APOE4 enhances age-dependent decline in cognitive function by down-regulating an NMDA receptor pathway in EFAD-Tg mice

Alzheimer’s disease (AD) is a progressive neurodegenerative disease that causes loss of memory and cognitive function, and is the most common cause of dementia in individuals over the age of 60. APOE4, the greatest genetic risk factor for sporadic Alzheimer’s disease (AD), increases risk ~3- and 15-fold with a single or double allele [1-11] compared to APOE3, whereas APOE2 decreases AD risk ~2-fold per allele [12-16]. The multifactorial mechanisms through which apolipoprotein E (apoE) affects AD risk ultimately converge on modulation of cognitive function. As well, the amyloid-β peptide (Aβ) [17-19], the proposed proximal neurotoxin in AD, is a major cause of impaired synaptic function, particularly soluble oligomeric forms of the peptide (oAβ) [20-23]. However, how human (h)-apoE interacts with Aβ to affect cognitive function, and the potential underlying neuronal signaling pathways, remains unclear, in part due to the lack of a tractable familial AD (FAD)-Tg mouse model. In addition, debate continues on whether apoE4 represents an overall loss of positive function or gain of toxic function, a distinction that significantly impacts therapeutic approaches for targeting not only APOE4-induced AD risk, but for the effects on all h-APOE genotypes.

In AD patients, APOE4 is associated with an earlier age of onset for cognitive deficits than APOE3 [6-11], and possibly a faster rate of cognitive decline [24,25], though results are conflicting regarding the latter. However, even in the absence of AD, older APOE4 carriers (60+ years of age) exhibit deficits in episodic memory and higher rates of cognitive decline compared to APOE3 carriers [26-29]. Although these data demonstrate greater apoE4-induced cognitive impairment compared to apoE3, it remains unclear whether this is a loss of positive function or gain of toxic function. This issue is highlighted by a recent case report of a 40-year-old male patient with an ablative frame shift mutation that results in a complete lack of apoE [30]. The patient is described as cognitively normal on gross functional tests (MMSE), raising the hypothesis that all the h-APOE genotypes are either a gain of toxic function, or are not required for cognitive function. However, sub-domain tests indicate deficits in memory, language, visual-spatial abilities and executive function, in addition to signs of dyslexia [30], supporting the loss of function hypothesis. Data from Tg mouse models on the role of apoE on cognitive decline are primarily derived from models that express h-apoE, but without h-Aβ pathology. As with non-AD patients, in APOE-TR mice, apoE4 is associated with cognitive deficits in both young (Morris water maze, Barnes maze) [31,32] and older mice (Morris water maze, Y-maze) [33,34]. Similar data were also observed in mice expressing h-apoE under the control of heterologous promoters (reviewed [35]). In FAD-Tg mouse models expressing h-apoE under the control of the NSE promoter, behavioral performances (water maze) follow the pattern apoE3 > apoE4 = apoE-knockout (KO), consistent with a loss of positive function for apoE4 [36]. However, as apoE is physiologically expressed by glia, the relevance of these data is unclear.

At the synaptic level, AD patients exhibit decreased levels of postsynaptic intracellular scaffold proteins, including postsynaptic density protein 95 (PSD95) and drebrin, suggesting post-synaptic disruption precedes loss of pre-synaptic proteins to initiate the cognitive deficits characteristic of the disease (reviewed in [37-39]). Importantly, decreased levels of PSD95 and drebrin can lead to decreased expression of N-methyl-D-aspartate receptor (NMDAR) subunits (N1, NR2A and NR2B) [37,38]. Clinically, in vivo and in vitro evidence indicate that AD, Aβ, inflammation and chronic vasculitis can result in chronic NMDAR activation, disrupting postsynaptic ionic gradients, long-term potentiation (LTP) and cognition [37-39]. Further, lower NDMA receptor levels may result in a decreased Ca²⁺-dependent activation of the calcium-calmodulin-II (CaMK-II)/cAMP response binding element peptide (CREB) pathway, leading to decreased production of the brain derived neurotropic factor (BDNF), critical for synaptic function and for increasing NMDAR levels via positive feedback [39-43]. Mechanistically, an apoE4-induced reduction in post-synaptic proteins may disrupt CaMK-II/CREB/BDNF signaling to impair cognitive function [44]. Similar effects are observed in long-term primary neuron-glia co-cultures, as apoE4 accelerates the loss of GluN1 levels and mature spines compared to apoE3 [45]. Further, by inducing intracellular sequestration, apoE4 reduces neuronal cell-surface expression of NMDA receptors in vitro [46]. However, little is known about the APOE genotype-specific effects on these processes in combination with AD pathology.

To assess whether apoE4 imparts a loss or gain of function requires a comparison to the absence of apoE (APOE-KO), not simply a comparison to apoE2/apoE3. For example, in vivo studies demonstrate that with LPS-induced inflammation and amyloid deposition, apoE4 is anti-inflammatory [47] and anti-amyloidigenic [35,48] compared to apoE-KO, though apoE3 is better than apoE4. In other data more directly related to synaptic dysfunction, no differences were observed between apoE4 and apoE-KO in measures including spine density and LTP [49,50], with apoE3 higher than both. Finally, apoE4 exhibits a gain of toxic function compared to apoE-KO for oAβ42-dependent attenuation of LTP [51] and oAβ42-induced neurotoxicity in neuron/glial co-cultures [52]. Thus, it is critical to determine the effect of h-apoE on postsynaptic protein expression and signaling in the EFAD and 5xFAD/APOE-KO mice.

As data indicate that the APOE4-induced risk for AD is significantly greater in females compared to males in both humans and APOE-TR mice [35,53-55], female EFAD-Tg mice [48] were used in this study to identify the effects of Aβ pathology on APOE genotype-specific modulation of behavior. EFAD mice are an AD-Tg mouse model with h-apoE expressed under the regulated control of the endogenous mouse (m)-apoE promoter (APOE-TR) [56] and h-Aβ42 over-expressed via the 5xFAD-Tg mice, an FAD-Tg mouse model [57]. In addition, E4FAD mice were compared to 5xFAD/APOE-KO to address whether apoE4 imparts a loss of positive or gain of toxic function. Finally, E2FAD, E3FAD, E4FAD and 5xFAD/APOE-KO mice at 2-, 4-, and 6-months of age were used as previous data demonstrated significant age-dependent (2-6 months), apoE isoform-specific (apoE4 > apoE3 = apoE2) effects on the development of Aβ pathology in EFAD mice [48,58,59]. Therefore, from a translational perspective, it is important to incorporate sex, APOE genotype, and Aβ pathology in a preclinical model. Using the recently developed, tractable EFAD-Tg mice and 5XFAD/APOE-KO mice, age-dependent changes in spatial recognition memory (Y-maze and Morris water maze), pre-synaptic (synaptophysin) and post-synaptic (PSD95 and drebrin) protein levels, and the NMDAR subunits levels and activation of the CaMK-II-CREB-BDNF pathway were measured in 2-, 4- and 6-month female mice. The results demonstrate a greater age-induced deficit in behavior and reduction in postsynaptic proteins in the E4FAD and 5XFAD/APOE-KO mice compared with E2FAD and E3FAD mice, consistent with an apoE4 loss of function. However, further results demonstrate that while phosphorylated CaMK-II (p-CaMK-II) followed the same apoE-specific pattern as cognition and synaptic protein levels, levels of phosphorylated CREB (p-CREB) and BDNF demonstrate an apoE4 toxic gain of function.

APOE4-induced AD risk is likely the result of multiple, overlapping mechanisms, both Aβ-dependent and Aβ independent (for review [74]). One challenge in understanding the effect of APOE genotype on various mechanistic readouts is determining whether apoE4 represents a loss of positive function or a gain of toxic function. Thus, we investigated the early, age-dependent APOE genotype-specific effects on cognitive functions and synaptic viability in EFAD-Tg mice [48,58,59], specifically female mice based on data in both humans [53-55] and Tg mice [33,35,36,75,76] that APOE4 females exhibit significantly increased cognitive impairment compared to APOE4 males and APOE3 females. In the Y-maze, a significant age-dependent decline in spatial recognition memory was observed only for the E4FAD mice from 2-4 months, indicating a more rapid decline at earlier stages of Aβ deposition compared to other genotypes (Figure 1). In the MWM, a measure of spatial learning and memory, the E4FAD and 5xFAD/APOE-KO mice were both slower to learn than the E2FAD and E3FAD mice during the 5-day training phase (Figure 2B). In addition, both the E4FAD and 5xFAD/APOE-KO mice exhibited an age-related increase in escape latency from 2-6 months during the training trials, suggesting the failure of some compensatory effect over time. As this is consistent with previous studies demonstrate higher anxiety levels in APOE4-TR and APOE-KO mice [76,77], we hypothesize that this elevated stress response may facilitate spatial learning in young E4FAD mice and mask adverse effects of apoE4 on spatial cognition. Indeed, it has been shown that normal aging can counteract stress-induced facilitation of cognitive processing in APOE4-TR mice, as measured by MWM, making phenotypic differences easier to detect in older mice [33,75]. This apoE4 effect in the EFAD mice is amplified by the overproduction of Aβ42 driven by the presence of the 5-FAD mutations. Indeed, the 5xFAD mice show progressive learning and memory deficits tasks as early as 3 months [78-82]. As deficits in spatial learning and memory due to apoE4 have mainly been reported in older and non-AD mice [33,75], our findings are consistent with synergistic effects between apoE4 and the aggressive Aβ42 pathology characteristic of the EFAD mice [48]. In addition, the use of only female EFAD mice also optimized the risk of cognitive deficits in the E4FAD mice. Indeed, sex interacts with APOE to affect cognitive function. Clinical data indicate that the APOE4-induced risk for AD is significantly greater, perhaps exclusive to, females [53-55]. These data are consistent with the greater cognitive impairment in female APOE4-TR mice compared to female APOE3-TR mice, and with both APOE3- and APOE4-TR females compared to APOE genotype-matched males (review [35]). Overall, as measured in this study, behavior appeared to be primarily an apoE4 loss of function, specifically: E2FAD = E3FAD > E4FAD ≈ 5xFAD/APOE-KO. However, further studies in humans and Tg-mouse models are critical to determine the role of potential interactive effects among Aβ pathology, APOE genotype and sex on memory and cognitive decline.

Spatial and learning memory performances are directly linked to synaptic function. ApoE4 is associated with progressive synaptic deficits in both AD patients and h-APOE-Tg mouse models [83-87]. Consistent with previous reports, age-induced reductions in synaptic proteins preferentially occurred in post-synaptic proteins compared particularly to synaptophysin, a pre-synaptic protein (Figure 4) [81,88-90]. In the current model, levels of drebrin and PSD95 were lower in E4FAD and 5xFAD/APOE-KO mice compared to E2FAD and E3FAD mice, consistent with an apoE4 loss of function. ApoE is also linked to long-term potentiation (LTP) and NMDAR-mediated signaling [50,51]. As the NMDAR component of synaptic transmission has been shown to decline during aging [91,92], NMDAR activation may provide a mechanistic pathway for understanding apoE-related memory impairment. Indeed, in this study apoE-related cognitive impairment correlates with a decrease in the levels of NMDR subunits and components of the signaling pathway (p-CaMK-II/p-CREB/BDNF). Levels of all three NMDAR subunits were reduced with age for all the genotypes with reductions greater in the E4FAD and 5xFAD/APOE-KO mice compared to E2FAD and E3FAD mice, consistent with an apoE4 loss of function (Figure 5). These apoE4-related deficits in NMDAR-dependent functions likely reflect changes to neuronal networks contributing to short and long-term memory, and their contribution to memory consolidation [93-96]. Reduction of BDNF levels through either genetic or pharmacological means not only impaired LTP and reduced the number of synapses, but also caused deficits in the formation and consolidation of memory [97-99]. However, the effects of APOE genotype on the p-CaMK-II, p-CREB, and BDNF signaling cascade are not consistent. Again, all the signaling components were reduced with age for all the genotypes (Figure 6). While the greater reduction in p-CaMK-II in E4FAD and 5xFAD/APOE-KO mice compared to E2FAD and E3FAD mice is consistent with an APOE4 loss of function, apoE4 represents a gain of toxic function for both the activated p-CREB transcription factor and its downstream target protein BDNF, as the reductions in p-CREB and BDNF levels are greater in E4FAD compared to 5xFAD/APOE-KO mice (summarized in Figure 7, right side). This gain of toxic function does not translate directly to the behavioral measures used in this study. One likely possibility for this seeming disconnect is that a number of signaling pathways are activated by changes in synaptic proteins, particularly the NMDAR, that eventually converge to modulate behavioral readouts. Thus, the final effect on behavior will be a composite of a number of signaling pathways (reviewed in [100]). Taken together, our findings support the notion that apoE4 exacerbates behavioral deficits in EFAD mice by decreasing synapse viability by reducing synapse-related proteins, particularly via down-regulation of NMDAR and NMDAR-mediated signaling via CaMK-II, CREB, and BDNF (Figure 7). Consistent with the APOE2 protective effect for AD risk [1-11], these results also demonstrate that E2FAD mice are consistently less susceptible to age-induced changes in the components of this cascade, from signaling to behavior.

ApoE is the primary ligand for the low-density lipoprotein (LDL) receptor (LDLR) family (apoE receptors), although Reelin is the primary ligand for ApoE-receptor 2 (ApoER2). ApoER2 and Reelin are important modulators of synaptic plasticity and NMDAR functions in vitro and in vivo [106-108]. Thus, the association between ApoER2, Reelin and NMDAR are critical for LTP, memory formation and retrieval. ApoE4 has been demonstrated to reduce the cell-surface levels of both ApoER2 and NMDAR via intracellular sequestration, thus inhibiting the ability of Reelin to facilitate glutamate-mediated synaptic plasticity [46]. The impaired recycling of apoE4 may contribute to this reduction in receptors at the cell surface [105]. Loss of ApoER2 reduces Reelin binding, thus further reducing activation of NMDAR via signaling by the Src family kinases [46,105].

While a number of Aβ-independent mechanisms likely contribute to the APOE-associated risk for AD [109], oAβ has been demonstrated to be preferentially synapotoxic (for review [37,110]). We have published the effects of APOE genotype on Aβ accumulation in the EFAD mice at 6-months of age, the age of the mice used for this study [48,58,59,111,112]. These results demonstrate amyloid deposition by IHC and total brain Aβ42 by ELISA is: 5xFAD > E4FAD > E3FAD = E2FAD. A three-step sequential protein extraction protocol using TBS (soluble), TBS + Triton X-100 (TBSX, detergent), and formic acid (FA, insoluble) was used for the hippocampus and cortex. In the soluble fractions of both brain regions, both Aβ42 and oAβ are: E4FAD > E3FAD = E2FAD. There is no APOE genotype difference in the levels of Aβ42 in the detergent fraction. In the insoluble fraction, Aβ42 is: E4FAD > E3FAD = E2FAD. As the EFAD mice are on the 5xFAD background, the amount of Aβ40 is difficult to detect; the primary species is Aβ42. Thus, Aβ levels (amyloid, soluble and insoluble) are greatest in the E4FAD mice. This association between APOE and Aβ accumulation is consistent with the functional changes reported herein. Therefore, a particularly relevant approach to interpreting the results of this study is to consider APOE modulation of soluble Aβ levels at the synapse (Figure 7, left side). Previous publications from our group and others demonstrate that apoE isoform-specific effects on Aβ clearance and interactions with apoE receptors likely play a role in this process at several levels. It has been specifically demonstrated that apoE4 both increases the levels of oAβ and directs it to the synapse [5]. ApoE isoforms may modulate oAβ levels through differential apoE/Aβ complex levels [113]. However, as isolation and analysis of the apoE/Aβ complex in vivo is technically challenging, data are conflicting as to the significance or even the existence of this complex [114-116]. Nevertheless, it is interesting to note that the levels of soluble apoE4/Aβ complex are lower than apoE3/Aβ and decrease in AD in human synaptosomes, CSF and EFAD-Tg mouse brains, the reverse of soluble oAβ levels [48,58,59,113]. ApoE receptors also play a key role, particularly ApoER2, as Reelin signaling can prevent the oAβ-induced inhibition of NMDAR at the synapse [117]. As well, neuronal LRP1 provides a significant mechanism for the clearance Aβ [101,102] and in vitro, Aβ clearance is impaired with apoE4, compared to apoE3 [103], consistent with a greater accumulation of intraneuronal Aβ [104]. To connect these Aβ-dependent processes to decreased synaptic function and impaired cognition requires, in part, a return to ApoER2, NMDAR and Reelin. ApoE4 induces deficits in ApoER2 and NMDAR signaling and recycling, as well as reducing Reelin binding to both receptors [46,105]. Ultimately, these apoE-mediated differential effects on apoE receptors and Aβ accumulation could contribute to the mechanisms responsible for synaptic dysfunction and cognitive decline characteristic of observed in the EFAD mice that, ultimately, will translate to AD patients.

Herein we provide evidence that the APOE4 genotype constitutes a loss of positive function contributing to age-related deficits in behavioral performances in the EFAD and 5xFAD/APOE-KO mice (Figure 7-right side; Table 1A). This loss of positive function with apoE4 was related to a decrease in post-synaptic proteins, including NMDAR subunits, leading to impaired NMDAR-related signaling. However, apoE4 represents a gain of toxic function for the final components measured in this pathway, activated p-CREB and its downstream target protein BDNF. It is our interpretation that multiple signaling pathways converge to determine the final synaptic transmission impairment and learning and memory deficits associated with apoE4. This conclusion requires further study to determine the potential contributions of other signaling components to either an apoE4-mediated loss or gain of function, data of high therapeutic significance. Table 1B provides a representative summary of CNS-relevant functions modulated by apoE4 that can be attributed to a loss of positive or gain of toxic function. Again, it is important to assess whether apoE4 imparts a loss of positive or gain of toxic function in comparison to the absence of apoE (APOE-KO), not simply a comparison to apoE2/apoE3. Table 1B provides a larger perspective on the interplay between among multiple functions that exhibit a loss of positive (for example, anti-inflammatory properties and amyloid deposition) or gain of toxic function (for example, toxic proteolytic fragments of apoE and oAβ-induced neurotoxicity in vitro and ex vivo). Only an understanding of the relative contribution of the functions measured in this paper, listed in Table 1A and the many others as yet undefined will enable a confident identification of APOE4-induced AD risk as a loss of positive or gain of toxic function.

Targeting the most potent genetic risk factor for AD appears a very attractive strategy and is still under intense study. If the hypothesis is that all apoE isoforms, particularly apoE4, represent a toxic gain of function, then reducing APOE expression and/or apoE levels is one therapeutic approach for AD. However, the potential dangers of this approach in the human brain are still subjected to debate [30]. Here, we provide additional insight into the mechanism by which APOE4 increases AD risk, in which apoE4 mainly appears as a loss of positive function. Accordingly, rather than APOE gene inactivation, therapies that correct the loss of positive function related to apoE4, such as increasing the lipidation of apoE4 containing lipoproteins [58] appear to be more appropriate.