The term HDL in fact refers to a highly heterogeneous range of particles within human plasma, ranging in size from 5 to 17 nm and density from 1.063–1.210 kg/l. Alongside cholesterol, lipidomic and proteomic techniques have identified over 200 additional lipids and more than 85 proteins which may reside within the HDL particle [16, 17], as well as multiple enzymes and even genetic material in the form of micro RNAs (Table 1). Significant variation in the relative quantities of these components results in a wide spectrum of circulating HDL particles that may differ in size, density, structure, and function. For ease of assessment, however, these are commonly separated into five broad subclasses, ranging from the largest and least dense (termed HDL_2b) to the smallest and most dense (termed HDL_3b). The major structural protein component of HDL is apolipoprotein A-I (apoA-I), formed in both the liver (~80%) and intestine (~20%) and secreted in a lipid-free state. Efflux of phospholipids and free cholesterol on to at least 2 molecules of apoA-I stimulate the biogenesis of HDL, resulting in the generation of a discoidal lipid-poor ‘nascent’ HDL particle. Subsequent incorporation of numerous enzymes such as lecithin-cholesterol acyltransferase (LCAT), cholesterol ester transfer protein (CETP) and phospholipid transfer protein (PLTP) into this nascent particle further alter the lipid and phospholipid composition—resulting in the generation of mature spherical particles of differing sizes and densities. These particles may be further modified through the incorporation of numerous other structural proteins and enzymes (shown in Table 1) all of which interact to determine the eventual functional capabilities of the HDL particle [18].

The concept of HDL as a shuttle for removing excess cholesterol from lipid-laden macrophages within peripheral tissues was first proposed over 50 years ago [19], and since then has become one of the most well-recognised pathways underlying HDL’s anti-atherogenic nature (Fig. 2). When cell cholesterol content is normal, this process is dominated via the passive aqueous diffusion of free cholesterol between cell membranes and HDL [20]. Although this pathway is bidirectional, esterification of free cholesterol within HDL by the enzyme LCAT maintains a concentration gradient promoting efflux out of the cells, thereby preventing its accumulation in cell membranes and maintaining homeostasis. In addition to passive diffusion, non-aqueous efflux may also occur through the binding of HDL to scavenger receptor class B1 (SR-BI), with this pathway shown to be enhanced when absolute HDL levels are low, or HDL particle size is large [21]. When cell cholesterol content is elevated, however—as is the case in atherogenic lipid-laden macrophages—a shift to active efflux pathways mediated through the upregulation of the ATP-binding cassette transporters ABCA1 and ABCG1 begin to predominate. The most important of these pathways involves the efflux of phospholipids and free cholesterol on to lipid-free or lipid-poor apolipoproteins via ABCA1, as demonstrated by the inability of ABCA1-knockout mice [20] or patients with Tangier’s disease (who lack ABCA1) to generate HDL [22]. Although all HDL-associated apolipoproteins are capable of accepting cholesterol from ABCA1, interactions with ApoA-I account for the majority of the response [23]. As is the case for passive diffusion, LCAT-mediated esterification of cholesterol on these newly formed nascent HDL particles reduces free extracellular free cholesterol levels by trapping them as esters within the HDL core; with this process both maintaining the concentration gradient and resulting in the development of mature spherical HDL particles capable of mediating further efflux through ABCG1 receptors [24]. Once transferred to HDL, cholesterol esters are then returned to the liver for biliary excretion in one of two ways—they may be exchanged for triglycerides in VLDLs and other remnant lipoproteins by CETP, thereby enabling their uptake and excretion via hepatic LDL receptors, or they may be selectively removed from HDL directly via its binding to hepatic SR-BI [25].

Diminished nitric oxide (NO) bioavailability is a hallmark of the vascular disease process, and is thought to underpin every stage of atherosclerosis [26]. The first evidence of HDL’s ability to directly promote endothelial NO production came in the early 2000s, where the addition of HDL to cultured endothelial cells was shown to directly stimulate endothelial NO synthase (eNOS) activation in vitro [27]. In this ground-breaking study, the authors demonstrated an essential role for both ApoA-I and SR-BI in NO generation, as removal of the former using antibodies, or the latter using SR-BI knockout mice, resulted in a loss of effect. However, the use of purified ApoA-I in place of HDL also failed to elicit a response, suggesting a necessary involvement of other components within the HDL particle for eNOS phosphorylation. Later research identified a number of HDL-bound bioactive phospholipids—namely sphingosylphosphorylcholine (SPC), sphingosine-1-phosphate (S1P), and lysosulfatide (LSF)—as at least partially responsible [28], with the binding of these to sphingosine-1-phosphate receptor 3 (S1P3) demonstrated to stimulate vasorelaxation via the activation of phosphatidylinositol 3-kinase(PI3K)/Akt and mitogen-activated protein kinase (MAPK) pathways [29]. In addition to ApoA-I and SR-BI, additional ABCG1 pathways are also likely contribute to this response. HDL-mediated efflux of 7-ketosterol via ABCG1 has also been shown to improve NO release by preventing the generation of NO-scavenging reactive oxygen species within the endothelium, while cholesterol efflux via the same receptor has been shown to elicit similar responses via a reduction in eNOS/caveolin interactions [30, 31].

An accumulation of foam cells within the vascular wall is another key process of atherosclerotic disease, initiated via the uptake of circulating LDL-c and its subsequent oxidation by reactive oxygen species (ROS) and pro-inflammatory enzymes such as MPO within the sub-endothelial space. This oxLDL is subsequently taken up by scavenger receptors on macrophages, triggering a cascade of further oxidative and inflammatory responses which encourage the promotion of endothelial adhesion molecule expression and increased monocyte infiltration within the vessel wall. HDL has been shown to exert multiple antioxidant and anti-inflammatory effects that may reduce this vicious circle of oxidation/monocyte infiltration, such as an ability to protect endothelial and smooth muscle cells from the cytotoxic effects of oxLDL [32‐34] in vitro. This effect is likely to be attributed to a variety of mechanisms. For example, increases in NO bioavailability (as described in the previous section) likely contribute at least in part to this response, as do the presence of multiple antioxidant enzymes carried within HDL’s protein cargo. The most studied of these enzymes is the antioxidant enzyme paraoxonase-1 (PON-1), although roles for other enzymes such as lipoprotein-associated phospholipid A2 (Lp-PLA₂) [35] and LCAT [36] have also been demonstrated. The presence of PON-1 has been shown to protect both HDL and LDL from oxidation in vitro [33, 37], while its absence (using PON-1 knockout mice) has been demonstrated to have the opposite effect [38]. Interactions with ApoA-I appear to be crucial for its activity, as demonstrated by the significantly increased capacity for PON-1 to prevent LDL oxidation and promote RCT in HDL particles containing ApoA-I as opposed to those containing ApoA-II or IV [39]. Additional antioxidant effects of ApoA-I also likely contribute to HDL’s antioxidant properties via its ability to directly bind and remove oxidised lipids from LDL particles within the vascular wall, as treatment of arterial cell walls with ApoA-I or an Apo-AI mimetic peptide in vitro prevents the oxidation of LDL, as does injection of ApoA-I into both mice and humans [40, 41]. HDL has also been shown in a number of studies to reduce superoxide production in endothelial cells treated with tumour necrosis factor alpha (TNF-α) [42‐44], possibly through inhibitory effects on nicotinamide adenine dinucleotide phosphate (NADPH)-oxidases mediated through HDL-associated lysosphingolipids and their interaction with S1P3 and SR-BI receptors [45]. Both this pathway and others have also been shown to have downstream effects on the production of numerous inflammatory-mediated adhesion molecules such as vascular and intercellular adhesion molecules (VCAM-1 and ICAM-1) [46], E-selectin [28], and monocyte chemoattractant protein-1 (MCP-1) [45, 47], reducing their expression and limiting monocyte transmigration across the vascular wall. Furthermore, ABCA1-mediated cholesterol efflux to ApoA-I may also provide additional suppression through the activation of anti-inflammatory signalling molecules during reverse cholesterol transport [48].

The concept that individuals with chronic disease may have structurally modified and potentially dysfunctional HDL was initially suggested in the mid-1990s, where evidence was produced for the first time showing the replacement of ApoA-I and paraoxonase-1 (PON-1) during an acute inflammatory response with acute phase proteins such as ceruloplasmin and serum-amyloid A (SAA) [47]. In this seminal study, the authors further noted that the antioxidant and anti-inflammatory vasoprotective properties of these modified HDL particles were also lost—or in certain cases—even completely reversed, suggesting that conformational changes in the HDL particle may have negatively affected its function. Since then, wide-ranging structural changes have been reported in a variety of inflammatory disease states, many of which have been implicated in the generation of a dysfunctional phenotype which may act to increase atherosclerotic risk. The most well-studied of these is the incorporation of acute phase proteins such as SAA, symmetric dimethylarginine (SDMA), lipopolysaccharide binding protein (LBP), alpha-1-antitrypsin (A1AT), or fibrinogen into HDL’s protein cargo [49]. These changes in turn result in reciprocal and detrimental reductions in ApoA-I, a decrease in the activity of HDL-associated antioxidant enzymes such as PON-1 and Lp-PLA₂, and an increased presence of inflammatory enzymes and lipid peroxidation products such as myeloperoxidase (MPO) and malondialdehyde (MDA) [49]. Furthermore, compositional changes in HDL’s lipid cargo—such as the enrichment with triglycerides commonly observed in hypertriglyceridaemic states—may further affect particle size and density, and therefore functional ability.

While a reduction in cholesterol efflux is a well-established aspect of the innate immune system during acute infections, it may have long-term negative implications under conditions of chronic inflammatory stress [15]. Numerous studies have shown that this vital homeostatic process may become impaired by a number of structural and conformational changes within HDL particles exposed to acute or chronic inflammatory conditions. The incorporation of SAA has been shown to impair cholesterol efflux in some [50‐53], but not all [42, 54], studies. These changes may arise through the displacement of atheroprotective components such as ApoA-I or PON-1 within the HDL particle itself [50], or by interactions with cell membrane-bound receptors responsible for binding HDL during the RCT process. In mice overexpressing SAA, reduced binding of HDL to SR-BI and a reduction in selective hepatic cholesterol ester uptake from HDL has been observed in conjunction with elevated levels of lipid-free SAA, suggesting a reduced capacity for the liver to remove cholesterol esters for excretion [53]. Further upstream in the RCT pathway, genetic ablation of SAA has been observed to preserve cholesterol efflux in mice following an endotoxin challenge, whereas the same challenge in humans resulted in progressive decreases in efflux capacity in line with elevations in the content of HDL-incorporated SAA [51]. Furthermore, an increased affinity for HDL to bind to proteoglycans at peripheral sites has also been demonstrated following SAA elevation in endotoxin-injected mice [55]. This effect is again virtually abolished following SAA knockout, suggesting that the incorporation of SAA into HDL may further impair RCT through the entrapment of HDL within vascular lesions, where its contents may subsequently be liable to oxidation or enzymatic modification that render it more atherogenic [55]. In support of this, a number of studies have demonstrated extensive structural and functional changes within HDL particles located within the vessel wall in comparison to those circulating in the bloodstream [56‐58], resulting in a substantially dysfunctional HDL phenotype which is unable to promote cholesterol efflux. For example, ApoA-I isolated from human atherosclerotic plaques has been shown to be lipid-poor, heavily oxidised by the inflammatory enzyme myeloperoxidase (MPO), and virtually devoid of any cholesterol acceptor activity [59]. This dysfunctional form of ApoA-I—which may also be caused by chlorination, nitration, or sulfoxidation of other amino acid residues—appears to exert its detrimental effects via multiple sites of action, including impairment of ABCA1 cholesterol efflux [58, 60, 61] and decreased ApoA-I-induced activation of LCAT [62]. Further inhibition of ABCA1 via an accumulation of the lipid peroxidation product MDA (a reactive carbonyl that has been shown to increase in HDL in chronic disease states) has also been demonstrated [63], while HDL-enrichment with triglycerides has been shown to impact RCT in some, but not all, studies. These latter findings may perhaps seem counter-intuitive given the finding that triglyceride-enriched HDL may in fact increase macrophage cholesterol efflux [64‐66]. However, they are likely explained by further downstream effects that may affect overall RCT, including a reduced ability for LCAT to convert free cholesterol to cholesterol esters [64], and a diminished capacity to deliver these cholesterol esters to hepatic cells via SR-BI [64, 67]. HDL remodelling due to elevated PLTP activity may also adversely affect cholesterol acceptor ability, and has been shown to result in increased atherosclerotic plaque formation in mice [68].

The ability of HDL to promote endothelial nitric oxide bioavailability has repeatedly been shown to be impaired in multiple chronic inflammatory and chronic disease states [42‐44, 69‐71]. HDL isolated from patients with either established coronary heart disease or acute coronary syndrome has a reduced capacity for NO generation that appears mediated, at least in part, by activation of endothelial lectin-like oxidised LDL receptor 1 (LOX-1). This upregulation of LOX-1 has been hypothesised to occur in response to an excess accumulation of the highly reactive lipid peroxidation compound MDA, possibly due to the displacement of antioxidant enzyme PON-1 from within the HDL particle. These changes in turn result in an increased activation of the eNOS-inhibitor endothelial PKCβII, ultimately leading to a reduced capacity for endothelial NO generation [71]. Inactivation of PON-1 in healthy subjects in vitro has been shown to result in decreased eNOS phosphorylation at serine residue 1177 (an eNOS activating site), increased phosphorylation at threonine residue 495 (a deactivating site), and a subsequent decrease in NO bioavailability. This relationship may be mediated at least in part by the oxidative transformation of PON-1 by increased levels of pro-inflammatory enzymes such as MPO [72], as they have been shown to display reciprocal relationships. Furthermore, MPO and other oxidative products within HDL may directly bind to and activate endothelial cells, thereby further reducing NO production via eNOS coupling [73, 74].

The first evidence underpinning the concept of HDL ‘dysfunction’ came from the finding that HDL isolated from rabbits undergoing an acute phase response was unable to prevent lipid peroxidation or MCP-1 expression [47]. Since then, multiple other pro-inflammatory and pro-oxidant properties have been attributed to dysfunctional HDL particles; including an increased capacity to stimulate NADPH oxidase activity [44], an increased capacity for endothelial superoxide production [42, 69], an increased expression of cell adhesion markers such as VCAM-1 [43], and increased eNOS uncoupling [73]. Much of these functional changes are likely closely linked to those previously described when addressing reductions in NO bioavailability, with the replacement of protective proteins such as ApoA-I and PON-1 with atypical components such as MDA and MPO of particular importance. For example, while in vitro work has highlighted the ability of ApoA-I to remove seeding molecules responsible for LDL oxidation [40], this ability is lost in HDL isolated from patients with established CHD [75]. PON-1 is frequently observed to be reduced in inflammatory conditions and chronic disease [42, 47, 69], and genetic-knockout mice lacking PON-1 have been shown to have both impaired antioxidant effects and increased risk of atherosclerosis [38].

The cardiovascular disease process has been shown to exert widespread changes in HDL structure and function which may accelerate, rather than reduce, future risk of major adverse events. In a series of studies involving the LURIC, 4S, and KORA cohorts, the association between HDL-c and CVD mortality was assessed in almost 9000 patients undergoing coronary angiography who were stratified by median levels of serum SAA and SDMA [76, 77]. Remarkably, while patients with below median levels of both acute phase proteins showed the traditional inverse relationship between HDL-c and CVD outcomes, those with above median levels were observed to have the complete opposite response—i.e. a significantly increased risk of CVD mortality as HDL-c levels increased. Similar findings have also been reported when stratifying risk by levels of C-reactive protein (CRP) in the PREVEND study [78], and ApoC-III in two prospective case-control studies analysed within the Nurses Health and Health Professionals Follow-Up Studies [79], suggesting that stratification by inflammatory burden in CVD may help to identify the independent effects of ‘biologically effective’ and ‘biologically ineffective’ HDL particles. When assessing HDL function directly, cholesterol efflux capacity has been shown to be decreased in patients with CHD [80], while numerous endothelial protective functions appear to be impaired. Reduced NO bioavailability is evident in HDL isolated from patients with both established coronary heart disease and acute coronary syndrome [71]. Furthermore, and in contrast to its usual antioxidant and anti-inflammatory functions, HDL isolated from patients with CHD has also been shown to promote rather than protect against LDL oxidation [47, 75], as well as increasing the expression of adhesion markers such as MCP-1 [47], and promoting endothelial superoxide production in the presence of TNF-α [71].

HDL isolated from patients with type 2 diabetes demonstrates many of the structural and functional alterations previously described for CVD. Structural changes include increased SAA levels due to underlying low-levels of chronic inflammation [81], increased triglyceride content due to insulin resistance and elevated CETP activity [82], decreased ApoA-I and S1P due to displacement by other proteins [83], and increased oxidised lipids and advanced glycation end products (AGE) due to increased levels of systemic oxidative and glycaemic stress [84]. Cholesterol efflux is reduced [85]—possibly due to an impairment of both ABCA1 and SR-BI mediated pathways through oxidation and AGE—and LCAT activity is impaired [86]. Widespread impairments in the endothelial protective effects of HDL are also evident, as demonstrated by a recent study in which HDL isolated from patient with type 2 diabetes demonstrated impaired NO bioavailabilty, increased MPO and NADPH oxidase activity, increased superoxide production, and increased lipid peroxidation [44].

The presence of reduced renal function in chronic kidney disease (CKD) leads to an accumulation of acute phase proteins within a circulating uremic milieu, many of which are capable of negatively altering HDL structure and function. HDL isolated from patients on haemodialysis is enriched with numerous components detrimental to HDL function (e.g. SAA, ApoC-III, A1AT, and triglycerides), and has been observed to have a reduced capacity to initiate RCT from macrophages [87]. Of particular importance in CKD is a significant increase in the acute phase protein symmetric dimethylarginine (SDMA), whose incorporation into the HDL particle is associated with impairments in HDL-mediated RCT and endothelial NO production in a dose-dependent manner [43, 70, 77]. This effect becomes progressively pronounced as kidney function declines, as demonstrated by a 40–60% decline in HDL-mediated endothelial NO bioavailability in patients with stage 5 CKD as opposed to a 20% increase in healthy controls [43]. Interestingly, these changes are largely reversed following kidney transplantation in children, but not adults—suggesting that HDL’s vascular protective functions may be at least partially restored in the young [43]. Using genetic knockout mice models to investigate the pathways underlying this relationship, Speer and colleagues identified toll-like receptor 2 (TLR2) as the likely signalling receptor responsible, suggesting that structural modifications to HDL caused by CKD may result in damage-associated molecular patterns within the HDL particle that trigger a chronic innate immune response [74]. Further post-translational modifications to other circulating proteins may also affect function. Albumin isolated from haemodialysis patients is heavily oxidised and markedly impairs cholesterol efflux in vitro; likely due to its high-affinity for SR-BI reducing binding sites for HDL-mediated RCT [88].

HDL dysfunction has recently been demonstrated in a range of autoimmune diseases, including type 1 diabetes [89], rheumatoid arthritis [90, 91], systemic lupus erythematosus [92], and primary antiphospholipid syndrome [69]. Despite being at increased risk of dyslipidaemia, patients with type 1 diabetes commonly demonstrate elevated—rather than reduced—levels of HDL-c. Multiple studies have reported the presence of multiple dysfunctional properties to these particles, however, including reduced cholesterol efflux capacity [93, 94], reduced PON-1 activity [95], increased MPO activity [96], and decreased levels of S1P [97]. These modifications appear to be independent of HbA1c [93], and may at least partially explain the increased risk of CVD in patients with the highest HDL-c levels [98]. HDL isolated from patients with rheumatoid arthritis, systemic lupus erythematosus, and antiphospholipid syndrome also display widespread structural changes, resulting in a pro-inflammatory phenotype that worsens as systemic inflammatory burden increases [69, 91]. Treatment with anti-inflammatory medications in these patients may partially reverse these structural and functional modifications, as evidenced by decreases in SAA and LBP content, increased NO bioavailability and PON-1 activity, and decreased SO production in individuals with rheumatoid arthritis treated with the monoclonal antibodies infliximab and adalimumab [99, 100].

Periodontal disease represents one of the most prevalent chronic inflammatory conditions in the world today, affecting approximately 20–50% of the global population [101]. Numerous vasoprotective functions of HDL have been shown to be impaired in periodontal patients compared to age-matched controls, including impaired endothelial NO bioavailability, increased superoxide production, and reduced PON-1 activity [42]. When using an intensive periodontal treatment intervention as an experimental model of acute inflammation, further transient decreases in both endothelial NO availability and PON-1 activity—but not RCT—are apparent in the following 24 h, highlighting the vulnerability of HDL to rapidly remodel when exposed to an inflammatory response. These changes were accompanied by substantial increases in acute phase reactants such as SAA, C3, and prothrombin, and returned to baseline levels following resolution of the inflammatory response.