The role of APOE in transgenic mouse models of AD

Abstract

Identified in 1993, APOE4 is the greatest genetic risk factor for Alzheimer’s disease (AD), increasing risk up to 15-fold compared to the common variant APOE3. Since the mid 1990’s, transgenic (Tg) mice have been developed to model AD pathology and progression, primarily via expression of the familial AD (FAD) mutations in the presence of mouse-APOE (m-APOE). APOE4, associated with enhanced amyloid-β (Aβ) accumulation, has rarely been the focus in designing FAD-Tg mouse models. Initially, FAD-Tg mice were crossed with human (h)-APOE driven by heterologous promoters to identify an APOE genotype-specific AD phenotype. These models were later supplemented with FAD-Tg mice crossed with APOE-knockouts (APOE^−/− or APOE-KO) and h-APOE-targeted replacement (h-APOE-TR) mice, originally generated to study the role of APOE genotype in peripheral lipid metabolism and atherosclerotic lesion development. Herein, we compare the m- and h-APOE multi-gene clusters, and then critically review the relevant history and approaches to developing a Tg mouse model to characterize APOE-dependent AD pathology, in combination with genetic (sex, age) and modifiable (e.g., inflammation, obesity) risk factors. Finally, we present recent data from the EFAD mice, which express 5xFAD mutations with the expression of the human apoE isoforms (E2FAD, E3FAD E4FAD). This includes a study of 6- and 18-month-old male and female E3FAD and E4FAD, a comparison that enables examination of the interaction among the main AD risk factors: age, APOE genotype and sex. While no single transgenic mouse can capture the effects of all modifiable and genetic risk factors, going forward, a conscious effort needs to be made to include the factors that most significantly modulate AD pathology.

Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder with no cure and few palliative treatments [1]. Since the mid 90’s, transgenic (Tg) mice have been developed to model the phenotypic pathology and progression of AD, including amyloid plaques (amyloid-β peptide/Aβ), intracellular neurofibrillary tangles (tau; NFT), neuroinflammation, neuritic dystrophy, frank neuronal loss, and learning/memory deficits (reviewed in [75, 76]. Although it accounts for only 2-3% of AD cases, familial AD (FAD) is the result of autosomal dominant mutations in amyloid precursor protein (APP), presenilin-1 (PS1) or presenilin-2 (PS2) that result in an increase in Aβ42 or the ratio of Aβ42/Aβ40 [82]. As wild type (wt) mouse APP/PS1/PS2 expression does not lead to plaque deposition, FAD mutations in the human genes are commonly used to induce AD pathology in Tg mouse models (FAD-Tg mice) (for review, [76]). As summarized in Figure 1, the first generation of FAD-Tg mouse models were developed to understand the progression of Aβ pathology, including either single (Fig. 1: #2,3) [21, 25] or multiple FAD mutations (Fig. 1: #7,9,10,11,12) [29, 52, 55, 56, 62]. Mutations in the MAPT gene expressing microtubule-associated protein were later combined with the FAD mutations to induce tau pathology (Fig. 1: #10) [56]. In 1993, APOE4 was identified as the greatest genetic risk factor for AD and found to be associated with accelerated Aβ accumulation [16], both as amyloid and soluble oligomeric forms of Aβ (oAβ), the latter considered a proximal neurotoxin [15, 20, 23, 31, 32, 36, 64, 65, 80]. Despite the 5- to 15-fold increased risk in APOE4 carriers, many FAD-Tg mice fail to model the effects of human apoE isoforms. Specifically, FAD-Tg models continue to use murine (m)-APOE instead of human (h)-APOE or limit the spatial and temporal expression of h-APOE via heterologous promoters (Fig. 1: #6,8) Consequently, FAD-Tg models fail to capture the scope of the APOE genotype-specific effects, particularly on Aβ deposition and neuroinflammation. In addition, female sex profoundly increases the APOE4 related risk for AD, though the nature of this synergistic risk remains unclear [6, 19]. Thus, an ideal Tg model for AD would exhibit the relevant spatial and temporal progression of AD pathology, mirroring the APOE4- and female sex-based risk in humans (reviewed in [75]). Herein we critically review the relevant history and approaches to APOE-dependent AD pathology in Tg-mouse models.

Figure 1. Timeline of development of FAD- and APOE-transgenic mouse models.

Example of FAD-, APOE- and crossed-transgenic mouse models developed over time (1992-2018). Boxes are numbered in chronological order for reference in text. Boxes are labeled: line 1- common name of transgenic model; line 2- genotype of mouse; line 3-promotor. Attributions: last name of first author, common abbreviation of journal, year of publication. NOTE: This timeline is not comprehensive but illustrative.

Murine vs human APOE: APOE^−/− mice

APOE encodes the protein apolipoprotein E (apoE), the primary apolipoprotein expressed in the brain. While apolipoprotein J and apolipoprotein D are also expressed in the brain and are components of lipoproteins, these apolipoproteins do not support the synthesis or survival of a lipoprotein [17, 34]. In humans, apoE has 3 naturally occurring isoforms that differ by a single amino acid change at residues 112 and 158 (apoE2^Cys,Cys, apoE3^Cys,Arg, apoE4^Arg,Arg). In contrast, m-apoE is expressed as a single isoform and differs from h-apoE at the protein level by ~100/300 amino acids. Based on the pioneering work of John Taylor and colleagues [4, 5, 18, 37, 57], h-APOE is now understood to be part of a ~58Kb multi-gene cluster on chromosome 19 that contains TOMM40/APOE/APOCI/APOCI’/APOCCIV/APOCII, while m-APOE is part of a ~37Kb gene cluster on chromosome 7 that contains TOMM40/APOE/APOCI/APOCCIV/APOCII [24, 83] (Fig. 2). Thus, APOE transcription is regulated by transcription factors in the promotor located up to −1000bp from the APOE transcription start site (TSS; reviewed in [47, 48]), as well as the interaction between these elements and far distal enhancers [3, 12, 58, 68-71]. These distal elements include the hepatic control regions (HCR) and multienhancer regions (ME) that contain, for example, liver X receptor (LXR) response elements (LXRE) that dictate cell-specific expression ([35, 67]; reviewed in [86]). Indeed, one of the unique aspects of APOE transcriptional regulation is that in the absence of distal enhancers, the promotor cannot direct gene transcription in vitro [35, 46, 79] or in vivo [3, 22, 35, 67, 71]. Recent concern has focused on the difference in the transcriptional regulation of the m- and h-APOE genes [47, 48], driven, in part, by the growing popularity of FAD-Tg mice with h-APOE, where h-APOE expression is driven by the m-APOE promotor. Of particular concern is the role of nuclear transcription factors in the expression of APOE and thus the number of, for example, LXRE consensus sequences in m- and h-APOE [13, 49, 86]. While these are important points to consider, a comprehensive review of this cluster regulation is beyond the scope of this review.

Figure 2. TOMM40/APOE/APOCI/APOCI’/APOCIV/APOCII multi-gene cluster.

Schematic representation of the multi-gene clusters containing the h-APOE and m-APOE genes. The size of the clusters are defined by the 5’ transcription start site (TSS) of TOMM40 and 3’ end of the coding sequence of APOCII. The 5’ promotor regions and distal regulatory elements, including the hepatic control regions (HCR) and multi enhancers (ME), are identified. The +1 TSS of the h- and m-APOE genes are the basis for the location of other genes in the cluster. The locations of the HCR and ME are based on consensus sequences from www.ensembl.org and www.ncbi.nlm.nih.gov/gene. The exons and introns of the h- and m-APOE gene are from www.ncbi.nlm.nih.gov/IEB/Research/Acembly/index.html. All sizes are approximate, scaled based on gene sequence. In general, this figure is designed to be illustrative and not comprehensive.

Initially, APOE-Tg mice, including APOE-knockouts (APOE^−/− or APOE-KO) (Fig. 1: #1) and APOE-targeted replacement (APOE-TR) (Fig. 1: #4) mice, were generated primarily to understand the role of apoE in peripheral lipid metabolism and atherosclerotic lesion development [73, 87]. Many FAD-Tg models continue to use murine (m)-APOE instead of human (h)-APOE or limit the spatial and temporal expression of h-APOE via heterologous promoters. However, APOE^−/− and APOE-TR mice are being exploited in FAD-Tg models to study the role of apoE in the CNS (for review, [76]). Genetic deletion of m-APOE in FAD-Tg mice delays Aβ accumulation and amyloid deposition, as first demonstrated in the landmark paper with PDAPP^+/+ APOE^−/− mice (Fig. 1: #5) [8], and replicated using other FAD-Tg mouse models. Introducing h-APOE into FAD-Tg/m-APOE^−/− mice further delays Aβ accumulation/amyloid deposition, with Aβ pathology eventually appearing in the sequence of h-apoE isoforms: apoE4 > apoE3 > apoE2. This temporal delay in pathology can be considerable, as with the PDAPP/APOE-TR mice where plaque deposition is delayed from 6- to 18-months (M) (Fig. 1: #13) [7]. These results underscore functional differences between m- and h-APOE on Aβ accumulation. Similar to amyloid deposition, apoE functioning varies in m-apoE vs. h-apoE in most phenotypic AD processes, including Aβ clearance, neuroinflammation and synaptic integrity (reviewed in [75]). Thus, in this comparison, the APOE^−/− mouse is functionally similar to mouse models with h-APOE compared to m-APOE; thus, FAD-Tg mouse models expressing endogenous m-APOE are less relevant and predictive of human AD pathology.

Role of apoE in Tau pathology

As reviewed by Rebeck and colleagues, the deposition of Aβ is apoE isoform-specific, while the accumulation and progression of tau pathology is not apoE isoform-specific [63]. While Aβ plaques found in FAD-Tg mouse brains are structurally similar to those found in human AD brains, these mice do not develop NFTs composed of tau aggregates comparable to those found in human AD brains. Therefore, the role of microtubule associated protein tau (MAPT) has been well studied in frontal temporal dementia (FTD) and other variants of frontotemporal lobar degeneration (e.g., progressive supranuclear palsy) [27, 60, 72]. In FTD-Tg mouse models with mutations in human tau (e.g., Tau P301L, Tau P301S), NFTs form by 3-6M and may be accompanied by motor deficits [77, 84]. The MAPT mutations used to induce NFTs and tauopathy in mice do not increase AD risk but are associated with the rare FTD’s in humans [30]. Recently, in Tau_P301S × APOE-TR mice (Fig. 1: #18), it was demonstrated that in the presence of apoE4, tau levels increased with apoE4 > apoE3 = apoE2 [66]. In contrast, adeno-associated viral (AAV) injections of human tau_P301L in neonatal APOE-TR mice show exacerbated tau pathology and behavioral deficits in the presence of apoE2 but not apoE4 [88]. Although inconsistent, the exacerbation of tau pathology by h-APOE4 in the Tau_P301S mouse is likely a response to the neurodegeneration and neuroinflammation induced by the mutant tau, not the product of a direct interaction between apoE4 with tau. Indeed, apoE4 likely acts as a “second hit” in cases where the CNS is already compromised by the “first hit” from causes that include TBI, neurodegeneration and even recovery from anesthesia [44]. Further evidence to consider in regard to apoE/tau interactions includes: 1. ApoE is not likely to influence tau pathology via a direct interaction under normal conditions in vivo as tau and apoE are not present in the same intracellular compartment. As a classic secretory protein, exocytosis of nascent apoE does not predict an interaction with tau, primarily a microtubule-associated protein found in the cytosol. 2. FAD-Tg mice do not mirror the temporal and pathological progression of tau pathology. Indeed, the only evidence for an apoE isoform-specific effect on tau hyperphosphorylation (apoE4 > apoE3) is in NSE-APOE mice (Fig. 1: #8), with relevance compromised by the use of a heterologous promoter that targets APOE expression to neurons, while apoE is primarily secreted by glial cells (for review, [76]). Regardless, these mice do not develop NFTs [61]. 3. Incorporating tauopathy in AD mouse models may not mirror human AD as MAPT mutations used to induce tauopathy in mice do not increase AD risk but are associated with the rare FTD’s in humans, as discussed above [30]. In conclusion, the utility of incorporating tau mutations in AD-Tg mouse models to determine if apoE exerts an isoform-specific effect on the development of tau pathology requires a careful interpretation and may have a limited relevance to AD as mutant tau causes FTD.

Temporal regulation of apoE expression (AAV delivery vs. Inducible models)

Crossing APOE-TR mice that express h-apoE isoforms with FAD-Tg mice results in apoE expression at physiological levels under the control of the m-apoE regulatory sequences. However, age-dependent mechanisms and critical periods for apoE isoform-specific modulation of AD pathology remain unclear. To address the stage of amyloid deposition where apoE4 has its greatest effect, both AAV-mediated gene delivery and Cre-mediated mouse models can be activated at different stages of pathology. AAV4-mediates overexpression of h-apoE in ependymal cells and astrocytes in the subventricular zone of FAD-Tg mouse brains via the ependymal system, while AAV8 driven by GFAP promotor leads to overexpression of h-apoE in all astrocytes of APOE-TR mice [26, 38]. Interestingly, in both studies Aβ-associated AD pathology was highest in apoE4-expressing mice. However, AAV-mediated CNS gene transfer in mice is not fully understood due to limited characterization of biodistribution profiles and immune responses with certain AAV serotypes. This limits the interpretation of these and future findings that use AAV-mediated gene delivery to study the effect of apoE expression on amyloid pathology at various stages. Crossing mice with the Cre transgene driven by the GFAP promotor with inducible apoE3 or apoE4 in a FAD-Tg mice provides astrocyte-specific expression (Fig. 1: #19). Thus, expression of apoE3 or apoE4 can be activated at different stages of AD pathology based on the Cre expression [41]. Alternatively, anti-sense oligos (ASOs) that target APOE3 or APOE4 mRNA expression in FAD-Tg mice lower apoE expression, again either before or after significant amyloid deposition [28]. In both studies, expression of apoE4 increased Aβ-associated AD pathology in the brain when expressed early in the development of AD pathology but not in the later stages. Thus, differential modulation of Aβ aggregation highlights the critical interaction between two AD risk factors: age and APOE.

Recent applications of APOE-Tg mouse models

In addition to the EFAD mice, a number of novel FAD/APOE-Tg models have been developed (Fig. 1). Other publications have relied on a comparison between FAD/m-APOE^+/+ and FAD/m-APOE^−/− to identify the function of “apoE,” actually m-apoE. The latter is concerning as this comparison, initially made by Bales in 1997 as discussed above, does not address the differential isoform-specific functions of h-APOE (for example [81]). Similarly, using a FAD/APOE-Tg mouse with mixed apoE4/m-apoE, differential effects on cerebral amyloid angiopathy (CAA) and parenchymal plaques were observed, suggesting opposing functions for m-apoE versus h-apoE4 (Fig. 1: #16) [40]. The question remains: What can be learned by comparing m-apoE to h-apoE4 that will inform the difference between the human apoE isoforms. With the FAD/APOE-Tg models, APOE4 has been linked to enhanced Aβ-induced deficits in insulin signaling [11] and support the use of an anti-h-apoE4 antibody to reduce apoE aggregation and plaque formation [39] [Fig. 1: #17]. Data from Koldamova and colleagues demonstrate the interactive effects of diet, APOE and sex on the immune response (Fig. 1: #15) [54]. Compared to normal diet, a high fat diet (HFD) increased amyloid deposition in ♂ and ♀ APOE4 carrier mice vs. ♂ and ♀ APOE3 carrier mice [53]. In addition, the HFD ♀APOE4 carriers also had the greatest microglial coverage surrounding plaques, thus establishing a link among diet, APOE, sex and the neuroimmune response [54].

EFAD mice: effects of modifiable and genetic risk factors on AD pathology

The EFAD-Tg mouse model (carriers: 5xFAD^+/−/APOE^+/+; non-carriers: 5xFAD^−/−APOE⁺⁺) was designed as a tractable preclinical FAD-Tg-mouse model for identifying the mechanisms underlying apoE isoform-specific modulation of AD pathology (Fig. 1: #14) [85]. From their first description in 2012, EFAD mice have enabled critical basic and therapeutic research [2, 9, 10, 14, 33, 42, 43, 51, 75, 89]. Recent publications have expanded the characterization of this model to include measures of cerebrovascular integrity and inflammation [50, 78]. In LPS-challenged ♂E4FAD carrier mice, peripheral inflammation increased cognitive decline and decreased post-synaptic protein levels [50]. Furthermore, chronic peripheral inflammation resulted in cerebrovascular deficits, suggesting that the interaction among APOE4, Aβ, and peripheral inflammation induced cerebrovascular damage and cognitive decline. Recent developments with the EFAD-Tg mouse presented in this review including the moderation effect of modifiable (e.g., inflammation, obesity) and genetic (e.g., sex, age) risk factors on AD pathology. Epidemiological evidence suggests that vascular risk and lifestyle factors (e.g., obesity, Western diet) increase AD risk in humans with increasing attention on the interplay of APOE genotype and sex [59]. In ♂EFAD carrier mice on a control diet, metabolic profiles for E3FAD = E4FAD, while the Western diet was associated with Aβ-associated pathology with E4FAD > EFAD3 [51]. This relationship differs by sex as recent findings show that in ♀EFAD carrier mice, Western diet yields E3FAD > E4FAD for select measures of both metabolic dysfunction and Aβ accumulation [14]. These results demonstrate a significant interaction between diet, APOE genotype and sex on AD pathology and an important gene-environment interaction [51].

Aged male and female EFAD mice

As an update on the EFAD mice, a recently completed study of an 18M cohort of ♀ and ♂ E3FAD and E4FAD mice was compared to 6M mice (n=8-12). The 6M male EFAD mouse data was adapted from [75]. All protocols follow the UIC Institutional Animal Care and Use Committee protocols. Breeding and colony maintenance was conducted at UIC. All the biochemical and immunohistochemical methods have been published previously [85]. Kaplan Meier survival curves for only the 18M cohort revealed a significance difference in survival: ♀E4FAD < (♂E4FAD ~ ♀E3FAD < (♂E3FAD (Fig. 3). Results from MWM in the 18M cohort demonstrated a significant memory loss in the ♀E4FAD compared to (♂E4FAD, ♀E3FAD, and (♂E3FAD (data not shown). The apoE and Aβ extraction profiles, using a 3-step sequential extraction with TBS, TBSX (TBS + 0.1 % TritonX) and FA (formic acid) were compared between 6M and 18M (Fig. 4 & 5). For apoE, total levels increased significantly from 6 to 18M in the 4 groups but with no difference between genotype and sex (Fig. 4A). However, in the TBSX fraction, the fraction where lipoprotein-associated apoE elutes, there was no age-induced increase in the ♀E4FAD mice, and ♀E4FAD < (♂E4FAD mice, and while there was an age-induced increase in the ♀E3FAD, ♀E3FAD < ♂E3FAD (Fig. 4B). Total Aβ increased from 6 to 18M in the 4 groups, with ♀E3FAD > ♂E3FAD and ♀E4FAD > (♂E4FAD (Fig. 5A). Importantly, soluble oAβ exhibited a similar increase from 6 to 18M, with the response at 18M: ♀E3FAD > ♂E3FAD and ♀E4FAD > (♂E4FAD (Fig 5B). There was no difference in the soluble Aβ42 levels at 18M among the 4 groups (Fig 5C). Immunohistochemistry for astrogliosis (GFAP) and Aβ deposition (Aβ mAb MOAB2) (Fig. 6A) revealed that at 18M, astrogliosis in the ♀E4FAD was > 6M, > (♂E4FAD and > ♀E3FAD, thus demonstrating significant age, sex and genotype effects (Fig. 6B). Aβ deposition in the cortex at 18M was also greatest in ♀E4FAD mice (Fig. 6C). These results indicate that measures of AD pathology in the EFAD mice develop with age, with 18M the calculated 75% survival rate for mice on their genetic background (https://www.jax.org/research-and-faculty/research-labs/the-harrison-lab/gerontology/life-span-as-a-biomarker), confirmed by the Kaplan Meier survival curve (Fig. 3). Both biochemical and immunohistochemical measures of AD pathology demonstrate significant interactive effects among age, sex and APOE. Aging studies with these mice are ongoing.

Figure 3. Kaplan Meier survival curves for 18M EFAD mice.

Survival rates of 18M ♀E4FAD, (♂E4FAD, ♀E3FAD, and ♂E3FAD plotted as % survival. n = 8-12 mice per cohort.

Figure 4. ApoE levels in 6M and 18M EFAD mice.

A) Total apoE levels; B) extraction profiles of apoE using a three-step sequential protein extraction (TBS, TBSX and FA) in the cortex of 6 and 18M EFAD mice measured by ELISA. Data expressed as mean ± S.E.M. and analyzed by two-way ANOVA followed by Bonferroni’s multiple comparison post hoc analysis. n=8-12, § p < 0.05 vs. 6M within sex and genotype (age effect); * p < 0.05 vs. males within genotype (sex effect); # p < 0.05 vs. E3FAD within sex (APOE genotype effect).

Figure 5. Aβ levels in 6M and 18M EFAD mice.

Levels of: A) total Aβ42; B) soluble oligomeric Aβ; and C) soluble Aβ42 in the cortex of 6-and 18M EFAD mice measured by ELISA. Data expressed as mean ± S.E.M. and analyzed by two-way ANOVA followed by Bonferroni’s multiple comparison post hoc analysis. n=8-12, § p < 0.05 vs. 6M within sex and genotype (age effect); * p < 0.05 vs. males within genotype (sex effect); # p < 0.05 vs. E3FAD within sex (APOE genotype effect).

Figure 6. Aβ-associated neuroinflammation in 6M and 18M EFAD mice.

A) Representative images of sagittal brain sections from 6- and 18M EFAD mice immunostained for astrogliosis (GFAP, green) and Aβ (MOAB2, red), x10 magnification (scale bar = 500 μm). Quantification of % area in cortex and hippocampus + subiculum covered by: B) astrocytes; C) Aβ deposition. Data expressed as mean ± S.E.M. and analyzed by two-way ANOVA followed by Bonferroni’s multiple comparison post hoc analysis. n=8-12, § p < 0.05 vs. 6M within sex and genotype (age effect); * p < 0.05 vs. males within genotype (sex effect); # p < 0.05 vs. E3FAD within sex (APOE genotype effect).

Conclusion

FAD-Tg mouse models are evolving in an effort to replicate the diverse pathology present in sporadic AD. However, information gleaned from these models is limited if basic risks that include age, APOE genotype, and sex are not considered. This is particularly critical for preclinical testing as the Tg model should mirror the human population, which includes, for example, APOE4 females. Furthermore, characterization of these effects should not ignore the potential contributions of peripheral phenotypes (e.g., Cre, AAV). In addition, the physiological relevance of certain models is still understudied and controversial. For example, while the use of inducible APOE-Tg models is a promising approach, it is still understudied. Comparisons between models with m-APOE and APOE^−/− are not mechanistically informative or predictive of future therapeutic approaches as a case study of an APOE^−/− human demonstrated deficits in learning, memory, language, visuospatial abilities, and executive functions as well as extensive xanthomas and a lipidemic plasma lipoprotein profile [45]. Further, m-apoE is not equivalent to the human apoE isoforms, thus requiring the introduction of h-APOE. In the end, no single transgenic mouse can capture the effects of all modifiable and genetic risk factors. Going forward, a conscious effort needs to be made to include the factors that most significantly modulate AD pathology. This approach will help to: 1) allow full interpretation of findings; and 2) validate a mouse model for testing therapeutics that translate to clinical trials.

Highlights:

• APOE transgenic mouse models of Alzheimer’s disease have evolved overtime to include various methods for deleting or introducing the expression of apolipoprotein E.

• Limitations of current APOE transgenic models include comparison with FAD/m-APOE with FAD/APOE^−/− mice to determine the effect of APOE on AD pathology.

• Few recent models are phenotypically characterized by APOE isoform with genetic (sex, age) and modifiable (e.g., inflammation, obesity) Alzheimer’s disease risk factors.

• New aging data presented that further characterizes the EFAD mice (5xFAD^+/−/APOE3^+/+ or 5xFAD^+/−/APOE4^+/+) supports the interactive and synergistic effects of APOE genotype with sex on FAD-induced pathology.