Functional interactions between membrane-bound transporters and membranes

Abstract

One key role of many cellular membranes is to hold a transmembrane electrochemical ion gradient that stores free energy, which is used, for example, to generate ATP or to drive transmembrane transport processes. In mitochondria and many bacteria, the gradient is maintained by proton-transport proteins that are part of the respiratory (electron-transport) chain. Even though our understanding of the structure and function of these proteins has increased significantly, very little is known about the specific role of functional protein-membrane and membrane-mediated protein-protein interactions. Here, we have investigated the effect of membrane incorporation on proton-transfer reactions within the membrane-bound proton pump cytochrome c oxidase. The results show that the membrane acts to accelerate proton transfer into the enzyme’s catalytic site and indicate that the intramolecular proton pathway is wired via specific amino acid residues to the two-dimensional space defined by the membrane surface. We conclude that the membrane not only acts as a passive barrier insulating the interior of the cell from the exterior solution, but also as a component of the energy-conversion machinery.

One characteristic feature of a living cell is an electrochemical potential that is maintained by ion transporters or pumps (generators) across many of its membranes. In this context the membrane acts as an insulating barrier defining separate compartments with different ion concentrations. Results from recent studies point also to more direct roles of the membrane in function of the cellular machinery. For example, the membrane may transmit external mechanical forces to ion channels (1), it may modulate function of membrane proteins, or specific lipid molecules may even act as “cofactors” of membrane proteins (2). Here, we focus on yet another aspect of membrane protein-membrane interactions, i.e., interactions between the protein and membrane surfaces that modulate proton-delivery rates to buried intraprotein sites.

The proton is the key player in energy conversion of most cells. In living cells protons are translocated by transmembrane electrochemical potential generators and the free energy stored in this potential is used either directly, linking the backflow of protons to transmembrane transport of other molecules or ions, or to synthesize ATP by the ATP synthase. Early functional considerations of this energy-conversion machinery suggested that biological membranes may facilitate proton transfer between proton pumps (potential generators) and ATP synthases, as well as other transport proteins (3). More recent experimental and theoretical studies indicate that rapid proton diffusion along surfaces of biological membranes plays a functional role (for review, see refs. 4–7).

In the present study we investigated functional interactions between a membrane-bound proton transporter, cytochrome c oxidase (CytcO) and the surrounding membrane. CytcO is an enzyme that catalyzes the reduction of molecular oxygen to water and uses part of the free energy to pump protons across the membrane from the negative (n) to the positive (p) side (reviewed in refs. 8–12). The CytcO from Rhodobacter sphaeroides (cytochrome aa₃) consists of four subunits (SUs), which harbor four redox-active metal sites located in SUs I and II (13, 14). During turnover electrons are transferred from an external electron donor, cytochrome (cyt.) c, to a copper center (Cu_A) in SU II, located near the p-side surface of CytcO, and then via a heme group (heme a) in SU I, to the catalytic site composed of heme a₃ and Cu_B, also located in SU I. Protons that are used to reduce O₂ to H₂O are taken up from the opposite (n) side of the membrane:

[1a]

[1b]

where the catalytic reaction in Eq. 1a is linked to proton pumping indicated in Eq. 1b. The subscripts n and p refer to the two sides of the membrane.

One of the proton-conducting pathways, called the K pathway, used to transfer protons from the n-side solution to the catalytic site, starts at the Glu^II101 (the superscript II indicates the SU number) residue in subunit II, near the protein-membrane interface (15, 16) (Fig. 1) and continues toward the catalytic site via a highly conserved Lys^I362. Proton transfer through this pathway can be studied time resolved upon flash photolysis of a carbon monoxide (CO) ligand from the heme a₃ iron in CytcO in which the catalytic site is reduced while heme a and Cu_A are oxidized (15, 17) (often referred to as the mixed-valence CytcO). Because the CO ligand stabilizes the reduced state of heme a₃, after flash photolysis of the complex the electron initially residing at heme a₃ is transferred to heme a over time scales < 3 μs (18–20). On a slower time scale, there is further electron transfer from heme a₃ to heme a (and to some extent also Cu_A), which is coupled to proton release from the catalytic site to solution on the n side of the membrane, via the K pathway, with a time constant of ∼1 ms (17). The proton donor is a water molecule at the catalytic site, which interacts electrostatically with heme a₃. Upon proton release, the formed hydroxide binds to the oxidized heme a₃ (21). Even though this proton release occurs in the opposite direction to that during turnover, the kinetics of the proton-coupled electron transfer (PCET) reflects both the proton uptake and release rates (the sum of these rates) through the K pathway (15, 17). Formation of the mixed-valence-CO (from the oxidized form) was found to be accompanied by the uptake of approximately two protons, and this value was almost independent of pH (22).

Fig. 1.

Structure of cytochrome c oxidase from R. sphaeroides. The overall structure (four subunits) is shown in the background (PDB ID code 1M56). The redox-active cofactors, Cu_A, heme a, heme a₃, and Cu_B, are shown together with a number of amino acid residues that participate in proton transfer through the K-proton-transfer pathway. Residue Ser^I299 is not part of the proton pathway, but it is hydrogen-bonded to Lys362 (via a water molecule), which is a key element of the pathway. The thick () and thin arrows () indicate rapid proton transfer in the membrane-reconstituted CytcO and slower proton transfer in detergent-solubilized CytcO, respectively. The arrow within CytcO indicates proton transfer via Lys^I362 within the K proton pathway.

Replacement of the Lys^I362 residue by, e.g., a Met essentially abolishes proton transfer through the K pathway, and the activity drops to < 2% of that of the wild-type CytcO (23). Proton transfer from solution to the catalytic site, via the Lys^I362 residue, requires reisomerization of the side chain to reach between the lower and upper parts of the pathway (24, 25). Thus, locking the Lys residue by introduction of an Asp residue at the position of Ser^I299 results in slowed proton transfer through the K pathway (15). Furthermore, replacement of the surface-exposed Glu^II101, at the K-pathway entrance, by either an Ala or Asp, leads to slowed proton transfer through the pathway (15, 16), which is manifested as a lower steady-state activity of CytcO (16, 26). This activity could be partially restored in the Glu^II101Ala mutant (also in combination with His^II96Ala) upon addition of arachidonic acid, cholic acid, or deoxy-cholate (27, 28), where deoxy-cholate has been shown to bind/interact near the K pathway with its carboxyl group pointing toward the position of Glu^II101 in the structure of the wild-type CytcO (27). Furthermore, Qin et al. reported an inhibitory cadmium-binding site near the entrance of the K pathway (14). These results indicate that the acid group near the K-pathway entrance is important for maintaining rapid proton delivery to the pathway. It has also been suggested that Glu^II101 may act as a gate controlling proton delivery into the K pathway (29). Furthermore, it has been shown that structural changes in the K pathway are orchestrated with transitions between catalytic intermediates in CytcO (30–32), which suggests that functional interactions between this pathway and the membrane are important for understanding function.

Results from recent studies have shown that in the Paracoccus denitrificans CytcO, the PCET after dissociation of CO from the two-electron reduced CytcO and the proton release through the K pathway accelerates by a factor of ∼7 upon insertion of the CytcO into lipid vesicles (33). It was suggested that the effect may be due to changes in the conductivity of the pathway or that the negatively charged membrane surface may act as a proton acceptor thereby increasing the rate of proton release. In the present study we used a combination of site-directed mutagenesis and kinetic studies to investigate the PCET in CytcO in a membrane environment. The results indicate that the K proton pathway is wired via an acidic residue (Glu^II101) to the two-dimensional space defined by the membrane surface.

Results

Fig. 2A shows absorbance changes associated with PCET from heme a₃ to heme a for CytcO in detergent solution and CytcO incorporated into the membrane of small unilamellar vesicles (SUVs). As seen in Fig. 2, upon reconstitution of CytcO into the membrane, the rate constant increased from ∼830 s^-1 to ∼5,900 s^-1 (at pH 9.5), i.e., by a factor of 7, in agreement with an earlier study with the P. denitrificans CytcO (33). We performed the studies at high pH (9.5) because at this pH, the extent of the PCET displays a maximum and also the effect of the membrane on the proton-transfer rate increases with increasing pH (34). To test whether the acceleration of the PCET was due to a change in conductivity of the proton pathway or due to specific interactions with the membrane surface, we investigated the PCET in a number of mutant CytcOs in which residues within the pathway or at the surface near the pathway orifice were modified (summarized in Table 1).

Fig. 2.

PCET for wild-type and mutant forms of CytcO in DDM and reconstituted into SUVs. Absorbance changes at 598 nm after flash photolysis of CO from the two electron reduced CytcO are shown. (A) Wild-type CytcO, (B) Ser^I299Glu mutant CytcO, (C) Ser^I299Gly CytcO, (D) Glu^II101Ala CytcO, and (E) Glu^II101Asp CytcO. Experimental conditions: 0.1 M Ches pH 9.5. For the detergent-solubilized enzyme 0.05% DDM was used, T ≅ 22 C. The traces have been normalized such that the total absorbance changes are equal (ca. 1 μM reacting CytcO).

Table 1.

Time constants of the PCET for the wild-type and mutant CytcOs in detergent solution and reconstituted into lipid vesicles measured in 0.1 M Ches at pH 9.5

	Detergent solution	Membrane
CytcO	Time constant (ms)		Ratio
Wild type	1.2	0.17	7
SerI299Gly	2.3	0.28	8
SerI299Glu	5.7	0.64	9
GluII101Ala	5.7	5.7	1
GluII101Asp	13	3.3	4

For the time constants errors are within 10% of the values.

First, we modified the Ser^I299 residue. It is buried within the CytcO and hydrogen-bonded to a water molecule that in turn is hydrogen-bonded to Lys^I362, which is a key element of the proton pathway (Fig. 1). Replacement of this residue by a Glu resulted in a decrease in the PCET rate constant by a factor of ∼5 to 175 s^-1. Upon incorporation of the Ser^I299Glu mutant CytcO into a membrane, the rate constant increased by a factor of ∼9 to 1,600 s^-1 (Fig. 2B); i.e., the acceleration in the rate was similar to that observed for the wild-type CytcO. A similar effect was observed when the Ser^I299 residue was replaced by Gly (Fig. 2C and Table 1). Thus, also when the structure of the K pathway was modified as a result of the mutations at Ser^I299, the effect of incorporation into the membrane was the same as on the wild-type CytcO, which indicates that the effect of membrane incorporation is not due to changes in the pathway structure.

When replacing the protonatable residue Glu^II101 (SU II), at the orifice of the proton pathway at the n-side surface, by an Ala, the proton release was slowed to 175 s^-1, i.e., by a factor of ∼5 (Fig. 2D). However, in this case incorporation of the Glu^II101Ala mutant CytcO into a membrane did not result in any acceleration of the PCET rate, and the rate constant was the same as that observed with detergent-solubilized CytcO (Fig. 2D).

We note that in all mutant CytcOs discussed above, in detergent solution, the PCET rates were significantly slowed compared to the wild-type CytcO. In the case of the Ser^I299Glu and the Glu^II101Ala mutant CytcOs, the PCET rate was slowed to the same extent [for both mutant CytcOs the rate constant was 175 s^-1 (τ = 5.7 ms) in detergent solution]. Yet, the effects of incorporation of these mutant CytcOs into a lipid membrane were different; in the “interior mutant” CytcO (Ser^I299Glu), the PCET was accelerated by about the same factor as with the wild-type CytcO, whereas in the Glu^II101Ala surface mutant there was no effect of membrane reconstitution. To test whether or not the increase in the PCET rate upon membrane reconstitution is linked to the connectivity to the membrane, we replaced the Glu^II101 residue by a protonatable (acidic) residue, an Asp. As with the Glu^II101Ala mutant CytcO, also the Glu^II101Asp mutation resulted in slower PCET; in detergent solution the rate constant was 77 s^-1 (Fig. 2E). However, in contrast to the Glu^II101Ala mutant CytcO, in this case reconstitution into vesicles resulted in an increase in the PCET rate by a factor of ∼4 (Fig. 2E); the rate constant was 300 s^-1.

The amplitudes of the PCET absorbance changes at 598 nm were the same, within ∼15%, for the wild type and all mutant CytcOs in detergent solution and reconstituted in membranes (Fig. 2), which indicates that the extent of proton release was the same.

Discussion

In the present study we investigated the effect of the membrane on a specific PCET, where the proton is transferred through the K pathway. During CytcO turnover, this proton transfer is linked to electron transfer into the catalytic site and immediately precedes binding of O₂ to the reduced heme a₃. The data show that the PCET accelerates upon incorporation of the wild-type CytcO into a membrane, and quantitatively the same effect was observed in mutant CytcOs, Ser^I299Gly and Ser^I299Glu, in which the structure of the K pathway is modified. The Ser^I299 residue itself is not part of the K pathway, but it interacts with Lys^I362 via a water molecule, which is a key element of the pathway. Results from earlier studies showed that any modification of the Lys^I362 residue results in dramatically slowed or completely impaired PCET. The slowed PCET as a result of mutations at Ser^I299 is presumably an indirect effect of structural changes around the Lys^I362 residue such that mutations at Ser^I299 modify the proton-transfer conductivity through the K pathway via Lys^I362 (25). The results from this study show that the acceleration in the PCET rate upon incorporation of the Ser^I299Glu/Gly mutant CytcOs into a membrane was approximately the same as with the wild-type CytcO even though in the mutant CytcOs the PCET was slowed. In other words, even though the structure of the K pathway was modified, the proton-transfer rate changed by about the same factor as with the wild-type CytcO upon reconstitution of the CytcO in a membrane. These results suggest that the effect of reconstitution of the CytcO into the membrane is not due to changes in the K-pathway structure (i.e., conductivity).

As seen in Table 1, in both the Ser^I299Glu and the Glu^II101Ala mutant CytcOs, the PCET rate was slowed to the same extent in detergent solution; the rate in detergent solution was ∼5 times slower than in the wild-type CytcO. However, although in the former case the rate was significantly accelerated upon incorporation of the CytcO into a membrane [by a factor of ∼9 to 1,600 s^-1 (τ ≅ 640 μs)], no effect of membrane incorporation was observed in the latter case. Taken together, the data discussed above indicate that the effect of membrane incorporation is due to interactions of the membrane with the proton-pathway orifice rather than due to effects on the pathway structure. This conclusion is further supported by the data with the Glu^II101Asp mutant CytcO. In this case the PCET rate was slowed as a result of the mutation, but the rate increased significantly upon incorporation of the mutant CytcO into a membrane. We speculate that, in the Glu^II101Ala mutant, introduction of the nonprotonatable residue results in loss of the connectivity between the proton pathway and the membrane surface and as a consequence proton uptake/release from/to solution proceeds independently of the membrane. Introduction of the Asp residue reestablished this contact.

In this context a question that arises is why the PCET is slower with the Glu^II101Asp than with the Glu^II101Ala mutant CytcO in detergent solution. The surface-exposed Glu^II101 residue is located at the entrance of the K pathway and presumably establishes a protonic contact between the interior of the pathway and the protein surface. With the Ala residue at the position of the Glu^II101, there is room for a water molecule replacing the position of the Glu, which would facilitate proton transfer into the pathway (35), but with less efficiency (36, 37). A similar situation was observed previously for proton uptake to the Q_B site in photosynthetic reaction centers (38). Similarly, in the P. denitrificans CytcO water molecules were suggested to act as the K-pathway entry point instead of the Glu (39). With the Glu^II101Asp mutant CytcO the proton contact is reestablished. However, the Asp side chain is shorter than that of Glu, which would increase the distance between the proton donor and acceptor within the pathway thereby slowing the overall PCET in this mutant CytcO.

The amplitudes of the PCET absorbance changes were approximately the same for CytcO in detergent solution and reconstituted into membranes (at a fixed pH), which indicates that the extent of proton release was the same. In other words, introduction of the membrane had only an effect on the kinetics of proton transfer and not the pK_a of the internal proton donor/acceptor (see Introduction). This finding is consistent with a mechanism where the membrane changes only the barriers for proton transfer and not the driving force (40).

Results from earlier studies have shown that the rate of the PCET in detergent solution is pH dependent with a maximum value of ∼2,000 s^-1 (τ = 500 μs) at low pH (< 7) (18). This behavior was attributed to a titration of an internal proton donor/acceptor within the K pathway (18, 41). Because this maximum rate is slower than that observed with the membrane-reconstituted wild-type CytcO at pH 9.5, the effect of the membrane cannot be a simple pK_a shift of this proton donor.

Results from earlier studies have shown that the PCET rate is independent of the composition of the buffer or its concentration (17, 21), which indicates that buffer molecules do not act as proton donors or acceptors for the protons released through the K pathway. As the observed PCET rate represents the sum of the release and uptake rates, it is not straightforward to calculate a second-order proton-uptake rate constant.

The Glu^II101 residue carries a negative charge at neutral pH (42). Even if in the P. denitrificans CytcO strong electrostatic interactions were found between the residue and the catalytic site 25 Å away (43), the residue is located near the protein surface and it is expected to be fully deprotonated at pH 9.5 (36). As seen in Fig. 1, the entry point of the proton pathway around Glu^II101 is located near the protein-membrane interface, and the distance between the residue and the membrane edge is estimated to be < 6 Å. At this distance the Coulomb cages of the Glu and the lipid phosphate groups merge, which allows efficient proton transfer between the two (44, 45). This structural composition would increase the probability of proton transfer from the Glu to the phosphate group, which is presumably faster than proton release from the Glu to solution (46). This effect would explain why introduction of the membrane accelerates proton uptake and release. The observation that upon removing Glu^II101 (Glu^II101Ala mutation) the PCET rate is the same without and with the membrane indicates that this single mutation disconnects the proton pathway from the membrane. The lack of alternative routes for proton transfer between the CytcO and membrane surface is evident from an inspection of the R. sphaeroides CytcO crystal structure (13, 14), which shows that there are no other carboxylic acids within 15 Å from Glu^II101.

The data from the present study are consistent with earlier results from studies using fluorescence correlation spectroscopy, which showed that the membrane acts to accelerate protonation of a surface-bound pH-sensitive fluorescent dye (47; for a theoretical analysis, see 48). In other words, proton transfer along the membrane surface is faster than proton exchange between a surface-bound protonatable group and solution. This effect was shown to display a pH dependence such that it was most pronounced at pH > 8 (34). Furthermore, we showed recently that incorporation of CytcO into a membrane results in acceleration of the protonation rate of a surface-exposed pH-sensitive fluorescent dye at the n side of the membrane (49). Also data from earlier studies show that proton migration along protein-membrane surfaces may be faster than proton exchange between surface residues and the bulk solution (50), which involves specific surface residues (51, 52).

In conclusion, the data from this study show that the membrane environment modulates a proton transfer that is part of the catalytic reaction. The modulation is not an effect of structural changes of the protein, but rather the membrane is a functionally active component of the membrane transporter-membrane system. Because the same structural components are found at the surfaces of many membrane-bound transporters (4), the mechanism can be generalized to the entire energy-converting system as well as membrane-bound enzymes in which proton uptake or release is part of the reaction cycle. Under the nonequilibrium conditions in a living cell, this structural-functional design would facilitate direct proton exchange between proton pumps and, for example, ATP synthases thereby increasing the energy-conversion efficiency of the cell.

Materials and Methods

Mutagenesis, Growth, and Purification.

Site-directed mutagenesis was performed as described (53). R. sphaeroides was grown in Sistrom medium and the His-tagged protein was purified on a Ni column as described (54). After elution, the buffer was exchanged to 0.1 M Hepes pH 7.4, 0.1% dodecyl β-d-maltoside (DDM) before freezing in liquid nitrogen. The enzyme was stored at -80 °C until use.

Reconstitution of CytcO into Small Unilamellar Vesicles.

Incorporation of CytcO was performed as reported earlier (55). Briefly, purified soybean lipids (L-α-phosphatidylcholine from soybean, Type II-S, containing 14–23% phosphatidylcholine, Sigma-Aldrich) were dissolved at a concentration of 40 mg/mL in 0.1 M 2-(N-cyclohexylamino)ethanesulfonic acid (Ches) pH 7.4, 100 μM EDTA, and 2% cholate. The lipid solution was sonicated (Ultrasonic Processor XL2020, Heat Systems) in 30-s-on/30-s-off cycles, at 20–30% of the maximum power, until it was clear and thereafter centrifuged to remove large particles, before mixing 1∶1 (vol∶vol) with the enzyme solution containing 4–5 μM CytcO, 0.1 M Ches pH 7.4, 100 μM EDTA, and 4% cholate. Detergents were removed using the biobeads method as described in ref. 55. The diameter of the vesicles prepared using this method was 25 ± 5 nm, determined using dynamic light scattering (Nano Z, Malvern Instruments).

Preparation of the Two-Electron Reduced State.

Detergent-solubilized CytcO was diluted to 2 μM in 0.1 M Ches pH 9.5, 0.05% DDM, 100 μM EDTA. The sample containing CytcO reconstituted in vesicles was diluted to ∼1 μM CytcO and the pH was adjusted to 9.5. The samples were transferred to a modified anaerobic Thunberg cuvette. The atmosphere was exchanged on a vacuum line first to N₂ and thereafter to CO after which the sample was incubated for approximately 1–2 h. During that time, CO donates two electrons at a time to CytcO (CO + H₂O → CO₂ + 2H⁺ + 2e^-), which results in formation of the “mixed-valence state” in which the catalytic site heme a₃/Cu_B is reduced with CO bound to the heme, while heme a and Cu_A remain oxidized. The progress of the reaction was monitored on a spectrophotometer (Cary 400 Bio UV-vis, Varian) and formation of the mixed-valence-CO complex was verified from the absorption spectrum. If necessary, the samples were titrated with ferricyanide to reoxidize a small fraction of reduced heme a.

Flash Photolysis of CO from the Two-Electron Reduced CytcO.

The CO ligand was dissociated from the reduced heme a₃ by a laser flash (8 ns at 532 nm, Brilliant B, Quantel, 200 mJ). The reaction after CO dissociation was monitored at different single wavelengths using time-resolved optical absorption spectroscopy (for a general description of the setup, see ref. 25). The CO dissociation and recombination, and the following electron transfer from heme a₃ to heme a, were monitored at 445 nm. The PCET at a millisecond time scale was monitored at 598 nm, which is an isosbestic point for the CO-recombination reaction. Typically, for CytcO in detergent solution or reconstituted in lipid vesicles 25 or 75 signals, respectively, were averaged. The 598-nm traces were normalized to the absorbance changes associated with CO dissociation, measured at 445 nm.