Researchers led by Takaomi Saido and Hiroki Sasaguri at the RIKEN Center for Brain Science, Wako, Japan, have developed a new, double knock-in model of amyloidosis. The mice produce human Aβ43 and Aβ42 in the brain, and they begin to deposit dense-core amyloid plaques in the cortex and hippocampus before they are 3 months old. They also develop rampant neuroinflammation. Saido believes the mice will change the landscape of experimental AD research. The paper was uploaded to bioRXiv on April 30.

“This new mouse model is a very welcome addition to the toolbox of AD mouse models. It avoids the Arctic mutation, which is helpful for some questions, but not necessarily for others,” wrote Stefan Lichtenthaler, German Center for Neurodegenerative Disease, Munich, to Alzforum (see comment below).

Saido and colleagues had previously developed APP knock-in mice as an alternative to a plethora of models that overexpress the human amyloid precursor protein and/or human presenilin. They argued that knock-ins more faithfully recapitulate the pathology that occurs in the human brain, since they are devoid of potential overexpression and genomic rearrangement artifacts (Apr 2014 webinar). 

Their APP-NL-F knock-ins express APP with the Swedish and Iberian mutations that cause autosomal-dominant AD. Because these need more than a year to develop an appreciable plaque load, the researchers also made a version containing the Arctic mutation. This glutamic-acid-to-glycine swap at position 22 of the Aβ sequence makes the peptide much more likely to aggregate. As such, these APP-NL-G-F mice begin to lay down plaques in the brain before they are 2 months old.

Alas, this Arctic mutation has its downsides. Proteases in the brain have a harder time digesting Aβ peptides with the mutation, and it may bind less avidly to Aβ antibodies being tested as immunotherapies. In short, changing the peptide also limits its relevance.

Enter the new double knock-ins. Co-first authors Kaori Sato, Naoto Wakamura, and colleagues crossed the APP-NL-F mice with a knock-in expressing presenilin 1 with the P117L mutation that causes familial AD. The APP-NL-F/PS1 P117L double knock-ins produce about 25-fold more Aβ than does the NL-F single knock-in, and they have prominent amyloidosis by 12 months. Plaques are often of the dense-core variety, which is associated with AD. The double knock-ins have about threefold and sixfold more dense-core plaques in the hippocampus and cortex, respectively, than do the APP-NL-F mice. Microglial inflammation is rampant in the cortex and hippocampus, as well.

Curiously, microglia are much more active in the hippocampi of the double knock-ins than in the hippocampi of the Arctic APP-NL-G-F single knock-ins, despite similar plaque loads. The authors believe higher levels of Aβ43 and dense-core plaques in the double knock-ins might explain the difference. Microglia might even be responsible for laying down dense-core plaques as a way to protect the brain (Apr 2021 news).—Tom Fagan


  1. This new mouse model is a very welcome addition to the toolbox of AD mouse models. It avoids the Arctic mutation, which is helpful for some questions, but not necessarily for others.

    It is great that the authors will make their mouse model easily available. With the recent mouse model generated by Denali that is freely available, it will be important to compare both new mouse models in detail to established ones, for example with regard to changes in the activity state of microglia and the structure of amyloid plaques (Feb 2021 news). Such a comparative analysis will be extremely helpful for researchers to decide which mouse model to use for their specific question.

    This is particularly important because no lab will want to import and handle many different mouse models in parallel. I am glad to see that the new mouse model again contains the I716F mutation which boosts the Aβ42/40 ratio. It is this mutation that I introduced into the field many years ago (Lichtenthaler et al., 1999) and that was later discovered in AD patients in Spain.


    . Mechanism of the cleavage specificity of Alzheimer's disease gamma-secretase identified by phenylalanine-scanning mutagenesis of the transmembrane domain of the amyloid precursor protein. Proc Natl Acad Sci U S A. 1999 Mar 16;96(6):3053-8. PubMed.

  2. Just a note. The double mutants are heterozygous for the PSEN1 mutation, making it easy to breed them. We expect homozygous mutants to show even more aggressive pathology.

  3. Alzheimer’s disease mouse modelling needs tricks, including overexpression of human genes such as APP, PSEN, or tau, and combinations of mutations in these genes, to induce relevant neuropathology. However, a complete model for AD, i.e., linking amyloid plaques to abnormal tau pathology, induction of neurofibrillary tangles, granulovacuolar neurodegeneration, and neuronal cell loss, has, as yet, not been achieved. While the field agrees that the existing models have limitations, and should be only used if taking these into account, most of the models also come with serious technical flaws, which are less acknowledged.

    The main flaw is the lack of good controls for the overexpression of the transgenes. Indeed, mice overexpressing the non-mutated transgenes are not available because it is next to impossible to recreate the same gene insertions.

    Further, several times over the last years Saido and colleagues have made the point that overexpression of proteins leads to artefacts, and they have shown that knock-in models largely circumvent this important issue. Nevertheless, the available knock-in models come with their own limitations, mainly because a combination of APP mutations is needed to produce amyloid plaques in a reasonable (experimental) time frame. 

    In the current manuscript, Saido and colleagues provide a solution for one of the issues, i.e., the Arctic mutation introduced in the middle of the Aβ sequence to hasten aggregation and plaque formation. This mutation makes Aβ less prone to antibody binding or to proteolytic degradation, among other problems.

    While removing this mutation and accelerating the amyloid pathology using a PSEN knock-in mutation is a step forward, it is a small step compared to the more fundamental issues that remain to be solved before a full model of AD will be achieved. For instance, the combined use of mutations at the β- and γ-secretase site in APP raise questions as to whether the model is useful to investigate the effects of drugs that interfere with these enzymes. Another problem is that the very strong increase in Aβ42 will affect the biophysical composition of the amyloid plaques (i.e., the composition of amyloid plaques in mice are different from the ones in human). Most importantly, we are still far from a comprehensive model of AD that develops tangle pathology and neurodegeneration without further manipulations of the tau gene. 

    Let’s hope that Saido and colleagues can, with further clever molecular tinkering, solve these issues over the coming years. We, in contrast, think that one or more essential human-specific factors are lacking in mice and have followed an alternative road by complementing the models of Saido and colleagues with transplanted iPSC-derived human brain cells. Indeed, we see indeed considerable additional pathology in human neurons compared to mouse neurons and therefore believe that the downstream events in the pathogenesis of Alzheimer’s disease are partially human-specific.

  4. In this study, the Saido lab continues its mission of generating new mouse knock-in models for Alzheimer’s disease that are more physiologically relevant and useful for studying disease mechanisms and testing candidate therapeutics. The new mouse produces wild-type secreted Aβ peptides, avoiding altered properties of these peptides due to the presence of the E22G “Arctic” mutation (Aβ numbering) that they used in their previously developed KI mouse model.

    In addition to the Arctic mutation, the previous “NL-G-F” KI mice contain the KM-to-NL Swedish double mutation near the β-secretase cleavage site and the I45F Iberian mutation within γ-secretase cleavage sites. The Swedish mutation increases β-secretase cleavage of APP to elevate the membrane-bound C-terminal fragment C99, which in turn serves as a substrate for γ-secretase in producing Aβ peptides. The Iberian mutation dramatically elevates the Aβ42/40 ratio that is widely considered critical to AD pathogenesis. Although the NL-G-F mice displayed early and robust AD pathology (e.g., amyloid plaques, gliosis) and cognitive impairment with physiological levels of APP expression, the presence of the Arctic mutation leads to Aβ deposits with altered biophysical properties and affects certain Aβ-targeting drug candidates.

    To produce mice that develop AD phenotypes early but with wild-type Aβ, here the Saido lab crossed homozygous “NL-F” KI mice (with only the Swedish and Iberian mutations) with mice carrying a knock-in of the presenilin-1 (PSEN1) familial AD mutation P117L. The resulting mice produce elevated Aβ42/40, similarly to NL-G-F mice, and they develop amyloid plaques and gliosis by 3 months of age but with wild-type Aβ secreted and deposited.

    Unfortunately, no cognitive testing is described in the new paper, but presumably this is work in progress. Another interesting unanswered question is the role of longer forms of Aβ of 45 to 49 residues in length. Like many FAD mutations, the Iberian I45F mutation can elevate these peptides (specifically Aβ46 for this mutation). The new mice may help address whether these membrane-anchored forms of Aβ contribute to pathogenesis.

    Takaomi Saido is to be commended for persevering over many years to systematically and rigorously work toward these important AD KI mouse models and for making these mice freely available to academic researchers. The new mouse model is an important step in the right direction and should speed the day when the world finally has effective agents for the prevention and treatment of AD.

Make a Comment

To make a comment you must login or register.


Webinar Citations

  1. Good-Bye Overexpression, Hello APP Knock-in. A Better Model?

Research Models Citations

  1. APP NL-F Knock-in
  2. APP NL-G-F Knock-in

Mutations Citations

  1. APP KM670/671NL (Swedish)
  2. APP I716F (Iberian)
  3. APP E693G (Arctic)
  4. PSEN1 P117L

News Citations

  1. Microglia Build Plaques to Protect the Brain

Further Reading

No Available Further Reading

Primary Papers

  1. . New App knock-in mice that accumulate wild-type human Aβ as rapidly as AppNL-G-F mice exhibit intensive cored plaque pathology and neuroinflammation. bioRxiv. April 30, 2021.