The concept that high-density lipoproteins (HDL) could protect against coronary heart disease (CHD) primarily originated from epidemiological studies of the healthy population, in particular the Framingham study [1]. Patients with manifest CHD frequently exhibit low HDL cholesterol (HDL-C) [2]. The relationship between HDL-C and cardiovascular risk is not linear; for instance, no further improvement in prognosis is seen with HDL-C levels above ~60 mg/dl (1.5 mmol/l) (Fig. 1) [3]. Retrospective investigations of the EPIC Norfolk and of the IDEAL study show that very high concentrations of HDL-C may be associated with increased risk [4]. A recent register study of more than 1 million US veterans found a U-shaped relationship between HDL-C and total mortality with 50 mg/dL (1.25 mmol/L) as the nadir associated with the lowest mortality [5]. A very recent analysis of the Framingham study reports that the predictive value of HDL-C is modified by LDL cholesterol (LDL-C) and triglycerides (TG): Compared to low HDL-C (defined as <50 mg/dl in women and <40 mg/dl in men) in isolation, risk increases when low HDL-C occurs together with high LDL-C and/or TG. Cardiovascular (CV) risk increases by 30% for LDL-C ≥ 100 mg/dl and TG < 100 mg/dl or LDL-C < 100 mg/dl and TG ≥ 100 mg/dl. When both TG and LDL-C are ≥100 mg/dl, CV risk increases by 60% [6].

HDL are the smallest (5–17 nm) and densest (1.063–1.210 kg/l) lipoproteins in the plasma. Apolipoprotein (Apo) A1, the major protein in HDL, is synthesised in the liver and the small intestine. The liver is the most important organ through which cholesterol is excreted, either directly or after being converted into bile acids. Excess cholesterol is transported from the periphery (e.g. from macrophages in blood vessel walls) to the liver. HDL play a key role in this pathway, known as reverse cholesterol transport (RCT) (Fig. 2) [8‐11]. In addition to HDL, LDL also significantly contribute to RCT. The overwhelming majority of HDL-C measured in the blood originates from the liver and the intestine. Therefore, the concentration of HDL-C in the plasma cannot be used as a measure of cholesterol efflux from vessel walls, or of the efficiency of RCT.

The RCT begins with the transfer of cholesterol from cell membranes to HDL. To date, four biochemical pathways have been described that are involved in this transfer [8]. As long as the cholesterol content of the cell is normal, more than two-thirds of the (poorly water-soluble, non-esterified) cholesterol leave the cells by passive diffusion along a concentration gradient between the cell membrane and preferably large, globular HDL. This concentration gradient is maintained by extracellular esterification of free cholesterol mediated by the lecithin cholesterol acyltransferase (LCAT).

Secondly, the passive, aqueous efflux of non-esterified cholesterol can be further enhanced by up-regulation of ATP-binding cassette transporter G1 (ABCG1). ABCG1 mobilises cholesterol from subcellular compartments.

Thirdly, passive, but non-aqueous transfer of free cholesterol from cells to larger HDL can be mediated by the scavenger receptor class B type I (SR-B1). SR-B1 promotes not only the cellular cholesterol efflux, but also the “selective” (i.e. non-endocytotic) delivery of cholesterol from HDL to liver cells [12].

Fourthly, cholesterol and phospholipids from macrophages and foam cells are actively transferred to ATP-binding cassette transporter A1 (ABCA1) to lipid-free apo A1 (pre-ß-HDL). This process is important for the formation of “nascent” HDL particles, (i.e. HDL discs), which contain Apo A1, phosphatidylcholine and non-esterified cholesterol. The importance of ABCA1 for the metabolic maturation of HDL was identified by the elucidation of the genetic cause of Tangier’s disease [13, 14]. The expression of ABCA1 is regulated by the intracellular cholesterol content, and efflux mediated by ABCA1 is therefore important in situations of cellular cholesterol excess.

The directed flux of cholesterol between cells and HDL is dependent on a concentration gradient between the cell membrane and acceptor lipoproteins. This gradient is maintained at the cellular level by the hydrolysis of cholesteryl esters and translocation of cholesterol, extracellularly by lecithin cholesterol acyltransferase (LCAT) and the cholesterol ester transfer protein (CETP). The LCAT esterifies free cholesterol (for example, cholesterol associated with pre-ß-HDL) with a fatty acid from the lecithin. Cholesteryl esters are more hydrophobic than free cholesterol and are located inside the HDL. This transforms HDL from discoidal into spherical, pseudomicellular particles (α-HDL). Maturation of HDL is incomplete in genetic or secondary deficiency of LCAT. In patients with familial LCAT deficiency, spherical HDL are absent.

CETP is a hydrophobic glycoprotein which is mainly secreted by the liver and is mostly bound to HDL in the blood [15]. It facilitates the transfer of cholesteryl esters, triglycerides and, to a smaller extent, phospholipids between HDL, LDL and VLDL. Thus, cholesteryl esters are channelled into the metabolism of LDL and delivered to the liver.

The standard method for the determination of cholesterol in LDL and HDL is the combined method of precipitation and ultracentrifugation (beta quantification) [16]. In the first step, VLDL is recovered by means of ultracentrifugation, then LDL is precipitated and finally HDL-C is determined enzymatically in the supernatant. This method is not suitable for use in the clinical laboratory because of high costs and complexity. In the late 1990s, “homogeneous” methods were introduced which allow the measurement of HDL-C in a single reaction vessel with no prior separation steps [17]. In these methods, the cholesterol in non-HDL particles is “masked” with antibodies, polymers or detergents, and in a second reaction step HDL-C is determined enzymatically. The commercially available homogeneous assays show a good agreement with the reference method in normolipidemic samples and in many cases of hypercholesterolemia. The determination of HDL-C is hardly affected by prior consumption of food and the results can usually well be interpreted in post-prandial samples [18]. Differences to the reference method and between the assays are, however, observed in hypertriglyceridemia (elevated concentrations of chylomicrons and/or VLDL), at low HDL-C (<20 mg/dl), and if atypical lipoproteins are present (type III hyperlipoproteinemia, liver disease, chronic kidney disease) [19‐21]. In these cases, HDL-C can be accurately measured by lipoprotein electrophoresis, however by staining lipoproteins using enzymatic cholesterol oxidation rather than unspecific lipophilic chemicals.

Due to the limitations of conventional lipoprotein analysis, researchers are considering the replacement of HDL-C and LDL-C determinations with the determination of Apo A1 and ApoB, respectively. Measurement of apolipoproteins can be readily standardised [22] and is not significantly distorted by high triglycerides. Apo B not only reflects the concentration of LDL, but all atherogenic lipoproteins taken together [VLDL, remnants, LDL, Lp(a)].

In recent years, assays have been introduced which reflect the functionality of HDL better than HDL-C. Of importance are the determination of HDL particle concentrations (and sizes) with nuclear magnetic resonance (NMR) spectroscopy [23], the measurement of HDL proteins (Apo A1, serum amyloid A) and the in vitro measurement of the cholesterol uptake capacity of HDL in cell culture models (cholesterol efflux) [24‐27]. While the latter will most likely remain a research method, NMR and apolipoproteins are in principle suitable for routine use, once their standardisation has been accomplished.

Table 1 provides an overview of rare monogenic disorders in the HDL metabolism. According to the results from Danish and American population studies, approximately 10% of individuals with HDL-C levels below the 5th percentile are heterozygous for mutations in the genes of APOA1, ABCA1 or LCAT [28, 29]. Data on the risk of atherosclerosis in these individuals are contradictory [30]. Some mutations of APOA1 have been associated with increased risk of myocardial infarction [31], and another one, Apo A1-Milano, may reduce the risk. Some mutations of the APOA1 gene also cause familial amyloidosis. In large Danish population studies, heterozygosity for HDL-C-lowering mutations of the ABCA1 or APOA1 gene was not associated with an increased risk of myocardial infarction. Interestingly, ABCA1 mutations increase the risk for Alzheimer’s disease. Data from large Dutch family studies suggest an increased cardiovascular risk of HDL-C-lowering mutations in the genes of APOA1, ABCA1 or LCAT. Italian studies found no evidence for an increased risk of atherosclerosis in heterozygous carriers of LCAT mutations [32‐34].

Homozygous and heterozygous carriers of mutations in these three genes are very rare and present themselves with pronounced or even complete HDL deficiency and characteristic clinical syndromes: patients with complete LCAT deficiency develop corneal opacities, anaemia and progressive renal failure, which ultimately requires renal replacement therapy. Patients with partial LCAT deficiency (“fish eye disease”) develop corneal opacities, but neither renal disease nor anaemia. In both cases, the remaining HDL are discoidal in shape. In classical LCAT deficiency, LDL include particles which are similar to Lipoprotein X and may be the cause of the characteristic nephropathy. The risk of atherosclerosis of mutation carriers is not increased or at most slightly increased [35], possibly because HDL can also transport free cholesterol directly to the liver [36].

Homozygosity for mutations in the ABCA1 gene leads to Tangier disease [13, 14]. As a result of the strongly reduced cellular cholesterol efflux, cholesterol accumulates in macrophages in various organs (enlarged tonsils, hepatosplenomegaly and peripheral neuropathy). Patients with Tangier disease often also have low LDL-C, which could attenuate the atherogenic effect of the low HDL concentration [Schaefer, 2010 # 21]. Homozygous nonsense mutations in the APOA1 gene were found to [28] be the cause of HDL deficiency in patients with severe xanthomatosis and early atherosclerosis.

Most patients with a deficiency of cholesteryl ester transfer protein (CETP) have been observed in Japan. Affected persons have very high concentrations of HDL-C. Pharmacological inhibition of CETP has therefore been tested as an approach to increase HDL. It is unclear whether the HDL produced through CETP inhibition have a normal function. We observed a slightly increased cardiovascular mortality at low concentrations of CETP in the blood despite high HDL-C [37].

Only recently, mutations of scavenger receptor (SR-BI) (P279S, P376L) have been described which are associated with both significant increases in HDL-C and increased cardiovascular risk [38‐40]. These mutations impair the selective uptake of cholesteryl esters by hepatocytes. In mice, the overexpression of SR-B1 leads to an increase in RCT, despite a drop in HDL-C [41]. On the other hand, lack of SR-BI increases HDL-C, but is atherogenic [42]. High HDL-C therefore does not provide protection from atherosclerosis in all cases.

This stands in contrast to genetic factors which increase LDL cholesterol, because monogenetic (familial hypercholesterolemia) and polygenetic factors increasing LDL cholesterol consistently lead to concordant changes in the CV risk (positive Mendelian randomisation) [43, 44].

Compared to genetically determined abnormalities in HDL metabolism, low HDL-C much more frequently occur in patients with metabolic syndrome or diabetes mellitus. Low HDL-C levels are also associated with systemic inflammation, e.g. with cigarette smoking, chronic inflammatory diseases or chronic kidney disease (Fig. 3) [45, 46]. In cases of extremely low HDL-C, rare diagnosis may be considered, e.g. neoplasia or an increased risk for sepsis [47, 48].

Evidence has long been available from animal experiments for anti-atherosclerotic effects of HDL. Administration of homologous HDL in addition to an atherogenic diet inhibits the formation of fatty streaks in rabbits [49]. Infusion of Apo A1 Milano (R173C, complexed with phospholipids) inhibits atherogenesis in cholesterol-fed rabbits [50] and apo E-deficient mice [51].

Alongside its prominent role in RCT, anti-oxidative and anti-inflammatory effects are also attributed to HDL, as well as improvement of endothelial function [9, 52, 53]. They inhibit the pathological attachment of monocytes to the endothelium and support endothelial repair mechanisms. Many of the endothelial effects of HDL are mediated by the endothelial SR-B1 receptor and by the bioactive lipid sphingosine-1-phosphate that can act via the S1P3 receptor on endothelial cells [54]. Investigations of the composition of HDL [55, 56] have found enzymes, acute-phase proteins, components of the complement system and protease inhibitors in addition to the characteristic apolipoproteins associated with HDL. The lipidome of HDL is even more heterogeneous, and HDL can also function as the vehicle for microRNA in the blood [57]. Paraoxonase 1 (PON1) is an esterase associated with HDL, which inhibits the formation of lipid peroxides in LDL [58, 59] and HDL. Compared to healthy people, the activity of HDL-associated PON1 is reduced in patients with CHD [60]. An anti-oxidative effect is also attributed to Apo A1 itself [61]. HDL are anti-inflammatory in several ways: they inhibit the expression of adhesion molecules (VCAM-I and ICAM-I) in the endothelium, inhibit the terminal complement complex, inhibit the synthesis of chemokines (MCP-1) and induce the expression of the anti-inflammatory cytokine transforming growth factor beta 2.

In addition, HDL can have antithrombotic and profibrinolytic effects [62]. HDL protect from cell damage, necrosis and apoptosis. These protective effects have also been demonstrated in pancreatic beta cells, implying that HDL might be able to improve insulin secretion and protect from the development of diabetes mellitus [63‐68].

In summary The anti-inflammatory, cytoprotective and wound-healing effects of HDL make them a part of the innate host defense system [69]. From an evolutionary point of view, the potential anti-atherogenic effects, if any of them exist, would likely be secondary.

Interestingly, several studies have provided evidence that the vascular effects of HDL are variable and hardly correlate with HDL-C concentrations in the plasma. In patients with diabetes mellitus, coronary disease, chronic renal insufficiency, cardiovascular risk factors and disorders, the function of HDL is impaired [9, 61, 70‐74]. In patients who underwent coronary angiography, we have observed that HDL-C correlates inversely with cardiovascular mortality in the absence of CHD, that the relationship of HDL-C to long-term prognosis was weakened in stable coronary heart disease, whereas in clinically unstable patients the relationship was completely abolished [74].

The protein composition of HDL changes significantly in the course of an acute-phase reaction. The “acute-phase HDL” are characterised, for example, by increased concentrations of serum amyloid A (SAA), secretory phospholipase A2 (sPLA2-IIa) and ceruloplasmin, and a lower proportion of Apo A1 [73]. Unlike HDL from healthy subjects, HDL from patients with CHD, kidney disease or diabetes mellitus have no protective vascular effects or may even cause paradoxical harmful effects [55, 56, 61, 70, 71, 75, 76]. The characterisation of the HDL’s “milieu” may offer an opportunity for determining the function of HDL. For example, HDL-C corrected by the SAA concentration correlates with risk better than HDL-C alone in several independent cohorts [77].

Another way to determine the functionality of HDL in the laboratory, although one that requires more efforts, is measuring cellular cholesterol efflux. In this method, cultivated cells are enriched with labelled cholesterol and incubated with Apo B-free serum from the patient, and the rate of transfer of labelled cholesterol into the patient serum is determined. The rate of cholesterol efflux has been associated with the prevalence of CHD, independent of the HDL-C concentration [24], with cardiovascular events and deaths (Fig. 4) [25, 26]. In another study, cellular efflux was inversely correlated with prevalent atherosclerosis, but not with the long-term risk of cardiovascular events [78].

The positive vascular effects of HDL have been an attractive target for the treatment of chronic or acute vascular disease. Approaches to increase HDL include lifestyle modifications, conventional pharmacological therapy and novel therapeutic modalities which are still under scrutiny.

HDL play a central role in RCT. Large epidemiological studies have found an inverse relationship between the concentration of HDL-C in plasma and CHD. Mendelian randomisation studies show no protective effects of a genetically high HDL-C concentration. Cholesterol transported with HDL neither has any protective function nor reflects the functionality of HDL. HDL particles as such, as well as specific proteins and lipids in HDL, contribute to the classical functions of HDL in RCT, or mediate their anti-oxidative and anti-inflammatory properties. HDL composition is modified in complex ways in acute and chronic diseases. In this sense, it is misleading to interpret HDL-C as being equivalent to HDL or “good cholesterol”. The determination of pathophysiologically relevant lipids or proteins associated with HDL may come to replace HDL-C in the future, but at present such assays are not yet available for routine clinical use [9, 115]. The following clinical consequences result from these new findings:

Clearly, the “HDL-Cholesterol story” is not over but the recent findings set the stage for future research on this complex lipopoprotein.