Dissociation and Recombination between Ligands and Heme in a CO-sensing Transcriptional Activator CooA

CooA from Rhodospirillum rubrum is a transcriptional activator in which a heme prosthetic group acts as a CO sensor and regulates the activity of the protein. In this study, the electronic relaxation of the heme, and the concurrent recombination between ligands and the heme at approximately 280 K were examined in an effort to understand the environment around the heme and the dynamics of the ligands. Upon photoexcitation of the reduced CooA at 400 nm, electronic relaxation of the heme occurred with time constants of 0.8 and 1.7 ps. The ligand rebinding was substantially completed with a time constant of 6.5 ps, followed by a slow relaxation process with a time constant of 173 ps. In the case of CO-bound CooA, relaxation of the excited heme occurred with two time constants, 1.1 and 2.4 ps, which were largely similar to those with reduced CooA. The subsequent CO recombination process was remarkably fast compared with that of other CO-bound heme proteins. It was well described as a biphasic geminate recombination process with time constants of 78 ps (60%) and 386 ps (30%). About 10% of the excited heme remained unligated at 1.9 ns. The dynamics of rebinding of CO thus will help us to understand how the physiologically relevant diatomic molecule approaches the heme binding site in CooA with picosecond resolution.

CooA, a transcriptional activator from Rhodospirillum rubrum, is a heme protein that acts as a CO sensor in vivo by binding CO (1)(2)(3)(4). CooA is the first example of a transcriptional regulator containing heme as a prosthetic group (1). Only CObound CooA activates transcription of the genes for the key CO-oxidizing enzymes (2)(3)(4)(5)(6). Although the ferrous heme in CooA is in a six-coordinate form (3,4), one of the heme axial ligands is replaced by exogenous CO upon the binding of CO (3,5,7), which triggers the conformational change in CooA required for specific binding to the target DNA (2,3,5,6,8).
Though CO has been widely used as a probe to study the biochemical and biophysical properties of heme proteins, it has been thought to have no physiological role. CooA is the first example of a heme protein in which CO has a physiological function. Analysis of the dynamics of binding and escape of the ligand in heme proteins provides information on the intrinsic reactivity of the site for heme iron biding with the ligand, and how the reactivity and the pathway of the ligand are controlled by the protein. Observation of the motion of ligands such as O 2 , NO, and CO within heme proteins is facilitated by the simple photodissociation of diatom-heme protein complexes (9 -11). The dynamics of geminate rebinding, escaping, and bimolecular rebinding of ligands can be studied by various spectroscopic methods over a wide time range.
Flash photolysis studies on CO-bound CooA will provide some useful information on the mechanisms of CO sensing by the heme and information on the regulation of CooA activity by CO. Measurement of transient absorption in the Soret band region on a tens of nanosecond or longer time scale has recently been carried out in studies on CooA, the results of which suggested a relatively high yield of geminate recombination (8). In this study, we have revealed the details of the geminate recombination dynamics upon photodissociation of CO-bound CooA with subpicosecond resolution. Accurate measurements of the transient signals of CooA in both the presence and absence of CO has allowed us to reveal the multiexponential recombination dynamics of the ligands and to compare the unique features of CooA with those of other heme proteins.

EXPERIMENTAL PROCEDURES
Sample-The expression and preparation of the recombinant CooA were performed as described previously (1,2). In brief, the recombinant CooA was expressed in Escherichia coli JM109 and purified by consecutive column chromatography steps, a Q-Sepharose ion-exchange column (Amersham Pharmacia Biotech) and a chelating-Sepharose column (Amersham Pharmacia Biotech, HR10/10). After removing salts by dialysis against an appropriate buffer, the sample was used for spectroscopic measurements. Reduced CooA was prepared by adding an excess amount of freshly prepared dithionite solution under an argon atmosphere to the protein solution. CO-bound CooA was prepared by introducing an excess amount of gaseous CO into the reduced sample.
Transient Absorption System-The transient absorbance changes were determined by measuring the transmission of a probe pulse through the samples under pump-on and pump-off conditions. The pump pulses centered at 400 nm were generated by the second harmonic generation of a femtosecond Ti/sapphire oscillator-amplifier system (800 nm, 1 kHz). The pump pulses were sent through computercontrolled delay optics to the sample. Femtosecond white light continuum pulses generated by the amplified 800-nm pulses were used to probe the absorbance changes in the 420 -510-nm region (Fig. 1).
Two different measurement schemes were used. In the first scheme (transient spectrum method), a dual photodiode array detector attached to the exit port of a monochromator with a fully opened output slit was used. One array monitored the probe pulse spectrum after transmission through the sample irradiated by the pump beam, and the other monitored the fluctuation of the probe pulse spectrum without transmission * This work was supported by Grants-in-aid for Scientific Research 40087507 (to K. Y.), 11554028 (to K. Y.), 12680631 (to S. A.), and 12019222 (to S. A.) from the Ministry of Education, Science, Sports, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. through the sample. The pump pulses were chopped at 5 Hz, and the spectra were synchronously recorded at 10 Hz.
In the second scheme (single wavelength method), the dual photodiode array was replaced by two photodiode detectors. A mechanical chopper at half the repetition rate of the pulses (500 Hz) modulated the pump beam with a fixed phase to the pulse train. One photodiode detector monitored the energy of the white continuum pulse at 440 (Ϯ5) nm after transmission through the sample irradiated by the pump beam, and the other monitored the energy fluctuation of the same portion of the white continuum pulse without transmission through the sample. The probe and reference signals were carefully balanced and integrated by gated integrators for each laser shot. The difference signal between the two outputs of the gated integrators was sent to a digital lock-in amplifier. The amplitude of the modulation, which is synchronous to the pump pulse train was used to calculate the transient absorbance changes. The sample was continuously irradiated by the laser beams for at most 2 s to avoid build-up of photoproducts. A dark period of at least 1 s was introduced before measuring the transient absorbance change at each time delay.
The instrument-response function and the time zero at each probe wavelength were estimated by optical Kerr effect (OKE) cross correlation between pump and white continuum pulses. Neat o-dichlorobenzene in the sample cell was used as the OKE medium. The full width at half-maximum of the instrument-response function was 0.55-0.65 ps in the 420 -510-nm region. The delay time where the OKE signal shows maximal intensity is defined to be Ϫ0.3 ps at each wavelength. With this definition, more than 90% of the instrumental-response limited transient absorption signal is developed at 0.0 ps. Curve-fitting analysis was applied to the data in the time region later than 0.0 ps. The angle between the pump and probe polarizations was set at the magic angle (54.7°) in all transient absorption measurements.
The sample was introduced under nitrogen gas atmosphere into a gas-tight cylindrical cuvette with a path length of 2.5 mm. The CooA concentration was typically ϳ25 M on the basis of heme content. This cuvette was cooled (ϳ280 K) and rotated at ϳ3 Hz during the measurements to ensure that each pair of pump-probe pulses would irradiate a new portion of the sample. The absorption spectrum before and after each measurement did not show any change. The energy of the pump pulse was 1.5 J in the transient spectrum method and 1 J in the single wavelength method. There was no essential change in the kinetics at 440 nm when the pump power was decreased to 0.4 J (data not shown). The probe pulse energy was Ͻ80 nJ. The diameters of the pump and probe beams at the sample positions were 0.3 and 0.2 mm, respectively. We estimated that at most 40% of the heme was excited by a 1 J pump pulse.

RESULTS
The electronic absorption spectra of the reduced and CObound CooA, together with the probe wavelength range for flash photolysis, are shown in Fig. 1. Transient absorption spectra at representative delay times are shown in Fig. 2, a and b. Similar changes in the transient spectral shapes occurred in the case of both the reduced and CO-bound CooA within 5 ps upon excitation. After 5 ps, the spectral shape remained almost constant but the amplitude decayed. The decay seemed to be much faster in the case of reduced CooA than that in the case of CO-bound CooA. Transient absorption kinetics at representative wavelengths are shown in Fig. 3. The signal at 427 nm was dominated by photobleaching of the absorption of the sixcoordinate heme in the ground state. Between ϳ430 and ϳ510 nm, the five-coordinate heme showed an absorption band (12). The appearance of a positive absorbance change above 480 nm was instrument-response limited (e.g. at 488 nm in Fig. 3, a  and b; Ref. 9). There were two phases in the transient absorbance changes from 0.2-5.6 ps. The first phase was the growth of the positive signal at around 440 nm and the negative signal at around 425 nm (Fig. 2, a and b). Evidence of this phase was the delayed peak of the transient signals at 427 and 443 nm compared with that at 488 nm (Fig. 3, a and b). The second phase was observed from 1.2-5.6 ps in which the wavelength that gave the maximum positive absorbance change shifted from 441 nm to Յ438 nm (Fig. 2, a and b). The difference between the reduced and CO-bound CooA became clear at 23 ps and later times, where relatively larger transient absorbance changes remained in the case of CO-bound CooA.
These complicated spectral changes were analyzed with the help of global curve fitting analysis. The kinetics over the whole wavelength region within 45 ps were simultaneously fit by the sum of triple exponential functions with common time constants. This yielded time constants of 0.8, 1.7, and 6.5 ps for the reduced CooA, and 1.1, 2.4, and 78 ps for the CO-bound CooA. For each time constant, the decay-associated spectrum (DAS), 1 which is a plot of the exponential amplitude against wavelength, is given in Fig. 4.
The single wavelength method was employed to study the long-lived transients in a more extended time region up to 1.9 ns (Fig. 5). The kinetic trace of the reduced CooA was fit by a sum of two exponential functions with time constants of 6.5 ps, 173 ps, and an offset (a very long time constant) (Fig. 5a). The kinetic trace of the CO-bound CooA was fit by a sum of three exponential functions with time constants of 4, 78, and 386 ps and an offset (Fig. 5b). The spectral shapes of the transient absorption spectra of the CO-bound CooA at time delays of 23, 110, and 630 ps were compared in Fig. 2c. All of the above parameters obtained by the curve fit are summarized in Table I. DISCUSSION Distinction between Relaxation of Heme and Ligand Rebinding-It has been reported that the dissociation of diatomic ligands from heme upon photoexcitation of cytochrome c oxidase (13), hemoglobin, or myoglobin (14) occurs on a ϳ0.1-ps time scale with a quantum yield close to unity. It is assumed that dissociation of ligands in CooA occurs within our instrumental response time with a quantum yield of unity. Rubtsov et al. 2 have studied the transient absorption changes in the mid-IR region after photolysis, and their study revealed an instrument-response limited bleaching of the vibrational band of the bound CO. This also supports our assumption. All of the time-resolved processes in this work are thus associated with relaxation of heme and/or rebinding of ligands.
For both reduced and CO-bound CooA, we have resolved three exponentially decaying components within 45 ps ( Table  I). The fastest two of the three processes seem to be common to both the reduced and CO-bound CooA. This is supported by the following key points of the data. Firstly, the shortest two time constants of the reduced CooA, 0.8 and 1.7 ps, are similar to those of the CO-bound CooA, 1.1 and 2.4 ps, respectively. Secondly, the shapes of the DASs of the 0.8 and 1.7 ps components are also similar to those of the 1.1 and 2.4 ps components, respectively (Fig. 4).
The dynamics of photodissociation and recombination between diatomic ligands and heme proteins have been extensively investigated for many heme compounds, such as myoglobin, hemoglobin, and cytochrome c oxidase (9 -12, 16, 17). It seems to be most informative to compare the observed kinetics in this study with those obtained by Petrich et al. (12), because their experimental conditions were similar to ours. They measured the transient absorption spectra in the Soret band region for many diatom-heme protein combinations (hemoglobin, myoglobin, and protoheme with CO, NO, O 2 or unligated) using 307-or 580-nm excitation pulses. They found 0.3-ps and 2.5-3.2-ps processes in the case of all heme protein complexes, including the unligated ones (12). The spectral features of the 0.8 -1.1-ps and 1.7-2.4-ps components observed here are in good agreement with those of the 0.3-ps and 2.5-3.2-ps components observed previously. The disagreement of the shortest time constants may result from our comparatively long instrument-response function. It thus follows that the 0.8 -1.1-ps components and 1.7-2.4-ps components are attributable to relaxation of electronically excited five-coordinate heme. The 1.1 and 2.4 ps DASs of the CO-bound CooA are, however, blue-shifted by about 6 nm from the corresponding ones of the reduced CooA. This blue-shift may reflect the difference in the energy level of the excited heme after the dissociation of the ligand. This idea is supported by the relatively strong positive absorbance changes around 480 nm in the case of reduced CooA as compared with that in the case of CO-bound CooA (Fig. 2, a  and b), because only highly excited five-coordinate heme shows absorption around 480 nm (12). It has been recently shown that a proximal ligand (His-77) is common to the reduced and CObound CooA (3,25). It is thus possible that the energy dissipation of the excited heme through CO dissociation in the CObound CooA is more effective than that of the distal ligand in the reduced CooA.
The slowest-decaying components in Fig. 4 are very different in terms of time constants comparing the reduced CooA (6.5 ps) and CO-bound CooA (78 ps). It should also be noted that the DASs of the 6.5 and 78 ps components are similar to the difference spectrum between five-and six-coordinate heme (12). These components are thus attributable to recombination of the ligand. Given the common proximal ligand for reduced and CO-bound CooA (3,25), the great difference in the time constant probably results from the different natures of the distal ligands. It should be noted that heme itself undergoes a relatively slow relaxation process, like vibrational population relaxation, which affects the absorption spectrum in the Soret band region. The exponential decay time of vibrationally hot states of the electronically relaxed heme could be as slow as 6 -16 ps (18,19). The DASs for the 6.5-ps time constant may be affected by vibrational cooling processes.
There is a minor component with a time constant of 173 ps in the kinetics of the reduced CooA ( Fig. 5a and Table I). In carbomonoxymyoglobin, there is a similarly slow relaxation process upon photodissociation of CO (20,21). The ligation state of the proximal histidine changes with a time constant of 100 ps (probed by (Fe-His) Raman band shift, Ref. 21), and the restructuring of the asymmetric protein environment requires about 300 ps (probed by circular dichroism at 355 nm) (20). An even slower relaxation on the time scale of nano-to microseconds has been detected by probing weak iron-porphyrin charge transfer transitions (22), though the major change is complete on the subnanosecond time scale. These are intrinsic to myoglobin, because such slow processes are not seen in the case of isolated heme in protein-free solution (21). It is thus likely that the 173-ps component observed here is indicative of a relaxation of the CooA polypeptide upon photoexcitation of the heme, which affects the heme absorption spectrum.
Ligand Rebinding Dynamics-The rebinding of the ligand to the reduced CooA is very fast (6.5 ps, Figs. 4a and 5a). This recombination time seems to be comparable with that (7-20 ps) of the cytochrome a center in reduced cytochrome c oxidase (13,17), where the heme is six-coordinate with the axial ligands being histidine residues (23). The 6.5-ps rebinding indicates that both the proximal and distal ligands of the ferrous heme in the reduced CooA are endogenous ligands of high inertia, not exogenous small molecules or ions, for which recombination times of at least tens of picoseconds are expected. These results are consistent with the proposed coordination structure of the heme in CooA (25). 3 The fastest component of 4 ps in the 440-nm kinetics of CO-bound CooA (Fig. 5(bЈ)) seems to be associated with the relaxation process of excited heme (Fig. 4 and Table I). There seems to be no great change in the transient spectral shape of CO-bound CooA from 23-630 ps (Fig. 2c). These findings lead us to conclude that the recombination dynamics of CO in CooA is multiexponential: 78 ps (60%) and 386 ps (30%). About 10% of the dissociated CO remains free at 1.9 ns. This overall rebinding dynamics is largely consistent with that obtained by mid-IR probing of the CO vibrational band (CO-bound CooA (0.5 mM) in D 2 O) 2 or by resonance Raman probing (25). A small but clear difference between our findings and the results re-  (b and b). Two-and three-exponential curves (solid lines) were sufficient and necessary to fit the data for reduced CooA and CO-bound CooA, respectively. Optimum time constants are shown in Table I. The insets (aЈ and bЈ) show the same kinetics on short time scales of up to 150 ps. ported by others is that more long lived five-coordinate heme was detected in the present study (25). 2 We have studied the dependence of the decay dynamics of the five-coordinate heme on solvents (H 2 O or D 2 O) and on excitation intensities (0.4 -1.0 J/pulse). Neither of these two experimental factors had effects strong enough to explain the discrepancy (data not shown). We have also observed that long continuous exposure (e.g. for a few minutes) of the CO-bound CooA sample to the excitation light changes the apparent decay dynamics of the five-coordinate heme (data not shown), and this is probably because of the accumulation of CO-free CooA in the sample cell.
The rate and yield of the geminate recombination of the CO in CooA (Table I) are strikingly faster and higher than those of CO in wild-type myoglobin (ϳ180 ns, 4%, Ref. 26). These parameters of CooA are almost comparable with those of NO in wild-type myoglobin, where the geminate rebinding time constants and amplitudes are 18.9 ps (41%), 126 ps (49%), and an offset (24). Regarding the much higher reactivity between NO and heme than that between CO and heme (15), there must be a special mechanism responsible for the fast rebinding of CO in CooA. The geminate recombination between a diatom ligand and heme on a picosecond time scale (10 -500 ps) has been best studied in the case of NO rebinding in myoglobin (10,15). Rapid NO recombination can be achieved by specific mutations in two ways: (i) removing steric restrictions directly adjacent to the iron atom and (ii) inhibiting ligand movement away from the iron atom by placing diffusional barriers. Recent results obtained by resonance Raman spectroscopic study and EXAFS analysis by this group have suggested that there are no distal amino acid residues that interact directly with the CO bound to the heme (8). 3 Thus, there seem to be no steric restrictions affecting the binding of CO to heme at its binding site. In addition, we propose that the distal side of the heme pocket in CooA provides a much smaller free volume than that of myoglobin, which should cause a decrease of the CO recombination time by three orders of magnitude as compared with that in the case of myoglobin (26). The axial ligand trans to His-77 in the reduced CooA is replaced by CO upon CO binding (3,25). 3 It is thus probable that the released ligand acts as a diffusional barrier against the dissociated CO.
The rebinding dynamics of CO in CooA is clearly nonexponential. There are at least three models by which one can explain the nonexponential recombination kinetics (24,15). (i) The multiple-site model invokes several intermediate sites for CO after the photodissociation, which differ in terms of the potential barriers to recombination with heme. The microscopic origin probably arises from the complicated structure of the protein environment into which CO is ejected by photolysis. (ii) The relaxation model incorporates a potential barrier that evolves as the heme and protein structure relax following photodissociation, which gives a time-dependent rate constant of CO-rebinding. (iii) In the inhomogeneous model, one assumes a distribution of rebinding rates arising from a static inhomogeneous energy barrier distribution.
The NO rebinding rate in the case of myoglobin on the picosecond time scale is primarily determined by the details of the distal heme pocket structure (24,15). The multiple-site model in which there are at least two intermediate states for the dissociated CO seems to be most plausible for the multiexponential rebinding dynamics of CO in CooA. The validity of this model, however, depends on the detailed structure of the distal heme pocket, which is not yet clarified. The applicability of the other two models is also considered as follows. In the case of reduced CooA, the 173-ps process seems to reflect heme protein relaxation. If this relaxation also occurs in the case of CO-bound CooA, it may change the reactivity between heme and CO or the structure of the distal heme pocket. The small change in the transient spectral shape at around 423 nm between 23 ps and 630 ps (Fig. 2c) may be because of such heme protein relaxation. The 78 and 386 ps components in the recombination kinetics (Fig. 5b and Table I) could then represent recombination during and after the relaxation process with a ϳ173-ps time constant, respectively (as in the relaxation model). Rubtsov et al. 2 have also found a similarly slow relaxation process in CO-bound CooA in D 2 O by probing CO vibrational bands, which also supports this model. It should be noted that the real rebinding dynamics could be caused by a mechanism consistent with the multiple-site and relaxation models combined. Conformers of CooA may show different recombination rates (as in the inhomogeneous model). However, there has been no report on CooA conformers with a population ratio comparable with the amplitude ratio of the 78 and 386 ps components of the CO recombination dynamics (Table I).
In summary, this study has shown that CO geminate recombination proceeds with a high yield (90%) on a picosecond time scale. These features may be relevant to the physiological function of CooA which should hold CO tightly while it is activated to regulate transcription of the target DNA. The fast CO rebinding is possibly because of the crowded heme pocket in which the ligand released from the ferrous heme acts as a barrier against CO diffusion. 3 The CO rebinding dynamics is multiexponential. This is explained by the presence of multiple intermediate states of the dissociated CO in the distal heme pocket and/or the time dependence of the reactivity of CO with the heme. Reduced 0-45 420-510 0.8 ps, 1.7 ps, 6.5 ps (see Fig. 4) Reduced 0-1900 440 (Ϯ5) 6.5 ps (93%), 173 ps (6%), offset (1%) CO-bound 0-45 420-510 1.1 ps, 2.4 ps, 78 ps a (see Fig. 4) CO-bound 0-1900 440 (Ϯ5) 4 ps b , 78 ps (60%), 386 ps (30%), offset (10%) a The 78-ps time constant was a fixed parameter in the global curve analysis, since it was more accurately determined by the kinetic trace probed at 440 nm.
b The amplitude of the 4-ps component is 15% of the total amplitude in the 440-nm kinetics (Figure 5a), but it is excluded from the percentage calculation here.