Ecto-F₁-ATPase: A moonlighting protein complex and an unexpected apoA-I receptor

Abstract

Mitochondrial ATP synthase has been recently detected at the surface of different cell types, where it is a high affinity receptor for apoA-I, the major protein component in high density lipoproteins (HDL). Cell surface ATP synthase (namely ecto-F₁-ATPase) expression is related to different biological effects, such as regulation of HDL uptake by hepatocytes, endothelial cell proliferation or antitumor activity of Vγ9/Vδ2 T lymphocytes. This paper reviews the recently discovered functions and regulations of ecto-F₁-ATPase. Particularly, the role of the F₁-ATPase pathway(s) in HDL-cholesterol uptake and apoA-I-mediated endothelial protection suggests its potential importance in reverse cholesterol transport and its regulation might represent a potential therapeutic target for HDL-related therapy for cardiovascular diseases. Therefore, it is timely for us to better understand how this ecto-enzyme and downstream pathways are regulated and to develop pharmacologic interventions.

INTRODUCTION

F₁F_o ATP synthase, the terminal enzyme of the oxidative phosphorylation pathway, is a complex molecular motor responsible for the large majority of ATP synthesis in all living beings, except in archaea which use a related enzyme to produce ATP, the A₁A_o ATP synthase[1]. It is located in the inner membrane of mitochondria, in the thylakoid membrane of chloroplasts in plants, and in the plasma membrane (PM) of certain bacteria. ATP synthase is highly concentrated in inner membrane cristae, where it is part of the so-called “ATP synthasome”, in association with an inorganic phosphate (Pi) carrier as well as the adenine nucleotide translocase (ANT) which exchanges ADP and ATP[2]. Experimental evidence also suggests that ATP synthase is organized in dimers[3], or even oligomers[4]. This supramolecular organization is involved in the generation of inner membrane cristae and depends on the e and g subunits[5].

Despite a few differences in subunit composition, the general structure of ATP synthase is highly conserved throughout evolution. In eukaryotic cells, it is composed of 16 unique subunits (α, β and c being present in multiple copies in the whole complex). It comprises the activity-bearing subunits (three αβ heterodimers), a rotor (subunits γ, δ, ε and the c ring) and a stator (subunits a, e, f, g, A6L, b, F6, d and OSCP) which holds the αβ heterodimers. It is organized into two major domains: a soluble F₁ domain (α, β, γ, δ, ε) and a membrane-associated F_o domain (Figure 1). The F₁ domain bears the catalytic activity while the F_o domain is a proton channel. The biogenesis of ATP synthase follows a complex and timed mechanism[6] which involves at least five assembly factors: Atp10p, Atp11p, Atp12p, Atp22p and Fmc1p[7-10]. So far only the human orthologs of Atp11p and Atp12p have been identified[11].

Figure 1.

Mitochondrial ATP synthase. In eukaryotic cells, F₁F_O ATP synthase is a complex molecular motor composed of 16 unique subunits (α, β and c being present in multiple copies in the whole complex). It comprises the activity-bearing subunits (three αβ heterodimers), a rotor (in orange: subunits γ, δ, ε and the c ring) and a stator (in green: subunits a, e, f, g, A6L, b, F6, d and OSCP) which holds the αβ heterodimers. It is organized into two major domains: a soluble F₁ domain (α, β, γ, δ, ε) and a membrane-associated F_O domain. The F₁ domain bears the catalytic activity while the F_O domain is a proton channel. The F_O domain uses the proton gradient created by the four other complexes of the respiratory chain to drive the rotation of the central stalk, which will alternatively change the conformation of the three αβ heterodimers, leading to the synthesis of ATP from ADP and Pi. In the case of the eukaryotic ATP synthase, the translocation of 10 protons through the c ring will lead to a full turn, and the synthesis of 3 ATP molecules. In the absence of a proton gradient, or if the F₁ domain is isolated from the F_O domain, the enzyme behaves as an ATP hydrolase.

The F_o domain uses the proton gradient created by the four other complexes of the respiratory chain to drive the rotation of the central stalk, which will alternatively change the conformation of the three αβ heterodimers, leading to the synthesis of ATP from ADP and Pi. The rotation of ATP synthase was demonstrated by a remarkable experiment in which the αβ heterodimers were immobilized on a solid support while the γ chain was coupled to a fluorochrome-labeled actin filament, and visualized under a microscope after addition of ATP[12]. ATP synthesis and hydrolysis involve a binding-change mechanism[13,14] in which catalytic sites undergo conformational changes driven by the interactions with the γ chain. In the case of the eukaryotic ATP synthase, the translocation of 10 protons through the c ring will lead to a full turn, and the synthesis of 3 ATP molecules. However, the exact mechanism of catalysis is still debated, as experimental data support both a bi-site mechanism, in which one of the catalytic sites is empty at any given moment[15], or a tri-site mechanism where all sites are filled with nucleotides[16]. The F₁ domain also bears three non-catalytic sites which have been proposed to enhance ATP synthesis/hydrolysis upon binding of nucleotides[17].

In the absence of a proton gradient, or if the F₁ domain is isolated from the F_O domain, the enzyme behaves as an ATP hydrolase. This potential energy loss is tightly regulated by the inhibitor factor 1 (IF₁) protein when the proton gradient collapses, for instance in the case of hypoxia as the respiratory chain stops functioning. Under normal conditions, IF₁ exists as a homotetramer which turns into a homodimer when the pH in the matrix compartment decreases[18]. This allows it to tightly bind ATP synthase by interacting with both the αβ dimers and the central stalk, thereby blocking the rotation of the latter and preventing ATP hydrolysis[19]. This inhibitory mechanism has recently been shown to be highly dependent on the catalytic activity of the enzyme[20]. For extensive reviews on the structure and rotary mechanism of ATP synthase[21-23].

Given its essential role in ATP synthesis, defects in ATP synthase are involved in extremely severe and often lethal diseases because of the marked reduction in ATP production and concomitant accumulation of reactive oxygen species (ROS)[24,25]. Such pathologies are mostly linked to mutations in ATP synthase subunits, the most frequently described being mapped to the mitochondrial ATP6 gene (encoding for subunit a) which is responsible for the maternal-inherited NARP syndrome (Neurogenic muscle weakness, Ataxia, Retinitis Pigmentosa) or Leigh syndrome (severe and fatal encephalopathy), depending on the extent of heteroplasmic mutations. Defects in ATP synthase biogenesis, due to mutations in genes encoding its assembly factors for instance, have also been described[26]. Finally, abnormal accumulations of ATP synthase subunits have been found in various neurodegenerative diseases, such as subunit c accumulation in neuronal ceroid lipofuscinosis[27,28], or subunit α accumulation in Alzheimer’s disease[29].

Over the past few years, various reports by many independent groups have described the presence, at the cell surface of mammalian cells, of various subunits of ATP synthase, which will also be referred to as ecto-F₁-ATPase throughout this review. The exact structure of this ectopically-expressed ATP synthase and whether it is exactly the same enzyme as the mitochondrial complex have not been definitely determined. Nevertheless, a number of pieces of experimental evidence suggest that it is a closely related entity. We will review here the various reports describing the unexpected cell surface expression of ATP synthase and its possible functions.

IS ECTO-F₁-ATPASE THE SAME ENTITY AS MITOCHONDRIAL ATP SYNTHASE?

Over the past few years, a growing number of reports, including our work[30,31], have described the presence of ATP synthase subunits at the surface of a wide variety of tumor as well as normal cells by fluorocytometry.

While most of these reports focused on the α and β subunits against which monoclonal antibodies are available[30-32], a few have described the presence of other subunits at the cell surface by using in-house manufactured polyclonal reagents analyzed by cytometry or confocal microscopy. Vantourout et al[33] have shown that high cell surface expression of major histocompatibility complex class I (MHC-I) antigens interferes with the detection of the α and β subunits using some monoclonal reagents, probably through molecular hindrance. In a recent work, Mangiullo et al[34] described the presence of F_OI-PVP (b subunit), γ and OSCP subunits at the cell surface of rat hepatocytes. The inhibitory protein IF₁ has also been shown to be present at the cell surface[35]. Recently, d and OSCP subunits were shown to be expressed on the surface of osteosarcoma cells[36]. In a number of studies, biochemical approaches have been used to confirm the presence of F₁-ATPase/ATP synthase in plasma membranes. Bae et al[37] have shown via a proteomics approach that multiple ATP synthase subunits are found in the lipid rafts of rat hepatocytes and HepG₂ cells, a human hepatocarcinoma cell line. F₁-ATPase components have been coprecipitated with cell surface proteins from hemopoietic tumors by Vantourout et al[33]. One could argue that pure plasma membrane preparations are extremely difficult to obtain and easily contaminated by membranes originating from different sub-cellular compartments. A very low amount of mitochondrial material could explain the presence of ATP synthase subunits in membrane preparations. Nevertheless, in some works using biochemical approaches[33,34] the absence of other mitochondrial proteins in samples has been carefully checked, ruling out a contamination by mitochondrial material. The observation of multiple components of F₁-ATPase by cytometry, and the identification of some of them by mass spectrometry and peptide sequencing[30,37,38], make very unlikely the hypothesis that ecto-F₁-ATPase detection could be due to cross-reactivity with other proteins bearing sequence similarities.

The presence of subunits other than α and β is also not completely unexpected. Indeed, since both these subunits do not have any membrane-anchoring domain, the cell surface expression of other subunits, especially from the F_O domain, would be required to allow such a localization of the complex.

To summarize, while definitive proof may still be lacking, it seems now highly probable that ecto-F₁-ATPase is a cell surface complex identical or closely related to the mitochondrial ATP synthase. The wide range of cellular types that have been shown to express ecto-F₁-ATPase (Table 1) certainly suggest that this ectopic expression is more likely to be a ubiquitous phenomenon rather than restricted to a few cell lines. Among all samples that we have analyzed for ecto-F₁-ATPase expression using different approaches, so far only one, namely 721.221 (an EBV-immortalized B cell line), seemed to not express ecto-F₁-ATPase, or at extremely low levels that are not detectable[33,46].

Table 1.

Ecto-F₁-ATPase subunit detection in various cell types

Cells/tissues	Subunits	Methods	Ref.
K562, A549, Raji	β	B, FC, S	[38]
HUVEC	α, β, IF1	B, FC, CM	[35,39,40]
HepG2, IHH, Human hepatocytes	α, β	SPR, FC, CM	[30]
3T3-L1	α, β	B, CM	[41]
Jurkat	β	RP + MS	[42]
Rat hepatocytes	α, β, δ, γ, b, d, e, F6, OSCP	CM	[37]
HaCat, Human skin	β	RP + MS, CM	[43]
Daudi, RPMI-8226, U937, SK-NEP, G401, G402	α and/or β	FC	[31]
Awells, ST-Emo, HeLa, fibroblasts, Rma-S	α, β	B, CM, FC	[33]
Rat tonsils	β	PM	[44]
Neurons, neuroblastoma, astrocytoma	α	B, CM	[45]
Osteosarcoma	α, β, d	CM	[36]

Studies reporting a cell surface expression of ATP synthase subunits are summarized. The experimental approaches used in these studies are mentioned. B: Cell surface biotinylation; FC: Flow cytometry; S: Sequencing; CM: Confocal microscopy; SPR: Surface plasmon resonance (Biacore); RP + MS: Rafts purification and mass spectrometry; PM: Plasma membrane extraction. Cell lines/tissues are of human origin unless otherwise noted. K562: Erythroleukemia; A549: Lung adenocarcinoma; RPMI-8226: B myeloma; Raji, Daudi: Burkitt lymphomas; Jurkat: T lymphoma; HUVEC: Human umbilical vein endothelial cells; HepG2: Hepatocarcinoma; IHH: Immortalized human hepatocytes; 3T3-L1: Murine fibroblasts (differentiated into adipocytes); HaCat: Immortalized human keratinocytes; U937: Monocytic leukemia; SK-NEP, G401, G402: Renal cell carcinomas; Awells, ST-EMO: EBV-immortalized B cells (ST-EMO was derived from a major histocompatibility complex classI deficient patient); HeLa: Cervix carcinoma; Rma-S: Murine T lymphoma.

AN UNEXPECTED AND SO FAR UNEXPLAINED LOCALIZATION

While many proteins are now known to have multiple sub-cellular localizations, a cell surface expression of mitochondrial proteins is quite unusual and puzzling. To our knowledge, there is so far no experimental work precisely deciphering the mechanism used by ATP synthase to reach the plasma membrane (PM).

In our view, two main hypotheses are conceivable: (1) ATP synthase subunits are routed to the PM instead of the mitochondria. The concept of multiple localization of a given protein is quite established now and several mechanisms can be involved (see Karniely et al[47] for an extensive review). The first is that different mRNA isoforms of ATP synthase subunits could hold a cell-surface signal peptide rather than mitochondrial targeting signal. Bovine subunit c has been reported to have two isoforms (encoded by two different genes) with slightly different signal peptides[48]. However, to our knowledge, this is an isolated case so far, and there is no experimental evidence suggesting that these different peptides lead to different localizations. Another explanation is that the known signal peptides of ATP synthase subunits are ambiguous and could lead to a mitochondrial or cell surface expression. We have analyzed some of ATP synthase subunit leader peptides with the SignalP software[49] and some leader peptides of ATP synthase subunits may share some predicted similarities with PM-targeting sequences. Finally, some proteins use two targeting signals on the same polypeptide to reach different destinations, such as Catalase A in yeast[50]. This latter possibility seems unlikely since no PM signal can be identified in the sequences of ATP synthase subunits. Even if this first hypothesis cannot be definitely ruled out yet, it is in our opinion the less likely for a number of reasons. Firstly, it would imply that all subunits display ambiguous signals, or have different isoforms, and so far there is no formal proof of this. Secondly, two subunits (namely A6L and a) are encoded by the mitochondrial genome, and it seems unlikely that they would be targeted to the cell surface instead of staying in the mitochondria. Finally, the ATP synthase assembly factors would also have to be expressed elsewhere [such as the endoplasmic reticulum (ER)] to allow the expression of a complete and correctly assembled complex at the cell surface; and (2) The simpler and thus more tempting hypothesis is that once assembled into the mitochondria, ATP synthase then reaches the cell surface. As a direct rerouting of the whole complex via transporters is improbable in this case, given the size of ATP synthase (around 600 kDa and a dimension of 19 nm × 11 nm), it is suggested that the complex reaches the cell surface via vesicular transport or fusion of mitochondrial membranes with the PM.

In healthy cells, mitochondria form a dynamic network and constantly divide and fuse, a process which is believed to maintain the integrity and quality of the mitochondrial function and implicates several recently identified proteins such as mitofusins or Drp1[51,52]. A direct fusion of mitochondrial membranes with the PM, resulting in a cell surface expression of ATP synthase, seems difficult to imagine mostly because mitochondria have two membranes and ATP synthase is located in the inner one. Unless the fusion occurs between PM and mitochondrial contact sites, which are distinct structures playing an important role in the mitochondrial metabolism and apoptosis[53], this hypothesis is unlikely.

However, there are now several lines of evidence indicating that mitochondria communicate with other cellular compartments. A recent report by de Brito et al[54] has shown that mitofusin 2 is implicated in the interaction between ER and mitochondria, allowing the exchange of Ca²⁺ between those organelles. In another recent report, Neuspiel et al[55] identified a new outer membrane mitochondria-anchored protein ligase named MAPL (mitochondria-anchored protein ligase) and characterized the sorting of MAPL into mitochondria-derived vesicles which then fuse with peroxisomes. Interestingly, Neuspiel et al[55] have also identified distinct vesicles containing another mitochondrial protein, Tom20 (translocase of outer membrane 20, a member of the complex importing proteins encoded by the nuclear genome into mitochondria). These Tom20-containing vesicles do not fuse with peroxisomes and their fate is currently under investigation. In addition, Soltys et al[56] have shown that Hsp60, a mitochondrial chaperone, can be found in vesicles that are distinct from major organelles and may communicate with the PM. Interestingly, Hsp60 has also been shown to be expressed at the cell surface[57,58]. In both cases, these uncharacterized vesicles could thus be part of a shuttle system between mitochondria and PM, allowing the cell surface expression of selected mitochondrial proteins. In this context, Rab32 or endophilin B1, involved in mitochondrial dynamics and morphology, respectively[59,60], might also be involved in the biogenesis of mitochondria-derived vesicles and the trafficking towards the cell surface, raising the question of the existence of a new intracellular pathway from mitochondria to cell surface responsible for the targeting of the F₁-ATPase.

A recent report has shown that a deficiency in palmitoyl protein thioesterase 1 (Ppt1), responsible for the infantile neuronal ceroid lipofuscinosis syndrome, was linked with an increase in ecto-F₁-ATPase expression in neurons[61]. This suggests that Ppt1 might be involved in the regulation of the subcellular localization of ATP synthase. In addition, Schmidt et al[45] have immunoprecipitated F₁-ATPase α chain from the synaptic membranes of murine neural cells. Strikingly, they reported that this subunit was at least in part N-glycosylated and that this post-translational modification was abrogated by treating cells with the Golgi-disrupting drug brefeldin A. They also identified a unique N-glycosylation site which is absent in the human α-subunit sequence. This suggests that at least some components of the mitochondrial F₁-ATPase can transit through the Golgi apparatus, although the mechanism is still enigmatic.

Another question regarding the cell surface expression of ATP synthase is its sub-localization at the PM. Some of its subunits, as well as a few other mitochondrial proteins, have often been found in lipid rafts in proteomic studies[37,42]. However, a recent report by Zheng et al[62] suggested that mitochondrial proteins, including ATP synthase subunits, were the best contaminants of raft preparations, mostly based on the fact that they were resistant to cholesterol depletion by methyl-β-cyclodextrin, whereas raft-resident proteins usually are sensitive to this treatment. Detection of ecto-F₁-ATPase by confocal microscopy usually reveals a patchy pattern[30,33], which is reminiscent of several proteins residing in lipid rafts. We also have experimental evidence showing at least a partial co-localization of ecto-F₁-ATPase subunits with the raft markers cholera toxin (unpublished data). This may indicate that the presence of ecto-F₁-ATPase in lipid rafts is not exclusive. It may also vary with the cell line or conditions of culture. Similarly, its detection in isolated rafts may depend on raft purification procedures. Thus, more extensive studies are required to definitively answer this question.

A NEW LOCALIZATION FOR MANY NEW ROLES: ECTO-F₁-ATPASE AS A MOONLIGHTING PROTEIN

The term “moonlighting protein” was introduced in 1999 by Jeffery[63] to describe proteins that could perform several unrelated roles. Their different functions can depend on different localizations, expression in different cell lines, ligand concentration, or on many other mechanisms.

ATP synthase appears as a typical example of a moonlighting protein. Its α subunit, for instance, has been shown to be part of a complex importing cytosolic tRNAs into mitochondria in Leishmania[64]. Subunit F6 has also been shown to be present in plasma and involved in hypertension through its binding on ecto-F₁-ATPase expressed on endothelial cells[65]. It is also now clear that the whole complex expressed at the cell surface is implicated in many physiological roles, depending on the ligand and the cell type (Table 2). So far, the most documented roles of ecto-F₁-ATPase are its involvement in HDL endocytosis by hepatocytes, endothelial cell survival, and as an immunological ligand for a subset of non-conventional T lymphocytes (Figure 2).

Table 2.

Ligands and roles of Ecto-F₁-ATPase

Cells/tissues	Ligands	Proposed roles	Ref.
K562, A549, Raji	?	NK/LAK-mediated tumor cell lysis	[38]
HUVEC	Angiostatin HDL/apoA-I	Cell survival through ATP production	[40,60,66]
	FC6	Cell survival through purinergic receptor activation
		Control of blood pressure
Hepatocytes	HDL/apoA-I	HDL endocytosis	[30]
3T3-L1, HaCat	?	Cell survival	[41,43]
Tumor cells	Vγ9/Vδ2 TCR, apoA-I, MHC-I	Tumor cell recognition by Vγ9/Vδ2 T cells/presentation of phosphoantigens	[31,33,46,67]
Tonsils (mouse)	Enterostatin	Food intake regulation	[44]
Neurons	APP, Amyloid β-peptide	Synaptic plasticity regulation	[45]

Major roles of Ecto-F₁-ATPase described in the literature, as well as ligands involved, are summarized here. MHC-I: Major histocompatibility complex class I; NK: Natural killer; LAK: Lymphokine activated killer; HDL: High density lipoprotein; apoA-I: Apolipoprotein A-I; FC6: Coupling factor 6; ATP: Adenosine triphosphate; APP: Amyloid precursor protein. Cells/tissues are of human origin unless otherwise noted. K562: Erythroleukemia; A549: Lung adenocarcinoma; Raji: Burkitt’s lymphoma; HUVEC: Human umbilical vein endothelial cells; 3T3-L1: Murine fibroblasts (differentiated into adipocytes); HaCat: Immortalized human keratinocytes.

Figure 2.

Model of the leading hypotheses regarding the distinct cellular events mediated by ecto-F₁-ATPase. (1) Model of F₁-ATPase/P2Y₁₃-mediated high density lipoproteins (HDL) endocytosis by hepatocytes. ApoA-I binding to ecto-F₁-ATPase stimulates extracellular ATP hydrolysis, and the ADP generated selectively activates the nucleotide receptor P2Y₁₃ and subsequent RhoA/ROCK I signaling that controls holo-HDL particle endocytosis. Conversely IF₁ inhibitor which inhibits ecto-F₁-ATPase activity or adenylate kinase (AK) activity which consumes ADP generated upon activation of ecto-F₁-ATPase, downregulates holo-HDL particle endocytosis. Thus, HDL endocytosis depends on the balance between the AK activity that removes extracellular ADP and the ecto-F₁-ATPase that synthesizes ADP. Activation of ADP synthesis by apoA-I unbalances the pathway and increases HDL endocytosis when required; (2), (3): Model of F₁-ATPase-mediated endothelial cell protection. (2): Hydrolytic activity of ecto-F₁-ATPase expressed on endothelial cells can be increased by apoA-I, stimulating cell proliferation. Conversely, inhibition by IF₁, angiostatin increases apoptosis and blocks proliferation both at the basal level and after stimulation by apoA-I. The downstream receptors and intracellular signaling are still not completely characterized, but ADP-sensitive receptors could be involved; (3): The anti-angiogenic activity of angiostatin is attributed to its ability to inhibit ATP production by ecto-F₁-ATPase, thus eliminating a significant source of energy and reducing proliferation. It is also proposed that this inhibition prevents the extrusion of protons by ecto-F₁-ATPase, thus lowering intracellular pH and hampering cell survival; and (4) Model of F₁-ATPase-mediated tumor cell recognition by Vγ9/Vδ2 T lymphocytes. Ecto-F₁-ATPase is found at the surface of tumor cells and plays a role in the presentation of adenylated derivatives of small phosphoantigens, which are classical inducers of cell lysis by Vγ9/Vδ2 T lymphocytes. Recognition of ecto-F₁-ATPase by the γ9/δ2 TCR is stimulated by apoA-I and requires a low expression of class I HLA-antigens, a situation commonly found in tumor cells.

Hepatic HDL-cholesterol uptake

Our work on high density lipoproteins (HDL) has identified ecto-F₁-ATPase as the receptor of apolipoprotein A-I (apoA-I), the major protein component of HDL. Indeed, binding of HDL on hepatocyte membranes displayed two binding sites with high and low affinities[68]. A high-affinity receptor for lipid-free apoA-I was purified by Biacore and identified as the β-chain of ATP synthase. As binding of lipid-free apoA-I to ecto-F₁-ATPase activated endocytosis of whole and fully lipidated HDL by the low-affinity binding sites, we have hypothesized that enzymatic activity of ecto-F₁-ATPase might play a role in this stimulatory process. Indeed, direct addition of ADP increased HDL endocytosis. We showed that binding of apoA-I on ecto-F₁-ATPase increased its hydrolase activity and led to an enhanced HDL endocytosis. Later, we identified the purinergic P2Y₁₃ receptor as the target of ADP produced by ecto-F₁-ATPase[69] and we have shown recently that the small GTPase RhoA and its effector ROCK I were activated downstream of P2Y₁₃[70]. We could demonstrate that this F₁-ATPase/P2Y₁₃-mediated HDL endocytosis pathway is controlled through cell surface adenylate kinase (AK) activity which consumes ADP generated upon activation of ecto-F₁-ATPase by apoA-I and therefore constitutively inhibits P2Y₁₃ and subsequent hepatic HDL endocytosis[71]. In addition, niacin (vitamin B3), which is well-known to increase plasma HDL levels in humans, has recently been shown to inhibit cell surface expression of ecto-F₁-ATPase in hepatocytes[72]. This study thus suggests the implication of ecto-F₁-ATPase in HDL metabolism.

Our recent work has confirmed the role of the P2Y₁₃ ADP-receptor in HDL-mediated reverse cholesterol transport (RCT) in vivo[73]. We showed that P2Y₁₃-deficient mice exhibited a decrease in hepatic HDL cholesterol uptake, hepatic cholesterol content, and biliary cholesterol output, although their plasma HDL and other lipid levels were normal. These changes translated into a substantial decrease in the rate of macrophage-to-feces RCT. Therefore, hallmark features of RCT are impaired in P2Y₁₃-deficient mice. Furthermore, cangrelor, a partial agonist of P2Y₁₃, stimulated hepatic HDL uptake and biliary lipid secretions in normal mice and in mice with a targeted deletion of scavenger receptor class B type I (SR-BI) in liver [hypomSR-BI-knockout(liver)] but had no effect in P2Y₁₃ knockout mice, which indicate that the P2Y₁₃-mediated HDL uptake pathway is independent of SR-BI-mediated HDL selective cholesteryl ester uptake. Pharmacological activation of the P2Y₁₃ receptor might thus represent a new therapeutic approach to improve HDL-cholesterol removal and/or turnover, which could be an alternative way to prevent and treat atherosclerosis.

Endothelial cell survival

Ecto-F₁-ATPase has also been characterized by Moser et al[32] as the receptor of angiostatin, an inhibitor of angiogenesis originating from the proteolysis of plasminogen. Later work established that ecto-F₁-ATPase was able to synthesize ATP at the cell surface of endothelial cells. This ATP production would facilitate the proliferation of these cells, especially under hypoxic conditions which are known to impair cell survival through the inhibition of oxidative phosphorylation. The anti-angiogenic activity of angiostatin is attributed to its ability to inhibit ATP production by ecto-F₁-ATPase, thus eliminating a significant source of energy and reducing proliferation. It was also proposed that this inhibition prevents the extrusion of protons by ecto-F₁-ATPase, thus lowering intracellular pH and hampering cell survival[39,66].

It is well known that HDL and apoA-I have a protective effect against the development of endothelial dysfunction, one of the first events in the pathogenesis of atherosclerosis, by enhancing endothelial cell survival. The mechanisms mediating this atheroprotection are not well understood, but interactions between HDL and apoA-I with cell surface receptors have been proposed to be essential contributors. Considering that both HDL and apoA-I bind ecto-F₁-ATPase, we have evaluated the hypothesis that the protective effects of apoA-I on endothelial cells are dependent on ecto-F₁-ATPase activity. This hypothesis is also supported by the anti-angiogenic effect of angiostatin.

As shown for hepatocytes[30], the hydrolytic activity of ecto-F₁-ATPase expressed on endothelial cells can be increased by apoA-I, stimulating cell proliferation and inhibiting apoptosis in a dose-dependent manner[40]. Conversely, inhibition by IF₁, angiostatin or antibodies directed against the β-chain of ecto-F₁-ATPase increases apoptosis and blocks proliferation both at the basal level and after stimulation by apoA-I. These results are in favor of a direct effect of the ecto-F₁-ATPase activity on endothelial cell survival through extracellular ADP production. The downstream receptors and intracellular signaling are still not completely characterized, but ADP-sensitive receptors could be involved.

This pathway thus provides a new mechanism of action of apoA-I in endothelial protection distinct from other HDL-mediated protective effects previously described[74]. Thus, depending on the cell type, distinct pathways seem to be involved downstream of F₁-ATPase activation by apoA-I and induce distinct cellular events such as stimulation of HDL uptake by hepatocytes or endothelial cell survival.

Tumor cell recognition

Vγ9/Vδ2 T lymphocytes are a non-conventional subpopulation of cytotoxic T cells unique to humans and higher primates. These cells are known to participate in the immune response against various intracellular pathogens as well as tumor cells, through the recognition of small alkyl diphosphate molecules termed phosphoantigens[75,76]. Experimental evidence suggests that recognition of these antigens requires a cell-cell contact, and thus may depend on a cell surface presentation molecule expressed on target cells but distinct from classical antigen presentation molecules such as major histocompatibility antigens and CD1[77]. This is confirmed by recent data showing that the recognition of phosphoantigens by a soluble recombinant Vγ9/Vδ2 T cell receptor (TCR) on the cell surface is dependent on a trypsin-sensitive structure[78]. We have initially identified apoA-I as an enhancer of the Vγ9/Vδ2 T cell response, and shown that these cells can actually recognize F₁-ATPase on the surface of tumor cells through their TCR[31]. We have also shown that ecto-F₁-ATPase interacts with MHC-I molecules on the cell surface of various cell lines, leading to a masking of epitopes which prevents the detection of ecto-F₁-ATPase by commercially available antibodies[33]. This suggests that cell surface expression of ATP synthase cannot be formally ruled out simply because it is not detected by flow cytometry or confocal microscopy. It is also of particular interest for the physiology of Vγ9/Vδ2 cells because, like natural killer cells, they express inhibitory receptors which bind MHC antigens and modulate TCR signaling and lymphocyte activation[79].

Our latest work has focused on the study of a novel ATP analog, ApppI [triphosphoric acid 1-adenosin-5’-yl ester 3-(3-methylbut-3-enyl) ester] in which the γ-phosphate of ATP is esterified with isopentenol. This metabolite has recently been identified in cells treated with aminobisphosphonate drugs which are known to potently sensitize cells to Vγ9/Vδ2 T cell cytotoxicity. As it is an ATP analog and has a structure reminiscent of nucleotidic phosphoantigens naturally produced by some bacteria, we thus used this molecule to explore the potential role of ecto-F₁-ATPase in phosphoantigen presentation to Vγ9/Vδ2 T cells. We found that classical targets of Vγ9/Vδ2 T cells can spontaneously produce ApppI[67]. Furthermore, it stably binds to purified ATP synthase and ApppI-loaded ATP-synthase efficiently activates Vγ9/Vδ2 T cells in vitro. Our current model proposes that phosphoantigens produced by tumor cells or intracellular pathogens are converted to nucleotidic derivatives, loaded onto ATP synthase, and presented to Vγ9/Vδ2 T cells when the complex is exported to the cell surface[46].

As ATP synthase is present at the cell surface of a number of normal cells, the potential self-reactivity of Vγ9/Vδ2 T cells is probably tightly controlled by the interaction between ecto-F₁-ATPase and MHC-I, as well as by phosphoantigen production by the target cell. Moreover, these observations suggest another beneficial role for apoA-I; as an immunomodulator in anti-tumor responses.

Since ecto-F₁-ATPase has already been detected in a wide variety of cells, with many different ligands, there is no doubt that it is probably implicated in many other cellular processes. In many instances, these roles seem to rely on the enzymatic activity of ATP synthase. While its hydrolase activity is undoubtedly possible since it does not need a proton flux nor requires a fully assembled F₁F_O complex, there are still conflicting reports regarding the ability of the cell surface complex to synthesize ATP.

TO SYNTHESIZE OR NOT TO SYNTHESIZE, THAT IS THE QUESTION

The first work showing that ecto-F₁-ATPase can synthesize ATP from ADP and Pi was reported in endothelial cells[39] and showed that angiostatin was able to inhibit both the synthase and hydrolase activities of ATP synthase. This was shown by bioluminescence measurement of ATP and a radioactive thin layer chromatography (TLC) assay using [³H]ADP and ³²Pi. In later work, the authors have shown that ATP production was enhanced when extracellular pH was low, hypothesizing that concomitantly with ATP synthesis, ecto-F₁-ATPase would extrude protons from the cytoplasm. It was also proposed that this process improves the resistance of tumor and endothelial cells to low pH. Similarly, it would also improve cell survival in hypoxic conditions, a hallmark of tumoral micro-environment. Since then, a number of independent studies have reported a synthase activity in many different cell types, including keratinocytes, adipocytes, or hepatocytes, among others[34,43,41]. Most of these works have used a bioluminescence assay to measure the production of ATP after addition of ADP in the culture media and underlined the importance of ecto-F₁-ATPase in this production through the use of oligomycin, a well known inhibitor of mitochondrial ATP synthase.

However, whether or not ecto-F₁-ATPase can synthesize ATP is still debated, as there are several conflicting reports. Several studies by our group and others seem to completely rule out a potential contribution of ecto-F₁-ATPase to extracellular conversion of ADP to ATP. Yegutkin et al[80] have shown that ATP production after addition of ADP to endothelial cells was not increased in the presence of Pi, which should be the case if ecto-F₁-ATPase was able to synthesize ATP. They concluded that ADP phosphorylation was mostly due to cell surface adenylate kinase (AK, which produces ATP and AMP from two ADP molecules) and/or nucleotide diphosphokinase (NDPK, which transfers terminal phosphate from a nucleotide triphosphate to a nucleotide diphosphate) enzymatic activities. This is in agreement with our results in hepatocytes as well as in endothelial cells. Indeed, we have observed that addition of [³H]ADP and ³²Pi to the cell medium does not lead to production of dual-labeled ATP, but only [³H]ATP[30,40,71], indicating that free inorganic phosphate is not used for ATP synthesis and excluding a role for ATP synthase in the process of ATP generation.

Most of the studies reporting a synthase activity have shown that oligomycin decreased extracellular ATP production after addition of ADP to culture media, thus suggesting that ecto-F₁-ATPase was indeed able to synthesize ATP. However, in our hands, oligomycin had no effect on ATP synthesis on the cell surface of human hepatocytes after addition of ADP[71]. The reason for these discrepancies is not clear. One should note that oligomycin has been shown to inhibit other enzymes, such as the non-gastric H,K-ATPase[81]. Thus it does not appear to be a strictly specific inhibitor of ATP synthase/F₁-ATPase and could act on other nucleotide-converting enzymes. One cannot exclude also that the activity of ecto-F₁-ATPase depends on particular experimental settings (culture conditions, cell type) and on the metabolic status of the cells. Nevertheless, in mitochondria, ATP production by ATP synthase is strictly dependent on the electrochemical proton gradient between the two sides of the inner mitochondrial membrane. This gives the chemical energy required for ATP synthesis. Whether a strong proton flux can occur at the cell surface and lead to ATP synthesis by ecto-F₁-ATPase is not certain and in the case of extracellular acidosis, the proton gradient would not be favorable for ATP synthesis. Mangiullo et al[34] have recently shown that ecto-F₁-ATPase was active in ATP synthesis and that it was correlated to an oligomycin-sensitive proton flux from the cytoplasm to the extracellular medium in the case of intracellular acidosis. Although, as discussed above, oligomycin might inhibit other proton pumps, this suggests that synthesis might occur in particular metabolic settings. It remains to be demonstrated whether these experimental conditions are relevant to physiological situations. Thus, the actual role of ecto-F₁-ATPase in proton transport across the PM remains arguable.

Accordingly, the synthase activity of ecto-F₁-ATPase is still an open issue, and there certainly is a need for reassessment of results obtained with the various techniques assessing ATP synthase activity as well as with the different inhibitors.

CONCLUSION

Over the past few years, data originating from many different groups have confirmed the cell surface expression of this mitochondrial enzyme. While this ectopic expression is now certainly widely accepted, there is still a need for a definitive characterization of its enzymatic activity, as well as new insights into the mechanism of its cell surface targeting.

In our opinion, the synthase activity of ecto-F₁-ATPase is still heavily challenged on both experimental and theoretical aspects. It is an important issue that requires additional work and may depend on the development of novel techniques which would be more specific and sensitive than those currently used. Nevertheless, the monitoring of dual-labeled ATP production from [³H]ADP and ³²Pi by high-performance liquid chromatography (HPLC) is probably so far the most reliable method to determine whether ecto-F₁-ATPase is active in ATP synthesis, since ATP synthase is the only known enzyme able to generate ATP directly from ADP and inorganic free phosphate. Therefore, it should be the method of choice to study the enzymatic properties of ecto-F₁-ATPase.

Understanding how ecto-F₁-ATPase reaches the cell surface is a challenging but extremely important point. As the use of siRNAs targeting ATP synthase subunits is hardly conceivable since it will undoubtedly rapidly lead to cell death, inhibiting its cell surface expression might help to better characterize its roles. Further studies are definitely required to decipher this mechanism.

ATP synthase is a fascinating molecular machine and a perfect example of a nanomotor. It has been one of the most studied enzymes over the past 40 years and garnered Professor Boyer PD and Sir Walker JE the Nobel Prize in Chemistry in 1997 for their outstanding work on the comprehension of its structure and the ATP synthesis mechanism[82-84]. Its central role in energy production is fundamental, an active person synthesizing their own weight in ATP every day. This essential role is now enhanced by its new implication in various important cellular processes, achieved through this ectopic localization and interactions with diverse ligands. We are only beginning to decipher these new functions.