The first atomic-level structure of γ-secretase reveals how individual amino acids interact within the four-unit complex that churns out Aβ. Described August 17 in Nature, the 3.4 Angstrom resolution cryo-electron microscopy (cryo-EM) structure illuminated the configuration of the 20 transmembrane regions that weave in and out of the cell, and also offered hints about how the complex contorts to snip substrates. The researchers, led by Yigong Shi at Tsinghua University in Beijing and Sjors Scheres at Cambridge University in England, used the structure to map out many of the known presenilin mutations linked to familial Alzheimer’s disease, pinpointing “mutation hotspots” that face the enzyme’s core. They measured Aβ production from 10 of the mutants, and found that while some trigger an uptick in γ-secretase activity and others turn it down, they all tipped the balance toward production of Aβ42 over Aβ40.

Sam Sisodia of the University of Chicago called the structure a tour de force. “A 3.4Å resolution structure via cryo-EM is virtually unheard of; it doesn’t get much better than this,” he said. Sisodia added that while the paper allows researchers to dissect the structure of the complex like never before, further studies that examine how the complex changes shape when bound to substrates or inhibitors will be crucial to understand how it catalyzes proteolysis in the membrane.

This sub-four Angstrom feat tops the 4.32Å structure unveiled by Shi’s group this spring (see May 2015 news). The researchers also obtained the previous structure using cryo-EM, which freezes proteins in place before zooming in on them with a beam of electrons. That structure allowed the researchers to assign each of the 20 transmembrane regions in the complex to one of the secretase’s four subunits: presenilin 1 (the catalytic core of the enzyme), nicastrin, APH-1, or PEN-2. 

Covered Horseshoe. The four subunits of γ-secretase make a lopsided horseshoe with a nicastrin lid (green). A previous study assigned each of the 20 transmembrane domains to a specific subunit, and the new atomic-level structure confirmed those assignments. [Courtesy of Bai et al., Nature 2015.] 

With the better resolution the researchers confirmed that γ-secretase forms a lopsided horseshoe-shaped structure with thick and thin ends, lying on its side in the membrane and adorned with an extracellular lid. Except for its single transmembrane domain jutting into APH1, nicastrin was primarily extracellular and formed a cap over the horseshoe. APH1 made up the bulk of the horseshoe’s thick end, and one of presenilin’s nine transmembrane (TM) regions fit snugly into a pocket made up of two APH1 TM domains. Presenilin rounded out the horseshoe and linked up on the thin end with PEN2, containing three TMs. A key finding from Shi’s previous paper was that the two TM domains (TM6/7) that contain PS1’s catalytic aspartate residues reside on the convex, not the concave, side of the horseshoe. This suggested that substrates enter the complex not through the middle of the horseshoe, which would be intuitively satisfying, but rather through a lateral route on the outside of the structure. The new structure confirmed this configuration, and offered up new insights about the enzymatic acrobatics that could be necessary to accommodate a substrate. 

To move beyond TM assignments and view γ-secretase at the atomic level, co-first authors Xiao-chen Bai at Cambridge and Guanghui Yang at Tsinghua ramped up the magnification of EM and bombarded millions of γ-secretase particles. The massive data set of more than 400,000 particles that revealed a clear structure yielded the 3.5Å map of the complex. The researchers then employed a new algorithm they developed to single out the most homogenous 160,000 particles for analysis. This allowed them to squeeze out that extra 0.1Å resolution from the data. In their prized 3.4Å structure, most of the amino acids in the transmembrane regions of the complex could be identified, except for much of TM2 and TM6 of presenilin, which were poorly resolved. The fuzzy resolution of these two TMs, which the researchers had also noted in their previous study, suggested these regions were highly flexible and could assume a variety of configurations. This flexibility could also help explain how the secretase manages to accommodate such a variety of substrates with wildly different sequences.

Activity with Help from a PAL?
The atomic structure revealed the surprising location of presenilin’s catalytic residues. Not only were they on the convex side of the horseshoe, as had been observed before, they were also more than 10Å apart: Asp257 was located in the middle of TM6 slightly toward the extracellular side, while Asp385 sat near the cytoplasmic side of TM7. This is much too far for the hydrogen bonding needed for catalysis to occur, which indicates that the complex must undergo a conformational change to bring the two residues together and snip a substrate. The authors speculated that this could occur via a proline-alanine-leucine motif that resides on TM9. The structure revealed that this motif, which was previously reported to facilitate substrate binding and activate the enzyme, lies between the two catalytic aspartates (see Wang et al., 2006Sato et al., 2008).

“Clearly, this is the model of an inactive γ-secretase complex that must go through extensive conformational changes to bring the catalytic aspartate residues to hydrogen bonding distance,” commented Lucia Chavez-Gutierrez of KU Leuven in Belgium. “This snapshot supports the view that γ-secretase exists as an ensemble of multiple conformations and their equilibrium is functionally relevant.”

The atomic structure of nicastrin revealed no surprises, and supported current hypotheses about this subunit’s role in substrate binding. Nicastrin consists of two lobes—a large one harboring a proposed substrate binding pocket containing Glu22 and Tyr337, and a small one containing a lid that stretches out to cover this pocket (see red sequence in image below). The researchers previously proposed that this lid rotates around a pivot point centered at Phe287. Now they report that a greasy hydrophobic pocket surrounds this phenylalanine, which may keep the hinge well-oiled. They proposed that entering substrates somehow trigger the lid to pivot and lift, which then opens up the binding pocket to position the substrate for processing.

Open Sesame. Researchers propose that nicastrin (small lobe, green; large lobe, blue) binds substrates via Tyr337 and Glu333. A lid covering this binding pocket (red) may flip open to allow substrates to bind (right). [Courtesy of Bai et al., Nature 2015.] 

The high-resolution structure unveiled many interactions between residues in the complex. Van der Waals contacts between hydrophobic residues made up the bulk of bonding within the membrane. In addition, two phospholipids—one wedged in between TM1/TM8 of PS1 and TM4 of APH1, and the other between APH1 and nicastrin’s lone TM—stabilize interactions between the subunits, suggest Bai and colleagues. Taisuke Tomita of the University of Tokyo wondered about the lipids’ potential role in γ-secretase function. “It is well known that composition of lipids also affects enzymatic activity,” he wrote. “It would be very interesting to test whether these lipids are essential for γ-secretase structure and activity in a similar manner to that observed in several channels and receptors.”

Lipid Linkage.

Two phospholipids are shown embedded between different subunits of γ-secretase. The researchers propose they bring subunits together. [Courtesy of Bai et al., Nature 2015.]

Mutation Hotspots
The researchers used the atomic structure to map out the locations of PS1 mutations linked to familial AD. Some 212 mutations affect 135 residues, of which 53 have more than one mutation. Of these most variable residues, 35 were located in TM domains and resolved clearly in the structure. The researchers found a majority of them fell within two mutational hotspots: TM2-5 and TM6-9. While the mutations were clearly scattered across different regions, most of them faced the inner core of the complex. Interestingly, this alignment of FAD mutations on specific facets of the helices had been proposed more than a decade ago (see image below and Hardy and Crook, 2001). While TM6 and 7 contain the catalytic residues of PS1 and thus could have obvious links to function, mutations in the other TMs could affect substrate recognition or positioning, the researchers speculated. 

Helical Skew.

This extracellular view of PS1 reveals that most of the 35 residues affected by FAD-linked mutations (in orange) lined up on the same side of each helix. [Courtesy of Bai et al., Nature 2015.]

To get an idea of how different mutations affected γ-secretase function, the researchers generated 10 PS1 mutants, purified them, and mixed them with the C99 fragment of APP to monitor processing in vitro. They found that four of the mutations led to strong reductions in Aβ40 and Aβ42 production, two producing no detectable amounts of Aβ40. Three other mutations had little impact on γ-secretase activity, while three others increased cleavage activity. The researchers concluded that these variations in proteolysis refuted the idea that an overall impairment in γ-secretase function leads to AD in these mutation carriers. All of the mutations that produced detectable levels of Aβ40 and Aβ42 elevated Aβ42/40 ratios to varying degrees, but the researchers said that this did not prove those heightened ratios were the cause of AD either. 

The fact that PS1 mutations have differing effects on γ-secretase activity is not a surprise, said John Hardy and Rita Guerreiro of University College London in a joint comment. Similar conclusions have been drawn from other biochemical studies (see Chavez-Gutierrez et al., 2012). 

Sisodia wondered why Shi and colleagues selected these 10 particular mutations to measure γ-secretase activity, as none were among the most well-characterized FAD mutations. Shi said the researchers were measuring activity of the remaining mutations using their in vitro assay. They also plan to examine the structure of the complex when bound to substrates as well as γ-secretase modulators, to see how the enzyme changes in response to binding. This information could also help guide rational design of γ-secretase modulators, which aim to block the production of Aβ42 while sparing processing of other important substrates.—Jessica Shugart


  1. Shi and colleagues delight us with the first atomic view of the γ-secretase complex!

    The atomic model shows many interesting features, but what I find most provocative and exciting is the plasticity/flexibility implied by this static view to make the structure compatible with an active and highly allosteric protease. Clearly, this is the model of an inactive γ-secretase complex that must go through extensive conformational changes to bring the catalytic Aspartate residues, located more than 10Å from each other, into hydrogen bonding distance. This snapshot thus highlights how important protein conformational changes are in the γ-secretase complex, which very well supports the view that γ-secretase exists as an ensemble of multiple conformations and that their equilibrium is functionally relevant (Elad et al., 2015).

    Furthermore, the active site of PS1 is accessible from the convex side of the transmembrane (TM) horseshoe, while the putative substrate binding pocket (Glu333) in the nicastrin extracellular domain (ECD) actually faces the concave side. Participation of the nicastrin ECD in substrate recruitment, as proposed by Shah et al., would require a large conformational rearrangement in the complex (Shah et al., 2005). However, it should be kept in mind that the nicastrin-substrate interaction model remains controversial, as it has been challenged by two independent groups (Chavez-Gutierrez et al., 2008; Zhao et al., 2010). The new atomic model will probably guide experimentalists to further resolve this controversial issue. 

    The reported atomic structure places about two-thirds of the residues mutated in familial Alzheimer disease (FAD) in two “hotspots” in PSEN1, each located at the inner core of a four-TM bundle (just a note: several FAD-linked mutations have been identified in TM1). How mutations in these hotspots translate into the common kinetic mechanism (impaired carboxypeptidase-like activity) observed for FAD-linked mutations (Chavez-Gutierrez et al., 2012) awaits further investigation. I am certain that the atomic model will serve as a structural frame for this task. 

    Finally, I would like to emphasize that the activity data provided by Shi and colleagues on FAD-linked mutations confirms that total γ-secretase loss of function is not commonly observed among disease-causing mutants (Chavez-Gutierrez et al., 2012) and therefore it cannot be the basis of the pathogenic mechanism.


    . The mechanism of γ-Secretase dysfunction in familial Alzheimer disease. EMBO J. 2012 May 16;31(10):2261-74. Epub 2012 Apr 13 PubMed.

    . Glu(332) in the Nicastrin ectodomain is essential for gamma-secretase complex maturation but not for its activity. J Biol Chem. 2008 Jul 18;283(29):20096-105. PubMed.

    . The dynamic conformational landscape of gamma-secretase Nadav. J Cell Sci. 2015 Feb 1;128(3):589-98. PubMed.

    . Nicastrin functions as a gamma-secretase-substrate receptor. Cell. 2005 Aug 12;122(3):435-47. PubMed.

    . Gamma-secretase composed of PS1/Pen2/Aph1a can cleave notch and amyloid precursor protein in the absence of nicastrin. J Neurosci. 2010 Feb 3;30(5):1648-56. PubMed.

  2. This paper by Drs. Scheres, Shi, and colleagues represents a milestone in research on the structural biology of γ-secretase. They have utilized an impressive array of methodologies that took tremendous effort. This study provides several molecular insights into the unique intramembrane protease.

    The structure fits very well with previous biochemical analyses in regard to intratransmembrane domains and intermolecule associations. They also confirmed novel intimate interactions by cross-linking methodologies. While we know we have an accurate global structure for γ-secretase, we await the structure within the lipid bilayer. It is well known that the composition of lipids in the bilayer also affects the enzymatic activity. Consistent with this, the researchers found phospholipids embedded in the γ-secretase complex. It would be very interesting to test whether these lipids are essential for γ-secretase structure and activity in a similar manner to that observed for several channels and receptors (reviewed in Cornelius et al., 2015; Taberner et al., 2015).

    Their study also highlights a yet-unsolved issue of the γ-secretase: structural dynamics during the endoproteolysis. We and others have reported conformational changes of the γ-secretase by biochemical experiments (Takeo et al., 2012; Takagi-Niidome et al, 2013; Takeo et al., 2014; Takagi-Niidome et al., 2015). Intriguingly, structures of some parts of PS (i.e., N-terminal region, TMD2, Loop 1, Loop 6) have not been resolved in this new structure. Also, the distance of the catalytic aspartates is farther than that of activated aspartic proteases. Further structural analysis of γ-secretase with inhibitors/modulators and molecular dynamics simulation would provide novel insights into the mechanistic action of the intramembrane proteolysis.


    . General and specific lipid-protein interactions in Na,K-ATPase. Biochim Biophys Acta. 2015 Sep;1848(9):1729-43. Epub 2015 Mar 16 PubMed.

    . TRP channels interaction with lipids and its implications in disease. Biochim Biophys Acta. 2015 Sep;1848(9):1818-27. Epub 2015 Mar 30 PubMed.

    . Contribution of the γ-secretase subunits to the formation of catalytic pore of presenilin 1 protein. J Biol Chem. 2012 Jul 27;287(31):25834-43. Epub 2012 Jun 11 PubMed.

    . Inhibition of γ-secretase activity by a monoclonal antibody against the extracellular hydrophilic loop of presenilin 1. Biochemistry. 2013 Jan 8;52(1):61-9. Epub 2012 Dec 14 PubMed.

    . Allosteric regulation of γ-secretase activity by a phenylimidazole-type γ-secretase modulator. Proc Natl Acad Sci U S A. 2014 Jul 22;111(29):10544-9. Epub 2014 Jul 9 PubMed.

    . Cooperative roles of hydrophilic loop 1 and the C-terminus of presenilin 1 in the substrate-gating mechanism of γ-secretase. J Neurosci. 2015 Feb 11;35(6):2646-56. PubMed.

  3. This study by Bai and colleagues resolves the γ-secretase structure to an atomic level (resolution of 3.4Ǻ). This type of detail allows the authors to map and analyze some of the reported PSEN1 disease-associated mutations and to start to develop clearer ideas about the mechanisms of the mutations. As predicted in an earlier study (Hardy and Crook, 2001), many of the mutations align along transmembrane (TM) helical faces. Bai and colleagues now analyzed 35 residues in the TMs that correspond to 101 Alzheimer’s mutations, and identified two mutational hotspots in TMs 2-5 and 6-9. In both cases the affected residues have their side chains facing the inner core of the TMs, with only one side of each TM helix being affected.

    This structural resolution of PSEN1 yields information that can be used when assessing pathogenicity of novel variants identified in the gene. We have previously proposed a decision tree, where the location of the variants influenced the prediction of pathogenicity (Guerreiro et al., 2010). When a new variant is identified, it should be modeled using this high-resolution structure to help in the decision regarding its pathogenicity.

    Bai and colleagues show that the mutations do not cause a consistent change in the protease activity of γ-secretase.  In fact, this was already known from biochemical analysis of the effects of presenilin mutations on γ-secretase (Chavez-Gutierrezet al. 2012). Whether this detailed knowledge of γ-secretase structure will lead to more or less work on developing inhibitors and modulators of this complex is debatable (De Strooper, 2014). And we have to remember that there are many different γ-secretases based on different presenilins, different APHs, and different splice forms of the enzyme. Clearly, therefore, this is a great step forward, but much more needs to be done.


    . The mechanism of γ-Secretase dysfunction in familial Alzheimer disease. EMBO J. 2012 May 16;31(10):2261-74. Epub 2012 Apr 13 PubMed.

    . Lessons from a failed γ-secretase Alzheimer trial. Cell. 2014 Nov 6;159(4):721-6. PubMed.

    . Genetic screening of Alzheimer's disease genes in Iberian and African samples yields novel mutations in presenilins and APP. Neurobiol Aging. 2010 May;31(5):725-31. Epub 2008 Jul 30 PubMed.

    . Presenilin mutations line up along transmembrane alpha-helices. Neurosci Lett. 2001 Jun 29;306(3):203-5. PubMed.

  4. I agree with Sam Sisodia that assessing the structure of other mutations would be useful. An obvious one is the D9 mutation which has a clear effect on function (Perez-Tur et al., 1995; Thinakaran et al., 1996).


    . A mutation in Alzheimer's disease destroying a splice acceptor site in the presenilin-1 gene. Neuroreport. 1995 Dec 29;7(1):297-301. PubMed.

    . Endoproteolysis of presenilin 1 and accumulation of processed derivatives in vivo. Neuron. 1996 Jul;17(1):181-90. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Gamma Secretase: Intramembrane Liaisons Revealed

Alzpedia Citations

  1. Presenilin-1 (PSEN1)

Paper Citations

  1. . C-terminal PAL motif of presenilin and presenilin homologues required for normal active site conformation. J Neurochem. 2006 Jan;96(1):218-27. PubMed.
  2. . The C-terminal PAL motif and transmembrane domain 9 of presenilin 1 are involved in the formation of the catalytic pore of the gamma-secretase. J Neurosci. 2008 Jun 11;28(24):6264-71. PubMed.
  3. . Presenilin mutations line up along transmembrane alpha-helices. Neurosci Lett. 2001 Jun 29;306(3):203-5. PubMed.
  4. . The mechanism of γ-Secretase dysfunction in familial Alzheimer disease. EMBO J. 2012 May 16;31(10):2261-74. Epub 2012 Apr 13 PubMed.

Other Citations

  1. mutations 

Further Reading


  1. . The dynamic conformational landscape of gamma-secretase Nadav. J Cell Sci. 2015 Feb 1;128(3):589-98. PubMed.

Primary Papers

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