Characterizing the proton loading site in cytochromec oxidaseJianxun Lu and M. R. Gunner1

Department of Physics, City College of New York, New York, NY 10031

Edited by Arieh Warshel, University of Southern California, Los Angeles, CA, and approved July 15, 2014 (received for review April 18, 2014)

Cytochrome c oxidase (CcO) uses the energy released by reductionof O2 to H2O to drive eight charges from the high pH to low pHside of the membrane, increasing the electrochemical gradient.Four electrons and protons are used for chemistry, while fourmore protons are pumped. Proton pumping requires that residueson a pathway change proton affinity through the reaction cycle toload and then release protons. The protonation states of all resi-dues in CcO are determined in MultiConformational ContinuumElectrostatics simulations with the protonation and redox statesof heme a, a3, CuB, Y288, and E286 used to define the catalyticcycle. One proton is found to be loaded and released from residuesidentified as the proton loading site (PLS) on the P-side of the pro-tein in each of the four CcO redox states. Thus, the same protonpumping mechanism can be used each time CcO is reduced. Calcu-lations with structures of Rhodobacter sphaeroides, Paracoccusdenitrificans, and bovine CcO derived by crystallography and molec-ular dynamics show the PLS functions similarly in different CcOspecies. The PLS is a cluster rather than a single residue, as differentstructures show 1–4 residues load and release protons. However,the proton affinity of the heme a3 propionic acids primarily deter-mines the number of protons loaded into the PLS; if their protonaffinity is too low, less than one proton is loaded.

MCCE | bioenergetics | pKa | proton transfer

The electrochemical gradient across cellular, mitochondrial, orchloroplast membranes powers ATP synthesis and other key

biological functions (1, 2). The gradient is formed by protonstransferred from the high (N-) to low (P-) pH side of themembrane through proteins that carry out exergonic electrontransfer reactions (3, 4). The protons move because the redoxreactions modify the proton affinity of residues and cofactors bypKa shifts of the redox active species, by changing the long-rangeelectrostatic potential or through induced conformational changes(4). The accessibility of sites to N- and P-sides of the membranemay also change to ensure that protons are taken up from thecorrect side in each step. Despite the importance of these reac-tions, it has proved to be very challenging to understand howproteins control proton transfer thermodynamics and kinetics tomove protons uphill (3, 5).CcO is the well-studied terminal electron acceptor in the

aerobic respiratory electron transport chain; it uses the chemicalenergy liberated by reducing O2 to H2O to drive protons againstthe electrochemical gradient (5, 6). The overall reaction is

4cytc2+P +O2 + 8H+N → 4cytc3+P + 2H2O+ 4H+

P :

Four electrons are donated by four reduced cytochromes c onthe P-side of the membrane (cytc2+P ), whereas eight protons aretaken up from the N-side (H+

N) through the D- and K-channels(SI Appendix, Fig. S1) (7, 8). Four protons are consumed in theredox reaction in the binuclear center (BNC) and the other fourare pumped from the N- to P-side of the membrane (H+

P ) (9, 10).The BNC contains a Cu center (CuB), a high-spin Fe-heme(heme a3), and a tyrosine (Y288, Rhodobacter sphaeroides num-bering is used here; Fig. 1).

CcO adds to the proton gradient by two parallel mechanisms(2, 3, 5, 11). Because the substrate oxygen has a very low pKa,and the product water has a high pKa, the reaction cannot occurwithout four protons being bound. These substrate protons arebound from the N-side of the membrane adding to the protongradient. CcO also functions as a proton pump (9). Here, residuesburied in the protein change their proton affinity in a sequencecoupled to each reduction of the protein, and this results in pro-tons being bound from the N-side and released to the P-side. Theprotein needs to ensure that protons are not taken from the P-sidewhere their concentration is higher, although the mechanism bywhich this accomplished is not understood (12, 13).The CcO reaction cycle is divided into four redox states, R, F, O,

and E (Fig. 2 and SI Appendix, Fig. S2). Between two states, oneelectron and one proton are transferred into the BNC, and oneproton is pumped across the membrane (11, 14–17). In the R statewhen both heme a3 and CuB are reduced, oxygen binds to theBNC. Then heme a3, CuB, and Y288 (or heme a) (2) contributefour electrons to break the O2 bond. The initial product state isdenoted the P state, which progresses to the F state after a pro-ton is transferred to the BNC. After O2 reduction, which oxidizesthe CcO cofactors, the redox centers are rereduced by fourcytochromes c, with one electron added to sequentially form theO, E, R, and then F states.Protons for pumping are taken up via the D-channel, whereas

protons are given to the BNC from the D- and K-channels (8).E286, located at the inner terminal of the D-channel, passeschemical protons to the BNC or pumps protons outward to theP-side of the membrane (18, 19). The D- and K-channels areseen in crystal structures as narrow chains of waters and polarresidues (SI Appendix, Fig. S1). In contrast, the proton transferpathway from E286 to the P-side of the membrane is a complexnetwork of polar and ionizable residues.Measurements of the change in transmembrane potential

suggest that there is a group, denoted the proton loading site(PLS) that transiently holds the proton as it moves from E286

Significance

The transmembrane proton gradient is the primary source ofcellular energy. Cytochrome c oxidase pumps protons uphillacross the membrane using the energy released by the re-duction of O2. Proton transport requires that residues changetheir proton affinity during the reaction cycle. Calculation ofproton binding identifies a cluster of residues that bind andrelease a proton on each reduction of the active site. Theseresidues can constitute a proton loading site. The heme pro-pionic acids play the central role in proton loading. If the pro-pionic acid proton affinity is too low, no protons will be loaded.The same proton-pumping mechanism can be used for each ofthe four redox steps carried out in the reaction cycle.

Author contributions: J.L. and M.R.G. designed research; J.L. performed research; J.L. andM.R.G. analyzed data; and J.L. and M.R.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: gunner@sci.ccny.cuny.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1407187111/-/DCSupplemental.

12414–12419 | PNAS | August 26, 2014 | vol. 111 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1407187111

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020

along the exit pathway above heme a and a3 on P-side of theprotein (20, 21). However, it is not known what residue binds theproton or if the same residue is used in each redox state (22–25).The propionic acids on the A (PRA) or D (PRD) rings of heme a3(26) or His A334, which is a ligand to CuB, have all been proposedto be the PLS (15, 24, 27, 28). The energetics and pathway ofproton pumping has been simulated assuming the PRD of hemea3 is the PLS (26, 29–32). Because there are many ionizable groupson the P-side, the PLS could also be made up of a cluster ofresidues rather than a single site (SI Appendix, Fig. S1).In each redox state there must be a sequence of individual

substates that order the reduction of CcO and transfer of a protonto the BNC and to the P-side of the membrane. Time-resolvedmeasurements of the R-to-F transition help to define an order(22, 33). Because each reduction of the BNC is associated withproton pumping (14, 30, 34–36), it is assumed that the order ofsubstates is the same in each redox state (5). Because this is a cycle,the starting substate is arbitrary (Fig. 2 and SI Appendix, Fig. S2and Table S1). The basic cycle assumes: (i) a new active site redoxstate is initiated with the electron transfer from heme a to the BNC;(ii) this increases the BNC proton affinity so a proton is bound viathe D- or K-channel that will be used for chemistry. Other residuescan substitute even when proton input into the D-channel is blockedby mutation (37, 38); (iii) BNC protonation leads to proton releasefrom the nearby PLS to the P-side; (iv) heme a is reduced by cy-tochrome c via CuA; (v) which leads to loading the nearby PLS byproton transfer from E286. The second electron transfer to theBNC moves the system to the next redox state restarting the se-quence. The reduction states of heme a3, CuB, and Y288 andprotonation states of the oxygens on heme a3, CuB, or Y288define the system as being in the R, F, O, or E state (39, 40).The sequence of substates i–iv does not specify if E286 is

deprotonated transiently or if there are metastable states withE286 deprotonated. E286 has a high proton affinity, with an op-erational pKa for CcO turnover of 9.4 (35). The choice taken hereis to have the reduction of the BNC before the reprotonation ofE286. In addition, calculations are carried out where E286 pro-tonation state is not fixed. Previous calculations have reprotonatedE286 before (30) or after (32) electron transfer to the BNC.Simulations by Warshel, who considered many possible electronand proton transfer sequences, found the lowest energy barrierwhen the electron transfer to the BNC occurs before reprotona-tion of E286 (29, 32). The protonation of the PRD of heme a3 mayopen a cavity near E286 (26); this is calculated to significantlylower the proton affinity of E286, stabilizing its deprotonated statefor proton transfer to the BNC. Other simulations have suggested

rotating E286 between up or down conformations may (30, 31) ormay not (26) change its proton affinity.Continuum electrostatic calculations have been used to evaluate

the protonation states of all residues and the Em of heme a with theBNC fixed in different ionization states (24, 41, 42). Analysis of theproton acceptors near the BNC in each transition showed with fourelectrons added to the BNC, only two protons not four were bound(41, 42). Thus, the protein does not appear to maintain electro-neutrality (43, 44). This earlier work did not divide each BNCredox state into multiple substates but rather added an electronand investigated the equilibrium protonation of the system.The work presented here uses MultiConformation Continuum

Electrostatics (MCCE), which carries out continuum electrostaticanalysis with Monte Carlo sampling of protein conformation andprotonation states to calculate the equilibrium protonation of all

Fig. 1. Key groups in CcO. The redox and protonation states of heme a, a3,CuB, Y288, and E286 are defined to change through the reaction cycle (Fig. 2and SI Appendix, Fig. S2 and Table S1). CuA is always oxidized here. PRA andPRD, which are the propionic acids of heme a and a3, are allowed to come toequilibrium with the imposed charges in each substate.

Fig. 2. (Upper) The four redox states of the BNC. In R state O2 binds thereduced BNC. Each state is separated by the addition of one electron fromheme a and one proton from E286 to the BNC. Red, reduced; green, moreoxidized; blue, most oxidized. (Lower) The substates for the F state. Thesequence shown starts at the end of the R state (R6) with heme a3 and CuB

reduced, O2 bound in the BNC, and E286 having released the proton to loadthe PLS. Electron transfer from heme a to the BNC triggers reduction of O2 togenerate the PR state (F1 and F2 here). E286 is then protonated via theD-channel. The proton is transferred from E286 to the BNC. In the F state theproton acceptor is the hydroxyl ligand to CuB. E286 is then reprotonated andheme a is reduced by cytochrome c via CuA. Then E286 loses a proton, whichshould be loaded into the PLS for pumping. This substate (F6) is now pre-pared for the next electron transfer to form the O1 substate as shown. Thetransitions i, ii, iv, and v are described in the text (5). Transition iii is protonrelease from the PLS. In the calculations, proton release is not fixed in anystep; rather, it results from the change in the electrostatic environment, sothis step is not explicitly included in the reaction cycle. Rather, the number ofprotons bound to the PLS is monitored in all substates to identify whenprotons are released. In calculations of transition v, a proton is removedfrom E286 without designating the proton acceptor. See SI Appendix, Fig. S2and Table S1 for a description of the entire reaction cycle.

Lu and Gunner PNAS | August 26, 2014 | vol. 111 | no. 34 | 12415

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020

residues throughout the CcO reaction cycle. The aim is to de-termine if long-range electrostatic interactions are sufficient to leadto a cycle of proton binding and release from residues that can beidentified as the PLS. The R, F, O, and E redox states are eachbroken into six substates that define the redox and protonationstates of heme a, a3, CuB, Y288, and E286 and the substrate/product oxygen ligands on CuB and heme a3 (Figs. 1 and 2 and SIAppendix, Fig. S2 and Table S1) (33). Different structures of Rb.sphaeroides, Paracoccus denitrificans, and bovine CcO are in-vestigated. The protonation states of residues that could contributeto the PLS are monitored in each imposed substate. It is foundthat residues in the designated PLS cluster can take up and releaseas much as one proton through the substate sequence in eachredox state, indicating one mechanism can be used for pumping inall redox states. Residues that may contribute to the PLS areidentified. The PLS cluster load protons when heme a is reducedand release them when the BNC is protonated as suggested pre-viously (29–32). However, the protons are found to be bound andreleased over several substates rather than in a single transition.

Results and DiscussionEach of the four BNC redox states (R, F, O, and E) is dividedinto six substates (Fig. 2 and SI Appendix, Fig. S2 and Table S1).The sequence follows the order of states (i–v) described in theintroduction (5, 14) with two additional clarifications. Firstthe protonation state of E286 is defined. Electron transfer to theBNC can occur before or after reprotonation of E286. Here it isassumed to occur first. This choice sustains a more negativelycharged CcO, so highlights the electrostatic control of the protonaffinity of the PLS (SI Appendix, Fig. S2 and Table S1) (29, 32).In addition, calculations are carried out when the E286 pro-tonation state is free to equilibrate with the protein. The con-sensus model suggests the PLS will be loaded when heme a isreduced, and protons will be released from the PLS when pro-tons are transferred into the BNC (3, 5). Here we are trying todetermine which steps support changes in PLS proton occu-pancy, not to impose them. Therefore, after heme a reduction, aproton is removed from E286 without specifying where it will go.The calculated CcO equilibrium protonation then shows if theresidues in the PLS cluster bind protons. Also, there is no stepthat specifies PLS unloading; rather, the proton occupancy in thePLS region is free to show which transitions lead to proton loss.The pattern of proton uptake and release is first described incalculations using the Rb. sphaeroides crystal structure 1M56 andthen compared with the results in other CcO structures (SI Ap-pendix, Table S1) (26).

Changes in PLS Protonation in the Rb. sphaeroides Structure 1M56.Fig. 3A shows the imposed charge in the cycle varies from +1 to−1 (red line). The total CcO charge equilibrated in each substate(black line) changes. In the 1M56 structure, the change in totalprotein charge is approximately half that imposed, indicatingthat protons are released to solution when the imposed charge ispositive and bound when it is negative. The protonation changesare summed for all of the ionizable residues in a 10-Å spherearound the PRD of heme a3, which are evaluated as membersof the PLS (green line, SI Appendix, Fig. S1). The rest of theprotein is denoted the non-PLS region (purple line). Thesecalculations do not follow the pathway of proton transfer, sothere is no way to know if these protons are loaded from the N-or P-side. What is shown is that the PLS region does bind andrelease protons as the electrostatic potential changes through theimposed reaction cycle.The PLS holds the fewest protons when the BNC has received

the chemical proton with heme a oxidized and E286 protonated(substate R4, F4, O4, E4, denoted X4; Figs. 2 and 3). Thissubstate, with the most positive imposed charge, is found to bethe end of the proton release and the beginning of the nextproton-loading process. Protons are added to the PLS whenheme a is reduced, as suggested previously (X5) (5, 29–32). Moreprotons are loaded when E286 is deprotonated without assigning

the proton acceptor (X6). The electron on heme a is thentransferred into the BNC to generate the next redox state (X1).This electron transfer does not change the net charge of thewhole protein yet induces additional proton uptake to the PLS.The reprotonation of E286 begins the loss of protons from thePLS (X2). The chemical proton is transported from E286 intothe BNC (X3); this is another step with no change in net imposedcharge, yet it is associated with proton release from the PLS.Finally, E286 is reprotonated and the PLS reaches its minimumprotonation level (X4).The pattern for proton loading and release is repeated for all

four BNC redox states, although there is some variation in theamount of protons bound (Fig. 3A). Changes of less than oneproton bound or released indicate that the site is near its pKa andso is partially protonated at the beginning and/or that the pKadoes not shift enough for the residue to change from fully pro-tonated to fully deprotonated (SI Appendix, Fig. S3). Though theprotons are not constrained to go to the PLS, the equilibratedCcO does bind and release protons into this region. Heme a isoxidized during proton loading and reduced during proton re-lease. Protonation of E286 and chemical proton transfer fromE286 to the BNC also contribute to proton release.

Identity of the Residues in the PLS That Change Protonation in EachSubstate Cycle. Twenty-two groups are monitored, providing anunbiased test of which ones are active members of the PLS (SIAppendix, Fig. S2 and Table S1). In 1M56, all four propionicacids of the heme a and a3 are involved in proton loading andrelease (SI Appendix, Fig. S4). These four acids are all deproto-nated when the PLS has the fewest protons bound; they are par-tially protonated as heme a is reduced and E286 deprotonated(X5 and X6). When the electron is transferred from heme a to theBNC (X1), the captured proton goes to the heme a3 propionates,with the heme a propionates returning to their deprotonated state.The heme a3 propionates then lose their protons over the fol-lowing steps (SI Appendix, Fig. S4).

Protonation Changes Outside of the PLS in 1M56. The non-PLSresidues also respond to buffer the imposed charge. In 1M56, thenon-PLS residues take up and release 0.6–0.8 protons in differ-ent redox states (Fig. 3A). Because these residues bind and re-lease protons from the same side of the membrane, they do notcontribute to pumping. Approximately 10–20 residues contributeto this distal protonation change (SI Appendix, Fig. S5 and TableS4). No individual non-PLS residue binds more than 0.3 protons.Most of these residues are relatively solvent-exposed on bothinner and outer surfaces; they start out partially protonated, withtheir pKa ∼pH 7, so a small change in the electrostatic potentialleads to changes in protonation. Other internal residues, such asGlu C90 and His C212, are in ion pairs, where proton transferoccurs within the pair and so makes little contribution to thetotal charge change.

Proton Loading Cycle in Different CcO Structures. The same CcO reac-tion cycle is imposed on five additional crystal structures, includingthree from Rb. sphaeroides, one from Paracoccus denitrificans, onefrom bovine as well as six Rb. sphaeroides Molecular Dynamics(MD) snapshots (SI Appendix, Table S3). The response of the PLSin these structures is divided into three groups (SI Appendix, Fig.S6). In the first group, proton uptake and release is seen in the PLScluster in all four redox states. Changes in protonation vary from∼ 0.6 to 1 proton, with more significant uptake in E and R than in Fand O portions of the cycle (SI Appendix, Fig. S6A). The secondgroup supports loading and release of ∼1 proton in E and R states,but with significantly less proton uptake in F and O—this includes1M56, which has been described above, several other Rb. sphaer-oides MD snapshots, and the bovine CcO structure 2OCC (SI Ap-pendix, Fig. S6B). The last group shows little or no proton loading inthe PLS; it includes the Rb sphaeroides structures 2GSM and 3FYE,and the P. denitrificans structure 1AR1 (SI Appendix, Fig. S6C).

12416 | www.pnas.org/cgi/doi/10.1073/pnas.1407187111 Lu and Gunner

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020

The Rb. sphaeroides structures 1M56, 2GSM, and 3FYE showa range of proton uptake into the PLS. The MD snapshotsderived from 1M56 can either take up more (1M56a) or less(1M56b) protons than the parent structure. The behavior is notcorrelated with the size of the cavity near E286 that controls theGlu proton affinity (16, 26). The behavior of the P. denitrificansand bovine CcO structures fits within the range found with theRb. sphaeroides CcO structures.In the eight structures that show proton uptake, a small subset of

residues monitored as the PLS contributes to proton loading. Indifferent Rb. sphaeroides structures, in addition to the four pro-pionic acids, Arg A52, 412, 481, and 482, and Lys B227 load at least0.1 proton in at least one structure. In 1M56, all four propionateschange charge (SI Appendix, Fig. S4). In 1M56f, only PRA of hemea3 takes up ∼0.9 protons. Five structures use three or four residues,whereas two structures use two residues and one uses a singleresidue. In P. denitrificans, Lys B191 and the heme a3 acids loadprotons. In the bovine structure, both heme a3 acids, Asp A52 andLys B171, contribute. Thus, the PLS is more of a cluster thana single site. Which residues are active is influenced by smalldifferences between structures, which are highlighted in thesecalculations where the protein backbone is fixed (SI Appendix,Tables S5 and S6). This suggests that in an experimental ensembleof structures, protons could be loaded to different sites. However,the PRA and PRD of heme a3 are most likely to play roles inproton uptake. These two groups will be examined to understandwhy more protons are loaded in the E and R states than in F andO, and why some structures load more protons than others.

Comparison of the Four BNC Redox States. In most CcO structuresinvestigated, there are more protons loaded in E and R statesthan in F and O. The net imposed charge and the heme a andE286 charge distributions are independent of redox state.

However, the BNC charge distribution is different in the fourredox states (Fig. 3A and SI Appendix, Fig. S2 and Table S1). Theinteraction of PRA and PRD of heme a3 with the BNC wascalculated in each structure (SI Appendix, Fig. S7). The elec-trostatic potential from the BNC at these two propionic acids is∼120 mV more negative in the E and R states than that in F andO, although the net charge is the same in all redox states. Thenegative potential raises the PRA and PRD proton affinity, in-creasing their pK′7 by ∼2 pH units in E and R states.

The Factors That Change the PLS Proton Uptake. In all structures, thefour propionic acids are fully ionized in the substate where thePLS has the fewest protons (X4). Fig. 4 compares the protonuptake in the E state, which generally has a large uptake, and theR state, which has lower uptake with the proton affinity of thePRA and PRD of heme a3 as given by their pK′7s (Eq. 1). In all ofthe structures that show significant proton loading, at least one ofthe heme a3 propionates has its pK′7 near or above 7. Comparingthese structures shows changes in interactions of the propionicacids with other PLS residues and with the backbone dipolescontribute to the difference in pK′7; this is not unexpected given thedensity of charged groups in the PLS—however, it indicates onespecial interaction does not decide between structures that loadprotons and ones that do not (SI Appendix, Tables S5 and S6). Theproton affinity of the propionic acids is lowered by as much as 3.6pH units when a proton has moved into the BNC. Thus, the long-range electrostatic interactions with the chemical proton in theBNC can push the proton out of the PLS (5, 10, 11, 30).To test the importance of the propionic acids, their proton

affinity is increased in 2GSM, a structure that does not loadprotons into the PLS (SI Appendix, Figs. S8 and S9). The pK′7 ofheme a and a3 propionic acids range from 0.1 to 5.5 in thesubstates where protons should be loaded. The PLS does not

Fig. 3. (A) The charge change in the CcO 1M56 structure as it goes through the defined substates reaction cycle (Fig. 2). Each summed charge is providedrelative to the charge in the R4 substate where the PLS has the fewest protons bound. R4 is formed by reprotonation of E286 with heme a oxidized. Red line,charge imposed on BNC (the summed charge of Y288, heme a3, CuB, and their ligands), E286, and heme a. Green line, net charge of the PLS. Purple line,charge of the non-PLS. Black line, total charge of the protein. (B) Substate cycle without fixing E286 in 1M56a. Now only substates X2, X4, and X5 are imposedfor each of the four redox states (brown circles). E286 remains neutral in all substates. The imposed charge, which now includes heme a, a3, CuB, and Y288 only(red line), varies from 0 to 1. The sum charge on the PLS with (green line)/without (brown line) E286 fixed is shown.

Lu and Gunner PNAS | August 26, 2014 | vol. 111 | no. 34 | 12417

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020

load protons in F and O states and binds ∼0.2 protons in E and R.When the proton affinity of heme a3 acids is increased to shiftthe pK′7 by four pH units, 0.6 protons are taken up in F and Ostates, and one proton E and R. Interestingly, if both heme a3 acidsare fixed in their neutral states in the cycle, the PLS loads oneproton. Arg A481 and Asp A407, which are hydrogen bonded toPRD and PRA of heme a3, respectively, become the sites for protonloading. Several residues in the PLS cluster have been mutated suchas D407 (45) and R481 (46, 47) with modest changes in pumping;this is consistent with the finding here that more than one residuemakes up the PLS and they can substitute for each other.

The Proton Affinity of E286. The protonation state of E286 is fixedin the imposed substates that generate the results shown in Fig.3A. The sequence chosen has reprotonation of E286 occurringafter electron transfer from heme a to the BNC (Fig. 2). Asfound previously, the E286 pK′7 is >10 in all substates for allcrystal structures and some MD snapshots (26). With such a highproton affinity, the deprotonated E286 may be very unstable sothe assumption made here that the PLS comes to equilibriumwith an anionic Glu may not be correct. The proton uptake intothe PLS was determined in the structure 1M56a, which showsrobust pumping, without fixing the protonation state of E286.There are now only three defined substates: X2, X4, and X5,because the substates with a fixed deprotonated E286 are notconsidered (Fig. 2). The free E286 remains protonated. How-ever, the proton loading and release in the PLS is still seen (Fig.3B and SI Appendix, Fig. S10); it is smaller in the F and O statesand less changed in the E and R states, which is not unexpected,because the charge distribution in the BNC raises the protonaffinity of the PLS less in the F and O states so the contributionof a negatively charged E286 is more important there. Thus,proton loading still occurs when E286 remains protonated.The simulation with free E286 was repeated in the 1M56e

structure, which has a large cavity near E286 that is opened whenMD simulations are run in the presence of a protonated PRD ofheme a3 (26). This structural change better solvates the ionizedE286, stabilizing the deprotonated state. Robust pumping is seenin the standard 24-substate cycle (SI Appendix, Fig. S6A). In thissimulation, the pK′7 of E286 and the PRA of heme a3 are bothclose to 7 (SI Appendix, Fig. S10). Thus, E286 and the PRA areboth partially protonated in all substates and so both respond tothe changes in electrostatic interactions imposed by changing thecharge on heme a and the BNC (SI Appendix, Fig. S11). With thelower E286 proton affinity, there is a mixture of states with either

E286 or the PLS protonated, revealing the possibility of protontransfer from E286 to the PLS (26).

ConclusionThe work presented here addresses the long-standing questionsabout the identity of the PLS in CcO and determines if long-rangeelectrostatic interactions are sufficient to lead to proton loadingand release. A cluster of residues play roles showing the PLS is nota single site; this is not unexpected given the number of ionizableresidues in the region on P-side of the protein. Different inputstructures show different amounts of proton loading. The protonaffinity of the heme a3 acids controls the number of protonsbound; if their pK′7s (Eq. 1) are much lower than 7, they stay fullyionized and the PLS does not load protons through the cycle. Thisbehavior would be seen in uncoupled CcO where O2 is reducedwith no proton pumping. When the propionate pK′7s are ≥7 in thesubstates where heme a is reduced and E286 deprotonated, pro-ton loading and release is seen. Interestingly, if the propionic acidsare fixed in their protonated states through the cycle, one proton isloaded in each CcO redox state, using residues that are hydrogenbonded to the now neutral acids.The behaviors of P. denitrificans and bovine CcO are highly

similar to that found for Rb. sphaeroides CcO. In addition, thesimilarity of proton loading and release shows that one mecha-nism suffices to allow pumping one proton each time the proteinis reduced in all A-type CcOs. The results are thus consistentwith previous models suggesting that protons should be loadedwhen heme a is reduced, and released when a proton is trans-ferred to the BNC (5). However, the free calculations showloading and release occur over several substates of the reactioncycle rather than being coupled to a single step of the reaction.

Materials and MethodsProtein Coordinates. Crystal structures 1M56 (48), 2GSM (49), 3FYE (50)(Rh. sphaeroides), 1AR1 (51) (P. denitrificans), and 2OCC (52) (bovine) wereanalyzed. In addition, MD snapshots initiated with 1M56 coordinates indifferent CcO redox states were studied (denoted 1M56a–1M56f). 1M56a–1M56d have a small cavity near E286 as found in the crystal structure.1M56e and 1M56f have a large and water-filled cavity near E286 (26). Allstructures have E286 in the downward direction toward the D-channel. Theredox states fixed in the MD simulation that produced the snapshots are givenin SI Appendix, Table S3.

Substates in the Calculation. The substate order relies on experimental studiesof the electron and proton transfer in the PR-to-F transition (5, 14, 33) and isthe same as used previously in other simulations (29). The PR state occursbetween R and F after electron transfer into the BNC breaking the O–Obond to form the feryl heme a3. PR is formed when heme a is reduced beforeoxygen reduction, and heme a, not Y288, provides the fourth electronneeded for O2 reduction (14, 53). In the substate cycle the PR state is denotedF1 and F2. Though there is only information about the substates in the PR/Fredox state, each of the four redox states will be assumed to proceed throughthe same electron and proton transfer sequence—the only difference is whichcofactor in the BNC is reduced and whether the aquo ligand to heme a3 or CuBor Y288 receives the proton (Fig. 2 and SI Appendix, Table S1).

MCCE Calculations. Though the redox and protonation states of heme a, a3,CuB, Y288, and E286 are fixed, the protonation states of the rest of theprotein are allowed to reach equilibrium. MCCE (54) is used to calculate theprotonation states of CcO in all substates. The method has been describedpreviously (41, 42), and additional details are provided in SI Appendix.

MCCE Analysis of the Proton Affinity. The proton affinity of a residue can becalculatedwhilemaintaining the ionization states of all other residues in theirequilibrium states at a given pH (26). For each residue equilibrated with thebulk at pH 7, the free energy of ionization is

ΔGðAH→A−Þ= 1:36ðpK7′ − 7Þkcal=mol: [1]

Residues That May Be in the PLS. All ionizable residues within 10-Å of thePRD of heme a3 are assessed as potential members of the PLS. In Rh.sphaeroides CcO, it includes 22 residues: Arg A52, 407, 408, 481, 482; Lys

Fig. 4. Dependence of the PLS proton uptake on the proton affinity of hemea3 acids (pK′7). The change in PLS protonation in the R1 (●) and E1 (○) substatesrelative to that found in the R4 substate, which has the fewest protons bound.The higher pK′7 of heme a3 acids is used for each CcO structure in R1 or E1substate. The linear regression line has an R2 of 0.729.

12418 | www.pnas.org/cgi/doi/10.1073/pnas.1407187111 Lu and Gunner

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020

B227; His A411; Asp A407, 412, and B229; Glu A54 and B254; Tyr A336, 409, 410,414, 415, 483, and the PRA and PRD of heme a and a3. The ligands to Mg2+,His A411, Asp A412, and Glu B254 are also included. The residues in the 10-Åsphere in P. denitrificans and bovine CcO are listed in SI Appendix, Table S2.The residue charges are summed in each substate of the reaction sequence.

ACKNOWLEDGMENTS. Helpful discussion with Qiang Cui and Puja Goyal andaccess to their MD snapshots are gratefully acknowledged. This work wassupported by National Science Foundation Grant MCB 1022208, withinfrastructure support from National Center for Research Resources Grant2G12RR03060 and National Institute on Minority Health and HealthDisparities Grant 8G12MD007603 from the National Institutes of Health.

1. Mitchell P, Moyle J (1965) Stoichiometry of proton translocation through the re-spiratory chain and adenosine triphosphatase systems of rat liver mitochondria. Na-ture 208(5006):147–151.

2. Babcock GT, Wikström M (1992) Oxygen activation and the conservation of energy incell respiration. Nature 356(6367):301–309.

3. Brzezinski P, Gennis RB (2008) Cytochrome c oxidase: Exciting progress and remainingmysteries. J Bioenerg Biomembr 40(5):521–531.

4. Gunner MR, Amin M, Zhu X, Lu J (2013) Molecular mechanisms for generatingtransmembrane proton gradients. Biochim Biophys Acta 1827(8-9):892–913.

5. Kaila VR, Verkhovsky MI, Wikström M (2010) Proton-coupled electron transfer in cy-tochrome oxidase. Chem Rev 110(12):7062–7081.

6. Brzezinski P, Adelroth P (2006) Design principles of proton-pumping haem-copperoxidases. Curr Opin Struct Biol 16(4):465–472.

7. Jünemann S, Meunier B, Gennis RB, Rich PR (1997) Effects of mutation of the conservedlysine-362 in cytochrome c oxidase from Rhodobacter sphaeroides. Biochemistry 36(47):14456–14464.

8. Konstantinov AA, Siletsky S, Mitchell D, Kaulen A, Gennis RB (1997) The roles of thetwo proton input channels in cytochrome c oxidase from Rhodobacter sphaeroidesprobed by the effects of site-directed mutations on time-resolved electrogenic in-traprotein proton transfer. Proc Natl Acad Sci USA 94(17):9085–9090.

9. Wikstrom MK (1977) Proton pump coupled to cytochrome c oxidase in mitochondria.Nature 266(5599):271–273.

10. Michel H (1999) Proton pumping by cytochrome c oxidase. Nature 402(6762):602–603.11. Rich PR, Jünemann S, Meunier B (1998) Protonmotive mechanism of heme-copper

oxidases. J Bioenerg Biomembr 30(1):131–138.12. Kim YC, Wikström M, Hummer G (2009) Kinetic gating of the proton pump in cyto-

chrome c oxidase. Proc Natl Acad Sci USA 106(33):13707–13712.13. Ferguson-Miller S, Babcock GT (1996) Heme/copper terminal oxidases. Chem Rev

96(7):2889–2908.14. Brändén G, Gennis RB, Brzezinski P (2006) Transmembrane proton translocation by

cytochrome c oxidase. Biochim Biophys Acta 1757(8):1052–1063.15. Belevich I, Bloch DA, Belevich N, Wikström M, Verkhovsky MI (2007) Exploring the

proton pump mechanism of cytochrome c oxidase in real time. Proc Natl Acad Sci USA104(8):2685–2690.

16. Wikström M, Verkhovsky MI, Hummer G (2003) Water-gated mechanism of protontranslocation by cytochrome c oxidase. Biochim Biophys Acta 1604(2):61–65.

17. Siletsky SA, et al. (2007) Time-resolved single-turnover of ba3 oxidase from Thermusthermophilus. Biochim Biophys Acta 1767(12):1383–1392.

18. Puustinen A, Wikström M (1999) Proton exit from the heme-copper oxidase of Es-cherichia coli. Proc Natl Acad Sci USA 96(1):35–37.

19. Kaila VR, Verkhovsky MI, Hummer G, Wikström M (2009) Mechanism and energeticsby which glutamic acid 242 prevents leaks in cytochrome c oxidase. Biochim BiophysActa 1787(10):1205–1214.

20. Seibold SA, Mills DA, Ferguson-Miller S, Cukier RI (2005) Water chain formation andpossible proton pumping routes in Rhodobacter sphaeroides cytochrome c oxidase: Amolecular dynamics comparison of the wild type and R481K mutant. Biochemistry44(31):10475–10485.

21. Wikström M, et al. (2005) Gating of proton and water transfer in the respiratoryenzyme cytochrome c oxidase. Proc Natl Acad Sci USA 102(30):10478–10481.

22. Faxén K, Gilderson G, Adelroth P, Brzezinski P (2005) A mechanistic principle forproton pumping by cytochrome c oxidase. Nature 437(7056):286–289.

23. Belevich I, Verkhovsky MI, Wikström M (2006) Proton-coupled electron transfer drivesthe proton pump of cytochrome c oxidase. Nature 440(7085):829–832.

24. Quenneville J, Popovi�c DM, Stuchebrukhov AA (2006) Combined DFT and electro-statics study of the proton pumping mechanism in cytochrome c oxidase. BiochimBiophys Acta 1757(8):1035–1046.

25. Sharpe MA, Ferguson-Miller S (2008) A chemically explicit model for the mechanismof proton pumping in heme-copper oxidases. J Bioenerg Biomembr 40(5):541–549.

26. Goyal P, Lu J, Yang S, Gunner MR, Cui Q (2013) Changing hydration level in an in-ternal cavity modulates the proton affinity of a key glutamate in cytochrome c oxi-dase. Proc Natl Acad Sci USA 110(47):18886–18891.

27. Kaila VR, Sharma V, Wikström M (2011) The identity of the transient proton loadingsite of the proton-pumping mechanism of cytochrome c oxidase. Biochim BiophysActa 1807(1):80–84.

28. Sugitani R, Medvedev ES, Stuchebrukhov AA (2008) Theoretical and computationalanalysis of the membrane potential generated by cytochrome c oxidase upon singleelectron injection into the enzyme. Biochim Biophys Acta 1777(9):1129–1139.

29. Olsson MH, Siegbahn PE, Blomberg MR, Warshel A (2007) Exploring pathways andbarriers for coupled ET/PT in cytochrome c oxidase: A general framework for exam-ining energetics and mechanistic alternatives. Biochim Biophys Acta 1767(3):244–260.

30. Fadda E, Yu CH, Pomès R (2008) Electrostatic control of proton pumping in cyto-chrome c oxidase. Biochim Biophys Acta 1777(3):277–284.

31. Yamashita T, Voth GA (2012) Insights into the mechanism of proton transport incytochrome c oxidase. J Am Chem Soc 134(2):1147–1152.

32. Pisliakov AV, Sharma PK, Chu ZT, Haranczyk M, Warshel A (2008) Electrostatic basisfor the unidirectionality of the primary proton transfer in cytochrome c oxidase. ProcNatl Acad Sci USA 105(22):7726–7731.

33. Morgan JE, Verkhovsky MI, Palmer G, Wikström M (2001) Role of the PR intermediatein the reaction of cytochrome c oxidase with O2. Biochemistry 40(23):6882–6892.

34. Wikström M, Verkhovsky MI (2002) Proton translocation by cytochrome c oxidase indifferent phases of the catalytic cycle. Biochim Biophys Acta 1555(1-3):128–132.

35. Namslauer A, Aagaard A, Katsonouri A, Brzezinski P (2003) Intramolecular proton-transfer reactions in a membrane-bound proton pump: The effect of pH on theperoxy to ferryl transition in cytochrome c oxidase. Biochemistry 42(6):1488–1498.

36. Johansson AL, et al. (2011) Proton-transport mechanisms in cytochrome c oxidase re-vealed by studies of kinetic isotope effects. Biochim Biophys Acta 1807(9):1083–1094.

37. Wikström M, Verkhovsky MI (2011) The D-channel of cytochrome oxidase: An alter-native view. Biochim Biophys Acta 1807(10):1273–1278.

38. Belevich I, et al. (2010) Initiation of the proton pump of cytochrome c oxidase. ProcNatl Acad Sci USA 107(43):18469–18474.

39. Kim YC, Hummer G (2012) Proton-pumping mechanism of cytochrome c oxidase: Akinetic master-equation approach. Biochim Biophys Acta 1817(4):526–536.

40. Siegbahn PE, Blomberg MR (2008) Proton pumping mechanism in cytochrome c oxi-dase. J Phys Chem A 112(50):12772–12780.

41. Song Y, Michonova-Alexova E, Gunner MR (2006) Calculated proton uptake on an-aerobic reduction of cytochrome C oxidase: Is the reaction electroneutral? Bio-chemistry 45(26):7959–7975.

42. Zhang J, Gunner MR (2010) Multiconformation continuum electrostatics analysis ofthe effects of a buried Asp introduced near heme a in Rhodobacter sphaeroides cy-tochrome c oxidase. Biochemistry 49(37):8043–8052.

43. Mitchell R, Rich PR (1994) Proton uptake by cytochrome c oxidase on reduction and onligand binding. Biochim Biophys Acta 1186(1-2):19–26.

44. Popovi�c DM, Stuchebrukhov AA (2004) Electrostatic study of the proton pumpingmechanism in bovine heart cytochrome C oxidase. J Am Chem Soc 126(6):1858–1871.

45. Qian J, et al. (1997) Aspartate-407 in Rhodobacter sphaeroides cytochrome c oxidase isnot required for proton pumping or manganese binding. Biochemistry 36(9):2539–2543.

46. Mills DA, Ferguson-Miller S (2002) Influence of structure, pH and membrane potentialon proton movement in cytochrome oxidase. Biochim Biophys Acta 1555(1-3):96–100.

47. Mills DA, et al. (2005) An arginine to lysine mutation in the vicinity of the hemepropionates affects the redox potentials of the hemes and associated electron andproton transfer in cytochrome c oxidase. Biochemistry 44(31):10457–10465.

48. Svensson-EkM, et al. (2002) The X-ray crystal structures of wild-type and EQ(I-286) mutantcytochrome c oxidases from Rhodobacter sphaeroides. J Mol Biol 321(2):329–339.

49. Qin L, Hiser C, Mulichak A, Garavito RM, Ferguson-Miller S (2006) Identification ofconserved lipid/detergent-binding sites in a high-resolution structure of the mem-brane protein cytochrome c oxidase. Proc Natl Acad Sci USA 103(44):16117–16122.

50. Qin L, et al. (2009) Redox-dependent conformational changes in cytochrome C oxi-dase suggest a gating mechanism for proton uptake. Biochemistry 48(23):5121–5130.

51. Ostermeier C, Harrenga A, Ermler U, Michel H (1997) Structure at 2.7 A resolution ofthe Paracoccus denitrificans two-subunit cytochrome c oxidase complexed with anantibody FV fragment. Proc Natl Acad Sci USA 94(20):10547–10553.

52. Yoshikawa S, et al. (1998) Redox-coupled crystal structural changes in bovine heartcytochrome c oxidase. Science 280(5370):1723–1729.

53. Einarsdóttir O, Szundi I (2004) Time-resolved optical absorption studies of cytochromeoxidase dynamics. Biochim Biophys Acta 1655(1-3):263–273.

54. Song Y, Mao J, Gunner MR (2009) MCCE2: improving protein pka calculations withextensive side chain rotamer sampling. J Comput Chem 30(14):2231–2247.

Lu and Gunner PNAS | August 26, 2014 | vol. 111 | no. 34 | 12419

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

29, 2

020