Download The

The Role of DHHC5-Mediated Palmitoylation of δ-Catenin in Cadherin Stability and Synapse Plasticity

Gian Stefano Brigidi

B.Sc., McGill University, 2007

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF

THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES

(Neuroscience)

THE UNIVERSITY OF BRITISH COLUMBIA

(Vancouver)

December 2014

Abstract

Synapses of the Central Nervous System are specialized junctions of cell-cell contact that transmit signals from one neuron to another in a rapid and efficient manner. Synapses are highly plastic structures that can be continually modified in response to fluctuations in neuronal activity. Changes in the number, size, and protein composition of synapses have been observed following alterations in neuronal activity in vitro and following the learning of specific tasks in vivo. Thus, elucidating the molecular mechanisms underlying activity-mediated trafficking of proteins to and from synaptic compartments is essential for our understanding of brain function. Previous work has demonstrated a requirement for the cadherin-adhesion complex in activity-induced enhancements in synapse strength, however the molecular mechanisms that translate synaptic activation into enhanced cadherin-based adhesion and synapse strengthening remain unknown. This dissertation discusses work that unravels how synaptic activity coordinates the enhancement of cadherin surface stabilization, enlargement of dendritic spines, and increased surface insertion of AMPA receptors. This work demonstrates that increased synaptic activity enhances the palmitoylation of a brain-specific component of the cadherin-adhesion complex, δ-catenin, which in turn causes δ-catenin to traffic toward the synaptic membrane in spines where it associates with and stabilizes surface N-cadherin. This results in enhancements in synapse structure and efficacy, and is correlated with the acquisition of contextual fear memories. Furthermore, we show that palmitoylation of δ-catenin is mediated by the palmitoyl acyl- transferase DHHC5, and that DHHC5 drives activity-induced increases in surface AMPA receptor levels through the palmitoylation of δ-catenin. Finally, we demonstrate that the activity- induced palmitoylation of δ-catenin by DHHC5 is accomplished through the rapid trafficking of DHHC5 out of the synapse and into the dendritic shaft where it can associate with and palmitoylate δ-catenin, resulting in δ-catenin’s synaptic recruitment. This work provides new insights into the cellular and molecular mechanisms that underlie activity-induced synapse plasticity.

Preface

The content in chapter 1.5, entitled “Cadherin-catenin Adhesion Complexes at CNS synapses” is a review that has been published as:

Brigidi, G.S., and S.X. Bamji. 2011. Cadherin-catenin adhesion complexes at the synapse. Current Opinion in Neurobiology 21:208-214.

The manuscript was written and figures prepared by me under the supervision of the senior author, Dr. Shernaz X. Bamji.

The work in chapter 2, entitled “Palmitoylation of δ-catenin by DHHC5 Mediates Activity-

Induced Synapse Plasticity” has been published as:

Brigidi, G.S., Y. Sun, D. Beccano-Kelly, K. Pitman, M. Mobasser, S.L. Borgland, A.J. Milnerwood, and S.X. Bamji. 2014. Palmitoylation of delta-catenin by DHHC5 mediates activity-induced synapse plasticity. Nature Neuroscience 17:522-532.

All experiments and data analysis for this manuscript was done by myself with following exceptions: Immunocytochemical experiments in Figure 2.12 were done by co-authors Yu Sun and Mahsan Mobasser. Electrophysiology experiments in Figure 2.8 were designed by co- authors and collaborators Dr. Austen Milnerwood and Dr. Stephanie Borgland, and done by co- authors Dayne Beccano-Kelly and Kimberley Pitman. The manuscript was written and figures prepared by me under the supervision of the senior author, Dr. Shernaz X. Bamji.

The work in chapter 3, entitled “Activity-Regulated Trafficking of the Palmitoyl Acyl-

Transferase DHHC5” is currently under revision as:

iii

Brigidi G.S., Santyr B, Shimell J, Jovellar B, and S.X. Bamji (2014). Activity-Regulated Trafficking of the Palmitoyl Acyl-Transferase DHHC5. Manuscript under revision.

All experiments and data analysis for this manuscript were done by myself with following exceptions: Immunocytochemical experiments and data analysis in Figures 3.1-3.5 were done with assistance from co-authors Brendan Santyr, Jordan Shimell, and Blair Jovellar. The experiment in Figure 3.3D was done by Brendan Santyr. The manuscript was written and figures prepared by me under the supervision of the senior author, Dr. Shernaz X. Bamji.

Table of Contents

Abstract…...... ii Preface…… ...... iii Table of Contents ...... v List of Tables ...... viii List of Figures ...... ix List of Abbreviations ...... xii Acknowledgements ...... xiv Chapter 1: Introduction ...... 1 1.1 The Hippocampus as a Model System ...... 1 1.2 Excitatory Synapses of the Central Nervous System ...... 3 1.2.1 Composition of the Presynaptic Compartment ...... 4 1.2.2 Composition of the Postsynaptic Compartment ...... 8 1.2.2.1 AMPA Receptors ...... 10 1.2.2.2 NMDA Receptors ...... 11 1.2.3 Dendritic Spines ...... 14 1.2.3.1 Spine Morphology ...... 15 1.2.3.2 Structure–Function Relationship ...... 18 1.2.3.3 Spine Abnormalities and Neuropsychiatric Disease ...... 20 1.3 Inhibitory CNS Synapses ...... 22 1.4 Synapse Plasticity ...... 25 1.4.1 Long-Term Potentiation ...... 26 1.4.1.1 Functional LTP ...... 27 1.4.1.2 Structural LTP ...... 28 1.4.2 Long-Term Depression ...... 31 1.4.2.1 Functional LTD ...... 31 1.4.2.2 Structural LTD ...... 32 1.4.3 LTP and LTD in Learning and Memory ...... 33 1.5 Cadherin-Catenin Adhesion Complexes at CNS Synapses ...... 35 1.5.1 Cadherins in Synapse Formation ...... 36 1.5.1.1 Cadherin Functions at Presynaptic Compartments ...... 37 1.5.1.2 Cadherin Functions at Postsnaptic Compartments ...... 39 1.5.2 Regulation of Cadherin-Based Adhesion at the Synapse ...... 43 1.6 δ-Catenin at the Synapse ...... 46 1.6.1 Postsynaptic Signaling Pathways of δ-Catenin ...... 46 1.6.2 Aberrant δ-Catenin Function and Cognitive Performance ...... 49 1.7 Neuronal Protein Palmitoylation ...... 50 1.7.1 Palmitoyl Acyl-Transferases: The DHHC Family of Proteins ...... 53 1.7.2 DHHC Proteins in Cognitive Function and Neurological Disease ...... 57 v

1.7.3 Dynamic Palmitoylation of Synaptic Substrates ...... 59 1.8 Rationale and Hypothesis ...... 61 Chapter 2: Palmitoylation of δ-Catenin by DHHC5 Mediates Activity-Induced Synapse Plasticity ...... 63 2.1 Introduction ...... 63 2.2 Materials and Methods ...... 65 2.2.1 Antibodies and cDNA Constructs ...... 65 2.2.2 Contextual Fear Conditoning ...... 68 2.2.3 Cell Cultures ...... 68 2.2.4 Neuronal Activation ...... 69 2.2.5 Immunoblot Assay ...... 69 2.2.6 Palmitoylation Assay ...... 70 2.2.7 Immunocytochemistry ...... 70 2.2.8 Confocal Imaging ...... 71 2.2.9 Fluorescence Recovery After Photobleaching (FRAP) ...... 71 2.2.10 Image Analysis and Quantification ...... 72 2.2.11 Electrophysiology ...... 73 2.2.12 Statistical Analysis ...... 73 2.3 Results ...... 74 2.3.1 Activity-Dependent Palmitoylation of δ-Catenin in Neurons ...... 74 2.3.2 Palmitoylation of δ-Catenin Regulates its Binding to N-Cadherin ...... 78 2.3.3 Palmitoylated and Cadherin-Binding Residues of δ-Catenin ...... 82 2.3.4 Palmitoylated δ-Catenin Stabilizes N-Cadherin at Synapses ...... 85 2.3.5 Palmitoylated δ-Catenin Regulates Spine Remodeling ...... 89 2.3.6 Palmitoylated δ-Catenin Mediates Changes in Synapse Efficacy ...... 92 2.3.7 δ-Catenin Palmitoylation Increases after Acqusition of Contextual Fear Memory .. 98 2.3.8 DHHC5 is Required for Palmitoylation of δ-Catenin ...... 100 2.4 Discussion ...... 107 Chapter 3: Activity-Mediated Trafficking of the Palmitoyl Acyl-Transferase DHHC5 .. 112 3.1 Introduction ...... 112 3.2 Materials and Methods ...... 114 3.2.1 Antibodies and cDNA Constructs ...... 114 3.2.2 Cell Cultures ...... 115 3.2.3 Neuronal Activation ...... 115 3.2.4 Acyl-Biotin Exchange (ABE) Assay ...... 116 3.2.5 Biotinylation Assay ...... 116 3.2.6 Immunoprecipitation (IP) ...... 116 3.2.7 Western Blot Analysis ...... 117 3.2.8 Immunocytochemistry ...... 117 3.2.9 Confocal Imaging ...... 118 3.2.10 Image Analysis and Quantification ...... 118 3.2.11 Statistical Analysis ...... 119 vi

3.3 Results ...... 120 3.3.1 Activity-Induced Mobilization of DHHC5 ...... 120 3.3.2 Activity-Induced Trafficking of DHHC5 ...... 123 3.3.3 DHHC5 Binds δ-Catenin and Mediates its Synaptic Recruitment ...... 126 3.3.4 Activity-Induced Endocytosis of DHHC5 and Trafficking on Recycling Endosomes ...... 128 3.4 Discussion ...... 133 Chapter 4: Conclusion ...... 137 4.1 Formation of a Stable Synaptic Adherens Junction ...... 138 4.2 Linkage of the Cadherin-Adhesion Complex with the AMPAR-Scaffold Complex by Palmitoylated δ-Catenin ...... 141 4.3 Transsynaptic Strengthening and Synaptic Tagging by Palmitoylated δ-Catenin ...... 142 4.4 Activity-Regulated Trafficking of DHHC Proteins ...... 144 4.5 Significance and Limitations ...... 145 Bibliography ...... 147

vii

List of Tables

Table 1.1. Selected CAZ Proteins Involved in Active Zone Formation and Synaptic Vesicle Cycling...... 6

viii

List of Figures

Figure 1.1: The Trisynaptic Loop of the Mouse Hippocampus ...... 2 Figure 1.2: Schematic Depiction of an Excitatory CNS Synapse ...... 4 Figure 1.3: Schematic Organization of Proteins at The PSD ...... 9 Figure 1.4: Schematic Depiction of Basal Synaptic Transmission at a Hippocampal Synapse ... 14 Figure 1.5: Morphological Classes of Dendritic Spines and a Filopodium ...... 16 Figure 1.6: Spine Structure and Expression of Functional AMPA Receptors ...... 19 Figure 1.7: Dendritic Spine Numbers through Human Development among Normal Individuals and Patients with Neuropsychiatric Disorders ...... 21

Figure 1.8: Schematic of a GABAA Receptor with associated Postsynaptic Cytosolic Proteins..24 Figure 1.9: Hippocampal Long-Term Potentiation ...... 26 Figure 1.10: LTP Induction Results in Morphological Enhancement of Spines ...... 29 Figure 1.11: LTD Induction Causes Spine Shrinakge and Elimination ...... 32 Figure 1.12: Schematic Depiction of the Classic Cadherin/Catenin Adhesion Complex at Excitatory Synapses ...... 42 Figure 1.13: Schematic Depiction of Mechanisms Regulating Cadherin-Based Adhesion at the Synapse ...... 45 Figure 1.14: Schematic Depiction of the known δ-Catenin Postsynaptic Signaling Pathways ... 48 Figure 1.15: Schematic Depiction of Palmitoylation and Common Lipid Modifications ...... 52 Figure 1.16: The Family of Mammalian DHHC Proteins and Representative Enzyme-Substrate Pairs ...... 56

Figure 2.1: δ-Catenin Palmitoylation and its Association with Synaptic N-Cadherin are Increased after Activity ...... 77 Figure 2.2: Homeostatic Enhancement of Synaptic Strength Enhances the Palmitoylation of δ- Catenin ...... 78 Figure 2.3: Activity Enhances the Clustering of δ-Catenin with Surface N-Cadherin at the Postsynaptic Membrane ...... 81

Figure 2.4: Palmitoylation of δ-Catenin Occurs at C960 and C961 and Requires K581 for Binding to N-Cadherin ...... 84 Figure 2.5: Expression of δ-Catenin Constructs does not Impact Basal N-Cadherin Levels within Spine Heads, nor does Expression of K581M and C960-1S Impact N-Cadherin Stability ...... 87 Figure 2.6: δ-Catenin Palmitoylation is Required for Activity-Induced Stabilization of N- Cadherin within Dendritic Spine Heads ...... 88 Figure 2.7: δ-Catenin Palmitoylation is Required for Activity-Induced Spine Remodeling ...... 91 Figure 2.8: δ-Catenin Palmitoylation is Required for Activity-Induced AMPA Receptor Insertion and Changes in mEPSCs ...... 95 Figure 2.9: Activity-Induced Insertion of AMPA Receptors Requires Cadherin-Binding by Palmitoylated δ-Catenin ...... 97 Figure 2.10: Context-Dependent Fear Conditioning Increases δ-Catenin Palmitoylation and N- Cadherin Associations in the Hippocampus ...... 99 Figure 2.11: DHHC5 and DHHC20 Palmitoylate δ-Catenin, but Activity-Induced Recruitment of δ-Catenin to N-Cadherin is Mediated by DHHC5 ...... 103 Figure 2.12: DHHC5 and DHHC20 Enhance the Recruitment of δ-catenin to N-cadherin under Basal Conditions ...... 105 Figure 2.13: DHHC5 is Required for Activity-Induced Palmitoylation of δ-Catenin and Surface AMPAR Insertion ...... 106

Figure 3.1: Changes in DHHC5 Function and Localization Following Increased Neuronal Activity ...... 122 Figure 3.2: Activity Enhances DHHC5 Trafficking from Spines ...... 125 Figure 3.3: Activity Enhances DHHC5 Association with δ-Catenin ...... 128 Figure 3.4: Activity-Induced Endocytic Cycling of DHHC5 ...... 129 Figure 3.5: DHHC5 is Trafficked on Recycling Endosomes (REs) ...... 132

Figure 4.1: Model of the Palmitoylation of δ-Catenin by DHHC5 Driving Activity-Induced Enhancements of Synapse Adhesion, Structure, and Efficacy ...... 138

List of Abbreviations

ABE – Acyl-biotin exchange ABP – Actin binding protein AD – Alzheimer’s Disease AKAP150 – A-kinase anchoring protein 150 AMPAR – α-amino-3-hydroxy-5-methyl-4-isoxazoleproprionic acid type glutamate receptor APT1 – Acyl-protein thioesterase 1 AP5 – D(–)-2-amino-5-phosphonovaleric acid ARCAD – Arcadlin ASD – Autism spectrum disorder Ca2+ – Calcium CAM – cell adhesion molecule CaMKII – calcium/calmodulin-dependent protein kinase II CAZ – Cytoskeletal active zone CA1 and CA3 – Cornu Ammonis Areas 1 and 3, respectively CRD – Cysteine rich domain CNV – Copy number variation CNS – Central Nervous System CoA – Palmitoyl Coenzyme A CR – Conditioned Response DG – Dentate Gyrus DHHC – Aspartate-Histidine-Histidine-Cysteine Motif DIV – Days-in-vitro EC – Entorhinal cortex EPSC and mEPSC – Excitatory postsynaptic current, and miniature EPSC, respectively EPSP – Excitatory postsynaptic potential FRAP – Fluorescence Recovery after Photobleaching GABA and GABAR – γ-aminobutyricacid and GABA receptor, respectively GAD – L-Glutamic acid decarboxylase GEF – Guanine nucleotide exchange factor GK – Guanylate Kinase GKAP – guanylate kinase-associated protein HD – Huntington’s Disease HFS – High frequency stimulation HIP14 – Huntingtin-interacting-protein 14 Hk – Hakai Ubiquitin Ligase IntDen – Integrated Density IRSp53 – insulin receptor substrate JMD – Juxtamembrane domain KCh – K+ channel MAGUK – Membrane-associated guanylate kinase protein LAP – Leucine-rich-repeat and PDZ domain LAR – Leukocyte antigen related LFS – Low frequency stimulation LTD and cLTD – Long-term depression and chemical LTD, respectively LTP and cLTP – Long-term potentiation and chemical LTP, respectively xii

mCh – mCherry mGluR – metabotropic glutamate receptor Mg2+ – Magnesium MPR – Membrane proximal region

NH2OH – Hydroxylamine NL-1 – Neuroligin-1 nNOS – neuronal nitric oxide synthase NMDAR – N-methyl-D-aspartate type glutamate receptor NSF – N-ethylmaliemide sensitive factor PAT – Palmitoyl acyl-transferase PKA – Protein kinase A PKC – Protein kinase C PDZ – PSD-95/Discs large/Zona-occludens PPT – Palmitoyl-protein thioesterase PP2B – protein phosphatase 2B PSD – Postsynaptic density PSD-95 – PSD protein 95 PS1 – Presenilin-1 PTV – Piccolo-Bassoon Transport Vesicle p120ctn – p120-catenin RE – Recycling Endosome RFP – Red fluorescent protein ROI – Region-of-interest RRP – Readily releasable pool RP – Reserve pool RTK – receptor tyrosine kinases SCRIB – Scribble SDS-PAGE – Sodium-dodecyl sulfate and polyacrylamide gel electrophoresis SEP – Super ecliptic pHluorin shRNA – Short hairpin RNA SH3 – Src family homology domain 3 SPAR – spine-associated RapGAP STED – Stimulated Emission Depletion imaging STV – Synaptic transport vesicle SV – Synaptic Vesicle SynGAP – Synaptic GTPase-activating protein SZ – Schizophrenia TfR – Transferrin Receptor TMD – Transmembrane domain TTX – Tetrodotoxin VGAT – Vesicular GABA Transporter VGluT – Vesicular Glutamate Transporter YFP – Yellow fluorescent protein 2-BP – 2-bromopalmitate

xiii

Acknowledgements

I would like to thank, first and foremost, my Ph.D. supervisor, Dr. Shernaz Bamji. She has been a kind and encouraging mentor, and her guidance and clarity were truly inspiring and essential for my work. She has been supportive in science and in life outside the lab, and I am truly grateful to have had such an excellent mentor whom I can always look up to and call a friend. I would also like to thank the members of my research advisory committee, Dr. Lynn

Raymond, Dr. Ann Marie Craig, and Dr. Tim O’Connor, for all their time, thoughtfulness, and support throughout my thesis.

I would like to thank all the members of the Bamji Lab, past and present, who provided a friendly and exciting atmosphere in the lab, and most especially for tolerating the inevitable ups and downs that I experienced along the way. I especially would like to thank my current and past colleagues Fergil Mills, Eugenia Petoukhov, Yu Sun, and Andrea Globa who will always be my friends in and out of the lab. I would like to also acknowledge my colleagues in neuroscience at

UBC, past and present, for their helpful scientific discussion, and for all the good times that I could not have possibly done without.

I especially want to thank my parents, for everything they have provided and sacrificed for me. It’s difficult to fully express how essential their advice, support, and love have been all these years. I simply would not be at this stage in life without them.

I cannot thank enough my beautiful girlfriend Khatereh, for the endless calm, love, and motivation she has always provided me. Her strength and trust have done more for me than she could ever possibly know. And through her, I want to thank my west coast family Mitra Saberi and Ghazaleh Aminoltejari, who truly made me feel like family, and at home in Vancouver.

xiv

Chapter 1: Introduction

1.1 The Hippocampus as a Model System

The hippocampus is a distinct anatomical structure within the medial temporal lobe of the human brain that has been favored and highly studied by neurobiologists for decades because of its importance in the formation of episodic and declarative memories. Two established lines of evidence that support the necessity of the hippocampus in learning and memory are that lesions to the human hippocampus selectively impair the formation of new memories1, and activity- induced synapse plasticity is a prominent feature of the synaptic connections among hippocampal neurons2. Therefore, detailed examinations of the circuitry and synapses that make up the hippocampus can potentially reveal the cellular and molecular mechanisms that underlie learning and memory.

The simple and invariant circuitry of the rodent hippocampus, which is typically depicted as a trisynaptic loop that receives its primary input from subpopulations of cells within the entorhinal cortex (EC; Fig. 1.1), has made it an attractive and highly utilized model system for studying the neurobiology of learning. There are three major excitatory neuronal subtypes within the hippocampus, including granule cells of the dentate gyrus (DG) subfield that send their axons to the pyramidal cells of the Cornu Ammonis (CA) area 3 (CA3), which in turn send their excitatory projections to the pyramidal cells of area CA1 along the Schaffer collateral tract (Fig.

1.1). The CA1 neurons convey the major output of the hippocampus back to the EC and other cortical regions, allowing many researchers to propose that the hippocampus may act as a ‘way- station’ in the brain for memory acquisition3.

Figure 1.1 The Trisynaptic Loop of the Mouse Hippocampus. The perforant path from layer II neurons of the entorhinal cortex (EC) conveys the major input to the hippocampus. Axons of the perforant path make excitatory synapses with the dendrites of granule neurons in the dentate gyrus (DG), which in turn send their excitatory axonal projections, called the mossy fibers, to innervate apical dendrites of CA3 pyramidal neurons. The CA3 neurons project to ipsilateral CA1 neurons using the excitatory Schaffer collateral axonal tract; these synapses between the Schaffer collateral axons and CA1 dendrites are among the most well studied in all of neurobiology. The CA1 neurons send excitatory projections back to distinct populations of cells in the EC. Note that some populations of EC cells send axons directly to CA1 neurons through the temporoammonic pathway, which is not discussed in this dissertation. Also, CA3 neurons make modulatory projections to contralateral CA3 and CA1 populations through their commissural fibers. Adapted with permission from Neves et al., 2008.

The simple laminar pattern of the neurons and pathways within the trisynaptic loop is readily accessible to electrophysiological recording techniques, which have revealed much of what is known about excitatory neurotransmission3. Indeed, the slice preparation of the rodent hippocampus lends itself perfectly to the study of synaptic transmission as it can be kept healthy for many hours, enables the washing on and off of pharmacological agents, and allows specific pathways to be stimulated and the evoked synaptic responses to be measured for prolonged periods of time2.

Hippocampal neurons have also widely been studied in vitro as dissociated cell cultures that maintain key phenotypic properties of central nervous system (CNS) synapses in vivo, can 2

be maintained for up to months, and are highly amenable to acute chemical stimulation and molecular manipulations2, 4, 5. For example, dissociated hippocampal neurons in culture have been shown to establish axonal and dendritic polarity, and form mature functional and structural synaptic connections over a well-defined time course6-8. Furthermore, the excitatory cells of the hippocampus represent a relatively homogenous population of neurons with properties typical of

CNS neurons in general, and the pyramidal cells comprise up to 90% of the total population5.

Thus, dissociated cultures of hippocampal neurons are ideal for studying the development, maintenance, and plasticity of excitatory hippocampal synapses. In this dissertation, primary cultured hippocampal neurons were used as a model system to elucidate the molecular mechanisms underlying activity-regulated synapse plasticity.

1.2 Excitatory Synapses of the Central Nervous System

CNS synapses are characterized by the neurotransmitter with which they employ to transduce signals from one cell to another. Glutamatergic synapses comprise the major class of excitatory synapses within the adult mammalian brain and will be termed “synapses” in this dissertation. Excitatory synapses consist of presynaptic compartments within small boutons along the length of a cell’s axon in close apposition to a postsynaptic membrane localized to the tip of spiny protrusions along a neighboring cell’s dendrite (Fig. 1.2).

Figure 1.2. Schematic Depiction of an Excitatory CNS Synapse. Glutamatergic CNS synapses contain multiple protein classes. Synaptic vesicles that contain glutamate cycle at the presynaptic active zone, which is maintained by active zone proteins organized into a scaffold. The presynaptic terminal is immediately apposed to the postsynaptic compartment with the synaptic cleft in between, and the two are tethered together by a variety of transsynaptic cell adhesion molecules. Glutamate receptors, predominantly AMPA and NMDA-type receptors, are localized to the postsynaptic membrane, and stabilized by vast network of scaffold and structural molecules in the postsynaptic density. In mature synapses, the postsynaptic compartment is usually housed in the tip of dendritic spine protrusions emanating from dendrites. Adapted with permission from McAllister, 2007.

1.2.1 Composition of the Presynaptic Compartment

The directional nature of synaptic transmission is exemplified by the asymmetrical structure and composition among pre- and postsynaptic compartments. The presynaptic compartment is dedicated to releasing the neurotransmitter glutamate through membrane fusion of synaptic vesicles (SVs). The synthesis of glutamate from glutamine occurs directly in the cytoplasm of presynaptic terminals and is thought to be catalyzed by the enzyme phosphate- activated glutaminase9. Uptake of glutamate into SVs is driven by a proton-dependent electrochemical gradient across the vesicular membrane, mediated by a glutamatergic synapse-

specific protein termed vesicular glutamate transporter (VGluT;10). Structurally, the presynaptic compartment is characterized by the presence of hundreds of neurotransmitter-filled SVs and by the active zone; a specialized region of the plasma membrane where SVs dock, fuse, and release their cargo of neurotransmitters across the synaptic cleft11. A subdomain within the presynaptic terminal immediately across the synaptic cleft from the postsynaptic membrane is the cytoskeletal active zone (CAZ), which consists of an electron-dense matrix of proteins organized in lattice-like structures that are linked by filamentous molecules and tightly associated with the actin cytoskeleton.

The CAZ is proposed to function as a platform for the regulated movement of SVs through the active zone to the plasma membrane, and defines the locale of SV fusion and neurotransmitter release11, 12. A large number of proteins have been shown to localize to the

CAZ, including cell adhesion molecules (CAMs), structural and organizing molecules, and the machinery responsible for SV exocytosis and endocytosis (For summary, see Table 1.1)11, 13.

Synaptic Vesicle Cycling

Synaptobrevin, Vamp, Components of SNARE machinery responsible for SNAP25, syntaxin synaptic vesicle docking and fusion. Synaptotagmin Ca2+-sensing regulator of synaptic vesicle cycle. Munc18, munc13, Rim1α, Regulate synaptic vesicle priming. Rab3a N- and P/Q-type Calcium Mediate Depolarizing Calcium (Ca2+) influx triggered Channels by action potentials. Clathrin, Dynamin, Complex of proteins that interact with actin to regulate Amphiphysin synaptic vesicle endocytosis and recycling. Active Zone Structural Molecules Bassoon and Piccolo Large structural and scaffolding molecules of the CAZ that interact with actin binding proteins. Synapsins Link synaptic vesicles to actin cytoskeleton and regulate reserve pool of vesicles. Spectrin Actin binding protein, links with cell adhesion molecules. Synaptic Adhesion Molecules N-cadherin Homophilic adhesion molecule; dynamically interacts with actin cytoskeleton to control SV recruitment. Neurexins Bind postsynaptic neuroligins; can trigger active zone formation and associate with Ca2+ channels for active zone localization. SynCAM Homophilic adhesive molecule that can induce functional active zone formation. Table 1.1 Selected CAZ Proteins Involved in Active Zone Formation and Synaptic Vesicle Cycling.

Within the CAZ terminal, SVs can be functionally organized into three distinct pools that are loosely associated with their localization with respect to the presynaptic membrane: the readily releasable pool (RRP), the recycling pool, and the reserve pool (RP)14. 1-2% of total SVs within the presynaptic terminal are localized within the RRP, and which are docked at the presynaptic membrane and immediately available for exocytosis, whereas the RP contains between 50-90% of total SVs and which are not available for exocytosis. The percentage of SVs within the recycling pool can vary from between 5-50%, and constitutes the pool that can be

readily utilized for exocytic cycling during synaptic stimulation. SVs in the RP may not even be used during high intensity stimulation, and the precise function of which remain to be understood14.

The regulated movement of SVs between the different pools, in addition to SV fusion and neurotransmitter release, is governed by a process known as the synaptic vesicle cycle. The synaptic vesicle cycle is comprised of several major steps: SV docking, priming, fusion, and endocytic recycling. Vesicle docking involves the movement through the CAZ and attachment to the presynaptic membrane, followed by a priming step where a number of reactions render a docked SV competent for fusion14, 15. Primed SVs are ready for fusion upon depolarization and

Ca2+ influx, and can subsequently be endocytosed and recycled for later use14.

The actin cytoskeleton is abundantly rich at the presynaptic terminal16 and has been proposed to maintain the integrity of vesicle pools and the cycling of SVs between them, both by functioning as a scaffold to restrict SV movement and also as a conduit to direct SV transfer14.

Actin filaments are also major components of the active zone17, where they may function to both guide docking vesicles and facilitate the localization of fusion, or serve as a barrier for the priming reactions and prevent inappropriate fusion14. Transsynaptic CAMs localized to the membrane adjacent to the active zone can also work with the actin cytoskeleton and lead to the recruitment SVs to the presynaptic terminal (Table 1.1)18, 19, and therefore represent important regulators of synapse function.

1.2.2 Composition of the Postsynaptic Compartment

The postsynaptic compartment of excitatory synapses is localized to the tip of spiny protuberances along the length of a cell’s dendrites and is dedicated to receiving and transducing the signals from presynaptic glutamate release through the positioning of glutamate receptors at the postsynaptic membrane. Immediately beneath the postsynaptic membrane on the cytosolic side is the postsynaptic density (PSD), a 30-40nm thick specialized membrane microdomain apposed specifically to the active zone across the synaptic cleft. The PSD consists of an electron- dense protein matrix of which mass spectroscopy studies estimate contains more than 1100 different proteins20. The vast complex of proteins can be roughly divided into several groups, including 1) glutamate receptors, 2) scaffolding molecules, 3) CAMs, 4) G-proteins and associated modulators, 5) cytoskeletal organizers, and 6) signal transduction molecules including kinases and phosphatases (Fig1.3)21.

The resulting complex assembly of the PSD, which is built from highly enriched PSD core proteins and proteins that are loosely associated with it, is the molecular basis of highly localized and distinct intracellular events that are important for synapse function (Fig. 1.3).

The relative abundance of the PSD within the brain has allowed detailed biochemical characterization using differential fractionation and sucrose gradient centrifugation to obtain synaptosome fractions22. This is followed by detergent extraction with nonionic detergents, such as Triton X-100, which can allow separation of prominent protein bands on gels by SDS-PAGE

(sodium-dodecyl sulfate and polyacrylamide gel electrophoresis) and western blotting techniques22. However, some PSD proteins cannot be extracted by non-ionic detergents, and require the use of strong ionic detergents like SDS or sarcosinate23. Aside from postsynaptic receptors, some of the major bands and consequently most abundant PSD-resident proteins that 8

were initially isolated included postsynaptic density-95 (PSD-95; also known as SAP90 and

DLG-4), Shank, Homer, calcium/calmodulin-dependent protein kinase II (CaMKII), densin-180, synaptic GTPase-activating protein (SynGAP), and β-actin22, 24, 25 (Fig. 1.3).

Figure 1.3. Schematic Organization of Proteins at the PSD. Only major families and certain classes of PSD proteins are shown. The domain structure for PSD-95 is indicated (PDZ domain, SH3 domain, GuK domain). Abbreviations: AKAP150, A-kinase anchoring protein; CAM, cell adhesion molecule; CaMKII, calcium/calmodulin-dependent protein kinase II; Fyn, a Src family tyrosine kinase; GKAP, guanylate kinase-associated protein; H, Homer; IRSp53, insulin receptor substrate; KCh, K+ channel; mGluR, metabotropic glutamate receptor; nNOS, neuronal nitric oxide synthase; RTK, receptor tyrosine kinases; SPAR, spine-associated RapGAP. Adapted with permission from Sheng and Hoogenraad, 2007.

PSD-95 is among the most abundant of PSD proteins and is a member of the membrane- associated guanylate kinase (MAGUK) protein family22. Its structure is organized into distinct modes (Fig. 1.3): Localized to its N-terminus are three type I PDZ domains, named after the

proteins in which these sequences were first discovered (PSD-95, discs large, zona occludens

1)26. PDZ domains are well-defined 90 amino acid sequences that bind to short PDZ-binding motifs at the C-terminus of many proteins, and therefore mediate protein-protein interactions. At its C-terminus is a Src homology 3 (SH3) protein interaction module, known to bind proline-rich domains in other proteins, followed by a non-catalytic guanylate kinase (GK) domain27. The variety of domains through which PSD-95 can associate with other molecules in the PSD illustrates its organizational role in PSD assembly. PSD-95 also serves to anchor glutamate receptors within the postsynaptic membrane through direct and indirect interactions22, playing an important role in synapse function. For example, overexpression of PSD-95 in cultured hippocampal neurons increases glutamate receptor clustering28, while knockdown reduces receptor clustering at the postsynaptic compartment29.

Among the many organizational and signal transduction functions of the PSD, it is arguable that none are more important than the clustering and transport of membrane-bound glutamate receptors. The predominant glutamate receptors localized to the postsynaptic membrane that mediate fast excitatory glutamatergic transmission and supported by the PSD are

α-amino-3-hydroxy-5-methyl-4-isoxazoleproprionic acid type receptors (AMPA Receptors), and

N-methyl-D-aspartate type receptors (NMDA Receptors).

1.2.2.1 AMPA Receptors

AMPA Receptors (AMPARs) mediate the majority of fast excitatory glutamatergic neurotransmission in the brain. At hippocampal synapses, the receptors are largely Ca2+- impermeable, display fast signaling kinetics, and mediate rapid basal synaptic signaling30. These characteristics depend on the composition of AMPAR subunits, and their modifications 10

introduced by alternative splicing and RNA editing. AMPARs assemble from 4 subunits, GluA1-

GluA4, in various combinations to form heterodimeric tetramers30. At hippocampal excitatory synapses the predominant AMPAR subunit composition is GluA1/2 heterodimers, with a smaller fraction of GluA2/3 heterodimers31, 32.

The presence of the GluA2 subunit has a profound impact on the biophysical properties of AMPAR heteromeric complexes such that the GluA2-containing AMPARs are Ca2+- impermeable with a linear current–voltage relationship while GluA2-lacking receptors are Ca2+- permeable and have an inwardly rectifying current–voltage relationship33.

AMPAR subunit composition can also control receptor trafficking, synaptic delivery, and postsynaptic membrane localization through the subunits’ cytosolic C-terminal tails. The C-tails of the subunits are either long (e.g., GluA1 and GluA4) or short (e.g., GluA2 and GluA3)34, 35, and can govern the synaptic delivery of the receptor complex. The long-tailed AMPARs in the hippocampus (i.e. GluA1-containing) are important for the activity-dependent insertion of

AMPARs to synapses during synaptic plasticity36, 37, whereas the short-tailed AMPARs are largely present at synapses under basal conditions and appear to constitutively recycle in and out of synapses in the absence of activity34, 36, 38-40. However, internalization of both forms of

AMPARs occurs during other forms of synaptic plasticity that involve activity-dependent synaptic weakening41. Synapse plasticity will be discussed in chapter 1.4.

1.2.2.2 NMDA Receptors

NMDA Receptors (NMDARs) are another major class of receptors mediating excitatory glutamatergic neurotransmission at hippocampal synapses. In contrast to AMPARs, NMDARs do not mediate much of the fast basal synaptic transmission, but have been termed “coincidence 11

detectors” (i.e. the coincidence of glutamate release and postsynaptic depolarization42) and as such have critical roles in excitatory synaptic plasticity, as well as excitotoxicity. The involvement of NMDARs in these diverse processes reflects their unique features, which include a voltage-sensitive block by extracellular Mg2+, a high permeability to Ca2+ and unusually slow

‘activation/deactivation’ kinetics. NMDARs are both ligand-gated and voltage-gated; glutamate and its co-agonist glycine binding are required for opening, adding an additional layer of control over NMDAR conductance.

NMDARs are comprised of 4 subunits, each with an extracellular N terminus and intracellular C terminus43. The channel-lining domain is formed by a P loop that dips into the membrane from the cytoplasm and contains an asparagine residue critical for Ca2+ permeability and Mg2+ block44, 45. There are seven types of subunits in total, one GluN1, four GluN2 (A-D), and two GluN3. The GluN1 subunit is obligatory and contains the binding site for glycine45, whereas the GluN2 subunits confer differences in channel gating and conductance kinetics and contain the glutamate binding site45. Similar to AMPARs, NMDARs are assembled as heterotetramers, and in the hippocampus they possess two GluN1 and two GluN2 subunits45.

The expression of NMDAR subunits varies with respect to brain region, developmental period, and activity46-49, and interestingly, early in development there is a switch from expression of GluN2B subunits to higher expression of GluN2A subunits46-48. The gradual replacement of

GluN2B to 2A during postnatal development has been implicated in the speeding of the decay of

NMDAR-mediated currents, and this phenomenon has been linked with the ability of neuronal circuits to exhibit enhanced synaptic plasticity during developmental stages50.

Importantly, NMDARs diffuse laterally within the membrane between synaptic and extrasynaptic pools 51, profoundly impacting their function. Ca2+-influx through synaptic 12

NMDARs is commonly associated with synaptic plasticity, and can lead to the expression of pro- survival genes42, 52. In contrast, activation of extrasynaptic NMDARs can lead to re-uptake of

Ca2+ into mitochondria and breakdown of their membrane potential, DNA damage, and cell death53-55. Perturbations in the balance between synaptic and extrasynaptic NMDAR activity can therefore contribute dramatically to neuronal dysfunction.

Within the hippocampus, NMDARs also form the basis for what are identified as “silent synapses”3. In the CA1 subfield, silent synapses are defined as such a synapse in which an excitatory postsynaptic current (EPSC) is absent at the resting membrane potential but becomes apparent upon depolarization. Commonly, silent synapses are unmasked during synapse plasticity and strengthening. The absence, or silence of EPSCs under basal conditions reflects the functional presence of NMDARs but not AMPARs at these synapses because only AMPARs can conduct current under basal conditions (i.e. not depolarized). Therefore, the absence of functional postsynaptic AMPARs renders these synapses unable to mediate synaptic transmission under physiological conditions3. The discovery of silent synapses was a critical step toward understanding the molecular mechanisms of synapse strengthening, which will be discussed in Chapter 1.4. Furthermore, the differences between AMPARs and NMDARs with respect to their role in neurotransmission illustrate the principals of excitatory neurotransmission in the CNS (Fig. 1.4).

Figure 1.4. Schematic Depiction of Basal Synaptic Transmission at a Hippocampal Synapse. (Top Panels) At these synapses, AMPAR and NMDAR glutamate receptors are both localized to the postsynaptic membrane. Both are permeable to Na+ and K+, and with reversal potentials close to 0 mV. However, NMDARs exhibit important interactions with divalent cations: whereas Ca2+ is highly permeant, mediating the important signaling functions of these receptors, Mg2+ gets stuck in the pore, producing voltage-dependent NMDAR blockade at negative membrane potentials (blue trace in bottom left panel). At the resting membrane potential (left-hand panels), synaptic glutamate will evoke an excitatory postsynaptic current (EPSC) that is mediated almost entirely by AMPARs (red trace in bottom left panel), and therefore they mediate the vast majority of fast basal synapse transmission. Depolarized potentials (for e.g. +40mV) will relieve the Mg2+ blockade, and EPSCs will subsequently contain contributions from both AMPARs and NMDARs (black trace in bottom right panel). The specific traces also highlight an important difference between the receptors’ kinetics: AMPAR-mediated currents activate quickly and decay within milliseconds, whereas NMDAR-mediated currents activate more slowly and decay over hundreds of milliseconds. It should be noted that under physiological conditions the membrane potential of the postsynaptic compartment never reaches positive values. Adapted with permission from Kirchner and Nicoll, 2008.

1.2.3 Dendritic Spines

The postsynaptic compartment of the glutamatergic synapse is typically housed in spiny protrusions emanating along the length of dendrites called dendritic spines. Spines vary in length

(typically 0.5-2µm), are rich in actin cytoskeletal filaments, and have highly diverse 14

morphologies56. The formation of spines is thought to initiate from dendrites as immature filopdia-like protrusions longer than 2 µm, which may subsequently give rise to mature spines housing functional postsynaptic complexes57. Live-imaging studies of spines in vivo have supported this hypothesis, as spine number increases as animals mature, accompanied by a decrease in filopodia58, 59. Similarly, early dendritic filopodia have been observed transitioning into shorter and less motile spiny protrusions as animals age60. A great deal of work has shown the critical role spines play in synapse development and function, which will be discussed in detail below.

1.2.3.1 Spine Morphology

Spine morphology is commonly, and loosely, classified as stubby spines, elongated thin spines, and mushroom spines. Filopodia are often classified as protrusions distinct from spines

(Fig. 1.5). Filopodia are elongated, thin, headless, and highly motile protrusions that lack postsynaptic machinery and emerge from dendrites in young neurons during development (Fig.

1.5). Spines are commonly characterized as containing two distinct structural elements, the actin cytoskeleton and the PSD28, 61, with filopodia lacking a PSD62, and the initiation of spine morphogenesis has been observed to precede the accumulation of PSD-95 in spine heads63. As such, the question of how spine morphogenesis occurs is often framed in terms of how the actin cytoskeleton is locally regulated within a filopodium such that PSD assembly can be initiated.

Therefore, it has been hypothesized that filopodia serve as precursors to mature spines housing a functional PSD (Fig. 1.5)62.

Figure 1.5. Morphological Classes of Dendritic Spines and a Filopodium. Protrusions emanating from dendrites of GFP-transfected hippocampal neurons at (A) 7 days-in-vitro or (B) 21 days-in-vitro are predominantly filopodia or spines, respectively. Scale bar = 5µm. (C) Schematic depiction of the common morphological classifications of spines relative to a filopodium. The gray disks at the tip of the protrusions denote the PSD, and the red chains denote actin filaments. Adapted with permission from Sekino et al., 2007.

Although direct evidence for filopodia functioning as precursors in spine morphogenesis has not been proven, evidence for a role of the actin cytoskeleton and actin-binding proteins

(ABPs) in driving the formation of functional spine-synapses has been demonstrated in cultured neurons62. Overexpression of Drebrin A, a brain-specific side-binding protein of actin filaments, accelerated spine formation64. During 2 weeks of development in cultured hippocampal neurons,

Takahashi et al., observed that filopodia density decreased concurrently with a significant increase in spine density and clustering of filamentous actin65. Takahashi et al., then observed that the clustering of Drebrin A in filopodia preceded that of PSD-95 and spine morphogenesis, and that acute knockdown of Drebrin A reduced the fraction of filopodia that developed into spines65. This study and another66 established that the recruitment of specific ABPs to particular

filopodia could drive the recruitment of PSD-95, and initiate PSD assembly and spine morphogenesis.

Spines are particularly rich in actin filaments that exhibit highly dynamic activity during spine morphogenesis and maintenance. Actin is required for synapse formation in young neurons, as treatment of 7 DIV cultures with Latrunculin A, a molecule that binds and sequesters actin monomers, completely disrupted synapse assembly and ultrastructure, whereas treatment of

14 DIV cultures did not significantly impact synapse number67. Mature synapses, therefore, transition to more stable scaffold-dependent structures67.

Actin is also highly dynamic during the maintenance of mature spines under basal conditions. The conserved mechanism of actin regulation is called actin “treadmilling.” Actin monomers within persistent filaments undergo a continuous process of exchange, with polymerization and addition occurring at the barbed growing end, and severing and depolymerization occurring simultaneously at the pointed end68, 69. In spines, most actin filaments undergo complete turnover of their constituent monomers within 60 seconds68, 70. The distribution and activity of actin filaments within spines is not completely uniform; targeted stimulation of photoactivatable-actin to the tip of spines demonstrated that higher rates of polymerization were occurring at the tip adjacent to the PSD, generating a net flow of actin monomers toward the spine base71. Moreover, the velocity of filament polymerization was significantly elevated in the vicinity of the PSD relative to the spine core and neck, and endocytic zones72, presumably to maintain a dynamic scaffold for AMPARs. These results were in agreement with previous work showing that treatment of mature cultured neurons with cytochalasin D, an inhibitor of actin polymerization (with limited actin filament severing activity), reduced AMPAR clustering in spines, but did not dramatically impact spine structure73. 17

Thus, the distribution of filamentous actin activity within spines is arranged as a dynamic network of hubs that can be independently controlled and balanced to maintain postsynaptic structure and PSD function.

1.2.3.2 Spine Structure–Function Relationship

Spines are enormously heterogeneous with regard to their overall morphology (Fig. 1.5), and a strong correlation has been observed for spines with large, wide heads and high postsynaptic AMPAR content74. For example, examination of spine morphology with an electron microscope level combined with immunogold labeling of AMPARs demonstrated that spines with larger PSDs exhibited significantly higher AMPAR immunoreactivity compared to spines with smaller PSDs75. By using light-induced release of caged glutamate, a later study was able to address the synaptic function of individual spines along the dendrites of CA1 hippocampal neurons, and observed that sensitivity to glutamate varied dramatically among individual spines

(Fig. 1.6)76. The amplitude of glutamate-induced currents was highest at spines with the largest heads, whereas thin spines and filopodia exhibited the lowest amplitude currents (Fig 1.6)76. No correlation was observed for the AMPAR content of a stimulated spine and its neighboring spines, indicating that spines can independently control the efficacy of their synaptic function76.

Figure 1.6. Spine Structure and Expression of Functional AMPA Receptors. (a–j) Fluorescence images (upper panels) and glutamate-sensitivity heat maps (lower panels; whole cell recording of EPSCs as evoked by 2-photon uncaging of MNI-Glutamate) for various dendritic spines of CA1 pyramidal neurons in rat hippocampal slices. The larger spines in panels f-j contain higher amplitude glutamate- induced currents than thin spines and filopodia in upper panels a-e, correlating spine size with synapse strength. Adapted with permission from Kasai et al., 2003.

The morphology of mature spines has been shown to compartmentalize biochemical and electrical signals77-79, however the extent to which compartmentalization is controlled by a particular parameter of spine geometry (i.e. spine head width, length, neck width, neck length, volume etc.) has been difficult to discern. A recent study used super-resolution microscopy

(Stimulated-Emission-Depletion Imaging; STED) in combination with Fluorescence Recovery after Photobleaching (FRAP) imaging in CA1 cells transfected with yellow fluorescent protein

(YFP) in order to accurately measure correlations between the morphological parameters of individual spines and their biochemical diffusion properties80. FRAP imaging of YFP-filled spines revealed that small changes in spine neck width had the largest impact on molecular diffusion, indicating that this parameter functions as a diffusional barrier and is a primary determinant of a spine’s biochemical compartmentalization80. Furthermore, this and multiple

other studies have demonstrated that the morphology of spines changes following synaptic activity, and is an essential process in synapse plasticity. This will be discussed further in

Chapter 1.4.

1.2.3.3 Spine Abnormalities and Neuropsychiatric Disease

Over the course of development and into adulthood, changes in the number and morphology of spines accompany synapse formation, maintenance, and plasticity, working to establish and fine-tune the connectivity of neural circuits58, 59. As spines are intrinsically linked with synapse efficacy, even subtle aberrations in their structure and number can have a profound impact on the patterns of connectivity within circuits.

Neuropsychiatric diseases like autism spectrum disorder (ASD), schizophrenia, and

Alzheimer’s Disease (AD) share a common severe deficit in information-processing that is accompanied by impaired neural connectivity and plasticity81. Although synaptic aberrations are not the only relevant pathological traits of these disorders, understanding the molecular mechanisms underlying aberrant spine structure and function may provide insight into their etiologies. A collection of studies that utilized Golgi-staining of postmortem brain tissue from human patients have characterized Schizophrenia and AD as involving advanced synapse and spine loss through adulthood, whereas ASD is characterized as increased spine and synapse formation through adolescence (Fig. 1.7)81.

Genome-wide studies have recently provided much insight to the molecular underpinnings of these diseases and the cellular basis for their associated spine perturbations.

Indeed, many of the identified genes encode for synaptic proteins. Rare point mutations in the genes for both Neuroligin and Neurexin CAMs have been linked to ASD82, and the expression of 20

these point mutants in cultured hippocampal neurons impact the number of excitatory synapses83.

Mutations and de novo copy number variations (CNVs) in the genes encoding the Shank family of the PSD proteins, which are important for spine maintenance and turnover84, 85, have also been

Fig. 1.7. Dendritic Spine Numbers through Human Development among Normal Individuals and Patients with Neuropsychiatric Disorders. The common emergence of symptoms and diagnosis of ASD, Schizophrenia (SZ), and Alzheimer’s Disease (AD) is indicated by the bars across the top of the graph. In normal individuals, the density of spines increases before and after birth, and is subsequently pruned and reduced as spines are selectively eliminated through the course of development into adulthood. In ASD, exaggerated spine formation or incomplete pruning may occur in childhood, leading to an aberrant increase in spine numbers. In schizophrenia, exaggerated spine pruning during late childhood or adolescence may disrupt neural circuits and lead to the emergence of symptoms during these periods. In Alzheimer's disease, spines are rapidly lost in late adulthood, suggesting perturbed spine maintenance that may underlie cognitive decline. Adapted with permission from Penzes et al., 2011.

identified in ASD patients86, 87. Furthermore, rare mutations in a Rap guanine nucleotide exchange factor (GEF) localized to synapses, Epac2, have been identified in ASD patients88 and also regulate spine density and morphology89. Interestingly, Neuroligins can associate with

Epac2 and enhances its activity89, and Shank proteins may also bind Neuroligins90. A complex of

Neuroligin-Neurexins, Shank proteins, and Epac2 comprise a common synaptic signaling pathway that may potentially be disrupted in ASD81.

Mutations in a number of synaptic proteins that control spine morphology and density have been identified in schizophrenic patients81, 91. In particular, many mutations have been found in genes that encode for PSD scaffold molecules or associated regulatory proteins. 1-2% of schizophrenia cases have involved microdeletions in a region of chromosome 22 that encode the gene for DHHC892, a palmitoyl acyl-transferase enzyme that can palmitoylate PSD-95 and regulate spine numbers93. Mutations to the DISC1 gene have been commonly identified in schizophrenic patients94, 95 and can regulate spine morphology through the Rac GEF Kalirin-796.

Interestingly, a reduction in the gene encoding Kalirin-7 has also been observed in the brains of some schizophrenic patients97. The DISC1 protein also interacts with PSD-95 in spines96, and together with the loss of DHHC8 function in chromosome 22 microdeletion patients92, 93, the disease-associated disruptions in the functions of DISC1 and Kalirin-7 could cause a loss of scaffolding molecules in spines, and contribute to the etiology of schizophrenia.

A well-identified risk factor for late-onset AD is the inheritance of the ε4 allele of the gene encoding apolipoprotein E98, 99. Expression of the ε4 allele in mice reduced spine number along pyramidal neurons in the cortex and granule neurons in the DG region of the hippocampus compared to mice expressing another allele100, 101. This suggests that the molecular mechanisms underlying a major genetic risk factor for AD might involve the regulation of spines.

1.3 Inhibitory CNS Synapses

The majority of inhibitory synapses in the mature mammalian CNS consist of γ- aminobutyricacid (GABA)-releasing axonal terminals apposed to dendritic shafts containing 22

GABA receptors (GABARs) and associated proteins required for GABAergic transmission.

Unlike glutamatergic synapses, which are predominantly formed on dendritic spines,

GABAergic contacts more often develop directly on the shafts of target dendrites. GABA- releasing terminals also commonly contact the neuronal soma and axon initial segments, placing them close to the sites of action potential initiation, where they can exert a proportionately stronger influence on neuronal firing.

Presynaptically, GABAergic terminals are morphologically similar to glutamatergic boutons, and share much of the same fundamental molecular machinery (Table 1.1). GABAergic terminals achieve the biosynthesis of GABA through the decarboxylation of glutamate via L- glutamic acid decarboxylase (GAD). Unlike other neurotransmitter systems, GABA synthesis is mediated by two different enzymes, GAD65 and GAD67102. Once GABA synthesis occurs, it is packaged into SVs by a distinct transporter protein termed vesicular GABA transporter

(VGAT)103.

Postsynaptic GABARs fall into three distinct classes: GABAA, GABAB and GABAC receptors. The ionotropic GABARs, GABAA and GABAC receptors, are members of the ligand- gated superfamily of ion channels, whereas GABAB receptors are metabotropic G-protein coupled receptors104. Many of the postsynaptic constituents of GABAergic synapses have remained elusive, due to the lack of a PSD-like structure and a well-established biochemical fractionation method for selective isolation. Despite the lack of an extensively defined composition of the molecular components of GABAergic synapses relative to glutamatergic synaptic contacts, an extensive protein network has been implicated in the function of these synapses (Fig. 1.8).

Figure 1.8. Schematic of a GABAA Receptor with Associated Postsynaptic Cytosolic Proteins. Associated scaffold molecules: Gephyrin, a trimeric scaffold molecule for GABAARs, associates with the γ2 subunit. N-ethylmaleimide sensitive factor (NSF) is associated through an interaction with another scaffold molecule, GABARAP. AP2, an adaptor protein for endocytic complexes, associates with β and γ subunits, and may be involved in receptor internalization. AKAP150/79, a scaffold for Protein Kinase A (PKA) and protein phosphatase 2B (PP2B), directly binds the β subunit. Associated signaling molecules: Protein Kinase C (PKC) is associated with the β subunit through an interacting protein, RACK1. Cytoskeletal associated proteins: GABARAP binds the β subunit. Motor proteins: a hypothesized motor protein, GRIF1, binds the β subunit and may traffic GABAARs to nascent inhibitory synapses. Ubiquitination and degradation pathways: PLIC1, which recruits GABAARs into the ubiquitination pathway, binds the β subunit. Adapted with permission from Collingridge et al., 2004.

Some of the specific proteins found associated with GABARs in the postsynaptic compartment are actually common to both synapse types, including cadherin and neuroligin

CAM family members, AKAP150/79, and protein kinases (Fig. 1.8)105, 106. However, molecules specific to inhibitory synapses have also been identified, like the cytoplasmic multidomain protein, gephyrin, which organizes another inhibitory postsynaptic-specific protein,

GABARAP107, and it itself is critical for GABAR clustering108 (Fig. 1.8). Gephyrin’s function involves assembly of an organizing scaffold for GABARs and other associated proteins106.

1.4 Synapse Plasticity

Synapses of the CNS are specialized junctions of cell-cell contact designed to rapidly and efficiently transmit information from one neuron to another. The morphology and strength of synapses are continually altered in response to neural activity, and therefore synapses are considered to be highly plastic structures. Synapse plasticity has been shown to be critical for the fine-tuning of brain development, as well as higher brain functions such as learning and memory, and fluctuations in neural activity can directly impact synapse plasticity. Therefore, an understanding of the molecular underpinnings of changes in synaptic strength and morphology in response to activity may reveal neural mechanisms of learning and memory. This dissertation will focus exclusively on the plasticity of excitatory synapses.

The hippocampus has long been understood as a brain region essential for the acquisition of new memories1, and indeed, one of the first cellular phenomena hypothesized to underlie learning and memory was observed in acute hippocampal slices more than 40 years ago109.

Experimenters delivered high frequency electrical stimulation to the perforant pathway innervating the granule neurons of the DG, and observed a significant increase in the amplitude of evoked responses in the granule cells that lasted for many days109. This long-lasting enhancement, or potentiation, of synaptic strength was termed long-term potentiation (LTP; Fig.

1.9), and has since been extensively studied among hippocampal neurons.

Figure 1.9. Hippocampal Long-Term Potentiation. Delivery of a presynaptic stimulus to the Schaffer collateral axonal tract paired with depolarization of postsynaptic CA1 cells results in long-lasting functional and structural synaptic enhancement. (a) The amplitude of excitatory postsynaptic potentials (EPSPs; mV) in a single CA1 neuron evoked by presynaptic stimulus before and after the pairing protocol (red bar). A stable and long-lasting enhancement of the postsynaptic response efficacy is observed after pairing. (b) 2-photon microscope images of a dendrite of the GFP-transfected and stimulated cell at the indicated time points before and after pairing. Red arrowheads denote new spines that appeared approximately 30 minutes after pairing and were maintained through the experimental endpoint. Scale bar = 5µm. Adapted with permission from Engert and Bonhoeffer, 1999.

1.4.1 Long-Term Potentiation

After the initial demonstration of LTP, a subsequent study provided experimental support for Donald Hebb’s theory of the neural mechanism for memory storage110, demonstrating that 1) the only synapses that were strengthened during LTP were those that were active during delivery of the depolarizing stimulus in the postsynaptic cell, and 2) inactive synapses were not 26

potentiated111. Therefore, collections of synapses that are activated in a coordinated manner and contribute together to firing an action potential in the postsynaptic neuron will be strengthened by an LTP stimulus. This provides a mechanism for linking various ensembles of neurons that encode the different features of an experience that are presented together, and forming associations to represent a memory112.

A number of subsequent studies established the basic cellular mechanisms underlying the expression of LTP and the strengthening of synapses among CA1 neurons of the hippocampus.

Induction of LTP typically involves the delivery of a high-frequency stimulus paired with depolarization of the postsynaptic cell, and requires the activation of NMDA receptors3.

NMDAR activation results in a strong postsynaptic inward Ca2+ current that activates a number of signaling kinases, including CaMKII, PKA, PKC, and mitogen-activated protein kinase, all of which have been shown to be required for LTP induction3, 113, 114. It should be noted that early debate in the LTP field centered upon the locale of an LTP mechanism at the synapse as being either presynaptic or postsynaptic. Although not discussed in this dissertation, the evidence to date strongly points to the postsynaptic compartment driving the majority of changes in synapse strength3.

1.4.1.1 Functional LTP Studies of LTP expression observed that GFP-tagged AMPARs were directly inserted into the postsynaptic membrane of transfected CA1 neurons following stimulation, and that this required an interaction between the GluA1 subunit and a postsynaptic PDZ-domain protein36, 37.

This and work by others established that activity-induced trafficking of AMPARs to the synapse is a critical determinant of the functional synaptic strengthening occurring during LTP.

The GluA1 subunit of AMPARs was demonstrated early to be required for LTP to occur.

At hippocampal synapses, GluA1/2 heteromers were specifically inserted into the postsynaptic membrane following activity stimulation, whereas GluA2/3 heteromers cycled to synapses constitutively in the absence of stimulation34. The long C-tail of the GluA1 subunit was found to dominate that of GluA2 and determine the trafficking characteristics of GluA1/2 heteromers during plasticity34. In agreement with this result, ablation of GluA1 in vivo completely inhibited hippocampal LTP115. However, the subcellular localization of the source from which GluA1/2 and GluA2/3 AMPARs were inserted at the postsynaptic membrane following activity and under basal conditions remains a contentious subject. Although not discussed in detail here, the work of many groups has established a consensus model for activity-induced AMPAR insertion that involves contributions from extrasynaptic AMPAR exocytosis, surface diffusion, and synaptic capture, in addition to direct exocytosis into spine heads116-122.

1.4.1.2 Structural LTP

Structural enhancement of dendritic spines was first observed in CA1 neurons in which the potentiated synapses along a specific region of a cell’s dendritic arbor were isolated from other blocked synapses123. The induction of LTP was accompanied by an enlargement of preexisting spines and the formation of new spines that lasted for hours, and was not observed along cells where NMDAR activation was inhibited 123. These results demonstrated a structural form of LTP (Fig. 1.10).

Figure 1.10. LTP Induction Results in Morphological Enhancement of Spines. (Top Panels) 2-photon images of GFP-transfected CA1 neuron’s dendrite before and at time points up to 12 hrs after a paired stimulation protocol. White arrows (top) denote the enlargement of a preexisting spine, and (bottom) the formation of a new spine. Scale bar = 3µm. (Bottom Panel) EPSP amplitude (mV) of the imaged cell before and after pairing (red bar) over the course of the experiment. Adapted with permission from Engert and Bonhoeffer, 1999.

Activity-induced spine enlargement and increased AMPAR surface levels are two highly correlated, but separate processes that occur during LTP. Experiments utilizing glutamate uncaging onto single spines reveled that stimulated spines undergo a transient period of substantial enlargement, lasting up to 15 minutes after stimulation, but only small spines exhibited a persistent enlargement lasting up to 100 min and that was dependent on CaMKII activation124. In another experiment in cultured neurons, overexpression of PSD-95 occluded the induction of functional LTP, but activity-induced spine enlargement was still observed125, suggesting that functional and structural LTP involve different cellular pathways. Live imaging demonstrated that spine enlargement preceded the arrival of AMPARs by several minutes126, and the expression of a GluA1 “channel pore-dead” mutant inhibited functional LTP and did not impact spine enlargement, however the expression of GluA1 with a mutation in its C-terminal

PDZ domain was sufficient to inhibit both functional and persistent structural LTP127. Thus, functional and structural LTP appear to be highly correlated but divergent processes that share 29

some common regulatory mechanisms, but much remains unclear about what coordinates the two phenomena.

The observation that mushroom spines with large, wide heads were positively correlated with high postsynaptic AMPAR content76 implied that the structural enlargement of spines observed during LTP must involve a substantial increase in spine head width. Indeed, head width is increased significantly following induction of LTP124, 128-130, and has since been considered to be the most prominent morphological parameter determining synaptic strength. However, the shortening and widening of spine necks was also observed in some experiments following LTP stimulation131, and recent work has established that widening of both spine heads and necks occurs in a concerted fashion to enhance spines’ biochemical compartmentalization80.

Regulation of actin dynamics within spines is necessary for activity-induced structural enlargement71, and the Rho family of GTPases, which catalyze the activity of actin binding proteins, has also been shown to be required132. In hippocampal neurons transfected with fluorescently tagged RhoA and cdc42, two Rho family members, glutamate uncaging over a single spine resulted in the rapid activation and trafficking of RhoA and cdc42 to the stimulated spine133. A small increase in RhoA was observed in spines adjacent to the stimulated one, however cdc42 activity was entirely restricted to the activated spine133, indicating that activity- induced structural enhancements require precise protein trafficking to the activated synapse.

Recently, the trafficking of entire protein complexes specifically to activated spines in distinct temporal phases was observed following LTP stimulation134, 135, suggesting cells can coordinately direct the trafficking of proteins to activated synapses. However, the molecular mechanisms underlying how such coordinated trafficking occurs remains unclear.

1.4.2 Long-Term Depression

Researchers investigating synaptic plasticity and LTP in the hippocampus observed that stimulation of schaffer collateral axonal tracts occasionally failed to potentiate the postsynaptic response in CA1 neurons, and could actually reliably weaken the efficacy of synaptic transmission for long periods of time. A protocol that could consistently induce the long-term diminishment of synaptic strength was identified and involved the delivery of long trains of low frequency stimuli, in contrast to the high frequency stimuli used for LTP induction, and was termed long-term depression (LTD)136. Paradoxically, just like LTP the induction of LTD at hippocampal synapses was shown to require the activation of NMDARs and inward Ca2+ flux136.

However, LTD results in the activation of protein phosphatases and the dephosphorylation of many substrate proteins that are modified by kinases during LTP137, 138. Subsequently, the dephosphorylation and surface removal of AMPARs was found to underlie the functional reduction in synapse strength139, 140.

1.4.2.1 Functional LTD

That NMDAR activation can paradoxically lead to LTP and LTD, two opposite synaptic outcomes, may be explained by differences in the inward Ca2+ flux141. It appears that LTP requires a high influx of Ca2+ to act upon low-affinity kinases like CaMKII, whereas LTD necessitates only a moderate Ca2+ flux to activate a high-affinity Ca2+-dependent phosphatase cascade, including Calcineurin and protein phosphatase 1137, 138, 141, and leads to the internalization of synaptic AMPARs139, 140. In contrast to the case with LTP, in LTD the GluA2

C-tail regulates constitutive receptor cycling and endocytosis through interactions with N- ethylmaliemide sensitive factor (NSF) and the clathrin adaptor protein AP2, respectively34, 142. 31

1.4.2.2 Structural LTD

The positive correlation between spine volume and postsynaptic AMPAR content, which had informed how spines may respond to LTP-inducing stimuli also held that small spines should contain less AMPARs than large ones (Fig. 1.6)76. Therefore, LTD stimuli that result in the removal of synaptic AMPARs may also result in spine shrinkage and elimination. Indeed, 2- photon time lapse imaging of CA1 neurons in hippocampal slices during low frequency stimulation confirmed this hypothesis143, 144, demonstrating that spines were reduced in volume and eliminated over 6 hours of imaging (Fig. 1.11). Blockade of NMDARs completely inhibited spine loss, showing that the structural component of NMDAR-dependent LTD involves a reduction and removal of spines143, 144.

Figure 1.11. LTD Induction Causes Spine Shrinkage and Elimination. 2-photon images of the dendrites of GFP-transfected CA1 neurons in a hippocampal slice at the time points indicated (minutes) during no stimulation (C) or low-frequency stimulation (LFS; D-F). Open arrowheads indicate stable spines during the imaging period. White arrows indicate spines targeted for shrinkage. White arrowheads indicate spines that were eliminated. Scale bar = 1µm. Adapted with permission from Zhou et al., 2004. 32

The correlation among functional and structural synaptic enhancement during LTP also holds true for LTD, and the molecular pathways underlying the two processes appear to share some common mechanisms. Simultaneous imaging and recordings from CA1 neurons revealed that LTD induction revealed that while both spine shrinkage and AMPAR removal require activation of the phosphatase calcineurin, inhibition of protein phosphatase 1 or NSF blocked

AMPAR removal and not spine shrinkage145. Peptide inhibition of ADF/cofilin, an actin filament depolymerizing protein, blocked LTD-induced spine shrinkage but not AMPAR removal145.

Intriguingly, treatment of cells with Jasplakinolide, an actin-stabilizing molecule that blocks all actin dynamics, inhibited both AMPAR removal and spine shrinkage145. LTD-induced AMPAR removal and spine shrinkage do appear to involve independent molecular pathways, however actin remodeling represents a common substrate involved in both phenomena that contributes to the physiological expression of LTD.

1.4.3 LTP and LTD in Learning and Memory

While much of the synaptic plasticity field focused on the cellular and physiological mechanisms underlying the LTP and LTD, others investigated if they occurred in vivo and were correlated with the acquisition and retention of new memories. The initial demonstration that

LTP induction requires NMDAR activation146 was supported when AP5, an NMDAR antagonist, was infused into the ventricles of rats before exploring a water maze and inhibited their ability to learn the location of a hidden platform147. This demonstrated that NMDAR activation is required for both LTP and the acquisition of spatial memory. A subsequent study conditionally ablated the GRIN1 gene, encoding for the obligatory GluN1 subunit of NMDARs, in the CA1 area of the hippocampus148, and demonstrated that NMDAR-mediated synaptic transmission was 33

completely absent in CA1 neurons, but intact in recordings from DG cells in the same slice148.

LTP was abolished in CA1, and the transgenic mice were significantly deficient in navigating a water maze compared to wildtype littermates148. The results established that NMDAR- dependent, LTP-like synapse plasticity in the hippocampus is required for the acquisition of new spatial memories.

Stronger evidence for the involvement of LTP in learning and memory was provided in rats in which LTP was observed in vivo during learning149. The animals were trained in a contextual fear paradigm, and a significant increase in an LTP-associated phosphorylation site along the C-tail of GluA1 was observed, demonstrating a biochemical correlation between the two phenomena149. The field activity of CA1 neurons in vivo was recorded while the animals behaved, and EPSPs were significantly enhanced in trained animals relative to their baseline synapse strength and to naïve controls, showing that LTP-like enhancements of synapse strength occur during learning149. Trained animals were then subjected to stimulation of the schaffer collateral axonal tract that innervates the CA1 neurons’ dendrites in order to electrically induce

LTP, but which was occluded in the trained animals and not in naïve controls149. These results allowed the authors to conclude that learning induces hippocampal LTP and supporting its role as a cellular correlate of learning and memory. Nevertheless, the evidence provided at that time was still largely viewed as correlative.

Recent work has come close to establishing a causal link between LTP/LTD and memory formation. Rats were trained using an auditory associative conditioning assay that requires the amygdala, a region of the brain involved in providing an emotional context to hippocampal- dependent memories150. The animals were injected with a virus encoding a variant of channelrhodopsin, a blue light-activated Na+ channel that induces depolarization and action 34

potential firing in the cells in which it is expressed, targeted to the auditory cortex that extends axons to the amygdala150. The animals were trained to associate the onset of an electrical foot shock with optical (i.e. blue light) stimulation in vivo presented in a paired manner, and this produced a conditioned response (CR;150). Animals that were subjected to a paired stimuli showed significantly increased AMPAR mediated EPSCs in amygdala neurons, indicating that the associative training induced an LTP-like effect150. The subsequent delivery of low frequency optical stimulation induced LTD in the cells of the amygdala that had been activated by the paired training, resulting in a disruption of the animals’ CR response150. Thus, LTD had inactivated the animals’ memory. Following this, high frequency optical stimulation induced

LTP in the same amygdala cells that were engaged by the paired training but inactivated by

LTD, and this resulted in the animals’ recovery of the CR150. Therefore, memory reactivation was driven by LTP, and this result demonstrated that LTP and LTD underlie memory acquisition and inactivation, respectively150. The results of this study suggest that neurons use LTP to form assemblies of strengthened synapses that together represent a memory, and that LTD may be used to disassemble them and inactivate, or forget, the memory. Therefore, studies examining the molecular mechanisms underlying synapse plasticity and LTP/LTD can provide important insight into the cellular underpinnings of learning and memory.

1.5 Cadherin–Catenin Adhesion Complexes at CNS Synapses

Classic cadherins function as key organizers during the formation and remodeling of synapses in the vertebrate central nervous system (reviewed in 151). Cadherins are Ca2+- dependent homophilic adhesion molecules whose adhesive strength can be regulated by

conformational changes, through cadherin’s association with intracellular binding proteins, and through cellular regulation of cadherin turnover and internalization.

Cadherin surface concentrations are critical to determining the strength of adhesion.

Intercellular interactions between two cadherin molecules (in trans) results in the formation of weakly-adhering cadherin monomers, whereas cis clustering of two or more cadherin monomers results in the formation of strongly-adhering strand dimers152. Both the extracellular and intracellular domains of cadherins are essential for cadherin-based adhesion152-154. At its C- terminus, cadherin interacts with β-catenin, which dynamically associates with α-catenin, a direct binding partner of filamentous actin155. β-catenin also acts as a scaffold to recruit a variety of proteins to cadherin adhesion sites including PDZ domain-containing proteins, protein kinases, and phosphatases (reviewed in 151). Adhesion is significantly weakened in cells expressing cadherin lacking the β-catenin binding site, underscoring the importance of this interaction in adhesion154. The juxtamembrane domain of cadherin is also important for adhesion through its regulation of cadherin cis-clustering as well as signaling pathways that modulate actin dynamics152, 154. Proteins that bind to this domain include p120-catenin (p120ctn), δ-catenin, p0071 and ARVCF.

1.5.1 Cadherins and Synapse Formation

Multiple cadherins are expressed in the CNS and exhibit distinct spatial and temporal patterns of expression156. The most widely expressed and best-studied cadherin in the brain is N- cadherin. The subcellular localization of N-cadherin within synaptic contact sites is developmentally regulated, perhaps reflecting differential roles of N-cadherin in young and mature synapses157. Similarly, the expression of N-cadherin at excitatory and inhibitory synapses 36

varies through development, with N-cadherin initially localizing to both excitatory and inhibitory synapses, and later localizing primarily to excitatory synapses158. Interestingly, β-catenin is localized to both mature excitatory and inhibitory synapses, suggesting that alternate cadherin isoforms may localize to and function at mature inhibitory synapses158. The following sections will largely focus on the role of cadherins and catenins in the formation of excitatory hippocampal synapses, the bulk of which has been studied using cultured hippocampal neurons in vitro.

1.5.1.1 Cadherin Function at Presynaptic Compartments

Presynaptic compartments are formed through the recruitment of preassembled clusters of synaptic proteins to sites of cell–cell contact11. One such preassembled cluster, the Piccolo–

Bassoon transport vesicles (PTV) transports proteins associated with the active zone, including cytoskeletal matrix proteins and proteins involved in synaptic vesicle exocytosis. The other identified preassembled cluster, the synaptic transport vesicle (STV), consists of pleiomorphic vesicles that transport proteins associated with the mature SV11. As a caution, the sharpness of these classifications remains somewhat controversial.

Accumulating evidence suggests that cadherin adhesion complexes play a central role in localizing SVs to presynaptic compartments, however its role in the recruitment and localization of PTVs is unclear. Disruption of intercellular cadherin interactions in hippocampal cultures perturbs the clustering of SVs159, but does not alter the localization of the PTV marker,

Bassoon160. The localization of SVs has been shown, in part, to depend on cadherin’s association with β-catenin161 and the subsequent recruitment of the PDZ protein, Scribble, to nascent synapses 162 (Fig. 1.12). Indeed, ablation of β-catenin in vivo and in vitro, and knockdown of 37

Scribble in cultured neurons results in the mislocalization of SVs along the axon, while Bassoon localization is primarily unaffected161, 162. It is unclear how Scribble regulates the localization of

SVs, however as Scribble is a member of the LAP (leucine-rich repeat and PDZ domain) protein family, it likely acts as a scaffold to recruit additional proteins that localize SVs 162 (Fig. 1.12).

Phosphorylation of β-catenin on tyrosine residue 654 can rapidly disrupt cadherin-β- catenin interactions, and mobilize SVs associated with synapses163. Tyrosine kinases and phosphatases that control β-catenin tyrosine phosphorylation also regulate SV localization164, 165.

The tyrosine kinase, Fer, is recruited to cadherin complexes through its association with p120ctn166. Fer phosphorylates and activates the tyrosine phosphatase, SHP-2, which in turn dephosphorylates β-catenin, thereby promoting cadherin-β-catenin interactions166 (Fig. 1.12).

Knockdown of Fer or SHP-2 (and the subsequent disruption of cadherin-β-catenin interactions) inhibits the clustering of SVs at synapses166. Interestingly, knockdown of Fer, and disruption of

Fer-p120ctn interactions also disrupts the proper localization of the PTV marker, Bassoon166

(Fig. 1.12). As cadherin-β-catenin complexes do not appear to be involved in the localization of

PTVs to synapses, it would be of interest to identify the pathway by which p120ctn, Fer and

SHP-2 mediates the recruitment of PTVs to synapses.

Phosphorylation of β-catenin is also regulated by leukocyte antigen related (LAR) tyrosine phosphatase, which is localized at synapses through its association with liprin-α1, the vertebrate homolog of the C. elegans protein Syd2167. Syd2 is important for the clustering of

SVs, and Syd2 mutants exhibit diffuse SV localization along the axon168. It is possible that liprin-α1 partially modulates vesicle localization by recruiting LAR tyrosine phosphatase to nascent synapses, where it maintains the integrity of the cadherin-β-catenin complex by dephosphorylating β-catenin. Further experimentation is required to test this hypothesis. 38

Recent work has demonstrated that postsynaptic cadherin can act trans-synaptically to localize SVs and enhance synapse formation through its ability to recruit and functionally interact with another postsynaptic cell adhesion molecule, neuroligin-1 (NL1)160, 169.

Overexpression of N-cadherin in postsynaptic cells enhances the clustering of postsynaptic NL1, and increases the clustering of SVs in young neurons synapsing onto N-cadherin overexpressing cells160, 169. Although NL1 expression is sufficient to enhance the clustering of SVs along the axon160 and induce the formation of new synapses169, expression of NL1 in an N-cadherin null background dramatically attenuates NL1’s effects160, 169. These two adhesion systems can be linked by their association with the postsynaptic scaffolding protein, S-SCAM. S-SCAM binds to

β-catenin170 and plays an important role in localizing NL1 to synapses171. The association between S-SCAM and β-catenin, as well as S-SCAM and NL1, is essential for NL1-mediated

SV clustering160 (Fig. 1.12).

1.5.1.2 Cadherin Function at Postsynaptic Compartments

Cadherins and their associated catenins are also key mediators of postsynaptic development. Cadherins play critical roles in regulating the formation and maturity of dendritic spines; excitatory postsynaptic specializations observed in many CNS neurons. Disruption of intercellular cadherin interactions in cultured hippocampal neurons results in the formation of immature, filopodia-like protrusions and a concomitant decrease in the accumulation of the postsynaptic protein, PSD-95159, 172, 173. Interestingly, postnatal ablation of N-cadherin in vivo does not affect baseline spine density or morphology, but does have profound effects on activity- dependent spine enlargement and LTP at mature CA1 synapses174. Induction of LTP promotes

the clustering of N-cadherin within spines and leads to the selective growth and stabilization of

N-cadherin-associated spines173.

Cadherin’s regulation of spine maturation in culture is mediated by its interaction with the catenins and their subsequent control of actin dynamics. Indeed, ablation of α-catenin, β- catenin, p120ctn, and δ-catenin each result in phenotypes similar to that observed following disruption of cadherin-based adhesion, including fewer dendritic protrusions with mature spine- like morphology, and a larger proportion of motile, filopodial-like protrusions175-178.

Overexpression of α-catenin substantially enhances the density of mature spines, through its association with β-catenin and actin175. However, the immature spine phenotype observed in

β-catenin knockdown cultures can be rescued by overexpression of wildtype β-catenin, but not β- catenin constructs deficient in cadherin binding, or one lacking the C-terminal PDZ domain176.

This suggests that cadherin- β-catenin can regulate spine morphology through its dynamic association with α-catenin as well as through β-catenin’s interactions with PDZ domain- containing postsynaptic molecules. Previous work has shown that enhanced activity drives β- catenin to spines, where it interacts with cadherin to influence the accumulation of postsynaptic proteins165. It would be of interest to determine whether activity-dependent changes in spine morphology require the translocation of β-catenin into spines.

Cadherin can also regulate spine morphology through its association with p120ctn and δ- catenin (Fig. 1.12). Loss of p120ctn in hippocampal pyramidal neurons reduces spine and synapse density and decreases spine head width177. The reduction in spine density can be attributed to aberrant RhoA-GTPase activity, whereas the decreased spine head width is due to reduced p120ctn-cadherin interactions, and a subsequent decrease in Rac1-GTPase activity177.

Thus, the dynamic association between cadherin and p120ctn may dictate the availability of 40

p120ctn to mediate Rho-GTPase signaling or cadherin-based adhesion and its subsequent involvement in spine density or size, respectively (Fig. 1.12).

δ-catenin shows high homology to p120ctn, but unlike p120ctn which is expressed in most tissues, δ-catenin is exclusively expressed in the brain179. Acute knockdown, and overexpression of δ-catenin both result in a high proportion of immature spines with filopodial morphology178, 180. Although this appears paradoxical, further work is required to understand the mechanism by which δ-catenin regulates spinogenesis. Several studies have demonstrated that δ- catenin can inhibit RhoA, and activate Rac1 and cdc-42 GTPases similar to p120ctn, resulting in changes in spine morphology181, 182 (Fig. 1.12). δ-catenin binds several PDZ domain-containing proteins, including a number of postsynaptic receptor scaffold molecules, such as AMPA-type receptor binding proteins ABP and GRIP, and PSD-95, and these interactions may also dictate δ- catenin’s role in postsynaptic development183, 184 (Fig. 1.12).

N-cadherin can also enhance spine maturation by recruiting the Rac guanine nucleotide exchange factor (GEF), kalirin-7, into spines through the adaptor molecule AF-6/afadin185.

Signaling between N-cadherin and AF-6 may be regulated by AF-6’s interaction with α- catenin185 (Fig. 1.12). Inhibition of intercellular cadherin interactions reduced N-cadherin association with AF-6 and kalirin-7, leading to reduced Rac1 activation and the formation of immature filopodia-like protrusions185.

Finally, cadherin can also regulate spine maturation independent of the catenins, through its direct association with the AMPA receptor subunit, GluR2186 (Fig. 1.12). Indeed, GluR2 is unable to enhance spine formation in N-cadherin knockdown cells, indicating that this interaction is essential for GluR2-mediated control of spine morphology186.

Figure 1.12. Schematic Depiction of the Classic Cadherin/Catenin Adhesion Complex at Excitatory Synapses. (1-3) Cadherins mediate the localization of vesicles to presynaptic compartments through multiple mechanisms. (1) p120ctn recruits Fer to cadherin sites. Fer phosphorylates and activates SHP-2, which dephosphorylates β-catenin and promotes cadherin – β-catenin interactions. Cadherin-bound β- catenin then recruits scribble (Scrib) which acts as a scaffold for the further recruitment of proteins that mediate the localization of SVs. (2) Transsynaptic interactions between multiple adhesion complexes can also localize SVs at synapses. Postsynaptic cadherin - β-catenin complexes recruit S-SCAM, which further clusters NL1 to these sites. Binding of NL1 with presynaptic neurexin clusters SVs at presynaptic compartments. (3) p120ctn and Fer are important for the localization of PTVs to synapses. (4-7) Cadherin - catenin complexes regulate spine head width, length and density through multiple pathways. (4) p120ctn bound to cadherin can activate Rac GTPase and enhance spine head width. (5) Recruitment of Kalirin-7 (Kal-7) to cadherin sites via AF-6 and α-catenin, activates Rac1 GTPase and enhances spine head width. (6) p120ctn and δ-catenin that are not associated with cadherin inhibit RhoA GTPase and regulate spine length and density. (7) The dynamic interaction between cadherin, β-catenin and α-catenin, also regulates actin polymerization and spine length. Bi-directional arrows denote dynamic interactions; black arrows denote interaction or activation; white arrows denote outcomes at synapses. Adapted with permission from Brigidi and Bamji, 2011.

1.5.2 Regulation of Cadherin-Based Adhesion at the Synapse

Synapses are highly dynamic structures that undergo reorganization following activity187.

The composition and stability of cadherin clusters is also altered following activity, suggesting cadherins as potential regulators of activity-dependent changes in synaptic architecture.

Evidence suggests that activity stabilizes cadherin at the membrane. Under basal conditions, N-cadherin is rapidly endocytosed, however following activation of NMDA receptors, the rate of N-cadherin internalization is dramatically reduced, effectively positioning more cadherin within the synaptic membrane188. This is mediated by inhibiting the phosphorylation of β-catenin, thereby enhancing cadherin - β-catenin associations and stabilizing cadherin at the plasma membrane188 (Fig. 1.13). Activity also enhances the formation of the strongly adhering cis-dimeric strand conformation that stabilizes cadherin at the surface, however this appears to be independent of β-catenin binding, suggesting that cadherin stabilization within the synaptic membrane occurs through β-catenin-dependent and independent mechanisms189 (Fig. 1.13). NMDA receptor activation is required for activity-dependent enhancement of spine size, indicating that NMDAR-mediated stabilization of cadherin at the membrane underlies spine morphogenesis188, 190.

Contrasting evidence suggests that activity can also diminish cadherin-based adhesion.

NMDA receptor activation can increase the activity of the transmembrane protein presenilin-1

(PS1), and enhance PS1’s ability to cleave N-cadherin within its transmembrane domain191 (Fig.

1.13). Enhanced neuronal activity also increases the expression of the protocadherin, arcadlin, which binds to the transmembrane domain of cadherin and induces endocytosis of the entire N- cadherin-arcadlin complex 192 (Fig. 1.13).

The binding of p120ctn to the cadherin juxtamembrane domain enhances cadherin stabilization at the membrane by; 1) blocking the binding of clathrin adaptor proteins that target it for endocytosis193, and 2) blocking the phosphorylation of two tyrosine residues within cadherin’s juxtamembrane domain that are critical for the recruitment of the E3 ubiquitin ligase,

Hakai, which targets E-cadherin for proteolytic degradation193-195 (Fig. 1.13). Single point mutations in p120ctn that specifically abolish its ability to bind the cadherin juxtamembrane domain enhance the internalization of cadherin and disrupt spine maturation193.

Recent work in developing Zebrafish Rohon-Beard neurons demonstrates how developmentally regulated proteolytic processing of cadherin may be a mechanism to control the kinetics of synapse formation196. In young neurons, the predominant form of N-cadherin contains a prodomain that inhibits adhesion (Fig. 1.13), whereas older neurons express mature N-cadherin lacking the prodomain. Overexpression of N-cadherin containing the prodomain blocks the formation of synapses in young neurons but does not affect synapse numbers in older neurons indicating that cleavage of the prodomain may serve as a mechanism to control the rate of synapse formation196.

Regulation of cadherin membrane stability is also regulated by the scaffold protein,

IQGAP1. IQGAP1 is highly colocalized with N-cadherin in spines, and has recently been shown to enhance N-cadherin stability and promote adhesion197. Disruption of intercellular cadherin interactions diminishes cadherin’s association with IQGAP1, resulting in a reduction of stable N- cadherin cis-clusters and an inhibition of memory acquisition197.

Considerable work has been done to examine how cadherin complexes regulate synaptic architecture, however several major questions remain unanswered with respect to activity- induced synapse plasticity. For example, how do fluctuations in neuronal activity translate to 44

changes in cadherin-based adhesion at the synapse? In turn, how might such changes regulate functional and structural synaptic efficacy? Much work is needed to elucidate the roles of the catenin molecules in regulating the synaptic stability of cadherin in response to activity.

Figure 1.13. Schematic Depiction of Mechanisms Regulating Cadherin-Based Adhesion at the Synapse. Cadherin membrane stabilization and adhesive activity can be enhanced by the activation of NMDA receptors (NMDAR), by: (1) inhibiting β-catenin phosphorylation and thereby enhancing cadherin – β-catenin interactions; and (2) by enhancing the formation of strongly-adhering cadherin cis- dimeric clusters. (3) Binding of p120ctn to the cadherin juxtamembrane domain (JMD) enhances cadherin stability within the membrane by blocking the binding and modification of cadherin’s JMD by endocytic factors. These include clathrin adaptor proteins (AP), tyrosine kinases, and the ubiquitin ligase Hakai (Hk), which normally induce cadherin endocytosis and degradation. Activation of NMDARs can also reduce cadherin membrane stability by: (4) increasing the association between cadherin and the protocadherin, arcadlin, thereby enhancing cadherin turnover; and (5) increasing presenilin-1 (PS1) activity, and enhancing PS1-mediated cleavage of cadherin. (6) Expression of pro-cadherin in immature cells blocks homophilic cadherin interactions and inhibits adhesion. Adapted with permission from Brigidi and Bamji, 2011.

1.6 δ-Catenin at the Synapse

The cadherin juxtamembrane domain is essential for cadherin clustering in cis152, 154, and has been shown to be important for adhesion in trans154, supporting an important role for the catenin proteins bind this domain. The juxtamembrane domain of N-cadherin has been shown to be bound by several members of the p120ctn subfamily of catenin proteins, however in the brain it is bound exclusively δ-catenin198. This suggests a role for δ-catenin in synapse function through the regulation of cadherin-based adhesion.

1.6.1 Postsynaptic Signaling Pathways of δ-Catenin

δ-catenin was first identified in a screen for molecules that interacted with the Presenilin1 component of the γ-secretase complex199, 200. In mature neurons, δ-catenin is exclusively localized to dendrites and enriched in spines179, where it links cadherin to the actin cytoskeleton through direct interactions with Cortactin180, the Rho-family GTPases Rac1 and cdc42182, and the

RhoA guanine nucleotide exchange factor p190RhoGEF181 (Fig. 1.14). δ-catenin has also interacts with the γ-secretase complex, which can cleave N-cadherin201 (Fig. 1.14), however the function of this association is unclear. Interestingly, cadherins compete with p190RhoGEF for δ- catenin binding, suggesting that δ-catenin shuttles between postsynaptic membrane-proximal regions and the cytosol, and potentially exerting different functions in each compartment202.

Moreover, serine phosphorylation of δ-catenin by Cdk5 drives it away from the synaptic membrane203 (Fig. 1.14). It is unclear what modification or signaling mechanism recruits δ- catenin to the synaptic membrane to associate with cadherins.

In addition to being a major binding partner of cadherins, δ-catenin has been shown to

associate with several postsynaptic scaffold molecules and AMPA receptor binding proteins in a

PDZ-domain dependent manner, including PSD-95 183, S-SCAM 204, and GRIP1/ABP 184, 205

(Fig. 1.14). When transfected into heterologous cells ABP exhibits a diffuse cytosolic distribution, however when co-transfected with δ-catenin it becomes highly localized at adherens junctions at the cell membrane, dependent on PDZ interactions184. The formation of a stable cadherin–δ-catenin–ABP complex at the cell surface also leads to the recruitment of the GluA2-

AMPARs184, which subsequently become destabilized as a result of δ-catenin trafficking to the cytosol, driven by Cdk5 phosphorylation203. This suggests that δ-catenin–cadherin complexes may serve as anchorage sites for AMPARs at the cell membrane via the PDZ interaction between

δ-catenin–GRIP1/ABP.

In addition to the cytoplasmic link between cadherin and AMPARs via δ-catenin and

ABP, the GluA2 subunit of AMPARs can directly bind to N-cadherin through their extracellular domains186. As an increase in the number of AMPARs inserted into the postsynaptic membrane is a hallmark characteristic of LTP37, it would be interesting to examine if δ-catenin enhances the surface levels of AMPARs at the postsynaptic membrane by linking ABP/AMPAR complexes with cadherin. However, such a role for δ-catenin has not been explored.

Overexpression of δ-catenin in cultured hippocampal neurons enhances the formation of dendritic spines with an elongated, filopodia-like morphology180, 181. Interestingly, acute knockdown of δ-catenin in cultured neurons also induces a similar immature spine phenotype178,

182, however it is unclear if aberrant changes in spine morphology are due to δ-catenin’s regulation of the cadherin adhesion complex, the actin cytoskeleton, or through δ-catenin’s PDZ- dependent interactions with postsynaptic scaffold molecules.

Figure 1.14. Schematic Depiction of known δ-Catenin Postsynaptic Signaling Pathways. δ-catenin is localized to the postsynaptic compartment, and not in axons. (1) δ-catenin can bind the juxtamembrane domain of N-cadherin, the function of which is unclear. Cadherins compete with Rho GTPase proteins for δ-catenin binding. (2) Serine phosphorylation by Cdk5 drives δ-catenin away from the membrane toward the cytosol. (3-5) δ-catenin localized to the cytosol can (3) undergo serine/tyrosine phosphorylation by an unknown kinase that drives it to enhance the activity of Rac1 and cdc42 GTPases. This in turn activates LIM kinase (not shown) to modulate actin dynamics, and leads to changes in spine head width and new spine formation, respectively. (4) δ-catenin can inhibit RhoA by sequestration of p190RhoGEF, which is catalyzed by Akt1-mediated threonine phosphorylation, and leads to inhibition of Rho-mediated spine elongation. (5) δ-catenin associates with cortactin, which recruits the Arp2/3 complex and leads to actin branching and spine growth. Tyrosine phosphorylation of δ-catenin by Src family kinases inhibits its association with cortactin. (6) δ-catenin localized to membrane associates with GRIP/ABP, two AMPAR binding proteins, and these interactions stabilize surface AMPARs at the postsynaptic membrane. δ- catenin also binds PSD-95, which is linked to NMDARs, however the function of this interaction is unknown. δ-catenin may be bound to N-cadherin and AMPAR binding proteins simultaneously, linking receptor-scaffold assemblies with the cadherin adhesion complex, and potentially leading to membrane stabilization. (7) δ-catenin associates with the Presenilin 1 (PS1) component of the γ-secretase complex that can cleave N-cadherin. The function of the δ-catenin–PS1 interaction is not clear. The white arrow denotes movement of δ-catenin from a cadherin-bound, membrane-associated localization toward the cytosol. Black arrows denote activation and/or association.

1.6.2 Aberrant δ-Catenin Function and Cognitive Performance

Mutations in δ-catenin have been shown to severely impair cognitive function, and may underlie some forms of mental retardation and schizophrenia206-209. In humans, the gene encoding δ-catenin is localized to a critical region of chromosome 5 that underlies the severe mental retardation of Cri-du-Chat syndrome. In Cri-du-Chat patients, the severity of mental retardation correlates with hemizygous loss of the δ-catenin gene209. The importance of δ-catenin in cognitive function and neural connectivity is also supported by the substantial deficits in learning, memory, and synaptic plasticity demonstrated in mice with a targeted ablation of the δ- catenin gene206. Additionally, δ-catenin knockout mice show normal neuronal structure and architecture 5 weeks after birth, but exhibit a progressive loss of dendritic complexity, spine density and stability, and visual cortex responsiveness by 10 weeks of age, implicating δ- catenin’s critical importance in maintaining neural connectivity210. Together, these findings suggest a critical role for δ-catenin in coordinating activity-induced enhancements in synapse adhesion, structure, and efficacy that are important for learning and memory. The function of δ- catenin in synapse plasticity remains unclear.

δ-catenin was recently identified as a substrate of protein palmitoylation (a post- translational lipid modification to be discussed in detail in chapter 1.7) in a large palmitoyl- proteomic screen of neuronal proteins211. However, the functional relevance of δ-catenin palmitoylation remains unknown. Because neuronal activity has the potential to regulate the dynamics of δ-catenin palmitoylation, and in turn control the recruitment of δ-catenin to the synaptic membrane and its association with N-cadherin, palmitoylation may represent a critical mechanism controlling δ-catenin trafficking and function during synapse strengthening.

1.7 Neuronal Protein Palmitoylation

Posttranslational modifications, such as phosphorylation, glycosylation, ubiquitination, and lipidation, confer additional functions to substrate proteins. These modifications have been designed for cells to regulate their homeostasis and to rapidly respond to extracellular signals.

Lipidation of substrates increases the hydrophobicity, and therefore contributes to their precise targeting, subcellular trafficking, and function with respect to cellular membranes212. N- myristoylation, S-palmitoylation, and prenylation are common lipid modifications occurring among cytoplasmic substrates (Fig. 1.15). N-myristoylation is a co-translational attachment of

14-carbon myristic acid to an N-terminal glycine residue through a stable, permanent amide linkage213, 214. Prenylation is the post-translational attachment of a prenyl group, which can involve a 15-carbon farnesyl or 20-carbon geranylgeranyl group, to a C-terminal cysteine- containing motif, CaaX (“a” and “X” represent aliphatic and any residues, respectively; Fig.

1.15)215. S-palmitoylation involves the addition of palmitate, a 16-carbon saturated fatty acid, to cysteine residues through a labile thioester linkage212 (Fig. 1.15), which makes S-palmitoylation unique among all other lipid modifications in that it is reversible. Among the three, S- palmitoylation is thought to most efficiently increase protein hydrophobicity and facilitate its substrates’ membrane association212. A small subset of proteins, including the secreted morphogen Sonic hedgehog, undergo an irreversible N-palmitoylation at glycine and cysteine residues, which occurs in the luminal face of the protein secretory pathway216. S-palmitoylation will be referred to as “palmitoylation” in this dissertation.

Among mammalian cell types, neurons exhibit an especially polarized morphology. Their elongated processes, axons and dendrites, project hundreds of microns from the cell body and house asymmetric synaptic compartments. Such a polarized morphology and specialized 50

synaptic compartmentalization require precise trafficking and distribution of proteins among these distal locales. Among the common lipid modification described above, palmitoylation is the most highly utilized modification among neuronal proteins211, 217.

Palmitoylation modifies numerous soluble cytoplasmic proteins and integral transmembrane proteins, including signaling proteins and enzymes, scaffold molecules, ion channels, CAMs, and glutamate receptors217. Palmitoylation-depalmitoylation cycling on neuronal proteins can be constitutive218-221, or dynamically regulated by cellular signals such as synaptic activity222-224.

Palmitate cycling on cytosolic proteins typically regulates a substrate’s trafficking among various cellular membranes. The proto-oncogene family members, N-ras and H-ras, are subjected to both prenylation and palmitoylation, and have been well studied with regard to their palmitoylation-dependent trafficking cycles225, 226. The ras proteins are palmitoylated at the somatic Golgi, which targets and stabilizes their association with the lipid membranes of vesicles that traffic them toward the plasma membrane. At the plasma membrane, they are depalmitoylated, released from their membrane association and into the cytoplasm where they return to the Golgi for another round of palmitoylation225, 226. Importantly, the Golgi and plasma membrane pools of ras proteins activate distinct downstream signaling cascades227. Therefore, palmitate cycling can modify the function of substrates through regulation of their localization.

Figure 1.15. Schematic Depiction of Palmitoylation and Common Lipid Modifications. (A) Chemical Structures of N-myristoylation, irreversible N-palmitoylation, thio (S) palmitoylation, and a form of prenylation, farnesylation. N-myristoylation involves the addition of a 14-carbon lipid, myristate and occurs at the N-terminal glycine residue via a stable amide linkage. N-palmitoylation is thought to occur initially via a thioester intermediate utilizing the thiol of the cysteine residue, followed by a spontaneous rearrangement to generate an irreversible amide linkage. S-palmitoylation involves the addition of 16- carbon palmitate fatty acid to a cysteine residue through a reversible thioester linkage. Prenylation is the addition of either a 15-carbon farnesyl (farnesylation) or a 20-carbon geranylgeranyl (geranylgeranylation; not shown) group, to a cysteine residue via a stable thioether linkage. (B) Mechanism of reversible S-palmitoylation (termed “palmitoylation). Palmitate is transferred from palmitoyl-CoenzymeA (CoA), which generated in the cell by acyl-CoA synthetase, to a substrate via palmitoyl acyl-transferases (PATs), which are not shown here. The palmitate modification can be cleaved by palmitoyl-protein thioesterases (PPTs). Gray boxes denote cysteine residues where palmitoylation occurs. Adapted with permission from Iwanaga et al., 2009.

Many integral membrane proteins are also palmitoylated, which has been shown to regulate their localization to subdomains of the plasma membrane, such as lipid rafts which are regions rich in sphingolipids and cholesterols228. Synaptic membranes in dendrites are abundantly enriched in lipid rafts, and blockade of lipid raft formation destabilizes AMPARs and leads to an overall breakdown of spines and loss of synapses229. More work is needed however to determine the precise role of lipid rafts at synapses.

1.7.1 Palmitoyl Acyl-Transferases: The DHHC Family of Proteins

Forward genetic studies in yeast first identified the enzymes responsible for mediating palmitate attachment, or palmitoyl acyl-transferases (PATs). A mutant Ras protein was generated in which the prenylation consensus sequence was mutated, leaving plamitoylation as the protein’s only lipid modification230. Yeast therefore solely depended upon palmitoylation to regulate the essential function of the Ras protein, and through a genetic lethality screen two enzymes responsible for mediating palmitate attachment were identified, Erf2 and Erf4230, 231. A subsequent study identified another PAT, Akr1 that palmitoylated yeast casein kinase 2 232. Erf2 and Akr1 were found to have a similar structure consisting of 4-6 transmembrane domains, and critically, share a homologous cysteine-rich domain (CRD) that contains an identical aspartate- histidine-histidine-cysteine (DHHC) motif 233. A large screen was then employed to identify the mammalian PATs, and pulled out a family of 23 DHHC motif-containing proteins that have been shown to be responsible for the attachment of palmitate in mammalian cells 234.

Mammalian DHHC proteins have been classified into several subfamilies according to their phylogenetic relationship with one-another 217 (Fig. 1.16), and were shown to be sufficient to mediate the palmitoylation of neuronal proteins 217, 235. The DHHC proteins are though to 53

catalyze the transfer of palmitate from palmitoyl-CoA to substrate proteins by binding their substrates through their diverse N- and C-termini (Fig. 1.15-16), undergoing transient autopalmitoylation within their consensus CRD and DHHC motif, and then transferring palmitate to specific cysteine residues of their substrates 217, 236 (Fig. 1.16). Evidence for the transient autoplamitoylation step comes from site-directed mutagenesis of the cysteine within the

DHHC motif to a serine, which has been used in a number of studies to generate dominant- negative DHHS clones that are not sufficient to catalyze palmitate attachment, and consequently disrupt the function of their substrates218-221, 235. Some DHHC proteins have been shown to oligomerize, forming multimeric-signaling complexes along the somatic Golgi membrane with decreased enzymatic activity relative to the monomeric forms218, 237. More work is needed to understand the function of DHHC multimerization.

No specific consensus sequence has been identified for substrates of palmitoylation, in sharp contrast that for myristoylation and prenylation212, 217. However, the DHHCs do exhibit substrate specificity with some overlap235, 238, and multiple neuronal substrates’ palmitoylated cysteines are near hydrophobic and basic residues235. Therefore, it was been proposed that the

DHHCs recognize a specific pattern for palmitoylation involving basic and hydrophobic resides proximal to cysteines212. Online palmitoylation-prediction software has been developed that examines potential substrates for such patterns (CSS-Palm 3.0)212, 239, and has provided useful starting points for identifying site(s) of palmitate attachment.

The DHHCs are localized to a variety of cellular membranes, with the majority resident on the somatic Golgi membrane240, but in neurons some DHHCs are localized to distal synaptic compartments far from the cell body. DHHC2 is a PAT for PSD-95, and has been well studied with respect to its neuronal localization and dynamic regulation of PSD-95 palmitoylation235, 241, 54

242. Following increased activity, DHHC2 enhanced the palmitoylation of PSD-95 and its localization at the PSD through an increase in its own localization to the postsynaptic membrane, and not through any change in its enzymatic activity242. Membrane-localized DHHC2 can locally maintain palmitate cycling on PSD-95 within dendritic spine heads241, regulating surface

AMPAR levels241, 242. Therefore, DHHC2 can traffic between various cell membranes (also see

243), demonstrating that neurons change the palmitoylation levels of various substrates by positioning their PATs in their proximal vicinity.

Such dynamic localization observed for DHHC2 is not the case for all the DHHCs; for example, DHHC3 also palmitoylates PSD-95 but is stably localized to the membrane of Golgi compartments235, 242. The observed changes in the localization of DHHC2 occurred following slow homeostatic regulation of neuronal activity over many hours241, 242, and to what extent, if any, that neurons can rapidly respond to acute changes in synaptic activity by regulating the subcellular localization of DHHCs is unclear.

Figure 1.16. The Family of Mammalian DHHC Proteins and Representative Enzyme-Substrate Pairs. (a) Schematic structural representation of DHHC3 (also known as GODZ), which contains 4 transmembrane domains and the conserved cysteine-rich domain (CRD) and DHHC motif facing the cytoplasmic side of cellular membranes. DHHCs are substantially divergent in their N- and C-termini, and DHHC3 contains a PDZ domain-binding motif at its C-terminus. (b) Phylogenetic organization to mouse DHHCs into subfamilies based upon alignment of DHHC-CRD domains. DHHC mouse clones were more recently classified as zDHHCs, referring to a common zinc-finger domain shared among the family. The zDHHC and DHHC nomenclature are used interchangeably, with the exception of the following discrepancies: DHHC10 (zDHHC11), DHHC11 (zDHHC23), DHHC13 (zDHHC24), DHHC22 (zDHHC13), and DHHC23 (zDHHC25). (c) Schematic representation of structures and identified substrates for selected DHHCs, as well as related diseases for which each DHHC is implicated. Beside the common DHHC-CRD domain, the DHHC proteins share transmembrane domains, and some contain PDZ-binding motifs, SH3 domains, or ankyrin repeats, which can mediate protein-protein interactions. Within the amino acid sequence of the consensus DHHC-CRD domain, letters highlighted in red and green denote conserved of the DHHC motif and CRD, respectively. “X” represents any amino acid. Adapted with permission from Fukata and Fukata, 2010. 56

The identification of the enzymes responsible for protein depalmitoylation, the PPTs, has been extremely limited to date217. Acyl-protein thioesterase 1 (APT1) has been shown to

244, 245 depalmitoylate Gαi and Ras proteins , and is localized to dendritic spines where it can regulate morphology246. PPTs like APT1 likely play critical roles in regulating palmitate cycling on neuronal proteins and in synapse function, and much work is needed to identify more PPTs for neuronal proteins.

1.7.2 DHHC Proteins in Cognitive Function and Neurological Disease

DHHC proteins have been shown to be involved in higher brain functions like learning and memory, supporting a role in cognitive function. DHHC5 has been identified as a PAT for two synaptic proteins, PSD-95 and Grip1b220, 236, 247, and mice homozygous for a hypomorphic allele of the ZDHHC5 gene (“DHHC5 gene-trap mice”) were significantly impaired in the acquisition of hippocampus-dependent contextual fear memories compared to wildtype littermates247. Indeed, DHHC5 (and its homologue DHHC8) exhibits higher expression in the brain relative to other DHHCs220, and is particularly enriched in the DG and CA1 areas of the hippocampus247, supporting its importance in hippocampal learning and memory through a potential regulatory role in synapse plasticity.

DHHC5 has also been implicated in neuropsychiatric disorders. Genome-wide association studies have reported the occurrence of mutations within a region of chromosome 11 containing ZDHHC5 in patients with bipolar disorders and schizophrenia248, 249. Furthermore, a de novo nonsense mutation in the DHHC5 protein has also recently been reported in schizophrenic patients91, indicating a possible involvement of DHHC5 in these neuropsychiatric disorders, and further supporting the importance of DHHC5 in regulating cognitive function. 57

The PAT DHHC17 (also known as Huntingtin-interacting protein 14; HIP14), is also enriched in the brain and was first identified as a PAT for the huntingtin protein that is mutated in mouse models of Huntington’s Disease (HD)250, 251, suggesting that deficient palmitoylation of

DHHC17 substrates may underlie some of the pathology of HD. Ablation of the ZDHHC17 gene in mice resulted in motor deficits and disruptions in spatial memory, recapitulating many phenotypic characteristics of HD-model mice252. Moreover, excitatory synapses in the striatum and hippocampus were significantly reduced, and hippocampal LTP was abolished252, 253.

Therefore, loss of DHHC17/HIP14 results in substantial neurophysiological deficits associated with the cognitive dysfunction observed in HD.

The gene encoding DHHC8, a close homologue of DHHC5, is disrupted in a large percentage of schizophrenic patients with chromosome 22 microdeletions92, and the specific loss of DHHC8 directly contributes to the risk of developing schizophrenia254. A mouse model of the specific chromosome 22 microdeletion that disrupts DHHC8 expression (Df(16)A+/-) exhibits significant reductions in dendritic spines and synapses, and in the frequency of miniature EPSCs, and this phenotype can be completely rescued by re-introduction of functional DHHC893.

Ablation of the ZDHHC8 gene resulted in a similar phenotype as the Df(16)A+/- mice, as well as alterations in the morphology of dendritic branches in the hippocampus, and a significant reduction in the palmitoylation of PSD-9593. The phenotypes of these mouse models suggest that

DHHC8 is critically involved in the regulation of the neural connectivity and may contribute to the cognitive and information-processing deficits observed in patients with chromosome 22 microdeletions.

A number of other DHHC proteins have been implicated in neurological disorders255-257 and neurodegenerative diseases251. The generation of additional loss-of-function transgenic mice 58

for the DHHC proteins implicated in synapse function or neurological malfunction will aid in modeling neurological disease and in revealing the role of this family of proteins in cognitive performance.

1.7.3 Dynamic Palmitoylation of Synaptic Substrates

Palmitoylation modifies a number of synaptic proteins, and regulates diverse aspects of their function at synapses. PSD-95 was among the first synaptic substrates of palmitoylation to be examined, and palmitate attachment is critical for its synaptic localization222, 242. Treatment of cultured hippocampal neurons with glutamate (which typically results in receptor internalization) caused the depalmitoylation of PSD-95, a reduction in AMPAR clustering, and reduced amplitude and frequency of miniature EPSCs 222. Conversely, a homeostatic increase in synaptic activity significantly increases PSD-95 palmitoylation and increases AMPAR surface levels 242.

Palmitoylation of gephyrin is required for its localization it to inhibitory postsynaptic compartments where it clusters GABAARs, in a somewhat analogous manner to PSD-95 at excitatory synapses223. Therefore, palmitoylation of PSD-95 and gephyrin maintains their localization at postsynaptic compartments to continually regulate neurotransmitter receptor localization.

The addition of palmitate to synaptic proteins has also been shown to inhibit their localization at synapses. The AMPAR subunits, GluA1-4, all contain a conserved cysteine residue proximal to transmembrane domain 2 (TMD2) within the ion channel-forming domain221. Transfection of hippocampal neurons with GluA1 and 2 subunits (forming a GluA1/2 heteromer) in which the TMD2 cysteine of both subunits was mutated did not impact assembly of the receptor or basal synaptic efficacy, whereas overexpression of DHHC3, which was found 59

to palmitoylate the TMD2 site, reduced surface levels of GluA1 and GluA2221. The authors concluded that palmitoylation of the GluA1 and 2 TMD2 site causes their accumulation in Golgi compartments and that depalmitoylation is needed for their synaptic localization.

Palmitoylation can regulate the state of other post-translational modifications upon synaptic proteins. Cysteine 811 of GluA1 is localized to the membrane proximal region (MPR) of the C-tail and is also palmitoylated258. The MPR is bound by 4.1N, a neuron-enriched actin binding protein that links actin filaments to the plasma membrane, and also contains two serine residues that are phosphorylated by PKC141, 258. Inhibition of palmitoylation increased the binding of GluA1 to 4.1N and enhanced GluA1 surface levels258, suggesting that binding of 4.1N enhances GluA1 surface localization and palmitoylation functions to prevent this association.

Deletion of the entire MPR significantly reduced surface AMPAR levels, whereas deletion of the distal C-tail but with the MPR intact had no impact258, demonstrating that the sites of palmitoylation and 4.1N binding were critical. Mutation of the two PKC phosphorylation sites as well as a PKC inhibitor reduced AMPAR surface levels, which led the authors to conclude that

PKC phosphorylation facilitates the binding of 4.1N and stabilization of surface AMPARs258.

Therefore, C811 palmitoylation drives AMPAR internalization by inhibiting PKC-mediated phosphorylation of nearby serines, functioning to prevent 4.1N binding and AMPAR surface stabilization.

Palmitoylation of synaptic proteins can also drive their localization to mobile vesicles that deliver synaptic cargo during plasticity. AKAP79/150 is a signaling scaffold molecule important for the postsynaptic anchoring of PKA and the induction of LTP224. Palmitoylation of

AKAP79/150 is enhanced by an acute chemical LTP stimulation, resulting in its recruitment to recycling endosomes and delivery to the postsynaptic membrane, where it assembles a PKA 60

signaling complex224. Localization at recycling endosomes and the postsynaptic membrane is disrupted in a palmitoylation-deficient AKAP79/150, which also cannot maintain LTP-induced synapse strengthening224. Palmitoylation of AKAP79/150 is therefore required for enhanced synapse efficacy, and demonstrates how neurons use palmitoylation to rapidly respond to fluctuations in synaptic activity.

δ-catenin was recently identified as a synaptic substrate of palmitoylation211, and accumulating evidence points towards its importance in activity-dependent synapse plasticity178,

184, 206, 210. It would therefore be of great interest to investigate the role of palmitoylated δ-catenin in the activity-induced regulation of synapse structure and efficacy.

1.8 Rationale and Hypothesis

Dendritic spines are highly dynamic structures, and changes in the structure and molecular composition of an individual spine have been shown to correlate with changes in the efficacy of the synapse74. Changes in spine morphology require the trafficking of a number of proteins to the postsynaptic compartment, including actin regulating proteins132, 133, entire complexes of scaffold molecules134, 135, glutamate receptors37, and components of the cadherin- catenin adhesion complex173, 185. Post-translational modifications like palmitoylation have been shown to tightly control synaptic protein trafficking224, 242, and therefore may be critical components of activity-induced structural and functional synaptic enhancement.

Previous studies have established the importance of cadherins in synapse plasticity. LTP induction promotes the clustering of N-cadherin within spines and leads to the selective growth and stabilization of cadherin-associated spines over ones lacking N-cadherin173, 185. Moreover, postnatal ablation of N-cadherin in vivo blocks activity-dependent spine enlargement and the 61

formation of LTP174. Therefore, activity-dependent spine remodeling appears to require the formation of stable cadherin-adhesion complexes at the synaptic membrane.

δ-catenin has been shown to play an important role in the development and maturation of dendritic spines and neural connectivity178, 210. δ-catenin knockout mice display a number of abnormalities in hippocampal short-term and long-term synaptic plasticity, and exhibit severe impairments in cognitive function206. Given δ-catenin’s association with the juxtamembrane domain of cadherin, its involvement in regulating spine size and density, and its status as a newly identified substrate for palmitoylation211, palmitoylated δ-catenin may function as an activity- dependent regulator of cadherin-mediated synaptic plasticity.

I hypothesize that 1) dynamic palmitoylation of δ-catenin plays an important role in mediating synaptic plasticity through regulation of cadherin stability at the synapse, and 2) that a subset of DHHC proteins are involved in the palmitoylation of δ-catenin, and localized at synapses where they can rapidly palmitoylate δ-catenin following enhanced activity. In chapter

2, I investigate the role of palmitoylated δ-catenin in activity-induced synapse plasticity and identify the PAT responsible for its palmitoylation. In chapter 3, I examine the mechanism by which changes in synaptic activity translate to enhanced palmitoylation of δ-catenin by its PAT.

Chapter 2: Palmitoylation of δ-Catenin by DHHC5 Mediates

Activity-Induced Synapse Plasticity

Synaptic cadherin adhesion complexes are known to be key regulators of synapse plasticity. However, the molecular mechanisms that coordinate activity-induced modifications in cadherin localization and adhesion and the subsequent changes in synapse morphology and efficacy remain unknown. We demonstrate that the intracellular cadherin binding protein, δ- catenin, is transiently palmitoylated by DHHC5 after enhanced synaptic activity and that palmitoylation increases δ-catenin–cadherin interactions at synapses. Both the palmitoylation of

δ-catenin and its binding to cadherin are required for activity-induced stabilization of N-cadherin at synapses and the enlargement of postsynaptic spines, as well as the insertion of GluA1 and

GluA2 subunits into the synaptic membrane and the concomitant increase in miniature excitatory postsynaptic current amplitude. Notably, context-dependent fear conditioning in mice results in increased δ-catenin palmitoylation, as well as increased δ-catenin–cadherin associations at hippocampal synapses. Together these findings suggest a role for palmitoylated δ-catenin in coordinating activity-dependent changes in synaptic adhesion molecules, synapse structure and receptor localization that are involved in memory formation.

2.1 Introduction

Synaptic adhesion molecules, and in particular the classic cadherins, have been well studied with respect to their regulation of synapse function. Indeed, cadherins have been shown to be essential for activity-dependent structural remodeling, including enlargement of spines 173,

174 and increases in synapse number259. Activation of NMDA receptors enhances the clustering 63

and stabilization of cadherins at synapses188, and it is believed that this is essential for the establishment and maintenance of long-term potentiation (LTP)174, 260 as well as the formation of long-term contextual memory197.

The stability and clustering of cadherin at the cell membrane is primarily dictated by its interaction with intracellular catenins. Cadherin–β-catenin interactions are essential for intercellular cadherin interactions18, whereas the clustering of cadherins within the membrane is largely mediated by interactions with δ-catenin, p120-catenin (p120ctn), p0071, and ARVCF, that all bind to the cadherin juxtamembrane domain18. The cadherin juxtamembrane domain also contains multiple endocytic motifs, which regulate its endocytic turnover261, and is predominantly bound by δ-catenin in the brain198, suggesting a crucial role for δ-catenin in regulating cadherin membrane localization.

In mature neurons, δ-catenin is enriched in dendritic spines, where it can link cadherin to the actin cytoskeleton182, as well as to several postsynaptic scaffold molecules, including PSD-95

183 and GRIP184. The importance of δ-catenin in cognitive function and neural connectivity is supported by deficits in learning and memory, as well as synapse plasticity and morphology in δ- catenin-mutant mice178, 206, 210. Importantly, mutations of δ-catenin have been associated with severe impairments of cognitive function262, and may underlie the mental disabilities associated with Cri-du-Chat syndrome and schizophrenia209, 263.

δ-catenin has been identified as a substrate for protein palmitoylation211, a reversible post-translational modification involving the addition of palmitate to cysteine residues, mediated by a family of palmitoyl-acyl transferase (PAT) enzymes containing a conserved Asp-His-His-

Cys (DHHC) motif217. Recent work has demonstrated that palmitoylation of synaptic protein

substrates by particular DHHC proteins can be enhanced following neuronal activity224, 242, and can increase the trafficking of these proteins to the synapse224, 242.

Previous work in vitro clearly point to a role for cadherins in activity-mediated synapse plasticity. However, the molecular mechanisms that translate enhanced neuronal activity to changes in cadherin-based adhesion and synaptic remodeling remain poorly understood. Here we demonstrate that the dynamic palmitoylation of δ-catenin by DHHC5 can coordinate activity- dependent changes in synapse adhesion, structure, and the efficacy of synaptic transmission.

Together, this suggests a key role for δ-catenin in what are widely regarded as fundamental molecular processes underlying learning and memory.

2.2 Materials and Methods

2.2.1 Antibodies and cDNA Constructs

Primary antibodies: δ-catenin (1:500 for WB, 5µg for IP; BD Transduction Laboratories

611536), β-actin for WB (1:10000; Sigma A1978), N-cadherin for WB and IP (1:500, 5µg; BD transduction Laboratories 610921), p120-catenin for WB (1:500, BD Transduction Labs

610133), PSD-95 for IF (1:500; Affinity BioReagents MA1-046), PSD-95 for IP (5µg;

Calbiochem CP35), GFP for IP (10µL; Synaptic Systems 132 002), GFP for WB (1:1000; Roche

11814460001) and Streptavidin-HRP (1:5000; Thermo Scientific 21126), DHHC5 for WB

(1:1000; Sigma Prestige HPA014670), DHHC20 for WB (1:500; Abcam ab110478), c-Myc for

WB and IF (1:1000, 1:500; Sigma M4439), FLAG for WB (1:500; Sigma F1804), and HA for

WB and IF (1:1000, 1:500; Sigma H9658). Secondary antibodies: IgG-HRP for WB (1:5000;

BioRad mouse 170-6516, and rabbit 170-6515), and Alexa Fluor 568 and 633 for IF (1:1000;

Molecular Probes A-11004).

GFP-δ-catenin, and Myc and FLAG-tagged human DHHC plasmids 2-22 were kind gifts from Dr. Alaa El-Husseini (University of British Columbia, Vancouver, BC, Canada). N- cadherin-GFP was a kind gift from Dr. Chin-Yin Tai (Academica Sinica, Taipei, Taiwan), respectively. SEP-GluA1 and SEP-GluA2 were obtained from Addgene (24000 and 24001, respectively), and originally developed by Dr. Roberto Malinow (University of California, San

Diego, CA). TfR-mCherry-SEP was a kind gift from Dr. Michael Ehlers (Duke University,

Durham, NC). HA-tagged DHHC5 and DHHC8, control shRNA (shRNA-D5c), shRNA against

DHHC5 (shRNA-D5), and shRNA-resistant DHHC5 (DHHC5*) were kind gifts from Dr.

Richard Huganir (Johns Hopkins University, Baltimore, MD).

To knock down endogenous δ-catenin expression, a short hairpin RNA (shRNA) was generated using a previously described target sequence (Arikkath et al., 2008) 5'-

GATCCCCGCAACTATGTCGACTTCTATTCAAGAGATAGAAGTCGACATAGTTGCTTT

TT-3’. A control shRNA with 3 mutations was generated and used as a control. Both sequences were cloned into a pSuper vector using BglII/HindIII. shRNA-resistant δ-catenin was generated by introducing 3 silent mutations into the target sequence of δ-catenin, using a QuickChange kit

(Stratagene). To knockdown endogenous DHHC20 expression in cultured rat neurons, Silencer

Select siRNA s157138 (siRNA-D20; cat. 4390771) targeted against the rat zDHHC20 gene

(Accession no. NM_001039336.1) was obtained from Invitrogen. Silencer Select negative control no. 1 siRNA (siRNA-D20c; cat. 4390843) was also obtained from Invitrogen, and used as a negative control. Human Myc-tagged DHHC20 (Accession no. NM_153251.3) was not targeted by either of the siRNAs, and was used as a rescue construct in rat neurons.

The GFP-δ-catenin plasmid described above was used to generate all δ-catenin mutants.

δ-catenin 18CS mutant: a C-terminal fragment of mouse δ-catenin isoform 2 (Accession no.

NM_008729) containing all sites downstream of an internal EcoN1 site at nucleotide 1446, and in which all cysteine codons were changed to code for serines, was synthesized by Genscript

(Piscataway, NJ) and cloned into a pUC57 vector. The corresponding fragment in mouse GFP-δ- catenin was cut out using EcoN1 and Sma1 sites, and replaced with the synthesized fragment.

Subsequently, the two N-terminal cysteine residues upstream of the fragment were mutated to serines using the QuickChange Kit, generating the 18CS construct. The NTD11CS construct was generated by cutting 18CS at an Nhe1 site at 2310, and replacing with the corresponding sequence from GFP-δ-catenin. The CTD7CS construct was generated by cutting GFP-δ-catenin with Nhe1, and replacing with the sequence from the 18CS construct. The NTD16S construct was generated using the QuickChange kit to mutate the two most-C-terminal serine residues of the 18CS construct back to cysteines. GFP-δ-catenin K581M, C791S, C804S, C827S, C853S,

C931S, and C960-1S were generated using the QuickChange kit.

RFP-δ-catenin constructs were generated by PCR of monomeric RFP using primers

CCGGC ACCGGT ATG GCCTCCT CCGAGGAC and CCGGC AGATCT GGCGC

CGGTGGAGTG, and replacing the GFP in GFP-δ-catenin constructs using

AgeI and BglII sites. N-cadherin-RFP was generated by excising N-cadherin out of N-cadherin-

GFP using XbaI and EcoRI sites, and then inserted into a pcDNA3 vector. Subsequently, RFP was PCR’d using primers ccggc GAATTCATGGCCTCCT CCGAGGAC, and

CCGGCAGATCTGGCGCCGGTGGAGTG from monomeric RFP, and inserted into the same pcDNA3 vector containing N-cadherin, using XbaI and EcoRI sites.

2.2.2 Contextual Fear Conditioning

Male mice (6-9 weeks) were trained as per197, 264. Briefly, mice were placed in conditioning cage with a shock floor (Coulbourn Instruments) and habituated for 3 min. The

Conditioned group then received a 0.3mA shock for 5 seconds, whereas the Naïve group received no foot shock, and both groups were kept within the cage for an additional 2 min. 1 hour or 24 hours later, mice were placed back into the conditioned chamber for 5 min and the time spent freezing quantified using FreezeFrame software (Coulbourn Instruments). Mice were euthanized immediately following testing, and hippocampi and cortices isolated. Data was analyzed using Prism software (GraphPad software, La Jolla, CA). All animal procedures were performed in accordance with protocols approved by the Canadian Council on Animal Care.

2.2.3 Cell Cultures

Primary hippocampal neurons: Hippocampi from embryonic day 18 (E18) Sprague-

Dawley rats of either sex were prepared as previously described265, and plated at a density of 130 cells/mm2. Neurons were transfected with Lipofectamine 2000 (Invitrogen) at 9-10 DIV according to the manufacturer’s recommendations, and used for experiments at 14-16 DIV.

Neurons were nucleofected at 0 DIV using an Amaxa nucleofector kit (Lonza; VPG-1003) according to the manufacturer’s optimized protocol (no. 101), and used for biochemical experiments at either 6 DIV or 12 DIV, as indicated.

HEK Cells: HEK293T cells were transfected using polyethylenimine (PEI; Sigma) as previously described266, in a 3:1 ratio with the total plasmid DNA to be transfected, per

condition. 293T cells were transfected at 70-80% confluency, and incubated for 24-48 hours before harvesting for biochemistry.

2.2.4 Neuronal Activation

Activity was enhanced using a previously described chemical LTP protocol130, 267.

Briefly, maintenance media was replaced with a Hank’s Balanced Salt Solution (HBSS; Gibco) containing 137.93mM NaCl, 5.33mM KCl, 4.17mM NaHCO3, 0.441mM KH2PO4, 0.338mM

Na2HPO4, and supplemented with 1.5mM CaCl2, 30mM D-Glucose, 0.5µM Tetrodotoxin (TTX;

Alamone Labs), and 20µM Bicuculline methiodide (Fluka BioChemika), for 20 minutes. The media was then replaced with the same media supplemented with 200µM glycine, or 200µM glycine plus 50µM AP5 (Sigma), or media alone for 3 minutes. The solution was then replaced with fresh media for the indicated times prior to experimentation.

To induce homeostatic scaling, cells were incubated with maintenance media plus 1µM

TTX or 20µM Bicuculline for 48 hours. To block protein palmitoylation HBSS-based media described above were additionally supplemented with 50µM 2-Bromohexadecanoic acid (2-

Bromopalmitate; Sigma) where indicated.

2.2.5 Immunoblot Assay

Western blotting was performed as previously described19. Brain tissue, primary hippocampal neurons, and 293T cells were homogenized in an ice-cold lysis buffer containing

1% IGEPAL CA-630 (Sigma), 50mM Tris-HCl pH 7.5, 150mM NaCl, 10% Glycerol, and supplemented with phenylmethanesulfonyl fluoride solution (PMSF) and a protease inhibitor cocktail with EDTA (Roche). Proteins were visualized using enhanced chemiluminescence

(Pierce Biotechnology) on a Bio-Rad Versadoc 4000 (Bio-Rad Laboratories). Blots were quantified using Image J software.

Synaptosomal Fractionation: Crude synaptosomal fractions (P2) were prepared as previously described19. Hippocampi harvested from adult male mouse brains were homogenized in an ice- cold buffer containing 320mM sucrose, 4mM HEPES, and 1mM EGTA, then centrifuged at

1312 x g for 10 minutes at 4°C. The resulting supernatant (S1) was then centrifuged at 14,481 x g for 15 minutes at 4°C. The resulting pellet (P1) was then re-suspended in homogenization buffer and centrifuged at 17, 522 x g for 15 minutes at 4°C to obtain the crude synaptosomal fraction, P2. P2 was resuspended in lysis buffer as described above.

Immunoprecipitation: Immunoprecipitation (IP) assays were performed as previously described19. Biochemical labeling of all surface proteins using Sulfo-NHS-SS-biotin (Thermo

Scientific) and Neutravidin-agarose (Thermo Scientific) was performed as previously described268.

2.2.6 Palmitoylation Assay

Protein palmitoylation was determined using the biotin-BMCC variation of the ABE assay, exactly as previously described211, 224, 238, 269, 270

2.2.7 Immunocytochemistry

Immunohistochemistry experiments were performed as previously reported19.

2.2.8 Confocal Imaging

All neurons were imaged using an Olympus Fluoview 1000 confocal microscope

(60×/1.4 Oil Plan-Apochromat). Identical acquisition parameters were used for all cells across all separate cultures within an experiment. For live imaging of GFP-δ-catenin trafficking, dendritic spines, and SEP-tagged AMPA receptor puncta, a region of interest along a primary dendrite of a transfected cell within 100µm of, and emanating directly from the cell body was chosen and imaged before, and again 40-60 minutes following 3 min activity induction (see above). To image dendritic spines, a z-projected stack of 3 slices of 0.37µm each was acquired, and the same z-stack was imaged before and after activity induction.

2.2.9 Fluorescence Recovery After Photobleaching (FRAP)

Dendritic spines within 100µm of the cell body were imaged every 5 seconds for 5 min prior to and after photobleaching. Spines were identified using GFP or GFP-δ-catenin. A 1µm circular region of interest (ROI) was photobleached within a spine head using the tornado function within Fluoview software (Olympus). The fluorescence of N-cadherin-RFP in the photobleached ROI was quantified over time using Fluoview software. The recovery of fluorescence intensity (R) was determined by normalizing the intensity at a specific time (Ft), using the formula: R = (Ft-F0)/(Fi-F0) where (F0) represents the fluorescence at the time of photobleaching, and (Fi) the fluorescence before bleaching.

To account for passive bleaching, a 1µm ROI in an adjacent spine was tracked over time and quantified. The fluorescence intensity within the photobleached ROI was normalized to the control, non-photobleached ROI for each scan. Normalized fluorescence recovery data was then collected in Prism software (GraphPad), analyzed, and fit to a single exponential model, which

generated plateau values for the mean R value among each group of cells. Plateau values ± SEM were statistically compared in Prism software.

2.2.10 Image Analysis and Quantification

Confocal images for a particular experiment were subjectively thresholded using ImageJ software, and the same threshold was used for all images obtained for a single experiment, throughout the experimental analysis. Puncta were identified as a thresholded fluorescence cluster with an area between 0.05-3µm2. Puncta area and Integrated Density (the product of area and mean gray value) were then determined using ImageJ. An Image J colocalization plugin was used to assess colocalization between GFP-δ-catenin and N-cadherin-RFP or PSD-95

(http://rsb.info.nih.gov/ij/plugins/colocalization.html). Points of colocalization were defined as regions of >4 pixels in size, with >50 intensity ratio between the two channels.

Dendritic spine morphology was determined by manually tracing GFP-labeled protrusions between 0.5µm to 10µm in length. Spines were defined as any protrusion with a length:width ratio of <4, and filopodia as >4. Only the proximal 100µm of primary dendrites were analysed. The integrated density of GluA1, GluA2, and TfR-mCherry tagged with SEP was determined in Image J. Only SEP-fluorescent puncta that were present along dendrites before activity induction were considered for analysis. Statistical analysis was performed in Prism software. The confocal micrographs shown in Figures 2, 4, 6, 11, and S4 were re-sampled to 300 pixels per inch using Photoshop CS6 (Adobe Systems, Inc.) to aid in resolution. Confocal micrographs shown in Figs. 2, 4, 6, 10, and S4 were subjected to a 1 pixel Gaussian blur. No images were resampled for data analysis.

2.2.11 Electrophysiology

General recording conditions were as previously described271. To obtain whole-cell patch recordings, cells were perfused (2-4 mL/min) with an extracellular recording solution (ECS) containing the following: 125mM NaCl, 5mM KCl, 2mM CaCl2, 1mM MgCl2, 5mM HEPES,

33mM D-Glucose, 0.5µM TTX, 20µM Bicuculline, pH 7.3, 290 mOsm/L, and heated to 30°C.

Pipette Resistance was 3-6MΩ when filled with 130mM Cs methansulfonate, 5mM CsCl, 4mM

NaCl, 1mM MgCl2, 5mM EGTA, 10mM HEPES, 5mM QX-314, 0.5mM GTP, 10mM Na2- phosphocreatine, 5mM MgATP, 0.1mM Spermine, pH 7.4, and 290-295 mOsm/L. Series resistances (Rs<25 MΩ) were uncompensated, and did not differ between groups of cells (mean

15.92 ± 0.63MΩ). Cells were discarded if Rs changed >10%, or if Ih exceeded -50pA. mEPSCs were isolated using Clampfit 10 (detection threshold 5pA; typically 2-3rms noise), and non- unitary events were suppressed from all amplitude analysis, but not frequency analysis. Chemical

LTP was induced as described above, but with the following modifications: cells were incubated in ECS described above for 15 min at 37°C, then the media was exchanged for ECS containing

0mM MgCl2, and ± 200µM glycine as indicated, for 3 min at 37°C, then removed and exchanged with standard ECS. The cells were incubated in ECS for 30 min at 37°C, at which point they were returned to room temperature, and recordings began as described above.

2.2.12 Statistical Analysis

All data values are expressed as mean ± SEM. For all imaging and electrophysiological experiments, “n” refers to the number of cells used per condition, over at least 3 separate cultures, with the exception of the analysis performed in Fig. 2.7g,h, where “n” refers to the number of spines, and is specified within the figure legends. No statistical analysis was used to

pre-determine the sample sizes used for experiments, however our sample sizes are similar to those reported in193, 220, 241. Data collection and analysis were not performed blind to the conditions of the experiment, with the exception of the electrophysiology experiments performed in Fig. 2.8, where data analysis was performed by a different experimenter than that who performed the actual cellular recordings and who was completely blind to the experimental groups. No specific randomization was used for data collection or analysis, and cells and animals were assigned to particular experimental groups with absolutely no bias and by different experimenters. All data were collected in a spreadsheet, analyzed in Prism software and met the assumption of normality by using a D’Agostino and Pearson omnibus normality test in Prism, with the exception of all biochemical data in which the “n” values were too small, and so a normal distribution was assumed and not formally tested. Statistical significance was determined by student’s t-test, paired t-test, one or two-way ANOVA with post hoc tests using Prism, where indicated. Statistical significance was assumed when p<0.05. In all figures, *p<0.05, **p<0.01, and ***p<0.001, as determined in Prism software. All figures were generated using Illustrator

CS6 software (Adobe Systems, Inc.)

2.3 Results

2.3.1 Activity-Dependent Palmitoylation of δ-Catenin in Neurons

We determined whether palmitoylation of δ-catenin is regulated by activityusing the acyl-biotin exchange (ABE) assay, which exchanges palmitoyl modifications with biotin217, 220,

224. ABE assays yield similar results to experiments that use (3H)-palmitate to assay palmitoylation, and have quickly become the assay of choice for determining protein palmitoylation211, 217, 220. A typical control for the specificity of biotin labeling is the exclusion of 74

hydroxylamine (NH2OH) from the assay. Indeed, hydroxylamine is required to cleave palmitate from cysteine residues and therefore essential for the appropriate biotinylation of these cysteine residues269.

15-16 days in vitro (DIV) hippocampal neurons were treated with a glycine/bicuculline solution for 3 minutes. This method has previously been used to induce LTP in hippocampal slices272, and enhance structural remodeling of excitatory synapses in cultured neurons by activating synaptic NMDA receptors130, 267, and will hereafter be referred to as chemical LTP

(cLTP). δ-catenin palmitoylation was significantly enhanced 40 min following cLTP stimulation.

This increase was blocked by D(–)-2-amino-5-phosphonovaleric acid (AP5; 50µM), demonstrating NMDA receptor activity is required for palmitoylation of δ-catenin (Fig. 2.1a,b).

We next examined the time-course of activity-induced δ-catenin palmitoylation and compared it to that of a well-studied palmitoylated synaptic protein, PSD-95 242. Although no significant differences were observed in δ-catenin expression, δ-catenin palmitoylation was significantly increased 20 min following cLTP, peaked at 40 min, and returned to basal levels by 180 min

(Fig. 2.1c,d). PSD-95 also exhibited increased palmitoylation following cLTP, however palmitoylation levels increased more slowly, exhibiting a significant increase 40 min after cLTP that was maintained up to 180 min (Fig. 2.2a,b), indicating that PSD-95 is stably palmitoylated, as previously suggested220. Thus, palmitoylation of δ-catenin is transiently increased following synaptic activation, and represents a distinct cellular event relative to another palmitoylated synaptic protein, PSD-95.

Long-term, homeostatic scaling of neuronal networks in vitro has been shown to regulate the palmitoylation of several synaptic proteins273. To examine if palmitoylation of δ-catenin is

similarly modulated at a network level, hippocampal neurons were treated for 48 hours with tetrodotoxin (TTX; 1µM) to enhance synaptic network strength (Fig. 2.2c,d). There was a pronounced increase in δ-catenin palmitoylation 48 hours following TTX treatment in agreement with that with our observation that activity enhances δ-catenin palmitoylation.

Figure 2.1. δ-Catenin Palmitoylation and its Association with Synaptic N-Cadherin are Increased after Activity. (a–d) ABE chemistry and western blotting for streptavidin–horseradish peroxidase (HRP) to determine the palmitoylation of immunoprecipitated proteins. Omission of NH2OH controlled for nonspecific incorporation of biotin. (a,b) δ-catenin palmitoylation at 40 min after treatment with glycine or glycine + AP5 (50 µM; n = 4, P = 0.017, F2,9 = 6.59). IP, immunoprecipitation. (c,d) δ-catenin palmitoylation at 20, 40 and 180 min after glycine treatment (n = 4, P = 0.001, F4,10 = 19.01). (e–g) Images of 14 DIV hippocampal neurons transfected with GFP–δ-catenin and N-cadherin–RFP immediately before and 20, 40 and 180 min after glycine treatment with staining for PSD-95 post hoc. The areas of GFP–δ-catenin (GFP–δ-cat) and N-cadherin–RFP (Ncad-RFP) puncta (GFP–δ-catenin: n = 19, P < 0.001, F3,18 = 10.26; N-cadherin–RFP: n = 19, P = 0.476, F3,18 = 21.69) (e), colocalization of δ- catenin and N-cadherin (glycine: n = 19, P < 0.001, F3,18 = 4.171; glycine + AP5: n = 9, P = 0.755, F3,8 =

Figure 2.1 Continued: 0.399) (f) and colocalization of δ-catenin and N-cadherin with PSD-95 (g) in cells treated with glycine (Gly) or glycine + AP5 (n = 19, 9, respectively, P = 0.001, Student's t test). (h,i) Interactions between δ- catenin and N-cadherin 40 min after glycine treatment or the application of 2BP (50 µM) from either 0 to 40 min or 40 to 180 min after stimulation (n = 7, P = 0.007, F5,24 = 4.15). n values indicate the number of separate blots from separate cell cultures (b,d,i) or the number of cells from three separate cultures (e–g). All graphs show the mean ± s.e.m. *P < 0.05, **P < 0.01, one-way ANOVA and Tukey's test post hoc (b,d,i); *P < 0.05, **P < 0.01, repeated measures one-way ANOVA and Tukey's test post hoc (e,f); **P < 0.01, Student's t test (g).

Figure 2.2. Homeostatic Enhancement of Synaptic Strength Enhances the Palmitoylation of δ- Catenin. 14-16 DIV hippocampal neurons were (a,b) treated with a glycine solution for the indicated time, or (c,d) TTX for 48 h and lysates immunoprecipitated with the indicated antibodies. ABE chemistry and western blotting for streptavidin-HRP was used to determine palmitoylation of immunoprecipitated proteins. Omission of hydroxylamine (NH2OH) was used to control for non-specific incorporation of biotin. (a,b) In contrast to the time-course for δ-catenin palmitoylation shown in Fig. 2.1c,d, PSD-95 palmitoylation peaked 180 min after glycine treatment (n=3, p=0.012, F3,8=7.05). (c,d) δ-catenin palmitoylation was increased 48 h after treatment with TTX. (n=4, p=0.01, F2,9=15.17). Graphs represent mean ± SEM. The n value indicates the number of separate blots from separate cultures. *p<0.05, **p<0.01, one-way ANOVA, Tukey's test post hoc.

2.3.2 Palmitoylation of δ-Catenin Regulates its Binding to N-Cadherin

Since the addition of palmitate to substrate proteins increases their hydrophobicity and their trafficking to the membrane220, 242, we examined the effects of activity on the subcellular

distribution of δ-catenin. Prior to stimulation, GFP-δ-catenin was diffusely distributed along dendrites (Fig. 2.3a). However, 40 minutes after cLTP, GFP-δ-catenin became increasingly clustered, as evidenced by a decrease in the area of δ-catenin puncta (Fig. 2.1 and 2.3a) with a concomitant increase in fluorescence intensity (Fig. 2.3b). This was abolished in cells treated with glycine plus AP5 (Fig. 2.3b,c). In contrast, no change was observed in the area or intensity of N-cadherin puncta (Fig. 2.1e, 2.3a-c). The colocalization of δ-catenin and N-cadherin also significantly increased 20-180 min after cLTP (Fig. 2.1f). Post hoc immunostaining demonstrated an increased colocalization of N-cadherin/δ-catenin clusters with PSD-95 indicating that activity enhances the recruitment of δ-catenin to synaptically localized cadherin clusters (Fig. 2.1g, 2.3e).

We next determined whether the time course for δ-catenin/N-cadherin interactions was similar to that of δ-catenin palmitoylation and δ-catenin/N-cadherin colocalization. The association of δ-catenin with N-cadherin was increased 20 mins after stimulation and was maintained up to 180 mins after stimulation (Fig. 2.3i,j). This was in sharp contrast to the time course of δ-catenin palmitoylation, which decreased to basal levels 180 mins after stimulation

(Fig.2.1c,d).

These results suggested that a transient increase in δ-catenin palmitoylation is required for activity-dependent recruitment of δ-catenin to cadherin clusters, but is not essential to maintain this interaction. To test this, we examined the association of δ-catenin and N-cadherin in the presence of the global palmitoylation blocker, 2-bromopalmitate (2BP; 50µM) at different time points. When cells were treated with 2BP 0-40 min following cLTP, the activity-induced increase in δ-catenin/N-cadherin interactions was abolished (Fig. 2.1h,i). However, when cells

were treated with 2BP from 40-180 min after cLTP treatment (a time frame during which δ- catenin is highly palmitoylated), the activity-induced increase in δ-catenin/N-cadherin interactions was maintained (Fig. 2.1h,i). This demonstrates that δ-catenin palmitoylation is not required for maintaining its interaction with N-cadherin.

To determine whether activity enhances the recruitment of δ-catenin specifically to surface cadherin, we isolated surface proteins through biotinylation and immunoprecipitation with Sulfo-NHS-SS-biotin at various time points and observed an activity-dependent increase in

δ-catenin co-immunoprecipitated with the surface fraction (Fig. 2.3k,l). In contrast, there was no significant increase in the recruitment of p120ctn to the surface fraction (Fig. 2k,l). Together, these data demonstrate that activity-induced palmitoylation of δ-catenin increases its recruitment to surface N-cadherin at postsynaptic membranes, but that this association can be maintained in the absence of δ-catenin palmitoylation.

Figure 2.3. Activity Enhances the Clustering of δ-Catenin with Surface N-Cadherin at the Postsynaptic Membrane. (a-d) Confocal images of 14 DIV neurons transfected with GFP-δ-catenin and N-cadherin-RFP before and 20, 40 and 180 min after (a,b,e) glycine or (c,d) glycine+AP5. (a,b) Glycine enhanced the IntDen of δ-catenin (GFP-δ-catenin: n=19, p<0.001, F3,18=15.79; N-cadherin-RFP: n=19, p=0.953, F3,18=60.91). Glycine+AP5 did not affect (c) puncta area (GFP-δ-catenin: n=9, p=0.446, F3,8=4.74; N-cadherin-RFP: n=9, p=0.105, F3,8=41.53) or (d) IntDen (GFP-δ-catenin: n=9, p=0.429, F3,8=41.29; N-cadherin-RFP: n=9, p=0.631, F3,8=76.03). Scale bar = 5µm.(e) Activity increased th 81

Figure 2.3 Continued: colocalization of GFP-δ-catenin with PSD-95, 40 min after glycine (n=21 cells, 3 separate cultures; p=0.003, student's t-test). (g-l) 14-16 DIV hippocampal neurons were incubated with a glycine solution for the indicated amount of time and lysates immunoprecipitated with the indicated antibodies. (g,h) 40 min after glycine treatment, the amount of N-cadherin associated with δ-catenin was increased (n=3, p=0.034). (i,j) δ-catenin/N-cadherin interactions were enhanced 40 min following glycine treatment and maintained for 180 min (n=4, p=0.041, F3,12=7.61). (h,l) Following incubation with glycine for the indicated times, neurons were biotinylated with Sulfo-NHS-SS-biotin, and lysates immunoprecipitated with neutravidin-coated beads to isolate all surface proteins. There was an increase in the amount of δ- catenin associated with surface proteins 40 and 180 min after glycine treatment, (n=3, p=0.008, F4,10=6.35). In contrast, there was no change in the association of p120ctn with the surface fraction (n=3, p=0.958, F4,10=0.151) or the amount of N-cadherin at the membrane (n=3, p=0.776, F4,10=0.441). The n values indicate (a-e) the number of cells from 3 separate cultures, and (g-l) indicate the number of separate blots from separate cultures. All graphs represents mean ± SEM. (a-d) **p<0.01, repeated measures one-way ANOVA, Tukey's test post hoc. (e,h) *p<0.05, **p<0.01, student's t-test. (j,l)*p<0.05,; one-way ANOVA, Tukey's test post hoc.

2.3.3 Palmitoylated and Cadherin-Binding Residues of δ-Catenin To determine the role of palmitoylated δ-catenin at the synapse, we sought to generate palmitoylation-defective mutants of δ-catenin, which first necessitated the identification of palmitoylated cysteine residue(s). We generated a number of δ-catenin mutants in which cysteines were mutated to serines and assayed for palmitoylation in HEK293T cells (Fig. 2.4a).

δ-catenin palmitoylation was abolished when all 18 cysteines were mutated to serines (18CS) and when the 7 C-terminal cysteines were mutated to serines (CTD7CS), but not when the 11 N- terminal cysteines were mutated to serines (NTD11CS), suggesting that one or more of the 7 C- terminal cysteine residues are the sites for palmitoylation (Fig. 2.4a-c).

We next mutated each of the 7 C-terminal cysteine residues to serines (GFP-δ-catenin

C791S, C804S, C827S, and C853S, C906S, C960-1S), and assayed their palmitoylation compared to WT and CTD7CS δ-catenin. Point mutants C791S, C804S, C827S, C853S, and

C906S were palmitoylated as robustly as WT, whereas palmitoylation of C960-1S was significantly reduced compared to WT (Fig. 2.4b,c). This strongly suggested that cysteines 960

and 961, which are also highly predicted to be palmitoylated (CSS-Palm 3.0239), are the sites of palmitoylation. To verify this we generated a δ-catenin mutant in which all cysteines, with the exception of cysteines 960 and 961, were mutated to serines (NTD16CS). Palmitoylation of

NTD16CS was similar to that of WT, demonstrating that cysteines 960 and 961 are necessary and sufficient for the palmitoylation of δ-catenin (Fig. 2.4b,c).

To determine whether palmitoylation of δ-catenin has a role beyond its regulation of δ- catenin/cadherin interactions, we sought to generate a δ-catenin mutant that could not bind to cadherin. Lysine residue 401 on p120ctn has been shown to be required for p120ctn’s association with cadherin 193 and we mutated the corresponding lysine on δ-catenin (lysine 581) to methionine (GFP-δ-catenin K581M). Immunoprecipitation assays in 293T cells showed minimal association between δ-catenin K581M and N-cadherin compared to WT δ-catenin (Fig. 2.4d,e).

Although the palmitoylation-deficient C960-1S mutant exhibited a significant reduction in its ability to bind cadherin compared to WT δ-catenin, it still exhibited significantly more binding compared to the K581M mutant (Fig. 2.4d,e). The addition of palmitate to δ-catenin therefore functions to target δ-catenin to the membrane where it associates with N-cadherin, but is not absolutely required for binding.

Finally, we also generated short hairpin RNAs against δ-catenin (shRNA) and a control mismatch (shRNA-c) using previously described target sequences274 and validated them in 293T cells (Fig. 2.4f,g).

Figure 2.4. Palmitoylation of δ-Catenin Occurs at Cysteines 960 and 961, and Requires Lysine 581 for Binding to N-Cadherin. (a) Schematic depiction of δ-catenin constructs N-terminally tagged with GFP (not shown here) and illustrating the approximate localization of all 18 cysteine residues (white circles), as well as cysteine-to-serine mutations (black circles). K581 (K) is located within the third Armadillo repeat domain (gray boxes) and the PDZ-binding motif (PDZb) at the C terminus. (b,c) 84

Figure 2.4 Continued: Palmitoylation, as determined by ABE labeling and probing with antibodies to streptavidin, of the specified GFP–δ-catenin constructs after transfection into HEK293T cells for 36–48 h and then immunoprecipitation of the lysates with antibodies to GFP (n = 3–5 blots from separate cultures, P < 0.001, F11,41 = 10.18). (d,e) Immunoprecipitation, using antibodies to N-cadherin, of the lysates of HEK293T cells that were transfected with the indicated GFP-tagged δ-catenin constructs plus N- cadherin–RFP (n = 3, P < 0.001, F2,6 = 32.87). (f,g) Immunoprecipitation of HEK293T cells that were transfected with a control shRNA (shRNA-c) or δ-catenin shRNA plus the indicated GFP–δ-catenin constructs (asterisks beside the protein names, for example, WT*, denote shRNA resistance) (n = 4, P = 0.005, F4,13 = 6.41). n numbers indicate the number of separate blots from separate cell cultures. All graphs show the mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001, one-way ANOVA and Tukey's test post hoc.

2.3.4 Palmitoylated δ-Catenin Stabilizes N-cadherin at Synapses

We next determined the role of δ-catenin palmitoylation in regulating cadherin stability within postsynaptic spine heads using fluorescence recovery after photobleaching (FRAP).

Hippocampal neurons were transfected at 10 DIV with N-cadherin-RFP to visualize cadherin, δ- catenin shRNA , and shRNA-resistant GFP-tagged δ-catenin WT, K581M, or C960-1S constructs. FRAP assays were done 5-6 days after transfection, as previously reported to ensure knockdown of endogenous δ-catenin274. Cells were imaged 40-60 minutes after cLTP, or treatment with a control buffer lacking glycine. We observed no differences in the initial fluorescence intensity of N-cadherin clusters in all groups analyzed (Fig. 2.5a).

N-cadherin-RFP clusters within dendritic spines were identified and a region of interest

(ROI) of 1µm diameter was photobleached using a 405nm laser. The fluorescence recovery of N- cadherin-RFP within the photobleached ROI was determined over 5 mins of time-lapse imaging, and normalized to a control ROI in an adjacent spine (Fig. 2.6a,b). In control neurons expressing

GFP and shRNA-c, the fluorescence recovery of N-cadherin-RFP plateaued at 50.6 ± 8.0%

(mean ± SEM) 5 mins after photobleaching (Fig. 2.6a,c,i), consistent with previous reports186, 275.

Activity has been shown to stabilize N-cadherin at the synapse188 and accordingly, glycine

treatment dramatically reduced the fluorescence recovery of N-cadherin-RFP to 14.9 ± 7.4%.

Treatment with glycine plus AP5 (50µM) abolished activity-induced stabilization of cadherin, resulting in fluorescence recovery similar to that of untreated cells (53.6 ± 4.8%; Fig. 2.6a,c,i).

Overexpression of WT δ-catenin increased the stability of N-cadherin under basal conditions and occluded further activity-dependent stabilization of N-cadherin (Fig. 2.6d,i). In contrast, overexpression of K581M and C960-1S did not impact the stability of N-cadherin under basal conditions suggesting that δ-catenin’s palmitoylation and binding to cadherin are required for N-cadherin stabilization (Fig. 2.5b,c). Acute knockdown of δ-catenin using shRNA substantially decreased cadherin stability under basal conditions and abolished activity- dependent stabilization of N-cadherin (Fig. 2.6e,i). Introducing shRNA-resistant δ-catenin

(WT*) to shRNA-expressing cells rescued the shRNA phenotype, demonstrating that this was not due to off-target effects (Fig. 2.6f,i). Our data showing a requirement for δ-catenin for the stabilization of N-cadherin at synapses is in accord with previous work showing that ablation of

δ-catenin reduces N-cadherin at the synapse178, 201, 206.

To examine whether δ-catenin binding to cadherin is essential for activity-induced cadherin stabilization, we knocked down δ-catenin and expressed the cadherin-binding mutant,

K581M*. Similarly, to examine whether palmitoylation of δ-catenin is important for activity- induced cadherin stabilization, we knocked down δ-catenin and expressed the palmitoylation mutant, C960-1S* GFP-δ-catenin. Abolishing δ-catenin’s association with cadherin (Fig. 2.6g,i) or δ-catenin palmitoylation (Fig. 2.6h,i) significantly reduced the stability of N-cadherin under basal conditions and abolished activity-dependent stabilization of N-cadherin (Fig. 2.6g,h,i).

Together, this demonstrates that the palmitoylation of δ-catenin, and its binding to cadherin are

required for the stabilization of N-cadherin at synapses both under basal conditions and following enhanced neuronal activity.

Figure 2.5. Expression of δ-Catenin Constructs does not Impact Basal N-Cadherin Levels within Spine Heads, nor does Expression of K581M and C960-1S Impact N-Cadherin Stability. (a-c) Hippocampal neurons were transfected at 10 DIV with the indicated constructs (*denotes shRNA- resistance) and fluorescence recovery after photoleaching (FRAP) determined at 15-16 DIV. (a) Fluorescence intensity of N-cadherin within spine heads before photobleaching was not impacted by expression of δ-catenin shRNA or δ-catenin constructs, nor by glycine treatment (n values indicated in Fig. 4, p=0.806, F12,140=0.639; one-way ANOVA). (b,c) Normalized fluorescence recovery of N-cadherin- RFP in cells expressing the indicated shRNAs and δ-catenin constructs. The dashed purple line represents the plateau for fluorescence recovery in control, untreated cells (Fig. 2.6c). (b) Points with error bars represent mean ± SEM, solid lines represent single exponential fit. Statistical tests compare plateau values from exponential fits ± SEM. The number of neurons used in each condition is indicated below, and represent cells obtained from at least 3 separate cultures: shRNA-c (n=11), shRNA-c + K581M (n=9), and shRNA-c + C960-1S (n=9); p=0.034, F2,26=3.85, one-way ANOVA, Tukey's test post hoc (no Tukey tests were significant). (c) The mobile fraction of N-cadherin-RFP (fluorescence within the ROI at the 5 min time point, normalized for photobleaching; mean ± SEM; p=0.897, F2,26=0.109; one-way ANOVA).

Figure 2.6. δ-Catenin Palmitoylation is Required for Activity-Induced Stabilization of N-Cadherin within Dendritic Spine Heads. (a) Fluorescence within a photobleached ROI (red circles) normalized to non-photobleached ROI in adjacent spines (blue circle) (cells were initially photobleached at 0 s, white asterisks, within a 1-µm diameter ROI). Scale bars, 1 µm. (b) Fluorescence recovery within photobleached and non-photobleached ROIs (from the top two rows in a) and normalization for bleaching. (c–h) Normalized fluorescence recovery of N-cadherin–RFP. Dotted lines represent the plateau for fluorescence recovery in control cells. Points with error bars represent the mean ± s.e.m., and 88

Figure 2.6 Continued: solid lines represent a single exponential fit. Statistical tests compare the plateau values from exponential fits ± s.e.m. Neurons were obtained from ≥3 separate cultures. n = 10 cells, −glycine; n = 15 cells, +glycine; n = 5 cells, glycine + AP5; P = 0.003, F2,27 = 7.15, one-way ANOVA (c); n = 11 cells, −glycine; n = 5 cells, +glycine; P = 0.991, Student's t test (d); n = 18 cells, −glycine; n = 8 cells, +glycine; P = 0.099, Student's t test (e); n = 18 cells, −glycine; n = 9 cells, +glycine; P = 0.003, Student's t test (f); n = 12 cells, −glycine; n = 9 cells, +glycine; P = 0.223, Student's t test (g); n = 25 cells, −glycine; n = 8 cells, +glycine; P = 0.92, Student's t test (h). (i) The mobile fraction of N-cadherin–RFP (fluorescence within the ROI at the 5-min time point normalized for photobleaching; mean ± s.e.m.; P < 0.001, F12,140 = 8.03, one-way ANOVA). *P < 0.05, one-way ANOVA and Tukey's test post hoc relative to control cells expressing shRNA-c in the absence of glycine; ##P < 0.01, Tukey's test post hoc relative to same transfection condition in the absence of glycine.

2.3.5 Palmitoylated δ-Catenin Regulates Spine Remodeling

As palmitoylated δ-catenin stabilizes N-cadherin within spines, we next investigated if it is important for activity-dependent spine remodeling. 10 DIV neurons were transfected with

GFP, shRNAs, and the indicated δ-catenin constructs. 5-6 days later, all spiny protrusions along primary dendrites were imaged before and 40-60 minutes after cLTP. Neurons expressing GFP and shRNA-c exhibited spine head width, length, and density consistent with previous descriptions of spine morphology in cultured hippocampal neurons186, 193 (Fig. 2.7a,g-i). cLTP treatment dramatically enhanced the width and density of protrusions (Fig. 2.7a,g,i), confirming previous reports of activity-mediated changes in spine morphology in culture and hippocampal slices130, 133. Overexpression of δ-catenin increased the overall length and total density of protrusions, consistent with previous reports178, 181. We also observed an increase in protrusion head width and the density of mature spines, with a reduction in filopodia density compared to control cells (Fig. 2.7b,g-i). None of these parameters were changed following cLTP.

Interestingly, knockdown of δ-catenin resulted in protrusions with a filopodial appearance, including a decrease in protrusion width, an increase in protrusion length and an overall increase in the proportion of filopodia compared to spines, consistent with previous

reports178, 182. Notably, cLTP did not alter any of these parameters (Fig. 2.7c,g-i). WT* δ-catenin rescued the phenotype observed in shRNA expressing cells, demonstrating that these morphological changes were not the result of off-target effects (Fig. 2.7d,g,h). Rescue with WT*

δ-catenin also increased basal spine density consistent with that observed following δ-catenin overexpression. Expression of K581M* (Fig. 2.7e) or C960-1S* (Fig. 2.7f) in a δ-catenin knockdown background resulted in a phenotype similar to that of δ-catenin knockdown (Fig.

2.7c), with the exception of spine/protrusion density, which was similar to control (Fig. 2.7i). δ- catenin knockdown has previously been shown to increase the density of spiny protrusions via the PDZ binding motif of δ-catenin, and not through cadherin binding 178, corroborating well with our data. Strikingly, cLTP failed to alter protrusion morphology or density (Fig. 2.7e,f,g-i).

Together, this demonstrates that palmitoylation of δ-catenin and its binding to cadherin are required for activity-dependent changes in spine morphology, but not density.

Figure 2.7. δ-Catenin Palmitoylation is Required for Activity-Induced Spine Remodeling. (a–h) Results from primary dendrites imaged before and 40–60 min after glycine treatment. n = 12 cells, 803 spines (a); n = 7 cells, 396 spines (b); n = 10 cells, 529 spines (c); n = 10 cells, 606 spines (d); n = 8 cells, 352 spines (e); n = 10 cells, 535 spines (f). Open arrowheads in a and d indicate remodeling of pre- existing protrusions, and filled arrowheads indicate the appearance of new protrusions. Scale bars, 2 µm. (g) Protrusion head width before and after glycine treatment (P < 0.001, F5,6430 = 127.57 (between groups effect); P < 0.001, F1,6430 = 43.13 (glycine treatment effect)). (h) Length of protrusions before and after glycine treatment (P < 0.001, F5,6430 = 213.45 (between groups effect); P < 0.001, F1,6430 = 11.57 (glycine treatment effect)). (i) The density of total protrusions (with the mean represented by the crosshatched bars plus the solid bars; the top error bars represent the total protrusion s.e.m.; P = 0.005, F5,102 = 2.23 (between groups effect); P = 0.04, F1,102 = 4.32 (glycine treatment effect); two-way ANOVA), filopodia (solid bars) and spines (crosshatched bars ± s.e.m.; P < 0.001, F5,102 = 65.05 (between groups effect); P < 0.001, F5,102 = 18.71 (glycine treatment effect); two-way ANOVA) before and after glycine treatment. Cells were obtained from ≥3 separate cultures. Graphs represent the mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA and Bonferroni's test post hoc relative to shRNA-c before glycine treatment; ###P < 0.001, Bonferonni's test post hoc relative to the same condition before glycine (g,h). In i, black asterisks above the bars compare total protrusions relative to shRNA-c cells before glycine treatment, white asterisks within the crosshatched bars compare spines relative to shRNA-c before glycine treatment, # above the bars compare total protrusions within groups before and after glycine treatment, and # within the crosshatched bars compare spines within groups before and after glycine treatment. #,*P < 0.05, ##,**P < 0.01, ***P < 0.001, two-way ANOVA and Bonferonni's test post hoc.

2.3.6 Palmitoylated δ-Catenin Mediates Changes in Synapse Efficacy

δ-catenin can associate with several postsynaptic scaffold molecules and AMPA receptor

(AMPAR) binding proteins in a PDZ-dependent manner 183, 184, 276. We therefore hypothesized that activity-induced palmitoylation of δ-catenin and the subsequent surface stabilization of cadherin in spine heads may be important for the insertion and stabilization of AMPARs at pre- existing synapses. To visualize surface AMPARs, GluA1 or GluA2 subunits were tagged with super-ecliptic pHluorin (SEP-GluA1, or SEP-GluA2, respectively). Neurons were transfected at

10 DIV with SEP-GluA1 or SEP-GluA2, shRNAs, and either RFP or the indicated RFP-δ- catenin constructs, and imaged before and after cLTP at 15-16 DIV. Clusters of SEP-GluA1

(Fig. 2.9a) and GluA2 (not shown) localized to spine heads under basal conditions in neurons expressing shRNA-c plus RFP. Under basal conditions the integrated density (intDen; product of mean grey value and area) of GluA1surface clusters were similar for all conditions examined

(p=0.082, F5,71=2.492; one-way ANOVA) (Fig. 2.9a). The intDen of GluA2 surface clusters was increased in δ-catenin overexpressing cells compared to control (1.48 ± 0.13 fold; p<0.05), in agreement with our observation of enlarged spine heads among these cells (Fig. 2.7b,g,h), but was similar among all other groups (p=0.039, F5,69=2.053; one-way ANOVA).

40-60 min after cLTP there was a significant increase in the intDen of GluA1 (Fig. 2.8a, and 2.9a) and GluA2 clusters (Fig. 2.9b), compared to the same clusters before activity, consistent with previous studies277, 278. This was abolished in the presence of AP5 (50µM; Fig.

2.8a, and 2.9b). Interestingly, both overexpression and knockdown of δ-catenin abolished activity-induced insertion of GluA1 and GluA2 into the synaptic membrane (Fig. 2.8a, and

2.9a,b). WT* δ-catenin, but not the K581M* and C960-1S* δ-catenin mutants, rescued the

shRNA phenotype (Fig. 2.8a, and 2.9a,b), indicating that δ-catenin palmitoylation and binding to cadherin are essential for activity-dependent insertion of AMPARs into the synaptic membrane.

δ-catenin overexpression (shRNAc + WT-δ-catenin) enhanced the colocalization of δ- catenin and SEP-tagged GluA1 and GluA2 subunits under basal conditions compared to cells expressing shRNA + WT-δ-catenin (wildtype rescue, control) (Fig. 2.8b, and 2.9c). This suggests that overexpressing δ-catenin increases its localization at synapses and accounts for the observed increase in surface GluA2 levels under basal conditions. In contrast, the localization of the cadherin-binding mutant (shRNA + K581M) at synapses was significantly decreased but not abolished (Fig. 2.8b, and 2.9c), indicating that other motifs including its PDZ-binding domain may also be involved. The δ-catenin palmitoylation mutant (shRNA + C960-1S) exhibited similar synaptic localization as controls (shRNA+WT-δ-catenin). Since this mutation does not abolish binding to cadherin entirely, the C960-1S mutant may localize to synapses through cadherin interactions as well as through other binding motifs including its PDZ-binding domain.

Following cLTP, we observed an increase in the recruitment of δ-catenin to synapses in control (shRNA + WT-δ-catenin) and δ-catenin overexpressing (shRNA-c + WT-δ-catenin) cells

(Fig. 2.8b, and 2.9c). In contrast, cLTP did not increase the recruitment of cadherin-binding deficient or palmitoylation-deficient δ-catenin to synapses.

To confirm that δ-catenin specifically regulates the membrane insertion and stabilization of AMPARs, and not overall exocytosis, we examined turnover of the transferrin receptor tagged with SEP and mCherry (TfR-mCherry-SEP)277. Activity enhanced TfR-mCherry-SEP florescence in control cells, as well as cells overexpressing δ-catenin, and knockdown cells (Fig.

2.9d), demonstrating that δ-catenin plays a specific role in regulating the surface insertion and stabilization of AMPAR cargo, and does not impact overall cellular exocytosis or trafficking.

We next determined whether δ-catenin palmitoylation is required for glycine-mediated enhancement of miniature excitatory post-synaptic currents (mEPSCs)130, 267. Neurons were transfected at 10 DIV with the indicated shRNAs and GFP or GFP-tagged δ-catenin constructs. 5 days later, we obtained whole-cell voltage clamp recordings, at 40 min after cLTP. Under basal conditions, we observed increased mEPSC amplitude and frequency in δ-catenin overexpressing cells relative to control (Fig. 2.8c-e), consistent with our observation of increased spine density

(Fig. 2.7i) and surface GluA2-AMPARs in these cells276. There were no significant differences in basal mEPSC amplitude and frequency between any of the other groups relative to control cells

(Fig. 2.8c-e). cLTP significantly increased the amplitude and frequency of mEPSCs in cells expressing shRNA-c (Fig. 2.8c,f,g). Overexpression and knockdown of δ-catenin abolished this effect, which was restored in knockdown neurons expressing δ-catenin WT (Fig. 2.8c,f,g).

Notably, expression of palmitoylation-deficient δ-catenin C960-1S did not rescue the glycine- induced increase in mEPSC amplitude and frequency (Fig. 2.8c,f,g). These findings demonstrate that palmitoylated δ-catenin is required for activity-induced enhancement of mEPSC amplitude and frequency.

Figure 2.8. δ-Catenin Palmitoylation is Required for Activity-Induced AMPA Receptor Insertion and Changes in mEPSCs. (a,b) Results from cells that were imaged before and 40–60 min after the indicated treatments. (a) IntDen of pre-existing SEP-GluA1 puncta normalized to the same puncta before treatment (dashed line). The cell numbers and P values from paired t tests were as follows: shRNA-c (n = 22, P < 0.001), shRNA-c + AP5 (n = 9, P = 0.91), shRNA-c + WT (n = 9, P = 0.865), shRNA (n = 10, P

Figure 2.8 Continued: = 0.942), shRNA + WT* (n = 14, P < 0.001), shRNA + K581M* (n = 13, P = 0.133) and shRNA + C960- 1S* (n = 9, P = 0.176). (b) Percentage δ-catenin and GluA1 colocalization (P < 0.001, F3,49 = 8.25, one- way ANOVA). # denote significance among the groups before treatment relative to shRNA + WT*, asterisks denote significance within groups before and after glycine: shRNA-c + WT (n = 12, P < 0.001), shRNA + WT* (n = 15, P < 0.001), shRNA + K581M* (n = 12, P = 0.552) and shRNA + C960-1S* (n = 14, P = 0.494). (c–g) Whole-cell recordings (held at −65 mV) 40 min after the indicated treatments. (c) Representative mEPSC traces. (d) Basal mEPSC amplitudes (P = 0.031, F4,45 = 2.928, one-way ANOVA). (e) Basal mEPSC frequencies (P < 0.001, F4,45 = 6.056, one-way ANOVA; NS, not significant). (f,g) Percentage mEPSC amplitude and frequency 40 min after glycine treatment normalized to the mean in untreated cells (dashed line). The n values for the −glycine and +glycine groups, respectively, and the P values from Student's t tests for amplitude and frequency, respectively, were as follows: shRNA-c (n = 17, 23; P < 0.001, 0.048), shRNA-c + WT (n = 9, 8; P = 0.857, 0.741), shRNA (n = 9, 7; P = 0.076, 0.440), shRNA + WT* (n = 7, 10; P = 0.016, 0.048), shRNA + C960-1S* (n = 8, 7; P = 0.801, 0.209). Graphs represent the mean ± s.e.m. ***P < 0.001, paired t test; #P < 0.05, one-way ANOVA and Tukey's test post hoc (a,b); #P < 0.05, ###P < 0.001, one-way ANOVA and Tukey's test post hoc; *P < 0.05, ***P < 0.001, Student's t test; NS, not statistically significant (d–g).

Figure 2.9. Activity-Induced Insertion of AMPA Receptors Requires Cadherin-Binding by Palmitoylated δ-Catenin. (a) Confocal images of 15-16 DIV primary hippocampal neurons transfected at 10 DIV with SEP-GluA1 plus the indicated shRNA and RFP or RFP-δ-catenin constructs (*denotes shRNA resistance). SEP-fluorescent puncta are pseudocolored in heat maps. Cells were imaged before and 40-60 min after glycine treatment. Scale bar = 5µm. (b) IntDen of pre-existing SEP-GluA2 puncta following treatment with glycine or glycine+AP5, normalized to the mean IntDen of the same puncta before treatment (dashed line). n denotes the number of cells, and p values from paired t-tests as follows: shRNA-c (n=17, p<0.001), shRNA-c+AP5 (n=11, p=0.945), shRNA-c+WT (n=13, p=0.289), shRNA (n=9, p=0.158), shRNA+WT* (n=10, p=0.006), shRNA+K581M* (n=14, p=0.314), and shRNA+C960- 1S* (n=11, p=0.056). (c) Percent colocalization of δ-catenin/GluA2 before, and 40 min after glycine treatment (p<0.001, F3,45=13.07; one-way ANOVA with Tukey's test post hoc). Crosshatches denote significance among “before” groups relative to shRNA+WT*, asterisks denote significance within groups 97

Figure 2.9 Continued: before and after glycine. n denotes the number of cells, and p values paired t-tests within groups as follows: shRNA-c+WT (n=12, p=0.006), shRNA+WT* (n=13, p<0.001), shRNA+K581M* (n=12, p=0.763), and shRNA+C960-1S* (n=12, p=0.273). (d) IntDen of SEP fluorescence in cells expressing SEP-TfR-mCherry normalized to pre-existing SEP-fluorescent puncta before treatment (shRNA-c: n=7, p<0.001; shRNA-c+WT: n=7, p=0.002; shRNA: n=6, p=0.01). n=cells from at least 3 separate cultures. Graphs represent mean ± SEM. *p<0.05, **p<0.01, ***p<0.001; paired t-test. #p<0.05, ##p<0.01, one- way ANOVA with Tukey's test post hoc.

2.3.7 δ-Catenin Palmitoylation Increases after Acquisition of Contextual Fear Memory

We next investigated whether palmitoylated δ-catenin is involved in learning and memory using a hippocampal-dependent, contextual fear-conditioning paradigm197 (Fig. 2.10a).

Conditioned mice exhibited enhanced freezing 1 h and 24 h after reintroduction to the context where they received a single foot shock (0.3mA, 5sec), demonstrating acquisition of contextual memory (Fig. 2.10a). Immediately following testing, mice were sacrificed and hippocampal and cortical tissue isolated for biochemical analyses. Palmitoylation of δ-catenin in the hippocampus was significantly increased 1 h after training, but returned to baseline 24 h after training (Fig.

2.10b,c), demonstrating transient palmitoylation similar that observed in hippocampal cultures following cLTP (Fig. 2.1c,d). In contrast, δ-catenin palmitoylation was unchanged in the cortex following fear conditioning (Fig. 2.10d,e), demonstrating specificity for δ-catenin palmitoylation in the hippocampus following the formation of contextual memories.

We next examined the effects of fear conditioning on δ-catenin/N-cadherin interactions at the synapse. Hippocampal P2 synaptosomal fractions were isolated, immunoprecipitated with an

N-cadherin antibody, and membranes immunoblotted for δ-catenin. There was a significant increase in δ-catenin/cadherin interactions at both the 1 and 24 h time points (Fig. 2.10f,g). This relatively stable increase in the association between δ-catenin and cadherin following fear conditioning was similar to that observed in cultured neurons following increased activity (Fig. 98

2.3i,j). Our findings are consistent with the interpretation that a transient increase in palmitoylation of δ-catenin targets it for a more stable association with N-cadherin at hippocampal synapses in vivo, and is correlated with the acquisition of contextual fear memories.

Figure 2.10. Context-Dependent Fear Conditioning Increases δ-Catenin Palmitoylation and N- Cadherin Associations in the Hippocampus. (a) Freezing behavior in 6- to 9-week-old male mice 1 h and24 h after contextual fear conditioning (conditioned group, C) compared to mice that did not receive a 99

Figure 2.10 Continued: foot shock (naive group, N) (n = 10 mice per group per time point; P < 0.001, F1,36 = 35.89 (treatment effect); P = 0.539, F1,36 = 0.38 (time point effect)). (b–e) Palmitoylation, determined by ABE chemistry, in hippocampal (b,c) or cortical (d,e) lysates from naive and conditioned mice immunoprecipitated with antibodies to δ-catenin or IgG. (b,c) Palmitoylation of δ-catenin in the hippocampus of conditioned mice (n = 5 blots from 5 separate animals; P = 0.034, F1,16 = 5.37 (treatment effect); P = 0.037, F1,16 = 5.17 (time point effect)). (d,e) Palmitoylation levels of δ-catenin in the cortex of naive and conditioned mice (n = 3 blots from 3 separate animals; P = 0.463, F1,8 = 0.59 (treatment effect); P = 0.978, F1,8 = 0.1 (time point effect)). (f,g) The amount of δ-catenin bound to N-cadherin in the hippocampus of conditioned mice 1 and 24 h after training, as determined by immunoprecipitation of P2 synaptosomes from isolated hippocampal lysates with antibodies to N-cadherin. The IgG lane in the blot on the right was cropped from another position within the same blot (n = 3 and 5 blots from 3 and 5 separate animals 1 and 24 h after training, respectively; P < 0.001, F1,12 = 24.95 (treatment effect); P = 0.492, F1,12 = 0.50 (time point effect)). All graphs represent the mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA and Bonferroni's test post hoc.

2.3.8 DHHC5 is Required for Palmitoylation of δ-Catenin

We next sought to identify the DHHC protein that palmitoylates δ-catenin in an activity- dependent manner. Although 23 mammalian DHHC proteins have been identified 217, only 17 of these have been shown to have PAT activity 236. We co-expressed each of these 17 DHHC proteins with δ-catenin in HEK293T cells and tested for δ-catenin palmitoylation (Fig. 2.11a,b).

DHHC5 and DHHC20 significantly increased δ-catenin palmitoylation relative to a vector control, indicating that these proteins are sufficient for δ-catenin palmitoylation (Fig. 2.11a,b).

We performed a second overexpression screen in neurons using DHHC proteins that are expressed in the brain, and which have known synaptic protein substrates (217, Allen Mouse

Brain Atlas; http://mouse.brain-map.org/). Overexpression of DHHC5 and 20 enhanced δ- catenin/ N-cadherin colocalization in the absence of cLTP (Fig. 2.11c, and 2.12a), and this enhanced clustering was abolished in cells expressing the δ-catenin C960-1S mutant (Fig. 2.11d, and 2.12b). Together, this indicates that DHHC5 and 20 are sufficient to enhance the recruitment of δ-catenin to cadherin clusters by palmitoylating δ-catenin in neurons.

100

We next determined whether DHHC5 and DHHC20 are necessary for activity-dependent palmitoylation of δ-catenin. The efficacy of DHHC5 shRNA (shRNA-D5; 220) and DHHC20 siRNA (siRNA-D20) were first validated in neurons (Fig. 2.11e,f). Activity-induced recruitment of δ-catenin to N-cadherin clusters was abolished in neurons expressing shRNA-D5 but not siRNA-D20 (Fig. 2.11g). These results indicate that although both DHHC5 and 20 are sufficient to palmitoylate δ-catenin, only DHHC5 is required for activity-induced palmitoylation of δ- catenin and the activity-dependent targeting of δ-catenin to cadherin.

We assayed δ-catenin palmitoylation in DHHC5 knockdown cells to confirm our observations. We first confirmed that the DHHC5 knockdown observed at 6 DIV (Fig. 2.11e) persisted for up to 12 DIV (Fig. 2.13a). The knockdown in individual cells was robust, with a

40% decrease in DHHC5 levels reflecting the transfection efficiency of Amaxa nucleofection

(approximately 50%). Basal levels of δ-catenin palmitoylation did not appear to be significantly reduced in cells expressing DHHC5 shRNA (Fig. 2.13a,b). However, activity-dependent increase in δ-catenin palmitoylation was abolished in these cells and restored in cells co-expressing shRNA-resistant DHHC5 (Fig. 2.13a,b). Together, this strongly indicates that DHHC5 is required for activity-dependent palmitoylation of δ-catenin.

DHHC5 has previously been shown to accelerate the constitutive delivery of AMPARs to the cell surface through palmitoylation of the δ-catenin binding partner, GRIP1b 184, 220.

However, it is unknown whether DHHC5 plays a role in activity-dependent recruitment of

AMPARs to the cell surface. As expected, we observed a robust increase in the intDen of SEP-

GluA1 clusters 40 min after cLTP in control neurons (vector + shRNA-c; Fig. 2.13c,d). Neurons overexpressing DHHC5 exhibited a significant increase in the basal intDen of SEP-GluA1,

101

consistent with a role for DHHC5 in constitutive AMPAR surface delivery 220, and also exhibited a further activity-dependent increase in AMPAR recruitment to the membrane (DHHC5 + shRNA-c; Fig. 2.13c,d). We observed a similar basal increase in SEP-GluA1 intDen in neurons expressing DHHC5 plus δ-catenin shRNA, in line with the observation that DHHC5 enhances the recruitment of AMPAR to the cell surface through palmitoylation of another substrate.

However, the activity-dependent increase in the intDen of SEP-GluA1was abolished in these cells. Expression of shRNA-resistant δ-catenin (δ-catenin WT*) rescued the effect of knocking down δ-catenin in DHHC5 expressing cells, whereas expression of palmitoylation-defective δ- catenin (C960-1S*) did not (Fig. 2.13c,d). Together, this clearly demonstrates that DHHC5 mediates activity-induced insertion of AMPARs to the cell surface through palmitoylation of δ- catenin. The increase in the intDen of SEP-GluA1 under basal conditions was maintained in δ- catenin knockdown cells, WT rescue cells, and C960-1S rescue cells (Fig. 8c,d), indicating that palmitoylation of δ-catenin by DHHC5 is not required for constitutive AMPAR surface delivery.

We also observed significantly more GluA1/ δ-catenin WT colocalization compared to

GluA1/ δ-catenin C960-1S colocalization at basal levels in cells expressing DHHC5. Moreover, activity significantly enhanced GluA1/ δ-catenin WT colocalization, but had no effect on GluA1/

δ-catenin C960-1S colocalization (Fig. 2.13e).

102

Figure 2.11. DHHC5 and DHHC20 Palmitoylate δ-Catenin, but Activity-Induced Recruitment of δ- Catenin to N-Cadherin is Mediated by DHHC5. a,b) Palmitoylation of GFP–δ-catenin co-transfected with the indicated DHHC constructs in HEK293T cells (n = 3–6 blots from separate cultures; P < 0.001, F19,36 = 7.154). (c) Colocalization of GFP–δ-catenin and N-cadherin–RFP in cells expressing the indicated DHHCs (P < 0.001, F11,834 = 15.4). The cell numbers from 3–10 cultures were as follows: 103

Figure 2.11 Continued: vector –glycine (n = 262), vector +glycine (n = 200), DHHC2 (n = 35), DHHC3 (n = 47), DHHC5 (n = 39), DHHC7 (n = 31), DHHC8 (n = 56), DHHC9 (n = 49), DHHC13 (n = 22), DHHC15 (n = 18), DHHC17 (n = 24), DHHC20 (n = 63). (d) The role of δ-catenin palmitoylation in DHHC5-induced (P = 0.004, F3,75 = 4.67) and DHHC20-induced (P = 0.005, F3,86 = 6.559) clustering of δ-catenin and N- cadherin. The numbers of cells from 3 cultures were as follows: DHHC5 experiment, WT + vector (n = 22), WT + DHHC5 (n = 17), C960-1S + vector (n = 20), C960-1S + DHHC5 (n = 20); DHHC20 experiment, WT + vector (n = 23), WT + DHHC20 (n = 20), C960-1S + vector (n = 12), C960-1S + DHHC20 (n = 20). (e,f) RNA interference–mediated knockdown of DHHC5 and DHHC20 in 6 DIV hippocampal neurons. n = 5 blots from 5 cultures; P = 0.022, F2,12 = 5.29 (asterisks next to the protein names denote shRNA resistance) (e); n = 3 blots from 3 cultures; P = 0.019, F2,6 = 8.29 (hDHHC20 denotes human DHHC20) (f). (g) Activity-induced increases in the colocalization of GFP–δ-catenin and N-cadherin–RFP after knockdown of DHHC5 or DHHC20. The cell numbers and P values from paired t tests before as compared to after activity were as follows: DHHC5 (P < 0.001, F5,78 = 5.29, one-way ANOVA); shRNA-D5c (n = 15, P < 0.001), shRNA-D5 (n = 13, P = 0.288), shRNA-D5 + DHHC5* (n = 14, P < 0.001); DHHC20 (P < 0.001, F5,66 = 13.65, one-way ANOVA); siRNA-D20c (n = 11, P = 0.004), siRNA-D20 (n = 13, P < 0.001), siRNA-D20 + hDHHC20 (n = 12, P = 0.012). Graphs represent the mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001, one-way ANOVA and Tukey's test post hoc (b–f); *P < 0.05, **P < 0.01, ***P < 0.001, paired t test; ##P < 0.01, one-way ANOVA and Tukey's test post hoc (g).

104

Figure 2.12. DHHC5 and DHHC20 Enhance the Recruitment of δ-Catenin to N-Cadherin under Basal Conditions. (a,b) Confocal images of hippocampal neurons transfected at 10 DIV with the indicated GFP-δ-catenin constructs, N-cadherin-RFP, and either an empty vector or the indicated Myc or HA-tagged DHHC constructs. Neurons were imaged at 14-16 DIV, 40 min after the indicated treatment with glycine or a control buffer lacking glycine. Scale bar = 20µm. (a) DHHC5 and DHHC20 are sufficient to cluster δ-catenin and enhance its colocalization with N-cadherin under basal conditions. Images for DHHC2 and DHHC8 are provided as negative controls. (b) Overexpression of DHHC5 and DHHC20 does not enhance the recruitment of palmitoylation-deficient (C960-1S) δ-catenin to N-cadherin clusters indicating that DHHC5 and DHHC20 enhance δ-catenin/N-cadherin colocalization by palmitoylating δ-catenin.

105

Figure 2.13. DHHC5 is Required for Activity-Induced Palmitoylation of δ-Catenin and Surface AMPAR Insertion. (a, top) DHHC5 levels at 12 DIV after the indicated treatment (shRNA-D5c, 1.0 ± 0.13; shRNA-D5, 0.62 ± 0.042; shRNA-D5 + DHHC5*, 0.89 ± 0.044; P = 0.042, F2,6 = 5.57, one-way ANOVA). (a, bottom, b) The effect of knockdown of DHHC5 on activity-induced palmitoylation of δ- catenin (n = 3 blots from 3 separate cultures; P = 0.002, F6,14 = 10.03). (c) Confocal images of 106

Figure 2.13 Continued: hippocampal neurons transfected with the indicated constructs before and 40–60 min after glycine treatment. SEP fluorescent puncta are pseudocolored in the heat maps. DHHC5-hemagglutinin (HA) overexpression was confirmed by post hoc immunostaining for HA. Scale bars, 5 µm. (d) IntDen of SEP- GluA1 puncta normalized to the same puncta before glycine treatment. The n values, indicating the cell numbers from ≥3 separate cultures, and the P values from paired t tests are as follows: vector + shRNA-c (n = 14, P < 0.001), DHHC5 + δ-catenin shRNA-c (n = 15, P = 0.006), DHHC5 + δ-catenin shRNA (n = 13, P = 0.857), DHHC5 + δ-catenin shRNA + δ-catenin WT* (n = 16, P = 0.002) and DHHC5 + δ- catenin shRNA + δ-catenin C960-1S* (n = 13, P = 0.161). DHHC5 overexpression increased basal SEP- GluA1 IntDen (P = 0.001, F9,132 = 4.141, one-way ANOVA). # denote significance among the groups before treatment relative to vector + shRNA-c, and asterisks denote significance within groups before and after glycine treatment. (e) Colocalization of the indicated δ-catenin constructs with GluA1 (P < 0.001, F3,54 = 19.29, one-way ANOVA). # denote significance among the groups before treatment relative to DHHC5 + shRNA + C960-1S*. Colocalization of WT* and GluA1 increased after activity (P = 0.005, paired t test), whereas colocalization of C960-1S* and GluA1 did not (P = 0.741, paired t test). Graphs represent the mean ± s.e.m. *P < 0.05, **P < 0.01, one-way ANOVA and Tukey's test post hoc (b); *P < 0.05, ***P < 0.001, paired t test; #P < 0.05, ##P < 0.01, one-way ANOVA and Tukey's test post hoc (d,e).

2.4 Discussion

Our study demonstrates that palmitoylated δ-catenin mediates activity-dependent changes in synapse adhesion, structure, and efficacy that are correlated with the formation of new memories. We have demonstrated that enhanced synaptic activity results in a transient increase in the palmitoylation of δ-catenin, which targets it for a lasting association with N-cadherin at the postsynaptic membrane. Notably, we observed that the binding of δ-catenin to N-cadherin confers increased stability to cadherin molecules within spine heads and in turn, allows the activity-dependent remodeling of spines and the insertion of synaptic AMPARs. Consistently, the palmitoylation of δ-catenin is essential for activity-dependent increases in mEPSC amplitude and frequency. Moreover, we demonstrate that the acquisition of context-dependent fear memory is correlated with a transient increase in the palmitoylation of δ-catenin and a concomitant increase in synaptic cadherin/δ-catenin interactions specifically within the hippocampus. This temporal agreement between our in vitro and in vivo results underscore our model that palmitoylation of δ-catenin drives changes in the structure and efficacy of synapses believed to 107

underlie learning and memory. Finally, we demonstrated that DHHC5 and DHHC20 are sufficient to palmitoylate δ-catenin, whereas only DHHC5 is required for activity-induced increase in δ-catenin palmitoylation. Interestingly, we demonstrate that DHHC5 mediates activity-induced surface insertion of AMPARs through palmitoylation of δ-catenin.

The stability of N-cadherin at the membrane is known to be regulated by the catenins18.

Binding of p120ctn to the cadherin juxtamembrane domain stabilizes surface cadherin by masking endocytic motifs193, 261, and it is possible that δ-catenin stabilizes cadherin by the same mechanism. Indeed, δ-catenin associates with the identical region of the cadherin juxtamembrane domain as p120ctn and both utilize a conserved lysine residue to bind cadherin193. Despite this, different protein interaction domains on these two proteins suggest that p120ctn and δ-catenin exert different functions when bound to cadherin. As an example, δ-catenin contains a PDZ- binding motif that enables it to interact with PDZ domain-containing proteins. Although it is tempting to speculate that palmitoylation of δ-catenin provides a competitive advantage for binding to cadherin following increased activity, we have observed increased cadherin/δ-catenin interactions with no significant changes in cadherin/p120ctn interactions. Activity-dependent recruitment of β-catenin to synaptic cadherin has also been shown to stabilize cadherin within the membrane188. As β-catenin interacts with the C-terminal domain of cadherin279, a region distinct from the site of δ-catenin interaction, it likely that activity-driven post-translational modifications of both β-catenin and δ-catenin are important for regulating cadherin surface stability.

Cadherin clustering within spine heads has been shown to be essential for activity- induced spine remodeling in cultured neurons173 and in vivo174. It is possible that the

108

palmitoylation of δ-catenin and its binding to cadherin promote spine remodeling by increasing the interactions of membrane-bound cadherin with the actin cytoskeleton through its association with α-catenin175. It is also possible that δ-catenin regulates spine morphology more directly, through its interaction with actin-binding proteins, including cortactin180 and several Rho-

GTPases182 (but see181). Interestingly, knockdown of δ-catenin enhances protrusion density through mechanisms involving its PDZ-binding domain and not by its ability to bind cadherin178.

Consistent with this report, we show that spine density is not significantly changed in neurons expressing δ-catenin K581M or C960-1S. Together, our results suggest that δ-catenin palmitoylation and cadherin-binding are primarily important for spine remodeling and maturation, whereas a δ-catenin’s interactions with actin-binding proteins and an unknown signaling mechanism via its PDZ domain can regulate protrusion density.

The reversible addition of palmitate to synaptic proteins can dramatically impact their trafficking and localization, however the regulation and temporal profile for synaptic protein palmitoylation varies greatly between substrates220, 242. For example, activity regulates the palmitoylation of PSD-95, resulting in a relatively stable modification of this protein220, 242 (but see 241). In contrast, palmitoylation of the synaptic adaptor protein, GRIP1b, is not activity- dependent and exhibits a strikingly high rate of palmitate turnover220. Thus, the transient activity- dependent increase in δ-catenin palmitoylation is distinctive and does not merely reflect generalized changes in synaptic protein palmitoylation.

It is interesting to speculate that δ-catenin may cooperate with GRIP1b to mediate surface insertion of AMPARs. Palmitoylation of δ-catenin by DHHC5 is specifically required for activity-dependent increases in surface AMPARs, whereas palmitoylation of GRIP1b by

109

DHHC5 mediates constitutive delivery of AMPARs to the cell surface220. We hypothesize that constitutive palmitoylation of GRIP1b regulates the mobilization and delivery of AMPARs to synapses, whereas transient activity-dependent palmitoylation of δ-catenin increases the stable surface population of synaptic AMPARs.

There are two possible mechanisms by which δ-catenin palmitoylation and cadherin binding regulate surface AMPAR localization and synapse strength. First, δ-catenin localization at synapses may stabilize AMPARs at the synaptic surface by directly interacting with the scaffold molecules GRIP/ABP184, 205 and PSD-95183 through its PDZ binding motif. Second, δ- catenin may regulate the stability of AMPARs at cell the surface by stabilizing N-cadherin within the membrane and enhancing interactions between cadherin and AMPAR subunits.

Indeed, N-cadherin has been shown to bind GluA2 directly through its extracellular domain, leading to the stabilization of GluA2-containing AMPARs186. Moreover, GluA1-containing

AMPARs have also been observed in complex with N-cadherin280. δ-catenin can therefore link

N-cadherin with AMPARs both by interacting with adaptor proteins intracellularly, and by positioning more N-cadherin in the membrane and enabling the direct extracellular interaction between AMPARs and cadherin. Although cells overexpressing δ-catenin exhibited activity- induced increases in the recruitment of δ-catenin to synapses, we report no activity-induced increase in GluA1 and GluA2 surface expression or mEPSC frequency and amplitude. We suggest that the high level of synaptic δ-catenin and surface GluA1 and GluA2 levels under basal conditions occludes any additional activity-induced increase in surface GluA1 and GluA2 (but not activity-induced recruitment of δ-catenin).

110

Knockout studies have demonstrated a requirement for N-cadherin as well as δ-catenin and β-catenin in the formation of contextual memories197, 206, 281. Indeed, acute blockade of N- cadherin-mediated adhesion in the hippocampus in vivo inhibits contextual memory formation197 and knockdown of β-catenin disrupts the consolidation of fear memories281. Interestingly, in stark contrast to what we observed for δ-catenin, fear conditioning resulted in a transient decrease in β-catenin/cadherin interactions associated with increased β-catenin phosphorylation281. Post-translational modifications for both catenin proteins occurred within minutes and lasted <2 hours, altering their association with synaptic cadherin. It is likely that activity-driven post-translational modifications to both β-catenin and δ-catenin are important for regulating cadherin surface stability. However, the distinct time course for the association of cadherin with these catenins following learning suggests different roles for these molecules in memory acquisition and consolidation. Our data demonstrates that by increasing surface

AMPARs and enhancing spine maturation and size, activity-driven palmitoylation of δ-catenin is a vital component for LTP, learning and memory.

111

Chapter 3: Activity-Mediated Trafficking of the Palmitoyl Acyl-

Transferase DHHC5

Neuronal activity can regulate the palmitoylation of proteins, which in turn controls the dynamic localization of proteins to and from synaptic compartments. Despite the importance of palmitoylation in synapse plasticity, it is unclear how activity alters the ability of palmitoyl-acyl transferases to palmitoylate their substrates. Here we demonstrate that DHHC5, a palmitoyl-acyl transferase for a number of key synaptic proteins, is constitutively active and that its ability to palmitoylate substrates in an activity-dependent manner is dependent on changes in its subcellular localization. Under basal conditions DHHC5 is primarily localized to dendritic spines, whereas its substrate, δ-catenin, is highly localized to dendritic shafts, resulting in the segregation of the enzyme/substrate pair. Neuronal activity enhances DHHC5 endocytosis and its translocation to dendritic shafts on recycling endosomes, positioning DHHC5 in the same compartment as its substrate. Following binding to, and palmitoylation of, δ-catenin, DHHC5 and δ-catenin are trafficked together back into spines, thereby regulating cadherin stabilization and synapse plasticity.

3.1 Introduction

Palmitoylation is a reversible post-translational modification involving the addition of palmitate onto cysteine residues that facilitates the trafficking of proteins to cell membranes212.

Protein palmitoylation is mediated by a family of multi-pass, transmembrane palmitoyl-acyl transferase (PAT) enzymes that contain a zinc-finger domain and a conserved Asp-His-His-Cys

(DHHC, also called zDHHC) motif that is required for palmitoyl-transferase activity217. In

112

neurons, DHHC proteins are localized to the Golgi, vesicular, or plasma membranes and palmitate cycling on substrate proteins can be constitutive or dynamically regulated by cellular signals241, 242. Recent work has identified a number of synaptic proteins that are substrates for palmitoylation and demonstrated that neuronal activity regulates the palmitoylation and trafficking of these proteins211, 217. Thus, palmitoylation is an important regulator of synapse plasticity.

DHHC5 is localized to postsynaptic compartments220 and can palmitoylate PSD-95236, 247,

Grip1b220, and δ-catenin282. DHHC5-mediated palmitoylation of Grip1b and δ-catenin increases synaptic delivery and surface stabilization of AMPA receptors (AMPARs), respectively, implicating DHHC5 in the regulation of synapse efficacy220, 282. The importance of DHHC5 in synaptic regulation is further supported by impaired synapse plasticity and performance on learning and memory tasks in ZDHHC5 mutant mice247.

We have previously shown that activity increases DHHC5-mediated palmitoylation of δ- catenin282. Here we demonstrate that this is not due to alterations in the enzymatic activity of

DHHC5, but rather its subcellular localization. Indeed, synaptic activity enhances the internalization of DHHC5 and its translocation from spines into dendritic shafts where it binds and palmitoylates δ-catenin. We demonstrate that DHHC5 is mobilized on recycling endosomes

(REs) and is subsequently re-trafficked back into spine synapses together with palmitoylated δ- catenin. Together, we demonstrate that activity-dependent regulation of DHHC protein trafficking provides a mechanism for the local control of protein palmitoylation and delivery to synapses.

113

3.2 Materials and Methods

3.2.1 Antibodies and cDNA Constructs

Primary antibodies: δ-catenin (1:500 for WB, 5µg for IP; BD Transduction Laboratories

611536), N-cadherin (1:500; BD Transduction Laboratories 610921), PSD-95 (1:500; Abcam ab2723), Gephyrin (1:500; Synaptic Systems 147 011), GFP for IP (10µL; Synaptic Systems 132

002), GFP for WB (1:1000; Roche 11814460001) DHHC5 (1:500 ICC, 1:1000 WB, 1µg IP;

Sigma Prestige HPA014670), Transferrin Receptor (1:500; Millipore GR09L), VPS-35 (1:500;

Abnova H000055737-M02), GluA1 (1:1000; Millipore 05-855R) and hemagglutinin (HA)

(1:500; Cell Signaling Technology C29F4). Secondary antibodies: IgG-HRP (1:5000; BioRad mouse 170-6516, and rabbit 170-6515), Streptavidin-HRP (1:5000; Thermo Scientific 21126),

Alexa-Fluor 568 goat anti-mouse, and 633 goat anti-rabbit (1:1000; Life Technologies A-11004 and A-21070, respectively).

GFP-δ-catenin and RFP-δ-catenin were generated as previously described (Brigidi et al.,

2014). TfR-mCherry was a kind gift from Dr. Michael Silverman (Simon Fraser University,

Vancouver, BC). HA-DHHC5 and HA-DHHS5 were kind gifts from Dr. Richard Huganir (Johns

Hopkins University, Balitmore, MD). The GFP-DHHC5 construct was generated by PCR of

DHHC5 from HA-DHHC5 (Accession no: NM_00139388.) using an EcoRI-tagged forward primer (CCGGCGAATTCTATGCCCGCAGAGTCTG) and a BamHI-tagged reverse primer

(CGAGATTTCTGTGTGAGGATCCCCGGC), and pasted into pEGFP-C1 using EcoRI and

BamHI sites.

114

3.2.2 Cell Cultures

Primary hippocampal neurons: Hippocampi from embryonic day 18 (E18) Sprague-Dawley rats of either sex were prepared as previously described265, and plated at a density of 130 cells/mm2.

A NeuroCult SM1 supplement (Stem Cell Technologies 05711) was used in the place of B27 in maintenance media. Neurons were transfected with Lipofectamine 2000 (Invitrogen) at 9-10 DIV according to the manufacturer’s recommendations, and used for experiments at 13-14 DIV.

HEK Cells: HEK293T cells were transfected using polyethylenimine (PEI; Sigma) as previously described266, in a 3:1 ratio with the total plasmid DNA to be transfected. 293T cells were transfected at 70-80% confluency, and incubated for 24-48 hours before harvesting for biochemistry.

3.2.3 Neuronal Activation

Neuronal activity was modified using previously described chemical LTP or chemical

LTD protocols130, 267, 282, 283. Briefly, maintenance media was replaced with an extracellular recording solution (ECS) containing the following: 125mM NaCl, 5mM KCl, 2mM CaCl2, 0mM

MgCl2, 5mM HEPES, 33mM D-Glucose, and supplemented with 0.5µM TTX, 20µM

Bicuculline, pH 7.3, 290 mOsm/L, for 10-15 minutes. For chemical LTP, this media was supplemented with 200µM glycine, or 200µM glycine plus 50µM AP5 (Sigma) for 3 minutes.

For chemical LTD, the media was supplemented with 20µM NMDA (Sigma) plus 10µM glycine for 3 minutes. The solution was then replaced with fresh ECS (containing 0mM MgCl2 for LTP, or 2mM MgCl2 for LTD) for the indicated times prior to experimentation. Cells were continually maintained at 37°C for the duration of activity stimulation.

115

3.2.4 Acyl-Biotin Exchange (ABE) Assay

ABE assays were performed as previously described270. Briefly, cells were lysed in lysis buffer supplemented with 50mM N-ethylmaliemide, and target proteins immobilized overnight on protein A/G-conjugated sepharose beads by IP at 4°C. Immunocomplexes were then washed and treated with lysis buffer at pH 7.2 and supplemented with 1M hydroxylamine (NH2OH) for 1 hr at room temperature, then washed and treated with lysis buffer at pH 6.2 and supplemented with 1µM biotin-BMCC (Thermo Scientific) for 1 hr at 4°C. Target proteins were then eluted in a 2x loading buffer containing 5mM dithiothreitol (DTT), and palmitoylated proteins analyzed by immunoblotting.

3.2.5 Biotinylation Assay

Biotinylation of all surface proteins was performed as previously described268. Briefly, cells were washed in phosphate-buffered saline (PBS) supplemented with 1mM MgCl2 and

0.1mM CaCl2 (PBS-CM) and treated with PBS-CM supplemented with 0.5mg/mL of Sulfo-

NHS-SS-biotin (Thermo Scientific) for 30 min at 4°C. Cells were then washed once with PBS-

CM, then twice with PBS-CM supplemented with 20mM glycine for 7 min each at 4°C. Cells were then lysed in lysis buffer of PBS containing 1% IGEPAL CA-630 at 4°C and supplemented with the inhibitors described above. Surface proteins were then immobilized on Neutravidin- agarose beads (Thermo Scientific) overnight at 4°C and analyzed by immunoblotting.

3.2.6 Immunoprecipitation

Immunoprecipitation (IP) assays were performed as previously described 19. Briefly, lysates were incubated overnight at 4°C with the indicated antibody. 50µL of protein A/G–

116

Sepharose (GE healthcare, Chicago, IL) was added to the lysates, and the bead-bound immunocomplexes were recovered after 2 hours, washed 4 times with lysis buffer, solubilized with loading buffer, separated by SDS–PAGE and analyzed by immunoblotting with the indicated antibodies.

3.2.7 Western Blot Analysis

Western blotting was performed as previously described 19. Brain tissue, primary hippocampal neurons, and 293T cells were homogenized in an ice-cold lysis buffer containing

1% IGEPAL CA-630 (Sigma), 0.5% Triton-X 100, 50mM Tris-HCl pH 7.5, 150mM NaCl, 10%

Glycerol, and supplemented with phenylmethanesulfonyl fluoride solution (PMSF) and a protease inhibitor cocktail with EDTA (Roche). Proteins were separated by SDS-PAGE, analyzed by immunoblotting with the indicated antibodies, and visualized using enhanced chemiluminescence (Pierce Biotechnology) on a Bio-Rad Versadoc 4000 (Bio-Rad

Laboratories). Blots were quantified using Image J software.

3.2.8 Immunocytochemistry

Immunocytochemistry experiments were performed as previously reported 19. Briefly, cells were fixed in a 4% paraformaldehyde/sucrose solution pre-warmed to 37°C for 10 min, then washed in PBS, treated with 0.1% Triton-X 100/PBS for 10 min at room temperature (RT), washed again with PBS, and then blocked with 10% goat serum (GS)/PBS for 1 hr at RT.

Following blocking, cells were incubated with primary antibodies in 0.1% GS/PBS overnight at

4°C. Subsequently, cells were washed in PBS 3x for 10 min each, incubated in secondary

117

antibodies in 0.1% GS/PBS for 1 hr at RT, washed again in PBS 3x, and mounted on microscope slides in Prolong Gold (Molecular Probes).

3.2.9 Confocal Imaging

All neurons were imaged using an Olympus Fluoview 1000 confocal microscope

(60×/1.4 Oil Plan-Apochromat). Identical acquisition parameters were used for all cells across all separate cultures within an experiment. For time-lapse imaging of GFP-DHHC5, a region of interest along a primary dendrite of a transfected cell within 100µm from the cell body was chosen and imaged before, and again after the 3 min of cLTP/cLTD at the indicated time points.

Dendritic spines were defined as any protrusion between 0.5µm to 10µm in length emanating from the dendritic shaft within the proximal 100µm from the cell body. Confocal images shown in the figures were subjected to a 1 pixel Gaussian blur. The levels and contrast of confocal images were moderately adjusted in Photoshop CS6 software (Adobe Systems, Inc.) using scientifically accepted procedures284.

3.2.10 Image Analysis and Quantification

Confocal images for a particular experiment were subjectively thresholded using ImageJ software, and the same threshold was used for all images obtained for a single experiment, throughout the experimental analysis. For live time-lapse imaging experiments, the same threshold was applied to the image acquired before stimulation, and to all the images acquired afterward at the indicated time points. Puncta were defined as a thresholded fluorescence cluster with an area between 0.05-3µm2. Puncta area and Integrated Density (IntDen; the product of area and mean gray value) were then determined using ImageJ. An Image J colocalization plugin was 118

used to assess colocalization between different channels282

(http://rsb.info.nih.gov/ij/plugins/colocalization.html). Points of colocalization were defined as regions of >4 pixels in size, with >50 intensity ratio between the two channels.

For analysis of GFP-DHHC5 fluorescence within spines, a circular region-of-interest

(ROI) was drawn around identified spines and the Integrated Density (IntDen) measured before and after stimulation.

For analysis of colocalization specifically within spines and dendrites, a mask was generated in Photoshop CS6 that outline fluorescence in spines and shafts. Using the paintbrush tool, the dendrite was manually highlighted, separating spines and dendrites and enabling the generation of spine-only or dendrite-only masks. Specific masks were then applied to all channels to selectively analyze colocalization within spines or dendrites. Colocalization analysis was then done as described above.

3.2.11 Statistical Analysis

All data values are expressed as mean ± SEM. For all imaging experiments, “n” refers to the number of cells used per condition, over at least 3 separate cultures, with the exception of the analysis performed in Fig. 3.2C, where “n” refers to the number of spines, and is specified within the figure legends. Our sample sizes are similar to those reported in the literature220, 241, 282. All data were analyzed in Prism software (GraphPad Software, Inc.) and met the assumption of normality by using a D’Agostino and Pearson omnibus normality test in Prism, with the exception of all biochemical data in which the “n” values were too small, and so a normal distribution was assumed and not formally tested. Statistical significance was determined by student’s t-test, one-way ANOVA or repeated measures one-way ANOVA with post hoc tests

119

using Prism, where indicated. Statistical significance was assumed when p<0.05. In all figures,

*p<0.05, **p<0.01, and ***p<0.001, as determined in Prism software. All figures were generated using Illustrator CS6 software (Adobe Systems, Inc.).

3.3 Results

3.3.1 Activity-Induced Mobilization of DHHC5

Previous work has shown that increased neuronal activity enhances DHHC5-mediated palmitoylation of its substrate, δ-catenin282. To further understand the molecular mechanism underlying this process, we first determined whether activity enhances protein palmitoylation by increasing the enzymatic activity of DHHC5. Protein S-palmitoylation proceeds by a two-step mechanism: the initial autopalmitoylation of a DHHC cysteine side chain followed by the transfer of palmitate to the substrate cysteine217, 236. The autopalmitoylation of DHHC proteins can therefore be used as a measure of enzymatic activity236, 242. We examined whether neuronal activity enhances the autopalmitoylation of DHHC5 using the acyl-biotin exchange (ABE) assay, which exchanges palmitoyl modifications with biotin211, 269, 270 (Fig. 3.1A). Exclusion of hydroxylamine (NH2OH) was used as a control for the specificity of biotin labeling as it is essential for the cleavage of palmitate from cysteines220, 224, 282. 14 days in vitro (DIV) hippocampal neurons were stimulated using a standard chemical LTP (cLTP) protocol involving a 3 min treatment with glycine/bicucculine that selectively activates synaptic NMDARs, leading to the recruitment of AMPARs, and the enhancement of synapse strength267, 282. There was no significant difference in the autopalmitoylation of DHHC5 10 and 40 min after stimulation with glycine, time points associated with activity-induced palmitoylation of δ-catenin by DHHC5282

(Fig. 3.1A,B), indicating that activity does not regulate the enzymatic function of DHHC5. 120

We next examined whether neuronal activity controls protein palmitoylation by modifying the subcellular localization of DHHC enzymes. Under basal conditions, 56.9 ± 5.8%, of endogenous DHHC5 localized to excitatory synapses (n= 14 cells) and 31.18 ± 2.53%, localized to inhibitory synapses (n= 32 cells) indicating that approximately 88% of DHHC5 is localized to synapses (Fig. 3.1C). Furthermore, 79.73 ± 4.92% of excitatory synapses and 46.78

± 1.77% of inhibitory synapses colocalize with DHHC5, indicating that the majority of synapses contain DHHC5 (Fig. 3.1C).

We stimulated 14 DIV neurons using cLTP267 or cLTD283 protocols in which cells were treated for 3 min with glycine/bicucculine or glycine/NMDA, respectively, and then returned to basal media for 40 min. In contrast to cLTP, cLTD activates both synaptic and extrasynaptic

NMDARs, resulting in AMPAR internalization and synaptic depression140, 283. The efficacy of these protocols in our cells were confirmed by determining the integrated density (IntDen; product of area and mean gray value) of PSD-95 puncta, which was significantly increased following cLTP and decreased following cLTD, in agreement with previous observations (Fig.

3.1D)118, 267, 283. Similarly, cLTP significantly increased the IntDen of DHHC5 within spines, whereas cLTD did not indicating that DHHC5 is trafficked to excitatory synapses following cLTP (Fig. 3.1F. In AU: –Gly, 14.51 ± 0.46, n=38 cells; +Gly, 19.02 ± 0.86, n=35;

+Gly/NMDA, 15.18 ± 1.07, n=14; p<0.001, F2,84=12.99, one-way ANOVA ).

To further confirm this, neurons were immunostained with antibodies against DHHC5, the excitatory postsynaptic marker, PSD-95, or the inhibitory postsynaptic marker, gephyrin 40 min after cLTP/cLTD (Fig. 3.1F). cLTP increased DHHC5/PSD-95 colocalization but did not impact DHHC5/gephyrin clocalization (Fig. 3.1G,I). Interestingly, cLTD did not affect DHHC5 colocalization with either synaptic marker (Fig. 3.1G,I). Together this indicates that cLTP 121

increases the recruitment of DHHC5 specifically to excitatory spine synapses, whereas cLTD does not affect DHHC5 localization at either excitatory or inhibitory synapses.

Figure 3.1. Changes in DHHC5 Function and Localization Following Increased Neuronal Activity. (A, B) Autopalmitoylation of DHHC5 is not altered following cLTP stimulation (p=0.107, F3,8=2.83, n=3 blots from 3 cultures). Exclusion of NH2OH was used as a control for the specificity of biotin labeling. (C) Confocal image of 14 DIV neurons demonstrating colocalization of DHHC5 and PSD-95. (D, E) The IntDen of PSD-95 puncta is altered 40 min after treatment with cLTP (+Gly) and cLTD (+Gly,NMDA), (p<0.001, F2,84=15.28, n=38, 35, 14), whereas the IntDen of gephyrin is not (p=0.354, F2,93=1.05, n= 37, 39, 20). (F-I) Confocal images of 14 DIV neurons (F, H) demonstrating increased colocalization of 122

Figure 3.1 Continued: DHHC5 with PSD-95 (p<0.001, F2,84=22.98) (F, G), but no change in colocalization of DHHC5 and gephyrin (p=0.114, F2,93=2.21) (H,I) 40 min after cLTP. (J) Confocal images of 14 DIV neurons transfected with GFP-δ-catenin and immunostained for PSD-95 and DHHC5 (n=45 cells). Scale bars (C) = 20µm, (F, H, J) = 5µm. n = cells from 3 separate cultures. Colocalized puncta are denoted by white arrowheads. All graphs display mean ± SEM. *p<0.05, ***p<0.001; one-way ANOVA; Tukey’s test post hoc.

3.3.2 Activity-Induced Trafficking of DHHC5

We next focused on how activity-mediated translocation of DHHC5 impacts its ability to palmitoylate substrates. Since activity-mediated palmitoylation of PSD-95 is dependent on

DHHC2 and not DHHC5235, 242, and Grip1b palmitoylation is not activity-regulated220, 242, we focused on δ-catenin, which is palmitoylated by DHHC5 following activity282. Under basal conditions, 55.79 ± 2.48% of DHHC5 and 28.57 ± 1.31% of GFP-δ-catenin are localized to spine synapses (GFP-δ-catenin was validated in282. Only 27.31 ± 0.05% of GFP-δ-catenin colocalized with DHHC5, with virtually all colocalization occurring at spine synapses (93.38 ± 6.34%

DHHC5/δ-catenin colocalized with PSD-95; Fig. 3.1J, n=45 cells). This demonstrates that

DHHC5 and δ-catenin are largely localized to separate compartments with a fraction of DHHC5 and δ-catenin localized to spines in accordance with the low level of δ-catenin palmitoylation and spine localization under basal conditions282.

To track the localization of both δ-catenin and DHHC5 following activity, we transfected cells with GFP-DHHC5 and RFP-δ-catenin. We first confirmed that GFP-DHHC5 is a faithful marker of DHHC5 by comparing its localization at synapses to that of endogenous DHHC5 (Fig.

3.2A). 49.4 ± 5.0% of GFP-DHHC5 localized to excitatory synapses (n=19 cells), which was statistically similar to the 55.79 ± 2.48% of endogenous DHHC5 localized to synapses (Fig.

3.1C; p=0.343, student’s t-test). GFP-DHHC5 puncta were localized to spine heads prior to

123

cLTP treatment, translocated out of spines 2-3 min after stimulation and then trafficked back into spines 4 min after stimulation with significantly more DHHC5 localized to spines 5-20 min after stimulation (Fig. 3.2B,C). Stimulation of cells in the presence of the NMDAR blocker, AP5

(50µM), abolished glycine-mediated translocation of GFP-DHHC5 (Fig. 3.2B,C).

As previously shown, expression of GFP-DHHC5 increased the localization of δ-catenin to spines even under basal conditions resulting in the localization of RFP-δ-catenin in both spines and shafts (Fig. 3.2E)282. 2-3 min after stimulation GFP-DHHC5 moved out of spines

(Fig. 3.2E) and was significantly less colocalized with RFP-δ-catenin in the spine but significantly more colocalized in the shaft (Fig. 3.2E,F). 3-4 min after stimulation, GFP-DHHC5 was re-trafficked back into spines together with RFP-δ-catenin (Fig. 3.2E) and their colocalization in spines significantly increased while their colocalization in shafts significantly decreased (Fig. 3.2F). Accordingly, RFP-δ-catenin accumulated in spines and was concomitantly depleted from shafts 4-20 min after stimulation (Fig. 3.2E,G). Together, this demonstrates that

DHHC5 is driven out of spines following stimulation and then returns into spines together with

δ-catenin.

124

Figure 3.2. Activity Enhances DHHC5 Trafficking from Spines. (A) Confocal images of 14 DIV neurons demonstrating partial colocalization of GFP-DHHC5 and PSD-95. (B) High magnification confocal images of GFP-DHHC5 fluorescence (pseudocolored as a heat map) and DsRed before and after 125

Figure 3.2 Continued: glycine stimulation. (C) There is a change in the IntDen of GFP-DHHC5 within spine heads after glycine stimulation (p<0.001, F9,220=42.45; n=221 spines, 8 cells) which is blocked in cells treated with glycine + AP5 (p=0.46, F9,141=0.974, n=142 spines, 6 cells). (D) Representative image of GFP-DHHC5 within masks made of spines (dashed white line) or dendrites (dashed yellow line). (E) High magnification confocal images of GFP-DHHC5 and RFP-δ-catenin (pseudocolored as a heat map) within a single spine and region of dendrite (traced with white and yellow dashed lines, respectively) before and after stimulation. (F) Significant changes were observed in GFP-DHHC5 and RFP-δ-catenin colocalization in spines (p<0.001, F9,6=91.14, n=7 cells) and dendrites (p<0.001, F9,6=58.24, n=7 cells) post-stimulation. (G) Changes in IntDen of RFP-δ-catenin within spines (p<0.001, F9,6=7.549, n=7 cells) and dendrites (p<0.001, F9,6=11.83, n=7 cells) post-stimulation. n = number of cells or spines from 3-5 separate cultures. Scale bars (A) = 20µm, (B, D, E) = 1µm. All graphs show mean ± SEM. (C, F, G) Asterisks and crosshatches (F, G) above data points indicate significance relative to before stimulation within spines or dendrites, respectively. *p<0.05, **/##p<0.01, ***/###p<0.001; repeated measures one-way ANOVA, Tukey’s test post hoc.

3.3.3 DHHC5 Binds δ-Catenin and Mediates its Synaptic Recruitment

DHHC proteins have been shown to bind to their substrates during palmitoylation220 so we determined whether the time course for DHHC5/δ-catenin interactions corresponded with the translocation of δ-catenin into spine synapses. We hypothesized that DHHC5 may bind to substrates, palmitoylate them, and orchestrate their translocation from one subcellular compartment to the next. We first demonstrated that DHHC5 and δ-catenin could associate with one another when co-expressed in HEK293T cells (Fig. 3.3A). We then examined whether activity modifies DHHC5/δ-catenin interactions in hippocampal neurons. 14 DIV neurons were stimulated using cLTP, and lysates immunoprecipitated with a δ-catenin antibody at the indicated time points (Fig. 3.3B). Under basal conditions, DHHC5/δ-catenin interaction was low, in agreement with the low levels of palmitoylated δ-catenin observed in the absence of activity

282. DHHC5/δ-catenin interactions were significantly increased 5-10 min after glycine treatment and then returned to basal levels by 15 min (Fig. 3.3B,C). The time course of the activity- induced association of DHHC5 and δ-catenin (Fig. 3.3C) coincides with trafficking of these two

126

proteins into spines (Fig. 3.2B-G) suggesting that the activity-induced palmitoylation and synaptic accumulation of δ-catenin is driven by a rapid and transient increase in its association with DHHC5.

We further demonstrated that the recruitment of δ-catenin to synapses is dependent on its palmitoylation. Indeed, overexpression of DHHC5 enhanced recruitment of δ-catenin to synapses, but mutant DHHS5 that lacks palmitoyl-transferase activity220 did not (Fig. 3.3D,E).

This is consistent with our observations that palmitoylation-deficient δ-catenin (C960-1S) cannot cluster in spines282. As palmitoylation is essential for the trafficking of δ-catenin to spines together with DHHC5, it is likely that palmitate is required to tether δ-catenin to vesicular membranes such as REs for its translocation.

127

Figure 3.3. Activity Enhances DHHC5 Association with δ-Catenin. (A) HA-DHHC5 and GFP-δ- catenin coimmunoprecipitate in HEK 293T cells (n=3 blots, 3 cultures). (B, C) There was a transient enhancement in DHHC5/δ-catenin interactions following cLTP stimulation in 14 DIV hippocampal neurons (p<0.001, F5,12=18.15, n=3 blots, 3 cultures). (D) Confocal image of 14 DIV neurons demonstrating punctate GFP-δ-catenin localization and enhanced GFP-δ-catenin/ PSD-95 colocalization in cells expressing HA-DHHC5 but not HA-DHHS5. Scale bar = 5µm. (E) Percentage of GFP-δ-catenin colocalized with PSD-95 clusters (p<0.001, F2,59=8.54; n= 24 (vector), 20 (HA-DHHC5), and 18 (HA- DHHS5) cells, 3 cultures). All graphs display mean ± SEM. *p<0.05, **p<0.051, ***p<0.001; one-way ANOVA; Tukey’s test post hoc.

3.3.4 Activity-Induced Endocytosis of DHHC5 and Trafficking on Recycling Endosomes

DHHC proteins have previously been shown to localize to both the cell and RE membranes220, 240, 241, 243. Using a biotinylaytion assay we demonstrated that DHHC5 is localized

128

to the insoluble, surface fraction indicating its presence at the cell surface under basal conditions

(Fig. 3.4A,B). cLTP stimulation decreased surface DHHC5 levels 3 min post-stimulation, followed by the reinsertion of DHHC5 into the membrane 3-20 min after stimulation. DHHC5 surface levels were significantly enhanced compared to baseline 20 min after cLTP (Fig.

3.4A,B), in accordance with the increased amount of DHHC5 observed in spine synapses 20 min after cLTP (Fig 3.1F,G). We also observed an accumulation of δ-catenin in the surface fraction from 3-20 min post-stimulation which reflects the enhanced association of δ-catenin with surface

N-cadherin as previously observed282. As expected, surface GluA1-containing AMPARs increased224, 285 while N-cadherin levels remained unchanged following stimulation282.

Figure 3.4. Activity-Induced Endocytic Cycling of DHHC5. (A, B) 14 DIV hippocampal neurons were stimulated and biotinylated at the indicated time points. Lysates were immunoprecipitated with

129

Figure 3.4 Continued: neutravidin-coated beads to isolate all surface proteins, and blots probed with the indicated antibodies. (B) Intensity of protein levels in the surface fraction, normalized to whole cell input levels. DHHC5: p<0.001, F3,8=21.88; δ-catenin: p=0.009, F3,8=7.891; GluA1: p=0.003, F3,8=10.83; N-cadherin: p=0.879, F3,8=0.221). n =3 blots, 3 cultures. Graph shows mean ± SEM. *p<0.05, **p<0.01; one-way ANOVA, Tukey’s test post hoc.

The internalization and recycling of surface proteins is a highly regulated process that can be modified by synaptic activity through regulation of dendritic endosomal pathways119, 139, 188,

277. Importantly, increased activity can increase/decrease the rate of internalization of specific synaptic proteins without impacting overall endocytosis, indicating that activity does not merely generate a bulk flux of membrane proteins188, 286. To determine whether DHHC5 is specifically endocytosed and trafficked by REs, we assayed DHHC5 localization with respect to the transferrin receptor (TfR), a known marker for REs118, 130, 277. We demonstrated that 39.74 ±

2.07% of DHHC5 colocalized with TfR (n= 38 cells) indicating that a fraction of endogenous

DHHC5 is localized to dendritic REs under basal conditions (Fig. 3.5A). Immediately following stimulation (0-1 min) TfR-mCherry (mCh) rapidly translocated into spines (Fig. 3.5B,C) and was significantly more colocalized with GFP-DHHC5 (Fig. 3.5B,D). 2-3 min after stimulation both

GFP-DHHC5 and TfR-mCh trafficked out of spines as observed by both a decrease in their

IntDen (Fig. 3.5B,C), and their decreased colocalization within spines (Fig. 3.5B,D). The two proteins trafficked together back into spines 3-4 min after stimulation as evidenced by their increased IntDen (Fig. 3.5B,C), and their increased colocalization within spines (Fig. 3.5B,D).

GFP-DHHC5 continued to accumulate in spines 5-20 min post-stimulation, whereas TfR-mCh was trafficked out. TfR-positive REs have previously been shown to rapidly traffic in and out of spines following enhanced activity118, 130. Our results also demonstrate rapid translocation of

130

TfR-positive REs and further demonstrate that endocytosed DHHC5 is transported to dendritic shafts on REs.

To confirm the transport of DHHC5 between plasma and RE membranes, we immunostained for VPS-35, a marker for the retromer complex that mediates trafficking of protein cargos between these two compartments287, 288. 49.04 ± 3.76% of DHHC5 was localized to VPS-35 puncta in dendrites (spines plus shafts) (Fig. 3.5E) providing further support that

DHHC5 traffics between the plasma membrane and REs.

Our study demonstrates that activity promotes the internalization of DHHC5 from the cell membrane and its trafficking to the dendritic shaft on REs (Fig. 3.5F). This activity-driven change in the subcellular localization of DHHC5 positions it closer to its substrate, δ-catenin, resulting in the association of this enzyme–substrate pair and the palmitoylation of δ-catenin.

Importantly, δ-catenin that is tethered to the membrane of REs by palmitate is then driven into synaptic spines where it has been shown to associate with cadherin, stabilize surface cadherin and AMPARs, and increase synapse efficacy282.

131

Figure 3.5. DHHC5 is Trafficked on Recycling Endosomes (REs). (A) Confocal image of 14 DIV hippocampal neurons demonstrating colocalization between the transferrin receptor (TfR) and DHHC5 (n=38 cells). (B) High magnification confocal images of GFP-DHHC5 and TfR-mCh fluorescence in a single spine (white dashed line) over time (n=6 cells, 3 cultures). (C) Changes in the IntDen of GFP- 132

Figure 3.5 Continued: DHHC5 (p<0.001, F9,45=7.21) and TfR-mCh (p=0.002, F9,45=4.71) within spines. Percent GFP-DHHC5 colocalized with TfR-mCh in spines (p<0.001, F9,45=7.21). */#p<0.05, **/##p<0.01, ***p<0.001; repeated measures one-way ANOVA, Tukey’s test post hoc. (E) Confocal image of hippocampal neurons demonstrating colocalization between VPS-35 and DHHC5 at 14 DIV (n=22 cells). Scale bars (A, E) = 5µm, (B) = 1µm. Graphs show mean ± SEM. (F) Model of activity-regulated trafficking of DHHC5. (1) Under basal conditions, DHHC5 is localized to the postsynaptic membrane. (2) DHHC5 is endocytosed and trafficked out of the spine on REs (2-3 min post-stimulation). (3) DHHC5 associates with δ-catenin and palmitoylates it (2-10 min post-stimulation). (4) DHHC5 and δ-catenin traffic back into spines (3-20 min post-stimulation). The palmitoylation of δ-catenin is required for its trafficking into spines. (5) DHHC5 dissociates from δ-catenin and is reinserted into the synaptic membrane (10-20 min post- stimulation). Significantly more DHHC5 is recruited to the synaptic membrane 20 min poststimulation compared to basal levels. δ-catenin binds and stabilizes N-cadherin, leading to enhancements in synapse structure and efficacy. At later timepoints (up to 180 min post-stimulation) and while still bound to N- cadherin, δ-catenin becomes depalmitoylated by an unknown protein palmitoyl-thioesterase.

3.4 Discussion

A number of key synaptic proteins including PSD-95242, δ-catenin282, gephyrin223,

AKAP79/150224, and cdc42211 been shown to be palmitoylated in an activity-dependent manner.

This regulates their clustering and trafficking, thereby modulating the plasticity and function of synapses. However, to date the molecular mechanism(s) underlying activity-mediated palmitoylation of synaptic proteins remains largely unknown. Our study demonstrates that activity promotes the internalization of DHHC5 from the cell membrane and its trafficking to the dendritic shaft on recycling endosomes (Fig. 3.5F). This activity-driven change in the subcellular localization of DHHC5 positions it closer to it substrate, δ-catenin, resulting in the association of this enzyme/substrate pair and the palmitoylation of δ-catenin. Importantly, palmitoylated δ- catenin that is bound to DHHC5 on REs is then driven into synaptic spines where it has been shown to associate with cadherin, stabilize surface cadherin and AMPARs, and increase synapse efficacy282. Together, this work uncovers a novel mechanism by which changes in synaptic activity can rapidly regulate protein palmitoylation and translocation of synaptic substrate proteins. 133

DHHC proteins have been shown to traffic between different subcellular membrane compartments, providing an additional layer of control to the palmitoylation of their substrates241-243. In non-neuronal cells, DHHCs were thought to exclusively reside at the somatic

Golgi in order to generate the bulk flow of substrate proteins toward cellular membranes, while unidentified palmitoyl-protein thioesterase enzymes were suggested to maintain the equilibrium of substrate trafficking through local depalmitoylation225. However, some DHHC proteins are known to localize to dendritic compartments far from the neuronal cell body242. In the case of

DHHC2, dendritic localization leads to the rapid palmitoylation of PSD-95 in response to extracellular signals such as activity242. It was previously unclear how distally localized DHHC proteins could regulate palmitoylation in response to changes in neuronal activity. Using an ABE assay, we observed no change in the autopalmitoylation levels of DHHC5 following activity, however the acyl-biotin switch enables quantification of the net palmitoylation levels of DHHC5 and may not reflect subtle changes in palmitate attachment to various cysteines within its cysteine-rich domain270. Our observations that activity does not impact the enzymatic activity of

DHHC5 but alters its subcellular localization to position it in the proximity of its substrate provides a novel mechanism underlying how neurons use the trafficking of DHHC proteins to rapidly control localized palmitoylation.

The endosomal trafficking of surface proteins is a tightly controlled process that can be modified by synaptic activity119, 139, 188, 277. Indeed, TfR-positive REs rapidly move in and out of spines following changes in synaptic activity, resulting in the insertion of AMPARs to the synaptic membrane following cLTP118, 119, 278 and the endocytosis of AMPARs following cLTD139, 140, 286. Specific synaptic proteins can therefore be targeted for surface insertion or internalization, while the overall rate of endosomal membrane cycling does not change188, 286. 134

Our demonstration of activity-induced internalization of DHHC5 from the plasma membrane and trafficking out of and back into spines on REs is similar to activity-mediated turnover of

AMPARs119, 286 but with a different timescale and magnitude. Thus, the endocytic trafficking of

DHHC5 represents a distinct signaling pathway that is important for the palmitoylation and synaptic delivery of substrate proteins.

Palmitoylated Grip1b and δ-catenin have distinct but potentially cooperative roles in regulating the surface localization of AMPARs220, 282. It is therefore intriguing that DHHC5 palmitoylates δ-catenin in an activity-dependent manner whereas Grip1b palmitoylation is not activity-dependent242, and the trafficking of DHHC5 may provide an explanation. Under basal conditions, Grip1b is enriched in the postsynaptic compartment220, and DHHC5-mediated palmitoylation targets it to dendritic REs where it continually traffics within synapses and maintains AMPAR delivery220, 289, 290. This is in agreement with the fast rate of palmitate turnover on Grip1b, which exhibits a half-life of approximately 35 min220, but is longer than the transient efflux of DHHC5 out of the synapse following activity. The majority of δ-catenin is localized to the dendritic shaft under basal conditions, largely separated from its PAT until activity-induced trafficking of DHHC5 out of the synapse. The trafficking of palmitoylated δ- catenin into spines results in the surface stabilization N-cadherin and AMPARs282. DHHC5 may regulate synapse plasticity by constitutively palmitoylating Grip1b for continuous synaptic

AMPAR delivery, and increasing palmitoylation of δ-catenin following activity for surface

AMPAR stabilization.

Mutations in the genes encoding DHHC5 and its homologue DHHC8 have repeatedly been linked to schizophrenia and other neuropsychiatric disorders91-93, 248, 249, 254. Furthermore,

135

both DHHC5247 and δ-catenin206, 210 have been implicated in higher-order brain functions like learning and memory, raising the possibility that aberrant functioning of this enzyme-substrate pair may disrupt synapse plasticity and contribute to the pathology underlying these conditions.

Our study demonstrates that the precise activity-regulated trafficking of DHHC5 is required for the synaptic targeting of δ-catenin, and may be an essential component of synapse function and learning and memory.

136

Chapter 4: Conclusion

The work presented in this dissertation demonstrates a novel mechanism underlying the coordinated enhancement of synapse adhesion, structure, and efficacy that is correlated with memory acquisition. This mechanistic pathway is initiated by the activation of synaptic

NMDARs, leading to the endocytosis of DHHC5 from the synaptic membrane and trafficking into the shaft where it associates with δ-catenin, catalyzes its palmitoylation, and drives the recruitment of palmitoylated δ-catenin into the synapse (Fig. 4.1). Palmitoylated δ-catenin then traffics to the synaptic membrane where it binds N-cadherin, resulting in its stabilization and the enlargement of dendritic spine heads, and enhances surface AMPAR levels and synapse efficacy by linking the cadherin-adhesion complex to receptor-scaffold complexes (Fig. 4.1).

Palmitoylation of δ-catenin and its association with synaptic N-cadherin were significantly increased in the hippocampus of mice that acquired a contextual fear memory. Together, the data presented in this dissertation suggests that the activity-induced palmitoylation of δ-catenin by

DHHC5 is involved in memory formation, and provides insight into the role of protein palmitoylation in driving plastic changes at the synapse.

137

Figure 4.1. Model of the Palmitoylation of δ-Catenin by DHHC5 Driving Activity-Induced Enhancements of Synapse Adhesion, Structure, and Efficacy. Under basal conditions, δ-catenin is predominantly localized to the postsynaptic cytosol and dendritic shafts. A large fraction of DHHC5 is localized to the postsynaptic membrane, and recycling endosomes (REs) constitutively cycle AMPARs and N-cadherin between the synaptic membrane and endosomal compartments. Within 0-10 min after LTP activity stimulation, DHHC5 is endocytosed from the synaptic membrane onto REs, traffics to the dendritic shaft and associates with δ-catenin to catalyze its palmitoylation. Palmitoylated δ-catenin is anchored to the lipid membrane of REs trafficking back into the postsynaptic compartment, leading to the recruitment of palmitoylated δ-catenin to the synaptic membrane. From 10-40 min after LTP, Palmitoylated δ-catenin binds synaptic N-cadherin, stabilizing the cadherin adhesion complex at the cell surface, and leading to the enlargement of dendritic spine heads and increased surface levels of AMPARs to enhance synapse efficacy. DHHC5 and AMPARs are delivered by REs and accumulate at the postsynaptic membrane. Stable cadherin-catenin complexes at the surface form clusters in trans and cis, cooperating to enhance adhesive strength. Arrows denote direction of protein trafficking, and line weight indicates the relative magnitude of trafficking.

4.1 Formation of a Stable Synaptic Adherens Junction

This dissertation has demonstrated that cadherin-based adhesion at the synapse is a critical determinant of synapse strength and morphology, based upon the previous studies discussed in the introduction combined with the work herein. We demonstrated that mutation of lysine 581 of δ-catenin abolishes its ability to associate with cadherin, providing the most detailed insight into the binding interface between δ-catenin and N-cadherin to date. However, many questions remain unanswered pertaining to the assembly of a stable cadherin-catenin complex at the cell surface that can participate in strong intercellular adhesion. For example, through what mechanism does binding of palmitoylated δ-catenin to the juxtamembrane domain 138

of N-cadherin result in stabilized adherens junctions at the synaptic membrane? Evidence from structural biology studies point to a potential role for δ-catenin in assembling cadherin cis interfaces in the brain.

The formation of surface cadherins adhered in trans absolutely requires the association of

β-catenin at cadherin’s C-terminus279, which dynamically interacts with an α-catenin–EPLIN complex to regulate actin dynamics155, 291, 292, and control spine morphology159, 175, 188. However, cadherin-mediated adhesion among distinct membranes has been observed in silico as a continuous lattice of cadherin molecules adhered in both trans and cis153. Mutation of the extracellular residues participating in the cis interface among cadherins in the same membrane do not abolish trans adhesion, but completely disrupt the formation of stable adherens junctions293. This indicates that trans and cis interfaces between cadherin molecules cooperate to form organized adhesive arrays between membranes294.

Biological evidence has demonstrated that the cadherin juxtamembrane domain is essential for cadherin clustering in cis, the organization of cadherin-adhesive junctions between membranes154, 293, and dendritic spine morphology193. The cadherin juxtamembrane domain contains a motif that is both recognized by p120ctn and δ-catenin and which contains residues that serve as an endocytic signal261. The binding of p120ctn in non-neuronal cells masks this region from access by endocytic factors, thereby preventing cadherin endocytosis261. δ-catenin and p120ctn bind the same core region of the juxtamembrane domain295, 296, however a proteomic screen demonstrated that δ-catenin is the exclusive binding partner of N-cadherin in neurons198. Thus, the binding of δ-catenin likely promotes cadherin surface stability in a similar

139

manner to p120ctn by preventing the association of endocytic factors, but further evidence points to an additional role for δ-catenin in regulating cadherin cis interactions.

The generation of a δ-catenin K581M cadherin-binding mutant construct in the work presented in chapter 2 (Fig. 2.4d,e) demonstrates that not only do p120ctn and δ-catenin bind the same region of cadherins, they also bind using a highly similar structural interface. The crystal structure of p120ctn bound to the juxtamembrane domain of E-cadherin revealed a critical electrostatic interaction involving lysine 401 of p120ctn, mutation of which completely inhibited p120ctn binding193, and we identified lysine 581 as the conserved residue in mouse δ-catenin.

Interestingly, the in silico complex of p120ctn bound to the cadherin juxtamembrane domain was observed forming head-to-tail multimers in cis with a periodicity that was similar to that observed in native cadherin junctions297, suggesting a biological role for p120ctn in assembling cis clusters of cadherins. δ-catenin exclusively binds N-cadherin in the brain by the same mechanism as p120ctn in other tissues, and loss of cadherin-binding by δ-catenin completely inhibited activity-induced spine enlargement and AMPAR insertion (Figs. 2.7-2.8). Moreover, ablation of δ-catenin in vivo significantly reduces synaptic N-cadherin levels201. Therefore, I hypothesize that the binding of δ-catenin stabilizes cadherins by assembling cis clusters at the synaptic membrane. This may organize synaptic adherens junctions in such a manner as to permit spine enlargement and increased AMPAR surface insertion. More work is needed to test this hypothesis.

It is tempting to speculate that the palmitoylation of δ-catenin confers a competitive advantage over p120ctn for cadherin-binding in neurons. It would be of great interest to examine

140

how palmitoylated δ-catenin may contribute to the formation of cis cadherin clusters at the synaptic membrane.

4.2 Linkage of the Cadherin Adhesion Complex with the AMPAR-Scaffold Complex by

Palmitoylated δ-Catenin

δ-catenin palmitoylation and cadherin-binding functions were necessary for activity- induced stabilization of synaptic cadherins, spine remodeling, and increased surface AMPAR content. The PDZ-binding motif of δ-catenin has also been shown to control some facets of spine morphology and AMPAR surface levels178, 184, 205. The work presented in this dissertation indicates that palmitoylated δ-catenin links cadherin-adhesion complexes with AMPA receptor- scaffold complexes at the synaptic membrane. This linkage may be essential for coordinating activity-induced changes in synapse adhesion and structure with changes in synapse efficacy.

N-cadherin can directly interact with the extracellular domain of the GluA2 subunit of

AMPARs, resulting in the surface stabilization of both molecules186, and can also interact through an intracellular δ-catenin–GRIP1/ABP linkage184, 205. Palmitoylated δ-catenin may enhance the surface stabilization of these two complexes by simply mediating the intracellular linkage, promoting the direct extracellular interaction, or a combination of both interactions, but an exact picture of the two linked complexes has not been shown.

Intriguingly, recent work has shown that GRIP1 directly binds a C-terminal region of N- cadherin, and mediates the trafficking of N-cadherin and AMPARs in the same vesicles for delivery to synapses298, adding yet another layer to the already complicated picture of the two linked complexes. Surprisingly, the GRIP1 binding site on N-cadherin is immediately adjacent to

141

the juxtamembrane domain core region bound by δ-catenin, and the impact of GRIP1–N- cadherin binding on synaptic surface levels of N-cadherin was not shown298. Our demonstration of the δ-catenin–N-cadherin binding mechanism together with the previous structural data of p120ctn–E-cadherin binding193 suggests that the association of δ-catenin with N-cadherin may provide steric hindrance to antagonize the direct GRIP1–N-cadherin interaction. Furthermore, the PDZ-binding motif of δ-catenin could potentially sequester GRIP1 from binding N-cadherin directly. The work presented in this dissertation has provided insight into the importance of linking synapse adhesion and receptor localization for coordinated regulation of synapse function. More work is needed to establish exactly how these complexes interact to promote synapse strengthening.

4.3 Transsynaptic Strengthening and Synaptic Tagging by Palmitoylated δ-Catenin

The current study demonstrates that chemical LTP stimulation enhances the palmitoylation of δ-catenin by DHHC5, and leads to changes in the postsynaptic compartment of the synapse. However, the question of whether the recruitment of palmitoylated δ-catenin to the postsynaptic membrane leads to transsynaptic strengthening and synapse growth in both compartments remains unanswered.

LTP is comprised of multiple phases; the induction and expression phases occur within an approximate window of 0-60 min after stimulation and primarily involve protein modifications and trafficking, whereas the late phase of LTP requires protein synthesis and involves transsynaptic growth299, 300. Interestingly, LTP induction is thought to only enhance the probability for a long-lasting enhancement of synapse efficacy, but not the commitment of

142

synapses to such a persistent change301. Indeed, previous studies have demonstrated that the early and late phases of LTP are dissociable and can become uncoupled302, 303. A “tag and capture” model of synaptic strengthening suggests that strong stimulation of a synapse leads to its

“tagging” by the addition of a protein or complex, which then functions to “capture” newly synthesized proteins and leads to late LTP specifically at the activated and tagged synapse299.

This model has been put forward to explain how activated synapses transition into persistent strengthening, and evidence suggests that an N-cadherin–δ-catenin complex may serve as a tag.

Under basal conditions, δ-catenin is largely localized to the dendritic shaft, but following synaptic activation it is rapidly palmitoylated and recruited to the postsynaptic membrane where it binds N-cadherin. N-cadherin has been shown to be particularly important for the late-phase of

LTP259, and LTP induction results in the synthesis of new N-cadherin and its subsequent synaptic insertion173, 259. Moreover, expression of a dominant-negative N-cadherin construct in a postsynaptic cell reduced presynaptic SV release probability304, whereas the forced stabilization of the cadherin–β-catenin interaction in spines enhanced presynaptic SV clustering165. Thus, the recruitment of postsynaptic N-cadherin leads to presynaptic enhancement through a transsynaptic mechanism. The results presented here suggest that stabilization of N-cadherin at the postsynaptic membrane through the binding of palmitoylated δ-catenin may potentially serve as a tag that can lead to long-term transsynaptic growth. It would be of great interest to examine if the trafficking of palmitoylated δ-catenin to a single activated synapse acts as a tag that can lead to the capture of newly synthesized N-cadherin and strengthening in both the pre- and postsynaptic compartments.

143

4.4 Activity-Regulated Trafficking of DHHC Proteins

Synaptic activity resulted in the rapid endocytosis and trafficking of DHHC5 into the dendritic shaft, followed by its delivery of palmitoylated δ-catenin to the synapse, and its own accumulation at the synaptic membrane. The current study is among the first to demonstrate the occurrence of activity-directed dynamic subcellular localization of DHHC proteins in neurons, and raises the question of what domains within the DHHC proteins specify their membrane targeting and sensitivity to activity.

DHHC2 had previously been observed colocalized with dendritic vesicles as well as the plasma membrane in neurons242. A slow homeostatic enhancement of synapse strength through activity blockade resulted in the increased insertion of DHHC2 into the postsynaptic membrane, where it locally maintained palmitoylation of PSD-95241, 242, but no cycling between membranes was observed. Another study demonstrated that DHHC2 cycles between different membrane compartments in non-neuronal cells, and that this depended upon its cytoplasmic C-terminus243.

This raises the possibility that distinct domains within the N- and C-termini of DHHC proteins confer specialized membrane trafficking and cycling properties upon particular family members.

DHHC5 is among a small group of DHHCs that contain a PDZ-binding motif at its C- terminus217. Synaptic activation resulted in the internalization of DHHC5 from the plasma membrane within 2 min of stimulation, suggesting that rapid NMDAR-mediated influx of Ca2+ triggered its membrane removal. It would be interesting to speculate that the association or dissociation of DHHC5 with a postsynaptic PDZ-domain protein initiates its membrane removal.

DHHC5 is strongly associated with PSD-95247, and although its sufficiency to palmitoylate it has been shown in one study236, another report demonstrated that DHHC2 and not DHHC5 is a

144

primary PAT for PSD-95235. Therefore, the association between DHHC5 and PSD-95 may primarily serve another function, such as to stabilize DHHC5 at the postsynaptic membrane.

Intriguingly, LTP induction results in the rapid and transient loss of PSD-95 from spine heads, followed by its trafficking back into spines over a similar time course as that observed for

DHHC585. Phosphorylation of PSD-95 by CaMKII, an enzyme that is rapidly activated by

NMDAR-mediated Ca2+ flux113, drove the activity-induced trafficking of PSD-95 out of the synapse85. It would be of great interest to examine if dissociation of DHHC5 from PSD-95 leads to its plasma membrane removal and targeting to recycling endosome membranes in an activity- dependent manner. More work is needed to identify the mechanisms controlling the subcellular localization of DHHC proteins in neurons.

4.5 Significance and Limitations

The work presented in this dissertation answers several questions pertaining to the cell biology of the synapse. First, this study provides a molecular mechanism through which activity enhances the structural and functional strength of the synapse. Second, this research has demonstrated how adhesion and receptor systems are coordinately linked and cooperate at the synapse to control plasticity. Third, this study provides further insight into the molecular components that contribute to the assembly of functional intercellular adherens junctions, which is important for general cell biology and biomedical sciences. Fourth, this work elucidates how neurons use the palmitoylation machinery to efficiently deliver protein cargo to activated synaptic compartments in a rapid manner, which is also important for the field of general cell biology. Fifth, in a larger context, the current study identifies a signaling pathway that is potentially critically involved with memory formation. 145

A limitation of the work described in this dissertation is that the majority of experiments were performed in vitro using dissociated neuronal cultures, and the findings may not completely represent all aspects of synapse function and plasticity in vivo. Furthermore, an increase in the palmitoylation of δ-catenin in the hippocampus was correlated with the acquisition of contextual memories, but a casual link has not yet been established. Future work involving the use of ex vivo hippocampal slices and the generation of transgenic knockin mice expressing the palmitoylation-deficient mutant form of δ-catenin could fully elucidate the importance of this signaling pathway in neural circuits and cognitive function. Nevertheless, δ-catenin and DHHC5 have both been implicated in cognitive performance206, 210, 247, and linked with neurological disorders and disabilities91, 209, 248, 249, 262, 263, suggesting that this enzyme-substrate pair are critically involved in brain function and plasticity. Further investigations stemming from this work will be beneficial to a better understanding of learning and memory.

146

Bibliography

1. Scoville, W.B. & Milner, B. Loss of recent memory after bilateral hippocampal lesions. Journal of neurology, neurosurgery, and psychiatry 20, 11-21 (1957). 2. Neves, G., Cooke, S.F. & Bliss, T.V. Synaptic plasticity, memory and the hippocampus: a neural network approach to causality. Nat Rev Neurosci 9, 65-75 (2008). 3. Kerchner, G.A. & Nicoll, R.A. Silent synapses and the emergence of a postsynaptic mechanism for LTP. Nat Rev Neurosci 9, 813-825 (2008). 4. Banker, G.A. & Cowan, W.M. Rat hippocampal neurons in dispersed cell culture. Brain research 126, 397-342 (1977). 5. Fletcher, T.L. & Banker, G.A. The establishment of polarity by hippocampal neurons: the relationship between the stage of a cell's development in situ and its subsequent development in culture. Developmental biology 136, 446-454 (1989). 6. Dotti, C.G., Sullivan, C.A. & Banker, G.A. The establishment of polarity by hippocampal neurons in culture. J Neurosci 8, 1454-1468 (1988). 7. Bartlett, W.P. & Banker, G.A. An electron microscopic study of the development of axons and dendrites by hippocampal neurons in culture. I. Cells which develop without intercellular contacts. J Neurosci 4, 1944-1953 (1984). 8. Bartlett, W.P. & Banker, G.A. An electron microscopic study of the development of axons and dendrites by hippocampal neurons in culture. II. Synaptic relationships. J Neurosci 4, 1954-1965 (1984). 9. Takamori, S. VGLUTs: 'exciting' times for glutamatergic research? Neuroscience research 55, 343-351 (2006). 10. Takamori, S., Rhee, J.S., Rosenmund, C. & Jahn, R. Identification of a vesicular glutamate transporter that defines a glutamatergic phenotype in neurons. Nature 407, 189-194 (2000). 11. Ziv, N.E. & Garner, C.C. Cellular and molecular mechanisms of presynaptic assembly. Nat Rev Neurosci 5, 385-399 (2004). 12. Phillips, G.R., et al. The presynaptic particle web: ultrastructure, composition, dissolution, and reconstitution. Neuron 32, 63-77 (2001). 13. Shupliakov, O. & Brodin, L. Recent insights into the building and cycling of synaptic vesicles. Exp Cell Res 316, 1344-1350 (2010). 14. Cingolani, L.A. & Goda, Y. Actin in action: the interplay between the actin cytoskeleton and synaptic efficacy. Nat Rev Neurosci 9, 344-356 (2008). 15. Murthy, V.N. & De Camilli, P. Cell biology of the presynaptic terminal. Annu Rev Neurosci 26, 701-728 (2003). 16. Landis, D.M., Hall, A.K., Weinstein, L.A. & Reese, T.S. The organization of cytoplasm at the presynaptic active zone of a central nervous system synapse. Neuron 1, 201-209 (1988). 17. Hirokawa, N., Sobue, K., Kanda, K., Harada, A. & Yorifuji, H. The cytoskeletal architecture of the presynaptic terminal and molecular structure of synapsin 1. J Cell Biol 108, 111-126 (1989). 18. Brigidi, G.S. & Bamji, S.X. Cadherin-catenin adhesion complexes at the synapse. Curr Opin Neurobiol 21, 208-214 (2011).

147

19. Sun, Y. & Bamji, S.X. beta-Pix modulates actin-mediated recruitment of synaptic vesicles to synapses. J Neurosci 31, 17123-17133 (2011). 20. Collins, M.O., et al. Molecular characterization and comparison of the components and multiprotein complexes in the postsynaptic proteome. Journal of neurochemistry 97 Suppl 1, 16- 23 (2006). 21. Boeckers, T.M. The postsynaptic density. Cell and tissue research 326, 409-422 (2006). 22. Sheng, M. & Hoogenraad, C.C. The postsynaptic architecture of excitatory synapses: a more quantitative view. Annual review of biochemistry 76, 823-847 (2007). 23. Kennedy, M.B. The postsynaptic density at glutamatergic synapses. Trends Neurosci 20, 264-268 (1997). 24. Cho, K.O., Hunt, C.A. & Kennedy, M.B. The rat brain postsynaptic density fraction contains a homolog of the Drosophila discs-large tumor suppressor protein. Neuron 9, 929-942 (1992). 25. Apperson, M.L., Moon, I.S. & Kennedy, M.B. Characterization of densin-180, a new brain-specific synaptic protein of the O-sialoglycoprotein family. J Neurosci 16, 6839-6852 (1996). 26. Kim, E. & Sheng, M. PDZ domain proteins of synapses. Nat Rev Neurosci 5, 771-781 (2004). 27. Kistner, U., Garner, C.C. & Linial, M. Nucleotide binding by the synapse associated protein SAP90. FEBS Lett 359, 159-163 (1995). 28. El-Husseini, A.E., Schnell, E., Chetkovich, D.M., Nicoll, R.A. & Bredt, D.S. PSD-95 involvement in maturation of excitatory synapses. Science 290, 1364-1368 (2000). 29. Elias, G.M., et al. Synapse-specific and developmentally regulated targeting of AMPA receptors by a family of MAGUK scaffolding proteins. Neuron 52, 307-320 (2006). 30. Greger, I.H. & Esteban, J.A. AMPA receptor biogenesis and trafficking. Curr Opin Neurobiol 17, 289-297 (2007). 31. Lu, W., et al. Subunit composition of synaptic AMPA receptors revealed by a single-cell genetic approach. Neuron 62, 254-268 (2009). 32. Wenthold, R.J., Petralia, R.S., Blahos, J., II & Niedzielski, A.S. Evidence for multiple AMPA receptor complexes in hippocampal CA1/CA2 neurons. J Neurosci 16, 1982-1989 (1996). 33. Anggono, V. & Huganir, R.L. Regulation of AMPA receptor trafficking and synaptic plasticity. Curr Opin Neurobiol 22, 461-469 (2012). 34. Shi, S., Hayashi, Y., Esteban, J.A. & Malinow, R. Subunit-specific rules governing AMPA receptor trafficking to synapses in hippocampal pyramidal neurons. Cell 105, 331-343 (2001). 35. Kohler, M., Kornau, H.C. & Seeburg, P.H. The organization of the gene for the functionally dominant alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor subunit GluR-B. J Biol Chem 269, 17367-17370 (1994). 36. Hayashi, Y., et al. Driving AMPA receptors into synapses by LTP and CaMKII: requirement for GluR1 and PDZ domain interaction. Science 287, 2262-2267 (2000). 37. Shi, S.H., et al. Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 284, 1811-1816 (1999). 38. Luscher, C., et al. Role of AMPA receptor cycling in synaptic transmission and plasticity. Neuron 24, 649-658 (1999). 148

39. Noel, J., et al. Surface expression of AMPA receptors in hippocampal neurons is regulated by an NSF-dependent mechanism. Neuron 23, 365-376 (1999). 40. Song, I., et al. Interaction of the N-ethylmaleimide-sensitive factor with AMPA receptors. Neuron 21, 393-400 (1998). 41. Luthi, A., et al. Hippocampal LTD expression involves a pool of AMPARs regulated by the NSF-GluR2 interaction. Neuron 24, 389-399 (1999). 42. Hunt, D.L. & Castillo, P.E. Synaptic plasticity of NMDA receptors: mechanisms and functional implications. Curr Opin Neurobiol 22, 496-508 (2012). 43. Behe, P., et al. Determination of NMDA NR1 subunit copy number in recombinant NMDA receptors. Proceedings. Biological sciences / The Royal Society 262, 205-213 (1995). 44. Cull-Candy, S.G. & Leszkiewicz, D.N. Role of distinct NMDA receptor subtypes at central synapses. Science's STKE : signal transduction knowledge environment 2004, re16 (2004). 45. Dingledine, R., Borges, K., Bowie, D. & Traynelis, S.F. The glutamate receptor ion channels. Pharmacological reviews 51, 7-61 (1999). 46. Williams, K., Russell, S.L., Shen, Y.M. & Molinoff, P.B. Developmental switch in the expression of NMDA receptors occurs in vivo and in vitro. Neuron 10, 267-278 (1993). 47. Sheng, M., Cummings, J., Roldan, L.A., Jan, Y.N. & Jan, L.Y. Changing subunit composition of heteromeric NMDA receptors during development of rat cortex. Nature 368, 144-147 (1994). 48. Watanabe, M., Inoue, Y., Sakimura, K. & Mishina, M. Developmental changes in distribution of NMDA receptor channel subunit mRNAs. Neuroreport 3, 1138-1140 (1992). 49. Monyer, H., Burnashev, N., Laurie, D.J., Sakmann, B. & Seeburg, P.H. Developmental and regional expression in the rat brain and functional properties of four NMDA receptors. Neuron 12, 529-540 (1994). 50. Sanz-Clemente, A., Matta, J.A., Isaac, J.T. & Roche, K.W. Casein kinase 2 regulates the NR2 subunit composition of synaptic NMDA receptors. Neuron 67, 984-996 (2010). 51. Tovar, K.R. & Westbrook, G.L. Mobile NMDA receptors at hippocampal synapses. Neuron 34, 255-264 (2002). 52. Lonze, B.E. & Ginty, D.D. Function and regulation of CREB family transcription factors in the nervous system. Neuron 35, 605-623 (2002). 53. Dau, A., Gladding, C.M., Sepers, M.D. & Raymond, L.A. Chronic blockade of extrasynaptic NMDA receptors ameliorates synaptic dysfunction and pro-death signaling in Huntington disease transgenic mice. Neurobiology of disease 62, 533-542 (2014). 54. Parsons, M.P. & Raymond, L.A. Extrasynaptic NMDA receptor involvement in central nervous system disorders. Neuron 82, 279-293 (2014). 55. Hardingham, G.E. & Bading, H. Synaptic versus extrasynaptic NMDA receptor signalling: implications for neurodegenerative disorders. Nat Rev Neurosci 11, 682-696 (2010). 56. Tada, T. & Sheng, M. Molecular mechanisms of dendritic spine morphogenesis. Curr Opin Neurobiol 16, 95-101 (2006). 57. Ziv, N.E. & Smith, S.J. Evidence for a role of dendritic filopodia in synaptogenesis and spine formation. Neuron 17, 91-102 (1996). 58. Zuo, Y., Yang, G., Kwon, E. & Gan, W.B. Long-term sensory deprivation prevents dendritic spine loss in primary somatosensory cortex. Nature 436, 261-265 (2005).

149

59. Zuo, Y., Lin, A., Chang, P. & Gan, W.B. Development of long-term dendritic spine stability in diverse regions of cerebral cortex. Neuron 46, 181-189 (2005). 60. Portera-Cailliau, C., Pan, D.T. & Yuste, R. Activity-regulated dynamic behavior of early dendritic protrusions: evidence for different types of dendritic filopodia. J Neurosci 23, 7129- 7142 (2003). 61. Sala, C., et al. Regulation of dendritic spine morphology and synaptic function by Shank and Homer. Neuron 31, 115-130 (2001). 62. Sekino, Y., Kojima, N. & Shirao, T. Role of actin cytoskeleton in dendritic spine morphogenesis. Neurochem Int 51, 92-104 (2007). 63. Okabe, S., Miwa, A. & Okado, H. Spine formation and correlated assembly of presynaptic and postsynaptic molecules. J Neurosci 21, 6105-6114 (2001). 64. Hayashi, K. & Shirao, T. Change in the shape of dendritic spines caused by overexpression of drebrin in cultured cortical neurons. J Neurosci 19, 3918-3925 (1999). 65. Takahashi, H., et al. Drebrin-dependent actin clustering in dendritic filopodia governs synaptic targeting of postsynaptic density-95 and dendritic spine morphogenesis. J Neurosci 23, 6586-6595 (2003). 66. Hering, H. & Sheng, M. Activity-dependent redistribution and essential role of cortactin in dendritic spine morphogenesis. J Neurosci 23, 11759-11769 (2003). 67. Zhang, W. & Benson, D.L. Stages of synapse development defined by dependence on F- actin. J Neurosci 21, 5169-5181 (2001). 68. Star, E.N., Kwiatkowski, D.J. & Murthy, V.N. Rapid turnover of actin in dendritic spines and its regulation by activity. Nat Neurosci 5, 239-246 (2002). 69. Pollard, T.D. & Borisy, G.G. Cellular motility driven by assembly and disassembly of actin filaments. Cell 112, 453-465 (2003). 70. Okamoto, K., Nagai, T., Miyawaki, A. & Hayashi, Y. Rapid and persistent modulation of actin dynamics regulates postsynaptic reorganization underlying bidirectional plasticity. Nat Neurosci 7, 1104-1112 (2004). 71. Honkura, N., Matsuzaki, M., Noguchi, J., Ellis-Davies, G.C. & Kasai, H. The subspine organization of actin fibers regulates the structure and plasticity of dendritic spines. Neuron 57, 719-729 (2008). 72. Frost, N.A., Shroff, H., Kong, H., Betzig, E. & Blanpied, T.A. Single-molecule discrimination of discrete perisynaptic and distributed sites of actin filament assembly within dendritic spines. Neuron 67, 86-99 (2010). 73. Allison, D.W., Gelfand, V.I., Spector, I. & Craig, A.M. Role of actin in anchoring postsynaptic receptors in cultured hippocampal neurons: differential attachment of NMDA versus AMPA receptors. J Neurosci 18, 2423-2436 (1998). 74. Kasai, H., Matsuzaki, M., Noguchi, J., Yasumatsu, N. & Nakahara, H. Structure-stability- function relationships of dendritic spines. Trends Neurosci 26, 360-368 (2003). 75. Nusser, Z., et al. Cell type and pathway dependence of synaptic AMPA receptor number and variability in the hippocampus. Neuron 21, 545-559 (1998). 76. Matsuzaki, M., et al. Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons. Nat Neurosci 4, 1086-1092 (2001). 77. Svoboda, K., Tank, D.W. & Denk, W. Direct measurement of coupling between dendritic spines and shafts. Science 272, 716-719 (1996).

150

78. Majewska, A., Tashiro, A. & Yuste, R. Regulation of spine calcium dynamics by rapid spine motility. J Neurosci 20, 8262-8268 (2000). 79. Araya, R., Jiang, J., Eisenthal, K.B. & Yuste, R. The spine neck filters membrane potentials. Proc Natl Acad Sci U S A 103, 17961-17966 (2006). 80. Tonnesen, J., Katona, G., Rozsa, B. & Nagerl, U.V. Spine neck plasticity regulates compartmentalization of synapses. Nat Neurosci 17, 678-685 (2014). 81. Penzes, P., Cahill, M.E., Jones, K.A., VanLeeuwen, J.E. & Woolfrey, K.M. Dendritic spine pathology in neuropsychiatric disorders. Nat Neurosci 14, 285-293 (2011). 82. Sudhof, T.C. Neuroligins and neurexins link synaptic function to cognitive disease. Nature 455, 903-911 (2008). 83. Chih, B., Afridi, S.K., Clark, L. & Scheiffele, P. Disorder-associated mutations lead to functional inactivation of neuroligins. Human molecular genetics 13, 1471-1477 (2004). 84. Roussignol, G., et al. Shank expression is sufficient to induce functional dendritic spine synapses in aspiny neurons. J Neurosci 25, 3560-3570 (2005). 85. Steiner, P., et al. Destabilization of the postsynaptic density by PSD-95 serine 73 phosphorylation inhibits spine growth and synaptic plasticity. Neuron 60, 788-802 (2008). 86. Durand, C.M., et al. Mutations in the gene encoding the synaptic scaffolding protein SHANK3 are associated with autism spectrum disorders. Nat Genet 39, 25-27 (2007). 87. Berkel, S., et al. Mutations in the SHANK2 synaptic scaffolding gene in autism spectrum disorder and mental retardation. Nat Genet 42, 489-491 (2010). 88. Bacchelli, E., et al. Screening of nine candidate genes for autism on chromosome 2q reveals rare nonsynonymous variants in the cAMP-GEFII gene. Molecular psychiatry 8, 916-924 (2003). 89. Woolfrey, K.M., et al. Epac2 induces synapse remodeling and depression and its disease- associated forms alter spines. Nat Neurosci 12, 1275-1284 (2009). 90. Bourgeron, T. A synaptic trek to autism. Curr Opin Neurobiol 19, 231-234 (2009). 91. Fromer, M., et al. De novo mutations in schizophrenia implicate synaptic networks. Nature 506, 179-184 (2014). 92. Stark, K.L., et al. Altered brain microRNA biogenesis contributes to phenotypic deficits in a 22q11-deletion mouse model. Nat Genet 40, 751-760 (2008). 93. Mukai, J., et al. Palmitoylation-dependent neurodevelopmental deficits in a mouse model of 22q11 microdeletion. Nat Neurosci 11, 1302-1310 (2008). 94. St Clair, D., et al. Association within a family of a balanced autosomal translocation with major mental illness. Lancet 336, 13-16 (1990). 95. Schumacher, J., et al. The DISC locus and schizophrenia: evidence from an association study in a central European sample and from a meta-analysis across different European populations. Human molecular genetics 18, 2719-2727 (2009). 96. Hayashi-Takagi, A., et al. Disrupted-in-Schizophrenia 1 (DISC1) regulates spines of the glutamate synapse via Rac1. Nat Neurosci 13, 327-332 (2010). 97. Hill, J.J., Hashimoto, T. & Lewis, D.A. Molecular mechanisms contributing to dendritic spine alterations in the prefrontal cortex of subjects with schizophrenia. Molecular psychiatry 11, 557-566 (2006). 98. Bettens, K., et al. Follow-up study of susceptibility loci for Alzheimer's disease and onset age identified by genome-wide association. Journal of Alzheimer's disease : JAD 19, 1169-1175 (2010). 151

99. Sleegers, K., et al. The pursuit of susceptibility genes for Alzheimer's disease: progress and prospects. Trends in genetics : TIG 26, 84-93 (2010). 100. Ji, Y., et al. Apolipoprotein E isoform-specific regulation of dendritic spine morphology in apolipoprotein E transgenic mice and Alzheimer's disease patients. Neuroscience 122, 305- 315 (2003). 101. Dumanis, S.B., et al. ApoE4 decreases spine density and dendritic complexity in cortical neurons in vivo. J Neurosci 29, 15317-15322 (2009). 102. Bu, D.F., et al. Two human glutamate decarboxylases, 65-kDa GAD and 67-kDa GAD, are each encoded by a single gene. Proc Natl Acad Sci U S A 89, 2115-2119 (1992). 103. McIntire, S.L., Reimer, R.J., Schuske, K., Edwards, R.H. & Jorgensen, E.M. Identification and characterization of the vesicular GABA transporter. Nature 389, 870-876 (1997). 104. Connolly, C.N. & Wafford, K.A. The Cys-loop superfamily of ligand-gated ion channels: the impact of receptor structure on function. Biochem Soc Trans 32, 529-534 (2004). 105. Collingridge, G.L., Isaac, J.T. & Wang, Y.T. Receptor trafficking and synaptic plasticity. Nat Rev Neurosci 5, 952-962 (2004). 106. Kneussel, M. & Loebrich, S. Trafficking and synaptic anchoring of ionotropic inhibitory neurotransmitter receptors. Biology of the cell / under the auspices of the European Cell Biology Organization 99, 297-309 (2007). 107. Kneussel, M. & Betz, H. Receptors, gephyrin and gephyrin-associated proteins: novel insights into the assembly of inhibitory postsynaptic membrane specializations. J Physiol 525 Pt 1, 1-9 (2000). 108. Kneussel, M., et al. Loss of postsynaptic GABA(A) receptor clustering in gephyrin- deficient mice. J Neurosci 19, 9289-9297 (1999). 109. Bliss, T.V. & Lomo, T. Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol 232, 331- 356 (1973). 110. Hebb, D.O. The organization of behavior; a neuropsychological theory (Wiley, New York,, 1949). 111. Wigstrom, H., Gustafsson, B., Huang, Y.Y. & Abraham, W.C. Hippocampal long-term potentiation is induced by pairing single afferent volleys with intracellularly injected depolarizing current pulses. Acta physiologica Scandinavica 126, 317-319 (1986). 112. Kandel, E.R., Dudai, Y. & Mayford, M.R. The molecular and systems biology of memory. Cell 157, 163-186 (2014). 113. Malenka, R.C. & Bear, M.F. LTP and LTD: an embarrassment of riches. Neuron 44, 5-21 (2004). 114. Lisman, J., Yasuda, R. & Raghavachari, S. Mechanisms of CaMKII action in long-term potentiation. Nat Rev Neurosci 13, 169-182 (2012). 115. Zamanillo, D., et al. Importance of AMPA receptors for hippocampal synaptic plasticity but not for spatial learning. Science 284, 1805-1811 (1999). 116. Passafaro, M., Piech, V. & Sheng, M. Subunit-specific temporal and spatial patterns of AMPA receptor exocytosis in hippocampal neurons. Nat Neurosci 4, 917-926 (2001). 117. Borgdorff, A.J. & Choquet, D. Regulation of AMPA receptor lateral movements. Nature 417, 649-653 (2002).

152

118. Wang, Z., et al. Myosin Vb mobilizes recycling endosomes and AMPA receptors for postsynaptic plasticity. Cell 135, 535-548 (2008). 119. Petrini, E.M., et al. Endocytic trafficking and recycling maintain a pool of mobile surface AMPA receptors required for synaptic potentiation. Neuron 63, 92-105 (2009). 120. Makino, H. & Malinow, R. AMPA receptor incorporation into synapses during LTP: the role of lateral movement and exocytosis. Neuron 64, 381-390 (2009). 121. Patterson, M.A., Szatmari, E.M. & Yasuda, R. AMPA receptors are exocytosed in stimulated spines and adjacent dendrites in a Ras-ERK-dependent manner during long-term potentiation. Proc Natl Acad Sci U S A 107, 15951-15956 (2010). 122. Granger, A.J., Shi, Y., Lu, W., Cerpas, M. & Nicoll, R.A. LTP requires a reserve pool of glutamate receptors independent of subunit type. Nature 493, 495-500 (2013). 123. Engert, F. & Bonhoeffer, T. Dendritic spine changes associated with hippocampal long- term synaptic plasticity. Nature 399, 66-70 (1999). 124. Matsuzaki, M., Honkura, N., Ellis-Davies, G.C. & Kasai, H. Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761-766 (2004). 125. Ehrlich, I. & Malinow, R. Postsynaptic density 95 controls AMPA receptor incorporation during long-term potentiation and experience-driven synaptic plasticity. J Neurosci 24, 916-927 (2004). 126. Kopec, C.D., Li, B., Wei, W., Boehm, J. & Malinow, R. Glutamate receptor exocytosis and spine enlargement during chemically induced long-term potentiation. J Neurosci 26, 2000- 2009 (2006). 127. Kopec, C.D., Real, E., Kessels, H.W. & Malinow, R. GluR1 links structural and functional plasticity at excitatory synapses. J Neurosci 27, 13706-13718 (2007). 128. Lang, C., et al. Transient expansion of synaptically connected dendritic spines upon induction of hippocampal long-term potentiation. Proc Natl Acad Sci U S A 101, 16665-16670 (2004). 129. Harvey, C.D. & Svoboda, K. Locally dynamic synaptic learning rules in pyramidal neuron dendrites. Nature 450, 1195-1200 (2007). 130. Park, M., et al. Plasticity-induced growth of dendritic spines by exocytic trafficking from recycling endosomes. Neuron 52, 817-830 (2006). 131. Tanaka, J., et al. Protein synthesis and neurotrophin-dependent structural plasticity of single dendritic spines. Science 319, 1683-1687 (2008). 132. Tashiro, A. & Yuste, R. Regulation of dendritic spine motility and stability by Rac1 and Rho kinase: evidence for two forms of spine motility. Mol Cell Neurosci 26, 429-440 (2004). 133. Murakoshi, H., Wang, H. & Yasuda, R. Local, persistent activation of Rho GTPases during plasticity of single dendritic spines. Nature 472, 100-104 (2011). 134. Bosch, M., et al. Structural and Molecular Remodeling of Dendritic Spine Substructures during Long-Term Potentiation. Neuron 82, 444-459 (2014). 135. Meyer, D., Bonhoeffer, T. & Scheuss, V. Balance and Stability of Synaptic Structures during Synaptic Plasticity. Neuron 82, 430-443 (2014). 136. Dudek, S.M. & Bear, M.F. Homosynaptic long-term depression in area CA1 of hippocampus and effects of N-methyl-D-aspartate receptor blockade. Proc Natl Acad Sci U S A 89, 4363-4367 (1992). 137. Mulkey, R.M., Herron, C.E. & Malenka, R.C. An essential role for protein phosphatases in hippocampal long-term depression. Science 261, 1051-1055 (1993). 153

138. Mulkey, R.M., Endo, S., Shenolikar, S. & Malenka, R.C. Involvement of a calcineurin/inhibitor-1 phosphatase cascade in hippocampal long-term depression. Nature 369, 486-488 (1994). 139. Beattie, E.C., et al. Regulation of AMPA receptor endocytosis by a signaling mechanism shared with LTD. Nat Neurosci 3, 1291-1300 (2000). 140. Lee, H.K., Kameyama, K., Huganir, R.L. & Bear, M.F. NMDA induces long-term synaptic depression and dephosphorylation of the GluR1 subunit of AMPA receptors in hippocampus. Neuron 21, 1151-1162 (1998). 141. Bredt, D.S. & Nicoll, R.A. AMPA receptor trafficking at excitatory synapses. Neuron 40, 361-379 (2003). 142. Lee, S.H., Liu, L., Wang, Y.T. & Sheng, M. Clathrin adaptor AP2 and NSF interact with overlapping sites of GluR2 and play distinct roles in AMPA receptor trafficking and hippocampal LTD. Neuron 36, 661-674 (2002). 143. Nagerl, U.V., Eberhorn, N., Cambridge, S.B. & Bonhoeffer, T. Bidirectional activity- dependent morphological plasticity in hippocampal neurons. Neuron 44, 759-767 (2004). 144. Zhou, Q., Homma, K.J. & Poo, M.M. Shrinkage of dendritic spines associated with long- term depression of hippocampal synapses. Neuron 44, 749-757 (2004). 145. Wang, X.B., Yang, Y. & Zhou, Q. Independent expression of synaptic and morphological plasticity associated with long-term depression. J Neurosci 27, 12419-12429 (2007). 146. Collingridge, G.L., Kehl, S.J. & McLennan, H. Excitatory amino acids in synaptic transmission in the Schaffer collateral-commissural pathway of the rat hippocampus. J Physiol 334, 33-46 (1983). 147. Morris, R.G., Anderson, E., Lynch, G.S. & Baudry, M. Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature 319, 774-776 (1986). 148. Tsien, J.Z., Huerta, P.T. & Tonegawa, S. The essential role of hippocampal CA1 NMDA receptor-dependent synaptic plasticity in spatial memory. Cell 87, 1327-1338 (1996). 149. Whitlock, J.R., Heynen, A.J., Shuler, M.G. & Bear, M.F. Learning induces long-term potentiation in the hippocampus. Science 313, 1093-1097 (2006). 150. Nabavi, S., et al. Engineering a memory with LTD and LTP. Nature 511, 348-352 (2014). 151. Arikkath, J. & Reichardt, L.F. Cadherins and catenins at synapses: roles in synaptogenesis and synaptic plasticity. Trends Neurosci 31, 487-494 (2008). 152. Shapiro, L., et al. Structural basis of cell-cell adhesion by cadherins. Nature 374, 327-337 (1995). 153. Boggon, T.J., et al. C-cadherin ectodomain structure and implications for cell adhesion mechanisms. Science 296, 1308-1313 (2002). 154. Yap, A.S., Niessen, C.M. & Gumbiner, B.M. The juxtamembrane region of the cadherin cytoplasmic tail supports lateral clustering, adhesive strengthening, and interaction with p120ctn. J Cell Biol 141, 779-789 (1998). 155. Yamada, S., Pokutta, S., Drees, F., Weis, W.I. & Nelson, W.J. Deconstructing the cadherin-catenin-actin complex. Cell 123, 889-901 (2005). 156. Bekirov, I.H., Needleman, L.A., Zhang, W. & Benson, D.L. Identification and localization of multiple classic cadherins in developing rat limbic system. Neuroscience 115, 213-227 (2002). 154

157. Yagi, T. & Takeichi, M. Cadherin superfamily genes: functions, genomic organization, and neurologic diversity. Genes Dev 14, 1169-1180 (2000). 158. Benson, D.L. & Tanaka, H. N-cadherin redistribution during synaptogenesis in hippocampal neurons. J Neurosci 18, 6892-6904 (1998). 159. Togashi, H., et al. Cadherin regulates dendritic spine morphogenesis. Neuron 35, 77-89 (2002). 160. Stan, A., et al. Essential cooperation of N-cadherin and neuroligin-1 in the transsynaptic control of vesicle accumulation. Proc Natl Acad Sci U S A 107, 11116-11121 (2010). 161. Bamji, S.X., et al. Role of beta-catenin in synaptic vesicle localization and presynaptic assembly. Neuron 40, 719-731 (2003). 162. Sun, Y., Aiga, M., Yoshida, E., Humbert, P.O. & Bamji, S.X. Scribble interacts with beta-catenin to localize synaptic vesicles to synapses. Mol Biol Cell 20, 3390-3400 (2009). 163. Bamji, S.X., Rico, B., Kimes, N. & Reichardt, L.F. BDNF mobilizes synaptic vesicles and enhances synapse formation by disrupting cadherin-beta-catenin interactions. J Cell Biol 174, 289-299 (2006). 164. Roura, S., Miravet, S., Piedra, J., Garcia de Herreros, A. & Dunach, M. Regulation of E- cadherin/Catenin association by tyrosine phosphorylation. J Biol Chem 274, 36734-36740 (1999). 165. Murase, S., Mosser, E. & Schuman, E.M. Depolarization drives beta-Catenin into neuronal spines promoting changes in synaptic structure and function. Neuron 35, 91-105 (2002). 166. Lee, S.H., et al. Synapses are regulated by the cytoplasmic tyrosine kinase Fer in a pathway mediated by p120catenin, Fer, SHP-2, and beta-catenin. J Cell Biol 183, 893-908 (2008). 167. Spangler, S.A. & Hoogenraad, C.C. Liprin-alpha proteins: scaffold molecules for synapse maturation. Biochem Soc Trans 35, 1278-1282 (2007). 168. Zhen, M. & Jin, Y. The liprin protein SYD-2 regulates the differentiation of presynaptic termini in C. elegans. Nature 401, 371-375 (1999). 169. Aiga, M., Levinson, J.N. & Bamji, S.X. N-cadherin and neuroligins cooperate to regulate synapse formation in Hippocampal cultures. J Biol Chem (2011). 170. Nishimura, W., Yao, I., Iida, J., Tanaka, N. & Hata, Y. Interaction of synaptic scaffolding molecule and Beta -catenin. J Neurosci 22, 757-765 (2002). 171. Iida, J., Hirabayashi, S., Sato, Y. & Hata, Y. Synaptic scaffolding molecule is involved in the synaptic clustering of neuroligin. Mol Cell Neurosci 27, 497-508 (2004). 172. Okamura, K., et al. Cadherin activity is required for activity-induced spine remodeling. J Cell Biol 167, 961-972 (2004). 173. Mendez, P., De Roo, M., Poglia, L., Klauser, P. & Muller, D. N-cadherin mediates plasticity-induced long-term spine stabilization. J Cell Biol 189, 589-600 (2010). 174. Bozdagi, O., et al. Persistence of coordinated long-term potentiation and dendritic spine enlargement at mature hippocampal CA1 synapses requires N-cadherin. J Neurosci 30, 9984- 9989 (2010). 175. Abe, K., Chisaka, O., Van Roy, F. & Takeichi, M. Stability of dendritic spines and synaptic contacts is controlled by alpha N-catenin. Nat Neurosci 7, 357-363 (2004).

155

176. Okuda, T., Yu, L.M., Cingolani, L.A., Kemler, R. & Goda, Y. beta-Catenin regulates excitatory postsynaptic strength at hippocampal synapses. Proc Natl Acad Sci U S A 104, 13479- 13484 (2007). 177. Elia, L.P., Yamamoto, M., Zang, K. & Reichardt, L.F. p120 catenin regulates dendritic spine and synapse development through Rho-family GTPases and cadherins. Neuron 51, 43-56 (2006). 178. Arikkath, J., et al. Delta-catenin regulates spine and synapse morphogenesis and function in hippocampal neurons during development. J Neurosci 29, 5435-5442 (2009). 179. Kosik, K.S., Donahue, C.P., Israely, I., Liu, X. & Ochiishi, T. Delta-catenin at the synaptic-adherens junction. Trends Cell Biol 15, 172-178 (2005). 180. Martinez, M.C., Ochiishi, T., Majewski, M. & Kosik, K.S. Dual regulation of neuronal morphogenesis by a delta-catenin-cortactin complex and Rho. J Cell Biol 162, 99-111 (2003). 181. Kim, H., et al. Delta-catenin-induced dendritic morphogenesis. An essential role of p190RhoGEF interaction through Akt1-mediated phosphorylation. J Biol Chem 283, 977-987 (2008). 182. Abu-Elneel, K., et al. A delta-catenin signaling pathway leading to dendritic protrusions. J Biol Chem 283, 32781-32791 (2008). 183. Jones, S.B., et al. Glutamate-induced delta-catenin redistribution and dissociation from postsynaptic receptor complexes. Neuroscience 115, 1009-1021 (2002). 184. Silverman, J.B., et al. Synaptic anchorage of AMPA receptors by cadherins through neural plakophilin-related arm protein AMPA receptor-binding protein complexes. J Neurosci 27, 8505-8516 (2007). 185. Xie, Z., et al. Coordination of synaptic adhesion with dendritic spine remodeling by AF-6 and kalirin-7. J Neurosci 28, 6079-6091 (2008). 186. Saglietti, L., et al. Extracellular interactions between GluR2 and N-cadherin in spine regulation. Neuron 54, 461-477 (2007). 187. Segal, M. Dendritic spines and long-term plasticity. Nat Rev Neurosci 6, 277-284 (2005). 188. Tai, C.Y., Mysore, S.P., Chiu, C. & Schuman, E.M. Activity-regulated N-cadherin endocytosis. Neuron 54, 771-785 (2007). 189. Tanaka, H., et al. Molecular modification of N-cadherin in response to synaptic activity. Neuron 25, 93-107 (2000). 190. Xie, Z., et al. Kalirin-7 controls activity-dependent structural and functional plasticity of dendritic spines. Neuron 56, 640-656 (2007). 191. Marambaud, P., et al. A CBP binding transcriptional repressor produced by the PS1/epsilon-cleavage of N-cadherin is inhibited by PS1 FAD mutations. Cell 114, 635-645 (2003). 192. Yasuda, S., et al. Activity-induced protocadherin arcadlin regulates dendritic spine number by triggering N-cadherin endocytosis via TAO2beta and p38 MAP kinases. Neuron 56, 456-471 (2007). 193. Ishiyama, N., et al. Dynamic and static interactions between p120 catenin and E-cadherin regulate the stability of cell-cell adhesion. Cell 141, 117-128 (2010). 194. Fujita, Y., et al. Hakai, a c-Cbl-like protein, ubiquitinates and induces endocytosis of the E-cadherin complex. Nat Cell Biol 4, 222-231 (2002).

156

195. Miyashita, Y. & Ozawa, M. Increased internalization of p120-uncoupled E-cadherin and a requirement for a dileucine motif in the cytoplasmic domain for endocytosis of the protein. J Biol Chem 282, 11540-11548 (2007). 196. Latefi, N.S., Pedraza, L., Schohl, A., Li, Z. & Ruthazer, E.S. N-cadherin prodomain cleavage regulates synapse formation in vivo. Dev Neurobiol 69, 518-529 (2009). 197. Schrick, C., et al. N-cadherin regulates cytoskeletally associated IQGAP1/ERK signaling and memory formation. Neuron 55, 786-798 (2007). 198. Tanaka, H., et al. Linkage of N-cadherin to multiple cytoskeletal elements revealed by a proteomic approach in hippocampal neurons. Neurochem Int 61, 240-250 (2012). 199. Zhou, J., et al. Presenilin 1 interaction in the brain with a novel member of the Armadillo family. Neuroreport 8, 2085-2090 (1997). 200. Zhou, J., et al. Presenilin 1 interaction in the brain with a novel member of the Armadillo family. Neuroreport 8, 1489-1494 (1997). 201. Restituito, S., et al. Synaptic autoregulation by metalloproteases and gamma-secretase. J Neurosci 31, 12083-12093 (2011). 202. Kim, H., Oh, M., Lu, Q. & Kim, K. E-Cadherin negatively modulates delta-catenin- induced morphological changes and RhoA activity reduction by competing with p190RhoGEF for delta-catenin. Biochem Biophys Res Commun 377, 636-641 (2008). 203. Poore, C.P., et al. Cdk5-mediated phosphorylation of delta-catenin regulates its localization and GluR2-mediated synaptic activity. J Neurosci 30, 8457-8467 (2010). 204. Ide, N., et al. Interaction of S-SCAM with neural plakophilin-related Armadillo-repeat protein/delta-catenin. Biochem Biophys Res Commun 256, 456-461 (1999). 205. Misra, C., et al. Regulation of synaptic structure and function by palmitoylated AMPA receptor binding protein. Mol Cell Neurosci 43, 341-352 (2010). 206. Israely, I., et al. Deletion of the neuron-specific protein delta-catenin leads to severe cognitive and synaptic dysfunction. Curr Biol 14, 1657-1663 (2004). 207. Kim, J.S., et al. Presenilin-1 inhibits delta-catenin-induced cellular branching and promotes delta-catenin processing and turnover. Biochem Biophys Res Commun 351, 903-908 (2006). 208. Liauw, J., Nguyen, V., Huang, J., St George-Hyslop, P. & Rozmahel, R. Differential display analysis of presenilin 1-deficient mouse brains. Brain Res Mol Brain Res 109, 56-62 (2002). 209. Medina, M., Marinescu, R.C., Overhauser, J. & Kosik, K.S. Hemizygosity of delta- catenin (CTNND2) is associated with severe mental retardation in cri-du-chat syndrome. Genomics 63, 157-164 (2000). 210. Matter, C., Pribadi, M., Liu, X. & Trachtenberg, J.T. Delta-catenin is required for the maintenance of neural structure and function in mature cortex in vivo. Neuron 64, 320-327 (2009). 211. Kang, R., et al. Neural palmitoyl-proteomics reveals dynamic synaptic palmitoylation. Nature 456, 904-909 (2008). 212. Iwanaga, T., Tsutsumi, R., Noritake, J., Fukata, Y. & Fukata, M. Dynamic protein palmitoylation in cellular signaling. Prog Lipid Res 48, 117-127 (2009). 213. Johnson, D.R., Bhatnagar, R.S., Knoll, L.J. & Gordon, J.I. Genetic and biochemical studies of protein N-myristoylation. Annual review of biochemistry 63, 869-914 (1994).

157

214. Wilcox, C., Hu, J.S. & Olson, E.N. Acylation of proteins with myristic acid occurs cotranslationally. Science 238, 1275-1278 (1987). 215. Zhang, F.L. & Casey, P.J. Protein prenylation: molecular mechanisms and functional consequences. Annual review of biochemistry 65, 241-269 (1996). 216. Pepinsky, R.B., et al. Identification of a palmitic acid-modified form of human Sonic hedgehog. J Biol Chem 273, 14037-14045 (1998). 217. Fukata, Y. & Fukata, M. Protein palmitoylation in neuronal development and synaptic plasticity. Nat Rev Neurosci 11, 161-175 (2010). 218. Fang, C., et al. GODZ-mediated palmitoylation of GABA(A) receptors is required for normal assembly and function of GABAergic inhibitory synapses. J Neurosci 26, 12758-12768 (2006). 219. Hayashi, T., Thomas, G.M. & Huganir, R.L. Dual palmitoylation of NR2 subunits regulates NMDA receptor trafficking. Neuron 64, 213-226 (2009). 220. Thomas, G.M., Hayashi, T., Chiu, S.L., Chen, C.M. & Huganir, R.L. Palmitoylation by DHHC5/8 Targets GRIP1 to Dendritic Endosomes to Regulate AMPA-R Trafficking. Neuron 73, 482-496 (2012). 221. Hayashi, T., Rumbaugh, G. & Huganir, R.L. Differential regulation of AMPA receptor subunit trafficking by palmitoylation of two distinct sites. Neuron 47, 709-723 (2005). 222. El-Husseini Ael, D., et al. Synaptic strength regulated by palmitate cycling on PSD-95. Cell 108, 849-863 (2002). 223. Dejanovic, B., et al. Palmitoylation of gephyrin controls receptor clustering and plasticity of GABAergic synapses. PLoS biology 12, e1001908 (2014). 224. Keith, D.J., et al. Palmitoylation of A-kinase anchoring protein 79/150 regulates dendritic endosomal targeting and synaptic plasticity mechanisms. J Neurosci 32, 7119-7136 (2012). 225. Rocks, O., et al. The palmitoylation machinery is a spatially organizing system for peripheral membrane proteins. Cell 141, 458-471 (2010). 226. Rocks, O., et al. An acylation cycle regulates localization and activity of palmitoylated Ras isoforms. Science 307, 1746-1752 (2005). 227. Fehrenbacher, N., Bar-Sagi, D. & Philips, M. Ras/MAPK signaling from endomembranes. Molecular oncology 3, 297-307 (2009). 228. Brown, D.A. & London, E. Functions of lipid rafts in biological membranes. Annual review of cell and developmental biology 14, 111-136 (1998). 229. Hering, H., Lin, C.C. & Sheng, M. Lipid rafts in the maintenance of synapses, dendritic spines, and surface AMPA receptor stability. J Neurosci 23, 3262-3271 (2003). 230. Mitchell, D.A., Farh, L., Marshall, T.K. & Deschenes, R.J. A polybasic domain allows nonprenylated Ras proteins to function in Saccharomyces cerevisiae. J Biol Chem 269, 21540- 21546 (1994). 231. Bartels, D.J., Mitchell, D.A., Dong, X. & Deschenes, R.J. Erf2, a novel gene product that affects the localization and palmitoylation of Ras2 in Saccharomyces cerevisiae. Mol Cell Biol 19, 6775-6787 (1999). 232. Roth, A.F., Feng, Y., Chen, L. & Davis, N.G. The yeast DHHC cysteine-rich domain protein Akr1p is a palmitoyl transferase. J Cell Biol 159, 23-28 (2002). 233. Mitchell, D.A., Vasudevan, A., Linder, M.E. & Deschenes, R.J. Protein palmitoylation by a family of DHHC protein S-acyltransferases. J Lipid Res 47, 1118-1127 (2006).

158

234. Roth, A.F., et al. Global analysis of protein palmitoylation in yeast. Cell 125, 1003-1013 (2006). 235. Fukata, M., Fukata, Y., Adesnik, H., Nicoll, R.A. & Bredt, D.S. Identification of PSD-95 palmitoylating enzymes. Neuron 44, 987-996 (2004). 236. Ohno, Y., et al. Analysis of substrate specificity of human DHHC protein acyltransferases using a yeast expression system. Mol Biol Cell 23, 4543-4551 (2012). 237. Lai, J. & Linder, M.E. Oligomerization of DHHC protein S-acyltransferases. J Biol Chem 288, 22862-22870 (2013). 238. Huang, K., et al. Neuronal palmitoyl acyl transferases exhibit distinct substrate specificity. FASEB J 23, 2605-2615 (2009). 239. Ren, J., et al. CSS-Palm 2.0: an updated software for palmitoylation sites prediction. Protein Eng Des Sel 21, 639-644 (2008). 240. Ohno, Y., Kihara, A., Sano, T. & Igarashi, Y. Intracellular localization and tissue-specific distribution of human and yeast DHHC cysteine-rich domain-containing proteins. Biochimica et biophysica acta 1761, 474-483 (2006). 241. Fukata, Y., et al. Local palmitoylation cycles define activity-regulated postsynaptic subdomains. J Cell Biol 202, 145-161 (2013). 242. Noritake, J., et al. Mobile DHHC palmitoylating enzyme mediates activity-sensitive synaptic targeting of PSD-95. J Cell Biol 186, 147-160 (2009). 243. Greaves, J., Carmichael, J.A. & Chamberlain, L.H. The palmitoyl transferase DHHC2 targets a dynamic membrane cycling pathway: regulation by a C-terminal domain. Mol Biol Cell 22, 1887-1895 (2011). 244. Duncan, J.A. & Gilman, A.G. A cytoplasmic acyl-protein thioesterase that removes palmitate from G protein alpha subunits and p21(RAS). J Biol Chem 273, 15830-15837 (1998). 245. Yeh, D.C., Duncan, J.A., Yamashita, S. & Michel, T. Depalmitoylation of endothelial nitric-oxide synthase by acyl-protein thioesterase 1 is potentiated by Ca(2+)-calmodulin. J Biol Chem 274, 33148-33154 (1999). 246. Siegel, G., et al. A functional screen implicates microRNA-138-dependent regulation of the depalmitoylation enzyme APT1 in dendritic spine morphogenesis. Nat Cell Biol 11, 705-716 (2009). 247. Li, Y., et al. DHHC5 interacts with PDZ domain 3 of post-synaptic density-95 (PSD-95) protein and plays a role in learning and memory. J Biol Chem 285, 13022-13031 (2010). 248. Fallin, M.D., et al. Genomewide linkage scan for bipolar-disorder susceptibility loci among Ashkenazi Jewish families. Am J Hum Genet 75, 204-219 (2004). 249. Schizophrenia Working Group of the Psychiatric Genomics, C. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421-427 (2014). 250. Huang, K., et al. Huntingtin-interacting protein HIP14 is a palmitoyl transferase involved in palmitoylation and trafficking of multiple neuronal proteins. Neuron 44, 977-986 (2004). 251. Yanai, A., et al. Palmitoylation of huntingtin by HIP14 is essential for its trafficking and function. Nat Neurosci 9, 824-831 (2006). 252. Singaraja, R.R., et al. Altered palmitoylation and neuropathological deficits in mice lacking HIP14. Human molecular genetics 20, 3899-3909 (2011). 253. Milnerwood, A.J., et al. Memory and synaptic deficits in Hip14/DHHC17 knockout mice. Proc Natl Acad Sci U S A 110, 20296-20301 (2013).

159

254. Mukai, J., et al. Evidence that the gene encoding ZDHHC8 contributes to the risk of schizophrenia. Nat Genet 36, 725-731 (2004). 255. Mansouri, M.R., et al. Loss of ZDHHC15 expression in a woman with a balanced translocation t(X;15)(q13.3;cen) and severe mental retardation. Eur J Hum Genet 13, 970-977 (2005). 256. Raymond, F.L., et al. Mutations in ZDHHC9, which encodes a palmitoyltransferase of NRAS and HRAS, cause X-linked mental retardation associated with a Marfanoid habitus. Am J Hum Genet 80, 982-987 (2007). 257. Tarpey, P.S., et al. A systematic, large-scale resequencing screen of X-chromosome coding exons in mental retardation. Nat Genet 41, 535-543 (2009). 258. Lin, D.T., et al. Regulation of AMPA receptor extrasynaptic insertion by 4.1N, phosphorylation and palmitoylation. Nat Neurosci 12, 879-887 (2009). 259. Bozdagi, O., Shan, W., Tanaka, H., Benson, D.L. & Huntley, G.W. Increasing numbers of synaptic puncta during late-phase LTP: N-cadherin is synthesized, recruited to synaptic sites, and required for potentiation. Neuron 28, 245-259 (2000). 260. Tang, L., Hung, C.P. & Schuman, E.M. A role for the cadherin family of cell adhesion molecules in hippocampal long-term potentiation. Neuron 20, 1165-1175 (1998). 261. Nanes, B.A., et al. p120-catenin binding masks an endocytic signal conserved in classical cadherins. J Cell Biol 199, 365-380 (2012). 262. Jun, G., et al. delta-Catenin Is Genetically and Biologically Associated with Cortical Cataract and Future Alzheimer-Related Structural and Functional Brain Changes. PLoS One 7, e43728 (2012). 263. Vrijenhoek, T., et al. Recurrent CNVs disrupt three candidate genes in schizophrenia patients. Am J Hum Genet 83, 504-510 (2008). 264. Radulovic, J., Kammermeier, J. & Spiess, J. Relationship between fos production and classical fear conditioning: effects of novelty, latent inhibition, and unconditioned stimulus preexposure. J Neurosci 18, 7452-7461 (1998). 265. Xie, C., Markesbery, W.R. & Lovell, M.A. Survival of hippocampal and cortical neurons in a mixture of MEM+ and B27-supplemented neurobasal medium. Free Radic Biol Med 28, 665-672 (2000). 266. Durocher, Y., Perret, S. & Kamen, A. High-level and high-throughput recombinant protein production by transient transfection of suspension-growing human 293-EBNA1 cells. Nucleic Acids Res 30, E9 (2002). 267. Lu, W., et al. Activation of synaptic NMDA receptors induces membrane insertion of new AMPA receptors and LTP in cultured hippocampal neurons. Neuron 29, 243-254 (2001). 268. Diering, G.H., Mills, F., Bamji, S.X. & Numata, M. Regulation of dendritic spine growth through activity-dependent recruitment of the brain-enriched Na(+)/H(+) exchanger NHE5. Mol Biol Cell 22, 2246-2257 (2011). 269. Drisdel, R.C., Alexander, J.K., Sayeed, A. & Green, W.N. Assays of protein palmitoylation. Methods 40, 127-134 (2006). 270. Brigidi, G.S. & Bamji, S.X. Detection of Protein Palmitoylation in Cultured Hippocampal Neurons by Immunoprecipitation and Acyl-Biotin Exchange (ABE). J Vis Exp (2013). 271. Tapia, L., et al. Progranulin deficiency decreases gross neural connectivity but enhances transmission at individual synapses. J Neurosci 31, 11126-11132 (2011). 160

272. Musleh, W., Bi, X., Tocco, G., Yaghoubi, S. & Baudry, M. Glycine-induced long-term potentiation is associated with structural and functional modifications of alpha-amino-3- hydroxyl-5-methyl-4-isoxazolepropionic acid receptors. Proc Natl Acad Sci U S A 94, 9451- 9456 (1997). 273. Turrigiano, G.G. The self-tuning neuron: synaptic scaling of excitatory synapses. Cell 135, 422-435 (2008). 274. Arikkath, J., et al. Erbin controls dendritic morphogenesis by regulating localization of delta-catenin. J Neurosci 28, 7047-7056 (2008). 275. El Sayegh, T.Y., et al. Phosphorylation of N-cadherin-associated cortactin by Fer kinase regulates N-cadherin mobility and intercellular adhesion strength. Mol Biol Cell 16, 5514-5527 (2005). 276. Ochiishi, T., Futai, K., Okamoto, K., Kameyama, K. & Kosik, K.S. Regulation of AMPA receptor trafficking by delta-catenin. Mol Cell Neurosci 39, 499-507 (2008). 277. Kennedy, M.J., Davison, I.G., Robinson, C.G. & Ehlers, M.D. Syntaxin-4 defines a domain for activity-dependent exocytosis in dendritic spines. Cell 141, 524-535 (2010). 278. Park, M., Penick, E.C., Edwards, J.G., Kauer, J.A. & Ehlers, M.D. Recycling endosomes supply AMPA receptors for LTP. Science 305, 1972-1975 (2004). 279. Huber, A.H. & Weis, W.I. The structure of the beta-catenin/E-cadherin complex and the molecular basis of diverse ligand recognition by beta-catenin. Cell 105, 391-402 (2001). 280. Nuriya, M. & Huganir, R.L. Regulation of AMPA receptor trafficking by N-cadherin. Journal of neurochemistry 97, 652-661 (2006). 281. Maguschak, K.A. & Ressler, K.J. Beta-catenin is required for memory consolidation. Nat Neurosci 11, 1319-1326 (2008). 282. Brigidi, G.S., et al. Palmitoylation of delta-catenin by DHHC5 mediates activity-induced synapse plasticity. Nat Neurosci 17, 522-532 (2014). 283. Li, D., et al. SAP97 directs NMDA receptor spine targeting and synaptic plasticity. J Physiol 589, 4491-4510 (2011). 284. Rossner, M. & Yamada, K.M. What's in a picture? The temptation of image manipulation. J Cell Biol 166, 11-15 (2004). 285. Jaafari, N., Henley, J.M. & Hanley, J.G. PICK1 Mediates Transient Synaptic Expression of GluA2-Lacking AMPA Receptors during Glycine-Induced AMPA Receptor Trafficking. The Journal of Neuroscience 34, 11618-11630 (2012). 286. Ehlers, M.D. Reinsertion or degradation of AMPA receptors determined by activity- dependent endocytic sorting. Neuron 28, 511-525 (2000). 287. Choy, R.W., et al. Retromer mediates a discrete route of local membrane delivery to dendrites. Neuron 82, 55-62 (2014). 288. Zhang, D., et al. RAB-6.2 and the retromer regulate glutamate receptor recycling through a retrograde pathway. J Cell Biol 196, 85-101 (2012). 289. Dong, H., et al. GRIP: a synaptic PDZ domain-containing protein that interacts with AMPA receptors. Nature 386, 279-284 (1997). 290. Setou, M., et al. Glutamate-receptor-interacting protein GRIP1 directly steers kinesin to dendrites. Nature 417, 83-87 (2002). 291. Drees, F., Pokutta, S., Yamada, S., Nelson, W.J. & Weis, W.I. Alpha-catenin is a molecular switch that binds E-cadherin-beta-catenin and regulates actin-filament assembly. Cell 123, 903-915 (2005). 161

292. Taguchi, K., Ishiuchi, T. & Takeichi, M. Mechanosensitive EPLIN-dependent remodeling of adherens junctions regulates epithelial reshaping. J Cell Biol 194, 643-656 (2011). 293. Harrison, O.J., et al. The extracellular architecture of adherens junctions revealed by crystal structures of type I cadherins. Structure 19, 244-256 (2011). 294. Wu, Y., et al. Cooperativity between trans and cis interactions in cadherin-mediated junction formation. Proc Natl Acad Sci U S A 107, 17592-17597 (2010). 295. Thoreson, M.A., et al. Selective uncoupling of p120(ctn) from E-cadherin disrupts strong adhesion. J Cell Biol 148, 189-202 (2000). 296. Lu, Q., et al. delta-catenin, an adhesive junction-associated protein which promotes cell scattering. J Cell Biol 144, 519-532 (1999). 297. Al-Amoudi, A., Diez, D.C., Betts, M.J. & Frangakis, A.S. The molecular architecture of cadherins in native epidermal desmosomes. Nature 450, 832-837 (2007). 298. Heisler, F.F., et al. GRIP1 interlinks N-cadherin and AMPA receptors at vesicles to promote combined cargo transport into dendrites. Proc Natl Acad Sci U S A 111, 5030-5035 (2014). 299. Sanhueza, M. & Lisman, J. The CaMKII/NMDAR complex as a molecular memory. Molecular brain 6, 10 (2013). 300. Bourne, J.N. & Harris, K.M. Coordination of size and number of excitatory and inhibitory synapses results in a balanced structural plasticity along mature hippocampal CA1 dendrites during LTP. Hippocampus 21, 354-373 (2011). 301. Redondo, R.L. & Morris, R.G. Making memories last: the synaptic tagging and capture hypothesis. Nat Rev Neurosci 12, 17-30 (2011). 302. Frey, U. & Morris, R.G. Synaptic tagging and long-term potentiation. Nature 385, 533- 536 (1997). 303. Redondo, R.L., et al. Synaptic tagging and capture: differential role of distinct calcium/calmodulin kinases in protein synthesis-dependent long-term potentiation. J Neurosci 30, 4981-4989 (2010). 304. Vitureira, N., Letellier, M., White, I.J. & Goda, Y. Differential control of presynaptic efficacy by postsynaptic N-cadherin and beta-catenin. Nat Neurosci 15, 81-89 (2011).

162