CryoEM of frozen membranes has revealed its most intimate folds. Biochemistry of tissue extracts says it bundles up with other secretases. But how does γ-secretase, the all-important enzyme that cleaves Aβ42 from its precursor, truly behave in the cell membrane? In the July 9 eLife, researchers led by Wim Annaert at KU Leuven, Belgium, offer a nanoscopic glimpse of this wily intramembrane protease. Adopting various models of super-resolution light microscopy, they spied the secretase gliding through the plasma membrane alone or in pairs, but almost never in larger complexes. It lingered with substrates, but rarely dallied with other secretases. The protein complex occasionally stopped by ‘hotspots,” possibly characterized by a distinct lipid microenvironment, where it may be less active.

  • Super-resolution microscopy tracks individual γ-secretase complexes.
  • Protease travels the cell membrane alone or in pairs.
  • No evidence for megadalton super complexes with other secretases.

“Super-resolution microscopy is the type of technology we need to move the field forward and better understand how and where intramembrane proteolysis occurs,” Annaert told Alzforum. “This study is cutting-edge, and the authors have performed a robust analysis of the data,” wrote Gopal Thinakaran, University of South Florida, Tampa.

First author Abril Angélica Escamilla-Ayala and colleagues used a slew of rainbow-colored, fluorescently tagged proteins to simultaneously track γ-secretase subunits and substrates at single-molecule resolution in the cell membranes of mouse embryonic fibroblasts. This allowed them to visualize, in real time, the movement of these proteins in a way that has never been achieved before.

“This is an important technical first step toward investigating the lateral surface diffusion of γ-secretase particles in neuronal synapses, to better understand the structural and functional dynamics of this protease in the context of synaptic activity and synaptic plasticity, both in healthy and disease/AD states,” wrote Patrick Fraering, Foundation Eclosion, Plan-les-Ouates, Switzerland.

Going Stag. Nicastrin subunits tagged with either green fluorescent protein (green) or SNAP (left) rarely overlap in fibroblast membranes (right), suggesting γ-secretase exists mainly as a monomer. [Courtesy of Escamilla-Ayala et al., eLife 2020.]

First, the authors established the stoichiometry of the complex’s subunits using fluorescent presenilin (PS)1 and nicastrin (NCT) chimeras. They introduced these into fibroblasts that had their own PS1, PS2, and nicastrin genes deleted. Escamilla-Ayala checked that that the fluorescent PS1 and NCT formed functional γ-secretase complexes with endogenous PEN2 and Aph1, the other two components of the γ-secretase tetramer. Then she used Structured Illumination Microscopy (SIM) to determine the PS1-NCT ratio of γ-secretase complexes on the cell surface. The scientists examined plasma membrane sheets supported on a silica matrix to eliminate interference from complexes in the Golgi, endosomes, or other subcellular organelles. They found that most PS1 molecules in the plasma membranes were within 50 nM of one NCT molecule, and vice versa. This indicates that each subunit binds in a 1:1 ratio, since a second partner would lie further away. The finding confirms, in this type of living cell, the 1:1 ratio found in prior biochemical analyses. It is in keeping with the accepted stoichiometry of one each of the four γ-secretase subunits per mature protease.

Next, the authors looked in the plasma membrane sheet for higher-order organization. Previously, researchers working in Dennis Selkoe’s lab at Brigham and Women’s Hospital, Boston, reported that they had isolated megadalton-sized complexes comprising multiple molecules of APP, α-, β, and γ-secretase from mouse brain extracts (Jan 2019 news). Could these mega-complexes be seen at the plasma membrane?

Complex Squiggles. Analyzing tracks (colored lines, left) made by individual γ-secretase complexes as they travel through the cell membrane reveals “hotspots” (white ovals, right), where complexes meet. [Courtesy of Escamilla-Ayala et al., eLife 2020.]

Escamilla-Ayala and colleagues found no evidence for this. A next-nearest neighbor analysis of SIM data indicated that nicastrin molecules tagged with two different fluorophores were randomly distributed and rarely came into contact with each other, suggesting γ-secretase travels by itself (see image above). Other evidence supported this. For example, it took only one photobleaching step to quench fluorescence of PS1 chimeras in the complex. This means that each contained but one fluorophore—hence only one PS1. In living cells, the authors found no evidence for mega complexes, either. Here they used a super-resolution technique called PhotoGate to bleach an area, then watched while fluorescent molecules repopulated it. By measuring the intensity of the fluorescence of the molecules diffusing into the field of view, the researchers were able determine how many presenilin molecules they contained, and hence how many γ-secretase complexes. They found that about 60 percent of γ-secretase diffused as a monomer and 40 percent as a dimer; there were hardly any higher-order structures.

What about complexes with substrates? Or with other secretases? Using fluorescent antibodies and fluorescent chimeras, the authors checked to see how close γ-secretase cozied up to two of its substrates, APP and N-cadherin, and to the secretases ADAM10 and BACE1. Next-nearest neighbor analysis of SIM data showed that γ-secretase associated with its substrates, but not with ADAM10 or BACE1.

Likewise, a method called single-particle tracking coupled to photo-activated localization microscopy (SptPALM) offered little indication that secretases formed mega complexes. SptPALM tracks individual molecules as they diffuse through the cell membrane and can superimpose those tracks to determine when molecules interact (see image above). Escamilla-Ayala saw that γ-secretase and BACE1 often crossed paths. Alas, when they did their diffusion properties stayed the same, suggesting that they were merely passing each other like so many ships in the night. On the other hand, γ-secretase seemed to linger as it passed ADAM10 hotspots, but while this could in theory indicate that they interact, the authors think it’s more likely that both merely stop at the same “hotspot” in the membrane on occasion, possibly to “transfer” a substrate for the final proteolysis step, Annaert thinks (see image below).

Loner Complex. Hotspots for presenilin (green) do not overlap with those for ADAM10 (purple, left) or BACE1 (purple, right). [Courtesy of Escamilla-Ayala et al., eLife 2020.]

What those hotspots are remains a mystery. The authors found that almost 90 percent of γ-secretase complexes are mobile in the membrane, albeit not all are traveling at the same speed. This heterogeneity might reflect different conformational states or interactions with various membrane components, they believe. Using inhibitors to lock γ-secretase in a “compact” conformation it adopts while cleaving substrates did not affect diffusion, suggesting that active secretase moves about the membrane just as much as inactive does.

Despite the motility, different secretase complexes tended to visit the same hotspots, where they often paused. Annaert is trying to characterize these spots. They may contain markers of lipid rafts, implying they might provide a specific environment. Interestingly, because γ-secretase inhibitors (GSIs) reduced the number of hotspots, the authors believe they are not “hot” in terms of γ-secretase activity; instead, they may be places where the inactive complex briefly rests.

How to reconcile this new data with reports of secretase megacomplexes? “We believe a major difference in the Escamilla-Ayala et al. study is their use of exogenous expression of tagged proteins instead of genome-edited endogenous tagging,” wrote Selkoe and Lei Liu, also from Brigham and Women’s, in a comment to Alzforum (see below). “Exogenous expression of Presenilin-1, Nicastrin, BACE1, and ADAM10 could lead to unavoidable artifacts,” they added. Annaert told Alzforum that his group went to great pains to rule out overexpression artifacts. Fraering agreed. “It should be noted that, in this study, the authors paid special technical attention to minimize the potential risk of overexpression artifacts,” he wrote.

Whether the same findings reported here will hold true in neurons remains to be seen. This study is a first step toward using these methods in more relevant cell types. “Now that we have the right constructs and know the best fluorescent tags, we plan to study γ-secretase in synapses, in dendrites, and in neurites to see what the effects are of APP mutations, inhibitors, and other molecules—not only on activity but also on localization and diffusion of the protease,” said Annaert. “Then we can really understand how genetics affects cell biology in Alzheimer’s disease.”—Tom Fagan

Comments

  1. In our 2015 (Chen et al., 2015) and 2019 (Liu et al., 2019) studies, we biochemically isolated protein complexes (roughly >5 MDa) that contain the four γ-secretase complex components and ADAM10/17 or BACE1. Furthermore, in our 2019 study, we confirmed this protease complex (from cells lines, iPSC-derived neurons and wild-type human brain tissue) as proteolytically active in de novo generation of a full panel of Aβ peptides from full-length APP by sequential cleavages within the complex itself. The resultant Aβ42/40 ratio was ~0.1, i.e., physiological, supporting correct APP processing by the isolated HMW complex in vitro.

    Importantly, this Aβ-generating activity could be correctly modulated by two γ-secretase modulators in that 2019 paper and three more in our follow-up work (unpublished). We further identified a heterocyclic compound, roburic acid, which could destabilize this complex and partially separate PS/γ-secretase from BACE to reduce Aβ in vitro production by the complex (Fig. 6 in ref. 2). We also showed an activity-dependent pattern of the HMW protease complex enzymatic activities in human iPSC-derived neurons. All these biochemical data support the existence of multi-secretase complexes in normal brain and cells as well as the actual function of such protease complexes.

    As a particularly relevant comparison to the new imaging data reported by Escamilla-Ayala et al., our 2019 study had also included STED nano-scopy with 14 nm pixel resolution, as well as a proximity ligation assay, in order to visualize the endogenous BACE1/Presenilin-1 complexes. And in recent preliminary follow-up studies, we have used a LEICA SP8 FALCON FLIM microscope to show the interaction of endogenous BACE1 and Presenilin-1 in HEK293 cells labeled by the respective antibodies.

    In light of the extensive findings summarized above on endogenous complexes, we believe a major difference in the Escamilla-Ayala et al. study is their use of exogenous expression of tagged proteins instead of genome-edited endogenous tagging. Exogenous expression of Presenilin-1, Nicastrin, BACE1 and ADAM10 could lead to unavoidable artifacts. A 2D Blue Native/SDS immunoblot of 1 percent CHAPSO cell extracts should be sufficient to tell a difference between an endogenous and an exogenous γ-secretase instead of using the more dissociating 0.5 percent DDM detergent. Also, as we have mentioned in two of our relevant studies (Liu et al., 2019; Liu et al., 2019), only ~10 percent of total cellular BACE1 appears to be incorporated into the high MW protease complexes. Even the “lowest” overexpression of BACE1 would likely not incorporate sufficiently into the high MW complex.

    We would emphasize the importance of investigating endogenous proteins (here, presenilin and BACE) and their quantitative functional readout, especially in the case of the sensitive and complicated γ-secretase complex.

    Overall, we believe our biochemical, cell biological, and imaging data strongly suggest the existence of a small but enzymatically active portion of endogenous BACE1 that is associated with γ-secretase in functional HMW complexes in cultured cells, human neurons, and wild-type mouse and human brain tissue.

    References:

    . Physical and functional interaction between the α- and γ-secretases: A new model of regulated intramembrane proteolysis. J Cell Biol. 2015 Dec 21;211(6):1157-76. PubMed.

    . A cellular complex of BACE1 and γ-secretase sequentially generates Aβ from its full-length precursor. J Cell Biol. 2019 Feb 4;218(2):644-663. Epub 2019 Jan 9 PubMed.

    . Multiple BACE1 inhibitors abnormally increase the BACE1 protein level in neurons by prolonging its half-life. Alzheimers Dement. 2019 Sep;15(9):1183-1194. Epub 2019 Aug 12 PubMed.

  2. This is an interesting, very well conducted, and solid study using super-resolution microscopy to visualize the spatial and regional distribution and dynamics of individual γ-secretase particles in membranes of living fibroblasts. This challenging task confirms, in membranes from living cells, the 1:1 stoichiometry of the two γ-secretase components PS and NCT, as previously shown with purified active γ-secretase complexes (Fraering et al., 2004; Sato et al., 2007; Bai et al., 2015). 

    It further suggests a monomeric and dimeric distribution of the complexes at the cell surface. Whether γ-secretase adopts monomeric or dimeric structural organizations has been/remains a long debate, mainly because the lipids associated with purified complexes (Ayciriex et al., 2016) increase the apparent molecular weight estimation of the monomer, but also provide a biophysical hydrophobic environment adapted and prone for dimer formation.

    Finally, this study confirmed, in live cells, the physical association of γ-secretase with substrates APP or N-cadherin, but was unable to provide further evidence of high-ordered cellular “secretase clusters” composed of large multiprotein complexes containing, among other proteins, the three major secretases (α/ADAM10-, β/BACE1- and γ-secretases) regulating APP processing and Aβ production. 

    It should be noted that, in this study, the authors paid special technical attention to minimizing the potential risk of overexpression artifacts. In the future, optimal physiological conditions are expected, for example by combining live-cell nanoparticle tracking and super-resolution imaging. In such an approach, coupling quantum dots to an antibody specific for mature and active γ-secretase (the selection of which is another challenging task) would allow the tracking of γ-secretase single particles expressed endogenously in primary neuronal cultures. Recently, a similar approach has successfully been used in Daniel Choquet’s lab to measure disturbances of AMPAR surface diffusion in various rodent models of Huntington's disease (Zhang et al., 2018). 

    Altogether, this is an important technical first step toward investigating the lateral surface diffusion of γ-secretase particles in neuronal synapses, to better understand the structural and functional dynamics of this protease in the context of synaptic activity and synaptic plasticity, both in healthy and disease/AD states.

    References:

    . Purification and characterization of the human gamma-secretase complex. Biochemistry. 2004 Aug 3;43(30):9774-89. PubMed.

    . Active gamma-secretase complexes contain only one of each component. J Biol Chem. 2007 Nov 23;282(47):33985-93. PubMed.

    . An atomic structure of human γ-secretase. Nature. 2015 Sep 10;525(7568):212-7. Epub 2015 Aug 17 PubMed.

    . The lipidome associated with the γ-secretase complex is required for its integrity and activity. Biochem J. 2016 Feb 1;473(3):321-34. PubMed.

    . Modulation of AMPA receptor surface diffusion restores hippocampal plasticity and memory in Huntington's disease models. Nat Commun. 2018 Oct 15;9(1):4272. PubMed.

  3. Of note, our study focused on the distribution of γ-sec specifically at the cell surface. This should be the site where both ADAM-10 shedding and intramembrane proteolysis by γ-sec preferentially occurs, whereas for BACE1 and γ-sec this should be endosomal compartments.

    We combined several independent super-resolution approaches including a very extensive image analysis demonstrating that most γ-sec complexes are mono- and dimeric. Notably, this includes sptPALM, which is a live single-molecule localization approach (as opposed to STED) demonstrating no clusterization. At most we find hot-spot areas, i.e., zones where in time more single tracks are found and suggesting specific environments featured by maybe specific lipid or protein organization. Even these hot-spot areas do not overlap between γ-sec and those of sheddases. We rather see γ-sec molecules frequenting temporarily other hot spot areas, and we favor a more fine-tuned spatial and temporal transient “interaction” between the different sheddases, for instance upon the availability/generation of the substrate.

    Even in the case of direct interaction, our data does not provide evidence for >5MD sized complexes. For sure, it cannot be the sum of tens of γ-secretase and/or sheddases, as this should have been immediately visible with SIM and PALM. The “>5MDa” definition seems to be rather derived from the fact that the HMW complexes are solely recovered in the void fraction, i.e., the range of the applied gel filtration cannot resolve anything. It cannot be excluded that here artifacts are generated inherent to extractions in 1 percent Chapso. It is to my modest opinion also not surprising that enzyme activities are found in such extracts as enzymes and substrates can freely encounter each other post-extraction.

    With respect to our model systems, we indeed use exogenous expressed proteins. But related to γ-sec we use lentiviral vectors to stably rescue the respective KO cells and subclone for only those clones that express the reintroduced subunits at nearly the same level as the endogenous counterparts. For γ-sec this can be faithfully quality controlled by avoiding the accumulation of FL-PSEN1 and the major formation of mature NCT (further corroborated by BN-PAGE). BN-PAGE using .5 percent DDM preserves γ-sec complexes as shown by several groups including theirs. In support, BN-PAGE analysis of ER, endosomal and plasma membrane fractions clearly shows an abundance of subcomplexes in early biosynthetic compartments, whereas late compartments only have the mature 440kDa complex (unpublished) underscoring even its usefulness in analyzing complex assembly regulation. For the sheddases, cells can tolerate higher expression levels without this leading to artifacts. But, again here, we only selected lowest expression cells. Even if this is higher as endogenous, it would likely have increased the chances to encounter them in very large complexes, but that was not the case either.

    Importantly, and overall, we established here novel tools and introduced SMLM to study in unprecedented detail γ-sec diffusion live in in situ lipid bilayers. We hope this will be a steppingstone for further studies (including in neurons and in an AD context) with the potential to bridge cell biology with the challenging field of structural biology.

  4. This fantastic work shows lateral trapping of γ-secretase forming transient hot spots on the plasma membrane. The comments posted so far are insightful in calling for more high-resolution studies to understand the mechanisms of action of how these hot spots could be important in health and disease. However, it was surprising to see no mention of two associated works that have shown similar clustering of APP molecules on the plasma membrane of non-neuronal (de Coninck et al., 2018) and neuronal cells (Kedia et al., 2020). 

    Kedia et al. indicated that proteolytic processing of APP is one of the widely researched areas in the cellular neuroscience community, largely due to its central role in AD pathology. Though the biochemical pathway resulting in its proteolysis is well understood, there is no clear consensus on the probability of product formation across neuronal sub-compartments. This is, in part, due to the lack of understanding of nanoscale organization of APP in neuronal sub-compartments, and the dearth of information on the instantaneous retention and recycling rates of APP with millisecond precision. Absence of such direct readouts of the real-time chemical composition of the enzymes and substrates at the molecular resolution has limited our understanding of the parameters that define local product formation in vivo.

    In Kedia at. al., we presented a model of dynamic molecular organization of APP at synapses by combining multiple paradigms of microscopy (widefield/confocal), nanoscopy (dSTORM/STED), and novel quantitative analysis. Additionally, we address the heterogeneity in the lateral diffusion using high-density single particle tracking (sptPALM) and Universal Point Accumulation of Nanoscale Topography (uPAINT), which aids in comprehending the evolution of these nanostructures on the neuronal processes. Furthermore, we incorporate this spatiotemporal detail in silico to generate nanoscale topography, encompassing the realistic heterogeneity of APP distribution in dendrites and synapses. Major observations of this paper of interest to this community are:

    1) Nanoscale compartmentalization and differential association of APP in functional zones of the synapse forms a focus for local processing of APP. Individual excitatory synaptic regions are segregated into different functional zones, namely Endocytic Zone (EZ) and Post Synaptic Density (PSD), which differ both spatially and functionally. We identify that APP is clustered into nanoscale structures (nanodomains) within these functional zones. To understand the nature of this organization, we quantified the morphological and biophysical properties of nanodomains like size, area, density, and number of molecules per nanodomain. Additionally, we also calculated the total number of APP molecules associated with different functional zones of the synapse. These molecular signatures varied between functional zones of the synapse as well as between individual synapses. This implies that such variability in the compositionality is directly linked to the molecular identity of the associated functional zones, controlling the locus of APP processing. 

    2) Lateral exchange and immobilization of APP molecules in and out of these nanodomains dictate an equilibrium between free and confined pools of APP in the membranes of live hippocampal neurons. We identify that the molecular signatures of nanodomains of APP varied across synapses. Using high-density single-particle tracking, we confirm that the association and dissociation of these domains are regulated by lateral diffusion. We show that laterally diffusing APP molecules are transiently immobilized in these clusters and there exists an equilibrium between nanodomain and extra-nanodomain APP molecules. Furthermore, we found that the immobilization kinetics differed between wild-type APP and the Swedish APP mutation implicated in familial AD, and the APP-Icelandic variant is considered protective. This was the case in both neurons and heterologous cell lines (see Mueller et al., 2017, for comprehensive review). These observations point out that even minor alterations in individual APP molecules can affect its lateral diffusion, influencing the instantaneous availability of APP molecules per unit area.

    3) Biophysical reconstruction of APP organization reveals the nanoscale heterogeneity of APP in dendrites and synapses. With the aid of realistic nanoscale reconstruction, we can now visualize the compositionality of APP with precision of a few tens of nanometers. Further, the mapping of nanoscale topography of APP allows us to unveil its variations in different functional zones of a synapse, implicating a physiologically regulatable transient organization of APP at each synapse. Extensive biochemical investigations have facilitated tremendous insight into the roles of different secretases in the processing of APP. However, the mechanism behind how the optimal concentrations of various constituents of amyloidogenic processing evolve in different subcellular and synaptic compartments remains elusive. In Kedia et al., we present a nanoscale model of an excitatory synapse with realistic parameters that highlight the heterogeneity of APP organization within each sub-compartment, thus illustrating the role of stochasticity of APP composition influencing its regulation and processing.

    We also point out that components of amyloidogenic machinery can form supramolecular clusters that could be controlled by lateral diffusion. This  supports the hypothesis derived in this current paper. These observations of local differences in the compositionality of diffusive versus immobilized APP point toward heterogeneity in the product formation in neurites and synaptic zones, which can be taken together with Escamilla-Ayala et al.

    In Kedia et al., we also show that this stochasticity in the organization is not only associated with molecular localization but also with evolution and regulation of these nanodomains dictated by random processes like lateral diffusion.

    Considering the crucial role of APP as a substrate for several enzymatic processing, and γ secretase as a key enzyme that is involved in proteolytic processing, it will be key to understand how this organization is controlled in neurons in health and disease. We hope both these studies set a new platform for using advanced microscopy to understand synaptic nano-organization and regulation within synapses.

    References:

    . Real-time nanoscale organization of amyloid precursor protein. Nanoscale. 2020 Apr 21;12(15):8200-8215. Epub 2020 Apr 7 PubMed.

    . Packing Density of the Amyloid Precursor Protein in the Cell Membrane. Biophys J. 2018 Mar 13;114(5):1128-1141. PubMed.

    . Not just amyloid: physiological functions of the amyloid precursor protein family. Nat Rev Neurosci. 2017 May;18(5):281-298. Epub 2017 Mar 31 PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. BACE and γ-Secretase Form Mega-Complex that Processes APP

Further Reading

Primary Papers

  1. . Super-resolution microscopy reveals majorly mono- and dimeric presenilin1/γ-secretase at the cell surface. Elife. 2020 Jul 7;9 PubMed.