VU Research Portal

The endolysosomal system in neuronal physiology and pathology Vazquez Sanchez, S.

2019

document version Publisher's PDF, also known as Version of record

Link to publication in VU Research Portal

citation for published version (APA) Vazquez Sanchez, S. (2019). The endolysosomal system in neuronal physiology and pathology.

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

E-mail address: [email protected]

Download date: 05. Oct. 2021 The endolysosomal system in neuronal physiology and pathology

Sonia Vazquez Sanchez Printed by: Ipskamp

Layout by: Sonia Vazquez Sanchez

Cover: Confocal microscopy image of a human neuron overexpressing TauP301L-GFP (green) and treated with tau fibrils. Compartments of the endolysosomal system are immunolabelled in red (CD63) and in yellow (LAMP1). The nucleus (DAPI) and the dendrites (MAP2) are labelled in blue. The bookmark shows a zoom of this neuron acquired with STED microscopy.

The publication of this thesis was financially supported by CNCR and Alzheimer Nederland.

© 2019 by Sonia Vazquez Sanchez

ISBN: 978-94-028-1522-1 VRIJE UNIVERSITEIT

THE ENDOLYSOSOMAL SYSTEM IN NEURONAL PHYSIOLOGY AND PATHOLOGY

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam, op gezag van de rector magnificus prof.dr. V. Subramaniam, in het openbaar te verdedigen ten overstaan van de promotiecommissie van de Faculteit der der Bètawetenschappen op maandag 17 juni 2019 om 15.45 uur in de aula van de universiteit, De Boelelaan 1105

door

Sonia Vazquez Sanchez

geboren te Madrid, Spanje promotor: prof.dr. M.Verhage copromotor: dr. J.R.T. van Weering CONTENTS

Chapter 1 7 General introduction

Chapter 2 29 VPS35 depletion does not impair presynaptic structure and function

Chapter 3 57 Sorting nexin 4 is an endosomal sorting protein located to synapses

Chapter 4 105 The seeding of tau pathology alters the endolysosomal system

Chapter 5 135 Summary, general discussion, and future directions

Chapter 1

General introduction

7 Chapter 1

THE ENDOLYSOSOMAL SYSTEM

The endolysosomal system consists of a complex network of interconnected membrane compartments with constant flux of material. The appearance of an endomembrane trafficking system was a key event in evolution for the prokaryote-to-eukaryote transition (see review (Dacks and Field, 2007)). The evolution of this trafficking system has been critical for the emergence and diversification of complex cell types such as neurons, and organisms such as humans. For example, it is required for neuronal development and function, mediating cell fate decisions, cell migration, axon outgrowth and polarity (see review (Yap and Winckler, 2012)). The endolysosomal system maintains cell homeostasis by cargo sorting, degradation and recycling. Different endolysosomal compartments are distinguished based on their function, molecular composition and structure (Figure 1 and Table 1). This distinction between compartments is very useful to study the endolysosomal system; however, it is important to note that they are not fixed and separate entities. Instead, they are extremely dynamic with constant exchange of material and with highly overlapping features.

Both the endolysosomal molecular machinery and ultrastructure are used to define endolysosomal compartments (Table 1). Rab proteins are widely used as organelle markers, including different endolysosomal compartments. Rab proteins are a family of small monomeric GTPases part of the RAS superfamily which regulate membrane trafficking by recruiting effector proteins in a GTP-bound conformation (see review (Stenmark, 2009)). On the early endosome, Rab4 and Rab5 mediate early endosomal fusion and biogenesis. On recycling endosomes, Rab11 is involved in trafficking cargo to the plasma membrane. On late endosomal compartments, Rab7 is involved in endosomal maturation, lysosome biogenesis and trafficking cargo away from the late endosome (see review (Galvez et al., 2012)). Phosphoinositides are phosphorylated forms of phosphatidylinositol (PI) that are also differently distributed on the endolysosomal membranes and regulate membrane trafficking. PI(3)P is enriched on early endosome membranes, while PI(4)P is on recycling endosomes and PI(3,5)P2 on late endosomes. In the lysosomal membrane, several phosphoinositides coexist including PI(3)P, PI(4)

P, and PI(4,5)P2 (see review (Wallroth and Haucke, 2018)). Other transmembrane or membrane associated proteins are frequently used to label the different endolysosomal compartments. For example, early endosome antigen 1 (EEA1) is used to label early endosomes because it localizes to early endosomal membranes to mediate endosome docking and fusion (Christoforidis et al., 1999a). Tetraspanin CD63 is used to label late endosomes/multivesicular bodies (MVBs) because it is highly enriched on intraluminal vesicles (ILVs) in MVB (see review (Pols and Klumperman, 2009)). For lysosomes, there are mainly three transmembrane proteins used as molecular markers due to their

8 General introduction

Endoplasmic Nucleus reticulum 1 Golgi apparatus Lysosome

trans-Golgi network

MVB

Early Endolysosome endosome Endocytic Recycling vesicle endosome

Plasma membrane Exosomes Extracellular space

Figure 1: Simplified schema of the main endolysosomal compartments and pathways: biosynthesis (purple arrows), recycling (green arrows) and degradation (blue arrows). Cargo (black dot) is defined as transmembrane proteins, their lipids and associated proteins. Inthe biosynthetic pathway, membrane proteins are synthesized on the endoplasmic reticulum, transported and modified at the Golgi apparatus, and sorted in the trans-Golgi network (purple arrows). In the retrograde pathway, endocytic vesicles from the plasma membrane fuse with early endosomes, where the fate of the cargo is determined: the endocytic cargo will be degraded or recycled. If the cargo is recycled, it will be sorted in membrane tubules which emanate from the endosome to other cell compartments such as the trans-Golgi network, the recycling endosome or directly to the plasma membrane. If the cargo is degraded, it will be included in the intraluminal vesicles inside the early endosome which accumulate during the process of maturation to late endosome. The late endosome or multivesicular body (MVB) fuses with the lysosome forming a hybrid organelle called endolysosome, in which degradation takes place. After that, the endolysosome can mature to a lysosome. Alternatively, the multivesicular body can fuse with the plasma membrane to release its content (black arrow). When the intraluminal vesicles are release to the extracellular space, they are called exosomes.

9 Chapter 1 predominantly lysosomal membrane localization: the lysosome-associated membrane protein 1 (LAMP1), involved in lysosomal stability, integrity and exocytosis (see review (Saftig and Klumperman, 2009)), the lysosome-associated membrane protein 2 (LAMP2), involved in chaperone-mediated autophagy (see review (Saftig and Klumperman, 2009)) and the lysosomal integral membrane protein 2 (LIMP2) which is a receptor for lysosomal transport of the acid hydrolase β-glucocerebrosidase (GC) (see review (Gonzalez et al., 2014)). Other molecules used to define endolysosomes are based on their function such as DQTM Red-BSA which produces a bright fluorescent product when it is hydrolyzed (Bright et al., 2016). An alternative approach to discern between distinct endosolysosomal compartments is the ultrastructural analysis by transmission electron microscopy (TEM), which provides the high resolution required to resolve the complex membrane structure of the endolysosomal system (see Table1 and review (Klumperman and Raposo, 2014)).

1 General sorting mechanism and machineries

The biosynthetic and retrograde pathway converge in the early endosome. The early endosomes constitute a sorting station in which cargo is sorted for degradation or retrieved to be targeted to a different location. To assure specificity in cargo sorting, the endosolysosomal system faces two main challenges: recognizing the fate of each cargo and separating it from neighboring cargo that needs to be trafficked by a distinct pathway. This specificity is achieved through the coordinated action of proteins, which can be viewed as molecular machineries constituted by protein sub-complexes. Here, we will review the general endosomal sorting principles using the retromer-mediated sorting machinery as an example. Other protein sub-complexes follow similar principles including retriever-mediated sorting or ESCRT sorting (see Figure 4) (see review (McNally and Cullen, 2018)). Briefly, in the early endosomal membrane there is both cargo which needs to be recycled and which needs to be degraded. Cargos contain a sorting signal which is recognized by cargo-recognition sub-complex. Selected cargo is then concentrated in membrane sub-domains that are incorporated into transport carriers through membrane deformation, stabilization and scission. The cargo enriched transport carriers traffic cargo to its destination.

1.1 Cargo recognition sub-complex: VPS26-VPS35-VPS29 trimer

Retromer is a protein complex that mediates cargo sorting from the endosome to the trans-Golgi network (TGN) (Seaman et al., 1998) and to the plasma membrane (Temkin et al., 2011). Retromer was first described in yeast (Seaman et al., 1998) and proved to be highly conserved across all eukaryotes (Koumandou et al., 2011). Retromer is critical for cell function, as a total lack of retromer subunits leads to embryonic lethality both in

10 General introduction EEA1 CD63

Other proteins 1 LAMP1 and 2,LIMP2 and LAMP1 2 2 PI(3)P PI(4)P PI(3,5)P Molecular markers Molecular Phosphoinositides PI(3)P, PI(4)P, PI(4,5)P PI(4)P, PI(3)P, Rabs Rab 11 In the right, endolysosomal molecular machinery differentially tubular Morphology spherical, electron dense globular,tubules, > 5-8 ILV, patches of clathrin coat Rab 7 Transmission Electron Microscopy pleomorphic,tubulates, few ILV, patches of clathrin coat Rab 4, Rab 5 m m Diameter 0.25 - 1 μ - 0.25 60 - 100 nm 60 - 0.20 - >1 μ >1 - 0.20 100 - 500 nm 100 - Lysosome Early endosome distributed among distinct endolysosomal compartments. In the left, morphological description of endolysosomal compartments based of Transmission Electron Microscopy (TEM) data. The diameter of the tubular structures fluctuate between 20–50 nm, and (Klumperman and Raposo, 2014)). review the diameter of between ILV 40-60 nm (see Table 1: Table Key features to discern and define distinct endolysosomal compartments. Recycling endosome Late endosome/MVB

11 Chapter 1 mammalian (Lee et al., 1992; Muhammad et al., 2008) and fly models (Zhou et al., 2011). In yeast, retromer is a pentameric protein complex formed by two essential modules: the cargo-selection subcomplex constituted by vacuolar protein sorting-associated protein 26, 35 and 29 (Vps26, Vps35 and Vps29), and the membrane deformation subcomplex constituted by Vps5 and Vps17 (Burd and Cullen, 2014; Seaman et al., 1998). In mammals, the VPS26-VPS35-VPS29 trimer (Vps26-Vps35-Vps29 orthologs) does not form a stable complex with the membrane deformation subcomplex (Norwood et al., 2011), constituted by Sorting nexin-Bin/Amphiphysin/Rvs (BAR) (SNX-BAR) proteins (Vps5 and Vps17 orthologs). Recently, the subunits of retromer in mammals have been redefined based on the cargo recognition function. The SNX-BAR complex can also mediate cargo recognition independently of retromer (VPS26-VPS35-VPS29), forming a distinct cargo recognition subcomplex (Kvainickas et al., 2017; Simonetti et al., 2017). Apart from yeast, retromer now means the cargo-selection subcomplex constituted by VPS35, VPS29 and VPS26A or VPS26B. The membrane deformation subcomplex of retromer is referred as retromer-associated SNX-BAR complex, formed by different pattern of dimerization of SNX-BAR proteins (SNX1, SNX2, SNX5, SNX6 and SNX32 in mammalians). Retromer can also bind to the early endosomal membrane by its association with SNX3 and SNX27 (which bind PI(3)P) or to late endosomal membranes through its association with Rab7 (see review (Cullen and Steinberg, 2018)).

1.2 Sorting signal: The sorting motif

Retromer dependent cargo contains a sorting motif which is an unstructured linear peptide sequence in the cytoplasmic tail of the cargo. Retromer recognizes and binds its cargo in a sequence-dependent manner both directly and through adaptor proteins. For example, VPS26 can directly bind the FANSHY sorting motif of the endocytic sortilin- related receptor SorLA, which binds to amyloid precursor protein (APP) (Fjorback et al., 2012). Retromer also associates indirectly with cargo through cargo adaptor proteins such as SNX3 and SNX27. SNX3 directly binds Wnt-binding protein Wntless which mediates Wnt signaling (Harterink et al., 2011). Through its PDZ domain, SNX27 binds to carboxy-terminal type I PDZ domain-binding sorting motifs such as the one present in the β2-adrenergic receptor (see review (Cullen and Steinberg, 2018)).

1.3 Cargo enriched sub-domain scaffolds: The WASH complex

The scaffold of retromer dependent cargo enriched subdomains is driven by branched actin polymerization. The Wiskott–Aldrich syndrome protein and SCAR homologue (WASH) complex stimulates the actin-related protein 2/3 (Arp2/3) complex to drive branched actin polymerization on the endosomal membrane. WASH complex is pentameric

12 General introduction protein complex constituted by WASHC1 (also known as WASH1), WASHC2A/B/C (also known as FAM21A/B/C), WASHC3 (also known as CCDC53), WASHC4 (also known as KIAA1033 or WASH interacting protein (SWIP)), and WASHC5 (also known as KIAA0196 or strumpellin). Retromer directly binds to FAM21 to recruit the WASH complex that 1 regulates the formation and maintenance of cargo enriched sub-domains (see review (McNally and Cullen, 2018)).

1.4 Membrane remodeling sub-complex: The retromer-associated SNX-BAR complex

SNX-BAR heterodimers are involved in endosomal tubule formation and stabilization in retromer-dependent sorting in mammals. Sorting nexins are classified by the presence of a particular type of phox-homology (PX) domain (phosphoinositide-binding phox homology (PX) domain), which binds predominantly phosphatidylinositol 3-phosphate (PI(3)P) in early endosomal membranes (Carlton et al., 2005). SNX-BAR subfamily is defined by the presence of a carboxy-terminal Bin/Amphiphysin/Rvs (BAR) domain, which binds to curved membranes upon dimerization (Carlton et al., 2004; Cullen, 2008). SNX-BAR dimers stabilize membrane tubules that act as cargo enriched transport carriers to traffic cargo to its target organelle. In mammals, different dimers constituted by SNX1, SNX2, SNX5, SNX6 and SNX32 form the membrane deformation subcomplex in retromer- dependent sorting (van Weering and Cullen, 2014).

1.5 Retromer-SNX-BAR assembly in membrane tubules

The crystal structure of different proteins of the sorting machinery have been keyto elucidate how they operate. In particular, the crystal structure of VPS29-VPS35 and SNX1 have been essential to understand the retromer mediated sorting (Hierro et al., 2007; van Weering et al., 2012a). In the SNX-BAR family the PX and BAR domains are targeted and tubulate phosphoinositide‐containing membranes. The BAR domain is a banana-shaped protein-dimerization domain that contains positively charged residues in its concave side which associates with membranes through electrostatic interactions and through insertion of amphipathic helix in the membrane (van Weering et al., 2012a). The PX and BAR domain together function as coincidence detectors of PI(3)P positively curved membranes (Carlton et al., 2004). In addition, the BAR domains oligomerizes to form helical arrays that induce and stabilize curved membrane structures such as tubules and vesicles (Figure 2) (van Weering and Cullen, 2014).

The assembly of the fungal Chaetomium thermophilum retromer (Vps26, Vps29, Vps35 and Vps5) was resolved in membrane tubules using cryo-electron tomography (Figure 2).

13 Chapter 1

In vitro, membrane tubules of 31 nm were formed by incubating Vps26, Vps29, Vps35 and Vps5 with liposomes. These tubules were decorated with a protein coat of 15 nm. Vps5 homodimers formed an oligomeric pseudo-helical array stabilized by BAR-BAR (tip-tip) interactions of consecutive Vps5 homodimers. Vps26 dimers also stabilize this

a convex side BAR1 side view BAR2 b - + BAR1 BAR2 +++ + + + concave side hydrophobic AH AH AH dimerization + - + interface + + + basic residues + - acidic residues AH

c’ c’’

+++ + + + +++ + + +

------

e d

f

Vps5 Vps35

Vps26

Vps29 Tubular lumen

Figure 2: Retromer/SNX-BAR assembly in membrane tubules. (a-e) Assembly model of the SNX- BAR sub-complex in membrane tubules. (a) SNX-BAR dimers consist in two SNX-BARs (BAR1 and BAR2) with rigid banana shape with basic residues and two amphipathic helix (AH) on the concave side. (b) SNX-BARs form dimers through hydrophobic and electrostatic interactions between residues in its hydrophobic dimerization interface of the BAR domain. (c’) The amphipathic helices insert in the membrane and the positively charge concave part interacts with the negatively charge membrane through electrostatic interactions which both senses and (c’’) induces membrane curvature. (d-e) SNX-BAR dimers form helical arrays that extend and stabilize membrane tubules through tip-tip interaction between consecutive dimers. (f) Pseudo-helical array of fungal Vps5 dimers is stabilized by Vps26 dimers that contact four Vps5 dimers. Each Vps26 interacts with the N-terminal of Vps35, that dimerizes through Ct-Ct interactions. Vps35 dimers form an arch-like structure in which Vps29 associates the opposite side of VPS35 dimerization. Adapted from (Kovtun et al., 2018; van Weering and Cullen, 2014).

14 General introduction array from the top in where each Vps26 dimer was in contact with four Vps5 dimers and the N-terminal of a Vps35 dimer. Vps35 formed dimers through Ct interaction with other Vps35 leading to an arch like structure in which in the opposite side of dimerization Vps29 was bound (Kovtun et al., 2018). Although resolving this assembly was a break through, 1 it is not clear how much it can be generalized to other non-fungus retromer complex assemblies (Simonetti and Cullen, 2018).

1.6 The SNX-BAR family in endosomal sorting

a

b

c

d

Figure 3: Dimerization pattern of the mammalian SNX-BAR subfamily. On the left side are the SNX-BAR members that can form homodimers and on the right side the ones that do not form homodimers. (a) The retromer-associated SNX-BARs form predominantly heterodimers, where SNX1 or SNX2 dimerize with SNX5 or SNX6 or SNX32. These SNX-BARs are involved in endosomal recycling to the trans-Golgi network and to the plasma membrane. (b) SNX4 is a hub of dimerization forming homodimers and heterodimers with SNX7 and SNX30 and mediate recycling from the early endosome to the plasma membrane through the recycling endosome. (c) SNX9, SNX18, and SNX33 form predominantly homodimers and are involved in endocytosis. (d) SNX8 forms homodimers that might regulate retromer-independent endosome-to-TGN traffic. Adapted from (van Weering and Cullen, 2014).

15 Chapter 1

Several members of the SNX-BAR family are involved in membrane tubule formation and stabilization. These SNX-BARs are able to remodel membranes when forming homo or heterodimers, whose pattern of dimerization overlaps with their tubulation ability (Figure 3) (van Weering et al., 2012a). These molecular distinct SNX-BAR tubules seem to mediate specific trafficking pathways (Figure 3). For example, SNX4 forms homodimers or heterodimers with SNX7 or SNX30 to mediate endosome-to-plasma membrane recycling. SNX4-decorated tubules emanate from the endosomes during the Rab5-Rab7 transition (early endosome to late endosome) and during Rab4-Rab11 transition (early recycling endosome to endosome recycling compartment) (van Weering et al., 2012b). In yeast, Snx4p (SNX4 ortholog) mediates the recycling to the plasma membrane of Scn1p (an exocytic v-SNARE) (Hettema et al., 2003). Similarly in HeLa cells, SNX4 recycles the transferrin receptor (TfnR) back to the plasma membrane through the recycling endosome (Traer et al., 2007). Hence, SNX4 is a hub for dimerization which mediates recycling back to the plasma membrane. SNX4 might constitute or be part of an undiscovered sorting endosomal machinery, which might follow similar principles as retromer-mediated sorting.

SNX9, SNX18 and SNX33 are SNX-BAR proteins containing an N-terminal Src homology 3 domain (SH3) and are involved in endocytosis. The SH3 domain mediates protein- protein interactions with for example dynamin-1 and -2, synaptojanin, and Neural Wiskott–Aldrich Syndrome Protein (N-WASP) (Shin et al., 2007), which are all part of the endocytic machinary. In neurons, both SNX9 overexpression and depletion impaired synaptic vesicle endocytosis (Shin et al., 2007). SNX9 is recruited to the neck of clathrin coated pits by phosphatidylinositol 3,4-bisphosphate (PI(3,4)P2) (Schöneberg et al., 2017b). There, SNX9 oligomerizes and narrows the neck, where dynamin associates to mediate the endocytic fission (Schöneberg et al., 2017b). SNX18 has been shown to function in the same processes as SNX9, such as transferrin endocytosis (Park et al., 2010). Hence, SNX18 and potentially SNX33 might also be involved in endocytosis in a similar way. SNX8 forms homodimers and it localizes in early endosomes (van Weering et al., 2012b). SNX8 depletion increased Shiga toxin transport and inhibit ricin transport to the trans-Golgi network where they localize (Dyve et al., 2009). Despite this discrepancy, SNX8 might mediate a recycling pathway to trans-Golgi network. Hence, different SNX- BAR proteins might constitute or be part of endosomal sorting machineries which govern different endosomal sorting pathways.

1.7 The sorting machinery for degradation

The endosomal sorting complexes required for transport (ESCRT) recognizes and sorts cargo to intraluminal vesicles (ILVs) for degradation in the endolysosome (see review (Cullen and Steinberg, 2018)).

16 General introduction

Sorting signal: Ubiquitylation is the sorting signal for degradation. In general, monoubiquitylation on lysine residues within the intracellular cytosolic domain of the cargo proteins (Clague et al., 2012). 1 Cargo recognition sub-complex: ESCRT-0 associates with early endosome membranes via the binding of HRS to PI(3)P. ESCRT-0 recognizes ubiquitinated cargo and forms clusters. Components of ESCRT-I (TSG101 and UBAP1) and ESCRT-II (VPS36) also bind ubiquitinated cargo, thereby ESCRT-0, ESCRT-I and ESCRT-II mediate the recognition and concentration of the cargo into membrane domain for degradation (Figure BOX2) (Babst et al., 2002b; Katzmann et al., 2001; Schöneberg et al., 2017a).

Cargo enriched sub-domain scaffolds: Flat clathrin coats seem to scaffold the ESCRT sub-domain for degradation (Raiborg et al., 2006).

Membrane remodeling sub-complex: ESCRT-III mediates membrane remodeling. ESCRT-III does not have a ubiquitin binding domain but it is recruited to the ESCRT-0, -I and -II complex to induce the inward membrane budding and ILV formation. The ESCRT- III complex polymerizes in organized filaments in a flat spiral on the membrane of the degradation subdomain. This spiral laterally arrests, compresses and compacts the degradation domain (Figure 4) (Babst et al., 2002a; Im and Hurley, 2008; Kostelansky et al., 2006). This spiral buckles into a three-dimensional spring which mediates inward membrane budding to form ILVs (Chiaruttini et al., 2015). a b c

Recycling cargo ESCRT-0 Ubiquitynilated cargo ESCRT-I ESCRT-III filament ESCRT-II ILV

Figure 4. Sorting for degradation: the ESCRT complexes. (a) In the early endosomal membrane cargo for recycling and for degradation (ubiquitinylated) coexist. (b) ESCRT-0, ESCRT-I and ESCRT- II recognize and concentrate ubiquitinated cargo into membrane domains for degradation where ESCRT-III is recruited. (c) ESCRT-III polymerizes into a spiral filament which mediates inward budding to form a cargo for degradation enriched ILV. Adapted from (Cullen and Steinberg, 2018).

2 Endolysosomal dysfunction and neurodegeneration: Alzheimer’s disease

Dysregulation of protein trafficking and degradation is a major aspect of most neurodegenerative disorders, including Alzheimer’s disease (AD) (Small and Petsko, 2015). Although endolysosomal genes are ubiquitously expressed, mutations in these

17 Chapter 1 genes are notably associated with neurodegenerative diseases (Small and Petsko, 2015). Endolysosomal gene sets have been specifically associated with AD; however no single gene can explain the variation significantly, consistent with the polygenic etiology of late-onset AD (Gao et al., 2018). AD is characterized by cognitive decline and memory loss (Alzheimer’s Association 2018), and it is the most common cause of dementia accounting for 60-80% of all dementia cases (Barnes and Yaffe, 2011). According to a recent Alzheimer’s Association estimation, more than 13.8 million people will be suffering from AD in 2050, and in 2017, AD care have cost more than $232 billion in the U.S.A. alone (Alzheimer’s Association, 2018). Finding effective therapies is a priority from both social and economic perspective, thus understanding AD mechanisms and thereby understanding how endolysosomal dysfunction contributes to AD pathogenesis is critical for successful therapeutic intervention.

2.1 Neuropathology of AD: Aβ pathology, tau pathology and ‘endosomopathy’

AD is characterized at the neuropathological level by the aggregation of amyloid beta and tau. Amyloid beta (Aβ) deposits extracellularly in senile plaques, while tau aggregates intracellularly, forming neurofibrillary tangles. Pathological studies revealed that both amyloid beta and tau aggregates spread through the brain in a stereotypical pattern (Braak and Braak, 1991). However, only tau spreading correlates strongly with the cognitive decline observed in AD patients (Aschenbrenner et al., 2018; Braak and Braak, 1991). Tau pathology starts in the entorhinal cortex, then propagates to limbic areas and finally to the cortex (Braak and Braak, 1991). AD is also neuropathologically characterized by endolysosomal aberrations (Cataldo et al., 1997; Colacurcio et al., 2017). In AD brains, endosome swelling is one of the first cellular symptoms observed (Cataldo et al., 2000), and endolysosomal proteins are upregulated, including Rab5, Rab7, Cathepsin D and LAMP1 (Ginsberg et al., 2010a; Ginsberg et al., 2010b). AD is also characterized by the appearance of granulovacuolar degeneration (GVD) bodies (see review (Köhler, 2016)). The charged multivesicular body protein 2B (CHMP2B) localizes to the core of the GVD, and lysosome-associated membrane protein 1 (LAMP1) is surrounding the GVD core (Funk et al., 2011), suggesting an endolysosomal origin. Although endolysosomal aberrations are a consistent feature of the AD brain, their cause and relation to the disease remain largely unknown. While the link between aberrant endolysosomal trafficking and Aβ pathology is becoming clearer (see below), the link with tau pathology is still poorly understood.

18 General introduction

2.2 Endolysosomal trafficking dysfunction in AD: Retromer dysfunction

One of the best studied endolysosomal sorting complexes in the context of AD is retromer. Retromer-associated sorting gene mutations are associated with higher risk of 1 AD, including SORL1, Rab7, members of the sorting nexin family and several subunits of retromer (Rogaeva et al., 2007; Vardarajan et al., 2012). Decreased retromer levels are also associated with Alzheimer disease (Small et al., 2005). In fact, increasing retromer levels has been proposed as a therapeutic target for AD, alleviating both Aβ pathology and tau pathology (Mecozzi et al., 2014; Young et al., 2018). Here, we review some aspects of retromer trafficking in neurons and how its dysregulation might play a role in AD.

2.2.1 Retromer and APP processing

The amyloid hypothesis of AD etiology postulates that the accumulation of Aβ in the brain is the primary cause of AD and that the other disease features are consequences of Aβ pathology (Hardy and Selkoe, 2002). In the endolysosomal system, the amyloid precursor protein (APP) is cleaved by β-secretase 1 (BACE1), which produces the C-terminal fragment (βCTF). Then, γ -secretase cleaves βCTF producing Aβ peptides (Small and Gandy, 2006). In AD brains, VPS35 levels are decreased (Small et al., 2005) and mice lacking one allele of VPS26 or VPS35 show increased Aβ levels and neurodegeneration (Muhammad et al., 2008; Wen et al., 2011). In vitro, VPS35 depletion increases the time that APP is localized to endosomes containing BACE1 and increases Aβ production (Bhalla et al., 2012; Small et al., 2005). Retromer regulates the trafficking of APP out of the endosomal system via the sorting receptor SorLA, and thereby Aβ production (Bhalla et al., 2012; Muhammad et al., 2008). In VPS26 heterozygote knock out mice, VPS35 levels are also reduced which suggests that the interaction among the individual proteins is critical for the complex stability (Muhammad et al., 2008). Motivated by this observation, R55 was designed as a pharmacological chaperon for retromer. R55 is a small molecule that stabilizes the trimeric VPS26–VPS35–VPS29 retromer structure in vitro and increases retromer protein levels in neuronal cultures. Increasing retromer levels enhances its function and APP is sorted away from endosomes, thereby reducing Aβ- levels in cultured neurons (Mecozzi et al., 2014).

2.2.2 Retromer and lysosomal function

Lysosomal dysfunction is not only a hallmark of AD but of most neurodegenerative diseases in which there is an abnormal accumulation of protein aggregates (Colacurcio et al., 2017; Nixon et al., 2000). Proper lysosomal hydrolytic activity is required for all autophagic processes which are critical for neuronal proteostasis: microautophagy

19 Chapter 1

(endosome-mediated), chaperon mediated autophagy (CMA) and macroautophagy (autophagosome-mediated) (Galluzzi et al., 2017). One of the first established functions of retromer is the recycling of the cation-independent mannose 6-phosphate receptor (CIMPR) from endosomes to the trans-Golgi network, which is involved in delivery of lysosomal hydrolases (Arighi et al., 2004; Seaman, 2004). Hydrolases dissociate from CIMPR in the acidic environment of the endosomal lumen. Several hydrolases undergo proteolytic maturation during endosomal-to-lysosome trafficking and become fully functional in the lysosome (Braulke and Bonifacino, 2009). After delivering the hydrolases to the endosomes, CIMPR is recycled back to the to the trans-Golgi network for a next round of hydrolases delivery. Retromer depletion leads to a decrease of endosomal recycling of CIMPR which results in mistrafficking of hydrolases thereby decreasing lysosomal proteolytic activity, alteration in lysosomal structure and impaired autophagy (Cui et al., 2018). Recently, it has been shown that SNX5 and SNX6 recycled the CIMPR independently of retromer through the recognition of a specific WLM endosome-to-TGN sorting motif (Kvainickas et al., 2017; Simonetti et al., 2017). Although there is discrepancy in the field on weather CIMPR recycling is dependent of the VPS26-VPS35-VPS29 trimer or the SNX-BAR dimer (Cui et al., 2018; Kvainickas et al., 2017; Seaman, 2018; Simonetti et al., 2017), there is consensus in that endosomal mistrafficking results in lysosomal dysfunction. Neuronal loss is a feature in several inherited pediatric lysosomal storage disorders (Futerman and Van Meer, 2004). Hence, lysosomal dysfunction might play a dual role in AD, promoting accumulation of the Aβ and tau, and neuronal loss.

2.2.3 Retromer and postsynaptic function

Neurotransmission is the process through neurons communicate and it takes place in specialized neuronal structures called synapses. The cognitive decline observed in AD, has been associated with early synaptic dysfunction and loss (Forner et al., 2017). Endosomal trafficking at synapses is critical for glutamate receptor recycling backto the plasma membrane, a process mediated by retromer. The depletion of VPS35 leads to decreased levels of glutamate receptor in the plasma membrane, reducing glutamate neurotransmission (Choy et al., 2014; Hussain et al., 2014; Tian et al., 2015). Upon VPS35 depletion spines are also reduced, which can be partially restored by overexpression of glutamate receptors (Tian et al., 2015). Synaptic loss of glutamate receptors is sufficient to produce loss of dendritic spines (Hsieh et al., 2006). Hence, postsynaptic retromer might be involved in both synaptic dysfunction and synapse loss which are key features in AD brains (Forner et al., 2017). These data highlight the importance of endolysosomal sorting in maintaining postsynaptic function and integrity, which may underlie neurodegeneration.

20 General introduction

3 Endolysosomal sorting in presynaptic terminals

In the presynaptic terminals, synaptic vesicles fuse with the plasma membrane to release its neurotransmitter content (Figure 5). Synaptic vesicle release is tightly regulated and 1 has been intensely studied over the past 50 years (see reviews (Südhof, 2013; Südhof and Rizo, 2011; Südhof and Rothman, 2009)). Immediately after exocytosis, membrane is retrieved to reform synaptic vesicles and to restore the extension and tension of the synaptic membrane (Lou, 2018; Maritzen and Haucke, 2017). Membrane retrieval occurs through different mechanisms such us clathrin mediated endocytosis, kiss-and-run, bulk endocytosis, and ultrafast endocytosis (see reviews (Gan and Watanabe, 2018; Milosevic, 2018)). In 1973, synaptic vesicle recycling was described for the first time, when two different mechanism were proposed in parallel. Heuser and Reese proposed that cisternae (endosome-like compartment) mediate synaptic vesicle recycling (Heuser and Reese, 1973). Ceccarelli et al. proposed that synaptic vesicles reform directly from the plasma membrane (Ceccarelli et al., 1973). Since then, more data have supported both models which might fulfill different synaptic requirements. Here, we review morphological, functional and compositional studies which support the involvement of endolysosomal trafficking in synaptic vesicle recycling.

presynaptic terminal degradation

synaptic endosome

clathrin coated pit 7

4. ultrafast 3. bulk endocytosis endosytosis 6 5

8 2.kiss docked/ and run 1.CME primed SV SV fusion

postsynaptic terminal

21 Chapter 1

Figure 5: Synaptic vesicle recycling. In the presynaptic terminal, synaptic vesicles are trafficked to the plasma membrane where they fuse to release neurotransmitters (blue arrows), which are sensed in the postsynaptic terminal. Both synaptic vesicle release and membrane retrieval are tightly coupled to assure synaptic homeostasis. Membrane retrieval can occur through different mechanisms including clathrin mediated endocytosis (CME, 1) kiss-and-run (2), bulk endocytosis (3), and ultrafast endocytosis (4). After membrane retrieval, synaptic vesicles need to be reformed to maintain neurotransmission; however, presynaptic recycling mechanism remains poorly understood. Different models of synaptic vesicle recycling have been proposed (green arrows): synaptic vesicles reform directly from membrane retrieved through CME (5) or bulk endocytosis (6), or vesicles reform from a synaptic endosome intermediate (7) and the ‘kiss and run’ model (8).

3.1 Endosome-like compartments are involved in the synaptic vesicle cycle

In the last decade, ultrafast endocytosis has emerged as a key form of endocytosis in presynaptic terminals (Figure 5). Ultrafast endocytosis and its associated recycling mechanism are visualized with exquisite temporal and spatial resolution by combining optogenetics with high pressure freezing and transmission electron microscopy (Watanabe et al., 2013a). Ultrafast endocytosis happens at physiological temperatures immediately after synaptic vesicle fusion (50–100 ms). After ultrafast endocytosis, a large endocytic vesicle transitions to a synaptic endosome (1s). From the synaptic endosome, coated vesicles appear (3 s), which reform synaptic vesicles (5–6 s after stimulation) (Watanabe et al., 2013a). When ultrafast endocytosis is inhibited by low temperature or actin disruption, synaptic vesicles are retrieved directly from the plasma membrane by clathrin-mediated endocytosis (Watanabe et al., 2013a; Watanabe et al., 2013b). Clathrin is also required to generate synaptic vesicles from the synaptic endosome after ultrafast endocytosis retrieval (Watanabe et al., 2013b). Ultrafast endocytosis is speed by synaptojanin and endophilin which are not required for the synaptic vesicle reformation from the endosome (Watanabe et al., 2018). Endosomal recycling constitutes a fast recycling pathway for synaptic vesicles at physiological conditions.

3.2 Endolysosomal machinery is involved in the synaptic vesicle cycle

Many endosomal molecules have been found both in presynaptic terminals and synaptic vesicles, including Rab proteins, endosomal SNAREs, lysosomal proteins, and coat proteins as AP-1 and AP-3 among others (Morgan et al., 2013; Takamori et al., 2006). However, some of the canonical endosomal proteins are not present in presynaptic terminals, such as the transferrin receptor (TfR) and EEA1 (Cameron et al., 1991; Wilson et al., 2000). These molecules can be grouped in early endosomal, late endosomal and lysosomal machinery.

22 General introduction

3.2.1 Early endosomal machinery

3.2.1.1 PI(3)P 1 Phosphatidylinositol-3-phosphate (PI(3)P) is enriched in the early endosomal membrane, where active Rab5 recruits phosphatidylinositol-3-kinases to trigger the local enrichment of PI(3)P (Christoforidis et al., 1999b). Rab5, EEA1 and other endosomal effectors contain a FYVE zinc-finger domain which specifically binds to PI(3)P. Hence, tag versions of 2xFYVE domain are used as a marker for the PI(3)P-containing endosomes (Wucherpfennig et al., 2003). PI(3)P-containing endosomes have been observed in Drosophila neuromuscular junction by combining GFP-2xFYVE domain and immunoelectron microscopy. In contrast with other endosomal markers such as Rab5, presynaptic PI(3)P seems to be restricted to synaptic endosomes (Wucherpfennig et al., 2003). This presynaptic GFP-2xFYVE-labeled endosomes disappear when synaptic vesicles are depleted by continuous stimulation at 30 Hz while blocking endocytosis. When the endocytic block is removed, this endosome is recovered (Wucherpfennig et al., 2003), suggesting membrane exchange between the synaptic vesicle pool and PI(3)P-containing endosomes.

Apart from being the only reported specific marker for synaptic endosomes, PI(3)P also seems to play a role in the synaptic vesicle cycle. Rizzoli et al. studied the impact of inhibiting phosphoinositide-3 kinases in the synaptic vesicle cycle (Rizzoli and Betz, 2002). Wortmannin is an irreversible inhibitor of phosphoinositide-3 kinases which inhibits FM dye uptake at the frog neuromuscular junction (Richards and Betz, 2000). In tetanized terminals treated with wortmannin, synaptic vesicles are depleted and there is an accumulation of cisternae (Rizzoli and Betz, 2002). These results were reproduced with a reversible inhibitor of phosphoinositide-3 kinases (LY294002). When PI(3)P formation is blocked, the number of synaptic vesicles is reduced while there is an accumulation of cisternal membrane (Rizzoli and Betz, 2002). Hence, phosphoinositide-3 kinases are required for synaptic vesicle reformation but not for endocytosis, suggesting a role of the PI(3)P-containing endosomes in synaptic vesicle reformation.

3.2.1.2 Endosomal SNARE proteins

The endosomal SNARE proteins Syntaxin 13, Vti1a, Syntaxin 6 and VAMP4 are involved in the fusion between the trans-Golgi network and endosomes. These SNAREs have been detected in synaptosomes and synaptic vesicle fractions by western blot, immunoelectron microscopy and proteomics (Rizzoli et al., 2006; Takamori et al., 2006). Endosomal SNAREs recycle as fast as the synaptic vesicle protein VAMP2/Synaptobrevin-2. The proportion of released vesicles containing endosomal SNARE (vs VAMP2/Synaptobrevin-2) decreases

23 Chapter 1 upon prolonged stimulation (20 Hz/30 s) when compared with a shorter stimulation (20 Hz/2 s). Hence, endosomal SNAREs seem to be mainly present on vesicles prone to fuse. In line with these results, endosomal SNARE proteins (Vti1a, Syx6, Syx13) are enriched in recently endocytosed vesicles, together with other endosomal markers such us Rab4, Rab5 and PI(3)P (Hoopmann et al., 2010). Finally, blocking fusion with endosomes using soluble Syntaxin 13 fragments strongly reduces the vesicles labeled with FM 4-64FX upon 20Hz/2s stimulation (Hoopmann et al., 2010). Therefore, the fusion of endocytic material with a synaptic endosome seems to increase the fusiogenicity of the reformed synaptic vesicles.

3.2.1.3 Rab5

Rab5 has been identified in both synaptic vesicles and in presynaptic endosomal structures (Meltsje et al., 1994; Shimizu et al., 2003; Stahl et al., 1994; Wucherpfennig et al., 2003). In the embryonic nervous system, loss of Rab5 implies the loss of PI(3) P-containing endosomes. In addition, when the function of Rab5 is lost exclusively in the nervous system by expression of a dominant-negative GDP-bound Rab5 mutant (Rab5S43N), PI(3)P-containing synaptic endosomes are disrupted (Wucherpfennig et al., 2003). Hence, Rab5 is required for the integrity of presynaptic PI(3)P-containing endosomes.

In presynaptic terminals, expression of a Rab5 dominant-negative mutant leads to an increase in big vesicles with a diameter of 70 nm (Shimizu et al., 2003; Wucherpfennig et al., 2003). These big vesicles have been interpreted as endocytic intermediates due to the lack of fusion with synaptic endosomes and as large synaptic vesicles, product of synaptic vesicle homotypic fusion (Shimizu et al., 2003; Wucherpfennig et al., 2003). Therefore, Rab5 seems to mediate the fusion of endocytic material with synaptic endosomes and to prevent synaptic vesicle homotypic fusion. Impaired Rab5 function decreases neurotransmitter release probability (Shimizu et al., 2003; Wucherpfennig et al., 2003), while Rab5 overexpression or expression of constitutive active Rab5 increases the neurotransmitter release efficacy (Uytterhoeven et al., 2011; Wucherpfennig et al., 2003). Therefore, Rab5 increases the efficacy of synaptic vesicle release, probably by mediating the fusion of endocytic material with endosomes.

3.2.1.4 Rab35

Rab GTPases are turned off (RabGDP-bound) by GTPase-activating proteins (GAPs). Sky is a GAP protein which mainly turns off Rab35, which is an endosomal Rab protein enriched at neuromuscular junction terminals. An increase in Rab35-GTP causes endocytic material

24 General introduction to cycle excessively to endosomes, which results in more clearance of dysfunctional synaptic vesicles proteins. In line with this mechanism, the dominant negative Rab35 strongly reduces neurotransmitter release and the constitutive active Rab35 leads to a vesicle protein pool in which synaptic proteins are more recently synthesized. Sky loss 1 results in increase neurotransmitter release mediated by an increase in Rab35-GTP (active). Hence, the Rab35-Sky pathway restricts endosomal trafficking of synaptic vesicle components, thereby regulating synaptic vesicle fusiogenicity (Uytterhoeven et al., 2011).

3.2.2 Late endosomal machinery

The late endosomal protein Rab7 is found in purified synaptic vesicles (Takamori et al., 2006). The expression of constitutive active mutant Rab7 inhibits synaptic vesicle reformation in Drosophila neuromuscular junction (Uytterhoeven et al., 2011), and its depletion restores the increased neurotransmission in Sky mutants (Fernandes et al., 2014). Snapin mediates retrograde transport of GFP-Rab7 late endosomes in axons, and lack of Snapin results in larger presynaptic terminals which contain more synaptic vesicles but release less (Di Giovanni and Sheng, 2015). Hence, Rab7 and late endosomal retrograde trafficking seem required to facilitate neurotransmitter release. The ESCRT complex sorts ubiquitinated proteins into multivesicular bodies (see Figure 4). The increase in neurotransmitter release in Sky mutants is rescued by the depletion of hrs, vps23, vps25, and vps32, members of ESCRT 0, ESCRT I, ESCRT II, or ESCRT III respectively. Hence, ESCRT-mediated sorting facilitates the increased neurotransmitter release upon loss of Sky function. Ubiquitinated proteins targeted to degradation are more effectively cleared upon loss of Sky function. However, if Sky and hrs defects are combined, the clearance is not increased suggesting a specific role of the ESCRT complex in clearance of dysfunctional synaptic vesicle proteins (Uytterhoeven et al., 2011). Hsc70-4 is a chaperon involved in targeting proteins to degradation via microautophagy (endosome-mediated autophagy) based in sorting motif signals. Hsc70- 4 loss results in less microautophagy and impaired neurotransmission, while Hsc70-4 overexpression promotes microautophagy and neurotransmitter release. Hence, Hsc70 seems to stimulate microautophagy and protein turnover, which results in a younger and more fusiogenic synaptic vesicle pool (Uytterhoeven et al., 2015). Taken together, the late endosomal machinery is involved in clearance of old, dysfunctional presynaptic proteins which regulates the fusiogenicity of the synaptic vesicle pool.

3.2.3 Lysosomal machinery

The homotypic fusion and vacuole protein sorting (HOPS) complex is involved in the

25 Chapter 1 fusion of cargo vesicles with the lysosome. The HOPS complex is formed by Vps11, 16, 18, 33, 39, and 41. Dor is an orthologue of Vps18 in Drosophila which loss reduces the ready releasable pool without affecting synaptic ultrastructure. In dor mutants, the synaptic vesicle protein-pool age is older and Pro–Cathepsin L accumulates, indicating an impairment in degradation of synaptic proteins. In addition, loss of dor or depletion of vps39 (both components of HOPS) rescues the neurotransmission increase in Sky mutants, indicating that trafficking of vesicles to endosomes is not sufficient to increase neurotransmitter release and that HOPS complex-dependent traffic to lysosomes is required as well (Fernandes et al., 2014).

4. Conclusions

Endolysosomal sorting assures cell homeostasis in eukaryotic cells. Across species, the endolysosomal sorting machinery and sorting mechanisms are conserved. Many endolysosomal gene mutations have been notably associated with neurodegenerative disorders, including AD. Endolysosomal sorting plays a key role at the postsynaptic side of the synapse, but its role at the presynaptic side is less understood. The synaptic endosome provides a fast membrane recycling pathway and the modulation of endolysosomal machinery can alter presynaptic neurotransmitter vesicle fusion. However, which endolysosomal sorting machineries and mechanisms are important for neuronal function in health and disease are largely unclear.

26 General introduction

5. Aim of the thesis

The general goal of this thesis was to study the neuronal endolysosomal system in presynaptic function and in the context of Alzheimer’s disease. 1

The first aim was to investigate the role of key endosomal sorting proteins in presynaptic terminals. We characterized the presynaptic distribution of key endosomal sorting proteins and investigated their function using a shRNA approach in mouse primary cultured neurons. Upon depletion of these presynaptic endosomal sorting proteins, we have evaluated presynaptic structure, function and composition. In Chapter 2, we studied VPS35, the core component of retromer. VPS35 was present in presynaptic terminals but its deletion did not affect presynaptic structure, function or composition. In Chapter 3, we developed a novel antibody against SNX4 to characterize its subcellular distribution in neurons. SNX4 was found in presynaptic terminals and its depletion dysregulated the neuronal proteome without affecting presynaptic ultrastructure or neuronal morphology. The localization of SNX4 and VPS35 as presynaptic proteins suggests a selective demand for endosomal recycling in presynaptic terminals.

The second aim of this thesis was to investigate the link between tau pathology and endolysosomal aberrations (Chapter 4). We evaluated how the seeding of tau pathology impacted the endolysosomal system. Seeding assays of tau pathology were implemented in HEK293 cell, mouse primary neurons and iPSC-derived human neurons, which recapitulated the main hallmarks of tau pathology including tau hyperphosphorylation, misfolding and insoluble aggregation. In these cellular models, the endolysosomal system was characterized using confocal microscopy and three-dimensional stimulated emission depletion microscopy (3D-STED).

Finally, Chapter 5 summarizes the main results of this thesis putting them in a broader discussion with existing literature and suggests future research directions.

27 28 Chapter 2 VPS35 depletion does not impair presynaptic structure and function

29 Chapter 2

VPS35 depletion does not impair presynaptic structure and function

Sonia Vazquez-Sanchez1, Sander Bobeldijk1, Marien P. Dekker2, Linda van Keimpema1, 3, and Jan R.T. van Weering 2, *

1Department of Functional Genomics, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, Vrije Universiteit (VU), Amsterdam, Netherlands

2Clinical Genetics, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, VU medical center, Amsterdam, Netherlands

3Sylics (Synaptologics BV), PO box 71033, 1008 BA, Amsterdam, The Netherlands

*Corresponding author: Jan R.T. van Weering, Center for Neurogenomics and Cognitive Research, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. Email: [email protected]

30 VPS35 in presynaptic terminals

ABSTRACT

The endosomal system is proposed as a mediator of synaptic vesicle recycling, but the molecular recycling mechanism remains largely unknown. Retromer is a key protein complex which mediates endosomal recycling in eukaryotic cells, including neurons. Retromer is important for brain function and mutations in retromer genes are linked to neurodegenerative diseases. In this study, we aimed to determine the role of retromer in presynaptic structure and function. We assessed the role of retromer by knocking down VPS35, the core subunit of retromer, in primary hippocampal mouse neurons. VPS35 depletion led to retromer dysfunction, measured as a decrease in GluA1 at the plasma membrane, and bypassed morphological defects previously described in chronic retromer depletion models. We found that retromer is localized at the mammalian presynaptic 2 terminal. However, VPS35 depletion did not alter the presynaptic ultrastructure, synaptic vesicle release or retrieval. Hence, we conclude that retromer is present in the presynaptic terminal but it is not essential for the synaptic vesicle cycle. Nonetheless, the presynaptic localization of VPS35 suggests that retromer-dependent endosome sorting could take place for other presynaptic cargo.

31 Chapter 2

INTRODUCTION

Retromer is a protein complex that regulates endosomal recycling in all eukaryotic cells (Koumandou et al., 2011). Retromer was first described in yeast (Seaman et al., 1998) and is highly conserved across all the lineages (Koumandou et al., 2011). The retromer complex is formed by two essential modules: the cargo-selection subcomplex, which binds to the protein that has to be transported, and the membrane deformation subcomplex, which binds to the endosomal membrane to produce the necessary membrane deformation for trafficking (see review (Burd and Cullen, 2014)). The cargo-selection subcomplex in mammals is constituted by VPS35, VPS29 and VPS26A or VPS26B (Haft et al., 2000). VPS35 is the largest protein and the central subunit of this trimetric complex (Hierro et al., 2007). The membrane deformation subcomplex is constituted of SNX-BAR proteins (SNX1, SNX2, SNX5, SNX6 and SNX32 in mammals). SNX-BAR proteins dimerize in different patterns which leads to a variety of retromer complexes, (although some variants of retromer complex do not contain SNX-BAR proteins, (see reviews (Cullen and Korswagen, 2012; van Weering and Cullen, 2014)). These two modules together transport the cargo from the endosome to the trans-Golgi network (Seaman et al., 1998) or to the plasma membrane (Temkin et al., 2011). Retromer is essential for a great variety of cell functions by specific sorting of membrane proteins: Retromer is involved inWnt- dependent development (Franch-Marro et al., 2008), epithelial polarity (Zhou et al., 2011), neuronal morphogenesis (Korolchuk et al., 2007; Tian et al., 2015; Wang et al., 2012), autophagy (Tang et al., 2015), nutrition (Arighi et al., 2004) and lysosomal degradation (Miura et al., 2014) among other cell processes. The central role of retromer is also highlighted by the observation that the lack of retromer is lethal during embryonic stages, both in mammalian (Lee et al., 1992; Muhammad et al., 2008) and fly models (Franch- Marro et al., 2008; Zhou et al., 2011).

Retromer dysfunction is linked with Parkinson’s and Alzheimer’s disease among other neurological disorders (see review (Small and Petsko, 2015)). In fact, increasing retromer stability has been proposed as a therapeutic target for these neurodegenerative diseases (Chu and Praticò, 2017; Follett et al., 2016; Mecozzi et al., 2014). Although retromer seems a promising drug target, very little is still known about the neurobiological function of retromer. Hence the physiological role of retromer in the brain needs to be addressed.

The most characteristic neuronal function is to communicate through neurotransmission, a process which takes place at synapses. Functional studies have established that retromer regulates adrenergic and glutamatergic neurotransmitter receptor trafficking to the postsynaptic plasma membrane (Choy et al., 2014; Hussain et al., 2014; Tian et al., 2015). Retromer has been found dynamically localized at the synapses in murine

32 VPS35 in presynaptic terminals neurons (Bhalla et al., 2012; Choy et al., 2014; Munsie et al., 2015), and VPS35 is found in synaptosomal membranes and synaptic vesicle enriched fractions isolated from rodent brain (Tsika et al., 2014). A recent study in Drosophila suggests that Vps35 is in presynaptic terminals at the edge of the active zone, where it regulates synaptic vesicle recycling (Inoshita et al., 2017). In both Parkinson’s and Alzheimer’s disease, central proteins that contribute to the pathology are found in presynaptic terminals (APP (Das et al., 2016; Laßek et al., 2015) and α-synuclein (Lashuel et al., 2013; Murphy et al., 2000)), which might suggest that retromer-dependent trafficking occurs at presynaptic terminals. However, to our knowledge there is no report investigating retromer role in the mammalian presynaptic terminal.

The aim of this study was to determine the role of retromer in presynaptic structure and 2 function. We first investigated the location of VPS35, the core subunit of retromer, in the presynaptic terminal with confocal and immuno-electron microscopy techniques in mouse hippocampal neurons. To address retromer function, we acutely depleted retromer subunit VPS35 to evaluate the impact of this depletion on the presynaptic ultrastructure using immunocytochemistry and electron microscopy, and the impact on synaptic vesicle release and retrieval using life cell imaging with pHluorin secretion reporters (synaptopHluorin (Granseth et al., 2006) and sypHy (Miesenböck et al., 1998)).

RESULTS

VPS35 is in the presynaptic terminal

We first characterized the distribution of retromer in mouse hippocampal synapses. We performed immunocytochemistry against VPS35 together with a synaptic marker (VAMP2/Synaptobrevin-2) in cultured wild-type neurons after 14 days in vitro (DIV14). We performed co-localization studies only in the neurites in order to exclude the endosomes present in the cell body from the analysis. Approximately 22% of VPS35 immunoreactivity showed also VAMP2/Synaptobrevin-2 reactivity (Manders’ coefficient M1: 0.22±0.01), while approximately 35% of synapses contained VPS35 signal (Manders’ coefficient M2: 0.35±0.02) (Fig. 1a, b). Therefore, the confocal data show that retromer can be found at synaptic locations in hippocampal mouse neurites. We immunostained free floating sections of wild-type mouse brain to test in which synaptic compartments VPS35 can be found using electron microscopy. We observed immunoreactivity in hippocampal presynaptic terminals, but not all synapses showed VPS35 immunoreactivity (Fig. 1c’). VPS35-positive presynaptic terminals showed a dark precipitate around the whole synaptic vesicle cloud (Fig. 1c’’, c’’’). VPS35 immunoreactivity was more frequently found in the presynaptic side (82.4%) than in the postsynaptic side (17.6%; SEM=1.8, Fig. 1d),

33 Chapter 2

Merge 1.0 a VAMP2 VPS35 VAMP2 VPS35 b 0.8 0.6 0.4 0.2 Mander's coefficient Mander's 0.0 VAMP2 VPS35 in VPS35 in VAMP2

d 17.6 % Post 82.4 % Pre

c’ c’’

T

PSD

c’’’

T

PSD

e’ e’’ f’ f’’

T T T T PSD PSD PSD PSD

Figure 1: VPS35 is present in presynaptic terminals. (a) Representative confocal microscopy images of hippocampal neurons immunolabeled for VAMP2/Synaptobrevin and VPS35. Arrowheads indicate co-localization between VAMP2/Synaptobrevin and VPS35 puncta. Scale bar of the neuron image=20 μm, scale bar of the zoomed neurite=3 μm. (b) Mander’s coefficients for the co-localization between VAMP2/Synaptobrevin and VPS35 in neurites (n=78 fields of view, N=3 animals). (c) Representative electron micrographs of hippocampal synapses from 3 independent wild-type mice. Each of the micrographs correspond to a different animal (N=3 animals). Scale bar=200 nm. (c’) In black arrowheads indicate the synaptic vesicle cloud of DAB positive presynaptic terminals and in white arrowheads DAB negative presynaptic terminals. (c’’, c’’’). Zoom in of DAB positive presynaptic terminals. (d) Percentage of synapses with VPS35 immunoreactivity in the presynaptic site (82.4% Pre) and postsynaptic site (17.6% Post). (e’, e’’) Immunoelectron micrographs of presynaptic terminals stained with a rabbit antibody against VPS35 labelled with Protein A-10nm gold conjugate. The images are representative of two independent experiments (N=2 animals). Scale bar=200 nm (f’, f’’) Immunoelectron micrographs of presynaptic terminals stained with a goat antibody against VPS35 labelled with a secondary antibody rabbit anti-goat and Protein A-10nm gold conjugate. The images are representative of two independent experiments (N=2 animals). Scale bar=200 nm. ‘PSD’ indicates postsynaptic side and ‘T’ the presynaptic terminal.

34 VPS35 in presynaptic terminals and it was not observed in negative controls (blocking peptide, Supplementary Fig. S2). The presynaptic location of VPS35 was verified by immuno-gold electron microscopy using two different antibodies against VPS35 detected by Protein A-gold 10nm. VPS35 immunosignal of both antibodies was detected inside presynaptic terminals (Fig. 1e-f, Supplementary Fig. S1), but not in the negative controls (absence of primary antibody, Supplementary Fig. S2). VPS35 immuno-gold signal was also present in some postsynaptic structures (Supplementary Fig. S3). Overall, these data show that VPS35 is present in presynaptic terminals, but not all synapses contain retromer.

Synaptic VPS35 is functionally knocked down by independent shRNAs

VPS35 is a crucial protein for maintaining the stability of all known retromer complexes 2 (Norwood et al., 2011; Mecozzi et al., 2014). In order to decrease retromer levels and to be able to define the role of retromer at the presynaptic terminal, we designed three short hairpin RNAs (shRNA) against VPS35. Previous studies documented that retromer depletion affects hippocampal development inducing defects in neurites and spine density (Tian et al., 2015; Wang et al., 2012); hence, we aimed to bypass possible changes in neuronal morphology, which may confound our presynaptic functional studies. Therefore, we infected cultured neurons with lentivirus particles containing shRNAs when the neurons already formed synapses (DIV7). We assessed the shRNA efficiency in reducing VPS35 protein in neurons at DIV14-DIV15 by quantifying VPS35 protein levels using both immunocytochemistry and western blot. The immunostaining of primary cortical neuron network cultures revealed that the three shRNA were able to reduce VPS35 expression: shVPS35-1 significantly reduced the VPS35 staining by 81%, shVPS35-3 significantly reduced VPS35 by 80%, while shVPS35-2 reduced the VPS35 signal just by 31%, which was not significantly different compared to control levels (Fig. 2a, c; Supplementary Table S1). These results were replicated by western blot (Fig. 2b, d). Therefore, we conclude that VPS35 protein level was reduced by independent shRNAs.

Next, we tested if the VPS35 depletion induced a functional impairment in retromer in neurons. Previous studies showed that impairment in retromer-SNX27 pathway results in a reduction of GluA1 at the neuronal surface (Hussain et al., 2014; Tian et al., 2015). In accordance to these studies, we find that the three shRNA against VPS35 lead to a significant reduction of GluA1 surface staining compared to control (GluA1 levels significantly reduced by 35% (shVPS35-1), 27% (shVPS35-2) and 41% (shVPS35-3) (Fig. 2e, f; Supplementary Table S1). We conclude that seven-day infection with all three independent shRNAs against VPS35 led to retromer dysfunction in mouse neurons.

35 Chapter 2

a b kDa shRNA Control shVPS35-1 shVPS35-2 shVPS35-3 VPS35 reporter VPS35 100

Actin lortnoC 40 c **** d ** 2.0 **** ** 1.5 3

1.0 2

0.5 1

Relative VPS35 (ICC) VPS35 Relative 0.0 (WB) VPS35 Relative

Control Control shVPS35-1 shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3 e shRNA reporter GluA1 (surface staining)

Control

shVPS35-1

shVPS35-2

shVPS35-2 shVPS35-3

f **** ** 2.0 ****

1.5

1.0

shVPS35-3 0.5

GluA1 normalize to control normalize GluA1 0.0

Control shVPS35-1shVPS35-2shVPS35-3

Figure 2: Synaptic VPS35 is functionally knocked down by three independent shRNA. (a) Representative confocal microscopy images of cortical neurons infected with control shRNA and the three shRNAs against VPS35. Left, mCherry signal reporting the expression of lentivirus containing the shRNAs coding sequences. Right, neurons immunolabeled for VPS35. Scale bar=50 μm. (b) Representative western blot showing the knock down of VPS35 by three independent shRNAs. Original uncropped blots are shown in Figure S1. (c) Quantification of VPS35 intensity in immunostainings (n=25±1 fields of view, N=2 animals) (d) Quantification of VPS35 levels normalized to total protein levels (assessed by TCE staining) in western blot. Values are presented as a ratio compared to the control condition. (N=5±1 blots/animals). (e) Representative confocal microscopy images of hippocampal neurons expressing either control shRNA or one of the three shRNAs against VPS35. Left, mCherry signal which reports expression of lentivirus containing the shRNAs. Right, surface immunolabelling of GluA1. Scale Bar=5 μm. (f) Quantitative analyses of GluA1 staining intensity (n=35±1 fields of view, N=3 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

Presynaptic structure is not affected by knocking down VPS35 in neurons

To test if the acute retromer depletion bypasses the previously reported morphological defects(Tian et al., 2015; Wang et al., 2012), we stained autaptic neurons at DIV14-

36 VPS35 in presynaptic terminals

a Merge b

MAP2 SMI-312 MAP2 SMI-312 VAMP2 Synaptophysin Bassoon

Control Control

shVPS35-1 shVPS35-1 shVPS35-2

shVPS35-2 2

shVPS35-3 shVPS35-3

** c 6 d 10 e 0.4 ** m mm) 8 µ 0.3 4 6 0.2 2 4 2 0.1 Axonal lengh (mm) lengh Axonal

Dendritic lengh ( lengh Dendritic 0 0 0.0 VAMP2 synapse/ VAMP2

Control Control Control f shVPS35-1shVPS35-2shVPS35-3 g shVPS35-1shVPS35-2shVPS35-3 h shVPS35-1shVPS35-2shVPS35-3 200 200 0.5 m µ 150 150 0.4 0.3 100 100 0.2 50 50 0.1 VAMP2 (a.u.) VAMP2 0 0 0.0 Synaptophysin (a.u.) Synaptophysin Bassoon synapse/ Bassoon

Control Control Control shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3

Figure 3: Neuronal morphology is not affected by knocking down VPS35 in neurons. (a, b) Representative confocal microscopy images of hippocampal autaptic neurons containing either control shRNA or one of the three shRNAs against VPS35. Scale bar=50 μm (a) Immunolabelling for MAP2 and SMI-312. Merge image of the MAP2 (green) SMI-312 (magenta). (b) Immunolabelling for VAMP2/Synaptobrevin, Synaptophysin and Bassoon. Quantification of (c) the dendritic length (n=56±8 neurons, N=5 animals); (d) axonal length n=38±8, N=3 animals); (e) VAMP2/Synaptobrevin- labelled synaptic density (n=34±7 neurons, N=3 animals); (f) of VAMP2/Synaptobrevin staining intensity n=34±7 neurons, N=3 animals). (g) Synaptophysin staining intensity (n=34±7 neurons, N=3 animals). (f) Bassoon staining intensity (n=20±1 neurons, N=2 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

37 Chapter 2

15 for dendritic (MAP2), axonal (SMI-312) and synaptic (VAMP2 /Synaptobrevin-2, Synaptophysin and Bassoon) markers. We used autaptic neuronal cultures as they allowed us to measure the neuritic arbor from a single neuron (Miesenböck et al., 1998). We analyzed the confocal images using SynD, a semi-automated image analysis routine (Schmitz et al., 2011). The overall neuron morphology did not change upon VPS35 knock down (Fig. 3a, b). The following morphological parameters did not differ between the control and VPS35 shRNA-expressing neurons: dendritic length, axonal length, synapse density calculated as Bassoon positive puncta per µm of neurite and, the expression levels of synaptic vesicle proteins such as VAMP2/Synaptobrevin-2 and Synaptophysin (Fig. 3c, d, f-h; Supplementary Table S1). We observed a significant difference in the number of VAMP2/Synaptobrevin-2 positive puncta per µm, which was reduced by 18% in shVPS35-2 and shVPS35-3 infected neurons compared to control (Fig. 3e; Supplementary Table S1). To verify that changes in neurite length can be detected in this assay, we compared an early time point (DIV4) of autaptic neuronal cultures, when neuron morphology is less complex, with the mature time point used in this study (DIV14). Quantitative analysis detected both axonal and dendritic length increased significantly during neuronal development (Supplementary Fig. S5). In addition, this methodology has been used in our laboratory previously (Arora et al., 2017; Keimpema et al., 2017; Melero et al., 2017; Schmitz et al., 2016). Together, these data show that acute knock down of VPS35 in mature neurons does not alter most aspects of neuronal morphology.

We next explored the presynaptic ultrastructure upon VPS35 knock down by Transmission Electron Microscopy (TEM) using aldehyde fixation at DIV14-15. The overall synaptic morphology of shRNA-expressing neurons did not show abnormalities (Fig. 4a). The active zone length, the total amount of synaptic vesicle and, the number of docked synaptic vesicles did not differ between the four groups (Fig. 4b-d). These data show that acute VPS35 depletion does not affect presynaptic ultrastructure.

Synaptic vesicle release is not altered by knocking down VPS35

We tested the potential role of VPS35 in presynaptic function by using fluorescent reporters of synaptic vesicle release and retrieval. First, we used sypHy, a pH-sensitive variant of GFP fused in the luminal domain of the synaptic vesicle protein Synaptophysin-1 (Granseth et al., 2006). This pH-sensitive reporter allows the visualization of both synaptic vesicle release and retrieval. Our protocol consisted of an electrical stimulation (100 AP,

40 Hz, 30 mA) to evoke synaptic vesicle release followed by an exposure to NH4Cl (de- quenching all sypHy) to quantify the total reporter pool (Fig. 5a, b). The reporter showed a punctate pattern upon NH4Cl application which is used to place regions of interest (ROIs) for sypHy measurements (Fig. 5a). In this experiment, we excluded neurons infected with

38 VPS35 in presynaptic terminals

a Control shVPS35-1 shVPS35-2 shVPS35-3

b c d 2.5 800 20

2.0 600 15 1.5 400 10 1.0 200 5

0.5 vesicles # synaptic

Active zone length (µm) length zone Active 2 0.0 0 vesicles synaptic # docked 0

Control Control Control shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3

Figure 4: Presynaptic nanostructure is not altered in VPS35 knock down neurons. (a) Typical examples of electron micrographs of hippocampal synapses from control and VPS35 knock down neurons. Scale bar=200nm. The quantitative parameters (b) active zone length (c) total number of synaptic vesicles and (d) docked synaptic vesicles are indicated as bar graphs. (n=162±3 synapses, N=3 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

shVPS35-1 because they failed to show a punctate sypHy pattern upon NH4Cl exposure, which is required for data analysis (data not shown). Control, shVPS35-2 and shVPS35-3 expressing neurons showed a similar time course of fluorescence intensity with a timed response to the electrical stimulation and recovery back to baseline fluorescence within 60 seconds after the stimulus (Fig. 5b). The synaptic vesicle release, which is measured as peak amplitude, was not significantly different between the control neurons and neurons expressing shVPS35-3, but it was increased compared with shVPS35-2 (Fig. 5c). The three groups showed the same percentage of active synapses defined as the percentage of ROIs that respond both to electrical stimulation and NH4Cl perfusion (Fig. 5d). The total pool of sypHy was significantly reduced in VPS35-depleted neurons compared with control (Fig. 5e; Supplementary Table S1). These data suggest that VPS35 knock down does not affect synaptic vesicle release or retrieval.

As the effect of shVPS35-1 could not be evaluated using sypHy, we tested also synaptopHluorin as an alternative reporter of synaptic vesicle release and retrieval. SynaptopHluorin is a pH-sensitive variant of GFP fused to the luminal domain of VAMP2/ Synaptobrevin-2 which works as sypHy, but it shows more cell surface expression which increases the background fluorescence (Miesenböck et al., 1998). Hence, at the end

39 Chapter 2

NH4Cl a b Control 1.0 shVPS35-2 Baseline shVPS35-3 0.2 Stimulus 0.5 0.1

F/Fmax Stimulus Recovery ∆ 0.0 30 35 40

NH4Cl 0.0

20 40 60 80 100 Time (seconds) c d e 0.6 150 2200 * * **

F/Fmax) 2000 ∆ 0.4 100 1800 1600

0.2 50 (a.u.) Flmax 1400 0.0 (%) synapses active 0 1200 peak amplitude ( amplitude peak

st Control Control Control 1 shVPS35-2shVPS35-3 shVPS35-2shVPS35-3 shVPS35-2shVPS35-3

Figure 5: VPS35 KD does not affect sypHy release or retrieval. (a) Representative sypHy fluorescence images of neurites during baseline, stimulation, the recovery period and the exposure to NH4Cl. Scale bar=10 µm. (b) Time course of sypHy fluorescence during the imaging protocol, plotted as ΔF/Fmax. The grey box indicates the electrical stimulation (100 AP, 40 Hz, 30 mV) and the black box the exposure to 10 seconds of NH4Cl (n=23±8 fields of view, N=4 animals). (c) Maximum response amplitude during the electrical stimulation plotted as ΔF/Fmax (d) Percentage of responsive synapses during the stimulation (e) Maximum response to the exposure to NH4Cl. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

of the protocol using this reporter, we exposed the neurons to a pH = 5.5 solution to calculate the fraction of synaptopHluorin that remained in the plasma membrane. We also added a second stimulation to the new protocol (identical to the first one) to measure if retromer depleted neurons were able to efficiently release synaptic vesicles after having been already electrically stimulated, which would indicate if retromer is involved in refilling the releasable synaptic vesicles pools. Using synaptopHluorin as reporter, all the experimental groups showed the typical puncta pattern when treated with NH4Cl (Fig. 6a); thus, all groups were included in analysis. All neurons showed a similar time course of fluorescence intensity during the protocol and peak amplitude to the first response (Fig. 6b, c). Compared to control, shVPS35-2 and shVPS35-3 expressing neurons showed the same number of active synapses and the same baseline fluorescence. However, shVPS35-1 infected neurons showed a significant decrease in these two parameters compared to control (Fig. 6d, f; Supplementary Table S1). Neurons infected with shVPS35-3 showed the same response to NH4Cl, but shVPS35-1 and shVPS35-2 infected neurons showed a decreased response compared to control (Fig. 6e). The ratio between the fluorescence peaks after stimulation was equal between all the groups (Fig. 6g). The

40 VPS35 in presynaptic terminals

Baseline Stimulus Recovery NH4Cl pH = 5.5 a

Control b shVPS35-1 NH4Cl pH = 5.5 1.0 shVPS35-2 shVPS35-3 0.2

0.1

0.5 0.0 30 35 40 F/Fmax

∆ Stimulus Stimulus 2 0.0 50 100 150

Time (seconds)

c -0.5 d e 0.25 150 15000 **

F/Fmax) 0.20 * ∆ 100 10000 0.15 0.10 50 Fmax(a.u.) 5000 0.05 active synapses (%) synapses active 0.00 0 0 peak amplitude ( amplitude peak st 1 Control Control Control f g h

shVPS35-1shVPS35-2shVPS35-3 ) shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3

6000 st 2.0 8000 /1

nd 1.5 6000 4000 ** 1.0 4000 2000 pH pH = (a.u.) 5.5

Baseline F (a.u.) Baseline 0.5 2000

0 0.0 0

Control Control Control Ratio peak amplitude (2 amplitude peak Ratio shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3 shVPS35-1shVPS35-2shVPS35-3

Figure 6: Repetitive stimulation does not induce presynaptic failure in VPS35 KD neurons. (a) Representative synaptopHluorin fluorescence images of neurites during the base line, the first stimulation, the first recovery period, the exposure to NH4Cl and the exposure to pH=5.5. Scale bar=40 µm. (b) Time course of synaptopHluorin fluorescence during the imaging protocol, plotted as ΔF/Fmax. The grey boxes indicate the electrical stimulation (100 AP, 40 Hz, 30 mV each), the black box the duration of the exposure to NH4Cl and the white box the duration of the exposure to pH=5.5. (n=24±2 fields of view, N=3 animals). (c) Maximum response amplitude during the electrical stimulation plotted as ΔF/Fmax (d) Percentage of responsive synapses during the stimulation (e) Maximum synaptopHluorin levels during exposure to NH4Cl. (f) Average fluorescence of synaptopHluorin during baseline recordings (g) Ratio of the maximum synaptopHluorin fluorescence amplitude between the first and the second electrical stimulation. (h) Minimum response to the exposure to pH=5.5. Detailed information (average, SEM, n and statistics) is available in Supplementary Table S1.

41 Chapter 2 fluorescence during the pH 5.5 wash was also similar between all experimental groups (Fig. 6h). To ensure that the life cell imaging methodology was working as described in literature (Granseth et al., 2006; Miesenböck et al., 1998) and as previously in our laboratory (Arora et al., 2017; Keimpema et al., 2017; Melero et al., 2017; Spangler et al., 2013), we assessed the effect of VPS35 depletion in presence and absence of calcium (Supplementary Fig. S6). When calcium is present, both control and the VPS35 knock down showed a similar amount of synaptic vesicle release, but when calcium is absent, synaptic vesicle release is significantly decreased in both groups. These results show that the assay can register changes in synaptic vesicle release. Together, the experiments performed with sypHy or synaptopHluorin show that acute VPS35 depletion does not affect synaptic vesicle release and retrieval, suggesting that retromer does not affect the synaptic vesicle cycle.

DISCUSSION

The present study addressed the effect of retromer dysfunction in presynaptic structure, and synaptic vesicle release and retrieval. To avoid the potential interference of VPS35 depletion during development, we acutely knocked down VPS35 in neurons after synapse formation. Acute VPS35 depletion resulted in retromer dysfunction, which was measure as a decrease in GluA1 receptors in the plasma membrane (Hussain et al., 2014; Tian et al., 2015). Our results show for the first time that retromer is present at the mammalian presynaptic terminal. VPS35 depletion did not affect most measured neuronal features: neuronal morphology (neurite length and synapse number), presynaptic ultrastructure, and synaptic vesicle release and retrieval. The data show that presynaptic retromer is not essential for basic presynaptic structure and function.

To determine the role of presynaptic retromer we have used a shRNA approach. shRNAs are widely used to acutely deplete proteins, but this method is susceptible to off target effects (see reviews (Fellmann and Lowe, 2014; Kaelin, 2012)). The off-target effects are those genes or processes which are affected by the shRNA that are not the target, in this case VPS35. The three shRNAs have been validated to functionally inhibit retromer by impairing GluA1 surface localization, which has been described to be retromer- dependent by several laboratories (Hussain et al., 2014; Tian et al., 2015). To avoid the phenotypic association with off-target effects, only phenotypes that are replicated by all three shRNAs against VPS35 are considered to be VPS35-dependent processes in this study.

Several studies have shown that chronically decreasing retromer levels causes defects in neuronal morphology and synapse formation (Liu et al., 2014; Tian et al., 2015; Wang

42 VPS35 in presynaptic terminals et al., 2012). We aimed to acutely induce retromer dysfunction in maturing neurons to circumvent these neuronal morphology defects, which might interfere in the evaluation of retromer role in the presynaptic terminal. We did not find changes in any of the measured parameters: dendritic length, axonal length, and the number of synapses or synaptic proteins, avoiding the interference of retromer dysfunction during the initial stages of neuronal network development. We did observe a small reduction of 18% in the number of VAMP2 /Synaptobrevin-2 positive puncta per µm of neurite (Fig. 3e) in shVPS35-2 and shVPS35-3, which might suggest that the number of synapses is reduced. However, this reduction was not observed for synapse marker Bassoon (Fig. 3h). In addition, shVPS35-1 conditions did not replicate this effect on VAMP2/Synaptobrevin-2 puncta (Fig. 3f). Together these results suggest that the reduction in VAMP2/Synaptobrevin-2 puncta does not represent a significant loss of synapses. Hence, we conclude that retromer 2 is not required for maintenance of existing synapses and formation of new synapses in mature neurons.

Retromer depletion did not affect presynaptic ultrastructure. Tian et al. (2015) reported an increase in synaptic vesicles in VPS35 haploinsufficient presynaptic terminals. This increase in synaptic vesicles was coupled with a decrease in neurotransmitter receptor at the postsynaptic sites. Hence, the alteration in presynaptic ultrastructure was interpreted

Table 1: Overview of the effect of the different shRNAs against VPS35 in the measured presynaptic variables compared to control. Significant decrease is note as ‘red arrow’, no significant difference is noted as "=" and not applicable as ‘empty cells’. Detailed information (average, SEM, n and statistics) is displayed in Supplementary Table S1.

Figure meassured variable shVPS35-1 shVPS35-2 shVPS35-3 3c Dendritic length (µm) = = = 3d Axonal length (µm) = = = 3f Synpases/µm (VAMP2) = 3g VAMP2 (a.u.) = = = ⬇ ⬇ 3h Synaptophysin-1 (a.u) = = = 3i Synpases/µm (bassoon) = = = 4b Active zone length (µm) = = = 4c # Synaptic vesicles/synapse = = = 4d # docked synaptic vesicles/synapse = = = 5c 1st peak amplitud (∆F/Fmax) = 5d % Active synapses = = ⬇ 5e Fmax (a.u.) 6c 1st peak amplitud (∆F/Fmax) = = = ⬇ ⬇ 6d % Active synapses = = 6e Fmax (a.u.) = = ⬇ 6f F Baseline (a.u) = = ⬇ 6g Ratio peak amplitud (2nd/1st) = = = ⬇ 6h F pH = 5.5 (a.u.) = = =

43 Chapter 2 as a compensatory mechanism to the impairment in the postsynapse. We observed the decrease in glutamatergic receptor labeling at the cell surface, but the number of synaptic vesicles was not altered, suggesting that such potential secondary effects were indeed circumvented by transient retromer depletion (Fig. 2e, f and 3a, c). In Drosophila, retromer depletion led to a decrease in the number of synaptic vesicles, which was interpreted as a defect in endocytosis and regeneration of the synaptic vesicles due to the lack of retromer (Inoshita et al., 2017). Previous work of this laboratory has shown that addressing presynaptic ultrastructure with TEM is sensitive to detect changes (He et al., 2017; Meijer et al., 2012; Meijer et al., 2017; Schmitz et al., 2016; Toonen et al., 2006; Wierda et al., 2007). However, upon VPS35 knock down we did not observe such an effect. The differences between literature and our data suggest that retromer function may be different in different organisms, or in different synapses, and/or it may change during developmental stages.

Retromer depletion did not affect synaptic vesicle release and retrieval as measured by pH-sensitive reporters. In contrast, VPS35 knock down reduced the total expression of these reporters compared to control (except for shVPS35-3 in SynaptopHluorin experiments) (Fig. 5e and 6e). In the shVPS35-1 condition, the sypHy reporter even failed to show a punctate synaptic localization. The reduced expression of these reporters was not mirrored by the endogenous proteins (VAMP2/Synaptobrevin-2 or Synaptophysin; Fig. 3f, g; Supplementary Table S1), or a general reduction in synaptic vesicles (Fig. 4c). These observations might suggest that retromer is involved in the targeting of newly or exogenous expressed proteins to synaptic terminals. Experiments using labelling of endogenous de novo synthesized proteins may shed light on this. However, the inconsistencies of the observations between groups suggest that the reduction in fluorescent reporter expression could be mediated by non-specific off-target effects of the shRNA approach.

Acute VPS35 depletion does not affect the presynaptic structure and synaptic vesicle release; however, retromer is present in presynaptic terminals. We used immuno-electron microscopy against VPS35 to address VPS35 localization in mammalian synapses for the first time. The nano-resolution of this technique allowed us to demonstrate that VPS35 is present in the murine hippocampal presynaptic terminal, which is in line with a recent study that showed that Vps35 is in the Drosophila presynaptic terminal (Inoshita et al., 2017). A main question remains unanswered: why retromer is present in mammalian presynaptic terminals but does not affect the dominant membrane recycling pathway in that region, the synaptic vesicle cycle. Our confocal data show that about 35% of hippocampal synapses contain retromer, suggesting that retromer may play a role in just a subset of synapses. We hypothesized that retromer plays a role in the modulation of presynaptic

44 VPS35 in presynaptic terminals communication. Recently, it has been shown that dopamine transporter availability at the cell surface is regulated by retromer (Wu et al., 2017). Dopamine transporter mediates the presynaptic reuptake of dopamine, which determines dopaminergic neurotransmission. Potentially retromer may also recycle neurotransmitter transporters in hippocampal terminals. Retromer also mediates G protein–coupled receptors (GPCRs) signaling (see review (van Weering and Cullen, 2011)). GPCRs can be localized at presynaptic terminals and are important regulators of synaptic communication (see review (Atwood et al., 2014)). Hence, through these two recycling pathways, which would not directly affect parameters tested in this study, retromer might be involved in the modulation presynaptic communication.

Our goal was to investigate the role of retromer in presynaptic terminals in terms 2 of structure and synaptic vesicle release. We found that retromer is present at the mammalian presynaptic terminal. However, acute retromer depletion did not affect any of the measured structural and functional parameters. The fact that retromer is present the presynaptic terminals could be further investigated to elucidate the role of retromer in recycling presynaptic membrane proteins such as neurotransmitter transporters or GPCRs.

MATERIALS AND METHODS

Plasmids

The target sequences of the shRNAs were as follows: CGT GTG GAC TAC GTC GAT AAA (shVPS35-1), CCA AAT CTT GAG TCC AGT GAA (shVPS35-2), GCT GTC ACC AAA GAG TTA CTA (shVPS35-3), TTC TCC GAA CGT GTC ACG T (shControl, scramble (Zhang et al., 2008)). The target sequences were cloned in to a lentiviral expression vector under the U6 promotor containing mCherry under Synapsin promotor, which was used as a reporter of the lentiviral infection.

To report synaptic vesicle release we used Synaptophysin-pHluorin under Synapsin promotor (sypHy) (Granseth et al., 2006) and Synaptobrevin-pHluorin under Synapsin promotor (synaptopHluorin) (Miesenböck et al., 1998).

Laboratory animals

Animal experiments were approved by the animal ethical committee of the VU University/ VU University Medical Centre (“Dier ethische commissie (DEC)”; license number: FGA 11-03) and, they are in accordance to institutional and Dutch governmental guidelines

45 Chapter 2 and regulations.

Primary cell culture

Mouse E18 hippocampi or cortices were dissected in Hanks balance salt solution (HBSS, Sigma) with 10mM HEPES (Life Technologies) and digested by 0.25% trypsin (20 minutes at 37 oC; Life technologies) in HBSS. The tissue disassociation was performed with fire- polished Pasteur pipettes in DMEM with FCS. The neurons were spun down and re- suspended in neurobasal medium with 2% B-27, 18 mM HEPES, 0.25% glutamax and 0.1% Pen-Strep (Life Technologies). For VPS35 protein quantification (western blot), 150,000/mL cortical neurons were plated in coated plates, and for immunocytochemistry 50,000/mL hippocampal neurons on coated coverslips with poly-L-ornithine (PLO, Sigma) and laminin (Sigma). For morphological characterization, 1,300/mL hippocampal neurons were plated on astrocyte micro-islands (Wierda et al., 2007). For electron microscopy and life cell imaging 25,000/mL hippocampal neurons were plated in a monolayer of o astrocytes. Neurons were maintained at 37 C and 5% CO2 until the day of the experiment.

Western blot

Neurons at DIV14-15 were washed with ice-cold phosphate-buffered saline (PBS), scraped and lysed in loading buffer. Samples (300.000 neurons each) were boiled for 10 minutes at 90 oC, run in SDS-PAGE (10% 1 mm acrylamide gel with 2, 2, 2-Trichloroethanol) and, transferred into Polyvinylideenfluoride (PVDF) membranes (Bio-rad) (1 hour, 0.3 mA, 4C). Membranes were blocked and incubated with primary antibodies (2 hours, room temperature) in PBS-T and 5%milk (VPS35 1:500, Abcam, Cat. No. ab10099), (actin 1:10000, Chemicon, Cat. No. MAB1501) incubated with secondary alkaline phosphatase conjugated antibodies (1:10000, Sigma) in PBS-T and 5% milk (1 hour, room temperature), incubated 5 minutes with AttoPhos (Promega) and, scanned with a FLA-5000 fluorescent image analyzer (Fujifilm).

Immunocytochemistry and Confocal Imaging

Neurons at DIV 14-15 were fixed in 4% paraformaldehyde in PBS, permeabilized with 0.5% Triton X-100 and, blocked with 2% normal goat serum and 0.1% Triton X-100 in PBS. The primary antibodies used were MAP2 (1:20000, Abcam, Cat. No. ab5392), SMI- 312 (1:5000, Abcam, Cat. No. ab24574), VAMP2/Synaptobrevin-2 (1:1000, Synaptic Systems, Cat. No. 104 211), Bassoon (1:500, Enzo Life Science, Cat. No. SAP7F407), Synaptophysin-1 (1:1000, SynapticSystems, Cat. No. 1011004), VPS35 (1:500, Abcam, Cat. No. ab97545), GluA1 (1:50, Merk Millipore, Cat. No. MAB2263). The secondary

46 VPS35 in presynaptic terminals antibodies were conjugated to Alexa dyes (1:1000, Molecular Probes). The cells were mounted on microscope slides with Dabco-Mowiol (Invitrogen). Image acquisition was performed on a Carl Zeiss LSM510 confocal microscope, with a Plan-Neofluar 40 x/1.3 oil objective. Colocalization analysis was performed using JACoB plugin in zoomed neurites (Bolte and Cordelieres, 2006). Morphological analysis was performed using SynD (Schmitz et al., 2011).

Electron microscopy

For pre-embedding immunolabelling of VPS35, whole wild-type mouse brains were immersion-fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH7.4), cryo-protected in 30% sucrose and frozen at −80°C. Endogenous peroxidase in 40 μm 2 cryosections was quenched by 0.3% H2O2 and 10% methanol in PBS. The sections were treated with 1 freeze–thaw cycle and blocked with 0.1% BSA in PBS. The primary anti-VPS35 antibody (1:250, Abcam, Cat. No. ab10099) labelled free-floating sections for 1h at room temperature and was detected by a biotinylated rabbit anti-goat antibody (Jackson ImmunoResearch, Cat. No. 305065003), avidin–biotin horseradish peroxidase complex formation (VECTASTAIN ABC kit; Vector Laboratories, Burlingame, CA), and 3′-3′-diaminobenzidine (DAB) precipitation (DAB Substrate Kit, Vector Laboratories). As a negative control, primary antibody was preincubated with blocking peptide (Abcam, Cat. No. ab23181 at a ratio of 5:1) for 30 minutes at room temperature with agitation. The sections were contrasted by 1% osmium tetroxide and 1.5% potassium ferricyanide, dehydrated though increasing ethanol concentrations (30%, 50%,70%,90%,96%,100%), and embedded in epoxy resin. Hippocampal regions were cut into 80 nm sections for transmission electron microscopy (TEM) analysis in a JEOL1010 electron microscope (JEOL, Tokyo, Japan). Digital images of regions with immunoreactivity were acquired by a side-mounted CCD camera (Morada; Olympus Soft Imaging Solutions, Münster, Germany) and iTEM analysis software (Olympus Soft Imaging Solutions). Two independent researchers counted the presence of immunoreactivity in the presynaptic or postsynaptic side to calculate the percentages of Figure 1d.

For ultrastructural characterization of VPS35 KDs, neurons at DIV14-15 were fixed and flat embedded. Cells were fixed for 1 hour with 2.5% glutaraldehyde (GA, Merck) in 0.1 M cacodylate buffer, pH 7.4, after cell were wash and stained 1 hour at room temperature with 1% OsO4/1% KRu(CN)6 in milliQ water. Then cells were embedded in epoxy resin and sectioned as described above. Cells were stained using in uranyl acetate and lead citrate in Ultra stainer LEICA EM AC20. Images were acquired at 60.000x magnification using the TEM set-up described above.

47 Chapter 2

For immuno-gold TEM, hippocampi of 2 months mice were fixed in 4% PFA with 0.1% GA in 0.1M PB and embedded in increasing concentrations of gelatin at 37°C. The hippocampi were infiltrated in 2.3 M sucrose at 4°C and frozen in liquid nitrogen. Seventy nm thick sections were obtained with a cryo-ultramicrotome (UC6, Leica), collected at −120°C in 1% methyl-cellulose in 1.2 M sucrose and transferred onto formvar/carbon- coated copper mesh grids. The sections were washed with PBS at 37°C treated with 0.1% glycine, and immunolabelled. VPS35 (1:200, Abcam, Cat. No. ab97545) was diluted in PBS with 0.1% BSA and VPS35 (1:200, Abcam, Cat. No. ab10099) was diluted in PBS with 0.1% of BSA and 0.1% cold water fish gelatin and detected by a rabbit anti- goat antibody (1:200, Jackson ImmunoResearch, Cat. No. 305065003). The antibodies were detected with Protein A-10 nm gold (CMC, UMC Utrecht, Netherlands). The negative controls were processed in parallel without primary antibody. The sections were counterstained with 0.4% uranyl acetate in 1.8% methyl-cellulose on ice and imaged on a Tecnai 12 Biotwin transmission electron microscope (FEI Company).

Live cell Imaging

Neurons at DIV 14-15 were placed in the imaging chamber containing Tyrode’s solution

(2 mM CaCl2, 2.5 mM KCl, 119 mM NaCl, 2 mM MgCl2, 30 mM glucose, 25 mM HEPES, 50μM AP5 and 10 μM DNQX at pH 7.4). The experiment was performed at room temperature with perfusion of 1 ml per minute of Tyrodes buffer. Images were acquired with the Axiovert II microscope (Zeiss, Oberkochen, Germany) with a 40x oil objective (NA 1.3). The filters were 488 ± 5 nm (emission) and 525±25 nm (excitation) for pHluorin, and 514±5 nm (emission) and 625±27,5 nm (excitation) for mCherry as shRNA reporter. The imaging protocols included 30 first seconds of base line recording, one or two identical stimulation (2,5 seconds at 40 Hz and 30 mA) followed by one minute of recovery time and a final 10 seconds perfusion of NH4 (2 mM CaCl2, 2.5 mM KCl, 119 mM NaCl, 2 mM

MgCl2, 30 mM glucose, 25 mM HEPES, 50 mN NH4Cl at pH 7.4). As specified in the result section also a final 10 seconds acid perfusion during was applied (2 mM CaCl2, 2.5 mM

KCl, 119 mM NaCl, 2 mM MgCl2, 30 mM glucose, 25 mM MES at pH 5.5). Fluorescence puncta during NH4 exposure (synaptic locations) were analyzed as regions of interest of 4 by 4 pixels’ radium (ROIs). Fluorescence during depolarization of neurons was normalized to baseline and the maximum fluorescence during NH4Cl perfusion. The results for each ROI were averaged for each field of view and presented as data points. Fields of view were excluded if a technical problem was detected that could disturb the results.

Statistical Analysis

Data are expressed as mean values ± standard error of the mean (SEM). The Shapiro-Wilk

48 VPS35 in presynaptic terminals normality test was used to evaluate the distribution of the data. Bartlett’s test was used to test homoscedasticity. In case data were normally distributed and homoscedastic, data were compared by one-way analysis of variance (ANOVA). Dunnets post-hoc tests were performed after a significant effect was detected by comparing the different knock down groups to the control. When data were not normality distributed and homoscedastic, the Kruskal-Wallis test was used with Dunn’s multiple test as post-hoc. When P-values were lower than 0.05, significance was noted in the figure as: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

Data availability 2 The datasets generated and analyzed during the current study are available from the corresponding author on request.

AUTHOR CONTRIBUTIONS

S.V.S. performed experiments and analyzed the data. S.B. collected and analyzed confocal images for the morphological characterization. M.P.D. collected and analyzed electron microscopy data. L.vK. collected the confocal images of developing neuronal cultures. S.V.S and J.R.T.vW designed the experiments and, wrote the manuscript.

ACKNOWLEDGMENTS

The authors thank Prof. dr. Matthijs Verhage for his suggestions and critical reading of the manuscript, Joke Wortel for housing and breeding the mice, Frank den Oudsten and Desiree Schut for providing cell cultures, and Robbert Zalm for cloning and lentiviral production. EM analysis was performed at the VU/VUmc EM facility (ZonMW 91111009). This work was supported by the EC under FP7-PEOPLE-2013 (607508).

COMPETING INTERESTS STATEMENT

The authors declare no competing financial interests.

49 Chapter 2

SUPPLEMENTARY FIGURES

a’ a’’

b’ b’’

Supplementary Figure S1: Uncropped electron micrographs of the presynaptic localization of VPS35. (a’ and a’’) Corresponding images from Figure 1 e’ and e’’ using the rabbit antibody against VPS35. N=2 animals. (b’ and b’’) Corresponding images from Figure 1 f’ and f’’ using the goat antibody against VPS35. N=2 animals. Scale bar=200 nm.

a b c

50 VPS35 in presynaptic terminals

Figure S2: Electron micrographs of the negative controls for immunoelectron microscopy against VPS35. (a) Negative control processed in parallel with the immunolabelling with VPS35 Cat. No. ab10099, but preincubating the primary antibody with the blocking peptide Cat. No. ab23181 at a ratio 5:1, Figure 1c-d. Scale bar=250nm. (b and c) Negative control processed in parallel with the immuno-gold labelling with VPS35, but without adding primary antibody, just bridging antibody and Protein A-gold, Figure 1e-f. Scale bar=200nm.

a b

T

PSD T 2 PSD

Supplementary Figure S3: Electron micrographs of the postsynaptic localization of VPS35. (a) Using the rabbit antibody against VPS35. N=2 animals. (b) Using the goat antibody against VPS35. N=2 animals. Scale bar=200 nm. ‘PSD’ indicates postsynaptic density and ‘T’ the presynaptic terminal.

a b

Brain lysateControl shVPS35-1shVPS35-2shVPS35-3 Brain lysate Control shVPS35-1 shVPS35-2 shVPS35-3

Supplementary Figure S4: Original uncropped blots for (a) VPS35 and (b) actin of the data shown in Figure 2b. The brain lysate condition serves as a technical control and is left out in Figure 2.

51 Chapter 2

a Merge MAP2 SMI-312 MAP2 SMI-312 DIV4 DIV14

b c 4 **** 3 **

(mm) 3 2

2

Axonal length (mm) 1

Dendritic length Dendritic 1

0 0

DIV4 DIV4 DIV14 DIV14

Supplementary Figure S5. Increase in dendritic and axonal length during development is detected using SynD. (a) Representative confocal microscopy images of hippocampal autaptic wild-type neurons at DIV4 and DIV14 stained with MAP2 and SMI-312. Scale bar=50 μm (b) Quantification of the dendritic length (n=20 neurons, N=2 animals). (c) Quantification of the axonal length (n=20 neurons, N=2 animals). Detailed information (average, SEM, n and statistics) is available in in Table S1.

52 VPS35 in presynaptic terminals

a NH4Cl pH = 5.5 0.08 1.0 0.06

0.04

0.02 0.5 0.00 30 35 40 F/Fmax Stimulus Stimulus ∆

0.0 50 100 150 Time (seconds)

Control shVPS35-3 -0.5 Control without calcium shVPS35-3 without calcium 2 b c ns ns 0.15 3 **** **** F/Fmax) ∆

0.10 F/F0) 2 ∆

0.05 1 Fmax ( 0.00 0 peak amplitude ( amplitude peak st 1 Control Control

shVPS35-3 shVPS35-3

Control without calcium Control without calcium shVPS35-3 without calcium shVPS35-3 without calcium

Supplementary Figure S6: SynaptopHluorin reports the absence of calcium in the extracellular medium. (a) Time course of synaptopHluorin fluorescence during the imaging protocol, plotted as Δ F/Fmax. The grey boxes indicate the electrical stimulation (100 AP, 40 Hz, 30 mV each one), the black box the duration of the exposure to NH4Cl and the white box the duration of the exposure to pH=5.5. (n=21±3 fields of view, N=2 animals). (b) Maximum response amplitude during the electrical stimulation plotted as ΔF/Fmax. (c) Maximum synaptopHluorin levels during exposure to NH4Cl. Detailed information (average, SEM, n and statistics) is available in in Table S1

53 Chapter 2

Supplementary Table S1: Summery of the mean, SEM, n/N numbers and statistic reports of all measured variables in the study. Independent field of view (n), independent animal (N), not applicable (empty cells).

Figure Meassured variable Group Mean ± SEM n/N Statistics p-value VAMP2 in VPS35 0.22 ± 0.01 78/3 1b Mander's coeficient VPS35 in VAMP2 0.35 ± 0.02 78/3 Control 1 ± 0.049 40/3 Relative VPS35 levels shVPS35-1 0.18 ± ˂ 0.01 40/3 H = 126.5, <0.0001 2c (ICC) shVPS35-2 0.68 ± 0.03 41/3 p<0.0001 0.0524 shVPS35-3 0.19 ± 0.01 42/3 <0.0001 Control 1 ± 0.14 7 Relative VPS35 levels shVPS35-1 0.25 ± 0.04 7 H = 16.1, 0.0053 2d (WB) shVPS35-2 0.58 ± 0.11 7 p=0.0011 0.4866 shVPS35-3 0.20 ± 0.06 5 0.0014 Control 1 ± 0.34 35/3 Relative GluA1 levels shVPS35-1 0.64 ± 0.15 35/3 H = 33.06, < 0.0001 2f (ICC) shVPS35-2 0.72 ± 0.31 36/3 p<0.0001 0.0029 shVPS35-3 0.58 ± 0.28 35/3 < 0.0001 Control 2122 ± 120 56/5 shVPS35-1 1750 ± 89 57/5 H = 7.61, 3c Dendritic length (µm) shVPS35-2 1747 ± 92 64/5 p=0.0547 shVPS35-3 1775 ± 139 48/5 Control 2952 ± 328 26/3 shVPS35-1 3142 ± 335 27/3 H = 0.23, 3e Axonal length (µm) shVPS35-2 3200 ± 330 33/3 p=0.9709 shVPS35-3 3022 ± 394 18/3 Control 0.19 ± < 0.01 41/3 Synpases/µm shVPS35-1 0.20 ± 0.01 29/3 H = 19.35, >0.9999 3f (VAMP2) shVPS35-2 0.15 ± < 0.01 31/3 p=0.0002 0.0011 shVPS35-3 0.16 ± < 0.01 28/3 0.0127 Control 74.06 ± 5.16 41/3 shVPS35-1 77.09 ± 6.59 29/3 H = 2.60, 3g VAMP2 (a.u.) shVPS35-2 69.47 ± 5.25 31/3 p=0.4562 shVPS35-3 80.64 ± 5.40 28/3 Control 82.58 ± 6.34 41/3 shVPS35-1 76.19 ± 8.24 29/3 H = 2.30, 3h Synaptophysin-1 (a.u) shVPS35-2 77.97 ± 6.57 31/3 p=0.5113 shVPS35-3 90.52 ± 6.63 28/3 Control 0.29 ± 0.01 18/2 Synpases/µm shVPS35-1 0.25 ± 0.01 18/2 H = 8.02, 3i (bassoon) shVPS35-2 0.27 ± 0.01 19/2 p=0.0454 shVPS35-3 0.29 ± 0.01 18/2 Control 0.54 ± 0.02 154/3 Active zone length shVPS35-1 0.49 ± 0.02 161/3 H = 7.81, 4b (µm) shVPS35-2 0.48 ± 0.01 163/3 p=0.0501 shVPS35-3 0.47 ± 0.01 160/3

54 VPS35 in presynaptic terminals

Control 90.14 ± 5.51 159/3 # Synaptic shVPS35-1 99.13 ± 7.03 164/3 H = 6.06, 4c vesicles/synapse shVPS35-2 88.10 ± 4.70 165/3 p=0.1085 shVPS35-3 105.40 ± 6.02 162/3 Control 4.91 ± 0.22 159/3 # docked synaptic shVPS35-1 4.95 ± 0.21 164/3 H = 0.03, 4d vesicles/synapse shVPS35-2 4.91 ± 0.20 165/3 p=9980 shVPS35-3 4.76 ± 0.17 162/3 Control 0.14 ± < 0.01 31/4 1st peak amplitud H =6.74, p = 5c shVPS35-2 0.19 ± 0.01 19/4 0.0211 (∆F/Fmax) 0.0343 shVPS35-3 0.16 ± 0.01 14/4 > 0.9999 Control 93.89 ± 1.09 31/4 H =0.05, p = 5d % Active synapses shVPS35-2 91.11 ± 2.55 19/4 0.9751 2 shVPS35-3 91.77 ± 2.52 14/4 Control 1584 ± 28 31/4 H =13.33, p = 5e Fmax (a.u.) shVPS35-2 1442 ± 28 19/4 0.0014 0.0013 shVPS35-3 1469 ± 29 14/4 0.0277 Control 0.15 ± <0.01 26/3 1st peak amplitud shVPS35-1 0.13 ± < 0.01 22/3 F(3,92) =1.5227, 6c (∆F/Fmax) shVPS35-2 0.13 ± <0.01 23/3 p=0.2127 shVPS35-3 0.13 ± <0.01 25/3 Control 89.03 ± 2.08 26/3 shVPS35-1 70.50 ± 4.58 22/3 H = 13.32, 0.0012 6d % Active synapses shVPS35-2 81.71 ± 3.62 23/3 p=0.0040 0.3297 shVPS35-3 87.11 ± 1.80 25/3 > 0.9999 Control 5726 ± 487 26/3 shVPS35-1 3711 ± 239 22/3 H = 14.25, 0.0034 6e Fmax (a.u.) shVPS35-2 4206 ± 258 23/3 p<0.0026 0.1957 shVPS35-3 5490 ± 491 25/3 > 0.9999 Control 2841 ± 117 26/3 shVPS35-1 2251 ± 132 22/3 H = 8.81, 0.0226 6f F Baseline (a.u) shVPS35-2 2355 ± 66 23/3 p<0.0318 0.4927 shVPS35-3 2863 ± 198 25/3 > 0.9999 Control 1.21 ± 0.03 26/3 Ratio peak amplitud shVPS35-1 1.08 ± 0.05 22/3 H = 2.26, 6g (2nd/1st) shVPS35-2 1.21 ± 0.03 23/3 p<0.4532 shVPS35-3 1.15 ± 0.04 25/3 Control 1881 ± 214 26/3 shVPS35-1 1621 ± 150 22/3 H = 6.56, 6h F pH = 5.5 (a.u.) shVPS35-2 1883 ± 173 23/3 p<0.0870 shVPS35-3 1787 ± 120 25/3 DIV4 0.40 ± 0.05 20/2 U = 36.00, S4b Dendritic length (mm) DIV14 1.55 ± 0.22 20/2 p<0.001 DIV4 0.66 ± 0.12 20/2 U = 85.00, S4c Axonal length (mm) DIV14 1.27 ± 0.18 20/2 p=0.0082

55 Chapter 2

Control 0.06 ±< 0.01 21/2 H = 46.45, control without p<0.0001 0.02 ±< 0.01 20/2 < 0.0001 1st peak amplitud (shVPS35-3 vs S5b calcium (∆F/Fmax) shVPS35-3 0.06 ±< 0.01 24/2 shVPS35-3 > 0.9999 shVPS35-3 without without calcium 0.02 ±< 0.01 18/2 calcium p< 0.0001) Control 1.04 ± 0.15 21/2 control without 0.83 ± 0.08 20/2 H = 0.89, S5c Fmax (a.u.) calcium shVPS35-3 0.99 ± 0.12 24/2 p=0.8256 shVPS35-3 without 0.92 ± 0.12 18/2 calcium

56 Chapter 3

Sorting nexin 4 is an endosomal sorting protein located to synapses

57 Chapter 3

Sorting nexin 4 is an endosomal sorting protein located to synapses

Sonia Vazquez-Sanchez1, Miguel A. Gonzalez-Lozano2, Marien P. Dekker3, Marieke Meijer3, Rozemarijn Jongeneel1, Alexarae Walfenzao1, Ka Wan Li2, and Jan R.T. van Weering 3, *

1Department of Functional Genomics, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, Vrije Universiteit (VU), Amsterdam, Netherlands

2Department of Molecular and Cellular Neurobiology, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, VU University, Amsterdam, The Netherlands

3Clinical Genetics, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, VU medical center, Amsterdam, Netherlands

*Corresponding author: Jan R.T. van Weering, Center for Neurogenomics and Cognitive Research, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. Email: [email protected]

58 SNX4 in presynaptic terminals

ABSTRACT

Sorting nexin 4 (SNX4) is an evolutionary conserved protein that mediates recycling from the endosomes back to the plasma membrane in yeast and mammalian cells. Local recycling is critical for synapse function in the brain, and while SNX4 has been detected in the brain, its neuronal localization and function have not been addressed. Using a new antibody, we found that endogenous SNX4 partially co-localized with both early and recycling endosomes in primary neurons, similar to the reported localization of SNX4 in non-neuronal cells. Moreover, SNX4 was accumulated in synapses and immuno- electron microscopy revealed that SNX4 was predominantly localized to presynaptic terminals. Using three different shRNAs, SNX4 depletion drastically impaired synaptic vesicle release. However, this phenotype was not restored by expressing a SNX4 variant resistant to shRNAs. SNX4 depletion dysregulated the neuronal proteome but not presynaptic ultrastructure or neuronal morphology. Mass spectrometry analysis revealed that synaptic communication-related proteins were downregulated upon SNX4 depletion. The identification of SNX4 as a novel presynaptic protein indicates a selective demand for endosomal recycling in presynaptic boutons that might regulate the local proteome.

3

59 Chapter 3

INTRODUCTION

Sorting nexin 4 (SNX4) is an evolutionary conserved protein that mediates endosomal recycling from endosomes back to the plasma membrane (Koumandou et al., 2011; van Weering et al., 2010). SNX4 is a member of the sorting nexin family (SNX) characterized by a phosphatidylinositol 3-phosphate binding domain (phosphoinositide-binding phox homology (PX) domain) (Carlton et al., 2005), which is necessary for peripheral membrane localization (Cullen, 2008; Teasdale et al., 2001). More specifically, SNX4 is part of the SNX-BAR subfamily characterized by having a carboxy-terminal Bin/Amphiphysin/ Rvs (BAR) domain, which binds to curved membranes upon dimerization (Carlton et al., 2004; Cullen, 2008). SNX4 forms tubules that emanate from the endosomes during the Rab5-Rab7 transition (early endosome to late endosome) and during Rab4-Rab11 transition (early recycling endosome to endosome recycling compartment)(van Weering et al., 2012b). Hettema et al. (2003) showed that silencing the yeast homologue of SNX4 (Snx4p) decreases Scn1p (an exocytic v-SNARE) in the plasma membrane and increases Scn1p degradation at the vacuole (the lysosome equivalent in yeast) (Hettema et al., 2003). In HeLa cells, a similar SNX4 pathway has been observed: SNX4 recycles back to the plasma membrane the transferrin receptor (TfnR), an iron-transporting receptor located to the plasma membrane, avoiding lysosomal degradation (Traer et al., 2007).

SNX4 is expressed in the brain (Kim et al., 2017). SNX4 protein levels are 70% decreased in Alzheimer’s disease brains in the highest Braak stages (Kim et al., 2017). Two recent studies propose that SNX4 dysregulation leads to a mis-sorting of beta-secretase 1 (BACE1) in Alzheimer’s disease (Kim et al., 2017; Toh et al., 2018). BACE1 is an enzyme involved in proteolytic processing of the amyloid precursor protein (APP), which leads to the formation of the pathological Aβ peptide. These recent studies show that SNX4 recycles BACE1 from the early endosome to the recycling endosome, thus preventing its degradation (Kim et al., 2017; Toh et al., 2018). When SNX4 was depleted, BACE1 was directed to the late endosome and Aβ levels were increased (Toh et al., 2018). Mis-sorting of BACE1 and increased Aβ production due to SNX4 dysregulation might be a process involved in Alzheimer’s disease etiology.

While SNX4 function is associated with pathological mechanisms in the brain, the physiological role and subcellular distribution of SNX4 in neurons remains unclear. First, we characterized the localization of endogenous SNX4 in primary mouse neurons using a new antibody. Endogenous SNX4 partially co-localized with both early and recycling endosome markers, which is in accordance with the previously established role of SNX4 in non-neuronal cells. Neuronal SNX4 accumulated specifically in synaptic areas and with a predominant localization to presynaptic terminals, suggesting that SNX4 fulfills

60 SNX4 in presynaptic terminals a specific role in this compartment. We addressed the impact of knocking down SNX4 on presynaptic ultrastructure, protein composition and synaptic vesicle fusion and endocytosis in primary mouse neurons.

RESULTS

SNX4 is expressed in the brain and in neurons

In order to characterize the localization of endogenous SNX4 in mouse neurons, we have developed a novel antibody. Commercially available antibodies against SNX4 only detected mouse SNX4 by western blot. This novel antibody was designed against the N-terminal region of mouse SNX4 in collaboration with Synaptic Systems (Cat. No. 392 003) (Supplementary Figure S1). This novel antibody detected a protein of ~50 kDa, which corresponds with the size of SNX4. Different brain regions were studied with this novel antibody to gain resolution on SNX4 distribution in the brain. The ~50 kDa SNX4 signal appeared in all the studied brain regions (Figure 1a, b), suggesting that SNX4 is ubiquitously expressed in the brain. The novel antibody was tested for immunocytochemistry in mouse neurons cultured on a feeding layer of rat astrocytes. Autaptic hippocampal neurons of 15 days in-vitro (DIV15) were stained with a dendritic marker (MAP2), a synaptic marker (Bassoon) and SNX4 antibodies. The novel SNX4 antibody showed signal both in the feeding layer of rat astrocytes and in the mouse primary neuron (Figure 1c). These data 3 a b 1.5

1.0

SNX4/actin 0.5 kDa cerebelum cortex hippocampusprefrontal cortexhypothalamusstriatum olfactory bulb

55 SNX4 0.0 40 Actin cortex striatum cerebelum hippocampushypothalamusolfactory bulb c prefrontal cortex MAP2 bassoon SNX4 Merge

Figure 1: SNX4 is expressed in the brain and in neurons. (a) Western blot of different mouse brain areas for SNX4 and actin. Original uncropped blots are shown in Supplementary Figure S9. (b) Quantification of SNX4 levels normalized to actin in western blot. (N=3±1 blots/animals). (c) Confocal microscopy of a hippocampal neuron on an astrocyte island immunolabelled with MAP2 (blue), Bassoon (green) and SNX4 (magenta). Scale bar of the neuron image=20 μm, scale bar of the zoomed neurites=5 μm. Representative image of n=25 neurons, N=3 animals.

61 Chapter 3 indicate that the novel antibody against SNX4 recognizes endogenous mouse and rat SNX4 both by western blot and immunocytochemistry. In addition, SNX4 seems to be ubiquitously expressed among different brain regions and cell types.

a c 3 DIV0 DIV3 DIV7 DIV14-15 Plate neurons SNX4 shRNAs Measurement 2

1

b SNX4/actin

0 kDa Control shSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 + R1 shSNX4-3+ R2 + R3

SNX4 55 Control shSNX4-1shSNX4-2shSNX4-3 40 Actin shSNX4-1shSNX4-2 + R1shSNX4-3 + R2 + R3 d f sh reporter Synaptophysin-1 SNX4 Merge 2.5 **** 2.0 Control 1.5 1.0

Synapses per µm Synapses 0.5 0.0

sh1 Control shSNX4-1shSNX4-2shSNX4-3

g 3 shSNX4-1shSNX4-2 +shSNX4-3 R1 + R2 + R3

2

sh1R1 1

0 Synaptophysin-1 (a.u) Synaptophysin-1

Control shSNX4-1shSNX4-2shSNX4-3

shSNX4-1shSNX4-2 +shSNX4-3 R1 + R2 + R3 sh2 h 3 **** **** 2 ****

SNX4 (a.u) 1

sh2R2 0

Control shSNX4-1shSNX4-2shSNX4-3

shSNX4-1shSNX4-2 +shSNX4-3 R1 + R2 + R3 i 8

sh3 6

4 **** **** *** 2

0

sh3R3 in Syph-1 puncta (a.u) SNX4 Control shSNX4-1shSNX4-2shSNX4-3

shSNX4-1shSNX4-2 +shSNX4-3 R1 + R2 + R3

62 SNX4 in presynaptic terminals

Novel SNX4 antibody specifically labels endogenous mouse SNX4 on western blot and immunocytochemistry.

To confirm that the novel antibody specifically detects SNX4, we developed three independent shRNAs against SNX4, and rescue constructs (Supplementary Figure S2). Cortical mouse neurons were lentiviral infected at DIV3 with the rescue SNX4 constructs (R1, R2 and R3), and at DIV7 with the three shRNA against SNX4 (shSNX4-1, shSNX4-2, and shSNX4-3) and the shRNA control (Control). At DIV14-15 the neurons were lysed or fixed, and SNX4 levels were evaluated using western blot and immunocytochemistry (Figure 2a). Using western blot, the band that appeared at ~50 kDa was decreased when using the three independent shRNAs against SNX4 and these levels were restored when the shRNA against SNX4 was combined with the SNX4 rescue constructs (Figure 2b, c). The same ~50 kDa band was observed using two commercially available antibodies (Supplementary Figure S1). However, a lower band of ~30 kDa (which does not correspond with SNX4 size) was also present in all samples. This lower band did not decrease in the neurons expressing shRNA against SNX4, suggesting that the antibody also recognizes another protein (Supplementary Figure S9).

For immunocytochemistry, mCherry was used as a transfection reporter of all shRNA constructs (both shRNA control and against SNX4). DIV15 cortical neurons were stained with SNX4 and sypnaptophysin-1 antibodies (synaptic marker) (Figure 2d). Both synaptic 3 density (number of synapses per µm, Figure 2e) and the total synaptophsyin-1 intensity (Figure 2f) were not changed upon modulation of SNX4 levels. However, the total SNX4 intensity was decreased upon SNX4 knock down and restored when the shRNA was combined with the rescue constructs (Figure 2g). The same decrease and rescue was observed for the SNX4 intensity in synapses (Figure 2h), showing that SNX4 is present in synaptic locations. Together, these data confirm that SNX4 is detected using the novel antibody both in western blot and immunocytochemistry and that cellular SNX4 levels can

Figure 2: Novel SNX4 antibody specifically labels endogenous mouse SNX4 on western blots and immunocytochemistry. (a) Experimental design timeline. (b) Representative SNX4 and actin western blot of control neurons, neurons transfected with shRNAs against SNX4 and neurons transfected with the shRNA and its rescue construct (description of the constructs in Supplementary Figure S2). Original uncropped blots are shown in Supplementary Figure S9. (c) Quantification of SNX4 levels normalized to actin in western blot. Values are presented as a ratio compared to the control condition. (N=3 blots/animals). (d) Confocal microscopy images of neurons infected with control shRNA, the three shRNAs against SNX4 and its respective rescue constructs. Left, mCherry signal reporting the transfection of the shRNAs coding sequences. Middle, Synaptophysin-1 labelling. Right, SNX4 labelling. (n=50±13 fields of view, N=4±1 animals). Scale bar=20 μm. (e) Quantification of synaptic density relative to control labelled as Synaptophysin-1 puncta. (f) Quantification of Synaptophysin-1 staining intensity relative to control. (g) Quantification of SNX4 staining intensity relative to control. (h) Quantification of SNX4 staining intensity in Synaptophysin-1 puncta relative to control. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

63 Chapter 3 be modulated by the three independent shRNA against SNX4 and its rescue constructs.

a b c Rab5 SNX4 Merge 2.0 ** 2.0 ****

1.5 1.5 lortnoC

1.0 1.0

0.5 0.5 Total Rab5 (a.u.) Total SNX4 (a.u.) 0.0 0.0

Control Control d shSNX4-2 e shSNX4-2 Rab5 in SNX4 1.0 **** 1.0 SNX4 in Rab5 **** **** 0.8 0.8 0.6

’s coeficient 0.6

shSNX4-2 0.4 0.4 0.2 0.2 Pearson Mander's Coefficient Mander's 0.0 0.0

Control Control Control f Rab11 SNX4 Merge g shSNX4-2 h shSNX4-2shSNX4-2 2.0 2.0

1.5 1.5 **** lortnoC

1.0 1.0

0.5 0.5 Total SNX4 (a.u.) Total Rab11(a.u.) 0.0 0.0

Control Control i shSNX4-2 j shSNX4-2 Rab11 in SNX4 0.8 **** 1.0 SNX4 in Rab11 **** * 0.6 0.8

shSNX4-2 0.6 0.4 0.4 0.2 0.2 Pearson coeficient Pearson Mander's Coefficient Mander's 0.0 0.0

Control Control Control shSNX4-2shSNX4-2 shSNX4-2 k l 1.5 m 1.5 * kDa Control shSNX4-1 shSNX4-2 shSNX4-3 TfnR 1.0 1.0 100- SNX4 55-

TfnR/total protein TfnR/total 0.5 0.5

Actin protein SNX4/total 40- 0.0 0.0

Control Control shSNX4-1 shSNX4-2 shSNX4-3 shSNX4-1 shSNX4-2 shSNX4-3

64 SNX4 in presynaptic terminals

SNX4 is located at neuronal early and recycling endosomes, but SNX4 depletion does not decrease TfnR levels in neurons

SNX4 has been found colocalizing with early and recycling endosomal markers in HeLa cells, where it coordinates recycling from early endosomes to the plasma membrane through the recycling endosomes (Traer et al., 2007). We hypothesized that the same pathway exists in neurons. To test if SNX4 also colocalizes with these endosomal makers in neurons, hippocampal mouse neurons at DIV14-15 were fixed and immmunostained for endogenous SNX4, Rab5 (early endosome marker) and Rab11 (recycling endosome marker). A validated shRNA against SNX4 (shSNX4-2) was used as control for the specific detection of SNX4. Upon SNX4 depletion, both the total neuronal levels of Rab5 and SNX4 were decreased (Figure 3a, b, c). About 58% of the Rab5 signal colocalized with SNX4 signal but this value dropped to about 41% upon SNX4 depletion (Pearson’s coefficient, Figure 3 a, d). Approximately 51% of SNX4 signal colocalized with Rab5, and approximately 64% of Rab5 colocalized with SNX4 signal (Mander’s coefficients, Figure 3a, e). Both Mander’s coefficients dropped to 29% and 54% respectively upon SNX4 knock down. Expression of shSNX4-2 decreased the levels of SNX4, but it did not affect the levels of Rab11 (Figure 3f, g, h). About 45% of the Rab11 signal colocalized with SNX4 signal but this value dropped to about 31% upon SNX4 depletion (Pearson’s coefficient, Figure 3f, i). Approximately 39% of SNX4 signal colocalized with Rab11 staining, and approximately 49% of Rab11 colocalized with SNX4 signal (Mander’s coefficients, Figure 3 3f, e). Both Mander’s coefficients dropped to 18% and 42% respectively upon SNX4 knock down. (Figure 3f, j). These data indicate that SNX4 is located to neuronal early and recycling endosomes.

Figure 3: Neuronal SNX4 is located to early and recycling endosomes but SNX4 depletion does not decrease TfnR levels in neurons. (a) Confocal microscopy images of control and SNX4 knock down neurons immunolabelled for Rab5 and SNX4. Merge image of Rab5 (green) and SNX4 (magenta). (n=21±2 neurons, N=3 animals). Scale bar of the neuron image=50 μm, scale bar of the zoomed neurite=5 μm. (b) Quantification of total Rab5 levels in the neuron normalized to control. (c) Quantification of total SNX4 levels in the neuron normalized to control. (d) Pearson’s coefficients for the co-localization of Rab5 and SNX4 in neurites. Mander’s coefficients for the co- localization of Rab5 and SNX4 in neurites. (f) Confocal microscopy images of control and SNX4 KD neurons immunolabelled with Rab11 and SNX4. Merge image of Rab11 (green) and SNX4 (magenta). (n=38±1neurons, N=3 animals). Scale bar of the neuron image=50 μm, scale bar of the zoomed neurite=5 μm. (g) Quantification of total Rab11 levels in the neuron normalized to control. (h) Quantification of total SNX4 levels in the neuron normalized to control. (i) Pearson’s coefficients for the co-localization of Rab11 and SNX4 in neurites. (j) Mander’s coefficients for the co-localization of Rab11 and SNX4 in neurites. (k) Western blot of neurons infected with control shRNA (Control), and the three shRNAs against SNX4 stained for TfnR, SNX4 and actin. Original uncropped blots are shown in Supplementary Figure S9. (l) Quantification of TfnR levels normalized to total amount of proteins (N=3±1). (m) Quantification of SNX4 levels normalized to total amount of protein in western blot (N=3±1). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

65 Chapter 3

In HeLa cells, SNX4 depletion leads to decreased levels of TfnR which can be restored by lysosomal inhibition (Traer et al., 2007). To test if SNX4 depletion also decreases TfnR levels in neurons, we measured the levels of TfnR upon SNX4 knock down using western blot (Figure 3k, Supplementary Figure S3). Upon shSNX4 expression, TfnR levels were not changed while SNX4 levels were decrease upon shSNX4-2 expression and showed

a Synaptophysin-1 SNX4 Merge e

SyMSySMP2 PSD lortnoC

f SyMSySMP2 PSD kDa

shSNX4-2 PSD95 100 55 SNX4 15 VAMP2

Syph-1 in SNX4 b c SNX4 in Syph-1 d g 25 1.0 **** 1.0 **** ** 2.0 **** 20 0.8 en t 0.8 1.5 15 10 0.6 oe ff ici 0.6 PSD95/TCE 1.0 5 0.4 0.4 0 Total SNX4

ea rs on’s coefficient P2 M r's c ande r's 0.5 SyS P 0.2 0.2 SyM PSD M 2.0 0.0 0.0 0.0 h 1.5

Control Control Control Control 1.0

shSNX4-2 shSNX4-2 shSNX4-2 shSNX4-2 VAMP2/TCE 0.5 j’ 0.0

P2 M SyS SyM PSD 5 i 4 T 3 2 SNX4/TCE 1 0

P2 M PSD SyS SyM PSD

j’’ k 7 **** 6 5 pa rticles PSD 4 T go ld 3 T # 2 PSD 1 T PSD

66 SNX4 in presynaptic terminals a strong trend towards reduction upon shSNX4-1 and shSNX4-3 (Figure 3l, m). SNX4 depletion in HeLa cells leads to abnormal Rab11 distribution (from juxtanuclear to peripherical localization)(Traer et al., 2007). We tested if upon neuronal SNX4 depletion the distribution of Rab11 in synapses was also changed. No difference was observed in the colocalization of synaptic markers with recycling endosome markers upon SNX4 depletion, suggesting that the peripherical distribution of Rab11 is normal upon SNX4 depletion (Supplementary Figure S4).

Together these data show that that SNX4 localizes with both early and recycling endosomes in neurons but that its depletion does not decrease TfnR levels.

Synaptic SNX4 is predominantly located to presynaptic terminals

SNX4 appeared to be localized at synapses (Figure 1 and Figure 2). To confirm this, we analyzed colocalization between SNX4 and synaptic markers. Control and SNX4 knock down hippocampal neurons at DIV14-15 were fixed and immmunostained for SNX4 and Synaptophysin-1 (synaptic marker) (Figure 4a). About 71% of the Synaptophysin-1 puncta colocalized with SNX4 puncta but this value dropped to about 55% upon SNX4 depletion (Pearson’s coefficient, Figure 4b). Approximately 63% of SNX4 immunoreactivity colocalized with Synapthophysin-1, and approximately 64% of synapses colocalized with SNX4 signal (Mander’s coefficients, Figure 4c). Both Mander’s coefficients dropped 3 to 48% and 52% respectively upon SNX4 knock down. In this experiment, SNX4 knock down was of about a 32% reduction (Figure 4d). This colocalization between synaptic markers and SNX4 was confirmed using VGluT1 and SNX4 (Supplementary Table S1)

Figure 4: Synaptic SNX4 is predominantly located to presynaptic terminals. (a) Confocal microscopy images of hippocampal neurons from control and SNX4 knock down neurons immunolabelled with Synatophysin-1 and SNX4. Merge image of Synaptophysin-1 (green) and SNX4 (magenta). Scale bar of the neuron image=50 μm, scale bar of the zoomed neurite=5 μm. (b) Pearson and (c) Mander’s coefficients for the co-localization between Synatophysin-1 and SNX4 in neurites. (d) Quantification of total SNX4 levels in the neuron normalized to control. (n=36 fields ofview, N=3 animals). (e) Representative western blot of hippocampal subcellular fractions (pellet 2 (P2), microsomal fraction (M), synaptosomes (SyS), synaptic membrane fraction (SyM), and PSD fraction (PSD)) stained with SNX4, VAMP2/Synaptobrevin-2, and PSD95. Original uncropped blots are shown in Supplementary Figure S9. (f) Total protein in each hippocampal subcellular fraction. Quantification of (g) PSD95, (h) VAMP2/Synaptobrevin-2, and (i) SNX4 levels normalized to total protein. Values are presented as a ratio compared to each total hippocampus lysate. (N=3 blots/animals). (j’,j”) Immunoelectron micrographs of synaptic terminals stained with SNX4 antibody labelled with Protein A-10nm gold conjugate. The images are representative of three independent experiments (N=3 animals). Scale bar=200 nm. ‘PSD’ indicates postsynaptic side and ‘T’ the presynaptic terminal. (k) Number of gold particles in the postsynaptic side and the presynaptic terminal in each synapse. (n=46 synapses, N= 3 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

67 Chapter 3 and overexpressed GFP-SNX4 and VAMP2/Synaptobrevin-2 (Supplementary Figure S5). As an independent confirmation for the synaptic localization of SNX4, we blotted for SNX4 in hippocampal subcellular fractions (Figures 4e-i). As expected, the PSD fraction Merge b a MAP2 SMI-312 MAP2 SMI-312 15

10

5 Control

Dendritic length (mm) length Dendritic 0

Control c 15 shSNX4-1shSNX4-2shSNX4-3

10 shSNX4-1

5

Axonal length (mm) Axonal length 0

Control e shSNX4-1shSNX4-2*** shSNX4-3 shSNX4-2 0.5 0.4 0.3 0.2 0.1

0.0 shSNX4-3 Synapses per µm (bassoon) per µm Synapses

b Control shSNX4-1shSNX4-2shSNX4-3 d Homer-1 bassoon f 3 g 2.5 2.0 2

1.5 lortnoC ass oon 1.0

b Control 1 Homer-1 0.5 0 0.0

Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 h 5 i 2.5 ***

4 2.0 shSNX4-1 in -1 3 1.5

oph ys in -1 2 1.0 1 0.5 Synapt 0 Synaptotagm 0.0

shSNX4-2 Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 j k 4 2.0 **** **** 3 1.5 **** in -1 2

MP 2 1.0 lortnoC VA

1 Syntax 0.5 shSNX4-3 0 0.0

Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3

68 SNX4 in presynaptic terminals

(PSD) was highly enriched in PSD95 and depleted for VAMP2/Synaptobrevin-2. SNX4 was found in all subcellular fractions except in the PSD fraction (PSD). The presence of SNX4 in synaptosomes (SyS) indicates a synaptic localization of SNX4. The presence of SNX4 in the synaptic membrane fraction (SyM) in combination with the absence in the PSD fraction suggests a SNX4 presynaptic localization.

To investigate the distribution of SNX4 within the synapse, immuno-gold electron microscopy was performed using Protein A-gold 10 nm to detect the SNX4 antibody. SNX4 immunosignal was detected inside presynaptic terminals and in the postsynaptic side, but not in the negative controls (blocking peptide) (Figure 4j’, j’’, Supplementary Figure S6). SNX4 immunosignal was more abundant in the presynaptic terminal than in the postsynaptic side (Figure 4k). Overall, these data show that SNX4 is present in both sides of the synapse but it is more abundant in presynaptic terminals.

SNX4 depletion does not affect neuronal morphology

To define the role of SNX4 at presynaptic terminals, we first tested if SNX4 depletion leads to abnormal neuronal morphology. Hippocampal autaptic neurons were infected with lentiviral particles containing shRNA against SNX4 or control shRNA at DIV7, when neurons are forming synapses, and fixed at DIV14-15, when neurons and synapses are mature. Neurons were stained for dendritic (MAP2), and axonal (SMI-312) markers, imaged 3 using confocal microscopy and analyzed using SynD (Figure 5a). Neuronal networks were stained with synaptic markers (Homer-1, Bassoon, Synaptotagmin-1, Synaptophysin-1, Syntaxin-1 and VAMP2/Synaptobrevin-2) (Figure 5d). SNX4 knock down did not affect the length of the dendritic arbor and the length of the axon (Figure 5a, b, c). The synaptic density (Bassoon puncta per µm) was not changed in neurons expressing shSNX4-1 and shSNX4-2 compared with control, but it was reduced in shSNX4-3 expressing neurons (Figure 5e). The intensity of Bassoon, Homer-1 and Synaptophysin-1 was not changed

Figure 5: SNX4 depletion does not affect neuronal morphology.(a) Confocal microscopy images of hippocampal autaptic neurons containing control and SNX4 shRNAs immunolabelled with MAP2 and SMI-312. Merge image of the MAP2 (green) SMI-312 (magenta). Scale bar=50 μm. (n=22, N=3 animals). (b) Quantification of the dendritic length, and (c) axonal length. (d) Confocal microscopy images of hippocampal neurons containing control and SNX4 shRNAs immunolabelled with Homer-1 and Bassoon. Scale bar=40 μm. (e) Quantification of synaptic density relative to control labelled as Bassoon puncta (n=32±1 fields of view, N=3 animals). Quantification of protein intensity normalized to control of (f) Bassoon (n=32±1 fields of view, N=3 animals), (g) Homer-1 (n=24±2 fields of view, N=2 animals), (h) Synaptotagmin-1 (n=26±4 fields of view, N=2 animals), (i) Synaptophysin-1 (n=26±4 fields of view, N= animals) (j) Syntaxin-1 (n=23±2 fields of view, N=2 animals), and(k) VAMP2/Synaptobrevin-2 (n=66±4 neurons, N=5 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

69 Chapter 3 in any of the SNX4 knock down groups compared with control (Figure 5f, g, h). The intensity of Synaptotagmin-1 and VAMP2/Synaptobrevin-2 was decreased in shSNX4-1 expressing neurons but did not change in shSNX4-2 and shSNX4-3 expressing neurons compared to control (Figure 5i, j). The intensity of Syntaxin-1 was reduced in shSNX4-1 and shSNX4-3 neurons, but not changed in shSNX4-2 neurons compared with control (Figure 5k). Overall, these data show that SNX4 knock down does not alter the neuronal morphology and levels of several synaptic proteins.

SNX4 depletion does not affect presynaptic ultrastructure

The effect of SNX4 depletion in presynaptic ultrastructure was evaluated by Transmission Electron Microscopy (TEM) in aldehyde fixed hippocampal neurons at DIV14-15. The overall synaptic morphology was not affected by SNX4 depletion (Figure 6a). Control

a Control shSNX4-1 shSNX4-2 shSNX4-3

b c d 2000 **** 800 **** 0.3 *** 1500 600 0.2

1000 SV # 400 0.1

500 200 SV docked/total Active zone length (nm) zone length Active 0 0 0

Control Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 e f g 40 **** 20 15 30 15 10 20 10 # tubules #

# big vesicles big # 5 10 5

0 0 0 % of synapses of % MVB with

Control Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3

Figure 6: SNX4 depletion does not affect presynaptic ultrastructure. (a) Electron micrographs of hippocampal synapses from control and SNX4 knock down neurons. (n=157±5 synapses, N=3 animals). Scale bar=200nm. (b) Quantification of active zone length. (c) Total number of synaptic vesicles. (d) Fraction of synaptic vesicles that are docked. (e) Number of big vesicles in presynaptic terminals. (f) Number of membrane tubules in presynaptic terminals. (g) Percentage of presynaptic terminals containing multivesicular bodies (MVB). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

70 SNX4 in presynaptic terminals and shSNX4-2 and shSNX4-3 expressing neurons did not differ in the main presynaptic parameters: active zone length, the total amount of synaptic vesicles, and the number of docked synaptic vesicles (Figure 6b-d). These groups did not differ either in parameters related with the endosomal system in presynaptic terminals: the number of vesicles bigger than synaptic vesicles, the number of membrane tubular structures and the presence of multivesicular bodies (MVB) in the presynaptic terminals (Figure 6e-g). However, shSNX4-1 expressing neurons showed an increase in both active zone length and in the number of synaptic vesicles, and a decrease in both docked and large vesicles. Together, these data show that SNX4 depletion does not affect presynaptic ultrastructure.

SNX4 depletion impairs synaptic vesicle release

Next, the impact of SNX4 depletion in presynaptic function was evaluated. Fluorescent reporters of synaptic vesicle release and retrieval were expressed in hippocampal neurons containing control shRNA and the three independent shRNAs against SNX4. First, neurons were imaged using a pH-sensitive variant of GFP fused to the luminal domain of the synaptic vesicle protein VAMP2/Synaptobrevin-2 (synaptopHluorin) (Miesenböck et al., 1998). The protocol consisted in the following elements: 30 seconds of initial recording (to use as a baseline), an electrical stimulation (100 AP, 40 Hz, 30 mA, to evoke synaptic vesicle release), an identical electrical stimulation after one minute of recovery (to measure if neurons were able to efficiently release synaptic vesicle after 3 having been already electrically stimulated, which would be a read out of efficiency refilling the releasable synaptic vesicle pools), an exposure to NH4Cl after one minute of recovery (to de-quench all synaptopHluorin to quantify the total reporter pool), and a final exposure to a pH=5.5 solution (to calculate the fraction of synaptopHluorin that remained in the plasma membrane) (Figure 7a, b). The total pool of synaptopHluorin (maximum fluorescence during the exposure 4to NH Cl) was the same in the four groups (Figure 7c). SynaptopHluorin fluorescence was quenched under resting conditions in all groups (fluorescence during the baseline), but this fluorescence was lower in neurons expressing shSNX4-1 (Figure 7d). Electrical stimulation triggered synaptic vesicle release in control group but synaptic vesicle release was almost absent in the three SNX4 depletion groups (first peak amplitude ΔF/Fmax, Figure 7e). About 50% reduction in the percentage of active synapses was observed in the three groups of SNX4 depleted neurons compared to control (Figure 7f). The ratio between the fluorescence peaks after stimulation was equal between all the groups (Figure 7g). The fluorescence during the pH=5.5 wash was decreased in shSNX4-1 and shSNX4-2 but not shSNX4-3 compared to control (Figure 7h). These data suggest that SNX4 loss impairs synaptic vesicle release.

To assure that the synaptic vesicle release impairment upon SNX4 knock down is SNX4

71 Chapter 3

a b 0.15

Baseline 0.10 1.0 Control shSNX4-1 Stimulus shSNX4-2 0.05 shSNX4-3

0.5 Recovery 0 30 35 40 45 F/Fmax ∆ NH4Cl 0.0 50 100 150 pH = 5.5 Time (seconds) c d e -0.5 **** 20000 0.25 * 8000 **** 0.20 15000 6000 * 0.15 10000 4000 0.10 Fmax (a.u.) 5000 Baseline F (a.u.) 2000 0.05 0 0 0.00 1st Peak Amplitud ( ∆ F/Fmax) Amplitud Peak 1st

Control ) Control st Control f shSNX4-1shSNX4-2shSNX4-3 g shSNX4-1shSNX4-2shSNX4-3 h shSNX4-1shSNX4-2shSNX4-3 /1

nd 2.5 4000 150 **** * * 2.0 ** **** 3000 100 1.5 1.0 2000 50 0.5

pH F (a.u.)pH = 5.5 1000

active synapses (%) synapses active 0.0 0.0 -0.5 0

Ratio peak amplitude (2 amplitude peak Ratio Control Control Control i shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 j 0.20 Baseline Control 1.0 0.15 shSNX4-1 0.10 shSNX4-2 Stimulus 0.05 0.5 0.00 shSNX4-3 32 34 36 38 40 Overexpression

Recovery F/Fmax

∆ shSNX4-1 + R1 0.0 NH Cl shSNX4-2 + R2 4 40 60 80 100 shSNX4-3 + R3 Time (seconds) n.s. k l m 20000 -0.5 8000 *** 0.4 F/Fmax) ∆ 15000 6000 0.3

ax (a.u.) 10000 4000 0.2 ne (a.u.) ase li ne Fm 5000 B 2000 0.1 0.0 0 0 Peak amplitude( Peak

Control Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3

shSNX4-1shSNX4-2shSNX4-3 + R1Overexpression + R2 + R3 shSNX4-1shSNX4-2 shSNX4-3+ R1Overexpression + R2 + R3 shSNX4-1shSNX4-2 shSNX4-3+ R1Overexpression + R2 + R3

72 SNX4 in presynaptic terminals dependent, we designed a new experiment with eight groups: control group, 3 groups of neurons expressing the three shRNAs against SNX4, 3 groups of neurons expressing the shRNAs against SNX4 and its rescue constructs, and a group of neurons overexpressing SNX4. Hippocampal mouse neurons were infected at DIV3 with the rescue constructs and at DIV7 with the shRNAs. At DIV14-15, the neurons were imaged using sypHy, a pH-sensitive variant of GFP fused in the luminal domain of the synaptic vesicle protein Synaptophysin-1 (sypHy) (Granseth et al., 2006b). This reporter works as synaptopHluorin, but it has an improved signal/noise ratio because it has lower signal at cell surface, reducing the background fluorescence. In this experiment, we simplified the protocol: 30 seconds of baseline recording, a single electrical stimulation (100 AP, 40 Hz, 30 mA), one minute of recovery, and an exposure to NH4Cl. The total pool of sypHy was the same in the four groups (Figure 7k). SynaptopHluorin fluorescence was quenched equally under resting conditions in all groups (Figure 7m). Compared to control, synaptic vesicle release was almost abolished in the three SNX4 knock down groups and in the 3 rescue groups, but not in the overexpression group (Figure 7l). Together, these data show that SNX4 knock down impairs SNX4 but that this effect was not restored by the re-introduction of SNX4.

Presynaptic function was independently addressed using whole-cell patch clamp recordings in cortical control neurons and neurons expressing shSNX4-1. Both groups showed the same average peak amplitude of spontaneous excitatory postsynaptic 3 currents (mEPSCs). However, the frequency of these mEPSCs was about 70% decreased in the knock down group, suggesting a defect in synaptic vesicle release but not in synaptic vesicle loading or neurotransmitter receptor function on the postsynaptic side

Figure 7: SNX4 depletion impairs synaptic vesicle release but it cannot be restored by SNX4 re-introduction. (a) Representative synaptopHluorin fluorescence images of neurites during the baseline, the first stimulation, the first recovery period, the exposure to 4NH Cl and the exposure to pH=5.5. Scale bar=10µm. (b) Time course of synaptopHluorin fluorescence during the imaging protocol, plotted as ΔF/Fmax. The grey boxes indicate the electrical stimulation (100 AP, 40 Hz,

30 mV each), the black box the duration of the exposure to NH4Cl and the white box the duration of the exposure to pH=5.5. (n=21±3 fields of view, N=3 animals). (c) Maximum synaptopHluorin levels during exposure to NH4Cl. (d) Average fluorescence of synaptopHluorin during baseline recordings. (e) Maximum response amplitude during the first electrical stimulation plotted as ΔF/ Fmax. (f) Percentage of responsive synapses during the stimulation. (g) Ratio of the maximum synaptopHluorin fluorescence amplitude between the first and the second electrical stimulation. (h) Minimum response to the exposure to pH=5.5. (i) Representative sypHy fluorescence images of neurites during baseline, stimulation, recovery period and exposure to NH4Cl. Scale bar=10 µm. (j) Time course of sypHy fluorescence during the imaging protocol, plotted as ΔF/Fmax. The grey box indicates the electrical stimulation (100 AP, 40 Hz, 30 mV) and the black box the exposure to

10 seconds of NH4Cl (17±7 fields of view, N=2/3 animals). (k) Maximum response to the exposure to NH4Cl. (l) Average fluorescence of synaptopHluorin during baseline recordings. (m) Maximum response amplitude during the electrical stimulation plotted as ΔF/Fmax. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

73 Chapter 3

(Supplementary Figure S7).

To investigate why SNX4 depletion decreases synaptic vesicle release, we measured the intracellular calcium at synaptic terminals upon electrical stimulation using Fluo-5F a b shSNX4-1 shSNX4-2

132 313 175 90 138 113

317 Control.1 Control.2 Control.3 Control.4 Control.5

shSNX4-3.4 shSNX4-3.5 shSNX4-3.3 shSNX4-3.1 shSNX4-3.2 shSNX4-1.3 shSNX4-1.4 shSNX4-1.5 shSNX4-1.1 shSNX4-1.2 shSNX4-2.3 shSNX4-2.1 shSNX4-2.2 shSNX4-2.4 shSNX4-2.5 shSNX4-3 shSNX4-3 shSNX4-1 Control shSNX4-2 c fold change (log2) FDR p-value moderated effect-size gene label shSNX4-1 shSNX4-2 shSNX4-3 shSNX4-1 shSNX4-2 shSNX4-3 shSNX4-1 shSNX4-2 shSNX4-3 SNX4 -1,9984 -2,3015 -1,6532 4,22E-05 2,20E-04 3,92E-04 -15,4237 -10,0344 -9,3945 SCN2A1 -1,8983 -0,9966 -1,4454 3,24E-04 9,25E-03 3,05E-03 -8,4662 -4,8936 -5,999 NPTX1 -1,6701 -0,6705 -1,4834 7,40E-04 0,032 4,62E-03 -7,2094 -3,6777 -5,2989 ACSBG1 -1,6007 -0,7454 -0,4429 1,36E-04 8,44E-03 0,033 -10,466 -5,0329 -3,3286 LGI1 -1,4974 -0,513 -0,3425 9,61E-05 0,0142 0,0433 -11,3557 -4,4277 -3,101 GAD1 -1,4139 -1,5257 -1,1528 1,88E-04 4,82E-04 2,28E-03 -9,8764 -8,8025 -6,4539 SLC1A2 -1,4104 -1,7706 -1,0806 4,55E-05 2,20E-04 2,63E-04 -13,1491 -10,1309 -11,0388 SLC38A3 -1,3397 -1,7519 -0,7253 1,88E-04 1,54E-04 0,0303 -9,5853 -13,671 -3,4071 PRKCG -1,2792 -1,4452 -0,7152 8,21E-04 0,0275 0,0424 -7,0029 -3,8204 -3,1227 HAPLN4 -1,2742 -1,3862 -0,8578 4,69E-04 0,0328 6,07E-03 -7,9051 -3,638 -4,9361 GSTM1 -1,167 -0,3839 -0,5334 1,11E-04 0,0407 0,0478 -10,856 -3,4366 -3,0192 CYP46A1 -1,0856 -0,5695 -1,1756 5,88E-04 0,0105 1,82E-03 -7,5576 -4,7465 -6,9129 NRCAM -1,0739 -0,8491 -0,7405 2,06E-04 0,0241 3,57E-03 -9,4001 -3,9498 -5,7204 DLG1 -1,0579 -1,1193 -1,3255 5,85E-03 0,0107 3,57E-03 -4,8872 -4,6983 -5,7382 GAD2 -0,9959 -0,6562 -0,7078 2,10E-04 0,0328 0,0173 -9,2643 -3,6482 -3,9062 ATP1A2 -0,9221 -0,4355 -0,7349 1,97E-03 0,0428 0,0429 -6,0553 -3,3848 -3,1115 SLC12A5 -0,9064 -0,8074 -2,2111 1,97E-03 0,0301 2,15E-04 -6,0527 -3,7404 -12,0952 HOMER1 -0,9022 -0,9154 -0,6216 5,74E-03 0,0256 0,0144 -4,9177 -3,8863 -4,0932 TTYH1 -0,8935 -0,7255 -1,4004 9,93E-03 2,49E-03 2,80E-03 -4,4403 -6,4596 -6,2162 HSD17B7 -0,8744 -0,9277 -1,1029 2,59E-04 2,19E-03 2,15E-04 -8,8037 -6,621 -12,1043 COX6B1 -0,8642 -0,5963 -1,3287 5,88E-04 4,25E-03 2,93E-03 -7,5646 -5,864 -6,1056 CNTNAP1 -0,8605 -0,9546 -1,5431 0,0109 0,0248 2,91E-03 -4,3535 -3,9127 -6,1456 CALU -0,8269 -0,4464 -1,1017 3,51E-03 3,85E-03 5,25E-03 -5,4294 -5,9709 -5,1073 MAL2 -0,8249 -0,7857 -0,3985 0,0139 4,53E-03 0,0423 -4,1144 -5,7028 -3,1284 EIF3J1;EIF3J2 -0,6799 -1,1254 -1,0949 2,71E-03 6,46E-03 0,0182 -5,7094 -5,3002 -3,8519 COX5A -0,6443 -0,5883 -0,6424 0,0269 4,53E-03 4,03E-03 -3,489 -5,7106 -5,4833 WRN -0,6072 -0,5017 -0,6195 0,015 4,41E-03 2,46E-03 -4,0276 -5,7815 -6,3521 CACNB4 -0,5732 -0,4249 -0,3754 0,0172 5,09E-03 0,0275 -3,8944 -5,5663 -3,5134 SLC25A31 -0,525 -1,0123 -0,7691 0,0145 0,0189 0,0383 -4,0653 -4,161 -3,2047 CADPS -0,5102 -1,056 -0,5688 0,0241 5,41E-04 8,38E-03 -3,5915 -8,5744 -4,5905 MPP2 -0,4727 -1,0437 -0,5593 0,0408 9,95E-03 0,0285 -3,1375 -4,8256 -3,4772 NSF -0,4629 -0,5834 -0,446 8,03E-04 4,00E-03 3,57E-03 -7,0544 -5,9341 -5,7084 UBQLN2 -0,4351 -0,6255 -0,5712 0,016 7,62E-03 0,0229 -3,9706 -5,1353 -3,6548 CASKIN1 -0,4286 -0,5819 -0,9158 0,0189 0,0121 6,03E-03 -3,8061 -4,5766 -4,9573 NDUFA10 -0,4281 -0,3017 -0,7369 3,65E-03 0,0241 3,30E-03 -5,3818 -3,9507 -5,8467 ARHGAP23 -0,3603 -0,6395 -0,7949 0,0396 0,0103 0,0221 -3,1755 -4,7732 -3,688 GSTP1 0,2805 0,3165 0,3704 0,0463 0,0239 0,0132 3,0287 3,9579 4,1698 EIF3L 0,2879 0,2759 0,3823 0,0202 0,0347 0,0215 3,7438 3,5813 3,7088 UCHL1 0,3848 0,2558 0,6416 0,0188 0,0243 2,25E-03 3,8164 3,937 6,5293 RUVBL2 0,4002 0,3941 0,4849 0,0439 0,0186 0,0217 3,0706 4,19 3,7019 ARCN1 0,4005 0,4288 0,7002 0,0309 0,0268 0,0179 3,3637 3,8389 3,8704 EIF3D 0,4341 0,5592 0,4746 0,0114 9,22E-03 0,0249 4,2927 4,8998 3,5927 NIPSNAP3B 0,4349 0,4354 0,5513 0,0141 9,13E-03 0,0377 4,0954 4,9104 3,2239 ARF1;ARF3 0,444 0,47 0,7139 9,00E-03 0,0107 0,0139 4,5216 4,7085 4,1177 CMAS 0,5274 0,5351 1,1777 0,0145 8,53E-03 3,92E-04 4,0636 5,0076 9,519 PHPT1 0,5432 0,5851 1,0529 0,0354 0,0451 3,12E-03 3,2601 3,3453 5,957 PRMT5 0,5907 0,7489 0,8382 5,36E-03 9,41E-04 0,0137 4,9873 7,8038 4,1375 SARS 0,6114 0,431 0,5456 8,03E-04 5,82E-03 7,33E-03 7,0467 5,411 4,7573 COTL1 0,6983 0,6134 0,6995 0,0138 0,0263 0,0275 4,1228 3,8553 3,512 GNS 0,7469 0,8501 0,6852 1,71E-03 3,58E-03 7,38E-03 6,2238 6,0424 4,7214 ATP5D 0,8292 0,9812 1,0355 2,98E-03 3,64E-03 0,0173 5,6123 6,0218 3,9064 CTSD 0,9413 0,7229 0,6115 4,92E-04 8,76E-03 7,42E-03 7,8237 4,963 4,7117 LAMP1 0,9597 0,4383 0,8584 1,88E-04 0,0107 1,79E-03 9,5817 4,7145 6,9593 UBE2K 1,0167 0,86 0,9268 8,84E-04 1,28E-03 0,0313 6,9185 7,3601 3,3755 ATP6V1G1 1,15 0,7426 1,2081 4,72E-03 0,0208 4,03E-03 5,1205 4,0787 5,4892 TRY10 1,2077 1,1499 1,4952 7,15E-04 6,43E-03 0,0433 7,2689 5,3091 3,0974 WASL 2,0974 2,5595 2,6242 1,88E-04 2,20E-04 3,57E-03 9,6109 10,5997 5,6921 -3 0 3 0.00001 0.01 0.05

74 SNX4 in presynaptic terminals

AM as a calcium indicator. The expression of Synapsin-ECFP allowed the visualization of synaptic terminals. Electrical stimulation (100 AP, 40 Hz, 30 mA) resulted in robust intracellular calcium signals with similar increase and decrease kinetics in the four groups (Supplementary Figure S8). This indicates that SNX4 depletion does not affect calcium dynamics after stimulation. Hence, defects in the calcium flux cannot explain the impairment in synaptic vesicle release upon SNX4 knock down.

The neuronal proteome is dysregulated upon SNX4 knock down

To investigate the impact of knocking down SNX4 in the neuronal proteome, the proteome of SNX4 knock down and control neurons was characterized and compared. Five independent cortical cultures were infected with control shRNA and three different shRNAs against SNX4 at DIV7. At DIV15, cells were harvested and the proteins were extracted and digested into peptides for subsequent identification and quantification using LC-MS/MS (Gillet et al., 2012; Koopmans et al., 2018). A total of 2531 proteins were identified and quantified from a total of 12027 peptides. Only peptides identified with high confidence were used (i.e., a Q-value ≤ 0.01 over all samples in at least one group, allowing for one outlier within each condition). The full list containing the 2531 proteins quantified in this study is in Supplementary Table S2. Hierarchical clustering was used to classify samples into groups according to similarities between them (Figure 8a). Each biological replicate from the same group (control, shSNX4-1, shSNX4-2 and shSNX4-3) 3 clearly clustered together, suggesting that the expression of each shRNA leads to a neuronal proteome with each own identity.

To study the dysregulation in the neuronal proteome upon SNX4 depletion, we focused in dysregulated proteins, which had moderated effect-size higher (upregulated) or lower (downregulated) than 3 compared with control (Supplementary Table S2 and Figure 8b, c). 313, 175 and 317 proteins were uniquely dysregulated compared with control, in shSNX4-1, shSNX4-2 and shSNX4-3 expressing neurons respectively. Cellular levels for 90 proteins were significantly different in all the three knock down groups compared with control (Figure 8b). Only statistically significant proteins for which dysregulation was

Figure 8: SNX4 depletion dysregulates the neuronal proteome. (a) Dendogram of the protein expression relationship between the neurons containing shRNA control and shRNA against SNX4. The hierarchical clustering reflects similarity between the samples. (b) Venn diagram showing the overlap among the dysregulated proteins in neurons containing shRNA against SNX4 with a moderated effect-size eBayes@limma > ±3 compared with control. (c) Heatmap of the protein expression of dysregulated proteins in SNX4 knock down neurons. The log2 of the fold change is color coded: Red indicates the log2 of the fold change of the maximum downregulation, green indicates the maximum upregulation and yellow no dysregulation. The p-value is color coded: Dark blue indicate the lowest p-value, and white p-value=0.05. The moderate effect size is not color coded.

75 Chapter 3 in same direction among the three knock down groups (upregulated or downregulated protein in the three groups) were considered. Hence, 36 proteins were downregulated and 21 were upregulated in the three knock-down groups (Figure 8c). To analyze these results, g:Profiler enrichment analysis was used (Supplementary Table S3) (Reimand et al., 2016). The downregulated proteins were i.e. enriched in proteins involved in synaptic signaling (ATP1A2, DLG1, LGI1, MPP2, SLC1A2, SLC12A5, PRKCG, GAD1, GAD2, HOMER1, CADPS, NPTX1, CACNB4), and GABAergic synapse functional groups (NSF, SLC12A5, PRKCG, GAD1, GAD2, SLC38A3).

a b c ** kDa 2.5 * 4 *** 100 ControlshSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 +shSNX4-3 R1Control + R2 + R3 .u. ) 2.0 LAMP1 3 .u. )

1.5 2

1.0 (a LAMP1 1

LAMP1 (a LAMP1 TCE 0.5 0 0.0

Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 d e f g shSNX4-1shSNX4-2 + R1shSNX4-3 + R2 + R3

* 1.5 ** 1.5 * kDa *** *** ControlshSNX4-1shSNX4-2shSNX4-1 + R1 3 *** 70 GAD1 GAD2 .u. ) 1.0 .u. ) 1.0 2 .u. ) (a 1 (a AD 1 2 (a AD 2 G 0.5 G 0.5

TCE GAD 1

0.0 0.0 0

Control Control Control shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2shSNX4-3 shSNX4-1shSNX4-2

shSNX4-1 + R1

Figure 9: Levels of selected proteins measured using proteomics and western blot. (a) Quantification of LAMP1 levels measured in proteomics. (b) Western blot for LAMP1 and total amount of protein measure with TCE. (c) Quantification of LAMP1 levels normalized to total amount of protein in stain-free gel. (d) Quantification of GAD1 levels measured in proteomics. (e) Quantification of GAD2 levels measured in proteomics. (f) Western blot for GAD1 and GAD2 and total amount of protein measure with TCE. (g) Quantification of GAD1 and GAD2 levels normalized to total amount of protein in stain-free gel. Original uncropped blots are shown in Supplementary Figure S9. (Proteomic data N=5 animals, western blot data N=3±2 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

Some of the dysregulated proteins in the mass spectrometry were analyzed by western blotting. GAD1 and GAD2 were reduced by 40% and 60% in the knock down groups in mass spectrometry (Figure 8c, 9d, e). Upon expression of shSNX4-1 and shSNX4-2, GAD1 and GAD2 average levels were bellow control and rescue SNX4 groups by western blotting (Figure 9f, g and Supplementary Table S1 for exact values). In addition, both

76 SNX4 in presynaptic terminals

Cathepsin D and LAMP1, which are canonical lysosomal protein makers, were 60% upregulated in SNX4 knock down groups (Figure 8c, 9a). In LAMP1 western blots, the average levels of LAMP1 knock down groups fall above the control average and the rescued groups (Figure 9b, c and Supplementary Table S1 for exact values). Although western blot data are not significantly different between groups, the average protein levels in SNX4 knock down groups moved in the same direction as the changes detected by mass spectrometry data.

Together, these data indicate that each SNX4-targetted shRNA alters the cellular levels of many proteins, including lysosomal proteins and proteins involved in synaptic transmission.

DISCUSSION

We have characterized the localization of SNX4 in murine neurons and its role in presynaptic structure, composition and function. A novel antibody was developed and validated for detecting endogenous mouse SNX4 in western blot, immunocytochemistry and immuno-electron microscopy. Endogenous SNX4 partially co-localized with both early and recycling endosomes in neurons, which is in accordance with the previously established role of SNX4 in non-neuronal cells (Traer et al., 2007). SNX4 accumulated in synaptic areas and immuno-electron microscopy revealed that SNX4 was predominantly 3 presynaptic within the synapse. SNX4 knock down impaired synaptic vesicle release without affecting the morphology of neurons or presynaptic ultrastructure. However, this impairment was not restored by re-expressing a knock-down resistant variants of SNX4. Synaptic communication-related proteins and lysosomal proteins were among the proteins with altered cellular levels upon SNX4 depletion. This study identifies SNX4 as a presynaptic protein and suggests that SNX4-dependent sorting is important at presynaptic terminals.

SNX4 is a presynaptic protein

In this study, SNX4 has been identified as a synaptic protein. Endogenous SNX4 was expressed in all brain areas tested and in brain cell types (Figure 1). This ubiquitous expression of SNX4 was expected, based in its evolutionary conservation across the eukaryotes (Koumandou et al., 2011; van Weering et al., 2010). At the neuronal level, both endogenous and overexpressed SNX4 highly colocalized with several synaptic markers (Bassoon, Synaptophysin-1, VGluT1 and VAMP2/Synaptobrevin-2) (Figure 4, Supplementary Figure S5 and Supplementary Table S1). Although the new antibody showed an unspecific band in western blot analyses, immunocytochemistry-labeling

77 Chapter 3

Table 1: Overview of the effect of the different shRNAs against SNX4 and SNX4 expression constructs in the measured variables compared to control. Significant decrease is note as ‘red arrow’, significant increase as ‘green arrow’, no significant difference is noted as “=” and not applicable as ‘empty cell’. Detailed information (average, SEM, n and statistics) is displayed in Supplementary Table S1.

shSNX4-1 shSNX4-2 shSNX4-3 Over- Figure Meassured variable shSNX4-1 shSNX4-2 shSNX4-3 +R1 +R2 + R3 expressed 2f Synapses per µm (Syph-1) = = = = = 2f Synaptophysin (a.u.) ======⬇ 3l TfnR (a.u.) ======5b dendritic length (mm) = = = 5c axonal length (mm) = = = 5e Synapses per µm (bassoon) = = = 5f Bassoon (a.u.) = = = 5g Homer-1 (a.u.) = = = 5h Synaptophysin-1 (a.u.) = = = 5i Synaptotagmin-1 (a.u.) = = 14j VAMP2/synaptobrevin-2 (a.u.) = = ⬇ 5k Syntaxin-1 (a.u.) = ⬇ 6b active zone length (nm) = = ⬇ ⬇ 6c # synaptic vesicles = = ⬆ 6d #docked vesicles = = ⬆ 6e # big vesicles = = ⬇ 6f # tubules = = = ⬇ 6g % synapses with MVB = = = 7c Fmax (a.u.) = = = 7d F Baseline (a.u) = = 7e 1st peak amplitude (∆F/Fmax) ⬇ 7f % Active synapses ⬇ ⬇ ⬇ 7g Ratio peak amplitude (2nd/1st) = = = ⬇ ⬇ ⬇ 7h F pH=5.5 (a.u.) = 7k Fmax (a.u.) ======⬇ ⬇ 7l F Baseline (a.u) ======7m 1st peak amplitude (∆F/Fmax) = 9a LAMP1 (a.u) proteomics ⬇ ⬇ ⬇ ⬇ ⬇ ⬇ 9c LAMP1 (a.u) WB ======⬆ ⬆ ⬆ 9d GAD1 (a.u) proteomics 9e GAD2 (a.u) proteomics ⬇ ⬇ ⬇ 9g GAD (a.u) WB = = = ⬇ ⬇ ⬇

of SNX4 in synapses proved to be specific: the signal was decreased upon SNX4 shRNA and restored upon introduction of a shRNA-resistant SNX4 variant (Figure 2d, i). Supporting this specific synaptic distribution, SNX4 was found in synaptic fractions by western blot where the specific SNX4 band was observed both in synaptosomes and in the synaptic membrane fraction (Figure 4e-i). Immuno-electron microscopy revealed that endogenous SNX4 is present in both the pre- and post- synaptic terminals, being most abundant in presynaptic terminals. Presynaptic SNX4 appeared in the synaptic vesicle cloud suggesting a synaptic vesicle localization, which has not been previously described (Takamori et al., 2006). Hence, SNX4 is a predominantly presynaptic protein, which suggests that SNX4-mediaded recycling may be required for presynaptic function.

78 SNX4 in presynaptic terminals

The shRNA approach: limitations

A shRNA approach was used to determine the role of SNX4 in presynaptic terminals. This approach is widely used to acutely deplete proteins to study their function; however, shRNAs may produce off target effects (see reviews (Fellmann and Lowe, 2014; Kaelin, 2012)). In our study, only phenotypes replicated by three validated shRNAs against SNX4 are considered SNX4 dependent. Furthermore, if a phenotype was produced by the three shRNAs, a rescue experiment was performed to assure SNX4-dependence by expressing shRNA-resistant SNX4. The proteomic analysis of SNX4 depleted neurons showed that each shRNA induced a different dysregulation of the neuronal proteome (Figure 8), while SNX4 was knocked down similarly by the three shRNAs against SNX4 (Figure 2). These uniquely dysregulated proteins seem SNX4 depletion independent and they might produce off-target effects, which might explain the different effects observed in Table 1.

SNX4 recycling pathway in neurons

Most SNX4 studies have been performed in mitotic cells, where SNX4 localizes in early and recycling endosomes (Leprince et al., 2003; Solis et al., 2013; Teasdale et al., 2001; Traer et al., 2007; van Weering et al., 2012b). In neurons, we found that SNX4 also co-localized with early (Rab5) and recycling (Rab11) endosomal markers, suggesting that neuronal SNX4 functions in the previously described recycling pathway from early endosome to 3 recycling endosome to the plasma membrane. In HeLa cells, SNX4 silencing leads to abnormal Rab11 puncta distribution: from juxtanuclear to peripherical distribution (Traer et al., 2007). In neurons, SNX4 depletion did not affect the distribution of Rab11 puncta in synapses, indicating that SNX4 depletion does not affect recycling endosomal distribution in synapses (Supplementary Figure S4).

In HeLa cells, SNX4 recycles cargo proteins from the early endosome back to the plasma membrane through recycling endosomes, avoiding its lysosomal degradation. Hence, when SNX4 is depleted, SNX4-dependent cargo is degraded at the lysosome (Traer et al., 2007). Upon SNX4 depletion, we found that the expression of some proteins involved in synaptic transmission were decreased. Although lysosomal processes were not detected as an affected functional protein group (Supplementary Table S3), the individual lysosomal proteins Cathepsin D and LAMP1 were increased upon SNX4 knock down. Potentially the SNX4 recycling pathway is also present in presynaptic terminals where it recycles presynaptic proteins such as GAD1 and GAD2 from endosomes avoiding degradation. The increase in two lysosomal proteins might suggest an increase of lysosomal compartments to cope with the increased degradation demand.

79 Chapter 3

The best described SNX4-dependent recycling cargo is the TfnR, which is decreased upon SNX4 depletion in HeLa cells (Traer et al., 2007). In neurons, SNX4 depletion did not decrease TfnR levels, which suggest that the SNX4-dependent recycling might be different in different cell types (Figure 3k-m and Supplementary Figure S3). The impact of SNX4 depletion on TfnR levels in neurons is different compared with HeLa cells, but it is not possible to exclude that the TfnR is not recycled by SNX4. For example, neuronal SNX4-dependent cargo might not be degraded in lysosomes upon SNX4 depletion but be mis-trafficked or accumulated in internal compartments where it cannot function without affecting the total protein level.

SNX4 might be involved in the insertion of proteins in the plasma membrane of synaptic terminals. In presynaptic terminals, the endosomal system is known to be involved in the insertion of metabotropic receptors such as G protein–coupled receptors (GPCRs) (see review (Irannejad and von Zastrow, 2014)). GPCRs can be localized at presynaptic terminals and are important regulators of synaptic communication (see review (Atwood et al., 2014)). On the postsynaptic side, the endosomal system is also involved in the insertion of neurotransmitter receptors in the plasma membrane which is an also known mechanism of synaptic plasticity (see review (Kneussel and Hausrat, 2016)). Hence, SNX4 might play a role regulating the surface localization of receptors and, therefore, synaptic plasticity.

SNX4 in Alzheimer’s disease

In neurons, SNX4 has been studied in Alzheimer’s disease and Aβ production. SNX4 depletion decreases APP levels and modulation of SNX4 levels caused a dysregulation of BACE1 and Aβ levels (Kim et al., 2017). BACE1 was not detected, but APP was identified and quantified in the proteomic analysis upon SNX4 depletion. APPlevels were only decreased upon shSNX4-3 expression, but not upon shSNX4-1 and shSNX4-2 (Supplementary Table S2, shSNX4-3 p=0.0172). Based in our data, SNX4 depletion does not decrease APP levels. The lack of rescue experiments in previous studies and the inconsistency with our data indicates that more research is required to establish the role of SNX4 in Aβ production.

The role of SNX4 in synaptic vesicle release

SNX4 depletion impaired synaptic vesicle release. However, this phenotype was not restored by SNX4 re-introduction. The expression of all shRNAs against SNX4 decreased synaptic vesicle release in all readouts: the peak amplitude using both fluorescent reporters and the frequency of mEPSCs (Figure 7e, m and Supplementary Figure S7).

80 SNX4 in presynaptic terminals

In combination with the shRNAs, the re-introduction of shRNA resistant SNX4 did not restore the synaptic vesicle release phenotype. SNX4 was decreased in synapses upon expression of the three shRNAs and increased upon re-expression of shRNA-resistant SNX4 constructs (Figure 2). In some neurons, SNX4 re-introduction resulted in higher SNX4 levels than the endogenous ones (Figure 2h). Overexpression of SNX4 in control neurons did not produce defects in synaptic vesicle release (Figure 7m). Hence, increased SNX4 levels in the rescue conditions cannot explain the lack of restoration in the release phenotype. The functionality of the re-expressed protein was not directly addressed in this study. Although very unlikely, there are scenarios in which re-expressed SNX4 might not be functional. We cannot exclude that presynaptic SNX4 might require alternative splicing or posttranslational modification, but there are no indications in the data that suggest this. Although it seems unlikely that three independent shRNAs against SNX4 produce the same off-target effect, the straight forward explanation of why the re-introduction of SNX4 cannot restore synaptic vesicle release is that this phenotype is due to an off-target effect of the shRNAs.

Understanding why SNX4 knock down impairs synaptic vesicle release

In this study, the possible explanations for the synaptic vesicle release impairment upon SNX4 knock down were narrowed by excluding frequent causes of synaptic vesicle release phenotypes. Synaptic vesicle release is a high energy demanding process (Harris 3 et al., 2012). The membrane retrieval and re-acidification of intracellular compartments which involves the active pumping of H+ also requires ATP (Harris et al., 2012). After exposure to ammonium, which neutralized the pH of the intracellular compartments, all neurons re-acidified its compartments (Figure 7). Calcium imaging showed that calcium dynamics, which are highly ATP-dependent (Harris et al., 2012), were similar among all the groups (SNX4 knock down and control, Supplementary Figure S8), indicating that the energy supply is not compromised in these neurons. Neuronal morphology was not altered in SNX4 knock down neurons which is a commonly described off-target effects associated with the expression of shRNAs in neurons (Alvarez et al., 2006). These observations indicate that the shRNAs against SNX4 do not produce a toxic effect.

Synaptic vesicle release impairment may be explained by reduced number of synaptic vesicles or reduced levels of crucial proteins for the release. The release reporters were equally expressed and the number of synaptic vesicles were equal among groups (Figure 7c, k and Figure 6c). The reporters were also equally quenched under resting conditions, and the amount of synaptopHluorin reporter in the plasma membrane was not modified, which are indications of proper retrieval and acidification (Figure 7b, d, l, h). Proteomic analysis showed that none of the known crucial proteins for synaptic vesicle release was

81 Chapter 3 dysregulated upon SNX4 depletion (Supplementary Table S2). It is not possible to exclude that the mis-traffic of these proteins is the cause of the vesicle release phenotype, but the reduction of any single protein upon SNX4 depletion cannot explain the synaptic vesicle release phenotype.

This study highlights the importance of the use of independent shRNAs and rescue experiments to study protein function. The novel identification of SNX4 at presynaptic terminals opens a new line of research on the role of endosomal sorting in presynaptic function. To advance this field, the generation of new tools such a conditional knock out mice line seems indispensable. Notwithstanding its limitations, this pioneer study demonstrates that SNX4 is in synapses, indicating a synaptic SNX4 demand.

MATERIALS AND METHODS

Plasmids

Short harping RNA (shRNA) against SNX4 was used to knock down SNX4. The target sequences were cloned in to a lentiviral expression vector under the U6 promotor. To report lentiviral infection the plasmid also contained mCherry under Synapsin promotor. The target sequences of the shRNAs were as follows: GGG AAT GAC TAC CAA ACT C (shSNX4-1), GCA GTG GAA TAG ATA CAT TAT (shSNX4-2), GCT GAT ATT GAA CGC TTC AAA (shSNX4-3), TTC TCC GAA CGT GTC ACG T (shControl, scramble) (Zhang et al., 2008). To carry out the rescue experiments, the mouse SNX4 cDNA was used (Supplementary Figure S2). To induce the silence mutagenesis to rescue shSNX4-1 the following primers were used: GAAGGGAATGACAACGAAGCTTTTTGGTCAAGAAACTCCAG (forward) and CTGGAGTTTCTTGACCAAAAAGCTTCGTTGTCATTCCCTTC (reverse). To induce the silence mutagenesis to rescue shSNX4-3 the following primers were used: GGGCTGATATCGAGCGCTTTAAAGAACAAAAG (forward) and CTTTTGTTCTTTAAAGCGCTCGATATCAGCCC (reverse). See Supplementary Figure S2.

To report synaptic vesicle release we used Synaptophysin-pHluorin (sypHy) (Granseth et al., 2006b) and Synaptobrevin-pHluorin (synaptopHluorin) (Miesenböck et al., 1998), both under Synapsin promotor. Synapsin-ECFP was used to labelled synapses in life cell imaging. This construct was obtained by replacing mCherry with ECFP of Synapsin- mCherry which was a kind gift of Dr. A. Jeromin (Allen Brain Institute, Seattle, USA) (Farina et al., 2015). Human GFP-SNX4 plasmid was a kind gift of Pete Cullen (University of Bristol, UK).

82 SNX4 in presynaptic terminals

Laboratory animals

Animal experiments were approved by the animal ethical committee of the VU University/ VU University Medical Centre (“Dier ethische commissie (DEC)”; license number: FGA 11-03) and, according to institutional and Dutch governmental guidelines and regulations.

Primary cell culture

Primary neurons were cultured from wild-type mouse E18 hippocampi or cortices. Briefly, tissue was dissected in Hanks balance salt solution (HBSS, Sigma) with 10mM HEPES (Life Technologies) and digested by 0.25% trypsin (20 minutes at 37 oC; Life technologies) in HBSS. The tissue disassociation was performed with fire-polished Pasteur pipettes in DMEM with FCS. The neurons were spun down and re-suspended in neurobasal medium with 2% B-27, 18 mM HEPES, 0.25% glutamax and 0.1% Pen-Strep (Life Technologies). Neurons were plated in coated coverslips with poly-L-ornithine (PLO, Sigma) and laminin (Sigma), on astrocyte micro-islands (Wierda et al. 2007) and astrocyte monolayer (for information of each experiment of specific tissue, neuronal density, and substrate, see o Supplementary Table S4). Neurons were maintained at 37 C and 5% CO2 until the day of the experiment.

Subcellular fractioning 3

Subcellular fractions were obtained from hippocampi from three-month-old wild-type mice as previously described (Pandya et al., 2017; Von Engelhardt et al., 2010). Isolated hippocampi were homogenized on a dounce homogenizer (potter; 12 strokes, 900 rpm) using homogenizer buffer (0.32 M Sucrose, 5 mM HEPES pH 7.4, Protease inhibitor cocktail (Roche)), and spun at 1000xg for 10 minutes at 4oC to obtain Supernatant 1 (S1). S1 was centrifuged at 20,000xg for 20 minutes to obtain pellet 2 (P2) and supernatant 2 (S2). S2 was ultracentrifuged at 100,000xg for 2 hours to obtain the pellet containing the microsomal fraction (M). S1 was ultracentrifugated in a 0.85/1.2 M sucrose density gradient at 100,000xg for 2 hours to obtain Synaptosomes (SyS) at the interface of 0.85/1.2M sucrose. SyS were exposed to a hypotonic shock of 5 mM HEPES pH 7.4 with protease inhibitor for 15 minutes, and sucrose gradient ultracentrifugated as stated above to obtain the synaptic membrane fraction (SyM) at the interface of 0.85/1.2M. SyS was also treated with 1% Tx-100 for 30 minutes, layered on top of 1.2/1.5/2M sucrose, centrifuged at 100,000xg for 2 hours, to obtain the PSD fraction (PSD) at the interface of 1.5/2M sucrose.

83 Chapter 3

Western blot

To characterize SNX4 expression in brain tissue, different mouse brain areas were dissected, weighted and homogenized in ice-cold PBS with protease inhibitors. The samples were spun down and the pellets were lysed in 100µl Laemmli sample buffer (2% w/v sodium dodecyl sulfate (SDS), 10% v/v Glycerol, 0.26 M β-mercaptoethanol, 60 mM Tris-HCl pH 6.8, and 0.01% w/v Bromophenolblue) per each mg of brain tissue. Samples were boiled for 10 minutes at 90 oC and per each brain area, 10 µl were loaded in SDS-PAGE (10% 1 mm acrylamide gel with 2,2,2-Trichloroethanol) and, transferred into Polyvinylideenfluoride (PVDF) membranes (Bio-rad) (1 hour, 0.3o mA,4 C). Membranes were blocked using 2% milk (Merck) with 0.05% of normal goat serum (NGS) in PBS-T (PBS with 0.1% Tween-20). Membranes were cut based on the molecular weight marker and incubated overnight at 4oC with the primary antibodies in PBS-T (see Supplementary Table S5 for primary antibody details). After three washes with PBS-T, membranes were incubated with secondary alkaline phosphatase conjugated antibodies (1:10000, Jackson ImmunoResearch) in PBS-T during 1 hour at 4oC, washed three times with PBS-T and, incubated 5 minutes with AttoPhos (Promega). The images were acquired with a FLA- 5000 fluorescent image analyzer (Fujifilm). ImageJ Gel Analysis method was usedto compare the intensity of signal both in the gel with TCE and in the western blot.

For protein quantification in SNX4 knock down neuronal cultures, cortical neurons at DIV14-15 were washed with ice-cold phosphate-buffered saline (PBS), scraped, lysed in Laemmli sample buffer and, boiled for 10 minutes at 90 oC. Per each condition, 300.000 neurons were loaded in SDS-PAGE (10% 1 mm acrylamide gel with 2,2,2-Trichloroethanol), and the western blot was continue as described above.

To quantify the total amount of protein in each subcellular fraction, Bradford assay was used. All fractions (including initial hippocampus) were lysed in Laemmli sample buffer, boiled for 10 minutes at 90 oC, and loaded (5µg of each sample) in SDS-PAGE (10% 1 mm acrylamide gel with 2,2,2-Trichloroethanol). Western blot was continue as described above. The total amount of loaded protein was quantified in the gel with TCE. Each sample was normalized to this value and to the value of the hippocampal sample in order to compare between experiments.

Immunocytochemistry and Confocal Imaging

Neurons at DIV 14-15 were fixed with 2% paraformaldehyde in PBS and cell culture media for 10 minutes followed by 4% paraformaldehyde in PBS for 30 minutes at room temperature. Then, neurons were washed three times with PBS, permeabilized with 0.5%

84 SNX4 in presynaptic terminals

Triton X-100 for 5 minutes and, blocked with 2% normal goat serum and 0.1% Triton X-100 in PBS for 40 minutes. Neurons were incubated at room temperature during 1 hour with primary antibodies, washed three times with PBS, incubated during 1 hour with secondary antibodies conjugated to Alexa dyes (1:1000, Molecular Probes), washed three times with PBS, and mounted on microscope slides with Dabco-Mowiol (Invitrogen). The antibodies were diluted in 2% normal goat serum and 0.1% Triton X-100 in PBS at its optimal dilution (for primary antibodies details see Supplementary Table S5).

For SynD analysis (Schmitz et al., 2011), confocal images were acquired using a Carl Zeiss LSM510 meta confocal microscope, with a Plan-Neofluar 40x/1.3 oil objectives. For colocalization analysis, confocal images were acquired in the same microscope but optical zooms of the neurites with 5 times of magnification were acquired and analyzed using JACoB plugin (Bolte and Cordelieres, 2006). For quantification of protein levels, images were acquired in a confocal microscope (Nikon Eclipse Ti) equipped with 63x/1.4 oil objective controlled by NisElements 4.30 software and analyzed measuring the intensity inside a neuronal mask (using mCherry) in ImageJ.

Electron microscopy

For immuno-gold detection of SNX4 in TEM, hippocampi of 2 months old mice were fixed in 4% PFA with 0.1% glutaraldehyde (GA, Merck) in 0.1M PB and embedded in increasing 3 concentrations of gelatin at 37°C (5 minutes 2% gelatin, 15 minutes 5% gelatin, 30 minutes 10% gelatin, 10 minutes 12% gelatin, 60 minutes 12% gelatin). The hippocampi were infiltrated in 2.3 M sucrose at 4°C and frozen in liquid nitrogen. Seventy nm thick sections were obtained with a cryo-ultramicrotome (UC6, Leica), collected at −120°C in 1% methyl-cellulose and 1.2 M sucrose and transferred onto formvar/carbon-coated copper mesh grids. The sections were washed with PBS at 37°C and treated with 0.1% glycine to quench aldehyde groups. The sections were blocked with 0.1% of BSA and 0.1% cold water fish gelatin and incubated during 2 hours at room temperature with SNX4 antibody (1:100, Synaptic Systems, Cat. No. 392 003) diluted in blocking solution. To detected the primary rabbit antibody, Protein A-10 nm gold (1: 25, CMC, UMC Utrecht, Netherlands) was incubated during 1hour at room temperature. The negative controls were processed in parallel without primary antibody and with primary antibody preincubated with the blocking peptide (SynapticSystems, Cat. No. 392-0P at a ratio of 1:10). The sections were counterstained with 0.4% uranyl acetate in 1.8% methyl-cellulose on ice and imaged on a Tecnai 12 Biotwin transmission electron microscope (FEI company).

For ultrastructural characterization of SNX4 knock down neurons, neurons at DIV14-15 were fixed for 1 hour with 2.5% GA in 0.1 M cacodylate buffer, pH 7.4, washedand

85 Chapter 3

stained 1 hour at room temperature with 1% OsO4/1% KRu(CN)6 in milliQ water. Then, cells were dehydrated though increasing ethanol concentrations (30%, 50%, 70%, 90%, 96%, 100%), followed by flat embedded with EPON 50% in 100% ethanol for 30 minutes, and 100% EPON for 48 hours at 65 oC. The coverslips were removed by bathing in liquid nitrogen and boiling water. Ultrathin sections (80 nm) were obtained by cutting parallel to the cell monolayer and collected on single-slot formvar-coated copper grids. Cells were stained using uranyl acetate and lead citrate in Ultra stainer LEICA EM AC20. Images were acquired at 60.000x magnification using a JEOL1010 transmission electron microscope at 60kV using a side-mounted CCD camera (Morada; EMSIS, Münster, Germany) and iTEM analysis software (EMSIS). Large vesicles were defined as clear core vesicles which did not fall in the category of synaptic vesicles.

Live cell Imaging

All live cell imaging experiments were carried with DIV14-15 neurons. Coverslips were placed in an imaging chamber containing Tyrode’s solution (2 mM CaCl2, 2.5 mM KCl, 119 mM NaCl, 2 mM MgCl2, 30 mM glucose, 25 mM HEPES, 50μM AP5 and 10 μM DNQX at pH 7.4). The experiments were performed at room temperature with perfusion of 1 ml per minute of Tyrodes buffer. Images were acquired with the Axiovert II microscope (Zeiss, Oberkochen, Germany) with a 40x oil objective (NA 1.3) using a Polychrome VI light source and a Photometrics Cascade camera. The filters were 488 ± 5 nm (emission) and 525±25 nm (excitation) for pHluorin or Fluo-5F, 514±5 nm (emission) and 625±27,5 nm (excitation) for mCherry as shRNA reporter, and and 457±5 nm (emission) and 480±10 nm (excitation) for Synapsin-ECFP. MethaMorph imaging software was used to control the microscope and record the images. To study synaptic vesicle release, images were acquired at 1Hz using the specified protocols in the result section. The protocols included 30 first seconds of base line recording, one or two identical stimulation (2,5 seconds at 40 Hz and 30 mA) followed by one minute of recovery time, 10 seconds perfusion of NH4 (2 mM CaCl2, 2.5 mM KCl, 119 mM NaCl, 2 mM MgCl2, 30 mM glucose, 25 mM HEPES, 50 mN NH4Cl at pH 7.4), and 10 seconds of acid perfusion (2 mM CaCl2, 2.5 mM KCl, 119 mM NaCl, 2 mM MgCl2, 30 mM glucose, 25 mM MES at pH 5.5). Fluorescence puncta during NH4 exposure (synaptic locations) were analyzed as regions of interest of 4 by 4 pixels’ radium (ROIs). Fluorescence during depolarization of neurons was normalized to baseline and the maximum fluorescence during NH4Cl perfusion. An active synapse was defined as a roi which responds to NH4Cl perfusion and to the electrical stimulation (a responsive ROI is the one that has and intensity higher than the double of the baseline plus two times the standard deviation). The results for each ROI were averaged for each field of view and presented as data points. For the calcium imaging assay, neurons were incubated 10 minutes with 2µM Fluo-5F, AM (Invitrogen, Cat. No. F14222) in media at 37°C, washed

86 SNX4 in presynaptic terminals with media during 15 minutes and imaged at 4Hz using the following protocol: 10 second of base line recording, one stimulation (100AP, 40 Hz, 30 mA), and 15 seconds of recovery recording. Regions of interest of 4 by 4 pixels’ radium (ROIs) were defined in the neurites using the puncta pattern of ECFP-Synapsin (synaptic locations). Fluorescence during depolarization of neurons was normalized to baseline, and maximum fluorescence during stimulation. The results for each ROI were averaged for each field of view and presented as data points. Fields of view were excluded if a technical problem was detected that could disturb the results.

Patch-clamp electrophysiology

Cortical autaptic neurons were infected at DIV7 with control shRNA and shSNX4-1. At DIV14-18, whole-cell voltage clamp electrophysiological recordings were acquired. Neurons were kept voltage-clamped at membrane potential Vm=-70 mV with borosilicate glass pipettes (2.5-4.5 mOhm) filled with 125 mM K+‐gluconic acid, 10 mM NaCl, 4.6 mM MgCl2, 4 mM K2‐ATP, 15 mM creatine phosphate, 10 U/ml phosphocreatine kinase, and 1 mM EGTA (pH 7.30, 300 mOsmol). External solution consisted of 10 mM HEPES,

10 mM glucose, 140 mM NaCl, 2.4 mM KCl, 4 mM MgCl2, and 4 mM CaCl2 (pH=7.30, 300 mOsmol). Recordings were acquired at room temperature using an Axopatch 200A amplifier (Molecular Devices), Digidata 1322A and Clampex 9.0 software. 3

Proteomics

Cortical neurons were plate at a density of 250.000 neurons/mL in laminin/poly-L-ornithine coated 6-well plates. At DIV7, neurons were lentiviral infected shRNA against SNX4 or Control shRNA. At DIV14-15, neurons were washed twice with ice-cold phosphate- buffered saline (PBS), and scraped twice in 500µl of PBS with protease inhibitor cocktail (Roche) per well. The scraped neurons were pelleted in 1.5mL tubes 5 minutes at 3000 g at 4°C and the supernatant was discarded. The neurons were re-suspended and lysed by pipetting up and down in 15 μL of loading buffer (0.05 M Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 0.1M DTT, 0.001% bromophenol). The samples were heated at 90ºC for 5 minutes and incubated with 3μL of 30% acrylamide at room temperature for 30 minutes to block cysteine residues. To normalize the total amount of proteins among samples, 1 μL of each sample was run in a SDS polyacrylamide gel (10% SDS polyacrylamide gel containing 0.5% 2,2,2-Trichloroethanol (TCE)) . The gel was scanned in a Gel Doc EZ Imager (Bio-Rad) and analyzed with Image Lab software to compare and correct the total protein amount between samples. Each protein sample (~500.000 neurons) was separated about 1cm on a 10% SDS polyacrylamide gel, fixed overnight and stained with colloidal Coomassie Brilliant Blue G. Each sample lane was cut into small fragments

87 Chapter 3 and transferred to the wells of a MultiScreen- HV 96 well filter-plate. The samples were distained (two times) with 150 μL 50% acetonitrile in 50 mM ammonium bicarbonate, dehydrated in 150 μL 100% acetonitrile and rehydrated with 150 μL 50 mM ammonium bicarbonate. The waste solution was collected by centrifugation at 200g for 1min. After the last dehydration in 100% acetonitrile, the dried fragments were re-swelled with 120 μL Trypsin/Lys-C Mix solution (Promega) and incubated overnight in a humidified chamber at 37°C. The peptides were extracted from the gel pieces twice with 150μL 50% acetonitrile in 0.1% TFA, and then once with 150μL 80% acetonitrile in 0.1% TFA. Finally, the peptides were dried in solution using a speedvac and stored at -20ºC.

The peptides were re-dissolved in 7 μL of 2% acetonitrile/0.1% formic acid solution containing iRT reference peptides and injected (6.3 μL of each sample) into the Ultimate 3000 LC system. The peptides were trapped on a 5 mm C18 PepMap 100 column for 5 minutes and separated on a homemade 200 mm C18 Alltima column. The reverse phase liquid chromatography was performed by linearly increasing the acetonitrile concentration in the mobile phase at a flow rate of 5 μL/minute: from 5 to 22% in 88 minutes, to 25% at 98 minutes, to 40% at 108 minutes and to 95% in 2 minutes. The separated peptides were electro-sprayed into the TripleTOF 5600 MS (Sciex) with a micro-spray needle (at a voltage of 5500 V). The mass spectrometer was set in data-independent acquisition at high sensitivity and positive mode under the following parameters: parent ion scan of 100 msec (mass range of 350-1250 Da), SWATH mass range between 450-770 m/z, SWATH window of 8 Da, MS/MS scan time of 80 msec per window (range 200-1800 Da), collision energy for each window was determine for a 2+ ion centered upon the window, with a spread of 15 eV.

Data was analyzed using Spectronaut 8.0 (Bruderer et al., 2015) and a spectral library created from merging two data-dependent analyses of wild-type hippocampal neuron cultures and hippocampal synaptosomes containing spike-in iRT peptides from Biognosys (He et al., 2018). The retention time prediction was set to dynamic iRT; the cross-run normalization based on total peak areas was enabled. The resulted peptide abundances were processed using R language for statistical computation. Protein abundances were computed using Spectronaut normalized peak area, and Loess normalized using the ‘normalizeCyclicLoess’ function from limma R package (fast method and 10 iterations) (Ritchie et al., 2015). Empirical Bayes moderated t-statistics with multiple testing correction by false discovery rate (FDR) was performed on log-transformed protein abundances as implemented by the ‘eBayes’ and ‘topTable’ functions from limma R package. To perform functional enrichment analysis only proteins fulfilling the following criteria were used: moderated effect-size eBayes@limma was superior to ±3, direction of the dysregulation was equal among the SNX4 knock down groups, and Empirical

88 SNX4 in presynaptic terminals

Bayes moderated t-statistics FDR was ≤ 0.05. These proteins were analyzed in g:Profiler (version: r1741_e90_eg37) using the total identified proteins with high confidence as a gen list background, and using default setting (including Homo sapiens as a default organism) (Reimand et al., 2016).

Statistical Analysis

Data are expressed as mean values ± standard error of the mean (SEM). The Shapiro-Wilk normality test was used to evaluate the distribution of the data. Bartlett’s test was used to test homoscedasticity. If comparing two homoscedastic and normal distributed groups, t-test was used. If comparing two groups, data were not homoscedastic and normal distributed, Mann-Whitney test was used. In case of comparing more than two groups, data were normally distributed and homoscedastic, data were compared by one-way analysis of variance (ANOVA). Dunnets post-hoc tests were performed after a significant effect was detected by comparing the different knock down groups to the control. In case of comparing more than two groups, data were not normality distributed and homoscedastic, the Kruskal-Wallis test was used with Dunn’s multiple test as post-hoc. When P-values were lower than 0.05, significance was noted in the figure as: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Data availability 3

The datasets generated and analyzed during the current study are available from the corresponding author on request.

AUTHOR CONTRIBUTIONS

S.V.S. performed experiments and analyzed the data. R.J., and A.W. collected and analyzed confocal images for the SNX4 colocalization studies. M.P.D. collected and analyzed electron microscopy data. M.M. performed electrophysiological recordings. M.A.G.L. and K.W.L produced and critically discussed the proteomic data. S.V.S. and J.R.T.vW. designed the experiments and, wrote the manuscript.

ACKNOWLEDGMENTS

The authors thank Prof. Dr. Matthijs Verhage and Prof. Dr. Peter J. Cullen for their critical reading of the manuscript, Joke Wortel for housing and breeding the mice, Frank den Oudsten and Desiree Schut for providing cell cultures, and Robbert Zalm and Joost Hoetjes for cloning and lentiviral production. EM analysis was performed at the VU/

89 Chapter 3

VUmc EM facility (ZonMW 91111009). This work was supported by the EC under FP7- PEOPLE-2013 (607508) and Alzheimer Nederland (WE.03_2016-05).

COMPETING INTERESTS STATEMENT

The authors declare no competing financial interests.

90 SNX4 in presynaptic terminals

SUPPLEMENTARY FIGURES

MEQAPPDPEKLLQPGPLEPLGGPGAVLEAAVGEENEGTREDGSGVDTMTGNNFWLKKIEISVSEAEKRTG RNAVNMQETYTAYLIETRSVEHADGQSVLTDSLWRRYSEFELLRNYLLVYYPHVVVPPLPEKRAEFVWHK LSADNMDPDFVERRRVGLENFLLRVASHPVLCRDKIFYSFLTQEGNWKETVNETGFQLKADSRLKALNAT FRVKNPDKRFTELRHYSDELQSVISHLLRVRARVADRLYGVYKVHGNYGRVFSEWSAIEKEMGDGLQSAG HHMDVYASSIDDILEDEEHYADQLKEYLFYAEALRAVCRKHELMQYDLETAAQDLAAKKQQCEELATGTV RTFSLKGMTTKLFGQETPEQREARIKVLEEQINEGEQQLKSKNLEGREFVKNAWADIERFKEQKNRDLKE ALISYAVMQISMCKKGIQVWTNAKECFSKM!

b e

ControlshSNX4-1shSNX4-2 kDa ControlshSNX4-1shSNX4-2 kDa 55 SNX4 55 SNX4

40 Actin 40 Actin

c d f g

ControlshSNX4-1shSNX4-2 kDa ControlshSNX4-1shSNX4-2 kDa ControlshSNX4-1shSNX4-2 ControlshSNX4-1shSNX4-2 180 180 130 130 100 100 70 70 55 55 40 40 35 35 25 25 15 15 3 Supplementary Figure S1: Epitopes of the different antibodies against SNX4 (a) Sequence of amino acids of mouse SNX4 (>gi|18017596|ref|NP_542124.1| sorting nexin-4 [Mus musculus]). The epitopes of the different antibodies are highlighted. In orange, the epitope of SNX4 antibody from cat. N. 392 003, Synaptic Systems (1-21 amino acids of mouse SNX4). In blue, epitope from cat. N. HPA005709, Sigma (238-386 amino acids of human SNX4). In yellow, epitope from cat. N. sc-271403, Santa Cruz (361-393 amino acids of human SNX4). The green is just the product of the overlapping sequences highlighted in yellow and blue. (b) Representative western blot of control neurons and neurons with shRNAs against SNX4 stained for SNX4 (N. sc-271403, Santa Cruz) and actin. Original uncropped blots for SNX4 (N. sc-271403, Santa Cruz) (c) and acting (d). (e) Representative western blot of control neurons and neurons with shRNAs against SNX4 stained for SNX4 (N. HPA005709, S) and actin. Original uncropped blots for SNX4 (N. HPA005709, S) (f) and actin (g).

91 Chapter 3

a G GGA ATG ACT ACC AAA CTC (shSNX4-1) b GCA GTG GAA TAG ATA CAT TAT (shSNX4-2) c GCT GAT ATT GAA CGC TTC AAA (shSNX4-3)

d ATGGAGCAGGCACCTCCGGACCCCGAGAAGCTCTTGCAGCCTGGACCCCTGGAGCCGCTGGGTGGCCCTGGCGCTGTGCTGGAGGC CGCGGTCGGTGAGGAGAACGAGGGCACCCGAGAAGACGGCTCAGGGGTCGACACGATGACGGGAAATAATTTTTGGTTGAAGAAAA TAGAAATCAGTGTTTCAGAAGCAGAGAAGAGAACCGGAAGGAACGCCGTGAACATGCAAGAAACGTACACTGCCTACCTCATCGAG ACTCGGTCAGTTGAGCATGCCGATGGTCAGAGTGTGCTCACAGACTCGCTGTGGAGGCGGTACAGTGAGTTCGAGTTGTTGAGAAA CTACCTTCTAGTGTACTACCCACATGTTGTTGTGCCACCTCTCCCAGAAAAGCGGGCAGAGTTCGTGTGGCATAAACTCTCTGCTA CAACATGGACCCAGACTTTGTGGAGAGACGACGCGTGGGCTTAGAAAACTTCCTCTTGAGGGTTGCTTCACATCCTGTCCTTTGTA GAGACAAAATCTTCTATTCATTTTTAACCCAGGAAGGTAACTGGAAGGAGACTGTGAATGAGACTGGATTTCAGCTGAAGGCAGAC TCCAGGTTAAAAGCGCTTAATGCAACATTCAGAGTGAAAAACCCAGACAAGAGGTTTACTGAGCTGAGGCACTACAGTGATGAGCT GCAGTCTGTCATCTCGCATCTCCTTCGAGTCAGAGCTAGAGTAGCAGATCGACTCTATGGTGTATATAAAGTACATGGGAATTATG GGAGAGTTTTTAGTGAATGGAGTGCCATCGAAAAAGAAATGGGGGATGGGCTGCAGAGTGCTGGGCATCACATGGACGTGTATGCA TCTTCTATTGATGATATTTTGGAAGATGAAGAGCACTATGCAGATCAGCTGAAGGAGTATCTGTTTTATGCAGAAGCACTTCGGGC TGTGTGCAGGAAGCATGAGCTTATGCAGTATGACCTGGAGACAGCTGCTCAAGACCTGGCTGCCAAGAAGCAGCAGTGCGAGGAGC TGGCCACCGGGACTGTGAGAACATTCTCGTTGAAGGGAATGACTACCAAACTCTTTGGTCAAGAAACTCCAGAGCAAAGAGAAGCC AGGATAAAGGTGCTAGAGGAGCAGATAAATGAAGGGGAACAGCAGCTGAAGTCTAAAAATCTGGAAGGCAGAGAATTTGTGAAAAA TGCATGGGCTGATATTGAACGCTTCAAAGAACAAAAGAACCGGGACCTAAAGGAAGCTCTCATCAGCTATGCTGTCATGCAGATCA GCATGTGCAAAAAGGGAATTCAGGTTTGGACCAATGCTAAAGAATGCTTCAGCAAGATGTAA

e G GGA ATG ACA ACG AAG CTT (R1) f GCT GAT ATC GAG CGC TTT AAA (R3)

Supplementary Figure S2: Sequences of nucleotides used for knocking down and rescue mouse SNX4. Target sequences of the shRNAs to knock down SNX4: (a) shSNX4-1, (b) shSNX4-2 and (c) shSNX4-3. (d) Sequence of the cDNA of mouse SNX4 used to rescue. The target sequence of shSNX4-1 is highlighted in red and the sequence of shSNX4-3 in yellow. The sequence of shSNX4-2 was design against a region of the 3’ UTR, therefore it is not highlighted in the cDNA and no silence mutagenesis was needed for the rescue. (e) Modified nucleotides by silence mutagenesis in the cDNA of mouse SNX4 to rescue the shSNX4-1 and (f) shSNX4-3. The mutated nucleotides are in bold.

* b 2.0 a **

kDa Control shSNX4-1 1.5 .u. ) 100- TfnR 1.0

55- SNX4 TfnR (a 0.5 40- Actin 0.0

Control shSNX4-1shSNX4-2shSNX4-3

Supplementary Figure S3: SNX4 depletion decreases TfnR levels in HeLa cells but not in neurons. (a) Western blot of HeLa cells infected with control shRNA (Control) and shRNA against SNX4 (shSNX4-1) stained for TfnR, SNX4 and actin. (b) Quantification of TfnR levels measured in proteomics upon SNX4 knock down and control.

92 SNX4 in presynaptic terminals

a Control shSNX4-1 shSNX4-2 shSNX4-3 Synaptophysin-1 Rab11 Merge

b c

1.0 1.0 * ** 0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Syph-1 in Rab11 (M 1) Rab 11 in Syph-1 (M 2) 0.0 0.0 3

Control Control shSNX4-1 shSNX4-2 shSNX4-3 shSNX4-1 shSNX4-2 shSNX4-3

Supplementary Figure S4: SNX4 depletion does not decrease the recycling endosomal marker Rab11 at synapses. (a) Confocal microscopy images of control and SNX4 KD neurons immunolabelled with Synaptophysin-1 and Rab11. Merge image of Synaptophysin-1 (green) and Rab11 (magenta). (n=21±1 neurons, N=2 animals). Scale bar of the neuron image=20 μm, scale bar of the zoomed neurite=4 μm. (b and c) Mander’s coefficients for the co-localization of Synaptophysin-1 and Rab11. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

93 Chapter 3

a b c 0.8 0.8 en t

0.6 0.6 oe ff ici 0.4 0.4

0.2 0.2 VAMP2 ea rs on ’s Coefficient ande r's C P M 0.0 0.0 bassoon bassoon SNX4 in

and SNX4 in SNX4 bassoon

d VGluT1 SNX4 Merge

lortnoC shSNX4-2

e f VGluT1 in SNX4 g SNX4 in VGluT1 0.8 **** 1.0 **** **** 2.0 **** en t en t 0.6 0.8 1.5 oe ff ici oe ff ici 0.6 0.4 1.0 ‘s c ntensity (a.u) ntensity

0.4 i

0.2 ande r's c 0.5 M ea rs on 0.2 SNX4 P

0.0 0.0 0.0

Control Control Control Control shSNX4-2 shSNX4-2 shSNX4-2 shSNX4-2

Supplementary Figure S5: SNX4 is located to synaptic areas. (a) Pearson’s and (b) Mander’s coefficients for the co-localization between Bassoon and SNX4 in neurites (n=25 neurons, N=3 animals). (c) Confocal microscopy images of wild-type neurons overexpressing GFP-SNX4 immunolabelled with VAMP2/Synaptobrevin-2. Merge image of VAMP2/Synaptobrevin-2 (red) and GFP-SNX4 (green) Scale bar=5µm. (d) Confocal microscopy images of hippocampal neurons containing from control and SNX4 KD neurons immunolabelled with VGluT1 and SNX4. Merge image of VGluT1 (green) and SNX4 (magenta). Scale bar of the neuron image=50 μm, scale bar of the

94 SNX4 in presynaptic terminals zoomed neurite=2.5 μm. (b) Pearson and (c) Mander’s coefficients for the co-localization between VGluT1 and SNX4 in neurites. (d) Quantification of total SNX4 levels in the neuron normalized to control. (n=34±1 fields of view, N=3 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

a’’

Supplementary Figure S6: Electron micrographs of the negative controls for immuno-gold labelling against SNX4 (a’, a’’). Negative control processed in parallel with the immunolabelling with SNX4 but preincubating the primary antibody with the blocking peptide at a ratio 10:1, Scale bar=200nm.

a b c 3 Control 50 40 *** 40 30

ud e (pA) 30 shSNX4-1 20 20

Am pli t 10 10 50pA 200ms

0 Frequency (events/s) 0 l

Contro Control shSNX4-1 shSNX4-1

Figure S7: SNX4 depletion impairs spontaneous synaptic vesicle release. (a) Representative traces of spontaneous mini excitatory postsynaptic currents (mEPSC) in Control and in shSNX4-1 neurons. (b) Quantification of the mEPSC amplitude and (c) frequency. (n=25±5 neurons, N=4 animals). Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

95 Chapter 3

b a 1.0 Control synapsin-ECFP 0.8 shSNX4-1 shSNX4-2 baseline 0.6 shSNX4-3

stimulus F/Fmax 0.4 ∆

recovery 0.2

0.0 3 4 5 6 7 8 9 10 11 12 Time (seconds)

Figure S8: SNX4 depletion does not affect calcium in/ex-flux. (a) Representative Fluo-5F fluorescence images of neurites labelled with Synapsin-ECFP during baseline, stimulation, and recovery period. Scale bar=10 µm. (b) Time course of Fluo-5F fluorescence during the imaging protocol, plotted as ΔF/Fmax. (n=28±1 fields of view, N=3 animals).

96 SNX4 in presynaptic terminals

a b c ControlshSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 + shSNX4-3R1 + R2 + R3 kDa ControlshSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 + shSNX4-3R1 + R2 + R3kDa cerebelumcortexhippocampusprefrontalhypothalamus cortexstriatumolfatory bulbs 180 180 130 130 kDa 100 100 70 other 180 70 130 55 100 other 55 SNX4 Actin 70 SNX4 40 40 55 Actin 35 40 35 35 25 25 25 15 15 15

d e f g kDa

HC P2 M SyS SyM PSD kDa Controlother shSNX4-1 Controlother shSNX4-1 Controlother shSNX4-1 kDa 180 180 130 130 180 100 100 130 70 70 TfnR 100 55 55 70 40 Actin TCE 55 SNX4 40 40 35 35 35 25 25 25 15 15 kDa 15 180 130 100 PSD95 70 55 SNX4 40 h i j 35 25 15 VAMP2 Control OverexpressionControl shSNX4-3 Control OverexpressionControlshSNX4-1 shSNX4-3 Control kDa OverexpressionshSNX4-1shSNX4-2shSNX4-3other shRNAother ControlshRNA shSNX4-1shSNX4-2 other shRNAother shRNAkDa shSNX4-2 other shRNAother shRNA

180 180 130 130 100 100 TfnR 70 70 55 55 40 Actin 40 35 35 3 25

k l m n ControlshSNX4-1shSNX4-2shSNX4-1 + R1 ControlshSNX4-1shSNX4-2shSNX4-1 + R1 kDa ControlshSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 + R1shSNX4-3 +Control R2 + R3 180 130 kDa ControlshSNX4-1shSNX4-2shSNX4-3shSNX4-1shSNX4-2 + R1shSNX4-3 +Control R2 + R3 100 70 180 55 130 40 100 70 35

Supplementary Figure S9: Original uncropped blots. (a) Original uncropped blots for SNX4 and actin of the data shown in Figure 1a. The top part was blotted for Fbxo41 which is not relevant for this study. (b) Western blot for SNX4 from Figure 2c. (c) Reblot for actin (low part of the blot) and for a non-relevant antibody (upper part of the blot) from Figure 2c. (d) Gel stained with TCE of the data shown in Figure 4e and original uncropped blots for PSD95, SNX4, and VAMP2/Synaptobrevin-2. The first line is full hippocampal lysate from which the subcellular fractions were obtained. Original uncropped blots from Supplementary Figure S3 for (e) TfnR, for (f) SNX4 and (g) actin. Original uncropped western blot of the data shown in Figure 3k stained for SNX4 (h) and for (i) TfnR and actin and (j) gel stained with TCE. Original uncropped data shown in Figure 9. Stain-free gel with TCE and western blots stained for for LAMP1 (k and l), and GAD1 and GAD2 (m and n).

97 Chapter 3

Supplementary Table S1: Summary of the mean, SEM, n/N numbers and statistics of measured variables in the study. Not applicable (empty cells).

Figure Meassured variable Group Mean ± SEM n/N Statistics p-value cerebelum 0.75 ± 0.46 3 cortex 0.68 ± 0.37 3 hippocampus 0.38 ± 0.16 3 1b SNX4/actin prefrontalcortex 0.25 ± 0.24 2 hypothalamus 0.50 ± 0.40 2 striatum 0.35 ± 0.15 3 olfactory bulb 0.17 ± 0.05 3 Control 1.00 ± 0.15 7 shSNX4-1 0.31 ± 0.08 3 > 0.9999 shSNX4-2 0.28 ± 0.13 3 0.616 H = 23.31, 2c (SNX4/Actin)/Control shSNX4-3 0.41 ± 0.22 3 > 0.9999 p=0.0055 shSNX4-1+R1 1.37 ± 0.51 3 > 0.9999 shSNX4-2+R2 1.99 ± 0.24 3 0.3567 shSNX4-3+R3 1.27 ± 0.26 3 > 0.9999 Control 1.00 ± 0.02 63/4 shSNX4-1 1.00 ± 0.01 49/3 > 0.9999 shSNX4-2 0.78 ± 0.01 43/3 < 0.0001 Synapses per H = 67.91 2f shSNX4-3 0.94 ± 0.02 39/3 0.5025 µm/Control p<0.0001 shSNX4-1+R1 0.99 ± 0.02 40/3 > 0.9999 shSNX4-2+R2 1.08 ± 0.07 37/3 0.2102 shSNX4-3+R3 1.01 ± 0.02 38/3 > 0.9999 Control 1.00 ± 0.02 63/4 shSNX4-1 1.04 ± 0.07 49/3 > 0.9999 shSNX4-2 1.08 ± 0.03 43/3 > 0.9999 Synaptophysin H = 24.46 2g shSNX4-3 0.76 ± 0.05 39/3 0.0649 (a.u.)/Control p=0.0004 shSNX4-1+R1 0.86± 0.09 40/3 > 0.9999 shSNX4-2+R2 1.00 ± 0.05 37/3 > 0.9999 shSNX4-3+R3 0.77 ± 0.10 38/3 0.2651 Control 1.00 ± 0.03 63/4 shSNX4-1 0.61 ± 0.03 49/3 < 0.0001 shSNX4-2 0.52 ± 0.02 43/3 < 0.0001 H = 189.81 2h SNX4(a.u.)/Control shSNX4-3 0.35 ± 0.02 39/3 < 0.0001 p<0.0001 shSNX4-1+R1 1.40 ± 0.10 40/3 0.5914 shSNX4-2+R2 1.35 ± 0.08 37/3 0.2777 shSNX4-3+R3 0.86 ± 0.06 38/3 0.1447 Control 1.00 ± 0.03 63/4 shSNX4-1 0.70 ± 0.04 49/3 0.0003 shSNX4-2 0.48 ± 0.01 43/3 < 0.0001 SNX4 in synaptophsyin H = 142.92 2i shSNX4-3 0.61 ± 0.03 39/3 < 0.0001 (a.u.) /Control p<0.0001 shSNX4-1+R1 2.18 ± 0.28 40/3 > 0.9999 shSNX4-2+R2 1.45 ± 0.11 37/3 0.1275 shSNX4-3+R3 2.13 ± 0.24 38/3 0.4394 Control 1.00 ± 0.03 39/3 U = 430.0, 3b Rab5 (a.u)/Control shSNX4-2 0.81 ± 0.03 42/3 p=0.0013 Control 1.00 ± 0.05 39/3 U = 120.0, 3c SNX4 (a.u)/Control shSNX4-2 0.50 ± 0.02 42/3 p<0.0001

98 SNX4 in presynaptic terminals

Pearson's Rab5 and Control 0.58 ± 0.01 39/3 U = 209.5, 3d SNX4 shSNX4-2 0.41 ± 0.01 42/3 p<0.0001 Mander's M1 Rab5 in Control 0.51 ± 0.02 39/3 U = 255.5, SNX4 shSNX4-2 0.29 ± 0.01 42/3 p<0.0001 3e Mander's M2 SNX4 in Control 0.64 ± 0.01 39/3 U = 413.0 Rab5 shSNX4-2 0.54 ± 0.01 42/3 p<0.0001 Control 1.00 ± 0.02 39/3 U = 538.0, 3g Rab11 (a.u)/Control shSNX4-2 0.93 ± 0.04 37/3 p=0.2194 Control 1.00 ± 0.02 39/3 U = 222.0, 3h SNX4 (a.u)/Control shSNX4-2 0.73 ± 0.04 37/3 p<0.0001 Pearson's Rab11 and Control 0.45 ± 0.01 34/3 U = 195.0, 3i SNX4 shSNX4-2 0.31 ± 0.01 37/3 p<0.0001 Mander's M1 Rab11 in Control 0.39 ± 0.01 39/3 U = 121.0, SNX4 shSNX4-2 0.18 ± 0.01 37/3 p<0.0001 3j Mander's M2 SNX4 in Control 0.49 ± 0.01 39/3 U = 444.5, Rab11 shSNX4-2 0.42 ± 0.03 37/3 p<0.0333 Control 0.99 ± 0.10 7 shSNX4-1 0.60 ± 0.16 3 H = 4.01, 3l Relative TfR shSNX4-2 0.89 ± 0.14 4 p=0.2776 shSNX4-3 0.76 ± 0.36 2 Control 0.96 ± 0.03 4 shSNX4-1 0.32 ± 0.10 3 H = 8.12, 0.2316 3m Relative SNX4 shSNX4-2 0.15 ± 0.07 3 p=0.0155 0.0264 shSNX4-3 0.26 ± < 0.01 2 0.1591 Pearson's Syph-1 and Control 0.71 ± 0.01 36/3 U = 153.0, 4b 3 SNX4 shSNX4-2 0.55 ± 0.01 36/3 p<0.0001 Mander's M1 Syph-1 in Control 0.63 ± 0.01 35/3 t=6.128,df=69, SNX4 shSNX4-2 0.48 ± 0.02 36/3 p<0.0001 4c Mander's M2 SNX4 in Control 0.64 ± 0.02 36/3 t=2.873,df=71, Syph-1 shSNX4-2 0.52 ± 0.02 37/3 p<0.0001 Control 1.00 ± 0.04 36/3 t=5.686,df=70, 4d SNX4 (a.u)/Control shSNX4-2 0.68 ± 0.03 36/3 p<0.0001 P2 4.23 ± 0.93 3 M 0.52 ± 0.26 3 4g (PSD95/TCE) / HC SyS 1.15 ± 0.79 3 SyM 1.00 ± 0.60 3 PSD 13.64 ± 8.55 3 P2 0.82 ± 0.41 3 M 0.75 ± 0.18 3 4h (SNX4/TCE) / HC SyS 1.39 ± 0.66 3 SyM 2.24 ± 1.73 3 PSD 0.02 ± 0.02 3 P2 1.18 ± 0.12 3 M 1.26 ± 0.08 3 4i (VAMP2/TCE) / HC SyS 1.47 ± 0.10 3 SyM 1.64 ± 0.16 3 PSD 0.14 ± 0.01 3

99 Chapter 3

T 1.69 ± 0.17 46/3 4k # gold particles PSD 0.63 ± 0.16 46/3 Control 2766 ± 166 22/3 shSNX4-1 4567 ± 574 22/3 H = 15.07, 0.4302 5b Dendritic length (µm) shSNX4-2 2517 ± 351 22/3 p=0.0017 0.4914 shSNX4-3 2303 ± 448 22/3 0.1011 Control 5631 ± 431 22/3 shSNX4-1 5241 ± 439 22/3 H = 5.72, 5c Axonal length (µm) shSNX4-2 4256 ± 385 22/3 p=0.1259 shSNX4-3 4808 ± 382 22/3 Control 0.32 ±< 0.01 32/3 Synapses per µm shSNX4-1 0.31 ± 0.01 33/3 H = 17.07, 0.4433 5e (Bassoon) shSNX4-2 0.30 ± < 0.01 32/3 p=0.0007 0.1569 shSNX4-3 0.27 ± 0.06 31/3 0.0001 Control 1.00 ± 0.09 32/3 shSNX4-1 0.79 ± 0.07 33/3 H = 13.36, 0.5179 5f Bassoon shSNX4-2 0.87 ± 0.04 32/3 p=0.0039 > 0.9999 shSNX4-3 1.08 ± 0.06 31/3 0.0997 Control 1.00 ± 0.07 22/2 shSNX4-1 0.89 ± 0.07 26/2 H = 0.18, > 0.9999 5g Homer-1 shSNX4-2 0.79 ± 0.05 25/2 p=0.0004 0.1652 shSNX4-3 1.29 ± 0.11 21/2 0.0864 Control 0.32 ±< 0.01 29/2 shSNX4-1 0.31 ± 0.01 30/2 H = 2.23, 5h Synaptophysin-1 shSNX4-2 0.30 ± 0.01 22/2 p=0.5252 shSNX4-3 0.27 ±< 0.01 25/2 Control 1.00 ± 0.05 29/2 shSNX4-1 0.67 ± 0.04 30/2 H = 22.93, 0.0002 5i Synaptotagmin-1 shSNX4-2 1.14 ± 0.11 22/2 p<0.0001 > 0.9999 shSNX4-3 0.99 ± 0.06 25/2 > 0.9999 Control 1.00 ± 0.03 70/5 shSNX4-1 0.41 ± 0.03 62/5 H = 88.73, < 0.0001 5j VAMP2/Synaptobrevin shSNX4-2 0.96 ± 0.07 68/5 p<0.0001 0.2004 shSNX4-3 1.46 ± 0.11 62/5 0.4519 Control 1.00 ± 0.04 25/2 shSNX4-1 0.52 ± 0.03 22/2 H = 35.73, < 0.0001 5k Syntaxin-1 shSNX4-2 0.81 ± 0.05 24/2 p<0.0001 0.1272 shSNX4-3 0.64 ± 0.04 22/2 < 0.0001 Control 463.4 ± 16.3 162/3 shSNX4-1 605.9 ± 21.2 156/3 H = 45.82, <0.0001 6b Active zone length (µm) shSNX4-2 449.0 ± 14.0 154/3 p=<0.0001 >0.9999 shSNX4-3 453.8 ± 15.3 151/3 >0.9999 Control 104.3 ± 5.5 162/3 # Synaptic shSNX4-1 200.4 ± 10.3 156/3 H = 81.92, <0.0001 6c vesicles/synapse shSNX4-2 121.3 ± 7.6 154/3 p=<0.0001 0.7435 shSNX4-3 108.2 ± 6.5 151/3 >0.9999

100 SNX4 in presynaptic terminals

Control 0.083 ± 0.004 162/3 # docked synaptic shSNX4-1 0.062 ± 0.004 156/3 H = 15.24, 0.0003 6d vesicles/synapse shSNX4-2 0.072 ± 0.004 154/3 p=0.0016 0.1189 shSNX4-3 0.077 ± 0.004 151/3 0.3055 Control 3.00 ± 0.30 162/3 shSNX4-1 1.95 ± 0.25 156/3 H = 33.46, 0.0059 6e #Big vesicles/synapse shSNX4-2 2.53 ± 0.25 154/3 p=<0.0001 >0.9999 shSNX4-3 4.55 ± 0.47 151/3 0.0184 Control 0.64 ± 0.09 162/3 # membrane shSNX4-1 0.62 ± 0.08 156/3 H = 4.05, 6f tubules/synapse shSNX4-2 0.85 ± 0.12 154/3 p=0.2558 shSNX4-3 1.05 ± 0.16 151/3 Control 0.060 ± 0.018 162/3 fraction of synapses shSNX4-1 0.070 ± 0.020 156/3 H = 1.26, 6g with MVB shSNX4-2 0.084 ± 0.024 154/3 p=0.7384 shSNX4-3 0.099 ± 0.026 151/3 Control 6230 ± 967 20/3 shSNX4-1 4494 ± 434 18/3 H = 0.57, 7c Fmax (a.u.) shSNX4-2 5511 ± 734 24/3 p=0.9010 shSNX4-3 5674 ± 929 18/3 Control 2610 ± 224 20/3 shSNX4-1 1819 ± 66 18/3 H = 9.04, 0.0369 7d F Baseline (a.u) shSNX4-2 2428 ± 204 24/3 p=0.0286 > 0.9999 shSNX4-3 2550 ± 334 18/3 > 0.9999 Control 0.13 ± 0.01 20/3 3 1st peak amplitud shSNX4-1 0.04 ±< 0.01 18/3 H = 42.19, < 0.0001 7e (∆F/Fmax) shSNX4-2 0.08 ± < 0.01 24/3 p=<0.0001 0.0202 shSNX4-3 0.06 ± 0.01 18/3 < 0.0001 Control 0.87 ± 0.02 20/3 shSNX4-1 0.41 ± 0.05 18/3 H = 33.79, < 0.0001 7f % Active synapses shSNX4-2 0.63 ± 0.06 24/3 p=<0.0001 0.0126 shSNX4-3 0.46 ± 0.05 18/3 < 0.0001 Control 1.11 ± 0.03 20/3 Ratio peak amplitud shSNX4-1 1.09 ± 0.10 18/3 H = 0.19, 7g (2nd/1st) shSNX4-2 1.07 ± 0.06 24/3 p=0.9783 shSNX4-3 1.06 ± 0.07 18/3 Control 1713 ± 80 20/3 shSNX4-1 1413 ± 38 18/3 H = 24.07 0.0012 7h F pH = 5.5 (a.u.) shSNX4-2 1515 ± 55 24/3 p=<0.0001 0.0411 shSNX4-3 1958 ± 147 18/3 > 0.9999 Control 3160 ± 479 25/3 shSNX4-1 3546 ± 777 16/2 shSNX4-2 2683 ± 693 10/2 shSNX4-3 4472 ± 853 16/2 H = 9.11 7k Fmax (a.u.) Overexpression 3429 ± 605 26/3 p=0.2445 sh-1+R1 29 07 ± 411 17/2 sh-2+R2 3540 ± 712 22/2 sh-3+R3 3466 ± 604 17/2

101 Chapter 3

Control 1591 ± 96 25/3 shSNX4-1 1688 ± 155 16/2 shSNX4-2 1622 ± 191 10/2 shSNX4-3 1985 ± 216 16/2 H = 12.09 7l F Baseline (a.u.) Overexpression 1552 ± 92 26/3 p=0.0976 sh-1+R1 1454 ± 56 17/2 sh-2+R2 1850 ± 226 22/2 sh-3+R3 1676 ± 113 17/2 Control 0.20 ± 0.01 25/3 shSNX4-1 0.03 ± < 0.01 16/2 < 0.0001 shSNX4-2 0.07 ± 0.01 10/2 0.0009 1st peak amplitud shSNX4-3 0.08 ± 0.01 16/2 H = 95.29 0.0001 7m (∆F/Fmax) Overexpression 0.17 ± 0.01 26/3 p<0.0001 > 0.9999 sh-1+R1 0.03 ± < 0.01 17/2 < 0.0001 sh-2+R2 0.09 ± 0.01 22/2 0.0001 sh-3+R3 0.08 ± < 0.01 17/2 0.0005 Control 1.00 ± 0.00 4 shSNX4-1 1.61 ± 0.36 4 shSNX4-2 1.56 ± 0.46 4 H = 9.01 9c LAMP1 (a.u.) shSNX4-3 2.35 ± 0.81 3 p=0.1730 shSNX4-1+R1 0.65 ± 0.16 3 shSNX4-2+R2 0.71 ± 0.10 3 shSNX4-3+R3 1.53 ± 0.70 2 Control 1.00 ± 0.04 7 shSNX4-1 0.52 ± 0.03 7 H = 10.83, 9g GAD (a.u.) shSNX4-2 0.81 ± 0.05 6 p=0.0935 shSNX4-1+R1 0.64 ± 0.04 5 Control 0.58 ± 0.02 20/2 Mander's M1 Syph-1 in shSNX4-1 0.57 ± 0.02 20/2 H = 11.87, > 0.9999 S4b Rab11 shSNX4-2 0.61 ± 0.02 19/2 p=0.0078 > 0.9999 shSNX4-3 0.49 ± 0.02 19/2 0.0100 Control 0.61 ± 0.02 20/2 Mander's M2 Rab11 in shSNX4-1 0.53 ± 0.02 20/2 F(3,74) = 1.43, 0.0885 S4c Syph-1 shSNX4-2 0.64 ± 0.02 19/2 p=0.0003 0.3692 shSNX4-3 0.49 ± 0.03 19/2 0.0047 S5a Pearson's Bassoon and SNX4 0.48 ± 0.02 25/3 basson in SNX4 0.41 ± 0.02 25/3 S5b Mander's SNX4 in bassoon 0.46 ± 0.02 25/3 Pearson's VGluT1 and Control 0.46 ± 0.29 35/3 t=6.61,df=66, S5e SNX4 shSNX4-2 0.29 ± 0.02 33/3 p<0.0001 Mander's M1 VGluT1 in Control 0.57 ± 0.02 35/3 U = 230.0, SNX4 shSNX4-2 0.39 ± 0.02 33/3 p<0.0001 S5f Mander's M2 SNX4 in Control 0.23 ± 0.01 35/3 U = 211.5, VGluT1 shSNX4-2 0.13 ± 0.01 33/3 p<0.0001 Control 1.00 ± 0.05 35/3 t=4.19,df=67, S5g SNX4 (a.u)/Control shSNX4-2 0.68 ± 0.04 33/3 p<0.0001 Control 19.41 ± 1.25 30/4 U = 308.0, S8b Amplitude (pA) shSNX4-1 19.28 ± 0.74 24/4 p=0.3713 Control 7.68 ± 1.71 30/4 U = 159, S8c Frecuency (events/s) shSNX4-1 1.86 ± 0.66 24/4 p=0.0006

102 SNX4 in presynaptic terminals

Supplementary Table S2: Avaliable on request with the first authour [email protected] or on https://docs.google.com/spreadsheets/d/1GR_gOjWdhYVkhNCE30JnuuV4FJ1ZO288Yw- K7NaruH0/edit?usp=sharing. Summary of quantified proteins with high confidence in each experimental group and its statistical analysis. Q&T list Q&T ARHGAP23 GAD1, GAD2 EIF3L, EIF3D EIF3L, EIF3D SARS, PRMT5, RUVBL2 ACSBG1, CNTNAP1,ACSBG1, GAD1, SLC38A3 WRN, NSF, SLC12A5, PRKCG, GAD1, GAD2, SLC38A3 GAD1, GAD2, HOMER1, CADPS, NPTX1, CACNB4 GAD1, GAD2 ,HOMER1, CADPS, NPTX1, CACNB4 GAD1, GAD2, HOMER1, CADPS, NPTX1, CACNB4 GAD1, GAD2, HOMER1, CADPS, NPTX1, CACNB4 ATP1A2, CYP46A1, DLG1, NRCAM, ACSBG1, LGI1, MAL2, SLC25A31, HOMER1, CADPS, WRN, TTYH1, CALU, GAD1, NDUFA10, HSD17B7, GSTM1, GAD2, NSF, DLG1, SLC1A2, PRKCG, GAD1, GAD2, CACNB4 CNTNAP1, MPP2, SLC1A2, SLC12A5, COX6B1, PRKCG, ATP1A2, DLG1, LGI1, MPP2, SLC1A2, SLC12A5, PRKCG, ATP1A2, DLG1, LGI1, MPP2, SLC1A2, SLC12A5, PRKCG, ATP1A2, DLG1, LGI1, MPP2, SLC1A2, SLC12A5, PRKCG, ATP1A2, DLG1, LGI1, MPP2, SLC1A2, SLC12A5, PRKCG, CASKIN1, NPTX1, COX5A, CACNB4, HAPLN4, SLC38A3, 1 1 1 1 t depth 1 3 5 1 3 1 2 1 6 1 7 1 4 3 4 1 4 2 4 4 t group tf tf BP BP BP BP rea rea MF cor cor keg t type

term ID 3 GO:0001046 KEGG:04727 REAC:888568 REAC:112315 1 0.176 0.079 0.333 Q&T/T 0.4 0.2 CORUM:742 0.4 0.2 CORUM:1097 0.147 0.185 TF:M06173_0 0.261 0.077 0.269 0.167 Q&T/Q 6 2 7 5 3 Q&T 5 2 5 2 Q 23 26 26 18 T 2 9 34 89 0.05 0.0341 10 0.0466 0.0292 27 34 0.0272 0.0341 10 0.00479 1314 34 31 0.912 0.024 TF:M03989_1 p-value 0.000349 178 34 13 0.382 0.073 GO:0098916 0.000373 179 34 13 0.382 0.073 GO:0099536 0.000687 0.000373 179 34 13 0.3820.000349 178 0.073 34 GO:0099537 13 0.382 0.073 GO:0007268 Headers descriptions are as follows: t name (term domain and name), p-value (enrichment p-value), T (total genes t name match class: 1 class: match match class: 0 class: match GABA synthesis synaptic signaling GABAergic synapse GABAergic trans-synaptic signaling EIF3S10, EIF3S8, EIF3S1, EIF3S7) chemical synaptic transmission synaptic chemical EIF3S10, EIF3S8, EIF3S1, EIF3S7, PCID1) eIF3 complex (EIF3S6, EIF3S5, EIF3S4, eIF3 complex (EIF3S6, EIF3S5, EIF3S4, Factor: motif: KGGCGGAAGM; ZNF85;

EIF3S3, EIF3S6IP, EIF3S2, EIF3S9, EIF3S12, anterograde trans-synaptic signaling trans-synaptic anterograde EIF3S3, EIF3S6IP, EIF3S2, EIF3S9, EIF3S12, core promoter sequence-specific DNA DNA sequence-specific promoter core

Factor: FLI1; motif: ACCGGAAATCCGGT;

Transmission across Chemical Synapses Chemical across Transmission

downregulated proteins downregulated proteins upregulated Supplementary Table S3: Supplementary g:ProfilerTable output for significantly downregulated and upregulatedproteins in neurons expressing the three shRNA against SNX4. associated to a functional term), Q (number of genes in input list), Q&T (fraction of (number genes in of the list genes with functional in term), Q&T/T fraction the of all list functional genes associated detected in to the list), functional term ID, t term), type (term Q&T/Q type or GO and Q&T list (genes in the associated to functional term). t depth (term in local hierarchy) (term group), domain), t group

103 Chapter 3

Supplementary Table S4: Specific tissue, neuronal density, substrate and plate format used in each experiment.

Figure Tissue Density (cell/mL) Substrate Plate format 1c hippocampi 1,300 astrocyte micro-islands 12 2b-c cortices 150,000 coated 6 2d-i hippocampi 70,000 coated 24 3a-j hippocampi 70,000 coated 24 3k-m cortices 150,000 coated 6 4a-d hippocampi 70,000 coated 24 5a-c hippocampi 1,300 astrocyte micro-islands 12 5d-k hippocampi 70,000 coated 24 6 hippocampi 25,000 astrocyte layer 12 7 hippocampi 25,000 astrocyte layer 12 8 cortices 250,000 coated 6 9 cortices 150,000 coated 6 S1b-g cortices 150,000 coated 6 S4 hippocampi 70,000 coated 24 S5a-c cortices 25,000 astrocyte layer 12 S5d-g hippocampi 70,000 coated 24 S8 cortices 1,300 astrocyte micro-islands 12 S9 hippocampi 25,000 astrocyte layer 12

Supplementary Table S5: Primary antibodies specifications and concentrations

Protein Company Catalog number WB ICC iEM Actin Chemicon MAB1501 1 : 10000 Bassoon Enzo Life Science SAP7F407 1 : 500 GAD1 and GAD2 Abcam ab11070 1: 1000 Homer-1 Synaptic Systems 160 004 1 : 200 1 : 300 LAMP1 Cell Signaling #3243 1: 1000 MAP2 Abcam ab5392 1 : 20000 PSD95 SynapticSystems 124 011 1 : 1000 Rab11 BD Transduction lab 610656 1 : 200 Rab5 Transduction labs 610281 1 : 50 SMI-312 Abcam ab24574 1 : 5000 SNX4 SynapticSystems 392 003 1 : 1000 1 : 500 1 : 100 SNX4 Santa Cruz sc-271403 1 : 500 SNX4 Sigma HPA005709 1 : 500 Synaptophysin-1 SynapticSystems 1011004 1: 1000 1 : 1000 Synaptotagmin-1 Verhage laboratory 1 : 1000 Syntaxin-1 Sudhof laboratory 1 : 1000 Tranferrin receptor Zymed 136800 1: 1000 VAMP2/synaptobrevin SynapticSystems 104 211 1 : 5000 1 : 1000 VGluT1 Millipore AB5905 1 : 5000

104 Chapter 4 The seeding of tau pathology alters the endolysosomal system

105 Chapter 4

The seeding of tau pathology alters the endolysosomal system

Sonia Vazquez-Sanchez1, Vera Wiersma1,2, Jeroen Kole3, Maarten P. Bebelman4, Rozemarijn Jongeneel1, Myrthe Flesseman1, Wiep Scheper1,2,5, and Jan R.T. van Weering 1,2, *

1Department of Functional Genomics, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, Vrije Universiteit (VU), Amsterdam, Netherlands

2Clinical Genetics, Center for Neurogenomic and Cognitive Research, Neuroscience Campus Amsterdam, Amsterdam UMC, Amsterdam, Netherlands

3Laboratory for Physiology, Institute for Cardiovascular Research, VU University Medical Center, Amsterdam, Netherlands.

4Department of Pathology, Cancer Center Amsterdam, VU University Medical Center, Amsterdam, The Netherlands; Division of Medicinal Chemistry, Amsterdam Institute for Molecules Medicines and Systems, VU University, Amsterdam, The Netherlands.

5 Alzheimer Center, Amsterdam UMC, Amsterdam, Netherlands

*Corresponding author: Jan R.T. van Weering, Center for Neurogenomics and Cognitive Research, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. Email: [email protected]

106 The seeding of tau pathology alters the endolysosomal system

ABSTRACT

Tau pathology and endolysosomal alterations co-occur in brains of Alzheimer’s Disease and pure tauopathy patients, but the relationship between these two pathological features is currently not understood. In vitro, tau pathology can be modeled by seeding with recombinant tau fibrils on human tau expressing cells, recapitulating tau hyperphosphorylation, misfolding and aggregation. The present study addressed how the seeding of tau pathology impacts the endolysosomal system in HEK293 cells, primary neurons and iPSC-derived human neurons. This study shows that the seeding of tau pathology induces abnormalities in the endolysosomal system of various cell models. In iPSC-derived human neurons, the seeding of tau pathology decreased the number, size and EEA1 labelling intensity of EEA1-positive early endosomes. While the seeding of tau pathology did not cause morphological changes in CD63-enriched late endosomes and LAMP1-labelled lysosomes, the data did indicate decreased proteolytic activity of these compartments. Thus, the seeding of tau pathology is causal to changes in the endolysosomal system, which may play a role in neurodegeneration.

4

107 Chapter 4

INTRODUCTION

Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by cognitive decline and memory loss (Alzheimer’s Association, 2018). Neuropathologically, AD is characterized by the deposition of pathological tau and amyloid-β proteins, which both seem to spread through the brain in a stereotypical pattern (Braak and Braak, 1991). Cognitive decline of AD patients correlates more strongly with tau pathology load than amyloid-β deposition (Aschenbrenner et al., 2018; Braak and Braak, 1991). Tau pathology is characterized by tau hyperphosphorylation, misfolding and aggregation, and neurofibrillary tangle formation (Ballatore et al., 2007). Tau pathology spreads cell-to-cell in a ‘prion-like’ manner in which pathological tau from an affected cell (seed) can template tau in another cell inducing tau pathology (Guo and Lee, 2014).

AD is also neuropathologically characterized by endolysosomal aberrations (Cataldo et al., 1997; Colacurcio et al., 2017). One of the first cellular symptoms observed in early AD brains is endosome swelling (Cataldo et al., 2000). In AD brains, the early endosomal protein Rab5 and the late endosomal protein Rab7 are increased both at mRNA and proteins levels (Ginsberg et al., 2010a; Ginsberg et al., 2010b), and the early endosomal phospholipid PI(3)P is decreased (Morel et al., 2013). The lysosomal enzyme Cathepsin D is also upregulated at mRNA and protein level in AD brains (Cataldo et al., 1995). Particularly, pro- and mature cathepsins B and D levels are increased within the enlarged endosomes (Cataldo et al., 1997). Immature autophagic vacuoles also accumulate in dystrophic AD neurites which suggests that their maturation to lysosomes is impaired (Nixon et al., 2005). AD neuropathology is also characterized by granulovacuolar degeneration (GVD) bodies in neurons affected by tau pathology (Köhler, 2016). The charged multivesicular body protein 2B (CHMP2B) localizes to the core of the GVD and the lysosome-associated membrane protein 1 (LAMP1) is surrounding the GVD core (Funk et al., 2011). This suggests that GVDs are, at least in part, of endolysosomal origin, and thus that endolysosomal trafficking might be affected in AD. Hence, aberrations at different levels of the endolysosomal system have been reported in AD brains. Notably, endolysosomal genes have also been associated with higher risk of AD (Gao et al., 2018; Karch and Goate, 2015; Naj et al., 2017; Rogaeva et al., 2007).

The Cathepsin D upregulation in AD brain positively correlates with tau pathology and neurofibrillary tangles (Chai et al., 2018). Similar to observations in AD brains, cathepsin D shows an abnormal subcellular distribution and LAMP1 accumulates in brains of primary tauopathies patients (Piras et al., 2016). Furthermore, GVD is a characteristic hallmark of various pure tauopathies (Nijholt et al., 2012). In addition, endosomal genes such as STX6 (involved in trans-Golgi-endosome fusion) have been associated with higher risk

108 The seeding of tau pathology alters the endolysosomal system

b a Control K18 Control K18 Rab5 EEA1 TauP301L-GFP TauP301L-GFP Merge + DAPI Merge + DAPI

c 2000 d 0.4 *** e 2.0 ) 2

1500 0.3 1.5

1000 0.2 1.0

# EEA1 puncta # EEA1 500 0.1 0.5 EEA1 puncta area (µm EEA1 EEA1 puncta intensity (a.u.) intensity puncta EEA1 0 0 0 Control K18 Control K18 Control K18 f 800 g 0.4 **** h 1.5 **** *** ) 2

600 0.3 1.0

400 0.2

0.5 # Rab5 puncta # Rab5 200 0.1 Rab5puncta area (µm Rab5 puncta intensity (a.u.) intensity puncta Rab5 0 0 0 4 Control K18 Control K18 Control K18

Figure 1: Seeding of tau pathology reduces the area of EEA1 and Rab5 puncta in HEK293 cells. (a) Confocal microscopy images of control and HEK293 cells with tau aggregates immunolabelled for EEA1. Merge image of EEA1 (magenta), TauP301L-GFP (green) and DAPI (blue). n=30±4 fields of view, N=3 experiments. Scale bar=20 μm. (b) Confocal microscopy images of control and HEK293 cell with tau aggregates immunolabelled for Rab5. Merge image of Rab5 (magenta), TauP301L-GFP (green) and DAPI (blue). n=18±1 fields of view, N=1 experiment. Scale bar=20 μm. (c) Number, (d) area and (e) normalized intensity of EEA1 puncta. (f) Number, (g) area and (h) normalized intensity of Rab5 puncta. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

109 Chapter 4

Control K18 a b c d

* 2.0 ** 1.5

20 ) 2 CD63 15 1.5 1.0

10 1.0

0.5 0.5

#CD63 puncta 5 CD63 puncta area (µm CD63 LAMP1 CD63puncta intensity (a.u.)

0 0.0 0.0 e Control K18 f Control K18 g Control K18 3 2.0 10 *** * ** ) 2

8 1.5 2 TauP301L-GFP 6 1.0 4 1 0.5 2 # LAMP1 puncta LAMP1 puncta area (µm Merge + DAPI 0 0 (a.u.) puncta intensity LAMP1 0 Control K18 Control K18 Control K18

LAMP1 h i CD63 in LAMP1 in CD63 **** 0.8 1.0 * **

0.8 0.6

0.6 0.4 0.4

0.2 coeficients Mander's

Pearson Correlation 0.2

0 0 Control K18 K18 K18 Control Control

Figure 2: Seeding of tau pathology increases the amount and area of CD63 and LAMP1 puncta in HEK293 cells. (a) Confocal microscopy images of control and HEK293 cell with tau aggregates immunolabelled for CD63 and LAMP1. Merge image of CD63 (cyan), LAMP1 (magenta), TauP301L- GFP (green) and DAPI (blue). Arrowheads indicate examples of colocalization between CD63 and LAMP1. n=45±1 fields of view, N=3 experiments. Scale bar=20 μm. (b) Number, (c) area and(d) normalized intensity of CD63 puncta. (e) Number, (f) area and (g) normalized intensity of LAMP1 puncta. (h) Pearson and (i) Mander’s coefficients for the co-localization between CD63 and LAMP1. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1. of the primary tauopathy progressive supranuclear palsy (PSP) (Höglinger et al., 2011). Together, these studies suggest that tau pathology and endolysosomal dysfunction are coupled, but the mechanistic connection between these two cellular phenotypes is currently not understood.

110 The seeding of tau pathology alters the endolysosomal system

This study examines how the seeding of tau pathology impacts the endolysosomal system. To address this question, the seeding of tau pathology was modeled in vitro using an assay in which recombinant pre-aggregated tau seeds induce tau pathology in HEK293 cells, in primary mouse neurons and in iPSC-derived human neurons. This model recapitulates the main hallmarks of tau pathology as observed in AD and pure tauopathies (tau hyperphosphorylation, misfolding and insoluble aggregation), making it a powerful tool to assess effects induced by the seeding of tau pathology. In these cellular models, the size, number and labelling intensity of the endolysosomal compartments was quantified using confocal microscopy and three-dimensional super-resolution stimulated emission depletion microscopy (3D-STED). This study shows that the seeding of tau pathology induces abnormalities in the endolysosomal system in various cell types.

RESULTS

The seeding of tau pathology alters the endolysosomal system in HEK293 cells

We first assessed the impact of the seeding of tau pathology on the endolysosomal system in an established HEK293 cell model of seeded tau pathology, in which human TauP301L- GFP is overexpressed and recombinant TauP301L K18 tau fibrils are added. K18 consists of the four microtubule binding repeats (residues Q244–E372 of the longest Tau isoform) which have the highest aggregation tendency (Mukrasch et al., 2005). The P301L mutation is causative of inherited frontotemporal dementia (FTD) tauopathy and it makes tau more prone to aggregation (von Bergen et al., 2001). Upon lipofectamine-mediated transduction of K18 fibrils, HEK293 cells overexpressing TauP301L-GFP developed significantly more pathological tau features, including pathological phosphorylation, misfolding and aggregation, compared to control buffer treated cells (Supplementary Figure S1), in line with previous studies (Guo and Lee, 2011).

Early endosome antigen 1 (EEA1) and Rab5 were used as canonical markers to label early 4 endosomes. Upon K18 treatment, the number of EEA1 puncta was not changed while the number of Rab5 puncta was 63% decreased. K18 treatment reduced the area of both EEA1 puncta and Rab5 puncta by 16%. EEA1 labelling intensity was not changed upon K18 treatment but Rab5 puncta labelling was 2% increased (Figure 1, Supplementary Figure S2). CD63 and LAMP1 were used to label late endosomes and lysosomes. Upon K18 treatment, the number and area of CD63 puncta was increased (198% and 11%, respectively), but CD63 puncta labelling intensity was not changed (Figure 2a-d). The number and area of LAMP1 puncta was also increased upon K18 treatment by 195% and 7% respectively, while the LAMP1 puncta labelling intensity was 11% decreased (Figure 2a, e-g). The colocalization between CD63 and LAMP1 was also increased upon

111 Chapter 4

K18 treatment measured with both Pearson’s correlation and Mander’s coefficients M1 and M2 (30%, 26%, 15% respectively; see Table S1 for detailed information) (Figure 2a, h-i). In conclusion, in HEK293 cells, the seeding of tau pathology reduces the size of early endosomes while increasing the size and number of late endosomes and lysosomes and increases the colocalization of CD63 and LAMP1, indicating a change in the molecular composition or maturation of the late endosomal compartments. LIMPII a LIMPII b DAPI DAPI TauP301L-GFP LIMPII TauP301L-GFP LIMPII TauP301L-GFP Control Control K18 K18

c d e

8000 5 2.5 ) 2

µ m 2.0 6000 4

3 1.5 4000 2 1.0

# LIMPII puncta LIMPII # 2000 1 0.5 LIMPII puncta area ( 0 0 0 Control K18 Control K18 puncta intensity (a.u.)LIMPII Control K18

Figure 3: Seeding of tau pathology does not affect the amount, area and intensity of LIMPII puncta in primary mouse neurons. (a) Confocal microscopy images of hippocampal neurons overexpressing TauP301L-GFP treated with control buffer or K18 fibrils, and immunolabelled for LIMPII. Merge image of LIMPII (magenta), TauP301L-GFP (green) and DAPI (blue). n=25±1 fields of view, N=3 experiments. Scale bar=20 μm. (b) Zoom in of (a). Scale bar=10 μm. (c) Number, (d) area and (e) intensity of LIMPII puncta. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1. The seeding of tau pathology does not alter lysosomes in mouse primary neurons

To assess potential changes in the endolysosomal system in a more AD relevant cell type, tau seeding was performed by addition of K18 fibrils to the culture medium of murine post-mitotic primary neurons to induce tau pathology. First, we confirmed that the K18 treatment also resulted in pathological tau phosphorylation, misfolding and aggregation in mouse neurons (Supplementary Figure S3), in line with previous rapports (Guo and Lee, 2013). The antibodies used in the previous experiments in HEK293 cells did not recognize

112 The seeding of tau pathology alters the endolysosomal system the mouse epitopes. Instead, lysosomal integral membrane protein 2 (LIMPII) was used to label lysosomes. K18 treatment did not affect the number, size and labeling intensity of LIMPII puncta in primary neurons containing TauP301L-GFP accumulations (Figure 3). Hence, the seeding of tau pathology did not morphologically affect LIMPII-positive lysosomes in these primary neurons.

EEA1 EEA1 EEA1 a TauP301L-GFP b TauP301L-GFP c puncta EEA1 MAP2 + DAPI EEA1 MAP2 + DAPI reconstruction Control Control Control K18 K18 K18

d e f 150 * 0.8 * 4 **** ) 3

0.6 3 100

0.4 2

50 # EEA1 puncta # EEA1 0.2 1 EEA1puncta volume (µm EEA1puncta EEA1 puncta intensity (a.u.) intensity puncta EEA1 0 0 0 Control K18 Control K18 Control K18

Figure 4: Seeding of tau pathology reduces the number, volume and labelling intensity of EEA1 puncta in human neurons. (a) Confocal microscopy images of human neurons overexpressing TauP301L-GFP treated without and with K18 fibrils, and immunolabelled for EEA1 and MAP2. Merge image of EEA1 (magenta), TauP301L-GFP (green), and DAPI and MAP2 (blue). Scale bar=10 μm. (b) Maximum intensity projection of 3D-STED microscopy z-stacks of the area squared in (a). Scale bar=1 μm. (c) 3D reconstruction of the EEA1 puncta squared in (b). (d) Number, (e) volume and (f) 4 normalized intensity of EEA1 puncta. n=45±1 fields of view, N=4 experiments. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

The seeding of tau pathology affects the endolysosomal system in human neurons

As differences between mouse and human species could affect tau pathology and endolysosomal trafficking, we adapted the tau pathology seeding assay to iPSC- derived human neurons, a state-of-the art cell model which is more relevant to study human disease. Upon treatment with K18 seeds in the culture medium, TauP301L- GFP expressing human neurons showed tau pathology features such as an increase in tau pathological conformation (Supplementary Figure S4a, d), as previously published

113 Chapter 4

(Verheyen et al., 2015). However, K18 treatment did not increase AT100 labelling which detects phosphorylation of Thr212 and Ser214 (Fig S4b, e). To accurately calculate the volume of different endolysosomal compartments, we implemented 3D-STED, which

114 The seeding of tau pathology alters the endolysosomal system enhances both X, Y and Z resolution, to characterize early endosomes (EEA1 enriched), late endosomes (CD63 enriched) and lysosomes (LAMP1 enriched). Upon K18 treatment, number, volume and labelling intensity of EEA1 punctae was decreased by 28%, 27% and 52% respectively (Figure 4, Supplementary Figure S5). The levels of endosomal sorting proteins SNX4 and VPS35 were 56% and 22% increased (Supplementary Figure S6). The number, volume and intensity labelling of CD63 and LAMP1 puncta was not changed by K18 treatment, nor the colocalization between CD63 and LAMP1 (Figure 5).

Although lysosomes were not morphologically different, their degradative capacity could be affected by the seeding of tau pathology. Human neurons were treatedTM withDQ Red-BSA prior to fixation which produces a bright fluorescent product when hydrolyzed in proteolytically active compartments, mainly the lysosomes. Upon K18 treatment, the intensity of DQTM Red-BSA inside LAMP1-positive puncta was 38% decreased and the number, area and intensity of DQTM Red-BSA puncta was decreased by 68%, 13% and 16%, respectively (Figure 6). Taken together, the seeding of tau pathology in human neurons decreased the number, size and EEA1 labelling intensity of early endosomes, and appeared to decrease lysosomal proteolytic activity without affecting the morphology or labelling intensity of late endosomes and lysosomes.

DISCUSSION

This study shows that the seeding of tau pathology induces abnormalities in the endolysosomal system. In HEK293 cells, the seeding of tau pathology reduced the size of early endosomes, while increasing the size and number of late endosomes and lysosomes. Furthermore, the colocalization of CD63 and LAMP1 was enhanced, suggesting a change in the maturity or molecular composition of the endolysosomal compartments. In mouse primary neurons, the seeding of tau pathology did not alter lysosome number and size. In iPSC-derived human neurons, the seeding of tau pathology decreased the number, size and EEA1 labelling intensity of early endosomes, but did not morphologically change 4 late endosomes and lysosomes. However, lysosomal proteolytic activity appeared to be

Figure 5: Seeding of tau pathology does not impact the number, volume and labelling intensity of CD63 and LAMP1 puncta in human neurons. (a) Confocal microscopy images of human neurons overexpressing TauP301L-GFP untreated and treated with K18 fibrils, and immunolabelled for CD63, LAMP1 and MAP2. Merge image of CD63 (magenta), LAMP1 (cyan), TauP301L-GFP (green), and DAPI and MAP2 (blue). Scale bar=5 μm. (b) Maximum intensity projection of 3D-STED microscopy z-stacks of the area squared in (a) and 3D reconstruction of the CD63 (white) and LAMP1 (yellow) puncta squared in the 3D-STED images. Scale bar=1 μm. (c) Number, (d) volume and (e) normalized intensity of CD63 puncta. (f) Number, (g) volume and (h) normalized intensity of LAMP1 puncta. (i) Pearson and (j) Mander’s coefficients for the co-localization between CD63 and LAMP1. n=34±3 fields of view, N=3 experiments. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

115 Chapter 4 reduced in human neurons upon the seeding of tau pathology (see summary Table 1).

Endolysosomal organelles are a continuum rather than separate compartments

A source of uncertainty in this study comes from the discrimination of endosolysosomal compartments based on molecular markers. This approach is widely used to study the endolysosomal system; however, it is important to note that endolysosomal compartments are not separate entities. Instead, they are extremely dynamic structures with a constant exchange of material and with highly overlapping features. EEA1 and Rab5 puncta are interpreted as early endosomes since these proteins are preferentially in early endosomes where they mediate early endosomal fusion (Christoforidis et al., 1999a; Stenmark, 2009). CD63 is found in the limiting membrane of the lysosome among others membranes but it is used as a late endosomal/multivesicular body (MVB) marker

a MAP2 + DAPI TauP301L-GFP DQ-BSA LAMP1 Merge Control K18

b c d e

* ) * 2.0 *** 500 *** 2.5 2 3

400 2.0 1.5 2 300 1.5 1.0 200 1.0 1

0.5 puncta # DQ-BSA 100 0.5 DQ-BSA puncta area (µm DQ-BSA

0 0 (a.u.) intensity puncta DQ-BSA 0 0 BQ-BSA in LAMP1 puncta (a.u.) BQ-BSA Control K18 Control K18 Control K18 Control K18

Figure 6: Seeding of tau pathology reduces the number, size and intensity of DQ-BSA puncta in human neurons. (a) Confocal microscopy images of human neurons overexpressing TauP301L- GFP treated without and with K18 fibrils, treated with DQ-BSA and immunolabelled with LAMP1 and MAP2. Merge image of DAPI and MAP2 (blue), TauP301L-GFP (green), DQ-BSA (magenta) and LAMP1 (cyan). Arrowheads indicate colocalization of DQ-BSA and LAMP1. Scale bar=10 μm. (b) Normalized intensity of DQ-BSA inside LAMP1 puncta. n=41±1 fields of view, N=3 experiments (c) Number, (d) area and (e) normalized intensity of DQ-BSA puncta. n=32±8 fields of view, N=3 experiments. Detailed information (average, SEM, n and statistics) is shown in Supplementary Table S1.

116 The seeding of tau pathology alters the endolysosomal system because it is seven-fold higher on ILVs in the MVB than in the limiting membrane (Pols and Klumperman, 2009). DQTM-BSA was used to label hydrolytic active, predominantly endolysosomal, compartments (Bright et al., 2016). For the visualization of lysosomes, we used antibodies against LAMP1 and LIMPII, which are involved in lysosomal integrity (Gonzalez et al., 2014; Saftig and Klumperman, 2009). However, these two proteins are also present on the membrane of late endosomes (Saftig and Klumperman, 2009).

The seeding of tau pathology differently affects the endolysosomal system in HEK293 cells and neurons

The seeding of tau pathology differently affected the endolysosomal system of HEK293 cells and neurons. In HEK293 cells, EEA1 labelled endosomes did not decrease in number or EEA1 labelling intensity but Rab5-positive endosomes did decrease in number, while slightly increasing its Rab5 labelling (Figure1). In human neurons, both the number and labelling of EEA1 endosomes were decreased. Late endosomes and lysosomes were not morphologically affected by the seeding of tau pathology in primary mouse as well as human neurons while these compartments were affected in HEK293 cells. Upon seeding of tau pathology in HEK293 cells, CD63-enriched late endosomes and LAMP1-positive lysosomes were increased in number and size. The differences between neuronal and HEK293 cell results might be explained by both cell type and experimental differences. The endolysosomal system of HEK293 cells might bear more resemblance to that of mitotic brain cells which can develop tau pathology, such us astrocytes (Kovacs et al., 2016), rather than brain post-mitotic neurons. Dividing cells can partition aggregates into daughter cells, while this is not possible for post-mitotic neurons (Guo et al., 2016). Therefore, the endolysosomal system might respond differently to the seeding of tau pathology in mitotic versus post-mitotic cells. In addition, in neurons, K18 fibrils were directly added to the cell medium during 11 days, while in HEK293 cells, K18 fibrils were transduced with Lipofectamine for 4 days to prevent overcrowding of the HEK293 culture. Hence, differences in tau seed internalization and tau pathology acquisition might have 4 different effects on endolysosomal compartments. In line with this, an enlargement of LAMP1/2 labelled lysosomes was observed in a human iPSC-derived neuronal model of tau pathology achieved through the cytosolic expression of tau repeats domains (with P301L and V337M mutations) without the addition of recombinant tau fibrils (Reilly et al., 2017).

The seeding of tau pathology reduces early endosome size

Early endosome size was decreased upon the seeding of tau pathology in both human neurons and HEK293 cells and using both molecular markers EEA1 and Rab5 (Figure 1a, b,

117 Chapter 4 d, g and Figure 4 a, b, c, e). In vitro, tau seeds can enter the cells via endocytosis (Calafate et al., 2016; Frost et al., 2009; Wu et al., 2013). Internalized tau seeds are trafficked through the endosomal pathway showing colocalization with Rab5 puncta (Calafate et al., 2016; Wu et al., 2013). These tau seeds can damage endosomal membranes producing leakage (Calafate et al., 2016; Flavin et al., 2017). Such damage can be sensed by galectins and trigger autophagy to clear the damaged organelles (Papadopoulos and Meyer, 2017), which might contribute to the reduction of early endosomes upon the seeding of tau pathology. In human neurons, the seeding of tau pathology reduced EEA1-puncta both in number and size, while in HEK293 only in size. Lipofectamine transduced seeds used in the HEK293 cells might take several entry routes, while in neurons, the naked seeds enter the cell mainly via endocytosis (Calafate et al., 2016; Frost et al., 2009), which could result in a more severe phenotype. In post-mortem AD brain, early endosome enlargement was reported using both Rab5 and EEA1 as molecular markers (Cataldo et al., 2000), while we find this compartment to be smaller. The different outcome might be due to the endpoint situation with more prolonged and complex pathology in the AD brain

Table 1: Overview of the effect of the seeding of tau pathology on the measured variables compared to control. Significant decrease is note as ‘red arrow’, significant increase as ‘green arrow’, no significant difference is noted as "=", and not applicable as ‘empty cell’. Detailed information (average, SEM, n and statistics) is displayed in Supplementary Table S1.

Change compared to control Parameter HEK293 cells mouse neurons human neurons number Rab5 size ⬇ labelling intensity ⬇ number = ⬆ EEA1 size ⬇ labelling intensity = ⬇ ⬇ number = ⬇ CD63 size = ⬆ labelling intensity = = ⬆ number = = ⬇ LAMP1/LIMPII size = = ⬆ labelling intensity = = ⬆ CD63/LAMP1 colocalization = ⬇ in LAMP1 puncta ⬆ number DQ-BSA ⬇ intensity ⬇ size ⬇ ⬇

118 The seeding of tau pathology alters the endolysosomal system

In AD brain, the levels of the endosomal sorting proteins VPS35 and SNX4 are decreased (Kim et al., 2017; Small et al., 2005), whereas in our human neurons VPS35 and SNX4 were increased upon the seeding of tau pathology. In an AD mouse model (APP/ PS1 transgenic), SNX4 levels were increased in 6-month-old brains but decreased in 24-month-old brains which suggested that SNX4 might be initially upregulated in AD pathogenesis but down regulated in late stages (Kim et al., 2017). Decreasing VPS35 protein in vitro increases Aβ levels (Small et al., 2005) and increasing the stability and levels of the trimeric VPS35–VPS29–VPS26 complex (retromer) decreases Aβ levels. Indeed, increasing retromer stability with pharmacological chaperons has been proposed as a therapeutic target for AD since it alleviates both Aβ and tau pathology (Mecozzi et al., 2014; Young et al., 2018). Therefore, endosomal sorting proteins levels are connected to AD pathogenesis but the cause and consequences of their dysregulation remain largely unclear.

The seeding of tau pathology diminishes proteolytic activity without affecting the morphology of late endolysosomal compartments

In mouse and human neurons, late endolysosomal and lysosomal compartments were not morphologically changed (Figure 3 and 5). However, the seeding of tau pathology in human neurons seemed to reduce proteolytic activity in these compartments (Figure 6). At this stage, we cannot exclude that reduced uptake of DQ-BSA in K18-treated neurons can also explain this phenotype. Given the reduction of early endosomal compartments (Figure 4), the delivery of DQ-BSA signal from early endosomes to LAMP1-compartments could also be impaired. This will result in less proteolytically de-quenched DQ-BSA- fluorescence. That said, the decreased of DQ-BSA in LAMP1 puncta (Figure 6) is also consistent with a model where the seeding of tau pathology impairs lysosomal proteolysis without affecting the overall morphology of LAMP1/LIMPII-positive compartments. Tau aggregates can damage the membrane of late endosomes and lysosomes in vitro (Flavin et al., 2017). Rupture of lysosomal membrane will lead to an increase in lysosomal pH 4 and leakage of the lysosomal content to the cytosol, resulting in decreased proteolytic activity, which would explain the decrease in DQ-BSA signal. This decrease in proteolytic activity might be related with the increased levels of TauP301L-GFP observed in the three cell models upon K18 treatment compared with control (Supplementary Figure S1, S3 and S4). Since leaked cytosolic cathepsins can trigger cell death (Gómez-Sintes et al., 2016), it would also be interesting to investigate if the seeding of tau pathology produces lysosomal leakage and triggers neurodegeneration in AD.

Late endosomal and lysosomal molecular markers such us Rab7 and LAMP1 accumulate in AD brains (Ginsberg et al., 2010b; Piras et al., 2016) as well as in brains with primary

119 Chapter 4 tauopathy (Piras et al., 2016). These measurements were done in brain lysates, while we addressed the effect of seeding of tau pathology on the labelling intensity of neuronal CD63, LAMP1- and LIMPII-positive compartments, where no increase is observed. This discrepancy may be explained by the contribution of other brain cell types present in the full brain lysates of the AD patients. In AD neurons, there is also an accumulation of autophagic vesicles and lysosome-related organelles (Nixon et al., 2005), but we did not observe it in cultured neurons. A diffuse pattern of Cathepsin D is reported in AD and primary tauopathy neuronal phenotypes indicating less degradative compartments (Piras et al., 2016), which is in line with the decrease in proteolytically processed-DQ-BSA compartments shown here. Hence, the seeding of tau pathology does not morphologically change CD63- and LAMP1-enriched compartments but it might decrease their proteolytic activity.

In conclusion, using different models that recapitulate human tau pathology, we showed that the seeding of tau pathology causes changes in the endolysosomal system. Treatment with tau fibrils might directly affect the endolysosomal system, since internalized tau aggregates can damage endosomal membranes producing leakage (Calafate et al., 2016; Flavin et al., 2017). The most prominent alteration was the reduction of early endosomal size, which was observed in both HEK293 cells and human neurons upon different seeding paradigms. The reported endolysosomal damage might reflect the endolysosomal damage produced by the seeding of tau pathology in the human brain which may play a role in loss of proteostasis, tau pathology-induced cell death and neurodegeneration.

MATERIALS AND METHODS

K18 fibrils

Tau-P301L-K18-Myc fibrils (K18) were obtained for research purposes under Material Transfer Agreement (MTA) from Janssen Pharmaceuticals (Beerse, Belgium). K18 synthetic fibrils were formed by in vitro fibrilization with heparin. K18 monomers (40 μM) were incubated 24h at 37°C with 10 μM low-molecular-weight heparin and 2 mM dithiothreitol (DTT) in sodium acetate buffer (100mM, pH=7.0). The stock of 40 µM K18 fibrils was diluted to 10 µM with sodium acetate buffer prior to sonication. An Eppendorf containing 200µL of 10 µM K18 fibrils was placed on ice and sonicated (25 pulses with 3 seconds of interval every 5 pulses). Sonicated K18 tau fibrils were stored at -80°C until usage.

120 The seeding of tau pathology alters the endolysosomal system

HEK293 cell culture

TauP301L-GFP doxycycline-inducible HEK293 cell line was obtained for research purposes under Material Transfer Agreement (MTA) from Janssen Pharmaceuticals (Beerse, Belgium). HEK293 cells were grown in DMEM supplemented with 10% Fetal Bovine Serum (FBS), 100 units/mL of penicillin and 100 µg/mL of streptomycin (P/S) (Gibco). 80% confluent cells were washed with PBS before trypsinization and plated o for each experiment. Cells were maintained at 37 C and 5% CO2 until the day of the experiment.

For confocal imaging, cells were plated on coated coverslips with poly-L-ornithine (PLO, Sigma) and laminin (Sigma) at 20.000 cell/mL with 100ng/mL Doxycycline to express TauP301L-GFP. For the seeding assay, cell medium was changed for DMEM without antibiotics after 1 day in culture. Optimem medium (Gibco), Lipofectamine-2000 (Invitrogen), sodium acetate buffer (for control) or K18 (final concentration 20nM) were incubated for 20 minutes at room temperature previous to addition to the cells. After 4-6 hours incubation, the transduction medium was replaced for normal medium (DMEM with 10%FBS and 1%P/S). Cells were fixed after 5 days in culture.

For sequential extraction and Western blot, cells were plated at 150.000 cells/mL and grown until confluence with 100ng/mL Doxycycline (Control) and without (WT). Tau pathology seeded cells (K18) consisted in already K18 transduced cells with Lipofectamine, which were expanded in culture while keeping 100ng/mL Doxycycline in the cell medium.

Laboratory animals

Animal experiments were approved by the animal ethical committee of the VU University/ VU University Medical Centre (“Dier ethische commissie (DEC)”; license number: FGA 11-03) and, according to institutional and Dutch governmental guidelines and regulations. 4

Primary mouse neuronal culture

Primary neurons were cultured from mouse E18 wild-type brains. Briefly, hippocampi or cortices were dissected in Hanks balance salt solution (HBSS, Sigma) with 10mM HEPES (Life Technologies) and digested by 0.25% trypsin (20 minutes at 37 oC; Life technologies) in HBSS. The tissue dissociation was performed with fire-polished Pasteur pipettes in DMEM with FCS. The neurons were spun down and re-suspended in neurobasal medium with 2% B-27, 18 mM HEPES, 0.25% glutamax and 0.1% P/S (Life Technologies). For confocal imaging, hippocampal neurons were plated on coated coverslips with poly-L-

121 Chapter 4 ornithine (PLO, Sigma) and laminin (Sigma) at 70.000 cells/ml. After three days in vitro (DIV3) cells were infected with hTau-P301L-EGFP under CMV promotor packed in lentiviral particles. At DIV7 K18 fibrils were added to neuronal medium (30nM) or the same volume of sodium acetate buffer to control neurons. At DIV10 medium was refreshed and at DIV18 cells were fixed. For sequential extraction and western blot, cortical neurons were plated at 300.000 cells/mL and tau pathology seeded as described above. Neurons o were maintained at 37 C and 5% CO2 until the day of the experiment.

IPSC-derived human neuronal culture

Neuronal Precursor Cells (NPCs) were generated from induced pluripotent stem cells (iPSCs) as previously described (Israel et al., 2012). NPCs were passaged using accutase and NPC base (DMEM/F12 with GX (Life technologies) with 0.005% N2 (Life technologies), 0.01% B27 (Life technologies), 1 unit/mL of penicillin and 1 µg/mL of streptomycin (P/S) (Life technologies) with 20ng/ml FGF (Peprotec). To generate neurons, NPCs were cultured on poly-L-ornithine and laminin coated plates. When cells were confluent, differentiation was started by changing the NPC base with 20ng/ml FGF to NPC differentiation medium (NPC base with 10ng/ml BDNF (Peprotech), 10ng/ml GDNF (Peprotech), 1µM cAMP (Sigma) and 1µg/ml laminin (Sigma). Medium was refreshed three times a week. After two weeks of differentiation, cells were washed with PBS and dissociated with accutase and accumax (1:1, innovative cell technologies), resuspended in sort buffer (NPC base with 1% FBS (Life technologies) and 2.5mM EDTA), filtered (100µm Fisherbrand) and spun down. Cells were resuspended in NPC differentiation medium and plated at 50.000cells/ mL onto confluent rat astrocytes previously treated with 2µM AraC (Sigma). Cells were treated once with 2µM AraC. For the tau pathology seeding assay, neurons were infected with lentiviral particles containing the TauP301L-EGFP construct 15 days prior fixation and treated with K18 at 30nM 11 days prior fixation. For the proteolytic activity assay, neurons were treated with 10 µg/mL DQTM Red-BSA (ThermoFisher) for 18h prior to fixation. Cells o were maintained at 37 C and 5% CO2, and 50% medium was refreshed two times a week until the termination of the experiment (9 weeks since the differentiation start).

Detergent extraction and western blot

Both HEK293 cells and mouse neurons were washed with ice-cold PBS (Gibco), scraped into ice-cold Triton lysis buffer (1% Triton in PBS with phosphatase (PhosSTOP, Sigma) and protease (SigmaFAST, Sigma) inhibitor cocktails). Lysed cells were pelleted at 100.000 g for 30 minutes at room temperature (RT) in a Sorvall M120-SE centrifuge with Sorvall S100-AT6 rotor in ultracentrifuge tubes (Seton). Supernatant was stored at -80oC as the soluble fraction. Pellets were washed once in Triton lysis buffer and again

122 The seeding of tau pathology alters the endolysosomal system centrifuged at 100.000 g for 30 minutes at RT. Subsequently, pellets were re-suspended into SDS lysis buffer (1% SDS in PBS with phosphatase and protease inhibitor cocktails) and centrifuged at 100.000 g for 30 minutes at RT. Supernatant was stored at -80oC as the insoluble fraction. The protein concentration of the soluble fraction was determined using BCA assay (Pierce). A volume containing 10µg of the soluble fraction and three times more volume of the insoluble fraction were loaded in a SDS-PAGE (Biorad 4-15% Miniprotean TGX Stain free pre-cast gels). Prior to loading, the samples were diluted in loading buffer and PBS to reach equal volumes and boiled at 96°C for 5 minutes. 2μL per lane of PageRuler standard was used as a molecular weight reference. After running in SDS-PAGE, Trichlorethanol (TCE) in gel was imaged using Biorad Gel Doc EZ imager to control for protein loading levels. Proteins were dry-transferred into nitrocellulose membranes (5min, 2.5 A) and blocked with 5% milk diluted in TBS-T (Tris-buffered saline with 0.05% Tween) during 30 minutes at room temperature. Membranes were incubated overnight at 4oC with the primary antibodies in TBS-T (see Supplementary Table S2 for primary antibody details). After four washes with TBS-T, membranes were incubated with secondary antibodies for 1h at room temperature (1:2000 horseradish peroxidase (HRP)- conjugated secondary antibodies (DAKO). After four washes, membranes were incubated Lumi-Light western blotting substrate (Sigma), imaged with a LI-COR Odyssey Imaging system and analyzed with Image Studio Lite Ver 5.2 software.

Immunocytochemistry and confocal imaging

Cells were fixed with 4% paraformaldehyde in PBS and cell culture media (1:1) for 10 minutes followed by 4% paraformaldehyde in PBS for 30 minutes at room temperature. Cells were washed three times with PBS, permeabilized with 0.5% Triton X-100 for 5 minutes and, blocked with 2% normal goat serum and 0.1% Triton X-100 in PBS for 40 minutes. Antibodies were diluted in 2% normal goat serum and 0.1% Triton X-100 in PBS. Cells were incubated overnight at 4oC with primary antibodies at its optimal dilution (see Supplementary Table S2). Cells were washed three times with PBS, incubated during 1 4 hour with secondary antibodies conjugated to Alexa dyes (1:1000, Molecular Probes), washed three times with PBS, incubated 10 minutes with DAPI 0.5μg/mL (Thermo Fisher Scientific) and mounted on microscope slides with Dabco-Mowiol (Invitrogen).

Single plane confocal images were acquired in a confocal microscope (Nikon Eclipse Ti) equipped with 63x/1.4 oil objective controlled by NisElements 4.30 software. JACoB plugin was used for colocalization analysis (Bolte and Cordelieres, 2006). For quantification of labelling intensity, neurons were analyzed measuring the intensity inside a neuronal mask (using MAP2) in ImageJ. For the characterization of the HEK293 endolysosomal system, randomly selected images were thresholded based on intensity and converted

123 Chapter 4 into a binary image to perform particle analysis in ImageJ. Particle size inclusion range was 0.1-5µm (Supplementary Figure S2). Same procedure quantified DQTM-BSA content of LAMP1 and DQTM-BSA puncta in control and Tau-P301L-GFP aggregated human neurons. For the characterization of the murine endolysosomal system, fields of view with evident Tau-P301L-GFP aggregated neurons were selected and analyzed using “Spots detection” feature of Imaris version 9.2 (Bitplane Inc.) with average diameter of 500 nm.

Immunocytochemistry and 3D-STED imaging

Cells were fixed with 4% paraformaldehyde in PBS for 10 minutes at room temperature. Then, cells were washed three times with PBS and blocked with 5% normal goat serum in PBS for 40 minutes. Neurons were incubated overnight at 4oC with primary antibodies, washed three times with PBS, incubated overnight at 4oC with secondary antibodies conjugated to STAR dyes (1:50, Aberrior), and washed three times with PBS. Cells were incubated 10 minutes with DAPI 0.5μg/mL (Thermo Fisher Scientific) and mounted on microscope slides with Dabco-Mowiol (Invitrogen). The antibodies were diluted in 1% BSA in PBS at its optimal dilution (see Supplementary Table S2).

Stimulated emission depletion (STED) microscopy was performed on a Leica TCS SP8 STED 3X microscope, Leica Microsystems (Wetzlar, Germany). Samples were excited with a pulsed white light laser at their maximum excitation efficiency. A continuous wave STED laser line at a wavelength of 592 nm was used for depletion of the TauP301L-GFP, reaching a lateral resolution of ~100 nm. A pulsed STED laser line at a wavelength of 775 nm was used for the depletion of the 580- and 635-nm fluorophores, reaching a lateral resolution of 80 nm. 50% of STED laser was used to performed 3D-STED reaching a Z resolution of ~200 nm. STED images were acquired using a dedicated oil objective with 100× magnification and a numerical aperture of 1.4 (Leica Microsystems). The signal was detected using a gated hybrid detector (HyD), Leica Microsystems, in photon-counting mode. The pinhole was set to 0.71 Airy Units (AU). Fields of view with evident Tau-P301L- GFP aggregated neurons were selected and 51 images z-stacks were sampled at 19 nm laterally and 80nm axially, optimized using Nyquist Calculator from SVI (Scientific Volume Imaging, Hilversum, the Netherlands).

Z-stacks were 3D reconstructed in Imaris version 9.2 (Bitplane Inc.) to analyze the endosomal compartments. Based on the intensity of the raw fluorescent signal, the “Surface” feature of Imaris was used to identify compartments and create three dimensional surfaces to derive quantitative values. Quality control of the data was performed by manually validating randomly selected z-stacks. Crops of the quality control checks provided by Imaris are presented in the typical examples as reconstructions in

124 The seeding of tau pathology alters the endolysosomal system

Figure 4 and Figure 5 and in the Supplementary Figure S5. Deconvolution was performed using the “Deconvolution Express” function in Huygens Professional, with the standard deconvolution profile.

Statistical Analysis

Data are expressed as mean values ± standard error of the mean (SEM). The Shapiro- Wilk normality test and Bartlett’s test were used to test normality and homoscedasticity respectively. t-test was used was used to compare homoscedastic and normal distributed groups, otherwise Mann-Whitney test was used. When P-values were lower than 0.05, significance was noted in the figures as: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author on request.

AUTHOR CONTRIBUTIONS

S.V.S. performed experiments and analyzed the data. V.W. and S.V.S implemented, collected and analyzed data for the validation of the seeding of tau pathology models. J.K. assisted on 3D-STED imaging. R.J. collected confocal images for the analysis of the endolysosomal system and M.F. analyzed them. M.B., V.W. and W.S. provide reagents and critically discussed the data. S.V.S. and J.R.T.vW. designed the experiments and wrote the manuscript.

ACKNOWLEDGMENTS

The authors thank Prof. Dr. Matthijs Verhage for his suggestions and critical reading of the 4 manuscript, Joke Wortel for housing and breeding the mice, Rik van der Kant and Desiree Schut for providing cell cultures, Robbert Zalm, Ingrid Saarloos and Joost Hoetjes for cloning and lentiviral production, Rene J.P. Musters and Marko Popovic for reagents, support and assistance in 3D-STED imaging and analysis, and Jeroen J.M. Hoozemans for his suggestions. This work was supported by Alzheimer Nederland (WE.03-2016-05).

COMPETING INTERESTS STATEMENT

The authors declare no competing financial interests.

125 Chapter 4

SUPPLEMENTARY FIGURES

a TauP301L-GFP MC1 Merge + DAPI b

15 * Control

10

MC1 (a.u.)MC1 5 K18 0 Control K18

c TauP301L-GFP AT8 Merge + DAPI d

10 ****

8 Control 6

4 AT8 (a.u.)AT8

2

K18 0 Control K18

e *** f 25

20 130 AT8 100 15 130 Tau5 10 100

5 GAPDH 40 TauP301L-GFP (a.u.) Sol Ins Sol Ins Sol Ins 0 Control K18 WT Control K18

Supplementary Figure S1: Seeding with K18 fibrils induces tau pathology in HEK293 cells overexpressing TauP301L-GFP. (a) Confocal microscopy images of HEK293 cells overexpressing TauP301L-GFP with tau aggregates and control immunolabelled for MC1 (misfolded tau-specific antibody (Jicha et al., 1997)). Merged image of MC1 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co-localization between TauP301L-GFP clustering and MC1. n=27±2 fields of view, N=2 experiments. Scale bar=20 μm.(b) Quantification of MC1 staining intensity relative to control. (c) Confocal microscopy images of HEK293 cells overexpressing TauP301L-GFP

126 The seeding of tau pathology alters the endolysosomal system with tau aggregates and control immunolabelled for AT8 (tau phospho-specific antibody for Ser202 and Thr205 (Goedert et al., 1995). Merge image of AT8 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co-localization between TauP301L-GFP and AT8 accumulations. n=39±2 fields of view, N=2 experiments. Scale bar=20 μm. (d) Quantification of AT8 staining intensity relative to control. (e) Quantification of TauP301L-GFP staining intensity relative to control. (f) Representative western blot for AT8 and Tau5 of sequential extraction of HEK293 cellular fractions recovered in 1% Triton X-100 lysis buffer and fraction solubilized in 1% SDS lysis buffer. GAPDH served as a loading control. (N=3 independent experiments).

EEA1 TauP301L-GFP a DAPI EEA1 b EEA1 puncta Control K18

Supplementary Figure S2: Quantification of EEA1 puncta in confocal images of HEK293 cells. (a) Confocal microscopy images of control and HEK293 cells with tau aggregates immunolabelled for EEA1. Merge image of EEA1 (magenta), TauP301L-GFP (green) and DAPI (blue). Scale bar=20 μm. (b) Zoom of the area squared in (a) showing the puncta used to derive quantitative values.

4

127 Chapter 4

a MAP2 TauP301L-GFP MC1 Merge Control K18

MAP2 TauP301L-GFP AT100 Merge b Control K18

c d e ** **** **** 1.25 1.25 20

1.00 1.00 15 0.75 0.75 10 0.50 0.50 5 0.25 0.25 MC1 intensityMC1 (a.u.) AT100 intensityAT100 (a.u.)

0 0 0 TauP301L-GFP intensityTauP301L-GFP (a.u.) Control K18 Control K18 Control K18

f 130 AT8 100 AT8 130 Tau5Tau5 100

40 GAPDH

Sol Ins Sol Ins

Control K18 Supplementary Figure S3: Seeding with K18 fibrils induces tau pathology in primary mouse neurons overexpressing TauP301L-GFP. (a) Confocal microscopy images of mouse primary neurons overexpressing TauP301L-GFP treated with K18 tau fibrils and control immunolabelled for MC1. Merge image of MC1 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co- localization between TauP301L-GFP clustering and MC1. n=45 fields of view, N=1 experiment. Scale bar=10 μm. (b) Quantification of MC1 staining intensity relative to control. (c) Confocal microscopy images of mouse primary neurons overexpressing TauP301L-GFP treated with K18 fibrils and control

128 The seeding of tau pathology alters the endolysosomal system immunolabelled for AT100 (tau phospho-specific antibody for Thr212 and Ser214 (Zheng‐Fischhöfer et al., 1998)). Merge image of AT100 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co-localization between TauP301L-GFP and AT100 accumulations. n=221±14 fields of view, N=3 experiments. Scale bar=10 μm. (d) Quantification of AT100 staining intensity relative to control. (e) Representative western blot for AT8 and Tau5 of sequential extraction of mouse primary neurons cellular fractions recovered in 1% Triton X-100 lysis buffer and fraction solubilized in 1% SDS lysis buffer. GAPDH served as a loading control. (N=3 independent experiments).

a MAP2 TauP301L-GFP MC1 Merge Control K18

MAP2 TauP301L-GFP AT100 Merge b Control K18

c **** d **** e 1.5 4 8 4

3 6 1.0

2 4

0.5 1 2 MC1 intensityMC1 (a.u.) AT100 intensityAT100 (a.u.)

TauP301L-GFP intensity TauP301L-GFP (a.u.) 0 0 0 Control K18 Control K18 Control K18

Supplementary Figure S4: Seeding with K18 fibrils induces tau pathology in human neurons overexpressing TauP301L-GFP. (a) Confocal microscopy images of human neurons overexpressing TauP301L-GFP with tau aggregates and control immunolabelled for MC1. Merge image of MC1 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co-localization between TauP301L-GFP clustering and MC1. n=60±1 fields of view, N=3 experiments. Scale bar=20 μm. (b)

129 Chapter 4

Confocal microscopy images of human neurons overexpressing TauP301L-GFP with tau aggregates and control immunolabelled for AT100. Merge image of AT100 (magenta), TauP301L-GFP (green) and DAPI (blue). Arrowheads indicate co-localization between TauP301L-GFP and AT100 accumulations. n=58±1 fields of view, N=3 experiments. Scale bar=20 μm.(c) Quantification of TauP301L-GFP intensity relative to control. (d) Quantification of MC1 staining intensity relative to control. (e) Quantification of AT100 staining intensity relative to control.

EEA1 puncta reconstruction a EEA1 MAP2 + DAPI TauP301L-GFP

b EEA1

c

130 The seeding of tau pathology alters the endolysosomal system

Supplementary Figure S5: 3D reconstruction used to quantify EEA1 puncta. (a) 3D view of a 4 μm z-stack of 51 3D-STED microscopy images of a human neuron overexpressing TauP301L-GFP (green) with tau aggregates, immunolabelled for EEA1 (magenta), and MAP2 (blue and DAPI). EEA1 positive puncta reconstructed is shown in grey. Scale bar=2 μm. (b) 3D view of the z-stack in (a) showing only EEA1 staining (magenta) and EEA1 reconstructed puncta (grey). Scale bar=2 μm. (c) Gallery of the three dimensionally reconstructed EEA1 puncta in (d) from which quantitative values are derived. Scale bar=1 μm.

a b MAP2 TauP301L-GFP VPS35 Merge

**** 3 Control

2

1 VPS35 intensity (a.u) K18 0 Control K18

MAP2 TauP301L-GFP SNX4 Merge c d ** 2.5

2.0 Control 1.5

1.0

0.5 SNX4 intensitySNX4 (a.u.)

K18 0 Control K18

Supplementary Figure S6: Seeding of tau pathology increases the levels of the endosomal 4 sorting proteins VPS35 and SNX4 in human neurons. (a) Confocal microscopy images of human neurons overexpressing TauP301L-GFP with tau aggregates and control immunolabelled for VPS35. Merge image of VPS35 (magenta), TauP301L-GFP (green) and DAPI (blue). n=60±1 fields of view, N=3 experiments. Scale bar=5 μm. (b) Quantification of VPS35 staining intensity relative to control. (c) Confocal microscopy images of human neurons overexpressing TauP301L-GFP with tau aggregates and control immunolabelled for SNX4. SNX4 polyclonal antibody shows unspecific nuclear staining. Merge image of SNX4 (magenta), TauP301L-GFP (green) and DAPI (blue). n=58±4 fields of view, N=3 experiments. Scale bar=5 μm. (d) Quantification of SNX4 staining intensity relative to control.

131 Chapter 4

Supplementary Table S1: Summary of the mean, SEM, n/N numbers and statistics of measured variables in the study.

Figure Meassured variable Group Mean ± SEM n/N Statistics Control 888.5 ± 69 30/3 1c EEA1 puncta number t=0.6667, df=62, p=0.5074 K18 946 ± 59 34/3 Control 0.24 ± <0.01 26/3 1d EEA1 puncta size (µm2) U=210.0, p=0.0006 K18 0.20 ±<0.01 34/3 Control 1.00 ± 0.03 29/3 1e EEA1 puncta intensity (a.u.) U=434.0, p=0.6701 K18 1.02 ± 0.04 32/3 Control 216 ± 38 18/1 1f Rab5 puncta number U=59.0, p=0.0007 K18 79 ± 8 19/1 Control 0.31 ± 0.01 18/1 1g Rab5 puncta size (µm2) t=4.395 df=35, p<0.0001 K18 0.26 ± <0.01 19/1 Control 1.00 ±<0.01 17/1 1h Rab5 puncta intensity (a.u.) t=6.414 df=33, p<0.0001 K18 1.02 ±<0.01 18/1 Control 1.00 ± 0.13 44/3 2b CD63 puncta number U=663.0, p=0.0074 K18 2.98 ± 0.65 45/3 Control 0.73 ± 0.03 44/3 2c CD63 puncta size U=696.0, p=0.0160 K18 0.79 ± 0.01 45/3 Control 1:00 ± 0.02 44/3 2d CD63 puncta intensity U=918.0, p=0.5574 K18 0.93 ± 0.04 45/3 Control 1.00 ± 0.11 44/3 2e LAMP1 puncta number U=517.0, p=0.0001 K18 2.85 ± 0.37 45/3 Control 0.81 ± 0.05 44/3 2f LAMP1 puncta size U=732.0, p=0.0346 K18 0.87 ± 0.03 45/3 Control 1.00 ± 0.02 44/3 2g LAMP1 puncta intensity U=605.0, p=0.0016 K18 0.89 ± 0.02 45/3 Control 0.33 ± 0.01 44/3 2h Pearson's CD63 and LAMP1 t=5.638 df=87, p<0.0001 K18 0.44 ± 0.01 45/3 Control 0.19 ± 0.01 44/3 2i Mander's CD63 in LAMP1 U=728.0, p=0.0322 K18 0.24 ± 0.01 45/3 Control 0.50 ± 0.02 44/3 2j Mander's LAMP1 in CD63 t=2.666 df=87, p=0.0091 K18 0.57 ± 0.01 45/3 Control 818 ± 204 24/3 3c LIMPII puncta number U=273.50, p=0.6030 K18 1260 ± 383 25/3 Control 3.11 ± 0.1 24/3 3d LIMPII puncta size (µm2) t=0.2747 df=46, p=0.7847 K18 3.14 ±0.06 25/3 Control 1.00 ± 0.10 24/3 3e LIMPII puncta intensity (a.u.) t=1.687 df=46, p=0.984 K18 0.77 ± 0.07 25/3 Control 26.27 ± 3.00 44/4 4d EEA1 puncta number U=719.0, p=0.0380 K18 18.48 ± 2.15 44/4 Control 0.29 ± 0.02 44/4 4e EEA1 puncta volume (µm3) U=728.0, p=0.0319 K18 0.21 ±0.01 45/4 Control 1.00 ± 0.09 44/4 4f EEA1 puncta intensity (a.u.) U=478.0, p<0.0001 K18 0.48 ± 0.03 45/4 Control 61.09 ± 9.62 32/3 5c CD63 puncta number U=507.5, p=0.6439 K18 69.35 ± 9.16 34/3

132 The seeding of tau pathology alters the endolysosomal system

Control 0.32 ± 0.04 32/3 5d CD63 puncta volume (µm3) U=487.0, p=0.9070 K18 0.35 ±0.06 31/3 Control 1.00 ± 0.08 31/3 5e CD63 puncta intensity (a.u.) U=449.0, p=4049 K18 0.94 ± 0.08 33/3 Control 58.21 ± 7.54 33/3 5f LAMP1 puncta number U=559.0, p=0.7685 K18 67.53 ± 9.37 36/3 Control 0.17 ± 0.01 31/3 5g LAMP1 puncta volume (µm3) U=491.0, p=0.4030 K18 0.18 ±0.02 36/3 Control 1.00 ± 0.04 33/3 5h LAMP1 puncta intensity (a.u.) U=514.0, p=0.3396 K18 1.01± 0.09 36/3 Control 0.60 ± 0.02 34/3 5i Pearson's CD63 and LAMP1 U=56150, p=0.4405 K18 0.56 ± 0.03 37/3 Control 0.40 ± 0.03 34/3 5j Mander's CD63 in LAMP1 t=1.264 df=69, p=0.2103 K18 0.34 ± 0.03 37/3 Control 0.64 ± 0.04 34/3 5j Mander's LAMP1 in CD63 t=0.5990 df=69, p=0.5511 K18 0.67 ± 0.03 37/3 Control 1. ± 0.05 42/3 6b DQ-BSA in LAMP1 puncta (a.u.) U=489.5, p=0.0007 K18 0.62 ±0.08 41/3 Control 74 .47 ± 15.92 38/3 6c DQ-BSA puncta number U=434.0, p=0.0006 K18 23.22 ± 7.58 41/3 Control 1.97 ± 0.10 34/3 6d DQ-BSA puncta area (µm2) U=273.5, p=0.0344 K18 1.71 ±0.07 24/3 Control 1.00 ± 0.05 34/3 6e DQ-BSA puncta intensity (a.u.) U=256.0, p=0.0168 K18 0.84 ± 0.02 24/3 Control 1.00 ± 0.30 37/2 S1b MC1 intensity (a.u.) U=506.0, p=0.0117 K18 1.40 ± 0.33 41/2 Control 1.00 ± 0.08 25/2 S1d AT8 intensity (a.u.) U=23.0, p<0.0001 K18 3.65 ± 0.36 28/2 Control 1.00 ± 0.11 38/2 S1e TauP301L-GFP intensity (a.u.) U=400.0, p=0.0003 K18 4.06 ± 0.75 40/2 Control 1.00 ± <0.01 45/1 S3c MC1 intensity (a.u.) U=261.0, p<0.0001 K18 1.023 ± <0.01 45/1 Control 1.00 ± 0.02 207/3 S3d AT100 intensity (a.u.) U=10845, p<0.0001 K18 3.36 ± 0.22 232/3 Control 1.00 ± <0.01 45/1 4 S3e TauP301L-GFP intensity (a.u.) U=660.0, p=0.0045 K18 1.023 ± <0.01 45/1 Control 1.00 ± 0.06 59/3 S4c TauP301L-GFP intensity (a.u.) U=931.0, p<0.0001 K18 1.50 ± 0.08 60/3 Control 1.00 ± 0.08 59/3 S4d MC1 intensity (a.u.) U=267.0, p<0.0001 K18 3.05 ± 0.20 60/3 Control 1.00 ± 0.02 57/3 S4e AT100 intensity (a.u.) U=1562, p=0.5111 K18 0.98 ± 0.02 59/3 Control 1.00 ± 0.03 59/3 S6d VPS35 intensity (a.u.) t=9.450 df=117 p<0.0001 K18 1.56 ± 0.04 60/3 Control 1.00 ± 0.02 57/3 S6e SNX4 intensity (a.u.) U=1130, p=0.0023 K18 1.22 ± 0.04 59/3

133 Chapter 4

Supplementary Table S2: Primary antibodies specifications and concentrations.Not applicable (empty cells).

Antibody Company Catalog number dilution STED dilution confocal dilution WB AT100 ThermoScientific MN1060 1 : 100 AT8 ThermoScientific MN1020 1 : 100 1:500 CD63 BD Bioscience 556019 1 : 100 1 : 1000 EEA1 Cell Signaling 2411 1 : 100 1 : 100 GAPDH EMD Millipore mab375 1:2500 LAMP1 Cell Signaling 9091S 1 : 100 1 : 100 LIMPII Novus Biologicals nb400-129 1 : 100 MAP2 Abcam ab5392 1 : 2000 1 : 20000 MC1 kind gift from Peter Davies 1 : 100 Rab5 Abcam ab18211 1 : 200 SNX4 SynapticSystems 392 003 1 : 500 Tau5 Abcam Ab80579 1:1000 VPS35 Abcam ab97545 1 : 500

134 Chapter 5 Summary, general discussion and future directions

135 Chapter 5

SUMMARY

The first aim of this thesis was to study the role of endosomal sorting in presynaptic terminals (Chapter 2 and Chapter 3). We used a shRNA approach to evaluate the impact of depleting key endosomal sorting proteins in the presynaptic terminals of mouse primary neurons. In Chapter 2, we show for the first time that VPS35 is present at the mammalian presynaptic terminal. VPS35 depletion did not affect neuronal morphology (neurite length and synapse number), presynaptic ultrastructure, and synaptic vesicle release and retrieval. In Chapter 3, we first set out to perform basic characterizations of SNX4 localization in neurons. We developed a novel antibody against SNX4 for western blot, immunocytochemistry and immunoelectron microscopy. SNX4 was present in neurons and it accumulated in synapses, mainly at presynaptic terminals. SNX4 depletion downregulated synaptic communication-related proteins without affecting presynaptic ultrastructure or neuronal morphology. Three shRNAs against SNX4 drastically impaired synaptic vesicle release. However, this phenotype was not restored by expressing a SNX4 variant resistant to the shRNAs. In Chapter 2 and Chapter 3, the identification of VPS35 and SNX4 as presynaptic proteins indicated a selective demand for endosomal sorting in presynaptic boutons.

The second aim of this thesis was to investigate if the endolysosomal aberrations observed in the Alzheimer’s disease (AD) brain can be a consequence of the seeding of tau pathology (Chapter 4). In iPSC-derived human neurons, the seeding of tau pathology decreased the number, size and EEA1 labelling intensity of early endosomes, and increased the levels of the endosomal sorting proteins VPS35 and SNX4. The seeding of tau pathology did not morphologically change late endosomes and lysosomes but lysosomal proteolytic activity appeared to be reduced. Thus, the seeding of tau pathology causally affects the endolysosomal system, which may play a role in neurodegeneration in AD.

GENERAL DISCUSSION

1.The shRNA approach to study endosomal sorting proteins in presynaptic terminals shRNAs are widely used to deplete proteins more acutely than genetic modifications, but this method has two main pitfalls. First, effective shRNAs deplete protein levels but do not abolish their expression. Therefore, it cannot be excluded that minor residual levels of target proteins are enough to carry their function, thus ‘no effect’ upon shRNA treatment should be interpreted with caution. In Chapter 2, we validated that the reduction in VPS35 expression was large enough to produce functional effects in a known VPS35

136 General Discussion postsynaptic function by measuring GluA1 surface localization (Ch.2 Figure 2) (Tian et al., 2015). Unfortunately, such validation assays are not available for all shRNA targets. Second, the shRNA approach is susceptible to off-target effects, which are defined as those proteins or processes which are affected by the shRNA that are independent of the target. To avoid the phenotypic association with off-target effects, only replicated phenotypes by three independent shRNAs against the target protein are considered potentially specific in this thesis. Furthermore, if a phenotype was produced by the three shRNAs, a rescue experiment by re-expressing non-shRNA-targetable constructs was performed to ensure specificity.

In Chapter 3, the synaptic vesicle release impairment phenotype produced by three independent shRNAs against SNX4 was not reversed upon SNX4 re-expression (Ch.3 Figure 7). The time line of the rescue experiments might play a critical role in determining the outcome, since depletion of the protein might lead to irreversible damage. Rescue experiments with different time lines were also performed but synaptic vesicle impairment upon SNX4 depletion was not restored (data not shown). The amount of re-expressed protein might also be crucial since overexpression of proteins might have an effect as well. We performed rescue experiments with different protein dosages but the synaptic vesicle release impairment was not restored (data not shown). Hence, the experimental design in terms of timeline and SNX4 dosage does not likely explain the lack of phenotypic restoration. The levels and distribution of the re-expressed SNX4 protein were validated (Ch.3 Figure 2), but re-expressed SNX4 functionality was not verified as no such assay is available in neurons. The re-introduced SNX4 constructs lack mRNA untranslated regions, which can determine the translation efficiency and folding of a protein (see review (Rodnina, 2016)). Re-expressed SNX4 might be differently folded which might tune SNX4 structure, posttranslational modifications and interaction with other proteins, and hence, its function and ability to restore the synaptic vesicle release phenotype. Potentially, SNX4 needs posttranslational modifications to function that are not present on re-introduced SNX4. Atg20 is the yeast homolog of SNX30, a SNX-BAR which requires SNX4 to form functional dimers (van Weering et al., 2012a). Atg20 architecture is fine- tuned by acetylation and phosphorylation in the N-terminal, which impacts its function in autophagy by allowing the interaction with the autophagic protein Atg11 (Popelka et al., 2017). SNX4 is predicted to undergo acetylation and phosphorylation in the N-terminal, as Atg20 (http://www.uniprot.org), which may affect its function. Although the data do not exclude these hypotheses, a common off-target effect on synaptic vesicle release of the three independent SNX4 shRNAs seems the most likely explanation for the lack of phenotypic rescue upon SNX4 re-introduction. 5

In Chapter 2 and 3, shRNAs were designed to only target murine VPS35 and SNX4

137 Chapter 5 mRNAs. The Basic Local Alignment Search Tool (BLAST) was used to assure the sequence specificity of the target sequence, and western blot verified protein depletion. Although shRNAs were designed with the same strategy, the shRNAs against VPS35 did not affect synaptic vesicle release (Ch.2 Figures 5 and 6), indicating that the synaptic vesicle release impairment is not a general off-target effect of the used shRNA approach. The proteomic analysis of SNX4 depleted neurons showed that each shRNA induced a different dysregulation of the neuronal proteome (Ch.3 Figure 8), while SNX4 was knocked down similarly by the three shRNAs against SNX4 (Ch.3 Figure 2). These uniquely dysregulated proteins seem SNX4 depletion-independent and they might be or produce off-target effects. At this stage, it is not possible to exclude that the synaptic vesicle release impairment upon SNX4 knock downs is produced by these SNX4-independent dysregulated proteins. Although the shRNA approach is widely used in the literature, there are few reports advocating for its controlled use. Off-target effects on neuronal morphology are common and depend on the shRNA sequence, independently of the targeted protein (Alvarez et al., 2006). Off-target effects on neuronal migration have also been reported using nine control (scramble) shRNAs, due to an alteration of endogenous miRNA pathways (Baek et al., 2014). Similar issues have been seen reported with the use of morpholinos (Eisen and Smith, 2008) and pharmacological tools due to unspecific effects. However, when tightly controlled, these unperfect tools have expanded our knowledge. The germinal knock out models are an alternative strategy but they are not always available or appropriate to study protein function. For example, the total lack of some endolysosomal proteins leads to toxicity, developmental problems and embryonic lethality, such us subunits of retromer or the WASH complex (Jahic et al., 2015; Xia et al., 2013) (Lee et al., 1992; Muhammad et al., 2008) (Zhou et al., 2011). Development of new tools seems necessary to investigate the role of these proteins, such as conditional knock outs in which the total lack of a protein can be timely controlled. Nevertheless, this approach is less affordable, also has drawbacks and rescue experiments are equally important.

The shRNA approach is not only a tool to study protein function, it is also a therapeutic strategy to deplete the mRNA of toxic proteins. shRNAs offer long lasting treatment (up to 3 years) and they can be delivered in vivo, making them a promising option for different diseases, including neurological disorders such as Huntington’s Disease (Aguiar et al., 2017). Huntington’s Disease is an inherited trinucleotide repeat disorder (> 36 CAGs) in the Huntingtin gene in which shRNA approach is undertaken but has not been successful yet (Aguiar et al., 2017). The shRNA approach might provide a powerful technology to deplete aberrant proteins. However, it depends entirely on specificity of the shRNA which is difficult to control in a clinical setting and cannot be theoretical predicted. The shRNAs need to be thoroughly validated as a therapeutic strategy, to reduce their off-target effects

138 General Discussion and to optimize the dosage to minimize toxicity while maximizing target mRNA depletion.

2.The role of presynaptic VPS35 and SNX4

VPS35 and SNX4 are present in mammalian presynaptic terminals but their depletion did not affect presynaptic ultrastructure (Ch.2 Figure 4 and Ch.3 Figure 6), which suggests that VPS35 and SNX4 are not required for synaptic vesicle recycling. Hence, a main question remains unanswered: why these recycling proteins are present in presynaptic terminals but they do not seem to participate in the dominant presynaptic membrane recycling pathway, the synaptic vesicle recycling pathway (Sudhof, 2004). Here, we explore some explanations for the results and alternative presynaptic recycling pathways in which VPS35 and SNX4 might participate.

2.1 VPS35 and SNX4 might participate in synaptic vesicle recycling under specific physiological conditions

Synaptic vesicle recycling can occur through different pathways depending on the stimulation paradigm. In 1973, the recycling of synaptic vesicles was followed for the first time in presynaptic nerve terminals of muscles of Rana pipiens by transmission electron microscopy in combination with horseradish peroxidase, which labels endocytic compartments. Synaptic vesicles disappeared upon a strong stimulation protocol while the plasma membrane length increased. When the plasma membrane length was restored, cisternae (endosome-like compartments) transiently appeared to reform synaptic vesicles (Heuser and Reese, 1973). Upon a milder stimulation (2Hz), synaptic vesicles were reformed directly from the plasma membrane (Ceccarelli et al., 1973), suggesting that the intensity of the stimulation can define the recycling pathway. The same amount of action potentials (200AP) given at different frequency can also stimulate different recycling pathways. At high frequency (40Hz) endosomal recycling is more prominent than at slower frequencies (5Hz) (Kononenko et al., 2014). Nevertheless, even after a single stimulus which elicits 1-2 action potentials, endosomal recycling can regenerate synaptic vesicles (Watanabe et al., 2014). Our ultrastructural studies were performed under resting conditions without any stimulation and without blocking spontaneous activity in network cultures. Hence, SNX4 and VPS35 are not required for basal synaptic vesicle recycling but it cannot be excluded that these proteins play a role in synaptic vesicle recycling under other circumstances such us upon intense stimulation. Investigating how different stimuli impact the presynaptic ultrastructure upon VPS35 and SNX4 depletion by combining high pressure freezing and transmission electron microscopy will aid to understand the 5 role of these proteins upon different physiological conditions with exquisite temporal and spatial resolution.

139 Chapter 5

2.2 Presynaptic VPS35 and SNX4 might participate in neurotransmission but not in membrane recycling of synaptic vesicles

Presynaptic ultrastructural changes do not always accompany alterations in synaptic vesicle release. For example, the lack of complexins dramatically impairs neurotransmitter release without affecting presynaptic ultrastructure (Reim et al., 2001). Electrophysiological recordings are preferred to measure neurotransmission because they provide excellent temporal resolution and the possibility to measure different release components and plasticity using well established protocols. VPS35 depletion decreases the surface expression of glutamate receptors (Ch.2 Figure 2), which, in turn, reduces AMPA and NMDA excitatory postsynaptic currents (Choy et al., 2014; Tian et al., 2015). Isolating the impact of presynaptic VPS35 from postsynaptic VPS35 effects in electrophysiological recordings might be challenging. Therefore, we chose fluorescent reporters of synaptic vesicle release (synaptopHluorin and sypHy) to address the role of presynaptic VPS35 in neurotransmission (Ch.2 Figure 5 and 6 and Ch.3 Figure 7). SNX4 was also detected in postsynaptic elements (Ch.3 Figure 4), and we anticipated that postsynaptic SNX4 might also traffic neurotransmitter receptors to the cell surface. Therefore, a synaptopHluorin/ sypHy approach was also adopted to address the impact of SNX4 depletion in presynaptic neurotransmission. Contrary to our prediction, the depletion of SNX4 did not affect the peak amplitude of spontaneous excitatory postsynaptic currents (Ch.3 Figure S7), indicating that SNX4 depletion does not affect the glutamate receptors at the surface of the postsynaptic terminal. In future investigations, electrophysiological recordings should be incorporated to assess the role of presynaptic SNX4 in synaptic function.

At this stage, the evidence for a role of SNX4 in synaptic vesicle release is not conclusive (Ch.3 Figure 7). Given the role of SNX4 in cargo sorting (Hettema et al., 2003; Traer et al., 2007), we speculate that SNX4 might be involved in endosomal sorting of presynaptic cargo. Blocking such a pathway would not automatically result in an altered synaptic vesicle membrane flow, in line with our observations upon SNX4 depletion (as in Ch.3 Figure 6). The decrease in endosomal sorting may result in dysfunctional synaptic vesicle machinery, consistent with the synaptic vesicle release impairment upon SNX4 depletion (Ch. 3 Figure 7). Uncovering these unexplored synaptic endosomal sorting pathways will provide insight in the homeostasis of the local presynaptic proteome, on which these distant synaptic terminals may depend.

140 General Discussion

2.3 VPS35 might participate in presynaptic plasticity by regulating signaling of G protein–coupled receptors (GPCRs)

Presynaptic long-term depression (LTD) is the reduction of synaptic efficacy by decreasing the probability of neurotransmitter release (Atwood et al., 2014). The predominant presynaptic LTD is mediated by presynaptic Gi-coupled metabotropic receptors such as metabotropic glutamate receptors, cannabinoid receptors and opioid receptors (Atwood et al., 2014). Retromer recycles GPCRs, such as β2 adrenergic receptor (β2AR), from endosomes to the plasma membrane controlling their surface abundance (Figure 1) (Choy et al., 2014). Apart from the plasma membrane, β2AR can initiate signaling from endosomes via the heterotrimeric G protein (Gs), which is limited by β2AR-retromer colocalization (Varandas et al., 2016). Hence, enrichment of cargo at retromer domains can regulate the GPCR signaling at endosomes (Varandas et al., 2016) (Figure 1d). Retromer modulates the type 2 of cannabinoid receptor signaling by modulation of the β-arrestin-1 mediated signaling at endosomes (Nogueras-Ortiz et al., 2017). In line with these pathways, presynaptic VPS35 might participate in presynaptic GPCR signaling,

a b c d Gs signalling

β-arrestin DAG cAMP cAMP GPCR ligand

Gq Gi Gs G Gα G proteins G βγ G Gα G βγ retromer

Figure 1: The potential role of VPS35 in presynaptic GPCR signaling. (a) GPCR signaling is activated at the plasma membrane by ligand binding to the GPCR which induces the exchange of GDP for GTP on Gα subunit of heterotrimeric G proteins. Gα subunits can be Gq (which activates phospholipase C producing diacyl glycerol (DAG), Gs or Gi which stimulates or inhibits (respectively) adenylyl cyclase, and thereby cAMP production. Both DAG and cAMP act as second messengers for several downstream targets. (b) GPCR responses desensitize as an adaptation upon prolonged or repeated ligand exposure. β-arrestins function as endocytic adapters to mediate the desensitization of GPCRs by trafficking them to endosomal compartments. (c) Retromer can recycle GPCRs from endosomes to the plasma membrane. Hence, retromer can regulate surface signaling of GPCRs by modulating their abundance in the membrane. (d) β-arrestin can also initiate GPCR signaling 5 from endosomes. This signaling can be terminated by retromer. Hence, retromer might also regulate GPCR signaling from intracellular compartments. Adapted from (van Weering and Cullen, 2011).

141 Chapter 5 both controlling surface targeting and endosomal signaling, and thereby VPS35 might modulate presynaptic communication. Future experiments addressing presynaptic LTD in VPS35-depleted cells are of interest to understand the role of presynaptic VPS35.

2.4 VPS35 and SNX4 might participate in APP trafficking and processing

VPS35 regulates Aβ production by trafficking APP away from endosomes (Chapter 1). VPS35 is localized presynaptically (Ch.2 Figure 1), and APP is enriched at presynaptic terminals where it is considered to play a role in cell adhesion and synaptic stability (Laßek et al., 2013; Müller et al., 2017). On the cell surface, APP is non-amyloidogenic cleaved, whereas the internalized APP is amyloidogenic cleaved by BACE1 and γ -secretase (Müller et al., 2017). APP is internalized with synaptic vesicle proteins (Marquez-Sterling et al., 1997) and endogenous Aβ is mainly localized to presynaptic vesicular organelles (Yu et al., 2018). Synaptic activity can modulate the levels of extracellular Aβ (Kamenetz et al., 2003), mainly due to synaptic vesicle release (Cirrito et al., 2005). In transgenic mice, 70% of released Aβ requires both clathrin-mediated endocytosis and synaptic vesicle release (Cirrito et al., 2008). Hence, the presynaptic terminal is a key compartment for Aβ production and release. Presynaptic VPS35 might be critical for sorting APP away from synaptic endosomes back to the plasma membrane, preventing its amylogenic processing. Decreased VPS35 levels, as observed in AD brains (Small et al., 2005), will result in enhanced amylogenic cleavage and Aβ production in synaptic endosomes.

Synaptic activity routes APP into BACE1-positive endosomes (Das et al., 2013), indicating that APP amyloidogenic cleavage can occur in these organelles. BACE1 can be recycled from the early endosome with fast kinetics by VPS35 or slow kinetics by SNX4 (Toh et al., 2018). Phosphorylation of the DXXLL-motif sequence DISLL in the cytoplasmic tail routes BACE1 to fast VPS35 recycling, reducing its residency time in the early endosome and thereby APP processing and Aβ production (Toh et al., 2018). Since both VPS35 and SNX4 are presynaptic localized (Ch.2 Figure 1, and Ch.3 Figure 4), they might play a role in presynaptic BACE1 recycling, and thus APP amylogenic cleavage. SNX4 depletion increases BACE1 in late endosomal compartments, modulates Aβ production (Toh et al., 2018) and decreases APP levels (Kim et al., 2017). In Chapter 3, APP levels were not changed upon SNX4 depletion (Ch3. Supplementary Table S2). Hence, although SNX4 may be involved in BACE1 trafficking, which might play a role in presynaptic APP amylogenic cleavage, SNX4 exact role in APP processing and Aβ production remains to be determined.

Taken together, presynaptic VPS35 and SNX4 might regulate presynaptic trafficking of a variety of cargo including proteins for synaptic secretion, GPCRs and APP. Thus,

142 General Discussion addressing the role of synaptic endosomal sorting is critical to understand physiological processes such as synaptic plasticity and maintenance as well as pathological processes such as APP amylogenic processing.

3. The seeding of tau pathology alters the endolysosomal system

Endolysosomal aberrations and tau pathology co-occur in brains of Alzheimer’s Disease and pure tauopathy patients, but the relation between these two pathological features is unexplored. We found that the seeding of tau pathology alters the endolysosomal system (Chapter 4). In neurons, the seeding of tau pathology decreases the number, size and EEA1 labelling intensity of early endosomes, upregulates the levels of VPS35 and SNX4, and appears to reduce lysosomal proteolytic activity while not morphologically changing late endosomes and lysosomes (Ch.4 Figures 4-6, Supplementary Figure S6). How the seeding of tau pathology leads to these endolysosomal aberrations is still an open question. A possible explanation for the results is that upon the seeding of tau pathology, endogenous tau loses its function which would result in the destabilization of microtubules. Consequentially, the trafficking and fusion of endolysosomal compartments will be impaired. In addition, tau aggregates might occupy cytosolic space which might also hamper the trafficking and fusion of endolysosomal compartments. Life cell imaging would be a suitable approach to explore if endolysosomal trafficking and fusion may explain the observed phenotypes upon the seeding of tau pathology. Another possible explanation is that the seeding of tau pathology damages these organelles, which would be cleared. In vitro, internalized tau seeds are trafficked through the endocytic pathway (Calafate et al., 2016; Wu et al., 2013) and damage endolysosomal compartments (Calafate et al., 2016; Flavin et al., 2017). Damaged endosomes can be sensed by galectins and trigger autophagy to clear the damaged organelles (Papadopoulos and Meyer, 2017). Hence, endosomal damage and clearance might explain the observed early endosomal diminishment upon the seeding of tau pathology (Ch.4 Figure 4).

Tau pathology can spread via exosomes (Asai et al., 2015), which implies that cytosolic tau seeds are packed into intraluminal vesicles (ILVs). Hyperphosphorylated tau can be sorted in ILVs in the multivesicular bodies (MVB) for lysosomal degradation through the Rab35/ESCRT pathway in a phospho-selective manner (Vaz‐Silva et al., 2018). For tau aggregates packed into ILVs to leak to the cytosol, both the membrane of the ILV and the MVB membrane need to be damaged, which might delay the damage and leakage of this organelle. MVBs fuse with lysosomes forming endolysosomes, the main hydrolytic compartment where degradation takes place (Bright et al., 2016). In the endolysosome, tau 5 seeds may rupture the membrane and leak into the cytosol, leading to an increase in the endolysosomal lumen pH, and thereby decreasing its degradation capability, consistent

143 Chapter 5 with the observed decrease in DQ-BSA signal (Ch. 4 Figure 6). Damaged endolysosomes cannot mature to lysosomes and will accumulate, which might compensate the reduced early endosomal influx and explain the lack of effect in number, size and labelling intensity of CD63 compartments (Ch. Figure 5). Hence, both internalized extracellular tau seeds and cytosolic tau aggregates might damage endolysosomal compartments. To further test this hypothesis, assays addressing organelle rupture, endolysosomal pH and transmission electron microscopy to discriminate between late endosomes, endolysosomes and lysosomes are recommended. Furthermore, in our experiments, we did not segregate the effects on the endolysosomal system from the K18 seed treatment or from the seed-induced tau pathology. Hence, it would be of interest to test if wild-type cells which do not develop tau pathology upon K18 treatment show the same changes in the endolysosomal system. Such experiments could form a possible starting point to dissect the underlying molecular mechanism of endolysosomal aberrations upon the seeding of tau pathology.

3.2 Upregulation of VPS35 and SNX4 as part of an autophagic response to the seeding of tau pathology

The seeding of tau pathology upregulated VPS35 and SNX4 (Ch.4 Figure S6), which might be related to the emerging role of these proteins in autophagy (Tang et al., 2015) (Zavodszky et al., 2014) (Popelka et al., 2017). Apart from being involved in hydrolase delivery to the lysosome (Chapter 1), VPS35 recycles LAMP2A to the trans-Golgi network preventing its degradation and promoting LAMP2-chaperone mediated autophagy (Tang et al., 2015). In addition, both VPS35 and yeast Snx4 facilitate macroautophagy through trafficking of autophagic proteins such as ATG9A, Atg11 and Atg27 (Suzuki and Emr, 2018) (Zavodszky et al., 2014). Tau can be degraded by all autophagic pathways: chaperone mediated autophagy, macroautophagy and microautophagy (Caballero et al., 2018). Autophagy sequesters and degrades cytosolic aggregated seeds, alleviating aggregated tau load and cell toxicity (Menzies et al., 2017). In addition, autophagy also clears damaged endolysosomal compartments (Papadopoulos and Meyer, 2017), preventing the leakage of lysosomal proteases to the cytosol which can trigger death programs (Boya and Kroemer, 2008). Hence, the upregulation of VPS35 and SNX4 might be part of a protective autophagic response to remove aggregated tau and/or damaged endolysosomal compartments, thereby preventing cell toxicity and death.

The measured upregulation of VPS35 and SNX4 upon the seeding of tau pathology (Ch.4 Figure S5) might have a presynaptic component, since both these proteins are located at presynaptic terminals (Ch.2 Figure 1, and Ch.3 Figure 4). In AD brains, pathologically phosphorylated and misfolded tau localizes to presynaptic terminals (Tai et al., 2014),

144 General Discussion and accumulates in synaptic vesicles (McInnes et al., 2018). The association of tau with synaptic vesicles through Synaptogyrin-3 impairs vesicle mobility and neurotransmitter release, and thereby impairs synaptic function (McInnes et al., 2018). In vitro, inducing synaptic contacts enhances tau pathology spreading, and reducing synaptic formation and activity decreases the spreading of tau pathology (Calafate et al., 2015). In vivo, increasing neuronal activity results in enhanced tau spreading, pathology and cell atrophy (Wu et al., 2016). Hence, the presynaptic terminal is a key compartment for tau pathology toxicity and spreading. Presynaptic VPS35 and SNX4 might be involved in decreasing presynaptic pathological tau load via autophagy, which might protect from tau pathology damage, spreading and disease progression.

FUTURE DIRECTIONS

Endosomal sorting proteins are ubiquitously expressed but their expression level vary between different brain regions and cell types. In the adult murine brain, VPS35is mainly expressed in excitatory pyramidal neurons in layers IV–V in the cortex and CA2–3 regions of the hippocampus whereas it is not enriched within dopaminergic neurons of the substantia nigra and interneurons (Tsika et al., 2014; Wen et al., 2011). Endogenous SNX4 was expressed in all tested brain areas and cell types and it appears enriched in cortex, cerebellum and hypothalamus (Ch.3 Figure 1). SNX4 is expressed in glutamatergic hippocampal neurons and synapses (Ch.3 Figure S5) but its depletion led to a decrease in both GAD1 and GAD2, suggesting a role in GABAergic neurons (Ch.3 Figure 8 and 9). Within synapses, VPS35 and SNX4 are more abundant in the presynaptic terminal, but they are also present at postsynaptic compartments (Bhalla et al., 2012; Inoshita et al., 2017) (Ch. 2 Figure 1 and Ch.3 Figure 4). Building an atlas to define the regional, cellular and subcellular localization of endolysosomal sorting proteins would aid to unravel their function.

The lack of a clear definition of the synaptic endolysosomal system is one ofthe complicating factors to study these organelles. Many endolysosomal proteins are located at presynaptic terminals (Ch. 1) (Takamori et al., 2006; Wilhelm et al., 2014), while some of the canonical endolysosomal proteins are not, such as the transferrin receptor and EEA1 (Cameron et al., 1991) (Wilson et al., 2000). Hence, the presynaptic endolysosomal machinery differs from the canonical one, likely to fulfill specific requirements. Characterizing the presynaptic endolysosomal machinery is challenging due to the small volume of the presynaptic terminal (0.37 µm3) (Wilhelm et al., 2014). Optical super resolution techniques such as STED can map organelles and proteins within presynaptic 5 terminals (Willig et al., 2006), making them a suitable tool for characterizing the presynaptic endolysosomal machinery. Another challenge in synaptic endosolysosomal studies

145 Chapter 5 is the potential diversity of the presynaptic endolysosomal machinery. For example, VPS35 depletion decreased the number of synaptic vesicles in Drosophila, which was interpreted as a defect in endocytosis and regeneration of synaptic vesicles (Inoshita et al., 2017). The differences between this study and our data suggest that retromer function may vary in different organisms or in different synapses, or it may change during development. Hence, we recommend to map the presynaptic endolysosomal machinery at different developmental points, and in different synapses. Furthermore, neurons can be challenged with different stimulation protocols to better understand the role of the endolysosomal machinery upon different presynaptic physiological demands. Life cell imaging with STED has been able to resolve synaptic nanoscale processes in real time (Westphal et al., 2008) and may provide the required dynamic information. Such data will aid to uncover how endolysosomal sorting fulfills presynaptic physiological demands and thus where endolysosomal failure can be expected to affect the brain in disease.

Understanding the presynaptic endolysosomal sorting machinery might be critical to unravel neurodegeneration mechanisms. Synapse loss is a hallmark of neurogenerative disorders that occurs early in disease progression (Forner et al., 2017). Lysosomal storage disorders (LSD) often undergo synaptic loss and neurodegeneration (Futerman and Van Meer, 2004), where lysosomal dysfunction leads to presynaptic alterations which precede synaptic loss (Sambri et al., 2017). In addition, presynaptic endolysosomal sorting might be key to regulate APP amylogenic cleavage and tau induced synaptic dysfunction as described above. Therefore, addressing presynaptic endolysosomal sorting might provide a better understanding of the pathogenesis of AD and other neurodegenerative disorders, thereby providing potential targets for therapeutic intervention at an early stage of disease progression that precedes irreversible cell death.

146 REFERENCES Aguiar, S., van der Gaag, B., and Cortese, F.A.B. (2017). RNAi mechanisms in Huntington’s disease therapy: siRNA versus shRNA. Translational neurodegeneration 6, 30. Alvarez, V.A., Ridenour, D.A., and Sabatini, B.L. (2006). Retraction of synapses and dendritic spines induced by off-target effects of RNA interference. J Neurosci 26, 7820-7825. Arighi, C.N., Hartnell, L.M., Aguilar, R.C., Haft, C.R., and Bonifacino, J.S. (2004). Role of the mammalian retromer in sorting of the cation-independent mannose 6-phosphate receptor. J Cell Biol 165, 123-133. Arora, S., Saarloos, I., Kooistra, R., van de Bospoort, R., Verhage, M., and Toonen, R.F. (2017). SNAP-25 gene family members differentially support secretory vesicle fusion. J Cell Sci 130, 1877- 1889. Asai, H., Ikezu, S., Tsunoda, S., Medalla, M., Luebke, J., Haydar, T., Wolozin, B., Butovsky, O., Kügler, S., and Ikezu, T. (2015). Depletion of microglia and inhibition of exosome synthesis halt tau propagation. Nat Neurosci. Aschenbrenner, A.J., Gordon, B.A., Benzinger, T.L., Morris, J.C., and Hassenstab, J.J. (2018). Influence of tau PET, amyloid PET, and hippocampal volume on cognition in Alzheimer disease. Neurology 91, e859-e866. Association, A.A.s. (2018). 2018 Alzheimer’s disease facts and figures. Alzheimer’s & Dementia 14, 367-429. Atwood, B.K., Lovinger, D.M., and Mathur, B.N. (2014). Presynaptic long-term depression mediated by G i/o-coupled receptors. Trends Neurosci 37, 663-673. Babst, M., Katzmann, D.J., Estepa-Sabal, E.J., Meerloo, T., and Emr, S.D. (2002a). Escrt-III: an endosome-associated heterooligomeric protein complex required for mvb sorting. Dev Cell 3, 271- 282. Babst, M., Katzmann, D.J., Snyder, W.B., Wendland, B., and Emr, S.D. (2002b). Endosome- associated complex, ESCRT-II, recruits transport machinery for protein sorting at the multivesicular body. Dev Cell 3, 283-289. Baek, S.T., Kerjan, G., Bielas, S.L., Lee, J.E., Fenstermaker, A.G., Novarino, G., and Gleeson, J.G. (2014). Off-target effect of doublecortin family shRNA on neuronal migration associated with endogenous microRNA dysregulation. Neuron 82, 1255-1262. Ballatore, C., Lee, V.M.-Y., and Trojanowski, J.Q. (2007). Tau-mediated neurodegeneration in Alzheimer’s disease and related disorders. Nature Reviews Neuroscience 8, 663-672. Barnes, D.E., and Yaffe, K. (2011). The projected effect of risk factor reduction on Alzheimer’s disease prevalence. The Lancet Neurology 10, 819-828. Bhalla, A., Vetanovetz, C.P., Morel, E., Chamoun, Z., Di Paolo, G., and Small, S.A. (2012). The location and trafficking routes of the neuronal retromer and its role in amyloid precursor protein transport. Neurobiol Dis 47, 126-134. Bolte, S., and Cordelieres, F. (2006). A guided tour into subcellular colocalization analysis in light microscopy. J Microsc 224, 213-232. Boya, P., and Kroemer, G. (2008). Lysosomal membrane permeabilization in cell death. Oncogene 27, 6434. Braak, H., and Braak, E. (1991). Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 82, 239-259. Braulke, T., and Bonifacino, J.S. (2009). Sorting of lysosomal proteins. Biochimica et Biophysica Acta (BBA)-Molecular Cell Research 1793, 605-614. Bright, N.A., Davis, L.J., and Luzio, J.P. (2016). Endolysosomes are the principal intracellular sites of acid hydrolase activity. Curr Biol 26, 2233-2245.

147 Bruderer, R., Bernhardt, O.M., Gandhi, T., Miladinović, S.M., Cheng, L.-Y., Messner, S., Ehrenberger, T., Zanotelli, V., Butscheid, Y., and Escher, C. (2015). Extending the limits of quantitative proteome profiling with data-independent acquisition and application to acetaminophen-treated three-dimensional liver microtissues. Mol Cell Proteomics 14, 1400-1410. Burd, C., and Cullen, P.J. (2014). Retromer: a master conductor of endosome sorting. Cold Spring Harb Perspect Biol 6, a016774. Caballero, B., Wang, Y., Diaz, A., Tasset, I., Juste, Y.R., Stiller, B., Mandelkow, E.M., Mandelkow, E., and Cuervo, A.M. (2018). Interplay of pathogenic forms of human tau with different autophagic pathways. Aging cell 17, e12692. Calafate, S., Buist, A., Miskiewicz, K., Vijayan, V., Daneels, G., De Strooper, B., de Wit, J., Verstreken, P., and Moechars, D. (2015). Synaptic contacts enhance cell-to-cell tau pathology propagation. Cell reports 11, 1176-1183. Calafate, S., Flavin, W., Verstreken, P., and Moechars, D. (2016). Loss of Bin1 Promotes the Propagation of Tau Pathology. Cell Reports 17, 931-940. Cameron, P.L., Südhof, T.C., Jahn, R., and De Camilli, P. (1991). Colocalization of synaptophysin with transferrin receptors: implications for synaptic vesicle biogenesis. The Journal of cell biology 115, 151-164. Carlton, J., Bujny, M., Peter, B.J., Oorschot, V.M., Rutherford, A., Mellor, H., Klumperman, J., McMahon, H.T., and Cullen, P.J. (2004). Sorting nexin-1 mediates tubular endosome-to-TGN transport through coincidence sensing of high-curvature membranes and 3-phosphoinositides. Curr Biol 14, 1791-1800. Carlton, J., Bujny, M., Rutherford, A., and Cullen, P. (2005). Sorting nexins–unifying trends and new perspectives. Traffic 6, 75-82. Cataldo, A.M., Barnett, J.L., Berman, S.A., Li, J., Quarless, S., Bursztajn, S., Lippa, C., and Nixon, R.A. (1995). Gene expression and cellular content of cathepsin D in Alzheimer’s disease brain: evidence for early up-regulation of the endosomal-lysosomal system. Neuron 14, 671-680. Cataldo, A.M., Barnett, J.L., Pieroni, C., and Nixon, R.A. (1997). Increased neuronal endocytosis and protease delivery to early endosomes in sporadic Alzheimer’s disease: neuropathologic evidence for a mechanism of increased β-amyloidogenesis. J Neurosci 17, 6142-6151. Cataldo, A.M., Peterhoff, C.M., Troncoso, J.C., Gomez-Isla, T., Hyman, B.T., and Nixon, R.A. (2000). Endocytic pathway abnormalities precede amyloid β deposition in sporadic Alzheimer’s disease and Down syndrome: differential effects of APOE genotype and presenilin mutations. The American journal of pathology 157, 277-286. Ceccarelli, B., Hurlbut, W., and Mauro, A. (1973). Turnover of transmitter and synaptic vesicles at the frog neuromuscular junction. The Journal of cell biology 57, 499-524. Chai, Y.L., Chong, J.R., Weng, J., Howlett, D., Halsey, A., Lee, J.H., Attems, J., Aarsland, D., Francis, P.T., and Chen, C.P. (2018). Lysosomal cathepsin D is upregulated in Alzheimer’s disease neocortex and may be a marker for neurofibrillary degeneration. Brain Pathol. Chiaruttini, N., Redondo-Morata, L., Colom, A., Humbert, F., Lenz, M., Scheuring, S., and Roux, A. (2015). Relaxation of loaded ESCRT-III spiral springs drives membrane deformation. Cell 163, 866-879. Choy, R.W.-Y., Park, M., Temkin, P., Herring, B.E., Marley, A., Nicoll, R.A., and von Zastrow, M. (2014). Retromer mediates a discrete route of local membrane delivery to dendrites. Neuron 82, 55-62. Christoforidis, S., McBride, H.M., Burgoyne, R.D., and Zerial, M. (1999a). The Rab5 effector EEA1 is a core component of endosome docking. Nature 397, 621. Christoforidis, S., Miaczynska, M., Ashman, K., Wilm, M., Zhao, L., Yip, S.-C., Waterfield, M.D., Backer, J.M., and Zerial, M. (1999b). Phosphatidylinositol-3-OH kinases are Rab5 effectors. Nat Cell Biol 1, 249. Chu, J., and Praticò, D. (2017). The retromer complex system in a transgenic mouse model of AD: influence of age. Neurobiol Aging.

148 Cirrito, J.R., Kang, J.-E., Lee, J., Stewart, F.R., Verges, D.K., Silverio, L.M., Bu, G., Mennerick, S., and Holtzman, D.M. (2008). Endocytosis is required for synaptic activity-dependent release of amyloid-β in vivo. Neuron 58, 42-51. Cirrito, J.R., Yamada, K.A., Finn, M.B., Sloviter, R.S., Bales, K.R., May, P.C., Schoepp, D.D., Paul, S.M., Mennerick, S., and Holtzman, D.M. (2005). Synaptic activity regulates interstitial fluid amyloid-β levels in vivo. Neuron 48, 913-922. Clague, M.J., Liu, H., and Urbé, S. (2012). Governance of endocytic trafficking and signaling by reversible ubiquitylation. Dev Cell 23, 457-467. Colacurcio, D.J., Pensalfini, A., Jiang, Y., and Nixon, R.A. (2017). Dysfunction of autophagy and endosomal-lysosomal pathways: Roles in pathogenesis of Down syndrome and Alzheimer’s Disease. Free Radic Biol Med. Cui, Y., Carosi, J.M., Yang, Z., Ariotti, N., Kerr, M.C., Parton, R.G., Sargeant, T.J., and Teasdale, R.D. (2018). Retromer has a selective function in cargo sorting via endosome transport carriers. J Cell Biol, jcb. 201806153. Cullen, P.J. (2008). Endosomal sorting and signalling: an emerging role for sorting nexins. Nature Reviews Molecular Cell Biology 9, 574-582. Cullen, P.J., and Korswagen, H.C. (2012). Sorting nexins provide diversity for retromer-dependent trafficking events. Nat Cell Biol 14, 29-37. Cullen, P.J., and Steinberg, F. (2018). To degrade or not to degrade: mechanisms and significance of endocytic recycling. Nature reviews Molecular cell biology. Dacks, J.B., and Field, M.C. (2007). Evolution of the eukaryotic membrane-trafficking system: origin, tempo and mode. J Cell Sci 120, 2977-2985. Das, U., Scott, D.A., Ganguly, A., Koo, E.H., Tang, Y., and Roy, S. (2013). Activity-induced convergence of APP and BACE-1 in acidic microdomains via an endocytosis-dependent pathway. Neuron 79, 447-460. Das, U., Wang, L., Ganguly, A., Saikia, J.M., Wagner, S.L., Koo, E.H., and Roy, S. (2016). Visualizing APP and BACE-1 approximation in neurons yields insight into the amyloidogenic pathway. Nat Neurosci 19, 55-64. Di Giovanni, J., and Sheng, Z.H. (2015). Regulation of synaptic activity by snapin‐mediated endolysosomal transport and sorting. The EMBO journal 34, 2059-2077. Dyve, A.B., Bergan, J., Utskarpen, A., and Sandvig, K. (2009). Sorting nexin 8 regulates endosome- to-Golgi transport. Biochem Biophys Res Commun 390, 109-114. Eisen, J.S., and Smith, J.C. (2008). Controlling morpholino experiments: don’t stop making antisense. Development 135, 1735-1743. Farina, M., van de Bospoort, R., He, E., Persoon, C.M., van Weering, J.R., Broeke, J.H., Verhage, M., and Toonen, R.F. (2015). CAPS-1 promotes fusion competence of stationary dense-core vesicles in presynaptic terminals of mammalian neurons. Elife 4. Fellmann, C., and Lowe, S.W. (2014). Stable RNA interference rules for silencing. Nat Cell Biol 16, 10-18. Fernandes, A.C., Uytterhoeven, V., Kuenen, S., Wang, Y.-C., Slabbaert, J.R., Swerts, J., Kasprowicz, J., Aerts, S., and Verstreken, P. (2014). Reduced synaptic vesicle protein degradation at lysosomes curbs TBC1D24/sky-induced neurodegeneration. J Cell Biol 207, 453-462. Fjorback, A.W., Seaman, M., Gustafsen, C., Mehmedbasic, A., Gokool, S., Wu, C., Militz, D., Schmidt, V., Madsen, P., and Nyengaard, J.R. (2012). Retromer binds the FANSHY sorting motif in SorLA to regulate amyloid precursor protein sorting and processing. J Neurosci 32, 1467-1480. Flavin, W.P., Bousset, L., Green, Z.C., Chu, Y., Skarpathiotis, S., Chaney, M.J., Kordower, J.H., Melki, R., and Campbell, E.M. (2017). Endocytic vesicle rupture is a conserved mechanism of cellular invasion by amyloid proteins. Acta Neuropathol 134, 629-653.

149 Follett, J., Bugarcic, A., Yang, Z., Ariotti, N., Norwood, S.J., Collins, B.M., Parton, R.G., and Teasdale, R.D. (2016). Parkinson Disease-linked Vps35 R524W Mutation Impairs the Endosomal Association of Retromer and Induces α-Synuclein Aggregation. J Biol Chem 291, 18283-18298. Forner, S., Baglietto-Vargas, D., Martini, A.C., Trujillo-Estrada, L., and LaFerla, F.M. (2017). Synaptic impairment in Alzheimer’s disease: a dysregulated symphony. Trends Neurosci 40, 347- 357. Franch-Marro, X., Wendler, F., Guidato, S., Griffith, J., Baena-Lopez, A., Itasaki, N., Maurice, M.M., and Vincent, J.-P. (2008). Wingless secretion requires endosome-to-Golgi retrieval of Wntless/ Evi/Sprinter by the retromer complex. Nat Cell Biol 10, 170-177. Frost, B., Jacks, R.L., and Diamond, M.I. (2009). Propagation of tau misfolding from the outside to the inside of a cell. J Biol Chem 284, 12845-12852. Funk, K.E., Mrak, R.E., and Kuret, J. (2011). Granulovacuolar degeneration (GVD) bodies of Alzheimer’s disease (AD) resemble late‐stage autophagic organelles. Neuropathol Appl Neurobiol 37, 295-306. Futerman, A.H., and Van Meer, G. (2004). The cell biology of lysosomal storage disorders. Nature reviews Molecular cell biology 5, 554. Galluzzi, L., Baehrecke, E.H., Ballabio, A., Boya, P., Bravo‐San Pedro, J.M., Cecconi, F., Choi, A.M., Chu, C.T., Codogno, P., and Colombo, M.I. (2017). Molecular definitions of autophagy and related processes. The EMBO journal 36, 1811-1836. Galvez, T., Gilleron, J., Zerial, M., and O’Sullivan, G.A. (2012). SnapShot: Mammalian Rab proteins in endocytic trafficking. Cell 151, 234-234. e232. Gan, Q., and Watanabe, S. (2018). Synaptic vesicle endocytosis in different model systems. Front Cell Neurosci 12, 171. Gao, S., Casey, A.E., Sargeant, T.J., and Mäkinen, V.-P. (2018). Genetic variation within endolysosomal system is associated with late-onset Alzheimer’s disease. Brain 141, 2711-2720. Gillet, L.C., Navarro, P., Tate, S., Röst, H., Selevsek, N., Reiter, L., Bonner, R., and Aebersold, R. (2012). Targeted data extraction of the MS/MS spectra generated by data-independent acquisition: a new concept for consistent and accurate proteome analysis. Mol Cell Proteomics 11, O111. 016717. Ginsberg, S.D., Alldred, M.J., Counts, S.E., Cataldo, A.M., Neve, R.L., Jiang, Y., Wuu, J., Chao, M.V., Mufson, E.J., and Nixon, R.A. (2010a). Microarray analysis of hippocampal CA1 neurons implicates early endosomal dysfunction during Alzheimer’s disease progression. Biol Psychiatry 68, 885-893. Ginsberg, S.D., Mufson, E.J., Counts, S.E., Wuu, J., Alldred, M.J., Nixon, R.A., and Che, S. (2010b). Regional selectivity of rab5 and rab7 protein upregulation in mild cognitive impairment and Alzheimer’s disease. J Alzheimers Dis 22, 631-639. Goedert, M., Jakes, R., and Vanmechelen, E. (1995). Monoclonal antibody AT8 recognises tau protein phosphorylated at both serine 202 and threonine 205. Neurosci Lett 189, 167-170. Gómez-Sintes, R., Ledesma, M.D., and Boya, P. (2016). Lysosomal cell death mechanisms in aging. Ageing research reviews 32, 150-168. Gonzalez, A., Valeiras, M., Sidransky, E., and Tayebi, N. (2014). Lysosomal integral membrane protein-2: a new player in lysosome-related pathology. Mol Genet Metab 111, 84-91. Granseth, B., Odermatt, B., Royle, S.J., and Lagnado, L. (2006). Clathrin-mediated endocytosis is the dominant mechanism of vesicle retrieval at hippocampal synapses. Neuron 51, 773-786. Guo, J.L., Buist, A., Soares, A., Callaerts, K., Calafate, S., Stevenaert, F., Daniels, J.P., Zoll, B.E., Crowe, A., and Brunden, K.R. (2016). The dynamics and turnover of tau aggregates in cultured cells: insights into therapies for tauopathies. J Biol Chem, jbc. M115. 712083. Guo, J.L., and Lee, V.M. (2013). Neurofibrillary tangle‐like tau pathology induced by synthetic tau fibrils in primary neurons over‐expressing mutant tau. FEBS Lett 587, 717-723. Guo, J.L., and Lee, V.M. (2014). Cell-to-cell transmission of pathogenic proteins in neurodegenerative diseases. Nat Med 20, 130.

150 Guo, J.L., and Lee, V.M.-Y. (2011). Seeding of normal Tau by pathological Tau conformers drives pathogenesis of Alzheimer-like tangles. J Biol Chem 286, 15317-15331. Haft, C.R., de la Luz Sierra, M., Bafford, R., Lesniak, M.A., Barr, V.A., and Taylor, S.I. (2000). Human orthologs of yeast vacuolar protein sorting proteins Vps26, 29, and 35: assembly into multimeric complexes. Mol Biol Cell 11, 4105-4116. Hardy, J., and Selkoe, D.J. (2002). The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics. Science 297, 353-356. Harris, J.J., Jolivet, R., and Attwell, D. (2012). Synaptic energy use and supply. Neuron 75, 762- 777. Harterink, M., Port, F., Lorenowicz, M.J., McGough, I.J., Silhankova, M., Betist, M.C., Van Weering, J.R., Van Heesbeen, R.G., Middelkoop, T.C., and Basler, K. (2011). A SNX3-dependent retromer pathway mediates retrograde transport of the Wnt sorting receptor Wntless and is required for Wnt secretion. Nat Cell Biol 13, 914. He, E., Lozano, M.A.G., Stringer, S., Watanabe, K., Sakamoto, K., den Oudsten, F., Koopmans, F., Giamberardino, S.N., Hammerschlag, A., and Cornelisse, L.N. (2018). MIR137 schizophrenia- associated locus controls synaptic function by regulating synaptogenesis, synapse maturation and synaptic transmission. Hum Mol Genet. He, E., Wierda, K., van Westen, R., Broeke, J.H., Toonen, R.F., Cornelisse, L.N., and Verhage, M. (2017). Munc13-1 and Munc18-1 together prevent NSF-dependent de-priming of synaptic vesicles. Nature communications 8, ncomms15915. Hettema, E.H., Lewis, M.J., Black, M.W., and Pelham, H.R. (2003). Retromer and the sorting nexins Snx4/41/42 mediate distinct retrieval pathways from yeast endosomes. The EMBO journal 22, 548-557. Heuser, J., and Reese, T. (1973). Evidence for recycling of synaptic vesicle membrane during transmitter release at the frog neuromuscular junction. The Journal of cell biology 57, 315-344. Hierro, A., Rojas, A.L., Rojas, R., Murthy, N., Effantin, G., Kajava, A.V., Steven, A.C., Bonifacino, J.S., and Hurley, J.H. (2007). Functional architecture of the retromer cargo-recognition complex. Nature 449, 1063-1067. Höglinger, G.U., Melhem, N.M., Dickson, D.W., Sleiman, P.M., Wang, L.-S., Klei, L., Rademakers, R., De Silva, R., Litvan, I., and Riley, D.E. (2011). Identification of common variants influencing risk of the tauopathy progressive supranuclear palsy. Nat Genet 43, 699. Hoopmann, P., Punge, A., Barysch, S.V., Westphal, V., Bückers, J., Opazo, F., Bethani, I., Lauterbach, M.A., Hell, S.W., and Rizzoli, S.O. (2010). Endosomal sorting of readily releasable synaptic vesicles. Proceedings of the National Academy of Sciences 107, 19055-19060. Hsieh, H., Boehm, J., Sato, C., Iwatsubo, T., Tomita, T., Sisodia, S., and Malinow, R. (2006). AMPAR removal underlies Aβ-induced synaptic depression and dendritic spine loss. Neuron 52, 831-843. Hussain, N.K., Diering, G.H., Sole, J., Anggono, V., and Huganir, R.L. (2014). Sorting Nexin 27 regulates basal and activity-dependent trafficking of AMPARs. Proceedings of the National Academy of Sciences 111, 11840-11845. Im, Y.J., and Hurley, J.H. (2008). Integrated structural model and membrane targeting mechanism of the human ESCRT-II complex. Dev Cell 14, 902-913. Inoshita, T., Arano, T., Hosaka, Y., Meng, H., Umezaki, Y., Kosugi, S., Morimoto, T., Koike, M., Chang, H.-Y., and Imai, Y. (2017). Vps35 in cooperation with LRRK2 regulates synaptic vesicle endocytosis through the endosomal pathway in Drosophila. Hum Mol Genet, ddx179. Irannejad, R., and von Zastrow, M. (2014). GPCR signaling along the endocytic pathway. Curr Opin Cell Biol 27, 109-116. Israel, M.A., Yuan, S.H., Bardy, C., Reyna, S.M., Mu, Y., Herrera, C., Hefferan, M.P., Van Gorp, S., Nazor, K.L., and Boscolo, F.S. (2012). Probing sporadic and familial Alzheimer’s disease using induced pluripotent stem cells. Nature 482, 216.

151 Jahic, A., Khundadze, M., Jaenisch, N., Schüle, R., Klimpe, S., Klebe, S., Frahm, C., Kassubek, J., Stevanin, G., and Schöls, L. (2015). The spectrum of KIAA0196 variants, and characterization of a murine knockout: implications for the mutational mechanism in hereditary spastic paraplegia type SPG8. Orphanet J Rare Dis 10, 147. Jicha, G.A., Bowser, R., Kazam, I.G., and Davies, P. (1997). Alz‐50 and MC‐1, a new monoclonal antibody raised to paired helical filaments, recognize conformational epitopes on recombinant tau. J Neurosci Res 48, 128-132. Kaelin, W.G. (2012). Use and abuse of RNAi to study mammalian gene function. Science 337, 421-422. Kamenetz, F., Tomita, T., Hsieh, H., Seabrook, G., Borchelt, D., Iwatsubo, T., Sisodia, S., and Malinow, R. (2003). APP processing and synaptic function. Neuron 37, 925-937. Karch, C.M., and Goate, A.M. (2015). Alzheimer’s disease risk genes and mechanisms of disease pathogenesis. Biol Psychiatry 77, 43-51. Katzmann, D.J., Babst, M., and Emr, S.D. (2001). Ubiquitin-dependent sorting into the multivesicular body pathway requires the function of a conserved endosomal protein sorting complex, ESCRT-I. Cell 106, 145-155. Keimpema, L., Kooistra, R., Toonen, R.F., and Verhage, M. (2017). CAPS-1 requires its C2, PH, MHD1 and DCV domains for dense core vesicle exocytosis in mammalian CNS neurons. Sci Rep 7, 10817. Kim, N.-Y., Cho, M.-H., Won, S.-H., Kang, H.-J., Yoon, S.-Y., and Kim, D.-H. (2017). Sorting nexin-4 regulates β-amyloid production by modulating β-site-activating cleavage enzyme-1. Alzheimers Res Ther 9, 4. Klumperman, J., and Raposo, G. (2014). The complex ultrastructure of the endolysosomal system. Cold Spring Harb Perspect Biol, a016857. Kneussel, M., and Hausrat, T.J. (2016). Postsynaptic neurotransmitter receptor reserve pools for synaptic potentiation. Trends Neurosci 39, 170-182. Köhler, C. (2016). Granulovacuolar degeneration: a neurodegenerative change that accompanies tau pathology. Acta Neuropathol 132, 339-359. Kononenko, N.L., Puchkov, D., Classen, G.A., Walter, A.M., Pechstein, A., Sawade, L., Kaempf, N., Trimbuch, T., Lorenz, D., and Rosenmund, C. (2014). Clathrin/AP-2 mediate synaptic vesicle reformation from endosome-like vacuoles but are not essential for membrane retrieval at central synapses. Neuron 82, 981-988. Koopmans, F., Ho, J.T., Smit, A.B., and Li, K.W. (2018). Comparative Analyses of Data Independent Acquisition Mass Spectrometric Approaches: DIA, WiSIM‐DIA, and Untargeted DIA. Proteomics 18. Korolchuk, V.I., Schütz, M.M., Gómez-Llorente, C., Rocha, J., Lansu, N.R., Collins, S.M., Wairkar, Y.P., Robinson, I.M., and O’Kane, C.J. (2007). Drosophila Vps35 function is necessary for normal endocytic trafficking and actin cytoskeleton organisation. J Cell Sci 120, 4367-4376. Kostelansky, M.S., Sun, J., Lee, S., Kim, J., Ghirlando, R., Hierro, A., Emr, S.D., and Hurley, J.H. (2006). Structural and functional organization of the ESCRT-I trafficking complex. Cell 125, 113-126. Koumandou, V.L., Klute, M.J., Herman, E.K., Nunez-Miguel, R., Dacks, J.B., and Field, M.C. (2011). Evolutionary reconstruction of the retromer complex and its function in Trypanosoma brucei. J Cell Sci 124, 1496-1509. Kovacs, G.G., Ferrer, I., Grinberg, L.T., Alafuzoff, I., Attems, J., Budka, H., Cairns, N.J., Crary, J.F., Duyckaerts, C., and Ghetti, B. (2016). Aging-related tau astrogliopathy (ARTAG): harmonized evaluation strategy. Acta Neuropathol 131, 87-102. Kovtun, O., Leneva, N., Bykov, Y.S., Ariotti, N., Teasdale, R.D., Schaffer, M., Engel, B.D., Owen, D.J., Briggs, J.A., and Collins, B.M. (2018). Structure of the membrane-assembled retromer coat determined by cryo-electron tomography. Nature, 1. Kvainickas, A., Jimenez-Orgaz, A., Nägele, H., Hu, Z., Dengjel, J., and Steinberg, F. (2017). Cargo- selective SNX-BAR proteins mediate retromer trimer independent retrograde transport. J Cell Biol, jcb. 201702137.

152 Lashuel, H.A., Overk, C.R., Oueslati, A., and Masliah, E. (2013). The many faces of α-synuclein: from structure and toxicity to therapeutic target. Nature Reviews Neuroscience 14, 38-48. Laßek, M., Weingarten, J., Einsfelder, U., Brendel, P., Müller, U., and Volknandt, W. (2013). Amyloid precursor proteins are constituents of the presynaptic active zone. J Neurochem 127, 48-56. Laßek, M., Weingarten, J., Wegner, M., and Volknandt, W. (2015). The amyloid precursor protein—a novel player within the molecular array of presynaptic nanomachines. Front Synaptic Neurosci 7. Lee, J.J., Radice, G., Perkins, C.P., and Costantini, F. (1992). Identification and characterization of a novel, evolutionarily conserved gene disrupted by the murine H beta 58 embryonic lethal transgene insertion. Development 115, 277-288. Leprince, C., Le Scolan, E., Meunier, B., Fraisier, V., Brandon, N., De Gunzburg, J., and Camonis, J. (2003). Sorting nexin 4 and amphiphysin 2, a new partnership between endocytosis and intracellular trafficking. J Cell Sci 116, 1937-1948. Liu, W., Tang, F.-L., Erion, J., Xiao, H., Ye, J., and Xiong, W.-C. (2014). Vps35 haploinsufficiency results in degenerative-like deficit in mouse retinal ganglion neurons and impairment of optic nerve injury-induced gliosis. Molecular brain 7, 1. Lou, X. (2018). Sensing Exocytosis and Triggering Endocytosis at Synapses: Synaptic Vesicle Exocytosis–Endocytosis Coupling. Front Cell Neurosci 12, 66. Maritzen, T., and Haucke, V. (2017). Coupling of exocytosis and endocytosis at the presynaptic active zone. Neurosci Res. Marquez-Sterling, N.R., Lo, A.C., Sisodia, S.S., and Koo, E.H. (1997). Trafficking of cell-surface β-amyloid precursor protein: evidence that a sorting intermediate participates in synaptic vesicle recycling. J Neurosci 17, 140-151. McInnes, J., Wierda, K., Snellinx, A., Bounti, L., Wang, Y.-C., Stancu, I.-C., Apóstolo, N., Gevaert, K., Dewachter, I., and Spires-Jones, T.L. (2018). Synaptogyrin-3 mediates presynaptic dysfunction induced by tau. Neuron 97, 823-835. e828. McNally, K.E., and Cullen, P.J. (2018). Endosomal retrieval of cargo: retromer is not alone. Trends Cell Biol. Mecozzi, V.J., Berman, D.E., Simoes, S., Vetanovetz, C., Awal, M.R., Patel, V.M., Schneider, R.T., Petsko, G.A., Ringe, D., and Small, S.A. (2014). Pharmacological chaperones stabilize retromer to limit APP processing. Nat Chem Biol 10, 443-449. Meijer, M., Burkhardt, P., De Wit, H., Toonen, R.F., Fasshauer, D., and Verhage, M. (2012). Munc18‐1 mutations that strongly impair SNARE‐complex binding support normal synaptic transmission. The EMBO journal 31, 2156-2168. Meijer, M., Dörr, B., Lammertse, H.C., Blithikioti, C., van Weering, J.R., Toonen, R.F., Söllner, T.H., and Verhage, M. (2017). Tyrosine phosphorylation of Munc18‐1 inhibits synaptic transmission by preventing SNARE assembly. The EMBO journal, e201796484. Melero, J.E., Nadadhur, A.G., Schut, D., Weering, J.V., Heine, V.M., Toonen, R.F., and Verhage, M. (2017). Differential Maturation of the Two Regulated Secretory Pathways in Human iPSC-Derived Neurons. Stem cell reports 8, 659-672. Meltsje, J., Huber, L.A., Stenmark, H., Williamson, E., Zerial, M., Parton, R.G., and Dotti, C.G. (1994). The involvement of the small GTP-binding protein Rab5a in neuronal endocytosis. Neuron 13, 11-22. Menzies, F.M., Fleming, A., Caricasole, A., Bento, C.F., Andrews, S.P., Ashkenazi, A., Füllgrabe, J., Jackson, A., Sanchez, M.J., and Karabiyik, C. (2017). Autophagy and neurodegeneration: pathogenic mechanisms and therapeutic opportunities. Neuron 93, 1015-1034. Miesenböck, G., De Angelis, D.A., and Rothman, J.E. (1998). Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins. Nature 394, 192-195. Milosevic, I. (2018). Revisiting the role of clathrin-mediated endoytosis in synaptic vesicle recycling. Front Cell Neurosci 12, 27.

153 Miura, E., Hasegawa, T., Konno, M., Suzuki, M., Sugeno, N., Fujikake, N., Geisler, S., Tabuchi, M., Oshima, R., and Kikuchi, A. (2014). VPS35 dysfunction impairs lysosomal degradation of α-synuclein and exacerbates neurotoxicity in a Drosophila model of Parkinson’s disease. Neurobiol Dis 71, 1-13. Morel, E., Chamoun, Z., Lasiecka, Z.M., Chan, R.B., Williamson, R.L., Vetanovetz, C., Dall’Armi, C., Simoes, S., Du Jour, K.S.P., and McCabe, B.D. (2013). Phosphatidylinositol-3-phosphate regulates sorting and processing of amyloid precursor protein through the endosomal system. Nature communications 4, 2250. Morgan, J.R., Comstra, H.S., Cohen, M., and Faundez, V. (2013). Presynaptic membrane retrieval and endosome biology: defining molecularly heterogeneous synaptic vesicles. Cold Spring Harb Perspect Biol 5, a016915. Muhammad, A., Flores, I., Zhang, H., Yu, R., Staniszewski, A., Planel, E., Herman, M., Ho, L., Kreber, R., and Honig, L.S. (2008). Retromer deficiency observed in Alzheimer’s disease causes hippocampal dysfunction, neurodegeneration, and Aβ accumulation. Proceedings of the National Academy of Sciences 105, 7327-7332. Mukrasch, M.D., Biernat, J., von Bergen, M., Griesinger, C., Mandelkow, E., and Zweckstetter, M. (2005). Sites of tau important for aggregation populate β-structure and bind to microtubules and polyanions. J Biol Chem. Müller, U.C., Deller, T., and Korte, M. (2017). Not just amyloid: physiological functions of the amyloid precursor protein family. Nature Reviews Neuroscience 18, 281. Munsie, L., Milnerwood, A., Seibler, P., Beccano-Kelly, D., Tatarnikov, I., Khinda, J., Volta, M., Kadgien, C., Cao, L., and Tapia, L. (2015). Retromer-dependent neurotransmitter receptor trafficking to synapses is altered by the Parkinson’s disease VPS35 mutation p. D620N. Hum Mol Genet 24, 1691-1703. Murphy, D.D., Rueter, S.M., Trojanowski, J.Q., and Lee, V.M.-Y. (2000). Synucleins are developmentally expressed, and α-synuclein regulates the size of the presynaptic vesicular pool in primary hippocampal neurons. J Neurosci 20, 3214-3220. Naj, A.C., Schellenberg, G.D., and Consortium, A.s.D.G. (2017). Genomic variants, genes, and pathways of Alzheimer’s disease: an overview. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 174, 5-26. Nijholt, D.A., van Haastert, E.S., Rozemuller, A.J., Scheper, W., and Hoozemans, J.J. (2012). The unfolded protein response is associated with early tau pathology in the hippocampus of tauopathies. The Journal of pathology 226, 693-702. Nixon, R.A., Cataldo, A.M., and Mathews, P.M. (2000). The endosomal-lysosomal system of neurons in Alzheimer’s disease pathogenesis: a review. Neurochem Res 25, 1161-1172. Nixon, R.A., Wegiel, J., Kumar, A., Yu, W.H., Peterhoff, C., Cataldo, A., and Cuervo, A.M. (2005). Extensive involvement of autophagy in Alzheimer disease: an immuno-electron microscopy study. J Neuropathol Exp Neurol 64, 113-122. Nogueras-Ortiz, C., Roman-Vendrell, C., Mateo-Semidey, G.E., Liao, Y.-H., Kendall, D.A., and Yudowski, G.A. (2017). Retromer stops beta-arrestin 1–mediated signaling from internalized cannabinoid 2 receptors. Mol Biol Cell 28, 3554-3561. Norwood, S.J., Shaw, D.J., Cowieson, N.P., Owen, D.J., Teasdale, R.D., and Collins, B.M. (2011). Assembly and solution structure of the core retromer protein complex. Traffic 12, 56-71. Pandya, N.J., Koopmans, F., Slotman, J.A., Paliukhovich, I., Houtsmuller, A.B., Smit, A.B., and Li, K.W. (2017). Correlation profiling of brain sub-cellular proteomes reveals co-assembly of synaptic proteins and subcellular distribution. Sci Rep 7, 12107. Papadopoulos, C., and Meyer, H. (2017). Detection and clearance of damaged lysosomes by the endo-lysosomal damage response and lysophagy. Curr Biol 27, R1330-R1341. Park, J., Kim, Y., Lee, S., Park, J.J., Park, Z.Y., Sun, W., Kim, H., and Chang, S. (2010). SNX18 shares a redundant role with SNX9 and modulates endocytic trafficking at the plasma membrane. J Cell Sci, jcs. 064170.

154 Piras, A., Collin, L., Grüninger, F., Graff, C., and Rönnbäck, A. (2016). Autophagic and lysosomal defects in human tauopathies: analysis of post-mortem brain from patients with familial Alzheimer disease, corticobasal degeneration and progressive supranuclear palsy. Acta neuropathologica communications 4, 22. Pols, M.S., and Klumperman, J. (2009). Trafficking and function of the tetraspanin CD63. Exp Cell Res 315, 1584-1592. Popelka, H., Damasio, A., Hinshaw, J.E., Klionsky, D.J., and Ragusa, M.J. (2017). Structure and function of yeast Atg20, a sorting nexin that facilitates autophagy induction. Proceedings of the National Academy of Sciences 114, E10112-E10121. Raiborg, C., Wesche, J., Malerød, L., and Stenmark, H. (2006). Flat clathrin coats on endosomes mediate degradative protein sorting by scaffolding Hrs in dynamic microdomains. J Cell Sci 119, 2414-2424. Reilly, P., Winston, C.N., Baron, K.R., Trejo, M., Rockenstein, E.M., Akers, J.C., Kfoury, N., Diamond, M., Masliah, E., and Rissman, R.A. (2017). Novel human neuronal tau model exhibiting neurofibrillary tangles and transcellular propagation. Neurobiol Dis 106, 222-234. Reim, K., Mansour, M., Varoqueaux, F., McMahon, H.T., Südhof, T.C., Brose, N., and Rosenmund, C. (2001). Complexins regulate a late step in Ca2+-dependent neurotransmitter release. Cell 104, 71-81. Reimand, J., Arak, T., Adler, P., Kolberg, L., Reisberg, S., Peterson, H., and Vilo, J. (2016). g: Profiler—a web server for functional interpretation of gene lists (2016 update). Nucleic Acids Res 44, W83-W89. Richards, D., and Betz, W. (2000). A possible link between phosphatidyl-inositol signaling and the actin cytoskeleton in synaptic vesicle cycling at the frog neuromuscular junction. Paper presented at: Soc Neurosci Abstr. Ritchie, M.E., Phipson, B., Wu, D., Hu, Y., Law, C.W., Shi, W., and Smyth, G.K. (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47-e47. Rizzoli, S.O., Bethani, I., Zwilling, D., Wenzel, D., Siddiqui, T.J., Brandhorst, D., and Jahn, R. (2006). Evidence for early endosome‐like fusion of recently endocytosed synaptic vesicles. Traffic 7, 1163-1176. Rizzoli, S.O., and Betz, W.J. (2002). Effects of 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4- one on synaptic vesicle cycling at the frog neuromuscular junction. J Neurosci 22, 10680-10689. Rodnina, M.V. (2016). The ribosome in action: Tuning of translational efficiency and protein folding. Protein Sci 25, 1390-1406. Rogaeva, E., Meng, Y., Lee, J.H., Gu, Y., Kawarai, T., Zou, F., Katayama, T., Baldwin, C.T., Cheng, R., and Hasegawa, H. (2007). The neuronal sortilin-related receptor SORL1 is genetically associated with Alzheimer disease. Nat Genet 39, 168. Saftig, P., and Klumperman, J. (2009). Lysosome biogenesis and lysosomal membrane proteins: trafficking meets function. Nature reviews Molecular cell biology 10, 623. Sambri, I., D’alessio, R., Ezhova, Y., Giuliano, T., Sorrentino, N.C., Cacace, V., De Risi, M., Cataldi, M., Annunziato, L., and De Leonibus, E. (2017). Lysosomal dysfunction disrupts presynaptic maintenance and restoration of presynaptic function prevents neurodegeneration in lysosomal storage diseases. EMBO Mol Med 9, 112-132. Schmitz, S.K., Hjorth, J.J., Joemai, R.M., Wijntjes, R., Eijgenraam, S., de Bruijn, P., Georgiou, C., de Jong, A.P., van Ooyen, A., and Verhage, M. (2011). Automated analysis of neuronal morphology, synapse number and synaptic recruitment. J Neurosci Methods 195, 185-193. Schmitz, S.K., King, C., Kortleven, C., Huson, V., Kroon, T., Kevenaar, J.T., Schut, D., Saarloos, I., Hoetjes, J.P., and de Wit, H. (2016). Presynaptic inhibition upon CB1 or mGlu2/3 receptor activation requires ERK/MAPK phosphorylation of Munc18‐1. The EMBO journal, e201592244. Schöneberg, J., Lee, I.-H., Iwasa, J.H., and Hurley, J.H. (2017a). Reverse-topology membrane scission by the ESCRT proteins. Nature reviews Molecular cell biology 18, 5.

155 Schöneberg, J., Lehmann, M., Ullrich, A., Posor, Y., Lo, W.-T., Lichtner, G., Schmoranzer, J., Haucke, V., and Noé, F. (2017b). Lipid-mediated PX-BAR domain recruitment couples local membrane constriction to endocytic vesicle fission. Nature communications 8, 15873. Seaman, M.N. (2004). Cargo-selective endosomal sorting for retrieval to the Golgi requires retromer. J Cell Biol 165, 111-122. Seaman, M.N. (2018). Retromer and the cation‐independent mannose 6‐phosphate receptor— Time for a trial separation? Traffic 19, 150-152. Seaman, M.N., McCaffery, J.M., and Emr, S.D. (1998). A membrane coat complex essential for endosome-to-Golgi retrograde transport in yeast. The Journal of cell biology 142, 665-681. Shimizu, H., Kawamura, S., and Ozaki, K. (2003). An essential role of Rab5 in uniformity of synaptic vesicle size. J Cell Sci 116, 3583-3590. Shin, N., Lee, S., Ahn, N., Kim, S.-A., Ahn, S.-G., YongPark, Z., and Chang, S. (2007). Sorting nexin 9 interacts with dynamin 1 and N-WASP and coordinates synaptic vesicle endocytosis. J Biol Chem 282, 28939-28950. Simonetti, B., and Cullen, P.J. (2018). Endosomal Sorting: Architecture of the Retromer Coat. Curr Biol 28, R1350-R1352. Simonetti, B., Danson, C.M., Heesom, K.J., and Cullen, P.J. (2017). Sequence-dependent cargo recognition by SNX-BARs mediates retromer-independent transport of CI-MPR. J Cell Biol, jcb. 201703015. Small, S.A., and Gandy, S. (2006). Sorting through the cell biology of Alzheimer’s disease: intracellular pathways to pathogenesis. Neuron 52, 15-31. Small, S.A., Kent, K., Pierce, A., Leung, C., Kang, M.S., Okada, H., Honig, L., Vonsattel, J.P., and Kim, T.W. (2005). Model‐guided microarray implicates the retromer complex in Alzheimer’s disease. Ann Neurol 58, 909-919. Small, S.A., and Petsko, G.A. (2015). Retromer in Alzheimer disease, Parkinson disease and other neurological disorders. Nature Reviews Neuroscience 16, 126-132. Solis, G.P., Hülsbusch, N., Radon, Y., Katanaev, V.L., Plattner, H., and Stuermer, C.A. (2013). Reggies/flotillins interact with Rab11a and SNX4 at the tubulovesicular recycling compartment and function in transferrin receptor and E-cadherin trafficking. Mol Biol Cell 24, 2689-2702. Spangler, S.A., Schmitz, S.K., Kevenaar, J.T., de Graaff, E., de Wit, H., Demmers, J., Toonen, R.F., and Hoogenraad, C.C. (2013). Liprin-α2 promotes the presynaptic recruitment and turnover of RIM1/ CASK to facilitate synaptic transmission. J Cell Biol 201, 915-928. Stahl, B., Von Mollard, G.F., Walch-Solimena, C., and Jahn, R. (1994). GTP cleavage by the small GTP-binding protein Rab3A is associated with exocytosis of synaptic vesicles induced by alpha- latrotoxin. J Biol Chem 269, 24770-24776. Stenmark, H. (2009). Rab GTPases as coordinators of vesicle traffic. Nature reviews Molecular cell biology 10, 513. Sudhof, T.C. (2004). The synaptic vesicle cycle. Annu Rev Neurosci 27, 509. Südhof, T.C. (2013). Neurotransmitter release: the last millisecond in the life of a synaptic vesicle. Neuron 80, 675-690. Südhof, T.C., and Rizo, J. (2011). Synaptic vesicle exocytosis. Cold Spring Harb Perspect Biol 3, a005637. Südhof, T.C., and Rothman, J.E. (2009). Membrane fusion: grappling with SNARE and SM proteins. Science 323, 474-477. Suzuki, S.W., and Emr, S.D. (2018). Membrane protein recycling from the vacuole/lysosome membrane. J Cell Biol 217, 1623-1632. Tai, H.-C., Wang, B.Y., Serrano-Pozo, A., Frosch, M.P., Spires-Jones, T.L., and Hyman, B.T. (2014). Frequent and symmetric deposition of misfolded tau oligomers within presynaptic and postsynaptic terminals in Alzheimer’s disease. Acta neuropathologica communications 2, 146.

156 Takamori, S., Holt, M., Stenius, K., Lemke, E.A., Grønborg, M., Riedel, D., Urlaub, H., Schenck, S., Brügger, B., and Ringler, P. (2006). Molecular anatomy of a trafficking organelle. Cell 127, 831-846. Tang, F.-L., Erion, J.R., Tian, Y., Liu, W., Yin, D.-M., Ye, J., Tang, B., Mei, L., and Xiong, W.- C. (2015). VPS35 in Dopamine Neurons Is Required for Endosome-to-Golgi Retrieval of Lamp2a, a Receptor of Chaperone-Mediated Autophagy That Is Critical for α-Synuclein Degradation and Prevention of Pathogenesis of Parkinson’s Disease. The Journal of Neuroscience 35, 10613-10628. Teasdale, R., Loci, D., Houghton, F., Karlsson, L., and Gleeson, P. (2001). A large family of endosome-localized proteins related to sorting nexin 1. Biochem J 358, 7. Temkin, P., Lauffer, B., Jäger, S., Cimermancic, P., Krogan, N.J., and Von Zastrow, M. (2011). SNX27 mediates retromer tubule entry and endosome-to-plasma membrane trafficking of signalling receptors. Nat Cell Biol 13, 715-721. Tian, Y., Tang, F.-L., Sun, X., Wen, L., Mei, L., Tang, B.-S., and Xiong, W.-C. (2015). VPS35- deficiency results in an impaired AMPA receptor trafficking and decreased dendritic spine maturation. Molecular brain 8, 1. Toh, W.H., Chia, P.Z.C., Hossain, M.I., and Gleeson, P.A. (2018). GGA1 regulates signal-dependent sorting of BACE1 to recycling endosomes, which moderates Aβ production. Mol Biol Cell 29, 191- 208. Toonen, R.F., Wierda, K., de Wit, H., Cornelisse, L.N., Brussaard, A., Plomp, J.J., and Verhage, M. (2006). Munc18-1 expression levels control synapse recovery by regulating readily releasable pool size. Proceedings of the National Academy of Sciences 103, 18332-18337. Traer, C.J., Rutherford, A.C., Palmer, K.J., Wassmer, T., Oakley, J., Attar, N., Carlton, J.G., Kremerskothen, J., Stephens, D.J., and Cullen, P.J. (2007). SNX4 coordinates endosomal sorting of TfnR with dynein-mediated transport into the endocytic recycling compartment. Nat Cell Biol 9, 1370-1380. Tsika, E., Glauser, L., Moser, R., Fiser, A., Daniel, G., Sheerin, U.-M., Lees, A., Troncoso, J.C., Lewis, P.A., and Bandopadhyay, R. (2014). Parkinson’s disease-linked mutations in VPS35 induce dopaminergic neurodegeneration. Hum Mol Genet 23, 4621-4638. Uytterhoeven, V., Kuenen, S., Kasprowicz, J., Miskiewicz, K., and Verstreken, P. (2011). Loss of skywalker reveals synaptic endosomes as sorting stations for synaptic vesicle proteins. Cell 145, 117-132. Uytterhoeven, V., Lauwers, E., Maes, I., Miskiewicz, K., Melo, M.N., Swerts, J., Kuenen, S., Wittocx, R., Corthout, N., and Marrink, S.-J. (2015). Hsc70-4 deforms membranes to promote synaptic protein turnover by endosomal microautophagy. Neuron 88, 735-748. van Weering, J.R., and Cullen, P.J. (2011). Signaling: Retromer arrests receptor on the run. Nat Chem Biol 7, 251-252. van Weering, J.R., and Cullen, P.J. (2014). Membrane-associated cargo recycling by tubule- based endosomal sorting. Paper presented at: Semin Cell Dev Biol (Elsevier). van Weering, J.R., Sessions, R.B., Traer, C.J., Kloer, D.P., Bhatia, V.K., Stamou, D., Carlsson, S.R., Hurley, J.H., and Cullen, P.J. (2012a). Molecular basis for SNX‐BAR‐mediated assembly of distinct endosomal sorting tubules. The EMBO journal 31, 4466-4480. van Weering, J.R., Verkade, P., and Cullen, P.J. (2010). SNX–BAR proteins in phosphoinositide- mediated, tubular-based endosomal sorting. Paper presented at: Semin Cell Dev Biol (Elsevier). van Weering, J.R., Verkade, P., and Cullen, P.J. (2012b). SNX–BAR‐Mediated Endosome Tubulation is Co‐ordinated with Endosome Maturation. Traffic 13, 94-107. Varandas, K.C., Irannejad, R., and von Zastrow, M. (2016). Retromer endosome exit domains serve multiple trafficking destinations and regulate local G protein activation by GPCRs. Curr Biol 26, 3129-3142. Vardarajan, B.N., Bruesegem, S.Y., Harbour, M.E., George-Hyslop, P.S., Seaman, M.N., and Farrer, L.A. (2012). Identification of Alzheimer disease-associated variants in genes that regulate retromer function. Neurobiol Aging 33, 2231. e2215-2231. e2230.

157 Vaz‐Silva, J., Gomes, P., Jin, Q., Zhu, M., Zhuravleva, V., Quintremil, S., Meira, T., Silva, J., Dioli, C., and Soares‐Cunha, C. (2018). Endolysosomal degradation of Tau and its role in glucocorticoid‐ driven hippocampal malfunction. The EMBO journal 37, e99084. Verheyen, A., Diels, A., Dijkmans, J., Oyelami, T., Meneghello, G., Mertens, L., Versweyveld, S., Borgers, M., Buist, A., and Peeters, P. (2015). Using human iPSC-derived neurons to model TAU aggregation. PLoS One 10, e0146127. von Bergen, M., Barghorn, S., Li, L., Marx, A., Biernat, J., Mandelkow, E.-M., and Mandelkow, E. (2001). Mutations of tau protein in frontotemporal dementia promote aggregation of paired helical filaments by enhancing local β-structure. J Biol Chem 276, 48165-48174. Von Engelhardt, J., Mack, V., Sprengel, R., Kavenstock, N., Li, K.W., Stern-Bach, Y., Smit, A.B., Seeburg, P.H., and Monyer, H. (2010). CKAMP44: a brain-specific protein attenuating short-term synaptic plasticity in the dentate gyrus. Science 327, 1518-1522. Wallroth, A., and Haucke, V. (2018). Phosphoinositide conversion in endocytosis and the endolysosomal system. J Biol Chem 293, 1526-1535. Wang, C.-L., Tang, F.-L., Peng, Y., Shen, C.-Y., Mei, L., and Xiong, W.-C. (2012). VPS35 regulates developing mouse hippocampal neuronal morphogenesis by promoting retrograde trafficking of BACE1. Biology open, BIO20122451. Watanabe, S., Liu, Q., Davis, M.W., Hollopeter, G., Thomas, N., Jorgensen, N.B., and Jorgensen, E.M. (2013a). Ultrafast endocytosis at Caenorhabditis elegans neuromuscular junctions. Elife 2, e00723. Watanabe, S., Mamer, L.E., Raychaudhuri, S., Luvsanjav, D., Eisen, J., Trimbuch, T., Söhl- Kielczynski, B., Fenske, P., Milosevic, I., and Rosenmund, C. (2018). Synaptojanin and Endophilin Mediate Neck Formation during Ultrafast Endocytosis. Neuron 98, 1184-1197. e1186. Watanabe, S., Rost, B.R., Camacho-Pérez, M., Davis, M.W., Söhl-Kielczynski, B., Rosenmund, C., and Jorgensen, E.M. (2013b). Ultrafast endocytosis at mouse hippocampal synapses. Nature 504, 242. Watanabe, S., Trimbuch, T., Camacho-Pérez, M., Rost, B.R., Brokowski, B., Söhl-Kielczynski, B., Felies, A., Davis, M.W., Rosenmund, C., and Jorgensen, E.M. (2014). Clathrin regenerates synaptic vesicles from endosomes. Nature 515, 228. Wen, L., Tang, F.-L., Hong, Y., Luo, S.-W., Wang, C.-L., He, W., Shen, C., Jung, J.-U., Xiong, F., and Lee, D.-h. (2011). VPS35 haploinsufficiency increases Alzheimer’s disease neuropathology. The Journal of cell biology 195, 765-779. Westphal, V., Rizzoli, S.O., Lauterbach, M.A., Kamin, D., Jahn, R., and Hell, S.W. (2008). Video- rate far-field optical nanoscopy dissects synaptic vesicle movement. Science 320, 246-249. Wierda, K.D., Toonen, R.F., de Wit, H., Brussaard, A.B., and Verhage, M. (2007). Interdependence of PKC-dependent and PKC-independent pathways for presynaptic plasticity. Neuron 54, 275-290. Wilhelm, B.G., Mandad, S., Truckenbrodt, S., Kröhnert, K., Schäfer, C., Rammner, B., Koo, S.J., Claßen, G.A., Krauss, M., and Haucke, V. (2014). Composition of isolated synaptic boutons reveals the amounts of vesicle trafficking proteins. Science 344, 1023-1028. Willig, K.I., Rizzoli, S.O., Westphal, V., Jahn, R., and Hell, S.W. (2006). STED microscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis. Nature 440, 935. Wilson, J.M., de Hoop, M., Zorzi, N., Toh, B.-H., Dotti, C.G., and Parton, R.G. (2000). EEA1, a tethering protein of the early sorting endosome, shows a polarized distribution in hippocampal neurons, epithelial cells, and fibroblasts. Mol Biol Cell 11, 2657-2671. Wu, J.W., Herman, M., Liu, L., Simoes, S., Acker, C.M., Figueroa, H., Steinberg, J.I., Margittai, M., Kayed, R., and Zurzolo, C. (2013). Small misfolded Tau species are internalized via bulk endocytosis and anterogradely and retrogradely transported in neurons. J Biol Chem 288, 1856-1870. Wu, J.W., Hussaini, S.A., Bastille, I.M., Rodriguez, G.A., Mrejeru, A., Rilett, K., Sanders, D.W., Cook, C., Fu, H., and Boonen, R.A. (2016). Neuronal activity enhances tau propagation and tau pathology in vivo. Nat Neurosci 19, 1085-1092.

158 Wu, S., Fagan, R.R., Uttamapinant, C., Lifshitz, L.M., Fogarty, K.E., Ting, A.Y., and Melikian, H.E. (2017). The Dopamine Transporter Recycles via a Retromer-Dependent Postendocytic Mechanism: Tracking Studies Using a Novel Fluorophore-Coupling Approach. J Neurosci 37, 9438-9452. Wucherpfennig, T., Wilsch-Bräuninger, M., and González-Gaitán, M. (2003). Role of Drosophila Rab5 during endosomal trafficking at the synapse and evoked neurotransmitter release. The Journal of cell biology 161, 609-624. Xia, P., Wang, S., Du, Y., Zhao, Z., Shi, L., Sun, L., Huang, G., Ye, B., Li, C., and Dai, Z. (2013). WASH inhibits autophagy through suppression of Beclin 1 ubiquitination. The EMBO journal 32, 2685-2696. Yap, C.C., and Winckler, B. (2012). Harnessing the power of the endosome to regulate neural development. Neuron 74, 440-451. Young, J.E., Fong, L.K., Frankowski, H., Petsko, G.A., Small, S.A., and Goldstein, L.S. (2018). Stabilizing the Retromer Complex in a Human Stem Cell Model of Alzheimer’s Disease Reduces TAU Phosphorylation Independently of Amyloid Precursor Protein. Stem cell reports 10, 1046-1058. Yu, Y., Jans, D.C., Winblad, B., Tjernberg, L.O., and Schedin-Weiss, S. (2018). Neuronal Aβ42 is enriched in small vesicles at the presynaptic side of synapses. Life Science Alliance 1, e201800028. Zavodszky, E., Seaman, M.N., Moreau, K., Jimenez-Sanchez, M., Breusegem, S.Y., Harbour, M.E., and Rubinsztein, D.C. (2014). Mutation in VPS35 associated with Parkinson’s disease impairs WASH complex association and inhibits autophagy. Nature communications 5. Zhang, L., Schessl, J., Werner, M., Bonnemann, C., Xiong, G., Mojsilovic-Petrovic, J., Zhou, W., Cohen, A., Seeburg, P., and Misawa, H. (2008). Role of GluR1 in activity-dependent motor system development. J Neurosci 28, 9953-9968. Zheng‐Fischhöfer, Q., Biernat, J., Mandelkow, E.M., Illenberger, S., Godemann, R., and Mandelkow, E. (1998). Sequential phosphorylation of Tau by glycogen synthase kinase‐3β and protein kinase A at Thr212 and Ser214 generates the Alzheimer‐specific epitope of antibody AT100 and requires a paired‐helical‐filament‐like conformation. Eur J Biochem 252, 542-552. Zhou, B., Wu, Y., and Lin, X. (2011). Retromer regulates apical–basal polarity through recycling Crumbs. Dev Biol 360, 87-95.

159 160