the cells of WT CooA or CooA variants grown in 1ϫ MOPS-buffered media in the presence of CO were harvested by centrifugation. Cell pellets were dissolved in 50 l of H 2 O, vortexed, treated with 50 l of 2 N HCl, vigorously vortexed, and then treated with 1000 l of a 7:2 acetone/ methanol solution. This solution was vigorously vortexed and centrifuged, and the spectra of the supernatant were immediately measured. The peak intensity at 383 nm, normalized for cell mass, was used for determination of the heme-containing CooA accumulation.
In Vitro DNA Binding Assays-In vitro DNA binding assays of WT CooA and CooA variants were performed using the fluorescence polarization technique with a Beacon 2000 fluorescence polarization detector (Panvera Corp., Madison, WI) as described previously (10). For the measurement of DNA binding of CO-, imidazole-and CN Ϫ -bound ⌬P3R4 CooA at pH 9.5, the following assay buffer (high pH anisotropy buffer) was used: 40 mM glycine-NaOH, pH 9.5, 6 mM CaCl 2 , 50 mM KCl, 5% (v/v) glycerol, and 5 mM dithiothreitol. As a fluorescence probe, a 26-bp target DNA containing P cooF was labeled with Texas Red on one end of the duplex and used at the concentration of 6.4 nM. Salmon sperm DNA at 6.4 M was included in the reaction mixture to eliminate possible nonspecific DNA binding. Dissociation constants (K d ) were calculated by fitting of the binding data to an equation that incorporated a fluorescence quenching factor upon DNA-protein interaction as described elsewhere (21).
UV-visible Absorption Spectroscopy-UV-visible absorption spectroscopy of CooA samples was performed at room temperature in quartz cuvettes using a Shimadzu UV-2401PC spectrophotometer. UV-visible spectra of CooA samples were routinely obtained using 25 mM MOPS buffer, pH 7.4, with 0.1 M NaCl, unless stated otherwise. The UV-visible spectra of CO-, imidazole-, and CN Ϫ -bound ⌬P3R4 CooA were obtained with high pH anisotropy buffer for proper comparison with DNA binding activities of these forms.

RESULTS AND DISCUSSION
Weakening the Pro 2 Ligand Allows CooA to Bind Other Exogenous Ligands Such as Imidazole and CN-As noted in the Introduction, WT CooA is only activated in response to CO. It has been demonstrated that other potential exogenous ligands cannot create the 6-coordinate CooA adduct; NO is the only small molecule, other than CO, that can form a stable adduct to the heme of CooA (16), but NO binding results in a 5-coordinate form that appears to be inactive (16). This has led to the hypothesis that CO specificity results from its unique ability to create a 6-coordinate heme in CooA. To test if any exogenous ligand resulting in a 6-coordinate CooA adduct leads to CooA activation, we investigated a CooA variant that allows creation of 6-coordinate CooA adducts with exogenous ligands other than CO.
The simultaneous deletion of Pro 3 and Arg 4 (termed ⌬P3R4 CooA) perturbs the Pro 2 ligand in Fe(III) CooA, presumably by limiting the ability of Pro 2 to reach the heme (20). The following results show that ⌬P3R4 CooA is perturbed in the Fe(II) form as well and that imidazole and CN Ϫ can bind to this form. In the 6-coordinate adducts produced, the small molecules have presumably displaced the perturbed Pro 2 ligand, although this has not been demonstrated. Fig. 2A shows the UV-visible spectra of various forms of ⌬P3R4 CooA in high pH anisotropy buffer. A comparison of the spectra of Fe(III) and Fe(II) ⌬P3R4 CooA ( Fig. 2A) with those of WT CooA (data not shown) measured in the presence of 0.2 M KCl suggested the existence of a population of proteins in both forms of the ⌬P3R4 CooA with an open heme coordination site. The addition of imidazole (final 0.2 M) to Fe(II) ⌬P3R4 CooA resulted in increase of signal intensity with concomitant red shifts of Soret, ␣, and ␤ bands to 426, 560, and 530 nm ( Fig. 2A), respectively, indicating that imidazole-bound 6-coordinate CooA heme was formed. Similarly, the addition of KCN (final 0.2 M) into Fe(II) ⌬P3R4 CooA resulted in shifts of Soret, ␣, and ␤ bands to 435, 567.5, and 538.5 nm ( Fig. 2A), characteristic of a 6-coordinate low spin CN Ϫ adduct of the heme.
This imidazole-and CN Ϫ -binding property is also observed in other CooA variants with perturbed Pro 2 ligation. These include ⌬P3-I7 CooA (data not shown), which lacks five residues from Pro 3 to Ile 7 , and G117I CooA, where Ile 117 sterically perturbs Pro 2 ligation (20). In each case, it appears that the Pro 2 perturbation creates a population with an open heme coordination, to which these exogenous ligands bind. We assume that imidazole and CN Ϫ first bind on the Pro 2 side to the low level of 5-coordinate species in Fe(II) form and shift the equilibrium to the 6-coordinate adduct with the small molecule adduct. This open heme coordination position would almost certainly be the Pro 2 side, because Pro 2 is the endogenous ligand that is perturbed in these CooA variants. This implies that those exogenous ligands are binding to the same side of the heme as does CO to WT CooA. This situation allowed us to test the ability of "properly bound" exogenous ligands other than CO to activate CooA.
Imidazole and CN Ϫ Binding Cannot Activate ⌬P3R4 CooA-We then tested whether Fe(II)-imidazole or Fe(II)-CN Ϫ ⌬P3R4 CooA showed DNA binding activity under the same conditions as were used to measure the spectra. As shown in Fig. 2B, Fe(II) ⌬P3R4 CooA responds to CO with a significant increase in signal by fluorescence anisotropy but shows no response to imidazole or CN Ϫ . When CO was then added to head space of Fe(II)-imidazole or Fe(II)-CN Ϫ ⌬P3R4 CooA, DNA binding activity was restored in each case. The appearance of partial activity in the sample with both CN Ϫ and CO (Fig. 2B) reflects the ability of high levels of CN Ϫ to compete for heme binding with the modest levels of CO used, as revealed by UV-visible spectrum (data not shown). Surprisingly, CO addition to Fe(II)imidazole ⌬P3R4 CooA actually resulted in DNA binding activity that is reproducibly higher than that of the Fe(II)-CO form. Under this condition, CO completely displaced imidazole from the heme, as revealed by UV-visible spectrum (data not shown), so that the stimulation by imidazole was not due to its being a ligand. The exact mechanism of this interesting secondary effect of imidazole is under further investigation. In the absence of added effectors, Fe(III) and Fe(II) ⌬P3R4 CooA failed to show any DNA binding activity at this CooA concentration (Fig. 2B).
These results indicate that imidazole or CN Ϫ binding to Fe(II) ⌬P3R4 CooA cannot trigger the conformational change that leads to DNA binding, and therefore that CO provides a level of specificity for the CooA activation process in addition to its ability to form a 6-coordinate CooA adduct. If the nature of the small molecule ligand is sensed, then the surfaces of the C-helices in the vicinity of the heme are the obvious candidates for this sensing function for the following reasons. (i) The C-helices of CooA are the closest residues to the heme iron other than Pro 2 , at least in the known Fe(II) structure (8). (ii) Resonance Raman analysis indicated that Ile 113 , Leu 116 , and Gly 117 are the residues close to the bound CO. 2 (iii) As discussed in the Introduction, there are several lines of evidence that C-helix repositioning occurs upon CO binding and that 2 C. M. Coyle, unpublished data. this is a critical aspect of CooA activation. It was therefore an attractive hypothesis that this repositioning results from a direct interaction with the CO-bound heme, and we examined the residues in this region of the C-helices for their roles in CO-specific activation of CooA.
Hydrophobic Residues at Positions 113 and 116 Are Important for a Normal CooA Function-Because all C-helix residues near the distal side of the heme are hydrophobic (Fig. 1C), it was our working hypothesis that specificity in activation by CO would be the result of steric interaction between the bound CO and these residues (8). Based on the known structure of Fe(II) CooA, Ile 113 and Leu 116 are within 8 Å of the heme iron on Pro 2 side (8), and we have already shown a functional importance for Gly 117 , although the basis for this remains unknown (20). We therefore created CooA variants with substitutions of small (Ala) and large (Phe) residues at positions 113 and 116. Table  I shows in vivo activities of these CooA variants with a CooAdependent ␤-galactosidase reporter system, showing that these variants were somewhat altered in their CO responsiveness of CooA. It should be noted that there is excess amount of CooA in the cells under the conditions of these assays, so that activities below 80% actually represent a meaningful loss of CooA functionality. The basis for the low but significant CO-independent activity of I113F CooA is unknown.
To change more dramatically the distal heme pocket volume, we created two CooA variants with double substitutions, I113A/L116A and I113F/L116F CooA variants. Although I113A/L116A CooA was severely perturbed in its CO-sensing function, I113F/L116F CooA was relatively normal (Table I). The synergistic effect of the combination of the I113A and L116A substitutions led to the hypothesis that these two residues might form a functional pair in some way, such that modification of a single residue might have only a modest effect. In order to understand the functional requirements of that pair, we simultaneously randomized the codons for both residues 113 and 116 and screened for those CooA variants with significant activity in the presence of CO, as well as seeking less functional variants for comparison. Such a randomization and screening procedure allows a clearer understanding of the requirements at both positions for CO responsiveness, because a large number of possibilities are tested. Approximately 6,000 colonies, with cooA randomized at these two codons, were screened in the presence of CO. CooA variants with high activity (blue colonies; ϳ10%), intermediate activity (pale blue, ϳ30%), and negligible activity (white colonies; ϳ60%) colonies were seen. Selected variants were then examined more quantitatively for ␤-galactosidase activity in the presence and absence of CO and their cooA genes sequenced (Table II). The first conclusion is that, although a variety of residues can support normal CooA function at positions 113 and 116, neither charged nor hydrophilic residues are acceptable for good activity in the presence of CO. Such residues are absent among the normally active variants, yet are common among those variants with little or no activity (Table II). The requirements for a functional CooA appear to be rather more restrictive at position 116 than at position 113, as a narrower range of residues is found at that position among normally active variants. This suggests that residue 116 has a more important role than does residue 113 for the in vivo CO response of CooA.
Because Leu at position 116 was so common in the double random mutagenesis, we wanted to probe further the acceptability of other residues at that position. We then randomized only the codon for residue 116, screened in the presence of CO for variants with a range of activities, identified causative changes, and measured quantitative ␤-galactosidase activity in the presence and absence of CO. As suggested by the 113/116 double randomization results, hydrophobic residues at position 116 such as Val and Phe gave high activity in the presence of CO, whereas hydrophilic residues such as His, Asn, Arg, and Gly allowed very low activity (Table III).
Because of the possible synergy between the residues at positions 113 and 116, we analyzed the variants from the double-codon randomization for informative patterns. There was no apparent correlation between CO-dependent activity and the amino acid volume of either residue or of the sum of their volumes (data not shown). However, there was a clear correlation between hydrophobicity at position 116 and in vivo activity, with a suggestive pattern for the combination of the 113/116 positions (Fig. 3). At position 116 (Fig. 3A), there are three exceptions to the pattern as follows: L116C, L116H, and L116R CooA variants. L116C CooA can be explained by the fact that the hydrophobicity of Cys can assume a range of hydrophobic natures, depending on context, and that value used in other reports would move it to a more consistent position in the figure (22). L116R CooA accumulates heme-containing protein at negligible levels, so its inactivity probably reflects an absence of that species, as discussed below. L116H CooA is intriguing because His could be a ligand in the Fe(II) form, as has strongly been suggested for Lys in L116K CooA (23). We previously showed that Lys ligation in this variant led to the unusually high activity of Fe(II) form and decrease of its activity upon CO binding (23). It is a reasonable hypothesis that His 116 ligation could be the origin for this phenotype in L116H CooA.
At position 113, although the correlation between hydrophobicity and CooA function is much less obvious (Fig. 3B), nonhydrophobic residues were found at this position in CooA variants lacking in CooA function. Asp, Ser, and Asn were found to be detrimental: I113D CooA lacked CooA function, and Ser and Asn at 113 position were responsible for perturbed activities of I113S/L116I and I113N/L116C CooA variants, respectively, because Ile and Cys are acceptable at position 116 (Tables II and  III). Whereas the analysis in Fig. 3, A or B, reveals an important role of hydrophobicity at individual residue 116 or 113 in the CO response of CooA in vivo, the data shown in Fig. 3C suggest the importance of overall hydrophobicity (sum of hydrophobicity at residues 113 and 116). For example, this analysis suggests why I113A/L116A CooA displays poor CooA function, although the individual I113A and L116A substitutions would seem to be acceptable (Fig. 3C). The functional dependence on overall hydrophobicity might be the core of the previously hypothesized synergistic relationship between positions 113 and 116. a % activity indicates the mean value of multiple measurements of activity relative to that of WT CooA in the presence of CO and showed variability Ͻ10%.
b Cells grown anaerobically were used for the activity. c The ratio will be informative only when considered with "Red ϩ CO" activity since small changes of "Red" activity (denominator) potentially result in large differences in this ratio.
However, despite the importance of hydrophobicity at 113 and 116 residues, the size range of acceptable residues at these positions appears to disprove the original hypothesis that CO specificity results from a precise interaction between the CO and the C-helix residues.
Hydrophobic Nature at Position 113 and 116 Is Important for Proper Accumulation of Heme-containing CooA-Based on resonance Raman analysis of CooA variants, Ile 113 and Leu 116 , together with Gly 117 , are primary distal heme pocket residues of the Fe(II)-CO form of CooA. 2 Heme pocket residues are known to affect greatly the heme stability in myoglobin (24,25). We therefore examined the in vivo accumulation of hemecontaining CooA in some CooA variants in order to test whether the introduction of hydrophilic or charged residues at position 113 and/or 116 affected that property. The accumulation of heme-containing CooA was measured by the method described under "Experimental Procedures" on the same culture samples as were used for the in vivo activity measurement (1ϫ MOPS-buffered media, in the presence of CO). As shown in Fig. 4, all of the CooA variants tested accumulated heme less well than did WT CooA, and hydrophilic CooA variants such as I113D, L116R, and L116N failed to accumulate detectable levels of heme under this condition. I113K CooA was also highly perturbed in accumulation of heme-containing protein. This result suggests that hydrophobic residues at these positions are important for heme stability when CO is present. Fig. 4 gives a hint about the relationship between in vivo heme-containing CooA accumulations and in vivo CooA activities. Whereas WT CooA without IPTG induction (WT, 0 M) did not accumulate heme-containing CooA to detectable levels with this method, its in vivo activity was quite high (73%). This could mean that the decreased accumulation of heme-containing CooA with variants such as I113D, I113K, I116R, and I116N is only secondarily responsible for the dramatic loss of in vivo activities in those CooA variants. Clearly, the modest activity seen even at rather good levels of accumulation (variants L116Q, L116T and L116F in Fig. 4) implies that these variants must be seriously perturbed functionally. The in vivo activity of CooA is governed by a variety of factors including  b Cells grown anaerobically were used for the activity. c The ratio will be informative only when considered with Red ϩ CO activity since small changes of Red activity (denominator) potentially result in large differences in this ratio.
DNA binding, proper interaction with RNA polymerase, and the accumulation of heme-containing CooA. Because of their internal location in CooA, we doubt that 113 and/or 116 substitutions alter its interaction with RNA polymerase. Therefore, the above examination implies that hydrophobicity at positions 113 and/or 116 has a role in both proper conforma-tional response to CO and normal heme retention.
A Representative Hydrophilic CooA Variant, I113D, Is Perturbed in DNA Binding Activity-In order to probe more directly the effects of a hydrophobic residue in this region on the conformational response of CooA to CO, we purified a CooA variant with a representative hydrophilic substitution, I113D CooA, and directly measuring DNA binding activity of the protein. We chose I113D CooA variant because of its good heme stability during the preliminary manipulations and its adequate CooA accumulation in a heme-containing form under aerobic growth conditions in rich medium (LB). (It is unknown why the same I113D CooA variant accumulates poorly hemecontaining CooA in 1ϫ MOPS-buffered media anaerobically in the presence of CO.) Originally, we planned to purify a CooA variant with hydrophilic residues at position 116 as well, but we failed because such variants were unstable. Instability of heme-containing CooA variant at this position has been already reported (23). Fig. 5A shows the titration of a Texas Red-labeled DNA probe with the purified I113D CooA. Although Fe(III) and Fe(II) forms of 113D CooA did not show any DNA binding activity up to 4,000 nM CooA, the Fe(II)-CO form showed highly perturbed DNA binding activity corresponding to a K d value of 1,320 nM, in contrast to the 23 nM K d value of Fe(II)-CO WT CooA (Fig. 5A). Upon CO binding I113D CooA is therefore highly defective in performing the conformational change necessary for DNA binding. However, UV-visible spectra of I113D CooA were normal in all three forms (Fig. 5B), indicating that the perturbed DNA binding of Fe(II)-CO I113D CooA is not a result of a dramatic change of heme ligation states. These data show that the presence of a charged residue in this region greatly affects the ability of CooA to undergo the proper conformational change in response to CO.
Proper Heme Positioning May Be Critical for CooA Response to CO-In the analysis of I113D CooA above, there was no evidence that perturbation of the vicinity of the bound CO was the basis for the altered activity, and we wondered if the problem might instead be one of heme positioning. Namely, it is known that the heme of CooA must move with respect to the protein during oxidation and reduction (17), and it is certainly possible that there is further heme movement upon CO binding. A role of residues 113 and 116 might therefore be in proper heme positioning. We therefore asked if an indirect perturbation of the heme position might mollify the problem caused by the I113D substitution. We have already found that the I95W substitution can increase the CO-responsive activity of a variety of CooA variants, 3 consistent with the notion that this residue on the His 77 side of the heme might sterically move the heme to a position consistent with that in CO-bound WT CooA. Ile 95 lies on the B-helix (8), and bulkier substitutions might push the heme toward the C-helices, thereby reducing the heme pocket size, and reorienting the heme close to Leu 116 and away from Ile 113 in the known Fe(II) structure. For this reason, we introduced the I95W substitution into a strain that already had the I113D substitution. As shown in Fig. 5A, the K d value of Fe(II)-CO I113D/I95W CooA was determined as 98 nM, which is ϳ13-fold lower than that of Fe(II)-CO I113D CooA itself. In the absence of CO, DNA affinity was not detected (data not shown). In an otherwise WT background, the I95W substitution afforded an ϳ2-fold increase in DNA binding affinity over WT CooA (Fig. 5A). Therefore, the I95W substitution can clearly assist Fe(II)-CO I113D CooA in achieving the active conformation. Although we do not know Fe(II)-CO CooA structure, it is highly unlikely that this I95W effect is due to the 3 H. Youn, unpublished data. direct interaction between Trp 95 and Asp 113 because they are so distant in the Fe(II) structure. Rather, the effect of I95W is almost certainly a direct one on the heme itself, although the UV-visible spectra of I113D/I95W CooA were not perturbed in any of the forms (data not shown), so there was no observable change in the heme ligation states. Therefore, we conjecture that the nature of the defect in I113D CooA might be a perturbed heme-C-helix interaction due to the mis-positioned heme that is cured by the I95W substitution.
Identification of Imidazole-activated CooA Variants-One approach to understanding the basis of CO specificity and the process of activation of CooA is to identify and analyze CooA variants that respond to other effectors. The physical barrier provided by the Pro 2 ligand is apparently not the only factor in specificity, because we have shown that imidazole and CN Ϫ binding fail to activate ⌬P3R4 CooA. Given the minor importance of the displaced Pro 2 in CO-sensing function (10), it seems likely that any additional level of CO specificity might be provided by C-helix residues in the distal heme pocket. To test this notion, we allowed the binding of imidazole by weakening the Pro 2 ligand, and we then asked if any residues at position 113 and 116 might support imidazole-responsive CooA activation. We pooled plasmids encoding all of the variants listed in Tables I-III and introduced the mutation encoding ⌬P3R4 into that mixture, so that ⌬P3R4-containing derivatives of all the variants were created, although mixed together. This mixed pool was then introduced into the reporter strain so that individual plasmids could be screened for their effects on in vivo activities anaerobically in the presence of 25 mM imidazole. Although most colonies remained white, some turned blue (active). These blue colonies were collected and then re-streaked onto agar plates with and without imidazole, and a comparison of the results identified imidazole-dependent CooA variants. Clones displaying imidazole-dependent ␤-galactosidase activity were sequenced. Table IV shows the quantitative in vivo activities of selected imidazole-responsive CooA variants with control, WT CooA, ⌬P3R4 CooA in "anaerobic," "anaerobic ϩ imidazole," and "anaerobic ϩ CO" conditions. The four imidazole-activated CooA variants showed 6 -18% full activity under the anaerobic ϩ imidazole condition, which is dramatically above the activity seen with WT or ⌬P3R4 CooA (Table IV). As explained earlier in this report, these levels of activity would be relatively modest in an assay where CooA was limiting, but the response is still striking. It is also interesting that ⌬P3R4 L116T CooA, which showed the highest in vivo activities in the presence of imidazole, is only slightly more active in the presence of CO. The negligible effector-free activity of ⌬P3R4 L116T CooA also indicates that the ligand-free form is not close to the active conformation, and imidazole induces a substantial conformational change in the CooA variant. The UV-visible spectra of the hydroxylapatite-batch preparations of the imidazole-activated CooA variants (listed in Table IV) in Fe(II) forms were all changed upon addition of 25 mM imidazole (data not shown), implicating that imidazole was a ligand in these conditions. Preliminary analysis of in vitro DNA binding by ⌬P3R4 I113F/L116F CooA in the presence of saturating imidazole showed DNA affinity below that of WT CooA with CO but well above background. These results demonstrate that imidazole binding in this variant does not lead to fully active CooA. However, this screening has hardly optimized the potential of CooA to respond to imidazole, but the fact that modification of positions 113 and 116 give readily detectable activity demonstrates a clear role of these residues in effector specificity once the physical barrier (Pro 2 ligation) was removed.  Summary and Working Hypothesis-CooA and CRP function as dimer. In each protein, the two C-helices provide a dimerization interface. The comparison of the structures of inactive Fe(II) CooA with active cAMP-bound CRP revealed the repositioning of the C-helices with respect to each other, and it became our working hypothesis that CO binding activates CooA by inducing this repositioning. The importance of the repositioning in CooA activation has been confirmed by the findings of CooA variants where alteration of positions 121-126 generated CooA variants with significant effector-independent activity (27).
The goal of this study was to determine the basis for the CO specificity of WT CooA and, by implication, the mechanism by which CO binding causes C-helix repositioning. The simple hypothesis that CO specificity was due to the selective ability of CO to displace on Pro 2 was disproved by the demonstration that other effectors could bind to ⌬P3R4 CooA, yet not lead to CooA activation. Structural and spectral evidence had indicated that positions 113 and 116 were close to the heme-bound CO in WT CooA (8), 2 and the results reported here show the significance of these residues in response to effectors. The nature of these residues is clearly important for activation in response to CO binding as well as for proper heme retention, with the specific residue at position 116 being particularly critical. As described below, we believe that these effects can be explained with a model whereby a hydrophobic pocket is critical for both heme retention and response to CO. We note that the C-helices in the CooA dimer assume a coiled-coil helical structure where Leu 116 is in the d position and Ile 113 is in the a position. The residue requirements for an optimal leucine zipper are also substantially consistent with the acceptable residues that we detect at positions 113 and 116. Nevertheless, those CooA variants with very good leucine zipper residues (including WT CooA) at these positions do not result in effectorindependent activity, indicating that the leucine zipper effect itself is not sufficient to afford the transition energy to active conformation.
Although these residues play an important role in activation, their role in CO specificity is less clear. We expected that specificity might be provided by steric contacts between the heme-bound CO and these residues, but the variety of acceptable residues at these positions for CO activation disproves that hypothesis. Nevertheless, an important role of these residues in effector specificity is certainly shown by the results with imidazole. The weakening of the Pro 2 ligand (by the ⌬P3R4 alteration) permits binding of imidazole, but it is not sufficient for activation. In contrast, perturbation of residues 113 and 116 in the ⌬P3R4 background allows significant activation in response to imidazole binding, although the molecular rules that govern this response have not been elucidated. Although it is not clear if the process of activation by imidazole in these variants is mechanistically similar to that of WT CooA by CO, it nevertheless establishes a potential role for these residues in sensing the heme-bound effector. It might well be that CO specificity is provided by the exclusion of other small molecules by an unknown mechanism and that a variety of residues have this property. Further analysis of the requirements of residues at 113 and 116 to various small molecules should clarify this important issue.
Our present working hypothesis for the results in this paper is the following. CO binding displaces Pro 2 , which is protonated and therefore expelled from the hydrophobic heme pocket. Therefore, CO binding will certainly expose the hydrophobic C-helix residues near the CO-bound heme in Fe(II)-CO CooA. The hydrophobic interaction between the CO-bound heme and residues 113 and 116 allows it to reposition within that hydrophobic cavity, which directly or indirectly affects C-helix repositioning. In this view, hydrophobicity at positions 113 and 116, and exclusion of water from the heme cavity of Fe(II)-CO CooA, would be critical for CO responsiveness of CooA. Proper positioning of the CO-bound heme by the residues at positions 113 and 116 is also important, as suggested by the ability of the I95W substitution to restore activity to certain variants, presumably through a different heme-positioning mechanism.
The results reported here show that C-helix residues 113 and 116 are important for heme retention, effector response, and effector specificity in CooA. Further analysis will better define the molecular basis of those effects.