. Intraneuronal APP, not free Aβ peptides in 3xTg-AD mice: implications for tau versus Aβ-mediated Alzheimer neurodegeneration. J Neurosci. 2011 May 25;31(21):7691-9. PubMed. RETRACTED


Please login to recommend the paper.


  1. Winton et al. use histochemistry and a variety of antibodies to conclude that the 3xTg-AD mice lack intracellular Aβ. This study also uses a genetic approach to conclude that Aβ does not modulate tau pathology in 3xTg-AD mice. We find that the data presented do not justify the conclusions. In the following discussion, we point out that the authors’ own data demonstrate the presence of intraneuronal Aβ in the 3xTg-AD mice. We and many others have reported multiple studies over the past decade that argue against these conclusions; however, Winton et al. failed to discuss these studies.

    We believe the clear discrepancies between Winton et al. and previous publications likely result from the relative affinities of the various neoepitope antibodies used. To try to detect intracellular Aβ, Winton et al, rely heavily on two neoepitope antibodies, BA27 and BC05. These antibodies are most commonly used for biochemical Aβ ELISA measurements, but it is unclear how effective they are for immunofluorescence analysis, as only low-powered images are provided in the paper. It is well established that antibody sensitivity and specificity can vary in biochemical versus histological uses. Thus, the lack of staining with these two antibodies proves little, especially when no positive control is shown in the fluorescent images. In contrast to these data, we and other groups have used different Aβ42-specific antibodies that clearly demonstrate intraneuronal Aβ in the 3xTg-AD mice (Oddo et al., 2006, Fig. 3), as well as the presence of intracellular oligomers and ADDLs, using multiple antibodies. Importantly, numerous laboratories have independently replicated our findings (to cite a few recent ones, see Khandelwal et al., 2011; Sudol et al., 2009; España et al., 2010). Most recently, one study demonstrated intracellular Aβ by Western blot in 3xTg-AD prior to plaque formation (Himeno et al., 2011, Fig. 3).

    Notably, we detect most intraneuronal Aβ within hippocampal CA1 neurons; Winton et al. focus mainly on cortical sections for Aβ, and hippocampal regions for tau analysis. Perhaps the most confusing piece of data presented by Winton et al. is depicted in Figure 5G-I showing 5685 and 4G8 immunofluorescence of 3xTg mice plus a merged image. Only low magnification images are presented, which seem inadequate for making a conclusion about intraneuronal staining. However, after zooming in on these images, one can clearly see cells and portions of cells in Fig. 5I that label with 4G8 only. The double-label presented depicts C-terminal APP labeling in red and Aβ/APP labeling in green. If all of the immunolabeling were the result of APP, one would expect 100 percent colocalization (all yellow). Instead, significant portions of the labeling appear green, indicating Aβ (Fig. 5I). This staining also appears reduced in the 3xTg-AD/BACE-/- mice, as one would expect (Fig. 5J-L); however, no quantification is provided. Surprisingly, the red APP immunolabeling appears decreased in the BACE knockout mice (J vs. G), despite biochemical data suggesting APP is increased in BACE knockouts in line with other groups’ findings.

    We have never asserted that there is no neuronal APP expression in a mouse that was specifically designed to overexpress APP in neurons. Rather, we previously showed, and assert here again, that a significant portion, although not all, of the immunolabeling in 3xTg-AD mice represents Aβ. We used similar co-labeling techniques in the past to demonstrate intraneuronal staining in the 3xTg-AD mice (McAlpine et al., 2009).

    The lack of immunoreactivity seen in the study by Winton et al. may also be due to the methodology employed, including fixation and pretreatments (we use paraformaldehyde fixation with formic acid pretreatments). Although formic acid and 4 percent PFA fixation are briefly mentioned, the methods section implies that the majority of the mice were fixed with 10 percent formalin, a far stronger fixation approach that is known to reduce antigen sensitivity. It is unclear which, if any, images are derived from formic acid-treated sections.

    In summary, it is correct that full-length APP can be detected in the 3xTg-AD mice. However, the data provided in this study does not convince us that no intraneuronal Aβ is present in 3xTg-AD mice.

    The second conclusion made by Winton et al. is that tau pathology develops independently of the presence of Aβ. This is based on Fig. 7 displaying that the 3xTg-AD/BACE-/- mice show similar somatodendritic accumulation of tau to 3xTg-AD mice. To our eyes, this figure actually seems to show enhanced AT8 immunoreactivity in 3xTg-AD/BACE-/- mice compared to the 3xTg-AD mice, although no statement is made to this effect. Unfortunately, the paper offers no quantification or time course to fully investigate tau pathology in these mice lacking Aβ. We are puzzled that a lab that established methods for the biochemical analysis of tau provides no biochemical support for this conclusion. As homogenates were prepared for these mice for ELISA measurements in Fig. 3, it would surely have been critical to perform biochemical confirmation of these immunohistochemical images. We asked for samples from the 3xTg-AD/BACE-/-, and were disappointed to find that the mice no longer exist and the lead author has left the lab (J. Trojanowski and V. Lee, personal communication).

    We, and many others, have published numerous studies highlighting the influence of Aβ on tau pathology, including in the 3xTg-AD mice. For example, we showed that immunotherapy against Aβ removes somatodendritic human tau (Oddo et al., 2004), a finding that has been replicated in other mouse models (Wilcock et al., 2009) and human Aβ immunotherapy trials (Boche et al., 2010).

    We also performed multiple genetic manipulations of the 3xTg-AD mice to alter Aβ production, in the same fashion as the generation of these 3xTg-AD/BACE-/- mice. We have removed the presenilin mutation, and shown that this reduces Aβ42 and leads to a reduction in tau phosphorylation and accumulation (Oddo et al., 2008). Consistent with this, we showed that modulating the ApoE allele in the 3xTg-AD mice alters Aβ accumulation and also provided quantitative biochemical evidence that it alters tau phosphorylation and accumulation (Oddo et al., 2009). Thus, none of these studies support the conclusion that tau pathology develops independently of Aβ in the 3xTg-AD mice.

    Perhaps the difference in findings is due to the analysis performed. Detailed biochemical analysis of tau pathology in the 3xTg-AD/BACE-/- as a function of age compared to the 3xTg-AD mice would help to clarify the issue. It is also possible that differences in strain background between 3xTg-AD and the 3xTg-AD/BACE-/- mice were not accounted for. We previously collaborated with one of the authors on this manuscript and detailed the importance of background strain on AD-related pathologies, as have other groups. Notably, we are preparing a manuscript at the moment in which we have produced a novel transgenic mouse that overexpresses both APP and human wild-type tau transgenes. We crossed this to the same BACE knockout mouse used by Winton et al., and detect clear biochemical reductions in hyperphosphorylated tau in even just heterozygous BACE knockout progeny. These findings again support a role for Aβ in facilitating tau pathology, as do many other independent studies.

    In summary, we disagree with the conclusions presented in this study.


    . A dynamic relationship between intracellular and extracellular pools of Abeta. Am J Pathol. 2006 Jan;168(1):184-94. PubMed.

    . Parkin mediates beclin-dependent autophagic clearance of defective mitochondria and ubiquitinated Abeta in AD models. Hum Mol Genet. 2011 Jun 1;20(11):2091-102. PubMed.

    . Generating differentially targeted amyloid-beta specific intrabodies as a passive vaccination strategy for Alzheimer's disease. Mol Ther. 2009 Dec;17(12):2031-40. PubMed.

    . Intraneuronal beta-amyloid accumulation in the amygdala enhances fear and anxiety in Alzheimer's disease transgenic mice. Biol Psychiatry. 2010 Mar 15;67(6):513-21. PubMed.

    . Apomorphine treatment in Alzheimer mice promoting amyloid-β degradation. Ann Neurol. 2011 Feb;69(2):248-56. PubMed.

    . Inhibition of soluble TNF signaling in a mouse model of Alzheimer's disease prevents pre-plaque amyloid-associated neuropathology. Neurobiol Dis. 2009 Apr;34(1):163-77. PubMed.

    . Abeta immunotherapy leads to clearance of early, but not late, hyperphosphorylated tau aggregates via the proteasome. Neuron. 2004 Aug 5;43(3):321-32. PubMed.

    . Vascular amyloid alters astrocytic water and potassium channels in mouse models and humans with Alzheimer's disease. Neuroscience. 2009 Mar 31;159(3):1055-69. Epub 2009 Jan 19 PubMed.

    . Reduction of aggregated Tau in neuronal processes but not in the cell bodies after Abeta42 immunisation in Alzheimer's disease. Acta Neuropathol. 2010 Jul;120(1):13-20. PubMed.

    . Blocking Abeta42 accumulation delays the onset and progression of tau pathology via the C terminus of heat shock protein70-interacting protein: a mechanistic link between Abeta and tau pathology. J Neurosci. 2008 Nov 19;28(47):12163-75. PubMed.

    . Genetically altering Abeta distribution from the brain to the vasculature ameliorates tau pathology. Brain Pathol. 2009 Jul;19(3):421-30. PubMed.

  2. Virginia Lee and coworkers have convincingly demonstrated that the perinuclear, intracellular Aβ immunoreactivity observed in a subset of neurons in the 3xTg-AD mouse is predominantly, if not exclusively, APP and not Aβ. Moreover, they demonstrate that this intracellular immunoreactivity does not require BACE activity, indicating that it does not come from the same pathway that produces secreted Aβ.

    Does this mean that intracellular accumulation of APP is insignificant and we can disregard it? As the authors state, these findings “warrant further study as to the role of aberrant APP accumulation.” The accumulation of APP is abnormal because it is age related and it only takes place in a subset of neurons. Similar intraneuronal accumulation of Aβ and APP immunoreactivity is also observed in humans, where it appears to be associated with the earliest signs of AD pathology, while it declines with advancing AD pathology and is largely absent in non-demented individuals that lack AD pathology (1). This suggests that intracellular accumulation of APP may be one of the earliest events of AD pathology. In addition, even if the intracellular Aβ immunoreactivity is APP, it appears to be APP that is misfolded in the same conformation as aggregated Aβ because it reacts with conformation-dependent antibodies that are specific for Aβ aggregates, but not Aβ monomers or APP, such as M16 (2,3) or OC or NU1 (4).

    A possible significance of the accumulation of intracellular APP is that it may be the result of seeding by the uptake of extracellular Aβ oligomers. We have previously reported that incubation of APP-expressing cells with Aβ oligomers results in the intracellular accumulation of APP and amyloidogenic fragments of APP in the detergent insoluble fraction of the cell (2). A fraction of this insoluble APP may be slowly converted to insoluble, intracellular Aβ containing “ragged” amino termini by non-specific degradation of the protease-sensitive parts of the insoluble APP (5). If the intracellular accumulation of APP is an early event in AD pathogenesis, it may provide a facile explanation for the recent failure of γ- secretase inhibitors in clinical trials. In those trials, cognitive decline was accelerated in the treated group, possibly because γ-secretase inhibition causes the accumulation of amyloidogenic APP fragments inside the cell. These results can also be explained by the fact that γ-secretase has many other substrates that would also be affected by γ-secretase inhibition; however, a pathogenic role for APP CTF accumulation is parsimonious.


    . Intraneuronal beta-amyloid accumulation and synapse pathology in Alzheimer's disease. Acta Neuropathol. 2010 May;119(5):523-41. PubMed.

    . Intracellular A beta 1-42 aggregates stimulate the accumulation of stable, insoluble amyloidogenic fragments of the amyloid precursor protein in transfected cells. J Biol Chem. 1995 Jun 16;270(24):14786-92. PubMed.

    . 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.

    . Intracellular Aβ-oligomers and early inflammation in a model of Alzheimer's disease. Neurobiol Aging. 2011 Mar 15; PubMed.

    . Intracellular accumulation of insoluble, newly synthesized abetan-42 in amyloid precursor protein-transfected cells that have been treated with Abeta1-42. J Biol Chem. 1999 Jul 16;274(29):20650-6. PubMed.

  3. As mentioned in the discussion, we and others had already noted that Aβ peptides were not a major component of the intraneuronal 6E10 immunoreactivity originally reported by the LaFerla lab in the paper by Kitazawa et al., where they showed this was accelerated by chronic systemic inflammation (Kitazawa et al., 2005). We used Dennis Selkoe's C9 anti-APP antibody to distinguish between Aβ and full-length APP/APP-derived fragments, and concluded that the species which accumulates intraneuronally in 3xTg-AD mice exposed to chronic systemic inflammogen (LPS) was likely to be β-CTF. Inhibition of TNF signaling appears to be driving the process (McAlpine et al., 2009). Our biochemical and immunohistological analyses did not reveal robust accumulation of Aβ peptides in 3xTg-AD mice under basal conditions or in response to chronic systemic inflammation. Still, those studies suggested that it may be feasible to selectively target soluble TNF therapeutically to prevent dysregulated APP turnover and/or transport, and perhaps β-CTF accumulation, induced by chronic neuroinflammation. The significance for the AD field is that plenty of evidence indicates that species other than intraneuronal Aβ can exert neurotoxicity, and their formation may be modulated by inflammatory processes.


    . Lipopolysaccharide-induced inflammation exacerbates tau pathology by a cyclin-dependent kinase 5-mediated pathway in a transgenic model of Alzheimer's disease. J Neurosci. 2005 Sep 28;25(39):8843-53. PubMed.

    . Inhibition of soluble TNF signaling in a mouse model of Alzheimer's disease prevents pre-plaque amyloid-associated neuropathology. Neurobiol Dis. 2009 Apr;34(1):163-77. PubMed.

  4. Like Frank LaFerla and Salvo Oddo, I have concerns with the findings and conclusions of Winton and colleagues. Their data clearly demonstrate that several APP and Aβ antibodies share similar patterns of intraneuronal immunoreactivity that persist even in a BACE-/- background that precludes Aβ formation. Although these findings provide compelling evidence that intraneuronal Aβ immunoreactivity in 3xTg-AD mice may represent APP species, there are also a few issues which argue against this conclusion.

    In some of my lab’s prior work with 3xTg-AD mice (Rosario et al., 2006; Carroll et al., 2007), we compared the patterns of immunoreactivity for antibodies against Aβ and APP C-terminal fragments to exclude the possibility that the observed intraneuronal Aβ immunostaining largely reflected APP species. We found that Aβ and APP-CTF exhibit qualitatively distinct immunoreactivities that differ both in cellular localization and temporal expression. APP-CTF immunoreactivity in 3xTg-AD mice is localized along the periphery of the soma and is clearly extranuclear, such that large, unstained nuclear regions are visible within neurons. Further, as expected with an APP transgene, the magnitude of APP-CTF expression in 3xTg-AD mice does not exhibit significant change with increasing age and is much higher than that observed in non-Tg mice. In contrast to the APP-CTF pattern, we observed that intraneuronal Aβ immunoreactivity in 3xTg-AD mice has a characteristic punctate appearance that is often present throughout the soma. Further, intraneuronal Aβ immunoreactivity exhibits robust age-related changes, becoming apparent at approximately age three months and strongly increasing at least through age 12 months.

    Examination of the figures in the paper by Winton et al. shows that their intraneuronal APP immunolabeling is very similar to our observations, with an even, extranuclear appearance that does not obviously increase over time (five to seven months vs. 10-12 months). Notably, their intraneuronal Aβ immunolabeling is unlike the pattern we observe. It is not punctate in appearance and does not increase with age. Rather, their Aβ immunostaining exhibits either an absence of intraneuronal labeling or intraneuronal labeling that matches the APP-CTF pattern. Thus, it appears that Winton et al. may not be observing the intraneuronal Aβ as reported by LaFerla, my lab, and several other groups. As suggested by LaFerla’s comments, one plausible explanation for the discordant findings of Winton et al. is that their Aβ antibodies and/or immunohistochemical protocol were not optimized for detecting intraneuronal Aβ.


    . Androgens regulate the development of neuropathology in a triple transgenic mouse model of Alzheimer's disease. J Neurosci. 2006 Dec 20;26(51):13384-9. PubMed.

    . Progesterone and estrogen regulate Alzheimer-like neuropathology in female 3xTg-AD mice. J Neurosci. 2007 Nov 28;27(48):13357-65. PubMed.

  5. Winton et al. demonstrated that the 3xTg mouse model suggests that APP accumulates in an age-dependent manner with no signs for intraneuronal Aβ up to the age of 12 months. In previous publications by the LaFerla group (1) and others, the 3xTg-AD mouse has been used as a model with early intraneuronal Aβ pathology. There has been always an issue whether epitopes of diverse Aβ antibodies also cross-react with its parent molecule APP. For example, Aho et al. demonstrated that often it is APP rather than Aβ that is detected intracellularly when using antibodies like 4G8, 6E10, and 82E1 clones that are unable to distinguish Aβ from APP (2). Unfortunately, these antibodies have been widely used in many studies, likely due to their high sensitivity in tissue sections.

    This issue has been a major concern also for us, as we have been one of the solicitors for the “intraneuronal Aβ hypothesis” (3). Therefore, we have used different approaches during the last years to analyze whether intraneuronal Aβ is a risk factor for AD.

    1. In the APP/PS1KI mouse model, there is early and robust accumulation of diverse intraneuronal Aβ peptides in CA1, frontal cortex, and some cholinergic nuclei. All these brain regions show age-dependent neuron loss between 25 and 50 percent, whereas regions with a comparable plaque load but without intraneuronal Aβ accumulation never develop neuron loss (4-5).

    2. In the 5xFAD model, the same is true for cortical layer 5 (6).

    3. The TBA42 model exclusively expresses N-truncated Aβ3-42 (without APP), and develops an age-dependent neurological phenotype and severe Purkinje cell loss following intracellular Aβ accumulation (7).

    4. Recently, we identified low molecular Aβ pE3 oligomers that can be detected by 9D5, a novel mouse monoclonal antibody against a conformational neoepitope (8). Importantly, we identified 9D5 by screening human AD brain tissue for a specific staining pattern, i.e., strong intraneuronal and only minor plaque staining. The antibody 9D5 did not cross-react with Aβ1-42 monomers, dimers, or other higher molecular structured aggregates, indicating that these oligomers present a unique and novel neoepitope. The therapeutic potential of 9D5 was demonstrated in passively immunized 5xFAD mice as plaque load, and Aβ levels were reduced and behavioral deficits normalized.

    We believe that these arguments support the “intraneuronal Aβ hypothesis.” Although most of the data have been obtained in transgenic mouse models, this hypothesis can be principally translated into human pathology, as the example of the 9D5 antibody has demonstrated.


    . 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.

    . Immunohistochemical visualization of amyloid-beta protein precursor and amyloid-beta in extra- and intracellular compartments in the human brain. J Alzheimers Dis. 2010;20(4):1015-28. PubMed.

    . A modified beta-amyloid hypothesis: intraneuronal accumulation of the beta-amyloid peptide--the first step of a fatal cascade. J Neurochem. 2004 Nov;91(3):513-20. PubMed.

    . Intracellular accumulation of amyloid-Beta - a predictor for synaptic dysfunction and neuron loss in Alzheimer's disease. Front Aging Neurosci. 2010;2:8. PubMed.

    . Neuron loss in transgenic mouse models of Alzheimer's disease. Int J Alzheimers Dis. 2010;2010 PubMed.

    . Motor deficits, neuron loss, and reduced anxiety coinciding with axonal degeneration and intraneuronal Aβ aggregation in the 5XFAD mouse model of Alzheimer's disease. Neurobiol Aging. 2012 Jan;33(1):196.e29-40. PubMed.

    . Intraneuronal pyroglutamate-Abeta 3-42 triggers neurodegeneration and lethal neurological deficits in a transgenic mouse model. Acta Neuropathol. 2009 Oct;118(4):487-96. PubMed.

    . Identification of low molecular weight pyroglutamate A{beta} oligomers in Alzheimer disease: a novel tool for therapy and diagnosis. J Biol Chem. 2010 Dec 31;285(53):41517-24. PubMed.

  6. The current debate catalyzed by the important paper of Winton et al. highlights a serious concern facing the AD field, namely, the relevance of transgenic animal models to the neurodegeneration of AD, the real cause of dementia. Most transgenic models currently used in AD research overexpress high levels of exogenous APP (many times over endogenous APP), an artificial condition not applicable to AD where there is no APP overexpression. Moreover, protein overexpression in brain often causes toxicities due to diverse reasons, including possible interference of the overexpressed protein with normal cellular trafficking, which may not be relevant to disease mechanisms (see also Robakis, 2011; Pimplikar et al., 2010).

    The APP-overexpressing transgenic animals may serve as models of amyloidosis. However, it should be kept in mind that although amyloidosis in these animals is driven by overproduction of Aβ peptides (due to overexpression of APP), there is little evidence of Aβ overproduction in sporadic AD, and the mechanism of Aβ fibrillation in human brain (both diseased and normal) remains unknown.

    Finally, research in the last quarter of a century has shown that, although both APP and presenilins play critical roles in the etiology of AD, the role of Aβ and AD amyloid, beyond its usefulness for the definition of the disease, remains unclear.


    . Mechanisms of AD neurodegeneration may be independent of Aβ and its derivatives. Neurobiol Aging. 2011 Mar;32(3):372-9. PubMed.

    . Amyloid-independent mechanisms in Alzheimer's disease pathogenesis. J Neurosci. 2010 Nov 10;30(45):14946-54. PubMed.

  7. The study by Winton et al. is a well-designed study that provides important new information on a frequently used animal model of Alzheimer’s disease. However, based on the results presented, I would be cautious to conclude that intraneuronal Aβ accumulation does not occur in 3xTg-AD. We have never had the opportunity to examine the neuropathology of 3xTg-AD, but we do have considerable experience with Tg-ArcSwe and Tg-Swe mice.

    Winton et al. used paraffin-embedded material (e.g., Fig. 1), and in our hands, it is difficult to detect intraneuronal Aβ in such tissues. Far greater sensitivity is reached by instead using frozen tissue sections, paraformaldehyde-fixed (4 percent) or unfixed, and immersing them in heated citric acid and formic acid for antigen retrieval. With N-terminal cleavage-specific antibodies, it is quite easy to detect intraneuronal Aβ, but with C-terminal cleavage-specific antibodies, the protocols need to be well optimized (especially with tissues from young mice). Moreover, age and anatomic location are important parameters to consider. We have noticed that intraneuronal Aβ first emerges in neurons with very high APP overexpression and is easiest to detect at the time when extracellular plaques begin to form. Winton et al. also used paraformaldehyde-fixed vibratome sections and 82E1 (an N-terminal neoepitope-specific Aβ-antibody), but, unfortunately, they do not show the results.

    Similar to Winton et al., we conducted genetic and pharmacological experiments. We generated double-transgenic Tg-ArcSwe x BACE-knockout mice, administered a γ-secretase inhibitor, and combined that with immunohistochemical detection techniques. Our data strongly suggest that free intraneuronal Aβ peptides do accumulate in Tg-ArcSwe mice (Philipson et al., 2009). The same age-dependent intraneuronal Aβ-phenotype develops in both Tg-ArcSwe and Tg-Swe (Tg-Swe is devoid of the Arctic mutation: Lord et al., 2006). Possibly, intraneuronal Aβ accumulation occurs in some APP transgenic models and not at all in other models, but it is more likely that the levels and thereby ease with which intraneuronal Aβ can be detected differs between models.

    Winton et al. highlight the fact that intraneuronal APP is far more abundant than intraneuronal Aβ in transgenic mouse brain. When studying intraneuronal Aβ, it is therefore essential to use neoepitope-specific Aβ antibodies, and to carefully validate and optimize the experimental protocols. Interestingly, 3xTg-AD mice, but not Tg2576, display strong intraneuronal APP-immunostaining, i.e., the phenotype does not simply depend on the extent of APP overexpression. Importantly, Winton et al. observe similar age-dependent tau phenotypes in 3xTg-AD mice with or without murine BACE (and thus human Aβ).


    . The Arctic Alzheimer mutation facilitates early intraneuronal Abeta aggregation and senile plaque formation in transgenic mice. Neurobiol Aging. 2006 Jan;27(1):67-77. PubMed.

    . Genetic and pharmacological evidence of intraneuronal Abeta accumulation in APP transgenic mice. FEBS Lett. 2009 Sep 17;583(18):3021-6. PubMed.

  8. How can mice with so many genetic defects possibly reflect the pathophysiology of the earliest, asymptomatic, pre-amyloid
    accumulation stages of human AD? If this question cannot be answered positively and persuasively, it matters little what form of APP is being analyzed by the different antisera.

  9. It is a healthy sign that the paper by Winton et al. has initiated a debate over intraneuronal Aβ, as more studies are appearing in the literature suggesting that intraneuronal Aβ could play a pathological role in Alzheimer’s disease (AD). This issue can be debated at the technical level in terms of antibody specificity and what a given antibody "sees" or "does not see" in brain sections. The debate can also commence at the biological level to see how the concept of intraneuronal Aβ fits with what we know about the subcellular locations where APP is processed by BACE and γ-secretase.

    As pointed out by Thomas Bayer and others, both 6E10 and 4G8 antibodies recognize the Aβ portion in full-length APP and APP-C-terminal fragments. When these antibodies are used in Western blotting, it is easy to distinguish Aβ from other fragments because of its size. Obviously, this is not possible in immunohistochemical analysis. Therefore, scientists rely on double-staining using one of these antibodies in conjunction with a C-terminal antibody (which should see other forms of APP except Aβ) and subtracting one signal from the other (see McAlpine et al., 2009) to get an indirect measure of intracellular Aβ. However, this approach is less than satisfactory, prone to artifacts and subject to interpretation (e.g., see LaFerla’s comments on Fig. 5 of Winton et al.) and might be the root cause of discordance between the findings of Winton et al. and Oddo et al. However, Winton et al. used end-specific antibodies (Ban50 and BC05; Fig. 1), and combined this approach with genetic ablation of BACE (Fig. 6), thereby providing a compelling argument in support of their interpretation. One curious observation, however, is that Winton et al. found no immunoreactivity using BC05 in Fig. 6K. Similarly, Philipson et al. (see comments by Nilsson above) also find no signal using the 40- or 42-specific antibodies that recognize neoepitopes at the C-terminus (Fig. 3 of Philipson et al.). Shouldn’t γ cleavage of α-CTF expose the end of p3 peptide which is recognized by these antibodies? Or is α-CTF poorly or not processed by γ-secretase at all (1)? The lack of this signal is an enigma, since BACE-knockout animals should retain α- and γ-secretase activities and produce large amounts of p3 fragment.

    It is also important to consider where in neurons β-CTF peptides are cleaved by γ-secretase, since this is precisely where Aβ peptides are "born." A generally accepted view is that shedding of the extracellular domain of APP is a prerequisite for γ-cleavage (2), and that BACE cleavage takes place in early endosomes (3,4), followed by γ-cleavage in a post-Golgi compartment (5). This makes it difficult to understand how so much intraneuronal Aβ signal is observed in the soma, which is far from the location of early endosomes. On the other hand, full-length APP is known to be accumulated in perinuclear space in the soma, increasing the likelihood that the conclusions of Winton et al. are correct. On a somewhat related note, there are published reports of the presence of Aβ in mitochondria (together with APP and γ-secretase (6,7), which is very difficult to reconcile with the current knowledge of biosynthetic pathways of membrane proteins.

    Finally, the sentiments expressed by Nik Robakis (since amyloid-independent mechanisms can also contribute to AD; see, e.g., Alzforum Webinar) or Vince Marchesi above are important. Marchesi’s comments will make some of us cringe uncomfortably, but he raises a valid point. Moreover, if Aβ does accumulate intracellularly, is it present in sufficient amounts to be able to contribute to the disease mechanism? Is it present free in cytosol or bound in membrane vesicles? I believe the intraneuronal Aβ Thomas Bayer is alluding to is the pyroglutamate Aβ, which is different from the one debated here. The Alzforum Webinar organized around this theme (16 June 2011) will be an interesting event where some of these questions might get answered. I hope the panel members will also discuss the "forest" view of intraneuronal Aβ in addition to examining the "trees." Knowing the panel members, there is no doubt that it will be a lively event.


    . Dual role of alpha-secretase cleavage in the regulation of gamma-secretase activity for amyloid production. J Biol Chem. 2010 Oct 15;285(42):32549-56. PubMed.

    . Requirements for presenilin-dependent cleavage of notch and other transmembrane proteins. Mol Cell. 2000 Sep;6(3):625-36. PubMed.

    . Amyloidogenic processing of the Alzheimer beta-amyloid precursor protein depends on lipid rafts. J Cell Biol. 2003 Jan 6;160(1):113-23. PubMed.

    . Efficient inhibition of the Alzheimer's disease beta-secretase by membrane targeting. Science. 2008 Apr 25;320(5875):520-3. PubMed.

    . The discrepancy between presenilin subcellular localization and gamma-secretase processing of amyloid precursor protein. J Cell Biol. 2001 Aug 20;154(4):731-40. PubMed.

    . Mitochondrial accumulation of APP and Abeta: significance for Alzheimer disease pathogenesis. J Cell Mol Med. 2009 Oct;13(10):4137-45. PubMed.

    . The mitochondrial secret(ase) of Alzheimer's disease. J Alzheimers Dis. 2010;20 Suppl 2:S381-400. PubMed.

  10. We agree with previous comments that Winton and colleagues provide interesting new evidence of APP accumulation and the possible lack of interaction between Aβ and tau in the triple-transgenic mouse. However, this paper is not enough to counter the hundreds of papers now on intraneuronal Aβ in Alzheimer’s disease, and the triple-transgenic mouse remains an important model of both β amyloid and tau pathogenesis. Dystrophic neurites in human AD show APP accumulation; the precise form of this APP isn’t fully clear. It is possible that the development of AD might be characterized by an increase of APP accumulation earlier than Aβ (such as occurs in head injury). Winton et al. also highlight the problem of using Aβ domain antibodies (such as 6E10, 4G8, etc.) to differentiate Aβ from APP.

    Our main point in this comment is that intraneuronal Aβ is not easy to see, particularly when you compare it to more easily detectable APP. Aβ is also a changing target. As we well know, it likes to aggregate, but the Aβ42 end-specific antibodies are remarkably specific for the monomer form (Takahashi et al., 2004), so that oligomers become invisible unless you use an oligomer-specific antibody.

    How do we know that Aβ42 antibodies are not just seeing APP? Immunofluorescence and immunoelectron microscopy clearly show that antibodies against Aβ42 have different subcellular labeling when compared to APP N- and C-terminal antibodies (Takahashi et al., 2002; Muresan et al., 2009; Gouras et al., 2010: Fig. 2, Aβ42 and APP in human primary neurons). In other words, high-resolution imaging clearly shows that APP and Aβ42 are in different places, which obviously does not fit the idea that they are one and the same. In addition, if Aβ C-terminal end-specific antibodies see APP, this would discount much of AD research on Aβ that has relied on these antibodies for ELISA.

    Moreover, the paper by Winton et al. in a way argues against Aβ antibodies seeing APP, since they do not even detect intraneuronal Aβ even when there is a lot of APP. There is another technical point: They see intraneuronal APP only faintly by immunohistochemistry and not at all by immuno-EM in Tg2576 mice. Although APP is normally an abundant protein in neurons, it is overexpressed in these mice, and it has been seen.

    Intraneuronal Aβ isn’t easy to see, but it is there. With a good confocal and antibodies, it is quite evident even in wild-type neurons. For recent rigorous technical papers on intraneuronal Aβ, we recommend (as Bayer and Wirths did): Aho et al. (2009) for postmortem human brain; Philipson et al. (2009) on AD transgenic mice; and an important biochemical paper on intraneuronal Aβ42 by Hashimoto et al. (2010).


    . Immunohistochemical visualization of amyloid-beta protein precursor and amyloid-beta in extra- and intracellular compartments in the human brain. J Alzheimers Dis. 2010;20(4):1015-28. PubMed.

    . Intraneuronal beta-amyloid accumulation and synapse pathology in Alzheimer's disease. Acta Neuropathol. 2010 May;119(5):523-41. PubMed.

    . Analysis of microdissected human neurons by a sensitive ELISA reveals a correlation between elevated intracellular concentrations of Abeta42 and Alzheimer's disease neuropathology. Acta Neuropathol. 2010 May;119(5):543-54. PubMed.

    . The cleavage products of amyloid-beta precursor protein are sorted to distinct carrier vesicles that are independently transported within neurites. J Neurosci. 2009 Mar 18;29(11):3565-78. PubMed.

    . Genetic and pharmacological evidence of intraneuronal Abeta accumulation in APP transgenic mice. FEBS Lett. 2009 Sep 17;583(18):3021-6. PubMed.

    . Intraneuronal Alzheimer abeta42 accumulates in multivesicular bodies and is associated with synaptic pathology. Am J Pathol. 2002 Nov;161(5):1869-79. PubMed.

    . 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.

  11. In my view, the paper by Winton et al. supports the key point that one can prove the
    existence of intraneuronal Aβ only by using end-specific Aβ
    antibodies (i.e., to Aβ Ile42 or Val40, and, to a lesser extent, Asp1) that cannot
    recognize any APP epitopes other than the free Aβ peptide. This would exclude numerous antibodies that are not truly end-specific and recognize epitopes in holoAPP, C99 and C83, and/or APPs-α.

    Also, more studies using sensitive biochemical assays to extract free intracellular Aβ from cells and quantify it would be useful, if this has not yet been widely done by multiple labs. One would need to use methods to eliminate any Aβ peptides bound to the outside surface of cells before lysing the cells and quantifying free Aβ inside the cells (or specifically within their membrane vesicles).

    Obviously, Aβ is normally generated by β- and γ-secretases within vesicles inside cells (in addition to on the plasma membrane), but these initially intracellular peptides are efficiently secreted under wild-type circumstances. The question is, How much of Aβ at steady state exists inside the cell versus in the extracellular space—in normal brain and then in AD brain. The Winton paper helps shed light on this important question in an APP-overexpressing mouse model.

  12. We would like to comment on a few issues raised by this paper and previous comments, starting with the now old observations from various groups, Gunnar Gouras’s in particular (1), that in the AD brain Aβ accumulates in a population of neurons in the vulnerable brain regions. Often, some Aβ also accumulates intraneuronally in the non-AD brain. This should not be surprising.

    Yet, these data are questioned again and again, mostly based on disbelief that the antibodies used for immunocytochemistry/immunohistochemistry (ICC/IHC) detect genuine Aβ. Indeed, this should be of concern in any ICC/IHC experiment, and this is why adequate controls are required. Antibodies usually recognize epitopes present in immunogens, mostly in the conformation present in the immunogen. Some antibodies work great on immunoblots but are useless in ICC/IHC. Others work well in ICC/IHC but detect the antigen poorly on immunoblots. Numerous other factors may lead to uncertainties as to what the molecular species the antibodies detect in situ really is. The problem is even more complicated with detection of Aβ because the epitopes contained in it are also present in APP and—some of them—in CTFs.

    Yet, no matter how one looks at it, Aβ is generated intracellularly. There is no question about this. Unlike α-secretase cleavage, which occurs to a large extent at the cell surface (2), β-secretase cleavage occurs, and Aβ is produced, in intracellular compartments. This happens mainly in endosomes and autophagic vacuoles, but also in early secretory compartments such as the endoplasmic reticulum or the ER-Golgi-intermediate compartment, as Virginia Lee's pioneering papers in the 1990s clearly showed (3,4).

    Normally, Aβ is either degraded via the lysosome/autolysosome/proteasome machineries inside cells, or secreted into the extracellular space—some via true secretory vesicles, and most from endosomes, including multivesicular bodies. However, in conditions of AD and in old age, the secretory/endocytic/degradative machineries could fail (see, e.g., the latest study from Randy Nixon’s lab [5]). As a consequence, Aβ is likely to accumulate within cells. A different pool of intracellular Aβ could form by reuptake of Aβ from the extracellular space. The re-internalized Aβ could easily end up in the same endosomes that contain unprocessed and partially processed APP. Thus, it is expected that cells contain a mixture of full-length APP and APP fragments—including Aβ—in the same intracellular compartments.

    Much is being discussed about the use of antibodies raised to internal epitopes of Aβ to detect the free peptide in ICC/IHC. In theory, they should detect not only Aβ, but also full-length APP and other APP-derived polypeptides that contain the epitope. Yet this is not exactly true. It boils down to the affinity of the antibody for the Aβ peptide versus APP and APP-derived fragments. As Charlie Glabe noted in his comment, antibodies made to the Aβ peptide preferentially detect the epitope in the conformation that was available in the antigen, i.e., the injected peptide. If so, antibody 6E10 is likely to bind with higher affinity to Aβ proper, compared to APP and APP-derived polypeptides containing the 6E10 epitope. This is particularly true with in-situ detection, such as ICC/IHC, where the native conformation is preserved to a greater extent than in immunoblotting assays, which denature the proteins.

    In our experience, the 6E10 antibody in ICC/IHC has a preference for the Aβ peptide compared to full-length APP or CTFβ. We usually rely on co-staining with antibodies to C- and N-terminal epitopes to determine what exactly the antibodies to internal Aβ epitopes stain. Validation with terminal end-specific antibodies should be always done, but these antibodies are usually less well characterized.

    The problem with LaFerla's 3xTg-AD mouse is more complicated, since it expresses mutant APP at above physiological levels. We recently showed that, upon overexpression, APP processing becomes abnormal, with much APP remaining uncleaved (6). This might be one explanation for the seemingly increased accumulation of full-length APP detected in the study by Winton et al. Still, we think the 3xTg-AD mouse also accumulates large amounts of genuine Aβ within neurons, as LaFerla and other groups reported over the years. While we did not study this particular mouse, we found intraneuronal accumulations of bona fide Aβ in a subpopulation of neurons from another model of AD—Bruce Lamb’s R1.40 mouse, which expresses a human APPswe mutant at slightly above-normal levels (7). Of course, there is also full-length APP in the neuronal soma in the brain of this mouse, as expected.

    We think that the apparent contradiction between the earlier and present studies with the 3xTg-AD mouse is easy to reconcile. Full-length APP and Aβ co-accumulate in certain neurons. We believe this is also true for the human AD brain. In our view, Winton et al. do not argue against the presence of intraneuronal Aβ in AD, and its relevance to the disease process. Virginia Lee was, in fact, an early pioneer of the idea of intracellular Aβ. Instead of being a repudiation of this idea, we think that the paper is a reminder of the difficulties in drawing conclusions when the real culprit may hide in a “sea” of potential culprits. In other words, is the culprit Aβ? Is it full-length APP? Or the CTFs, including AID?

    Maybe none of them. We would like to conclude by referring to Nikos Robakis's comment. Like him, we are not convinced that so far the data connecting APP and Aβ to AD support firm conclusions on their roles in the disease process. No one can doubt that they accompany the disease in most cases of bona-fide AD, familial or sporadic. They are part of this disease. Therefore, the focus on APP and Aβ is fully justified, as is the work with animal models of disease (even if they are not perfect), as long as this is not the only focus.

    See also: Muresan, V., B.T. Lamb, and Z. Muresan, DISC1 is required for the formation of intracellular Aβ oligomers, suggesting a link between schizophrenia and Alzheimer’s disease. Annual Meeting of the Society for Neuroscience, San Diego, November 13-17, 2010.


    . Intraneuronal Abeta42 accumulation in human brain. Am J Pathol. 2000 Jan;156(1):15-20. PubMed.

    . Amyloid precursor protein trafficking, processing, and function. J Biol Chem. 2008 Oct 31;283(44):29615-9. PubMed.

    . Alzheimer's A beta(1-42) is generated in the endoplasmic reticulum/intermediate compartment of NT2N cells. Nat Med. 1997 Sep;3(9):1021-3. PubMed.

    . Detection of a novel intraneuronal pool of insoluble amyloid beta protein that accumulates with time in culture. J Cell Biol. 1998 May 18;141(4):1031-9. PubMed.

    . Lysosomal proteolysis inhibition selectively disrupts axonal transport of degradative organelles and causes an Alzheimer's-like axonal dystrophy. J Neurosci. 2011 May 25;31(21):7817-30. PubMed.

    . The cleavage products of amyloid-beta precursor protein are sorted to distinct carrier vesicles that are independently transported within neurites. J Neurosci. 2009 Mar 18;29(11):3565-78. PubMed.

  13. The paper by Winton et al. is a well-written and comprehensive study of the nature of intracellular Aβ-immunoreaction (IR) as observed in very representative transgenic (Tg) animal models. The proposition that only APP material is present intraneuronally in such models is indeed thought provoking. Most of us in the field will accept that the intraneuronal Aβ immunoreactive material might represent a variety of molecular species, notably the APP C-terminal fragment (C-99), APP (as well illustrated in this publication), and a variety of Aβ species—monomeric, soluble oligomeric, and fibrillar forms. The report does not, however, invalidate the comprehensive and well-supported work done by the groups of LaFerla, Gouras, Bayer, our own lab, and others.

    In my view, it is risky to assert that a given molecule (Aβ), which exists as N-terminally and C-terminally truncated forms (and is conformationally diverse in its presentation) is not present because it cannot be demonstrated in certain immunohistochemical conditions. In this case, 10 percent formaldehyde fixation, a procedure which will render extensive protein-peptide cross-linking, occluding many antigenic sites and creating steric hindrance to antibody recognition, will make it difficult to label weaker and not abundant epitopes. The bulk of the studies supporting the presence of intracellular Aβ immunoreactive material have been performed in milder formaldehyde (4 percent) fixation procedures.

    Another problem with this report is the assertion that Aβ-IR as revealed with 4G8 with confocal microscopy in the triple Tg is fully colocalized with that of antibody 5685. Such colocalization cannot be resolved with the "merge" function with a very low magnification micrograph, as reported in the paper. This can be more convincingly resolved at much higher magnification. Even in Figure 5.I, it is quite clear that some neurons only reveal signal for 4G8 and not for 5685, revealing the holoAPP problem. Even if colocalizations were apparently 100 percent, it should be kept in mind that it will simply mean that both molecules are in the same compartment but not necessarily that two distinct antibodies are recognizing the same molecular species. A single focal plane in confocal microscopy is in the order of 0.7 μm, giving room for thousands of molecules in the same "merged" image. For that reason, proper colocalization analysis must be performed (Manders et al. 1993). Indeed, given that the images provided do not show the same intensity distribution between the two channels (the neurons with the highest 4g8 immunoreactivity are often not the ones with the highest 5685 immunoreactivity), it is likely that colocalization analysis reveals that the antibodies detect different proteins.

    Despite these limitations, and in line with LaFerla's and Pike's comments, our lab has been able to differentiate the APP versus Aβ immunoreactions. In our recently described Tg rat (Leon et al., 2010) and mouse (Ferretti et al., 2011) models of AD-like amyloid pathology, we dissociated APP-immunoreactivity with the standard APP N-terminus antibody 22C11 from that of McSA1, which preferentially recognizes Aβ-immunoreactivity independently of state of oligomerization or conformation. With simultaneous immunolabeling with 22C11 and McSA1, and analyzing high magnification and resolution images, we detected a differential pattern of immunoreactivities (as illustrated in Fig. 3 of Ferretti et al., 2011). The 22C11-immunoreactivity appeared diffusely distributed, and that of McSA1-IR (Aβ) was largely associated to granule-like structures in the perinuclear region and in the axon. Little to no subcellular colocalization was observed, suggesting that McSA1 does not recognize the holo-APP protein in 4 percent paraformaldehyde fixed free-floating tissue sections. Furthermore, such intracellular material was also reactive to the Nu1 antibody (Lambert et al., 2007) well characterized for the identification of Aβ oligomers (ADDLs). Again, in our hands, in the McGill-Thy1-APP Tg mice dual immunostaining revealed a differential intracellular staining pattern for Nu1 and APP (Fig. 7 of Ferretti et al., 2011) . Likewise, in the McGill-Thy1-R-APP rat transgenic model of the full amyloid pathology, where we identify a pre-plaque stage of intracellular β amyloid. That immunoreaction with the McSA1 antibody can also be differentiated with confocal microscopy from that of APP, as revealed by the 22C11 antibody. Furthermore, to exclude possible cross-reactivity of the human Aβ-specific antibody used (McSA1) with human holo-AβPP or N-terminal products thereof (e.g., soluble AβPPα or sAβPPα), we pre-absorbed (Fig. S2) McSA1 with synthetic Aβ1-42. This completely abolished intracellular immunostaining, while the same molar concentration of synthetic soluble sAβPPα did not show any significant effect (see supplementary data of Leon et al., 2010.)

    The report of Winton et al. postulates that overexpression of holo-APP protein rather than intracellular accumulation of Aβ molecular species is behind the often-reported intracellular Aβ immunoreactivity in transgenic models. It is certainly possible that fairly aggressive transgenesis would facilitate some accumulation of the holo-APP molecule within neurons, but it does not exclude the occurrence of Aβ as recognized by a number of fairly Aβ-specific antibodies. In the McGill-R-Thy1-APP rat transgenic model, that proposition does not apply, as the full amyloid pathology is achieved in the Tg rat with a single APP transgene per allele, making it closer to the progressive, slowly evolving, human pathology.

    Winton et al. also postulate that the presence of mutated presenilin protein might provoke perikaryal accumulation of APP due to a vesicular impairment. As our rat and mice transgenic models only carry human mutated forms of APP, that would not apply. Importantly, in both the McGill mice and rat Tg models, we consistently detect Aβ40 and Aβ42 by ELISA in pre-plaque stages, as it has also been reported in other Tg models. This occurrence of Aβ40 and Aβ42 fragments at the stage of intracellular accumulation of Aβ immunoreactivity has been validated in the rat Tg model independently with biochemical analyses by Drs. Lisa Munter and Gerd Multhaup (Freie University of Berlin), colleagues who have a solid record in Aβ biochemistry.

    In closing, it is likely that the Aβ-immunoreactivity in transgenic models represents not only Aβ itself, evolving from monomeric to oligomeric to fibrillar as suggested by the report of Ferretti et al. (Ferretti et al., 2011), but APP as well (as illustrated by the work of Winton et al., as well as other APP fragments, notably C-99). The main challenge would be to determine which are the relative proportions of these entities and, ultimately and most importantly, which would be their relative contribution to CNS dysfunction.


    . Measurement of colocalization of objects in dual colour confocal images. J. Microsc. 1993 Mar;169(3):375-82.

    . A novel transgenic rat model with a full Alzheimer's-like amyloid pathology displays pre-plaque intracellular amyloid-beta-associated cognitive impairment. J Alzheimers Dis. 2010;20(1):113-26. PubMed.

    . Transgenic mice as a model of pre-clinical Alzheimer's disease. Curr Alzheimer Res. 2011 Feb 1;8(1):4-23. PubMed.

    . Monoclonal antibodies that target pathological assemblies of Abeta. J Neurochem. 2007 Jan;100(1):23-35. PubMed.

  14. The intraneuronal pool of Aβ may have a double origin, which should be incorporated in the discussion raised by findings by Winton et al. that question the existence of intraneuronal Aβ in the 3xTg mouse model.

    Seminal findings by laboratories headed by Beyreuther and Glabe reported more than 15 years ago that exogenous Aβ42 is taken up from the extracellular space and may result in accumulation of amyloidogenic fragments in different cell lines (Knauer et al., 1992; Ida et al., 1996). Interestingly, Ida et al. found an Aβ-specific clearance activity inside of neuroblastoma SH-SY5Y cells. Cellular uptake of Aβ was also reported to occur in rat microglia cells when removal of Aβ from the medium and an accumulation in the cells was observed (Ard et al., 1996).

    Although the molecular mechanisms for neuronal Aβ internalization have remained unclear, the involvement of numerous receptors in endocytosis of Aβ was described, as well as an accumulation of internalized Aβ in the endosomal/lysosomal system or late endosomes/multivesicular bodies (MVBs) (for a review, see Mohamed and Posee de Chaves, 2011). In glial cells, the purpose of uptake may reflect mainly clearance activity, whereas in neurons Aβ uptake may represent an alternative mechanism in the development of AD. Novel findings from our group strongly support this view.

    As presented by Christian Barucker et al. at the last ADPD Conference (Barucker et al., 2011. A novel function of Aβ in gene regulation as an alternative pathway of toxicity. ADPD Conference, Barcelona, Spain, 9-13 March 2011), shorter and longer forms of Aβ are rapidly internalized by SH-SY5Y cells and transported into the nucleus. There, all Aβ peptides accumulated in a time dependent manner. However, we found only specific genes affected by toxic Aβ42 peptides (manuscript in preparation). Also, our findings indicate that this nuclear signaling function is unique for Aβ42 wild-type, and imply that deregulation of Aβ target genes could be an alternative pathway for Aβ-induced neurotoxicity.


    . Intracellular accumulation and resistance to degradation of the Alzheimer amyloid A4/beta protein. Proc Natl Acad Sci U S A. 1992 Aug 15;89(16):7437-41. PubMed.

    . Rapid cellular uptake of Alzheimer amyloid betaA4 peptide by cultured human neuroblastoma cells. FEBS Lett. 1996 Sep 30;394(2):174-8. PubMed.

    . Scavenging of Alzheimer's amyloid beta-protein by microglia in culture. J Neurosci Res. 1996 Jan 15;43(2):190-202. PubMed.

  15. The current debate on intraneuronal amyloid β (Aβ) in Frank LaFerla’s mice is thought provoking to the entire Alzheimer’s disease (AD) research community because Aβ is a key factor in progression and pathogenesis. In view of that, I read the paper by Winton et al. with great interest. Dr. Lee and colleagues have characterized AD mouse models (3xTg-AD, 3xTg-AD/BACE-/-, BACE-/-, and non-transgenic WT mice) in order to determine the existence of intraneuronal Aβ, and reported that intraneuronal Aβ is absent in 3xTg-AD mice. In our experience, immunostaining analysis using fixed tissues is unreliable and requires extensive optimization. Thus, the conclusion that intraneuronal Aβ is not present in 3xTg-AD mice needs further investigation.

    We extensively characterized postmortem brains from AD patients and control subjects, cortical tissues from APP/PS1 mice, and primary neurons from Tg2576 mice using 6E10, 1-40, 1-42, and A11 antibodies. We found excessive amounts 4 kDa Aβ, and also oligomeric Aβ and full-length APP and other APP derivatives in AD patients, APP/PS1 mice, and Tg2576 mice (see Fig. 3 in our recent paper Manczak et al., 2011). Based on immunoprecipitation, Western blotting, and immunohistochemistry data, we concluded that the 6E10 antibody is immunoreactive to full-length APP and its derivatives.

    Hyperphosphorylation of tau has been found to be induced by several factors: 1) chronic oxidative stress (Su et al., 2010); 2) Aβ-mediated oxidative stress (Garwood et al., 2011); 3) reduced insulin-like growth factor 1 (Cheng et al., 2005); and 4) mutations in tau gene (Oddo et al., 2003; Lewis et al., 2001; Gotz et al., 2001). The hyperphosphorylation observed in 3xTg-AD mice may likely be caused by a combination of Aβ-mediated oxidative stress and tau mutation.

    Several recent studies reported that intraneuronal Aβ is associated with mitochondrial membranes, causing mitochondrial dysfunction and neuronal damage (Caspersen et al., 2005; Crouch et al., 2005; Manczak et al., 2006; Devi et al., 2006; Hansson Petersen et al., 2009; Yao et al., 2009). Several groups have examined the 3xTg-AD for mitochondria abnormalities and oxidative stress (e.g., Yao et al., 2009, Resende et al., 2008). They found that Aβ is associated with mitochondria and increased oxidative stress and decreased respiration (supporting the presence of intraneuronal Aβ in 3xTg-AD mice).

    Overall, the data overwhelmingly support intraneuronal Aβ in AD brains and AD mouse models, and support the view that Aβ causes neuronal damage. Therefore, at this point, given the state of Aβ research in the AD field, I believe the conclusions by Winton et al. are questionable.


    . Tau is hyperphosphorylated in the insulin-like growth factor-I null brain. Endocrinology. 2005 Dec;146(12):5086-91. PubMed.

    . Copper-dependent inhibition of human cytochrome c oxidase by a dimeric conformer of amyloid-beta1-42. J Neurosci. 2005 Jan 19;25(3):672-9. PubMed.

    . Accumulation of amyloid precursor protein in the mitochondrial import channels of human Alzheimer's disease brain is associated with mitochondrial dysfunction. J Neurosci. 2006 Aug 30;26(35):9057-68. PubMed.

    . Cyclophilin D deficiency attenuates mitochondrial and neuronal perturbation and ameliorates learning and memory in Alzheimer's disease. Nat Med. 2008 Oct;14(10):1097-105. PubMed.

    . Astrocytes are important mediators of Aβ-induced neurotoxicity and tau phosphorylation in primary culture. Cell Death Dis. 2011;2:e167. PubMed.

    . Formation of neurofibrillary tangles in P301l tau transgenic mice induced by Abeta 42 fibrils. Science. 2001 Aug 24;293(5534):1491-5. PubMed.

    . The amyloid beta-peptide is imported into mitochondria via the TOM import machinery and localized to mitochondrial cristae. Proc Natl Acad Sci U S A. 2008 Sep 2;105(35):13145-50. PubMed.

    . Enhanced neurofibrillary degeneration in transgenic mice expressing mutant tau and APP. Science. 2001 Aug 24;293(5534):1487-91. PubMed.

    . ABAD directly links Abeta to mitochondrial toxicity in Alzheimer's disease. Science. 2004 Apr 16;304(5669):448-52. PubMed.

    . Mitochondria are a direct site of A beta accumulation in Alzheimer's disease neurons: implications for free radical generation and oxidative damage in disease progression. Hum Mol Genet. 2006 May 1;15(9):1437-49. PubMed.

    . Impaired mitochondrial dynamics and abnormal interaction of amyloid beta with mitochondrial protein Drp1 in neurons from patients with Alzheimer's disease: implications for neuronal damage. Hum Mol Genet. 2011 Jul 1;20(13):2495-509. PubMed.

    . Chronic oxidative stress causes increased tau phosphorylation in M17 neuroblastoma cells. Neurosci Lett. 2010 Jan 14;468(3):267-71. PubMed.

    . Mitochondrial bioenergetic deficit precedes Alzheimer's pathology in female mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A. 2009 Aug 25;106(34):14670-5. PubMed.

  16. The paper by Winton et al. has raised a debate on the significance of intraneuronal Aβ. I generally agree with the opinions by Frank LaFerla, Salvo Oddo, Christian Pike, and Gunnar Gouras et al., arguing that the data are insufficient to disprove the presence of intraneuronal free Aβ.

    Visualization of intraneuronal Aβ by endo-specific antibodies requires appropriate immunodetection methods, including pretreatment for antigen retrieval (e.g., treatment with formic acid). Therefore, it is possible that the absence of immunopositive signals with endo-specific Aβ antibodies (BC05 and BA27) was due to inefficient antigen retrieval. The authors compared 3xTg-AD mice with 3xTg-AD-BACE knockout mice, but the comparison did not yield information about whether BACE deficiency has any effect on intraneuronal Aβ.

    I'd like to emphasize that intraneuronal Aβ, especially Aβ42, is detectable in our mouse model without APP overexpression (Chui et al., 1999). Intracellular Aβ is evident in many cellular models including ours (Takeda et al., 2004). However, it is true that intracellular Aβ is fairly difficult to be detected, as claimed by Gouras et al. Thus, cautious research is necessary in dealing with intraneuronal Aβ.


    . Transgenic mice with Alzheimer presenilin 1 mutations show accelerated neurodegeneration without amyloid plaque formation. Nat Med. 1999 May;5(5):560-4. PubMed.

    . Intraneuronal Abeta42 accumulation in human brain. Am J Pathol. 2000 Jan;156(1):15-20. PubMed.

    . Enhanced generation of intracellular Abeta42 amyloid peptide by mutation of presenilins PS1 and PS2. Eur J Neurosci. 2004 Jan;19(2):258-264. PubMed.

  17. So APP might also exist inside cells: Didn’t we know this already? The question now appears to be whether there is more APP than shorter forms of Aβ inside cells.

    We evaluated accumulation of mitochondrial Aβ in mitochondrial fractions in the presence of digitonin to diminish contamination with other organelles. To investigate whether Aβ was within the organelle (i.e., within a membrane-enclosed compartment) or adsorbed to the outer mitochondrial membrane, we performed protease protection to investigate whether Aβ was within the organelle (i.e., within a membrane-enclosed compartment) or adsorbed to its outer membrane. The data indicated that most of mitochondria-associated Aβ was in a protected, membrane-bound compartment (8). Western blotting, confocal microscopy, and immunoelectron microscopy have shown in human AD patients that Aβ is present intracellularly, including mitochondria (2,8,10,14).

    Hansson Petersen et al. demonstrated that Aβ is localized in mitochondria of human aged brain from surgical biopsy tissue by immunogold EM with a specific Aβ antibody. Intracellular proteins that bind to this internal Aβ have been identified in animal models and humans (2,10-11,13), and these proteins were not found to be binding APP. There have been papers showing that these intracellular Aβ interactions or accumulation of intracellular Aβ can be perturbed with reversing of both biochemical and physiological changes (2,9,11-13,15-18).

    Thus, it seems that one paper cannot overwhelm a series of findings derived from many pathological and biochemical studies.

    Comparing the images presented by Winton et al. with those in other papers using the same mouse model at similar ages (19-20), we notice substantially fewer Aβ plaques in Winton et al. (Fig.1B). Possibilities underlying this discrepancy are, first, Winton used 3xTg-AD mice demonstrating less Aβ production/deposition at the ages up to 12 months, and/or Winton’s immunostaining method was less sensitive because of the methodology. Some appropriate controls would have included pre-immune or non-immune IgG for immunostaining to confirm the specificity of immunoreaction, and qualification of all images. Biochemical approaches to show the size of the Aβ immunoreactive band would have helped to verify the presence of Aβ.

    In any event, one cannot conclude that intracellular Aβ is of limited significance in AD-related disturbances because its level is low. For example, Aβ has been shown to directly interact with intracellular proteins ABAD (2), cyclophilin D (10), and Drp1 (18), exacerbating neuronal and mitochondrial stress. These data suggest intracellular Aβ, though its level might be low, plays a role in Aβ-induced intracellular perturbations in AD pathogenesis.


    . 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.

    . ABAD directly links Abeta to mitochondrial toxicity in Alzheimer's disease. Science. 2004 Apr 16;304(5669):448-52. PubMed.

    . Distinct presenilin-dependent and presenilin-independent gamma-secretases are responsible for total cellular Abeta production. J Neurosci Res. 2003 Nov 1;74(3):361-9. PubMed.

    . Abeta localization in abnormal endosomes: association with earliest Abeta elevations in AD and Down syndrome. Neurobiol Aging. 2004 Nov-Dec;25(10):1263-72. PubMed.

    . Beta-amyloid accumulation impairs multivesicular body sorting by inhibiting the ubiquitin-proteasome system. J Neurosci. 2006 Apr 19;26(16):4277-88. PubMed.

    . Loss of endosomal/lysosomal membrane impermeability is an early event in amyloid Abeta1-42 pathogenesis. J Neurosci Res. 1998 Jun 15;52(6):691-8. PubMed.

    . RAGE-mediated signaling contributes to intraneuronal transport of amyloid-beta and neuronal dysfunction. Proc Natl Acad Sci U S A. 2009 Nov 24;106(47):20021-6. PubMed.

    . Mitochondrial Abeta: a potential focal point for neuronal metabolic dysfunction in Alzheimer's disease. FASEB J. 2005 Dec;19(14):2040-1. PubMed.

    . Mitochondria are a direct site of A beta accumulation in Alzheimer's disease neurons: implications for free radical generation and oxidative damage in disease progression. Hum Mol Genet. 2006 May 1;15(9):1437-49. PubMed.

    . Cyclophilin D deficiency attenuates mitochondrial and neuronal perturbation and ameliorates learning and memory in Alzheimer's disease. Nat Med. 2008 Oct;14(10):1097-105. PubMed.

    . Mitochondrial bioenergetic deficit precedes Alzheimer's pathology in female mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A. 2009 Aug 25;106(34):14670-5. PubMed.

    . Mitochondrial amyloid-beta levels are associated with the extent of mitochondrial dysfunction in different brain regions and the degree of cognitive impairment in Alzheimer's transgenic mice. J Alzheimers Dis. 2010;20 Suppl 2:S535-50. PubMed.

    . Inhibition of amyloid-beta (Abeta) peptide-binding alcohol dehydrogenase-Abeta interaction reduces Abeta accumulation and improves mitochondrial function in a mouse model of Alzheimer's disease. J Neurosci. 2011 Feb 9;31(6):2313-20. PubMed.

    . The amyloid beta-peptide is imported into mitochondria via the TOM import machinery and localized to mitochondrial cristae. Proc Natl Acad Sci U S A. 2008 Sep 2;105(35):13145-50. PubMed.

    . ABAD enhances Abeta-induced cell stress via mitochondrial dysfunction. FASEB J. 2005 Apr;19(6):597-8. PubMed.

    . Interaction of amyloid binding alcohol dehydrogenase/Abeta mediates up-regulation of peroxiredoxin II in the brains of Alzheimer's disease patients and a transgenic Alzheimer's disease mouse model. Mol Cell Neurosci. 2007 Jun;35(2):377-82. PubMed.

    . Endophilin I expression is increased in the brains of Alzheimer disease patients. J Biol Chem. 2008 Feb 29;283(9):5685-91. PubMed.

    . Impaired mitochondrial dynamics and abnormal interaction of amyloid beta with mitochondrial protein Drp1 in neurons from patients with Alzheimer's disease: implications for neuronal damage. Hum Mol Genet. 2011 Jul 1;20(13):2495-509. PubMed.

    . Lack of pathology in a triple transgenic mouse model of Alzheimer's disease after overexpression of the anti-apoptotic protein Bcl-2. J Neurosci. 2008 Mar 19;28(12):3051-9. PubMed.

    . Females exhibit more extensive amyloid, but not tau, pathology in an Alzheimer transgenic model. Brain Res. 2008 Jun 24;1216:92-103. PubMed.

  18. I agree with Dr. Reddy's comments. The mitochondrial localization of various Aβ species constitutes direct evidence of intraneuronal Aβ accumulation rather than APP. We have used the 6E10 antibody to detect multiple Aβ species in purified mitochondrial fractions by Western blot. In the 3xTg-AD mouse model, we observed at least three distinct bands with molecular weight at full APP size (~130 kD, ~56 kD, and a 16 kD size band). Although we did not observe the 4 kD amyloid monomer in our mitochondrial prep, we believe the 16 kD band is likely to represent the Aβ tetramer and cannot be explained by abnormally folded full-length APP.

    Similarly, in the Tg2576 model, Reddy's group demonstrated a 4 kD Western band in a purified brain mitochondrial prep. In mutant APP-expressing N2A cell lines, this group identified oligomers of Aβ at various sizes (mainly 16 kD, 27 kD, and ~40 kD).

    The small immuno-bands (4 kD, 16 kD, 27 kD, 56 kD) observed in isolated mitochondrial fraction from the 3xTg-AD mice, Tg2576 mice, and other APP transgenic models more or less provide evidence of intraneuronal, or, more specifically, mitochondrial accumulation of Aβ oligomers.

    Noteworthy: APP is also observed in the same mitochondrial prep in our study in the 3xTg-AD mouse model and in other models.


    . Mitochondrial bioenergetic deficit precedes Alzheimer's pathology in female mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A. 2009 Aug 25;106(34):14670-5. PubMed.

    . Ovarian hormone loss induces bioenergetic deficits and mitochondrial β-amyloid. Neurobiol Aging. 2011 Apr 22; PubMed.

    . Mitochondria are a direct site of A beta accumulation in Alzheimer's disease neurons: implications for free radical generation and oxidative damage in disease progression. Hum Mol Genet. 2006 May 1;15(9):1437-49. PubMed.

    . ABAD directly links Abeta to mitochondrial toxicity in Alzheimer's disease. Science. 2004 Apr 16;304(5669):448-52. PubMed.

  19. There is compelling evidence that Aβ generated from the sequential cleavage of APP by β-secretase and PS/γ-secretase is both secreted at the extracellular space and accumulated intracellularly in cultured cells and brains from transgenic mouse models and AD. The general picture raised by a large number of studies from independent labs, including our group, is that intracellular Aβ is commonly detected in the brains of several AD transgenic mouse models including 3xTg-AD, APP/PS1KI, 5xFAD, R1.40, APPSw,Ind, and APPInd (Tampellini et al. 2010; Oddo et al., 2003; Casas et al., 2004; Oakley et al., 2006; Gimenez-Llort et al., 2007; España et al., 2010; Manczak et al., 2011).

    Winton and colleagues add a new perspective and raise doubt about previous studies in 3xTg-AD mice. Our recent studies of Aβ staining with 2G3 antibody, specific for Aβ40, and an Aβ42 antibody, revealed robust intraneuronal Aβ in the basolateral amygdala (BLA), and to a lesser extent in CA1/CA3 hippocampus (20-40 percent of cells) and layer V of the neocortex (20 percent of neurons of this layer) in six-month-old 3xTg-AD mice (España et al., 2010). Staining was not detected in non-transgenic controls. This indicates that intracellular Aβ staining is almost absent in the majority of cortical layers in 3xTg-AD mice at six months (Fig. 1A, C of Winton et al.), but is detected within neurons of the hippocampus and amygdala. Winton et al. do not show pictures of BLA and hippocampus at six months.

    We observed an age-dependent increase of intracellular Aβ/APP staining with 2G3 (Aβ40) and 6E10 antibodies, but unchanged APP/APPs labeling with 8E5 antibody. Since full-length APP expression detected by Western blotting and 8E5 staining is unchanged in 3xTg-AD mice during aging, 2G3 and 6E10 stainings seem to represent intracellular Aβ species. In agreement with Christian Pike´s comment, our studies indicate that APP staining was found to be widespread in the soma, whereas Aβ staining was punctuate and perinuclear. The differential staining pattern for Aβ/APP in 3xTg-AD mice of the present report clearly differs from the previous reports, raising doubt in my mind whether the lack of BA27 (Aβ40) and BC05 (Aβ42) staining truly represents absence of intracellular Aβ. It is possible that BA27 and BC05 (regretfully JFR/c40 and c42 staining is not shown) do not recognize the conformational state of intracellular Aβ but are able to detect extracellular aggregated Aβ in plaques (Fig. 1A-D). This may also explain why BA27 and BC05 labeling disappears in 10- to 12-month-old 3xTg-AD;BACE-/- mice (Fig. 4A-D).

    From the studies with 3xTg-AD; BACE-/- mice, it seems that the experimental staining conditions with 6E10/4G8 antibodies used here (dilution of 6E10 and 4G8 antibodies may be also important) allow detection of human APP but not soluble Aβ. Still, the lack of staining with BA27 and BC05 antibodies in the 3xTg-AD neocortex, a brain region not abundant in Aβ-containing neurons, does not necessarily prove the absence of intraneuronal Aβ accumulation in 3xTg-AD mice.

    We performed many control experiments to rule out the possibility of APP detection with Aβ-specific antibodies. Using these approaches, we also detected APP expression (8E5) and intraneuronal Aβ staining with 6E10 (APP/Aβ), Aβ40 (2G3), and Aβ42 antibodies, but not amyloid plaques in the hippocampus, specific neocortical layers, and BLA in two other transgenic lines (APPInd and APPSw,Ind mice from Lennart Mucke´s lab). The consequences of abnormal intracellular accumulation of Aβ in these models and their relevance for AD pathology need further investigation.


    . Massive CA1/2 neuronal loss with intraneuronal and N-terminal truncated Abeta42 accumulation in a novel Alzheimer transgenic model. Am J Pathol. 2004 Oct;165(4):1289-300. PubMed.

    . Intraneuronal beta-amyloid accumulation in the amygdala enhances fear and anxiety in Alzheimer's disease transgenic mice. Biol Psychiatry. 2010 Mar 15;67(6):513-21. PubMed.

    . Modeling behavioral and neuronal symptoms of Alzheimer's disease in mice: a role for intraneuronal amyloid. Neurosci Biobehav Rev. 2007;31(1):125-47. PubMed.

    . Impaired mitochondrial dynamics and abnormal interaction of amyloid beta with mitochondrial protein Drp1 in neurons from patients with Alzheimer's disease: implications for neuronal damage. Hum Mol Genet. 2011 Jul 1;20(13):2495-509. PubMed.

    . Intraneuronal beta-amyloid aggregates, neurodegeneration, and neuron loss in transgenic mice with five familial Alzheimer's disease mutations: potential factors in amyloid plaque formation. J Neurosci. 2006 Oct 4;26(40):10129-40. PubMed.

    . 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.

    . Effects of synaptic modulation on beta-amyloid, synaptophysin, and memory performance in Alzheimer's disease transgenic mice. J Neurosci. 2010 Oct 27;30(43):14299-304. PubMed.

Make a Comment

To make a comment you must login or register.

This paper appears in the following:


  1. Intraneuronal Aβ: Was It APP All Along?


  1. Mistakes Prompt Retraction of Controversial Paper, and Publication Ban