Ligand Dynamics and Early Signaling Events in the Heme Domain of the Sensor Protein Dos from Escherichia coli*

In the heme-based sensor Dos from Escherichia coli, the ferrous heme is coordinated by His-77 and Met-95. The latter residue is replaced upon oxygen binding or oxidation of the heme. Here we investigate the early signaling processes upon dissociation of the distal ligand using ultrafast spectroscopy and site-directed mutagenesis. Geminate CO rebinding to the heme domain DosH appears insensitive to replacement of Met-95, in agreement with the notion that this residue is oriented out of the heme pocket in the presence of external ligands. A uniquely slow 35-ps phase in rebinding of the flexible methionine side chain after dissociation from ferrous DosH is completely abolished in rebinding of the more rigid histidine side chain in the M95H mutant protein, where only the 7-ps phase, common to all 6-coordinate heme proteins, is observed. Temperature-dependence studies indicate that all rebinding of internal and external ligands is essentially barrierless, but that CfigsO escape from the heme pocket is an activated process. Solvent viscosity studies combined with molecular dynamics simulations show that there are two configurations in the ferrous 6-coordinate protein, involving two isomers of the Met-95 side chain, of which the structural changes extend to the solvent-exposed backbone, which is part of the flexible FG loop. One of these configurations has considerable motional freedom in the Met-95-dissociated state. We suggest that this configuration corresponds to an early signaling intermediate state, is responsible for the slow rebinding, and allows small ligands in the protein to efficiently compete for binding with the heme.

The iron atom in heme proteins can coordinate two axial ligands. One of these ligands is almost invariably a histidine residue from the protein backbone. The other binding site can be unoccupied, occupied by an external ligand like O 2 or H 2 O, or occupied by another amino acid residue. In many heme proteins, changes in heme coordination play an important role in functional processes. A rapidly expanding group of proteins can stably adopt 6-coordinate (6-c) 3 configurations with either an internal residue or an external ligand bound to the 6th position (1)(2)(3)(4)(5)(6). In particular, heme-based sensor proteins have been identified in which the exchange of an internal ligand and an external signaling molecule at the heme binding site leads to changes in the activity of an associated enzymatic domain.
Dos from Escherichia coli is such a heme-based sensor protein. It is composed of a heme domain of known structure, DosH (7,8), and a phosphodiesterase domain. The former contains a b-type heme and is coordinated by histidine (His-77) and methionine (Met-95) in the ferrous unliganded form (Fig. 1). Small gaseous molecules like CO, NO, and O 2 can replace Met-95. As DosH shows high sequence similarity with the heme domain of the oxygen sensor FixL (which by contrast possesses a 5-coordinate (5-c) heme in the ferrous unliganded form), it was proposed to be a direct oxygen-sensing domain as well (2). However, in the ferric form the heme binds a H 2 O molecule, and the activity of the holoprotein with cyclic AMP as substrate strongly decreases upon oxidation. Consequently, Dos has also been proposed to act as a redox sensor (9). Very recently, a role as a general gas sensor has also been put forward on the basis of moderate up-regulation of activity by CO, NO, and O 2 with cyclic diGMP as substrate (10), and a crucial role of Met-95 in this regulation has been demonstrated (11). The ligand-induced up-regulation in Dos contrasts with the O 2 -induced inactivation of kinase activity in FixL.
In the ferrous form of the heme, the ligand at the 6th position is photolabile. This is the case for the small diatomic gaseous ligands (12,13) but also for amino acid residues (14 -16). This property allows triggering of heme-ligand dissociation using ultrashort light pulses and following the initial dynamics of the system spectroscopically (17). For internal ligands such dynamics have been studied for a very limited number of systems, i.e. methionine in cytochrome c (16,18) and proline in the bacterial CO sensor CooA (19). In both systems, the attachment to the protein backbone keeps the dissociated internal residue close to the heme cofactor so that it fully recombines with heme in a monoexponential way in ϳ6 ps.
The early dynamics of ligands in wild-type (WT) DosH are unusual in several aspects (20). First, in contrast to the above examples, the dissociated methionine recombines in a clearly biexponential way: not only was a 7-ps phase observed as in other systems but also an additional slower phase of ϳ35 ps with similar amplitude. Because dissociation of methionine must be the initial step changing the activity in the holoprotein from that associated with the unliganded reduced form to that associated with external ligand (or to that of the oxidized waterbound form in the case of redox sensing), this finding opens the exciting perspective that several steps in the early signaling pathway can be followed. Second, the dynamics of external ligands has particular characteristics, and points at a restrained heme pocket. Apart from O 2 recombination, which predominantly occurs within ϳ5 ps as in FixLH (21) and possibly even faster (22), CO geminate recombination occurs to a substantial amount (60%) in ϳ1.5 ns (20). The present work aims at exploring the possibility of monitoring initial signaling events in DosH, and the role of the relatively flexible side chain of the crucial methionine 95 residue, by combining site-directed mutagenesis, femtosecond spectroscopy, and molecular dynamics simulations.
Early site-directed mutagenesis studies have identified Met-95 as the displaceable axial residue in 6-c reduced DosH (23)(24)(25). This assignment was confirmed by x-ray crystallography studies (7,8). Indeed, replacement of Met-95 by alanine, leucine, or isoleucine leads to a 5-c ferrous heme. These findings also indicate that in the absence of Met-95, no other residue occupies the binding site. This contrasts with cytochrome c, where replacement of the methionine ligand can lead to the coordination of closely located lysine residues (26). In DosH, replacement of Met-95 by histidine leads to a bis-histidine 6-c form (24,25). Met-95 is part of the flexible and solvent-exposed FG loop. This loop has been implicated in intra-protein signal propagation toward the enzymatic domain, both in the ferrous deoxy to oxy transition (2,8) and in the ferrous to ferric transition (7).
In the present study we compare WT DosH with M95I and M95H mutant proteins. Isoleucine, the corresponding residue in FixL (23), has a volume similar to methionine and cannot form a bond with the heme. Histidine does bind to the heme but has a less flexible side chain.

DNA Manipulations, Protein Expression, and Purification-
Chromosomal DNA from E. coli HB101 was isolated following standard protocols. The DNA fragment corresponding to codons 1-152 from E. coli yddU (renamed dos (2)), coding for the oxygen-sensing PAS domain (DosH), was amplified using the primers 5Ј-ATG AAG CTA ACC GAT GCG GAT A-3Ј (forward) and 5Ј-C GAC CGA CCG GTG ATT GTC CTC-3Ј (reverse) and cloned into a pCRT7/NT cloning vector (Invitrogen) under control of an isopropyl 1-thio-␤-D-galactopyranoside-inducible T7 promoter. The recombinant gene fragment corresponding to the Dos heme domain in E. coli was used as template for site-directed mutagenesis reactions. The substitutions M95I and M95H were introduced following the QuikChange site-directed mutagenesis protocol (Stratagene) by replacing the codon ATG by ATC and CAC, respectively. All constructs were confirmed by DNA sequencing prior to further analysis. The final constructs were transformed into E. coli strain BL21DE3 for expression. Protein expression and purification were performed as described in a previous study (20).
Sample Preparation-The proteins were prepared to a sample concentration of 50 -70 M in a gastight optical cell, in 50 mM Tris-HCl, pH 7.4. Ferrous deoxy-DosH was prepared by degassing the sample and reducing it with sodium dithionite. To obtain the CO-bound form, it was subsequently equilibrated with 1 atm of CO gas. For the O 2 -and NO-bound form, samples were degassed, reduced with sodium ascorbate and equilibrated with 1 atm of O 2 or 0.01 atm of NO, respectively. For the experiments in the presence of glycerol, concentrated reduced and degassed DosH solutions were mixed in the sample cell with degassed glycerol.
Spectroscopy-Steady-state spectra were recorded using a Shimadzu UV-visible 1601 spectrophotometer. Multicolor femtosecond absorption spectroscopy (27) 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. Multiple time window acquisition schemes were used, with windows ranging from 4 ps to 4 ns full scale. Unless otherwise indicated the experiments were performed at 20°C; for some experiments the sample holder was thermostatted in the range of 8 -40°C. For most experiments the sample was continuously moved perpendicular to the beams to ensure sample renewal between subsequent pulse pairs. This was not the case for the experiments comparing reduced unliganded glycerol-containing samples (which could not be fully homogenized) with aqueous solutions. It was verified, however, that for the buffer-only samples, the kinetics values for each were identical with and without movement.
Molecular Dynamics Simulation-A model of the Dos heme domain was constructed using the structure from PDB file code 1V9Z (7) with the molecular modeling program CHARMM32 (28). Two identical polypeptide chains of 113 residues (amino acid 20 -132), 2 hemes, 196 water molecules from the structure, and 1411 additional water molecules composed the model. Hydrogen atoms were added using the CHARMM HBUILD command and a relative dielectric constant equal to unity was applied to the entire structure. The heme iron atom was bonded to the His-77 N⑀ 2 and Met-95 S ␦ . For the trajectories shown, the bond length and force constant for the latter bond were taken to be those used previously for cytochrome c (29). For comparison, a model was also used with a weaker bond (30 kcal/mol/ Å 2 ), consistent with density functional theory calculations of the iron protoporphyrin-imidazole-dimethyl sulfide system, that we performed.
Non-bonded interactions were shifted to zero between 8 and 14 Å. The energy of the structure was minimized with SD and ABNR algorithms, and the temperature of the model was raised to 300 K over 30 ps in 30 steps of 10K. A 300-ps equilibration phase, with atomic velocities assignment randomly chosen in a Boltzmann distribution at 300 K each picosecond, was applied to the model. A 650-ps free dynamics trajectory was then performed. This trajectory was used to generate different initial structural conditions before suppression of the heme-Met-95 bond. The root mean square deviation of backbone atoms from the PDB structure remains Ͻ1.2 Å.
Dissociation of Met-95 from heme was simulated by suppression of the Met-95 S ␦ -heme iron-bonded interactions and switching to a five coordinate heme parameter set. Three 500-ps dissociation simulations were performed. The model structure was saved each picosecond for further analysis. Fig. 2 shows steady-state spectra of ferrous deoxy-WT, M95I, and M95H DosH. These are in agreement with previously published spectra (2,23,25), demonstrating that the M95I mutant is 5-c, whereas M95H is 6-c, as WT DosH. Fig. 2 also outlines the spectral choices for pump-probe spectroscopy, with excitation in the ␣ band and probing in the Soret band, where large differences between 5-c and 6-c species are observed. The spectra of the carboxylated complexes are very similar for all three complexes and typical for 6-c CO complexes (not shown).

Steady-state Spectroscopy of DosH Mutant Proteins-
Internal Ligand Rebinding in DosH-Ferrous deoxy-M95I DosH is 5-c. Excitation of this species leads to spectral changes (not shown) very similar to those observed in deoxymyoglobin (30) and deoxy-FixLH (21). They evolve with time constants up to ϳ3 ps, which can be understood in terms of excited-state photophysics (12, 31).  (20). The rebinding kinetics in M95H DosH are single exponential, with a time constant of 7.6 ps, close to that of the fastest phase in WT (and to that found in other 6-c heme proteins). Thus, it appears that the biexponential kinetics observed in WT are specific for dissociated methionine.
We note that both in M95H and WT DosH a small long-lived spectral feature remains that does not reflect formation of a 5-c ferrous species, but rather a few percent of heme-oxidation (20). This feature is not observed upon photoexcitation of the 5-c M95I mutant.
External Ligand Rebinding-The kinetics of CO rebinding in WT DosH is unusual in that ϳ60% of dissociated CO rebinds to the heme on the nanosecond timescale, prior to leaving the protein (20). This implies that direct rebinding of dissociated CO effectively competes with its migration out of the heme pocket. Fig. 4 shows that, whereas small but significant changes in the CO rebinding kinetics are observed upon replacement of Met-95 by isoleucine or histidine, the WT characteristics do not strongly change. In particular, we note that the kinetics in the M95I mutant does not resemble the kinetics observed in FixLH (where very little decay occurs 4 ), which has isoleucine at the equivalent position in wild type (23). All kinetics could be reasonably described by a fit to a monoexponential decay and an asymptotic value (see Table 1 for fit parameters). These findings indicate that the residue in position 95 does not strongly 4 We previously reported that CO dissociated from WT FixLH does not recombine with heme up to 4 ns (20). Recent improvements in the alignment procedure of the delay line have allowed assessing that in fact a small (ϳ10%) recombination in ϳ400 ps does occur.  influence the heme pocket volume and surface when it is replaced by an external ligand. Transient absorption experiments on the oxycomplexes of the DosH M95I and M95H mutant proteins gave very similar results to those of the WT protein (21), indicating a very low yield of dissociated oxygen on the timescale of a few picoseconds. Similar observations have been made in FixLH (21,22).
Temperature Dependence Studies-The unusual kinetics of CO rebinding provides a unique opportunity to investigate potential activation barriers involved in CO rebinding from the heme pocket and in migration out of the heme pocket, by determined the temperature dependence. Fig. 5A shows that the overall decay increases moderately with decreasing temperature, with both the rate and relative amplitude of the decay phase varying. This indicates that thermodynamic barriers are involved in the CO dynamics. We analyzed these data in terms of Scheme 1,   Table 1.  Fig. 5B shows an Arrhenius plot of both rates k 1 and k 2 . It is clear that the temperature dependence of the kinetics arises essentially from a decrease of k 2 with temperature. Thus CO migration out of the heme pocket has an enthalpy barrier of 45 meV (1.0 kcal/mol) Ϯ 10 meV, and CO rebinding to the heme (with a time constant of ϳ2.8 ns) appears essentially barrierless.
Similar experiments were performed for reduced DosH without external ligands (Fig. 6A). In this case it was found that the kinetics of methionine rebinding to the heme did not change within experimental error in the investigated temperature range. Thus, both rebinding phases are essentially barrierless. Similarly, the rebinding kinetics of His-95 to the heme in the M95H mutant was found to be temperature-independent (not shown).
Glycerol Effect-To investigate the influence of the environment of the heme domain on the dynamics of dissociated methionine and to be able to discriminate between different models of interconversion and rebinding of the two methio-nine-dissociated states (see "Discussion"), we determined the effect of the solvent viscosity on the dynamics of reduced DosH by varying the glycerol concentration. The overall rebinding kinetics increases speed with glycerol concentration (Fig. 6B). Analysis of the kinetic traces indicates that, as with temperature, the rates are independent of glycerol concentration (Fig.  6B, inset). However, the relative amplitudes of the phases do vary: the fast phase increases with glycerol concentration (Fig.  6B). This suggests that two distinct configurations of methionine can be adopted directly after its dissociation from the heme, the probability of adopting either configuration being solvent-dependent, but the interconversion rate not (see "Discussion").
For comparison, the effect of varying the glycerol concentration on the rebinding kinetics of NO and CO was also determined. For NO, full geminate recombination was observed with a main phase of 5 ps (85%) and a minor phase of 20 ps (20), without significant escape from the heme vicinity. This kinetics does not alter significantly when 60% glycerol is present (not shown). This observation is consistent with previous observations on other heme proteins that the time constant of the fastest phase of geminate NO recombination is independent of glycerol concentration (32,33) and supports the idea that these phases reflect predominantly barrierless rebinding.
By contrast, geminate recombination of CO, which is incomplete and occurs on a much longer timescale, does depend on glycerol concentration (Fig. 5C). With increasing viscosity, the overall rebinding becomes faster, which is generally consistent with previous studies on NO rebinding in other heme proteins (32,33), with CO rebinding to protoheme (34,35) and with a barrier for CO escape in DosH as deduced from the temperature-dependence studies (Fig. 5B). A quantitative analysis in terms of Scheme 1 is complicated by the fact that the kinetics also becomes markedly multiexponential in the presence of glycerol. This finding may indicate that in the presence of glycerol a distribution of distinct configurations exists that interconverts on a timescale longer than a few nanoseconds.
Molecular Dynamics Simulations-To get insight into the nature of the methionine-dissociated states in reduced WT DosH, we performed classic molecular dynamics simulations. In these simulations, trajectories of the 6-c methionine-bound state were calculated, and bond breaking was simulated by instantaneously suppressing the Fe-Met bond. The rebinding process itself was not directly taken into account in these classic simulations. In the trajectory of the 6-c form, two main different configurations of Met-95 were observed. Configuration I was like the x-ray structure (pink configuration in Fig. 7A), and in configuration II a concerted rotation over the 1 and 3 dihedral angles occurred (blue configuration in Fig. 7A). Similar configurational changes also occurred after dissociation of Met-95 (Fig. 7B). In configuration II of the dissociated form (Fig. 7B), the sulfur atom of Met-95 is located at a larger distance (average of 5.1 Å) from the heme iron than in configuration I (average of 3.4 Å). Configuration II was infrequently populated (for ϳ200 ps on a total of 2150 ps), and its dwell time was in the order of 50 ps (Fig. 7, C and D). However, we found that its relative occupation time was dependent on the force constant for the Fe-Met bond and increased substantially when lower values were taken (see "Materials and Methods").
Changes between the two configurations occurred within 1 ps. Remarkably, changes in Met-95 configuration did not occur upon dissociation of the residue from the heme iron. Thus, the population distribution of the two configurations can be considered as almost "static" on the timescale of the methionine rebinding experiments of Fig. 3.
Finally, Fig. 7 (A and B) also show that the difference in configurational change in Met-95 extends to movement of the backbone of the neighboring residues in a loop that is solvent-exposed (cf. Fig. 1). Thus, changes in solvent composition (in our simulation composed only of water) may influence the population of the two configurations. This finding is consistent with the observed experimental glycerol effect on the population of the two methionine configurations discussed above.

DISCUSSION
The Met-95 residue in Dos plays a crucial role in intramolecular signal transmission as it is ligated to the heme iron in the reduced form without external ligands, and it swaps to a position pointing out of the heme pocket when an external ligand coordinates to the heme (7,8). This situation contrasts with the very similar FixL heme domain, where the corresponding isoleucine residue (Ile-218 in the Bradyrhizobium japonicum enzyme) points out of the heme pocket in all ligation forms (36,37). In our present experiments we have investigated the initial heme-ligand dynamics, and the role of the residue in position 95 in DosH, under conditions where it is either in the heme pocket or pointing out of it.
CO Rebinding Dynamics-CO dissociation from the heme in WT DosH leads to substantial recombination in ϳ1.5 ns (20); this is remarkable because significant geminate recombination on this timescale rarely occurs in heme proteins and, in particular, is small in the related protein WT FixLH (21,38). In FixL the residue corresponding to Met-95 in Dos is isoleucine. However, the M95I replacement in DosH does not lead to kinetics for CO rebinding similar to those in FixL, where very little rebinding occurs in the first few nanoseconds (Fig. 4). This finding implies that Dos Met-95 is not the major factor determining the CO-rebinding kinetics. Overall, the finding that substitution of methionine 95 leads to only moderate changes in the CO rebinding kinetics indicates that in DosH Met-95 is folded out of the heme pocket in the CO-bound state, similarly as observed in the x-ray structure of the O 2 -bound state (7,8). The same holds true for the corresponding residue Ile-218 in CO-bound FixLH from B. japonicum (36,37). Whereas the distal heme pocket is very hydrophobic for both Dos and FixL, the steric properties of the residues composing it are not the same. For instance, the equivalent residues of Ile-215 and Leu-236 in B. japonicum FixLH, both located at ϳ3 Å from the CO oxygen atom in CO-bound FixLH, are Val and Phe, respectively, in DosH (2). Thus, different steric constraints may be at the origin of the differences between CO dynamics in the heme vicinity in DosH and FixLH. For instance, the higher rebinding efficiency in DosH compared with FixLH (see above) may arise from the possibility for CO in DosH to adopt a perpendicular orientation close to the heme more easily, favoring rebinding.
The substantial yet relatively simple (monoexponential) kinetics of CO rebinding in DosH (in buffer) makes it a convenient system to study the thermodynamic properties of nanosecond CO dynamics in a heme pocket. The experimental result, that both the relative amplitude and the apparent rate of the rebinding phase vary with temperature, can be analyzed in terms of a scheme. (Scheme 1 and Fig. 5B) of competition between essentially barrierless (Ͻ10 meV) rebinding ( Ϸ 2.8 ns) and activated (barrier enthalpy 45 Ϯ 10 meV) migration of CO out of the direct heme environment with a similar rate around room temperature. The 2.8-ns rebinding phase is substantially slower than the fastest phases of CO rebinding observed in several other heme proteins (19,39,40) that are on the sub-100-ps timescale. Therefore, we suggest that this phase does not reflect the intrinsic heme-CO rebinding rate, but rather is limited by very low barrier motions of CO in the heme pocket in adopting a favorable configuration for rebinding.
A picture emerges where initially the heme and the dissociated CO are not in a favorable configuration for direct rebinding. In principle this can be due to a lack of heme relaxation after disruption of the bond or to the configuration of the dissociated CO. In view of the faster rebinding in other heme-CO systems (see above) we do not favor the former possibility. Instead we suggest that the dissociated CO is initially ejected into a site from which rebinding is hindered. For instance, the CO angle may not be perpendicular to the heme, a prerequisite for binding. We note that for NO and O 2 that both rebind predominantly within a few picoseconds (20,21), the bent bound configuration allows for greater freedom of orientation. Subsequently, CO explores the heme pocket in an essentially barrierless manner and can find a configuration favorable for rebinding, in competition with, thermally activated, escape from the heme pocket.
The (microseconds to seconds) CO rebinding to Mb has been extensively studied at cryogenic temperatures, where it is strongly non-exponential and where population of different protein substates plays a role (41)(42)(43)(44)(45). One result from modeling such kinetics is a significant (ϳ3 kcal/mol, 120 meV) enthalpy for CO binding to the heme from the pocket. By contrast, the corresponding process for NO is barrierless, as inferred from direct assessment of the temperature independence of the ϳ10 ps fastest phase of NO rebinding to Mb (43). Our results imply that, for DosH, at least in the physiological temperature range, enthalpic barriers do not limit CO rebind-ing directly from the heme pocket. The same holds true for rebinding of the methionine residue (Fig. 6, see below). Altogether, in the DosH heme pocket, direct formation of the hemeligand bond from the close-by partners is essentially unrestricted by thermal conformational changes of the heme or the protein environment. Likewise, ligand motions relating to early sensing within the heme pocket occur in this regime.
Methionine Rebinding-In WT Dos, switching from the deoxy state to the oxy state is initiated at the level of the heme by dissociation of the distal methionine residue. Using photodissociation, this process is synchronized in macroscopic samples. In the absence of nearby external ligands to bind to the heme (as in our experiments) the internal methionine residue subsequently rebinds. In WT DosH, this occurs bi-exponentially, with two similar-amplitude phases, one of ϳ7 ps, as in all other studied 6-c heme proteins, and a unique slow 35-ps phase only observed in DosH (20). Our key observation, that in the M95H mutant the slower phase completely disappears, demonstrates that this phase is specific for the presence of a methionine residue. As will be discussed below, we strongly suggest that the flexibility of the Met side chain plays a major role.
A maximum entropy analysis of the kinetics of Met-95 rebinding in WT DosH indicated that the kinetics should be considered as two distinct exponential processes rather than a more continuous distribution of processes (20). Two main kinetic schemes can be put forward to explain a bi-exponential rebinding. The first one is an extension of the equivalent scheme used for the analysis of the CO kinetics (Scheme 1). It is of the form in Scheme 2, in which the initially dissociated Met is close to the heme (state A), and a competition exists between direct rebinding and a conformational rearrangement to a different state (B) for which rebinding to the heme is slower (in Scheme 2 this recombination occurs via state A-direct recombination can also be envisaged). The second possibility is that two distinct configurations are populated directly upon dissociation and do not interchange on the timescale of the experiment.
Variational approaches were attempted to discriminate between these possibilities. First, within the investigated range, the rebinding kinetics was independent of temperature ( Fig.  6A). This implies that, whatever the kinetic model, the rebinding processes (and, in Scheme 2 the conformational interchange) are essentially barrierless (see "Discussion" above), but it does not help to discriminate the models.
The second approach, variation of the viscosity of the aqueous solvent, does alter the kinetics: the relative amplitudes of the phases change, but the rates do not (Fig. 6B). This result is not readily compatible with Scheme 2: solving the kinetic equations associated with this Scheme shows that the relative amplitudes of the phases only change if the rate constants are also altered. However, our observation is in agreement with two distinct configurations, in which the relative populations depend on the viscosity, but the rates do not. Our molecular dynamics simulations point indeed at the existence of two classes of configurations of Met-95, differing in particular with regard to the orientation of the solvent-exposed backbone. Also, our simulations do not predict changes in the population of these configurations induced by dissociation of Met-95 from the heme, as Reaction 2 would predict. We will first briefly compare results with the viscosity effect observed in picosecond ligand rebinding in other systems.
In previous studies on the effect of glycerol on biexponential picosecond NO geminate rebinding with the globin proteins Mb (33) and dehaloperoxidase (32) an increase in the amplitude (but not the rate) of the fast phase of rebinding was observed, but in these cases it was associated with an increase in the rate (on the (10 2 ps) Ϫ1 scale) of the slow phase. These results were interpreted in terms of immediate population of two dissociated NO rotamers, only one of which would directly rebind. Here glycerol, in addition to influencing the initial population distribution, was suggested to lower the barrier of rotamer interconversion, in general agreement with the idea that for NO rebinding barriers exist for the slower phases but not for the fastest (ϳ7 ps) (46). In this light, the lack of dependence of the rates observed in our study on Met rebinding in DosH is consistent with the finding that both processes are barrierless (Fig.  6A). The viscosity effect on the rate in the above globin proteins indicates that the influence of the solvent glycerol would be effective in the heme pocket, where NO is confined. In the case of methionine dynamics in WT DosH, where no rate effect is observed, this influence may be restricted to the protein surface.
Indeed, in the deoxy state, the backbone of the Met-95 residue is exposed to the solvent (Fig. 1). The two configurations identified by our molecular dynamics simulations (Fig. 7) show substantial changes in the backbone position; therefore they are consistent with the experimental finding that the viscosity influences the relative population of the two configurations. Thus a picture emerges where DosH with a 6-c ferrous heme can exist in two configurations that interchange on the time scale of hundreds of picoseconds. In one configuration (presumably configuration I in Fig. 7), thermal dissociation of Met-95 leads to rebinding in ϳ7 ps, as in all other investigated 6-c heme proteins. In configuration II, due to the isomerized Met-95 structure, dissociated Met-95 is on average located in a less favorable position to reform the Fe-S bond, and must sample a larger configurational space to optimize conditions for rebinding, which therefore takes longer (35 ps). Interestingly, indications for relatively large configurational flexibility of the dissociated Met-95 side chain in configuration II are apparent from the large angular fluctuations in Fig. 7C.
Potential Role of the Two Configurations-As schematically illustrated in Fig. 8, we propose that, in the presence of external ligands such as O 2 in the protein moiety, it is from configuration II that ligand replacement will most likely take place, as such ligands will compete most efficiently with Met-95 binding to the heme. Thus, in the presence of external ligands, Met-95 will have a significantly lower probability to recombine with heme on the picosecond timescale. Instead, signal-related rearrangement of Met-95 takes place toward the position where it points out of the heme pocket, as in the x-ray structure of the oxycomplex (8). This rearrangement presumably occurs on a much longer time scale, as for instance the reverse rearrangement after CO dissociation takes ϳ100 s (20). Consequently,  (7)). Thermal dissociation is indicated by red arrows. Interactions between O 2 , Met-95, and heme are indicated by violet curved arrows. Differences in thickness of arrows qualitatively indicate differences in rates. Met-95, His-77, and the solvent-exposed FG loop were depicted using PyMOL. the dissociated configuration II can be seen as a first intermediate in signal transmission within the protein.
In configuration I, the close presence of the Met-95 sulfur atom to the heme iron and the shorter lifetime of the Met-95 dissociated state will hinder efficient competition of binding of external ligands (Fig. 8). Because the distribution between configurations I and II is sensitive to the protein environment (Fig.  6), it is possible that configuration I, which we suggest to be physiologically less active, is less populated in the presence of the enzymatic domain. In addition, we can speculate that additional external factors may influence the distribution between the two configurations and that this may represent a mechanism for the regulation of the effective working range of the sensor.
Altogether, we have shown that the unusually slow phase of Met-95 rebinding to heme in ferrous WT DosH is associated with the flexibility of the Met-95 side chain. A unique configuration can thus be populated from which, after thermal dissociation, Met-95 does rebind in a quasi-barrierless way, while exploring considerable configurational space in the heme pocket. Such a configuration cannot, for instance, be populated by dissociated methionine in the rigid electron-transfer protein cytochrome c (18). For Dos, this situation should allow "switching ligands" to compete in heme binding and to initiate further Met-95 rearrangements along the signaling pathway, involving the flexible FG loop. The movement of the FG loop is thought to be involved in the signaling leading to changes in the activity of the enzymatic domain (47).