# THE PHYSIOLOGICAL FUNCTIONS OF THE AMYLOID PRECURSOR PROTEIN GENE FAMILY

EDITED BY: Ulrike C. Müller and Thomas Deller PUBLISHED IN: Frontiers in Molecular Neuroscience

#### *Frontiers Copyright Statement*

*© Copyright 2007-2018 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-355-9 DOI 10.3389/978-2-88945-355-9

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **THE PHYSIOLOGICAL FUNCTIONS OF THE AMYLOID PRECURSOR PROTEIN GENE FAMILY**

Topic Editors: **Ulrike C. Müller,** Universität Heidelberg, Germany **Thomas Deller**, Universität Frankfurt, Germany

3D-reconstruction of an EGFP-labeled hippocampal CA1 pyramidal cell.

We thank Prof. Dr. Mario Vuksic, Frankfurt/Zagreb for the original image and Annika Mehr for the artwork.

The amyloid precursor protein APP plays a key role in the pathogenesis of Alzheimer's disease (AD), as proteolytical cleavage of APP gives rise to the Ab peptide which is deposited in the brains of Alzheimer patients. Despite this, our knowledge of the normal cell biological and physiological functions of APP and the closely related APLPs is limited. This may have hampered our understanding of AD, since evidence has accumulated that not only the production of the Ab peptide but also the loss of APP-mediated functions may contribute to AD pathogenesis. Thus, it appears timely and highly relevant to elucidate the functions of the APP gene family from the molecular level to their role in the intact organism, i.e. in the context of nervous system development, synapse formation and adult synapse function, as well as neural homeostasis and aging.

Why is our understanding of the APP functions so limited? APP and the APLPs are multifunctional proteins that undergo complex proteolytical processing. They give rise to an almost bewildering array of different fragments that may each subserve specific functions. While Ab is aggregation prone and neurotoxic, the large secreted ectodomain APPsα - produced in the non-amyloidogenic α-secretase pathway - has been shown to be neurotrophic, neuroprotective and relevant for synaptic plasticity, learning and memory. Recently, novel APP cleavage pathways and enzymes have been discovered that have gained much attention not only with respect to AD but also regarding their role in normal brain physiology. In addition to the various cleavage products, there is also solid evidence that APP family proteins mediate important functions as transmembrane cell surface molecules, most notably in synaptic adhesion and cell surface signaling. Elucidating in more detail the molecular mechanisms underlying these divers functions thus calls for an interdisciplinary approach ranging from the structural level to the analysis in model organisms. Thus, in this research topic of Frontiers we compile reviews and original studies, covering our current knowledge of the physiological functions of this intriguing and medically important protein family.

**Citation:** Müller, C. U., Deller, T., eds. (2018). The Physiological Functions of the Amyloid Precursor Protein Gene Family. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-355-9

# Table of Contents

#### **Editorial:**

*06 Editorial: The Physiological Functions of the APP Gene Family* Ulrike C. Müller and Thomas Deller

#### **Section 1: Secretases – substrates, regulation and functions**


Kristina Endres and Thomas Deller

*65 The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein*

Lisa Seipold and Paul Saftig

*72 Corrigendum: The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein* Lisa Seipold and Paul Saftig

#### **Section 2: APP/APLP structure and transmembrane signaling**


#### *134 LRP1 Modulates APP Intraneuronal Transport and Processing in Its Monomeric and Dimeric State*

Uta-Mareike Herr, Paul Strecker, Steffen E. Storck, Carolin Thomas, Verena Rabiej, Anne Junker, Sandra Schilling, Nadine Schmidt, C. Marie Dowds, Simone Eggert, Claus U. Pietrzik and Stefan Kins

*151 APP Function and Lipids: A Bidirectional Link*

Marcus O. W. Grimm, Janine Mett, Heike S. Grimm and Tobias Hartmann

#### **Section 3: Functions during development, at the synapse and for neuroprotection**


Suzanne Guénette, Paul Strecker and Stefan Kins

*219 Region-Specific Differences in Amyloid Precursor Protein Expression in the Mouse Hippocampus*

Domenico Del Turco, Mandy H. Paul, Jessica Schlaudraff, Meike Hick, Kristina Endres, Ulrike C. Müller and Thomas Deller


# Editorial: The Physiological Functions of the APP Gene Family

Ulrike C. Müller <sup>1</sup> \* and Thomas Deller <sup>2</sup>

<sup>1</sup> Department of Functional Genomics, Institute for Pharmacy and Molecular Biotechnology, Universität Heidelberg, Heidelberg, Germany, <sup>2</sup> Institute of Clinical Neuroanatomy, Neuroscience Center, Goethe University Frankfurt, Frankfurt, Germany

Keywords: amyloid precursor protein (APP), APP like proteins, synaptic plasticity, development, spine density, neuroprotection and neuronal repair, synaptogenesis, secretases

**Editorial on the Research Topic**

#### **The Physiological Functions of the APP Gene Family**

The amyloid precursor protein APP plays a key role in the pathogenesis of Alzheimer's disease (AD), as proteolytical cleavage of APP gives rise to the β-amyloid peptide Aβ, which is deposited in the brains of AD patients (Selkoe and Hardy, 2016). In contrast to this key role in AD, the reviews and original papers in this Special Issue entitled "The physiological functions of the APP gene family" aim to shed some light on the "bright side" of APP, which exhibits important physiological functions during brain development, for adult brain plasticity and protection against injury. This change of perspective is timely, since accumulating evidence suggests that disease symptoms are caused both by an overload of toxic substances, e.g., "too much Aβ," as well as deficits of protective molecules, e.g., "not enough APPsα."

Unraveling APP functions has not been trivial, since APP undergoes complex processing. APP processing is initiated either by α-secretase cleavage within the Aβ region, or by β-secretase (BACE) cleavage at the N-terminus of Aβ, leading to the secretion of large soluble ectodomains, termed APPsα and APPsβ, respectively. Subsequent processing of the C-terminal fragments (CTFα or CTFβ) by γ-secretase results in the production of Aβ, p3 and the APP intracellular domain (AICD). This processing—as well as processing along non-canonical pathways (see Müller et al., 2017, for review) results in numerous fragments, which have different and partially opposite functional properties. Furthermore, APP functions are in part shared by APP-like proteins 1 and 2 (APLP1 and 2), which confounds some experimental approaches. Finally, expression changes over time and with aging add additional levels of complexity. In short, understanding APP gene family functions is challenging and this special issue provides a broad overview of the state-of-the art in this field.

Several reviews (Seipold and Saftig; Endres and Deller; Yan; Becker-Pauly and Pietrzik) focus on the properties of canonical and non-canonical α-, and β-secretases, their substrates, regulation, and neurobiological functions in health and disease. Müller et al. give a systematic overview over proteomic methods to systematically identify the substrates of membrane proteases. The knowledge of these substrates is crucial to identify the physiological and pathological functions of secretases and to assess potential risks of their pharmacological impairment to treat AD (Endres and Deller; Yan). In addition, there is evidence that the secretases which are transmembrane proteases can form larger complexes with other cell surface proteins that may modulate their activity including members of the tetraspannin family (Seipold and Saftig). APP processing is further modulated by the lipid composition of the plasma membrane and accumulating evidence suggests that Aβ and the AICD play an important role in regulating lipid homeostasis (Grimm et al.). Likewise, lipoprotein receptors may interact with APP to control developmental processes and synaptic function (Pohlkamp et al.). They have been shown to not only regulate Aβ uptake and

Edited and reviewed by: Nicola Maggio, Sackler Faculty of Medicine, Tel Aviv

> University, Israel \*Correspondence:

Ulrike C. Müller u.mueller@urz.uni-heidelberg.de

Received: 26 September 2017 Accepted: 02 October 2017 Published: 23 October 2017

#### Citation:

Müller UC and Deller T (2017) Editorial: The Physiological Functions of the APP Gene Family. Front. Mol. Neurosci. 10:334. doi: 10.3389/fnmol.2017.00334 degradation, but also APP processing and APP trafficking. In this regard, employing live cell imaging in primary neurons Herr et al. demonstrate that low-density lipoprotein receptorrelated protein 1 (LRP1) modulates the axonal transport of APP monomers and dimers.

There is a large body of evidence indicating that APP family proteins are multimodal proteins that can function as ligands via their secreted fragments, in particular APPsα, or as cell surface proteins mediating signal transduction and synaptic adhesion (as reviewed by Müller et al., 2017). Wild et al. discuss how metal (Cu and Zn) binding affects the structure of the APP extracellular domain and may modulate its role as a synaptic adhesion molecule. As APP family proteins have no enzymatic activities, signal transduction relies on interactions with other membrane proteins and/or adaptors. The role of the Fe65 adaptor family is summarized by Guenette et al. Fe65 binding to the APP C-terminus involves its phosphotyrosine-binding (PTB) domain 2 which can also mediate the formation of cytosolic Fe65 dimers, as shown by X-ray crystallography (Feilen et al.). The importance of heteromeric G-protein interactions with the APP C-terminus for physiological APP signaling and AD pathogenesis is reviewed by Copenhaver and Kogel.

Major insight into the in vivo functions of APP family proteins has been obtained from animal models. Drosophila expresses only one APP protein called APP-like (APPL) and two reviews (Cassar and Kretzschmar; Preat and Goguel) deal with APPL functions in flies. In mice the analysis of APP functions is complicated by partially overlapping functions within the gene family and lethality of double and triple knockout mice (Han et al.). To circumvent early postnatal lethality mice with conditional floxed alleles have been generated (Müller et al., 2017). Together, the analysis of engineered mouse models indicated that APP family

#### REFERENCES


proteins and their proteolytic fragments are important during nervous system development for neuronal migration, neurite outgrowth, axonal pathfinding, and synaptogenesis (Müller et al., 2017). Proteomic studies, reviewed by Weingarten et al. established APP family proteins as important components of the active zone. Lazarevic et al. demonstrated that low amounts of Aβ are involved in the regulation of neurotransmitter release. It should be noted, however, that APP family proteins have also been localized at postsynaptic sites including the neuromuscular junction. In addition, APP family proteins have important functions in the adult hippocampus, where they are differentially expressed in all subregions (Del Turco et al.) and regulate synaptic plasticity and memory (Ludewig and Korte). The recently identified function of APP for structural spine plasticity is summarized by Montagna et al. In particular APPsα holds great therapeutic potential for AD as reviewed by Mockett et al. Finally, Hefter and Draguhn highlight the role of APP and APPsα as a protective factor for acute neuronal insults including hypoxia.

We thank all contributors for their interesting and informative articles and the reviewers for their constructive and thoughtful suggestions.

#### AUTHOR CONTRIBUTIONS

UM and TD wrote the manuscript and both authors approved the final version for publication.

#### FUNDING

The authors thank the Deutsche Forschungsgemeinschaft for their support within programme FOR1332.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Müller and Deller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Proteomic Substrate Identification for Membrane Proteases in the Brain

Stephan A. Müller 1,2† , Simone D. Scilabra1,2† and Stefan F. Lichtenthaler 1,2,3,4 \*

<sup>1</sup> German Center for Neurodegenerative Diseases (DZNE), Munich, Germany, <sup>2</sup> Neuroproteomics, Klinikum rechts der Isar, Technische Universität München, Munich, Germany, <sup>3</sup> Institute for Advanced Study, Technische Universität Munich, Garching, Germany , <sup>4</sup> Munich Cluster for Systems Neurology (SyNergy), Munich, Germany

Cell-cell communication in the brain is controlled by multiple mechanisms, including proteolysis. Membrane-bound proteases generate signaling molecules from membranebound precursor proteins and control the length and function of cell surface membrane proteins. These proteases belong to different families, including members of the "a disintegrin and metalloprotease" (ADAM), the beta-site amyloid precursor protein cleaving enzymes (BACE), membrane-type matrix metalloproteases (MT-MMP) and rhomboids. Some of these proteases, in particular ADAM10 and BACE1 have been shown to be essential not only for the correct development of the mammalian brain, but also for myelination and maintaining neuronal connections in the adult nervous system. Additionally, these proteases are considered as drug targets for brain diseases, including Alzheimer's disease (AD), schizophrenia and cancer. Despite their biomedical relevance, the molecular functions of these proteases in the brain have not been explored in much detail, as little was known about their substrates. This has changed with the recent development of novel proteomic methods which allow to identify substrates of membrane-bound proteases from cultured cells, primary neurons and other primary brain cells and even in vivo from minute amounts of mouse cerebrospinal fluid (CSF). This review summarizes the recent advances and highlights the strengths of the individual proteomic methods. Finally, using the example of the Alzheimer-related proteases BACE1, ADAM10 and γ-secretase, as well as ADAM17 and signal peptide peptidase like 3 (SPPL3), we illustrate how substrate identification with novel methods is instrumental in elucidating broad physiological functions of these proteases in the brain and other organs.

# Keywords: proteomics, degradomics, protease, BACE, ADAM10, ADAM17, Alzheimer's disease

# PROTEOLYTIC PROCESSING IN ALZHEIMER'S DISEASE

Proteolysis is a biological process playing an essential role in all organisms and tissues, including the brain. For example, proteolysis regulates numerous cell functions, spanning from degradation of faulty proteins to post-translational generation of active signaling molecules, neurite outgrowth and modeling of the extracellular matrix. Therefore, protease activity must be tightly regulated and, conversely, aberrant proteolysis is associated with several pathological conditions ranging from inflammation to cancer and neurodegeneration. A prime example is Alzheimer's disease (AD), where deregulation of proteolysis leads to neurodegeneration. AD is the most common type of dementia, a syndrome characterized by loss of memory and cognitive decline. AD causes

#### Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Robert Vassar, Northwestern University, USA Stefan Kins, Kaiserslautern University of Technology, Germany Ulrich Auf Dem Keller, ETH Zurich, Switzerland

\*Correspondence: Stefan F. Lichtenthaler stefan.lichtenthaler@dzne.de

†These authors have contributed equally to this work.

Received: 10 August 2016 Accepted: 21 September 2016 Published: 13 October 2016

#### Citation:

Müller SA, Scilabra SD and Lichtenthaler SF (2016) Proteomic Substrate Identification for Membrane Proteases in the Brain. Front. Mol. Neurosci. 9:96. doi: 10.3389/fnmol.2016.00096 a substantial loss of neurons and synapses in the brain, leading to an overall loss in brain weight. Additional neuropathological hallmarks of the disease are the amyloid-β (Aβ) plaques, consisting of the mostly 42 amino acid long Aβ peptide (Aβ42), and the intraneuronal accumulation of neurofibrillary tangles, consisting of hyperphosphorylated forms of the microtubule-associated protein tau (Huang and Mucke, 2012). According to the widely accepted amyloid cascade hypothesis (Selkoe and Hardy, 2016), Aβ forms neurotoxic oligomers, which initiate an inflammatory response involving the activation of microglia and astrocytes. Subsequently tau becomes aberrantly phosphorylated and aggregates in neurofibrillary tangles, leading to synaptic loss, neuronal death, and ultimately dementia (Selkoe and Hardy, 2016).

Aβ derives from the transmembrane protein amyloid precursor protein (APP; Dislich and Lichtenthaler, 2012; **Figure 1A**) through sequential cleavage by two proteases, the β- and γ-secretase (Haass and Selkoe, 2007). The β-secretase was identified in 1999 by five independent research groups, and is referred to as β-site APP cleaving enzyme 1 (BACE1; Hussain et al., 1999; Sinha et al., 1999; Vassar et al., 1999; Yan et al., 1999; Lin et al., 2000). BACE1 cleavage releases a soluble extracellular fragment of APP (sAPPβ) and generates a carboxy (C)-terminal membrane-tethered fragment known as C99 (**Figure 1**). C99 undergoes a subsequent intramembrane cleavage by γ-secretase, a multi-subunit protease complex comprising four transmembrane proteins: presenilin, nicastrin, Pen2 and Aph1 (De Strooper et al., 2010). The γ-secretase cleavage of C99 generates Aβ and releases intracellularly the APP intracellular domain (AICD). APP can undergo an alternative cleavage, mediated by a disintegrin and metalloproteinase 10 (ADAM10; Lammich et al., 1999; Kuhn et al., 2010), also known as α-secretase, that releases its soluble ectodomain (sAPPα) and generates a membrane-tethered fragment, C83 (**Figure 1B**). Importantly, the subsequent cleavage by γ-secretase releases a truncated form of Aβ, which is non-toxic. Three other proteases emerged to be involved in the processing of APP. Asparagine endopeptidase (AEP), known as the δsecretase, is a cysteine proteinase that mediates APP processing in an age-dependent manner and is linked to AD pathogenesis (Zhang et al., 2015). Furthermore, the membrane-tethered metalloproteinase (MT5-MMP) cleaves APP at amino acids 504–505, initiating a proteolytic processing that leads to the generation of APP fragments (Aη-α), which lower neuronal activity (Ahmad et al., 2006; Willem et al., 2015; **Figure 1C**). Loss of MT5-MMP ameliorates pathology and behavioral deficits in a mouse model of AD (Baranger et al., 2016). A member of the meprin family of metalloproteases, meprin β, was also shown to cleave APP, with the cleavage site being identical to that of the β-secretase or in close proximity to it. This shedding event is followed by the γ-secretase cleavage and leads to the generation of Aβ or truncated variants of Aβ (i.e., Aβ 2–40; Bien et al., 2012). Additionally, meprin β can process APP at the N-terminus, releasing two N-terminal fragments of APP of 11 and 22 kDa, namely APP11 and APP22 (Jefferson et al., 2011).

#### REGULATED INTRAMEMBRANE PROTEOLYSIS

The proteolytic processing of APP is a prime example for a proteolytic process referred to as regulated intramembrane proteolysis (RIP; **Figure 2**). RIP frequently comprises two proteolytic cleavages, namely shedding and intramembrane proteolysis. Shedding is mediated by membrane-tethered proteases, referred to as ''sheddases'', which cleave their transmembrane substrates, thereby releasing their soluble ectodomains into the extracellular milieu (**Figure 2**). Most sheddases cleave their substrates at peptide bonds outside of the membrane, but at a short distance from the lumenal or extracellular membrane surface. Shedding can be followed by a second cleavage within the substrates' transmembrane domain. This cleavage results in release of the intracellular domain (ICD) into the cytosol and the extracellular secretion of the small remaining peptide. As it occurs for APP, α- and β-secretase function as sheddases, and their activity can be coupled with the action of γ-secretase to perform RIPping of the remaining membrane-tethered protein fragment.

Shedding and intramembrane proteolysis initiate a sequence of extracellular and intracellular events that control a broad range of physiological processes in the brain, including cell-cell communication, cell differentiation and development (Murphy et al., 2008; Lichtenthaler et al., 2011; Weber and Saftig, 2012). For instance, the tumor necrosis factor-α (TNF), a proinflammatory cytokine, is generated as a transmembrane protein that needs to be shed by ADAM17 from the cell surface in order to trigger immune responses (Black et al., 1997). Interestingly, the remaining membrane-bound fragment can be further cleaved by SPPL2a or SPPL2b within the membrane, releasing the TNF ICD which acts as an additional signaling molecule (Friedmann et al., 2006). Similarly to TNF, several growth factors, including EGF-like growth factors and neuregulins, are inactive when bound to the membrane and get activated by proteolytic shedding (Blobel, 2005).

Sheddases do not only modulate the availability of ligands, but also regulate the activity of signaling receptors. Notch is a clear example of cell surface receptor that requires RIPping to initiate its signaling pathway and control cell-differentiation (Hartmann et al., 2002). For other substrates, RIP is a mechanism to terminate a protein's function. For example, shedding shuts down the signaling function of TNF receptors (D'Alessio et al., 2012; Deng et al., 2015) or the adhesive functions of cell adhesion proteins (Solanas et al., 2011).

#### SHEDDASES AND INTRAMEMBRANE PROTEASES

Members of several different families of proteases have been shown to function as sheddases, including several ADAMs, BACE proteases, membrane-type metalloproteinases (MT-MMPs) and rhomboids (**Figure 2**; Blobel, 2005; Vassar et al., 2014; Itoh, 2015). In addition, signal peptide peptidase like 3

a disintegrin and metalloproteinase 10 (ADAM10; yellow arrow), beta-site APP cleaving enzyme 1 (BACE1; red arrow), γ-secretase (orange arrows), asparagine endopeptidase (AEP; blue arrows), membrane-type matrix metalloproteases (MT5-MMP; fuchsia arrow) and meprin β (green arrows). (B) APP can undergo amyloidogenic processing when cleaved by BACE1. Cleavage of APP by BACE1 results in generation of sAPPβ. Subsequent cleavage of the remaining transmembrane domain by γ-secretase releases amyloid-β (Aβ). (C) Conversely, cleavage of APP by ADAM10 favors the non-amyloidogenic pathway, releasing sAPPα. Subsequent γ-secretase cleavage releases a non-toxic truncated form of the Aβ peptide, called p3. (D) In addition, APP can be cleaved by MT5-MMP, which results in the release of sAPPη. Consecutively, C-terminal fragment (CTF)-η can be cleaved by ADAM10 or BACE1 that release Aη-α and Aη-β, respectively. The recently identified δ-secretase cleaves APP a few amino acids N-terminally to the BACE1 cleavage site (not shown in the figure).

(SPPL3) from the SPP family and site-1 protease (S1P) can also act as sheddases (Lenz et al., 2001; Voss et al., 2012). ADAM and BACE proteases cleave substrates in their extracellular domain, at a short distance from the membrane, and need the sequential cleavage of an intramembrane proteinase in order to perform RIPping. Conversely, rhomboids and SPPL3 are intramembrane proteases that cleave their substrates within or close to the transmembrane domain. As a consequence of such cleavage, regardless whether it occurs extracellularly or within the transmembrane domain, the ectodomain of substrates is released into the extracellular milieu. This is of note, as the secreted form of transmembrane proteins can acquire functions different from that of the membranebound counterpart. MT-MMPs can act as sheddases. However, compared to the related family of ADAMs, MT-MMPs can cleave their substrates more distantly from the cell surface and on different sites, thereby releasing truncated forms of protein ectodomains or lower molecular weight fragments (Selvais et al., 2011; Fu et al., 2013; Willem et al., 2015).

# FUNCTION OF PROTEASES IS DETERMINED BY SUBSTRATES

Proteases have been well characterized in pathophysiology of disease as key players in the development of several pathological conditions, including neurodegenerative diseases. Thus, protease inhibition has been widely targeted for drug development. Unfortunately, in the vast majority of cases, therapies based on protease inhibition have failed in clinical trials. Indeed, there are critical limitations to the development of therapies targeting proteases. First, distinct members of a protease family share structural features, thus drug-based inhibition of a specific protease can affect the activity of homologs. For example, BACE1 inhibitors have been developed to reduce Aβ production in the brain and are tested for treatment and prevention of AD. However, they also block the homologous protease BACE2, which has critical functions in pigmentation (Rochin et al., 2013; Neumann et al., 2015). In fact, mice treated with such inhibitors get a gray fur color and patients treated with these drugs need to go for regular dermatology testing (Yan, 2016). More importantly, proteases often do not target a specific substrate, but they can cleave an array of diverse proteins. As a consequence, their inhibition can deregulate a number of cellular processes, and inhibition-based therapies can lead to mechanism-based side effects that often are more pronounced than amelioration of the pathology itself. For instance, due to its central contribution to the pathogenesis of AD, γ-secretase has been extensively targeted for drug development. A number of γ-secretase inhibitors have been generated and tested for their ability to reduce Aβ production in vitro and in vivo. One of them, called Semagacestat, was terminated in clinical trial Phase III, as it was associated with worsening of patient cognition and with higher incidence of skin cancer (De Strooper, 2014). These mechanism-based side effects were linked to the chronic inhibition of Notch cleavage by γ-secretase.

A deep understanding of protease functions and their roles in cell biology is necessary for developing effective therapeutic strategies. As the biological function of proteases depends on their substrate spectrum, the identification of the substrate repertoire is essential to understand the function of a specific protease and to predict potential side-effects of their therapeutic inhibition. In the last years, a number of proteomics-based methods have been developed in order to identify the substrate repertoire of specific proteases. In this review, we summarize the most commonly used and other suitable methods and give examples of their applications with a focus on sheddases and intramembrane proteases, in particular on BACE1, ADAM10 and γ-secretase in AD.

#### METHODS FOR MASS SPECTROMETRY BASED SUBSTRATE IDENTIFICATION OF MEMBRANE PROTEASES IN THE BRAIN

Mass spectrometry (MS) based proteomics offers powerful methods to identify membrane protease substrate candidates in vitro and in vivo. Especially, non-targeted quantification of protein cleavage products in the secretome of brain-derived primary cells or cell lines, as well as cerebrospinal fluid (CSF) are suitable for protease substrate identification. In this context, the secretome comprises all proteins released by cells into body fluids or into the conditioned medium of cultured cells. For sheddases such as BACE1 and ADAM10, the ectodomain of their substrates is released into the extracellular space. Therefore, usually a loss of function condition, such as protease KO, knockdown (KD), or inhibition, is quantitatively compared with related control conditions to identify substrates. At loss of function conditions, substrate cleavage is fully or partly prevented which leads to a reduced abundance of the related cleavage products in the secretome.

Additionally, some substrates accumulate in the cell membrane when the target protease does not cleave them. Therefore, membrane protease substrate candidates might also be identified by quantitative proteomics due to an increased abundance within the cell membrane. Alternatively, also gain of function conditions such as overexpression of the target protease can be used which leads to increased cleavage activity and subsequently to increased abundance of substrate cleavage products in the secretome.

Here, we will provide a short overview of the main methods for MS-based protease substrate identification with a focus on methods for sheddase substrate identification. In the first section, methods are described that are used to identify substrates in the secretome or on the cell surface (**Figure 3**). In the second section, methods are described that also allow protease cleavage site determination (**Figure 4**). Usually, bottom-up proteomics is used for this purpose. Briefly, in all protocols, secreted or membrane proteins are digested with a protease, usually trypsin, to create proteolytic peptides. In most cases those peptides are separated by C18 reversed phase liquid chromatography (LC) prior to MS analysis. The MS raw data is searched against a protein database to identify proteotypic peptides. Relative peptide and protein quantification can be done by different methods. According to the different protocols for protease substrate identification, label-free and label-based quantification methods are used. A detailed explanation of different quantification methods can be found in several review

articles (Bantscheff et al., 2007; Schulze and Usadel, 2010; Bakalarski and Kirkpatrick, 2016).

#### Methods for Protease Substrate Identification

#### Glyco-capture

Most substrates of sheddases are single-pass transmembrane or GPI-anchored proteins, which are usually glycosylated within their ectodomain. According to UniProt reference database of Homo sapiens (date: 2016-06-30), 92% of all transmembrane type I (1125 out of 1228, term: SL-9905) and 83% of all transmembrane type II (347 out of 420, term: SL-9906) proteins are annotated as glycoproteins (term: KW-0325). Upon membrane protein cleavage, a part of the ectodomain is secreted (**Figure 2**). Glyco-capturing (**Figure 3**) facilitates specific enrichment of glycoproteins. In the first step, cis-diol groups of N- and/or O-linked carbohydrates are oxidized to aldehydes using periodate. At 1 mM periodate

mainly sialic acid residues are oxidized whereas all cis-diol groups can be oxidized with higher concentrations such as 20 mM. The aldehydes can be covalently coupled to a hydrazide resin (Zhang et al., 2003). Alternatively, aminooxybiotin can be used to biotinylate the oxidized sugars for subsequent pull-down with avidin or streptavidin beads (Zeng et al., 2009). Both, hydrazone and oxime ligations are catalyzed by aniline (Dirksen and Dawson, 2008). After glycoprotein pull-down, proteins are digested with trypsin for MS analysis.

When glycoproteins are covalently coupled to a hydrazide resin, the digestion is performed directly on the beads. In this case, the remaining glycosylated peptides can be released using peptide-N-glycosidase F (PNGaseF) and analyzed as separate fraction to further reduce the sample complexity (Zhang et al., 2003; Stützer et al., 2013). For example, this technique was used to identify substrates of BACE1 and 2 in pancreatic β-cells by quantification of N-glycopeptides of cell supernatants and lysates (Stützer et al., 2013). Glycoprotein labeling is even possible on the cell surface of living cells (Wollscheid et al., 2009; Zeng et al., 2009). Here, glycoproteins are labeled covalently with amino-oxybiotin or biocytin-hydrazide for subsequent enrichment of glycoproteins or glycopeptides using streptavidin beads.

A drawback of glyco-capturing is that secretome analyses usually have to be performed under serum-free conditions because many serum proteins, such as immunoglobulins, are also glycosylated. Thereby, peptides from secreted proteins might be masked by the presence of high abundant peptides from serum protein. However, glyco-capturing might also be used to identify protease substrates in vivo using plasma or CSF samples (**Table 1**).

#### Secretome Protein Enrichment with Click Sugars (SPECS)

SPECS was developed to overcome the difficulty of secretome analysis in the presence of serum or other protein containing culture media additives (Kuhn et al., 2012; **Figure 3**). For example, primary neurons have to be cultured using additives such as B27 which have a high protein concentration, in particular of albumin. Quantitative proteomics usually covers a dynamic concentration range of 3–4 orders of magnitude. Fetal bovine serum (FBS) has a total protein concentration of 30–45 mg/ml which includes 17–34 mg/ml of albumin


#### TABLE 1 | Advantages and disadvantages of different techniques for membrane protease substrate identification.

(FBS-BBT, Rocky Mountain Biologicals, Inc., Missoula, MT, USA). Thus, conditioned cell culture medium with 10% FBS contains 1.7–3.4 mg/mL whereas the concentration of secreted proteins is three orders of magnitude lower (in the µg/mL range). For example, a concentration of 7.0–7.5 µg/mL was reported for the J774 murine macrophage cell line cultured in 20 mL serum-free medium (Chevallet et al., 2007). Therefore, high concentrations of protein supplements lead to a dramatically decreased quantification of cell-derived secreted proteins.

For SPECS, azido sugars are used for metabolic labeling of glycoproteins in cell culture. ManNAz (N-azidoacetylmannosamine-tetraacylated) is taken up by cells, converted to N-azidoacetyl-sialic acid and mainly incorporated into N-linked glycosylation but also into O-glycosylation (Sletten and Bertozzi, 2011). GlcNAz (N-azidoacetylglucosaminetetraacylated) and GalNAz (N-azidoacetylgalactosaminetetraacylated) are primarily used to label O-glycans. After metabolic labeling for usually 24–48 h, shed glycoproteins in cell supernatant and/or surface glycoproteins are modified by copper-free alkyne-azide click chemistry with dibenzocyclooctyne (DBCO)-containing biotinylation reagents. The biotinylated glycoproteins can be efficiently pulled down with avidin- or streptavidin-coupled beads for further analysis. The protocol requires SDS-PAGE fractionation of purified glycoproteins and in-gel digestion because bovine albumin within the conditioned medium cannot be completely removed. However albumin abundance is reduced more than 50-fold and several hundred glycoproteins were identified in the cell supernatant (Kuhn et al., 2012). Alternatively, on-bead tryptic digestion of secreted glycoproteins was reported for low-serum conditions using alkyne beads for covalent coupling (Roper et al., 2013). Here, the secretome of stromal cell lines was directly analyzed under serum-free conditions and compared to glycoprotein enrichment after ManNAz labeling using serum-free and low serum (1%) conditions. Overall, only 75 and 100 proteins were identified using SPECS at serum-free and low serum conditions, respectively. Compared to the whole secretome digestion (193 proteins), a significant enrichment for glycoproteins was reported and 46 additional proteins with lower abundance could be identified.

SPECS has been used to identify substrates of BACE1 (Kuhn et al., 2012) and ADAM10 (Kuhn et al., 2016) in murine primary neuronal cell cultures and of SPPL3 in two different cell lines (Kuhn et al., 2015). SPECS is a technique that is well-suited for substrate identification of sheddases in any cell culture system. Quantification is performed on many peptides in contrast to enrichment of neo N- or C-termini which relies on quantification by one or two peptides per protein. Hence, SPECS offers increased reliability of protein quantification. Yet, as the method does not enrich specifically terminal peptides, cleavage sites cannot be automatically inferred from the MS data (**Table 1**). However, for some substrates semi-tryptic peptides were identified and allowed determination of the cleavage site.

Importantly, SPECS identifies secreted, cell-derived proteins, regardless of whether they are soluble, secreted proteins or proteolytically derive from membrane proteins. Thus, SPECS can also be used for the identification of secreted proteins as biomarkers. When the research goal is to identify sheddase substrates, the hit list is simply filtered for membrane proteins and thus yields the list of substrate candidates.

A variant of SPECS has also been used to label membrane proteins at the surface of ADAM10-deficient neurons. Compared to wild-type neurons, a number of membrane proteins were found to be enriched, suggesting that they may be ADAM10 substrates. In fact, several of them also showed reduced ectodomain release into the conditioned medium and were validated as ADAM10 substrates (Kuhn et al., 2016). While changes in the secretome are mostly used to identify shedding substrates, these results demonstrate that the enrichment of substrates in the membrane may be an alternative approach.

#### Azidohomoalanine Labeling

Azidohomoalanine (AHA) labeling (Dieterich et al., 2006) is an alternative labeling method of newly synthesized proteins and is similar to the SPECS method. AHA is an azidecontaining analog of methionine which is incorporated into proteins via the methionyl-tRNA more slowly than methionine (Kiick et al., 2002). Eichelbaum et al. (2012) established a proteomic method using AHA labeling for secretome analysis in the presence of serum supplements. AHA labeled proteins in the secretome are covalently bound to alkyne resin via Cu(I) catalyzed cycloaddition reaction between azide groups and a terminal alkyne. After stringent washing of the beads, an on-bead digestion is performed. Proteolytic peptides are fractionated followed by LC-MS/MS analysis. This method was used in combination with pulsed stable isotope labeling by amino acids in cell culture (SILAC) to monitor protein synthesis and secretion during macrophage activation (Eichelbaum and Krijgsveld, 2014). This method might also be used with copper-free alkyne-azide click chemistry with DBCO similar to SPECS. While SPECS enriches for N-glycosylated proteins, AHA labeling facilitates capturing of all secreted proteins. However, a drawback is that cellular toxicity has been observed for AHA labeling, which is not the case for azido-sugar labeling. The reason appears to be that AHA, which is not identical to methionine, but is incorporated into the amino acid backbone of proteins, may slightly alter the conformation of numerous cellular proteins leading to cellular toxicity. In contrast, the modified sugars in SPECS are located at the outside of the protein structure and are less likely to affect protein conformation. For the AHA method a careful titration of the AHA concentration can minimize the cellular toxicity for every cell type. The AHA method has not yet been used for protease substrate identification, but may be well suited for determining sheddase substrates. Compared to the other methods for protease substrate identification, AHA labeling and SPECS share similar advantages and disadvantages (**Table 1**).

#### Surface Biotinylation

An alternative approach for the enrichment of cell surface proteins is their biotinylation using N-hydroxysuccinimide (NHS) chemistry. Proteins are labeled at amino-groups of lysine residues and protein N-termini with NHS-biotin for subsequent pull-down. Even though cell surface biotinylation is frequently used for proteomics, there are no publications for membrane protease substrate identification available. However, this technique was already used to validate MS-based substrate identifications via immunoblotting (Stützer et al., 2013). Unfortunately, this approach is not suitable for secretome analysis because all proteins within the conditioned medium would be labeled, i.e., also serum proteins and not just the cellderived proteins (**Table 1**). Additionally, labeled lysine residues are no more accessible for tryptic cleavage which results in long peptides. To overcome this latter issue, peptides can be further digested with other proteases such as GluC to get more peptides with a suitable length for LC-MS analysis.

#### Cerebrospinal Fluid Proteomics

CSF is the only body fluid that is in direct contact with the brain. Therefore, CSF proteomics is the only method that facilitates in vivo secretome analysis of the brain. Ideally, KO mice are the system of choice to study membrane proteases, as the proteolytically released substrate ectodomains will be found in the CSF. While milliliters of CSF can be sampled from humans, only 5–15 µl can be collected from mice (Liu and Duff, 2008). This makes proteomic analysis of murine CSF challenging.

Furthermore, sampling of murine CSF is very susceptible to contaminations by cells or blood. The quantification of more than 50% of all CSF proteins is affected even by low levels of blood contamination (Aasebø et al., 2014). Therefore, sampling of murine CSF is the most critical part of the proteomic workflow (**Table 1**).

An immunodepletion kit from Agilent is available for three most abundant proteins in murine blood which might also work for murine CSF. However, depletion or even fractionation of CSF can lead to sample losses, especially for minute sample amounts. Many CSF proteins bind to albumin and get codepleted (Holewinski et al., 2013) or might bind unspecifically to the depletion beads or plastic. Additionally, multi-use of immunoaffinity depletion columns require efficient stripping of bound proteins to reduce sample carryover and to maintain the binding capacity (Gundry et al., 2009). Another study performed a proteomic analysis of murine CSF that was immunodepleted with IgY-14 resin which is designed to remove the 14 most abundant human serum proteins (Cunningham et al., 2013). Yet, 100 µl of pooled CSF was used, but overall only 289 proteins were identified. Consequently, direct in-solution digestion without depletion is the method of choice for murine CSF.

Recently, the workflow for murine CSF proteomics was optimized and allowed identification and label-free quantification of BACE1 substrates in mouse brains using individual wild-type and BACE1−/<sup>−</sup> mice (Dislich et al., 2015). This shows the suitability of this approach to identify protease substrates in vivo by proteomics (**Table 1**). Additionally, it was shown that quantitative CSF proteome analysis of individual mice is possible using only 5 µl CSF resulting in 522 relatively quantified proteins. This was a considerable improvement in comparison to 128 identified proteins from individual mice (Smith et al., 2014) and 103 relatively quantified proteins using pooled murine CSF (Cunningham et al., 2013) in previous studies.

# Methods for Protease Substrate Identification and Cleavage Site Determination

#### Terminal Amine-based Isotope Labeling of Substrates (TAILS)

Protease cleavage creates neo N- and C-termini. If the protease of interest is inhibited or knocked-out, the neo N- and C-terminal peptides are no longer generated. Thus, identification of neo N- and C-terminal peptides allows both substrate and cleavage site identification at the same time. Besides cleavage site determination, the methods for N-termini identification are also suitable for identification of N-termini of whole proteins (Vaca Jacome et al., 2015; Berry et al., 2016). Terminal amine-based isotope labeling of substrates (TAILS) is a method for specific enrichment of the terminal peptides. Two different protocols for enrichment of either protein/peptide N- or C-termini, called N- and C-TAILS are available (Schilling et al., 2010; Kleifeld et al., 2011). The first step of N-TAILS is labeling of α- and ε-amines with methyl groups (dimethyl labeling) or other amino group-reactive isobaric labeling reagents for proteomics, such as iTRAQ or TMT reagents (**Figure 4**). All free protein/peptide N-termini including the neo N-termini as well as lysine residues are modified. Up to three different conditions can be relatively quantified by using stable isotope dimethyl labeling (Boersema et al., 2009) while up to 10 samples can be relatively quantified with isobaric labeling.

After labeling of amines, samples from the different conditions, i.e., protease inhibition and vehicle control, are mixed. In the next step, the labeled proteins and peptides are digested by trypsin and/or other endoproteases. This leads to peptides derived from the former N-term, the C-term, as well as internal regions called ''internal peptides''. The peptides of the former N-termini have no free amino group whereas all other proteolytic peptides have a new, free amino group. Dendritic polyglycerol aldehyde polymers are used to remove all ''internal'' peptides with free amino groups under mild reductive conditions using cyano-borohydride. In the last step, the remaining N-terminal peptides are analyzed by MS.

C-TAILS is the counterpart of N-TAILS which facilitates the identification of neo C-termini. After dimethyl-labeling, similar to the N-TAILS protocol, carboxyl groups are protected with ethanolamine. A tryptic digestion is used to generate new free N- and C-termini of ''internal'' peptides. Again, labeling of α-amines is performed. The ''internal'' peptides are removed by coupling the free carboxyl groups to a polyallylamine polymer. The remaining C-terminal peptides are analyzed by MS.

TAILS is a powerful technique to identify the exact cleavage site of a protease. However, the methods require working at serum-free or low-serum conditions in cell culture for secretome analysis. For example, secreted proteins of different cell lines, which were cultured in serum-free medium, were incubated with recombinant meprin α and β for substrate identification by N-TAILS (Jefferson et al., 2011, 2013). On the other hand, in vivo analysis of proteins of cell lysates or membrane fractions is possible (Sabino et al., 2015; Prudova et al., 2016). Yet, quantification is based on only one peptide at the N- or C-terminus of the cleavage site and additionally the N-terminal peptide of the intact protein. To overcome this issue, routinely a whole secretome analysis using the labeled peptides after proteolytic digestion is performed which is used for relative protein quantification (Prudova et al., 2016). Moreover, for C-TAILS it has been difficult to achieve complete labeling of the carboxy-terminal groups, which is a disadvantage for the analysis of type I membrane protein shedding substrates, where the N-terminal ectodomain is released into the conditioned medium and would need to be detected with C-TAILS (**Table 1**).

#### Combined Fractional Diagonal Chromatography (COFRADIC)

Combined fractional diagonal chromatography (COFRADIC) is an umbrella term for different multistep chromatographic methods that include peptide derivatization, fractionation and isolation of modified peptides. With different types of modifications, it is possible to separate terminal peptides from neo N- and C-termini from other peptides (Van Damme et al., 2010). Usually, two conditions, with and without protease are differentially labeled in cell culture with SILAC using isotopic labeled arginine. Proteins are reduced, alkylated at cysteines and acetylated with NHS-(D3)acetate at α- as well as ε-amines. After tryptic digestion, N-pyroglutamate residues are enzymatically removed by pyroglutamyl aminopeptidases (Abraham and Podell, 1981) because peptides carrying those residues are usually not charged at pH 3.

Strong cation exchange (SCX) cartridges are used at pH 3 to remove the majority of peptides with a free N-terminus, while most peptides with acetylated N-termini have a net-charge of zero because the C-terminal arginine is positively charged but the carboxyl group at the C-term is mostly deprotonated (Staes et al., 2008; **Figure 4**). C-terminal peptides, which contain no C-terminal arginine are also not positively charged at pH 3 and elute with the flowthrough (Staes et al., 2008). Exceptions are histidine containing peptides, because histidine residues are positively charged at a pH of 3. Thus, those peptides are retained by SCX cartridges.

In the next step, hydrogen peroxide can be used to uniformly oxidize all methionines. Now, peptides are fractionated by C18 RP chromatography. The free α-amines of C-terminal and internal peptides in all fractions of both conditions are either differentially labeled with isotopic variants of NHS-butyrate ( <sup>12</sup>C4, <sup>13</sup>C4) for isolating both (neo) C- and N-terminal peptides. On the other hand, also isobaric tags might be used to label primary amines. Alternatively, free α-amines can be labeled with 2,4,6-trinitrobenzenesulfonic acid (TNBS) which introduces a very hydrophobic trinitrophenyl label, for isolating (neo) N-terminal peptides only (Staes et al., 2011).

In the former, the matching fractions of condition 1 and 2 are combined and a second C18 RP chromatography run is used for isolating terminal peptides for LC-MS based quantification. For TNBS labeling, trinitrophenyl containing internal or C-terminal peptides elute later than in the first C18 RP run which allows efficient separation of the (neo) N-terminal peptides derived by protease cleavage.

Relative quantification of N-terminal peptides is done by SILAC labeling with arginine whereas quantification of C-terminal peptides is based on the butyrate (or alternative) labeling. Different studies have been carried out to identify substrates and cleavage specificities of proteases. For example, substrate specificities of the granzyme tryptases A and K were identified (Plasman et al., 2014). In a more general approach, the secretome of gastric cancer-associated myofibroblasts was analyzed and identified activation of matrix metalloproteinases (Holmberg et al., 2013).

COFRADIC facilitates the identification and quantification of neo N- and C-termini which allows identification of protease substrates and their cleavage sites. However, extensive HPLC fractionation as well as LC-MS analysis of many fractions is very time-consuming. Additionally, histidine containing peptides are lost during the SCX chromatography step (**Table 1**).

#### Subtiligase Method

The subtiligase protocol enables enrichment of free protein N-termini as well as protease cleavage derived neo N-termini (**Figure 4**). A peptide ester which contains a biotin, a TEV cleavage site and an Abu-tag (α-aminobutyric acid) is enzymatically coupled to free protein N-termini with subtiligase. The reaction is specific for α- over ε-amines (Braisted et al., 1997). Mostly neo N-termini are modified because 68% of the yeast and 85% of the human proteins are acetylated at the protein N-term (Van Damme et al., 2011). Labeled proteins/peptides are pulled down with avidin or streptavidin conjugated beads. After tryptic digestion and washing, N-terminal peptides are released using TEV protease. The Abu-tag enables the discrimination between labeled N-terminal and background peptides. Quantification can be done label-free or with other label-based methods (Wiita et al., 2014). Like with TAILS and COFRADIC, exact cleavage sites of proteases can be analyzed. A drawback of the method is that typically a high protein amount of 30–300 mg of cell lysate is used according to Wiita et al. (2014; **Table 1**). The subtiligase method was used for different studies, such as to identify caspase substrate profiles (Agard et al., 2010) and to analyze cell apoptosis (Crawford et al., 2013).

#### Summary of Methods

All described methods with the exception of CSF analysis are suitable for any cell culture experiment including primary cells as well as cell lines or bacterial cells. However, cells that require serum or other high protein containing supplements are best analyzed using SPECS or AHA labeling to enrich selectively for cell derived proteins. The other methods are better suited for serum free or low serum conditions.

In vivo samples can be analyzed using all methods except the metabolic labeling methods SPECS and AHA labeling which would cause extensive costs for in vivo labeling. In the case of body fluids, glyco-capturing has the advantage to enrich for glycosylated proteins which include 89% of type 1 and 2 transmembrane proteins according to UniProt.

TAILS, COFRADIC and the subtiligase method have the advantage to facilitate protein cleavage site determinations. Thus, those methods are well suited for cell-free in vitro cleavage assays such as incubation of cell secretomes, potential substrates, peptide libraries, or even whole cell lysates with the protease of interest. Such an approach was reported e.g., for meprin α and β substrate identification by N-TAILS (Jefferson et al., 2013).

#### IDENTIFICATION OF MEMBRANE PROTEASE SUBSTRATES USING PROTEOMICS

In the following paragraphs we will describe the application of several of the proteomic methods described above to the identification of substrates for the Alzheimer-related proteases BACE1 and BACE2, ADAM10, ADAM17 as well as γ-secretase and its distant homolog SPPL3.

#### β-Site Amyloid Precursor Protein Cleaving Enzyme (BACE) 1 and 2

The beta secretase BACE1 is known to shed the ectodomain of APP which leads to the release of the sAPPβ fragment. Subsequently, cleavage of the APP C-terminal fragment (CTF) by γ-secretase generates amyloid β peptides which can form plaques in the brain, a pathological hallmark of AD (Selkoe and Hardy, 2016). Therefore, BACE1, which is highly expressed in neurons, is a major drug target to inhibit Aβ generation and thus delaying or preventing the onset of AD. Different pharmaceutical companies have developed BACE1 inhibitors (Vassar, 2014). However, several BACE1 inhibitors have failed in the clinic because of side effects that may not be related Müller et al. Brain Proteomics

to BACE1 inhibition (Barão et al., 2016). The inhibition of BACE1 might also lead to mechanism-based side-effects because it also cleaves other transmembrane proteins. This is further emphasized by the finding that BACE1−/<sup>−</sup> mice show various phenotypes (Vassar, 2014). Hence, it is essential to identify BACE substrates and to characterize the biological function of the full-length proteins as well as the resulting BACE1 cleavage products.

In recent years, different MS-based proteomic studies were carried out with the goal to identify BACE substrates in an unbiased manner. In 2009, Hemming et al. performed a study with HEK and HeLa cells overexpressing BACE1 and compared the secretome with cells transfected with a control vector (Hemming et al., 2010). Cells were cultured and metabolically labeled in serum-free SILAC medium to enable a MS-based secretome analysis. This study identified 69 putative BACE1 substrates (65 TM type I, 1 TM type II and 3 GPI anchored proteins) that were enriched in the secretome of HEK and/or HeLa cells upon BACE1 overexpression. Different hits were further validated by immunoblotting.

A similar approach was used in a different study (Ivankov et al., 2013). However, even though well-validated BACE1 substrates such as APP, APLP1 and APLP2 were identified, overexpression of BACE1 is known to lead to artificial cleavage of some membrane proteins. One reason is that overexpressed BACE1 can be active in the endoplasmic reticulum (Huse et al., 2002), whereas under endogenous conditions it cleaves in acidic cellular compartments such as trans-Golgi network and endosomes (Vassar et al., 1999). One example is the protein LRP1, which was identified as a BACE1 substrate candidate upon BACE1 overexpression (von Arnim et al., 2005), but did not show any change in cleavage upon inhibition of endogenous BACE1, at least not in primary neurons (Kuhn et al., 2012).

In 2012, two proteomic studies were published which used primary neurons treated with a BACE inhibitor to identify proteins with reduced abundance in the secretome. Thus, these studies were based on endogenous levels of BACE1 and its substrates. The study of Zhou et al. (2012) used primary neurons cultured in Neurobasal medium without protein supplements. N-propionylation was used to differentially label secreted proteins of control and BACE inhibitor treated samples. Finally, 13 putative BACE substrates could be identified that showed reduced abundance in the secretome of inhibitor treated neurons. Additionally, several experiments were carried out to validate L1 and CHL1 as BACE1 and γ-secretase substrates.

For the second study of 2012, SPECS was used for primary neurons treated with a BACE inhibitor or DMSO (Kuhn et al., 2012). This led to the identification of 34 BACE substrate candidates. Seven of them were also validated by immunoblots of BACE inhibitor treated neurons, BACE1−/<sup>−</sup> neurons and BACE1−/<sup>−</sup> brain homogenates.

One of the substrates identified in both proteomic studies in 2012 is the cell adhesion protein CHL1. Subsequent studies further validated CHL1 as a BACE1 substrate in vivo and demonstrated that BACE1-cleavage of CHL1 is required for correct axon targeting in the olfactory bulb and the hippocampus of mice (Hitt et al., 2012; Barão et al., 2015).

Another proteomic study used the pancreatic β islet cell line Min6 to identify BACE1 and BACE2 substrates (Stützer et al., 2013). This study employed Min6 cells that were overexpressing BACE1 and/or BACE2, cells with single or double KDs of BACE1 and BACE2 as well as control cells. Substrates were identified by glyco-capturing. For validation of hits from the first screen, the results of seven proteins were further validated by immunoblotting. The same study also used primary islets from BACE1−/−, BACE2−/<sup>−</sup> and BACE double KO (DKO) mice as well as BACE inhibitor treated islets for validation. Relative protein quantification of 56 candidates was performed using targeted proteomics (selected reaction monitoring). Finally, 40 candidates showed an accumulation in cell lysates and/or reduced abundance in cell supernatants (≥1.25-fold) in at least one of the BACE KO or inhibition conditions. The proteins SEZ6L, SEZ6L2 and TMEM27 were further validated as BACE2 substrates in murine primary islet cells (WT vs. BACE1−/−, BACE2−/<sup>−</sup> and BACE DKO) by immunoblotting.

Finally, a label-free quantitative proteomic analysis of CSF from BACE1−/<sup>−</sup> was performed to identify BACE1 substrates in vivo (Dislich et al., 2015). In this study, 10 BACE1 substrates or substrate candidates showed to have a significantly lower abundance in BACE1−/<sup>−</sup> CSF (APP, APLP1, APLP2, CHL1, CNTN2, NCAM1, PLXDC2, PAM, PTPRN2, SEZ6L2) indicating that CSF proteomics is able to identify and validate BACE1 substrates in vivo. Furthermore, APLP1 and 2 were validated by immunoblotting and PTPRN2, PLXDC2 as well as ENPP5 were confirmed by in vitro assays as BACE1 substrates.

Taken together, the long list of BACE1 substrates demonstrates a central role for BACE1 in basic neurobiology. Whether the substrates and their functions have an impact on the suitability of BACE1 as a drug target in AD, remains to be carefully monitored in future studies.

#### ADAM10

ADAM10 acts as the constitutively active APP α-secretase and is a drug target for AD (Jorissen et al., 2010; Kuhn et al., 2010; Saftig and Lichtenthaler, 2015). ADAM10 is a ubiquitously expressed metalloprotease of the adamalysin family that regulates through shedding the function of several transmembrane proteins, thereby playing a crucial role in cellsignaling and development. The early embryonic lethality of ADAM10-deficient mice has been associated with loss of Notch signaling, that emerged to be a major ADAM10 substrate (Hartmann et al., 2002). Moreover, mice with a conditional knock-out of ADAM10 in neurons, show postnatal lethality at about 3 weeks and display numerous phenotypes in the brain, including impaired synaptic function and disorganized laminar architecture of the neocortex. However, the underlying substrates were largely unknown. A recent study used SPECS and identified around 90 substrate candidates for ADAM10 in primary murine neurons (Kuhn et al., 2016). Several of them were validated by immunoblots. One of the substrates is the cell adhesion protein NrCAM for which it was demonstrated that its loss of cleavage in ADAM10-deficient mice correlates with deficits in axon targeting in the olfactory bulb in mouse brains (Kuhn et al., 2016). More of the newly identified ADAM10 substrates are likely to be assigned in the future to the numerous phenotypes in ADAM10-deficient mice and will enhance our understanding of the broad neurobiological functions of this protease.

#### ADAM17

The metalloprotease ADAM17 is a homolog of ADAM10. When activated, it can act as an additional α-secretase and may reduce Aβ levels (Caccamo et al., 2006). ADAM17 is also known as TNF-α converting enzyme (TACE) and was the first sheddase to be identified, as the enzyme responsible for releasing the soluble ectodomain of TNF (Moss et al., 1997). ADAM17 plays a crucial role in cellcell communication, being able to release not only TNF, but also several other transmembrane proteins, including cytokines, adhesion molecules, receptors and growth factors. ADAM17 deficient mice display several abnormalities at birth, including open eyes and skin defects, that phenocopy mice lacking EGF receptor or a number of its ligands, which are known substrates of ADAM17 (Blobel, 2005). Most ADAM17 substrates have been identified through candidate approaches (Qian et al., 2016). Proteomics has not been extensively used to uncover ADAM17 substrates. However, a few secretome analyses have been performed which searched for transmembrane proteins undergoing shedding in response to specific stimuli, such as lipopolysaccharide (LPS) and 12-O-tetradecanoylphorbol 13-acetate/Phorbol 12-myristate 13-acetate (TPA/PMA), which are known to also activate ADAM17. One study identified a number of transmembrane proteins, such as CSF1R and Sema4D, that are shed by metalloproteinases in response to LPS or TPA in macrophage-like cells (Shirakabe et al., 2014). In order to investigate proteomic changes induced by LPS in macrophages, one study used a method similar to SPECS (Eichelbaum and Krijgsveld, 2014), whereas another group performed secretome analysis from a small number of cells cultured without serum (Meissner and Mann, 2014). Together with a list of known substrates of ADAM17, these studies identified a number of proteins that can also potentially be cleaved by ADAM17. However, these proteins were not validated as ADAM17 substrates so far. Another study specifically investigated changes in the secretome of ADAM17−/<sup>−</sup> mouse embryonic fibroblasts (MEFs) by using SILAC or label-free based approaches (Kawahara et al., 2014). Label free secretome analysis identified 179 proteins, which were significantly down-regulated in ADAM17-deficient MEF cell supernatants. Transmembrane proteins, including TNFR2 and syndecan-4, were strongly reduced in the secretome of ADAM17−/<sup>−</sup> MEFs, suggesting that they are ADAM17 substrates. Furthermore, a proteomic study of ADAM17 deficient epidermis was performed which showed pronounced changes in a number of proteins involved in barrier formation,

Functions of ADAM17 in the brain have been little explored so far and no proteomic study has as yet been done to address this issue specifically. Yet, the role of ADAM17 in inflammation suggests that ADAM17 is also involved in various neuroinflammatory conditions.

#### γ-Secretase

γ-secretase has been a major drug target in AD in the past. It is a protease complex that cleaves transmembrane type 1 proteins within or close to their transmembrane domain. While γ-secretase only directly sheds the ectodomain of a single, naturally short substrate (Laurent et al., 2015), it typically requires shedding of its substrates in order to cleave them within the transmembrane domain. In 2008, a proteomic study was performed to identify γ-secretase substrates in HeLa cells (Hemming et al., 2008). Therefore, cells were differentially labeled with the SILAC method and treated with the γ-secretase inhibitor DAPT or DMSO as a control. Since γ-secretase cleavage usually requires previous shedding by other proteases (Struhl and Adachi, 2000), such as BACE1 or ADAM10, substrates are commonly identified by an accumulation of the CTF upon γ-secretase inhibition. Hence, SDS-PAGE of membrane fractions was applied for proteomic γ-secretase substrate profiling to separate CTFs from full-length proteins (Hemming et al., 2008). The gels were cut into 10 slices and in-gel digestion was performed with trypsin. Relative quantification between DAPT and DMSO was done separately for each fraction. CTFs with a DAPT/DMSO intensity ratio larger than 1.86 were considered as enriched. Overall, CTFs of 13 proteins, among them APP and APLP2 showed enrichment for DAPT treatment. Very likely, this approach missed to identify more γ-secretase substrates as CTFs of proteins with a short cytoplasmic domain are hard to quantify. Additionally, low molecular weight peptides and proteins offer just few tryptic peptides and are often lost during washing steps of the in-gel digestion protocol (Klein et al., 2007; Müller et al., 2010).

#### Signal Peptide Peptidase-Like 3 (SPPL3)

The signal peptide peptidase (SPP) family has five members, SPP, SPPL2A, 2B, 2C and 3. They are distant homologs of γ-secretase and belong together with γ-secretase to the intramembranecleaving aspartic proteases. The SPP family cleaves type II transmembrane proteins within or close to their transmembrane domain (Voss et al., 2013). Similar to most substrates of γ-secretase, ectodomain shedding by another protease is required to enable cleavage by SPP, SPPL2A and B (Voss et al., 2013), whereas it has not been investigated so far if SPPL2C is proteolytically active and which biological role it has. An exception to the other family members is SPPL3, which does not require prior shedding of its substrates by another protease (Krawitz et al., 2005; Voss et al., 2013). Therefore, type II transmembrane proteins can be directly shed by SPPL3, which is mostly localized in the Golgi apparatus (Krawitz et al., 2005; Voss et al., 2013).

A global secretome analysis of HEK cells overexpressing SPPL3 as well as MEF SPPL3−/<sup>−</sup> cells were used for unbiased substrate identification (Kuhn et al., 2015). For this purpose, the SPECS method was applied to enrich and quantify secreted glycoproteins. The majority of identified SPPL3 substrates are involved in modifying N- or O-glycosylation. Hence, SPPL3 has a fundamental role in regulating different protein glycosylation pathways. Whether and how this role impacts the brain, needs to be studied in the future.

#### CONCLUSION AND OUTLOOK

The first substrates of sheddases and intramembrane proteases were largely identified by candidate approaches, partially driven by the phenotypes of the corresponding protease knock-out mice. One example is the loss-of-Notch-function phenotype which allowed to identify Notch as a substrate for ADAM10 and γ-secretase (De Strooper et al., 1999; Struhl and Greenwald, 1999; Hartmann et al., 2002). Another example is the hypomyelination of BACE1-deficient mice which led to the identification of neuregulin-1 as a BACE1 substrate (Hu et al., 2006; Willem et al., 2006).

In the last 10 years the focus has shifted to the use of proteomics as an unbiased method for the systematic substrate identification of sheddases— on which we focus in this review, but also of other proteases, including metalloproteases in inflammation and caspases in apoptosis. Generally, the field of proteomics dealing with proteases, protease inhibitors and protein degradation is referred to as degradomics (López-Otin and Overall, 2002). Given the advance in mass spectrometric instrumentation and the development of powerful degradomic methods as described here, we are likely to see many more systematic substrate identification studies being published over

#### REFERENCES


the next years. This will include many of the over 40 sheddases and intramembrane proteases, where substrates and functions are little understood to date. The degradomics methods are likely to be further improved to allow analysis of lower sample amounts, and analyses will be increasingly done with in vivo material, such as a tissue samples and body fluids.

Several of the sheddases and intramembrane proteases—such as BACE1, ADAM10, ADAM17 and γ-secretase—are major drug targets for neurodegeneration or inflammatory diseases. Other proteases of these families will likely turn out to be drug targets for additional diseases. Thus, degradomic studies of these exciting protease families will not only allow us to understand their basic functions in the brain and other tissues, but will also enable us to better evaluate their therapeutic potential and to predict possible side effects of drugs modulating the protease activity. Additionally, the cleaved ectodomains of the protease substrates in body fluids, such as CSF, hold the potential to be used as companion diagnostics for monitoring whether and how patients respond to protease inhibitors. Thus, the new proteomic methods have paved the way for even faster discovery in basic and applied neuroscience and other research fields.

#### AUTHOR CONTRIBUTIONS

SAM, SDS and SFL co-wrote the review.

#### ACKNOWLEDGMENTS

Research in our laboratory is funded by the Deutsche Forschungsgemeinschaft (FOR2290), Fonds voor Innovatie door Wetenschap en Technologie (IWT), the Centers of Excellence in Neurodegeneration Research (CoEN), the Helmholtz Israel program, the Research Award of the Breuer Foundation and the Bundesministerium für Bildung und Forschung (JPND). SDS is supported by a Marie Sklodovska-Curie Individual Fellowship.

that promotes amyloid pathology and cognitive decline in a transgenic mouse model of Alzheimer's disease. Cell Mol. Life Sci. 73, 217–236. doi: 10. 1007/s00018-015-1992-1


implications for proteomic examination of the albuminome and albumindepleted samples. Proteomics 9, 2021–2028. doi: 10.1002/pmic.200800686


dominates over proteolytic processing by cathepsins in pancreatic tumors. Cell Rep. 16, 1762–1773. doi: 10.1016/j.celrep.2016.06.086


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Müller, Scilabra and Lichtenthaler. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Physiological Functions of the β-Site Amyloid Precursor Protein Cleaving Enzyme 1 and 2

#### Riqiang Yan\*

Department of Neurosciences, Lerner Research Institute, Cleveland Clinic Foundation, Cleveland, OH, USA

BACE1 was discovered as the β-secretase for initiating the cleavage of amyloid precursor protein (APP) at the β-secretase site, while its close homology BACE2 cleaves APP within the β-amyloid (Aβ) domain region and shows distinct cleavage preferences in vivo. Inhibition of BACE1 proteolytic activity has been confirmed to decrease Aβ generation and amyloid deposition, and thus specific inhibition of BACE1 by small molecules is a current focus for Alzheimer's disease therapy. While BACE1 inhibitors are being tested in advanced clinical trials, knowledge regarding the properties and physiological functions of BACE is highly important and this review summarizes advancements in BACE1 research over the past several years. We and others have shown that BACE1 is not only a critical enzyme for testing the "Amyloid Hypothesis" associated with Alzheimer's pathogenesis, but also important for various functions such as axon growth and pathfinding, astrogenesis, neurogenesis, hyperexcitation, and synaptic plasticity. BACE2 appears to play different roles such as glucose homeostasis and pigmentation. This knowledge regarding BACE1 functions is critical for monitoring the safe use of BACE1 inhibitors in humans.

#### Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Stefan Lichtenthaler, German Center for Neurodegenerative Diseases (DZNE), Germany Claus Pietrzik, Johannes Gutenberg-Universität Mainz, Germany

> \*Correspondence: Riqiang Yan yanr@ccf.org.

Received: 19 October 2016 Accepted: 22 March 2017 Published: 19 April 2017

#### Citation:

Yan R (2017) Physiological Functions of the β-Site Amyloid Precursor Protein Cleaving Enzyme 1 and 2. Front. Mol. Neurosci. 10:97. doi: 10.3389/fnmol.2017.00097 Keywords: Alzheimer's disease, amyloid plaques, amyloid precursor protein, secretase, BACE1, BACE2, aspartic protease, BACE substrates

#### INTRODUCTION

For over 30 years, the study of β-amyloid (Aβ) peptides has been the largest field of research geared toward understanding Alzheimer's pathogenesis and therapeutic intervention. After the molecular cloning of amyloid precursor protein (APP; Kang et al., 1987; Tanzi et al., 1987), it became clear that Aβ is a small fragment of APP that is located in the region partially spanning the transmembrane (TM) domain. The excision of Aβ from APP requires the sequential cleavage of APP by both β- and γ-secretase. In 1999, four groups independently reported identification of membrane-anchored aspartic protease as the β-secretase (Hussain et al., 1999; Sinha et al., 1999; Vassar et al., 1999; Yan et al., 1999), while the γ-secretase consists of presenilin-1 or -2, which forms a complex with additional multi-TM proteins nicastrin, pen2, and Aph1 (De Strooper et al., 1998; Wolfe et al., 1999; Li et al., 2000; Yu et al., 2000; Francis et al., 2002). After initial discovery, the β-secretase was named as BACE1, meaning β-site APP converting enzyme (Vassar et al., 1999). For the past 17 years, extensive efforts have been focused on the development of compounds that specifically inhibit BACE1 activity for Alzheimer's disease (AD) therapy, and several major hurdles of producing brain-penetrable small molecular inhibitors have been overcome. Several highly potent BACE1 inhibitors have been developed by pharmaceutical and biotech companies

and have been advanced to phase II/III clinical trials (see reviews by Ghosh and Osswald, 2014; Oehlrich et al., 2014; Vassar, 2014; Yan, 2016). Concurrently, BACE1 has also been shown to cleave multiple membrane substrates and its physiological roles in neuronal functions continue to be revealed (Vassar et al., 2014; Yan and Vassar, 2014; Hu et al., 2015a; Barao et al., 2016). Because of the importance of BACE1 inhibitors for therapeutic benefits in AD, this review will focus on summarizing the growing body of knowledge regarding the biological functions of BACE1.

#### BACE1 IS A TYPICAL ASPARTIC PROTEASE

In the initial molecular cloning of β-secretase, the Pharmacia group was exploring whether an aspartic protease functions as such a secretase (Yan et al., 1999). Two other groups had also screened for β-secretase activity through their aspartic protease collections (Hussain et al., 1999; Lin et al., 2000). Independently, all five groups demonstrated that the β-secretase is a type I TM aspartic protease having a classical bilobal structure with two active aspartate motifs (D93TG and D289SG). Although not broadly cited, this enzyme was also named to be memapsin 2 based on the standard nomenclature for aspartic proteases (Lin et al., 2000). The crystal structure of BACE1 shows gross similarity to other aspartic proteases, but the catalytic pocket is more open and less hydrophobic than that of other human aspartic proteases (Hong et al., 2000).

BACE1 is synthesized in the endoplasmic reticulum (ER) as a precursor protein having pre- (residues 1–21), pro- (residues 22–45), core protease- (residues 46–460), TM- (residues 461–477), and C-terminal domains (residues 478–501). BACE2 has 518 amino acids and almost identical structural organization (**Figure 1**). Both proteins share 59% identity and are two known aspartic proteases docked on the membrane through the type I TM domain (Hussain et al., 1999; Yan et al., 1999; Bennett et al., 2000; Lin et al., 2000). In the aspartic protease family, the prodomain usually assists in protein folding (Baker et al., 1993) and can flip to block the active pocket by conferring zymogen-like properties (Khan and James, 1998). Therefore, this prodomain is normally removed by furin-like proprotein convertases during maturation in the Golgi compartment to produce active enzyme. Interestingly, this prodomain has weak inhibitory effects and proBACE1 is enzymatically active (Shi et al., 2001). Consistent with this, BACE1 is active in the ER compartment (Yan et al., 2001a). While inhibiting prodomain removal has a weak effect on blocking BACE1 activity, enhancing shedding of BACE1 near the ectodomain region impacts its cleavage of APP. It is known that docking on the lipid bilayer is essential for BACE1 to cleave APP at the β-secretase site, as removing this TM domain abolishes cleavage of APP at the β-secretase site in cells (Yan et al., 2001a). This is consistent with observations that many soluble aspartic proteases cleave APP peptides at the β-secretase site in vitro, but not in vivo (Brown et al., 1996; Chevallier et al., 1997; Mackay et al., 1997; Gruninger-Leitch et al., 2000; Turner et al., 2002; Tomasselli et al., 2003).

BACE1 also undergoes other multiple post-translational modifications: it is N-glycosylated on four sites (N-153, N-172, N-223, and N254; Haniu et al., 2000), acetylated on seven Lys residues (Lys-126, Lys-275, Lys-279, Lys-285, Lys-299, Lys-300, and Lys-307) in the ER (Costantini et al., 2007), ubiquitinated at Lys-501 for the control of endocytosis to lysosomes for degradation (Tesco et al., 2007; Kang et al., 2012) and at Lys-203 and Lys-382 for the proteasomal degradation of BACE1 (Wang R. et al., 2012), palmitoylated in four C-terminal Cys residues (Cys474/478/482/485) for lipid raft localization (Benjannet et al., 2001; Vetrivel et al., 2009; Bhattacharyya et al., 2013), and phosphorylated at Ser-498 (Walter et al., 2001), which is linked to BACE1 cellular trafficking (Pastorino et al., 2002; He et al., 2005). Phosphorylation of BACE1 at Thr252 by the p25/Cdk5 complex appears to increase BACE1 activity (Song et al., 2015). A recent study suggests that glycol modifications of BACE1 by N-acetylglucosamine (GlcNAc), a sugar-bisecting enzyme highly expressed in brain, regulates BACE1 stability (Kizuka et al., 2015). Loss of GlcNAc will lead to enhanced degradation of BACE1 by increased trafficking of BACE1 to lysosomes from the late endosomes. This is reminiscent of deubiquitinylation by ubiquitin-specific peptidase 8 (USP8), an endosome-associated deubiquitinating enzyme. Studies have shown that RNAimediated depletion of USP8 increased BACE1 ubiquitination on Lys-501, promoted BACE1 accumulation in the early endosomes and late endosomes/lysosomes, and decreased levels of BACE1 in the recycling endosomes (Yeates and Tesco, 2016). It should be noted that most post-translational modifications, except for the disulfide linkage, can regulate BACE1 activity but are not necessary for BACE1 proteolytic activity per se, as recombinant BACE1 produced in bacteria lacks these modifications but is sufficiently active.

#### CELLULAR TRAFFICKING OF BACE1

BACE1 is first synthesized in the ER and then is distributed to various cellular compartments such as the Golgi network, endosomes, and cell surface, where the luminal BACE1 catalytic domain will cleave its cellular substrates such as APP. Like other aspartic proteases, the catalytic activity of BACE1 is elevated in more acidic environments (Shimizu et al., 2008). Because of this preferential activation, altered localization or cellular trafficking of BACE1 in cellular compartments impacts generation of Aβ from the cleavage of APP (Vassar et al., 2009).

Several proteins have now been shown to bind BACE1 and to alter cellular localization. Golgi-localized γ-ears containing proteins from the ADP ribosylation factor-binding (GGA) family were first shown to bind to BACE1 via the dileucine motif, and this binding impacts not only BACE1 endosomal trafficking but also cellular stability (He et al., 2002, 2005; Wahle et al., 2005; Tesco et al., 2007; Santosa et al., 2011; Walker et al., 2012; von et al., 2015). Depletion of both GGA1 and GGA3 induces a rapid and robust elevation of BACE1, and such an effect is likely inhibited by flotillin, which can compete with GGA proteins for binding to the same dileucine motif in the BACE1 tail (John et al., 2014). Reticulon (RTN) proteins, mainly

localized in the ER, have been shown to bind BACE1 and this binding induces retention of BACE1 in the ER, which has a relatively neutral pH environment and thus is less favorable for APP cleavage by BACE1 (Sharoar and Yan, 2017). On the other hand, increased trafficking of BACE1 to the more acidic endosomes by cellular trafficking proteins such as the Vps10p domain-sorting receptor sortilin (Finan et al., 2011), the small GTPase ADP ribosylation factor 6 (ARF6; Sannerud et al., 2011), Rab-GTPases Rab11 (Udayar et al., 2013), and Sorting nexin 12 (Zhao et al., 2012) results in significant increases in Aβ generation.

In neurons, BACE1 is also targeted to axons and presynaptic terminals (Kandalepas et al., 2013) and its axonal transport is regulated by altered levels of calsyntenin-1 (Steuble et al., 2012; Vagnoni et al., 2012), retromer vps35 (Wen et al., 2011; Wang C.L. et al., 2012), RTN3 (Deng et al., 2013), Rab11 and Eps15 homology domain proteins (Buggia-Prevot et al., 2013, 2014; Udayar et al., 2013). The enhanced localization of BACE1 at synaptic sites is suggested to increase release of Aβ by the synaptic terminals and directly facilitates amyloid deposition in AD patients (Sadleir et al., 2016).

#### IDENTIFIED BACE1 SUBSTRATES

BACE1 cleaves many cellular substrates other than APP, so its biological functions will be affected by altered cleavage of these substrates. Various biochemical and proteomic approaches have been employed to search for BACE1 substrates. Initially, optimal BACE1 cleavage sites were explored (Gruninger-Leitch et al., 2002; Turner et al., 2002; Tomasselli et al., 2003), but this effort together with the use of bioinformatic tools was not successful in determining potential substrates. Instead, several BACE1 substrates were identified through candidatebased characterizations. For example, neuregulin-1 (Nrg1) was identified via the finding that BACE1 plays a role in regulating myelination (Hu et al., 2006; Willem et al., 2006).

Using unbiased proteomic analysis of cultured media from cell lines with or without overexpression of BACE1, Selkoe and his colleagues reported 68 putative BACE1 substrates (Hemming et al., 2009). By using the developed secretome protein enrichment with click sugars (SPECS) method, Lichtenthaler and his colleagues identified 34 membrane-associated proteins as potential BACE1 substrates (Kuhn et al., 2012). This group has also compared cerebrospinal fluids from BACE1-null vs. wild-type mice using label-free quantitative proteomics, and they identified additional novel substrates while validating several previously reported substrates (Dislich et al., 2015). Among these reported BACE1 substrates, the proteins listed in **Table 1** have gained the most attention and/or are fully validated.

#### BIOLOGICAL FUNCTIONS ATTRIBUTABLE TO BACE1-CLEAVABLE SUBSTRATES

As outlined above, the list of BACE1 substrates has grown in recent years (**Table 1**). It is highly important to understand the biological functions associated with BACE1 cleavage of these individual substrates. The following sections summarize studies on this topic that have been published in recent years.

TABLE 1 | Partial list of characterized BACE1 substrates.


↓ BACE2 cleavage site. <sup>∨</sup>α-secretase cleavage site.

# Astrogenesis and Neurogenesis

BACE1 was found to cleave Jagged-1 (Jag1), a type I TM ligand for Notch receptors (Hu et al., 2013). BACE1 mainly cleaves Jag1 at the A1050–A<sup>1051</sup> site near the TM domain (He et al., 2014), and abolished cleavage in BACE1-null mice causes elevation of full-length Jag1, which in turn enhances Notch activation by producing high levels of Notch intracellular domain (NICD; Hu et al., 2013). Notch is highly expressed during neonatal stages and then gradually declines, with persistent low levels of expression in adulthood. BACE1 and Jag1 expression concurrently have the exact same patterns: high levels in neonatal stages and gradual reduction thereafter. Such parallel expression patterns in early developmental stages imply indispensable roles during development. Indeed, BACE1 deficiency causes enhanced astrogenesis and reduced neurogenesis, which is restricted to the hippocampal dentate gyrus (Hu et al., 2013). BACE1 and Jag1 are mainly expressed by pyramidal neurons, while Notch is highly expressed in neural stem cells in the subgranular zone. In earlier studies, high NICD activity was shown to inhibit neurogenesis in the postnatal dentate gyrus and to act as a switch from neurogenesis to gliogenesis (Morrison et al., 2000; Breunig et al., 2007). Hence, it appears that BACE1 regulates Notch signaling, via cleavage of Jag1, to control proliferation and differentiation of multi-potent neural precursor cells into neurons or astrocytes in the early developmental hippocampus.

#### Myelination and Remyelination

The role of BACE1 in the control of myelination during development and of remyelination in the adult appears to occur through its cleavage of Nrg1 (Fleck et al., 2012; Hu et al., 2015a). Nrg1, which is one of the largest genes in the human genome with 33 spliced isoforms due to specific uses of six different transcriptional initiation sites as well as multiple splicing isoforms, is typically recognized by the presence of exons coding for the epidermal growth factor (EGF)-like domain (Holmes et al., 1992; Chang et al., 1997). Although Nrg1 isoforms with six different membrane topology types are found in the brain, types I and III β1 Nrg1 isoforms are mainly expressed by neurons and have been established as BACE1 substrates (Hu et al., 2006, 2008; Willem et al., 2006; La et al., 2011; Fleck et al., 2013). BACE1 specifically cleaves Nrg1 at the F-M site, which is located 10 residues before the TM domain and is shared by both types I and III β1 isoforms (Hu et al., 2008; La et al., 2011; Fleck et al., 2013). These two Nrg1 isoforms can also be cleaved by ADAM10 and ADAM17 at the F-Y site, which is seven residues upstream of BACE1 cleavage site. After cleavage by either BACE1 or ADAM10/17, type I Nrg1 releases its N-terminal fragment (Nrg1-ntf) to the extracellular space, where Nrg1-ntf binds to ErbB receptors (ErbB2 and ErbB3 heterodimers and ErbB4 homodimers) on nearby cells in a paracrine fashion, while type III Nrg1-ntf, which remains tethered to the lipid bilayer due to a hydrophobic CRD in its N-terminus, signals to adjacent cells in a juxtacrine fashion (Warren et al., 2006).

Activated Nrg1 signaling, mainly initiated by type III Nrg1, is critical for optimal myelination: mice with reduced Nrg1 signaling activity exhibit hypomyelination of peripheral nerves during development (Michailov et al., 2004; Taveggia et al., 2005). BACE1-null mice also display hypomyelination in their sciatic nerves, which are typically ensheathed by Schwann cells (Hu et al., 2006; Willem et al., 2006). This phenocopy is not only seen in BACE1-null mice but also in zebrafish (van Bebber et al., 2013) and rat (Weber et al., 2017) knockout (KO) models, and is consistent with reduced Nrg1 activity, which leads to decreased downstream signaling events such as reduced Akt phosphorylation and transcriptional expression of myelin genes like myelin basic proteins. Although type III Nrg1 is abundantly expressed in brain neurons (Liu et al., 2011), it has less effect on entheathing axons in the central nervous system, and both Nrg1 heterozygous mice and BACE1-null mice display weak hypomyelination phenotype in central nerves (Hu et al., 2006; Taveggia et al., 2008). Hypomyelination was observed in BACE1 null optic nerves, but not in broad brain regions of BACE1-null mice (Hu et al., 2006).

Interestingly, inhibition of BACE1 produces more dramatic suppression of myelination in a co-culture myelination system than pan inhibition of ADAM proteases (Luo et al., 2011). This is likely related to the presence of an additional BACE1 cleavage site between L-Q, located 16 residues upstream of the EGF-like domain in type III Nrg1 (Fleck et al., 2013). Cleavage of type III Nrg1 at these two BACE1 sites will release the EGF-like domain for signaling through ErbB receptors. This observation further supports the importance of BACE1-dependent Nrg1 signaling in myelination.

When peripheral nerves are severely injured, myelin on proximal segments of damaged axons can be removed due to Wallerian degeneration and regrowing axons will be remyelinated by Schwann cells via contacting regenerating axons in the proximal band of Büngner. BACE1 has been shown to be indispensable for remyelination in nerve crush experiments, as remyelinated axons remained hypomyelinated in BACE1 null sciatic nerves (Hu et al., 2008). In nerve transplantation experiments, it has been further demonstrated that nerve injury induces expression of BACE1 in Schwann cells and that this increased expression of BACE1 in Schwann cells is required for remyelination (Hu et al., 2015b). nerve region leads to shortened internode as well as reduced nerve conduction. BACE1 is also required for initial and optimal remyelination of corpus callosum axons, as demonstrated in cuprizone-induced demyelination experiments (Treiber et al., 2012).

In peripheral nerves, BACE1 cleavages of type I Nrg1 in Schwann cells and type III Nrg1 in axons contribute to normal remyelination. This conclusion is supported by mouse genetic studies using conditional deletion of Nrg1 isoforms or transgenic mice overexpressing either type I or type III Nrg1 (Fricker et al., 2011, 2013; Stassart et al., 2013). These studies show that Schwann cell-derived type I Nrg1 is dispensable for developmental myelination and myelin maintenance, but is required for an autocrine signaling function for remyelination, as loss of Nrg1 expression in Schwann cells severely impairs remyelination after nerve crush. More recently, it is shown that BACE1 can cleave Jag1 and Delta1 in axons and Schwann cells, and abrogated cleavage of these two Notch ligands enhances Schwann cell proliferation (Hu et al., 2017). This abnormally increased Schwann cell density within the given sciatic Hence, by cleaving types I and III Nrg1 as well as Jag1/Delta1 in different cell types, BACE1 controls these two signaling pathways to regulate optimal myelination and remyelination.

#### Epileptic Seizures

BACE1-null mice develop convulsive and spontaneous behavioral seizures in an age-dependent manner, beginning at a young age and becoming more frequent with aging (Kobayashi et al., 2008; Hitt et al., 2010; Hu et al., 2010). Longsustained epileptic seizures in BACE1-null mice likely contribute to neuronal loss in the 2-year-old mouse hippocampus, although neurodegeneration in this region was not evident in young mice (Hu et al., 2010). The molecular mechanism underlying this epileptic seizure activity remains elusive, and cleavages of multiple BACE1 substrates may each contribute. It has been shown that BACE1 cleaves voltage-gated sodium channel β subunits (Kim et al., 2005; Wong et al., 2005; Huth et al., 2011). Voltage-gated sodium channels consist of a heterotrimeric complex of one 260 kDa α-subunit and one or two auxiliary β subunits (Catterall, 2000), and abolished cleavage in BACE1-null mice likely increases surface expression of ion-conducting, channel-forming α-subunits through cellular trafficking (Isom, 2002; Yu et al., 2005). Sodium channel Nav1.2 protein was found to be elevated in BACE1-null hippocampal mossy fiber regions (Hu et al., 2010; Kim et al., 2011). This increase is consistent with greater neuronal excitability, as manifested by more frequent firing with larger amplitude in BACE1-null brain slices and a significant shift of the inactivation curve in the direction of depolarization (Dominguez et al., 2005; Hu et al., 2010). However, the BACE1- and subsequently γ-secretase-cleaved β subunit is also known to enhance gene expression of Na<sup>v</sup> α subunits (Kim et al., 2007) and BACE1 deficiency will reduce the level of Na<sup>v</sup> α subunits (Kim et al., 2011). Pharmacological blockage of sodium channel activity was not sufficient to reduce seizure activities (Hitt et al., 2010). Hence, the altered activity of sodium channel activity is unlikely to be the only explanation for seizure activity.

BACE1 can also cleave KCNE1 and KCNE2, two auxiliary subunits of voltage-gated potassium channels (Sachse et al., 2013). Both KCNE1 and KCNE2 are expressed in brains and altered functions of these two proteins are linked to epileptic seizures (Goldman et al., 2009; Heron et al., 2010). On the other hand, BACE1 deficiency may cause epilepsy through a noneenzymatic mechanism, as BACE1 interacts with an M-currentproducing KCNQ (Kv7) family member, resembling the function of a β-subunit (Hessler et al., 2015). The loss of M-current due to BACE1 deficiency enhances neuronal excitability, which could also contribute to epileptic seizures.

Another family of proteins, the seizure-related gene 6 (Sez6) and its family member Sez6L, was identified as BACE1 substrate through an unbiased proteomic approach and was recently validated as a strong substrate of BACE1 (Kuhn et al., 2012). Their levels in BACE1-null CSF are significantly lowered, reflecting the abrogated cleavage by BACE1 (Pigoni et al., 2016). Although Sez6 KO mice have not been shown to have seizures, Sez6 is suggested to be a susceptibility gene for febrile seizures (Mulley et al., 2011). It remains to understand whether the abolished cleavage of Sez6 in BACE1-null mice contributes to epileptic seizures. Taking all of these findings into consideration, it is reasonable to postulate that multiple factors, including epigenetic factors, contribute to epileptic seizures in BACE1-null mice, which display variable spiking patterns on electroencephalography (Hitt et al., 2010; Hu et al., 2010).

#### Muscle Spindle Defects

Muscle spindles, which are composed of specialized intrafusal muscle fibers, are innervated by afferent axons extending from sensory neurons (Hunt, 1990). Nrg1 in sensory neurons transduces its signals through ErbB2/ErbB3 receptors in muscles to control the formation of muscle spindles (Andrechek et al., 2002; Hippenmeyer et al., 2002; Leu et al., 2003). BACE1 deficiency impairs coordinated muscle function between forelimbs and hindlimbs, resulting in a swaying walking pattern

as well as a reduction in the number of muscle spindles (Cheret et al., 2013). Such an ambulatory defect, likely due to dysfunctional proprioception governed by muscle spindles, is more dramatic in newborns while less severe in BACE1-null adult mice or heterozygous mice (Cheret et al., 2013). This role of BACE1 in reduced muscle spindle maintenance is due to abrogated or reduced cleavage of Ig domain-containing type I β1 Nrg1 (IgNrg1β1) isoforms, which are preferentially expressed by proprioceptive sensory neurons and are sufficient to induce muscle spindle differentiation in animals (Hippenmeyer et al., 2002). Consistently, transgenic mice overexpressing IgNrg1β1 develop supernumerary muscle spindles (Rumsey et al., 2008). If wild-type mice are treated with the BACE1 inhibitor Ly2811376 for 29 days, up to 40% of muscle spindles are lost (Cheret et al., 2013). As discussed previously (Hu et al., 2015a), abolished cleavage of Nrg1 reduces the expression of transcription factors in the early growth response (Egr) family. Egr3, in particular, controls expression of the muscle spindle-specific genes necessary for forming muscle spindle fibers. Hence, BACE1-dependent type I IgNrg1β1 signaling is critical for motor coordination.

# Axonal Growth and Neuronal Migration Defects

The neural cell adhesion molecule close homolog of L1 (CHL1), which is a type I membrane protein and a component of Sema3A receptors, is a natural BACE1 substrate with identified cleavage sites located between Y<sup>1086</sup> and E1087, 18 residues upstream of the TM domain (Kuhn et al., 2012; Zhou et al., 2012). After BACE1 cleavage, which is inducible by Sema3A, CHL1-ntf, and CHL1-ctf are released. The CHL1-ctf appears to induce growth cone collapse in thalamic neurons (Barao et al., 2015), while the soluble CHL1-ntf may interact with neuropilin-1 to influence axon guidance (Hitt et al., 2012; Kuhn et al., 2012; Zhou et al., 2012). BACE1-null mice show axon pathfinding defects with mistargeting olfactory sensory neuron projections to glomeruli in the olfactory bulb and a shortened and disorganized infrapyramidal bundle of the mossy fiber projection from the dentate gyrus to CA3 in the hippocampus (Rajapaksha et al., 2011; Cao et al., 2012; Hitt et al., 2012). It should also be noted that BACE1 and CHL1 are co-localized in the terminals of hippocampal mossy fibers, olfactory sensory neuron axons, and growth cones of primary hippocampal neurons, and that axonal defective phenotypes in BACE1-null mice and CHL1-null mice are correlated, confirming the importance of BACE1 cleavage of CHL1 in neuronal development.

#### Synaptic Dysfunctions

The effects of BACE1 inhibition on APP- or Aβ-mediated synaptic functions has recently been summarized in a separate review (Yan et al., 2016). As mentioned above, BACE1 can cleave type I Nrg1, which is highly important for synaptic functions through the signaling of ErbB4 receptors in the brain (see recent comprehensive review by Mei and Nave, 2014). More relevantly, Nrg1 has been identified as a susceptible gene in schizophrenia, which is a disease of synaptic dysfunction (Stefansson et al., 2002) and BACE1-null mice display schizophrenia-like behaviors, which include positive (hyperactivity and pre-pulse inhibition), negative (social withdrawal), and panel (cognitive functions) behaviors (Savonenko et al., 2008). In recent years, altered functioning of Nrg3, a member of the Nrg gene family, has also been found in association with schizophrenia pathogenesis. Nrg3 is an identified BACE1 substrate (Hu et al., 2008), further showing the importance of BACE1-dependent Nrg1 signaling functions. While hypo-function of Nrg1 is linked to schizophrenia-like behaviors in BACE1-null mice, enhanced expression of either BACE1-cleaved Nrg1-ntf fragment or of full-length Nrg1 surprisingly also induces schizophrenia-like behaviors (Kato et al., 2010; Luo et al., 2013; Yin et al., 2013; Agarwal et al., 2014). This is in line with clinical observations that increased Nrg1 or ErbB4 transcripts and proteins are found in schizophrenia patients (Harrison and Law, 2006; Geddes et al., 2011), supporting the importance of balanced BACE1-cleaved Nrg1 in synaptic functions.

BACE1-null mice also exhibit other synaptic dysfunctions. By electrophysiological recording of brain slices, it was demonstrated that hippocampal activity-dependent long-term potentiation at mossy fibers to CA3 is impaired, while long-term depression is increased (Wang et al., 2008, 2014). Intriguingly, a recent study reported that mice treated with the BACE1 inhibitors SCH1682496 and LY2811376 show impaired cognitive functions (Filser et al., 2015). As the list of identified BACE1 substrates has continued to grow (Hemming et al., 2009; Kuhn et al., 2012; Dislich et al., 2015), many of these potential substrates have been shown to control synaptic plasticity and their roles in BACE1-dependent synaptic functions have begun to gain attention. One such example discussed earlier is Sez6, which has been shown to play a role in synaptic function (Gunnersen et al., 2007). In Sez6 KO mice, dendritic spines are significantly shorted and excitatory synapses are thinner in the deep-layer pyramidal neurons of the somatosensory cortex, showing the importance of Sez6 in forming dendritic arbors and controlling synaptic plasticity. More roles of other BACE1 substrates are likely to emerge over the coming years.

#### Retinopathy

The role of BACE1 in retinal pathophysiology has gained increasing attention in recent years, as several BACE1 inhibitors have been found to cause retinal thinning, lipofuscin accumulation, and vascular dysfunction, which terminated clinical trials (Fielden et al., 2015). BACE1 was initially suggested to mediate these retinopathies in a report that BACE1-null mice were found to develop retinal thinning, apoptosis, reduced retinal vascular density, and an increase in age pigmentation and lipofuscin (Cai et al., 2012). The mechanism is linked to the BACE cleavage of vascular endothelial growth factor receptor (VEGFR1). However, the retinal phenotypes in BACE1-null mice are controversial and are not seen in all lines of BACE1-null mice or in BACE1-null rat (Fielden et al., 2015), suggesting possible off-target toxicity. While retinopathy is closely monitored in BACE1 inhibitor clinical trials, recent studies have shown that it is likely due to cross-inhibition of cathepsin D by BACE1 inhibitors (Zuhl et al., 2016). Likely, this side effect can be


#### TABLE 2 | Characterized BACE2 substrates.

fnmol-10-00097 April 15, 2017 Time: 15:24 # 7

↓ BACE2 cleavage site. <sup>∨</sup>α-secretase cleavage site.

mitigated by developing BACE1 inhibitors with minimal offtarget inhibition of other aspartic proteases such as cathepsin D and E, both of which are important in lysosomal functions.

On the other hand, specific inhibition of BACE1 is likely to benefit retinal functions, as BACE1 activity in retina is elevated in response to stress conditions such as mitochondrial respiratory inhibition or oxidative stress (Xiong et al., 2007). It has been suggested that changes in BACE1 expression appear earlier in the retina than in the brain and precede behavioral deficits, and abnormal expression of BACE1 in the retina appears to be an early pathological change in APP/PS-1 transgenic mice (Li et al., 2016). BACE in the adult retina is mostly present in the plexiform layers, consistent with localization of this enzyme to synaptic terminals (Xiong et al., 2007). In AD, Aβ levels are elevated in neurodegenerative retinas, and this potentially causes damage in retinal function in aging (Dentchev et al., 2003; Gupta et al., 2016; Masuzzo et al., 2016). In this sense, inhibition of BACE1 will be beneficial to retinal function.

#### BACE2 SUBSTRATES AND ITS BIOLOGICAL FUNCTIONS

While BACE2 was discovered simultaneously with BACE1 (Vassar et al., 2014), the functional importance of BACE2 has emerged after the finding that BACE2 cleaves the proproliferative plasma membrane protein Tmem27 and PMEL (see summary in **Table 2**). In pancreatic MIN6 cells treated with a BACE2 inhibitor or siRNA, BACE2 was initially shown to mediate insulin receptor β-subunit (IRβ) expression and surface trafficking (Casas et al., 2010). A separate BACE2 silencing study in murine and human β cells reveals Tmem27, known to promote the preservation of functional β-cell mass, as a BACE2 substrate (Esterhazy et al., 2011). Mice with BACE2 deficiency have been shown to correlatively increase β-cell mass, and improved control of glucose homeostasis is associated with increased insulin levels. Hence, BACE2 inhibition should be beneficial to diabetic patients by controlling β-cell maintenance and glucose metabolism.

To maintain glucose homeostasis, islet amyloid polypeptide (IAPP) in pancreatic β cells needs to co-secrete with insulin, and the formation of IAPP amyloid is a hallmark pathological feature of type 2 diabetes (Mukherjee et al., 2015). BACE2 was recently shown to cleave IAPP at two ectodomain sites (Rulifson et al., 2016) and loss of BACE2 cleavage likely increases IAPP homodimer formation and subsequent production of cytotoxic oligomers and amyloid fibrils. Hence, this study suggests that BACE2 inhibition may lead to β-cell dysfunction due to IAPP accumulation in proteinaceous plaques in and around pancreatic islets. These two controversial aspects will be further resolved in more detailed future studies.

On the other hand, BACE2 inhibition can cause loss of pigmentation, as BACE2 cleaves the integral membrane form of PMEL within the juxtamembrane domain and exerts its role in melanosome biogenesis (Rochin et al., 2013). Although BACE1 is expressed in pigment cells, the level of BACE2 is 37-fold higher as that in retinal pigment epithelial cells, suggesting that BACE2 is the major BACE homolog in pigment cells. Consistently, mice with BACE2 deficiency show loss of pigment in skin and retina. However, BACE2 depletion reduces neither the number of stage IV melanized melanosomes nor the total melanin content. Instead, the loss of BACE2-cleaved PMEL N-terminal fragment impairs the organization of PMEL fibrils into parallel sheets, with a threefold decrease in the number of fibrillar stage II and III melanosomes and a sixfold increase in the number of round organelles containing unstructured aggregates. Hence, BACE2 is required for the formation of PMEL amyloid fibrils and for melanosome morphogenesis, consistent with demonstrations by pharmacological inhibition of BACE1 and BACE2 (Shimshek et al., 2016).

It is also demonstrated that Sez6L and Sez6L2 are effectively cleaved but in rate limiting proteolytic manner in pancreatic islet β-cells by BACE2 (Stutzer et al., 2013). Although Sez6L is also a BACE1 substrate, it is not cleaved by BACE1 in pancreatic cells (Pigoni et al., 2016). Additional BACE2 substrates, explored through proteomic approaches, include CD200, IGF2R, LAMP2, MPZL1, and SORT1 (Stutzer et al., 2013). The functional importance of BACE2 cleavages of these proteins remains to be established.

#### SUMMARY

Inhibition of BACE1 is one the most promising therapeutic targets for treating AD, and five drugs have currently entered into clinical trials (Vassar, 2014; Yan, 2016). While there is great promise for BACE1 inhibition in benefiting Alzheimer's patients, it also raises caution regarding mechanism-based side effects associated with long-lasting inhibition of this enzyme. In addition to the fact that BACE1 is indispensable for proper astrogenesis, axonal growth and migration, myelination and remyelination, neuronal excitation, and synaptic plasticity, BACE1-null mice are also found to be susceptible to early lethality (Dominguez et al., 2005; Weber et al., 2017). Reduced body weight is seen in BACE1-null mice but not in rats. With more efforts, the available BACE1-null mice and rats will identify more shared phenotypes, and these phenotypes will have to be taken into consideration when BACE1 is inhibited for long terms. Moreover, BACE1 inhibition may also cause cross-inhibition with BACE2, as some compounds such as MK8931 appear to be more potent in blocking BACE2 activity (see reviews by Yan, 2016). Although BACE1 and BACE2 exhibit distinct cleavage specificity, substrates like APP, Jag1 and Sez6 family proteins are shared by these two enzymes. The number of studies using BACE2-null mice is increasing, and inhibition of BACE2 may alter glucose homeostasis and pigmentation. Future studies are expected to provide more knowledge regarding the

#### REFERENCES


biological functions of BACE1 and BACE2 in brains and other tissues.

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and approved it for publication.

# FUNDING

This study was supported by grants from the National Institutes of Health to RY (NS074256, AG025493, AG046929, and NM103942) and a grant from the National Multiple Sclerosis Society to RY (RG 4012A1/1).


motif are negative regulators of cleavage by Bace2. Biol. Chem. 393, 473–484. doi: 10.1515/hsz-2012-0104




distribution, and genetic linkage near the Alzheimer locus. Science 235, 880– 884. doi: 10.1126/science.2949367



**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Yan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Metalloprotease Meprin β Is an Alternative β-Secretase of APP

#### Christoph Becker-Pauly <sup>1</sup> \* and Claus U. Pietrzik <sup>2</sup> \*

<sup>1</sup> Unit for Degradomics of the Protease Web, Institute of Biochemistry, University of Kiel, Kiel, Germany, <sup>2</sup> Institute for Pathobiochemistry, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany

The membrane bound metalloprotease meprin β is important for collagen fibril assembly in connective tissue formation and for the detachment of the intestinal mucus layer for proper barrier function. Recent proteomic studies revealed dozens of putative new substrates of meprin β, including the amyloid precursor protein (APP). It was shown that APP is cleaved by meprin β in distinct ways, either at the β-secretase site resulting in increased levels of Aβ peptides, or at the N-terminus releasing 11 kDa, and 20 kDa peptide fragments. The latter event was discussed to be rather neuroprotective, whereas the ectodomain shedding of APP by meprin β reminiscent to BACE-1 is in line with the amyloid hypothesis of Alzheimer's disease, promoting neurodegeneration. The N-terminal 11 kDa and 20 kDa peptide fragments represent physiological cleavage products, since they are found in human brains under different diseased or non-diseased states, whereas these fragments are completely missing in brains of meprin β knock-out animals. Meprin β is not only a sheddase of adhesion molecules, such as APP, but was additionally demonstrated to cleave within the prodomain of ADAM10. Activated ADAM10, the α-secretase of APP, is then able to shed meprin β from the cell surface thereby abolishing the β-secretase activity. All together meprin β seems to be a novel player in APP processing events, even influencing other enzymes involved in APP cleavage.

#### Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Oliver Schilling, University of Freiburg, Germany Ulrich Auf Dem Keller, ETH Zurich, Switzerland

#### \*Correspondence:

Christoph Becker-Pauly cbeckerpauly@biochem.uni-kiel.de Claus U. Pietrzik pietrzik@uni-mainz.de

Received: 26 October 2016 Accepted: 09 December 2016 Published: 05 January 2017

#### Citation:

Becker-Pauly C and Pietrzik CU (2017) The Metalloprotease Meprin β Is an Alternative β-Secretase of APP. Front. Mol. Neurosci. 9:159. doi: 10.3389/fnmol.2016.00159

#### Keywords: meprin β, N-terminal truncated Aβ, APP, shedding, proteolysis

#### INTRODUCTION

To date, more than 35,000 research articles dealing with the amyloid precursor protein (APP) are annotated in Pubmed (13.10.2016) and most of these papers are related to Alzheimer's disease. Nevertheless, APP is still an enigma in terms of its physiological and pathophysiological functions.

APP is a multi-domain glycosylated type 1 transmembrane protein. Earlier studies reported that the ectodomains of APP family proteins have zinc- (Bush et al., 1993) and copper bindingproperties (Simons et al., 2002) and that APP is able to reduce bound Cu2<sup>+</sup> to Cu<sup>+</sup> (Multhaup et al., 1996). Moreover, APP has been proposed to bind extracellular matrix proteins like heparin and collagen (Small et al., 1994), and to have a receptor-like function (Beher et al., 1996). In this context, it became more and more challenging, whether APP can form cellular cis-dimers (Scheuermann et al., 2001), reminiscent of classical receptor dimerization described for the EGF receptor (Schlessinger, 2002). However, there is accumulating evidence from biochemical and structural data that APP can form homodimers (Scheuermann et al., 2001; Kaden et al., 2009; Isbert et al., 2012) as well as heterodimers with its homologs APLP1 and APLP2 (Soba et al., 2005).

To date, at least three domains have been reported to promote APP dimerization: first the E1 domain containing the N-terminal Growth factor like domain (GFLD) and Copper binding domain (CuBD) (Soba et al., 2005). The second dimerization interface is represented by the E2 domain (amino acids 365–699), the largest subdomain of the APP ectodomain, containing the carbohydrateand the juxtamembrane region. Crystallographic and X-ray structure modeling revealed that the E2 region can reversibly dimerize in an antiparallel orientation in solution (Wang and Ha, 2004) and it has been reported that binding of extracellular matrix components, such as heparin, to this domain may also regulate dimerization (Gralle et al., 2006). However, in contrast to Wang and colleagues a study by Dulubova and colleagues could not confirm that the E2 domain does dimerize in solution (Dulubova et al., 2004). A third dimerization interface is located at the extracellular juxtamembrane/transmembrane (JM/TM) boundary, where APP contains three consecutive glycine-xxxglycine (GxxxG) motifs (Munter et al., 2007; Gorman et al., 2008; Kienlen-Campard et al., 2008) one embedded within the Aβ sequence.

Interestingly, detection of APP dimerization in vivo showed a possibility that the efficient processing of APP by α- and βsecretases (see below) may depend on its oligomerization state that results in cooperative effects for these allosteric enzymes (Schmidt et al., 2012).

Although the German psychiatrist Alois Alzheimer was the first to demonstrate a relationship between specific cognitive changes, neurological lesions in the human brain, and clinical history (Alzheimer, 1907), much later the amyloid cascade hypothesis attributed these observations to the presence of the cleavage products of APP in the brain (Hardy and Selkoe, 2002). Alzheimer reported the results of an autopsy on a 55-yearold woman named Auguste Deter and noted the presence of two distinct pathological lesions in Deters brain, which now define Alzheimer's disease (AD): first, the neurofibrillary tangles (NFTs), which accumulate intraneuronal (later shown to be composed of paired helical filaments (PHFs) containing the microtubule-associated protein tau; Goedert et al., 1988, 1989); second, extracellular amyloid deposits in the form of diffuse or neuritic senile plaques (Price et al., 1997). Senile plaques accumulate extracellular and were isolated and purified in 1984 by Glenner and Wong, who showed that it was a ∼4 kDa peptide (Aβ), primarily 40 or 42 amino acids in length, which they speculated was cleaved from a larger precursor (Glenner and Wong, 1984). Subsequently, it has been demonstrated that this peptide fragment originated from a larger precursor protein, named the amyloid-β precursor protein (AβPP, or APP as used here) and was characterized from the analysis of a full-length cDNA encoding a translational product of 695 residues (Kang et al., 1987).

#### CONVENTIONAL APP PROCESSING

Multiple enzymes have been shown to process APP during its lifetime. The non amyloidogenic pathway, in which APP is cleaved within the sequence of the amyloid peptide by a generally named enzyme group called α-secretase, precludes the formation of the full-length Aβ which is found in the amyloid core of senile plaques (Zheng and Koo, 2006). One other pathway leads to the production of Aβ peptides from its precursor after the initial cleavage by a generally named enzyme group called β-secretase (Hussain et al., 1999; Sinha et al., 1999; Vassar et al., 1999; Yan et al., 1999). The first β–secretase identified was then named β-site APP-cleaving enzyme (BACE-1). BACE-1 is a type I membrane-bound aspartyl protease located in the endosomal/lysosomal compartment (Sinha et al., 1999; Vassar et al., 1999). Cleavage of APP by BACE-1 (Vassar, 2002) occurs between methionine 596 and aspartate 597 of APP695 (**Figure 1**), producing two fragments, the secreted N-terminal ectodomain sAPPβ and a 10 kDa, 99-amino-acid-long fragment C99, encompassing the Aβ peptide and the remaining C-terminal part. The optimal pH of BACE-1 activity is ∼4.5, suggesting that the β-site cleavage of APP occurs preferentially in more acidic compartments, such as in endosomes and lysosomes (Vassar et al., 1999).

After α- or β-cleavage, the carboxyl terminal fragments (CTFs) of APP, known as αCTF (C83) and βCTF (C99), respectively, remain membrane-associated and are further cleaved by the γsecretase-complex (Edbauer et al., 2003). The γ-secretase is an aspartyl protease complex (Wolfe et al., 1999), which unlike αand β-secretases, acts within the membrane and cleaves APP at multiple sites (Zhao et al., 2004), releasing either, Aβ and intracellular C-terminal domain fragments (ICDs) or p3 and ICDs (**Figure 1**). This process is called regulated intramembrane proteolysis (RIP) (Brown et al., 2000). However, while the two predominant forms of Aβ and p3 terminate at valine 637 (Aβ40 and p3/40) and alanine 639 (Aβ42 and p3/42) (Haass et al., 1992a), some isolated ICDs are shorter than expected and begin at sites 9–10 amino acid downstream of those residues (Gu et al., 2001).

BACE-1 is described to be the major Aβ generating β-secretase (Hussain et al., 1999; Sinha et al., 1999; Vassar et al., 1999; Yan et al., 1999; Lin et al., 2000). This was convincingly shown when a genetic knock-out of the protease in mice abolished Aβ generation almost completely (Luo et al., 2001; Roberds et al., 2001; Dominguez et al., 2005). In accordance to that, BACE-1 was found to be upregulated in brains of sporadic AD patients (Fukumoto et al., 2002). However, there is strong evidence that certain amounts of Aβ are generated independently of BACE-1. This was supported, when using potent BACE-1 inhibitors in vitro and in vivo (Asai et al., 2006; Nishitomi et al., 2006; Hussain et al., 2007; Stanton et al., 2007; Sankaranarayanan et al., 2008). Interestingly, some studies showed that by inhibition of Aβ1-x generating β-secretase activity, alternative N-terminally truncated Aβ peptides increase (Haass et al., 1995; Schrader-Fischer and Paganetti, 1996; Takeda et al., 2004; Schieb et al., 2010; Mattsson et al., 2012). Analysis of Aβ species in BACE-1 knock-out mice likewise revealed that the generation of Aβ1-x peptides was completely abolished while N-terminally truncated Aβ variants could still be generated (Nishitomi et al., 2006). These N-terminally truncated Aβ peptides are also found in the cerebrospinal fluid, brain tissue, and human blood plasma (Wiltfang et al., 2001; Lewczuk et al., 2004; Takeda et al., 2004;

Güntert et al., 2006; Lewis et al., 2006; Maler et al., 2007; Murayama et al., 2007). Later it was demonstrated that BACE-1 invariably generates two Aβ variants beginning with the aspartate in p1 or p11, therefore other proteases might account for the production of N-terminally truncated peptides (Citron et al., 1995; Vassar et al., 1999). Indeed, heterogeneity of alternative β-secretase cleavage events has been described (Golde et al., 1992; Haass et al., 1992b; Seubert et al., 1992; Busciglio et al., 1993) leading to alternative Aβ peptides other than Aβ1/11-x (Vigo-Pelfrey et al., 1993; Asami-Odaka et al., 1995; Wang et al., 1996), which could also be found in amyloid plaques in vivo (Masters et al., 1985; Güntert et al., 2006). It is not clear whether N-terminally truncated Aβ species are generated via cleavage of APP by yet unknown proteases or via truncation of Aβ1-x after its γ-secretase mediated release, e.g., by aminopeptidase A (Sevalle et al., 2009). In contrast to further subsequent cleavage of already released Aβ peptides, Cathepsin B (Hook et al., 2005, 2014; Kindy et al., 2012), S and L (Schechter and Ziv, 2011) have been discussed to be directly involved in Aβ generation, acting as alternative β-secretases. The enzymatic cleavage events of cathepsins on APP are not fully understood since some groups showed that cathepsins are rather involved in Aβ degradation lowering total Aβ burden (Mueller-Steiner et al., 2006; Letronne et al., 2016).

The amyloid peptides Aβ2-40/42 cannot be assigned to BACE-1 activity and are most likely generated due to an alternative β-secretase cleaving APP between 672Asp/673Ala (Wiltfang et al., 2001; Schieb et al., 2010, 2011). Aβ2-x might act as a precursor and can likewise be processed to Aβ3-x by the alanyl-aminopeptidase activity of aminopeptidase N (Hosoda et al., 1998). This is supposed to occur even under physiological conditions due to activity of cortical aminopeptidase N (Kuda et al., 1997; Wiltfang et al., 2001). It was also discussed that N-terminally truncated Aβ peptides arise when Aβ is degraded by a variety of Aβ degrading enzymes e.g., myelin basic protein, neprilysin, and angiotensin-converting enzyme (Saido and Leissring, 2012). But until recently no proof about the exact mechanisms leading to N-terminally truncated Aβ variants could be given, which changed by the identification of the metalloprotease meprin β as an alternative β-secretase described below.

#### ALTERNATIVE APP PROCESSING

In the last years, more and more focus has been put on modified N-terminally truncated Aβ variants. Increased levels of Aβ2- 42 were detected in AD brains (Wiltfang et al., 2001). This is in line with results showing decreased levels of Aβ2-42 in Becker-Pauly and Pietrzik Meprin β Cleavage of APP

CSF of AD patients (Bibl et al., 2012). Since BACE-1 is not capable in directly generating this peptide, a suggested model for the emergence of N-terminal truncation is the subsequent cleavage of the N-terminus of BACE generated Aβ1-x by either Aβ degrading enzymes like insulin-degrading enzymes (IDE) or neprilysin or the aminopeptidase A (APA) (Arai et al., 1999; Wiltfang et al., 2001; Wang et al., 2006). A candidate directly generating N-terminally truncated Aβ independent of BACE-1 is the metalloprotease meprin β. Meprin β is a multi-domain type I transmembrane protein, member of the astacin family of zinc-endopeptidases that is predominantly present as a dimer at the cell surface (Arolas et al., 2012; **Figure 2**). In 2011 meprin β was introduced as an alternative enzyme involved in APP processing (Jefferson et al., 2011). In 2012, N-terminally truncated Aβ2-40 peptides generated by meprin β (**Figure 1**), dependent on subsequent cleavage of the γ-secretase, but independent of BACE-1, were detected in supernatants of overexpressing cells (Bien et al., 2012). Interestingly, increased mRNA levels of meprin β were measured in AD brain homogenates supporting a potential role for this enzyme in neurodegeneration. Various posttranslational modifications of Aβ peptides have been described ranging from oxidation (Hou et al., 2002; Palmblad et al., 2002) to phosphorylation (Kumar et al., 2011, 2012), nitration (Kummer et al., 2011), glycosylation (Halim et al., 2011) or pyroglutamation of Glu3 of Aβ3-40 (Russo et al., 2002; Wittnam et al., 2012). These modifications have been shown to have an effect on the properties of the peptide. The oxidation at Met35 for example impedes the formation of protofibrils and fibrils from monomers (Hou et al., 2002). Nitration and pyroglutamation both increase the aggregation of Aβ (Schilling et al., 2004; Kummer et al., 2011). Meprin β was demonstrated to cleave APP at p3 position in a peptide derived in vitro assay (Bien et al., 2012), which would eventually lead to the release of Aβ3-40 peptides, containing an N-terminal pyroglumate modification. This cleavage site for meprin β, however, was so far only found in peptide cleavage assays and not in coexpression experiments with full length APP in cellular systems (Bien et al., 2012).

Several mutations within the APP sequence have been shown to have an impact on β-secretase cleavage by BACE-1. The recently described APP mutation A673T that has been shown to protect against AD as well as against cognitive decline in the elderly independent of AD was analyzed (Jonsson et al., 2012). The mutation is located at p2 of Aβ (Aβ-A/T) and has been shown to reduce BACE-1 mediated Aβ generation by 40% using synthetic peptides as substrates. Moreover, a significantly decreased Aβ production in human APP A673Toverexpressing primary neurons has been observed (Benilova et al., 2014; Maloney et al., 2014). Additionally, a decreased aggregation propensity of Aβ-A/T could be measured, which is showing the complexity of the protective effects of the substitution. As meprin β was shown to be involved in APP processing close to the BACE-1 cleavage site Schoenherr and colleagues investigated the effect of the APP A673T mutation on meprin β activity (Schönherr et al., 2016). The authors revealed a significant decrease of ∼70% in the Aβ2-40/1-40 ratio compared to wildtype APP sequence in meprin β transfected cells and in a peptide cleavage assay using the APP A673T constructs. The decreased cleavage of APP by meprin β in the presence of the A673T substitution can nicely be explained by the cleavage preference of meprin β revealed by proteomics (Becker-Pauly et al., 2011). Here, a preference of alanine over threonine in P1' position was observed. As the activity of meprin β on APP processing varies with mutations around the original BACE-1 cleavage site Schoenherr and colleagues investigated whether the Swedish mutation of APP (K670N/M671L; APPswe) may affect meprin β cleavage activity. Surprisingly, Aβ2-x variants were completely missing in cells overexpressing meprin β and APP bearing the Swedish double mutation K670N/M671L (APPswe) which is located in close vicinity of the β-secretase cleavage site (**Figure 1**). This clearly shows a significant influence of amino acid substitutions around the β-secretase cleavage site for meprin β mediated Aβ generation.

Although BACE-1 is clearly the most prominent enzyme responsible for the generation of Aβ1-40 and Aβ1-42 peptides from the APP wildtype or APPswe sequences, meprin β may be responsible for generating small amounts of N-terminal truncated Aβ2-40 and Aβ2-42 peptides. N-terminal truncated Aβ peptides are almost exclusively generated by meprin β from the complete APP wildtype sequences or from APP carrying familiar Alzheimer disease mutations at the γ–secretase cleavage site but bearing the wildtype sequence around the β-cleavage site.

## AD MOUSE MODELS

To analyze AD in an in vivo situation, different mouse models were already generated in the 1990's. However, these mouse models always show potential weaknesses which have to be considered before translating the results obtained from the mouse studies into the human situation. The major drawback is that cleavage of endogenous murine APP via the amyloidogenic pathway was never observed to lead to an AD-like phenotype. Hence, overexpression of different human APP forms in mice was and still is the most promising way to establish appropriate animal models. There are common models to study Aβ plaque pathology that all bear the APP Swedish mutation, such as 5xFAD mice, carrying mutations in the APP and PSEN1 genes [APP K670N/M671L (Swedish), APP I716V (Florida), APP V717I (London), PSEN1 M146L, and PSEN1 L286V; (Oakley et al., 2006)], J20 mice, carrying mutations only in the APP gene [K670N/M671L (Swedish) and the APP V717F (Indiana; Mucke et al., 2000)], or the 3xTg mice, carrying mutations in the APP, PSEN1, and the MAPT genes [K670N/M671L (Swedish), MAPT P301L, and PSEN1 M146V; (Oddo et al., 2003)]. These models all manifest an amyloid pathology although varying between animal models as well as differential learning and memory deficits. Thus, they appear to be appropriate models to mimic AD phenotypes at first sight. Notably, the human sequence of the Swedish familiar Alzheimer disease mutation (APPswe) is used in almost all AD animal models as it serves as a better substrate for BACE-1, thereby increasing production of total Aβ and specifically 1-X

cleft of meprin β as shown in (A). Positively charged amino acid residues important for the cleavage specificity are highlighted in blue. Part of the APP that builds the

Aβ peptides (Citron et al., 1992; Cai et al., 1993). However, in light of the result put forward by Schoenherr and colleagues Aβ2- 42 peptides which have been detected in brains of AD patients will not be generated in these mouse models. Therefore, it is likely that the actual effect of meprin β has been overlocked in many studies focusing on APP processing. This issue must be considered when analyzing the results from the ongoing clinical trials, using BACE-1 inhibitors for the treatment of AD patients.

Aβ peptide is displayed as surface model. Glycans in meprin β are depicted as stick models.

#### MEPRIN β AND APP BEYOND AD

As mentioned above in it has been shown that meprin β additionally cleaves APP apart from the Aβ sequence resulting in N-terminal APP fragments (NTF) (Jefferson et al., 2011). These fragments were also detected in human brain homogenates suggesting that this interaction not only occurs in overexpressing cell systems, but probably also under endogenous levels in the human brain. The in vivo relevance for this proteolytic event was further supported by analyzing brain lysates from meprin β deficient mice where this particular N-APP cleavage was abolished (Jefferson et al., 2011). Interestingly, Tessier-Lavigne and colleagues showed that an N-terminal APP fragment found in AD patients binds the death receptor 6 (DR6) thereby inducing neurodegeneration (Nikolaev et al., 2009). Thus, it was speculated whether meprin β might be the responsible protease in this regard. However, purification and characterization of the meprin β generated N-APP fragments showed neither negative nor positive influence on neuronal cell viability (Jefferson et al., 2011). Therefore, it is likely that APP cleavage by meprin β in the N-terminal region has rather protective function.

# PHYSIOLOGICAL FUNCTIONS OF MEPRIN β

Meprin β is strongly expressed in the intestinal epithelium and in kidney proximal tubular cells, and to minor levels in several other tissues, e.g., in skin, certain immune cells, and the brain (Broder et al., 2013). Besides many potential substrates analyzed in vitro only few in vivo functions of meprin β have been reported so far. In the intestine, where meprin β is found at the apical site of epithelial cells, the protease is responsible for the detachment of the mucus by cleaving mucin 2, an important step for proper barrier function (Schütte et al., 2014). Along the same line, meprin β cleaves type 1 pili of adherent-invasive E. coli, thereby preventing colonization of these bacteria in the ileal mucosa of Crohn's disease patients (Vazeille et al., 2011). Several other studies provide evidence for an important immunological function of meprin β (Banerjee and Bond, 2008; Bylander et al., 2008; Banerjee et al., 2009, 2011; Yura et al., 2009; Broder and Becker-Pauly, 2013; Zhang et al., 2015). As known for other members of the astacin family, namely BMP-1 (bone morphogenetic protein 1) and tolloids, meprin β is involved in the maturation of procollagens I and III (Kronenberg et al., 2010; Broder et al., 2013; Prox et al., 2015). Collagen, the most abundant protein in human body, is a crucial factor for the integrity of connective tissue, tendon, and bone. To prevent fibril assembly already in intracellular compartments, collagens contain C- and N-terminal prodomains that need to be removed proteolytically by extracellular proteases. Meprin β is such an enzyme, and Mep1b−/<sup>−</sup> mice show severe impairments of the connective tissue in skin characterized by reduced tensile strength and decreased collagen deposition (Broder et al., 2013). On the other hand, under pathological conditions, overexpression of meprin β is associated with fibrotic diseases, such as keloids of the skin (Kronenberg et al., 2010) and pulmonary hypertension (PH) (Biasin et al., 2014). PH is a severe fibrotic condition of the lung with very bad prognosis for the patients that die 2– 3 years after diagnosis. In genetic screens of lung tissues from patients and a mouse model of PH meprin β was found amongst the most up-regulated genes (Biasin et al., 2014). Here, AP-1 transcription factor complex was identified as an inducer of Mep1b mRNA expression. Whether meprin β is only involved in the progression of fibrosis by collagen maturation and deposition, or if the protease also contributes to the onset of the disease as a pro-inflammatory enzyme has to be further investigated.

# REGULATION OF MEPRIN β

As meprin β associated pathologies, such as fibrosis, cancer, and AD, are mostly based on increased expression and activity of the protease, information about the regulation of the enzyme is important.

# ACTIVATION

Meprin β is expressed as an inactive zymogen and requires proteolytic removal of its propeptide to gain full enzymatic activity. Several tryptic serine proteases have been identified as activators of latent meprin β, amongst them kallikreins (KLKs) 4, 5, and 8, as well as pancreatic trypsin (Ohler et al., 2010). The latter is supposed to be the physiological activator in the intestine, thereby contributing to the mucus-cleaving activity of meprin β (Schütte et al., 2014), whereas KLKs may rather be important in skin and mesenchymal tissues (Ohler et al., 2010). Based on the crystal structure of the ectodomain of human meprin β it became evident that the activation site at amino acid position Arg61 is in very close proximity to the cell surface (Arolas et al., 2012). Therefore, it was doubtful whether the previously described soluble tryptic activators, which were identified in in vitro assays using recombinant soluble promeprin β, are capable of activating the membrane bound meprin β. Indeed, not even trypsin was able to cleave off the propeptide of full length meprin β, which led to the assumption that possible candidates are most likely membrane bound serine proteases. In this regard, matriptase-2 (MT-2), a type 2 transmembrane protein, was found to fully activate meprin β at the cell surface (Jäckle et al., 2015). Consequently, MT-2 mediated activation of meprin β resulted in increased APP shedding and subsequently decreased sAPPα levels. If this proteolytic interaction may have impact on neurodegenerative disorders has to be shown. Surprisingly, however, in a different study MT-2 was found to directly cleave neuronal APP695, but was effectively inhibited by the Kunitz protease inhibitor (KPI) domain present in other APP isoforms (APP751 and APP770) from the periphery (Beckmann et al., 2016). Of note, the additional domains in APP751 (KPI) and APP770 (KPI/OX2) do not lead to altered proteolytic processing by meprin β (Jefferson et al., 2011). This demonstrates how complex the proteolytic processing of APP can be and how important it is to elucidate the time-dependent and site-specific cleavage events with regard to the different proteases, such as ADAM10, BACE-1, meprin β, or MT-2.

# INHIBITION

The tissue inhibitors of metalloproteinases (TIMPs) are effective regulators of the catalytic activity of matrix metalloproteases (MMPs) and ADAMs (Yamamoto et al., 2015). TIMPs, however, do not inhibit meprin β, and so far only one rather unspecific endogenous inhibitor was identified, namely fetuin-A (Kruse et al., 2004; Hedrich et al., 2010). Interestingly, calcium was found to inhibit the proteolytic activity of meprin β by binding to a cluster of negatively charged amino acids in close proximity to the active site, thereby inducing conformational changes (Arnold et al., 2015). However, the inhibition constant of calcium for meprin β is about 11 mM, which resembles the concentration in the endoplasmic reticulum and not at the cell surface. The amino acid residues forming the calcium binding site in meprin β contribute to correct folding of the protease. Mutations within the calcium binding site resulted in protein that stacks to the ER and is not properly secreted (Arnold et al., 2015). The calcium concentration needed for the inhibition of meprin β is rather not relevant for extracellular inhibition, at least under physiological conditions. Thus, regulation of meprin β's activity must occur on a different level. One possibility is the ectodomain shedding of meprin β by ADAM10 or ADAM17 (Jefferson et al., 2013). Very importantly, only membrane bound meprin β is capable of generating aggregation prone Aβ2-x peptides and not the shed solubilized protease (Bien et al., 2012). Therefore, ADAM10 does not only prevent Aβ release by cleaving APP at the α-secretase site, but additionally by the shedding of meprin β and thereby preventing its activity toward the β-secretase site.

#### LOCALIZATION

As mentioned above, shedding of APP by meprin β occurs predominantly at the cell surface and thus competes with ADAM10 for the substrate (Schönherr et al., 2016). Recent studies demonstrated that ADAM10 localization and maturation is influenced by tetraspanins (TSPANs), building microdomains of protein clusters at the cell surface (Prox et al., 2012). In a yeasttwo-hybrid approach TSPAN8 was identified as an interaction partner of meprin β, which was further proven by split-RFP and luciferase complementation assays (Schmidt et al., 2016). It was further demonstrated that APP together with meprin β is located in TSPAN8 enriched microdomains. However, overexpression of

#### REFERENCES


TSPAN8 had no obvious influence on meprin β activity and APP cleavage. Nevertheless, orchestration of proteases and substrates at the cell surface by regulatory factors has to be further studied to fully understand the complex proteolytic processing of APP by different enzymes.

Concluding, the protease meprin β appears as an important candidate for further studies on APP processing and Aβ generation and may have a contributing role to the physiological and pathophysiological function of APP itself.

### AUTHOR CONTRIBUTIONS

All authors (CP, CB) contributed to the concept and drafting the work. The authors ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

# FUNDING

This work was supported by the Deutsche Forschungsgemeinschaft, grant BE 4086/2-2 (to CB), and grant PI 379 6-2 (to CP).


cells. Proc. Natl. Acad. Sci. U.S.A. 90, 2092–2096. doi: 10.1073/pnas.90.5. 2092


proteolytic processing of mutant and wild-type beta-amyloid precursor protein. J. Biol. Chem. 270, 6186–6192. doi: 10.1074/jbc.270.11.6186


are strikingly different from APP and APLP2. J. Cell Sci. 122, 368–377. doi: 10.1242/jcs.034058


A673T allele of amyloid precursor protein. J. Biol. Chem. 289, 30990–31000. doi: 10.1074/jbc.M114.589069


disease therapeutics. Hum. Mol. Genet. 10, 1317–1324. doi: 10.1093/hmg/10.12. 1317


to monitor kidney response to high dose morpholino oligonucleotides in the mdx mouse. Toxicol. Rep. 2, 838–849. doi: 10.1016/j.toxrep.2015. 05.008


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Becker-Pauly and Pietrzik. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Regulation of Alpha-Secretase ADAM10 In vitro and In vivo: Genetic, Epigenetic, and Protein-Based Mechanisms

Kristina Endres <sup>1</sup> \* and Thomas Deller <sup>2</sup>

*<sup>1</sup> Clinic of Psychiatry and Psychotherapy, University Medical Center Johannes Gutenberg-University Mainz, Mainz, Germany, 2 Institute of Clinical Neuroanatomy, Neuroscience Center, Goethe-University, Frankfurt/Main, Germany*

ADAM10 (A Disintegrin and Metalloproteinase 10) has been identified as the major physiological alpha-secretase in neurons, responsible for cleaving APP in a non-amyloidogenic manner. This cleavage results in the production of a neuroprotective APP-derived fragment, APPs-alpha, and an attenuated production of neurotoxic A-beta peptides. An increase in ADAM10 activity shifts the balance of APP processing toward APPs-alpha and protects the brain from amyloid deposition and disease. Thus, increasing ADAM10 activity has been proposed an attractive target for the treatment of neurodegenerative diseases and it appears to be timely to investigate the physiological mechanisms regulating ADAM10 expression. Therefore, in this article, we will (1) review reports on the physiological regulation of ADAM10 at the transcriptional level, by epigenetic factors, miRNAs and/or protein interactions, (2) describe conditions, which change ADAM10 expression *in vitro* and *in vivo*, (3) report how neuronal ADAM10 expression may be regulated in humans, and (4) discuss how this knowledge on the physiological and pathophysiological regulation of ADAM10 may help to preserve or restore brain function.

#### Edited by:

*Ashok Hegde, Georgia College and State University, USA*

#### Reviewed by:

*Eckart D. Gundelfinger, Leibniz Institute for Neurobiology, Germany Baojin Ding, University of Texas Southwestern Medical Center, USA*

#### \*Correspondence:

*Kristina Endres kristina.endres@unimedizin-mainz.de*

> Received: *21 December 2016* Accepted: *20 February 2017* Published: *17 March 2017*

#### Citation:

*Endres K and Deller T (2017) Regulation of Alpha-Secretase ADAM10 In vitro and In vivo: Genetic, Epigenetic, and Protein-Based Mechanisms. Front. Mol. Neurosci. 10:56. doi: 10.3389/fnmol.2017.00056* Keywords: ADAM10, aging, alpha-secretase, Alzheimer's disease, mouse models, promoter, transcription factors, spine

#### ADAM10 - PORTRAIT OF A BIOLOGICALLY VERSATILE PROTEASE

#### Introduction

ADAM10 (A Disintegrin and Metalloproteinase 10) was identified in vitro as a key proteinase in the processing of the amyloid precursor protein (APP) more than 15 years ago (Lammich et al., 1999). The zinc-dependent proteinase cleaves APP within the A-beta sequence, thus preventing the production of this peptide. Furthermore, APP-cleavage by ADAM10 liberates APPs-alpha, which has neuroprotective properties and is involved in the regulation of synaptic plasticity and learning and memory (reviewed in Kögel et al., 2012). In line with these findings, overexpression of ADAM10 in mice revealed elevated APPs-alpha levels and demonstrated a robust in vivo activity of ADAM10 (Postina et al., 2004). Overexpression of ADAM10 was also effective in animal mouse models of Alzheimer's disease (AD) and reduced plaque load as well as deficits in learning and memory (Postina et al., 2004; Schmitt et al., 2006). Subsequent investigations of RNAi-mediated knock-down of the enzyme in primary cortical neurons (Kuhn et al., 2010) as well as conditional knock-down in mice (Jorissen et al., 2010) consolidated the enzymes' role in APP processing in vivo. Collectively, these data point to ADAM10 as being the most important physiological alpha-secretase involved in the processing of APP in neurons.

The central role of ADAM10 in APP processing has made ADAM10 an interesting target for AD therapy. It has been proposed (e.g., Fahrenholz and Postina, 2006; Vincent and Govitrapong, 2011) that similar to the situation in intact animals (Postina et al., 2004) an increase in ADAM10 could result in decreased A-beta load and improved learning and memory in AD patients. For this approach to be effective and safe, however, the cell biology of ADAM10 and its cellular functions need to be better understood. ADAM10 is a versatile protease which cleaves not only APP but also several other proteins (see paragraph 2). Therapeutic strategies for AD focusing on ADAM10 as a target have to keep these additional substrates in mind. In the present review we will summarize the extant literature on ADAM10 and focus on what is known about its regulation in vitro and in vivo. Understanding the regulation of this enzyme may be a necessary step toward understanding its usefulness in therapeutic contexts.

#### Domain Structure, Cellular Synthesis, and Maturation of ADAM10

ADAM10 is a catalytically active member of the ADAM family of proteinases. The ADAMs are grouped together as a family because they share structural features with snake venom disintegrin proteases (Wolfsberg et al., 1995a,b). ADAM10 is co-translationally synthesized via the rough ER, matures and is transported via the Golgi apparatus. Maturation includes removal of the prodomain (**Figure 1**: 1), which keeps the enzyme in an inactive state. A cleavage site for proprotein convertases such as PC7 (Anders et al., 2001) is mandatory for production of the catalytically active enzyme as shown by analyzing mutated ADAM10. However, the prodomain has not a mere inhibitory function but is also needed as an intramolecular chaperon for correct folding (Anders et al., 2001). This is reflected by the fact that a large proportion of ADAM10 has been found to be localized in the Golgi apparatus in AR breast carcinoma cell line by confocal microscopy (Gutwein et al., 2003). The mature form of ADAM10 of about 68 kDa was found in the Golgi compartment as well as in the ER/plasma membraneenriched fraction of postnuclear supernatant and at least cleavage of another substrate of ADAM10—the L1 adhesion molecule seems to occur in both. Recent investigations suggested by administering the inhibitor RVKR for up to 8 h before measuring shedding activity that cleavage by proprotein convertases might be dispensable for rapid stimulation of ADAM10 (Maretzky et al., 2015). However, as the half-life time of ADAM10 is rather long

(>72 h; Mezyk-Kopec et al., 2009), this result may need to be interpreted with some caution.

The catalytic domain of ADAM10 (**Figure 1**: 2) contains the characteristic zinc-binding consensus motif (HEXGHXXGXXHD) of active members of the proteinase family. A point mutation within this motif (E384A) results in a dominant negative acting protein and a decreased APPs-alpha secretion, as could be shown shown in HEK cells and mice (Fahrenholz et al., 2000; Postina et al., 2004).

The catalytic and the proximal disintegrin domain contain high-mannose as well as complex-type N-glycan attachment sites (Escrevente et al., 2008). The disintegrin domain of ADAM10 (**Figure 1**: 3) does not appear to be essential for ADAM10 protease activity in cell culture experiments (Fahrenholz et al., 2000). Rather, the short intracellular C-terminus seems to play an important role: Epidermal growth factor (EGF) cleavage has been reported to be partially impaired in ADAM10−/− cells overexpressing a cytoplasmic domain deletion mutant of the proteinase (Horiuchi et al., 2007). However, the cytoplasmic domain of ADAM10 has also been reported to negatively influence constitutive shedding through an ER retention motif: an ADAM101cyto mutant displayed increased catalytic activity compared to ADAM10 Wt with regard to Betacellulin cleavage (Maretzky et al., 2015). The cytoplasmic domain of ADAM10 contains several binding sites that may be involved in regulatory events, such as an IQ consensus binding site for calmodulin (Horiuchi et al., 2007) and two proline-rich putative Src homology 3 (SH3) binding domains. The juxtamembrane binding site affects basolateral localization of ADAM10 in epithelial cells (Wild-Bode et al., 2006), while in neurons the SH3 binding domains direct ADAM10 to the postsynaptic membrane (Marcello et al., 2007). Using a phage library analysis comprising 305 human SH3 domains, 38 candidate binding proteins for the ADAM10 C-terminus were identified, including endophilin-A2, Lck, or ZDHHC6 (Ebsen et al., 2014). Although the biological relevance of many of these putative ADAM10 binding partners needs to be determined, this finding suggests that regulatory interactions at the C-terminus could play a major role regarding the cellular localization as well as the activity of the proteinase.

## Developmental and Adult Expression of ADAM10 in Mouse and Human Brain

ADAM10 is expressed in various tissues in mice (Marcinkiewicz and Seidah, 2000). Its presence in the developing as well as in the adult CNS underscores its importance for normal brain development and function. Since ADAM10 can only process a putative substrate if both, protease and substrate are expressed at the same time and in the same cellular compartments, it is important to know the temporospatial pattern of ADAM10 expression in the brain. This pattern can then be compared to corresponding data of putative substrates or binding partners.

The distribution of ADAM10 was studied in mouse cerebral cortex from embryonic day (E) 14.5 to postnatal day (P) 1 using in situ hybridization analysis. This revealed ADAM10 expression within the ventricular zone and the cortical plate from E17.5 to P1 (Ma et al., 2013; see also **Figure 2**). These data on ADAM10 mRNA were corroborated by immunofluorescence analyses which detected ADAM10 protein in developing cerebral cortex from E14.5 to E18.5 (Ma et al., 2013).

ADAM10 plays an essential role during development. Animals with a conventional ADAM10 knock-out die on E9.5 (Hartmann et al., 2002), which underlines the general importance of this protease. More recently generated conditional Nestin-Cre-ADAM10 knock-out mice with a cell-specific inactivation of ADAM10 in neural progenitor cells (NPCs), NPC-derived neurons and glial cells prolonged the life span of the mice to

FIGURE 2 | Distribution of ADAM10 mRNA in the murine brain. Sagittal section of C57Bl6/J mouse brain (male) at E18.5 (A1,A2; Image credit: Allen Institute; http://developingmouse.brain-map.org/experiment/show/100055949, ©2016. Allen Institute for Brain Science) and P56 (B1,B2; Image credit: Allen Institute; http:// developingmouse.brain-map.org/experiment/show/69514738, ©2016. Allen Institute for Brain Science C1,C2: magnification of hippocampal area of the adult brain). ADAM10 mRNA expression is revealed by *in situ* hybridization [A1–C1, upper row ISH; A2–C2, lower row expression energy (cells with highest probability of gene expression)]. CA1-3, *Cornu Ammonis* regions; Cb, cerebellum; Ctx, cerebral cortex; DG, *Dentate Gyrus*; H, hippocampus; ob, olfactory bulb; SC, Superior Colliculus; Th, thalamus; vmh, ventral mid-/hindbrain

a perinatal time point (Jorissen et al., 2010). These mutants displayed a disrupted neocortex and a severe reduction of the ganglionic eminence. Knock-out of ADAM10 in the postnatal CNS using a CaMKII-alpha-Cre driver finally allowed investigation of adult mice (Prox et al., 2013). This conditional mutant did not show gross morphological abnormalities but exhibited synaptic dysfunction, increased early perinatal lethality, altered behavior, and epileptic seizures. Similar results were reported by another group which independently established an adult ADAM10 knock-out model (Zhuang et al., 2015). Taken together, these studies indicate that ADAM10-deficiency results in major developmental phenotypes. Lack of the protease at later stages is compatible with life but results in a number of dysfunctions.

The cellular expression pattern of ADAM10 was also investigated in some of these studies. Interestingly, ADAM10 protein expression partially correlated with both, S100β and Tuj1 expression (E16.5 to P1), which indicates a relevance of ADAM10 for glial as well as neuronal cell function during late embryonic cerebral cortex development stages (Ma et al., 2013). In the developing brain of chicken, ADAM10 shows a weak but widespread expression at E12 in most gray matter areas (Lin et al., 2008). Expression intensity decreased from E14 to E19, with the exception of the telencephalon and the cerebellum. Some ADAM10-positive non-neuronal cells may be oligodendrocytes, since they were shown to co-express galactocerebroside, which is a marker for oligodendrocytes at late stages of chicken embryogenesis (Lin et al., 2008).

ADAM10 expression has also been studied in developing human brain: Bernstein and colleagues compared the amount of ADAM10 in temporal cortex of stillborn children with those of normal aged adults and found a general increase (Bernstein et al., 2003). In a follow-up study, a weak expression of ADAM10 in the cytoplasm of pyramidal as well as nonpyramidal neurons was confirmed in pre- and perinatal human brains (Bernstein et al., 2009). In contrast to these findings, analysis of total human fetal brain RNA obtained from a commercial source (Clontech, Kaczur et al., 2007) failed to detect a prominent ADAM10 expression using microarray analysis. Differences in the stage of fetal development (which has not been reported) or technical issues, e.g., detection of ADAM10 in total mRNA preparations, may explain these discrepancies. Analysis of transcripts from a human fetal brain library, however, revealed two types of ADAM10 cDNAs: one encoding a 748 amino acid protein [designated Kuzbanian (Kuz)L] and a second one (KuzS, encoding a 568 amino acid protein), which lacks the cysteine-rich, transmembrane, and cytoplasmic domain (Yavari et al., 1998). Fetal human brain expressed substantially more of the short than of the long variant, while fetal lung predominantly contained the longer variant. In adult human brain tissue, Northern blot demonstrated the persistence of both forms in different amounts (Yavari et al., 1998). The presence of the short transcript appears to be unique to humans as the transcript could not be detected in adult and embryonic tissue of mice. Whether it is functionally relevant, i.e., whether it is translated into a biologically active protein is unclear.

The expression of ADAM10 in the adult brain has been studied in rodents and humans. By using northern blot technique it could be shown that adult human amygdala, caudate nucleus and corpus callosum contain relatively high amounts of ADAM10 transcripts whereas mRNA levels in the subthalamic nucleus and the thalamus were comparably low (Yavari et al., 1998). In the adult rodent brain, ADAM10 mRNA was reported to be moderately expressed throughout the whole brain, including the olfactory bulb, the hippocampus or the subthalamic region (mouse and rat: Kärkkäinen et al., 2000, see also **Figure 2**: P56). Semiquantitative evaluation of ADAM10 mRNA using ISH analysis revealed only the Pontine nuclei as a brain structure not expressing the protease (Kärkkäinen et al., 2000). These findings have been confirmed by a recent investigation that also described positive ISH-stainings for neurons of the cerebral cortex, hippocampus, thalamus, and cerebellar granular cells in the CNS of adult mice (Guo et al., 2016).

# ADAM10—PHYSIOLOGICAL SUBSTRATES AND FUNCTIONS

#### Physiological Substrates of ADAM10

ADAM10 is probably best known for its ability to process APP. ADAM10 cleaves APP at the alpha-secretase cleavage site and in vitro as well as in vivo studies have implicated ADAM10 as the biologically most relevant neuronal alpha-secretase (e.g., Postina et al., 2004; Jorissen et al., 2010; Kuhn et al., 2010). Of note, the regional and cellular overlap of ADAM10 and APP, which is necessary for ADAM10 to process APP in tissues, is age-dependent: at early developmental stages the mRNA distributions of ADAM10 and APP are not fully congruent but with aging the overlap increases (Marcinkiewicz and Seidah, 2000). This finding—but also the wealth of data on other substrates of ADAM10 (see below)—suggests that ADAM10 substrates may change: During development and in the young brain ADAM10 may preferentially cleave substrates other than APP and the role of ADAM10 as alpha-secretase of APP may emerge with aging.

Presently, a rather large number of ADAM10 substrates have been identified in different experimental settings (e.g., reviewed for proteomic approaches in Müller et al., 2016). Of notice, ADAM10 substrates belong to type I as well as type II transmembrane but also Glycosylphosphatidylinisotol (GPI)-anchored proteins, indicating a considerable flexibility of the protease with regard to substrate recognition. Consensus cleavage motifs for proteases are commonly deduced from the amino acids surrounding the naturally occurring cleavage sites within protein substrates. This approach failed in the case of ADAM10 because it lacks a well-defined consensus sequence: for ADAM10 leucine was found to be preferred (and tyrosine accepted) in the P1' position (immediately downstream of the cleavage site) in an investigation using oriented peptide mixture libraries which gives evidence of a shallow or deep S1' site (John et al., 2004). ADAM10's preference for larger residues at P1' has been confirmed but acceptance of aromatic amino acids and even glutamine were also reported (Caescu et al., 2009). This tolerance for aromatic residues in P1' may be the most relevant difference in cleavage site specificities between ADAM10 and its close relative ADAM17 (Tucher et al., 2014). In the early investigation tyrosine was found to be favored at P1 (immediately upstream of the cleavage site; John et al., 2004). In contrast to this, selectivity for small residues such as alanine at the P1 was described by Caescu et al. (2009) and specificities for proline and basic residues were recently reported (Tucher et al., 2014). In sum, these reports show the methodological limitations and uncertainties involved in pinpointing cleavage site specificities from linear unmodified peptide libraries. In addition, the activity state of the cell may also influence shedding capacity, as is the case for NG2 (Sakry et al., 2014) as well as the synaptic marker neuroligin 1 (Suzuki et al., 2012), further complicating work in this direction.

At present, a wide variety of substrates has been identified for ADAM10 and some of them have been confirmed not only in primary culture but also in vivo. In line with its ubiquitous expression, ADAM10 substrates are linked to a number of biological systems and physiological as well as pathological functions (c.f. chapter 4), including the immune and nervous system but also cancerogenesis (e.g., Vincent and Checler, 2012). In their review Pruessmeyer and Ludwig reported on the "good, the bad and the ugly" ADAM10 substrates (Pruessmeyer and Ludwig, 2009). Since then, a number of additional ADAM10 substrates were identified and have resulted in a more complete, albeit even more complex picture of ADAM10 (**Table 1**).

#### TABLE 1 | ADAM10 substrates identified within the last years.


*Putative ADAM10 substrates (ordered from newest to oldest publication date) identified since 2009 or not included in Pruessmeyer and Ludwig (2009) are listed. (PubMed search 29-11-2016: "ADAM10 and substrate" or "ADAM10 and proteolysis").*

# Functions of ADAM10 at the Synapse and in Non-neuronal Cells

ADAM10 processes other proteins and thus, influences the functions of its substrates by in-/activating them or by liberating biologically active fragments. Thereby, the biological effects of ADAM10 activity are tightly linked to the functions of the substrates and their cleavage products. Because of the large number of ADAM10 substrates identified to date, we focus in this review on those which are known to have an important impact on brain function and which are likely to co-localize with ADAM10 at the synapse or in glial cells.

The earliest study on the distribution of ADAM10 at synapses was based on immunocytochemistry and suggested that ADAM10 co-localizes with the postsynaptic scaffold protein Synapse-associated protein 97 (SAP-97) but not with the presynaptic vesicle protein synaptophysin (Marcello et al., 2007). However, a more recent study using the sensitive proximity ligation assay reported proximity of the enzyme with synaptophysin in mouse primary hippocampal neurons (Lundgren et al., 2015). This suggests that ADAM10 can be present in both parts of a synapse. One example where this could be functionally relevant is the neurexin-neuroligin-interaction: neurexins and neuroligins are cell-adhesion molecules which form transsynaptic complexes (e.g., Tsetsenis et al., 2014). They appear to be important for normal synapse specification and function (Jedlicka et al., 2011, 2015). For the postsynaptic protein Neuroligin 1, ADAM10 has been found to act as the major sheddase, as could be shown by pharmacological and genetic means in primary rat cortical neurons (Suzuki et al., 2012). NMDA receptor activation as well as prolonged epileptic seizure condition increased shedding, suggesting a role for neuronal activity in this context. Interestingly, shedding of Neuroligin 1 could be induced by soluble neurexin 1α or β derived from overexpressing HEK293 cells (Suzuki et al., 2012), indicating that ligand binding at the cell surface also regulates Neuroligin 1 shedding. Similar observations have been made for the Notch-Delta complex where Notch1 cleavage by ADAM10 is induced by Delta binding (e.g., reviewed in Van Tetering and Vooijs, 2011). Intriguingly, Notch 1 as well as its ligands - Delta or Jagged have been found to be cleaved by ADAM10 (for example: Pan and Rubin, 1997; Lavoie and Selkoe, 2003). A recent publication regarding systemic characterization of ADAM10 substrates from neurons highlighted that ADAM10 is also in principle capable of shedding the Neuroligin ligands Neurexins 2 and 3, although deletion of the proteinase resulted only in a comparably mild reduction of the shedding (Kuhn et al., 2016). If this role for ADAM10 in the cleavage of major anchoring proteins can be verified in vivo and in human brain, interfering with ADAM10 activity may indeed be a powerful tool to influence synaptic structure and function.

ADAM10 has also been found to process substrates of non-neuronal cells. Since neurons and glial cells are highly interdependent and jointly regulate synapse functions, ADAM10 may also influence network activities through glial cells. For example, the marker transmembrane proteoglycan nerveglia antigen 2 (NG2), commonly found on the so-called "NG2-glial cells" (Eugenín-Von Bernhardi and Dimou, 2016), has also been identified as a substrate of ADAM10 (Sakry et al., 2014). Similar to what has been reported for Neuroligin 1, shedding of NG2 is also regulated by neuronal activity. Moreover, neurons from NG2-knock-out mice exhibited diminished amplitudes of AMPA receptor-currents which could be rescued by application of the partial NG2 ectodomain (Sakry et al., 2014). This suggests that an NG2-cell derived ectodomain produced by ADAM10 processing regulates synaptic activity and plays a role in neuron-glia communication.

Another non-neuronal substrate of ADAM10 with implications for glial and neuronal function is the microglial surface protein triggering receptor expressed on myeloid cells 2 (TREM2). TREM2 has been suggested to play a role in phagocytosis and has been recently recognized as a genetic risk-factor for AD (Frank et al., 2008; Colonna and Wang, 2016). Using HEK293 Flp-In cells and enzymatic inhibitors, the release of soluble TREM2 ectodomain was demonstrated to depend on ADAM10 but not on ADAM17 or beta-secretase (beta-site amyloid precursor protein cleaving enzyme 1; BACE1) (Kleinberger et al., 2014). Reduction of cell surface TREM2 decreases the ability of microglia to phagocytose and remove cellular debris or apoptotic neurons (Kleinberger et al., 2014). Of note, TREM2 ligands were identified on Neuro2A cells and on cultured cortical and dopamine neurons (Hsieh et al., 2009), suggesting an impact of a non-neuronal shedding event on neurons.

Finally, there is growing evidence for a role of exosomes in neuron-glia communication (Frühbeis et al., 2013). In this regard it is of interest that functionally active ADAM10 has been found in exosomes from ovarian carcinoma cells where it contributes to L1 and CD44 cleavage (Stoeck et al., 2006) and in exosomes of primed B-cells (Padro et al., 2013). Whether microglial or neuronal cells also use exosomes to deliver ADAM10 or shedded substrates among themselves is currently unknown.

# REGULATION OF ADAM10

ADAM10 is a multifunctional protease active throughout the life of an organism and its regulation is controlled at transcriptional, epigenetic, translational and post-translational levels. These different levels of regulation allow a cell to adapt ADAM10 levels rapidly to functional perturbations as well as to slower changes induced by aging and/or maturation.

# Transcriptional Regulation of ADAM10

The human ADAM10 gene is localized on chromosome 15, whereas its murine homolog is found on chromosome 9 (Yamazaki et al., 1997a,b). Both genes are comprised of about 160 kb with high sequence preservation within the first 500 bp upstream of the translation initiation site (Prinzen et al., 2005). The human core promoter is positioned at −508 to −300 bp and contains no TATA box but several functional binding sites for common transcription factors such as Sp1 and USF (Prinzen et al., 2005). SNPs in the human promoter region at position −279 and −630 indicated no association with AD (Prinzen et al., 2005), whereas a SNP located at −644 was correlated with CSF APPs-alpha levels (Bekris et al., 2011). The 5′ UTR of the human gene was located 444 bp upstream of the start codon (Lammich et al., 2010), the 3′ UTR up to 1254 bp downstream of the stop codon (Augustin et al., 2012).

Even before the promoter of human ADAM10 was described, several pathways regulating the enzyme's expression had been identified: for example, in the prostate cancer cell line LNCaP insulin-like growth factor I combined with 5 alphadihydrotestosterone increased mature and immature ADAM10 protein amounts (McCulloch et al., 2004). Similarly, EGF led to the up-regulation of ADAM10 mRNA and protein in those cells. In addition, the Tcf/Lef-family of transcription factors which is known to interact with beta-catenin (Wisniewska, 2013) also seems to be involved: Wang et al. demonstrated in transgenic AD mice the induction of Wnt signaling by huperzine A. This was accompanied by elevated beta-catenin levels and increased ADAM10 protein levels (Wang et al., 2011). These findings were corroborated by the observation that NMDA receptor activation in primary neurons similarly increased ADAM10 via Wnt/MAPK signaling (Wan et al., 2012).

Using different cell systems Paired Box Genes (PAX) were similarly identified as putative ADAM10 regulators. In melanoma cells chromatin immunoprecipitation assay and overexpression as wells as siRNA-mediated knock-down gave evidence that PAX2 can regulate ADAM10 expression (Lee et al., 2011). Downregulation of PAX2 via siRNA in A498 (renal carcinoma), EAhy (endothelial), T98G (glioblastoma), and SKOV3ip (ovarian carcinoma) cells revealed a nearly total loss of ADAM10 protein as demonstrated by Western blot analysis (Doberstein et al., 2011). Therefore, PAX2 seems to play an important role in ADAM10 expression control—at least in cancer cells. Interestingly, the related PAX4 has been shown to regulate ADAM10 post-transcriptionally (see paragraph "Regulation of ADAM10 at the translational level").

Another signaling pathway that increases ADAM10 amount within the cell via gene regulation requires melatonin. It has been reported that melatonin elevates ADAM10 level in HEK293 and neuronal SH-SY5Y cells via G protein-coupled receptorinduced PKC/Erk activation (Panmanee et al., 2015; Shukla et al., 2015). This effect seems to depend on human ADAM10 promoter region −1193 to −555 as a respective deletion construct failed to respond in a reporter gene assay (Shukla et al., 2015). The authors of the report discuss that the binding sites of cAMP response element-binding protein (CREB) and octamerbinding transcription factor 1 (Oct-1) which were described earlier (Prinzen et al., 2005) might contribute to the regulation or that a yet unidentified Hypoxia-inducible factor 1 (HIF-1) binding site might be responsible. The regulation of ADAM10 via the sleep hormone melatonin seems highly interesting as sleep disturbances are considered characteristic symptoms of AD (for example Sung et al., 2017).

Agonists specific for Peroxisome Proliferator-Activated Receptor alpha (PPARalpha) but not PPARbeta, delta, or gamma elevated ADAM10 protein amount in primary murine hippocampal neurons (Corbett et al., 2015). Seven PPAR responsible elements were identified by in silico analysis and the specific agonist GEM led to enrichment of PPARalpha and its heterodimer Retinoid X Receptor alpha (RXRalpha) binding partner at two direct repeat 1 PPAR responsive elements (PPRE) located in the ADAM10 promoter in wild type, but not in PPARalpha knock-out hippocampal neurons. 9-cis retinoic acid failed to synergistically increase ADAM10 amount in this context, therefore a non-permissive PPARalpha/RXRalpha heterodimer seems to regulate ADAM10 expression similar to the RARalpha/beta/RXR dimer from earlier investigations (Tippmann et al., 2009). PPARalpha is known to be involved in fatty acid metabolism. In this regard, it is of interest that lowering the cholesterol amount of cells increased ADAM10's catalytic but not transcriptional activity (Kojro et al., 2010) and that various fatty acids and lipids such as Docosahexaenoic acid (DHA) interfere with the balance of APP processing (e.g., Eckert et al., 2011; Grimm et al., 2016). Additionally, Sex-determining region Y-box 2 (Sox2), a major factor of adult tissue homeostasis and regeneration control, was recently identified to upregulate ADAM10 expression in HEK293 cells using overexpression experiments (Sarlak et al., 2016).

The retinoic acid receptor (RAR) family is particularly interesting with regard to ADAM10 regulation because of its therapeutic potential. Both, RAR alpha and beta are capable of inducing human ADAM10 promoter activity (Tippmann et al., 2009). Moreover, the commercially available drug acitretin which intracellularly liberates retinoic acid (Ortiz et al., 2013), shifts APP processing in AD model mice toward the alpha-secretase cleavage pathway (Tippmann et al., 2009). The neuroprotective property of RARalpha agonists has been shown in cortical cultures, an AD mouse model (Tg2576 mice) (Jarvis et al., 2010), as well as in hippocampal tissue of aged SAMP8 mice (Kitaoka et al., 2013). Cilostazol-stimulated N2A cells with overexpression of human mutated APP also displayed ADAM10 elevation which was significantly attenuated by a RARbeta inhibitor and RARbeta-gene silencing (Lee et al., 2014). The effect of cilostazol on ADAM10 expression could be antagonized by sirtinol and by Sirtuin 1 (SIRT1)-gene silencing, suggesting that RARbeta and this class of deacetlyase together act on the ADAM10 promoter.

For a systematic approach on transcription factors relevant to ADAM10 regulation, we performed a screening approach (Reinhardt et al., 2014). **Figure 3** sums up transcription factors that showed a significant influence on ADAM10 expression in these investigations. One has to consider that the screening approach was performed in human neuronal SH-SY5Y cells and only included single expression plasmids for 704 human transcription factors. Therefore, accessory proteins for single factors might not have been present in the cell line or combinations of transcription factors might be needed for full activation. However, we identified 11% transcription factors with a comparably strong influence on promoter activity of ADAM10 with nine factors inhibiting and 74 factors increasing transcriptional activity (**Figure 3**). Starting from this screening we were able to further characterize regulation of ADAM10 via one of the strongest inducers—X-Box binding protein 1 (XBP-1, Calfon et al., 2002). The active transcription factor is built upon ER stress sensor Inositol requiring enzyme 1 alpha (IRE1 alpha) activation and leads to increase in ADAM10 mRNA as well as protein and subsequent release of the APP cleavage product APPs-alpha (Reinhardt et al., 2014). Interestingly, we also found

the amount of XBP1-mRNA to be decreased in Alzheimer model mice at higher age and also in Alzheimer's disease patients.

#### Epigenetic Regulation of ADAM10

Currently, little is known about the epigenetic regulation of ADAM10. The 5′ -untranslated region of the human ADAM10 gene contains a large GC-rich domain at −700 to +200 bp. The GC content of the first 600 bp upstream of the ATG of the human ADAM10 gene is 67% and nine CpG islands have been predicted (Prinzen et al., 2005). This abundance of CpGs suggests that cytosine methylation could play a role in regulating the proteinases' expression. SIRT1 an evolutionarily conserved NAD+-dependent deacetylase pivotal for metabolic control has been identified to increase ADAM10 expression (Lee et al., 2014). SIRT1 is involved in histone deacetylation and methylation, promoter CpG island methylation, and inactivation of non-histone transcription factors (Zhang and Kraus, 2010). Conceivably, SIRT1 is also involved in deacetylation of RAR or in chromatin modifications upon recruitment by the receptor but currently this has not yet been demonstrated. Investigations into these regulatory mechanisms are non-trivial and complicated by the fact that SIRT1 also acts on the cellular retinoid binding protein II (CRABPII) and also has a more general effect on RA signaling (Tang et al., 2014).

In transgenic AD model mice (5 × FAD) a significant increase in global DNA methylation, measured by 5-methyl cytosine, has been reported and additional changes in e.g., demethylase Dnmt3b or enzymes of histon acetylation/ deacetylation such as Hdac2, Jarid1a, or G9a (Griñán-Ferré et al., 2016). Surprisingly, no changes of ADAM10 expression were observed when using whole brain mRNA preparations. Although ADAM10 was found within the top CpG sites of an epigenomic analysis of psychiatric tic-diseases using peripheral blood samples (cg00785856, Zilhão et al., 2015), the methylation site did not reach significance at the genome-wide threshold.

Finally, as melatonin seems to be able to increase the level of deacetylase in young and aged primary neurons (Tajes et al., 2009), the observed induction of ADAM10 by melatonin (Panmanee et al., 2015; Shukla et al., 2015) might also rely on deacetylase activation.

# Regulation of ADAM10 at the Translational Level

Besides regulation on the transcriptional/epigenetic level, translational modifiers can regulate the amount and availability of ADAM10: RNA structure, RNA-binding proteins (RBPs), and miRNAs have been reported to play a role.

The working group of Christian Haass explored a suppression of ADAM10 expression by its 5′UTR (Lammich et al., 2010) and identified a stable G-quadruplex structure of ADAM10 mRNA (Lammich et al., 2011). The stability of a G-quadruplex structure depends in part on binding proteins, such as fragile X mental retardation protein (FMRP; Oostra and Willems, 1995) and indeed, FMRP immunoprecipitated from cortical mouse tissue revealed bound ADAM10 mRNA (Pasciuto et al., 2015). Mice lacking FMRP displayed a shift of APP processing toward the non-amyloidogenic pathway during early stages of development, which subsequently led to synaptic and behavioral deficits (Pasciuto et al., 2015). Lack of FMRP could increase ADAM10 levels because FMRP stabilizes the G-quadruplex structure and can thus perturb translation initiation, as has been previously suggested for two other mRNAs (MAP1B and PP2A) that are FMRP targets (Lu et al., 2004; Castets et al., 2005). Another RNA-binding protein found to regulate ADAM10 is the neuronal ELAV protein: nELAV was shown by using immunoprecipitation to bind ADAM10 mRNA via an adenine- and uridine-rich element (Amadio et al., 2009). This might result in an increase in amyloidogenic APP processing. Since A-beta peptides have been found to inhibit ELAV-binding to ADAM10 mRNA (Amadio et al., 2009), this could reduce the ADAM10 amount even further, potentially leading to a vicious cycle.

miRNAs can silence cytoplasmic mRNAs either by triggering degradation or by promoting translation repression. For ADAM10 a prominent example for such a regulatory mechanism is hepatic miR-122, which decreased ADAM10 protein in human hepatic cancer cell lines (Bai et al., 2009). Using a systematic approach, i.e., a combination of different bioinformatics tools, we identified several candidate miRNAs that should act on ADAM10 and evaluated three of them via reporter gene assay—miR-103, -107, and -1306 (Augustin et al., 2012). Additionally, miR-144/451 which has been shown to be induced by A-beta peptide in SH-SY5Y cells decreased ADAM10 protein amount (Cheng et al., 2013). This regulation might be indirect and based on the transcription factor PAX4 (Zhang et al., 2015). In gastric cancer tissue miR-448 (Wu et al., 2016) and in tumor initiator cells of head and neck squamous cell carcinoma miR-494 (Chang et al., 2015) were also identified as novel regulators of ADAM10.

# Post-translational Regulation of ADAM10: Maturation and Interaction Partners

After their synthesis, membrane proteins mature along the secretory pathway; they are transported to distinct compartments of the cell and finally, they locally interact with proteins and lipids of the phospholipid-bilayer. Eventually, they are degraded. Protein synthesis and removal are in homeostasis and thus determine the concentration of functional intramembranous proteins. In principle, ADAM10 can be regulated at all of these stages, offering possibilities for intervention.

The ADAM10 zymogen is cleaved by proprotein convertases within the secretory pathway to yield the active enzyme (see paragraph 1). Removal of the prodomain of ADAMs likely involves a canonical consensus site for the proprotein convertase Furin (Roebroek et al., 1994), which is located between the proand the catalytic domain of ADAM10 (Anders et al., 2001). More recently, a novel cleavage site upstream of the prodomain has been identified (Wong et al., 2015). ADAM10 has four potential N-glycosylation sites of which three are located in the metalloprotease domain (N267, N278, and N439) and one in the disintegrin domain (N551). In bovine ADAM10 all four have been found glycosylated and required for full in vivo activity (Escrevente et al., 2008).

Binding of ADAM10 to synapse associated protein 97 (SAP97) is required for inserting ADAM10 into the synaptic membrane (Marcello et al., 2013). Interaction of SAP97 with ADAM10 is mediated via a protein kinase C (PK C) phosphorylation site within the SAP97 SRC homology domain (Saraceno et al., 2014). Removal of ADAM10 from excitatory synapses occurs by clathrin-mediated endocytosis in human hippocampal tissue (Marcello et al., 2013). This is mediated by the clathrin adaptor protein AP2 which interacts with the ADAM10 C-terminal domain. In addition to control of surface concentrations of ADAM10 by transport mechanisms, further cleavage events may occur: the ectodomain of ADAM10 can be processed by ADAM9/15 or gamma-secretase (Cissé et al., 2005; Parkin and Harris, 2009; Tousseyn et al., 2009). Using recombinant mouse ADAM9 prodomain as a competitive inhibitor of ADAM9, Moss et al. demonstrated an increase of ADAM10-dependent APP processing in human neuronal SH-SY5Y cells (Moss et al., 2011). However, a truncated soluble ADAM10 construct was incapable of shedding cell-associated amyloid precursor protein while earlier reports described that shedded ADAM10 had the ability to cleave endogenous Prion protein in fibroblasts (Cissé et al., 2005).

The intensity of ADAM10 cleavage may further depend on the cytoskeleton: a dominant negative dynamin I mutant not only increased surface expression of both, immature, and mature ADAM10 but also strongly increased the amount of the C-terminal cleavage product of ADAM10 (Carey et al., 2011). In addition to its role as a protease acting at the cell surface it has been speculated that the soluble ADAM10 Cterminus could act as a signaling molecule, facilitating nuclear entry of other proteins (Endsley et al., 2014). A protein class which is deeply involved in for example cytoskeletal anchoring and protein-trafficking is the tetraspanin family (Charrin et al., 2014). Several tetraspanins have been identified to interact with ADAM10: tetraspanin 12 (Tspan 12) binds to ADAM10 in a palmitylation-dependent mechanism and increases non-amyloidogenic shedding of APP by increased enzymatic maturation of the protease (Xu et al., 2009). Co-immunoprecipitation experiments also identified specific ADAM10 interactions with Tspan5, Tspan10, Tspan14, Tspan15, Tspan17, and Tspan33/Penumbra (Haining et al., 2012), which all led to enhanced enzyme maturation. Interestingly, only overexpression of Tspan15 resulted in a reduction of ligandinduced Notch-1 processing by ADAM10 (Jouannet et al., 2016). This led to the assumption that the tetraspanins might differentially influence compartimentalization of ADAM10. Indeed, the apparent diffusion coefficient of ADAM10 was higher in cells overexpressing Tspan15 as compared to control cells or Tspan5 overexpressing cells and also decreased the coimmunoprecipitation of proteins of the tetraspanin web with ADAM10 (Jouannet et al., 2016). Tspan12 and 17 also seem to stabilize a high molecular weight protein complex that tethers ADAM10 to the gamma-secretase allowing rapid sequential processing of substrates (Chen et al., 2015).

ADAM10 is known to be mainly located outside of lipid rafts and alpha-secretase cleavage of APP occurs in non-raft domains (Kojro et al., 2001). Targeting ADAM10 artificially into lipid raft domains of the plasma membrane resulted in impaired enzymatic activity in human neuroblastoma cells (Harris et al., 2009; Kojro et al., 2010). Depletion of one of the constituents of lipid rafts, i.e., cholesterol, enhanced ADAM10 activity in different cellular models (Kojro et al., 2001, 2010; Matthews et al., 2003). The sigma-1 receptor contains a cholesterol recognition domain in its C-terminus and is able to remodel lipid rafts by changing the relative distribution of cholesterol between raft and non-raft fractions (Takebayashi et al., 2004). Interestingly, overexpression of sigma-1 in HEK293 or COS cells diminished Betacellulin cleavage by ADAM10 further substantiating the lipid-sensitivity of the enzyme (Li et al., 2012). Several investigations also report on influence of different lipid species such as trans fatty acids on APP processing balance (e.g., Eckert et al., 2011; Grimm et al., 2012) but in this regard it is not clear if this has a direct influence on ADAM10 or whether indirect mechanisms are involved.

# ROLES OF ADAM10 IN NEURAL HOMEOSTASIS AND PATHOLOGY

ADAM10 has a number of physiological functions (see above) contributing to brain development or neural homeostasis. Diseases challenge this physiological state and the brain reacts to such perturbations with adaptations at the molecular, cellular, and functional level. The picture that is currently emerging from studies using animal models and human brains suggests a two-faced role of ADAM10 in diseases: beneficial as well as detrimental effects can be attributed to the protease depending on the specific setting and the substrates involved. In the following we review some of the conditions and diseases in which ADAM10 has been implicated.

#### ADAM10 in Aging and Alzheimer's Disease

ADAM10's role in these contexts is of particular interest because of its function as in vivo alpha secretase (Jorissen et al., 2010; Kuhn et al., 2010). Cleavage of APP along the nonamyloidogenic pathway yields APPs-alpha, which is important for neuroprotection (Kögel et al., 2012), learning and memory (Taylor et al., 2008; Hick et al., 2015; Xiong et al., 2016), and the structural integrity of neurons (Lee et al., 2010; Tyan et al., 2012; Weyer et al., 2014; Hick et al., 2015). Because cleavage of APP along the non-amyloidogenic pathway decreases with aging (Kern et al., 2006) and reduced APPs-alpha levels were found in CSF of some AD patients (Lannfelt et al., 1995; Sennvik et al., 2000), it is likely that insufficient APPs-alpha levels could contribute to the cognitive deficits of AD patients.

What is known about age-dependent changes in ADAM10 levels or activity in human brain? Unfortunately, with the notable exception of a publication from Bernstein et al. (2003) who compared still-born children with normal aged adults and found an increase in ADAM10 amount, data on ADAM10 in human brain are scarce. To study ADAM10 in humans, peripheral surrogate markers have been used, although it is unclear how comparable they are to CNS expression levels. A study aiming at comparing brain and leukocyte APP processing reported that while ADAM10 is present in brain it remains undetectable in the blood leukocyte fraction (Delvaux et al., 2013). Others, however, demonstrated ADAM10 expression in peripheral mononuclear blood cells as well as in platelets (Colciaghi et al., 2002). Using three groups of cognitively healthy subjects, we recently described an elevation of ADAM10 protein amount as well as catalytic function with cognitively healthy aging (Schuck et al., 2016). The reason why ADAM10 should be up-regulated is unclear. It is conceivable that it is a reaction to age-dependent changes in stress signatures (such as ER stress; e.g., Taylor, 2016) and thus represents a protective response. Although more data are needed, "healthy agers" show an ADAM10-increase whereas AD patients show a decrease (see below). In the former case APPs-alpha could be present in sufficient amounts protecting the brain whereas in the latter case APPs-alpha levels might be insufficient.

Using animal models of AD the role of ADAM10 as a protective protease has been demonstrated: overexpression of the protease at low level (30% above endogenous expression) was sufficient to nearly abolish plaque deposition in APP/PS1 AD model mice (Postina et al., 2004). These changes went hand-inhand with improvements of learning and memory. In line with this gain-of-function approach, overexpression of a dominant negative ADAM10 mutant reduced alpha-secretase activity and worsened cognitive deficits (Schmitt et al., 2006). Interestingly, investigations using peripheral platelets of AD patients and healthy controls reported a decreased ADAM10 amount in AD patients (Colciaghi et al., 2002). Furthermore, ADAM10 levels in patient platelets were highly correlated with performance of the patients in psychological tests (Manzine et al., 2013, 2014). Together, these data suggest that normalizing or even increasing ADAM10 levels in AD could have a disease-modifying or at least disease-protracting effect.

Finally, it should be kept in mind that ADAM10 is multifunctional and that some effects of ADAM10 in the context of aging and AD could depend on ADAM10-mediated cleavage of other substrates than APP, such as Klotho (Chen et al., 2007; Bloch et al., 2009). This protein is linked to longevity (Kurosu et al., 2005) and soluble Klotho (s-Klotho) may be cardioprotective (Xie et al., 2012). Lower CSF s-Klotho levels have also been associated with endothelial dysfunction and neuronal damage in neuropsychiatric systemic lupus erythematosus patients (Ushigusa et al., 2016). Thus, lower ADAM10 levels in the aged brain may have detrimental effects on several levels involving APP processing as well as the processing of other ADAM10 substrates.

# ADAM10, Dendritic Spines and Fragile X Syndrome

The level and/or activity of ADAM10 affect neuronal structures in the adult brain, in particular dendritic spines. This was shown using conditional ADAM10-deficient mice (Prox et al., 2013), which exhibited hippocampal neurons with fewer and abnormally shaped spines. The effect of ADAM10 on spines may depend on several substrates involved in the regulation of spine density, geometry and dynamics, including APP, N-cadherins, Neurexins, Neuroligins, and Nectin-1 (Prox et al., 2013). These substrates act as cell adhesion molecules and are known to influence spine morphology as well as synaptic transmission.

Of particular interest in this context is again the link between ADAM10 and APP. APP and in particular its cleavage product APPs-alpha have been shown to regulate dendritic complexity as well as spine numbers of hippocampal neurons (Lee et al., 2010; Tyan et al., 2012; Weyer et al., 2014). This effect appears to be agedependent: whereas young APP-deficient mice had normal spine numbers, older APP-deficient mice showed a decrease in their spine density (Tyan et al., 2012). It may also depend on the brain region, since APP levels may show regional variations (Del Turco et al., 2016). Since APPs-alpha is generated by ADAM10 cleavage of APP, it is likely that some of the structural effects on spines seen in conditional ADAM10 knock-out mice (Prox et al., 2013) are the result of reduced APPs-alpha levels. Indeed, Prox et al. (2013)reported a reduction of APPs-alpha in brain of conditional ADAM10 knock-out mice to 5% of control levels. Since aging is also associated with reduced dendritic complexity and spine densities (Dickstein et al., 2007), it is attractive to speculate that reduced ADAM10 levels/activity and reduced APPs-alpha levels could play a role in this context (Lannfelt et al., 1995; Sennvik et al., 2000).

Whereas, reduction of ADAM10 may contribute to conditions in which fewer dendritic spines are observed, too much ADAM10 could contribute to diseases with the opposite phenotype, i.e., too many spines. Fragile X syndrome (FXS) is a good example for such a disease and is characterized by increased spine numbers and abnormally long spines. Mice with a fragile X mental retardation protein (FMRP) knock-out at an early adult age (P21, Pasciuto et al., 2015) showed a parallel increase in the expression of APP and mature ADAM10, suggesting that ADAM10 processing of APP could play a role. Indeed, primary fibroblasts obtained from adolescent and adult patients with FXS showed similar changes (Pasciuto et al., 2015), suggesting that an upregulation of ADAM10 and APP could also occur in brains of FXS-patients. In line with these findings, overexpression of APP (Lee et al., 2010) caused FXS-like spine changes in vitro.

In sum, under healthy conditions ADAM10 and its processing of cell adhesion molecules at synapses is in a homeostatic balance. Reduction of ADAM10 levels may cause a reduction in spine densities. Conversely, an increase in ADAM10 levels may increase the density of spines. Normalizing ADAM10 levels could be a potential therapeutic strategy.

## Synaptic Function and Epilepsy

Dendritic spines and excitatory synaptic neurotransmission are intimately linked (Kasai et al., 2010). It is, therefore, in line with the effects of ADAM10 on dendritic spines that conditional ADAM10 knock-out mice show functional abnormalities at excitatory synapses: electrophysiological analysis of hippocampal CA1 neurons revealed almost normal basal synaptic transmission and short-term-plasticity but a grossly impaired induction of long-term-potentiation (Prox et al., 2013). These electrophysiological abnormalities were accompanied by reduction of postsynaptic density protein-95 (PSD-95) and several NMDA-receptor subunits, suggesting a severe disruption of synaptic architecture and function. Spatial learning was impaired at the behavioral level (Prox et al., 2013). Mechanistically, the impairment of synaptic plasticity and learning could be linked to several of the substrates of ADAM10 at the synapse. Again, APP is one of the more interesting candidates because its fragment APPs-alpha has been shown to be involved in synaptic plasticity, as well as learning and memory in the hippocampus (Taylor et al., 2008; Hick et al., 2015). At present it is unknown whether some of the abnormalities of the conditional ADAM10 mice could be rescued by recombinant APPs-alpha. Answering this question could help to better understand the relative importance of APP in this context.

Gain-of-function experiments resulted in an increased susceptibility of neurons for seizures: in mice overexpressing ADAM10 under the Thy1 promoter (∼postnatal day 1), kainatetreatment evoked stronger and longer episodes of seizures as compared to wild type mice (Clement et al., 2008). Moreover, a dominant negative variant of ADAM10 seemed protective against this form of experimental epilepsy as shown e.g., by decreased neuronal damage score. The role of ADAM10 in epilepsy is complex, however, since conditional ADAM10 knock-out mice also showed seizures. However, these seizures may be linked to the gliosis observed in these mice (Prox et al., 2013). Thus, different disease mechanisms could play a role and additional work is needed before the role of ADAM10 in epilepsy can be assessed.

#### ADAM10 and Traumatic Brain Injury

ADAM10 is upregulated at injury sites (Zohar et al., 2011) and in denervated areas of the brain following brain injury (Warren et al., 2012; Del Turco et al., 2014). Reactive astroglia but not microglia has been shown to upregulate the protease following denervation (Warren et al., 2012; Del Turco et al., 2014). ADAM10's role in traumatic brain injury is still poorly understood and different modes of action have been proposed, which may not be mutually exclusive. First, the protease could be involved in the reorganization of the extracellular matrix of denervated regions (e.g., Deller et al., 2000), which may be a requirement for denervation-induced synaptic reorganization to occur (Warren et al., 2012). Secondly, ADAM10 could process synaptic adhesion molecules such as N-cadherin (Malinverno et al., 2010; Warren et al., 2012), neuroligins (Suzuki et al., 2012), or ephrins (Janes et al., 2005), which could tether degenerating terminals to their postsynaptic membranes. Cleaving the transsynaptic molecular bridge could be a necessary first step for re-innervation. Thirdly, ADAM10 could cleave APP and liberate APPs-alpha (Del Turco et al., 2014), which is neuroprotective in vitro (Kögel et al., 2012) and protects neurons in vivo following brain injury (reviewed in: Plummer et al., 2016). Of note, in these contexts upregulation of ADAM10 is associated with a short-term neural "defense"-reaction. This reaction seems to be transient and ADAM10 levels return to normal within a few days. In contrast, under lesioning conditions resulting in longterm upregulation of ADAM10 (Warren et al., 2012) synaptic reorganization failed and functional deficits persisted. Under

these conditions, pharmacological blockade of ADAM10 helped to restore function, suggesting that long-term upregulation of ADAM10 is detrimental for brain rewiring. Collectively, these findings suggest the following model for ADAM10's role in brain injury: ADAM10 plays a plasticity-enhancing and neuroprotective role during the first phase following injury. It shapes the extracellular environment for sprouting fibers, clears synaptic sites, and liberates neuroprotective APP fragments. During the second phase, however, sprouting of surviving fibers occurs and new synapses form. If ADAM10 is still upregulated at this time point it could interfere with the stabilization of new synapses by cleaving the molecular bridge that binds pre- and postsynaptic structures (**Figure 4**).

### Stroke and Psychiatric Diseases

A positive association between the rs653765 polymorphism of ADAM10 and atherosclerotic cerebral infarction has been found in a Chinese population cohort (Li et al., 2013). Patients that carried the rs653765 C > T mutation also showed increased ADAM10 mRNA in PBMCs as did aged patients in comparison to younger patients or healthy controls (>70 years). As already mentioned, an association between CpG-site methylation in the ADAM10 locus and psychiatric tic-disorders has been identified (Zilhão et al., 2015). Beside this epigenetic association, ADAM10 has been characterized as one of the candidates within a low density GWA study for conduct disorder (Jian et al., 2011). Further associations with psychiatric disorders are conceivable, since ADAM10 processes neuroligins, which have also been

FIGURE 4 | ADAM10's potential two-faced role under conditions of brain injury. Whereas a transiently increased activity/amount of ADAM10 seems to be part of a protective and restorative response to mild neural lesions, a persistent upregulation of ADAM10 as seen following severe lesions may be deleterious.

identified as candidate genes in autism spectrum disorders and schizophrenia (e.g., Sun et al., 2011; Chen et al., 2014). In this regard it is of interest that Ray and colleagues recently reported alterations of APP processing and amount not only in FXS patients but also in autism spectrum disorder patients (Ray et al., 2016). However, they also reported age-dependent elevation of ADAM17 in the latter so that this protease might be due to observed changes instead of ADAM10.

### Brain Tumors

ADAM10 may have deleterious effects for patients with brain tumors because it may promote the spreading of tumor cells. Reduced motility of glioblastoma cells treated with ADAM10 targeted siRNA has been observed (Kohutek et al., 2009) and invasiveness of pituitary adenomas correlated with ADAM10 expression level (Pan et al., 2012). Both publications suggests that ADAM10 may process putative barriers restricting tumor cells. With regard to cancer stem cells Bulstrode et al. reported that ADAM10 promotes the self-renewal of brain tumor sphere forming cells (Bulstrode et al., 2012). Additionally, treatment with inhibitors specific for ADAM10 or ADAM17 increased immune recognition of glioblastoma-initiating cells by natural killer cells (Wolpert et al., 2014). This seemed to be due to enhanced cell surface expression of UL16-binding protein 2 (ULBP2), which is shedded by both proteinases. In sum, stimulating ADAM10 expression, as suggested for AD patients (see below), may not be an option for oncology patients.

# CONCLUSION AND OUTLOOK—ADAM10-TARGETING DRUGS AS NOVEL THERAPEUTICS?

ADAM10 is a biologically multifunctional protease involved in many important processes. It is expressed almost ubiquitously in the body. High amounts of ADAM10 are found in neural tissue during development, maturation and aging. Under conditions of neuronal activity and under some pathological conditions, ADAM10 expression is altered which, in turn, leads to changes in the processing of its substrates. The biological activity of these substrates and their cleavage products lead to measurable changes in function, biochemistry and even neural structures.

How can ADAM10 be considered a target for therapy in spite of the large number of substrates with multiple functions? First of all it has to be kept in mind that the majority of data on ADAM10 was obtained using in vitro systems. Although these studies can show putative interactions, such in vitro interactions require in vivo verification. ADAM10 can only cleave putative substrates if protease and substrate are in the same microcompartment at the same time. Since the availability of substrates and their distribution changes during development and aging, it is likely that changes in ADAM10 expression result in different effects depending on the age and stage of development of an organism. Of particular importance for the use of drugs targeting ADAM10 is the fact that ADAM10 shows an increasing overlap with its substrate APP with age (Marcinkiewicz and Seidah, 2000), suggesting that ADAM10-mediated APP cleavage may become more relevant at later stages in life. Regardless of these considerations, a rational approach to therapy development will take all these possibilities into account and will look at the net biological effects changes in ADAM10 expression induce in neural tissue. Complex in vitro systems, such as organotypic slice cultures (e.g., Gähwiler et al., 1997; Del Turco and Deller, 2007) and in vivo models (e.g., Postina et al., 2004) will help to address these questions.

The duration of ADAM10 expression changes may also play a critical role during the course of a disease. The enzyme can be briefly upregulated or persistently increased, depending on the specific conditions. Thus, ambivalent or even opposite outcomes can be expected for ADAM10 effects on brain structure and function, as has been shown for its role in brain injury (see **Figure 4**). Finally, patients may have different genetic predispositions or constitutively elevated ADAM10 levels, which might also harm the brain as has been shown for infarction and cancerogenesis (Pan et al., 2012; Li et al., 2013).

In sum, there are drug safety-issues which need to be explored before ADAM10 targeting drugs can be considered for therapy. The complex expression patterns and time courses of ADAM10 and its substrates may constrain the use of ADAM10-targeting drugs to specific situations, aged patients or some diseases. A clinical pilot study using acitretin was, however, promising

#### REFERENCES


(Endres et al., 2014). Acitretin (Neotigason), which increases APP processing along the non-amyloidogenic pathway in vitro, in primary cells, and in AD model mice (Tippmann et al., 2009; Reinhardt et al., 2016), was given to patients with mild to moderate AD for 4 weeks. Compared to the placebo group, treated patients showed a significant increase in their CSF APPs-alpha levels. Acitretin-treatment was well-tolerated and considered overall safe (Endres et al., 2014). Longer and larger trials will now be needed to evaluate the potential of acitretin as a novel AD-therapeutic. In any case, the pilot study raises hopes that at least for some AD patient groups ADAM10-targeting therapies may eventually prove to be useful.

#### AUTHOR CONTRIBUTIONS

KE and TD contributed equally to all aspects of this review, including development of the overall concept, writing and creating the figures.

## FUNDING

Grant sponsor: Deutsche Forschungsgemeinschaft (DFG, FOR 1332 to TD) and (NGFN, FKZ01GS08130) and the Alfons Geib-Stiftung to KE.

Down syndrome brains. Age (Dordr). 31, 11–25. doi: 10.1007/s11357-008- 9076-x


suppressed by its 5′ -untranslated region. J. Biol. Chem. 285, 15753–15760. doi: 10.1074/jbc.M110.110742


expression correlates with clock drawing test scores in Alzheimer's disease. Int. J. Geriatr. Psychiatry 29, 414–420. doi: 10.1002/gps.4020


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Endres and Deller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein

Lisa Seipold and Paul Saftig\*

Institut für Biochemie, Christian-Albrechts-Universität zu Kiel (CAU), Kiel, Germany

Tetraspanins are a family of ubiquitously expressed and conserved proteins, which are characterized by four transmembrane domains and the formation of a short and a large extracellular loop (LEL). Through interaction with other tetraspanins and transmembrane proteins such as growth factors, receptors and integrins, tetraspanins build a wide ranging and membrane spanning protein network. Such tetraspanin-enriched microdomains (TEMs) contribute to the formation and stability of functional signaling complexes involved in cell activation, adhesion, motility, differentiation, and malignancy. There is increasing evidence showing that the tetraspanins also regulate the proteolysis of the amyloid precursor protein (APP) by physically interacting with the APP secretases. CD9, CD63, CD81, Tspan12, Tspan15 are among the tetraspanins involved in the intracellular transport and in the stabilization of the gamma secretase complex or ADAM10 as the major APP alpha secretase. They also directly regulate, most likely in concert with other tetraspanins, the proteolytic function of these membrane embedded enzymes. Despite the knowledge about the interaction of tetraspanins with the secretases not much is known about their physiological role, their importance in Alzheimer's Disease and their exact mode of action. This review aims to summarize the current knowledge and open questions regarding the biology of tetraspanins and the understanding how these proteins interact with APP processing pathways. Ultimately, it will be of interest if tetraspanins are suitable targets for future therapeutical approaches.

#### Edited by:

Thomas Deller, Goethe-University, Germany

#### Reviewed by:

Monica DiLuca, University of Milan, Italy Claus Pietrwik, University of Mainz, Germany Jochen Walter, University of Bonn, Germany

#### \*Correspondence:

Paul Saftig psaftig@biochem.uni-kiel.de

Received: 31 October 2016 Accepted: 05 December 2016 Published: 21 December 2016

#### Citation:

Seipold L and Saftig P (2016) The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein. Front. Mol. Neurosci. 9:149. doi: 10.3389/fnmol.2016.00149 Keywords: tetraspanin, Alzheimer disease, membrane microdomains, amyloid precursor protein, secretases, amyloid beta

# INTRODUCTION

The neurotoxic amyloid beta (Aβ) peptide is a major component of senile plaques in Alzheimer's Disease (AD) and derives from its precursor the amyloid precursor protein (APP). Despite an intensive effort and increasing understanding of its role in AD the physiological function of APP is not completely understood. APP and its relatives amyloid-like protein-1 (APLP1) and amyloid-like protein-2 (APLP2) are proteolytically processed, ubiquitously expressed and share overlapping functions. APP has been linked with trophic roles in neurons and synapses, axon pruning, intracellular signaling and apoptosis (Muller and Zheng, 2012). How APP interaction with other proteins is defined, how its proteolytic processing is controlled and how signaling events are regulated by APP is poorly understood. Proteomics-based approaches and yeast-two-hybrid screens have been used to identify the protein interaction network of APP (Kohli et al., 2012; Yu et al., 2015) and of the proteases known to cleave APP (Wakabayashi et al., 2009; Jeon et al., 2013). Among others, members of the tetraspanin family have been identified. Tetraspanins have been characterized as scaffold for protein interactions establishing tetraspanin-enriched microdomains (TEMs) and are involved in grouping APP and functional important protein partners.

This review focuses on the emerging role of tetraspanins in the regulation of the proteases involved in the proteolytic processing of APP. The available knowledge about how tetraspanins regulate processing and intracellular trafficking of APP and APP-cleaving secretases is summarized. It is discussed why tetraspanins are attractive novel drug targets. There are some excellent reviews covering different aspects of tetraspanin biology thereby providing a useful overview about their diverse functions (Berditchevski and Odintsova, 2007; Yanez-Mo et al., 2009; Charrin et al., 2014).

#### WHAT ARE TETRASPANINS?

Tetraspanins are compact and glycosylated transmembrane proteins, that span cell membranes four times. Two extracellular domains, one larger and one smaller loop are separated from three cytosolic domains, one short loop and one N-terminal and C-terminal end, respectively. Intracellular cysteine residues of the tetraspanins can be modified by lipidation, i.e., addition of palmitate, possibly contributing to the establishment of tetraspanin microdomains and the regulation of intracellular signaling events (Berditchevski et al., 2002; Charrin et al., 2002; Yang et al., 2002). The large extracellular loop (LEL) and the transmembrane domains play a role in mediating proteinprotein interactions (Hemler, 2003; Charrin et al., 2009). The structure of the isolated LEL of human CD81 was solved. It looks mushroom-shaped and it consists of a conserved subdomain, including three helices and a more variable one with two helices, possibly involved in the binding to other membrane proteins (Kitadokoro et al., 2001; Seigneuret et al., 2001). The full CD81 structure revealed a cone-like structure, where the LEL harbors an intramembrane cavity which is supposed to bind cholesterol (Zimmerman et al., 2016). It is speculated that the cholesterol bound structure favors a closed structural state of this tetraspanin with less tightly bound partner proteins.

Thirty three members of tetraspanins have been described. They can be mainly found at the plasma membrane and within endocytic membranes. Co-immunoprecipitation and crosslinking experiments revealed a high affinity of tetraspanins to interact with each other and other transmembrane proteins. These are in particular integrins, but also members of the immunoglobulin superfamily, signaling receptors, enzymes such as proteases and many other integral proteins residing in TEMs (Yanez-Mo et al., 2009).

#### FUNCTIONS OF TETRASPANINS

The function of tetraspanins is mainly defined by their ability to interact with other transmembrane proteins. Due to the great variety of partner proteins, tetraspanins are involved in various cellular processes like migration, adhesion, signaling and pathogen infection (Boucheix and Rubinstein, 2001; Lammerding et al., 2003; Barreiro et al., 2008). By regulating cell motility and different signaling pathways, tetraspanins play an important role in cancer progression and metastasis (Boucheix and Rubinstein, 2001; Wang et al., 2011). For example, the tetraspanins CD9 and CD151 contribute to cancer cell invasion by interacting with different integrins and signaling enzymes, like protein kinase C (PKC) and phosphoinositide 4-kinase (PI4K) (Zhang et al., 2001; Wang et al., 2011). Tetraspanins also modulate intracellular signaling pathways by coordinating ligand-receptor binding at the cell surface. This is exemplified by the observation that tetraspanin 3 promotes binding of the NogoA ligand to the receptor sphingosine-1-phosphate-receptor-2 (S1PR2), which activates an intracellular signaling cascade leading to the inhibition of neurite outgrowth (Thiede-Stan et al., 2015). Most tetraspanins regulate the functions of their partner proteins by modulating their spatiotemporal distribution at the plasma membrane and organizing them together with other functional proteins (e.g., enzymes and substrates) (Odintsova et al., 2003; Haining et al., 2012; Thiede-Stan et al., 2015). Recent studies, demonstrated that the interaction with tetraspanins influences the motility of their partner proteins and their association with other molecules within the plasma membrane (Yang et al., 2012; Mattila et al., 2013; Jouannet et al., 2015). In addition, some tetraspanins directly control the trafficking of their partner proteins (Berditchevski and Odintsova, 2007). For example CD63, facilitates endocytosis of the HIV receptor C-X-C chemokine receptor type 4 (Yoshida et al., 2008).

Further genetic and in vivo studies demonstrated the importance of tetraspanins in various physiological and pathophysiological processes. In the central nervous system, the knockout of CD81 increased brain size and number of glial cells in mice (Geisert et al., 2002). Tspan7 regulates spine maturation and AMPA receptor trafficking by interacting with the protein interacting with C-kinase 1 (PICK1) in rat hippocampal neurons (Bassani et al., 2012). Moreover, loss of CD9 in mice impaired formation of axoglial paranodal junctions and caused myelination deficits in the peripheral nervous system (Ishibashi et al., 2004). Also other tetraspanins like Tspan5 (Garcia-Frigola et al., 2001) and Tspan3 (Seipold et al., 2016) are highly expressed in the brain and in neuronal cells. However, their physiological roles remain unclear. Tetraspanin knockout mice additionally revealed the importance of CD9, CD81, CD37, and CD151 for fertilization, brain and peripheral nerve development and the immune response. However, analysis of tetraspanin functions by loss-of-function approaches in mice has been hampered, due to compensatory effects and their redundant functions.

In human, mutations of tetraspanin 7, CD151 and the retinal tetraspanin Peripherin/RDS are associated with X-linked mental retardation, skin and kidney diseases, deafness and retinal degeneration (Kohl et al., 1998; Zemni et al., 2000; Karamatic Crew et al., 2004).

#### TETRASPANINS AS REGULATORS OF α-SECRETASE ACTIVITY

Several tetraspanins associate with the APP secretases and regulate their activity. In particular, the membrane localized αsecretase ADAM10 associates with multiple tetraspanins. Using mild detergent conditions CD9, CD53, CD81, CD82, and CD151 were identified to associate with ADAM10. CD9, CD81, CD82 were able to stimulate ADAM10-dependent TNFα and EGF shedding (Arduise et al., 2008). In an independent study the association of ADAM10 with tetraspanin 12 (Tspan12) caused an accelerated ADAM10 maturation, i.e., the cleavage of the pro-ADAM10 to the mature and active protease, followed by an increased ADAM10-dependent APP processing (Xu et al., 2009). It was postulated that Tspan12 activated proprotein convertases and stabilized the mature form of ADAM10. Coimmunoprecipitation experiments, performed under stringent detergent conditions, suggested that CD9, CD81, CD82, and CD151 did not directly interact with ADAM10 (Dornier et al., 2012). It was concluded that these tetraspanins associate with ADAM10 through interactions mediated by other members of the tetraspanin web. However, tetraspanins belonging to the TspanC8 subfamily still interacted with ADAM10 under stringent immunoprecipitation conditions, indicating that these tetraspanins directly bind to the protease (Dornier et al., 2012).

This evolutionary related subgroup of TspanC8 tetraspanins (**Figure 1**) includes the tetraspanins 5, 10, 14, 15, 17, and 33, which all contain eight conserved cysteine residues within their LEL. Analysis of the TspanC8-ADAM10 interaction revealed that overexpression of individual TspanC8 tetraspanins promoted ADAM10 maturation in human cells and Drosophila melanogaster (Haining et al., 2012). With exception of Tspan10 and Tspan17, TspanC8 overexpression also increased ADAM10 surface localization. Heterologous Tspan10 and Tspan17 expression led to a localization of ADAM10 to late endosomes (Dornier et al., 2012). Although, the C8 tetraspanins exert similar

ClustalOmega and is presented as dendrogram.

effects on ADAM10 maturation and trafficking (except Tspan10 and Tspan17), they have different impact on the cleavage of ADAM10 substrates (Prox et al., 2012; Noy et al., 2016). The overexpression of the TspanC8s members Tspan5 and Tspan14 promoted ligand induced shedding of the Notch receptor. In contrast, expression of Tspan15 reduced Notch processing (Dornier et al., 2012). Tspan15 was the only TspanC8 member, that increased ADAM10-mediated N-cadherin shedding after overexpression in human embryonic kidney (HEK293) and monkey fibroblast-like Cos7 cells (Prox et al., 2012; Noy et al., 2016). It was also shown that the generation of APP C-terminal fragments was differentially affected by certain TspanC8s. While expression of Tspan14 and Tspan33 slightly reduced the appearance of APP C-terminal cleavage products (Jouannet et al., 2015), Tspan5 expression had no effect on the production these fragments. Tspan15 expression in the human osteosarcoma cell line U2OS-N1 (Jouannet et al., 2015) also reduced APP processing, but increased it in murine neuroblastoma (N2a) and HEK293 cells (Prox et al., 2012). Since, tetraspanins act in concert with other tetraspanins in TEMs the conflicting data may be explained by the different composition of tetraspanins in the different cellular systems used.

There is increasing evidence that TspanC8s mediate substrate specificity by a direct interaction with ADAM10 and modulation of its association with other membrane components, e.g., integrins (Jouannet et al., 2016). Using ADAM10 chimeric and truncation constructs, it was demonstrated that the TspanC8s differentially favor the interaction with the ADAM10 membrane proximal stalk region, cysteine-rich domain and disintegrin domain (Noy et al., 2016). TspanC8s may constrain and stabilize ADAM10 in defined conformations (Noy et al., 2016). The expression of TspanC8 tetraspanins had different impact on the membrane environment of ADAM10. Mass-spectrometric analysis of ADAM10-associated proteins revealed that in Tspan5 expressing cells ADAM10 preferably associated with classical components of the tetraspanin web such as the α3β1 integrin, CD9P1 and CD9, which was reduced in Tspan15 transfected cells (Jouannet et al., 2015). Tspan5 expression enhanced ADAM10's localization at the cell periphery, while Tspan15 expression did not (Jouannet et al., 2015).

Tspan3, a TspanC6 tetraspanin, was identified, as another ADAM10 and APP interacting tetraspanin in cells and in the murine brain (Seipold et al., 2016). Tspan3 expression did not obviously influence ADAM10 maturation or trafficking but increased ADAM10-mediated APP cleavage. Tspan3 is likely involved in this process as a scaffold protein, which stabilizes ADAM10 and APP at the cell surface.

An interaction of tetraspanins with ADAM17, which is closely related to ADAM10 and under certain circumstances also exerts α-secretase activity toward APP (Buxbaum et al., 1998), has only been described for CD9. Heterologous expression of CD9 or treatment with CD9-specific antibodies inhibited phorbol ester (PMA)-stimulated shedding of the ADAM17 substrates TNFα and ICAM-I, while CD9 knockdown increased it (Gutierrez-Lopez et al., 2011). In the same manner, treatment with neutralizing anti-CD9 monoclonal antibodies reduced ADAM17-mediated shedding of LR11 (SorLa) in human Seipold and Saftig Tetraspanins and APP

leukocytes, while CD9 expression increased it (Tsukamoto et al., 2014). Most likely CD9 inhibits ADAM17 sheddase activity by affecting the membrane compartmentalization of ADAM17 itself or its substrates. CD9 may even interact with both, ADAM10 and ADAM17, but exerting opposite effects on their activity with regards to TNFα shedding (Gutierrez-Lopez et al., 2011).

# TETRASPANINS AS REGULATORS OF γ-SECRETASE ACTIVITY

Next to ADAM10 also the γ-secretase complex interacts with tetraspanins. Wakabayashi et al. showed that Presenilin-1 and Presenilin-2 associate with the tetraspanins CD9, CD81, Upk1b as well as with the tetraspanin associated proteins EWI-F, EWI-2 which connect the tetraspanin web with the actin cytoskeleton (Sala-Valdes et al., 2006) and CD98hc (Fenczik et al., 2001), a regulator of integrin signaling and amino acid transport (Wakabayashi et al., 2009). The activity of the γ-secretase complex was strongly decreased upon knockdown of CD81, EWI-F, and CD98hc, which correlated with a decrease in Aβ production. Inhibition of γ-secretase activity was also observed in CD9 and CD81-deficient mouse embryonic fibroblasts revealed by an accumulation of C-terminal fragments of the γ-secretase substrates APP, APLP-2 ADAM10, N-Cadherin and Syndecan-3. Treatment with anti-CD9 monoclonal antibodies reduced Aβ levels in HEK293 cells (Wakabayashi et al., 2009). Additionally, siRNA mediated knockdown of Tspan33 in HeLa cells reduced the γ-secretase dependent cleavage of a constitutively active, truncated form of Notch1 and that of T-cell acute lymphoblastic leukemia (T-ALL) Notch1 oncogenic mutants (Dunn et al., 2010).

Independent studies analyzing the interactome of the γsecretase complex identified the tetraspanins CD63 and Tspan3 in the network of the presenilin interacting proteins (Jeon et al., 2013; Seipold et al., 2016). CD63 is one of the few tetraspanins which is found on late endosomal and lysosomal membranes (Rous et al., 2002). CD63 associates with CD9, CD81, and CD82 within the tetraspanin web. However, the functional consequence of this interaction for the γ-secretase complex has not been elucidated. Due to its functions in the endosomal sorting complex required for transport (ESCRT)-independent formation of intraluminal vesicles (van Niel et al., 2011) CD63 could control the degradation of the γ-secretase complex.

#### TETRASPANINS AS A PART OF A MULTI-SECRETASE COMPLEX

Another mechanism by which tetraspanins regulate both, α- and γ-secretase, activity has recently been reported by Chen et al. (2015). It was shown that α- and γ-secretase associate in an active multiprotease complex at the plasma membrane (**Figure 2**). Assembly of this multi-secretase complex seems to be modified by Tspan12 and the TspanC8 tetraspanins Tspan 5, 14, 12, and 17. Knockdown of Tspan5 and Tspan14 decreased ADAM10 association with the γ-secretase complex, which correlated with a reduced presence of mature ADAM10. Knockdown of Tspan12 and Tspan17 also decreased the association of ADAM10 with the

γ-secretase complex and the α-secretase-dependent generation of soluble sAPPα, but did not alter ADAM10 maturation. Moreover, Tspan12 and Tspan17 seem to contribute to an α-/γ-secretase feedback mechanism. This feedback mechanism is related to γ-secretase inhibition and causes an increase of sAPPα at the expense of sAPPβ. This is also accompanied by an increase of APP and BACE1 surface levels. This effect was less effective after knockdown of Tspan12 and Tspan17 (Chen et al., 2015).

In conclusion, tetraspanins are potent regulators of αand γ-secretase activity, which modulate maturation, complex assembly, trafficking and substrate specificity. In regards to βsecretase cleavage, no direct interaction of tetraspanins with the β-secretase BACE1 has been reported. The metalloprotease Meprin β, which cleaves APP in a manner similar to BACE1 (Bien et al., 2012), interacts with Tspan8 and resides together with APP in TEMs (Schmidt et al., 2016). However, Tspan8 had no impact on the proteolytic activity of Meprin β towards APP.

# TETRASPANINS AS THERAPEUTIC TARGETS

The treatment with monoclonal antibodies (mAbs) against CD81 diminished the development of neurological symptoms in a multiple sclerosis mouse model (Dijkstra et al., 2008) and prevented hepatitis C virus (HCV) infection after prophylactic injection in mice (Meuleman et al., 2008). Application of anti-CD9 antibodies reduced tumor growth and progression in gastric cancer mouse xenografts (Nakamoto et al., 2009). Stimulatory CD151 antibodies promoted cell adhesion and thereby reduced immobilization of tumor cells and metastasis (Zijlstra et al., 2008). The humanized anti-CD37 IgG fusion protein Otlertuzumab (TRU-016) is a potential drug for the treatment of lymphoid B-cell malignancies (Robak et al., 2009) and was tested in phase 1 clinical trials for the treatment of chronic lymphocytic leukemia (Byrd et al., 2014). It is proposed that mAbs directed against tetraspanins inhibit lateral associations or cause the formation of tetraspanin aggregates, which disrupt TEMs and cause a downregulation of the targeted

Seipold and Saftig Tetraspanins and APP

tetraspanin and its partner protein(s) (Hemler, 2008). Also recombinant soluble large extracellular domains (sLEL) may inhibit tetraspanin-dependent functions. Similar to mAbs, CD81 sLELs reduced HCV infectivity and blocked HIV-1 entry into macrophages (Flint et al., 2006; Ho et al., 2006).

The potential of tetraspanins to modulate γ-secretase activity in AD was demonstrated by RNAi-mediated knockdown experiments. The downregulation of CD81 and tetraspaninassociated proteins EWI-F and CD98hc reduced the secretion of neurotoxic Aβ in HEK cells, stably overexpressing a mutated form of APP, which favors amyloidogenic processing. Likewise, treatment with anti-CD81 and anti-CD9 mAbs decreased Aβproduction in HEK293 cells (Wakabayashi et al., 2009). However, most therapeutic approaches targeting γ-secretase activity were accompanied by severe side effects, like skin cancer development, gastrointestinal toxicity and infections (Doody et al., 2013), due to the physiological role of its substrates, for example Notch1 (Haapasalo and Kovacs, 2011). Interestingly, the individual knockdown of CD9 and CD81 in HeLa cells had no significant effect on the activity of different leukemic mutant forms of Notch1 (Dunn et al., 2010). It was further shown that human Tspan33 promotes γ-secretase cleavage of Notch and that depletion of Tspan33 might be a potential target in T-ALL, a rare yet aggressive form of lymphoblastic leukemia, which is associated with activating mutations of Notch1 (Weng et al., 2004; Dunn et al., 2010). Since CD9, CD81, and Tspan33 are also regarded as regulators of ADAM10- and ADAM17 (Arduise et al., 2008; Gutierrez-Lopez et al., 2011; Haining et al., 2012; Jouannet et al., 2016), the effects of potential therapeutics have to be studied carefully.

By reducing Aβ- and increasing sAPPα-production the upregulation of ADAM10 expression had beneficial effects in an AD mouse model (Postina et al., 2004). ADAM10 is another promising target for the treatment of AD, as demonstrated by a recent study using the synthetic retinoid acitretin to increase ADAM10 expression in AD patients (Endres et al., 2014). Targeting specific members of the TspanC8s, which enhance ADAM10 activity, but have different impact on its substrate specificity, could possibly reduce side effects of a global ADAM10 activation. Moreover, most of the TspanC8s are not expressed in all cell-types (Dornier et al., 2012; Jouannet et al., 2016), indicating that targeting these tetraspanins could regulate ADAM10 activity in a cell-type specific manner. With regard to AD, enhancing ADAM10 non-amyloidogenic APP processing could be achieved by stimulation of Tspan12, Tspan15 and Tspan33 using agonistic monoclonal antibodies, sLELs or small molecular drugs that increase the promoter activity and protein expression of these tetraspanins.

ADAM10 is also associated with tumor progression, metastasis and inflammation by site-specific cleavage of several adhesion molecules and cytokines. In this case a downregulation of its proteolytic activity could be of therapeutic benefit (Moss et al., 2008; Saftig and Reiss, 2011). ADAM10-mediated N-cadherin shedding was associated with cancer cell migration (Kohutek et al., 2009) promoting tumor progression and metastasis. In this regard downregulation of Tspan5 and Tspan15, which predominantly promote ADAM10-mediated N-cadherin shedding (Noy et al., 2016), by antagonistic mAbs, sLELs or RNAi, could be a therapeutic option. By sharing several substrates with ADAM10, inhibition of ADAM17 is also effective in different kinds of cancer and inflammatory disorders (Saftig and Reiss, 2011). The expression of CD9 reduced ADAM17 dependent TNFα shedding (Gutierrez-Lopez et al., 2011), which is a main factor in inflammation and involved in rheumatoid arthritis, psoriasis and inflammatory bowel disease.

To evaluate the full therapeutic potential of tetraspanins, the exact mechanisms and consequences of potential tetraspanindirected therapeutics need to be further investigated. Due to their multiple interaction partners and the complex organization in TEMs, tetraspanins also have opposing functions, which might depend on the cellular system. While downregulation of Tspan33 in HeLa cells decreased Notch1 signaling (Dunn et al., 2010), its overexpression in U2OS-N1 cells reduced Notch1 activity (Jouannet et al., 2016). Targeting such tetraspanins could cause severe adverse effects such as cancer development and inflammation. Moreover, redundancy of tetraspanin functions and compensatory effects might decrease the clinical activity of potential therapeutics.

## CONCLUSION

In summary, tetraspanins are potent regulators of APP cleaving enzymes. In particular, tetraspanins came into focus as cell-type and substrate specific regulators of the α-secretases ADAM10 but also of the γ-secretase complex.

Their specific functions and localization make tetraspanins an interesting target for the treatment of AD and possibly other diseases. However, first approaches trying to target tetraspanins have not succeeded, which could be explained by their functional redundancy. It will be necessary to better understand how tetraspanins exactly work and how their redundancy is regulated. Another central aspect is how tetraspanin expression is regulated and if tetraspanin dysfunctions are associated with the development of AD.

# AUTHOR CONTRIBUTIONS

Both authors wrote the review manuscript. LS designed the figure art.

# FUNDING

Work in the laboratory of PS has been funded through grants of the Deutsche Forschungsgemeinschaft (DFG) in the SFB877, A3, and A12; through support of the VERUM foundation, the Alzheimer Research Price of the Breuer Foundation and the Interuniversity Attraction Poles Program IUAP P7/16 of the Belgian Federal Science Policy Office.

#### REFERENCES


and promotes glioblastoma cell migration. J. Neurosci. 29, 4605–4615. doi: 10.1523/JNEUROSCI.5126-08.2009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Seipold and Saftig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Corrigendum: The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein

#### Lisa Seipold and Paul Saftig\*

**A corrigendum on**

Institut für Biochemie, Christian-Albrechts Universität zu Kiel, Kiel, Germany

Keywords: tetraspanin, Alzheimer disease, membrane microdomains, amyloid precursor protein, secretases, amyloid beta

#### Edited and reviewed by:

Thomas Deller, Goethe-University, Germany

\*Correspondence: Paul Saftig psaftig@biochem.uni-kiel.de

Received: 25 January 2017 Accepted: 31 January 2017 Published: 10 February 2017

#### Citation:

Seipold L and Saftig P (2017) Corrigendum: The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein. Front. Mol. Neurosci. 10:37. doi: 10.3389/fnmol.2017.00037

#### **The Emerging Role of Tetraspanins in the Proteolytic Processing of the Amyloid Precursor Protein**

by Seipold, L., and Saftig, P. (2016). Front. Mol. Neurosci. 9:149. doi: 10.3389/fnmol.2016.00149

There was a mistake of a statement on page 5 (4th paragraph). It should read: "... In this regard downregulation of Tspan15, which predominantly promotes ADAM10-mediated N-cadherin shedding (Noy et al., 2016), by antagonistic mABs, sLELs, or RNAi, could be a therapeutic option....". The authors apologize for the mistake.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Seipold and Saftig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Structure and Synaptic Function of Metal Binding to the Amyloid Precursor Protein and its Proteolytic Fragments

#### Klemens Wild<sup>1</sup> \*, Alexander August <sup>2</sup> , Claus U. Pietrzik <sup>3</sup> and Stefan Kins <sup>2</sup> \*

<sup>1</sup>Heidelberg University Biochemistry Center (BZH), University of Heidelberg, Heidelberg, Germany, <sup>2</sup>Division of Human Biology and Human Genetics, Technical University of Kaiserslautern, Kaiserslautern, Germany, <sup>3</sup> Institute for Pathobiochemistry, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany

Alzheimer's disease (AD) is ultimately linked to the amyloid precursor protein (APP). However, current research reveals an important synaptic function of APP and APP-like proteins (APLP1 and 2). In this context various neurotrophic and neuroprotective functions have been reported for the APP proteolytic fragments sAPPα, sAPPβ and the monomeric amyloid-beta peptide (Aβ). APP is a metalloprotein and binds copper and zinc ions. Synaptic activity correlates with a release of these ions into the synaptic cleft and dysregulation of their homeostasis is linked to different neurodegenerative diseases. Metal binding to APP or its fragments affects its structure and its proteolytic cleavage and therefore its physiological function at the synapse. Here, we summarize the current data supporting this hypothesis and provide a model of how these different mechanisms might be intertwined with each other.

#### Edited by:

Thomas Deller, Goethe University Frankfurt, Germany

#### Reviewed by:

Stefan Lichtenthaler, DZNE - German Center for Neurodegenerative Diseases, Germany Michael R. Kreutz, Leibniz-Institute for Neurobiology, Germany

#### \*Correspondence:

Klemens Wild klemens.wild@bzh.uni-heidelberg.de Stefan Kins s.kins@biologie.uni-kl.de

> Received: 25 October 2016 Accepted: 16 January 2017 Published: 31 January 2017

#### Citation:

Wild K, August A, Pietrzik CU and Kins S (2017) Structure and Synaptic Function of Metal Binding to the Amyloid Precursor Protein and its Proteolytic Fragments. Front. Mol. Neurosci. 10:21. doi: 10.3389/fnmol.2017.00021 Keywords: amyloid precursor protein (APP), zinc, copper, synaptic transmission, Alzheimer's disease

# INTRODUCTION

Alzheimer's disease (AD) is a fatal neurodegenerative disorder and a severe burden of our aging societies (Schaller et al., 2015). One pathological hallmark of AD is the formation of senile plaques deposited in the brain concomitant with a massive decline of neuronal mass and therewith of memory and cognitive abilities (Selkoe and Hardy, 2016). On the molecular level, the amyloid precursor protein (APP) is fundamental to the pathology as the plaques predominantly consist of its proteolytic breakdown product, which is the neurotoxic amyloid-beta peptide (Aβ; Selkoe, 2011; Haass et al., 2012). APP and the paralogous APP-like proteins (APLP1 and 2) are expressed in various tissues (APLP1 only in the brain) and in alternative splice forms (Walsh et al., 2007; Müller and Zheng, 2012) and are concentrated in the synapses of neurons. They are single-span type I transmembrane proteins with a large extracellular domain (ectodomain) and a short cytoplasmic tail APP intracellular domain (AICD; Coburger et al., 2014). APP is a prime example for ectodomain shedding by α-, β-, or γ-secretases and for regulated intramembrane proteolysis (RIP) by the γ-secretase complex (Thinakaran and Koo, 2008; Haass et al., 2012). In the first step, the majority of the extracellular domain is shed off at distinct sites within the juxtamembraneous domain by different proteases, involving ADAM10 and BACE. The resulting extracellular cleavage products are released in form of soluble fragments, designated according to the cleavage site as sAPPα and sAPPβ, respectively (Brunholz et al., 2012). More recently, an additional more N-terminally located η–cleavage site was described, possibly mediated by cleavage of MT5-MMP (Willem et al., 2015). This causes an even more complex picture of APP processing, including besides sAPPη different extracellularly released fragments, such as Aη–α or Aη–β peptides. In fact, accumulating evidence suggests that the APP processing is likely even more complex, as additional cleavage sites within the Aβ domain by e.g., Meprin β have been described (Bien et al., 2012). However, the different residual carboxyl-terminal fragments (CTFs) are subsequently cleaved by the γ-secretase complex, which causes in case of preceded BACE1 shedding release of various Aβ peptides (Aβ1–36 to Aβ1–43) and the AICD (Steiner et al., 2008). The regulation of cleavage is a key event for both physiological and pathological processes, whereby the pathophysiological relevance of all the different peptides is yet not well understood, and depends on the localization of APP, on post-translational modifications, and on its oligomerization state (Kienlen-Campard et al., 2008; Eggert et al., 2009; Haass et al., 2012; Muresan and Ladescu Muresan, 2015; Winkler et al., 2015). APP is in equilibrium between monomeric and dimeric species (Soba et al., 2005; Isbert et al., 2012) and underlies a rapid turnover from the cell surface into endosomal compartments, the presumed major place of Aβ generation (Thinakaran and Koo, 2008; Haass et al., 2012).

APP localization, oligomerization state and processing are influenced by direct binding to copper and zinc ions (Acevedo et al., 2011; Baumkötter et al., 2014; Mayer et al., 2014) and dysregulation of copper and zinc homeostasis are an apparent feature of neurodegenerative diseases, including AD. However, as APP binds copper and zinc with affinities in the nano and micro molar range, respectively, and as local concentrations of copper and zinc can vary quiet a lot, it appears reasonable that metal binding occurs only under certain pathophysiological conditions. Here, we summarize recent advances in the molecular details of copper and zinc binding to APP and their potential impact on APP processing and their potential role in the pathological as well as in the physiological context.

### APP AND ITS FRAGMENTS AS METALLOPROTEINS

APP is a multi-domain membrane protein with a single transmembrane domain (TMD) and several unstructured regions (APP numbering in the following corresponds to APP770, UniProtKB: P05067; **Figure 1**). The large extracellular ectodomain (residues 18–699) is divided into the N-terminal growth factor-like domain (GFLD, 18–123), the copper-binding domain (CuBD, 124–189), an unstructured acidic domain (AcD, 190–289), a Kunitz-type protease inhibitor (KPI) domain with an Ox2 region (290–364, not present in neuronal APP695 splice form), the E2 domain (365–575), and the juxtamembrane region (JMR, 576–699). GFLD and CuBD form a structural unit termed E1 domain. The JMR harbors the sites for secretase cleavage during the process of ectodomain shedding.

Metal binding is well documented to APP and to Aβ peptides (Hesse et al., 1994; Talmard et al., 2007; Kong et al., 2008; Dahms et al., 2012; Baumkötter et al., 2014). Copper binding to APP has been first structurally characterized for the CuBD (Barnham et al., 2003; Kong et al., 2007). Copper (II) binds with high affinity (K<sup>D</sup> of 10 nM; note: binding constants to copper and zinc ions strongly depend on the applied method and conditions and values have to be evaluated critically) in a slightly distorted square pyramidal geometry (a type 2 non-blue site) to three protein ligands (His147, His151 and Tyr168) and two water molecules (**Figure 2A**; Kong et al., 2008). The distorted type 2 geometry is consistent with an observed redox-activity of APP, and CuBD was found to also bind copper (I) (with one water ligand lost) although no conformational change within the protein could be detected. Interestingly, although not interpreted in the original articles, a nearby disulfide bridge (Cys144- Cys174) is partly reduced in the high-resolution structures. The mechanism of how copper-binding to the CuBD could

FIGURE 1 | Schematic overview of amyloid precursor protein (APP770). The domain architecture is given with residues involved in metal binding and cleavage sites of various secretases. The E1 domain contains two copper binding sites (orange lines) with clustered residues (four histidines and one tyrosine), one in the growth factor like domain (GFLD, blue) and one in the copper binding domain (CuBD, green). The E1 domain is followed by an acidic domain (AcD) and a Kunitz protease inhibitor (KPI) domain (gray). The E2 domain (yellow) binds copper via four histidine residues (orange lines). Zinc binding involves three residues of the copper binding site (His457, His507 and His511, while His388 is replaced by a water molecule). The residues corresponding to the recently identified zinc binding site in APLP1 are highlighted in blue. Three of four histidines are conserved while His450 is replaced by Met531 in APP770. The juxtamembrane region (JMR) and transmembrane domain (TMD, gray) harbor the Aβ region (red). After β-secretase cleavage at Met671 and further γ-secretase cleavage at various sites within the TMD, Aβ is liberated into the extracellular space, whereas the APP intracellular domain (AICD) is released into the cytosol. Cleavage by α-secretases occurs within the Aβ region at Lys687. Additionally, η-secretase cleaves the ectodomain at position Asn579.

decrease Aβ production as described in CHO cells (Borchardt et al., 1999) and various mouse studies (Bayer et al., 2003) however remained enigmatic. Specific binding of zinc to two C-terminal cysteines (Cys186 and Cys187) within the CuBD has also been reported (Bush et al., 1993). However, data had been acquired with tryptic and synthetic peptides and in available structures the two cysteines are not available for zinc coordination. Crystallographic analyses of the CuBD with bound zinc would help to clarify this point and would build the basis for follow-up studies, addressing the impact of zinc binding on APP pathophysiology.

Recently, we reported specific copper binding to the N-terminal GFLD (**Figure 2B**; Baumkötter et al., 2014). Here, copper (II) was found to adopt the same ligand geometry with however two protein ligands (His108 and His110) in a flexible hairpin loop also involved in Heparin binding and putatively three water ligands (mimicked in the crystal by an aspartate of a symmetry contact). Binding affinity as determined by isothermal titration calorimetry (ITC) is in the low nanomolar range (K<sup>D</sup> of 28 nM) almost as high as to CuBD. Copperbinding to the GFLD again correlates with the reduction of a neighboring disulfide bridge (Cys98-Cys105) pointing to redoxactivity also of this site. Furthermore, the match of geometry and functionality is suggestive for a complementation of the two copper-binding sites by either a conformational change within the E1 domain into a ''closed'' conformation (**Figure 2C**) or by dimerization of two ''open'' E1 domains in trans (**Figure 2D**). In line, conformational flexibility within the E1 domain has been validated recently (Hoefgen et al., 2015). In contrast to the CuBD and E2 domains, no binding to the GFLD could be detected for zinc ions (Baumkötter et al., 2014).

The E2 domain (Dahms et al., 2012), which like E1 binds strongly to copper (K<sup>D</sup> of 13 nM) and with low affinity to zinc (K<sup>D</sup> of 3.9 µM; with some uncertainties of the ITC measurements) harbors two or three different metal binding sites, here designated as M1–3 (also see **Table 1**). Both metals bind to the same evolutionary conserved site (denoted as M1) within the APP family and due to the more than 100 times lower affinity, zinc (II) cannot compete for copper (II) binding in vitro. In X-ray structures of the metal-bound E2 domain the M1 site is found central to the coiled-coil like fold and consists of four histidines (His388, His457, His507 and His511 for APP770 numbering) spread on three helices (αB, αC and αD; **Figures 3A,B**; Dahms et al., 2012). The ligand coordination has been described as tetrahedrally-distorted square planar geometry (although a fifth ligand is missing here compared to the E1 copper binding sites). Notably, in contrast to the GFLD and CuBD, no redox-activity has been observed for the M1 site. The central location of the M1 site enables a metal dependent conformational switch within the E2 domain with a 12◦ bending of the helical rod (**Figure 3C**) and bending correlates with an increased rigidity and thermostability of the E2 domain.

Copper binding to the E2 domain was reported to stimulate heparin binding, which might be of physiological importance by modulating APP binding to the extracellular matrix during brain development (Dienemann et al., 2015). However, in an X-ray structure of dimeric human APLP1 E2 domain bound to a heparin-hexasaccharide, the M1 site is destroyed and the four histidines are involved in carbohydrate binding instead (Lee et al., 2011). Here, the E2 domain is bent in an orthogonal direction by 13◦ due to dimer formation that allows the accommodation of the ligand in the dimer interface in a 2:1 protein to ligand ratio (**Figure 3D**).

Binding to the M2 site within the E2 domain has been analyzed by competitive cadmium displacement studies similar to analysis of M1 (Dahms et al., 2012). Thereby, cadmium was only partially replaced by zinc, but not by copper. However, no further functional studies validated its functional or structural relevance yet.


TABLE 1 | Metal binding properties of amyloid precursor protein (APP)/APP-like proteins (APLPs).

a Isothermal titration calorimetry; <sup>b</sup>X-ray crystallography; <sup>c</sup>amino acids indicated in gray represent conserved positions w/o experimental data validating copper or zinc binding; <sup>d</sup> reported APLP1 specific Zn binding site, here designated as M3 site. <sup>1</sup>Baumkötter et al. (2014), <sup>2</sup>Hesse et al. (1994), <sup>3</sup>Kong et al. (2007), <sup>4</sup>Dahms et al. (2012), <sup>5</sup>Dienemann et al. (2015) and <sup>6</sup>Mayer et al. (2014)

Two other more recent studies suggested a zinc binding site in the APLP1 E2 domain (between M1 and M2 of APP E2: APLP1 residues His430, His433, His450 and His452) to regulate homo- and hetero-dimerization with APP and APLP2 (Mayer et al., 2014, 2016). Notably, only three of the four histidine residues involved in binding are conserved in APP and APLP2 (**Table 1**). However, as mutations of single histidine residues only had a minor impact on APLP1 cell adhesion features, it appears reasonable to postulate that binding of zinc to the M3 site might affect APP and APLP2 function in a similar manner. Moreover, specific dimerization implies complementation of metal coordination in trans, which would require a different dimerization as observed in the X-ray structure with heparin. Such zinc mediated dimerization was observed in a crystal contact of another recent APLP1 E2 structure bound to a heparin dodecasaccharide (Dahms et al., 2015), which however, was mediated again by other surface exposed histidines. Therefore, the structural and physiological consequences of copper and zinc binding to the E2 domain appear still highly controversial and need further investigations.

Since copper and zinc binding to APP is predominantly mediated by histidine and due to pKa values of this amino acid, the interaction of APP with copper and zinc is impossible under acidic conditions. APP is trafficked through the secretory pathway to the cell surface up to endosomes and lysosomes. Thereby it passes cell compartments with different pH values.

The pH of the ER is near neutral, while the downstream compartments (cis-and trans-Golgi, secretory vesicles) become progressively more acidic down to pH 6 in the trans-Golgi and <5 in lysosomes (Casey et al., 2010). Thus, it is likely that APP binding to copper and zinc varies between different subcellular compartments, possibly altering its structure and binding properties. In line with this, it was reported that APP can adopt different conformations depending on pH (Hoefgen et al., 2015). Interestingly, the altered structure at acidic pH is stabilized by hydrogen bonds involving His147 and Tyr168 in the CuBD that mediate copper binding under neutral pH conditions.

Most attention has been given to copper and zinc binding to Aβ peptides due to the direct influence on pathological processes and metal accumulation (millimolar range) in the amyloid plaques (for review see Tõugu et al., 2011; Tiiman et al., 2013). From a structural viewpoint, the peptide-ligand complexes are difficult to tackle, as both metals rapidly precipitate Aβ and induce a multitude of Aβ conformations and oligomeric assemblies. Furthermore, Aβ conformations strongly depend on the conditions (e.g., pH and solvent) used for structure determination. The role of metal ions in the self-assembly of Aβ has been reviewed in detail recently (Faller et al., 2013). In consensus, copper and zinc bind to the N-terminus of Aβ(1–16; numbering for Aβ-peptide only) in a 1:1 ratio mostly involving the aspartate and alanine at the very N-terminus (Asp672 or Asp1 in Aβ nomenclature, and Ala673), glutamate 11, and various histidines (His6, His13 and His14). Furthermore, redox-active copper is able to induce reactive oxygen species (ROS) that lead e.g., to aggregation-prone cross-linked Aβ dimers by oxidization of Tyr10 (Smith et al., 2007). Respective residues 1–10 are missing in APLP1 and also APLP2 has a two-residue deletion including His6. Affinities for copper vary significantly between attomolar and micromolar values, although careful analyses suggest values in the range of 30–60 nM (Tõugu et al., 2011; Tiiman et al., 2013). No X-ray or NMR structure has been reported for an Aβ-copper complex. At a physiological pH of 7.4 a square-pyramidal coordination is assumed including Asp1 and His6 and His13 or His14 (Faller et al., 2013). Interestingly, this coordination is exactly as found for the X-ray structure of the GFLD-Cu(II) complex, which might therefore be regarded as first atomic model of an Aβ-Cu(II) complex (**Figure 4A**). Ligands might originate from the same or different Aβ peptides, which reflects the ability of copper (II) to bind to all forms of monomeric, fibrillary and non-fibrillary Aβ species. Zinc binds to the same site as copper, but binding is again weaker with affinities in the lower micromolar range. The commonly accepted coordination includes Asp1, Glu11 and the three histidines. Zinc immediately precipitates Aβ and NMR structures are deposited only for zinc-mediated dimeric Aβ-zinc (II) species of a mutant human Aβ(1–16) and rat Aβ(1–16) revealing part of this coordination and different dimerization patterns (**Figures 4B,C**). Aggregation increases the affinity of Aβ for copper and zinc leading to the apparent high concentrations in the amyloids. However, the structural understanding of this process is just at the beginning despite all efforts taken.

In summary, the biometals copper and zinc bind to different folded regions of the APP ectodomain and to various Aβ species. Affinities for all sites are about 100 times higher for copper, which challenges the idea of zinc binding to APP in the synaptic cleft despite its excess. In all complexes the metals seem to be involved in diverse conformational changes of APP and APLP1 as well as APLP2, inducing dimerization or oligomerization in different ways and thus might influence specific physiological and pathological processes.

# THE ROLE OF COPPER AND ZINC IONS IN ALZHEIMER'S DISEASE

#### Zinc and Alzheimer's Disease

Sustained elevation of zinc levels are detrimental for neurons and have therefore been implicated in AD pathogenesis (Koh et al., 1996). A multitude of studies have analyzed overall zinc levels within AD vs. control brains with rather inconsistent outcome (Ayton et al., 2013). In consensus, however, it has been clearly demonstrated that zinc levels are increased in amyloid plaques of AD patients (Bush et al., 1994c). The elevated zinc levels concern only a small fraction of the total cortical volume, and the total tissue zinc concentrations only rise during advanced pathology (Religa et al., 2006). Closer analysis revealed zinc binding to the residues 6–28 of the Aβ peptide (Bush et al., 1993, 1994a,b,c), where zinc ions may bind to histidines 6, 13 and 14 (Damante et al., 2009) as described above. Binding of zinc ions induces rapid precipitation of Aβ into insoluble aggregates in humans. However, since the rat and mouse Aβ sequences differ from the human Aβ sequence in three amino acids, differences in the affinity for zinc binding may explain why mice and rats do not develop amyloid pathology (Bush et al., 1994c). Zinc is released by glutamatergic fibers in the cortex and hippocampus where the synaptic vesicle located zinc-transporter 3 (ZnT3) loads the metal into synaptic vesicles (Frederickson et al., 2000). In addition to the increased zinc concentrations in amyloid plaques the expression of ZnT3 is closely associated to Aβ deposition. Ten different zinc-transporters have been identified with ZnT3 showing its highest expression within the brain (Palmiter et al., 1996; Huang and Tepaamorndech, 2013). Low ZnT3 expression reduces interstitial (Lee J.-Y. et al., 2002) and vessel-wall (Friedlich et al., 2004) amyloid but increases the levels of soluble Aβ in the brains of the APP transgenic x ZnT3 KO mice (Lee J.-Y. et al., 2002). Zinc does not only affect Aβ aggregation but also influences APP processing and function. APP processing and subsequent Aβ generation is dependent on ADAM10 activity. ADAM10 requires zinc binding for proteolysis and therefore subtle changes in ADAM10 activity influence Aβ production (Lammich et al., 1999).

#### Copper and Alzheimer's Disease

In healthy brain tissue the total copper levels have been reported to increase from youth to adulthood followed by a constant decrease during aging (Maynard et al., 2002). Aged but healthy brain tissue contains approximately 80 µM copper, whereas copper levels are decreased by approx. 30%–40% in affected brain regions of AD patients (Deibel et al., 1996; Adlard and Bush, 2006; Religa et al., 2006; Magaki et al., 2007). In contrast, copper levels (and zinc levels) are enriched in extracellular amyloid plaques (Lovell et al., 1998; Dong et al., 2003; Miller et al., 2006; Leskovjan et al., 2009). Thus, either elevated levels of copper directly associated with the Aβ peptide or decreased levels of copper in amyloid plaques surrounding brain tissue might affect the course of AD. Copper binds with high affinity to Aβ and promotes its oligomerization and neurotoxicity (Atwood et al., 1998, 2004; Huang et al., 1999; Masters and Selkoe, 2012). Moreover, cellular copper deficiency promotes the amyloidogenic processing of APP leading to increased Aβ levels (Cater et al., 2008; Hung et al., 2009), whereas an elevation of intracellular copper levels promotes the non-amyloidogenic pathway and attenuates Aβ generation in cells (Borchardt et al., 1999; White et al., 2006; Donnelly et al., 2008) as well as in transgenic mice (Cherny et al., 2001; Bayer et al., 2003; Phinney et al., 2003; Adlard et al., 2008). However, the molecular mechanisms leading to elevated Aβ production in the presence of decreased cellular copper levels are currently elusive.

It has been suggested that copper induces conformational changes in APP that influence the monomer/dimer equilibrium and thus affects the proteolytic processing of APP (Kong et al., 2008; Baumkötter et al., 2014). In line, it has been shown that enhanced APP dimerization leads to reduced Aβ generation (Kienlen-Campard et al., 2008; Eggert et al., 2009). On the other hand, it has been reported that attenuated APP dimerization correlates with reduced Aβ levels (Munter et al., 2007; Kaden et al., 2008; Richter et al., 2010). Further studies will be required to solve this discrepancy.

An alternative mechanism underlying elevated Aβ production in the presence of decreased cellular copper levels might concern the influence of copper on APP trafficking (Hung et al., 2009; Acevedo et al., 2011), either by altered conformational changes (Spoerri et al., 2012) or more indirectly by regulation of APP phosphorylation (Acevedo et al., 2014). Additionally, copper modifies Aβ and accelerates its aggregation. The copperinduced Aβ oligomerization was found to contain a membranepenetrating structure (Curtain et al., 2001; Smith et al., 2006). As for other Aβ aggregates, the copper-bound Aβ oligomers complex have been shown to exhibit cytotoxic properties and vice versa general Aβ toxicity in tissue culture is partially dependent on copper (You et al., 2012).

Since more than a decade it is known, that copper may play also a detrimental role in AD due to its interaction with Aβ-peptides leading to amyloid fibrilization, its influence on Tau and GSK3β and induction of oxidative stress (Kenche and Barnham, 2011). Based on these observations Bush and Tanzi (2008) have proposed the ''Metal Hypothesis of AD''. Therefore potential therapeutic strategies have been developed either targeting Aβ copper interactions by selectively occupying the metal binding site on Aβ or through development of peptides effectively competing with Aβ-peptides for the metal ions. Such an approach has been developed e.g. for Wilson's disease (WD) to target a metal overload using metal chelators. Such chelators as desferrioxime, penicillamine and trientine have very high metal binding affinities and are hydrophilic. Therefore such chelators are inappropriate to tackle a brain disease like AD since they will not be able to cross the blood brain barrier (BBB). In contrast to these hydrophilic compounds metal protein attenuating compounds have been developed like clioquinol (CQ, 5-chloro-7-iodo-8 hydroxyquinoline), which are indeed capable to cross the BBB. Although some approaches have been promising in early clinical trials none of these compounds have clinically proven to be effective in AD (Adlard et al., 2008; Lannfelt et al., 2008). One potential drawback might be the unwanted side effects on APP copper interactions and its subsequent physiological consequences.

#### COPPER AND ZINC IN THE INTERPLAY WITH APP AT THE SYNAPSE

Based on the observation that zinc and copper bind APP, APLP1 and APLP2 with high affinities and the high abundance of these metals in the synaptic cleft during neurotransmission, the impact of copper and zinc on the trans-cellular dimerization properties of APP during cell adhesion and synaptogenesis went into focus of current research.

#### Influence of Copper on Brain Function

The transition state metals copper and zinc, as well as iron, are essential for the catalytic activity of a variety of enzymes, including amongst others the cuproenzymes cytochrome c oxidase, superoxide dismutase (SOD), ceruloplasmin, tyrosinase and dopamine β hydroxylase (Turski and Thiele, 2009). However, copper is also toxic to cells and thus, its distribution in the cell has to be tightly regulated. In neurons, the copper transporter 1 (CTR1) mediates copper import (Lutsenko et al., 2010). Interestingly, only reduced copper (I) is transported by CTR1 (Macreadie, 2008). Thus, extracellular copper (II) has to be reduced to copper (I) by membrane bound metalloreductases prior to uptake into the cell by CTR1. The assumed mammalian metalloreductase remains to be identified (Lee J. et al., 2002). Potential candidates are the Steap proteins (Ohgami et al., 2006) as well as APP since it binds with high affinity to copper and is also able to reduce copper (Multhaup et al., 1996). However, elevated APP levels cause a decrease and APP depletion an increase of intracellular copper concentrations (White et al., 1999; Maynard et al., 2002, 2006; Phinney et al., 2003; Bellingham et al., 2004; Treiber et al., 2004), arguing against a function of APP in copper influx.

Copper imported via CTR1 is delivered by specific copper chaperones directly to different cuproenzymes. For example, the copper chaperone CCS mediates transfer of copper from CTR1 to the SOD (McCord and Fridovich, 1969; Culotta et al., 1997). Delivery of copper to cuproenzymes in the secretory pathway is mediated by the copper chaperone ATOX1, which transfers copper (I) from CTR1 to the intracellular copper transporters ATP7A and ATP7B, located in different intracellular compartments (Kim et al., 2008). Besides these principal metallochaperone mediated copper transport pathways, cytoplasmic copper also binds to glutathione immediately after entry into the cell and is subsequently transferred to metallothionein (Hung et al., 2010). Due to the described mechanisms, the intracellular concentration of free copper is maintained at exceedingly low levels (Rae et al., 1999).

Copper is found all over the brain and is most abundant in the basal ganglia (Madsen and Gitlin, 2007). The data on chelatable copper concentrations in extracellular fluids and intracellular compartments is a matter of debate. In some particular neurons copper is released at the synapse (Hartter and Barnea, 1988; Brown et al., 1997), estimated to reach upon depolarization and activation of N-methyl-D-aspartate (NMDA) receptors (Rajan et al., 1976; Hartter and Barnea, 1988; Kardos et al., 1989; Peters et al., 2011; Tamano and Takeda, 2011) micromolar concentrations (Kardos et al., 1989) to approx. 15 µM (Hopt et al., 2003). Moreover, transient copper concentrations above 100 µM at the synaptic cleft have been reported (Kardos et al., 1989; Gaier et al., 2013). Consistently, the synaptic cleft is the only reported microenvironment within the brain where chelatable copper might be easily excessible (Roberts et al., 2012). Interestingly, the activation of NMDA receptors leads to relocalization of ATP7A from the trans golgi-network (TGN) to neuronal processes and thus in turn contributes to increased copper concentrations at the synapse (Schlief et al., 2005).

Loss-of-function mutations in copper transporters, such as ATP7A and ATP7B, lead to hereditary diseases, Menkes disease (MD), or WD. Loss of ATP7A function causes growth failure, brittle hair, hypopigmentation, arterial tortuosity and neuronal loss most prominent in the hippocampus and cerebellum (Okeda et al., 1991; Chelly et al., 1993; Mercer et al., 1993; Vulpe et al., 1993). The phenotypes are mostly due to consequences of specific cuproenzymes dysfunction resulting from reduced cellular copper levels (D'Ambrosi and Rossi, 2015). Loss of ATP7B, which is primarily expressed in the liver, leads to copper overload in the liver and later in the brain, possibly due to impaired copper transport at the BBB (Huster and Lutsenko, 2007; Kaler, 2011). Although copper levels are mainly increased in cerebrospinal fluid and in the basal ganglia (Südmeyer et al., 2006), WD patients show widespread neuronal cell loss and white matter abnormalities, causing symptoms that include parkinsonism, seizures and mental disorders (Gitlin, 2003).

The molecular mechanisms underlying copper induced neurodegeneration are only partially understood. Mechanism(s) discussed involve S-nitrosylation, oxidation and allosteric modulation, increased anchorage of the neurotransmitter receptors to the membrane, and modulation of neurotransmitter receptor function (Weiser and Wienrich, 1996; Kim and Macdonald, 2003; Schlief and Gitlin, 2006; El Meskini et al., 2007; Huidobro-Toro et al., 2008; Peters et al., 2011; Gaier et al., 2013).

#### Influence of Zinc on Brain Function

In contrast to copper ions there is a substantial amount of zinc loosely bound to biomolecules, designated as reactive or chelatable zinc, which is implicated in neuronal signaling. Reactive zinc is largely distributed within presynaptic vesicles in some axon terminals throughout the telencephalon and co-localizes with a subset of glutamatergic neurons (Frederickson et al., 2000). Cytosolic reactive zinc levels are in the picomolar range and are estimated to rise to micromolar levels in the synaptic cleft and in intracellular compartments, such as mitochondria, secretory vesicles and lysosomes (Sensi et al., 2009). While it is evident that zinc is released during synaptic activity, there is little consensus on the amount or duration of its existence in the synaptic cleft (Watt et al., 2010). Zinc homeostasis is mainly maintained by regulated activities of ''zinc importing'' ZnT transporters (SLC30 family), ''zinc exporting'' ZIP transporters (SLC39 family), and zinc buffering proteins, including metallothioneins (Sensi et al., 2011). At the synapse, vesicular released zinc interacts with various neuronal ion channels (NMDA, (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid, AMPA), GABA<sup>A</sup> (γ-aminobutyric acid type A) receptors), Glycine and other surface receptors, such as TrkB. Furthermore, zinc can bind to and regulate ProSAPs/Shanks scaffolding proteins of the postsynaptic density (PSD), involved in synaptic signaling. Indeed, in some cerebral areas nearly 50% of the glutamatergic synapses are actually ''glu-zinc-ergic'' (Watt et al., 2010). Therefore, zinc is considered as an important synaptic modulator, affecting neurotransmission at inhibitory as well as excitatory synapses.

## Influence of Copper/Zinc on APP Synaptic Function

Synaptic function of APP is discussed in detail in other chapters of this special issue. Briefly, loss of APP function leads to a reduced number of dendritic spines (Watt et al., 2010; Tyan et al., 2012) and to impairments in the structural plasticity (Zou et al., 2016). A possible mechanism infers an important trans-synaptic adhesion molecules like function for membrane anchored APP (Siddiqui and Craig, 2011) and major neurotrophic roles of the secreted sAPPα ectodomain (Soba et al., 2005; Bell et al., 2008; Jimenez et al., 2011; Aydin et al., 2012; Caldwell et al., 2013; Baumkötter et al., 2014). Loss of APLP2 (von Koch et al., 1997) had no consequences on brain function (Weyer et al., 2011, 2014; Midthune et al., 2012), whereas APP/APLP2 and APLP1/APLP2 double KO mice exhibit severe deficits in formation of the neuromuscular junction and die early after birth. This suggests that APP/APLPs share some overlapping functions, but also have distinct synaptic functions that are not compensated by the other family members.

As pointed out before, copper binds APP at different sites within the E1 and E2 domain, causing structural changes and altered dimerization properties and heparin binding characteristics (Baumkötter et al., 2012). Copper binding to the GFLD coincides with structural rearrangements in the heparinbinding loop region (Baumkötter et al., 2014), possibly also representing an APP dimerization interface (Kaden et al., 2008). In line, copper-induced trans-directed in vitro interaction of APP, and mutations, abolishing copper-binding to the GFLD, reduce APP synaptogenic activity in a cellular assay system (Baumkötter et al., 2014). Therefore, it appears reasonable that copper might also modulate the neurotrophic function of secreted APP forms. Thus, copper modulation in the synaptic cleft upon synaptic activity might contribute to APP transsynaptic signaling in context of synaptic maturation (**Figure 5**). Actually, a decrease of D-serine in brains of APP knockout mice was reported to contribute to synaptic deficits in aged mice (Zou et al., 2016). Most likely the decrease in D-serine is explained by a loss of function of APP in calciumdependent release of D-serine from astrocytes (Martineau et al., 2014). Thus, as both D-serine and zinc can bind and modulate NMDA receptor function in antagonistic ways, it is tempting to speculate that APP might sense changes in zinc concentrations that in turn could affect D-serine secretion in the synaptic cleft, antagonizing the antidepressant-like effects of zinc and thereby contributes to homeostasis of synaptic activity. A validation of this tempting hypothesis is however hampered by the technical limitation for the quantification of reactive copper in the synaptic cleft. Knock-in approaches testing different APP mutants might be a way to disentangle

the functional relation between copper and APP at the synapse.

Different reports suggest binding of APP, APLP1 and to a minor extent also APLP2 to zinc with affinities in the low micromolar range (Bush et al., 1993; Mayer et al., 2014, 2016) and it was shown by FRET analysis of heterologous expressed APP/APLPs that addition of zinc can induce clustering of APP, APLP1 (EC50: 10 µM), and APLP2 (EC50: 300 µM) as well as of all types of heterotypic APP/APLPs combinations (Mayer et al., 2016). Interestingly, combinations of APLP1 with APLP2 also exhibited an EC50 at about 50 µM. As reactive zinc concentrations can reach high micromolar and possibly also millimolar concentrations in the synaptic cleft, all types of oligomers (including homo- and heterotypic APLP2 containing oligomers) are likely formed also under in vivo conditions at the synapse. Consistently, addition of 50 µM zinc to cells expressing APLP1 caused a lateral concentration at cell-cell contact sites (Mayer et al., 2016), as formerly described by Soba et al. (2005) in presence of copper (Baumkötter et al., 2014). Notably, alanine mutations of one of four histidines involved in zinc binding at the M3 site (H430/H433/H450/H452) did not abolish zinc-induced oligomerization and only lowered the affinity about two fold. This suggests that APP and APLP2 transdirected dimerization might also be affected by zinc in a similar way, possibly in interplay with heparin binding and APP/APLPs dimerization/oligomerization (Bush et al., 1994a).

Despite some major gaps in our understanding of APP/APLPs synaptic function, the current data as presented in this review article strongly suggest that activity-dependent changes in zinc and copper concentrations in the synaptic cleft can be sensed by the APP/APLP family. In turn, they seem to modulate neurotransmission by different pathways including neurotrophic activity of sAPPα or trans-cellular dimerization/signaling. However, one major gap in our current understanding, especially in respect to the function of copper, is the limitation of available sensitive sensors, allowing determination of local transient changes in copper concentration. In this regard, live-cell optical imaging with fluorescent sensors offers a potentially powerful approach for interrogating aspects of labile copper accumulation, speciation, trafficking, and redox function in living systems at the molecular level. Such reagents have greatly facilitated the study of calcium and zinc in cell biology, but analogs tools for cellular copper remain underdeveloped (Zeng et al., 2006; Dean et al., 2012). Therefore, the most promising way might actually be, to use mutant APP impaired in copper and/or zinc binding in different cellular assays, allowing to estimate the

#### REFERENCES


pathophysiological impact of copper and zinc on APP function and its role in AD.

#### AUTHOR CONTRIBUTIONS

KW, AA, CUP and SK wrote this review article.

#### ACKNOWLEDGMENTS

Research in our laboratories was funded by the Deutsche Forschungsgemeinschaft (FOR1332, to CUP, KW and SK) and the Stiftung Rheinland-Pfalz für Innovation (to CUP and SK). KW acknowledges generous support by Irmgard Sinning at the Heidelberg University Biochemistry Center (BZH).


that encodes a potential heavy metal binding protein. Nat. Genet. 3, 14–19. doi: 10.1038/ng0193-14


cerebral amyloid angiopathy in an animal model of Alzheimer's disease. J. Neurosci. 24, 3453–3459. doi: 10.1523/JNEUROSCI.0297-04.2004


amyloid precursor-like protein 1 (APLP1) oligomerization. J. Biol. Chem. 289, 19019–19030. doi: 10.1074/jbc.M114.570382


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Wild, August, Pietrzik and Kins. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Fe65-PTB2 Dimerization Mimics Fe65-APP Interaction

Lukas P. Feilen1† , Kevin Haubrich1,2† , Paul Strecker <sup>3</sup> , Sabine Probst <sup>4</sup> , Simone Eggert <sup>3</sup> , Gunter Stier <sup>1</sup> , Irmgard Sinning<sup>1</sup> , Uwe Konietzko<sup>4</sup> , Stefan Kins <sup>3</sup> , Bernd Simon<sup>2</sup> and Klemens Wild<sup>1</sup> \*

<sup>1</sup>Heidelberg University Biochemistry Center (BZH), University of Heidelberg, Heidelberg, Germany, <sup>2</sup>European Molecular Biology Laboratory (EMBL), Structural and Computational Biology, Heidelberg, Germany, <sup>3</sup>Division of Human Biology and Human Genetics, University of Kaiserslautern, Kaiserslautern, Germany, <sup>4</sup> Institute for Regenerative Medicine (IREM), University of Zurich, Zurich, Switzerland

Physiological function and pathology of the Alzheimer's disease causing amyloid precursor protein (APP) are correlated with its cytosolic adaptor Fe65 encompassing a WW and two phosphotyrosine-binding domains (PTBs). The C-terminal Fe65-PTB2 binds a large portion of the APP intracellular domain (AICD) including the GYENPTY internalization sequence fingerprint. AICD binding to Fe65-PTB2 opens an intramolecular interaction causing a structural change and altering Fe65 activity. Here we show that in the absence of the AICD, Fe65-PTB2 forms a homodimer in solution and determine its crystal structure at 2.6 Å resolution. Dimerization involves the unwinding of a C-terminal α-helix that mimics binding of the AICD internalization sequence, thus shielding the hydrophobic binding pocket. Specific dimer formation is validated by nuclear magnetic resonance (NMR) techniques and cell-based analyses reveal that Fe65-PTB2 together with the WW domain are necessary and sufficient for dimerization. Together, our data demonstrate that Fe65 dimerizes via its APP interaction site, suggesting that besides intra- also intermolecular interactions between Fe65 molecules contribute to homeostatic regulation of APP mediated signaling.

#### Edited by:

Thomas Deller, Goethe University Frankfurt, Germany

#### Reviewed by:

Tommaso Russo, University of Naples Federico II, Italy Kwok-Fai Lau, The Chinese University of Hong Kong, Hong Kong

#### \*Correspondence:

Klemens Wild klemens.wild@bzh.uni-heidelberg.de

†These authors have contributed equally to this work.

Received: 22 February 2017 Accepted: 25 April 2017 Published: 11 May 2017

#### Citation:

Feilen LP, Haubrich K, Strecker P, Probst S, Eggert S, Stier G, Sinning I, Konietzko U, Kins S, Simon B and Wild K (2017) Fe65-PTB2 Dimerization Mimics Fe65-APP Interaction. Front. Mol. Neurosci. 10:140. doi: 10.3389/fnmol.2017.00140 Keywords: Fe65, phosphotyrosine-binding domain (PTB), homodimerization, amyloid precursor protein (APP), AICD, Alzheimer's disease

# INTRODUCTION

The Fe65s (Fe65, Fe65L1 and Fe65L2) are a family of conserved eukaryotic adaptor proteins involved in a variety of biological processes (Russo et al., 1998; McLoughlin and Miller, 2008; Minopoli et al., 2012). Special attention has been given to the brain-enriched Fe65 as its expression pattern parallels the amyloid precursor protein (APP; Guenette et al., 2006). Accordingly, the physiological functions of the two proteins are interdependent and knockout studies resulted in markedly similar phenotypes (Zambrano et al., 2002; Guenette et al., 2006; Strecker et al., 2016). APP is a single-spanning type-1 transmembrane protein (Coburger et al., 2014) with numerous neuronal functions especially in the developing brain (Müller and Zheng, 2012). Sequential regulated proteolysis of APP by different secretases (Lichtenthaler et al., 2011; Haass et al., 2012) results in multiple break-down products including soluble ectodomains, the Aβ-peptides forming the amyloids in Alzheimer's disease, and the APP intracellular domain (AICD) that is released into the cytosol (Selkoe and Hardy, 2016). The AICD is an intrinsically disordered peptide of 47 residues (Ramelot et al., 2000) and includes the GYENPTY internalization sequence that besides Fe65 binds also to many other adaptor proteins (Russo et al., 1998) with a variety of physiological functions and pathological implications (Müller et al., 2008; Pardossi-Piquard and Checler, 2012).

Fe65 determines localization and nuclear signaling of APP and modulates APP processing and Aβ-peptide generation (McLoughlin and Miller, 2008). Fe65 is a multidomain protein including an N-terminal α-helical domain and three proteinprotein interaction modules: a WW domain and two consecutive C-terminal phosphotyrosine-binding (PTB) domains (**Figure 1A**). The WW domain binds to the Mena protein (Ermekova et al., 1997) involved in actin dynamics and cell motility thus regulating neuronal positioning in the developing brain. Fe65-PTB1 has been mainly implicated as central module of a ternary AICD/Fe65/Tip60 complex responsible for transcriptional activity of APP (Cao and Südhof, 2001), with the histone acetyltransferase Tip60 being a key regulator of genome expression and stability. Further data suggested Fe65 to provide a dominant role for nuclear signaling (Yang et al., 2006). The analysis of the AICD/Fe65/Tip60 interaction revealed that only membrane-bound AICD in context of APP and not on its own is a potent transactivator of transcription (Cao and Südhof, 2004). The distinction had been interpreted by a membrane association dependent transition of Fe65 from a closed to an open and active conformation, involving its WW and PTB2 domain.

Most attention has been given to Fe65-PTB2 as it directly interacts with the AICD and thus functionally joins the two proteins (Borg et al., 1996). The interaction is phosphotyrosineindependent and untypically for PTB-interactions (Uhlik et al., 2005) includes an extended interface of 28 AICD residues including two α-helices (αN and αC; **Figure 1B**; Radzimanowski et al., 2008c). The GYENPTY internalization sequence is recognized in a rather hydrophobic crevice with GYE involved in a PTB-typical β-augmentation manner and NPTY starting helix αC and placing the canonical PTB-relevant tyrosine in its binding pocket. Unique for the AICD/Fe65- PTB2 complex is the N-terminal binding helix αN within AICD that is capped by the T668PEE-motif. Phosphorylation of threonine T<sup>668</sup> regulates the interaction and has been identified as sensitive checkpoint switching between physiological and pathological APP related pathways (Ando et al., 2001).

Here we present structural and functional data on Fe65-PTB2 revealing the domain as flexible module forming a homodimer in vitro and ex vivo in the absence of APP. Dimerization mimics the AICD-interaction and at the same time shields the hydrophobic crevice. The interaction competes with AICD binding and therefore with APP signaling depending on its cellular context.

#### MATERIALS AND METHODS

#### Protein Production and Characterization for X-ray Structure Analysis

Human Fe65-PTB2 (residues 534-667; UniPROTKB: APBB1\_HUMAN, O00213) was expressed and purified for crystallization as described previously (Radzimanowski et al., 2008a). To avoid precipitation of concentrated and pure Fe65-

FIGURE 1 | Fe65 and amyloid precursor protein (APP). (A) Domain architecture of human Fe65 with numbering of domain boundaries. (B) Schematic for Fe65-mediated APP-signaling by the APP intracellular domain (AICD)/Fe65-phosphotyrosine-binding domains 2 (PTB2) complex at the cell membrane. Structural details for the interaction are depicted as follows: αN and αC: AICD helices; T and Y: AICD sequence fingerprints (T: T668PEE, Y: N684PTY) as part of AICD helices, GYE: AICD region involved in β-augmentation with Fe65-PTB2. APP-cleavage sites by secretases are indicated by Greek symbols. (C) X-ray structure of the Fe65-PTB2 dimer of dimers. The dimer is constituted by a "complementing" subunit (blue) with a transition of the C-terminus to strand β ct (dark blue), while the "accommodating" subunit (yellow) contains the entire helix α3 (orange). The second dimer symmetrically attached by β-augmentation is shown with gray subunits. The central disulfide bond connecting the dimer of dimers is shown in magenta. (D) Close-up on the C-terminal Fe65-transition. According regions (L656-D663) of the complementing (dark blue, β ct) and accommodating (orange, α3) subunits are given with side chains and numbering.

PTB2, 5% (v/v) glycerol was added in the final size exclusion chromatography (SEC) buffer. Multi angle light scattering (MALS) was performed in line with SEC and monitored by refractive index measurements (Wyatt technology). The protein (5–20 mg/mL) was crystallized within 3 days in an automated platform at 18◦C by mixing equal amounts (200 nL) of protein solution and a reservoir containing 1.6 M ammonium sulfate, 0.08 M sodium acetate pH 4.6 and 20% (v/v) glycerol in a sitting drop setup. The high glycerol concentration allowed direct flash-cooling in liquid nitrogen for X-ray structure analysis. X-ray data collection was done at beamline ID29 of the European Synchrotron Radiation Facility (ESRF). Data was integrated with program XDS (Kabsch, 2010) and scaled and merged with program AIMLESS (Evans and Murshudov, 2013) from the CCP4-package (Winn et al., 2011). The structure was solved by the Molecular Replacement method (PHENIX package; Adams et al., 2010) using a monomeric Fe65-PTB2 molecule taken out of the Fe65-PTB2/AICD complex (PDB entry: 3dxc). Iterative model building, refinement and validation were performed with programs COOT (Emsley et al., 2010) and PHENIX. All structural figures were prepared using PyMOL (Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC)<sup>1</sup> .

#### NMR Measurements

Sequences for wildtype (wt) Fe65-PTB2 and the C633E mutant were cloned into a pETHis vector using NcoI/BamHI restriction enzymes. The proteins were expressed in E. coli BL21(DE3) Rosetta pLysS grown in LB media or for <sup>15</sup>N- or <sup>13</sup>C/15Nlabeling in M9 media by induction with 0.5 mM IPTG overnight at 22◦C. Pellets were lysed by sonication in 20 mM Tris pH 8.0, 150 mM NaCl, 0.2% (v/v) Nonidet P-50 and 2 mM DTT, and the proteins purified by nickel affinity chromatography. Spin-labeling of the C633E mutant was performed by incubation with a five-fold molar excess of 3-(2-Iodoacetamido)-proxyl free radical dissolved in methanol over night at 4◦C. Free spin-label was removed by buffer exchange or SEC into 20 mM Na2HPO<sup>4</sup> pH 6.5 and 150 mM NaCl. Nuclear magnetic resonance (NMR) spectra were acquired on Bruker Avance III 600 and 800 spectrometers with a cryogenic triple resonance probe and a Bruker Avance III 700 with a triple resonance probe at concentrations of 0.1–0.5 mM in the same buffer at 300 K. Data where processed with NMRPipe (Delaglio et al., 1995) and analyzed using NMRView (Johnson and Blevins, 1994). The transfer of backbone assignment from the wt protein (Dietl et al., 2014) was confirmed by analyzing HNCA, HNCACB and CBCA(CO)NH spectra of the C633E mutant. Chemical shift based secondary structure predictions and structure based chemical shift predictions where done using the programs TALOS+ (Shen et al., 2009) and SPARTA+ (Shen and Bax, 2010). Model-free Liparai-Szabo parameters derived from the <sup>15</sup>N relaxation data of the C633E mutant were analyzed and compared to hydrodynamic diffusion tensors using the programs ROTDIF and ELM (Berlin et al., 2013). Paramagnetic Relaxation Enhancements where measured and analyzed as described (Simon et al., 2010). SAXS measurements were carried out at the BM29 beamline at ESRF in Grenoble (Pernot et al., 2013). Samples were measured in NMR buffer (20 mM Na2HPO<sup>4</sup> pH 6.5, 150 mM NaCl, 2 mM DTT) at concentrations between 0.25 and 6 mg/mL, a temperature of 300 K and a wavelength of 1 Å. Data was processed using the ATSAS suite (Petoukhov et al., 2012).

# Pull-Down Experiments

The coding sequence for full-length human Fe65 was inserted into the pUKBK vector system (Kohli et al., 2012) by standard cloning techniques in order to attach either a streptavidin-binding peptide (SBP) together with a myc-tag or a mCherry (mChe)-tag to the protein N-terminus. Thereof, the following deletion constructs were generated: ∆PTB2 (Fe65(1- 532)-(665-710)), ∆WW (Fe65(1-253)-(286-710)), and ∆WW- ∆PTB2 (Fe65(1-253)-(286-532)-(665-710)). After transfection with Lipofectamine 2000 (ThermoFisher Scientific) for 22 h, HEK293 cells were lysed in homogenization buffer consisting of 140 mM KCl, 20 mM HEPES pH 7.2, 10 mM NaCl, 5% (v/v) glycerol, 2 mM MgSO4, 1% (v/v) Triton-X100, 2 mM DTT, EDTA-free Protease-Inhibitor Cocktail (Roche), and 2 mM Phenantrolen. Pull-down assays were performed with Dynabeads<sup>r</sup> M-280 Streptavidin (ThermoFisher Scientific) and bound proteins were eluted with biotin and further separated on NovexTM 10%–20% Tricine Protein Gels. Antibodies used for detection were the c-myc antibody (1:1000, 9E10, Roche), mCherry antibody (1:1000, 5F8, Chromotek), and GAPDH antibody (1:5000, Meridian Life Science). ECL detection was performed with the ImageQuant LAS 4000 (GE Healthcare Life Sciences). Quantification was done on the latest exposure before saturation of the brightest band on the blot, using the ImageQuant TL software.

#### Co-Immunoprecipitation

Co-Immunoprecipitation (Co-IP) experiments were performed as described before (Baumkotter et al., 2014). Briefly, HEK293 cells were transfected with pcDNA3.1 constructs containing FE65-HA, FE65-Flag or APP-myc using JetPRIME (Polyplus transfection). Twenty to twenty-two hours after transfection cells were harvested and lysed in 150 mM NaCl2, 50 mM Tris/HCl pH 7.5, 2 mM EDTA, 1% (v/v) NP40 and freshly added Complete Protease Inhibitor mix (Roche) for 20 min on ice. After centrifugation at 16,000× g for 10 min the supernatant was pre-cleared with protein A Sepharose beads (GE Healthcare). Then the supernatant was incubated over night with anti-HA agarose beads (Roche) to allow binding of HA-tagged FE65. After washing bound proteins were eluted by denaturation with SDS sample buffer at 95◦C. Samples were separated on 8% Tris/glycine gels and probed via immunoblotting for HA-, Flagand myc-tagged constructs.

#### Subcellular Fractionation

Subcellular fractionation was performed according to Abcams subcellular fractionation protocol. HEK293 cells were transfected as described before. Twenty to twenty-two hours post transfection cells were resuspended in 1 mL of fractionation

<sup>1</sup>http://www.pymol.org

buffer (250 mM Sucrose, 20 mM HEPES, 10 mM KCl, 2 mM MgCl2, 1 mM EDTA and 1 mM EGTA with freshly added Complete Protease Inhibitor mix (Roche)) and passed 10 times through a 27 gauge needle. After differential centrifugation at 720× g and 10,000× g for 5 min and 100,000× g for 1 h the supernatant (cytosolic fraction) was transferred and kept on ice for further analysis. The sediment (membrane fraction) was resuspended by pipetting and pass through 10 times a 27-gauge needle. Protein concentration of membrane and cytosolic fraction was determined using the BCA assay (Sigma).

#### Blue Native Gel Electrophoresis

For Blue Native Gel analysis 100 µg protein of the cytosolic and membrane fraction was diluted in 1.5 M amino caproic acid, 0.05 M Bis-Tris, pH 7, 1.25% (w/v) dodecyl maltosidase and 5% (w/v) Coomassie Brilliant Blue G250, as described in detail before (Eggert et al., 2009). Afterwards, samples were separated on a 4%–15% (w/v) Tris-HCl gel (Biorad), transferred on a PVDF membrane and probed via immunoblotting for HA- and myc-tagged constructs.

# RESULTS

#### Fe65-PTB2 Dimerization

Recombinantly expressed human Fe65-PTB2 (residues 534- 667) is difficult to purify as it precipitates at higher protein concentrations in the mg/mL range. Instability is related to the exposure of a hydrophobic crevice that corresponds to the AICD binding site and complex formation dramatically enhances solubility about a 100-fold (Radzimanowski et al., 2008c). When purified via SEC, Fe65-PTB2 partitions in monomeric, dimeric and tetrameric species as validated by multi-angle light scattering (MALS) and on SDS-PAGE the protein appears as detergent-resistant dimer (Supplementary Figure S1). Unspecific aggregation of Fe65-PTB2 at concentrations in the mg/mL range can be prevented by the addition of glycerol and we subsequently crystallized the domain and solved its crystal structure by molecular replacement at 2.6 Å resolution (**Table 1**).

Fe65-PTB2 crystallizes as dimer of dimers with a continuous central β-sheet (**Figure 1C**). Dimerization occurs via a structural transition of the C-terminal α-helix α3 within one Fe65-PTB2 subunit (the ''complementing'' subunit) in respect to the conformation as seen in the previously solved AICD/Fe65- PTB2 complex (rmsd of 0.85 Å for 123 Cα-atoms; **Figure 1D**, Supplementary Figure S2; Radzimanowski et al., 2008c). The last two helical turns dissolve (starting at L656) and adopt an extended β-conformation that complements the ''accommodating'' subunit in trans (dimer interface: 585 Å<sup>2</sup> ). The interface is classified just about stable (Krissinel and Henrick, 2007). The newly formed β-strand (defined here as β ct) quasisymmetrically mediates also the dimer of dimer contact with the tetrameric assembly being stabilized by a disulfide bridge between respective cysteine (C661) residues (**Figure 1C**).

#### TABLE 1 | Data collection and refinement statistics.


∗ for 5% of all data. Statistics for the highest-resolution shell are shown in parentheses.

# Fe65-PTB2 Dimer Structure in Solution

Having solved the crystal structure of Fe65-PTB2, we had to make sure that the observed interactions did not represent a crystallographic artifact and are also present in solution. We therefore first performed concentration dependent (0.25–6 mg/mL) small angle X-ray scattering (SAXS) measurements under reducing conditions to avoid the covalent and likely non-physiological cysteine bridge. The data showed a sharp increase in intensity at very small scattering angles that becomes more pronounced with higher concentrations and thus confirming the observation of the presence of aggregation (**Figure 2A**). Accordingly, the deduced molecular masses showed a strong concentration dependence that reflects the monomerdimer transition. Calculating the theoretical scattering curves of the monomer, dimer and tetramer structures and fitting them against the experimental data, revealed the best fit to correspond to the crystallographic dimer (Supplementary Figure S3A), which holds true for the whole concentration range and also when the data are interpolated to zero concentration. Calculations of monomer and dimer content based on fitting linear combinations of two structures range from more than 20% of monomer to almost exclusively dimer at higher concentrations, but should be taken as rough estimates with the given data quality and the insecurity of especially the dimer structural model. In accordance with these data, a dissociation constant could be estimated by preliminary isothermal titration calorimetry (ITC) measurements to be in the low micromolar range (data not shown).

In order to obtain high resolution structural information for Fe65-PTB2 dimerization in solution, we performed an

subunit is shown in color if the average intensity ratio of the observed <sup>1</sup>H-15N peak in the paramagnetic and diamagnetic NMR spectra of the corresponding residue and its two neighbors is smaller than 0.7 and thus identifies amino acids that are close to the paramagnetic center. Residues in yellow are bleached for molecules that are simultaneously <sup>15</sup>N and nitroxide labeled, while residues in red are also bleached when exclusively <sup>15</sup>N and nitroxide labeled proteins are mixed. The spin-label carrying C<sup>661</sup> residues are highlighted for the monomer (on the C-terminal α-helix) and the crystallographic dimer (on the extra β-strand).

extended NMR characterization. Overall, we observe a high consistency between the backbone chemical shift data and the dihedral angles observed in the crystal structure (**Figures 2B**, Supplementary Figure S3B). For the C-terminus, the chemical shifts predict the existence of a helix until Y<sup>658</sup> and indicate an increase in backbone flexibility starting from M655. Interestingly, the observed secondary Cα-Cβ chemical shift differences for the C-terminus are in between the values predicted for the accommodating (long C-terminal α-helix) and complementing (β-sheet augmentation) subunits of the crystal structure.

A more detailed picture for the internal dynamics and dimerization was obtained by the analysis of <sup>15</sup>N relaxation data. The average ratio of transverse and longitudinal relaxation rates measured at 300 K indicated a rotational correlation time τ<sup>c</sup> of 10.6 ns. This value compares to 8.9 ns (for complementing subunit) and 15.7 ns (for crystallographic dimer) as calculated from the coordinates. Assuming a rapid exchange between rigid monomers and dimers the experimental value would suggest a high percentage of monomers in solution. However, since the intermolecular interaction is mediated by the flexible C-terminus, we envision a dimer with a rather flexible connection between the monomers and thus with faster effective rotational correlation time than expected for a rigid dimer. This model is supported by the observation of quickly reducing backbone order parameters S<sup>2</sup> for the C-terminal residues starting from Y<sup>658</sup> (**Figure 2C**, Supplementary Figure S3C).

To further characterize the oligomerization in solution, we introduced nitroxide spin-labels covalently attached to cysteine residues to measure paramagnetic relaxation enhancements (PREs). The presence of the electron spin leads to signal broadening of nuclear spins in spatial proximity (less than ∼20 Å) to the nitroxide and can assist NMR protein structure determination. Due to the r <sup>−</sup><sup>6</sup> dependence of the induced relaxation, the signal bleaching can also be used to structurally and dynamically characterize specific encounter complexes (Clore, 2015). Since Fe65-PTB2 contains six native cysteines and the evaluation of the experiment requires a single spin-label attached to each molecule, we performed an extensive mutational analysis to determine the accessibility and structural importance of all native cysteines. In the end, only two cysteines (C<sup>633</sup> and C <sup>661</sup>) were solvent exposed to be efficiently paramagnetically labeled. Of particular interest are the PRE results for the C633E mutant, which positions the spin-label solely on C<sup>661</sup> at the C-terminus in the center of the oligomerization region. We measured the intensity ratios in <sup>15</sup>N-1H heteronuclear correlation spectra (HSQC) between the paramagnetic and diamagnetic state of the molecule (Supplementary Figures S3D,E). Due to the instability and precipitation of the molecule in solution during the measurements, a number of intensity ratios larger than one for residues that are not in proximity of the nitroxide were observed. Therefore, and because of the difficulties to accurately model the spin-label being attached to a flexible C-terminus, we resign from a detailed quantitative analysis of the data. A qualitative picture however can be obtained by plotting the experimental Ipara/Idia ratios onto the X-ray structure. The lowest ratios are observed for residues in the C-terminal helix and the loops and secondary structure elements in the vicinity of the C-terminus. To disentangle intra- and inter-molecular contributions, we performed a second experiment with a mixed sample of <sup>14</sup>N-paramagnetic and <sup>15</sup>N-diamagnetic molecules. The observed PREs are exclusively due to inter-molecular proximity of the radical. Bleaching was observed for patches adjacent to the hydrophobic crevice and on surface loops consistent with the presence of the dimer and tetrameric species in solution (**Figure 2D**). Strongest bleaching with Ipara/Idia ratios below 0.3 occurred for residues L<sup>609</sup> and F <sup>611</sup> that also are in closest contact within the crystallographic dimer and for C<sup>661</sup> itself that also bridges the observed dimer of dimers.

Taken all NMR measurements together, a transient dimer formation as seen in the crystal structure is validated as homotypic interaction in solution. The tetrameric and covalent linkage of two dimers seems to be favored only under high concentrations and oxidizing conditions as seen in the crystallographic array.

#### Fe65-PTB2 Mimics the AICD

The central part within the AICD/Fe65-PTB2 interface has been previously shown to be constituted by antiparallel β-augmentation of the PTB domain with the G681YE sequence

FIGURE 3 | Fe65 mimicry of AICD binding. (A) Left: X-ray structure of the AICD/Fe65-PTB2 complex (PDB: 3dxc; Radzimanowski et al., 2008c). The central interacting part of the AICD is detailed: G681YE in dark blue, N684PTY in cyan. Right: same view and coloring of the Fe65 dimer with the AICD replaced by the accommodating subunit. The geometry and type of interactions mimic the AICD/Fe65-PTB2 complex. Matching sequences are given in the alignment. Coloring as in Figure 1D. (B) Surface potential (±5 kBT/e; blue: positive, red: negative) of the Fe65-PTB2 dimer. Dimerization results in an extended positively charged groove with tightly bound sulfate ions originating from crystallization. (C) Coordination and electron densities (2mFo-DFc, 1.0 σ) for the centrally bound sulfate ions (magenta). Binding occurs next to strand β ct and the N-terminus of Fe65-PTB2 (green). Same orientation as in Panel B as indicated by the red rectangle.

fingerprint of the AICD (APP695 numbering; **Figure 3A**, left panel; Radzimanowski et al., 2008c). The glycine presents an essential hinge that places the N-terminally located helix αN of the AICD almost perpendicular to the C-terminal helix α3 of Fe65-PTB2 whereas the tyrosine residue (Y682) is imbedded in a hydrophobic pocket formed by residues of helix α3. Glutamate E<sup>683</sup> is involved in an intramolecular salt bridge with a lysine (K688) following the NPTY<sup>687</sup> sequence. In the crystal structure of the Fe65-PTB2 dimer, the induced strand β ct with the C661LD sequence directly matches to the AICD strand (**Figure 3A**, right panel). Cysteine C<sup>661</sup> occupies the glycine position although due to the restrained main chain flexibility it does not introduce a similar hinge. The hydrophobic leucine L <sup>662</sup> superposes with the tyrosine and aspartate D<sup>663</sup> forms an AICD-equivalent intramolecular salt-bridge with arginine R<sup>665</sup> . Thus, the complementing Fe65-PTB2 mimics the interacting AICD in space and charge. Of note, the accommodating Fe65-PTB2 subunit does not show the structural transition. The hydrophobic crevice of the complementing subunit is therefore still available, however, the adjacent C-terminal binding site for helix αC of the AICD is destroyed by the helical unwinding and the respective space is occupied by the accommodating subunit (Supplementary Figure S2C). In summary, Fe65-PTB2 dimerization results in a structural change that blocks the AICD binding site either fundamentally in the accommodating subunit or partially in the complementing subunit.

# A Basic Cluster Next to the Dimerization Site

In order to evaluate changes of the surface properties due to dimerization we calculated surface charge potentials. The analysis revealed a pronounced positively charged patch (R<sup>605</sup> , R <sup>607</sup>, R657, K<sup>660</sup> and R665) in the center of the dimer directly located at the transition site of the C-terminal helix (**Figures 3B,C**). Due to its location, the shape of the patch differs between an extended (complementing subunit with extended strand β ct) and a condensed form (accommodating subunit with folded C-terminal helix; Supplementary Figure S4). Fe65-PTB2 was crystallized in sulfate conditions and we find sulfate ions bound to both the condensed and extended patches. Most strikingly, in the elongated patch next to the dimer interface we find three adjacent sulfate ions (**Figures 3B,C**). The spatial arrangement of the ions perfectly match to the three phosphoryl-groups of the head-group (IP3) of phosphatidyl-inositol-4,5-bisphosphate (PIP2; Supplementary Figure S4), which has been found earlier to bind to Fe65 by liposome flotation assays (Cao and Südhof, 2004). PIP2-binding is a recurrent and functionally important feature of many PTB domains due to their juxtamembrane location and always occurs in similar basic clusters (Uhlik et al., 2005). Of note, also the N-terminus of Fe65-PTB2, and thus the PTB1-PTB2 linker region implicated in the intramolecular closure by binding to the WW-domain (Cao and Südhof, 2004), locates next to the basic cluster.

# Fe65 Dimerization In Vivo

All structural studies have been performed in vitro with isolated Fe65-PTB2 at rather high protein concentrations and they do not necessarily reflect the in vivo situation in context of the full-length protein and the cellular environment. We therefore set out to determine its relevance by testing Fe65 dimerization in the cellular context. HEK293 cells expressing Fe65 full-length protein fused N-terminally to a SBP and deletion variants missing either the WW domain (Fe65∆WW), the PTB2 (Fe65∆PTB2) domain or both (Fe65∆WW/PTB2; **Figure 4**), were subjected

FIGURE 4 | Deletion of the PTB2 domain impairs Fe65 dimerization in cells. (A) HEK293 cells expressing streptavidin-binding peptide (SBP)-myc-Fe65 (SBP-Fe65) and mCherry-Fe65 (mChe-Fe65) as wildtype (wt) or deletion constructs were subjected to pulldown analyses. Total cell lysates (L) and eluates (E) were analyzed with antibodies against myc, mCherry and GAPDH. (B) Levels of co-precipitated mChe-Fe65 constructs in the eluate are significantly reduced in both constructs harboring a deletion of the PTB2 domain. (C) Confirmation of similar levels of mChe-Fe65 in the lysate. (D) Similar amounts of SBP-Fe65 are eluted in all experiments. No GAPDH signal is seen in the eluate (not shown). Mean ± SEM of n = 3 are shown (∗p < 0.05, ∗∗p < 0.01, t-test).

to streptavidin-based isolation. Indeed, all precipitates of SBP-tagged Fe65 also recovered mCherry-tagged Fe65 in the eluate, and thus proving Fe65 dimerization in a cellular context (**Figures 4A,B**). Deletion of exclusively the PTB2 domain resulted in a strong reduction of the interaction with full-length Fe65 and the same was true for a Fe65 deletion mutant lacking the PTB2 and WW domains. In contrast, deletion of solely the WW domain did not significantly interfere with Fe65 dimerization. The negative control of the input of SBP- and mCherry-tagged Fe65 validates the dimerization event (**Figures 4C,D**). These results show that Fe65 dimerization takes place in a cellular environment and implement the PTB2 domain being mainly responsible for dimer formation.

Furthermore, we tested via Blue Native Gel analyses, if Fe65 migrates as a dimer. The analyses revealed a single band with a molecular weight of about 200 kDa pointing indeed to a full-length Fe65 dimer (**Figure 5A**). In HEK cells, Fe65 partitions into a major cytosolic and a minor membrane-bound fraction,

whereas co-expression of APP caused a strong repartitioning of Fe65 towards the membrane fraction. Co-expression of APP did not alter electromobility of Fe65 in the native gel analysis. However, as APP and Fe65 have very similar molecular weights, the native gel analysis does not allow for differentiating homotypic from heterotypic complexes. In the next step we tested, if APP co-expression might affect Fe65 dimerization. For this purpose, we analyzed HEK293 cells expressing Flag- and HA-tagged Fe65 and myc-tagged APP and performed co-immunoprecipitation studies with anti-HA antibodies from total cell extracts (**Figure 5B**). For control we used cells expressing Flag-Fe65 and myc-APP only. The analyses revealed interactions of HA-Fe65 with both Flag-Fe65 and myc-APP. No clear reduction was observed for HA-Fe65/Flag-Fe65 interaction upon co-expression with myc-APP. However, these data again did not allow for differentiating between a trimeric complex of APP with dimeric Fe65 and two separate dimeric complexes either consisting of HAand Flag-tagged Fe65 or HA-Fe65 and myc-APP. Therefore, we repeated the Co-IP of Fe65-HA, Fe65-Flag and mycAPP from HEK293 cell extracts from the membrane fraction. In this fraction only minor Fe65 amounts are present and we could not detect any Fe65 dimer. Upon co-expression of APP, Fe65 was shifted into the membrane fraction as expected from the known and strong APP-Fe65 interaction (Radzimanowski et al., 2008c). Interestingly, under these conditions we succeeded to precipitate the two differently tagged Fe65 molecules (HA and Flag) and APP (**Figure 5B**). These data show that Fe65 at least to some extend can still dimerize in presence of APP and even a trimeric species might be formed.

#### DISCUSSION

Fe65 is a versatile protein-adaptor with an interactome list of increasing size and complexity. It participates in various neuronal processes, including neurogenesis, neuronal migration and positioning, neurite outgrowth, synapse formation and plasticity, and finally in learning and memory (McLoughlin and Miller, 2008; Minopoli et al., 2012; Strecker et al., 2016). The most studied function concerns the gene transactivation complex together with APP and the histone acetyltransferase Tip60, although the pathway that at least in parts parallels Notch signaling and its gene targets are far from being understood (Cao and Südhof, 2001; Pardossi-Piquard and Checler, 2012). However, it is possible that in ageing and sporadic Alzheimer's disease there is an increase of nuclear signaling concomitant with amyloidogenic processing of APP and the accumulation of the Aβ-peptide (Fukumoto et al., 2002; Yang et al., 2003; Goodger et al., 2009). Inline, it was found that an alternate splice variant of Fe65 (Fe65a2 isoform) lacking the last exon confers resistance against very late onset of AD (Hu et al., 2002). The exon codes for residues starting at the C-terminal end of helix α3 of Fe65-PTB2 and therefore is impaired in AICD binding. Soon after the first description of the signaling pathway, it was found that complex formation with APP includes a membrane-associated initiation process that enables Fe65 to act as transactivator of transcription once the AICD is cleaved-off (Cao and Südhof, 2004). This process was associated with an opening of Fe65 by the release of a WW-PTB2 domain interaction eventually triggered by a membrane-associated factor.

The AICD/Fe65-PTB2 contact is of hydrophobic character and recombinant expressed Fe65-PTB2 is aggregation prone (Radzimanowski et al., 2008a). Here we show by X-ray crystallography and extended NMR measurements including spin-labeling PRE techniques that homotypic dimerization of the Fe65-PTB2 domain mimics AICD binding and effectively shields the hydrophobic surface. The shielding may reflect the physiological need of chaperoning this surface in case the binding partner is not present or binding is to be prevented for functional reasons. This intermolecular protection does not contradict the predicted intramolecular WW-PTB2 interaction, which involves the PTB1-PTB2 boundary and could occur at the same time inhibiting downstream signaling via the WW-domain (Cao and Südhof, 2004). Interestingly, the interaction of the Fe65 WW domain and full length Fe65 is inhibited by excess of AICD, indicating that AICD binding to the PTB2 domain affects the interaction of PTB1-PTB2 boundary with the WW domain (Cao and Südhof, 2004). Homotypic dimerization might also impact pathological pathways including the AICD/Fe65 interaction. Of note, the Fe65a2 isoform conferring very late onset AD resistance (Hu et al., 2002) lacks the dimerization sequence and thus excludes the self-association. However, all interactions of Fe65 distinct to the dimerization site and independent of APP binding are likely to be unaffected by the homotypic Fe65-PTB2 interaction.

We demonstrate by co-immunoprecipitation assays performed in transfected HEK293 cells in the presence of Fe65/APP overexpression that at the membrane a Fe65-dimer complex still co-exists with APP, which could correlate with the previously described Fe65-activating state of a ''primed complex'' (Cao and Südhof, 2004). While there is no indication yet for an additional membrane-associated protein factor, activation seems to be guided by the lipid PIP2, which plays an important role in many endocytic events. PIP2-binding most likely occurs via the epitope identified by multiple sulfate ion binding in our dimeric Fe65 crystal structure and/or via Fe65-PTB1 (Radzimanowski et al., 2008b). As the epitope is in direct proximity to the dimer interface, membrane association could also have a direct influence on the monomer-dimer equilibrium. As also the PTB1-PTB2 linker region is directly adjacent, the WW-domain is likely be involved in this process as also indicated by our pull-down assays, which show at least some influence of the WW-domain on Fe65 dimerization. The WW-domain recognizes polyproline stretches (Meiyappan et al., 2007) and might bind to two proline residues close to Fe65-PTB2 and therefore to the PIP2-epitope. Inline, it had been found that the AICD and the WW-domain cannot bind simultaneously to the PTB-domains including the linker (Cao and Südhof, 2004).

We therefore propose the following integrated scenario for Fe65-mediated gene transactivation (**Figure 6**): Fe65 is the central adaptor for APP nuclear signaling as validated earlier. Without its upstream signal, consisting of the AICD in context of membrane-associated APP, Fe65 resides in a

closed conformation. This conformation occurs in the cytosol and might avoid futile cycles and ensure efficient recycling of Fe65 pools from the nucleus back to the endomembrane system or the cell membrane. The closed conformation favors homotypic dimerization via the structural transition of the C-terminal helix α3 to strand β ct that performs substratemimicry. At the membrane, APP and potentially other protein and lipid factors like PIP2, induce an opening of Fe65 and the homodimer finally dissociates. Therefore, membrane association via the basic cluster and subsequent APP binding would also result in the opening and activation of Fe65. Similarly, it appears well feasible that other functions of Fe65, involving interaction via the WW-domain with Mena or via the PTB1 domain with other surface receptors such as LRP1 might also go along with changes in the Fe65 monomer/dimer equilibrium. Further research will be required to understand these processes in more detail.

Upon ε-cleavage of APP by γ-secretase, the AICD is released from the membrane into the cytosol and the Fe65-AICD complex translocates to the nucleus. Very recent results indicate that the PTB2 rather than the WW domain is important for the nuclear localization of Fe65 (Koistinen et al., 2017). Secretase cleavage is influenced by various aspects like APP cellular localization (Haass et al., 2012), APP dimerization (Winkler et al., 2015) and APP and Fe65 phosphorylation (Bukhari et al., 2016). Due to the tight and extended interaction involving 2/3 of the AICD (Radzimanowski et al., 2008c) and co-localization studies (von Rotz et al., 2004), we favor co-migration without degradation of the AICD. Fe65-PTB1 then binds to Tip60 or other transcription factors like CP2/LSF/LBP1 (Zambrano et al., 1998). The WW-domain in the open Fe65 conformation could finally engage with downstream components as found for the nucleosome assembly factor SET (Telese et al., 2005) or the AICD might interact with Med12 from the transcriptional mediator complex (Xu et al., 2011) essential for starting transcriptional activation processes.

In summary, our structural and biochemical dissection of the molecular properties of the multiprotein-adapter Fe65 reveal the details of an essential regulatory circuit of APP signaling. The importance of APP signaling in health and disease make it worth revisiting Fe65 and its different functional conformations as target for further pharmacological investigations.

#### ACCESSION NUMBERS

Coordinates and structure factors have been deposited at the Protein Data Bank (PDB) with accession number 5NQH.

#### AUTHOR CONTRIBUTIONS

LPF, KH, PS, SP, SE, GS, BS and KW performed the experiments. All authors analyzed the data and contributed to writing of the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by the Deutsche Forschungsgemeinschaft (FOR1332 to KW and SK), the Alzheimer Forschung Initiative (to SK), and the Schweizerischer Nationalfonds (SNF 31003A\_146532 to UK). We thank

#### REFERENCES


Jürgen Kopp and Claudia Siegmann from the BZH/Cluster of Excellence: CellNetworks crystallization platform. We acknowledge access to the beamlines at the European Synchrotron Radiation Facility (ESRF) in Grenoble and the support of the beamline scientists.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fnmol.2017.00 140/full#supplementary-material


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Feilen, Haubrich, Strecker, Probst, Eggert, Stier, Sinning, Konietzko, Kins, Simon and Wild. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Role of APP Interactions with Heterotrimeric G Proteins: Physiological Functions and Pathological Consequences

Philip F. Copenhaver<sup>1</sup> \* † and Donat Kögel<sup>2</sup>†

<sup>1</sup> Department of Cell, Developmental and Cancer Biology, Oregon Health & Sciences University, Portland, OR, USA, <sup>2</sup> Experimental Neurosurgery, Goethe University Frankfurt, Frankfurt am Main, Germany

Following the discovery that the amyloid precursor protein (APP) is the source of β-amyloid peptides (Aβ) that accumulate in Alzheimer's disease (AD), structural analyses suggested that the holoprotein resembles a transmembrane receptor. Initial studies using reconstituted membranes demonstrated that APP can directly interact with the heterotrimeric G protein Gαo (but not other G proteins) via an evolutionarily G protein-binding motif in its cytoplasmic domain. Subsequent investigations in cell culture showed that antibodies against the extracellular domain of APP could stimulate Gαo activity, presumably mimicking endogenous APP ligands. In addition, chronically activating wild type APP or overexpressing mutant APP isoforms linked with familial AD could provoke Go-dependent neurotoxic responses, while biochemical assays using human brain samples suggested that the endogenous APP-Go interactions are perturbed in AD patients. More recently, several G protein-dependent pathways have been implicated in the physiological roles of APP, coupled with evidence that APP interacts both physically and functionally with Gαo in a variety of contexts. Work in insect models has demonstrated that the APP ortholog APPL directly interacts with Gαo in motile neurons, whereby APPL-Gαo signaling regulates the response of migratory neurons to ligands encountered in the developing nervous system. Concurrent studies using cultured mammalian neurons and organotypic hippocampal slice preparations have shown that APP signaling transduces the neuroprotective effects of soluble sAPPα fragments via modulation of the PI3K/Akt pathway, providing a mechanism for integrating the stress and survival responses regulated by APP. Notably, this effect was also inhibited by pertussis toxin, indicating an essential role for Gαo/i proteins. Unexpectedly, C-terminal fragments (CTFs) derived from APP have also been found to interact with Gαs, whereby CTF-Gαs signaling can promote neurite outgrowth via adenylyl cyclase/PKA-dependent pathways. These reports offer the intriguing perspective that G protein switching might modulate APP-dependent responses in a context-dependent manner. In this review, we provide an up-to-date perspective on the model that APP plays a variety of roles as an atypical G protein-coupled receptor in both the developing and adult nervous system, and we discuss the hypothesis that disruption of these normal functions might contribute to the progressive neuropathologies that typify AD.

Keywords: Alzheimer's disease, amyloid precursor protein, APPL, Drosophila, Gαo, Manduca, migration, stress signaling

#### Edited by:

Detlev Boison, Legacy Health, USA

#### Reviewed by:

Kristina Endres, University of Mainz, Germany Angele Parent, University of Chicago, USA

> \*Correspondence: Philip F. Copenhaver copenhav@ohsu.edu

†These authors have contributed equally to this work.

> Received: 26 October 2016 Accepted: 05 January 2017 Published: 31 January 2017

#### Citation:

Copenhaver PF and Kögel D (2017) Role of APP Interactions with Heterotrimeric G Proteins: Physiological Functions and Pathological Consequences. Front. Mol. Neurosci. 10:3. doi: 10.3389/fnmol.2017.00003

# APP AS AN UNCONVENTIONAL G PROTEIN-COUPLED RECEPTOR: HISTORICAL PERSPECTIVE

Members of the APP family share many of the structural features that distinguish type-1 transmembrane receptors, including evolutionarily conserved extracellular domains capable of binding a variety of candidate ligands, plus highly conserved intracellular domains that can mediate interactions with numerous cytoplasmic adapter and signaling proteins (Turner et al., 2003; Jacobsen and Iverfeldt, 2009; Deyts et al., 2016b). In addition, APP is also capable of both homodimeric binding (to itself) and heterodimeric interactions with two APP-like proteins (APLP1 and APLP2) and other membrane-associated proteins (Scheuermann et al., 2001; Soba et al., 2005; Wang et al., 2009; Kaden et al., 2012), consistent with the perspective that APP and its orthologs can function as neuronal receptors that modulate both physiological and pathological responses. Whereas receptors with the topology of APP are most commonly associated with the activation of intracellular kinases (Heldin et al., 2016; Trenker et al., 2016), a growing number of singlepass receptors have now been shown to function as authentic G protein-coupled receptors (GPCRs) that mediate cellular responses via heterotrimeric G proteins, including Fibroblast Growth Factor and Epidermal Growth Factor Receptors (Patel, 2004; Hawkes et al., 2007). Based on the identification of a short motif in Insulin-like Growth Factor II receptor that binds the heterotrimeric G protein Gi (Okamoto et al., 1990), Nishimoto et al. (1993) identified a similar motif in APP (**Figure 1A**; described below), suggesting that APP might also function as G protein-interacting receptor. Specifically, they identified a 20 amino acid peptide ("peptide 20") within the intracellular domain (His657-Lys676; numbering in APP695) that could directly bind and activate heterotrimeric G proteins containing Gαo but not other Gα subunits (including Gαs, Gαi1, Gαi2, and Gαi3) in reconstituted membranes (**Table 1**). This effect was blocked by PTX (a selective inhibitor of the Gαo/i subfamily). They also demonstrated that the alpha subunit of Go (Gαo) but not Gαi could be co-immunoprecipitated with APP from concentrated brain membranes, an interaction that was inhibited by adding excess peptide 20. Using membrane preparations from transfected SF9 cells, they then showed that APP<sup>695</sup> could be co-immunoprecipitated with purified bovine Go, in contrast to mutated forms of APP lacking the peptide 20 domain (Nishimoto et al., 1993). Of note is that Gβ could also be detected in these immunoprecipitates, consistent with the model that APP normally interacts with Go as a heterotrimeric complex (similar to conventional GPCRs). Lastly, Gαo was shown to specifically mediate the effects of peptide 20 on GTP hydrolysis, while pretreatment with GTPγS blocked this interaction (Lang et al., 1995), indicating that the activation state of Go regulates its interaction with APP (again consistent with conventional GPCRs).

In related experiments, Ikezu et al. (1996) co-expressed APP with chimeric Gα subunits to demonstrate that the last five amino acids of Gαo are necessary for its interactions with APP, whereas chimeras containing the cytoplasmic domains of other Gα subunits were ineffective (**Table 1**). This result is consistent with extensive evidence that C-terminal residues within Gα subunits control the specificity of their interactions with conventional GPCRs (Hamm et al., 1988; Herrmann et al., 2004). In collaboration with other groups, they also showed that soluble peptide 20 could regulate Go-dependent exocytosis but had no effect on Gs-dependent membrane fusion events, further validating the model that APP specifically interacts with the C-terminal region of Gαo (Colombo et al., 1994; Lang et al., 1995). These results provide strong evidence that the juxtamembrane G protein-binding domain in APP promotes functional interactions with Gαo (but not other G proteins), suggesting that APP might indeed function as an atypical Gocoupled receptor.

Subsequent studies explored whether stimulating APP with an antibody against its extracellular domain (22C11; to mimic ligand binding) could induce Gαo activity. In liposomes containing reconstituted APP<sup>695</sup> and bovine Go, treatment with 22C11 induced the activation of Go (but not Gi2) in the absence of other proteins (Okamoto et al., 1995, 1996). Although the 22C11 antibody can also detect APLP2 (Slunt et al., 1994), other antibodies targeting different epitopes in APP (but not APLP1 or APLP2) were also found to induce Go-associated responses, including α-1680 and Alz90 (Sudo et al., 2000). In this regard, several groups also tested whether the effects of APP on Gαo signaling might be recapitulated by APLP1 or APLP2. Although one study showed that antibody activation of either APP or APLP2 could induce similar cytotoxic responses to 22C11 (Mbebi et al., 2002), other investigators used APP knockout lines to show that only re-expression of APP rescued Gαodependent responses, whereas expression of APLP1 and APLP2 did not (Sola Vigo et al., 2009; Milosch et al., 2014). Thus, these studies provided intriguing evidence that only APP can function as an unconventional Go-coupled receptor, albeit under rather artificial conditions.

# ABERRANT APP-Go SIGNALING CAN PROVOKE NEURODEGENERATION

How might the misregulation of normal APP-Go signaling contribute to the pathology of AD? To address this issue,

**Abbreviations:** Aβ, beta-amyloid peptide derived from APP; AC, adenylyl cyclase; AD, Alzheimer's disease; AICD, APP intracellular domain cleavage fragments of APP family proteins; Akt, target of PI3K (also called Protein kinase B); APP, amyloid precursor protein; APP695, predominant isoform of APP in mammalian neurons (695 amino acids); APLP1 and APLP2, APP-Like-Proteins 1 and 2 (additional APP family members expressed in the mammalian brain); APPL, APP-Like, the insect ortholog of human APP; BiFC, bimolecular fluorescence complementation; CaMKIV, calcium/calmodulin-dependent protein kinase IV; cAMP, cyclic adenosine monophosphate; CREB, cAMP response element binding protein; CTX; cholera toxin; FAD, familial AD; Gαi, alpha subunit of the heterotrimeric G protein Gi; Gαo, alpha subunit of the heterotrimeric G protein Go; Gβγ, beta/gamma dimeric subunits of heterotrimeric G proteins; GSK3β, glycogen synthase kinase 3 beta; JNK, c-Jun N-terminal kinase; pCREB, phosphorylated CREB; PI3K, phosphatidylinositol-4,5-bisphosphate 3-kinase; PKA, protein kinase A; PS-1, presenilin-1; PS-2; presenilin 2; PTX, pertussis toxin, a selective inhibitor of Gαi/Gαo family of heterotrimeric G proteins; sAPPα, secreted ectodomain fragments of APP generated by α-secretase cleavage. sAPPL, secreted ectodomain fragments of insect APPL (equivalent to sAPPα).

and APLP1 that were found to be necessary for interactions between membrane-tethered AICDs or CTF fragments of the holoproteins and Gαs (Deyts et al., 2012). (D) Deletions in APPL that interfere with Gαo-associated motile responses in developing neurons (1D1, 1D2) and prevent direct binding between APPL and Gαo (1D2). Citations describing each deletion construct are as follows: <sup>a</sup>Nishimoto et al., 1993; <sup>b</sup>Okamoto et al., 1996; c Ikezu et al., 1996; <sup>d</sup>Yamatsuji et al., 1996a; <sup>e</sup>Yamatsuji et al., 1996b; <sup>f</sup>Hashimoto et al., 2000; <sup>g</sup>Sudo et al., 2001; <sup>h</sup>Sola Vigo et al., 2009; <sup>i</sup>Milosch et al., 2014; <sup>j</sup>Shaked et al., 2009; <sup>k</sup>Torroja et al., 1999b; <sup>l</sup>Ramaker et al., 2013.

Yamatsuji et al. (1996a,b) used COS cells expressing Go to compare the responses elicited by wild type APP<sup>695</sup> versus APP containing missense mutations that are known to cause early onset FAD. In contrast to wild type APP695, expression of these "FAD-APP" mutant isoforms (including V642I, V642F, V642G) induced a dramatic increase in DNA fragmentation and apoptosis. This effect was blocked by PTX treatment (indicating Gαo/i proteins) or by expressing a dominant-interfering form of Gαo (**Table 1**), but was not affected by CTX (an activator of Gαs) and was absent in COS cells lacking Go. Notably, treatment with either synthetic Aβ<sup>40</sup> or Aβ<sup>42</sup> did not induce apoptotic responses in this assay, nor did conditioned medium harvested from cell cultures expressing the V<sup>642</sup> mutant isoforms (which produce abundant Aβ42). In combination, these studies supported the model that mutated forms of APP linked with FAD can indeed function as constitutively active Go-coupled receptors. Moreover, they suggested that the pathophysiological effects of FAD-APP mutations might be caused by aberrant hyperactivation of Go-dependent signaling, rather than simply promoting the accumulation of neurotoxic Aβ. An appealing corollary to this model is that the downstream pathways regulated by Go could provide novel biomarkers or therapeutic targets for treating AD.

Unfortunately, attempts to identify these downstream pathways produced paradoxical results. For example, using COS cells co-expressing chimeric Gα subunits with different

#### TABLE 1 | Evidence for functional interactions between APP family proteins and heterotrimeric G proteins.


Summary of published evidence that APP interacts with Gαo (but usually not other G proteins, including Gαs, Gαz, and Gαi isofroms). The table includes studies on both wild type APP695, isolated peptide 20 constructs (containing the G protein-binding domain of APP695), and FAD-associated mutant forms of APP with altered residues at V<sup>642</sup> (indicated in the left-hand column). Studies that showed direct binding between Gαo and APP/APPL are indicated with an asterisk (<sup>∗</sup> ). Studies that used PTX to indicate the involvement of Gαo/i family proteins are indicated with a cross (†). Studies that used CTX to indicate the absence of Gαs-dependent signaling is indicated with a hash mark (#). Study that showed direct binding between Gαs and constructs containing the G proteinbinding domain is indicated with a double asterisk (∗∗). Citations for each set of results are shown in the right-hand column. <sup>∗</sup>Studies that showed direct binding between Gαo and APP/APPL (or the G protein-binding domain). †Studies that demonstrated sensitivity to PTX, indicating the involvement of Gαo/i. #Studies that tested sensitivity to CTX, indicating the absence of Gαs-dependent signaling. ∗∗Study that showed direct binding between Gαs and peptides containing the G protein-binding domain.

variants of APP, Ikezu et al. (1996) found that FAD-APP isoforms inhibited cAMP response element (CRE)-mediated transcription in a Gαo-specific manner. Curiously, this effect was independent of adenylyl cyclase (AC) activity, while inhibitors of Gβγ signaling (rather than Gαo) blocked apoptotic responses in this assay (Giambarella et al., 1997). From these studies, the authors concluded that APP signaling normally regulates both Gαo- and Gβγ-dependent pathways, whereby Gαo regulates CRE-dependent transcriptional responses, while Gβγ regulates other effectors (as yet undefined) that can induce apoptosis when chronically activated. More perplexing were the results from another group, who found that 22C11 treatment in brain membrane fractions actually inhibited Gαo-dependent responses (Brouillet et al., 1999), leading to the proposal that unknown proteins expressed by neurons but not glial-derived cells (or in reconstituted membranes) might regulate Gαo activation by APP (Brouillet et al., 1999; Sudo et al., 2000). How the misregulation of Gαo- versus Gβγ-dependent pathways might contribute to AD remained an open question.

#### NEUROTOXIC MECHANISMS OF MISREGULATED APP- Gαo SIGNALING: CONFLICTING MODELS

Subsequent investigations have generated an unexpectedly complicated (and often contradictory) view of how the APP-Go pathway might function in the diseased nervous system. Using a variety of transfected cell lines, Nishimoto et al. (1993) first confirmed that the induction of APP-Gαo signaling (by antibody crosslinking or induced dimerization) required transmembrane APP (Sudo et al., 2000; Hashimoto et al., 2003a), and that hyperactivation of this pathway could induce apoptotic responses in cultured mouse neurons (see also Rohn et al., 2000). Both groups described classic features of neuronal apoptosis in their assays, including neurite degeneration, nuclear condensation, internucleosomal DNA cleavage, and activation of pro-apoptotic caspases (including caspase 3, 7, and 9). Treatment with inhibitors of glutathione metabolism or NADPH oxidase (as well as incubation with antioxidants) effectively blocked the cell death response, suggesting that hyperstimulation of the APP-Gαo pathway induces a chronic elevation of reactive oxygen species (ROS), resulting in the induction of caspasedependent apoptosis. Moreover, expressing FAD-APP isoforms induced the same cytotoxic responses caused by hyperstimulating wild type APP, including activation of ASK1 (Apoptosis Signal-Regulating kinase) and its downstream effector JNK that resulted in chronic upregulation of NADPH oxidase, elevated ROS levels, and activation of pro-apoptotic caspases (Hashimoto et al., 2003b; Niikura et al., 2004). A similar response could be induced by expressing a chimeric protein containing the dimerization domain of the EGF receptor fused with the APP cytoplasmic domain, providing a plausible explanation for how the hyperstimulation of normal APP-Go signaling with crosslinking antibodies could provoke neuronal death in an Aβ-independent manner. By comparison, the neurotoxic effects of FAD-associated mutations within a different region of APP (K595/M596) were found to be independent of Go, suggesting that different disease-associated mutations in APP might perturb a variety of signaling pathways that affect neuronal viability (Hashimoto et al., 2000). Collectively, these results bolstered the

argument that the aberrant APP-Go signaling might contribute to both late-onset AD and some forms of FAD.

However, it should be noted that enforced dimerization of APP (with crosslinked antibodies or chimeric fusion proteins) involves rather artificial methods that may not recapitulate authentic physiological or pathophysiological interactions. Moreover, it is difficult to reconcile these results with more recent evidence that ∼65% of membrane-bound APP in healthy cells is normally present in a dimeric configuration (Gralle et al., 2009). Nevertheless, these cytotoxic effects could be recapitulated by overexpressing an FAD-APP isoform (V642I-APP) in both neuroblastoma cells and primary neurons (Niikura et al., 2000, 2004), independent of Aβ-associated toxicity (Sudo et al., 2001). Alternatively, other groups have suggested that forced dimerization of APP might provoke Go-dependent apoptotic responses via a variety of other pathways, including PAK3-dependent re-entry into the cell cycle (McPhie et al., 2003), misregulation of Src-dependent actin dynamics and focal adhesion turnover (Xu et al., 2009), and calpain/calcineurin-dependent proteolysis of CaMKIV, resulting in the misregulation of CREB (Mbebi et al., 2002). Also problematic is the mechanism by which the APP-Go pathway might actually stimulate JNK: although both the α and βγ subunits of a number of heterotrimeric G proteins (including Go) can modulate JNK activity in different contexts, these responses typically require a cascade of other kinases and adapter proteins that have not been implicated in APP-Go signaling (Goldsmith and Dhanasekaran, 2007; Bromberg et al., 2008; Yu et al., 2016). Lastly, all of these studies focused on pathological outcomes that could be induced by aberrant APP-Gαo signaling, but the authentic functions of this pathway in the healthy nervous system remained largely unexplored. As described below, recent studies from the Kögel laboratory have now indicated that APP-Gαo signaling may actually antagonize the JNK pathway under physiological conditions, whereby the induction of APP signaling counteracts cellular stress responses via the PI3K cascade, providing a mechanism that promotes neuronal survival (Kögel et al., 2012; Milosch et al., 2014).

#### IS APP-Gαo SIGNALING ALTERED IN HUMAN PATIENTS WITH AD?

Whether the misregulation of APP-Go signaling actually plays a role in provoking AD remains unknown. However, a variety of studies have offered intriguing hints that support this hypothesis. Initial reports using human brain samples revealed that the expression patterns of many heterotrimeric G proteins are altered in late sporadic AD, particularly within the most vulnerable brain regions (including cortex and hippocampus). These changes also correlate with a general reduction in G protein-dependent GTP hydrolysis at stages that precede the onset of clinical disease (O'Neill et al., 1994; Cowburn et al., 2001; Garcia-Jimenez et al., 2002). Similarly, using reconstituted membrane preparations from human brain samples, Mahlapuu et al. (2003) found that the induction of G protein activity by APP-derived peptides was significantly reduced in postmortem elderly AD patients compared to age-matched controls. Recapitulating the original studies by Nishimoto et al. (1993), they also found that membrane-tethered constructs of the Go domain (peptide 20 plus the transmembrane T639-L<sup>649</sup> sequence) induced more robust [35S]GTPγS binding than soluble peptide 20 (Mahlapuu et al., 2003). Curiously, adding the transmembrane peptide alone (T639-L649) also affected [35S]GTPγS binding, while equivalent peptides containing V<sup>642</sup> APP-FAD mutations were even more effective (Karelson et al., 2005), although how these hydrophobic constructs might interact with G proteins when applied to isolated membranes is unclear. Nevertheless, these results provided indirect evidence that disease-associated changes in the GPCR-like function of APP might contribute to both FAD and late-onset AD (as noted by the authors).

Perhaps because it is the most abundant G protein in the brain (Strittmatter et al., 1990; Jiang and Bajpayee, 2009), the overall levels of Gαo do not appear to be altered in either FAD or late-onset sporadic AD (O'Neill et al., 1994; Shaked et al., 2009), but several studies suggest that Gαospecific responses are progressively disrupted in both familial and late sporadic forms of the disease. For example, using membrane preparations from human brain samples, Reis et al. (2007) found that the effects of FAD-APP-derived peptides on G protein activity were blocked by PTX, while another report showed that Aβ peptides could activate Gαo in lipid vesicles (Rymer and Good, 2001), although it is unclear whether the topology of these assays recapitulates authentic Gαo-Aβ interactions. More compelling are two studies showing that APP-Go signaling might be directly altered by neurotoxic Aβ in neurons. Based on previous evidence that APP can bind neurotoxic Aβ fibrils (Lorenzo et al., 2000; Van Nostrand et al., 2002), Lorenzo et al. (2000) also showed that APP overexpression rendered hippocampal neurons more vulnerable to Aβ-induced degeneration, an effect that was abrogated by deletion of the Go-binding domain in APP or treatment with PTX (Sola Vigo et al., 2009). Notably, expressing a PTX-insensitive form of Gαo restored the toxic effects of Aβ treatment, but only in the presence of an intact Go-binding domain. Subsequent work by Masliah and colleagues demonstrated that treatment with Aβ reduced APP-Gαo interactions (corresponding to Go activation) and induced cell death in transfected neuroblastoma lines, and again this effect was PTX-dependent (Shaked et al., 2009). Aβ treatment also provoked a significant increase in calcium (Ca2+) influx in a Go-dependent manner, consistent with earlier studies suggesting that hyperactivation of APP signaling could provoke Ca2<sup>+</sup> overload and cell death. Most notably, they showed that APP-Gαo interactions declined in patients suffering from progressive stages of AD, corresponding to an overall increase in G protein activation (though not specifically Gαo).

In the course of their cell culture assays, the authors found that mutating a particular residue within the cytoplasmic domain of APP (D664A) blocked the ability of Aβ to affect APP-Gαo interactions (Shaked et al., 2009). Noting that this residue is required for caspase-dependent cleavage of APP to generate a cytotoxic C31 fragment (Lu et al., 2003), they proposed a mechanism by which Aβ binding induces caspase-dependent

cleavage of APP, resulting in the release of a C31-Go complex that could stimulate Gαo in some undefined fashion. However, other investigators have noted that the D664A mutation (located within the Go domain) is equally likely to disrupt interactions between APP and other cytoplasmic proteins (Galvan et al., 2007), the most obvious candidate being Gαo. Thus, mutations at this site might perturb key structural features that permit APP to function as a Go-coupled receptor, although the steric rearrangements that lead to the activation of Gαo remain unexplored. Paradoxically, Shaked et al. (2009) also reported that deletion of the C-terminal YENPTY domain mitigated the effects of Aβ on Gαo activation, contradicting several previous studies demonstrating that this motif is not required for direct interactions between APP and Gαo (Nishimoto et al., 1993; Kawasumi et al., 2004; King and Scott Turner, 2004; Sola Vigo et al., 2009). Nevertheless, these results offered the most compelling evidence that APP-Go signaling is altered over the course of AD, consistent with the model that elevated Aβ might induce the aberrant activation of Gαo-dependent pathways that provoke neuropathological responses.

Recently, Fogel et al. (2014) used fluorescence resonance energy transfer (FRET)-based protocols to demonstrate a close association between APP and Gαo that was modulated by APP activation. They also showed that Aβ<sup>40</sup> induced structural rearrangements in the presynaptic APP/Go complex by promoting APP dimerization, which in turn resulted in G protein-dependent Ca2<sup>+</sup> influx and glutamate release (Fogel et al., 2014). Both aspects of this response were found to critically involve the E1 extracellular domain of APP, suggesting that Aβ<sup>40</sup> can mimic the effects of endogenous ligands. Based on these findings, the authors proposed that excessive APP activation by amyloid peptides might contribute to hippocampal hyperactivity under pathological conditions, supporting the hypothesis that normal APP-Gαo interactions are altered in AD. An added dimension to this model is that Gαo may also functionally interact with presenilins, essential components of the γ-secretase complex that are involved in generating Aβ peptides and AICD fragments and are also mutated in some forms of FAD (Walter et al., 2001; Jayne et al., 2016). For example, Smine et al. (1998) showed that presenilin-1 (PS-1) could be co-immunoprecipitated with Gαo (but not Gαi2) when overexpressed in COS-7 cells, and that a C-terminal fragment (CTF) of PS-1 could activate Gαo (but not Gαi2) in a PTX-sensitive manner. Likewise, overexpressing FAD mutant forms of Presenilin-2 (PS-2) in neuroblastoma cells induced apoptotic responses that were inhibited by PTX and restored by expressing a PTX-resistant variant of Gαo but not Gαi (Wolozin et al., 1996; Abe et al., 2004). Whether presenilins actually modulate Gαo-dependent pathways in neurons and how this might affect APP-Gαo interactions remains to be explored. Nevertheless, it is possible that multiple factors associated with AD might contribute to the pathological misregulation of APP-Gαo signaling (including FAD-linked mutations in both APP and the presenilins), as well as the accumulation of neurotoxic amyloid peptides that can hyperactivate this pathway.

# STRUCTURE, SPECIFICITY, AND EVOLUTIONARY CONSERVATION OF THE Go-BINDING DOMAIN IN APP FAMILY PROTEINS

As noted earlier, Nishimoto et al. (1993) first identified the G protein-binding domain in APP, based on their previous discoveries that several type-1 transmembrane proteins directly bind Gα subunits via short peptide sequences containing BBXB or BBXXB motifs, where B is a basic amino acid residue and X is any non-basic residue (Okamoto et al., 1990, 1991; Okamoto and Nishimoto, 1992). From this analysis, they identified "peptide 20" in APP<sup>695</sup> (H657-L676), which contains two N-terminal basic residues (HH) and terminates in a BBXXB motif (**Figure 1A**; magenta region). In a meticulous series of experiments using reconstituted liposomes and isolated membrane fractions, they then showed that this "peptide 20" domain (subsequently designated the Go activator domain) was both necessary and sufficient for directly binding and activating Gαo, but not Gαs, Gαi1, Gαi2, or Gαi<sup>3</sup> (**Table 1**). Removing either the N-terminal histidines (**Figure 1A**, asterisks) or the C-terminal BBXXB motif from peptide 20 (RHLSK) greatly attenuated its ability to simulate Gαo in GTPase activation assays, although membrane-tethered versions of the Go domain were considerably more potent than soluble forms. Thirdly, they demonstrated that interactions between full-length APP and Gαo required this domain: a deletion that removed both the Go domain and the C-terminal YENPTY motif precluded APP-Gαo interactions (His657-N695; **Figure 1B**1), whereas a deletion encompassing only the YENPTY did not (**Figure 1B**2). These results provide strong evidence that the juxtamembrane G protein-binding domain in APP promotes functional interactions with Gαo but not other G proteins (Nishimoto et al., 1993).

Using similar methods, Nishimoto et al. (1993) subsequently showed that full-length APP binds and stimulates Gαo (but not Gαi2) following antibody activation in reconstituted vesicles (Okamoto et al., 1995; Ikezu et al., 1996), while the apoptotic effects of FAD-APP isoforms (mutated at V642) were both PTXsensitive and required the Go domain: FAD-APP constructs lacking only the Go domain (**Figure 1B**3) failed to induce Gαodependent cytotoxic responses, whereas deletions encompassing the YENPTY domain (**Figure 1B**2) had no effect (Okamoto et al., 1996; Yamatsuji et al., 1996a; Hashimoto et al., 2000; Niikura et al., 2000; Sudo et al., 2001). This apoptotic response could also be blocked with dominant-interfering forms of Gαo (GαoG204A) but not Gαi<sup>2</sup> (**Table 1**; Yamatsuji et al., 1996b). Using Myc-tagged constructs for in vitro pull-down assays, Brouillet et al. (1999) subsequently confirmed that the cytoplasmic domain of APP could bind Gαo but not Gαi2, and that this interaction was reduced when the N-terminal H657H<sup>658</sup> doublet was replaced with hydrophobic residues. Sudo et al. (2001) and Hashimoto et al. (2003a) then showed that that apoptotic effects of APP stimulation were prevented by deleting the Go interaction domain (**Figure 1B**3) but not the YENPTY domain (**Figure 1B**2), and that they were mediated specifically by Gαo but not Gαi. Similarly, based

on evidence that Aβ might induce neurotoxic responses via the APP-Gαo pathway, Lorenzo and colleagues showed that this effect also required the Go domain (Sola Vigo et al., 2009): deleting the entire cytoplasmic domain (**Figure 1B**4) precluded the activation of Gαo-dependent responses to Aβ, as did complementary deletions targeting different portions of the Go domain (**Figure 1B**5**,**6), whereas a deletion encompassing the YENPTY motif did not (**Figure 1B**7). In a more physiological context, the Kögel group recently demonstrated the importance of the Go domain in mediating APP-dependent neuroprotective responses to sAPPα: a deletion that removed the conserved PEERH motif within this domain (**Figure 1B**9) prevented APPdependent signaling that was also blocked by PTX (implicating Gαo/i proteins), whereas two different deletions targeting the YENPTY motif (**Figure 1B**10**,**11) had no effect (as summarized below).

In contrast to the foregoing studies, Shaked et al. (2009) reported that Gαo could still be co-immunoprecipitated with APP lacking the C31 cytoplasmic region (including both the Go-binding domain and the YENPTY motif; **Figure 1B**12), but that deleting this region prevented APPdependent activation of Gαo pathways in cell culture. They also found that over-expressed C99 fragments could be co-immunoprecipitated with Gαo (the only report of this interaction). Curiously, deletion of only the YENPTY motif (**Figure 1B**13) also blocked Gαo-dependent responses in this assay, in contrast to many other studies demonstrating that this domain is not required for APP-Gαo interactions. Based on these observations, the authors postulated that the transduction of APP-Gαo signaling might involve the YENPTY motif as well as the Go domain (either directly or indirectly), possibly in response to Aβ-induced cleavage of APP (Shaked et al., 2009). Whether this response also involves internalization responses mediated by the YENPTY motif remains to be explored (cf. Lai et al., 1995; Deyts et al., 2016b).

Other members of the APP family also contain Go-like domains, albeit with some sequence variations (**Figure 1C**). Both APLP1 and APLP2 contain only one N-terminal histidine that aligns with the HH doublet in APP<sup>695</sup> (highlighted in yellow), and only APLP2 also possesses an intact C-terminal BBXXB motif (boxed region). As summarized above, only APP<sup>695</sup> has been shown to activate Gαo, although a rigorous analysis of potential interactions between APLP1/2 and Gαo has not been conducted in vivo. Likewise, the Go domains in both nematode APL-1 and insect APPL contain only a single N-terminal histidine and lack complete BBXXB motifs. Nevertheless, studies in several insect models have shown that APPL does functionally interact with Gαo both in vitro and in vivo, whereby deleting different portions of the Go domain in APPL (**Figure 1D**1,2) disrupted Gαo-associated responses in the developing nervous system (Torroja et al., 1999b; Ramaker et al., 2013; and described below). How these structural variations within the Go domain might affect the dynamics of Gαo activation/inactivation under physiological conditions remains to be explored.

# PHYSIOLOGICAL ROLE OF APP-Gαo INTERACTIONS IN STRESS SIGNALING AND NEUROPROTECTION

Based on early work suggesting that APP might regulate both cell adhesion and excitoprotective responses (Mattson et al., 1993; Schubert and Behl, 1993), a variety of in vitro and in vivo assays demonstrated that both full-length APP and its sAPPα ectodomain fragments (produced by α-secretase cleavage) could have potent neuroprotective activity under different conditions (reviewed in Kögel et al., 2012; Nhan et al., 2015). For example, deletion of the sole APP ortholog in nematode (APL-1) caused larval lethality that could be rescued by expressing extracellular domain fragments equivalent to sAPPα (Hornsten et al., 2007; Ewald et al., 2016), while overexpressing sAPPα rescued some behavioral deficits in mice lacking members of the APP family (Ring et al., 2007; Weyer et al., 2011). From these and other experiments emerged a complex scenario whereby both APP and sAPPα might independently confer beneficial responses under physiological conditions. However, elevated sAPPα levels can also have unwanted effects on cell proliferation and tumorigenesis, potentially due to interactions with receptors whose roles in neuroprotection is unclear (Adlerz et al., 2007; Zhou et al., 2011). More recently, Kögel and colleagues have provided new evidence that transmembrane APP and sAPPα interact as a ligand/receptor pair in neurons to modulate stress signaling, via activation of the pro-survival PI3K/Akt pathway (Milosch et al., 2014). Using a variety of experimental strategies, they demonstrated that both APP and sAPPα antagonize the activation of the JNK-dependent stress signaling pathway, which (as noted earlier) is a key upstream modulator of mitochondriadependent apoptosis (Kögel et al., 2005; Copanaki et al., 2010; Eckert et al., 2011). Conversely, several groups have now shown that the protective function of APP requires activation of the PI3K/Akt pathway (Cheng et al., 2002; Copanaki et al., 2010; Eckert et al., 2011; Jimenez et al., 2011). Since Akt negatively regulates several JNK-activating kinases, including ASK1 and mixed lineage kinase 3 (MLK3), these findings suggest that APP modulates a dynamic interplay between stress and survival pathways (Kögel et al., 2012).

To define the role of full-length APP in this response, Milosch et al. (2014) showed that the protective effects of both sAPPα and a recombinant fragment containing only the E1 domain of APP were completely abrogated in neurons from APP knockout animals or in APP-depleted SH-SY5Y cells. These results clearly demonstrated that expression of membrane-bound holo-APP was required for sAPPα-dependent Akt activation and neuroprotection in these assays, supported by other evidence that sAPPα can regulate the dimerization of transmembrane APP in cell culture (Gralle et al., 2009; Kaden et al., 2012). Likewise, studies in Drosophila have shown that sAPPL ectodomain fragments (equivalent to sAPPα) bind full-length APPL, and that the neuroprotective effects of sAPPL require the presence of the holoprotein (Wentzell et al., 2012). More recently, a behavioral analysis demonstrated that full-length APPL and secreted sAPPLα act together to promote memory formation in

adult Drosophila (Bourdet et al., 2015), consistent with the model that APP-sAPPα interactions may serve a variety of physiological functions in the nervous system.

Although the foregoing experiments demonstrated that the C-terminal domain of APP was required for the neuroprotective effects of the holoprotein, the last 15 amino acids were dispensable (as summarized in **Figure 1B**8−11): sAPPα-dependent activation of Akt was unaffected in neurons from APP-1CT15 mice, which express a mutant form of APP lacking the cytoplasmic YENPTY motif (Milosch et al., 2014). As noted in other reviews, this domain mediates interactions with a plethora of cytoplasmic proteins but not Gαo (Nishimoto et al., 1993; King and Scott Turner, 2004; Sola Vigo et al., 2009). To further map the specific regions in APP that are required for this activity, APP-KO cells were transfected with an APP construct lacking the PEER motif within its Go-binding domain (1PEERH). In contrast to the YENPTY mutant, the 1PEERH mutant did not rescue sAPPα-induced Akt activation following trophic factor deprivation. In addition, treatment with PTX completely abolished the ability of sAPPα to promote Akt activation and cell survival, further implicating a role for Go in this response. Lastly, activation of the PI3K/Akt pathway by sAPPα induced the phosphorylation of glycogen synthase kinase 3β (GSK3β), which is a well-known mechanism for inhibiting GSK3β-induced apoptotic responses (Watcharasit et al., 2003; Hanumanthappa et al., 2014). Whereas PI3K/Akt signaling was originally linked with receptor tyrosine kinase activation, numerous studies have shown that heterotrimeric G proteins also play a critical role in regulating PI3K activity under both physiological and pathological conditions (Murga et al., 1998; Murga et al., 2000; New and Wong, 2007; Yanamadala et al., 2009). Since PTX selectively inhibits members of the Gαo/i family, while APP only interacts with Gαo and potentially Gαs (as noted below), these results argue that APP/sAPPα interactions induce the PI3K/Akt pathway specifically via Gαo.

Based on these findings, we propose that transmembrane APP mediates sAPPα-induced neuroprotection via Gαo-coupled activation of the PI3K/Akt pro-survival pathway (**Figure 2A**). In turn, activation of Akt phosphorylates and inhibits GSK3β, as well as other pro-apoptotic targets (Datta et al., 1997; Endo et al., 2006; Jover-Mengual et al., 2010). We also propose that this response requires direct interactions between sAPPα and holo-APP as a ligand-receptor pair. These results offer a resolution to paradoxical findings from previous investigations, demonstrating that holo-APP and sAPPα are equally important in mediating neuroprotective responses. Conversely, factors that interfere with this function would render neurons more susceptible to cellular stress during brain aging and AD. The model that APP-Gαo signaling serves a neuroprotective function under physiological conditions contrasts with the cytotoxic response elicited by hyperactivating this pathway in AD models (as summarized above). Of note is that treatment with Aβ might also interfere with the neuroprotective effects of sAPPα, resulting in the disinhibition of GSK3β and consequent upregulation of apoptotic pathways (Jimenez et al., 2011). Since GSK3β activity is increased in the AD brain (Crews and Masliah, 2010; Jimenez et al., 2011; Llorens-Martin et al., 2014), we hypothesize that the decline in sAPPα levels associated with both sporadic AD and FAD contributes to this phenomenon (Almkvist et al., 1997; Sennvik et al., 2000), thereby promoting tau hyperphosphorylation (Deng et al., 2015) and sensitizing neurons to stress and apoptosis. In summary, these studies provide new insight into the mechanisms by which APP-Go signaling regulates neuronal stress responses under physiological conditions, and how the loss of this function might render neurons more susceptible to cellular stress during normal brain aging and AD.

# APP-Gαo SIGNALING IN THE CONTROL OF NEURONAL MOTILITY: VIEWS FROM A NON-MAMMALIAN SYSTEM

Although APP was originally identified in humans, it is actually a member of an evolutionarily ancient family of proteins that may serve similar roles in the developing nervous systems of many organisms (Coulson et al., 2000; Ewald and Li, 2012; Lazarov and Demars, 2012; Shariati and De Strooper, 2013). Studies using a variety of insect models have shown that APPL shares both structural and functional conservation with human APP695, including homologous extracellular and intracellular motifs that regulate interactions with other proteins (Cassar and Kretzschmar, 2016). In particular, several groups have demonstrated a role for APPL-Gαo signaling in neuronal development. Using genetic methods, Torroja et al. (1996, 1999a) first showed that APPL plays an important role in regulating neuronal growth and maturation, and that this activity requires the conserved Go-binding domain shared by APP<sup>695</sup> and APPL. Replacing endogenous APPL with a mutant form lacking this domain (**Figure 1D**1) disrupted the normal maturation of synaptic boutons at the neuromuscular junction, potentially caused by the loss of ligand-dependent APPL-Go signaling (Torroja et al., 1999b). Subsequent investigations into this response suggested a role for the homophilic cell adhesion receptor Fasciclin II (Fas II; the insect ortholog of NCAM), whereby trans-synaptic interactions mediated by Fas II could promote APPL signaling, in part via the activation of Gαo. Whether Fas II acts as a ligand as well as a co-receptor for APPL remains to be explored, as does the role of downstream Gαo effectors in regulating synaptic maturation. Nevertheless, this work offered compelling evidence that the APP-Go pathway is conserved in both invertebrate and vertebrate nervous systems.

Using Manduca sexta (tobacco hornworm) as a complementary model, the Copenhaver laboratory has also explored the role of APPL-Gαo signaling in the developmental control of neuronal motility. Unlike Drosophila, the formation of the embryonic nervous system in Manduca involves an extended period of neuronal migration (Copenhaver and Taghert, 1989; Copenhaver, 2007), analogous to the more complex waves of migration that typify mammalian brain development (Ayala et al., 2007; Tabata and Nagata, 2016). Notably, APPL colocalizes with Gαo in the leading processes and growing axons of migratory neurons in Manduca (Swanson et al., 2005), similar to the colocalization of APP and Gαo in cultured mammalian neurons (Ramaker et al., 2013).

In addition, co-immunoprecipitation assays showed that endogenously expressed APPL and Gαo functionally interact in a manner that is regulated by Gαo activation (Ramaker et al., 2013). By co-expressing fusion constructs of APPL and Gαo containing complementary portions of Venus fluorescent protein in transfected COS7 cells, bimolecular fluorescence complementation (BiFC) assays were used to demonstrate that transmembrane APPL directly bound Gαo (but not Gαi or Gαs), while APP<sup>695</sup> also directly bound Gαo, similar to conventional GPCRs (Marinissen and Gutkind, 2001; Oldham and Hamm, 2008). More importantly, expressing these constructs in transgenic Drosophila lines revealed that APPL bound Gαo in healthy neurons, providing the first demonstration of direct interactions between an APP family protein and Gαo in vivo. Notably, this interaction could be readily visualized within synaptic regions of the brain by BiFC, whereas deleting the Go domain in APPL (**Figure 1D**2) eliminated APPL-Gαo binding (Ramaker et al., 2013). In combination, these studies substantiate the model that APP family proteins can indeed function as unconventional GPCRs, specifically regulating Gαo-dependent responses.

By adapting an embryo culture assay that permits targeted manipulations of migratory neurons in Manduca (Horgan and Copenhaver, 1998), the Copenhaver laboratory subsequently showed that APPL-Gαo signaling plays an important role in regulating neuronal motile behaviors: inhibiting either APPL expression or Gαo activity induced a distinctive pattern of ectopic growth and migration, while hyperstimulating the APPL-Gαo pathway induced collapse-stall responses (Ramaker et al., 2013). These effects were analogous to the striking pattern of ectopic neuronal migration reported in the brains of mice deleted for all three APP family proteins (Herms et al., 2004), and recapitulated earlier studies in Manduca showing that activated Gαo inhibits migration via the induction of voltage-independent currents

(Horgan et al., 1995; Horgan and Copenhaver, 1998). More recent studies have identified Manduca Contactin (MsContactin) as a candidate ligand for APPL (Ramaker et al., 2016b). Specifically, experiments in cultured embryos indicated that GPI-linked MsContactin (expressed by adjacent glial cells) activates APPL-Gαo signaling in the migratory neurons to induce local retraction responses (**Figure 2B**), thereby preventing ectopic outgrowth. This discovery was supported by reports that multiple Contactin family members in mammalian systems can interact with APP and its orthologs both in cis and trans (Ma et al., 2008; Osterfield et al., 2008; Tachi et al., 2010; Osterhout et al., 2015). In summary, our experiments provide new evidence that APP family proteins regulate key aspects of neuronal development during embryogenesis, in part via activation of Gαo-dependent pathways. Still to be determined are the downstream effectors that transduce the effects of APPL-Gαo signaling on neuronal behavior. Likewise, whether mammalian Contactins might regulate APP-Gαo signaling in migratory cortical neurons, and whether modulation of the PI3K-Akt pathway or GSK3β activity also contributes to this response within the developing nervous system remains to be explored (e.g., Morgan-Smith et al., 2014).

# APP MAY ALSO REGULATE NEURONAL MOTILITY via Gαs-DEPENDENT PATHWAYS

Most studies support the model that transmembrane APP normally binds and activates Gαo in response to a variety of ligands (including sAPPα and MsContactin), suggesting that APP cleavage (by secretases or caspases) is likely to terminate APP-Gαo signaling rather than activating it. In support of this model, we recently showed that blocking α-secretase activity in the migratory neurons of cultured Manduca embryos significantly increased membrane-associated APPL levels, while inducing the same collapse/stall responses caused by hyperactivating APPL-Gαo signaling with Contactin fusion proteins (Ramaker et al., 2016a,b). Likewise, our analysis of endogenously expressed APP family proteins showed that Gαo could be readily coimmunoprecipitated with both full length APP<sup>695</sup> (from mouse and human brain lysates) and APPL (from Manduca and Drosophila lysates), whereas we did not detect their CTF or AICD fragments in the immunoprecipitated complexes (Ramaker et al., 2013). These results are also consistent with past work focusing on the functional interactions between transmembrane APP<sup>695</sup> and Gαo (e.g., Okamoto et al., 1995; Hashimoto et al., 2003a; Sola Vigo et al., 2009). However, as noted above, several reports have shown that Gαo can also interact with membrane-tethered peptide 20 domains (mimicking CTFs that contain the Gobinding domain), and one study showed that Gαo could be coimmunoprecipitated with C99 fragments (normally generated by β-secretase cleavage) when overexpressed in neuroblastoma cells (Shaked et al., 2009). Whether Gαo actually continues to interact with CTFs following α- or β-cleavage of the holoprotein in neurons, and whether these interactions might affect downstream pathways regulated by APP-Gαo signaling under physiological conditions, is still unknown.

In contrast, recent studies by Parent and colleagues have indicated that a different G protein (Gαs) may be activated by CTFs derived from the APP holoprotein (Deyts et al., 2012). Specifically, they found that overexpressing a membranetethered AICD construct (mAICD) or experimentally elevating intracellular APP-CTF levels dramatically increased neurite outgrowth in both neuroblastoma cells and transfected cortical neurons. This response required AC-dependent activation of protein kinase A (PKA) and corresponded to the phosphorylation of two PKA targets (CREB and GSK3β), both of which can regulate neuronal motility. To test the involvement of Gαs (a canonical activator of AC), they also showed that HA-tagged Gαs could be co-immunoprecipitated with mAICD from transfected cells, whereas dominant-negative Gαs (lacking its palmitoylation site) prevented mAICD-induced outgrowth. Focusing on the BBXXB motif in APP that was originally identified by Nishimoto et al. (1993) (**Figure 2C**, asterisks), Deyts et al. (2012) found that mutating this site prevented interactions between the mAICD construct and HA-Gαs. Curiously, they also demonstrated an interaction between Gαs and an equivalent construct derived from APLP1, which (like insect APPL) lacks a BBXXB motif (**Figure 1C**, boxed region), suggesting that this motif may not be strictly required for functional interactions between APP family proteins and Gα subunits within intact neurons.

More recently, the Parent group conducted a series of carefully controlled experiments in both cultured neurons and transgenic mice, demonstrating that elevating APP-CTF levels (by a variety of methods) induced exuberant neurite outgrowth, coincident with enhanced PKA and CREB phosphorylation (Deyts et al., 2016a). Consistent with their earlier work, they found that overexpressing β-CTF fragments of APP (C99) also stimulated outgrowth, whereas a C99 construct with a mutated BBXXB motif did not. Lastly, they showed that treatment with an AC inhibitor prevented increased outgrowth and phosphorylated CREB levels in their assays, again implicating Gαs-dependent signaling. Whether Gαs endogenously interacts with APP-CTFs in healthy neurons and whether this interaction is perturbed over the course of AD remains to be explored. Nevertheless, given available evidence that Gαo normally interacts with full-length APP but not its fragments in neurons (as summarized above), these results support the intriguing view that APP cleavage might induce a novel type of G protein switching (Tucek et al., 2002; Woehler and Ponimaskin, 2009), whereby the holoprotein signals as a transmembrane receptor specifically via Gαo, while its CTF fragments can selectively regulate Gαs-dependent pathways (**Figure 2C**). In the context of neuronal development, this model might also help explain how APP-dependent signaling can promote neuronal motility in some contexts while restricting it in others.

## CONCLUSION AND PERSPECTIVE: LIGAND-DEPENDENT MODULATION OF APP-Gαo SIGNALING

Despite considerable efforts to establish a role for aberrant APP-Gαo signaling in AD, proof for this model has been hindered

by incomplete understanding of the mechanisms that normally regulate this pathway in the brain. Because past studies often relied on rather artificial assays and overexpression systems, it is still unclear whether hyperstimulating this pathway results in the misregulation of endogenous signaling responses or produces novel gain-of-function effects that normally do not occur in the brain. Our laboratories have now approached this issue using complementary strategies, with the goal of understanding how this evolutionarily conserved signaling pathway regulates neuronal functions in both the developing and mature nervous system. As summarized in **Figure 2A**, sAPPα ectodomain fragments are clearly able to activate the PI3K/Akt pathway and modulate neuronal stress signaling, a response that undoubtedly plays important roles in both the developing and adult brain (Kögel et al., 2012; Milosch et al., 2014). By comparison, Contactin-dependent activation of APP-Go signaling can regulate the motile behavior of developing neurons (**Figure 2B**), in part by modulating Ca2<sup>+</sup> influx and downstream effectors that modulate cytoskeletal dynamics (Horgan and Copenhaver, 1998; Copenhaver and Ramaker, 2016). Evidence that CTF fragments might also regulate neuronal behavior via Gαs (**Figure 2C**) suggests that G protein switching could also contribute to the refinement of APP-dependent motile responses (Deyts et al., 2012, 2016a).

We postulate that our different experimental preparations have revealed an important aspect of APP-Go signaling: namely, that the integration of this pathway with alternative or complementary effectors can be strongly influenced by particular combinations of ligands and co-receptors for APP that are expressed in a context-dependent manner. As has been reviewed elsewhere, APP family proteins can interact with a wide variety of candidate binding partners (Hoe et al., 2009; Jacobsen and Iverfeldt, 2009; Rice et al., 2013; Deyts et al., 2016b), although most of these interactions have yet to be validated in vivo. For example, experiments using cultured neurons have shown that stimulation with sAPPα can promote APP-dependent outgrowth via interactions with members of the integrin and L1CAM families (Osterfield et al., 2008; Young-Pearse et al., 2008), a response that can be further modulated by extracellular proteins like Reelin, F-spondin, and Semaphorin 3A (Ho and Sudhof, 2004; Hoe et al., 2009; Magdesian et al., 2011). More recently, elegant work by Young-Pearse and colleagues showed that different members of the pancortin family can both promote and inhibit APP-dependent responses in migrating cortical neurons, possibly via a combination of direct and indirect interactions (Rice et al., 2012). Whether these interactions also regulate Go-dependent aspects of motility remains to be explored. Outside the nervous system, APP family proteins are strongly upregulated by keratinocytes during wound healing (Herzog et al., 2004), while treatment with sAPPα stimulates their motile behavior (Kirfel et al., 2002), although it is unclear if this response is transduced by APP or other receptors. From a developmental perspective, ample precedent for this model of APP-Go signaling can be found in the responses elicited by other neuronal guidance receptors that can both stimulate and inhibit outgrowth, depending on a variety of interacting factors (Nishiyama et al., 2003; Egea and Klein, 2007; Yoshida, 2012; Finci et al., 2014; Kaplan et al., 2014). Likewise, whether activation of APP-Gαo signaling induces neuroprotective or neurotoxic responses might be strongly affected by convergent input from physiological stimuli (particularly sAPPα) or pathological factors (including Aβ<sup>42</sup> oligomers).

Lastly, it should be noted that APP expression is significantly altered in a variety of other diseases besides AD. In Down syndrome (DS), trisomy 21 results in a triplication of the gene encoding APP (as well as many other genes; Antonarakis et al., 2004), and most DS patients exhibit accelerated Aβ accumulation and develop AD-like neurological pathologies (Millan Sanchez et al., 2012; Castro et al., 2016). APP expression is also dramatically upregulated in the brain following traumatic brain injury (Plummer et al., 2016; Acosta et al., 2017) and in lesions associated with epilepsy and multiple sclerosis (Noebels, 2011; Matias-Guiu et al., 2016). Whether APP serves a neuroprotective function or promotes degenerative responses in these diseases is still unknown; hence, determining how APP-Gαo signaling is altered in AD should also be relevant to other conditions in which this pathway might be misregulated. Only by fully defining the normal mechanisms of APP-Go signaling in the brain will it be possible to resolve how the misregulation of this pathway may contribute to the pathological sequelae that give rise to AD.

#### AUTHOR CONTRIBUTIONS

PC and DK contributed equally to all aspects of this review, including development of the overall concept, writing and correcting the text, and creating the table and figures included in the review.

# FUNDING

Work from the Copenhaver laboratory was funded in part by NIH grants NS078363 and AG025525 to PC, who also received support from OHSU Presidential Bridge Funding Award. Work from the Kögel lab was funded by the Deutsche Forschungsgemeinschaft (DFG, grants KO 1898/6-1 and 10/1). The authors declare no competing financial interests.

#### ACKNOWLEDGMENT

We thank Dr. Doris Kretzschmar for critical input on this review.

# REFERENCES


precursor causes sustained apoptosis signal-regulating kinase 1/c-Jun NH2 terminal kinase-mediated neurotoxic signal via dimerization. J. Pharmacol. Exp. Ther. 306, 889–902. doi: 10.1124/jpet.103.051383



precursor protein in the cerebrospinal fluid of Alzheimer's disease patients. Neurosci. Lett. 278, 169–172. doi: 10.1016/S0304-3940(99)00929-5


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Copenhaver and Kögel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Functional Roles of the Interaction of APP and Lipoprotein Receptors

Theresa Pohlkamp1,2† , Catherine R. Wasser 1,2† and Joachim Herz 1,2,3,4 \*

<sup>1</sup>Department of Molecular Genetics, UT Southwestern Medical Center, Dallas, TX, USA, <sup>2</sup>Center for Translational Neurodegeneration Research, UT Southwestern Medical Center, Dallas, TX, USA, <sup>3</sup>Department of Neuroscience, UT Southwestern Medical Center, Dallas, TX, USA, <sup>4</sup>Department of Neurology and Neurotherapeutics, UT Southwestern Medical Center, Dallas, TX, USA

The biological fates of the key initiator of Alzheimer's disease (AD), the amyloid precursor protein (APP), and a family of lipoprotein receptors, the low-density lipoprotein (LDL) receptor-related proteins (LRPs) and their molecular roles in the neurodegenerative disease process are inseparably interwoven. Not only does APP bind tightly to the extracellular domains (ECDs) of several members of the LRP group, their intracellular portions are also connected through scaffolds like the one established by FE65 proteins and through interactions with adaptor proteins such as X11/Mint and Dab1. Moreover, the ECDs of APP and LRPs share common ligands, most notably Reelin, a regulator of neuronal migration during embryonic development and modulator of synaptic transmission in the adult brain, and Agrin, another signaling protein which is essential for the formation and maintenance of the neuromuscular junction (NMJ) and which likely also has critical, though at this time less well defined, roles for the regulation of central synapses. Furthermore, the major independent risk factors for AD, Apolipoprotein (Apo) E and ApoJ/Clusterin, are lipoprotein ligands for LRPs. Receptors and ligands mutually influence their intracellular trafficking and thereby the functions and abilities of neurons and the blood-brain-barrier to turn over and remove the pathological product of APP, the amyloid-β peptide. This article will review and summarize the molecular mechanisms that are shared by APP and LRPs and discuss their relative contributions to AD.

Keywords: LRP, APOE, LDL receptor gene family, neuromuscular junction, synapse, glutamate receptors, trafficking, amyloid beta

## LIPOPROTEIN RECEPTORS

#### Structure and General Physiological Properties

Besides the important role in lipid metabolism, members of the low-density lipoprotein (LDL) receptor family take part in a broad range of pre- and post-developmental functions in brain and play key roles in the pathogenesis of Alzheimer's disease (AD). Much like the amyloid precursor protein (APP), members of the LDL receptor family are type-I membrane receptors with single-pass transmembrane (TM) domains that can be endocytosed, proteolytically processed and participate in a variety of protein interactions both inside and outside of the cell, including direct interactions with APP (May et al., 2005; Dieckmann et al., 2010). Lipoprotein receptors are involved in various mechanisms of APP-processing and Aβ-clearance in several cell types including neurons, astrocytes, endothelial cells of the blood brain barrier (BBB), and ependymal cells of the blood cerebrospinal fluid (CSF) barrier (BCSFB; reviewed by Hoe and Rebeck, 2008; Marzolo and Bu, 2009; Wagner and Pietrzik, 2012; Lane-Donovan et al., 2014).

#### Edited by:

Thomas Deller, Goethe-University, Germany

#### Reviewed by:

Eckart Förster, Ruhr University Bochum, Germany Claus Pietrzik, University of Mainz, Germany

> \*Correspondence: Joachim Herz

joachim.herz@utsouthwestern.edu

†These authors have contributed equally to this work.

Received: 04 January 2017 Accepted: 16 February 2017 Published: 01 March 2017

#### Citation:

Pohlkamp T, Wasser CR and Herz J (2017) Functional Roles of the Interaction of APP and Lipoprotein Receptors. Front. Mol. Neurosci. 10:54. doi: 10.3389/fnmol.2017.00054

In the peripheral and central nervous system, lipoprotein receptors and APP interact to control developmental processes and synaptic function. These lipoprotein receptors are highly conserved—at least as far back in evolution as C. elegans (Yochem and Greenwald, 1993)—and are related by both structure and function (Krieger and Herz, 1994; **Figure 1**). The seven core members of this receptor family are the LDL receptor (Ldlr), Apolipoprotein E (ApoE) receptor 2 (Apoer2/Lrp8), very-LDL receptor (Vldlr), LDL receptor-related protein 1 (Lrp1), Lrp1b, Lrp2/Megalin and multiple epidermal growth factor (EGF) repeat containing protein 7 (Megf7/Lrp4; Dieckmann et al., 2010). Structurally, the extracellular domain (ECD) of each of the core LDL receptor family members is composed of a combination of two types of conserved domains: (1) ligand binding-type repeat domains (LBDs); and (2) EGF-precursor homology domains. The amino-terminal LBD domain confers ligand specificity, consisting of cysteine-rich complement-type ligand binding-type repeats (LBRs, sometimes called type A repeats). The EGF-precursor domains participate in the pH-dependent release of bound ligands after endocytosis and contain a mixture of EGF receptor-like repeats (EGF-repeats) and YWTD (Tyr-Trp-Thr-Asp) β-propeller repeats (Beglova and Blacklow, 2005; Andersen et al., 2013; reviewed in Li et al., 2001). The intracellular domain is less conserved between the family members, but each of the core members contain at least one NPxY (Asn-Pro-X-Tyr) motif that functions in protein interaction/signal transduction (Trommsdorff et al., 1998; Howell et al., 1999; Gotthardt et al., 2000) and endocytosis (Chen et al., 1990).

The smaller receptors within the LDL receptor family, Ldlr, Vldlr and Apoer2, contain only one EGF-precursor domain and have a juxtamembraneous domain rich in serine and threonine residues, which serve as sites for O-linked glycosylation (Kingsley et al., 1986; Sakai et al., 1994; Christie et al., 1996; Kim et al., 1996). This O-linked sugar (OLS) domain is alternatively spliced in both Apoer2 and Vldlr (Sakai et al., 1994; Kim et al., 1997; Clatworthy et al., 1999), and inclusion of the OLS-domain hinders the proteolytic processing of the receptors (Magrané et al., 1999; May et al., 2003; Wasser et al., 2014). However, for Apoer2 it was shown that exclusion of the OLS-domain produces ''cleavage-resistant'' Apoer2 splice variants, as the OLS-domain is likely the site of the initial extracellular cleavage that precedes further processing by γ-secretase (Wasser et al., 2014).

Additional somewhat distant members are Lrp5 and Lrp6 as well as the Sortilin-related receptor with LDLR class A repeats (SorLA; **Figure 1**). Lrp5 and Lrp6 (called arrow in D. melanogaster) encode four EGF-precursor domains but lack N-terminal LBDs and intracellular NPxY-motifs (Brown et al., 1998; Nakagawa et al., 1998; Wehrli et al., 2000). SorLA (SorL1/LR11/Lrp11), with multiple LBDs and one EGF-precursor domain, is a hybrid-LDL receptor family member in that it has an additional Vps10p-sorting domain and Fibronectin repeats (Jacobsen et al., 1996). In addition, SorLA has one NPxY-related retromer binding motif (FANSHY; Phe-Ala-Asn-Ser-His-Tyr; Fjorback et al., 2012). Containing three to five LBRs and no other typical LDL receptor domains, the most distant relatives are Lrad3 (Ranganathan et al., 2011) as well as Lrp3 (Ishii et al., 1998), Lrp10 (murine Lrp9; Sugiyama et al., 2000) and Lrp12 (ST7/Mg13; Battle et al., 2003), which have two additional CUB domains.

#### Genetics

Despite the high degree of homology between the receptors and the overlapping expression pattern and function, the majority of these receptors are indispensable for survival or proper brain function. In fact, deletion of Lrp1 (Herz et al., 1992), Lrp1b (Dietrich et al., 2014), Lrp2 (Willnow et al., 1996), Lrp4 (Weatherbee et al., 2006) or Lrp6 (Pinson et al., 2000) in the mouse lead to embryonic or postnatal death with complete or high penetrance. While mice lacking Lrp5 (Fujino et al., 2003), Ldlr (Shimada et al., 1996), Apoer2 or Vldlr (Trommsdorff et al., 1999), or the distant member SorLA (Andersen et al., 2005) survive, they all have abnormalities in cholesterol homeostasis and/or brain development. Of the most distant relatives, gene silencing of Lrp12 leads to defects in brain lamination (Grote et al., 2016), yet to date in vivo knockouts or knockdowns of the more distant members Lrp3, Lrp10 and Lrad3 have not been reported.

# LIPOPROTEIN METABOLISM AND ALZHEIMER'S DISEASE

One percent of all AD cases are early onset (EOAD) generally manifesting from mutations in APP or APP processing genes and leading to increased production of the toxic APP cleavage product, amyloid β (Aβ). The other 99% of cases are late-onset AD (LOAD) with increased Aβ-levels and deposition that are apparently independent from EOAD-like mutations in APP/APP processing genes. Instead, the leading cause in LOAD appears to be an imbalance between Aβ production and clearance from the brain (Weller et al., 2008; Mawuenyega et al., 2010). Thus, it is important to understand the various mechanisms by which LDL receptor family members and their ligands clear Aβ.

Aside from age, the most important risk modifier for developing LOAD is ApoE (Corder et al., 1993). ApoE is a major cholesterol transporter in the brain and in the circulation. In humans there are three ApoE alleles: ε2, ε3, and ε4 (ApoE2, 3 and 4, respectively). ApoE3 is the most abundant allele and understood as the neutral isoform with regards to ADphysiology, the least abundant isoform ApoE2 appears to be protective against AD (Corder et al., 1994; Conejero-Goldberg et al., 2014). Importantly, the ε4 allele of ApoE (ApoE4) dramatically reduces the age of AD onset and is carried by >50% of those afflicted with the disease (Corder et al., 1993), despite an allele frequency of only ∼15% in the general population (Utermann et al., 1980). Therefore ApoE4 is the most prevalent, biomedically important risk allele for LOAD.

The brain is the most cholesterol-rich organ, containing approximately 25%–30% of the body's total cholesterol (Dietschy and Turley, 2001), and high serum cholesterol levels correlate with cognitive impairment and AD (Zambón et al., 2010; Di Paolo and Kim, 2011). Interestingly, evidence from in vivo studies suggests that altered serum cholesterol levels affect the processing of APP as well as the neurotoxicity and clearance

NPxY-lacking Lrp5/Lrp6 and hybrid SorLA with additional Fibronectin repeats (pink) and importantly the VPS10p-sorting motif (green). Four very distant "far side" proteins (right, Lrp3, Lrp10, Lrp12, and Lrad3) only encode ligand binding-type repeats. Lrp3, Lrp10 and Lrp12 also contain atypical CUB-domain (binds Complement, Uegf and Bmp1). In addition to the OLS domains of Apoer2 and Vldlr, alternative splicing of Apoer2 produces splice variants lacking N-terminal ligand binding type repeats (repeats 4–6; Brandes et al., 2001; gray).

Apoer2 express an additional extracellular O-linked sugar (OLS) domain adjacent to the transmembrane (TM) segment. The more distant members (middle) are the

of Aβ (Reed et al., 2014). Despite this, the role of cholesterol metabolism in the pathogenesis of AD is not well understood.

The cholesterol metabolism link to AD pathogenesis is further supported by additional genome-wide association studies that implicate other apolipoproteins and their receptors as AD risk factors. In addition to ApoE, a variety of SNPs in ApoJ/Clusterin from several populations are associated with LOAD (Harold et al., 2009; Bagyinszky et al., 2014). Other apolipoprotein polymorphisms associated with AD have been reported in ApoA-I (Shibata et al., 2013), ApoA-IV (Császár et al., 1997), ApoC-I (Ki et al., 2002; Zhou et al., 2014; Shang et al., 2015), ApoC-II (Schellenberg et al., 1992), ApoC-III (Sun et al., 2005) and ApoD (Shibata et al., 2013). Among the LDL receptor family members, mutations in SorLA (Meng et al., 2007; Bagyinszky et al., 2014) appear to impart the most dramatic risk for developing AD. Aside from SorLA, Lrp1 (Kang et al., 1997), Lrp1b (Shang et al., 2015), Lrp2 (Wang et al., 2011), Lrp4 (Vargas et al., 2010), Lrp6 (De Ferrari et al., 2007) and Apoer2 (Ma et al., 2002) have been associated with AD risk. Furthermore, a non-LDL receptor family member, Trem2 (triggering receptor expressed on myeloid cells 2), is an alternative receptor for apolipoproteins, including ApoE and ApoJ/Clusterin, and has recently been identified as high risk factor for LOAD (Jin et al., 2015). In sum, cholesterol metabolism and the homeostasis/signaling of lipoprotein receptors and their ligands appear to be inextricably linked to the pathogenesis of LOAD.

With diverse functions including gathering nutrients and clearing toxic, useless debris from the extracellular space, as well as mediating intracellular trafficking/signaling and even transcription, the indispensable nature of many of the lipoprotein receptors is not surprising. Most of these receptors play some part in APP processing or clearance of Aβ, affecting the balance between Aβ-production and clearance. Understanding how these lipoprotein receptors and their ligands influence the homeostasis of Aβ production/clearance individually, as well as in unison, will prove crucial for not only elucidating mechanisms of AD pathogenesis, but also the design of potential therapeutic interventions to counteract the disease. In this chapter, we will focus on lipoprotein receptors and their role in AD pathogenesis through regulating APP processing and Aβ clearance.

## Ldlr

## Structure and General Physiological Properties

Ldlr, the founding member of the LDL receptor family, is ubiquitously expressed throughout the body, where it plays a key role in regulating cholesterol homeostasis (reviewed in Go and Mani, 2012). The receptor clusters after binding cholesterol-rich LDL particles and mediates cholesterol uptake through clathrin-mediated endocytosis of the lipoprotein-bound receptor (reviewed in Brown and Goldstein, 1979). Mutations in the Ldlr gene are responsible for familial hypercholesterolemia (FH), a disease in which Ldlr function is impaired, leading to increased plasma cholesterol concentrations and causing premature cardiovascular disease (Hobbs et al., 1990; Fass et al., 1997).

#### Genetics

While impaired Ldlr function in humans leads to elevated plasma cholesterol and premature cardiovascular disease due to reduced uptake of cholesterol-rich LDLs (Hobbs et al., 1990; Fass et al., 1997), the effect in mice is similar yet less severe (Ishibashi et al., 1993; Osono et al., 1995). In the CNS, where Ldlr is expressed higher in astrocytes than in neurons, Ldlr also plays a role in cholesterol homeostasis in the brain. Ldlr knockout mice display some synaptic and learning deficiencies (Mulder et al., 2004, 2007; de Oliveira et al., 2011, 2013, 2014; Moreira et al., 2012). Interestingly, murine ApoE expression is elevated in the CSF of mice lacking Ldlr, and this phenotype is even more dramatic in mice carrying the human ApoE3 and ApoE4 isoforms of ApoE (Fryer et al., 2005). Ldlr deficiency also leads to elevated neuroinflammatory responses and oxidative stress (Thirumangalakudi et al., 2008; Katsouri and Georgopoulos, 2011), which might be further exacerbated by a high cholesterol diet (Ettcheto et al., 2015).

#### Biochemistry and Cellular Function

As cholesterol metabolism is linked to AD and regulated by Ldlr, Ldlr knockout mice have been used as a model organism to study the interplay between cholesterol and Aβ-deposition in several studies. While Ldlr has no known direct or indirect interaction with APP or APP processing, Ldlr binds to Aβ and mediates its clearance by degradation in astrocytes, but does not alter APP processing (Kim et al., 2009). Ldlr knockout mice are more susceptible to Aβ-induced neurotoxicity, when Aβ is injected into the hippocampus (de Oliveira et al., 2014). Aβ-deposition is exacerbated with Ldlr-deficiency in AD mice (Tg2576 and APP/PS1; Cao et al., 2006; Katsouri and Georgopoulos, 2011) and is attenuated with Ldlr overexpression on an APP/PS1 background due to enhanced clearance (Kim et al., 2009). The additional knockout of ApoE does not affect the Aβ levels in Ldlr-deficient AD mice (APP/PS1; Katsouri and Georgopoulos, 2011), and this was confirmed by an in vitro study in astrocytes demonstrating that the clearance of Aβ is independent of ApoE (Basak et al., 2012). This suggests that the Ldlr-dependent glia response in Aβclearance is independent of ApoE despite Ldlr being a strong ApoE receptor (Katsouri and Georgopoulos, 2011; Basak et al., 2012). Nonetheless, Castellano et al. (2011) showed that Aβ turnover in the mouse brain in vivo is strongly dependent upon ApoE isoform, indicating that other mechanisms besides Ldlr-mediated Aβ removal are responsible for Aβ homeostasis in the intact brain.

#### Lrp1

# Structure and General Physiological Properties

The second receptor identified in the LDL receptor family, Lrp1 (Herz et al., 1988) is one of the largest (∼600 kDa) and most versatile members as it is known to bind over 100 different ligands (Herz and Strickland, 2001; Gonias and Campana, 2014). Lrp1 can be processed by the same enzymes as APP: ADAM10 (Nakajima et al., 2013), BACE1 (von Arnim et al., 2005) and γ-secretase (May et al., 2002; May and Herz, 2003; Zurhove et al., 2008). The sequential processing of Lrp1 first produces a soluble Lrp1-ECD, followed by a γ-secretase-mediated release of the Lrp1-ICD (May et al., 2002). The Lrp-ECD is capable of binding Lrp1 ligands (Quinn et al., 1997), and the Lrp1-ICD can translocate to the nucleus and regulate gene transcription (Zurhove et al., 2008). Of note, this Lrp1-ICD-mediated transcriptional regulation might be relevant to neuroinflammation (Zurhove et al., 2008), which is emerging as a common factor in many neuropathological conditions including AD (Heneka et al., 2015; Chen et al., 2016). Lrp1 also undergoes rapid, constitutive recycling; despite the two NPxY motifs in the Lrp1 cytoplasmic tail, a YxxL motif in the intracellular domain of Lrp1 is the dominant and main mediator of Lrp1 endocytosis—unlike other lipoprotein receptors, where the NPxY motifs mediate this process (Li et al., 2000). In addition to the liver and vasculature, Lrp1 is highly expressed in the brain (Rebeck et al., 1993) where it plays essential roles in signal transduction and endocytosis (Herz and Strickland, 2001; May et al., 2004). During brain development, it modulates radial glia stem cell proliferation, survival and differentiation (Safina et al., 2016). Importantly, Lrp1 can regulate the amyloidogenic processing of APP as well as the clearance of Aβ, which implicates Lrp1 as a key participant in the pathogenesis of AD (Kounnas et al., 1995; Ulery et al., 2000; Van Uden et al., 2000).

#### Genetics

Global Lrp1 knockout mice are embryonically lethal (Herz et al., 1992, 1993). Lrp1 gene polymorphisms have been associated with a premature risk of cardiovascular disease in patients with familial hypercholesterolemia/FH (Aledo et al., 2012) and abnormal inflammatory responses in fibroblasts (Klar et al., 2015).

#### Biochemistry and Cellular Function

Lrp1 directly interacts with APP extracellularly and regulates the localization and processing of APP (Kounnas et al., 1995). In several cell lines, depletion of the rapidly recycling Lrp1 reduced Aβ production (Ulery et al., 2000; Pietrzik et al., 2002). In vivo, overexpression of a minireceptor of Lrp1 (EGF-precursor domain-II, TM-domain, and ICD-domain) in an AD mouse model (PDAPP) increased soluble brain Aβ (Zerbinatti et al., 2004); however, reduced levels of Lrp1 in hippocampal neurons of another AD mouse model (APP/PS1) had no effect on Aβ production (Xu et al., 2012).

The extracellular interaction of Lrp1 and APP only occurs with APP isoforms containing the Kunitz protease inhibitor (KPI) domain and promotes the internalization of APP (Kounnas et al., 1995; Billnitzer et al., 2013). The KPI domain is present in the longer APP isoforms (APP<sup>770</sup> and APP751) but not in the shortest, principally neuronal isoform (APP695), which is the dominant isoform in the brain (reviewed in Nalivaeva and Turner, 2013). This Lrp1-APP interaction can be blocked with the chaperone and Ldlr receptor family member antagonist, RAP (receptor-associated protein; Kounnas et al., 1995; Kinoshita et al., 2001). In hippocampal neurons, RAP treatment inhibited axonal branching due to increased APP on the cell surface that signals via complex formation with Fe65 and Mena (Ikin et al., 2007; Billnitzer et al., 2013). In APP knockout neurons, which have increased axonal branching compared to wildtype, RAP treatment had an additive Erk2-associated effect on branching (Billnitzer et al., 2013).

Intracellular interactions with APP and Lrp1 also appear important in modulating the amyloidogenic processing of APP. Both Fe65 and Dab1 interact with Lrp1 NPxY motifs and modify intracellular signal transduction (Trommsdorff et al., 1998; Gotthardt et al., 2000; Kinoshita et al., 2001; Pietrzik et al., 2004). These adaptors also bind APP (Fiore et al., 1995; Trommsdorff et al., 1998). The cytoplasmic adaptor protein, Fe65, links APP to Lrp1 and enhances amyloidogenic processing of APP (Pietrzik et al., 2002; Kinoshita et al., 2003; Yoon et al., 2005; Klug et al., 2011). Dab1 can interfere with this Lrp1/Fe65/APP complex by competing with Fe65 for Lrp1 binding, thereby reducing amyloidogenic APP processing (Kwon et al., 2010). Of note, the ICD of APP along with Fe65 translocates to the nucleus where it suppresses Lrp1 transcription (Liu et al., 2007). APP and Lrp1 also share other cytoplasmic interactions, one of which is with the endosomal sorting nexin 17 (Snx17). Snx17 interacts with the NPxY motifs in Lrp1 and APP to regulate their recycling from early endosomes back to the cell surface (Lee et al., 2008; Donoso et al., 2009; Farfán et al., 2013).

Despite promoting neuronal Aβ production, Lrp1 participates in Aβ clearance (reviewed in Kanekiyo and Bu, 2014). Lrp1 binds Aβ, with higher affinity for Aβ<sup>40</sup> than Aβ<sup>42</sup> (Shibata et al., 2000; Storck et al., 2016). Within the brain, Lrp1 endocytoses Aβ from the extracellular space and directs it to the lysosome for degradation (Kanekiyo et al., 2013). Lrp1 is also expressed in astrocytes and microglia where it is involved in Aβ-clearance (reviewed in Ries and Sastre, 2016). Another major Aβ clearance mechanism involves the transcytosis of Aβ from the brain to the circulation via the BBB (Marques et al., 2013). Lrp1 gene silencing reduced the clearance of intracerebroventricularlyinjected Aβ across the BBB in wildtype mice (Jaeger et al., 2009). Furthermore, an endothelial (brain and choroid plexus) specific Lrp1 knockout revealed that Lrp1 preferentially clears Aβ40, as these mice accumulated Aβ<sup>40</sup> faster and demonstrated reduced spatial memory (Storck et al., 2016), which is a common phenotype observed with high levels of Aβ. Moreover, Lrp1 cleavage by ADAM10 has opposing effects as well; whereas soluble Lrp1 in the brain inhibits Aβ clearance, in the periphery it could provide a sink for Aβ monomers. Inhibition of ADAM10 reduces Lrp1 ectodomain shedding, thereby promoting Aβ-clearance across the BBB, especially Aβ<sup>40</sup> (Shackleton et al., 2016); however, ADAM10 cleavage of Lrp1 also leads to the segregation of soluble Lrp1 into the periphery where it has been described to prevent the reentering of Aβ monomers into the brain (Sagare et al., 2007). Recently it was found that another AD risk gene, PICALM, plays a central role in BBB transcytosis of Aβ, and it has been reported that extracellular binding of Aβ to Lrp1 induces an intracellular conformational change allowing for PICALM binding and endocytosis of the entire complex (Zhao et al., 2015).

Importantly, both the Vldlr- and Lrp1-mediated Aβ clearance mechanisms via the BBB are differentially slowed down by ApoEisoforms: ApoE4 > ApoE2 or ApoE3 (Deane et al., 2008). Besides clearance of Aβ, Lrp1 can compete with APP for BACE1 (von Einem et al., 2010) and γ-secretase (Lleó et al., 2005) cleavage. Taken together, it appears that Lrp1 contributes to the Aβhomeostasis in two opposing ways: whereas Lrp1 promotes intraneuronal APP processing towards Aβ (**Figure 2**), Lrp1 also provides an important clearance mechanism of Aβ across the BBB and/or BCSFB (**Figure 3**).

# Lrp1b (LRP-DIT)

# Structure and General Physiological Properties

Lrp1b is very similar to Lrp1 in overall structure and sequence (∼59% identical). Where Lrp1b differs most from Lrp1 is an extra LBR in the ECD and a 33 amino acid insert in the ICD (Liu et al., 2000). Lrp1b was first associated with tumorigenesis, but is also highly expressed in the adult brain (Liu et al., 2000; Haas et al., 2011) and retains APP at the cell surface reducing Aβ production (Cam et al., 2004).

#### Genetics

Mutations in Lrp1b are associated with multiple different types of cancer (Liu et al., 2000; Langbein et al., 2002; Sonoda et al., 2004), including gliomas (Roversi et al., 2006). Lrp1b-deficiency leads to embryonal lethality (Dietrich

Lrp1 exceeding that of Lrp1b by many-fold. Both bind Fe65, connecting them in a complex APP, and have opposite effects on APP processing. The fast endocytosis rate of Lrp1 increases the exposure of APP to the endosomal β- (BACE1, β) and γ-secretase (γ), producing Aβ (green tears) and soluble APPβ (sAPPβ) fragment. Another intraendosomal sorting receptor of the LDL receptor family, SorLA, can bind and reroute receptors from the endosome back to the trans-Golgi network (TGN), where it is either sequestered, sorted back to the cell surface, or sent to the lysosome for degradation. Apoer2, which also recycles slowly, binds Fe65 via its NPxY-motif, promoting APP surface stability and decrease amyloidogenic processing. Additionally, simultaneous binding of the secreted, extracellular ligand, F-spondin, to the ECDs of APP and Apoer2 also promotes APP stability at the surface.

et al., 2010). Like Lrp4 knockins expressing a truncated ECD (see ''Lrp4'' Section for details), a similar truncation of Lrp1b allows animals to survive, be fertile and develop mostly normal. However, in contrast to Lrp4-ECD (Pohlkamp et al., 2015) mice, synaptic plasticity in hippocampal field recording is not affected in Lrp1b-ECD mice (Marschang et al., 2004).

# Biochemistry and Cellular Function

Lrp1b binds to fibrinogen and ApoE carrying proteins (Haas et al., 2011). In total, Lrp1 and Lrp1b share numerous ligands. Lrp1b also binds APP at the extracellular KPI-containing domain (Cam et al., 2004). With an internalization rate of more than 10 min for Lrp1b, the rate of endocytosis is much slower than Lrp1, which has a rate of less than 30 s (Liu et al., 2001). In contrast to overexpression of Lrp1 in a cell culture system, overexpression of Lrp1b increased APP surface expression, resulting in enhanced non-amyloidogenic α-secretase cleavage and reduced Aβ production (Cam et al., 2004). Based on these in vitro findings, a model for the Lrp1- vs. Lrp1b-effect on APP processing was proposed by Wagner and Pietrzik (2012), where fast Lrp1 uptake shifts APP processing from α-cleavage towards the endosomal toxic

#### FIGURE 3 | Continued

plexus functions to produce and filter CSF. This filtration removes metabolic waste, excess neurotransmitters and foreign/toxic particles, such as Aβ, which is mainly produced by neurons (see Figure 2). Apolipoproteins, such as ApoE and ApoJ/Clusterin (yellow dots), mainly secreted from astrocytes ("Astro"), bind circulating interstitial Aβ. These Aβ-laden apolipoproteins then bind lipoprotein receptors (red) and mediate their cellular uptake. ApoJ/Clusterin is eliminated rapidly across the BCSFB by ependymal Lrp2 (light red), facilitating the clearance of Aβ via lysosomal degradation in ependymal cells and subsequent exocytosis into the CSF, where soluble Lrp2 (sLrp2) has been detected (Spuch et al., 2015). BACE1 is the enzyme that processes Lrp2 and Lrp1 to release sLrp2 and sLrp1, respectively. BACE1 is also found in the choroid plexus (Crossgrove et al., 2007; Liu et al., 2013). Other lipoprotein receptors (dark red, most notably Lrp1) then transport Aβ and the apolipoproteins across the endothelial cells from the CSF to the blood vessels of the choroid plexus. sLrp1 can also be detected in plasma, albeit its origin there is mainly peripheral.

β- and γ-cleavage-pathway, whereas Lrp1b-APP interaction results in prolonged surface time and increased α-cleavage of APP (**Figure 2**). However, it is important to note that while Lrp1, not Lrp1b, is likely to promote intracellular Aβproduction, it is conversely important for Aβ-clearance across the BBB.

#### Apoer2 (Lrp8) AND Vldlr

#### Structure, General Physiological Properties and Genetics

Both Apoer2 and Vldlr are quite similar in size and domain composition to Ldlr (**Figure 1**; Kim et al., 1996). The sequence identity between Vldlr and Apoer2 is approximately 50% (Kim et al., 1996 and reviewed in Reddy et al., 2011). Apoer2 has seven ligand-binding repeats, one less than Vldr, and contains a unique alternatively-spliced proline-rich domain not found in Vldlr (Kim et al., 1997; Clatworthy et al., 1999; Sun and Soutar, 1999). In the brain, Apoer2 only contains five ligandbinding domains due to alternative-splicing of exon 5 (Kim et al., 1997; Clatworthy et al., 1999; Sun and Soutar, 1999). The site of least homology between the Apoer2 and Vldlr is the OLS domain (Kim et al., 1996). As mentioned above, the OLS domain is alternatively-spliced in both receptors. For both receptors, splice variants containing the OLS domain are highly glycosylated, and this glycosylation inhibits proteolytic processing (Magrané et al., 1999; May et al., 2003; Wasser et al., 2014). For Vldlr, splice variants lacking this glycosylated domain undergo rapid proteolytic cleavage (Magrané et al., 1999). Unlike Vldlr, the OLS domain is required for the initial extracellular cleavage of Apoer2 (presumably due to loss of the extracellular cleavage site), so Apoer2 variants lacking the OLS domain are actually resistant to proteolysis (Wasser et al., 2014).

Apoer2 and Vldlr are almost exclusively expressed in the brain where they act as receptors not only for ApoE but also for the neuromodulator Reelin (D'Arcangelo et al., 1999; Trommsdorff et al., 1999). Ligand binding increases the proteolytic processing of both receptors (Hoe and Rebeck, 2005). The proteolytic fragments of Apoer2 can inhibit further signaling, whereby the soluble ECD fragment acts

ependymal cells of the choroid plexus also facilitate Aβ removal. The choroid

(Continued)

as a dominant negative receptor (Koch et al., 2002) and the released ICD translocates to the nucleus and represses Reelin transcription (Balmaceda et al., 2014; Telese et al., 2015).

The signaling initiated by Reelin binding to Apoer2 and Vldlr plays essential roles during the development of the CNS and neuronal function through adulthood (Förster et al., 2010). During development, Reelin expressed and secreted from Cajal-Retzius cells modulates the cytoskeleton and mobility of migrating neurons (Frotscher et al., 2009) and ensures proper cortical, hippocampal and cerebellar lamination (Trommsdorff et al., 1999).

Apoer2 and Vldlr double knockout leads to a phenotype comparable to Reelin or Dab1 deficiency: mice develop strong ataxia, a smaller cerebellum, and defective lamination of cerebellum, cortex and hippocampus (Trommsdorff et al., 1999).

Cortical Cajal-Retzius cells die out after birth and the amount of hippocampal Cajal-Retzius cells dramatically thins out later during postnatal hippocampal maturation (Chowdhury et al., 2010). In total, the expression pattern changes so that in the cortex and hippocampus Reelin is now expressed in a more distributed fashion, mainly by subtypes of GABAergic interneurons (Drakew et al., 1998; Pesold et al., 1998; Pohlkamp et al., 2014). Besides neuronal migration, Reelin-signaling plays parts in both axo- (Leemhuis et al., 2010) and dendritogenesis (Assadi et al., 2003; Niu et al., 2004; Jossin and Goffinet, 2007; Zhang et al., 2007; Kawauchi and Hoshino, 2008; Matsuki et al., 2008; Chai et al., 2009; Ventruti et al., 2011) as well as synapse formation and function (Glantz and Lewis, 2000; Sinagra et al., 2005; Groc et al., 2007; Qiu and Weeber, 2007; Niu et al., 2008; Campo et al., 2009; Dumanis et al., 2011; Hellwig et al., 2011; Bal et al., 2013). In the adult brain, Reelin regulates synaptic function, plasticity and spatial learning and fear memory (Weeber et al., 2002; Beffert et al., 2005; Herz and Chen, 2006; Wasser et al., 2014).

Apoer2 and Vldlr bind Reelin and cluster together resulting in the phosphorylation of Dab1 and Src-kinase-mediated phosphorylation of NR2 subunits of the NMDA receptor (Hiesberger et al., 1999; Arnaud et al., 2003; Bock and Herz, 2003; Strasser et al., 2004), which requires a unique 59-amino acid insert in the Apoer2 cytoplasmic tail through direct interaction with PSD-95 (Beffert et al., 2005). Reelin-mediated NMDAR phosphorylation increases Ca2+-influx through NMDAR, resulting in increased activation of cAMP-response element binding protein (CREB; Chen et al., 2005) and the potent enhancement of long-term potentiation (LTP; Weeber et al., 2002). Hippocampal LTP is modestly reduced or severely perturbed in mice lacking Vldlr or Apoer2, respectively, and LTP is not enhanced by acute Reelin treatment in either mutant (Weeber et al., 2002).

There are several lines of evidence that implicate Reelin signaling as protective against AD pathogenesis. First, Reelinsignaling can counteract Aβ-induced synaptic suppression (Durakoglugil et al., 2009) by enhancing synaptic LTP, an effect that requires a unique alternatively spliced exon in the ICD of Apoer2 (Beffert et al., 2005). Interestingly, the AD-risk factor ApoE4 actually prevents this protective effect by sequestering the ApoE receptors along with other synaptic receptors in the endosome (Chen et al., 2010), and postnatal loss of Reelin exacerbates the cognitive deficits in AD mouse model (Lane-Donovan et al., 2015). In AD mice, Apoer2 and its ligand Reelin are localized in fine granular structures and reactive astrocytes surrounding Aβ plaques (Wirths et al., 2001; Motoi et al., 2004). Furthermore, both humans with AD and a transgenic AD mouse model have higher expression of the Apoer2 splice variant that lacks the alternatively spliced CTD, which would be predicted to impair the Reelin-mediated suppression of Aβ-toxicity (Hinrich et al., 2016). Treating these AD mice with antisense oligonucleotides designed to increase the inclusion of the alternatively spliced proline-rich domain in Apoer2 restored the expression of the functional Apoer2 variant and rescued their AD-related memory deficits (Hinrich et al., 2016).

#### Biochemistry and Cellular Function

Both Apoer2 and Vldlr interact with APP-binding proteins and influence the amyloidogenic processing of APP (reviewed Hoe and Rebeck, 2008; Marzolo and Bu, 2009; Wagner and Pietrzik, 2012; Lane-Donovan et al., 2014). Of the two receptors, Apoer2 interacts with a larger number of APP-binding proteins. Both APP and Apoer2 bind F-spondin (Ho and Südhof, 2004; Hoe et al., 2005) and Reelin (Hoe et al., 2009) extracellularly, as well as the intracellular adaptor proteins X11α/β (Borg et al., 1996; He et al., 2007), Fe65 (Fiore et al., 1995; Borg et al., 1996; Hoe et al., 2006a), Snx17 (Lee et al., 2008; Sotelo et al., 2014), Dab1 (Homayouni et al., 1999; Howell et al., 1999), and Dab2 (Cuitino et al., 2005; Lee et al., 2008). To date, Vldlr is known to directly interact with both Reelin and Fe65 (Dumanis et al., 2012) and immunoprecipitation results supported that Fe65 increases the interaction between APP and Vldlr in vivo, suggesting that Vldlr is involved in APP trafficking (Dumanis et al., 2012).

Ligand binding to Apoer2 induces homotypic clustering as well as clustering with other receptors, including APP (Divekar et al., 2014). The clustering of Apoer2 is weaker with ApoE binding compared to the clustering upon binding either Reelin or F-spondin (Divekar et al., 2014). ApoE inhibits γ-secretase cleavage of Apoer2 and APP (Irizarry et al., 2004; Hoe et al., 2006b), and ApoE3 imparted a greater inhibition than ApoE4 preventing the release of the Apoer2-ICD and APP intracellular domain (Hoe et al., 2006b). Interestingly, Apoer2 deficient mice express more ApoE and have elevated levels of the aggregation prone form of Aβ (Aβ42; Petit-Turcotte et al., 2005).

F-spondin is an extracellular ligand for both Apoer2 (Hoe et al., 2005) and APP (Ho and Südhof, 2004). This secreted extracellular protein, F-spondin, is composed of an amino-terminal Reelin and F-spondin domains followed by a thrombospondin domain, which contains six thrombospondin repeats (TSRs; reviewed in Feinstein and Klar, 2004). The central portion of the APP-ECD binds within the amino-terminal Reelin and F-spondin domains, while the LBD of Apoer2 binds the first four TSRs of F-spondin (Hoe et al., 2005). F-spondin stabilizes Apoer2 and APP at the cell surface, promoting α-cleavage of both proteins and reducing Aβ formation (Hoe et al., 2005). Of note, other LDL receptor family members-Vldlr, Lrp4 and Lrp2—also bind the first four TSRs of F-spondin (Zisman et al., 2007).

Like Lrp1, the NPxY domain of Apoer2 binds the cytosolic adaptor protein Fe65. While Lrp1 and Fe65 enhance Aβ production, Fe65 increases the interaction of APP and Apoer2 and decreases APP processing by stabilizing them at the cell surface (Hoe et al., 2006a). As Apoer2 and Lrp1 interact within the same region of Fe65, these two receptors may compete with each other for Fe65 binding and differentially influence APP processing (Hoe et al., 2006a). Dab1 also binds the NPxY motifs of Apoer2 and APP, and Aβ is decreased with Dab1 overexpression and increased in Dab1-deficient primary neurons (Hoe et al., 2006c).

Apoer2 directly interacts with APP extracellularly (Fuentealba et al., 2007). In Lrp1-deficient cells, Apoer2 promotes the cell surface retention of APP. This stabilization of APP requires cytoplasmic domain of Apoer2 (Fuentealba et al., 2007). Co-expression of Apoer2 with APP promotes APP surface expression and the lipid raft association of APP dependent on the Apoer2 CTD, but unexpectedly increased Aβ formation (Fuentealba et al., 2007). In contrast, X11α/βbinding to Apoer2 mediates ApoE induced endocytosis of APP and β-secretase resulting in APP processing and Aβ production (He et al., 2007), and Reelin can interrupt this interaction between X11α/β and Apoer2 (Minami et al., 2010), indicating another protective role of Reelin against Aβ toxicity.

## Lrp2 (MEGALIN/gp330)

# Structure and General Physiological Properties

Lrp2 is structurally very similar to Lrp1b and one of the most studied lipoprotein receptors in conjunction with AD. Similar to Lrp1, Lrp2 undergoes proteolytic processing to release the ECD followed by γ-secretase cleavage to release the ICD (Zou et al., 2004; Biemesderfer, 2006). The Lrp2-ICD contains sorting signals including three NPxY and a PPPSP motif that control Lrp2 surface expression specifically at cholesterol- and glycosphingolipid-rich regions (Marzolo et al., 2003). Besides binding to APP and ApoE, Lrp2 is also an important receptor for ApoJ/Clusterin, which is another genetic risk factor for AD. Lrp2 is expressed on endothelial cells of different organs, including capillaries in the brain and the ependymal cells of the choroid plexus, where it controls cholesterol homeostasis and Aβ-clearance (Willnow et al., 1996; Hammad et al., 1997; Chun et al., 1999; Bell et al., 2007). Besides its expression in endothelial and ependymal cells, Lrp2-expression has also been reported in dying neurons of postmortem brains of AD patients and cultured astrocytes (LaFerla et al., 1997; Bento-Abreu et al., 2008).

During neural tube formation and forebrain development Lrp2 is required for the dorsal to ventral gradient of the bone morphogenic protein 4 (BMP4) and sonic hedgehog (Shh). Lrp2 mediates endocytosis of Bmp4 for degradation and Bmp4 levels are increased in Lrp2-deficient mice (Spoelgen et al., 2005). Lrp2 is also a required co-receptor for Shh, ligand-binding induces a positive feedback loop and increased Shh-expression, thus Lrp2-deficiency leads to the loss of Shh expression in the ventral neuroepithelium (Christ et al., 2012). Finally, the loss of the Bmp4-Shh gradient in the neural tube causes holoprosencephaly, the failure of the brain to develop into two hemispheres (Spoelgen et al., 2005; Christ et al., 2012). Moreover, Shh and Lrp2 signaling regulates oligodendrocyte progenitor migration and proliferation in the optic nerve (Ortega et al., 2012) and glial cell specification during neural development (Wicher et al., 2005). The role of Lrp1 and Lrp2 in regulating neural stem cell and progenitor cell function has been reviewed in detail elsewhere (Auderset et al., 2016). However an implication of APP for these mechanisms has not been described.

#### Genetics

Lrp2-deficient mice die shortly after birth due to respiratory insufficiency. Lrp2 function is critical during neural tube formation, as it acts to organize Shh-mediated forebrain development during neurulation (Christ et al., 2012). Besides malfunctioning of endothelial tissues including lung and kidney, Lrp2-deficiency in neuroepithelium leads to impaired proliferation and forebrain fusion (Willnow et al., 1996). Endothelial cell specific Lrp2 deletion leads to impaired Aβ-clearance, which is described in more detail in the next section.

#### Biochemistry and Cellular Function

In the adult brain, Lrp2, facilitated by its ligand ApoJ/Clusterin, mediates Aβ clearance from the CSF (Hammad et al., 1997; Bell et al., 2007; **Figure 3**). As a part of the blood-CSF barrier (BCSFB), the choroid plexus takes part in the production and filtration of the CSF, including clearance of Aβ (**Figure 2**). Lrp2 is expressed within the choroid plexus, where it is sorted to the apical surface of ependymal cells within the lateral ventricles (Zheng et al., 1994; Chun et al., 1999; Willnow et al., 1999; Carro et al., 2005; Alvira-Botero and Carro, 2010). Despite a lack of AD pathology, mice lacking Lrp2 within these ependymal and endothelial cells display cognition deficits that mimic those in AD mice with elevated Aβ production (Dietrich et al., 2014). Of note, ApoJ/Clusterin also binds to Lrp1 (Gil et al., 2013), Vldlr, and Apoer2 (Andersen et al., 2003; Leeb et al., 2014) and alternative receptors Trem2 (Yeh et al., 2016) and Plexin A4 (Kang et al., 2016), yet it is not known how ApoJ/Clusterin interactions with the other LDL receptor family members affects AD pathology.

Lrp2 expression decreases with age, which goes along with a reduced clearance rate of Aβ (Carro et al., 2005). In brains of AD-patients, damaged neurons express more Lrp2 (LaFerla et al., 1997), and the transcription of Lrp2 mRNA is repressed by microRNA-146a (Zhang et al., 2016). Genetically, a single nucleotide polymorphism (SNP) in the Lrp2 promoter that reduces Lrp2 expression by 20% is considered a risk factor for AD (Vargas et al., 2010; Wang et al., 2011). Additionally, much like Lrp1, Lrp2 forms a complex with APP and Fe65 to control neurite branching and APP processing (Alvira-Botero et al., 2010).

#### Lrp4 (MEGF7)

### Structure and General Physiological Properties

One of the shorter members of the LDL receptor family, Lrp4, is critical for survival in that LRP4 knockout mice die after birth due to defects in the neuromuscular junction (NMJ; Weatherbee et al., 2006). Lrp4 is also involved in the development of both the kidneys and limbs as Lrp4 knockout mice display abnormal limb morphology and renal agenesis (Johnson et al., 2005; Simon-Chazottes et al., 2006; Karner et al., 2010; Tanahashi et al., 2016). Additionally, Lrp4 regulates chondrocyte and osteoblast homeostasis during cartilage and bone growth (respectively) through binding the ligands Wise/Sostdc1, Dickkopf and Sclerostin (Choi et al., 2009; Asai et al., 2014). As Lrp4-deficient mice die due to abnormal NMJ formation, Lrp4 plays a pivotal role during development at the NMJ where Lrp4 along with its ligand, the heparan-sulfate proteoglycan (HSPG) Agrin, and co-receptors muscle-specific tyrosine receptor kinase (MuSK) and APP act together to orchestrate NMJ formation (Kim et al., 2008; Zhang et al., 2008; Choi et al., 2013). The Lrp4 ligand, Agrin, similar to the Apoer2 and Vldlr ligand Reelin, which also interacts with APP, is a large extracellular matrix protein with multiple binding domains. On the muscle fiber membrane, MuSK and Lrp4 form a functional receptor complex for Agrin. Upon Agrin binding to Lrp4, MuSK is phosphorylated resulting in Rapsyn-dependent focal clustering of nicotinic Acetylcholine receptors (nAChR; Shen et al., 2014). Recent evidence suggests that these components, which are also expressed in the adult brain, also play a role in synaptic plasticity and/or AD pathogenesis (Glenner and Wong, 1984; Berzin et al., 2000; Gomez et al., 2014; Pohlkamp et al., 2015; Sun et al., 2016).

# Genetics

Deficiency in Lrp4, MuSK, Agrin, APP and APLP2, or the intracellular scaffold Rapsyn lead to neonatal lethality, due to failure to form NMJs (Gautam et al., 1999; Wang et al., 2005; Weatherbee et al., 2006). At central synapses, these components do not appear critical for synapse formation; however, a recent report demonstrated that Agrin, Lrp4 and MuSK act together on the astrocyte to control synaptic plasticity (Sun et al., 2016). Lrp4, like APP, is a substrate for ADAM10 secretase and γ-secretase and undergoes proteolytic processing by these enzymes to release soluble ECD and ICD fragments of Lrp4, respectively (Dietrich et al., 2014). Targeted expression of various Lrp4 truncations in mice revealed a differential dependence of membrane anchoring and the presence of the ICD for Lrp4-mediated mechanisms. Knockins expressing secreted Lrp4-ECD survive, but display impaired LTP and develop only partially functional NMJs with abnormal limb development. Alternatively, in mice expressing a membrane-anchored Lrp4 with deleted ICD limb development is only mildly affected and LTP is normal (Johnson et al., 2005; Choi et al., 2013; Pohlkamp et al., 2015).

Studies at the NMJ also revealed important insights how different members of the APP-family interact (Choi et al., 2013). In APP/APLP2 mutants, NMJ endplate patterning is severely impaired, whereas APLP1/APLP2 mutants develop normal endplate patterning with reduced size and apposition of preand postsynaptic specializations. APLP1 seems to be exclusively expressed in the neuronal ending of the NMJ whereas APP and APLP2 are present on both, the muscle and the neuronal sides (Klevanski et al., 2014). In addition, Fe65/Fe65L1 double knockout mice show severe motor impairments, NMJ preand postsynaptic appositions, and impaired hippocampal LTP (Strecker et al., 2016). Fe65 interacts with Apoer2, Vldlr, Lrp1, Lrp1b, Lrp2, but binding to Lrp4 has so far not been examined.

#### Biochemistry and Cell Biology

On the muscle fiber membrane, MuSK and Lrp4 form a functional receptor complex for Agrin. Upon Agrin binding to Lrp4, MuSK is phosphorylated resulting in Rapsyn-dependent focal clustering of nAChR (Shen et al., 2014). APP, and presumably APLP2, present on the muscle fiber surface and along with APLP1 on the neuron, also binds to Lrp4 and Agrin, which is required for the localized clustering of AChR on the muscle fiber where nerves terminate to allow a functional NMJ to form (Kim et al., 2008; Choi et al., 2013; **Figure 4**). Interestingly, unlike Lrp1 and Lrp1b, in vitro experiments show that Lrp4 binding to APP does not require the KPI domain in APP (Choi et al., 2013).

Similar to the lipoprotein receptor ligand Reelin, multiple functions have been described for Agrin in shaping and maintaining neuronal activity in the brain. Agrin stimulates filopodia formation to allow structural plasticity (McCroskery et al., 2009) and inhibits astrocytic ATP release resulting in enhanced synaptic glutamate release (Sun et al., 2016). Agrin also regulates the strength of GABAergic synapses during network inactivation (Pribiag et al., 2014), reduces Aβ-levels (Rauch et al., 2011), and contributes to acetylcholine receptor clustering (Rauch et al., 2011). However, as of now, it is unknown if these functions require Lrp4-mediated endocytosis and trafficking. For example, Lrp4 does not require endocytic activity to promote NMJ formation (Willnow et al., 2012). Agrin binds not only Lrp4 but also to multiple other receptors and ligands such as heparin (Wallace, 1990), NCAM (Storms et al., 1996), Integrins (Martin and Sanes, 1997), α-dystroglycan (Bowe et al., 1994), Na+/K+ATPase (Hilgenberg et al., 2006) and notably APP (Choi et al., 2013). Moreover, presynaptic activity dependent release and postsynaptic activity- dependent activation of the protease Neurotrypsin regulates Agrin cleavage at α- and β-sites (Reif et al., 2007; Stephan et al., 2008; Gisler et al., 2013). Specifically the short C-terminal fragment of Agrin potentially promotes filopodia outgrowth via α-dystroglycan (Gisler et al., 2013).

Lrp4 also contributes to synaptic plasticity. Mice lacking Lrp4 or expressing a truncated Lrp4 retaining the ECD (Lrp4- ECD) in the brain have impaired hippocampal LTP and impaired

memory (Gomez et al., 2014; Pohlkamp et al., 2015). Importantly, Sun et al. (2016) showed that the astrocyte-specific knockout of Lrp4 (using GFAP-Cre) extinguishes all brain Lrp4 expression and enhances the release of ATP from astrocytes, which may be causative for the described impairment in LTP. Of note, GFAP-Cre expression is not restricted to astrocytes and found in some neuronal populations as well. However, the authors also demonstrated that Agrin, by binding to Lrp4 and activating MuSK, controls the ATP release from astrocytes (Sun et al., 2016). The impaired LTP in Lrp4-ECD mice (Pohlkamp et al., 2015) suggests that anchoring of Lrp4 to the astrocytic membrane is required for normal synaptic potentiation. Neurons exclusively express the TM-Agrin (Neumann et al., 2001) that contains the alternatively spliced Z+ insert required for Lrp4 binding. TM-Agrin, by binding to Lrp4 could mediate a direct interaction of astrocytes and neurons. Furthermore, activity-driven neurotrypsin cleavage would allow the release of the Agrin C-terminal Lrp4-binding domain, which then can diffuse and bind to Lrp4/MuSK complexes on the astrocytic surface to control ATP release. It needs to be determined if this pathway requires APP or APLP1/2 in the complex, which are mainly/exclusively expressed by neurons. The astrocytic Agrin/Lrp4/MuSK complex together with APP or APLP2 on the neuronal surface might also be relevant for astrocyte-neuron interactions.

In the hippocampus, besides neurons, astrocytes express functional α7-type AchRs (Shen and Yakel, 2012), which is increased in the brain of AD-patients (Yu et al., 2005). Importantly Aβ binds to hippocampal α7AchR expressed on astrocytes, resulting in increased Ca2<sup>+</sup> permeability (Pirttimaki et al., 2013). Activation of α7AchR on astrocytes triggers AMPA receptor recruitment to glutamatergic synapses, a mechanism also involved in converting silent synapses to functional ones (Wang et al., 2013a). At the NMJ Agrin/Lrp4/MuSK/APP complex formation appears to be required to effectively cluster AchRs. So far, however, astrocytic α7AchR function has not been shown to require the formation of an Agrin/Lrp4/MuSK/APPcomplex. However, total AChR clustering in TM-Agrin knockout mouse brains, expressing only 20% of the Lrp4-binding Z+ Agrin form, is 4- to 5-fold reduced (Rauch et al., 2011).

Heparan sulfate proteoglycans (HSPG) inhibit BACE1 mediated APP cleavage (Scholefield et al., 2003). Thus, Agrin, as the major HSPG accumulating in plaques of AD-brains (Verbeek et al., 1999) might be a relevant inhibitor of BACE1. Agrin has also been described to be relevant for the function of the BBB (Rauch et al., 2011; Steiner et al., 2014). However, Aβ-clearance via Agrin and Lrp4 in astrocytes is unlikely, since in the neuron-specific TM-Agrin knockout, which expresses only 20% of Z+ Lrp4-interacting Agrin, Aβ clearance is not affected. By contrast, endothelial-specific knockout of Agrin does reduce Aβ-clearance (Rauch et al., 2011).

#### Lrp5/6

#### Structure and General Physiological Properties

Lrp5 and Lrp6 share 71% homology and are more distantly related members of the family. Despite encoding three LBRs and four EGF-precursor homology domains, compared to the core members, the domains appear in an inverse order with the ligand-binding domains adjacent to the TM segment rather than at the N-terminus. Additionally, their ICDs lack NPxY motifs. Both receptors have important functions in Wnt/βcatenin signaling, whereby Wnt and the Frizzled-receptors, mediate intracellular β-catenin translocation to the nucleus for transcriptional control of target gene expression (reviewed by Joiner et al., 2013). Similar to Lrp4, Lrp5 and Lrp6 are involved in bone growth (Lara-Castillo and Johnson, 2015), recently Lrp6 has also been suggested to have a role in AD and APP processing (De Ferrari et al., 2007).

#### Genetics

Lrp5 deficiency causes osteoporosis and bone fracture in mice due to reduced osteoblast proliferation and low bone mass (Kato et al., 2002), and point mutations have been found in human patients with altered bone mass. Lrp5 knockout also leads to defects in cholesterol and glucose metabolism. Lrp5 and ApoE double knockout mice suffer from hypercholesterolemia, fat intolerance, and atherosclerosis (Fujino et al., 2003; Magoori et al., 2003). Mesenchymal specific Lrp5 and Lrp6 double mutants resembled β-catenin knockouts, with severe skeletal development defects (Joeng et al., 2011). Whereas Lrp5 deficiency primarily affects bone density, Lrp6 deficiency severely affects brain development. Lrp6 deletion leads to death after birth, similar to Wnt mutants they have a caudal truncation of the body axis, excess neural tissue, defects in neural tube closure, loss of paraxial mesoderm, and mid- and hindbrain defects (Pinson et al., 2000). A point mutation in an EGF repeat of Lrp6 causes coronary artery disease with high LDL-levels by affecting Wnt signaling (Mani et al., 2007). A SNP in a highly conserved region of LRP6, initially genetically associated with low bone mass, has now been associated with AD (De Ferrari et al., 2007).

#### Biochemistry and Cellular Function

Wnt signaling via Lrp6 has been implicated in neuronal differentiation (Jeong et al., 2014), commissural axon guidance (Avilés and Stoeckli, 2016), and adult neurogenesis in the hippocampal niche (Schafer et al., 2015). Neuronal deletion of Lrp6 in the forebrain of the mouse leads to defects in synaptic integrity and memory formation. Furthermore crossing these mice with APP/PS1 mice led to increased APP processing to Aβ that in turn inhibited Wnt signaling, resulting in a synergistic effect on synaptic dysfunction (Liu et al., 2014). Wnt signaling is also compromised in brains of patients with AD (Liu et al., 2014).

# SorLA (Sorl1/LR11/LRP11)

### Structure and General Physiological Properties

SorLA is a hybrid-type receptor, as the only member of the LDL receptor family with a Vps10p (yeast vacuolar protein sorting 10 protein) domain and six Fibronectin repeats (**Figure 1**). SorLA is predominantly expressed in the brain, especially in neurons (Jacobsen et al., 1996; Yamazaki et al., 1996), where it acts as an intracellular sorting receptor transporting cargo, including APP, between different intracellular compartments in the cell (Andersen et al., 2005). In addition to familial mutations linked to AD (Meng et al., 2007), SorLA is reduced in postmortem AD brains (Scherzer et al., 2004) and in the CSF of AD patients (Ma et al., 2009).

#### Genetics

Defective homeostasis of SorLA and its cargo disrupts cellular function and causes AD, atherosclerosis and obesity (Caglayan et al., 2014). In mice, SorLA knockout leads to increased Aβlevels in the brain, whereas neuronal SorLA overexpression causes a redistribution of APP to the Golgi, which results in decreased Aβ production (Andersen et al., 2005).

#### Biochemistry and Cellular Function

The ICD of SorLA is important for retrograde trafficking from endosomes to the trans-Golgi network (TGN) by binding to the retromer complex and anterograde trafficking by interacting with clathrin-adaptors (Jacobsen et al., 2002; Seaman, 2007; Fjorback et al., 2012). SorLA binds APP and Aβ to control their transport from endosomes either to the TGN to prevent proteolytic APP-breakdown or to lysosomes for Aβ-degradation, which recently has been reviewed in detail by Schmidt et al. (2016). The mosaic receptor has different extracellular binding domains: an N-terminal Vps10p domain followed by an EGF-precursor homology domain and 11 LBRs. Whereas the LBRs are important for APP binding and rerouting away from the proteolytic pathway (Andersen et al., 2005), the Vps10p domain is responsible for Aβ-binding and the final lysosomal degradation (Caglayan et al., 2014). The Vps10p domain consists of a ten-bladed β-propeller fold with a large tunnel that has a propensity for ligands with a β-sheet formation. An internal ligand derived from the SorLA propeptide binds in this tunnel, extends the domain by one β-propeller blade, and presumably blocks ligand binding (Kitago et al., 2015). The SorLA propeptide is removed in late Golgi compartments by furin (Munck Petersen et al., 1999). SorLA and its interaction with APP have recently been reviewed in detail by Schmidt et al. (2016).

#### VERY DISTINCT AND SHORT RECEPTORS CONTAINING LBRs

Lrp3, Lrp10 (murine Lrp9) and Lrp12 (ST7/Mig13) share high homology (Battle et al., 2003) and have two ligand-binding CUB domains, Lrad3 does not have CUB domains (**Figure 1**). Even though in the literature all four receptors have been claimed to be members of the LDL receptor family, the domain composition puts them into a different class of mosaic proteins. All four receptors lack EGF-precursor homology domains found in all other members of the LDL receptor family. All four receptors have three to five LBRs (**Figure 1**), but lipoprotein binding remains to be confirmed, and their CTDs encode intracellular sorting motifs. Lrad3 and Lrp10 have been shown to interact with APP, thus we briefly review them in this section.

**Lrp3**, discovered in 1998 is expressed in a wide range of human tissues, including the brain, with the highest expression in skeletal muscle and ovary. Interestingly, in contrast to other LDL receptor family members, Lrp3 does not seem to bind to RAP (Ishii et al., 1998).

**Lrp10 (murine Lrp9)** is expressed in various tissues, including the brain. Little is known about its function; only one publication describes its involvement in APP processing. Lrp10 is located in endosomes and in the TGN (Sugiyama et al., 2000). The cytoplasmic tail interacts with clathrin adaptors that coordinate shuttling between endosomes and TGN (Boucher et al., 2008; Doray et al., 2008). Recently, in vitro data showed that APP interacts with the ECD of Lrp10, and both proteins colocalize at the TGN. Lrp10 expression in brains of AD patients is reduced. In cell culture, Lrp10 overexpression induces the accumulation of APP in the TGN, which results in reduced APP-surface expression and processing. Conversely, knockdown of Lrp10 led to increased processing of APP to Aβ (Brodeur et al., 2012).

**Lrp12 (ST7/MG13)** has been annotated as a member of the LDL receptor family in 2003 (Battle et al., 2003). The Lrp12s ICD contains several motifs implicated in endocytosis and signal transduction. Lrp12 is important during CNS development where it controls the formation of the cortical plate, neuronal polarity, and migration (Schneider et al., 2011; Wang et al., 2013b). It is also involved in tumorigenesis including epilepsyassociated gangliogliomas (Garnis et al., 2004; Robens et al., 2016). Silencing of Lrp12 in primary neurons leads to increased dendritic branching, silencing of Lrp12 in the mouse brain during brain development leads to cortical dyslamination and seizure sensitization (Grote et al., 2016). As of today, no role in AD has been described. However, Lrp12 is expressed in neurons and astrocytes of the adult brain (Grote et al., 2016).

**Lrad3** has the shortest ECD of all receptors (**Figure 1**), with only three LBRs. Lrad3 is found in the brain and is expressed in microvascular endothelial cells and neurons (Otsuki et al., 2005; Ranganathan et al., 2011). In cell culture, the results of Lrad3 overexpression were similar to those of Lrp1: Lrad3 promoted the pathogenic proteolytic pathway of APP, shifting it away from the α-secretase pathway towards the endosome, resulting in enhanced Aβ production. While Lrad3 does not interact with Aβ, the receptor does interact with the central APP fragment (C99) that contains the ICD, the TM-domain, and a short ECD (Ranganathan et al., 2011). The Lrad3-ICD contains two PPxY motifs to which WW-domain containing proteins, e.g., ubiquitin ligases, bind (Ingham et al., 2004). More recently, it was found that Lrad3 is a component of the ubiquitin proteasome system by activating the E3 ubiquitin ligases Itch and Nedd4 (Noyes et al., 2016). However, a direct role of Lrad3 regulation of ubiquitination to APP processing has not been established.

#### LIPOPROTEIN RECEPTORS AND APP BEYOND ALZHEIMER's

The function of APP and Aβ beyond AD is not well understood and understudied, especially in conjunction with lipoprotein receptors. Different chapters of this series discuss the physiological role of APP and its cleavage products from various physiological perspectives. APP and its trafficking and processing plays a role in neurite outgrowth and synaptogenesis, APP-deficiency decreases dendritic spine numbers and impairs LTP, which can be rescued by sAPPα but not sAPPβ (Tyan et al., 2012). APP function is largely occluded in single APP mutants, since its paralogs APLP1 and APLP2 can partially compensate for APP-loss. Characterization of combined knockouts of APP and its close relatives APLP1 and APLP2 provides additional insights into the trophic functions of APP: whereas single knockouts and APLP1/APP double knockouts are viable and fertile, combined APLP2/APP or APLP1/APLP2 knockouts display reduced viability (Heber et al., 2000). This suggests that APLP2 carries the most essential physiological functions that can be partially compensated by redundancy in the other family members. APP and APLP2 are expressed ubiquitously, while APLP1 expression is restricted to the nervous system (Lorent et al., 1995). Lrp4, MuSK, Agrin and APP/APLP2 are essential components of a functional complex that recruits and clusters acetylcholine receptors at the NMJ (reviewed in the ''Lrp4'' Section). Additionally, Lrp4 does not require the KPI domain to bind APP (Choi et al., 2013).

APP trafficking and processing is controlled by a large variety of proteins, but little is known about their physiological relevance. APP interacts with numerous type-I TM receptors, many of which are lipoprotein receptors, and several other ligands, adaptor and scaffolding proteins, which together provide a protein-protein network involved in signaling, processing of various receptors, partially through endocytic pathways.

# CONCLUDING REMARKS

APP processing to Aβ and in particular the accumulation of the amyloidogenic Aβ<sup>42</sup> product, either from increased production or impaired clearance, are initiating events in AD, and ApoE genotype is the most important late onset risk factor for AD. Both APP and ApoE interact with LDL receptor family members to regulate APP trafficking, processing and elimination. Therefore, it is all but certain, that LDL receptor family members play a pivotal role in the pathogenesis of AD.

As a result of the work reviewed in this article, we have learned much about the potential molecular mechanisms that these lipoprotein receptors play in AD pathogenesis, yet the relative importance of each individual event is still unclear. Continuing work on the biology of LDL receptor related genes and their ligands on the physiology of the APP processing machinery holds great promise not only to greater understanding of the disease process but also for the identification of novel and effective therapeutic approaches.

# AUTHOR CONTRIBUTIONS

TP and CRW jointly wrote the article and designed the figures under JH guidance and JH edited the manuscript.

# ACKNOWLEDGMENTS

This work was supported by grants from the NHLBI (R37 HL063762), the NIA (RF AG053391), the NINDS and NIA (RO1 NS093382), as well as, the Consortium for Frontotemporal Dementia Research (A108400), and the Brightfocus Foundation (A2016396S). We would like to thank Nancy Heard and Barbara Dacus for their help in preparing the figures.

# REFERENCES


growth factor I. J. Neurosci. 25, 10884–10893. doi: 10.1523/JNEUROSCI.2909- 05.2005


defines minimum requirements for GGA binding. FEBS Lett. 511, 155–158. doi: 10.1016/s0014-5793(01)03299-9


splicing and transcription units of the human apolipoprotein E receptor 2 gene. J. Biol. Chem. 272, 8498–8504. doi: 10.1074/jbc.272.13.8498


cell proliferation and presynaptic bouton density in the murine hippocampus. Neurosci. Res. 59, 251–256. doi: 10.1016/j.neures.2007.07.004


interneuron subtypes in the rat barrel cortex. Cereb. Cortex 24, 3046–3058. doi: 10.1093/cercor/bht161


Genet. B Neuropsychiatr. Genet. 153B, 895–902. doi: 10.1002/ajmg.b. 31056


gp330/megalin. Proc. Natl. Acad. Sci. U S A 93, 8460–8464. doi: 10.1073/pnas. 93.16.8460


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Pohlkamp, Wasser and Herz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# LRP1 Modulates APP Intraneuronal Transport and Processing in Its Monomeric and Dimeric State

Uta-Mareike Herr 1 †, Paul Strecker 2 †, Steffen E. Storck <sup>1</sup> , Carolin Thomas <sup>2</sup> , Verena Rabiej <sup>1</sup> , Anne Junker <sup>1</sup> , Sandra Schilling<sup>2</sup> , Nadine Schmidt <sup>2</sup> , C. Marie Dowds <sup>2</sup> , Simone Eggert <sup>2</sup> , Claus U. Pietrzik <sup>1</sup> \* † and Stefan Kins <sup>2</sup> \* †

*1 Institute of Pathobiochemistry, Molecular Neurodegeneration, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany, <sup>2</sup> Division of Human Biology and Human Genetics, Technical University of Kaiserslautern, Kaiserslautern, Germany*

#### Edited by:

*Thomas Deller, Goethe-University, Germany*

#### Reviewed by:

*Matthias Kirsch, Albert Ludwig University of Freiburg, Germany Hyunsoo Shawn JE, Duke NUS Graduate Medical School, Singapore*

#### \*Correspondence:

*Claus U. Pietrzik pietrzik@uni-mainz.de Stefan Kins s.kins@biologie.uni-kl.de*

*† These authors have contributed equally to this work.*

Received: *20 November 2016* Accepted: *10 April 2017* Published: *27 April 2017*

#### Citation:

*Herr U-M, Strecker P, Storck SE, Thomas C, Rabiej V, Junker A, Schilling S, Schmidt N, Dowds CM, Eggert S, Pietrzik CU and Kins S (2017) LRP1 Modulates APP Intraneuronal Transport and Processing in Its Monomeric and Dimeric State. Front. Mol. Neurosci. 10:118. doi: 10.3389/fnmol.2017.00118* The low-density lipoprotein receptor-related protein 1, LRP1, interacts with APP and affects its processing. This is assumed to be mostly caused by the impact of LRP1 on APP endocytosis. More recently, also an interaction of APP and LRP1 early in the secretory pathway was reported whereat retention of LRP1 in the ER leads to decreased APP cell surface levels and in turn, to reduced Aβ secretion. Here, we extended the biochemical and immunocytochemical analyses by showing via live cell imaging analyses in primary neurons that LRP1 and APP are transported only partly in common (one third) but to a higher degree in distinct fast axonal transport vesicles. Interestingly, co-expression of LRP1 and APP caused a change of APP transport velocities, indicating that LRP1 recruits APP to a specific type of fast axonal transport vesicles. In contrast lowered levels of LRP1 facilitated APP transport. We further show that monomeric and dimeric APP exhibit similar transport characteristics and that both are affected by LRP1 in a similar way, by slowing down APP anterograde transport and increasing its endocytosis rate. In line with this, a knockout of LRP1 in CHO cells and in primary neurons caused an increase of monomeric and dimeric APP surface localization and in turn accelerated shedding by meprin β and ADAM10. Notably, a choroid plexus specific LRP1 knockout caused a much higher secretion of sAPP dimers into the cerebrospinal fluid compared to sAPP monomers. Together, our data show that LRP1 functions as a sorting receptor for APP, regulating its cell surface localization and thereby its processing by ADAM10 and meprin β, with the latter exhibiting a preference for APP in its dimeric state.

Keywords: amyloid precursor protein (APP), dimerization, transport, low density lipoprotein receptor-related protein 1 (LRP1), processing

# INTRODUCTION

The amyloid precursor protein (APP) is a type I transmembrane protein that has first been identified related in association with Alzheimer's disease (AD) as representing the precursor of amyloid β (Aβ) peptides (Kang et al., 1987). Those peptides generated by sequential cleavage of APP by β- and γ-secretases were shown to be a major component of senile plaques found in the brains of AD patients (Merz et al., 1983; Masters et al., 1985). Besides its role in AD pathogenesis, APP has been implicated in physiological functions including intracellular signaling, trophic activity in neurons and synapses as well as in synaptic and cell adhesion processes (Baumkötter et al., 2012; Müller and Zheng, 2012). Recent studies revealed that APP can dimerize or oligomerize in cis- as well as in transorientation (Scheuermann et al., 2001; Soba et al., 2005; Munter et al., 2007; Kaden et al., 2009; Wang et al., 2009; Isbert et al., 2012; Baumkötter et al., 2014; Klevanski et al., 2014; Stahl et al., 2014). Remarkably, APP dimers were detected in mouse brains (Soba et al., 2005; Schmidt et al., 2012), indicating that dimer formation occurs in vivo under physiological conditions. Trans-cellular APP dimerization is assumed to modulate synapse organization (Soba et al., 2005; Wang, 2005; Wang et al., 2009; Isbert et al., 2012; Baumkötter et al., 2014; Klevanski et al., 2014; Stahl et al., 2014). In contrast, APP cis-dimerization, that has been shown to occur as early as in the endoplasmic reticulum (ER) (Isbert et al., 2012), has been implicated in processing of APP by α- , β-, and γ-secretases (Munter et al., 2007, 2010; Kaden et al., 2008; Eggert et al., 2009; Libeu et al., 2012; Schmidt et al., 2012; So et al., 2012; Jung et al., 2014). Recently, it has been claimed that efficient processing of APP by α- and β-secretases may depend on its oligomerization state that results in cooperative effects for these allosteric enzymes, influenced by SorLA and possibly also LRP1 (Schmidt et al., 2012). However, whether sAPP dimers are generated in vivo in neurons, which secretases are required and what might be the role of LRP1 in this context, is unknown yet.

LRP1, a member of the low density lipoprotein receptor (LDLR) family (Krieger and Herz, 1994), was shown to interact with APP via the N- and C-terminal domain and to affect its processing (Ulery et al., 2000; Pietrzik et al., 2002, 2004). This effect is presumably based on the impact of LRP1 on APP endocytosis (Knauer et al., 1996; Ulery et al., 2000; Pietrzik et al., 2002; Cam et al., 2005). In addition, APP can interact with LRP1 before it is cleaved by furin in the TGN, implying an interaction of APP with LRP1 early in the secretory pathway (Pietrzik et al., 2004). This hypothesis was confirmed in 2008 (Waldron et al., 2008), by using a truncated LRP1-construct (LRP-CT) (Pietrzik et al., 2002) containing a dilysine ER-retention motif (KKAA) capable of binding to APP. The retention of LRP1 in the ER leads to a decrease in Aβ secretion as well as to a decrease in full length APP and CTF levels at the plasma membrane (Waldron et al., 2008).

Here, we extend the analysis of APP transport characteristics and show that LRP1 plays a crucial role in trafficking and processing of monomeric as well as dimeric APP.

### MATERIALS AND METHODS

#### Cell Culture

Human Embryonic Kidney cells (HEK 293T) were cultured in Dulbecco's Modified Eagle's Medium (DMEM; Thermo Fisher Scientific) supplemented with 10% fetal calf serum (FCS), 1 mM sodium pyruvate (Sigma-Aldrich), 100 units/ml penicillin and 0.1 mg/ml streptomycin (Thermo Fisher Scientific).

Chinese Hamster Ovary cells, either CHO K1 or LRP-deficient CHO 13-5-1 (FitzGerald et al., 1995), were grown in Alpha Minimum Essential Medium (α-MEM; Lonza) supplemented equally.

Primary neurons were extracted from cortices of C57BL/6J or 5xFAD/Lrp1flox/flox mouse embryos at embryonic day 14 as described previously (Maier et al., 2013). Cells were seeded on poly-L-ornithine (100 µg/ml; Sigma-Aldrich) coated 6-well plates or 6 cm dishes, respectively, in a density of 600,000 cells per well or 1,000,000 cells per dish. They were cultured in Neurobasal Medium (Thermo Fisher Scientific) complemented with 100 units/ml penicillin and 0.1 mg/ml streptomycin, 1 x B27 supplement and 1 x GlutaMAX (all Thermo Fisher Scientific).

Primary cortical neurons (PCN) were prepared using E14 embryos from C57BL/6J mice (Janvier) or 5xFAD/Lrp1flox/flox mice as described before (Stahl et al., 2014; Hermey et al., 2015). PCN dissolved in DB1 medium [DMEM with 10% FBS, 0.79% D-glucose and 1 x GlutaMAX (Thermo Fisher Scientific)] were plated on poly-L-lysine (Sigma-Aldrich) coated fluorodishes in a density of 6<sup>∗</sup> 10<sup>5</sup> /cm<sup>2</sup> . Six hour post plating DB1 was changed and PCN were cultivated in neurobasal medium supplemented with B27 and GlutaMAX (Thermo Fisher Scientific).

Primary hippocampal neurons (PHN), used for APP/LRP live cell imaging, were prepared from P0 pups of C57BL/6J mice and treated in the same way as described for PCN.

All cell types were cultivated at 37◦C in an incubator maintaining a relative humidity of over 80% and a CO<sup>2</sup> level of 5%.

#### DNA Constructs and Cloning

For analyzing the properties of APP cis-dimers a human APP695 construct with a dimer-bearing amino acid exchange from lysine (K) to cysteine (C) at position 587 (APP695 K587C) was generated for transient and stable transfections. The plasmid consisting of the human APP695 CDS with the triplet mutation (AAG to TGT) at position 1,761 as well as a C-terminal myc-tag in the vector pLBCX was developed by an overlap extension PCR as described by Isbert et al. (2012). The restriction sites for HindIII and ClaI, which are flanking the myc-tagged, mutated APP sequence, enabled the subcloning of this DNA fragment into the vector pLHCX resulting in the pLHCX-APP695 K587C construct. Hence this construct has the same vector backbone as the also used pLHCX-APP695 wt plasmid (Jäger et al., 2009). APP dimer constructs exhibiting a mutation in the APP internalization motif "YENPTY" (Lai et al., 1995; MarquezSterling et al., 1997) were generated performing a standard PCR followed by restriction and ligation into the pLHCX vector backbone. The plasmid pLHCX-APP695 K587C served as template for PCR using the forward primer 5 ′ -CCCAAGCTTATGCTGCCCGGTTTG-3′ , which contains a 5′ HindIII restriction site and the reverse primer 5′ -CC ATCGATGGTTACAGATCCTCTTCTGAGATGAGTTTTTGTT CGTTCTGCATCTGCTCAAAGAACTTTTCGTAGCCGTTTTC GTAG-3′ exhibiting the mutation in the internalization motif, the myc-epitope and a 3′ ClaI restriction site. The described mutation results in an amino acid exchange from NPTY to NGYE at the C-terminus of the expressed APP695 K587C protein. The amplified DNA fragment was subcloned in frame into the pLHCX vector backbone via the HindIII and ClaI restriction

sites. Sequencing of the generated construct authenticated its accuracy. To study the processing of monomeric and dimeric APP by meprin β, HEK 293T cells were co-transfected with either APP695 wt or APP695 K587C and the meprin β HA construct in pLBCX (Schönherr et al., 2016).

For generation of the expression construct encoding the LRP1-GFP fusion protein, the EGFP cDNA was amplified from pcDNA3.1 APP-GFP (Szodorai et al., 2009) using the oligos 5′ -TGAGCAGATGCAGAACGTCG-3′ and 5′ - GCACAGTCGAGGCTGATCAGC-3′ . The PCR product was cloned via flanking BamHI/NotI sites in frame into pLBCX myc-LRP1 and the resulting construct, pLBCX-myc-LRP1-GFP, was validated by sequencing.

#### Infections and Transfections

The infection of primary cortical neurons (PCN) with an adenoviral vector encoding human APP695 (Yuan et al., 1999) was performed at DIV 7. Cells were incubated with 100 plaqueforming units per cell for 6 h. In contrast, for live cell imaging, PCN or PHN were transiently transfected at DIV 6 using calcium phosphate transfection. A neurobasal medium containing 2% B27 (transfection medium) was prepared and incubated for at least 30 min at 37◦C and 5% CO2. Meanwhile, the following transfection mix was pipetted (sufficient for two fluorodishes): Solution A containing 75 µl H2O dd, 9.5 µl 2.5 M CaCl<sup>2</sup> and 20 µg DNA; Solution B containing 75 µl 2 × HBS pH 7.07 (274 mM NaCl, 10 mM KCl, 1.4 mM Na2HPO4, 15 mM D-Glucose, 42 mM HEPES pH 7.1). Solution A was added to Solution B, immediately vortexed for 10 s at maximum speed and incubated for 20 min at RT. Meanwhile, the medium of the cultured neurons was replaced by 2 ml of the previously prepared transfection medium. The old medium was collected for later usage. Afterwards, 89.75 µl of the transfection mix were added per neuronal culture dish. The neuronal cells were incubated for 3 h at 37◦C until precipitates were formed. To remove the precipitates, the cells were washed twice with 2 × HBS. Therefore 1 ml prewarmed 2 x HBS was added to the transfected neurons before 1 ml was removed. This step was repeated once and the medium-HBS mix was afterwards removed completely. To provide important growth factors for neuronal growth, 2 ml of the collected old medium were added to each dish. The cells were incubated at 37◦C for 18–20 h and analyzed by live cell imaging.

For transient transfection of HEK and CHO cells with different APP695 constructs or the meprin β construct a transfection mixture containing 8 µg polyethylenimine (PEI) and 2 µg DNA in 120 µl serum-free medium was added to the cells for 4 h.

Stable CHO cells were generated as described previously (Isbert et al., 2012) using pLHCX-APP695 K587C and 350 µg/ml Hygromycin B (Thermo Fisher Scientific) for selection.

#### Antibodies

The antibody mix 1G75A3 of the two monoclonal antibodies 1G7 and 5A3, both directed against the APP ectodomain, was provided by Dr. Koo (UC San Diego School of Medicine, USA) and enabled the detection of all forms of full-length APP (mature, immature or dimerized) in cell lysates as well as soluble APP in the conditioned medium. This antibody mix was used for Western Blotting and for immunoprecipitation of APP. For detection of LRP1 in Western Blotting the polyclonal antibody 1,704 (Pietrzik et al., 2002) directed against the C-terminus of LRP1 was used. Y188 (Abcam) directed against the C-terminus of APP was used to detect monomeric and dimeric APP in Blue Native Gel Electrophoresis. Aβ was detected by IC16, a monoclonal antibody recognizing the amino acids 1 to 16 of the human Aβ sequence (Jäger et al., 2009). The polyclonal anti-actin antibody and the secondary HRP-conjugated goat anti-rabbit antibody were purchased from Sigma-Aldrich. The secondary donkey antibody against mouse, also HRP-conjugated, was obtained from Jackson ImmunoResearch.

#### Western Blotting

After collecting the conditioned medium cells were harvested and lysed either in RIPA (50 mM Tris-Cl (pH 8), 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, 10 mM NaF, 1 mM β-glycerophosphate, 0.5% sodium deoxycholate) regarding neurons or, concerning HEK and CHO cells, in NP-40 lysis buffer (500 mM Tris (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, 0.02% NaN3) both containing 1 x protease inhibitor cocktail (PI; Roche). Debris were pelleted by centrifugation with 18,600 × g for 20 min at 4◦C. The protein concentrations were measured using the PierceTM BCA Protein Assay Kit (Thermo Fisher Scientific) to determine equal amounts of total protein for lysate analysis. For comparable protein amounts of the conditioned media volumes were adjusted to the protein concentration in the corresponding lysates. After addition of 4 x SDS sample buffer with (Roti <sup>R</sup> -Load 1; Roth) or without (40% glycerol, 200 mM Tris-HCl (pH 6.8), 0.08% bromphenol blue, 8% SDS in VE-H2O) β-mercaptoethanol (βME) samples were boiled for 5 min at indicated temperatures. Proteins were separated by gel electrophoresis in 6 or 7% Bis-Tris gels and transferred onto nitrocellulose membranes (GE Healthcare Life Sciences) via wet blot. To block non-specific binding membranes remained for 1 h in 5% (w/v) non-fat dry milk dissolved in TBS containing 0.05% Tween 20 (Roth) before incubation with the appropriate primary and secondary antibodies. The protein detection was carried out using the Immobilon Western HRP Substrate (Millipore) resulting in chemiluminescence, which was recorded by the LAS-3000 mini (Fujifilm).

#### Immunoprecipitation and Detection of Aβ Peptides

The immunoprecipitation of Aβ peptides was performed as described by Schönherr et al. (2016). Proteins were separated by Urea SDS-PAGE corresponding to the approach of Klafki et al. (1996) and transferred to PVDF membranes via semidry Western Blotting (Biorad) at 47 mA per gel. Afterwards membranes were boiled for 3 min in 1 x PBS before blocking nonspecific binding in 5% (w/v) non-fat dry milk in TBST for 30 min. Membranes were incubated over night at 4◦C with IC16 antibody (1:500). After washing with TBST the secondary HRP-cojugated mouse antibody was added for 1 h at room temperature. Protein detection and recording were performed as described above.

#### Blue Native Gel Electrophoresis

Blue native gel electrophoresis was performed as described before (Eggert et al., 2009). Briefly, transfected cells were resuspended in 1 ml of homogenization buffer (250 mM sucrose in 20 mM HEPES, pH 7.4, with protease inhibitors) and then sheared by passing through a 27 gauge needl. Postnuclear supernatant was collected after a centrifugation step at 1,000 × g for 15 min. After sedimentation at 100,000 × g for 1 h the membrane fraction was washed once with 200 µl of homogenization buffer followed by another centrifugation at 100,000 × g. The pellet was resuspended in 200 µl homogenization buffer. 100 µg of protein were solubilized with Blue Native sample buffer (1.5 M amino caproic acid, 0.05 M Bis-Tris, 10% n-dodedecyl- \_-D-maltoside, and protease inhibitor at pH 7). The samples were separated on gradient gels Thyroglobulin (669 kDa), apoferritin (443 kDa), catalase (240 kDa), aldolase (158 kDa), and bovine serum albumin (66 kDa) were used as molecular weight standards.

# Live Cell Imaging

Fluorophore tagged LRP1 and APP fusion proteins were tracked by imaging of living cells, as described before (Szodorai et al., 2009; Hermey et al., 2015). Briefly, during live cell imaging transfected cells were temperature-controlled (37◦C) and CO2 controlled (5%). Images were taken every 200 ms over a period of 30 s. GFP-tagged proteins were excited with 470 nm and RFP fusion proteins with 550 nm wave length using a matching filter and fast changing LED's. Kymographs were created using Image J software (1.46r) in combination with the Multiple-Kymograph plugin. The slope of the traces is a direct measure for the velocity of the vesicles (v = cotan(α), where α is the angle relative to the x-axis). Single tracks with an angle 0◦ < α < 90◦ were defined as anterograde, and tracks with a slope 90◦ < α < 180◦ were defined as retrograde transport vesicles. Tracks with slopes of 90◦ (parallel to the time axis) were determined as stationary vesicles. For vesicle distribution all lines of one kymograph were counted as individual transport vesicles and the sum of all anterograde, retrograde and stationary vesicles was set to 100% (given as relative amount of vesicles). For calculation of total amount of vesicles per neurite segment, again all traces of individual kymographs were counted as single vesicles (stationary, anterograde and retrograde vesicles) and related to a neurite length of 1 µm.

#### Immunocytochemistry

Primary cortical neurons (PCN) were differentiated for 7 days in vitro and then subjected for immunocytochemical analysis. PCN were fixed for 10 min at 37◦C in 4% (w/v) PFA with 4% (w/v) sucrose and permeabilized for 10 min with 0.1% (v/v) NP40 in 1 x PBS. For detection of LRP1 and APP we used the polyclonal antibody 1,704 and monoclonal antibody C1/6.1, respectively. Secondary antibodies were Alexa Flour 488 and Alexa Flour 594 (1:1,000, Invitrogen). Hoechst (33258, Thermo Fisher Scientific) was used as nuclear counterstaining. Imaging was performed with microscope Axio Observer Z.1 (Zeiss with apotome) and z-stacks were taken in 0.2 µm steps.

#### Pulse-Chase Assay

To examine the expression and stability of APP dimers, a pulsechase assay was performed with CHO K1 and CHO 13-5-1 cells 48 h after seeding on 6 cm dishes. Cells were starved in DMEM without methionine and cysteine complemented as described above, which was replaced after 1 h by 1 ml of the same medium containing 150 µCi35S/ml (EasyTagTM EXPRESS35S Protein Labeling Mix; PerkinElmer). Following 15 min incubation at 37◦C the medium was substituted to 2 ml α-MEM supplemented as outlined above but with addition of 40 mM HEPES (Lonza). Cells were maintained in this medium at 37◦C for indicated time spans before being harvested and lysed in NP-40 buffer with 1 x PI as detailed previously. For immunoprecipitation of APP, lysates and conditioned media were incubated over night at 4◦C with protein G agarose beads (Roche) and the 1G75A3 antibody mix against the APP ectodomain. Beads were washed as described above, pelleted and finally boiled in 4 x SDS sample buffer at 80◦C for 5 min. The accordingly recovered proteins were separated on 4–12% NuPAGE gradient gels (Invitrogen). After electrophoresis gels were incubated in fixation buffer (10% acetic acid and 20% ethanol in VE-H2O) for 15 min and washed for 1 h with VE-H2O thereby renewing the water every 20 min. Gels were dried onto chromatography paper (Whatman) for 2 h at 65◦C using the Model 583 Gel Dryer (Bio-Rad). Exposure of the film was carried out over night at room temperature in an exposition cassette. Radioactivity was detected by a phosphor imager (Cyclone Plus Storage Phosphor System; PerkinElmer) and visualized via the OptiQuant software.

# Tat-Cre Treatment

PCN of 5xFAD/Lrp1flox/flox mouse embryos (E14) were treated with Cre-recombinase fused to a basic protein translocation peptide derived from HIV-TAT (Tat-Cre) (provided by Dr. Roosmarijn E. Vandenbroucke; Inflammation Research Center, VIB, Ghent, Belgium; Department of Biomedical Molecular Biology, Ghent University, Ghent, Belgium) at DIV 4. Therefore, the culture medium was reduced to 2 ml and the recombinase was added in a final concentration of 200 nM. As control, PCN were treated with the Tat-Cre buffer (20 mM HEPES, 0.6 M NaCl, pH 7.4). Cells were incubated with the Tat-Cre recombinase or the vehicle alone for 48 h at 37◦C before cell lysis at DIV 6.

#### Inhibitor Treatment

To study the processing of APP dimers at the cell surface CHO cells stably expressing APP695 K587C were treated with ADAM10 inhibitor (GI254023X; provided by Dr. Ludwig, TU Aachen) 24 h after seeding on 6 cm dishes. Therefore, the medium was reduced to 2 ml fresh medium containing 10 µM ADAM10 inhibitor (stock solution: 10 mM in DMSO stored at −20◦C) or DMSO as vehicle control. Cells were maintained under these conditions for 24 h.

#### Cell Surface-Biotinylation

The surface levels of APP dimers were examined 24 h after reduction of the medium and inhibitor treatment. After collecting the conditioned medium cells were rinsed 3 times with ice-cold PBS. Surface proteins were biotinylated by addition of 0.25 mg/ml Sulfo-NHS-LC-LC-Biotin (Thermo Fisher Scientific) dissolved in 1 x PBS for 40 min at 4◦C thereby refreshing the biotin solution after 20 min. To quench unconjugated biotin, cells were washed 4 times with 50 mM NH4Cl in ice-cold 1 x PBS. Cells were lysed in NP-40 buffer containing 1 x PI. Equal protein amounts were incubated over night at 4◦C with NeutrAvidin Agarose Resin (Pierce). Unbound proteins were removed in 3 washing steps with NP-40 buffer and centrifugation at 4◦C with 24 × g for 2 min. Beads were boiled at 80◦C in 4 x SDS sample buffer for 5 min to elute proteins, which were separated on 6% Bis-Tris gels.

#### CSF Isolation

Cerebrospinal fluid (CSF) was harvested from 4 months old 5xFAD/Lrp1flox/flox and 5xFAD/Lrp1BE <sup>−</sup>/<sup>−</sup> mice by puncture of the cisterna magna as described previously (Vandenbroucke et al., 2012; Storck et al., 2016). Cell free CSF was obtained by centrifugation at 800 × g for 10 min at 4◦C. 2 µl of CSF were diluted in water and mixed with equal amounts of 2 x loading dye (0.72 M Bis-Tris, 0.32 M Bicine, 30% (w/v) sucrose, 2% SDS, 0.02% bromophenol blue without βME). Samples were denatured at 70◦C for 5 min to maintain putative dimerization of sAPP. Samples were separated by SDS-PAGE on 7% polyacrylamide SDS gels, transferred onto nitrocellulose membranes (Amersham Hybond ECL) and then blocked in 5% (w/v) non-fat dry milk in TBST (20 mM Tris, 137 mM NaCl, 0.1% (v/v) Tween-20). The antibody mix 1G75A3 (1:300) was used to detect sAPP.

#### Animals

In vivo analyses were performed with tamoxifen-inducible 5xFAD mice lacking Lrp1 in brain endothelial and choroid plexus epithelial cells (5xFAD/Lrp1BE <sup>−</sup>/−) (described in detail in Storck et al., 2016). 5xFAD mice, which represent a well-established AD model harboring 3 APP mutations and 2 PSEN1 mutations that are linked to FAD, served as LRP1 expressing controls. All animal studies were conducted in compliance with European and German guidelines for the care and use of laboratory animals and were approved by the Central Animal Facility of the University of Mainz and the ethical committee on animal care and use of Rhineland-Palatinate, Germany. Mice were housed on a 12-hlight cycle and had ad libitum access to water and a standard laboratory diet. To induce knock-out of Lrp1 in CSF-secreting epithelial cells of the choroid plexus in 5xFAD/Lrp1BE <sup>−</sup>/−, 12 week-old animals were injected i.p. with 2 mg tamoxifen (T5648, Sigma-Aldrich) for 7 consecutive days as described in Storck et al. (2016). After tamoxifen injection the standard laboratory diet was changed to chow supplemented with 400 mg tamoxifen citrate per kilogram dry weight (CRE Active TAM400, LASvendi) to maintain Cre-mediated recombination.

#### Quantification and Statistical Analysis

Western Blots and phosphor imager results were quantified by densitometry using ImageJ (1.44 or 1.46r) or Multi Gauge V3.0, respectively. The Graph Pad Prism 4 software (Graph Pad; La Jolla) provided the basis for compilation of the shown graphs and for statistical analysis. Data were analyzed by Student's t-test or one-way ANOVA followed by Tukey's post-hoc test. For live cell analysis at least 5 kymographs were analyzed. Student's t-test was used when comparing only two sets of data or one-way ANOVA followed by Bonferroni post-hoc test when comparing three sets of data and given the data were normaly distributed, respectively. The Kruskal-Wallis-Test followed by Dunn's Multiple Comparison Test was used to assess statistical differences between three sets of data given that data weren't normally distributed or variance was significantly different. The level of significance was set at p < 0.05 (<sup>∗</sup> ), p < 0.01 (∗∗) and p < 0.001 (∗∗∗).

# RESULTS

#### APP Dimers Are Generated and Processed in Cortical Neurons

As described before 30–50% of APP are present in a dimerized form in human brain (Munter et al., 2007; Schmidt et al., 2012). To investigate, whether APP695 dimer formation can be analyzed in a neuronal system, we infected primary cortical neurons of C57BL/6J mouse embryos (E14) with an adenovirus driving expression of human APP695. Indeed, we were able to detect APP dimers in the lysate of DIV 8 mouse neurons (**Figure 1A**) comparable to the expression of human APP dimers in HEK cells (**Figure 1B**). Likewise, by analyzing the supernatant of the

FIGURE 1 | APP dimer generation and processing takes place in primary cortical neurons. (A) Murine primary cortical neurons (DIV 7) were infected with an adenoviral vector encoding human APP695 while (B) HEK 293T cells were transiently transfected with the pLHCX-APP695 wt construct. 24 h post infection or transfection, respectively, conditioned media (CM) were collected and cells were lysed in RIPA (PCN) or NP-40 (HEK) lysis buffer. Via the antibody mix 1G75A3 (1:3,000) APP was detected in lysates (upper blots) and conditioned media (lower blots). PCN show similar APP dimer expression in the lysate as HEK cells and also generate soluble APP dimers. Under reducing conditions using β-mercaptoethanol (βME) and heating at 95◦C the dimer band disappeared. All lanes of lysate or conditioned medium are on the same blot but were rearranged for better presentation.

same cells, we observed soluble APP dimers in the conditioned medium suggesting that APP dimers are not only generated but also processed in neuronal cells as well as in human kidney cells (**Figures 1A,B)**. To verify the existence of APP dimers linked by disulfide bridges as we have described before (Isbert et al., 2012), the samples were boiled in sample buffer containing β-mercaptoethanol (βME). Note, that in the samples under reducing conditions the disulfide bonds were dissolved and the dimer band signal decreased, whereas the signal intensity for monomeric APP increased (**Figures 1A,B**). These data show that disulfide-bound sAPP dimers, formed most likely in the ER, are anterogradely transported and shed by secretases in a dimerized status.

#### Monomeric and Dimeric APP Show Similar Transport Characteristics

As we found that neurons secrete disulfide-bound dimerized sAPP, we wondered if monomeric and dimeric APP are transported along the secretory pathway in the same or diverse types of transport vesicles. For this purpose we used an inducible FK501-binding-protein (FKBP) -based dimerization system (Rollins et al., 2000), previously used for analysis of APP processing in dependence of APP dimerization (Eggert et al., 2009). For live cell imaging expression constructs encoding for C-terminal tagged GFP APP-FKBP fusion proteins were generated (APP-F1-GFP) (Video in Supplementary Material 1). For control, we first verified that APP-GFP and non-dimerized APP-F1-GFP exhibit identical transport characteristics (**Figure 2A**). Futhermore, as GFP has a weak tendency to self-dimerize (Chalfie and Kain, 2005), we tested if APP-GFP might exhibit in comparison to APP altered dimerization properties, by using the blue-native gel system (Eggert et al., 2009). Notably, we observed for APP-GFP no increase in dimerization properties (Supplementary Figure 1). In the next step, the transport of non-dimerized (APP-F1-GFP + EtOH) and dimerized APP (APP-F1-GFP + dim.) was compared (**Figure 2A**). Surprisingly, the induction of APP dimerization had no significant influence on APP transport velocities in anterograde or retrograde direction, respectively (**Figures 2B,C**). The majority of APP vesicles moved with a velocity between 0.5 and 2.5 µm/s in both directions, independent of their dimerization status. These data suggest that monomeric and dimeric APP are transported by the same kinesin dependent transport machinery.

#### LRP1 Deficiency Leads to Accelerated Trafficking of APP Dimers

Since APP cis-homodimers and monomers show similar transport characteristics, we assumed that both follow the same principle. Previously, we demonstrated that LRP1 influences monomeric APP transport along the secretory pathway (Waldron et al., 2008). Therefore, we wanted to analyze now, whether a lack of LRP1 may also affect trafficking of APP dimers. For this purpose, we generated a cDNA construct providing the continuous expression of SDS-stable APP cis-dimers. The expression construct exhibits a triplet mutation in the coding

#### Herr et al. APP Trafficking/Dimerization Involves LRP1

#### FIGURE 2 | Continued

After 18–20 h and 1 h prior live cell recording of axonal vesicle movements, APP-F1-GFP expressing neurons were either treated with 100 nM AP20187 (dimerizer) or for control with the vehicle of the dimerizer, ethanol (negative control). (A) Representative primary neuron and kymographs of cells expressing APP GFP or APP-F1-GFP treated with dimerizer or ethanol respectively. The ROI is marked by a rectangle. Bar: 20µm. (B) Vesicle distribution and (C) anterograde and retrograde transport velocities of APP-GFP, non-dimerized APP-F1-GFP (ethanol control) and APP-F1-GFP dimerized vesicles. No differences among APP variants could be observed (one-way ANOVA followed by Bonferroni post hoc test). Bars represent mean values ± SEM, *n* = 3 (≥16 cells per approach).

sequence of APP695, which leads to an amino acid exchange from lysine (K) to cysteine (C) at position 587 (APP695 K587C) (**Figure 3A**). This mutation enabled the formation of APP cis-dimers by disulfide bridges between the cysteine residues in the E2 domain of two mutant APP molecules. According to our expectation a stepwise increase of temperature up to 95◦C showed only a slight decrease of APP K587C dimers, indicating that most of the APP K587C dimers are stabilized by intramoleclular disulfide bonds (**Figures 3B,C**). To get further insights on the generation and processing of APP cis-dimers in regard to LRP1, we performed a pulse-chase assay with CHO K1 and LRP1-deficient CHO 13-5-1 cells stably expressing APP695 K587C dimers. This assay revealed that sAPP dimers were already immunoprecipitated after a 30 min chase in CHO 13-5-1 cells while in CHO K1 cells shed APP dimer fragments were first detectable after a 1 h chase (**Figure 3D**). Quantification of the sAPP dimer/APP dimer ratio showed an increase of this ratio in LRP1-deficient cells (**Figure 3E**) suggesting an earlier availability of APP dimers for shedding at the cell surface. Thus, these results point to an accelerated trafficking of APP dimers in the absence of LRP1.

#### LRP1 Alters APP Transport Characteristics

We observed that anterograde transport of monomeric and dimeric APP is affected by LRP1, suggesting that LRP1 and APP might be sorted in common transport vesicles and that LRP1 might be required for APP sorting. To address this hypothesis, we performed co-stainings of endogenous APP and LRP1 in primary cortical neurons (PCN) and used again a live cell imaging approach.

Immuncocytochemical analysis of PCN differentiated for 7 days in vitro using anti-APP and anti-LRP1 antibodies revealed a strong cytoplasmic staining within the cell body and a punctate staining of LRP1 and APP in neurites, at least in part representing transport vesicles (Supplementary Figure 2). Interestingly, we observed only a low co-localization rate, arguing that LRP1 and APP are mostly transported in different transport vesicle types.

An expression construct encoding an N-terminal myc tagged LRP-mini-receptor (Rabiej et al., 2015) was used for generation of a C-terminally GFP tagged LRP-mini-receptor (LRP1-GFP). After verification that the newly generated LRP1-GFP fusion protein was expressed as full-length protein and that the GFPtag did not alter the subcellular localization (data not shown) the construct was used for live cell imaging. First, we wanted to analyze transport velocities of APP-RFP and LRP1-GFP in single transfected primary hippocampal mouse neurons (PHN). Time lapse series of 30 s were recorded at an interval of 200 ms/frame and vesicle movement was quantified based on the analysis of kymographs (**Figures 4A,B;** Video in Supplementary Material 2). Quantification showed that the largest fraction (68%) of anterograde LRP1-GFP-positive vesicles was transported with a velocity of 1–2 µm/s in contrast to APP-RFP that was mostly (66%) transported in vesicles faster than 2 µm/s (**Figure 4C**). Also for retrograde moving vesicles, a clear difference in transport characteristics was observed (**Figure 4D**). Although most of the retrograde transport vesicles containing APP-RFP and LRP1-GFP moved with a velocity of 1–2 µm, a fraction of APP-RFP positive transport vesicles showed retrograde transport characteristics with velocities >2 µm/s, which was not observed for LRP1-GFP containing vesicles. These data suggest that the majority of LRP1 and APP are transported in distinct anterograde transport vesicles, whereas a larger fraction of LRP1 and APP is co-transported in retrograde transport vesicles.

Further, we tested if LRP1 and APP co-expression might affect APP transport characteristics and vice versa. For this purpose, we performed live cell imaging analyses of PHNs co-expressing LRP1-GFP and APP-RFP 18 to 20 h post transfection, as described above (representative kymographs **Figures 4E,F**). Quantification revealed that LRP1-GFP and APP-RFP are co-transported in common anterograde and retrograde transport vesicles (**Figures 4G,H**). Most interestingly, coexpression of LRP1-GFP caused a change of APP-RFP transport characteristics, that was highly similar to those observed in single transfected cells for LRP1-GFP (**Figures 4A–D**), whereas LRP1-GFP transport upon co-expression of APP remained unchanged (**Figures 4G,H**). This holds true for APP/LRP1 anterograde (**Figure 4G**) as well as retrograde (**Figure 4H**) transport. Accordingly, also the mean velocities were strongly reduced upon co-expression of LRP1 (**Figure 5E**). Notably, the relative amount of anterograde, retrograde and stationary vesicles remained unchanged, arguing that LRP1 not simply holds back APP in the Golgi. Instead, our data indicate that co-expression of LRP1 may cause a recruitment of APP into common transport vesicles, that exhibit different transport characteristics.

To further validate that LRP1 modifies APP intraneuronal transport, we analyzed APP-RFP transport in primary neurons with reduced LRP1 levels. For this purpose, we used PCN of Lrp1flox/flox mouse embryos, treated with 200 nM Crerecombinase fused to a basic protein translocation peptide derived from HIV-TAT (Tat-Cre) for 48 h prior live cell imaging. Reduced LRP1 expression of about 2-fold was validated by Western Blot analysis (**Figures 5A**, **7A**). Interestingly, we observed in contrast to LRP1 co-expression only a tendency toward increased APP transport velocity in anterograde direction (p = 0.051) (**Figures 5C,F**) and no change in retrograde direction or for the amount of stationary vesicles (**Figures 5D,F**). In contrast, LRP1 deficiency caused a significant (p = 0.011) decrease of stationary and an increase (p = 0.011) of moving transport vesicles (**Figure 5F**). Seperation of moving vesicle data into anterograde and retrograde transport revealed due to lower n-number not the significance levels (p = 0.06) (**Figure 5F**).

Together, our data show that elevated LRP1 expression causes a decrease of APP transport rate whereas reduced levels of LRP1 cause an increase of APP transport rates.

#### LRP1 Expression Affects Processing of APP695 K587C Dimers

Showing that the expression of LRP1 alters trafficking of monomeric as well as dimerized APP, we assumed that LRP1 may also affect processing of APP dimers. We previously demonstrated that internalization of APP (mostly monomeric and possibly also dimeric) from the cell surface is reduced in the absence of LRP1 resulting in an increase in sAPPα secretion (Pietrzik et al., 2002). To investigate, whether a similar effect is obtained also for covalently bound APP homodimers, we used CHO K1 and LRP1-deficient CHO 13-5-1 cells both expressing APP695 K587C exogenously. In Western Blot analyses we first compared APP dimer expression and sAPP dimer secretion of both cell lines (**Figure 6A**). Here, we detected lower APP dimer expression in the lysate of CHO 13-5-1 cells compared to CHO K1 despite a comparable total protein load. However, the ratio of sAPP dimers to dimeric APP of LRP1-deficient CHO cells was approximately 3-fold stronger than in CHO K1 cells. To test, if this difference may be explained by increased processing due to decreased internalization of APP dimers from the cell surface in LRP1-deficient CHO 13-5-1 cells, we transfected CHO K1 and CHO 13-5-1 cells with an APP695 dimer construct, exhibiting a mutation in the APP internalization motif "YENPTY" (Lai et al., 1995; MarquezSterling et al., 1997). The amino acid exchange from NPTY to NGYE leads to a reduced internalization of APP dimers from the cell surface mimicking LRP1 deficiency. We expected that exogenous expression of APP695 K587C NGYE compared to APP695 K587C should increase sAPP dimer

Anterograde and (D) retrograde transport vesicles containing APP-RFP (white columns) or LRP-GFP (black columns). Note the change of APP-RFP transport characteristcs in (G) anterograde and (H) retrograde direction (light gray columns) upon co-expression of (G,H) LRP1-GFP (dark gray columns). Bars represent mean values ± SEM, *n* > 5 (≥254 vesicles); Student's *t*-test, *p* < 0.05 (\*), *p* < 0.01 (\*\*).

secretion in CHO K1 cells, whereas expression in CHO 13-5-1 cells, that already show reduced APP dimer internalization due to the absence of LRP1, should cause no further increase in sAPP secretion. According to our expectation, we detected higher amounts of soluble APP dimers in the conditioned medium of CHO K1 cells, but no significant difference in sAPP dimer secretion in CHO 13-5-1 cells (**Figure 6B**). This further supports the hypothesis that LRP1 affects internalization of APP monomers and dimers in a comparable manner.

# Lrp1 Knock-Out in PCN Affects APP Dimer Processing

As we could show that APP dimers are formed and processed in primary cortical neurons (**Figure 1**), we wanted to analyze the effect of a Lrp1 knock-out on APP dimer processing in neuronal cells. Hence, PCN from 5xFAD/Lrp1flox/flox mouse embryos (E14) were treated with Tat-Cre recombinase to induce the excision of Lrp1 via recombination of the loxP recognition sites flanking this gene. Western Blot analysis revealed a 2-fold reduction of LRP1 expression in PCN treated with Tat-Cre for 48 h compared to neurons incubated with the vehicle (**Figure 7A**). We assume that the incomplete reduction of LRP1 was due to its long half-life (24 h) (Reekmans et al., 2010). Interestingly, the sAPP dimer/APP dimer ratio of neurons with a partial Lrp1 knock-out showed a more than 2-fold increase in comparison to the buffer treated PCN (**Figure 7B**). These observations are similar to the effects seen in LRP1-deficient CHO 13-5-1 cells and might be explained by a faster transport rate of APP dimers to and/or less internalization from the cell surface. This may result in an elevated APP processing by the active secretases at this site due to earlier and/or prolonged substrate availability.

# Processing of APP Cis-Dimers by Meprin β and ADAM10

The presence of soluble APP dimers indicates that APP cisdimers are enzymatically cleaved thereby releasing soluble dimerized fragments. Thus, we wanted to investigate, whether processing of APP cis-dimers can be attributed to the same secretases known to be responsible for cleavage of monomeric APP. Regarding the processing of monomeric APP at the cell surface, the metalloproteinases ADAM10 (Weidemann et al., 1989; Lammich et al., 1999) and meprin β (Jefferson et al., 2011; Bien et al., 2012; Schönherr et al., 2016) are implicated. As the metalloproteinase meprin β itself occurs in a dimerized form (Bertenshaw et al., 2003; Kruse et al., 2004), we first focused on the role of meprin β in APP dimer cleavage. To address this point, we co-transfected HEK 293T cells with either APP695 K587C or APP695 wt and the meprin β construct to analyze processing of APP cis-dimers in comparison to monomeric APP cleavage by this secretase. As a control for meprin β activity, cells were solely transfected with the wt APP or the dimer-bearing APP construct. As expected, analysis of the conditioned medium of transfected HEK 293T cells revealed

higher sAPP dimer levels in cells expressing the dimer-bearing APP construct, compared to those transfected with wildtype APP695 (**Figure 8A**). Co-transfection with meprin β resulted in a decrease in the signal for monomeric as well as dimerized soluble APP irrespective of the APP construct used for transfection. Interestingly, the reduction of sAPP dimers was considerably stronger than that of monomeric sAPP. To quantify these observations, we calculated the ratio of sAPP dimers secreted from APP/meprin β expressing cells to sAPP dimer secretion of solely APP expressing cells as well as APP K587C/meprin β to APP K587C expressing cells. APP695 wt expressing cells showed no significant difference in the ratio for monomeric and dimerized sAPP, possibly due to the weak signal for sAPP dimers (**Figures 8A,B**). In contrast, processing of APP695 K587C by meprin β was significantly increased compared to cleavage of monomeric sAPP (**Figures 8A,B**). In line with this, meprin β co-transfection resulted also in an increase of Aβ secretion (Supplementary Figure 3). Together, these data suggest a higher affinity of meprin β for dimerized than for monomeric APP695.

To investigate the role of α-secretase cleavage in APP cisdimer processing, CHO K1 and CHO 13-5-1 cells expressing APP695 K587C were treated with the ADAM10 inhibitor GI254023X (Ludwig et al., 2005). Quantification of APP dimer expression in the lysates of CHO K1 and CHO 13-5-1 cells after incubation with GI254023X revealed an increase of APP dimers of 50 or 56%, respectively (**Figure 9A**). In line with this, we detected an average decrease in sAPP dimer secretion of 38% for inhibitor treated CHO K1 cells compared to those incubated with the vehicle DMSO alone (**Figure 9B**). A similar result (55% reduction of sAPP dimers after ADAM10 inhibition) was observed in CHO 13-5-1 cells (**Figure 9B**). As ADAM10 cleaves APP at the cell surface (Lammich et al., 1999), we expected an accumulation of the mature cell surface exposed APP, after treatment with GI254023X. Indeed, cell surface biotinylation assays using APP695 K587C expressing CHO cells revealed after

ADAM10 inhibition an increase of mature cell surface APP dimers of 43% in comparison to DMSO controls (**Figure 9C**). In CHO 13-5-1 cells treated with GI254023X the surface expression of APP dimers was increased by 54% compared to DMSO controls (**Figure 9C**), underlining our assumption that LRP1 deficiency accelerates availability of APP dimers for processing by ADAM10.

Together, these data indicate that processing of APP cisdimers can be attributed to the same secretases (meprin β and ADAM10) with meprin β shedding APP preferentially in the dimeric form.

# LRP1 Expression Affects sAPP Dimer Secretion In vivo

As secretion of monomeric sAPP fragments into the cerebrospinal fluid (CSF) of AD patients has been shown previously (Van Nostrand et al., 1992; Sennvik et al., 2000; Olsson et al., 2003; Brinkmalm et al., 2013), we wanted to study the generation of dimeric sAPP in vivo. To investigate, whether LRP1 expression also affects processing of APP dimers in vivo, we analyzed the cerebrospinal fluid of 5xFAD mice expressing LRP1 and of 5xFAD mice with a tissue-specific Lrp1 knock-out in brain endothelial cells and the choroid plexus epithelial cells. In 5xFAD mice only very little amounts of sAPP dimers could be detected (**Figure 10A**). However, in 5xFAD mice lacking LRP1 in CSF-secreting epithelial cells of the choroid plexus an about 4-fold stronger immunoreactivity for sAPP dimers was observed (**Figure 10B**). In line with our previous results showing preferred clavage of APP dimers by meprin β (**Figure 8**), the immense increase in dimerized APP fragments may be explained by the involvement of meprin β besides ADAM10 in APP cleavage at the surface of epithelial cells. These in vivo data underline that LRP1 preferentially affects sAPP dimer secretion.

#### DISCUSSION

Our data show that LRP1 recruits APP into common fast axonal transport (FAT) membrane bound organelles (MBOs), suggesting that LRP1 functions as a sorting receptor. Thereby, increased levels of LRP1 slow down APP anterograde transport and decrease its endocytosis rate. This in turn causes an increase of surface APP and thus accelerates secretion of sAPP. Interestingly, we observed the same influence for APP monomers and dimers. However, Lrp1 knock-out in choroid plexus cells increased sAPP monomer secretion, but much more pronounced sAPP dimer secretion in the CSF. This is likely explained by different processing properties of cell surface APP monomers/dimers, as we found that meprin β preferentially cleaves APP dimers.

Our live cell imaging analyses in primary neurons show that LRP1 is anterogradely transported with a median velocity of 1–2 µm/s, (**Figure 4**), indicating that LRP1 anterograde transport is mediated by the fast axonal kinesin dependent transport (FAT) machinery. Time lapse analysis of APP from our group and others revealed transport velocities of 2–10 µm/s (**Figures 2**, **4**; Kaether et al., 2000; Szodorai et al., 2009; Hermey et al., 2015). Those types of transport vesicles with velocities above 2 µm/s have only been observed very rarely for LRP1 positive vesicles. In line with the low extend of co-localization of LRP1 and APP in neurites, these data indicate that APP and LRP1 are transported in distinct membrane bound organelles (MBO), associated with different FAT machineries. Interestingly, reduced levels of LRP1 in primary neurons caused an increase of APP transport vesicles (**Figure 5**), whereas co-expression of LRP1 and APP caused an approximation of both transport characteristics, changing APP transport toward velocities observed for LRP1. These data corroborate our previous assumption that LRP1 causes a sorting of APP into LRP1 bearing MBOs (Waldron et al., 2008). As monomeric and dimeric APP are transported with very similar transport characteristics (**Figure 2**), and as a knock-out of Lrp1 caused an increase of monomeric as well as dimeric sAPP (Waldron et al., 2008; **Figure 3**) we assume that LRP1 affects monomeric and dimeric APP in a similar way. In contrast, other sorting receptors of APP, such as SorLA are assumed to affect the equilibrium of APP dimerization, causing different processing kinetics of monomeric and dimeric APP, as indicated by elegant mathematical modeling (Schmidt et al., 2012). Notably, these analyses were performed in cells lacking LRP1. Thus, it would be interesting for future studies to investigate the interplay of LRP1, SorLA and APP dimerization in more detail.

Since LRP1 recruits APP to transport vesicles and the velocity of vesicles carrying APP dimers is similar to APP monomer carrying vesicles we wondered whether APP dimers are released in a similar LRP1 dependent manner as APP monomers (Waldron et al., 2008). In a pulse-chase analysis of LRP1 expressing CHO K1 and LRP1-deficient CHO 13-5-1 cells both stably expressing APP dimers, we detected faster sAPP dimer

release in LRP1-deficient cells (**Figure 3D**). As APP dimerization already takes place in the ER and as those dimers remain stable throughout their transport to the plasma membrane (Isbert et al., 2012; Khalifa et al., 2012), an interaction of dimeric APP with LRP1 early in the secretory pathway may lead to a decelerated APP dimer trafficking, similar as shown for monomeric APP (Waldron et al., 2008). Thus, cell surface processing may be affected, resulting in the delayed occurrence of dimerized sAPP in LRP1 expressing cells.

The interaction of APP with LRP1 also plays an important role at the cell surface as monomeric APP is internalized in a complex with LRP1 by clathrin-mediated endocytosis, thereby affecting its processing (Knauer et al., 1996; Ulery et al., 2000; Pietrzik et al., 2002, 2004; Cam et al., 2005). Hence, we estimated that the accelerated generation of soluble APP dimers in LRP1 deficient cells compared to LRP1 expressing cells (**Figure 6A**) may result from a reduced APP dimer internalization due to LRP1 deficiency that in turn causes higher APP levels for processing at the cell surface. This assumption is strengthened by the fact that APP dimer constructs, harboring a mutated internalization motif caused an increase in sAPP levels (**Figure 6B**), as shown before for internalization deficient monomeric APP (Perez et al., 1999). Importantly, the sAPP dimer levels of the internalization deficient mutant were not increased in LRP1 lacking cells. To investigate, whether the LRP1 regulatory effects are similar in neuronal cells, we analyzed PCN with a knock-out of Lrp1 by Cre recombination. Similar as shown for LRP1-deficient CHO 13-5-1 cells we observed a decrease of APP in the lysates accompanied by an accelerated sAPP generation for monomeric as well as dimeric APP (**Figure 7**), when LRP1 expression was knocked out in primary neurons. The more than 2-fold increase of the sAPP dimer to APP dimer ratio further supports our hypothesis that LRP1 regulated effects on APP dimer transport and internalization take place in neuronal cells. A resulting earlier and/or prolonged availability of APP may therefore provide more substrate for the cell surface-active shaddases ADAM10 and meprin β. Thus, our data strongly suggest that monomeric and dimeric APP trafficking is equally affected by LRP1 (Ulery et al., 2000; Pietrzik et al., 2002; Cam et al., 2005) and that this process is also important in the neuronal system.

Due to the similar characteristics of monomeric and dimeric APP transport we assumed that APP dimers might also be processed by the same secretases as monomeric APP. Here, we concentrated on cell surface APP. It has been well documented over the last decade that ADAM10 is the most prominent sheddase of APP at the cell surface (Weidemann et al., 1989; Lammich et al., 1999; Jorissen et al., 2010; Kuhn et al., 2010). Recently, the metalloproteinase meprin β was identified to be also capable to process APP at the cell surface (Jefferson et al., 2011; Bien et al., 2012; Schönherr et al., 2016). Here, we show that meprin β as well as ADAM10 are implicated in dimer processing (**Figures 8**, **9**). Inhibition of ADAM10 resulted in a surface accumulation of APP dimers as shown by protein surface biotinylation. This was accompanied by an increase of APP dimers in the lysate and a decrease in soluble APP dimers in the conditioned medium of the tested cells (**Figures 9A–C**) as shown for monomeric sAPPα (Woods and Padmanabhan, 2013). The role of LRP1 in APP dimer processing is again highlighted by ADAM10 inhibition in LRP1-deficient cells as the demonstrated effects (in CHO K1 cells) were considerably stronger in CHO 13-5-1 cells (**Figures 9A–C**).

To investigate the role of meprin β in cleavage of APP dimers, we analyzed the sAPP monomer and dimer ratios of HEK cells transfected with APP alone to cells co-expressing meprin β. Interestingly, the sAPP dimer ratio was significantly increased compared to the sAPP monomer ratio (**Figure 8**) when the stabilized APP homodimer was generated. This indicates that meprin β processes APP, with a preference for APP dimers. Although we do not understand the underlying molecular mechanism yet, it appears reasonable that the higher affinity of meprin β for dimeric vs. monomeric APP

is connected to the fact that this secretase itself exists in form of a dimer (Bertenshaw et al., 2003; Kruse et al., 2004). Thus, dimerized APP may offer a cooperative effect on enzymatic activity of meprin β, as recently postulated for α- and β-secretase for APP dimer processing (Schmidt et al., 2012).

Due to the fact that a large fraction of APP occurs in a dimerized form in the brain (Munter et al., 2007; Schmidt et al., 2012) and that APP dimers are processed in PCN of C57BL/6J mice (**Figures 1**, **7**), we asked whether the regulatory effect of LRP1 in APP dimer processing does also play a role in vivo. Therefore, we analyzed the CSF of 5xFAD mice and 5xFAD mice with an induced Lrp1 knock-out in brain endothelial and choroid plexus epithelial cells. Indeed, we were able to detect sAPP dimers to an about 4-fold greater extent in the CSF of mice with the tissue-specific Lrp1 knock-out than of 5xFAD control littermates (**Figures 10A,B**). As the choroid plexus epithelial cells express APP (Kalaria et al., 1996; Bergen et al., 2015) and are the main producers of CSF (Brown et al., 2004), sAPP dimers in the CSF presumably originate from these cells. Thus, the increased amount of sAPP dimers in the CSF of 5xFAD/Lrp1BE <sup>−</sup>/<sup>−</sup> mice is likely due to reduced internalization of APP from the surface of the choroid plexus epithelial cells. This may provide more APP dimers for enzymatic cleavage at the cell surface. We could show that ADAM10, the main sheddase of APP at the cell surface (Weidemann et al., 1989; Lammich et al., 1999;

K1 and 54% for CHO 13-5-1 cells.

Jorissen et al., 2010; Kuhn et al., 2010), is aslo implicated in the cleavage of APP dimers (**Figure 9**). Thus, the processing of APP dimers in epithelial cells of the choroid plexus may be at least partially performed by this secretase. Furthermore, the especially high levels of sAPP dimers may point to a prominent role of meprin β in this context as we could show that this protease exhibits a preferred activity for dimeric vs. monomeric APP (**Figure 8**).

#### REFERENCES


Altogether, our studies show that LRP1 affects trafficking of APP monomers and dimers and that APP dimers are preferentially cleaved by ADAM10 and meprin β. Hence, dimerized APP may affect physiological as well as pathogenic functions of APP by different transport and processing characteristics and should be included in future studies regarding the interplay with other sorting receptors than LRP1 or the generation of Aβ species.

# AUTHOR CONTRIBUTIONS

CP and SK conceived and supervised the study. UH, PS, CP, and SK designed the experiments. UH, PS, SES, CT, VR, AJ, SS, NS, CD, and SE performed experiments. UH, PS, SES, CP, and SK analyzed the data. UH and PS wrote and revised the manuscript. CP and SK reviewed the manuscript.

# FUNDING

Research in our laboratories was funded by the Deutsche Forschungsgemeinschaft (FOR1332, to CP and SK; PI379/6-2, KI819/6-2) and the Stiftung Rheinland-Pfalz für Innovation (to CP and SK).

#### ACKNOWLEDGMENTS

We thank Prof. Dr. E. Koo from UC San Diego School of Medicine (USA) for providing the antibody mix 1G75A3 and Prof. Dr. A. Ludwig from the University Medical Center RWTH Aachen for providing the ADAM10 inhibitory compound GI254023X. We further thank Roosmarijn E. Vandenbroucke from the Flanders Institue for Biotechnology (VIB; Belgium) for providing the Tat-Cre recombinase.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fnmol. 2017.00118/full#supplementary-material


receptor-related protein (LRP). Annu. Rev. Biochem. 63, 601–637. doi: 10.1146/annurev.bi.63.070194.003125


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Herr, Strecker, Storck, Thomas, Rabiej, Junker, Schilling, Schmidt, Dowds, Eggert, Pietrzik and Kins. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# APP Function and Lipids: A Bidirectional Link

Marcus O. W. Grimm1,2,3 \* † , Janine Mett 1† , Heike S. Grimm<sup>1</sup> and Tobias Hartmann1,2,3

<sup>1</sup>Experimental Neurology, Saarland University, Homburg/Saar, Germany, <sup>2</sup>Neurodegeneration and Neurobiology, Saarland University, Homburg/Saar, Germany, <sup>3</sup>Deutsches Institut für DemenzPrävention (DIDP), Saarland University, Homburg/Saar, Germany

Extracellular neuritic plaques, composed of aggregated amyloid-β (Aβ) peptides, are one of the major histopathological hallmarks of Alzheimer's disease (AD), a progressive, irreversible neurodegenerative disorder and the most common cause of dementia in the elderly. One of the most prominent risk factor for sporadic AD, carrying one or two aberrant copies of the apolipoprotein E (ApoE) ε4 alleles, closely links AD to lipids. Further, several lipid classes and fatty acids have been reported to be changed in the brain of AD-affected individuals. Interestingly, the observed lipid changes in the brain seem not only to be a consequence of the disease but also modulate Aβ generation. In line with these observations, protective lipids being able to decrease Aβ generation and also potential negative lipids in respect to AD were identified. Mechanistically, Aβ peptides are generated by sequential proteolytic processing of the amyloid precursor protein (APP) by β- and γ-secretase. The α-secretase appears to compete with β-secretase for the initial cleavage of APP, preventing Aβ production. All APP-cleaving secretases as well as APP are transmembrane proteins, further illustrating the impact of lipids on Aβ generation. Beside the pathological impact of Aβ, accumulating evidence suggests that Aβ and the APP intracellular domain (AICD) play an important role in regulating lipid homeostasis, either by direct effects or by affecting gene expression or protein stability of enzymes involved in the de novo synthesis of different lipid classes. This review summarizes the current literature addressing the complex bidirectional link between lipids and AD and APP processing including lipid alterations found in AD post mortem brains, lipids that alter APP processing and the physiological functions of Aβ and AICD in the regulation of several lipid metabolism pathways.

#### Edited by:

Thomas Deller, Goethe-University, Germany

#### Reviewed by:

Anthony J. Turner, University of Leeds, UK Lorena Perrone, Université Grenoble Alpes, France Joachim Herz, University of Texas Southwestern Medical Center, USA

#### \*Correspondence:

Marcus O. W. Grimm marcus.grimm@uks.eu

†These authors have contributed equally to this work.

Received: 31 October 2016 Accepted: 24 February 2017 Published: 10 March 2017

#### Citation:

Grimm MOW, Mett J, Grimm HS and Hartmann T (2017) APP Function and Lipids: A Bidirectional Link. Front. Mol. Neurosci. 10:63. doi: 10.3389/fnmol.2017.00063 Keywords: lipids, APP processing, AICD, Abeta cholesterol, sphingolipids, PUFA, sulfatides, gangliosides

#### ALZHEIMER'S DISEASE

Worldwide currently there are more than 46 million people suffering from dementia and the number of affected individuals is estimated to double every 20 years. Alzheimer's disease (AD) is a devastating neurodegenerative disorder, which is the most common cause of dementia in the elderly population. Clinically AD is characterized by a progressive loss of cognitive brain functions leading to memory dysfunction, impaired judgment, disorientation and finally to a total loss of memory and personality (Plassman et al., 2007; World Alzheimer Report, 2015). AD-patients typically die in average within 3–10 years after diagnosis due to secondary disorders (Zanetti et al., 2009). The clinical symptoms of AD might be caused by an extensive loss of synapses and neurons leading to a strong hippocampal and cortical atrophy (Scheff and Price, 1993; Gómez-Isla et al., 1996; Mouton et al., 1998; Dickerson et al., 2001). The characteristic neuropathological hallmarks of the disease are intracellular neurofibrillary tangles (NFTs) and extracellular localized amyloid plaques. While the NFTs are composed of the microtubuli-associated protein tau in a hyperphosphorylated state (Grundke-Iqbal et al., 1986a,b), the amyloid plaques are mainly built up of amyloid-β (Aβ) peptides. Aβ-peptides are hydrophobic, 38–43 amino acid long products generated by the sequential proteolytic processing of the amyloid precursor protein (APP; Glenner and Wong, 1984; Masters et al., 1985; Kang et al., 1987). The significant cerebral accumulation of Aβ, starting several years prior to the first symptoms, is respected to trigger the disease process (Glenner and Wong, 1984; Glenner, 1989; Hardy and Higgins, 1992; Hardy and Selkoe, 2002). Especially the accumulation of Aβ42 (indicating 42 amino acids), which is the major Aβ species found in neuritic plaques, is considered to initiate AD progression (Iwatsubo et al., 1994; Tamaoka et al., 1995). Due to the additional hydrophobic amino acids isoleucine and alanine Aβ42 has a higher tendency to aggregate compared to the more prevalent Aβ40 (indicating 40 amino acids; Jarrett et al., 1993). Increasing evidence suggests small oligomers of Aβ to represent the most toxic form of the peptide (Lambert et al., 1998; Lesné et al., 2006; Shankar et al., 2008). Several mechanisms are discussed to contribute to Aβ neurotoxicity, among them the induction of inflammatory processes, a disruption of calcium homeostasis and membrane integrity, cholinergic and mitochondrial dysfunction and increased oxidative stress (Grimm and Hartmann, 2012).

There are two forms of AD, the more common sporadic AD with a disease onset after the age of 65 (late onset AD, LOAD) and the genetically based form (familial AD, FAD) with an earlier manifestation of symptoms. The two variants are basically distinguishable from each other in clinical and neuropathological terms. Less than 5% of all AD-cases belong to FAD which is caused by mutations in the genes encoding for APP and the presenilins (PS) 1 and 2, proteins involved in proteolytic APP-processing (Levy et al., 1990; Goedert et al., 1994; Levy-Lahad et al., 1995; Sherrington et al., 1995; Tanzi, 2012). Besides aging, hypercholesterolemia, hypertension, atherosclerosis, homocysteinemia, diabetes mellitus and obesity are discussed as non-genetic risk factors for LOAD (Barnes and Yaffe, 2011; Polidori et al., 2012). The ε4 allele of the apolipoprotein E (ApoE) has been identified as the most important genetic risk factor for the sporadic form of the disease (Corder et al., 1993; Strittmatter et al., 1993).

As already mentioned, Aβ is generated by proteolytic processing of the precursor protein APP. APP is a ubiquitously expressed type I-transmembrane protein cycling between the plasma membrane and acidic intracellular compartments (Haass et al., 1992; Koo and Squazzo, 1994; Thinakaran and Koo, 2008). It consists of a large ectodomain, a single transmembrane domain and a short intracellular part. APP belongs to an evolutionary conserved protein family including the APP-like proteins 1 and 2 (APLP1, APLP2) in mammals. APP can be sequentially cleaved via two different pathways (Haass et al., 1992; Thinakaran and Koo, 2008; De Strooper, 2010; **Figure 1**). In the predominant non-amyloidogenic processing pathway the generation of Aβ is precluded. It is initiated by the α-secretase dependent cleavage of APP within the Aβ-domain shedding off the soluble ectodomain sAPPα and generating the membrane-anchored C-terminal fragment (CTF) C83 (indicating 83 amino acids). Members of the ADAM (a disintegrin and metalloprotease) protein family have been identified as catalytically active α-secretases with ADAM10 representing the physiologically relevant, constitutive α-secretase in neurons (Lammich et al., 1999; Kuhn et al., 2010). In contrast, the aspartyl protease β-site APP cleaving enzyme 1 (BACE1) initiates the amyloidogenic APP-processing pathway generating the membrane-spanning CTF C99 (indicating 99 amino acids) and releasing sAPPβ into the extracellular space (Vassar et al., 1999). The two alternative pathways differ in their subcellular localization: due to the acidic pH-optimum of BACE1 the amyloidogenic APP-processing is localized in acidic intracellular compartments, while non-amyloidogenic APP-processing mainly takes place at the cell surface (Parvathy et al., 1999; Grbovic et al., 2003; Carey et al., 2005). In both pathways the CTFs are subsequently processed by the γ-secretase complex, which consists of the proteins PS1 or PS2 as the catalytic core, Aph1 (anterior pharynx defective 1) a or b, PEN2 (presenilin enhancer 2) and nicastrin (Baulac et al., 2003; Edbauer et al., 2003; Kimberly et al., 2003). The γ-secretase possesses the unusual property to cleave its substrates within their transmembrane domains after shedding off the ectodomain, a process called regulated intramembrane proteolysis (RIP; Brown et al., 2000; Lichtenthaler et al., 2011). This catalytic activity leads to the generation of the non-toxic peptide p3 out of C83 and of Aβ out of C99 combined with the release of APP intracellular domain (AICD) into the cytosol in both processing pathways (Passer et al., 2000; Kakuda et al., 2006; Grimm and Hartmann, 2012). Due to multiple γ-secretase cleavage sites within the transmembrane domain of APP, the generated Aβ- and AICD-peptides can vary in length (Funamoto et al., 2004; Qi-Takahara et al., 2005; Kakuda et al., 2006). AICD is reported to translocate to the nucleus and to regulate the transcription of target genes, among them the genes encoding for APP, BACE1, the Aβ-degrading protease neprilysin (NEP) as well as several enzymes involved in lipid metabolism (Cao and Südhof, 2001; von Rotz et al., 2004; Grimm et al., 2011b,d, 2012c, 2013, 2015a).

#### LINK BETWEEN LIPIDS AND AD

A link between AD pathology and lipids was already observed more than a century ago by Alois Alzheimer, who described a higher occurrence of ''adipose inclusions'' or ''lipoid granules'' in post mortem AD-brain tissue as a third pathological hallmark of the disease (Foley, 2010). In the meantime the content of several lipid classes and fatty acids has been found to be altered in the brain of AD-patients. A physiological function of Aβ and AICD in the regulation of several lipid metabolism pathways has been reported, possibly explaining the altered cerebral content of some lipid species in AD-affected brain tissue. Inversely, APP-processing is strongly influenced by the

surrounding lipid environment indicating a bidirectional link between APP-proteolysis and lipid metabolism (Grimm et al., 2012b; Mett et al., 2014).

The link between lipid homeostasis and AD-pathology is strengthened by the identification of the ApoEε4-allele as the most important genetic risk factor for LOAD. ApoE is a lipoprotein involved in the transport of cholesterol and other lipids in the central nervous system (CNS). In humans there are three different ApoEε alleles encoding for the isoforms ApoEε2, ApoEε3 and ApoEε4 (Weisgraber et al., 1981; Mahley et al., 1996; Holtzman et al., 2012). The ApoEε4-allele is associated with an increased AD-risk, earlier disease onset and enhanced cerebral plaque load (Corder et al., 1993; Kuusisto et al., 1994; Breitner et al., 1999; Tiraboschi et al., 2004). In contrast, ApoEε2 carriers have a reduced risk of developing AD (Corder et al., 1994). These associations might be explained by an isoformdependent binding of ApoEε (ε2 > ε3 > ε4) to Aβ-peptides influencing the clearance and aggregation of the peptide (Ma et al., 1994; Deane et al., 2008; Castellano et al., 2011; Holtzman et al., 2012).

A strong impact of the surrounding lipid bilayer on APP-processing is given by the fact that APP as well as all secretases are transmembrane proteins and that γ-secretase dependent APP cleavage even takes place in the hydrophobic membrane environment. For example, the exact position of γ-secretase cleavage and hence the length of the generated Aβ-peptides depends on membrane thickness (Grziwa et al., 2003; Winkler et al., 2012). In addition, the membrane fluidity influences APP-processing. Increased membrane fluidity seems to stimulate the non-amyloidogenic APP-processing by reducing APP internalization (Kojro et al., 2001). In this context it is important to note that APP-processing is also influenced by the subcompartmentalization of the membrane. Lipid raft microdomains are compact, dynamic assemblies of membrane proteins enriched in cholesterol, gangliosides and other sphingolipids. They are detergent-resistent and strongly differ in their lipid composition from the surrounding non-raft domains. Implications of lipid rafts in the intracellular protein trafficking, protein-lipid and protein-protein interactions as well as transmembrane signaling have been reported (Brown and Rose, 1992; Lingwood and Simons, 2010). The generation of Aβ has been shown to mainly take place in lipid rafts due to the co-localization of APP with BACE1 and the γ-secretase complex within these membrane microdomains (Lee et al., 1998; Riddell et al., 2001; Ehehalt et al., 2003; Vetrivel et al., 2004). In contrast, the non-amyloidogenic APP-proteolysis seems to occur predominantly in non-raft regions (Ehehalt et al., 2003; Harris et al., 2009). All these details indicate that a modulation of the membrane lipid composition might provide the opportunity of influencing Aβ-generation.

In the following sections of this article, the impact of several lipids and fatty acids on Aβ-associated AD-pathology is reviewed as well as the regulation of the corresponding metabolic pathways by APP-processing.

# THE IMPACT OF CHOLESTEROL ON AD

The brain is the most cholesterol-rich organ in the body (23 mg/g), it contains 23% of the total body sterol while only accounting for 2.1% of the total body weight (Dietschy and Turley, 2004). Within brain tissue cholesterol is mainly present in myelin sheaths and in the membranes of glial cells and neurons in its unesterified form. Due to the limited transport of cholesterol across the blood-brain barrier the cerebral cholesterol level is mainly dependent on de novo synthesis by oligodendrocytes, astrocytes and to a lesser extent by neurons. The conversion of 3-hydroxy-3-methylglutaryl-CoA to mevalonate catalyzed by the hydroxymethylglutaryl-CoA reductase (HMGCR), which is inhibited by statins, is the rate-controlling step in cholesterol biosynthesis (Martins et al., 2009; Di Paolo and Kim, 2011). The first evidence for a link between AD-pathogenesis and cholesterol metabolism was provided in 1994 by the observation that dietary cholesterol increases Aβ-production in rabbits (Sparks et al., 1994). Today there are many lines of evidence arguing for a connection between the pathology of AD and cholesterol homeostasis which are summarized below.

In several epidemiological studies elevated serum/plasma cholesterol contents have been identified as a risk factor for developing AD. Especially high serum cholesterol level in midlife are associated with a higher AD-risk (Pappolla et al., 2003; Solomon et al., 2009; Matsuzaki et al., 2011; Meng et al., 2014). Additionally, enhanced level of low-density lipoprotein (LDL) cholesterol and reduced level of high-density lipoprotein (HDL) cholesterol in serum correlate with the cerebral amyloid deposition in living human beings (Reed et al., 2014). In line, in human post mortem AD-brains cholesterol was found to be elevated and highly enriched in amyloid plaques (Cutler et al., 2004; Xiong et al., 2008; Panchal et al., 2010).

Most cell culture studies revealed that increasing cellular cholesterol level lead to an enhanced Aβ production whereas a depletion or reduction of cholesterol by e.g., cyclodextrin or statins shows the opposite effect (Simons et al., 1998; Fassbender et al., 2001; Maulik et al., 2013). The Aβ increasing property of cholesterol is based on a direct activation of β- and γ-secretase proteolytic activity (Kalvodova et al., 2005; Grimm et al., 2008; Osenkowski et al., 2008). Cholesterol is enriched in lipid raft membrane microdomains, in which amyloidogenic APP-processing mainly takes place. Thus modulating cellular cholesterol content inevitably affects membrane structure, membrane fluidity as well as cellular processes associated with lipid raft microdomains. Cholesterol depletion leads to the disruption of lipid rafts and therefore to a reduced association of APP, BACE1 and the components of the γ-secretase complex to lipid raft membrane microdomains, resulting in decreased amyloidogenic APP processing. Vice versa, an increase of cellular cholesterol leads to a higher lipid raft content of the membranes and hence to Aβ-overproduction (Simons et al., 1998; Hao et al., 2001; Hicks et al., 2012). High membrane cholesterol levels additionally promote APP endocytosis leading to enhanced Aβ-production in acidic intracellular compartments (Cossec et al., 2010). Conversely, APP is primarily localized at the cell surface in cholesterol-depleted cells leading to increased α-secretase-dependent non-amyloidogenic APP processing (Kojro et al., 2001). Beside the cholesterolmediated effects on APP-proteolytic processing, cholesterol has been shown to promote Aβ-aggregation and -toxicity (Schneider et al., 2006; Ferrera et al., 2008; Abramov et al., 2011).

A strong correlation between hypercholesterolemia and enhanced Aβ level has also been observed in several animal models (Sparks et al., 1994; Refolo et al., 2000; Maulik et al., 2013). Inversely, a reduction of accumulated Aβ-peptides along with improved behavioral memory was achieved in animal models after administration of cholesterol-lowering drugs including statins (Fassbender et al., 2001; Refolo et al., 2001; Kurata et al., 2012). It should be noted, that there are also a few studies in which statins had no or oppositional effects on the cerebral Aβ-content in vivo (Park et al., 2003; Cibickova et al., 2009).

The impact of statins on AD has also been analyzed in observational studies and randomized controlled trials leading to inhomogeneous results. Statin intake is associated with a reduced incidence of AD or dementia in general in most, but not all of these studies (Wolozin et al., 2000, 2007; Rea et al., 2005; Arvanitakis et al., 2008; Haag et al., 2009). Especially the reduction of serum cholesterol level by the intake of statins in midlife might have a preventive effect towards the development of AD (Kivipelto et al., 2002; Pappolla et al., 2003; Shinohara et al., 2014). In strong contrast, most clinical trials failed to observe any benefit of statins in individuals already suffering from AD (Feldman et al., 2010; Sano et al., 2011; McGuinness et al., 2014), indicating cholesterol-lowering drugs to have rather a protective than a therapeutic potential in respect to AD.

Beside the described influence of cholesterol on APPproteolysis, there is also an impact of APP-processing on cholesterol homeostasis. APP/APLP2- and PS1/PS2-deficient fibroblasts have a significantly increased cellular cholesterol content, which can be reversed by the supplementation of Aβ40-peptides. In line with this, enhanced cerebral cholesterol concentrations were found in APP- and PS-deficient mice (Grimm et al., 2005; Umeda et al., 2010). Analysis of the underlying mechanisms revealed that Aβ40 reduces cholesterol de novo synthesis by inhibiting HMGCR activity (Grimm et al., 2005).

#### Summary

The existence of a regulatory feedback cycle, in which Aβ-production is stimulated by cholesterol while cholesterol de novo synthesis is inhibited by high cellular Aβ40 concentrations is indicated.

#### Future Directions

The heterogeneous results of studies analyzing the impact of statins on the incidence of AD denote the existence of responders and non-responders. For the future it will be important to find biomarkes to identify patients that might profit from statins.

### THE IMPACT OF DOCOSAHEXAENOIC ACID (DHA) ON AD

Docosahexaenoic acid (DHA, 22:6) is a polyunsaturated fatty acid (PUFA) naturally occurring in high amounts in marine food, especially in fish oil (Mann et al., 2010). It accounts for 30%–40% of all esterified fatty acids in neuronal plasma membrane phospholipids and for 8% of the brain dry weight, thus belonging together with α-linolenic acid (ALA, 18:3) and eicosapentaenoic acid (EPA, 20:5) to the most important ω3-fatty acids in the CNS (Lauritzen et al., 2001; Muskiet et al., 2006). As endogenous DHA-biosynthesis is highly limited in humans, the main part of this fatty acid is provided by dietary intake (Pawlosky et al., 2001). DHA is efficiently transported across the blood brain barrier (Ouellet et al., 2009; Nguyen et al., 2014) and rapidly incorporates into phospholipids of cellular membranes leading to increased membrane fluidity (Horrocks and Farooqui, 2004; Yang et al., 2011).

The DHA content is reported to be reduced in the serum/plasma of AD-patients as well as in certain regions of post mortem AD-brains (Söderberg et al., 1991; Conquer et al., 2000; Tully et al., 2003). Because of its six double-bonds DHA is very susceptible to lipid-peroxidation resulting in oxidative stress known to be involved in AD pathogenesis (Smith et al., 1994; Yatin et al., 1998; Fam et al., 2002; Cai et al., 2011). Indeed, the levels of PUFA oxidation products are elevated in AD-affected brains, indicating the reduced DHA content in these tissues to be caused by increased oxidative damage (Sayre et al., 1997; Markesbery and Lovell, 1998; Montine and Morrow, 2005; Grimm et al., 2016a).

Several epidemiological trials found the dietary intake of DHA or higher DHA serum/plasma levels to be associated with a reduced risk of developing AD indicating a potential of DHA in AD-prevention (Kalmijn et al., 1997; Barberger-Gateau et al., 2002; Morris et al., 2003b). However, other studies failed to find an association between PUFAs and AD-risk (Engelhart et al., 2002; Kröger et al., 2009; Jicha and Markesbery, 2010; Mett et al., 2014).

We and others analyzed the impact of DHA on APP-processing revealing the fatty acid to reduce Aβ-levels via pleiotropic mechanisms. DHA reduces β- and γ-secretase activity and stimulates α-secretase-dependent APP-cleavage. In addition to direct effects, the activities of γ- and β-secretase are reduced by DHA due to a PS1-displacement out of lipid rafts and a reduced BACE1 internalization. The stimulated α-secretase activity in presence of DHA is based on the enhanced gene expression and protein stability of ADAM17. Altogether these effects lead to a shift from amyloidogenic to non-amyloidogenic APP-processing and thus to reduced total Aβ-level. DHA additionally has cholesterol-lowering effects further inhibiting Aβ-production. It reduces cholesterol de novo synthesis via inhibition of HMGCR and disturbs lipid raft integrity by shifting cholesterol out of these membrane microdomains (Hashimoto et al., 2005a; Stillwell et al., 2005; Grimm et al., 2011c). Beside the described effects on APP-processing an impact of DHA on Aβ-degradation and -aggregation has also been reported. We recently observed a highly enhanced insulin-degrading enzyme (IDE)-dependent Aβ-degradation in neuroblastoma cells after the supplementation of DHA- and EPA-containing phosphatidylcholine (PC; Grimm et al., 2016b). Others reported an increased microglial phagocytosis of Aβ as well as a reduction of Aβ-fibrillation and Aβ-induced toxicity in the presence of DHA (Hossain et al., 2009; Hjorth et al., 2013).

A protective effect of dietary DHA with regard to cerebral Aβ-level and amyloid plaque load could be further confirmed in vivo in several animal models (Lim et al., 2005; Green et al., 2007; Perez et al., 2010). In line with this, higher cognitive performances were observed in AD-animal models after DHA supplementation (Hashimoto et al., 2002, 2005b; Calon et al., 2004). However, others failed to find any beneficial effect of DHA in AD transgenic mice (Arendash et al., 2007).

A possible therapeutic use of DHA regarding AD has been investigated in several clinical trials showing inconsistent results. Some studies revealed a beneficial effect of daily DHA treatment in patients with very mild cognitive dysfunctions (Freund-Levi et al., 2006; Kotani et al., 2006; Chiu et al., 2008). Others did not observe any influence of DHA on AD-biomarkers and cognitive decline in AD patients (Freund-Levi et al., 2009; Quinn et al., 2010). It should be mentioned, that oxidized DHA species and the lipid-peroxidation products of PUFAs are able to increase amyloidogenic APP-processing and hence Aβ-generation. In a recent study we demonstrated, that only 1% oxidized DHA reverts the positives effects of DHA on Aβ-production indicating that PUFAs have to be prevented from oxidation in nutritional approaches (Grimm et al., 2016a). In such approaches DHA often is combined with E-vitamins due to their high antioxidative properties acting as scavengers of radicals and peroxides (Kamal-Eldin and Appelqvist, 1996). However, we demonstrated in two recent studies that several tocopherol and tocotrienol species have beside their protective antioxidative properties the undesirable effect of increasing amyloidogenic APP processing and reducing the enzymatic degradation of Aβ-peptides (Grimm et al., 2015b, 2016c).

#### Summary

Despite the inhomogeneous results of clinical studies there are several epidemiological and molecular indications for a beneficial effect of DHA in preventing AD and possibly halting its progression, at least at very early disease stages. The fact that its oxidation products are able to reverse the beneficial effects of DHA might partially explain the divergent outcomes of clinical DHA studies and underlines the need to prevent DHA from oxidation in such trials.

#### Future Directions

Because of the controversial effects of several E-vitamins regarding the molecular mechanisms of AD, the identification of further molecules for the prevention of DHA from oxidative damage in nutritional approaches without side effects on APP processing might be valuable. Additionally, the combination of DHA with precursors/cofactors for membrane synthesis and synaptogenesis as for example uridine-monophosphate, choline and phospholipids might further strengthen its beneficial effects on cognition as demonstrated in a transgenic mouse model of AD (Koivisto et al., 2014).

### THE IMPACT OF TRANS FATTY ACIDS ON AD

Trans fatty acids (TFAs) are unsaturated fatty acids, which are characterized by having at least one double-bond in transconfiguration. This means that the two hydrogen atoms are, in contrast to cis-configuration, localized on opposite sides of the double-bond. Because of their straighter shape compared to the cis-counterparts, TFAs have higher melting points and lead to a decreased fluidity of biological membranes (Roach et al., 2004; Ibrahim et al., 2005). TFAs in our diet arise from industrial procedures and to a lesser extent from biological processes in the digestive tract of ruminant animals. The key source of TFAs is commercially prepared food due to hydrogenation or thermal treatment of oils (Bhardwaj et al., 2011). Accumulation of these fatty acids in the body as well as incorporation in brain tissue has been reported indicating an impact of TFAs on cerebral biochemistry (Laryea et al., 1990; Teixeira et al., 2012).

Studies analyzing the relationship between TFAs and AD-risk or the progression of cognitive decline came to inconsistent results. A positive correlation between dietary TFA intake and AD-risk was found in one study while others reported the AD-risk not to be influenced by TFAs (Engelhart et al., 2002; Morris et al., 2003a). Similarly, some authors observed the TFA intake to result in a higher rate of cognitive decline in women with type 2 diabetes, in persons with high copper consumption and in older persons in general while others failed to find a relationship between TFA intake and cognitive decline in women (Morris et al., 2004, 2006; Devore et al., 2009; Naqvi et al., 2011; Okereke et al., 2012).

We investigated the effects of TFAs on APP-processing and Aβ-generation in neuroblastoma cells compared to their cis-counterparts. In presence of TFAs, we found a shift from non-amyloidogenic to amyloidogenic APP-processing accompanied by a significant increase in Aβ-production. TFA supplementation increases the activity of β- and γ-secretase due to direct effects and an enhanced gene expression of BACE1 and the γ-secretase complex components (Grimm et al., 2012a). The direct effect on γ-secretase activity was confirmed by others demonstrating the activity of purified γsecretase to be stimulated by an increased trans/cis-ratio of supplemented fatty acids (Holmes et al., 2012). In contrast, non-amyloidogenic APP-processing is reduced in TFA-treated cells because of enhanced APP-internalization and a reduction in ADAM10 gene expression. Additionally, we found TFAs to stimulate Aβ-aggregation in vitro (Grimm et al., 2012a).

The impact of TFAs on cerebral Aβ-levels and cognition has also been investigated in vivo with less clear results. In a study by Phivilay et al. (2009), Aβ- and tau-pathology was unaltered in the brain tissue of an AD-mouse model after dietary supplementation of TFAs. Another study reported a declined spatial learning performance of mice fed with a TFA- and monosodium glutamate-rich diet (Collison et al., 2010).

As TFAs are reported to be linked to cholesterol and DHA homeostasis they might also affect APP-processing and Aβ-generation via indirect mechanisms. The dietary intake of TFA leads to an inauspicious enhanced ratio of LDL/HDL plasma cholesterol (Mensink and Katan, 1990; Judd et al., 1994), which might be associated with a higher AD-risk as described above. Furthermore, high TFA consumption was shown to modify the fatty acid profile of murine brain tissue with a reduction in DHA content. Nevertheless, in this study the cerebral Aβ-levels were unaltered as already mentioned (Phivilay et al., 2009).

#### Summary

Due to the dissimilar results of studies analyzing the impact of TFAs on AD-risk and Aβ-associated pathology in vivo, further trials are necessary to clarify the role of these fatty acids in ADpathogenesis.

#### Future Directions

If the negative effects of TFA on AD-risk can be confirmed in vivo, a stronger reduction of TFA intake should be recommended, particularly because of the accumulation of these fatty acids in the human body over time and their incorporation into brain tissue (Laryea et al., 1990; Teixeira et al., 2012).

# THE IMPACT OF PLASMALOGENS ON AD

Plasmalogens (PL) are commonly occurring phospholipids accounting for 22% of the total phospholipid mass in human brain tissue. They are characterized by an enol ether double-bond at the sn1-position, which links an alkenyl chain to the glycerol backbone. At the sn2-position they are enriched in PUFAs including DHA and arachidonic acid (AA, 20:4). Phosphatidylethanolamine (PE) and PC are the most common polar head groups of PLs, which have a high susceptibility to oxidative stress due to their enol ether double-bond (Broniec et al., 2011; Braverman and Moser, 2012). PL-biosynthesis takes place in peroxisomes and the endoplasmic reticulum. The initial committed step reaction of PL de novo synthesis is catalyzed by the peroxisomal enzyme alkyl-dihydroxyacetonephosphatesynthase (AGPS; De Vet et al., 1999). PL level in the human body are mainly modulated by PL metabolism, but to a lesser extent also by the dietary consumption of PL-rich meat and fish (Blank et al., 1992).

While one study by Pettegrew et al. (2001) did not detect an AD-dependent alteration in the cerebral PL-content, we and others found a reduction of PE-PLs and PC-PLs in human post mortem AD-brains (Ginsberg et al., 1995; Han et al., 2001; Grimm et al., 2011a; Igarashi et al., 2011; Rothhaar et al., 2012). In line with this, a reduced PE-PL content was also observed in the serum and in erythrocyte membranes of AD-patients (Goodenowe et al., 2007; Oma et al., 2012).

The reduction of PL content in AD-affected brain tissue might be explained by enhanced PL degradation due to increased oxidative stress and a stimulated activity of phospholipases in presence of Aβ-peptides (Sanchez-Mejia et al., 2008). Additionally, we demonstrated PL biosynthesis to be regulated by APP-processing. Under physiological conditions AGPS gene expression and hence PL biosynthesis is upregulated by AICD. In contrast, under pathological conditions the Aβ-induced reactive oxidative species impair AGPS protein stability leading to a decreased PL de novo synthesis (Grimm et al., 2011d).

Because of the altered PL content in AD-brain tissue, we analyzed the impact of PLs on APP-processing. Our results demonstrate that PLs reduce γ-secretase activity in living cells as well as in purified membranes derived from neuroblastoma cells and murine brain tissue. Compared to the corresponding phospholipids lacking the enol ether, all tested PC-PL- and PE-PL-species independent of the bound fatty acid directly inhibited γ-secretase activity. In contrast, the activities of α- and β-secretase remained unchanged after PL-supplementation (Rothhaar et al., 2012). The direct inhibitory effect of PE-PLs on the γ-secretase complex has been recently confirmed by others (Onodera et al., 2015). Interestingly, in our study the addition of PLs to cellular membranes derived from human AD-brains also resulted in a decreased γ-secretase activity. This indicates the rebuilding of a normal PL level to have a positive impact in the pathologic situation of AD (Rothhaar et al., 2012). However, such ex vivo experiments have their clear limitations and further studies are necessary to analyze the in vivo relevance of PLs on APP-processing. In addition to Aβ-production, there is also an impact of PLs on the aggregation of Aβ-peptides. PE-PL has been reported to eliminate the neurotoxicity-associated Aβ-oligomerization phase while allowing fibril formation (Lee et al., 2011).

#### Summary

In the pathologic situation of AD a vicious cycle between PLs and Aβ-generation can be postulated: the accumulation of Aβ results in a reduced cerebral PL content stimulating γ-secretase activity and hence leading to a further increased Aβ-production.

#### Future Directions

In the future the in vivo-relevance of the effects of PLs on the generation of Aβ-peptides should be analyzed. An interesting human model for such trials could be cells derived from patients affected by Zellweger syndrome, which show deficient PL-levels due to a defective peroxisome assembly (Styger et al., 2002; Saitoh et al., 2009).

# THE IMPACT OF SPHINGOLIPIDS ON AD

Sphingolipids are an inhomogeneous group of lipids characterized by a backbone consisting of the amino alcohol sphingosine. Sphingolipid biosynthesis is initiated by the serine palmitoyl-CoA transferase (SPT) catalyzing the condensation of palmitoyl-CoA and L-serine to 3-ketosphinganin, which is further metabolized to ceramide. Ceramide is the most important branching point within the sphingolipid metabolism pathways serving as precursor for the generation of sphingosine, sphingomyelin (SM) and more complex glycosphingolipids.

All sphingolipids are anchored in the membrane bilayer via their ceramide moiety, besides cholesterol they represent major components of lipid raft membrane microdomains (Posse de Chaves and Sipione, 2010). The first evidence for a role of sphingolipids in neurodegeneration came from the observation of lysosomal storage diseases, inherited disorders characterized by the lysosomal accumulation of different sphingolipids. These diseases are associated with early dementia and the development of AD-related Aβ- and tau-pathology (Tarasiuk et al., 2012). The link between sphingolipid metabolism and AD-pathogenesis is further strengthened by alterations of several sphingolipids in post mortem AD-brain tissue and their potential to modulate APP-processing and Aβ-aggregation summarized below. Additionally, SPT gene expression and hence total sphingolipid biosynthesis is downregulated by the APP-processing product AICD (Grimm et al., 2011b).

#### Ceramide

As already mentioned, ceramide is generated by de novo synthesis and by hydrolysis of various more complex sphingolipids. Ceramides are pro-apoptotic and neurotoxic signaling molecules, additionally participating in the regulation of cellular proliferation and differentiation (Dawson et al., 1998; Toman et al., 2002).

The ceramide level has been reported to be increased in different brain regions and in the cerebrospinal fluid (CSF) of AD-patients. As the increase in ceramide content is already present at the earliest clinical stages of AD, it might be speculated that it is involved in disease development (Han et al., 2002; Satoi et al., 2005; Katsel et al., 2007; He et al., 2010; Filippov et al., 2012). Such a relationship is supported by a 9-year-follow-up study reporting an association between elevated baseline serum ceramide levels and an enhanced risk for developing AD (Mielke et al., 2012).

As reported by Katsel et al. (2007) the accumulation of ceramide in AD-affected individuals might be explained by multiple gene expression abnormalities. The authors found an increased cerebral expression of genes involved in ceramide de novo synthesis along with a reduced expression of genes required for glycosphingolipid formation out of ceramide. Another explanation for the increased ceramide content in AD-brain tissue is the Aβ-mediated activation of sphingomyelinases (SMases) catalyzing the brake down of SM to ceramide. We and others found Aβ-peptides to directly stimulate neutral SMase (nSMase)-activity (Jana and Pahan, 2004; Lee et al., 2004; Grimm et al., 2005), a stimulation of acidic SMase

Frontiers in Molecular Neuroscience | www.frontiersin.org March 2017 | Volume 10 | Article 63

(aSMase) by Aβ has also been observed (Malaplate-Armand et al., 2006). The resulting enhanced ceramide level is reported to be a mediator of Aβ-induced apoptosis. Besides a probable involvement in Aβ-induced cell death, ceramide also affects APP-cleavage. Accumulation of endogenous ceramide levels in cultured cells by the use of cell-permeable C6-ceramide or by nSMase treatment promotes amyloidogenic APP-processing. The resulting ceramide-induced enhanced Aβ biogenesis is caused by a post-translational stabilization of the β-secretase BACE1 due to elevated acetylation of the protein (Puglielli et al., 2003; Ko and Puglielli, 2009).

In their entirety these facts indicate the existence of a feed-forward cycle between ceramide and Aβ under the pathological conditions in AD-brain tissue: enhanced ceramide level lead to an increased Aβ-production resulting in the activation of SMases and hence in a further elevation of ceramide content, which stimulates Aβ-production and might be involved in the induction of apoptotic cell death.

# Sphingomyelin

SM accounts for approximately 10% of mammalian cellular lipids and is highly enriched in myelin sheets. It is produced out of ceramide by the activity of SM-synthases, SMases catalyze the catabolic break down of SM back to ceramide.

The already mentioned increased ceramide content and the upregulation of SMases in post mortem AD brains (Katsel et al., 2007; He et al., 2010) suggests that SM concentrations might be reduced in these tissues. However, the results of studies analyzing the SM content in AD-affected brains are inhomogenous (Pettegrew et al., 2001; Cutler et al., 2004; Bandaru et al., 2009; He et al., 2010). In addition, SM level were found to be significantly increased in the CSF of individuals with prodromal AD while there was a slight, but not significant reduction of SM in the CSF of patients with mild and moderate AD (Kosicek et al., 2012). In an epidemiological study by Mielke et al. (2011) higher SM concentrations and an enhanced SM/ceramide-ratio in plasma was found to correlate with a decelerated disease progression among AD-patients.

In strong contrast to ceramide, SM was demonstrated to inhibit Aβ-production. Increasing SM content of cultured cells either by direct exposure or nSMase inhibition leads to a significant decrease of Aβ-peptides caused by an inhibition of γ-secretase dependent APP-processing. In this study we additionally identified the already mentioned direct stimulation of nSMase by Aβ42 (Grimm et al., 2005).

Accordingly, the Aβ-induced elevation of SMase-activity in AD-brain tissue results in an enhanced ceramide/SM-ratio. The increase in γ-secretase activity due to lowered SM-level in combination with the ceramide-dependent activation of β-secretase further promotes Aβ-production might result in a futile cycle.

#### Sphingosine and Sphingosine 1-Phosphate

Ceramidases catalyze the conversion of ceramide to sphingosine, which is phosphorylated by sphingosine kinase (SK) generating the anti-apoptotic and neuroprotective molecule sphingosine 1-phosphate (S1P). S1P has been demonstrated to induce cell survival and proliferation and to antagonize Aβ- and ceramideinduced cell death (Cuvillier et al., 1996; Gomez-Brouchet et al., 2007; Czubowicz and Strosznajder, 2014). In contrast, sphingosine seems to have a role in apoptosis, cooperatively or independently from ceramide signaling (Sweeney et al., 1998; Lepine et al., 2004).

In line with an increased acid ceramidase expression and activity, the sphingosine content has been found to be elevated in post mortem AD-brains (Huang et al., 2004; He et al., 2010). It should be mentioned, that there is also another study reporting a decreased acid ceramidase gene expression in AD-brain tissue (Katsel et al., 2007). In contrast, the cerebral S1P-content seems to be declined in AD-affected individuals and to negatively correlate with the level of Aβ and phosphorylated tau protein (He et al., 2010). In line with these observations, γ-secretase activity is reduced in cells devoid of S1P-lyase degrading intracellular S1P (Karaca et al., 2014). Contrariwise, S1P has been shown to increase the production of Aβ-peptides by directly stimulating β-secretase activity in another study (Takasugi et al., 2011). Therefore, further studies are necessary to clarify the role of sphingosine and S1P in APP-processing and AD-pathogenesis.

#### Sulfatides

Sulfatides are complex glycosphingolipids generated from ceramide by the addition of a galactose moiety and a sulfate group catalyzed by ceramide galactosyltransferase (CGT) and cerebrosidesulfotransferase (CST), respectively. They are highly enriched in myelin sheaths and mainly synthesized by oligodendrocytes.

Several studies reported the cerebral sulfatide content to be dramatically decreased in AD-patients compared to cognitive normal controls. These alterations were already observed in the earliest recognizable states of the disease (Han et al., 2002; Bandaru et al., 2009; Cheng et al., 2013). However, there are two other studies which failed to find a significant alteration in sulfatide content in AD-brain tissue (Cutler et al., 2004; Chan et al., 2012). CSF sulfatide level are also strongly reduced in AD-patients as reported by Han et al. (2003b) who suggested the sulfatide/phosphatidylinositol ratio in the CSF to be a potential AD-biomarker.

Interestingly, there seems to be a link between sulfatide homeostasis and ApoE: sulfatides are associated with ApoE-containing particles in the CSF and ApoE is involved in the modulation of cellular sulfatide content in an isoformdependent manner. This possibly provides an explanation for the genetic association between ApoE and AD (Han, 2010). A role of ApoE in the regulation of cerebral sulfatide level has been demonstrated by Cheng et al. (2010). In this study the age-dependent decline in cortical sulfatide concentrations of APP transgenic mice was found to be totally abolished in ApoE-knockout animals. The sulfatide content in murine brain tissue was further demonstrated to be dependent on ApoE-genotype. In comparison to human ApoEε3 and wildtype ApoEε, the human ApoEε4-isoform is associated with a strong sulfatide depletion in the brain of transgenic mice (Han et al., 2003a). Additionally, sulfatides seem to be involved in ApoE-dependent Aβ-clearance. Treatment of cultured cells with sulfatides results in a strong reduction of Aβ-peptides in the culture media. The underlying mechanism was identified as a facilitated ApoE-mediated Aβ-clearance through an endocytotic pathway in response to elevated sulfatide levels (Zeng and Han, 2008).

Their robust depletion in post mortem AD-brain tissue and their potential to strongly reduce Aβ-levels in vitro indicate that sulfatides might be an attractive target in AD research. Further studies are necessary to investigate the role of this lipid class in the molecular mechanisms of the disease.

#### Gangliosides

Gangliosides, sialic acid containing glycosphingolipids, represent 6% of the total lipid content in brain. They are abundant in the luminal leaflet of cellular organelles and the outer leaflet of the plasma membrane, where they are localized in lipid raft microdomains. Important functions of gangliosides in the development, proliferation and differentiation of neuronal cells have been reported. The glycosylceramide synthase (GCS) catalyzes the first step of ganglioside biosynthesis by glycosylating ceramide. Dependent of the number of sialic acid residues gangliosides are classified into four catagories, the o-, a-, band c-series. In brain tissue the most common gangliosides are GM1, GD1a, GD1b and GT1b belonging to the a- and b-series. GM3 is the precursor of all a- and b-series gangliosides, which are segregated by the GD3-synthase (GD3S)-catalyzed addition of sialic acid to GM3 (Busam and Decker, 1986; Lahiri and Futerman, 2007; Yu et al., 2011).

In AD-brain tissue there is a reduction of total ganglioside content along with significant regional differences in the distribution of specific ganglioside species. In brains affected by FAD and LOAD the total ganglioside level is decreased in several brain regions (Kalanj et al., 1991; Svennerholm and Gottfries, 1994; Gottfries et al., 1996). Kracun et al. (1991) reported a reduction of all major brain gangliosides combined with an increase in the more simple GM2 and GM3 in the cortex of AD-patients. In line with this, the GM1 and GM2 level were found to be elevated in the lipid raft fraction derived from cortical regions of AD brains (Molander-Melin et al., 2005). In summary, in AD-affected brains complex gangliosides tend to decrease while there is an elevation of simple ganglioside species.

Interestingly, in post mortem AD-brains GM1 and GD1a have been found to be associated with Aβ-plaques forming GAβcomplexes exhibiting early pathological changes of AD. This indicates a role of these ganglioside species in Aβ-aggregation (Nishinaka et al., 1993; Yanagisawa et al., 1995). Indeed, GM1 induces a conformational transition of Aβ from random coil to β-sheet structure and triggers the formation of toxic Aβ-fibrils (Choo-Smith et al., 1997; Hayashi et al., 2004; Okada et al., 2007). Further studies demonstrated an accumulation and aggregation of Aβ in GM1-enriched lipid rafts leading to an increased cytotoxicity (Wakabayashi et al., 2005).

Besides Aβ-aggregation, APP-processing and hence Aβ-generation is also influenced by GM1 and other gangliosides. Direct administration of total ganglioside extract to purified γ-secretase leads to an enhanced enzyme activity and increases the ratio of generated Aβ42 to Aβ40 peptides (Holmes et al., 2012). In line, the inhibition of GCS and hence total ganglioside biosynthesis results in a significant reduction of Aβ-production in various cell lines. The addition of exogenous brain gangliosides reverses these effects indicating the reduction of total ganglioside biosynthesis to be beneficial in AD. In this study, the authors found glycosphingolipids to affect APP-processing via regulating the subcellular APP-transport in the secretory pathway (Tamboli et al., 2005). In our own study we demonstrated GCS gene expression to be regulated by PS and APP. Deficiency in these proteins or the inhibition of γ-secretase activity results in an increased GCS gene expression and hence in increased glycosylceramide and total ganglioside level in vitro and in vivo. We showed that GCS is upregulated in the brain tissue of an AD-mouse model and of patients suffering from LOAD. Accordingly, total ganglioside de novo synthesis is modulated by APP-processing and deregulated in the pathological situation of AD (Grimm et al., 2014).

The treatment of neuroblastoma cells with GM1 has been shown to stimulate Aβ-generation and to reduce the sAPPα level without affecting sAPPβ (Zha et al., 2004). In strong contrast to this, peripheral injections of GM1 reduce the cerebral Aβ-burden in an AD-mouse model, possibly due to the promotion of Aβ-degradation in the periphery (Matsuoka et al., 2003). In another study the impact of GD3S deficiency, which results in a loss of b-series gangliosides and an accumulation of GM3, GM1 and GD1a, on the cerebral Aβ-levels in an AD-mouse model has been analyzed. Compared to the control animals, the GD3S-depleted mice showed an almost completely eliminated Aβ-associated neuropathology and no cognitive decline (Bernardo et al., 2009). In line with this, we found the generation of Aβ in cultured cells to be reduced after GM3 supplementation while the addition of the GD3S-product GD3 stimulated Aβ-release. In this context it is important to mention that we also found a regulation of GD3S by APPprocessing. The activity of GD3S is inhibited by a direct interaction of Aβ with GM3 leading to a reduced substrate availability and hence to an impaired conversion of GM3 to GD3. Additionally, the gene expression of GD3S is downregulated by AICD. These results indicate the existence of a regulatory feedback cycle, in which Aβ and AICD increase the GM3/GD3 ratio leading to a reduction of amyloidogenic APP-processing (Grimm et al., 2012c).

All these data indicate a strong link between ganglioside homeostasis and AD. As the single ganglioside species differ in their amyloidogenic potential, further studies are necessary to identify the most promising molecular target in ganglioside metabolism for developing therapeutic approaches regarding AD.

#### Summary

In post mortem AD-brain tissue there are alterations in the content of several sphingolipid species, which can be partially explained by an impact of Aβ and AICD on enzymes involved in sphingolipid homeostasis. Several sphingolipid classes have been shown to affect the proteolytic processing of APP and Aβ-clearance: ceramides, total gangliosides, GM1 and GD3 are associated with an increased Aβ-level while SM, sulfatides and GM3 have the opposite effect.

#### Future Directions

The fact that ceramide is associated with an increased amyloidogenic APP processing while an increase in SM-levels results in a decreased Aβ-generation indicates SMases to be interesting pharmacological targets regarding AD. Hence, the impact of SMase-inhibitors as for example fluoxetine, maprotiline or desipramine (Kölzer et al., 2004; Kornhuber et al., 2008) on the proteolytic processing of APP and on cognitive functions should be analyzed in suitable models. Another molecular target might be the GD3S, whose inhibition results in an enhanced GM3/GD3-ratio leading to a reduction in amyloidogenic APP proteolysis. In this context it should be mentioned, that mice lacking the GD3 synthase gene show abnormalities in the sciatic nerve and in peripheral nerve regeneration along with impaired neurogenesis and behavioral deficits (Ribeiro-Resende et al., 2014; Wang et al., 2014). This phenotype indicates that pharmacological interventions in ganglioside homeostasis might be associated with severe side effects.

# LIPIDS AS POTENTIAL BIOMARKERS FOR AD

Regarding therapeutic interventions for AD an early diagnosis of the disease and hence the identification of biomarkers, which can be used for the in vivo diagnosis prior to the first symptoms, is important. So, the identification of early AD-biomarkers with a high specificity and reliability is a central topic in AD research (Fiandaca et al., 2014). The lipid alterations connected to AD, which are partially detectable at the very early disease stages

as described above, might have the potential to be used as biomarkers for early AD diagnosis by lipidomic approaches. For example, Mapstone et al. (2014) discovered a set of eight PC species and two acylcarnitines in the peripheral blood that predicts the development of mild cognitive impairment or AD within 2–3 years with an accuracy of more than 90%. However, further studies are needed to identify combinations of lipidomics-based biomarkers which can be used for the detection of preclinical AD with the required sensitivity and specificity.

#### CONCLUSION

In conclusion all these findings demonstrate a close link of APP, APP processing and AD to lipid homeostasis. It could be demonstrated that APP processing and especially AICD has a physiological function in in the regulation of several lipid metabolic pathways. Inversely, APP-processing is strongly dependent on the lipid microenvironment indicating a

#### REFERENCES


bidirectional link between APP-proteolysis and lipid metabolism. This results in tightly connected complex regulatory cycles (**Figure 2**). Under pathological situations such as AD, this balanced regulation might be disrupted leading to pathological alterations in lipid homeostasis and Aβ peptide overproduction, resulting in increased neurodegeneration.

#### AUTHOR CONTRIBUTIONS

MOWG, JM, HSG and TH wrote the manuscript.

#### FUNDING

According to the author guidelines, funding for the research leading to these results were received from: the EU FP7 project LipiDiDiet, Grant Agreement No. 211696. Moreover funding for MOWG and TH was provided by Fundació la Maratò de TV3 and by JPND MindAD 1ED1508.

AD increases, then declines: the cache county study. Neurology 53, 321–331. doi: 10.1212/WNL.53.2.321


development and in the mature animal. J. Lipid Res. 45, 1375–1397. doi: 10.1194/jlr.R400004-JLR200


amyloidogenic β- and γ-secretase cleavage regulates neprilysin expression. Front. Aging Neurosci. 7:77. doi: 10.3389/fnagi.2015.00077


amyloid precursor protein. Brain Res. 1296, 203–215. doi: 10.1016/j.brainres. 2009.07.105


pathology in a transgenic mouse model of Alzheimer's disease. Neurobiol. Dis. 8, 890–899. doi: 10.1006/nbdi.2001.0422


dementia three decades later. Dement. Geriatr. Cogn. Disord. 28, 75–80. doi: 10.1159/000231980


by the transmembrane aspartic protease BACE. Science 286, 735–741. doi: 10.1126/science.286.5440.735


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Grimm, Mett, Grimm and Hartmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Analysis of Amyloid Precursor Protein Function in Drosophila melanogaster

#### Marlène Cassar and Doris Kretzschmar\*

Oregon Institute of Occupational Health Sciences, Oregon Health and Science University, Portland, OR, USA

The Amyloid precursor protein (APP) has mainly been investigated in connection with its role in Alzheimer's Disease (AD) due to its cleavage resulting in the production of the Aβ peptides that accumulate in the plaques characteristic for this disease. However, APP is an evolutionary conserved protein that is not only found in humans but also in many other species, including Drosophila, suggesting an important physiological function. Besides Aβ, several other fragments are produced by the cleavage of APP; large secreted fragments derived from the N-terminus and a small intracellular C-terminal fragment. Although these fragments have received much less attention than Aβ, a picture about their function is finally emerging. In contrast to mammals, which express three APP family members, Drosophila expresses only one APP protein called APP-like or APPL. Therefore APPL functions can be studied in flies without the complication that other APP family members may have redundant functions. Flies lacking APPL are viable but show defects in neuronal outgrowth in the central and peripheral nervous system (PNS) in addition to synaptic changes. Furthermore, APPL has been connected with axonal transport functions. In the adult nervous system, APPL, and more specifically its secreted fragments, can protect neurons from degeneration. APPL cleavage also prevents glial death. Lastly, APPL was found to be involved in behavioral deficits and in regulating sleep/activity patterns. This review, will describe the role of APPL in neuronal development and maintenance and briefly touch on its emerging function in circadian rhythms while an accompanying review will focus on its role in learning and memory formation.

#### Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Peter Soba, University of Hamburg, Germany Valerie Goguel, Centre National de la Recherche Scientifique, France

#### \*Correspondence:

Doris Kretzschmar kretzsch@ohsu.edu

Received: 03 June 2016 Accepted: 13 July 2016 Published: 26 July 2016

#### Citation:

Cassar M and Kretzschmar D (2016) Analysis of Amyloid Precursor Protein Function in Drosophila melanogaster. Front. Mol. Neurosci. 9:61. doi: 10.3389/fnmol.2016.00061

#### Keywords: Drosophila melanogaster, amyloid precursor proteins, neuronal outgrowth, neuronal survival, synaptogenesis

The Amyloid precursor protein (APP) is a key factor in Alzheimer's Disease (AD) because, as the name implies, it is the precursor from which the neurotoxic Aβ peptides are generated (Glenner and Wong, 1984; Masters et al., 1985). APP is a type-one membrane-spanning protein consisting of a large extracellular N-terminal domain and a small intracellular C-terminal domain in addition to the Aβ region (Goldgaber et al., 1987; Kang et al., 1987; Robakis et al., 1987; Tanzi et al., 1987). Alternative splicing of the APP gene produces three major isoforms (695aa, 751aa, and 770aa), with APP<sup>695</sup> being the major form found in the nervous system (Tanaka et al., 1989; Lorent et al., 1995). In addition to APP, vertebrates express two closely related proteins called Amyloid Precursor-Like Proteins (APLP) 1 and 2 (Coulson et al., 2000; Turner et al., 2003).

Over the last decade, transgenic Drosophila expressing either human APP<sup>695</sup> or Aβ have been extensively used to study the pathogenic function of APP (Cowan et al., 2010; (Iijima-Ando and Iijima, 2010; Moloney et al., 2010; Wentzell and Kretzschmar, 2010; Prüßing et al., 2013; Bouleau and Tricoire, 2015). However, insects also express an ortholog of APP which was named APP-like or APPL. APPL is about 30% overall identical to human APP<sup>695</sup> but a much higher degree of conservation is found in the extracellular E1 and E2 domains and especially in the C-terminal intracellular domain or AICD (Rosen et al., 1989; Swanson et al., 2005; **Figure 1**). Five isoforms of APPL are described in Drosophila that range from 830aa to 890aa (Attrill et al., 2016), however it is unknown whether these isoforms are functionally different. In contrast to the human protein, which is also expressed in non-neuronal cells (Sandbrink et al., 1994a,b), APPL is only expressed in neurons, starting at stage 13 of Drosophila embryogenesis (Luo et al., 1990; Martin-Morris and White, 1990). Interestingly, APPL lacks the Kunitz-like domain and is therefore more closely related to APP<sup>695</sup> than other isoforms (Arai et al., 1991). Like APP, APPL is processed by several secretases, resulting in secreted fragments, a neurotoxic Aβ-like peptide, and an intracellular AICD (Luo et al., 1990; Carmine-Simmen et al., 2009; Bolkan et al., 2012). However, in comparison to APP, the cleavage sites for the α- and β-secretase are reversed in APPL, with the β-site being more proximal to the transmembrane region and the α-site being more distal (Carmine-Simmen et al., 2009; Stempfle et al., 2010). The evolutionary conservation of APPL and its processing not only suggests that this protein has important physiological functions but also that studies in Drosophila can provide insights into the normal functions of human APP and its proteolytic fragments.

#### APPL AND THE DEVELOPMENT OF THE PERIPHERAL NERVOUS SYSTEM

Flies that completely lack APPL (Appl<sup>d</sup> , Luo et al., 1992) are viable but show a loss of sensory bristles on the sternopleuron and scutellum, parts of the adult thorax (Merdes et al., 2004). The same phenotype was observed when knocking down the Appl mRNA during development via RNA-interference. These mechano-sensory organs (MSOs) are derived from a sensory organ precursor cell (SOP), which is determined by lateral inhibition via Notch signaling. They consist of a shaft, a socket, a sheath cell, the sensory neuron, and a supporting glial cell (Lai and Orgogozo, 2004). Because not only the sensory neuron is missing in Appl<sup>d</sup> flies, but also the external cell types of the MSO, this indicates that APPL plays a role in SOP linage formation (Merdes et al., 2004). This result implies that in the peripheral nervous system (PNS) APPL is expressed in neuronal precursor cells, possibly playing a role in the determination of the MSOs, whereas in the central nervous system (CNS) it is restricted to differentiated neurons (Luo et al., 1990).

In addition, APPL is required for the correct development of the enteric nervous system (ENS) in insects, more specifically in the migration of enteric neurons. During embryonic development of Manduca sexta, the neurons in the enteric plexus (EP cells) align with the muscle bands on the midgut and foregut

and subsequently migrate along these pathways (Copenhaver and Taghert, 1989). APPL expression is detectable in the EP cells starting shortly before the onset of migration (Swanson et al., 2005) and knocking down APPL caused the EP neurons to ectopically migrate onto the interband regions (Ramaker et al., 2013). In Drosophila, the enteric neurons do not migrate along the gut and therefore this function of APPL does not play a role in flies. If APPL can act as a neuronal guidance receptor in cell migration of other neurons in flies remains to be determined.

# APPL AND NEURONAL OUTGROWTH

The expression of APPL during embryonic development correlates with the onset of axonal outgrowth and it is especially abundant in growing axons and in areas of synapse formation (Luo et al., 1990; Martin-Morris and White, 1990). Initially, no gross abnormalities were described in the larval or adult CNS of Appl<sup>d</sup> flies. However, they showed behavioral deficits in the fast-phototaxis assay, which is based on visual input and startle-induced locomotion (Luo et al., 1992). Later studies revealed that the loss of APPL does have effects on neuronal outgrowth, although the phenotypes are more subtle. Using cultures derived from embryonic neuroblasts, Li et al. (2004) found that the loss of APPL did not affect the initial outgrowth but resulted in significantly shorter neurites when cultured for 6 days. Surprisingly, overexpression of APPL or a secreted N-terminal fragment reduced neurite length whereas expression of a secretion-deficient form (APPLsd) or a variant that in addition lacks the intracellular C-terminus (APPLdelCT) increased neurite length. Thus, in cell culture secreted APPL seems to function as a growth limiting ligand for a yet unknown receptor, whereas full-length APPL may act as a receptor that promotes neurite growth. Focusing on specific cell types, changes in axonal outgrowth and arborization were also observed in vivo. Induction of APPL in the lateral neurons, a group of neurons that play a key role in the regulation of circadian rhythms, promoted axonal arborization, as did expression of human APP (Leyssen et al., 2005). Interestingly, in these experiments the C-terminus appeared to be required for the axonal outgrowth. Deleting the C-terminus of APP or only the YENPTY motif, which mediates the interaction with various proteins like X11α or Fe65 (Turner et al., 2003; Poeck et al., 2012), prevented these phenotypes.

Similarly, affecting the levels of APPL in the mushroom bodies caused changes in its morphology. The mushroom bodies are considered to be the center for learning and memory in flies. They consist of the calyx, which contains the dendrites and is localized in the dorsal-posterior part of the brain, and the peduncle, which is formed by the axons which project as a bundle from dorsocaudal to rostroventral (Heisenberg, 2003). These axons then separate and form five lobes with the α/α 0 lobes projecting dorsally whereas the β/β 0 and γ-lobes are horizontally orientated towards the midline of the brain. APPL is prominently expressed in the mushroom bodies, especially in the neurons that form the α and β lobes (Soldano et al., 2013). A function of APPL in these neurons was first suggested by Li et al. (2004) who showed that expressing additional APPL in the mushroom bodies resulted in a fuzzy appearance of the β-lobes, though only detectable in some flies. The authors suggested that this could be probably due to a loosened fasciculation of these axons. A more prominent phenotype was observed more recently by Soldano et al. (2013) analyzing Appl<sup>d</sup> flies. Although still not fully penetrant, 14% of these flies showed a complete loss of an α-lobe and 12% a loss of a β-lobe (Soldano et al., 2013). Interestingly, it turned out that APPL function is cell-autonomously required for the development of the β-lobe whereas its function in the α-lobe is non-autonomous. Rescue experiments showed that the C-terminus was required for the axonal outgrowth of the β-lobe (Soldano et al., 2013), as was suggested for the axonal growth of lateral neurons (Leyssen et al., 2005). In both cell types the function was mediated by the Abelson kinase, which binds to the C-terminus of APPL via the adapter protein disabled (Leyssen et al., 2005; Soldano et al., 2013). Result from the studies in mushroom body neurons suggested that this then regulates the activity of the Planar Cell Polarity signaling pathway (Soldano et al., 2013), a pathway that has been shown to regulate neuronal outgrowth in flies and vertebrates (Lyuksyutova et al., 2003; Ng, 2012). Notably, whereas these in vivo experiments show a requirement of the C-terminus, suggesting that APPL acts as a receptor in axonal outgrowth, the cell culture experiments indicated that the C-terminus is not needed to promote outgrowth. This might be due to the special conditions in culture or alternatively different neuronal subtypes use different fragments and signaling pathways for proper outgrowth.

It has also been shown that the loss of APPL affects the outgrowth of photoreceptors. APPL is expressed in all photoreceptors but a more prominent expression can be detected in the R7 and R8 subtype, whereby the expression depends on Ras signaling (Mora et al., 2013). R7 and R8 project into the medulla, the second optic neuropil in Drosophila, where they target different layers (Meinertzhagen and Hanson, 1993). Focusing on R7, Mora et al. (2013) found that 2% of the R7 cells do not reach their target field. Although this is a relatively mild phenotype, it nevertheless has physiological consequences because Appl<sup>d</sup> flies exhibited a reduced preference for UV light, which is detected by this photoreceptor subtype. Using a knock-down strategy for APPL, another group observed changes in the symmetrical arrangement of the photoreceptors in the adult eye combined with an occasional loss of R7 photoreceptors (Singh and Mlodzik, 2012). The authors also show that these phenotypes were enhanced by a knock down of hibris (hbs), which is a family member of the immunoglobulin cell adhesion proteins (Johnson et al., 2012). HBS seems to exert its function by affecting the γ-processing of APPL because it can promote the cleavage of Presenilin (PSN) into its active form (Singh and Mlodzik, 2012). As in vertebrates, the fly γ-secretase consists of NCT, APH1, PEN2, and the catalytically active PSN (Hu and Fortini, 2003; Stempfle et al., 2010) and expression of Drosophila PSN was shown to promote APPL cleavage (Carmine-Simmen et al., 2009). The interaction of HBS with APPL therefore suggests that its function in photoreceptor development and outgrowth requires the C-terminus or more specifically C-terminal cleavage of APPL.

Together, these experiments show that APPL does have a function in neuronal development and outgrowth, most likely acting as a receptor for a so far unknown ligand. However, its loss neither prevents axonal growth nor are the phenotypes fully penetrant. This indicates that APPL acts more like a ''robustness'' factor that supports the correct outgrowth instead of initiating or allowing it.

#### APPL FUNCTION IN SYNAPTOGENESIS AND AXONAL TRANSPORT

In addition to affecting axonal growth, APPL has also been shown to interfere with synapse formation. During larval development, different types of synaptic boutons are added along the axonal terminus, forming the stereotyped pattern of neuromuscular junctions (NMJ) at the body wall muscles (Gramates and Budnik, 1999). Appl<sup>d</sup> mutant larvae revealed a significant reduction in bouton numbers whereas overexpression of APPL induced additional boutons of different sizes; large ''parent'' boutons and small ''satellite'' boutons that are connected to the parent boutons (Torroja et al., 1999). The C-terminus was required to induce this phenotype and interestingly a deletion of the YENPTY domain prevented the formation of satellite boutons. In contrast, a deletion of the G<sup>0</sup> binding site (**Figure 1**) prevented the induction of additional parent boutons. These experiment suggest that APPL also acts as a receptor at the NMJ. Additional experiments showed that to fulfil its function at the NMJ, APPL interacts with the cell adhesion molecule Fasciclin II (Fas II; an neural cell adhesion molecule (NCAM) homolog; García-Alonso et al., 1995) and the PDZ-domain containing dX11/Mint protein (Hase et al., 2002; Ashley et al., 2005). Because dX11/Mint binds to the YENPTY domain, this would explain the requirement of the C-terminus of APPL for bouton formation (Ashley et al., 2005). dX11/Mint binding seems to regulate the localization of APPL because a loss of dX11/Mint or expression of a dX11/Mint construct with a deletion in the APPL binding site resulted in an increase in the levels of APPL at the boutons (Ashley et al., 2005). A role of dX11/Mint in regulating APPL localization was confirmed in mushroom body neurons where the loss of dX11/Mint caused a depletion of APPL from the axons in the peduncle and the lobes while mis-localizing it to the calyx, which contains the dendrites from which it is normally excluded (Gross et al., 2013).

As with photoreceptors, the defects in the formation of the NMJ may not be very dramatic in Appl<sup>d</sup> but they do have physiological consequences; the loss of APPL resulted in a reduction in the amplitude of evoked excitatory junctional potentials (EJPs) when recording from body wall muscles of larvae (Ashley et al., 2005). Performing whole-cell patch clamp measurements on embryonic cells in culture revealed that both, the loss and overexpression of APPL increased A-type K<sup>+</sup> currents, suggesting a role of APPL in modulating synaptic function (Li et al., 2004). Additional studies by the same group suggest that this is mediated via the secreted ectodomain (sAPPL) and a similar finding has been made in mammals using cultured hippocampal neurons treated with sAPPα (Furukawa et al., 1996).

A role of APPL in axonal transport was suggested by the finding that overexpression of APPL caused transport defects detectable by the accumulation of vesicles or mitochondria, whereby this phenotype required the presence of the C-terminus (Torroja et al., 1999; Gunawardena and Goldstein, 2001; Shaw and Chang, 2013). Changes in axonal trafficking have also been described after the loss of APPL (Gunawardena and Goldstein, 2001), indicating that the role in axonal transport is a physiological function of APPL. This is also supported by the observation that a dominant-negative mutation of Drosophila Tip60, a histone acetyltransferase that has been shown to bind to the C-terminus of APP proteins, also induced axonal trafficking defects (Johnson et al., 2013). In addition, this mutation enhanced transport defects induced by APP. Another manipulation that enhanced the trafficking defects caused by APP and also by APPL is a knock down of nebula while overexpression of Nebula suppressed this phenotype (Shaw and Chang, 2013). Manipulating Nebula alone had no effect and therefore its function in axonal trafficking under normal physiological conditions is unclear. Interestingly, Nebula is the fly homolog of Down syndrome critical region 1 (DSCR1) and almost all Down syndrome patients develop AD (Wisniewski et al., 1985). At this point the role of DSCR1 in AD is not understood; however, due to DSCR1 being overexpressed in Down syndrome (Fuentes et al., 2000) one would expect a suppression of possible transport defects caused by the third copy of the APP gene. Interestingly, overexpression as well as loss of Nebula affects synaptic function and memory formation in flies (Chang et al., 2003; Chang and Min, 2009).

# APPL AND NEURONAL SURVIVAL

The experiments described above reveal that changes in APPL can interfere with neuronal development. But APPL has also been demonstrated to play a role in the integrity of the adult nervous system. Appl<sup>d</sup> flies have a significantly reduced life span, shortened to approximately two thirds of the survival span of wild type flies, and they show signs of neurodegeneration when aged (Wentzell et al., 2012). This was detectable by the formation of spongiform lesions in the brains of 3 week old Appl<sup>d</sup> flies and although they are not very numerous, such lesions do not occur in age-matched wild type brains. Furthermore, the loss of APPL can aggravate the neurodegeneration caused by mutations in other genes, like yata and löchrig (loe). Yata belongs to a family of pseudokinases, found in almost all eukaryotes, that play a role in vesicle trafficking of secretory proteins and the export of tRNA from the nucleus (Anamika et al., 2009). yata mutant flies show progressive degeneration that affects the brain and retina (Sone et al., 2009). This phenotype was enhanced by the loss of APPL whereas overexpression of APPL ameliorated it, suggesting a neuroprotective function of APPL. Similarly, combining the loe mutation with Appl<sup>d</sup> significantly worsened the neurodegeneration that is observed in the brain of loe mutants (Tschäpe et al., 2002). loe encodes the γ-subunit of AMP-activated protein kinase (AMPK), a key enzyme in regulating energy homeostasis (Kemp et al., 1999). AMPK also regulates protein prenylation and loe mutant flies show an increase in Rho prenylation and activity and changes in actin dynamics (Cook et al., 2012, 2014). Interestingly, the Rho pathway has also been connected with modulating Aβ production in vertebrates (Tang and Liou, 2007). In contrast to the enhancing effect of the Appl<sup>d</sup> mutant, overexpressing APPL suppressed the degeneration in loe mutant flies and the same effect was achieved by expressing the secreted sAPPL (Wentzell et al., 2012). However, the latter was only protective in

the presence of endogenous APPL and co-immunoprecipitation experiments showed that sAPPL can bind to full-length APPL. This suggests that sAPPL acts as a ligand that binds to full-length APPL as a receptor (**Figure 2**). The protective function appears to be mediated specifically by the α-cleaved ectodomain because additional expression of Kuzbanian (KUZ) was also protective (Wentzell et al., 2012). KUZ is homologous to ADAM10 and like its vertebrate ortholog it acts as an α-secretase (Carmine-Simmen et al., 2009). In contrast, increasing β-cleavage by inducing Drosophila β-secretase (dBACE; Bolkan et al., 2012) expression enhanced the degeneration in loe. A neuroprotective function of the α-cleaved sAPP was also described in mice and like in flies it required the presence of full-length APP (Milosch et al., 2014). Together with findings that expression of APPL ameliorated the degenerative phenotype in a Drososophila RasGAP (vap) mutant and flies mutant for the microtubule binding protein MAP1B (futscholk) (Wentzell et al., 2012), this further supports a neuroprotective function of APP proteins and their α-cleaved ectodomains. Interestingly, in the case of loe a reduction in sAPPLα may be part of the mechanism leading to the degenerative phenotype in this mutant because loe mutant flies showed a decrease in APPL processing whereas additional LOE expression promoted cleavage (Tschäpe et al., 2002).

That the cleavage and generation of specific fragments is important for the protective function is also supported by studying mutations in proteins that affect APPL processing. Transmembrane and Coiled-coil domain 2 (TMCC2) is a vertebrate protein that can form a complex with APP and ApoE and promote APP cleavage (Hopkins et al., 2011). Its Drosophila homolog is encoded by dementin (dmtn) and loss of neuronal DMTN caused neuronal degeneration in the adult brain and a reduced live span (Hopkins, 2013). It also interfered with the processing of APPL, resulting in the production of an abnormal 50 kD fragment. Similarly, the loss of dBace in photoreceptors resulted in degeneration but in this case of glial cells in the lamina, the main target region of photoreceptors (Bolkan et al., 2012). That this is indeed due to an effect on APPL and not another target of dBace was shown by the result that this phenotype was suppressed in the Appl<sup>d</sup> background. In contrast, expressing secretion-deficient APPL (APPLsd) enhanced the glial degeneration, supporting the hypothesis that cleavage of APPL is required for glial survival. These findings reveal that APPL not only plays a role in the survival of both, neurons and glia. However, for glia additional full-length APPL seems to be deleterious and the cleavage by dBACE prevents the glial cell death (**Figure 2**).

Lastly, APPL was found to be upregulated after injury (Leyssen et al., 2005). However whether this is connected to a protective mechanism, like a possible axonal sprouting of neighboring neurons after neuronal loss, is so far unclear. Although an upregulation of APP after injury has also been observed in mammals, this mostly seems to have negative consequences because it can increase the risk to develop AD or other neurodegenerative diseases (Shi et al., 2000; Gupta and Sen, 2016; Ułamek-Kozioł et al., 2016).

#### BEHAVIORAL DEFICITS AND APPL

As mentioned before, changes in APPL levels also affect behavior, including memory (see accompanying review by V. Goguel). Furthermore, Appl<sup>d</sup> flies also show a significantly reduced performance in the fast-phototaxis assay (Luo et al., 1992), a test that can be used to measure general fitness, locomotion, and visual orientation (Benzer, 1967). The phototaxis phenotype may be due to the loss of secreted APPL fragments because expression of full-length APPL could restore this function whereas secretion-deficient APPLsd could not (Luo et al., 1992). Interestingly, also the overexpression of APPL induced phototaxis phenotypes that were further enhanced by expression of dBACE (Carmine-Simmen et al., 2009). The latter suggests that the deficits in the phototaxis assay after APPL overexpression are due to the generation of the neurotoxic dAβ cleaved from the full-length protein. This is supported by the finding that expression of only dAβ also causes phototaxis defects that are even more severe (Carmine-Simmen et al., 2009). In the case of APPL overexpression, the behavioral deficits could be a consequence of the degeneration and neuronal cell death that is detectable after APPL expression. In contrast, the Appl<sup>d</sup> deletion mutant shows very subtle morphological changes and modestly increased cell death is only detectable late in life. Therefore the loss of APPL may directly interfere with neuronal function, possibly by affecting synaptic functions.

Finally, recent experiments suggest a function of APPL in the regulation of circadian rhythms, due to the observation that increasing APPL levels prevented the age-related decline in rhythmicity (Blake et al., 2015). This function seems to be specifically mediated by the full-length protein because expressing additional dBACE or KUZ resulted in a disruption of the rhythmic activity pattern. In addition to supporting a protective role for the full-length APPL this also indicates that a cleavage product is deleterious for rhythmicity. Because dBACE and KUZ expression disrupted the circadian activity pattern, this appears to be due to a fragment produced by both cleavage events, excluding dAβ and the N-terminal fragment. However, both β- and α-cleavage promote processing by the γ-secretase and therefore the production of the AICD. Confirming the role of the AICD in circadian rhythmicity, expressing only the AICD pan-neuronally or specifically in the central pacemaker neurons disrupted rhythmicity in an age-dependent manner (Blake et al., 2015). Like humans, flies are diurnal animals and this rhythmicity is regulated by the circadian clock. The clock generates a circa 24 h periodicity by an autoregulatory negative feedback loop of four core clock genes and their proteins; Clock and Cycle (BMAL1 in mammals) are the positive elements which promote transcription of the negative elements Period and Timeless (Hardin and Panda, 2013). These proteins are transcriptional regulators that generate circadian rhythms in downstream clock-controlled genes, providing a temporal coordination of cellular and physiological processes with the environment. Supporting a direct role of APPL in regulating circadian rhythms, altering the cleavage pattern of APPL interfered with the rhythmic expression pattern of Period in the central pacemaker cells while not affecting the survival of these neurons (Blake et al., 2015). Because the AICD has been shown to play a role in transcriptional regulation in vertebrates (Cao and Sudhof, 2004; von Rotz et al., 2004), this function of APPL may be an effect of the AICD on the transcription of Period. Not being a transcription factor itself, the AICD forms a ternary complex Fe65 and Tip60. Intriguingly, the loss of Drosophila Tip60 induces sleep disturbances and reduces the axon length of central pacemaker neurons (Pirooznia et al., 2012), providing another hint that the AICD may regulate the circadian clock and rhythmicity.

#### CONCLUSION

The studies described above show that full-length APPL can act as a receptor that promotes neurite growth and synaptogenesis in vivo. This function appears to require the C-terminus which, together with various interaction factors, can activate

#### REFERENCES


downstream signaling pathways, similary to what has been suggested for vertebrate APP (Deyts et al., 2016). For some of these neurodevelopmental functions, cell adhesion molecules like Fas II may act as the activating signals. Fas II has been shown to be required for the function of APPL at the larval NMJ but Fas II is also enriched in mushroom body neurons. Therefore, an interaction between APPL and Fas II might also be required for the correct formation of the mushroom body lobes. Because the mushroom body neurons are crucial for memory formation, this raises the possibility that the Fas II-APPL interactions take part in synaptic plasticity and memory formation, an issue that has not been explored so far.

However, APPL can also act as a ligand via its secreted ectodomains, whereby the α- vs. the β-cleaved fragment seem to play different, even opposing roles. Expression of the secreted sAPPL promotes correct α-lobe formation in Appl<sup>d</sup> mutants and neuronal survival in loe, whereby the protective function appears to be mediated specifically by the α-cleaved ectodomain whereas the β-cleaved form is neurotoxic. Such opposing functions of the ectodomains have also been described in vertebrates with sAPPα connected to neuroprotective functions (Araki et al., 1991; Mattson et al., 1993; Goodman and Mattson, 1994) whereas sAPPβ was shown to be deleterious for neuronal survival (Nakagawa et al., 2006; Nikolaev et al., 2009). Lastly, the experiments in Drosophila showed that APPL can activate its receptor function by binding to its own ectodomain and recently, a similar finding was reported for mammals where sAPPα protected cells from serum-starvation induced cell death only in the presence of full-length APP (Milosch et al., 2014).

Although the studies in Drosophila and other models have provided important insights into the functions of APP proteins and their fragment, we are still far away from understanding the various roles of this protein. Drosophila provides a variety of tools and assays to study the physiological functions of APP proteins in vivo and future experiments including these model will hopefully unravel the functions of APP and the pathways it is involved in. In turn, this can then provide the basis to determine whether and how disruptions of these functions contribute to the deleterious effects seen in Alzheimer patients.

#### AUTHOR CONTRIBUTIONS

MC provided literature. DK wrote review.

#### FUNDING

This work was supported by the National Institute of Health, NINDS project grant (NS096332).

disease and control subjects. Ann. Neurol. 30, 686–693. doi: 10.1002/ana. 410300509


protein dX11/Mint. J. Neurosci. 25, 5943–5955. doi: 10.1523/JNEUROSCI. 1144-05.2005


Biochem. Biophys. Res. Commun. 120, 885–890. doi: 10.1016/s0006-291x(84) 80190-4


interferes with cholesterol homeostasis and Appl processing. EMBO J. 21, 6367–6376. doi: 10.1093/emboj/cdf636


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Cassar and Kretzschmar. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Role of Drosophila Amyloid Precursor Protein in Memory Formation

#### Thomas Preat and Valérie Goguel\*

Genes and Dynamics of Memory Systems, Brain Plasticity Unit, Centre National de la Recherche Scientifique (CNRS), ESPCI Paris, PSL Research University, Paris, France

The amyloid precursor protein (APP) is a membrane protein engaged in complex proteolytic pathways. APP and its derivatives have been shown to play a central role in Alzheimer's disease (AD), a progressive neurodegenerative disease characterized by memory decline. Despite a huge effort from the research community, the primary cause of AD remains unclear, making it crucial to better understand the physiological role of the APP pathway in brain plasticity and memory. Drosophila melanogaster is a model system well-suited to address this issue. Although relatively simple, the fly brain is highly organized, sustains several forms of learning and memory, and drives numerous complex behaviors. Importantly, molecules and mechanisms underlying memory processes are conserved from flies to mammals. The fly encodes a single non-essential APP homolog named APP-Like (APPL). Using in vivo inducible RNA interference strategies, it was shown that APPL knockdown in the mushroom bodies (MB)—the central integrative brain structure for olfactory memory—results in loss of memory. Several APPL derivatives, such as secreted and full-length membrane APPL, may play different roles in distinct types of memory phases. Furthermore, overexpression of Drosophila amyloid peptide exacerbates the memory deficit caused by APPL knockdown, thus potentiating memory decline. Data obtained in the fly support the hypothesis that APP acts as a transmembrane receptor, and that disruption of its normal function may contribute to cognitive impairment during early AD.

Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Peter Soba, Center for Molecular Neurobiology (ZMNH), University of Hamburg, Germany Doris Kretzschmar, Oregon Health and Science University (OHSU), USA

> \*Correspondence: Valérie Goguel

valerie.goguel@espci.fr

Received: 26 October 2016 Accepted: 28 November 2016 Published: 08 December 2016

#### Citation:

Preat T and Goguel V (2016) Role of Drosophila Amyloid Precursor Protein in Memory Formation. Front. Mol. Neurosci. 9:142. doi: 10.3389/fnmol.2016.00142 Keywords: Drosophila melanogaster, amyloid precursor protein, learning, memory, mushroom bodies, conditional expression, amyloid peptide

#### DROSOPHILA AS A MODEL TO STUDY THE ROLE OF THE APP PATHWAY IN MEMORY

The initial events leading to Alzheimer's disease (AD) are still unknown. Amyloid deposits, a hallmark of AD, are formed by the aggregation of amyloid peptides (Aβ) resulting from proteolytic processing of the amyloid precursor protein (APP; Turner et al., 2003). APP is a transmembrane protein that is subjected to two exclusive proteolytic pathways: the non-amyloidogenic pathway initiated by the α-secretase producing a secreted APP form (sAPPα), and the amyloidogenic pathway initiated by the β-secretase leading to the production of Aβ. For many years, the amyloid hypothesis put Aβ at the center of the etiology of AD (Hardy and Selkoe, 2002). However, several studies have shown that APP plays a positive role in memory (Meziane et al., 1998; Ring et al., 2007), raising the possibility that aside from Aβ toxicity, APP loss-of-function may participate in AD, particularly during the early stages of the disease characterized by memory impairment. The physiological role of APP is difficult to assess due to its numerous proteolytic metabolites. Functional studies of the APP pathway in rodents are also limited due to the redundancy of the three APP-related genes and the lethality of the triple knockout (Heber et al., 2000; Herms et al., 2004). In addition, mouse studies have been essentially performed using constitutive mutants, making it hard to discriminate developmental functions from direct roles in the adult brain. The Drosophila melanogaster genome contains homologs of 75% of human disease-related genes (Fortini et al., 2000; Reiter et al., 2001). Interestingly, the fly expresses a single non-essential APP ortholog, called APP-Like (APPL). APPL is a neuronal-specific protein particularly expressed in the axonal neuropil of the adult mushroom bodies (MB; Torroja et al., 1996).

Appl-deficient flies (Appl<sup>d</sup> ) display phototaxis deficits that are alleviated upon ectopic expression of human APP (hAPP), which is the first demonstration that APPL is an APP ortholog (Luo et al., 1992). APPL/hAPP sequence comparison found homology regions at the E1 and E2 ectodomains and at the C-terminal intracellular domain (Rosen et al., 1989). APPL protein (887 aa) is substantially longer than hAPP (695 aa), largely due to having longer sequences between E1 and E2 domains and between E2 and Aβ sequences. Aβ sequences are manifestly not conserved between APP and APPL, and amyloid peptides are not described in wild-type Drosophila. However, a Drosophila Aβ-like peptide (dAβ) was identified in old flies overexpressing APPL (Carmine-Simmen et al., 2009). Indeed, APPL overexpression in old age leads to Thioflavin-Spositive aggregates that are associated with neurodegeneration, suggesting that APPL processing produces an analog of human Aβ (Carmine-Simmen et al., 2009). Importantly, APPL undergoes similar proteolytic pathways to APP (Poeck et al., 2011), and the homologs of all mammalian secretases have been characterized in the fly (Rooke et al., 1996; Boulianne et al., 1997; Hong and Koo, 1997; Carmine-Simmen et al., 2009).

Despite its relative simplicity, the fly brain is highly structured and drives sophisticated behaviors. In particular, it is extensively used as a model system to study associative memory. Molecular mechanisms underlying memory are conserved from flies to mammals (McGuire et al., 2005), and the neuronal structures involved are well described (Heisenberg, 2003; Waddell, 2010; Aso et al., 2014a,b). The MB are known as the central integrative brain structure for olfactory associative memory (de Belle and Heisenberg, 1994; Pascual and Préat, 2001; Gerber et al., 2004; Krashes et al., 2007; Gervasi et al., 2010). The MB are a bilateral structure composed of 4000 intrinsic neurons, the Kenyon cells, classed into three subtypes whose axons form two vertical (α and α') and three medial (β, β' and γ) lobes (Crittenden et al., 1998). Using a classical conditioning paradigm in which an odorant is paired with the delivery of electric shocks, the fly is capable of forming six discrete aversive memory phases reflected at neural network level (Bouzaiane et al., 2015). Learning and short-term memory are measured immediately after a single conditioning, while middle-term memory (MTM) is assessed 1–3 h later. The fly can also produce two antagonistic forms of consolidated memory (Isabel et al., 2004). Long-term anesthesia-resistant memory (LT-ARM) is

Many human neurodegenerative diseases can be modeled in Drosophila (Bilen and Bonini, 2005). In particular, transgenic flies have been generated to analyze human Aβ-induced toxicity. Expression in the Drosophila brain of human Aβ42 resulted in defects similar to that observed in the mouse (Finelli et al., 2004; Greeve et al., 2004; Iijima et al., 2004, 2008; Crowther et al., 2005; Zhao et al., 2010). Thus, similarities between Aβ-induced neurotoxic biochemical pathways in flies and humans make Drosophila a relevant model to study the molecular basis of AD pathogenesis. Neuronal expression of human Aβ42 leads to a learning deficit in young flies, and MTM deficit in older flies (Iijima et al., 2004, 2008; Fang et al., 2012). Likewise, neuronal overexpression of hAPP alters learning and MTM in young flies and these deficits become more pronounced as the fly ages (Sarantseva et al., 2009). hAPP expression in the MB was also shown to alter LTM (Goguel et al., 2011).

APP overexpression-related memory deficits likely result from accumulation of amyloid peptides, especially as it was suggested that the fly secretases can cleave APP (Greeve et al., 2004). Furthermore, the above-cited results were obtained using constitutive overexpression, creating conditions under which APP and/or Aβ accumulate over the entire life of the fly, thus increasing their toxic potential, particularly during developmental stages. In fact, when dAβ overexpression is achieved in the MB of adult flies for only 2 days, no MTM deficit is observed (Bourdet et al., 2015a). Taken together, the data suggest that memory impairments observed with constitutive expression of APP result from a developmental defect and/or general neuronal dysfunction rather than from some specific alteration of the molecular mechanisms required to sustain memory formation.

# APPL IS REQUIRED FOR SPECIFIC MEMORY PHASES

Aβ toxicity has been a focus of research for years, but it now appears essential to better understand APP function in brain physiology. Early on, it was shown that Appl<sup>d</sup> flies do not form normal associative learning, but it was impossible to conclude that APPL was involved in this process as the Appl<sup>d</sup> flies did not react normally to electric shock exposure, which was the unconditioned stimulus used for the study (Luo et al., 1992). It was later shown that Appl disruption leads to slight abnormalities in the morphology of the MB lobes (Li et al., 2004). More recently, a study demonstrated the role of APPL in brain wiring (Soldano et al., 2013). Thus, functional studies need to rule out possible roles during brain development. One of the major advantages of the Drosophila model is that it can be used to implement inducible loss-of-function studies. Indeed, the expression of any gene of interest can be controlled both spatially (Brand and Perrimon, 1993) and temporally (McGuire et al., 2003).

Using conditional RNA interference, it was demonstrated that APPL expression in the adult MB is required for the proper formation of specific memory phases. APPL silencing in the MB of adult flies was shown to disrupt MTM and LTM, but neither learning nor ARM formation was affected (Goguel et al., 2011; Bourdet et al., 2015b). MTM and LTM are two memory phases known to share identical neuronal circuits (Bouzaiane et al., 2015), indicating a role for APPL in these structures. These memory phenotypes are reminiscent of the pattern displayed by amnesiac mutants (Quinn et al., 1979; Feany and Quinn, 1995; DeZazzo et al., 1999; Yu et al., 2006), suggesting that APPL and Amnesiac, a predicted neuropeptide precursor showing homology to an adenylate cyclase-activating peptide (Feany and Quinn, 1995), are involved in the same molecular pathways.

Memory deficits are thus caused by loss of APPL function, independent of the amyloid pathway toxicity. This data further supports the hypothesis that APP downregulation might contribute to early cognitive impairment in AD. To further assess which APPL fragment is required for memory processes, two APPL-mutant forms were used: a constitutively-secreted APPL protein (APPL<sup>s</sup> ) and a non-cleavable secretion-defective form (APPLsd). The Appl<sup>s</sup> sequence contains a stop codon that generates a soluble 788-amino-acid N-terminal fragment of APPL, whereas APPLsd is deleted from the α and β cleavage sites, thus preventing its processing (Torroja et al., 1996, 1999). Consequently, APPLsd is exclusively expressed as a transmembrane protein. Overexpression of APPL<sup>s</sup> in the adult MB rescued the MTM deficit caused by a reduction of endogenous APPL levels, indicating that a secreted fragment of APPL is involved in memory (Bourdet et al., 2015b). This is consistent with mammalian studies showing a role for sAPPα in memory (Meziane et al., 1998; Bour et al., 2004; Ring et al., 2007; Taylor et al., 2008). Unexpectedly, however, overexpression of the fly α-secretase KUZ (Rooke et al., 1996), thought to increase sAPPLα levels, did not rescue the memory deficit caused by APPL partial loss-offunction, and even further exacerbated the MTM impairment (Bourdet et al., 2015b). Interestingly, KUZ overexpression in this context was shown to decrease full-length APPL (fl-APPL) protein levels, prompting the hypothesis that the exacerbation of the memory phenotype resulted from a reduction of fl-APPL levels. Supporting this hypothesis, transient APPLsd expression in the MB was also able to restore wild-type MTM in an APPL knockdown background (Bourdet et al., 2015b). Interestingly, neither APPL<sup>s</sup> nor APPLsd overexpression rescued the LTM phenotype of APPL partial loss-of-function flies. Although negative results are difficult to interpret, they may indicate distinct molecular APPL requirements for MTM and LTM.

Taken together, the data indicate that both fl-APPL and sAPPL are involved in MTM. This apparently contradicts a previous study showing that sAPPα could rescue the spatial learning defect of APP knockout mice (Ring et al., 2007). However, the APPL proteins APLP1 and APLP2 were preserved in that study. As the three APP homologs show some functional redundancy (Anliker and Müller, 2006), disruption of full-length APP functions might have been partially fulfilled by APLP1 or APLP2. Memory function cannot therefore be attributed exclusively to sAPPα.

Interestingly, APP may be a receptor for sAPPα (Young-Pearse et al., 2008; Gralle et al., 2009). In Drosophila, sAPPL was shown to act as a soluble ligand for neuroprotective functions (Wentzell et al., 2012). Moreover, co-immunoprecipitation experiments from transfected Kc cells uncovered an interaction between fl-APPL and sAPPL, suggesting that sAPPL could be a ligand for fl-APPL (Wentzell et al., 2012). It is thus tempting to speculate that APPL is involved in MTM processes through a sAPPL/fl-APPL ligand/receptor interaction.

### Aβ EXACERBATES THE MEMORY DEFICIT CAUSED BY APPL PARTIAL LOSS-OF-FUNCTION IN DROSOPHILA

Neurotoxic effects of Aβ accumulation have been well documented, and studies have shown that the β-APP cleavage enzyme, Beta-secretase 1 (BACE1), has a negative impact on memory. In mice models of AD, BACE1 deficiency rescues memory deficits (Ohno et al., 2004, 2007), and conversely, expression of hBACE1 was shown to worsen learning and memory deficits (Rockenstein et al., 2005; Chen et al., 2012). In normal mice, hBACE1 gene knock-in caused AD-relevant cognitive impairment (Pluci´nska et al., 2014). The authors concluded that low hBACE1 levels were sufficient to cause the formation of toxic Aβ oligomeric assemblies. It is important to note here that hBACE1 knock-in mouse also generated decreased full-length APP levels (Pluci´nska et al., 2014), a feature that could participate in cognitive impairment. In Drosophila, overexpression of the fly β-secretase (dBACE; Carmine-Simmen et al., 2009) in the adult MB did not impact MTM (Bourdet et al., 2015a). In contrast, it exacerbated the memory deficit of low-APPL-level flies (Bourdet et al., 2015a). One possibility is that, similar to KUZ overexpression, an increase of dBACE-mediated processing reduces fl-APPL levels, thus aggravating the memory deficit caused by APPL knockdown. Interestingly, similar results were observed with dAβ: dAβ expression in adult MB neurons impaired MTM only in an APPL partial loss-of-function background (Bourdet et al., 2015a). It was hypothesized that both dBACE and dAβ expression exacerbate the memory deficit caused by a reduction of APPL levels through similar mechanisms mediated by an increase in dAβ production that has knock-on effects on APPL function. Memory would thus be affected by two related processes—APPL downregulation and Aβ toxicity—uncovering a functional link between APPL and Aβ.

# A PHYSIOLOGICAL ROLE FOR Aβ IN MEMORY?

Several reports have shown that at very low physiological concentrations, Aβ modulates synaptic strength (Kamenetz et al., 2003; Abramov et al., 2009) and enhances memory (Puzzo et al., 2008, 2012; Garcia-Osta and Alberini, 2009; Morley et al., 2010). Aβ appears to be a modulator of synaptic activity requiring a fine balance between production and removal. Neprilysins are the major Aβ-degrading enzymes (Iwata et al., 2001) and, as such, are thought to be key to AD. Neprilysin proteins are zinc-dependent endopeptidases known to inactivate small peptides. Their active site faces the extracellular space, and they can be present at presynaptic sites (Fukami et al., 2002; Iwata et al., 2004). Neprilysins play a major role in brain function by terminating neuropeptide signaling at the cell surface, and they are involved in many neuronal processes from axonal regeneration and synaptic plasticity to neuro-inflammation, while at the behavioral level neprilysins have been implicated in motor function, anxiety, circadian rhythms and sleep (Nalivaeva et al., 2012).

The issue of whether neprilysins are involved in memory in non-pathological conditions has been addressed in Drosophila. Four neprilysins are expressed in adult Drosophila brain (Meyer et al., 2011), and we have shown using inducible RNA interference that they are all required for MTM and LTM (Turrel et al., 2016). We have proposed that these neprilysins target several neuropeptides involved in memory processes (Turrel et al., 2016). An attractive hypothesis is that one of these targets might be Aβ peptide derived from physiological APPL processing. Consistently, the memory phenotypes observed after neprilysin silencing are reminiscent of the specific pattern in APPL mutants: only MTM and LTM are impaired, suggesting a functional interaction between neprilysins and APPL. Neprilysin 2 would be a good candidate here, since several studies have shown that it is capable of degrading human Aβ42 (Finelli et al., 2004; Cao et al., 2008).

in memory. (A) Secreted APPL (sAPPL) and full-length APPL (fl-APPL) interact to induce middle-term memory (MTM) formation via a signaling pathway such as G protein activation. (B) After sAPPL and fl-APPL have interacted, secretases produce APPL intracellular domain (AICD). After translocation into the nucleus, AICD activates the transcription required for long-term memory (LTM) formation. (C) dAβ inhibits APPL: Drosophila Aβ-like peptide (dAβ) binding to APPL promotes APPL cis-dimerization thus preventing sAPPL/fl-APPL interaction.

#### CONCLUSIONS

In the fly, both sAPPL and fl-APPL are required for MTM, raising the possibility that sAPPL is a ligand of its own precursor (**Figure 1**). A good candidate acting downstream of APPL could be the G<sup>o</sup> signaling pathway (Nishimoto et al., 1993; Okamoto et al., 1995; Ramaker et al., 2013). It remains to be determined whether sAPPL and fl-APPL are expressed by the same Kenyon cells, which would point to an autocrine mechanism, or whether fl-APPL is expressed in one specific cell type while sAPPL is secreted from another.

It has been reported that the normal physiological function of APP may be compromised by Aβ (Bignante et al., 2013). Direct interactions between Aβ fibrils and APP were described, with Aβ acting to enhance APP multimerization, a potentially toxic mechanism (Lorenzo et al., 2000; Van Nostrand et al., 2002; Lu et al., 2003; Shaked et al., 2006; Sola Vigo et al., 2009; Kedikian et al., 2010). dAβ expression enhances APPL knockdown memory impairment, raising the possibility that dAβ-induced toxicity may be caused, at least in part, by a physical dAβ/APPL interaction. Such a direct interaction could thus promote APPL cis-dimerization, a process that would compromise its function in memory (**Figure 1**). Furthermore, sAPPL and dAβ could have opposite functions, as sAPP was shown to disrupt APP dimers (Gralle et al., 2009). Under physiological conditions, dAβ may also interact with APPL, for example to balance and/or terminate APPL signaling. Characterization of dAβ as a neprilysin substrate would support the hypothesis of a physiological role for dAβ in memory.

It is not known whether distinct memory phases are supported by distinct APPL-mediated mechanisms. LTM is the only memory phase to depend on transcription regulation (Dubnau et al., 2003; Didelot et al., 2006). Given that APP intracellular domain (AICD), the cleavage product of APP by γ-secretase, could function as a transcription factor (Cao and Südhof, 2001; Kimberly et al., 2001; Müller et al., 2007), it would be important to know whether the APPL intracellular domain plays a specific role in LTM formation (**Figure 1**). It has already been reported that AICD production correlates to enhanced plasticity and memory in a TgAPP mice background (Ma et al., 2007).

The vast majority of AD cases are late-onset, happening to people at age 65 and older. Even though AD has not been described in Drosophila, the fly nonetheless undergoes an age-related memory impairment (AMI). This AMI exclusively concerns MTM and LTM, as none of the other memory phases decline with age (Tamura et al., 2003; Tonoki and Davis, 2012). Most strikingly, the memory phases affected by APPL knockdown in the MB are precisely those that are lost during fly aging. It is not known whether AMI could be linked to an aging-induced APPL dysfunction. It would be valuable to learn whether APPL expression and/or processing are modified during fly aging. To further explore these issues, the fly offers a suitable simplified system to decipher APP physiological function in learning and memory.

#### AUTHOR CONTRIBUTIONS

VG wrote the manuscript, TP discussed with VG and reviewed the manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the Fondation pour la Recherche Médicale (DEQ20140329540).


redundant functions of amyloid precursor protein family members. J. Neurosci. 20, 7951–7963.


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Preat and Goguel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Physiological Concentrations of Amyloid Beta Regulate Recycling of Synaptic Vesicles via Alpha7 Acetylcholine Receptor and CDK5/Calcineurin Signaling

#### Edited by:

Ulrike C. Müller, Heidelberg University, Germany

#### Reviewed by:

Andreas Vlachos, Albert Ludwig University of Freiburg, Germany Gael Barthet, UMR5297 Institut Interdisciplinaire de Neurosciences (IINS), France

> \*Correspondence: Anna Fejtova

anna.fejtova@uk-erlangen.de

#### †Present address:

Vesna Lazarevic, Department of Clinical Neuroscience, Center for Molecular Medicine, Karolinska Institutet, Stockholm, Sweden

> ‡These authors have contributed equally to this work.

> > Received: 30 March 2017 Accepted: 26 June 2017 Published: 21 July 2017

#### Citation:

Lazarevic V, Fienko S, ´ Andres-Alonso M, Anni D, Ivanova D, Montenegro-Venegas C, Gundelfinger ED, Cousin MA and Fejtova A (2017) Physiological Concentrations of Amyloid Beta Regulate Recycling of Synaptic Vesicles via Alpha7 Acetylcholine Receptor and CDK5/Calcineurin Signaling. Front. Mol. Neurosci. 10:221. doi: 10.3389/fnmol.2017.00221 Vesna Lazarevic1,2,3† , Sandra Fienko ´ 1‡ , Maria Andres-Alonso1‡ , Daniela Anni 4‡ , Daniela Ivanova<sup>1</sup> , Carolina Montenegro-Venegas <sup>2</sup> , Eckart D. Gundelfinger 2,3,5,6 , Michael A. Cousin<sup>7</sup> and Anna Fejtova1,2,4,5 \*

<sup>1</sup>RG Presynaptic Plasticity, Leibniz Institute for Neurobiology, Magdeburg, Germany, <sup>2</sup>Department of Neurochemistry and Molecular Biology, Leibniz Institute for Neurobiology, Magdeburg, Germany, <sup>3</sup>German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany, <sup>4</sup>Molecular Psychiatry, Department of Psychiatry and Psychotherapy, University Hospital, University of Erlangen-Nuremberg, Erlangen, Germany, <sup>5</sup>Center for Behavioral Brain Sciences, Otto von Guericke University, Magdeburg, Germany, <sup>6</sup>Medical Faculty, Otto von Guericke University, Magdeburg, Germany, <sup>7</sup>Centre for Integrative Physiology, University of Edinburgh, Edinburgh, United Kingdom

Despite the central role of amyloid β (Aβ) peptide in the etiopathogenesis of Alzheimer's disease (AD), its physiological function in healthy brain is still debated. It is well established that elevated levels of Aβ induce synaptic depression and dismantling, connected with neurotoxicity and neuronal loss. Growing evidence suggests a positive regulatory effect of Aβ on synaptic function and cognition; however the exact cellular and molecular correlates are still unclear. In this work, we tested the effect of physiological concentrations of Aβ species of endogenous origin on neurotransmitter release in rat cortical and hippocampal neurons grown in dissociated cultures. Modulation of production and degradation of the endogenous Aβ species as well as applications of the synthetic rodent Aβ<sup>40</sup> and Aβ<sup>42</sup> affected efficacy of neurotransmitter release from individual presynapses. Low picomolar Aβ<sup>40</sup> and Aβ<sup>42</sup> increased, while Aβ depletion or application of low micromolar concentration decreased synaptic vesicle recycling, showing a hormetic effect of Aβ on neurotransmitter release. These Aβ-mediated modulations required functional alpha7 acetylcholine receptors as well as extracellular and intracellular calcium, involved regulation of CDK5 and calcineurin signaling and increased recycling of synaptic vesicles. These data indicate that Aβ regulates neurotransmitter release from presynapse and suggest that failure of the normal physiological function of Aβ in the fine-tuning of SV cycling could disrupt synaptic function and homeostasis, which would, eventually, lead to cognitive decline and neurodegeneration.

Keywords: amyloid beta, acetylcholine receptors, synaptic vesicle recycling, neurotransmitter release, CDK5, calcineurin

#### INTRODUCTION

Amyloid beta (Aβ) peptide arises by processing of the extracellular domain of the amyloid precursor protein (APP) mediated by the β-and γ-secretase proteolytic activity. Aβ is famous for its connection with Alzheimer's disease (AD), the most common form of neurodegeneration, characterized by progressive cognitive decline, memory impairment and formation of amyloid plaques in the brains of affected patients. Aβ is the main component of these plaques (Glenner and Wong, 1984). This fact together with a genetic link of earlyonset familiar forms of AD to mutations interfering with the proteolytic processing of APP (Goate et al., 1991) advocate the key role of extracellular Aβ in the pathogenesis of AD. However, the mechanisms by which dysregulated extracellular Aβ disrupts the brain function are still not completely understood. While there is no clear relationship between the number of senile plaques and disease severity, progressive synapse dismantling occurring before formation of any amyloid deposits emerged as the best correlate of the cognitive decline in AD patients and animal models (Terry et al., 1991; Mucke et al., 2000). As synapse loss in AD is preceded by defects in neuronal transmission and plasticity, it has been suggested that extracellular Aβ might be involved in the regulation of these processes (Chapman et al., 1999; Hsia et al., 1999; Walsh et al., 2002; Palop and Mucke, 2010). A large body of evidence supports an inhibitory effect of Aβ on synaptic function. Elevated (high nanomolar and low micromolar) extracellular Aβ reduces neurotransmission mainly by postsynaptic mechanisms including increased internalization or desensitization of postsynaptic glutamate receptors and downstream signaling (Walsh et al., 2002; Hsieh et al., 2006). Recently, it has also been demonstrated that high nanomolar Aβ affects endocytosis of presynaptic neurotransmitter vesicles indicating that the inhibitory effect of Aβ on membrane trafficking is not restricted only to the postsynaptic compartment (Park et al., 2013).

In healthy brain, Aβ is released into the extracellular space depending on neuronal activity (Kamenetz et al., 2003; Cirrito et al., 2005). Thus, due to the widely accepted inhibitory nature of Aβ on the neurotransmission, it has been proposed that endogenous Aβ functions as a negative modulator of synaptic strength in a physiological feed-back mechanism preventing over excitation of brain circuits. Dysregulation of this homeostatic mechanism would trigger synaptic destabilization eventually leading to the development of AD (Palop and Mucke, 2010). However, extracellular concentrations of Aβ in normal brain have been estimated to low picomolar levels, far lower than the concentrations used in most studies showing the Aβ-induced synaptic depression and neurotoxicity (Cirrito et al., 2003; Puzzo et al., 2011). Unexpectedly, several studies investigating the impact of physiological concentration of Aβ revealed a positive effect on neuroplasticity and learning (Puzzo et al., 2008, 2011; Garcia-Osta and Alberini, 2009). Hippocampal long-term potentiation (LTP) and learning were improved upon application of synthetic mouse or human Aβ<sup>42</sup> in picomolar concentrations (Puzzo et al., 2008, 2011; Garcia-Osta and Alberini, 2009), whereas high nanomolar Aβ, in the same experimental setting, led to well-established reduction of LTP suggesting a hormetic nature of Aβ on synaptic plasticity (Puzzo et al., 2008). Interestingly, these effects were sensitive to α-bungarotoxin, a selective antagonist of α7 nicotinic acetylcholine receptor (α7nAChR), and absent in α7nAChR knockout mice, which implies that functional α7nAChRs are required for Aβ42-induced neuroplasticity (Puzzo et al., 2008, 2011). This is consistent with the reported high-affinity binding of Aβ to α7nAChR (Wang et al., 2000a,b) and increased calcium influx through α7nAChRs in isolated hippocampal synaptosomes upon application of picomolar Aβ<sup>42</sup> (Dougherty et al., 2003). It has been proposed that the positive impact of Aβ<sup>42</sup> on neurotransmission is mediated by potentiating neurotransmitter release from presynapse (Puzzo et al., 2008; Abramov et al., 2009). Abramov et al. (2009) convincingly demonstrated modulation of presynaptic release probability in rat and mice hippocampal neuronal cultures treated with thiorphan (Th), an inhibitor of the rate-limiting peptidase neprilysin involved in the extracellular clearance of Aβ species. Later on, the same laboratory described the modulation of presynaptic release probability in the same rodent cultures upon application of picomolar amounts of human Aβ1–40 peptide (Abramov et al., 2009; Fogel et al., 2014). In contrast to ex vivo electrophysiological experiments in hippocampal slices and behavioral analyses (Puzzo et al., 2008, 2011), both studies in cultured cells argued against the contribution of α7nAChR to the effect of Th or Aβ1–40 on neurotransmission and proposed an alternative pathway involving APP homodimerization and signaling via heteromeric Gi/<sup>o</sup> proteins (Fogel et al., 2014).

Thus, it is unclear, whether different species of endogenous Aβ peptides exert the same effect on presynapse, what is the contribution of α7nAChRs, and what signaling connects putative Aβ receptors to the regulation of neurotransmitter release form presynapse. To address these questions, we tested systematically presynaptic effects of Th and rodent Aβ1–40 as well as Aβ1–42 in low to intermediate picomolar and low micromolar concentrations in cultured cortical neurons. To this end we visualized and quantified synaptic vesicle (SV) recycling within individual presynaptic boutons in living cells and investigated the contribution of α7nAChRs and their downstream signaling to the Aβ-mediated regulation of presynaptic function. Our data have potential implications for the pathophysiology of AD. Since Aβ modulates neurotransmission at very low extracellular concentrations, this physiological function would be directly affected already upon minor changes in extracellular Aβ levels occurring in early phases of AD and thus might contribute to cognitive impairments far before formation of amyloid plaques.

#### MATERIALS AND METHODS

#### Antibodies

For immunocytochemical stainings (ICC) and for Western blots (WB) following primary antibodies were used from rabbit: anti-CDK5 (WB 1:1000, C-8 Santa Cruz), anti-homer1 (ICC 1:1000, Synaptic Systems), anti-VGLUT1 (ICC 1:1000, Synaptic Systems), anti-VGAT (ICC 1:1000, Synaptic Systems), Lazarevic et al. Aβ Controls Synaptic Vesicles via α7nAChR

anti-VGAT lumenal domain Oyster550-labeled (ICC 1:200, Synaptic Systems), from mouse: anti-synaptotagmin1 lumenal domain Oyster550-labeled (ICC 1:250, Synaptic Systems), antiβ-tubulin isotype III (WB 1:2000, Sigma), anti-Aβ17–24 (4G8) (5 µg/ml, Signet), and from guinea pig: anti-synaptophysin (ICC 1:1000, Synaptic Systems). For ICC Alexa Fluor 488- (1:2000), Cy3- (1:2000) and Cy5- (1:1000) fluorescently labeled secondary antibodies were purchased from Jackson ImmunoResearch. For WB secondary antibodies labeled with Alexa Fluor 680 (1:20,000, ThermoFisher Scientific/Molecular Probes) and IRDye 800CW (1:20,000, Rockland) were used.

#### Chemical Reagents

Thiorphan (Th), FK-506 monohydrate, TMB8 and Choline chloride were purchased from Sigma-Aldrich. β-Secretase inhibitor IV, InSolution γ-Secretase inhibitor L-685, 458, InSolution Roscovitine, α-Bungarotoxin and Bafilomycin A1 from Calbiochem. Aβ1–42 and Aβ1–40 peptides, D-(-)-2-Amino-5-phosphonopentanoic acid (APV), 6-Cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX), BAPTA-AM and PNU 120596 from Tocris. Aβ was diluted according to the manufacturer's instruction. Th was diluted to 1 mM stock solution in artificial cerebrospinal fluid (ASCF) supplemented with 1 mM ascorbic acid to prevent Th oxidation (Iwata et al., 2001; Abramov et al., 2009). In all experiments control cells were treated with ascorbic acid in ASCF.

#### Animals

Breeding of animals and experiments using animal material were carried out in accordance with the European Communities Council Directive (2010/63/EU) and approved by the local animal care committees of Sachsen-Anhalt and Middle-Franconia/Germany.

#### Primary Neuronal Cultures

Primary cultures of cortical neurons were prepared as described previously (Lazarevic et al., 2011). In brief, rat embryos at day 18–19 after fertilization (E18–E19) were sacrificed by decapitation. The brains were removed and deprived of meninges. After treatment with 0.25% trypsin for 15 min and mechanical trituration cell suspension was plated in DMEM containing 10% fetal calf serum (FCS), 1 mM glutamine and antibiotics (100 U/ml penicillin, 100 µg/ml streptomycin) onto poly-D-lysine coated glass coverslips (Sigma, 18 mm diameter). Twenty-four hours after plating, the medium was exchanged for Neurobasal medium supplemented with B27 (Life Technologies), antibiotics, and 0.8 mM glutamine. The cells were maintained in a humidified incubator with 5% CO2. Primary hippocampal cultures were prepared according to a modified original protocol from Banker (1980) as described in Frischknecht et al. (2008). Briefly, rat embryos were sacrificed at E18–E19, brains were removed, hippocampi extracted and subjected to trypsin digestion and mechanical trituration. Thereafter, cells suspended in DMEM containing 10% FCS, 1 mM glutamine and penicillin/streptomycin were plated onto poly-L-lysine-coated glass coverslips. After 1 h, coverslips with primary hippocampal neurons were transferred into a Petri dish containing an astrocytic monolayer in Neurobasal medium supplemented with B27, antibiotics and glutamine, as described before, and placed in a humidified incubator. Neurons cultured for 18–21 days in vitro (DIV) were used for all analyses. For immunocytochemistry and Aβ<sup>42</sup> ELISA cells were plated on poly-D-lysine-coated glass coverslips at a density of 50,000 cells/12 mm coverslip in 24-well plates in 0.5 ml growth media. Aβ<sup>40</sup> ELISA was done on cells plated at density 100,000 cells/18 mm coverslips in 12-well plates in 1 ml of growth media. For imaging experiments cells were plated on poly-L-lysine-coated glass coverslips at a density of 30,000 cells/18 mm coverslip in a 60-mm Petri dish. For biochemical experiments cells were plated in 6-well plates at a density of 300,000 cells/well.

# ELISA Measurements of Aβ

The concentration of Aβ peptides in extracellular medium was assessed by sandwich ELISA using Human/Rat Aβ(42) high sensitive kit and Human/Rat Aβ(40) kit II purchased from Wako. The measurements were done according to manufacturer's protocol always using fresh medium.

### Immunocytochemistry and Synaptotagmin1 Luminal Domain Antibody Uptake

Neurons were fixed with 4% paraformaldehyde, 4% sucrose in PBS pH 7.4, for 3 min at RT. Prior to immunostaining, the cells were blocked and permeabilized with PBS containing 10% FCS, 0.1% glycine and 0.3% TritonX-100 for 30 min. Subsequently, primary antibodies were applied overnight at 4◦C. After three washing steps with PBS at RT coverslips were incubated with secondary antibodies for 1 h at RT. Both, primary and secondary antibodies were diluted in PBS containing 3% FCS. Coverslips were mounted on slides with Mowiol (Calbiochem) and kept at 4 ◦C until microscopic analysis. Synaptotagmin1 luminal domain antibody uptake was done as described (Lazarevic et al., 2011). Cells were briefly washed with freshly prepared Tyrode's buffer (119 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 30 mM glucose, 25 mM HEPES pH 7.4) and incubated with fluorescently-labeled syt1 antibody diluted in the same buffer, either for 20 min at 37◦C to monitor network activity driven uptake, or for 4 min at RT in Tyrode's buffer containing 50 mM KCl and 71.5 mM NaCl to assess evoked syt1 Ab uptake. Thereafter, samples were fixed and stained. In each experiment, at least two coverslips per treatment were processed in parallel. The results are representative of 2–5 independent experiments.

# Lentiviral Particles Production

The original mRFP-synaptophysin-pHluorin (sypHy) construct was obtained T. Oertner (Rose et al., 2013). cDNA of SypHy was inserted into FUGW backbone vector by standard cloning. Lentiviral particles were generated in HEK293T cell line (ATTC, Manassas, VA, USA) using FUGW-based transfer, psPAX2 packaging and pVSVG pseudotyping vectors (Lois et al., 2002). HEK293T cells were grown in media containing 10% FCS to 60% confluence in 75 cm<sup>2</sup> flasks. The cells were transfected using the calcium phosphate method (Fejtova et al., 2009). Molar ratio of FUGW: psPAX2: pVSVG was 2:1:1. Twenty-four hours later the content of FCS was reduced to 4%. After 48 h, the viruscontaining media was collected and cleared from large cellular debris by centrifugation for 20 min at 2000 g. The supernatant was aliquoted and stored at −80◦C until further use.

#### Synapto-pHluorin Imaging

SypHy was delivered to rat hippocampal neurons cultured for 4 DIV by the means of lentiviral infection and subjected to imaging at DIV17-18. Coverslips were incubated with Aβ1–42 or water as a control in a conditioned media for 1 h in the incubator. Coverslips were installed in an imaging chamber (Warner Instruments) and imaged at RT on an inverted microscope (Observer. D1; Zeiss) endowed with an EMCCD camera (Evolve 512; Delta Photometrics) controlled by MetaMorph Imaging (MDS Analytical Technologies). A 63× objective and GFP/mCherry single band exciters ET filter set (exciter 470/40, exciter 572/35, emitter 59022m, dichroic 59022BS) were used. Transduced neurons were identified by RFP expression. Neurons were stimulated in the presence of bafilomycin A1 (1 µM) to prevent vesicle reacidification and APV (50 µM) and CNQX (10 µM) to block recurrent network activity with 40 AP at 20 Hz, followed by 2 min recovery period. Afterwards, a stimulation with 900 AP at 20 Hz was delivered and a pulse of 60 mM NH4Clcontaining solution applied (Burrone et al., 2006). Electrical stimulation was delivered by a S48 stimulator unit (GRASS Technologies). A stream of images was acquired at 10 Hz and 5 s of the baseline was recorded before each stimulation, followed by imaging of the recovery phase for another 60 s. Synaptic boutons responsive to stimulation were selected by subtracting the first 10 frames of the baseline (before stimulation) from the 10 frames directly after the stimulus. Only neurons showing ≤20% increase in the fluorescence after NH4Cl application were considered as viable and metabolically active and included for the analysis. The mean IF intensities were measured in the circular regions of interest (ROIs with a diameter of 8 × 8 pixel) placed over each responding synapse using Time Series Analyzer V2.0 plugin in ImageJ and plotted after bleaching correction using GraphPad Software. The relative sizes of the RRP and the RP were expressed as fractions of the total sypHy-expressing pool detected after addition of NH4Cl. RP was quantified by averaging the mean of 50 values per each cell (representative of the frames 390–440 corresponding to time points 39–44 s on the XY graph). The results are representative of 3–5 independent experiments.

#### Image Acquisition and Analysis

Images were acquired with Zeiss Axio Imager A2 microscope with Cool Snap EZ camera (Visitron Systems) and MetaMorph Imaging software (MDS Analytical Technologies). For each pair of coverslips (treated vs. control) the same exposure time was taken. Per each experimental condition two coverslips were individually treated and processed. Images were captured from at least 3–5 visual fields (= cells) per coverslip and further analyzed using NIH ImageJ and OpenView software (Tsuriel et al., 2006). Upon appropriate background subtraction, immunoreactive puncta were counted along the 20 µm of proximal (≥10 µm and ≤50 µm distance from the cell body) or distal dendrite (≥50 µm distance from the cell body). The synaptic immunofluorescence intensities (IF) were assessed in a region of interest (ROI) set by the mask in the channel for synaptophysin (sph), which was used as synaptic marker. The mask was created semiautomatically using OpenView software.

#### Quantitative Western Blot

Control and treated 3 weeks old cortical neurons were washed with ice cold PBS and lysed in lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA) supplemented with Complete protease inhibitors (Roche), phosphatase inhibitor cocktail PhosStop (Roche) and Calpain inhibitor PD150606 (Tocris). Precleared cell lysates were mixed with SDS loading buffer. Equal amount of proteins was loaded on SDS-PAGE and electrotransferred to Millipore Immobilon-FL PVDF membranes. Membranes were incubated over night at 4◦C with primary antibodies and 1 h at RT with fluorescently labeled secondary antibody. Immunodetection and quantification was carried out using Odyssey Infrared Imagine System and Odyssey software v2.1 (LI-COR). After appropriate background subtraction all values were normalized using βIII-tubulin as a loading control.

# CDK5 Immunoprecipitation and Activity Assay

The kinase assay was performed as described in Crews et al. (2011). Briefly, 3 weeks old cortical neurons were washed with ice cold PBS and lysed in CDK5 IP buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA) supplemented with Complete protease inhibitors, phosphatase inhibitor cocktail PhosStop and Calpain inhibitor PD150606. Precleared cell lysates were subjected to immunoprecipitation using GammaBind Plus Sepharose beads (GE Healthcare) coupled with rabbit polyclonal antibody against CDK5 (C-8; Santa Cruz Biotechnology). After 3 h incubation at 4◦C, immunoprecipitates were washed three times with CDK5 IP buffer and resuspended in 50 µl of CDK5 kinase buffer (25 mM Tris pH 7.5, 10 mM MgCl2) in the presence of 90 µM ATP and 0.1 mM CDK5 substrate, Histone H1 (PKTPKKAKKL; sc-3066 Santa Cruz). Samples were then incubated for 30 min at 30◦C and the reaction was stopped by adding 50 µl of Kinase glo plus reagent (Promega). Luminescent signal was measured on FLUOstar Omega microplate reader (BMG Labtech).

#### Calcineurin Activity Assay

Calcineurin activity was assessed using calcineurin cellular activity assay kit (Calbiochem, Cat. No. 207007) according to the manufacturer's instructions. In brief, 3 weeks old primary cortical neurons were lysed in the buffer supplied by the manufacturer (25 mM Tris-HCl pH 7.5, 0.5 mM dithiothreitol, 50 µM EDTA, 50 µM EGTA, 0.2% Nonidet P-40). Upon removal of free phosphate samples were incubated with calcineurin substrate, RII phosphopeptide (DLDVPIPGRFDRRVpSVAAE), in the assay buffer containing 100 mM NaCl, 50 mM Tris-HCl (pH 7.5), 6 mM MgCl2, 0.5 mM CaCl2, 0.5 mM dithiothreitol, 0.05% Nonidet P-40. After 30 min incubation at 30◦C, reactions were terminated by adding 100 µl GREEN reagent. A<sup>620</sup> was measured using a microtiter plate reader. After background subtraction from each sample, the activity of calcineurin was determined as the difference between total phosphatase activity and the phosphatase activity in the EGTA containing buffer.

#### Statistics

All statistical analyses were performed with Prism 5 software (GraphPad Software). The normal distribution of the data was assessed using D'Agostino-Pearson omnibus test. Accordingly, parametric or non-parametric test was applied (as indicated in each experiment). In all graphs numbers within bars depict the numbers of analyzed individual visual fields (= cells) or independent wells/samples obtained from at least two different cell culture preparations. All data are always normalized to the mean of the control group and expressed as mean ± SEM. The level of statistical significance is indicated as <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 in all the graphs.

#### RESULTS

#### Th Modulates Recycling of SVs at Excitatory and Inhibitory Synapses

In order to investigate the role of endogenously secreted Aβ in the regulation of neurotransmitter release, we applied Th to the cultured rat cortical neurons at DIV18-21. Th is an inhibitor of metalloprotease neprilysin, which mediates the rate-limiting step in Aβ degradation (Iwata et al., 2001; Abramov et al., 2009). Treatment with Th for 1 h led to 1.6 and 1.2 fold elevation in the extracellular concentrations of Aβ<sup>42</sup> and Aβ40, respectively compared to untreated control, as assessed by sandwich ELISA of freshly collected neuronal growth media (**Figure 1A**; [Aβ42]: control: 57 ± 11 pM vs. Th: 94 ± 15 pM, **Figure 1B**; [Aβ40]: control: 130 ± 11 pM vs. Th: 153 ± 11 pM). Low pM concentrations of endogenously produced Aβ<sup>42</sup> and Aβ<sup>40</sup> peptides measured here as well as their Th-mediated elevation are in line with previous reports and comparable with concentrations obtained in vivo (Cirrito et al., 2005; Abramov et al., 2009).

To examine the effect of Th on presynaptic function, we monitored the efficacy of SV recycling at the level of individual synapses using synaptotagmin1 antibody (syt1 Ab) uptake (Kraszewski et al., 1995; Lazarevic et al., 2011). This assay utilizes a fluorophore-labeled antibody that recognizes a luminal domain of the SV protein syt1. This epitope becomes accessible for the antibody, added to neuronal media, only after fusion of SVs for neurotransmitter release before they undergo rapid compensatory endocytosis. Network activity-driven syt1 Ab uptake was significantly increased in cultures treated with Th compared to untreated control (**Figures 1C,D**; Th 138 ± 5% of control). Interestingly, application of Th led to an increased network activity-driven recycling of SV in both excitatory and inhibitory synapses as shown by quantification of syt1 Ab uptake in puncta immunoreactive for glutamatergic markers homer and vesicular glutamate transporter 1 (VGLUT1) or for a marker of inhibitory synapses, vesicular GABA transporter (VGAT; **Figures 1E,F**; Homer, Th: 146% ± 12%; VGLUT1, Th: 145 ± 10%; VGAT, Th: 162 ± 13% of control). Th-mediated increase of SV recycling in inhibitory synapses was further confirmed by performing uptake assay with an antibody against luminal domain of VGAT (Th: 151 ± 9% of control).

#### Extracellular Aβ**<sup>40</sup>** and Aβ**<sup>42</sup>** Exert Hormetic Effect on SV Recycling

To confirm that Th-induced changes in SV recycling rely on the modulation of the extracellular concentration of the endogenously secreted Aβ peptides, we treated neurons with Th in the presence or absence of 4G8 monoclonal antibody (5 µg/ml) that specifically binds Aβ peptides. While incubation with Th alone clearly increased SV recycling, this effect was completely prevented by chelation of extracellular Aβ using 4G8 antibody (**Figure 2A**; Th: 138 ± 5%; 4G8/Th: 110 ± 4% of control). Next, we inhibited production of Aβ, either by blocking β-secretase (BACE inhibitor IV, 0.5 µM) or γ-secretase (GAMMAinh, L-685, 458, 0.2 µM). Treatment with these inhibitors for 6 h led to a notable decrease of the syt1 Ab uptake, pointing to the role of endogenously released Aβ in the modulation of basal SV recycling (**Figure 2A**; βinh: 83 ± 7%; γinh: 73 ± 5% of control). Moreover, pre-treatment with secretase inhibitors, 5 h prior to the treatment with Th for 1 h, fully blocked the Th-induced increase in SV recycling (βinh/Th: 76 ± 6%; γinh/Th: 71 ± 6% of control). Taken together these experiments strongly support involvement of endogenously secreted Aβ in the Th-induced increase in the syt1 Ab uptake.

To strengthen this assumption, we added synthetic Aβ<sup>42</sup> and Aβ40, in physiological concentrations of 200 pM, to the growth media for 1 h, which caused a substantial increase in the syt1 Ab uptake (**Figures 2B,C**; Aβ40: 154 ± 9%; Aβ42: 157 ± 9% of control). In contrary, 1-h treatment with peptides from identical preparation at 1 µM concentration, widely used to induce neurotoxic effects (Park et al., 2013), had an opposite effect and led to a significant decrease in syt1 Ab uptake (**Figures 2B,C**; Aβ40: 69 ± 4%; Aβ42: 77 ± 3% of control). None of the treatments affected the density of synapses contacting proximal dendrites assessed as number of puncta immunoreactive for synaptic vesicle protein sph (**Figure 2D**; control: 43 ± 1; Th: 43 ± 1; 200 pM Aβ40: 42 ± 2, Aβ42: 41 ± 1; 1 µM Aβ40: 47 ± 3, Aβ<sup>42</sup> 38 ± 2 synapses). The number of active synapses assessed as sph puncta with over-threshold immunofluorescence for syt1 Ab uptake was decreased in cells treated with 1 µM Aβ<sup>42</sup> but unchanged in all other conditions (**Figure 2D**, control: 36 ± 1; Th: 39 ± 2; 200 pM Aβ40: 37 ± 3, Aβ42: 37 ± 2; 1 µM Aβ40: 36 ± 3, Aβ<sup>42</sup> 27 ± 2 active synapses). The percentage of active synapses calculated as proportion of active synapses out of sph positive synapses for each analyzed visual field differed between Th-treated cells and cells treated with 1 µM Aβ peptides further indicating that elevated physiological and supraphysiological concentrations of Aβ<sup>40</sup> and Aβ<sup>42</sup> exert opposite effects on presynaptic function (**Figure 2D**, control: 83 ± 2%; Th: 89 ± 4%; 200 pM Aβ40: 87 ± 3%, Aβ42: 89 ± 3%; 1 µM Aβ40: 79 ± 4%, Aβ<sup>42</sup> 73 ± 5%). These results are in line with a hormetic effect of Aβ<sup>40</sup> and Aβ<sup>42</sup> peptides, with low, physiological (high pM) concentration potentiating SV recycling, and high, supraphysiogical (low µM) concentration having a negative impact in the identical experimental readout.

To study durability and reversibility of Aβ-mediated effect on SV recycling, cells were treated with 200 pM Aβ<sup>42</sup> for 1 h prior to the washout and replacement with the conditioned medium. In this experiment, cells monitored 1 h after 200 pM Aβ<sup>42</sup> treatment showed a significant increase in syt1 Ab uptake. In contrast, cells assayed 23 h after the Aβ<sup>42</sup> washout showed recycling indistinguishable from untreated controls (**Figures 2E,F**; Aβ42: 141 ± 5%; Aβ42/washout: 113 ± 8% of control). This suggests that modulation of Aβ production and clearance might represent a physiological mechanism, inducing rapid changes in the recycling of SVs.

### Aβ Potentiates Basal Neurotransmission via an Increase in the Recycling Pool of SVs

The observed increase in the network activity-driven SV recycling points to changes in presynaptic properties, but could also reflect elevated network activity in Th- or Aβ-treated cultures. To test whether this effect relies on a modulation of presynaptic mechanisms, we performed syt1 Ab uptake during pulse-application of 50 mM KCl, which depolarizes neuronal membranes and induces release of all releasable SVs, i.e., recycling pool (RP) of SVs (Alabi and Tsien, 2012). Depolarization-induced syt1 Ab uptake was upregulated by 40% in Th-treated cultures (**Figures 3A,B**; Th: 142 ± 8% of control) suggesting the role of the endogenous Aβ in the regulation of the recycling pool of SV. Accordingly, application of 200 pM Aβ<sup>42</sup> exerted similar effect (**Figure 3B**; Aβ42: 137 ± 9% of control). Interestingly, high 1 µM concentration of synthetic Aβ<sup>42</sup> significantly reduced the depolarization–induced recycling confirming a hormetic nature of Aβ-mediated modulation of the presynaptic properties (**Figure 3B**; Aβ42: 78 ± 4% of control).

Since syt1 Ab uptake enables only end point measurement of SV recycling and fully relies on the endogenous expression of syt1, we sought to assess whether SV turnover was increased by an alternative method. To this end, we moved to hippocampal neurons, where the potentiating effect of Th, Aβ<sup>40</sup> and Aβ<sup>42</sup> was originally described by others and reproduced by us in the synapses contacting proximal dendritic segments (**Figures 3C,D**; Th proximal: 139 ± 13% of proximal control; Th distal: 119 ± 15% of distal control). We performed live imaging of SV turnover implementing genetically encoded pH-sensitive probe called synaptophysin-pHluorin (sypHy; Burrone et al., 2006; Granseth et al., 2006; Rose et al., 2013) expressed in

comparison test, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Scale bar, 5 µm.

FIGURE 3 | Picomolar Aβ increases the size of the functional recycling pool. (A) Representative images of depolarization-induced syt1 Ab uptake in control and Th-treated neurons. (B) Quantification of evoked syt1 Ab uptake after application of Th and pM or µM Aβ42. Numbers represent the number of cells used for analyses derived from three (Th) and two (Aβ) independent cultures. (C) Representative images of syt1 Ab uptake from control or Th-treated hippocampal neurons contacting proximal as well as distal dendritic segments. Proximal and distal areas where the boutons were analyzed are boxed. (D) Quantification of the data in (C). Values within columns correspond to the number of analyzed visual fields pooled from two independent cultures. (E) Averaged traces from hippocampal cells expressing sypHy and treated with control or 200 pM Aβ42. Intensities are normalized to the maximal NH4Cl response. (F) Mean values of the recycling pool. Numbers within columns represent the number of the analyzed cells from at least three independent cell culture preparations. Values are expressed as the mean ± SEM. Statistical significance was assessed by Kruskal-Wallis test followed by Dunn's multiple comparison test (B), unpaired two-tailed Student t test (D) or Mann Whitney test (F) <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Scale bars are10 µm in overview and 5 µm in insets.

cultured hippocampal neurons using lentiviral vectors. SypHy fluorescence is quenched at low, acidic pH found inside the SVs, but increases upon SV fusion and exposure of their lumen to the neutral pH of the extracellular media. The imaging was performed in the presence of bafilomycin, an inhibitor of the vesicular proton pump that prevents reacidification of SVs upon their compensatory endocytosis. To visualize readily releasable pool (RRP), which corresponds to the morphologically docked SVs, we delivered 40 AP at 20 Hz (Burrone et al., 2006). After 2 min pause to allow for the recovery of RRP, 900 AP at 20 Hz were delivered to release all releasable SVs (RP). SV refractory to stimulation (resting pool, RtP) were visualized by the alkalization of the SVs lumen using NH4Cl pulse, which enables accurate assessment of RP relative to all SVs present at the individual synapse (Burrone et al., 2006). In neurons treated with Aβ<sup>42</sup> no changes in RRP, but a significant increase of RP (**Figures 3E,F**; Aβ42: 112 ± 1% of control) and correspondingly a decrease in RtP were observed. These experiments ultimately demonstrate that Aβ modifies presynaptic function by the regulation of the turnover of presynaptic SV.

#### Role of Calcium Signaling via α7nAChR in the Aβ-Mediated Potentiation of SV Recycling

Previous studies proposed that, at picomolar concentrations, Aβ could bind and activate presynaptic α7nAChR (Wang et al., 2000b; Dougherty et al., 2003). Thus, we sought to determine the role of these receptors in the Aβ-mediated regulation of the presynaptic neurotransmitter release. Treatment with α-bungarotoxin (BgTx, 50 nM), a specific blocker of α7nAChRs, for 90 min did not significantly affect syt1 Ab uptake in untreated cells. However, the same treatment completely prevented increase in syt1 Ab uptake induced by Th- or 200 pM Aβ<sup>42</sup> application for 1 h, revealing a critical role of α7nAChR in the Aβ42-mediated regulation of SV cycling (**Figures 4A,B**; BgTx: 91 ± 5%; Th: 151 ± 11%; BgTx/Th: 109 ± 8%; Aβ42: 127 ± 6%; BgTx/Aβ42: 101 ± 9% of control). Interestingly, BgTx application was unable to completely block decreased syt1 Ab uptake mediated by supraphysiological concentration (1 µM) of Aβ<sup>42</sup> (**Figures 4A,B**; BgTx: 89 ± 6%; Aβ42: 68 ± 5%; BgTx/Aβ42: 81 ± 4%). This indicates that the effect mediated by higher concentrations of the peptide may involve not only α7nAChRs-dependent signaling, but also other types of either pre- or postsynaptic receptors (Lauren et al., 2009; Nikolaev et al., 2009). Next, we wondered whether activation of α7nAChR is sufficient to mimic Th-induced upregulation of syt1 Ab uptake. To this end, we applied choline (Ch, 500 µM), an agonist of α7nAChR (Alkondon et al., 1997), 20 min prior to Th or control treatment. Interestingly, choline had no effect on network activity-driven syt1 Ab uptake in control cells, however it completely blocked Th-induced presynaptic activity when applied together with Th (**Figure 4C**; Ch: 110 ± 9%; Th: 149 ± 9%; Ch/Th: 91 ± 5% of control). This might be due to the wellknown, fast desensitization of AChRs upon choline binding. To confirm this, we tested the impact of PNU120596, an allosteric modulator of α7nAChR, which has been shown to increase mean opening time of these receptors and thereby interfere with channel desensitization (Hurst et al., 2005). PNU120596 (3 µM, 1 h), had no effect on syt1 Ab uptake (**Figure 4C**; PNU: 114 ± 6% of control). Co-application of PNU120596 with choline fully mimicked Th-induced response (**Figure 4C**; PNU/Ch: 155 ± 11% of control) but PNU120596 was unable to further potentiate the effect of Th (**Figure 4C**; PNU/Th: 145 ± 15% of control), suggesting that endogenous Aβ might influence both channel opening and stabilization. Importantly, the Th-mediated increase in the depolarization-induced syt1 Ab

FIGURE 4 | Picomolar Aβ acts via α7nACh receptors. (A) Representative images of syt1 Ab uptake from Th-, Aβ<sup>42</sup> 200 pM or Aβ<sup>42</sup> 1 µM-treated neurons pre-incubated with BgTx to investigate the effect of α7nAChR. (B) Quantification of syt1 Ab uptake from (A). Numbers within columns indicate numbers of analyzed cells from two (Th; Aβ<sup>42</sup> 200 pM) or three (Aβ<sup>42</sup> 1 µM) cell culture preparations. (C) Statistical analysis of syt1 Ab uptake from cortical neurons treated with control, Th or/and ortho- and allosteric modulators of α7nACh receptors. (D) Quantification of depolarization-induced syt1 Ab uptake after control, Th and/or BgTx application. The numbers within bars (C,D) indicate the number of analyzed cells obtained from two separate cell culture preparations. In all graphs the values are expressed as the mean ± SEM. Statistical significance was evaluated by Kruskal-Wallis test followed by Dunn's multiple comparison test (B,C) or one-way ANOVA with Bonferroni post hoc test (D); <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Scale bar 5 µm.

uptake was completely precluded by preincubation of cells, for 30 min, with BgTx before Th application for further 1 h (**Figure 4D**; Th: 140 ± 7%; BgTx: 108 ± 8%; BgTx/Th: 100 ± 7% of control). Our data reveal that the Aβ-induced potentiation of presynaptic function relies on the modulation of α7nAChRs.

One of the most prominent features of α7nACh receptors is their high Ca2<sup>+</sup> conductance, which substantially contributes to synaptic Ca2<sup>+</sup> signaling. To investigate whether influx of extracellular Ca2<sup>+</sup> plays a role in the potentiation of SV recycling mediated by Th, we incubated both control and Th-treated cells in Ca2+-free medium for 1 h and subsequently performed live staining with syt1 Ab in usual imaging media containing 2 mM Ca2+. Incubation of cells in the Ca2+-free medium produced no significant effect on the network activity-driven syt1 Ab uptake (**Figure 5A**; Ctrl/noCa2+: 99 ± 8% of control) nor on the size of RP assessed by syt1 Ab uptake upon pulse application of 50 mM KCl (**Figure 5B**; Ctrl/noCa2+: 89 ± 4% of control). However, the impact of Th treatment on both network activity-driven and depolarization-induced syt1 Ab uptake was fully abolished in the Ca2+-free medium (**Figures 5A,B**; network activity driven syt1 Ab uptake: Th: 136 ± 8%; Th/noCa2+: 96 ± 7% of control; syt1 Ab uptake upon KCl depolarization: Th: 139 ± 8%; Th/noCa2+: 89 ± 4% of control).

α7nAChRs are also essentially involved in the activation of the calcium-induced calcium release pathway, which governs calcium efflux form cellular internal stores via stimulation of ryanodine receptors (RyR; Sharma and Vijayaraghavan, 2001; Dajas-Bailador et al., 2002). To test the involvement of this signaling in Th-mediated presynaptic strengthening, we applied a RyR blocker, TMB-8 (100 µM, 1 h), to the control and Th-treated cells and quantified their network activity-driven syt1 Ab uptake. The TMB-8-mediated block of Ca2<sup>+</sup> release from the intracellular stores efficiently impeded Th-induced effect on the SV recycling (**Figure 5C**; Th: 153 ± 11%; TMB-8/Th: 76 ± 5%; TMB-8: 107 ± 9% of control). Furthermore, application of cell-permeable calcium chelator, BAPTA-AM (10 µM, 1 h), also prevented the increase in SV recycling in Th-treated cells (**Figure 5D**; Th: 133 ± 8%; BAPTA-AM/Th: 89 ± 7%; BAPTA-AM: 76 ± 5% of control), suggesting contribution of calcium-dependent pathways in Aβ-driven presynaptic strengthening.

#### Involvement of CDK5/Calcineurin Balance in the Aβ-Mediated Regulation of SV Pools

In the recent years, cyclin dependent kinase 5 (CDK5) and phosphatase calcineurin emerged as two major players controlling the efficacy of neurotransmitter release by modulation of the size of RP of SVs (Kim and Ryan, 2010, 2013; Marra et al., 2012). Moreover, CDK5 and calcineurin are well described targets of calcium signaling and were already previously linked to downstream signaling via α7nAChR (Patrick et al., 1999; Lee et al., 2000; Stevens et al., 2003). To address potential contribution of CDK5 and calcineurin in Aβ-mediated alternations in SV recycling, we pharmacologically blocked these enzymes and carried out syt1 Ab uptake under high KCl conditions in which all recycling vesicles undergo exocytosis and are labeled. Application of Roscovitine (1 h, 50 µM), a potent CDK5 inhibitor (Meijer et al., 1997) led to a considerable increase of syt1 Ab uptake, yet combined treatment (Th/Roscovitine) did not exert any further effect (**Figure 6A**; Th: 136 ± 6%; Roscovitine: 161 ± 8%; Th/Roscovitine: 132 ± 7% of control). In contrast, inhibition of calcineurin by FK506 (1 µM, 1 h) induced a significant reduction in the depolarization–induced syt1 Ab uptake in both, control and Th-treated cells (**Figure 6B**; Th: 129 ± 7%; FK506: 86 ± 6%; Th/FK506: 105 ± 5% of control). These results are in line with possible involvement of CDK5/calcineurin signaling in Aβ-driven regulation of SV recycling pool. To further explore this hypothesis we performed CDK5 and calcineurin activity assay. Analysis of kinase activity confirmed that cells treated with Th or 200 pM Aβ42, show significant reduction in CDK5 activity (**Figure 6C**; Th: 80 ± 1%; Aβ42: 85 ± 1% of control), with no change in the total protein levels (**Figure 6D**). Furthermore, Th-induced effect was completely prevented when α7nAChR were blocked (**Figure 6C**; BgTx/Th:

101 ± 2% of control). On the other hand, a phosphatase activity assay revealed significantly higher calcineurin activity in the neurons treated with Th (**Figure 6E**; Th: 179 ± 14% of control), corroborating that balancing the activity of these enzymes has an important role in Aβ-driven regulation of the recycling vesicles. Altogether, these results support a view that Th-mediated regulation of SV turnover involves modulation of CDK5/calcineurin phosphohomeostasis downstream of α7nAChR receptors.

#### DISCUSSION

Despite the central role of Aβ peptide in the etiopathogenesis of AD, its physiological function in the healthy brain is still poorly understood. During the past decades, Aβ emerged as a vital factor that regulates neurotransmission and studies exploring effects of physiological i.e., low-intermediate picomolar concentrations of Aβ suggested presynaptic locus of its action. In this work, we directly tested the role of Th, Aβ<sup>40</sup> and Aβ<sup>42</sup> species in the regulation of neurotransmitter release from the presynapse and examined the underlying cellular signaling.

# Th and Endogenous Aβ<sup>40</sup> and Aβ<sup>42</sup> Converge on Regulation of SV Recycling

Using quantification of syt1 Ab uptake at levels of individual synapses, we confirmed previously reported effects of Th on SV recycling in cultured rat cortical and hippocampal neurons. The authors of the original publication argued that Th effect is based on the elevation of extracellular concentrations of endogenously produced Aβ species and provided numerous, yet indirect, evidence for their claim (Abramov et al., 2009). In a following study, presynaptic strengthening was induced in the same neuronal cultures by human Aβ40, which, however, differs from its murine ortholog in its N-terminal sequence (Fogel et al., 2014). Thus, the presynaptic effect of endogenous Aβ<sup>40</sup> and Aβ<sup>42</sup> has never been explicitly demonstrated and compared. Our experiments extend previous findings and show that Th and 200 pM Aβ<sup>40</sup> or Aβ42, in parallel experiments, exert the same effect, namely an increase in the turnover of SVs. All treatments could be efficiently blocked by inhibition of α7-nAChRs, which strongly speaks for their common mechanism of action. Interestingly, we observed smaller relative increase in the size of the recycling pool of SVs compared to the effect seen in the depolarization-driven syt1 Ab uptake assays, which might reflect differences in vesicle origin released by the chemical and electrical stimuli. Accelerated compensatory endocytosis of released SVs could also contribute to this effect and should be tested in future studies. In contrast to the previous studies that argued against any impact of physiological Aβ on the inhibitory synapses, we observed an Aβ treatment-induced increase in the network activity-driven recycling of SVs in the inhibitory synapses (Abramov et al., 2009). However, we cannot exclude that effect shown here (**Figure 1F**) simply reflects an increase in the overall network activity upon Th treatment and not direct regulation of release at inhibitory synapses.

# Calcium Influx via α7nAChR is Required for Aβ-Mediated Increase in SV Recycling

Involvement of α7nAChR in the modulation of neurotransmission by Aβ is controversial. The human and rodent Aβ42-induced memory enhancement and LTP increase were absent in the knock out for the α7 subunit of nAChR and Aβ42-induced LTP was also sensitive to the antagonists of α7nAChR BgTx and mecamylamine (Puzzo et al., 2008, 2011). In contrast, the presynaptic strengthening shown by imaging of activity-induced styryl dye destaining upon treatment with Th and human Aβ<sup>40</sup> was not sensitive to pharmacological block of these receptors (Fogel et al., 2014). In the experiments described here, the effect of Th and picomolar concentrations of Aβ<sup>42</sup> on network activity- and depolarization-driven SV recycling was fully prevented upon pretreatment with α7nAChR competitive antagonist-BgTx. In line with a requirement of the α7nAChR-mediated calcium influx for the Th-induced increase in SV recycling the effect of Th was absent upon depletion of extracellular and/or intracellular calcium and upon interference with calcium-induced calcium release from cellular internal stores. The effect of Th was also lost in the cells pretreated with choline, an agonist of α7nAChR that at high concentrations induces a rapid desensitization of the channel. Interestingly, a co-application of desensitizing concentration of choline and the allosteric modulator of α7nAChR PNU120596, known to increase opening time of agonist-bound receptor and decrease the receptor desensitization was comparable to the Th-mediated increase in SV recycling. Co-application of PNU120596 and Th did not further potentiate SV recycling suggesting that Th might affect receptor desensitization kinetics similarly as PNU120596. However, it is still unclear, whether Aβ<sup>40</sup> and Aβ<sup>42</sup> act as agonists or modulators of α7nAChR and what are the cofactors required for their action. While the impact of 200 pM Aβ<sup>40</sup> and Aβ<sup>42</sup> was hindered by pharmacological interference with α7nAChR, a blockage of these receptors did not fully prevent the effect of 1 µM Aβ42. This might imply that at higher concentrations Aβ<sup>42</sup> acts via different cell receptors. Nevertheless, displacement measurements of Aβ<sup>42</sup> and BgTx on radiolabeled α7nAChR suggested a competition of the two compounds on the same binding site with Aβ<sup>42</sup> having a 4000-fold higher affinity as compared to BgTx (Wang et al., 2000b). Therefore, our result might be also explained by an incomplete blockage of α7nAChR with BgTx in the presence of 1 µM Aβ42.

# Endogenous Aβ Modulates Recycling SV via Modulation of CDK5 and Calcineurin Activity

Quantification of depolarization-induced syt1 Ab uptake and sypHy imaging in living neurons shown here strongly implies that Th, Aβ<sup>40</sup> and Aβ<sup>42</sup> control neurotransmitter release via regulation of the recycling of SVs. Recently, CDK5 and calcineurin were suggested to govern the recycling of SVs by setting a balance in the local phospho- and dephosphorylation (Kim and Ryan, 2010, 2013; Marra et al., 2012). We have shown that application of Th or Aβ<sup>42</sup> leads to a rapid decrease of CDK5 activity in the cell lysates. This decrease in CDK5 activity required normal α7nAChR signaling substantiating their function upstream of CDK5 in the regulation of SV recycling by Aβ. Pharmacological inhibition of CDK5 by roscovitine mimicked Th-induced increase in the depolarization-induced syt1 Ab uptake and co-application of Th and roscovitine had no additive effect, suggesting that they share common pathways in regulation of the SV turnover. Th application increased the calcineurin activity and a pharmacological inhibition of calcineurin activity interfered with the Th-induced increase in the depolarization-induced recycling of SVs. At present, we can only speculate about signaling pathways connecting calcium influx via α7nAChR and regulation of calcineurin and CDK5 activity. Previously reported activation of calcineurin by calcium influx through α7nAChRs is compatible with our observations (Stevens et al., 2003). Cleavage of p35, an activator of CDK5, was observed upon application of micromolar Aβ<sup>42</sup> and resulted in a formation of CDK5/p25 hyperactive complex (Patrick et al., 1999; Lee et al., 2000). This scenario contradicts our results that demonstrate calcium influx-induced decrease of CDK5 activity. It is possible that a slight elevation of intracellular calcium upon Aβ-mediated activation of α7nAChR has different consequences for CDK5 regulation than a massive calcium influx induced by a prolonged application of micromolar Aβ42.

Taken together, our data support the function of endogenous Aβ species in the regulation of neurotransmitter release. The described modulation of presynaptic function by Aβ was rapid and reversible. Moreover, a depletion of endogenous Aβ upon interference with its production and application of elevated Aβ concentrations led to a decrease in the synaptic strength, which is in accordance with previously proposed hormetic regulation of neurotransmission by endogenous Aβ (Puzzo et al., 2008; Abramov et al., 2009). Thus, our results corroborate on function of endogenous Aβ as a physiological regulator of presynaptic efficacy. Compellingly, the fast modulation of release by Aβ might act in tuning of synaptic strength at the level of individual synapses in processes of synaptic plasticity. The observation that intracellular signaling cascades, shown here to be involved in the physiological regulation of SV recycling by Aβ, are also implicated in AD pathogenesis supports the hypothesis that failure of the physiological function of Aβ in the tuning of SV recycling could impair synaptic homeostasis and initiate synaptic dysfunction leading to cognitive decline and neurodegeneration in AD.

#### AUTHOR CONTRIBUTIONS

VL and AF conceptualized the study and curated the data. VL, SF, MA-A and DA performed investigations, formal analysis,

#### REFERENCES


methodology development and validation. CM-V, DI, MAC provided methodologies. AF and EDG provided infrastructure and resources. VL, SF and MA-A executed visualization of data. AF supervised the study. VL, SF and AF have written original draft. All authors reviewed and edited the final manuscript.

#### FUNDING

This study was supported by grants from following agencies: Deutsche Forschungsgemeinschaft (FE1335/1 and SFB779/A06), People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme FP7/2007-2013/ NPlast under REA grant agreement no. [289581], Federal State of Saxony-Anhalt and the European Regional Development Fund (ERDF) (CBBS, FKZ: ZS/2016/04/78120), Leibniz Association (LGS Synaptogenetics, SAW 2013-15 and SAW 2014-2016). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

#### ACKNOWLEDGMENTS

We thank I. Herbert, K. Hartung, B. Kracht and S. Müller for help with primary neuronal cultures, T. Oertner for the sypHy construct, N. Ziv for providing the OpenView software, P. Lewczuk for recommendations on ELISA measurements and R. Frischknecht and all lab members for conceptual discussions.

in models of Alzheimer's disease. Cell Death Dis. 2:e120. doi: 10.1038/ cddis.2011.2


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Lazarevic, Fienko, Andres-Alonso, Anni, Ivanova, Montenegro- ´ Venegas, Gundelfinger, Cousin and Fejtova. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# APP—A Novel Player within the Presynaptic Active Zone Proteome

Jens Weingarten<sup>1</sup> , Melanie Weingarten<sup>1</sup> , Martin Wegner <sup>2</sup> and Walter Volknandt <sup>1</sup> \*

1 Institute for Cell Biology and Neuroscience, Biologicum and BMLS, Goethe University, Frankfurt am Main, Germany, <sup>2</sup>Department of Molecular Bioinformatics, Goethe University, Frankfurt am Main, Germany

The amyloid precursor protein (APP) was discovered in the 1980s as the precursor protein of the amyloid A4 peptide. The amyloid A4 peptide, also known as A-beta (Aβ), is the main constituent of senile plaques implicated in Alzheimer's disease (AD). In association with the amyloid deposits, increasing impairments in learning and memory as well as the degeneration of neurons especially in the hippocampus formation are hallmarks of the pathogenesis of AD. Within the last decades much effort has been expended into understanding the pathogenesis of AD. However, little is known about the physiological role of APP within the central nervous system (CNS). Allocating APP to the proteome of the highly dynamic presynaptic active zone (PAZ) identified APP as a novel player within this neuronal communication and signaling network. The analysis of the hippocampal PAZ proteome derived from APP-mutant mice demonstrates that APP is tightly embedded in the underlying protein network. Strikingly, APP deletion accounts for major dysregulation within the PAZ proteome network. Ca<sup>2</sup><sup>+</sup>-homeostasis, neurotransmitter release and mitochondrial function are affected and resemble the outcome during the pathogenesis of AD. The observed changes in protein abundance that occur in the absence of APP as well as in AD suggest that APP is a structural and functional regulator within the hippocampal PAZ proteome. Within this review article, we intend to introduce APP as an important player within the hippocampal PAZ proteome and to outline the impact of APP deletion on individual PAZ proteome subcommunities.

#### Edited by:

Thomas Deller, Goethe-University, Germany

#### Reviewed by:

Imre Vida, Charité Universitätsmedizin Berlin, Germany Eckart D. Gundelfinger, Leibniz Institute for Neurobiology, Germany

#### \*Correspondence:

Walter Volknandt volknandt@bio.uni-frankfurt.de

Received: 24 November 2016 Accepted: 07 February 2017 Published: 20 February 2017

#### Citation:

Weingarten J, Weingarten M, Wegner M and Volknandt W (2017) APP—A Novel Player within the Presynaptic Active Zone Proteome. Front. Mol. Neurosci. 10:43. doi: 10.3389/fnmol.2017.00043 Keywords: amyloid precursor protein, hippocampus, neuronal network, presynaptic active zone, synapse

## INTRODUCTION

The development of a neuronal circuit requires precise coordination of billions of neurons, with up to 100,000 synaptic connections each, forming a stable but plastic network and persisting over the lifespan of an organism (Turrigiano, 2008, 2012). The key word deciphering this phenomenon is ''homeostasis'' and was introduced by Walter Cannon in the early 1930's (Cannon, 1932). Within the neuronal network numerous homeostatic mechanisms ensure physiological activity in a spatio-temporal manner on various groups of synapses (Turrigiano, 2008; Yu and Goda, 2009). Maintenance of synaptic homeostasis demands on a coordinated proteomic response at both—pre- and postsynaptic sites (Schanzenbächer et al., 2016).

**Abbreviations:** AD, Alzheimer's disease; APLP1/2, amyloid precursor like proteins1 and 2; APP, amyloid precursor protein; CA, cornu ammonis; HD, Huntington's disease; PAZ, presynaptic active zone; PD, Parkinson's disease; SNAP25, synaptosomal associated protein 25; SV2, synaptic vesicle protein 2; VAMP2, vesicle associated membrane protein2/synaptobrevin2.

Consecutive steps of processing the arrival of an action potential into a chemical signal by recruiting a subset of individual proteins that fuse synaptic vesicles with the presynaptic plasma membrane, release of their neurotransmitter into the synaptic cleft, which further react with their specific receptor at the postsynapse (**Figure 1**), demands on a rather stringent progression. Key players within this network comprise prominent candidates like synaptic vesicle protein 2 (SV2), synaptotagmin-1, synaptosomal associated protein 25 (SNAP25), syntaxin and vesicle associated membrane protein2/synaptobrevin2 (VAMP2; Südhof and Rizo, 2011; Südhof, 2012; Laßek et al., 2015). The unique set of proteins regulating, mediating and controlling proper presynaptic physiology was recently complemented by a yet unappreciated companion—the amyloid precursor protein (APP).

#### APP—A BRIEF PROFILE

The APP belongs to an evolutionary conserved gene family with specific expression pattern in C. elegans, Drosophila and mammals (reviewed in Coulson et al., 2000; Jacobsen and Iverfeldt, 2009). Discovered during the 1980s as precursor protein of A-beta (Aβ)—the main constituent of senile plaques—much effort has been made to understand the pathophysiology of Alzheimer's disease (AD) and the physiological function of APP (Glenner and Wong, 1984; Kang et al., 1987). The progression of AD is characterized by a massive loss of synapses especially within the hippocampus. Extracellular senile plaques and intracellular neurofibrillary tangles induce and promote successive degeneration of neurons manifested by severe impairments in learning and memory and behavioral changes (Grundke-Iqbal et al., 1986; Supnet and Bezprozvanny, 2010). Enzymatic processing of APP is initiated by either β-secretase/γ-secretase (amyloidogenic pathway) cleavage, or α-secretase/γ- secretase (non-amyloidogenic pathway). The proteolytic processing of amyloid precursor like proteins 1 and 2 (APLP1 and APLP2) is comparable to that of APP, however, only the amyloidogenic pathway can induce the formation of Aβ-peptides (Eggert et al., 2004). Until now, little is known about the shift in enzymatic processing of the APP leading to

Frontiers in Molecular Neuroscience | www.frontiersin.org February 2017 | Volume 10 | Article 43

acidic domain highlighted in blue). (C) Schematic cartoon of a chemical synapse highlighting APP at the PAZ.

the accumulation of Aβ-peptides and the formation of oligomers and fibrils. Since senile plaques consist mainly of Aβ-fibrils, it was of great interest how these structures are organized (Lu et al., 2013). The analysis of these fibrils derived from AD patients revealed an individual molecular structure. These variations were suggested to correlate with the severity of impairments in the individual pathogenesis of AD in patients (Lu et al., 2013).

APP is a type 1 transmembrane protein with a large N-terminal domain, a single transmembrane region and a short C-terminal domain (**Figure 1**). The N-terminal domain is subdivided into an E1 domain comprising a heparinbinding/growth factor-like domain (HBD/GFLD), a copper and zinc-binding domain (CuBD/ZnBD) followed by an acidic region (DE), optionally a KPI-domain (not present in the neuronal specific isoform APP695), and an E2 domain consisting of a second HBD (HBD2) a collagen-binding region and N-glycosylation binding sites (Jacobsen and Iverfeldt, 2009; Kaden et al., 2012). The APP intracellular domain (AICD) contains the highly conserved YENPTY motif involved in the internalization of APP and phosphorylated or dephosphorylated tyrosine mediated binding of adaptor proteins like FE65 (**Figure 1**), Dab1 and X11a (munc-18 interacting protein, Mint; Jacobsen and Iverfeldt, 2009). All APP family members reveal a high structural overlap except the Aβ domain that is only present in mammalian APP.

#### APP AT THE SYNAPSE

Multiple isoforms were described for mammalian APP (e.g., 695aa, 770aa), but only APP695aa is expressed in neurons. Within neurons, APP was discussed as bona fide SV (Groemer et al., 2011) and constituent of the presynaptic plasma membrane (Marquez-Sterling et al., 1997; Lyckman et al., 1998). In addition, APP was described as a constituent of endocytosed synaptic vesicles, but being sorted away from bona fide synaptic vesicles (Marquez-Sterling et al., 1997). On the contrary, Groemer et al. (2011) reported a small copy number of APP to synaptic vesicles as a result of endosomal synaptic vesicle recycling processes. However, they emphasized that the majority of APP was immunodetected in fractions containing the plasma membrane, and only a small amount was present in purified synaptic vesicle fractions (Groemer et al., 2011). In our studies, we clearly demonstrated that APP and its family members are constituents of the presynaptic plasma membrane and that APP is virtually absent from synaptic vesicles (Laßek et al., 2013). Within the presynaptic nerve terminal, a small section characterized by an assembly of electron dense material, is responsible for Ca2+-triggered exocytosis of synaptic vesicles. This section is called presynaptic active zone (PAZ; Gray, 1963; Südhof, 2012). The composition of the PAZ proteome identified the release site as dynamic focal hot spot, providing the prerequisite for structural and functional changes also in the adult nerve terminal. Neuronal communication and signal transduction depends not only on the concerted action of individual proteins within the PAZ but also on proper energy supply (Boveris and Navarro, 2008). Besides the glycolytic chain associated with synaptic vesicles (reviewed in Burré and Volknandt, 2007), mitochondria are the main source for the production of ATP at the presynaptic terminal. Therefore, mitochondria are essential in maintaining presynaptic homeostasis and phosphorylation reactions and are highly involved in synaptic plasticity (reviewed in Mattson et al., 2008). The allocation of APP to the proteome of this highly dynamic substructure of the presynapse, identified APP as yet unknown player within the neuronal communication and signaling network (Laßek et al., 2013, 2016).

To address the question which physiological function APP is executing in the central nervous system (CNS), a variety of genetically designed mouse models has been generated (Heber et al., 2000; Ring et al., 2007; Hick et al., 2015). It turned out that loss of APP causes an age-dependent phenotype with no severe physiological impairments in younger mice but impairments in learning and memory in the elderly (Phinney et al., 1999; Ring et al., 2007). At postsynaptic sites, reduced dendritic length and branching accompanied by a total spine density reduction was characteristic for old APP-KO mice and indicates a physiological role of APP in maintaining spine density (Tyan et al., 2012). This was further supported by Weyer et al. (2014) demonstrating a specific role of APP in sustaining spine structure and density. Classification of spine structure can be morphologically addressed revealing stubby, thin and mushroom spines. In APP-KO cornu ammonis 1 (CA1) neurons this spine subtype distribution is altered by a significant decrease in mushroom spines (Weyer et al., 2014). Interestingly, substantial changes of the proteomic composition of neurotransmitter release sites are already detectable in younger mice (Laßek et al., 2014, 2016). Since APP plays an essential role during the development of the neuronal circuit (Lazarov and Demars, 2012), it was suggested that the APLP2 compensates for the loss of APP (Weyer et al., 2011; Laßek et al., 2016). Screening immunopurified PAZs derived from individual total mouse brain revealed prominent players to be affected by APP deletion. Candidates like SV2A, synaptotagmin-1 and synaptophysin turned out to be differentially regulated. It is worth mentioning, that the opposite effect was observed for deletion of either APLP1 or APLP2 (Laßek et al., 2014). Moreover, deletion of the family members did not result in any morphological alterations in CNS or overall impairments in learning and memory (Heber et al., 2000; Weyer et al., 2011). Memory formation requires a variety of network oscillations that are regularly synchronized between hippocampal CA1 and CA3 region (Korte et al., 2012). As inhibitory interneurons play an essential role in this coordinated action of synchronization, their oscillations can affect a large population of pyramidal neurons, inhibiting specific input pathways and guarantee for a high background-to-noise ratio (Mann and Paulsen, 2007). Accompanied with the observed shift in the excitatory-inhibitory ratio in APP-KO mice, it was suggested that GABAA receptormediated inhibition is altered in aged APP-KO mice and that these changes contribute to the reduction in LTP in aged APP-KO mice. This assumption was further sustained by LTP-rescue experiments employing pharmacological blockade of GABA<sup>A</sup> receptors (Fitzjohn et al., 2000; Korte et al., 2012). Synaptic plasticity requires persistent changes within the entire network. Thereby, the strength of a neuronal connection is individually adjusted (up or down) dependent on homeostatic synaptic scaling (Turrigiano, 2012; Davis, 2013). Homeostasis implies the capability to restore individual baseline functions upon continued input. This is achieved by controlling and modulating the expression and trafficking of specific proteins and protein complexes. Initially, synaptic scaling was described as bidirectional modulation of neurotransmitter receptor abundance at individual synapses. In this context, it was suggested that this effect stabilizes neuronal excitability while sustaining learning-related information (Turrigiano, 2008, 2012; Davis, 2013). At presynaptic sites, homeostasis encompasses the fast, long-lasting and accurate modulation of synaptic vesicle fusion (Davis, 2013). Alteration at protein level as response to homeostatic scaling in hippocampal neurons was recently analyzed by Schanzenbächer et al. (2016). They uncovered the necessity of new protein synthesis upon upor down scaling induced by pharmacological treatment. More than 300 proteins (e.g., neurotransmitter receptors, scaffolding and signaling proteins) were affected by this stimulation. Strikingly, genes affected by the stimulation, encode for proteins critically involved in neurological diseases like AD, Parkinson's disease (PD) or schizophrenia. Proteins identified and regulated by homeostatic scaling in this approach provide a starting point to examine how their dysregulation might contributes to a variety of neuronal disorders (Schanzenbächer et al., 2016).

#### APP AND THE HIPPOCAMPUS

APP is functionally integrated into the hippocampal PAZ proteome and fits into the evolutionary conserved active zone protein complex, comprising prominent constituents like ELKS, CASK bassoon, RIM and Munc18 (Südhof, 2012; Laßek et al., 2016). Embedding APP into the entire PAZ proteome unraveled APP as a context-sensitive regulator with impact on synaptic vesicle cycle, cytoskeletal organization and Ca2+-homeostasis. Deletion of APP significantly affects those proteins serving as mediator (e.g., α-synuclein) within the PAZ but not their central players (e.g., SNARE-machinery). It was obvious, that loss of APP accounts for individual rearrangements of the entire network structure with no current effect on presynaptic functionality (Laßek et al., 2016). Interestingly, these massive alterations in protein abundance within the PAZ proteome did not account for impairments in learning and memory pointing to a yet unknown compensatory mechanism in young APP-KO mice (Ring et al., 2007). The most important guarantors for sufficient energy metabolism, calcium- and redox homeostasis are mitochondria (Yin et al., 2014; Grimm et al., 2016). They support the intracellular energy demand by producing ATP, affect redox-sensitive kinases via second messengers H2O<sup>2</sup> and NO and regulate the NAD+/NADH homeostasis, involved in maintenance of mitochondrial energy statues (Yin et al., 2014). A recent proteome study on young and old APP-KO mice revealed drastic changes in mitochondrial protein abundance at hippocampal neurotransmitter release sites. These results indicated that old APP-KO mice display a dysregulation in their bioenergetics metabolism accompanied by hyperphosphorylation of CaMKII (Laßek et al., 2017). It is tempting to speculate that during the induction of LTP CaMKII becomes over-activated, which has a negative impact on synaptic plasticity, and prevents proper learning and memory consolidation. Recently, over-activation of CaMKII was described in hippocampal neurons following synaptic stimulation and increased intracellular Ca2+-levels. As a kind of protective mechanism, CaMKII is able to form clusters (spherical clusters, identical in size and shape) preventing excessive protein phosphorylation, independent of the autocatalytic center, due to an imbalance in Ca2+-homeostasis (Dosemeci et al., 2007). Therefore, cognitive impairments in old APP-KO mice might be associated with an imbalance in mitochondrial function and phosphorylation-activity of the serine/threoninespecific kinases CaMKII as observed during the progression of AD (Grimm et al., 2016). Thermodynamic imbalance and compensatory mechanisms acting in impaired neurons, will further induce a competition for energy substrates and finally shift formerly healthy neurons into affected ones (Demetrius et al., 2015).

The expression pattern of APP in the hippocampus and especially at PAZs has been further analyzed in detail by Rodrigues et al. (2014), demonstrating that APP is most abundant on glutamatergic neurotransmitter release sites as compared to GABAergic ones. Their findings further revealed that less than half of hippocampal synapses were immunopositive for APP (Rodrigues et al., 2014). Strikingly, deletion of APP accounts for an increase in activity of GABAergic synapses. This dysregulation in balance between inhibitory and excitatory neurons was induced by a diminished endocytosis of VDCC in GABAergic hippocampal neurons (Yang et al., 2009). In this context it is worth mentioning the idea of graded molecular profiles of hippocampal neurons (here CA1 neurons) stated by Cembrowski et al. (2016). The hippocampal formation has been attributed with regional-specific functions. Whereas the dorsal hippocampus is known to be associated with cognitive functions (like memory and spatial navigation) the ventral region is basically associated with behavior (like stress and emotion; Fanselow and Dong, 2010; Strange et al., 2014). Illustrating the gene expression profile along the dorsal-ventralaxis of CA1 neurons revealed unique profiles of decay. These findings make it rather interesting to figure out how those CA1 neurons perform their region-specific functions. In line with these findings similar considerations were made for the diversity of presynaptic performance. Atwood and Karunanithi (2002) described various models of functional differentiation of presynaptic neurons. (1) Different amounts of strengthregulating presynaptic proteins or a variable combination of more than one presynaptic protein can be induced or attracted by a postsynaptic neuron. (2) Neuronal activity or specific input to presynaptic neurons can induce a differential occurrence of presynaptic proteins in different neurons (Atwood and Karunanithi, 2002). However, if the molecular profile and physiology of those neurons is different, what about the profile of individually expressed proteins? The abundance of several PAZ proteins differs considerably between brain regions, presumably reflecting region-specific functional adaptions. This can be of vital importance to understand the impact of therapeutic drugs (e.g., prevalence or therapy of AD) on their targets and to elucidate their subsequent effects on the PAZ proteome. The identification of individual PAZ protein components is a prerequisite for further functional investigations and also provides a solid basis for evaluating their interaction. Therefore, differences in the PAZ proteome reflect specific adaptions to regional neuronal circuitries and the functional and structural dynamics of their corresponding release sites (Weingarten et al., 2015). Salient findings by Schwenk et al. (2014) are in accordance with this dynamic and functional diversity of proteomes. They demonstrated that a large multiprotein complex provides an individual assembly of its core-subunits and a regional specific architecture over space and time (Schwenk et al., 2014). This perspective is indispensable for those proteins sharing a differential expression pattern (like APP)—not only in specific brain regions but also in neurons and individual synapses, respectively. A current study by Counts et al. (2014) went a step further, performing molecular profiling of CA1 neurons derived from patients with mild cognitive impairment (MCI) and AD—MCI, is a prodromal stage of AD. It is widely accepted that early pathological events triggering the outcome of AD are associated with CA1 neurons. Compared to control groups (NCI, no cognitive impairments), expression of genes involved in proper synaptic function in CA1 neurons, is severely dysregulation in MCI, whereas no further changes were observed in AD. Interestingly, APP,

REFERENCES


APLP1 and APLP2 transcripts were not altered at any stage in CA1 neurons. Molecular profiling of CA1 neurons revealed that early changes in synaptic elements provide susceptibility to cognitive decline in aged patients. These findings point to an early onset of synaptic failure that becomes manifested in the dysregulation of the hippocampal neuronal circuit (Counts et al., 2014).

#### CONCLUDING REMARKS

Our proteomic profiling of PAZs derived from total mouse brain revealed a summary of all alterations due to loss of APP. Going a step further, looking only at the hippocampus, the picture of presynaptic changes was impressively refined. Therefore, the next step should include proteomic studies addressing the molecular profiling of individual neurons within the hippocampus. Combining our approaches with new technologies will provide novel insights into the biological function of APP within the CNS. Moreover, interdisciplinary approaches and sustained exchanges of information can facilitate new perspectives within the challenging APP research field.

#### AUTHOR CONTRIBUTIONS

The authors MWei, JW, MWeg and WV contributed equally to this review article.

#### ACKNOWLEDGMENTS

We are grateful to Herbert Zimmermann for valuable suggestions.


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a shared affiliation, though no other collaboration, with the authors JW, MWei, MWeg and WV, and the handling Editor states that the process met the standards of a fair and objective review.

Copyright © 2017 Weingarten, Weingarten, Wegner and Volknandt. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# APP Protein Family Signaling at the Synapse: Insights from Intracellular APP-Binding Proteins

Suzanne Guénette<sup>1</sup> \*, Paul Strecker <sup>2</sup> and Stefan Kins <sup>2</sup> \*

1 Independent Researcher, 84 3099 Uplands Dr., Ottawa, ON, Canada, <sup>2</sup>Department of Biology, Division of Human Biology, University of Kaiserslautern, Kaiserslautern, Germany

Understanding the molecular mechanisms underlying amyloid precursor protein family (APP/APP-like proteins, APLP) function in the nervous system can be achieved by studying the APP/APLP interactome. In this review article, we focused on intracellular APP interacting proteins that bind the YENPTY internalization motif located in the last 15 amino acids of the C-terminal region. These proteins, which include X11/Munc-18-interacting proteins (Mints) and FE65/FE65Ls, represent APP cytosolic binding partners exhibiting different neuronal functions. A comparison of FE65 and APP family member mutant mice revealed a shared function for APP/FE65 protein family members in neurogenesis and neuronal positioning. Accumulating evidence also supports a role for membrane-associated APP/APLP proteins in synapse formation and function. Therefore, it is tempting to speculate that APP/APLP C-terminal interacting proteins transmit APP/APLP-dependent signals at the synapse. Herein, we compare our current knowledge of the synaptic phenotypes of APP/APLP mutant mice with those of mice lacking different APP/APLP interaction partners and discuss the possible downstream effects of APP-dependent FE65/FE65L or X11/Mint signaling on synaptic vesicle release, synaptic morphology and function. Given that the role of X11/Mint proteins at the synapse is well-established, we propose a model highlighting the role of FE65 protein family members for transduction of APP/APLP physiological function at the synapse.

#### Edited by:

Thomas Deller, Goethe University Frankfurt, Germany

#### Reviewed by:

Lisa M. Munter, McGill University, Canada Thorsten Müller, University Hospitals of the Ruhr-University of Bochum, Germany

#### \*Correspondence:

Suzanne Guénette suzanne.guenette@gmail.com Stefan Kins s.kins@biologie.uni-kl.de

Received: 23 December 2016 Accepted: 13 March 2017 Published: 30 March 2017

#### Citation:

Guénette S, Strecker P and Kins S (2017) APP Protein Family Signaling at the Synapse: Insights from Intracellular APP-Binding Proteins. Front. Mol. Neurosci. 10:87. doi: 10.3389/fnmol.2017.00087 Keywords: amyloid precursor protein (APP), FE65, FE65L1, X11/Mint proteins, synaptic signaling

# INTRODUCTION

The amyloid precursor protein (APP) can be processed to generate the amyloid β (Aβ) peptides, which aggregate to form senile plaques, one of the major pathological hallmarks found in Alzheimer's disease (AD; Masters and Selkoe, 2012). APP is a ubiquitously expressed type I transmembrane protein with a large ectodomain, a single membrane spanning domain, and a short cytoplasmic tail. The ectodomain comprises two highly conserved E1 and E2 domains, involved in metal (copper and zinc) and heparin binding (Baumkötter et al., 2012; Müller and Zheng, 2012).

APP has important physiological functions at the synapse (Zheng and Koo, 2011). Aged mice deficient in APP show impairments in behavior (Müller et al., 1994; Phinney et al., 1999; Ring et al., 2007), long-term potentiation (LTP; Seabrook et al., 1999; Ring et al., 2007), dendritic branching and synaptic density (Zheng et al., 1995; Dawson et al., 1999; Phinney et al., 1999; Seabrook et al., 1999; Lee et al., 2010; Tyan et al., 2012; Weyer et al., 2014). No synaptic deficits are present in APP-like protein 2 (APLP2) knockout (KO) mice (Midthune et al., 2012; Weyer et al., 2014). Yet mice doubly deficient for APP and APLP2 or for APLP1 and APLP2 exhibit early postnatal lethality and show deficits in neuromuscular junction (NMJ) formation, including incorrect apposition of pre- and postsynaptic sites (von Koch et al., 1997; Heber et al., 2000; Wang et al., 2005; Klevanski et al., 2014). These data suggest genetic redundancy of APP family members for synapse formation.

Interestingly, the introduction of either sAPPα (APPsα-KI) or APP with a mutation in the intracellular domain (APPY682G) onto an APLP2-deficient background produced a partial rescue of the phenotypes presented in the doubly deficient mice (Barbagallo et al., 2011; Weyer et al., 2011). In addition, it was shown that learning deficits in Drosophila lacking APPL, the only homolog of APP in the fruit fly, are partially rescued by secreted sAPPL (788 amino acid soluble N-terminal fragment) or a non-cleavable full-length APPL (Bourdet et al., 2015; Cassar and Kretzschmar, 2016). These data indicate that APP function depends on both the activity of secreted sAPP, likely functioning as a ligand, and on full-length APP, possibly working as a receptor or co-receptor. Interestingly, exogenous Adenovirusmediated expression of sAPPα in aged AD model transgenic mice (APPswe/PS1∆E9) restored synaptic plasticity and partially rescued spine density deficits (Fol et al., 2016). These data, along with those from many other studies, suggest that sAPPα may function as a neurotrophic factor (Meziane et al., 1998; Bour et al., 2004; Taylor et al., 2008; Claasen et al., 2009; Weyer et al., 2014; Hick et al., 2015; Kundu et al., 2016; Plummer et al., 2016). Although many different extracellular binding partners of APP are reported, including different heparin sulfate proteoglycans (HSPG; Aydin et al., 2012; Reinhard et al., 2013), none of the identified proteins have been reported to function as sAPP receptors. In the case of full length APP, it was proposed that APP might be involved in trans-synaptic signaling, similar to other synaptic modulators such as Neuroligin, Neurexin and LRRTMs (Siddiqui and Craig, 2011; Baumkötter et al., 2012). Several studies provide experimental evidence consistent with this notion. Dimerization of APP can occur in a trans-orientation (Soba et al., 2005; Kaden et al., 2008; Wang Z. et al., 2009; Baumkötter et al., 2012; Klevanski et al., 2014) and inactivation of APP at either the pre- or postsynaptic sites of the NMJ in APLP2 KO mice causes defects similar to the combined germline deletions of APP and APLP2 (Wang Z. et al., 2009). Moreover, expression of APP bearing an intact E1 domain in human embryonic kidney cells co-cultured with primary hippocampal neurons promotes the presynaptic differentiation of contacting axons (Wang Z. et al., 2009; Baumkötter et al., 2014; Stahl et al., 2014). Dendritic spine formation is also increased by heterologous expression of APP in primary hippocampal neurons (Lee et al., 2010). Conversely, a loss of endogenous APP causes a decrease in spine density (Lee et al., 2010). Although the molecular mechanisms are not yet fully understood, the current knowledge clearly suggests an essential physiological function of trans-interacting full length APP in synapse organization.

Despite the well-documented essential functions of APP/APLPs at the synapse, there is little knowledge of the molecular signals activated by APP/APLPs either functioning as putative ligands or as cell surface-associated receptors. The identification of receptor(s) responsible for sAPP-dependent signaling may shed light on the molecular mechanism underlying sAPP function at the synapse. In contrast, studies of intracellular APP-binding proteins have already provided some interesting insights on the molecular mechanisms by which full-length APP may transmit synaptic signals. Here, we summarize current knowledge of the synaptic functions of APP-binding proteins. Protein-based studies used to identify APP interactors have yielded a long list of candidate proteins involved in many different pathways. Aside from a few interesting reports highlighting the putative interaction of APP with G-protein mediated signaling (Milosch et al., 2014; Ramaker et al., 2016), the proteins most commonly identified in these studies bind the YENPTY APP internalization motif. This review is a discussion of our knowledge of the synaptic role of YENPTY APP-binding proteins.

## APP/APLP BINDING PROTEINS AND SYNAPTIC FUNCTION

Synapse formation and maintenance involves homo- and heterotypic interactions of Synaptic Cell Adhesion Molecules (SAM), including APP/APLP (Siddiqui and Craig, 2011), extracellular matrix components, extracellular ligands such as soluble APP fragments and other growth factors, as well as their adjacent receptors (Deyts et al., 2016). Herein, we present the signaling pathways involved in synapse formation, synaptic plasticity and synaptic neurotransmission in which APP-binding proteins participate, with a particular focus on the signaling events in which APP intracellular YENPTY domain binding proteins may play a role to alter synaptic function. This includes their role in well-established signal transduction pathways and their impact on cellular pathways, such as endocytosis, that are known to participate in signaling at the synapse (Fassio et al., 2016). The APP YENPTY domain binding proteins discussed include the X11/Munc-18-interacting proteins (Mints), FE65 proteins, Dab1, Numb/Numbl and Gulp1/CED-6, all capable of binding APP and other receptors through phosphotyrosine binding (PTB) domains (King et al., 2004; Wolfe and Guénette, 2007; Hao et al., 2011).

#### Reelin Signaling

The large extracellular protein reelin is best known for its role in neuronal migration in the developing cortex. Reelin interaction with the lipoprotein receptors apolipoprotein E receptor 2 (ApoER2) and very low-density lipoprotein receptor (VLDLR) initiates a signaling cascade through tyrosine phosphorylation of bound Dab1, an adaptor protein that is essential for neuronal positioning in the developing mouse brain (D'Arcangelo et al., 1999; Hiesberger et al., 1999; Howell et al., 1999; Trommsdorff et al., 1999). Dendritic morphogenesis and excitatory synapse formation are also regulated by the reelin/ApoER2/VLDLR signaling pathway (Niu et al., 2004; Groc et al., 2007; Qiu and Weeber, 2007). In the adult brain, reelin signaling through ApoER2 alters the activity of postsynaptic glutamate receptors in hippocampal slices, affecting LTP and synaptic plasticity (Weeber et al., 2002; Beffert et al., 2005). These events are also dependent on tyrosine phosphorylation of the Dab1 adaptor protein and the subsequent recruitment of Src family kinases to phosphorylated Dab1, known as the canonical reelin signaling pathway (reviewed in Bock and May, 2016). In the adult hippocampus, Dab1 regulates synaptic plasticity (Trotter et al., 2013). The adult forebrain specific and excitatory neuron specific conditional Dab1 KO mice, used to demonstrate this role for Dab1, display deficits in associative (fear conditioning) and spatial learning, while demonstrating no other developmental abnormalities previously associated with loss of this protein (Trotter et al., 2013). However, spine area measurements of hippocampal CA1 apical dendrites were reduced in these conditional KO mice. Furthermore, impairments in both hippocampal LTP and reelin-induced LTP were observed and these were associated with deficits in the sustained activation of ERK2 following synaptic potentiation (Trotter et al., 2013). Thus, Dab1-mediated reelin signaling is important for synaptic plasticity.

Several lines of evidence support a functional interaction between APP and reelin signaling (Hoe et al., 2006, 2009; Pramatarova et al., 2008; Rice et al., 2013; Divekar et al., 2014). Despite, the identification of Dab1 as a cytosolic binding protein for APP (Homayouni et al., 1999), and evidence for a genetic interaction between APP and Dab1 (Pramatarova et al., 2008), there is no evidence supporting a role for APP binding to Dab1 in the transmission of an APP-dependent reelin signal. However, the increase in APP binding to ApoER2 and the postsynaptic density (PSD)-95 protein in primary cortical neurons treated with reelin suggests that APP may participate in reelin signaling as a co-receptor (Divekar et al., 2014). Given that ApoER2 association with itself is increased by reelin treatment and because receptor clustering is a known mechanism for activation of intracellular signaling cascades for other receptors such as EGFR, Trk and Ephrin, the reelin-dependent increase in APP binding to ApoER2 may play a role in reelin synaptic signaling (Divekar et al., 2014). Dab1 binding to both APP and ApoER2 may modulate downstream signals. In addition, FE65, which also binds the NPXY recognition motif in ApoER2, may compete with Dab1 in this cellular context, adding another level of complexity to the regulation of this signaling cascade (Hoe et al., 2006).

#### Notch Signaling

The canonical Notch signaling pathway involves γ-secretase cleavage of Notch to produce the Notch intracellular domain (NICD) fragment, which is transcriptionally active. Notch signaling is regarded as a developmental signaling pathway for regulating stem cell maintenance and differentiation (Hori et al., 2013). It also plays a role in neurite outgrowth, dendritic arborization in immature neurons and synaptic plasticity in the adult brain (reviewed by Ables et al., 2011; Giniger, 2012). In mature pyramidal neurons, Notch signaling plays a role in regulating filopodia and spine densities (Dahlhaus et al., 2008; Alberi et al., 2011). Synaptic activity leads to an Arc-dependent increase in Notch and NICD levels (Alberi et al., 2011). Furthermore, downregulation of Notch in the hippocampus leads to impaired LTP and enhanced long-term depression (LTD; Wang Y. et al., 2004; Alberi et al., 2011), suggesting a role for Notch signaling in synaptic plasticity. Spatial learning deficits in the Morris Water Maze (MWM) and memory deficits in the Y-maze were also reported for mice in which Notch is knocked out in mature neurons (Alberi et al., 2011). Collectively, these data suggest that Notch signaling plays a role in hippocampal synaptic function.

Notch signaling is highly regulated, with the outcome being partly dependent on crosstalk with other signaling pathways and the type of cell receiving the Notch activation signal. One example of this crosstalk occurs between the Notch and reelin signaling pathways, with stabilization of NICD resulting from reelin-Dab1 signaling (Hashimoto-Torii et al., 2008). Moreover, NICD overexpression is able to rescue the neuronal migration phenotype of mice lacking reelin (Hashimoto-Torii et al., 2008). This seems to be due to the effect of Notch signaling on the morphology adopted by neural precursor cells to facilitate cellular migration. Whether crosstalk between Notch and reelin signaling plays a role in synaptic plasticity is presently unclear.

Evidence for interaction between Notch and APP signaling pathways also exists. Several studies have reported a physical interaction between APP and Notch (Fassa et al., 2005; Fischer et al., 2005; Oh et al., 2005, 2010; Chen et al., 2006). The YENPTY domain of Notch as well as APP interact with Numb and Numb-like (Numbl; Roncarati et al., 2002). Numb is an endocytic accessory protein that regulates clathrin-mediated endocytosis of its cargo proteins (reviewed in Yap and Winckler, 2015) and the absence of Numb and Numbl reduces Notch endocytosis producing higher levels of Notch and Notch signaling. Numb was identified in Drosophila as a Notch binding protein that regulates cell fate determination through inhibition of Notch signaling (Gulino et al., 2010). However, the consequence of altering Numb levels differs between vertebrates and Drosophila, as it's absence in vertebrates produces morphogenesis defects rather than the predicted increase in neurogenesis resulting from increased Notch signaling (Kuo et al., 2006; Rasin et al., 2007). In the absence of Numb and Numbl, adherens junctions are lost in radial glial cells due to abnormal cadherin localization. This alters cell polarity, producing detachment and ectopic localization of radial glial cells in the developing cortex (Rasin et al., 2007). Thus, Numb mediated trafficking of N-cadherin in the endocytic pathway participates in the maintenance of adherens junctions. Numb, which is also expressed in the adult mammalian cortex, hippocampal pyramidal cell layer and cerebellum (Stump et al., 2002), may participate in the regulation of the endocytic trafficking of its cargo proteins at the synapse. In support of this possibility, Numb has recently been shown to participate in mGlu1 mediated LTD in Purkinje cells (Zhou et al., 2015).

Numb/Numbl binding to the APP intracellular domain, AICD, alters nuclear signaling by repressing Notch activity (Roncarati et al., 2002). Further evidence supporting crosstalk between APP and Notch signaling comes from promoterreporter activation experiments showing that AICD in the presence of FE65 can trans-activate Hes-1, a Notch1 target gene, while NICD can trans-activate KAI-1, a putative AICD target gene, in HEK293 cells (Fischer et al., 2005). Interestingly, NICD trans-activation of the Hes-1 promoter can also be enhanced by FE65 expression (Fischer et al., 2005). However, opposing effects of APP on Notch signaling were reported for different cell types, indicating that the APP/Notch signaling crosstalk is context dependent (Oh et al., 2010). This may be due to cell-type dependent splicing of Numb, since alternatively spliced isoforms of Numb differentially affect APP internalization into the endocytic pathway (Kyriazis et al., 2008) and thus AICD generation. It may also be due to the cellular complement of adaptor proteins shared by Notch and APP, such as Numb and FE65, as competition of these adaptor proteins for APP or Notch/NICD may alter downstream signals. Furthermore, APP binding to Notch may modulate Notch signaling strength by preventing Notch receptor ligand interactions (Roncarati et al., 2002; Oh et al., 2005; Chen et al., 2006). Further studies are needed to assess whether crosstalk between Notch and APP signaling plays a role in synaptic structure and/or plasticity.

The association of FE65 proteins with receptors such as ApoER2 and Notch/NICD is shown in **Figure 1** as a possible mechanism by which FE65 may function at the synapse.

# Adhesion Proteins

Adhesion proteins that form complexes with APP, such as N-cadherin and calsyntenin/alcadeins, are implicated in synaptic contact formation and synaptic plasticity (Tang et al., 1998; Togashi et al., 2002; Arikkath and Reichardt, 2008; Pettem et al., 2013; Ster et al., 2014).

The classical cadherins participate in cell adhesion and communicate with their intracellular binding partners, the catenins, to link adhesion to intracellular pathways. The cadherin/catenin complex localizes to synapses where it regulates activity dependent spine remodeling (Arikkath and Reichardt, 2008; Bian et al., 2015). Although co-immunoprecipitation experiments demonstrate N-cadherin binding to APP (Asada-Utsugi et al., 2011), a role for N-cadherin/APP interactions in cadherin-modulated synaptic events has not been reported. However, N-cadherin binds the APP YENPTY binding protein, Numb, which plays a role in mGlu1 mediated LTD in Purkinje cells (Zhou et al., 2015). Thus, the ratio of APP-Numb and N-cadherin-Numb interactions may alter synaptic transmission.

In addition to classical adhesion molecules, there are a number of synaptic adhesion complexes that induce synaptic differentiation, a classic example is presynaptic neurexin binding to postsynaptic neuroligin. The cadherin related protein family member, Calsyntenin-3/Alcadein β, which is highly expressed in interneurons, forms a functional complex with α-neurexin that promotes calsyntenin-3 mediated presynaptic differentiation of inhibitory synapses (Pettem et al., 2013; Um et al., 2014). Calsyntenin-1 and -2 do not share this effect (Um et al., 2014). However, the observation that knockdown of all three calsyntenin proteins is necessary for decreased inhibitory synaptic transmission in both cultured hippocampal neurons and layer II/III somatosensory cortical neurons in situ suggests that all three family members redundantly regulate inhibitory synapse formation and function (Um et al., 2014).

Calsyntenin-1/Alcadein α forms a ternary complex with APP and the APP YENPTY binding protein, X11L. The formation of this ternary complex suppresses secretase cleavage of both APP (Araki et al., 2003) and Calsyntenin 1/Alcadein α (Araki et al., 2004). Furthermore, the γ-secretase cleavage product of Alcadein α, AlcαICD, competes with APP for FE65 binding and FE65 stabilizes AlcαIACD, similar to its stabilization of AICD (Kimberly et al., 2001; Araki et al., 2004). This competition may lead to regulation of AICD-mediated signaling. Although, the impact of Calsyntenin-1/Alcadein α cleavage on synaptic function is unknown its putative binding to FE65 at the synapse is highlighted in **Figure 1**.

Given that APP/APLP trans-dimerization is implicated in establishing synaptic contacts (Soba et al., 2005; Wang Z. et al., 2009; Prox et al., 2013; Klevanski et al., 2014), while factors that increase APP processing such as shedding (Stahl et al., 2014) or activity-dependent Aβ generation may be important for synaptic remodeling, a better understanding of the integration of Notch and reelin signaling on APP processing, signaling and metabolism at the synapse and the role of cross-talk between APP and other synaptic adhesion molecules seems warranted, as these may contribute to synaptic plasticity.

#### Gulp1 and Endocytosis

Gulp1/CED-6, a YENPTY APP-binding protein, is a neuronal protein found in synaptosome-enriched fractions of rat brain, where it co-localizes with clathrin-coated vesicles (Martins-Silva et al., 2006). Gulp is involved in trafficking in the endocytic pathway enhancing APP processing and Aβ generation when overexpressed (Kiss et al., 2006; Hao et al., 2011). Furthermore, Gulp associates with and positively regulates ADP-ribosylation factor 6 (Arf6; Ma et al., 2007), a small GTPase that regulates clathrin and caveolin-independent endocytic trafficking of BACE1 in the somatodendritic compartment of neurons, where BACE1 encounters APP (Ma et al., 2007; Sannerud et al., 2011). Thus, Gulp/APP interactions might regulate synaptic levels of APP and its proteolytic products by regulating APP intracellular trafficking at the synapse.

#### Regulation of APP Intracellular Complexes through Phosphorylation

APP-dependent modulation of synaptic structure and function may occur through alternative splicing of APP or phosphorylation of the APP C-terminus, thereby altering interaction of APP/APLP with their binding proteins (Kyriazis et al., 2008; Tamayev et al., 2009; Dunning et al., 2016). Alternative splicing of APP and its homologs is complex, but detailed investigations in the context of the putative physiological functions of APP are lacking (Pandey et al., 2016). The APP intracellular tail encompasses three Tyr and five Ser/Thr putative phosphorylation sites, of which two of the Tyr residues (Tyr682 and Tyr687) and three of the Ser/Thr sites (Thr654, 668 and Ser655) can exist in a

Rac1, Ras-related C3 botulinum toxin substrate 1; VGLUT1, vesicular glutamate transporter 1; SV2A, synaptic vesicle glycoprotein 2A; LRP1, low-density lipoprotein receptor-related protein; ApoER2, apolipoprotein E receptor 2; VLDLR, very low-density lipoprotein receptor; APP, amyloid precursor protein; APLPs, APP-like proteins (1 and 2); P2X2, P2X purinergic receptor 2; NCAM2, neural cell adhesion molecule 2; L1CAM, neural cell adhesion molecule L1; PSD, post-synaptic density; PTB1 and PTB2, phosphotyrosine-binding domain 1 and 2; WW, protein domain containing two tryptophans.

phosphorylated state. These phosphorylation sites are docking sites for different adaptor proteins and at least for some of these, Tyr682 and Thr668, an influence on the binding of Shc and Grb2 or FE65, respectively, with full-length APP, and/or the α- and β-secretase derived APP C-terminal fragments has been documented (for review, see Schettini et al., 2010). The AICD interactome was also found to differ depending on phosphorylation of Tyr682 and Thr668 (Tamayev et al., 2009). Likely, the physiological relevance of the different sites can only be understood in specific signaling contexts and should include

Frontiers in Molecular Neuroscience | www.frontiersin.org March 2017 | Volume 10 | Article 87

analysis of both APP phosphorylation and phosphorylation of their interacting proteins. More research will be required for a better understanding of these networks in the context of synapse formation and function.

The studies described above provide information on how APP-binding adaptor proteins contribute to signaling pathways implicated in synaptic function. Interestingly, X11 and FE65 proteins modulate signaling in these pathways. The remainder of this review focuses on the X11 and FE65 proteins and discusses their significance for synaptic function in light of recent KO studies. The synaptic phenotypes identified offer specific contexts in which to study the interplay between APP-binding PTB adaptor proteins and the additional ligands they bind on synaptic signaling.

#### X11/MINT MUTANT MICE

The APP-interacting proteins, X11, X11L and X11L2, bind to the YENPTY motif in the cytoplasmic region of APP (Borg et al., 1996; McLoughlin et al., 1999; Tanahashi and Tabira, 1999; Tomita et al., 1999). Interaction of X11 and X11L with Munc-18 (Munc-18-interacting protein (Mint)), a protein mediating membrane-vesicle fusion, was also reported (Okamoto and Südhof, 1997). Hence, multiple nomenclatures exist for this protein family: X11/X11α/Mint1, X11L/X11β/Mint2, and X11L2/X11γ/Mint3. In this review article, we refer to these proteins by their original nomenclature—X11, X11L and X11L2 (Duclos et al., 1993).

All three X11 proteins contain a conserved C-terminus, which consists of a phosphotyrosine-interaction/binding (PTB) and two PDZ (PSD95, Drosophila disc large tumor suppressor (Dlg1), and zonula occludens-1 protein (zo-1)) domains, mediating different types of protein–protein interactions. The X11 proteins diverge in the N-terminus, where X11 and X11L display an additional Munc-18 interacting domain and where only X11 bears a CASK-interacting domain (Okamoto and Südhof, 1997, 1998; Butz et al., 1998; Borg et al., 1999). Further, X11 and/or X11L associate with different interaction partners, including Kalirin-7 and XB51/NECAB3 (Lee et al., 2000; Jones et al., 2014). X11L is exclusively expressed in neurons, whereas X11 is found predominantly in the brain, but is also expressed in the pancreas, testis and paranephros (Motodate et al., 2016). Notably, some neurons, such as Purkinje cells showed only expression of X11, whereas X11L2 was found ubiquitously expressed, with substantial amounts in the brain (Motodate et al., 2016). The X11 family proteins regulate intracellular trafficking of APP as well as other NPXY motif containing transmembrane proteins (Araki et al., 2003; Saito et al., 2008, 2011; Gross et al., 2013; Sullivan et al., 2014) and affect APP processing, including generation of the Aβ peptide (Borg et al., 1998; Tanahashi and Tabira, 1999; Tomita et al., 1999; Shrivastava-Ranjan et al., 2008; Caster and Kahn, 2013). Interestingly, X11L2 and to a lesser extent X11L are distributed between the cytosolic and nuclear fractions, whereas X11 is recovered mostly in the cytosolic and membrane fractions. Thus, X11L2 might function as a transcriptional co-activator (Sumioka et al., 2008).

In a recent study, it was shown that all X11 family proteins are involved in activity dependent regulation of surface APP levels (Sullivan et al., 2014). Neuronal activity was associated with APP endocytosis followed by increased APP levels at the surface. This is highly interesting, as elevated APP cell surface levels were shown to increase APP synaptogenic activity (Stahl et al., 2014). In addition, X11 overexpression increases excitatory synaptic activity and activity dependent APP endocytic trafficking and Aβ generation (Sullivan et al., 2014). These data are consistent with the hypothesis, that X11/APP interactions may regulate activitydependent synaptic remodeling.

X11 loss of function analyses revealed movement impairments and a decrease in GABAergic neurotransmission in KO mice (Ho et al., 2003, 2006). Further, X11-KO mice showed alterations in dopaminergic neurotransmission (Mori et al., 2002). X11L and X11L2 single KO mice revealed no obvious deficits, but X11L is functionally redundant for X11, as 80% of X11/X11L DKO mice die early after birth and the surviving mice exhibit increased growth and aggravated motor impairments (Ho et al., 2003, 2006; Sano et al., 2009). Furthermore, mouse X11/X11L mutants exhibited impairments in presynaptic neurotransmitter release, as indicated by lowered basal neurotransmission and reduced miniature excitatory post-synaptic current (mEPSC) frequency (Ho et al., 2006). As paired pulse facilitation was decreased and synaptic density was unchanged, these data can be explained by a decrease in synaptic vesicle release probability in X11/X11L DKO neurons (Ho et al., 2006). These data argue that the impaired synaptic vesicle release might be due to loss of interaction between X11/X11L and Munc-18. Consistently, the additional loss of X11L2, a family member lacking the Munc-18 binding site, did not aggravate the synaptic phenotype of X11/X11L DKO mice (Ho et al., 2006).

Interestingly, X11 single KO mice exhibit an increased paired-pulse depression at inhibitory synapses (Ho et al., 2003), consistent with an increased release probability, whereas analysis of X11/X11L/X11L2 KO neurons suggests a decreased release probability at excitatory synapses. This observation suggests that X11 may play a more specialized function at inhibitory synapses, whereas at excitatory synapses X11 and X11L might exhibit overlapping functions. Consistently, X11 is highly expressed in interneurons (Ho et al., 2003). However, other compensatory mechanisms may occur, for example, X11/X11L/X11L2-deleted neurons show increased levels of FE65, FE65L1 and FE65L2 proteins suggesting that X11 and FE65 proteins are functionally related (Ho et al., 2006). As X11 and FE65 proteins both contain a PTB domain, mediating binding to APP, it is conceivable that X11 and FE65 proteins are partially redundant for an APP-mediated function at the synapse (Ho et al., 2006). However, in a recent study no alterations in paired pulse facilitation were observed in FE65/FE65L1 DKO mice (Strecker et al., 2016). Alternatively, the functional overlap of X11 and FE65 may occur in dendritic spines. Levels of the AMPA-type glutamate receptor, GluR1, are increased in cortical neurons with acute deletion of X11 protein family members and the postsynaptic localization of the AMPA-type receptor GLR1 of Caenorhabditis elegans is impaired in Lin-10/X11 mutant interneurons (Rongo et al., 1998). Furthermore, X11 localizes to the mobile fraction of the PSD in excitatory cortical neurons where it interacts with Kalirin-7, a guanine-nucleotide exchange factor (GEF) that regulates Rac1 localization and function (Jones et al., 2014).

#### FE65/FE65L1 MUTANT MICE

The FE65 protein family, consisting in mammals of FE65, FE65-like 1 (FE65L1) and FE65-like 2 (FE65L2), are scaffolding/adaptor proteins able to form multi-molecular complexes that function in many cellular processes, such as calcium homeostasis (Nensa et al., 2014), actin remodeling and nuclear signaling (recently reviewed in Chow et al., 2015). All three FE65 proteins share conserved protein-protein interaction/binding motifs, namely the N-terminal WW-domain and the two C-terminal phosphotyrosine-binding domains 1 and 2 (PTB1, PTB2; Meiyappan et al., 2007; Radzimanowski et al., 2008a,b). The complexity of the FE65 protein family is further increased by the existence of several splice variants (p90FE65, p60FE65), polymorphisms within FE65 and cleavage products driven by proteases (p65FE65, which has an up to 40-fold higher affinity for APP than p97FE65; Hu et al., 1999, 2002, 2005; Domingues et al., 2011; Saeki et al., 2011; Golanska et al., 2013; Loosse et al., 2016). However, little is known about the specific localization and functions of these FE65/FE65L1/FE65L2 isoforms. Future experiments with specific antibodies against the different FE65 family members as well as their individual splice variants and processing products, might help clarify these questions.

FE65 and its family members interact with the intracellular domains of APP/APLPs (Fiore et al., 1995; Guénette et al., 1996; Duilio et al., 1998). As FE65 is predominantly expressed in the brain, similar to APP695, it has been studied more extensively than the more widely distributed FE65L1 and FE65L2 (Kesavapany et al., 2002; Guo et al., 2012). However, during mouse brain development FE65 expression clearly differs from APP. Whereas APP is upregulated during development until the first postnatal week, FE65 levels begin to decline after embryonic day 15 and increase again progressively from post-partum day 10 to adulthood (Sandbrink et al., 1997; Kesavapany et al., 2002). Interestingly, histological examination of FE65 or FE65L1 KO mouse brains revealed no abnormalities, while mice lacking both FE65 and FE65L1 resemble the APP/APLP1/APLP2 triple-KO (TKO) mouse phenotypes, exhibiting among other phenotypes, ectopic neurons and axonal pathfinding defects (Herms et al., 2004; Guénette et al., 2006). These data suggest that FE65 proteins mediate APP protein function in the developing brain possibly through transmission of an APP-dependent signal necessary for brain development. An alternative possibility is that loss of the FE65 proteins leads to APP-dependent sequestration of PTB-binding adaptor proteins essential for brain development.

The FE65 interaction with Mena/Vasp proteins, regulators of actin dynamics, is of interest because Mena KO mice have axonal pathfinding defects and improper positioning of neurons in the developing brain that bear resemblance to phenotypes observed in FE65/FE65L1 DKO and the APP/APLP1/APLP2 TKO mice (Lanier et al., 1999; Goh et al., 2002; Herms et al., 2004; Guénette et al., 2006). Recovery of a tripartite complex of FE65, Mena and APP and the co-localization of these proteins in growth cones and synapses suggest a neuronal function for this complex (Sabo et al., 2003; Ikin et al., 2007). Adenovirus-mediated expression of interaction-deficient FE65, bearing mutations that either abrogates PTB2 domain interactions (APP) or WW domain interactions (Mena/Vasp), altered axon branching (Ikin et al., 2007) suggesting a role for such complexes in neurite outgrowth. Functional analyses to determine whether APP-FE65-Mena/Vasp or FE65/Mena/Vasp complexes are present at the synapse would be a first step towards addressing a putative role for this complex in synaptic function.

Our recent detailed in vivo study examining FE65 protein family function using learning behavior analyses, immunohistological staining and electrophysiological measurements of different FE65/FE65L1 protein family KO mice provides further insights into the role of FE65 protein family members in the central and peripheral nervous system (CNS, PNS) that again show phenotypes similar to APP protein family KO mice (Strecker et al., 2016). Impairments in the maintenance of LTP in the Schaffer collateral pathway of FE65/FE65L1 DKO mice suggest that these proteins play a role in synaptic plasticity (Strecker et al., 2016). Although the FE65 single KO mice showed a trend towards decreased post-tetanic potentiation, maintenance of LTP was not significantly different from WT and no deficits were observed in FE65L1KO mice. A previous study of the isoform specific p97FE65 KO mice (lacking the longest FE65 isoform, p97, but simultaneously overexpressing six-times more of the shorter isoform, p60) reported early-phase LTP dysfunction (Wang Y. et al., 2009). Collectively these data support overlapping functions for FE65 and FE65L1 in synaptic neurotransmission. Interestingly, comparable potentiation rates have been observed in LTP measurements of acute hippocampal slices of APP∆CT15-DM mice (APP lacking the last 15 amino acids KI—APLP2 KO mice; Klevanski et al., 2015) pointing towards a shared function for FE65 and APP at the synapse. A role for FE65 proteins at the synapse is further supported by FE65 interaction with SV2, a synaptic vesicle protein, as well as sarcoplasmatic/endoplasmatic reticulum calcium ATPase (SERCA) and ryanodine receptor (RYR; Nensa et al., 2014), involved in calcium release/homeostasis in synapses under normal physiological conditions (reviewed in Mendoza-Torreblanca et al., 2013; Del Prete et al., 2014; Elaïb et al., 2016). Interestingly, dysregulation of calcium homeostasis is discussed in pathological conditions of AD (reviewed in Small, 2009), which may involve dysregulation of this aspect of FE65 protein function.

p97FE65 KO mice displayed deficits in cognitive behavior in non-spatial learning tasks and showed significant impairments in hidden platform and reversal learning in the MWM spatial learning test (Wang B. et al., 2004; Wang Y. et al., 2009). However, no memory deficits were observed for these mice. In contrast, memory deficits were observed in the MWM test for the FE65 KO (lacking both p60 and p97 isoforms) and FE65L1 KO mice in our study (Strecker et al., 2016). Confounding behaviors for locomotion analyses and possible loss of vision in FE65/FE65L1 DKO mice (Suh et al., 2015; Strecker et al., 2016) made it impossible to interpret the MWM spatial learning deficits observed for FE65/FE65L1 DKO mice.

Additional insights into the molecular mechanisms by which loss of the FE65 proteins results in the observed phenotypes comes from our knowledge of the function of their binding partners (Chow et al., 2015). The functions in which FE65 protein family members may participate include effects on actin cytoskeleton dynamics (Ermekova et al., 1997; Perkinton et al., 2004; Ward et al., 2010), Ca2<sup>+</sup> homeostasis (Nensa et al., 2014), APP mediated signaling (discussed in more detail below), nuclear signaling via Tip60 (Cao and Südhof, 2001) and the response to DNA damage (Minopoli et al., 2012).

Early studies suggesting a role for FE65 in transcriptional regulation came from the identification of the histone acetyltransferase, Tip60 and the transcription factors, CP2/LSF/LBP1 and SET, as FE65 PTB1 domain binding proteins (Zambrano et al., 1998; Cao and Südhof, 2001; Telese et al., 2005). Despite intensive studies addressing AICD/FE65-regulated gene expression, there is a lack of consensus for most of the identified target genes (Hébert et al., 2006; Waldron et al., 2008). For a list of gene targets including those supported by promoter binding studies see Pardossi-Piquard and Checler (2012). These conflicting results may be due to the different experimental systems studied and the possibility that FE65 transcriptional regulation only occurs in specific physiological contexts. In support of this possibility, a recent study showed that FE65 is involved in epigenomic regulation of specific transcriptional programs implicated in the response to DNA damage (Ryu et al., 2015).

With respect to the role of FE65 proteins in synaptic function, the small GTPase, ARF6, which influences endocytic and membrane trafficking in neurons, is an intriguing FE65 interactor that may form an APP-FE65-Arf6 tripartite complex (Sannerud et al., 2011; Cheung et al., 2014; Tang et al., 2015). FE65 preferentially binds to ARF6 in its inactive GDP-bound form and stimulates the activation of ARF6 (Cheung et al., 2014). ARF6 is involved in synaptic function via regulation of AMPA receptor trafficking and synaptic plasticity during NMDA receptor-mediated LTP (Oku and Huganir, 2013). It also participates in NMDA-dependent LTD (Scholz et al., 2010) and regulates the cycling and readily releasable pool (RRP) of synaptic vesicles at the presynaptic site (Tagliatti et al., 2016). A recent study points towards a bi-directional function for ARF6 in spine formation and maintenance that is dependent on neuronal maturity and activity (Kim et al., 2015). In immature neurons expression of genes involved in cell motility and actin cytoskeleton organization are up-regulated by ARF6, while in mature neurons expression of genes important for neuronal activity such as synaptic transmission are up-regulated by ARF6 (Kim et al., 2015). Furthermore, synaptic activity reverses these effects indicating that ARF6 mediated signaling may play a role in synaptic plasticity (Kim et al., 2015). Interestingly, the interaction of FE65 and ARF6 influences ARF6 signaling to Rac1 (Cheung et al., 2014), which is implicated in neuronal outgrowth and spine structural plasticity (Cheung et al., 2014; Kim et al., 2015). In addition, Rac1 was previously reported to interact with FE65 and regulates its expression (Wang et al., 2011). Both Arf6 and Rac1 are included in **Figure 1** as FE65 binding proteins that may contribute to FE65 function at the synapse. Knockdown of ARF6 also affects neuronal migration in the developing



Phenotypes from X11 or FE65 mutant mice exhibiting similar phenotypes as APP/APLP2 mutant mice, lacking the FE65 or X11 interaction site are highlighted in dark gray. <sup>a</sup> Isoform specific FE65 KO, expressing higher levels of the p60 isoform, <sup>b</sup>APP1CT knockin on an APP-like protein 2 (APLP2) KO background or APP1CT knockin with familial Alzheimer's disease (AD) mutations in the humanized Aβ domain on an APLP2 KO background (Li et al., 2010) or APPY682G knockin on an APLP2 KO background (Barbagallo et al., 2011). <sup>c</sup>Abbreviations: LTP, Long-term potentiation; PTP, Post-tetanic potention; PPF, Paired pulse facilitation; mEPSC, miniature excitatory post-synaptic current; RRP, readily releasable pool; NMJ, neuromuscular junction; n.d., not determined. References: <sup>1</sup>Strecker et al., 2016; <sup>2</sup>Suh et al., 2015; <sup>3</sup>Ho et al., 2006; <sup>4</sup>Klevanski et al., 2015; <sup>5</sup>Wang B. et al., 2004; <sup>6</sup>Wang Y. et al., 2009; <sup>7</sup>Guénette et al., 2006; <sup>8</sup>Li et al., 2010; <sup>9</sup>Barbagallo et al., 2011.

cortex (Hara et al., 2016). Thus, the FE65/ARF6 interaction and its effects on ARF6 signaling are consistent with many of the phenotypes observed in FE65/FE65L1 DKO and APP/APLP1/APLP2 TKO mouse brains. Further research in this direction should help determine the contribution of the FE65/ARF6 pathway to the phenotypic similarities observed between FE65/FE65L1 DKO and APP/APLP1/APLP2 TKO synaptic defects in the hippocampus.

FE65/FE65L1 KO and various APP KO mouse models share common impairments in NMJ formation with reduced preand postsynaptic areas and deficits in apposition of the preand postsynapse (Li et al., 2010; Weyer et al., 2011; Klevanski et al., 2014, 2015; Strecker et al., 2016). These are aggravated in FE65/FE65L1 DKO compared to FE65 or FE65L1 KO mice NMJs (Strecker et al., 2016), possibly leading to muscle degeneration/denervation (Suh et al., 2015) and the locomotion deficits and impairments in strength observed in these mice (Strecker et al., 2016).

APP interaction with low-density lipoprotein receptorrelated protein 4 (LRP4), a component of the postsynaptic LRP4/MUSK/Agrin complex, is important for Acetylcholinereceptor patterning and stabilization at postsynaptic sites of the NMJ (Choi et al., 2013). Given that FE65 interaction with the intracellular domain of many lipoprotein receptors has been demonstrated (Gotthardt et al., 2000; Hoe et al., 2006; Alvira-Botero et al., 2010; Dumanis et al., 2012), the observation that the LRP4 ectodomain is sufficient for pre- and post-synaptic differentiation of the NMJ indicates that any contribution FE65 may have to this pathway may be via its interaction with APP in an APP/LRP4 tripartite complex (Gomez and Burden, 2011).

#### REFERENCES


#### CONCLUDING REMARKS

To gain insights into the molecular mechanisms by which APP functions at the synapse, we have re-examined the cytosolic APP interactome literature. Taking the different interaction partners into account, we highlighted some putative signaling pathways, involving Reelin, Notch and cell adhesion proteins, in which APP-interactors may participate to modulate synaptic function. PTB-containing interactors that bind the YENPTY motif in the APP-C terminus are the most prominently studied. Comparisons of several APP mutant mouse models that either lack or bear a mutation in the YENPTY motif, to X11 or FE65 KO mouse models reveal a surprisingly high degree of similarity between APP mutant mice and FE65 protein family KO mice (**Table 1**). Therefore, we conclude, that Fe65 family proteins play a pivotal role in APP function and have outlined possible cellular events in which APP-FE65 signaling may operate at the synapse (**Figure 1**). In future it will be important to evaluate those putative pathways and to investigate in more detail the regulation of APP-FE65 interactions at the synapse.

#### AUTHOR CONTRIBUTIONS

SG, PS and SK co-wrote the review.

#### ACKNOWLEDGMENTS

Research was funded by the Deutsche Forschungsgemeinschaft (FOR1332, KI-819/5-2 to SK), the Stiftung Rheinland-Pfalz für Innovation (to SK) and the Alzheimer Forschung Initiative (to SG and SK).


YENPTY motif of amyloid precursor protein. Mol. Cell. Biol. 16, 6229–6241. doi: 10.1128/mcb.16.11.6229


protein-interacting proteins in brain development. EMBO J. 25, 420–431. doi: 10.1038/sj.emboj.7600926


FE65 with resistance to very late onset dementia of the Alzheimer type. Hum. Mol. Genet. 11, 465–475. doi: 10.1093/hmg/11.4.465


fashion with glypican and syndecan proteoglycans. J. Cell Sci. 126, 4856–4861. doi: 10.1242/jcs.137919


associative learning. J. Neurosci. 33, 15652–15668. doi: 10.1523/JNEUROSCI. 2010-13.2013


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Guénette, Strecker and Kins. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Region-Specific Differences in Amyloid Precursor Protein Expression in the Mouse Hippocampus

Domenico Del Turco<sup>1</sup> \*, Mandy H. Paul <sup>1</sup> , Jessica Schlaudraff <sup>1</sup> , Meike Hick 1,2 , Kristina Endres <sup>3</sup> , Ulrike C. Müller <sup>2</sup> and Thomas Deller <sup>1</sup>

1 Institute of Clinical Neuroanatomy, Neuroscience Center, Goethe-University, Frankfurt, Germany, <sup>2</sup> Institute of Pharmacy and Molecular Biotechnology (IPMB), Heidelberg University, Heidelberg, Germany, <sup>3</sup> Clinic for Psychiatry and Psychotherapy, University Medical Center Mainz, Mainz, Germany

The physiological role of amyloid precursor protein (APP) has been extensively investigated in the rodent hippocampus. Evidence suggests that APP plays a role in synaptic plasticity, dendritic and spine morphogenesis, neuroprotection and—at the behavioral level—hippocampus-dependent forms of learning and memory. Intriguingly, however, studies focusing on the role of APP in synaptic plasticity have reported diverging results and considerable differences in effect size between the dentate gyrus (DG) and area CA1 of the mouse hippocampus. We speculated that regional differences in APP expression could underlie these discrepancies and studied the expression of APP in both regions using immunostaining, in situ hybridization (ISH), and laser microdissection (LMD) in combination with quantitative reverse transcription polymerase chain reaction (RT-qPCR) and western blotting. In sum, our results show that APP is approximately 1.7-fold higher expressed in pyramidal cells of Ammon's horn than in granule cells of the DG. This regional difference in APP expression may explain why lossof-function approaches using APP-deficient mice revealed a role for APP in Hebbian plasticity in area CA1, whereas this could not be shown in the DG of the same APP mutants.

#### Edited by:

Nicola Maggio, The Chaim Sheba Medical Center, Israel

#### Reviewed by:

Alessandro Vercelli, University of Turin, Italy Michael R. Kreutz, Leibniz-Institute for Neurobiology, Germany

#### \*Correspondence:

Domenico Del Turco delturco@em.uni-frankfurt.de

Received: 04 August 2016 Accepted: 15 November 2016 Published: 29 November 2016

#### Citation:

Del Turco D, Paul MH, Schlaudraff J, Hick M, Endres K, Müller UC and Deller T (2016) Region-Specific Differences in Amyloid Precursor Protein Expression in the Mouse Hippocampus. Front. Mol. Neurosci. 9:134. doi: 10.3389/fnmol.2016.00134 Keywords: APP, dentate gyrus, CA1, immunostaining, western blotting, laser microdissection, in situ hybridization, RT-qPCR

# INTRODUCTION

Amyloid precursor protein (APP) is an integral membrane protein involved in the pathogenesis of Alzheimer's disease (AD). It is processed by proteases and cleaved into several biologically active fragments (e.g., Turner et al., 2003; Müller and Zheng, 2012; Zhang et al., 2012). Of note, proteolysis of APP by beta- and gamma-secretases generates the amyloid-ß (Aß) peptide, which oligomerizes, interferes with synaptic functions, and eventually aggregates into extracellular amyloid plaques, one of the neuropathological hallmarks of AD (Selkoe and Hardy, 2016). In contrast, proteolysis of APP by α-secretases (e.g., Postina et al., 2004; Yang et al., 2006; Fahrenholz, 2007; Prinzen et al., 2009; Saftig and Reiss, 2011; Kuhn et al., 2016), generates soluble APP-α (sAPPα), which is neuroprotective and important for neuronal plasticity (Turner et al., 2003; Ring et al., 2007; Aydin et al., 2012; Kögel et al., 2012). In the latter case, the Aß-peptide is not formed because α-secretases cleave APP within the Aß region of the protein. In AD the balance of this processing by secretases shifts towards the amyloidogenic pathway, which increases Aß production and leads to a lack of sAPPα (Endres and Fahrenholz, 2012) resulting in an impairment of cognition.

A region of the brain which is of particular interest in the context of AD is the hippocampus. Since the hippocampal formation and hippocampus-dependent learning and memory are affected early during the course of the disease (Braak and Braak, 1991) the hippocampus has been used as a model brain region to study the role of APP and its cleavage products in synaptic plasticity, learning and memory and neuroprotection (e.g., Turner et al., 2003; Ring et al., 2007). Interestingly, our physiological investigations of APP−/<sup>−</sup> mice revealed remarkable differences between the subregions of the hippocampus: whereas APP was necessary for long-term potentiation (LTP) at the CA3–CA1 synapse (Ring et al., 2007; Weyer et al., 2011; Hick et al., 2015) it was not essential for LTP at the entorhinal cortex-granule cell (EC-GC) synapse in the dentate gyrus (DG; Jedlicka et al., 2012). We speculated that regional differences in basal APP expression or APP processing could explain these phenotypic differences. This interpretation would be in line with a recent publication, which reported APP to be predominantly expressed by interneurons in the DG (Wang et al., 2014).

To provide first evidence for this hypothesis and to reliably quantify differences in APP expression between granule cells of the DG and pyramidal cells of area CA1, we studied layerspecific expression levels of APP in the principal cell layers using laser microdissection (LMD) in combination with quantitative polymerase chain reaction (qPCR) and western blot analysis (e.g., Burbach et al., 2003; Del Turco et al., 2014). Since APP is alternatively spliced into three major isoforms (Kang et al., 1987; Tanzi et al., 1988; Sisodia et al., 1993; Rohan de Silva et al., 1997), i.e., APP-770, APP-695 and APP-751, assays detecting all major isoforms were employed. Furthermore, we used an antibody for western blotting, which is highly specific for APP and does not show staining on APP−/<sup>−</sup> brain tissue (Guo et al., 2012) to quantify APP levels and to study its cellular distribution. The selection of the antibody appeared to be especially important, since some antibodies show unspecific background staining on tissue sections and may cross-react with APP-related proteins, such as the APP-like-proteins 1 or 2 (Anliker and Müller, 2006; Kaden et al., 2012; Müller and Zheng, 2012). Together with in situ hybridization (ISH) data for APP, our results show that APP is expressed exclusively by hippocampal neurons under physiological conditions. It is ∼1.7 fold higher expressed by CA1 pyramidal cells compared to dentate granule cells, which may contribute to the regional differences seen in electrophysiological studies of APP−/<sup>−</sup> mice (Ring et al., 2007; Jedlicka et al., 2012).

#### MATERIALS AND METHODS

#### Animals

Adult (3–5 months old) male C57BL/6J mice (Janvier, France) and APP-deficient mice obtained from the colony at Heidelberg University (e.g., Li et al., 1996; Jedlicka et al., 2012) were used for experimental analysis. Animal care and experimental procedures were performed in agreement with the German law on the use of laboratory animals (animal welfare act; TierSchG). Animal welfare was supervised and approved by the Institutional Animal Welfare Officer.

#### Immunofluorescence

Mice were deeply anesthetized with an overdose of pentobarbital (300 mg/kg body weight) and transcardially perfused with 0.9% sodium chloride (NaCl) followed by 4% paraformaldehyde (PFA) in phosphate-buffered saline (pH 7.4). Brains were removed, post-fixed for 4–24 h in 4% PFA and sectioned in the coronal plane (40 µm) using a vibratome (VT1000 S, Leica Microsystems). Free-floating sections were incubated in a blocking buffer containing 0.5% Triton X-100 and 5% bovine serum albumin (BSA) in 0.05 M Tris-buffered saline (TBS) for 30 min at room temperature followed by incubation in the primary antibody (diluted in 0.1% Triton X-100 and 1% BSA in 0.05 M TBS) overnight at 4 ◦C. The following primary antibodies were used: mouse anti-APP (22C11, immunogen: 66–81 amino acids (aa) of purified recombinant Alzheimer precursor A4 fusion protein (N-terminus); MAB348, Chemicon), rabbit anti-APP (CT20, immunogen: synthetic peptide corresponding to 751–770 aa of human APP (C-terminus); 171610, Calbiochem), rabbit anti-APP (Y188, immunogen: synthetic peptide corresponding to C-terminus of human APP (YENPTY motif); ab32136, Epitomics), mouse anti-NeuN (A60, immunogen: purified cell nuclei from mouse brain; MAB377, Chemicon) and rabbit anti-GFAP (immunogen: GFAP isolated from cow spinal cord; Z0334, Dako). After several washes, sections were incubated with the appropriate secondary Alexa-conjugated antibodies (1:2000, Invitrogen, Waltham, MA USA) for several hours at room temperature, counterstained with Hoechst 33242 (Invitrogen) or DRAQ5 (Thermo Fisher Scientific, Waltham, MA, USA), and finally mounted in DAKO Fluorescent Mounting Medium (Dako).

### Western Blotting

For protein extraction, 10× volume of homogenization buffer (20 mM Tris, 500 mM NaCl, 0.5% CHAPS, 5 mM EDTA) was added to freshly dissected tissue samples, i.e., whole hippocampus as well as microdissected CA1 pyramidal cell layer (pcl) and dentate granule cell layer (gcl). Homogenization was performed with a pestle (Wheaton, Montgomery, MD, USA). After centrifugation at 4◦C for 30 min (22,000 rpm, Sorvall WX Ultra Series, Thermo Electron Corporation), protein concentration was quantified with a Qubit<sup>r</sup> 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA) using Qubit<sup>r</sup> Protein Assay Kit (Life Technologies, Carlsbad, CA, USA). Samples were denatured for 5 min at 95◦C and immediately cooled down on ice. For gel electrophoresis, protein amounts (approx. 30 µg for hippocampal tissue, 5–6 µg for microdissected tissue) were loaded onto 8% SDS–polyacrylamide gels and were separated at 120 V for 15 min followed by 160–180 V for 45 min. Subsequently, gels were blotted to nitrocellulose membranes at 15 V for 75 min. Blots were then washed twice in TBS and incubated with Odyssey Blocking Buffer (LI-COR Biosciences) at room temperature for 60–120 min. Blots were washed again in TBS and incubated overnight at 4◦C with the appropriate primary antibody diluted in 1:1 Odyssey Blocking Buffer with TBS and 0.1% Tween20. Blots were washed in TBS with 0.1% Tween20 and incubated with an IRDye800CW conjugated secondary antibody (LI-COR Biosciences) at room temperature for 45 min. For normalization mouse anti-GAPDH antibody (Calbiochem) in combination with an IRDye680 conjugated goat anti-mouse antibody (LI-COR Biosciences) was used. Two-color imaging was performed using Odyssey<sup>r</sup> Infrared Imaging System (LI-COR Biosciences). Densitometric analysis for each protein band was done using the Image Studio Software (LI-COR Biosciences). Each protein quantification was first normalized against GAPDH (loading control) from the same gel (intra-blot analysis), before comparisons for changes were made (inter-blot comparisons). The results (x-fold) are presented as means and standard deviations (SD) of three independent experiments. Statistics were analyzed using Student's t-test. P values of ≤ 0.05 were considered statistically significant.

#### In situ Hybridization

An ISH probe specific for all major App isoforms was designed to detect the juxtamembrane region of APP. To this end, a cDNA fragment encoding aa 492–623 of APP695 was cloned into the pcDNA3 vector. Prior to in vitro transcription, the plasmid was linearized and gel-purified using a gel extraction kit (Qiagen). in vitro transcription of DIG-labeled antisense RNA probe from the SP6 promoter was performed using the Roche DIG RNA labeling kit (SP6/T7), following the manufacturer's instructions. Probes were subsequently purified using RNase-free ChromaSpin 100 columns (Clontech). The quantity of labeled and purified probe was estimated by Dot blot as described in the DIG RNA labeling kit manual.

Whole mouse brains were dissected and immediately placed on dry ice until they were thoroughly frozen. Brain slices (14 µm) were cut on a cryostat (Zeiss Hyrax C50), collected on Superfrost plus slides (Thermo Scientific) and dried at 56◦C for 30 min. Sections were fixed for 10 min in 4% PFA in PBS, washed thrice in diethyl pyrocarbonate (DEPC)—treated PBS, and then permeabilized and acidified in triethanolamine hydrochloride (TEA-HCl)—acetic anhydride for 10 min. After three washing steps with DEPC-PBS, slices were dehydrated in an ethanol series (50%, 75%, 95%, 100%; 5 min each) and dried for at least 2 h at 56◦C. Anti-sense probe was diluted in hybridization buffer to the final concentration of approx. 400 pg/µl and heated to 80◦C for 10 min. After cooling down on ice, 100 µl of hybridization solution were applied to each slide, which was then covered with parafilm. Hybridization was done overnight at 56◦C. On the next day, slides were placed in 4× SSC for 10 min to wash off excess probe. Stringent washing steps were 30 min in 0.2× SSC at 60◦C, followed by another 90 min in fresh 0.2× SSC at 60◦C, followed by 10 min in 0.2× SSC at room temperature. For probe detection, slides were equilibrated in P1DIG (100 mM Tris-HCl; 150 mM NaCl) for 10 min and blocked in blocking solution (P1DIG + 0.5% BSA + 1% Blocking reagent, Roche) for 30 min. Brain slices were encircled with PAP PEN and anti-DIG-AP antibody (80 µl, diluted 1:500 in blocking solution) was pipetted on every brain slice. Antibody incubation was done overnight at 4◦C in a humidified chamber. The next day, all slides were washed twice for 15 min in P1DIG, then equilibrated in P3DIG (100 mM Tris-HCl; 100 mM NaCl; 50 mM MgCl2, pH 9.5) for 2 min. Slides were incubated in substrate solution (NBT/BCIP, diluted 1:50 in P3DIG) overnight at room temperature until color development was sufficient. Slides were then washed in PBS, fixed for 10 min in 4% PFA in PBS, washed in P4DIG (10 mM Tris-HCl; 1 mM EDTA, pH 8.0) for 10 min, then air dried for 2 h and finally mounted in Mowiol (Polysciences).

#### Digital Illustrations

Figures were prepared digitally using commercially available graphics software (Photoshop, Adobe Inc., San Jose, CA, USA). Fluorescent images were acquired using a digital camera (Digital Sight DS-M5c, Nikon, Germany) or confocal microscopy (Eclipse C1 Plus, Nikon). Single fluorescent images of the same section were digitally superimposed. The contrast, brightness and sharpness of images were adjusted as needed for each section. No additional image alteration was performed.

#### Laser Microdissection

Mice were killed by an overdose of isoflurane (Abbott). Brains were rapidly removed from the cranium, embedded in tissue freezing medium and immediately flash-frozen in −70◦C isopentane cooled by dry ice. Cryostat sections (8 µm for RNA analysis, 20 µm for western blotting) were mounted on polyethylene naphthalene (PEN) or polyester (POL) slides (Leica Microsystems). For RNA analysis, sections were fixed shortly in −20◦C cold acetone, stained with 1% cresyl violet staining solution and dehydrated in 75% and 100% ethanol. Using a Leica LMD6500 system (Leica Microsystems), defined tissue samples of the dentate gcl and of CA1 pcl were collected separately from the same brain sections and transferred to −80◦C until further processing.

#### RNA Isolation and Reverse Transcription

Total RNA was isolated using the RNeasy Plus Micro Kit (Qiagen) according to the manufacturer's recommendations. RNA integrity was assessed using the Agilent 2100 Bioanalyzer system and Agilent RNA 6000 Pico Kit (Agilent Technologies), and then reverse transcribed using High Capacity cDNA Reverse Transcription Reagents Kit (Applied Biosystems) following the manufacturer's recommendations.

# Quantitative Polymerase Chain Reaction (qPCR)

cDNA was amplified using TaqMan<sup>r</sup> Fast Universal PCR Master Mix (Applied Biosystems) and the StepOnePlus Real-Time PCR System (Applied Biosystems). PCR products were checked on Agilent DNA 1000 Chips (Agilent Technologies) with the Agilent 2100 Bioanalyzer system to verify product specificity and amplicon size. Quantification of the gene expression of candidate reference genes was carried out using SYBR<sup>r</sup> GreenERTM qPCR Supermix Universal (Invitrogen, Waltham, MA, USA) following the manufacturer's recommendations. Primer efficiencies and quantification cycle (Cq) values were calculated using LinRegPCR Software (Tuomi et al., 2010). To determine the most stable reference genes and the minimum number for accurate normalization, NormFinder (Andersen et al., 2004) and geNorm (Vandesompele et al., 2002) were used according to the developer's manuals. qPCR data were tested for statistical significance using one-way ANOVA followed by Bonferroni post hoc test to correct for multiple comparisons, <sup>∗</sup>p ≤ 0.05.

## RESULTS

#### APP is Differentially Expressed in the Principal Cell Layers of the Hippocampus

For immunohistochemical detection of APP protein in the adult mouse hippocampus, widely used antibodies against APP were selected which recognize the major isoforms of APP in the rodent brain, i.e., APP-770, APP-695 and APP-751. To address the specificity of these antibodies, we tested the antibodies on wild type (APP+/+) and APP deficient (APP−/−) brain tissue sections. Two of the antibodies, i.e., Y188 and CT20, that both recognize C-terminal APP epitopes showed immunoreactivity only in APP+/<sup>+</sup> brain sections but virtually no staining in APP−/<sup>−</sup> hippocampal tissue (**Figures 1A–J**). Using these antibodies, a considerably stronger fluorescence signal was observed in the principal layers of the Ammon's horn compared to the dentate gcl (**Figures 1A,B,D,F,G,I**). Non-specific immunoreactivity was moderately higher in APP−/<sup>−</sup> sections using CT20 compared to the Y188 antibody (**Figures 1C,E,H,J**). In contrast, the 22C11 antibody did not show specific staining (**Figures 1K–O**).

To quantify protein levels and to corroborate our immunofluorescence data, we performed double-fluorescence western blot analysis using whole hippocampal homogenates as well as laser microdissected tissue samples of CA1 pcl and DG gcl. Holo-APP (∼95–100 kDa) was recognized by all three APP antibodies in wild type but not in APP deficient tissue samples (**Figure 2**). CT20 and 22C11 demonstrated additional fragments of smaller size in both genotypes (**Figure 2**). The Y188 antibody appeared to be the most specific of the three, which was in line with our recent western blot results using this antibody indicating that it primarily detects full-length APP whereas the relative abundance of C-terminal stubs that are detected by this antibody is much lower (Fol et al., 2016). Based on these results, we chose Y188 to quantitatively determine APP in laser microdissected samples of hippocampal subregions (**Figure 3**).

In line with our immunofluorescence labeling, quantitative western blot analysis of microdissected tissue revealed a significantly higher APP protein level (approximately 1.7 fold) in the pcl of CA1 compared to the gcl of the DG (**Figure 3**). These

FIGURE 1 | Specificity of amyloid precursor protein (APP) antibodies tested on hippocampal sections of adult wild type and APP deficient mice. (A) Immunofluorescence of the dorsal hippocampus of wild type (+/+) mice using the Y188 antibody. The dentate gyrus (DG) shows only a weak signal, whereas a more intense labeling is seen in Ammon's horn (CA1–3). (B,C) Immunofluorescence is detectable in the granule cell layer (gcl) and molecular layer (ml) of the DG in wild type but not in brain sections of APP deficient mice using the Y188 antibody. (D,E) Principal cell layer (pcl) of CA1 shows a strong signal in wild type mice. Some immunofluorescence is also seen in stratum radiatum (sr). In contrast, staining is absent in APP-deficient hippocampal tissue. (F–J) Immunofluorescence of the hippocampus of wild type and APP deficient mice using the CT20 antibody. (F,G,I) Similar to the results with Y188 a stronger signal can be seen in Ammon's horn (CA1–3) of wild type mice compared to the DG. (H,J) Background staining is slightly higher in APP-deficient tissue sections compared to the background seen with the Y188 antibody (in C,E). (K–O) Immunostaining using the 22C11 antibody shows similar staining in wild type and APP-deficient tissue, suggesting that this antibody is not sufficiently specific to identify APP in tissue sections (K), DG (L,M) and CA1 (N,O). Scale bars: (K) 500 µm; (M,O) 25 µm.

data confirmed our initial impression that APP is differentially expressed in these two hippocampal subregions.

# APP is Predominantly Expressed by Neurons in the Adult Mouse Hippocampus

To elucidate, which hippocampal cell types produce relevant amounts of APP protein, we performed confocal doubleimmunofluorescence analysis using Y188 in combination with the neuron-specific marker NeuN (neuronal nuclear antigen) or the astrocytic marker GFAP (glial fibrillary acidic protein; **Figure 4**). We performed this staining since earlier publications, which were in part performed in tissue cultures, had also suggested an astroglial expression of APP (Golde et al., 1990; Haass et al., 1991; LeBlanc et al., 1991). In our preparations, we found that APP is predominantly expressed by hippocampal neurons (**Figures 4A–D**). In contrast, we did not detect an astroglial APP expression (**Figures 4F–J**). Of note, APP-positive

neurons were not only found in the principal cell layers of the hippocampus, i.e., in pcl of Ammon's horn and, to some weaker extent, in the dentate gcl, but also in adjacent layers, e.g., hilus, stratum radiatum (sr) or stratum lacunosum-moleculare (slm; **Figures 4A,C**).

To also identify App mRNA-expressing cells in the hippocampus, we next performed non-radioactive ISH using a digoxygenin-labeled riboprobe. This probe detects an mRNA sequence corresponding to the juxtamembrane region of APP, which is present in all major APP isoforms, i.e., APP-770, APP-751 and APP-695, but not conserved in the related APLPs. Strongly App mRNA-expressing cells were detected in the pcl of Ammon's horn and in the hilus of the DG, whereas only a comparatively weak ISH signal was observed in the dentate gcl (**Figures 5A,C**). Outside the principal cell layers, only few App mRNA-expressing cells could be found in the adjacent layers, e.g., sr or slm (**Figure 5D**), which is an expression pattern that corresponds to the expression of APP by interneurons (Wang et al., 2014) but not by astroglia. Hippocampal tissue of APP deficient mice served as negative control and was stained using the same anti-sense riboprobe. This experiment revealed only weak non-specific background (**Figure 5B**).

Together, the results obtained by ISH, immunofluorescence and western blot analysis suggest that App mRNA as well as APP protein are predominantly expressed by principal neurons but not by astroglial cells in the adult mouse hippocampus.

# Quantitative Analysis of App mRNA Expression in Hippocampal Subregions

By using qPCR in combination with LMD, we aimed to compare App mRNA expression levels in the pcl of CA1 compared to the gcl of the DG. For this purpose, only high quality

RNA samples (RIN-values: ∼9) of laser microdissected cell layers were used (**Figures 6A–C**). To more reliably analyze possible differences in gene expression, we first validated a panel of suitable reference genes (see **Table 1** for details) for both hippocampal subregions in order to achieve robust qPCR data. Two established and widely accepted algorithms, i.e., geNorm and NormFinder, were used for the expression stability ranking of reference genes for CA1 and DG (**Table 2**). As determined by pairwise variation using geNorm and accumulated SD analysis according to NormFinder, the most stable reference genes as well as the minimal number necessary for accurate normalization were determined (**Figures 6D,E,G,H**). Of note, both algorithms showed a comparable ranking for all of the candidate reference genes tested (**Table 2**; **Figures 6D,G**).

Based on this data set, we used a normalization index out of the two most stable reference genes as well as the best combination of suitable genes, i.e., Gapdh and Sdha for geNorm, and Gapdh and Pgk1 for NormFinder, respectively. For App gene expression analysis, two different qPCR assays specific for all major App isoforms were selected, which detected almost identical gene expression levels. Using this strategy, we determined a significantly higher expression of App mRNA (1.5- to 1.7-fold) in CA1 pyramidal cells compared to granule cells of the DG using either of the reference gene indices for accurate normalization of qPCR data (**Figures 6F,I**).

#### DISCUSSION

In the present study, we analyzed the expression of APP at the protein and mRNA level in the gcl and CA1 pcl of the

(F–J) Double-immunofluorescence staining for APP (Y188, red) and the astrocytic marker GFAP (green) revealed no APP expression by this glial cell population in the adult mouse hippocampus. DRAQ5 was used to visualize cell nuclei (blue). Scale bar: (A,F) 200 µm; (B,G) 25 µm; (E,J) 12.5 µm.

adult mouse hippocampus using confocal immunofluorescence, ISH and LMD in combination with qPCR or western blot analysis. Our main findings can be summarized as follows: full-length APP is expressed by neurons under physiological conditions. APP expression is ∼1.7× stronger at both mRNA and protein level in CA1 pyramidal cells compared to dentate granule cells. We propose that these differences in basal APP expression may contribute to the regional differences in APP function we reported in earlier studies using APP−/<sup>−</sup> animals (Ring et al., 2007; Jedlicka et al., 2012).

## Endogenous Full-Length APP Levels in the Mouse Hippocampus—Methodological Considerations

Quantification of endogenous APP levels in the brain is confounded by the fact that some commercially available antibodies recognize not only full-length APP but also APP cleavage products and/or other protein fragments (Guo et al., 2012). In addition, antibodies may cross-react with the highly homologous APLPs, which further limits antibody specificity (Slunt et al., 1994). Thus, we ensured using tissue of APP−/<sup>−</sup> mice that the antibodies we used for immunohistochemistry and western blot analysis in this study are highly specific and can be employed to detect holo-APP in the mouse

hippocampus with high reliability. Furthermore, since APP is expressed in different isoforms in the nervous system and the brain (Kang and Müller-Hill, 1990; Sisodia et al., 1993), we designed probes for ISH and primers for qPCR which detect the three major isoforms of APP. Choice of these tools for our quantitative analysis make us confident that we predominantly measured total full-length APP mRNA and protein in our study.

lacunosum-moleculare (slm). Scale bars: (A) 100 µm; (C,D) 50 µm.

Furthermore, since we were specifically interested in the neuronal expression of APP in these two regions and since our immunohistochemistry revealed a neuron-specific expression pattern of APP in the hippocampus (see below) we used LMD to selectively harvest the neuronal cell layers, i.e., the gcl of the DG and the CA1 pcl, respectively. This approach makes our quantification quite specific for granule cells and CA1 pyramidal cells, since the number of principal cells by far exceeds the number of cells of other cell types in these layers. Thus, we are confident that we here report robust and reliable data on the relative expression of APP mRNA and protein in the principal neurons of two major subfields of the hippocampus.

# Full-Length APP is Expressed by Neurons in the Mouse Hippocampus

In the adult rodent CNS, three major APP isoforms encoded by alternatively spliced transcripts have been described. In line with Guo et al. (2012), our data indicate that in tissue of intact and

Representative section of the dorsal hippocampus (coronal plane, cresyl violet staining) before (A) and after (B) LMD is shown. Scale bar: 250 µm. (C) RNA integrity analysis of total RNA isolated from the dissected gcl (red) and from pcl (blue) demonstrating highly intact RNA (RIN-values: ∼9; Agilent 2100 Bioanalyzer). (D,E) Average expression stability values (M) and evaluation of the optimum number of candidate reference genes for CA1 pcl and for DG gcl according to geNorm software. Pairwise variation (V) of candidate reference genes indicates that the use of the two most stable genes is sufficient to obtain an accurate normalization index for quantitative PCR (qPCR) analysis. (F) A significantly higher gene expression for App (1.5- to 1.7-fold) was detected in CA1 pcl relative to DG gcl using two different App-specific TaqMan assays (A: Mm\_01344172\_m1, B: Mm\_00431830\_m1) after normalization to a reference gene index calculated by geNorm. (G,H) Gene expression stability values (S) and accumulated SD analysis using NormFinder. The minimal number of reference genes required for effective normalization is highlighted. (I) Comparable to the results obtained by geNorm algorithm, a significantly higher App expression (1.6- to 1.7-fold) was detected in CA1 pcl relative to DG gcl after normalization to the reference gene index estimated by NormFinder. Data (N = 5–6 mice) were tested for statistical significance using one-way ANOVA followed by Bonferroni post hoc test to correct for multiple comparisons, <sup>∗</sup>p ≤ 0.05.

otherwise untreated mouse brain endogenous APP is expressed selectively by neurons but not astroglia: neither immunostaining with the APP-specific antibody Y188 nor ISH with App-specific riboprobes revealed a glial expression pattern. Similarly, doublelabeling for neuronal and astroglial markers revealed a highly selective neuronal expression. In culture, however, previous studies reported that both astrocytes and microglia express APP (Haass et al., 1991; LeBlanc et al., 1991; Forloni et al., 1992; Mönning et al., 1995) and during aging Aß production has also been reported from non-neuronal sources in transgenic APP overexpressing mice. Thus, the possibility exists that glial cells express APP under reactive conditions in vivo. This issue was previously addressed by Guo et al. (2012), who used a traumatic brain injury model and an AD mouse model and failed to detect APP-positive astrocytes using APP-specific antibodies. They concluded that in vivo APP levels in astrocytes may be too low for detection even under reactive conditions (Guo et al., 2012). In our own investigations, in which we used entorhinal cortex lesions (Lynch et al., 1978; Steward, 1994; Deller and Frotscher, 1997) to denervate the DG, we also failed to see an




increase in App mRNA in the denervated outer molecular layer (Del Turco et al., 2014). In this layer, reactive glia are particularly abundant (Deller et al., 2000, 2007; Del Turco et al., 2014). Although these two reports cannot rule out the possibility that under some other conditions APP is expressed in vivo by glial cells, they certainly suggest that glial APP is not the primary source of APP in the intact or injured brain.

#### Endogenous Neuronal App mRNA Levels are Tightly Controlled

It has been pointed out by others that App is regulated very much like a housekeeping gene (Dawkins and Small, 2014). The fact that the App promoter lacks TATA and CAAT boxes but contains sites for several transcription factors regulating the expression of proteins associated with cell proliferation and differentiation suggests that App mRNA levels are primarily regulated during development (Izumi et al., 1992; Clarris et al., 1995; for review see Dawkins and Small, 2014). In the adult brain App mRNA levels may be much more tightly controlled to supply neural tissue with a constant level of APP protein for further processing.

However, a certain degree of transcriptional regulation has been reported for APP and App mRNA in adult neurons following brain injury (Murakami et al., 1998; Van Den Heuvel et al., 1999, 2007; Itoh et al., 2009). This appears to be of relevance, since head trauma is considered a risk factor for AD (e.g., Mortimer et al., 1991; Szczygielski et al., 2005). Concerning this lesion-induced regulation, the experimental literature is somewhat controversial (Szczygielski et al., 2005). By hindsight this is not surprising since many different antibodies and probes were used and some of them may not have been tested for specificity. In our own investigations using the entorhinal cortex lesion model we initially failed to observe an increase in App mRNA using screening methods. Only after using the sensitive LMD/qPCR approach (Burbach et al., 2003), which made it possible to measure App mRNA within microdissected cell and fiber layers did we detect a ∼1.3 fold increase of App mRNA in denervated granule cells at



Expression stability values of candidate reference genes for CA1 pyramidal cell layer (pcl) and for granule cell layer (gcl) of the dentate gyrus (DG) calculated by geNorm (M-values) and NormFinder (S-values) algorithms.

7 days post lesion (Del Turco et al., 2014). We conclude that neuronal App expression is tightly regulated and even under extreme conditions, e.g., brain injury, App gene expression changes range between 1- to 2-fold. If translated 1:1 into protein, as our present study suggests, such an increase in App mRNA may, however, be biologically and pathophysiologically relevant.

Finally, it should be kept in mind that transcriptional regulation of App is only one regulatory step under physiological and pathophysiological conditions, likely limiting the amount of full-length APP protein available for downstream processing. Post-transcriptional regulation by miRNAs has also been recently described (Schonrock et al., 2012). Most importantly, however, the amount, availability and activity of the secretases eventually decide the ''biological fate'' of the full-length protein by liberating its biologically active fragments. In contrast to the levels of App mRNA, which appear to be tightly controlled and provide neurons with a basal supply of APP, the activity and/or expression of secretases is regulated by neuronal activity and many other conditions, which have been reviewed elsewhere (Endres and Fahrenholz, 2012; Sun et al., 2012; Vassar et al., 2014; Vincent, 2016).

# Basal Expression of APP is Higher in CA1 Pyramidal Neurons Compared to Dentate Granule Cells

ISH against endogenous App mRNA revealed a weaker labeling of dentate granule cells compared to the pyramidal cells of Ammon's horn. This made us wonder whether this reflected a true regional difference between the DG and the other hippocampal subfields. Since non-radioactive ISH cannot be reliably used for quantitative analysis, we quantified App mRNA using qPCR. Using LMD, the principal cell layers were harvested, which reduced dilution effects. The careful selection of reference genes using current recommendations for qPCR (Vandesompele et al., 2002; Andersen et al., 2004) ensured a very robust reference for the subregional comparison. In sum, this revealed a 1.5- to 1.7-fold difference in App mRNA expression between the DG and area CA1. The difference in App mRNA level translates into protein, since we used the same LMD approach to obtain the tissue for western blot analysis and found a comparable difference for APP protein, i.e., ∼1.7-fold more protein in area CA1 compared to the DG.

## Regional Differences in APP Expression May Contribute to Regional Differences in Synaptic Plasticity

The physiological role of APP has been investigated in the hippocampus using APP−/<sup>−</sup> mice. This loss-of-function approach revealed a robust role for APP in synaptic plasticity at the CA3-CA1 synapse (Dawson et al., 1999; Ring et al., 2007). Animals lacking APP showed an impaired LTP, which went hand-in-hand with memory dysfunctions. In contrast, using similar stimulation protocols for synaptic strengthening, the same line of APP−/<sup>−</sup> mice did not show an LTP-defect at the EC-GC synapse in vivo (Jedlicka et al., 2012). This was a somewhat surprising result and we suggest—based on the data reported in this article—that differences in APP expression level between the two regions might contribute to the functional differences seen in our recordings.

How could different APP levels in neurons contribute to differences in synaptic function? Although the physiological role of APP is not yet fully understood several recent publications have suggested important functions for APP and its cleavage products at central synapses. With regard to full-length APP, it has been shown that it can act as a cell-adhesion molecule in trans, i.e., linking pre- and postsynapse, thus affecting the stability of synapses (Soba et al., 2005; Stahl et al., 2014). On the presynaptic side, APP regulates the abundance of synaptic vesicle proteins and may impact on synaptic transmission (Laßek et al., 2013, 2014, 2016a,b; Fanutza et al., 2015). This presynaptic effect is in line with our own findings, which indicate that lack of APP causes presynaptic changes at the EC-GC (Jedlicka et al., 2012) as well as the CA3-CA1 synapses (Hick et al., 2015). On the postsynaptic side, sAPPα, which is generated from APP by α-secretase cleavage, appears to be required for synaptic strengthening. Experiments using sAPPα-binding antibodies and recombinant sAPPα (Turner et al., 2003; Taylor et al., 2008) as well as our own approaches using mouse genetics (Ring et al., 2007; Hick et al., 2015) revealed an essential function of this fragment in Hebbian-plasticity at both synapses. Most likely, the APP effect on synaptic plasticity is caused by an increased delivery of NMDAR to synapses (Cousins et al., 2009; Hoe et al., 2009), resulting in increased NMDAR currents (Taylor et al., 2008). In conclusion, APP and its cleavage products influence synaptic function at both pre- and postsynapse. It is thus highly likely that regional differences in APP levels could impact on the effect size experimenters can observe using APP-KO mice.

Unraveling and understanding the role of APP at central synapses is non-trivial and may ultimately require synapsespecific answers. In addition to the above discussed differences in APP levels, differences in APP processing and thus the abundance of specific fragments such as sAPPα between brain regions may also play an important role. Likewise, regional differences in the expression of APP-like proteins, i.e., APLP1 or APLP2, which can partially compensate for a loss of APP could affect the interpretation of loss-of-function experiments (von Koch et al., 1997; Heber et al., 2000; Weyer et al., 2011; Hick et al., 2015; Vnencak et al., 2015). Regardless of all these considerations, however, APP can only play a role in synaptic plasticity of a synapse if it is present. If not, other factors will predominate. Thus, we feel confident to conclude that APP plays a greater role for synaptic plasticity in area CA1 compared to the DG in mice. This finding, which implies that some effects of APP are region-specific, may be of relevance for future studies on APP and may also

#### REFERENCES


affect the design and analysis of APP-related animal models of AD.

#### AUTHOR CONTRIBUTIONS

DDT, MHP, JS, and MH performed experiments. DDT and TD conceived the study. All authors were involved in data interpretation. DDT and TD wrote the manuscript with contributions from all other authors.

#### FUNDING

This research was funded by the Deutsche Forschungsgemeinschaft (DFG FOR 1332 to TD and UCM).


RNA in response to traumatic brain injury: an ovine head impact model. Exp. Neurol. 159, 441–450. doi: 10.1006/exnr.1999.7150


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Del Turco, Paul, Schlaudraff, Hick, Endres, Müller and Deller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Novel Insights into the Physiological Function of the APP (Gene) Family and Its Proteolytic Fragments in Synaptic Plasticity

#### Susann Ludewig<sup>1</sup> and Martin Korte1, <sup>2</sup> \*

*<sup>1</sup> Division of Cellular Neurobiology, Zoological Institute, TU Braunschweig, Braunschweig, Germany, <sup>2</sup> Helmholtz Centre for Infection Research, AG NIND, Braunschweig, Germany*

The amyloid precursor protein (APP) is well known to be involved in the pathophysiology of Alzheimer's disease (AD) via its cleavage product amyloid ß (Aß). However, the physiological role of APP, its various proteolytic products and the amyloid precursor-like proteins 1 and 2 (APLP1/2) are still not fully clarified. Interestingly, it has been shown that learning and memory processes represented by functional and structural changes at synapses are altered in different APP and APLP1/2 mouse mutants. In addition, APP and its fragments are implicated in regulating synaptic strength further reinforcing their modulatory role at the synapse. While APLP2 and APP are functionally redundant, the exclusively CNS expressed APLP1, might have individual roles within the synaptic network. The proteolytic product of non-amyloidogenic APP processing, APPsα, emerged as a neurotrophic peptide that facilitates long-term potentiation (LTP) and restores impairments occurring with age. Interestingly, the newly discovered η-secretase cleavage product, An-α acts in the opposite direction, namely decreasing LTP. In this review we summarize recent findings with emphasis on the physiological role of the APP gene family and its proteolytic products on synaptic function and plasticity, especially during processes of hippocampal LTP. Therefore, we focus on literature that provide electrophysiological data by using different mutant mouse strains either lacking full-length or parts of the APP proteins or that utilized secretase inhibitors as well as secreted APP fragments.

#### Edited by:

*Jason D. Shepherd, University of Utah, USA*

#### Reviewed by:

*Inna Slutsky, Tel Aviv University, Israel Nicolas Sergeant, French Institute of Health and Medical Research (Inserm), France*

\*Correspondence:

*Martin Korte m.korte@tu-bs.de*

Received: *01 November 2016* Accepted: *14 December 2016* Published: *20 January 2017*

#### Citation:

*Ludewig S and Korte M (2017) Novel Insights into the Physiological Function of the APP (Gene) Family and Its Proteolytic Fragments in Synaptic Plasticity. Front. Mol. Neurosci. 9:161. doi: 10.3389/fnmol.2016.00161* Keywords: amyloid precursor protein, amyloid precursor-like protein, long-term potentiation, synaptic plasticity

### INTRODUCTION

The amyloid precursor protein (APP) gene is localized in humans on chromosome 21 and its expression gives rise to three major isoforms (APP695, APP751, APP770; around 170 kDa) generated via alternative splicing. APP695 is the predominant isoform in neurons (Robakis et al., 1987; Yoshikai et al., 1990). APP is translated in the endoplasmatic reticulum (ER) where it forms stable dimers which are transported through the secretory pathway via the Golgi apparatus to the cell surface (Isbert et al., 2012; Tan and Evin, 2012). APP is classified as a type I transmembrane glycoprotein with one membrane spanning domain, a large extracellular N-terminus and a small intracellular C-terminus (Dyrks et al., 1988). The mammal APP is part of a larger gene family including the homologs amyloid precursor-like proteins 1 and 2 (APLP1 and APLP2), both of which are expressed throughout the body nervous system (brain, spinal cord, retina), immune system (thymus, spleen), muscle (smooth, cardiac, and skeletal), kidney, lung, pancreas, prostate gland, and thyroid gland (Wasco et al., 1993; Liu et al., 2008; Aydin et al., 2012). Interestingly, the APP and APLP2 proteins are found at particularly high levels in the brain where their expression patterns largely overlap in pyramidal neurons of the cortex and hippocampus (Bendotti et al., 1988; Lorent et al., 1995). Thereby, the APP isoform APP<sup>695</sup> is especially found in excitatory neurons as well as in GABAergic interneurons while the expression of the other two isoforms, 751 and 770, is assigned to other cell types (Wang et al., 2014; Hick et al., 2015). In vitro studies revealed APP expression in astrocytes and microglia that is increased following brain injury (LeBlanc et al., 1997; Rohan de Silva et al., 1997). On the other hand a more recent study reported that APP expression is restricted to neurons and cannot be found in major glial cells like astrocytes or microglia under basal as well as neuroinflammatory conditions (Guo et al., 2012). These contradictory results are possibly due to the lack of APP specific antibodies. The highly homologous APP family members differ only slightly in their peptide domain structure and hence are displaying a similar proteolytic processing. The relatively short intracellular part of the C-terminus of APP and related proteins contains a YENPTY peptide motif which was shown to promote clathrin mediated endocytosis, modulate Aβ generation, interfere with Ca2<sup>+</sup> homeostasis, and interact with multiple kinases, and adapter proteins (Perez et al., 1999; Leissring et al., 2002; Ring et al., 2007; Jacobsen and Iverfeldt, 2009). The extracellular part of APP is composed of the large E2 and E1 domains containing interaction sites for multiple binding partners like F-spondin, LRP1, Nogo-66 receptor, Notch 2, Netrin, Alcadein, sorL1/LR11, and extracellular matrix components (Müller and Zheng, 2012). Additionally, the E1 domain could be demonstrated to be crucial for the homo- and heterodimerization of APP family members (Soba et al., 2005). Interestingly, the Aβ motive, which is highly conserved in mammals and zebrafish is unique for APP. The APLPs lack this sequence.

Although the structure of both APP and APLPs are well known, the precise cellular function of these proteins remains elusive. For instance, extensive posttranslational modifications and the various cleavage products of APP and APLP processing complicate precise investigations. Nevertheless, several studies assessed putative cellular functions of the APP family members during development and in the adult nervous system (Jacobsen and Iverfeldt, 2009). Certainly, one of the most intriguing discoveries in this respect is the involvement of APP and its cleavage products in processes of synaptic plasticity (Korte et al., 2012) at which activity patterns generated by experience are able to modify neuronal function and structure. These include activity-dependent alterations of the efficacy of synaptic transmission and changes in the structure and number of synaptic connections (for a review see Korte and Schmitz, 2016). Part of the pathophysiology of Alzheimer's disease (AD) is related to the malfunction of synapses (Selkoe, 2002) and the application of amyloid beta (Aß) oligomers has been shown to directly impair synaptic plasticity (Shankar et al., 2008). Despite a huge amount of data which looked at the pathophysiological role of Aß plaques, it is less clear what the physiological function of APP and its fragments (including Aß) might be. In addition to APP, it is also important to further the understanding of the putative physiological functions of the related APLP1 and APLP2 proteins and their cleavage products. In this review we concentrate on the role of APP, APLP1, APLP2, and their proteolytic fragments in processes of synaptic transmission and in particular synaptic plasticity under physiological conditions (**Table 1**).

# ROLE OF FULL-LENGTH APP PROTEINS AT THE SYNAPSE

Gene targeting of APP protein family members provides a powerful tool to investigate the proteins functions. Studying adult APP and APLP2 single KOs in synaptic plasticity revealed only subtle phenotypes (von Koch et al., 1997) mainly due to the overlapping ubiquitous expression of the two proteins in mammals and their similar processing (see **Figure 1**). Under steady state conditions, the majority of full-length APP is located in the Golgi apparatus and in the trans-Golgi network (Thinakaran and Koo, 2008). When present at the plasma membrane APP and APLPs were shown to form homo- and heterotypic cis interactions and have been proposed to mediate cell–cell interactions in trans (Soba et al., 2005; Kaden et al., 2009; Baumkötter et al., 2012; Mayer et al., 2016). Synaptic adhesion by APP might not only be crucial to build and maintain synaptic contacts, but also to regulate synaptic plasticity (see **Figure 2**). Highest expression levels at the membrane were observed for APLP1 suggesting that it might be the family member with the upmost potential to mediate cell contacts (Kaden et al., 2009). Recently, the study of Mayer et al. (2016) identified APP and APLP2 to exhibit basal adhesive properties while APLP1 mediated neuronal adhesion is dynamic and regulated by zinc. Copper was instead shown to induce cis- and transdimerization of APP at its E1 domain (Baumkötter et al., 2014). Importantly enhanced trans or cis interaction of APPs or APLPs is accompanied by a reduction of ectodomain shedding of the proteins (Stahl et al., 2014; Mayer et al., 2016) and might therefore interfere with the ability to modulate synaptic function.

# APP-KO

The well-studied constitutive KO of APP in mice leads to an age-related deficit in synaptic plasticity, mainly in long-term potentiation (LTP, see **Box 1** for definition). LTP reflects the increase in synaptic strength that lasts for at least 1 h and is paralleled by alterations at the contact sites between nerve cells, the presynapse (axonal boutons) and postsynapse (dendritic spines). No alterations in synaptic plasticity, the cellular correlate for learning and memory (Stuchlik, 2014) were found in young mice accompanied by normal basal synaptic transmission properties and short-term synaptic plasticity (STP) paralleling the intact behavioral learning of adult and impaired performance of aged mice (Seabrook et al., 1999; Ring et al., 2007; and reviewed by Turner et al., 2003; Korte et al., 2012). The age-dependent LTP defect is further supported by the electrophysiological

#### TABLE 1 | Electrophysiological characteristics of the APP protein family members and their proteolytic domains.


*(Continued)*

TABLE 1 | Continued


FIGURE 1 | Proteolytic processing of APP. Full-length APP can be processed by α-, ß-, η-, and γ-secretases in three different pathways. The left panel illustrates the η-secretase processing of APP. Initially η-secretase cleavage releases the soluble APPsη, while CTFη remains embedded in the membrane. It is further processed by α- or ß-secretase at the extracellular side generating An-α or An-ß. Shedding of CTFη within the transmembrane domain by γ-secretase yields the APP intracellular domain (AICD) containing the highly conserved interaction motif (YENPTY, yellow box) or the short extracellular peptides Aß seen in the amyloidogenic or p3 within the non-amyloidogenic pathway. The non-amyloidogenic pathway depicted in the middle is driven by the α-secretase liberating APPsα in the extracellular space. Subsequently processing of membrane tethered CTFα by γ-secretase generates the p3 peptide and cytoplasmic AICD. The right panel illustrates APP processing in the amyloidogenic pathway by ß-secretase resulting initially in the release of the APPsß ectodomain. Following γ-secretase shedding of the membrane tethered CTFß the Aß peptide is secreted along with AICD in the cytoplasm.

measurements of murine organotypic hippocampal slice cultures (OHCs) from APP-KO pups prepared at postnatal day zero. No differences in the Input–Output characteristics and STP of APP-KO in comparison to wild-type OHCs were observed (Weyer et al., 2014). In agreement, the loss of APP does not impair synaptic plasticity in the adult organism and thereby APLP2 and maybe APLP1 are considered to perform redundant functions, but fail to compensate for APP deficiency with age.

#### APLP2-KO

The function of APLP2 in synaptic plasticity has also been addressed in detail since this protein shares the highest degree of sequence homology with APP within the gene family. Furthermore, the spatial and temporal expression pattern of APLP2 is highly reminiscent to that of APP (Wasco et al., 1993). APP and APLP2 are ubiquitously expressed in the nervous tissue and at the neuromuscular junction (NMJ, Slunt et al., 1994; Lorent et al., 1995) as well as in pyramidal and GABAergic neurons of the hippocampus and cortex (Wang et al., 2014; Hick et al., 2015). In contrast to APP-KO mice, young and aged APLP2 single KOs behave like wild-type mice showing no impairments in LTP, STP, PPF, or basal synaptic transmission (Weyer et al., 2011; Midthune et al., 2012). These observations go in line with normal learning and memory performance in cognitive tasks like the Morris-Water-Maze (MWM) or the passive avoidance test (Heber et al., 2000; Guo et al., 2012). The functional effects are consistent with investigations of dendritic spine numbers at excitatory neurons, reflecting the number of excitatory synapses. Whereas, the spine density assessed in vivo was affected in aged APP-KO animals, it was unaltered in APLP2-KO mice as well as in APLP2 OHCs in vitro (Lee et al., 2010; Midthune et al., 2012; Weyer et al., 2014). It seems likely that endogenous APP is able to compensate for the genetic ablation of APLP2 with age, while vice versa APLP2 is incapable to compensate the loss of APP in aged animals. This implicates that APP has either different or dominant neuronal functions compared to APLP2.

#### BOX 1 | Term definitions.

Synaptic plasticity designates the activity-dependent alterations of the efficacy of synaptic transmission and changes in the structure as well as number of synaptic connections whereby activity patterns are generated by experience. Synaptic connections build the contact sites between nerve cells and alterations at these contact sites provide the basis to store memories and information within neuronal networks (Korte and Schmitz, 2016).

LTP—Long-term potentiation is defined as a persistent increase in synaptic strength lasting for at least 1 h (Bliss and Lomo, 1973). It consists of an induction phase, including processes that trigger the alterations leading to the changes in synaptic efficacy followed by the expression or maintenance phase of LTP. LTP can be divided in different types: LTP lasting from 1 to 3 h is independent of transcription and translation and named early or E-LTP; if it lasts longer than 3 h, it is generally dependent on altered gene expression and referred to as late LTP (L-LTP, Bliss and Collingridge, 1993; Kandel, 2001).

LTD—Long-term depression is the counterpart of LTP and therefore defined as a persistent reduction in synaptic strength. LTD prevents excessive synaptic activity (Korte and Schmitz, 2016).

STP—Short term synaptic plasticity is a form of synaptic plasticity that is NMDA-R dependent, but presynaptically expressed. It depends on the frequency of induction as well as subsequent activity and lasts from ms to min (Zucker and Regehr, 2002; Volianskis and Jensen, 2003).

PPF—Paired-pulse facilitation is a NMDA receptor-independent form of short-term plasticity and a typical presynaptic phenomenon. The facilitation is caused in the process of re-establishment of intracellular Ca2<sup>+</sup> levels after repetitive Ca2<sup>+</sup> influx into the presynaptic terminal. PPF can be investigated by applying two single stimuli spaced by a defined time interval. Depending on the length of the Inter-Stimulus-Interval and type of stimulus used the second signal is facilitated or depressed (Paired-pulse depression, PPD). At shorter ISIs of <20 ms PPD is observed whereas larger ISIs >20 ms lead to PPF (Zucker and Regehr, 2002).

Spine density—Spines are small membrane protrusions from dendrites often with a neck-head structure building the postsynaptic elements of glutamatergic synapses (Korte and Schmitz, 2016). Their density can therefore be seen as correlate of the amount of excitatory synapses and often represents functional changes in synaptic strength.

#### APLP1-KO

Despite the generation and first characterization of the conventional APLP1-KO mouse in 2000 by Heber and colleagues, the function of this homolog has been less attended in synaptic plasticity Since APLP1 is the only APP family member with restricted expression to the brain (Lorent et al., 1995; Thinakaran and Koo, 2008; Klevanski et al., 2014), it is intriguing to speculate that APLP1 has a unique neuronal role and therefore might also be of particular importance for synaptic plasticity. However, Heber et al. (2000) described only minor (if any) distinct phenotypes of APLP1-KO. The ablation of the APLP1 gene function did not result in impaired cognitive behavioral performance in the MWM task but rather. However, during the behavioral paradigm it has been noted that depletion of APLP1 resulted in an improvement of acquisition learning. The in vivo analysis at the perforant path-granule cell synapse (PP-DG) in young adult mice (16–20 weeks old) revealed unaltered STP and LTP, associated to enhanced excitatory transmission (Vnencak et al., 2015). The authors argued that maybe a larger number of perforant path synapses or an increased synaptic strength in APLP1-deficient mice may cause this enhancement, but final clarification is missing. Furthermore, the paired-pulseinhibition (PPD) paradigm of the population spike points toward decreased GABAergic network inhibition in APLP1-KOs, an effect observed also for other APP-KO models.

#### ROLE OF THE APP PROTEIN FAMILY IN SYNAPTIC INHIBITION

The hippocampus is comprised of 95% excitatory and 5% inhibitory neurons, both expressing the APP family proteins (Hick et al., 2015). It is well established that the GABAergic system is especially important during the induction of LTP (Bliss and Lomo, 1973) and that excitation and inhibition must be tightly balanced for a well-coordinated network. This notion is supported by the finding that the inhibition of GABA<sup>A</sup> receptors facilitates LTP and leads to hyperexcitability causing epileptic

Frontiers in Molecular Neuroscience | www.frontiersin.org January 2017 | Volume 9 | Article 161

seizures (Gustafsson and Wigström, 1988; Casasola et al., 2004). Hippocampal hyperactivity is a hallmark of neurological diseases like mild cognitive (MCI, Bakker et al., 2012) and AD (Palop et al., 2007). Several studies suggest that the hyperactivity is caused by APP overexpression (Born et al., 2014) while others assume Aß to be the trigger (Busche et al., 2008; Minkeviciene et al., 2009). The APP family proteins seem to be closely involved in regulating GABAergic transmission as both APLP1-KO and aged APP-KO mice exhibit reduced GABAergic mediated PPD responses (Seabrook et al., 1999; Vnencak et al., 2015) and in addition increased susceptibility for kainite-induced seizures (Steinbach et al., 1998). Moreover, supporting the role of APP within the GABAergic network are the chronic reduction of GABA<sup>A</sup> receptors and the lowered number of GABA<sup>B</sup> autoreceptors mediating PPD of inhibition in the absence of APP (Fitzjohn et al., 2000) as well as the identified interaction of APP with GABA<sup>B</sup> receptors in vitro (Norstrom et al., 2010) as well as recently in vivo (Schwenk et al., 2016). Like in APP-KO, in mice expressing only the secreted APPsα on an APLP2 deficient background (APPsa-DM; Weyer et al., 2011), the neutralization of GABA<sup>A</sup> receptors by picrotoxin rescues impaired LTP presumably due to a facilitation of postsynaptic depolarization. Moreover, while addressing oscillatory activity by recording local field potentials (LFPs) in the dorsal hippocampus revealed normal theta- and gamma-frequency bands the coupling of gamma amplitude to the theta phase was diminished in around 9 months old APP-KO mice (Zhang et al., 2016). This observation indicates the presence of alterations within the local inhibitory networks (Zhang et al., 2016) thereby preventing a coordinated neuronal communication. Investigations by Yang et al. (2009) yielded that deletion of APP in hippocampal neurons increased L-type voltage gated Ca2<sup>+</sup> channel (LTCC) levels and function underlying an altered GABAergic STP. Likewise, a recent report implied APP possibly via the APPsα fragment to stabilize Ca2<sup>+</sup> homeostasis by regulating inhibition of LTCCs (Hefter et al., 2016). Nevertheless, APLP1 deficiency causes no LTP deficit even though GABAergic inhibition is affected in APLP1-KO mice. The related proteins, APP and APLP2, might

exhibit similar interactions at the presynaptic membrane and thus possibly compensate for the functional loss of APLP1 at the postsynaptic density (PSD) during LTP induction and maintenance.

#### APP AND APLP2 DOUBLE KO

The high content of APP and APLP2 especially in pyramidal cells of the cortex and hippocampus (Lorent et al., 1995) and their localization at synaptic sites (Laßek et al., 2013) suggest a role in synaptic transmission and synaptic plasticity. To address the function of these redundantly expressed proteins, combined KO models are necessary. Unfortunately, APP and APLP2 double KO (DKO) mice die perinatally (von Koch et al., 1997; Heber et al., 2000) indicating an indispensable role for both these proteins during development. The lethal phenotype of these DKO mice is most likely due to important functions of APP and APLP2 at the NMJ (Wang et al., 2005; Weyer et al., 2011) and reviewed by Caldwell et al. (2013). Neuromuscular transmission is severely impaired due to a reduced amount of synaptic vesicles and their impaired release. While the Knock-In of APPsα in the APP/APLP2-DKO mouse (APPsα-DM) rescued the lethal phenotype it resulted in muscular weakness and severe alterations in NMJ morphology (Ring et al., 2007). While the above study indicated that at the NMJ of APP and APLP2 DKO mice most alterations are found presynaptically, the role of the APP family members and their fragments at synapses within the CNS still remained open. The conditional approach used by Hick et al. (2015) opened the possibility to address the function of APP and APLP2 in the CNS leaving the PNS unaffected. Crossing of APPflox/flox on an APLP2 null background to NexCre-deleter mice generates viable double mutants (cDKO). In these mice the depletion of APP is initiated from embryonic stage 11.5 onwards in excitatory neurons of the forebrain, while APLP2 is constitutively not expressed allowing the investigation of neurodevelopmental effects. Young adult mice show a pronounced deficit in LTP induction and maintenance as well as impairments in PPF. Alterations during the initial phase of LTP, the so-called post-tetanic potentiation and also STP provided a hint toward an impaired presynaptic function. In contrast, the functionality of the postsynapse remained unaffected as basal synaptic transmission was unaltered (Hick et al., 2015). Another study using young conventional APP/APLP2 deficient mice (APP/APLP2-DKO, surviving escape mutants) described increased PPF and synaptic frequency facilitation (FF, Fanutza et al., 2015), supporting the assumption that APP and APLP2 are involved in presynaptic function.

#### PRESYNAPTIC FUNCTION OF APP FAMILY PROTEINS

Short-term plasticity (STP) depends on the release probability of synaptic vesicles, their recycling and content in the presynapse as well as on the activity of calcium sensor kinases. APP and APLP2 show a variety of possible interactions with the synaptic vesicle release machinery: Biochemical approaches showed that APP is associated with synaptic vesicle proteins (Del Prete et al., 2014; Laßek et al., 2014) and that it can be cleaved within vesicles by BACE-1 (Del Prete et al., 2014). Especially the intracellular regions of APP, APLP2, and CTF-ß have been shown to interact with presynaptic vesicle proteins like Rab, AP-2 subunits, the Ca2<sup>+</sup> sensors synaptotagmins, clathrin, and complexin (Del Prete et al., 2014; Fanutza et al., 2015). Results from APP-KO animals point toward a role of APP in controlling synaptic vesicle protein content in the presynaptic active zone as synaptophysin, synaptotagmin-1, and SV2A protein levels are reduced in APP KO mice. In contrast, when beside APP also APLP1 or APLP2 are gene targeted, the abundance of synaptic vesicle proteins is increased (Laßek et al., 2014). The increase in SV2A and synaptotagmin-1 has also been observed in the conditional APP/APLP2 mutant mice generated by Hick et al. (2015) and recently analyzed (Laßek et al., 2016). In that study, Lassek and colleagues further show that APP deletion disturbs Ca2<sup>+</sup> homeostasis, due to a misregulation of calmodulin and neuromodulin but not of the expression of CaMKII or Ca2<sup>+</sup> channels. APLP1 is also localized at the presynaptic active zone (Laßek et al., 2016), but beside the function as mediator of neuronal adhesion (Kaden et al., 2009; Mayer et al., 2016) and its potential involvement in GABAergic neurotransmission (Vnencak et al., 2015) no other role or interaction partners have been attributed so far.

# POSTSYNAPTIC FUNCTION OF APP FAMILY PROTEINS

In addition to a possible function at the presynapse in the developing and mature CNS, all APP family members have been suggested to play a role at the postsynapse. In particular an interaction with N-methyl-D-aspartate receptors (NMDA-R) has been shown especially for the GluN1/GluN2A and GluN1/GluN2B subunits (Cousins et al., 2015). APP, APLP1, and APLP2 are further involved in the regulation of the cell surface expression of NMDA-Rs thus controlling NMDA-R homeostasis (Cousins et al., 2015).

Addressing the role of APP and APLP2 at the postsynapse with the whole cell patch clamp method (measuring miniature excitatory postsynaptic currents (mEPSCs) yielded conflicting results. The study of Fanutza et al. (2015) using conventional APP/APLP2 double mutants, described a decreased mEPSC frequency and an increased mEPSC decay time leading to the assumption of redundant mediated function of APP and APLP2. In contrast, Hick et al. (2015) investigated a conditional APP/APLP2 KO (cDKO) and found no alterations in spontaneous synaptic mEPSCs and in their frequencies. Moreover, the analysis of the NMDA-R subunit composition further points toward unchanged postsynaptic transmission in the cDKO mice (Hick et al., 2015). In this context it is important to note that around 80% of the APLP2−/−APP−/<sup>−</sup> mice die within the first weeks after birth and only 0.3% survive until weaning (von Koch et al., 1997; Heber et al., 2000). Therefore, the mice studied by Fanutza et al. (2015)

were so called "escape-mutants" and their results need to be interpreted with care. It might be that the surviving conventional DKOs developed adaptation mechanisms e.g., an upregulation of synaptic proteins accounts for these controversial results. APLP1 is supposed to accumulate at the postsynapse (Kim et al., 1995) and was also shown to regulate NMDA-R content (Cousins et al., 2015). APLP1, like the other two family members contains the highly conserved YENPTY interaction motif and in thus able to initiate downstream signaling cascades in the postsynaptic compartment supporting synaptic plasticity (activation of intracellular signaling cascades and their contribution to synaptic plasticity is discussed below).

### PROTEOLYTICALLY GENERATED PEPTIDES—APPSα, APPSß, Aß, AN-α, AN-ß

Gene targeting of APP family members using single and double mutants provided evidence about the possible involvement of these proteins in synaptic plasticity, but it could not answer the question of whether the observed effects arose from the action of the full-length proteins or from the absence of their secreted fragment(s).

Evidence pointing to a role of APP fragments in processes of synaptic plasticity arose from the observation that APP processing by α- and ß-secretase is activity-dependent (Nitsch et al., 1993; Fazeli et al., 1994; Kamenetz et al., 2003; Gakhar-Koppole et al., 2008) and can thus be potentiated by neuronal depolarization or high frequency stimulation (HFS). Consequently, the released domains may be especially involved during processes of synaptic activity.

Depending on their site of release, extra- and/or intracellularly, they might have functions as signaling molecules or initiate signaling by binding to different types of receptors. Proteolytic processing of APP is depicted in **Figure 1** and was shown to be similar for APLP1 and APLP2 except for the release of Aß as its coding sequence is absent in the APP homologs (Eggert et al., 2004; Walsh et al., 2007). The current view allows differentiation between three different pathways initiated by the α-, ß-, or η-secretase (see **Figure 1**). In the non-amyloidogenic pathway the α-secretase cuts within the Aβ domain liberating the large APPsα ectodomain and a membrane-anchored C-terminal fragment α (CTF α). The latter is further cut by the γ-secretase releasing the p3 fragment extracellularly and the remaining APP intracellular domain (AICD) into the cytoplasm. The amyloidogenic processing by the ß-secretase yields the APPsß ectodomain and the membrane-tethered CTF ß. Afterwards the activity of the γ-secretase generates the AICD peptide along with Aß. Recently Willem et al. (2015) identified a η-secretase cleavage site in the extracellular domain of APP releasing a short extracellular APPsη ectodomain. Subsequent processing of the remaining membrane anchored CTF η by the α- or ß-secretase generates two new peptides, Aη-α and Aη-ß (Willem et al., 2015). Importantly, APP processing is not restricted to the plasma membrane, but was also shown to occur within synaptic vesicles (Del Prete et al., 2014).

# APPSα PROMOTES SYNAPTIC PLASTICITY

Numerous studies showed that the α-secretase released ectodomain APPsα exerts a role in neuroprotection, synaptic plasticity, and within neuronal networks (Ring et al., 2007; Weyer et al., 2011; Kögel et al., 2012). The acute synaptic function of endogenous APPsα in the adult brain was shown by using APP/APLP2 cDKO mice (Hick et al., 2015). One hour incubation with 10 nM recombinant APPsα peptide (recAPPsα) rescued the severe LTP deficit in acute slices of the mutants indicating that the soluble ectodomain acts on a rapid time-scale. These results were in line with previous findings of Taylor et al. (2008) reporting that intrahippocampal infusion of recAPPsα in the dentate gyrus (DG) of anesthetized rats enhances LTP recorded at the PP-DG pathway in vivo. Moreover, a recent study showed that recAPPsα is able to rescue age-dependent LTP deficits in vitro (Moreno et al., 2015). In addition, we showed that virus driven long-term expression of APPsα restores impaired synaptic plasticity in a mouse model of AD (Fol et al., 2016). It is by now not clear how APPsα mediates the rescue and which receptor might be activated. Overall there is good evidence for a prominent role of APPsα at the postsynapse, in particular by influencing NMDA-R function and synaptodendritic protein synthesis (Taylor et al., 2008; Claasen et al., 2009).

# MODULATION OF POSTSYNAPTIC FUNCTION BY APPSα

One possible mechanism of APPsα action at synapses might be the facilitation of evoked NMDA-R currents at the postsynapse as shown in the study of Taylor et al. (2008). These results were confirmed by acute application of recAPPsα on acute slices of APP/APLP2 cDKO mice or aged rats restoring the LTP induction deficit and highlighting that APPsα modulates synaptic plasticity and regulates early events of the LTP processes (Hick et al., 2015; Moreno et al., 2015). Both studies further report that exogenous applied APPsα does not affect basal synaptic transmission or glutamate release. NMDA-Rs may stimulate αsecretase cleavage of APP during high-frequency stimulation (HFS) or HFS activates metabotropic glutamate (mGluRs) or muscarinic acetylcholine receptors (mAChRs) to promote APPsα release. Notably, the processing must be tightly regulated as high APPsα concentrations reduce LTP induction by activation of inhibitory signaling pathways (Taylor et al., 2008). The concentration dependent action of APPsα to increase NMDA-R currents could further be linked to D-serine availability at the synapse (Moreno et al., 2015). D-serine is the main coagonist required for NMDA-R activation (for details see review Billard, 2012) and APPsα stimulates it's production and release. A recent study further showed that APP deficiency is linked to alterations in D-serine levels accompanied by impaired structural plasticity of dendritic spines (Zou et al., 2016). Facilitation of LTP expression by APPsα might also be mediated through the induction of a subset of plasticity-associated immediate early genes (Ryan et al., 2013), with de novo protein synthesis taking place in synaptoneurosomes mainly by activation of protein kinase G (Claasen et al., 2009). Among APPsα activated signaling cascades are furthermore the phosphatidylinositol-3 kinase (PI(3)K)-Akt kinase signaling pathway (Cheng et al., 2002; Milosch et al., 2014) and the mitogen-activated protein (MAP) kinase signaling pathway (Greenberg et al., 1995; Cheng et al., 2002).

Taken together, APPsα initiates several intracellular signaling cascades to support synaptic activity with an impact on NMDA-R currents, but still the APPsα-specific receptor triggering the effect on NMDA-Rs remains so far elusive. At least the experiments performed by Reinhard et al. (2013) could show that APPsα binding to a cell surface receptor involves two different subdomains. The N-terminal growth factor like domain (GFLD) of APPsα mediates the binding of protein and receptor, while the E2 domain interacts with membrane-anchored heparin sulfate proteoglycans (HSPG) and thus enhances the affinity to the APPsα-receptor. Among the potential receptors for which an interaction with the APP ectodomain is suggested are the lowdensity lipoprotein receptor-related protein (LRP1, Hoffmann et al., 1999; Goto and Tanzi, 2002), the sortilin-related receptor SORLA (Andersen et al., 2006; Hartl et al., 2013), Nogo-66 (Park et al., 2006), and the p75 neurotrophin receptor (Hasebe et al., 2013).

# INHIBITION OF APPSα MEDIATED FUNCTIONS

In-line with the results following exogenous application of APPsα on LTP in vitro and in vivo are the opposite effects observed after α-secretase inhibition (which leads to a reduction in APPsα production). The conditional KO of the major α-secretase ADAM-10 resulted in strongly impaired LTP and altered STP (Prox et al., 2013). Within this study no differences in basic synaptic transmission were found. Interestingly, hippocampal network activity recorded in vivo in the CA1 region of the hippocampus of ADAM-10 cDKO mice was severely impaired and 20% of the animals showed electrographic seizures (Prox et al., 2013). A modulatory role for APPsα on network activity in the hippocampus and cortex has further been observed with regard to aging by Sánchez-Alavez et al. (2007) which recorded electroencephalographic activity. In addition, the key role of APPsα and APLP2sα for LTP induction and maintenance was shown by experiments using the ADAM-10 inhibitor in OHCs (Weyer et al., 2011) or by in vivo LTP recordings in the dentate gyrus after infusion of the α-secretase inhibitor TAPI-1 (Taylor et al., 2008). Due to the lack of ADAM-10 or its inhibition, APP processing by the ß-secretase is favored resulting in higher amounts of Aß peptides and APPsß which may further impair LTP, especially at nano- to micromolar levels see review Wang H. et al. (2012).

#### APPSß DOES NOT MODULATE SYNAPTIC FUNCTION

Only a few studies addressed the physiological action of the ß-secretase which leads to the release of the ectodomain APPsß (see **Figure 1**). APPsß is only 16 amino acids shorter than APPsα, but it is not as neuroprotective as APPsα. This was demonstrated by the Knock-In of the two soluble domains in the perinatal APP/APLP2 DKO mutant model. Only APPsα <sup>+</sup>/+APLP2−/−, but not APPsß+/+APLP2−/<sup>−</sup> mice were viable (Li et al., 2010; Weyer et al., 2011). With regard to synaptic plasticity, APPsß cannot restore the LTP defect of APP/APLP2 cDKO mice (Hick et al., 2015) and does not facilitate LTP recorded in vivo within the DG of rats (Taylor et al., 2008). APPsß was further shown to have no influence on synaptic protein synthesis (Claasen et al., 2009). Consistent with the functional readout on synapses, Tyan et al. (2012) showed that only APPsα but not APPsβ partially rescued defects in dendritic spine number and morphology of primary hippocampal neurons from APP-KO mice.

#### Aß DOMINANTLY ACTS AT THE PRESYNAPSE

At physiological, picomolar concentrations Aß was shown to modulate presynaptic vesicle release (Puzzo et al., 2008; Abramov et al., 2009; Wang H. et al., 2012). It functions via binding to presynaptic APP homodimers (Fogel et al., 2014) or by activating α7-nAChRs (Tong et al., 2011). The study by Lawrence et al. (2014) highlighted that the N-terminal domain of Aß contains this agonist-like activity of the Aß peptide. It was further suggested that successive α- and ß-secretase activity will release the short functional domain, named Aß1–15 (or Aß1– 16, Portelius et al., 2011). With regard to synaptic plasticity, Aß1–15 significantly enhances PTP and LTP without altering baseline synaptic transmission at femtomolar concentrations, while higher amounts had no effect on hippocampal LTP (Lawrence et al., 2014). During LTD, Aß was shown to have a facilitating role through mGluR and NMDA-R due to the altered glutamate recycling at synapses (Li et al., 2009; Chen et al., 2013). The pathological effects of Aß, especially Aß42, are discussed in detail elsewhere (Mucke and Selkoe, 2012; Wang H. et al., 2012; Ripoli et al., 2014; Salgado-Puga and Pena-Ortega, 2015) We only want to mention that under pathological conditions Aß has the opposite effects on synaptic plasticity: it facilitates LTD, depresses LTP, causes dendritic spine loss and leads to hippocampal hyperactivity (Selkoe, 2002; Busche et al., 2008; Shankar et al., 2008; Mucke and Selkoe, 2012; Fol et al., 2016).

#### AN-α AND AN-ß, THE NEW PLAYERS IN THE FIELD

The recently identified η-secretase releases a short extracellular APP-η ectodomain (Willem et al., 2015). The CTFη cleavage product remains anchored to the plasma membrane and subsequently is further processed by α- or ß- secretases to produce two small peptides, Aη-α and Aη-ß (see **Figure 1**; Willem et al., 2015). Willem and colleagues assessed the synaptic function of these peptides by measuring LTP in vitro. While both peptides had no influence on baseline synaptic transmission, hippocampal LTP was severely impaired by Aη-α but not by Aηß. The only structural difference between the two molecules is a C-terminal elongation of the Aη-α peptide by 16 additional Ludewig and Korte Role of APP in Synaptic Plasticity

amino acids (**Figure 1**). Interestingly, the same 16 amino acids are also present at the C-terminus of the APPsα fragment and, similar to Aη-ß, are lacking in the truncated APPsß form (**Figure 1**). This short peptide sequence contains a predicted neuroprotective domain and a heparin binding site (Furukawa et al., 1996). Indeed, neuroprotective properties have been reported for the APPsα peptide. However, and in contradiction to a favorable cellular function of this amino acid sequence, it has been found that Aη-α mediates neurotoxic effects (Willem et al., 2015). The adverse action of Aη-α was also observed by in vivo Ca2<sup>+</sup> imaging experiments performed in the study of Willem et al. (2015) in which Aη-α strongly suppressed the activity of hippocampal neurons. In line with these findings are the observations for both ß-derived peptides. It seems unlikely that these fragments are involved in synaptic plasticity since both Aη-ß and APPsß lacked any modulatory effects on synaptic transmission when bath-applied to acute-hippocampal slices of APP/APLP2 cDKO mice at CA3-CA1 synapses (Hick et al., 2015) or when added during mossy fiber LTP recordings (Taylor et al., 2008). The different modes of action might be a consequence of a conformational change caused by the 16 additional amino acids at the carboxy-terminus of the Aη-α/APPsα cleavage products and/or by specific post-translational modifications (PTMs) like glycosylation or phosphorylation (Walter and Haass, 2000). In the study of Willem et al. (2015) Aη-α conditioned medium or 100 nM synthetic Aη-α showed a reduction in LTP, while only lower concentrations of 1–11 nM recombinant APPsα increased LTP. Moreover, the application of higher APPsα amounts had no effect or resulted even in reduced LTP (Taylor et al., 2008; Hick et al., 2015; Moreno et al., 2015). It would be interesting to know if APPsα can additionally be cleaved by η-secretase and if the released Aη-α could act as a co-player for Aß or APPsα and would therefore provide a modulatory mechanism.

# KNOCK-IN OF APPSα, APPSß, AND THE APP INTRACELLULAR DOMAIN (AICD)

Beside the acute application of APP functional domains as peptides, gene targeting allows their re-introduction on APP or APLP2 null backgrounds. These conditional approaches or Knock-In (KI) mice opened the possibility of the functional characterization of the APP/APLP proteins during development as the constitutive triple KO and nearly all DKOs are embryonic lethal (von Koch et al., 1997; Heber et al., 2000). The study of Ring et al. (2007) analyzed the role of two APP functional domains by generating C-terminally truncated KI alleles of APP. APPsα-KI mice produce only APPsα, whereas APP1CT15- KI mice lack the last 15 amino acids, including the highly conserved YENPTY motif. The phenotypes of both KI lines were similar to WT littermates. LTP as well as learning and memory assessed in behavioral tasks were normal presumably due to the constitutive expression of APLP2. The subsequent combination of both KI mice with APLP2 null mutants generated partially viable offsprings, whereas APPsβ-DM mice die (Li et al., 2010). APPsα-DMs were characterized in detail by Weyer et al. (2011) and APP1CT15-DMs in the study of Klevanski et al. (2015). Both DM strains display alterations at PNS and CNS synapses. The mice suffer of muscular weakness due to altered morphology of the NMJ synapse and impaired transmitter release. Still, the APPsα-DMs reveal more severe electrophysiological impairments at the NMJ by additional reduced quantal content and alterations in the frequency of miniature endplate potentials (MEPP) compared to single mutants investigated by Ring et al. (2007). Hence different motifs account for a normal physiological function in the DMs. With regard to the CNS, both DMs are an impaired induction and maintenance of LTP paralleled by severely altered hippocampusdependent behavior. STP between CA3/CA1 pyramidal cells was unchanged, while only APP1CT15-DMs have altered postsynaptic properties and a trend toward defective proteinsynthesis dependent Late-LTP.

# AICD Is Crucial at Both Sites of the Synapse

The sole expression of AICD on an APP/APLP2 deficient background revealed alterations in synaptic plasticity. This might be a consequence of the abolished interaction of the intracellular domain with several adaptor proteins (Klevanski et al., 2015). For instance, APP interaction partners like Dab1, Shc, Grb, and Mint/X11 proteins mediate not only clathrin-mediated endocytosis of APP, but are also involved in the translocation of APP to the cell-surface (Aydin et al., 2012; van der Kant and Goldstein, 2015). Of particular importance might be the interaction with the adapter protein family FE65.I Interestingly FE65/FE65L1double deficient mice show a similar phenotype of cortical dysplasia as APP triple KO animals (Guénette et al., 2006). The FE65 proteins colocalize with APP in the ER and Golgi and facilitate the translocation of the precursor protein to the cell surface (Sabo et al., 1999). In addition, these proteins also regulate the shuttling of a multimeric complex of AICD/FE65/Tip60 into the nucleus to regulate gene transcription (Cao and Südhof, 2001). Long-lasting strengthening of synaptic transmission is impaired in APP1CT15-DMs perhaps by impaired FE65/AICD mediated postsynaptic transcriptional activity (Klevanski et al., 2015). Interestingly, the analysis of FE65-KO, FE65L1-KO, and FE65/FE65L1-DKO mice revealed similar CNS phenotypes with impairments in LTP and dysfunctions in hippocampal learning tasks in double transgenic animals (Strecker et al., 2016). Accordingly, the APP-FE65 interaction might be crucial for synaptic function, but also for precise ectodomain shedding. In APP1CT15-DMs mice, processing of APP via the amyloidogenic pathway is heavily impaired (Klevanski et al., 2015). That might have a positive effect with regard to Aß accumulation but also a negative outcome since picomolar amounts of Aß positively regulate the presynaptic vesicle release probability and facilitate learning and LTP in the hippocampal CA1 region by activating α7-nAChRs (reviewed by Wang H. et al., 2012). Collectively, these studies highlight an essential function for the 15 C-terminal amino acids including the YENPTY motif for transmembrane signaling and the ectodomain APPsα for proper synapse function.

#### CONCLUSION

The majority of experimental data provided so far indicate a requirement for APP and APLP2 in synaptic plasticity which is in particular mediated by their proteolytic derived domains. The diverse functions of the APP protein family during either preor postsynaptically initiated processes of synaptic plasticity and under basal conditions are summarized in **Figure 2**. According to this model, APP full length proteins mediate stability of synaptic structures by their cell adhesion properties when integrated into the plasma membrane (Kaden et al., 2009; Baumkötter et al., 2012) and thus maintain appropriate spine numbers, especially via the APPsα domain (Tyan et al., 2012; Weyer et al., 2014). The insertion of full-length proteins is regulated by electrical activity or gradients of ions like zinc. The APLP1 protein shows the highest presence at the cell surface among all APP protein family members (Kaden et al., 2009; Mayer et al., 2016). As indicated **Figure 2B** depicts the APP protein family function at the presynaptic site, where the Aß, Aß-15 and possibly the APPsα domain interfere with glutamate release by activating nAChRs and enhancing intracellular Ca2<sup>+</sup> levels (Puzzo et al., 2008; Wang Z. et al., 2012; Lawrence et al., 2014). It is further hypothesized that homodimerized APP acts as a G-Protein coupled receptor which is activated by Aß and might be involved in neurotransmitter release following enhanced Ca2<sup>+</sup> influx. Especially the intracellular domains of APP and APLP2 seem to be associated with proteins of the synaptic vesicle release machinery regulating the molecular composition of synaptic vesicles at the presynaptic active zone (Del Prete et al., 2014; Fanutza et al., 2015; Laßek et al., 2015). At the postsynaptic compartment (**Figure 2C**) patterns of synaptic activity modulate APP family protein processing. HFS enhances the amount of secreted APPsα possibly linked to mGluRs or AChRs activation (Nitsch et al., 1992, 1997). Released APPsα was shown to facilitate NMDA-R currents (Taylor et al., 2008; Weyer et al., 2011) by increasing the NMDA-R agonist D-serine (Moreno

#### REFERENCES


et al., 2015) or by up-regulating signaling cascades downstream of NMDA-Rs- promoting synaptic plasticity like the CamKII pathway (Claasen et al., 2009) or by inducing the expression of immediate early genes involved in synaptic plasticity (Ryan et al., 2013). In this regard only APPsα was shown to have trophic functions while APPsß mediates neither positive nor negative effects with respect to baseline synaptic function or synaptic plasticity (Taylor et al., 2008; Hick et al., 2015). Several lines of evidence indicate that under physiological conditions structural and functional synaptic modulation is mediated by APPsα. What still needs to be investigated, however, is the mechanism by which APPsα exerts its trophic action, particularly which receptor might be activated and if the recently discovered Aη peptides might function as regulators of APPsα mediated synaptic plasticity and homeostasis. Identifying the cellular site of η-secretase cleavage within neurons and answering whether the secretion of Aη peptides is linked to neuronal activity will reveal the roles of the peptides in processes of synaptic plasticity.

Overall elucidating the physiological function of APP family members and fragments is an important step to understand brain function as well as brain dysfunction, also with respect to a possible treatment of neurodegenerative disorders like AD. It is important to acknowledge, that rational therapeutic approaches need to take into account the functional role of disease associated proteins.

#### AUTHOR CONTRIBUTIONS

SL: wrote the review and prepared the figures. MK: designed the review and wrote the paper.

#### FUNDING

This work was supported by the Deutsche Forschungsgemeinschaft Grants (KO 1674/3-1, 3-2) to MK.


plaques in a mouse model of Alzheimer's disease. Science 321, 1686–1689. doi: 10.1126/science.1162844


via activation of the MAPK pathway. Eur. J. Neurosci. 28, 871–882. doi: 10.1111/j.1460-9568.2008.06398.x


affect hippocampal neuron morphology or function. Mol. Cell. Neurosci. 49, 448–455. doi: 10.1016/j.mcn.2012.02.001


abnormalities of APP-deficient mice. J. Neurosci. 27, 7817–7826. doi: 10.1523/JNEUROSCI.1026-07.2007


hippocampal NMDA receptor function, long-term potentiation and spatial memory. Neurobiol. Dis. 31, 250–260. doi: 10.1016/j.nbd.2008.04.011


to influence GABAergic short-term plasticity. J. Neurosci. 29, 15660–15668. doi: 10.1523/JNEUROSCI.4104-09.2009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Ludewig and Korte. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Role of APP in Structural Spine Plasticity

#### Elena Montagna<sup>1</sup> , Mario M. Dorostkar <sup>2</sup> and Jochen Herms 1,2,3 \*

<sup>1</sup>Department for Translational Brain Research, German Center for Neurodegenerative Diseases (DZNE), Ludwig-Maximilian-University Munich, Munich, Germany, <sup>2</sup>Center for Neuropathology and Prion Research, Ludwig-Maximilian-University Munich, Munich, Germany, <sup>3</sup>Munich Cluster of Systems Neurology (SyNergy), Ludwig-Maximilian-University Munich, Munich, Germany

Amyloid precursor protein (APP) is a transmembrane protein highly expressed in neurons. The full-length protein has cell-adhesion and receptor-like properties, which play roles in synapse formation and stability. Furthermore, APP can be cleaved by several proteases into numerous fragments, many of which affect synaptic function and stability. This review article focuses on the mechanisms of APP in structural spine plasticity, which encompasses the morphological alterations at excitatory synapses. These occur as changes in the number and morphology of dendritic spines, which correspond to the postsynaptic compartment of excitatory synapses. Both overexpression and knockout (KO) of APP lead to impaired synaptic plasticity. Recent data also suggest a role of APP in the regulation of astrocytic D-serine homeostasis, which in turn regulates synaptic plasticity.

Keywords: APP, dendritic spines, synaptic plasticity, in vivo, d-serine

#### Edited by:

Thomas Deller, Goethe University Frankfurt, Germany

#### Reviewed by:

Johannes Vogt, Johannes Gutenberg-Universität Mainz, Germany Andreas Vlachos, University of Düsseldorf, Germany

#### \*Correspondence:

Jochen Herms jochen.herms@med.unimuenchen.de

Received: 19 December 2016 Accepted: 21 April 2017 Published: 10 May 2017

#### Citation:

Montagna E, Dorostkar MM and Herms J (2017) The Role of APP in Structural Spine Plasticity. Front. Mol. Neurosci. 10:136. doi: 10.3389/fnmol.2017.00136 STRUCTURAL PLASTICITY

Structural synaptic plasticity refers to morphologically observable changes of synapses which accompany the classical electrophysiological events during synaptic plasticity. Most prominent among them are dynamic changes in the number and shape of dendritic spines, which correspond to the postsynaptic compartment of glutamatergic excitatory synapses. Dendritic spines are small (1–2 µm long) protrusions of the dendritic shaft, which receive excitatory synaptic input and compartmentalize calcium (Majewska et al., 2000; Yuste and Bonhoeffer, 2001; Yuste, 2011) and therefore dictate the biophysical characteristics of a postsynapse. They are fundamental players in establishing and maintaining the neuronal network as well as other complex functions such as learning and memory. Conventionally, dendritic spines are classified according to their morphology into three different groups: thin spines, which are fine and long but have a discernible head; stubby spines, with a large head and an indiscernible neck and mushroom spines with big head and thin neck (Yuste and Bonhoeffer, 2004; Alvarez and Sabatini, 2007; Herms and Dorostkar, 2016). Additionally, filopodia are very motile protrusions that can transform themselves into mushroom or thin spines (Alvarez and Sabatini, 2007). However, a STED and EM based study revealed a higher degree of heterogeneity of both spine size and morphology (Tønnesen et al., 2014). These morphologies reflect different functional properties: for example, thin spines are more dynamic and more plastic than mushroom and stubby spines, which are thought to be more stable (Yuste and Bonhoeffer, 2001; Knott et al., 2006). A fraction of spines are continuously retracted and newly formed, and this process, expressed as turnover rate (TOR), is accelerated during learning and memory formation (Fu and Zuo, 2011). Montagna et al. APP in Synaptic Plasticity

Dendritic spines were discovered by Ramon y Cajal, who used Golgi's silver staining method to visualize dendrites and their processes (Yuste and Bonhoeffer, 2001). While essentially the same technical approach is still used today, modern research on spines is typically conducted on transgenic animals expressing a fluorophore in a sparse subset of neurons (Feng et al., 2000). This allows visualization of spines on confocal microscopes, and, more importantly, in vivo observation of the dynamic changes comprising structural plasticity.

#### AMYLOID PRECURSOR PROTEIN IS A SYNAPTIC PROTEIN

Amyloid precursor protein (APP) is a member of a family of conserved type I membrane proteins which also includes APP like one protein (APLP1) and APP like two protein (APLP2; Wasco et al., 1992, 1993; Slunt et al., 1994). The major APP isoform expressed in neurons is 695 amino acids long, while longer isoforms are expressed in other tissues. Full-length APP consists of four main domains: the extracellular domains E1 (Dahms et al., 2010) and E2; a transmembrane sequence (Dulubova et al., 2004; Keil et al., 2004; Dahms et al., 2012); and the APP intracellular domain (AICD; Kroenke et al., 1997; Radzimanowski et al., 2008; Coburger et al., 2014; **Figure 1**). APP can be cleaved by a large number of proteases, which are grouped into α-, β- and γ-secretases, depending on the cleavage site. However, proteases which cleave APP outside these three sites also exist (Vella and Cappai, 2012; Willem et al., 2015; Zhang et al., 2015; Baranger et al., 2016). Depending on the combination of proteases which process APP, a vast number of different cleavage products may be generated, which have various biological properties (Nhan et al., 2015; Andrew et al., 2016). Among them are, for instance, amyloid β fragments which are generated by the action of β, and γ-secretases and which are known to be involved in the pathogenesis of Alzheimer's disease. Other proteolytic products, such as the soluble fragment sAPPα and CTFs have been shown to be neuroprotective (Chasseigneaux and Allinquant, 2012; Hick et al., 2015; Andrew et al., 2016). Furthermore, in vitro evidence suggests that CTFs induce axonal outgrowth by interacting with G-protein αs subunits, which in turn activate adenylyl cyclase/PKA-dependent pathways (Copenhaver and Kögel, 2017), although these findings have not been corroborated in vivo.

In the brain, APP reaches its highest expression level during early postnatal development (from P1 to P36 in mice) and is preferentially localized at pre- and postsynapses (De Strooper and Annaert, 2000). During this period, synaptogenesis occurs and neuronal connections are formed (Hoe et al., 2009; Wang et al., 2009). Accordingly, many studies described putative roles of APP in the modulation of neurite outgrowth and synaptic connectivity (Moya et al., 1994; De Strooper and Annaert, 2000; Herms et al., 2004; Wang et al., 2009; Hoe et al., 2012; Müller and Zheng, 2012; Weyer et al., 2014; Hick et al., 2015). Synaptogenesis and neurite outgrowth may be mediated by full-length APP, which has been shown to exhibit cell adhesion- and receptor-like properties (Qiu et al., 1995; Ando et al., 1999; Turner et al., 2003; Soba et al., 2005; Müller and Zheng, 2012; Coburger et al., 2014; Deyts et al., 2016): there is convincing evidence that two distinct extracellular E1 domains from neighboring molecules of APP, APLP1 and APLP2 (Soba et al., 2005; Baumkötter et al., 2012; Deyts et al., 2016) can interact via their heparin binding domains (HBDs), and form a so-called heparin cross-linked dimer (Coburger et al., 2014). The interaction of the E2 domains with heparin cross-linked dimers further strengthens the dimerization process (Wang et al., 2009; Hoefgen et al., 2014). As APP is present both on preand postsynaptic terminals, a dimerization across the synapse may be relevant for synapse formation and stabilization (Wang et al., 2009; Baumkötter et al., 2014; Stahl et al., 2014). Moreover, the interaction of E1 and E2 domains with extracellular matrix components, like collagen, heparin, laminin, glypican, F-spondin and β1- integrin reinforces APP dimerization, and may further modulate the stability or plasticity of dendritic spines (Beher et al., 1996; Williamson et al., 1996; Rice et al., 2013; Wade et al., 2013).

Furthermore, growth factors and receptor-like proteins have been shown to interact with the APP-extracellular domains (Reinhard et al., 2005; Coburger et al., 2014; Deyts et al., 2016). Thus, activation of growth factor receptors could be an alternative mode of action of how APP affects spine plasticity. Additionally, the intra-cellular domain AICD itself may mediate receptor-like activity (Cao and Südhof, 2001, 2004; McLoughlin and Miller, 2008; Müller et al., 2008; Klevanski et al., 2015). Here, an intracellular response is triggered by the interaction of AICD-cleavage products with effector and adaptor proteins from the cytosolic compartment (Okamoto et al., 1990; Timossi et al., 2004; Deyts et al., 2012; **Figure 1**).

In addition to developmental processes, APP has also been shown to be involved in synaptic plasticity of mature synapses. For instance, some AICD-proteolytic products can be directly translocated into the nucleus and activate several transcription factors, like CP2/LSF/LBP1 or Tip60 (Müller et al., 2008; Schettini et al., 2010; Pardossi-Piquard and Checler, 2012), which are known to be involved in the regulation of dendritic spine plasticity.

# APP IS INVOLVED IN STRUCTURAL SPINE PLASTICITY

Two main bodies of evidence support a role of APP in structural plasticity. On one hand, overexpression of APP, which is often used to model Alzheimer's disease, may alter dendritic spines independently of typical Alzheimer's disease pathology. These findings are described later in this section. On the other hand, knockout (KO) of APP alters spine dynamics: in the hippocampus, APP KO causes a range of synaptic alterations, depending on the model and paradigm studied. For instance, in cultured hippocampal neurons of APP KO animals, we found enhanced amplitudes of evoked AMPA- and NMDA-receptor-mediated EPSCs, which were reduced by pre-conditioned wildtype

medium. Additionally, we found an increased density of synaptophysin-positive presynaptic puncta (Priller et al., 2006). The number of dendritic spines, in contrast was reduced (Tyan et al., 2012) in APP KO neurons, while it was increased in APP overexpressing neurons (Lee et al., 2010). In organotypic slice cultures APP-KO neurons showed a pronounced decrease in spine density and reductions in the number of mushroom spines, which was rescued by sAPPα expression (Weyer et al., 2014). These results suggest that soluble sAPPα modulates synaptic function in the neonatal hippocampus. A study in hippocampal slices of adult APP KO mice found decreased paired-pulse facilitation in the dentate gyrus, while granule cell excitatory transmission was unaltered (Jedlicka et al., 2012). These contrasting findings may be the result of region-specific differences in APP expression in the hippocampus (Del Turco et al., 2016).

We recently studied dendritic spines of layer V pyramidal neurons of the somatosensory cortex in 4 month old APP-KO × GFP-M mice (Zou et al., 2016), which is accessible to chronic in vivo imaging. The density and the TOR of dendritic spines were monitored for a period of 9 weeks in comparison to GFP-M control mice (**Figure 2**). No differences were detected in the overall spine densities between the groups, whereas the fate of individual spines over time exhibited significant changes in their elimination and formation rates, resulting in reduced spine TOR (Zou et al., 2016). Since an alteration in spine plasticity is often correlated with alteration in spine morphology, we performed morphological analyses and found a decrease in the fraction of thin spines and an increase in the fraction of mushroom spines (Zou et al., 2016). These findings mirror the dynamic changes in TOR as thin spines are typically less stable than mushroom or stubby spines. In an earlier article (Bittner et al., 2009), in contrast, we had found an

(B) Statistical summary of alterations in relative spine density over time. WT mice respond with increased spine density and turnover, while APP KO mice do not. Treatment with D-serine restores EE-induced synaptic plasticity in APP KO. (WT, n = 5; APP KO, n = 6; APP KO + D-serine, n = 4). Figured modified from Zou et al. (2016).

increased number of spines in APP KO, while turnover was not analyzed in detail. Two main factors may explain this apparent discrepancy: first, the data from the 2009 article were recorded almost a decade earlier, on an older generation multiphoton microscope. Modern microscopes have become considerably better at resolving thin spines. APP KO changes the morphology from thin to mushroom spines, which are more voluminous and thus easier to detect. Since thin spines used to be harder to detect, the results may have been interpreted as an apparent increase in spine densities. Second, the 2009 study used the YFP-H mouse line to label neurons, while the 2016 study used the GFP-M line. Although the populations of neurons which are labeled in both lines overlap, they are not identical. Thus, the subset of neurons analyzed in the earlier study may have had a different response to APP KO or it may have had a relatively higher fraction of thin spines, thus aggravating the first factor.

In order to understand whether the reduced TOR in APP-KO can be increased by physiological stimuli, we exposed APP KO mice to enriched environment (EE) which enhances the spine plasticity in several brain regions and increases TOR (Berman et al., 1996; Kozorovitskiy et al., 2005; Nithianantharajah and Hannan, 2006; Mora et al., 2007; Jung and Herms, 2014; Sale et al., 2014). However, APP KO mice exposed to EE for 5 weeks did not exhibit the physiological increase in spine density which was observed in WT controls (Zou et al., 2016). Thus, loss of APP leads to impaired adaptive spine plasticity (**Figure 2**).

In order to elucidate which domain of APP modulates dendritic spine plasticity, spine density and TOR were investigated in APP-∆CT15 mice (Ring et al., 2007). These mice express a truncated form of APP, lacking 15 amino acids at the C-terminus, which correspond to the AICD. It was shown that several other phenotypes of APP-KO mice were rescued in APP-∆CT15 mice, such as growth rates, brain weight, grip strength, locomotor alterations and spatial learning associated with long term potentiation (LTP) impairment in aged mice (Müller et al., 1994; Zheng et al., 1995; Dawson et al., 1999; Magara et al., 1999; Ring et al., 2007).

To further elucidate the role of APP in spine dynamics, our team conducted a study on 4–5 month old APP 23-GFP-M mice by 2-photon microscopy in vivo. APP 23-GFP-M mice overexpress human APP (isoform 751) with the Swedish (KM670/671NL) mutation under the murine Thy1 promoter (Sturchler-Pierrat et al., 1997). This leads to the formation of amyloid β deposits starting at 6 months of age and therefore this mouse line is considered to be a model of amyloidosis. However, our study revealed a significant decrease in dendritic spine density of layer V neurons of the somatosensory cortex (Zou et al., 2015) before the appearance of Aβ plaques, which was correlated with the amount of intracellular APP accumulating in neurons. Intracellular APP accumulation has been shown to mediate neuro- and synaptotoxicity in a number of publications (Neve et al., 1992; Fukuchi et al., 1994; Oster-Granite et al., 1996; Lu et al., 2003). Thus, it is crucial to distinguish between these different causes of synaptotoxicity when studying models of amyloidosis, as they do not all necessarily reflect human disease.

# APP REGULATES SPINE PLASTICITY BY MODULATION OF ASTROCYTIC D-SERINE

An additional mechanism for APP-mediated spine-arrangement is suggested by its modulation of astrocytic D-serine homeostasis, which is a modulator of synaptic NMDA receptors (Engert and Bonhoeffer, 1999; Hering and Sheng, 2001; Lai and Ip, 2013). The calcium-dependent astrocytic release of D-serine modulates NMDA-dependent LTP (Henneberger et al., 2010). It has been shown that full-length APP and its fragments modulate D-serine secretion (Wu and Barger, 2004; Wu et al., 2007) as well as astrocytic calcium homeostasis (Hamid et al., 2007; Linde et al., 2011). More recently, biosensor measurements in the cortex of 4–6 month old APP KO mice revealed decreased extracellular D-serine levels, while total D-serine was increased (Zou et al., 2016). These results suggest an alteration of D-serine homeostasis in APP deficient mice may underlie the altered regulation of spine dynamics. Treatment with exogenous D-serine for 5 weeks, supplemented in drinking water of standard housed and EE mice, restored extracellular D-serine levels and normalized the concentrations of total D-serine and L-serine in APP-KO brain (Zou et al., 2016). Furthermore, the administration of D-serine rescued the impaired dendritic structural plasticity in APP-KO mice: D-serine treated APP-KO mice had restored spine dynamics under standard housing conditions. Moreover, upon environmental enrichment, the fraction of thin spines was enhanced, while fraction of mushrooms spines was decreased (**Figure 2**). Although these data do not contest the synaptic role played by APP, they suggest a new interaction between APP and the D-serine homeostasis which is involved in spine dynamics and plasticity.

#### CONCLUSIONS

Several mechanisms by which APP may modulate spine plasticity have been identified (summarized in **Figure 1**): structural properties of the full-length protein may help stabilizing synapses, while binding of ligands to the extracellular part may trigger intracellular cascades, similar to a classical receptor molecules. Additionally, recent findings demonstrate that APP modulates astrocytic D-serine homeostasis, which interacts with

#### REFERENCES


NDMA receptors to modify synaptic function. Lastly, neurotoxic and neuroprotective APP fragments may trigger or alleviate pathophysiological mechanisms involved in neurodegenerative diseases. Thus, APP seems to regulate synaptic plasticity at several levels. Yet, the relative importance of each of these mechanisms in physiology and disease remains to be elucidated.

#### AUTHOR CONTRIBUTIONS

EM and MMD wrote the review article, prepared figures. JH wrote the review article.

#### ACKNOWLEDGMENTS

This work was funded by the European Commission within the 7th framework (Extrabrain–606950).


spectral variants of GFP. Neuron 28, 41–51. doi: 10.1016/S0896-6273(00) 00084-2


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Montagna, Dorostkar and Herms. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Amyloid-precursor Like Proteins APLP1 and APLP2 Are Dispensable for Normal Development of the Neonatal Respiratory Network

Kang Han<sup>1</sup> , Ulrike C. Müller<sup>1</sup> \* † and Swen Hülsmann2,3 \* †

1 Institute of Pharmacy and Molecular Biotechnology, Heidelberg University, Heidelberg, Germany, <sup>2</sup> Klinik für Anästhesiologie, Universitätsmedizin Göttingen, Göttingen, Germany, <sup>3</sup> Center for Nanoscale Microscopy and Molecular Physiology of the Brain (CNMPB), Göttingen, Germany

Recent studies using animal models indicated that the members of the amyloid precursor protein (APP) gene family are important for the formation, maintenance, and plasticity of synapses. Despite this, the specific role of the APP homologs APLP1 and APLP2 within the CNS and PNS is still poorly understood. In contrast to the subtle phenotypes of single mutants, double knockout mice (DKO) lacking APP/APLP2 or APLP1/APLP2 die within the first day after birth. Whereas APP/APLP2-DKO mice show severe deficits of neuromuscular morphology and transmission, the underlying cause of lethality of APLP1/APLP2-DKO mice remains unclear. Since expression of both proteins was confirmed by in situ hybridization, we aimed to test the role of APLP1/APLP2 in the formation and maintenance of synapses in the brainstem, and assessed a potential dysfunction of the most vital central neuronal network in APLP1/APLP2-DKO mice by analyzing the respiratory network of the medulla. We performed in vivo unrestrained whole body plethysmography in newborn APLP1/APLP2-DKO mice at postnatal day zero. Additionally, we directly tested the activity of the respiratory network in an acute slice preparation that includes the pre-Bötzinger complex. In both sets of experiments, no significant differences were detected regarding respiratory rate and cycle variability, strongly arguing against central respiratory problems as the primary cause of death of APLP1/APLP2-DKO mice. Thus, we conclude that APLP1 and APLP2 are dispensable for the development of the network and the generation of a normal breathing rhythm.

Keywords: amyloid-precursor like proteins, pre Bötzinger complex, medullary slice

# INTRODUCTION

Amyloid precursor-like proteins 1 (APLP1) and 2 (APLP2) are type I transmembrane proteins belonging to the evolutionary conserved amyloid precursor protein (APP) gene family (Coulson et al., 2000; Muller et al., 2017). APP has been intensely studied with regard to Alzheimer's disease (AD) pathogenesis, as proteolytic processing of APP gives rise to the Aβ peptide that is deposited in extracellular plaques in the brains of Alzheimer patients (Selkoe and Hardy, 2016). Although the Aβ region is unique for APP the two APLPs share with APP an overall similar structural organization and several conserved domains (Fox et al., 2007). Moreover, they are processed by the same set of α-, β- and γ- secretases yielding a complex array of proteolytic fragments

#### Edited by:

Andreas Vlachos, Albert Ludwig University of Freiburg, Germany

#### Reviewed by:

Lisa M. Munter, McGill University, Canada Gregory D. Funk, University of Alberta, Canada

#### \*Correspondence:

Ulrike C. Müller u.mueller@urz.uni-heidelberg.de Swen Hülsmann shuelsm2@uni-goettingen.de †These authors have shared senior authorship.

> Received: 28 February 2017 Accepted: 29 May 2017 Published: 22 June 2017

#### Citation:

Han K, Müller UC and Hülsmann S (2017) Amyloid-precursor Like Proteins APLP1 and APLP2 Are Dispensable for Normal Development of the Neonatal Respiratory Network. Front. Mol. Neurosci. 10:189. doi: 10.3389/fnmol.2017.00189

(Slunt et al., 1994; Scheinfeld et al., 2002; Eggert et al., 2004; Endres et al., 2005; Kuhn et al., 2016). During development and in adult rodents APP and APLP2 are ubiquitously expressed in a largely overlapping pattern in many tissues with particularly high expression in the nervous system (brain, spinal cord, retina, ganglia) including the neuromuscular junction (NMJ) (Slunt et al., 1994; Lorent et al., 1995; Sarasa et al., 2000; Wang et al., 2005; Caldwell et al., 2013; Klevanski et al., 2014; see also footnote<sup>1</sup> ).

In contrast, APLP1 is specifically expressed in neurons (Lorent et al., 1995). APP and APLPs are found in the somatodendritic and axonal compartment (Schilling et al., 2017) and have been localized to the presynaptic active zone (Kim et al., 1995; Walsh et al., 2007; Lassek et al., 2013). APP family proteins can form homotypic and heterotypic dimers and have been implicated in transcellular and synaptic adhesion in vitro and in vivo at the NMJ (Soba et al., 2005; Wang et al., 2009; Baumkotter et al., 2014). Their physiological functions have been studied using knockout (KO) mice lacking either individual family members or in all possible combinations of DKO and triple KO mice (Aydin et al., 2012; Mockett et al., 2017; Muller et al., 2017). All three single KO mice exhibit rather subtle phenotypes and display no apparent alterations in brain morphology (Zheng et al., 1995; von Koch et al., 1997; Magara et al., 1999; Heber et al., 2000). APP-KO mice, which have been most thoroughly studied, show reduced brain and bodyweight (15–20%), reduced grip strength and locomotor activity and increased susceptibility to brain injury (Zheng et al., 1995; Ring et al., 2007; Hefter et al., 2016; Plummer et al., 2016). Reduced theta-gamma coupling in APP-deficient mice (Zhang et al., 2016) points toward alterations in the connectivity of central networks (Zhang et al., 2016), but only aged (12-month-old) APP-KO mice exhibit reduced spine density in cortex and hippocampus, impairments in long term potentiation (LTP) at CA3/CA1 synapses of the hippocampus and impairments in spatial learning (Dawson et al., 1999; Seabrook et al., 1999; Ring et al., 2007; Lee et al., 2010; Tyan et al., 2012). Apart from subtle retinal abnormalities (Dinet et al., 2016) APLP2-KO mice show a wild type like phenotype with normal spine density and synaptic plasticity even at old age (von Koch et al., 1997; Heber et al., 2000; Weyer et al., 2011; Midthune et al., 2012). APLP1-KO have been studied in much less detail compared to APP and APLP2 deficient mice. Similar to APP-KO mice, APLP1-KO mice show reduced body weight but normal locomotor activity and grip strength (Heber et al., 2000). Electrophysiological analysis of perforant pathgranule cell synapses of the dentate gyrus revealed decreased network inhibition but no alterations in LTP in APLP1-KO mice (Vnencak et al., 2015). In line with this, morphological analysis of CA1 neurons in organotypic hippocampal cultures revealed normal spine density and dendritic branching (Weyer et al., 2014). However, recent analysis of aged (1-year-old) APLP1-KO mice showed reduced spine density and frequency of miniature excitatory synaptic currents in the hippocampus pointing toward compensatory mechanism that may fail with aging (Schilling et al., 2017).

Genetic evidence indicates that the above mentioned- rather minor- phenotypes of single KOs are likely due to functional compensation within the gene family. APP/APLP2 DKO mice, APLP1/APLP2-DKO and triple KO mice die within the 1st day after birth (von Koch et al., 1997; Heber et al., 2000; Herms et al., 2004). Interestingly, APP/APLP1-DKO mice proved viable, indicating that APLP2 has unique properties that are required when either APP or APLP1 are lacking (Heber et al., 2000). Together these data suggest that APP family proteins can serve overlapping functions in tissue in which they are co-expressed (Muller et al., 2017).

Indeed, recent data suggest a functional compensation between APP and APLP2 in the CNS. Conditional, forebrainspecific APP/APLP2-DKO mice exhibited reduced spine density and branching of hippocampal neurons, impaired synaptic plasticity and pronounced impairments in hippocampus dependent behavior that were found already in young adult mice (Hick et al., 2015). Despite this, the specific role of the APP homologs APLP1 and APLP2 within the CNS and PNS is still poorly understood. Conditional APLP1/APLP2-DKO mice have not been generated so far, which precludes detailed analysis of neuronal network functions of APLP1 and APLP2 in the adult brain. However, analysis of the networks in the brainstem, that are already developed at birth, especially the respiratory network (Feldman et al., 2012; Dick et al., 2015) is possible and thus allowed us to gather information about basic synaptic connectivity in otherwise lethal APLP1/APLP2- DKO mice. Analysis of this vital network appears even more reasonable, since the morphology of the NMJ appears normal in APLP1/APLP2-DKO mice. Unlike, APP/APLP2-DKO mice, which show a strongly altered morphology of NMJs at the diaphragm and severely impaired neurotransmission (Wang et al., 2005, 2007, 2009; Caldwell et al., 2013; Klevanski et al., 2014), APLP1/APLP2-DKO mice showed normal endplate patterning and only very subtle morphological abnormalities of individual synapses, which strikingly contrasts with the highly penetrant lethality of these mutants [less than 0.5% of APLP1/APLP2-DKO pups survive up to weaning (Klevanski et al., 2014)].

Therefore, we analyzed the respiratory network of newborn mice using whole body plethysmography and direct electrophysiological recordings from brain stem slices. However, we did not observe significant differences between APLP1/APLP2-DKO pups and littermates, suggesting that APLP1 and 2 are not essential for respiratory network formation and function.

#### MATERIALS AND METHODS

#### Animal Ethics

Experiments were conducted in accordance with the guidelines of the German Physiological Society, the European Communities Council Directive and the law of Federal Republic of Germany. Breeding of perinatally lethal APLP1/APLP2-DKO mice (Heber et al., 2000) to obtain tissue samples and brain slices has been approved by the Regierungspräsidium Karlsruhe (35-9185.81/ G-82/14).

<sup>1</sup>http://developingmouse.brain-map.org

## In Situ Hybridization

fnmol-10-00189 June 20, 2017 Time: 18:17 # 3

The gene sequence of the APLP1 probe corresponds to nucleotides 1431–1940 of the murine APLP1 mRNA as previously described (Vnencak et al., 2015). The gene sequence of the APLP2 probe corresponds to nucleotides 1554–2100 of APLP2 mRNA (GenBank accession number NM\_001102456.1, for full sequence see **Table 1**). DIG labeled RNA probes were in vitro transcribed using the Roche DIG RNA labeling kit (SP6/T7) and purified using RNAse free ChromaSpin 100 columns (Clontech). The quantity of labeled and purified probe was estimated by Dot blot as described in the DIG RNA labeling kit manual. Brains of P0 mice were fixed in 4% PFA/DEPC-PBS over night at 4◦C followed by three washes in PBS (3 min each), and dehydrated through an ascending sucrose series (10%; 15%; 30%) diluted in DEPC-PBS. OCT (Tissue –Tek) embedded brains were frozen on dry ice, and finally stored at −80◦C. Brains were cut on a cryostat (Zeiss Hyrax C50) at a thickness of 30 µm. Brain sections were post-fixed with 4% PFA/DEPC-PBS for 20 min, and treated with Proteinase K (10 µg/ml in 20 mM Tris/HCl, 1mM EDTA, pH 7.2) for 10 min, washed in 3 times of DEPC-PBS (10 min each), and then put horizontally into a chamber humidified with 50% formamide/4× SSC. Each slide was covered with 100 µl hybridization buffer plus 400 pg labeled probe. Hybridization was carried out over night at 56◦C in the tightly sealed humidified chamber. On the next day, coverslips were floated off in 5× SSC to wash away excess probe (10 min). Stringency washes were 20 min in 5× SSC for 3 times and 40 min in 0.5× SSC/20% formamide at 60◦C in a water bath. Slides were cooled down to RT in 0.5× SSC/20% formamide at RT, and washed 15 min in NTE (0.5 M NaCl, 10 mM Tris pH 7.0, 5 mM EDTA) at 37◦C, treated with 10 µg/ml RNase A/NTE for 30 min at 37◦C, followed by a 15 min wash in NTE. After a further 40 min wash in 0.5× SSC/20% formamide at 60◦C slides were equilibrated in P1 DIG (100 mM Tris/HCl; 150 mM NaCl) for 10 min and afterwards blocked in blocking solution (P1DIG + 0.5% BSA + 1% Blocking reagent, Roche) for 30 min. Brain slices were encircled with PAP PEN, anti-DIG-AP antibody (Roche) was diluted 1:500 in blocking solution, and 80 µl were pipetted onto every brain slice. Antibody incubation was overnight at 4◦C in a humidified chamber. The next day, all slides were washed twice for 15 min in P1 DIG and then equilibrated in P3 DIG (100 mM Tris/HCl; 100 mM NaCl; 50 mM MgCl2, pH 9.5) for 2 min. 80 µl substrate solution for alkaline phosphatase (NBT/BCIP stock solution, diluted 1:50 in P3 DIG) were pipetted onto each brain slice, incubation was overnight at RT until color development (due to the formation of the insoluble, violet NBT diformazan) was sufficient. Slides were then washed in PBS-T, fixed for 10 min in 4% PFA in PBS, washed in P4 DIG (10 mM Tris/HCl; 1 mM EDTA, pH 8.0) for 10 min, air dried for 2 h and finally mounted in Mowiol.

#### Microscopy and Image Processing

Images from ISH were taken with a Keyence BZ-9000 microscope, using a 20× objective. Tiled images were automatically generated from 20 high resolution images TABLE 1 | Amyloid precursor-like proteins 2 probe in situ hybridization.



(1340×1024 pixel CCD sensor) using the Image Joint Function of BZ-II Analyzer software (Keyence). Scaling was performed for data reduction leading to final images (1760 × 1805 Pixel). Final images were exported to tif-format (8bit-RGB) and composed to final figures in a graphic program (CorelDraw). The "Paxinos Atlas of the Developing Mouse Brain" (Paxinos, 2007) was used to identify the brainstem structures depicted in **Figures 1**, **2**.

#### Unrestrained Whole-Body Plethysmography

Resting ventilation was measured using unrestrained wholebody plethysmography (Bartlett and Tenney, 1970) adapted for use with neonatal animals: Individual animals were placed in a chamber (5–10 ml) that was connected to one side of a differential pressure transducer (model DP103-14, Validyne Engineering, Northridge, CA, United States). The chamber communicated with atmospheric pressure through a slow leak (27 gauge hypodermic needle) to minimize pressure differences between the chambers because of fluctuations in atmospheric pressure during measurements. The analog signal from the transducer was demodulated (model CD-15 carrier demodulator, Validyne Engineering), amplified, filtered and recorded on thermal chart recorder. Additionally, data were digitized ( ≥ 1kHz) using an interface (ITC-16; HEKA, Lambrecht, Germany) and then captured and stored to disk by Apple computer (Acquire, Bruxton Corporation, Seattle, WA, United States or Axograph 4.0, Axon Instruments, Foster City, CA, United States). The ventilatory pattern was recorded from each animal for 2–3 min on postnatal day 1 during the first 8 h after birth. Measurements were performed at room temperature (Gomeza et al., 2003), however, the brief time away from the nest would have minimal effect on body temperature. Moreover, WT and KO animals were treated in exactly the same way, since the experimenter was blind

negative control we used a section from an APLP1-KO mouse (F) High magnification images of the ventral lateral medulla and ventral respiratory column (VRC) are shown in (A0–F<sup>0</sup> ). The region of the pre Bötzinger Complex (prBö) is shown in panels (D,D<sup>0</sup> ). Scale bars correspond to 100 µm. IO = inferior olive, Amb = nucleus ambiguous (encircled). Note the violet staining due to the alkaline phosphates mediated formation of the NBT diformazan (see Materials and Methods).

to the genotypes. To remove drift from the recordings a digital filtering was performed (Band pass 0.5–20 Hz; LabChart 8).

#### Slice Preparation

Immediately after the plethysmography, mice were anesthetized with ether. The brain and upper cervical spinal cord were isolated in ice-cold artificial cerebrospinal fluid (aCSF), which was saturated with carbogen (95% O2–5% CO2). The brainstem was isolated and glued with cyano-acrylate to an agar block with its rostral end directed upward. Brainstem slicing was started from the rostral end with the neuroaxis inclined by 20◦ to the plane of the blade. This configuration preserved most projections from the pre-Bötzinger complex to the nucleus of hypoglossus and left the hypoglossal (XII) rootlets intact (Ramirez et al., 1996). A 650–750 µm thick cut was made to obtain one slice containing the functionally intact respiratory center (Gomeza et al., 2003). This preparation includes the pre-Bötzinger complex (preBötC), a region which is essential for the generation of the respiratory

section from an APLP2-KO mouse (F). High magnification images of the ventral lateral medulla and VRC are shown in (A0–F<sup>0</sup> ). The region of the pre Bötzinger Complex (prBö) is shown in panels (C,C<sup>0</sup> ). Scale bars correspond to 100 µm. IO = inferior olive, Amb = nucleus ambiguous (encircled). Note the violet staining due to the alkaline phosphates mediated formation of the NBT diformazan (see methods).

rhythm (Smith et al., 1991). The slice was transferred to a recording chamber and stabilized by a platinum frame with nylon fibers. The XII rootlet was drawn into a suction electrode, or alternatively an extracellular electrode filled with aCSF was placed in the hypoglossal nucleus. The concentration of extracellular K <sup>+</sup> in aCSF saturated with carbogen at 30◦C was increase to 8 mM to maintain respiratory network activity. Extracellular neuronal activity was amplified 5'000–10'000 times, band-pass (0.25–1.5 kHz) filtered, rectified, and integrated (Paynter filter with a time constant of 50–100 ms) (Hulsmann et al., 2000). Hypoglossal activity, which corresponds to the inspiratory phase of the respiratory cycle (Smith et al., 1990) can be used as an index of the central respiratory rhythm (Smith et al., 1991). Rootlet discharges and their integrals were digitized at 5 kHz using an interface (ITC-16; Instrutech Corp., Great Neck, NY, United States) and Axograph 4.0 software (Axon Instruments, Inc., Foster City, CA, United States). Data were stored on hard disk for off-line analysis. Burst interval and amplitude of the integrated

rootlet signal were measured with Axograph 4.0 software. Burst frequency was calculated as the reciprocal of the mean inter-burst interval. The coefficient of variation (CV) of the amplitude of the integrated burst was used as an additional parameter of the overall network activity. Results are expressed as means ± SE. One Way Analysis of Variance (ANOVA) tests were used to determine the significance of changes using Sigma Plot software (SPSS Inc., Chicago, IL, United States). Results were considered significant if P < 0.05. The aCSF contained (in mM) 118 NaCl, 3 KCl, 1.5 CaCl2, 1 MgCl2, 1NaH2P04, 25 NaHCO<sup>3</sup> and 30 <sup>D</sup>-glucose (pH = 7.4, 310 mosmol/l) at a temperature of 30◦C. Substances were purchased from Sigma (Deisenhofen, Germany) unless otherwise indicated.

#### Data Handling and Statistical Analysis

Figures were assembled using the graphic program CorelDraw. Statistical analysis was performed with SigmaPlot software using One Way Analysis of Variance (ANOVA) tests. Significance was assumed if p < 0.05.

# RESULTS

In this paper we addressed how a deficiency of both APLP1 and APLP2 affects the respiratory network of newborn mice and ask the question whether APLP1/APLP2 DKO mice die because of a functional defect of the respiratory network in the medulla oblongata. To this end, we first assessed the level of APLP1 and APLP2 expression in the medulla. Second, we measured breathing using whole body plethysmography to test if the animals are able to ventilate. Last, we tested the neuronal activity of the kernel of the respiratory network, the pre-Bötzinger complex, which allows to test if the neuronal interaction in the network is grossly intact (Gomeza et al., 2003; Rahman et al., 2015).

#### Expression of APLP Proteins in the Respiratory Network of the Medulla

As a baseline for further experiments we studied the expression pattern of APLP1 and 2 in the medulla of newborn wild type mice. As reliable antibodies that work in immunohistochemistry are not available for APLPs, we assessed mRNA expression by in situ hybridization (ISH). Brain sections from APLP1-KO and APLP2- KO mice were processed in parallel and served as a negative control. We found substantial and largely overlapping expression of both APLPs in the ventrolateral medulla. High level expression of APLP1 was detected in motor nuclei of the medulla especially in the facial nucleus, at the nucleus ambiguus and in the inferior olive (**Figure 1**). Along the ventral respiratory column (VRC) substantial expression is also found in the Bötzinger Complex (BötC), the Pre-Bötzinger Complex (preBötC) and in the rostral Ventral respiratory group (rVRG). Additionally, neurons in the nucleus of the solitary tract, which is part of the dorsal respiratory column, also expressed APLP1. A similar expression pattern was observed for APLP2 (**Figure 2**), however, the expression in the facial nucleus and inferior olive was less prominent as compared to the neighboring structures. Similar to APLP1 substantial expression of APLP2 mRNA was found in the VRC (including BötC, preBötC, and rVRG). In contrast, only a weak background staining was observed in APLP1-KO and APLP2-KO sections.

#### Breathing of Newborn Mice

In total we obtained 4 litters with 42 offspring (APLP1+/<sup>+</sup> APLP2−/<sup>−</sup> (APLP1-WT): n = 7; APLP1+/−APLP2−/<sup>−</sup> (APLP1 heterozygous): n = 18; APLP1−/−APLP2−/<sup>−</sup> (DKO): n = 17). Among these, we found 4 dead offspring in the cage from which 3 newborns were genotyped as DKO and one as APLP1 heterozygous.

Having established that APLP1 and APLP2 are co-expressed in the respiratory brain stem we now aimed to identify a potential central respiratory insufficiency that could be causal for the early death of DKO mice at postnatal day zero (Heber et al., 2000). Therefore, breathing of the neonates was measured (within the first 8 h after birth) with unrestrained wholebody plethysmography. From 14 analyzed APLP1−/−APLP2−/<sup>−</sup> mice, two showed extremely long apneic intervals ( ≥ 10 s; not shown). However, shorter apneas (2–10 s) were also found in the other genotypes (APLP1+/+APLP2−/<sup>−</sup> mice = two animals; APLP1+/−APLP2−/<sup>−</sup> mice = 7 animals; APLP1−/−APLP2−/<sup>−</sup> mice = 4 animals; **Figure 3**). The average duration of the longest apnea (or cycle interval) detected during the recordings were not significantly different between APLP1+/+APLP2−/<sup>−</sup> mice (1.9 ± 0.5 s mean ± SEM; median: 1.5 s), APLP1+/−APLP2−/<sup>−</sup> mice (1.7 ± 0.3 s; median: 1.5 s) and APLP1−/−APLP2−/<sup>−</sup> mice (4.0 ± 1.1 s; median: 1.6 s; ANOVA on Ranks; p = 0.578).

Further quantitative analysis of the breathing rate of surviving mice was, however, not different between genotypes (**Figure 3A**). APLP1+/+APLP2−/<sup>−</sup> had on average a rate of 1.58 ± 0.11 Hz (mean ± SEM). APLP1+/−APLP2−/<sup>−</sup> mice (n = 7) ventilated with an average rate of 1.64 ± 1.3 Hz and APLP1−/−APLP2−/<sup>−</sup> mice with a rate of 1.29 ± 0.18 Hz (p = 0.206). There was a trend toward a higher variability of the respiratory cycle length in APLP1−/−APLP2−/<sup>−</sup> mice. Although the variability of the respiratory cycle tended to be larger in DKO mice, the difference of the coefficient of variation (CV) of the inter burst interval between APLP1+/+, APLP2−/<sup>−</sup> (29.4 ± 5.4), APLP1+/−, APLP2−/<sup>−</sup> (34.3 ± 6.2) and DKO (50.0 ± 10.3) remained non-significant (p = 0.489). Similarly, also no significant differences for the CV of the amplitude was detected between APLP1+/+APLP2−/<sup>−</sup> (43.2 ± 7.5), APLP1+/−APLP2−/<sup>−</sup> (49.7 ± 7.4) and APLP1−/−APLP2−/<sup>−</sup> (41.8 ± 5.8). These data show that APLP1−/−APLP2−/<sup>−</sup> DKO mice are able to breath at birth. Thus, a primary problem resulting from a defect in the development of the respiratory network appears unlikely.

# Analysis of Respiratory Network Function In Situ

To substantiate our interpretation and to investigate the brainstem respiratory network independent of arousal or other central neuronal factors that might influence breathing we also analyzed central respiratory network activity in the rhythmic slice preparation from the caudal brainstem including the

genotypes. (B–D) Quantification of statistical evaluation the respiratory rate (B), and variability of the breathing. The Coefficient of Variation (CV) of the respiratory interval is shown in (C), the CV of the amplitude in (D). In (B–D), the lower boundary of the box is the 25th percentile, lines within a box represents the median, the higher boundary of the box provides 75th percentile. Error bars indicate the 10th and 90th percentiles, respectively. Number of animal analyzed are given on top of the boxes.

pre-Bötzinger complex. We recorded rhythmic hypoglossal motoneuron pool discharges, which are known to occur in synchrony with periodic bursts of neurons in the pre-Bötzinger complex. The frequency and the CV of the burst discharge of hypoglossal motoneuron pool was very similar in all three groups of mice (**Table 2** and **Figure 4**). Taken together, these findings suggest that the anatomy and the connectome of neonatal respiratory network is intact and that the mice do not die from a failure of respiratory rhythm generation.



#### DISCUSSION

In situ expression analysis demonstrated that both APLP mRNAs are expressed in brain stem areas important for respiratory function. This is well in agreement with previous studies that reported expression of APLP1 in the spinal cord and brain stem of E18.5 wild type mice<sup>2</sup> . However, the resolution was insufficient to unequivocally demonstrate expression, e.g., in nuclei involved in the control of respiratory rhythm generation. Expression of APLP2 mRNA had only been shown in spinal cord at E15, but not assessed in brain stem or close to birth (Lorent et al., 1995). Thus, our findings that both APLPs show an overlapping expression pattern in the medulla were in line with our initial hypothesis that a combined lack of both APLPs might disturb breathing. Our subsequent findings do, however, not support this notion.

Since breathing frequency is not different in APLP1/APLP2- DKO mice in comparison to the APLP1 positive mutants, we can argue that a developmental defect in the respiratory network is unlikely to be responsible for the early postnatal death of the double knock out mice. Unlike in mice that have been shown to die from a central respiratory failure, presenting with either no

<sup>2</sup>http://developingmouse.brain-map.org

Values are depicted as average ± SEM. Number of animal/slices analyzed are given inside the bars.

breathing movements (Rahman et al., 2015) or extremely long apneas (Blanchi et al., 2003; Gomeza et al., 2003) immediately after birth, breathing of APLP1/APLP2-DKO mice was neither slower nor strikingly more irregular than in viable littermates that were heterozygous or wild type for APLP1. Moreover, the observation of a normal breathing rhythm is in line with a normal development of diaphragm innervation as demonstrated by a normal endplate distribution and branching pattern of the phrenic nerve (Klevanski et al., 2014).

Irregular breathing patterns with periods of apnea (**Figure 3**) are not uncommon in wild type mice at postnatal day zero (Robinson et al., 2000). Thus, longer apneas, which were observed in 2 APLP1/APLP2-DKO pubs, should be considered secondary to a yet unknown cause of deterioration of the mouse and might therefore reflect the agony of the dying animal. Further assessment of the respiratory network properties in respiratory rhythmic slice preparation containing the pre-Bötzinger Complex supported this notion. Frequency of inspiratory burst as recorded from the hypoglossal motoneurons in APLP1/APLP2-DKO mice was indistinguishable from APLP2 single knock out and heterozygous APLP1+/−APLP2−/<sup>−</sup> littermates (**Figure 4**). Thus, there is no evidence for a general disturbance of the respiratory network as an elementary cause of the death. Moreover, the persistence of respiratory activity argues against a substantial disturbance of synaptic interaction in the respiratory network, which is in line with the absence of obvious ultrastructural changes of brain stem synapses in APLP1/APLP2- DKO mice (Heber et al., 2000) although a quantitative analysis has not been performed. In this regard it is interesting that for APP/APLP2-DKO mice reduced synaptic vesicle density and active zone size was reported for submandibular ganglion synapses (Yang et al., 2005). Nevertheless it is unlikely that APP family proteins are essential for basal synaptic transmission of CNS neurons, as excitatory neurons derived in vitro from triple KO embryonic stem cells showed normal spontaneous mEPSC frequencies and amplitudes (Bergmans et al., 2010) consistent with unaltered basal synaptic transmission as demonstrated by normal input/output strength recorded at the CA3/CA1 pathway in forebrain-specific APP/APLP2- DKO mice (Hick et al., 2015). More recent data point toward a regulatory role of APP family proteins to facilitate neurotransmitter release, as proteins of the release machinery including Munc18 and synaptotagmins have been found to interact with APP and the APLPs (Weyer et al., 2011; Fanutza et al., 2015).

All three APP family proteins have been shown to interact with NMDA receptors and enhance their cell surface expression in transfected cells (Cousins et al., 2009, 2015). From this one might expect synaptic alterations in the forebrain and/or brain stem of APLP1/APLP2-DKO mice. Unfortunately, such data are

#### REFERENCES

Aydin, D., Weyer, S. W., and Muller, U. C. (2012). Functions of the APP gene family in the nervous system: insights from mouse models. Exp. Brain Res. 217, 423–434. doi: 10.1007/s00221-011-2861-2

currently unavailable and will await the generation conditional APLP1/APLP2-DKO mice. We also cannot rule out at present, whether APP may compensate for the loss of APLPs in the brain stem, although it is not sufficient to confer postnatal survival. With respect to a potential alteration of NMDA-receptor expression in APLP1/APLP2 KO mice, an obvious similarity to the phenotype of NMDAR1-deficient mice needs to be pointed out (Forrest et al., 1994). NMDAR1-deficent newborn mice also develop a lethal phenotype with prolonged apneas and death, but the respiratory activity in medullary slice preparation, is indistinguishable from controls (Funk et al., 1997).

In summary, in the presence of APP, amyloid precursorlike proteins APLP1 and APLP2 are dispensable for a normal organization of the respiratory network during embryonic development. Thus, alteration of synaptic function in other brain areas like the arousal system, or even non-neuronal and metabolic effects, such as hypoglycaemia previously shown for APP/APLP2-DKO mice (Needham et al., 2008), might be a potential cause of neonatal death. Finally, our findings may also warrant to re-examine APLP1 and APLP2 mediated functions at the NMJ. Although morphological alterations in APLP1/APLP2- DKO were very subtle this does not preclude functional defects, e.g., for transmitter release as previously detected specifically at neuromuscular synapses of APP/APLP2-DKO mice (Wang et al., 2005, 2009).

#### AUTHOR CONTRIBUTIONS

SH and UM designed the study. SH and KH performed experiments. SH, KH, and UM wrote the manuscript.

#### FUNDING

SH received funding from DFG-Research Center for Nanoscale Microscopy and Molecular Physiology of the Brain (CNMPB). UM was supported by grants from the Deutsche Forschungsgemeinschaft (MU-1457/8-2 and 9-2) and the Else Kröner-Fresenius-Stiftung (Az2014\_A22). KH was supported by the China Scholarship Council. The authors acknowledge financial support by the Open Access Publication Funds of the Göttingen University.

#### ACKNOWLEDGMENT

The authors are grateful to G. Mesuret for technical instruction regarding neonatal slice preparation and D. W. Richter for support.


depend on copper binding to the growth factor-like domain. J. Neurosci. 34, 11159–11172. doi: 10.1523/JNEUROSCI.0180-14.2014


glial glycine uptake in glycinergic inhibition. Neuron 40, 785–796. doi: 10.1016/ S0896-6273(03)00672-X


fnmol-10-00189 June 20, 2017 Time: 18:17 # 10

Paxinos, G. (2007). Atlas of the Developing Mouse Brain : at E17.5, PO, and P6. Amsterdam: Elsevier.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Han, Müller and Hülsmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fnmol-10-00189 June 20, 2017 Time: 18:17 # 11

# APP as a Protective Factor in Acute Neuronal Insults

#### Dimitri Hefter 1,2 and Andreas Draguhn<sup>1</sup> \*

1 Institute of Physiology and Pathophysiology, Heidelberg University, Heidelberg, Germany, <sup>2</sup>Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany

Despite its key role in the molecular pathology of Alzheimer's disease (AD), the physiological function of amyloid precursor protein (APP) is unknown. Increasing evidence, however, points towards a neuroprotective role of this membrane protein in situations of metabolic stress. A key observation is the up-regulation of APP following acute (stroke, cardiac arrest) or chronic (cerebrovascular disease) hypoxic-ischemic conditions. While this mechanism may increase the risk or severity of AD, APP by itself or its soluble extracellular fragment APPsα can promote neuronal survival. Indeed, different animal models of acute hypoxia-ischemia, traumatic brain injury (TBI) and excitotoxicity have revealed protective effects of APP or APPsα. The underlying mechanisms involve APP-mediated regulation of calcium homeostasis via NMDA receptors (NMDAR), voltage-gated calcium channels (VGCC) or internal calcium stores. In addition, APP affects the expression of survival- or apoptosis-related genes as well as neurotrophic factors. In this review, we summarize the current understanding of the neuroprotective role of APP and APPsα and possible implications for future research and new therapeutic strategies.

#### Edited by:

Thomas Deller, Goethe-University, Germany

#### Reviewed by:

Dirk Isbrandt, DZNE Bonn & University of Cologne, Germany Wickliffe C. Abraham, University of Otago, New Zealand Maximilian Lenz, University of Düsseldorf, Germany

#### \*Correspondence:

Andreas Draguhn andreas.draguhn@physiologie. uniheidelberg.de

Received: 07 November 2016 Accepted: 16 January 2017 Published: 02 February 2017

#### Citation:

Hefter D and Draguhn A (2017) APP as a Protective Factor in Acute Neuronal Insults. Front. Mol. Neurosci. 10:22. doi: 10.3389/fnmol.2017.00022 Keywords: Alzheimer, ischemia, calcium toxicity, cell death, amyloid precursor protein, stroke, traumatic brain injury, neuroprotection

# INTRODUCTION

Amyloid precursor protein (APP) has been first described in 1987 as a potential substrate of pathological deposits in the nervous system (Kang et al., 1987). By now, there is good evidence from multiple lines of research that specific domains of APP do indeed contribute to amyloid plaques as found in patients with Alzheimer's disease (AD; Hardy and Selkoe, 2002). On the other hand, the function of this ubiquitously expressed protein in healthy brains remains poorly understood. Recent evidence from neurological patients and from different disease models hint towards a potential neuroprotective function of APP under conditions of acute cellular insult: APP is up-regulated following hypoxia, ischemia or traumatic brain injury (TBI; Van den Heuvel et al., 1999; Pottier et al., 2012). This reaction coincides well with some known interactions between APP and other proteins which are relevant for homeostatic regulation of cell integrity under stressful conditions, such as certain glutamate receptors, calcium channels or gene-regulatory networks (Russo et al., 2005). With respect to the underlying molecular mechanisms it is important to note that the integral membrane protein APP can give rise to both, protective and potentially damaging molecules following cleavage by different secretases (Brunholz et al., 2012). These cleavage processes keep a balance between different amyloidogenic and non-amyloidogenic products of APP, including the protective APPsα fragment which is secreted into the extracellular space (Mattson et al., 1993a). Together, APP or its fragments may well have a neuroprotective role during acute challenges of neuronal integrity, and it may exert this function by regulating neuronal calcium homeostasis and cell survival. Novel findings on APP-related neuroprotective mechanisms open promising new therapeutic strategies in stroke, AD and TBI.

In the present review article, we summarize the evidence for a neuroprotective function of APP in the adult brain. After a brief introduction of the protein and its metabolites, we summarize current knowledge from clinical, animal and in vitro studies about its role in stroke, brain injury and neurodegeneration. Finally, we discuss possible mechanisms and point out several promising therapeutic targets.

#### APP STRUCTURE, EXPRESSION, TRAFFICKING, CLEAVAGE AND SUBCELLULAR LOCALIZATION

APP is a type-1 transmembrane protein comprising a long extracellular N-terminal domain, a transmembrane region and an intracellular C-terminal domain, APP intracellular domain (AICD; Kang et al., 1987). Alternative splicing of the APP gene, which is located on chromosome 21, produces three isoforms containing 695, 751 and 770 amino acids, respectively (Beyreuther et al., 1993). While APP751 and APP770 are expressed almost ubiquitously, APP695 can be found nearly exclusively in neurons. Depending on the isoform, the APP extracellular domain consists of up to six different subdomains with specific structural motives and various binding partners such as extracellular matrix proteins (heparine, collagene, laminine, proteoglycans), metals (copper, zinc) and regulatory proteins (LDL-receptor-related protein, F-spondin; Gralle and Ferreira, 2007; Müller and Zheng, 2012). After translation in the endoplasmic reticulum (ER), APP undergoes various post-translational modifications in the Golgi complex before it is transported to the cell membrane (Caporaso et al., 1994). The mature membrane protein can be processed by different membrane-associated proteolytic enzymes, beginning with cleavage of the transmembrane domain by γ-secretase. Subsequent cleavage by α-secretase results in three fragments: AICD, a short p3 fragment and the secreted soluble APP α (APPsα). Alternatively, cleavage by the β-secretase BACE-1 (Beta-site APP Cleaving Enzyme 1) releases APPsβ and the neurotoxic amyloid β (Aβ) peptide (refer to Haass et al., 2012 for review on processing of APP). Under normal conditions, only a small fraction of the expressed APP is secreted, and cleavage by α-secretases outweighs the amyloidogenic pathway by far (Hick et al., 2015).

In neurons, APP is found in somatodendritic and axonal compartments as well as in the presynaptic active zone (Laßek et al., 2016) which it reaches by fast axonal transport (Brunholz et al., 2012). Its intracellular trafficking involves four different neuronal trafficking adaptors including Mint1 and is regulated by tyrosine phosphorylation (Dunning et al., 2016). Expression, trafficking and processing of APP are complexly regulated, including prominent changes during pathological states. APP expression is upregulated under conditions of metabolic stress (Hoyer et al., 2005), ischemia (Pottier et al., 2012), brain injury (Van den Heuvel et al., 1999) and inflammation (Herbst-Robinson et al., 2015). APP processing and degradation differ under conditions of acute stress. In response to increased levels of intracellular calcium, APP is degraded via the ubiquitinproteasome proteolytic pathway (Jung et al., 2015). Facilitated degradation might counteract overexpression of APP under conditions of acute stress, prevent accumulation of misfolded protein and its processing into Aß. As an additional adaptive mechanism, cleavage of the protein is regulated by synaptic activity, affecting the balance between amyloidogenic and non-amyloidogenic pathways (Kamenetz et al., 2003; Cirrito et al., 2005). Intriguingly, APP is expressed and cleaved heterogeneously in different types of neurons and in astrocytes and in different brain areas, which might contribute to variable susceptibility to insults between brain regions and cell types (Del Turco et al., 2016; Liao et al., 2016). Activated by proinflammatory cytokines, astrocytes were shown both to contribute to Aß production as well as to stimulate the secretion of APPsα, suggesting a significant contribution of glia cells to production and cleavage of APP and a tight coupling between APP processing and the immune system (Zhao et al., 2011; Yang et al., 2015). While still quite superficially understood, this activity- and stress-dependent multi-level relation of APP in neural, glial and immune cell response strongly suggests a role as an acute phase protein with functions in cellular survival under metabolically challenging conditions.

#### FUNCTIONS OF APP AND ITS METABOLITES

APP is highly conserved across different phyla including mammals, insects and nematodes, suggesting that the protein has advantageous effects on survival and reproduction of animals (Müller and Zheng, 2012; van der Kant and Goldstein, 2015). Indeed, in the nematode C. elegans knock-out of APP-like protein (APL-1) is lethal (Hornsten et al., 2007). Drosophila lacking the APP ortholog APPL exhibit severe memory deficits (Bourdet et al., 2015). Most knowledge on systemic functions of APP has been gained from studies of genetically modified rodents. Remarkably, mice lacking APP are viable, fertile, and exhibit a relatively mild phenotype. Alterations include reduced body and brain weight and several neurological symptoms like reduced grip strength (Weyer et al., 2011; Caldwell et al., 2013), deficits in spatial memory (Puzzo et al., 2011), and increased susceptibility to seizures (Steinbach et al., 1998). This phenotype may be related to changes at the cellular and network level like reduced numbers of dendritic spines, reduced hippocampal LTP and altered short-term plasticity (Seabrook et al., 1999; Weyer et al., 2011; Jedlicka et al., 2012; Korte et al., 2012). The absence of more severe deficits is likely due to the existence of homologous proteins, called APLP1 and APLP2, which can compensate the lack of APP due to overlapping functions (Aydin et al., 2012). Indeed, double knockout mice lacking two of the three homologous proteins are much more heavily affected: mice lacking APP and the globally expressed APLP2 as well as APLP1-KO/APLP2-KO mice die perinatally Hefter and Draguhn APP as a Neuroprotective Factor

due to impaired neuromuscular transmission (Wang et al., 2005), while mice deficient for APP and APLP1, which is predominantly expressed in the brain, survive birth but exhibit rather severe deficits (Heber et al., 2000). Triple knock-out mice die during embryonic development or shortly after birth and show lissencephaly-like cortical malformations (Herms et al., 2004), pointing towards a role for APP and its homologs in essential developmental mechanisms like neuronal migration, neurite outgrowth and synaptogenesis. Detailed studies at the cellular and molecular level revealed several further functions of APP. The protein is involved in regulation of synaptic vesicle exocytosis (Kohli et al., 2012) glutamatergic, GABAergic and cholinergic synaptic transmission (Wang et al., 2005, 2014; Schrenk-Siemens et al., 2008) and synapse formation (Priller et al., 2006). Interestingly, it also regulates endosomal phosphoinositide metabolism and prevents neurodegeneration (Balklava et al., 2015), and it interacts with a large variety of survival-related cascades (Russo et al., 2005; Venezia et al., 2007).

# APPsα and APPsβ

Several functions of APP seem to be mediated by its soluble cleavage product APPsα. Selective expression of APPsα in mice with APP−/<sup>−</sup> background abolishes most of their deficits, rescuing LTP as well as the typical anatomical and behavioral abnormalities (Ring et al., 2007; Hick et al., 2015). Mice selectively expressing APPsα on APP-KO/APLP2- KO background (which by itself is lethal) survive well into adulthood and show only a mildly altered phenotype, similar to simple APP-KO animals (Zhang et al., 2013). Enhancing APPsα levels by over-expression of ADAM-10 increases cortical synaptogenesis in vivo (Bell et al., 2008). Intraventricular application of APPsα enhances memory in mice (Meziane et al., 1998). Altogether, there is strong evidence that APPsα mediates many of the effects of APP on brain development and supports several cognitive functions. In addition, the APPsα fragment has been shown to mediate a variety of neuroprotective and trophic effects (Hick et al., 2015; Fol et al., 2016; Hefter et al., 2016; Plummer et al., 2016), as discussed in following sections. It is important to note that secretion of APPsα is regulated by neuronal activity (Kirazov et al., 1997; Gakhar-Koppole et al., 2008) and by activated astrocytes (Yang et al., 2015). This may point towards state-dependent functions of the protein, in line with the neuroprotective effects described below. The trophic effects of APPsα are dosedependent, beginning as low as 100 pM, reaching an optimum at 10 nM and decreasing at higher doses (Demars et al., 2011).

Notably, APPsβ fails to mimic the beneficial effects of APPsα, although there is only a difference of 16 amino acids between both proteins (Hick et al., 2015). In other studies, however, trophic effects of APPsβ were detected, albeit less potent than those of APPsα (Chasseigneaux et al., 2011). Interestingly, APPsβ was found to undergo further proteolytic cleavage and bind to ''death receptor 6'', activating caspase-6 and thus contributing to neurodegeneration (Nikolaev et al., 2009).

# APP Intracellular Domain (AICD)

The intracellular domain of APP, termed AICD, interacts with various cytosolic signaling cascades including glycogen synthase kinase 3 (GSK-3), Ras proteins and MAPK pathways and is able to translocate to the nucleus after forming a complex with the adaptor protein Fe65 (Schettini et al., 2010). There, it is involved in regulation of genes associated with survival and apoptosis (Müller et al., 2008; Multhaup et al., 2015). Indeed, overexpression of AICD was found to induce apoptosis by interaction with the p53-pathway (Ozaki et al., 2006; Nakayama et al., 2008). Moreover, AICD modulates intracellular calcium homeostasis and ATP content (Hamid et al., 2007) and affects synaptic plasticity and hippocampus-dependent learning by increasing LTP (Klevanski et al., 2015).

# Amyloid ß

Resulting from APP cleavage by BACE-1, Aβ peptides can accumulate extracellularly as soluble oligomers or in amyloid plaques, promoting neurodegeneration in AD (Haass, 2010). Interestingly, Aβ can be internalized by neurons and accumulates in the cytosol as well as within endosomes/lysosomes and mitochondria (Chen and Yan, 2006). It exerts neurotoxic effects via a variety of mechanisms, such as disruption of calcium homeostasis (Berridge, 2010), overactivation of mGluR5 (Zhang et al., 2015), impairment of synaptic transmission, plasticity and network function (Kamenetz et al., 2003; Palop and Mucke, 2010), mitochondrial dysfunction (Chen and Zhong, 2013) and apoptosis (Umeda et al., 2011). Remarkably, it is also able to translocate into the nucleus and influence apoptosis-related gene transcription (Barucker et al., 2014; Multhaup et al., 2015). The APP fragment has also been suggested to form channel-like pores in neuronal membranes, but the underlying mechanisms are currently unknown (Barucker et al., 2014).

#### LINKS BETWEEN ISCHEMIA, BRAIN INJURY AND NEURODEGENERATION—RESULTS FROM HUMAN STUDIES

Sporadic AD is the most common cause of dementia and constitutes one of the most imminent medical problems in developed countries (Prince et al., 2015). Cognitive deficits in AD are caused by progressive loss of neurons, beginning in the temporal lobe and resulting in severe global brain atrophy (Fox and Schott, 2004). The neuronal loss is linked to pathological accumulation of amyloid and tau protein, as first described by Alzheimer (1906). No causal treatments exist at this stage of the disease. However, irreversible macroscopic pathology and cognitive decline are preceded by functional deficits such as disturbance of cellular calcium- and energy-homeostasis (Mattson, 1994), mitochondrial dysfunction (Swerdlow and Khan, 2004; Rönnbäck et al., 2016), synaptic failure (Selkoe, 2002) and activation of pro-apoptotic pathways (Mattson, 2000), offering an opportunity for detection and intervention at the preclinical stage (Jack and Holtzman, 2013). Interestingly, various lines of evidence suggest that molecular pathomechanisms in AD such as amyloid deposition and disrupted calcium homeostasis overlap with those in hypoxiaischemia (Peers et al., 2009) and TBI (Magnoni and Brody, 2010).

In many pathologies of the CNS such as TBI and stroke the brain-blood barrier (BBB) is disturbed which results in extravasation of blood-derived proteins including albumin and inflammatory mediators into the brain tissue (Zhao et al., 2015). Presence of albumin in the extracellular space may act epileptogenic and promote degeneration (Friedman et al., 2009). Inflammatory cytokines may regulate secretase activity and both facilitate non-amyloidogenic cleavage as well as Aß deposition (Zhao et al., 2011; Yang et al., 2015), possibly contributing to the development of AD later on (Sastre et al., 2008). Alternatively, amyloid may directly diffuse from vessels into the brain tissue through a malfunctioning BBB (Pluta et al., 2009). Amyloid plaques, in turn, are well known to evoke a strong inflammatory response with activation of microglia, astrocytes and inflammatory mediators.

Amyloid accumulates and deposits into plaques if its intracellular degradation and extracellular clearance are disturbed. Proteolytic degradation is inhibited by lack of energy substrates and oxidative stress, while extracellular degradation requires intact interstitial and cerebrospinal fluid flow and BBB function and is impaired in inflammation (Iliff et al., 2015; Tarasoff-Conway et al., 2015). As these processes are disturbed in stroke and TBI, both conditions may lead to impaired amyloid clearance and AD development.

In line with these pathomechanisms, a history of TBI (Fleminger et al., 2003; Sivanandam and Thakur, 2012), stroke (Thiel et al., 2014) and cardiac arrest (de la Torre, 2006) are risk factors for developing AD. Below, we will describe the similarities between these conditions in detail. **Figure 1** shows the pathophysiological cascades leading from acute insult to long-term neurodegeneration.

#### TRAUMATIC BRAIN INJURY, APP AND AD

### TBI Leads to Amyloid Pathology and Strongly Increases the Risk for AD and Cognitive Decline

TBI is a debilitating and life-threatening condition which is the leading cause of disability in people under 35 years in industrial countries (Feigin et al., 2013). Besides acute primary damage, TBI promotes secondary neurodegeneration and increases the risk for developing AD by ∼2-fold (Mortimer et al., 1991; Mayeux et al., 1993; Schofield et al., 1997; Guo et al., 2000; Fleminger et al., 2003). Following TBI, diffuse Aβ deposits can be found in the temporal cortex as early as 2 h after the insult (Ikonomovic et al., 2004). Furthermore, post mortem histological analysis shows that deposition of amyloid ß-protein in the brain occurs in approximately one-third of individuals who die shortly after a severe head injury (Roberts et al., 1994). Aβ levels are altered in cerebro-spinal and interstitial cerebral fluid in patients with TBI (Magnoni and Brody, 2010; Tsitsopoulos and Marklund, 2013) and correlate with clinical outcome (Magnoni and Brody, 2010). A history of TBI prior to the onset of dementia correlates with greater amyloid burden in patients with mild cognitive deficits (Mielke et al., 2014) and is associated with faster rates of cognitive decline in AD patients (Moretti et al., 2012; Gilbert et al., 2014). In TBI patients, APP transcription is upregulated and its axonal transport is interrupted due to diffuse axonal injury, which results in deposition of APP and its products in axonal ''bulbs'' (Hayashi et al., 2015). These results from human studies are in line with a large body of evidence from various models of TBI in mice, rats and sheep, where APP overexpression following TBI has been extensively studied. In models of focal cerebral injury inflicted by stabbing or weight drop local APP immunoreactivity increased in neurons as well as in astrocytes (Otsuka et al., 1991; Lewén et al., 1995, 1996). In the midline fluid percussion model of diffuse brain injury in adult rats APP expression was globally elevated in cortex and hippocampus within hours following the insult (Murakami et al., 1998); in a lateral fluid percussion model APP was overexpressed as early as 1 h after the insult (Pierce et al., 1996). In a weight fall model of brainstem injury in adult rats APP mRNA levels rose as soon as 1 h post-impact, peaked 3 h after the injury at almost twofold baseline level and declined to baseline within 24 h (Yang et al., 2014). Similarly, in an ovine TBI model APP mRNA was up-regulated as soon as 30 min post-impact (Van den Heuvel et al., 1999).

# Protective Function of APP and APPsα in TBI

While over-expression of APP following mechanical insults has been observed several decades ago, the functional effects remained unclear until recently. By now, evidence from different animal models points towards an acute neuroprotective effect of APP and APPsα in TBI (Plummer et al., 2016). In diffuse traumatic injury in rats, intraventricular administration of APPsα 30 min after the insult reduced axonal injury and apoptosis and improved motor and cognitive outcome (Thornton et al., 2006). In the same model of TBI, mice lacking APP suffered from greater cognitive and motor impairment in correspondence with larger lesions and increased hippocampal cell loss as compared to WT, again suggesting a protective role of APP in TBI (Corrigan et al., 2012a). Once again, posttraumatic application of exogenous APPsα mitigated these deficits (Corrigan et al., 2012b). These protective effects were found to be mediated by the heparin-binding D1 and D6a domains of APPsα (Corrigan et al., 2011). In an additional study conducted by the same group, the neuroprotective site was pinned down to the APP96-110 sequence in D1, which, applied intraventricularly post-trauma, was enough to significantly improve histological and functional outcome (Corrigan et al., 2014).

At first glance, these findings seem to contradict exacerbation of amyloid pathology and increased risk of AD following TBI. However, there are several possibilities how the two mechanisms may be reconciled. First, APPsα may exert neuroprotective functions independent from the detrimental effects of Aß or

FIGURE 1 | Pathophysiological changes in neurons following acute ischemic and traumatic insults. Micro- and macroscopic focal strokes, global hypoxia-ischemia and traumatic brain injury (TBI) lead to abruption of extracellular glucose and oxygen supply and excessive glutamate release. One major shared pathomechanism is NMDAR-mediated excitotoxicity, or over-activation of NMDAR by glutamate, which facilitates sodium and calcium influx. Due to excessive ion influx, the cellular membrane potential is depolarized, which leads to activation of voltage gated calcium channels such as LTCC, initiating a vicious cycle of ion influx, calcium overload, depolarization and aberrant activity. Successively calcium from intracellular calcium stores, particularly mitochondria and the ER, is released, increasing calcium levels to up to 200-fold of ∼100 nM during resting. Calcium activates secondary messengers that are able to translocate to the nucleus and modulate gene transcription. Long-lasting or severely elevated calcium levels may lead to activation of caspases and apoptosis. Following the osmotic gradient, water enters the cell and leads to cell swelling and brain edema. Due to glucose and oxygen shortage and excessive formation of reactive oxygen species (ROS), mitochondrial function is compromised and ATP production halts. Malfunction of the energy demanding ion pumps, predominantly the sodium potassium pump, ultimately leads to breakdown of the membrane potential, a phenomenon known as anoxic or hypoxic spreading depolarization or spreading depression (due to depression of network activity in the field potential recording). Given the energy supply is timely restored, this stage can be reversed without long-lasting morphological damage. If the insult is protracted, neurons might undergo (dependent of insult's severity) necrotic or apoptotic death or degenerate with a delay of days to decades due to synaptic or metabolic malfunction. Acute cell death and delayed degeneration contribute to brain atrophy and development of dementia. Glu, glutamate; Gluc, glucose; LTCC, L-Type calcium channel; NMDAR, NMDA receptor; AMPAR, AMPA receptor; M, mitochondrion; NCL, nucleus; ER, endoplasmic reticulum; ROS, reactive oxygen species.

amyloid plaques. Second, APPsα could prevent deposition of Aß and further growth of plaques. Third, APPsα may promote clearance of plaques. There is evidence for all three mechanisms. APPsα was shown to modulate BACE activity, possibly inhibiting amyloidogenic cleavage (Obregon et al., 2012). As described in sections ''APPsα and APPsβ'' and ''Mechanisms of Neuroprotection by APP and APPsα Protection in Hypoxia-Ischemia, Excitotoxicity, Degeneration'', APPsα counteracts Aßmediated excitotoxic damage and delayed degeneration by various trophic and regulatory effects on calcium homeostasis, synaptic function and survival pathways. Recently Fol et al. (2016) discovered that APPsα can ameliorate amyloid pathology by recruitment of microglia, underlining its involvement in clearance of amyloid.

# BRAIN ISCHEMIA, APP AND AD

#### Over-Expression, Amyloidogenic Processing of APP and Increased Risk of AD

Cardiovascular diseases and ischemic stroke share overlapping genetic and metabolic risk factors such as hypertension, dyslipidemia, glucose intolerance or diabetes and adipositas (Arboix, 2015). Recently these risk factors were established to also increase the odds of developing AD (Orehek, 2012; Wiesmann et al., 2013; Traylor et al., 2016). Moreover, hypoxicischemic conditions of the brain such as in ischemic stroke (Honig et al., 2003), heart arrest (de la Torre, 2006), and cerebral small vessel disease (Cai et al., 2015) directly correlate with AD risk, suggesting that cerebrovascular dysfunction is one possible cause of the neurodegenerative disease (Humpel, 2011; Orehek, 2012). Data from a large meta-analysis (Zhou et al., 2015) and a longitudinal study with over 6500 participants (Tosto et al., 2016) show that ischemic stroke increases AD risk by about 1.6 to 2.2-fold, respectively. Several studies indicate that, vice versa, AD patients have an increased risk to develop ischemic (Chi et al., 2013) and hemorrhagic (Chi et al., 2013; Tolppanen et al., 2013; Zhou et al., 2015) stroke and have a higher prevalence of cerebrovascular lesions (Jellinger, 2010). Other studies, however, did not find an increased risk of ischemic stroke in patients with AD (Imfeld et al., 2013; Tolppanen et al., 2013; Zhou et al., 2015). Not surprisingly, cerebrovascular disease and AD contribute additively to cognitive impairment in patients (Hohman et al., 2015) and mouse models (Pimentel-Coelho et al., 2013), possibly forming a vicious cycle of ischemia and neurodegeneration (Pluta et al., 2013).

It has been suggested that cerebrovascular disease, vascular dementia and AD share common pathophysiological cascades such as altered APP processing (Selnes et al., 2010), perturbed energy metabolism (Chen and Zhong, 2013) and pathological immune response (Brod, 2000). These common pathways may then result in overlapping histopathological findings (de la Torre, 2002; Pluta et al., 2009, 2012; Attems and Jellinger, 2014). APP overexpression and Aβ deposition likely play a pivotal role in these processes. In ischemic stroke patients, expression of APP is indeed increased (Pottier et al., 2012) and serum Aβ levels are elevated, correlating with infarct size and clinical outcome (Lee et al., 2005). Likewise, patients who suffered from hypoxia during a cardiac arrest present with increased Aβ levels, which—again—correlate with clinical outcome (Zetterberg et al., 2011). Increased age-related deposition of Aβ was also shown in chronic cerebral vascular disease in rats (Schreiber et al., 2014). However, elevation of Aβ following ischemia is transient. A recent study employing Pittsburgh Compound-B positron emission tomography (11C-PiB-PET; an in vivo imaging method of amyloid), revealed no accumulation of Aβ in patients 18 months after ischemic stroke (Sahathevan et al., 2016).

## Neuroprotective Role of APP in Ischemia in Animal Studies

At the first glance, findings concerning APP in conditions of hypoxia/ischemia seem to be contradictory. On the one hand, the pathological role of APP is supported by multiple animal studies. On the other hand, several studies show beneficial effects of APP in animal models of hypoxia-ischemia. It can be assumed that these opposing effects are mediated by the different cleavage products of APP.

On the one hand, ischemia and oxidative stress enhance BACE-1 and γ-secretase activity, resulting in increased Aβ deposition in rats and mice (Sun et al., 2006; Guglielmotto, 2009; Li et al., 2009). APP accumulates in regions of neurodegeneration following focal cerebral ischemia in the rat (Stephenson et al., 1992). Stroke in rats with Aβ pathology leads to aggravated comorbidity, hippocampal atrophy, and cognitive impairment, similar to the consequences of stroke in AD patients (Amtul et al., 2014).

On the other hand, postischemic intraventricular application of APPsα increases neuronal survival in a model of transient focal ischemia in rats (Smith-Swintosky et al., 1994). APP-KO as well as BACE-KO mice are unable to maintain cerebral blood flow and experience drastically increased acute mortality in a model of global cerebral ischemia (Koike et al., 2012). Overexpression of APP provides neuroprotection following middle cerebral artery occlusion in rats (Clarke et al., 2007). There is compelling evidence that APP acts as a potent anti-thrombotic agent (Van Nostrand, 2016). Moreover, it is required for effective immune and glial cell responses to inflammatory stimuli (Carrano and Das, 2015). With glutamate excitotoxicity being a key pathomechanism of ischemic neuronal damage (Broughton et al., 2009), activation of ADAM10, and thus facilitation of APPsα production, provides neuroprotection against excitotoxic stress in vivo (Clement et al., 2008). These findings are in line with models of AD where expression of APPsα protects against neurodegeneration and rescues synaptic function (Fol et al., 2016).

Taken together, these data support the importance of balance between the beneficial APPsα and the neurotoxic amyloidogenic pathway, thus resolving the initially contradictive results.

# MECHANISMS OF NEUROPROTECTION BY APP AND APPsα PROTECTION IN HYPOXIA-ISCHEMIA, EXCITOTOXICITY, DEGENERATION

As outlined in previous sections, ischemia, traumatic injury and degeneration share some common pathological cascades leading to neuronal death (see also **Figure 1**). One common mechanism of damage is dysregulation of calcium homeostasis (Mattson et al., 1993b; Webster et al., 2006). Intracellular calcium levels at rest are around 100 nM, and fluctuations in cytosolic calcium concentration are tightly coupled to metabolic and synaptic activity (Berridge et al., 2003). Neuronal calcium homeostasis is disturbed in AD, with strong evidence pointing towards a pivotal role of Aβ in destabilizing the balance between Hefter and Draguhn APP as a Neuroprotective Factor

mechanisms increasing and decreasing free intracellular calcium (Khachaturian, 1994; LaFerla, 2002; Green and LaFerla, 2008; Berridge, 2010). Similarly, TBI as well as ischemic-hypoxic insults lead to drastic elevations of cellular calcium of up to 20 µM (Yao and Haddad, 2004; Sun et al., 2008). Such acute, strong increases, as well as longer-lasting mild perturbations of calcium levels initiate a plethora of pathological cascades and can, finally, activate caspases and initiate apoptosis (Mattson and Chan, 2003; Orrenius et al., 2003).

APP and its metabolites, most of all APPsα, intervene with these cascades on multiple levels and exert neuroprotective effects under various conditions of cellular stress, revealing novel possible therapeutical leverage points (Kögel et al., 2012). APPsα was shown to mediate neuroprotection and stabilize intracellular calcium levels in in vitro models of excitotoxicity (Mattson et al., 1993a; Ma et al., 2009). The secreted form of APP also protects against Aβ-mediated toxicity in rat hippocampal cell cultures by attenuating Aβ-mediated calcium elevation (Goodman and Mattson, 1994). In a recent study on acute hippocampal slices (Hefter et al., 2016) we showed that APP protects neuronal function in acute hypoxia and promotes recovery of neuronal activity. The protective effects were largely exerted by the APPsα fragment and mediated by inhibition of L-type calcium channels (LTCC). These voltage-gated calcium channels (VGCC) are beside other calcium-permeable membrane channels such as NMDA receptors (NMDAR) and internal stores major sources of intracellular calcium (Yao and Haddad, 2004; Thibault et al., 2007), thus contributing to traumatic/ischemic neuronal damage as well as to the pathophysiology underlying AD. **Figure 2** summarizes major neuroprotective mechanisms of APP and APPsα as discussed below.

#### Modulation of NMDA Receptors

Traumatic and ischemic injury is marked by aberrant neuronal activity and excessive glutamate release from neurons and glia, mediating excitotoxicity through enhanced activation of glutamate receptors including the calcium-permeable NMDAR. These processes form a vicious cycle of excessive cation influx, further depolarization, opening of more channels and, eventually, breakdown of the membrane potential, osmotic cell swelling and death (Broughton et al., 2009; McAllister, 2011; see also **Figure 1**). Application of NMDAR blockers is an established neuroprotective strategy in models of excitotoxicity, hypoxiaischemia and TBI (Kubo et al., 2001) models. Remarkably, APPsα suppresses NMDAR-mediated currents (Furukawa and Mattson, 1998), potently attenuating calcium responses and thus providing protection against NMDAR-mediated excitotoxicity in hippocampal cell culture (Furukawa et al., 1996; Furukawa and Mattson, 1998; **Figure 2**). Seemingly contradicting these results, APPsα was shown to enhance LTP in acute hippocampal slices (Ring et al., 2007) as well as in vivo, where intrahippocampal application of the protein increased NMDAR currents, rescued LTP and memory performance (Taylor et al., 2008). This apparent discrepancy may be due to activation of different NMDAR subtypes which, dependent on their subcellular localization (synaptic vs. extrasynaptic) may promote either synaptic potentiation or proapoptotic effects (Hardingham

APP is activity-dependent and α-secretases are stimulated by NMDAR, generating the neuroprotective APPsα fragment. APPsα acts inhibitory on NMDAR and LTCC. This negative feedback mechanism may breach the vicious cycle of excitotoxicity and constitute an important protective mechanism in response to acute insults. Several further trophic, regulatory and anti-apoptotic functions of APP and APPsα are listed. They may contribute to acute neuroprotective effect on multiple levels. Since exact mechanisms of interaction are oftentimes not known, this ambiguity is represented by dashed arrows. The triple period below indicates that the list makes no claims of being complete since many more mechanisms are being discussed.

et al., 2002; von Engelhardt et al., 2007). In our experiments slices from wildtype mice showed postischemic potentiation of evoked population responses whereas synaptic transmission in slices from APP-KO mice was drastically reduced (Hefter et al., 2016). As this kind of plasticity also depends on NMDAR (Maggio et al., 2015), it may also be modulated by APP.

However, we could not observe involvement of NMDAR in APP-mediated protection from hypoxia. Taking into account that secretion of APPsα is activity-dependent (Kirazov et al., 1997; Gakhar-Koppole et al., 2008), its effects on NMDAR may provide a negative-feedback loop on extrasynaptic NMDAR in excitotoxicity or a positive feedback loop on subsynaptic NMDAR in LTP and learning.

# Modulation of L-Type Calcium Channels

LTCC belong to the family of VGCC, which are—depending on the membrane potential—almost exclusively conductive for calcium (Zuccotti et al., 2011). They are one of the major sources of extracellular calcium influx in ischemia (Cataldi, 2013) and contribute to neurodegeneration in AD when over-activated by Aβ (Webster et al., 2006). The subtype Cav1.2 has been identified as a potential pharmacotherapeutical target (Anekonda and Quinn, 2011). Several studies point towards beneficial effect of LTCC blockers in AD patients (Anekonda and Quinn, 2011; Hefter and Draguhn APP as a Neuroprotective Factor

Lovell et al., 2015) as well as in animal models of ischemia and neurodegeneration (Gholamipour-Badie et al., 2013). APP was shown to interact directly with Cav1.2 in cultured hippocampal and striatal inhibitory interneurons, with lack of APP resulting in aberrant activity of Cav1.2 and altered short-term plasticity (Yang et al., 2009). In primary cultures of rat cortical neurons expression of human APP inhibited calcium oscillations by modulation of LTCC, suggesting a pivotal role in control of neuronal excitability (Santos et al., 2009). In line with these results we recently found an important role of LTCC for APP-mediated neuroprotection in hypoxia (Hefter et al., 2016). These studies suggest that regulation of LTCC function and thereby cytosolic calcium levels by APP may be neuroprotective. However, the molecular mechanisms underlying regulation of LTCC by APP, the function in healthy neurons and the role in ischemia and degeneration remain elusive.

## Effect of APP on Intracellular Calcium Stores

Internal calcium stores, most importantly the ER and mitochondria, play a major role in the regulation of intracellular calcium homeostasis and contribute to elevations of calcium levels under pathological conditions (Mattson et al., 2000). Regulation of store-related calcium homeostasis appears to be mediated by the intracellular domain of APP, i.e., AICD. In cell culture studies, AICD-deficient cells show increased cytosolic calcium concentrations, decreased ability of the ER to buffer calcium and decreased levels of ATP (Hamid et al., 2007). Although a direct binding of AICD to ER receptors such as ryanodine or inositol triphosphate (IP3) receptors has not been described, indirect effects on ER stores and on calcium signaling in general are discussed. One such mechanism is modulation of phosphoinositide-regulated signaling by regulation of the PIKfyve complex, an essential kinase that synthesizes phosphatidylinositol-3,5-bisphosphate. Its loss of function results in neurodegeneration (Balklava et al., 2015; Currinn and Wassmer, 2016). APP was also proposed to modulate IP3 by affecting the transcription of GSK 3b (Hamid et al., 2007). Moreover, AICD may affect calcium levels by binding to X11, BP1, ShcA and other adaptor proteins which might link it to calcium signaling pathways (LaFerla, 2002) and regulate the expression of genes involved in calcium homeostasis such as S100a9 (Leissring et al., 2002; Pardossi-Piquard and Checler, 2012).

# Effect on Survival/Apoptosis Signaling Pathways and Gene Expression

In severely compromised tissue, such as the ischemic core in stroke, neurons undergo necrosis due to osmotic swelling, lack of energy metabolites and breakdown of ion gradients (Lo et al., 2003). Under milder and longer-lasting metabolic stress, such as in the ischemic penumbra zone or in chronic cerebral hypoperfusion, the balance between anti- and proapoptotic pathways inclunding NF-κB and p53-pathways may tilt towards apoptotic death (Dirnagl et al., 1999; Broughton et al., 2009). In studies on cultured cells, APPsα was shown to exert anti-apoptotic effects by mechanisms such as upregulation of immediate early gene transcription factors, activation of CREB and NF-κB, genes related to cell survival (Guo et al., 1998; Ryan et al., 2013), phosphorylation of glycogen synthase kinase 3β (GSK-3β; Jimenez et al., 2011) or regulation of expression of cyclin-dependent kinase 5 (CDK-5; Hartl et al., 2013). Effects may be mediated by binding of APPsα to several different receptor proteins which are not yet unambiguously identified (Gustafsen et al., 2013). Potential targets include membranebound APP itself (Milosch et al., 2014), and direct inhibition of BACE-1 by APPsα which would counteract Aβ-mediated neurotoxicity (Obregon et al., 2012). Further protection of amyloid toxicity by APPsα was mediated by increased expression of the neuroprotective proteins transthyretin and insulin-like growth factor 2 and subsequent inhibition of the proapoptotic BAD (Stein et al., 2004). AICD was described to interact with more than 20 adaptor proteins including Fe65 proteins members of the Mint/X11 family and members of the JIP family (c-jun-Nterminal kinase interacting protein, JIP1b and JIP2), translocate to the nucleus and interact with survival-related genes (Pardossi-Piquard and Checler, 2012).

### Effects on Neurogenesis and Proliferation

Recent years have provided evidence for neurogenesis in several regions of the adult human central nervous system including the dentate gyrus, striatum and olfactory bulb. This mechanism can, in principle, enhance cognitive functions and support recovery from neuronal damage (Inta and Gass, 2015). APPsα was found to stimulate proliferation of neuronal progenitor cells in the subventricular zone (Caillé et al., 2004) and in the hippocampus (Baratchi et al., 2012), whereas AICD was reported to have antiproliferative effects (Zhou et al., 2011). An imbalance between these APP products and thus between neurogenesis and degeneration may contribute to the development of AD (Zhou et al., 2011). Recently APP was shown to control adult hippocampal neurogenesis through GABAergic interneurons, regulating GABAergic synaptic transmission (Wang et al., 2014). APP's known trophic effects on neuronal viability, cell adhesion, axonogenesis, dendritic arborization and dendritic spines may also contribute to recovery from traumatic and metabolic insults and counteract degeneration (Perez et al., 1997; Lee et al., 2010). APPsα-mediated trophic effects are activity-dependent and stimulated by activation of 5-HT<sup>4</sup> and NMDAR (Gakhar-Koppole et al., 2008; Cochet et al., 2013), suggesting them to be a feasible adaptive strategy in LTP, plasticity and excitotoxicity.

# POTENTIAL THERAPEUTIC STRATEGIES

Currently available pharmacological therapies in AD are mostly based on acetylcholine esterase inhibitors such as donepezil and rivastigmine or NMDAR blockers like memantine. They are far from eliminating the (unknown) primary cause of the disease, but do only alleviate symptoms and delay disease progression (Huang and Mucke, 2012). Likewise, immunotherapeutic approaches with antibodies against Aβ reduce amyloid burden, but show only limited success in the prevention of cognitive decline in ongoing phase III clinical trials (Reiman, 2016). One

reason for the slow and tedious progress in therapy of AD may be that cognitive deficits in AD arise not only due to an excess of toxic metabolites, but also from loss of function of protective APP products. Likewise, the past two decades of research on neuroprotective strategies in ischemic stroke and TBI have been hampered by failures to translate results from bench to bedside (Hoyte et al., 2004; O'Collins et al., 2006). Mechanistic understanding of APP's role in these diseases may help to break this streak. Restitution of the perturbed balance between harmful and beneficial APP metabolites emerges as a promising neuroprotective strategy. Although the hope to find a ''cure for all'' seems delusive, shared pathological mechanisms in ischemia, injury and AD imply that discovery of common leverage points for novel drugs may be feasible. We will briefly discuss such potential therapeutic strategies which may comprise activation of α-secretases, inhibition of ß- and γ-secretases, exogenous administration of APPsα or amyloid antibodies, control of cellular calcium levels by block of LTCC and NMDAR, activation of neuroprotective mechanisms and inhibition of proapoptotic downstream targets of APP (**Figure 3**; Selkoe, 2011).

#### Activation of α-Secretases

A variety of potential strategies may shift the balance towards non-amyloidogenic cleavage of APP, including modulation of expression, trafficking and regulation of ADAM10 (Postina, 2012). Direct activation of α-secretases by etazolate has been shown to be beneficial in TBI in mice (Siopi et al., 2013). As an indirect mechanism, activation of muscarinic M1 acetylcholine receptors has been reported to increase α-secretase cleavage of APP and decrease Aß levels (Beach et al., 2001). Another activator of α-secretase and inhibitor of β- und γ-secretase is melatonin (Mukda et al., 2016). Increased dimerization of APP via specific compounds such as disulfiram was shown to shift the balance towards non-amyloidogenic cleavage products of APP and thus may present a novel therapeutic approach (Libeu et al., 2012). However, overexpression or activation of ADAM-10 may also have harmful consequences due to effects on other substrates of this enzyme (Clement et al., 2008).

#### Inhibition of β- and γ-Secretases

In mice, a selective γ-secretase inhibitor has already been successfully tested, reaching a 33% reduction of Aβ levels within 1 week, without causing severe side effects (Basi et al., 2010). In a mouse model of TBI, pharmacological inhibition of γ-secretase activity reduced post-traumatic tissue loss and improved motor and cognitive recovery (Loane et al., 2009). Both γ- and β-secretase process various other substrates than APP, complicating the use of respective inhibitors (John et al., 2003). However, strategies to specifically inhibit APP cleavage by BACE-1 do exist (Ben Halima et al., 2016) and first BACE-1 inhibitors made it into clinical trials (Vassar et al., 2014).

#### Regulation of Calcium Homeostasis

Regulation of intracellular calcium is a promising neuroprotective strategy (Duncan et al., 2010). As discussed above, APP stabilizes calcium homeostasis by interacting with LTCC, NMDAR and other signaling pathways, offering some feasible pharmacological leverage points. LTCC blockers of the dihydropyridine family such as the common antihypertensive drugs nimodipine and nifedipine attenuate progression of dementia in humans, inhibit Aβ formation in cell culture (Lovell et al., 2015), counteract Aβ-mediated calcium increase and excitotoxicity (Anekonda and Quinn, 2011) and alleviate Aβrelated memory deficits in animal models (Gholamipour-Badie et al., 2013). The NMDAR blocker memantine is not only an established drug for treatment of AD (Danysz and Parsons, 2012), but has also protective effects against excitotoxicity in small doses, being potentially beneficial in patients with high risk of ischemic stroke (Trotman et al., 2015). Other potential calcium-stabilizing approaches target downstream pathways of AICD (Nagase and Nakayama, 2014).

# Delivery of Exogenous APPsα

As proven in rodent models of TBI, intraventricular application of exogenous APPsα or its heparin binding domain promote neuronal survival and improve functional outcome (Corrigan et al., 2014). Following TBI or malignant stroke, patients often receive a decompressive craniotomy including ventricular drainage or insertion of an intracranial pressure probe. Application of APPsα through these entries seems feasible. However, these results are highly preliminary and it remains to be proven whether this technique is safe, beneficial and technically feasible in human patients.

# CONCLUSION

The immense multitude and complexity of APP interactions and functions discovered in recent decades may seem overwhelming and evoke the concern to miss the forest for the trees. Nevertheless, some common principles have emerged from recent studies. APP is more than the mother molecule of amyloid, and AD is more than an amyloido-tauopathy. In this review article, we present convergent evidence from human studies, animal models and in vitro experiments for a neuroprotective role of APP in ischemia, brain injury and neurodegeneration. Most studies suggest that these neuroprotective and trophic effects are mainly conducted by the extracellularly secreted fragment APPsα, whereas amyloidogenic cleavage leads to various harmful consequences. We hypothesize that under pathological conditions the cleavage balance of APP is disturbed, and loss of its neuroprotective function may contribute to disease development. While the pathological role of APP in AD may result from an overshoot of pathological products of APP (Aß), production of the neuroprotective soluble fragment APPsα may, in turn, reflect the normal, beneficial reaction of the organism to metabolic challenges. Therefore shifting this balance towards APPsα secretion may be a promising treatment strategy in AD, stroke

and TBI. Causal treatments are urgently needed in these conditions. Novel therapeutic targets arise from unraveling the mechanisms of APP-mediated neuroprotection such as regulation of cellular calcium levels by LTCC and NMDAR

#### REFERENCES


inhibition, regulation of survival and apoptosis signaling pathways, trophic effects on synapto- and neurogenesis, synaptic function, plasticity and memory formation. However, current understanding of these highly complex processes and the specific contributions of APP is far from complete, and successful translation into clinic is still a major challenge. One of the reasons might be the predominant focus on histopathological endpoints in most studies in the field, largely neglecting longitudinal functional studies. Deeper comprehension of APP-related processes in living tissue, employing functional electrophysiological and imaging techniques should complement morphological studies. Combined (interventional) functional and structural evidence may help to develop new neuroprotective therapies.

#### AUTHOR CONTRIBUTIONS

DH and AD designed, drafted, wrote and revised this work and approved this version to be published. DH designed and created the figures.

#### FUNDING

This study was supported by the DFG Research Group 1332.


on brain aging outcomes. Neurobiol. Aging 36, 2501–2508. doi: 10.1016/j. neurobiolaging.2015.05.021


targets for traumatic brain injury. Nat. Med. 15, 377–379. doi: 10.1038/ nm.1940


secretases expression in aged mouse hippocampus. Neurosci. Lett. 621, 39–46. doi: 10.1016/j.neulet.2016.04.013


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Hefter and Draguhn. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Therapeutic Potential of Secreted Amyloid Precursor Protein APPsα

Bruce G. Mockett 1† , Max Richter 2† , Wickliffe C. Abraham1‡ and Ulrike C. Müller <sup>2</sup> \* ‡

<sup>1</sup>Department of Psychology, Brain Health Research Centre, Brain Research New Zealand, University of Otago, Otago, New Zealand, <sup>2</sup>Department of Functional Genomics, Institute for Pharmacy and Molecular Biotechnology, Heidelberg University, Heidelberg, Germany

Cleavage of the amyloid precursor protein (APP) by α-secretase generates an extracellularly released fragment termed secreted APP-alpha (APPsα). Not only is this process of interest due to the cleavage of APP within the amyloid-beta sequence, but APPsα itself has many physiological properties that suggest its great potential as a therapeutic target. For example, APPsα is neurotrophic, neuroprotective, neurogenic, a stimulator of protein synthesis and gene expression, and enhances long-term potentiation (LTP) and memory. While most early studies have been conducted in vitro, effectiveness in animal models is now being confirmed. These studies have revealed that either upregulating α-secretase activity, acutely administering APPsα or chronic delivery of APPsα via a gene therapy approach can effectively treat mouse models of Alzheimer's disease (AD) and other disorders such as traumatic head injury. Together these findings suggest the need for intensifying research efforts to harness the therapeutic potential of this multifunctional protein.

#### Edited by:

Oliver Wirths, University of Göttingen, Germany

#### Reviewed by:

Natalia N. Nalivaeva, University of Leeds, UK Davide Tampellini, Institut National de la Santé et de la Recherche Médicale (INSERM), France

> \*Correspondence: Ulrike C. Müller u.mueller@urz.uni-hd.de

†These authors have contributed equally to this work. ‡Joint senior authors.

Received: 09 December 2016 Accepted: 25 January 2017 Published: 07 February 2017

#### Citation:

Mockett BG, Richter M, Abraham WC and Müller UC (2017) Therapeutic Potential of Secreted Amyloid Precursor Protein APPsα. Front. Mol. Neurosci. 10:30. doi: 10.3389/fnmol.2017.00030 Keywords: Alzheimer's disease, amyloid precursor protein, APPsα, synaptic plasticity, neuroprotection

# INTRODUCTION

Secreted amyloid precursor protein-alpha (APPsα, also known as soluble APPα), when generated from the neuronally expressed APP695 isoform by the action of α-secretase (**Figure 1**), is a 612 amino acid protein that was first shown in the mid-1990s to promote the survival and growth of cultured neurons under physiological and non-physiological conditions (e.g., glucose and oxygen deprivation, amyloid-β (Aβ) toxicity; Mattson et al., 1993; Barger and Mattson, 1996a; Furukawa et al., 1996). These observations have been supported and extended by myriad reports over the intervening years (Ryan et al., 2013; Hefter et al., 2016) and has generated suggestions that the promotion of α-secretase cleavage of APP and increasing APPsα levels could be a therapeutic strategy for the treatment of Alzheimer's disease (AD; Turner et al., 2003; Ring et al., 2007; Postina, 2012; Hick et al., 2015; Fol et al., 2016; Habib et al., 2016) and possibly other neurological disorders. The purpose of this review is to consider the extent to which APPsα generation may be disrupted in AD, and summarize the many positive functions of APPsα that could be lost in the disease. In addition we will discuss the potential that either enhancement of non-amyloidogenic processing of APP or upregulating the expression of APPsα by other means has for preventing or at least slowing the progression of AD as well as treating other neurological disorders.

# APP PROCESSING

APP is a single pass type I transmembrane protein that undergoes complex proteolytical processing by several enzymes termed secretases. In the amyloidogenic pathway, APP processing is initiated by β-secretase (β-amyloid cleaving enzyme, BACE-1), a transmembrane aspartate-type protease (for review see Vassar et al., 2014) that cleaves APP at the N-terminus of Aβ, leading to the secretion of the soluble ectodomain APPsβ (**Figure 1A**). In the competing and physiologically predominant non-amyloidogenic pathway, α-secretase cleaves APP within the Aβ region (**Figure 1A**), in a process that can be stimulated by neuronal and synaptic activity (Hoey et al., 2009; Hoe et al., 2012). This not only prevents the formation of Aβ peptides but also leads to the secretion of the ectodomain APPsα, which is only 16 amino acids longer than APPsβ (**Figure 1B**), into the extracellular space. Several members of the ADAM (a disintegrin and metalloprotease) family including ADAM9, ADAM10 and ADAM17, transmembrane Zn-proteases located at the cell surface, are able to cleave APP at the α-secretase site in vitro (reviewed by Saftig and Lichtenthaler, 2015). In neurons ADAM10 serves as the major physiological α-secretase as demonstrated by pharmacological inhibition and knockdown in vitro, as well as brain-specific knockout (KO) in vivo (Kuhn et al., 2010; Colombo et al., 2013; Prox et al., 2013). Subsequent processing of the membrane tethered C-terminal fragment resulting from β-secretase activity (CTFβ) by γ-secretase liberates Aβ and the APP intracellular domain (AICD), while CTFα processing yields the p3 fragment. γ-secretase cleavage occurs within the membrane by a complex of transmembrane proteins containing as a catalytic core presenilin (PS) 1 or 2. In wild-type neurons the predominant Aβ species generated is Aβ40, whereas familial forms of AD (FAD) linked to PS1 mutations result in a higher proportion of longer, more aggregation prone Aβ species including Aβ42 and Aβ43 that are believed to trigger plaque deposition (Veugelen et al., 2016).

#### ALZHEIMER'S DISEASE

AD is a progressive neurodegenerative disease for which aging is the most significant risk factor. It has traditionally been diagnosed by the appearance of functional deficits that frequently begin with self-reporting of impaired episodic memory (Dubois et al., 2007). Definitive diagnosis, however, requires post-mortem confirmation, although in recent times a number of biomarkers are providing new ways of diagnosing in life, such as medial temporal lobe atrophy with hippocampi volume loss, abnormal cerebrospinal fluid levels of the neurotoxic Aβ peptide and tau protein, plus positron emission tomography evidence for amyloid plaques and reduced glucose metabolism (Jack and Holtzman, 2013). While the proximal causes of sporadic AD are largely unknown, the familial forms arise when any one of several autosomal dominant mutations in genes regulating the production and clearance of Aβ are present (Dubois et al., 2007, 2010).

The post-mortem neuropathology of AD is characterized by the extensive development of extracellular plaques containing Aβ that are generated by amyloidogenic processing of APP (**Figure 1A**), intraneuronal hyperphosphorylated tau leading to neurofibrillary tangles, neuroinflammation and cell loss. Moreover, accumulation of intraneuronal Aβ has been observed as an early event in transgenic animal models (Kumar et al., 2013) and may contribute to pathogenesis (Zou et al., 2015; Ji et al., 2016). Sub-clinical progression of AD may occur over 15–20 years prior to diagnosis (Jack and Holtzman, 2013). This early phase of the disease is characterized by the formation of soluble oligomeric forms of Aβ that cause neuronal dysfunction and toxicity that may underpin early cognitive deficits. At the center of this early dysfunction in particular is impairment of synaptic function. Investigations in both AD patients and in mouse models of AD have revealed significant reductions in dendritic spine density in both cortical and subcortical regions early in the disease that are highly correlated with the appearance of cognitive deficits (Scheff et al., 1990, 2006; Terry et al., 1991; Spires-Jones and Knafo, 2012). Compensatory enlargement of remaining synapses has been reported and may mitigate some of the early losses in spine density; as AD progresses, however, spine loss exceeds synaptic growth leading to a net reduction in synaptic transmission (Scheff et al., 1990). Further progression of AD results in loss of dendritic complexity (reduced length, less branching, changes in dendrite diameter) and eventually cell death (Alpár et al., 2006).

An important pathology associated with synaptic dysfunction is the impairment in the synaptic plasticity mechanisms hypothesized to underpin learning and memory. The most extensively studied form of synaptic plasticity, long-term potentiation (LTP), is reliably impaired in most animal models of AD and can also be caused by extracts obtained from post-mortem AD brain (Oddo et al., 2003; Shankar et al., 2008; Li et al., 2011). The impairment of LTP observed in animal models and from raised Aβ levels may in part relate to altered transmission and loss of dendritic spines (reviewed by Spires-Jones and Knafo, 2012), as well as impairments in N-methyl-D-aspartate (NMDA) receptor expression and inhibition of LTP-associated de novo protein synthesis (Snyder et al., 2005; Li et al., 2011).

The treatment of AD has proven to be extremely challenging. Despite an exhaustive array of clinical trials that now number in the hundreds (Schneider et al., 2014), no disease-modifying treatments have proven effective for clinical use, although there is renewed hope arising from a recent study that has given very promising results from anti-Aβ antibody treatment (Sevigny et al., 2016). On the other hand, a lack of significant cognitive improvements was recently reported for a phase III clinical trial in patients with mild AD (EXPEDITION-3) using Solanezumab, an anti-Aβ antibody that binds only soluble Aβ (Hawkes, 2016<sup>1</sup> ). Thus, at present only two classes of drugs have been approved by the Food and Drug Administration for AD treatment and these only address the symptoms of the disease (Geldenhuys and Darvesh, 2015). Acetylcholinesterase inhibitors (e.g., donepezil)

<sup>1</sup>http://www.ctad-alzheimer.com/live-webcast

target the reduced cholinergic innervation of the hippocampus and cortex resulting from the loss of basal forebrain cholinergic neurons (Whitehouse et al., 1982), and memantine targets the increased tonic activation of extrasynaptic NMDA receptors that leads to activation of apoptotic pathways and neuronal death (Hardingham and Bading, 2010). While these treatments provide some symptomatic relief, their efficacy invariably reduces over time and ultimately they fail to halt or reverse the progression of the disease. Therefore, it is vital that new treatment options continue to be explored.

# A SHIFT IN THE BALANCE OF α-SECRETASE VERSUS β-SECRETASE ACTIVITY?

The amyloid cascade hypothesis has been the most widely supported explanation of the pathology that drives the progression of AD (De Strooper and Karran, 2016; Selkoe and Hardy, 2016), although other elements of the neuropathology are gaining increasing attention (Herrup, 2015; Rius-Pérez et al., 2015; Briggs et al., 2016). The amyloid cascade hypothesis contends that there is either a shift in APP processing towards the amyloidogenic pathway, or there is a reduction in Aβ clearance which results in the excessive accumulation of Aβ and a shift in the ratio of the various Aβ species to favor Aβ42. There is also evidence that BACE1 is upregulated during aging and AD, thus favoring amyloidogenic APP processing (Fukumoto et al., 2002, 2004; Holsinger et al., 2002; Yang et al., 2003; Li et al., 2004; Ahmed et al., 2010).

With the firm focus on increased levels of both soluble and insoluble Aβ in the brain and cerebrospinal fluid (CSF) in AD, relatively little attention has been given to a possible associated reduction in α-secretase activity and thus a shift away from the production of APPsα that might amplify the toxic effects of Aβ, hyperphosphorylated tau and other neuropathologies. However, the evidence for a reduction in APPsα levels in AD is mixed. Measuring mixed alpha and beta forms of secreted APP, Kibbey et al. (1993) reported that levels of APPs in the CSF of AD patients were 3.5 times lower than that in healthy controls. Subsequent studies specifically measuring APPsα in CSF supported this finding (Lannfelt et al., 1995; Almkvist et al., 1997; Sennvik et al., 2000), and positive correlations between reduced APPsα levels and diminished performance in cognitive testing in both AD patients (Almkvist et al., 1997) and normal aged rats (Anderson et al., 1999) have been reported. The loss of cholinergic innervation from the basal forebrain to the cortex and hippocampus in the very earliest stages of AD may underlie the loss of APPsα production and this may be the driver for the shift to amyloidogenic processing of APP (Obregon et al., 2012).

On the other hand, there is also evidence that APPsα levels may not be changed in the early stages of sporadic AD (Perneczky et al., 2014). Several studies using newly developed methodologies have reported that APPsα CSF and blood plasma levels are unchanged in sporadic AD patients (Olsson et al., 2003; Perneczky et al., 2011, 2013; Rosén et al., 2012; Brinkmalm et al., 2013) with decreases only in advanced AD (Rosén et al., 2012) and in AD patients carrying the ApoE-ε4 allele (Olsson et al., 2003). One study has even reported an increase in APPsα levels in the CSF of AD patients (Rosén et al., 2012). Thus, a complete understanding of the pattern of APPsα production in AD and its significance will require more detailed study of AD patients and testing in animal models of the disease.

While the production of APPsα in the brain still needs to be fully understood, evidence from studies in humans and animals indicates that reduced APPsα levels can exacerbate AD symptoms. A mutation at the α-secretase cleavage site of human APP (APP770K687N) was found to cause early onset dementia. The mutation severely reduced α-cleavage and thus APPsα production, but at the same time led to the production of highly toxic Aβ species, hampering a clear interpretation of the specific impact of low APPsα levels (Kaden et al., 2012). However, Epis et al. (2010) demonstrated that hippocampal ADAM10/SAP97 levels (a complex required for synaptic ADAM10 localization) are reduced in AD patients, while activity-attenuating mutations in the prodomain of the human ADAM10 gene have been associated with AD (Kim et al., 2009; Suh et al., 2013). Reducing ADAM10 activity in adult mice by impairing its trafficking (Epis et al., 2010) or through forebrain-specific conditional ADAM10 KO (Prox et al., 2013) shifted APP processing towards Aβ production. Together these data suggest that reduced APPsα levels may contribute to the early stages of sporadic AD.

# PROPERTIES AND FUNCTIONS OF APPsα

The possible significance of any impairments in ADAM10 activity or in the expression of APPsα becomes quickly apparent when one considers that this protein exerts a large number of growth factor-like properties when applied exogenously to neural tissue. Understanding the functionality of this protein, and its mechanisms of action, is crucial not only for understanding its biology in normal tissue, but also for providing critical information that will underpin any attempt to harness its potential therapeutic benefits (**Figure 2**).

# NEUROPROTECTION

APPsα has strong neuroprotective properties that mitigate in cultured neurons the effects of a range of pro-apoptotic insults including hypoglycemia and glutamate toxicity (Mattson et al., 1993; Furukawa et al., 1996) and, importantly, Aβ-induced

toxicity (Goodman and Mattson, 1994; Barger and Mattson, 1996a,b; Furukawa et al., 1996; Guo et al., 1998). More recently, we and others have demonstrated that the effects of other disease-associated insults such as excessive NMDA receptor activation (Ryan et al., 2013) and proteasomal impairments (Copanaki et al., 2010; Kundu et al., 2016) can be mitigated by APPsα administration. APPsα inhibits the upregulation of the co-chaperone BAG3 and suppresses BAG3-mediated aggresome formation under conditions of proteasomal stress (Kundu et al., 2016). Moreover, APPsα is a key activator of the PI3K/Akt survival signaling pathway that is triggered in response to serum withdrawal in cultured neurons (Milosch et al., 2014). Although the mechanisms conferring neuroprotection are only partially understood (for review see Kögel et al., 2012), some of these effects depend on the binding of APPsα to cell surface APP, that via its C-terminal domain can interact with G<sup>0</sup> protein to trigger the pro-survival Akt kinase pathway (Milosch et al., 2014).

While most previous studies focused on cell death, the impact of APP on cellular and neuronal network functions during metabolic stress remain largely unknown. In this regard, Hefter et al. (2016) recently studied hypoxia-induced loss of function and recovery upon re-oxygenation in mouse hippocampal slices. While APP-KO mice showed impaired functional recovery after transient hypoxia, this could be largely rescued by APPsα expression or by pharmacological block of L-type calcium channels. Voltage-gated Ca2<sup>+</sup> channels are, in addition to NMDARs and internal Ca2<sup>+</sup> stores, major sources of intracellular calcium contributing to traumatic/ischemic insults and AD pathogenesis. These data indicated that APP, in particular APPsα, supports neuronal resistance against acute hypoxia by regulating calcium homeostasis (Hefter et al., 2016).

In addition to these in vitro studies, there is also evidence that APPsα may protect against acute forms of brain injury in vivo. Smith-Swintosky et al. (1994) demonstrated that APPsα ameliorates neuron loss in the hippocampus under conditions of transient ischemia, consistent with subsequent findings that APP-KO mice show increased acute mortality upon ischemia (Koike et al., 2012). In addition, a series of recent experiments have shown a protective effect of APPsα in traumatic brain injury (reviewed by Plummer et al., 2016). Intracerebroventricular (ICV) administration of APPsα following traumatic injury in rats significantly reduced cell and axonal death and improved motor outcomes (Corrigan et al., 2014). While APP-KO mice are more vulnerable to traumatic brain injury this could be rescued by recombinant APPsα or peptides derived from it (Corrigan et al., 2014; Plummer et al., 2016). Together these data indicate that endogenous APPsα is neuroprotective under injury conditions and suggest that these properties may be exploited in a therapeutic setting.

In a positive feedback cycle, APPsα may promote the further production of APPsα by blocking the amyloidogenic pathway through binding to and inhibiting the β-secretase BACE1 (Peters-Libeu et al., 2015), leading to a reduction in Aβ production (Obregon et al., 2012; but see also Fol et al., 2016). Further protection against AD-related toxicity by APPsα may arise from the inhibition of the tau phosphorylating enzyme GSK3β, thus reducing tau hyperphosphorylation and the subsequent production of NFTs (Deng et al., 2015).

# TROPHIC FUNCTIONS: CELL PROLIFERATION AND ADULT NEUROGENESIS

In addition to neuroprotection, APPsα exerts trophic functions both in vitro and in vivo. Early studies indicated that APPsα restores the growth of fibroblasts in which endogenous APP expression had been attenuated (Saitoh et al., 1989), stimulates thyroid epithelial cell growth (Pietrzik et al., 1998) and enhances the proliferation of rat fetal neural stem cells (Hayashi et al., 1994; Ohsawa et al., 1999). While these trophic functions appear beneficial under physiological conditions, enhanced APPsα expression has been detected in different tumors including glioblastoma (for review see Chasseigneaux and Allinquant, 2012). APPsα has also been implicated in adult neurogenesis. APP knockdown in adult mice resulted in reduced numbers of neurospheres that could be cultured form the ventricular zone (Caillé et al., 2004) and an APP-Fc fusion protein (Fc domain of IgG fused to the APP ectodomain) was shown to bind to the subventricular zone of adult mice (Caillé et al., 2004), suggesting that APPsα may stimulate neuronal stem/progenitor cell proliferation. Consistent with these findings, APPsα infusion into the lateral ventricle increased the number of EGF-responsive progenitor cells (Caillé et al., 2004), while pharmacological blockade of α-secretase by infusion of the inhibitor batimastat decreased the number of neuronal progenitors in vivo (Caillé et al., 2004). This was further corroborated by in vitro studies indicating that APPsα stimulates the proliferation of cultured neuronal precursor cells (NPCs) from the adult subventricular zone even in the absence of EGF (Demars et al., 2011) and also NPCs from the dentate gyrus (Baratchi et al., 2012). Consistent with the latter finding, transgenic overexpression of ADAM10 led to increased hippocampal neurogenesis (Suh et al., 2013). In addition, intraventricular injection of APPsα rescued the age-dependent decline in the number of NPCs in vivo (Demars et al., 2013). Taken together these findings indicate a prominent role of APPsα in adult neuronal progenitor cell proliferation.

# ROLE FOR NEURITE OUTGROWTH, SYNAPTOGENESIS AND SPINE DENSITY

Several in vitro studies indicated that APPsα can promote neurite (Clarris et al., 1994; Small et al., 1994) and axonal outgrowth (Young-Pearse et al., 2008). Several APP domains important for these functions have been identified including the N-terminal APP96–110 region located in the first heparin-binding domain and the APP319–335 region which contains the RERMS motif (Ninomiya et al., 1994). Studies from animal models also indicate a crucial role for APPsα in synaptogenesis and modulation of spine density. Using organotypic hippocampal cultures we have demonstrated a pronounced decrease in spine density and reductions in the number of mushroom spines thought to represent mature synapses in CA1 pyramidal neurons of APP-KO mice. Interestingly, APPsα expression alone was sufficient to prevent the defects in spine density observed in APP-KO mice, as APPsα knock-in mice that lack transmembrane APP and express solely the secreted APPsα fragment exhibited unaltered spine density and spine type distribution (Weyer et al., 2014). In line with this, APPsα could also partially restore spine density deficits of cultured APP-KO neurons (Tyan et al., 2012). In turn, these findings imply that autocrine or paracrine APPsα signaling, important for spine formation and/or maintenance, involves a so far unknown receptor distinct from APP itself. Further support for a synaptotrophic role of APPsα comes from transgenic mice with moderate overexpression of human wild-type APP (Mucke et al., 1996) or indirect up-regulation of APPsα by transgenic expression of the α-secretase ADAM10 (Bell et al., 2008), which both led to increased synaptic density. In Tg2576 mice, expression of mutant hAPP increased spine density in CA1 and cortical neurons of young mice prior to plaque deposition presumably via APPsα, whereas spine density was decreased in aged animals, likely due to Aβmediated synaptotoxic effects (Lee et al., 2010). This suggests that APPsα might counteract Aβ-mediated effects on spine density during early stages of pathogenesis. Recent evidence indicates that APP also regulates spine plasticity. Using two-photon in vivo microscopy, (Zou et al., 2016) analyzed cortical spine dynamics in vivo and reported decreased spine turnover rates (formation of new spines or loss of established spines) in APP-KO mice. Moreover, when housed under environmental enrichment, APP-KO mice failed to respond with an increase in spine density (Zou et al., 2016), suggesting that not only a reduction in spine numbers but also alterations in spine dynamics could contribute to deficits in synaptic plasticity and behavior found in APP mutant mice (Dawson et al., 1999; Seabrook et al., 1999; Ring et al., 2007). It remains to be seen which domains of APP or which proteolytic fragment is important for this function. The mechanism underlying the effects of APPsα on spines is presently unknown, although NMDARs could play a crucial role. APP-KO mice have decreased levels of extracellular D-serine (Zou et al., 2016), an essential endogenous co-factor of NMDAR activation (Panatier et al., 2006). Taken together these findings indicate important synaptogenic and synaptic modifying properties of APPsα that may be of therapeutic value (Fol et al., 2016; see also below).

#### SYNAPTIC PLASTICITY

Synaptic plasticity phenomena, such as LTP and long-term depression (LTD), are fundamental to learning and memory and are thus also central to normal cognitive function. In mouse models of AD, LTP is consistently impaired in an age-dependent fashion (Oddo et al., 2003; Vigot et al., 2006), and in some cases LTD is facilitated (Megill et al., 2015), while humans with diagnosed AD also show impaired synaptic plasticity (Trebbastoni et al., 2016). It is interesting to note then that APPsα has the capacity to facilitate LTP and thus has the potential to counter the LTP-impairing effects of Aβ. In an early study, Ishida et al. (1997) demonstrated that APPsα increased the frequency dependency of LTD induction in CA1 from 1 Hz to 10 Hz and facilitated LTP expression induced by 100 Hz stimulation, possibly by a protein kinase G (PKG)-dependent mechanism. Moreover, we showed in anesthetized rats that exogenously applied APPsα exerted an inverted U-shaped dose-dependent facilitation of LTP in the dentate gyrus, although too high a dose impaired LTP (Taylor et al., 2008). Moreover, APPsα antibodies as well as an α-secretase inhibitor impaired LTP, and the latter effect could be rescued by exogenous APPsα but not by APPsβ, despite its lacking only the 16 C-terminal residues when compared to APPsα (**Figure 1B**). The inhibition of LTP appeared to be mediated, at least in part, through a reduction of NMDAR currents generated during the high-frequency stimulation (HFS). No effects on basal AMPA or NMDA receptor currents were observed, suggesting that endogenous APPsα may be released during the HFS to contribute to LTP. However this point requires further study, as the effect of α-secretase inhibition on tetanic NMDA receptor currents was small, and other studies have reported both a decrease (Furukawa and Mattson, 1998) and an increase in single NMDA receptor currents (Moreno et al., 2015) in response to exogenous APPsα delivery. More recently, we generated conditional APP/APLP2 double KO (termed NexCre cDKO) mice that lack APP expression and thus APPsα secretion in excitatory forebrain neurons on a global APLP2-KO background (Hick et al., 2015). Consistent with findings by Taylor et al. (2008), this led to impairments in hippocampal LTP that were also reflected in impairments in hippocampus-dependent learning and memory tasks, including deficits in Morris water maze and radial maze performance (Hick et al., 2015). Interestingly, we demonstrated that acute treatment of brain slices with nanomolar amounts of recombinant APPsα, but not APPsβ, rescued the impairment of LTP (Hick et al., 2015). These findings indicate a crucial ability specifically for APPsα to support synaptic plasticity of mature hippocampal synapses on a rapid time scale. Similar differential effects of APPsα vs. APPsβ have been reported in assays of neuroprotection, with APPsβ being far less effective (reviewed by Chasseigneaux and Allinquant, 2012). Thus, the crucial functional domain of APPsα may reside within terminal APPsα-CT16 residues, and/or their presence alters the conformation of APPsα in a critical way. Indeed, there is evidence from recent structural analysis by small angle X-ray diffraction studies that the threedimensional structure of APPsα is very different from APPsβ (Peters-Libeu et al., 2015). This study further suggested that the N-terminal E1 domains folds back towards the C-terminal juxtamembrane domain in APPsβ (Peters-Libeu et al., 2015). Thus, epitopes that are accessible in APPsα or when provided as peptides may become masked in APPsβ. This may have important functional implications as distinct 3D structures may enable or prevent binding to different receptors. Although the receptor(s) mediating the acute effects of APPsα on synaptic plasticity are currently unknown, they are not the endogenous APP and APLP2 that are both lacking in NexCre cDKO mice (Hick et al., 2015).

APPsα also appears to play an important role in processes of natural aging. Not only is memory performance correlated with APPsα levels (Anderson et al., 1999), but aging-related deficits in both LTP and cognitive behavior can be rescued by exogenous APPsα (Moreno et al., 2015; Xiong et al., 2016).

### GENE EXPRESSION AND PROTEIN SYNTHESIS

Full expression of LTP requires gene expression and de novo protein synthesis, and this raises the question of whether APPsα itself directly regulates protein synthesis and the processes of translation and transcription that underlie it. Barger and Mattson (1996a) suggested that APPsα could regulate transcription through activation of the transcription factor NF-kappa B (NFκB), and extensive gene expression responses to relatively brief delivery of exogenous APPsα have been reported (Stein et al., 2004; Ryan et al., 2013). Gene expression responses occurred in as little as 15 min and these slowly changed from predominantly upregulation to predominantly downregulation during 24 h of APPsα treatment (Ryan et al., 2013). Upregulation occurred for immediate early gene transcription factors and for NFκB- and CREB-regulated genes, as well as regulation of late response genes known to be involved in cell survival, inflammatory responses, apoptosis and neurogenesis. These findings were further corroborated by Aydin et al. (2011).

Although APPsα can regulate coupled transcriptional and translational processes, it can also directly regulate protein synthesis. Claasen et al. (2009) found, using rat hippocampal synaptoneurosomes that are not transcriptionally competent, that APPsα initiated de novo protein synthesis in the dendritic compartment that was sensitive to the translation inhibitor cycloheximide. This effect was: (1) dose-dependent with higher concentrations failing to affect baseline protein synthesis; (2) age-dependent with a much reduced effect in tissue from aged rats; and (3) abolished by a PKG inhibitor and partially blocked by inhibitors of calcium/calmodulin protein kinase II (CaMKII), and mitogen-activated protein kinases (MAPKs). It appears likely therefore that at least part of the LTP facilitation by APPsα is through regulated transcriptional and translational processes, but this hypothesis has yet to be directly tested.

# MEMORY

Intracerebral administration of antibodies against the APPsα region of APP is able to cause learning and memory impairments in rat inhibitory avoidance (Doyle et al., 1990; Huber et al., 1993) as well as chick inhibitory avoidance (Mileusnic et al., 2000) tasks. Similarly, inhibition of α-secretase impaired rat spatial watermaze memory (Taylor et al., 2008) while APP knock-out impaired mouse spatial learning (Ring et al., 2007). Although these treatments are not specific manipulations of APPsα, it is notable that memory deficits could be prevented in a number of experiments by acute administration of either full-length APPsα (Taylor et al., 2008) or APPsα fragments (Mileusnic et al., 2000), or by genetic over-expression of full-length APPsα (Ring et al., 2007). APPsα and its fragments have also been used to rescue memory under other conditions of impairment, such as caused by the muscarinic receptor antagonist scopolamine (Meziane et al., 1998), Aβ (Mileusnic et al., 2004), head injury (Corrigan et al., 2012), and aging (Xiong et al., 2016). Moreover, viral vector mediated over-expression of APPsα rescued memory in a mouse model of AD (Fol et al., 2016).

There is also evidence that normal memory can be enhanced by exogenous APPsα or peptide fragments. Full-length APPsα enhanced go-no-go discrimination and operant lever pressing in rats (Meziane et al., 1998) while a 17-mer fragment (derived from the heparin binding domain located in the conserved E2 domain) facilitated spatial memory in the watermaze task for aged but non-memory impaired rats (Roch et al., 1994). A 5-mer peptide internal to that fragment converted short-term avoidance memory to long-term memory in chicks (Mileusnic et al., 2004). These findings need to be treated with caution, however, because transgenic over-expression of APPsα from gestation has been shown to lead to the development of autism-like markers such as hypoactivity and impaired sociability (Bailey et al., 2013), as well as aberrant T-lymphocyte development and function (Bailey et al., 2011).

# APPsα AS A THERAPEUTIC TARGET

The neurotrophic, neuroprotective, neurogenic, synaptogenic as well as neuronal plasticity and memory enhancing properties establish APPsα as an attractive therapeutic target during the early stages of AD and possibly also later. In this regard it should be kept in mind that due to the highly plastic nature of synapses, their dysfunction and loss are reversible processes. Thus, synaptic repair stimulated by trophic APPsα may ameliorate pathophysiology and improve clinical outcome as a complementary approach to eliminating toxic factors.

APPsα levels may either be enhanced by shifting APP processing towards the non-amyloidogenic pathway or by direct delivery/expression of exogenous APPsα (**Figure 1**). Inhibiting amyloidogenic APP processing, e.g., by targeting the Aβ-generating secretases has been a major focus of AD research over last two decades (e.g., Yan and Vassar, 2014; Geldenhuys and Darvesh, 2015) and several advanced BACE inhibitors are in phase 3 clinical trials (Cumming et al., 2012). However, using systematic proteomic approaches, it has become clear that all secretases have numerous substrates besides APP (Saftig and Lichtenthaler, 2015; Kuhn et al., 2016). Pharmacological inhibition of secretases may therefore have serious drawbacks due to mechanismbased side effects on other targets that are important for normal brain physiology. These concerns were further fueled by recent findings demonstrating that BACE inhibition upregulates non-canonical APP processing and production of Aη fragments that impair neuronal activity and LTP (Willem et al., 2015).

With respect to the alternative approach, enhancement of non-amyloidogenic APP processing may be achieved by upregulating α-secretase expression at the transcriptional level or by modulating its subcellular trafficking or activity (for review see Endres and Fahrenholz, 2012; Postina, 2012; Saftig and Lichtenthaler, 2015; Habib et al., 2016).

# Transcriptional Activation of ADAM10

The human ADAM10 promoter contains two retinoic acid response elements and ADAM expression can be upregulated at the transcriptional level by the vitamin A analog acitretin in cells and in transgenic AD mouse models leading to increased APPsα and reduced Aβ production (Tippmann et al., 2009). In a small clinical trial with AD patients, acitretin, that is already approved to treat psoriasis, was well tolerated and caused a significant increase in APPsα levels that was detectable in CSF samples of treated patients (Endres et al., 2014). Long-term studies with larger patient cohorts are planned. Melatonin, which decreases during aging and in AD patients, has been shown to efficiently decrease Aβ levels when administered at early stages of pathogenesis in Tg2576 AD mice (Matsubara et al., 2003). Recently, detailed in vitro studies indicated that the underlying mechanism involves plasma membrane-located melatonin receptor activation, and ERK1/2 phosphorylation leading to increased APPsα levels via transcriptional activation of ADAM10 and ADAM17 (Shukla et al., 2015). Moreover, and in line with data from Moreno et al. (2015) and Xiong et al. (2016), melatonin partially restored APPsα levels and spatial learning in aged mice (Mukda et al., 2016).

# Post-Transcriptional Activation of α-Secretase

Although the precise mechanisms of activation are not fully understood, it is clear that α-secretase activity, as judged by enhanced APPsα levels, can be directly or indirectly upregulated via ion channels, G-protein coupled receptors (GPCRs) and receptor tyrosine kinases. In particular, receptor-activated protein kinase C, MAP kinases, PI3 kinase and Ca2<sup>+</sup> signaling have been shown to contribute to α-secretase activation. In many cases, however, it has not been directly studied which enzymes mediate increased APPsα production. In these cases processing may involve ADAM10 and/or ADAM17 and possibly further metalloproteases that have been shown to have APP cleaving activity in vitro (Saftig and Lichtenthaler, 2015). As a detailed description of these various pharmacological approaches is beyond the scope of this review, the reader is referred to a series of recent reviews (see Postina, 2012; Saftig and Lichtenthaler, 2015; Habib et al., 2016; Spilman et al., 2016).

Upregulation of α-secretase activity was reported for etazolate, an allosteric activator of GABA<sup>A</sup> receptors, which increased APPsα in rat cortical neurons and guinea pig brain (Marcade et al., 2008), improved memory in aged rats (Drott et al., 2010) and proved protective against traumatic brain injury (Siopi et al., 2013). The neuropeptide pituitary adenylate cyclase-activating polypeptide (PACAP) potently increased APPsα levels, an effect that was abrogated by an antagonist of the GPCR PAC1, by a hydroxamate-based ADAM inhibitor and by inhibitors of MAP kinases and PI3 kinases (Kojro et al., 2006). In vivo, APPsα production in the brain was stimulated by long-term intranasal PACAP application. The effects of PACAP application were not limited to increased APPsα levels but were instead pleiotropic, including upregulation of the PAC1 receptor, BDNF and of the anti-apoptotic Bcl-2 protein (Rat et al., 2011). While these in vivo effects, including improved object recognition in transgenic AD model mice (Rat et al., 2011), appear favorable for treatment, Gardoni et al reported that PACAP treatment of primary hippocampal neurons led to postsynaptic ADAM10 accumulation and N-cadherindependent reductions in spine head volume and reduced postsynaptic GluR1 expression (Gardoni et al., 2012). Thus a more detailed in vivo characterization appears warranted.

Activation of serotonin type 4 receptors (5-HT4Rs), another class of neuronally expressed GPCR, promotes the activity of ADAM10 and APPsα generation. The 5-HT4R was shown to directly interact with the mature form of ADAM10 and agonist stimulation of the receptor accelerated ADAM10 activity by cAMP/Epac (cAMP-responsive Rap1 guanine nucleotide exchange factor) signaling (Cochet et al., 2013). Tesseur et al. (2013) reported that chronic 5-HT4 receptor activation lowered Aβ production in transgenic hAPP/PS1 AD model mice but the authors found no evidence for a direct activation of ADAM10. The underlying mechanism appears more complex and may involve decreased APP and BACE-1 expression and elevated astroglial and microglial responses. More recently, donecopride, a promising synthetic multitargeted ligand that functions both as a partial agonist of 5-HT4R and as an acetylcholinesterase inhibitor, has been developed and shown to have memory enhancing ability (Lecoutey et al., 2014).

# Direct Expression of APPsα in the CNS

Although α-secretase-targeting pharmacological strategies are potentially promising, there remains the concern regarding lack of specificity (see for example Gardoni et al., 2012). Acitretin may induce other genes with retinoid response elements in their promoters and upregulation of α-secretase activity (ADAM10, 17 or others) will likely lead to the processing of many additional substrates. In this regard, Kuhn et al. (2016) recently demonstrated that ADAM10 has over 40 neuronal substrates including some involved in tumorigenesis. Thus, it is still unclear whether increasing α-secretase activity in neural tissue will ultimately be of therapeutic benefit for patients. An approach to circumvent these problems is the direct delivery of APPsα into the brain.

While previous studies demonstrated the neuroprotective properties of APPsα against acute forms of brain injury (Van Den Heuvel et al., 1999; Thornton et al., 2006; Corrigan et al., 2012, 2014; Plummer et al., 2016) the situation is quite different in neurodegenerative diseases such as AD, characterized by chronic production and accumulation of neurotoxic molecules including Aβ. Another challenge is the need for sustained expression of neurotrophic/neuroprotective factors. This calls for a gene therapy approach. During recent years gene therapy approaches to neurological disorders including AD have been explored in preclinal studies and also entered phase I/II clinical trials (Tuszynski et al., 2015; Choudhury et al., 2016b; Fol et al., 2016; Hocquemiller et al., 2016). For the CNS, adeno-associated virus (AAV) vector systems have been most commonly used due to their safety, non-pathogenic nature, the ability to transduce dividing and non-dividing cells, particularly neurons in vivo, the wide volumetric distribution of vector particles in tissue and the ability to mediate long-term gene expression in vivo (Choudhury et al., 2016a; Hocquemiller et al., 2016).

Recently, we employed AAV9-mediated gene transfer of APPsα into the brain to explore its potential to ameliorate or rescue structural, electrophysiological and behavioral deficits of transgenic APP/PS1 AD model mice. A single bilateral injection of AAV-APPsα particles was sufficient to mediate long-lasting APPsα expression over 5 months that was well tolerated without apparent side effects. Interestingly, sustained APPsα overexpression in aged APP/PS1 mice with already preexisting pathology and amyloidosis restored LTP, ameliorated spine density deficits and also rescued spatial reference memory assessed by the Morris water maze. Moreover, we demonstrated a significant reduction of soluble Aβ species and plaque load. In addition, APPsα treatment induced the recruitment of microglia with a ramified morphology into the vicinity of plaques and upregulated IDE and TREM2 expression suggesting enhanced plaque clearance (Fol et al., 2016). These data further corroborate the therapeutic potential of APPsα for AD that raises hope to translate these findings into clinical application.

To this end further experimental studies including different routes of viral vector application, dose optimization and studies in lager animals are needed. Several routes of vector administration to the CNS have been developed: (i) direct injection into the brain parenchyma; (ii) CSF-based delivery using ICV, cisternal or lumbar intrathecal (IT) administration; and (iii) intravascular (e.g., intravenous) administration. Intracranial injection has been explored not only for diseases with anatomically restricted pathology such as Parkinson's disease (reviewed in Choudhury et al., 2016b; Hocquemiller et al., 2016), but also for neuropathic lysosomal storage diseases (LSD) that affect large brain regions. For LSDs, multiple intraparenchymal injections were used in phase I/II clinical trials that showed the safety of the approach and also lead to encouraging clinical outcome (Leone et al., 2012; Tardieu et al., 2014). Vector delivery via the CSF', in particular intracisternal and IT is a less invasive alternative strategy that is particularly promising for the delivery of secreted proteins such as growth factors and lysosomal proteins, and has been successfully used to express Apolipoprotein E in AD model mice (Hudry et al., 2013). Systemic, intravascular administration is the least invasive route and has the potential to enable wide spread vector distribution as every cell in the brain being a maximum of 40 microns from the microvasculature (Wong et al., 2013). In this regard encouraging progress has been made, as serotype AAV9 and AAVrh.10 have been shown to cross to some extent the blood brain barrier (BBB; reviewed in Hocquemiller et al., 2016; Saraiva et al., 2016), apparently by active transcytosis through endothelial cells (Merkel et al., 2017). The development of modified AAV vectors with re-engineered capsids should improve this further (Choudhury et al., 2016a; Deverman et al., 2016; Jackson et al., 2016). One of the main challenges for AD gene therapy is the widespread pathology that affects several anatomic regions involved in learning and memory. Thus, protocols that either target regions affected early during disease and/or widespread gene delivery to several anatomical regions are required.

#### REFERENCES


A non-invasive alternative to viral vector mediated gene transfer are formulations of recombinant APPsα protein, sub-domains or active peptides that enable transport across the BBB. This includes intranasal delivery that has been successfully used in preclinical models of CNS diseases (Lochhead and Thorne, 2012). Examples for AD are the intranasal delivery of insulin (Mao et al., 2016) or PACAP (Rat et al., 2011) to enhance non-amyloidogenic APP processing in transgenic mouse models. Liposomes and nano-particle based approaches are emerging as further options (Kreuter, 2014; Khalin et al., 2016). Finally, transient opening of the BBB by transcranial focused or scanning ultrasound in combination with microbubbles might be used to further enhance delivery of viral vectors, proteins such as APPsα (or active fragments) and nano-particles (Thévenot et al., 2012; Leinenga et al., 2016).

Collectively, these various approaches all appear to merit further investigation. However it needs to be kept in mind that many challenges lie ahead for translating these approaches to the human brain, especially given its size and thus the widespread volume of brain tissue that needs to be treated. Moreover, the preclinical animal models being currently used do not fully recapitulate the human disease features, and thus successes in animal models need to be treated with caution. Nonetheless, despite these challenges the neuroprotective and synaptic repair inducing properties of APPsα make it a worthy target for future research aiming to treat AD, as well holding other neurological disorders.

#### AUTHOR CONTRIBUTIONS

BGM, MR, WCA and UCM co-wrote this review. MR designed the figures.

#### ACKNOWLEDGMENTS

The authors acknowledge support from the Deutsche Forschungsgemeinschaft (FOR 1332) to UCM, the Alzheimer Research Price of the Breuer Foundation to UCM, the Health Research Council of New Zealand to BGM and WCA, the Neurological Foundation of New Zealand to WCA and the University of Otago Research Committee to BGM.

amyloid precursor protein in aged rats: correlation with spatial memory deficits. Neuroscience 93, 1409–1420. doi: 10.1016/s0306-4522(99) 00244-4


differentiation of adult hippocampal neural progenitor cells. Hippocampus 22, 1517–1527. doi: 10.1002/hipo.20988


through ADAM10-N-cadherin signaling pathway. J. Cell Sci. 125, 1401–1406. doi: 10.1242/jcs.097576


NMDA receptors. J. Neurosci. 31, 6627–6638. doi: 10.1523/JNEUROSCI.0203- 11.2011


and tangles: intracellular Aβ and synaptic dysfunction. Neuron 39, 409–421. doi: 10.1016/S0896-6273(03)00434-3


knockouts reveals reduced spine density in APP-KO mice that is prevented by APPsα expression. Acta Neuropathol. Commun. 2:36. doi: 10.1186/2051-59 60-2-36


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Mockett, Richter, Abraham and Müller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.