1. Introduction
Cytochrome c Oxidase (CcO) is a membrane protein and part of the respiratory chain (also known as complex IV). Receiving electrons from cytochrome c, it undergoes a redox cycle in which it transforms dioxygen and protons into water. In addition to four protons, thus consumed, CcO pumps four additional protons from the inside of the cell to the outside of the membrane and against a proton gradient [
1]. The resulting increase in electrochemical membrane potential can then be used by ATPases to synthesise ATP.
Uptake of the protons and transport to the bi-nuclear redox centre (BNC) for water production (“chemical” protons), and uptake of those protons pumped can occur via two distinct pathways, the D- and K-channels, named after important residues, D132 and K362 (Rh. sphaeroides numbering), respectively. The D-channel connects the N-side of the membrane from D132 to E286. The K-channel spans from E101 or H96 on the N-side to Y288 at the BNC.
Both “chemical” and pumped protons pass only through the D-channel in the oxidative phase (states A to F in
Figure 1) [
2,
3,
4]. In the reductive phase of the catalytic cycle (states O to R; see
Figure 1), the K-channel transports “chemical” protons, while the pumped protons pass through the D-channel [
5].
For CcO to function as both an oxidase enzyme and a proton pump, the transport of protons through one or the other channel must be highly regulated. We have previously shown that in the
state (between
and F of the oxidative phase, and the first state in which an electron has arrived at the BNC; see
Figure 1), proton passage through the D-channel is auto-regulated, preventing the transfer of an unsolicited second proton before the first one has left the D-channel towards the P-side [
7]. This regulation is mainly achieved by a protonated E286 at the end of the D-channel, leading to directionality of the hydrogen bonds in a water network that favours the closed conformation of the so-called asparagine gate (formed by residues N121 and N139), thereby preventing the formation of a hydrogen-bonded chain of water molecules that allows proton passage to the upper part of the D-channel. The asparagine gate and, with that, proton transport in the D-channel are further affected by the K-channel such that the protonation of K362, located about half-way up the K-channel, favours longer-lived hydrogen bonds between N139 and water molecules and, with that, the formation of a hydrogen-bonded water wire, only with the excess proton located in the lower half of the D-channel.
In a structure of oxidised CcO, there is no connection between the end of the D-channel and the BNC by water molecules [
8], confirming that the transport of “chemical” protons to the BNC in the reductive phase has to go through the K-channel. The proton passage through the K-channel in the O→E redox state, after the arrival of the first electron in the reductive phase, depends on whether the channel contains sufficient water molecules such that a hydrogen-bonded network of protein residues and water molecules can span all the way from the channel entrance to the BNC. The hydration level, in turn, depends on the location of the excess proton. Our previous simulation study of CcO in the O→E redox state showed the transfer of a proton past S365 (at about the lower third of the K-channel) to be determinant for an increase in channel hydration that allows proton transport further up the K-channel [
6].
There is an ongoing controversy in the literature about the order of events, that is, electron transfer and transport of the protons, in the O→E transition [
9]. Based on kinetic absorption spectra and electrometric techniques [
10], it was proposed that the pumped proton passes the D-channel before the reducing electron reaches the BNC. This idea was challenged, since this proton transfer to the proton loading site could take place without a reducing electron being transferred to the BNC [
11]. Therefore, Wikström and co-workers proposed electron transfer to the BNC prior to the arrival of the proton at the proton loading site, enhanced by a positive charge in the K-channel close to the BNC [
11]. This charge has further been proposed to be located on K362 in a conformation pointing towards the BNC (“up” conformation). Time-resolved optical spectroscopy and EPR experiments on a K362M mutant showed electron transfer to the BNC but not fully from heme a3 to CuB [
12]. There is a consensus, though, that proton delivery through the K-channel can only be achieved with K362 present and a proton arriving at the BNC expels the (then pumped) proton from the proton loading site.
In studies on the pH gradient across the membrane, it was shown that the states of protonation of key residues in the two channels are not independent of each other [
13]. In particular, when Y288 has low proton affinity, the protonation states of E286 and K362 are anti-correlated, which can be understood as only one channel being active at a time. Upon the protonation of Y288, this anti-correlation was found to be lost, suggesting that both channels could operate simultaneously. However, a protonated Y288 renders the simultaneous protonation of K362 unlikely [
13] and corresponds to a completed passage of a proton through the K-channel. Since the simultaneous protonation of Y288 and E286 has also been found to be unlikely [
13], the D-channel becomes inactive upon the protonation of Y288. The passage of a proton through the D-channel can, therefore, only take place before the arrival of the “chemical” proton at Y288.
The transport of protons through one or the other channel must thus take place in the correct order. Should the “chemical” proton pass through the K-channel, reach the BNC, and be consumed in the redox reaction to water before a proton has reached the proton loading site, the enzyme would have missed the possibility to pump a proton across the membrane. It is, however, unclear how the two channels are orchestrated to achieve the correct order of proton transfers.
In this work, we used molecular dynamics simulations to simulate CcO in different protonation states of the D- and K-channels and analysed the communication within and between the protein residues of the two channels ultimately affecting the proton transfer probability in one or the other channel.
3. Discussion
According to the literature, the order of transfer events of electron, pumped proton, and “chemical” proton is still unclear. Electron transfer to the BNC may take place prior to the transfer of the proton to be pumped, or both transfers are fully coupled. There is also controversy on whether this electron transfer requires the “chemical” proton to be already at K362 or not [
9,
11,
15].
In our simulations, the electron is already at the BNC, and the (unknown) proton loading site is not yet occupied by a proton. Rather, this proton is still somewhere in the D-channel (models 10*, 01*, and 11*) or has not yet entered the D-channel (models 00*). Therefore, our models cannot fully reflect the scenario after proton transport through the D-channel to the proton loading site and before proton transport through the K-channel. Our models rather assume that electron transfer to the BNC is before proton transfer to the proton loading site and has just occurred.
According to our data, both channels show individual regulation according to their own protonation state. The D-channel has a higher hydration level and is likely more proton conductive when E286 is not protonated. The persistent hydrogen bond between the two residues of the asparagine gate, N121 and N139, in models with protonated E286 suggests that a “closed” gate prevents water molecules from entering the D-channel. However, also in some models with unprotonated E286, there is a probability of finding a hydrogen bond between N121 and N139, corresponding to a closed asparagine gate. Yet, in those models, the number of water molecules in the D-channel is high. This leaves the possibility of water molecules having entered the D-channel by avoiding the asparagine gate or of the asparagine gate having opened (from the closed state in the crystal structure) and closed again during the equilibration phase of the simulation. In the latter case, i.e., with frequent opening and closing, the asparagine gate would not have a regulating role. However, charged E286 seems to exert higher attraction on water molecules than protonated E286, which is supported by the higher number of hydrogen bonds between E286 and water molecules when this residue is charged. Note that a similar effect was observed for D132 and E101, i.e., these residues show more hydrogen bonds with water molecules when unprotonated. This implies that the protonation state of E286 would regulate the likelihood of proton transport through the D-channel by attracting (or not) sufficient water molecules for proton transfer via hydrogen-bond connections. At least the hydrogen-bond connection between E286 and N121 is indeed more probable when E286 is not (yet) protonated.
The K-channel assumes its proton conductivity via the excess proton moving with its own hydration shell (see also [
6]). With the excess proton located below S365, the hydration level of the K-channel is too low to render proton transport likely. This is in agreement with previous proposals of K362 protonation being the rate-determining step in proton transport through the K-channel [
16]. From K362 to Y288, proton transport via hydrogen-bond connections is likely and even more so with K362 in an “up” conformation, as also shown in [
17].
E101 protonation, even together with protonated K362, leads to a low hydration level in the K-channel, but does not harm proton transfer from K362 to Y288. It hinders, however, back transfer to the lower part of the K-channel due to low hydration and consequently low hydrogen-bond connectivity in the lower part of the K-channel. Furthermore, our previous work [
6] has shown that hydrogen-bond connectivity and the associated proton transfer probability in the K-channel favour proton transfer “upwards” as soon as K362 is protonated and in an “up” conformation [
18].
With K362, E101, and E286 protonated (model 0111), however, hydration and hydrogen-bond connectivity in the lower part of the K-channel suggest a non-negligible, albeit low, probability of back transfer. Still, the transfer of a proton from K362 to the protonated entry of the K-channel seems unlikely.
The changes in the protonation state of the D-channel can in principle be “sensed” by the K-channel through electrostatic interactions with either the ion or the titratable residues K362 and E101, and by Y288. Likewise, the D-channel is “informed” about the protonation state of the K-channel via electrostatic interactions. Since the D-channel appears not to respond to those differences in K-channel protonation, at least not in the quantities observed in this work, any signalling between the two channels would then have to be from the D-channel to the K-channel.
The protonation of the D-channel, however, has hardly any influence on the conformation of K-channel residues or on the hydration level and hydrogen-bond connections in the K-channel. Two noteworthy exceptions are models 0111 and 1100b, which show hydrogen-bond connections between the upper part and lower part of the K-channel that are not found in the other models with the same K-channel protonation. In both cases, there are (more) water molecules in the K-channel, which make these hydrogen-bond connections possible (see
Figure 15 and
Figure S43). It is unclear whether these observations can be related to CcO function, although it is tempting to consider model 1100b as a state in which the D-channel has transported its protons, albeit not fully, to the proton loading site. In this state, the K-channel can now become active and the likely rate-determining transition to a protonation state with the proton past S365 and further up to K362 (such as models 1100c and 1110) can take place. If, however, a completed proton passage through the D-channel is required to “activate” the K-channel, model 0100b should show an effect similar to that of model 1100b, but this is not the case. Moreover, the probability of finding a hydrogen-bond connection to K362 is also rather weak in model 1100b (∼0.1). There is also no obvious explanation why only in model 1100b, water molecules are more probable at the height of S365 and slightly above, thereby making hydrogen-bond connections and eventually proton passage through the K-channel possible.
The apparent independence of the two proton-conducting channels of CcO, as manifested in the analysed properties (conformational preference for key protein residues, hydration level, and hydrogen-bond connections, with the latter being perhaps most important for proton transport), does not rule out that the activity of the two channels is regulated by proton transfer events. Our data suggest, however, that such regulation is not achieved via direct communication between the two channels (“sensing” their respective protonation states). Another possible way is indirect communication not, or not solely, via the location of the excess proton but rather the injected electron and the associated regulation of proton transfer events.
After the injection of the first electron in the reductive phase and its transfer from CuA to heme a, a proton is transported through the D-channel to the proton loading site (somewhere above the BNC). This is coupled to (partial) electron transfer from heme a to the BNC. This electron transfer is denoted here as partial, since spectroscopic data suggest only 60% of the electron to be transferred from heme a to the BNC [
10,
12].
The negative charge at the BNC then triggers proton transfer through the K-channel. The communication of the two channels is thus not performed directly through their protonation states. Rather, the first proton transfer through the D-channel is likely independent of the K-channel’s protonation state. However, the accomplishment of proton delivery through the D-channel (to the proton loading site) and the electron transfer coupled therewith are signalled to the K-channel.
A mechanism in which the pumped proton passes the D-channel before the electron reaches the BNC has indeed been previously proposed [
10]. This proposal is also in agreement with the effect of K362M mutation on the rate of reduction of heme a3, as measured using time-resolved optical spectroscopy and EPR [
12]. In that work, even with an inactive K-channel, one electron could reduce the BNC but was found to be shared between heme a3 and CuB in the
Rh. sphaeroides oxidase. Complete proton pumping, however, requires expelling the proton from the proton loading site, which is only achieved with a proton arriving at the BNC, delivered through the K-channel.
The notion of K362 being already protonated, required for and facilitating electron transfer to the BNC, as mentioned in [
11], is not supported by our data. All simulations with protonated K362 show a high hydration level of the K-channel that is in contradiction with the few water molecules observed in the crystal structure [
8]. Even in the models in which E101 is also protonated (models *11), which exhibit a significantly lower hydration level, there are enough water molecules present in the upper half of the channel leading to hydrogen-bond connectivity that allows proton transfer from K362 to Y288. All that could prevent this proton transfer from happening before proton transport through the D-channel is the fast electron transfer to CuB and a proton transfer to the proton loading site that is faster than the proton transfer from K362 to the BNC. Models in which the proton transfer through the D-channel and further to the proton loading site via water molecules and hydrogen-bond connections above E286 is much faster have indeed been proposed [
19,
20,
21]. However, simulations of proton transfer from E286 to the putative proton loading site, albeit in the oxidative phase, show barriers of about 10 kcal/mol [
22,
23] and are thus comparable to that of proton transfer from K362 to Y288, which has been calculated to be about 11 kcal/mol [
17].
K-channel protonation, at least in the upper part of the channel (at K362 or at position e or f), can be anticipated to be less likely without the electron at CuB. Though not fully comparable, simulations of this proton transfer found indeed a higher barrier in the oxidised Pr state than in the reduced O→E state [
17]. One can envisage, however, that protons can enter the K-channel up to position b (just below S365) without further attraction by an electron at the BNC.
The (partial) electron transfer to the BNC coupled to or followed by proton transfer to the proton loading site can thus be regarded as the “go” signal for the K-channel. Only with the electron there, the proton affinity of the BNC is high enough to enable the proton and its accompanying water molecules to travel beyond S365. Once this residue has been passed, further hydration and proton passage to K362 and all the way up to Y288 become feasible.
Since there is no pronounced effect of D-channel protonation on the K-channel, the K-channel is “insensitive” to the proton to be pumped or a proton to re-load E286 while making its way through the D-channel. Any “sensitivity” that regulates K-channel activity must, therefore, be towards the redox state of the BNC (and perhaps to a proton at the proton loading site, since that would counteract the negative charge at the BNC to some extent).