The heme pocket afforded by Gly117 is crucial for proper heme ligation and activity of CooA.

CooA, a CO-sensing homodimeric transcription activator from Rhodospirillum rubrum, undergoes a conformational change in response to CO binding to its heme prosthetic group that allows it to bind specific DNA sequences. In a recent structural study (Lanzilotta, W. N., Schuller, D. J., Thorsteinsson, M. V., Kerby, R. L., Roberts, G. P., and Poulos, T. L. (2000) Nat. Struct. Biol. 7, 876-880), it was suggested that CO binding to CooA results in a modest repositioning of the C-helices that serve as the dimer interface. Gly(117) is one of the residues on the C-helix within 7 A of the heme iron on the Pro(2) side of the heme in CooA. Analysis of a series of Gly(117) variants revealed altered CO-sensing function and heme ligation states dependent on the size of the substituted amino acid at this position; bulky substitutions perturbed CooA both spectrally and functionally. A combination of spectroscopic and mutagenic studies showed that a representative Gly(117) variant, G117I CooA, was specifically perturbed in its Pro(2) ligation in both Fe(III) and Fe(II) forms, but comparison with other CooA variants indicated that perturbation of Pro(2) ligation is not the basis for the lack of CO response in G117I CooA. These results have led to the hypothesis that (i) the heme and the C-helix region move toward each other following CO binding and the interaction of the heme with the C-helix is crucial for CooA activation, and (ii) this event occurs only when a properly sized heme pocket is afforded.

Proteins that sense small gaseous molecules, such as NO, O 2 , and CO, play important roles in biological systems ranging from prokaryotes to mammals. These include soluble guanylyl cyclase, which binds NO and exerts a variety of physiological responses, including neurotransmission and smooth muscle vasodilation (1)(2)(3), and FixL, which binds O 2 and regulates gene expression in nitrogen-fixing rhizobia (4,5). The only CO-sensing protein described thus far, CooA, was found in the photosynthetic bacterium Rhodospirillum rubrum, which can utilize CO as a sole energy source (6). CooA uses a b-type heme to bind CO, whereupon the protein undergoes a conformational change that allows it to bind DNA and activate the transcription of genes encoding the CO-oxidation system (7).
CooA is a homodimeric protein containing an effector-binding domain, which contains the heme, and a DNA-binding domain linked by a hinge region. The heme of CooA is sixcoordinate and low spin in its Fe(III), Fe(II), and Fe(II)-CO forms, implying that CO must displace one of the axial ligands of Fe(II) CooA. Recently, the x-ray crystal structure of Fe(II) CooA was solved to 2.6-Å resolution and revealed that Pro 2 of one monomer serves as one ligand to the heme iron of the other monomer and that the ligand trans to Pro 2 is His 77 (8). Previous studies, including electron paramagnetic resonance (EPR) 1 analysis and site-directed mutagenesis showed that one of the ligands in the Fe(III) form is Cys 75 in the thiolate form (7,9), and Pro 2 is assumed to be the trans ligand, based on indirect evidence (10,11). Thus, there is an unusual ligand switch upon reduction of the heme, with His 77 replacing Cys 75 (7,12). CO binding has been shown to displace Pro 2 (10,11,13), but a variety of Pro 2 variants turned out to be CO-responsive, indicating that Pro 2 is not critical for activation of CooA in response to CO (10). Cys 75 is also not essential for activity, because C75S CooA shows CO-dependent activity, although the heme is unstable in the Fe(III) form (7). The nature of the residue at position 77 is, however, absolutely critical for CO-dependent conformational changes of CooA (7). CooA variants altered at this position still bind CO but fail to be activated. In at least one such variant, CO is apparently bound to the "wrong" side of the heme (13), which might be the basis for the lack of CO activation.
CooA is a member of the cAMP receptor protein (CRP)/ fumarate and nitrate reductase activator protein (FNR) family of transcriptional regulators (14). These proteins exert their function by binding a cognate DNA sequence after responding to their specific effectors (cAMP for CRP; reducing conditions for FNR; CO for CooA). The only structure known for CRP is the cAMP-bound form (15), and a variety of analyses have suggested that a substantial conformational change occurs upon effector binding (16 -18). No structure has been obtained for FNR, although it appears to be activated by a very different mechanism of a monomer-dimer transition rather than a conformational change within a dimer, as in the case of CRP and CooA (19). The structural comparison of Fe(II) CooA in the absence of effector (CO) with CRP bound to its effector (cAMP) suggested a number of differences between them, among which was a modest repositioning of the long C-helices that serve as the dimer interface in both CRP and CooA (8). This has lead to the hypothesis that in CooA, effector binding causes a change in the relative positions of the C-helices, which transmits a signal to the DNA-binding domains. Presumably this repositioning either destabilizes the inactive conformation, stabilizes the active conformation, or both.
It is unclear how CO binding might affect the repositioning of the C-helix, although it seems plausible that CO displacement of Pro 2 allows the heme, still tethered to His 77 , to interact with the C-helices. To test this hypothesis, we examined the surface of the C-helices in the structure of Fe(II) CooA and noted that there are relatively few residues that are in a position to interact with the heme (8): Gly 117 , Leu 112 , Ile 113 , and Leu 116 . In this paper, we report the importance of the heme pocket affected by substitution at the Gly 117 position and the analysis provides a number of insights into the response of wild-type (WT) CooA to CO.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-WT CooA and CooA variants were constructed in an Escherichia coli overexpression system and a ␤-galactosidase reporter system as described previously (7).
In Vivo ␤-Galactosidase Assay-In vivo ␤-galactosidase activity in an E. coli reporter system was monitored as described previously (7) and quantitated using the standard protocol (20).
Purification of WT CooA and CooA Variants-The purification of WT CooA and CooA variants to ϳ95% purity was performed as described previously (7). The heme content of CooA preparations was quantitated by the reduced pyridine-hemochromogen method (21).
Preparation of Hydroxyapatite Batch-treated CooA Samples-For the initial screening of the heme coordination states of different CooA variants, a 50-ml culture of the cells was harvested, resuspended in 5 ml of 25 mM MOPS buffer, 0.2 M NaCl, 10% glycerol, pH 7.4, broken with a French pressure cell (ϳ120 MPa), and centrifuged for 30 min at 11,700 ϫ g. The supernatant was then mixed with 0.3 g of solid hydroxyapatite resin. After unbound materials were removed from the resin, a high salt buffer containing 25 mM MOPS, pH 7.4, 10 mM potassium phosphate, pH 7.4, 1.2 M KCl, and 5% glycerol was added, and the resin was washed twice. CooA was then eluted with high phosphate buffer containing 25 mM MOPS, pH 7.4, 160 mM potassium phosphate, pH 7.4, 50 mM KCl, and 5% glycerol. The eluent was precipitated with ammonium sulfate with a final saturation of 50% and stored at Ϫ20°C until use. CooA prepared with this procedure resulted in an enrichment of CooA to ϳ10% of total protein.
Spectroscopic Measurements-Electronic absorption spectra of CooA variants were measured with a Shimadzu UV-2401 spectrophotometer. Anaerobic reduction of CooA samples and anaerobic addition of endogenous ligands such as CO, KCN, and imidazole were carried out as described previously (7,22,23). EPR spectra were recorded on a Varian E-15 spectrometer with an Oxford Cryostat 3120 system to regulate and monitor the temperature at a microwave frequency of 9.25 GHz at 4 K and 200 microwatts (W) for high spin signals and 22 K and 20 W for low spin signals. EPR spectra at pH 7. In Vitro Fluorescence Polarization Assay-In vitro DNA binding assays were performed using the fluorescence polarization assay described previously (10). Target DNA containing pCooF (purchased from Genosys) was labeled with the fluorescent dye, Texas Red, on one end of the duplex and used at the concentration of 6.4 nM. Dissociation constants (K d ) were calculated by fitting the binding data to a non-linear fitting equation described by Lundblad et al. (24).

Ligation Structure and in Vivo Activity of Gly 117 Variants
Are Closely Related with Residue Size-The crystal structure of Fe(II) CooA revealed that relatively few residues are present in the vicinity of the heme on the side of Pro 2 , one of the axial ligands in Fe(II) CooA (8). These residues include Gly 117 , and we tested the hypothesis that the portion of the pocket provided by this residue might be important for heme interaction with the C-helix upon CO binding (Fig. 1). The set of Gly 117 variants studied here included G117A, G117S, G117V, G117I, and G117H. As shown in Table I, the activities of these Gly 117 variants in response to CO are severely perturbed in our in vivo assay. As the size of substitution at the position 117 increased (see "mole vol " column in Table I), CO-sensing function was reduced. This CO-sensing function is indicated in Table I by the ratio of the activity of each CooA variant in the presence of CO to that in the absence of CO; this ratio is an approximate measure of the -fold activation caused by effector binding. There is also a less striking correlation between residue size and effector-independent activity: Larger residues at position 117 provided greater effector-independent activity in the Fe(III) and Fe(II) forms (Table I). It is important to note that the differences in activity in vivo of the variants are not the result of poor accumulation of heme-containing CooA, because we have shown that ϳ5% active CooA (compared with WT CooA) is sufficient for maximal in vivo activity (25) and all the Gly117 variants accumulate Ն25% of WT CooA (Table I).
The Soret wavelength maxima of Gly 117 variants are also listed in Table I. Although substitutions at position 117 with relatively smaller side chains such as Ala and Ser showed wavelength maxima that were indistinguishable from that of WT CooA, G117V CooA and G117I CooA revealed a blueshifted Soret peak at 387 nm in the Fe(III) form, characteristic of a five-coordinate high spin heme. The Fe(II) forms of G117V CooA and G117I CooA are also severely perturbed in terms of Soret maxima, and the detailed spectral properties of purified G117I CooA will be discussed below. The extent of the perturbation of both UV-visible spectra and the CO-sensing function in Gly 117 variants was well correlated with the size of the substitutions (Table I), with the exception of G117H, which will be discussed below. This result indicates that the portion of the heme pocket created by Gly 117 is important for the ligation structure and CO-sensing function of CooA.
Cys 75 , but Not Pro 2 , Is a Ligand in Fe(III) G117I CooA-As shown in Fig. 2B, the UV-visible spectrum of purified Fe(III) G117I CooA is typical for a thiolate-ligated five-coordinate high spin heme with a Soret peak at 387 nm and ligand-to-metal charge-transfer peak appearing at ϳ640 nm. The EPR spectrum of Fe(III) G117I CooA corroborated the results of the UV-visible spectrum, displaying only high spin signal with a single population corresponding to g x ϭ 7.15, g y ϭ 5.02, and g z ϭ 1.93 ( Fig. 3 and Table II). This result indicates that one of the two axial ligands in the six-coordinate Fe(III) WT CooA ( Fig.  2A) is selectively perturbed in the Fe(III) G117I CooA variant. The rhombicity (expressed as E/D (26)) of G117I CooA is comparable to that of a model heme-thiolate complex (Table II  ( 27)), although it is a bit lower than those of Fe(III) fivecoordinate, high spin thiolate-ligated hemoproteins such as cytochrome P450cam (P-450cam (28)), endothelial nitric oxide synthase (26), and chloroperoxidase (29) (Table II). On the other hand, the rhombicity of G117I CooA is significantly higher than those of nitrogen-based ligand high spin signal such as FixL (30) and soluble guanylyl cyclase (31). The results of UV-visible and EPR spectroscopies strongly suggest that the retained ligand in Fe(III) G117I CooA is a thiolate.
Because Cys 75 is a normal ligand in Fe(III) WT CooA, we tested the hypothesis that the thiolate ligand seen above was due to this residue. We reasoned that altering Cys 75 in a G117I background would have a profound effect on the protein if Cys 75 was the ligand but should have no effect if another Cys residue served that role. A strain with both G117I and C75A substitutions in CooA accumulated Ͻ1% the normal level of CooA, consistent with the hypothesis that Cys 75 is the ligand in Fe(III) G117I CooA (Table I)  is that these variants display a significant amount of sixcoordinate heme under conditions where the altered residues would have served as the normal ligand (7,10). This indicates that there are adventitious ligands that partially substitute for the altered ligands. UV-visible and EPR spectra of Fe(III) P2Y CooA, but not of Fe(III) WT CooA, are pH-dependent at the range from ϳ7 to ϳ10 (10), which presumably reflects a property of an adventitious/exogenous ligands. However, the Fe(III) G117I CooA has no such adventitious ligand, and we therefore expected that it would also lack pH effects in spectral analyses. As anticipated, the UV-visible spectra of Fe(III) G117I CooA were not changed in the pH range of 3.4 to 9.5 (data not shown). Consistent with the results of UV-visible spectra, EPR spectra of Fe(III) G117I CooA were not changed at pH 6.5 and 9.5, although the EPR spectrum at pH 3.4 showed a slight amount of another five-coordinate high spin population, indicated by a shoulder at g x ϭ 7.77 (Fig. 3), which is not significant. Addition of imidazole or cyanide anion (CN Ϫ ) failed to change the UVvisible spectrum of Fe(III) G117I CooA (data not shown) which indicates that exogenous ligands are also precluded. It is our hypothesis that the substituted Ile at position 117 is so close to the heme iron that it sterically precludes native Pro 2 ligation as well as potential adventitious/exogenous ligands.
Ligation States of Fe(III) G117I and Fe(III) ⌬P3R4 CooA Are Different-Because Pro 2 is precluded from ligating the heme iron in Fe(III) G117I CooA, we also investigated the ligation state of Fe(III) ⌬P3R4 CooA, another variant perturbed in Pro 2 ligation, to know if the perturbation of Pro 2 ligation is sufficient to result in the spectral properties of Fe(III) G117I CooA. As expected, the UV-visible spectra of ⌬P3R4 CooA revealed the highly perturbed ligation state of the Fe(III) form (Fig. 2D).
However, unlike G117I CooA, the UV-visible spectrum of Fe(III) ⌬P3R4 CooA at pH 7.4 exhibited a mixed-spin state with a five-coordinate high spin heme as a major population, indicated by a 387-nm Soret peak and small amount of sixcoordinate low spin heme, indicated by a shoulder of 423 nm (Fig. 2D). EPR spectra of Fe(III) ⌬P3R4 CooA showed two kinds of high spin signals, in which the major one roughly corresponded to that of Fe(III) G117I CooA, whereas the minor one was similar to that of other high spin thiolate-ligated hemes such as P-450cam, endothelial nitric oxide synthase, and chloroperoxidase (Table II). Unlike Fe(III) G117I CooA, the spin state of Fe(III) ⌬P3R4 CooA was found to be pH-dependent based on UV-visible (data not shown) and EPR spectra (Fig. 3).
Finally, the open coordination site of Fe(III) ⌬P3R4 CooA was accessible to exogenous imidazole and CN Ϫ as judged by the changes in the UV-visible spectrum (data not shown). Thus, although Pro 2 ligation appears to be perturbed in the Fe(III) forms of both G117I and ⌬P3R4 CooA, the different effects of pH and exogenous ligands on the spectra indicate that the perturbations are not identical.
Pro 2 Ligation Is pH-dependent in Fe(III) WT CooA-We had previously found that WT CooA appeared to start precipitating as the pH was lowered to ϳ6, so our ability to lower the pH of G117I CooA to 3.4 initially surprised us. We therefore wondered if WT CooA could also be analyzed at this pH and found that a rapid transition to a lower pH allowed WT CooA to remain in solution. We therefore asked how low pH perturbed the ligation state of WT CooA. As shown in Fig. 4A, as the pH was decreased from 7.4, the fraction of five-coordinate high spin form of Fe(III) WT CooA increased as indicated by the Soret band increase at 387 nm in the UV-visible spectrum. Above pH 7.4, no significant spectral changes were observed, which is consistent with a previous study (7). Fitting the absorbance at 387 nm as a function of pH using the Henderson-Hassalbach equation showed a single protonation process with a pK a of ϳ5.5 (Fig. 4B). This is most easily rationalized by supposing that one of the ligands is being protonated and displaced at acidic pH. To identify the retained ligand at lower pH, we measured the EPR spectrum of Fe(III) WT CooA at pH 3.4. As shown in Fig. 3 and Table II, the EPR spectrum of Fe(III) WT CooA at pH 3.4 displayed a high spin rhombic symmetry with g values of g x ϭ 7.29, g y ϭ 4.99, and g z ϭ 1.92 similar to that of Fe(III) G117I CooA, suggesting that the retained ligand of Fe(III) WT CooA at this pH is Cys 75 , not Pro 2 . Our apparent ability to protonate Pro 2 at low pH and cause its deligation suggests that it is the neutral form of Pro that serves as the normal ligand, in at least the Fe(III) form. The EPR spectrum of Fe(III) WT CooA also exhibited a minor high spin signal having axial symmetry with g values of g x ϭ g y ϳ6 and g z ϭ ϳ2 (Table II), which might reflect free heme.
His 77 Is a Retained Ligand and Pro 2 Is a Perturbed Ligand in Fe(II) G117I CooA-The known CooA Fe(II) structure (8) suggested that bulky substitutions at position 117 might produce steric hindrance between Pro 2 and the substituted residue, so we anticipated that the Fe(II) form of G117I CooA would be highly perturbed. As expected, the UV-visible spectrum of Fe(II) G117I CooA is significantly different from that of Fe(II) WT CooA, characterized by a reduction of the peak intensity in Soret, ␣, and ␤ bands with a concomitant appearance of a shoulder at ϳ440 nm (Fig. 2B), which suggests a five-coordinate high spin Fe(II) heme. To test this hypothesis, we investigated the effect of pH on spectral features and the ability of exogenous ligands such as imidazole and CN Ϫ to bind the heme iron. As shown in Fig. 5B, upon raising the pH to 9.5 the peak intensity in Soret, ␣, and ␤ bands was increased at the expense of the ϳ440-nm shoulder. Also, addition of imidazole (Fig. 5A)  or CN Ϫ (Fig. 5B) to Fe(II) G117I CooA produced significant changes in the UV-visible spectra. These results are consistent with the existence of an open coordination site in Fe(II) G117I CooA, strongly suggesting that the Fe(II) form of G117I CooA contains a significant amount of five-coordinate heme and that Ile 117 does not preclude exogenous ligands in this redox state, in contrast to the Fe(III) state.
The two ligands in Fe(II) WT CooA are Pro 2 and His 77 (8). To identify the retained and perturbed ligand of Fe(II) G117I CooA, we compared the UV-visible spectrum of Fe(II) G117I CooA with those of G117I ⌬P3R4 (data not shown) and G117I H77A CooA (Fig. 2D). The summary listed in Table I shows that the UV-visible spectrum of Fe(II) G117I CooA corresponds exactly to that of G117I ⌬P3R4 CooA but is markedly different from that of G117I H77A CooA. This result suggests that His 77 is the retained ligand, and that Pro 2 is the perturbed ligand, in Fe(II) G117I CooA.
Bulky Residues at Position 117 Do Not Appear to Interact with Bound CO-In contrast to the significant perturbation of ligation states in Fe(III) and Fe(II) G117I CooA, the UV-visible spectrum of Fe(II)-CO G117I CooA was highly similar to that of WT CooA (Fig. 2B), with a similar A Fe(II)-CO /A Fe(III) ratio (Table  I). This similarity is surprising, given the extreme functional perturbation seen in vivo (Table I). To test the hypothesis that bulky residues at position 117 interact with the bound CO but fail to perturb the UV-visible spectrum, we performed a preliminary resonance Raman spectroscopic analysis of purified G117H CooA. We reasoned that G117H has a similar bulk, as does G117I, and causes similar effects on in vivo activity (Table  I), but histidine is more likely to reveal an interaction between the position 117 residue and the CO in resonance Raman spectroscopy. The analysis showed the same Fe-CO stretching band (487 cm Ϫ1 ) as did WT CooA, 2 suggesting that the bound CO in G117H CooA is in a similar environment as it is in WT CooA and is therefore apparently not binding to the opposite side of the heme as has been proposed to be the basis for the inactivity of H77Y CooA (13). The absence of an effect by His 117 itself on this frequency indicates that this introduced residue does not interact effectively with the bound CO. We did not pursue resonance Raman analysis of G117I CooA because its frequency shift of the Fe-CO stretching band would be much lower than that of G117H CooA. This result implies that the lack of CO-sensing function of G117H CooA (and possibly G117I CooA) is not caused by either hindering the CO-binding process or changing the environment of bound-CO.
In   fluorescence polarization method. As shown in Fig. 6A, the Fe(II)-CO form of purified G117I CooA showed only negligible DNA binding above 1 M CooA. The failure of G117I CooA to bind DNA in the presence of CO is not due to a poor affinity for CO; consistent with its open heme coordination state, its K d for CO is actually lower than that of WT (data not shown). This result indicates that the functional perturbation of Fe(II)-CO G117I CooA is due to a poor ability to bind target DNA. Surprisingly, Fe(III) G117I CooA showed a concentration-dependent increase of anisotropy value at somewhat high protein concentrations, corresponding to a K d of 1.8 M. This property of Fe(III) G117I CooA is different from that of Fe(III) WT CooA (Fig. 6B), implying that some structural change that is caused by the Ile substitution at the Gly 117 position partially mimics the DNA-binding conformation of the CO-bound form of WT CooA. Although the Fe(III) form of G117I CooA is much more active than the Fe(II) and Fe(II)-CO forms in terms of in vitro DNA binding, the in vivo activity of Fe(III) G117I CooA is very low, similar to those of Fe(II) and Fe(II)-CO G117I CooA. We were surprised by this low activity, because other variants with this level of affinity for DNA typically display substantial activity in vivo (data not shown). This result therefore suggests that, although the structure of Fe(III) G117I CooA supports DNA binding, it does not allow the proper interactions with RNA polymerase that are necessary for transcription activation (32). It is our hypothesis that CO binding to WT CooA causes conformational changes that properly position the DNAbinding domains and also the regions of CooA that interact with RNA polymerase; we are actively pursuing this latter issue.
We were concerned that the substitutions at position 117 might affect DNA binding indirectly by precluding CooA dimerization, because Gly 117 lies on the C-helices that form the dimer interface. However, gel filtration analysis of Fe(III) G117I CooA showed that it migrated as a single peak at the position of dimeric WT CooA (data not shown).
Functional Perturbation of Gly 117 Variants Does Not Originate from the Alteration of Pro 2 Ligation-Because some of the Gly 117 variants appeared to prevent Pro 2 from ligating the heme in the Fe(III) and Fe(II) forms, we investigated the in vitro DNA binding properties of ⌬P3R4 CooA. In contrast to the case of G117I CooA, ⌬P3R4 CooA showed effective DNA binding in the Fe(II)-CO form, corresponding to a K d of 158 nM (Fig.  6C). Under the same experimental conditions, Fe(II)-CO WT CooA displayed a K d of 16.2 nM (Fig. 6B), a value roughly similar to that reported in a previous study (10). These results indicate that the direct perturbation of Pro 2 ligation in ⌬P3R4 CooA does not eliminate CO-responsive DNA binding but does decrease the affinity of the CO-bound form for DNA, implying that the CO-bound form of ⌬P3R4 CooA is not optimal. Although the deletion of Pro 3 and Arg 4 residues must perturb Pro 2 ligation, it is possible that Pro 2 is not eliminated as a ligand, and we therefore investigated the activity of a deletion variant, which lacks five N terminus residues (from Pro 3 to Ile 7 ) and is highly unlikely to have Pro 2 in the vicinity of the heme. As shown in Table I, ⌬P3-I7 CooA still retained some ability to respond to CO in vivo, in contrast to the lack of response in G117I CooA. This phenotype is quite similar to those of variants completely missing Pro 2 such as P2Y CooA (10) or CooA ⌬N5 (Pro 2 -Phe 5 deleted (11)). This result substantiates the notion that the perturbation of Pro 2 ligation seen in some Gly 117 variants is not the basis of the lack of function in these variants. The severe functional perturbation of G117A CooA and G117S CooA (Table I) despite apparent normal ligation in the Fe(III) and Fe(II) forms also supports this.
To rule out the steric interaction between the substituted amino acids at position 117 and Pro 2 as the basis for the lack of CO-sensing function of Gly 117 variants, we also constructed G117A, G117H, and G117I CooA variants in the background of ⌬P3R4. Like single Gly 117 variants, these Gly 117 ⌬P3R4 double variants also showed loss of CO-sensing function (Table I) size of the substituted amino acid at this position, although even an Ala substitution was found to be sufficient for some perturbation of CO-sensing function. Because Gly ranks next to Pro as a helix breaker (33), and even Ala perturbs CO-responsiveness, we cannot completely rule out helix perturbation as a functional aspect of the Gly 117 variants. However, we note that there is no helix bending at Gly 117 in the solved Fe(II) CooA structure (8). Furthermore, given the fact that Ala is statistically the best helix former (33), but is the second best in terms of CO-responsiveness, the role of Gly 117 being a "helix breaker" seems unlikely. Fig. 7 provides a schematic representation of our proposed ligation states of WT, G117I, and ⌬P3R4 CooA. As noted in the figure, ⌬P3R4 CooA has a mixture of five-and six-coordinate hemes in both the Fe(II) and Fe(III) states, but the identity of the residues that provide the sixth ligand is unknown; it might be Pro 2 itself, the N terminus, or some adventitious ligand. In contrast, the Fe(III) form of G117I CooA contains only fivecoordinate high spin heme with Cys 75 as one axial ligand. Upon reduction, G117I CooA undergoes a ligand switch in which His 77 replaces Cys 75 . Given that Pro 2 ligation is totally precluded in Fe(III) G117I CooA, Pro 2 is not critical for the ligand switch in G117I CooA. In this regard, the relatively low rhombicity of high spin signals in the EPR spectrum of Fe(III) G117I CooA is intriguing. Other high spin thiolate-ligated hemes such as P-450cam, endothelial nitric oxide synthase, and chloroperoxidase display higher rhombicity than G117I CooA and do not undergo any ligand switch. The low level of interaction between the sulfur of Cys 75 and iron valence orbitals indicated by the low rhombicity of Fe(III) G117I CooA may make the ligand switch possible upon reduction, but the paucity of other ligand switch examples makes this difficult to test. The Fe(II) form of G117I CooA also contains a significant amount of five-coordinate heme with the Pro 2 ligand being highly perturbed. Taken together, the G117I substitution perturbs only the Pro 2 ligand and is therefore superficially similar to the situation in ⌬P3R4 CooA.
Despite these similarities, the functional properties of G117I and ⌬P3R4 CooA are significantly different. Although the G117I CooA shows DNA-binding activity in the Fe(III) form, it does not undergo CO-dependent activation. On the other hand, ⌬P3R4 CooA appears to be functionally rather normal in vivo (i.e. only active in the Fe(II)-CO form), even though its DNAbinding affinity in vitro is approximately ten times lower than that of WT CooA. It is therefore unlikely that the lack of CO-sensing function in G117I CooA originates from merely perturbing the Pro 2 ligation.
The strictly five-coordinate high spin heme in Fe(III) G117I CooA and the inability of any adventitious/exogenous ligand to access the open coordination site of Fe(III) G117I CooA indicates that the substituted Ile is very close to the heme iron. We assume that this close proximity is the basis for the surprising DNA-binding activity of Fe(III) G117I CooA. It is our hypothesis that in the Fe(III) state of G117I CooA, the heme/C-helix interaction, partially mimics that of heme-CO/C-helix in WT CooA, although by a different specific mechanism. In contrast, the steric interaction of the substitution Ile at position 117 with the CO-bound heme is presumably responsible for the lack of CO-sensing function of G117I CooA.
Smaller substitutions such as Ala or Ser at 117 position do not perturb the ligation states of CooA based on UV-visible spectra. However, the G117A substitution does exert a severe effect on CO-sensing function of CooA as demonstrated by the absence of noticeable DNA binding up to 1 M heme in the fluorescence polarization assay (data not shown). It is possible that Ala 117 of G117A CooA might actually be closer to the heme in the Fe(II)-CO form than in the Fe(II) form and that steric hindrance between the substituted Ala and the heme is responsible for the functional perturbation of G117A CooA. The apparently distinct effects of the G117A substitution on different ligation states of CooA would not be surprising, because it is well established in WT CooA that the heme must move with respect to the protein matrix upon reduction (8).
Taking these results together, we suggest that (i) CO binding is followed by heme or/and the protein matrix movement toward each other, and that the interaction of the heme with the C-helix is crucial for CooA activation; and (ii) this event occurs only when there is a sufficient heme pocket afforded by Gly 117 . This hypothesis is consistent with the previous prediction based on the structural comparison between effector-free CooA and effector-bound CRP that the repositioning of the two Chelices might be required for CooA activation (8). This proposed close proximity of the heme to the C-helix in the Fe(II)-CO form might create the small CO pocket suggested by the rapid recombination rate of CO rebinding after CO photolysis (34).
The results with G117H CooA are surprising. Although this variant is functionally similar to G117I CooA (Table I), it is spectrally distinct, showing apparently normal UV-visible spectra under all conditions. Although we do not know the basis for this difference, it is our working hypothesis that His 117 is serving as a ligand in this variant, a function that would not be possible for the similarly sized Ile 117 .
In summary, this study shows that the heme pocket afforded by Gly 117 is crucial for proper ligation and activity of CooA. The molecular basis for the functional perturbation in G117I CooA is probably steric interaction between the substituted Ile and heme, which might hinder heme movement and proper interaction with the C-helix required for CooA activation after CO binding. It remains to be determined if the CO bound to the heme is directly involved in such proper interaction.