CETP (cholesteryl ester-transfer protein) is essential for neutral lipid transfer between HDL (high-density lipoprotein) and LDL (low-density lipoprotein) and plays a critical role in the reverse cholesterol transfer pathway. In clinical trials, CETP inhibitors increase HDL levels and reduce LDL levels, and therefore may be used as a potential treatment for atherosclerosis. In this review, we cover the analysis of CETP structure and provide insights into CETP-mediated lipid transfer based on a collection of structural and biophysical data.
CETP (cholesteryl ester-transfer protein) transfers neutral lipids, including CE (cholesteryl ester) and TAG (triacylglycerol), between HDL (high-density lipoprotein) and LDL (low-density lipoprotein) or VLDL (very-low-density lipoprotein) particles [1–5]. By transferring CE from HDL to LDL, CETP decreases atheroprotective HDL levels [2,6]. Several reviews describing CETP biology and CETP inhibition are available [7,8]; however, there is no review describing the structural and biophysical insights into CETP function. Furthermore, a recent crystal structure has uncovered important details of lipid transfer and inhibition. This review will focus on structural and biophysical contributions towards understanding the role of CETP in lipid transfer. Here we provide a brief description of the biological role of CETP and rationale for inhibition, the parsimonious evolutionary path for the CETP protein family and a distillation of data regarding mechanisms of lipid transfer.
CETP and CVD (cardiovascular disease)
Biological role of CETP
HDL levels are an inverse predictor for CVD [9–13], elevated HDL levels being associated with significant reductions of carotid artery intima-media thickening, coronary artery atherosclerosis and overall cardiovascular risk [14–18]. The anti-atherogenic property of HDL is proposed to be due to its role in the reverse cholesterol transport pathway, where CE is moved from peripheral tissue, including macrophages and foam cells in the arterial wall, to the liver for biliary excretion . CETP transfers scavenged CE from HDL to LDL or VLDL, supporting the major route (approximately 70%) for CE clearance by liver absorption .
Although CE-rich LDL or VLDL can be cleared by hepatic uptake, it is possible for these lipoproteins to return scavenged CE back to peripheral tissues . Thus, by transferring CE to LDL or VLDL, CETP potentially facilitates the pro-atherogenic deposition of CE into peripheral tissue. In cases of persistent hypertriglyceridaemia, LDL and VLDL levels are elevated and the majority of scavenged CE is transferred from HDL resulting in net movement of CE to peripheral tissue. Consistent with this, the bulk of evidence supports a pro-atherogenic role for CETP .
Rationale for CETP inhibition in CVD
Owing to the strong inverse correlation of HDL levels with heart disease [9–13] and studies showing that rodents and humans naturally lacking CETP have elevated HDL and a resistance to diet-induced atherosclerosis [21–23], CETP has become a potential target for pharmacological inhibition. In support of the idea that reduced CETP activity will aid patients suffering from atherogenic dyslipidaemia, the two major agents used to treat dyslipidaemia, statins and fibrates, appear to share a common mechanism of action by indirectly decreasing plasma CETP levels (approximately 35 and 26% reduction respectively) [2,24–26].
Three CETP inhibitors have been brought to late-stage clinical trials: anacetrapib, dalcetrapib and torcetrapib. Torcetrapib raises patient HDL levels (approximately 70%), yet also increases the likelihood of CVD (approximately 25%)  because of a CETP-independent, off-target interaction [28,29] (reviewed in ) that increases systolic blood pressure (approximately 2.5–5.5 mmHg) [27,30–32]. Consistent with the pressor effect being CETP-independent, anacetrapib and dalcetrapib increase HDL without an increase in blood pressure [28,33]. Recent data from anacetrapib clinical trials are promising for CVD treatment, showing a 40% decrease in LDL levels with a concomitant 138% increase in HDL levels . Further results from ongoing dalcetrapib trials are awaited.
CETP is a medium-sized protein (476 residues in Homo sapiens, approximately 74 kDa) with four N-glycosylation sites (Asn88, Asn240, Asn341 and Asn396) . Fully glycosylated CETP has reduced activity (approximately 40%) relative to a partially glycosylated form without glycosylation at Asn341 .
CETP contains no independently active subunits , although a number of residues have been identified as functionally important (reviewed in ). Based on deletion studies and monoclonal antibody binding experiments, residues in the C-terminus (approximately residues 460–475) are involved in lipid transfer [37–41]. Furthermore, binding of the monocolonal antibody T2 to the C-terminus enhances CETP–HDL affinity , as does binding of anacetrapib, dalcetrapib or torcetrapib [42–44], suggesting that small-molecule inhibition may be due to the prevention of CETP–HDL dissociation.
The structure of CETP
CETP is a banana-shaped protein with approximate dimensions of 130 Å×30 Å×35 Å (1 Å=0.1 nm) (Figure 1a) . It has two structurally similar domains (the N- and C-domains) connected by a linker (residues 240–259). The structure contains 16 β-sheets and six α-helices arranged into four structural units, an N-barrel (N-domain), a C-barrel (C-domain), a central β-sheet at the interface of the N- and C-domains and a C-terminal helix. The N- and C-domains overlay well [2.71 Å3 RMSD (root mean square deviation) over 159 residues]; the N-domain contains β-strands β1–β8 and α-helices α1 and α2, whereas the C-domain contains equivalent secondary structure elements (β1′–β8′, α1′ and α2′) and an additional helix between β2′ and β3′ (ασ1) .
Structure of CETP
Interestingly, the α-helices and β-strands of CETP contain several π-helices and β-bulges (Figure 1A). π-Helices or β-bulges usually manifest in α-helices or β-strands respectively, due to the insertion of a single residue without accommodation by the original secondary structure . Since insertion is typically unfavourable and associated with protein destabilization (approximately 3–6 kcal/mol; 1 kcal=4.184 kJ) [47–50], retention of a π-helix or β-bulge implies the insertion may confer a selective advantage outweighing the cost of destabilization. Consistent with this, π-helices are correlated with functional sites [51,52]. Most of the π-helices and β-bulges in CETP are conserved between members of the CETP family (Figure 2).
π-Helix conservation in the CETP family
The crystal structure of CETP shows two CE and two PL (phospholipid) molecules bound within two large hydrophobic chambers, corresponding to the centres of the N- and C-domains. The two chambers are connected by a ‘neck’, creating a true tunnel through CETP with openings approximately 25 Å apart on the concave surface of the protein. The PL molecules bind with their hydrophobic tails buried towards the centre of the chambers, sheltering the hydrophobic core of CETP, while their phosphate heads remain solvent-exposed at the tunnel entrance (Figure 1b). Although not present in this structure, these chambers are presumably also the site of TAG binding.
CETP inhibitors bind in the N-domain hydrophobic chamber
No structure of a CETP–inhibitor complex is available; however, given that dalcetrapib is known to form a disulfide bond with Cys13 within one of the hydrophobic chambers [44,53], and anacetrapib, dalcetrapib and torcetrapib compete for CETP binding , it is likely that all known CETP inhibitors bind near Cys13 within the N-domain hydrophobic chamber . Proximal anacetrapib-, dalcetrapib- and torcetrapib-binding sites imply that these inhibitors share a common mechanism of inhibition.
If anacetrapib and torcetrapib bind near Cys13, three potential mechanisms of inhibition are suggested: direct blockage of CE binding, a decrease in CETP solubility because of the loss of PL or blockage of the neck-connecting chambers. Torcetrapib is non-competitive against CE binding, arguing inhibition is not because of direct blockage of CE binding . T2 or small molecules induce a stable complex between CETP and HDL [38,42–44]; PL loss would effectively make CETP more hydrophobic and change the partitioning of CETP to favour association with HDL, and thus these results are consistent with the loss of PL. Inhibition due to blockage of the neck is supported by mutations in the neck decreasing CE and TAG transfers .
Evolution of CETP and related proteins
CETP sequence conservation
Sequence alignment of 26 CETP orthologues reveals 47 residues with 100% conservation, 51 residues with≥96% conservation and 44 residues with≥92% conservation. Strikingly, all of the 100% and nearly all the 96% conserved residues are located in the N-domain, whereas the 92% conserved residues are more evenly distributed across both domains (Figure 1c). The reason for the stronger conservation in the N-domain is unknown.
Evolution of the CETP family of proteins
CETP belongs to a family that may have evolved from an ancestral lipid binding protein; members of this family include BPI (bactericidal/permeability-increasing protein), LBP (lipopolysaccharide-binding protein) and PLTP (phospholipid-transfer protein) . BPI was the first member of the CETP family to have its structure determined, first at 2.4 Å resolution  and later extended to 1.7 Å resolution . The structures of CETP and BPI are highly similar, with the N-barrel, C-barrel and central β-sheet individually overlaying well, but with different relative orientations. Similar to CETP, BPI contains two hydrophobic chambers, each plugged by a PL molecule, but does not have CE molecules bound.
The clear symmetry of the N- and C-domains of CETP family members implies these domains developed from gene duplication and fusion of an ancient single-domain progenitor. Consistent with this, there exists a single-domain CETP family member that also binds its cargo in a hydrophobic chamber . Furthermore, BPI retains bactericidal activity when its C-domain is removed, demonstrating that functionally BPI can be considered a fusion of two single-domain proteins [57–60]. Although it is possible that BPI, CETP, LBP and PLBP (phospholipid-binding protein) diverged from a single-domain progenitor and then individually underwent gene duplication and fusion, the parsimonious path places divergence after a common single-domain ancestor had become a two-domain protein. Consistent with this, there is a higher overall sequence identity between BPI and CETP (22%) than between the N- and C-domains of CETP (11%).
Proposed lipid transfer mechanisms
In the shuttle mechanism, CETP acts as a cargo-carrying shuttle picking up lipids from one lipoprotein and depositing them at a second. This mechanism is supported by kinetic characterization by two groups [61,62] and the interpretation of the first structure of CETP , where the authors note the concave surface of CETP is the appropriate size for the protein to dock on to HDL. Docking would require changes in the C-terminal helix, consistent with the demonstrated importance of C-terminal residues [39,41]. Once docked, dissociation of PL would allow bound lipids to exit and new lipids to bind. Docking between CETP and LDL or VLDL would require the concave surface of CETP to flatten as either particle is significantly larger than HDL . This model can be used to describe transfer between one type of lipoprotein (homoexchange) and or transfer between two different types of lipoprotein (heteroexchange).
Lipid transfer through an HDL–CETP–LDL complex is supported by early kinetic experiments . There is no information on the architecture of an HDL–CETP–LDL complex; however, because HDL and LDL do not directly bind each other [61–63], and are so much larger than CETP, a reasonable arrangement would be for CETP to separate the two lipoproteins by interacting with one at each of its two domains. In such an arrangement it may be possible for CETP to act as a tunnel directly linking the lipid cores of the two lipoproteins. Consistent with this, non-connecting internal cavities extend from the defined hydrophobic chambers to the tips of CETP (Figure 1b); structural changes could result in the connection of these internal pockets to create a longer, tip-to-tip tunnel for lipid transfer. Structural changes may be facilitated by the conserved π-helices and β-bulges on the ends of the N- and C-domains (Figures 1A and 2), either by inherently destabilizing these regions or through peristalsis-like shifts [46,64]. If CETP is forming a tunnel between HDL and LDL, this would explain why both N- and C-domains are necessary for function , and why mutations or small-molecule inhibitors blocking the neck are able to arrest lipid transfer while still allowing CE binding [45,65].
Structural and biophysical characterization of the CETP system has led to insights on the mechanism of lipid transfer and the evolutionary history of the CETP family of proteins. The recent results of the anacetrapib clinical trials demonstrate inhibition of CETP could play an important role in CVD prevention , underscoring the importance of continued research of CEPT-facilitated lipid transfer. Current and future studies will provide further insights into the CETP transfer mechanism and aid the rational design of newer and better inhibitors to complement anacetrapib and dalcetrapib.
Proteins with a BPI/LBP/PLUNC-Like Domain: Revisiting the Old and Characterizing the New: A Biochemical Society Focused Meeting held at New Business School, University of Nottingham, U.K., 5–7 January 2011. Organized and Edited by Colin Bingle (Sheffield, U.K.) and Sven-Ulrik Gorr (University of Minnesota School of Dentistry, Minneapolis, MN, U.S.A.).
We thank Dr Andrea Hall, Dr Ann Aulabaugh and Dr Kieran Geoghegan for critical evaluation of this paper. J.H. and X.Q. are employees of Pfizer Inc., a pharmaceutical company that discovered torcetrapib and markets drug products.