Aβ production depends on the rendezvous between the amyloid precursor protein (APP) and β-secretase (BACE-1). Two new studies explore the dynamics of this encounter. Scientists led by Subhojit Roy, University of California, San Diego, report in the August 7 Neuron, that while APP and BACE-1 are mostly shuttled about the cell separately, neuronal activity drives them together in the endosome. The finding may help explain why overactivity in neural networks fuels Aβ production. In the second study, Scott Small and Gil Di Paolo of Columbia University Medical Center find that a drop in a master regulator of endosomal trafficking strands APP in the endosomal membrane, making it a sitting duck for amyloidogenic cleavage by BACE-1 and γ-secretase. Their results are outlined in the August 2 Nature Communications.

Much evidence suggests that the amyloidogenic processing of APP occurs in endosomes (for two reviews, see Rajendran and Annaert, 2012 and Nixon, 2005), and that Aβ accumulates in the organelles early in AD, before plaques have accumulated (see Cataldo et al., 2004). However, most previous studies examined the trafficking of the relevant proteins in non-neuronal cells or in neuron-derived cell lines that lack the sub-cellular organization of neurons, such as neuroblastoma cells. Roy and colleagues wanted to examine APP and BACE-1 trafficking in cultured hippocampal neurons.

To visualize the movement of these two proteins, first author Utpal Das and colleagues transfected neurons with plasmids encoding APP and BACE-1 DNA tagged with green and red fluorescent proteins, respectively, and examined their trafficking within the somatodendritic compartment. After synthesis, the two briefly appeared together in the endoplasmic reticulum and Golgi, but then went their separate ways. Most of the BACE-1 moved about in acidic, recycling endosomes, while most of the APP traveled in Golgi-derived vesicles. There was only a small overlap between the two proteins. These results suggest that for the most part, the two proteins are shuttled separately under basal conditions.

How then, do the two proteins come together? Because recent research suggests that neuronal activity boosts Aβ production (see ARF related news story) and that this requires endocytosis (see ARF related news story), the scientists stimulated the neurons with glycine, potassium, or the GABA antagonist picrotoxin to see if it affected co-localization. After stimulation, APP entered BACE-1-containing recycling endosomes in the dendrites. The amount of β C-terminal fragments—which result from BACE-1 cleavage of APP—soon rose. If Das and colleagues blocked clathrin-mediated endocytosis from the membrane with an inhibitor called dynasore, APP and BACE-1 stayed apart despite the stimulation, suggesting co-localization depended on APP endocytosis. Roy explained that while only a small amount of APP finds its way into recycling endosomes under basal conditions, upon stimulation, much more APP moves to the plasma membrane and is endocytosed to meet BACE-1 in recycling endosomes.

Connecting the dots with human disease, the researchers examined membrane samples from postmortem brains and found that APP co-localized more with higher-density fractions enriched with BACE-1 in AD tissue, than it did in control samples.

These results suggest that the neuron has evolved to separate APP and BACE-1 and prevent their interaction, but that neuronal activity, known to trigger amyloidogenesis, breaks down that trafficking. "Our studies suggest a mechanistic pathway by which neuronal stimulation gives rise to Aβ," said Roy. "It's been known for a while that neuronal stimulation makes more Aβ, but how exactly this happens was unknown."

"This study clearly shows an activity-induced convergence of APP and BACE-1 in specialized recycling endosomes," Di Paolo wrote Alzforum in an email. "It provides additional evidence to support the idea that neuronal activity promotes the amyloidogenic processing of APP and the generation of Aβ."

Gunnar Gouras, Lund University, Sweden, who was not involved in either project, wondered how physiologically relevant the stimulation paradigms were. Typically, naturally occurring excitation is briefer than these experimental simulations, he said. He also wondered how this co-localization process changes with age to enhance pathology.

Wim Annaert, KU Leuven, Belgium, pointed out that this in vivo imaging requires overexpression of the fluorescently labeled protein. He suggested developing techniques that would elicit more physiologically relevant levels (see full comment below). He also cautioned that tags on APP's C-terminal fragment could interfere with sorting motifs. Roy commented that his group limited the amount of protein overexpression by testing neurons hours, rather than days, after transfection. He also said that the labeled BACE-1 and APP were previously shown to cleave and be cleaved like their endogenous counterparts, so their behavior and sorting likely mimic the unmodified proteins. In these experiments, they could even see that the fluorescently labeled APP co-localized with the endogenous form.

In their Nature Communications paper, Small and Di Paolo zeroed in on one particular stage of trafficking—the engulfment of APP from the endosome's membrane into vesicles within the 'late endosome.' This is an important step for delivery of cargo to the lysosome for degradation. The lipid phosphatidylinositol-3-phosphate (PI3P) is a master regulator of endosomes that helps sort cargos such as APP into these vesicles. The researchers observed low levels of this lipid in both post-mortem brains of AD patients and in mouse models of AD, including PS1, APP, and PS1/APP mice (see Chan et al., 2012). Human brain regions most affected by AD—the prefrontal cortex and entorhinal cortex—contained 40 and 25 percent less PI3P, respectively, compared to control brains. Could those low levels cause a pileup of APP in the endosomal membrane?

To find out, first authors Etienne Morel and Zeina Chamoun recreated the stunted PI3P levels in cultured mouse primary cortical neurons using short hairpin RNAs to silence a lipid kinase responsible for PI3P synthesis, Vps34. With Vps34 suppressed, APP remained on the membrane of the endosomes instead of moving into intraluminal vesicles. Endosomes grew larger and A production doubled, suggesting that more time spent on the membrane led to more amyloidogenic processing. These results suggest that targeting PI3P regulating enzymes could be a new therapeutic strategy for AD, wrote the authors.

"Our study shows that if you interfere with specific endosomal pathways, you enhance the time that APP spends with BACE-1 and γ-secretase, promoting amyloidogenic processing," Di Paolo told Alzforum. "This [disruption] creates a traffic jam in the endosomal compartment." He and the authors wrote that the study links two important defects implicated in AD pathology—lipid metabolism and endosomal sorting.

The study met with enthusiasm from several researchers. "The authors elegantly demonstrate that PI3P deficiency, as observed in AD brains and in the brains of APP transgenic mice, leads to enhanced amyloidogenic processing of APP," wrote Lawerence Rajendran to Alzforum in an email (see full comment below). "Since PI3P levels are decreased in late-onset AD cases, this offers a convincing explanation for the increased Aβ levels." He proposed examining whether APP might be secreted via exosomes, like Aβ, and how determining how β- and γ-secretase sorting are affected by a PI3P deficiency.

Ralph Nixon, Nathan Kline Institute in Orangeburg, New York, pointed out that the late endosome has not typically been a focus of AD research, but has been implicated in a number of other neurodegenerative diseases, including frontotemporal dementia, amyotrophic lateral sclerosis, and hereditary spastic paraplegia (for a review, see Nixon, 2013. "The growing literature in the neurological field supports the view that these compartments, when disrupted, are pathogenic in their own right, independently of whatever APP contributes, he told Alzforum.

Gouras agreed that the study offers much needed insight into the intricate workings of the endosomal system and its role in amyloidogenic processing of APP. That mouse models of AD show a drop in PI3P in the brain suggests that APP mutations are sufficient to cause the dearth, he suggested, making him question whether PI3P is actually the root of Aβ overproduction.—Gwyneth Dickey Zakaib


  1. This is a spectacular study, linking lipid metabolism, membrane trafficking and Alzheimer's disease. It is carefully done, combining elegant cell biology in neurons with biochemistry. For me, the most important aspect of the study is that the findings suggest that in the absence of any genetic abnormalities in the retromer components, misregulation in lipid levels could phenocopy late-onset AD symptoms.

    The role of endocytosis in the generation of Aβ peptides was identified long ago (Koo et al., 1994;. Cataldo et al., , 1997(). We previously showed that Rab5-EEA1 positive endosomes are involved in the beta-secretase processing of APP (Rajendran et al., 2006; Rajendran et al., 2008). Wim Annaert's lab demonstrated that BACE1 traffics to the early endosomes via a Arf6 mediated pathway (Sannerud et al., 2011) and we showed in an earlier work that APP endocytosis depends on cholesterol and flotillin (Schneider et al., 2008). These studies established sorting of both APP and BACE1 to early endosomes as key events in the generation of Aβ. Now, the authors looked carefully at the role of endosomal lipids, such as phophatidylinositol-3-phosphate lipids, in the sorting of APP. They demonstrate that PI3P deficiency, as observed in AD brains and in the brains of APP transgenic mice, leads to enhanced amyloidogenic processing of APP. They attribute this to altered sorting of APP.

    The authors show that APP can be sorted to the intraluminal vesicles (ILVs) of multivesicular bodies (MVBs)- a process whereby membrane proteins are either marked for degradation via lysosomes or for the removal by secretion through exosomes. They show that APP, if ubiquitinated, can be sorted to these ILVs as a means for their removal from early endosomes, where they are processed by BACE1. Misregulation of this sorting to ILVs, due to PI3P deficiency or mutation in ubiquitination signals, leads to enhanced Aβ secretion, thus linking ILV sorting of APP to downregulation of amyloidogenic processing of the protein. Since PI3P levels are decreased in late-onset AD cases, this work offers a convincing explanation for the increased Aβ levels in AD. Interestingly, early work by Anne Cataldo and Ralph Nixon has shown that the early endosomes are enlarged in AD patient brains and the current work lends supportive evidence that PI3P deficiency could contribute to this abnormal morphology and decreased ILV sorting of APP.

    Of note, since the authors discover that APP is sorted to the intra-luminal vesicles (ILVs) of the multivesicular endosomes, it would be interesting to see if some of the APP is secreted via exosomes (the vesicles that are derived from the fusion of MVBs with the plasma membrane thereby releasing their ILVs as exosomes). Since Aβ has been shown to be released via exosomes (Rajendran et al., 2006), it would be interesting to see if the precursor protein is secreted via exosomes as well. Of course, an open question is what happens to beta and gamma-secretase sorting under these conditions? Interesting times ahead for more cell biology studies in AD.


    . Evidence that production and release of amyloid beta-protein involves the endocytic pathway. J Biol Chem. 1994 Jul 1;269(26):17386-9. PubMed.

    . Increased neuronal endocytosis and protease delivery to early endosomes in sporadic Alzheimer's disease: neuropathologic evidence for a mechanism of increased beta-amyloidogenesis. J Neurosci. 1997 Aug 15;17(16):6142-51. PubMed.

    . Alzheimer's disease beta-amyloid peptides are released in association with exosomes. Proc Natl Acad Sci U S A. 2006 Jul 25;103(30):11172-7. PubMed.

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

    . Flotillin-dependent clustering of the amyloid precursor protein regulates its endocytosis and amyloidogenic processing in neurons. J Neurosci. 2008 Mar 12;28(11):2874-82. PubMed.

    . ADP ribosylation factor 6 (ARF6) controls amyloid precursor protein (APP) processing by mediating the endosomal sorting of BACE1. Proc Natl Acad Sci U S A. 2011 Aug 23;108(34):E559-68. PubMed.

    View all comments by Lawrence Rajendran
  2. The study by Das et al. was meticulously conducted and provides an important platform for further work in understanding how Aβ production normally occurs. It has been known for some time that Aβ secretion is activity-dependent, but the precise mechanisms are only now starting to be delineated. The prevailing model for APP processing was that it took place in axons/synaptic boutons and was in some way related to synaptic vesicle recycling. This study and a few other studies from the Holtzman, Malinow, and Worley groups (Verges et al., 2011; Wu et al., 2011; and Kamenetz et al., 2003) now highlight the role of APP processing and Aβ production in dendrites. Indeed, one of the putative physiological functions proposed for Aβ is to regulate postsynaptic AMPA type glutamate receptor levels (Kamenetz et al., 2003) in a negative feed-back cycle. Synaptic dysfunction in Alzeheimer’s disease thus may result from poor neuronal homeostasis.


    . Opposing synaptic regulation of amyloid-β metabolism by NMDA receptors in vivo. J Neurosci. 2011 Aug 3;31(31):11328-37. PubMed.

    . Arc/Arg3.1 regulates an endosomal pathway essential for activity-dependent β-amyloid generation. Cell. 2011 Oct 28;147(3):615-28. PubMed.

    . APP processing and synaptic function. Neuron. 2003 Mar 27;37(6):925-37. PubMed.

  3. This is a very nice study using high-level microscopy to provide evidence for BACE1 and APP converging in dendritic endosomes of cultured neurons by live cell imaging following synaptic activation. It is less clear what this work means for Alzheimer’s disease. Synaptic activity is of course fundamental to brain function. The modulation of APP processing and Aβ by synaptic activity is of considerable interest, although there are many questions. One concern is how physiological the experimental stimulation is to what occurs in brain; physiological stimulation is more rapid than what is typically employed experimentally (including in our own studies in primary neurons). A major question is why a normal process (modulation of APP trafficking and processing with synaptic stimulation) becomes abnormal with age and AD. Some of these issues could have been discussed a bit more. It is also quite interesting that this study shows greater enrichment of BACE1 with APP in human AD brain fractionation studies.

    There are some interesting aspects of this manuscript that the authors might have addressed more, for example in Fig. 3B one observes movement of APP into spines upon synaptic activation, which is not commented on. In Fig. 4D, while BACE and APP are said to come together with activity, in the kymograph BACE1 and APP actually appear together from the get-go. The authors could consider adding glycine directly during the imaging (as we have done), to provide live imaging evidence of dynamic convergence. One also wonders if transferrin receptor trafficking might be altered by activity.

    View all comments by Gunnar Gouras
  4. In their Neuron paper, Roy and co-workers used a combination of polarized neurons and in vivo imaging to scrutinize the transport routes utilized by APP and BACE1. Their rationale is that substrate and enzyme are best spatially separated to avoid default processing leading to aberrant production of proteolytic fragments. I am enthusiastic about the work, not only because of the high quality and live imaging approaches but also because the work beautifully confirms our initial proposal that transport regulation is likely to be more central in the pathology than anticipated, as we outlined some years ago in a review ((Sannerud and Annaert, 2009; and updated in (Rajendran and Annaert, 2012) and reported that for APP and BACE1 that this is indeed the case (Sannerud et al., 2011). Therein we demonstrated that APP majorly follows along a clathrin-dependent internalization route while BACE1 reaches early endosomes via a clathrin-independent route governed by the small GTPase ARF6. The authors now extend this towards neurons and support our findings that APP and BACE1 separately get internalized to endosomal compartments.

    What I find surprising is that the authors only inquired about APP's internalization, essentially confirming its dependence on clathrin. On the other hand, only one group, ours, found evidence that BACE1 and APP follow distinct internalization routes. Moreover, we identified the major route of BACE1 internalization, namely a clathrin-independent route regulated by the small GTPase ARF6. I would have expected that in this paper the authors would have confirmed this. It is reassuring that in their view much of the transport regulation and processing appears to occur in the dendritic compartments, confirming and extending our data in SCG neurons in a compartmentalized culture system.

    Although I fully agree that more cell biology is needed in neurons, at the end of the day, the major conclusions relate to transport regulations that are no different from neuronal mechanisms. A major added value is that the authors studied the effect of neuronal activity on APP-BACE1 convergence, which nicely complements earlier studies and increased awareness of the importance of activity in amyloidogenic processing.

    Much work remains to be done and further optimization of the cellular models is required to fully understand how transport regulation is intermingled with APP processing. For instance, although in vivo imaging is appealing, it requires overexpression of fluorescently labeled proteins. To achieve stable physiologically relevant levels of reporter proteins, we likely have to use viral vectors. Also the location of reporter tags can jeopardize the readout because the C-terminus of BACE1 harbors sorting motifs that may not work properly with a large tag next to it; hence live imaging should go along with confirmations of the location of endogenous proteins. Another intriguing aspect is the abundant co-localizations that the authors observe in endosomes, particularly in activity-induced circumstances. From our experience, if one co-expresses APP with BACE1 there processing occurs rapidly, as becomes evident when incubating cells with BACE1 inhibitors. Hence, the authors cannot rule out that a major part of the colocalizations are caused by the APP-CTF (hooked up to GFP) rather than full-length APP, arguing that maybe shedding has occurred earlier, closer to, or in early endosomes. Although the authors controlled this by using BACE1 inhibitors, their western blots demonstrate α-CTF as the overwhelming proteolytic fragment and hence inhibiting the small contributions of BACE1 shedding may not affect the extent of colocalization significantly. It would be interesting to repeat some of the key experiments with a dual-tagged APP to monitor in real time in neurons the actual shedding events.

    Finally, the extrapolation to AD brain needs further scrutiny. It is questionable whether endosomal (and other subcellular) compartments remain properly preserved in frozen brain sections. In summary, this paper, overall, supports the growing appreciation that endosomal compartments, and not, for instance, Golgi, are the primary sites of BACE1-mediated APP processing and that enzyme and substrate remain separated until they find each other in the “shedding compartments”.


    . Trafficking, a key player in regulated intramembrane proteolysis. Semin Cell Dev Biol. 2009 Apr;20(2):183-90. PubMed.

    . Membrane trafficking pathways in Alzheimer's disease. Traffic. 2012 Jun;13(6):759-70. PubMed.

    . ADP ribosylation factor 6 (ARF6) controls amyloid precursor protein (APP) processing by mediating the endosomal sorting of BACE1. Proc Natl Acad Sci U S A. 2011 Aug 23;108(34):E559-68. PubMed.

    View all comments by Wim Annaert
  5. This paper by Das et al. extends previous studies that indicated that endosomes—specifically, the early and recycling endosomes—are major sites of cleavage of Aβ precursor protein (APP) by β-secretase. It also provides additional support for the notion that the amyloidogenic processing of APP is potentiated by neuronal activity. Using live imaging of fluorescently tagged APP and BACE-1 (the major β-secretase), the authors also provide data supporting the hypothesis that APP is processed by diverting it to recycling endosomes, where BACE-1 normally accumulates. We agree that this is one of several ways by which amyloidogenic processing of APP could occur, and wholeheartedly recommend this paper. Here, we would like to comment on the intraneuronal sites, other than the synapse, where AD-relevant APP processing could occur.

    Das et al. focus on the processing of APP in dendrites, and the clear suggestion is that APP cleavage by β-secretase occurs within the dendrites, at postsynaptic sites. While there is no doubt that processing of APP in neuronal processes is robust, we would like to stress that the endosomes localized in neurites are not the exclusive sites of intraneuronal APP processing, and that compartments in the soma also contribute to the cleavage of APP by β-secretase. Newly synthesized APP is extensively transported to the plasma membrane, and is thereafter retrieved by endocytosis and delivered to early and recycling endosomes in the soma that contain large amounts of β-secretase. In our studies, we also find that the levels of BACE1 in the soma by far exceed those present in neurites (see Muresan and Mursean, 2006). In addition to endosomes, β-secretase is present throughout the early secretory pathway: in the endoplasmic reticulum (ER), the ER-Golgi intermediate compartment (ERGIC), the Golgi apparatus, and in the trans-Golgi network (TGN) (see Thinkaran and Koo, 2008, Vassar et al., 2009). Das et al. also find that APP and β-secretase colocalize at the level of ER and Golgi. Although the general view is that β-secretase activity in these less-acidic compartments would be low, it is hard to believe that it is completely suppressed. In fact, earlier studies have reported APP processing in the ER and ERGIC (see Cook et al., 1997).

    Those of us who listened to Charles Glabe’s talk at the AD/PD Conference in Florence earlier this year should have been impressed with the strong data suggesting that significant proteolytic processing of APP in Alzheimer’s disease could in fact occur in the neuronal soma, rather than at the synaptic terminal. To us, the interpretation that Aβ-loaded cell bodies could also be responsible for the initiation of the neuritic plaques is likely correct. Our own immunohistochemical data on paraffin sections from AD brain and on cryosections from a mouse model of AD (Bruce Lamb’s R1.40 mouse) support this interpretation. This is not to say that significant APP cleavage, and generation of Aβ, does not also occur at the synapse. We only want to draw attention to the possibility that the processing of APP, and the generation of AD-relevant APP fragments (CTFβ, sAPPβ, in addition of Aβ), also occur—perhaps to a large extent—in the neuronal soma. Although the synapse is a major site of the neuronal pathology in AD, and the processing of APP (with generation and subsequent secretion of Aβ) is likely regulated by synaptic activity (see Cirrito et al., 2008, Tampellini and Gouras, 2010), the neuronal activity could also affect processing of APP in the soma, via retrograde signaling, for example. It would be naive to rule out this latter possibility.

    Glabe, C., Conformational Diversity of Amyloid in Human AD Brain. The 11th International Conference on Alzheimer’s & Parkinson’s Disease, Florence, Italy, March 6-10, 2013. ARF related story


    . Neuritic deposits of amyloid-beta peptide in a subpopulation of central nervous system-derived neuronal cells. Mol Cell Biol. 2006 Jul;26(13):4982-97. PubMed.

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

    . The beta-secretase enzyme BACE in health and Alzheimer's disease: regulation, cell biology, function, and therapeutic potential. J Neurosci. 2009 Oct 14;29(41):12787-94. 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.

    . Endocytosis is required for synaptic activity-dependent release of amyloid-beta in vivo. Neuron. 2008 Apr 10;58(1):42-51. PubMed.

    . Synapses, synaptic activity and intraneuronal abeta in Alzheimer's disease. Front Aging Neurosci. 2010;2 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Do Overactive Brain Networks Broadcast Alzheimer’s Pathology?
  2. Link Between Synaptic Activity, Aβ Processing Revealed

Paper Citations

  1. . Membrane trafficking pathways in Alzheimer's disease. Traffic. 2012 Jun;13(6):759-70. PubMed.
  2. . Endosome function and dysfunction in Alzheimer's disease and other neurodegenerative diseases. Neurobiol Aging. 2005 Mar;26(3):373-82. PubMed.
  3. . 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.
  4. . Comparative lipidomic analysis of mouse and human brain with Alzheimer disease. J Biol Chem. 2012 Jan 20;287(4):2678-88. PubMed.
  5. . The role of autophagy in neurodegenerative disease. Nat Med. 2013 Aug;19(8):983-97. PubMed.

Other Citations

  1. PS1

Further Reading


  1. . 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.
  2. . Endosome function and dysfunction in Alzheimer's disease and other neurodegenerative diseases. Neurobiol Aging. 2005 Mar;26(3):373-82. PubMed.
  3. . Membrane trafficking pathways in Alzheimer's disease. Traffic. 2012 Jun;13(6):759-70. PubMed.

Primary Papers

  1. . Phosphatidylinositol-3-phosphate regulates sorting and processing of amyloid precursor protein through the endosomal system. Nat Commun. 2013;4:2250. PubMed.
  2. . Activity-Induced Convergence of APP and BACE-1 in Acidic Microdomains via an Endocytosis-Dependent Pathway. Neuron. 2013 Aug 7;79(3):447-60. PubMed.