Investigations of heme ligation and ligand switching in cytochromes P450 and P420

Abstract

It is generally accepted that the inactive P420 form of cytochrome P450 (CYP) involves the protonation of the native cysteine thiolate to form a neutral thiol heme ligand. On the other hand, it has also been suggested that recruitment of a histidine to replace the native cysteine thiolate ligand might underlie the P450→P420 transition. Here we discuss resonance Raman investigations of the H93G myoglobin (Mb) mutant in the presence of tetrahydrothiophene (THT) or cyclopentathiol (CPSH), and on pressure-induced cytochrome P420_cam (CYP101), that show a histidine becomes the heme ligand upon CO binding. The Raman mode near 220 cm⁻¹, normally associated with the Fe-histidine vibration in heme proteins, is not observed in either reduced P420_cam or the reduced H93G Mb samples, indicating that histidine is not the ligand in the reduced state. The absence of a mode near 220 cm⁻¹ is also inconsistent with a generalization of the suggestion that the 221 cm⁻¹ Raman mode, observed in the P420-CO photoproduct of inducible nitric oxide synthase (iNOS), arises from a thiol-bound ferrous heme. This leads us to assign the 218 cm⁻¹ mode observed in the 10 ns P420_cam-CO photoproduct Raman spectrum to a Fe-histidine vibration, in analogy to many other histidine bound heme systems. Additionally, the inverse correlation plots of the ν_Fe-His and ν_CO frequencies for the CO adducts of P420_cam and the H93G analogs provide supporting evidence that histidine is the heme ligand in the P420-CO bound state. We conclude that, when CO binds to the ferrous P420 state, a histidine ligand is recruited as the heme ligand. The common existence of a HXC-Fe motif in many CYP systems allows the C→H ligand switch to occur with only minor conformational changes. One suggested conformation of P420-CO involves the addition of another turn in the proximal L helix so that, when the protonated Cys ligand is dissociated from the heme, it can become part of the helix and the heme is ligated by the His residue from the adjoining loop region. In other systems, such as iNOS and CYP3A4 (where the HXC-Fe motif is not found) a somewhat larger conformational change would be necessary to recuit a nearby histidine.

Introduction

The cytochrome P450 enzyme family (CYP) is composed of a broad range of heme-containing proteins that are involved in drug metabolism, toxicity, xenobiotic degradation, and biosynthesis¹. One key structural feature of these proteins is the coordination of the thiolate anion of cysteine (Cys) to the heme iron as the fifth ligand in the active P450 form^2-5. The biologically inactive conformation of a cytochrome P450 protein is typically denoted as the “P420” form and is characterized by a CO bound Soret peak near 420 nm, which is blue-shifted with respect to the peak at ~450 nm found in the active cytochrome P450. The inactive conformation can be formed from all known types of P450 using various methods^6-10. It has recently been accepted that this spectral change is due to protonation of the cysteine thiolate, resulting in thiol ligation to the heme iron^10-12. On the other hand, it has also been suggested that the P450 → P420 transition involves a ligand switch from a cysteine (Cys) to a histidine (His) ligated heme¹³. The similarity of the absorption spectra between the P420 form of cytochrome P450 and other proximal histidine ligated heme proteins¹⁴ reinforces such a correlation. Wells et al.¹³ provided key evidence of histidine ligation in the CO bound form of P420 by observing a strong ν_Fe-His mode at 218 cm⁻¹ in the 10 ns transient Raman spectra of the P420_cam-CO photoproduct with an intensity equivalent to that of MbCO. Moreover, the equilibrium resonance Raman spectrum of the P420_cam-CO adduct is virtually identical to that of MbCO, lending further support to the histidine ligation model, at least when CO is bound.

On the other hand, heme model compound studies^{4, 5} along with spectroscopic comparisons between chloroperoxidase and cytochrome P420_cam have led to suggestions¹² that a thiolate-thiol transition might accompany reduction of the ferric P420 (even though a residual low-spin thiolate population is also observed¹²). More recently, Perera et al.¹¹ showed that the proximal ligand mutant (H93G) of deoxy myoglobin (Mb) can bind thiol and thioether compounds with high affinity, K_d ~ 10 μM. This conclusion is based on the observation of changes in the absorption spectra of the reduced H93G Mb mutant upon titration with tetrahydrothiophene (THT) and cyclopentathiol (CPSH)¹¹. However, it must be pointed out that the changes of the absorption spectra upon ligand binding are quite small and this could be due to either direct heme ligation or to perturbations of the electrostatic environment surrounding the heme^{15, 16}. The observed optical changes are not large enough to provide unambiguous evidence for heme ligation by either THT or CPSH.

Recently, Sabat et al.¹⁰ suggested that thiol could be the proximal heme axial ligand in the inactive (P420) form of inducible nitric oxide synthase (iNOS). This protein is analogous to the P450 class, based upon thiolate ligation to the heme in its active state. The conclusion to exclude histidine as the proximal ligand in the 5 ns transient Raman spectra of the CO adduct was based on the absence of the expected ~1 cm⁻¹ H/D isotopic shift for the 221 cm⁻¹ mode, which is usually assigned to the Fe-His vibration^17-22. Based on the absence of a H/D isotopic shift, it was concluded that the 221 cm⁻¹ mode observed in the inactive iNOS-CO photoproduct spectrum was neither a Fe-His nor a Fe-SH stretching mode (the latter ligation state was expected to generate an even larger isotopic shift). As a result, Sabat et al. assigned the 221 cm⁻¹ mode to a new mode associated with the thiol-ligated ferrous heme chromophore. However, if this assignment is correct, the 221 cm⁻¹ heme mode should be observed in the thiol-bound reduced state of the inactive P420 iNOS and other reduced P420 systems or analogs. Because the Raman spectrum of reduced P420 iNOS is not available¹⁰, we turned to the Raman spectrum of reduced P420_cam to try and find the predicted thiol-bound heme mode at 221 cm⁻¹. However, the Raman spectra of reduced P420_cam from either this or a prior¹³ study does not reveal the presence of such a mode.

Therefore, in order to resolve the various characterizations of the heme proximal ligand in P420 systems, we have used resonance Raman spectroscopy to study the CO bound and ferrous forms of P420_cam and H93G Mb; the latter, in the presence and absence of THT and CPSH. There is no 221 cm⁻¹ mode observed in ferrous P420_cam or in the reduced H93G Mb with or without the THT or CPSH ligands. Moreover, it is also noteworthy that the experiments show no difference, within the resolution of ±1cm⁻¹, between the resonance Raman spectra of reduced H93G Mb with or without the THT and CPSH. These observations do not yield supportive evidence that either THT nor CPSH directly ligates the heme iron of the reduced H93G Mb. Additional experiments are presented that probe the inverse correlation of the ν_Fe-CO and ν_CO Raman modes and are consistent with histidine as the proximal heme ligand in CO bound P420. As discussed below, we conclude that the ~220 cm⁻¹ modes observed in P420-CO, iNOS-CO, and H93G-CO photoproduct Raman spectra are, in fact, signatures of Fe-His ligation in the CO bound complexes.

We also construct a kinetic scheme that describes the photostationary states of P420-CO and H93G-CO and accounts well for the various Raman observations consistent with previously determined rate constants. A fundamental hypothesis underlying the kinetic model is that CO binding to heme greatly increases the “acidity” of the ferrous iron atom so that it efficiently recruits the strong sigma donating histidine ligand. Earlier work has shown that the affinity of CO bound heme for imidazole is ~10⁵ times larger than its affinity for a weak ligand such as water²³. This concept has been used previously, both in the context of low pH MbCO ligation kinetics²⁴ and in prior studies of ligand switching in H93G-CO^{25, 26}; it is extended here to account for Raman observations in the P420 system.

Materials and Methods

All chemicals used in this study were purchased from Sigma Aldrich. Imidazole-free sperm whale H93G myoglobin was prepared as previously described^{11, 27}. High purity P420_cam was prepared in the absence of camphor by pressure treatment of P450_cam as previously reported^{8, 9, 12}. We have shown in previous studies that high hydrostatic pressure will dissociate bound substrate²⁸ and that the pressures required to initiate P420 inactivation are less in the absence of camphor. Thus, in order to generate a clean sample of P420, the substrate free form of the protein was used. The reduced H93G Mb and P420_cam samples were prepared in 0.05 M KPi pH 7.0 buffer and the protein concentrations were adjusted to 100 μM. A small amount of saturated sodium dithionite solution (1% by sample volume) was used to reduce the samples. In order to prepare the samples of H93G Mb with thiol or thioether, either 20 mM of tetrahydrothiophene (THT) or cyclopentanethiol (CPSH) in ethanol stock solution was added to the reduced H93G Mb solution¹¹. The final concentration of THT and CPSH in the H93G Mb solution was 200 μM, which should lead to >95% binding with <1% bis-thiol formation, assuming the K_d=10 μm as taken from the work of Perera et al.¹¹.

Resonance Raman spectra were obtained using a standard setup with 90° light-collection geometry and a single grating monochromator model SP-2500i, Princeton Instruments, Acton, MA. An optical polarization scrambler was inserted in front of the monochromator to obtain the intensity of the scattered light without bias from the polarization-sensitive grating. The monochromator output was coupled to a thermoelectrically cooled charge-coupled detector (PIXIS 400B, Princeton Instruments). To improve detection in the low-frequency region of Raman shifts, an interferometric notch filter (Kaiser Optical Systems, Ann Arbor, MI) was used to extinguish the elastically and quasi-elastically scattered laser light. Samples were excited with a 413.1 nm laser line generated by a krypton laser (Innova 300, Coherent) using a power of 11 mW at the sample or with a 442 nm laser line from a HeCd laser (Melles Griot) at powers up to 32 mW. In order to study the photolabile CO adducts, a cylindrical quartz cell with 10 mm diameter was mounted to a home-built spinning system and used for the Raman measurement. The spinning speed was set at 6000 rpm for all experiments except static measurements. All Raman spectra were frequency calibrated using pure fenchone with ~ 1cm⁻¹ spectral resolution.

Results

Resonance Raman spectra of deoxy H93G and its THT and CPSH adducts are compared with ferrous P420_cam in Fig. 1. The corresponding high frequency region (1300 cm⁻¹ - 1700 cm⁻¹) region, including the ν₄, ν₃, ν₂ and ν₁₀ bands, is shown in Fig. S2 of the Supporting Information. All spectra in Fig. 1 are normalized to the ν₇ band. No differences are observed in the Raman spectra of reduced H93G Mb, H93G(THT) Mb and H93G(CPSH) Mb, although the absorption spectra in the Soret region show subtle, but clear, differences¹¹ (see also Fig. S1, Supporting Information). The concentrations of THT and CPSH used in these measurements should lead to ~ 95% binding according to K_d =10 μM as reported by Perera R. et al.¹¹. Based on prior Raman studies of ligand binding to the H93G Mb mutant²⁹, we expect to observe changes in the resonance Raman spectra when ligand binding to the heme takes place. Given the fact that there is no observable change in the resonance Raman spectra upon THT or CPSH binding, it seems possible that these thioether and thiol compounds might be binding to a site in the H93G protein that is close enough to affect the Soret band shape and position (~2 nm shift), but perhaps they are not replacing the heme water ligand that is normally present in reduced H93G Mb.

Figure 1.

Low frequency resonance Raman spectra of reduced H93G Mb, its THT and CPSH adducts, and ferrous P420_cam. The excitation wavelength is 413 nm and the laser power at the sample is 11mw for the H93G samples, while for the P420 sample the excitation is 20mw at 413nm (blue), 25 mW at 420 nm (cyan) and 32mw at 442 nm (magenta). The sample cell is spinning at 6000 rpm. All spectra are normalized to the ν₇ band. The 492 cm⁻¹ feature also appears in Fig. 3 and is not assigned to a ν_FeCO mode.

The Raman spectra of reduced P420_cam and the H93G derivatives also show no evidence of a mode near ~220 cm⁻¹. Sabat et al. ¹⁰ observed a Raman mode at 221 cm⁻¹ in the 5 ns photoproduct Raman spectrum of the inactive iNOS P420-CO adduct and, because a H/D isotopic shift was not detected, they assigned the 221 cm⁻¹ mode to a heme vibration activated by thiol ligation. However, if thiol is ligated to the heme in the reduced state, one would expect this mode to be present in the equilibrium Raman spectrum of the reduced P420 sample (i.e., the equilibrium species should display essentially the same modes, although slightly shifted, when compared to the 5 ns transient photoproduct species). Note that the P420 sample was probed at several wavelengths (413 nm, 420 nm¹³, and 442 nm) and there is no evidence of a mode near 220 cm⁻¹. The absence of a 221 cm⁻¹ mode in the Raman spectrum of the equilibrium reduced P420_cam and the H93G P420 analog samples does not support its assignment to a thiol-bound reduced heme mode¹⁰. On the other hand, if the usual assignment of this mode to the Fe-His vibration is made, its observation in the 10 ns transient Raman spectra provides strong evidence for a heme-histidine bond in the CO-bound forms of P420¹³ and, by analogy, iNOS¹⁰. Such an assignment is also consistent with previous transient Raman studies of the H93G-CO photoproduct²⁶.

The low frequency resonance Raman spectra of CO bound H93G, H93G(CSPH), and H93G(THT) Mb, excited at 413 nm, are shown in Fig. 2. All spectra are normalized to the ν₇ band. There is essentially no difference between the three samples, except for small changes in the relative amplitudes of the 507 cm⁻¹ and 522 cm⁻¹ modes, which are associated with the Fe-CO stretching frequency. The broad mode at 492 cm⁻¹ seen in both Figs. 1 and 2 is not isotopically sensitive (Fig. 3) and therefore not assigned to a Fe-CO mode. The spectra of the corresponding high frequency region, including the ν₄, ν₃, ν₂ and ν₁₀ bands, are displayed in Fig. S3 of the Supporting Information and are identical for all three samples.

Figure 2.

Resonance Raman spectra of CO bound H93G, H93G (CSPH), and H93G (THT) excited at 413 nm. The incident laser power at the sample is 11mw. The sample is spinning at 6000 rpm. All spectra are normalized to the ν₇ band.

Figure 3.

Resonance Raman spectra of the ν_Fe-CO and ν_CO modes of ¹²CO and ¹³CO bound H93G, H93G (THT) and H93G (CPSH). The low frequency Fe-CO stretching and bending region is in the left panels. The high frequency CO stretching is in the right panels and the ¹³CO – ¹²CO difference is shown in blue. Excitation wavelength is 413 nm and laser power at the sample is 11mw. The sample was in a cell spinning at 6000 rpm. Dashed and dotted lines indicate ν_FeCO modes of H93G(His)CO and H93G(H₂O)CO respectively.

Figure 3 shows the resonance Raman spectra for the ν_Fe-CO and ν_CO modes of the ¹²CO (black) and ¹³CO (red) adducts of H93G, H93G(THT), and H93G(CPSH) Mb. The lower frequency Fe-CO stretching region is shown in the left panels, while the higher frequency CO stretching is displayed in the right panels. The isotopic shifts confirm that there are two peaks corresponding to the ν_CO stretching. For the ¹²CO sample there is a strong ν_Fe-CO peak at 522 cm⁻¹ that shifts to 517 cm⁻¹ and a broader feature at 507 cm⁻¹ where the shift is less obvious, but becomes apparent upon fitting the data as discussed in the Supporting Information (Fig. S4). The corresponding ν_CO stretching modes are located at 1960 cm⁻¹ and 1942 cm⁻¹, respectively. The ν_CO Raman frequencies agree very well with the infrared measurements reported previously²⁵. The 507 cm⁻¹ mode is associated with the histidine bound Fe-CO mode and its shift upon isotopic labeling is not so easily seen due to its breadth and weaker resonance enhancement. (Note that the Soret absorption band of CO bound heme is blue-shifted by approximately 5 nm when a weak ligand such as water replaces imidazole³⁰. This means that, based on the Raman excitation profile of MbCO¹⁶, the resonance enhancement at 413 nm will favor the Fe-CO mode of the water bound heme relative to that of the histidine bound population). There is also interference from the broad feature near 492 cm⁻¹, as seen in Fig. 1 and in the fit to the CO bound lineshape (Fig. S4). This peak has been previously reported for WT deoxyMb^{31, 32} and its H93G mutant²⁵ and, since it shows no isotopic shift, it is not assigned to a Fe-CO mode. On the other hand, there is photolytic activity in the region near 507 cm⁻¹ (vide infra), so we are confident that it represents the position of a Fe-CO oscillator. Upon ¹³CO substitution, the respective ν_CO modes, 1960 cm⁻¹ and 1942 cm⁻¹, show clear shifts down to 1915 cm⁻¹ and 1897 cm⁻¹, respectively.

Figure 4 shows the resonance Raman spectra of the same samples as in Fig. 3, but it demonstrates the effect of photolysis when the spinning cell is stopped. The spectra in black have the sample in the quartz cell spinning at 6000 rpm, while the red spectra are accumulated in a static cell. The panels on the left side show the 460-550cm⁻¹ ν_Fe-CO stretching region, while the panels on the right side display the 1870~2000 cm⁻¹ ν_CO stretching bands. As expected, the ν₄ bands demonstrate that the ratio of the CO dissociated (5C) to the CO bound (6C) species is increased in the static cell (see Fig. S5 in Supporting Information). In the static-cell spectra, the intensities of the 507 cm⁻¹ band and the 1942 cm⁻¹ band decrease simultaneously relative to the 522 cm⁻¹ and 1960 cm⁻¹ bands. Thus, we associate the 507 cm⁻¹ ν_Fe-CO peak with the 1942 cm⁻¹ ν_CO peak as belonging to the same FeCO species. A similar association holds for the 522 cm⁻¹ and 1960 cm⁻¹peaks. Moreover, the data indicate that the 507/1942 cm⁻¹ FeCO species has a slower CO geminate rebinding rate (and thus a smaller relative population in the photostationary state of the static cell) than the species at 522 cm⁻¹ and 1960 cm⁻¹.

Figure 4.

Resonance Raman spectra of CO bound H93G with and without THT/CPSH. Spectra in black are taken with the sample spinning at 6000 rpm, while the red spectra are with a static cell. Panels on the left side show the 460-550cm⁻¹ ν_Fe-CO stretching band and the difference spectra in blue reveal the presence of the 507 cm⁻¹ mode because of its increased photolysis compared to the 522 cm⁻¹ species. Panels on the right side display the ν_CO stretching band and the analogous increased photolysis of the 1942 cm⁻¹ species compared to the 1960 cm⁻¹ species. The excitation wavelength is 413 nm and the laser power at the sample is 11mw.

The resonance Raman spectra of the P420_cam ¹²CO and ¹³CO adducts, showing the ν_Fe-CO and ν_CO peaks, are displayed in Fig. 5a. To our knowledge, these are the first ν_Fe-CO and ν_CO frequency correlations that have been made for a P420 system. The ν_CO difference spectrum between ¹³CO-¹²CO is shown in blue. For ¹²CO, the ν_Fe-CO stretching peak is found at 494 cm⁻¹, while the ν_CO mode is at 1965 cm⁻¹. The ν_CO mode from the resonance Raman spectrum matches the IR measurments very well^{33, 34}. In Fig. 5b we show the stationary vs. spinning cell comparisons where a loss of Fe-CO and CO band intensities is observed in the stationary cell. Again for P420, as observed in the H93G adducts, the ratio of the CO dissociated (5C) to the CO bound (6C) species is increased in the stationary cell. This is evidenced by the appearance of a ν₄ band at 1361 cm⁻¹ in Fig. S5 (Supporting Information) that confirms the presence of significant 5C photoproduct (rather than a 4C photoproduct)³¹. There is no sign of a mode near ~220 cm⁻¹ in the photostationary state data (Fig. S5). Since this mode is clearly observed and assigned to ν_Fe-His in the 10 ns transient Raman spectrum¹³, its absence in Fig. S5 is attributed to a ligand switch between His and another ligand with a rate that is faster than the cw photoexcited CO escape rate into solution. It should also be noted that the ν₄ band is located at 1354 cm⁻¹ in the 10 ns transient spectrum¹³ and this band is shifted to 1361 cm⁻¹ in Fig. S5. This is also consistent with a ligand switch and probably involves water replacement as an intermediate heme ligand in the photocycle. A kinetic model that accounts for these observations, along with the previously observed kinetic rates for CO binding to P420³⁵, is given in Scheme I below.

Figure 5.

(a) Resonance Raman spectra of the P420 ¹²CO and ¹³CO adducts at λ=413nm. The power at the spinning sample cell is 11mw. The ν_Fe-CO region is on the left and the ν_CO is on the right with the ¹³CO-¹²CO Raman difference spectra shown in blue. (b) Resonance Raman Spectra of P420 CO. Spectra in black are taken with the sample spinning at 6000 rpm, while the red spectra are with a static cell.

Scheme I.

Kinetic model for photostationary states in H93G-CO and P420-CO with H indicating a proximal histidine ligand and L indicating a water (or, on longer timescales a thiol) proximal ligand. The A-states have CO bound, the B-states have CO in the distal pocket, while CO has escaped into solution in the C-states. The H93G-CO sample reveals poplulations of A_H, A_L, and C_L while only A_H and C_L populations are observed in P420-CO (unless the states A_H and A_L have identical FeCO frequencies). This kinetic scheme is reduced to an effective three state system and analyzed more completely in the Supporting Information.

Discussion

The Raman spectra of the CO adducts of the H93G Mb mutant with and without added THT and CPSH are presented in Figs. 3 and 4. From the variation in photoactivity, we conclude that the ν_Fe-CO modes at 507 cm⁻¹, and 522 cm⁻¹ are associated with the ν_CO modes found at 1942 cm⁻¹ and 1960 cm⁻¹, respectively. These data points, along with others^{30, 36-38}, are plotted on a ν_Fe-CO vs ν_CO correlation diagram shown in Fig. 6. This allows us to compare the CO adducts of P420_cam and the H93G compounds with the large data base of other CO bound heme species that have been studied using the correlation method^{16, 36}.

Figure 6.

Correlation plot of ν_FeCO and ν_CO with data from ref. 29-32. The red color open squares are for histidine (or imidazole) ligated heme systems. The open green triangles represent thiolate ligated heme systems. The open blue diamonds represent heme systems with a weak or absent proximal ligand. Three solid lines labeled 5C/6C(water), His, and Cys are the lines fitted to the data points. The solid blue and red dots are data points for H93G-CO and P420-CO from this work. The notation H93G(H₂O)CO and H93G(His)CO correspond to the histidine and water ligated populations of H93G-CO observed in the H93G Raman spectra.

The ν_Fe-CO and ν_CO modes typically follow an inverse π back-bonding relationship³⁶ as shown in Fig. 6. Back-donation of iron d_π electrons to the CO π* orbitals strengthens the Fe-CO bond and weakens the C-O bond. Trans ligands with a stronger σ donation will weaken the Fe-CO bond more than expected, due to the change in π backbonding resulting from sigma donor competition for the iron d_Z² orbital³⁷. Thus, CO adducts with a strong trans ligand lie lower on the plot. For the same reason, CO adducts with weaker trans ligand lie higher on the plot (e.g., the water-ligated FePPIX-CO data point³⁰ is denoted by the arrow). Mb variants with a neutral trans His ligand and different distal pocket mutations are spread along the line labeled His. The position of the various CO adducts on the line reflects the polarity of their distal binding pocket. CO adducts with distal residues that produce strong H-bond interactions, like V68N, or the Mb “closed” distal pocket (A₁) state¹⁶ lie higher on the line, whereas those with nonpolar distal residues, like H64V or the Mb “open” distal pocket (A₀) state¹⁶, lie lower on the line. The 507/1942 cm⁻¹ modes of the H93G MbCO complex are assigned to 6-coordinate low-spin histidine bound forms with a “closed” distal pocket, while the 522/1960 cm⁻¹ modes are assigned to a water bound form²⁵ (In principle, the 522/1960 cm⁻¹ modes could also result from a five coordinate CO bound species; however, the photostationary state conditions do not produce a 4C photoproduct ν₄ band. Rather, a ν₄ band at 1359 cm⁻¹ is observed, which is consistent with the assignment²⁵ of water as the ligand trans to CO in these complexes).

The ν_Fe-CO and ν_CO modes of cytochrome P420_cam are found at 494 cm⁻¹ and 1965 cm⁻¹. These values agree with independent Raman and IR data^{13, 33}. The 494/1965 cm⁻¹ point is found on the inverse correlation diagram at a position that is consistent with a trans His ligand and an open distal pocket. This supports the assignment of histidine as the proximal ligand in P420_cam-CO, but it is not fully conclusive because of recent work on model ferrous porphyrin complexes in thioether solvents that has suggested thiol complexes may have a similar back-bonding correlation³⁷.

Interestingly, as seen in Fig. 5, the Fe-CO and CO frequencies of P420 do not change upon photoexcitation in a stationary cell even though a significant photoproduct population is created (as evidenced by the ν₄ band at 1361 cm⁻¹ observed in Fig. S5). Analogous to the H93G system, the P420-CO shows no indication of a Fe-His mode under the photostationary conditions (Fig. S5). This suggests that, as for H93G-CO^{25, 26}, there is a rapid loss of the His ligand upon CO photodissociation (i.e., k_Ex >> k_γ (k_out/k_BA^H+k_out+k_Ex) in Scheme I below). Moreover, there must be a relatively rapid reset rate, k_Re, to the initial His bound ligation state (A_H) following CO binding.

In contrast to H93G-CO, water does not appear as a photostationary state ligand (L) in P420-CO, although it probably participates as an intermediate in the ligand exchange process denoted by k_Ex in Scheme I. Assuming water functions as a transient ligand (L) in the photocycle, it is potentially detectable as a trans ligand in the CO bound state, A_L, when the system is driven into a photostationary state equilibrium. In such a case, we might expect to see intermediates with frequencies (analogous to the 522/1960 cm⁻¹ modes) that appear on the upper line in Fig. 6. However, the CO geminate recombination rate and geminate yield are very large for P420²⁹, possibly exceeding those observed for CooA³⁹. (The missing amplitude in the early kinetic work³⁰ makes precise evaluation difficult, but the actual geminate amplitude for CO binding in P420 may approach 99%). This has the effect of reducing the effective rate of C_L formation from A_H, as well as the rate of A_L formation from A_H via the B_H-B_L channel in Scheme I. Under this condition very little A_L population will exist in the photostationary state (see Supporting Information for a more detailed analysis).

The other possibility to account for the fixed positions of ν_Fe-CO and ν_CO in the stationary cell (Fig. 5b) would be for the state A_L in Scheme I to have Fe-CO and CO frequencies that are identical to the histidine-ligated CO bound state (A_H). In principle, the latter possibility might be realized if thiol (not water) was the ligand (L). However, it would be highly coincidental if thiol and histidine ligation lead to exactly the same FeCO frequencies. Moreover, direct thiol ligation (without a water ligand intermediate) would also require very rapid protein rearrangements during the ligand switching process. Finally, the hypothesis of a single thiol ligand can also be ruled out because of the strong Fe-His mode observed in the 10 ns transient spectra of P420-CO. We can use the ν₇ band as a reference to determine that the strength of the transient Fe-His mode in P420CO is equivalent to that of MbCO¹³. This observation is the “smoking gun”, demonstrating that thermal equilibrium must favor the A_H state in the P420_cam system.

In the myoglobin H93G mutant, the native proximal ligand is replaced by Gly and the only possible candidates for histidine ligation are His97 on the proximal side and His64 on the distal side. The binding of His64 was previously excluded by resonance Raman spectra of the CO bound H64V/H93G double mutant, which is very similar to that of H93G MbCO²⁵. Based on transient Raman spectra that detects the 220 cm⁻¹ Fe-His mode, His97 was assigned as the prime candidate to be the trans ligand when CO binds and acidifies the heme iron²⁶. When CO is photolyzed, the time resolved step-scan infrared data indicate²⁶ that the histidine ligand is replaced by water on a time-scale that is faster than ~10⁶ s⁻¹ and that the histidine ligand does not rebind until CO bimolecular rebinding takes place²⁵. Thus, in the static cell, the population of the His bound form will be less than that in the spining cell because the continuous photolysis leads to a larger proportion of the fast rebinding water bound species. The water bound heme has a lower proximal barrier and rebinds CO much more rapidly than the His bound heme⁴⁰. This is why only the ν_CO and ν_FeCO peaks, associated with the water bound heme population (A_L), are observed in the photostationary state when the spinning cell is stopped.

The model for the CO photolysis and H₂O-hisitidine ligand exchange in the H93G system in Scheme I is very similar to the ligand switch model presented in an earlier study²⁵, but measurements of H93G-CO kinetics²⁹ suggest that CO escape into solution must compete with the rate for histidine exchange with water in the pocket. The photostationary equilibrium spectra for H93G-CO display strong evidence that water ligates to the CO bound heme as suggested by earlier work²⁵. Moreover, the rate of histidine recruitment by the water bound CO state (A_L) to form A_H is on the order of, or smaller than, the rate of A_L and C_L production from A_H (see Supporting Information). The different photostationary state behavior of H93G compared to P420 can be traced to the geminate rebinding rate of A_H in the respective systems and the fact that H93G retains a distal barrier that significantly slows its geminate rebinding compared to P420. We also note that the photostationary populations of A_H and A_L in Scheme I appear favor the A_L state in H93G MbCO, even under spinning conditions. More details of the kinetic analysis in spinning and stationary conditions can be found in the Supporting Information.

The frequencies of the ν_Fe-CO and ν_CO modes in H93G MbCO are unaffected by the addition of THT or CPSH. Comparison of the absorption spectra of H93G deoxyMb in 200 μM CPSH and THT solution to the pure H93G deoxyMb shows a clear difference, as can be seen in Fig. S1. These results are consistent with prior work¹¹, but one can interpret the small absorption spectral change as THT and CSPH binding to the protein, close enough to affect the heme electrostatic environment^{15, 16}, yet without direct ligation to the heme. While it is conceivable that both the thiol and thioether compounds bind to heme and precisely mimic the histidine and/or water-bound heme ligation states, it seems much more likely that these compounds are not actually ligating the heme iron. Rather, they could have a binding site nearby, close enough to account for the 2 nm shift in the Soret band, which provides the only evidence that these ligands are binding to the reduced H93G Mb system (recall from Figs 1, 2, S2, and S3 that there is no binding effect registered in the heme-specific resonance Raman spectra). A non-heme binding site for THT and CPSH is also consistent with the observation of two H93G MbCO species. Two H93G MbCO states (histidine and water) are observed in the Raman spectra as described above. If THT and CPSH were binding to the heme iron in the expected 1:1 stoichiometry, one would expect that a single set of ν_Fe-CO and ν_CO modes would be observed.

In Fig. 2, the 522 cm⁻¹ peak shows a somewhat lower relative intensity compared to the 507 cm⁻¹ peak when THT and CPSH are added to H93G MbCO. This indicates that the His bound form, characterized by the 507 cm⁻¹ peak intensity, has more relative population when the CPSH or THT are added to the solution. The system is undergoing a complex photon driven dynamics that involves competition between the photoexcitation rate, CO escape and bimolecular entry into the heme pocket, and the rate of recruitment of His97 as an iron ligand during the time CO is bound to the heme. Ligand switching models involving histidine and water have been proposed previously in the context of both H93G and low pH Mb^{24, 25, 31, 41}. Upon CO binding to the water-ligated H93G Mb state, the iron seeks to bind a strong sigma donating ligand²³ and recruits His97. Depending upon the photoexcitation rate, the CO escape and entry into the pocket, and the very different geminate rebinding rates for the water and His97 bound heme⁴⁰, the two CO-bound populations will reach a photostationary equilibrium (e.g., see Scheme I). The presence of CPSH and THT in the heme pocket might be expected to modify the equilibrium between A_L and A_H, leading to the subtle changes in Fe-CO populations observed in Figs. 2 and 3.

In Fig. 1 there is no 220 cm⁻¹ mode observed in the resonance Raman spectra of either reduced H93G Mb or reduced P420_cam. This indicates that, in the absence of CO, histidine is not ligated to the reduced heme. The cysteine thiol ligand is the obvious candidate for ligation in the reduced state of P420. Although there is no direct spectroscopic evidence, the thiol ligation assignment in the ferrous form of P420 has been discussed in the context of similarities with CPO^{11, 12}. The absence of water ligated CO bound signals in the photostationary state Raman spectra indicate that water does not form a particularly stable intermediate in P420-CO; however, it does not eliminate water as a possible transient ligand in states C_L and A_L of the photocycle.

Nanosecond transient Raman spectra of P420_cam-CO¹³, iNOS P420-CO¹⁰ and H93G-CO²⁵ adducts all show a strong 220 cm⁻¹ mode, which is very similar to the MbCO 10 nanosecond transient Raman spectra¹³ and has been assigned in many heme protein systems as the Fe-His stretching vibration^{17-19, 22}. If the 221 cm⁻¹ mode, observed in the 5 ns photoproduct Raman spectrum of the P420 form of iNOS-CO, was a heme mode¹⁰ (rather than the Fe-His vibration), we should be able to observe it in the Raman spectrum of the reduced P420 samples/analogs in Fig. 1. The absence of a 221 cm⁻¹ Raman mode in Fig. 1 is inconsistent with its assignment to a thiol ligated heme mode. This indicates that the transient photoproduct Raman spectra of P420 systems is actually revealing the presence of a proximal histidine ligand, which has been recruited in the CO bound state as shown in Scheme I. Such an analysis is consistent with the prior assignment for the 220 cm⁻¹ mode observed in the H93G-CO system²⁶. Thus, in thermal equilibrium (i.e., no photoexcitation or between the ns laser pulses), the histidine ligated P420-CO state is favored.

Since the ν_Fe-His mode at ~220 cm⁻¹ is not active in 6-coordinate CO adducts³⁶ and the CO dissociated population rapidly replaces the histidine ligand with water, the photostationary experiments are not able to observe a ν_Fe-His mode with either a spinning or a static cell. Thus, the 5 ns transient resonance Raman spectrum of P420-CO, which displays a 220 cm⁻¹ mode similar in intensity to the MbCO photoproduct, provides the most direct evidence to assign histidine as the trans ligand in the CO bound P420 systems¹³. We suggest that the apparent absence of an isotopic shift for the 221 cm⁻¹ mode in the inactive iNOS P420-CO photoproduct spectrum is due to the fact that the H/D shift is only expected¹⁰ to be 0.7 cm⁻¹. The signal-to-noise of the transient Raman spectra¹⁰ of the iNOS P420-CO was significantly worse than that of the control experiment on MbCO where a ~ 1 cm⁻¹ H/D shift was observed. The peak shift algorithm developed previously²⁰ indicates that a 25 cm⁻¹ (full width at half-height) Gaussian band that shifts by 0.7 cm⁻¹ will generate a maximum-to-minimum in the Raman difference spectrum that is 8% of the measured peak height. Since the noise level of the transient difference spectrum between the protonated and deuterated iNOS P420-CO photoproduct exceeds this value by approximately a factor of two¹⁰, we do not believe that the 0.7 cm⁻¹ isotopic shift of the Fe-His mode could be detected for iNOS P420 and therefore that the ~0.7 cm⁻¹ H/D shift is present, but undetectable.

Finally, one must address the issue of whether there are His ligands available to undergo ligand switching with Cys in various P450 systems. Using the protein data base, we find that there is at least one His residue within 10 −15 Å of the heme iron in membrane bound P450 3A4, P450cam, and iNOS as shown in Figs. S7, S8, and S9 of the Supporting Information. These nearby histidine residues are candidates for binding to heme when P420 undergoes the tertiary structural changes associated with the binding of CO and the loss of the thiol ligand. These distances also have implications regarding the extent and type of the conformational fluctuations that are taking place in these systems as they become inactivated. In other heme proteins, such as chloroperoxidase⁴² and Rr-CooA⁴³, axial thiolate ligands have been shown to undergo a ligand-switch to a nearby histidine following reduction of the heme iron. It has also been reported that the nitrophorin 1 H60C mutant loses its heme thiolate cysteine ligand upon heme reduction⁴⁴. Generally, upon reduction of the heme iron, the neutralized heme core could trigger protonation of the thiolate ligand leading to its dissociation and the subsequent structural rearrangements that facilitate a functional and biologically relevant ligand switch^{43, 45}. A variety of thiolate-heme proteins that have evolved to become heme-sensor proteins with functional ligand switching reactions have recently been reviewed⁴⁶ and the P450/P420 reaction has many similarities.

Dunford et al.⁴⁷ recently reported that CYP121 from M. Tuberculosis can undergo reversible conversion from its P420 form back to the P450 form when the pH is raised from 6.5 to 10.5. The P420 form dominates at lower pH, while the P450 form dominates at higher pH. These observations were interpreted using a simple two-state model⁴⁷ involving the reversible protonation of the proximal cysteine ligand. There are 4 histidine residues less than 15 Å from the heme iron in CYP121. However, the H343 residue is only 2 amino acids away from the proximal Cys ligand, forming a HXC-Fe sequence that is shared with both CYP101 and CYP51 (For most CYP51s, this sequence is conserved⁴⁸, including the CYP51 from M. Tuberculosis studied by Dunford et al.). The proximity of His343 to the heme in CYP121 might facilitate its ligation to the iron atom in the event that protonation of Cys345 leads to its dissociation in the CO bound state. Upon deprotonation at higher pH, the Cys345 thiolate will again become a strong ligand that can potentially displace His343 so that the P450 form is reversibly recovered. The reversibility of this process will depend sensitively upon the strength of the His-heme-CO ligation and the new structural motif that is formed in the P420 state (vide infra). An important point is that a Cys(thiol)→His→Cys(thiolate) ligand switch mechanism provides an alternative interpretation of the pH dependent reversible P420/P450 conversion in CYP121 and would help to explain the irreversibility observed in CYP51 (assuming the P420 structure formed in CYP51 is particularly stable). Moreover, the relatively slow (0.72±0.02 s⁻¹) and pH independent transition rate that is observed for P450→P420 conversion⁴⁷ might actually be more consistent with a ligand switch process as the rate limiting step, rather than a simple thiolate protonation reaction. (In the context of Scheme I, this means that an additional column of states A_T, B_T, and C_T, where T represents thiol, must be added in slow exchange with the photocycle states A_L, B_L, and C_L, where L=H₂O.) Additional investigations specific to the P420 forms of CYP121 and CYP51 are clearly needed to identify the proximal ligand in the CO bound forms at acid and alkaline pH.

Along with CYP121 and CYP51, CYP101 shares the HXC-Fe motif with X representing Phe, Arg and Leu, respectively. The Cys loop HXC sequences in the heme proximal pocket of these proteins are shown in Fig. 7. Here we recognize that the His residue is actively engaged in H-bonding with the nearby propionate group of the heme. Some possible proximal pocket conformational changes that would allow binding of the nearby histidines to the heme iron are depicted in Fig. 8. Examination of the helix-loop transition region in CYP101, which lies between residues G359(helix) and L358(loop), suggests that upon dissociation of the protonated Cys357, the three adjacent loop residues L358, C357, and L356, could coil into the alpha helix, leaving His355 in prime position to bind to the iron atom as depicted in Fig. 8. Moreover, we find that another histidine (H352) lies downstream in the loop region and that it can be easily positioned to recover the necessary H-bond to the heme propionate. Thus, a relatively simple and energetically favorable conformational fluctuation, slightly extending the L-helix, could lead to the histidine-heme ligation following CO binding in CYP101. Larger scale renditions of this transition can be seen in section S10 of the Supporting Information.

Figure 7.

Crystal structure of CYP51 (PDB#1EA1), CYP121(PDB#2IJ7), CYP101(PDB#2CPP) and CYP3A4(PDB#3NXU). The HXC motif in CYP51, CYP121 and CYP101 are highlighted. The right bottom panel shows the loop (magenta) of the heme proximal pocket in CYP3A4. The C442 and H402 residues are shown with a space filling model. There is no HXC motif in CYP3A4.

Figure 8.

The three panels on the left side show the crystal structure of CYP101 (PDB#2CPP), CYP51(PDB#1EA1), and CYP121(PDB#2IJ7). The L-helix is extended by 5 residues past the Cys ligand in the CYP51 and CYP121 structures. The helix structures on the left and right side of the proximal cysteine ligand are labeled as A and B, respectively. The red dashed line indicates the hydrogen bond between “near” propionate group and H355, H392 and H343 in the three structures, respectively. The three panels on the right show the possible CO bound P420 structures for the three proteins. The proposed structures are formed by extending the A helix by three amino acids (highlighted in red). A remnant of the B helix structure is preserved in the proposed P420-CO structure for CYP51 and CYP121. In the proposed structures, H352, R391, and Q342 replace the histidine residues of the native structures by forming a hydrogen bonds (red dashed line) with the propionate group. Larger renditions and different angles showing the possible structures underlying the ligand switch can be found in the SI.

It is interesting to compare the CYP121 and CYP51 proteins, which also share the HXC motif. CYP121 displays a reversible pH dependent P450/P420 conversion while CYP51 does not^{47, 49}. Evidently, CYP51 converts to its P420 form immediately upon reduction, even in the absence of CO ⁴⁹, suggesting that its Cys heme ligand has a pK that is somewhat higher than for CYP121. A simple thiol/thiolate pH titration model for P420 conversion would predict that, upon raising the pH high enough (say above 10), the putative thiol Cys ligand in CYP51 should ultimately deprotonate and revert to a thiolate so that a reversible P420/P450 should also be observed in this system. On the other hand, the ligand switch model can easily explain the irreversible behavior. For example, if a particularly stable P420 structure is formed upon CO binding, the simple deprotonation of the Cys residue may not be enough to energetically reconfigure the protein structure and recover thiolate ligation to the heme iron.

In Fig. 8 we have shown a possible structural change for CYP51 conversion to P420. Here we find that, upon extension of the L-helix to include Cys394 and the orientation of His392 as the axial ligand, Arg391 is naturally prepositioned to form a strong H-bond with the heme propionate. Thus, the Cys(thiol)→His ligand switch in CYP51 may form a very stable alternative structure (e.g., by incorporating the Cys(thiol) into the adjacent alpha helix and forming a strong Arg391 H-bond with the heme propionate) leading to a situation that is energetically stable and not reversible by pH back-titration⁴⁷. It should also be noted that the P420 form of CYP51 displays a partial reconversion to P450 upon loss of CO and re-oxidation⁴⁹. Since the iron-histidine bond is weakened following CO dissociation and iron re-oxidation, it is evidently possible for the cysteine thiolate residue to successfully compete to once again become a heme ligand in the ferric state of CYP51. In contrast to CYP51, CYP121 undergoes a reversible pH titration between P420 and P450^{47, 49}. The structures that we found for the P420-CO state in this system appeared to have adequate, but less satisfying, H-bonding to the heme propionate suggesting that it might have less energetic stabilization and therefore be more likely to undergo a reversible transition following pH back titration.

Although many P450 systems share the HXC-Fe motif, some do not and their potential histidine ligands are not found as close to the proximal heme ligation site. Two important examples discussed above are iNOS and CYP3A4, where a larger tertiary structure change is necessary to bring a histidine close enough for heme ligation to occur. Figures 7, S7, S9, and S10 (Supporting Information) show nearby histidine residues that are ligation candidates for these systems. For iNOS, H661 (~10 Å away) and H407 (~ 16 Å away) are potential candidates for a ligand switch with Cys 415. For CYP3A4, the H402 (~14 Å away) is a possibility for ligand switching with Cys442 as seen in the lower right quadrant of Fig. 7. The fact that a strong 220 cm⁻¹ mode is observed in the nanosecond transient Raman spectra of iNOS-CO indicates that, even for proteins without the HXC-Fe motif, a histidine ligand can possibly be recruited to form the final P420 state following CO binding. We also note that, in systems without the obvious HXC-Fe motif, it may be possible that other strong sigma donating ligands can act as a substitute for histidine binding to the CO ligated heme.

In summary, the conversion of the P450 thiolate to thiol appears to be an important step in the conversion of P450 to P420. However, upon CO binding the 10 ns transient Raman spectra demonstrate that a nearby histidine must be involved in a ligand switching equilibrium with the cysteine thiol. All indications are that this equilibrium favors a histidine bound P420CO ground state in CYP101. When there are His residues near the heme, as in the HXC-Fe loop, only relatively small tertiary structural changes are needed for the ligand switch (e.g., incorporation of C and X into the nearby helix structure can properly position the His for heme binding). The structures that are formed in these P420-CO states may have variable stability. In some cases, the process might be irreversible with pH titration (e.g., CYP51), depending upon the strength of the alpha helix interactions, the His-Fe-CO bond formation, and the heme propionate H-bond that is formed. On the other hand, if the P420-CO states are less stable, reversibility via pH titration would be more likely (e.g., CYP121). If the ligand switch requires a very large tertiary structural change, the process would be likely to proceed at a slower rate and may not be easily reversible.

Conclusion

We have used resonance Raman spectroscopy to study P420_cam and the H93G Mb mutant, with and without the addition of THT and CPSH compounds. There is no evidence from the vibrational spectra to indicate that THT or CPSH ligate the heme iron. There is no 220 cm⁻¹ Fe-His mode observed in the reduced samples, which indicates that histidine is not the heme ligand in the ferrous state. Thiol is the likely heme ligand in the thermally equilibrated ferrous state of the P420 systems, while water appears as the ligand in the reduced H93G system. The transient Raman spectra of the CO bound species indicate that a Fe-His bond is formed when CO binds to P420 systems and acidifies the heme iron. A histidine ligand for P420-CO is also indicated by the position of the υ_Fe-CO and υ_CO frequencies on the inverse correlation plots although, strictly speaking, thiol cannot be ruled out by this correlation³⁷. On the other hand, the invariance of the residual P420 FeCO frequencies when photostationary photoproduct states are formed indicates that the equilibrium between A_H and A_L in Scheme I favors the histidine ligated state, A_H, unless a thiol ligated state, A_L, yields exactly the same frequencies. The photostationary state Raman spectra of H93G MbCO indicate that, following CO binding to the water ligated heme, the rate for histidine recruitment is ~ 10⁴ s⁻¹ (see Supporting Information). Several P450 systems offer a common HXC-Fe motif that may facilitate Cys-His ligand switching, while other P450 systems must undergo slower and larger tertiary changes in order for the acidic iron of the CO-bound heme to replace the weakly bound thiol ligand with a nearby histidine or with some other strong sigma bonding donor.

The above considerations bring into focus some motivations for trying to better understand the P450-P420 reaction. For example, if a reversible helix-loop transition triggered by CO photolysis underlies the P450-P420 conversion, in even a sub-set of P450 systems, they would present interesting models for the study of protein conformational transitions using time-resolved spectroscopies. It is also conceivable that reversible transitions of this type might play a functional role in the regulation of the monoxygenase function. Finally, if it is possible to engineer the destabilization of the P420 form of the protein without similarly destabilizing the P450 form, one might envision P450 mutants that are more robust and less likely to convert to the P420 form. Mutations that remove propionate H-bonding in the P420 form (but not in the P450 form) provide one interesting target.

Abbreviations

THT

tetrahydrothiophene

CPSH

cyclopentathiol

His

histidine

Mb

myoglobin