Water may inhibit oxygen binding in hemoprotein models

Abstract

Three distal imidazole pickets in a cytochrome c oxidase (CcO) model form a pocket hosting a cluster of water molecules. The cluster makes the ferrous heme low spin, and consequently the O₂ binding slow. The nature of the rigid proximal imidazole tail favors a high spin/low spin cross-over. The O₂ binding rate is enhanced either by removing the water, increasing the hydrophobicity of the gas binding pocket, or inserting a metal ion that coordinates to the 3 distal imidazole pickets.

Hemoglobin (Hb), myoglobin (Mb) and cytochrome c oxidase (CcO) are hemoproteins that bind O_2, subsequently transporting or reducing it in a 4e-/4H+ process (1, 2). The kinetics of oxygen binding to the active sites of these biomolecules are tuned by stabilizing interactions between the oxygen complex and the immediate environment (3, 4). In monometallic proteins such as myoglobin or hemoglobin, a distal histidine stabilizes the oxygen complex by hydrogen bonding (5, 6). A distal copper tris-imidazole ligand in CcO also provides enthalpic stabilization (7). In addition to structural factors affecting oxygen adduct stability in Mb and Hb, it has been suggested that the molecules of water present in the binding pocket could contribute to the relative oxygen affinities (4, 8, 9) in a model dubbed the “water displacement model.” This proposed model could logically be extended to CcO where water is the product of the 4e-/4H+ reduction of O₂. Water is expected to be present in larger quantities in CcO than in Hb or Mb, and it is removed from the binding site by water channels (10–13). Here, we describe well-defined biomimetic hemoprotein models 1–4 (Fig. 1) that demonstrate the role of water in slowing the binding of O₂. The rates of reaction correlate with the hydrophobicity of the distal pocket (tris-pivalamido- or picket fence in 2, tris-imidazole in 3) and the presence or absence of a distal bound metal [Cu(I) or Zn(II) in 4ab].

Fig. 1.

Porphyrin models 1–4. Distal features affecting O₂ binding: pocket, hydrophobicity, and metal [Cu(I), Zn(II)]. (A) (Inset) Superimposed Myoglobin and Hemoglobin active sites (8). (B) (Inset) Cytochrome c oxidase active site (14).

Results and Discussion

The reaction of oxygen with ferrous porphyrins 1–4 was carried out by injecting an anaerobic solution of porphyrin into O₂-saturated dichloromethane solution (10 mM) under 1 atm of O₂. Under the initial conditions of the reaction it is assumed, based on earlier studies (15), that the ligand association rate k_on(O₂) is several orders of magnitude faster than the ligand dissociation rate k_off(O₂) and consequently the dissociation rate can be assumed negligeable. The O₂ binding rates were obtained by monitoring the change in ε of the Soret and Q bands in the UV/Vis spectrum (Fig. 2) that shifted from 426 nm to 421 nm, and from 535 nm to 550 nm, respectively. The oxygen complex was characterized at various stages of the reaction by 2 resonance Raman stretches: (i) an oxygen isotope sensitive band observed at 570/544 cm⁻¹ that corresponds to the Fe-O stretch and (ii) a ν4-band (a spin state and redox state marker band) at 1,370 cm⁻¹ that is typical of a ferric-superoxo species (16–19). Second order k_on(O₂) rate constants measured with tris-imidazole porphyrin models 3–4ab were in the 1–18 M⁻¹·s⁻¹ range (Table 1). They appeared to be 7 to 8 orders of magnitude smaller than those reported for CcO, Mb, and Hb (20–26). However, the binding was too fast to measure a rate with 2, which is consistent with the 10⁸ rate reported with the α4 analog of 2 by flash photolysis of its CO complex (15).

Fig. 2.

Monitoring the absorption at 426 nm upon reaction of models 1–4a with oxygen. The rate of O₂ binding to iron porphyrins is enhanced by either increasing the hydrophobic character of the distal pickets (Imidazole-C2-H 3a < Imidazole-C2-Pr 3b < t-Bu 2) (A) or adding a distal metal [Cu(I) in 4a] or drying the cavity within the distal pocket in 3a (B).

Table 1.

Binding of dioxygen to hemoprotein models 1–4 (in dichloromethane): rates, redox potential, absorption spectrum, spin state, activation parameters

Model	konO2, M−1·s−1	E0 (mV) MeCN vs NHE	UV/Vis Deoxy, nm	Spin state*	Ref(s).
1	5 × 107	180	428	HS/LS	15
2 (α3)	Too fast	90	427–429	HS/LS	This study
(α4)	4.3 × 108	—	426	HS	15
3a (H) Wet	1	87	426	LS	This study
Dry	Too fast	—	435	HS/LS
3b (Pr)	18	—	426	LS	This study
4a Cu(I) Wet	6.0	123	426	LS	This study
4b Zn(II)	6.5	96	426	LS	This study
Hb, Mb	107 to 108	—	435	HS	24–26
CcO	107 to 108	—	444	HS	20, 21, 31

Comparison with hemoproteins Hb, Mb (in buffer), and CcO.

*, as shown by Resonance Raman and ¹H-NMR (Fig. 3 A–F).

Because O₂ binding involves an electron transfer from Fe(II) to form a ferric superoxide (24, 27, 28), the O₂ binding rates were examined in light of the redox potential of the hemes (Table 1 and Fig. S1) (29), because it appears that the distal features (superstructure and/or metal) have an effect on the redox potential of the heme. When no superstructure is present in flat porphyrin 1, the redox potential is 180 mV, whereas it is 90 mV and 87 mV for Picket fence (2) and tris-imidazole (3), respectively (vs. NHE). The small difference between 2 and 3 may be due to the difference of the donating character of t-Bu compared with that of imidazole. However, despite significant differences in the redox potential between 1 and 2, the rates of O₂ binding reported with these species were in the same range (15) and very different from the slow rate found with 3.

The presence of a distal metal has an effect on the redox potential of the heme. It shifts from 87 mV for an iron only species 3 that does not bear a distal metal, to 123 mV [with a distal Cu(I)], to 96 mV for a distal Zn(II). The rates of O₂ binding are markedly affected by the presence of a distal metal, but are surprisingly unaffected when Cu_B(I) is replaced by Zn_B(II). Previous reports had suggested enthalpic stabilization of the oxygen complex in bimetallic systems (7) and suggested that oxygen may bind first to Cu(I) before binding to iron (30).

In summary, we observed the O₂ binding rates to be independent of the redox potential of the iron porphyrins, but dependent on the distal porphyrin structure. Increased rates were observed commensurate with increasing hydrophobicity of the distal pocket (t-butyl in 2 vs. imidazole in 3; simple C2-H-imidazole in 3 vs. C-2 alkylated imidazole in 3b) and to a lesser extent, in the presence of a distal metal. The presence of water in the distal pocket may explain these differences.

In picket fence species 2 a molecule of water bound to iron may undergo hydrogen bonding with other molecules of water that in turn could hydrogen bond with the 3 distal amides of each picket. In tris-imidazole species 3ab, more hydrogen bonding can occur because of the nitrogen atom of the imidazole ring, resulting in a bigger cluster of water in the distal pocket than in 2. Once a distal metal is loaded in 3a to form 4a, a reorganization of the pocket may occur resulting in fewer molecules of water present in the distal pocket (a water cluster might still bind to the distal metal instead of the imidazole nitrogen).

To examine this hypothesis, tris-imidazole porphyrin 3a was carefully dried by a series of distillations in dry dichloromethane at room temperature, which resulted in an increase of the O₂ binding rate by an order of magnitude. If drying was carried out 3 times by azeotropic distillation from anhydrous refluxing toluene, the binding rate of O₂ became too fast to be measured (Fig. 2). Reintroducing H₂O into a previously dried aliquot returned the O₂ binding rate to the same value as that obtained before drying. When the reintroduced water was D₂O, the O₂ binding rates demonstrated a weak but significant isotope effect (1.1), yielding additional evidence that the presence of protons in the distal pocket may be involved in the transition states. Together, these facts are consistent with the concept that the rate of O₂ binding to an iron porphyrin is dramatically decreased when a cluster of water molecules is present in the distal pocket bound to iron.

Spectroscopic studies of the ferrous porphyrins 1–4 help explain the differences in reactivity with O₂ (Fig. 3). In the presence of presumably wet toluene, porphyrin 3a is 6-coordinate (6C) with a Soret band at 426 nm. After rigorous drying and distillation of the toluene, the Soret band in the UV/Vis spectrum shifts to 435 nm, typical of a 5-coordinate (5C) species (32) indicative of the absence of a bound distal water cluster. Washing the 5C species with H₂O regenerates the 6-coordinate species (Soret 426 nm). The ν4 and ν8 spin state marker bands in the resonance Raman (rR) spectrum of 1 (flat) and 2 (picket fence) demonstrate 2 bands corresponding to a mixture of high spin (HS) S = 2 (ν4: 1,342 cm⁻¹; ν8: 336 cm⁻¹) and low spin (LS) S = 0 (ν4: 1,356 cm⁻¹; ν8: 380 cm⁻¹) (16, 19) iron. However, only 1 stretch at 1,355 cm⁻¹ indicative of a low spin species is present when a water-cluster bound tris-imidazole species 3–4ab is evaluated. As a result of being low spin, species 3–4ab displayed mostly diamagnetic NMR features (the off-rate of water being obviously slower than the NMR time scale). However, the ¹H-NMR spectrum of these species is not as well-defined as with CO-bound or O₂-bound species (33, 34) suggesting that multiple species with different spin states may be present. Upon drying the sample by azeotropic distillation with toluene the following observations were made: (i) paramagnetic signals develop first at 15 ppm then at 50–60 ppm corresponding to the signals of β-pyrrole protons (35–38) and (ii) the ν4 band of a high spin species develops (1,342 cm⁻¹) to give a 3:7 ratio ν4 band(HS)/ν4 band(LS). However, washing a dry sample with water resulted in the disappearance of both the ¹H-NMR paramagnetic signals and the resonance Raman ν4 high spin marker band. Upon azeotropic distillation under anhydrous conditions the amount of water in the distillate was titrated with the Karl Fisher reagent (39). A correlation with the amount of porphyrin, showed that 6 molecules of water were removed from the pocket in 3a. An approximation of the volume of half of a cone formed by the tris-imidazole porphyrin system [bottom radius (6 Å) × height (6 Å) × top radius (3 Å)] gives 2 × 10⁻¹⁶ nL. This corresponds to ≈6 molecules of water, suggesting full occupancy of the distal pocket. At this level of dryness (λ_max = 435 nm (5C), rR 30% HS) OH stretch resonances at 3,392 cm⁻¹ are still observed in the infrared spectrum suggesting that some molecules of water are still present (bending modes at 1,650 cm⁻¹ and 900 cm⁻¹ overlapped with the porphyrin bands and were not observed). Infrared bands that have been reported as a characteristic of strongly hydrogen-bonded water were also found at 6,700 cm⁻¹, 5,200 cm⁻¹ and 3,220 cm⁻¹ (40–42) (Fig. S2). It is expected that the water molecule bound to Fe(II) has the lowest free energy of all water molecules in the pocket and would be the most difficult to remove. Even under prolonged exposure to vacuum, water could still be detected in IR: This fact may be related to the importance of effective water channels that operate in CcO to remove water from the active site. The binding of water to iron is expected to be tighter in the ferric than the ferrous case, because the former is a better Lewis acid; and the smaller size of the ferric ion may have extra room for an even bigger water cluster.

Fig. 3.

A-F. Spectroscopic analyses of ferrous imidazole-tailed- tris-imidazole picket porphyrin 3ab before (trace a) and after (trace b) drying illustrating the low-spin/high- spin cross over due to the presence of a cluster of 6 molecules of water. (A) UV/Vis spectrum. (B) Resonance Raman. (C and D) Comparison resonance Raman ν4 and ν8 bands for compounds 1–3. (E) ¹H-NMR. (F) IR.

Interestingly, a 50:50 ratio of HS:LS was found at room temperature in a dry sample of the ferrous flat tail porphyrin 1 that also displays paramagnetic and diamagnetic signals (Fig. S3). At low temperature, resonance Raman indicates it is mainly low-spin (Fig. S4). Although model 1 can still bind water, it does not have a distal superstructure capable of hosting a cluster of water molecules. The fact that 1 still displays LS character strongly suggests that the proximal tail itself, possibly because of its rigid design with a 1.3 disubstituted phenyl linker between the imidazole ring and the porphyrin, has a strong coordinating effect that induces a spin equilibrium making the 5C ferrous complex 1 borderline high spin/low spin [a covalently attached tail can be considered as being equivalent to ca 40 equiv. of free imidazole (34)]. This tail is rigid, unlike previous “floppy” imidazole tails where the imidazole was linked to the porphyrin by several methylene linkages (15, 32, 34). Altogether the kinetic and spectroscopic data along with the presence of a distal water cluster in the tris-imidazole environment correlate with borderline high-spin/low-spin character of ferrous heme (induced by the strong donating character of the proximal imidazole tail). In addition, the distal cluster of water has a donating effect sufficiently strong to make the iron porphyrin low spin. The resulting low-spin character favors tighter binding of water resulting in a decrease of the free energy of bound water. This phenomenon is reminiscent of the high-spin/low-spin cross-over found in ferric cyt.P450 (43–45), that is induced by 2 cooperative factors: the coordination of 6 molecules of water in the substrate binding domain with other molecules of water and an arginine residue. Molecular mechanics and INDO studies on the latter system suggested that the presence of several water ligands decreased the energy difference between HS and LS from 75 kJ/mol to 16 kJ/mol. Moreover, according to the ligand-field theory, a thiolate (as in cytP450) is not as strong a ligand as imidazole (as in 3–4) (46). This further explains the capacity of these particular imidazole-tailed porphyrins 3–4 to undergo spin cross-over upon binding to a distal water cluster with a possible cooperative effect from the distal imidazole pickets in 3ab or from the distal metal-bound trisimidazole in 4ab.

The large cluster of water molecules bound to iron not only induces steric hindrance inhibiting oxygen binding, but may also create a spin state barrier to binding: A low-spin ferrous heme is diamagnetic, whereas dioxygen is paramagnetic. As a result, oxygen binding is not facile when a water cluster is present. In such a situation, the rates of O₂ binding may actually mirror the off rate of water and the spin equilibrium, which may explain why 2 steps are observed during O₂ binding. The fast first step corresponds to the binding to a 5C species, whereas the slower step corresponds to the binding to a LS (water-cluster-bound) 6C species (Fig. 4).

Fig. 4.

Oxygen binding in a bimetallic Fe(II)Cu(I) model bearing a water cluster in the distal tris-imidazole pocket.

Dichloromethane used for the kinetic studies may be a good mimic of the hydrophobic environment in membrane bound CcO. The CcO enzyme system has water channels composed of ≈4 Å pores to expel water and protons (10–13), whereas displacement of water from model 3–4 is expected to meet ca 6 kcal/mol solvation barrier in a hydrophobic medium consequently slowing down the rate of O₂ binding relative to that seen in CcO. When O₂ binding studies with 3 were carried out in a pH 7 buffered-acetonitrile mixture, a 5-fold increase in O₂ binding rates was found [k_on(O₂) = 30 M⁻¹·s⁻¹], which suggests that unlike dichloromethane this solvent mixture may mimic the hydrophilic character of water channels of CcO. In addition to solvating the displaced water cluster, it is also expected that the hydrogen bonding network inside the bound water cluster can be extended to the external aqueous environment, resulting in a decreasedfree energy barrier to removal of the iron-bound water cluster. These results are consistent with other reports of oxygen binding with water soluble porphyrins (47, 48). They also highlight the possible relevance of water channels in CcO, which allow the removal of water before oxygen binds. The O₂ binding rate that could not be measured with dry samples may be related to that found in CcO. An earlier report suggested that O₂ binding in free CcO is in the 10⁷ range (20), whereas O₂ binding to flash photolyzed CO bound CcO is in the 10⁸ range (21, 49). This difference may account for bound water in the free CcO, that is not present in the CO bound CcO. Upon photolysis of the CO bound species, a competition may occur between O₂ and water binding. Nevertheless these fast rates are consistent with the high spin state found in ferrous deoxy heme a3 in CcO (31, 50, 51). The present results imply that efficient water pumps must operate in CcO to remove tightly bound molecules of water and to prevent the formation of a low spin state in CcO. In the Fe(III)/Cu(II) resting state the water pumping should be even stronger because (i) the ionic diameter of iron in the ferric state is 1.2 times smaller than that in the ferrous state, which will leave more room for a bigger cluster and (ii) it is a stronger Lewis acid than ferrous iron, which should result in a tighter binding with a water cluster.

In conclusion, using simple ferrous tris-imidazole picket porphyrins we have shown that water can dramatically slow the rate of O₂ binding, which addresses the importance of an efficient removal of water in CcO to achieve high O₂ binding rates.

Materials and Methods

Iron porphyrins 1-4 were synthesized as described in ref. 52. Solvents from a Solvent Purifier System were passed through alumina columns. For water titration experiments solvents were stored over activated sieves and basic alumina, and the glassware was flame-dried. Reactions with oxygen were performed by injecting a solution of porphyrin (25 μL) in 1.1 mL of O₂-saturated dichloromethane solution (≈10 mM; resulting porphyrin concentration 2 × 10⁻⁶ M⁻² × 10⁻⁵ M) at 11 °C, 25 °C, and 35 °C.

Spectra.

UV-Vis spectra were recorded on a Hewlett Packard apparatus 8,452 in glass cuvettes (1-cm path) sealed with a 14/24 septum. The time traces of distal metal were obtained using the kinetics mode available in the Agilent Chemstation Software at time intervals of 0.5 s under constant stirring.

Cyclic voltamograms were collected on a Pine potentiostat (Robinson). Redox potentials reported are vs. Ag/AgCl reference electrode. Concentration of catalyst was 1–2 mM; supporting catalyst wastert-butyl ammonium bromide; scan rate 200 mV/s.

¹H-NMR spectra were collected on a Varian 500-MHz apparatus in CDCl₃ in PTFE-caped NMR tubes.

Resonance Raman spectra were obtained by using a Princeton instrument ST-135 back-illuminated CDD detector on a Spex 1,877 CP triple monochromator with 1,200, 1,800, and 2,400 grooves per millimeter holographic spectrograph gratings. Excitation was provided by a Dye Laser (Stilbene 599; Coherent) that was energized by a Coherent Innova Sabre 25/7 Ar+ CW ion laser. The laser line at 425 nm (≈10 mW) was used for excitation. The spectral resolution was <2 cm⁻¹. Sample concentrations were ≈800 μM in Fe. The samples were either run at room temperature, or cooled to 77 K in a quartz liquid nitrogen finger dewar (Wilmad) and hand spun to minimize sample decomposition during scan collection.

Water Titration Experiments.

In the glove box, 10.5 mg of porphyrin 3a (7.6 × 10⁻⁶ mol) in solution in anhydrous toluene (SPS, stored over activated sieves, 20 mL) was placed in a 100 mL-capacity flame-dried Young flask. The flask was sealed, transferred outside the box and connected to the one end of a T-shaped glass, whereas the other end was connected to another empty Young flask. The T and empty Young glasses were flame dried under vacuum. The porphyrin solution in Flask A was frozen and put under vacuum. Then flask A was closed, the solution was melted and plunged in an oil bath at 120 °C for 30 min. Flask B was plunged in liquid N₂ and the main vacuum tap was closed. The tap in Flask A was carefully opened and distillation of the toluene occurred. When all of the solvent in Flask A was distilled and collected in Flask B, the tap in both flasks was closed. The solution in flask B was slowly melted by plunging the vessel in water. Both flasks were transferred in the glove box. The titration of water from the content of Flask B was carried on a Metler Toledo DL39 Karl Fischer Coulometer, using the coulometric method (39). A quality check of the toluene used for the extraction showed it was anhydrous (<1 ppm water). Flask A was refilled with dry toluene in the box and the operation was repeated to ensure that no water was present. The first distillate contained 36 ppm of water, which corresponds to ≈5 molecules of water per tris-imidazole porphyrin, the second distillate 11 ppm (1.3 molecules), the third distillate 2 ppm (no water was detected). IR spectra were recorded on a Mattson Galaxy 4030 FT-IR spectrometer on 1-cm-thick KBr pellets by evaporation of a solution of porphyrin in THF.