|Home | About | Journals | Submit | Contact Us | Français|
The interaction of lipids with subunit c from F1F0 ATP synthase is studied by biophysical methods. Subunit c from both Escherichia coli and Streptococcus pneumoniae interacts and copurifies with cardiolipin. Solid state NMR data on oligomeric rings of F0 show that the cardiolipin interacts with the c subunit in membrane bilayers. These studies offer strong support for the hypothesis that F0 has specific interactions with cardiolipin.
Biological membranes, which play many key roles in cellular physiology, are often depicted as inert matrices anchoring essential proteins, for example as described in the fluid mosaic model . However, there is a growing variety of evidence that the role of lipids goes beyond the one of an inert support. Arguably, tight interactions between membrane-embedded proteins and the phospholipids are implicitly required to maintain chemical potential gradients and keep the bilayer effectively sealed. Lipids have also been observed as structural members in membrane complexes including ion channels or membrane receptors [2–5]. The membrane-embedded F1F0-ATP synthase, which produces the majority of the ATP used by the cell, has been suggested to also bind lipids through its membrane-embedded portion F0 [6–9]. The question of a preferential interaction with one particular type of lipid remains however up for discussions.
Several aspects of lipid–protein interactions have been probed in the past. The localization of the protein in the bilayer , the electrostatic component of the interaction , and the role of lipids in protein oligomerization  have been addressed by fluorescence correlation spectroscopy. Insights about the immersion of a protein in the bilayer can be probed by the combine use of protein labels and EPR spectroscopy . Oriented solid-state NMR, with in particular the PISEMA experiment , is another powerful tool to characterize the immersion and tilt angle of a membrane protein inside the bilayer [15–18]. Magic-angle spinning (MAS) NMR offers the possibility to design an unlimited number of experiments [19,20]. The location of membrane-embedded systems can be assessed by measuring the exposure of the protein to paramagnetic labels [21,22]. Other experiments monitoring the efficiency of 1H spin-diffusion between the lipids and the proteins, or water and the protein, have been successfully used to probe respectively the membrane-embedded or solvent-exposed portions of the protein [23–25]. The NMR-sensitive 31P and 2H nuclei are very accessible probes of the physical state of the membrane, allowing access to the effect of the protein on the phase and the dynamics of the lipids, and reciprocally [26–32]. These methods have been used to study several systems of highly relevant biological interest. One can quote for example the mechanism by which a virus merges with cell membranes , the action of antimicrobial peptides , and the implication of lipids in the mechanism of membrane proteins, such as the regulator cardiac muscle Phospholamban [35,36], or the potassium channel KcsA [37,38]. Implication of these interactions in non-viral pathologies, such as Huntingtin disease  or blood clotting , has also been investigated. But the question of specificity, implying a reconnaissance between one specific type of lipids and one specific site of the protein, remains difficult to evidence.
Depending on the strength of their interaction with the protein and their consequent rate of exchange, lipids have been referred to in different terms. The bulk lipids exchange very rapidly, and interact non-specifically with the protein mostly through the physical properties of the membrane such as its fluidity, lateral pressure, or the charge of its surface. The shell of lipids at the surface of the protein is usually referred to as annular lipids. The exchange rate of these lipids with the bulk is generally around 1 to 2.107 s−1 at 37 °C, which is at least a factor of 4 slower than the rate of exchange between two lipids of the bulk (8.107 s−1) . These exchange rates are all fast on an NMR timescale (> 104 s−1). Other lipids can be found buried within a membrane protein, in deep grooves in the protein surface, between transmembrane α-helices, or at the protein–protein interface of an oligomer. These lipids are referred to as non-annular lipids by contrast to the one at the annular sites [42,43]. They establish much stronger and specific interactions with the protein, as shown by their significantly lower exchange rate. Their longer lifetime at the binding site can allow their identification by X-ray crystallography . Understanding lipid–protein interactions can have biological or medical implications. For example, non-annular lipids at the interface between transmembrane α-helices in helix bundles have been proposed to play a key role allowing relative motions of helices during protein function. Such motions can be modulated using hydrophobic inhibitors of Ca2+-ATPase, small molecules that bind to the lipid binding cavities such as t-butyl-hydroquinone, cyclopiazonic acid, and thapsigargin [45,46].
ATP synthase is a molecular complex containing several subunits divided between the soluble F1 portion and the membrane-embedded part F0. The soluble F1 part, where the enzymatic reaction occurs, has been well characterized. Far less is known about the transmembrane proton-pump F0. Solid-state NMR turns out to be a very informative technique for the study of this membrane protein, as shown by recent studies of the active site structure . Subunit c is a major constituent of F0 and takes the form of an oligomeric ring. In the case of Escherichia coli, the c ring is made of 10 protomers of 79 amino acids each. The phospholipid composition of E. coli is dominated by phosphatidylethanolamine (69%), phosphatidylglycerol (19%), cardiolipin (CL) (6.5%) and other minor species such as phosphatidylserine and phosphatidic acid . Streptococcus pneumoniae phospholipid composition was found to be dominated by phosphatidylglycerol and cardiolipin in a ratio close to 1:1 . CL has a non-canonical structure: it consists of two phosphatidylglycerols connected by a glycerol moiety. It was shown to interact with or be essential for the functions of several mitochondrial proteins, including complexes involved in oxidative phosphorylation . Complex I (NADH:ubiquinone oxidoreductase) and Complex III (ubihydroquinone:cytochrome c oxidoreductase) are inactive in the absence of CL [51,52]. The activity of ATP synthase (complex V) also depends on the presence of CL . The existence of an interaction between CL and bovine ATP synthase, as well as Thermus thermophilus ATP synthase, was identified by solution NMR and mass spectrometry [9,8]. Here we study the c ring of E. coli and S. pneumoniae ATP synthases and their interactions with cardiolipin by MAS-NMR.
A similar protocol was used for the preparation of E. coli and S. pneumoniae subunit c. The subunit c gene (atpE) cloned into a pET-17b vector was overexpressed in E. coli BL21(DE3) cells. After transformation, the cells were inoculated into 5 mL LB medium fractions containing 100 mg/L of ampicillin for over-night pre-culture. The cells were then grown in 1 L LB medium fractions containing 100 mg/L of ampicillin, at 37 °C with shaking at 250 rpm, until the OD600 (optical density at 600 nm) reached 0.8–1. The cells were pelleted and washed in P minimal medium , and then transferred into 250 mL fractions of P minimal medium containing 100 mg/L of ampicillin and enriched with 3 g/L 15NH4Cl and 4 g/L U-[13C]glucose. After 30 min of growth at 37 °C with shaking at 250 rpm, the expression of subunit c was induced with 1 mM IPTG and continued for 18 h. The cells were then harvested and suspended in a 50 mM NH4Ac buffer at pH 7 before extraction and purification.
The protein was extracted and purified thanks to organic solvent . Six volumes (relative to the cell suspension) of 1:1 CHCl3:MeOH were added, the cells were lysed by grinding with a blender for a few minutes, and then incubated at 4 °C for 2 h. The extract was then centrifuged and filtered to remove cell debris. Water and chloroform were then added to adjust the final CHCl3:MeOH:H2O ratio to 8:4:3. The aqueous (H2O/MeOH) and organic (CHCl3) phases were separated by centrifugation, the upper aqueous phase was removed by aspiration and the surface of the lower organic phase was washed several times with 3:47:48 CHCl3:MeOH:H2O fractions. The organic phase was then collected, one volume of HCl3 was added and MeOH was adjusted to keep the protein from precipitating. This solution was then rota-evaporated to near dryness and suspended in a 5 mL fraction of 2:1 CHCl3:MeOH. Subunit c was then precipitated by addition of 8 volumes of chilled ether, and left to precipitate at −20 °C for 48 h. The ether suspension was centrifuged at −4 °C and the resulting pellets were dried under nitrogen stream before being dissolved in a minimal amount of 2:1 CHCl3:MeOH. The solution was then loaded on a CM52 cation-exchange column in the case of E. coli subunit c and a DE52 column in the case of S. pneumoniae subunit c. The column was washed with loading buffer and 1:1 CHCl3:MeOH and subunit c was then eluted with 5:5:1 CHCl3:MeOH:H2O fractions. The yield of protein was about 50 mg per liter minimum media.
Only E. coli subunit c was reconstituted in lipid bilayers. The subunit c ring was reconstituted in POPC liposomes with a protein-to-lipid ratio of 1:1 (w:w). 10 mg of protein and 100 mg of β-octylpyranoside detergent were mixed with 10 mg of POPC in CHCl3. The solvent was evaporated under nitrogen to form a film of mixed protein-detergent-lipids. The dried film was then dissolved in aqueous buffer (20 mM Tris–HCl, 10 µM DTT, 20% glycerol, 0.03% NaN3, pH 8) thanks to 10 cycles of 20 min of freeze–thaw cycles at 40 °C with sonication and − 20 °C. The solution was then diluted to reach the CMC of the detergent and dialyzed against 8 L of buffer (10 mM HEPES, 10% glycerol, 0.1 mM TCEP, 0.2% NaN3, pH 5) at room temperature for 48 h. The precipitate was then collected by centrifugation and packed into Bruker 3.2 mm rotors.
The solid-state experiments were carried out on a Bruker Avance 900 MHz spectrometer using a 3.2 mm triple tuned (1H, 13C, 15N) E-Free CP-MAS probe with a 20 kHz MAS frequency. The ordered temperature was set to 260 K. The DARR experiment was acquired with 100 ms mixing times . Magnetization transfer between 1H and 13C nuclei was performed with an echo of 60 ms during the INEPT experiment . The signal was acquired on the 13C channel under Spinal-64 heteronuclear decoupling  with a r.f. field strength of 90 kHz. Quadrature detection was obtained with TPPI. The maximum acquisition time was set to 7 ms in t1 and 9 ms in t2 for the DARR experiment. The maximum acquisition time was set to 12 ms in t1 and 9 ms in t2 for the INEPT experiment. Data were processed using zero-filling up to 2048 points in t1, and 4096 points in t2, with a square cosine filter with a baseline correction in both dimensions. Processing was performed using the software Topspin 3.1. Referencing was made with respect to DSS, based on an external reference of adamantane.
Lipids were isolated from the purified protein samples by lyophilization followed by dissolution in methanol. Contrary to phospholipids, subunit c is indeed not soluble in methanol. The isolated lipids were then analyzed by solution NMR and mass spectrometry.
The solution experiments were carried out on a Varian Inova 600 MHz spectrometer using a CryoHCN probe. All HSQC were acquired with a maximum acquisition time of 12 ms in t1 and 140 ms in t2. Data were processed with a square cosine filter with a baseline correction in both dimensions. Processing was performed using the software Topspin 3.1. Referencing was made with respect to DSS, assuming a difference between TMS and DMS of 1.9 ppm.
The lipids isolated from S. pneumoniae subunit c sample were analyzed by ESI–MS on a Thermo Finnigan LTQ mass spectrometer using flow injection analysis and negative ion mode. The lipid extract (1.4 to 14 µM) in chloroform was re-suspended in 50% methanol at a ratio of 1:5 and 10 µL was injected into a continuous flow of the same solvent (50% methanol) at 20 µL/min. The ESI source spray voltage was at 5 kV and the heated capillary at 275 °C. ESI–MS data were collected at profile mode, 2 microscans, 100 ms maximum injection time and from 200 or 600–1800 m/z.
The spectra of the experiments recorded at a proton field of 900 MHz on the uniformly enriched E. coli subunit c show some correlation peaks which don't belong to the protein and that we assign to lipids. In the high resolution magic angle spinning 13C–13C spectra of uniformly enriched c subunit reconsitituted in lipid bilayers, some correlation peaks can be unambiguously assigned to cardiolipin. This observation was made primarily based on correlations in a 13C–13C correlation DARR  spectrum (Fig. 1) recorded at an effective temperature of 280 K, a temperature for which the POPC bilayer is expected to be in a liquid disordered Lα phase. In this spectrum, the region between 64 and 74 ppm clearly shows several correlation peaks: an isolated cross peak between 69.3 ppm and 72.5 ppm, and three partially resolved cross peaks between 65.9 ppm and 73.4 ppm (Table 1). The chemical shifts of these peaks were respectively assigned to the central and peripheral glycerol groups of CL (Table 1) . Based on the strength of the cross peaks, it is very likely that this signal arises from a molecule of uniformly and highly 13C-isotopically enriched CL from the E. coli growth. No other strong cross-peaks were observed for these samples in the chemical shift range corresponding to the glycerol functionalities of phospholipids, suggesting that CL is the only endogenous lipid present in large (stoichiometric or greater) amounts.
The specificity of the interaction between subunit c and cardiolipin was verified on two levels: the presence of cardiolipin was confirmed and the absence of other lipids was proved.
The assignment of CL was confirmed by 1H–13C correlation NMR spectroscopy both in the solid-state and in solution. In the solid-state, we recorded an INEPT spectrum on the sample of E. coli subunit c reconstituted in POPC (Fig. 2). In contrast to the DARR spectra described above, INEPT spectra are based on long through-bond 1H–13C magnetization transfers and show enhanced signals for most mobile nuclei with longer transverse coherence life times. The INEPT spectrum of the c subunit preparations at 280 K displays 1H and 13C chemical shift correlations assignable to the lipids, whereas the signals from the less mobile protein are not visible. The strongest signals are from natural abundance 13C nuclei in the POPC molecules from the reconstituted membrane. Weaker peaks, whose 13C chemical shifts correspond to the ones assigned to CL on the DARR spectrum, are also clearly visible. The 1H chemical shifts and the additional 13C shifts are in good agreement with those reported for CL previously (Table 1). The presence of the CL in INEPT spectrum suggests that some of the CL molecules are in rapid conformational exchange or rotate rapidly on the microsecond timescale. We assign these peaks to molecules of CL which exchanged with the bulk, or to some residual mobility in the bound lipids.
The assignment of CL and the absence of other lipids were further confirmed by solution NMR on isolated lipids samples (Fig. 3). 13C–1H HSQC experiments were recorded on commercial samples of POPC (Fig. 3a), POPG (Fig. 3b), CL (Fig. 3c) and compared to samples of lipids isolated from S. pneumoniae subunit c (Fig. 3d) and E. coli subunit c (Fig. 3e). Even if these spectra were proton-detected, they are displayed in a way allowing a direct comparison with the solid-state INEPT spectrum of Fig. 2. A close-up view of the glycerol region (Fig. 3f) shows a perfect superposition of the E. coli and S. pneumoniae lipids spectra with the commercial CL spectrum. This observation further confirms the assignment of the endogenous lipids to CL. The glycerol region displayed in Fig. 3f also shows that POPC and POPG are absent from the natural lipids spectra. In the natural lipids spectra, the absence of peaks at 74–4 ppm, 65–3.9 ppm and 69–3.9 ppm, 62–4.2 ppm, 56–3.4 ppm shows the absence of POPG and POPC respectively. This confirms the specificity of the subunit c interaction with cardiolipin.
The nature of the endogeneous lipids was further investigated by mass spectrometry. The ESI–MS spectra of a sample of lipids isolated from S. pneumoniae subunit c are shown in Fig. 4. The general pattern is very similar to the one found in the literature for cardiolipin . The most intense peak at m/z 1428 can be assigned to the [M − H]− species. It brings further information regarding the dominant lipid hydrophobic chain since 1428 corresponds exactly to the mass of a (16:0/16:0)(16:0/16:0)-CL enriched in 13C. The breadth of the peak is attributed to partial isotope labeling and distribution of the lipid chain length. The second major peak at m/z 724 is assigned to the [M − 2H]2− species. The loss of fatty acyl ketene as shown in Fig. 4 leads to singly and doubly-charged fragment ions at m/z 1176 (1428–253) and m/z 588 (1428–253)/2 respectively. This overall aspect of the MS spectrum confirms the strong and specific interaction of ATP synthase subunit c with cardiolipin and brings additional insight on the nature of the dominant hydrophobic chain.
3C–13C DARR spectra with very long mixing times can be used to detect long-distance correlations peaks (up to 6–7 Å ). DARR spectra of 350 and 450 ms mixing times recorded on the E. coli subunit c sample reveal interesting potential lipid–protein contacts (Fig. 5). Specifically peaks at 23.9 ppm, 35.0 ppm and 60.3 ppm correlate with the peak at 72.5 ppm uniquely identified as C′2 nuclei of the central glycerol of CL. These correlations may support a recent hypothesis of an interaction between CL and Lys34 [61,62], since the chemical shifts (60.3 ppm, 35.0 ppm and 23.9 ppm) are consistent with possible assignment of Lys Ca, Cd and Cg and Lys34 is the only Lys present in the E. coli subunit c sequence. This last result should be taken with high precaution though, since the 72.5–60.3 ppm correlation is very weak and the 35.0 ppm peak is ambiguously assigned due to similar chemical shifts of lipids hydrophobic chains.
Even if the exact localization of the interaction is uncertain, two criteria indicate that the CL is “non-annular” or bound to the protein. First, we clearly demonstrated that they are endogenous molecules of CL that co-purified with the c subunit protein. This in turn implies that some CL in the E. coli cells were bound tightly enough to subunit c to resist the chloroform extraction used during the preparation of the sample. During this protocol, the c protomer is first isolated and the ring is then reconstituted in detergents. These lipids are therefore probably not simply trapped inside the ring, but are likely involved in a strong interaction with the c protomer. It also implies that the exchange rate of these lipids is slow on an NMR time scale. We conclude that subunit c of ATP synthase contains some non-annular binding sites that are specific for CL. As previously mentioned, the existence of an interaction between the protein and this lipid has recently been suggested by Walker . His hypothesis involves a methylated K34. Our data also are suggestive of the structural and functional involvement of this residue, perhaps facilitating the motion of the c-ring necessary to the proton pump.
We have shown using solid-state NMR that the subunit c of ATP synthase is involved in a non-annular interaction with CL. This interaction was proved to be specific and conserved among several organisms (E. coli and S. pneumoniae). These observations overall emphasize the importance of structural studies in a bilayer environment as close as possible to its native one, preserving the activity of ATP synthase . It also opens the possibility to consider using hydrophobic inhibitors binding to annular lipids site of subunit c of ATP synthase for medical applications. Subunit c is notably known for being the drug target of major diseases such as tuberculosis.
We are deeply grateful to Boris Itin and Mikael Goger from the New York Structural Biological Center for the technical support, and to Jennifer Aguilan for performing the ESI–MS experiments. We acknowledge financial support from the European Molecular Biology Organization (EMBO ALTF 1452-2010) and the Fulbright Organization, NIH R01GM072085 and NIH R01GM088724.