Two recent papers detail stumbles and strides for tau models. In the June 6 Nature Communications, scientists led by Michael Koob and Karen Ashe, University of Minnesota, Minneapolis, reported that most of the rampant neurodegeneration seen in the rTg4510 tauopathy mouse model comes not from overexpression of the human tau, as once thought, but from disruption of endogenous mouse genes. The result could be a tough pill to swallow for researchers who use this mouse, because it suggests much of the neurodegeneration is completely unrelated to tau.

  • The rTg4510 tauopathy model has massive neuron loss at seven months.
  • Much of the neurodegeneration is due to genomic disruption, not to mutant tau.
  • Tau propagates more quickly when human tau is knocked into the mouse locus.

Meanwhile, in the July 4 Journal of Biological Chemistry, Takaomi Saido, RIKEN Center for Brain Science, Wako, Japan, reports on a new knock-in mouse in which human tau replaces the mouse version. Saido and colleagues report that pathological tau seeds derived from human Alzheimer’s brains propagate faster in these mice than in animals expressing the murine protein (see image below). Crossed with an amyloid model, the knock-ins accumulate more tau near dystrophic neurites, suggesting Aβ incites tau pathology there.

The rTg4510 mouse model has been widely used since 2005 (Ramsden et al., 2005). Ashe and colleagues created it by injecting the P301L tau—which causes familial frontotemporal dementia—into a single-celled embryo such that it integrated randomly into the genome. The mutant human tau transgene was spliced downstream of a tetracycline operon control element. By crossing these mice into a line with a tetracycline (tTA) transactivator under control of the CaMKIIα promoter, the researchers drove human tau gene expression in the forebrain. Use of the tTa transactivator allowed transgene expression to be turned off by feeding doxycycline to the animals. These mice exhibited tau pathology by 4 months of age and severe atrophy well before one year. They lost 20 percent of their brain mass by 7 months.

Pass It On. When patient-derived pathological tau seeds are injected into mouse brain, tau pathology (bright orange) propagates more robustly in tau KI mice with a human tau (bottom) than in wild-type mice (top). [Courtesy of Saito et al., 2019. JBC.]

It has been assumed that the overexpression of the P301L tau isoform in the forebrain was the cause of the severe phenotype in this mouse. Ashe set out to discover why P301L mutant tau was so detrimental. However, first she needed a version in which the mutant tau gene was placed into a specific location, instead of being randomly inserted into the genome. That way, she could introduce other tau variants into the same spot and directly compare them.

Koob and first author Julia Gamache used a targeted-insertion technique called Flp/Frt recombination. They placed the P301L tau in a position in the genome known to allow gene expression without disrupting endogenous mouse genes. The researchers crossed these T2 mice, as they were called, to the same tTA-driver line used in rTg4510 mice so that the tau transgene would be specifically expressed in the forebrain.

Though homozygous T2 mice produced more tau than rTg4510, they had a milder, more delayed phenotype. Tau pathology accumulated later and to a lesser degree in rT2/T2 mice, with paired helical filaments appearing in the rTg4510 hippocampus at 2 months, and in the T2/T2 mice at 5 months. At 7 months, when rTg4510 mice have already lost considerable brain volume, T2 brain mass held steady, though it dropped off after 10 months to levels comparable to the rTg4510 mice.

Does something besides tau cause the earlier and more severe phenotype in the rTg4510 mice? The researchers used whole-genome sequencing to look for genomic disruption. They found that 70 copies of the tau transgene had displaced 263,608 base pairs of the gene for fibroblast growth factor 14 (Fgf14) on chromosome 14. This removed the promoters and first exons of most isoforms of Fgf14, vastly decreasing its expression. Similarly, WGS revealed that in both mouse models, seven copies of the tTa transgene inserted on chromosome 12, disrupting five functional, endogenous genes: Vipr2, Ptprn2, Wdr60, Esyt2, and Ncapg2. The gene disruptions confirm results from a molecular characterization of transgene integration sites for 40 commonly used transgenic mouse lines from the Jackson lab published earlier this year (Goodwin et al., 2019). 

The results suggest that experiments using the rTg4510 mouse could be misleading. “It can be problematic if you don’t define the genetics behind these phenotypes,” Koob said. “That needs to be done if we want to work with the right disease-relevant models.”

“This sobering analysis is important for the field,” said Li Gan, University of California, San Francisco. “We should all be aware of the limitations of transgenic models.” However, the paper does not invalidate all results from the rTg4510 mouse or any other animal model, it just means researchers have to confirm their results in a second model, she said.

Joanna Jankowsky, Baylor College of Medicine, Houston, echoed those sentiments, cautioning that researchers shouldn’t throw the baby out with the bathwater. “I think there is still value in the rTg4510 model; it just depends on the question you’re asking,” she told Alzforum. For example, the model could help answer questions about tau pathology and neurodegeneration, but perhaps not about therapeutics. She also argued that targeted-insertion methods to create new mouse lines were costly, and suggested that embryonic injection was still a legitimate way to create mouse models, as long as any findings hold in multiple mouse models.

“We believe that the issue of insertion effects—via Fgf14 or other mechanisms—is significant, and that the effect of transgene insertion is an essential consideration when interpreting data from transgenic models,” wrote Marc Aurel Busche and Samuel Harris, University College London, to Alzforum. Busche and Brad Hyman used the rTg4510 mouse to study how Aβ and tau interact to affect neuronal activity (Dec 2018 conference news). “In this regard, application of suitable controls, including the use of alternate mouse lines, is crucial to ensuring any conclusions obtained from models of neurodegenerative disease are accurately drawn.” 

Ashe agreed, saying that ideally all results should be confirmed in multiple mouse lines, as well as in vitro and in human tissue where possible. “I will believe a hypothesis only if findings from all three lines of investigation—mice, cells and humans—converge.”

In pursuit of a model that more faithfully recapitulates human disease, Takashi Saito and colleagues developed the MAPT knock-in mouse they described in the JBC paper. This lab has argued strongly for using knock-in rather than overexpression models, and many researchers now use the APP knock-in models they developed (Dec 2016 news; Apr 2019 conference news). Saito used homologous recombination to insert a human tau gene into the genome of tau-deficient C57BL/6J mice. When the researchers injected pathological tau aggregates from human AD brains into the brains of these mice, tau pathology propagated faster than in mice with the murine version of the protein (see image above). This suggested that pathological human tau interacts better with human than murine tau. The researchers then crossed the MAPT knock-in mice with the APP knock-ins (Sep 2016 news). In the presence of amyloidosis, tau accumulation intensified near dystrophic neurites, suggesting that the Aβ exacerbates tau pathology.

Koob praised the model, saying it better approximated what was going on in human patients. However, he pointed out that the researchers only replaced the coding region of the tau gene, without the associated DNA and RNA regulators. Koob says he has knocked complete human tau genes into mice and is in the process of doing the same thing for APP.

Ashe said she would like to see a mouse in which murine APP, MAPT and APOE4 are all replaced with human genes. She thinks their neuropathological, molecular, and functional phenotypes at different ages would be informative. “Theoretically, it should be possible to develop a knock-in mouse that develops catastrophic brain organ failure as occurs in humans with AD by humanizing the right genes,” she wrote to Alzforum. “Such a mouse would be an ideal model for testing experimental therapies.”—Gwyneth Dickey Zakaib


  1. We applaud this careful genetic and phenotypic analysis of the rTg4510 model and encourage similarly comprehensive investigations of other widely used mouse lines. Along this line, we recently demonstrated that in the hAPP-J20 mouse model, the transgene array is inserted into intron one of the Zbtb20 gene, causing a deletion of ~41kb of intronic sequence. Studies are in progress to determine whether this genetic alteration contributes in any way to the phenotype of this model.

    It is likely that the genomes of other transgenic mouse models used in dementia research are also altered by transgene insertions. Depending on how they were constructed, even knockout and knock-in models can have genomic alterations that could contribute to their phenotype, for example, through the inadvertent deletion of noncoding RNAs.

    We therefore strongly encourage all researchers to undertake a full genetic characterization of their lines, including determining the integration site(s) of all transgenes and the careful analysis of all DNA sequences that were deleted or modified in any way.


    . The integration site of the APP transgene in the J20 mouse model of Alzheimer's disease. Wellcome Open Res. 2017;2:84. Epub 2018 Oct 10 PubMed.

  2. The overexpression paradigm is less problematic for modeling tauopathy than for Aβ amyloidosis because limited proteolysis is not a prerequisite for physiological and pathological processing of tau protein. Overexpression of APP results in overproduction of APP, APPsβ, CTFβ, Aβ, AICD, APPsα, CTFα, p3 fragment, APPsη, CTFη, Aη-β and Aη-α, where Aβ is the major component of plaques.

    It is probably important to note that most anti-Aβ antibodies bind to CTFβ and APPsα. This requires a caution in interpreting the effect of immunotherapy on the APP-transgenic mice and of human subjects.

Make a Comment

To make a comment you must login or register.


Research Models Citations

  1. rTg(tauP301L)4510

Mutations Citations

  1. MAPT P301L

News Citations

  1. Tau Silences, Aβ Inflames; Hitting Excitatory Synapses Hardest
  2. Knock-In Alzheimer’s Mice Catch on More Broadly in the Field
  3. Parsing How Alzheimer’s Genetic Risk Works Through Microglia
  4. Do APP Knock-ins Call Overexpression Models of AD into Question?

Paper Citations

  1. . Age-dependent neurofibrillary tangle formation, neuron loss, and memory impairment in a mouse model of human tauopathy (P301L). J Neurosci. 2005 Nov 16;25(46):10637-47. PubMed.
  2. . Large-scale discovery of mouse transgenic integration sites reveals frequent structural variation and insertional mutagenesis. Genome Res. 2019 Mar;29(3):494-505. Epub 2019 Jan 18 PubMed.

Further Reading

Primary Papers

  1. . Factors other than hTau overexpression that contribute to tauopathy-like phenotype in rTg4510 mice. Nat Commun. 2019 Jun 6;10(1):2479. PubMed.
  2. . Humanization of the entire murine Mapt gene provides a murine model of pathological human tau propagation. J Biol Chem. 2019 Aug 23;294(34):12754-12765. Epub 2019 Jul 4 PubMed.