Antiphospholipid reactivity against cardiolipin metabolites occurring during endothelial cell apoptosis

Patients with antiphospholipid syndrome (APS) show high levels of circulating antiphospholipid antibodies (aPLs) and are prone to arterial and venous thrombosis, recurrent abortions, and/or foetal loss [1‐3]. aPLs are heterogeneous antibodies binding to phospholipids, proteins, or phospholipid-protein complexes. Historically, cardiolipin (CL) was used as an antigen for aPL determination [4]. In 1990, different reports described the requirement of beta₂-glycoprotein I (β₂-GPI) for the binding of anticardiolipin antibodies (aCLs) in solid-phase immunoassays [5‐7]. Although β₂-GPI represents the best target antigen in the pathogenesis of APS, other phospholipid-binding proteins have been described as phospholipid cofactors [8‐12]. Moreover, it has been reported that in infectious diseases aCL recognises CL in the absence of β₂-GPI [13‐16].

CL is a unique anionic phospholipid composed of two phosphate groups and four fatty acid chains and represents the most unsaturated (susceptible of oxidation) lipid of the body [17, 18]. In particular, the degree of unsaturation of the acyl chains in CL influences the binding of β₂-GPI to CL, whereas (hydro)peroxidation of CL has been shown to be essential for enhancing the binding of aPLs [19]. Although CL is predominantly associated with the inner mitochondrial membrane (where CL synthase is located), its rapid re-modelling into the highly unsaturated species that are most common in adult tissues (for example, tetra-linoleyl-CL) occurs in other membranes. CL re-modelling, in fact, involves relocation to the outer mitochondrial membrane, as well as to extra-mitochondrial compartments [17, 18], with rapid de-acylation into mono- and di-lysocardiolipin (with three and two acyl chains, respectively). These metabolites are transported to the endoplasmic reticulum (ER) for efficient re-acylation into the mature forms of CL found in mitochondria in a process that seems to be facilitated by lipid transfer proteins like Bid [20]. Consistent with such a multi-organelle cycle of CL re-modelling, we have recently shown that CL is exposed on the plasma membrane (PM) of cells undergoing apoptosis induced by death receptors like Fas and tumour necrosis factor-alpha (TNF-α) [21, 22]. This translocation onto the cell surface implies a leakage of CL (and/or of its metabolites) from the normal re-modelling cycle [20], probably as a consequence of an apoptosis-mediated increase of ER and secretory membranes. This leak might well represent an in vivo trigger for the generation of aCL [21, 22]. Interestingly, mass spectroscopy analysis has demonstrated an early degradation of mitochondrial CL into its immediate metabolite, monolysocardiolipin (MCL), during Fas-induced apoptosis [23]. This finding has been subsequently confirmed in human pro-monocytic U937 cells [22].

The recent data have increased our attention on the role of CL metabolites and, in general CL acylation, on the generation and properties of aCL, a subject that has been analysed by a limited number of studies so far [24, 25]. In the present study, we demonstrate that the number of acyl chains in CL derivatives is important not only for the binding of β₂-GPI, but also for the generation of epitopes of 'pure' aCL for specific CL metabolites like MCL. In addition, we show that CL and its key derivative, MCL, re-locate to the PM of human umbilical vein endothelial cells (HUVECs) undergoing apoptosis. We suggest that aCL and antimonolysocardiolipin antibody (aMCL) might derive from alteration in the metabolism of CL as a possible consequence of enhanced apoptosis.

In this study, we report for the first time that sera from APS and SLE patients display strong reactivity toward MCL, a key metabolite of CL. MCL is the immediate product of CL degradation, a process that has been shown to occur either in mitochondria (both membranes) as part of the re-modelling cycle of the mature lipid or in lysosomes [20, 42]. CL is among the most resistant phospholipids toward phospholipase A2 hydrolysis in membranes of healthy cells. However, significant levels of MCL exist in healthy tissues and increase after stimulation of physiological apoptotic pathways such as Fas/FasL, which are relevant to white blood cells and immunological response [22, 23]. In this context, our finding of autoantibodies that cross-react with CL and MCL could provide a pathologically relevant link between the metabolic cycle and membrane traffic of CL and apoptosis. Indeed, CL is known to undergo degradation (including peroxidation) during many pathways of cell death [42].

Recent findings have suggested that CL and its metabolites may be transported by apoptosis regulator proteins like Bid, especially after caspase cleavage [23, 43]. On the other hand, we recently demonstrated that CL becomes exposed onto the PM of myelomonocytic cells undergoing apoptosis in vitro, suggesting that intracellular traffic of CL may enhance CL immunogenicity in vivo [21] as previously described for other autoantigens [44]. Of relevance to this issue, Casciola-Rosen and coworkers [45] have shown aCL binding to surface blebs of apoptotic cells, which would be consistent with the clustering of aCL immunostaining in focal surface regions that we detected in apoptotic endothelial cells. This indicates that cells undergoing apoptosis expose CL on their surface in segregated membrane regions that could enhance the binding of circulating autoantibodies. In this view, we consider that our previous findings of CL reactivity on the cell surface of apoptotic cells [21] could reflect a combined recognition of exposed CL and MCL. Indeed, our original findings in the present work suggest a prominent role of both CL and MCL in eliciting a pathological autoimmune reactivity in APS and related conditions. Moreover, we demonstrated for the first time the relocation of CL and MCL to the PM of apoptotic HUVECs, which are largely involved in the pathogenesis of APS. In contrast, in our extensive mass spectrometry (MS) studies of membrane lipids and their changes after the induction of apoptosis, we detected very low levels of MS signature ions for DCL species in the PM of apoptotic cells, suggesting that in most cells DCL is present in trace amounts with respect to CL and MCL, most likely reflecting a combination of rapid re-modelling turnover in the ER and lysosomal degradation in by-products of CL catabolism [22]. These findings might explain the low DCL reactivity that we found in sera of both APS and SLE patients.

Our data also demonstrate that the chemistry of the acyl chains of CL is important not only for the binding to β₂-GPI, but also for the intrinsic immunogenicity of the CL molecule. We also demonstrate here that the serum protein β₂-GPI shows a differential binding to CL derivatives (MCL > CL > DCL), which, although similar, does not entirely match the reactivity exhibited by SLE and APS sera. In fact, by using both ELISA and TLC immunostaining, we found that sera can also bind with high efficiency to MCL and CL in the absence of β₂-GPI. Thus, our data support the notion that aCL and anti-β₂-GPI represent two distinct populations of antibodies [40] with overlapping but not coincident antigenic recognition. In this respect, pathogenic human monoclonal antibodies that bind phospholipids in the absence of β₂-GPI have been described [46, 47]. Likewise, the binding of plasmatic β₂-GPI to CL and MCL on the cell surface after apoptotic stimuli is possible. Thus, we cannot exclude the possibility that aPLs may also be directed to this phospholipid-protein complex.

After an early report showing different reactivity of various CL derivatives with Wassermann antibody [48], only a few investigations have been published on the role of the acyl chains of CL in binding of aPLs [24, 25, 49‐51]. Notably, Qamar and co-workers [49] suggested a potential role of the fatty acyl chains in the binding of autoantibodies to lysophosphatidylethanolamine. Another study concluded that aPL binding as monitored with an ELISA system depends on the fatty acid moiety of phospholipids but is also influenced by the nature of their polar head and phosphodiester groups [24]. Interestingly, the authors of this work suggested that in vivo aPLs may bind to intermediate ('transition') membrane phospholipids, which could expose their fatty acid chains that contributed to antibody recognition. In accordance with recent insights into the role of CL and its re-modelling in apoptosis [23, 42, 43], we propose that both CL and MCL are the 'transition' membrane lipids suggested earlier [24]. In agreement with this possibility, Berger and colleagues [25] suggested that the number of acyl chains in CL may represent a significant factor in the binding of aPLs. However, they detected a significant decrease of binding to MCL in 12 SLE aCL-positive sera [25]. The discrepancy with our data may be due to the selection of patients based on their high reactivity to CL or to the small number of patients tested as well as to the different dilution of the sera [25]. More recently, antibodies to lysophosphatidylcholine (LPC) have been identified indicating that LPC may also contribute to the antigenicity of oxidised low-density lipoproteins [49, 50]. Moreover, a strong correlation between antibodies directed to LPC and CL was found [50]. It is noteworthy that LPC is a by-product of CL re-modelling, PC being the predominant acyl donor for MCL during re-modelling [18]. On the other hand, it has also been suggested that CL, after oxidation, could be hydrolysed by serum phospholipase, thus creating molecular structures similar to either LPC or MCL [50]. Interestingly, phospholipase A2 activity is enhanced in both autoimmune and inflammatory diseases [51], as well as during apoptosis [20]. Consistent with our study, it is thus possible that CL hydrolysis of one fatty acid chain to yield MCL enhances CL antigenicity. In addition, MCL has hybrid properties between a diacyl-lipid (bilayer-forming) and a lyso-lipid (micelle-forming), which may physically facilitate autoantibody reactivity. In contrast, there are fundamental physico-chemical reasons why autoantibodies fail to bind to DCL because this lipid has an unusually large and hydrophilic polar head with respect to any other (di-acyl) phospholipid. DCL may have as many as five free hydroxyl groups in its polar head, some of which are likely to produce intramolecular hydrogen bonds that would change the geometry and overall structure of the lipid. Then, in the likely event of partial peroxidation of the acyl chains, additional hydrogen bonds could be formed between the polar head groups and the oxygenated fatty acids, producing large changes that alter the overall structure of the molecule. Such changes will strongly modify the docking properties and on/off constants of lipid binding to the cognate hydrophobic pockets in serum immunoglobulins (or other CL-interacting proteins). In the case of APS serum proteins, their binding to CL is critically determined by the conformation and chemistry of the acyl chains, as demonstrated here by the lack of binding to the hydrogenated CL analogue, which shows baseline levels comparable with those of DCL. Whereas MCL could maintain the overall structural determinants that enable the recognition of CL, the further loss of an acyl chain in DCL destroys the binding affinity because of the large structural alterations that distinguish this lipid from its related metabolites and, overall, from most other di-acyl lipids.

In this study, we describe a high reactivity of APS and SLE sera to MCL, the immediate degradation product of mitochondrial CL. We propose a model in which apoptosis or inflammatory processes enhance the hydrolysis (and membrane traffic) of CL into MCL, which can then escape the metabolic cycle of lipid re-modelling. MCL can then become exposed to the cell surface and consequently to the immune system, either directly or through the interaction with a plasma protein such as β₂-GPI. Interestingly, these events may occur in vivo in cells directly involved in the pathogenesis of APS, such as monocytic and endothelial cells, as demonstrated by us in in vitro experiments. Hence, a deranged process of CL metabolism could stimulate autoantibody reactivity by the synergistic binding to MCL.