Many studies point to endosomes as the sites of Aβ production, but the big picture may be more complicated, according to a paper in the December 7 Nature Neuroscience. Researchers led by Subhojit Roy at the University of California, San Diego, detected the convergence of APP and BACE1 in hippocampal neurons using protein chimeras that fluoresced only when the two proteins came together. Because BACE1 begins the process of snipping APP into Aβ, the results point to potential sites of Aβ production. Surprisingly, the researchers saw many encounters between APP and BACE1 in the Golgi apparatus, as well as during axonal transport. In dendrites, the researchers localized APP and BACE1 meetings specifically to recycling endosomes, which shuttle internalized proteins back to the cell surface. In addition, a rare APP variant that lowers the risk of Alzheimer’s disease associated more weakly with BACE1 in these assays, suggesting a mechanism for its protective effect. The study provides perhaps the first direct look at how BACE1 interacts with APP in neurons, the authors said.

Roy believes this fluorescence complementation assay may aid preclinical research. “I hope we can use it to discover new targets or drugs that interfere with the APP/BACE1 interaction, and thus inhibit Aβ production,” Roy wrote to Alzforum.

Other scientists liked the method as well. “This is a rigorous study with fantastic imaging,” Claudia Almeida at NOVA Medical School, Lisbon, Portugal, wrote to Alzforum. Gunnar Gouras at Lund University, Sweden, found the differences between axons and dendrites intriguing. “This is elegant neuroscience,” he said. Nonetheless, commenters noted that the study does not detect Aβ production, nor did it examine a time course of APP internalization, which might provide more clues to where Aβ is most likely to form. More work will be needed to pin down the dynamics of APP movements and Aβ generation, they added.

Researchers have long implicated endosomes as the primary sites of Aβ generation (for review, see Nixon, 2005; Rajendran and Annaert, 2012). There are many types of endosomes, however, and the details of Aβ production were hazy. Previously, Roy and colleagues labeled APP and BACE1 and found that synaptic activity stimulated the two proteins to come together in recycling endosomes of dendrites (see Aug 2013 news). Nonetheless, this type of co-localization did not prove a direct interaction, which would be needed for Aβ production.

Cozying Up in the Golgi. APP and BACE chimeras bind to reconstitute Venus protein fluorescence (green) in the Golgi (red; overlay appears yellow) in a primary hippocampal neuron (left); blocking export from the Golgi strengthens this association (right). [Courtesy of Das et al., Nature Neuroscience.]

To demonstrate a physical association, first author Utpal Das used a fluorescence complementation assay, a well-established tool of cell biologists. He conjugated one-half of Venus fluorescent protein to wild-type human APP, and the other half to mouse BACE1. The authors then transfected both constructs into mouse primary hippocampal neurons. The first fluorescence appeared about five hours after transfection, at which point APP and BACE1 protein levels remained low, so results were likely not an artifact of overexpression, the authors noted. Cells fluoresced only when the two proteins bound each other. They remained dark if the authors expressed an APP variant that did not bind BACE1.

About half of all APP-BACE1 interactions in these neurons occurred in dendritic spines. Some fluorescence also appeared in cell bodies around the nuclei, and along axons. To determine what organelles were involved, the authors co-labeled cultures with various vesicle markers. They found that the cell-body fluorescence associated mainly with the Golgi apparatus, where proteins are processed. Cooling the cultures to 20°C, which blocks export from the Golgi, enhanced this localization (see image above). In axons, Venus fluorescence also occurred in Golgi-derived vesicles. Most of these vesicles associated with presynaptic boutons, implicating synaptic compartments as the main site for Aβ production. The authors also saw movement of these vesicles along axons, suggesting that APP and BACE1 remain in close contact during axonal transport. Previously, data conflicted on whether APP and BACE1 were transported together (see Dec 2001 news; Lazarov et al., 2005; May 2006 conference news). 

In dendritic spines, 60 percent of Venus fluorescence occurred in vesicles that contained the GTP-binding protein Rab11 and the transferrin receptor. These markers distinguish recycling endosomes. Only about 30 percent of fluorescence co-labeled with the early endosomal marker Rab5, and even less occurred in late endosomes. Together, these data suggest that recycling endosomes might be more important sites of synaptic β-cleavage than early endosomes, the authors noted. Recycling endosomes are acidic, and thus have the right pH for BACE1 activity.

The authors wondered how much APP from the cell membrane ends up in recycling endosomes. They labeled cell-surface APP and measured where it appeared 30 minutes later. About half of it was in vesicles that contained transferrin receptor, supporting the idea that a large fraction of APP shuttles through recycling endosomes. However, 70 percent of cell-surface APP ended up in vesicles containing the late endosome marker LAMP1. The percentages add up to more than 100 because some vesicles may have both markers, Roy noted. Because few APP-BACE1 interactions occurred in late endosomes or lysosomes, this degradative pathway may protect APP from cleavage, the authors speculated.

The Icelandic A673T mutation in APP is known to stymie BACE1 cleavage and lower the risk of Alzheimer’s (see Jul 2012 news). The authors transfected this APP variant, conjugated with the N-terminal end of Venus fluorescent protein, into hippocampal neurons along with a BACE1-Venus C-terminal chimera. They saw much less fluorescence than in assays using wild-type APP, although APP trafficking was unaffected. These findings suggest that the protective mutation lessens the association of BACE1 with APP, the authors noted.

Some commenters had reservations about the study’s conclusions, however. Almeida pointed out that in the authors’ earlier study, relatively little APP ended up in recycling endosomes, belying the new findings. In the Venus fluorescence assay, once APP and BACE1 bind, they remain tethered, and Almeida wondered if this artificially stable interaction could override the normal sorting of APP. “It remains unclear how much the irreversible binding of APP and BACE1 in the Golgi altered their trafficking and their sites of physiological interaction,” she wrote to Alzforum.

Roy agreed that the irreversibility of the APP/BACE1 interaction does limit this assay’s ability to detect normal trafficking. However, he noted that the irreversibility is also a strength of the assay. APP and BACE1 normally associate very briefly, and so their encounters would be difficult to detect in living cells. The fluorescence assay captures these transient interactions and makes them visible, Roy wrote.

In future work, Roy plans to dissect the trafficking pathways that precede the convergence of APP and BACE1, and look for small molecules that inhibit their interaction. Such molecules might have therapeutic potential for lowering Aβ levels. “We hope that our approach will bypass some of the toxicity issues associated with enzymatic BACE inhibitors,” he wrote to Alzforum.—Madolyn Bowman Rogers.


  1. The paper by Das et al. uses a fluorescence complementation assay to visualize APP-BACE-1 interactions in cultured neurons transfected with APP and BACE-1, each tagged with one of two complementary fragments of Venus fluorescent protein. In this way, only the APP:BACE-1 complexes—not the separate proteins—become detectable (by reconstitution of the fluorescent holoprotein). The logic behind this approach is that the sites of interaction of APP with BACE-1 are also probable sites where APP undergoes cleavage by β-secretase. As expected, APP:BACE-1 complexes were detected at more than one intraneuronal location, both in somatodendritic and axonal compartments. Interestingly, APP and BACE-1 appeared to interact not only in endocytic compartments, but also in typical transport vesicles along the axon, as was proposed in earlier studies from the Goldstein lab (Kamal et al., 2001). In the soma, the detected APP:BACE-1 complexes localized to a perinuclear compartment, which the authors tentatively identify as the trans-Golgi network (TGN), a major “station” along the secretory pathway where sorting into transport vesicles is thought to occur.

    To us, the finding that APP is also proteolytically cleaved in soma, early along the secretory pathway, was not at all surprising, and confirmed results that we published since 2009, in several papers (Muresan and Muresan, 2012; Muresan et al., 2009; Muresan et al., 2013; Villegas et al., 2014). Other studies also provided strong data supporting that the perinuclear region in brain neurons in situ is a major site of accumulation of APP fragments (Pensalfini et al., 2014), and that a majority of the fragments secreted from axons in cultured human neurons (differentiated from human embryonic stem cells) are actually generated in the soma (Niederst et al., 2015).

    With regard to the identification of the perinuclear compartment where APP is cleaved, we initially explained our results by a scenario in which APP fragments are either generated in the TGN (as the Das et al. data suggest), or delivered (from another compartment, such as the endosome) to the TGN, for selective packaging into transport vesicles (Muresan et al., 2009). Yet more recent results obtained from studies of the endogenous APP, or APP tagged with small tags (but not large tags of the GFP type), strongly suggested that NH2-terminal APP fragments (NTFs) are generated at, and accumulate inside, the endoplasmic reticulum (ER) (Muresan and Ladescu Muresan, 2016). Surprisingly, using neuronal, locus coeruleus-derived CAD cells, we found that the NTFs are present within ER projections that extend into neurites, reaching the growth cone. These results suggest that the NTFs are transported to the synaptic terminal, while still inside the ER, not in typical, post-TGN transport vesicles. Indeed, we found that the accumulation of NTFs, and of bona fide ER marker proteins, such as Reticulon 4, is not at all sensitive to prolonged treatment with Brefeldin A (BFA), an agent that blocks ER-to-Golgi transport, leading to cessation of all vesicular transport along the classical secretory route (Klausner et al., 1992). Based on these, and other results, we proposed that a subdomain of the ER could function exclusively in the long-distance transport of membrane, membrane-associated, and secretory proteins, such as the NTFs (Muresan and Ladescu Muresan, 2016). The ER has in fact been implicated in the transport of RNA-binding proteins in the Xenopus oocyte, many years ago (Deshler et al., 1997). 

    Like APP, BACE-1 is present throughout the neuron, although the two proteins co-localize only at certain locations. Many studies have focused on the co-localization of BACE-1 with APP at the synapse (see, for example, Del Prete et al., 2014). In the soma, BACE-1 is preferentially localized to a perinuclear region, which may include the TGN (as shown by Das et al.). Yet, endogenous BACE-1 also accumulates at the ER, and could be active at this location (Muresan and Muresan, 2012; Muresan and Ladescu Muresan, 2016), even though the pH in the ER lumen might not optimal for its enzymatic activity (Muresan and Ladescu Muresan, 2015). One should not forget that the reticulons, which are bona fide, structural ER proteins, are major BACE-1 interacting proteins, which modulate not only BACE-1 activity, but also the generation of Aβ (He et al., 2004). 

    In fact, numerous studies (not cited here) implicate—in one way or another—the ER in the pathogenic process in AD and other neurodegenerative diseases.   

    To conclude, while APP cleavage could occur at the TGN (as the Das et al. paper suggests), other locations in the soma, such as the ER, endosomes, and lysosomes, are also candidate sites for APP processing. APP trafficking is highly dynamic, and subjected to complex regulation by factors that vary according to the physiological challenges of the neuron. As a consequence, APP transport and proteolytic processing differ not only between different types of neurons, but also between neighboring neurons of the same type, both in situ and in cell culture. As we summed up in a recent review article (Muresan and Ladescu Muresan, 2015), the real problem with the elucidation of APP transport and cleavage is that intracellular APP is an intractable protein with the current methodology. What do antibodies detect? Full-length APP, or APP fragments? What fragments? What do the tags report, the presence of full-length APP, or presence of APP fragments? Does tagged APP faithfully reproduce the biology of APP, at least with regard to processing and transport? These questions are difficult to answer. Certainly, the generation of improved constructs, with the tags placed in ways that do not interfere with the complex biology of APP, is essential.

    Zoia Ladescu Muresan contributed to this comment.


    . Kinesin-mediated axonal transport of a membrane compartment containing beta-secretase and presenilin-1 requires APP. Nature. 2001 Dec 6;414(6864):643-8. PubMed.

    . A persistent stress response to impeded axonal transport leads to accumulation of amyloid-β in the endoplasmic reticulum, and is a probable cause of sporadic Alzheimer's disease. Neurodegener Dis. 2012;10(1-4):60-3. 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.

    . Functional interaction between amyloid-β precursor protein and peripherin neurofilaments: a shared pathway leading to Alzheimer's disease and amyotrophic lateral sclerosis?. Neurodegener Dis. 2014;13(2-3):122-5. Epub 2013 Sep 4 PubMed.

    . Dual-tagged amyloid-β precursor protein reveals distinct transport pathways of its N- and C-terminal fragments. Hum Mol Genet. 2014 Mar 15;23(6):1631-43. Epub 2013 Nov 7 PubMed.

    . Intracellular amyloid and the neuronal origin of Alzheimer neuritic plaques. Neurobiol Dis. 2014 Nov;71:53-61. Epub 2014 Aug 1 PubMed.

    . Axonal amyloid precursor protein and its fragments undergo somatodendritic endocytosis and processing. Mol Biol Cell. 2015 Jan 15;26(2):205-17. Epub 2014 Nov 12 PubMed.

    . Shared Molecular Mechanisms in Alzheimer's Disease and Amyotrophic Lateral Sclerosis: Neurofilament-Dependent Transport of sAPP, FUS, TDP-43 and SOD1, with Endoplasmic Reticulum-Like Tubules. Neurodegener Dis. 2016;16(1-2):55-61. Epub 2015 Nov 26 PubMed.

    . Brefeldin A: insights into the control of membrane traffic and organelle structure. J Cell Biol. 1992 Mar;116(5):1071-80. PubMed.

    . Localization of Xenopus Vg1 mRNA by Vera protein and the endoplasmic reticulum. Science. 1997 May 16;276(5315):1128-31. PubMed.

    . APP is cleaved by Bace1 in pre-synaptic vesicles and establishes a pre-synaptic interactome, via its intracellular domain, with molecular complexes that regulate pre-synaptic vesicles functions. PLoS One. 2014;9(9):e108576. Epub 2014 Sep 23 PubMed.

    . Amyloid-β precursor protein: Multiple fragments, numerous transport routes and mechanisms. Exp Cell Res. 2015 May 15;334(1):45-53. Epub 2015 Jan 6 PubMed.

    . Reticulon family members modulate BACE1 activity and amyloid-beta peptide generation. Nat Med. 2004 Sep;10(9):959-65. Epub 2004 Aug 1 PubMed.

  2. A better understanding of axon versus dendrite interactions between APP and BACE1 is important. Das and colleagues use a new technique to identify the sites of interaction—a bicomplementation strategy that induces the irreversible binding of APP and BACE1 upon transient interaction. This technique showed that APP can transiently interact with BACE in the Golgi and remain coupled throughout its cellular itinerary. Importantly, the authors controlled for processing of the APP:VN, which is one half of the complementation assay, by BACE:VC—the other half. The authors found APP-BACE1 co-localized more with recycling endosomes than they had previously found upon synaptic activation (Das et al., 2013). I wonder if this difference is due to the artificially stable interaction between BACE1:VC and APP:VN: Because BACE1 is efficiently sorted for recycling, its sorting signal could override the APP sorting signal for degradation in the lysosome. Supporting this hypothesis, the authors observed that endocytosed APP alone (i.e., not bound to BACE1) co-localized more with Lamp1 positive late-endosomes/lysosomes than with markers of recycling endosomes. Moreover, the authors found that APP alone localized to recycling endosomes (Fig. 7f) less than when bound to BACE1 (Fig. 3d). I would have liked to see APP localization at shorter times of endocytosis. Typically, after 10 minutes APP should co-localize with early endosomes, from where it can be sorted for degradation reaching late-endosomes at a later time point and eventually lysosomes where it is quickly degraded.

    Interestingly, the authors suggest that the mechanism for the reduction of β-amyloid generation by the Icelandic APP protective mutation is via reduced interaction of APP and BACE1. Surprisingly, APP “Artic” and “London” mutations did not increase the interaction signal, possibly because the APP:VC-BACE1:VN interaction of the wild-type protein chimeras was maximal. It still remains unclear how much the irreversible binding of APP and BACE1 in the Golgi that occurs upon complementation altered their trafficking and the sites of physiological interaction of APP with BACE1. Overall, this is a very rigorous study with fantastic imaging that adds important information in elucidating the biology of neuronal APP and BACE1 in dendrites versus axons.


    . 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.

  3. We thank our colleagues for taking the time to read our paper and comment.

    Dr. Almeida's points are well taken; it is possible that there is some interference of normal trafficking after complementation (as we pointed out in the manuscript and in our comments to Alzforum). However, we note a few points.

    1) The association of BifC fragments is non-covalent, so in principle these interactions are not permanent.

    2) These techniques have been used with other proteins that traffic via the ER→Golgi (citations in the article), so this is not the first time. Clearly the VN/VC tagged proteins are not "trapped" in the ER/Golgi.

    3) Key findings—for instance axonal co-transport of APP and BACE-1—have been verified independently (without using the BifC techniques).

    4) A critical point being overlooked is that the short times after transfection (four to six hours)—barely enough for protein expression—was the best we could do to avoid prolonged associations. In most studies, transfected proteins are overexpressed for days.

    The stabilization of APP/BACE-1 interactions allowed us to see what was going on in the neuron, so this can be seen as a weakness or a strength. We tried other methods that do not stabilize these interactions and were unable to see anything meaningful. Thus we presume that these interactions are transient and require methods that stabilize the complex, which is not unusual for enzyme-substrate reactions.

    Of course every assay has strengths and weaknesses, and this one is no different. We think that when used appropriately, the assay provides useful information that can guide future research. In particular, this assay should be valuable for probing trafficking events that precede APP/BACE-1 interactions, since those pathways would not be influenced by complementation. Our conclusions are simply based on our observations, with little interpretation. We have shared these constructs with many scientists worldwide and will soon make them available through Addgene.

    Regarding Dr. Muresan’s comments, we never claimed that we were the first to see APP and/or BACE-1 in ER/Golgi. The early papers showing APP/BACE-1 in the ER/Golgi are cited in our manuscript. On the topic of citations, the journal restricted the number of articles we could cite, and many citations in the original manuscript had to be taken out during proofs. Also, we think that a “proposal” that APP/BACE interact in axons is different from actually seeing it, especially in light of conflicting evidence.

    Finally, while anyone can argue that the GFP/mCherry tags are affecting APP trafficking/cleavage; there is little evidence to support these claims. We have spent quite some time characterizing these constructs biochemically. Of note, many aspects of physiologic membrane-trafficking—stuff of textbooks—have been clarified using GFP/mCherry tagged proteins, and a carte blanche rejection of all work using tagged APP/BACE constructs is perhaps not reasonable.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Neural Activity Tips Endosomal Balance, Hastens Amyloid Pathology
  2. Suspects for Aβ Generation Spotted Together, En Route to Nerve Terminal
  3. Have APP, Will Travel
  4. Protective APP Mutation Found—Supports Amyloid Hypothesis

Paper Citations

  1. . Endosome function and dysfunction in Alzheimer's disease and other neurodegenerative diseases. Neurobiol Aging. 2005 Mar;26(3):373-82. PubMed.
  2. . Membrane trafficking pathways in Alzheimer's disease. Traffic. 2012 Jun;13(6):759-70. PubMed.
  3. . Axonal transport, amyloid precursor protein, kinesin-1, and the processing apparatus: revisited. J Neurosci. 2005 Mar 2;25(9):2386-95. PubMed.

Further Reading

Primary Papers

  1. . Visualizing APP and BACE-1 approximation in neurons yields insight into the amyloidogenic pathway. Nat Neurosci. 2016 Jan;19(1):55-64. Epub 2015 Dec 7 PubMed.