One of the basic research flavors that wafted pungently through the 9th International Conference on Alzheimer’s Disease and Related Disorders last week in Philadelphia is the growing sense that intraneuronal Aβ may have more of a hand in Alzheimer’s disease than previously appreciated. No one doubts that it must exist; after all, APP cleavage occurs in intracellular membrane compartments. But the discovery in the early 1990s that neurons secrete Aβ, as indeed do cells in most tissues of the body, drew much attention to its extraneuronal forms. What’s more, intraneuronal Aβ is rarely detected, whereas extraneuronal Aβ plaques are there in plain sight. (Similarly, the field’s overwhelming focus for many years on plaques rather than smaller species, such as oligomers, owes much to the fact that plaques are visible and thus amenable to study, while oligomers are maddeningly elusive in vivo.) This meeting report summarizes some of the presentations on this invisible potential foe inside neurons.

A Mouse Whose Neurons Perish in Droves
A vexing shortcoming of most mouse lines based on the amyloid hypothesis has long been that they do not show the massive neurodegeneration that marks the human disease they are made to model. Explanations abound (“mice are different,” “they don’t live long enough,” “they lack a tau component”), and neurodegeneration has even been made to seem secondary as the field shifted its focus on synaptic dysfunction, which existing mouse models do display to a small degree. But this smacks a bit of changing the topic. Even after years of recording LTP deficits and subtle behavioral phenotypes in the existing lines, critics of the amyloid hypothesis persistently point to the conspicuous absence of neurodegeneration from most models. Laurent Pradier of Aventis Pharma in Paris, with Thomas Bayer at Saarland University in Homburg, Germany, and colleagues, have given amyloid aficionados what they need—but with a twist.

In a pre-meeting symposium sponsored by the Alzheimer Research Consortium, and again at the conference, Pradier and Bayer presented a mouse line that develops dramatic neurodegeneration in the hippocampus. The point that generated the buzz is that these mice accumulate Aβ42 extensively inside their neurons, and that their neurodegeneration appears tied to this early form of Aβ42, not to the extracellular amyloid pathology that also develops in the mice.

The Franco-German team placed two familial AD mutations into the mouse presenilin gene and bred the resulting PS1 knockin mouse to the transgenic APP751 mouse. These APP751/PS1KI mice produce almost exclusively Aβ42, the most aggregation-prone form of the peptide. Normally, Aβ42 is a minor product of APP processing, and in some other APP/PS mouse models the ratio of Aβ42 to total Aβ increases but remains below 0.5. In this new line, this ratio exceeds 0.85.

Remarkably, the mice begin losing large numbers of pyramidal neurons in the CA 1/2 region of the hippocampus at six months of age, and this happens in parenchymal areas that have no plaques. The scientists also detected Aβ inside such neurons beginning at two months of age. The intraneuronal Aβ42 appeared to be aggregated, as thioflavin S, a marker of amyloid, produced granular staining.

“We think the intraneuronal Aβ42 leads to the death of neurons, not the plaques,” Pradier said. This would make Alzheimer’s more similar to other neurodegenerative diseases, where the protein accumulates to toxic levels inside neurons, for example, Parkinson’s (a-synuclein), prion diseases, and Huntington’s. It would also create a renewed sense of urgency about clarifying how Aβ and tau interact.

Presenilin Dysfunction: Not a Fringe View Anymore
One of the intriguing tie-ins of this work is how it dovetails with another recent study suggesting that some human presenilin mutations exert their effect through a partial loss of function (Saura et al., 2004.) Pradier and colleagues found that the APP CTF fragment that is the substrate for γ-secretase accumulates in their APP/PS KI mice, suggesting that the enzyme complex is partially inhibited. The hypothesis is that presenilin does not function properly, but what function it does have shifts the ratio of Aβ forms toward a relative increase in Aβ42. This idea would also explain the odd finding that, at low doses, some γ-secretase inhibitors actually increase the production of Aβ42, while at higher doses they shut down the enzyme complex. With regard to γ-secretase inhibitor drugs, Pradier said his data suggest that one needs a subtle modulator that nudges the enzyme complex away from producing Aβ42. In the context of presenilin dysfunction, it is interesting to note Bart Dermaut and Samir Kumar-Singh’s presentations on a PS1 mutation that leads to a splicing error and causes a tauopathy with intraneuronal Aβ but no plaques (see ARF related Philadelphia story).

Broadly, all this fits with earlier observations by Iva Greenwald, Ralf Baumeister, Christian Haass, Bart de Strooper, and others. These groups over the years all have found that some FAD PS mutants cannot reconstitute γ-secretase activity in C. elegans, yeast, and other systems. If such mutations acted strictly by a gain of function, they should be able to. In his presentation, Mark Fortini noted that some AD-relevant PS1 mutations act through partial loss of function. Nikolaos Robakis detailed in a plenary lecture the consequences for signal transduction and transcription of a loss of presenilin function caused by some FAD mutations, though he focused on pathways independent of Aβ.
The presenilin loss-of-function idea appears to become more widely accepted, but exactly how it relates to a role for intracellular Aβ is less clear. Pradier noted that this loss of function became apparent in his latest mouse strain because he kept the expression of the knockin PS1 gene at physiological levels. Generally, human transgenes are overexpressed, which can mask a subtle loss of function.

Not a New Idea, Just an Unloved One
Pradier’s and Bayer’s study gives a boost to those in the field who for years have been making a case for the role of intracellular Aβ. As early as 1985, Colin Masters and Konrad Beyreuther had written that “the amyloid is first deposited in the neuron, and later in the extracellular space.” (See Masters et al., 1985.) Eight years later, Virginia Lee, of the University of Pennsylvania, Philadelphia, demonstrated intraneuronal Aβ (Wertkin et al., 1993). Yesterday at the conference, Lee suggested in a lecture that different pools of Aβ exist inside neurons. One gets made in the trans- Golgi network and en route to nerve terminals, where it is secreted, she said. A second pool, however, accumulates over time inside the neuron. Lee hypothesized that this Aβ is actually made by a different γ-secretase, not the known complex containing presenilin.

What brought intracellular Aβ to the fore in this mouse model? The whole trick lay in shifting Aβ production toward the 42 variety, Pradier said, because this produced so much intraneuronal Aβ42 that it began aggregating to the point where thioflavin S picked it up.

Researchers have on occasion reported intraneuronal Aβ in human AD cases but, generally speaking, detecting it has been difficult. In Philadelphia, Charles Duyckaerts of Inserm in Paris, demonstrated intraneuronal Aβ with electron microscopy. Many labs have tried to do this, but few were able to. This French group met success by adding the Aβ antibody to brain sections of APP- and APP/PS1-transgenic mice, then fixing the sections and applying the secondary antibody with gold particles afterwards (gold is visible in the EM). Duyckaerts found that the Aβ42 accumulated within multi-vesicular bodies inside neurons, which also contained the lysosomal marker cathepsin D. This confirms an earlier finding by Gunnar Gouras, whose previous description of EM-immunogold localization of intraneuronal Aβ in mouse and human brain showed that it accumulated with age in the membranes of multi-vesicular bodies (Takahashi et al., 2002; Takahashi et al., 2004; ARF Live Discussion). Pradier’s group, too, has repeatedly described intraneuronal Aβ in mice (e.g., Wirths et al., 2001). But, like Gouras’s, it was not thioflavin-positive, nor was there massive neurodegeneration.

Changiz Geula, of Beth Israel Deaconess Medical Center in Boston, was intrigued enough by this evolving topic to look for intraneuronal Aβ in the brains of a group of primates he has been studying carefully for years. He looked for intraneuronal Ab accumulation in aged monkey brains, non-demented human brains, and four AD brains, and compared that with young monkey and human brains. In Philadelphia, he told his colleagues that he indeed found Aβ accumulation in hippocampal pyramidal and non-pyramidal neurons of old rhesus monkeys. He also found it in basal forebrain cholinergic neurons, which die early in AD, but much less so in other brain areas. Young rhesus monkeys and young human brains did not have this staining, but the brains of the AD patients did.

Not all postmortem studies jibe with this, however. Jerzy Wegiel’s group at the New York State Institute for Basic Research in Developmental Disabilities, Staten Island, reported that intraneuronal Aβ detection in control, Down’s, and AD cases argues against a role for it in AD pathogenesis. These scientists found Aβ inside neurons of infants and young people as well as people with AD. They suggest that, per se, intraneuronal Aβ expression is not related to neuronal death or neurofibrillary pathology. Other scientists cautioned in general terms that it is notoriously difficult to ensure that an antibody thought to bind AAβ inside neurons doesn't also recognize the Abeta sequence within APP.

How Do the Neurons Die?
On this question, Pradier would only reveal that the process differs from apoptosis, as the researchers looked for caspase activation but did not see it. He pointed out, however, that this mouse can be used to test neuroprotective agents directly in vivo, which is not possible with standard models of amyloidosis.

And how does this relate to tau? After all, neurofibrillary pathology is tied to neurodegeneration. In his mice, hippocampal and cortical neurons die without tau pathology, but the mice constitute an aggressive and artificial model, Pradier cautioned. In Alzheimer’s disease, it is more likely that an age-related build-up of tau pathology inside neurons sensitizes them, and that added Aβ42 accumulation pushes them over the edge, Pradier speculated.

By this hypothesis, plaques would remain behind after neurons have died. In fact, the death of a neuron could release pre-aggregated Aβ, which could perhaps act as a seed to accelerate the aggregation of extracellular Aβ into plaques. Plaques would then trap other proteins into binding to them and exert secondary toxicities in their own right, such as distorting neurites, Pradier said.

The question of whether direct links tie together Aβ—both inside and outside of neurons—and tau is still a nebulous one. The vapors appeared to lift a bit with Frank LaFerla’s presentation on what happens in his group’s triple transgenic mice (Oddo et al., 2003) after they receive shots of anti-Aβ antibodies into their hippocampus. In brief, the scientists showed that following treatment, the extracellular amyloid deposits disappeared first, then intraneuronal Aβ diminished and, a few days later, tau pathology vanished, as well. A month after the injection, amyloid pathology reappeared, followed by tau pathology. To LaFerla’s mind, this relationship between amyloid and tau pathology means that amyloid-removing therapies might indeed clear both amyloid and tau pathologies, but only if given before tau pathology has matured. This was also the conclusion of a plenary lecture delivered by Karen Hsiao Ashe. LaFerla’s work is slated to appear as the cover article in Neuron on August 5, when Alzforum will report on it separately in greater detail.

There were yet more presentations in support of the idea that intraneuronal Aβ figures prominently in the development of Alzheimer’s. They include recent work on the arctic mutation presented by members of Lars Lannfelt’s team at the University of Uppsala in Sweden. This group had discovered a family with early onset AD in Northern Sweden, whose APP mutation resides inside the actual Aβ sequence and makes the peptide more prone to aggregate. It appeared to point to a role of Aβ oligomers and protofibrils in driving disease (Nilsberth et al., 2001). In Philadelphia, Lars Nilsson presented data suggesting that combined SwedishArcticAPP-transgenic mice accumulate Aβ42 and form protofibrils and fibrils inside neurons. And Charlotte Stenh, also in Lannfelt’s group, said that previous, and puzzling, reports of a decrease in Aβ42 with this mutation were wrong because arctic Aβ42 aggregates so quickly that traditional ELISA tests do not quantify it correctly. That is because ELISA tests do not disaggregate protofibrils to record a signal from each Aβ unit. Denaturing the protofibrils first and then using antibodies against monomeric Aβ units gives a more precise measurement, and this method revealed increased levels of this peptide both extra- and intraneuronally, Stenh said.

What, then, does Aβ do inside the neurons? On this question, Claudio Cuello, of McGill University in Montreal, presented new data in support of the idea that physiological levels of it play a role in synaptic plasticity via a modest activation of CREB-directed gene expression. A pathological increase in intraneuronal Aβ would cause abnormal phosphorylation patterns that end up dysregulating these pathways. Cuello described work in transgenic rats that express intraneuronal Aβ in pyramidal neurons of the hippocampus and neocortex, as well as cell-culture work (see, for example, Echevarria et al., 2004). His research suggests that accumulating Aβ causes an upregulation of ERK, and that this kinase then phosphorylates a number of proteins, including tau. In t his way, this study suggests a further link between Aβ and tau pathways. ERK frequently pops up in gene expression comparisons between APP-transgenic and wild-type mice.

Cuello also told the audience that when intraneuronal Aβ levels rose in the rats, CREB phosphorylation diminished, uncoupling it from its normal function in regulating the expression of certain genes. CREB-driven gene expression has been found to be important for learning and memory; Cuello’s rats tested poorly in the water maze. Exploring this idea further in PC12 cells, the scientists found that CREB-driven gene expression depended on Aβ levels, but not on AICD. (AICD has drawn interest for its possible role in AD-related gene expression. But in Philadelphia, Thomas Südhof presented data suggesting that AICD may not do so itself, but merely induce a conformational change in the protein Fe65, readying it for its own gene regulatory function.) Cuello identified the Rap1/B-raf/MEK/ERK2 pathway as key to Aβ function, and said that it can become dysregulated by high concentrations of both intraneuronal and extraneuronal forms of the offending peptide.

To be sure, a great burden of proof remains before the field at large will be persuaded that intraneuronal Aβ accumulation damages neurons in early Alzheimer’s. But the idea is not languishing in obscurity any more. “Back in 1998, when I first presented data on intraneuronal Aβ at this conference, barely anyone paid attention,” Gouras remembered. At least that is no longer the problem.—Gabrielle Strobel.


  1. The idea of intraneuronal Aβ contributing to AD pathology certainly
    appears to be generating momentum, and this was evident at last
    month’s conference. Compelling evidence from new and established
    Aβ-related transgenic mouse models demonstrated that intraneuronal
    Aβ42 is an early event which precedes, and appears to correlate
    with, subsequent neuronal death. However, conflicting findings on
    postmortem AD brains on whether intraneuronal Aβ is an early event
    in AD, and is correlative with neuronal toxicity, needs to be resolved.

    The evidence presented from the aforementioned transgenic mouse studies
    gives rise to a number of important questions that may have significant
    implications in understanding AD pathology. The first question that
    comes to mind is, what is the site and mechanism of toxicity induced by
    intraneuronal Aβ? And how does this differ from toxicity induced by
    the extracellular version? In addition, with increasing evidence for
    Aβ acting upstream of neurofibrillary tangle formation, it will be
    interesting to examine whether the neurons containing intracellular
    Aβ aggregates are prone to neurofibrillary tangle formation, as well.
    Despite the known spatial disparity between amyloid plaques and
    neurofibrillary tangles, the paradigm for intracellular Aβ
    pathology does not preclude the notion that amyloid and neurofibrillary
    pathology may exist in one and the same neuron.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Philadelphia: Dementia Types Just Won’t Fit into Little Boxes

Paper Citations

  1. . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.
  2. . Neuronal origin of a cerebral amyloid: neurofibrillary tangles of Alzheimer's disease contain the same protein as the amyloid of plaque cores and blood vessels. EMBO J. 1985 Nov;4(11):2757-63. PubMed.
  3. . Human neurons derived from a teratocarcinoma cell line express solely the 695-amino acid amyloid precursor protein and produce intracellular beta-amyloid or A4 peptides. Proc Natl Acad Sci U S A. 1993 Oct 15;90(20):9513-7. PubMed.
  4. . Intraneuronal Alzheimer abeta42 accumulates in multivesicular bodies and is associated with synaptic pathology. Am J Pathol. 2002 Nov;161(5):1869-79. PubMed.
  5. . Oligomerization of Alzheimer's beta-amyloid within processes and synapses of cultured neurons and brain. J Neurosci. 2004 Apr 7;24(14):3592-9. PubMed.
  6. . Intraneuronal Abeta accumulation precedes plaque formation in beta-amyloid precursor protein and presenilin-1 double-transgenic mice. Neurosci Lett. 2001 Jun 22;306(1-2):116-20. PubMed.
  7. . 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.
  8. . The 'Arctic' APP mutation (E693G) causes Alzheimer's disease by enhanced Abeta protofibril formation. Nat Neurosci. 2001 Sep;4(9):887-93. PubMed.
  9. . Rat transgenic models with a phenotype of intracellular Abeta accumulation in hippocampus and cortex. J Alzheimers Dis. 2004 Jun;6(3):209-19. PubMed.

Other Citations

  1. ARF Live Discussion

External Citations

  1. Alzheimer Research Consortium

Further Reading