Different relaxations in myoglobin after photolysis

Abstract

To clarify the interplay of kinetic hole-burning (KHB), structural relaxation, and ligand migration in myoglobin (Mb), we measured time-resolved absorption spectra in the Soret region after photolysis of carbon monoxide Mb (MbCO) in the temperature interval 120-260 K and in the time window 350 ns to 200 ms. The spectral contributions of both photolyzed (Mb^*) and liganded Mb (MbCO) have been analyzed by taking into account homogeneous bandwidth, coupling to vibrational modes, and static conformational heterogeneity. We succeeded in separating the “time-dependent” spectral changes, and this work provides possibilities to identify the events in the process of ligand rebinding. KHB is dominant at T <190 K in both the Mb^* and the MbCO components. For MbCO, conformational substates interconversion at higher temperatures tends to average out the KHB effect. At 230-260 K, whereas almost no shift is observed in the MbCO spectrum, a shift of the order of ≈80 cm^-1 is observed in Mb^*. We attribute this shift to protein relaxation coupled to ligand migration. The time dependence of the Mb^* spectral shift is interpreted with a model that enables us to calculate the highly nonexponential relaxation kinetics. Fits of stretched exponentials to this kinetics yield Kohlrausch parameter values of 0.25, confirming the analogy between proteins and glasses.

The complexity of proteins is reflected in their heterogeneous structure: they exhibit a distribution of conformational substates (CS) (1, 2). As a consequence, all physical parameters that are determined or influenced by the structure are in general also distributed.

Ligand rebinding to myoglobin (Mb) after flash photolysis has produced a large amount of information about the nature of protein reactions. However, even this apparently simple reaction is surprisingly complicated, and a full understanding still has not been achieved. If CO is the ligand (state A, carbon monoxide Mb, MbCO), photolysis breaks the bond between the heme iron and CO, the protein goes to a metastable intermediate (Mb^*), and CO moves away from the heme. The subsequent rebinding of CO can be followed spectroscopically.

Below ≈180 K the population of CS is frozen: the protein ensemble is distributed with distributed properties. After photolysis, the ligand cannot escape from the heme pocket; it rebinds to the heme iron from a docking site (state B) in a single step (geminate rebinding). From the x-ray structure of the Mb^*+CO photoproduct at low temperatures (3-5) the overall structure of the B state is known: the photolyzed CO lies on top of the heme, parallel to it, and at a distance of ≈4 Å from the iron. The simple B→A reaction is not exponential in time: proteins in different CS have different activation enthalpies for rebinding (1). The reaction can be described by a temperature-independent distribution of enthalpies, g(H), up to ≈180 K (6). Absorption measurements on Mb have demonstrated that its spectral lines are inhomogeneously broadened (7-9). This spectral heterogeneity is coupled to the enthalpic distribution, and kinetic hole-burning (KHB), a spectral change during rebinding, is observed (8, 10-14). Below 100 K KHB has been detected both for Mb^* and MbCO (14).

At physiological temperatures, fluctuations between CS are rapid, and parameters related to processes slower than the rate of fluctuations average out; rebinding becomes exponential. The characterization of the dynamics of protein fluctuations and the relations to the biological function (15) are of utmost importance to a full understanding of their physiological role.

In the 180- to 240-K interval fluctuations between CS gradually set in with increasing temperature, and the transition between the low-temperature behavior and that observed at physiological temperatures occurs. A detailed description of this transition is still lacking, because of its complex nature. Ligand rebinding slows down as the temperature increases (inverse temperature effect), and the rebinding kinetics, while remaining highly nonexponential, present characteristic “bumps” (1, 16, 17). The original interpretation was that rebinding involves several steps and can be described by the scheme: A ⇐ B ⇔ C ⇔ D ⇔ S, where S denotes the state with the CO in the solvent and C and D are intermediate states with the CO in different docking sites within the protein (1). No structural information about states C and D was available at that time, so their existence remained speculative. More recently, x-ray diffraction studies of intermediates trapped at low temperatures after flash photolysis of horse WT MbCO (18) and sperm whale Mb mutants L29W (19) and YQR (20) enabled identification of the C and D intermediates as states in which the CO is trapped within the xenon cavities (21). The photolyzed CO molecule was clearly seen to migrate from the B state to Xe-4 (still in the distal heme pocket) and Xe-1 (in the proximal heme pocket). A fundamental role in determining ligand rebinding kinetics is therefore assigned to ligand migration from the main docking site in the distal heme pocket (state B) to the various xenon cavities and the external solvent (22, 23). In addition to the sequential model, a “side path” model has been proposed (24), in which the proximal docking site Xe-1 is a dead end in the rebinding reaction, and escape to the solvent occurs only from the B state via the histidine gate (25).

Protein relaxation and ligand migration affect not only ligand rebinding kinetics, but also the spectral properties of photolyzed Mb: spectral shifts in both band III (7, 16, 26, 27) and the Soret (13) that cannot be ascribed to KHB have been reported. It was also shown that such spectral shifts highly depend on solvent viscosity (28) and are even prevented from taking place at very high viscosities (29).

With the onset of protein fluctuations a number of events happen: (i) the ligand can migrate from the B docking site to other cavities inside the protein and eventually escape into the solvent; (ii) protein relaxation starts; (iii) KHB disappears; and (iv) the nonexponential kinetics become exponential when CS interconversion becomes faster than rebinding. Although all of the above phenomena require protein fluctuations, their interconnection is not clear. The reactions are not easy to separate either spectrally or kinetically, and there is no general agreement on their relevance to rebinding.

In this work, time-resolved and spectrally resolved optical absorption spectroscopy in the Soret region and in the temperature interval of 120-260 K is applied to study the interplay of KHB, ligand migration, and protein structural relaxation. Previous time-resolved studies of the spectral evolution of Mb after flash photolysis used mainly band III, a small band in the near-IR region observed in the deoxy state only (12, 16, 17, 26, 30, 31). This band, however, enables one to investigate only the unliganded molecules and prevents following the liganded ones. Analogous studies on the much more intense Soret band are complicated by the need of separating the overlapping contributions of deoxygenated and CO ligated proteins. Previous time-resolved studies on native and mutant Mbs in the Soret region were limited to the 250- to 310-K temperature region and used singular value decomposition (28, 29, 32) or polynomial fit (13, 33, 34) to determine the Mb^* peak position as a function of time; no information was obtained on the CO spectral component. The time evolution of the CO component could be particularly useful in that it is expected to be unaffected by ligand migration and/or protein relaxation, so that it can be safely used to investigate KHB and the onset of protein fluctuations.

The Soret band spectral deconvolution developed by Cupane et al. (9) enabled us to accurately measure spectral shifts of both the deoxy and CO components in the same experiment. Hence the onset of fluctuations, responsible for the disappearance of KHB in the Soret band of MbCO and Mb^*, and the time and temperature dependence of ligand migration and protein structural relaxation in Mb^* could be characterized.

Materials and Methods

Sample. Sperm whale Mb (Sigma) from an aqueous solution of 10 mM sodium phosphate, pH 7 was mixed with 75% glycerol to yield a final glycerol concentration of 65% (vol/vol). The sample was then repeatedly saturated with CO and reduced with Nadithionite immediately before the experiment. The final Mb concentration was ≈10 μM, so as to give an absorbance of 0.4 at the maximum of the Soret band before photolysis.

Experimental Apparatus. The sample in a 2-mm-thick sealed cuvette was mounted in a two-stage helium cryo-cooler (Air Products and Chemicals, Allentown, PA). The temperature was maintained constant to within 0.5 K in the 120- to 260-K range by a homemade controller.

Photolysis was achieved by an excimer laser (JATE XEL604, Szeged, Hungary) pumped dye laser (540 nm, 20 ns, ≈1 mJ) at sufficiently low frequencies (0.8-1.5 Hz, depending on the temperature) to allow virtually full rebinding between consecutive flashes. The actinic and measuring lights were quasicolinear, and the spot size of ≈2-mm diameter resulted in ≈2.7 × 10¹⁵ incident photons in the active volume containing ≈3.6 × 10¹³ hemes. With the absorbance at 540 nm being 0.028 this meant complete photolysis without photoselection, so the kinetics of ligand rebinding were not expected to be disturbed by tumbling of unphotolyzed molecules. The white continuum measuring light was dispersed by a Jobin Yvon (Longjumeau, France) HR320 spectrograph (spectral resolution 0.2 nm, corresponding to 11 cm^-1 at 425 nm), and detected by an optical multichannel analyzer consisting of a gated intensified photodiode array (IRY-512) a gate pulse generator (PG-10), and a detector controller (ST-120) from Princeton Instruments (Trenton, NJ). A 60-nm spectral range covering the Soret band was imaged on the 512-element diode array. Difference spectra were collected by accumulating 100 scans per spectrum at 10 time delays per decade in a logarithmically equidistant fashion from 350 ns to 200 ms. In a control experiment the different delays were applied in a random order, but the time dependence of the signal amplitude remained the same. This process rules out any systematic pumping effect of the actinic laser.

Singular value decomposition of the matrices of consecutive difference spectra at selected temperatures yielded orthonormal spectral and kinetic eigenvectors and corresponding singular values. Comparison of the singular values with the noise of the data and analysis of the autocorrelation of the eigenvectors allow estimating the rank of the data matrices. The gradual small shift of the positive and negative bands in the difference spectra, caused by KHB and/or protein relaxation, would theoretically render the rank of the data matrix equal to the number of columns. However, the magnitude of the shift is small enough to permit reconstruction of the data within noise by using the first two or three singular value decomposition components only. These noise-filtered spectra were used in the analysis.

Spectral Analysis. The output is ΔA(v,t), i.e., the absorption difference spectrum between the sample at time t after photolysis and the sample before photolysis: ΔA(v,t) = A(v,t) - A_CO(v). Fig. 1a shows difference spectra measured at various times after photolysis at T = 120 K. Time-dependent spectral variations of both the deoxy (positive peak) and CO (negative peak) components are seen already from the raw data as evidenced by the contour plots in Fig. 1 b and c. The absorption spectrum of the sample at time t after photolysis, A(v,t), which can be obtained by adding the absorption spectrum A_CO(v) of the unphotolyzed sample to ΔA(v,t),¶ is a mixture arising from both the liganded and unliganded species: A(v,t) = A_CO(v,t) + A_deoxy(v,t), where A_deoxy(v,t) and A_CO(v,t) are the spectral contributions of proteins that have not yet rebound or have already rebound the CO ligand, respectively.

Fig. 1.

Typical experimental data. (a) Time-resolved absorption difference spectra measured after photodissociation of sperm whale MbCO in 65% glycerol/water solution at 120 K; positive signals arise from the deoxy component and negative signals from the CO component. (b and c) To evidence the time evolution of the peak and the width, contour plots of spectra normalized to unit peak height are reported around the maximum (0.995) and at half-maximum for the deoxy component (b) and the CO component (c).

To fit the “reconstructed” spectra, A(v,t), we use the convolution of three functions (9): , where α = CO, deoxy. M_α is proportional to the matrix element of the electric dipole moment. The superposition of Lorentzians L_α(v) accounts for high frequency (hv_αh >> kT) normal modes and the Gaussian G_α(v) for low frequency (hv_αl ≤ kT) normal modes, respectively, vibronically coupled to the electronic transition. P_α(v) describes the spectral heterogeneity caused by CS, as reflected in the different frequencies of the electronic transition. The expression for P_deoxy(v) was first introduced by Champion and coworkers (7), whereas P_CO(v) is a further Gaussian that does not alter the overall shape of the curve, but simply adds a constant term to the Gaussian width of the Soret band.

The fitting procedure optimizes the following parameters: integrated absorbancies (M_CO, M_deoxy), peak frequencies (v_CO, v_deoxy), and Gaussian widths (σ_CO, σ_deoxy).

The other parameters have been taken from previous works (35, 36) and have been kept constant in the fittings. This choice is justified by the fact that spectral variations involve essentially intensity variations, peak shifts, and band broadenings; it is also justified “a posteriori” by the excellent quality of the fits. In fact, residuals were always <1%.

Results and Discussion

The fraction of Mb molecules that have not rebound the CO ligand at time t after photolysis, N(t), has been computed by dividing the absorbance change at 432 nm (≈23,150 cm^-1), measured at the various temperatures and times after photolysis, by ΔA_max, i.e., the absorbance change at the same wavelength, assumed to be temperature independent, at T = 120 K and t = 350 ns (Fig. 2). Indeed, at 120 K and 350 ns after photolysis virtually no CO molecules had yet rebound to the heme iron, so that N ≈ 1; moreover, at 432 nm the signal-to-noise ratio is optimal and ΔA_max, measured in static spectra, is almost temperature independent (35).

Fig. 2.

Rebinding kinetics of sperm whale MbCO in 65% glycerol/water solution. N(t) is the fraction of proteins that do not have a ligand bound at time t after photodissociation.

Peak frequency values of the CO and deoxy components (v_CO and v_deoxy), at selected temperatures, are reported as a function of time in Fig. 3. A blue shift of both components is observed; the effect is larger for the deoxy component and is almost vanishing for the CO component at higher temperatures. The rather large “noise” observed at short times for v_CO and at long times for v_deoxy is caused by the small amplitude of the corresponding spectral components, which makes the determination of the peak frequencies subject to rather large errors. Peak frequency shifts for the deoxy component measured at 260 K can be compared with analogous data (13) obtained at 250 K and in 75% glycerol/water. Although the overall time dependence looks very similar, a larger spectral shift is systematically observed in the present experiment [e.g., at 350 ns we measured v_deoxy(t) - v_deoxy(∞) ≈ -90 cm^-1 as compared to -60 cm^-1 reported in ref. 13]. The discrepancy can be attributed to the better spectral resolution of our experiment (0.2 nm as compared to 1 nm) or the different spectral analysis (the polynomial fit used in ref. 13 underestimates the spectral shift). At low temperatures (T ≤ 190 K) the Gaussian width of the CO component increases with time (at 120 K σ_CO is 138 cm^-1 at 10 μs and 146 cm^-1 at 10 ms), whereas σ_deoxy decreases (from 105 cm^-1 at 1 μs to 92.5 cm^-1 at 5 ms). At higher temperatures both quantities remain constant.

Fig. 3.

Time evolution at selected temperatures of the Soret band peak frequencies after photolysis of MbCO. (a) CO component. (b) Deoxy component. Typical error bars are indicated. Symbols are as in Fig. 2.

The blue shift and line narrowing of the deoxy component are in agreement with results at low temperature on both band III (37) and the Soret band (12). This behavior has been attributed to KHB: Mb molecules that contribute to the red side of the absorption spectrum rebind faster than those that contribute to the blue side; as the rebinding proceeds, holes are progressively “burned” from the red side and the band shifts to the blue while narrowing and decreasing in intensity. The behavior of the CO component of the Soret band confirms this interpretation. At t = 0⁺, all Mb molecules are unliganded and have a deoxy-like Soret band; as the rebinding proceeds, the holes burned from the red side of the deoxy band “fill” the red side of the MbCO Soret region first, resulting in a CO component of increasing amplitude that shifts to the blue and broadens.

We define the spectral shifts as Δv_deoxy(t) = v_deoxy(t) - v_deoxy(∞) and Δv_CO(t) = v_CO(t) - v_CO(∞). The reference value v_CO(∞) can be measured directly from the spectrum at the longest delay from photolysis, i.e., 20-200 ms depending on the temperature. The temperature dependence of v_CO(∞) measured in the present experiment closely parallels the analogous behavior obtained from static measurements on equilibrium MbCO (35). The reference value v_deoxy(∞) cannot be measured directly from the spectra because at long times the deoxy component becomes vanishingly small. It has been obtained by assuming that at all temperatures v_deoxy(∞) = v_deoxy(equilibrium), i.e., equal to the value measured in equilibrium measurements on ligand-free Mb (deoxy-Mb) at the same temperature. Values of v_deoxy(equilibrium) have been taken from ref. 35, after suitable calibration of the different spectrometers used.

KHB is evidenced by the “universal” plot of the peak frequency shifts as a function of the fraction of still unliganded Mb molecules after photolysis (11). In the absence of relaxation phenomena and if only KHB is present, the spectral shifts should be related only to the fraction of molecules that have not rebound the ligand, and all data points should fall on the same universal curve of Δv vs. N(t), irrespective of the temperature. Such behavior is to be expected in both deoxy and CO spectral components.

Fig. 4 shows universal plots for the CO and deoxy peak frequency shifts. Below ≈200 K a temperature-independent behavior, within the experimental error, is observed (Fig. 4 a and c). This finding is a clear indication that all of the shifts observed are consistent with a pure KHB effect and therefore fluctuations between the relevant CS are not present. The observation of KHB for both the deoxy and CO components implies a strict correlation between the CS of deoxy and CO Mb. A correlation between the A and B taxonomic substates has already been evidenced (38); our data indicate that this correlation is also extended to the statistical CS responsible for KHB.

Fig. 4.

Universal plot at various temperatures for the CO component (a and b) and the deoxy component (c and d). (a and c) Temperature interval 120 -200 K. (b and d) Temperature interval 220 -260 K; data relative to 120 and 140 K are also reported in b and d, for comparison. The continuous line in c is a fit to the data in the temperature interval 120 -180 K and represents the pure KHB behavior. Typical error bars are shown. Symbols are as in Fig. 2.

As the temperature is raised well above the solvent glass transition temperature (≈180 K), both Δv_CO and Δv_deoxy start to deviate from the universal behavior (Fig. 4 b and d), but in a profoundly different manner. Indeed, the deviation from the universal behavior observed in the Δv_CO vs. N(t) plot (Fig. 4b) is a direct consequence of the onset of fluctuations between CS. At 260 K and N ≈ 0.4 (time delay ≈ 1 μs) a shift <5 cm^-1 is observed in the CO component, indicating that in the microsecond time scale fluctuations between the relevant CS are able to average out almost every spectral shift caused by KHB. One should relate the observations to the taxonomic A and B states. In MbCO (the A states) the KHB plot (Fig. 4b) shows fluctuational averaging above 220 K. Comparing the data to the pressure jump relaxation experiments (39), we see that A states interconversion is in general slower than the averaging seen here (although they overlap because of the broad distributions and the different rates for different A states). Consequently, the present observations represent relaxations primarily within the A substates.

The behavior of Δv_deoxy is much more complex: even at the disappearance of KHB the deoxy component remains substantially red-shifted with respect to equilibrium deoxy-Mb (e.g., at 260 K and N ≈ 0.4 the magnitude of the red shift is still ≈80 cm^-1). Because, as shown by the behavior of Δv_CO, Mb is able to fluctuate among the relevant CS at temperatures above ≈200 K, we can safely conclude that the observed residual shift is related to the structural differences between the Mb^*+CO photoproduct and equilibrium deoxy-Mb that, therefore, can be monitored by its time evolution.

Analogous shifts of both band III and the Soret band of the photoproduct have been studied by several authors, and their structural origin is the subject of a long-standing debate. Srajer and Champion (12) and Ahmed et al. (40), based mainly on the similarity with the time dependence of the iron-proximal histidine frequency measured in resonance Raman experiments, attributed the observed shifts to proximal effects. The iron-porphryrin distance, also linked to other proximal coordinates such as the tilt and azimuthal angles of the iron-proximal histidine linkage, was considered as the relevant conformational coordinate. This view was supported by room temperature band III studies by Anfinrud and coworkers (31, 41) in the time range from 2 ps to 56 μs.

Lambright et al. (33) and Franzen and Boxer (13), by examining the Soret band shifts in photoproducts of both proximal and distal Mb mutants at nearly physiological temperatures, suggested that in the nanosecond-to-millisecond time range proximal contributions are negligible; they attributed the observed shifts to distal effects, likely related to the electrostatic field around the heme. In particular, studies on the proximal mutant H93G indicated that band III and the Soret band shifts are attributable to motions of amino acid residues in the distal pocket after CO photolysis. Recently, Nienhaus et al. (26) have studied the effect of ligand dynamics on band III through temperature-derivative spectroscopy between 10 and 100 K. They attribute the strong KHB to the interaction of the heme group with the photolyzed CO molecule; moreover, they show that migration of CO away from the B site causes a substantial blue shift of band III and therefore they assign the slow phase of the band III shift to ligand migration out of the B docking site and to structural relaxations around the heme group. The relevance of the electrostatic environment of the heme group (in connection with the presence of a water molecule in the distal heme pocket and with the position of the distal histidine H64) in affecting the peak frequency of the heme electronic absorption bands has also been suggested (ref. 42 and J. Berendzen, personal communication). Shifts in the position of the distal histidine, in connection with CO migration out of the B docking site, are also seen in nanosecond time-resolved crystallographic studies (ref. 43 and M. Wulff, personal communication).

Our data, although clearly showing the presence of a spectral relaxation different from KHB in the photoproduct, do not provide direct evidence about the structural origin of the observed shifts. We note, however, that while we see dynamic hole filling (DHF) in MbCO (i.e., the dynamic disappearance of KHB at ≈220 K caused by CS interconversion, see Fig. 4b), we do not see DHF in the photoproduct [i.e., a spectral red shift because of recovery of the initial CS distribution, as demonstrated by time-resolved studies on band III in hemoglobin (17)]. This observation could be caused by the fact that CS interconversion and DHF occur, in the photoproduct, on a time scale faster than our temporal resolution (unlikely, in view of the MbCO data) or by the fact that, as fluctuations set in, the deoxy proteins can “relax” to a set of CS other than those giving rise to KHB and in which the interaction between heme and CO is weakened. Thus, we favor an interpretation in terms of a protein relaxation leading to ligand migration and an altered electrostatic environment in the distal heme pocket.

We therefore divide the N(t) fraction into a subset, γ(t), of unrelaxed molecules in which the CO ligand is still in the main docking site, and a subset, η(t), of relaxed molecules in which the CO has migrated to other Xe cavities. At low temperature, where no ligand migration effect is present, η(t) = 0 and N(t) = γ(t). We postulate also that, while the spectrum of the γ(t) population shifts according to the universal behavior, the spectrum of the relaxed η(t) population is at the equilibrium deoxy-Mb position.

From the above model it follows that:

1

As already stated, v_deoxy(∞) is the peak frequency measured for equilibrium deoxy-Mb, and Δv_KHB(N) is the temperature-independent peak frequency shift observed at low temperature in the deoxy component after photolysis. We stress that, in view of our definition of spectral shifts, Δv_KHB is not the spectral shift caused by KHB, but rather the overall spectral shift of the deoxy component relative to its equilibrium position measured at temperatures where only KHB is present.

Rearranging Eq. 1 for the peak frequency shift:

2

Eq. 2 shows that, according to our model, it is possible to evaluate the fraction Φ(t) ≡ γ(t)/N(t) = Δv_deoxy(t)/Δv_KHB(N) of unliganded and unrelaxed Mb molecules in which the CO ligand is still in the main docking site in the distal heme pocket with respect to the number of Mb molecules that have not rebound a CO molecule at time t from photolysis. At each temperature Φ(0) = 1 because at t = 0 no protein has relaxed. At temperatures where fluctuations are effective Φ(∞) = 0, i.e., all proteins have relaxed and CO molecules have migrated to some Xe cavity or the solvent. In Fig. 5 we plot Φ(t) as a function of time above 210 K. Δv_KHB at all N(t) values has been obtained from a polynomial fit of all data shown in Fig. 4c. It is evident that Φ(t) is not a single exponential; this clearly shows that protein relaxation and ligand migration are characterized by a distribution of activation enthalpies. A power law, (1 + κt)^-η, did not fit the data; reasonable fits were obtained with stretched exponentials exp[-(kt)^β] (dashed lines in Fig. 5). Rate constants of 100, 8.5 × 10³, and 1.26 × 10⁵ s^-1 were obtained at 220, 240, and 260 K, respectively. Values of the Kohlrausch parameter β were 0.25 ± 0.02, independent of temperature; low β values are typical of glassy systems and confirm the already suggested (44) analogy between protein and glasses.

Fig. 5.

Fraction of unrelaxed proteins as a function of time after photolysis on a log-log scale. The continuous lines are fits in terms of stretched exponentials. Symbols are as in Fig. 2.

Results in Fig. 5 can be compared with a time-resolved crystallographic study (43), which accounts for the location of all CO molecules ≈100 ns after photolysis of MbCO at room temperature; 11% occupy the main docking site in the distal heme pocket, 26% have migrated to the proximal Xe-1 site, 63% have rebound geminately (or were never dissociated, in their experiment), and 0% have escaped to the solvent. In our notation this means Φ (≈100 ns) = 0.3 at room temperature. From Fig. 5 we obtain Φ (≈350 ns) = 0.55 ± 0.05 at T = 260 K, in qualitative agreement with the x-ray data.

Conclusions

1. During ligand rebinding to Mb time-dependent spectral shifts in the Soret region are observed; with the deconvolution procedure developed in our laboratory the shifts of the deoxy and CO components can be resolved and studied separately.

2. Below ≈200 K, KHB is solely responsible for the shifts of both the CO and deoxy components that are characterized by a universal behavior, reported here for MbCO above 100 K. This finding suggests that the correlation between B and A substates already observed at the level of taxonomic substates of tier 0 must be extended to the level of statistical substates.

3. Below ≈200 K, as rebinding goes to completion and N(t) → 0, the MbCO spectrum tends to the equilibrium one and Δv_CO → 0; this finding shows that, as expected, MbCO after photolysis is not structurally different from equilibrium MbCO and that at full rebinding the entire equilibrium distribution of statistical substates is recovered. On the other hand, the spectrum of Mb^* remains distinctly red-shifted with respect to equilibrium Mb.

4. Above 190 K KHB for MbCO starts to disappear and at ≈230 K virtually no spectral shift is observed within our time window: interconversion between the relevant A statistical substates sets in and averages out the spectral shifts because of KHB.

5. Above 190 K, new spectral shifts appear for Mb^*: the heme pocket becomes mobile, the tertiary structure of the protein relaxes, and Mb^* tends to equilibrium Mb. Spectral red shifts caused by dynamic hole filling are not observed for the photo-product, which means that in the relaxed state the deoxy proteins experience a set of substates that do not give rise to KHB. The new spectral shifts can be tentatively attributed to ligand migration linked to protein relaxation and alterations of the electrostatic environment of the heme.

6. A simple model can relate the observed spectral shifts to the fraction of unrelaxed proteins in which the photolyzed CO molecule, at a given time and temperature, is still in the main docking site in the distal heme pocket. Our data are in qualitative agreement with time-resolved crystallographic data.

7. The observed relaxation is a nonexponential process whose time dependence can be described by a stretched exponential with a Kohlrausch parameter β = 0.25.