Apolipoprotein E and clusterin inhibit the early phase of amyloid-β aggregation in an in vitro model of cerebral amyloid angiopathy

Sporadic cerebral amyloid angiopathy (CAA) is characterized by amyloid-β (Aβ) deposition in the cortical and leptomeningeal arteries, which leads to lobar hemorrhage, infarction, encephalopathies and dementia [40, 41]. Recently, several groups reported that various proteins are deposited with Aβ in the vessel walls of postmortem, autopsied CAA patients [14, 17, 20]. Apolipoprotein E (apoE), clusterin (CLU, also called apolipoprotein J. apoJ), serum amyloid-P component, tissue inhibitor of metalloproteinases-3, sushi repeat-containing protein 1 (SRPX1) and other proteins are the major Aβ-associated proteins in the vessel walls of CAA patients. However, the effects of these proteins on the development of CAA are not clear except for SRPX1, which may enhance the cerebrovascular degeneration induced by Aβ [17].

In this study, we first performed proteome analysis of biopsied leptomeningeal and cortical vessels obtained from 6 CAA patients and 5 non-CAA patients who underwent surgery for large lobar hemorrhages [18]. We found that expression of apoE and CLU is significantly increased in the vessels of acute-onset, symptomatic CAA patients as compared to non-CAA patients. Next, we examined the effects of apoE and CLU on the early phase of Aβ aggregation, using a powerful, physiologically relevant in vitro model of CAA [10]. This model was previously established to test the hypothesis that intramural periarterial drainage (IPAD) flow carries Aβ to the vascular basement membrane (BM), which may trap Aβ and induce amyloid fibril formation in vivo [1, 22].

In this study, we first performed proteome analysis of the Aβ-deposited leptomeningeal and cortical vessels (Tables 1 and 2). To the best of our knowledge, this is the first report of the proteome analysis of vessels biopsied from living symptomatic, clinical CAA patients. In previously published reports, proteome analysis was performed with leptomeningeal vessels obtained from postmortem, autopsied cases, in which Aβ deposition was histopathologically confirmed [14, 17, 20]. Although CAA is known to be more severe in posterior brain regions compared to anterior brain regions [41] and postmortem proteome analysis was performed using the vessels derived from occipital lobes [14, 17, 20], frontal lobes were involved in 3 CAA patients and parietal and temporal lobes were involved in other 3 CAA patients (Table 1). This difference in the brain regions from which the vessels were obtained may affect the results of proteome analysis (Table 2 vs. [14, 17, 20]). For example, this may explain why we found no significant upregulation of tissue inhibitor of metalloproteinases-3 (Table 2), in contrast to the finding of Manousopoulou et al. [20]. Additionally, the discrepancy in male/female ratio between CAA and non-CAA patients (male/female: 0/6 and 4/1, respectively) may also affect the results of our proteome analysis. Both apoE and CLU are representative signature proteins in various types of systemic amyloidosis [2]. Thus, it is worth noting that apoE and CLU are representative Aβ-associated proteins in the vessel walls of clinical CAA cases (Table 2), as well as in those of pathological CAA cases [14, 17, 20].

Carare and coworkers proposed the IPAD pathway model [1, 22, 34]. This model constitutes that instead of the conventional lymphatics, interstitial fluid and solutes are drained from the brain parenchyma to cervical lymph nodes along BMs in the walls of cerebral capillaries and tunica media of leptomeningeal arteries. Through this pathway, Aβ peptides, especially Aβ(1–40) produced by cortical neurons are carried away from the brain parenchyma [1, 22]. Reduced Aβ trafficking through this pathway may result in the aggregation of Aβ amyloid in the cerebrovascular BMs, leading to the manifestation of pathological as well as clinical CAA [41]. However, the driving mechanisms of this pathway and the molecules affecting the trafficking and aggregation of Aβ in this pathway are not fully understood [41, 43]. We previously established a simple yet powerful in vitro model of CAA, which recapitulates the IPAD flow draining Aβ and other solutes, as well as the vascular BM as a scaffold for Aβ aggregation in vitro [10]. To the best of our knowledge, this is the first, and the only in vitro system which recapitulates the pathogenesis of CAA in a physiologically relevant manner (Fig. 1). First, we completely removed the air-water interface, which is a hydrophobic-hydrophilic interface that potently induces protein denaturation and amyloid formation [21]. Complete removal of the air-water interface makes it possible to detect weak effects of BM molecules on the induction of amyloid formation in vitro. Second, we reduced Aβ concentration as low as possible (5 μM) to inhibit the spontaneous aggregation of Aβ in the reaction mixture. We also added HSA (4.5 μM) to mimic the CSF environment. Finally, we reproduced the IPAD flow in vitro by gently rotating the plate at 1 rpm. As Matrigel-coated beads slowly sink from the top to the bottom of a well, their surfaces are exposed to the relative countercurrent of the reaction mixture to induce the interaction of Aβ with BM molecules. Using this model, we have demonstrated that apoE and CLU inhibit the early phase aggregation of Aβ in vitro (Figs. 4 and 5).

Although apoE is considered as a key player in the pathogenesis of Alzheimer’s disease (AD) and CAA [16, 32], the effects of apoE on the aggregation of Aβ in vitro and in vivo are controversial. While some groups reported that apoE accelerates Aβ fibril formation in vitro [31, 35], we and other groups reported that apoE inhibits Aβ aggregation in vitro [6, 8, 15, 24, 37]. These opposite effects may be partly due to the difference in the concentrations of Aβ used in each experiment. We previously reported that apoE may inhibit or enhance Aβ amyloid fibril formation in a concentration-dependent manner [24]. When 50 μM Aβ(1–40) was incubated with 50–500 nM apoE, apoE dose-dependently inhibited Aβ amyloid fibril formation. In contrast, when 300 μM Aβ(1–40) was incubated with 3 μM apoE, apoE slightly enhanced Aβ aggregation. Similarly, while some groups reported that apoE promotes Aβ amyloid deposition in vivo [12], other groups showed that apoE delays Aβ amyloid deposition in vivo [4, 7, 13, 15, 32]. LaDu and coworkers produced EFAD mice, which are a tractable familial AD-transgenic (FAD-Tg) mouse model expressing human APOE rather than mouse APOE [32]. Consistent with our data (Fig. 4), they showed that introduction of human APOE to EFAD mice delays extracellular Aβ accumulation (not only plaque deposition but also CAA) from ~ 2 to 6 months compared with the control 5xFAD mice expressing mouse APOE [32]. They suggested that the mouse apoE is structurally and functionally distinct from human apoE.

The pathogenesis of AD and CAA is affected by apoE isoform-dependently [16, 32]. Robust data confirmed that ε4 allele of APOE is not only the risk factor of AD but also that of nonhemorrhagic-type CAA [3, 16, 32, 40]. In contrast, while the ε2 allele of APOE is protective for the manifestation of AD, it is a risk factor of hemorrhagic-type CAA [3, 16, 32, 40]. Tai et al. reported that the ability of human APOE to delay the extracellular Aβ accumulation in EFAD mice was in the order of 5xFAD < E4FAD < E3FAD ≤ E2FAD [32]. Consistent with this in vivo observation, Hori et al. reported that the in vitro conversion of Aβ protofibrils to fibrils progressed more slowly upon co-incubation with apoE2 or apoE3 as compared to the case with apoE4 [15]. In contrast, we found that the inhibitory effect of apoE3 on the kinetics of early phase Aβ aggregation was not significantly different from that of apoE4 (Fig. 6). It is hypothesized that apoE affects the pathogenesis of AD and CAA through a variety of mechanisms, including the effects on Aβ aggregation, Aβ transport and clearance from the interstitial/cerebrospinal fluid, and cellular metabolism of Aβ [16]. Thus, the linkage of ε4 allele of APOE to the manifestation of nonhemorrhagic-type CAA might result from mechanisms other than the direct effects of apoE on Aβ aggregation. Future studies are eagerly awaited to resolve this issue.

The pathogenesis of AD and CAA is affected by CLU [11, 36, 38]. Wilson and coworkers reported that CLU inhibits Aβ aggregation in vitro [26, 39, 42]. Consistent with our data (Fig. 5), Yerbury et al. reported that CLU significantly inhibits Aβ amyloid fibril formation at a molar ratio of CLU:Aβ =1:100 [42]. The effects of CLU on the aggregation of Aβ in vivo are controversial. While DeMattos et al. reported that CLU promotes amyloid plaque formation in vivo [5], Qi et al. reported that CLU reduces Aβ plaques as well as the severity of CAA in vivo [30]. Interestingly, DeMattos et al. reported that apoE^−/−/Clu^−/− PDAPP mice had both earlier onset and marked increase of Aβ deposition, suggesting that apoE and CLU cooperatively lower the Aβ level and suppress deposition [4]. Importantly, Wojtas et al. found a marked decrease in cortical plaque deposition but an equally striking increase in CAA in the brains of APP/PS1;Clu^−/− mice as compared to Clu^+/+ mice. They proposed that CLU facilitates Aβ clearance along the IPAD pathway by preventing binding of Aβ to cerebrovascular BMs [36]. This model is consistent with our data indicating that CLU acts as an extracellular chaperone to prevent the manifestation of CAA.

Finally, we consider how apoE and CLU inhibit amyloid formation in vitro. Based on a nucleation-dependent polymerization model [19, 23], we confirmed that apoE and CLU inhibit the early phase of Aβ aggregation (Figs. 4 and 5) as well as the seeded aggregation of Aβ amyloid fibrils (Fig. 3). Hori et al. reported that apoE interacts with Aβ protofibrils in the order of apoE2 = E3 > E4, leading to the inhibition of the conversion of Aβ protofibrils to fibrils [15]. Narayan et al. reported that CLU binds Aβ oligomers formed during the aggregation of Aβ monomers, thereby inhibiting the further growth of these oligomers into mature amyloid fibrils [26]. These reports may indicate that apoE and CLU interact with Aβ nuclei/on-pathway oligomers formed on the Matrigel-coated beads, leading to the inhibition of the successive aggregation of these species into amyloid fibrils (Figs. 4 and 5). Additionally, we previously reported that, like CLU, apoE inhibits the in vitro growth of Aβ amyloid fibrils by binding and sequestering Aβ monomers [24] (Fig. 3).

We performed proteome analysis with the vessels biopsied from symptomatic, clinical CAA patients and confirmed that the expression of both apoE and CLU are significantly increased in the vessels of CAA patients as compared to non-CAA patients. Next, we used a unique in vitro model of CAA to confirm that apoE and CLU specifically inhibit the early phase of Aβ aggregation on the surface of BM-coated beads. The interaction of Aβ with vascular BMs may be a promising therapeutic target for CAA. Future studies are essential to develop the therapeutics for CAA.