Shutting off transgenic Aβ production in middle-aged mice whose brains are jam-packed with amyloid pathology was able to rescue some aspects of cognition, according to a study in the February 27 Journal of Neuroscience. In this first publication of behavioral phenotypes in an Alzheimer's mouse model expressing tetracycline-regulated amyloid precursor protein (APP), researchers report that suppressing APP restored spatial and working memory, even while brain amyloid deposits and soluble Aβ42 remained. "At this very advanced stage of amyloidosis, we could still rescue some part of cognition," said Alena Savonenko of Johns Hopkins University School of Medicine, Baltimore, Maryland, who led the study with David Borchelt of the University of Florida, Gainesville. While some scientists say the work sustains hope that therapies regulating APP processing might be able to help people with advanced amyloidosis, others question the human relevance of data obtained in a mouse model with such rampant plaque pathology.

To address what happens when Aβ production is blocked after plaques form, Borchelt, Savonenko, and colleagues created a mouse in which they could suppress forebrain expression of a mutant APP transgene by adding doxycycline (DOX), a tetracycline derivative, to the animals' diet. Back in 2005, the scientists reported that turning off APP expression for three to six months kept Aβ pathology from progressing but did not dissolve existing plaques in these mice (see ARF related conference story and Jankowsky et al., 2005).

In the current paper, first author Tatiana Melnikova of Johns Hopkins and colleagues report the long-awaited cognitive data in this model. Why did it take so long? The inducible APP mice from the 2005 paper were unusually hyperactive; their constant scampering interfered with behavioral tasks that use novelty-induced motor activity as a readout for memory. Reducing the hyperactivity required changes in breeding strategy and DOX protocol. The scientists bred the APP and tetracycline-transactivator (tTA) transgenic lines 10 generations onto C57BL/6J and FVB/N backgrounds, respectively, then mated these to create double-transgenic F1 hybrids. (In tTA mice, behavior varies with strain background [see Han et al., 2012], and this particular combination seemed to work best.) The second "trick" was providing doxycycline during the first four to six weeks of life, instead of letting the animals develop DOX-free from birth as they had done in the 2005 study. Putting the animals on a DOX diet during their first few weeks of life "delayed expression of the transgene, allowing the mice to develop normally without too much APP flying around," Savonenko said.

After four to six weeks on DOX, the scientists switched the mice to DOX-free chow, unleashing APP expression until 12 to 13 months of age, by which time the mice had advanced amyloidosis. DOX was then reintroduced to shut off the transgene. "The whole cortex is flooded with plaques. In the hippocampus, you can barely see the structure," Savonenko said.

Yet after just seven days on DOX, the mice performed as well as non-transgenic and vehicle-treated controls in tests of short- and long-term spatial memory and working memory. The DOX-treated animals still struggled with tasks requiring episodic memory, and had more trouble adjusting to changes in experimental paradigms than did control mice.

As expected based on the 2005 findings, suppressing APP did not seem to influence Aβ load. "Everything we can see that's related to Aβ—soluble peptide, Aβ-immunoreactive species that run in gels the size of oligomers, amyloid plaques—all those things don't change when you turn the transgene off," Borchelt said. On the other hand, full-length APP, soluble APP ectodomains, and C-terminal APP fragments nearly disappeared after shutting off the transgene.

"This study clearly indicates that Aβ burden could be dissociated from cognitive decline associated with AD," Angèle Parent of the University of Chicago, Illinois, wrote in an e-mail to Alzforum. "If we could extrapolate these results to AD patients, it suggests that reducing expression of 'bad' APP (and its β C-terminal fragment accumulation) may be sufficient to improve cognitive function in a relatively short period of time." Recent work from her lab and others suggests that non-Aβ APP fragments play key roles in neuron health (see ARF related news story).

The data are curious in that memory improved in DOX-treated mice despite unchanged levels of what could be Aβ oligomers, which are widely seen as the most neurotoxic molecules in AD. However, the authors and other scientists were not certain they were measuring Aβ oligomers in this study. Though their immunoblots showed PBS-insoluble structures that ran around the same size as synthetic Aβ oligomers, "we could not easily detect soluble oligomers. If they are there, the levels are very low," Borchelt said. Nevertheless, those Aβ species persisted, and yet the mice got better.

Karen Ashe and Peng Liu of the University of Minnesota, Minneapolis, propose that the reduced levels of transgenic APP may contribute to the memory improvement in the mice, which have abundant neuritic plaques. The scientists have submitted a paper suggesting that neuritic plaques are neurotoxic "because the Aβ fibrils in the plaque core are a source of oligomers that diffuse away from the core," and that plaques can "act like space-occupying lesions in the brain and disrupt function," they wrote to Alzforum (see also Ashe and Aguzzi, 2013). In another submitted manuscript, Ashe and Liu show that oligomers within neuritic plaques require APP to exert their toxic effects. "DOX-mediated suppression of transgenic APP may block the APP-Aβ oligomer interaction, in turn ameliorating memory deficits," they suggest.

Another question is how well the inducible APP mice model AD, because their neuritic plaque load well exceeds that attained in the human disease, Ashe and Liu noted (see full comment below). Savonenko and Borchelt said the team is currently studying the effects of transgenic APP suppression in younger mice with less advanced amyloidosis. They are also exploring whether intracellular location of the presumed Aβ oligomers influences cognition in this model.—Esther Landhuis


Make a Comment

To make a comment you must login or register.

Comments on this content

  1. Regarding the conundrum of cognitive improvement despite lingering Aβ oligomers in the inducible APP mice, there are several issues to discuss.

    <p>First, it may be premature to consider the immunoreactive smears to be Aβ oligomers. Neither 6E10 (Figs. 6A, B) nor 82E1 (Figs. 6E, F) antibodies specifically recognize Aβ. Furthermore, besides measuring the molecular size of Aβ aggregates on SDS-PAGE, additional experiments are needed to confirm that the aggregates are oligomers. Our lab performs oligomer dissociation assays using protein denaturants (e.g., hexafluoroisopropanol), immunoblots using anti-oligomer antibodies (e.g., A11), and multiple Aβ-specific antibodies. Unfortunately, this paper provides no further supporting evidence to prove that the bands between 17-95 kDa (Figs. 6A, B, E, F) are Aβ oligomers.

    </p><p>Second, doxycycline (DOX)-mediated reduction of transgenic APP levels may explain the improvement of memory function in the mice. The mice in this study have an exceedingly high density of neuritic (thioflavin S-positive) plaques. These are neurotoxic because the Aβ fibrils in the plaque core are a source of oligomers that diffuse a limited distance away from the core (~50 microns). We have submitted a manuscript showing this.

    </p><p>We speculate that given enough neuritic plaques, they will act like space-occupying lesions in the brain and disrupt function. It turns out that the oligomers within neuritic plaques require a receptor to exert their toxic effects, and we have shown that this receptor is APP (we have submitted another manuscript showing this). So, DOX-mediated suppression of transgenic APP blocks the APP-Aβ oligomer interaction, in turn ameliorating memory deficits.

    </p><p>We would like to note some concern about the relevance of this mouse model to AD, because of the exceedingly high density of neuritic plaques. Such high densities of neuritic plaque are rarely, if ever, attained in AD. David Bennett's group showed that the total plaque load in AD is ~4 percent, which is the sum of neuritic and diffuse plaque loads (Bennett et al., 2004). In the current paper, the neuritic plaque load is not provided, but Euclidean geometry enables us to estimate the neuritic plaque load from the thioflavin S load; the neuritic plaque load (i.e., the volume of the cortical mantle occupied by neuritic plaques) is about 10 times the thioflavin S load. Looking at the thioflavin S photomicrographs in this paper, the thioflavin S load looks to be about 4 to 5 percent, which translates into a neuritic plaque load of 40 to 50 percent. We are doubtful about whether the cortical mantle in AD patients is occupied by this many neuritic plaques, and speculate that the mechanisms underlying deficits in AD and these mice may differ in important ways.


    . Neurofibrillary tangles mediate the association of amyloid load with clinical Alzheimer disease and level of cognitive function. Arch Neurol. 2004 Mar;61(3):378-84. PubMed.

  2. The interpretation of preclinical findings for human disease must always be careful.

    It is intriguing that all measurable Aβ species, including soluble Aβ42, remained unchanged after transgene suppression. The deposited amyloid could be a reservoir that leaches some soluble Aβ back into the interstitial fluid.

    The most pronounced changes were observed in APP fragments, some of which could be responsible for the improved cognition in DOX-treated mice. β-CTF has been implicated in cognitive and cellular toxicities and is suspected to contribute to the low efficacy of γ-secretase inhibitors. Thus, reduced levels of β-CTF may contribute to the cognitive improvement.

    Lack of significant change in the levels of Aβ oligomers during short-term DOX treatment is one of the most interesting findings of this study, since we did observe cognitive improvement during this time frame. However, it is difficult to know whether these oligomers are the same ones that were around before DOX treatment, or whether there is some new pool of oligomers derived, for example, from plaques. Are these "persistent" oligomers distributed in the same tissues as prior to DOX treatment? We are beginning to characterize these Aβ species in greater detail, and more studies are needed to answer these questions.

  3. Reply to comment by Karen Ashe and Peng Liu
    Regarding the detection of oligomeric Aβ from tissue extracts of tetAPPsi mice, we appreciate the comments of Dr. Ashe on methods to validate the identity of putative Aβ oligomers on immunoblots. While we agree that including a denaturing step could provide additional evidence that an immunoreactive band that migrates to a size consistent with an Aβ oligomer could indeed be an oligomer, it is worth noting that denaturation only provides evidence that the Aβ immunoreactivity was bound to a larger molecule. Such larger molecules could be oligomers, or complexes of Aβ with other proteins.

    In regard to the use of oligomer-specific antibodies for tissue extracts, we are unaware of any report that demonstrates that a band detected in immunoblots is indeed an oligomer. Validation would require some means of analyzing the immobilized entity to confirm that what is bound by the antibody is an oligomer. In preparations of purified Aβ peptide that have been analyzed by immunoblot, one can be confident that a band migrating at higher molecular weight is an oligomer, but in preparations from tissue, it is extraordinarily difficult to validate any particular immunoreactive band as being a true oligomer. In that regard, a key problem is that, unlike fully denatured proteins, which migrate to a consistent relative molecular weight, the size of the bands representing undenatured oligomers has been variable in the scientific literature. The origins of this variability are unknown, and may reflect the effects of tissue preparation or misinterpretation of the data. Thus, in regard to immunoblot detection of oligomers, the binding of an antibody to an entity of a particular size on an immunoblot does not carry the weight of evidence as the binding of antibody to a band of the expected molecular weight for fully denatured proteins (e.g., full-length APP, APP β-CTFs, or sAPPβ). What is currently lacking is precise molecular and structural characterization of an oligomer isolated from tissues, coupled with the knowledge of where such entities migrate in SDS-PAGE, and a consistent detection of such entities with the same ease with which other derivatives of APP are detected, such as full-length APP, APP β-CTFs, or sAPPβ.

    In addition to the uncertainties that face investigators who seek to link a particular oligomeric form of Aβ to a cognitive phenotype is the inability to know how the preparation of the tissue influences the structures they subsequently detect (by any method). We made every attempt to state clearly in our report that the bands labeled as oligomers in membrane fractions were only identified as such based on their relative migration rates, and that such bands could, in fact, be Aβ bound to other molecules or fragments of APP other than Aβ. We also clearly stated that such putative oligomers could have been created by tissue preparation. We note that all Aβ immunoreactive entities that we can detect in tetAPPsi mice, except full-length APP, APP β-CTFs, and sAPPβ, persist for many weeks after new APP synthesis has been depressed. No one would have been more pleased than we would be to have identified an Aβ reactive band of oligomeric size, or a particular Aβ species, that specifically disappeared as mice improved cognition. The tetAPPsi mice are available through the Jackson Laboratory, and we invite all interested parties to use these mice to identify the short-lived APP derivative that seems to mediate the more severe cognitive impairment observed in this model. If this entity turns out to be an oligomer of Aβ, none would be happier than we would be to have the tetAPPsi model used in such a discovery.

    In her comment, Dr. Ashe also suggests that mice that have very high burdens of thioflavin amyloid plaques may not be good models of cognitive symptoms associated with Alzheimer’s disease because the mechanism by which such plaques could disrupt cognition may not be relevant to human AD. While we agree that this could be true, we point out that our data show that the tetAPPsi mice can show remarkable improvements in performance on short-term memory tasks and more demanding spatial memory tasks in spite of the persistence of neuritic plaques at high densities. Thus, our data would argue that such plaques do not cause the severe memory deficits observed in this model. It is possible that the less severe cognitive deficits (lower cognitive flexibility) that persist along with the neuritic plaques in the tetAPPsi mice result from mechanisms that are unrelated to those that drive the same type of early cognitive impairment in humans. We wonder, however, how one would know what mechanisms in mice are truly relevant, as we are unaware of a definitive description of the basis for memory impairment in human AD.


News Citations

  1. SfN: How to Dispose of Plaques? Closing Spigot Won’t Do; Enzymes Nibble
  2. Beyond Aβ: Other APP Fragments Affect Neuron Health and Disease

Paper Citations

  1. . Persistent amyloidosis following suppression of Abeta production in a transgenic model of Alzheimer disease. PLoS Med. 2005 Dec;2(12):e355. Epub 2005 Nov 15 PubMed.
  2. . Strain background influences neurotoxicity and behavioral abnormalities in mice expressing the tetracycline transactivator. J Neurosci. 2012 Aug 1;32(31):10574-86. PubMed.
  3. . Prions, prionoids and pathogenic proteins in Alzheimer disease. Prion. 2013 Jan 1;7(1):55-9. PubMed.

Further Reading


  1. . Prions, prionoids and pathogenic proteins in Alzheimer disease. Prion. 2013 Jan 1;7(1):55-9. PubMed.

Primary Papers

  1. . Reversible pathologic and cognitive phenotypes in an inducible model of Alzheimer-amyloidosis. J Neurosci. 2013 Feb 27;33(9):3765-79. PubMed.