A Novel APP Knock-In Mouse Model to Study the Protective Effects of the Icelandic Mutation In Vivo

Alzheimer's disease (AD) is the most common form of dementia among people over 65 years old. AD is characterized by the accumulation of β-amyloid (Aβ) plaques in the brain parenchyma and within cerebral vessels (cerebral amyloid angiopathy), hyperphosphorylated intraneuronal neurofibrillary tangles, and loss of synapses. In more than 90% cases, AD occurs sporadically and late in life. The pathological mechanisms for this late-onset AD are still not completely elucidated, but genetic and environmental factors appear to play a significant role. Aging and the presence of the apolipoprotein E (ApoE) ε4 allele are considered the most common risk factors for the disease, but other conditions, including diabetes, high cholesterol, stroke, obesity, and pre-existent cardiovascular diseases, are also significant contributors (Rostagno et al., 2010). In contrast, familial forms of AD (FAD) comprise less than 5–10% of the total AD cases and typically appear before 65 years of age. FAD cases are characterized by mutations in three key genes involved in the generation of Aβ: presenilin (PSEN) 1 and 2 and amyloid precursor protein (APP). More than 300 mutations in these genes have been reported in AD (Rostagno et al., 2010). The dominant linkage of these genes to AD have helped support the original amyloid cascade hypothesis of AD pathogenesis, which considers parenchymal accumulation of Aβ as the crucial initial event that triggers other pathological events. Recent findings have suggested that soluble oligomers of Aβ, rather than monomers or insoluble amyloid fibrils, may be the key initiators of the pathological cascade of events responsible for synaptic dysfunction in the brains of AD patients (Hardy and Selkoe, 2002).

Aβ is a peptide generated by sequential cleavage of amyloid precursor protein (APP), first by β-site APP-cleaving enzyme (BACE1) and then by γ-secretase enzymes (amyloidogenic pathway). The main Aβ isoforms involved in neurotoxicity are Aβ40 (40 amino acids) and Aβ42 (42 amino acids). Aβ40 is mainly released from endothelial cells and platelets contributing to the vascular Aβ deposits, while Aβ42 is the main form found in neuronal plaques within the brain parenchyma (Herzig et al., 2004). Aβ42 is considered more toxic than Aβ40 being highly insoluble and more aggregation prone. But APP can also be processed by a nonamyloidogenic pathway. Physiologically, APP is cleaved by α-secretase before being further cleaved by γ-secretase, which produces a soluble form of APP (sAPPα) and C-terminal fragments (CTF-α). It is important to mention that APP, sAPPα, and CTF-α have neuroprotective properties. For example, APP regulates normal brain development and neuronal maturation while sAPPα and CTF-α promote reduction of neuronal injury and improvement in memory performance (Hardy and Selkoe, 2002).

Most reported AD-linked APP mutations are clustered within the N-terminal region of the protein which contains the BACE1 cleavage site (exons 16–17). Among the most studied of these mutations, the Swedish double mutation was first described in a Swedish family, and it involves a double mutation (K670N/M671L), in which a G to T substitution in codon 670 results in a switch from lysine to asparagine in the protein and an A to C substitution in codon 671 results in a switch from methionine to leucine. These mutations are adjacent to the BACE1 cleavage site of APP, and they result in elevated production of Aβ peptides (Mullan et al., 1992). Another mutation close to the β-cleavage site is the Italian variant (A673V), in which a C to T transition translates into an alanine-to-valine substitution at position 673 (Aβ position 2). The homozygous presence of this mutation results in enhanced production and aggregation of Aβ40 and Aβ42, while heterozygous individuals are unaffected, likely because the wild-type Aβ interferes with aggregation of the mutated counterpart (Di Fede et al., 2009).

Interestingly, transition of G to A within codon 673 of APP results in a different mutation, the Icelandic variant (A673T), the first APP variant associated with reduced AD risk. This mutation shifts APP processing toward the nonamyloidogenic pathway, resulting in minimal amyloid deposition and protection against age-related cognitive decline (Jonsson et al., 2012). While the protective effects of this mutation have been extensively studied in vitro (Jonsson et al., 2012; Colombo et al., 2017), however, challenges persist in assessing its activity in vivo because most mouse models of AD are generated by introduction of the Swedish mutations, which may interfere with the protective effect of the Icelandic variant due to the induction of APP overexpression. To address this limitation, Shimohama et al. (2024) recently created a novel APP Knock-in (KI) mouse model (APP^G-F-A673T).

The APP^G-F-A673T mouse is based on a previously reported KI model by the same group, the APP^G-F mouse, which harbors the pathogenic Arctic (E693G) and Beyreuther/Iberian (I716F) substitutions, thus avoiding the APP overexpression caused by the Swedish double mutation (Watamura et al., 2022). The expression of the E693G Arctic mutation leads to an increased propensity to form Aβ protofibrils, while the I716F Beyreuther/Iberian substitution enhances neurofibrillary changes and amyloid deposition. Although cognitive changes were not assessed in the APP^G-F mouse model, it exhibited age-dependent Aβ pathology (starting at 6 months) within the cortex and hippocampus, neuroinflammation, activated glia, and reactive astrocytosis (Watamura et al., 2022).

APP^G-F-A673T mice were generated by knocking-in the A673T substitution into APP^G-F mice via CRISP-Cas9 technology. Although Aβ42 levels did not change in the insoluble fraction of APP^G-F-A673T cortices, Aβ40 and Aβ42 levels were reduced in the soluble cortical fraction at 3 months of age. At this early age, APP^G-F-A673T mice showed a decreased CTF-β/CTF-α ratio suggesting a decreased APP susceptibility to BACE1 cleavage before the onset of pathology. No change in the expression level of APP or APP processing enzymes were reported (Shimohama et al., 2024). In older animals, when pathology was present, APP^G-F-A673T mice had fewer cortical and hippocampal plaques (at 8 months of age) and attenuated neuroinflammation with fewer activated astrocytes, glial cells, and dystrophic neurites (at 12 months) compared with age-matched APP^G-F (Shimohama et al., 2024).

In summary, the APP KI AD model mice exhibited protection against AD pathology in vivo consistent with the protective properties of the Icelandic mutation. The mice thus overcome the challenges caused by models incorporating the Swedish double mutation that were unsuitable to study the A673T substitution. Moreover, the results support the hypothesis that the Icelandic mutation causes a shift of APP processing from the amyloidogenic to the nonamyloidogenic pathway, leading to reduced production of toxic Aβ species (Jonsson et al., 2012). Another contributor to the protective effect may be that the substitution of alanine, a nonpolar, hydrophobic amino acid, for the polar threonine residue, may alter the Aβ aggregation properties and slow down the formation of beta-sheet assemblies (Colombo et al., 2017).

Despite the relevance and potential applications of the novel APP KI AD murine model, two limitations should be addressed in future work. The first limitation relates to the presence of the Artic and the Iberian mutation from the original APP^G-F parent mouse. These mutations alter not only the properties of Aβ42, but also γ-secretase activity, resulting in an increased Aβ42/Aβ40 ratio. As Aβ40 is the main contributor to the formation of vascular Aβ deposits leading to vascular damage and cerebral amyloid angiopathy onset, it will be important to determine whether the increased Aβ42/Aβ40 ratio alters the onset of vascular pathologies in the described models (Herzig et al., 2004). A second limitation is that it is unclear whether APP^G-F-A673T mice exhibit cognitive changes. In another APP KI model, the APP^{NL-G-F/NL-G-F} harboring the Swedish, Arctic, and Iberic mutations reported by Sakakibara et al. (2018), mice showed an anxiolytic-like behavior in elevated plus maze and only a subtle decline in spatial learning ability, retaining normal memory functions in the Barnes maze. Based on these findings, the authors speculated that the APP^{NL-G-F/NL-G-F} model might have been more suitable for studying AD prevention rather than treatment after neurodegeneration. Introducing the Icelandic mutation into mice models that exhibit cognitive deficits, including those with mutations in proteins other than APP, could provide valuable insights into the broader applicability and potential therapeutic benefits of the Icelandic mutation.

In conclusion, Shimohama et al. (2024) present a novel KI mouse model that overcomes the limitations imposed by the presence of the Swedish double mutation in previous models, and they demonstrate that the A673T substitution results in attenuated amyloid pathology in vivo. The potential protective effect of the Icelandic mutation might be applicable not only in familial cases but also within sporadic AD. Indeed, the A673T substitution could be introduced into the APP gene in young individuals at risk of AD or in ApoE4 carriers through genome editing technologies. However, potential pathway interactions between the genetically modified APP and the ApoE will require further investigations. Thus, further research should aim at designing more accurate in vivo models lacking disease-modifying familial mutations to better reflect human conditions and lead to a deeper understanding of sporadic AD pathogenesis and the design of novel therapeutic approaches.