Relationship between lipoproteins, thrombosis, and atrial fibrillation

Abstract The prothrombotic state in atrial fibrillation (AF) occurs as a result of multifaceted interactions, known as Virchow’s triad of hypercoagulability, structural abnormalities, and blood stasis. More recently, there is emerging evidence that lipoproteins are implicated in this process, beyond their traditional role in atherosclerosis. In this review, we provide an overview of the various lipoproteins and explore the association between lipoproteins and AF, the effects of lipoproteins on haemostasis, and the potential contribution of lipoproteins to thrombogenesis in AF. There are several types of lipoproteins based on size, lipid composition, and apolipoprotein category, namely: chylomicrons, very low-density lipoprotein, low-density lipoprotein (LDL), intermediate-density lipoprotein, and high-density lipoprotein. Each of these lipoproteins may contain numerous lipid species and proteins with a variety of different functions. Furthermore, the lipoprotein particles may be oxidized causing an alteration in their structure and content. Of note, there is a paradoxical inverse relationship between total cholesterol and LDL cholesterol (LDL-C) levels, and incident AF. The mechanism by which this occurs may be related to the stabilizing effect of cholesterol on myocardial membranes, along with its role in inflammation. Overall, specific lipoproteins may interact with haemostatic pathways to promote excess platelet activation and thrombin generation, as well as inhibiting fibrinolysis. In this regard, LDL-C has been shown to be an independent risk factor for thromboembolic events in AF. The complex relationship between lipoproteins, thrombosis and AF warrants further research with an aim to improve our knowledge base and contribute to our overall understanding of lipoprotein-mediated thrombosis.


Introduction
Atrial fibrillation (AF) is a multi-systemic condition that is associated with serious complications including thromboembolism, dementia, and heart failure, resulting in impaired quality of life, significant morbidity, and increased mortality. [1][2][3][4][5] The prevalence of AF rises with age and concomitant comorbidities. 6,7 At present, there is an upward trajectory to the global incidence and prevalence of AF. 8,9 Indeed, every individual has a one-in-four lifetime risk of developing this condition, 10,11 with a greater burden amongst those with risk factors. 12 By 2060, it is projected that at least 17.9 million people in Europe will be affected by AF. 13,14 The mechanism by which AF occurs is complex but has previously been described in detail. 15 Management of patients with the condition is primarily focused on the prevention of thromboembolism due to the presence of a prothrombotic state with this arrhythmia. The prothrombotic or hypercoagulable state in AF occurs as a result of multifaceted interactions, known as Virchow's triad of hypercoagulability, structural abnormalities, and blood stasis. 16 Despite considerable research in this area, the precise mechanisms by which AF contributes to a prothrombotic state remains ill-defined.
There is emerging evidence that lipoproteins are implicated in thrombogenesis, beyond their traditional role in atherosclerosis. In this review, † These authors are Joint first authors.
we provide an overview of the various lipoproteins and explore their relationship with AF, haemostasis, and the potential contribution to thrombogenesis.

Lipoproteins
Lipids (also known as 'fat') are naturally occurring compounds serving numerous biological functions including the formation of plasma membranes or signalling molecules, and as a source of energy. They exist in several forms including free fatty acids, glycerolipids (GL), glycerophospholipids (GPL), sphingolipids, and sterol lipids. Each of these lipid subtypes have different molecular structures and basic properties ( Figure 1). As a brief overview, fatty acids form the fundamental category of biological lipids and therefore the basic building blocks of more complex lipids. Their chemistry consists of a hydrocarbon chain with a terminal carboxylic acid group and may be defined as saturated or unsaturated depending on the maximum possible number of bonds or hydrogen atoms. 17,18 GL consist of a single glycerol molecule which acts as the backbone for attachment to fatty acid chains. The most relevant example of GL are triglycerides (TG), which contain three fatty acid chains and play an important role in metabolism as energy sources and sources of dietary fat. 18,19 Sterol lipids consist of four fused rings of hydrocarbon to which other molecules attach. A major type of sterol lipid is cholesterol which serves as a precursor for the synthesis of other steroids as well as serving as structural support for plasma membranes. 20,21 Dietary cholesterol is often stored and transported in the form of a cholesterol ester (CE), which chemically represents a cholesterol molecule joined to a fatty acid via an ester bond. 22 One common feature that lipids share as a group is their insolubility in water. Consequently, they must be transported with proteins in the circulation ('lipoproteins'). 23 Lipoproteins are complex structures consisting of a central hydrophobic core primarily composed of CE and TG which is surrounded by a hydrophilic membrane comprising of GPL, free cholesterol, and apolipoproteins. 23,24 There are several types of lipoproteins based on size, lipid composition, and apolipoprotein category, namely: chylomicrons, very low-density lipoprotein (VLDL), low-density lipoprotein (LDL), intermediate-density lipoprotein, and high-density lipoprotein (HDL). When elevated, all lipoproteins confer a proatherogenic risk, apart from HDL which is anti-atherogenic. 23 Each lipoprotein contains numerous types of lipid species and proteins, whose composition varies even between individual lipoproteins of the same type ( Figure 2).
LDL is the main transporter for cholesterol in the circulation and every LDL particle contains one apolipoprotein B100 molecule. LDL exists in a spectrum that varies in size and density with the three major density subclasses being small dense LDL (sdLDL), intermediate LDL, and large buoyant LDL. 25 Small dense LDLs are considered more atherogenic and pro-coagulant compared to the other subtypes of LDL for various features as they have decreased affinity for LDL receptors and hence remain longer in the circulation, more readily enter the arterial intima where they are engulfed by macrophages to become foam cells, and are more susceptible to oxidation than its larger counterpart. [26][27][28] There is also increasing evidence that the number of ApoB-rich particles or the concentration of apolipoprotein B may contribute to atherogenic risk. 29 Modern lipidomic techniques, with the aid of liquid chromatography coupled to mass spectrometry, have allowed for detailed characterization of the LDL lipidome. 30 This has revealed over 300 different lipid species residing within the interior or phospholipid membrane of the LDL particle. Each of these may have specific associations with various pathologies and interactions with traditional risk factors, thereby adding to its complexities. 31,32 Oxidative modification of LDL, predominantly by nonenzymatic processes, leads to the formation of oxidized LDL (OxLDL) particles. These particles have altered structure and content, containing oxidized proteins and lipids (particularly GPL), and leading to a more atherogenic phenotype. 33 Furthermore, the susceptibility of LDL to aggregation and proteoglycan binding has provided a deeper insight into the atherogenicity of LDL. 34 Lipoprotein(a) [Lp(a)] is a specialized form of LDL assembled in the liver from LDL and apolipoprotein(a) attached to apolipoprotein B100 via a disulphide bridge ( Figure 2). 35 Lp(a) has been implicated in atherogenesis by enhancing endothelial cell adhesion and molecule expression, promoting the formation of foam cells by binding to macrophages with high affinity and interfering with vascular permeability. 36 Furthermore, the Lp(a) constituent, apolipoprotein(a), shares many structural similarities with plasminogen which has been reported to cause interference with the physiological fibrinolysis process and to contribute to a prothrombotic phenotype. 37 3. Lipoproteins and atrial fibrillation 3.1 The paradoxical inverse relationship between cholesterol and the incidence of AF The association between serum cholesterol and coronary heart disease has been described since early 1964. 38 There is an increased risk of coronary heart disease with elevated total cholesterol (TC) and LDL cholesterol (LDL-C), and reduced HDL cholesterol (HDL-C) levels. 39,40 A longitudinal analysis over a 35-year period of patients from the Framingham study confirmed that long-term exposure to these lipid abnormalities led to a greater risk of atherosclerotic cardiovascular disease and mortality. 41 Moreover, both the LDL particle and LDL-C are now considered causal for atherosclerotic cardiovascular disease. 42 In turn, atherosclerotic disease is an established independent risk factor for incident AF. 43,44 As such, elevated levels of TC and LDL-C may have been expected to increase the risk of incident AF. However, current evidence does not support this and in contrast, several well-conducted observational studies have described a paradoxical inverse relationship between TC and LDL-C, and incident AF ( Table 1).
A health survey performed by Iguchi et al. 61 found that hypercholesterolaemia, defined by TC >220 mg/dL or the use of cholesterollowering agents, was related to reduced new-onset AF. Reduced levels of LDL-C has also been linked to increased prevalence of AF. 56 In one study of 88 785 patients, for example, TC and LDL-C levels were inversely linked to incident AF over a follow-up period of 7 years. 48 The authors reported no significant association between incident AF, and HDL-C or TG. However, the overall incidence of AF was extremely low at 0.52 per 1000 person-years. 48 Similar findings were described in the ARIC (Atherosclerosis Risk in Communities) cohort which was validated even when analysing lipid levels as time-dependent variables. 59 An ancillary study to ALLHAT (Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial) demonstrated that low HDL-C was associated with a significant increase in incident AF. 62 In a Japanese cohort, Watanabe et al. 60 also found that both TC and LDL-C were inversely related to incident AF. Furthermore, reduced levels of HDL-C was independently associated with greater incidence of AF in females, but not males. The former had a 28% higher risk of AF with each 10% decrease in HDL-C. Results from the SPCCD (Swedish Primary Care Cardiovascular Database) showed that each unit (mmol/L) increase in TC and LDL-C were associated with a 19% and 16% lower risk of incident AF, respectively; also, HDL-C and TG were not related to incident AF. In contrast to the previous study, Mourtzinis et al. 49 found no sexspecific differences in outcomes based on lipid abnormalities.
The relationship (or lack of) between the aforementioned measures of lipid abnormalities and incident AF has also been demonstrated among patients with ST-elevation myocardial infarction 46 and chronic heart failure. 50 In a small study of patients who had AF ablation, TC and LDL-C were inversely associated with a higher risk of AF recurrence. 65 However, subgroup analysis demonstrated that these factors were only significant in females but not males. The levels of HDL-C and TG were not related to AF recurrence post-ablation. 65 The inverse relationship between AF, and TC and LDL-C are further supported by the fact that use of lipid-lowering medications does not reduce the risk of incident AF. 59,66 It is worth noting that conflicting results have been demonstrated in few studies. A combined analysis of the MESA (Multi-Ethnic Study of Atherosclerosis) and Framingham Heart Study cohorts found that raised HDL-C and TG were independently associated with a lower risk of new-onset AF. 57 However, the authors reported that TC and LDL-C were not important risk factors for new-onset AF. In a communitybased cohort of Korean males, Kim et al. 52 found that although the presence of metabolic syndrome led to greater incidence of AF over a follow-up period of 8.7 years, this was driven primarily by central obesity, and neither TG or HDL-C were risk factors for incident AF. Similar results were obtained from a historical Japanese population. 53 Different study designs, populations, lifestyles and age ranges may partly explain some of the inconsistencies of previous studies.
Nonetheless, the current literature strongly indicates that both TC and LDL-C have an inverse relationship with incident AF. This is supported by a recent meta-analysis of nine large cohort studies. 67 Overall, these findings are important as they imply that a reduction in TC and LDL-C, may have unintended consequences for the risk of incident AF. The role of TG and HDL-C, and whether there are sex-specific responses to lipid abnormalities with regards to AF need further investigation.
In addition to the measures of lipids described above, several others have been explored in relation to incident AF. Aronis et al. 54 found that Lp(a) levels above 50 mg/dL (compared to <10 mg/dL) were not associated with incident AF. Monocyte to HDL-C ratio has also been described as a novel biomarker of inflammation that may be useful to predict new-onset AF in patients undergoing percutaneous coronary intervention 51 or coronary artery bypass grafting. 55

Underlying mechanisms
In general, there is limited research on mechanisms that underpin the relationship between lipoproteins and AF. In a report from the Women's Health Study, Mora et al. 58 conjectures that the inverse relationship may be due to the stabilizing effect of cholesterol on myocardial cell membranes. This may occur through the effects of cholesterol on the regulation of ion channels and sensitivity of volume-regulated anion current to osmotic gradients. [68][69][70][71] Furthermore, cholesterol depletion has been found to impair cardiomyocyte contractility by deregulation of calcium handling, adrenergic signalling and the myofibrillar architecture. 72 The link between cholesterol levels and development of AF may also be related to inflammation. It has been shown that TC, LDL-C and HDL-C levels were decreased while TG was increased during inflammation. 73 Therefore, reduced levels of cholesterol may be reflective of underlying inflammatory processes within the host that contributes to AF. Furthermore, lipoproteins influence the course of sepsis by binding to Lipoproteins, thrombosis, and atrial fibrillation bacterial endotoxins and attenuate the harmful effects of inflammatory responses. 74 It was reported that the effects of lipoproteins on incident AF extended beyond the cholesterol content to include the number of lipoprotein particles for LDL and VLDL. 58 In this regard, it was the smaller particles for each of these lipoproteins that were the actual driving force contributing to the inverse relationship with AF as larger cholesterolrich LDL-particles, total HDL-C, Lp(a) and TG were not associated with incident AF. 58 In a small study of female patients undergoing catheter ablation, those with AF had smaller lipoprotein particles with increased oxidation, glycation and TG content compared to controls in sinus rhythm. 75 Similar findings have been reported elsewhere among male patients. 76 Overall, these changes resulted in enhanced foam cell formation via accelerated phagocytosis by macrophages, and reduced antioxidant ability of HDL. 75 These changes are important as HDL particles have been shown to be more protective against cardiovascular events, 77,78 which are known to contribute to AF. Furthermore, foam cells are known to initiate a wide range of bioactivities including inflammatory processes 79-81 that may be linked to the pathogenesis of AF.
Sex differences in the association of lipoproteins and AF that were observed in some studies may be attributable to hormones, especially oestrogen, and differences in body fat distribution or insulin sensitivity. 82 Moreover, a fall in testosterone levels among ageing males may influence oxidative modification of LDL-C. 85 It is worth mentioning that the effects of specific lipoproteins may vary under certain conditions. For example, injection of VLDL extracted from patients with metabolic syndrome into mice resulted in excess lipid accumulation and apoptosis in the atria, and significantly greater left atrial dilatation compared to VLDL from healthy volunteers. 86 Thus, VLDL may contribute to the development of atrial cardiomyopathy and subsequent vulnerability to AF through direct cytotoxicity, altered action potentials, disrupted calcium regulation, delayed conduction velocities, modulated gap junctions and derangements in sarcomere proteins (Figure 4). 87 This highlights the fact that focusing on the quantity of lipoproteins on its own may limit our understanding of the mechanisms underlying the paradoxical inverse relationship of lipoproteins and AF.

Lipoproteins and thrombosis
The role of lipoproteins in modulating thrombosis and haemostasis to produce fibrin clots is well described. 88 LDL and VLDL have been shown to increase thrombin generation and inhibit fibrinolysis. 89,90 An inverse relationship of VLDL to fibrin clot permeability and fibre mass-length ratio has previously been demonstrated. 91 In addition to the coagulation system, platelets seem to be affected by lipoproteins as well. To start with, there is evidence that patients with excessive LDL, such as those in familial hypercholesterolaemia that is characterized by lack or defective LDL receptors, display enhanced platelet reactivity with increased a-granule secretion, 92 fibrinogen binding, 93 and aggregation. 94 In contrast, patients with abetalipoproteinaemia that is characterized by a lack of all apolipoprotein B-containing lipoproteins (chylomicrons, VLDL and LDL) have reduced platelet activation. 95 Furthermore, LDL has been shown to promote excess platelet activation which may contribute to the higher incidence of thrombosis in hyperlipidaemia. 96,97 Certain subclasses of LDL may be more harmful than others. For instance, sdLDL was shown to be independently associated with both thrombotic and haemorrhagic strokes. 98 A potential mechanism could include increased susceptibility to oxidation which leads to a substantial increase in thrombin generation compared to the larger native LDL. 99,100 In addition to identifying the lipid subclasses and oxidative states, evaluating the effects of individual lipid species may be of importance. For instance, Klein et al. 101 demonstrated that VLDL was capable of activating the contact pathway in the presence of platelets, thereby causing an increase in the rate and amount of thrombin generation. A subsequent detailed lipoprotein analyses revealed that this was driven by phosphatidylethanolamine (PE). Interestingly, PE is also responsible for oxLDL-induced thrombin generation. 102

OxLDL and haemostasis
Despite many decades of research into oxLDL, definitions of what it contains and method of detection vary between groups and publications. 33 Perhaps the most encompassing definition for oxLDL is 'A particle derived from circulating LDL that may have peroxides or their degradation products generated within the LDL molecule or elsewhere in the body associated with the particle'. 33 Such particles therefore may include lipid peroxides, hydroxides or aldehydes such as malondialdehyde in addition to protein oxidation products. These biochemical changes give oxLDL altered properties which may facilitate its detection and separation on the basis of density, negative charge and monoclonal antibody (mAb). The latter method utilizes antibodies to oxidized epitopes on the surface of oxLDL such as EO6 for oxidized phosphatidylcholine (oxPC) 103 and 4E6 for oxidized apoB. 104 Given the variation in detection methods of oxLDL and possible consequences on interpretation of the evidence, this review specifies the method of detection of oxLDL where appropriate.
Elevated oxLDL levels (detected by 4E6 mAb) are independently associated with several cardiovascular risk factors including increasing age, male gender, raised body mass index, abdominal obesity, hypertension, raised C-reactive protein, renal dysfunction, hyperuricaemia, and smoking. 105 These risk factors are important in AF, which has also been shown to be directly associated with elevated 4E6-measured oxLDL levels. 106 Oxidized LDL (4E6 mAb) correlates to thrombogenesis by interfering with the coagulation system and clot formation. In this regard, patients with acute coronary syndrome demonstrate a positive correlation between oxLDL and tissue factor levels in plasma. 110 Activation of T lymphocytes by oxLDL, prepared by chemical oxidation of native LDL with copper sulfate, via the lectin-type oxLDL receptor 1 (LOX-1) has also been shown to increase the expression of tissue factor on the surface of leukocytes. 111 Furthermore, oxLDL generated with copper oxidation was noted to inhibit fibrinolysis, modify fibrin clot structure and increase thrombin generation. 102,112 Finally, oxLDL (detected by 4E6) correlated to reduced clot permeability and prolonged clot lysis time. 113 OxLDL generated in vitro by copper oxidation has been shown to cause activation and aggregation of platelets via CD36 and LOX-1, [114][115][116] as well as impair endothelial regeneration by reducing the release of nitric oxide. 117 Furthermore, platelet reactivity in cardiovascular disease can be related to dyslipidaemia, 118,119 which is characterized by accumulation of oxLDL as measured by LDL isolation, lipid extraction and subsequent high performance liquid chromatography. 120 In turn, platelet reactivity is an important determinant of fibrin clot structure and effective platelet inhibition is associated with a weaker, more permeable fibrin network. 121 Therefore, oxLDL may indirectly influence fibrin clot properties through its effects on platelet reactivity. To complicate matters, recent evidence suggests that oxLDL activation of platelets promotes further oxLDL uptake by platelets (detected with the polyclonal orb10973 anti-oxLDL antibody), augmenting the pro-oxidative thrombogenic phenotype. 122 Finally, there is evidence suggesting that activated platelets contribute to the formation of oxLDL species and modification of lipoprotein function. 123 Putting it together, the evidence points towards a cycle of oxLDL-induced platelet activation leading to further oxLDL formation and uptake by platelets.

Lp(a) and haemostasis
In addition to its recognized atherogenic properties, 124 Lp(a) appears to have a direct prothrombotic effect by interfering with platelets and the fibrinolysis system. Although it has been found to interact with platelets, the target receptor remains unclear. 125 Furthermore, literature surrounding the nature of interaction between Lp(a) and platelets is conflicting, with evidence to suggest that it may have both activating and inhibiting effects. 126 Lp(a) has been shown to facilitate platelet activation through thrombin-related activating hexapeptide, but not thrombin or adenosine diphosphate. 127 On the contrary, some studies reported an inhibitory effect of Lp(a) to platelet activation by collagen or thrombin. 125 Less controversial is the ability of Lp(a) to impair platelet-mediated fibrinolytic reactions by interfering with the binding of plasminogen, which shares structural similarities to apolipoprotein(a), and tissue plasminogen activator to the platelet surface. 128 This is compounded by the ability of Lp(a) to inactivate tissue factor pathway inhibitor which may promote thrombosis through the extrinsic coagulation pathway. 129 However, evidence in genetic studies on the contribution of Lp(a) to venous thrombosis have been negative, 130,131 suggesting that the primary prothrombotic effects of Lp(a) may be limited to atherothrombosis (arterial) or anti-fibrinolysis. 132 Additional studies describing the association between lipoproteins and thrombotic conditions are summarized in Table 2.

The effects of lipid-modifying therapy on thrombosis and haemostasis
The role of lipoproteins in haemostasis is further supported by the fact that application of lipid-modifying therapy is associated with changes in haemostasis. 142 Specifically, atorvastatin may exert antiplatelet effects by interfering with redox signalling. 143 It has also been shown that statins are able to reduce fibrin clot lysis time, independent of warfarin. 144 For example, a randomized controlled trial by Undas et al. 145 confirmed the effects of statins and also showed similar results with the use of other lipid-modifying therapy, specifically fenofibrate. The authors reported increased fibrin clot permeability and reduced lysis time with the use of these agents compared to pre-treatment values, potentially through its effects on thrombin generation. Turbidity analysis also showed that use of these drugs resulted in thicker fibres that were more prone to effective fibrinolysis.
A further randomized controlled trial of patients with type 1 diabetes mellitus and dyslipidaemia found that the beneficial effects of statins on fibrin clot properties may be related to reduced expression of glycoprotein IIIa, tissue factor, and P-selectin. 146 Finally, the use of statins has been associated with risk reduction of both venous and arterial thromboembolisms. [147][148][149][150][151] Therefore, it is tempting to speculate that the statininduced protective effects may be related to its influence on reduction of pro-coagulant lipoproteins or enhancement of anti-coagulant lipoproteins. 90 A prospective, case-controlled study of patients with stable coronary artery disease and hypercholesterolaemia found that use of pravastatin was associated with reduced thrombus formation at both high and low shear rates. 152 As expected, there was a significant decrease in TC and LDL-C levels with pravastatin. Thrombus formation was also assessed after 1 week of treatment with pravastatin, prior to any significant reduction in TC and LDL-C levels, and it was found that this was unchanged compared to pre-treatment. As a result, the authors concluded that the beneficial effects of pravastatin on thrombogenicity was due to its effects on lipids/lipoproteins. 152 Interestingly, other studies have reported that the anti-coagulant effects of statin therapy, in terms of thrombin generation and platelet activation, were seen as early as 3 days following treatment. 153,154 Nonetheless, it should be noted that there currently remains insufficient evidence to conclude whether the protective effects of statins are related to its lipid-modifying effects or otherwise. 148 In contrast to the aforementioned studies, Dangas et al. 155 showed a reduction in thrombogenicity among patients after 6 months of treatment with pravastatin, regardless of change in LDL-C. Furthermore, despite a similar reduction in LDL-C between subgroups of patients treated with pravastatin compared to dietary advice only, the anti-thrombotic benefit was only demonstrated among those receiving pravastatin. Additionally, a study by Undas et al. 156 found that the use of simvastatin was associated with a reduction in thrombin generation, independent of changes in lipid profile. Overall, there may be various pathways by which lipid-modifying therapy, in particular statins, may interact with the haemostatic process.

Lipoproteins and thromboembolism in AF
Given the effects of lipoproteins on haemostasis, their contribution to thromboembolic events may be expected. Indeed, lipoprotein abnormalities have been shown to be an independent risk factor for stroke and venous thromboembolism. [157][158][159][160] However, few studies have explored this relationship in the context of AF ( Table 3).

Low-density lipoprotein cholesterol
LDL-C has been implicated in thromboembolic events among patients with AF. Wu et al. 165 found that LDL-C was an independent risk factor for both a history of ischaemic stroke and future stroke risk among patients with AF. Similar findings were reported in a case-controlled study, whereby raised LDL-C was shown to be an independent predictor of ischaemic stroke in patients with AF, irrespective of the CHA 2 DS 2 -VASc score. 164 Furthermore, this association demonstrated a dose-response pattern. A later study confirmed the relationship between LDL-C and ischaemic stroke, and observed that lowering LDL-C may be particularly beneficial among AF patients with a low CHA 2 DS 2 -VASc score (less than two in males and three in females). 161 . Interestingly, LDL-C appears to have an opposite influence on the risk of incident AF and subsequent thromboembolic risk which highlights the importance of regular monitoring and treatment adjustments in clinical practice.

Lipoprotein(a)
There are conflicting reports on the effects of Lp(a) on thromboembolic risk in AF. Igarashi et al. 166 demonstrated that serum Lp(a) was an independent risk factor for left atrial thrombus detected on transoesophageal echocardiogram in patients with chronic AF. Additionally, left atrial thrombus was present in 48% of AF patients with a Lp(a) level > _30 mg/dL, suggesting that this may be a useful biomarker to identify patients at high risk of thromboembolism. However, a limitation of this study was that relatively few patients (19%) were on anticoagulation therapy. 166 More recently, higher Lp(a) levels were found to be independently associated with clinically confirmed thromboembolic events in nonvalvular AF patients with a CHA 2 DS 2 -VASc score of less than two. 162 Curiously, Aronis et al. 54 found that elevated levels of Lp(a) was associated with an increased stroke risk among non-AF patients, but not in those with AF. In support of the latter, we previously demonstrated that there was no correlation between Lp(a) and D-dimer, as a marker of thrombogenesis. 167 Overall, the inconsistent results on Lp(a) may suggest the existence of different Lp(a) phenotypes that contribute differently to thrombogenesis 168 and therefore, sole measurement of total Lp(a) levels may be inadequate for this purpose. In this regard, the measurement of oxidized lipids may have an important role to increase our understanding on the potential impact of Lp(a) on atrial function and risk of AF. [169][170][171]

Other measures of lipoproteins
In a sub-study of the ARISTOTLE trial, higher levels of Apolipoprotein A1 were independently associated with a lower composite risk of ischaemic stroke, systemic embolism, myocardial infarction, and cardiovascular mortality. 163 When analysed separately, Apolipoprotein A1 was found to be a risk factor for each of the individual outcomes apart from myocardial infarction. In reverse, the authors reported that Apolipoprotein B was not associated with the risk of composite outcomes but that it was a risk factor for myocardial infarction. Decker et al. 172 demonstrated that low HDL and high TG were not independently associated with ischaemic stroke among AF patients over a follow-up period of 14.8 173 In this context, the effects of statins may be related to a reduction of oxLDL levels that promote its anti-inflammatory properties, 174,175 which has been shown to reduce the endogenous thrombin potential in patients with AF. 176 He et al. 177 found that prior use of statins resulted in lower plasma oxLDL levels at baseline and at 3-month follow-up among patients presenting with an ischaemic stroke. Furthermore, pre-stroke statin use was associated with reduced short-term mortality [odds ratio (OR) 0. 38

Gaps and limitations
Despite a wealth of evidence on the role of lipoproteins in thrombosis and AF, it is recognized that these molecules are heterogeneous, containing numerous subclasses and lipid species with variable effects. 178 In this regard, much of the conflicting evidence and paradox in prior studies may be due to the usage of crude methods of classification that undermines the complexity of lipoproteins. Given recent advancements in our ability to accurately analyse lipoprotein subclasses and lipid species, future studies should focus on identifying the relationship of these molecules with incident AF and thromboembolic complications. Moreover, the mechanism by which this occurs also warrants further investigation. With better understanding in this area, the development of targeted treatment approaches for high-risk subgroups may be possible. Furthermore, ongoing clinical trials such as the Lp(a)HORIZON study (ClinicalTrials.gov NCT04023552) are examining novel agents targeting Lp(a) levels and may provide more data on the association of Lp(a), incident AF and thrombotic events.
One group of lipids which is emerging as a key player in haemostatic reactions is oxidized GPL. These molecules have been shown to play a role in thrombotic disorders and are primarily generated enzymatically by platelets and leukocytes. 179,180 The presence of these molecules in lipoproteins has not been conclusively studied, particularly in light of newer lipidomic technologies. The majority of previous studies of oxidized GPL in lipoproteins had relied on antibodies that bind oxPC, demonstrating their presence as a defining feature of oxLDL 181 and Lp(a). 182 It is not known whether the presence of oxPC, or other oxidized GPL, on lipoproteins enhance coagulation reaction in a similar way to enzymatically-generated oxPC on the surface of activated cells. 179 The growth in the lipidomics field and availability of increasingly sensitive techniques may pave the way for studies in this area.
Moving forward, the role of genetics in lipoproteins should also be considered. Elevated Lp(a) is prevalent in approximately 20% of the population, 183 and strongly influenced by genetic variability. 184 Much of the variation is related to the apo(a) protein, which consists of kringle domains that vary in molecular weight and therefore size of the Lp(a) particle. 185,186 The genetic variation in the LPA locus has enabled Mendelian randomization studies to demonstrate that both the Lp(a) concentration and the smaller apo(a) isoform are independently causal for some cardiovascular diseases. 183,[187][188][189][190][191] While a large UK-based population study by Zanetti et al. 188 found no causal relationship between Lp(a) and AF, further Mendelian randomization studies are needed to confirm this finding in other cohorts.

Conclusion
There is a paradoxical relationship between TC and LDL-C, and incident AF. The mechanism by which this occurs is poorly defined but may be related to changes in the regulation of ion channels and inflammatory processes. To complicate matters, excess lipoproteins promote thrombin generation, inhibit fibrinolysis and enhance platelet activation. In this regard, LDL-C has been shown to be an independent risk factor for thromboembolic events in AF. Overall, the complex relationship between lipoproteins, thrombosis and AF warrants further research. An improved knowledge base in this area may unlock important mechanistic pathways that contribute to our overall understanding of haemostasis and guide our clinical approach in the treatment of prothrombotic conditions.

Data availability
The data that support the findings of this review are available from the corresponding author upon reasonable request.