Deciphering structural aspects of reverse cholesterol transport: mapping the knowns and unknowns

Atherosclerosis is a major contributor to the onset and progression of cardiovascular disease (CVD). Cholesterol‐loaded foam cells play a pivotal role in forming atherosclerotic plaques. Induction of cholesterol efflux from these cells may be a promising approach in treating CVD. The reverse cholesterol transport (RCT) pathway delivers cholesteryl ester (CE) packaged in high‐density lipoproteins (HDL) from non‐hepatic cells to the liver, thereby minimising cholesterol load of peripheral cells. RCT takes place via a well‐organised interplay amongst apolipoprotein A1 (ApoA1), lecithin cholesterol acyltransferase (LCAT), ATP binding cassette transporter A1 (ABCA1), scavenger receptor‐B1 (SR‐B1), and the amount of free cholesterol. Unfortunately, modulation of RCT for treating atherosclerosis has failed in clinical trials owing to our lack of understanding of the relationship between HDL function and RCT. The fate of non‐hepatic CEs in HDL is dependent on their access to proteins involved in remodelling and can be regulated at the structural level. An inadequate understanding of this inhibits the design of rational strategies for therapeutic interventions. Herein we extensively review the structure–function relationships that are essential for RCT. We also focus on genetic mutations that disturb the structural stability of proteins involved in RCT, rendering them partially or completely non‐functional. Further studies are necessary for understanding the structural aspects of RCT pathway completely, and this review highlights alternative theories and unanswered questions.


I. INTRODUCTION
Cholesterol accumulation in coronary arteries is the leading cause of morbidity worldwide (Farzadfar, 2019). This is a striking outcome for a common membrane lipid that maintains bilayer fluidity, integrity, and permeability (de Meyer & Smit, 2009). Cholesterol in the membrane is associated with sphingolipids and glycosylphosphatidylinositol (GPI)-anchored proteins, which form lipid rafts that regulate membrane trafficking, signal transduction and play a role in host-pathogen interactions (Silvius, 2003;Vieira et al., 2010). Given its central role in diverse physiological processes, cholesterol homeostasis is tightly regulated through an interplay among the mechanisms of its synthesis, dietary uptake, storage, and export (Luo, Yang & Song, 2020). While genes involved in both cholesterol biosynthesis and uptake are ubiquitously expressed, most cells are unable to catabolise it. These cells must resort to storing excess cholesterol as esters (inside lipid droplets) or employ efflux mechanisms that can remove the surplus. This net movement of excess cholesterol from peripheral tissues occurs by packaging it within highdensity lipoprotein (HDL) and constitutes the reverse cholesterol transport (RCT) pathway. RCT has long been considered to be the only pathway that reduces cholesterol load in the peripheral compartment by increasing its flux towards the liver where it can be catabolised. RCT can be broadly divided into four stages (Fig. 1). The first stage consists of efflux of cholesterol and phospholipids through ATP-binding cassette transporter A1 (ABCA1) present on the cell membrane to interact with the HDL precursor, apolipoprotein A1 (ApoA1) (Phillips, 2018). This interaction leads to the formation of a nascent disc-shaped HDL particle (dHDL). In the second stage, this disc is acted upon by lecithin-cholesterol acyltransferase (LCAT) which catalyses the esterification of free cholesterol (FC) contained within HDL to cholesteryl ester (CE). LCAT-mediated maturation of HDL forms spherical HDL (sHDL) with a lipid core primarily consisting of CE (Shih, Sligar & Schulten, 2009). In the third stage, while in plasma enroute to the liver, HDL is acted upon by several proteins that modulate its size and lipid composition. These proteins enable HDL to take up more cholesterol from peripheral cells, and an ensemble of such interactions between these proteins and mature HDL lead to HDL remodelling (Gursky, 2015). Finally, cholesterol-laden HDL unloads its cargo onto scavenger receptor, class B type 1 (SR-B1) expressed on hepatocyte membrane and is taken up by the liver. In the liver, cholesterol is either used to synthesise bile acids or is excreted along with the bile (Dikkers & Tietge, 2010).
HDL occupies a pivotal position in RCT, and dysregulation in its genesis/maturation is associated with dyslipidaemia manifesting as either hypertriglyceridemia [LCAT and CE transfer protein (CETP)-associated mutations] (Hovingh et al., 2005) and/or hypoalphalipoproteinaemia (Tangier disease caused by ABCA1 mutations) (Aiello et al., 2002), thereby significantly increasing the risk of atherosclerosis. HDL exerts a prominent anti-inflammatory effect by directly inhibiting the expression of adhesion molecules on the vascular wall (Cockerill et al., 1995), thereby reducing the inflammatory response prevalent in atherosclerotic cardiovascular disease (ASCVD). Furthermore, secretion of paraoxonase stimulated by HDL inhibits the oxidation of low-density lipoprotein (LDL) (Mackness et al., 2004). HDL has been speculated to prevent lesion progression by acting as a lipid acceptor in the vicinity of cholesterol-loaded foam cells (Cuchel & Rader, 2006).
Given the spectrum of cardioprotective effects of HDL, clinical efforts have focused on increasing plasma HDL levels (Superko, 2006). Despite being a promising approach, most interventions that raised HDL levels while simultaneously lowering LDL failed to deliver (Kingwell et al., 2014). This has shaken faith in the HDL hypothesis which correlates HDL levels with reduced risk of cardiovascular disease (CVD). Recently, genome-wide association studies and investigations involving families with hereditary HDL disorders have identified several factors that control HDL cholesterol levels (Weissglas-Volkov & Pajukanta, 2010;Brunham & Hayden, 2015). Furthermore, dynamic rates of RCT constituting synthesis and degradation of HDL could be clinically more relevant than steady-state HDL levels. This has stimulated interest in the RCT pathway and its key regulators. Insights into molecular mechanisms that regulate HDL synthesis, its remodelling, and catabolism could potentially identify therapeutic targets and diagnostic markers.
In this review, we revisit the distinct structure-function relationships of HDL with all the other components involved Deciphering structural aspects of RCT in the RCT pathway. We discuss how these stages are coordinated to maintain cholesterol homeostasis. We consider each of the four stages of the RCT pathway, including the protein-protein and protein-lipid interactions that drive this pathway while tracing the journey of HDL from its genesis to its final fate. Special emphasis is given to studies that explore the biophysical mechanisms of protein action. We also provide a comprehensive review of advances in this field to date and highlight potential future research directions.

II. STAGE 1: CHOLESTEROL EFFLUX AND DISCOIDAL HDL FORMATION
The genesis of HDL begins at the membrane interface, where cholesterol is translocated across the membrane in either an ATP-independent (aqueous diffusion) or ATPdependent (through ABCA1) manner. This efflux is impacted by cellular cholesterol status, transporter activity, and by the nature, concentration, and composition of the extracellular acceptors. This stage results in the formation of dHDL that is heterogeneous in size (diameter 8-12 nm), surface charge, and in the number of associated apolipoprotein molecules (Duong et al., 2006;Phillips, 2018). dHDL particles have a short half-life in plasma. In this section we discuss protein interactions that result in HDL genesis.
(1) Structure of ABCA1 ABCA1 plays an indispensable role in dHDL formation. Mutations characteristic of Tangier disease and familial hypoalphalipoproteinaemia have been shown to involve functional defects in lipid efflux by ABCA1 (Singaraja et al., 2006). ABCA1 in the plasma membrane self-associates as dimers or tetramers (Fig. 2) (Trompier et al., 2006). However, upon binding ApoA1, this dimer disassociates into its constitutive monomers (Nagata et al., 2013). A putative PDZ-binding motif at the C-terminus and the 2210-VFVNFA-2215 sequence have been implicated in dimer formation (Trompier et al., 2006). ABCA1 activity in the plasma membrane leads to lipid reorganisation, making FC more accessible to cholesterol oxidase. ABCA1 localises to both Fig. 1. A schematic representation of the reverse cholesterol transport (RCT) pathway. The first stage involves discoidal high-density lipoprotein (dHDL) formation by interaction between ATP-binding cassette transporter A1 (ABCA1) and non-lipidated apolipoprotein A1 (ApoA1). This is followed by conversion of cholesterol carried in dHDL to cholesteryl ester (CE) by lecithincholesterol acyltransferase (LCAT), maturing it to its spherical form (sHDL). sHDL is then acted upon by several proteins in plasma that alter its composition [cholesteryl ester transfer protein (CETP), phospholipid transfer protein (PLTP), hepatic lipase (HL), and endothelial lipase (EL)] and size (PLTP). Finally, CEs carried within high-density lipoprotein (HDL) are delivered to the liver through scavenger receptor, class B type 1 (SR-B1), where they are either utilised or excreted in bile with the help of heteromeric dimer composed of ATB-binding cassette transporter G5 and G8 (ABCG5/ABCG8). FC, free cholesterol; LDL, low-density lipoprotein; VLDL, very low-density lipoprotein. raft and non-raft domains; however, its activity affects overall lipid packing of the plasma membrane (Raducka-Jaszul et al., 2021;Sunidhi et al., 2022). ABCA1 consists of a single polypeptide chain that forms six domains: two transmembrane domains (TMDs), two nucleotide-binding domains (NBDs), and two large extracellular domains (ECDs). TMDs tether the protein in the membrane while providing a route to a tunnel enclosed within the ECDs. However, these two cavities are not continuous (Qian et al., 2017;Sun & Li, 2022). Movement of lipid substrate from the cavity within the TMDs to that within the ECDs requires a pronounced conformational change (Qian et al., 2017;Sun & Li, 2022). Recently, investigation of the ATP-bound ABCA1 structure revealed that the TMD and NBD regions are in close contact via salt bridges. This structure also revealed a conformational change in the ECD-TMD interface, causing ECDs to rotate by 30 (Sun & Li, 2022). By superimposing TMD1 of the ATP-bound and unbound states, TMD2 in the ATP-bound state was observed to shift towards the centre of the transporter, thereby narrowing the cavity between the two TMDs. Therefore, changes in the ECD-TMD interface created by movement of TMD helices upon ATP binding initiate the translocation of substrate from TMD to ECD (Sun & Li, 2022). Five aromatic residues (Phe273, Trp278, Trp458, Phe466, and Tyr482) make up a hydrophobic tunnel within the ECD, wherein an endogenous cholesterol molecule has also been found. This is hypothesised to be the exit tunnel for cholesterol. Of these, Tyr482 and Trp278 are essential for cholesterol binding. Additionally, Arg565, located at the ECD-TMD interface also plays a critical role in cholesterol efflux (Sun & Li, 2022). The illustration depicts the existing conundrum regarding the assembly of functional ABCA1 on the membrane: either dimer assembly takes place on the plasma membrane, or alternatively it may pre-assemble in the endoplasmic reticulum and then be transported to the membrane. HDL, high-density lipoprotein; LDL, low-density lipoprotein. Deciphering structural aspects of RCT ABCA1 in its resting state (nucleotide free) sequesters cholesterol/substrates into the cavity between TMDs in the cytosolic leaflets. ATP binding brings about closure of the NBDs, followed by the movement of TMDs towards the centre of the transporter and rotation of the ECD. Salt bridges stabilise this closed conformation, consequently pushing the substrate towards the ECD exit tunnel. Finally, ATPase activity causes NBDs to release ADP, reverting the transporter back to its resting state (Sun & Li, 2022). ABCA1 binds two molecules of ATP at Glu1058 and Glu2070 (Sun & Li, 2022. Classifying the mechanism of action of this transporter as simply an alternating access mechanism is difficult as the TMD end opens into a cavity within the ECD. Therefore, despite a conformational change upon NBD dimerisation, the substrate will not be effluxed through a pore. The fate of the lipid to be effluxed may be dependent on the interaction between ApoA1 and ABCA1 such that the presence of ApoA1 in vicinity of ABCA1 imposes an 'appropriate' conformation for lipid efflux. Additionally, the mechanism of substrate entry into the hydrophobic tunnel within the TMDs is still unexplored. Substrate entry may follow the 'credit-cardswipe' model, in which the substrate enters the tunnel laterally from the grooves of TM helices while its polar group remains shielded (Perez et al., 2015;Qian et al., 2017).
(2) Primary lipid acceptor: ApoA1 ApoA1 is a principal component of HDL constituting 70% of the total HDL proteins in plasma. It can stabilise all subclasses of HDL (formed as HDL matures or is remodelled) owing to its remarkable conformational adaptability. Its lipid-binding property is ascribed to its amphipathic α-helices as opposed to specific residues. This simple structural organisation of ApoA1 has been harnessed to design mimetic peptides (Wolska et al., 2021). ApoA1 contains a N-terminal globular domain composed of two 11-amino acid and eight 22-amino-acid tandem repeats (Segrest et al., 1994). These tandem repeats are separated by proline residues and make up the amphipathic α-helices (Segrest et al., 1994). Alterations in the sequence of these repeats or their repeat order affects lipid affinity and the molecular structure of HDL formed (Del Giudice et al., 2017). The C-terminal domain (residues 180-243) is unstructured and consists mostly of interspersed helices and random coils (Ray et al., 2018).
Helices in lipid-free ApoA1 unfold and refold over a very small timescale (seconds), contributing to its dynamism (Phillips, 2013;Ray et al., 2018). In the lipid-free state, the C-terminal domain (residues 180-243) sits on the N-terminal helical domains (H1 and H5) such that the two termini lie close to each other (Fig. 3A, B) (Melchior et al., 2017;Ray et al., 2018). The C-terminal domain plays a critical role in lipid-binding interaction with ABCA1 and its deletion dramatically decreases ApoA1 functionality (Chroni et al., 2007).
Mutation of residues 218-222 disrupts the interaction between ApoA1 and ABCA1 producing defective pre-β particles that fail to mature into α-HDL (Fotakis et al., 2013a); mutation of residues 225-230 results in the accumulation of immature pre-β particles that fail to activate LCAT (Fotakis et al., 2013b). The interactions of ApoA1 with other HDLremodelling proteins are non-specific, involving one or more epitopes that are formed as HDL particles take shape, grow, and mature. The decrease in functionality observed in deletion mutants may be due to a decline in conformational adaptability.
(3) Mechanism of ABCA1-mediated efflux and lipidation-induced changes in ApoA1 The nature of interaction between ApoA1 and ABCA1 and the mechanism of substrate delivery from ABCA1 to ApoA1 to generate nascent HDL has long been debated. Regardless of cell type, the binding of ApoA1 with ABCA1 is increased upon enhanced expression of ABCA1 (Lyssenko et al., 2013). This binding is achieved by direct contact between the two proteins ( Fig. 4A) and involves amphipathic helices of ApoA1 and the ECDs of ABCA1 (Fitzgerald et al., 2004). Elimination of positively charged lysine residues on ECDs of ABCA1 decreases ApoA1 binding (Nagao, Kimura & Ueda, 2012). Similarly, the role of extracellular loops of ABCA1 in binding with ApoA1 is illustrated by Tangier disease-causing ABCA1 mutants which cannot undergo ApoA1-stimulated cholesterol efflux owing to the absence of direct contact between ABCA1 and ApoA1 (Fitzgerald et al., 2002).
Given substantial evidence for contact between ABCA1 and ApoA1, one theory of nascent HDL formation states that ABCA1 directly loads the substrate onto a bound pool of ApoA1 molecules; the translocated phospholipids are temporarily stored in a reservoir formed by ECDs of ABCA1. Two models have been proposed. First, ABCA1 monomers constantly diffuse on the membrane surface. Once they sequester enough phospholipids, they undergo a conformational change leading to their dimerisation and cytoskeletonmediated stabilisation. This forms a binding site for nonlipidated ApoA1 dimers which, on accepting phospholipids, disassociate to form dHDL. ABCA1 dimers having lost their sequestered phospholipids disassociate into constitutive monomers and continue diffusing on the membrane surface (Nagata et al., 2013). The second model used a combination of computational and experimental approaches to conclude that floppase activity of ABCA1 moves phospholipids from the cytoplasmic to exoplasmic face of the cell (Segrest et al., 2015). This leads to increased surface density on the outer leaflet resulting in pleats, which form the interface for interaction with ApoA1 amphipathic helices.
These models both have shortcomings. The existence of a 'reservoir' implies a threshold for phospholipids that can be contained within the ABCA1 dimer, which should thus generate homogeneous dHDL particles; this is not observed in practice. A molecular dynamic simulation has suggested that lipid-free ApoA1 grows slowly by capturing phospholipids one at a time (Catte et al., 2006). The heterogeneity in dHDL composition in plasma in this case may be an outcome of variable time of interaction between ABCA1 and ApoA1.
The theory of indirect interaction between ApoA1 and ABCA1 is based on empirical evidence that approximately 10% of ApoA1 is bound to ABCA1 (low-capacity site), and the majority interacts with the adjacent plasma membrane (high-capacity site) (Phillips, 2014). Moreover, ApoA1 with its amphipathic helices (specifically, the two helices nearest to the C-terminus) has been shown to solubilise the membrane (Gillotte et al., 1999). Regions in the plasma membrane that are loosely packed and exhibit high lateral compressibility (such as the domain boundaries) are preferred sites for micro-solubilisation (Gillotte et al., 1999). The indirect theory does not disregard ApoA1-ABCA1 interaction completely, but it states that ApoA1 only binds at two sites. It first interacts with ABCA1 at the low-capacity binding site and stabilises ABCA1. This induces the translocase activity of ABCA1, causing net accumulation of phospholipid in the exofacial leaflet. Consequently, a membrane protrusion is formed which constitutes the site for ApoA1 binding (Wang et al., 2000) (Fig. 4B). Subsequently, the exovesiculated membrane is solubilised (rate-limiting step), and dHDL particles are released (Phillips, 2014).
The indirect theory relies on membrane protrusions formed by the movement of phospholipids from the inner to outer leaflet owing to ABCA1 activity. However, recent steered molecular dynamics has indicated that the phospholipid translocase activity imparted by the TMDs to ABCA1 removes phospholipids from the outer leaflet of the membrane (Segrest et al., 2022), which contradicts the indirect theory.
Lipidation through contact with ABCA1 changes the overall helical content of ApoA1 from 50% to 85% (Oda et al., 2013b;Ray et al., 2018). The earliest model of lipid binding by ApoA1 proposed a two-step mechanism; the C-terminal amphipathic helices mediate binding to the lipid surface, which is followed by the opening of the N-terminal helix bundle. The latter step enables ApoA1 to accommodate more lipids (Lagerstedt et al., 2007). The C-terminal domain further undergoes a conformational change from a random coil to an α-helix upon lipidation (Oda et al., 2013b). Deciphering structural aspects of RCT Interestingly, ABCA1 has been implicated in N-terminal unfolding of ApoA1 (Wang et al., 2013). ApoA1 mutants (L38G/K40G) that destabilised the N-terminal helix bundle showed an enhanced ability to form dHDL when incubated with ABCA1-expressing cells (Liu et al., 2019). This understanding of ApoA1-ABCA1 interaction has led to the development of short synthetic peptides that mimic the behaviour of apolipoproteins and promote ABCA1-dependent HDL formation (Islam et al., 2018). Additionally, HDL formed by these peptides can activate LCAT and associate with other HDL-remodelling enzymes. They can bind to oxidised lipids, and confer anti-inflammatory and antioxidant responses similar to endogenous HDL (Leman, 2015). Despite initial failures in their clinical development, further studies in this area may produce promising outcomes (Wolska et al., 2021).

(4) Formation of discoidal HDL
The interaction of ApoA1 with ABCA1 generates dHDL of heterogeneous size and chemical composition; however, these particles are poor substrates for subsequent lipidation by ABCA1 (Mulya et al., 2007). The structure of dHDL has eluded biophysicists for decades owing to its large size (160 kDa) and heterogeneity in shape and protein-lipid composition (Bibow et al., 2017). This is an important component of the puzzle, which will facilitate a better understanding of both the preceding and succeeding steps in RCT. Unfortunately, full-length lipidated ApoA1 (which forms the basis of dHDL) has been resistant to crystallisation. However, crystals of two truncated forms of ApoA1 and spectroscopic and biochemical techniques on reconstituted HDL (rHDL) particles in solution have enhanced our understanding of this intermediate form of HDL in plasma ( Fig. 3C-G) (Segrest et al., , 2012Mei & Atkinson, 2011).
Two predominant models exist for the orientation of dimeric ApoA1 around the surface of dHDL. According to the first 'picket fence' model, trans-membrane helices of ApoA1 are arranged perpendicular to the bilayer; the second model presumes them to run parallel to the surface of the bilayer (Jones et al., 2010;Gogonea, 2016).
How these models can accommodate changes in particle size upon the addition of more lipids is an interesting issue. Using reporter Trp fluorescence on rHDL particles, the region between residues 134-174 was found to be dynamic upon exposure to lipid. This domain was interpreted to form a 'hinge' on the ApoA1 molecule, which could buffer the surface area in response to changes in lipid composition of HDL (Bhat et al., 2005). The presence of a hinge domain has long been speculated, although its position has varied (residues 90-100 based on limited proteolysis; 99-143 based on immunoreactivity; and 121-165 in sHDL particles). Part of this dynamic region (residues 133-146) was further shown to form a looped belt (Gogonea, 2016), which could also facilitate interaction with LCAT (Martin et al., 2006). Additionally, the central domain of ApoA1 was proposed to wrap around the bilayer with its N-terminus looping back on itself. On particle expansion, this hairpin may straighten to accommodate more lipids, functioning like a 'belt buckle' (Bhat et al., 2005).
Since dHDL molecules contain two ApoA1 molecules, a functional model is needed to explain the formation of ApoA1 dimers. Fluorescence resonance energy transfer (FRET) and fluorescence lifetime measurements have suggested that ApoA1 monomers could exist in a variable-helix to helix registry with respect to each other in rHDL and sHDL, each of which may play distinct functional and regulatory roles He et al., 2019).
Therefore, ApoA1 might conform to multiple arrangements in a single dHDL particle. These orientations may be static and form during the generation of dHDL depending on the initial mode of interaction of ApoA1 monomers. Alternatively, these different registries may be in a dynamic equilibrium and interconvert as the particle matures and is remodelled. Specific registries may play regulatory roles. For instance, LCAT binds to rHDL particles in 5/5 and 5/2 helical registries but is only activated in the former case (Cooke et al., 2018). Accordingly, particle diameter, lipid composition, and other modulating factors, like the presence of ApoA2, could drive a shift in the helix registry. While such alternating registries could explain the dynamic functions of HDL, we still require evidence for the mechanism underlying these shifts. Mechanistically, the two molecules may simply slide past each other. Since the two monomers are held in place by a large number of salt bridges , this seems energetically unfeasible. Another, 'inch-worm' mechanism states that a small region of the protein might disassociate from its partner first and shift a small distance, simultaneously reorganising or redefining its associations (Cooke et al., 2018). Further studies are necessary to elucidate the exact mechanism involved. Several models have proposed various structures of dHDL; however, pertinent questions about dHDL generation remain unanswered. Is ApoA1 functional as a monomer, or does it require oligomerisation to act as a lipid acceptor? More importantly, are the ApoA1 molecules oligomerised prior to, during, or after lipidation? The structure of dHDL in solution shows two ApoA1 molecules wrapped around a small pool of lipids, and theories of expansion build upon the orientation and movement of these monomers; however, the spontaneity of this oligomerisation and the role of ABCA1 in this process (if any) are not well understood.
The heterogeneity observed in size and protein-lipid composition of dHDL particles is suggestive of a differential lipid efflux activity of ABCA1; however, regulating factors other than plasma membrane composition have not yet been identified. Lastly, pre-β-HDL or dHDL is believed to be a precursor HDL molecule that can accommodate more lipids upon expansion. Is it only destined to act as a lipid carrier? Trials involving pre-β-HDL infusion have failed to reverse coronary atherosclerosis (Sacks & Jensen, 2018). Other trials using pre-β-HDL are in progress (Didichenko et al., 2016;Sacks & Jensen, 2018). However, additional factors that activate or regulate the ability of dHDL to transport lipids may be present.

III. STAGE 2: LCAT-MEDIATED MATURATION OF DISCOIDAL HDL INTO SPHERICAL HDL
dHDL in plasma acts as a substrate for LCAT, a key enzyme in RCT that converts FC in HDL to (more hydrophobic) long-chained CEs. The increased hydrophobicity, accompanied by ApoA1 rearrangement, drives the conversion of dHDL to its more abundant and larger spherical form. This step increases the cholesterol-carrying capacity of HDL. Cholesterol, once esterified, is trapped inside the particle and cannot freely diffuse back (Jonas, 2000). In this section, we focus on esterification of cholesterol by LCAT leading to the maturation of HDL particles.
(1) Structure and activity of LCAT LCAT is a 416-amino acid protein present in plasma in its heavily glycosylated form. These LCAT glycans increase its solubility and prevent non-specific binding to cells (Jonas, 2000). LCAT primarily drives maturation of dHDL to sHDL (α-LCAT activity) following ApoA1-mediated activation. It also acts on cholesterol contained within ApoBcontaining particles such as LDL (β-LCAT activity), where it is activated by ApoE instead of ApoA1 (Zhao et al., 2005).
LCAT controls the intravascular turnover rate of cholesterol. Studies on patients with familial LCAT deficiency (FLD) show a lipid profile characteristic of CVD with high triglyceride (TG) and cholesterol levels (Ossoli et al., 2016). These studies have highlighted the role of LCAT in RCT and have provided evidence in favour of clinical therapies targeting/mimicking LCAT activity.
The first high-resolution (2.65 Å) crystal structure of LCAT showed three domains: a central domain that adopts an α/β hydrolase (ABH) fold, subdomain 1, and subdomain 2. The subdomains lie on top of the core hydrolase α/β domain (Piper et al., 2015). The ABH fold consists of the catalytic triad (made up of Ser181, Asp345, and His377) borne on loops with a conserved arrangement (Piper et al., 2015).
(2) ApoA1-mediated lid retraction and activation of LCAT LCAT binds to HDL and drives HDL maturation by catalysing transesterification of cholesterol to CE. The initial binding of LCAT to the HDL surface is mediated by the availability of lipid interface and electrostatic interactions only (Cho et al., 2001;Piper et al., 2015). This is followed by a specific interaction with ApoA1, which leads to LCAT activation. ApoA1 central domains, including residues 95-121 (Oda et al., 2013a;Gu et al., 2016), 134-145 (Martin et al., 2006), and 159-170 (Roosbeek et al., 2001;Wu et al., 2007) have been implicated in LCAT binding. Fig. 5 depicts various predicted LCAT-binding sites based on ApoA1 organisation around a dHDL molecule. Recent studies on the LCAT-HDL complex revealed that LCAT selectively interacts with the edge of dHDL through helix 4/6 regions of the double belt. This positional specificity allows LCAT to access acyl tails of lipids at the edge of the protein-lipid bilayer (Manthei et al., 2020).
Given that LCAT is larger in size than ApoA1, LCAT may associate with a hybrid epitope generated by ApoA1 oligomerisation instead of specific residues of a monomer (Cooke et al., 2018). That ApoA1 is an important structural determinant of the LCAT-cholesterol-phospholipid substrate complex is known, as LCAT tested on ApoA1-free liposomes shows no baseline activity (Gunawardane et al., 2016).
LCAT shuttles between an open and closed conformation. In the closed state, residues 226-236 of the lid region extend over the active site, hindering substrate and solvent entry (Manthei et al., 2017). The lid also blocks Leu64, Phe67, and Leu117 (of the interfacial domain) from accessing the membrane although the rest of the interfacial domain (subdomain 1) and N-terminus of LCAT are still available. LCAT activation due to interaction with ApoA1 leads to a rearrangement of the loop region between Cys50-Cys74 (subdomain 1) to flip out and insert into the membrane. This change in conformation to an 'open' state allows increased access to the active site (Piper et al., 2015). The anchorage of the lid region of LCAT to ApoA1 helices may result in the formation of a hydrophobic 'hood' that shields the active site of LCAT against solvents, allowing the enzyme to process hydrophobic substrates (Manthei et al., 2018;Laurenzi et al., 2021).
LCAT-mediated reaction requires binding of structurally diverse ligands, including phosphatidylcholine, acylated intermediates, and cholesterol, at the catalytic site. Therefore, the lid, through its dynamic conformations, may provide substrate specificity.
LCAT anchors itself to the surface of HDL by utilising hydrophobic residues in the interfacial domain followed by opening of the active site tunnel. Surprisingly, this interaction with membrane lipids is also allowed in its closed conformation (through its hydrophobic N-terminal residues) (Manthei et al., 2020); however, the lack of tunnel opening prevents access to the substrate. The membrane-anchoring hydrophobic amino acids of LCAT have been shown to attract adjacent cholesterol molecules (Casteleijn et al., 2018). Therefore, both ApoA1 and hydrophobic residues in the interfacial domain may play a role in transferring cholesterol to the active site.
(3) LCAT-mediated reaction on HDL surface After initial binding of LCAT to the surface of dHDL, ApoA1 facilitates migration of substrate (cholesterol) from the bilayer into the catalytic site of LCAT by forming an amphipathic tunnel (Jones et al., 2009a;Mei & Atkinson, 2011;Segrest et al., 2012;Gorshkova, Mei & Atkinson, 2018). LCAT combines both lipase and transferase activities. In the first half of the reaction, through its phospholipase A2-like activity, the sn-2 ester bond of phosphatidylcholine is cleaved to form an acyl intermediate and Fig. 6. Lecithin-cholesterol acyltransferase (LCAT)-mediated maturation of discoidal high-density lipoprotein (dHDL) to spherical high-density lipoprotein (sHDL). HDL maturation proceeds with LCAT binding to the dHDL surface through its interfacial domain. This initial binding is followed by a specific interaction with apolipoprotein A1 (ApoA1) that leads to changes in LCAT conformation (lid retraction) such that the active site of the enzyme is accessible to substrates. This is depicted in the figure by unhinging of the black loop (lid) leading to increased accessibility to the active site (between subdomains 1 and 2 shown in green and blue, respectively). The LCAT-mediated reaction proceeds with the conversion of cholesterol to cholesteryl ester (CE), which moves to the core of the molecule. This leads to a change in shape of the complex from a disc to a sphere. Conformational constraints caused by changes in helical registries and binding of ApoA1, finally lead to inhibition of the LCAT reaction and its disassociation.
Biological Reviews 98 (2023)  Deciphering structural aspects of RCT lysolecithin (rate-limiting step) (Jonas, 2000). Asp345 and His377 activate Ser181 which cleaves the sn-2 ester bond of lecithin. The tetrahedral intermediate, stabilised by Phe103 and Leu182 in the vicinity, forms lysolecithin and a fatty acyl intermediate. This fatty acyl moiety, instead of being directly transferred to cholesterol, is transacylated at the sulphur atom of a cysteine. This cysteine thioester subsequently transfers its fatty acyl group to cholesterol, forming CE while regenerating the non-acylated enzyme (Zhou et al., 1991). Fig. 6 provides a schematic summary of the LCAT-mediated reaction that matures dHDL to sHDL.
Since cholesterol esterification in HDL is linked to an increase in its cholesterol-carrying capacity, induction of LCAT expression/activity may enhance the rate of RCT in patients with acute coronary syndrome. However, injectable recombinant LCAT can only enhance cholesterol efflux synergistically (Yang et al., 2022). Alternatively, modifying LCAT activity by inducing Cys31Tyr mutation or using small molecular activators, such as ApoA1 mimetics, has shown promising results (Yang et al., 2022). More recently, a dramatic shift has taken place towards a functional HDL hypothesis that emphasises the effect of chronic inflammatory stress (specifically in ASCVD) that can change the structure and composition of HDL, rendering it dysfunctional. Dysfunctional HDL characterised by oxidised ApoA1 (Martínez-L opez et al., 2019) cannot activate LCAT and has impaired cholesterol acceptance (Chiesa & Charakida, 2019). This may be harnessed to develop a biomarker for distinguishing patients with a proinflammatory HDL phenotype such that their treatment options can be improved.
(4) Maturation of discoidal HDL into spherical HDL sHDL is the most abundant form of circulating HDL and has the highest amount of CE. In plasma, sHDL constitutes a heterogenous mixture of particles with 8.8-11 nm diameter, associated with 2-7 molecules of ApoA1 and a variable amount of packed CE and TG (Huang et al., 2011).
Conversion of discoidal to spherical HDL takes place due to LCAT activity leading to the accumulation of neutral CE, which intercalates between bilayer leaflets. Because of the increased load of neutral lipid, amphipathic helices of ApoA1 might spread out across the sphere, extricating proteinprotein contacts observed in dHDL. However, the overall secondary structure and global arrangement of ApoA1 is the same in both dHDL and sHDL (Segrest, Jones & Catte, 2013;Gogonea, 2016). Residues 115-158 have been proposed to be sensitive to the surface-packing density of ApoA1, thereby allowing the protein to accommodate varying amounts of lipids (Gogonea, 2016).
During particle maturation, two ApoA1 strands remain in a 'double-belt' configuration while the proteins dynamically bend into and out of the saddle positions (Shih et al., 2009). Interestingly, the N-and C-terminal helices of ApoA1 strands are more flexible and mobile than those of the central region (Shih et al., 2009). These regions are also sensitive to phospholipid concentrations, triggering fusion/remodelling of HDL particles (Jones et al., 2009b(Jones et al., , 2011. This flexibility forms a natural hinge between the terminal region and central domain, allowing the incorporation of a third ApoA1 molecule (Jones et al., 2009b;Shih et al., 2009).
According to the trefoil model, three ApoA1 chains are arranged along the surface such that they subtend an angle of 120 to each other, dividing the entire sphere into three equal portions (Gogonea, 2016). This model has been extended to sHDL particles containing more than three ApoA1 molecules such that ApoA1s twist around the sphere causing unfolding, which can be utilised during the increment of particle size (Huang et al., 2011). During sHDL maturation, a growing CE core and increased exposure of the lipid surface to solvent, thermodynamically destabilises sHDL (Wu et al., 2009). This is rescued by interaction with the third ApoA1 monomer to form a stable sHDL particle.
Simulation of sHDL using the full lipid composition in human plasma indicates that cholesterol contained within the inner core interacts with the aromatic residues of ApoA1 through its sterol ring (Vuorela et al., 2010). This interaction contributes to ApoA1 rigidity and stabilises sHDL (Catte et al., 2008). This may also have a significant impact on the availability of sterols to other remodelling proteins or during cholesterol delivery to the liver (Vuorela et al., 2010).
Our understanding of the formation of sHDL is in its early stages. While several models have explored the dynamics of ApoA1 reorganisation upon particle expansion, fundamental questions remain. How does sHDL accommodate lipids effluxed by ATP binding cassette subfamily G member 2 (ABCG2)/SR-B1? How are these absorbed into the core? Once esterified and partitioned into the core, how do the different remodelling proteins, such as cholesteryl ester transfer protein (CETP), gain access to these lipids? sHDL fractions are heterogenous in size, composition, and number of ApoA1 particles. Does this heterogeneity reflect the dynamics of lipid balance/imbalance? Furthermore, HDL was shown to be secreted in all the four sized subfractions, and these particles circulated in plasma within their secreted size range for 1-4 days (Mendivil et al., 2016). This has challenged the size expansion hypothesis, which states that sHDL accumulates neutral lipids and grows (Sacks & Jensen, 2018). It is possible that LCAT activity may not affect the size of HDL at all, but simply changes its shape such that these particles are not cleared rapidly (Mendivil et al., 2016). Whether these HDL subfractions play specific functional roles or have different prognostic efficacy in chronic heart disease (CHD) remains elusive. Additionally, the diversity in size and composition may serve as an invitation to other HDL-binding proteins, particularly the remodelling proteins, which may lead to fusion or shedding of ApoA1 particles (Sacks & Jensen, 2018).

IV. STAGE 3: HDL REMODELLING
HDL interacts with several proteins that impact its structure and metabolic turnover. This generates a dynamic mixture Biological Reviews 98 (2023)  of distinct subfractions varying in size, shape, and proteinlipid ratio. Based on density, HDL can be partitioned into two subfractions: HDL2 comprising large HDL particles (8.8-12.9 nm, density 1.063-1.125 g/ml) and HDL3 approximated by the sum of small and medium HDL particles (7.2-8.8 nm, density 1.125-1.210 g/ml) (Rothblat & Phillips, 2010). The functional importance, if any, of these subfractions remains to be characterised (Sacks & Jensen, 2018). Moreover, HDL can be subdivided into several other minor subspecies based on presence of a single protein. These species constitute approximately 10% or less of total ApoA1-HDL concentration. These include those associated with other apolipoproteins, such as ApoA4, ApoC1, ApoC2, ApoC3, ApoE, ApoL1, and ApoJ, or other immunomodulatory proteins (Sacks et al., 2020). This section considers the ensemble of proteins that interact with mature HDL, effectively remodelling it into an efficient carrier of cholesterol.
(1) HDL remodelling by CETP CETP transfers neutral lipids (CE and TG) between HDL, LDL, and very low-density lipoprotein (VLDL). It can facilitate bidirectional transfer of the same lipid (homo-exchange) or transfer TG in exchange for CE (hetero-exchange) (Zhang et al., 2012). CETP reduces the size of HDL particles, enhancing their stability (Rye, Hime & Barter, 1995). Additionally, CE removal from HDL induces CE formation through LCAT reaction (Oliveira et al., 1997).
CETP plays a pro-atherogenic role by exchanging CE between HDL and LDL, leading to its deposition in peripheral tissues. Moreover, CETP-deficient humans present with increased plasma HDL levels (Koizumi et al., 1985). Several therapies have been designed to target this step of the pathway (Tall & Rader, 2018). Despite the anti-atherogenic potential of CETP inhibition, clinical trials have yielded disappointing results, with increased CVD incidence in patients undergoing CETP inhibition (Tall & Rader, 2018). This has led to renewed interest in understanding the mechanism of lipid transfer by CETP. It is now believed that CETP inhibition affects CVD due to its LDL-lowering effects rather than its HDL-elevating effects (Sacks & Jensen, 2018). CETP activity is stimulated by several factors, such as surface availability of lipids in lipoprotein, overall particle size and charge, and presence of ApoC1 (Bruce et al., 1995;Morton & Greene, 2003a,b;Dumont et al., 2005). Interestingly, ApoC1-bound HDL is strongly associated with a reduced risk of CVD (Sacks et al., 2020). Although no direct interaction of apolipoproteins has been reported, lipoproteins with a high number of apolipoproteins bind more CETP molecules, entirely due to their size . CETP activity is also increased upon lipase activity due to the accumulation of lipolytic products (Sammett & Tall, 1985).
CETP is a boomerang-shaped molecule with three structural domains: a N barrel (at the N-terminus) that opens to a central linker region (residues 240-259), followed by a C barrel (at the C-terminus). Each barrel contains a twisted β-sheet and two helices while the central linker region is arranged into six antiparallel β-strands (Qiu et al., 2007). The barrels at each terminus open into a 60-Å-long tunnel wherein the N-terminal opening is narrower than the C-terminal opening (Qiu et al., 2007).
In the crystal structure, two CE molecules are buried inside the tunnel while the termini of the barrels close to the neck each are plugged by a phospholipid. These phospholipids block solvent entry into the tunnel, maintaining structural integrity and plasticity of CETP (Qiu et al., 2007;Revanasiddappa, Sankar & Senapati, 2018).
(2) Mechanism of lipid transfer by CETP Two models describe the mechanism of neutral lipid transfer through CETP. (i) In the shuttle mechanism, CETP shuttles lipids between donor and acceptor lipoproteins (Barter & Jones, 1980). (ii) In the ternary complex mechanism, CETP forms a bridge between two lipoproteins, thereby facilitating lipid transfer (Ihm et al., 1982). Both types of lipid exchange may require a cycle of unloading and reloading of phospholipid plugs when CETP binds a lipoprotein, however, most of the details of this process remain unclear.
The non-specific nature of the tunnel, combined with the observation that neck-blocked CETP mutants show reduced activity, suggests that lipid traverses the entire tunnel during its transfer. Therefore, CETP may admit lipid from one opening while using the other opening for deposition (Qiu et al., 2007).
Furthermore, CETP interacts with LDL through its Cterminus while the N-terminus engages with HDL, although this functional polarity might not be lipoprotein-selective (Zhang et al., 2012;Charles & Kane, 2012). Trp105, Trp106, and Trp162 at the N-terminus of CETP play an important role in anchoring and penetration of CETP into HDL (Cilpa-Karhu, Jauhiainen & Riekkola, 2015). Furthermore, binding studies with CETP inhibitors (torcetrapib and anacetrapib) have indicated that one end of CETP, when bound to the first lipoprotein, can trigger a conformational change at the other end, facilitating binding with the second lipoprotein (Zhang et al., 2017).
Once CETP binds to the HDL surface, CE molecules are pulled to the surface where the N-terminus of CETP interacts with HDL, forming a hydrophobic patch (Cilpa-Karhu et al., 2015). Hydrophobic interactions between CE fatty tails and CETP residues (Phe115 and Phe167) at the beginning of the tunnel and the 'peristaltic' movement of CETP drive lipid transfer through the tunnel (Lei et al., 2016;Dixit, Ahsan & Senapati, 2019). Deciphering structural aspects of RCT facilitates remodelling by modulating the size and composition of HDL. Treatment of cholesterol-laden fibroblasts with exogenous PLTP enhances cholesterol and phospholipid efflux (Wolfbauer, Albers & Oram, 1999). PLTP may work with microsomal triglyceride transfer protein in coordinating the assembly and secretion of ApoB-containing lipoprotein particles rich in TGs (Luo et al., 2010). More importantly, PLTP and CETP interact with each other. Purified PLTP in presence of CETP, enhances CE transfer from HDL3 to VLDL (Jiang, 2018).
Although a structure of PLTP is not available, homology models (lacking C-terminal residues 464-494) have been constructed, based on its sequence homology with bacterial permeability-increasing protein. The modelled structure predicts a boomerang-shaped protein exactly like CETP, with two barrel-type structures on each end and a central β-sheet connecting them.
Several models have been proposed to define the sequence of events involving PLTP-mediated HDL remodelling. Binding of PLTP to the HDL surface and subsequent phospholipid transfer leads to dissociation of ApoA1, forming an unstable HDL that fuses to form a larger stable particle (Lusa et al., 1996). This large unstable intermediate again dissociates its excess surface lipids with ApoA1 spontaneously, to release pre-β-HDL. The remaining particle either stabilises or further dissociates into smaller HDL particles (Settasatian et al., 2001). TG-rich rHDL particles are remodelled more rapidly than CE-rich rHDL particles by PLTP (Settasatian et al., 2001). Therefore, CETP and PLTP could act in concert to remodel larger unstable HDL into smaller particles, improving their lipid-carrying capacity while simultaneously generating pre-β-HDL that can further stimulate cholesterol efflux through ABCA1.

(4) HDL remodelling by hepatic and endothelial lipases
Hepatic lipase (HL) and endothelial lipase (EL) belong to the TG lipase family. They modulate HDL metabolism via their differing hydrolytic activities, substrate specificity, and tissuespecific gene expression. EL is mostly synthesised by vascular endothelial cells, thyroid epithelial cells, and hepatocytes. It possesses phospholipase activity with a very little triglyceridase activity that modulates HDL composition selectively (McCoy et al., 2002), whereas HL exerts equal phospholipase and triglyceridase activities on all lipoproteins (Fan et al., 1994). These lipases reduce HDL size (Annema & Tietge, 2012).
The role of lipases in lipoprotein metabolism has been established by several clinical and genetic knockout studies in mice. Patients with HL deficiency present with increased cholesterol and TG levels in plasma owing to the accumulation of VLDL, chylomicron remnants, TG-rich LDL, and HDL (Breckenridge et al., 1982). Moreover, complete HL deficiency is associated with a reduced catabolic rate of ApoA1 (Ruel et al., 2004). In addition to its hydrolytic action, HL also acts as a ligand that facilitates the uptake of lipoproteins possibly by interacting with SR-B1 (Bamberger, Glick & Rothblat, 1983).
EL overexpression reportedly increases ABCA1-mediated cholesterol efflux (Qiu & Hill, 2009), accelerates renal ApoA1 catabolism, and increases hepatic cholesterol uptake through SR-B1 . Inhibition of EL activity in HL-deficient mice further exacerbates HDL retention in plasma, indicative of the additive role of lipases in HDL metabolism (Brown et al., 2010).
Proteins belonging to the lipase family have a heparinbinding domain (Lys297, Lys298, Arg300, Lys 337, Lys436, and Arg 443 in the case of HL) and possess an α/β-hydrolase fold, which bring the key active site residues (Ser-Asp-His catalytic triad) together. They contain a lid element that provides access to the catalytic site, an inter-facial domain that aids binding to the lipid surface (Khan et al., 2017), and additional cofactor-binding sites for enzymatic activation. Cofactors for HL and EL have not been identified to date.
Structures for HL and EL are not available; however, a homology model has been constructed for both proteins based on their sequence identity (33% and 35%, respectively) with pancreatic lipase. Although HL and EL belong to the same subfamily and share significant homology, their binding pockets are quite different. Modelling studies have identified Ser168, Asp194, and His279 as the catalytic triad in vertebrate HL, a predicted lid region formed by residues 255-276, and a hinge region Arg332-Ser333-Lys334-Ser335. In EL, the hinge region is predicted to be Arg327-Asn328-Lys329-Arg330 (Holmes, VandeBerg & Cox, 2011).
Hepatic and endothelial lipases have a significant impact on lipoprotein metabolism. Therefore, obtaining molecular structural information to elucidate how these lipases exert their effects will be important. Because of their diverse range of functions in maintaining lipoprotein homeostasis and involvement in the pathophysiology of hyperlipidaemia, these proteins can serve as attractive biomarkers and potential therapeutic targets. In fact, the presence of a common polymorphism in the promoter region (−514 C to T) of the HL that modulates hepatic lipase activity has been proposed as a biomarker for cardiometabolic parameters and risk factors of CVD (Zambon et al., 2003). The presence of a T allele at this locus increases the risk of atherosclerosis marginally, whereas a C allele is associated with an increased thickness of carotid intima-media and unstable plaques. Patients with a CC genotype clinically benefit from intensive lipidlowering therapy (Zambon et al., 2003).

(5) Lipid efflux by ABCG1/SR-B1
While ABCA1 supplies phospholipids and cholesterol to lipid-poor ApoA1, it does not supply them to mature HDL (HDL2 and HDL3 particles) (Thuahnai et al., 2004). Two additional transporters, ABCG1 (also ABCG4) (Wang et al., 2004) and SR-B1, were subsequently identified that work in tandem with ABCA1 to mediate lipid efflux to HDL (Thuahnai et al., 2004). The pathophysiological role of ABCG1 has been successively evaluated in ABCG1 knockout mouse models. On administration of a diet high in fat and cholesterol, these mice showed no induction in plasma lipids but had a substantial accumulation of neutral lipids and phospholipids in hepatocytes and macrophages. Overexpression of human ABCG1 protected these tissues from high-fat diet-induced lipid accumulation (Kennedy et al., 2005). Moreover, patients with type 2 diabetes presented with reduced ABCG1 expression and cholesterol accumulation, thereby highlighting the role of ABCG1 in insulin resistance and CVD (Mauldin et al., 2006).
The crystal structure of ABCG1 has been found to have the same fold present in other ABCG subfamily members (Skarda, Kowal & Locher, 2021). The TMDs consist of six TM helices and three extracellular loops, of which, TM2 and TM5a line the path used for substrate translocation. Functional ABCG1 exists as a homodimer, but the two monomers are not connected by intermolecular disulphide bonds (Skarda et al., 2021).
The TMDs (TM2 and TM5 of each G1 monomer) enclose two cavities, a larger cavity that opens at the cytosolic end (substrate-binding cavity) and a smaller cavity that faces the extracellular milieu. Based on the alternating access and release mechanism, this conformation is inward facing (open). The larger cavity, by virtue of its shape, excludes large globular molecules and only allows binding of flat polycyclic molecules such as cholesterol. Residues facing the cavity are aliphatic or aromatic. These include Phe455, Met459, and Leu463 in TM2 and Phe555, Pro558, Val559, and Ile562 in TM5 of the other G1 molecule. Additionally, Glu551 located at the cytoplasmic end of TM5a, is also part of the substrate-binding pocket and interacts with the hydroxyl group of cholesterol (Sharpe et al., 2015;Skarda et al., 2021).
Although both ABCA1 and ABCG1 serve as lipid efflux transporters in RCT and share a similar TMD fold, ABCG1 lacks the extended extracellular domain that is characteristic of ABCA1. The role of ApoA1 in lipid efflux and nascent HDL generation through ABCA1 is well characterised; however, absence of such an interaction pinpoints efflux selectivity towards only mature HDL particles. These interactions may not be specific.
Unlike ABCA1, which is localised to non-raft domains, ABCG1 and its highly homologous member, ABCG4 localise to raft domains and disrupt them. Since newly synthesised or LDL-derived cholesterol localises to raft domains, ABCG1 might play a role in 'on-demand' removal of cholesterol from these meso-domains, while ABCA1 functions in the 'housekeeping' removal of cholesterol in macrophages. ABCG1 is palmitoylated at Cys25, 150, 311, 390, and 402 residues. Inhibiting the palmitoylation of Cys311 significantly affects its membrane translocation .
HDL constitutes a complex constellation of proteins and phospholipids with diverse functional roles. It undergoes dynamic changes in composition that influence local inflammatory and oxidative events. HDL remodelling specifically through CETP is pro-atherogenic and has been targeted for therapeutic intervention. However, consistent failures in this regard necessitate re-evaluation of CETP function and the hetero-exchange it facilitates. Moreover, the interplay among remodelling proteins and their combined effect on the lipoprotein population may be significant. Future trials involving CETP inhibition may benefit by studying HDLassociated metabolic profiles in addition to free and esterified cholesterol contained within HDL and LDL (Singh et al., 2021). Fig. 7 summarises the interplay amongst proteins involved in HDL remodelling. Recently, dHDL was observed to fuse with circulating sHDL to form a heterogenous population of enlarged HDL particles, lipid-poor ApoA1 particles, and lipid-free ApoA1 (Navdaev et al., 2020). HDL-remodelling proteins, due to their impact on HDL stability, size, and composition, may positively influence such dHDL-sHDL fusions. Their combined actions may influence HDL-SR-B1 binding or ApoA1 renal clearance (Lamarche et al., 1999). Moreover, since SR-B1 is involved in a bi-directional flux of cholesterol in and out of the cells, it may be important to distinguish between HDL particles that facilitate cholesterol influx or efflux. HDL remodelling may be responsible for such a distinction.

V. STAGE 4: CHOLESTEROL DELIVERY TO THE LIVER, ITS METABOLISM, AND EXCRETION
Cholesterol delivery to the liver serves two purposes: (i) CE can be metabolised and excreted; and (ii) depletion of CE from HDL allows HDL to accept more CE, driving RCT forward (Shen, Azhar & Kraemer, 2018). The rate of RCT depends on both the efflux of cholesterol from peripheral tissues to HDL and its removal from the plasma component through hepatic uptake by SR-B1 (Wang et al., 1996). In this section, we consider the mechanisms involved in cholesterol delivery to the liver.
(1) Structure of SR-B1 and mechanism of cholesterol uptake Studies on synthetic HDL particles containing [ 3 H] cholesteryl ether and [ 14 C] sucrose octaoleate showed that only cholesteryl ether was taken up by HepG2 cells. This implied a non-endocytic mechanism of HDL uptake. SR-B1 was later identified as the receptor that facilitated CE exchange. SR-B1 belongs to the family of class B scavenger receptors, which also includes the CD36 family and lysosomal integral membrane protein (LIMP)-2. It is well conserved across several species and ubiquitously expressed in mammals. SR-B1 mediates both cholesterol efflux from cells and CE uptake from HDL into cells (Thuahnai et al., 2004).
Mice with a targeted null mutation (−/−) in SR-B1 show a dramatic increase in cholesterol levels and enlarged lipoproteins in plasma (Varban et al., 1998). Moreover, decreased hepatic SR-B1 expression leads to reduced CE uptake by the liver (Varban et al., 1998) Deciphering structural aspects of RCT SR-B1 in the liver of mice with early and advanced atherosclerosis decreases lesion size (Kozarsky et al., 2000).
No SR-B1 structure is available to date; however, based on its sequence homology with LIMP-2, a homology model predicts two short TMDs (12-32 and 441-461 residues). Residues 33-440 form a large ectodomain consisting of an anti-parallel β-barrel core with short α-helical segments that form a bundle on top. The ectodomain joins the TMD by a small linker region (Neculai et al., 2013). Accordingly, the apex region of SR-B1 is made up of positively charged residues. Electrostatic interactions between these residues and HDL aid in a firm attachment to the receptor (Neculai et al., 2013). The β-barrel acts as a hydrophobic channel through which lipids are discharged. This hydrophobic tunnel is stabilised by a disulphide bridge between Cys274-Cys329 and Cys321-Cys323 (Neculai et al., 2013).
Additionally, Cys384 located in the tunnel lumen has been implicated in CE uptake by SR-B1 (Nieland et al., 2008).
SR-B1 is a heavily glycosylated protein. Glycosylation at Asn108 and Asn173 are important for its surface expression and lipid uptake, but not for HDL binding. It exists as dimers on the cell surface although higher-order oligomeric forms have also been reported (Reaven et al., 2006;Hu et al., 2011). A conserved glycine motif G15-X-X-G18-X-X-X-A-X-X-G25 in its N-terminal TMD is required for its dimerisation and lipid uptake activity (Gaidukov et al., 2011) while the C-terminal TMD interacts with the plasma membrane, acting as a cholesterol sensor. In addition to the plasma membrane, SR-B1 localises in early and late endosomes, suggesting its possible role in trafficking cholesterol out of these compartments (Ahras, Naing & McPherson, 2008). Two SR-B1 point mutations, S112F Fig. 7. High-density lipoprotein (HDL) remodelling in plasma. Discoidal HDL (dHDL) acquires cholesterol from ATP-binding cassette transporter A1 (ABCG1) and scavenger receptor, class B type 1 (SR-B1) expressed on peripheral tissues. HDL is continuously acted upon by lecithin-cholesterol acyltransferase (LCAT) driving its maturation. An increased cholesteryl ester (CE) core in HDL leads to expansion in its size. Furthermore, it is acted upon by cholesteryl ester transfer protein (CETP) which transfers CE from HDL to very low-density lipoprotein/low density lipoprotein (VLDL/LDL) in exchange for triglycerides (TGs). This TG-rich HDL is broken down by phospholipid transfer protein (PLTP) into smaller HDL particles (HDL3), releasing preβ-HDL particles. These particles can associate with ABCG1 and SR-B1 again to form larger HDL particles (HDL2). Hepatic lipase (HL) and endothelial lipase (EL) act on HDL in plasma, changing its lipid composition. and T175A in individuals with high plasma HDL have been found to be associated with defective HDL binding to SR-B1, influencing both its cholesterol efflux and CE uptake functions (Chadwick & Sahoo, 2012).
While SR-B1 has long been accepted as an HDL receptor for CE uptake, the mechanism remains elusive. A simplistic two-step model may involve HDL binding to the receptor followed by CE internalisation (Thuahnai et al., 2004). CE moves down its concentration gradient via a non-aqueous pathway (Reaven, Tsai & Azhar, 1996). Based on cellular uptake of apolipoprotein and CE, SR-B1 has been proposed to 'nibble' on HDL-CE, gradually decreasing its size, with subsequent release of lipid-free ApoA1 as opposed to removing CE all at once ('gobbling') (Gillard et al., 2017). This first step requires binding of HDL and appropriate organisation of bound HDL to extracellular domains of the receptor. This interaction prevents fusion of the lipid donor with the plasma membrane (Thuahnai et al., 2001). Therefore, HDL binding is highly specific. Competitive binding assays have indicated that SR-B1 has either two independent binding sites or one site that exhibits negative cooperativity (Thuahnai et al., 2003;Nieland et al., 2011).
(2) Fate of cholesterol in the liver The majority of cholesterol brought in as CE by HDL undergoes biliary secretion. The hepatocytes in the canalicular membrane (site of biliary secretion) are polarised such that HDL receptors are located on the basolateral side, while biliary cholesterol secretion occurs through the apical surface. Therefore, internalised cholesterol traverses to the other end of the cell. The detailed mechanism of intracellular transport of cholesterol in hepatocytes is not completely understood. However, Niemann-Pick type C protein 1 (NPC1) is an important mediator as it moves FC from the late endosomal compartment to the cytosol (Temel et al., 2007).
FC is utilised in synthesising bile acids that are ultimately secreted into the bile duct. A variety of transporters belonging to the ABC family take part in this process. ATPase class I type 8B member 1 (ATP8B1) flips phosphatidylserine from the outer to inner leaflet to maintain canalicular membrane integrity. ABCB4 translocates phospholipids to the outer membrane so that they can be added to cholesterol to form mixed micelles (Dikkers & Tietge, 2010). However, the major cholesterol transport out of hepatocyte is mediated by an ABCG5/ABCG8 heterodimer. The crucial role of this obligate heterodimer has been determined in studies with ABCG5 −/− and ABCG8 −/− mice, wherein targeted disruption of these genes led to an extremely low level of biliary cholesterol (sitosterolaemia) (Yu et al., 2002). An inward-facing and nucleotide-unbound structure is available for the ABCG5/ ABCG8 heterodimer, which consists of vestibules that are flanked by TM helices on the two ends and helices of the ECD on top. Based on their conservation across eukaryotes, these cavities may serve as sterol entryways into the heterodimer core (Lee et al., 2016). Nucleotide binding to ABCG5/ ABCG8 leads to an inward movement of NBDs, which is relayed onto TMDs (TMD polar relay) leading to a conformational switch to the pore-open form (Lee et al., 2016). Additionally, SR-B1 also mediates biliary cholesterol secretion independent of ABCG5/ABCG8 but dependent on phospholipid translocase activity of ABCB4 . Utilisation of cholesterol in bile acid synthesis and secretion of excess into bile, results in its removal through faeces. Fig. 8 provides a schematic summary of the removal of excess cholesterol delivered by HDL and its subsequent excretion into the bile duct.
(3) HDL-ApoA1 catabolism The levels of ApoA1 in plasma are continuously renewed through its synthesis-degradation cycle. Rates of human ApoA1 turnover, however, are variable. The majority of ApoA1 is catabolised by the liver while only one third is catabolised by the kidneys (Glass et al., 1983). Spontaneous exchange of ApoA1 and HDL remodelling constantly generates lipid-poor ApoA1 particles. While these particles can associate with ABCA1 to generate dHDL, they are also filtered in the glomerulus and subsequently catabolised in cells lining the proximal renal tubule (Rader, 2006). Since mature HDL particles are too large to be filtered out, only nonlipidated ApoA1 follows this route. Cubilin, in kidney proximal cells, along with megalin, facilitates endocytosis of poorly lipidated ApoA1 (Kozyraki et al., 1999;Hammad et al., 2000). However, cubilin deficiency does not impact ApoA1 or HDL levels in plasma, suggesting an alternative catabolic pathway (Christensen & Gburek, 2004).
The mechanism of HDL uptake by the liver and subsequent apolipoprotein degradation are poorly understood. TG enrichment of HDL during remodelling is implicated in increased HDL uptake by the liver (Lamarche et al., 1999). Furthermore, altered HDL composition as a result of enhanced intravascular remodelling leads to more rapid clearance. Therefore, the rate of HDL clearance, rather than just its production, is an important determinant of plasma HDL concentrations.

VI. REVERSE CHOLESTEROL TRANSPORT: A DISEASE PERSPECTIVE
Lipoprotein metabolism influences a cluster of complex disorders called metabolic syndromes. Atherogenic dyslipidaemia is a noteworthy phenotype of such disorders and is characterised by elevated levels of TGs with small LDL particles and decreased levels of HDL particles in plasma. Therefore, factors affecting lipoprotein synthesis, their compositional remodelling, and catabolism can contribute indirectly to atherosclerotic pathogenesis. This section summarises the nonsynonymous single nucleotide polymorphisms (nsSNPs) that impact RCT. Fig. 9 depicts the pathogenic mutations obtained from the single nucleotide polymorphism database (dbSNP) Biological Reviews 98 (2023)  Deciphering structural aspects of RCT (Sherry et al., 2001) of the three key players of RCT with roles in HDL formation and maturation.
The lipid efflux transporter ABCA1 that is involved in nascent HDL generation was linked to Tangier disease two decades ago. Since then, multiple studies have identified several SNPs in ABCA1, which are associated with hyperalphalipoproteinaemia, HDL deficiency, and other related disorders. Patients with Tangier disease lack the ability to release phospholipids and FC to apolipoproteins (Rust et al., 1999). Excess cholesteryl esters in reticuloendothelial tissues result in splenomegaly, enlarged tonsils, and peripheral neuropathy. Studies on ABCA1 mutant mice also report diminished LCAT activity (Aiello et al., 2002), a complete lack of α-migrating HDL particles, and pre-β HDL particles rich in TGs with an altered phospholipid distribution (Aiello et al., 2002). ABCA1 mutations have been broadly classified into four groups. The first category consists of maturationdefective mutants that have impaired cellular trafficking and consequently, negligent lipid efflux activity. These mutations are mostly localised on ECD1 and ECD2 (R587W, Q597R, and S1506L). The second category is that of lipid translocation-defective mutants (such as W590S in ECD1) that correctly localise to the membrane but have impaired membrane-remodelling activity and therefore, diminished Cholesteryl ester (CE) contained within high-density lipoprotein (HDL) is transported to the cytosol either directly through interaction with scavenger receptor, class B type 1 (SR-B1) or through an SR-B1-mediated endocytic mechanism. Niemann-Pick type C protein (NPC1) is implicated in cholesterol transfer from the endosome to the cytoplasm of hepatocytes. Free cholesterol (FC) in the cytoplasm resulting from the action of acyl-coenzyme A:cholesterol acyltransferase (ACAT) is utilised by cytochrome P450 (CYP450) to synthesise bile acid. The excess cholesterol is excreted into the bile canaliculi as mixed micelles. ATP-binding cassette transporter A1 (ABCB4) translocates phospholipids to the outer membrane aiding in assembly of micelles. ATP-binding cassette transporter B11 (ABCB11) pumps in bile salts while a heterodimer of ATP-binding cassette transporters G5 and G8 (ABCG5/ABCG8) and SR-B1 are implicated in selective excretion of cholesterol into the bile canaliculi. ATPase phospholipid transporting 8B1 (ATP8B1) transports phosphatidylserine from the outer to the inner leaflet to maintain membrane integrity.  (Wang & Smith, 2014). Two SNPs (G1050V and S1067C) in the NBD domain are predicted to disrupt structural flexibility in NBD1, thereby significantly altering the structure of the ATP-binding site (Dash et al., 2020). Moreover, majority of the pathogenic mutations of ABCA1 cluster in the NBDs (Fig. 9A, D).
Since ABCA1-ApoA1 interaction results in dHDL formation, mutations disrupting activity of either of the components produce a phenotype similar to that of hypoalphalipoproteinaemia. In the case of C-terminal ApoA1 mutants (residues 220-231), diminished ABCA1 binding and subsequent generation of α-migrating particles has been observed. These particles are poor substrates for LCAT (Chroni et al., 2003). Many known SNPs of ApoA1 affect its lipid-acceptor activity in plasma. Additionally, the N-terminal domain of ApoA1 hosts a majority of the pathogenic mutations (Fig. 9B, E). A95D selectively affects self-association of ApoA1 (Sviridov et al., 2002) while S36A, P143R, and E136K are linked with abnormal activation of LCAT (Weers et al., 2011). Naturally occurring mutations in the vicinity of helix six result in low plasma HDL levels and reduced LCAT activity (Mei & Atkinson, 2011). Homozygous mutants of ApoA1 (L141RPisa) have the aforementioned traits (Miccoli et al., 1997). Heterozygotes for ApoA1 (L159RFIN) result in a dominant-negative phenotype owing to impaired secretion and increased proteolytic degradation of ApoA1 (Miccoli et al., 1997). ApoA1 (L178P) heterozygotes have been detected with a decreased plasma HDL level and a remarkable susceptibility to CVD (Hovingh et al., 2005), whereas individuals with heterozygous ApoA1 (L144R) had reduced plasma HDL levels without any risk of ischemic heart disease Deciphering structural aspects of RCT (Haase et al., 2011). Surprisingly, heterozygous ApoA1 (A164S) carriers show normal plasma HDL levels despite being prone to ischemic heart disease (Haase et al., 2011). These contradictory results clearly indicate the need for understanding the changes in ApoA1 structure and their effects on RCT and the occurrence of CVD. Additionally, chemical modification-induced changes in ApoA1 structure by neutrophil myeloperoxidase lead to a loss of ability to activate RCT (Huang et al., 2014). Pro-atherogenic mutations mostly affect the central region of ApoA1, an essential region for interaction with LCAT and its subsequent activation, which may explain the phenomenon of impaired HDL maturation of these HDL-constituting mutants.
FLD is a rare monogenic disorder that occurs due to the presence of two mutant LCAT alleles that either abolish LCAT production or result in a loss-of-function of LCAT. These patients are diagnosed with a lipid profile characteristic of CVD with high TG and cholesterol levels, presence of negligible amounts of CE, and no HDL or pre β-HDL in plasma (Norum & Gjone, 1967). FLD manifests with corneal opacity, anaemia, and impaired renal function. Mutations in LCAT, which lead to partial LCAT deficiency (abolishing α-LCAT activity only) are inherited as Fish-eye disease (FED) and also result in corneal opacity later in life (Klein et al., 1993). Differentiating FLD from FED is critically based on the extent of cholesterol esterification by plasma samples. A large number of FLD-causing mutations have been found to cluster around the substrate-binding pocket (Fig. 9C, F), destabilising backbone regions around the catalytic core, such as G183S and L209P (that affect the position of Ser181 of the triad), T321M and G344S (that affect the position of Asp345), and L327R (that affects the position of His377) (Piper et al., 2015). Despite numerous structural analyses, a clear understanding of phenotypes associated with naturally occurring mutations has not yet been possible as LCAT activity, apart from its interactions with apolipoproteins, depends on several other factors including substrate structure, changes in enzyme conformation upon binding of substrates, and effects of activators. The structural analysis of such enzymes while determining their functionality must not rely only on simple reactions but also consider the physiological complexities that may impact the thermodynamic characteristics of these biomolecules.

VII. CONCLUSIONS
(1) Promotion of RCT by engaging HDL drives the transfer of excess lipids to the liver. This has emerged as the major pathway of athero-protection that can mitigate cardiovascular risk.
(2) Formation of HDL is dependent on functionally active ABCA1 and ApoA1. These proteins influence membrane dynamics during cholesterol and lipid transportation to form dHDL. ApoA1 influences ABCA1 activity and therefore evidence for both a direct and indirect interaction between these two proteins involved in dHDL formation exists.
(3) A dHDL moiety contains more than one ApoA1 molecule. Different biophysical studies have proposed different helical registries for organisation of ApoA1 monomers with respect to each other in a dHDL particle. These models mainly account for how this particle expands on addition of more lipids and interacts with remodelling proteins.
(4) It is likely that ApoA1 can exist in multiple helical registries and these are in dynamic equilibrium with each other. The conformational plasticity of ApoA1 allows for binding, activation, and subsequently inhibition, of HDL binding proteins thereby enabling the multitude of functions attributed to HDL. (5) LCAT binds HDL and drives its maturation by catalysing trans-esterification of cholesterol to CE, following ApoA1-mediated activation. This activation may be nonspecific in nature (involving interactions with amphipathic helices of ApoA1). Different orientations of ApoA1 may influence LCAT activity, either enabling or impeding cholesterol esterification. (6) LCAT activity on dHDL causes it to take up a spherical conformation to form sHDL, which has a higher cholesterol-carrying capacity. The respective orientation of ApoA1 molecules on sHDL determines the binding and accessibility of CE to other HDL remodelling proteins. (7) Different HDL particle sub-populations exist in plasma containing varying amounts of cholesterol and TGs. The complex inter-relationships between these particles and HDL remodelling proteins drive the pathway forward. These interactions ensure ApoA1 recycling, maintenance of HDL sub-populations and may even influence HDL interaction with SR-B1, expediting cholesterol delivery to liver. Continued investigations in this area are necessary to completely understand the repercussions of HDL remodelling.