Characterization of Variants Altered at the N-terminal Proline, a Novel Heme-Axial Ligand in CooA, the CO-sensing Transcriptional Activator*

CooA, the carbon monoxide-sensing transcription factor from Rhodospirillum rubrum, binds CO through a heme moiety resulting in conformational changes that promote DNA binding. The crystal structure shows that the N-terminal Pro2 of one subunit (Met1 is removed post-translationally) provides one ligand to the heme of the other subunit in the CooA homodimer. To determine the importance of this novel ligand and the contiguous residues to CooA function, we have altered the N terminus through two approaches: site-directed mutagenesis and regional randomization, and characterized the resulting CooA variants. While Pro2appears to be optimal for CooA function, it is not essential and a variety of studied variants at this position have substantial CO-sensing function. Surprisingly, even alterations that add a residue (where Pro2 is replaced by Met1-Tyr2, for example) accumulate heme-containing CooA with functional properties that are similar to those of wild-type CooA. Other nearby residues, such as Phe5 and Asn6 appear to be important for either the structural integrity or the function of CooA. These results are contrasted with those previously reported for alteration of the His77 ligand on the opposite side of the heme.

The sensing of dissolved gas molecules by proteins in biology has recently attracted considerable biochemical interest. The role of nitric oxide in a variety of important biochemical processes (1,2), and its receptor, soluble guanylyl cyclase (sGC) 1 have been well documented in eukaryotic systems (3,4). FixL, which modulates the expression of genes responsible for nitrogen fixation in rhizobia, is an example of an oxygen sensor (5,6). An oxygen sensor in Escherichia coli, termed DOS ("direct oxygen sensor"), has been reported although its physiological role remains undefined (7). Finally, for carbon monoxide (CO), CooA, the CO-oxidation activator protein, modulates the expression of genes required for the utilization of CO as a sole energy source in the photosynthetic bacterium Rhodospirillum rubrum (8). All of the above mentioned proteins have in common a heme prosthetic group to which their respective gas molecules bind. The binding event is then followed by a conformational change in the protein that effects activity.
Numerous studies have clearly demonstrated the physiological importance of CO in a wide variety of processes (9 -11), and although sGC has been implicated in sensing CO (12)(13)(14), direct evidence of a CO-receptor in eukaryotic signal transduction systems is lacking. CooA senses CO through a heme moiety and represents the current model system for biological CO-sensing. CooA belongs to the cAMP receptor protein (CRP) and fumurate and nitrate reductase activator protein superfamily of transcriptional activators (15). CooA is an ϳ50-kDa homodimeric protein, with each monomer possessing a heme that is six-coordinate under all oxidation and ligation states (16). Interestingly, an unusual redox-mediated axial ligand exchange occurs in CooA in that a cysteine ligand (Cys 75 ) in the oxidized (Fe III ) form is replaced by a histidine ligand (His 77 ) upon reduction of the heme-iron to the Fe II form (17)(18)(19)(20). The identification of the axial ligand trans to the Cys 75 /His 77 pair, which is believed to be provided by the protein based on the observation that the Fe III and Fe II forms of CooA are sixcoordinate and low-spin, had been unclear although spectroscopic studies have suggested a neutral nitrogen ligand (19,20). Finally, the ligand that is displaced upon binding CO remains speculative.
Recently, the three-dimensional structure of Fe II CooA has been solved by x-ray diffraction techniques (21). This report showed that the general folding topology of CooA was indeed similar to that of CRP (22). In addition to the verification of His 77 as one of the heme-axial ligands in Fe II CooA, inspection of the structure identified the other axial ligand as an Nterminal proline residue (Pro 2 ; Met 1 is removed by processing) from the other subunit of the dimer. This structural environment represents an unprecedented axial ligation arrangement for a heme protein.
In a previous study (18), we altered His 77 and found that the UV-visual spectra of these variants was normal in the Fe III form (when Cys 75 is the ligand on that side of the heme) but was perturbed in the Fe II form, when His 77 would normally be the ligand. Surprisingly, while the UV-visible absorption spectra of the Fe II -CO forms of these His 77 variants was similar to that of wild type, they were unable to undergo the conformational change in response to CO that is necessary for DNA binding (18). In addition, His 77 variants in the Fe II form bind cyanide with high affinity (23). Alteration of His 77 therefore perturbs the Fe II form of CooA as well as its ability to properly respond to CO binding. In this present study, we have altered the other side of the heme face in Fe II CooA, i.e. the Pro 2 ligand and contiguous residues, and have found that variants in that region show surprising functional similarity to WT CooA. We then purified a selected variant and characterized its biochemical parameters in detail. Implications concerning the mechanism of activation in the WT CooA protein are discussed.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The construction of E. coli strains expressing WT CooA, CooA variants altered in the region of Pro 2 by site-directed mutagenesis, and of strains expressing WT CooA and CooA variants in combination with a chromosomally encoded ␤-galactosidase reporter system have been described previously (18). Site-directed variants were constructed in a pKK233-3-based expression plasmid (24), while region randomization variants were constructed in a pEXT20-based expression plasmid (25) as the screening methodology (described below) depends on the tighter control of CooA expression provided by this vector.
In Vivo ␤-Galactosidase Activity Assays-Strains containing the P cooF -lacZ reporter fusion were grown in rich medium containing 100 g/ml ampicillin (18). These inocula were then diluted 40-fold into anaerobic 120-ml stoppered serum vials containing 20 ml of MOPSbuffered medium (23) supplemented with 100 g/ml ampicillin and 25 M isopropyl-␤-D-thiogalactopyranoside. For CO-induced cultures the headspace was made 2% (v/v) with CO gas. Cultures were grown with shaking at 30°C to mid to late-log phase (A 600 0.7 to 1.5), cell pellets were prepared and frozen, and ␤-galactosidase activities measured (26).
"Region Randomization" of the Pro 2 Region-To determine the functional importance of residues near the N terminus of CooA, as well as to create some novel variants for future analysis, we employed an unusual mutagenic approach of the complete randomization of certain portions of the cooA gene. This approach will be described in more detail elsewhere, but the salient features are the following. A library of cooAexpressing clones was created, by PCR amplification with an oligonucleotide fragment bearing the randomized region (described below) and then the resulting cooA library was ligated into pEXT20 and transformed into the P cooF -lacZ reporter host. The resulting clones were screened for CooA-dependent expression of ␤-galactosidase in the presence of CO. Only clones that provided a reasonable level of function were then sequenced and this provided a sense of the range of acceptable residues throughout the randomized region. The method is similar to that of phage display (27) except that our approach uses a genetic screen for interesting clones, whereas phage display uses a physical screen.
Simultaneous randomization of residues 2-3 or 2-6 was achieved using multiple sets of primers (Genosys) and three separate rounds of PCR using Pfu turbo DNA polymerase (Stratagene). Randomization primers were synthesized with a mixture of all four nucleotides at each base in the region to be randomized; this region was flanked by wildtype sequence. All other primers used in the process were non-mutagenic. In the first round of PCR, a second primer was used with the randomization primer to create a pool of DNA fragments, each completely random at the same multiple specified positions (i.e. the positions coding for residues 2-3 or 2-6) but otherwise encoding a WT portion of CooA; for technical reasons, the template contained a frameshift mutation within the region to be randomized. In the second round of PCR amplification, a second pair of primers was used to produce a DNA fragment bearing the remainder of WT CooA; this fragment overlapped with a WT tail of the randomized fragment. The fragments from the first and second rounds of PCR were then annealed at their overlapping regions, extended, and amplified in a third round of PCR, resulting in a pool of DNA fragments each containing the entirety of cooA with a randomized region of interest. Restriction sites incorporated into the external primers allowed directional ligation into pEXT20-based plasmids (25) with subsequent high-efficiency electroporation into the aforementioned ␤-galactosidase reporter host (UQ1639). Transformed cells were spread on 1 ϫ MOPS plates (23) and active variants were identified after anaerobic growth in the presence of 1% CO at 30°C and were retested for CO-dependence. The randomized regions of these variants were sequenced.
Purification of WT and P2Y CooA-The purification of WT CooA and the P2Y variant (Ͼ95% homogeneity) were performed using procedures described previously (18). The heme content of CooA preparations were quantified using the reduced pyridine-hemochromogen method (28). N-terminal analysis of purified P2Y CooA was performed at the Macromolecular Structure Facility of Michigan State University. UV-visible Absorption Spectroscopy of Cell-free Lysates-Preparation of cell-free lysates and their analysis by UV-visible absorption spectroscopy from E. coli strains expressing WT and variant CooAs were performed as described (18).
Approximate Comparisons of CooA Accumulation-Cell pellets from 1.5-ml culture were lysed enzymatically at room temperature for 20 min in a buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, 1 mg/ml lysozyme) and subsequently treated with DNase for 10 min in this buffer amended to 100 mM NaCl, 20 mM MgCl 2 , and 5 g/ml DNase I. Samples were centrifuged at 20,000 ϫ g for 10 min. The Soret peak intensity, corrected for a ϪCooA extract blank, was measured in 150 l of supernatant and compared with the accumulation of wt CooA.
UV-visible Absorption Spectroscopy of Isolated Proteins-In this study, all spectra were acquired as described (23). Dependence of the spectra of P2Y Fe III CooA on the pH of the solution were performed using ϳ4 M isolated P2Y Fe III CooA in the following buffers: pH 6.5, 7.0, and 7.5; 0.1 M MOPS, pH 8.0, 8.5, and 9.0; 0.1 M Tris HCl, pH 9.5, 10.0, and 10.5; 0.1 M glycine/NaOH. Absorbance changes as a function of pH in the 388-nm Soret peak (high-spin component) and in the 423-nm Soret peak (low-spin component) were used to estimate the pK a of the spectral transition.
Electron Paramagnetic Resonance Spectroscopy-Isolated P2Y Fe III CooA was buffered by mixing 200 l of the isolated protein (145 M) with 50 l of 500 mM buffer (pH 7.0, MOPS, and pH 10.0, glycine/ NaOH) to provide a final buffer concentration equal to 100 mM and a final heme concentration ϳ120 M. Samples were then degassed on an argon manifold to remove dissolved oxygen, frozen, and stored at 77 K. Spectra were recorded on a Varian E-15 spectrometer, with an Oxford Cryostat 3120 system to monitor and regulate the temperature. The magnetic field was measured using a Varian 929801 gaussmeter with a Tektronix type RM 503 oscilloscope and the microwave frequency (9.2 GHz, X-band) was monitored using a Hewlett Packard 5255A frequency counter. The spectra were recorded at 4 K and 200 microwatt power for the low-field region (30 -2030 Gauss), and at 23 K and 50 microwatt power for the high-field region (1850 -3850 Gauss). Spectral analysis was performed using a "rhombogram" computer program 2 for the calculation of theoretical g values.
Fluorescence Polarization Assay for DNA Binding-We have developed an assay for detecting DNA binding by the CooA⅐CO complex, based on fluorescence polarization (FP) (30,31), in order to test quantitatively the activity (i.e. DNA binding in response to CO binding to the heme of CooA) of our CooA protein preparations. The extent of FP of a freely tumbling molecule that has been conjugated with a flourescent tag is directly related to molecular volume when viscosity and temperature are held constant. The FP technique exploits changes in the molecular volume of a target molecule as a result of macromolecular interactions. Thus, the extent of FP of a fluorescently labeled DNA fragment (containing the promoter sequence of cooF, termed P cooF ) increases upon binding by active CooA. The value of FP is measured as anisotropy, which is a dimensionless parameter related to the intensity of parallel and perpendicular absorbed light components compared with the total intensity (31). The FP technique has been applied successfully to other DNA-binding protein systems, including CRP (32) and the tryptophan repressor (33). The following conditions were arrived at after testing effects of ionic strength, cation concentration, and protein and probe concentration (data not shown).
Equimolar amounts of two complementary 26-base pair oligonucleotides (containing the cooF promoter, P cooF : 5Ј-ATAACTGTCATCTGGC-CGACAGACGG-3Ј), one of which was 5Ј end-labeled with Texas Red (Genosys), were mixed and heated at 95°C for 2 min in a water bath and then allowed to anneal by cooling slowly to room temperature over a 2-h time period. Sufficient hybridization (Ͼ90%) was verified by resolving the double-stranded probe using non-denaturing polyacrylamide gel electrophoresis (10% (w/v); in Tris borate-EDTA, pH 8.0) against a 26-base pair single-stranded oligonucleotide control. Binding assays were performed in the following buffer components at their respective final concentrations: Tris-HCl, pH 8.0 (20 mM), CaCl 2 (6 mM), KCl (50 mM), glycerol (5% (v/v)) and dithiothreitol (1 mM). The DNA probe (6.4 nM) and varying concentrations of purified CooA samples (ϳ0.5-800 nM) were mixed in 5 ϫ 60-mm test tubes that were fitted with rubber septa. Samples were then reduced by the addition of sodium dithionite to 2 mM. Binding assays were initiated by the addition of a small volume of CO-saturated water determined to fully saturate CooA, followed by incubation at room temperature for 5 min to achieve equilibrium. Fluorescence polarization (anisotropy) was measured at 25°C using a Beacon 2000 fluorescence polarization detector (PanVera Corp., Madison, WI) using 594 nm excitation and 620 nm emission filters designed for measuring the Texas Red fluorescence. Dissociation constants (K d ) were calculated by non-linear curve fitting of the binding data (corrected for quenching) using the scheme of Lundblad et al. (31).

Analysis of CooA Variants Altered in the Pro 2 Region of
CooA-Because of the novelty of the Pro 2 ligand, we created a small set of CooA variants that were expected to perturb that ligand. It is known that the terminal Met is removed when the second residue is small, but not when it is large (34), so a variety of changes were made as noted in Table I.
Initial in vivo analyses of these variants, using a CooA-dependent ␤-galactosidase reporter system, showed that these Pro 2 region variants all retained CO-responsive CooA activity (Table I). Because the levels of WT CooA in the reporter strain are in excess of that necessary to fully saturate the reporter, CooA variants with poor accumulation but high specific activity are readily detected. By this assay, all Pro 2 region variants examined have a high level of CO responsive activity. In contrast, CooA variants altered at His 77 show virtually no response to CO, although some display a modest level of CO independent activity (Table I), which has been noted previously (17,23). Some Pro 2 region variants accumulate much less well than does WT CooA based on heme spectra, and we assume that this reduced level of accumulation reflects heme instability. It is interesting that the Pro 2 region variants in Table I that accumulate best are those with an "extended" N terminus, resulting from the presumed retention of Met 1 ; the retention of the terminal Met 1 was confirmed by N-terminal sequencing of purified P2Y CooA (data not shown). Variants at position 77 generally accumulate to fairly normal levels. In the Fe II and Fe II -CO forms, the Pro 2 region variants studied displayed Soret absorbance maxima that were very similar to that of WT CooA (data not shown). However, in the Fe III spectrum of all of these variants there was an additional Soret maxima to approximately 388 nm, which suggests a mixture of normal WT-like six-coordinate, low-spin species (denoted by a Soret peak at ϳ423 nm) and high-spin, five-coordinate species (denoted by a Soret peak at ϳ388 nm); there was variation in the proportion of these two forms among the various Pro 2 region variants (Table I). This spectral feature is further described below with a purified protein.
These results clearly indicate that, while Pro 2 appears to be optimal for heme accumulation and activity, a wide variety of alterations of the N terminus allow significant accumulation of heme-containing protein and retention of CO responsive activity. The observation that CooA variants altered at either His 77 or Pro 2 display some fraction of six-coordinate, low-spin heme, suggests that at least one adventitious ligand is available on each side of the heme and this issue is discussed later.
Functional Requirements for CooA Residues 1-6 -The sitedirected mutagenesis results provide an insight into the requirements for the N terminus, but were necessarily limited to the analysis of a small number of changes. We therefore employed a different approach, which we term region randomization, in which a library of cooA-containing clones was created in which codons 2-6 (or, in a second analysis, codons 2-3) were completely randomized as described under "Experimental Procedures." Clones expressing CO-responsive CooA were screened for those that accumulated detectable levels of hemecontaining CooA (based on cell pellet color compared with that of WT), and the cooA gene sequenced.
In the randomization of codons 2-6, approximately 39,000 randomized clones were screened and 46 displayed a reasonable level of ␤-galactosidase in response to CO. These were screened for detectable heme accumulation and the five clones with readily detectable heme-containing CooA are shown in Table II. While screening 39,000 clones does not test all possible amino acid sequences when 5 codons are randomized, it does reveal positions where there are, or are not, strong selections for heme accumulation and CO-responsiveness. In the randomization of codons 2-3, a much smaller number of clones was screened to obtain those in Table II, but technical problems prevent an estimation of frequency. The results from both randomizations demonstrate that a variety of possibilities are satisfactory as positions 2 and 3 (Table II). It is important to note that bulky residues at position 2 are also expected to retain the Met 1 , so that these functional clones differ in N-terminal length as well as specific sequence.
The crystal structure shows that Arg 4 is about 4 Å from a heme propionate (21). We therefore anticipated that this residue would be important and were surprised by its absence among randomized clones (Table II, 2-6 randomization, and data not shown). We therefore created the variants R4A and R4S (as R4S was fairly common among the functional randomized variants). These variants displayed modest heme accumulation and significant CO responsive activity (Table I), suggesting the nature of the residue at this position is of utility although not critical. We subsequently constructed ⌬Pro 3 -⌬Arg 4 , whose heme accumulation and activity are consistent with this hypothesis (Table I).
Phe 5 was clearly selected for in the codons 2-6 randomization and the crystal structure suggests that the presence of this residue might simply be a structural requirement. Among randomized variants that had significant function, but relatively less heme, other aromatic residues were common at this position, consistent with this structural requirement (data not shown). Finally, Asn (the residue in WT CooA) or Ser have clearly been selected at position 6, although we cannot exclude the possibility of other satisfactory residues at that position because of the limited sample size. The reason for the apparent preference for these residues may be structural as the crystal structure (21) shows that position 6 is the start of a helix and this residue may be involved in a hydrogen bond with Asn 9 .
These results indicate that a variety of substitutions at positions 2 and 3, even those that cause the addition of a residue at position 1, can result in reasonable heme accumulation and CO-dependent response, although the WT sequence appears optimal. The adventitious ligand that replaces Pro 2 in these variants is unknown and the possibility of more than one such ligand cannot be disproved. The N-terminal amine remains a possibility although the functionality of the single and double residue deletions near the N terminus suggests that this might be less likely in these variants.
Spectral Analysis of Fe II and Fe II -CO P2Y-To extend the above analysis of the Pro 2 region variants in cell-free lysates, we isolated P2Y CooA to study its biochemical and functional properties in detail. This variant was chosen because it accumulated reasonable levels of heme-containing CooA and because we expected it to be significantly different in structure from WT CooA, due to the retention of Met 1 , as well as the replacement of Pro 2 with Tyr 2 . The UV-visible absorption spectra of P2Y CooA compared with those of WT CooA are depicted in Fig. 1. Surprisingly, there are no detectable perturbations in the absorbance maxima of the Fe II and Fe II -CO forms of the P2Y CooA when compared with those of WT CooA. This result indicates that the Fe II and Fe II -CO spectral forms of P2Y CooA are similar to WT CooA in that they are both six-coordinate and low-spin, indicating a strong-field trans-axial ligand. There were also no observable changes in the UV-visible absorption spectrum of the P2Y Fe II and Fe II -CO forms of CooA as a function of pH (data not shown). Both of these results are in striking contrast to those seen for Fe III P2Y described and discussed below.
The fact that Fe II P2Y is six-coordinate and low-spin implies that a relatively strong field adventitious ligand has replaced Pro 2 and the absence of a pH effect would appear to suggest that this ligand is not water or hydroxide. P2Y Fe III CooA was also reduced with titanium-(III) citrate and yielded the same reduced spectrum (data not shown) which should also eliminate the possibility that the adventitious ligand was an oxidation product of the dithionite typically used in the reduction of CooA. Combined with the mutational results above, it is therefore our working hypothesis that the adventitious ligand in the a Predicted protein sequence of residues the N terminus. For most variants, residues 2-6 are shown, but Met 1 is shown where the sequence predicts its retention. The randomization of codons 2-3 (middle data set) does not perturb positions 4 -6 and these are shown in brackets.
b % accumulation relative to that of WT CooA based on the reduced spectrum for Pro 2 region variants. Fe II P2Y is some other unknown protein ligand.
Functional Analysis of Fe II -CO P2Y-Because the various Pro 2 region variants have substantial functionality in vivo, yet have certainly been perturbed in the vicinity of the heme, we analyzed the functional properties of purified P2Y in greater detail, using a quantitative assay for CooA binding to the P cooF promoter DNA sequence based on fluorescence polarization. When very low amounts of purified WT Fe II -CO CooA, which had been shown previously to be active in DNase I footprinting experiments (16), are added to the assay, the value for anisotropy is relatively low (Fig. 2). Upon addition of higher concentrations of WT Fe II -CO CooA to the assay, CooA-DNA interactions increase the molecular volume of the complex and the value of anisotropy increases until saturation. The intensity of the Texas Red-labeled DNA samples decreased slightly during the assay with increasing addition of Fe II -CO CooA (ϳ10% at saturating protein levels; data not shown), which was attributed to fluorescence quenching. Therefore, we included a quench correction factor into our non-linear curve fitting equation according to Lundblad et al. (31). By this analysis, WT Fe II -CO CooA binds one of its cognate promoters (P cooF used in this study) with a reasonable affinity (K d ϭ 12.7 Ϯ 2.3 nM, K a ϭ 7.9 ϫ 10 7 M Ϫ1 ), while Fe II -CO P2Y CooA has an order of magnitude decrease in its affinity (K d ϭ 102 Ϯ 2.3 nM, K a ϭ 9.8 ϫ 10 6 M Ϫ1 ). Spectroscopic analysis of P2Y Fe II -CO CooA (and WT) showed that CO was saturating under the conditions of this analysis, eliminating the possibility that the lower DNA affinity was actually a result of poor affinity for CO. These results are consistent with the in vivo results in which P2Y has substantial activity; as noted before, the excess of CooA in the reporter strain precludes our ability to detect modest reductions of CooA activity. The basis for the reduced DNA affinity of P2Y CooA is discussed below. As a control, purified H77Y Fe II -CO, which had been shown to be incapable of binding DNA in previous DNase I footprinting experiments (18) and is not responsive to CO in vivo (Table I) (18) also showed no increase in anisotropy values up to protein concentrations of 1 M under the same conditions (data not shown). Although the affinity of WT Fe II -CO CooA is lower than that of (cAMP) 1 -CRP (i.e. one cAMP bound per dimer) for site-specific DNA (K d ϭ 1.2 nM, K a ϭ 8.4 ϫ 10 8 M Ϫ1 ) (32), the direct comparison of DNA affinities for active CooA and active CRP for target DNA is problematic because of the different experimental conditions used for each assay. These results are consistent with the hypothesis that Pro 2 is not critical for the proper response of CooA to CO.
Analysis of P2Y Fe III CooA-Although the Fe III form of CooA is not involved in DNA binding, its analysis will be important as it is critical for the proper redox-dependent ligand switch in CooA (18). The perturbation of the UV-visible spectra (Tables I  and II, Fig. 1) is also consistent with the view that changes in the Pro 2 region perturb Fe III CooA.
In the UV-visible absorption spectrum of P2Y Fe III CooA at pH 7.4 (Fig. 1A), a spin equilibrium is evident in that there is a split Soret peak at 388 and 423 nm, denoting five-coordinate, high-spin and six-coordinate, low-spin fractions, respectively. In addition, there is an enhanced peak with the Fe III P2Y CooA variant at 640 nm (relative to that of WT CooA), often associated with high-spin, five-coordinate hemes. Low-spin Fe III heme proteins, such as cyano-Met sperm whale Mb (35) and b-type cytochromes (36) typically have Soret peaks from 422 to 429 nm, while high-spin Fe III heme proteins, such as those found in catalase (37) and FixL (38) have Soret peaks from 391 to 417 nm. While the data suggests that some fraction of the P2Y Fe III CooA exists as a high-spin, five-coordinate thiolate linkage, other hypotheses are possible and this issue is pursued below with EPR spectroscopy.
We examined if this spin-mixture equilibrium could be perturbed by the pH of the solution. Fig. 3 shows that upon raising the pH of the solution, there is a decrease in the Soret ABS 388 nm (high-spin species) and a concomitant increase in the Soret ABS 423 nm (low-spin species). However, even at pH 10.5, a significant fraction of the five-coordinate, high-spin species remained. The pH titration spectra display an isosbestic point at 410 nm, consistent with a simple two-state transition as a function of pH. We could not measure the effect of pH changes in P2Y Fe III CooA below pH 6.5 as the protein becomes unstable and begins to aggregate (data not shown). These results suggest a simple water-to-hydroxide ligation transition as a function pH in P2Y Fe III CooA. An alternative explanation for the pH dependence of P2Y Fe III CooA is that there is a deprotonation of some residue near the vicinity of the Fe III heme that can now act as a ligand in the absence of the natural Pro 2 ligand. However, the fact that the curve fitting above pH 8 is relatively poor implies that there is apparently another unknown transition in this region (see below).
EPR spectroscopy of P2Y Fe III CooA corroborated the results with UV-visible spectroscopy. An increase in the pH of the solution resulted in a dramatic decrease in the intensity of high-spin forms of P2Y Fe III CooA and a concomitant increase in the spin-quantity of the low-spin forms (as determined quantitatively by comparison to a CuEDTA standard) from ϳ0.2 to ϳ0.8 spins/heme (Fig. 4). Inspection of the high-spin region of the spectrum of P2Y Fe III CooA revealed the presence of two distinct S ϭ 5/2 systems. Analysis of the g values using a rhombogram analysis (29) for an S ϭ 5/2 system with g ϭ 2 indicates the spectrum is comprised of a system with theoretical g values of g z ϭ 1.63 (assumed; see below), g y ϭ 3.59, and g x ϭ 8.07, and a system with theoretical g values of g z ϭ 1.95 (assumed; see below), g y ϭ 5.04, and g x ϭ 6.89 (Table III). The high-spin system has unusual relaxation properties in that it is only observable at relatively high power and very low temperature (200 microwatts and 4 K). Under these conditions, the low-spin (S ϭ 1/2) features are saturated and appeared as a dispersion line shape that dominated the high-field signals arising from the S ϭ 5/2 system (data not shown). At lower powers and higher temperature (20 -50 microwatts, 23 K), the low-spin system can be observed as well defined derivativeshaped features, although the high-spin features are completely unobservable under these conditions. Therefore, the g x values from the high-field region (S ϭ 5/2) systems can only be assumed from the rhombogram analysis.
The EPR data of P2Y Fe III CooA in the low-field region can be interpreted as a mixture of two major ligation states. The g values observed for the two high-spin systems are absent in WT CooA (18) (Table III). The g values for one of the major highspin systems (g z ϭ 1.63, g y ϭ 3.59, and g x ϭ 8.07) are similar to those that are found in Fe III high-spin, five-coordinate thiolateligated hemes such as P-450cam (39), H93C myoglobin (39,40), and the H175C/D235L double mutant of cytochrome c peroxidase (41). The g values for the other major high-spin system (g z ϭ 1.95, g y ϭ 5.04, and g x ϭ 6.89) are similar to those found in Fe III high-spin, five-coordinate histidine-ligated hemes such as FixL (38) and sGC (42,43) and five-coordinate tyrosine-ligated hemes, such as H93Y myoglobin (44) and catalase (37) at low temperatures.
Both of these high-spin signals decrease as a function of pH. At neutral pH, the five-coordinate neutral nitrogen species represents a higher fraction when compared with the fivecoordinate thiolate species. However, at high pH, these two species are roughly equivalent. This observation implies that during the pH titration, the five-coordinate thiolate species may be an intermediate to the formation of the low-spin species that predominates at high pH. Because Cys 75 is the normal ligand in WT Fe III CooA, it is very likely that this residue is the source of the thiolate signal.
The six-coordinate, low-spin system exists in equilibrium with the two five-coordinate, high-spin systems and is evident even at neutral pH. This signal increases with pH and represents a thiolate/strong field ligation, based on the g values that are identical to those of WT CooA (Table III). The identity of the strong-field ligand trans to thiolate can either be H 2 O or hydroxide, which is observed in P-450 (45), Tyr 2 , or presumably a neutral nitrogen donor ligand provided by another residue in P2Y Fe III CooA. Based on these observations, we believe that there are a number of different liganding forms that exist in equilibrium in P2Y Fe III CooA. While the specific ligands remain unresolved in many of these forms, the results strongly support the notion of extreme flexibility in the heme:protein contacts, as suggested by the analysis of the crystal structure (21).
The spectral data of P2Y Fe III CooA suggested that a portion of the material exists as a five-coordinate species, and therefore  might be able to coordinate exogenous ligands, which should provide some insight into that open coordination position. Upon addition of exogenous imidazole to ϳ1000-fold molar excess, there were substantial spectral changes, with an isosbestic point at ϳ410 nm, from a signature of a five-coordinate, highspin heme to that of a six-coordinate, low-spin heme, indicative of imidazole binding to the heme of P2Y Fe III CooA (Fig. 5). However, the complete conversion of P2Y Fe III CooA to the imidazole-adduct was not observed even at high concentrations (ϳ2000-fold molar excess) as judged by the remaining fraction of the five-coordinate, high-spin form. Interestingly, P2Y Fe III CooA failed to coordinate either cyanide or azide even under large molar excess of ligand (1000-fold) and after long (1 h) incubation times (data not shown). Not surprisingly, the P2Y Fe III CooA was inactive in the FP assay in the presence or absence of imidazole (data not shown). This differential binding result is interesting in light of the fact that Fe III P-450cam and Fe III FixL are high-spin, five-coordinate hemeproteins that contain cysteine and histidine ligands, respectively, and bind both cyanide and azide (42,43,45). These results suggest that charge repulsion by an anionic environment might be the mechanism preventing cyanide and azide binding. However, because we do not know which side of the heme is bound by imidazole in these experiments, it is premature to speculate further on this observation. To address the concern that the five-coordinate form of P2Y Fe III CooA might be an intermediate form that is in the process of losing its heme, stability experiments were performed and the heme of P2Y showed no detectable loss from the Fe II , Fe III , or Fe II -CO forms (data not shown).
Conclusions-The major conclusions from this work are the following (i) Pro 2 has a non-critical role in the response of CooA to CO, as an unknown adventitious ligand can apparently serve as a reasonable substitute. (ii) There is great flexibility in the ligation state of the heme, as revealed by the complex properties of P2Y Fe III CooA, where at least three forms are detected in EPR analysis. (iii) Because of the perturbation of the P2Y Fe III CooA in the Pro 2 region variants, it is highly likely that Pro 2 serves as the normal ligand in WT Fe III CooA. (iv) While other residues in the vicinity of the N terminus appear to be important for function, such as Phe 5 and perhaps Asn 6 , it is our hypothesis that they are involved in creating a stable hemecontaining protein rather than in the actual response to CO; indeed, stabilizing the heme might be a major role of the N terminus of CooA. The results are consistent with the hypotheses proposed with the recent x-ray crystal structure of WT Fe II CooA (21), whereby the positioning of the heme upon CO binding is critical for a CO response.
As argued in the paper presenting the crystal structure of WT Fe II CooA, it appears highly likely that activation in response to CO involves a movement of the two CooA monomers with respect to each other through a repositioning of the two long helices (termed the "C" helices by analogy with CRP). This repositioning is presumably stimulated in some way by repositioning of the heme of CooA after CO binding. At present, it is unclear whether CO binding displaces His 77 or Pro 2 , but the moderate impairment of function of the CO response of P2Y CooA seen with FP can be rationalized as follows. If His 77 is normally displaced, then the adventitious ligand in the Pro 2 region variants must be slightly defective in positioning the heme to "signal" the C helix. On the other hand, if Pro 2 is normally displaced, then either the adventitious ligand in the Pro 2 region variants interferes with the CO-bound heme to prevent proper positioning or the released Pro 2 region itself normally affects the C helix and the Pro 2 region variants are altered in their ability to perform this interaction.