bibfall hekter dayshawna

[Edilma Howard]

In contrast to the overexpression models that have been used for decades, the new mice tend to have subtle phenotypes, which show up only after age, or other risk factors such as a high-fat diet, have exacted a toll. Scientists hope these mice will more accurately reflect the changes that occur in people with AD, Michael Sasner at the Jackson Laboratory (JAX) told Alzforum.

For example, for cognitive testing, the scientists use not the water maze but electroencephalography and touchscreen tests. Similar to human tests of attention, the mice are rewarded for touching a bright spot on the screen with their nose (Mar 2011 news).

Mice with AD-like changes such as plaques, phosphorylated tau, neuroinflammation, dendritic spine loss, or attention deficits go on to preclinical testing, where researchers determine whether drugs that have been studied in people, such as levetiracetam, verubecestat, and aducanumab, have similar effects in mice as they do in people. This serves to establish a sense of their predictive validity.

The first lines in this project used key AD genes: APP, tau, APOE, and TREM2. Because the well-known pathological APP and tau mutations are absent in late-onset AD, the researchers knocked in wild-type humanized sequences. Rather than humanize the entire APP gene, they edited only the Aβ42 sequence. Human Aβ42 varies by three amino acids from the mouse peptide, and it is more neurotoxic. For tau, the researchers humanized the entire gene so that it would be processed the way it is in human brain. Human tau forms six isoforms, compared to three in mice. The scientists hope that humanizing the gene will bring the ratio of 3R to 4R isoforms closer to what it is in human brain. In separate mouse lines, researchers knocked in each of the common APOE alleles. Although other groups had previously created APOE knock-ins, those were not widely or freely available, Sasner noted. For TREM2, they inserted the LOAD variants R47H and Y38C, or deleted the gene. See below for a list of available models to date.

LOAD2 and LOAD3 themselves constitute backgrounds into which geneticists can drop additional variants. Some of the first such polygenic lines express pathogenic variants of the lipid transporter ABCA7, the metabolic gene MTHFR, or the immune gene PLCG2. PLCG2 has a well-known protective variant, but MODEL AD researchers instead tested the risk variant M28L.

How well do these polygenic mouse lines represent AD pathology? The mice start out healthy, and gene expression in their brains bears little resemblance to patterns seen in AD brain as per postmortem bulk transcriptomic data from AMP-AD (Wan et al., 2020). However, by 1 year of age, gene expression in specific brain regions exhibits AD-like profiles (Preuss et al., 2020).

Unlike 5XFAD mice, which deposit amyloid starting at 2 months of age, the newer models start developing problems much later in life. JAX rears the mice to 24 months to fully assess their subtle phenotypes.

That said, layering on environmental risk factors amplifies things. For example, a high-fat diet jacked up proinflammatory cytokines in year-old PLCG2-M28L mice (Oblak et al., 2022). In LOAD2 mice, a high-fat diet boosted insoluble Aβ42, neuroinflammation, and plasma NfL by one year, and cortical neuronal loss in females by 18 months, but caused no plaque deposition or memory problems. The scientists are considering adding other insults, such as heavy metals and ozone to model the effects of pollution (Mar 2004 news; Feb 2012 news; May 2020 news).

Researchers throughout the field have been slow to shift toward using MODEL-AD mice, Sasner told Alzforum. A first publication is in the works and may generate interest. JAX is providing mice to several research groups at pharma companies, which are eager to move away from using transgenic models and toward knock-ins.

Knock-in Familial and Protective Mutations
For those who want more dramatic AD phenotypes than the LOAD models afford, fear not. In a separate initiative, JAX collaborates with academic labs to create FAD and other knock-ins. One project inserts multiple pathogenic APP SNPs into the mouse genome against a background of humanized Aβ42. For example, the SAA model, which was made in collaboration with Denali Therapeutics, combines the Swedish, Arctic, and Austrian mutations. The mice develop parenchymal amyloid in the cortex and hippocampus starting at 4 months old (Xia et al., 2022).

The field already has some APP knock-in options with the APPNL-F and APPNL-G-F models. They feature a different mix of mutations and were not widely available, though the RIKEN Center for Brain Science in Wako, Japan, which supplies the mice, recently loosened restrictions (Sep 2022 news).

For aficionados of APOE, JAX researchers have generated specialized mice, including ones that carry the protective Christchurch and Jacksonville mutations on an APOE3 or E4 background (Nov 2019 news; Oct 2021 news; Sep 2022 news).

In collaboration with Michael Koob at the University of Minnesota, Minneapolis, JAX scientists are making pathogenic tau variants on a human tau background. So far, these include frontotemporal dementia variants N279K and P301L. These are not AD genes, but good mouse models of AD will require tau pathology, and Sasner and colleagues plan to cross these with other models. One such cross combines SAA mice with N279K tau and APOE4 in an attempt to generate amyloid, tau, and vascular pathology in one animal. These mice, like many of the lines, have not been fully characterized yet.

Studying Biology Via Reporter Strains
The goal behind many mouse models is to test therapeutics, but that is not their only purpose. Equally important is having strains that allow scientists to dissect the biology of the disease, particularly the role of glial cells in amyloidosis. Despite intensive research on these cells, scientists have no consensus on what factors make microglia and astrocyte responses helpful or harmful, and how that varies by disease stage. Reporter models will help track the timing, extent, and type of glial activation. They could also help reveal the effects of therapeutic interventions on the glial response. JAX is generating numerous such reporter models (see table above).

There is no doubt that we need more murine models for AD research, but much of the work with existing models is being misinterpreted. The vast majority of studies in mouse models of amyloidosis or tauopathy treat mice with test agents very early in a prevention type of paradigm. Most studies in humans have treated long after the pathology has matured, in a therapeutic paradigm. These are apples and oranges.

One of the few treatments successful in APP mice in a therapeutic paradigm, immunotherapy, works even when started long after amyloid has matured to remove pathology and rescue memory loss (Wilcock et al., 2004; DeMattos et al., 2012). The mouse models also predicted ARIA, via micro-hemorrhages and increased angiopathy (Pfeifer et al., 2002; Wilcock et al., 2004). Thus far, immunotherapy is the only treatment that can remove amyloid and slow cognitive decline when applied in a therapeutic paradigm in humans (aducanumab, donanemab, lecanemab). It is likely that as more of the "failed" approaches in AD therapeutic paradigms are tested in prevention paradigms, some will also have benefit preventing AD, as they did in APP mice.

A final comment: Any mouse that does not have amyloid plaques or tau tangles is not a model for AD, as these pathologies define the disease. Perhaps the models without amyloid represent individuals carrying risk genes who do not get disease, but by definition, they are not AD models.

I am awed by the sophistication of the scientists making these models and the insights they will bring as we develop more agents for AD. That said, we need not denigrate the successes of the existing models to justify the development and study of new and, possibly, improved ones. Anti-amyloid immunotherapy was the first intervention to show success in APP mice. It is the first intervention to slow decline in people living with AD. It just took two decades to get here.

Wilcock DM, Rojiani A, Rosenthal A, Subbarao S, Freeman MJ, Gordon MN, Morgan D.Passive immunotherapy against Abeta in aged APP-transgenic mice reverses cognitive deficits and depletes parenchymal amyloid deposits in spite of increased vascular amyloid and microhemorrhage. J Neuroinflammation. 2004 Dec 8;1(1):24. PubMed.

There seems to be some misunderstanding about our App (Saito et al., 2014; Watamura et al., 2022), MAPT (Hashimoto et al., 2019), and Psen1 (Sasaguri et al., 2018) knock-in mice, some of which are even unreferred-to. These mice have been freely provided to all the basic researchers worldwide. As a result, the number of mouse users exceeded 600 a long time ago: the only continents where the mice do not exist are Africa and Antarctica. Consistently, our original paper is one of the most highly cited according to Clarivate Analytics. We chose to deposit our mice to RIKEN BioResource Center rather than to the Jackson Laboratory.

I have some comments that may help the students and postdocs who may be using the LOAD models. As Dr. Morgan indicated, SNPs in most GWAS-identified genes are unlikely to exert substantial effects, except for ApoE and TREM2, because each of their odds ratios is very small. There are more than 70 risk-associated genes thus far identified (Bellenguez et al., 2022), so it would be very interesting to analyze the mice that carry most, if not all, of these SNPs, although the number of mouse lines to create would be astronomic. Here is a tip for humanized tau mouse user. We have generated a number of mutant MAPT knock-in mice. One thing for sure is that the knock-in mice carrying the P301S mutation showed no tau pathology up to 24 months.

OWASP Cornucopia is a mechanism in the form of a card game to assist software development teams identify security requirements in Agile, conventional and formal development processes. It is language, platform and technology-agnostic.

c01484d022