Journal of Cell Science Accepted...

32
1 Title: The dynamic conformational landscape of γ-secretase Authors: Nadav Elad 1,2 , Bart De Strooper 1,2, 5,# , Sam Lismont 1,2 , Wim Hagen 3 , Sarah Veugelen 1,2 Muriel Arimon 4 , Katrien Horré 1,2 , Oksana Berezovska 4 , Carsten Sachse 3,# and Lucía Chávez- Gutiérrez 1,2,# 1 VIB Center for the Biology of Disease, 3000 Leuven, Belgium 2 Center of Human Genetics, University Hospitals Leuven & Department of Human Genetics, KU Leuven, and Leuven Research Institute for Neuroscience and Disease (LIND), 3000 Leuven, Belgium 3 European Molecular Biology Laboratory, Structural and Computational Biology Unit, Meyerhofstr.1, 69117 Heidelberg, Germany 4 Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA 5 UCL Institute of Neurology, Queen Square, London, UK # Corresponding authors: Bart De Strooper, MD, PhD: VIB center for the biology of disease, Herestraat 32 3000 Leuven, Belgium. +3216373246 [email protected] Carsten Sachse, PhD: [email protected] Lucía Chávez-Gutiérrez, PhD: [email protected] Journal of Cell Science Accepted manuscript © 2014. Published by The Company of Biologists Ltd. JCS Advance Online Article. Posted on 12 December 2014

Transcript of Journal of Cell Science Accepted...

Page 1: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

1

Title:

The dynamic conformational landscape of γ-secretase

Authors:

Nadav Elad1,2

, Bart De Strooper1,2, 5,#

, Sam Lismont1,2

, Wim Hagen3, Sarah Veugelen

1,2 Muriel

Arimon4, Katrien Horré

1,2, Oksana Berezovska

4, Carsten Sachse

3,# and Lucía Chávez-

Gutiérrez1,2,#

1 VIB Center for the Biology of Disease, 3000 Leuven, Belgium

2 Center of Human Genetics, University Hospitals Leuven & Department of Human Genetics,

KU Leuven, and Leuven Research Institute for Neuroscience and Disease (LIND), 3000 Leuven,

Belgium

3 European Molecular Biology Laboratory, Structural and Computational Biology Unit,

Meyerhofstr.1, 69117 Heidelberg, Germany

4 Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston,

MA, USA

5 UCL Institute of Neurology, Queen Square, London, UK

# Corresponding authors:

Bart De Strooper, MD, PhD: VIB center for the biology of disease, Herestraat 32 3000 Leuven,

Belgium. +3216373246 [email protected]

Carsten Sachse, PhD: [email protected]

Lucía Chávez-Gutiérrez, PhD: [email protected]

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t© 2014. Published by The Company of Biologists Ltd.

JCS Advance Online Article. Posted on 12 December 2014

Page 2: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

2

Abstract

The structure and function of the γ-secretase proteases are of vast interest because of their critical

roles in cellular and disease processes. We established a novel purification protocol for γ-

secretase complex that involves a conformation and complex-specific nanobody, yielding highly

pure and active enzyme. Using single particle electron microscopy, we analyzed the γ-secretase

structure and its conformational variability. Under steady state conditions the complex adopts

three major conformations, which are different in overall compactness and relative position of the

nicastrin ectodomain. Occupancy of the active or substrate binding sites by inhibitors

differentially stabilize sub-populations of particles with compact conformations, whereas a

Familial Alzheimer Disease-linked mutation results in enrichment of extended-conformation

complexes with increased flexibility. Our study presents the γ-secretase complex as a dynamic

population of inter-converting conformations, involving rearrangements at the nanometer scale

and high level of structural interdependence between subunits. The fact that protease inhibition or

clinical mutations, which affect Aβ generation, enrich for particular subpopulations of

conformers indicates the functional relevance of the observed dynamic changes, which are likely

instrumental for highly allosteric behavior of the enzyme.

Keywords:

γ-secretase / Alzheimer disease / Conformational landscape / Regulated intramembrane

proteolysis / Single particle electron microscopy

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 3: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

3

Introduction

γ-Secretase complexes have the remarkable property to cleave integral membrane proteins within

the lipid bilayer. Proteolytic products provide signals to the extra- and the intracellular

environment, coordinating physiological and pathological processes at both sides of the cell

membrane (De Strooper and Annaert, 2010; Selkoe and Wolfe, 2007). Best studied are the roles

of γ-secretases in the processing of the amyloid precursor protein (APP) and the secretion of the

neurotoxic amyloid β (Aβ) peptides in the context of Alzheimer disease (AD). However, other

substrates, including Notch, N-cadherin and ErbB4, link γ-secretase activities to development,

cancer and immunity (De Strooper and Annaert, 2010; Selkoe and Wolfe, 2007).

Presenilin 1 (PS1), nicastrin (NCT), presenilin enhancer 2 (PEN-2) and anterior pharynx

defective 1A (APH-1A) assemble in a tetrameric complex (Fig. 1A) (De Strooper, 2003) and PS1

autoproteolysis results in an active pentameric γ-secretase (Thinakaran et al., 1996), in which the

catalytic center is structured at the interface between the amino- and the carboxy-terminal

fragments of PS (NTF and CTF respectively) (Esler et al., 2000; Li et al., 2013; Li et al., 2000;

Wolfe et al., 1999) and connected to the intracellular aqueous environment (Sato et al., 2006;

Tolia et al., 2006). NCT, a type 1 integral membrane glycoprotein, plays a critical role in

complex maturation and stabilization (Chávez-Gutiérrez et al., 2008) and might be involved in

substrate binding (Shah et al., 2005; Zhang et al., 2012), although this is debated (Chávez-

Gutiérrez et al., 2008). The recent crystal structure of the NCT ectodomain (ECD) shows a two-

lobed architecture and an interface of van der Waals and hydrogen bonds between them (Xie et

al., 2014). APH-1 and PEN-2 contribute seven and two transmembrane domains (TMDs) to the

protease complex respectively and their roles are poorly understood. The first high resolution

(4.5Å) structure of γ-secretase was very recently obtained using single particle electron

microscopy (EM) (Lu et al., 2014). The model presents NCT ectodomain and traces the 19 TMDs

organized in a horseshoe shaped membrane core, the flexible loops connecting the TMDs and the

hydrophilic N- and C- termini of PS1 and PEN-2 were not visualized due to their high flexibility.

Although it represents a major step towards the molecular understanding of γ-secretase

mechanisms, additional functional-structural information on the complex is needed to understand

how γ-secretase cleaves different substrates and how Alzheimer causing mutations affect its

activity.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 4: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

4

γ-Secretase is a highly regulated allosteric enzyme with at least two conformations under

equilibrium: differential subunit composition (Serneels et al., 2009), aging-related

posttranslational modifications by oxidative stress (Guix et al., 2012; Wahlster et al., 2013),

pathogenic mutations in PS1 (Berezovska et al., 2005), and interaction with substrates or γ-

secretase modulators (GSMs) (Uemura et al., 2010), all shift γ-secretase conformational

equilibrium and regulate protease activity. With the exception of Li and colleagues (Li et al.,

2014), previous single-particle EM studies reported individual structures providing static views

of the complex (Lazarov et al., 2006; Li et al., 2014; Lu et al., 2014; Ogura et al., 2006;

Osenkowski et al., 2009; Renzi et al., 2011). Li and colleagues (Li et al., 2014) reported

differences in the relative position of the NCT ECD in the presence or absence of a non-active

site inhibitor (compound E), providing the first glimpse of structural changes triggered by a non-

active site inhibitor. However, the study focuses on the characterization of single conformations,

rather than on the distribution of (multiple) conformations under varying conditions (“allosteric

ensemble”).

Based on the highly allosteric behavior of γ-secretase, we postulate that this multimeric complex

adopts multiple conformations in steady state conditions and that ligand (inhibitor, substrate,

allosteric modulators) binding or presence of a pathological mutation (mutant PS) in its active

site alters the equilibrium between conformers. In this view, the catalytic activity of γ-secretase is

defined by the “average activity” of a set of co-existing conformers, which equilibrium depends

on presence of ligands, mutations, lipids, etc. To test our hypothesis we purified γ-secretase

using a complex-specific nanobody (Nb) and analyzed the resulting pure and active γ-secretase

by single particle EM. We show that wild-type γ-secretase complex exists as an ensemble of at

least three distinct major conformations (compact, intermediate and extended), which are in

dynamic equilibrium. Inhibition, substrate binding or the PS1 Familial Alzheimer Disease

(FAD)-causing mutation delta Exon 9 remodel the conformational landscape of the protease by

enriching specific subpopulations that already preexist in the ligand-free state.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 5: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

5

Results

A novel purification protocol based on an anti-γ-secretase complex-specific nanobody

results in highly pure and active enzyme

We expressed γ-secretase in insect cells using a baculovirus expression system. We used one step

purified γ-secretase complex (Fig. 1B, lane 1 and 1C, lane 1) (Acx et al., 2014) to generate

single chain camelid anti-γ-secretase nanobodies (Nbs) (Pardon et al., 2014). We identified the

conformational Nb30, which immunoprecipitates γ-secretase in 1% CHAPSO, a condition that

preserves integrity and activity of the complex, but fails to immunoprecipitate any of the γ-

secretase subunits in 1% Triton X-100 (TX-100) (complex dissociating conditions) (Capell et al.,

1998) (Fig. 1E, F). Nb30 does not affect γ-secretase activity in an in vitro activity assay (data not

shown). Therefore it is unlikely that binding of Nb30 changes γ-secretase conformation, although

we cannot discard this possibility. We then incorporated Nb30 in the purification of active γ-

secretase (see material and methods). The protocol yields highly pure and active untagged γ-

secretase complex (Fig. 1B, lane 2 and 1C, lane 2). We also screened for detergents that keep γ-

secretase active and stable. Gel filtration to exchange CHAPSO to Lauryl Maltose-Neopentyl

Glycol (LMNG) results in active (Fig. 1D) and stable γ-secretase complex. A typical EM

micrograph of negatively-stained particles (Fig. 2A) shows that this purification procedure results

in monodisperse particles of the expected size.

In order to verify the integrity of the complex, we performed antibody labeling with antibodies

against NCT ECD, NCT C-terminus (NCT CT), PS1 CTF, APH-1 and PEN-2, imaged the bound

complexes by EM and performed 2D analysis. Class averages of antibody-bound enzyme and

free γ-secretase were aligned, enabling identification of the bound antibody as an extra density

protruding from the γ-secretase projection (Fig. 2B-D, more examples in supplementary material

Fig. S1). By analyzing the 2D class averages, we identify the extended density as the NCT ECD

(Fig. 2C and D, indicated in dark green). As expected, NCT CT antibodies (light green) were

found at opposing position to the NCT ECD antibodies. In the membrane core, APH-1 CT

domain antibodies were localized directly beneath the NCT ECD antibodies and the C-terminal

part of PS1-CTF antibodies close to the APH-1 epitope, but distant from the NCT ECD

antibodies. PEN-2 antibodies localize at a distant position from the extended domain. Notably, all

antibodies bound at 1:1 stoichiometric ratio, confirming the integrity of the complex.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 6: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

6

Wt γ-secretase complex exists as an ensemble of different conformers

We investigated the architecture of the γ-secretase complex by single particle EM. Data sets are a

priori low-pass filtered to 25 Å in order to avoid any bias of the alignment from noise. EM

images were analyzed in 2D using reference-free alignment and classification. The class averages

show variability in the membrane core and in the relative position of the extended domain density

(Fig. 3A). Since variability in 2D projections can result from different conformations or different

orientations, multiple initial 3D maps were reconstructed using the random conical tilt (RCT)

method (Radermacher et al., 1987) resulting in 3D structures with resolutions of 23 to 25 Å (Fig.

3B-D, supplementary material Fig. S2A-F). From a total of 30 3D classes we found three

different structural states: extended, intermediate and compact which account for 17%, 30% and

17% of the aligned structures respectively (Fig. 3E-G). Classes differ in overall compactness of

the complex, mainly seen as a large hinge rotation of NCT ECD towards the membrane core. We

obtained the same three significant structures using unsupervised 3D classification with RELION

software (Scheres, 2012) (supplementary material Fig. S2G), confirming both the overall

architecture of γ-secretase structures and the proposed conformational heterogeneity in the

preparation. The compact structure resembles previously published EM structures (Li et al., 2014;

Lu et al., 2014), including a membrane core, surrounded by a detergent belt and the NCT ECD.

Differences were found in the detergent belt, which appeared rougher and thicker in our

structures. This can be attributed to the different detergents used in addition to the negative stain

method. However, we noted that our obtained structures vary considerably in dimensions and

domain architecture, indicating that the complex fluctuates between different conformations

under steady state conditions.

We defined the length of the major axis as a good and simple indication of the population

diversity (Fig. 3H). Indeed, a dramatic 30% (5 nm) difference between the compact and extended

structures is observed along this axis. Extended, intermediate and compact are 16-17 nm, 14-15

nm and 12-13 nm long, respectively. Importantly, we provide evidence below that, similarly,

large differences in the conformation of the complex are observed under physiological conditions

in the cell membrane (Fig. 5 and 6).

Inhibitor binding to the γ-secretase complex shifts its conformational equilibrium

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 7: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

7

We hypothesized that the equilibrium between the different conformations observed in our

experiments (defines protease activity at the macromolecular level) is shifted upon ligand binding

to the protease. We tested this by evaluating the shifts in the equilibrium between the different

conformers induced by binding of inhibitors to the protease. The transition-state analog inhibitor

X (inhX, L-685,458) and inhibitor peptide15 (pep15) bind the catalytic site and the initial

substrate site, respectively. Inhibitor binding, according to our hypothesis will stabilize particular

conformations and thus shift the conformational equilibrium. Indeed, class averages of the γ-

secretase-inhX complex (Fig. 4A, supplementary material S3B) appear rounded and less variable

at the NCT ECD density when compared to untreated γ-secretase (Fig. 3A). Accordingly, inhX

narrows the distribution and shifts the average length of the major axis from 15.3 nm (untreated,

Fig. 3H) to 12.8 nm (Fig. 4B, C, supplementary material S3A, B), indicating that inhibitor bound

γ-secretase is more compact and less structurally heterogeneous. The reduced structural

heterogeneity observed in the inhX-treated particles enabled us to validate the overall correctness

of the structure by tilt-pair analysis (Rosenthal and Henderson, 2003) (supplementary material

Fig. S4).

Next, we used a γ-secretase inhibitor pep15, which is a helical peptide that binds to the putative

initial substrate binding site in the complex (Das et al., 2003). Similar to inhX, pep15 enriches for

a compact conformation (Fig. 4D-F, supplementary material S3C, D), but NCT ECD is

positioned in the majority of the structures at the edge of the membrane core (Fig. 4E).

Accordingly, the average of the length of the major axis is 13.7 nm (Fig. 4F).

We asked whether the large conformational changes we observe in our in vitro system could also

be observed in γ-secretase complexes in their native state, when embedded in the cell membrane.

We therefore studied the effect of inhX on the conformation of the γ-secretase using

fluorescence-lifetime imaging microscopy (FLIM) on intact cells. We hypothesized that

treatment with inhX would increase the proximity of the NCT ECD to the membrane core and

therefore to its intracellular domain (Fig. 5A). Thus, we followed the interaction of antibodies

targeting NCT ECD and NCT CT (cytoplasmic domain) by FRET. Increased FRET efficiency in

HEK293 cells treated with 1 μM inhX, compared to non-treated cells, confirms that NCT ECD

approaches the membrane core (Fig. 5B).

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 8: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

8

We wondered whether the increased compactness of the γ-secretase observed in the presence of

these two inhibitors of very different classes would affect the stability of the complex. We

therefore performed co-immunoprecipitation (coIP) experiments of endogenous γ-secretase,

using a nanobody against NCT ECD (Nb28) (Fig. 1E, F), in CHAPSO, which keeps the complex

together, or TX-100, which normally dissociates the complex (Fig. 6 lane 4). In agreement with

Barthet et al. (2011) (Barthet et al., 2011) we find that γ-secretase is stable in the presence of

inhX and partially stable in pep15 in TX-100 (Fig. 6 lanes 5 and 6). Thus, remodeling the

conformational landscape of γ-secretase upon inhibitor binding is observed in EM and translates

into γ-secretase stabilization in TX-100.

FAD mutation remodels the γ-secretase conformational landscape by enriching an extended

conformation

We wondered what the effect of a FAD mutation would be on the conformational landscape of

the complex. FAD mutation PS1 delta exon9 (Δ9) misses part of the extracellular loop between

PS1-CTF and -NTF and is therefore expected to provide visible alterations to the complex. It

should be noted that, similar to other FAD mutations, Δ9 results in an altered Aβ production

profile (Chávez-Gutiérrez et al., 2012; Perez-Tur et al., 1995). We expressed and purified FAD

Δ9 γ-secretase complex, collected and processed EM images as indicated above. Windowed

particles and class averages show the same features as wt-γ-secretase, indicating that the Δ9

complex maintains grossly the features of wild type γ-secretase (Fig. 7A). NCT ECD was visible

as a separate density in more than a third of the class averages and appeared remarkably mobile,

adopting different shapes and variable positions with respect to the membrane core (Fig. 7A and

B). This made 3D reconstruction very difficult and therefore, we limited our investigation by

analyzing the conformational variability in the γ-secretase-Δ9 using 2D eigenimage analysis (Fig.

7B-D) (Elad et al., 2008; Elad et al., 2007). Using this approach the global variability in the data

set is reduced by classifying the particles first based on their overall shape, allowing the

identification of more subtle local variability within classes as density peaks in eigenimages.

Interestingly, the occurrence of significant positive and negative density peaks indicates

systematic structure variation in the region of the NCT ECD, even in class averages that

contained no obvious density for the NCT ECD (due to high domain flexibility) (Fig. 7C, Δ9 last

image). Wt γ-secretase and inhX bound γ-secretase data sets show significantly weaker density

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 9: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

9

peaks in eigenimage, although some variation in the NCT ECD region is observed in extended-

conformation classes of wt γ-secretase (Fig. 7C, yellow square). Finally, we confirmed an

enrichment of intermediate to extended conformations due to a highly mobile NCT ECD, as seen

in the length of the major axis (Fig. 7E, supplementary material S3E-G).

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 10: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

10

Discussion

The “ensemble model” for allostery states that the diversity and distribution of conformations

present in solution underlies enzyme activity (Boehr et al., 2009; Motlagh et al., 2014). In this

view, the conformational landscape, rather than specific ‘active’ or ‘inactive’ conformations,

provides the basis for the biochemical function of proteins. The relevance of this hypothesis has

been demonstrated in structural studies, for example of the regulation of G protein-coupled

receptors (GPCRs) (Nygaard et al., 2013), the ribosome (Fischer et al., 2010), the transcription

factor TFIID (Cianfrocco et al., 2013), or the integrins (Takagi et al., 2002).

We present here a first characterization of the γ-secretase structure dynamics. The most

remarkable conclusion form our work is that γ-secretase exists as an ensemble of conformers

(Fig. 8). The equilibrium between conformers is modulated by inhibitor binding to the active or

substrate binding sites providing functional relevance to this observation. It is also interesting that

a mutation causing Alzheimer Disease increases the conformational heterogeneity of the

complex. Therefore, we suggest that clinical mutations in Presenilin might cause an overall

destabilization of the structure leading to increased flexibility. Such destabilization could

speculatively result in destabilization of the enzyme-substrate state and promote product release,

providing an explanation for the higher release of longer Aβ peptides from the complex in the

presence of FAD mutations (Chávez-Gutiérrez et al., 2012).

Importantly, independent biochemical experiments show that inhX and pep15 binding, which

resulted in increased compactness of γ-secretase in the EM images, dramatically affected the

biophysical stability of the complex in a dissociating detergent (TX-100). In addition, our FLIM

data supports the inference that such large conformational shifts of NCT ECD happen in intact

cells. In agreement, NCT ECD is critical for the assembly, maturation and stability of γ-secretase

(Chávez-Gutiérrez et al., 2008). Furthermore, several publications using FLIM showed how γ-

secretase undergoes conformational changes in the intact cell membrane (Berezovska et al., 2005;

Li et al., 2014; Lleo et al., 2004; Shelton et al., 2009; Uemura et al., 2010). Thus, molecular

interactions affect γ-secretase function by changing the frequencies of subpopulations in the

conformational landscape. Allosteric behavior, substrate affinity and the catalytic mechanisms

are therefore determined by the ensemble-average. Thus, the observed conformational landscape

truly reflects the inherent dynamics of γ-secretase.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 11: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

11

The observed dynamics have particular implications for future structural and functional studies.

Structural homogeneity is determinant for protein crystallization and the generation of high

resolution data by EM. It should be noted that the recently published high resolution structure (Lu

et al., 2014) is based on a careful selection of about 5% to 10% of the particles from the purified

material. We suggest that such ”in silico” purification and the use of amphipol masked potential

structural heterogeneity enabling to build the high resolution model. The observed heterogeneity

might also explain why different γ-secretase structures published until now (Lazarov et al., 2006;

Li et al., 2014; Ogura et al., 2006; Osenkowski et al., 2009; Renzi et al., 2011) are rather

divergent. The current study suggests that pre-incubation with inhX may help to reduce

heterogeneity. In addition, other ligands, e.g. nanobodies that stabilize the protease complex in

particular conformations may also help studies aiming to elucidate the atomic structure of the

complex.

On the functional side, the conformational fluctuations observed in the present study indicate that

a more complete description of the γ-secretase catalysis has to take into account the

interconverting conformers that may exist at each step of the catalytic reaction. i.e. the

conformational fluctuations give rise to ensembles of Michaelis-Menten Enzyme-Substrate (ES),

Enzyme-Product (EP) complexes and their corresponding transition states, and these ensembles

will undergo equilibrium shifts depending on environment. In fact, the concept that multiple

enzyme conformers catalyze reactions in parallel is studied by “catalytic networks” (Benkovic et

al., 2008). Incorporation of dynamics in the studies of γ-secretase intramembrane proteolysis is

thus critical for understanding its function and elucidating its structure at high resolution.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 12: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

12

Materials and methods

Antibodies

The following antibodies were used for labeling: affinity purified rabbit polyclonal anti-APH-1a

(B80.3) and affinity purified monoclonal anti-NCT CT 9C3 have been described (Annaert et al.,

2001; Esselens et al., 2004). Affinity purified rabbit polyclonal against N-terminal amino acids

103-124 of NCT (anti-NCT ECD, ab24741) and affinity purified rabbit polyclonal against 13

amino acids at the C-terminus of PEN-2 (anti-PEN-2, ab62514) were purchased from Abcam.

Rabbit polyclonal against C-terminal amino acids 450-467 of PS-1 (anti-PS1 CT, S182) was

purchased from Sigma. The following antibodies were used for detection in WB: Rabbit anti

PEN-2 (B126.1) (Annaert et al., 2001; Esselens et al., 2004), Anti human PS1–loop (MAB5232),

purchased from Chemicon, mouse anti-NCT from BD Transduction Laboratories and rat anti-PS1

NTF (MAB1563) from Millipore, anti-FLAG M2 from Sigma.

Immunoprecipitation with nanobodies

Hi5-cells overexpressing the γ-secretase complex were lysed in either 1% CHAPSO or 1% Triton

X-100 for 1h on ice. Unsoluble material was removed by ultracentrifugation 45 min at 100 000g.

100μg solubilized material protein was incubated with the indicated nanobodies, which were

previously covalently coupled to NHS-beads (GE Healthcare) accordingly to manufacturer

instructions. 25 µl Nb beads (0.3 mg Nb/ml beads) were used per condition.

Immunoprecipitations were carried out in PBS buffer containing 1% CHAPSO or 1% Triton X-

100 in a total volume of 250μl, overnight at 4°C. Bound and unbound fractions were analyzed by

SDS-PAGE and western blot.

Overexpression and purification of wt and Δ9 γ-secretase

γ-Secretase complex was overexpressed in Hi5 insect cells from a single vector carrying the four

subunits. NCT was expressed as a GFP fusion. Purification followed three steps (Fig. 1B, C).

Expression and first step purification using an anti-GFP nanobody were performed as recently

published(Acx et al., 2014). A second immune-affinity step, performed with a conformational

nanobody His tagged (Nb30) that recognizes the γ-secretase complex and not the individual

subunits (Fig. 1E, F), enriched for mature and active protease complex. Nb30/γ-secretase

complex was purified by immobilized metal ion affinity chromatography (IMAC) and eluted with

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 13: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

13

250 mM imidazole. Finally, gel filtration in a Superose6 column (GE-Healthcare) removed

unbound Nb and exchanged detergent to 0.1% LMNG. Purified γ-secretase complexes were

tested in an in vitro activity assay as previously described (Acx et al., 2014).

FLIM-FRET

HEK293 cells were seeded at 150,000 cells per well on glass coverslips in 12-well. After 24h

medium was refreshed and 1 μM of L-685,458 (inhX, Calbiochem) or DMSO was added. After

overnight incubation, cells were fixed in 4% paraformaldehyde for 10 min, blocked (2% BSA,

2% FBS and 1% gelatin in PBS supplemented with 5% serum) and incubated with primary

antibodies (anti-NCT ECD ab24741 and anti-NCT CT 9C3) overnight at 4°C. Cells were

subsequently incubated with fluorophore (Alexa488 or Alexa555)-conjugated secondary

antibodies for 1 h at room temperature, washed with TBS and mounted in Vectashield H-1000

(Vector Labs).

The relative proximity between fluorophore labeled NCT ECD and CT was assessed by the

previously validated Fluorescence Lifetime Imaging Microscopy (FLIM) analysis as described

(Berezovska et al., 2005; Uemura et al., 2010). Briefly, the baseline lifetime (t1) of the Alexa488

fluorophore (negative control, FRET absent) was measured in the absence of an acceptor

fluorophore. Upon excitation of the donor fluorophore in the presence of Alexa555 acceptor

fluorophore, the donor lifetime shortens (t2) due to FRET if the donor and acceptor are less than

5-10 nm apart. The degree of donor fluorophore lifetime shortening (FRET efficiency) correlates

with the change in proximity between the two fluorescently labeled domains of NCT molecule.

The percent FRET efficiency is calculated using the following equation: %EFRET=100*(t1-t2)/t1.

A mode-locked Chameleon Ti:Sapphire laser (Coherent Inc., Santa Clara, CA) sending a

femtosecond pulse every ~12.5 nsec was used to excite the donor fluorophore (two-photon

excitation at 800 nm wavelength). Donor fluorophore lifetimes were recorded using a high-speed

photomultiplier tube (MCP R3809; Hamamatsu, Bridgewater, NJ) and a fast time-correlated

single-photon counting acquisition board (SPC- 830; Becker & Hickl, Berlin, Germany).

SPCImage software (Becker&Hickl, Berlin, Germany) was used to determine the donor

fluorophore lifetimes by fitting the data to one (negative control) or two (acceptor present)

exponential decay curves.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 14: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

14

Sample Preparation for EM

Purified γ-secretase sample was diluted 1:10 to 0.035 mg/ml in buffer without detergent (25 mM

PIPES pH 7.4, 150 mM NaCl, 5% Glycerol) immediately prior to fixation on the EM grid. We

diluted the detergent in order to reduce background noise in EM micrographs. γ-Secretase was

stable and active at the final concentration of LMNG (0.01%), which is 10x the critical micelle

concentration (Chae et al., 2010). Carbon-coated 400 mesh grids were glow-discharged and then

incubated 1 min with a 3 μl drop of 0.1 mg/ml inert nanobody, which has no affinity to γ-

secretase and no effect on the structural analysis due to its small size (15 kDa). We found that this

pretreatment of the carbon surface improved staining quality. 3 μl of purified sample was then

applied and negatively stained with 2% uranyl acetate. For preparation of the γ-secretase-inhX

and γ-secretase-pep15 complexes, purified sample was diluted in buffer containing 10 μM of

inhX or 1 μM pep15 respectively, and incubated for 1 h at room temperature. Antibody labeling

for EM imaging was performed during the second step of the purification while γ-secretase was

bound on the Nb30 beads. Beads with bound γ-secretase were incubated overnight at 4°C with

the specified antibodies followed by washing and elution with 250 mM imidazole.

EM Data Collection and 2D/3D image processing

Wt γ-secretase, γ-secretase-pep15 tilt pair (0° and 55°) and antibody-labeled γ-secretase data sets

were collected on a Gatan US4000 CCD camera at 47,000 magnification (1.8 Å/pixel) using a

Titan Krios (FEI, Eindhoven) operated at 200 kV and controlled using SerialEM (Mastronarde,

2005). γ-Secretase-inhX data set was collected on a TVIPS 4k x 4k CCD camera at 83,333

magnification (1.8 Å/pixel) using a Phillips CM120 Biotwin Ice operating at 120 kV. γ-Secretase

Δ9 data set and γ-secretase-inhX tilt pairs (0° and 25°, for tilt pair validation) were collected on

TVIPS F224HD 2k x 2k CCD camera (Gauting, Germany) at 113,744 magnification (2.1

Å/pixel) using a FEI T12 operating at 120 kV. Imaging was done using a defocus range of 0.7–

1.3 µm and an electron dose of 30 e-/Å

2. The contrast transfer function (CTF) was determined for

each micrograph using CTFFIND3 (Mindell and Grigorieff, 2003). Particles from micrograph of

all samples except antibody labeled data sets were selected interactively using EMAN2 (Tang et

al., 2007). Particles of the antibody labeled data sets were selected automatically using a template

matching algorithm implemented SPIDER (Frank et al., 1996; Rath and Frank, 2004). Particles

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 15: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

15

were extracted to yield boxes of 256 x 256 pixel dimension, and phase corrected using SPIDER.

After CTF correction, the box size was cropped to 160 x 160 pixels, images were binned by a

factor of 2 and band-pass filtered between 25 Å and 200 Å. Filtered particles were subjected to

10 iterations of reference-free alignment and classification using IMAGIC (van Heel et al., 1996),

starting with classes of 20 images and gradually increasing the number of images per class to

about 500, depending on sample homogeneity.

For RCT reconstruction of the wt γ-secretase 9,774 tilt-pairs were selected interactively using

EMAN2. The 0° tilt particles were subjected to reference-free alignment and classification,

resulting in 30 classes. A total of 30 3D RCT maps were reconstructed from the corresponding

tilted particles and each 3D map was refined internally against its tilted particles using IMAGIC.

Initial RCT structures were then refined iteratively with decreasing angular increment against a

data set of 36,093 0° tilt particles using SPIDER. A cross-correlation threshold was applied every

round to select best matching images for each structure. Additionally, in order to minimize the

effect of overpopulated orientations on the angular distribution, the number of images per

orientation was limited for 3D reconstruction. In total 6000-8000 molecular views (~20% of the

data set) were included in each structure. The FSCs from ‘gold standard’ refinement half-sets

(Scheres and Chen, 2012) indicate resolutions from 23 to 25 Å at FSC=0.143 for wt structures

(supplementary material Fig. S2A, C, E). Structures that refined until convergence, showed the

common features as described in the results above, and displayed good match between class

averages and re-projections were subsequently aligned in 3D and subjected to 3D classification.

Most of the structures featured preferred orientation with the γ-secretase extracellular side

contacting the carbon support. Therefore, structures could be roughly aligned by manual rotation

about the axis perpendicular to membrane plane. Manual alignment was followed by 4 iterations

of automated 3D alignment using UCSF Chimera (Pettersen et al., 2004), in which a 3D global

average was calculated and individual structures aligned to the global average structure.

Structures showing poor cross-correlation were discarded during the process. Aligned 3D

structures were subjected to multivariate statistical analysis (MSA)-based hierarchical ascendant

classification using IMAGIC. Unsupervised 3D classification in RELION (Scheres, 2012) was

performed using the 3D global average, filtered to 80 Å, as an initial model. The low-pass filtered

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 16: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

16

data was classified into 10 3D classes, which showed the same architecture and heterogeneity as

above, therefore independently confirming the results.

γ-secretase-inhX and γ-secretase-Δ9 data sets contained 9,267 and 11,443 particles respectively.

The 30 initial RCT models from the wt data (filtered to 50 Å) were used as starting models for

3D reconstruction of both data sets. We used the same refinement procedure as for the wt, except

that about 3000-4000 images were included in each structure. 22 and 25 of the γ-secretase-inhX

γ-secretase-Δ9 structures respectively were aligned and classified as described above for the wt γ-

secretase structures. Structures of all data sets were aligned to the same coordinates.

γ-Secretase-pep15 data set was collected and processed as described for wt γ-secretase above.

6777 tilt pairs were selected interactively and classified into 15 classes using reference-free

alignment, which were then used to reconstruct 15 3D maps using the RCT method. 13 maps

were used for 3D alignment and classification.

Antibody labeled data sets contained 10,000 particles each. Data sets were processed separately

by reference-free alignment and classification as described above, followed by classification into

classes of 20 images. Class averages of poor quality were discarded, and the remaining class

averages were aligned to class averages of the wt γ-secretase (non-labeled) data set using

SPIDER. We then selected class averages from the labeled data set showing the same γ-secretase

features as in the matching class average of the non-labeled data set, and containing an extra

density at the periphery. Only extra density appearing consistently at the same position relative to

γ-secretase density was interpreted as bound antibody.

Tilt pair test (Rosenthal and Henderson, 2003) for validation of the overall correctness of the

inhX-treated structure was performed on 46 particles taken from a single micrograph using

EMAN2 e2tiltvalidate procedure.

Fitting of the high resolution map as a rigid body (Lu et al., 2014) was done automatically using

UCSF Chimera (Pettersen et al., 2004), following by small manual adjustments of the NCT ECD

atomic coordinates position.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 17: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

17

Acknowledgements

This work was supported by the Fund for Scientific Research Flanders (FWO); research fund KU

Leuven; the Hercules Foundation, Federal Office for Scientific Affairs (IAP P7/16); a

Methusalem grant of the Flemish Government, VIB, IWT, the European Research Council (AdG

to BDS), BDS is the Arthur Bax and Anna Vanluffelen chair for Alzheimer disease. NE was

supported by Marie Curie and European Molecular Biology Organization (EMBO) fellowships,

and by the Foundation for Alzheimer Research (SAO-FRA). OB was supported by NIH

AG15379 grant. We also acknowledge an anonymous foundation.

Figure legends

Figure 1 - γ-Secretase purification using complex-specific Nb30.

(A) Membrane topology of the five γ-secretase subunits. Subunits include Presenilin 1 amino-

terminal (PS1-NTF) and carboxy-terminal fragment (PS1-CTF), which contain the catalytic

aspartates. Nicastrin (NCT), is a type 1 integral membrane glycoprotein of 130 kDa comprising a

large ectodomain (ECD), Anterior pharynx defective 1 (APH-1), is a 30 kDa protein of seven

TMDs topology, and presenilin enhancer 2 (PEN-2) is a 12 kDa hairpin-like protein with two

TMDs.

(B,C) Coomassie-stained SDS-PAGE and western blot following the three purification steps: (1)

Anti-GFP Nb affinity chromatography and elution by protease cleavage, (2) Binding to complex-

specific Nb30-His and IMAC, (3) Size-exclusion chromatography. Colored arrow heads mark the

different subunits according to the color scheme in panel (A).

(D) Activity assay of the purified γ-secretase as measured by AICD generation from APP

C99x3FLAG substrate. Increasing amounts of AICD product (lower lanes) is seen with

increasing concentration of purified γ-secretase (lanes 2-5). Additionally, purified γ-secretase is

inhibited specifically by transition-state analogue inhX (lane 1). E and F).

(E, F) Anti γ-secretase complex-specific Nb30. Nb30 immunoprecipitates (IPs) only the

assembled γ-secretase complex, whereas it fails to bind γ-secretase under dissociating conditions

(TX-100). Hi5 insect cells overexpressing γ-secretase complexes were solubilized in either 1%

CHAPSO (E) or 1% Triton X-100 (F) and used in IP experiments with Nbs 28, 30 and 125. 1%

CHAPSO keeps active γ-secretase complex, while 1% Triton X-100 dissociates the complex.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 18: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

18

Shown are SDS-PAGE/Western blot analyses of input, unbound and bound fractions.

Immunostainings for γ-secretase subunits indicate that Nb30 binds γ-secretase when the complex

is stable (1% CHAPSO, E), but not the individual subunits (1% Triton X-100, F). In contrast,

Nb28 binds both γ-secretase complex and dissociated NCT subunit. Nb125 is a non-interacting γ-

secretase Nb used as negative control.

Figure 2 - EM characterization of purified γ-secretase complex.

(A) Representative micrograph of the purified γ-secretase. Scale bar: 50 nm.

(B,C) Class averages of free (B) and antibody-labeled (C) γ-secretase showing corresponding

orientations. Extra density resulting from the bound antibody is pointed by an arrowhead for each

of the different antibodies in (C). The density of the anchored Fab is enhanced in class averages,

whereas other domains of the antibody, which move independently, are averaged out.

(D) Outlines of the densities in (B), γ-secretase is in grey and antibody density in colors.

Additional class averages in supplementary material Fig. S1.

Figure 3 - Variable conformation of wt γ-secretase.

(A) Representative class averages used for 3D reconstruction by the RCT method. Outlines of the

densities are shown next to class averages, membrane core in grey and NCT ECD in green.

Separation into domain is based on the reconstructed 3D maps.

(B-D) Representative structures from the different 3D classes - extended, intermediate and

compact. Structures are displayed from the side (viewing into the membrane plane). Cryo-EM

structure from (Lu et al., 2014) filtered to 30 Å is shown in the inset in grey for comparison.

(E-G) 3D class averages calculated by MSA-based hierarchical ascendant 3D classification. The

proportion of aligned RCT structures represented by each class is indicated.

(H) Length distribution of the γ-secretase structures based on measurements along their major

axis.

Figure 4 – Differential stabilization of the compact conformation induced by inhX and

pep15.

(A, D) Representative class averages from the γ-secretase-inhX and γ-secretase-pep15 data sets

respectively.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 19: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

19

(B, E) 3D averages and representative RCT structures from the γ-secretase-inhX and γ-secretase-

pep15 data set respectively, displayed from the side as the untreated γ-secretase structures (Fig

3B-G).

(C, F) Length distribution of the different γ-secretase-inhX nd γ-secretase-pep15 structures

respectively, based on measurement along their major axis.

Figure 5 - InhX-induced NCT ECD movement observed in HEK293 cells using FRET-

FLIM.

(A) The experimental setup is illustrated in the cartoon: FRET-FLIM was used to monitor the

distance between γ-secretase NCT ECD and NCT CT in HEK293 cells using antibodies carrying

donor and acceptor fluorophores.

(B) FRET efficiency was considerably higher following inhX treatment, indicating that the NCT

ECD has moved closer to the NCT CT on the cytosolic side. P = 0.0003 by unpaired Student’s t

test, bars represent std. error.

Figure 6 - InhX and pep15 treatments differentially increase the association of γ-secretase

subunits in HEK293 cells.

Co-IP of γ-secretase subunits from HEK293 cells using an anti-NCT nanobody (Nb28, Fig 1). IP

was either performed in Triton X-100 buffer following O/N incubation with 1 μM inhX (lanes 2

and 5) or 10 μM pep15 (lanes 3 and 6). InhX treatment stabilized the γ-secretase complex against

dissociation in TX-100, whereas pep15 induces partial stabilization compared to DMSO treated,

supporting the EM results.

Figure 7 - Conformational landscape altered by PS1-Δ9 mutation.

(A) Class averages comprising a small number of images (~10) per class, showing the expected

membrane core and NCT ECD domains as observed in the wt. The position of the NCT ECD

varies considerably.

(B) Illustration of the technique for identification of local variability in 2D classes: data-sets are

sorted into classes of similar features (represented in 2D class-averages). Then, in order to assess

local variability within classes, the images within each class are subjected to MSA and the

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 20: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

20

resulting eigenimages examined. Strong positive and negative density peaks in eigenimages

indicate local variability.

(C) Representative class averages from the γ-secretase-Δ9 data set (left panel, first row) and

corresponding eigenimage analysis of the images within classes (left panel, second row). Green

squares mark the position of significant positive and negative density peaks in eigenimages,

indicating systematic local variation at these positions. The same analysis was done for the class

averages of the wt γ-secretase and γ-secretase-inhX (right panel), resulting in significantly

weaker density peaks in eigenimages, although some peaks were observed in the extended

conformation class averages (yellow square).

(D) Quantification of local variability in the wt γ-secretase (yellow), γ-secretase-inhX (blue) and

γ-secretase-Δ9 (green). Reference-free aligned images from the three data sets were classified

into 30 classes and eigenimages calculated within each class. Density of the highest and lowest

peaks were measured in eigenimages 2 and 3 (eigenimage 1 represents the total average). γ-

secretase-Δ9 shows significantly stronger density in peaks compared to wt γ-secretase and γ-

secretase-inhX, indicating local variations.

(E) Length distribution of the different γ-secretase-Δ9 structures (supplementary material Fig.

S3E-G) based on measurement along their major axis.

Figure 8 - Energy landscape view of γ-secretase

Schematic energy landscape view of the conformational diversity in wt γ-secretase and its

rearrangement in γ-secretase-inhX, γ-secretase-pep15 and γ-secretase-Δ9 populations. Wt γ-

secretase fluctuates between three major conformations. InhX and pep15 differentially stabilize

the most compact conformation observed in the untreated γ-secretase, whereas the Δ9 mutation

increases conformational flexibility and enriches the subpopulations with intermediate to

extended conformations. The overall conformational landscape is multidimensional, however

here it is reduced to one dimension, concentrating on the length of the major axis.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 21: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

21

References

Acx, H., Chavez-Gutierrez, L., Serneels, L., Lismont, S., Benurwar, M., Elad, N. and De Strooper, B. (2014). Signature amyloid beta profiles are produced by different gamma-secretase complexes. J. Biol. Chem. 289, 4346-55. Annaert, W. G., Esselens, C., Baert, V., Boeve, C., Snellings, G., Cupers, P., Craessaerts, K. and De Strooper, B. (2001). Interaction with telencephalin and the amyloid precursor protein predicts a ring structure for presenilins. Neuron 32, 579-89. Barthet, G., Shioi, J., Shao, Z., Ren, Y., Georgakopoulos, A. and Robakis, N. K. (2011). Inhibitors of gamma-secretase stabilize the complex and differentially affect processing of amyloid precursor protein and other substrates. FASEB J. 25, 2937-46. Benkovic, S. J., Hammes, G. G. and Hammes-Schiffer, S. (2008). Free-energy landscape of enzyme catalysis. Biochemistry 47, 3317-21. Berezovska, O., Lleo, A., Herl, L. D., Frosch, M. P., Stern, E. A., Bacskai, B. J. and Hyman, B. T. (2005). Familial Alzheimer's disease presenilin 1 mutations cause alterations in the conformation of presenilin and interactions with amyloid precursor protein. J. Neurosci. 25, 3009-17. Boehr, D. D., Nussinov, R. and Wright, P. E. (2009). The role of dynamic conformational ensembles in biomolecular recognition. Nat. Chem. Biol. 5, 789-96. Capell, A., Grunberg, J., Pesold, B., Diehlmann, A., Citron, M., Nixon, R., Beyreuther, K., Selkoe, D. J. and Haass, C. (1998). The proteolytic fragments of the Alzheimer's disease-associated presenilin-1 form heterodimers and occur as a 100-150-kDa molecular mass complex. J. Biol. Chem. 273, 3205-11. Chae, P. S., Rasmussen, S. G., Rana, R. R., Gotfryd, K., Chandra, R., Goren, M. A., Kruse, A. C., Nurva, S., Loland, C. J., Pierre, Y. et al. (2010). Maltose-neopentyl glycol (MNG) amphiphiles for solubilization, stabilization and crystallization of membrane proteins. Nat. Methods 7, 1003-8. Chávez-Gutiérrez, L., Bammens, L., Benilova, I., Vandersteen, A., Benurwar, M., Borgers, M., Lismont, S., Zhou, L., Van Cleynenbreugel, S., Esselmann, H. et al. (2012). The mechanism of gamma-Secretase dysfunction in familial Alzheimer disease. EMBO J. 31, 2261-74. Chávez-Gutiérrez, L., Tolia, A., Maes, E., Li, T., Wong, P. C. and de Strooper, B. (2008). Glu(332) in the Nicastrin ectodomain is essential for gamma-secretase complex maturation but not for its activity. J. Biol. Chem. 283, 20096-105. Cianfrocco, M. A., Kassavetis, G. A., Grob, P., Fang, J., Juven-Gershon, T., Kadonaga, J. T. and Nogales, E. (2013). Human TFIID binds to core promoter DNA in a reorganized structural state. Cell 152, 120-31. Das, C., Berezovska, O., Diehl, T. S., Genet, C., Buldyrev, I., Tsai, J. Y., Hyman, B. T. and Wolfe, M. S. (2003). Designed helical peptides inhibit an intramembrane protease. J. Am. Chem. Soc. 125, 11794-5. De Strooper, B. (2003). Aph-1, Pen-2, and Nicastrin with Presenilin generate an active gamma-Secretase complex. Neuron 38, 9-12. De Strooper, B. and Annaert, W. (2010). Novel research horizons for presenilins and gamma-secretases in cell biology and disease. Annu. Rev. Cell Dev. Biol. 26, 235-60. Elad, N., Clare, D. K., Saibil, H. R. and Orlova, E. V. (2008). Detection and separation of heterogeneity in molecular complexes by statistical analysis of their two-dimensional projections. J. Struct. Biol. 162, 108-20. Elad, N., Farr, G. W., Clare, D. K., Orlova, E. V., Horwich, A. L. and Saibil, H. R. (2007). Topologies of a substrate protein bound to the chaperonin GroEL. Mol. Cell 26, 415-26. Esler, W. P., Kimberly, W. T., Ostaszewski, B. L., Diehl, T. S., Moore, C. L., Tsai, J. Y., Rahmati, T., Xia, W., Selkoe, D. J. and Wolfe, M. S. (2000). Transition-state analogue inhibitors of gamma-secretase bind directly to presenilin-1. Nat. Cell Biol. 2, 428-34.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 22: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

22

Esselens, C., Oorschot, V., Baert, V., Raemaekers, T., Spittaels, K., Serneels, L., Zheng, H., Saftig, P., De Strooper, B., Klumperman, J. et al. (2004). Presenilin 1 mediates the turnover of telencephalin in hippocampal neurons via an autophagic degradative pathway. J. Cell Biol. 166, 1041-54. Fischer, N., Konevega, A. L., Wintermeyer, W., Rodnina, M. V. and Stark, H. (2010). Ribosome dynamics and tRNA movement by time-resolved electron cryomicroscopy. Nature 466, 329-33. Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M. and Leith, A. (1996). SPIDER and WEB: processing and visualization of images in 3D electron microscopy and related fields. J. Struct. Biol. 116, 190-9. Guix, F. X., Wahle, T., Vennekens, K., Snellinx, A., Chavez-Gutierrez, L., Ill-Raga, G., Ramos-Fernandez, E., Guardia-Laguarta, C., Lleo, A., Arimon, M. et al. (2012). Modification of gamma-secretase by nitrosative stress links neuronal ageing to sporadic Alzheimer's disease. EMBO Mol. Med. 4, 660-73. Lazarov, V. K., Fraering, P. C., Ye, W., Wolfe, M. S., Selkoe, D. J. and Li, H. (2006). Electron microscopic structure of purified, active gamma-secretase reveals an aqueous intramembrane chamber and two pores. Proc. Natl. Acad. Sci. U. S. A. 103, 6889-94. Li, X., Dang, S., Yan, C., Gong, X., Wang, J. and Shi, Y. (2013). Structure of a presenilin family intramembrane aspartate protease. Nature 493, 56-61. Li, Y., Lu, S. H., Tsai, C. J., Bohm, C., Qamar, S., Dodd, R. B., Meadows, W., Jeon, A., McLeod, A., Chen, F. et al. (2014). Structural interactions between inhibitor and substrate docking sites give insight into mechanisms of human PS1 complexes. Structure 22, 125-35. Li, Y. M., Xu, M., Lai, M. T., Huang, Q., Castro, J. L., DiMuzio-Mower, J., Harrison, T., Lellis, C., Nadin, A., Neduvelil, J. G. et al. (2000). Photoactivated gamma-secretase inhibitors directed to the active site covalently label presenilin 1. Nature 405, 689-94. Lleo, A., Berezovska, O., Herl, L., Raju, S., Deng, A., Bacskai, B. J., Frosch, M. P., Irizarry, M. and Hyman, B. T. (2004). Nonsteroidal anti-inflammatory drugs lower Abeta42 and change presenilin 1 conformation. Nat. Med. 10, 1065-6. Lu, P., Bai, X. C., Ma, D., Xie, T., Yan, C., Sun, L., Yang, G., Zhao, Y., Zhou, R., Scheres, S. H. et al. (2014). Three-dimensional structure of human gamma-secretase. Nature 512, 166-70. Mastronarde, D. N. (2005). Automated electron microscope tomography using robust prediction of specimen movements. J. Struct. Biol. 152, 36-51. Mindell, J. A. and Grigorieff, N. (2003). Accurate determination of local defocus and specimen tilt in electron microscopy. J. Struct. Biol. 142, 334-47. Motlagh, H. N., Wrabl, J. O., Li, J. and Hilser, V. J. (2014). The ensemble nature of allostery. Nature 508, 331-9. Nygaard, R., Zou, Y., Dror, R. O., Mildorf, T. J., Arlow, D. H., Manglik, A., Pan, A. C., Liu, C. W., Fung, J. J., Bokoch, M. P. et al. (2013). The dynamic process of beta(2)-adrenergic receptor activation. Cell 152, 532-42. Ogura, T., Mio, K., Hayashi, I., Miyashita, H., Fukuda, R., Kopan, R., Kodama, T., Hamakubo, T., Iwatsubo, T., Tomita, T. et al. (2006). Three-dimensional structure of the gamma-secretase complex. Biochem. Biophys. Res. Commun. 343, 525-34. Osenkowski, P., Li, H., Ye, W., Li, D., Aeschbach, L., Fraering, P. C., Wolfe, M. S., Selkoe, D. J. and Li, H. (2009). Cryoelectron microscopy structure of purified gamma-secretase at 12 A resolution. J. Mol. Biol. 385, 642-52. Pardon, E., Laeremans, T., Triest, S., Rasmussen, S. G., Wohlkonig, A., Ruf, A., Muyldermans, S., Hol, W. G., Kobilka, B. K. and Steyaert, J. (2014). A general protocol for the generation of Nanobodies for structural biology. Nat. Protoc. 9, 674-93. Perez-Tur, J., Froelich, S., Prihar, G., Crook, R., Baker, M., Duff, K., Wragg, M., Busfield, F., Lendon, C., Clark, R. F. et al. (1995). A mutation in Alzheimer's disease destroying a splice acceptor site in the presenilin-1 gene. Neuroreport 7, 297-301.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 23: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

23

Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt, D. M., Meng, E. C. and Ferrin, T. E. (2004). UCSF Chimera--a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605-12. Radermacher, M., Wagenknecht, T., Verschoor, A. and Frank, J. (1987). Three-dimensional reconstruction from a single-exposure, random conical tilt series applied to the 50S ribosomal subunit of Escherichia coli. J. Microsc. 146, 113-36. Rath, B. K. and Frank, J. (2004). Fast automatic particle picking from cryo-electron micrographs using a locally normalized cross-correlation function: a case study. J. Struct. Biol. 145, 84-90. Renzi, F., Zhang, X., Rice, W. J., Torres-Arancivia, C., Gomez-Llorente, Y., Diaz, R., Ahn, K., Yu, C., Li, Y. M., Sisodia, S. S. et al. (2011). Structure of gamma-secretase and its trimeric pre-activation intermediate by single-particle electron microscopy. J. Biol. Chem. 286, 21440-9. Rosenthal, P. B. and Henderson, R. (2003). Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy. J. Mol. Biol. 333, 721-45. Sato, C., Morohashi, Y., Tomita, T. and Iwatsubo, T. (2006). Structure of the catalytic pore of gamma-secretase probed by the accessibility of substituted cysteines. J. Neurosci. 26, 12081-8. Scheres, S. H. (2012). RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519-30. Scheres, S. H. and Chen, S. (2012). Prevention of overfitting in cryo-EM structure determination. Nat. Methods 9, 853-4. Selkoe, D. J. and Wolfe, M. S. (2007). Presenilin: running with scissors in the membrane. Cell 131, 215-21. Serneels, L., Van Biervliet, J., Craessaerts, K., Dejaegere, T., Horre, K., Van Houtvin, T., Esselmann, H., Paul, S., Schafer, M. K., Berezovska, O. et al. (2009). gamma-Secretase heterogeneity in the Aph1 subunit: relevance for Alzheimer's disease. Science 324, 639-42. Shah, S., Lee, S. F., Tabuchi, K., Hao, Y. H., Yu, C., LaPlant, Q., Ball, H., Dann, C. E., 3rd, Sudhof, T. and Yu, G. (2005). Nicastrin functions as a gamma-secretase-substrate receptor. Cell 122, 435-47. Shelton, C. C., Zhu, L., Chau, D., Yang, L., Wang, R., Djaballah, H., Zheng, H. and Li, Y. M. (2009). Modulation of gamma-secretase specificity using small molecule allosteric inhibitors. Proc. Natl. Acad. Sci. U. S. A. 106, 20228-33. Takagi, J., Petre, B. M., Walz, T. and Springer, T. A. (2002). Global conformational rearrangements in integrin extracellular domains in outside-in and inside-out signaling. Cell 110, 599-11. Tang, G., Peng, L., Baldwin, P. R., Mann, D. S., Jiang, W., Rees, I. and Ludtke, S. J. (2007). EMAN2: an extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38-46. Thinakaran, G., Borchelt, D. R., Lee, M. K., Slunt, H. H., Spitzer, L., Kim, G., Ratovitsky, T., Davenport, F., Nordstedt, C., Seeger, M. et al. (1996). Endoproteolysis of presenilin 1 and accumulation of processed derivatives in vivo. Neuron 17, 181-90. Tolia, A., Chavez-Gutierrez, L. and De Strooper, B. (2006). Contribution of presenilin transmembrane domains 6 and 7 to a water-containing cavity in the gamma-secretase complex. J. Biol. Chem. 281, 27633-42. Uemura, K., Farner, K. C., Hashimoto, T., Nasser-Ghodsi, N., Wolfe, M. S., Koo, E. H., Hyman, B. T. and Berezovska, O. (2010). Substrate docking to gamma-secretase allows access of gamma-secretase modulators to an allosteric site. Nat. Commun. 1, 130. van Heel, M., Harauz, G., Orlova, E. V., Schmidt, R. and Schatz, M. (1996). A new generation of the IMAGIC image processing system. J. Struct. Biol. 116, 17-24. Wahlster, L., Arimon, M., Nasser-Ghodsi, N., Post, K. L., Serrano-Pozo, A., Uemura, K. and Berezovska, O. (2013). Presenilin-1 adopts pathogenic conformation in normal aging and in sporadic Alzheimer's disease. Acta Neuropathol. 125, 187-99.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 24: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

24

Wolfe, M. S., Xia, W., Ostaszewski, B. L., Diehl, T. S., Kimberly, W. T. and Selkoe, D. J. (1999). Two transmembrane aspartates in presenilin-1 required for presenilin endoproteolysis and gamma-secretase activity. Nature 398, 513-7. Xie, T., Yan, C., Zhou, R., Zhao, Y., Sun, L., Yang, G., Lu, P., Ma, D. and Shi, Y. (2014). Crystal structure of the gamma-secretase component nicastrin. Proc. Natl. Acad. Sci. U. S. A. 111, 13349-54. Zhang, X., Hoey, R. J., Lin, G., Koide, A., Leung, B., Ahn, K., Dolios, G., Paduch, M., Ikeuchi, T., Wang, R. et al. (2012). Identification of a tetratricopeptide repeat-like domain in the nicastrin subunit of gamma-secretase using synthetic antibodies. Proc. Natl. Acad. Sci. U. S. A. 109, 8534-9.

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 25: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

Figure 1

A

E

F

B C

D

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 26: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

C

B

D

A

Figure 2 Jo

urna

l of C

ell S

cien

ceA

ccep

ted

man

uscr

ipt

Page 27: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8

0

2

4

6

8

Length of major axis (nm)

H

# s

tru

ctu

res

Figure 3

D B

10

nm

2D class averages

3D class averages

C

Individual RCT structures

G E F

A

Extended 17%

Intermediate 30%

Compact 17%

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 28: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

A

Figure 4

B E

10

nm

# st

ruct

ure

s

Length of major axis (nm)

# st

ruct

ure

s

Length of major axis (nm)

γ-secretase + pep15 γ-secretase + inhX

D

F C

3D class average

RCT structure

3D class average

RCT structure

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 29: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

D M S O + In h X0

2

4

6

8

1 0

%E

FR

ET

***

Figure 5

A B

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 30: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

NCT

PS1 CTF

PS1 NTF

PEN-2

DM

SO

Inh

X

Pep

15

Input

IP: anti-NCT

Bound

DM

SO

Inh

X

Pep

15

Figure 6 Jo

urna

l of C

ell S

cien

ceA

ccep

ted

man

uscr

ipt

Page 31: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

Figure 7

# st

ruct

ure

s

Length of major axis (nm)

Δ9

2D class- averages

Eigenimages

C

Eigenimage 2

Positive Negative

D

Den

sity

Positive Negative

Eigenimage 3

B Local variability

Eigen- image

2D class- average

NCT ECD Core

A 1

0 im

g/cl

ass

E

wt inhX

Jour

nal o

f Cel

l Sci

ence

Acc

epte

d m

anus

crip

t

Page 32: Journal of Cell Science Accepted manuscriptjcs.biologists.org/content/joces/early/2014/12/07/jcs.164384.full.pdf · nm and 12-13 nm long, respectively. Importantly, we provide evidence

Ener

gy

Conformations En

ergy

Ener

gy

Conformations

Ener

gy

pep15

Figure 8 Jo

urna

l of C

ell S

cien

ceA

ccep

ted

man

uscr

ipt