|Home | About | Journals | Submit | Contact Us | Français|
Despite close structural similarity, the ferric and ferrous forms of cytochrome c (cyt c) differ greatly in terms of their ligand binding properties, stability, folding and dynamics. The reduced heme iron binds diatomic ligands such as CO only under destabilizing conditions that promote weakening or disruption of the native methionine-iron linkage. This makes CO a useful conformational probe for detecting partially structured states that cannot be observed in the absence of endogenous ligands. Heme absorbance, circular dichroism and NMR were used to characterize the denaturant-induced unfolding equilibrium of Fe2+ cyt c in the presence and absence of CO. In addition to the native state (N), which does not bind CO, and the unfolded CO-complex (U-CO), a structurally distinct CO-bound form (M-CO) accumulates to high levels (~75% of the population) at intermediate guanidine hydrochloride concentrations. Comparison of the unfolding transition for different conformational probes reveals that M-CO is a compact state containing a native-like helical core and regions of local disorder in the segment containing the native Met80 ligand and adjacent loops. Kinetic measurements of CO binding and dissociation under native, partially denaturing and fully unfolded conditions indicate that a state, M, that is structurally analogous to M-CO is populated even in the absence of CO. The binding energy of the CO ligand lowers the free energy of this high-energy state to such an extent that it accumulates even under mildly denaturing equilibrium conditions. The thermodynamic and kinetic parameters obtained in this study provide a fully self-consistent description of the linked unfolding/CO-binding equilibria of reduced cyt c.
Metalloproteins often depend on their cofactor not only for function, but also for efficient folding and stabilization of their native structure. For example, in the absence of heme, myoglobin folds into a marginally stable state lacking some of the native α-helices and tertiary interactions,1 and cytochrome b562 assumes a dynamic, molten globule-like, conformation2. An even more extreme case is cytochrome c (cyt c), which is initially synthesized as a largely disordered apoprotein3,4 and folds into a stable globular structure only after covalent attachment of the heme.5,6 These observations are not inconsistent with the notion that the native structure of a protein is encoded in its sequence, but suggest that the cofactor carries some of the information defining the native structure.
It has long been known that the heme and its axial ligands has a profound influence on the folding process of horse cyt c.7–17 Similar findings have been reported for c-type cytochromes from other species, including yeast and photosynthetic bacteria,18,19 but we focus here on horse cyt c, which has been studied especially thoroughly. Under typical denaturing conditions (e.g., 6 M GuHCl, pH >4), one of the two axial heme ligands, the imidazole nitrogen of His18, remains bound to the heme iron due to the fact that the adjacent Cys17 is covalently bound to the heme. However, coordination of the second axial ligand, the sulfur of Met80, is inherently less stable and readily dissociates under partly or fully denaturing conditions. In unfolded cyt c, the vacant heme coordination site can bind alternative ligands, including extraneous ligands, such as imidazole or, in the case of the reduced (Fe2+) form, carbon monoxide, as well as intramolecular ligands, such as His, Lys or the amino terminus in their deprotonated states. Detailed studies of such ligand exchange reactions in both iron oxidation states have provided a rich source of information on the conformational propensities and dynamics of the denatured state of cyt c.12,20–27 The predominant 6th iron ligand in the unfolded state of oxidized (Fe3+) cyt c is His33,28 which can become trapped during refolding and leads to accumulation of relatively long-lived (~100 ms) intermediate states that feature both native-like as well as non-native structural features.10,13,14
Although the structures of oxidized and reduced forms of mitochondrial cyt c, as determined by X-ray crystallography29–31 or NMR32–35 are very similar, the two forms differ greatly in terms of stability, dynamics and folding kinetics.12,36–41 The dramatic stabilization of the protein upon reduction of the heme is not fully understood, but appears to be due to a combination of electrostatic effects (the reduced heme is electrically neutral while the oxidized heme carries a net charge of +1), differential affinity of the heme iron for the axial ligands between the native and unfolded conformations in both oxidation states, and dynamic/entropic contributions. The fact that only non-native forms of Fe2+ cyt c bind a CO ligand with high affinity to form a photolabile ligand complex has opened unique opportunities for manipulating the conformational transitions of the protein and probing its dynamics on the µs to ms time scale.12,20,21,23,25,27,42,43 In their initial equilibrium characterization of Fe2+ cyt c in the presence of CO, Jones et al.12 already noticed that this system does not undergo a simple two-state unfolding transition. When observing the GuHCl-induced unfolding transition for Fe2+ cyt c at 40 °C using tryptophan fluorescence, they measured a midpoint concentration, Cm = 3.7 M (m = 2.6 kcal mol−1 M−1) in the presence of CO (1 atm, corresponding to ~1 mM CO in solution) and Cm = 5.1 M (m = 3.6 kcal mol−1 M−1) in the absence of CO. These parameters indicate that addition of CO results in a nearly 9 kcal mol−1 decrease the stability of native Fe2+ cyt c. A similar unfolding transition was observed using far-UV CD spectroscopy. However, changes in heme absorbance indicative of CO binding were observed already between 1 and 3 M GuHCl where the fluorescence and far-UV CD signals remain at their native levels.12 This is a clear indication that the unfolding transition in the presence of CO cannot be adequately described in terms of a two-state mechanism.
Despite this earlier evidence for non-cooperative behavior, Bhuyan and colleagues concluded that reduced cyt c undergoes a two-state unfolding transition, both in the presence and absence of CO, based on their equilibrium and stopped-flow studies of folding and unfolding.40,41,44,45 In order to resolve this controversy, we further characterized unfolding equilibrium of Fe2+ cyt c in the presence and absence of CO using optical techniques and NMR. The results provide clear evidence that, in addition to the native state (N), which does not bind CO, and the unfolded CO-complex (U-CO), a structurally distinct CO-bound form (M-CO) accumulates at intermediate denaturant concentrations. Based on its optical and NMR properties, M-CO is a folded state with a native-like helical core and regions of local disorder in the segment containing the native Met80 ligand and adjacent loops. Kinetic measurements of CO binding and dissociation under native and denaturing conditions confirm our hypothesis that a state, M, which is structurally analogous to M-CO, is populated even in the absence of CO. Even though M does not accumulate to detectable levels at equilibrium, its presence can limit the rate constant of unfolding, giving rise to a downward curvature in the log(rate) vs. [denaturant] plot at high denaturant concentration similar to that previously observed for oxidized c-type cytochromes.13,18,46 The binding energy of the CO ligand lowers the free energy of this transient intermediate to such an extent that it accumulates even under mildly denaturing equilibrium conditions. The thermodynamic and kinetic parameters obtained in this study provide a fully self-consistent description of the linked unfolding/CO-binding equilibria of reduced cyt c.
The changes in heme coordination that accompany unfolding and CO binding give rise to major changes in the heme absorption spectrum of cyt c.12,41,47 Figure 1 shows absorption spectra of horse Fe2+ cyt c in the folded and the GuHCl-unfolded forms. Addition of CO to unfolded cyt c (U) results in a large increase in the extinction coefficient for Soret absorption consistent with the formation of a hexacoordinate low-spin complex (U-CO) with CO displacing the native Met80 sulfur at the 6th iron coordination site (because of the covalent linkage of the heme via Cys14 and Cys17, the native His18 ligand remains bound to the 5th coordination site under the denaturing conditions used here7). To measure the effect of CO on the unfolding equilibrium, we recorded absorbance spectra on a series of Fe2+ cyt c samples at different GuHCl concentrations equilibrated under 1 atm CO (~1 mM). Figure 2(a) shows the absorbance changes at selected wavelengths in the Soret region as a function of GuHCl concentration. At certain wavelengths (e.g., 410 nm) gradual changes in absorbance occur already at GuHCl concentrations between 1 and 2.7 M while a single steeper transition centered around 3.6 M GuHCl is observed at other wavelengths (most pronounced at 418.5 nm). The two transitions are especially well resolved in the curve monitored at 416.5 nm, which shows two resolved phases with opposite sign resulting in a bell-shaped curve. These observations provide clear evidence for the presence of an intermediate state with absorbance properties distinct from the equilibrium states populated at low (N) and high (U-CO) denaturant concentrations. The solid lines in Figure 2(a) were obtained by fitting a three-state mechanism (Scheme 1) to the transition curves measured at 0.5 nm intervals over the range from 380 to 430 nm, using global fitting procedures described by Latypov et al.48 The equation used to describe Scheme 1 for the global analysis involves a total of 8 parameters; the global parameters, Cm1, m1, Cm2 and m2 listed in Table 1 describe the two coupled denaturant-induced unfolding transitions, and the local parameters εN, εM-CO, εU-CO and sU describe the wavelength-dependence of the data. A slope sU = d(εU-CO)/dc was included to account for the effect of GuHCl on the absorbance of the fully unfolded CO complex. The spectral parameters, εN, εM-CO and εU-CO, plotted in Figure 2(b) vs. wavelength, represent the intrinsic absorption spectra of the three conformational states. εN is identical to the spectrum of the native state while εU-CO represents the spectrum of U-CO extrapolated to 0 M GuHCl, which is similar to the spectrum of U-CO measured at high GuHCl concentrations (cf. Figure 1). The Soret band of the M-CO state is more intense than both N and U-CO, and its absorption maximum (λmax = 413.5 nm) falls in between those of N (414.5 nm) and U-CO (412.5 nm).
The GuHCl-induced absorbance changes in the visible region of the spectrum (500–570 nm) were analyzed using the same procedure (see Supplementary Material). The native absorbance bands at 520 and 550 nm disappear gradually from 1 to 3.5 M GuHCl reflecting primarily the transition from N to M-CO. The spectrum of M-CO is close to that of U-CO with a minor contribution of a native-like spectrum (cf. Figure 1). We also used far-UV CD spectroscopy to monitor the denaturant-induced loss in native helical secondary structure of reduced cyt c samples in the presence of CO. A plot of the raw CD data (ellipticity at 225 nm) vs. GuHCl concentration can be found in Supplementary Material. With increasing GuHCl concentration, the CD signal increases (becomes less negative) in two distinct steps, including a shoulder around 2.5 M accounting for ~20% of the total change and a sharp increase centered at 3.6 M GuHCl (in addition, there is a ~5% linear decrease between 0 and 1.5 M GuHCl). Initially, we used the equilibrium parameters (Cm and m-values) obtained by global fitting of the Soret absorbance data (Figure 2) as a constraint in fitting the visible heme absorption bands and ellipticity at 225 nm. The quality of the fits obtained (Supplemental Material) demonstrates the self-consistency of the data. The final set of thermodynamic parameters listed in Table 1 were obtained by globally fitting all of the available data, including the full absorbance spectra in the Soret region (380–430 nm), the visible bands at 520 and 550 nm, and the CD signal at 225 nm.
To compare the unfolding transitions monitored by different optical probes, we calculated normalized curves using the fitted intercepts and slopes of the native and U-CO states for each individual curve (Figure 3(a)). The dispersion among the different curves reflects the fact that the relative contribution of the M-CO intermediate to the equilibrium transition is different for different optical parameters. For example, the absorbance at 412.5 nm (green squares) corresponds to an isosbestic point between the M-CO and U-CO states (Figure 2(b)), and thus measures the gradual loss of the native population with increasing GuHCl concentration. In contrast, the curves at 406.5 (green up triangles) and 418.5 nm (not shown) coincide with isosbestic points between N and M-CO (Figure 2(b)), and thus reflect the population of the U-CO state, which builds up in a sharp transition centered at 3.6 M GuHCl. The ellipticity at 225 nm (diamonds) is dominated by a major transition at 3.6 M GuHCl with only minor changes (~20%) below 3 M, indicating that the M-CO state retains much of the native secondary structure consistent with earlier reports.12,41 In Figure 3(b) the populations of the three states derived from the global equilibrium parameters (Table 1) are plotted vs. GuHCl concentration. M-CO reaches a maximum population of 73% at 3.3 M GuHCl. Also shown in Figure 3(a) are normalized unfolding curves for reduced cyt c in the absence of CO measured in the Soret and visible regions of the absorption spectrum, as well as the CD signal at 225 nm. The fact that all optical probes can be fitted using a two-state model and yield nearly identical equilibrium parameters confirms that unfolding of reduced cyt c is a cooperative process.37,40 Global fitting of a two-state model to the optically monitored unfolding transitions of Fe2+ cyt c at 20 °C (in 0.1 M sodium phosphate, pH 7.0) yields an average midpoint Cm = 5.22 ± 0.06 M and m-value mNU = 3.25 ± 0.24 kcal mol−1M−1, corresponding to a free energy ΔGNU = 17.0 ± 1.3 kcal mol−1. Similar unfolding free energies (18.0–19.3 kcal mol−1) have been reported previously for reduced cyt c under similar conditions.49,50
For a more detailed structural characterization of the unfolding equilibrium, we recorded a series of 1D 1H NMR spectra as a function of GuHCl or urea concentration on Fe2+ cyt c samples before and after addition of CO (in D2O with amide groups fully deuterated). Figure 4 shows representative spectra obtained vs. urea concentration after equilibration with 1 atm CO. The resonances assigned to the Met80 side chain (Figure 4(b)) begin to lose intensity at lower denaturant concentrations (7–8.5 M urea) than the methyl resonances assigned to Leu32 and Ile57; the latter persist to higher denaturant concentrations before disappearing over a narrow range of urea concentrations centered around 8 M (similar behavior is seen for a number of other resolved resonances). These spectral changes clearly indicate that the native Met80-Fe coordination is disrupted at relatively low denaturant concentrations, where the M-CO state is populated, while other regions of the protein become unfolded only at higher urea concentrations. Figure 4(a) shows the low-field region of the 1H NMR spectrum, which contains resonances assigned to the four meso protons of the heme.51 Although unambiguous assignments are available only for the native state, the complex urea-dependent spectral changes are clearly inconsistent with a cooperative (two-state) equilibrium. For example, the resonance assigned to the δ-meso proton in the N-state disappears at lower urea concentrations than the α-meso resonance, indicating that the former tracks the population of the N-state while the latter reflects the combined populations of the N- and M-CO states. Another resonance near 9.6 ppm appears only at intermediate urea concentrations and shows the bell-shaped profile expected for the population of the M-CO state. Finally, two groups of resonances near 9.8 and 10.1 ppm, respectively, gain intensity at urea concentrations above 7 M, and can thus be assigned to the fully unfolded state. For resolved resonances, normalized peak intensities were obtained by taking the ratio of the peak area between the spectra measured in the presence and absence of CO, respectively (the latter shows no evidence for urea-induced unfolding up to the highest urea concentration measured; data not shown). A plot of relative peak intensities vs. urea concentration (Figure 5(a)) shows the behavior expected for a three-state unfolding equilibrium, including peaks assigned to the native state that disappear at relatively low urea concentration (tracking the N-state; red symbols), others that persist up to higher denaturant concentration (tracking the combined population of N and M-CO; yellow symbols), a third group that follows the population of the M-CO intermediate (green symbols), and a fourth group of lines that appear only at high denaturant concentrations, where the U-CO state accumulates (blue symbols). The fact that the Met80 CεH3 resonance is among the first group decaying at low urea concentration confirms that the native Met80 ligand is displaced by the CO ligand. A corresponding set of 1H NMR spectra were obtained for Fe2+ cyt c in the presence and absence of CO as a function of GuHCl concentration (data not shown for brevity; cf. ref. 37). The relative peak intensities of resolved N-state peaks vs. GuHCl concentration show again behavior characteristic of a three-state unfolding equilibrium (red and yellow symbols in Figure 5(b)). Global analysis of the NMR data using Scheme 1 yields equilibrium parameters similar to those obtained by CD and absorbance (cf. Table 1). In contrast to urea, GuHCl is a sufficiently strong denaturant to reach the fully unfolded state even in the absence of CO (cf. Figure 3). All of the resolved resonances in reduced cyt c decrease over a narrow range of GuDCl concentrations near 5 M (gray symbols in Figure 5(b)) consistent with a fully cooperative (two-state) unfolding transition.37,40
Although the CO ligand does not directly bind to the native state of reduced cyt c, a metastable CO complex is known to exist under native conditions.25,37,41 To prepare this state, we initially equilibrate Fe2+ cyt c under 1 atm CO in the presence of 6 M GuHCl, forming the U-CO complex. The denaturant is then removed by gel filtration, or its concentration is lowered by dilution, which allows the protein to refold while the CO molecule remains tightly bound to the heme, preventing formation of the Met80 sulfur-Fe2+ bond. In the absence of light, the CO complex thus obtained, M-COms, is metastable and corresponds to the M-CO equilibrium intermediate in the absence of denaturant. Thermal dissociation of the CO ligand occurs on a time scale of hours or longer at or below room temperature.25,41 At 15 °C the lifetime of the CO complex is about 4 h, which is sufficient to record a 2D NMR spectrum on M-CO with minimal thermal dissociation of the ligand.
A uniformly 15N-labeled recombinant H33N variant of horse cyt c48,52–54 was initially prepared in the U-CO form by equilibrating a 1 mM solution of dithinoite-reduced protein under 1 atm CO in the presence of 6 M GuHCl (in 0.1 M phosphate buffer at pH 7). The denaturant was removed at 4 °C by passing the sample over a spinning gel filtration column (Sephadex G25) equilibrated with buffer (0.1 M sodium phosphate in 95% H2O / 5% D2O, pH 7.0). A 15N-1H heteronuclear single quantum correlation (HSQC) spectrum was recorded at 15 °C. The spectrum of the M-COms state thus obtained is shown in Figure 6 (green contours) along with that of the N-state (red contours) recorded on the same sample after dissociation of the CO ligand (~8 h incubation at 15 °C). The spectrum of the refolded sample is virtually identical to that of a fresh Fe2+ cyt c sample that has never been unfolded and treated with CO. Backbone resonance assignments for the native reduced form were obtained using standard 15N-based 2D and 3D NMR methods.55–57 Our assignments (Figure 6) are fully consistent with those reported by Liu et al.54 The HSQC spectrum of M-COms shows similar chemical shift dispersion and line widths to that of the N-state indicative of a tightly folded state. However, the spectra of the two forms show widespread differences with a majority of residues experiencing significant chemical shift changes. Although independent resonance assignments for M-COms cannot be obtained readily because of its metastable nature, many of the resolved cross peaks can be assigned to nearby peaks in the native spectrum. In other cases, where an N-state peak has no obvious partner in the CO-bound form, we were able to obtain a lower-limit estimate for the chemical shift changes in each dimension from the distance to the nearest possible unassigned peak. In Figure 7(a) the normalized 1H/15N chemical shift changes (see figure caption) are plotted vs. residue number. Arrows indicate residues for which only a lower-limit estimate is available. The largest concentration of residues undergoing major shift changes spans residues 72 through 85 with a second cluster in the 50s. Other than a few isolated residues in the 40s and 60s, only moderate chemical shift perturbations are observed throughout the remainder of the sequence, including the N-terminal segment up to residue 39 and the region containing the C-terminal α-helix (residues 87–104). In Figure 7(b), residues undergoing significant chemical shift changes upon CO binding are mapped onto the native cyt c structure. In this view, all of the perturbed residues are located on the left side of the structure, which contains the native Met80-heme linkage that is disrupted by competitive binding of a CO ligand.
Our NMR evidence that M-COms has a well ordered structure distinct from that of the native state indicates that a substantial conformational change is required to allow binding of a CO ligand under non-denaturing conditions. This conclusion is consistent with the well-known fact that CO does not directly bind the native state of Fe2+ cyt c, unless it is destabilized in the presence of moderate concentrations of denaturant or extreme pH values.12,58,59 Thus, we postulate that a native-like intermediate, M, exists even in the absence of CO. This hypothesis, which will be further justified below, leads to the following expanded scheme to describe the equilibrium between native, intermediate and unfolded forms of cyt c and cyt c-CO (Scheme 2):
Equilibrium constants KNM, KMU and KMU-CO describe the conformational transitions, and KaM and KaU are second-order CO binding constants for the binding-competent intermediate and unfolded states, respectively. The fact that the equilibrium absorbance data in Figure 2 and the NMR-detected unfolding transitions in Figure 5 are well reproduced by a three-state mechanism (Scheme 1) indicates that the two unligated non-native states in Scheme 2, M and U, are not populated at equilibrium in the presence of CO due to its high affinity for partially and fully unfolded states.
For a complete thermodynamic description of the coupled unfolding/ligand binding equilibrium of reduced cyt c, we need information on the binding constants of CO with the fully and partially unfolded states, KaU and KaM. The former characterizes the U + CO U-CO equilibrium and can be determined directly by measuring the heme absorbance changes as a function of CO concentration under fully denaturing conditions. Figure 8(a) shows the Soret region of the heme absorbance spectrum (1.2 µM sample of Fe2+ cyt c in 6.5 M GuHCl, pH 7.5, 20 °C) recorded as a function of CO concentration ranging from 0 to 50 µM. Binding of a CO ligand to the sixth iron coordination site (His18 remains bound to the fifth position) is accompanied by a large increase in absorbance 414 nm (A414), which is plotted in Figure 8(b) vs. total CO concentration, using a log scale. The observation of a clear isosbestic point (at 418 nm) indicates that the assumption of a two-state binding equilibrium (U + CO U-CO) is well justified. Least-squares fitting of a 1:1 binding equilibrium (see Methods, Eq. 2) accurately reproduces the data in Figure 8(b) and yields a dissociation constant Kd = 14 ± 8 nM. The relatively large error is due to the fact that reliable absorbance measurements can be obtained only at protein concentrations much higher than the dissociation constant, which limits the accuracy of the fit.
The kinetics of the CO binding reaction with the GuHCl-denatured state of reduced cyt c (U + CO → U-CO) was measured by laser flash photolysis. The absorbance-detected kinetics was monitored at 550 nm following flash photolysis of unfolded Fe2+ cyt c-CO (in 6 M GuHCl), using a Nd-YAG laser to generate 8 ns pulses at 532 nm. The overall kinetics consists of three exponential phases and is fully consistent with previous studies in the Soret region.12 The initial increase in absorbance with time constants τ1 = 14 µs and τ2 = 0.4 ms is attributed to the binding of Met65/80 and His26/33, respectively.12 The absorbance subsequently decreases with a rate that varies with CO concentration and is attributed to rebinding of the CO ligand to the heme. A plot of the rate of rebinding measured at 10 °C increases linearly as a function of CO concentration up to ~100 µM (Figure 9(a)). The slope yields a second-order rate constant for CO binding of 1.76±0.08 × 105 M−1s−1 (Table 2), which is comparable to the value reported for myoglobin (5 × 105 M−1s−1),60,61 but significantly lower than the CO binding rate of 2–4 × 108 M−1s−1 found for model hemes in partially aqueous solvents.62,63 The temperature dependence of the apparent CO binding rate (measured at fixed CO concentration of 135 µM) shows Arrhenius behavior with a pre-exponential factor of 3 × 109 s−1 and activation energy, Ea = 11.2 kcal mol−1 (Figure 9(b)). This activation energy is more than twice the value for CO binding to myoglobin (5 kcal mol−1).60 The low second-order rate constant and large Ea observed in our case indicate that structural rearrangements at the heme, probably associated with the dissociation of non-native His or Met ligands, are required to allow the binding of CO to unfolded Fe2+ cyt c.
The rate of CO dissociation from the fully unfolded U-CO state was measured by using imidazole as a competitive heme ligand. Both ligands share the 6th ligation site of the reduced heme iron, and when the U-CO (or M-CO) complex dissociates in the presence of a large excess of imidazole, an imidazole molecule rapidly occupies the open ligation site. Given the high solubility of imidazole, we can easily reach conditions where its rate of association is much faster than the CO dissociation rate. The rate-limited step is then dissociation of CO from U-CO state, and the optically monitored kinetics yields the off-rate for the CO ligand (koffU). The kinetics of CO dissociation was measured at 20 °C by monitoring the absorbance changes at the peak of the Soret band (413 nm) of U-CO in the presence of imidazole at concentrations raging from 0.4 to 2.25 M imidazole. The reaction was initiated by 2000-fold dilution of a CO-saturated solution of 1 mM Fe2+ cyt c containing 6.5 M GuHCl (0.1 M Tris-HCl, pH 8.0) with the same buffer containing imidazole. The dead time of manual mixing was approximately 30 s. The absorbance at 413 nm decays exponentially over a time period of about 30 min. Figure 9(c) shows that the observed rate initially decreases with increasing imidazole concentration and reaches a constant value of 3.41±0.15 × 10−3 s−1 between 1.4 and 2.25 M.
Our kinetic data for binding and dissociation of CO from the unfolded state of reduced cyt c are fully consistent with the equilibrium binding data in Figure 8. Given the rate of CO binding to the U-state, kon U, measured by flash photolysis at 10 °C and the corresponding activation energy (Table 2), we estimate a value of 3.47 × 105 M−1s−1 at 20 °C. Together with the U-CO dissociation rate, koff U, measured by imidazole competition at 20 °C, we obtain an equilibrium dissociation constant Kd U = koff U/ kon U of 10 nM, which agrees within error with the value of 14 ± 8 nM measured by equilibrium CO binding (Figure 8).
Because of the low population of the M-state in the absence of CO, the equilibrium constant for the M M-CO transition and the rates of CO binding and dissociation with the M-state cannot be measured directly. However, the apparent rates of CO binding and dissociation are slow under conditions where the M-CO intermediate is the predominant non-native equilibrium state (~1 to 3 M GuHCl) and can be measured by manual mixing. The CO binding rate at 2.5 M GuHCl (pH 7.0) was measured at temperatures ranging from 15 °C to 30 °C by monitoring the time-dependent absorbance changes in the Soret region (414 nm) following addition of CO. The reaction was initiated by manually mixing an Fe2+ cyt c stock solution in the absence of CO into a 50-fold larger volume of CO-saturated buffer, yielding a final CO concentration near 1 mM. The observed kinetics is well represented by a single exponential with an apparent rate kapp=1.95±0.1 × 10−4 s−1 at 20 °C. The observed kinetic amplitude accounts for the total absorbance change expected from equilibrium spectra in the absence and presence of CO, ruling out the presence of any rapid processes occurring during the dead time of this manual mixing experiment (~90 s). Figure 10(a) shows an Arrhenius plot of the observed rate of CO binding at 2.5 M GuHCl under CO-saturated conditions, yielding an activation energy of 30.3±1.8 kcal mol−1.
To measure the rate of CO dissociation from the M-CO state populated at moderate denaturant concentrations, we first prepared U-CO in a CO saturated solution containing 6 M GuHCl and generated the M-CO state by dilution with CO-free buffer to final GuHCl concentrations ranging from 0.075 to 4 M. Heme absorption spectra in the visible range (500–600 nm) were recorded at time intervals from 2 min to several hours after 80-fold dilution of the U-CO sample with buffer. At early times, the spectrum is very similar to that of U-CO at equilibrium (cf. Figure 1, inset), indicating that the CO ligand is still bound. The absorbance bands at 520 and 550 nm increase with time due to CO dissociation and formation of native cyt c. Our previous laser photolysis results indicate that the late folding event (M → N) triggered by dissociation of the trapped CO ligand occurs at a rate of ~1 × 105 s−1.25 Thus, the kinetics observed at the low light levels used here is limited by thermal dissociation of the CO ligand rather than photolysis or the subsequent late-stage folding events. A logarithmic plot of the CO dissociation rate vs. GuHCl concentration (Figure 10(b)) shows that the rate remains nearly constant from 0 to 3.3 M GuHCl, followed by a sharp increase. The plateau corresponds to the region where M-CO accumulates with negligible population of U-CO (Figure 3(b)), and the observed rate (~3×10−5 s−1 at 10 °C; ~1.5×10−4 s−1 at 20 °C) represents that of direct dissociation of the CO ligand from M-CO. The low denaturant dependence confirms that this process does not involve a major unfolding transition. Above 3 M GuHCl, where the concentration of U-CO increases sharply, the observed rate is increasingly determined by unfolding of M-CO followed by dissociation from U-CO. To test this model further, we used numeric methods to solve the kinetic equations corresponding to Scheme 2, as described previously.46 The solid line in Figure 10(b) represents the slowest observable rate constant (eigenvalue) predicted by the simulation, using the elementary rate constants shown as dashed lines. The fact that we can quantitatively reproduce the complex denaturant dependence of the observed rate for the coupled CO-dissociation/unfolding process provides strong support for our model. The temperature dependence of the CO dissociation rate at low GuHCl concentrations (75 mM) shows Arrhenius behavior with an activation energy Ea = 26±3 kcal mol−1 (data not shown). The activation energy is similar to the value (23 kcal mol−1) previously reported for Fe2+ cyt c-CO at pH 7.4 prepared by dilution from alkaline solution.59 We and others43 have consistently observed a small decrease in dissociation rate between 0 and 3 M GuHCl, but the origin of this effect is not clear at this time (possible explanations include ionic strength and/or viscosity effects in addition to local structure formation following CO dissociation).
Comparison of the kinetic parameters for CO binding and dissociation at low and high denaturant concentrations (Table 2) reveals striking differences. Under stable conditions (<3 M GuHCl), where N and M-CO are the only equilibrium states populated in the absence and presence of CO, respectively, the apparent rate of CO binding is lower by a factor 2×106 and the activation energy is nearly 3-fold higher in comparison to the kinetics of CO binding to the unfolded state. Both observations indicate that the rate-limiting process for CO binding under these conditions involves a major conformational change of the protein rather than an intrinsic chemical barrier for binding.
According to Scheme 2, CO binding under non-denaturing conditions can be expressed in terms of the following reaction mechanism (Scheme 3) where the rate constants kNM and kMN describe the structural transition between the N and M-states, kon M is the second-order CO binding rate to the M-state and koff M represents the corresponding dissociation rate. We already know one of the rate constants in this scheme, namely that of the late folding event, kMN = 8.3 × 104 s−1, which was measured by Pabit et al. 25 in their flash photolysis experiments starting from the metastable M-COms state. Since the N M equilibrium is strongly displaced towards N at low denaturant concentrations, we have kMN kNM. Judging from the magnitude of the CO binding rate to U (Table 2), we also expect that kMN kon M. Under these conditions (analogous to hydrogen exchange in the EX2 limit), the observed rate for CO binding, kobs, can be approximated as
where kNM/kMN is the equilibrium constant for the N M pre-equilibrium, KNM, and kon M is the apparent rate for CO binding to the M-state at a given CO concentrations (e.g., 1 mM in a saturated solution at 1 atm). Since KNM 1 under native conditions, Eq. 1 explains how the apparent rate of CO binding at low denaturant concentration can be very slow, even if the intrinsic CO binding rate to the M-state is relatively fast.
In the subsequent analysis, we make the plausible assumption that the intrinsic affinity of the U- and M-states is independent of denaturant concentration. While it is difficult to directly test this assumption because of the limited range of denaturant concentrations over which these states are populated, it is consistent with our data on the rate of CO dissociation from the M-state in Figure 10(b). For the cyclic portion of Scheme 2, the following free energy relationship holds:
From the CO binding/dissociation rates to the U-state (Table 2), we obtain ΔGa U = −RT ln(kon U/koff U) = −6.7 kcal mol−1, and the equilibrium unfolding measurements in the presence of CO yield a value of 11.3 kcal mol−1 for ΔGMU-CO (Table 1). Eq. 2 thus reduces to
The free energy for the N M-CO transition provides an additional constraint:
By adding Eq. 3 and Eq 4, we obtain a value of 22 kcal mol−1 for the unfolding free energy in the absence of CO, ΔGNU = ΔGNM + ΔGMU. By comparison, the values for ΔGNU obtained by two-state analysis of the reduced cyt c unfolding equilibrium (Figure 3; cf. refs. 49,50) range from 17.0 to 19.3 kcal mol−1. This discrepancy is related in part to the unusually high midpoint for the GuHCl-induced unfolding transition of Fe2+ cyt c (5.2 M), which makes it difficult to define the post-transition baseline and introduces considerable uncertainty in ΔG due to the long extrapolation, making the analysis highly sensitive to errors in the m-value. In contrast, our present estimate for ΔGNU is based on more accurate thermodynamic parameters for the destabilized CO-bound form and reliable kinetic measurements for CO binding and dissociation. The total m-value measured in the presence of CO, mNU = mNM + mMU = 4.8 kcal mol−1M−1 (Table 1) is above the range of values (3.3–3.7 kcal mol−1 M−1) measured in the absence of CO (Fig. 3; ref. 49), suggesting that the latter is underestimated.
The available equilibrium data are insufficient for an independent determination of the free energies for all five transitions in Scheme 2. However, we were able to solve this problem by considering the kinetics of unfolding (a more detailed account of the kinetic mechanism of reduced cyt c will be presented elsewhere; Maki, Latypov, Hagen and Roder, in preparation). The rate of unfolding of Fe2+ cyt c in the absence of CO, obtained by absorbance-detected stopped-flow measurements under strongly denaturing conditions (5–7.5 M GuHCl), is included in Figure 10(b) (open circles) along with elementary rate constants (dashed lines) and the predicted rate of unfolding (solid line). The curvature in the chevron plot at high GuHCl concentrations can be explained as follows. Below 6 M GuHCl, the slope of the chevron plot is determined by kMU, which represents a global unfolding transition and increases steeply with denaturant concentration. However, above 6 M GuHCl, the observed rate approaches kNM, which becomes rate-limiting for unfolding under strongly denaturing conditions where kMU > kMN. Because the N->M transition is a local unfolding transition associated with the loss of the native Met80-heme linkage, the slope (kinetic m-value) associated with kNM is lower than that of kMU. Quantitative kinetic modeling of the data, using Scheme 1, fully reproduces the observed unfolding behavior (solid line in Figure 10(b)). The thermodynamic and kinetic parameters obtained are listed in Table 3. CO binding to the M- and U-states is expressed in terms of apparent rates and free energies at a CO concentration of 1 mM.
Bhuyan and Kumar41 previously observed a slight curvature in the unfolding rate of reduced cyt c, which they attributed to a denaturant-induced shift in the barrier position within the framework of a two-state model.64 In contrast, we attribute the curvature in the unfolding branch of the chevron plot to a change in the rate-limiting barrier for unfolding due to the presence of a high-energy intermediate state. We previously proposed such a mechanism to describe the unfolding kinetics of oxidized cytochrome c2, which shows a more pronounced roll-over effect due to its lower stability.46 Subsequent observations of non-linear unfolding profiles for a number of proteins were also found to be consistent with a mechanism involving a high-energy unfolding intermediate.65–67 Thus, the kinetics of unfolding we and others observed for reduced cyt c confirms the existence of state M in Scheme 1, which represents an obligatory unfolding intermediate, both in the presence and absence of CO.
The biological activity of cyt c as a mitochondrial electron transport shuttle critically depends on maintaining the native His/Met heme coordination,68 and binding of alternative ligands is strongly disfavored under physiological conditions. However, CO and other exogenous ligands have such a strong affinity for binding 5-coordinate ferrous heme that they can displace the native methionine ligand under partially denaturing conditions. Our equilibrium and kinetic analysis of CO binding to unfolded Fe2+ cyt c (Figure 8 and Figure 9, Table 2) yields a dissociation constant of ~15 nM, which corresponds to a binding free energy of 6.7 kcal mol−1 at 1 mM CO. Thus, CO acts as a denaturant by dramatically lowering the free energy of non-native conformations (those with weakened or disrupted sulfur-iron bond) relative to that of the native state. This made it possible to detect a structural intermediate, M, which is difficult to observe in the absence of exogenous ligand. As illustrated schematically in Figure 11, in the absence of CO (solid lines) M is a high-energy state that affects the kinetics of unfolding, but is not detectable at equilibrium. Addition of CO (dashed lines) lowers the free energy of M by nearly 6 kcal mol−1 (Table 3), which makes it observable as an equilibrium intermediate at moderate denaturant concentrations (Figure 11(b)). Our equilibrium unfolding measurements using optical probes (Figure 2 and Figure 3) and NMR (Figure 4–Figure 6) provide unambiguous evidence for accumulation of a structural intermediate, M-CO, which has a tightly folded, but non-native conformation (Figure 7). The structural and energetic properties of Fe2+ cyt c in the presence of CO are in marked contrast to those of myoglobin, which binds diatomic ligands under native conditions and undergoes minimal structural rearrangement upon ligand binding.69 Thus, we feel that the term “carbonmonoxycytochrome c” coined in ref. 41 is misleading.
Efforts to interpret the effect of CO on the unfolding transition of Fe2+ cyt c within a structural two-state framework (i.e., a cyclic scheme involving N, U, U-CO and N-CO states),41 have led to contradictions or require the unsatisfactory assumption that the free energy for binding CO to the N-state is positive (unfavorable) at low denaturant concentrations and becomes negative with increasing GuHCl concentration. In contrast, our five-state mechanism (Scheme 2) provides a fully self-consistent thermodynamic description of the linked CO-binding/ unfolding equilibria of Fe2+ cyt c (Table 3, Figure 11) and accounts for all of our equilibrium results and kinetic data on CO binding and dissociation under partly and fully denaturing conditions (Table 2). A key hypothesis underlying this analysis is that structurally equivalent intermediate states exist both in the presence and absence of the CO ligand (Scheme 2), which is supported by several lines of evidence. (i) NMR spectral differences (Figure 6) indicate that displacement of Met80 by a CO ligand leads to widespread structural changes. (ii) The unfolding kinetics of Fe2+ cyt c is consistent with the presence of an obligatory unfolding intermediate, which limits the rate of unfolding at high GuHCl concentrations (Figure 10). (iii) The free energy associated with the main structural unfolding transitions, M U, in the absence of ligand (12.2 kcal mol−1) is comparable to that obtained in the presence of CO (11.3 kcal mol−1). (iv) The M-state binds CO with similar affinity as the U-state and the association/dissociation rates are only weakly dependent on denaturant concentration, as expected for a state with a largely exposed 6th heme coordination site (Table 3).
Accumulation of native-like states (late folding or early unfolding intermediates) appears to be a more general phenomenon. Equilibrium studies of oxidized cyt c using NMR, optical spectroscopy and scattering methods have revealed a non-native state that structurally resembles the M-CO state.48,70,71,72 For both oxidation states, intermediates lacking the native Met80-iron linkage accumulate under moderately denaturing conditions. Binding of small-molecule ligands such as imidazole, cyanide or azide to oxidized cyt c also results in some local unfolding of the structure on the distal (Met80) side of the heme and increased mobility.73–75 The complex kinetics of unfolding observed for a number of proteins, including oxidized c-type cytochromes,18,46,76 staphylococcal nuclease,65 tendamistat66 and several other proteins,67 has been interpreted in terms of mechanisms in which conversion of the native state into a high-energy intermediate can become a rate-limiting step for unfolding under strongly denaturing conditions.
Horse heart cytochrome c from Sigma-Aldrich Co. (St. Louis, MO) was used without further purification. GuHCl and urea were ultra-pure grade from MP Biomedicals (Solon, OH). All the other chemicals are reagent grade. GuHCl concentration was determined by refractive index measurements using a Reichert-Jung refractometer (Leica, Bannockburn, IL). Concentration of oxidized cyt c was determined spectrophotometrically using an extinction coefficient of 1.06 × 105 M−1cm−1 at 410 nm. Protein solutions were equilibrated for a minimum of 30 min prior to measurement. Absorbance measurements were performed using a Perkin-Elmer or Hitachi spectrophotometers using 10 or 2 mm quartz cuvettes (0.5–25 µM cyt c solutions). The pH of solutions was adjusted by addition of either HCl or NaOH and buffered with 0.1 M sodium phosphate (pH 7.0) or 0.1 M Tris-HCl (pH 7.5–8.0). Stock solutions of 1 M sodium dithionite were prepared in glass tubes with serum stoppers which were first purged with argon or nitrogen (highest purity available) to remove oxygen. Reduced (Fe2+) cyt c was prepared by anaerobically adding dithionite to solutions of oxidized cyt c under an argon or nitrogen atmosphere. Dithionite concentrations were 0.5 to 5 mM for optical measurements and 5–20 mM for NMR experiments. All buffers and protein stock solutions were purged with oxygen-free argon, nitrogen or CO gas in glass tubes or cuvettes sealed with serum rubber stoppers, and solutions were transferred by using gas-tight Hamilton syringes.
The helix content of cyt c was measured at 225 nm in an Aviv 62DS spectropolarimeter using thermostatted cells with an optical path length of 2 mm. Ellipticity readings were time averaged for up to 2 min at 60 points/s. The bandwidth was set to 2 nm and the dynode voltage was kept below 400 V. The cyt c concentration was 25 µM. For the reduction of cyt c the dithionite concentration was 0.5 mM, and despite the significant dithionite absorbance the CD signal was only slightly degraded. For CO binding the air space which comprised about 1/4 of the total volume of the sample cell was filled with CO gas at 1 atm. CO samples were equilibrated for 30 min at 20 °C.
For the measurement of the CO binding equilibrium with unfolded Fe2+ cyt c, CO concentrations were adjusted by mixing two buffers (0.1 M Tris-HCl at pH 7.5 containing 6.5 M GuHCl and 3 mM dithionite) saturated with CO (1.0 mM) and nitrogen, respectively, at atmospheric pressure. One-fifth volume of Fe2+ cyt c stock solution in the same buffer (purged with nitrogen before reduction) was added to a final protein concentration of 1.2 µM. Absorbance spectra in the Soret region (390 – 450 nm) were measured at 20 °C using a 1-nm bandwidth. The path length of the optical cuvette was 10 mm.
The dissociation constant (Kd) for the bimolecular reaction U-CO U + CO is defined as
where [U], [CO] and [U-CO] are concentrations of Fe2+ cyt c in the U-state, free CO and the unfolded CO complex, respectively. Binding curves were obtained from the absorbance change at 414 nm, ΔA414, corresponding to the maximum in the CO-bound form (Figure 1 and Figure 8(a)), using the following expression:
where A0 414 is the absorbance change in the absence of CO, [CO]total is the total CO concentration in solution and C0 is the protein concentration. Absorbance data were fitted to Eq. 3 using non-linear least squares fitting.
Cyt c solutions were placed in thermostatted cells with path lengths of 0.2 cm for CO binding and 1 cm for CO dissociation. Temperature control was achieved by placing the cuvettes in a water-jacketed brass holder. Absorbance changes were monitored at 414 or 550 nm using the light from a tungsten lamp. Absorbance changes due to thermal dissociation of CO were measured for up to 15 h. CO dissociation from the unfolded cyt c (U-CO) was measured in the presence of 0.4-2.25 M imidazole as a competitive ligand for the sixth iron coordination site. 1–2 mM Fe2+ cyt c in a CO saturated denaturing buffer (6.5 M GuHCl in 0.1 M Tris-HCl, pH 8.0) was diluted down to 0.5 µM protein in 1–3 ml of the same denaturing buffer in the absence of CO. For CO binding the gas space of the cell was filled with 1 atm CO. For dissociation, 10 µL of 1 mM cyt c-CO in CO saturated 6 M GuHCl were diluted with 0.5 ml of CO-free, pH 7.0 buffer (0.1 M sodium phosphate) to lower the GuHCl concentration. Interference from photolysis was minimized by using the smallest slit width, 0.25 nm, for the monitoring light. From the Arrhenius temperature dependence of observed rates it was verified that photolysis due to the monitoring light is negligible. Solutions containing variable concentrations of CO were prepared by adding aliquots of a CO-saturated solution to argon or nitrogen-saturated samples in air-tight cuvettes with less than 1 µL gas space.
Laser flash photolysis experiments were performed using a nanosecond transient absorbance apparatus, equipped with a Nd-YAG laser, at the Regional Laser and Biotechnology Laboratories (RLBL), University of Pennsylvania. Solutions of 2 µM Fe2+ cyt c-CO were placed in a 1 cm × 1 cm, thermostatted cuvette. Laser pulses were 8 ns in duration with an energy of 150 µJ at 532 nm. Absorbance changes following photolysis were monitored at 550 nm.
Kinetic measurements of CO binding to the M state were initiated by mixing a protein solution containing 125 µM cyt c in the absence of CO with a buffer saturated with CO using a gas-tight Hamilton syringe. The mixing ratio was 1:49 at a final protein concentration of 2.5 µM. The dead time of the mixing was approximately 90 s. Both solutions contained 0.1 M Tris-HCl (pH 7.5), 2.5 M GuHCl and 5 mM dithionite. The protein stock solution and the buffer were purged with nitrogen and CO, respectively, before adding dithionite anaerobically. The reaction was monitored by recording time-dependent change in Soret absorbance at 414 nm on the spectrophotometer at temperatures ranging from 15 °C to 30 °C.
The kinetic traces for CO binding and dissociation were analyzed by non-linear least-squares fitting of a single-exponential function. The activation energies were determined using the Arrhenius relationship,
where R and T are the gas constant and absolute temperature. kobs, A0 and Ea are the observed binding/dissociation rate constant, the pre-exponential factor and the activation energy, respectively.
For 1D NMR, labile amide protons were exchanged in D2O in order to resolve resonances of amide side chain protons in the NH region of the NMR spectrum. A 0.5 mM cyt c solution in oxidized form (95%) prepared in 0.1 M phosphate, pD 10, D2O buffer was incubated at 60 °C for an hour. Then the pH was adjusted to 7 and the solution lyophilized. Concentration of urea (or GuHCl) was varied by mixing aliquots of pH 7 stock solutions of cyt c in D2O buffer and in 9.5 M urea (or 6.5 M GuHCl). TSP (3-[Trimethylsilyl] propionic acid) was also added as a reference for NMR chemical shift. A stock solution of 1 M sodium dithionite was prepared in 0.1 M phosphate, pD 7, D2O buffer. Cyt c was reduced anaerobically in NMR tubes by the injection of 10 µL dithionite stock under continuous argon flow. The NMR tubes were tightly sealed with conical rubber stoppers. The dithionite reduction procedure caused no substantial change in denaturant concentration. For CO binding studies, sealed NMR tubes containing cyt c solution and 1 atm CO were shaken at room temperature (22 °C) for about 2 hours to allow equilibration. NMR spectra were acquired on a Bruker DMX 600 MHz spectrometer. A typical spectrum was obtained by averaging 512 transients of 16K data points over a spectral width of 11.1 kHz. Multidimensional NMR spectra for uniformly 15N-labeled H33N variant of horse cyt c (details of sample preparation are described in ref.48) were collected at 15 °C using a 5-mm x,y,z-shielded pulsed-field gradient triple-resonance probe. To confirm resonance assignments, 15N edited NOE-HSQC and TOC-HSQC56, HNHA57, and HNHB77 spectra were collected. 1H-15N HSQC experiments were run with 256 experiments in 15N dimension (t1) consisting of 40 scans and 4096 data points in 1H dimension (t2). The spectra processing and contour peak integration were done by using Felix (FelixNMR, San Diego, CA. USA).
The work was supported by NIH grants R01 GM 056250 and CA06927, and an Appropriation from the Commonwealth of Pennsylvania to the Institute for Cancer Research. We thank the Spectroscopy Support Facility at FCCC for maintaining the NMR and optical spectrometers, and the Regional Laser and Biotechnology Laboratory for access to nanosecond laser flash photolysis.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.