1. Introduction
Cytochrome c oxidase (CcO), also known as complex IV of the respiratory chain, is a proton pumping membrane protein. It is located in the mitochondrial inner membrane where it uptakes molecular oxygen, electrons and protons. In a redox cycle (see
Figure 1), four protons are pumped across the mitochondrial inner membrane. The necessary energy for the pumping mechanism is gained by water formation from oxygen and four “chemical” protons [
1,
2]. The proton uptake from the N-side to the bi-nuclear centre, the location where redox chemistry takes place, is mediated via two water-filled channels, named D- and K-channels according to essential residues within the channels, D132 and K362 (
R. sphaeroides numbering). Furthermore, research has shown proton uptake through the K-channel becomes relevant only during the reductive half of the catalytic cycle (states O to R, see
Figure 1). Especially in the transition of redox states O→E, the K-channel delivers a chemical proton to the reaction [
3,
4]. In contrast, both “chemical” and pumped protons pass through the D-channel in the oxidative phase (states A to F in
Figure 1) [
3].
The crystal structure of
R. sphaeroides CcO clearly shows a wire of water molecules inside the D-Channel [
5], and simulations support water chains allowing proton transfer. Moreover, a self-regulating mechanism controlling the hydration level was observed [
6,
7,
8]. The amount of water present in the D-channel and thus its ability to form water wires is regulated by an interplay of the “asparagine gate”, formed by residues N121 and N139, and the terminal residues D132 and E286 of the channel. The passage of a proton through the “asparagine gate” via a chain of hydrogen-bonded water molecules has been confirmed by simulation studies [
7,
8] as the rate-determining step of proton transport through the D-channel. Particularly, the protonation state of E286 shows a major impact on the conformation of the “asparagine gate” and thus the channel regulation [
6].
In contrast, the number of crystal water molecules found in the K-channel of CcO is too low to enable a purely water-mediated proton transfer. It has therefore been argued that the name-giving lysine K362 may act like a lever elevating the proton by flipping from a downwards to an upwards position [
9] (where “downwards” means towards the channel entrance at the N-side of the membrane, and “upwards” refers to towards the bi-nuclear centre and thus towards the P-side of the membrane). Furthermore, mutation studies emphasise the importance of K362 [
10,
11], and various simulation studies feature a significant probability of K362 to carry a proton [
9,
12]. Other important titratable residues in the K-channel of CcO, H96 and E101, are located at the channel entrance. The solvent-exposed H96 can occur in neutral and protonated form and would change between these two forms when participating in proton uptake into the channel. E101, which has been discussed as the initial proton acceptor, can also exist in protonated and unprotonated form according to pKa calculations [
12].
In case of protonation of K362, proton transfer to Y288, the end of the K-channel, becomes feasible [
13,
14]. In the presence of an excess charge, the tertiary structure around the K-channel has been observed to undergo conformational changes, including a widening of the channel [
9], favouring or disfavouring a change in the hydration within the channel [
13,
15].
An increased hydration level would allow the excess proton to travel as hydronium ion within a sphere of water molecules, implying that the water cloud rearranges upon the proton transfer between two adjacent water molecules, such that there is always a “leading” water molecule, i.e., a water molecule ahead of the hydronium ion [
16]. Moreover, a Grotthuss mechanism along a water wire up to K362 would be conceivable. However, whether such wetting effects indeed occur and thus enable proton transport in the K-channel of CcO still remains unclear.
Hydration levels in the cavities of the BNC and in the proton-conducting channels in CcO have been studied by prediction of internal water sites and molecular dynamics simulations [
17,
18,
19]. The formation of water wires or water chains, along which a proton transport can also take place in the channels [
6,
14,
20] or from the BNC to the P-side and thus the exterior of the protein has been probed by simulations [
21]. Higher hydration levels indeed suggest higher connectivity along water chains. In all these studies [
18,
20], one fixed protonation state of the channel residues has been simulated. and while an effect of the hydration on connectivity by water chains could be observed, the effect of different locations of the excess proton in the K-channel has, to the best of our knowledge, not been studied, yet.
In this paper, we address this question with molecular dynamics simulations of CcO in the O→E redox state, that is, the first transition of a proton through the K-channel in the reductive phase of the catalytic cycle. By varying the position of the (excess) proton in the channel and analysing the hydration pattern, the hydrogen-bonded networks and the conformational dynamics of the protein residues, we find a delicate interplay between hydration and protonation dynamics.
Figure 1.
Redox cycle in cytochrome c oxidase. Figure adapted from [
22].
Figure 1.
Redox cycle in cytochrome c oxidase. Figure adapted from [
22].
4. Discussion
Our molecular dynamics simulations of CcO with an excess proton located at different positions in the K-channel, either on protein residues or as hydronium ions, afforded us the chance to explore the dependence of the hydration level of the K-channel on the proton position.
The probabilities of hydrogen-bonded connections and the lifetimes of hydrogen bonds, along with the probability for a proton transfer to take place along hydrogen-bonded connections, highly correlate with the number of water molecules in the channel. According to our simulation data, proton transfer across the channel is likely when the proton is located above S365 (model 0100c), at about the height of K362 (model 0100d) or when K362 is protonated but E101 is not (models 0110 and 0110H). These are also the models with the largest volume and highest number of water molecules inside the channel.
Models in which the proton-carrying water molecules, i.e., the hydronium ion, are in the lowest part of the channel (models 0100a and 0100b) exhibit a rather low hydration level, with particularly few water molecules in the upper part (above S365) of the channel. Located in the lower part of the channel, the hydronium ion drags its hydration cloud only as far as it has reached itself. As our projection of the water density shows, there is little probability density for a water molecule ahead of the hydronium ion (except for the positions of crystal water molecules). This further suggests that the presence of a positive charge alone is not sufficient to fill the channel with water, and the attraction of further water molecules dragged by the proton is limited to the immediate hydration cloud.
Comparison of models 0100b, 0100c and 0100d in which a hydronium ion works its way up the channel (from below S365, to above S365, and to the height of K362) and drags a water cloud with it, suggests the hydration level becomes high before the proton is passed on to K362 and remains high with the proton at K362 (model 0110). Models with a hydronium ion above K362 (0100e and 0100f) show a well-connected upper part of the K-channel but lower hydration level than models with protonated K362 or those with the hydronium ion further below, 0100c and 0100d. The lower hydration level in models 0100e and 0100f can be explained as the excess charge being located too far away from the channel entrance to attract water molecules from the bulk. Moreover, in the course of the simulation of these models, the channel was never wide at the height of K362. It is unlikely that a passage of the excess proton from K362 to the upper part of the channel leads to a loss of water molecules and the channel to “dry”. Rather, water molecules that are already inside the upper part of the K-channel remain, and the actual number is higher than that observed in the simulated models 0100e and 0100f. It is quite possible that the MD simulation time of 200 ns is too short to allow full sampling of major structural changes such as (unfavourable) channel widening and conformational changes of protein residues. Most residues show a clear preference for one or a few dihedral conformations in a particular protonation model. Likewise, different channel widths and hydration levels can be observed for different protonation models, suggesting that the individual models are sufficiently equilibrated to show the interplay of protonation state, local protein conformation and, importantly, proton transport and hydration level in the channel.
The hydration level is highest in the 0110 model, in which K362 points “up”, indicating that this conformation can lead to a wider channel, allowing more water molecules inside. Channel widening as a pre-requisite for a flip of K362 to an “up” position is unlikely since model 0111 shows an “up”-like K362 conformation and a narrow channel. Models 0110H and 0110 with a protonated K362 in “up” conformation, in contrast, exhibit a large channel volume and a higher hydration level. Hence, it appears more reasonable that the higher number of water molecules in 0110 and 0110H accounts for the wider channel, at least to some extent and not the conformation of K362.
The importance of a channel widening to increase the hydration level has been shown by experiments with a fluorescent marker at the channel entrance [
15]. When Helix 6 (containing H96) and Helix 8 (containing E101) widen, they give space for water molecules to enter the channel. While this explains how a higher hydration level at the channel entrance can be reached, the cause and consequence of the wider channel, higher channel volume and higher number of water molecules in the K-channel cannot be explained this way. In the lowest part (below S365) of the channel, the channel width is sufficient to accommodate larger numbers of water molecules, or in other words, an increased hydration level in the lower part of the channel does not lead to a larger polyhedron volume and/or widening of the channel. For the part of the channel above S365, the channel widens, and the volume increases with the hydronium ion and its water cloud moving up (as can be envisaged as transitions from models 0100a, 0100b, 0100c to 0100d). This suggests the moving water cloud to be the reason for the channel widening.
When a proton is located on E101, the side chain of E101 has a tendency to point “upward”. Since this is the neutral, or not-negative form, one can argue that it is not repelled by the negatively charged Y288 and can therefore point “up” towards Y288. Since Y288 is rather distant, though (>20Å), it is more likely that the hydrophobic (“dry”) interior of the K-channel is more favourable for E101’s side chain than the solvent-exposed channel entrance. This “up” conformation of E101 narrows the lower part of the channel, as manifested by short distances to P315, preventing (further) water molecules from entering. Correspondingly, even with protonated K362, the channel hydration is rather low when E101 is also protonated (model 0111).
According to mutation experiments [
35,
36], E101 is essential for the proton transport through the K-channel and has therefore been discussed to be the residue responsible for initial proton uptake. According to our simulations, as long as E101 is protonated, the channel has only a few water molecules inside and is unlikely to fill further. For the proton transport to proceed, the proton must thus be transferred to a nearby water molecule, of which there are so few that such a transport does not go farther than E101, just to a state corresponding to 0100a.
Alternatively, H96 can act as the residue that initially accepts a proton from the bulk. Then, E101 must play another critical role. One such role could be the amplification of H96’s proton affinity, E101 and H96 together forming a “proton acceptor diad”. This idea is supported by the 0100H model showing the shortest distances between the two residues (note that the distances listed in
Table 3 are not donor–acceptor distances but measured between the C
1 and the C
atom of H96 and E101, respectively).
The protonation model 0100H reflects the scenario of H96 initially accepting a proton from the bulk. This model allows enough water molecules in the lower part of the channel to directly transfer a proton via hydrogen-bonded chains up to S365, or at least to a state corresponding to model 0100b. Though the water density beyond S365 is low in model 0100b, there is a small probability of forming hydrogen-bonded connections between the hydronium ion and K362, as well as between S365 and K362, indicating a (small) chance for water molecules to pass S365. It is therefore conceivable that a state, corresponding to model 0100c can be formed, albeit with low probability. As soon as the excess proton is beyond S365, the hydration level and the hydrogen-bonded connectivity increase, rendering proton transfer through the channel feasible. Passage of S365, first by a water molecule and then by the proton itself, is therefore likely the rate-limiting step of proton transport through the K-channel.
Though even a hydrogen-bonded connection from the hydronium ion to T359, and therefrom to Y288, is observed in models 0100c and 0100d, when the hydronium is past S365, the water distribution still renders a pathway via K362 more probable. The “down” conformation of K362 observed in model 0100c allows proton uptake from the lower part of the channel by K362. In model 0100d, the orientation of K362 represents an intermediate state between a “down” and an “up” conformation that allows K362 to accept a proton from the hydronium ion and possibly subsequent proton donation to water molecules above K362. Reaching such a conformation is likely supported by the increased channel width in this model. Both factors are related, since a lack of water molecules between S365 and K362 does not allow the formation of a hydrogen-bond connection between the two residues and thus the transport of a proton from the lower to the upper part of the channel.
Once the proton has arrived at K362, it points “up” to the negatively charged Y288. In such a conformation, the high hydrogen-bonded connectivity renders proton transfer via a Grotthuss mechanism along hydrogen-bonded chains likely, in agreement with simulation studies of explicit proton transfer [
13].
Getting the proton to K362 has been considered as an important step/pre-requisite for proton transport through the K-channel [
9,
13,
37]. While a number of previous simulation studies suggest a water-mediated hydrogen-bonded network between K362 and Y288 [
17,
18,
20,
38], partially via T359, regardless of K362’s protonation state, with neutral K362 no connectivity with the lower part of the K-channel has been observed [
18,
38] but with a protonated K362 there is enough water in the lower part to render water chains between K362 and E101 conceivable [
20]. Our simulation data agree in as much as protonated K362 leads to a high connectivity of hydrogen bonds in the upper and lower part of the channel. Moreover, a hydronium ion located above S365 or at the height of K362, can lead to water-mediated hydrogen-bonded networks, “simply” due to the increase of water molecules present.
This increased water level could only be observed because of the (artificial) placement of hydronium ions and a molecular dynamics simulation that allows the protein structure to change such that a channel widening and with that a further filling of the channel can take place. Water insertion methods that work with static structures but also grand canonical Monte Carlo sampling may not be able to fully capture this “filling” of the channel (an increased number of water molecules also in the upper part of the channel) if it requires some channel widening (due to extra water molecules) to allow further water molecules to enter, attracted by an excess proton.
Our data show that the location of the excess proton is decisive for the hydration level. In particular, the region between S365 and K362 remains “dry” unless a hydronium ion is located there. For a hydronium ion to move beyond S365, or the proton to hop over to a water molecule beyond S365, the hydronium ion must have already moved with its hydration cloud up to just below S365, and one water molecule must be above S365. This passage of a water molecule is thus crucial for proton transport up to K362. Once this has been achieved, the channel fills up with sufficient water to facilitate further proton transport up to Y288 via conformational changes and hydrogen-bonded connections.