Functional importance of Glutamate-445 and Glutamate-99 in proton-coupled electron transfer during oxygen reduction by cytochrome bd from Escherichia coli

Abstract

The recent X-ray structure of the cytochrome bd respiratory oxygen reductase showed that two of the three heme components, heme d and heme b₅₉₅, have glutamic acid as an axial ligand. No other native heme proteins are known to have glutamic acid axial ligands. In this work, site-directed mutagenesis is used to probe the roles of these glutamic acids, E445 and E99 in the E. coli enzyme. It is concluded that neither glutamate is a strong ligand to the heme Fe and they are not the major determinates of heme binding to the protein. Although very important, neither glutamate is absolutely essential for catalytic function. The close interactions between the three hemes in cyt bd result in highly cooperative properties. For example, mutation of E445, which is near heme d, has its greatest effects on the properties of heme b₅₉₅ and heme b₅₅₈. It is concluded that 1) O₂ binds to the hydrophilic side of heme d and displaces E445; 2) E445 forms a salt bridge with R448 within the O₂ binding pocket, and both residues play a role to stabilize oxygenated states of heme d during catalysis; 3) E445 and E99 are each protonated accompanying electron transfer to heme d and heme b₅₉₅, respectively; 4) All protons used to generate water within the heme d active site come from the cytoplasm and are delivered through a channel that must include internal water molecules to assist proton transfer: [cytoplasm]→E107→E99 (heme b₅₉₅)→E445 (heme d)→oxygenated heme d.

Graphical abstract

1. Introduction

The tri-heme cytochrome bd family of respiratory oxygen reductases (cyt bd’s) are present in many in bacteria and archaea but not in eukaryotes [1]. Included in the cyt bd family are the so-called cyanide-insensitive oxidases or CIOs [1, 2] in which the active-site heme d is replaced by heme b, though the enzymes otherwise appear unchanged. All of the characterized cyt bd’s are quinol:O₂ oxidoreductases, catalyzing the 2-electron oxidation of (usually) ubiquinol or menaquinol and the 4-electron reduction of O₂ to H₂O [3–6]. Cyt bd’s typically have a high affinity for oxygen, and appear to be particularly important for oxygen-scavenging under micro-aerobic growth conditions [7, 8]. Although they are not proton pumps, cyt bd’s generate a proton motive force and are energy-conserving. The oxidation of the quinol molecules releases protons on the periplasmic side of the membrane (electrically positive side) while the protons used to generate water at the enzyme active site are taken from the cytoplasmic side of the membrane (electrically negative). This results in the electrogenic net transfer of 4 protons across the membrane per O₂ [9].

$$2Q{H}_2+O_2+4H_{in}^{+}\to 2\mathrm{Q}+2H_2\mathrm{O}+4H_{\mathit{out}}^{+}$$

Most studies of cyt bd have focused on E. coli cyt bd-I, which is one of two cyt bd’s encoded by E. coli, the other being cyt bd-II [10, 11]. A major advance facilitating research on cyt bd was the recently reported structure of a cyt bd from Geobacillus thermodenitrificans [12], an enzyme that is a homologue of E. coli cyt bd-I. The Gt-cyt bd contains three subunits: CydA and CydB each contain 9 transmembrane helices, and CydS has one transmembrane helix (Figure 1A). These subunits are equivalent to E. coli CydA, CydB and CydX [13–15]. There are three hemes present, all within CydA (Figure 1). Heme b₅₅₈ is a low spin heme located near the quinol binding site and serves for the input of electrons [4–6]; 2) Heme d is a high spin chlorin that is unique to this family of enzymes and is the site where O₂ binds and is reduced to water [3, 16–19]; 3) Heme b₅₉₅ is a high spin heme which relays electrons from heme b₅₅₈ to heme d [20] and which is in van der Waal’s contact with heme d [21]. The structures of heme b and heme d are shown in Supplementary Figure S1. Previous evidence of the strong interactions between heme d and heme b₅₉₅ indicated that the two hemes functioned as a bimetallic active site where O₂ is reduced to water [1, 21, 22], analogous to the binuclear center at the active site of cytochrome oxidases in the family of heme-copper respiratory oxygen reductases [23]. The structure, however, showed that the two heme Fe’s do not form a bimetallic center [12].

Figure 1. Views of the three hemes in the crystal structure of cyt bd from Geobacillus thermodenitrificans (PDB code: 5IR6 from [12]).

(A) View from the side showing subunit I, containing the hemes, and subunit II (light blue). (B) Closer view from the side of the three hemes. (C) View from the periplasm of the three hemes, also showing the glutamate ligands GtE378 and GtE101, which correspond to EcE445 and EcE99, respectively, from E. coli cyt bd. (Figures made with UCSF package Chimera)

More remarkably, the structure reveals that GtE101 (EcE99) and GtE378 (EcE445) are adjacent to the Fe components of heme b₅₉₅ and heme d, respectively (Figure 1). (Note that all residues referred to in this work are located within the CydA subunit). In each case, the glutamic acid carboxyl group is close enough to be considered an axial ligand to the respective adjacent heme, 2.1 Å to heme b₅₉₅ and 2.7 Å to heme d [12]. Although these two hemes are in van derWaal’s contact at their periphery, the heme irons and associated glutamates for heme b₅₉₅ and heme d are well separated (Figure 1). There are no other examples of native heme proteins in which glutamate or aspartate is a heme axial ligand. Previous studies have suggested that EcE107 (GtE108) and EcE99 (GtE101) play a role in facilitating the proton transfer from the cytoplasm to the catalytic active site, providing protons for the formation of water [9, 24, 25], and it has been proposed that EcE99 is the axial ligand to heme d [26]. However, all the previous studies, performed using Ec-cyt bd-I, were performed prior to the realization that CydX (equivalent to Gt-CydS) is a third subunit in addition to CydA and Cyd B [14, 15]. In the previous studies, recombinant enzyme was obtained using plasmid-encoded cydAB, and the protein was isolated using traditional column chromatography, since affinity tags on CydA or CydB perturb the enzyme or prevent assembly. Quite likely, variable amounts of chromosome-encoded CydX were present in the old preparations of the enzyme, though the isolated enzymes contained the full complement of bound hemes and had reasonable activity. In the current work, the Ec-cyt bd-I is isolated using plasmid-encoded cydABX and isolated using a strep-tag on CydX using a recently developed protocol [15].

The improved protocol results in higher purity, greater stability and higher specific activity of the isolated enzyme, thus justifying a re-examination of previous mutations in both EcE99 and EcE445 as well as examining different mutants not previously studied. Several features not previously noted are discussed, including a salt bridge between EcE445 and EcR448 (GtE378 and GtR381) at the surface of heme d.

2. Materials and methods

2.1 Site directed mutagenesis of cytochrome bd from Escherichia coli

The construct previously prepared for purification of WT cytochrome bd with a strep-tag was used as a template [15]. Mutagenesis primers (Supplementary Information Table S1) were designed for point mutations and the Quikchange mutagenesis kit (Stratagene) was used for PCR amplification. The mutated plasmids were transformed into XL-10 Gold cells and incubated overnight at 37°C to recover colonies. The plasmid sequence was verified by sequencing (ACGT Technologies).

2.2 Cell Growth and protein purification

A single colony was inoculated into 5 ml of LB with 100 μg/ml ampicillin and incubated with shaking at 37 °C. Yeast extract and tryptone were purchased from Acumedia and NaCl from Sigma-Aldrich. The following day, the 5 ml culture was inoculated in 300 ml LB with 100 μg/ml ampicillin and grown overnight at 37 °C. On the third day, 10 ml of the secondary culture was inoculated into twenty four 2-L flasks containing 1 L LB with 100μg/ml ampicillin, each. The flasks were incubated at 37 °C while shaking at 200 rpm, until the OD₆₀₀ of the culture reached 0.6. IPTG was added to a final concentration of 0.5 mM. The temperature was then lowered to 30°C, and the culture was incubated for 8 hrs or overnight.

The fully grown cultures were pelleted by centrifugation at 8000 rpm for 8 minutes, in 500-ml centrifuge bottles. The harvested cells were resuspended in 100 mM Tris-HCl, 10 mM MgS0₄, pH 8, with DNaseI and a protease inhibitor cocktail from Sigma. The cells were then homogenized using a Bamix Homogenizer, and passed three times through a Microfluidizer cell at 100 psi to lyse the cells. The soluble fraction of the lysate was separated from the insoluble by centrifugation at 8000 rpm. Membranes were obtained by centrifugation at 42000 rpm using a Beckman 45 Ti rotor in a Beckman ultracentrifuge for 4 hours.

Membranes were resuspended in 50 mM sodium phosphate, 300 mM NaCl, pH 8, and then solubilized with 1% DDM (dodecyl maltoside) or 1% SML (sucrose monolaurate). The solubilized membranes were pelleted by centrifugation at 42000 rpm for 45 minutes to remove unsolubilized membranes. The supernatant was stirred with Strep-tactin resin (QIAGEN) for 1 hr and then loaded onto a column. The column was washed with 20 column volumes of 50 mM sodium phosphate, 300 mM NaCl, pH 8, 0.05% DDM, and the protein was eluted with 2.5 mM desthiobiotin in the same buffer. The fractions with protein were pooled and concentrated using an Amicon concentrator with a 100 kDa molecular weight filter. The concentrated protein was exchanged into 50 mM sodium phosphate, 0.05% DDM, pH 7, using the concentrator. This protein was used for further studies.

2.3 UV-visible spectroscopy

Spectra of the protein were obtained using an Agilent DW-2000 Spectrophotometer in the UV-visible region with a 1-cm pathlength cuvette. The spectrum of this air-oxidized enzyme is referred to as that of the “air-oxidized” or “as isolated” protein. The enzyme was reduced with dithionite to obtain the “reduced” spectrum. It is necessary to incubate the enzyme with an oxidant to obtain the spectrum of the fully oxidized protein.

2.4 Pyridine hemochrome and heme analysis

For the wild type or mutant enzymes, 35 μl of the enzyme solution (1–10 μM) was mixed with an equal volume of 40% pyridine with 200 mM NaOH. The “oxidized” or “reduced” spectra were obtained by adding excess ferricyanide or the dithionite, respectively, and the concentration of heme b was calculated using the method described in [27]. Since the pyridine hemochrome complex of heme d is not stable, the concentration of heme d was estimated, instead, from the spectrum of the dithionite-reduced enzyme using the difference extinction coefficient ε_{(629–670 nm)} = 25 mM⁻¹ cm⁻¹[28]. The “reduced” spectrum is preferred for this purpose rather than the reduced minus oxidized spectrum which has wider peak due to heme d that is more susceptible to perturbations by mutations. The concentration of heme b₅₅₈ was estimated using the difference extinction coefficient ε_{(561–580 nm)} = 21 mM⁻¹cm⁻¹ for the reduced minus air-oxidized difference spectrum [28]. It was assumed that all the mutants contained one equivalent of heme b₅₅₈ so, in essence, the concentration of cyt bd is based on the measurement of heme b₅₅₈.

2.5 Measurement of oxygen reductase activity

Oxygen reductase activity was measured using the Mitocell Miniature Respirometer MT200A (Harvard Apparatus). Ubiquinol-1 (Q₁, 250 μM) was used as the electron donor to measure the rate of reduction of O₂ by the wild type and mutant enzymes, and the Q₁ was maintained reduced by the presence of 5 mM dithiothreitol (DTT). KCN (300 μM) was used to test the cyanide-sensitivity of the enzymes since this will eliminate any activity due to contaminating cytochrome bo₃. Such contamination was found not to be an issue.

2.6 Electrochemical measurement of midpoint potential of the hemes

The absorption differences between selected wavelengths [29] were used to determine the redox state of heme d and heme b₅₅₈ as a function of the solution potential. A potentiometric cell was connected to an Agilent DW-2000 Spectrophotometer to simultaneously measure the solution potential and optical spectra in a set up similar to the one described [30]. A Calomel electrode was connected to a transparent cuvette and the electrode connected to a potentiometer. The cuvette contained 2.5 ml of a solution of the purified enzyme and a cocktail of redox mediators, each at a final concentration of 10 μM: anthraquinone-2-sulfate (E_m=−225 mV), duroquinone (E_m=+5 mV), p-benzoquinone (E_m=+280 mV), (2,6)-dimethylbenzoquinone (E_m=+180 mV), hexammine ruthenium (III) chloride (E_m=+50 mV), and (α-methyl)ferrocenyl methanol (E_m=+430 mV). The solution was poised at a redox potential of 440 mV (SHE) by adding small increments of potassium ferricyanide and then successively reducing the sample using aliquots of a fresh solution of sodium dithionite. At each successive addition of reductant, the UV-visible spectrum was obtained to monitor the reduction of the hemes. The data were analyzed by fitting the absorbance at selected wavelengths to the following function using OriginPro 2016 or Matlab R2015a, based on the Nernst equation.

$$y=\frac{A_1+A_0}{1+{10}^{\frac{n(X-E_{m1})}{59}}}-A_0$$

where y is the absorbance change measured as a function of the fraction of the reduced species; A₁ and A₀ are parameters related to the absorbance of the fully reduced and fully oxidized heme species; n=1, the number of electrons in the ferric/ferrous transition; X is the solution potential; and E_m1 is the midpoint potential of the species being reduced. For simplicity, the redox interactions between the hemes, which are generally small [29], were not included in the analysis.

3. Results

Four mutations of E445 were made and examined: E445A, E445Q, E445H and E445C. The structure [12] shows that the equivalent residue in Gt-cyt bd (GtE378) is about 2.7 Å from the heme d Fe. There is no reliable way to obtain with certainty the values for the heme contents of the isolated proteins. The mutations can alter the UV-visible spectra of heme d and heme b₅₉₅, and heme d cannot be assayed by extraction of the pyridine hemochrome complex due to instability. The measurement that is most reliable is the pyridine hemochrome measurement of extracted heme b (heme b₅₉₅ plus heme b₅₅₈). It is assumed that the mutations do not result in changing the amount of the low spin heme b₅₅₈ in the preparations, so all enzymes are assigned a content of 1 equivalent of heme b₅₅₈. This is reasonable since heme b₅₅₈ is the most stable of the three hemes in the enzyme, and can be present even in the complete absence of both heme d and heme b₅₉₅ [4]. If the total heme b present is greater than 1 but less than 2 equivalents per heme b₅₅₈, it is assumed the additional heme b is bound as high spin heme b₅₉₅. If the total amount of extracted heme b is more than 2 equivalents per heme b₅₅₈, it is assumed that the excess above 2 equivalents is bound at the site where heme d is normally bound in the wild type enzyme. The heme d content is measured from the UV-visible spectrum of the purified protein, subject to error due to spectroscopic perturbations due to the mutations. This information is shown in Table 1.

Table 1.

Heme content of WT, E99 mutants and E445 mutants of E. coli cyt bd

Sample	heme b558	heme b595	excess heme b	heme d	Quinol oxidase activity (%)
WT	1	1	–	0.92	100
E445A	1	1.0	–	1.511	1
E445C	1	0.85	–	1.311	3
E445H	1	0.59	–	0.74	<3
E445Q	1	0.81	–	0.59	15
E99A	1	1.0	–	0.43	72
E99H	1	1	0.653	< 0.352	< 3
E99Q	1	1	0.633	< 0.372	< 3

¹Apparent excess heme d is attributed to perturbation of the extinction coefficients of the protein-bound heme due to the mutations. The pyridine hemochrome of heme d is not stable and quantitation by this approach is unreliable.

²The spectra of protein-bound heme d in these mutants is very perturbed making quantitation by the standard extinction coefficients unfeasible. It is assumed that the excess heme b must be bound to the site where heme d normally binds and the amount of heme d bound to that site cannot be more than that needed to account for full occupancy of that binding site.

³The amount of heme b extracted from these mutants is more than can be accounted for one heme bound to each of the two heme b binding sites (heme b₅₅₈ and heme b₅₉₅). It is assumed that each of these binding sites is fully occupied and the excess heme b binds elsewhere, presumably in the site that normally binds to heme d.

3.1 Estimated heme content of the E445 mutants

Figure 2 shows the absorption spectra of the isolated proteins and indicates that some of the mutants exhibit spectral perturbations. For example, the absorption peak for heme d appears broad in the air-oxidized spectrum of E445A (Figure 2b). Furthermore, an absorption peak at 606 nm is present in the spectra of both the air-oxidized and dithionite-reduced spectra of E445A. This peak is persistent upon removal of oxygen from the sample. (Figure 2i). It was previously reported that in E445A, heme b₅₉₅ is “locked” in the ferric state (i.e., not reducible by dithionite) [28]. It is, therefore, possible that the 606 nm peak is a perturbed form of ferric b₅₉₅, and an altered spectrum of heme b₅₉₅ could, in principle, interfere with the spectroscopic quantitation of heme b₅₅₈. With these caveats, the estimation of the amounts of each heme in the isolated enzymes are shown in Table 1. Despite the quantitative uncertainties, the data show that all of the E445 mutants have a significant amount of heme d present. Two of the mutants, E445A and E445C appear to have more than 1 mol of heme d per mole of heme b₅₅₈. Since the heme d peak in the UV-visible spectra is altered and appears wider, it is reasonable to suppose that the measured excess of heme d in these mutants is due to perturbed extinction coefficients for protein-bound ferrous heme d. It appears that E445H and E445Q each have a lower but substantial amount of heme d.

Figure 2. (a-h) Air-oxidized and dithionite-reduced spectra of WT Ec-cyt bd and of EcE445 mutants and EcE99 mutants.

Insets have been used to highlight unique and unexpected spectroscopic features in the UV-visible spectra.

The simplest interpretation is that heme d is perturbed but the amount bound to the enzyme is unchanged when E445 is changed to a small residue (E445A, E445C). Substitution of an isosteric residue, E445Q, or bulkier residue, E445H, both of which can provide nitrogen ligation to heme d, reduces the amount of heme d that is bound, but not drastically. Previous examination of E445Q showed no heme d present [31], but the improved protocols to obtain the enzyme in a more stable form in the current work clearly indicates the heme d remains in this mutant to a significant extent.

Previous studies on E445A [31], on the other hand, reported results similar to those presented here. The presence of heme b₅₉₅ is not obvious from the spectra of any of the mutants, but the total content of heme b by pyridine hemochrome analysis shows more heme b beyond the 1 equivalent of heme b₅₅₈. It is reasonably assumed that this “excess” heme b is due to heme b₅₉₅. By this measure, all the E445 mutants have a significant amount of heme b₅₉₅, with E445H having the lowest content of heme b₅₉₅ (0.59) (Table 1).

3.2 Quinol oxidase activity of the E445 mutants

All of the E445 mutants have much lower specific activity (normalized per heme d) than the WT, shown in Table 1. Essentially, E445A, E445C and E445H are inactive. The only mutant with significant activity is E445Q, which retains about 15% of the specific activity. Hence, although the importance of E445 for enzyme function is evident in this and in previous studies, the fact that some activity is retained in the damaged E445Q mutant demonstrates that this residue is not absolutely essential for function.

3.3 Heme content of the E99 mutants

Previous studies of E99 mutants indicated the absence of heme d [25, 26, 29], leading to the conclusion that E99 is the axial ligand to heme d [26]. The current work shows that heme d is present, however, in the E99A, E99H and E99Q mutants. The spectroscopic features of E99A (Figure 2f) are like WT cyt bd so the estimated the content of heme d (0.43) is likely reliable. The spectra of the E99H and E99Q mutants indicate that heme d is present, but the spectrum of the dithionite-reduced form is very red-shifted (peak at 665 nm instead of 630 nm) and not possible to quantify. An upper bound to the amount of heme d present in the E99H and E99Q mutants (Table 1) was estimated by assuming the “excess” heme b (beyond 2 equivalents per mol enzyme) is bound to the heme d-binding site, and assuming the capacity of that site is 1 equivalent (heme b plus heme d). This is discussed further below.

A significant portion of the “as isolated” air-oxidized form of WT cyt bd has O₂ bound to ferrous heme d, indicated by a peak in the absorption spectrum at 646 nm. When the air-oxidized form of the WT cyt bd is made anaerobic (e.g., vacuum-treated or after passing Argon over the sample) the amplitude of the peak at 646 nm diminishes (Supplementary Figure S2). The air-oxidized (as isolated) form of both the E99H and E99Q mutants also have a peak at 646 nm, but when the samples are made anaerobic by evacuating the air, the absorption at 646 nm is not diminished. Upon reduction by dithionite, the absorption peak shifts from 646 nm to 665 nm (Supplementary Figure S3). The origin of this spectroscopic feature is not known. Perhaps the oxygen-bound ferrous heme d in the “as isolated” form is not dithionite-reducible in these mutants or perhaps the spectrum is red-shifted due to altered hydrogen bonding.

Both the E99Q and E99H mutants have a large excess of heme b (>2.6) over the expected maximum of 2 equivalents per mol of enzyme. One possible explanation is that a substantial portion of the site that normally binds heme d is bound, instead, to high spin heme b. This is what is found for the so-called cyanide-insensitive oxidases (CIOs), a subset of cyt bd enzymes which lacks heme d and only contains heme b [1, 2]. This is consistent with the hypothesis that heme d is formed in situ from bound heme b by hydroxylation of porphyrin ring (see Supplementary Figure 1). Disruption of proton and/or electron transfer in the E99H and E99Q mutants may inhibit the conversion of heme b to heme d. This proposal is consistent with the observation that the isolated, air-oxidized forms of both the E99H and E99Q mutants have a substantial fraction of heme b₅₅₈ in the reduced (ferrous) state, suggesting obstruction of electron transfer from heme b₅₉₅ to heme d. The estimated amounts of heme d (Table 1) in both the E99H and E99Q mutants is based on the assumption that the amount of total heme b that is in excess of that accounted for by full occupancy of the heme b₅₅₈ and heme b₅₉₅ sites is due to binding of heme b to the site that normally binds to heme d, and the remaining capacity to bind heme d at that site is limited to a total capacity of one heme at that site. It is noted that the data show that neither E99Q or E99H form strong axial ligands with heme b₅₉₅ since that would result in a low spin heme that would be optically indistinguishable from heme b₅₅₈ which is used to measure the enzyme concentration. The consequence would be that the amount of the total extracted heme b beyond that assigned to heme b₅₅₈ would be very low, which is the opposite of what is observed with the E99Q and E99H mutants (Table 1).

The dramatic influence of the E99 mutants on the stability and properties of the heme d is consistent with previous work [25, 26, 29], which concluded that E99 is the axial ligand to heme d. The X-ray structure [12], however, shows that E99 is adjacent to heme b₅₉₅.

3.4 Quinol oxidase activity of the E99 mutants

Of the three E99 mutants examined, only E99A has ubiquinol-1 oxidase activity (Table 1). The specific activity is quite high when expressed per equivalent of heme d, about 72% of the wild type value. Previous work also demonstrated that both the E99A and E99L mutants have significant enzyme activity when assayed within E. coli membranes (13% and 7%, respectively) despite the reported content of heme d being too low to be quantified [25].

As concluded for E445, the current work demonstrates that although E99 is very important to the function of cyt bd, it is not absolutely required.

3.5 Perturbation of the midpoint potentials of the hemes by the E445Q, E445A and E445C mutants

Spectro-electrochemical redox titrations were performed to determine the midpoint potentials (pH 7) of heme b₅₅₈ and heme d due to mutating E445. Previous work has shown that the midpoint potential of heme b₅₅₈ is sensitive to the environment and depends on the detergent that is present when examining the isolated proteins. Hence, measurements were made with two different detergents, dodecylmaltoside (DDM) and sucrose monolaurate (SML). The midpoint potentials of heme d and heme b₅₅₈ were resolved, but not that of heme b₅₉₅, except for the WT cyt bd. The data are shown in Table 2 and in Figure 3.

Table 2. Midpoint potentials of heme d and heme b₅₅₈ of WT and E445 mutants of E. coli.

cyt bd determined in detergents dodecylmaltoside (DDM) or sucrose monolaurate (SML). Midpoint potentials are vs the Standard Hydrogen Electrode (SHE).

Sample	heme b558 (mV)	heme d (mV)
in DDM detergent
WT	88	246
E445A	46	265
E445Q	183	199
in SML detergent
WT	183	289
E445C	156	222

Figure 3. Spectro-electrochemical titrations of the WT Ec-cyt bd and the EcE445Q and EcE445C mutants.

The redox state of heme d was monitored by the absorption difference (ΔA) as a function of solution potential (panels A, C and D). ΔA was measured by using the following linear combination of wavelengths (A_{629 nm} – 0.5*(A_{660 nm} + A_{608 nm})). Similar trends were observed with the wavelength combinations proposed in [29]. The redox state of heme b₅₅₈ was monitored by the absorption difference (ΔA) as a function of solution potential (panel B), where ΔA = A_{561 nm} – 1.6*A_{573 nm} + 0.06*A_{629 nm}). Again, all trends were verified by comparison to midpoint potentials calculated using other wavelength combinations mentioned in [29]. Data were fit to a 1-electron reduction following the Nernst equation and the results are summarized in Table 2.

The most striking observation is that the midpoint potential of heme b₅₅₈ is sensitive to the mutation of E445, most dramatically for E445Q in the presence of DDM where the mutation results in a 100 mV increase in the midpoint potential of heme b₅₅₈. The E445A and E445C mutants result in significantly lower midpoint potentials of heme b₅₅₈. Not surprising, the mutations also perturb the midpoint potential of heme d. The midpoint potential of E445Q in DDM is lowered from 246 mV to 199 mV, which is the expected direction of change if one assumes E445 is deprotonated and E445Q resembles the protonated species. Protonation of E445 is expected to stabilize the reduced form of heme d. Similarly, in SML detergent, the midpoint potential of heme d drops from 289 mV to 222 mV for E445C compared to WT.

4. Discussion

The X-ray structure of cyt bd from G. thermodenitrificans [12] revealed several features that were unexpected from previous biochemical and biophysical studies, mostly on the E. coli homologue. The most unusual feature is that two of the three heme components of the enzyme, heme d and heme b₅₉₅, each have a glutamic acid located at positions where the carboxyl side chains could ligate to the heme irons: GtE101 (EcE99) adjacent to the Fe of heme b₅₉₅ and GtE378 (EcE445) adjacent to the Fe of heme d. Considering that no other native heme protein has been reported to have either Asp or Glu as an axial ligand, the current work is aimed at determining the functional roles of these two glutamic acid residues.

4.1 The glutamates are important but not essential for activity

The ubiquinol-1 oxidase activities of the mutants, shown in Table 1 demonstrate that despite the evident importance of these two glutamate residues, they are not absolutely required for activity. The E445Q mutant retains 15% of the activity, normalized per heme d, despite the moderately decreased amounts of both heme d and heme b₅₉₅. It is also noted that there is a clade of cyt bd in which E445 is replaced by a cysteine, although the E445C mutant of Ec-cyt bd is not functional. Table 1 also shows that the E99A mutant, which appears to have less than half of the heme d present, retains 72% of the ubiquinol-1 oxidase activity (per heme d). The new affinity-tag based protein preparation makes it apparent that although the E99A and E445Q mutants are significantly perturbed, they do maintain substantial ubiquinol-1 oxidase activity.

4.2 The glutamates are not strong ligands to the hemes

The heme analyses of the mutant enzymes are also summarized in Table 1. Despite the ambiguities due to mutation-induced changes in the heme d extinction coefficients, with the reasonable assumption that the content/spectra of heme b₅₅₈ are unchanged, the data show that none of the mutations completely eliminate either heme d or heme b₅₉₅. E445A and E445C appear to have excess heme d, which likely is due to altered extinction coefficients, and the larger residue substitutions, E445Q and E445H, have less heme d, 74% and 59%, respectively.

Similarly, the mutations of E99, which is adjacent to heme b₅₉₅, do not result in eliminating the heme b at this site. E99A has the same total content of heme b as does the wild type, and both E99H and E99Q have excess heme b, possibly indicating that the site that normally binds heme d is partially filled by heme b, as occurs in the so-called “cyanide-insensitive oxidase” variants of cyt bd [1]. The heme d that is bound to the E99H and E99Q mutants have spectra that are highly distorted compared to the WT, precluding quantitation. Oddly, the perturbations due to the E99 mutations appear to have a greater influence on the content and properties of the heme present at the heme d binding site than on the heme b₅₉₅ site, consistent with previous data that was used to support the assignment of EcE99 (GtE101) as the axial ligand to heme d [26]. Likewise, the most significant consequence of the E445A mutation is not on the properties of heme d, but rather the inability to reduce heme b₅₉₅, which is locked in the ferric state (i.e., not reduced by dithionite) [28]. The perturbation of the adjacent heme by these glutamate mutations could be mediated either through the protein or directly by the very close interaction of hemes b₅₉₅ and heme d, which are in van der Waals contact at the porphyrin edges [12] (Figure 4). One manifestation of the intimate electronic coupling between the two hemes is that photolysis of CO from ferrous heme d changes the absorption spectrum of heme b₅₉₅ within a few picoseconds [22].

Figure 4. Proximity of the three hemes in Gt-cyt bd.

Hemes d and heme b₅₉₅ are in Van der Waal’s contact. Heme b₅₅₈ is adjacent to heme d and separated by the indol ring of GtW374 (EcW441). Also shown are GtE101 (EcE99) and GtE378 (EcE445) that form weak axial ligands to heme b₅₉₅ and heme d, respectively. (Figure made with VMD. [60])

It is concluded that E99 and E445 are not primarily responsible for the stability of the hemes to which they are adjacent. This can be understood in terms of the relatively weak interactions between the heme Fe and carboxylate moiety which is likely to be primarily ionic rather than covalent due to the low basicity of the carboxyl group [32]. These conclusions are entirely consistent with previous spectroscopic studies [33, 34].

4.3 Previous studies show that heme b₅₉₅ and heme d are high spin

Resonance Raman studies indicate that heme b₅₉₅ from Ec-cyt bd is 5-coordinate high spin and that His19 is the strong axial ligand to heme b₅₉₅[34, 35]. In addition, CN⁻ does not interact with heme b₅₉₅ [34] and only 5% of ferrous b₅₉₅ binds CO when CO is present in solution [36]. In Gt-cyt bd however, CO can bind to heme b₅₉₅ and make it 6-coordinate. The extent of excitonic coupling between heme b₅₉₅ and heme d in Gt-cyt bd is also reduced, indicating structural differences in the vicinity of heme d and heme b₅₉₅ for the enzymes from these organisms [37]. Hence, heme b₅₉₅ has one strong histidine ligand on one side of the heme and is blocked from binding to exogenous ligands on the opposite side to varying extents depending on the ligand and the source of the enzyme. The structure suggests that the block against ligand binding to heme b₅₉₅ is the weak glutamate endogenous ligand.

In both Ec-cyt bd and Gt-cyt bd, heme d has been shown to be high spin and always 5-coordinate [34], while EPR and Raman spectra of Ec-cyt bd indicate that the endogenous ligand to heme d is weak and likely displaced by oxygen [34, 35, 38]. There is no evidence of any ligand on the opposite side of heme d from the weak glutamate ligand.

All of the data lead to the conclusion that E445 and E99 are not strong ligands to heme d and heme b₅₉₅, respectively, conclusions bolstered by studies examining models of heme proteins with carboxylate-containing compounds as ligands.

4.4 Model studies of glutamate as a heme ligand

The H93G mutant of sperm whale myoglobin has been used to replace the proximal ligand (H93) with several different exogenously added small molecule ligands that can be diffused into the cavity, including acetate[32] and benzoate[33]. The acetate adduct is 6-coordinate high spin with water ligating the heme Fe³⁺ on the distal side, stabilized by hydrogen bonding to the distal H64. The geometry of the Fe³⁺-acetate binding is inconsistent with an interaction between an electron pair forming a Fe-O bond. The Fe-O-C angle is 152° and the Fe is not in-plane, defined by the O=C-O carboxylate. The acetate binding is weak, as measured by the dissociation constant. The Fe-O-C bond angle of E99 and heme b₅₉₅ is also 152° and the carboxylate plane also does not include the heme Fe. The Fe-O bond length is about 2.1 Å for E99 interacting with heme b₅₉₅, similar to the 2.2 Å for the proximal acetate-heme Fe-O bond in myoglobin. The interaction between E445 and heme d appears to be even weaker, with the Fe-O bond length of about 2.7 Å and an Fe-O-C angle of 87°, indicating that the plane of the carboxylate is nearly parallel to the heme d plane, not consistent with a strong Fe-O bond.

Iron-strapped porphyrin model compounds with carboxylic acid groups hanging over the coordination site have also been examined [39]. A pyridyl residue serves as the proximal ligand to the Fe²⁺ heme and carboxyl groups are dangling within the distal pocket. Model compounds were examined in which the overhanging carboxyl can directly interact with the Fe²⁺ as well as compounds where the geometry precludes direct carboxyl interaction with the Fe²⁺. Compounds with hanging carboxyls which can interact directly with the Fe²⁺ are 6-coordinate high spin. Of particular interest is that these compounds bind O₂ and CO. It is concluded that O₂ and CO can displace one of the axial ligands in the 6-coordinate model compounds and that it is the carboxylate that is almost certainly the displaced ligand. Furthermore, it is suggested that the hanging carboxyl in the distal pocket can stabilize bound O₂ by either dipolar interactions and/or hydrogen bonding. These studies support a possible role of E445 to stabilize O₂ bound to ferrous heme d.

4.5 On which side of heme d does O₂ bind?

Although it is clear that O₂ binds to heme d, significant questions remain about the nature of O₂ binding in the cyt bd active site. The primary question is whether O₂ binds on the hydrophilic side of heme d, displacing EcE445, or whether O₂ binds on the hydrophobic side of the heme. Cyt bd has a high affinity for oxygen, 0.28 μM for Ec-cyt bd [17] and 0.5 μM for cyt bd from Azotobacter vinlandii (Av-cyt bd) [8], comparable to sperm whale myoglobin (1 μM) but less than Ascaris hemoglobin (3 nM) [40]. The very high affinity of O₂ to Ascaris hemoglobin is primarily due to a very slow rate constant for dissociation, 0.004 s⁻¹ vs 15 s⁻¹ for sperm whale myoglobin [40]. The second order rate constant for formation of the O₂ adduct is actually slower for Ascaris hemoglobin (1.5 × 10⁶ M⁻¹s⁻¹) than for sperm whale myoglobin (1.7 × 10⁷ M⁻¹s⁻¹). In comparing these cases with cyt bd, it is clear that high affinity for O₂ of cyt bd is attributable to the very large apparent rate constant for formation of the complex, 2 × 10⁹ M⁻¹s⁻¹, which is close to the diffusion limit [1, 8, 17, 41] and 100-fold faster than for sperm whale myoglobin. The rate constant for the dissociation of O₂ from the mixed valence state of Ec-cyt bd $(b_{558}^{3+}{b}_{595}^{3+}{d}^{2+})$ is 78 s⁻¹ [42], 5-times faster than for sperm whale myoglobin (12 s⁻¹) [43]. Effectively, there is no significant barrier impeding O₂ from getting from solution to the heme d Fe²⁺. The extremely rapid rate constant for association is consistent with the short pathway (6 Å to 8 Å) to the surface and the apparently open channel that is observed in the structure [12] (Figure 5). Visually, both sides of heme d appear accessible (Figure 5), aside from the block provided by GtE378 (EcE445) on the hydrophilic face of the heme [12]. The accessibility can be addressed quantitatively using molecular dynamics as it has been applied to other O₂ reductases [44].

Figure 5. Views of the oxygen binding site in Gt-cyt bd.

(A) The surface representation of subunits I (light green and II (blue) shows the opening leading from the periplasm to the cavity containing heme d. (B) A more detailed view of the opening, with heme d and its Fe (red) visible and with GtR11 (EcR9) indicated on the rim of the opening. (C) A view showing the α-helix that contains both GtR11 (EcR9) and the axial ligand GtH21 (EcH19) to heme b₅₉₅. This helix runs over the hydrophobic side of heme d. Also shown are GtR381 (EcR448) and GtE378 (EcE445) that form a salt bridge on the hydrophilic side of heme d. (Figure made with Chimera)

High affinity for O₂ can be mediated by hydrogen bond interactions, as is the case in myoglobin where several residues in the distal pocket form hydrogen bonds with O₂ when it binds to Fe²⁺[40, 45]. Genetic mutations to generate a totally hydrophobic distal pocket of myoglobin does not preclude O₂ binding, but the affinity is very low [43, 45]. However, the relatively high affinity for O₂ exhibited by cyt bd does not in itself exclude the possibility of binding to the hydrophobic side of the heme. The affinity for O₂ also depends on the degree to which the Fe²⁺ is available to react. In myoglobins the plane of the imidazole of the proximal axial histidine ligand eclipses the porphyrin nitrogens of the heme [46]. This restrains the movement of the Fe²⁺ with respect to the heme plane and substantially reduces the reactivity towards ligands on the distal side because the Fe²⁺ is pulled away from this side of the heme. The lack of such steric constraints due to the orientation of the proximal histidine is invoked to explain the high O₂ affinities of leghemoglobin and the L16A mutant of cyt c′ [47, 48]. Since the glutamate ligand on the opposite face of heme d is a weak ligand, it is unlikely to draw Fe²⁺ towards itself and limit the access of Fe²⁺ to O₂ on the hydrophobic side.

In the case of cyt c′, the wild type cyt c′ protein does not bind to O₂ because the Fe²⁺ binding site is blocked by the side chain of L16 and, when a space is generated by the L16A mutation, O₂ binds with very high affinity [49]. In Gt-cyt bd, there is a leucine (L18) on the hydrophobic side of heme d whose side chain is 3.9 Å from the heme d Fe which could restrict O₂ binding to the hydrophobic side of heme d, but this residue is not conserved, however, and is replaced by an alanine in the Ec-cyt bd.

Despite the structure showing a much more open binding site for O₂ on the hydrophobic side of heme d and GtE378 (EcE445) blocking access to the Fe on the hydrophilic face of the heme, it is not likely that the catalytic active site is on the hydrophobic side of heme d. There are no residues on the hydrophobic side, conserved or otherwise, to provide protons for the chemistry of reducing O₂ to water or to stabilize catalytic intermediates. Accompanying electron transfer, timely and rapid proton delivery to the oxygenated states of heme d that are intermediates during the catalytic cycle is essential. Furthermore, the highly oxidized oxoferryl state of the heme is sufficiently stabilized by the protein that, in the absence of a reductant, the enzyme can be isolated in this state [18, 19]. Cyt bd also has significant quinol peroxidase activity, reducing hydrogen peroxide to water [50]. Considering the conserved sets of catalytically critical residues in the distal pockets of peroxidases, catalases and in other O₂ reductases, the expectation that there must be conserved residues involved in the generation and stabilization of transient oxygenated states of heme d during the catalytic cycle is compelling.

If O₂ binds to heme d Fe²⁺ on the same side as EcE445 (GtE378), it must displace EcE445 (GtE378). It is likely that the structure [12] represents the fully oxidized state of cyt bd $(b_{558}^{3+}{b}_{595}^{3+}{d}^{3+})$ and it is possible that GtE378 (EcE445) is displaced upon reduction of heme d to the ferrous state. Even if this is not the case, there are numerous examples of O₂ binding to hexacoordinate heme proteins by displacement of a weakly bound distal ligand [51, 52], and this is also speculated to be the case for O₂ binding in the iron-strapped porphyrin models [39].

4.6 The hydrophilic face of heme d has a conserved GtE378-R381 (EcE445-R448) salt bridge

On the polar side of heme d, Ec E445 (GtE378) is part of a highly conserved region of CydA: ⁴⁴⁰GWFMT⁴⁴⁵EFGRQP⁴⁵¹W in Ec-cyt bd and ³⁷³GWYLA³⁷⁸EVGRQP³⁸⁴W in Gt-cyt bd. The X-ray structure shows (Figure 6) that this region of the protein is a helix that sits on the hydrophilic face of heme d. Besides EcE445 (GtE378), this stretch contains EcW441 (GtW374) which has been suggested [31] as being important for electron transfer between heme b₅₅₈ and heme d, though direct electron transfer has not been demonstrated. What has not been noted is that GtR381 (EcR448) forms a salt bridge with GtE378 (EcE445) at the surface of heme d (Figure 6A). Although EcE445 is replaced in by a cysteine in a small clade of organisms (e.g., Bacillus cereus), this arginine (EcR448) is invariant. The modification is specific to certain organisms of the order Bacilalles and phylum Cyanobacteria. It is not apparent what function this mutation would serve but in our experiments, this modification is not functional. We suspect that some corresponding mutations that are required to make this mutation viable are missing in our mutations but we are not able to identify what these could be from sequence alignments. The EcE445-R448 (GtE378-R381) salt bridge in cyt bd can be expected to partially neutralize the negative charge on the carboxylate of EcE445 and further weaken the interaction between EcE445 with the heme d Fe. It is also plausible that one role of this arginine (EcR448) is to help to stabilize the negative charge on the oxygen moiety in the oxoferryl intermediates ( $F{e}_d^{4+}=O^{2-}[{\text{Por}}^{+•}]$ and $F{e}_d^{4+}=O^{2-}$) [18] that are equivalent to compounds 1 and 2 of peroxidases [53]. In the case of myoglobin from Aplysia limacina, an arginine in the distal pocket provides the stabilizing hydrogen bond for O₂ bound to the ferrous heme Fe²⁺[46]. EcE445 (GtE378) and EcR448 (GtR381) could also interact and provide catalytically important hydrogen bonds, as occurs in the proximal pocket of catalases[53]. Furthermore, the side chain of GtN150 (EcN147), another highly conserved residue, is only 3.8Å from the carboxylate of GtE378 (EcE445) (Figure 6) and could be functionally important upon O₂ binding and displacement of GtE378 (EcE445). It is also likely that water molecules are present on the hydrophilic face of heme d and these could participate in important hydrogen bond networks and proton wires. An X-ray structure with improved resolution and of different redox states of the protein will be necessary to clarify this.

Figure 6. Views of GtE378 and GtR381 in Gt-cyt bd.

(A) The salt bridge between GtE378 (EcE445) and GtR381 (EcR448) at the hydrophilic side of heme d. (B) GtE378 is 6.89 Å from the propionate of heme b₅₅₈. (C) The conserved α-helix that runs along the hydrophilic side of heme d, ³⁷³GWYLA³⁷⁸EVGRQP³⁸⁴W in Gt-cyt bd. (D) The helix is colored in yellow; GtR381, GtE378, GtW374and GtI370 are indicated along with GtN150 (EcN147). (Figure made with Chimera)

If O₂ were to react on the hydrophobic side of heme d, the axial ligation of EcE445 (GtE378) hydrogen bonded to EcR448 (GtR381) would provide little electron donation (push) to stabilize the oxoferryl oxidation states ( $F{e}_d^{4+}=O^{2-}[{\text{Por}}^{+•}]$ and $F{e}_d^{4+}=O^{2-}$) of heme d that are intermediates in the reaction catalyzed by the enzyme[18]. In contrast, the “pull” provided by E445/R448 for O₂ reacting on the same face of heme d would favor formation of the high oxidation states of the heme d Fe.

Until further data are available, it seems reasonable to model the O₂ binding to heme d on the same face as EcE445 by displacing the glutamate, which then plays an important role in hydrogen bonding to oxygenated states of heme d and in proton transfer during the formation of water at the active site.

4.7 Redox state of heme b₅₉₅ controls O₂ access to heme d

While O₂ has easy access to heme d, the interaction between heme d and heme b₅₉₅, adds a level of complexity to gas ligand binding. The crystal structure shows the two hemes are in van der Waals contact [12] (Figure 4), and this close contact is manifested by both the excitonic coupling apparent in CD spectra [37], as well as rapid spectroscopic changes of heme b₅₉₅ upon photolysis of CO bound to reduced heme d [54]. Interestingly, it was shown with Ec-cyt bd that CO does not bind heme d and heme b₅₉₅ simultaneously, and CO-photolysis studies suggest negative cooperativity in ligand binding between these sites [55].

An additional manifestation of this heme-heme interaction is the fate of CO that is photolyzed from ferrous heme d. When CO is photolyzed from the fully reduced form ( $b_{558}^{2+}{b}_{595}^{2+}{d}^{2+}-CO$) of cyt bd, all of the CO is expelled from the protein and recombines with heme d mono-exponentially with a time constant of 12 μs [54]. When the same experiment is performed with the mixed valence form ( $b_{558}^{3+}{b}_{595}^{3+}{d}^{2+}-CO$), 70% of the bound CO is not completely expelled from the protein but, instead, exhibits geminate recombination to one or more sites within the protein with a time constant of 14 ns. These results are explained by the existence of cyt bd in an “open” form and a “closed” form. The fully reduced enzyme is expected to be in “open” form, as demonstrated by the free exchange of gas ligands and fast rate of thermal dissociation of CO and NO. In the mixed valence form, it appears that the enzyme is in equilibrium between the “open” and “closed” states. This supposition is supported by the kinetics of O₂ binding to the 30% of the CO-photolyzed mixed-valence form of the enzyme from which CO exits the enzyme [8]. The rate of O₂ binding is hyperbolic with respect to the concentration of O₂, consistent with a model in which the rate of O₂ binding is limited by the equilibration between closed and open states [8]. Experiments with the Av-cyt bd suggest that about 67% of the enzyme is in the “closed” state in the mixed valence form of the enzyme, and that the time for interconversion between the open and closed states is in the microsecond range [8]. Importantly, the geminate CO recombination kinetics of the EcE445A mutant, in which heme b₅₉₅ is “locked” in the ferric state, resembles the mixed valence state of the enzyme. Hence, it is specifically the redox state of heme b₅₉₅ that regulates ligand access to heme d.

One possibility is that the opening of the apparent O₂ delivery channel [12] is occluded when heme b₅₉₅ is oxidized. The EcH19 (GtH21) axial ligand to heme b₅₉₅ is part of an α-helix that includes a highly conserved EcR9 (GtR11) which forms part of the rim of the opening of the putative O₂ channel [12], providing a plausible link between the redox state of heme b₅₉₅ and access of O₂ to heme d (Figure 5C). A change in conformation of EcR9 (GtR11) in response to the redox state of heme b₅₉₅ could alter the access of O₂ to heme d. In addition, the same α-helix passes over the hydrophobic face of heme d and contains GtL18 that is about 3.9 Å away from the Fe²⁺. Movement of this helix due to a redox change in heme b₅₉₅ could also open a pocket on the hydrophobic side of heme d where photolyzed CO could enter and then rapidly re-bind to heme d at the physiologically relevant hydrophilic side of heme d. This could explain the 70% geminate recombination of CO that is observed [54].

The presence of the conserved arginine EcR9 (GtR11) on the rim of the opening to the O₂ binding site on heme d (Figure 5B) also might prevent protons from the bulk solution on the periplasmic side of the membrane gaining access to GtE378 (EcE445). This is a requirement for the enzyme to generate a proton motive force, utilizing protons only from the cytoplasm for the chemistry of reducing O₂ to water.

4.8 Proton transfer and the catalytic cycle

The voltage that is generated during turnover of cyt bd is primarily due to proton transfer from the cytoplasmic side of the membrane to the active site at heme d [28]. The proton pathway has been partially sketched out by previous mutagenesis studies [24, 25, 28, 56, 57] as well as by observation of the X-ray structure [12] (Figure 7). The structure shows that EcE445 (GtE378) is the most likely proton donor to oxygenated states during catalysis leading to the formation of water. EcE445 (GtE378) also likely forms a hydrogen bond to stabilize the oxygenated states of heme d (e.g., ferrous-oxy and oxoferryl states).

Figure 7. Proposed proton channel in Gt cyt bd.

A series of conserved hydrophilic residues along with the propionate of heme d and internal water molecules that are not observed in the X-ray structure line a proton channel conveying protons from the cytoplasm to GtE378 (EcE445) and from there to the oxygenated species at the active site. (Figure made with VMD)

Three conserved glutamatic acids are proposed to be critical components of the proton channel pathway: EcE99 (GtE101), EcE107 (GtE108) and EcE445 (GtE378). EcE107 (GtE108) has been shown to be protonated in both the fully reduced and fully oxidized forms of the enzyme [56] and essential for catalytic activity [57]. It is likely that EcE107 (GtE108) is the proton donor to EcE99 (GtE101), which we propose to become protonated upon reduction of heme b₅₉₅. Protonated EcE99 (GtE101) then becomes the proton donor to EcE445 (GtE378) accompanying electron transfer from heme b₅₉₅ to heme d. Protons from Ec445 are used for forming water at the heme d active site. The pathways for proton transfer must involve water molecules that are not resolved in the structure, as well as propionate A of heme d, which lies midway between EcE99 (GtE101) and EcE445 (GtE378), about 7 Å from each in the structure [12] (Figure 7). Once EcE445 (GtE378) is displaced by O₂ it could move closer to the propionate, but water molecules, perhaps hydrogen bonded to EcN147 (GtN150) would still be required to provide a pathway for protons to diffuse using the Grotthus-type mechanism, e.g., [58].

It is speculated that during catalytic turnover, the only source of protons to the active site must come through this pathway from the cytoplasm. Since EcE107 (GtE108) has been shown to be protonated in both the oxidized and reduced state of cyt bd [56], it is always prepared to provide a proton to EcE99 (GtE101) whenever heme b₅₉₅ receives an electron from heme b₅₅₈. Once heme b₅₅₈ is reduced, the electron goes preferably to heme b₅₉₅ because a proton is readily provided to EcE99 (GtE101) to stabilize the ferrous form of heme b₅₉₅. Presumably, EcE107 (GtE108) is rapidly reprotonated from the cytoplasmic channel. The net proton transfer from the cytoplasm to EcE99 (GtE101) is responsible for the voltage generation accompanying this step of the reaction [9, 28]. Once the proton and electron are at heme b₅₉₅ and its associated glutamate, EcE99 (GtE101), both the electron and proton can be transferred to heme d and its associate glutamate EcE445 (GtE378), and then to the oxygenated species. In this way, the sequence of electron transfer from heme b₅₅₈ to heme d is never direct, but requires heme b₅₉₅, and every electron that goes to heme d is accompanied by a proton that comes through the proton channel from the cytoplasm. Electron transfer between heme b₅₉₅ and heme d is not electrogenic because the electron is always accompanied by a proton exchanged between EcE99 (GtE101) and EcE445 (GtE378). The reverse electron transfer from heme d following photolysis of the CO adduct of the one-electron reduced cyt bd ( $b_{558}^{3+}{b}_{595}^{3+}{d}^{2+}-CO$

) also goes initially to heme b₅₉₅ before equilibrating with heme b₅₅₈ [20, 28] for the same reason. Electron transfer from heme d to heme b₅₉₅ is accompanied by proton transfer from EcE445 (GtE378) to EcE99 (GtE101) without generating a voltage [28]. Electron transfer from heme b₅₉₅ to heme b₅₅₈ requires proton transfer via EcE107 (GtE108) to the cytoplasm, which is slower and does generate a voltage [28]. The very close interaction between heme d and heme b₅₉₅ is the most probable explanation for the distorted spectra of heme d in the E99H and E99Q mutants. In E99A, it is likely that a water molecule takes the place of E99 in the proton channel, maintaining the supply of protons to the active site.

Figure 8 shows a schematic of the species in the catalytic cycle with the speculated protonation states of the two glutamates included. This builds on previous models that have been proposed [1, 12, 18] and includes speculation based on the considerations presented in this work. Further studies will be required to address the many unanswered questions and test this model.

Figure 8. Proposed catalytic mechanism of cyt bd.

Residue numbers are those of Ec-cyt bd. It is assumed that electron transfer to heme d is always accompanied by proton transfer from EcE99 to EcE445, and electron transfer to heme b₅₉₅ is always accompanied by proton transfer from EcE107 to EcE99, and EcE107 is rapidly reprotonated by protons delivered through the proton channel from the cytoplasm (in). Protons released by oxidation of ubiquinol (QH₂) are released to the periplasm (out). Species A₁ is the stable one-electron reduced oxy form; Species A₃ is an unobserved, transient 3-electron reduced species with the O₂ still bound to ferrous heme d.; Compound I (Cpd I) is the oxoferryl species with a porphyrin cationic radical at heme d; Species F is Compound II, the oxoferryl form following reduction of the porphyrin radical. Species A₁, Cpd I and F have been observed along with the mixed valence species (MV).

4.9 Role of E445 in the interaction between heme d and heme b₅₅₈

In Table 2, it is shown that mutations of EcE445 (particularly E445A and E445Q) affect the midpoint potential of heme b₅₅₈. It has also been shown that reduction by ubiquinol-1 or by TMPD (N,N,N′,N′-tetramethyl-p-phenylenediamine) via heme b₅₅₈, of the fully oxidized form of Ec-cyt bd ( $b_{558}^{3+}{b}_{595}^{3+}{d}^{3+}$) is 1000-fold slower than reduction of the oxoferryl form ( $b_{558}^{3+}{b}_{595}^{3+}{d}^{4+}=O^{2-}$), and much too slow to be compatible with the rate of steady state catalysis [59]. The rate of reduction of the mixed valence oxy complex ( $b_{558}^{3+}{b}_{595}^{3+}{d}^{2+}-O_2$ or $b_{558}^{3+}{b}_{595}^{3+}{d}^{3+}-O_2^{-}$) by ubiquinol must also be much faster than observed for the all-ferric form of the enzyme to be compatible with steady state turnover since this is also an obligatory step in the catalytic cycle (Figure 8). The implication is that the redox state of heme d controls the rate of reduction of heme b₅₅₈. Although electrostatics might play a role in some of these observations, conformational coupling appears likely. The structure shows that the edge-to-edge distance between the two hemes is only about 6 Å, with a tryptophan indol ring (GtW374 or EcW441) sandwiched between them (Figure 4). A perturbation of the position or orientation of heme d could be readily sensed by heme b₅₅₈. The structure suggests two other plausible sources of physical perturbation. The crystal structure places GtE378 (EcE445) 6.9 Å from the propionate group of heme b₅₅₈ (Figure 6B). Oxygenation of heme d will displace GtE378 (EcE445) in the direction of heme b₅₅₈ and where it could make a hydrogen bond to its propionate group of heme b₅₅₈, perhaps with an intervening water. Stabilizing reduced heme b₅₅₈ with a hydrogen bond could conceivably increase the midpoint potential of heme b₅₅₈ and accelerate its rate of reduction by a reductant. A second possibility is that any perturbation of GtE378 (EcE445) by either reduction of heme d or by mutation, could be conveyed to heme b₅₅₈ through the α-helix that contains GtE378 (EcE445) which contains two nearby residues that are in direct contact with heme b₅₅₈, GtW374 (EcW441) and GtI370 (EcI437) (Figure 6C).

5. Conclusion

The conclusion from the mutagenesis work presented in this study is that neither EcE445 (GtE378) nor EcE99 (GtE101) are strongly ligated to heme d and heme b₅₉₅, respectively, and they are not primarily responsible for the stability of the bound hemes. It is concluded that it is most likely that O₂ binds to the face of heme d on which EcE445 (GtE378) is located and that, therefore, EcE445 (GtE378) must be displaced upon O₂ binding. We also point out for the first time the presence of a conserved salt bridge between EcE445 (GtE378) and EcR448 (GtR441) on the hydrophilic face of heme d. Both of these residues are likely important to stabilize oxygenated states of heme d by electrostatics and/or hydrogen bonding. We speculate that reduction of heme d is accompanied by protonation of EcE445 (GtE378), which would break the salt bridge and be linked to a conformational change of both EcE445 (GtE378) and EcR448 (GtR441). We also suggest that EcE445 (GtE378) is the direct source of all protons required to convert O₂ to water. The protons all originate from the cytoplasm and reach the active site by a Grotthus-type mechanism that involves EcE107 (Gt108), EcE99 (Gt101) and EcE445 (GtE378) as well as water molecules that are not resolved in the structure.

Cyt bd is unique in utilizing weak axial ligands as proton acceptors for charge compensation when the heme d (EcE445/GtE378) and heme b₅₉₅ (EcE99/GtE101) act as electron acceptors. These two heme-associated glutamates, in addition of EcE107/GtE108 provide a pathway for the transport of protons, one per electron, to the oxygenated species bound to heme d, including the oxy, oxoferryl and hydroxide-bound species. In this way, EcE445/GtE378 and EcE99/GtE101are critical components for the proton-coupled electron transfer between two hemes and assure that each electron that reaches the active site of the enzyme is delivered by E99 through heme b₅₉₅ and is accompanied by a proton that originates from the bacterial cytoplasm. Now that a structure is available [12], further experiments can be designed to address the many aspects of this model that are postulated.

Highlights.

• Cytochrome bd is a respiratory quinol:O₂ oxidoreductase present in many pathogens

• The enzyme contains 3 hemes in close contact, 2 of which have glutamate axial ligands

• Cytochrome bd is the only known heme protein with glutamic acid as a heme ligand

• Cytochrome bd contributes to the proton motive force and membrane potential

• Glutamate heme ligands are part of the proton-conducting pathway across the membrane