Evidence for Fast Electron Transfer between the High-Spin Haems in Cytochrome bd-I from Escherichia coli

Cytochrome bd-I is one of the three proton motive force-generating quinol oxidases in the O2-dependent respiratory chain of Escherichia coli. It contains one low-spin haem (b558) and the two high-spin haems (b595 and d) as the redox-active cofactors. In order to examine the flash-induced intraprotein reverse electron transfer (the so-called ''electron backflow''), CO was photolyzed from the ferrous haem d in one-electron reduced (b5583+b5953+d2+-CO) cytochrome bd-I, and the fully reduced (b5582+b5952+d2+-CO) oxidase as a control. In contrast to the fully reduced cytochrome bd-I, the transient spectrum of one-electron reduced oxidase at a delay time of 1.5 μs is clearly different from that at a delay time of 200 ns. The difference between the two spectra can be modeled as the electron transfer from haem d to haem b595 in 3–4% of the cytochrome bd-I population. Thus, the interhaem electron backflow reaction induced by photodissociation of CO from haem d in one-electron reduced cytochrome bd-I comprises two kinetically different phases: the previously unnoticed fast electron transfer from haem d to haem b595 within 0.2–1.5 μs and the slower well-defined electron equilibration with τ ~16 μs. The major new finding of this work is the lack of electron transfer at 200 ns.


Introduction
Cytochrome bd-I is one of the three terminal oxidases in the aerobic electron transport chain of Escherichia coli [1][2][3][4]. The enzyme catalyzes the reduction of molecular oxygen to water with quinol [5]. The energy released in this redox reaction is stored in the form of a proton electrochemical gradient [6][7][8][9][10][11][12]. Apart from energy conservation, a bd-type oxidase serves other vitally important physiological functions [13,14] including its contribution to bacterial resistance to nitric oxide, hydrogen peroxide, peroxynitrite [15][16][17][18][19], nitrite [20,21], and sulfide [22]. Cytochrome bd-I contains three subunits, CydA (57 kDa), CydB (43 kDa) and CydX (4 kDa). A newly discovered small polypeptide CydX is thought to be required for maintenance of enzymatic activity and stabilization of the haems [23][24][25][26][27]. Cytochrome bd-I carries one lowspin haem (b 558 ) and the two high-spin haems (b 595 and d). In the course of a catalytic reaction, an electron from quinol transfers to haem b 558 and then to the oxygen reductase active site. The organization of the oxygen reductase site remains unclear. It certainly comprises haem d at which O 2 is bound, activated and reduced to 2H 2 O. Whether the site also contains the second haem, b 595 , is not known with certainty. A large body of spectroscopic data suggests that haem b 595 indeed can form with haem d a di-haem active site [8,10,[28][29][30][31][32][33][34][35][36][37][38][39] and that cytochromes b 595 and b 558 are each oxidised as the 'oxy' form decays (650 nm) during the low-temperature reaction with oxygen [40]. However, according to other reports, haem b 595 has an alternative or additional function [2,[41][42][43]. As to a possible additional role for haem b 595 , it's also worth mentioning that the observed catalase activity of cytochrome bd-I from E. coli could be associated with this haem [44]. Thus the exact functions of the high-spin haem b 595 remain to be clarified. To better understand the role of haem b 595 in the intraprotein electron transfer and the oxygen reduction reaction, we have applied time-resolved transient absorption spectroscopy and spectral modeling to the interhaem electron backflow reaction in cytochrome bd-I.
Flash photolysis of CO from the "mixed-valence" form of a terminal oxidase is a method by which haem-to-haem electron transfer can be measured. In cytochrome oxidase, initially the haem a 3 /Cu B binuclear site is trapped in its reduced ferrous/cuprous state and stabilized by CO binding to the haem iron whereas the two other redox sites, haem a and Cu A , are oxidized. Flash photolysis of CO results in ultrafast displacement of CO from haem a 3 (within a fraction of a picosecond) to bind Cu B which in turn gives rise to the three phases of the reverse electron transport (the so-called ''electron backflow'') at neutral pH. The electron re-equilibration should occur first by the electron transfer to haem a (in two steps with τ~1.4 ns and~3 μs) and then from both haems to Cu A (~35 μs). The 3-μs phase was reported to be rate-limited by CO dissociation from Cu B [45], whereas the 1.4-ns and 35-μs phases are rate-limited by the electron tunneling between haem a 3 and haem a and between haem a and Cu A , respectively [46][47][48]. The nanosecond interhaem electron transfer was detected in both cytochrome c and quinol oxidases of the haem-copper superfamily [46,49,50]. Under similar conditions, the 16μs component of the photolysis-induced electron backflow was resolved in cytochrome bd-I [39]. In this work, we provide, to the best of our knowledge for the first time, evidence for the existence of a faster electron backflow component, on a submicrosecond time range, induced by photodissociation of CO from ferrous haem d in one-electron-reduced state of cytochrome bd-I.

Biological materials
Membrane vesicles were prepared by passing the E. coli cells (strain GO105/pTK1) through a French press according to [51]. Cytochrome bd-I was isolated and purified as reported in [51,52].

Sample preparation
One-electron-reduced CO-bound cytochrome bd-I (b 558 3+ b 595 3+ d 2+ -CO) was prepared as described in [33,37,39]. To do this, the as-prepared cytochrome bd-I that is mostly in the oneelectron-reduced O 2 -bound state i.e. b 558

Spectroscopy
For CO photolysis, nanosecond pulses were used with time duration 5-15 ns and excitation wavelengths at 532 nm (from a Nd YAG laser) and 640 nm (from a Nd YAG pumped dye laser), the latter wavelength being near the haem d α-band maximum. The photoinduced absorption changes were measured with a single beam home-built spectrophotometer with submicrosecond time resolution (for details see [39,[53][54][55]) and with a nanosecond spectrophotometer, as described in [37,56].

Data analysis
Origin (OriginLab Corporation) and MATLAB (The Mathworks, South Natick, MA) were used for data manipulation and presentation.

Results and Discussion
Laser flash-photolysis of one-electron-reduced CO-poised cytochrome bd-I causes immediate dissociation of CO from ferrous haem d. CO photolysis is followed by its recombination with cytochrome bd-I and the latter process can be fitted by three exponentials with apparent firstorder rate constants (k) of 6.5×10 4 s -1 , 5.5×10 3 s -1 and 3.3 ×10 1 s -1 . The kinetic trace at 432 nm at which all the transitions can be seen is depicted in Fig 1. Previous work [39] allowed one to compose and assign the spectra of these kinetic steps. The rapid phase (k = 6.5×10 4 s -1 at 1 mM CO) is assigned to bimolecular recombination of CO to haem d plus backflow of the electron from haem d to haem(s) b. The intermediate phase (k = 5.5×10 3 s -1 ) is due to return of the electron from haems b to haem d and bimolecular recombination of CO in that enzyme fraction. The slow phase (k = 3.3 ×10 1 s -1 ) is complex but dissociation of an unidentified ligand (L) from haem d is possibly a major contributing factor [39]. This spectrum reflects the almost pure electron transfer reaction. As suggested by a previous report [39], in this redox state of the enzyme, dissociation of CO from the distal side of haem d may be accompanied by simultaneous binding of a ligand of unknown structure (L) to the proximal side of haem d. Thus, the absorption changes caused by these two processes mostly cancel each other out. L is then dissociated from haem d in the slow phase to return the haem ligation state to that before photolysis.
The identity of L is to be determined. To date, we favour a hypothesis that L is an internal protein ligand. Under this assumption, the easiest option could be that L is the proximal ligand to haem d. The haem d axial ligand is not yet identified. We assume that it should be a weak ligand like a glutamate residue [57] that could be easily detached from the haem iron rather than a strongly coordinated histidine residue [58]. This proposal is consistent with an earlier suggestion that the native ligand of haem d is replaced with cyanide to produce the five-coordinate high-spin cyano adduct [59,60]. The fact that absorption changes that possibly reflect exchange or binding of an internal ligand to haem d in the course of reduction of the bd enzyme are similar to the changes caused by CO [61] is also in line with this proposal. The 1.5-μs transient spectrum evidently is not identical to that at a delay time of 200 ns after CO photolysis from the same initial, one-electron-reduced state of cytochrome bd-I (Fig 2). The 200-ns spectrum displays a double-humped band with the two local maxima, at 415 and 432 nm, but has no clear minimum. Obviously, some event characterized by the change in absorption of the haems has occurred between 200 ns and 1.5 μs. . The difference can be reasonably fitted by the spectrum of electron transfer from haem d to haem b 595 (Fig 3, line 2). Thus, photodissociation of CO from haem d is followed by fast (between 0.2 and 1.5 μs) electron backflow from haem d to haem b 595 . The use of extinction coefficients for the difference (reduced-minus-oxidized) spectra of haem b 595 and haem d [62] makes it possible to gain an estimate of the extent of the fast electron transfer: within 0.2-1.5 μs 3-4% of haem d is oxidized and the same amount of haem b 595 is reduced. Earlier, the electron backflow from haem d to haems b could only be observed on the slower, microsecond time scale with the one-electron-reduced bd-type oxidase from Azotobacter vinelandii [63] and E. coli [7] at low CO concentrations.
Transient absorption spectra at delay times of 200 ns and 1.5 μs following CO photodissociation from the fully reduced cytochrome bd-I are shown by Fig 4 as a control. The spectra are very similar. They have a typical "W" (reversed) line shape with two maxima at ca. 425 and 445 nm and a minimum at about 435 nm. Such a peculiar "W" line shape is typical of static difference absorption spectra of CO binding to haem d in the fully reduced bd-type oxidase [32,34,64,65]. Therefore, these photoinduced spectral changes (Fig 4) can be associated with the removal of CO from haem d triggered by flash photolysis of the haem d Fe-CO bond. This control shows that, in contrast to the one-electron-reduced cytochrome bd-I, photodissociation of CO from haem d in the fully reduced oxidase does not lead to any intraprotein electron transfer between the metal-containing redox cofactors. This is expected because, in the dithionitereduced cytochrome bd-I, all the haem groups are in the ferrous oxidation state (Fe 2+ ).
Thus, the new observation is the lack of the photoinduced electron transfer between haem d and haem b 595 at 200 ns in one-electron-reduced cytochrome bd-I. Internal electron-transfer reactions seem to occur on a slower time scale.
It should be noted that the amplitude of the reverse electron transfer depends on the difference in apparent midpoint redox potentials (E m ) of haems d and b 595 . The E m values of haems d and b 595 obtained by spectroelectrochemical equilibrium redox titration were reported to be +265 and 165 mV respectively [62]. Accordingly, their difference (ΔE m ) is about 100 mV. According to the Nernst equation, such a large difference should correspond to a very small extent (~2%) of the electron backflow expected. The amplitude of the electron backflow observed in this work is in agreement with the estimated one. Taking the measured value (~3.5%) into account, the difference in the transient E m values of the haems is calculated to be 84 mV.
In haem-copper oxidases, the total amplitude of the reverse electron transfer (including the 1.4-ns and 3-μs phases, and the 35-μs phase of the electron transfer from a low-spin haem and a binuclear center to Cu A ) varies from~4% (cytochrome ba 3 from Thermus thermophilus [66]) to~25-35%,~50% and~75%-in cytochromes aa 3 from bovine heart [67], Paracoccus denitrificans [68] and Rhodobacter sphaeroides [47,69], respectively. The amplitude of the electron backflow of the 1.4-ns phase itself is~6% in cytochrome bo' from E. coli [49] and 11-16% in bovine heart cytochrome aa 3 [46], and as in case of cytochrome bd-I, is modulated by the difference in redox potentials of electron carriers. Thus, the observed extent of the electron backflow in cytochrome bd-I is not due to a damaged fraction of the preparation but is defined by ΔE m of the two haems.
The rate constant for electron transfer from haem d to haem b 595 should be the sum of the microscopic constants of forward (k for ) and backword (k back ) reactions. Assuming that the phase of electron transfer corresponds to the full equilibration between haem d and haem b 595 , 3.5% amplitude of the reaction yields the equilibrium constant, K of~28. Taking into account that the rate constant is the sum of the microscopic rate constants of forward (b 595 ! d) and backward (d ! b 595 ) electron transfer (k for + k back ) and that the equilibrium constant is K = k for /k back , the values for k for and k back can be estimated to be 0.48×10 6 -4.8×10 6 s -1 and 0.17×10 7 -1.7×10 7 s -1 , respectively. Thus, the apparent value of the rate constant should be close to the k for .
For a long time, in the haem-copper terminal oxidases, the 3-μs phase of the electron backflow was considered as the most rapid interhaem electron transfer process [48,[70][71][72]. Subsequently, it was proved that the 3-μs event is preceded by a much faster 1.4-ns phase of the reverse electron transfer from haem a 3 to haem a in bovine heart cytochrome c oxidase [46] or 1.2-ns phase of that from haem o' to haem b in the related quinol oxidase bo' of E. coli [49].
A phase of the rapid (0.2-1.5 μs) electron transfer between the haems in cytochrome bd-I reported here can be compared with these two consecutive kinetic components of the electron backflow reaction in the haem-copper oxidases. As shown earlier, the value of the rate constant of the 1.4-ns phase of the electron backflow in haem-copper oxidases is in fundamental agreement with the calculations for the activationless electron tunneling between the haems, based on the semiempirical Moser-Dutton rule [73,74] and the known distance between the haem cofactors, the specific low value of the reorganization energy and the correctly calculated the average protein packing density [46,49,75,76].
As noted before [77], when the donor-acceptor distance becomes short, as is the case for the haem-haem electron transfer in the haem-copper oxidases, the local heterogeneity of the coupling medium and the significantly reduced number of effective electron transfer pathways should be taken into account [49,77,78]. A more detailed model based on the quantum mechanical calculation of the coupling medium between donor and acceptor (the so-called pathway model) combined with the molecular dynamics enabled researchers to identify and describe theoretically the main pathways of the coupling between haem a and haem a 3 of the bovine enzyme [79,80].
With the present data on cytochrome bd-I, only very rough correlation between the rates of the fast interhaem electron transfer in the bd oxidase and the aa 3 oxidase and the distances between the corresponding haems can be made. Apart from the unknown values of the reorganization energies, packing density and the coupling medium in cytochrome bd-I, there is an ambiguity in defining distance metrics between the haems (edge-to edge or Fe-to-Fe) caused by the fact that atoms on the periphery of the large porphyrin rings are not always well coupled to the central metal [76,81,82].
As pointed out earlier [81], because of the different efficiency of the coupling of the intervening polypeptide medium between the redox centers, the electron transfer rates at the same distance can differ by as much as a factor of 10 3 , whereas the donor/acceptor distances that differ by as much as 5 Å can yield identical rates. Since the fast reverse electron transfer reaction in cytochrome bd-I is about 2 orders slower than that in the haem-copper oxidases, the rate constant of the electron transfer between haem d and haem b 595 is in agreement with the predicted Fe-Fe centre distance between the haems d and b 595 (~10 Å) [36] that is close to the Fe-Fe centre distance between the haems a and a 3 in haem-copper oxidases (~13 Å) [83].
An alternative possibility could be that the fast electron transfer between haem d and haem b 595 in cytochrome bd-I is not limited by the pure electron tunneling between the haems, as in the case of the 3-μs phase of electron backflow in the haem-copper oxidases. In contrast to the 1.4-ns phase, the 3-μs electron transfer is the phase with a significant activation barrier [45]. Its rate is limited by the CO migration from Cu B out of the protein that is accompanied by a slight modification of the a/a 3 redox equilibrium [46]. It would be expected that since cytochrome bd-I is lacking Cu B , the dissociation of CO should not limit this component of the electron transfer. However, the absence of the copper ion does not exclude a possibility of a transient trapping of CO on its way from haem d to the outside which, if coupled to the structural rearrangement, may theoretically control change in redox potentials of the haems d and b 595 and the electron transfer between them.
According to the modeling of excitonic interactions between ferrous haems d and b 595 in the absorption and circular dichroism spectra [36] and the data from this work, the Fe-Fe distance between haems d and b 595 should be much larger than that between haem a 3 and Cu B in cytochrome c oxidase (4.5 Å [84] or 5.2 Å [85]). If this is the case, we cannot consider haem d and haem b 595 as forming a binuclear center in cytochrome bd for O 2 reduction in which haem b 595 is a functional analogue of Cu B . However, haem b 595 could play a role to provide haem d with an electron [40] and possibly a proton in the course of the O 2 reduction reaction. Fig 5 shows the proposed scheme of electron backflow reaction in cytochrome bd-I that is based on the present and previous reports [37,39]. In brief, a laser flash induces dissociation of CO from the one-electron-reduced state of the enzyme that is followed by electron redistribution between the haem groups in 3-4% of the oxidase population. The electron from haem d first moves to haem b 595 and then reaches haem b 558 . Subsequently, CO recombines back to haem d that is coupled to the return of the electron from haems b 558 and b 595 to haem d.
Our present results thus indicate that the electron transfer between the two catalytically relevant haem groups (haem d and haem b 595 ) of cytochrome bd-I cannot itself limit or control Reverse electron transfer within one-electron-reduced cytochrome bd-I following photodissociation of CO. This scheme is based on the current work and on the works of Rappaport et al. [37] and Siletsky et al. [39]. Circles denote haems. Open circle shows that haem is oxidized. Black filled circle shows that haem is reduced. Gray filled circle shows that a small fraction of haem (4%) is transiently reduced. Laser flash causes time-unresolved dissociation of CO from haem d (1!2 transition). At CO concentration of about 1 mM, in most of the hemoprotein molecules in which ultrafast geminate recombination of CO with chlorin has not occurred [33,35,37] CO returns to haem d with the apparent first-order rate constant (k) of 6.5×10 4 s -1 at 1 mM CO (2!5 transition, right branch). However, in 3-4% of the enzyme population photodissociation of CO is followed by electron transfer from haem d to haem b 595 with k in the range 9.7×10 5 to 7.2×10 6 s -1 (2!3 transition). Then the electron transfers to haem b 558 with k of 6.5×10 4 s -1 (3!4 transition). After that, the electron is probably redistributed between the haems in accordance with their midpoint redox potentials [62]. Eventually, in this small fraction of the hemoprotein the electron returns from haem b 558 and haem b 595 to haem d simultaneously with recombination of CO with chlorin with the apparent first-order rate constant (k) of 5.5×10 3 s -1 (4!5 transition). catalysis of the oxygen reduction, and that the mechanism of nanosecond interhaem electron transfer may be universal not only in the family of haem-copper oxidases but also in the bdtype terminal oxidases. The placement of haem b 595 close enough to the haem d catalytic site is to assure that electron transfer into the site will be rapid enough to provide the fast reduction of molecular oxygen without the release of potentially harmful partly-reduced oxygen intermediates. Besides, the rapid electron transfer may also help to control the potential reactivity of the haem d porphyrin radical that arises in the catalytic cycle upon formation of the 'P' state [9, 86] by quickly reducing it in the energy-coupled electron transfer.

Conclusions
The reverse electron transfer from haem d to the haems b induced by flash-photolysis of oneelectron-reduced CO-bound cytochrome bd-I of E. coli consists of two kinetically different phases. The well-defined electron equilibration between the haems with τ~16 μs appeared to be preceded by the previously unnoticed fast electron tunneling from haem d to haem b 595 in the time interval between 0.2 and 1.5 μs. The major novel finding of this study is that there is no electron transfer at 200 ns. These data are consistent with the suggestion that haem b 595 is required for rapid electron donation during catalysis and provide insights into mechanisms of enzymatic O 2 reduction coupled to energy conservation.