Ligand Dynamics in an Electron Transfer Protein

Substitution of the heme coordination residue Met-80 of the electron transport protein yeast iso-1-cytochrome c allows external ligands like CO to bind and thus increase the effective redox potential. This mutation, in principle, turns the protein into a quasi-native photoactivable electron donor. We have studied the kinetic and spectral characteristics of geminate recombination of heme and CO in a series of single M80X (X = Ala, Ser, Asp, Arg) mutants, using femtosecond transient absorption spectroscopy. In these proteins, all geminate recombination occurs on the picosecond and early nanosecond time scale, in a multiphasic manner, in which heme relaxation takes place on the same time scale. The extent of geminate recombination varies from >99% (Ala, Ser) to ∼70% (Arg), the latter value being in principle low enough for electron injection studies. The rates and extent of the CO geminate recombination phases are much higher than in functional ligand-binding proteins like myoglobin, presumably reflecting the rigid and hydrophobic properties of the heme environment, which are optimized for electron transfer. Thus, the dynamics of CO recombination in cytochrome c are a tool for studying the heme pocket, in a similar way as NO in myoglobin. We discuss the differences in the CO kinetics between the mutants in terms of the properties of the heme environment and strategies to enhance the CO escape yield. Experiments on double mutants in which Phe-82 is replaced by Asp or Gly as well as the M80D substitution indicate that such steric changes substantially increase the motional freedom-dissociated CO.

Electron transfer reactions between and within proteins bearing redox-active constituents are abundant in the cell. Most physiological electron transfer reactions take place in the sub-millisecond time scale (1). Experimentally, it is not possible to study the dynamics of such reactions by mechanically mixing the reactants (e.g. stopped-flow spectroscopy). Such dynamics may, however, be studied by the use of light pulses. This approach is obvious for systems where light is a natural trigger, such as photosynthetic systems, but in some cases this approach can also be used with other pigmented proteins. The method we intend to develop relies on the fact that the gaseous ligand carbon monoxide (CO) 4 has a high affinity for ferrous heme proteins but does not bind to the ferric heme counterpart. Therefore, the effect of CO is to increase the apparent redox potential of any heme protein that has a vacant sixth coordination position and is thus able to bind this ligand. The CO molecule can be photo-cleaved from heme with a quantum yield close to unity (i.e. one CO molecule photodissociated for each photon absorbed by the heme group). Thus, a light pulse will dissociate the CO from the heme group, leaving this as a 5-coordinate high spin ferrous species. This has the effect of rapidly switching the redox potential of the protein that stabilizes the ferrous form (the CO adduct) into a good electron donor (5-coordinate ferrous heme). Provided CO does not rebind rapidly (i.e. Ͼ10 Ϫ5 s), photo-induced electron transfer can thus take place. These features have been exploited to investigate intraprotein electron transfer from photodissociated 5-coordinated heme a 3 in the active site of native cytochrome c oxidase, where CO acts as an inhibitor (2), and recently used to determine a rate as high as ϳ10 9 s Ϫ1 (3).
This study focuses on cytochrome c (cyt. c), a ubiquitous, small, soluble electron transfer protein carrying one heme cofactor. In native cyt. c, the heme iron is coordinated by two internal axial ligands, His-18 and Met-80, and cannot bind CO. However in methionine-modified forms (4,5), like carboxymethylated cytochrome c (cm-cyt. c), chemical modification of Met-80 precludes formation of the heme iron-Met-80 bond and allows CO to bind. In a sizeable fraction of cm-cyt. c (12-26% depending on the modification protocol (6)), CO rebinding is slow enough to allow flash-induced oxidation of the heme. This technique has been successfully used to study electron transfer from cm-cyt. c to cytochrome c oxidase (4) and to plastocyanin (7). These studies have provided proof of principle for the method, but the use of cm-cyt. c, although useful, is not ideal. This is primarily because the carboxymethylation reaction is not specific and can also lead to chemical modification of important "docking" residues on the surface of the protein.
With the objective of developing a method for studying interprotein electron transfer that employs proteins in "quasi-native" states, we have made a series of Met-80 mutants of yeast iso-1-cytochrome c, namely M80A, M80S, M80T, M80D, and M80E (8). Such mutants bind CO (8,9). Fig. 1 shows the solution structure of yeast cyano-M80A ferri-cyt. c (10) with labeled and some important amino acid residues that line the distal heme pocket where CO would normally bind to the ferrous heme iron (see under "Discussion"). The replacement residues for the native Met-80 were selected as those that were not expected to coordinate to the heme iron, especially in the ferrous redox state. However, for electron transfer experiments, these mutants have one drawback, namely that the apparent quantum yield () for photon-induced CO escape from the protein is very small (8). This is because of rapid and efficient "geminate" recombination of CO and the heme. We have recently shown that in cm-cyt. c such recombination is multiphasic and predominantly takes place on the picosecond time scale (6), i.e. much faster than inter-protein electron transfer (typically ϳ10 5 s Ϫ1 ).
In contrast to NO recombination, extensive geminate recombination of CO and heme is seldom observed in native natural ligand-binding proteins like myoglobin (Mb), and in particular not on the sub-nanosecond time scale (11). The fact that such recombination, by contrast, dominates in cyt. c can be understood in general terms by the absence of a "natural" ligand exchange pathway and by a different regime of protein dynamics in this electron transfer protein. For efficient low driving force electron transfer to occur, the reorganization energy of the environment of the redox partner (in the case of cyt. c the heme) should be small. Indeed, in cyt. c the heme group is surrounded by a hydrophobic and relatively rigid pocket (12,13), and the protein contribution to the reorganization energy of cyt. c is thought to be weak (14,15). Structural flexibility is not advantageous for that function. In contrast, proteins where ligand transfer is essential, structural flexibility is required to create temporary passages through which ligands may diffuse. In a general sense, the constructed CO-binding cyt. c mutants offer a unique chance to study these dynamics-function relations by comparison with functional ligand-binding proteins like Mb.
Studying the picosecond/nanosecond recombination dynamics of CO and heme in cyt. c proteins incorporating an M80X mutation can give insight into the heme environment. In a similar way, in wild type and mutant Mb, multiphasic picosecond heme-NO recombination kinetics have proven a sensitive tool for assessing the dynamic, steric, and electrostatic role of the heme environment (16 -22). As mentioned above, in Mb, CO geminate recombination is usually less extensive and takes place on a much longer time scale (23)(24)(25) and in the native protein in a monophasic way ( ϭ ϳ150 ns). To gain insight into the dynamic properties of the hydrophobic heme-binding core of cyt. c and in particular to investigate ways to optimize the escape probability of photodissociated CO by modifying the heme environment, we have measured CO recombination kinetics in a number of cyt. c mutants. Using spectrally resolved femtosecond spectroscopy, we also determined the heme spectra associated with different decay phases, which are indicators of the interaction of the heme with its environment, and in particular with the dissociated nearby ligand.
This initial study concerns a limited number of mutants. By substituting Met-80 the idea was to investigate the effects of introducing both steric and hydrophobic constraints. In addition to proteins bearing Ala, Ser, and Asp at this position, which were previously shown to have a low millisecond CO escape efficiency ( Ͻ 0.05) (8), we investigate the effect of the more bulky, charged residue Arg. Furthermore, starting from the M80D mutation, the effect of the additional substitution of the close-lying Phe-82 by a smaller residue (Asp or Gly) is explored.

MATERIALS AND METHODS
Mutagenesis-Site-directed mutagenesis, protein expression in Escherichia coli, and purification of yeast iso-1-cytochrome c mutants bearing single (M80A, M80S, M80D, or M80R) mutations in the heme environment were performed as described previously (8). The double mutations (M80D/F82D or M80D/ F82G) were introduced using a single-step PCR approach (8), and in addition using the M80D mutated pBPCYC1(wt)/3 plasmid construct as template DNA. In these proteins, the Cys-102 residue is also changed to a Thr to prevent dimerization through formation of disulfide bridges.
Sample Preparation for Ultrafast Spectroscopy-The samples were prepared to a heme concentration of ϳ50 M in a gas-tight optical cell with an optical path length of 1 mm. Unless indicated otherwise, the buffer was 20 mM Hepes, pH 8.0. For the deoxy form, the samples were de-gassed and reduced with 1 mM sodium dithionite. For the carboxy form, the deoxy samples were equilibrated with 1 atm (1 atm ϭ 101.3 kPa) CO. For the nitrosyl form, de-gassed samples were reduced with 10 mM sodium ascorbate and equilibrated with 0.01 atm NO. For the oxy form, 10 mM sodium ascorbate was added to an air-equilibrated sample. Clearly shown is the heme group with His-18 (proximal) and CN Ϫ (distal) placed in the 5th and 6th coordination sites of the central iron atom (bonds to iron not shown). The amino acids that are contained on the loop Gly-77 to Ile-85 (see "Discussion") are all shown, and the residues considered most important are labeled, including Gly-77, Lys-79, Ala-80, Phe-82, and Ile-85. Tyr-67 is also present, a residue that lies within the distal pocket and has been shown to modulate ligand binding to the heme iron as well as protein conformational changes. Swiss Pdb Viewer was used to create the figure (57).
Ultrafast Spectroscopy-Multicolor femtosecond absorption spectroscopy (16) was performed with a 30-fs pump pulse centered at 565 nm and a Ͻ30-fs white light continuum probe pulse, at a repetition rate of 30 Hz. Full spectra of the test and reference beams were recorded using a combination of a polychromator and a CCD camera. All experiments were carried out at 21°C. The sample was continuously moved perpendicular to the beams to ensure sample renewal between shots.
Basic data matrix manipulations and presentation were performed using Matlab (The Mathworks, South Natick, MA). The absorbance changes were treated using the SPLYMOD algorithm (26), with a Matlab interface (27).
Determination of Quantum Yield of CO Escape-The apparent quantum yield of dissociated CO on the millisecond time scale was also measured by the "pulsed" method developed by Brunori and co-workers (28) and as described previously (8). The method determines the value of for a given protein relative to a standard, generally Mb, for which the value is assigned as ϭ 1 (i.e. unity quantum yield).

Bimolecular Binding of Gaseous Ligands to the Ferrous Forms of M80A (CO, O 2 , and NO) and M80D (CO and NO) by Flow
Mixing-Protein concentrations were typically 3-5 M final after flow mixing. The mutant proteins were reduced prior to mixing by additions of sodium dithionite. For the CO and NO experiments, an excess of sodium dithionite was added to remove any dissolved O 2 from solution assumed to be present at a concentration of ϳ280 M at 20°C (29), therefore eliminating competition between gaseous ligands for the ferrous iron (see below for O 2 ). The temperature was maintained at 20°C in all flow experiments. The observed rates (k obs , s Ϫ1 ) of ligand binding were fitted to single exponential fits and plotted, for each ligand, as a function of their concentration (M). The data points for each ligand for M80A were then fitted with a rectangular hyperbola (see "Discussion"). The data points for each ligand for M80D were fitted to straight lines (see "Discussion").
Preparation and Calibration of CO Solutions for Bimolecular Binding Studies-Sodium phosphate buffer (100 mM, pH 7) was placed into a glass tonometer, and the solution was degassed thoroughly with vigorous shaking. CO was passed into the tonometer with mild shaking until the gas had completely equilibrated with the solution, and a slightly positive pressure was obtained. The CO solution was transferred from the tonometer to an Agla glass syringe connected to a micrometer screw gauge. A Mb sample of an absolute known concentration (ϳ10 M) was prepared using an extinction coefficient of 121,000 M Ϫ1 cm Ϫ1 at 435 nm for the deoxy form (30). The Mb was reduced with a slight excess of sodium dithionite and placed in an ϳ3-ml cuvette that had been previously weighed in addition to a suba-seal. The sample was sealed, and no gas phase was present. The cuvette and sample were then re-weighed assuming 1 g ϭ 1 ml. Spectra (375-650 nm) were recorded of the Mb sample titrated with 2-l additions of the CO solution until saturation of the protein with CO had been achieved (i.e. no further changes in the absorbance on additions of CO). The suba-seal was then removed, and ϳ5 ml of pure CO gas was bubbled through the sample so that complete saturation of the solution was achieved. The final spectrum was then recorded. Each addition of 2 l of CO caused an increase in absorbance at 424 nm, and so the absorbance change at 424 nm minus at 650 nm was plotted against the volume of CO added. The first point on this plot where subsequent additions of CO produced no further change in absorbance was taken as the 1:1 binding of heme with CO (assuming no free heme present), and so with the concentration of Mb exactly known as well as the volume of the solution, the concentration of CO in the sample was deduced and hence that of the stock CO solution. For the stopped-flow reactions of M80A and CO, the stock CO solution was prepared in sodium phosphate buffer (100 mM, pH 7). For the stopped-flow reactions of M80D and CO, the stock CO solution was prepared in sodium acetate buffer (100 mM, pH 4.8), the CO concentration being determined by titration with Mb at pH 7.
Preparation of NO Solutions for Bimolecular Binding Studies-The NO solutions for direct use in the stopped-flow apparatus were prepared using stock solutions of Proli NONOate (Axxora) stored in 25 mM NaOH. The concentration of a stock Proli NONOate solution was determined using the extinction coefficient of 8,600 M Ϫ1 cm Ϫ1 at 248 nm (commonly 12-14 mM). For each Proli NONOate molecule, 2ϫ NO was released on lowering of the pH. A known amount of stock Proli NONOate solution was added to a glass syringe containing either sodium phosphate buffer (100 mM, pH 7) for M80A or sodium acetate buffer (100 mM pH 4.8) for M80D prior to mixing. The highest concentration of NO in solution prior to mixing was ϳ1 mM and was therefore 0.5 mM in NO after mixing with protein.
Both buffer solutions were de-gassed prior to addition of the stock Proli NONOate solution so as to prevent unwanted side reactions of NO and O 2 . NO solutions were subsequently diluted with de-gassed buffer solutions.
Preparation of Fe 2ϩ -M80A and O 2 Solutions for Bimolecular Binding Studies-The concentration of dissolved oxygen in sodium phosphate buffer (100 mM, pH 7) was assumed to be ϳ280 M at 20°C prior to dilution into de-gassed phosphate buffer and flow mixing. M80A was reduced with a minimum amount of sodium dithionite (10 -20 M total) so as to prevent the removal of the O 2 being introduced on flow mixing. M80A is very resistant to autooxidation by air, and so can readily form a stable ferrous iron/oxygen adduct. The rate of autooxidation of M80A has been determined as k ϭ ϳ1.8 days Ϫ1 (31). M80S can also form a stable ferrous iron/oxygen adduct with k ϭ ϳ0.7 days Ϫ1 . However, all other mutant proteins in this study autooxidize sufficiently quickly so that the oxygen adducts cannot be observed in the time frame of the stopped-flow apparatus. Fig. 2A shows the transient spectra in the Soret band region at selected delay times after CO dissociation from the M80A mutant. The general shape of the spectra is typical of a 5-coordinate minus 6-coordinate ferrous heme spectrum (32) and denotes a red-shift of the Soret band after photolysis. The initial spectrum decays in a multiphasic time course on the picosecond and early nanosecond time scale, reflecting recombination of CO with the heme. Along with this decay the shape of the spectra also changes, and in particular the induced absorption maximum shifts from 424 nm at t ϭ 1 ps to ϳ432 nm on the hundreds of picoseconds time scale. These features are qualitatively similar to those observed for cm-cyt. c (6), and presumably, they reflect relaxation of the heme configuration during the recombination process. The analysis, in terms of spectra associated with exponential decay processes (decay-associated spectra), is shown in Fig. 2B where these relaxation processes are highlighted. In particular, the fastest decay phase, 14 ps, is associated with a much smaller red-shift than the later decay phases, 85 ps and 1.2 ns. In contrast to cm-cyt. c, at t ϭ 4 ns, virtually all CO has recombined ( Figs. 2A and 3). This finding is consistent with the reported very low ( ϭ 0.004) quantum yield of CO escape from the protein (8) ( Table 1) and implies that the geminate heme-CO recombination, which is at the origin of this loss in quantum yield, is completely covered by the 4-ns time window of this study.

CO Rebinding in Met-80 Single Mutants-
The transient spectra of the M80S and M80D mutants are qualitatively similar to the M80A mutant (not shown), but the kinetics are somewhat different (Fig. 3). For M80S, the overall picosecond rebinding kinetics are somewhat faster, and in particular no phase having a value longer than 100 ps is discerned. For this mutant also virtually all CO rebinds, in agreement with the low ( Table 1). The kinetics for CO rebinding in M80D need to be described by three distinct exponential phases, as in M80A, but the time constants for all phases are significantly longer, and at t ϭ 4 ns, ϳ4% of the CO has not rebound, in good agreement with the reported ϭ 0.04 for this mutant (8) ( Table 1).
For the M80R mutant, both the transient spectra ( Fig. 2C) and the decay kinetics ( Fig. 3) stand apart from the Ala, Asp, and Ser mutants. The overall kinetics are relatively slow and can be described by two phases, the fastest of which has a time constant of 80 ps and no sub-50-ps phase. Importantly, the amplitude of the constant, non-rebinding phase (A 0 ), is also much higher, 0.25, which is more than twice that observed in our previous experiments on cm-cyt. c (6). In addition, the spectral evolution during the decay is far less pronounced (a moderate effect can be discerned from the decay-associated spectra of the 80-ps component; see Fig. 2D), and the shift feature is much more symmetric than in the other mutants, where the induced absorption band is relatively broad at all delay times. Although for these proteins a steady-state 5-coordinate spectrum cannot be generated for direct comparison (see below), this latter finding suggests that in the M80R the heme rapidly adopts a more relaxed deoxy-like configuration after CO dissociation. Indeed, the spectrum is qualitatively similar to the steady-state CO dissociation spectra for Mb (33) and the c-type heme-undecapeptide microperoxidase-11 (5).
CO Rebinding in Double Mutants-To explore whether one may influence the probability of photodissociated CO escaping into bulk by further modifying the protein environment of the heme, we have generated double mutants in which in addition to the Met-80 to Asp mutation Phe-82 was replaced by the less bulky residues Gly or Asp. Both double mutants displayed very similar kinetics (Table 1), which were substantially slower than

TABLE 1 Fitted components of CO rebinding
The decrease in the bleaching, at 414 nm, was fitted to a 2-or 3-exponential decay function (⌺ i A i exp(Ϫt/ i )) and a constant A 0 ( i are decay times 1/k i ). The sums of all the amplitudes are normalized to 1, and A 0 is the fractional amplitude of the spectrum that has not decayed at the end of 4 ns. The fraction of CO release was determined from the amplitude of the millisecond bimolecular CO rebinding phase (8). The slight differences between A 0 and are presumably due to the spectral changes of the heme during recombination. Here we choose to use the kinetics at 414 nm, as the shape of the bleaching part of the spectrum does not change much during the recombination phases. Somewhat higher values of A 0 are obtained from kinetics in the induced absorption lobe of the spectrum. for the M80D single mutant (Fig. 4), and also the constant (A 0 ) and non-rebinding phase was significantly higher. These findings indicate that the CO escape pathway indeed involves the close surroundings of Phe-82. Interestingly, the transient spectra are more symmetric for the double mutants (Fig. 4, inset), already within picoseconds after CO dissociation. This indicates that the free volume created by the second mutation allows reduction of the interaction of the dissociated CO with heme. Taken together with the spectral and kinetic properties of M80R mutant (for both the differences with respect to the M80D single mutant are more pronounced), these data strongly suggest a correlation between the transient spectra and the speed and yield of heme-CO rebinding. NO Rebinding-In many natural ligand-binding heme proteins, NO rebinds predominantly on the picosecond time scale and in a multiphasic way that is very sensitive to the structure of the heme pocket (11). We have measured the spectral evolution upon excitation of the M80A-NO complex (Fig. 5), and we found that NO rebinding occurs fully in a single exponential phase with a time constant of 7 ps, i.e. in a very similar way as NO rebinding to wild type horse heart cyt. c (34) (where NO can replace the internal Met-80 ligand). Our present results thus indicate that in cyt. c dissociated NO is not sensitive to changes in the heme environment, does not explore multiple configurations in the heme pocket, and is kept close to the heme.

Mutant
O 2 Rebinding-Excitation of the oxycomplex of the M80A mutant leads to dissociation of O 2 from the heme, and rebinding is essentially completed on the picosecond time scale (Fig. 5, C and D). The main rebinding occurs with a time constant of 8 ps; a minor (ϳ10%) ϳ50-ps phase was also observed. For comparison, in Mb a rebinding phase of ϳ5 ps was also observed, but a substantial fraction of the dissociated O 2 did not rebind on the picosecond time scale (33,(35)(36)(37). Thus, in the M80A mutant, the heme environment appeared to act as an effective "cage" not only for CO and NO but also for O 2 .
Internal Ligand Rebinding-In the absence of external ligands and at elevated pH values, the ferrous heme in all M80X mutants coordinates the nitrogen of an amine group, most likely the side chain of Lys-79 (8). This ligand-iron bond is light-sensitive (6), and we have investigated the rebinding of this internal ligand upon photodissociation for the M80D mutant at pH 8. The apparent pK (pK app ) for the high to low spin transition on binding of this intrinsic lysine side chain in the ferrous M80D protein has been estimated previously at ϳ6.3 (8,38). Therefore, at pH 8 the M80D mutant is predominantly low spin with a lysine coordinated to the ferrous heme iron. As reported previously for cm-cyt. c (6), rebinding occurs in this mutant on the picosecond time scale ( ϭ ϳ7 ps; data not shown).
We also questioned whether, for the M80D mutant after photodissociation of CO from the heme iron, binding of the lysine side chain competes effectively with geminate recombination of CO. This particular mutant was chosen to investigate competition between the CO and lysine ligands at elevated pH values as we had previously studied these reactions for this protein under very similar conditions (8). Spectral analysis follow-  ing photodissociation of the CO derivative at pH 8, where binding of lysine is expected, showed that only CO bound and not lysine at t Ͻ4 ns. We therefore conclude that in the CO derivative the lysine is displaced from the vicinity of the iron and is probably not within the distal heme pocket but is located more toward the protein surface. On photodissociation, only the CO that migrates to bulk solution and rebinds on the millisecond time scale allows enough time for the refolding of protein on the distal side bringing the lysine close enough to the iron to bind. This conclusion is supported by comparing these results with those obtained for M80D at pH ϳ4.7, a pH value well below the alkaline transition for the lysine binding, where the lysine does not bind to heme iron (39). The results were identical at both pH values. We conclude that lysine replacement of photodissociated CO occurs on a time scale longer than 4 ns. This finding is also in agreement with observation of lysine binding with a rate constant of k ϭ 5750 s Ϫ1 for M80D at pH ϳ8 following photodissociation of CO from the ferrous iron (8). Lysine binding has been observed in the hundreds of microseconds for cm-cyt. c (8), with this time frame being more general for internal ligand binding in 6-coordinate heme proteins (40 -42).
Yield of CO Dissociation on the Millisecond Time Scale-For the M80A, M80D, and M80S mutants, the yield of CO released from the protein to bulk, as determined from the amplitude of the (millisecond) bimolecular phase of CO rebinding, has been determined previously (8). These amplitudes have also been measured for the other mutants and are tabulated in Table 1. As pointed out above, the observed values are in good agreement with the relative amplitudes of the constant phases determined by femtosecond spectroscopy. Together these findings indicate that the geminate rebinding phases of CO with 5-coordinate heme are essentially covered by the 4-ns time window.
Bimolecular Binding of Gaseous Ligands-The ligand concentration dependences of the pseudo first-order rate constants (k obs s Ϫ1 ) for combination of CO, O 2 , and NO with ferrous M80A are shown in Fig. 6A. As expected for such reactions, the values of k obs increased with increasing ligand concentration in an almost linear manner at low ligand concentrations. At these lower concentrations the trend observed for the rates of ligand combination to M80A is NO Ͼ O 2 Ͼ CO. At higher ligand concentrations, k obs values approached a plateau and all appear to be rate-limited at approximately the same value of ϳ70 s Ϫ1 (each fitted to a rectangular hyperbola, see "Discussion").
The ligand concentration dependences of the pseudo firstorder rate constants (k obs , s Ϫ1 ) for combination of CO and NO with ferrous M80D at low pH, where lysine does not bind to the heme (see below) are shown in Fig. 6B. As expected the values of k obs increased with increasing ligand concentration, with this trend being linear even up to the highest concentrations of both gaseous ligands CO and NO used (ϳ500 M). The trend observed for the rates of ligand combination to M80D is NO Ͼ CO. The values of the second-order rate constants for CO and NO combination to M80D were calculated as ϳ2.5 ϫ 10 5 and ϳ8 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively.
The inset to Fig. 6B shows k obs for combination of CO to M80E at low pH (pH ϳ5, see figure legend for details). The k obs values for this mutant also show a linear dependence with respect to the CO concentration up to the highest value available to us of ϳ1 mM. Shown in Fig. 6C are the steady-state spectra of the ferrous forms of M80A, M80D, and M80E at the low pH (pH Ͻ Ͻ pK app ) conditions of the bimolecular ligand binding experiments. They are high spin showing a large decrease in the ␣, ␤, and ␥ spectral bands with respect to the high pH low spin forms (see below). We suggest that in these only a weak field ligand or no ligand (shoulders at ϳ430 nm) is bound to the ferrous heme iron.
For all of the Met-80 mutants we have constructed, the ferrous proteins have identical spectra at pH Ͼ Ͼ pK app (Fig. 6C), indicating low spin iron and hence the presence of a bound strong field ligand (␣ ϭ 550, ␤ ϭ 520 and ␥ ϭ 417 nm) (8). Strong evidence that this ligand is a lysine is provided by the 1 H NMR of cm-cyt. c and M80E (8).

DISCUSSION
In wild type cyt. c, at pH values a few units either side of physiological, CO does not replace the endogenous heme iron ligands Met-80 or His-18. However, replacement of Met-80 by another amino acid, not expected to coordinate to the heme iron in the ferrous redox state, allows CO to bind (8,43). Our present results show that replacement by different residues effectively leads to engineering of a variety of "heme pockets" as sensed by the kinetics and spectral properties of heme-CO rebinding on the picosecond and early nanosecond time scale. The results are of interest on the one hand for the understanding of heme-ligand interactions in proteins, and on the other hand for developing efficient and "pseudo-native" photoactivable electron-injection proteins.
General Ligand Rebinding Properties-In general, recombination of CO with ferrous heme is slower than for O 2 and NO, presumably predominantly because of the difference in electronic interaction between the diatomic ligands and heme (44). Ligand rebinding in cyt. c is no exception in this respect (Figs. 3 and 5). However, although in many naturally ligand-binding heme proteins CO rebinding does not occur on the picosecond time scale and NO rebinding is a more sensitive tool for the heme environment, the inverse is the case in the modified cyt. c proteins studied here. In general terms, this observation can be rationalized by the notion that cyt. c is an electron transfer protein, rather than a ligand-binding protein, with the corresponding differences in flexibility and structure of the heme environment; in cyt. c, a functional ligand-accommodating heme pocket and a ligand exchange pathway with the aqueous environment are absent. For dissociated NO, which has a very high intrinsic affinity to heme, there is no effective competing pathway for rebinding to the heme. Indeed the 7-ps time constant of the unique phase of NO rebinding in the M80A mutant corresponds to the fastest phase of NO rebinding observed in many other heme proteins, including Mb (22,45,46), the NO sensor guanylate cyclase (47), and the sensor protein EcDos (41) (where external ligands replace an internal methionine heme ligand), and presumably corresponds to the intrinsic activationless rate of NO binding (48).
The intrinsic affinity of heme for CO is lower than for NO. In the heme-bound configuration CO is oriented perpendicular to the heme (49). In Mb, dissociated CO rapidly rotates (50), and although on the picosecond time scale it does not move away from the heme pocket (51,52), it does interchange between different orientational configurations (53), and it does not rebind to the heme in this time scale. It has been suggested that this is because a configuration of CO along the heme normal with the carbon atom pointing toward the iron, favorable for transition to the heme-bound state, is not frequently adopted (54). In the cyt. c mutants, dissociated CO does undergo geminate recombination to the heme with a high yield. This suggests that CO initially retains, or can easily adopt, an orientation close to the heme normal. Taking this view, at least initially the rotational freedom of CO appears restricted. To further investigate this issue, polarized transient infrared studies are in progress.
In contrast to NO rebinding, the CO rebinding kinetics are multiphasic, with distinct time constants, in all mutants. This observation implies that in competition with ligand rebinding, CO although partially restricted can nevertheless adopt a range of different configurations from which rebinding to the heme still occurs, but at a lower rate. The evolution of the transient spectra associated with the different decay phases indicates heme relaxation associated with progressively slower CO rebinding. As discussed previously (6), it is possible that the initial, more blue-shifted spectra correspond to hemes that are only partially domed, because of strong interactions with dissociated but closely sequestered CO. A picture thus emerges in which, after dissociation, heme-CO interactions are associated with fast rebinding and, in the heme-CO pairs that do not rebind, the progressive decrease in interaction increases their escape probability. Ϫ1 s Ϫ1 for CO and NO, respectively. Inset, CO concentration dependence of the observed rate constants for CO binding to M80E (sodium acetate, pH 5.0, 20 mM). Data points up to and including ϳ500 M CO were measured by stopped-flow spectroscopy and at 1 mM CO by laser flash photolysis of the CO adduct (8). C, the steady-state spectra of the low pH (pH Ͻ Ͻ pK app ) ferrous forms of M80A, M80D, and M80E: buffers used for spectra recorded, M80A (sodium phosphate, pH 7.0, 100 mM), M80D (sodium acetate, pH 4.8, 100 mM), and M80E (sodium acetate, pH 5.0, 20 mM). Shown also is a representative steady-state spectrum of the high pH (pH Ͼ Ͼ pK app ) ferrous forms of the M80X mutants (actual spectrum shown is of M80A at pH 11.0). All mutants display identical spectra at high pH values (see "Results").
The escape probability after the (sub)nanosecond CO-heme geminate recombination phases is Ͻ1% for two of the investigated cyt. c mutants (M80A and M80S, see Table 1). This is substantially lower than for any native ligand-binding protein, the lowest reported yields in this class of heme proteins being on the order of 5% for the CO sensor CooA from Rhodospirillum rubrum (55,56) and for the truncated hemoglobin HbO from Mycobacterium tuberculosis. 5 In addition, the fastest CO binding phases in the modified cyt. c electron transfer proteins (14 ps-30 ps, see Table 1) are, to our knowledge, faster than any other intra-protein heme-CO rebinding phase (1). These comparisons are consistent with the general idea that structural flexibility is more advantageous for functional ligand-binding heme proteins than for functional electron-transfer proteins, as indicated in the Introduction. Apart from our finding of variation of the CO rebinding kinetics in differently engineered proteins, it is clear that the protein matrix must be responsible for the fast rebinding, as geminate rebinding to the 5-coordinate heme iron in microperoxidase does not occur at low concentrations (25) (it does in aggregates (25,58)).
Effects of the Residue at Position 80 on CO Rebinding Kinetics and Yield-Comparison of the asymptotic values obtained from fitting the picosecond/nanosecond decay kinetics and the quantum yield of CO release on the millisecond time scale indicates that essentially all geminate rebinding takes place in the time window of our experiments. In native cyt. c, the heme environment is highly hydrophobic and rigid (12,13). The highest yield of geminate recombination (Ͼ99%) is found with Ala and Ser, which are both the smallest residues tested and the ones that are uncharged. These residues keep the hydrophobic nature of the heme pocket intact and leave presumably enough space for CO to be accommodated in the heme pocket without major rearrangement (8), and thus without rendering the protein more flexible so as to allow dissociated CO to undergo substantial translational and rotational motion. Asp at position 80 adds occupied volume as well as a negative charge to the heme environment, and the CO escape probability increases to about 4% (i.e. A 0 ϭ 0.04 and ϭ 0.04) ( Table 1). The fact that the millisecond escape probability is very similar with the somewhat bulkier Glu residue (8) suggests that the introduced charge plays an important role in the disruption of the rigidity of the heme pocket leading to enhanced CO escape.
By far the largest increase in CO escape was observed by introducing the long, positively charged side chain of Arg. This large increase is associated with transient heme spectra that indicate less interaction between CO and the heme already at short times after dissociation (Fig. 2). For this mutant, the cavity is clearly disrupted so as to allow substantial motion of the dissociated CO. The Arg residue occupies a much larger volume than the native Met residue. In principle the following are possible: (a) the Arg side chain points toward the heme and forces rearrangement of the other constituents of the heme pocket; (b) the pocket is rearranged so that the Arg side chain points toward the more hydrophilic protein surface, opening a crevice for CO motion. We believe the latter possibility is more likely, as a very crowded cavity would be expected to put strain on CO as well as the heme, which is not observed in the transient spectra, and in this case little freedom of motion for CO would be expected. An interesting comparison can be made with hemoglobin Zürich, where the distal histidine is replaced by arginine. In the CO complex, this substitution leads to an outward movement of the guanidinium group, which opens up the heme pocket and enhances rates of ligand exchange (59).
Further support for our interpretation of the effect of the M80R substitution in cyt. c is provided by NMR analysis of the solution structures of horse heart cyanoferricytochrome c (60) and yeast cyano-M80A ferri-cyt. c (10), and the x-ray structure of yeast iso-1-cytochrome c (61). Yao et al. (60) reported that on CN Ϫ binding to the ferric heme iron of cyt. c Met-80 is displaced, and the rupture of this Met-80 -iron bond increases the flexibility of residues 79 -82. This in turn leads to residues 77-85 (see Fig. 1) rotating away from the heme leaving Met-80 partially solvent-exposed. A different picture is given by the NMR structure of yeast cyano-M80A ferri-cyt. c (Fig. 1), which shows a very similar backbone conformation of the 77-85 segment to that of yeast iso-1-cytochrome c (61). Yao et al. (60) rationalized these findings by suggesting that in the M80A mutant a larger cavity exists on the distal side compared with the Met-80 protein, the shorter side chain of alanine providing sufficient room to allow binding of an external ligand (in this case CN Ϫ ) to the iron without major conformational change (60). In the Met-80 protein, however, the distal cavity cannot accommodate both the CN Ϫ bound to the iron and the side chain of the methionine leading to the conformational change described above. We believe that a similar conformational change as described for cyanoferricytochrome c (60) may be taking place in the M80R mutant. The distal pocket cannot accommodate both the long, positively charged side chain of an arginine residue at position 80 and a CO molecule bound to the heme iron, leading to the rotation of residues 77-85 ( Fig. 1) to leave the arginine side chain, at least partially, solvent-exposed. It may be that this conformational change makes the major contribution to the protein's large increase in over all other mutant proteins tested so far. This argument also explains the relative values of for the "minimally" modified form of cmcyt. c and M80X mutants Ala, Glu, and Asp. The value of for this form of cm-cyt. c is ϳ0.12 compared with M80A ϳ0.004 and M80D and Glu ϳ0.04 (i.e. a 30-and 3-fold difference, respectively). We explain these findings by noting that for M80D and M80E a negative charge is introduced into the distal cavity that is not present in the M80A mutant, and this makes the cavity more flexible allowing CO to diffuse more readily to bulk. With the introduction of bulk and a negative charge, i.e. a carboxymethylated methionine as is the case for cm-cyt. c, the disruption of the cavity is more radical than for the acidic 80 mutants alone, and there is a conformational change of the protein backbone resulting in the rotation of residues 77-85 (Fig. 1). For M80R, the value of is twice that of cm-cyt. c. We rationalize this finding on the basis that the bulky and positively charged side chain of the arginine residue leads not only to rotation of residues 77-85, rendering the guanidinium group of Arg-80 solvent-exposed, but also further disruption of the distal cavity. We therefore propose that a combination of steric and electrostatic factors control whether on binding an exogenous ligand the residue at position 80 remains within or is expelled from the distal pocket. This in turn dictates the value of . A systematic screening of amino acids, in combination with structural characterization, is required for a detailed understanding of the parameters governing the CO recombination dynamics. However, the remarkable finding that substitution of a single residue at position 80 results in variation in (the yield of CO escape from the cavity) over almost 2 orders of magnitude implies that the CO dynamics are a very sensitive probe of the heme pocket of cyt. c.
As suggested above by the kinetic argument (see "Results"), we consider it unlikely that the distal heme cavity of cyt. c with an exogenous ligand bound to the iron atom can accommodate bulky residues, especially if these are charged. Thus, for the CO-bound forms of the ferrous Met-80 mutants at pH values where the lysine side chain is able to bind to the iron (pH Ͼ pK app for the alkaline transition), we nevertheless expect the lysine side chain to be solvent-exposed. In agreement with this reasoning, competition between geminate rebinding of photolysed CO and binding of lysine to the iron is not observed. As in other 6-coordinate heme proteins, the residue can bind only on the 100-s time scale after CO dissociation and escape (40 -42).
Effect of Residue at Position 82-To explore ways to increase the quantum yield of CO escape, we designed double mutants starting from the M80D protein. One possibility to reduce steric hindrance for CO is to remove bulky residues close to the heme. Examination of the NMR structure of the M80A mutant (10) (see Fig. 1) suggests that suitable substitution of Phe-82 may facilitate diffusion of photodissociated CO from the close proximity of the iron. The Phe-82 residue is highly conserved in the primary sequences of eukaryotic cyt. c molecules and has been shown to influence significantly protein stability, heme reduction potential, and oxidation state-dependent conformational changes (62). In general, mutation of Phe-82 in yeast iso-1-cytochrome c destabilizes the native conformation of the protein by facilitating the ligand exchange reactions associated with the alkaline transition of the ferricytochrome (63). Singly modified Phe-82 mutants have already been shown to retain native structure but possess a more flexible heme pocket (64). Indeed when, in addition to the Met-80 to Asp substitution, Phe-82 was replaced by either of the smaller residues Gly or Asp, geminate recombination became less efficient, and the yield of CO escape increased 2-3-fold, implying that these substitutions allow CO to move away from the heme. The decrease of the rate and amplitude of the initial recombination phase (Table 1), and the changes in transient spectrum at short delay times, with respect to the M80D single mutant, all indicate that in the double mutants the interaction of CO with the heme is decreased already at short times after dissociation. The effect can be considered as the inverse of the observed increase in recombination of NO in Mb upon substituting a distal Val by Phe (18,19). Combining this mutation with the M80R substitution may also lead to further increase of the CO escape yield with respect to the single mutation, although the relevance for such a double mutant for photoinduced electron transfer studies may be limited (see below).

Bimolecular Combination of Gaseous Ligands with M80X
Mutants; Protein Dynamics Are Rate-limiting-The rate constants for binding of CO, O 2 , and NO to M80A are rate-limited at ϳ70 s Ϫ1 , whereas combination of M80D with CO and NO and M80E with CO (at least) are not rate-limited even at high concentrations of ligand (see Fig. 6). This clear distinction requires explanation.
The high to low spin transition induced by coordination of a lysine residue to the heme group of M80A has a pK app ϳ 9.4, and thus the ferrous iron of this particular mutant is predominantly high spin at pH 7, the pH used in these studies (see Fig.  6C) (8). We can therefore eliminate the possibility that the rate of binding of gaseous ligands is limited by dissociation of the lysine residue from the iron. Furthermore, studies on the kinetics of CO binding to M80A at pH values well above the pK app values indicated that the rate of CO binding is now indeed limited by the rate of lysine dissociation, a plateau at k ϭ ϳ 0.02 s Ϫ1 being reached at 250 M CO (31). This limit may be contrasted with that of 70 s Ϫ1 as we observed for CO binding at pH 7 and which therefore must have a different structural basis.
The ligand concentration dependences for combination to M80A reported in Fig. 6A can most readily be explained by invoking the argument already advanced to account for the high geminate yields, namely the highly caged and rigid environment afforded by the distal heme pocket. We propose that the bulk of the M80A protein is essentially "gas-tight" and that ligand access is only afforded by protein dynamics that allow entry. A simple mechanism may be depicted in Scheme 1, where C represents the protein conformation that is closed to ligand entry; O represents that sub-set of "open" conformations (possibly very few in number) that allow ligands (L) to approach the ferrous iron from bulk solution, and A represents the adduct form where L is bound to the heme iron. Thermal ligand dissociation is, for simplicity, ignored, as we know that the "offrates" of these ligands, NO, O 2 , and CO, from ferrous iron are relatively low. As we also know from our stopped-flow experiments that binding follows a simple single exponential process, we can conclude that, in the absence of light-induced ligand dissociation (indicated as a dashed arrow in Scheme 1), the concentration of O is small (i.e. k Ϫ1 Ͼ Ͼ k 1 ), and d(O)/dt will thus also be very small at all times. Given these constraints, the mechanism predicts a rate constant k obs for overall bimolecular ligand binding as shown in Equation 1, The rate limit, determined to be ϳ70 s Ϫ1 from Fig. 6A, is thus the rate constant (k 1 ) for the protein conformational change that allows ligand binding to the heme iron. The initial slope of SCHEME 1 the graph shown in Fig. 6 now corresponds to k 2 /K (where K ϭ k Ϫ1 /k 1 ) and thus is directly proportional to k 2 , the second-order combination of the ligand to the iron. The values of these slopes are ϳ5, 4, and 1 ϫ 10 5 M Ϫ1 s Ϫ1 for NO, O 2 , and CO, respectively. The order of these rate constants (i.e. NO Ͼ O 2 Ͼ CO) is the same as that expected for Mb (65) and, given the upper limit of k 2 for binding of gaseous ligands to heme on the order of 10 8 M Ϫ1 s Ϫ1 (66), suggests a lower limit for the value of K of ϳ100. It is interesting to note that the order of the slopes k 2 /K (NO Ͼ O 2 Ͼ CO) is also the same as for the global time scales of geminate recombination. As it is reasonable to assume K ligand-independent, one might speculate that the ligand dependence of the bimolecular rate k 2 is at least in part determined by the latter. However, this order may also be due to differences in ligand migration to the heme pocket from the solvent, for which, in contrast to Mb (67) no information is available for cyt. c.
If CO escape to bulk solution following its photodissociation from the heme iron were to be "gated" in a similar manner to bimolecular combination (as suggested in Scheme 1), then the estimate for the value of K suggests the upper limit for the value of (the quantum yield). For M80A at least, this proposal is attractive as only ϳ1% of the protein would exist in the open conformation consistent with the value of ϭ 0.005 (i.e. the percentage of CO escape following photodissociation is 0.5%).
This model we presume also applies to the other M80X mutants but with altered values of the rate and equilibrium constants governing the conformational change that permits ligand approach from bulk. For example in M80D and M80E (also high spin) the pseudo first-order rate constant for CO combination from bulk is linear, within our experimental error, up to ϳ125 and 175 s Ϫ1 , respectively (Fig. 6B). For NO recombination to M80D, rates are linear up to ϳ375 s Ϫ1 . This implies that any rate limit is probably of the order 10 3 s Ϫ1 (or greater). The apparent second-order constants for CO binding to M80D and M80E taken from Fig. 6 are ϳ2.5 and 1.7 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively. Given that these constants are approximately an order of magnitude lower than CO binding to unfolded cyt. c, we conclude from the argument advanced above that for these mutants the value of K is ϳ10, consistent with the value of ϭ 0.04 for both mutants (Table 1) (8). At pH values Ͼ ϾpK app , the rate of CO binding to the low spin ferrous heme in each mutant is limited, presumably in these cases by the off-rate of the lysine ligand bound to the iron.
Recently, much interest has focused on the many small and rapid molecular motions that take place in proteins that, when combined, bring about major conformational changes related to the function of the protein. In this context ligand binding to Mb is probably one of the most studied systems. Combining x-ray diffraction structure determination with rapid photolysis of CO on single crystals of Mb have allowed the reaction intermediates immediately following photodissociation of a CO molecule from the heme iron to be studied and the trajectory of CO from the distal pocket to bulk mapped (68 -71). These studies have provided insight into the way ligands diffuse away from the buried heme group following the rupture of the iron-ligand bond, indicating that "breathing" of the protein structure facilitates ligand migration and only certain conformations in a particular time window allowing ligand passage to bulk. The crystal structures of Mb show that gaseous ligands appear to have very restricted access from bulk solution to the heme iron and vice versa. However, gaseous ligands do freely exchange between bulk and the heme iron in Mb because the protein can undergo many small and localized molecular motions that when combined create passages for ligand exchange processes. A specific example of an important yet small conformational change controlling ligand access to the heme is that involving the distal histidine residue (72) and termed the His-gate in Mb (73). It has recently been estimated that for Mb essentially all CO exits via the His-gate following photodissociation from the heme iron (74,75). Our analysis of bimolecular CO binding suggests that a qualitatively similar (but far less efficient) mechanism may determine the CO escape yield from the distal pocket in cyt. c.
No substantial geminate binding of CO occurs after a few nanoseconds (at most in the case of M80R) ( Table 1). This finding suggests that leaving the heme pocket (presumably in competition with rebinding to the heme as discussed above) constitutes the major barrier for CO migration out of the protein, and no further cavities allow re-access of the heme pocket.
Relevance for Photo-induced Electron Transfer and Further Mutational Studies-The percentage of CO that escapes from M80R, M80D/F82D, and M80D/F82G per absorbed photon is ϳ31, 11, and 13, respectively (see Table 1). As the CO recombination phase is completed within ϳ1 ns, long and intense light pulses, such as the several nanosecond pulses from standard Nd:YAG lasers, can be used to release a higher proportion of CO per flash via multiple excitations of each heme. For example, when cm-cyt. c (% CO escape for the maximally modified form is ϳ25) is associated with an electron acceptor, such as cytochrome c oxidase, and CO is photodissociated by a Nd:YAG laser then the quantum yield of electron delivery to the acceptor has been shown to be close to unity (4,7).
Thus, in principle, we have constructed suitable candidates for investigation of photo-induced electron transfer. The cyt. c mutants produced bind CO that may be driven into bulk solution by a laser pulse of a few nanoseconds duration permitting electron transfer to an associated acceptor. The importance of the structures of such proteins being close to native as possible has been paramount in the design of these mutants (cf. cm-cyt. c). As discussed above, it is probable that for M80R the native structure is not as well retained as we would like, especially compared with M80A. It appears that in order to produce cyt. c mutants that allow relatively free passage of CO to bulk, there may be a "trade off" at the expense of retaining the native structure. However, it would seem that for the double mutants M80D/F82D and M80D/F82G, both proteins retain native structure and possess appreciably higher values of quantum yield ( ϭ Ͼ0.1) compared with the M80X mutants Ala, Ser, and Asp.
In view of our results, strategies to increase the value of the quantum yield () will include mutating residues that, like the Phe-82 mutants, appear to create only localized changes in protein structure and facilitate CO migration to bulk by creating a more flexible heme pocket and/or creating free volume that allows a reduction of the interaction of the dissociated CO with heme.
Conclusions-We have shown that geminate recombination of CO and heme in mutant cyt. c occurs very rapidly, predominantly on the picosecond time scale, and in a multiphasic manner. We propose that these unusual characteristics are related to the relative inflexibility and hydrophobicity of the heme environment, properties required for efficient electron transfer. However, by substituting Phe-82 by Gly or Asp in addition to the Met-80 to Asp substitution, the yield of CO escape may be sufficiently high for these mutants to be useful as near-native, photoactive electron transfer proteins.