William Honer of the University of British Columbia, Vancouver, reviewed the existing literature on synaptic pathology in Alzheimer’s as viewed from the perspective of presynaptic markers. As usual in this field, the findings are contradictory. While 10 years ago, the relationship between the levels of the marker synaptophysin and cognitive impairment seemed clear, today the picture is more mixed, Honer said. He asked the following questions:
- Are all presynaptic proteins equally affected in Alzheimer’s? Six studies comparing the same markers (mostly synaptophysin, syntaxin, SNAP-25, synaptotagmin, VAMP) in different brain areas suggest that, no, different proteins go down in concentration at different times. He cautioned, however, that as a rule, existing research on these proteins remains far less sophisticated than that done on tau, for example.
- Are all brain regions equally affected? As expected, the hippocampus, temporal and frontal cortex are most affected. Eighteen studies have confirmed this.
- What is the relationship with pathological markers? In the perforant pathway, the correlation of synaptophysin with plaque number is weak, with tangle number a little stronger. Of 21 studies on the subject using other presynaptic markers, most find the same result. Some studies relate a decline in presynaptic proteins to soluble Aβ, glutamate transporter loss, and microglial activation.
- What is the relationship with cognitive impairment? Most of the 16 existing studies on this question found a link between cognitive testing and loss of presynaptic proteins. However, the simplicity here deceives. In mild dementia, presynaptic proteins tend to be increased relative to control, reflecting, perhaps, a temporary compensatory effort by the brain. In other mild cases, Mini Mental State Exam scores did not correlate with frontal cortex synaptophysin.
- What is the relationship with phase of illness? A large study (Tiraboschi et al., 2000) looking at synaptophysin showed a small, insignificant difference in the mildest cases. Choline acetyl transferase levels did not differ statistically in mild cases, either. (See also related news.) Honer said that his group, as well, did not find a robust decrease in synaptophysin, syntaxin, or SNAP-25 levels until the more severe stages. In all, 11 studies on the question did not find consistent results.
Honer cautioned that, while this last finding might seem to weaken the notion that synapses die very early on in AD, it might also reflect inconsistency in the ways mild cases are assessed and diagnosed. As an interesting aside, Honer mentioned ongoing work showing that antibodies against presynaptic proteins can visualize plaques almost as well as standard methods, such as thioflavin S. VAMP, especially, is highly present in plaques. Finally, Honer threw out another bit of food for thought: In APP23 transgenic mice, the concentration of most presynaptic proteins assessed showed a temporary increase at 12 months, followed by a decrease below, but not far below, control levels. All of this fits the speculative notion that the pathology comes first, and sometime later, once synaptic damage begins, cognition suffers. Most people do not get diagnosed until this has progressed quite far.
Valina Dawson of Johns Hopkins University, Baltimore, Maryland, laid out a series of studies aimed at answering the question of how specific populations of neurons die. She suspected death pathways other than the caspase pathways were at work in neurodegeneration, in part because classic apoptotic players were first described in cell types that are programmed to turn over frequently. Neurons, by contrast, are designed to stay alive through a person’s lifespan and get replaced very slowly, if at all.
Dawson described elucidation of the following death pathway: Glutamate overly excites NMDR receptors; this activates neuronal nitric oxide synthase. Nitric oxide diffuses and, together with superoxide anion produced in mitochondria, leads to the formation of peroxinitrite radicals, damaging DNA and thus triggering expression of the nuclear protein PARP. This enzyme somehow causes the release of apoptosis-inducing factor (AIF) from the outside of mitochondria, where it is normally anchored. AIF then translocates to the nucleus and induces nuclear condensation. AIF, then, is the endpoint of a caspase-independent mode of cell death. This new pathway is an alternative to death by caspase-cytochrome C, Dawson pointed out.
She further described work using the MPTP model of Parkinson’s to suggest that this synapse-mediated mechanism of cell death might indeed be active in disease, and suggested a focus on developing therapeutic agents that bind AIF and redirect it away from the nucleus and toward proteasomal degradation. (See also related news; related news.)
Roberto Malinow, at Cold Spring Harbor Laboratory in Long Island, New York, presented new work probing a potential normal function of Aβ. Could it be that the peptide, once released from the synapse, acts as a negative feedback signal to keep neuronal hyperactivity in check? This might normally occur under conditions of high activity, but become dysregulated when APP or Aβ concentrations rise, Malinow said. His student Flavio Kamenetz prepared organotypic slices of transgenic mice carrying the APP Swedish mutation, cultured it with high or low activity, and then analyzed the Aβ content with an ELISA. He found that a decrease of activity with tetrodotoxin or a benzodiazepine led to a decrease of Aβ content by half; an increase of activity with the substance PTX correspondingly increased Aβ in medium.
By analyzing the different APP cleavage products, Kamenetz found that the BACE reaction is the controlling step. Then the scientists expressed GFP-APP delivered to neurons in the slice by injection of a viral construct and recorded the electrophysiological output of synapses from single cells to ask how extra APP affected their activity. Seen in the dendritic spines, the APP depressed excitatory transmission of AMPA and NMDA receptors, though the GABA transmission remained unaffected, Malinow said. Expressing a form of APP with a mutation in the Aβ sequence that prevents formation of the peptide abrogated the depression, as did γ-secretase inhibitors. This indicates that Aβ, not APP, mediates the synaptic depression. Other APP cleavage products were not necessary to produce the effect.
The interaction is a bit circular: Aβ leads to synaptic depression, and synaptic activity leads to higher Aβ secretion, Malinow reported, probably via an effect on BACE. Kamenetz also recorded from individual neurons surrounded by infected, Aβ-releasing cells and compared its transmission with that from a neuron surrounded by noninfected cells. Lo and behold! Transmission was down in the noninfected region, indicating that secreted Aβ acts on neighboring cells.
In hippocampal slices of normal, non-transgenic mice, this phenomenon is visible only after intense stimulation. Under conditions of strong LTP activation, Aβ depresses transmission somewhat and presenilin-inhibitors increase transmission, Malinow said.
How could this play out in AD? Malinow said that two speculative scenarios came to mind. An unknown factor causes an increase in the amount of Aβ, and the synapses disappear once they have been repressed long enough. Again, use it or lose it. Alternatively, synapses could burn out and crash. Conditions of prolonged intense excitation would induce intense APP processing, and the Aβ effect might somehow turn into a positive feedback signal. David Holtzman commented that this latter notion would fit well with studies finding that synaptic markers go early in transgenic mice and also are not decreased overall in early cognitive impairment. It also fits with clinical data that people who die with pathology but without dementia symptoms do not yet have synaptic changes.
Lennart Mucke, University of California, San Francisco, focused on transgenic mice studies suggesting how Aβ accumulation (but not plaques), aging, and ApoE isoform might lead to cognitive dysfunction. His lab has a paper coming out in the November 15 J. Neuroscience on the topic, which Alzforum will summarize in this space, and other research on calcium regulation and behavioral changes in transgenic mice will appear later.
David Holtzman of Washington University in St. Louis, Missouri, asked how Aβ could possibly cause LOAD when there is no obvious overproduction of the peptide in this vast majority of cases. PDAPP mice allow him to address this question, since they also do not have a longstanding gradual buildup of Aβ early in life. Aβ increases only when plaques start forming. "Why does Aβ convert to forms that are toxic?" Holtzman asked. He then reviewed work on factors influencing Aβ fibrillization that was recently summarized on Alzforum (see news story; news story).
In addition, Holtzman presented new ways to study the dystrophic, swollen neurites that occur around amyloid plaques of APP-transgenic mice. Holtzman crossed with PDAPP mice a strain of YFP-transgenic mice that express yellow fluorescent protein throughout the neuronal cytosol, reaching into distal neurite tips. He found that the mice have numerous swollen neurites, but only where there are mature plaques. Holtzman emphasized how widespread the occurrence of swollen neurites in an areas containing amyloid, suggesting that neuritic dystrophy, i.e., synaptic damage in still-living neurons, is much more extensive than previously thought. Preliminary studies of organotypic slice cultures of these mice indicate that this method can visualize dendrites, dendritic spines, and axons, and observe them over a period of several days.
Frank LaFerla of the University of California, Irvine, gave a whirlwind tour of the molecular, pathological, and electrophysiological characterization of his new triple-transgenic mouse model of AD. LaFerla tried to recapitulate more of the AD pathology than do current models, and do so with a method that avoids the confounding effects that can occur after breeding strains from different backgrounds. Instead of crossing strains, LaFerla took a mutant human PS1 knockin mouse developed by Mark Mattson and injected into its single-cell embryos both a mutant human tau and a mutant human APP transgene. Luckily, both transgenes inserted into the same site, enabling the mice to breed as if they were a single strain, said LaFerla.
Tau and APP are expressed in the CNS, mostly in AD-vulnerable regions, and in the spinal cord, and Aβ accumulates with age, LaFerla said. Aβ accumulation inside neurons at six months of age is the first detectable sign of pathology, LaFerla added, saying this supports a role of cytoplasmic Aβ early in the disease. By nine months, extracellular deposits show up in cortex and hippocampus. At six months, tau is not visible with immunoreactivity (not just tangles; there is no tau pathology at all), but by 12 months, initial signs of tau pathology appear, and by 15 months, the mice have extensive tau. "We believe that Aβ accumulation is the initiating trigger for sporadic and familial AD. We think Aβ is upstream of tau in the pathologic cascade," LaFerla said.
Electrophysiologic recording from the CA1 region of the hippocampus showed deficits in synaptic transmission and in LTP even at six months, when there was no extensive pathology yet. At one month of age, these parameters were normal. This leads LaFerla to propose that synaptic deficits occur very early in this mouse.
What could be the molecular mechanisms underlying this? LaFerla described experiments to tease apart the effect of presenilin mutations on Aβ generation from their other effect on disrupting calcium homeostasis (see related news story). LaFerla described some unpublished data suggesting that the intracellular tail of APP, AICD, affects the transcription of the calcium-related gene SERCA, which further increases calcium levels in ER stores already overfilled as a consequence of presenilin. This might further increase Aβ levels. LaFerla ascribes the age-dependence of LOAD to the waning activity of Aβ-degrading enzymes. (LaFerla; related news).
Virginia Lee of the University of Pennsylvania, Philadelphia, urged the audience to consider a link between tau and α-synuclein (see related news). She also reported that an α-synuclein-transgenic mouse recently developed in her lab (see related news) unexpectedly showed tau pathology as well, suggesting that the two proteins might interact and facilitate each other’s pathology.
Robert Edwards of University of California, San Francisco, reported that a small proportion of α-synuclein occurs in lipid rafts, specialized membrane regions that are high in cholesterol and are also known to harbor APP and Aβ. One of the two α-synuclein mutations known to cause familial, early onset Parkinson’s disease, A30P, abrogates this localization.
In the final talk, Don Price of Johns Hopkins University pulled together recent findings in his division on BACE around the theme of how Aβ secretion from synapses could damage synapses. He suggested a model where Aβ deposits around synapses, the presynaptic side detaches and swollen neurites form, glia proliferate and move in, and the axon degenerates. One set of findings supporting this notion, Price said, was that APP, its cleaving enzymes, and Aβ all are found in and around terminals and neurites. Price, with Vassilis Koliatsos, and Sam Sisodia’s group each have a paper coming out on this topic in J. Neuroscience on November 15, which Alzforum will cover in this space.
Price summarized knockout studies of the components of APP processing. Mice lacking nicastrin or presenilin 1 die in utero or shortly after birth, respectively, but BACE knockouts are healthy and breed well. Even when crossed with APP/PS1 transgenic mice, BACE knockout mice produce no Aβ at all, Price said. Preliminary evidence shows that these mice perform as well in the Morris water maze as do wild-type mice, even though they massively overproduce APP and APP/PS-1 mice that do have BACE perform poorly, Price said. This adds to existing evidence making a case for the development of BACE inhibitors, which are being pursued in most major pharmaceutical companies, said Price. He did not reveal any information on their status.
Price said that the expression levels of the APP-processing enzymes reveal why Alzheimer’s is a brain disease even though all cells make APP. α-secretase, which prevents Aβ generation by cutting in the middle of its sequence, is barely expressed in neurons but abundant in peripheral organs, as is BACE 2, which also makes an amyloid-preventing cut. BACE1, however, is abundant in brain but few peripheral organs checked so far. (The pancreas expressed BACE 1 but there, alternative splicing creates an isoform that cannot create Aβ.) See news story; news story; news story; and news story.
The symposium ended with the drawing of a bottle of champagne as the prize for filling in an evaluation. Allen Butterfield of the University of Kentucky, Lexington, was the lucky winner. The only pet peeve about the symposium may have been that after two days of unrelenting excitation of the audience’s collective synapses, everyone could have used a rejuvenating glass of fizz. All talks in this event were interesting. Those that did not get much space in this on-site summary were either covered on Alzforum previously, or were presented when the writer’s brain, regrettably, had temporarily filled up! Since there was no time for fact-checking, I especially invite all speakers to send corrections to firstname.lastname@example.org.—Gabrielle Strobel
- Revise Mechanism for Alzheimer's Drug? Acetylcholine Increased in Earliest Stage of Disease
- Oxidative Stress Triggers Neuronal Cell-Cycle Reentry
- In <I>C. Elegans</I>, Aspartyl Proteases and Calpain, Not Caspases, Play the Grim Reaper
- Ocean Waves Stimulate Brain Waves: Meeting Report from Caribbean
- Stockholm: Clusterin (or Beware the Evil Chaperone!)
- Society for Neuroscience Annual Meeting: The Calcium-Presenilin Connection
- Does the Intracellular AβPP Fragment Regulate Calcium?
- Meeting Report: American Society for Neurochemistry (22-26 June 2002)
- Presenting: A Not-Quite-Parkinson's Mouse Model
- BACE Above Base in Alzheimer’s Patients
- More on BACE1, the New Darling of Alzheimer's Disease Drug Developers
- Another Not-Quite-Parkinson's Model
- β-Secretase a Clean Therapeutic Target
No Available Further Reading
- Kamenetz F, Tomita T, Hsieh H, Seabrook G, Borchelt D, Iwatsubo T, Sisodia S, Malinow R. APP processing and synaptic function. Neuron. 2003 Mar 27;37(6):925-37. PubMed.
- Oddo S, Caccamo A, Shepherd JD, Murphy MP, Golde TE, Kayed R, Metherate R, Mattson MP, Akbari Y, LaFerla FM. Triple-transgenic model of Alzheimer's disease with plaques and tangles: intracellular Abeta and synaptic dysfunction. Neuron. 2003 Jul 31;39(3):409-21. PubMed.