This is Part 2 of a two-part series. See also Part 1.
2 November 2012. Animal models of Alzheimer’s disease have provided key insights into the disorder, but researchers agree they have limitations for studying human disease. At the Society for Neuroscience 2012 annual meeting, held 13-17 October in New Orleans, Louisiana, several presentations raised questions about current animal models and their relationships to disease mechanisms. For example, researchers reported that in mice, overexpression of amyloid precursor protein (APP) during development—something that does not occur in human disease—blunts the ability of the adult brain to respond to anti-amyloid therapy. This hints that current transgenic animal screens could miss useful drugs. Other presentations fueled doubts about the toxicity of Aβ42, looked at the effects of mouse background strain on amyloid pathology, and described an unusual strategy for targeting tau pathways. Several talks noted that amyloid plaques by themselves are insufficient to weaken cognition in mice, which seems true in people as well (see, e.g., ARF related news story; ARF news story). Two of the studies were chosen for discussion at an SfN press conference on AD, while others were part of nanosymposia on animal models and Aβ (see Part 1 of this series for other press conference coverage).
In most amyloid-based models of AD, mutant gene expression occurs throughout the life of the animal. This may not reliably reflect human AD, in which Aβ deposits accumulate only late in life, said Alena Savonenko at Johns Hopkins University, Baltimore, Maryland, in the press conference. She wondered what would happen if APP overexpression, a mainstay of most mouse mimics of AD, turned on only in adulthood. To explore this, she used the TetO-APPSwe/Ind mouse created in 2005 by Joanna Jankowsky, then at David Borchelt’s lab at Johns Hopkins (see Jankowsky et al., 2005). In this animal, the human APP transgene is under the control of the TetO promoter and can be turned off by feeding the mice doxycycline (see description of Tet-Off System). The scientists first used the model to test what happened when they shut down APP production late in life, after plaques had formed (see ARF related news story). Savonenko, who was a coauthor on the earlier paper, tried the opposite approach. She turned off the transgene during development to more closely mimic the human situation.
Savonenko and colleagues fed half the TetO-APP mice doxycycline for their first month of life to suppress the APP gene (these mice are called late expressors), while the other half produced APP throughout development (early expressors). After the first month, all mice were allowed to express the transgene. At 13.5 months old, both groups had similar plaque loads and levels of soluble Aβ in their brains, and both showed impaired spatial memory in the Y-maze, Savonenko reported. The researchers then switched off APP for one week in all mice to simulate the effects of anti-amyloid therapy given late in life. Plaque burden did not change, but in the late expressors, spatial memory bounced back to wild-type levels, while the early expressors showed no improvement. This implies that the presence of APP or one of its metabolites during development weakens the ability of the adult brain to respond to anti-amyloid treatment, Savonenko told the audience. Since most current APP mouse models overexpress the transgene throughout life, they might miss the effects of experimental drugs, she said.
The researchers saw other differences. Early expressors became more hyperactive than late expressors, although both groups returned to normal activity levels after APP was shut off. The brains of the early group may become sensitized to the effects of APP, said Jankowsky, who is now at Baylor College of Medicine, Houston, Texas, in a nanosymposium talk. She noted that when the transgene stays silent for the first six weeks of life, adult APP mice have normal activity levels (see Rodgers et al., 2012). This points to a developmental effect of APP on motor circuits that control activity, she suggested. Early expressors also show abnormalities on electroencephalograms, displaying sharp discharges that get worse with age. Late expressors, by contrast, look normal by EEG. Other researchers have found abnormal EEG activity linked to seizures in some strains of AD mice (see ARF related news story).
In a separate talk, Borchelt, now at the University of Florida, Gainesville, discussed whether APP or one of its breakdown products causes the observed cognitive and behavioral impairments in TetO-APP mice. In one experiment, he gave 13-month-old mice doxycycline to turn off APP expression, and their cognition rapidly improved compared to control animals. Borchelt ruled out relief from Aβ monomers as the cause of the improvement, noting that their levels stayed the same whether mice were on or off doxycycline, even after four weeks of treatment. He believes that soluble Aβ oligomers are not culpable, either, since he could not detect them by Western blotting of extracts from TetO-APP mice either on or off doxycycline, even using antibodies 6E10 and 4G8, which are sensitive to nanogram quantities. Oligomers do remain in the insoluble membrane fraction, but show no change on or off doxycycline, he added. By contrast, APP, its C-terminal fragments (CTFs), and sAPPα and β levels all drop in animals on doxycycline, correlating with memory improvements. The results suggest that sAPP, APP, or CTFs could be more than just bystanders in cognitive decline, and might interact with Aβ to weaken cognition, Borchelt said. Other researchers have found evidence for APP/Aβ interactions (see ARF related news story) and β-CTF toxicity (see ARF related news story; ARF news story; and ARF news story), or have pointed to a pyroglutamate-modified form of Aβ as the bad seed (see ARF related news story).
In the same nanosymposium, Chris Janus at the University of Florida, Gainesville, also raised the question of the toxic Aβ entity. He used BRI-Aβ42 mouse models created by Todd Golde and colleagues at the Mayo Clinic in Jacksonville, Florida (see ARF related news story on McGowan et al., 2005). These animals express Aβ42 fused to the BRI protein. The enzyme furin cleaves off Aβ and releases it into the extracellular matrix without the need for APP overexpression or processing. The mice deposit Aβ and develop florid plaques in forebrain by 17 months of age. Nonetheless, they show no behavioral impairments, Janus reported. He tested them on a wide variety of assays, including fear conditioning, open field anxiety tests, rotarod tests of motor coordination, water maze memory tests, and conditioned taste aversion, but found no abnormalities. Chronic exposure to aggregated extracellular Aβ by itself is not sufficient to cause cognitive decline, Janus concluded. The findings echo those in the TetO-APP mice, and even in humans, where plaques do not seem to directly affect cognition. An audience member suggested that Aβ localization might make a difference, pointing out that in traditional mouse models, Aβ gets released at synaptic clefts in response to activity, putting it in the right place to harm transmission. Further muddying the waters, a rat model that features viral overexpression of BRI-Aβ42 and/or BRI-Aβ40 in the hippocampus does show cognitive impairments (see Lawlor et al., 2007).
Other researchers suggested that strain differences confound AD mouse models. Takashi Morihara at Osaka University, Japan, leveraged such differences to find genes that interact with APP to modify pathology in the hopes that this might cast light on causes of sporadic AD. He crossed Tg2576 mice, which express APP with the Swedish mutation, into three backgrounds: C57BL/6, SJL, and DBA/2. While APP expression remained the same on all backgrounds, DBA mice had lower levels of cortical Aβ42 than did the other two strains.
To identify genes responsible, Morihara and colleagues used a transcriptomics approach to compare RNA expression from the three strains. Among the 54 genes that varied among the mice, four showed changes in expression that correlated with Aβ levels. Two of those candidates were splice variants of kinesin light chain 1 (KLC1). This protein associates with microtubules and plays a role in organelle transport and trafficking from the Golgi apparatus. Morihara found that DBA mice produce less KLC1 variant E. This splice form modifies Aβ accumulation, Morihara concluded. He speculated that variant E might affect APP localization and, thus, processing. Other work has linked Aβ production to APP localization within endosomes (see, e.g., ARF related news story and ARF news story). Mouse KLC1 splice patterns are conserved in people, Morihara said. He found increased levels of KLC1 variant E mRNA in AD brains, suggesting that in humans, too, it may contribute to Aβ burden.
In a change of pace from these Aβ-centered talks, Fred Van Leuven at KULeuven, Belgium, described a tau-based treatment strategy. At the press conference, he pointed out that although phosphorylation gets the lion’s share of the attention, tau can undergo several types of modifications that might affect its behavior. Van Leuven focused on O-GlcNAcylation, the addition of a sugar molecule to a protein. Other groups have reported that O-GlcNAcylation of tau and several other proteins drops in the Alzheimer’s brain (see, e.g., Liu et al., 2009; Dias and Hart, 2007; Robertson et al., 2004), and that pumping up O-GlcNAcylation can slow neurodegeneration in tau model mice (see Yuzwa et al., 2012, and Yu et al., 2012).
Van Leuven and colleagues tested this idea in aged P301L tau mice and in young biAT mice, which express mutant human APP as well as tau, and typically die at four months old (see Terwel et al., 2008). To accelerate O-GlcNAcylation, the researchers fed the animals an inhibitor of O-GlcNAcase, the enzyme that removes sugars from proteins. In agreement with their hypothesis, treated animals had significantly better motor skills, higher body weight, and longer lifespan compared to controls. The surprise came when the researchers isolated tau from the mouse brains, and found no sugars on it in either treated or untreated mice. This means tau is not O-GlcNAcylated in mouse brain, Van Leuven concluded. The finding conflicts with reports of O-GlcNAcylated tau in P301L-JNPL3 mice, which express a slightly different form of mutant tau (see Yuzwa et al., 2012). The data suggest the drug acted on some other protein, Van Leuven said. About 250 proteins get O-GlcNAcylated in mouse brain, and he speculated that the culprit probably lies downstream of tau in the pathological cascade. He is now looking for the target, and believes it will provide insight into neurodegenerative pathways downstream of tau.—Madolyn Bowman Rogers.
This is Part 2 of a two-part series. See also Part 1.