HDL Measures, Particle Heterogeneity, Proposed Nomenclature, and Relation to Atherosclerotic Cardiovascular Events

BACKGROUND

A growing body of evidence from epidemiological data, animal studies, and clinical trials supports HDL as the next target to reduce residual cardiovascular risk in statin-treated, high-risk patients. For more than 3 decades, HDL cholesterol has been employed as the principal clinical measure of HDL and cardiovascular risk associated with low HDL-cholesterol concentrations. The physicochemical and functional heterogeneity of HDL present important challenges to investigators in the cardiovascular field who are seeking to identify more effective laboratory and clinical methods to develop a measurement method to quantify HDL that has predictive value in assessing cardiovascular risk.

CONTENT

In this report, we critically evaluate the diverse physical and chemical methods that have been employed to characterize plasma HDL. To facilitate future characterization of HDL subfractions, we propose the development of a new nomenclature based on physical properties for the subfractions of HDL that includes very large HDL particles (VL-HDL), large HDL particles (L-HDL), medium HDL particles (M-HDL), small HDL particles (S-HDL), and very-small HDL particles (VS-HDL). This nomenclature also includes an entry for the pre-β-1 HDL subclass that participates in macrophage cholesterol efflux.

SUMMARY

We anticipate that adoption of a uniform nomenclature system for HDL subfractions that integrates terminology from several methods will enhance our ability not only to compare findings with different approaches for HDL fractionation, but also to assess the clinical effects of different agents that modulate HDL particle structure, metabolism, and function, and in turn, cardiovascular risk prediction within these HDL subfractions.

Conventionally, cardiovascular prevention strategies emphasize therapeutic reductions in LDL cholesterol (1, 2). However, increasing attention is being focused on HDL cholesterol as a secondary prevention target to address residual cardiovascular disease (CVD)¹¹ risk (3, 4). Low HDL cholesterol is widely prevalent in Westernized countries, and is independently predictive of CVD risk (5, 6), even at low concentrations of LDL cholesterol (7). However, low concentrations of HDL cholesterol are often accompanied by increased concentrations of small, cholesterol-depleted LDL particles and increased concentrations of cholesterol-enriched triglyceride remnants. Thus, the CVD risk associated with low HDL cholesterol values is difficult to separate from that of other associated lipoprotein abnormalities (8).

HDL particles are heterogeneous in size and composition. Despite the substantial epidemiological data suggesting a cardioprotective role for HDL, much remains unknown about the antiatherothrombogenic properties of different particles that comprise this class of lipoproteins. Some of the attributes of HDL in mitigating atherosclerosis include its role in reverse cholesterol transport, oxidation, and inflammation (9, 10). Several genetic mutations can affect the structure and function of HDL, but it is unclear what impact these have on CVD risk (11). Moreover, development of diagnostic and treatment strategies to study and target HDL metabolism must involve not only the absolute concentration of HDL cholesterol, but also the functional qualities of HDL particles (10, 12).

Methods for measurement of HDL subfractions (13) as well as compositional and functional assays may be superior to HDL cholesterol in predicting coronary heart disease (CHD) risk (10, 14). Thus, there is a need to propose a new framework to encompass and highlight the structural, compositional, and functional diversity of these particles.

In this special report we discuss the advantages and disadvantages of current analytical measures of HDL and provide insights into newer research methods that characterize HDL on the basis of its heterogeneous physicochemical and functional properties. There is an increasing need to understand, validate, and quantify the diverse roles of HDL particles in the atherosclerotic process to improve diagnosis, prevention, and treatment of CVDs (9, 10). The information in this report is intended to serve as a foundation to foster improved understanding of the pathophysiology of atherosclerosis and direct the future course of research and the design of interventions that effectively reduce residual risk in various patient populations. Finally, we present a uniform nomenclature for HDL subfractions and propose a paradigm to define the dynamic process of HDL metabolism with multiple static laboratory measurements. We acknowledge that the various HDL methods measure different physical and chemical properties of HDL and that the use of static measures for dynamic processes has inherent limitations, but we also recognize the inconsistency of existing nomenclature for HDL subclasses and the need for a uniform structure that allows clinician-scientists to interrelate the various methods into a functional construct.

HDL Cholesterol and Cardiovascular Risk

The cholesterol content of HDL is conventionally used to assess the multifarious antiatherothrombotic and immune-related functions of HDL particles. The reliance on HDL cholesterol in clinical practice partly derives from its use as a principal component of the Friedewald equation used to estimate LDL cholesterol (15).

HDL cholesterol has been evaluated as a risk marker in 68 long-term population-based studies involving more than 300 000 individuals (16). In multivariate models adjusted for both nonlipid and lipid (triglycerides and non-HDL cholesterol) risk factors, HDL cholesterol is inversely associated with CHD events. For every 0.39 mmol/L (15 mg/dL) increase in HDL cholesterol concentration, the risk of a CHD event was reduced by 22% (95% CI, 18%–26%). Low concentrations of HDL cholesterol predict CHD mortality equally well in nondiabetic and diabetic patients (17).

Evidence for the utility of HDL cholesterol as a risk marker in patients treated with lipid-altering therapy has been conflicting depending on the extent of covariate adjustments. In a metaanalysis involving 90 056 participants from 14 randomized trials of therapy with statins (hydroxy-methyl-glutaryl coenzyme A–reductase inhibitors), the Cholesterol Treatment Trialists' Collaboration investigators reported the proportional effects of various baseline prognostic risk factors on vascular events including HDL cholesterol (18). The 5-year incidence of major CVD events was higher among individuals with the lowest HDL cholesterol concentrations. The use of statins reduced the risk of CHD events by 22% in individuals in the lowest tertile of HDL cholesterol [<0.9 mmol/L (35 mg/dL)] and by 21% in individuals in the middle tertile [0.9–1.1 mmol/L (35–42 mg/dL)] and upper tertile [≥1.1 mmol/L (42 mg/dL)]. However, participants with the lowest HDL cholesterol concentrations had higher absolute risk (22.7%, 18.2%, and 14.2% for the low, middle, and high tertiles, respectively), and thus experienced the largest absolute risk reduction.

Similarly, on-trial concentrations of HDL cholesterol are predictive of recurrent CVD events in most prospective clinical trials (7, 19). The increased risk associated with low concentrations of HDL cholesterol persists even in statin-treated patients with LDL cholesterol <1.8 mmol/L (70 mg/dL). However, this concept was recently challenged in a metaanalysis of 95 trials involving nearly 300 000 individuals, the results of which suggested that on-trial HDL cholesterol concentrations were not significantly related to CHD events (20). Limitations of the study included use of pooled data rather than individual study participant data, and lack of consideration of baseline triglyceride concentrations. Furthermore, the majority of studies included in this metaanalysis showed minimal (<3%) differences in HDL cholesterol concentrations between the treatment groups, whereas the analytical variation for direct HDL cholesterol assays is frequently >10% (21).

Considering that high-risk individuals are often treated with statins, HDL measurements that extend beyond its cholesterol content may provide more useful information for risk stratification in potentially high-risk individuals and particularly in those patients treated with lipid-altering therapy.

Analytical Limitations of HDL Cholesterol Measurements

The earliest methods for measurement of HDL cholesterol involved preparative ultracentrifugation (22) for isolation of HDL with densities between 1.063 and 1.21 g/mL. It was not until the advent of chemical-based precipitation methods, with reagents such as dextran sulfate in the early 1970s, that it became practical to measure HDL cholesterol in clinical laboratories. In the past 10 years, most laboratories have switched to direct (homogenous) assays that do not involve physical separation of HDL from other lipoproteins. There are currently 7 different direct HDL cholesterol assays, which use several different approaches for either shielding or selectively consuming cholesterol on non-HDL lipoprotein fractions (Table 1). Direct HDL cholesterol assays are fully automated and precise and require less labor. Therefore, they have largely replaced older assays. It remains uncertain whether direct HDL cholesterol assays have clinical utility comparable to that of chemical-based precipitation methods (23–25). In a recent study of 175 individuals with a wide variety of lipid disorders, none of the 7 current direct assays met the minimum total error goal of less than 12% established by the National Cholesterol Education Program (21). Furthermore, inaccurate HDL cholesterol results from direct assays were found to significantly compromise the accurate classification of CVD risk based on estimated LDL cholesterol.

Commercially available direct HDL-cholesterol assays.

Table 1.

Commercially available direct HDL-cholesterol assays.

Precipitation

Heparin-Mn2+

0.46 mmoL (Lipid Research Clinics method)

0.92 mmoL (for EDTA plasma)

Dextran sulfate (50 kDa) Mg+2 (designated comparison method)

Phosphotungstate-Mg2+

Polyethylene glycol (does not precipitate apo E–rich HDL)

Facilitated precipitation

Polymedco (magnetic beads conjugated with dextran sulfate-Mg2+)

Direct (homogenous methods)

Denka Seiken (selective elimination)

Kyowa Medex (polyethylene glycol–modified enzymes/cylodextrin)

Sekisui Medical (formerly Daiichi) (synthetic polymer/detergent)

Serotec

Sysmex International Reagents (immunoinhibition)

UMA

Wako Pure Chemical Industries (immunoinhibition)

Classification of HDL by Physicochemical Properties

ANALYTICAL ULTRACENTRIFUGATION

The earliest method used for quantification of HDL involved analytical ultracentrifugation using schlieren optics. In the late 1940s, Gofman, Lindgren, and colleagues at Donner Laboratory in Berkeley, California, identified HDL subclasses as a function of size and density based on their ultracentrifugal flotation rate (F_1.2) in a high-salt solution (26). These studies established that most HDL particles have buoyant density between 1.063 and 1.21 g/mL, and this information provided the basis for standard preparative ultracentrifugal isolation of HDL (21, 26) (Fig. 1). Moreover, smaller, denser HDL3 (F_1.2 0–3.5) were well distinguished from larger, more buoyant HDL2 (F_1.2 3.5–9) on the basis of a distinct shoulder in the optical schlieren profile. Larger HDL (F_1.2 9–20), designated HDL1, represented a relatively minor species in most individuals. Using first principles of physics, we converted the analytical ultracentrifuge schlieren profiles to mass concentrations of lipoprotein particles. This gold standard method was the first to be used in a prospective study to demonstrate an inverse relation of plasma HDL concentration to CHD risk (27). Recently, results of long-term follow-up (29 years) of 1905 men in this study have demonstrated that both HDL2 and HDL3 are independently related to CHD risk (28).

Analytical ultracentrifugation.

Fig. 1.

Measurement of HDL by analytical ultracentrifugation. Major HDL-particle subclasses are distinguished by flotation rate in a salt solution of density 1.2 g/mL (F_1.2) and total mass, represented by the area under the curve (AUC), is determined by first principles of physics from schlieren optics. Initially, 3 major HDL subclasses were identified. HDL1, with the highest flotation rate, is generally not detectable in substantial concentrations in human plasma. A curve-fitting procedure was later developed to resolve 2 subclasses of HDL2 (HDL2a and HDL2b).

NONDENATURING GRADIENT GEL ELECTROPHORESIS

Gradient gel electrophoresis in conjunction with automated densitometry was applied in 1981 by Nichols and colleagues at Donner Laboratory (29) to identify 5 HDL subspecies separable on the basis of particle diameter: HDL3c (7.2 to 7.8 nm), HDL3b (7.8 to 8.2 nm), HDL3a (8.2 to 8.8 nm), HDL2a (8.8 to 9.7 nm), and HDL2b (9.7 to 12.9 nm) (Fig. 2). Results of subsequent studies indicated that HDL2b, which is strongly correlated with total HDL cholesterol, was most strongly inversely related to CHD risk (30), and that increased HDL3b was associated with an atherogenic lipoprotein phenotype characterized by increased triglycerides and small, dense LDL, together with reduced HDL2b (31). As described below, the use of 2-dimensional (2-D) electrophoresis has demonstrated that particles corresponding to HDL2b are independently and inversely related to CHD risk (32).

Nondenaturing gel electrophoresis.

Fig. 2.

Separation of HDL subclasses from 4 representative plasma samples by gradient gel electrophoresis. HDL is isolated from plasma by ultracentrifugation at density 1.21 g/mL, and electrophoresed in a 4%–30% nondenaturing gradient gel. After the protein is stained with Coomassie Blue and the gels are scanned by densitometry, the size distribution is determined by calibration using protein standards (right lane). This procedure resolves 5 distinct subclasses, although the smallest, HDL3c, is generally present at low concentrations. Std, standard.

Density Gradient Fractionation of Plasma Lipoproteins

Precise and reproducible fractionation of the major HDL particle subpopulations (HDL2b, -2a, -3a, -3b, and -3c) in human plasma is based on an isopycnic equilibrium method developed by Chapman and colleagues (33–36) (Fig. 3). By use of a gradient tube of a swing-out rotor, the gradient is constructed by consecutive layering of 4 salt solutions of distinct densities that have been adjusted accurately at the same temperature as that of the ultracentrifugal separation (+15° C). The major disadvantage of this method is the same as that of other ultracentrifugal separations; because lipoproteins are subject to high ionic strength and centrifugal force (57 × 10⁷g average/min), shear forces are reduced by use of a swing-out rotor.

Density gradient ultracentrifugation.

Fig. 3.

Representative electrophoresis profiles and mean particle sizes of HDL subfractions (HDL2b, HDL2a, HDL3a, HDL3b and HDL3c) from normolipidemic human plasma separated by isopycnic, single-spin, density gradient ultracentrifugation [1-dimensional electrophoresis was performed in nondenaturing gradient polyacrylamide gel (4%–20%)]. Human plasma is separated by isopycnic, single-spin, density gradient ultracentrifugation. The plasma or serum sample (3 mL) adjusted to a density of 1.21 g/mL is layered on a cushion of NaCl-KBr solution of density 1.24 g/mL at the base of the gradient tube; the discontinuous gradient is then completed by layering successive density solutions of 1.063, 1.019, and 1.006 g/mL onto the latter. The procedure involves a single ultracentrifugal step, allows almost quantitative recovery of highly resolved HDL fractions of defined hydrated density and physicochemical properties, avoids major contamination with plasma proteins, and facilitates HDL isolation in a nondenatured, nonoxidized state. Gradients are fractionated with a precision pipette from the meniscus downwards, thereby avoiding contamination with plasma proteins >1.25 g/mL present in the residue at the base of the tube. Peak diameter was determined at the maximum absorption intensity of each band by using Kodak 1D software filters following staining with Coomassie Brilliant Blue. **Size by negative stain electron microscopy provided smaller estimates (HDL2b + HDL2a, mean diameter 9.6 nm and range 10.8–7.2 nm; HDL3a + HDL3b + HDL3c, mean diameter 7.3 nm and range 9.0–5.4 nm) reflecting the nonhydrated state.

VERTICAL ANALYTICAL PROFILE

Vertical auto profile (VAP) is another HDL subfractionation method based on ultracentrifugation (37). Unlike most other ultracentrifugation methods, VAP is done in a vertical rotor, which makes the method relatively fast and more practical for the analysis of routine clinical specimens. With regard to HDL, VAP measures the cholesterol content of its 2 major density subfractions, namely HDL2 and HDL3 (38).

VAP is relatively precise, with intraassay CVs for lipoprotein subfractions that range from 4% to 10% (39). Specific limitations of ultracentrifugation methods including the VAP technique are described in online Appendix 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol57/issue3.

Limited studies have been done to compare the VAP method to other lipoprotein subfraction procedures, but so far the results of these studies have shown relatively poor agreement (40). This is perhaps not surprising because of lack of standardization of the different fractionation methods, as well as the fact that the various lipoprotein subfractionation methods are based on different physical and chemical properties of HDL.

2-D GEL ELECTROPHORESIS

HDL can be separated on the basis of size and charge (Figs. 4 and 5) (13, 41). The concentrations of these particles are expressed in milligrams per liter of apolipoprotein A-I (apo AI), and as a percentage of total plasma apo AI concentration. Five major HDL particles have been identified: (a) very small discoidal precursor HDL of pre-β mobility (pre-β-1 HDL, diameter about 5.6 nm), which contains apo AI and phospholipid; (b) very small discoidal HDL of α mobility (α-4 HDL, diameter about 7.4 nm), which contains apo AI, phospholipid, and free cholesterol; (c) small spherical HDL of α mobility (α-3 HDL, diameter about 8.0 nm), which contains apo AI, apo AII, phospholipid, free cholesterol, cholesteryl ester, and triglyceride; (d) medium-sized spherical HDL of α mobility (α-2 HDL, diameter about 9.2 nm), which contains the same constituents as α 3 HDL; and (e) large spherical HDL of α mobility (α-1 HDL), which contains the same constituents as α-3 and α-2 HDL, except for the near absence of apo AII (Fig. 4). Adjacent to the α particles are pre-α particles that are of similar size, but present in lower amounts and do not contain apo AII. In addition, there are large pre-β migrating HDL known as pre-β-2 HDL (42).

2-D gel electrophoresis.

Fig. 4.

The apo AI–containing HDL subpopulation profiles of a CHD patient (A) and a healthy individual (control) (B), with a schematic diagram of all of the apo AI–containing HDL particles shown on the right (C). Below panel (A) is a plot of a densitometric scan across the α-migrating HDL-particle region indicating the presence of 4 α-migrating HDL particles ranging in mean particle diameter from very large α-1 HDL (11.0-nm diameter) to very small pre-β-1 HDL (5.6-nm diameter). In the schematic diagram α-migrating apo AI–containing particles in the α-2 region (9.2-nm diameter) and in the α-3 region (8.1-nm diameter) contain both apo AI and apo AII (more heavily shaded), whereas all other particles containing apo AI, including small α-4 HDL (7.4-nm diameter), do not contain appreciable amounts of apo AII (less heavily shaded). The asterisk marks the serum albumin or α front. Based on their composition, very small pre-β-1 HDL and small α-4 HDL are discoidal particles that do not contain cholesteryl ester or triglyceride, whereas medium, large, and very large α-3, -2, and -1 HDL are spherical and contain cholesteryl ester and triglyceride in their cores. CHD patients in general in the untreated state tend to have significant decreases in the levels of apo AI in very large and large α-migrating HDL and modest increases in apo AI in very small pre-β-1 HDL and small α-4 HDL. In (C), 1, 2, 3, and 4 refer to α-1, -2, -3, and -4, respectively. Asztalos et al. (181).

2-D gel electrophoresis patterns after apo AI immunoblotting.

Fig. 5.

The 2-D gel electrophoresis patterns after apo AI immunoblotting observed for whole plasma are shown in the left panel, for lipoproteins of density (d) <1.125 g/mL as separated by ultracentrifugation are shown in the center panel, and for lipoproteins of density 1.125–1.24 g/mL are shown in the right panel. These data indicate that apo AI–containing HDL of density <1.125 g/dL are comprised mainly of very large and large α-migrating HDL, whereas apo AI–containing HDL particles of density 1.125–1.24 g/mL contain mainly medium and small α-migrating HDL and very small pre-β-1 HDL. Asztalos et al. (181).

Pre-β-1 HDL particles are most efficient in interacting with the ATP binding cassette transporter A1 (ABCA1) to promote cholesterol efflux from cells, whereas large α-1 HDL are the most efficient in interacting with the liver scavenger receptor B1 for delivery of cholesterol to the liver (43, 44). Intermediate-sized α-3 HDL are the most efficient in interacting with the ATP transporter G1 (ABCG1) to promote cellular cholesterol efflux onto spherical HDL particles containing both apo AI and apo AII (44). Delipidated HDL or apo AI_Milano complexed with phospholipids, which have been reported to promote regression of coronary atherosclerosis when delivered by infusion, consist of pre-β-1 HDL particles (45). There are other HDL particles containing apo E without apo AI (very large pre-β migrating HDL) and small HDL containing apo AIV without apo AI (43). The functions of these latter particles have not been well defined.

2-D gel electrophoresis of plasma followed by apo AI immunoblotting allows for the accurate diagnosis of disorders of HDL metabolism. apo AI deficiency is characterized by the absence of apo AI–containing HDL particles, and patients with apo AI deficiency often develop xanthomas and early-onset CHD (46). apo AI is nearly absent in Tangier disease, which is characterized by lack of functional ABCA1 transporters and cholesteryl ester deposition in macrophages throughout the body. These patients have only pre-β-1 HDL, and they usually develop premature CHD (47). Familial lecithin:cholesterol acyltransferase (LCAT)-deficient patients have only pre-β-1 and α-4 HDL particles, cannot esterify their cholesterol, and can develop severe corneal opacification, increased LDL, and renal failure (41). Lipoprotein lipase–deficient patients have marked hypertriglyceridemia that places them at high risk for pancreatitis. These patients also have low concentrations of HDL cholesterol that is carried only in pre-β-1 and α-4 HDL particles (48). Hepatic lipase–deficient patients have increased remnant lipoproteins, a decrease in α-2 HDL, and increased risk for premature CHD (48). Cholesteryl ester transfer protein (CETP)-deficient patients have very large α HDL that contains apo AI, apo AII, and apo E (49). The presence of apo E on these large HDL particles may be important to drive cholesterol efflux by the ABCG1 transporter.

When the concentration of apo AI in α-1 HDL is less than 140 mg/L, the individual is at increased risk of developing CHD (32). CHD patients also often have small discoidal HDL particles and decreased large α-1 and α-2 HDL particles. Concentrations of these particles are superior to HDL cholesterol concentrations for use in CHD risk prediction (50, 51).

Large α-1 particles increase with weight loss, niacin, certain statins (atorvastatin, rosuvastatin), and CETP inhibitors (52–57). Increasing the concentrations of apo AI in α-1 HDL to >200 mg/L (0.52 mmol/L) with a simvastatin/niacin combination has been associated with lack of progression and in some individuals with regression of coronary atherosclerosis (51).

HDL Particle Concentration

NUCLEAR MAGNETIC RESONANCE SPECTROSCOPY

Unlike other methods for analyzing HDL particles, nuclear magnetic resonance (NMR) spectroscopy does not require a physical separation step because the protons (the nuclei of hydrogen atoms) within lipoprotein particles of different sizes have a natural magnetic distinctness arising from their unique physical structure (58). As a result, lipoprotein particles of different sizes in unfractionated plasma or serum give rise to distinguishable lipid NMR signals that have characteristically different frequencies (Fig. 6, left panel) (59, 60). The NMR signal frequencies (chemical shifts) of HDL subfractions, compared to LDL and VLDL subfractions, are particularly well differentiated (Fig. 6, right panel).

NMR spectroscopy.

Fig. 6.

Relationship of lipoprotein subclass diameter to lipid methyl group NMR chemical shift and frequency (left panel) and methyl signal line shape and chemical shift of 5 isolated HDL subclasses of differing diameter (right panel). The natural magnetic distinctness of lipoprotein particles of different size makes it possible in theory to use any NMR instrument in any laboratory for lipoprotein analysis, but in practice such analysis requires dedicated instrumentation. The subclass NMR signals highly overlap, making it necessary to computationally “deconvolute” the plasma NMR signal envelope to extract the amplitudes of the subclass signals that are used to calculate subclass concentrations. Accurate and reproducible deconvolution is possible only if the NMR conditions (such as magnetic field strength and temperature) used to generate the subpopulation reference signal library exactly match the conditions used subsequently to measure (in approximately 1 min) each patient plasma sample. The NMR LipoProfile-3 method currently employed by LipoScience models the plasma signal as the sum of the contributing signals of 26 subpopulations of HDL as well as 47 subpopulations of LDL, VLDL, and chylomicrons (Chylos). Given the limited measurement precision of the derived concentrations of each of the many subpopulations, they are grouped for routine reporting purposes into “large,” “medium,” and “small” subclass categories.

The lipoprotein NMR signals employed for lipoprotein quantification are those from the terminal lipid methyl group protons, because they are unresponsive to, and therefore unaffected by, fatty acid and other chemical compositional differences (60). Furthermore, to a close approximation the number of methyl protons in a lipoprotein particle of given diameter is constant even in the face of significant variations in core cholesterol ester and triglyceride content. These properties render the detected subclass methyl-signal amplitudes directly proportional to subclass particle number and enable NMR-derived concentrations of HDL to be given in particle number units (μmol/L) (60). Although NMR analysis provides a new and potentially clinically advantageous way to quantify HDL and its subfractions, the NMR lipid methyl signal is inherently unable to provide HDL chemical compositional information.

The current NMR method models the plasma signal as the sum of the contributing signals of 26 subpopulations of HDL as well as 47 subpopulations of LDL, VLDL, and chylomicrons. Given the limited measurement precision of the derived concentrations of each of the many subpopulations, they are grouped for routine reporting purposes into “large,” “medium,” and “small” subclass categories. For research studies, it is not difficult to produce alternative groupings of the 26 HDL subpopulations to make the reported subfractions more similar to those produced by other analytic methods such as density gradient ultracentrifugation and gradient gel electrophoresis.

Investigations are ongoing to establish CVD relations with NMR-determined concentrations of HDL particles, HDL subclasses (large HDL particles, 9.4–14 nm; medium HDL particles, 8.2–9.4 nm; and small HDL particles, 7.3–8.2 nm), and HDL particle size. Among published studies are those documenting associations with age and sex (61, 62), longevity (63), insulin resistance and diabetes (64–67), CHD (62, 68–74), and changes brought about by exercise (75) and treatment with various drugs (76–84). An important consideration in interpreting the clinical significance of observed univariate disease associations with individual HDL subclasses or HDL size is the confounding that arises from the strong inverse correlation between large and small HDL subclasses and the even stronger inverse associations of large HDL particles (and HDL size) with total (and small) LDL particle concentrations (60, 65, 74). Without conducting statistical analyses that address the confounding caused by these correlations, misleading conclusions may be reached about the clinical importance and potential functional differences among HDL subclasses (62, 73, 74).

ION MOBILITY

Ion mobility, a gas-phase differential electrophoresis macromolecular mobility-based method, was developed by Benner and colleagues at Lawrence Berkeley National Laboratory (85). In this high-throughput procedure, lipoprotein size is determined by first principles of physics, and particles are counted directly after equalization of charge and separation by time of flight through a voltage gradient. Albumin contamination of the HDL size region is first reduced by incubation with blue dextran and a short ultracentrifugation in the absence of salt. In its current configuration, the method is designed to separate HDL2b from smaller HDL; methods are in progress to resolve and measure HDL2a and the HDL3 subspecies. This method has recently demonstrated that the large HDL2b subspecies is strongly inversely correlated with coronary disease risk in the prospective Malmo Diet and Cancer Study (86). The association of these large HDL particles with CHD risk was related to their inclusion in 2 independent principal components determined from ion mobility measurements of all lipoprotein fractions. One of these corresponds to the atherogenic lipoprotein phenotype, which includes increased concentrations of triglycerides and smaller LDL particles, and the second includes smaller HDL particles. Results of genetic analyses performed in this study indicated that these components have differing underlying determinants, and this may indicate 2 independent mechanisms for the cardioprotective effects of HDL.

HDL Particle Heterogeneity Viewed through Lipid and Protein Composition: apo AI and Cardiovascular Risk

apo AI is the major HDL protein (87) and is synthesized by both hepatic and intestinal cells (88). apo AI has been considered a more precise biomarker than HDL cholesterol on the basis of its functional role in mediating cholesterol mobilization via the ABCA1 transporter in peripheral cells, including arterial macrophages. Indeed, early studies suggested that apo AI measurements were superior to HDL cholesterol as a risk marker (89, 90). Subsequently, these 2 measures of HDL, HDL cholesterol and apo AI, have been directly compared in 2 large cohorts (74, 91). In the European Prospective Investigation into Cancer and Nutrition–Norfolk study, which included 2349 individuals, the non–lipid-adjusted risk of a major CHD event per 1-SD change was 0.78 (0.70–0.87) for HDL cholesterol and 0.79 (0.71–0.87) for apo AI (74). A recent analysis of the Women's Health Study showed that the magnitude of risk for a major cardiovascular event was higher for low levels of HDL cholesterol than for apo AI (91).

Among high-risk individuals treated with statin therapy, on-trial measures of apo AI provide incremental information on CHD risk beyond that obtained with HDL cholesterol (92), whereas in the stable CHD patients enrolled in the Incremental Decrease in Endpoints through Aggressive Lipid Lowering (IDEAL) study, these measures provide equivalent prognostic data (74). In the Air Force-Texas Coronary Atherosclerosis Prevention Study, low on-trial concentrations of apo AI at 1 year were predictive of major CHD events, whereas no significant predictive value was seen for on-trial HDL cholesterol concentration (92). In contrast, there were no differences between HDL cholesterol and apo AI concentrations with regard to risk of recurrent events in statin-treated CHD patients enrolled in the IDEAL study (74) or fibrate-treated CHD patients enrolled in the Veterans Administration HDL Intervention Trial (VA-HIT) (71).

apo AI Measurements

Because apo AI is an abundant serum protein, it is relatively easy to measure by either nephelometric or turbidometric methods that are available on most standard clinical chemistry analyzers (93). Usually, nonionic detergents are added to the assay buffer to disrupt HDL and expose apo AI antigenic sites, which ameliorates background problems from turbid specimens. Oxidized forms of apo AI (94–96) have also been evaluated as potential CHD risk biomarkers and may stimulate a reconsideration of the value of apo AI in cardiovascular risk assessment.

Classification of HDL by Apolipoprotein Composition of Lipoprotein Particles

Plasma lipoproteins can be separated and classified on the basis of their apolipoprotein composition. Alaupovic and colleagues used apolipoprotein composition as a basis for separating human plasma lipoproteins into particles, which included lipoprotein B (LpB) (LpB, LpB:C, LpB:C:E), LpA (LpAI, LpAII, LpAI:AII), LpC (LpCI:CII:CIII), LpE, and LpD (97, 98).

apo AI AND apo AI:AII PARTICLES

LpAI and LpAI:AII are the major HDL lipoproteins, containing approximately 35% and 65% of plasma apo AI, respectively (99). LpAI is initially secreted as a lipid-poor apo AI:phospholipid complex, which interacts with the ABCA1 transporter and facilitates cholesterol efflux resulting in the formation of pre-β HDL (100). The cholesterol in pre-β HDL is esterified to cholesteryl esters by LCAT, with conversion of pre-β HDL–containing to α HDL–containing LpAI particles (100, 101). Hepatic secreted apo AII associates with LpAI to form LpAI:AII. The role of apo AII in HDL metabolism has not been definitively established, although apo AII has been reported to decrease HDL remodeling (102) and reduce cholesterol uptake by hepatic scavenger receptor class B type I (SR-BI) (103). α HDL–containing LpAI and LpAI:AII drive efflux of cholesterol after interaction with the ABCG1 transporter (104–106). Thus, a dual pathway for efflux of cholesterol from cholesterol-loaded cells involves lipid-poor AI particles interacting with the ABCA1 transporter, and larger LpAI/LpAI:AII particles interacting with the ABCG1 transporter (107–111). In plasma both LpAI and LpAI:AII are heterogeneous and can be separated into subfractions on the basis of lipid composition, density, size, and charge.

The cardioprotective roles of LpAI and LpAI:AII have been controversial. In an initial report LpAI but not LpAI:AII was able to drive cholesterol efflux from Ob1771 adipocytes in culture, suggesting that LpAI but not LpAI:AII was antiatherogenic (110). Results of subsequent clinical studies in which LpAI and apo AI:AII were evaluated in CHD patients have revealed a decrease in HDL cholesterol and LpAI:AII (110) and a selective decrease in LpAI in individuals with HDL cholesterol <40 mg/dL (1.03 mmol/L) (112); In the Etude Cas-Témoins sur l'Infarctus du Myocarde trial, LpAI was decreased in CHD patients in Northern Ireland; however, both LpAI and LpAI:AII were decreased in patients in France (113). In the Prospective Epidemiologic Study of Myocardial Infarction study of 8784 individuals from France and Northern Ireland, regression logistic analyses showed that apo AI was a stronger predictor for the risk of CHD than HDL cholesterol, LpAI, and LpAI;AII (114). Quantification of LpAI and LpAI:AII in the Framingham Offspring study and VA-HIT clinical trial did not differentiate a subset of individuals with increased risk of CHD after adjustment for lipid and non-lipid CHD risk factors (32, 50). The variability in the results observed in analyses of LpAI and LpAI:AII in the various clinical studies may reflect potential heterogeneity in HDL particles in different patient groups as well as the diverse methods used to quantify HDL subclasses. A general conclusion resulting from these studies is that increased HDL cholesterol concentrations involving an increase in the large HDL subclasses containing both LpAI and LpAI:AII is associated with decreased CHD risk, whereas the reduction in lipid-poor LpAI and pre-β HDL is associated with increased CHD risk (115).

In healthy individuals, LpAI is catabolized at a faster rate than LpAI:AII (116). The major sites of catabolism of the protein components of LpAI and LpAI:AII are the liver and kidney, and the majority of HDL cholesterol is transported to the liver. Kinetic models incorporating the different rates of catabolism of LpAI and LpAI:AII have been developed (117, 118). The differential rates of metabolism of LpAI and LpAI:AII have been proposed to be related to the decreased ability of apo AI–containing lipoproteins to reassociate with HDL particles following cholesteryl ester delivery to the liver via the SR-BI receptors. Lipid-poor apo AI particles, in contrast to apo AII–containing HDL particles, are rapidly catabolized by the kidney, leading to an increased fractional catabolic rate (119).

Genetic absence of plasma LpAI resulting from a molecular defect in apo AI results in increased CVD (120–122), whereas a genetic defect in apo AII resulting in LpAI:AII deficiency was not associated with a major clinical phenotype (11, 123). Increased catabolism of LpAI and LpAI:AII leading to low HDL cholesterol concentrations is characteristic of Tangier disease and LCAT deficiency. Genetic deficiency of the ABCA1 transporter (Tangier disease) is associated with decreased cholesterol efflux and poor lipidation of pre-β HDL, resulting in accelerated LpAI catabolism and increased CVD (124). In contrast, in LCAT deficiency there is effective cholesterol efflux from cholesterol-loaded macrophages followed by defective maturation of pre-β HDL to α HDL; accelerated catabolism, primarily of LpAI:AII as well as LpAI; and renal disease, but no increased risk of CVD (125, 126). A unique lipoprotein particle, LpAI:AII:E, with attenuated catabolism relative to LpAI, is present in patients with CETP deficiency and markedly increased plasma HDL cholesterol concentrations (127, 128). Decreased catabolism of LpAI and LpAI:AII with increased HDL cholesterol concentrations has also been reported in patients treated with CETP inhibitors (52). Decreased plasma levels of LpAI and LpAI:AII are seen in hypertriglyeridemic patients because of increased catabolism of triglyceride-enriched LpAI and LpAI:AII (129, 130).

Statin drugs are associated with a modest 5%–7% increase in HDL cholesterol concentrations due to complex alterations in LpAI and LpAI:AII metabolism, including increased synthesis and decreased catabolism of apo AI as well as a decrease in CETP activity (131–133). In an analysis of the Voyager database that included 37 randomized clinical trials involving 32 258 patients, the percentage increase in HDL cholesterol and apo AI associated with the administration of atorvastatin, simvastatin, and rosuvastatin were similar (134). In human clinical studies fenofibrate administration resulted in an increased synthesis of apo AII with a minimal increase in apo AI synthesis; overall, increases in LpAI:AII predominated (131, 133, 135). Recent human kinetic studies with niacin have demonstrated both an increased synthesis and catabolism of apo AI and apo AII resulting in the preferential formation and accumulation of large LpAI particles and increased HDL cholesterol concentrations (53).

apo E

apo E is the most avid ligand for the LDL receptor, and as such drives catabolism and clearance of apo B– containing lipoproteins (136, 137). apo E is also a prominent HDL protein with unique functions in the HDL compartment (138, 139). For example, swine and dogs fed a high-fat diet accumulate large apo E–rich HDL that can deliver cholesterol to the liver directly via interaction with the LDL receptor (140, 141). In the presence of apo E, HDL particles undergo a core expansion due to their enhanced capacity to carry cholesterol (142). In addition, apo E interacts with ABCA1 to extract cellular cholesterol out of the cell, and drives the formation of larger HDL-sized particles from macrophage foam cells (143, 144). Because atheromatous plaques are only partially permeable to plasma solutes, such as apo AI, but rich in locally secreted proteins, such as apo E, arterial macrophages may represent cells in a unique microenvironment in which cholesterol efflux is directed more toward apo E–containing particles rather than toward classical apo AI–containing particles (145, 146).

In humans, apo E HDL concentrations are lower than in animals lacking CETP and vary with fasting state and apo E phenotype (147–149). Interestingly, both CETP-deficient patients (149, 150) and individuals treated with a CETP inhibitor (151) have increased apo E HDL concentrations (48, 150). The apo E–rich HDL of CETP-deficient individuals appears to be a strong acceptor of ABCG1-derived cholesterol from loaded macrophages. Finally, apo E inhibits the displacement of hepatic lipase from the endothelial surface (152), thus reducing hepatic lipase-mediated triglyceride hydrolysis of HDL and the affinity of HDL for its docking receptor, SR-BI (153). These observations suggest a scenario in which, in conditions of diminished CETP activity, the HDL uses apo E for core expansion and for direct hepatic delivery of lipid cargo via the LDL receptor (154). However, it remains unclear whether HDL particles present in CETP-deficient patients or generated by use of CETP inhibitors are removed by the LDL receptor, SR-BI, or a combination of both receptors.

The role of apo E HDL in atherosclerosis is difficult to assess, mostly because of the dominant effects of apo E on whole-body cholesterol metabolism and lipoprotein disposition (155). In experimental atherosclerosis, apo E is a potent antiatherogenic agent, not exclusively because of its effects on plasma lipids. Small amounts of macrophage-derived apo E completely correct both the dyslipidemia and the susceptibility to atherosclerosis in the apo E–deficient mouse model (156). More importantly, introduction of apo E in the plasma or vascular wall in quantities that are insufficient to change plasma lipid concentrations still produces significant vascular protection, thereby suggesting local effects within the atheroma (157). In contrast, apo E–deficient patients have not been reported to manifest early onset of accelerated atherosclerosis (158–160). In support of the contention that apo E is highly expressed in the atheroma, a recent evaluation of HDL composition by proteomics analysis has shown apo E enrichment in the HDL3 particles of individuals with CHD (161). This result may provide impetus for the development of methods aiming at validating HDL apo E as a biomarker to predict the presence of atheroma.

apo M HDL

Genetic manipulation studies in mice indicate that apo M plays an important role in the remodeling of plasma HDL, formation of pre-β HDL, and reverse cholesterol transport and it is a potent antiatherogenic protein (162, 163). apo M is a minor apolipoprotein found in approximately 5% of total HDL (see section on Proteomics below) and approximately 2% of LDL. apo M is positively correlated with HDL and LDL cholesterol in both healthy individuals and CVD patients (164). In 2 case control studies, however, no significant differences were observed in plasma apo M concentrations in apparently healthy individuals and CHD patients (165).

apo B HDL

apo B peptides have been found during shotgun proteomics analyses of isolated human HDL (161, 166), but their presence has been dismissed as due to either contaminating LDL or the presence of lipoprotein (a), the hydrated density of which overlaps that of HDL. Recent studies focused on hepatic microsomal triglyceride transfer protein (MTP) activity in mouse models have shown that the phospholipid transfer activity of MTP helps in assembly and secretion of VLDL- and HDL-sized particles containing apo B100 and apo B48 (167). Low concentrations of these particles are secreted and may be detected in the plasma.

PROPOSED NOMENCLATURE

As discussed above, use of different techniques and procedures has led to different terms in defining HDL species. To provide guidelines for future studies and to compare and contrast published data obtained by use of different methods, we propose a new HDL nomenclature based on density and size of the particles (Table 2). In addition, we compare these terms with other designations available in the literature. In this nomenclature, HDL particles are termed very large, large, medium, small, and very small.

Classification of HDL by physical properties.^a

Table 2.

Classification of HDL by physical properties.^a

Proposed term	Very large HDL (HDL-VL)	Large HDL (HDL-L)	Medium HDL (HDL-M)	Small HDL (HDL-S)	Very small HDL (HDL-VS)
Density range, g/mL	1.063–1.087	1.088–1.110	1.110–1.129	1.129–1.154	1.154–1.21
Size range, nm	12.9–9.7	9.7–8.8	8.8–8.2	8.2–7.8	7.8–7.2
Density gradient ultracentrifugation	HDL2b	HDL2a	HDL3a	HDL3b	HDL3c
Density range, g/mL	1.063–1.087	1.088–1.110	1.110–1.129	1.129–1.154	1.154–1.170
Gradient gel electrophoresis	HDL2b	HDL2a	HDL3a	HDL3b	HDL3c
Size range, nm	12.9–9.7	9.7–8.8	8.8–8.2	8.2–7.8	7.8–7.2
2-D gel electrophoresis	α-1	α-2	α-3	α-4	Pre-β-1 HDL
Size range, nm	11.2–10.8	9.4–9.0	8.5–7.5	7.5–7.0	6.0–5.0
NMR	Large HDL-Pb	Medium HDL-P		Small HDL-P
Size range, nm	12.9–9.7	9.7–8.8	8.8–8.2	8.2–7.8	7.8–7.2
Ion mobility	HDL2b	HDL2a and HDL3
Size range, nm	14.5–10.5	10.5–7.65

a

One-dimensional electrophoresis was performed in nondenaturing gradient polyacrylamide gel (4%–20%).

b

HDL-P, HDL particle.

Proteomics and Lipidomics: An Integrated View of HDL Biology

PROTEOMICS

The advent of the wider availability of mass spectrometric technologies, and their applicability to analysis of multicomponent protein mixtures, has heralded a surge of interest in the proteome of human HDL particles in health and disease.

Several factors are of central importance when studies of the HDL proteome are undertaken, and these include the nature of the starting biological material and its conservation, the method used for HDL particle separation and purification, and the type of mass spectrometric analysis applied. No systematic study of the impact of such factors on HDL isolation by different methods, and thus potentially on the HDL proteome, has been undertaken to date.

The working criterion used to define the HDL fraction under study is a key determinant of the HDL proteome. The choice of isolation/fractionation procedures is listed in Table 3; the precise nature of the HDL isolated by any of these techniques requires rigorous analysis prior to initiation of proteomic studies. Indeed, HDL isolated by fast-performance liquid chromatography is heavily contaminated by high molecular weight plasma proteins that coelute with HDL (168). Until now, ultracentrifugation has been the predominant method used for HDL isolation to study the HDL proteome.

Preparative HDL isolation/fractionation techniques.

Table 3.

Preparative HDL isolation/fractionation techniques.

Flotational ultracentrifugation Zonal ultracentrifugation Density gradient ultracentrifugtion (isopynic, NaCl/KBr; D2O, sucrose) Precipitation Size exclusion chromatography Ultracentrifugation/HDL density 1.063–1.21 Fast protein liquid chromatography/whole plasma Ion exchange chromatography Immunoaffinity chromatography Electrophoresis (2-D gel electrophoresis, etc.)

Once HDL is purified, the mass spectrometric technologies that have been employed to define the HDL proteome include SELDI-TOF, MALDI-TOF, shotgun and nano–liquid chromatography electrospray ionization mass spectrometry, and most recently, a shotgun approach involving linear ion trap quadrupole–Fourier transform ion cyclotron resonance mass spectrometry with a nanoelectron spray source. To different degrees, difficulty in quantifying proteins (as tryptic peptides), particularly those in low abundance, is a limitation of all of these technologies.

In one of the first comprehensive analyses of the HDL proteome, Vaisar et al. (161) identified some 50 protein components in human HDL3 isolated by ultracentrifugation. The biological activities of these proteins suggested that HDL contributes not only to lipid metabolism and cholesterol homeostasis, but also to complement regulation, the acute-phase response, and inhibition of proteolytic enzymes. Several other reports have confirmed the presence in HDL of multiple apolipoproteins (AI, AII, AIV, B, (a), CI, CII, CIII, CIV, D, E, F, H, J, L1, and M), in addition to α1-antitrypsin inhibitor; albumin; complement C3 and C4; fibrinogen; haptoglobin-related protein; paraoxonase 1 and 3; serum amyloid A1, A2, and A4; and transthyretin [data summarized in (169)].

Intriguingly, the plasma abundance of most of these proteins is insufficient to allow 1 copy per HDL particle, thereby suggesting that specific proteins may be bound to distinct particle species that are differentially distributed across the HDL particle spectrum. On this basis, it was plausible to expect that the multiple biological functions of HDL may be mediated by distinct particle subspecies defined by specific cluster(s) of bound proteins, and that such protein clusters cofractionate upon isolation of HDL subpopulations. As an initial step toward assessment of this hypothesis, plasma HDL from normolipidemic individuals was subfractionated by isopycnic density gradient ultracentrifugation into 5 subfractions (Fig. 3); the composition of their proteome was then evaluated by tandem mass spectrometry technology (166).

Five distinct patterns of distribution of individual protein components were observed across the HDL density subfractions; the most interesting of these identified small dense HDL (HDL3c) as a particle subpopulation in which 7 proteins occurred predominantly: apo J, apo L1, apo F, paraoxonase 1/3, phospholipid transfer protein, and platelet-activating factor acetylhydrolase (also termed lipoprotein-associated phospholipase A₂). The HDL3c proteome also contained apo AI; apo AII; apo D; apo M; serum amyloids A1, A2, and A4; apo CI; apo CII; and apo E.

The unique proteome of HDL3c has functional implications because this particle species exhibited the greatest potency among HDL subpopulations to protect LDL against oxidation. Such activity was highly correlated with the presence of apo J, apo M, serum amyloid A4, apo D, apo L1, and paraoxonases 1/3 in HDL3c.

These data should not be interpreted to indicate that all of the proteins detected in HDL3c are present on the same lipoprotein particle; indeed, the isolation of a unique particle containing the trypanosome lytic factor apo L1, plus apo AI and haptoglobin-related protein in the HDL3 density range suggests that this is certainly not the case (169), and that the HDL3c fraction consists of several species of HDL particles with distinct proteomes. On the basis of these findings, it may be concluded that: (a) the detected protein clusters described herein are potentially indicative of distinct subspecies of HDL particles that display specific biological function(s); (b) the proteomic analyses of defined HDL subspecies isolated by isopycnic density gradient ultracentrifugation from normolipidemic plasma samples have identified small dense HDL3c as a distinct particle subset(s); and (c) specific lipid and protein components of HDL3c endow them with potent antioxidative activities. Finally, these data support the concept that HDL may serve as a platform for the assembly of certain protein components that perform specific function(s), and that (apolipo)proteins may form the basis for functional heterogeneity of HDL.

It is of special interest to know whether the proteome of HDL may be altered in metabolic diseases characterized by dyslipidemia and increased CVD risk. If this were the case, then specific proteins could be used as biomarkers of altered HDL function. Indeed, it is now established that several of the major biological, antiatherogenic activities of HDL are attenuated in type 2 diabetes and metabolic syndrome, each of which is associated with high CVD risk (12). Furthermore, under conditions of acute inflammation, HDL particles are enriched in serum amyloid A, resulting in defective antiinflammatory activity of HDL (12).

Vaisar et al (161) and Greene et al (170) reported the first studies to detect modification of the HDL proteome in patients displaying incident CHD, and the most consistent finding was an increment of 150% in apo E content. This alteration in the proteome was normalized by combined treatment with a statin and niacin. These studies open new horizons not only for identification of protein biomarkers of altered HDL metabolism and function but also for well-targeted pharmacotherapies to correct them.

In summary, the precise nature of the HDL proteome critically depends on the method of HDL isolation and purification, and equally on the mass spectrometric technology employed for protein analysis and tryptic peptide quantification; structural and functional analysis of HDL particle subspecies may prove more informative than analyses of traditional total HDL; agreement on standardization of methods for HDL isolation from human plasma is especially important to define the key characteristics of HDL particles.

LIPIDOMICS

When HDL particles are evaluated for their content of cholesteryl ester and phosphatidylcholine (PC), cholesteryl linoleate predominates among cholesteryl ester, whereas the 18:2/16:0, 18:2/18:0, and 20:4/16:0 represent the most common PC fatty acids (171). Consistent with the above data, particle content of cholesteryl ester, free cholesterol, and phospholipid subclasses including PC, phosphatidylethanolamine, phosphatidylinositol, sphingomyelin (SM), and lysoPC progressively decreases with increase in hydrated density from HDL2b to HDL3c (171). However, no such differences are evident between HDL subspecies when data for cholesteryl ester, PC, phosphatidylethanolamine, phosphatidylinositol, and lysoPC are expressed as a percentage of total lipids, suggesting that their molecular species are in dynamic equilibrium between HDL subpopulations. In a similar fashion, when lipid moieties of HDL are analyzed on the basis of their total fatty acid content, the percentage distributions of saturated, monounsaturated, and polyunsaturated n-6 and n-3 fatty acids are indistinguishable between HDL particle subspecies (171).

However, the proportion of SM relative to total lipids decreases progressively in parallel with HDL density from 12.8% in HDL2b to 6.2% in HDL3c. Consequently, the SM/PC molar ratio decreases from 0.38 in HDL2b to 0.18 in HDL3c. The distinctly low SM content in HDL3c suggests that this pool is not in equilibrium with that of other HDL subpopulations, consistent with the slow rate of exchange of SM between lipoproteins and cell membranes (172). The low SM/PC ratio may reflect a distinct cellular origin(s) of small HDL as suggested by the low SM content of small nascent HDL particles secreted by J774 macrophages, which originate from the exofacial leaflet of the plasma membrane (173).

Similarly to SM, free cholesterol content decreases twofold from HDL2b to HDL3c (171, 174). As a result, the cholesteryl ester/free cholesterol ratio significantly increases with HDL density, supporting the contention that small HDL constitutes a major site of cholesterol esterification within the HDL particle spectrum (175). Increased LCAT activity and diminished SM/PC ratio in HDL3c are consistent with this proposal, because SM functions as a physiological inhibitor of LCAT (176, 177).

Among the minor bioactive lipid components, the abundance of sphingosine-1-phosphate (S1P) per HDL particle is asymmetric across the HDL spectrum, with preferential enrichment in HDL3 (40–50 mmol/mol HDL) compared to HDL2 subfractions (15–20 mmol/mol) (171, 177, 178). Enrichment of small HDL3 in S1P might be mechanistically related to the potent capacity of such particles to acquire polar lipids of cellular origin (171).

The heterogeneity of the HDL lipidome may translate into distinct functionality of HDL subpopulations. Indeed, small, dense HDL3 exhibit more potent antioxidative activity and antiinflammatory activities compared to large, light HDL2 independently of the concentration basis employed for such comparison (total protein, total mass, or particle number) (179). Furthermore, small, dense HDL3 exhibit more potent capacity to protect human microvascular endothelial cells from apoptosis induced by oxidized LDL compared to large, light HDL2 irrespective of comparison method (177, 178). Finally, small HDL particles represent more avid acceptors of cellular cholesterol via the ABCA1-dependent pathway (173, 175).

Studies of the mechanistic aspects of the potent antioxidative activity of HDL3 particles have revealed that their capacity to inactivate LDL-derived lipid hydroperoxides critically depends on the surface lipid fluidity that is primarily determined by the HDL lipidome, and notably by the SM/PC ratio (179). The enhanced fluidity of the surface monolayer of small HDL3 particles related to the low abundance of SM may equally contribute to their increased cellular cholesterol efflux capacity. Finally, the potent capacity of HDL3 to protect endothelial cells from apoptosis may in part reflect HDL3 enrichment in S1P, a minor bioactive lipid (171, 179).

Available data thereby suggest that lipidomic analyses of HDL particles can be used to obtain information regarding antiatherogenic functionality of HDL. Data are presently available on the relationship between the HDL lipidome and cardiovascular risk, and this information may prove useful for the evaluation of new HDL-increasing agents.

Summary

A growing body of evidence from epidemiological data, studies in animal models, and available clinical trial data supports the contention that HDL represents the next therapeutic target for reduction in the residual risk typical of statin-treated, high-risk cardiovascular patients. Measurement of HDL cholesterol has been employed as the principal method to assess the role of HDL as a risk factor for CVD. The physicochemical and functional heterogeneity of HDL presents an important challenge to the cardiovascular field for development of more effective laboratory and clinical methods to quantify HDL with a predictive value in assessing cardiovascular risk. Moreover, the metabolic and clinical associations between low concentrations of HDL cholesterol and increased concentrations of cholesterol-depleted HDL particles and cholesterol-enriched triglyceride remnants must be considered in CVD risk assessment.

The initial gold standard for the separation of plasma HDL was analytical ultracentrifugation with separation of HDL initially into HDL2 and HDL3 with further resolution into HDL2a, HDL2b, and HDL3. The identification of the density profile of HDL provided the information essential for the development of preparative methods to isolate, subfractionate, and characterize HDL. Concomitantly, characterization of HDL particles by size was achieved with gradient gel electrophoresis, which separates HDL into HDL2b, HDL2a, HDL3a, HDL3b, and HDL3c.

Further resolution of HDL particles by 2-D gel electrophoresis into pre-β HDL and α1–α4 HDL has been extremely useful in the characterization of HDL from animal models, clinical trials using different drugs, and genetic defects in lipoprotein metabolism. The observation of the metabolic maturation of pre-β HDL into HDL α1 to α4, the identification of the predictive value of reduction in α1 and cardiovascular risk, and the HDL profile in the genetic dyslipoproteinemias provided important new insights into HDL structure and metabolism.

A major advance in the assessment of HDL has been the development of methods to quantify HDL particle number. The new ion mobility method to quantify plasma apo B lipoproteins as well as HDL is rapidly progressing and will be useful in the clinical assessment of the plasma lipoproteins. NMR holds great promise for the quantification of the number of HDL particles in clinical samples, analogous to the particle number quantification of apo B–containing lipoproteins by using apo B immunoassay or NMR. The ability to correlate HDL particle number with HDL cholesterol, as well as the possibility of eliminating the confounding influence of the apo B–containing lipoproteins in the assessment of the potential association with clinical events, will provide new insights into the role of HDL in cardiovascular disease. Recent data from the VA-HIT and Multi-Ethnic Study of Atherosclerosis cIMT (carotid intima-media thickness) clinical trials substantiate the potential importance of this new approach to HDL particle quantification and CVD risk.

The development of new mass spectrometric methods has provided the unique opportunity to determine the protein composition of HDL and its constituent subfractions. In addition to the classic apolipoprotein, HDL contains proteins associated with inflammation, clotting, and complement regulation as well as proteolytic enzymes. Of particular interest was the finding that there were clusters of proteins on separate HDL particles, which gave rise to the hypothesis that specific subsets of HDL particles may exert specific function(s). In this regard, the unique proteome of HDL3c is particularly effective in protecting LDL against oxidation. The lipid components of HDL also exhibit marked heterogeneity. Thus the ratio of cholesteryl ester/free cholesterol and PC/SM differs among HDL subfractions, and the ratio of PC/SM in the smallest HDL3c particles markedly influences LCAT activation, the surface rigidity of the HDL particle, and, potentially, protein composition. In addition, bioactive lipid components such as S1P are preferentially associated with the HDL3c particle subset.

The combined findings from multiple analytical procedures used to characterize HDL support the concept that the marked physicochemical heterogeneity of HDL particles is the underlying basis for their functional heterogeneity. Further structural and compositional analysis of HDL particles may provide additional information on the identification of HDL particles with specific unique functions. Equally, studies at the molecular level possess the potential not only to reveal new risk biomarkers, but also to identify new pharmacotherapeutic targets to reduce atherosclerosis and CVD.

To facilitate the future characterization of HDL subfractions, development of a new uniform nomenclature for the subfractions of HDL is critical, and is proposed in this manuscript (Table 3). This classification system defines 5 HDL subclasses on the basis of physical and chemical properties, and assigns very large HDL particles to the largest subclass, and large HDL, medium HDL, small HDL, and very small HDL to the smallest and most dense subclasses. The very small HDL subclass includes pre-β, discoidal, or nascent HDL. The nomenclature will be tested through analysis of split samples by using the various methods described in this report. We anticipate that the development of a uniform nomenclature for HDL subfractions will increase our ability to compare data obtained with different methodologic approaches for HDL fractionation, and to assess the clinical effects of different agents, which modulate HDL particle structure, metabolism, and function, and thus CVD risk. Prospective studies will be essential in establishing the associations between HDL subclasses and CVD that will be revealed by using these various methodologies (180).

11 Nonstandard abbreviations:

 
• CVD

  cardiovascular disease

 
• CHD

  coronary heart disease

 
• 2D

  2-dimensional

 
• VAP

  vertical auto profile

 
• apo AI

  apolipoprotein A-I

 
• ABCA1

  ATP binding cassette transporter A1

 
• ABCG1

  ATP transporter G1

 
• LCAT

  lecithin:cholesterol acyltransferase

 
• CETP

  cholesteryl ester transfer protein

 
• NMR

  nuclear magnetic resonance

 
• HDL-P

  HDL particles

 
• IDEAL

  Decrease in Endpoints through Aggressive Lipid Lowering

 
• VA-HIT

  Veterans Administration HDL Intervention Trial

 
• LpB

  lipoprotein B

 
• SR-BI

  scavenger receptor class B type I

 
• CE

  cholesteryl ester

 
• PC

  phosphatidylcholine

 
• SM

  sphingomyelin

 
• S1P

  sphingosine-1-phosphate.

Author Contributions:All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.

Authors' Disclosures or Potential Conflicts of Interest:Upon manuscript submission, all authors completed the Disclosures of Potential Conflict of Interest form. Potential conflicts of interest:

Employment or Leadership: M.J. Chapman, European Atherosclerosis Society; M.M. Hussain, Chylos; J.D. Otvos, LipoScience; E.J. Schaefer, Boston Heart Laboratory.

Consultant or Advisory Role: R.S. Rosenson, Abbott Labs, Anthera, LipoScience, Residual Risk Reduction Initiative, Grain Foods Board, and Roche Genentech; H.B. Brewer, Jr., Merck, Merck-Schering-Plough, Schering-Plough, Roche, AstraZeneca, and Lilly; M.J. Chapman, Merck, Pfizer; S. Fazio, Merck; R.M. Krauss, Merck, Roche, Metabolex, Corcept Pharmaceuticals, Celera, and Gilead; E.J. Schaefer, AstraZeneca, Arisaph, DuPont, Merck, Unilever, and Vatera.

Stock Ownership: R.S. Rosenson, LipoScience; J.D. Otvos, LipoScience; E.J. Schaefer, Boston Heart Laboratory.

Honoraria: R.S. Rosenson, Abbott Labs, Anthera Pharmaceuticals, LipoScience, Residual Risk Reduction Initiative, Roche-Genentech, XOMA, and Grain Foods Board; H.B. Brewer, Jr., Merck, Merck-Schering-Plough, Schering-Plough, Roche, AstraZeneca, and Lilly; S. Fazio, Merck; M.M. Hussain, Merck, GlaxoSmithKline, and Pfizer; A. Kontush, Novo Nordisk; R.M. Krauss, Merck, Roche, Metabolex, Corcept Pharmaceuticals, Celera, and Gilead; E.J. Schaefer, AstraZeneca, Arisaph, DuPont, Merck, Unilever, and Vatera.

Research Funding: M.J. Chapman, Merck; S. Fazio, ISIS Genzyme; M.M. Hussain, NIH; A. Kontush, Pfizer, AstraZeneca, and GlaxoSmithKline; E.J. Schaefer, DuPont.

Expert Testimony: None declared.

Other: R.M. Krauss, coinventor of 2 patents for lipoprotein subfraction analysis.

Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.