Thematic Review Series: Lipid Transfer Proteins Is ABCA1 a lipid transfer protein?

ABCA1 functions as a lipid transporter because it mediates the transfer of cellular phospholipid (PL) and free (unesterified) cholesterol (FC) to apoA-I and related proteins present in the extracellular medium. ABCA1 is a membrane PL translocase and its enzymatic activity leads to transfer of PL molecules from the cytoplasmic leaflet to the exofacial leaflet of a cell plasma membrane (PM). The presence of active ABCA1 in the PM promotes binding of apoA-I to the cell surface. About 10% of this bound apoA-I interacts directly with ABCA1 and stabilizes the transporter. Most of the pool of cell surface-associated apoA-I is bound to lipid domains in the PM that are created by the activity of ABCA1. The amphipathic α-helices in apoA-I confer detergent-like properties on the protein enabling it to solubilize PL and FC in these membrane domains to create a heterogeneous population of discoidal nascent HDL particles. This review focuses on current understanding of the structure-function relationships of human ABCA1 and the molecular mechanisms underlying HDL particle production.

In this review, information is presented to answer the question of whether ABCA1 is a lipid transfer protein. It will become apparent that because the transporter mediates the transfer of PL (and FC) from a cell PM (donor) to apoA-I (acceptor), the simple answer is "yes". However, this response should be qualified because, as reviewed below, ABCA1 is not a traditional lipid transfer protein because it does not transport lipid molecules by binding them and diffusing from donor to acceptor lipid particle to transfer them. This review focuses on the molecular structure of ABCA1 and its PL translocase activity, as well as the role of its partner protein, apoA-I, in the mechanism of producing a heterogeneous distribution of nascent HDL particles. ABCA1 STRUCTURE AND ENZYME ACTIVITY ABC transporters are a family of integral membrane proteins that couple transport of chemically diverse substrates across the PL bilayer of cell membranes to the hydrolysis of ATP. The human genome codes for approximately 50 such transporters that are divided into seven subfamilies, designated ABCA through ABCG, based on organization of domains and amino acid homology (10). The 12 member ABCA subfamily, which includes ABCA1, is involved in moving lipid molecules across membranes (11). Figure 1A shows the general molecular architecture of ABC transporters and it is apparent that they comprise four core domains, namely, two transmembrane domains (TMDs) and two nucleotide binding domains (NBDs). ABCA1 is a full transporter and these four domains are formed by a single 2,261-residue polypeptide chain. ABCA Abstract ABCA1 functions as a lipid transporter because it mediates the transfer of cellular phospholipid (PL) and free (unesterified) cholesterol (FC) to apoA-I and related proteins present in the extracellular medium. ABCA1 is a membrane PL translocase and its enzymatic activity leads to transfer of PL molecules from the cytoplasmic leaflet to the exofacial leaflet of a cell plasma membrane (PM). The presence of active ABCA1 in the PM promotes binding of apoA-I to the cell surface. About 10% of this bound apoA-I interacts directly with ABCA1 and stabilizes the transporter. Most of the pool of cell surface-associated apoA-I is bound to lipid domains in the PM that are created by the activity of ABCA1. The amphipathic -helices in apoA-I confer detergent-like properties on the protein enabling it to solubilize PL and FC in these membrane domains to create a heterogeneous population of discoidal nascent HDL particles. This review focuses on current understanding of the structure-function relationships of human ABCA1 and the molecular mechanisms underlying HDL particle production. - Supplementary key words ATP binding cassette transporter A1 • high density lipoprotein • apolipoprotein A-I • cholesterol • phospholipid ABCA1 is a protein that plays a major role in HDL biosynthesis and cellular cholesterol homeostasis. It was discovered because mutations in ABCA1 can cause an autosomal recessive disorder called Tangier disease; this disease is characterized by low plasma HDL levels, deposition of cholesterol in tissue macrophages, and prevalent atherosclerosis (1). The tissue distribution of ABCA1 is ubiquitous, but its activity in hepatocytes and enterocytes is primarily responsible for plasma HDL production (2,3). The transporter mediates the release of phospholipid (PL) and free (unesterified) cholesterol (FC) from the plasma membrane (PM) of these cells to apoA-I in the extracellular medium, thereby producing nascent HDL particles (4)(5)(6). The ability of ABCA1 and other related ABC transporters to promote cellular FC efflux and enhance reverse choles-transporters, in general, and ABCA1, in particular, contain two extracellular domains (ECDs) that are glycosylated and contain some disulfide bonds ( Table 1); these accessory domains typically serve a regulatory role and are involved in protein-protein interactions (12). As summarized in Fig. 1B and Fig. 2, the structure and function of the NBDs is conserved across ABC transporters. High-resolution structures of monomeric (Escherichia coli vitamin B12 importer, BtuCD) and dimeric (Staphylococcus aureus multidrug exporter, Sav 1886) NBDs are available (13). The ATPase activities of the isolated NBDs of human ABCA1 have been compared (14) and the activity in the isolated transporter has been shown to be stimulated preferentially by choline-containing PL (15). Binding of two ATP molecules causes two NBDs to form a tight dimer with the ATP molecules at the interface; this event leads to a conformational change so that the TMDs form an outward-facing cavity (Fig. 2). Hydrolysis of the ATP to give ADP causes the NBDs to separate and drives conversion to an inwardfacing cavity. The generally accepted mechanism by which ABC transporters function to translocate substrate is the so-called "alternating access" model ( Fig. 2). The transport substrate selectivity is determined by the organization of the TMDs and the cavity formed by them. As an example, the high-resolution structure of the cavity formed by the bacterial ABC transporter, Sav 1886, has been solved (16).
Regarding human ABCA1, 12 transmembrane (TM) helices have been predicted (Table 1) and the membrane topology is summarized in Fig. 3 (17)(18)(19). The cryoelectron microscopy-derived structure of human ABCA1 (20) shown in Fig. 4 exhibits a broadly consistent TMD organization. However, the TMDs are exposed on all sides to lipids and do not form an obvious inwardfacing cavity, as would be expected for an ABC transporter in the absence of bound ATP (cf. Fig. 2). On the face of it, this conformation is inconsistent with the alternating Fig. 1. Representation of the general molecular structure of ABC transporters. A: Such transporters contain two TMDs and two conserved cytoplasmic ABCs (also designated as NBDs). Any given transporter is either an exporter (e.g., ABCA1) or importer. The two NBDs interact and, at the interface, dashed lines and red halfcircles represent the ABC signature motifs and P loops, respectively. Coupling helices transmit conformational changes between the NBDs and the TMDs. B: Schematic of a single NBD showing the locations of conserved and functionally critical motifs. The juxtaposition of conserved motifs in two NBDs creates two ATP-binding sites, each involving the P loop of one NBD and the signature motif of the other. The position of a bound ATP molecule is shown in the diagram. The and -phosphate groups of ATP are bound to the P loop (Walker-A motif) and the switch histidine contacts the -phosphate. The purine ring of adenine packs against an aromatic amino acid sidechain in the A loop. The Walker-B motif provides a catalytic glutamate involved in conversion of ATP to ADP and the signature LSGGQ motif holds and orients the ATP molecule during hydrolysis. The D loop is involved in dimerization of the NBDs and plays a role in coupling hydrolysis to transport. A groove in the NBD surface forms the contact interface with the coupling helix of the TMD. Reproduced with permission from Ref. 173.  (79) access model for PL translocation across the membrane. It is possible that this inconsistency is due to ABCA1 in digitonin micelles being in a different conformation from that in a PL bilayer where PL translocation is occurring. The fact that the ATPase activity of the transporter in digitonin micelles is very low supports this possibility (20). Importantly, the ABCA1 structure in Fig. 4 reveals that the ECDs adopt a novel protein fold. ECD1 and ECD2 form a hydrophobic tunnel that contains helices from both domains; this tunnel is not directly accessible from the membrane and is unlikely to be directly involved in mediating PL efflux. Because apoA-I interacts with the ECDs (see the section "apoA-I Binding to ABCA1-Expressing Cells" below), it is possible that separation of ECD1 and ECD2 and exposure of the hydrophobic surface lining the tunnel creates the binding site for apoA-I. More studies are required to establish the molecular details of the mechanistic links between the structure of ABCA1 shown in Fig. 4 and the known functions of the transporter with respect to PL translocation and nascent HDL particle assembly.
Because human ABCA1 is a PL exporter, the substrate PL binds with a higher affinity, probably via specific anchor points, to the conformation in which the translocation cavity is facing inwards to the cytoplasm. PL molecules in the inner leaflet of the PM diffuse into the cavity and are then translocated, which involves a 180° reorientation, as summarized in Fig. 2. Once in the outward-facing cavity, the PL molecule is bound with lower affinity (perhaps due to juxtaposition of groups with the same charge?) and can diffuse away into the exofacial leaflet. Potential modes of lipid translocation by ABC transporters have been reviewed recently (21). When reconstituted into liposomes, human ABCA1 can transport several classes of PL; the preference is phosphatidylcholine (PC) > SM ~ phosphatidylserine (PS) (22). Known Tangier disease-associated mutations, which are located in the ECDs and NBDs of human ABCA1, reduce PL translocase activity with the effect being greater for PS than for PC (22). Early studies of ABCA1 reported that it can translocate PS (23) and more recently another acidic PL, phosphatidylinositol (4,5)-bisphosphate (PIP2), has been shown to be a substrate (24). Mutations that cause Tangier disease (W590S in ECD1 and C1477R in ECD2) Fig. 2. Schematic of the alternating access mechanism of TM movement of substrate by an ABC transporter such as ABCA1. The TMDs form a transport substrate-binding cavity that can open to either the cytoplasmic leaflet or the exofacial leaflet of the cell PM. In the absence of bound nucleotide, the NBDs are far apart (open conformation) and the cavity faces the cytoplasm (left structure in diagram). When transport substrate (red circle) binds in this cavity and ATP binds to the NBDs causing them to move together and dimerize into the closed conformation (the power stroke), the coupled TMDs change conformation and the cavity shifts from inward-facing to outward-facing. This change causes the substrate to be translocated and become available for release on the other side of the membrane (right side of the diagram). Subsequent ATP hydrolysis (usually of both bound ATP molecules) and dissociation of ADP causes the NBDs to separate and adopt the open conformation, and the transporter returns to the starting conformation. Both exporters and importers seem to use the same mechanism, but differ in which state binds the transport substrate with higher affinity. As a PL exporter, ABCA1 binds PL more tightly with the cavity open to the cytoplasmic leaflet of the membrane. For the purpose of 2D illustration, the ATPase sites are shown as above one another, but in reality they are equidistant from the membrane. Reproduced with permission from Ref. 174.  Table 1. Two ECDs are formed by amino acids located between H1 and H2 and between H7 and H8. The two cytoplasmic NBDs are close together in 3D space (Figs. 1A, 4). have different effects on PS and PIP2 translocation. Because the W590S mutation reduced PS transport without affecting PIP2 transport and the C1477R mutation had the opposite effect, it was concluded that these two PL translocase activities of ABCA1 are independent of each other (24). Better understanding of PL substrate specificity requires more detailed knowledge of the TMD organization in human ABCA1.

SUBCELLULAR DISTRIBUTION OF ABCA1
ABCA1 is present mainly in the PM of cells (25)(26)(27) and this localization is dependent upon it being appropriately palmitoylated (28). Human ABCA1 is palmitoylated at four cysteine residues ( Table 1) and reduction of palmitoylation due to mutation of cysteine residues decreases the amount of transporter in the PM. Functional ABCA1 in the PM selfassociates and has been reported variously to form dimers and tetramers. The majority of ABCA1 in human skin fibroblasts is tetrameric and apoA-I binds to the transporter in this state (29). However, ABCA1 is predominantly dimeric in both Hela cells and mouse macrophage RAW cells, with this assembly taking place in the endoplasmic reticulum (ER) (30). Single molecules of ABCA1-green fluorescent protein in the PM of Hela cells have been visualized by total internal reflection fluorescence microscopy (31). Most of the transporter molecules are present as immobile dimers and this self-association requires its ATPase activity. Interestingly, two apoA-I molecules bind to the ABCA1 dimer to generate nascent HDL particles and this event is followed by dissociation of the ABCA1 dimers to monomers (see the section "ABCA1-Mediated PL/apoA-I Interaction" below). It is reported that apoA-I cannot bind to monomeric ABCA1 (29,31).
The presence of active ABCA1 in the PM leads to lipid reorganization. Thus, membrane FC is redistributed to cholesterol oxidase-accessible pools (32), most likely due to movement from raft domains (enriched in FC and SM) to nonraft domains (33,34). This rearrangement, which includes movement of SM, is due to destabilization of raft domains and is accompanied by partitioning of ABCA1 into nonraft domains (35). The question of how the transporter is distributed between raft and nonraft membrane domains has been investigated extensively, but a clear answer is lacking. When detergents have been used to separate the membrane domains, ABCA1 has been reported to be either mostly in nonraft domains (34,36,37) or in raft domains (38). Application of nondetergent methods to separate raft and nonraft domains has shown variously that ABCA1 can be enriched in either type of membrane domain (35,39,40). The lack of agreement is due to variations in experimental methods; for instance, the type of detergent used to fractionate raft and nonraft domains, the cell type, and the level of ABCA1 expression. As pointed out in a review of these issues (41), at this stage it is difficult to draw firm conclusions about ABCA1 location in the PM. It will be important to resolve what effects, if any, the composition and physical state of the lipids in the boundary layer around an ABCA1 molecule have on its functionality.

INTRACELLULAR TRAFFICKING OF ABCA1
Besides residing on the cell surface, ABCA1-green fluorescent protein in Hela cells has been shown to traffic between the PM and intracellular vesicles, including early and late endosomes (27,42). The presence of ABCA1 in intracellular sites affects vesicle trafficking and the distribution Fig. 4. Structure of human ABCA1 determined by single-particle cryoelectron microscopy. The nominal resolution for the overall structure is 4.1 Å and 3.9 Å for the ECD. Individual ABCA1 molecules were visualized by solubilizing them in small digitonin micelles that contain 60 molecules of detergent. The various domains in the protein are colored differently and glycosyl groups are shown as black sticks. The picture shows the structure of the transporter in the absence of ATP. The ABCA1 conformation was stable in the digitonin micelles, but the ATPase activity was very low. The presence of two NBD and two TMD domains is characteristic of ABC transporters in general (cf. Fig. 1). The regulatory domain (R domain) contains protein recognition sites (Table 1) from both halves of the transporter (amino acids 1182-1251 and 2155-2220, respectively) and is located near the two NBDs, which are juxtaposed. For the two TMDs, both the topology with respect to the membrane and the helix locations in the amino acid sequence are largely consistent with Fig. 3 and Table 1. Coupling helices that are orientated parallel to the membrane and link the conformations of the NBDs and TMDs (cf. Fig. 1A) are located between residues 3-20, 663-678, 1327-1344, and 1678-1696. The two ECDs characteristic of ABCA transporters are juxtaposed and cofolded in a twisted fashion with ECD1 positioned above TMD2 and ECD2 positioned above TMD1. The overall ECD has a hollow interior because helices in the two domains enclose an 60 Å high tunnel that has a predominantly hydrophobic interior. ECD1 and ECD2 contain three and one disulfide bonds, respectively (Table 1). Reproduced with permission from Ref. 20. of FC between pools in the ER, Golgi, and PM (43)(44)(45). Cholesterol in all of these sites contributes to overall ABCA1-mediated cholesterol efflux from the cells (46). In macrophages, FC in late endosomes/lysosomes is a preferential source for efflux (47). The rapid internalization of PM constituents to endocytic compartments (t 1/2 = 5-10 min) and recycling of the endosomes to the PM facilitates such cholesterol efflux (48). ABCA1-mediated mobilization of lysosomal FC requires functional Niemann-Pick C2 protein, but not Niemann-Pick C1 protein, consistent with ABCA1 being able to mobilize FC from the membrane, but not the lumen, of lysosomes (49).
Because, as outlined above, ABCA1 can undergo rapid retroendocytosis, any apoA-I bound to the transporter also cycles between the PM and cell interior. Indeed, apoA-I retroendocytosis has been proposed to contribute to nascent HDL production (50). Support for this idea comes from experiments showing that reduction of ABCA1 endocytosis induced by either mutant forms of the transporter (51) or the use of inhibitors of clathrin-dependent endocytosis (52) can decrease nascent HDL formation. However, counter to the view that the retroendocytosis pathway is significant, it has been shown that most internalized apoA-I does not colocalize with ABCA1 and is degraded in lysosomes. Consequently, insufficient intact apoA-I is resecreted to contribute significantly to the observed overall level of nascent HDL production (53,54). Furthermore, blocking endocytosis and apoA-I internalization using pharmacological agents has negligible effects on cholesterol efflux and HDL biogenesis. It follows that ABCA1 located in the PM is primarily responsible for lipidation of apoA-I and production of HDL particles (53,54).

apoA-I BINDING TO ABCA1-EXPRESSING CELLS
It is well-established that, regardless of cell type, expression of ABCA1 increases the binding of apoA-I to the cells (26,(55)(56)(57)(58)(59)(60)(61)(62). This binding process involves, at least in part, direct protein-protein contact between apoA-I and ABCA1 because the two proteins can be covalently cross-linked using linkers with short spacer arms. Inspection of the membrane topology of ABCA1 (Fig. 3) indicates that the ECDs (which, as mentioned earlier, serve a regulatory role in ABCA transporters) are the most likely interaction sites for apoA-I in the extracellular medium. apoA-I does, in fact, interact with the two ECDs in ABCA1 (Table 1) because binding is sensitive to mutations of these domains (57). The binding is not highly specific and seems only to require amphipathic -helices in the ligand because other apolipoproteins and peptides containing this structural motif bind to ABCA1 (59). apoA-I variants containing different combinations of amphipathic -helices also bind effectively to the transporter, further indicating that the binding does not require highly specific interactions. The disulfide bonds C75-C309 and C1465-C1477 in ECD1 and ECD2, respectively (Table 1), are required for ABCA1 to be fully functional (63). The conformations of the ECDs and the consequent apoA-I binding are dependent upon effective ATP hydrolysis at both NBD1 and NBD2 (Table 1) (64). Monoclonal antibodies to regions in either ECD1 or ECD2 inhibit apoA-I binding consistent with a key role for these domains (65). Interestingly, one of the monoclonal antibodies binds to residues 1370-1450 in ECD2 (Table 1), which is close to amino acid C1477 that mutagenesis studies (18) suggest is in a region crucial for apoA-I binding. apoA-I binding to ABCA1 requires appropriately charged lysine residues in the ECDs because chemically modifying these residues to eliminate the positively charged sidechain amino group reduces apoA-I binding (66).
Expression of the ABCA1 gene is regulated by transcription factors, the most important being the liver X-receptor (LXR) (67,68), which is sensitive to increases in cell cholesterol levels and upregulates ABCA1 under this condition. Posttranslational regulation of ABCA1 expression is also important for modulating its activity. ABCA1 turns over rapidly because, as described above, it is endocytosed quickly and becomes subjected to intracellular proteolytic degradation. Importantly, when apoA-I is bound to ABCA1, the transporter is stabilized by being protected from this proteolysis (69). Tall and colleagues showed that this proteolysis, which is inhibited by bound apoA-I, is mediated by calpain and dependent on a PEST sequence in ABCA1 (Table 1) (70). This inhibitory action of apoA-I is due to dephosphorylation of amino acids T1286 and T1305, which are constitutively phosphorylated; this dephosphorylation inhibits degradation by calpain, which apparently cuts ABCA1 at or near the PEST sequence (71). This calpain-mediated proteolysis occurs in early endosomes and, when it is inhibited by the presence of bound apoA-I, ABCA1 recycles to the cell surface, thereby enhancing nascent HDL biogenesis. This regulatory function of apoA-I binding to ABCA1 is of clear physiological importance.
The activity of ABCA1 is also sensitive to other factors. Thus, apoA-I and other amphipathic -helical molecules can induce overall phosphorylation and stabilization of ABCA1 (72,73). The activities of several protein kinases have been reported to modulate ABCA1 functionality (72,74,75). Besides stabilization of the transporter, binding of apoA-I to ABCA1 activates multiple cellular signaling pathways that regulate activity of the transporter (76). ABCA1 can interact with intracellular proteins, which modulate its activity. For instance, a calmodulin binding site is located near the PEST sequence ( Table 1) and binding of calmodulin protects ABCA1 against calpain-mediated degradation (77). In contrast, caveolin-1 binds to ABCA1 and enhances its degradation by increasing movement into the cell interior (78). In the absence of apoA-I, ABCA1 turns over rapidly with a half-life of 1-2 h and interaction of the PDZ-binding protein, -syntrophin, reduces this rate (79). The PDZ binding domain in ABCA1 is located at the extreme C terminus (Table 1) (79,80). Another possible C-terminal protein binding site encompasses residues 2215-2220 (Table 1); this VFVNFA motif, which is located in the cytoplasm, is required for the cholesterol efflux and apoA-I binding activities of ABCA1 (81).

MUTATIONS IN ABCA1
Genetic variation in ABCA1 contributes to plasma HDL cholesterol levels in the general population (82) and mutations that lead to large reductions in HDL (as seen in Tangier disease) are associated with increased risk of coronary artery disease (83). Studies of 15 missense mutations seen in patients with Tangier disease and familial hypoalphalipoproteinemia indicate that the severity and nature of the defects in ABCA1 function correlate with the clinical lipid phenotypes observed in these subjects (84). Generally, Tangier disease-associated mutations lead to parallel reductions in ABCA1-mediated efflux of cellular PL and FC (85). Mutations that disrupt glycosylation and interfere with the trafficking of ABCA1 within cells and prevent its localization in the PM fail to promote apoA-I binding to the cells and HDL particle production. Some mutations allow the transporter to locate in the PM, but perturb the conformation so that proper apoA-I binding is prevented. For example, the variant C1477R (cf. Ref. 18), which is in ECD2 (Table 1), exhibits this behavior (84) presumably because appropriate disulfide bridge formation in ECD2 does not occur. The mutation W590S in ECD1 is associated with Tangier disease and has been studied in several laboratories (18,57,(85)(86)(87) and there is agreement that this variant ABCA1 reaches the PM, but exhibits defective cellular PL and FC efflux. However, there is controversy regarding the effects of this mutation on total apoA-I binding to the cells; Vaughan, Tang, and Oram (85) report a 50% reduction relative to WT ABCA1, whereas others (87) report that apoA-I binding is unaffected and suggest that the defect is probably in lipid translocation. More work is required to elucidate the mechanistic basis for the impaired functionality of the ABCA1 W590S variant. The importance of appropriate glycosylation of residues in ECD1 for ABCA1 trafficking is demonstrated by the fact that the Tangier variants, R587W and Q597R, are associated with impaired processing of oligosaccharide from high mannose type to complex type and do not localize in the PM, but remain in the ER (86). Interestingly, the Tangier disease mutation, R1068H, in NBD1 (Table 1) alters the conformation of this domain, also resulting in improper trafficking to the PM (88). Mutagenesis has also been applied to investigate the effects of TM helix structure on ABCA1 activity. Strikingly, the mutation L834R in TM helix 6 ( Table 1) decreases PL and FC efflux without altering the amount of transporter in the PM (89). However, apoA-I binding is reduced, presumably due to conformational changes in the ECDs caused by changes in the structure of TM helix 6 and its orientation in the PM. Future mutagenesis studies can be expected to provide important mechanistic insights into ABCA1 structure-function relationships.

FACTORS INFLUENCING ABCA1 ACTIVITY
In addition to the effects of changes in its primary structure, exogenous factors can modulate ABCA1 activity. These factors may be present in the extracellular medium, in the PM, or in the cytoplasm and can either promote or inhibit FC efflux. As mentioned earlier, transcription factors such as LXR can increase ABCA1 expression and enhance FC efflux. However, LXR- can interact directly with ABCA1 at C-terminal residues 2247-2251 and inhibit ATP binding and the activity of NBD2 in cholesterol-normal cells (90). In contrast, in cholesterol-enriched cells that contain elevated levels of oxysterols, the LXR- dissociates from ABCA1 increasing its activity and FC efflux (91). These studies demonstrate that LXR is involved in both transcriptional and posttranscriptional regulation of ABCA1. The SM catabolite, ceramide, stabilizes ABCA1 and increases its concentration in the PM, thereby increasing FC efflux (92). The oxidation products of probucol, spiroquinone, and diphenoquinone also stabilize the transporter by protecting it from calpain-mediated degradation, which results in increased HDL biogenesis (93). The underlying mechanism involves disruption of the caveolin-ABCA1 interaction by these oxidation products and reduced internalization of the transporter and exposure to calpain (78). ABCA1-mediated FC efflux is sensitive to the SM content of the PM, with a reduction in SM leading to increased FC efflux (94). The increased efflux upon SM depletion is due, at least in part, to increased levels of PS in the exofacial leaflet of the PM; this redistribution is due to impaired inward PS translocation upon SM depletion (95). The effect of PL transfer protein (PLTP) is an example of a factor in the extracellular medium that influences ABCA1 activity; PLTP interacts with ABCA1 and stabilizes it (96), thereby promoting ABCA1-mediated FC and PL efflux as well as nascent HDL particle remodeling (97).
Various pharmacological agents inhibit ABCA1 activity. One of the most intensively studied is probucol, which inactivates ABCA1 in the PM (without changing its concentration), thereby reducing FC efflux and HDL production (98). These effects occur despite a reduction in calpainmediated proteolysis of ABCA1 in probucol-containing cells. In contrast to these findings, another study showed that impaired translocation of ABCA1 to the PM contributes to the reduced lipid efflux measured in the presence of probucol (99). The immunosuppressant drug, cyclosporine A, increases the ABCA1 content of the PM by reducing its cycling into the cell interior, but, despite the increased availability of the transporter, FC efflux is strongly inhibited (100). This inhibition occurs because cyclosporine A binds directly to ABCA1, but where it binds and the underlying mechanism are not known at this point (101). Pharmacological agents that can bind PL have also been shown to inhibit ABCA1-mediated FC efflux (102). Another change in cell lipids that can inhibit ABCA1 is the presence of unsaturated fatty acids (103); their presence leads to phosphorylation of serine residues and destabilization of the transporter. An example of an extracellular factor that inhibits ABCA1 activity is acidic conditions; at pH 5.5, FC efflux to apoA-I is reduced within 1 h of acidification, presumably due to pH-induced changes in the conformation of ABCA1 (104). Such inhibition of ABCA1 may occur in advanced atherosclerotic plaque where the intimal fluid is acidic.  (105). PL translocation by ABCA1 (large yellow arrow) is proposed to increase the surface pressure within the exofacial monolayer of the PM bilayer and induce formation of a discoidal PL bilayer pleat that becomes the growing lipid reservoir. apoA-I then binds to the extracellular bilayer pleat to create a discoidal nascent HDL (dHDL) particle. Reproduced with permission from Ref. 105.

ABCA1-MEDIATED PL/apoA-I INTERACTION
Interaction of apoA-I with cellular PL is essential for formation of nascent HDL particles and the interaction is associated with cholesterol efflux from cells. While it is established that the PL translocase activity of ABCA1 causes reorganization of the PM and increased apoA-I binding to cells (see the section "apoA-I Binding to ABCA1-Expressing Cells" above), how apoA-I/PL interactions contribute to the increased binding has been controversial. Because direct apoA-I/ABCA1 interaction occurs, early models of nascent HDL biogenesis have been based on the idea that the transporter loads translocated PL molecules onto this pool of bound apoA-I. Two models describing this type of mechanism are summarized in Fig. 5. In both cases, the translocated PL is held by the ECDs as a "reservoir" from which bound apoA-I can acquire PL to form discoidal nascent HDL particles. Ueda and colleagues showed that the stoichiometry of apoA-I/ABCA1 interaction is 1/1 mol/ mol and formation of ABCA1 dimer is critical for HDL particle production (Fig. 5A) (31). This model provides a mechanism for formation of discoidal HDL particles containing two apoA-I molecules, but not those containing three or four apoA-I molecules (see below). Segrest et al. (105) proposed a variation of the reservoir model in which the pool of PL interacting with the ECDs is created by pleating of the exofacial leaflet of the PM (Fig. 5B). These models were derived before the high-resolution structure of the ECDs (Fig. 4) was known and will need to be put in the context of this information. A limitation of the reservoir models of nascent HDL particle formation is that they do not account for PL bilayer solubilization being involved similarly in discoidal HDL particle formation in both cell and cell-free systems (see the section "Characterization of Nascent HDL" below).
Implicit in the reservoir models of nascent HDL biogenesis is the idea that all of the additional apoA-I that binds to cells when ABCA1 is expressed is interacting directly with the transporter. However, detailed cross-linking and binding studies of apoA-I interacting with macrophages (62) and fibroblasts (61) have shown that only about 10% of the bound apoA-I is bound directly to ABCA1. Most of the apoA-I binds with high affinity to lipid sites in the PM created by the PL translocase activity of the transporter. These lipid domains in the PM to which apoA-I binds protrude into the extracellular space (105)(106)(107) and are thought to be the sites of HDL particle assembly. It has been suggested that apoA-I must interact with ABCA1 before it can interact with PL in the PM (108). Such a "hand-off" mechanism cannot be excluded, although the highly flexible apoA-I molecule (109) possesses excellent lipid-binding capabilities without any activation.
Binding of apoA-I to the PM in ABCA1-expressing cells leads to concurrent release of PL and FC (110,111); prior depletion of FC from the PM by treatment with cyclodextrin leads to dissociation of FC and PL efflux (55). The release of PL and FC from the PM involves PL bilayer solubilization due to the detergent-like properties of the apoA-I molecule (112,113). The solubilization process involves insertion of amphipathic -helices between the PM PL molecules, which destabilizes and fragments the bilayer (114). Information about the properties of apoA-I and ABCA1 was integrated to develop the mechanism of nascent HDL particle formation presented in Fig. 6 (107). A central feature of this mechanism is that PL translocation via ABCA1 induces bending of the PM bilayer to create high curvature sites to which apoA-I can bind. This binding to exovesiculated PM domains leads to bilayer destabilization and solubilization of PL and FC to create discoidal nascent HDL particles. The latter step is rate-limiting for the overall process of nascent HDL particle formation. Variations in apoA-I structure exert similar effects on the rates of formation (and sizes) of HDL particles created by either spontaneous solubilization of PL vesicles or the ABCA1mediated efflux of cellular lipids (107,115). The rate of solubilization of PL bilayers is enhanced by the presence of acidic PL (116,117), so any accumulation of molecules, such as PS and PIP2, in the exofacial leaflet of the PM (23, 24) may also enhance nascent HDL particle production.
apoA-I can solubilize membrane PL and FC because it contains amphipathic -helices (109). Consequently, other exchangeable apolipoproteins containing this structural motif also create nascent HDL particles when incubated with ABCA1-expressing cells (118); -synuclein, which contains amphipathic -helices, is also able to participate effectively in ABCA1-mediated cellular FC efflux (119). Modifications of the amphipathic -helix content and structure of the apoA-I molecule that alter its lipid binding properties also alter its ability to partner with ABCA1 to create nascent HDL particles (120,121). In particular, the hydrophobicity of the C-terminal -helix in human apoA-I is critical for this function (122)(123)(124). Indeed, the C-terminal segment of human apoA-I (residues 209-241) has the appropriate molecular features to be able to mediate FC efflux via ABCA1 as an isolated peptide (125). Consistent with the idea that specific amino acid sequences are not required for functionality, various nonhomologous peptides that contain amphipathic -helices perform like apoA-I in supporting ABCA1-mediated lipid efflux from cells (126)(127)(128). There is no stereospecific requirement for the participation of amphipathic -helices in ABCA1-mediated efflux of cellular PL and FC because the same peptide containing either l-amino acids (right-handed -helix) or d-amino acids (left-handed -helix) functions equally well (127). Unlike apoA-I, in the absence of ABCA1 activity particularly hydrophobic amphipathic -helical peptides with high lipid affinity can promote efflux of cellular PL and FC by membrane solubilization, which is usually accompanied by some cell lysis (127). Small amphipathic -helical peptides that are agonists for ABCA1 are being developed as possible therapeutic approaches for combatting diseases such as atherosclerosis, diabetes, and Alzheimer's disease (129).
The physical state of apoA-I in the extracellular medium strongly influences the ability of the protein to support ABCA1-mediated lipid efflux and create nascent HDL particles. apoA-I is the most efficient acceptor of cellular lipid when it is present as a lipid-free monomer [K m 0.1 M apoA-I (120)]; the relative efficiency of self-associated apoA-I is less than 50% (130). apoA-I has to accept PL from the cell membrane to create discoidal HDL particles, so this ability is lower when the apoA-I already contains a Fig. 6. Summary of the molecular mechanism by which ABCA1 activity in the PM of cells promotes efflux of PL and cholesterol to extracellular apoA-I and formation of nascent HDL particles. As shown at the top of the diagram, direct apoA-I/ABCA1 interaction and apoA-I/membrane lipid interactions occur with the former leading to transporter stabilization and the latter to HDL particle assembly. The PM-activated lipid domain to which apoA-I binds is created as a consequence of the PL translocation induced by ABCA1. As shown in the lower part of the figure, the activated lipid domain is formed by membrane bending and comprises an exovesiculated segment of the PM. Amphipathic -helices in the apoA-I molecule confer detergent-like properties on the protein, allowing it to solubilize PL by binding to lattice defects in highly curved PL bilayer surfaces, thereby inducing bilayer fragmentation and formation of discoidal nascent HDL particles. These particles comprise small segments of PL/cholesterol bilayer (containing on the order of 100 PL molecules) that are most frequently stabilized by either two or three apoA-I molecules (the smaller and larger nascent HDL particles are formed simultaneously). The membrane solubilization step mediated by apoA-I is rate-limiting for the overall efflux of PL and FC from the cell. The catalytic efficiency (V max /K m ) of apoA-I is highest for the lipid-free protein so that its efficiency is reduced by prior phospholipidation. See the text for further details. Reproduced with permission from Ref. 6. Fig. 7. Demonstration of the size heterogeneity of nascent HDL particles typically created by the incubation of apoA-I with ABCA1expressing cells. Western blots of 2D native gel electrophoresis of apoA-I-containing nascent HDL particles generated by incubation of J774 macrophages and human skin fibroblasts with human plasma apoA-I are shown. After incubation of cells with human plasma apoA-I (5 g/ml) for 24 h at 37°C, medium was electrophoresed in the first dimension in a 0.7% agarose gel followed by electrophoresis in the second dimension in a 2-36% concave polyacrylamide gel. The nascent HDL bands from the 2D gel were transferred onto a PVDF membrane and probed with a polyclonal anti-apoA-I antibody. A: J774 macrophage whole medium. B: Human skin fibroblast whole medium. Molecular size markers (hydrodynamic diameter in nanometers) are indicated. Reproduced with permission from Ref. 135. complement of PL at the time it is added to the extracellular medium (131). Consistent with the idea that prior phospholipidation reduces the efficiency with which apoA-I participates in ABCA1-mediated cellular lipid efflux, larger HDL particles that contain more PL are relatively poor acceptors (132,133). The measured ABCA1-mediated FC efflux to HDL particles is likely to contain a contribution from apoA-I that has dissociated from the HDL particles (134).

CHARACTERIZATION OF NASCENT HDL
Incubation of lipid-free apoA-I with ABCA1-expressing cells leads to the production of apoA-I-containing nascent HDL particles and the concomitant release of larger cholesterol-containing, but apoA-I-free, microparticles (135)(136)(137). Two-dimensional native gel electrophoresis of the conditioned medium shows that, regardless of cell type, the apoA-I-containing HDL particle population is heterogeneous with respect to size and surface charge ( Fig. 7) (135,(138)(139)(140)(141)(142)(143); the various particles are produced simultaneously (131,135,138). The particles are discoidal in shape and the predominant species have hydrodynamic diameters in the 8-12 nm range. These particles contain two, three, or four apoA-I molecules per disc (135). Under certain circumstances a lipid-poor apoA-I particle (pre1-HDL) comprising a single apoA-I molecule and three to four PL molecules and one to two FC molecules is produced (144). The discoidal HDL particles consist of a segment of PL/FC bilayer stabilized by apoA-I wrapped around the edge (113) and contain various species of PL molecules (40,60,135,138,(145)(146)(147)(148)(149). PC is the predominant class of PL incorporated into these nascent HDL particles followed by SM and various types of acidic PL to a lesser extent. The exact PL composition varies with cell type and does not simply reflect the PL composition of the PM or the raft and nonraft domains contained therein. Due to differences in PL molecular shape and charge, lateral domains containing different molecular species can form in a membrane due to variations in local curvature (150,151). Thus, the PM PL molecules will be sorted to form the highly curved exovesiculated domain created by ABCA1 activity (Fig.  6); more work is required to understand the details of this process. Up to now, the apoA-I-binding PM microdomains created by the activity of ABCA1 have not been isolated. This problem is technically challenging (152), although recent studies of PM microdomains in ABCA1expressing cells from the laboratory of Genest and colleagues have led to successful characterization of a novel apoA-I-binding domain that contains desmocollin 1 (153). Interestingly, sequestration of apoA-I by desmocollin 1-containing PM microdomains reduces the availability of apoA-I for HDL particle production, thereby inhibiting FC efflux (152,153).
Factors controlling the size distribution of nascent HDL particles created by incubating apoA-I with ABCA1expressing cells are listed in Table 2. Because the solubilization of PL bilayers by apoA-I occurs similarly in cell-free systems and at the surface of ABCA1-expressing cells, parallel effects are observed in both cases. For instance, a shift to formation of larger particles containing more PL and apoA-I molecules is seen when the ratio of available PL to apoA-I is increased in both a dimyristoyl PC multilamellar vesicle/apoA-I system (154) and cell systems where ABCA1 is active (155). Variations in apoA-I structure and physical properties also exert similar effects on the sizes of HDL particles created by PL solubilization in cell-free and cell systems (115). The presence of the hydrophobic C-terminal domain of the human apoA-I molecule is critical for formation of smaller HDL particles ( Table 2). The relative production of larger and smaller nascent HDL particles is sensitive to the PL acyl chain composition and the FC content of the PM ( Table 2). The relative incorporation into HDL of PL from the exofacial and cytoplasmic leaflets of the cell PM has not been studied. However, there may not be much discrimination between these two pools of PL because apoA-I simultaneously solubilizes PL from the outer and inner leaflets of the bilayer in PL vesicles (156).
The size of discoidal nascent HDL particles influences the distribution of lipids contained in them because of molecular packing constraints in the nanoscale particles. The size-dependent distribution of PL molecules is due to varying amounts of PL being sequestered in a boundary layer with apoA-I at the disc edge (40). For instance, the presence of a relatively large boundary layer in a smaller discoidal HDL particle promotes preferential distribution of PS to such particles. Nascent HDL particles formed by the action of ABCA1 contain FC and this process promotes efflux of excess cholesterol from cells. The FC/PL ratio is higher in larger discoidal nascent HDL particles than in smaller ones (40,135,138,148). The same pattern of FC distribution is seen in HDL particles created by apoA-I-mediated PL vesicle solubilization in a cell-free system (40,157). The molecular mechanism responsible for the preferential distribution of FC to larger particles is summarized in Fig. 8. The FC content of discoidal HDL particles influences their reactivity with LCAT (147,158). It follows that, in vivo, if larger and smaller nascent HDL particles interact directly with LCAT, they would be expected to be remodeled at different rates when they enter the plasma compartment. However, the remodeling of nascent HDL particles in plasma is complex, so that these discoidal particles are not simply converted to mature spherical HDL particles by direct interaction with LCAT (141-143).

PHYSIOLOGICAL CONTRIBUTIONS OF NASCENT HDL CREATED BY ABCA1
The heterogeneity of nascent HDL summarized above underlies the heterogeneity of plasma HDL. The HDL particles of different sizes produced by the activity of ABCA1 at the surfaces of hepatocytes and enterocytes enter the circulation and are metabolized as discrete species (159). The various subspecies of HDL mediate cellular cholesterol efflux to different extents (6) and exert different effects on the overall rate of the reverse cholesterol transport pathway (160). Consequently, the various HDL subspecies possess different anti-atherogenic capabilities, which are not simply correlated with their cholesterol contents (161,162). The ability of HDL to remove cholesterol from cholesterol-loaded macrophages (foam cells) is key for reducing atherosclerosis and requires expression of ABCA1 together with ABCG1 (163). ABCA1 and ABCG1 synergize to remove cholesterol from cells because the nascent HDL particles created by the interaction of apoA-I with Fig. 8. Molecular mechanism responsible for the preferential distribution of cholesterol to larger nascent HDL particles during formation by membrane solubilization. Molecular packing constraints in nanoscale discoidal HDL particles of different sizes modify the cholesterol-solubilizing capacity. The molecular packing of the PL molecules adjacent to the -helices of the apoA-I molecules wrapped around the circumference of the discs is constrained so that their ability to solvate cholesterol molecules is reduced. The diagram shows a top down view of four HDL discs with their diameters drawn to scale. As explained in Ref. 40, the areas of each bilayer leaflet occupied by boundary PL and the remaining PL (labeled as PL available for cholesterol) were calculated for 9, 12, 14, and 17 nm HDL discs. The fraction of PL forming the boundary layer varies inversely with diameter for the discoidal HDL particles so that the fractional availability of PL for solvating cholesterol is low in small HDL particles. It is apparent that approximately doubling the particle diameter from 9 to 17 nm leads to a 13-fold increase in the amount of PL into which cholesterol can dissolve (relative available PL area). Reproduced with permission from Ref. 40.
Nascent HDL heterogeneity is demonstrated in Fig. 7. The hydrodynamic diameters of the larger and smaller HDL particles are approximately 11 and 8 nm, respectively. The relative production of larger and smaller nascent HDL particles is controlled by the ratio of available cellular PL to concentration of apoA-I in the extracellular medium, with a higher ratio favoring production of larger particles (155).
a The number of + and  signs indicates the direction and degree of change relative to either the chosen starting condition or WT human ABCA1 or apoA-I. ABCA1-expressing cells can effectively promote cholesterol efflux via ABCG1 (164,165); the latter process occurs by the aqueous diffusion mechanism (6,166). ABCA1 plays the major role in effluxing cholesterol from cholesterolloaded macrophages to extracellular apoA-I and HDL, and this pathway determines the capacity of serum samples to support this process (167). The so-called cholesterol efflux capacity, which is a measure of HDL functionality, is inversely associated with the likelihood of coronary artery disease, independently of HDL cholesterol levels (168)(169)(170)(171).
An essential feature of the central role of ABCA1 in the above processes is its ability to promote PL and FC efflux from cells to extracellular acceptors, such as apoA-I. Therefore, on this basis and as mentioned in the Introduction, ABCA1 functions as a lipid transfer protein. The enzymatic activity of ABCA1 leads to the transfer of PL from the cytoplasmic leaflet of the cell PM to the exofacial leaflet. The net transfer of PL (and FC) out of the cell and formation of nascent HDL particles requires the membrane-solubilizing capabilities of the amphipathic -helices in the major acceptor protein, apoA-I.