Background
Accumulation of β-amyloid (Aβ) in the brain is a culprit early in Alzheimer’s disease (AD) pathogenesis, and triggers down-stream neurodegenerative cascades including inflammatory responses, intraneuronal neurofibrillary pathology, and synaptic and neuronal loss (reviewed in [
1‐
3]). Susceptibility to sporadic AD is foremost modulated by
APOE genotype, which also influences the load of Aβ parenchymal plaques and vascular Aβ (VAβ) deposits. A single copy of the
APOE ε4 allele endows a ~3 fold increase in AD risk, and 2
APOE ε4 copies result in a ~15 fold risk increase, while an
APOE ε2 allele halves AD risk relative to 2 copies of ε3 (reviewed in [
4]). Autopsy series and more recently positron emission tomography (PET) imaging studies of fibrillar Aβ plaque load in AD patients have shown ε4> > ε3 > ε2 allele gradation effect on Aβ deposition [
1‐
3]. To counteract down-stream neurodegenerative effects triggered by Aβ accumulation, development of anti-Aβ therapeutic strategies including anti-Aβ immunotherapy have been proposed and pursued. Several anti-Aβ monoclonal antibodies (mAbs) have been tested in clinical trials in AD patients and were found to significantly reduce load of fibrillar Aβ as demonstrated using PET Aβ imaging [
5‐
8]. Whether
APOE genotype also differentially modulates degree of Aβ plaque load reduction in response to anti-Aβ passive immunization remains unknown due to limited clinical data and because preclinical testing of anti-Aβ mAbs has been exclusively conducted in AD transgenic (Tg) mice models expressing wild type, murine apoE [
9‐
12]. The main adverse effects associated with administration of certain anti-Aβ mAbs during clinical trials were amyloid related imaging abnormalities (ARIA) identified on magnetic resonance imaging (MRI) scans. These included vasogenic edema (ARIA-E) and cerebral microhemorrhages (ARIA-H) that in about 20% of cases are associated with clinical symptoms and signs [
13,
14]. Frequency of ARIA events and in particular ARIA-E was significantly higher among
APOE ε4 allele carriers compared to non-carriers, making the
APOE ε4 allele a risk factor for vascular complications of anti-Aβ immunotherapy [
5,
8,
14,
15]. In view of these considerations, we sought to re-examine the effects of passive immunization in APP
SWE/PS1
dE9 Tg mice with targeted replacement of the murine
Apoe gene for various human
APOE alleles, which expression remains controlled by the native murine
Apoe promoter [
16,
17]. These Tg mouse lines, hereafter designated as APP/ε2, APP/ε3, and APP/ε4, reflect the differential effect of
APOE alleles on the load of Aβ parenchymal plaques and VAβ known from AD patients [
18]. The vaccination experiment was started in 12 months. old mice with advanced load of Aβ deposits [
19,
20], to emulate the stage of human disease in regard to Aβ deposition in which anti-Aβ immunotherapy is currently occurring. We used mAb 10D5 direct against Aβ
3–7 epitope [
10], which is known to penetrate the blood–brain-barrier (BBB) and directly binds to deposited Aβ triggering microglial cells to clear Aβ plaques through Fc receptor-mediated phagocytosis [
9]. Our study in
APOE humanized APP
SWE/PS1
dE9 mice provides evidence for differential effect of
APOE alleles on response to anti-Aβ immunotherapy and occurrence of vascular complications associated with thereof.
Discussion
APOE genotype critically affects the burden of Aβ pathology in sporadic AD. ApoE isoforms encoded by various
APOE alleles differentially influence the rate of soluble Aβ clearance from the brain interstitial space and formation of Aβ parenchymal plaques and VAβ [
32,
33]. Our study provides novel evidence that
APOE genotype also differentially affects multiple aspects of response to Aβ immunotherapy. Firstly, we observed that
APOE genotype is associated with variable reduction in the load of Aβ parenchymal plaques and APP/ε4 mice had greater reduction in absolute Aβ plaque load values than APP/ε2 and APP/ε3 mice. The reduction concerned both Th-S positive, (fibrillar) and immunopositive Aβ (total) plaque loads with the later metric showing greater change than the former across all
APOE genotypes. As the immunopositive Aβ plaque load is inclusive of Th-S positive plaques this suggests that stimulated by 10D5 mAb microglia clear diffuse component of Aβ plaques more effectively than they do fibrillar Aβ aggregates. However, despite enhanced treatment effect in APP/ε4 mice the post-treatment Aβ plaque load in these mice and especially the fibrillar Aβ plaque load, remained significantly higher than in APP/ε2 and APP/ε3 mice, what is a result of distinguishably greater Aβ plaque load, associated with the ε4 allele. This observation may inform design of future passive immunization experiments in APP Tg model mice of AD and clinical trials suggesting specific tailoring of the anti-Aβ passive immunization protocol in
APOE ε4 allele carriers to achieve end-point plaque load reduction level compared to that expected among non-ε4 allele carriers. Potential modifications of the immunization protocol may include increasing antibody dose, earlier commencement of the treatment and extension of treatment duration. However one needs to be mindful that escalation of anti-Aβ mAbs dose in
APOE ε4 allele carriers may trigger increased rate of vasculotropic adverse events associated with immunization [
5,
8,
14]. With the advancement of certain newer anti-Aβ immunotherapies such as Gantanerumab and Aducanumab that appear to lower Aβ plaque load in humans to a greater extent than seen previously [
6,
8], it will be interesting to see if the effects of
APOE genotype noted here in animal models are also present in humans.
The primary mechanism of action of 10D5 mAb is based on its direct binding to deposited Aβ and stimulating Fc receptor-mediated Aβ plaque removal by activated microglia and possibly blood derived macrophages [
9,
34]. To compare microglial response across
APOE genotypes we immunostained brain sections against Iba1 and CD68 antigens and counterstained them with Th-S for fibrillar Aβ. Iba1 is a microglia specific marker, which expression increases with microglia activation and macrophage transformation [
30], while CD68 is a lysosomal/endosomal membrane glycoprotein up regulated in actively phagocytic cells [
31]. The load of Iba1 and CD68 positive cells was significantly higher in Age control and TY11-15 control APP/ε4 mice compared to APP/ε2 and APP/ε3 controls. Both Iba1 and CD68 loads correlated with the load of Th-S positive fibrillar deposits and when adjusted for it, they were comparable across all
APOE genotypes in control animals. Following 10D5 mAb treatment both Iba1 and CD68 loads increased across all
APOE genotypes, however when adjusted for the post-treatment fibrillar plaque load, they were significantly higher in APP/ε4 mice than in APP/ε2 and APP/ε3 animals. This finding illustrates increased effector function of 10D5 mAb exerted on microglia in the setting of the
APOE ε4 allele. In particular exaggerated post treatment CD68/Th-S ratio in APP/ε4 mice implies less effective Aβ degradation, which can be tied to a recent discovery showing that excessive apoE accumulation within Aβ plaques impairs ability of microglia to contain plaque load [
35]. In fact, we have previously demonstrated that APP/ε4 and PDAPP/ε4 mice have greatly increased content of apoE within Aβ deposits compared to the same Tg mice lines expressing other human
APOE alleles [
19,
36]. Though microglial cells play pivotal role in maintaining immunoproteostasis through phagocytosis and degradation of misfolded proteins, their protective function is inseparably associated with immune activation exerting a chronic, potentially harmful effect on the regional milieu of neuronal networks [
35,
37‐
39]. Although, APP/ε4 mice show the greatest absolute reduction in the load of parenchymal Aβ plaques in response to anti-Aβ immunization as compared to mice of other
APOE genotypes, this occurs at the expense of unparalleled microglia response making the post treatment Iba1 and CD68 loads the highest among all experimental groups. Thus both higher level of microglia activation driven by higher Aβ plaque load and exaggerated microglia effector response exerted by anti-Aβ immunotherapy may constitute independent deleterious effects associated with the ε4 allele, in AD pathogenesis. Increased microglia activation and associated reduction in Aβ plaque load was found only in 10D5 mAb immunized groups, while administration of the isotype control antibody TY11-15 changed neither expression of microglia markers nor parenchymal plaque load as compared to untreated age-matched animals of matching
APOE genotypes.
Both CAA and CCA variants of VAβ pathology have been described in AD autopsies and their relative preponderance has been linked to various
APOE genotypes [
40‐
43]. CAA comprises of fibrillar Aβ sandwiched between the adventitia and the tunica media of cerebral arteries, while CCA comprises of non-fibrillar Aβ deposits closely associated with the capillary outer basement membrane. Both development of CAA and CCA is primarily driven by neuron derived Aβ, while Aβ produced by myocytes locally in the tunica media further contributes to CAA development (reviewed in [
42]). Presence of apoE within CAA lesions is critical for formation of Aβ fibrillar assemblies [
25]. In this study we found that
APOE genotype modulates development of both CAA and CCA pathology in APP Tg mice. In accordance to previously published data concerning ε4 allele effect, we found increased CAA load in APP/ε4 mice [
25,
44‐
46]. Our novel observation here is that the ε2 allele was associated with greatly reduced CAA load, but also with enhanced incidence of CCA. Reduced CAA load in APP/ε2 mice likely reflects limited effect of apoE2 isoform on promoting Aβ fibrillar assemblies within arterial walls as compared to other human apoE isoforms, which are associated with higher load of CAA and fibrillar Aβ plaques [
47]. In turn, greater incidence of CCA may be related to increased amount of soluble Aβ cleared across the BBB in APP/ε2 animals, which deposit less Aβ in the brain while maintaining the same level of Aβ production [
32]. In fact increased rate of soluble Aβ clearance across the BBB has been experimentally confirmed in the setting of the
APOE ε2 allele [
32,
48]. Vascular disease risk factors including hypercholesterolemia and volume overload occurring in aged APP/ε2 mice effect deterioration of capillary wall structure and function and likely in tandem with greater load of soluble Aβ passing across the BBB contribute to CCA development.
The largest percentage of AD patients with both CAA and CCA was found among ε4 allele carriers, while CAA without CCA was more frequent among ε2 and ε3 carriers [
40,
42]. In these studies ε3 allele carriers and in particular ε3 allele homozygotes constituted the largest percentage among CAA/CCA free cases. Accordingly, we found that APP/ε3 mice have the lowest incidence of CCA among all
APOE genotypes and their CAA load is significantly lower than that in APP/ε4 mice demonstrating protective effect of the ε3 allele against development of Aβ angiopathy. However, mechanisms underlying VAβ deposition including preferential development of CAA and CCA remain elusive and require further elucidation. They likely are a resultant of multiple genetic features extending beyond
APOE genotype as no
APOE allelic variant confers absolute protection against VAβ deposition [
42,
43].
VAβ is detrimental to the function of cerebrovascular circulation through compromising blood–brain-barrier integrity [
49], effecting hemodynamic dysfunction [
29,
50], and facilitating occurrence of spontaneous hemorrhages [
25,
26,
36]. We found a significant number of perivascular hemosiderin deposits, reflecting ensued microhemorrhages, in brains of control mice representing all
APOE genotypes. They were associated both with CAA and CCA affected vessels. Strikingly the lowest incidence of perivascular hemosiderin deposits was observed among untreated age-matched and TY11-15 control APP/ε3 mice. In APP/ε4 and APP/ε2 controls the incidence of perivascular hemosiderin deposits was comparable and statistically significant higher than that in APP/ε3 controls. In human subjects prevalence of spontaneous microhemorrhages is high and positively correlates with advancing age. MRI studies detected microbleeds in 17.8% of subjects aged 60–69 years and in 38.3% in subjects who were more than 80 years old [
51,
52]. For comparison an autopsy study reported histological evidence of ensued microhemorrhages in 92.9% of elderly individuals with mean age of 81.1 ± 10.8 years, which included both AD and non-demented subjects [
53]. Evidence independently supports link of both the
APOE ε4 and ε2 allele with increased risk for microhemorrhages in humans [
51,
52,
54‐
57]. Both VAβ and hypertensive or atherosclerotic microangiopathy are recognized clinical risk factors for cerebral microbleeds, with the former correlating stronger with bleeding in cortical/subcortical distribution [
58,
59] while the later with bleeding to deep brain structures including basal ganglia and thalamic nuclei [
60,
61].
APOE ε2 targeted replacement mice develop type III hyperlipoproteinemia and spontaneous atherosclerosis [
16] as well as volume overload as a function of their obesity, all of which are recognized vascular risk factors and may explain increased preponderance to thalamic localization of hemosiderin deposits in APP/ε2 mice found in our study. However, these striking vasculotropic effects observed here in APP/ε2 mice may not be robustly seen in human subjects, where ε2 homozygotes constitute only 1% of the population, with less than 10% of them developing type III hyperlipoproteinemia, what is linked to distinctly different contribution of apoE to VLDL formation between humans and mice [
62].
Aβ immunotherapy can ameliorate VAβ burden and improve compromised vascular reactivity [
12,
21,
63], but it is well known to escalate incidence of brain microhemorrhages [
21,
26]. Both CAA load and CCA incidence were reduced in 10D5 mAb treated mice with the strongest effect on CAA load reduction seen in APPε4 animals and the strongest effect on CCA incidence seen in APP/ε2 mice. We also found a profound effect of
APOE genotype on Aβ immunization related microhemorrhages. Surprisingly not the ε4 allele but the ε2 allele was associated with the greatest increase in the number of perivascular hemosiderin deposits. 10D5 mAb treated APP/ε2 mice had the highest increase in the number of all perivascular hemosiderin deposits and also in the number of deposits in small and large (≥15 μm in diameter) subcategories. The increase in the total number of hemosiderin deposits in APP/ε2 mice was approximately two-fold higher than these in APP/ε3 and APP/ε4 mice. APP/ε2 mice represented the only
APOE genotype where 10D5 mAb immunization increased incidence of large hemosiderin deposits, bringing their number to three-fold value of the control group. Several human autopsy studies have shown that the ε2 allele is an exacerbating factor of microhemorrhages among subjects with VAβ pathology and that this effect is linked to fibrinoid necrosis of vascular wall specifically associated with the ε2 allele [
55‐
57]. As mentioned above APP/ε2 mice also develop hypercholesterolemia and atherosclerosis, hence increased susceptibility to cerebral microhemorrhages and in particular high incidence of large hemosiderin deposits in APP/ε2 mice associated with immunotherapy can be a resultant of prevalent VAβ pathology, anti-Aβ mAb effect, and cerebrovascular risk factors all cooperatively compromising vascular wall integrity. In contrast, 10D5 mAb treated APP/ε3 and APP/ε4 mice showed only an increase in the number of small hemosiderin deposits. Again the post-10D5 treatment number of hemosiderin deposits in APP/ε3 mice was the lowest among all
APOE genotypes, indicating a relative vasculoprotective effect of the ε3 allele, which has not been previously described in AD patients or APP Tg mice expressing human
APOE alleles.
Exacerbation of VAβ associated microhemorrhages appears to be associated with propensity of anti-Aβ mAb to bind deposited Aβ. Thus, application of N-terminus specific mAbs, which like10D5 binds epitopes exposed in Aβ parenchymal plaques and VAβ, but not mAbs, which like m266 bind the Aβ central domain and do not bind to deposited Aβ effect increased hemorrhage incidence [
64,
65]. However mAbs binding deposited Aβ are generally more effective in clearing VAβ [
12]. Time course analysis of VAβ-related events during immunotherapy in PDAPP mice, revealed that a time window for increased incidence of microhemorrhages can last up to six months, but then their incidence diminishes drastically, which correlates with achieving significant clearance of both CAA and CCA [
21]. Thus, the risk of brain microhemorrhages related to Aβ immunotherapy may be a transient complication and possibly reduced in individuals with modest VAβ burden, who also have lower prevalence of spontaneous microbleeds [
58,
59].