Lipoprotein (a): truly a direct prothrombotic factor in cardiovascular disease?

Elevated plasma concentrations of lipoprotein (a) [Lp(a)] have been determined to be a causal risk factor for coronary heart disease, and may similarly play a role in other atherothrombotic disorders. Lp(a) consists of a lipoprotein moiety indistinguishable from LDL, as well as the plasminogen-related glycoprotein, apo(a). Therefore, the pathogenic role for Lp(a) has traditionally been considered to reflect a dual function of its similarity to LDL, causing atherosclerosis, and its similarity to plasminogen, causing thrombosis through inhibition of fibrinolysis. This postulate remains highly speculative, however, because it has been difficult to separate the prothrombotic/antifibrinolytic functions of Lp(a) from its proatherosclerotic functions. This review surveys the current landscape surrounding these issues: the biochemical basis for procoagulant and antifibrinolytic effects of Lp(a) is summarized and the evidence addressing the role of Lp(a) in both arterial and venous thrombosis is discussed. While elevated Lp(a) appears to be primarily predisposing to thrombotic events in the arterial tree, the fact that most of these are precipitated by underlying atherosclerosis continues to confound our understanding of the true pathogenic roles of Lp(a) and, therefore, the most appropriate therapeutic target through which to mitigate the harmful effects of this lipoprotein.

The LBSs in plasminogen have been shown to be important for both lysine-dependent interactions with substrates, such as fibrin and cell-surface receptors (18)(19)(20), as well as for intramolecular interactions that maintain the closed native conformation of plasminogen (21,22).
Of the KIV types in apo(a), only KIV 5 , KIV 6 , KIV 7 , KIV 8 , and KIV 10 have LBSs, with all the other KIV types containing one or more key amino acid substitutions that inactivate their LBSs ( Fig. 1) (4,23). However, the LBSs in KIV 5 -KIV 8 have been demonstrated to be of a lower affinity than the one in KIV 10 , due to some conservative amino acid substitutions (24,25). As such, the LBSs in apo(a) KIV 10 and KIV 5 -KIV 8 have been termed as "strong" and "weak" LBSs, respectively (Fig. 1), although all of these LBSs are of a lower affinity for lysine than the kringle I LBSs in plasminogen (26). Of the lysine-binding apo(a) kringles, only KIV 10 is available for interaction with lysinecontaining substrates. This is because the LBSs in KIV 5 -KIV 8 are masked when bound to apoB-100 in the Lp(a) particle (27,28), with KIV 7 -KIV 8 having been explicitly shown to participate in noncovalent interactions with specific lysine residues on apoB-100 that precede covalent Lp(a) formation (29).
The protease-like domain in apo(a) is catalytically inactive, despite having an intact Ser-His-Asp catalytic triad (30). An Arg to Ser substitution at the location analogous to the site on plasminogen that is cleaved by plasminogen activators ensures that an activating cleavage of apo(a) cannot occur (4). In addition, several other amino acid substitutions relative to plasminogen, as well as a key nineamino acid deletion in apo(a), have been proposed to render the protease-like domain in apo(a) inactive (30).
The gene encoding apo(a) has been proposed to have arisen from duplication of the gene encoding plasminogen comparatively late in primate evolution ( Fig. 1) (31). Indeed, Lp(a) is only present in Old World monkeys, apes, and humans. The apo(a) from all these species generally shares the isoform size heterogeneity and kringle organization observed in humans, with the exception that some species lack KV (32)(33)(34)(35). What is most interesting in examining the various sequences is that independently occurring mutations have ensured that the protease-like domain is inactive and, most relevant to the role of Lp(a) in thrombosis, KIV 10 is not able to bind to lysine-containing substrates in the context of Lp(a). With respect to KIV 10 , some species, such as chimpanzees, have a key amino acid substitution that abolishes the LBSs (34). In other species, such as baboon, the LBS is intact (in most individuals), but the lack of KV in this species prevents binding of Lp(a) to lysine-Sepharose, perhaps by promoting a conformation of apo(a) within the Lp(a) particle that masks the KIV 10 LBSs (36). Crucially, human Lp(a) can bind to lysine-Sepharose and other lysine-containing substrates, as it contains both KV and an intact LBS in KIV 10 , and is unique in this respect among the primates examined, except for orangutans (35). Therefore, it can be argued that human Lp(a) is uniquely pathogenic, although a correlation between Lp(a) levels and the extent of diet-induced atherosclerosis was observed in a study of rhesus and cynomolgus macaques domain ( Fig. 1) (7). Kringles are autonomously folding domains lacking any helical secondary structure and containing only a few short stretches of -strand (8,9). The overall kringle structure is defined as a tri-looped arrangement stabilized by the presence of three invariant disulfide bonds (8,9). Kringles are present in several other proteases involved in coagulation and fibrinolysis, including prothrombin (two kringles), Factor XII (one kringle), tissue-type plasminogen activator (two kringles), and urokinase-type plasminogen activator (one kringle) (10). Several other proteins also contain kringles, most notably the nonprotease hepatocyte growth factor, which contains four (11). The apo(a) consists of 10 different types of kringle domains, differing in amino acid sequence, that are most homologous to plasminogen kringle IV (KIV), as well as a single plasminogen kringle V (KV)-like domain and a protease-like domain ( Fig. 1) (4). Of the 10 KIV types in apo(a), 9 are present in single copy in all apo(a) isoforms (12), while KIV type 2 (KIV 2 ) is encoded in a variable number of tandemly repeated copies by the apo(a) gene (LPA), giving rise to the existence of a series of differently-sized LPA alleles and, hence, apo(a) isoforms in the human population (13). Known alleles encode as few as 1 and as many as 34 KIV 2 repeats, giving rise to apo(a) isoforms containing between 10 and 43 KIV-like domains, and polypeptide molecular masses between 200 and 800 kDa (14).
The biological role of kringles is considered to be in ligand interactions (8), often with lysine-containing substrates. Several of the kringles in plasminogen contain lysine binding sites (LBSs), defined structurally by a hydrophobic trough, lined by two or three key aromatic side chains, that binds the aliphatic backbone of the lysine side chain and that is flanked on either end by a cationic and anionic center (15,16). LBSs can therefore bind to both internal lysines, as well as carboxyl-terminal lysines. Of the LBS in plasminogen, the LBS in kringle I has the highest affinity for lysine analogs, followed by KIV and KV (17). The apo(a) lacks plasminogen kringles I-III and contains single copies of sequences similar to plasminogen KV and protease (P), although the protease domain of apo(a) is nonfunctional. The apo(a) contains 10 different types of kringle related to plasminogen KIV. The apo(a) KIV 2 is present in differing numbers of copies in different apo(a) alleles. The apo(a) KIV 9 contains a free cysteine (-SH) that mediates covalent coupling to apoB-100 in Lp(a). The apo(a) KIV 5 -KIV 8 each contain a weak LBS (white asterisks) while KIV 10 contains a strong LBS (red asterisk). tPA with fibrin result in the formation of a ternary complex that results in efficient production of plasmin (38). Plasminogen binding to fibrin converts the protein from a closed to an open conformation that makes it a better substrate for tPA (39). Moreover, partial degradation of fibrin by plasmin results in the formation of carboxylterminal lysine residues that mediate positive feedback in the fibrinolytic cascade by: i) promoting plasminogen binding (40); ii) promoting plasmin-mediated conversion of native Glu 1 -plasminogen to Lys 77 -plasminogen, which lacks the tail domain (see Biochemistry of apo(a): Homology to Plasminogen, above) and is a better substrate for tPA (41); and iii) binding to plasmin and thus protecting it from consumption by antiplasmin (42).
The first studies to examine the functional implications of the homology between apo(a) and plasminogen demonstrated the ability of Lp(a) and apo(a) to inhibit binding of plasminogen to cell surface receptors on monocytes and endothelial cells (5,6). The presumed competition between apo(a)/Lp(a) and plasminogen for carboxylterminal lysine-containing cell surface receptors, such as -enolase and annexin AII tetramer, was speculated to inhibit cell-surface (pericellular) plasminogen activation. Direct proof of an effect of apo(a)/Lp(a) on pericellular plasminogen activation was only very recently presented by our group (43); interestingly, we found that carboxylterminal lysines played essentially no role in the inhibition by apo(a). Inhibition of pericellular plasminogen activation by Lp(a) is probably not a large factor in thrombolysis in the context of atherothrombotic effects, as this would only generate plasmin on the periphery of the thrombus rather than on or in the thrombus. However, inhibition of pericellular plasminogen activation by apo(a) may contribute to the atherosclerotic process through persistence (37). The true biological role of Lp(a) is unknown, although evolution has exerted pressure to maintain an inactive protease-like domain and, less so, a lack of lysine binding.
Because human Lp(a) lacks protease activity while retaining the ability to bind to lysine-containing substrates, it can readily be hypothesized that Lp(a) may interfere with the functions of plasminogen through molecular mimicry. A great many studies conducted in vitro have confirmed this concept, while offering mechanistic insights that may prove useful to test in vivo.

EVIDENCE FOR ANTI-FIBRINOLYTIC ROLES OF Lp(a): IN VITRO STUDIES
A large body of evidence has been accumulated concerning potential pathophysiological functions of Lp(a), both proatherosclerotic and prothrombotic (Fig. 2). It is notable, however, that none of these potential mechanisms has been explicitly proved to be occurring in human patients. Moreover, because atherosclerosis and subsequent thrombosis are mechanistically interlinked, it is not clear whether the direct procoagulant/antifibrinolytic effects of Lp(a) may be at play in increasing risk for atherothrombotic events (Fig. 2). In this section and the two subsequent sections, we summarize the evidence that Lp(a) is indeed procoagulant/antifibrinolytic.

Interference with functions of plasminogen
The lysine-binding function of plasminogen is crucial to its fibrinolytic role. Activation of plasminogen by tissuetype plasminogen activator (tPA) does not occur at a meaningful rate in the absence of a fibrin surface, whereas lysine-dependent interactions between plasminogen and feedback step of plasmin-mediated Glu-to Lys-plasminogen conversion in the context of fibrin ( Fig. 3) (49). This last result is noteworthy in that it had been previously found that apo(a) was unable to inhibit fibrinolysis in a system of purified components in which the Glu-to Lys-plasminogen step was bypassed by the addition of only Lys-plasminogen (46). In addition, apo(a) was not able to influence the activity of plasmin itself in degrading fibrin (46). These findings are supported by studies of thrombolysis in rabbit jugular vein (50) and transgenic mice expressing apo(a) (51). In addition, induction of carotid of mural thrombi or through effects in the vascular wall, such as extracellular matrix breakdown, cell migration, and angiogenesis (43). Embedded in this notion is an emerging theme that we will return to in this review: the difficulty of distinguishing, in an ontological sense, the proatherosclerotic and procoagulant/antifibrinolytic effects of Lp(a).
Studies performed in vitro indicate clearly that apo(a) and Lp(a) are capable of inhibiting tPA-mediated clot lysis and inhibiting tPA-mediated plasminogen activation ( Fig. 3) (44)(45)(46)(47)(48). In addition, apo(a) is able to attenuate the positive Thrombin formed from prothrombin cleaves fibrinopeptides (A, B) from soluble fibrinogen to form fibrin monomers, which then polymerize to form insoluble fibrin. The apo(a) forms a ternary complex with fibrin, plasminogen (Plg), and tPA to prevent activation of plasminogen to form plasmin (Pln) (C). Plasmin proteolyzes fibrin to form soluble fibrin degradation products (FDPs). The apo(a) prevents plasmin-mediated conversion of Glu-plasminogen to Lys-plasminogen (D), the latter of which is a better substrate for tPA. Lp(a) alters fibrin structure to yield thinner and more dense fibers, potentially by binding to the C domain of fibrin and, hence, preventing lateral association of fibrin fibrils (E). Lp(a) concentrations for CVD events (60). It was speculated that the discordance between the two studies may be explained by increased statin use among subjects with smaller Lp(a) isoforms.
Meanwhile, some biochemical studies have sought to directly assess whether smaller apo(a)/Lp(a) isoforms have greater pathogenic effects. In fact, the majority of such studies have focused on the antifibrinolytic effects of Lp(a). It has been reported that Lp(a) containing smaller apo(a) isoforms has a higher affinity for fibrin (53,54,61,62), is more effective in inhibiting plasminogen binding to fibrin, and inhibits plasminogen activation on fibrin to a greater extent (53,63). Interestingly, however, the effect of apo(a) isoform size on affinity for fibrin was only seen in the context of Lp(a) particles, as there was no difference in fibrin affinity seen in a panel of recombinant apo(a) species varying in KIV 2 repeat number, but not complexed to apoB-100 (54). These findings are in agreement with other functional studies in which isoform size did not affect the ability of apo(a) alone to inhibit plasminmediated conversion of Glu-plasminogen to Lys-plasminogen (49). Likewise, binding of Lp(a) to mononuclear cells (THP-1 cell line) was influenced by apo(a) isoform size, with a smaller isoform displaying a higher affinity and a greater ability to inhibit plasminogen binding (64). On the other hand, recombinant apo(a) variants of two different sizes showed no differences in their ability to inhibit plasminogen activation on the surface of THP-1 monocytes, THP-1 macrophages, or human umbilical vein endothelial cells (43). In one study, however, it was reported that larger apo(a) isoforms, whether in the context of Lp(a) or not, resulted in a greater inhibition of plasmin generation in an in vitro fibrinolysis assay (48). It must be noted that many of these studies used a limited number of recombinant apo(a) isoforms or Lp(a) isoforms of various sizes in human subjects. Thus, it cannot always be concluded with certainty that the effect observed (or not observed) is not attributable to an insufficient range of apo(a) variants or subject-dependent variations in the properties of Lp(a).
Taken in sum, while it is clear that important mechanistic questions as well as the role of Lp(a) isoform size remain unresolved, the existing body of literature consistently demonstrates that Lp(a)/apo(a) can inhibit fibrinolysis as well as plasminogen activation in the context of fibrin clots or on the vascular cell surface.

EVIDENCE FOR PROTHROMBOTIC EFFECTS OF Lp(a): IN VITRO STUDIES
Several studies have investigated the possibility that Lp(a) or apo(a) might influence platelet aggregation. These studies were motivated by observations that some lipoproteins induce platelet aggregation and that apo(a) mediates the binding of Lp(a) to plasminogen receptors on the platelet surface (65,66). The available studies, however, reflect contradictory findings. A study from our group found that apo(a) and Lp(a) enhanced platelet aggregation artery thrombosis in cynomolgus monkeys showed a relationship between Lp(a) levels in the animals and the extent of cessation of flow (52).
Studies aimed at uncovering the mechanism underlying the effect of apo(a) on tPA-mediated plasminogen activation have been somewhat in conflict, with some reports suggesting that apo(a) directly competes with plasminogen for binding to fibrin (53,54), while another described a mechanism whereby apo(a) forms a quaternary complex with plasminogen, tPA, and fibrin that has a much lower turnover number than the ternary complex lacking apo(a) (Fig. 3) (55).

Structural determinants on apo(a) for anti-fibrinolytic effects
The modular structure of apo(a) has given rise to many studies that have attempted to ascribe specific functions to particular domains in apo(a), largely through the use of deletion mutants and point mutants of apo(a). In particular, our own laboratory has amassed a large collection of recombinant apo(a) variants that allows us to systematically screen the various domains in apo(a) for functional roles. Through these studies, we have found that several domains are required for maximally-efficient inhibition of tPA-mediated plasminogen activation or of plasminmediated Glu-to Lys-plasminogen conversion (47,49). The strong LBSs in KIV 10 , as well as the KV, are required for both inhibitory activities, and additional sequences within KIV 5 -KIV 8 are also required to inhibit plasminogen activation. Although the exact role of each of these domains has not been uncovered, it is likely that apo(a), by virtue of its complex modular structure, makes a series of contacts with fibrin, plasminogen, and, possibly, tPA to effect its antifibrinolytic functions (55).
An additional area of interest has been to assess the role of apo(a) isoform size on its antifibrinolytic ability. There is a general inverse correlation between apo(a) isoform size and plasma Lp(a) concentrations (56). Several epidemiological studies have attempted to determine whether small apo(a) isoforms are more harmful in their own right, i.e., independent of their association with elevated plasma Lp(a) concentrations. While evidence both for and against this notion has been presented (57)(58)(59), epidemiological studies remain inherently ambiguous about the mechanism underlying any associations. A possible exception to this is the frequently cited example of the Bruneck study, which determined that small apo(a) isoform sizes predicted advanced atherogenesis (as assessed by ultrasonically determined carotid stenosis) even after adjustment for Lp(a) concentrations (57). However, only Lp(a) concentrations predicted early atherogenesis, and then only in the presence of elevated LDL-cholesterol levels (57). It was concluded that the effects of Lp(a) on advanced atherogenesis reflected plaque thrombosis [and thus the antifbrinolytic role of Lp(a)], while the effects on early atherogenesis reflected purely atherogenic effects, possibly exacerbating the effects of LDL. However, a more recent 15 year follow-up of the Bruneck subjects revealed that Lp(a) isoforms did not add to the predictive power of structure are fibrin fiber diameter and pore sizes within the fibrin network. Dense clots have been shown to be correlated with smaller pores and fibrin fibers of reduced diameter, and with increased fibrin stiffness (85). Because thinner fibers dissolve at a faster rate than larger diameter fibers, this suggests that the rate of fibrin clot lysis is largely a function of pore size. As such, smaller pores correlate with more dense clots and increased clot lysis times, which, in turn, has been associated with increased CHD risk.
Elevated Lp(a) levels have been associated with an altered fibrin clot structure that is, in turn, accompanied by reduced fibrin clot permeability and impaired fibrinolysis (86,87). This has been interpreted to suggest a causal role for Lp(a) in altering clot structure in coronary artery disease (CAD) patients compared with healthy controls, thereby contributing to CAD risk. This may reflect the ability of the apo(a) component of Lp(a) to bind with high affinity to the fibrin(ogen) C region (88). This interaction may disrupt the lateral association of fibrin protofibrils, thereby resulting in the formation of thinner fibrin fibers (Fig. 3) (83). Additionally, the negative charge of apo(a) (89) may also underlie the ability of Lp(a) to alter fibrin clot structure, as has been demonstrated for other negatively charged substances (90).
The LPA gene contains a SNP (rs3798220) that results in an Ile to Met substitution at amino acid 4399 within the protease-like domain of apo(a) (91). The allele encoding Met has been associated with elevated plasma Lp(a) levels, small apo(a) isoform sizes, and increased risk for CHD (91). Interestingly, in the Women's Heart Study, individuals heterozygous for the Ile4399Met variant exhibited elevated Lp(a) levels and an increased risk for CAD, and benefitted more from aspirin therapy than noncarrier subjects (92). This has been interpreted to suggest a possible prothrombotic role for the Ile4399Met polymorphism. Recently, it has been reported that Caucasians heterozygous for the Ile4399Met variant exhibit elevated Lp(a) levels, increased clot density, and increased clot lysis times, while non-Causasian carriers showed increased clot permeability and shorter lysis times, with no significant increase in Lp(a) levels (93).
Clearly, mechanistic studies are required to determine the role of apo(a)/Lp(a), as well as the Ile4399Met variant, in altering fibrin clot architecture and the molecular basis for resultant alterations in clot lysis times. Demonstration that apo(a) overexpression in plasminogen knockout mice resulted in a higher incidence of thrombosis in these mice suggests a plasminogen-independent contribution of Lp(a) to prothrombotic events (94). As such, the ability of Lp(a) to alter fibrin clot architecture with attendant effects on clot lysis times merits further study in this context.

INDIRECT EFFECTS OF Lp(a) ON THROMBOSIS
As mentioned above, Lp(a) has many effects on cells and molecules of the vasculature (Fig. 2). Some of these effects could be considered directly prothrombotic (italicized and granule release mediated by the thrombin receptor activating peptide SFLLRN, but not ADP (Fig. 3) (67). Another study found that apo(a) and Lp(a) evoke platelet aggregation in response to doses of arachidonic acid that would normally be subaggregant (68). These effects were the apparent result of specific binding of apo(a) to lysinecontaining receptors on the platelet (68). On the other hand, several studies have shown that Lp(a) or apo(a) actually decreases platelet activation induced by collagen, ADP, or platelet-activating factor (69-72). One provocative conclusion from these studies might be that an antiaggregation effect of Lp(a) may exist to counterbalance its antifibrinolytic effects (71). Overall, however, it appears that the effect of Lp(a) on platelet responsiveness depends on the balance of activating factors present.
A different prothrombotic effect of Lp(a) relates to its ability to bind to tissue factor pathway inhibitor (TFPI) (Fig. 3) (73). TFPI squelches the extrinsic pathway of coagulation through its ability to bind to Factor Xa and then the tissue factor/Factor VIIa complex (74). By binding of the apo(a) moiety to TFPI through lysine residues in the carboxyl-terminal portion of TFPI, Lp(a) is able to inhibit the activity of both free and cell-bound TFPI, thus reducing the anti-thrombotic effects of this inhibitor (Fig. 3) (73). Two studies found that plasma TFPI concentrations were higher in patients with elevated Lp(a) (75,76), potentially reflecting the effects of their mutual interaction on clearance rates from plasma; one of these studies also found that TFPI activity was no different between subjects with high and low Lp(a), consistent with an inhibitory consequence of their interaction (76). Recent genome-wide association study data indicate a significant association between plasma Lp(a) levels and a small region of chromosome 2 that harbors the gene encoding TFPI (77). However, no association between 12 SNPs in the gene encoding TFPI and Lp(a) levels were detected, casting doubt on whether the chromosome 2 association is attributable to TFPI (77).

EFFECT OF Lp(a) ON FIBRIN CLOT PROPERTIES
Fibrin is the structural framework for fibrin clots in both normal physiology as well as thrombi formed as a result of vascular disease processes (78). Fibrin fibers grow both longitudinally and laterally, and branch to form a threedimensional network (79). Initially, fibrin monomers assemble into half-staggered double-stranded protofibrils; in a second step, these protofibrils associate laterally to form fibrin fiber networks (80). A variety of factors have been shown to alter fibrin clot structure, including negatively charged substances (81). Variations in clot structure have been correlated with both atherothrombotic diseases, such as CHD, as well as purely thrombotic disorders, including venous thromboembolism (81,82).
Clot structure is important in determining both the stability and fibrinolytic potential of the fibrin clot, which can, in turn, have implications for abnormal thrombolysis (83,84). The most widely used measures of fibrin clot by angiography in survivors of MI. In an interesting and underappreciated study, Moliterno et al. (108) found that elevated Lp(a) was a significant predictor of resistance to endogenous thrombolysis. Notably, typical predictors of fibrinolytic potential were not significant predictors, including levels of plasminogen activator inhibitor type 1 (PAI-1) (108). Similar findings were reported by two other studies in the ensuing years (109,110). However, one very small study (32 subjects) found that Lp(a) levels were higher in MI patients exhibiting intermittent occlusion of the infarct-related artery than patients with persistent occlusion, a result seemingly at odds with an anti-fibrinolytic role for Lp(a) (111). The rise of interventional cardiology likely means that larger prospective studies of the effect of Lp(a) on the evolution of coronary artery thrombosis will no longer be feasible, yet the available findings are a strong indicator that Lp(a) has an antifibrinolytic effect in arterial thrombosis in vivo.

Thrombolytic therapy
A potential opportunity to assess the specific antifibrinolytic effects of Lp(a) that could also represent a target for interfering with the effects of Lp(a) is in the setting of thrombolytic therapy. In the early 1990s, this question was addressed in studies of different thrombolytic agents in acute MI, most frequently derivatives of tPA, such as alteplase. Surprisingly, these studies uniformly reported that Lp(a) concentrations did not predict the success of thrombolysis in this setting (112)(113)(114)(115)(116). Given the strong in vitro findings that apo(a) can interfere with tPA-mediated fibrinolysis and plasminogen activation (see above), it is fair to ask why these clinical results may have been observed. It is possible that at the very high local concentrations of thrombolytic agent achieved, any effect of apo(a) might be overwhelmed. It is also notable that the numbers of patients in the trials were quite small (rarely exceeding 50 per arm). Therefore, there would not have been many patients with elevated Lp(a), or with the very high concentrations of Lp(a) that may be required to overcome the effects of a large concentration of thrombolytic agent. One study using anistreplase found that although Lp(a) did not affect the outcome of thrombolysis, higher levels of Lp(a) were associated with smaller decreases in plasminogen levels after therapy; this is consistent with Lp(a) promoting an antifibrinolytic state under these conditions (116).
While most studies used derivatives of tPA, two studies examined the impact of Lp(a) on thrombolysis by streptokinase (115,117). This is interesting because tPA-mediated plasminogen activation, one of the reactions that apo(a) inhibits, would not be a significant contributor. On the other hand, plasmin-mediated conversion of Glu-plasminogen to Lys-plasminogen, which is also inhibited by apo(a) (49), would still be occurring and contributing to positive feedback. However, Lp(a) levels did not affect the outcome of thrombolytic therapy in this setting either (115,117); although, in one study, a correlation between Lp(a) concentrations and the atherosclerotic burden was noted (117).
One of the complications of thrombolytic therapy is bleeding elsewhere in the vasculature, most notably and items in Fig. 2), while others would be indirect. For example, we and others have shown that apo(a)/Lp(a) can promote endothelial cell dysfunction (95)(96)(97)(98)(99)(100). This may serve to suppress the native anticoagulant function of the endothelium, while provoking a procoagulant phenotype. An important emerging functional determinant of Lp(a) pathogenicity is the presence of oxidized phospholipids (oxPLs) (101), which, in part, are phosphorylcholine-containing adducts covalently linked to, as-yet poorly defined, sequences in apo(a) (102,103). It is not known whether the oxPLs on apo(a) influence its prothrombotic effects, a question complicated by the fact that variants of apo(a) lacking the key strong LBS in KIV 10 specifically lack oxPL addition (35). Interestingly, plasminogen contains oxPLs (104,105), and the presence of this adduct actually increases the fibrinolytic potency of plasminogen (105). However, there was no correlation between the extents of modification of plasminogen and Lp(a) by oxPLs.

Arterial thrombosis
While acute coronary syndromes almost all feature thrombotic complications of atherosclerotic disease, it remains a mystery as to whether Lp(a) contributes to the former or latter processes, or both (Fig. 2). Addressing this question in a direct way is a nontrivial undertaking, not least because a comparatively large number of subjects need to be included in a study to achieve sufficient statistical power with respect to Lp(a). Moreover, some of the most dangerous lesions, vulnerable plaques, are not readily detectable by many imaging methods, and it cannot be excluded at this time that Lp(a) does not have a strong association with the development of vulnerable plaques. In theory, though, it could be determined whether Lp(a) contributes directly to atherosclerosis, such as by detecting an association between Lp(a) concentrations and angiographically or ultrasonically detectable CAD. In fact, a relatively small number of studies have attempted to do such a study, albeit in the more easily monitored carotid artery. For example, Spence and coworkers found that Lp(a) was a significant predictor of carotid stenosis (presumably reflecting atherothrombotic events), but not total plaque area (presumably reflecting the underlying atherosclerotic process) (106). Gardener et al. (107) found that Lp(a) was a significant predictor of the number of lesions within the carotid, and just barely failed to be a significant predictor of the presence of plaque. LDL-cholesterol, however, was a significant predictor of both of these parameters (107). Clearly, additional studies that specifically capture the effect of Lp(a) on lesion development without the complication of atherothrombotic stenosis need to be performed.
Other studies have examined whether Lp(a) levels influence the spontaneous resolution of coronary thrombi (i.e., not aided by administration of thrombolytic therapy) through measurement of the patency of coronary arteries disorders (132). When applied to assessment of associations between genetically elevated Lp(a) levels and different forms of venous thrombosis, however, the results have been uniformly negative (129)(130)(131). These findings quite definitely rule out a role for elevated Lp(a) in the etiology of venous thrombosis.
In the wake of these findings, it is natural to ask why, despite strong in vitro evidence, Lp(a) does not influence this set of "pure" thrombotic disorders? A strong possibility is that venous thrombosis reflects impaired anti-coagulation/ accelerated coagulation rather than impaired fibrinolysis, and likewise, that Lp(a) does not promote coagulation in vivo as much as it impairs fibrinolysis. In the genetic studies that were negative for venous thrombosis, genetically elevated Lp(a) concentrations were a strong risk factor for atherosclerotic burden in a variety of locations in the arterial tree (129,130). It has been concluded, therefore, that elevated Lp(a) promotes atherosclerotic events through promotion of atherosclerosis, not through promotion of thrombosis/inhibition of fibrinolysis. However, the notable differences in the nature of venous and arterial thrombi suggest that this conclusion be tempered.

Children
While thrombosis of any kind is rare in children, these subjects represent a special case from the point of view of understanding risk. Genetic risk factors would play a greater role in children, as other types of risk factors related to age and lifestyle (smoking, diabetes, etc.) would not yet be at play. This may explain why studies of the role of Lp(a) in thrombosis in children have consistently shown positive associations (133)(134)(135). The forms of thrombosis studied included both venous (venous thromboembolism) and arterial (incident and recurrent ischemic stroke, where the thromboembolism would not have an atherosclerotic origin). A study that did not find a significant association between elevated Lp(a) and recurrent venous thrombosis in children did, however, report that a combination of elevated Lp(a) and at least one other thrombophilic risk factor was a significant predictor of risk (136). This finding may reflect the fact that because Lp(a) affects fibrinolysis more than coagulation, its effects would only be observed when some other factor precipitated thrombosis. In other words, elevated Lp(a) may affect the penetrance of procoagulant thrombophilic risk factors. One recent study reported that elevated Lp(a) was associated with risk for recurrent, but not incident, stroke (137). Another study found no relationship between elevated Lp(a) and thromboembolism (of either venous or arterial origin) in Thai children (138). In general, the majority of the epidemiological studies on Lp(a) have focused on Caucasian populations, with some also focusing on persons of sub-Saharan African descent. As differences in the genetic architecture of LPA between different ethnic groups (such as prevalence of differently sized isoforms) are incompletely understood (139), it will be important to consider this variable in the design of future studies. devastatingly in the brain. To our knowledge, however, the impact of Lp(a) concentrations on the rate of such complications was never examined.
A more recently-emerging indication for thrombolytic therapy is for the treatment of acute ischemic stroke. One study found a trend toward higher Lp(a) levels in patients who did not achieve recanalization 1 h after initiation of thrombolysis, although this did not achieve significance; notably, levels of another fibrinolysis inhibitor, PAI-1, were significantly associated with recanalization (118). This result is interesting, as a study regarding spontaneous recanalization in coronary arteries, cited above (108), found that Lp(a) levels, but not PAI-1 levels, were significantly associated with decreased recanalization. It has also been reported that Lp(a) levels are not associated with the rate of occurrence of the most feared complication of thrombolytic therapy for stroke, symptomatic intracranial hemorrhage (119). It is fair to conclude, based on the small number of studies reported to date and the relatively small numbers of patients involved in the reported studies, that further work to definitively rule out a role for Lp(a) in influencing the outcome of thrombolytic therapy is warranted.

Venous thrombosis and thromboembolism
Thrombosis in the venous compartment, observed clinically as deep vein thrombosis or venous thromboembolism, has a very different etiology to thrombosis in large arteries. Venous thrombosis normally is not precipitated by atherosclerotic disease, as the veins are generally resistant to atherosclerosis. Compared with arterial thrombi, venous thrombi are: i) more fibrin-rich and platelet-poor; ii) generally the result of stasis or as a complication of surgery, vascular access complications, or trauma; and iii) better predicted by risk factors for thrombophilia, such as Factor V Leiden or antithrombin deficiency (120). Against this backdrop, it might be expected that elevated plasma concentrations of Lp(a) would be a risk factor for various types of venous thrombosis. Historically, the results of such studies have been mixed, with evidence both for (121)(122)(123) and against (124-128) a role for elevated Lp(a) as a risk factor for different kinds of venous thrombosis in adult populations.
More recently, however, genetic approaches have provided strong evidence against a role for Lp(a) in venous thrombosis (129)(130)(131). Because certain SNPs, such as rs3798220 (Ile4399→Met) and rs10455872 (intronic A/G polymorphism), are strongly associated with plasma Lp(a) concentrations, they can serve as surrogate markers for Lp(a) concentrations while also facilitating measurement in very large prospective cohorts where measurement of Lp(a) concentrations would be either cost-prohibitive or otherwise impossible. Moreover, being genetic markers, use of these SNPs reflects lifelong elevation in plasma Lp(a) while eliminating the artifact of "reverse causality" where the condition under study, in fact, affects Lp(a) levels, rather than the converse. Therefore, positive results from "Mendelian randomization" and similar types of studies using these informative SNPs have been considered to provide evidence that Lp(a) is a causal risk factor for CHD and other atherosclerotic and atherothrombotic an authentic mechanism through which Lp(a) causes atherothrombotic events.

PERSPECTIVES FOR THE FUTURE
While a role for Lp(a) in venous thrombosis in adults can largely be ruled out, there remains considerable evidence that Lp(a) may contribute to atherothrombotic disease, both through promotion of atherosclerosis and promotion of thrombosis/inhibition of fibrinolysis. It cannot have escaped the attention of the reader that many of the basic science and clinical/epidemiological studies cited in this review are more than 20 years old. This reflects, in part, a long fallow period in Lp(a) research prior to the publication of large genetic studies that pointed to a causal role for Lp(a) in the development of CAD (132). The time is now ripe to revisit the prothrombotic/antifibrinolytic potential of Lp(a) with a view toward understanding its mechanistic contribution to atherothrombotic events. Specific foci for future research are as follows.

The development of better animal models for Lp(a) to better address crucial biochemical and pathophysiological issues
Mice expressing high levels of bona fide human Lp(a) were reported; these animals express both human apo(a) (a truncated 8-kringle version of the protein) and human apoB-100 from transgenes (140). The effect of the expression of Lp(a) on atherosclerosis in these animals has not yet been reported. A recent study of similar mice containing a 17-kringle apo(a) transgene found that expression of human Lp(a) resulted in an increase in aortic root atherosclerosis in the setting of uremia (141). Crucially, mouse models of the thrombotic complications of atherosclerosis have recently been developed, where plaque rupture in carotid arteries of fat-fed Apoe / mice is provoked by ultrasound and thrombosis monitored by intravital fluorescence microscopy (142). These studies have illuminated roles for both platelets and the intrinsic pathway of coagulation in atherothrombosis (142)(143)(144).

More effective identification of individuals at greater risk of events due to elevated Lp(a)
Further study of the relationship of elevated Lp(a) to the development of atherosclerosis per se (including "vulnerable" plaques) is required (106). This may pave the way for the development of a prognostic "signature" specific for Lp(a), in conjunction with certain biomarkers (such as elevated coagulation factors). It may also allow for assessment of the contribution of elevated Lp(a) to events after adjustment for its effect on atherosclerosis.

Development of therapeutic interventions targeted at Lp(a) that might mitigate the contribution of this lipoprotein to events
This will require a better elucidation of the mechanisms by which Lp(a) promotes coagulation and inhibits fibrinolysis, in order to illuminate a viable therapeutic target [for example, fibrin binding by Lp(a)]. Clearly, there remains much work to be done in clarifying the interactions between Lp(a) and coagulation and fibrinolysis while, ultimately, providing proof that promotion of thrombosis is