Spécialité Neurosciences Et Pharmacologie Présentée Par

THESE DE DOCTORAT DE L’UNIVERSITE PARIS DESCARTES Spécialité Neurosciences et Pharmacologie

Présentée par Delphine Ladarré

Pour l’obtention du grade de DOCTEUR DE L’UNIVERSITE PARIS DESCARTES

Neuronal polarization shapes the targeting and signaling of G-protein coupled receptors (GPCRs):

Type-1 cannabinoid receptors and 5-HT1B serotonin receptors show highly contrasted trafficking and signaling patterns in axons and dendrites.

Soutenance le 3 octobre 2014, devant le jury composé de:

Pr Catherine Marchand-Leroux Présidente du Jury Pr Michel Hamon Rapporteur Dr Régine Hepp Rapporteure Dr Xavier Nicol Examinateur Dr Zsolt Lenkei Directeur de thèse

A mon grand-père, Dr Alain Bullet

Remerciements

Pour commencer, je remercie l’ensemble des membres du jury d’avoir accepté d’évaluer mes travaux de thèse. J’espère que ce manuscrit vous intéressera et vous donnera satisfaction.

Je souhaite remercier Zsolt de m’avoir accueilli dans son équipe dès mon stage de 3A et d’avoir cru en moi en m’offrant la possibilité de continuer en master puis en doctorat. Au sein d’un laboratoire de neurosciences, tu m’as permis d’axer mon projet sur une étude de pharmacologie, un sujet qui me tenait à cœur. Au cours de nos discussions, parfois mouvementées, tu m’as donné l’occasion d’exposer mes théories et tu m’as laissé prendre des initiatives afin de les confirmer ou de les infirmer. L’autonomie que tu m’as accordée m’a permis d’acquérir une bonne organisation et la capacité à gérer un projet.

Je remercie également les membres de l’équipe « Biologie systémique du neurone », devenue par la suite « Dynamique et structure neuronale » qui m’ont grandement aidé au cours de ma thèse. Merci à Alex qui a mis en place le vidéomicroscope, merci pour ton encadrement au cours de mes stages de 3A et de M2 et pour ta grande pédagogie ! Merci à Anne qui m’a appris les techniques essentielles de culture de neurones, de transfection, d’immuno… merci aussi pour ta patiente et ta disponibilité, même après être partie du labo, pour répondre à toutes mes questions (« C’est rangé où ?? »). Merci à Damien pour les grandes discussions, notamment au sujet de Gertrude, mais aussi pour avoir gentiment corrigé une partie de ce manuscrit et pour m’avoir formée à l’électroporation in-utero ainsi qu’à la BM (pour le dernier point, c’est peut-être pas aussi sincère…). Un grand merci à Sophie, arrivée il y a deux ans dans l’équipe, et qui m’a grandement soutenue et aidée dans les moments difficiles. Merci à Ana (Anita) pour sa bonne humeur et son énergie débordante, merci surtout pour les fou-rires au labo quand il commençait à se faire tard. Merci à Lucie qui m’a aidé en BM. Merci aussi à Maureen et Jérémy, récemment arrivés dans l’équipe, mais qui m’ont aidé au cours de la rédaction. Merci aussi aux stagiaires adorables que j’ai eu le plaisir d’encadrer et qui ont fait un travail formidable : Natalia, Naomie, Flore, Stefan, Elodie et Pauline.

Bien évidemment, je remercie également tous les autres membres du laboratoire « Plasticité du Cerveau », autant les permanents que les étudiants. En particulier, je remercie les directeur de l’unité, Serge Birman au début de ma thèse et Thomas Préat actuellement, pour m’avoir accueilli au sein du laboratoire. Les étudiants (anciens et actuels) ont également été d’une grande aide. Je pense notamment à Alexandra, Raouf, Romain, Emna (pour les pauses clopes « éclair »), Isa, PYM, Ghislain, Gaetan, Marie, Quentin, Séverine, Marlène (et Jean !)... (et j’en oublie sûrement !) On aura quand même passé des bons moments ensemble ! Parmi les permanents chercheurs, je remercie Thierry, Armelle (Champagne !), Tania, PYP et Alice pour les discussions animées à la cafét. Je remercie également Marcel, Honorine, Alexandre, Léna, Sylvie, Sopharith, Malika et Abdoul. Et surtout, je remercie le « club couture » du labo, j’ai nommé Hélène et Aurélie ! Merci pour tous les déjeuners et cafés. Bien que vous m’ayez fourvoyée dans des lieux de perditions, vous m’avez toujours soutenu dans les moments difficiles et vous avez su me mettre des coups de pieds aux fesses pour me remettre sur le droit chemin (celui du travail…).

Le travail effectué en collaboration avec l’équipe de Bruno Goud, à l’institut Curie, a été très enrichissant. Merci donc à Jean-Baptiste, Sabine et Stéphanie. Je souhaite également remercier Pascal Dournaud pour ce début de collaboration au sujet du récepteur SST2. De plus, au cours de ma thèse, j’ai pu organiser trois réunions « TAC meeting » qui m’ont grandement aidé dans mes projets. Je remercie donc mes tuteurs, Pierre Vincent et Christophe Leterrier pour leur disponibilité et leurs idées.

Le laboratoire « plasticité du cerveau » n’étant pas le seul labo de l’ESPCI, au cours de ma thèse, j’ai eu le plaisir de côtoyer et de travailler avec des étudiants et permanents d’autres labos. Merci aux membres du laboratoire de spectro (Joëlle, Manu, Iman, Séga, Yann, Giovanni…) pour les discussions autour des micro-ondes. La salle de culture a beaucoup changé au cours de ma thèse et je souhaite remercier les utilisateurs de « l’ancienne petite » salle de culture pour leur bonne humeur, le partage de la salle et la motivation dont vous avez fait preuve lors du déménagement de la salle ; je pense notamment à petite Sophie, Alexandra et Marianna du LPEM, à Anne et Antoine du MMN, mais aussi à Raphaël et Matteo du LBC qui ont mis en place la nouvelle salle de culture et ont gentiment repris la gestion de cette salle. Mais que serait l’ESPCI sans ses pannes d’ascenseurs et de climatisation, ses fuites d’eau et ses coupures d’électricité interminables ? Ou comment monter une table optique au 4ème étage sans ascenseur ? Merci donc au personnel du STML actuel pour leur disponibilité et leur efficacité.

Frontières du Vivant est une école doctorale hors-norme. Elle m’a permis de découvrir de nombreux sujets de recherche dans des domaines scientifiques variés et de rencontrer des personnes formidables. Je remercie particulièrement l’équipe du « Imaging Thematic Club » (Marie, Sophie, Eugenia et Daria) pour les japs à volonté rue Bertholet.

L’ESPCI n’est pas seulement l’endroit où j’ai fait ma thèse, mais c’est aussi l’école d’ingénieur que j’ai intégrée après la prépa et où j’ai fait la connaissance de personnes extraordinaires. Merci aux PC1 (des jeunes aux vieux cons d’anciens) pour ces bons moments passés au foyer qui m’ont permis de décompresser pendant ma thèse. Merci surtout de m’avoir demandé « Ca avance ta thèse ? » ou « Ca y est ? Tu as rendu ton manuscrit ? » dans les meilleurs moments ! Remerciement spécial aux huit générations de barmen qui m’ont toujours gardé un ice-tea au frais.

Je remercie aussi mes amis de longue date et je m’excuse de ne pas pu les voir aussi souvent que je le souhaitais ces trois dernières années (promis, je me rattraperai bien vite !) : Gégé, Vivi, Michel, petite Alicia (et ses beaux sourires), ainsi que les Boudinées® (Florine, Astrid, Aurélia et Andréa).

Au cours de ma thèse, j’ai également continué la danse et le théâtre et j’ai eu l’occasion de m’investir en politique. Ces activités m’ont aidée à conserver un équilibre qui m’a permis de tenir ces trois années de doctorat. Je remercie donc l’ensemble des personnes qui m’ont entourée dans ces activités.

Pour finir, je remercie ma famille : les chTarbais, tante Rachel, les tantines de Tournay, les Américains, les Suisses, les Bullè et leurs trois bouts de chou, mais aussi et surtout ma maman qui m’a toujours soutenue et, sans qui, je ne serai pas là !

Table of contents

FIGURES, TABLES AND EQUATIONS ...... 5

ABBREVIATIONS...... 7

GENERAL INTRODUCTION...... 8

1. Neurons: polarization and traffic ...... 9 1.1. Neurons are highly specialized and polarized cells...... 9 1.1.1. General points...... 9 1.1.2. Synaptic transmission...... 9 1.1.3. Neuronal polarization ...... 10 1.2. Neuronal culture as study model ...... 11 1.3. Establishment of neuronal polarity...... 12 1.3.1. Breaking the symmetry...... 12 1.3.2. Molecular signaling ...... 14 1.4. Maintenance of neuronal polarity...... 14 1.4.1. Differential composition in proteins and lipids between dendrites and axon ...... 14 1.4.2. The axon initial segment as a diffusion barrier...... 16 1.4.3. Polarization of membrane lipids: a controversial subject ...... 16 1.5. Targeting of neuronal membrane proteins...... 16 1.5.1. Dendritic targeting...... 16 1.5.2. Axonal targeting ...... 17

2. G-protein coupled receptors...... 20 2.1. Definition and structure...... 20 2.2. GPCR classification ...... 21 2.2.1. Rhodopsin receptors ...... 21 2.2.2. Secretin receptors ...... 22 2.2.3. Adhesion receptors...... 22 2.2.4. Glutamate receptors...... 22 2.2.5. Frizzled/Taste2 receptors...... 22 2.3. GPCR activation...... 24 2.3.1. Energy landscape diagram...... 25 2.3.2. Activation-induced conformational changes ...... 26 2.4. Constitutive GPCR activity ...... 27 2.4.1. Constitutive activity theories and models...... 28 2.4.1.1. Energy landscape diagram ...... 28 2.4.1.2. The probabilistic model ...... 28 2.4.1.3. Ternary complex model ...... 29 2.4.1.4. Extended ternary complex model (ETC) and cubic ternary complex model (CTC)...... 30 2.4.2. Physiological constitutive activity...... 31 2.5. G proteins structure, activation and subtypes...... 32 2.5.1. Gi/o and Gs proteins...... 33 2.5.2. Gq/11 proteins...... 33 2.5.3. G12/13 proteins ...... 34 2.6. Adenylyl cylclases: structure, activation and subtypes ...... 35 2.7. GPCR regulation ...... 37 2.7.1. Desensitization ...... 38 2.7.2. Endocytosis...... 39 2.7.3. GPCR recycling and degradation ...... 39

3. Questions...... 41

4. Type-1 cannabinoid receptor (CB1R) ...... 42

1 4.1. Identification, structure and distribution ...... 42 4.2. Other cannabinoid receptors...... 43 4.3. CB1R pharmacology ...... 43 4.3.1. Ligands ...... 43 4.3.1.1. Phytocannabinoids...... 44 4.3.1.2. Synthetic agonists ...... 44 4.3.1.3. Endogenous agonists...... 45 4.3.1.4. Synthetic inverse agonists...... 45 4.3.1.5. Endogenous inverse agonists ...... 45 4.3.2. Binding sites ...... 45 4.3.3. Signaling...... 46 4.3.3.1. Calcium and potassium channels ...... 46 4.3.3.2. Mitogen-activated protein kinase (MAPK)...... 46 4.3.3.3. Nitric Oxide Synthase (NOS) ...... 47 4.3.4. Physiological roles...... 47 4.3.4.1. Synaptic transmission ...... 47 4.3.4.2. Pain ...... 48 4.3.4.3. Neuroprotection ...... 48 4.3.4.4. Neuronal development...... 49 4.3.4.5. Appetite regulation ...... 49 4.3.5. Constitutive CB1R activity...... 49 4.3.5.1. Structural determinants enabling CB1R constitutive activity ...... 50 4.3.5.2. Constitutive endocannabinoid tone...... 50

PART 1: INFLUENCE OF NEURONAL POLARITY ON THE TRAFFIC OF TYPE-1 CANNABINOID RECEPTOR...... 52

1. Introduction...... 53

2. Results: Article 1 (published): Activation-dependent plasticity of polarized GPCR distribution on the neuronal surface. Journal of Molecular Cell Biology, 2013...... 55

3. Main results and remaining questions ...... 84

PART 2: INFLUENCE OF NEURONAL POLARITY ON THE PHARMACOLOGY OF TYPE-1 CANNABINOID RECEPTOR...... 86

1. Emerging concepts of GPCR pharmacology: allosteric modulation, signaling bias and modulation by membrane lipids...... 87 1.1. Allosteric modulation...... 87 1.1.1. General points...... 87 1.1.2. Signaling bias ...... 88 1.2. Lipids and GPCR activity...... 89 1.2.1. Plasma membrane composition ...... 89 1.2.2. Modulation of GPCRs activity by membrane lipids...... 90 1.2.3. Lysophosphatidic acid receptor LPAR and sphingolipid receptor S1PR ...... 91 1.2.4. Endocannabinoid biochemistry ...... 92 1.2.4.1. AEA ...... 93 1.2.4.2. 2-AG ...... 93 1.2.4.3. 2-AG basal tone: a non-classical neuromodulator ...... 94 1.2.5. Are endocannabinoids genuine CB1R ligands or lipidic modulators? Relation to constitutive CB1R activity...... 95 1.2.6. CB1R allosteric modulation and signaling bias...... 96

2. Measure of cAMP/PKA signaling pathway in live neurons at the sub-cellular scale...... 98 2.1. cAMP/PKA pathway downstream of G proteins...... 98 2.1.1. cAMP structure and signaling ...... 98 2.1.2. PKA structure and signaling...... 99

2 2.1.2.1. Structure and activation ...... 99 2.1.2.2. PKA isoforms ...... 99 2.1.2.3. Sub-cellular localization ...... 100 2.1.2.4. Role of neuronal PKA...... 100 2.2. Förster Resonance Energy Transfer (FRET) imaging ...... 101 2.2.1. Principle and history...... 101 2.2.1.1. Fluorescence ...... 101 2.2.1.2. FRET...... 101 2.2.2. Detection of cAMP concentration by FRET...... 103 2.2.2.1. PKA-based cAMP FRET biosensors ...... 103 2.2.2.2. Epac-based cAMP FRET biosensors ...... 104 2.2.3. Detection of PKA activity by FRET...... 105 2.3. Setting up and improvement of FRET imaging to measure modulation of basal cAMP concentration and PKA activity downstream of endogenous CB1Rs...... 107 2.3.1. Equipment...... 107 2.3.2. Data analysis...... 108 2.3.3. Statistical analysis...... 111 2.3.4. Protocols and first results...... 111 2.3.4.1. Forskoline EC50 shift in somata induced by activation or blockade of overexpressed CB1Rs.111 2.3.4.2. Modulation of basal PKA activity induced by activation or blockade of overexpressed CB1Rs in somata and dendrites...... 115 2.3.4.3. Modulation of basal PKA activity in axons downstream of endogenous CB1Rs...... 118

3. Results: Article 2 (submitted): Polarized cellular patterns of endocannabinoid production and detection shape cannabinoid signaling in neurons...... 119

4. Main results and remaining questions ...... 151

PART 3: INFLUENCE OF NEURONAL POLARITY ON THE PHARMACOLOGY OF SEROTONINERGIC 5-HT1B RECEPTORS...... 152

1. Introduction...... 153

2. Materials and Methods...... 154

3. Results ...... 156 3.1. Overexpressed 5-HT1B receptors constitutively inhibit PKA activity in all neuronal compartments...... 156 3.2. Activation of 5-HT1B receptors leads to a stronger decrease of PKA activity in axons as compared to dendrites...... 157

4. Figures and legends...... 158

5. Discussion...... 159

PART 4: STRUCTURAL DETERMINANTS FOR THE CONSTITUTIVE ACTIVITY OF TYPE-1 CANNABINOID RECEPTORS - ROLE OF T210...... 161

1. Introduction...... 162

2. Materials and Methods...... 162

3. Results ...... 163 3.1. Overexpressed wild-type CB1Rs constitutively inhibit PKA activity in all neuronal compartments independently of 2-AG presence...... 163 3.2. Overexpressed hypoactive CB1Rs do not display constitutive activity, even in the somatodendritic compartment where 2-AG is present...... 163 3.4. Activity pattern of endogenous CB1Rs is similar to overexpressed WT-CB1Rs in dendrites and to overexpressed T210A-CB1Rs in axons...... 164

3 4. Figures and legends...... 165

5. Discussion...... 166 5.1. Receptor overexpression changes activation profile...... 166 5.2. T210 is required for CB1R activation by 2-AG...... 167 5.3. Responses to WIN in neurons expressing T210A-CB1Rs are due to endogenous CB1R activation...... 168

PART 5: ROLE OF TYPE-1 CANNABINOID RECEPTORS IN NEURONAL DEVELOPMENT AND THE ESTABLISHMENT OF NEURONAL POLARITY...... 169

1. Introduction...... 170

2. Review (published): Type-1 cannabinoid receptor signaling in neuronal development. Pharmacology, 2012...... 171

GENERAL DISCUSSION AND PERSPECTIVES...... 195

ANNEXE 1: ARTICLE 3 (UNDER REVIEW): FAST RETROGRADE SIGNALING IN DROSOPHILA AXONS...... 200

ANNEXE 2: SOMATOSTATIN SST2 RECEPTORS MODULATE BOTH CGMP PATHWAY AND PKA ACTIVITY IN HIPPOCAMPAL NEURONS...... 232

1. Introduction...... 233

2. Materials and Methods...... 234

3. Results ...... 236 3.1. Activation of endogenous SST2Rs decreases PKA activity in somata but not in the axons...... 236 3.2. Construction and validation of THPDE5VV, a new cGMP FRET probe...... 238 3.3. Activation of endogenous SST2Rs potentiates guanylate cyclases activation...... 241

4. Discussion...... 242

REFERENCES ...... 244

4 Figures, Tables and Equations

Figure 1: Neuronal structure p.9 Figure 2: Establishment of neuronal polarity P12 Figure 3: Model for axon specification in developing neurons p.13 Figure 4: Three main processes for dendritic and axonal targeting p.18 Figure 5: Structure of rhodopsin-like GPCR CXCR1 p.20 Figure 6: Diversity of GPCRs p.21 Figure 7: Phylogenetic analysis of the Rhodopsin family in rat genome p.23 Figure 8: Phylogenetic analysis of other GPCR families p.24 Figure 9: Theoritical energy landscape of a GPCR p.25 Figure 10: GPCR molecular switches p.26 Figure 11: Constitutive GPCR activity p.27

Figure 12: Theoritical energy landscape of a GPCR with low or high basal p.28 activity Figure 13: Probabilistic model p.29 Figure 14: GPCR activation models (1) p.30 Figure 15: GPCR activation models (2) p.31 Figure 16: Functional cycle of G protein activity p.33 Figure 17: Effectors dowstream of G proteins p.35 Figure 18: Adenylyl cyclase structure p.36

Figure 19: GPCR reglation : desensitization, recycling and downregulation p.38 (degradation) Figure 20: CB1R distribution in the CNS p.43 Figure 21: CB1R signaling p.46

Figure 22: How to distinguish constitutive activity from endogenous ligand p.51 binding? Figure 23: CB1R targeting p.54 Figure 24: Allosteric modulators p.87

Figure 25: Signaling bias and GPCRs multiple active states. p.88 Figure 26: Role of lipid raft domains in GPCR signaling p.90 Figure 27: LPARs and S1PRs subtypes and signaling p.92 Figure 28: Main pathways of eCBs synthesis and degradation p.93 Figure 29: 2-AG basal tone p.95 Figure 30: cAMP synthesis p.98 Figure 31: PKA structure and activation p.99

5 Figure 32: Conditions to obtain efficient FRET p.102 Figure 33: FlCRhR probe p.103 Figure 34: Epac1 structure and activation by cAMP binding p.104 Figure 35: AKAR probes p.106 Figure 36: Subcellular targeting of AKAR4 p.107 Figure 37: Channels splitting p.108 Figure 38: Matlab quantification p.110 Figure 39: Statistical analysis p.111 Figure 40: Protocols p.112

Figure 41: CB1R constitutively inhibits Forskoline-stimulated adenylyl p.114 cyclase in somata (1) Figure 42: CB1R constitutively inhibits Forskolin-stimulated adenylyl p.115 cyclase in somata (2) Figure 43: CB1Rs modulate basal PKA activity in somata p.116 Figure 44: CB1Rs modulate basal PKA activity in dendrites p.117

Figure 45: Serotoninergic receptors 5-HT1B modulate PKA activity p.158 differently in axons and dendrites Figure 46: T210 is a structural determinant for CB1R constitutive activity p.165

Figure 47: Activation of endogenous SST2Rs decreases PKA activity in p.237 somata but not in the axons Figure 48: cGMP biosensor construction p.239 Figure 49: Validation of THPDE5VV probe p.240

Figure 50: Activation of endogenous SST2Rs potentiates guanylate cyclases p.241 activation Figure 51: How do SST2Rs potentiate guanylate cyclase activity? p.243

Table 1: Differential composition in organelles and proteins between axon p.15 and dendrites Table 2: Responses of ACs to various modulators p.37 Table 3: Cannabinoid ligands p.44

Equation 1: FRET efficiency p.101 Equation 2: Definition of FRET Ratio p.102

Abbreviations - 2-AG: 2-arachidonoylglycerol - 5-HT: 5-hydroxytryptamine (= serotonin) - AC: Adenylyl Cyclase - AEA: N-arachidonoylethanolamide (= anandamide) - AIS: Axon Initial Segment - AKAP: A-kinase anchor protein - cAMP: cyclic Adenosine Monophosphate - CB1R: Type-1 cannabinoid receptor - cGMP: cyclic Guanosine Monophosphate - CME: Clathrin-mediated endocytosis - CNS: Central Nervous System - CFP: Cyan Fluorescent Protein - Δ9-THC: Δ9-Tetrahydrocannabinol - DAG: Diacylglycerol - DAGL: Diacylglycerol Lipase - DIV: Day In Vitro - DSE: Depolarization-induced Suppression of Excitation - DSI: Depolarization-induced Suppression of Inhibition - eCB: endocannabinoid - EC loop: Extracellular loop - Epac: guanine-nucleotide Exchange Proteins Activated by cAMP - FRET: Förster (or Fluorescence) Resonance Energy Transfer - Fsk: Forskoline - GC: Guanylate Cyclase - GFP: Green Fluorescent Protein - GPCR: G-protein coupled receptor - IC loop: Intracellular loop - LPA: Lysophosphatidic acid - LTD: Long-Term Depression - LTP: Long-Term Potentiation - PDE: Phosphodiesterase - PKA: cAMP-dependent protein kinase A - S1P: Sphingosine-1-phosphate - SST2R: somatostatin type-2 receptor - THL: Tetrahydrolipstatin - TM: Transmembrane - YFP: Yellow Fluoresent Protein

General introduction

8 1. Neurons: polarization and traffic

1.1. Neurons are highly specialized and polarized cells 1.1.1. General points The nervous system is divided in two parts: - Peripheral Nervous System (PNS) insures the connection between the brain and peripheral organs. It is composed of Sensory Nervous System that processes sensory information and Motor Nervous System that controls muscles and movements. - Central Nervous System (CNS) integrates information and elaborates responses to stimuli. It is composed of the encephalon and the spinal cord. Neurons constitute functional cells of nervous system. These highly polarized cells are composed of a cell body, called soma, and two types of extensions: one axon and one or several dendrites. The axon leads the nervous impulse, called action potential, and contains presynaptic terminals that release neurotransmitters. Dendrites oppose postsynaptic specializations to afferent axons (Figure 1).

Figure 1: Neuronal structure Hippocampal cultured neurons of rat embryos were transfected with cytoplasmic DsRed (red), fixed and immunostained with MAP2 (green), a specific marker of somatodendritic compartment, and Hoechst (blue), specific marker of nucleus.

1.1.2. Synaptic transmission Information is driven by action potential, and coded in frequency modulation. It is then transferred to the next neuron through Synaptic Transmission. A synapse is composed of a presynaptic part on axon of afferent neuron and a postsynaptic part on dendrites of efferent neuron. Afferent neurons stock neurotransmitter vesicles that are released in synaptic cleft and bind postsynaptic receptors. In the CNS, the predominant neurotransmitters are glutamate, associated with excitatory synapses (Excitatory Postsynaptic Potential, EPSP), and GABA, associated with inhibitory synapses (Inhibitory Postsynaptic Potential, IPSP).

9 CNS is a very complex system that is able to adapt. Thus, one stimulus leads to different responses depending on environmental factors and/or adaptative processes. To do so, neuronal transmission is modulated by various mechanisms, allowing synaptic plasticity. - Short-term synaptic plasticity: This reversible process appears during synaptic transmission. Activation of postsynaptic calcium channels leads to a release of postsynatptic neuromodulators, such as endocannabinoids, that retrogradely cross synaptic cleft, activate presynaptic receptors and induce synaptic silencing. This process is called Depolarization-induced Suppression of Excitation (DSE) for glutamatergic transmission and Depolarization-induced Suppression of Inhibition (DSI) for GABAergic transmission (Diana & Marty, 2004). - Long-term synaptic plasticity: Long lasting synaptic modulation is able to either strengthen (Long-Term Potentiation, LTP) or, on the contrary, diminish (Long-Term Depression, LTD) synaptic activity. These processes lead to protein synthesis and are involved in memory formation (Malenka & Bear, 2004).

1.1.3. Neuronal polarization Most of differentiated cells organize distinct domains, containing various proteins and lipids types, in their plasma membrane. This compartmentalization is critical for proper cell functioning. Neurons are highly specialized cells. Axons of mature neurons contain thousands of synaptic terminals and although they are very thin (less than 1µm), they can extend over longer distances than millions times soma diameter. The resulting high surface to volume ratio of this structure requires specific cellular organisation constraints that are not encountered in other cell types. How does neuronal polarization establish? How do neurons maintain this polarization? These questions have been asked in several studies and are still investigated. Development of neuronal cultures has helped a lot in these investigations, enabling the observation of individual developing and mature neurons. Indeed, Gary Banker and colleagues have developed a method to obtain in vitro primary culture of hippocampal neurons of rat embryos (Banker & Cowan, 1977). This method has been and is still largely used to study individual neurons.

10 1.2. Neuronal culture as study model Biological studies largely use cell culture models, notably immortal cell lines that are able to divide infinitely. Among these cell lines, neuroblastoma Neuro2A cells display neuron-like phenotype. These undifferentiated cells, after stimulation by retinoic acid, develop neuritic extensions (Sidell, 1982). However, Neuro2A do not display somatodendritic/axonal polarity and can not be used to precisely study this neuronal property. PC12 cells, isolated in 1976 from a rat pheochromocytoma, also differentiate in presence of Nerve Growth Factor (NGF) but do not establish neuronal polarity (Greene & Tischler, 1976). Thus, no cell line displays all characteristics of neurons from the CNS: protein expression (receptors, ionic channels and neurotransmitters), differentiated somatodendritic and axonal compartments and establishment of functional synapses. Thus, primary hippocampal cultures of rat embryos have been widely used to study neurons as individual cells. This method consist in removing hippocampi from brains of rat embryos between 17 to 19 embryonic days and to dissociate them before plating them at low density (100 to 500 cells/mm2) on coverslips previously coated with proteins enabling cell adhesion (Banker & Cowan, 1977). Cultured neurons develop themselves following well defined development stages (Dotti et al, 1988) (Figure 2). Plated cells rapidly adhere to the coverslip and grow lamellipodia around cell body (stage 1). After a few hours, lamellipodia condense and form neurites (stage 2). These neurites are alike but differ from the soma through differences in microtubules organization and absence of Golgi apparatus or endoplasmic reticulum. Within 24 hours, one of the neurites increases strongly its growth rate and starts its differentiation into an axon (stage 3). Thus, stage 3 is the first demonstration of polarization. The other neurites continue to grow slower and differentiate into dendrites after 4 to 6 Days In Vitro (DIV) (stage 4). Finally, after DIV10, axons and dendrites start to connect forming functional synapses (stage 5). At this stage, specific neuronal markers, such as MAP2 for somatodendritic compartment and Tau for axons, are well segregated. Maturation continues with formation of dendritic spines and development of axonal network. Moreover, GABAergic and glutamatergic synaptic connections develop (Bartlett & Banker, 1984). Neuronal cultures are predominantly composed of glutamatergic pyramidal neurons and a low percentage (about 10%) of GABAergic interneurons. Lastly, development of specific transfection protocols enabling to express various proteins or markers in live cultured neurons, has transformed primary neuronal cultures into a powerful tool to study neuronal physiology (Ohki et al, 2001).

Figure 2: Establishment of neuronal polarity (Dotti et al, 1988) Dotti and colleagues described five stages in neuronal development for hippocampal cultured neurons of rat embryos. Stage 1: Lamellipodia surround cell body. Stage 2: Equivalent neurites (minor processes) start growing. Stage 3: One neurite grows faster and differentiate into axon. Stage 4: Other neurites grows and differentiate into dendrites. Stage 5: During maturation, dendrites can develop dendritic spines and formation of synapses appears.

1.3. Establishment of neuronal polarity As described above, neuronal polarity is established early by distinction of one neurite that differentiate into axon and other neurites into dendrites. How is the axon chosen among the neurites? Which are the processes involved in order to avoid the development of several axons? In this part, I will report some results that partly answer these questions.

1.3.1. Breaking the symmetry Axotomy of cultured neurons at stage 3 leads to axonal regeneration. If the cut axon remains longer than the other neurites, it will grow again as an axon, while other neurites will grow as dendrites. Interestingly, if cut axon length is below a certain threshold, an other neurite will grow as the axon and the cut axon will develop as a dendrite (Goslin & Banker, 1989). Thus, all neurites at stage 2 are equivalent and their fate is determined by intrinsic determinants. At stage 2, in vitro cultured neurons display a symmetric morphology with equivalent neurites. One of these neurites will grow in axon while the others will differentiate into dendrites. To do so, neurons have to perform what physicists call spontaneous symmetry breaking. Asymetry emerges due to instability of the initially symmetric system (Andersen & Bi, 2000). Instability comes from combination of positive and negative feedbacks (Figure 3) (Arimura & Kaibuchi, 2007). Positive feedbacks are composed of membrane recruitment, microtubule assembly and increase of protein transport and of F-actin dynamics. All these

12 processes increase growth cone dynamics. Indeed, Bradke and Dotti showed that at stage 2, the neurite displaying the highest growth cone dynamics will start to grow faster and differentiate into the axon. Moreover, application of actin depolymerisation agents Cytochalasin D or Latrunculin B resulted in formation of multiple axons (Bradke & Dotti, 1997; Bradke & Dotti, 1999). This result demonstrates that neuronal polarization depends on the uniqueness of the neurite displaying positive feedback processes. To regulate other neurites and avoid development of several axons, neurons display negative feedbacks sent by the growing axon. A question remains: How does symmetry breaking occur? In 1952, Alan Turing proposed a model for morphogenesis. In this model, a symmetric (homogenous) system can become asymmetric (inhomogenous) due to local instability of reaction-diffusion processes: both local chemical reactions and widespread molecule diffusion contribute to the distribution pattern of a given molecule. In neurons, both positive and negative feedbacks are present. Turings theory predicts that these processes create local inhomogeneities. Thus, at a point, one neurite can overtake positive feedbacks threshold, initiates its differentiation into axon and sends negative feedbacks to other neurites reinforcing negative processes and avoiding their differentiation into axon (Cheng & Poo, 2012). Figure 3: Model for axon specification in developing neurons (Arimura & Kaibuchi, 2007) (a) At stage 2, neurites keep the same length through competition between positive and negative growth processes (PI3K: Phosphatidylinositol-3- kinase) (b) At stage 3, axon specification starts with increased positive processes in this neurite and strengthened negative processes in other neurites to avoid development of multiple axons

13 1.3.2. Molecular signaling Although extracellular cues have been shown to favour the axonal differentiation of one neurite, the fact that polarization can be also established in vitro shows robustness of the intracellular processes. Many studies reported positive contribution in axonal growth of various cytoskeleton proteins (Cheng & Poo, 2012). Given that actin polymerization plays a major role in positive regulation, Rho GTPases, such as cdc42 and Rac, are important effectors for axonal differentiation. Indeed, their inhibition with bacterial toxin B caused formation of multiple axons (Bradke & Dotti, 1999). However, the predicted balance between positive and negative regulators has been experimentally verified only for cAMP (cyclic adenosine monophosphate) and cGMP (cyclic guanosine monophosphate) effectors. Indeed, Shelly and colleagues reported that elevated cAMP concentration favoured axon formation, while elevated cGMP concentration favoured dendrite formation. Moreover, they reported that elevation of cAMP in one neurite of developing cultured neurons decreases cGMP concentration in this same neurite, while it increases cGMP concentration and decreases cAMP concentration in the other neurites (Shelly et al, 2010).

1.4. Maintenance of neuronal polarity

1.4.1. Differential composition in proteins and lipids between dendrites and axon Organelles and proteins are not uniformly distributed in neurons, due to neuronal polarity (Table 1). Moreover, some uniformly distributed proteins do not display the same organization in dendrites and axon. For example, while axonal microtubules all point out their plus-ends toward the distal part of the axon, in dendrites, there are microtubules that point out their plus-ends toward distal part of the dendrites and microtubules with plus-ends toward the soma. This polarized distribution and organization enable the maintenance of neuronal polarity.

14 Table 1: Differential composition in organelles and proteins between axon and dendrites. (Craig & Banker, 1994)

15 1.4.2. The axon initial segment as a diffusion barrier The axon initial segment (AIS) is localized close to the soma. This site is enriched in voltage- gated channels and ankyrin-G and constitutes the starting point of action potentials. Its specific protein assembly starts from DIV3-DIV4 (Leterrier & Dargent, 2013). Interestingly, AIS acts as a diffusion barrier, avoiding somatodendritic proteins to diffuse freely to the axon. Indeed, while 10-kD proteins can diffuse freely, 70-kD proteins remain somatodendritic due to their blockade in AIS (Song et al, 2009). Thus, proteins can traffic to the axon only by active transport. Traffic of neuronal membrane proteins plays a major role in maintenance of neuronal polarity and I will detail these processes in the next paragraphs.

1.4.3. Polarization of membrane lipids: a controversial subject Kobayashi and colleagues inserted fluorescent phospholipids in polarized cultured neurons. They found that only axons were labelled and concluded that diffusion barrier is responsible for the lack of diffusion of lipids between axonal and somatodendritic plasma membrane (Kobayashi et al, 1992). However, using modeling, Anthony Futerman calculated that axonal lipids would reach soma in a time scale higher than the observation of Kobayashi and colleagues (Futerman et al, 1993). Thus, it is unclear if polarized distribution of lipids exists in neurons.

1.5. Targeting of neuronal membrane proteins Importantly, proteins are constantly renewed and polarity is maintained due to specific targeting processes. Synthesized in the Golgi apparatus localized in the soma, protein sorting is performed after. For example, specific carriers, such as kinesin KIF5-driven carriers, target proteins in the axon, while others, such as KIF17-driven carriers, are not allowed to enter the axon. This specificity explains the dendritic segregation of NMDA receptor subunit NR2B (Song et al, 2009).

1.5.1. Dendritic targeting After their synthesis, most dendritic proteins are targeted to the plasma membrane through direct selective delivery. For example, cargoes transporting the dendritic transferrin receptor (TfR) from Golgi apparatus to plasma membrane are totally excluded from axon (Burack et al, 2000). Thus, dendritic proteins are driven by specific cargoes excluded from the axon leading to an exclusively dendritic delivery. The assembly to dendritic cargoes is driven by

16 specific structural protein patterns. Rivera and colleagues showed that somatodendritic voltage-gated K+ channels Kv4.2 display a dileucine-containing motif in C-terminal tail responsible for their dendritic targeting. Indeed, chimeras of axonal Kv1.3 and Kv1.4 containing this pattern were targeted to somatodendritic membrane (Rivera et al, 2003). Moreover, metabotropic glutamate receptors mGluR2 are excluded from axons but chimera construction with addition of C-terminal tail of axonal mGluR7 leads to both dendritic and axonal presence of mGluR2 (Stowell & Craig, 1999). Interestingly, serotonin receptors 5-

HT1A and 5-HT1B share highly similar structure. However, while 5-HT1A receptors are segregated in the somatodendritic plasma membrane, 5-HT1B receptors are localized predominantly on the axonal plasma membrane. Our group demonstrated that the third intramembrane loop of these receptors is responsible for their specific targeting (Carrel et al, 2011) (see also Part 3). Thus, dendritic proteins are sorted after their synthesis in function of specific patterns, targeted to somatodendritic membrane and remain excluded from the axon.

1.5.2. Axonal targeting Three main axonal targeting processes have been described (Figure 4). Firstly, after their synthesis, proteins can be sorted at the level of Golgi apparatus into carriers and delivered to the axon through a direct selective delivery. While this process is predominant in dendritic targeting, it seems that only a few axonal proteins reach the axon through this mechanism. Indeed, selective delivery relies on different microtubule orientation: while dendritic microtubules can project their plus-end both toward soma and distal dendritic end, axonal microtubules all orient their plus-end toward distal axonal part from the soma. Thus, axonal cargoes driving toward microtubule plus-ends can also traffic in dendrites. As carriers of axonal proteins can traffic both to dendrites and to the axon, selective fusion or retention are required to achieve specific axonal targeting. Selective fusion consists in protein ability to fuse with axonal membrane but not with dendritic membrane. Selective retention assumes that carriers can fuse to the dendritic membrane but proteins are rapidly removed by endocytosis and delivered to the axon in endosomes through transcytosis (Lasiecka & Winckler, 2011). Experimentally, the difference between both mechanisms is hard to detect but recent studies tend to demonstrate that transcytosis is the predominant axonal targeting mechanism. In 1997, Dottis group showed that somatodendritic transferrin receptors (TfR) are driven to the axon through transcytosis after binding to transferrin. Axonal receptor number was also increased after addition of excitatory neurotransmitters, while number of somatodendritic receptors increased in presence of inhibitory neurotransmitters (Hemar et al,

17 1997). However, this dendroaxonal transcytosis is due to TfR activation. In 2001, the group of Bénédicte Dargent demonstrated that specific C-terminus sequence of neuronal sodium channel Nav1.2 was recognized by the clathrin-dependent endocytic pathway in the somatodendritic compartment and is responsible for its transcytosis and targeting to the axon (Garrido et al, 2001). It was the first evidence that selective retention is a physiological pathway for axonal targeting of proteins. Next studies showed that this process is involved in targeting of various axonal proteins: Vesicle-associated membrane protein 2 (VAMP2), involved in synaptic vesicle fusion with the presynaptic membrane (Sampo et al, 2003), tropomyosin-related kinase A (TrkA) receptors, involved in growth and survival of developing neurons (Ascano et al, 2009), contactin-associated protein 2, a cell adhesion molecule (Bel et al, 2009) and β-site APP cleaving enzyme 1 (BACE1), cleaving enzyme of amyloid precursor protein (Buggia-Prevot et al, 2014). Interestingly, studies on adhesion molecule NgCAM traffic showed that both selective delivery and selective retention are involved in its axonal targeting (Sampo et al, 2003; Wisco et al, 2003). Thus, these processes are not exclusive.

Figure 4: Three main processes for dendritic and axonal targeting. (Horton & Ehlers, 2003)

(A) Selective Delivery consists in protein sorting directly after synthesis in Golgi apparatus. Different carriers are used to deliver axonal and dendritic proteins. (B) Same carriers are used to transport both dendritic and axonal proteins but selective fusion is applied meaning that while dendritic proteins can bind dendritic plasma membrane, axonal proteins can only bind axonal plasma membrane. (C) Selective retention assumes that axonal proteins can bind somatodendritic plasma membrane but are rapidly endocytosed and targeted to the axon through transcytosis.

Why are axonal proteins rapidly removed from the somatodendritic membrane? This question still remains unanswered. Some structural motives could favour their internalization (Carrel et

18 al, 2011; Garrido et al, 2001) but since neurons display high differences between somatodendritic and axonal membrane composition, it could be possible that membrane composition is responsible for this somatodendritic-specific endocytosis. During my PhD, I tried to answer this question, focusing my study on a particular protein family, for which traffic information is lacking: G-protein coupled receptors.

19 2. G-protein coupled receptors

2.1. Definition and structure G-protein coupled receptors (GPCRs) are seven transmembrane proteins that are coupled to heterotrimeric GTP binding proteins, called G proteins. They are made up of seven transmembrane (TM) alpha-helix domains linked by three intracellular (IC) loops, three extracellular (EC) loops, an extracellular N-terminal extremity and an intracellular C-terminal extremity (Figure 5). GPCRs bind extracellular molecules, called ligands, and activate downstream intracellular signaling pathways, leading to cellular response. GPCRs ligands are numerous and diverse: ions, peptides, proteins, lipids and also light can relay their message through these proteins (Figure 6) (Marinissen & Gutkind, 2001). Moreover, GPCRs are involved in various physiological functions, such as cell proliferation, hormonal regulation, synaptic transmission It has been estimated that 80% of neurotransmitters and hormones are active through GPCR activation of a GPCR (Birnbaumer et al, 1990). Thus, GPCRs are choice therapeutic targets and represent 45% of all treatment targets (Drews, 2000).

Figure 5: Structure of rhodopsin-like GPCR CXCR1 (Park et al, 2012) (A) CXCR1 topology (ECL: Extracellular loop; ICL: Intracellular loop) (B) Ensemble of 10 lowest energy structures of CXCR1 aligned in the membrane (n: bilayer normal)

Figure 6: Diversity of GPCRs (Marinissen & Gutkind, 2001) GPCRs can bind diverse ligands (5-HT: serotonin ; LPA: lysophosphatidic acid; PAF: platelet-activating factor; FSH: follicle-stimulating hormone; LH: leuteinizing hormone; TSH: thyroid-stimulating hormone) and modulate various signaling pathways (cAMP: cyclic adenosine monophosphate; DAG: diacylglycerol; PLC-β: phospholipase C-β; PKC: protein kinase C; PI3Kγ: phosphoionositide-3-kinase-γ; RhoGEF: Rho guanine nucleotide exchanged factor).

2.2. GPCR classification More than 1,000 GPCRs are encoded in the mammalian genome (Wettschureck & Offermanns, 2005). Thus, they constitute the largest family of membrane-bound receptors in mammals and represent 3% of the genome (Fredriksson et al, 2003). Based on phylogenetic criteria, Fredriksson et al. proposed a classification containing five main groups: Glutamate, Rhodopsin, Adhesion, Frizzled/Taste2 and Secretin (shortened to the acronym GRAFS) (Fredriksson et al, 2003; Lagerstrom & Schioth, 2008). (Figure 7 and Figure 8).

2.2.1. Rhodopsin receptors The family of Rhodopsin receptors is the largest family with 85% of the GPCRs. These receptors present some common characteristics, such as a NSxxNPxxY pattern in the TMVII domain or a DRY pattern (or D(E)-R-Y(F)) between the TMIII and the IC2 loop. This family contains notably receptors to opsines (e.g. rhodopsin receptor), prostanoid receptors (e.g. thromboxane receptor TP), serotonin receptors (e.g. 5-HT1A and 5-HT1B receptors),

21 somatostatin receptors (e.g. SST2R), cannabinoid receptors (e.g. type-1 cannabinoid receptor CB1R)

2.2.2. Secretin receptors Secretin family is a very small family containing GPCRs that bind peptidic hormones. The binding is due to an extracellular binding domain called Hormone Binding Domain (HBD) in the long N-terminal tail. This class contains notably receptors to glucagon (GCGR) and to growth-hormone-releasing hormone (GHRHR).

2.2.3. Adhesion receptors This family gathers GPCRs with a long N-terminal tail presenting adhesion-like patterns, such as EGF-like repetitions, mucin-like regions and cystein-rich patterns (Hayflick, 2000; Yona et al, 2008).

2.2.4. Glutamate receptors GPCRs of this family bind through an interaction with the N-terminal tail that has been compared to a Venus flytrap mechanism where two lobes of this region shape a cavity where the ligand binds enabling the receptor activation (Kunishima et al, 2000). This family contains notably metabotropic glutamate receptors and receptors to GABAB.

2.2.5. Frizzled/Taste2 receptors Frizzled receptors are activated through secreted glycoproteins called Wnt. The family name is a reference to the twisted form of the Wnt ligand. The Taste2 family includes taste receptors. Indeed, Taste2 receptors can be activated by cycloheximide and 6-n-propyl-2-thiouracil, molecules responsible for bitter taste (Chandrashekar et al, 2000).

Figure 7: Phylogenetic analysis of the Rhodopsin family in rat genome (Gloriam et al, 2007) Ligand types are indicated with colours: blue=peptide, lilac=amine, green=lipid-like, brown=purine, turquoise=opsin, black=other, red=orphan.

However, about 800 GPCRs are identified only through their sequence: neither their ligand nor their physiological role is known. These receptors are called orphan receptors (Wettschureck & Offermanns, 2005). Thus, many studies are currently conducted in order to identify ligands and functions of these GPCRs with high pharmacological potential (Oh et al, 2006). All families described above, most notably the Adhesion receptor family, contain some orphan receptors.

Figure 8: Phylogenetic analysis of other GPCR families (Gloriam et al, 2007) (a) Glutamate (b) Adhesion (c) Frizzled (d) Secretin (e) Taste2

2.3. GPCR activation Because of the diversity of their ligands, GPCR binding sites are various. While many small organic ligands bind within the TM segments, peptides, hormones and proteins often bind to the N-terminus and EC loops (Ji et al, 1998). However, ligand binding induces intracellular conformational changes that are similar for all GPCRs enabling the activation of G-proteins. The rhodopsin receptor is the first GPCR whose high-resolution structure has been described. Thus, currently, conformational changes induced by rhodopsin receptor activation are well

24 known. As Rhodopsin-like GPCRs share a similar structure, it is possible to extrapolate to this family. However, for GPCRs classified in other families, the extrapolation is less obvious. Here, I will briefly describe conformational changes induced by rhodopsin receptors activation. Photoactivation of rhodopsin induces a rotation and tilting of TM6 relative to TM3 (Farrens et al, 1996). Morever, changes in the cytoplasmic domain spanning TM1 and TM2 and also in the cytoplasmic end of TM7 have been observed (Altenbach et al, 2001). These changes lead to the coupling and activation of G proteins. However, this activation mechanism corresponds to bimodal switches as rhodopsin receptors display one inactive state that switches to an active state after activation by a photon and it is now well known that GPCRs are not bi-stable rigid structures. Indeed, while the TM domains are quite rigid, EC loops and IC loops enable high conformational flexibility. Thus, GPCRs can adopt a multitude of conformations and activation models have to take it into account.

2.3.1. Energy landscape diagram The energy of each conformation can be plotted in function of the conformation itself creating a continuum (Figure 9a, 9b) (Kobilka & Deupi, 2007). At the basal state (receptor alone), inactive conformations display a lower energy level than active conformations that favours GPCR inactive conformations. When an agonist binds the receptor, it stabilizes active conformations by reducing their energy. Thus, active conformations display a lower energy level than inactive conformations that favours GPCR active conformations. Therefore, efficacy of an agonist depends on its ability to decrease energy level of active conformations energy.

Figure 9: Theoritical energy landscape of a GPCR (Kobilka & Deupi, 2007) (a) Conformational states of an unbound GPCR (b) Effect of agonist binding on energy landscape (c) Effect of inverse agonist binding on energy landscape

25 2.3.2. Activation-induced conformational changes TM segments are rigid structures. At basal state, nonconvalent interactions maintain the conformation. Disrupting these interactions lead to destabilization and conformational changes. For Rhodospin-like GPCRs, two main disruption mechanisms, called Molecular Switches, have been identified (Figure 10). - The Rotamer Toggler Switch has been established with β2-adrenergic receptors using mutagenesis studies and computer simulations. Rotameric positions of Cys2856.47, Trp2866.48 and Phe2906.52 modulate the bend angle of TM6 around Pro2886.50. Agonist binding changes the angle of the helical kink formed by Pro2886.50 resulting in movement of the cytoplasmic end of TM6 (Shi et al, 2002). - The Ionic Lock involves Asp3.49/ Asp3.50 pair in the highly conserved (D/E)RY motif and Glu6.30. This ionic interaction holds together cytoplasmic ends of TM3 and TM6 at basal state (Ballesteros et al, 2001). GPCR activation leads to disruption of this interaction.

(a) Figure 10: GPCR molecular switches (Deupi & Kobilka, 2007)

(a) Rotamer toggle switch: Agonist binding changes the angle of the helical kink formed by Pro2886.50 resulting in movement of the cytoplasmic end of TM6.

(b)

(b) Ionic lock: Agonist binding disrupts ionic interactions between TM3 and TM6.

26 2.4. Constitutive GPCR activity Theoretical energy landscapes show that multitude of GPCR states leads to an energy continuum. Thus, at the basal state, even if active conformations are not favoured due to high energy levels, it is not excluded that some receptors display active conformations. Therefore, a basal activity exists, called constitutive activity. This concept leads to define ligands in function of their ability to favour active or inactive states (Figure 11). At basal state, some receptors are active leading to a basal biological response. An agonist is a ligand stabilizing receptor active states and its binding leads to an increase of active receptors and to a downstream biological response. Conversely, an inverse agonist is a ligand stabilizing inactive receptor conformations decreasing the basal biological response. A neutral antagonist binds the receptor without inducing changes in the conformational state. In the next part, I will present first how the constitutive activity has been included in theoretical pharmacological models, then, I will present studies demonstrating the physiological relevance of constitutive activity.

Figure 11: Constitutive GPCR activity

(a) At basal state, constitutively active receptors are distributed between active and inactive states. (b) Agonist stabilizes active state leading to an increase of GPCR number in active state. (c) Inverse agonist stabilizes inactive conformation increasing GPCR number in inactive state. (d) Neutral antagonist does not change basal state distribution between active and inactive conformations but prevent ligand binding. (Meye et al, 2014)

e R : inactive state R* : active state (e) Biological responses induced by R R* agonist constitutive activity (Leterrier, 2006, R R* neutral antagonist PhD manuscript)

Biological response Biological R R* inverse agonist Ligand concentration (log)

27 2.4.1. Constitutive activity theories and models 2.4.1.1. Energy landscape diagram The energy landscape diagram (described above) can be used to have a general vision of constitutive activity. Indeed, as shown in Figure 9c, binding of an inverse agonist stabilizes inactive conformations by decreasing their energy. Thus, energy difference between active and inactive states, called activation energy barrier, is amplified leading to lower probability for active states to appear. Moreover, all GPCRs do not display the same level of constitutive activity. Basal activity depends on two parameters: activation energy barrier level and number of active basal conformations (Figure 12) (Kobilka & Deupi, 2007). On the one hand, due to a greater conformational flexibility, activation energy barrier is diminished leading to an increase of activated receptors at basal state (Figure 12b). On the other hand, lowest energy conformation at basal state can be allocated to partially active intermediate conformations (Figure 12c). Depending on the GPCR, one or both of these theories can explain constitutive activity.

Figure 12: Theoritical energy landscape of a GPCR with low or high basal activity (Kobilka & Deupi, 2007) (a) Conformational states of a GPCR with low constitutive activity (b) Conformational states of a GPCR with high basal activity owing to a low activation barrier (c) Conformational states of a GPCR with high basal activity owing to more active basal conformations

2.4.1.2. The probabilistic model This model represents a given GPCR in the conformational space (Figure 13). Each conformation is defined by its probability to appear. We can link this model to the energy landscape diagram. Indeed, the lower is the energy of a conformation, the higher is its probability to appear. The addition of a ligand modifies this probability, differently for each conformation. Thus, the Gaussian distribution of conformations is shifted after ligand binding.

Figure 13: Probabilistic model (Kenakin, 2004) (a) Probabilistic view of unbound GPCR conformations. (b) Probabilistic view of ligand-bound GPCR conformations. (c) Gaussian distribution of conformational states is shifted in presence of a ligand. The pharmacological activity of a ligand is given by the quantity and type of receptor conformations that the ligand stabilizes in conformational space. An agonist stabilizes active conformations increasing their probability to appear and its efficacy is higher if it stabilizes more active conformations or preferentially active conformations able to activate G proteins. Conversely, an inverse agonist stabilizes inactive conformations increasing their probability to appear.

2.4.1.3. Ternary complex model Another point of view on GPCR activation consists in models based on linkage theory (Figure 14a, 14b). Clarks occupancy theory published in 1933 describes agonist binding on a GPCR as a chemical reaction with a dissociation constant KA. However, this model considers that receptor activation leads to 100% of efficacy. 23 years later, Stephenson and Furchgott modified this model to include the efficacy (Stephenson, 1956). However, this model lacks implication of G proteins in GPCR activation and the increasing interest for allosterism led De Lean and colleagues to propose a model taking into account the interaction between GPCR and G proteins, called ternary complex model (De Lean et al, 1980) (Figure 14c). Thus, this model also includes the availability of G proteins in the environment and only the ternary complex agonist-receptor-G protein can activate signaling pathways.

29 Figure 14: GPCR activation models (1)

(a) Clarks occupancy theory (A: ligand; R: receptor ; KA: equilibrium constant; Q: biological response) (b) Stephensen-Furchgott model (A: ligand; R: receptor ; KA: equilibrium constant; e: efficiency; S: second messenger; Q: biological response) (Park, 2012)

(c) Ternary complex model (H: ligand; R: receptor; X: additional component (G protein); E: effector) (De Lean et al, 1980)

2.4.1.4. Extended ternary complex model (ETC) and cubic ternary complex model (CTC) In 1989, Costa and Herz described for the first time antagonists of δ-opioid receptors with a negative intrinsic activity, defining the term of inverse agonism (Costa & Herz, 1989). Thus, models had to evolve to take into account that GPCRs can adopt active conformations and signal in absence of agonist. Therefore, in 1993, the extended ternary complex (ETC) model has been described (Samama et al, 1993) (Figure 15a). In this model, receptors can spontaneously adopt an active conformation (Ra) and couple G proteins inducing constitutive signaling. However, in 1996, Weiss and colleagues proposed cubic ternary complex (CTC) model considering that not only active but also inactive receptor conformations can couple to G proteins (Weiss et al, 1996) (Figure 15b). Relevance of this model was demonstrated later. Indeed, Vasquez and Lewis showed in 1999 that the inactive form of type-1 cannabinoid receptor, stabilized by the inverse agonist SR141716A, can sequester Gi/o proteins (Vasquez & Lewis, 1999). Thus, inverse agonists stabilize inactive G protein-bound non signaling conformations leading to G proteins sequestration avoiding remaining active receptors to couple and activate signaling pathways.

30 Figure 15: GPCR activation models (2)

(a) Extended ternary complex model (ETC)

(b) Cubic ternary complex model (CTC)

(A: ligand; Ri: inactive receptor; Ra: active receptor; G: G protein; L/K indicate various equilibrium constants; α, β, γ indicate modulators of equilibrium constants). (Kenakin, 2004)

The CTC model is convenient to give access to molecular binding contribution but only for each of eight possibilities, while the probalistic model forecasts a multitude of conformations. Moreover, another parameter is lacking: the agonist efficacy.

Finally, it is important to distinguish both terms constitutive activation and constitutive activity. On the one hand, constitutive activation refers to receptor ability to switch spontaneously between inactive and active conformations due to structural mobility. However, active conformations do not always activate downstream signaling pathways but induce only receptor desensitizing or endocytosis (Kobilka & Deupi, 2007). On the other hand, constitutive activity is a pharmacological term implying a spontaneous recruitment of signaling pathways, independently of ligand binding.

2.4.2. Physiological constitutive activity Overexpression of receptors in heterologous systems has been widely used to study constitutive activity (Seifert & Wenzel-Seifert, 2002). Indeed, detecting constitutive activity of endogenous receptors can be difficult, due to the lack of sensitivity. Overexpression allows

31 to increase both inactive and active receptors at the basal state and to detect endogenously low constitutive activity. In 1989, Costa and colleagues demonstrated that ICI-174,864, an antagonist of δ-opioid receptors, not only prevents agonist-induced GTPase increase, but decrease by itself GTPase activity in NG108-15 cells (Costa & Herz, 1989). ICI-174,864 showed also inverse agonist effects by potentiating forskolin responses in HEK293 cells, while δ-opioid receptor agonists inhibit this pathway (Chiu et al, 1996). Also, overexpression of β2-adrenergic receptors revealed the first proof of in vivo constitutive activity (Bond et al, 1995). However, physiological relevance of constitutive activity was shown with in vivo experiments: in vivo constitutive activity has been demonstrated for endogenous histamine H3 receptors (Morisset et al, 2000), serotoninergic receptors 5-HT2A and 5-HT2C (Berg et al, 2005; De Deurwaerdere et al, 2004). Also, Agouti-Related Protein (AgRP) has been identified in vivo as an endogenous inverse agonist of melacortin receptors (Ollmann et al, 1997) and plays a major role in appetite control (Adan & Kas, 2003).

2.5. G proteins structure, activation and subtypes G proteins are heterotrimeric proteins composed of an α subunit, that binds and hydrolyses GTP, and two indissociable βγ subunits. At basal state, the α subunit binds a GDP molecule and is associated to the βγ complex. This heterotrimer can couple to an activated GPCR. This coupling leads to the exchange of a GDP molecule with a GTP molecule. The α subunit dissociates from the activated receptor and from the βγ complex. Thus, both subunits are able to modulate the activity of various second messengers (see below) that regulate the cellular activity. Finally, GTP is hydrolysed in GDP, all the subunits are then able to associate and couple to an other activated receptor (Figure 16) (Lambert, 2008).

32 Figure 16: Functional cycle of G protein activity

G protein couples to a receptor and is activated when the receptor is activated by an agonist (Ag), leading to the dissociation of α subunit and βγ complex. (Wettschureck & Offermanns, 2005)

Several subtypes of G proteins exist in mammals. The α subunit type determines the basic properties of a given G protein and can be divided in four families: Gαs, Gαi/Gα0, Gαq/Gα11 and Gα12/Gα13. The βγ complex is composed of an association of one of the five β subunits subtypes and one among the twelve γ subunits subtypes. Most receptors are able to activate more than one G protein subtype. The pattern of G proteins activated by a given receptor determinates the cellular and biological response after receptor activation. In almost all G proteins, βγ complex activate the phospholipases C-β (PLC-β), phosphoionositide-3-kinases (PI3K) and G protein-regulated inward rectifier potassium channels (GIRK) and inhibit P/Q- type and N-type calcium channels. Moreover, it can modulate the activity of adenylyl cyclases (ACs) but their action depends on AC subtype (see below). On the contrary, depending on α subunit subtype, G proteins activation leads to various modulations of effectors activity (Figure 17).

2.5.1. Gi/o and Gs proteins

αi/o subunits regulate ACs activity decreasing their activity and thus inhibiting the synthesis of cAMP. On the contrary, αs subunits increase AC activity leading to an increased cAMP concentration in the cytosol.

2.5.2. Gq/11 proteins

αq and α11 subunits activate PLC-β leading to an increase of diacylglycerol (DAG) that activates the protein kinases C (PKC).

2.5.3. G12/13 proteins

α12 and α13 subunits activate Rho-guanine nucleotide exchange factor leading to an activation of the small GTPase RhoA that modulate actomyosin-based structures contractility.

Thus, the diversity of ubiquitous G proteins enables to modulate differently numerous signaling pathways involved in various cellular processes, depending on cell type. In the CNS, G-proteins are highly expressed and play a major role, notably in synaptic transmission (Wettschureck & Offermanns, 2005):

- Gs proteins couple to β1- and β2-adrenoreceptors, dopamine D1 and D5 receptors and adenosine A2 receptors.

- Gi/o proteins couple to μ- and δ-opioid receptors, α2-adrenoreceptor, GABAB receptors, CB1R. Indeed, their activation leads to inhibition of presynaptic N- and P/Q-type calcium channels and also to the activation of post-synaptic GIRKs leading to an inhibition of neurotransmitter release.

- Gq/11 proteins couple to metabotropic glutamate receptors mGluR1 and mGluR5.

Importantly, recent studies have demonstrated the existence of G-protein independent signaling dowstream of GPCRs. This signaling involves β-arrestins (Azzi et al, 2003). Thus, some researchers now replaced the term G-Protein Coupled Receptor by heptameric receptor.

During my PhD, I focused on receptors coupled to Gi/o proteins and their action on ACs. Thus, in the next paragraph, I will present the structure and subtypes of ACs.

Figure 17: Effectors dowstream of G proteins Different signaling pathways are modulated depending on the α subunit type of G proteins. The βγ complex modulate the same effectors. GIRK: G protein-regulated inward rectifier potassium channel; PLC-β: phospholipase C-β; PI-3-K: phosphoionositide-3- kinase; PIP2: phosphatidylinositol 4,5-bisphosphate; IP3: ionositol 1,4,5-triphosphate; DAG: diacylglycerol; PKC: protein kinase C; Rho-GEF: Rho-guanine nucleotide exchange factor. (Wettschureck & Offermanns, 2005)

2.6. Adenylyl cylclases: structure, activation and subtypes Membrane-bound ACs are composed of two repeats of a six transmembrane module (M1 and M2) and two cytoplasmic domains of approximaticely 40kDa (C1 and C2) in between (Figure 18) (Pierre et al, 2009).

35 Figure 18: Adenylyl cyclase structure

ACs are composed of two six-transmembrane domains (M1 and M2) and two cytosolic domains in between (C1 and C2). (Pierre et al, 2009)

The catalytic active site is created at the interface of C1 and C2. Gαs proteins bind to the outside of C2, inducing a conformational change and the rearrangement of both domains. A single molecule of forskolin can bind in a pocket at the interface of C1 and C2 domains inducing the same conformational change. By contrast, Gi/o proteins bind the C1 domain and interfere with this conformational change, preventing activation. To be converted in cAMP, ATP binds in a catalytic pocket formed at C1-C2 interface that contains two magnesium ions that contribute to the catalysis. Nine isoforms of membrane-bound ACs have been identified (AC1-AC9) and also one soluble form of AC (sAC). Membrane-bound AC isoforms are not sensitive to the same ligands and/or do not induce the same response (Table 2) (Hanoune & Defer, 2001). These ACs can be classified in function of their response to calcium and calmodulin (CaM): - AC1, AC3 and AC8 are positively regulated by calcium, - AC5 and AC6 are negatively regulated by calcium and AC9, the only forskolin- insensitive isoform, is inhibited by calcineurin, - AC2, AC4 and AC7 are insensitive to calcium. Furthermore, βγ subunits modulate differently the various AC isoforms: - AC2, AC4 and AC7 are stimulated by βγ subunits, - AC1, AC5 and AC6 are inhibited by βγ subunits, - AC3, AC8 and AC9 are insensitive to βγ subunits. ACs are ubiquitous proteins but depending on the tissue, not all isoforms are expressed and/or not at the same level. For example, in the CNS, all ACs are expressed but AC7 is expressed at very low levels and AC4 is only expressed in blood vessels (Visel et al, 2006) and their expression level varies depending on developmental stage. Notably, AC1 is highly expressed

36 in developing neurons but its expression is reduced in mature neurons and, on the contrary, AC8 expression is low during development and becomes high in the adult brain (Nicol & Gaspar, 2014).

Table 2: Responses of ACs to various modulators FSK: forskolin; CaM: calmodulin (Hanoune & Defer, 2001).

2.7. GPCR regulation Importantly, signaling downstream of a GPCR is not constant but depends on previous activations, due to cellular memory (Hausdorff et al, 1990). Indeed, while high stimulation leads to reduced system activability, a phenomenon called desensitization, low previous activation leads to increased response, called sensitization. Therefore, cellular response to an agonist depends on previous stimulations. Indeed, receptor activity (and activability) is regulated by cellular processes: desensitization, endocytosis and sequestration, recycling or degradation (Figure 19). These mechanisms enable a turn-over in membrane-bound receptors. In this paragraph, I will present these processes.

Figure 19: GPCR regulation : desensitization, recycling and downregulation (degradation) (Luttrell & Lefkowitz, 2002)

β-arr: β-arrestin; Dyn: Dynein; AP-2: Activating Protein 2; GRK2: G-protein coupled receptor kinase 2.

2.7.1. Desensitization Desensitization plays a major role in the control of signaling duration. Indeed, signaling induced by GPCR activation is stopped by this process, preventing toxic effects of prolonged cell activation. Two complementary mechanisms are involved in desensitization: decoupling from G proteins, signaling arrest, and endocytosis, which reduces membrane-bound activable receptors. First, GPCR Kinases (GRKs) phosphorylate the C-terminal tail domain or intracellular loops of active GPCRs (Pao & Benovic, 2005; Ribas et al, 2007). GPCR phosphorylation enables β- arrestin recruitment in charge of desensitization. Indeed, β-arrestins interact with GPCRs and favour G proteins decoupling through steric effects. Moreover, they group activated receptors and recruit necessary proteins for clathrin-mediated endocytosis, such as Assembly Protein-2 (AP-2) and clathrin (Lohse et al, 1990). There are also heterologous desensitization processes that do not require GPCR activation. They involve other kinases, such as Protein Kinase C (PKC), Protein Kinase A (PKA) or

38 Casein Kinase II (CKII) that phosphorylate and desensitize receptors (Hanyaloglu et al, 2001; Martin et al, 2004; Namkung & Sibley, 2004).

2.7.2. Endocytosis Endocytosis is a cellular process consisting in transport of solid or liquid matter from extracellular to intracellular compartments. It is also used in order to internalize ligand-bound G protein-decoupled receptors. Thus, it enables receptor sorting by removing non-activable receptors from the plasma membrane. Several endocytosis mechanisms exist. Here, I will present the main process involved in GPCR internalization: clathrin-mediated endocytosis (CME). While desensitization begins within seconds after agonist exposure, GPCR internalization occurs more slowly, within several minutes after agonist exposure (Luttrell & Lefkowitz, 2002). CME involves the concentration of activated agonist-bound receptors into a clathrin- coated pits on the plasma membrane. Other scaffold proteins, such as AP-2, slot into pits between clathrin network and plasma membrane (Conner & Schmid, 2003). Membrane invagination requires a lipidic rearrangement in the bilayer, involving various partners such as epsin, endophilin and synaptophysin. Cholesterol also plays a major role in this step, either facilitating interactions between plasma membrane and AP-2, or enabling local membrane destabilisation necessary for invagination (Subtil et al, 1999). The last endocytosis step is membrane fission transforming clathrin-coated pits into intracellular clathrin-coated vesicles. The GTPase dynamin interacts with the clathrin network and pinches off clathrin- coated pits from the plasma membrane. Once a clathrin-coated vesicle is internalized from the plasma membrane, the clathrin network dissociates and the vesicle heads toward sorting endosomes.

2.7.3. GPCR recycling and degradation Endosomes are divided into three compartments: early endosomes, late endosomes and recycling endosomes. They are notably distinguished by markers, such as Rab GTPases (Stenmark, 2009). - Early endosomes: after internalization, vesicles reach early endosomes through a Rab5-dependent mechanism. From there, receptors are either directly recycled (rapid recycling) and reappear on plasma membrane through a Rab4-dependent mechanism, or they are guided to recycling endosomes through Rab15-dependent process, or they stay in the early endosomes which will mature to late endosomes.

39 - Late endosomes: mature early endosomes become late endosomes marked by Rab7 and Rab9. All receptor-containing vesicles in late endosomes are guided to lysosomes in order to be degraded. - Recycling endosomes: Receptors present in recycling endosomes are guided back to the plasma membrane through a Rab11- and Rab35-dependent process (slow recycling). Therefore, after internalization, GPCRs are either degraded or recycled back to the plasma membrane and become activable again. How this decision is taken is still unclear. On the one hand, there are receptors known to recycle easily, such as β2-adrenergic receptors (Pippig et al, 1995), μ-opioid receptors (Tsao & von Zastrow, 2000) and gonadotropin-releasing hormone (GnRH) receptors (Vrecl et al, 1998). On the other hand, agonist concentration can be a criterion. Indeed, studies showed that low CB1R activation leads to receptor recycling, while high agonist concentration can associate CB1R with GASP1 protein, leading to its degradation (Martini et al, 2007; Tappe-Theodor et al, 2007).

To conclude, GPCR regulation involves various cellular processes. These processes enable a fine control and regulation of both the level and the efficiency of GPCRs on the plasma membrane. Interestingly, endocytosis and recycling are also involved in GPCR targeting.

40 3. Questions Dr Kenakin pointed out that extrapolating pharmacological data from one screening system to another one is almost impossible due to high dependence of GPCR pharmacology on tissue type (Kenakin, 2009). Notably, he proposed that only four pharmacodynamic parameters can be independent of the studied system: affinity, relative efficiency of two different agonists, orthosteric/allosteric binding and rate of dissociation of the ligand from the receptor. Because neuronal polarization is established and maintained through highly controlled protein targeting, several, if not most, neuronal GPCRs display a polarized distribution. However, their traffic and pharmacology are generally studied in non-polarized cell lines that do not give access to the in vivo GPCR localization and avoid extrapolation. Moreover, due to the differences between the composition of the somatodendritic and the axonal plasma membrane, GPCRs could display differential pharmacology, depending on their sub-neuronal localization. Focusing on one of the most abundant GPCR in the brain, the type-1 cannabinoid receptor (CB1R), that displays polarized distribution and constitutive activity (see below), we proposed to answer the following questions:

- How does neuronal polarity influence CB1R traffic? - How does neuronal polarity modify CB1R pharmacology compared to non-polarized cell-lines? - How do neuronal polarity and structural determinants modify constitutive CB1R activity? - Can results obtained on CB1Rs be generalized to other axonally polarized neuronal

GPCRs, such as 5-HT1B serotonin receptors? - Reciprocally, how do CB1Rs modify neuronal polarity?

In this manuscript, I will detail the required technical development and the results I obtained during my PhD that give some answers to these questions, starting by an introduction about what is currently known about CB1R targeting and signaling.

41 4. Type-1 cannabinoid receptor (CB1R)

4.1. Identification, structure and distribution Both hashish and marijuana are extracts of the same plant: Cannabis Sativa. For more than 3000 years, cannabis has been used worldwide as a recreational and therapeutic drug. Its psychoactive substance, the Δ9-Tetrahydrocannabinol (Δ9-THC), has been discovered only in 1967, enabling the understanding of its pharmacological action (Mechoulam et al, 1967; Mechoulam & Gaoni, 1967). However, it is only in 1990 that a coding sequence has been isolated and a cannabinoid receptor, called type-1 cannabinoid receptor (CB1R), has been identified (Matsuda et al, 1990). CB1R is a GPCR containing 473 amino acids, including 117 amino acids in the extracellular N-terminal tail. Interestingly, rat and human CB1Rs share 98% of common amino acids (Gerard et al, 1990). CB1R is one of the most abundant GPCRs in the CNS. Its distribution has been studied with various techniques: autoradiography, in situ hybridization and immunohistochemistry. Important receptor density was found in cerebral cortex, hippocampus, amygdala, basal ganglia and cerebellum (Figure 20A, 20B). These regions are in accordance with behavioural effects of cannabis (Herkenham et al, 1991). Interestingly, CB1Rs are highly expressed in GABAergic neurons and display a lower but significant expression in glutamatergic neurons (Marsicano & Lutz, 1999). Development of antibodies specifically the against C-terminal tail (Egertova & Elphick, 2000) or the N-terminus (Pettit et al, 1998) enabled determination of sub-neuronal CB1R localization. While first studies reported presynaptic localization of CB1Rs (Katona et al, 1999; Katona et al, 2006), latter studies demonstrated an intracellular distribution on endosomes in the somatodendritic compartment and high presence in extrasynaptic regions of axons (Leterrier et al, 2006; Matyas et al, 2006; McDonald et al, 2007) (Figure 20C, 20C).

42 C

Figure 20: CB1R distribution in the CNS (A) Binding sites of the agonist CP55,940 in the rat brain (Freund et al, 2003) (B) CB1R mRNA localization in rat brain (Freund et al, 2003) (C-C) Sub-neuronal localization of CB1R in hippocampal cultured neurons of rat embryos (Leterrier et al, 2006)

4.2. Other cannabinoid receptors In 1997, a second cannabinoid receptor, called CB2R, has been cloned. While CB1Rs are predominantly expressed in the CNS, and notably in neurons, CB2Rs are predominantly expressed in immune cells. However, one study reported presence of CB2Rs in brainstem neurons (Van Sickle et al, 2005). Importantly, CB2Rs share 44% of identity with CB1Rs in humans (Munro et al, 1993). These similarities lead to common pharmacological properties and CB1Rs ligands are generally also able to bind CB2Rs. Moreover, evidence for non-CB1R and non-CB2R cannabinoid effects have shown the existence of an other receptor able to bind Δ9-THC and other endocannabinoids: the orphan GPCR GPR55 (Baker et al, 2006). GPR55 mRNA is present in both human and rat brains, predominantly in caudate nucleus and putamen, but displays low level in hippocampus, frontal cortex or cerebellum (Sawzdargo et al, 1999).

4.3. CB1R pharmacology

4.3.1. Ligands Several ligands have been reported to bind CB1Rs. In Table 3, I listed some of them with indicative inhibition constant Ki for CB1R and CB2R. These ligands can be divided in five groups.

Compound Ki on CB1R (nM) Ki on CB2R (nM) Phytocannabinoids Δ9-THC 39.5 40 HU 210 0.73 0.22 Synthetic agonists CP 55,940 0.6 0.7 WIN 55,212-2 1.89 0.28 anandamide 89 > 279 Endogenous 2-arachidonoylglycerol 58.3 145 agonists Noladin ether 21.2 >3000 SR141716A 11.5 1640 Synthetic inverse AM251 7.5 2290 agonists AM281 12 4200 Endogenous Hemopressin NE NE inverse agonist

Table 3: Cannabinoid ligands. 3 3 Inhibition constant Ki was evaluated in vitro with [ H]CP 55,940 or [ H]HU 243 (Palmer et al, 2002)(Pertwee, 2010, Cannabinoid receptors ligands, Tocris Bioscience) (Heimann et al, 2007) NE: Not Evaluated

4.3.1.1. Phytocannabinoids Phytocannabinoids are cannabinoids present in the cannabis plant. The most common is Δ9- THC, an agonist of both CB1R and CB2R. Other phytocannabinoids exist, such as cannabidiol, a non-psychoactive compound that displays weak antagonist effect on CB2Rs (Ki=4.5µM) (Thomas et al, 2007).

4.3.1.2. Synthetic agonists CP 55,940 was the first cannabinoid synthesized by Pfizer that acts as an agonist on CB1Rs (Little et al, 1988). This molecule showed similar effects as Δ9-THC and its tritiated form [3H]CP 55,940 enabled the discovery of CB1R by radiolabeling (Devane et al, 1988). Both CP 55,940 and HU 210 are synthetic agonists derived from Δ9-THC and display high CB1R affinity. Later, other synthetic compounds were found to bind CB1Rs, even though their structures are not derived from Δ9-THC. The aminoalkylindole WIN 55,212-2 (WIN) is one of them (D'Ambra et al, 1992).

4.3.1.3. Endogenous agonists Endogenous agonists, called endocannabinoids, are lipidic molecules derived from eicosanoids. The first discovered endocannabinoid was called anandamide (AEA) from Sanskrit word ananda meaning bliss (Devane et al, 1992). Then, other endocannabinoids have been reported: homo-γ-linoenoylethanolamide (HEA) and docosatetraenoylethanolamide (DEA), both analogs of anandamide (Hanus et al, 1993), 2-arachidonoylglycerol (2-AG) (Mechoulam et al, 1995; Sugiura et al, 1995) and 2-arachidonoyl glyceryl ether (noladin, 2- AGE) (Hanus et al, 2001). To date, anandamide and 2-AG are the most studied endocannabinoids.

4.3.1.4. Synthetic inverse agonists Several CB1R antagonists, presenting inverse agonist effects, have also been developed: SR141716A, also called Rimonabant (Rinaldi-Carmona et al, 1994), AM251 (Gatley et al, 1996) and AM281 (Gatley et al, 1998).

4.3.1.5. Endogenous inverse agonists Recently, peptide hemopressin was shown to behave as a specific CB1R inverse agonist (Heimann et al, 2007). Even though hemopressin is an endogenous molecule, its physiological role as CB1R antagonist is still unclear.

4.3.2. Binding sites Because the three-dimensional CB1R structure has not been established yet, exact ligand binding sites are still unclear. However, a recent study proposed a binding mechanism of 2- AG for CB2R: as 2-AG is a lipid molecule, able to diffuse in membrane bilayer, one can assume that it reaches the receptor at the TM6/7 interface and enter in the binding pocket (Hurst et al, 2013). Due to similarities between CB1R and CB2R, a similar mechanism could be extended to CB1R. Among synthetic agonists, WIN binding requires an interaction with the second extracellular loop of CB1R (Bertalovitz et al, 2010) and mutation studies demonstrated that WIN does not bind to the same site as CP 55,940 or HU 210 (Kapur et al, 2007). Thus, CB1R may display several binding sites but we still do not know how for most identified CB1R ligands how they bind the receptor and modulate its activity.

4.3.3. Signaling Even before CB1R cloning, Howlett and colleagues demonstrated that Δ9-THC decreases cAMP concentration in neuroblastoma cells via activation of Gi proteins (Howlett & Fleming,

1984; Howlett et al, 1986). Later, CB1R was thus identified as Gi/o-proteins coupled receptor. In addition, CB1Rs modulate various signaling pathways (Figure 21).

4.3.3.1. Calcium and potassium channels CB1R activation leads to inhibition of various calcium channels: L-type calcium channels in cat cerebral arterial smooth muscle cells (Gebremedhin et al, 1999), Q-type in CB1R transfected AtT-20 cell line (Mackie et al, 1995) and N-type in neuroblastoma cells (Mackie & Hille, 1992) and CB1R transfected neurons (Pan et al, 1996). All these effects are sensitive to pertussis toxin (PTX) that inhibits Gi/o proteins and involve βγ subunit (Ikeda, 1996). Moreover, CB1Rs activate G protein-coupled inwardly-rectifying potassium channels (GIRKs) through the βγ subunit (Lujan et al, 2009) in neuroblastoma cells (Mackie et al, 1995) and in neurons (Bacci et al, 2004).

Figure 21: CB1R signaling β-arr: β-arrestin; GIRK: G protein-coupled inwardly- rectifying potassium channels; PI3K: Phosphoinositide 3-kinase; MAPK: Mitogen-activated kinase; JNK: c-Jun N-terminal kinase ; cAMP : cyclic adenosine monophosphate ; PKA : cAMP- dependent protein kinase A ; ERK: extracellular signal- regulated kinase; nNOS : neuronal NO synthase ; iNOS inducible NO synthase.

4.3.3.2. Mitogen-activated protein kinase (MAPK) Several studies reported that CB1Rs modulate MAPK activity, through activation of Phosphoinositide 3-kinase (PI3K) (Gomez Del Pulgar et al, 2002). Indeed, CB1Rs activate

46 p38 MAPK and c-Jun N-terminal kinase (JNK) (Derkinderen et al, 2001; Liu et al, 2000;

Rueda et al, 2000). However, the activation of p42/p44 MAPK (ERK1/2) is sensitive to PTX and a PKA inhibitor, demonstrating implication of the αi/o subunit (Davis et al, 2003). Moreover, recent studies reported that β-arrestin 2, involved in GPCR desensitization, induces late signaling after GPCR activation. Thus, modulation of MAPK signaling downstream of CB1Rs could also be due to late β-arrestin 2 activation (Luttrell & Lefkowitz, 2002).

4.3.3.3. Nitric Oxide Synthase (NOS) Some studies also reported modulation of nitric oxide synthesis downstream of CB1Rs. Indeed, CB1R activation leads to a decrease of inducible NOS (iNOS) activity in rat microglial cells (Waksman et al, 1999) and also a decrease of neuronal NOS (nNOS) activity both in cerebellar granule cells (Hillard et al, 1999) and mice cerebral cortex (Kim et al, 2006), through the decrease of PKA activity.

4.3.4. Physiological roles The high density of CB1Rs in the CNS predicts their implication in numerous physiological processes. Thus, the endocannabinoid system is involved in various disorders, such as pain, inflammation, stroke, multiple sclerosis, movement disorders (Parkinsons and Huntingtons disease), epilepsy, mental disorders (schizophrenia, anxiety and depression), drug addiction, and obesity. It appears as an emergent target of pharmacotherapy (reviewed in (Pacher et al, 2006)). Here, I will present some of physiological roles of CB1Rs.

4.3.4.1. Synaptic transmission Endocannabinoids are involved in synaptic plasticity (see above). Indeed, they induce DSE (Kreitzer & Regehr, 2001) and DSI (Ohno-Shosaku et al, 2001; Wilson & Nicoll, 2001) through retrograde signaling and inhibition of pre-synaptic calcium channels. They also mediate LTD at both excitatory (Gerdeman et al, 2002; Robbe et al, 2002) and inhibitory (Chevaleyre & Castillo, 2003; Marsicano et al, 2002) synapses. These processes require retrograde signaling of endocannabinoids and activation of pre-synaptic CB1Rs that inhibits the cAMP/PKA pathway (Chevaleyre et al, 2006; Heifets & Castillo, 2009). However, as endocannabinoids are lipidic molecules, it is difficult to explain how they are released and cross the synaptic cleft. This process is still poorly understood. Moreover, 2-AG is involved in slow self-inhibition (SSI) of low-threshold-spiking (LTS) interneurons (Bacci et al, 2004). Synapse activation leads to an increase of intracellular post-

47 synaptic calcium, stimulating 2-AG synthesis. Consequently, post-synaptic CB1Rs are activated by the 2-AG leading to synapse silencing (Marinelli et al, 2008). Interestingly, this process involves post-synaptic CB1Rs and does not require 2-AG release into the synaptic cleft.

4.3.4.2. Pain One of the earliest use of cannabis as pharmacological treatment was to treat pain and numerous studies demonstrated the effect of cannabinoids in pain relief in animal models (reviewed in (Walker & Huang, 2002)). Cannabinoid antinociception is mediated by effects on CNS, spinal cord and peripheral sensory nerves. Interestingly, pain induced by subcutaneous injections of the chemical irritant formalin in rats increases the release of AEA in the periaqueductal gray (PAG) (Walker et al, 1999). However, clinical trials conducted to evaluate efficiency of cannabinoids as painkillers showed they are not more effective than codeine and induced dose-dependent side effects, such as CNS depression, avoiding their use at high doses (Campbell et al, 2001). Recently, the Bayer laboratory developed and commercialized a new treatment, called Sativex, containing both Δ9-THC and cannabidiol that counteracts side effects of Δ9-THC. Sativex induces pain relief in patients suffering from multiple sclerosis or spinal cord injury and demonstrated efficiency in decreasing muscle spasms and spasticity (Wade et al, 2003). Thus, Sativex has been recently approved in France for the treatment of neuropathic pain associated with multiple sclerosis.

4.3.4.3. Neuroprotection Discovery of increased endocannabinoid levels after ischemic strokes (Berger et al, 2004; Schmid et al, 1995) or other types of brain injuries (Hansen et al, 2001; Sugiura et al, 2000) indicated that CB1R could be involved in neuroprotection. Later, several studies reported benefits of various cannabinoids after strokes in animal models. For example, Leker and colleagues showed that HU-210 reduced the size of infarct after focal ischemia induced by occlusion of the middle cerebral artery. Moreover, HU-210 improved motor disability and its action was reversed after SR141716A treatment, demonstrating the implication of CB1R in this process (Leker et al, 2003). Authors proposed that neuroprotection is due to cannabinoid- induced hypothermia. However, other studies reported also that cannabinoid-induced neuroprotection involves protection against excitotoxicity. Indeed, excitotoxicity is driven by high release of glutamate and, as shown above, cannabinoids decrease synaptic release of glutamate (Gerdeman et al, 2002; Robbe et al, 2002). Moreover, CB1R activation can

48 stimulate cytoprotective pathways, such as activation of nNOS (Oh et al, 2006), leading to neuroprotection. Numerous studies are currently conducted in order to evaluate the contribution of cannabinoids in neuroprotection because it could help finding new efficient treatments for strokes, and neurodegenerative diseases (Parkinsons disease, Alzheimers disease).

4.3.4.4. Neuronal development CB1R is present in embryonic brains, even before synapse formation. In mice, it is highly expressed in glutamatergic neurons from embryonic day 12.5 and is downregulated at approximately 5 days after birth (Vitalis et al, 2008). Thus, the role of CB1Rs during brain development was widely studied and showed important implications in migration and differentiation of neuronal progenitor cells, neuritogenesis, neurite outgrowth, axonal tracts structure and synapses establishment. During my PhD, we published a review on this subject (Gaffuri et al, 2012) (see Part 5). Importantly, studies have demonstrated various human disorders after maternal cannabis use during pregnancy, such as impulsive and hyperactive behaviours with increased inattention (Huizink & Mulder, 2006).

4.3.4.5. Appetite regulation In humans, it is well known that cannabis increases appetite. Accordingly, in rodents, Δ9-THC also induces hyperphagia (Williams et al, 1998), while the specific CB1R inverse agonist SR141716A decreases both appetite and body weight (Mackie, 2006). It was suggested that the effect is mediated by peripheral CB1R effects on lipolysis (Jbilo et al, 2005). Thus, SR141617A, also called Rimonabant, was commercialized by Sanofi-Aventis under the commercial name Accomplia in 2006 as an anti-obesity drug (Van Gaal et al, 2005). However, this promising drug induced serious psychiatric side-effects on patients, such as depression, suicide attempts and completed suicides and thus was withdrawn from the market in 2008 (Boekholdt & Peters, 2010). This shows the importance to increase our knowledge on the constitutive activity of neuronal CB1Rs.

4.3.5. Constitutive CB1R activity Studies on SR141716A demonstrated that this CB1R antagonist not only blocked agonist effects but also led to proper effects, contrary to agonist effects. For example, SR141716A has an effect on behavioural tests, such as arousal enhancement (Santucci et al, 1996) and also on synaptic transmission (Hoffman & Lupica, 2000). However, as endocannabinoids are lipidic

49 molecules present in membranes, it is difficult to determine whether constitutive CB1R activation is due to presence of endocannabinoids or to structural determinants (genuine constitutive activity). Here, I will present studies in favour of both theories.

4.3.5.1. Structural determinants enabling CB1R constitutive activity In 1998, Pan and colleagues demonstrated that SR141716A increases calcium current in superior cervical ganglion neurons. Moreover, they showed that CB1R constitutive activation is independent of the presence of 2-AG or AEA (Pan et al, 1998). This study strongly suggests that structural determinants are responsible for constitutive CB1R activity. Latter studies confirmed this hypothesis. For example, the truncation of CB1R distal C-terminal tail enhances its constitutive activity, while mutation of aspartate in the second transmembrane domain into aspargine (CB1R-D164N) decreases its constitutive activity (Nie & Lewis, 2001). Moreover, the group of Kendall investigated the contribution of threonine 210, an amino acid well conserved among CB1R family but absent in other GPCRs. T210 is proximal to the DRY motif, a highly conserved residue in GPCRs localized in the third intramembrane loop that is involved in GPCR activation. One point mutation of this residue by either alanine (CB1R-T210A) or isoleucine (CB1R-T210I) changes constitutive receptor activity. Indeed, compared to WT-CB1R (wild type CB1R), CB1R-T210A displays reduced constitutive activity, while CB1R-T210I displays increased constitutive activity (D'Antona et al, 2006). Interestingly, both agonists and inverse agonists are able to bind these mutants but binding affinities are modified. The hypoactive mutant CB1R-T210A displays reduced affinity for agonists and increased affinity for inverse agonists compared to WT-CB1R. On the contrary, CB1R-T210I displays an increased affinity for agonists and decreased affinity for agonists compared to WT-CB1R. However, this study was performed in HEK293 cells that are not endocannabinoid free. Thus, one can ask whether the decreased constitutive activity of CB1R- T210A is due to structural modifications preventing the receptor to adopt constitutively active states or to a diminished endocannabinoid binding. And, on the contrary, does CB1R-T210I display increased constitutive activity due to conformational changes or to an increased affinity for endocannabinoids? (see Part 4)

4.3.5.2. Constitutive endocannabinoid tone AEA and 2-AG, major endocannabinoids, are lipidic molecules and synthesized in lipid bilayer of cell plasma membrane (see Part 2 for detailed biochemistry). Thus, one can assume that endocannabinoids reach CB1Rs through two-dimensional diffusion in the lipid bilayer,

50 leading to cell-autonomous receptor activation. Indeed, the orientation of AEA in the bilayer was investigated by NMR (Nuclear Magnetic Resonance) and demonstrated that its fatty acid chain orients parallel to membrane acyl chain with terminal methyl close to the centre of the bilayer (Tian et al, 2005). Moreover, increasing evidence show that 2-AG enters the binding pocket between TM6 and TM7, through membrane bilayer (Hurst et al, 2013). With this unusual ligand binding, one can hypothesize that observed constitutive CB1R activity can be due to constitutive endocannabinoid tone in the plasma membranes. Indeed, blocking the synthesis of 2-AG, the major endocannabinoid in the mammal brain (Stella et al, 1997), abolishes (or strongly decreases) both constitutive CB1R signaling (Turu et al, 2007) and internalization (Leterrier et al, 2004; Leterrier et al, 2006) in heterologous systems.

Thus, due to endocannabinoid characteristics, it seems very difficult to study a system totally free of endocannabinoids. It could be possible to determine the contribution of structural determinants and endocannabinoid tone in CB1R constitutive activity by measuring the signaling induced by neutral antagonist binding (Figure 22). Neutral antagonist prevents agonist binding without blocking GPCR structural constitutive activity. Unfortunately, at present, such agonist has not been validated for CB1R.

Figure 22: How to distinguish constitutive activity from endogenous ligand binding? Neutral agonist does not affect constitutive activity but prevents from endogenous ligand binding. Thus, in the case of constitutive activity, treatment with neutral agonist leads to the same biological response as compared to vehicle. In presence of endogenous ligands, treatment with neutral agonist abolishes basal biological response. (Adapted from Leterrier, 2006, PhD manuscript).

Part 1: Influence of neuronal polarity on the traffic of type-1 cannabinoid receptor.

52 1. Introduction During his PhD in Zsolt Lenkeis team, Christophe Leterrier investigated the traffic of CB1R and notably the consequences of constitutive activity on CB1R traffic. First, transfected CB1Rs in HEK-293 cells are present at plasma membrane but also 85% of receptors are internalized. Interestingly, after inverse agonist AM281 treatment receptor internalization is highly reduced and an increase of receptors at the plasma membrane is observed. Thus, in this study, Christophe Leterrier demonstrated that origin of internalized receptors is a constitutive endocytosis, which is mediated by Rab5. Moreover, he showed that internalized receptors are recycled back to the plasma membrane through a Rab4-dependent mechanism, suggesting rapid recycling (Leterrier et al, 2004). Working with both overexpressed and endogenous CB1Rs in cultured hippocampal neurons, he demonstrated that, at basal state, somatodendritic CB1Rs are mostly present on endosomes while axonal CB1Rs are mostly localized at the plasma membrane. Interestingly, treatment with AM281 leads to an increase of membrane CB1Rs in the somatodendritic compartment, similarly to inhibition of endocytosis using either methyl-beta-cyclodextrin (MBCD) or dominant-negative proteins (Leterrier et al, 2006). Taken together, these results strongly suggest that CB1Rs are targeted to axonal plasma membrane through transcytosis. However, more experiments were required in order to confirm this mechanism. This work was performed by Anne Simon during her PhD in Zsolt Lenkeis team. Using microfluidic devices, she demonstrated that after their synthesis, CB1Rs are unspecifically targeted and can transiently appear on somatodendritic plasma membrane, from where they are rapidly internalized by constitutive endocytosis and targeted to the axonal plasma membrane, where they accumulate (Figure 23). Moreover, she demonstrated clearly that endocytosis is lower in axons compared to dendrites, even after CB1R pharmacological activation. As endocytosis seems important to obtain neuronal GPCR activation and as GPCR activation leads generally to their internalization, one can ask if chronic pharmacological activation or blockade of GPCRs can modify their polarized distribution. Working on both CB1Rs and somatodendritic somatostatin SST2 receptors, Anne Simon demonstrated that GPCR distribution is modified after chronic pharmacological treatments. During my master and my PhD, I took part to this study by performing some of the published experiments that enabled me to learn various experimental procedures, such as neuronal culture, manufacture of microfluidic devices and quantitative imaging. I also performed experiments asked by reviewers, notably the supplementary figure S4 where I demonstrated that internalized CB1Rs show very limited co-localization with lysosome

53 marker LAMP1 in somata indicating that somatodendritic internalized receptors are mostly recycled back to the somatodendritic membrane or targeted to the axonal plasma membrane and not degraded.

Figure 23: CB1R targeting After their synthesis in the soma, CB1Rs appear both on axonal (blue arrows) and somatodendritic (orange arrows) plasma membrane. However, on somatodendritic plasma membrane, CB1Rs are rapidely removed by constitutive endocytosis and recycled back to the axonal membrane through transcytosis. (Leterrier, 2006, PhD manuscript).

2. Results: Article 1 (published): Activation-dependent plasticity of polarized GPCR distribution on the neuronal surface. Journal of Molecular Cell Biology, 2013. Anne C. Simon, Claude Loverdo, Anne-Lise Gaffuri, Michel Urbanski, Delphine Ladarre, Damien Carrel, Isabelle Rivals, Christophe Leterrier, Olivier Benichou, Pascal Dournaud, Bela Szabo, Raphael Voituriez, and Zsolt Lenkei

55 Journal of Molecular Cell Biology Advance Access published May 24, 2013 doi:10.1093/jmcb/mjt014 Journal of Molecular Cell Biology (2013), Vol no. 0, 1–16 | 1 Article Activation-dependent plasticity of polarized GPCR distribution on the neuronal surface

Anne C. Simon1, Claude Loverdo2, Anne-Lise Gaffuri1, Michel Urbanski3, Delphine Ladarre1, Damien Carrel1, Isabelle Rivals4, Christophe Leterrier1,5, Olivier Benichou2, Pascal Dournaud6, Bela Szabo3, Raphael Voituriez2, and Zsolt Lenkei1,*

1 Laboratoire de Neurobiologie, CNRS UMR7637, ESPCI-ParisTech, 10 Rue Vauquelin, Paris 75005, France 2 Laboratoire de Physique Theórique de la Matie`re Condenseé – CNRS-UMR7600 – Universite´ Pierre et Marie Curie, 4 Place Jussieu, Paris 75005, France 3 Institute of Experimental and Clinical Pharmacology and Toxicology – Albert-Ludwigs University, Albertstrasse 25, Freiburg 79104, Germany 4 Equipe de Statistique Appliqueé, ESPCI-ParisTech, 10 Rue Vauquelin, Paris 75005, France 5 Present address: CRN2M, CNRS – Aix-Marseille Universite´ UMR 7286, Marseille F-13916, France 6 INSERM U676, Hoˆpital Robert-Debre´, 48 Boulevard Se´rurier, Paris 75019, France * Correspondence to: Zsolt Lenkei, Tel: +(33)140795184; E-mail: [email protected] Downloaded from

Directionality of information flow through neuronal networks is sustained at cellular level by polarized neurons. However, specific targeting or anchoring motifs responsible for polarized distribution on the neuronal surface have only been identified for a few neuronal G- protein-coupled receptors (GPCRs). Here, through mutational and pharmacological modifications of the conformational state of two

model GPCRs, the axonal CB1R cannabinoid and the somatodendritic SSTR2 somatostatin receptors, we show important conform- http://jmcb.oxfordjournals.org/ ation-dependent variations in polarized distribution. The underlying mechanisms include lower efﬁciency of conformation-dependent GPCR endocytosis in axons, compared with dendrites, particularly at moderate activation levels, as well as endocytosis-dependent transcytotic delivery of GPCRs from the somatodendritic domain to distal axonal portions, shown by using compartmentalized micro- ﬂuidic devices. Kinetic modeling predicted that GPCR distribution polarity is highly regulated by steady-state endocytosis, which is conformation dependent and is able to regulate the relative amount of GPCRs targeted to axons and that axonally polarized distribution is an intermediaryphenotype thatappears atmoderate basal activation levels.Indeed,weexperimentallyshowthatgradualchanges in basal activation-dependent endocytosis lead to highly correlated shifts of polarized GPCR distribution on the neuronal surface, which

can even result in a fully reversed polarized distribution of naturally somatodendritic or axonal GPCRs. In conclusion, polarized distri- by guest on May 29, 2013 bution of neuronal GPCRs may have a pharmacologically controllable component, which, in the absence of dominant targeting motifs, could even represent the principal regulatorof sub-neuronal distribution. Consequently, chronic modiﬁcations of basal GPCR activation by therapeutic or abused drugs may lead to previously unanticipated changes in brain function through perturbation of polarized GPCR distribution on the neuronal surface.

Keywords: targeting, transcytosis, endocytosis, constitutive, internalization, diffusion

Introduction have not yet been identified for the majority of neuronal GPCRs. Polarized distribution of G-protein-coupled receptors (GPCRs) to In addition, high-resolution electron microscopy has revealed one of the main neuronal sub-domains—soma, dendrites, or that many,if not most, GPCRs are localized outside of pre-and post- axons—has important functional consequences. Considerable synaptic specializations (Shigemoto et al., 1997; Charara et al., effort has been directed toward both classifying GPCRs as pre- or 1999; Fritschy et al., 1999; Meshul and McGinty, 2000; Riad postsynaptic and identifying protein motifs that assure the target- et al., 2000; Muly et al., 2003; Nyiri et al., 2005). Finally, there is ing and anchoring of receptors to presynaptic (axonal) and postsy- a remarkable and yet unexplained flexibility in distribution pheno- naptic (somatodendritic) specializations. However, decades of types, so GPCRs may show different polarized distribution patterns research have yielded the discovery of only a limited number of tar- depending on the neuronal cell type (Muly et al., 2003, 2010; geting motifs (reviews in Lasiecka and Winckler, 2011; Winckler Dumartin et al., 2007; Nathanson, 2008), or the neurochemical en- and Yap, 2011), so precise targeting and anchoring elements vironment (Dournaud et al., 1998; Decossas et al., 2003; Bernard et al., 2006). Thus, identification of additional mechanisms regulating sub-neuronal GPCR segregation remains an eminent research objective. Received October 11, 2012. Revised January 10, 2013. Accepted January 30, 2013. # The Author (2013). Published by Oxford University Press on behalf of Journal of Recent insights into the neuronal cell biology of polarized Molecular Cell Biology, IBCB, SIBS, CAS. All rights reserved. protein distribution suggest an indirect mechanism of axonal 2 | Journal of Molecular Cell Biology Simon et al. targeting, where after an initial indiscriminate delivery to the whole responsiveness to agonist and inverse agonist ligands (D’Antona neuronal surface, selective endocytosis eliminates proteins from et al., 2006). the somatodendritic but not axonal surface, followed either by deg- Neuronal expression of the three CB1R forms (wild-type or radation (Garrido et al., 2001) or by dendrite-to-axon transcytosis WT-CB1R, T210A-CB1R and T210I-CB1R), each labeled with the (Wisco et al., 2003). In both cases, the resulting steady-state distri- FLAG epitope at the extracellular N-terminus and GFP at the intra- bution is significantly polarized to the axonal surface. In some cellular C-terminus, yields highly different surface distribution pat- instances, the initial polarization of the somatodendritic mem- terns (Figure 1). Selective labeling of surface WT-CB1R shows brane might be reinforced by direct targeting (Yap et al., 2008), polarized distribution to axons (Figure 1A and B), as described pre- and specific axonal accumulation is also reinforced by specific re- viously (Coutts et al., 2001; Leterrier et al., 2006). The hyperactive tention/anchoring (Sampo et al., 2003; Fache et al., 2004; Xu T210I-CB1R is largely intracellular, but residual membrane labeling et al., 2006). In contrast to this default-mode endocytosis, is still polarized to axons. By contrast, the hypoactive T120A-CB1R agonist-induced endocytosis of GPCRs is a highly regulated shows mostly non-polarized distribution over the entire neuronal dynamic cellular process whose most prominent role is the degrad- surface, with clear somatodendritic polarization in a subset of ation or resensitization/recycling of GPCRs, which are typically neurons (Figure 1A). Quantitative evaluation of surface labeling in- desensitized following receptor activation. GPCRs are versatile sig- tensity shows that neurons expressing the three CB1R receptor naling molecules, displaying a flexible and dynamic three- forms at comparable levels are segregated in three clusters dimensional structure. Agonist binding modifies the energy land- (Figure 1B), which display significantly different mean CB1R ex- scape and leadsto the adoption of an active receptorconformation, pression levels on both the axonal and dendritic surface. In Downloaded from leading to activation of intracellular signaling pathways and, ultim- axons, the characteristic proximo-distal gradient of WT-CB1R ex- ately, to endocytosis through the classical clathrin-mediated endo- pression on the axonal surface (Coutts et al., 2001; Leterrier cytic pathway (Gainetdinov et al., 2004). Importantly, chronic et al., 2006) is maintained by the residual surface population of activation is often correlated with an elevated intracellular pool T210I-CB1R (Figure 1C). In striking contrast, T210A-CB1R expres- of GPCRs in non-polarized cells (Morisset et al., 2000; Whistler sion shows an inverted gradient, with a progressive diminution of http://jmcb.oxfordjournals.org/ et al., 2002; Leterrier et al., 2004; Marion et al., 2004; Morris surface labeling along the axon (Figure 1C) similar to the distribu- et al., 2004; Jacquier et al., 2006; Holliday et al., 2007; tion of the type-2 somatostatin receptor (SSTR2), a predominantly Mohammad et al., 2007; Chanrion et al., 2008). In neurons, acute somatodendritic GPCR (Csaba et al., 2007; Lelouvier et al., 2008). activation of GPCRs also results in endocytosis and recycling, Importantly, the remarkable phenotype shift of the T210A-CB1R both in vitro and in vivo (Riad et al., 2000; Bernard et al., 2006; mutant to a predominantly uniform/somatodendritic distribution Scherrer et al., 2006; Lelouvier et al., 2008; von Zastrow, 2010), was accompanied by the emergence of a novel somatodendritic but effects of chronic GPCR activation on sub-neuronal distribution CB1R functionality. The presynaptic CB1R inhibits neuronal Ca2+ have not been investigated yet. currents (Pan et al., 1996), but agonist activation of WT-CB1R We have previously shown that two axonal GPCRs, the CB1 can- (Supplementary Figure S1) and T210I-CB1R (not shown) did not by guest on May 29, 2013 nabinoid receptor and the 5-HT1B serotonin receptor, are also tar- inhibit increase of intracellular calcium in the soma or dendrites, geted initially to the somatodendritic plasma membrane and that in accordance with the lack of somatodendritic CB1R effects the subsequent establishment of correct polarized distribution reported previously using brain slices (Freiman et al., 2006). By relies on activation-dependent endocytosis (Leterrier et al., 2006; contrast, agonist-mediated activation of T210A-CB1R significantly Carrel et al., 2011). However, we do not know whether activated inhibited the entrance of calcium both in the soma and dendrites, GPCRs are preferentially endocytosed from the somatodendritic and this effect was inhibited by the inverse agonist Rimonabant plasma membrane compared with the axonal membrane and (Supplementary Figure S1). whether transcytotic axonal delivery, previously shown for the Taken together, T210-CB1R mutants show important alterations tyrosine-kinase receptor TrkA (Ascano et al., 2009), also plays a in the distribution of functionally competent CB1Rs on the neuronal role in GPCR targeting. More generally, currently we do not know surface. the extent of the activity-dependent plasticity of neuronal GPCR T210 mutants function as conformational CB1R isoforms distribution. In the present work, we developed quantitative in neurons methods to characterize and model the molecular dynamics of In the somatodendritic compartment, WT-CB1R displays a pre- GPCRs at single-neuron level in order to investigate the role of dominantly intracellular localization (Figure 2), T210I-CB1R is activity-dependent endocytosis in the regulation of sub-neuronal almost completely internalized, whereas T210A-CB1R displays a GPCR targeting. more prominent plasma membrane localization (Figure 2A), similar to the distribution of T210 mutants in non-polarized Results HEK-293 cells (D’Antona et al., 2006). Consequently, cells expres- Mutations of threonine 210 of CB1R lead to highly different sing the three receptor forms at comparable levels are segregated polarized distribution patterns into three clusters, which are significantly different in their surface

Point-mutations of the T210 residue, which is located in the 3rd labeling intensity (Figure 2B and B′). Remarkably, treatment of transmembrane helix, to alanine (T210A) or isoleucine (T210I) T210I-CB1R expressing neurons with the inverse agonist AM281 result in changes in the conformational state of CB1Rs, yielding or treatment of T210A-CB1R expressing neurons with the agonist hypo- and hyperactive receptor forms, respectively, which display WIN22,512-2, respectively, resulted in a signiﬁcant redistribution signiﬁcantly different levels of constitutive activity but preserve toward the WT phenotype (Figure 2A, C, and C′), suggesting that Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 3 Downloaded from http://jmcb.oxfordjournals.org/ by guest on May 29, 2013

Figure 1 Mutations of T210 transform both axonal/dendritic polarity and the proximo-distal axonal gradient of CB1R distribution on the plasma membrane. DIV8 cultured neurons were transfected with three forms of Flag-CB1R-eGFP: WT-CB1R, T210I-CB1R, and T210A-CB1R (GFP—green) and labeled for the surface population (Alexa568—red). (A) WT-CB1R displays highly polarized distribution on the axonal surface. T210I-CB1R is mostly intraneuronal but presents a residual membrane labeling to axons, whereas the hypoactive T120A-CB1R shows mostly non-polarized distribution over the entire neuronal surface. Arrowheads: axon; arrows: dendrite. Scale bar: 50 mm. (B) Surface labeling intensity of individual neurons (crosses) for the three CB1R forms in axons and dendrites. Square, circle, and diamond represent the mean fluorescence intensity and dotted line indicates a hypothetical equal rate of surface labeling intensity in axons and dendrites (i.e. non-polarized distribution). Data shown are mean + SEM, 32–78 neurons per condition, **P , 0.01, ***P , 0.001, two-way ANOVA, Bonferroni post hoc test. (C) The gradient of surface labeling intensity along the axon from proximal to distal. Inbox: gradient normalized to the intensity of the first pixel. WT-CB1R and T210I-CB1R present a typical augmentation of labeling intensity from proximal to distal, whereas the hypoactive mutant T210A-CB1R and the non- active somatostatin SSTR2 receptor display an inversed gradient. The gray area corresponds to the initial segment of the axon. Data shown are mean + SEM, 6–12 neurons per condition. 4 | Journal of Molecular Cell Biology Simon et al. Downloaded from http://jmcb.oxfordjournals.org/ by guest on May 29, 2013 Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 5 distribution of CB1R within the somatodendritic region is well cor- inverse agonist ligands, while treatment of T210I-CB1R expressing related with receptor conformation and that T210-CB1R mutants neurons with an inverse agonist or treatment of T210A-CB1R are true conformational CB1R isoforms, as proposed previously expressing neurons with an agonist returned these values toward (D’Antona et al., 2006). WT levels (Figure 3B–D). Together, mutational and pharmacologic- In conclusion, changes in signaling and in steady-state distribu- al modulation of CB1R yielded a large spectrum of activation levels, tion phenotypes confirm the notion that mutations of threonine resulting in considerable variation of both absolute and relative 210 to isoleucine (T210I) or alanine (T210A) result in functional (dendrite vs. axon) endocytosis levels. Notably, the abundance GPCR isoforms that closely approximate the active and inactive and labeling intensityof endosomes wasalwayshigher in dendrites conformational states of CB1R, respectively. By permanently than in axons. This difference was most pronounced at moderate modulating the conformational state of CB1R, mutations of T210 activation levels and was mainly due to the variability in endosome provide valuable tools to precisely investigate the relation abundance (Figure 3B) and, to a lesser extent, of labeling intensity between GPCR activation and sub-neuronal distribution, without of individual endosomes (Figure 3C). Finally, CB1R activation- putative non-desirable effects often associated with long-term dependent endocytosis is qualitatively similar between axons pharmacological treatment. and dendrites, with comparable response amplitude and half-

Activation-dependent endocytosis is highly different between maximal effective activation (EC50(dendrite) ¼ 112.6% + 2.3% and axons and dendrites EC50(axon) ¼ 117% + 2.6%, normalized to basal WT-CB1R activity The above-reported differential basal distribution of CB1R iso- in HEK-293 cells), showing a constantly negative offset of axonal forms may reflect differences in steady-state endocytosis that, in endocytosis efficiency at all activation levels (Figure 3E). In order Downloaded from turn, may depend on steady-state activation. Indeed, inverse to investigate whether this negative offset is due to anchoring to agonist treatment significantly inhibits spontaneous somatoden- scaffolding proteins in the axonal plasma membrane, we analyzed dritic endocytosis of endogenous CB1R in mature cultured hippo- the diffusion of WT-CB1R and T210A-CB1R by fluorescence recov- campal neurons (Supplementary Figure S2). In order to directly ery after photobleaching (FRAP) and compared it with the diffusion investigate the relationship between the conformational state of of the somatodendritic SSTR2 receptor, which is not expected to be http://jmcb.oxfordjournals.org/ CB1R and the relative level of endocytosis in different neuronal specifically anchored in axons. We observed that the diffusion co- sub-domains, we measured sub-domain specific endocytosis of efficient and the fraction of mobile receptors are similar for the the three CB1R isoforms (Supplementary Figure S2C; see three receptors (Supplementary Figure S3). In addition, diffusion Supplementary material for experimental details). In neurons of all receptors was rather rapid, close to the values reported for expressing WT-CB1R, the number and intensity of labeled endo- the GPCR-like bacteriorhodopsin in reconstituted membranes somes containing recently endocytosed receptors were higher in (Cherry et al., 1982). These results show that CB1Rs are dendrites than in axons (Figure 3A), by almost 400 and 150%, re- not anchored to scaffolding proteins in the axonal plasma mem- spectively (Figure 3B). The cumulative effect of these two para- brane and suggest that axonal accumulation of CB1R does not meters led to a more than 4-fold difference of total internalized rely on specific retention, but rather on reduced endocytic elimin- by guest on May 29, 2013 receptors in dendrites compared with axons. In T210I-CB1R expres- ation. In conclusion, our results show significant differences in sing neurons, both parameters were significantly elevated, more conformational-state-dependent endocytosis between axons and prominently in axons than in dendrites, leading to a diminished dif- dendrites. ference between the total amount of internalized receptors in the Transcytotic delivery of CB1R to the axonal surface two main neuronal sub-domains (Figure 3A–D). Conversely, in Our results show important sub-neuronal differences of neurons expressing the hypoactive T210A-CB1R, both the activation-dependent endocytosis. Polarized distribution could number and the intensity of labeled endosomes diminished, be reinforced if, after elimination from the somatodendritic leading to a significantly lower amount of total internalized recep- plasma membrane, CB1Rs were recycled to the axonal plasma tors both in axons and dendrites (Figure 3A–D). Similar shifts in membrane, as reported for various axonal proteins (Lasiecka and endosome number and labeling intensity were produced after Winckler, 2011). However, to our knowledge, such transcytotic treatment of WT-CB1R expressing neurons with agonist or transport of endocytosed GPCRs has not been reported previously.

Figure 2 Mutations of T210 yield conformational CB1R isoforms with different intracellular distribution in the somatodendritic domain. (A) WT-CB1R displays a predominantly intracellular localization with a low level of labeling at the plasma membrane. T210I-CB1R presents an almost completely internalized distribution, whereas T210A-CB1R displays more prominent plasma membrane localization. Neurons expressing T210I-CB1R and T210A-CB1R and treated for 3 h with the inverse agonist AM281 (AM 1 mM) or agonist WIN55,212-2 (WIN 1 mM), respectively, display a distribution phenotype similar to WT-CB1R. Scale bar: 25 mm. (B) Plotting of individual WT-CB1R (black squares), T210I-CB1R (red circle), and T210A-CB1R (blue triangles) expressing cells results in 3 clusters, showing comparable levels of receptor expression but significantly different in surface labeling intensity (P , 0.001, two-way ANOVA). Solid and dotted lines represent linear regression and 95% confidence intervals foreachcluster, respectively. Inset B’: normalized surface/total fluorescence ratio (mean + SEM, 57–68 neurons percondition, ***P , 0.001 vs. WT-CB1R, one-way ANOVA). (C) Pharmacological rescue of somatodendritic T210 phenotypes. 3 h of treatment of T210I-CB1 and T210A-CB1 expressing neurons with 1 mM of inverse agonist AM281 (T210I-CB1R—orange circles) or agonist WIN55,212-2 (T210A-CB1R—light blue triangles), respectively, restore the somatodendritic WT-CB1R (black squares) phenotype, as shown by comparable levels of surface labeling. Solid and dotted lines represent linear regression curves and 95% confidence intervals for each cluster, respectively. Inset C′: normalized surface/ total fluorescence ratio (mean + SEM, 18–41 neurons per condition, ns for non-significant, one-way ANOVA). 6 | Journal of Molecular Cell Biology Simon et al. Downloaded from http://jmcb.oxfordjournals.org/ by guest on May 29, 2013

Figure 3 Activation-dependent endocytosis is different between axons and dendrites. (A) ‘Feeding’ live neurons for 1 h with an antibody directed against the extracellular FLAG epitope, followed by an acid wash to eliminate surface-bound antibody, labels endosomes incorporating recently endocytosed receptors. WT-CB1R expressing neurons display an elevated number of brightly labeled endosomes in soma and dendrites, whereas axons show few and weakly labeled endosomes. T210I-CB1R neurons contain numerous brightly labeled endosomes both in axons and dendrites, whereas T210A-CB1R neurons present only few weakly labeled endosomes. Arrowheads indicate axons, arrows label dendrites. Scale bar: 50 mm. (B–D) Endosomes per neurite length (B) and mean labeling intensity (C) of individual endosomes in axons versus dendrites. Multiplying endosomes per neurite length with mean endosome intensity yields total internalized receptors per neurite length (D). Dotted lines indicate different hypothetical ratios in axons and dendrites (mean + SEM). (E) Correlation between endocytosis rate and relative activity of the different receptor forms, measured by the inhibition of forskolin-induced cAMP mobilization in HEK-293 cells. Activity and endocytosis display a highly signiﬁcant correlation, well-ﬁtted with a sigmoid curve (r2 ¼ 0.98 both for dendrites and axons) and indicate a constant offset between dendritic and axonal endocytic response to activation (mean + SEM).

Importantly, axons display anterogrademoving endosomes incorp- pathway. In order to directly show that CB1Rs, endocytosed from orating recently endocytosed WT-CB1R (Figure 4A and B), as the somatodendritic surface, are inserted to the axonal plasma expected for GPCRs recruited in the transcytotic axonal targeting membrane after transcytotic delivery, we incubated specifically Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 7 the somatodendritic compartment of living FLAG-WT-CB1R expres- GPCR targeting. Our model (Figure 5A) uses a limited number of sing embryonic cortical neurons in a microfluidic device, which parameters, measured, calculated, or fixed (see Materials and allows separation of somatodendritic and axonal domains (Taylor methods for details). In brief, receptors are transported inside et al., 2005), with anti-FLAG antibodies. After allowing endocytic the axon starting from the soma using one of the two transport uptake and transcytotic transport for 4 h at 378C, the presence of routes. For simplicity, we assume that the transport velocity for transcytosed anti-FLAG antibodies was detected on the surface of each route is the same, but the rate at which receptors are released 69 2 1 4 ¼ 13 1 a . % of CB R-expressing axons (Figure C and D, n , from to the membrane is different. Theflux ofreceptorsusingroute (n1 ) two independent experiments). This labeling showed a clear disto- only depends on the synthesis in the Golgi apparatus and is consid- proximal distribution gradient (Figure 4E), and was highly dimin- ered constant at the time scale of the experiment. In contrast, the ished after Dynasore-inhibited (80 mM) somatodendritic endocyto- flux of receptors using route 2 is proportional to the rate of soma- 11 ¼ 9 4 s sis ( %, n ) (Figure C). These results unambiguously show todendritic endocytosis (ki ). The fate of endocytosed receptors is 1 a that CB R is transcytosed to the distal axonal surface following either local recycling (kr) or transcytotic delivery (n2), because, somatodendritic endocytosis. for simplicity, receptor degradation was considered negligible. A kinetic model of neuronal GPCR targeting Indeed, previous in vivo ultrastructural analysis localized only a mi- The sumof the above-reported results indicatesthat somatoden- nority of intracellular CB1Rs to lysosomes in hippocampal neurons dritic endocytosis and subsequent transcytotic delivery are puta- (Thibault et al., 2012), and wedid not find notable co-localization of tive regulators of axonal targeting, as suggested previously for intracellular CB1R mutants with the lysosomal marker LAMP1 the WT-CB1R(Leterrier et al., 2006). In axons, long-range antero- (Supplementary Figure S4). Finally, receptors can freely diffuse Downloaded from grade transport preferentially delivers cargo to distal axonal on both the somatodendritic and axonal plasma membranes, or regions such as growth cones (Craig et al., 1995) or synaptic term- they may be ‘reflected’ from the tip of the growth cone and are par- inals (Ahmari et al., 2000), and GPCRs containing specific targeting tially ‘reflected’ from the diffusion barrier at the axonal initial domains, such as Sushi domain bearing GABAB receptor subtypes, segment (Nakada et al., 2003). were also reported to arrive directly to distal axonal portions When the resulting equations (see Supplementary material) are http://jmcb.oxfordjournals.org/ without previous insertion into the plasma membrane (Biermann solved at steady state, they yield explicit expression for the number 2010 2012 s et al., ; Valdes et al., ). However, in order to be able to of receptors on the somatodendritic plasma membrane (nm), the a account for the above-reported different distribution gradients of number of receptors on the axonal plasma membrane (nm), and 1 1 s a a CB R mutants on the axonal plasma membrane (Figure ), we the total numberof internalized receptors (ns + n1 + n2), as a func- s propose that GPCRs without specific axonal targeting signals tion of the steady-state level of somatodendritic internalization (ki ) reach the axonal plasma membrane through two parallel routes. (Figure 5A). This allows plotting the theoretical polarization of a s The first one corresponds to a non-specific post-Golgi pathway GPCR distribution on the neuronal surface (nm over nm) as a func- 1 s (route ), in which transport vesicles deliver GPCRs mainly to the tion of ki . The resulting theoretical curve is a hyperbole somatodendritic plasma membrane but are not actively excluded (Figure 5B). Interestingly, the plot of the polarized distribution of by guest on May 29, 2013 from the axon. However, these vesicles, which are not specialized the three CB1R isoforms against their measured somatodendritic for axonal delivery, do not progress very efficiently in axons and are internalization rate is also well fitted (r2 ¼ 0.77) with a hyperbole likely to fuse with the plasma membrane of the proximal axon (Figure 5C). Superposition of the experimental fit with the theoret- before advancing significantly. Receptors delivered through this ical curve allows translation of the dimensionless, theoretical s route thus are expected to show a proximo-distal distribution gra- steady-state somatodendritic internalization rate (ki ) into experi- dient on the axonal plasma membrane, and this route is not likelyto mentally measured rates of the somatodendritic endocytosis lead to efficient GPCR delivery to distal axonal portions. (Figure 5C). This finally allows plotting the effect of somatodendri- Conversely, route 2 originates from somatodendritic recycling tic endocytosis rate on the distribution of CB1R into the three major endosomes and delivers GPCRs, in vesicules specialized for long- sub-neuronal populations: CB1R on the axonal plasma membrane, range axonal delivery, directly to distal portions of axons. CB1R on the somatodendritic plasma membrane, and internalized Receptors delivered through this transcytotic route thus show a CB1R (Figure 5D). Strikingly, intermediate levels of steady-state disto-proximal distribution gradient on the axonal plasma mem- somatodendritic internalization, which correspond approximately brane. The relative proportion of receptors arriving through these to the measured internalization rate of WT-CB1R, result in a major- two pathways depends on the amount of receptors entering soma- ity of CB1R localized on the axonal plasma membrane (Figure 5D). todendritic recycling endosomes; it is thus proportional to somato- Lowering levels of somatodendritic internalization will first result in dendritic endocytosis. Since somatodendritic endocytosis is a balance (uniform distribution) between somatodendritic and regulated by the activation state of GPCRs, changes in receptor ac- axonal CB1R approximately at the measured rate of T210A-CB1R tivation state shift the relative amount of receptors addressed to endocytosis, and then further decrease results in a predominantly the axonal membrane through these two routes, ultimately somatodendritic localization. By contrast, increase in somatoden- leading to changes both in polarized distribution and in the dritic internalization will result in a steady decrease of CB1R axonal distribution gradient. Thus, distribution of GPCRs on the levels both on the somatodendritic and on the axonal plasma mem- neuronal plasma membrane is the result of a dynamic balance branes accompanied by an increase of internalized CB1R, which and shows high plasticity, correlated with changes in GPCR activa- population becomes dominant approximately at the measured tion state. endocytosis rate of T210I-CB1R. In another set of calculations we Using this theoretical framework, we analyzed the kinetics of investigated the influence of the somatodendritic internalization 8 | Journal of Molecular Cell Biology Simon et al. Downloaded from http://jmcb.oxfordjournals.org/ by guest on May 29, 2013

Figure 4 Transcytotic CB1R delivery to the distal axonal surface. (A) Anterograde transport of endosomes containing recently internalized CB1R in axons. Flag-WT-CB1R transfected neurons were incubated for 5 min with fluorescent anti-FLAG antibody. After 15 min, anterogradely moving labeled endosomes are present in the axon (white arrowheads), identified as an MAP2-immunonegative neurite by post hoc immunodetection. Scale bar: 50 mm. (B) Anterograde progression of a labeled endosome in the axon (labeled with red arrowheads on A) Scale bar: 50 mm. (C) Red-labeled puncta (white arrowheads) indicate transcytosed anti-FLAG antibodies on 69.2% of living, CB1R-expressing (green) axons (n ¼ 13), grown in a compartmentalized microfluidic device, as shown below. This labeling is either absent in control experiments with omitted Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 9 rate on the axonal distribution gradient. At low levels of somato- of the agonist WIN22,512-2 or the inverse agonist AM281 for 24 h s dendritic internalization (ki ), the theoretical curve shows a following transfection, the time scale required to establish polar- proximo-distal gradient corresponding to a gradual diminution of ized axonal distribution in cultured hippocampal neurons CB1R expression toward the end of the axon. The gradient is (Leterrier et al., 2006). Neurons were then classified based on s 1 5 reversed at higher levels of somatodendritic endocytosis (high ki surface CB R distribution (Supplementary Figure S ), as described values), leading to a higher level of CB1R expression in distal previously (Carrel et al., 2011) (see also Supplementary material). parts of the axon. After adjusting a limited number of parameters This experimental approach is able to detect population-wide (see Materials and methods), we were able to fit the theoretical changes in polarized CB1R distribution after modification of the curves to experimental values, measured for the three CB1R spontaneous endocytosis rate, as demonstrated by expressing a forms in neurons selected for having an axon of an approximate dominant negative (S34N) mutant of the small GTPase Rab5 length of 600 mm, in order to standardize experimental values. (Supplementary Figure S6), which significantly inhibits polarized Our results qualitatively show that the observed concentration gra- CB1R distribution in individual neurons (Leterrier et al., 2006). dients of receptors are compatible with our kinetic model Neurons expressing WT-CB1R, treated either with vehicle (data (Figure 5E). The distribution of the somatodendritic SSTR2, not shown) or with very low concentrations of agonist or inverse assumed to be transported only via route 1, can also be fitted agonist ligands (1 × 10 – 11 M), mainly displayed the Axonal pheno- (Figure 5E). It should be noted that the concentration of CB1R type (55.89% + 3.27%, Figure 6A), while T210I-CB1R expressing and SSTR2 on the first 50–100 mm of the axons is not well captured neurons belonged mainly to the Internalized phenotype by the model. Notably, 50–70 mm is the typical scale of the axon (Figure 6B) and those expressing T210A-CB1R displayed a predom- Downloaded from initial segment (Nakada et al., 2003), which corresponds to a inantly Uniform-Somatodendritic or Axonal/Uniform distribution region with highly reduced diffusion of lipids and membrane pro- (Figure 6C), similar to non-treated neurons shown in Figure 1A. teins but in our present model we did not attempt to model GPCR Interestingly, in addition to these predominant distribution pheno- dynamics inside of this special axonal sub-segment. types, each phenotypic category was represented for every recep- In conclusion, we are able to provide a relatively simple mathem- tor form, indicating an important natural variability of the polarized http://jmcb.oxfordjournals.org/ atical model to describe the relationship between GPCR activation CB1R distribution pattern in cultured hippocampal neurons. and polarized distribution between the somatodendritic and Increasing concentrations of agonist or inverse agonist gradually axonal plasma membranes as well as the relationship between transformed the WT-CB1R distribution pattern into T210I-CB1R GPCR activation and the axonal distribution gradients. This and T210A-CB1R phenotypes, respectively. Inverse agonist treat- model suggests that for GPCR lacking dominant targeting and/or ment of T210I-CB1R expressing neurons or agonist treatment of anchoring motifs, pharmacological activation state is an important T210A-CB1R expressing neurons led also to gradual transforma- determinant of the neuronal distribution phenotype. Interestingly, tions of distribution patterns, allowing an almost complete our model also provides two non-trivial predictions. First, by rescue toward the wild-type phenotype at 1 × 10 – 5 M of AM281 showing that polarized GPCR distribution forms a continuous spec- for T210I-CB1R and 1 × 10– 8 M of WIN22,512-2 for T210A-CB1R. by guest on May 29, 2013 trum of phenotypes from entirely somatodendritic through axonal Principal component analysis confirmed that pharmacological or to fully internalized, our model predicts an important activation- mutational modifications of CB1R activation result in highly over- dependent plasticity of distribution phenotypes, where gradual lapping distribution patterns (Figure 6D). However, there is a changes in steady-state receptor activation are translated into notable divergence between the distribution pattern of gradual changes of polarized plasma membrane distribution. In T210A-CB1R and inverse agonist treated WT-CB1R (compare other words, it should be possible to obtain similar distribution black and blue lines at the left-hand side of Figure 6D). phenotypes for different neuronal GPCRs, such as the three CB1R Interestingly, when newly synthesized WT-CB1Rs are exported to isoforms, through carefully adjusted chronic pharmacological the plasma membrane in a synchronized manner after Brefeldin A treatment. Second, the model predicts the possibility to transform (BFA) treatment (Misumi et al., 1986) in the presence of 1– a naturally somatodendritic GPCR into an axonally polarized distri- 10 mM inverse agonist AM281, the proportion of neurons bution through prolonged, moderateactivation. Weexperimentally showing uniform/somatodendritic distribution is 60%–80% after verified these predictions below. 24 h (Leterrier et al., 2006; and data not shown), which overlaps Polarized CB1R distribution patterns show important wellwith the T210A phenotype. The reason for this experimental di- activation-dependent plasticity vergence is currently not known. In order to verify the first theoretical prediction of the above Taken together, these results confirm the prediction of our model, we induced persistent and gradual changes in the confor- model by showing highly elevated activation-dependent plasticity mations of the three CB1R forms by using different concentrations of sub-neuronal distribution. This validates the notion that anti-FLAG antibody (0% labeling, n ¼ 6) or highly diminished after Dynasore-inhibited (80 mM) somatodendritic endocytosis (11%, n ¼ 9). Scale bars: 10 mm. (D) Experimental setup used to directly demonstrate transcytotic delivery of CB1R to the axonal surface. Embryonic cortical neurons expressing FLAG-CB1R-eGFP are cultivated in a two-compartment microfluidic device, which allows separate incubation of the somatodendritic and axonal domains. The somatodendritic compartment is first incubated with mouse monoclonal anti-FLAG antibodies for 4 h at 378C, and then the axonal compartment is incubated with fluorescently labeled anti-mouse antibody. (E) Proximo-distal distribution profile of transcytosed CB1Rs, established by counting the number of Alexa-Fluor 568 labeled puncta (white arrowheads on C) on 25 mm long axon segments, starting from the distal axon tip. Data were obtained from 13 neurons (mean + S.E.M, *P , 0.05, **P , 0.01 vs. 0–25 mm, Student’s t-test). 10 | Journal of Molecular Cell Biology Simon et al. Downloaded from http://jmcb.oxfordjournals.org/ by guest on May 29, 2013

Figure 5 A kinetic model of neuronal GPCR targeting. (A) A simple model derived from a Fokker–Planckequation, relates dendritic endocytosis rate to polarized distribution patterns by using a limited number of measured, calculated, or ﬁxed parameters to describe sub-neuronal GPCR distribution. Distribution gradients of receptors arriving atthe axonal surface through eachroute arerepresented in pink. The redcircle indicatesthe key s control parameter, the steady-state somatodendritic internalization rate (ki ).(B) The theoretical curve of polarized distribution of GPCRs on the a s s neuronal surface (nm over nm) as a function of steady-state somatodendritic internalization rate (ki ). Fitting parameters are s s a s kl = 0.65, kr = 6ki , km = 3.5, ki = 0.45 + 0.55ki .(C) Plot of the measured level of surface polarization (A/D) as a function of experimentally measured rates of somatodendritic endocytosis (black curve) with the superposed theoretical curve (red). The dotted line shows the calculation of the corrected experimental somatodendritic endocytosis rate for WT-CB1R. (D) Fitting curves of CB1R distribution on the neuronal surface as a function of the experimentally measured somatodendritic endocytosis rate. Sub-neuronal distribution displays a predominantly axonal polarization for intermediate endocytosis rates. Measured somatodendritic endocytosis rates of the three CB1R isoforms are presented as mean (solid lines) and SEM (dotted lines), respectively. (E) Measured (black) and theoretical (red) axonal distribution gradients of different receptors. The fol- WT ¼ 500 WT ¼ 225 WT ¼ 6 SST lowing were the ﬁt parameters lm = 2000 mm; lc mm; ( je/j1) ;( jout/j1) ; lm = 500 mm (see Supplementary material for equation details). Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 11

pharmacologically reversible conformational changes critically regulate CB1R distribution on the neuronal surface. Prolonged moderate activation leads to axonal redistribution of SSTR2s In order to verify the second theoretical prediction, we used SSTR2 for three reasons. First, SSTR2 showslittle constitutiveactiv- ity and is mostly localized to the plasma membrane in heterologous expression systems and in neurons. Second, SSTR2 shows a predominantly somatodendritic distribution in adult neurons, but do not appear to be actively excluded from axons as shown by the axonal localization of SSTR2 in certain developmental stages. Third, SSTR2s are efficiently internalized and recycled in neurons following activation by the agonist octreotide (Sarret et al., 1998; Csaba et al., 2007; Lelouvier et al., 2008; Le Verche et al., 2009). The previously characterized FLAG-SSTR2-GFP construct (Lelouvier et al., 2008) was exported to the plasma membrane in a synchronized manner, by using BFA washout, at different concentrations of octreotide. We used the BFA release protocol in order Downloaded from to shorten the incubation period to 12 h, because the peptide ligand octreotide may degrade during longer incubation periods at 378C. As expected, surface SSTR2s displayed a predominantly somatodendritic localization (Figure 7A) and high concentrations of octreotide (1 × 10– 5 M) resulted in a mostly internalized pheno- http://jmcb.oxfordjournals.org/ type (Figure 7C). In contrast, intermediate levels of octreotide treatment resulted in the striking appearance of neurons showing axonally polarized distribution of SSTR2s (Figure 7B and C). Interestingly, the concentration of octreotide needed for the appearance of neurons showing axonally localized SSTR2 (1 × 10– 8 M) corresponds approximately to the octreotide concentration range where SSTR2 internalization reaches steady-state CB1R internalization levels in HEK-293 cells (Figure 7D). In conclusion, pharmacological adjustment of the steady-state by guest on May 29, 2013 internalization rate of somatostatin SSTR2 receptors to the level of WT-CB1 partially transforms the natural somatodendritic phenotype of SSTR2 into an axonal phenotype, as predicted by our model. This finding further confirms the notion that pharmacological activation state of GPCRs may be an important determinant of sub-neuronal distribution. Figure 6 Polarized CB1R distribution patterns show important activation-dependent plasticity. Sub-neuronal distribution of the Discussion three CB1R forms in the presence of various concentrations of By applying quantitative measures of receptor endocytosis and agonist and inverse agonist ligands, classified into five phenotypic cat- subcellular localization in neurons followed by mathematical mod- egories (see Supplementary material). (A) WT-CB1R shows dominantly eling, we used an integrative approach to investigate the relation- axonal polarization in control conditions. Agonist activation of ship between the activation state and subcellular distribution of a WT-CB1R leads to a fully internalized distribution pattern, whereas model GPCR, the type-1 cannabinoid receptor. Our kinetic model inverse agonist treatment leads to a higher proportion of neurons dis- ( s) playing uniform surface labeling. (B)T210I-CB1R displays almost com- reveals the role of the somatodendritic endocytosis rate ki , pletely internalized distribution and could be rescued to WT-CB1R which is directly proportional to GPCR activation, to establish phenotypes with the appropriate concentration of inverse agonist. (C) experimentally observed distribution phenotypes. Experimental T210A-CB1R presents a mostly non-polarized distribution which can verification of two non-trivial predictions of the model indicates be restored to a mostly axonal phenotype with 10–100 nM of agonist. that distribution of functional GPCRs on the neuronal plasma mem- Bars show mean+ SEM, with total cell count at the top. (D) PCA plot brane shows previously unappreciated plasticity. The key deter- of the distribution of the three receptors for each treatment condition. minant of the normal targeting of CB1R to distal axonal portions Tendency curves show the effects of different concentrations of is the transcytotic targeting pathway. In order to enter this specific ligands on the distribution of each receptor form. Mutational or pharma- transport pathway, receptors must first be retrieved from the soma- cological modification of receptor activation levels leads to highly over- todendritic plasma membrane by activation-dependent endocyto- lapping distribution patterns, with the notable exception of inverse sis. Once delivered to the axonal plasma membrane, the loss due to agonist treated WT-CB1R, which presents a mixed phenotype. spontaneous endocytosis is limited because of a difference 12 | Journal of Molecular Cell Biology Simon et al. Downloaded from http://jmcb.oxfordjournals.org/

Figure 7 Prolonged moderate activation leads to axonal redistribution of SSTR2.(A and B) DIV8 neurons were transfected with Flag-SSTR2-eGFP, and protein export from the Golgi apparatus was blocked by treatment with BFA for 12 h. BFA washout was used to induce synchronized release of newly synthesized receptors in the presence of various concentrations of the SSTR2 agonist octreotide. After 12 h of treatment, neurons were labeled for surface SSTR2 and classified into five phenotypic categories (see Supplementary material). In control conditions and at low ligand con- by guest on May 29, 2013 centrations, neurons display typical accumulation of surface SSTR2 in the somatodendritic region (arrow) but not on the axonal surface (arrow- head). (C) Distribution patterns of sub-neuronal SSTR2 localization in the presence of various concentration of octreotide. Chronic treatment with 50 nM octreotide induces the appearance of neurons showing SSTR2 distribution polarized to the axonal surface. Total cell number is shown at the top of each bar. Each class is presented by mean + SEM. (D) Endocytosis rate of SSTR2 in HEK-293 cells measured by flow-cytometry. Endocytosis of SSTR2 is concentration dependent and reaches the endocytosis rate of WT-CB1R (horizontal gray line) at an approximate concentration of 1 × 10 – 8 M (mean + SEM). between dendritic and axonal endocytosis rates, which is morepro- axonal plasma membrane resulting in a predominantly internalized nounced at moderate activation levels. phenotype, as displayed by the hyper-active T210I-CB1R mutant or According to our model, at low activation levels receptors remain WT-CB1R and SSTR2 following treatment with high agonist concen- at the site of the initial indiscriminate delivery through route 1 on trations. the somatodendritic and axonal plasma membrane, resulting in Our model accounts remarkably well for the observed experi- predominantly somatodendritic or uniform distribution. In the mental phenotypes but has several limitations. First, this mechan- axon, the lack of efficient anterograde transport results in a ism is likely important only for the distribution of GPCRs which gradual diminution of receptor density toward more distal seg- lack dominant targeting and/or anchoring motifs, provided that ments, which receptorsreach through diffusion in the plasmamem- these receptors efficiently enter the recycling/transcytotic brane. The hypo-active T210A-CB1R mutant and SSTR2 show this pathway following activation. Currently, we do not know the pro- pattern of distribution. Gradual increase of GPCR activation leads portion of directly targeted and/or anchored receptors among to a gradual decrease of receptors on the somatodendritic mem- the total population of neuronal GPCRs, but methods similar to brane, through endocytic elimination. A high proportion of endocy- the FRAP approach used in our study may provide in the future tosed receptors enter specific anterograde transport vesicles and a relatively simple approach to estimate whether interactions are efficiently transported to distal axons. Consequently, receptors with multi-protein complexes may significantly influence steady- will be shifted from the somatodendritic membrane toward distal state localization of a given GPCR. Second, putative GPCR- axonal segments, as shown for the wild-type CB1R or a significant activation-dependent regulation of the recycling, the transcytotic proportion of agonist-activated SSTR2. Finally, further increase of and the degradative pathways, relatively poorly known in receptor activation leads to endocytic depletion also from the neurons, is not included. However, the fit of the theoretical and Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 13 experimental curves suggests that such putative differences in inverse agonist treatment (Figure 6D), which we report here but degradation and in the efficiency of recycling and transcytosis was absent in T210A-C1BR expressing neurons, similar to our pre- are not likely to represent important factors, which would be neces- vious report by a more rigorous BFA release protocol before inverse sary to explain the significant qualitative differences in sub- agonist treatment (Leterrier et al., 2006). Notably, by using two neuronal distribution of the three CB1R forms. We also directly closely related but differentially distributed serotonin receptors, demonstrate that all the three CB1R forms show similar and rela- we recently demonstrated also higher constitutive activation and tively low lysosomal localization. Third, we obtained the experi- spontaneous endocytosis for the axonal 5-HT1BR, compared with mental power necessary to acquire high-quality quantitative the somatodendritic 5-HT1AR, both in non-neuronal cells and in data, a prerequisite for modeling, by the use of highly resolved in neurons (Carrel et al., 2011). Activation-dependent constitutive vitro methods, including transfected tagged GPCRs and microflui- endocytosis of 5-HT1BR is crucial for axonal targeting, since dic devices. Consequently, the quantity of endogenous proteins inverse agonist treatment, which prevents constitutive activation, that regulate GPCR endocytosis, such as arrestins, GRKs, and cla- leads to atypical accumulation of newly synthesized 5-HT1BRs on thrin, might be limiting, leading to patterns of GPCR distribution the somatodendritic plasma membrane. and dynamics that might not be representative of endogenous What is the fate of endocytosed somatodendritic CB1R? By using receptors. However, as reported previously, our experimental spatial separation of somatodendritic and axonal domains, herewe methods lead only to moderate receptor overexpression show that at least a portion of endocytosed CB1R is transported (Leterrier et al., 2006) and in a recent in vivo study (Thibault anterogradely in axons. Although such anterograde transcytotic et al., 2012) we were able to qualitatively reproduce previous transport has been shown for several proteins such as L1/ Downloaded from activation-dependent patterns of in vitro CB1R distribution NgCAM (Wisco et al., 2003; Yap et al., 2008), the transferrin recep- (Leterrier et al., 2006). In vivo traffic of the SSTR2 receptor tor (Hemar et al., 1997), and the TrkA neurotrophin receptor (Csaba et al., 2007) is also similar to what we reported in vitro (Ascano et al., 2009), the molecular composition of putative specif- (Lelouvier et al., 2008). Collectively, these data suggest at least ic transport intermediates (recently reviewed by Lasiecka and the qualitative pertinence of our in vitro model. Winckler, 2011) is not well known yet, so identification of antero- http://jmcb.oxfordjournals.org/ Dendritic endocytosis grade transcytotic partners for neuronal GPCRs is an important Elimination through spontaneous endocytosis from the somato- subject of future research. dendritic plasma membrane is important for the correct, polarized Axonal endocytosis distribution of several neuronal proteins, such as voltage-gated CB1R displays a reduced endocytic response in axons at all acti- sodium channels, VAMP2, nicotinic acetylcholine receptor alpha vation levels compared with dendrites. This reduced endocytic re- 4 subunit, NgCAM, TrkA receptor, and Caspr2 (Garrido et al., sponse to activation may be the result of either a general reduction 2001; Sampo et al., 2003; Wisco et al., 2003; Fache et al., 2004; of endocytosis in axons or a CB1R-specific reduction of internaliza- Xu et al., 2006; Ascano et al., 2009; Bel et al., 2009). We have pre- tion. Indeed, anchoring of CB1R to scaffolding proteins such as viously reported similar endocytic elimination of CB1R from the CRIP1 was previously proposed to explain axonal accumulation by guest on May 29, 2013 somatodendritic membrane (Leterrier et al., 2006). Notably, we of CB1R(McDonald et al., 2007; Niehaus et al., 2007). However, have also found that this constitutive endocytosis was blocked in the FRAP-based measurements presented in the present study, the presence of the inverse agonist, suggesting that, as observed as well as a previous report (Mikasova et al., 2008), show that in non-polarized cells (Leterrier et al., 2004), constitutive activation CB1Rs are relatively free to diffuse in the axonal plasma membrane. of CB1R was responsible for constitutive endocytosis in neurons. Thus, specific interactionthroughanchoring to scaffolding proteins We could also reduce constitutive activation and endocytosis of is unlikely. More probably, reduced internalization of CB1R is CB1R by inhibition of DAG-lipase, suggesting that activation is par- caused by a general internalization impairment in axons compared tially due to cell-autonomous production of the endocannabinoid with dendrites. To our knowledge, such neural sub-domain specific 2-AG (Turu et al., 2007). 2-AG is synthesized by diacylglycerol reductions in GPCR endocytosis have not yet been reported. (DAG) lipase activity, which is restricted to the dendritic region of However, it has been shown that the plasma membrane of cultured adult neurons (Bisogno et al., 2003). Thus, it is likely that newly hippocampal neurons is endocytosed throughout dendrites but synthesized CB1Rs delivered to the somatodendritic plasma mem- only in presynaptic terminals and varicosities in axons (Parton brane are permanently activated by the presence of the et al., 1992). In a separate study, plasma membrane endocytosis cell-autonomously produced lipid mediator 2-AG. Our results was estimated to be about eight times faster in dendrites than in have also suggested that steady-state endocytosis is due to mobil- axons of young cultured hippocampal neurons (Ye et al., 2007). ization of similar pathways that are activated following activation In the present study, in addition to confirming that CB1Rs are by agonists, but this notion remained controversial (McDonald retained in axons due to the reduced endocytic response to consti- et al., 2007; Harkany et al., 2008). Now we provide further experi- tutive activation (Leterrier et al., 2006), we were also able to mental evidence to show that the pharmacological conformational provide a quantitative insight into the nature of this reduced endo- state of CB1Rs is a major regulator of steady-state somatodendritic cytic response. We found that each investigated activation level endocytosis and transcytotic axonal targeting. Possible explana- yields lower number of endosomes in axons than in dendrites tions for the divergent results may be related to the important and that axonal endosomes contain less receptors. However, at natural variability of polarized distribution patterns of WT-CB1R higher activation levels the number and labeling intensity of in cultured hippocampal neurons (Figure 6A) and to the yet unex- axonal endosomes surpassed steady-state dendritic values and plained mixed phenotype of WT-CB1R expression following approached activated dendritic values, showing that the axonal 14 | Journal of Molecular Cell Biology Simon et al. endocytic deficit is not ‘hard-wired’, but rather a quantitative differ- through the microchannels toward the somatodendritic compart- ence between dendritic and axonal endocytic response to a given ment. Before fixation, the axonal compartment was incubated activation level. The mechanism of this difference is currently with anti-mouse IgG Alexa-Fluor 568 (1:400) for 10 min at 378C to unknown, and further research is clearly needed to identify the reveal transcytosed CB1R. We quantified the proximo-distal distri- cause of different endocytosis rates between somatodendritic bution profile of transcytosed CB1Rs by counting the number of and axonal GPCRs. Alexa-Fluor 568 labeled spots on 25 mm-long axon segments, start- What is the physiological relevance of the above findings? Our ing from the distal axon tip. results suggest a previously unanticipated high functional plasti- Kinetic model of neuronal GPCR targeting city for several neuronal GPCRs, predicting that receptors, which The model is depicted in Figure 5A and involves the following are usually highly polarized to one neuronal sub-domain, may variables: have significant effects on other parts of the neuron. The sub- s neuronal targeting model could be useful to interpret plastic ns: number of receptors in the soma changes in activation-dependent GPCR distribution in pathological s nm: number of membrane receptors in the soma contexts, such as epilepsy-induced intraneuronal redistribution na: number of receptors following the ‘slow’ path 1 after intern- reported for both CB1R(Karlocai et al., 2011) and SSTR2 (Csaba 1 alization in the soma et al., 2005). The model also proposes a relatively simple a 2 reason—i.e. different levels of steady-state activation—to n2: number of receptors following the ‘fast’ path after intern- explain why GPCRs may show different polarized distribution pat- alization in the soma Downloaded from terns depending on the neuronal cell type (Dumartin et al., 2007; a nm: number of membrane receptors in the axon Nathanson, 2008) or the neurochemical environment (Dournaud ks , k :kineticconstantsthatcharacterize the exchange between et al., 1998; Bernard et al., 2006). Indeed, the intriguing and not m r ns and ns yet elucidated redistribution of M2 muscarinic receptors from the s m s somatodendritic plasma membrane to the axonal surface in ki : internalization rate in the soma, which yields receptors fol- http://jmcb.oxfordjournals.org/ acetylcholinesterase-deficient mice (Decossas et al., 2003) may lowing path 2 be explained by the moderate chronic activation of M2 receptors k1: rate at which receptors follow path 1 by chronically elevated acetylcholin levels of this experimental km : rate at which receptors following path i reach the membrane strain, in analogy with our in vitro results obtained by moderate i a chronic activation of the SSTR2 receptor. ki : internalization rate in the axon In conclusion, our results indicate that activation and targeting vi: average speed of receptors following path i of GPCRs may be intimately interconnected in neurons. l , l : characterize the exchange through the diffusion barrier Accordingly, it is likely that long-term changes in endogenous 1 2

between the soma and axon membrane by guest on May 29, 2013 ligand levels as well as clinical pharmacological treatments may also result in redistribution of several neuronal GPCRs. Thus, besides proposing a previously unappreciated physiological regu- Supplementary material latory mechanism, our model also raises the possibility that Supplementary material is available at Journal of Molecular Cell treatment-induced receptor redistributions may contribute to Biology online. both wanted and unwanted effects of therapeutic drugs acting on neuronal GPCRs. Acknowledgments The authors thank Dr Matthieu Piel (Institut Curie, Paris) for Materials and methods helpful discussions, Dr Frederick Saudou (Institut Curie, Paris), Dr Experimental procedures, model dynamics, and suppliers are Patrick Tabeling, Dr Herve´ Willaime, and Fabrice Monti (MMN listed in detail in Supplementary material. lab-ESPCI-ParisTech, Paris) for the help with microfluidic technol- Cortical neuronal culture in the microfluidic device ogy, and also Alexandre Roland, Wigo Bertrams, and Se´bastien 18 Cortexes of rat embryos were dissected at embryonic Day . Nicoulaud, who participated in this project. We thank Prof. Dr After trypsinization, dissociated cells were electroporated with Dieter K. Meyer and PD. Dr Jost Leemhuis (Albert-Ludwigs 5 1 mg of Flag-CB R-eGFP plasmid using the Amaxa system (Lonza) University,Freiburg,Germany) for providinguswithrat hippocampal and grown in a polydimethylsiloxane two-chamber microfluidic neurons for the calcium measurements. We are grateful to Natalia device which allows for separate incubation of the somatodendritic Velez-Alicea (MIT, Boston, MA) for correction of the English syntax. and axonal domains (Taylor et al., 2005). Dissociated neurons were distributed into the somatodendritic chamber and spontaneously Funding 450 grew axons through the mm long microchannels into the Z.L. was supported by a grant from the French Agence Nationale 7 axonal chamber. At DIV , a mix of mouse monoclonal anti-FLAG de la Recherche (ANR- 09-MNPS-004-01). This work was funded by 1 500 1 4 antibodies ( / ) and % of BSA was added for h to the soma- INSERM, CNRS, ESPCI-ParisTech, the ‘Fondation pour la Recherche 4 37 todendritic chamber. Incubation for h at 8C allowed endocytic Me´dicale’ (FRM) and the ‘Fondation ARC pour la Recherche sur le uptake and transcytotic transport. To eliminate the possibility of Cancer’. antibody leaking into the axonal chamber, we added a higher volume of medium into the axonal chamber to induce a slow flow Conflict of interest: None declared. Activation-dependent GPCR distribution in neurons Journal of Molecular Cell Biology | 15

References determinant in the C-terminus of the sodium channel Na(v)1.2. EMBO J. 20, Ahmari, S.E., Buchanan, J., and Smith, S.J. (2000). Assembly of presynaptic active 5950–5961. zones from cytoplasmic transport packets. Nat. Neurosci. 3, 445–451. Harkany, T., Keimpema, E., Barabas, K., et al. (2008). Endocannabinoid functions Ascano, M., Richmond, A., Borden, P., et al. (2009). Axonal targeting of Trkrecep- controlling neuronal specification during brain development. Mol. Cell. tors via transcytosis regulates sensitivity to neurotrophin responses. Endocrinol. 286,S84–S90. J. Neurosci. 29, 11674–11685. Hemar, A., Olivo, J.C., Williamson, E., et al. (1997). Dendroaxonal transcytosis of Bel, C., Oguievetskaia, K., Pitaval, C., et al. (2009). Axonal targeting of Caspr2 in transferrin in cultured hippocampal and sympathetic neurons. J. Neurosci. 17, hippocampal neurons via selective somatodendritic endocytosis. J. Cell Sci. 9026–9034. 122, 3403–3413. Holliday, N.D., Holst, B., Rodionova, E.A., et al. (2007). Importance of constitu- Bernard, V., Decossas, M., Liste, I., et al. (2006). Intraneuronal trafficking of tive activity and arrestin-independent mechanisms for intracellular trafficking G-protein-coupled receptors in vivo. Trends Neurosci. 29, 140–147. of the ghrelin receptor. Mol. Endocrinol. 21, 3100–3112. Biermann, B., Ivankova-Susankova, K., Bradaia, A., et al. (2010). The Sushi Jacquier, V., Prummer, M., Segura, J.M., et al. (2006). Visualizing odorant recep- domains of GABAB receptors function as axonal targeting signals. tor trafficking in living cells down to the single-molecule level. Proc. Natl Acad. J. Neurosci. 30, 1385–1394. Sci. USA 103, 14325–14330. Bisogno, T., Howell, F., Williams, G., et al. (2003). Cloning of the first sn1-DAG Karlocai, M.R., Toth, K., Watanabe, M., et al. (2011). Redistribution of CB1 can- lipases points to the spatial and temporal regulation of endocannabinoid sig- nabinoid receptors in the acute and chronic phases of pilocarpine-induced naling in the brain. J. Cell Biol. 163, 463–468. epilepsy. PLoS One 6, e27196. Carrel, D., Simon, A., Emerit, M.B., et al. (2011). Axonal targeting of the 5-HT1B Lasiecka, Z.M., and Winckler, B. (2011). Mechanisms of polarized membrane serotonin receptor relies on structure-specific constitutive activation. Traffic trafficking in neurons—focusing in on endosomes. Mol. Cell. Neurosci. 48, 12, 1501–1520. 278–287. Chanrion, B., Mannoury la Cour, C., Gavarini, S., et al. (2008). Inverse agonistand Lelouvier, B., Tamagno, G., Kaindl, A.M., et al. (2008). Dynamics of somatostatin

neutral antagonist actions of antidepressants at recombinant and native type 2A receptor cargoes in living hippocampal neurons. J. Neurosci. 28, Downloaded from 5-hydroxytryptamine2C receptors: differential modulation of cell surface ex- 4336–4349. pression and signal transduction. Mol. Pharmacol. 73, 748–757. Leterrier, C., Bonnard, D., Carrel, D., et al. (2004). Constitutive endocytic cycle of Charara, A., Blankstein, E., and Smith, Y. (1999). Presynaptic kainate receptors in the CB1 cannabinoid receptor. J. Biol. Chem. 279, 36013–36021. the monkey striatum. Neuroscience 91, 1195–1200. Leterrier, C., Laine, J., Darmon, M., et al. (2006). Constitutive activation drives Cherry, R.J., Godfrey, R.E., and Peters, R. (1982). Mobility of bacteriorhodopsin in compartment-selective endocytosis and axonal targeting of type 1 cannabin-

lipid vesicles. Biochem. Soc. Trans. 10, 342–343. oid receptors. J. Neurosci. 26, 3141–3153. http://jmcb.oxfordjournals.org/ Coutts, A.A., Anavi-Goffer, S., Ross, R.A., et al. (2001). Agonist-induced internal- Le Verche, V., Kaindl, A.M., Verney,C., et al. (2009). The somatostatin2A receptor ization and trafﬁcking of cannabinoid CB1 receptors in hippocampal neurons. is enriched in migrating neurons during rat and human brain development and J. Neurosci. 21, 2425–2433. stimulates migration and axonal outgrowth. PLoS One 4, e5509. Craig, A.M., Wyborski, R.J., and Banker, G. (1995). Preferential addition of newly Marion, S., Weiner, D.M., and Caron, M.G. (2004). RNA editing induces variation synthesized membrane protein at axonal growth cones. Nature 375, in desensitization and trafﬁcking of 5-hydroxytryptamine 2c receptor iso- 592–594. forms. J. Biol. Chem. 279, 2945–2954. Csaba, Z., Pirker, S., Lelouvier, B., et al. (2005). Somatostatin receptor type 2 McDonald, N.A., Henstridge, C.M., Connolly, C.N., et al. (2007). An essential role undergoes plastic changes in the human epileptic dentate gyrus. forconstitutive endocytosis, but not activity, in the axonal targeting of the CB1 J. Neuropathol. Exp. Neurol. 64, 956–969. cannabinoid receptor. Mol. Pharmacol. 71, 976–984.

Csaba, Z., Lelouvier, B., Viollet, C., et al. (2007). Activated somatostatin type 2 Meshul, C.K., and McGinty, J.F. (2000). Kappa opioid receptor immunoreactivity by guest on May 29, 2013 receptors traffic in vivo in central neurons from dendrites to the trans Golgi in the nucleus accumbens and caudate-putamen is primarily associated with before recycling. Traffic 8, 820–834. synaptic vesicles in axons. Neuroscience 96, 91–99. D’Antona, A.M., Ahn, K.H., and Kendall, D.A. (2006). Mutations of CB1 T210 Mikasova, L., Groc, L., Choquet, D., et al. (2008). Altered surface trafficking of produce active and inactive receptor forms: correlations with ligand affinity, presynaptic cannabinoid type 1 receptor in and out synaptic terminals paral- receptor stability, and cellular localization. Biochemistry 45, 5606–5617. lels receptor desensitization. Proc. Natl Acad. Sci. USA 105, 18596–18601. Decossas, M., Bloch, B., and Bernard, V. (2003). Trafficking of the muscarinic m2 Misumi, Y., Misumi, Y., Miki, K., et al. (1986). Novel blockade by brefeldin A of autoreceptor in cholinergic basalocortical neurons in vivo: differential regula- intracellular transport of secretory proteins in cultured rat hepatocytes. tion of plasma membrane receptor availability and intraneuronal localization J. Biol. Chem. 261, 11398–11403. in acetylcholinesterase-deficient and -inhibited mice. J. Comp. Neurol. 462, Mohammad, S., Baldini, G., Granell, S., et al. (2007). Constitutive traffic of 302–314. melanocortin-4 receptor in Neuro2A cells and immortalized hypothalamic Dournaud, P., Boudin, H., Schonbrunn, A., et al. (1998). Interrelationships neurons. J. Biol. Chem. 282, 4963–4974. between somatostatin sst2A receptors and somatostatin-containing axons Morisset, S., Rouleau, A., Ligneau, X., et al. (2000). High constitutive activity of in rat brain: evidence for regulation of cell surface receptors by endogenous native H3 receptors regulates histamine neurons in brain. Nature 408, somatostatin. J. Neurosci. 18, 1056–1071. 860–864. Dumartin, B., Doudnikoff, E., Gonon, F., et al. (2007). Differences in ultrastruc- Morris, D.P., Price, R.R., Smith, M.P., et al. (2004). Cellular trafficking of human tural localization of dopaminergic D1 receptors between dorsal striatum alpha1a-adrenergic receptors is continuous and primarily and nucleus accumbens in the rat. Neurosci. Lett. 419, 273–277. agonist-independent. Mol. Pharmacol. 66, 843–854. Fache, M.P., Moussif, A., Fernandes, F., et al. (2004). Endocytotic elimination and Muly, E.C., Maddox, M., and Smith, Y. (2003). Distribution of mGluR1alpha and domain-selective tethering constitute a potential mechanism of protein seg- mGluR5 immunolabeling in primate prefrontal cortex. J. Comp. Neurol. 467, regation at the axonal initial segment. J. Cell Biol. 166, 571–578. 521–535. Freiman, I., Anton, A., Monyer, H., et al. (2006). Analysis of the effects of canna- Muly, E.C., Maddox, M., and Khan, Z.U. (2010). Distribution of D1 and D5 dopa- binoids on identified synaptic connections in the caudate-putamen by paired mine receptors in the primate nucleus accumbens. Neuroscience 169, recordings in transgenic mice. J. Physiol. 575, 789–806. 1557–1566. Fritschy, J.M., Meskenaite, V., Weinmann, O., et al. (1999). GABAB-receptor Nakada, C., Ritchie, K., Oba, Y., et al. (2003). Accumulation of anchored proteins splice variants GB1a and GB1b in rat brain: developmental regulation, cellular forms membrane diffusion barriers during neuronal polarization. Nat. Cell distribution and extrasynaptic localization. Eur. J. Neurosci. 11, 761–768. Biol. 5, 626–632. Gainetdinov, R.R., Premont, R.T., Bohn, L.M., et al. (2004). Desensitization of G Nathanson, N.M. (2008). Synthesis, trafficking, and localization of muscarinic protein-coupled receptors and neuronal functions. Annu. Rev. Neurosci. 27, acetylcholine receptors. Pharmacol. Ther. 119, 33–43. 107–144. Niehaus, J.L., Liu, Y., Wallis, K.T., et al. (2007). CB1 cannabinoid receptor activity Garrido, J.J., Fernandes, F., Giraud, P., et al. (2001). Identification of an axonal is modulated by the cannabinoid receptor interacting protein CRIP 1a. Mol. 16 | Journal of Molecular Cell Biology Simon et al.

Pharmacol. 72, 1557–1566. subcellular distribution patterns of CB1 cannabinoid receptors in the rat fore- Nyiri, G., Cserep, C., Szabadits, E., et al. (2005). CB1 cannabinoid receptors are brain. Cereb. Cortex. doi: 10.1093/cercor/bhs240. enriched in the perisynaptic annulus and on preterminal segments of hippo- Turu, G., Simon, A., Gyombolai, P., et al. (2007). The role of diacylglycerol lipase campal GABAergic axons. Neuroscience 136, 811–822. in constitutive and angiotensin AT1 receptor-stimulated cannabinoid CB1 re- Pan, X., Ikeda, S.R., and Lewis, D.L. (1996). Rat brain cannabinoid receptor mod- ceptor activity. J. Biol. Chem. 282, 7753–7757. ulates N-type Ca2+ channels in a neuronal expression system. Mol. Valdes, V., Valenzuela, J.I., Salas, D.A., et al. (2012). Endoplasmic reticulum Pharmacol. 49, 707–714. sorting and kinesin-1 command the targeting of axonal GABAB receptors. Parton, R.G., Simons, K., and Dotti, C.G. (1992). Axonal and dendritic endocytic PLoS One 7, e44168. pathways in cultured neurons. J. Cell Biol. 119, 123–137. von Zastrow, M. (2010). Regulation of opioid receptors by endocytic membrane Riad, M., Garcia, S., Watkins, K.C., et al. (2000). Somatodendritic localization of traffic: mechanisms and translational implications. Drug Alcohol Depend. 5-HT1A and preterminal axonal localization of 5-HT1B serotonin receptors in 108, 166–171. adult rat brain. J. Comp. Neurol. 417, 181–194. Whistler, J.L., Enquist, J., Marley, A., et al. (2002). Modulation of postendocytic Sampo, B., Kaech, S., Kunz, S., et al. (2003). Two distinct mechanisms target sorting of G protein-coupled receptors. Science 297, 615–620. membrane proteins to the axonal surface. Neuron 37, 611–624. Winckler, B., and Yap, C.C. (2011). Endocytosis and endosomes atthe crossroads Sarret, P., Botto, J.M., Vincent, J.P., et al. (1998). Preferential expression of sst2A of regulating trafficking of axon outgrowth-modifying receptors. Traffic 12, over sst2B somatostatin receptor splice variant in rat brain and pituitary. 1099–1108. Neuroendocrinology 68, 37–43. Wisco, D., Anderson, E.D., Chang, M.C., et al. (2003). Uncovering multiple Scherrer, G., Tryoen-Toth, P., Filliol, D., et al. (2006). Knockin mice expressing axonal targeting pathways in hippocampal neurons. J. Cell Biol. 162, fluorescent delta-opioid receptors uncover G protein-coupled receptor dy- 1317–1328. namics in vivo. Proc. Natl Acad. Sci. USA 103, 9691–9696. Xu, J., Zhu, Y., and Heinemann, S.F. (2006). Identification of sequence motifs that Shigemoto, R., Kinoshita, A., Wada, E., et al. (1997). Differential presynaptic lo- target neuronal nicotinic receptors to dendrites and axons. J. Neurosci. 26,

calization of metabotropic glutamate receptor subtypes in the rat hippocam- 9780–9793. Downloaded from pus. J. Neurosci. 17, 7503–7522. Yap, C.C., Wisco, D., Kujala, P., et al. (2008). The somatodendritic endosomal Taylor, A.M., Blurton-Jones, M., Rhee, S.W., et al. (2005). A microﬂuidic culture regulator NEEP21 facilitates axonal targeting of L1/NgCAM. J. Cell Biol. 180, platform for CNS axonal injury, regeneration and transport. Nat. Methods 2, 827–842. 599–605. Ye, B., Zhang, Y., Song, W., et al. (2007). Growing dendrites and axons differ in Thibault, K., Carrel, D., Bonnard, D., et al. (2012). Activation-dependent their reliance on the secretory pathway. Cell 130, 717–729. http://jmcb.oxfordjournals.org/ by guest on May 29, 2013

Supplementary Figure S1: T210A mutation leads to a new CB1R-specific gain of somatodendritic function (related to Figure 1). Cultured hippocampal neurons, transfected with WT-CB1R, T210I-CB1R or T210A-CB1R, were loaded with the calcium- sensitive fluorescent dye Fura-4F AM and superfused. Calcium increases were evoked by pressure ejection of potassium (500 mM; 5 s pressure pulse; note that the concentration of potassium was greatly diminished during diffusion to the recorded neuron) from a pipette positioned in the vicinity of the neurons. Changes in background-corrected fluorescence ratios were expressed as percentages of the initial fluorescence (F/F0). Aa. A neuron expressing WT-CB1R (top) or T210A-CB1R (bottom) is shown. Ab, Ac. Potassium-evoked fluorescence changes in the neuron shown in A during superfusion of vehicle or WIN55212-2 (WIN); WIN inhibits potassium-evoked calcium increase. B. Temporal pattern of potassium-evoked calcium changes during superfusion with vehicle (Ve) (see Ab), WIN (see Ac) or the voltage-gated calcium channel blocker cadmium (Cd2+). C. Statistical evaluation of the effects of WIN55212-2 and cadmium on potassium- evoked calcium increases in the soma and dendrites (the evoked maxima were evaluated; see panel B. The number of independent experiments was: n=5 for WT-CB1R, n=6 for T210A-CB1R and n=7 T210I-CB1R. A filled symbol indicates a significant difference vs. the initial reference period PRE and *: p <0.05, **: p <0.01 vs T210I-CB1R, one-way ANOVA. D. The effect of the cannabinoid agonist WIN55212-2 on potassium-evoked increases in intracellular calcium concentration in cultured hippocampal neurons is blocked by the CB1R antagonist rimonabant. Neurons expressing WT-CB1R (top) and T210A-CB1R (bottom) are shown. Db and Dc) Potassium-evoked fluorescence changes in the neuron shown in D during superfusion of rimonabant (RIM) and RIM plus WIN55212-2 (RIM + WIN); RIM pretreatment abolishes the effects of WIN on potassium-evoked calcium increase both in soamta and dendrites. E. Temporal pattern of potassium-evoked calcium changes during superfusion of RIM (see Db), superfusion of RIM plus WIN (see Dc) and superfusion of the voltage-gated calcium channel blocker cadmium. F. Statistical evaluation of the effects of WIN55212-2 in the presence of RIM and the effects of cadmium on potassium-evoked calcium increases in the soma and dendrites (the evoked maxima were evaluated). The superfusions of RIM, WIN and cadmium are indicated by bars. The number of experiments was: n=9 (CB1R), n=9 (CB1R-T210A) and n=11 (CB1R-T210I).

Supplementary Figure S2: Spontaneous endocytosis of endogenous CB1R and the scheme of quantitative evaluation of endocytosis rates in axon and dendrites (related to Figure 3). A. One hour of antibody feeding with an antibody directed against the N-terminus of endogenous CB1R was realized on DIV15 neurons. Internalized (acid-strip resistent) anti-CB1R antibodies were detected in numerous cytoplasmic endosome- like organelles in about 7% of neurons (in our cultures, 5%-10% of mature cultured hippocampal neurons express CB1R endogenously (Leterrier et al., 2006), mostly in neuronal cell bodies and neurites expressing the dendritic marker Microtubule Associated Protein 2 (MAP2). B. Pre-treatment with the inverse-agonist AM281 resulted in a highly significant decrease in the number of neurons displaying endocytosed antibody. Shown data represents two pooled independent experiments (Mean ± S.E.M, * for p<0.05, Student t-test). Scale bar: 25 µm. C. Schematic representation of the quantitative evaluation of endocytosis rates in axon and dendrites, as described in Supplementary materials and methods. D. Mean diameter of axons and dendrites was determinated from soluble YFP transfected neurons and did not show significant differences. Values are presented as boxplots and represent data from 16 parts of axons and 48 parts of dendrites of at least three pooled independent experiments (Student's T-test).

Supplementary Figure S3: CB1R is not anchored in the axon (related to Figure 4). A. Representative time series from a FRAP experiment, showing the axon of a cultured hippocampal neuron expressing Flag-WT-CB1R-eGFP. Focused photo-bleaching (red square) is followed by rapid recovery of fluorescence. General decrease of fluorescence due to imaging-related photo-bleaching is measured in the yellow square. Scale bar: 2 µm. B. Diffusion coefficients of WT-CB1R (n = 28), T210A-CB1R (n = 21) and SSTR2 (n = 19) measured in axons and in somatodendritic areas for SSTR2 (n = 42). Diffusion rates appear to be close to free diffusion rates, suggesting the absence of receptor anchoring in neuronal membranes. C. Mobile fraction of receptors after photobleaching. Data shown represents three pooled independent experiments (Mean ± S.E.M, * for p<0.05 and *** for p<0.001 vs. SSTR2 soma- dendrites, one-way Anova).

Supplementary Figure S4: The three CB1R forms show only limited co-localization with the lysosome marker LAMP1. DIV8 cultured neurons expressing three forms of Flag-CB1R-eGFP: WT-CB1R (A), T210I-CB1R (B) and T210A-CB1R (C) (GFP - green) and labeled for lysosome marker LAMP-1 (Alexa568 - red). Representative optical sections of the somatodendritic area show only limited co-localization. Scale bar: 10 µm. Inset scale bar : 500 nm.

Supplementary Figure S5: Examples for receptor distribution phenotypes in DIV7 neurons expressing different CB1R forms (related to Figure 6). Three hours after transfection, neurons were incubated with various concentrations of ligand for 24 hours and were selectively labeled for surface localized CB1R in living neurons. All transfected neurons on the coverslip were classified according to the distribution of surface labeling by a collaborator blind to the experimental condition. Classification categories are "Internalized" (no detectable CB1R surface labeling), "Internalized / Axonal" (a weak labeling is detectable on the axonal surface), "Axonal" (corresponds to the natural phenotype of CB1R, which is polarized mostly to the axonal plasma membrane, similar to untreated neurons), "Axonal / Uniform" (CB1R is labeling is clearly detectable on the somatodendritic surface, but is less intense than the labeling of the axonal surface) and "Uniform or somatodendritic" (somatodendritic surface labeling is equivalent to, or more intense than labeling of the axon surface). Scale bar: 50 µm.

Supplementary Figure S6: Inhibition of endocytosis by dominant-negative rab5 (S34N) leads to significant redistribution of surface WT-CB1R population. Hippocampal neurons at DIV8 were co-transfected with Flag-WT-CB1R-eGFP (green) and with soluble eCFP (blue) as a control (A) or fusions of eCFP with wild-type rab5 (WT) (B) or dominant-negative rab5 (S34N) (C). Twenty-four hours later, neurons were stained for surface CB1R (red) and fixed. D. For each condition, all transfected neurons were classified according to the distribution of surface labeling as described in Sup Figure 5. eCFP or wild-type rab5 did not change sub- neuronal distribution of CB1R, whereas dominant-negative rab5 (S34N) strongly increased the Uniform or Somatodendritic population of neurons. Shown data represented two pooled, independent experiments (Mean ± S.E.M). Left panel scale bar: 50 µm. Right panel scale bar: 20 µm. Supplementary materials and methods:

Chemicals, antibodies and DNA constructs CB1R agonist (R)-(+)-[2,3-dihydro-5-methyl-3-(4-morpholinylmethyl)pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl]-1- naphthalenylmethanone mesylate [WIN55,212-2 (WIN)] and CB1R inverse agonist 1-(2,4-dichlorophenyl)-5-(4-iodophenyl)- 4-methyl-N-4-morpholinyl-1H-pyrazole-3-carboxamide (AM281) were obtained from TOCRIS. All culture media and additives were purchased from PAA Laboratories. The anti-CB1R N-terminal (N-Ter) antibody was characterized elsewhere (Leterrier et al., 2004). Monoclonal mouse anti-microtubule-associated protein 2 (clone HM-2), monoclonal mouse anti- FLAG antibody (clone M1), Polyclonal rabbit anti-LAMP1 antibody, bovine serum albumin (BSA), Dynasore, Poly-Allyl amine (PAA) and BFA were from Sigma-Aldrich. Alexa-Fluor conjugated secondary antibodies, Alexa-Fluor® 647 Monoclonal Antibody Labeling Kit and Lipofectamine 2000 were from Invitrogen. Fluorescent protein encoding plasmids were from Clontech. The FLAG-T210I-CB1R-enhanced green fluorescent protein (eGFP) and FLAG-T210A-CB1R- eGFP constructs were obtained by inserting a point mutation into the validated FLAG-WT-CB1R-eGFP sequence (Leterrier, et al., 2006) using the QuikChange mutagenesis system (Stratagene). All constructs were verified by full-length sequencing. The Ras-related GTP-binding protein (rab)-enhanced cyan fluorescent protein (eCFP) plasmids were gifts from Dr Robert Lodge (Université du Quebec, Montréal Quebec, Canada). Measure of CB1R and/or SSTR2 traffic and signaling in HEK-293 cells Human embryonic kidney (HEK)-293 (ATCC CRL-1573) cells were cultivated in DMEM with 10% fetal calf serum. For transient transfections, HEK-293 cells were transfected with 0.8 µg of FLAG-T210x-CB1R-eGFP or FLAG-SSTR2A-eGFP plasmid DNA using Effectene (Qiagen). For flow cytometry-based endocytosis assays, adherent living cells were incubated with serum-free DMEM medium containing 0.1% BSA supplemented with anti-FLAG antibody, directly labeled by using Alexa647 Monoclonal Antibody Labeling Kit. After this antibody feeding step, a short trypsin treatment serves both to detach cells and to eliminate non- internalized FLAG-antibodies. Trypsin was inhibited by adding serum-containing culture medium, and cells were diluted and fixed into buffer containing paraformaldehyde (final concentration of 2.5%). Following an optional single centrifugation and re-dilution step to remove paraformaldehyde, cells (about 10000 cells/sample) were directly analyzed in the flow cytometer for GFP and Alexa647 fluorescence levels. Each data point was realized in duplicate, and experiments were carried out in duplicate. Alexa647 intensity values were normalized by the expression level of receptors (GFP mean). A highly sensitive gene reporter assay was used to measure intracellular cAMP level of CB1R, as described previously (Lelouvier et al., 2008) using the Great EscAPe SEAP Chemiluminescence kit 2.0 (Clontech) following the protocol provided by the manufacturer. Briefly, HEK-293 cells were transfected with pCRE-SEAP (gift of Dr L. Devi, Mount Sinai School of Medicine, NY, USA) alone or with each form of T210x-CB1R receptors in 0.1 mg/ml PAA pre-coated 24-well plates using Effectene (Qiagen). Cells were starved overnight in serum-free medium, and exposed to 10 µM of forskolin for 4–6 h. 25 µl of supernatant were transferred in a Nunclon™ Surface 96-well plate (Nalge Nunc International) in addition of dilution buffer. The plate was incubated for 30 min at 65°C to eliminate endogenous alkaline phosphatase activity. Next, after an equilibration to room temperature, SEAP substrate solution was added and incubated for 30 min at room temperature. The intensity of the chemiluminescence signal was determined by a luminometer (Mediator PhL; Aureon Biosystems GmbH). Each value point, realized in triplicate, was obtained by dividing the fluorescence intensity in forskolin-treated cells by that of forskolin untreated cells. Each experiment was normalized to WT-CB1R transfected cells. Assays were performed in triplicate. Assays in cultured primary neurons Hippocampal neuronal cultures were performed essentially as described previously (Leterrier, et al., 2006). Briefly, hippocampi of rat embryos were dissected at embryonic day 18, trypsinized and dissociated with a fire-polished Pasteur pipette. Cells were counted and plated on poly-D-lysine-coated glass coverslips and then cultivated in glia-conditioned Neurobasal medium. Neurons were transfected after 7–10 days in vitro (DIV) using Lipofectamine 2000. They were treated and processed at 24h after transfection. For pharmacological treatments, concentrated ligands dissolved in dimethylsulfoxide (DMSO) were directly added to the culture medium. The highest final concentration reached was 0.1% DMSO, and control experiments with up to 0.5% DMSO have not shown effects on neuronal morphology and on the cellular distribution of CB1R (data not shown). For standard immunocytochemistry, neurons were briefly rinsed with Dulbecco’s PBS (DPBS; Invitrogen, catalog #14040- 083) and fixed in DPBS containing 4% paraformaldehyde and 4% sucrose. After permeabilization with a 5 min incubation in DPBS containing 0.1% Triton X-100 and blocking for 30 min in antibody buffer (DPBS supplemented with 2% BSA and 3% normal goat serum), neurons were incubated with primary antibodies diluted 1:400–1:3000 in antibody buffer for 1 h at room temperature. After DPBS rinses, neurons were labeled with secondary antibodies diluted to 1:400 in antibody buffer for 30 min at room temperature. For live surface CB1R staining, anti-FLAG M1 antibody was added at 1:500–1:1000 to the culture medium and incubated for 5 min at 37°C. Neurons were briefly rinsed with DPBS, fixed in DPBS containing 4% paraformaldehyde, 4% sucrose, and incubated for 30 min in secondary antibody diluted to 1:400. For antibody feeding experiments, living neurons were incubated for 1 h at 37°C with anti-FLAG M1 antibody (1:800) or N-ter antibody before fixation, permeabilization, and staining with secondary antibody (1:400). During agonist treatment, neurons were incubated simultaneously with the anti- FLAG antibody, whereas inverse agonist treated neurons were pretreated for two hours before the antibody feeding. To strip anti-FLAG antibody at the surface, neurons were subjected to a 4 min incubation at 4°C with a solution of 0.5 M NaCl and 0.2 M acetic acid, pH 3.5 (Strip Acid - SA solution), before being fixed. Endocytosed antibodies were revealed by 30 min of incubation in secondary antibody diluted to 1:400. Coverslips were finally mounted in Mowiol (Calbiochem) eventually containing Hoechst 33432 diluted to 1:1000. Widefield images were taken on a Leica DM-R using dry 20× NA 0.7, oil immersion 40× NA 1.2, and oil-immersion 100× NA 1.4 objectives and Coolsnap HQ (Roper Scientific), QImaging QICAM (Burnaby). Emission and excitation filters proper to each fluorophore were used sequentially, and the absence of cross talk between different channels was checked with selectively labeled preparations. For measurement of CB1R labeling intensity in axons and dendrites, widefield 20× images from either the green channel (total CB1R) or the red channel (surface CB1R) of transfected neurons were used as described previously (Leterrier, et al., 2006). Segments were traced along morphologically-identified axons and dendrites on the surface-labeled image using NeuronJ (Meijering et al., 2004) as a plugin for the ImageJ software (V1.44). For the measure of the axonal gradient of CB1R labeling intensity, axons were traced from proximal to distal regions using NeuronJ. The intensity of each pixel in this section was measured, and axons of similar length (100 µm of variability) were selected and normalized to calculate the mean fluorescence intensity of the surface distribution of receptors. For quantification of the surface-to-total ratio (S/T ratio), 40× images from the green channel (total FCB1R-eGFP fluorescence) and red channel (surface CB1R labeling) were used. A mask of each cell was generated from green images using ImageJ software and then used to calculate each channel intensity of fluorescence. For the quantification of antibody-feeding experiments, intensity of each individual endosome was measured in morphologically identified axon and dendrites. Three parameters were measured; the number of endosomes (normalized by the length of neurites), the mean intensity of endosomes and the total internalized receptors (obtained by multiplying the number and the mean intensity of endosomes). Widefield images from soluble YFP transfected neurons were used to measure the mean diameter of axons (four measurements by selected neurite) and dendrites (three measurements by selected neurite) using imageJ software. For the measure of surface distribution patterns, DIV7 neurons were transfected with the three forms of Flag-CB1R-eGFP for three hours. Then, cells were treated for 24 h with various concentrations of ligands in culture medium containing 0.1% of BSA to reduce non-specific ligand adsorption. Next, surface populations of living cells were specifically labeled and each transfected neuron was blindly classified into one of 5 categories depending on the distribution of surface labeling (Supplementary Figure S4). The categories were "internalized" (no detectable CB1R surface labeling), "Internalized / Axonal" (a low level of detectable labeling on the axonal surface), "Axonal" (corresponding to the wild-type phenotype of CB1R, which is polarized mostly to the axonal plasma membrane, similar to untreated neurons), "Axonal / Uniform" (CB1R present on both the somatodendritic and axonal surfaces, but mostly axonal) and "Uniform or somatodendritic" (somatodendritic surface labeling equivalent to, or more intense than labeling of the axon surface). The proportion of neurons in the different classes was calculated with respect to the total number of transfected neurons per coverslip. For FRAP experiments, neurons expressing Flag-CB1R-eGFP (WT-CB1R or T210A-CB1R) or Flag-SSTR2A-eGFP

(SSTR2) receptors were maintained at 37°C in a temperature-controlled chamber with 5% CO2. Images were obtained using an oil-immersion objective lens (Plan-Apochromat 63/1.4) on a Zeiss Observer inverted microscope equipped with a LSM 5 Exciter confocal scanning system (Carl Zeiss). A 488-nm laser was used for both imaging and bleaching. The first 10 images of regions of interest were recorded before bleaching to establish a baseline using 2% laser power. Membrane areas were then bleached with 100% laser power for 1 sec (10 cycles) followed by 2 min postbleached recordings. Fluorescence intensity was measured with ImageJ software on each movie using a square whose dimensions were adapted to the size of the bleached area. The non-stimulated decrease of fluorescence was measured as far as possible from the bleached area within a 2.71 × 0.88 µm2 region-of-interest. Next, the steering ratio of the linear regression (called a) was used to calculate the corrected intensity of the bleached area : I(t)corrected  I(t)measured  at . After normalization

t 1 using GraphPad Prism 4 (San Diego), the half-life time ( 2 ) of fluorescence recovery and the maximum intensity recovered after photobleaching ( I ) were measured. 0,275L2 D  t 1 The diffusion coefficient was calculated using the equation 2 , valid for diffusion in a narrow cylinder open at both ends, with L equaling the length of the bleached area (Konzack et al., 2007). The mobile fraction corresponds to

100  I. For videomicroscopy, DIV7 neurons were transfected with the Flag-CB1R-eGFP and were incubated with anti-FLAG antibody coupled with AlexaFluor® 647 fluorochrome for 5 min. At the end of the incubation, neurons were washed and imaged after 15 min. Images were obtained with a CCD CoolSNAPHQ2 camera with a 63× 1.4 NA objective and binning of 2. Three images per minute were taken using 200 msec exposure time for 20 min, after which neurons were fixed in 4% of paraformaldehyde and immunolabeled for the dendritic marker MAP2. To measure SSTR2 distribution kinetics, BFA was used to accumulate newly synthesized SSTR2 in a mixed Golgi-ER intracellular compartment, and to allow a pulse of receptor transport after BFA washout (Wisco et al., 2003; Fache et al., 2004). Neurons were transfected as described above with Flag-SSTR2A-eGFP. After 3 h, the transfection medium was removed and BFA was added to 0.75 µg/ml of final concentration. After 12–18 h, coverslips were washed three times in conditioned medium and incubated with fresh conditioned medium, either with vehicle (0.1%) or various concentration of the agonist octreotide. Twelve hours later, coverslips were removed, stained for surface SSTR2 with anti-FLAG M1 antibody as described above and fixed. All transfected neurons for each condition were classified based on their surface SSTR2 distribution, as described above. The proportion of neurons in each of the different classes was calculated with respect to the total number of transfected neurons per coverslip. Fluorescence measurement of calcium concentrations in cultured hippocampal neurons Primary hippocampal neuronal cultures were prepared as described previously (Henle et al., 2006). After 8–10 days in culture, neurons were transfected with wild-type receptor (WT-CB1R), the CB1 receptor mutant T210A-CB1R or the CB1 receptor mutant T210I-CB1R using the calcium phosphate transfection method (Jiang and Chen, 2006). Fluorescent calcium imaging was carried out 1–3 days after transfection. Neurons were loaded with Fura-4F AM 1×10-6 M (incubation for 30 min at 37°C). After the incubation, neurons were transferred to the recording chamber and superfused at room temperature with a buffer of the following composition (mM): NaCl 126, NaH2PO4 1.2, KCl 3, MgCl2 1, CaCl2 2.5, NaHCO3

26, glucose 10, pH 7.3–7.4 (after the solution was gassed with 95% O2 / 5% CO2). Imaging was carried out with a system consisting of Polychrome IV monochromatic light source, a cooled IMAGO VGA CCD camera and TILLvision imaging software (TILL Photonics). Fura-4F was excited at 340 and 380 nm wavelengths, and a 410 DCLP dichroic filter and a 470 LP emission filter were used (Omega Optical). Fluorescence values were corrected for background fluorescence. Ratios of emissions in response to 340 and 380 nm excitation were calculated. These emission ratios were used for further calculations. Stimulation-evoked changes in fluorescence ratios (F) were related to baseline fluorescence ratios (F0) (F/F0). F/F0 values within one experiment were normalized to the initial reference value (PRE; see Supplementary Figure S1). Kinetic model of neuronal GPCR targeting In addition of all parameters described in Materials and Methods, we will also make use of the following useful quantities: m - li vi /ki is the average length covered by receptors of route i before reaching the axon membrane. l  D /k a - m m i is the average diffusion length of axonal receptors before internalization.

- j1 is the flux of receptors following route 1 (see expression below) ? - j 2 is the rescaled flux of receptors following route 2 (see expression below)

- je is the flux of receptors entering the axon membrane from the soma membrane through the diffusion barrier (see expression below)

- jout is the flux of receptors internalized in the axon (see expression below) Model dynamics In the soma, we assume that "fresh" receptors , which actually correspond to receptors that have been recycled after internalization from the soma membrane with rate or from the axon membrane with rate , reach the soma membrane

( ) with rate or the axon through the slow pathway 1 ( ) with rate . In turn, receptors of the soma membrane

( ) are internalized at rate and go to the axon through the fast pathway 2 ( ), or are recycled to the soma with rate

. The corresponding dynamics reads:

(1)

(2)

Axonal receptors (resp. ) are assumed to follow a transport pathway 1 (resp.2) characterized by a velocity (resp.

). They pass to the axon membrane with rate (resp. ), which defines the transport length (resp.

). This dynamics reads:

(3)

In turn, receptors on the axon membrane diffuse with diffusion coefficient and are internalized at rate , which defines a diffusion length . This dynamics reads:

(4) These dynamical equations are completed by boundary conditions which ensure the conservation of the total amount of receptors. At the soma/axon boundary (x = 0) it reads: (5) and (6)

At the axon tip the conservation reads: . (7) Steady state distribution Previous equations are solved at steady state, which yields explicit expression for , the distribution of membrane axonal receptors and the total number of internalized receptor . The total number of axonal receptors on the membrane is straightforwardly deduced. The corresponding populations of somatodendritic surface, axonal surface, and internalized receptors are plotted in Figure 5B for values of the parameters obtained by the fitting s s procedure detailed below, which gives km /k1  7.5 and kr /ki  0.5 (other parameters are determined below). Previous equations also give access to the profile along the axon, which reads at steady state:

(8) with , , and is the flux of receptors through the diffusion barrier between the soma and axon membrane. We assume that is small compared to the other length scales in the problem, so that the path 1 can be treated as a source at . The flux can be written , and is used here as a fitting parameter. This prediction is compared to the observed fluorescence profiles rescaled by the mean fluorescence along the axon as follows. We use the WT receptor as a reference and fit the observed profile with fitting parameters . Here is obtained directly by conservation of the receptors, which reads . While this could seem poorly constrained, these 4 parameters actually allow us to fit also the fluorescence profiles of mutants T210I-CB1R and T210A-CB1R with no extra fitting. Indeed, the relative value of for the mutants with respect to the WT-CB1R can be deduced from the observed relative quantity of endocytosed receptors in the soma from Figure 3B. We assume that is small compared to the other fluxes to the soma membrane, so that at steady state the quantity of receptors endocytosed in the soma is equal to . The relative value of for the mutants with respect to the WT-CB1R can be deduced by a direct measure of the fluorescence gradient in the range x = 0.100 µm. The relative value of can be deduced from the observed relative quantity of endocytosed receptors in the axon (see Figure 3B). Lastly, the value of is assumed to be unchanged for all species and the relative value of can be deduced from the relative value of the endocytosis rate in the axon obtained from Figures 3B and 1. The corresponding fits are shown in Figure 5C, where the fitting parameters for WT-CB1R are : , , ; . For the SSTR2, the only fitting parameter is . Statistical Analysis All experiments were performed from two to four independent times. All measurements were plotted and analyzed using GraphPad Prism software. The significance of differences between various conditions was calculated using one-way ANOVA with Newman–Keuls or Bonferroni post-tests for computing p estimates or two-way ANOVA. For significance symbols, one symbol means P < 0.05, two means P < 0.01, and three means P < 0.001. For fluorescence measurements of calcium concentrations in cultured hippocampal neurons, the two-tailed Mann-Whitney test was used for comparisons between groups; significant differences are indicated by *. The two-tailed Wilcoxon signed-rank test was used for comparisons within groups (drug vs. PRE); significant differences are indicated by filled symbols. Principal component analysis (PCA) was used to compare the distribution of larger sets of wild-type and mutant CB1 receptors after treatment with various concentrations of agonist or antagonist. The descriptors of the receptors and chimeras were the percentages of the 5 distribution classes, and the principal components were computed on their covariance matrix (Johnson, 2002). The projection was made onto the subspace spanned by the first two principal components, and the percentage of the total variance they account for was systematically computed and shown in parenthesis on the PCA plots. The fact that their sum was close to 100% indicates that only a negligible amount information was lost by using only two components.

References :

Fache, M. P., Moussif, A., Fernandes, F., et al. (2004). Endocytotic elimination and domain-selective tethering constitute a potential mechanism of protein segregation at the axonal initial segment. J. Cell Biol. 166, 571-578. Henle, F., Leemhuis, J., Fischer, C., et al. (2006). Gabapentin-lactam induces dendritic filopodia and motility in cultured hippocampal neurons. J. Pharmacol. Exp. Ther. 319, 181-191. Jiang, M., and Chen, G. (2006). High Ca2+-phosphate transfection efficiency in low-density neuronal cultures. Nat. Protoc. 1, 695-700. Johnson, R. A., Wichern, D.W. (2002). Applied Multivariate Statistical Analysis Upper Saddle : Prentice Hall ed. Konzack, S., Thies, E., Marx, A., et al. (2007). Swimming against the tide: mobility of the microtubule-associated protein tau in neurons. J. Neurosci. 27, 9916-9927. Lelouvier, B., Tamagno, G., Kaindl, A. M., et al. (2008). Dynamics of somatostatin type 2A receptor cargoes in living hippocampal neurons. J.Neurosci. 28, 4336-4349. Leterrier, C., Bonnard, D., Carrel, D., et al. (2004). Constitutive endocytic cycle of the CB1 cannabinoid receptor. J. Biol. Chem. 279, 36013-36021. Leterrier, C., Laine, J., Darmon, M., et al. (2006). Constitutive activation drives compartment-selective endocytosis and axonal targeting of type 1 cannabinoid receptors. J. Neurosci. 26, 3141-3153. Meijering, E., Jacob, M., Sarria, J. C., et al. (2004). Design and validation of a tool for neurite tracing and analysis in fluorescence microscopy images. Cytometry A. 58, 167-176. Wisco, D., Anderson, E. D., Chang, M. C., et al. (2003). Uncovering multiple axonal targeting pathways in hippocampal neurons. J. Cell Biol. 162, 1317-1328.

3. Main results and remaining questions In this study, we demonstrated an unexpectedly high activation-dependent plasticity of polarized GPCR distribution and we characterized several important points of neuronal CB1R trafficking and targeting. First, using microfluidic devices, we showed that CB1Rs transiently present on the somatodendritic plasma membrane are constitutively endocytosed and transported to the axon through transcytosis. The decrease of basal CB1R activity leads to an increase of receptors on the somatodendritic plasma membrane and to a decrease of axonal receptors. On the contrary, increasing basal activity conducts to the inverse effect. Thus, constitutive activity is essential to obtain correct, i.e. axonal, CB1R polarization. Secondly, prolonged pharmacological treatment modifies neuronal GPCR distribution. Notably, the somatostatin SST2 receptors are predominantly present on somatodendritic plasma membrane and absent from axon in mature neurons. Their prolonged activation leads to their internalization and transcytosis to the axon, reversing their polarized distribution. Finally, measure of both plasma membrane levels and endocytosis rates demonstrated that CB1Rs are mostly internalized in the somatodendritic compartment, while they are predominantly localized to the plasma membrane in the axon. This difference in basal internalization may contribute to differential pharmacological properties between dendrites and axon. Consequently, we asked several questions: - Are CB1Rs, transiently present on the somatodendritic plasma membrane, activable by exogenous cannabinoids and do they recruit signaling pathways in this compartment? - Is constitutive somatodendritic endocytosis related to a constitutive modulation of signaling pathways? - To what extend endocannabinoids contribute to constitutive somatodendritic CB1R activation? - Is the reduced axonal endocytosis linked to different CB1R pharmacology in this compartment, as compared to dendrites?

During my PhD, I could obtain some answers to these questions. In the next part, I will present my results concerning the influence of neuronal polarity on the pharmacology of CB1R. I will first introduce some advanced pharmacological concepts useful to understand

84 some of our unexpected results. Then, I will present the technique we developed and used to answer the above questions.

Part 2: Influence of neuronal polarity on the pharmacology of type-1 cannabinoid receptor.

86 1. Emerging concepts of GPCR pharmacology: allosteric modulation, signaling bias and modulation by membrane lipids

1.1. Allosteric modulation

1.1.1. General points As described above, GPCRs display several binding sites. The binding site for the endogenous ligand is called the orthosteric binding site. Competitive agonists, antagonists and inverse agonists may also bind on this site. Interestingly, some ligands are able to bind GPCR on a site different from the orthosteric site and do not induce signaling changes downstream of the bound-receptor. However, binding of these ligands can modify orthosteric binding site conformation and thus change the affinity and/or the efficiency of agonist binding at orthosteric site (Wootten et al, 2013). These ligands, called allosteric modulators, bind to allosteric binding sites. Three types of allosteric modulators exist (Figure 24): - Positive allosteric modulators (PAMs): their binding enhances the affinity and/or the efficiency of agonist binding at orthosteric site. - Negative allosteric modulators (NAMs): their binding decreases the affinity and/or the efficiency of agonist binding at orthosteric site. - Neutral allosteric ligand (NALs): their binding does not affect the orthosteric binding site.

Figure 24: Allosteric modulators (Wootten et al, 2013)

Allosteric modulation is becoming increasingly important in pharmacology and drug design. Indeed, allosteric modulators are generally more specific to a particular GPCR when compared to classical ligands and can offer a finer modulation of receptor activity.

87 1.1.2. Signaling bias Another emerging concept in GPCR pharmacology addresses to the multitude of active conformations that a GPCR can adopt. Why should two different active states activate the same signaling pathway? (Figure 25) This theoretical prediction has been verified by experiments demonstrating that activation of a GPCR can lead to differential modulation of signaling pathways modulation, contingent of the agonist and/or the GPCR environment. Depending on authors, this concept is called signaling bias, protean agonism, functional selectivity, ligand bias or biased agonism. For example, while angiotensin type 1A receptor activation with angiotensin II activate Gq/11 proteins, its activation by an analogue molecule fails to recruit G proteins but activate ERK1/2 through β-arrestin recruitment (Wei et al, 2003). Also, ICI118551 and propanolol had been defined as inverse agonists of β2-adrenergic receptors due to their action on Gs proteins. This definition seems incorrect since these molecules induce partial agonist responses for the activation of ERK1/2 through a G-protein independent pathway involving β-arrestin activation (Azzi et al, 2003). These studies demonstrate the existence of several active conformations of GPCRs and strongly suggest that each conformation has a specific effect on downstream signaling pathway recruitment.

Figure 25: Signaling bias and GPCRs multiple active states.

(A) If we consider a GPCR as a two- state receptor, one signaling pathway is favoured due only to higher strength of signal.

(B) The multiple active states model offers a better understanding of GPCR signaling: each signaling pathway is favoured in function of its interaction to the corresponding active conformation (blue, green, red, orange or yellow). Thus, two agonists of the same GPCR can induce different cellular responses due to the active conformation they favour. (Park, 2012)

88 Interestingly, Kenakin pointed out that extrapolating pharmacological data from one screening system to another one is almost impossible due to high dependence of GPCR pharmacology on tissue environment (Kenakin, 2009). Thus, local context can strongly modify GPCR signaling. Notably, differences in membrane composition could change the pharmacology of these membrane-bound proteins by changing the conformations they can adopt. Consequently, since neurons are polarized cells displaying different membrane composition between the somatodendritic compartment and the axon, one can assume that neuronal GPCR pharmacology differs between soma, dendrites and axon. In the next paragraph, I will detail how lipidic membrane composition may change GPCR pharmacology and how neuronal CB1Rs could be impacted by these effects.

1.2. Lipids and GPCR activity

1.2.1. Plasma membrane composition Cellular plasma membranes are composed of both proteins and lipids. Their structures enable both compartmentalization of the cell within the extracellular matrix and communication with this environment. In 1972, Singer and Nicolson proposed a model for the cellular membrane, called the Singer-Nicolson model. In this model, the lipid bilayer is a neutral two-dimensional fluid in which proteins freely diffuse (Singer & Nicolson, 1972). However, later studies demonstrated that membranes are heterogenous due to high local protein concentrations (Phillips et al, 2009) and presence of various microdomains (Engelman, 2005). Notably, lipid raft microdomains are membrane regions enriched in cholesterol and sphingolipids that sequester proteins (Simons & Ikonen, 1997). Interestingly, several evidences suggest that lipid rafts spatially and temporally modulate activity of membrane proteins, such as GPCRs (Allen et al, 2007). For example, in bovine hippocampal neurons, cholesterol depletion

(associated to raft disruption) decreases agonist binding and G-protein coupling to 5-HT1A receptors (Pucadyil & Chattopadhyay, 2004). Here, raft regions enable association of three required partners for signaling: agonist, GPCR and G protein (Figure 26A). On the contrary, in CHO cells, κ-opioid receptors are localized in lipid rafts and cholesterol depletion strongly increases agonist binding (Xu et al, 2006). Here, GPCR association with lipid rafts prevents form maximal receptor activation (Figure 26B). However, the existence and physiological role of lipid raft domains remain controversial. Indeed, most results were obtained by indirect measurements (membrane extracts or

89 cholesterol depletion), notably due to the small size of these microdomains (about 10 nm) that is below optical microscopy resolution (Allen et al, 2007). Thus, their in vivo relevance has still never been clearly demonstrated.

Figure 26: Role of lipid raft domains in GPCR signaling (Allen et al, 2007)

1.2.2. Modulation of GPCRs activity by membrane lipids As described above, membrane composition, such as cholesterol-enriched raft domains, can modulate GPCR activity. Thus, several studies investigated the role of membrane lipids in GPCR activity. Because cholesterol constitutes about 25% of total lipidic bilayer (van Meer & de Kroon, 2011) and its depletion can be obtained easily using methyl-β-cyclodextrin, most studies focused on GPCR activity modulation by cholesterol.

Cholesterol depletion decreases agonist binding of the 5-HT1A receptors (Pucadyil & Chattopadhyay, 2004), increases ligand binding of κ-opioid receptors (Xu et al, 2006), decreases adenylyl cyclase mediated signaling downstream of activated μ-opioid receptor (Levitt et al, 2009) and decreases ligand binding of both oxytocin receptors and cholecystokinin (CCK) receptors (Gimpl et al, 1997). How does cholesterol modulate GPCR pharmacology? Gimpl and colleagues demonstrated the existence of two distinct mechanisms. Cholesterol enables CCK receptor activation by changing the membrane fluidity (experiments conducted with other steroids had the same effect), while its effect on oxytocin receptors is due to a highly specific molecular interaction (Gimpl et al, 1997). Moreover, cristallography studies reported that β2-adrenergic receptors (β2AR) can bind two molecules of cholesterol (Hanson et al, 2008). Moreover, existence of a cholesterol recognition/interaction amino acid consensus (CRAC) has been found in proteins interacting with cholesterol. CRAC motif was

90 identified in several GPCRs, such as 5-HT1A receptor, β2AR and rhodopsin receptor (Jafurulla et al, 2011). Interestingly, lipids other than cholesterol can also modify GPCR pharmacology. For example, after light activation, rhodopsin receptors adopt an equilibrium between inactive metarhodopsin I (MI) and metarhodopsin II (MII) conformations. Interestingly, membranes containing phospholipids with phosphatidylethanolamine headgroups (PE) or with docosahexanoyl acyl (DHA) shift the equilibrium to MII conformation, due to changes in membrane elastic properties (Soubias et al, 2010). Taken together, all these results suggest that lipid membrane composition plays a major role in GPCR pharmacology. Interestingly, some GPCRs are known to predominantly bind membrane lipids: lysophosphatidic acid (LPA) receptor (LPAR), sphingosine-1-phosphate (S1P) receptor (S1PR) and CB1R. Because LPAR and S1PR pharmacology is better understood than CB1R pharmacology, I will present in the next part these two receptors in order to later establish some analogies with CB1R.

1.2.3. Lysophosphatidic acid receptor LPAR and sphingolipid receptor S1PR Lysophospholipids are membrane lipids, including LPA and S1P. These latter have been widely studied for their signaling through GPCRs. To date, eleven GPCRs, either binding

LPA (LPAR1-6) or S1P (S1PR1-5) have been reported (Choi & Chun, 2013). Depending on their subtype and tissue localization, LPARs and S1PRs can couple to various G proteins

(G12/13, Gq/11, Gi/o or Gs) (Figure 27). Thus, they are implicated in various physiological functions. In the CNS, they are involved in brain development and many studies reported their implication in various diseases, such as multiple sclerosis, neuropathic pain, brain ischemic stroke, and schizophrenia (reviewed in (Choi & Chun, 2013)). Notably, the pro-drug FTY720 (also called fingolimod) has been approved by FDA in 2010 as a treatment of relapsing- releasing multiple sclerosis (Brinkmann et al, 2010). FTY720 is phosphorylated by sphingosine kinase to produce FTY720P that binds S1PRs leading to inhibition of lymphocyte recirculation (Chun & Hartung, 2010).

Importantly, the X-ray-crystal structure of S1PR1 has been established (Hanson et al, 2012). It is the first (and to date, the only) GPCR with lipid ligands whose structure has been described. The structure reveals notably a closed extracellular domain, avoiding binding of extracellular ligands. Moreover, a gap exists between TM1 and TM7 in membrane bilayer through which a lipid ligand can reach the binding pocket. Thus, S1P should diffuse in the two-dimensional membrane bilayer to bind its receptor. This model is coherent with previous suggestions

91 excluding the unlikely extracellular release of lipid ligands. However, a recent study also demonstrated that Nogo-A-Δ20 directly binds S1PR2 inhibiting neurite outgrowth and cell spreading through G13 and RhoA activation (Kempf et al, 2014). This binding requires the interaction of Nogo-A-Δ20 with EC2 and EC3 of S1PR2 suggesting another binding mechanism than the one of S1P. Interestingly, inhibition of S1P synthesis does not affect the effect of Nogo-A-Δ20 but addition of S1P increases its effect. Thus, these data suggest positive allosteric effects between membrane lipid ligand S1P and extracellular Nogo-A-Δ20 peptide.

Figure 27: LPAR and S1PR subtypes and signaling (Choi & Chun, 2013)

1.2.4. Endocannabinoid biochemistry As presented above (Part 1), several endocannabinoids (eCBs) have been reported to date. Among them, AEA and 2-AG, both derived from arachidonic acid (Figure 28), have been

92 implicated in various physiological processes and appear as major eCBs. In this paragraph, I will first describe synthesis and degradation pathways of these molecules. Then, as 2-AG levels are 170-fold higher than of AEA in the CNS (Stella et al, 1997) and as my results demonstrate (see below) high implication of 2-AG in the pharmacology of neuronal CB1Rs, I will focus on this molecule.

1.2.4.1. AEA AEA belongs to the N-acyl-ethanolamines (NAEs) family. Its precursor, the N-acyl- phosphatidyethanolamines (NAPEs) is hydrolyzed by a phospholipase D selective for NAPEs (NAPE-PLD). Its degradation is due to hydrolysis realized by the enzyme FAAH producing arachidonic acid and ethanolamine (Bisogno et al, 2005)(Figure 28).

1.2.4.2. 2-AG 2-AG is produced by hydrolysis of diacylglycerols (DAGs). This reaction is catalysed by DAG Lipases (DAGLs) that are specific for sn-1 position. Two isoforms of DAGLs have been cloned: DAGLα and DAGLβ (Bisogno et al, 2003). For degradation, 2-AG can be hydrolysed by FAAH, such as AEA, producing arachidonic acid and glycerol. However, in FAAH knockout mice, no increase of 2-AG level was reported (Lichtman et al, 2002). Thus, another pathway has been identified: hydrolysis by monoglycerol lipases (MAGLs) (Bisogno et al, 2005) (Figure 28). Recently, the serine hydrolase α-β-hydrolase domain 6 (ABHD6) has been identified as potent degradation enzyme for 2-AG (Marrs et al, 2010).

Figure 28: Main pathways of eCB synthesis and degradation. (Francischetti & de Abreu, 2006)

Among the two cloned DAGLs, DAGLα produces about 80% of the 2-AG in the brain and in DAGLα knock-out mice, all forms of eCB signaling (DSI, DSE and eCB-mediated-LTD) were removed. Moreover, DAGLβ knock-out mice did not display this effect (Gao et al,

93 2010; Tanimura et al, 2010). Thus, DAGLα seems the more important enzyme for 2-AG production and physiological effects. Consequently, I focused my study on DAGLα (see below). DAGLα is localized in the somatodendritic plasma membrane of mature neurons in several brain areas, such as the cerebellum (Bisogno et al, 2003), striatum (Uchigashima et al, 2007), hippocampus (Katona et al, 2006; Yoshida et al, 2006) and amygdala (Yoshida et al, 2011). This localization fits with post-synaptic production of 2-AG, necessary for retrograde signaling. Interestingly, MAGL is found in cytoplasm of pre-synaptic compartment (Yoshida et al, 2011), enabling 2-AG degradation after retrograde signaling. However, how 2-AG, a lipidic molecule, is able to cross the synaptic cleft is still unclear. Indeed, several studies demonstrated the existence of a transporter molecule that could enable 2-AG to be released from postsynaptic compartment and uptake in the presynaptic compartment but identification of such a transporter is still lacking (Hermann et al, 2006).

1.2.4.3. 2-AG basal tone: a non-classical neuromodulator Many studies reported an increase of 2-AG level after synaptic activation, due to postsynaptic calcium increase that activates DAGLα. These results suggest that 2-AG is synthesized on demand (Marsicano et al, 2003; Piomelli, 2003). Unlike classical retrograde signaling molecules, the lipidic nature of 2-AG precludes its storage in membrane-limited vesicles and complicates its release into the aqueous extracellular medium. Indeed, on the one hand, using in vivo microdialysis in nucleus accumbens, Caillé and colleagues reported that 2-AG concentration in extracellular medium is about 5 nmol/L (Caille et al, 2007). On the other hand, Stella and colleagues measured 2-AG levels in brain lipid extracts and found a concentration of 2.5 nmol per gram of brain tissues, making 2.5 µmol/L by extrapolating brain tissue at density 1kg/L (Stella et al, 1997). Thus, 2- AG seems to be 500-fold more concentrated in cell membranes than in extracellular medium. In addition, since 2-AG is a precursor for arachidonic acid, it is used in several other physiological functions than eCB signaling and DAGLα knock-out removes 80% of basal 2- AG. Also the presence of membrane-bound ABHD6 in the somatodendritic compartment indicates physiological relevance of controlling post-synaptic 2-AG level. Taken together, these date suggest that high 2-AG basal tone is maintained in neurons due to high basal DAGLα activity (Figure 29). This basal tone could explain why CB1R constitutive activity has been measured: 2-AG is continuously present in membrane, activating CB1Rs.

Figure 29: 2-AG basal tone. In resting neurons, data suggest that there is a basal 2-AG

tone, due to basal DAGLα activity. A little amount of this 2-AG pool can be released in extracellular medium. Neuronal stimulation increases calcium level in postsynaptic compartment leading to extra-activation of DAGLα and additional 2-AG synthesis that

is released for synaptic inhibition. (Alger & Kim, 2011)

1.2.5. Are endocannabinoids genuine CB1R ligands or lipidic modulators? Relation to constitutive CB1R activity. eCBs are non-classical agonists due to their lipidic structure. How do they bind CB1R? Plasma membrane localization of DAGLα favours the accumulation of the lipid 2-AG in the membrane, that could reach CB1Rs through two-dimensional diffusion. Using analogy with

S1PR1, Hurst proposed that 2-AG accesses the CB2R binding pocket through entry between TM6 and TM7 (Hurst et al, 2013). As CB1R amino acid sequence is very close to that of CB2R, we can assume that their structure, binding sites and agonist entry mode are analogous. Moreover, Sugiura and colleagues have tested cellular responses to 2-AG and several structural analogues, including 1-AG and 3-AG. This study demonstrates that 2-AG is the most potent to activate CB1R (Sugiura et al, 1999). Taken together, these data strongly suggest that 2-AG is a genuine CB1R agonist. However, the number of identified eCBs, their lipidic structure, ubiquitous distribution and involvement in other physiological pathways suggest that eCBs could modulate the conformation of plasma membrane proteins through changes in the lipidic membrane environment, such as cholesterol modulates the activity of several GPCRs. For example, AEA activates not only CB1R and CB2R but also the transient receptor potential vanilloid type-1 (TRPV1), present in neurons at postsynaptic sites. Thus, it is possible that AEA modulates synaptic transmission acting as anterograde or autocrine mediator (Di Marzo & De

Petrocellis, 2012). On the other hand, 2-AG was recently shown to potentiate GABAA activation by GABA (Sigel et al, 2011). This positive allosteric modulation is enabled by binding of 2-AG on GABAA receptors β2-subunit (Baur et al, 2013). Notably, double knock- out mice for CB1R and CB2R display reduced motility, while β2-knock-out mice display hypermotility. Moreover, in pancreatic β-cells, 2-AG blocks KATP channels through a CB1R

95 independent mechanism, given that CB1R-specific inverse agonist AM251 did not remove the effect (Spivak et al, 2012). Interestingly, before the cloning of CB1R and CB2R, several studies reported that cellular action of Δ9-THC is due to its presence in the plasma membrane that alters the physical properties of phospholipid bilayer (Bruggemann & Melchior, 1983; Hillard et al, 1985; Hillard et al, 1990). Taken together, these data suggest that eCBs are non- classical agonists, and as it is still unclear how they bind CB1Rs, it is possible that they modulate CB1R activity through changes in the lipidic membrane environment that enable CB1R activation. This mechanism is maybe possible due to constitutive activity of this receptor. Indeed, as described in Part 1, some structural determinants have been identified for CB1R constitutive activation. The genuine constitutive CB1R activity is difficult to test due to the ubiquitous presence of eCBs. Taken together, my hypothesis is that CB1R is a destabilized GPCR that can be activated only by membrane lipid composition changes and that presence of eCBs enables this change. However, I do not exclude the possibility that a binding site exists for 2-AG. Thus, in my opinion, for CB1R activation, a small amount of 2- AG can bind the receptor, but a big amount of 2-AG and other eCBs favour the activation by changing the lipidic membrane environment. Results that I will present in this part and Part 4 tend to validate this hypothesis.

1.2.6. CB1R allosteric modulation and signaling bias There is evidence that CB1R displays several binding sites for agonists. Indeed, studies have notably shown that WIN55,212-2 binds at a different site as HU-210 and CP-55,940 (Kapur et al, 2007; Song & Bonner, 1996). Moreover, synthetic allosteric modulators of CB1R have been identified, such as Org27569, Org29647 and PSNCBAM-1 (Ross, 2007). Interestingly, these compounds enhance the agonist CP55-940 binding affinity and decrease the binding affinity of the inverse agonist SR141716A. However, they decrease the signaling efficiency of CP55-940. Notably, Org27569 decreases the effect of CP55-940 on cAMP signaling but has no effect on ERK1/2 phosphorylation and displays a reduced effect on WIN55-212,2 signaling efficiency (Baillie et al, 2013). Moreover, alone, Org27569 displays signaling bias: it is able to enhance ERK1/2 phosphorylation and to activate Gs-protein leading to an increase of cAMP concentration. Interestingly, under certain conditions, CB1R agonist HU-210 can also display signaling bias. Indeed, activation of CB1Rs with this agonist in cultured striatal neurons increases cAMP concentration through activation of Gs proteins when pertussis toxin, an inhibitor of Gi/o proteins, is applied (Glass & Felder, 1997). Very recently, a study confirmed that CB1Rs display signaling bias. Indeed, various CB1R agonists (2-AG, AEA, Δ9-THC,

96 WIN55-212,2 and CP55,940) do not lead to the same efficiency of signaling pathway activation (Laprairie et al, 2014). Taken together, CB1R displays several binding sites, some molecules are known as allosteric modulator of this receptor and display signaling bias. These data strongly suggest that, such as other GPCRs, CB1R display several active conformations. Since membrane composition can influence strongly GPCR pharmacology and since neurons display membrane differences between somatodendritic compartment and axon, we asked if CB1R pharmacology is different between dendrites and axon. To do so, we proposed to measure the recruitment of cAMP/PKA pathway downstream of endogenous CB1Rs in cultured neurons of rat hippocampi with a technique enabling sub-neuronal resolution. In the next paragraphs, I will present the cAMP/PKA pathway and the technique I developed in order to answer this question.

97 2. Measure of cAMP/PKA signaling pathway in live neurons at the subcellular scale

2.1. cAMP/PKA pathway downstream of G proteins

2.1.1. cAMP structure and signaling 3’-5’-cyclic adenosine monophosphate (cAMP) is a cellular second messenger widely studied since its discovery in 1958 (Rall & Sutherland, 1958). It is synthesized from ATP in adenylyl cyclase catalytic pocket (Figure 30). cAMP can then bind cyclic-nucleotide-gated ion channels (CNG) (Matulef & Zagotta, 2003), guanine-nucleotide exchange proteins activated by cAMP (Epac) (Kawasaki et al, 1998) and cAMP-dependent protein kinase A (PKA) (Walsh et al, 1968). cAMP signaling is regulated by activation of phosphodiesterases (PDEs) that arrest its signaling by degrading cAMP (Manganiello & Degerman, 1999).

Figure 30: cAMP synthesis. cAMP is known to regulate several cellular processes, such as gene transcription (Yamamoto et al, 1988), cell migration (Burdyga et al, 2013) and proliferation (Stork & Schmitt, 2002). In neurons, it is notably involved in LTP through activation of PKA (Malenka & Bear, 2004) leading to long-term memory formation (see Annexe 1) but also in neurotransmitter release through activation of Epac (Kaneko & Takahashi, 2004). The variety of cAMP implication in the modulation of cellular processes requires a strong spatial and temporal regulation of cAMP concentrations that leads to subcellular microcompartmentalization of cAMP signaling. The temporal regulation is ensured, notably, by the presence of PDEs (see review (Lefkimmiatis & Zaccolo, 2014)). The spatial compartmentalization is ensured by various processes, including presence of A-kinase anchor proteins (AKAPs) that control spatial PKA organization. I will give more details on these proteins in the next paragraph, after presentation of PKA structure and signaling.

98 2.1.2. PKA structure and signaling 2.1.2.1. Structure and activation PKA is a tetrameric holoenzyme, consisting of two catalytic subunits and two regulatory subunits. In absence of cAMP, these four subunits are bound. Regulatory subunits contain a N-terminal dimerization domain, an interaction domain for one catalytic subunit in C-terminal tail and two cooperative cAMP binding sites, called A and B. Site B has high affinity for cAMP, while site A displays low affinity. When PKA is inactive, cAMP can only binds site B, leading to conformational changes increasing cAMP affinity for site A. When all binding sites are occupied, both regulatory subunits stay bound together with four cAMP molecules, while active catalytic subunits are released (Figure 31). Then, catalytic subunits phosphorylate specific serine and threonine residues of substrate proteins, such as S133 of cAMP-Response Element-binding protein (CREB) (Taylor et al, 1990). PKA activation is regulated by an endogenous inhibitor, called PKI (Scott et al, 1985; Walsh et al, 1971). PKI displays a high affinity for PKA catalytic subunits (Kd=0.2 nM in presence of ATP) and binds the catalytic site, acting as a competitive inhibitor. Moreover, PKI binding increases the kinetics of the nuclear export of the catalytic subunit, decreasing PKA effects on gene transcription (Wen et al, 1994; Wen & Taylor, 1994). Moreover, dephosphorylation of PKA- targeted proteins occur rapidly through phosphatases activity (Heijman et al, 2013).

Figure 31: PKA structure and activation. (R): regulatory subunit (C): catalytic subunit (Skalhegg & Tasken, 2000)

2.1.2.2. PKA isoforms Initially, two PKA isoforms had been identified (PKAI and PKAII) from their pattern of elution from DEAE-cellulose columns (Corbin et al, 1975; Reimann et al, 1971). Following studies demonstrated that these two PKA isoforms contained two different regulatory subtypes (RI and RII). Later, molecular biology techniques reported numerous heterogeneous isoforms for both regulatory and catalytic subunits. Thus, to date, RI subunits can be separated in two groups: RIα and RIβ, as well as RII subunits: RIIα and RIIβ. RI and RII differ by their cAMP affinity: in vitro, RI displays an

99 activation constant (Ka) of 100nM, while RII Ka equals about 400nM (Ishikawa & Homcy, 1997; Taylor et al, 2004). Three types of catalytic subunits have been identified to date: Cα, Cβ and Cγ. Splice variants of Cα (Cα1, Cα2 and Cα-s) have been reported. While all catalytic Cα subunits are ubiquitous, Cβ subunits are expressed in specific tissues, such as brain and Cγ has been found in testis.

2.1.2.3. Sub-cellular localization cAMP is produced on the plasma membrane by ACs but low level of cAMP is found in the cytosol due to the high amount of PDEs (Neves et al, 2008). Thus, PKA is mainly activated at the plasma membrane and translocate then to the nucleus in order to activate CREB. Interestingly, PKA can be targeted to specific sub-cellular domain through its binding with AKAPs. More than 50 AKAPs have been identified to date. AKAPs contain a binding site for the regulatory subunit of PKA, a targeting domain and an association with various proteins generally involved in signaling pathways. Microtubule associated protein 2 (MAP-2) has been the first identified AKAP. In mature neurons, MAP-2 is segregated in the somatodendritic compartment. Thus, binding PKA, AKAPs address it into a specific sub-cellular compartment. This targeting enables spatial regulation of cAMP/PKA pathway signaling (Scott et al, 2013).

2.1.2.4. Role of neuronal PKA Studies on hippocampal cultured neurons revealed that cAMP/PKA pathway is involved in the formation of synapses. Indeed, during neuronal development, a decrease of cAMP concentration leads to a decrease of synaptic clusters (Kavalali et al, 1999) and in mature neurons, increasing cAMP concentration induces an increase of active presynaptic terminals after protein synthesis (Ma et al, 1999). Moreover, PKA is involved in the formation of memory. In order to form long-term memory, activation of PKA in axon terminals should translocate to the nucleus in order to induce gene transcription and protein synthesis. Thus, one can assume that retrograde cAMP/PKA signaling occurs in neurons. Using Drosophila Melanogaster neuronal culture developed by Anne-Lise Gaffuri and live FRET imaging that I set up in the laboratory, we confirmed this retrograde signaling that occurs through retrograde cAMP diffusion (see Annexe 1).

100 2.2. Förster Resonance Energy Transfer (FRET) imaging

2.2.1. Principle and history

2.2.1.1. Fluorescence Fluorescence is a light emission due to deexcitation of a molecule, called fluorophore. At basal state, flurophore energy corresponds to its ground state (S0). It can absorb photons of a specific wavelength, contained in its absorption spectrum, and pass to an excited state (S1), corresponding to elevated energy (Figure 32a). Deexcitation enables the molecule to return to

S0 through an energy emission, corresponding to photon emission at a specific wavelength, contained in its emission spectrum. The latter process is called flurorescence. As shown in Figure 32a, shortly after absorption and before emission, relaxation occurs, leading to a decrease of energy. Thus, for a given fluorophore, emission energy is always lower than absorption energy and, as wavelength is inversely proportional to energy, emission wavelength is higher than absorption wavelength.

2.2.1.2. FRET According to Förster theory, FRET is defined as a non radiative energy transfer resulting from a dipole-dipole interaction between two flurorophores: one donor and one acceptor. Thus, deexcitation of the donor, instead of photon emission, corresponds to energy transfer to the acceptor, with a corresponding excitation wavelength. Then, acceptor deexcitation occurs with photon emission at its own wavelength (Figure 32a). Thus, emission spectrum of the donor has to overlap with the excitation spectrum of the acceptor (Figure 32b). Moreover, FRET efficiency is given by the following equation:

Equation 1: FRET efficiency R: distance between fluorophores R0: Förster distance of a given fluorophore pair

R0 corresponds to the distance between fluorophores enabling 50% of maximum FRET efficiency and depends on overlap integral of the donor emission spectrum with the acceptor absorption spectrum and their mutual molecular orientation. Thus, FRET efficiency increases when spectrum overlap is higher (Figure 32b) and when fluorophores are parallel (Figure 32d). Moreover, efficiency decreases rapidly when the distance between both fluorophores

101 becomes superior to R0. As R0 is generally comprised between 1 and 10 nm, FRET is efficient when two fluorophores are separated by a maximum of 10 nm (Figure 32c).

Figure 32: Conditions to obtain efficient FRET. (Broussard et al, 2013) (a) Jablonski diagram of FRET (see text). (b) Spectral overlap between donor emission spectrum and acceptor excitation spectrum. (c) Distance between the two fluorophores should not exceed 10 nm. (d) Fluorophores orientation: FRET occurs only between parallel dipoles.

The development of Green Fluorescent Protein (GFP) mutants has resulted in several fluorophores with various absorption/emission spectra. For example, the Cyan Fluorescent Protein (CFP) absorbs photons around 405 nm and emits around 477 nm, corresponding to blue color. The Yellow Fluorescent Protein (YFP) absorbs photons around 514 nm and emits around 528 nm, corresponding to yellow color. YFP has been improved by one-point mutations in order to reduce chloride sensitivity, faster maturation, and to increase brightness, leading to new fluorophores called Citrine, Venus, and Ypet. The pair CFP/YFP has been widely used to construct FRET probes, due to the overlap of CFP emission spectrum and YFP absorption spectrum. In these constructions, CFP act as a donor and YFP as an acceptor. Thus, to perform FRET imaging with these fluorophores, live cells are illuminated at 405 nm (CFP absorption) and both 477 nm (CFP emission) and 528 nm (YFP emission) channels are recorded. The FRET ratio is defined in Equation 2. Using this formula, FRET ratio increases when FRET efficiency increases.

YFPemission FRETratio  Equation 2: Definition of FRET Ratio. CFPemission

102

Using the strongly increased FRET efficiency when fluorophores are close, genetically encoded FRET biosensors have been developed to detect interesting modulation of signaling pathways, such as cAMP concentration and PKA activity. I will detail this development in the two next sections.

2.2.2. Detection of cAMP concentration by FRET

2.2.2.1. PKA-based cAMP FRET biosensors The first cAMP FRET biosensor, called FlCRhR, consisted in PKA catalytic subunit labelled with fluorescein and PKA regulatory subunit labelled with rhodamine (Figure 33) (Adams et al, 1991). When PKA is not activated, energy transfer can occur between fluorescein (donor) and rhodamine (acceptor). When cAMP binds the regulatory subunit, PKA activation leads to dissociation between regulatory and catalytic subunits and separation of both fluorophores. Thus, increase of cAMP concentration leads to a decrease of FRET with this probe. Using this probe in Aplysia sensory neurons, Bacskai and colleagues demonstrated that bath application of serotonin induces stronger cAMP elevation in neurites than in cell body. Moreover, perinuclear increase of cAMP concentration leads to translocation of PKA catalytic subunit in the nucleus (Bacskai et al, 1993). However, this chemical biosensor was difficult to use in vertebrate neurons. Thus, the later development of genetically encoded probes enabled to express cAMP biosensors in these cells.

Figure 33: FlCRhR probe (Adams et al, 1991).

Zaccolo and colleagues developed a genetically encoded FRET biosensor to detect cAMP concentration in living cells (Zaccolo et al, 2000). To perform that, they fused RII mouse PKA subunit with an improved GFP mutant and catalytic PKA subunit with a blue GFP

103 mutant, called enhanced Blue Fluorescent Protein (EBFP). BFP acts as a donor and GFP as an acceptor. This probe reports cAMP concentration the same way as FlCRhR: FRET ratio decreases when cAMP concentration increases. Finally, Nikolaev and colleagues developed cAMP FRET biosensors that contain both fluorophores on the same construction (Nikolaev et al, 2004). Notably, they constructed the PKA-camps probe, where CFP and YFP are linked by PKA RIIβ subunit. They also constructed cAMP reporters based on Epac.

2.2.2.2. Epac-based cAMP FRET biosensors Epac is a guanine nucleotide exchange factor for Rap1 and is activated by direct binding of cAMP. Its recent discovery showed a new signaling pathway activated by cAMP, different from PKA activation. Two types of Epac proteins have been identified and showed very similar structures. Epac1 contains four parts: a cAMP-binding domain, a Guanine nucleotide Exchange Factor (GEF) that activates Rap1, a Dishevelled, Egl-10, Pleckstrin domain (DEP), involved in membrane localization, and Ras-Exchanger Motif (REM) that interacts with GEF (Figure 34). Epac2 has the same structure but contains a second cAMP binding domain at its N-terminus, called binding domain A, while shared binding domain with Epac is called binding domain B. Interestingly, cAMP binding to the cAMP-binding domain induces Epac conformation changes. Notably, Val-Leu-Val-Leu-Glu sequence (where Val is valine, Leu is leucine and Glu is glutamate), localized between cAMP-binding domain and REM, is important for conformational change of Epac. This conformal change increases the distance between N- and C-termini (Figure 34).

Figure 34: Epac1 structure and activation by cAMP binding. (Bos, 2003) cAMP binding induces Epac conformational changes.

GEF: Guanine nucleotide Exchange Factor; REM : Ras-Exchanger motif; DEP: Dishevelled, Egl-10, Pleckstrin domain; VLVLE : Val-Leu-Val-Leu-Glu sequence (where Val is valine, Leu is leucine and Glu is glutamate)

104 Taking advantage of this conformational change after cAMP binding, Nikolaev and colleagues constructed genetically encoded FRET biosensors that detect cAMP concentration (Nikolaev et al, 2004). Indeed, they fused CFP and YFP on N- and C-termini, respectively, of either Epac1 (Epac1-camps sensor) or truncated Epac2, using only binding domain B (Epac2- camps sensor). Thus, with these constructions, cAMP concentration increase leads to a FRET ratio decrease. Many other Epac-based cAMP biosensors have been developed using these two constructions and improving either cAMP binding efficiency or flurophore resistance to photobleaching or temperature and pH sensitivity. Notably, the group of Dr Kees Jalink developed CFP- Epac(ΔDEP-CD)-YFP probe that contains truncated Epac1 protein: the DEP domain was removed in order to obtain a cytosolic probe and two mutations were introduced (T781A and F782A) to avoid Rap1 activation by the probe (Ponsioen et al, 2004). Recently, this group also constructed TEpacVV probe containing truncated Epac(ΔDEP-CD) fused with a very effective CFP mutant, called mTurquoise, to replace CFP and two Venus proteins to replace YFP (Klarenbeek et al, 2011). This improvement has strongly enhanced probe sensitivity to cAMP small concentration changes. During my PhD, I used TEpacVV probe to measure cAMP concentration changes downstream of endogenous CB1Rs in cultured hippocampal neurons of rats (see below).

2.2.3. Detection of PKA activity by FRET In 2001, Zhang and colleagues published the first A-Kinase Activity Reporter (AKAR1) probe (Zhang et al, 2001). This genetically encoded probe contains an ECFP and a YFP linked by a phosphoaminoacid binding domain (14-3-3τ) and a kemptide, synthetic sequence specifically phosphorylated by PKA (LRRASLP). When phosphorylated by PKA, phospho-serine (pS) binds to 14-3-3τ leading to conformational changes of the probe and the fluorophores move closer. Thus, an increase of PKA activity is detected by an increase of the FRET ratio (Figure 35A, 35B). However, pS binding to 14-3-3τ is so strong that cellular phosphatises are unable to dephosphorylate the probe. Thus, the lack of reversibility of AKAR1 has conducted to the development of a new AKAR probe, called AKAR2 (Zhang et al, 2005). In this construction, 14-3-3τ was replaced by a Forkhead associated domain 1 (FHA1). Because FHA1 binds phospho-threonine (pT), kemptide has been modified to replace serine by a threonine (LRRATLVD). This probe shows good reversibility. Moreover, in order to increase FRET efficiency, a 14 amino acid sequence was added between FHA1 domain and kemptide that

105 corresponds to the ideal length for efficient conformational changes. Finally, in this construction, YFP was replaced by a Citrine (Figure 35C). Development of AKAR probes continued with modifications of fluorophores in order to enhance brightness and dynamic range. Thus, AKAR3 contains an ECFP and cpVenus (cp for circulary permuted), a Venus variant that demonstrated high FRET efficiency (Figure 35D) (Allen & Zhang, 2006). For the construction of AKAR4, ECFP was replaced by cerulean (Figure 35E) (Depry et al, 2011). Figure 35F illustrates conformational changes for the AKAR4 probe.

Figure 35: AKAR probes. (A) Conformational changes induced by AKAR1 phosphorylation and dephosphorylation (Zhang et al, 2001). (B) AKAR1 sequence (Zhang et al, 2001). (C) AKAR2 sequence (Zhang et al, 2005)(figure adapted from (Allen & Zhang, 2006)). (D) AKAR3 sequence (Allen & Zhang, 2006). (E) AKAR4 sequence (Depry et al, 2011) (figure adapted from (Allen & Zhang, 2006)). (F) Conformational changes induced by AKAR4 phosphorylation and dephosphorylation (adapted from (Saucerman et al, 2006)).

In order to detect PKA activity in specific sub-cellular domains, targeting motifs have been added on AKAR probes. For example, fusion of Lyn kinase N-terminus with AKAR4 enables AKAR4 to be targeted to raft domains of plasma membrane (Figure 36). This construction is

106 called Lyn-AKAR4. On the other hand, combination of polylysine motif, adapted from K- Ras, and CaaX sequence (C=cysteine, a=aliphatic amino acid and X=any amino acid) anchors proteins in non-raft domains. Thus, fusion of this sequence with AKAR4 led to a PKA reporter targeted to non-raft lipidic domains of the plasma membrane (Figure 36). This probe is called AKAR4-Kras.

Figure 36: Sub-cellular targeting of AKAR4. (Depry et al, 2011)

(A) Sequences of Lyn-AKAR4 and AKAR4-Kras.

(B) Plasma membrane localization of Lyn-AKAR4 and AKAR4-Kras in HEK293 cells. Scale bar: 10µm.

In order to measure PKA activity modulation downstream of CB1Rs, I used AKAR3, AKAR4, Lyn-AKAR4 and mostly AKAR4-Kras. During my master, I began to work with AKAR3 to set up FRET imaging in cultured hippocampal neurons. In the next section, I will present technical developments I performed to finally be able to measure basal cAMP/PKA pathway modulation downstream of endogenous CB1Rs in all neuronal compartments, including thin mature axons.

2.3. Setting up and improvement of FRET imaging to measure modulation of basal cAMP concentration and PKA activity downstream of endogenous CB1Rs

2.3.1. Equipment Alexandre Roland, PhD student in Zsolt Lenkeis team, started to set up live FRET imaging on cultured hippocampal neurons of rats before I arrived in the laboratory. A motorized Nikon Eclipse Ti-E/B inverted microscope with the Perfect Focus System (PFS) in a 37°C thermostated chamber was installed with an oil immersion CFI Plan APO VC 60X,

107 NA 1.4 objective (Nikon) combined with a 1.5x magnification lens. PFS enables to realize multi-field acquisitions. Acquisitions were realized at the excitation wavelength of the CFP (434nm +/- 15nm) using an Intensilight (Nikon). Emitted light passed through an Optosplit II beam-splitter (Cairn Research) equipped with a FF509-FDi01 dichroïc mirror, a FF01-483/32- 25 CFP filter and a FF01-542/27-25 YFP filter and was collected by an EM-CCD camera (Evolve 512, Photometrics), mounted behind a 2x magnification lens. Acquisitions were performed by piloting the set-up with Metamorph 7.7 (Molecular Devices). All filter sets were purchased from Semrock. With this system, it is possible to image several neurons in parallel at subneuronal scale. When I arrived in the laboratory, I set up data analysis and statistical analysis, using cultured cultured hippocampal neurons of rats expressing AKAR3 probe. In the next two paragraphs, I will detail this development.

2.3.2. Data analysis During acquisition, one image is recorded at each time point (each minute or two minutes, depending on the protocol (see below)). Thus, for each stage, a stack is obtained, composed of both CFP and YFP channels on each image. The first data treatment consists in splitting each image of the stack in two parts to obtain individually CFP and YFP channels (Figure 37), using the ImageJ plug-in Cairn Image Splitter.

Figure 37: Channels splitting. CFP and YFP channels are recorded on a same image (1). Thus, this image is split in two to obtain CFP channel on the one hand (2) and YFP channel on the other hand (3). A realignement is applied to obtain channels superposition for each time point (4).

Quantification was performed using a Matlab program developped in collaboration with Pierre-Yves Plaçais (Brain Plasticity Unit, ESPCI-ParisTech). Set up of this program was realized by quantifying data obtained after 50 µM Forskoline injection (Figure 38). This

108 program presents the first YFP channel image of a stack (Figure 38-1). The user defines a background zone and the region where FRET ratio has to be calculated, called Region of Interest (ROI) (Figure 38-2). The program calculates then for each stack image the mean intensity of background region and ROI on YFP channel image and corresponding regions on CFP channel. Then, it reiterates the same operations for each YFP/CFP couple images of the stack, keeping the same regions (background and ROI) (Figure 38-3). Then, it calculates the crude FRET ratio for each time point (Figure 38-4). The user selects the injection time point and defines the baseline (time points before treatment) in order that the program calculate the normalized FRET ratio (Figure 38-5). This normalization enables places all neurons baselines at zero, enabling to remove basal variabilities. This process improves strongly the detection sensitivity. Indeed, as presented below, for detection of basal PKA modulation, FRET ratio variations are in the range of a few percents that could be completely lost in the noise due to basal inter-individual variability.

109

Figure 38: Matlab quantification (1-2) On the first YFP channel image, user defines a background (B) and region of interest (ROI). (3) The program calculates mean intensity of the ROI normalized by the background for each time point on both CFP (dotted line) and YFP channels (full line). Addition of Fsk induces a decrease of CFP intensity due to increase of FRET efficiency. (4) Crude FRET ratio (Rc) is then calculated for each time point, using the equation: IY - BY Rc  IC - BC With : IY : Mean intensity of the ROI in YFP channel BY: Mean intensity of the ROI background in YFP channel IC : Mean intensity of the ROI in CFP channel BC: Mean intensity of the ROI background in CFP channel (5) Normalized FRET ratio (Rn) is calculated: Rb  Ro Rn  100 Ro With : Ro : Mean of crude ratio on baseline points.

However, some lateral movements occur during acquisition and stack images are not always well aligned. Because the program uses the same background regions and ROIs for all images of the stack, it was necessary to obtain well-realigned images. Thus I developed a macro on ImageJ to automate realignment of stacks, based on Stackreg plug-in. This macro is applied before Matlab quantification.

110 2.3.3. Statistical analysis In order to perform statistical analysis, I created two Matlab programs. The first one enables to plot curves of normalized FRET ratio for all neurons treated with the same drug in function of the time (Figure 39-1). For each time point, mean +/- SEM are represented using Graphpad Prism. The second program calculates the mean response to the treatment for each neuron, enabling to plot histograms on Graphpad prims and perform statistical tests (Figure 39-2).

Figure 39: Statistical analysis (1) Mean +/- SEM of normalized FRET ratio of neurons treated with the same drug is calculated for each time point. Here is presented curve of somata treated with Fsk 50µM (N=5). (2) FRET response to the treatment is calculated for each neuron and histogram is plotted on Graphpad prism. Here is presented mean +/- SEM of somata responses to Fsk 50µM (N=5).

2.3.4. Protocols and first results

2.3.4.1. Forskoline EC50 shift in somata induced by activation or blockade of overexpressed CB1Rs.

Generally, measure of the signaling of Gi/o protein-coupled receptors is performed by activation of the GPCR followed by addition of Forskoline (Fsk) that activates directly ACs.

Using this protocol, activation of Gi/o proteins leads to a decreased response to Fsk compared to vehicle. Thus, beginning my master internship, I used cultured hippocampal neurons cotransfected at DIV6 with AKAR3 and wild-type CB1R (WT-CB1R). After selection of neurons (somatodendritic parts), I started the acquisition with one image per minute. Neurons were pretreated 3 hours before starting the acquisition with either CB1R agonist WIN55,212- 2 (WIN) 1µM or inverse agonist AM281 (AM) 1µM. After recording 10 minutes of baseline, Fsk 50µM was added to the medium (Figure 40, Protocol 1). However, using this protocol, treatment with WIN 1µM did not induce any decrease of Fsk-induced response and on the

111 contrary, treatment with AM 1µM did not potentiate Fsk response. One possible explanation could be that the Fsk dose we used (50µM) corresponds to saturating dose, enabling to overcome Gi/o inhibition on ACs.

Figure 40: Protocols. Protocol 1: Treatments with CB1R agonist WIN or inverse agonist AM are applied on neurons 3 hours before aquisition starts. During the acquisition, after recording 10 min baseline, Fsk 50µM is applied. Protocol 2: Treatments with CB1R agonist WIN or inverse agonist AM are applied on neurons 3 hours before aquisition starts. During the acquisition, after recording 10 min baseline, increasing concentration of Fsk are applied each 5 min (100nM, 300nM, 1µM, 3µM, 10 µM and 100µM). Protocol 3: After recording 30 min baseline, treatments with CB1R agonist WIN or inverse agonist AM are applied. 30 min later, Fsk is added (internal positive control). Eventual pretreatment can be applied before acquisition (Pertussis toxin (PTX) blocks Gi/o proteins and is applied overnight before aquisition; Tetrahydrolipstatin (THL) and RHC80267 (RHC) inhibit DAGL activity and are applied 3h before aquisition.)

Thus, I set up a protocol using Fsk dose-response (Figure 40, Protocol 2). Cultured neurons were transfected with AKAR3 and WT-CB1R at DIV6 and imaged from DIV8 to DIV10. After selection of neurons (somatodendritic parts), I started the acquisition with one image per minute. Neurons were pretreated 3 hours before starting the acquisition with either CB1R agonist WIN55,212-2 (WIN) 1µM or inverse agonist AM281 (AM) 1µM. After recording 10

112 minutes of baseline, increasing doses of Fsk were added in the medium each 5 minutes, ranging from 100 nM to 100 µM. Results are presented in Figure 41. The increase of Fsk concentrations led to progressive increases of FRET ratio (Figure 41-1, 41-2). I developed a Matlab program to calculate the mean of responses for each Fsk dose, enabling to obtain for each soma, a dose-response curve (Figure 41-3). Using Graphpad Prism, calculation of EC50 value was performed for each soma. Thus, I plotted mean dose-response curves for each treatment (Figure 41-4) and compared EC50 of each group (Figure 41-5). Interestingly, in presence of WIN 1µM, Fsk dose-response curves were shifted to higher Fsk concentration in somata, corresponding to a significantly increased EC50 as compared to Vehicle. Thus, CB1Rs present in somata are activable by exogenous cannabinoids and their activation leads to a decreased activity of ACs. Moreover, treatments with AM 1µM diminished EC50 to Fsk, indicating that CB1Rs constitutively inhibit ACs in somata.

113

Figure 41: CB1R constitutively inhibits Forskolin-stimulated adenylyl cyclase in somata (1)

(1) (2) Variation of FRET Ratio in a soma after stimulation with increasing Fsk doses. (3) Forskoline dose-reponse curve in a soma enabling the calculation of EC50. (4) (5) Comparison of somata Fsk dose-response curves of neurons treated with either AM

(1µM), WIN (1µM) or Vehicle (DMSO 0,02%). Calculation of EC50 and statistical analysis have been performed on Graphpad Prism. Significance of differences between various conditions was calculated using one-way ANOVA with Newman Keuls post-tests for computing p estimates. For significance symbols (asterisks), three symbols mean p < 0.001.

Moreover, I also used Fsk dose-responses protocol in order to evaluate EC50 to Fsk in somata of neurons expressing AKAR3 and WT-CB1R or hyperactive mutant T210I-CB1R or hypoactive mutant T210A-CB1R (Figure 42). Our results show that CB1R hyperactivation increases EC50 to Fsk in somata, while overexpression of the hypoactive mutant decreases

EC50. These results are coherent with previous results obtained with pharmacological treatments, demonstrating notably that overexpressed WT-CB1Rs constitutively inhibit AC activity in somata.

114

Figure 42: CB1R constitutively inhibits Forskolin- stimulated adenylyl cyclase in somata (2) Comparison of forskoline dose-response curves for somata of neurons cotransfected with AKAR3 and either CB1R Wild Type (WT), hyperactive CB1R (T210I) or hypoactive CB1R (T210A). Calculation of EC50 and statistical analysis have been performed on Graphpad Prism. Significance of differences between various conditions was calculated using one-way ANOVA with NewmanKeuls post-tests for computing p estimates. For significance symbols (asterisks), one symbol means p < 0.05, and three mean p < 0.001.

2.3.4.2. Modulation of basal PKA activity induced by activation or blockade of overexpressed CB1Rs in somata and dendrites.

Use of Fsk is a common method to study Gi/o-protein coupled receptors but leads to high increase of PKA activity as compared with its basal level. Thus, we sought to measure effect of direct activation or blockade of CB1R without using Fsk stimulation to verify that above results are not only due to artificial high increase of PKA activity. Thirty minutes after beginning of acquisition, cultured neurons cotransfected with WT-CB1R and AKAR3 were treated with either WIN (1µM), AM (1µM) or vehicle solution (DMSO 0,02%). 30 minutes after treatment addition, Fsk was added at 50µM, used as an internal positive control (Figure 40, protocol 3). After the injection of WIN, we observed a significant decrease of FRET ratio in somata as compared to vehicle (Figure 43-1, 43-2). This result suggests that somatic CB1Rs are activable and their activation leads to an inhibition of PKA activity. Moreover, in order to verify the implication of Gi/o proteins, neurons were pre-incubated with Pertussis Toxin (PTX) at least 12 hours before acquisition where WIN was added (Figure 43-3). Results show that PTX blocks WIN effects on PKA activity confirming that CB1R activation induces PKA activity decrease via Gi/o proteins that inhibit ACs. Moreover, the injection of AM led to an increase of FRET ratio in somata indicating that CB1Rs exert a constitutive inhibition on PKA activity that is removed by inverse agonist treatment (Figure 43-1, 43-2).

115

Figure 43: CB1Rs modulate basal PKA activity in somata. (1) FRET ratio in somata with injections of either WIN (1µM), AM (1µM) or Vehicle (DMSO 0,02%). (2-3) Mean of FRET responses in somata after treatment injection are

compared on GraphPad Prism. Significance of differences between various conditions was calculated using one-way ANOVA with NewmanKeuls post-tests for computing p estimates. For significance symbols (asterisks),

one symbol means p < 0.05, two means p < 0.01, and three means p < 0.001, NS means p>0.05.

In order to quantify FRET ratio changes in dendrites, I measured the ratio in several dendrites per neuron by defining several ROIs (Figure 44-1, 44-2). Then, in order to compare dendritic responses, I calculated the mean FRET ratio for each time point for all dendrites of the same neuron. Then, these mean curves are pooled together to compare effects of various treatments. Thus, numbers (N) indicated on Figure 44-3 and 44-4 correspond to the number of neurons and not to the number of dendrites. I automated this part in Matlab. Using this protocol, data show that similarly to somata, activation of CB1Rs decrease basal PKA activity in dendrites .

116 On the contrary, blockade of CB1Rs increases basal PKA activity, demonstrating that CB1Rs constitutively inhibit PKA activity in dendrites (Figure 44-3 and 44-4).

Figure 44: CB1Rs modulate basal PKA activity in dendrites. (1-2) Dendritic responses are quantified using the Matlab program described above. Several ROIs can be selected in order to measure FRET ratio in several dendrites (1). Then FRET ratio is calculated for each ROI (2). (3) Normalized FRET ratio are shown as mean +/- SEM in function of time. During acquisition, WIN (1µM), AM (1µM) or Vehicle (DMSO 0,02%) were added. (4) Mean of FRET responses in dendrites after treatment injection are compared on GraphPad Prism. Significance of differences between various conditions was calculated

using one-way ANOVA with NewmanKeuls post-tests for computing p estimates. For significance symbols (asterisks), three symbols mean p < 0.001.

117 2.3.4.3. Modulation of basal PKA activity in axons downstream of endogenous CB1Rs. Cytosolic localization of AKAR3 did not allow to measure PKA modulation in axons, due to the very low fluorescence in these thin neurites, which have a highly elevated surface-to- volume ratio. Thus, I used AKAR4-Kras probe that is targeted to plasma membrane. This subcellular targeting enabled to obtain a good fluorescence level in axons and to reliably measure basal PKA modulation in all neuronal compartments downstream of overexpressed WT- CB1Rs (see Part 4) by imaging a reasonable number of neurons in each experiment. We also observed that Fsk 10µM induced an increase of about 30% of normalized FRET ratio in somata using AKAR4-Kras probe, while Fsk 100 µM increases FRET ratio of about 8% only with AKAR3. We sequenced the AKAR3 probe used in the laboratory, and discovered that the fluorophores of this probe were not CFP/YFP as we supposed, but GFP/m-Cherry, another FRET pair. Our emission filters were optimized for CFP and YFP, capable to detect only a minor part of the GFP emission spectrum. Interestingly, when FRET occurs, a deformation of the donor emission spectrum is observed. Thus, I suppose that with our system we measured previously only FRET-induced changes in GFP emission, explaining why the amplitude of Fsk-induced FRET changes detected by AKAR3 was very low. Because of this experimental problem, we decided not to publish the data previously obtained in the somatodendritic domain and to repeat these experiments by using AKAR4-Kras. On the other hand, this unsuspected technical problem helped us to fine-tune the sensitivity of our system at the beginning of my PhD project, by trying to replicate the amplitude of published FRET changes. Interestingly, the strong efficiency we gained using AKAR4-Kras enables also the detection of endogenous CB1R signaling. Results obtained are presented in the next paragraph that is a submitted article titled Polarized cellular patterns of endocannabinoid production and detection shape cannabinoid signaling in neurons. Alexandre Roland, second author, has set up the FRET system and supervised my master project, Stefan Biedzinski, third author, has performed all the experiments using TEpacVV probe during his six-months internship in the laboratory, and Ana Ricobaraza, post-doctorant in our team, performed the DAGLα immunostaining experiment. I also used this technique to validate a new construction of cGMP sensor that I developed and to measure the recruitment of signaling pathways downstream of endogenous somatostatin SST2 receptors in cultured hippocampal neurons of rats (see Annexe 2).

118

3. Results: Article 2 (submitted): Polarized cellular patterns of endocannabinoid production and detection shape cannabinoid signaling in neurons. Delphine Ladarre, Alexandre Roland, Stefan Biedzinski, Ana Ricobaraza, and Zsolt Lenkei

119 The Journal of Neuroscience

http://jneurosci.msubmit.net

JN-RM-2596-14

Polarized cellular patterns of endocannabinoid production and detection shape cannabinoid signaling in neurons

Zsolt Lenkei, ESPCI-ParisTech Delphine Ladarre, ESPCI-ParisTech Stefan Biedzinski, ESPCI-ParisTech Alexandre Roland, ESPCI-ParisTech Ana Ricobaraza, ESPCI-ParisTech

Commercial Interest: No

This is a confidential document and must not be discussed with others, forwarded in any form, or posted on websites without the express written consent of The Journal for Neuroscience.

1 Polarized cellular patterns of endocannabinoid production and detection shape 2 cannabinoid signaling in neurons 3 4 Ladarre Delphine1,2, Roland Alexandre1,2,3, Biedzinski Stefan1,2, Ricobaraza Ana1,2, Lenkei 5 Zsolt1,2 6 1Brain Plasticity Unit, ESPCI-ParisTech, 75005 Paris, France 7 2CNRS UMR 8249, 75005 Paris, France 8 3FAS Center for Systems Biology, Harvard University, Cambridge MA 02138, USA 9 10 11 Corresponding author: Zsolt Lenkei M.D., Ph.D., Brain Plasticity Unit, ESPCI-ParisTech, 12 75005 Paris, France, email: [email protected] 13 14 15 Running title: Polarized cannabinoid signaling in neurons 16 Number of pages: 25 17 Number of figures: 5 18 Abstract: 203 words 19 Introduction: 497 words 20 Discussion: 1500 words 21 22 Acknowledgements: This work was supported by a grant from the ANR (L' Agence Nationale 23 de la Recherche) to Zsolt Lenkei (ANR-09-MNPS-004-01). We thank Dr. Christophe 24 Leterrier (Marseille) and Dr. Pierre Vincent (Paris) for discussions and advice and Maureen 25 McFadden for the help with the English syntax. 26 27 Conflict of Interest: The authors declare that they have no conflict of interest. 28 29

30 Abstract 31 32 Neurons display important differences in plasma membrane composition between 33 somatodendritic and axonal compartments, potentially leading to currently unexplored 34 consequences in G-protein-coupled-receptor signaling. Here, by using highly-resolved 35 biosensor imaging, we measured local changes in basal levels of key signaling components 36 downstream of endogenous type-1 cannabinoid receptors (CB1R), in individual axons and

37 dendrites of cultured rat hippocampal neurons. CB1R activation led to rapid, Gi/o-protein- and 38 cAMP-mediated decrease of cyclic-AMP-dependent protein kinase (PKA) activity in the 39 somatodendritic compartment. In axons, PKA inhibition was significantly stronger, in line 40 with axonally-polarized distribution of CB1Rs. Conversely, inverse agonist AM281 produced 41 marked rapid increase of basal PKA activation in somata and dendrites, but not in axons, 42 removing constitutive activation of CB1Rs generated by local production of the 43 endocannabinoid 2-arachidonoylglycerol (2-AG). Interestingly, somatodendritic 2-AG levels 44 differently modified signaling responses to CB1R activation by Δ9-THC, the psychoactive 45 compound of marijuana, and by the synthetic cannabinoids WIN55,212-2 and CP55,940. 46 These highly contrasted differences in sub-neuronal signaling responses warrant caution in 47 extrapolating pharmacological profiles, which are typically obtained in non-polarized cells, to 48 predict in vivo responses of axonal (i.e. presynaptic) GPCRs. Therefore, our results suggest 49 that enhanced comprehension of GPCR signaling constraints imposed by neuronal cell 50 biology may improve the understanding of neuropharmacological action. 51 52 53 54

55 Introduction

56 Polarized neuronal architecture maintains the directionality of information flow through 57 neuronal networks. Accordingly, protein and lipid composition of the plasma membrane 58 greatly differs between axons and the somatodendritic compartment (Horton and Ehlers, 59 2003). Local interaction between cell membrane components is increasingly considered as a 60 key dynamic component in sensory and signaling pathways. Notably, the highly regulated 61 lipid environment may control the structure, conformation and function of embedded proteins 62 (Phillips et al., 2009). A major brain G-protein coupled receptor (GPCR) that may be 63 particularly sensitive to the lipid composition of the plasma membrane is the type-1 64 cannabinoid receptor (CB1R). Predominantly localized in axons and specific presynaptic 65 nerve terminals, CB1R is the neuronal target of endocannabinoid lipids (eCBs) and of Δ9- 66 tetrahydrocannabinol (THC), the major psychoactive substance of marijuana. CB1Rs may 67 show elevated tonic (constitutive) activation in neurons (Pertwee, 2005), potentially resulting 68 from a combined effect of conformational instability (D'Antona et al., 2006) and ubiquitously 69 present membrane-borne eCBs, such as 2-arachidonoylglycerol (2-AG), which is the most 70 prominent brain eCB (Alger and Kim, 2011; Howlett et al., 2011) as well as an important 71 intermediate in the production of several other bioactive lipids (Nomura et al., 2011). 2-AG is 72 released from cell membrane phospholipids by the subsequent action of phospholipase C and 73 diacylglycerol lipases (DAGLα and DAGL). eCBs are generally considered to be retrograde 74 signals, being produced in the postsynaptic cell and travelling backwards across the 75 synaptic cleft to activate CB1Rs on presynaptic nerve terminals (Freund et al., 2003; Kano et 76 al., 2009). However, in addition to this retrograde synaptic signaling effect, eCBs synthetized 77 in the somatodendritic membrane may also have cell-autonomous effects on local CB1Rs, 78 such as endocannabinoid-mediated somatodendritic slow self-inhibition (SSI) (Bacci et al., 79 2004; Marinelli et al., 2009) or somatodendritic-endocytosis driven transcytotic targeting 80 (Leterrier et al., 2006; Simon et al., 2013). These findings suggest that locally produced 2-AG 81 may activate somatodendritic CB1Rs, although such CB1R-induced somatodendritic 82 signaling has not yet been shown directly..

83 CB1R activation, through coupling to Gi/o heterotrimeric proteins, leads to inhibition of cyclic 84 adenosine monophosphate (cAMP) production and inhibition of cyclic-AMP-dependent 85 protein kinase (PKA) activity (Howlett, 2005). cAMP and PKA regulate essential biological 86 functions in neurons such as excitability, efficacy of synaptic transmission and axonal 87 growth/pathfinding. Therefore, CB1R coupling to this major signaling pathway may have

88 important consequences on neuronal function. However, in absence of direct measurement of 89 somatodendritic and axonal CB1R signaling, whether and how differences in local CB1R 90 density and local 2-AG content regulate signaling responses to cannabinoids remains 91 unknown. 92 More generally, it is not currently known how the highly-polarized neuronal membrane 93 environment may shape GPCR signaling. This information may be important to better 94 understand neuronal effects of therapeutic or abused drugs, whose pharmacological properties 95 are usually studied in non-polarized heterologous expression systems, such as immortalized 96 cell lines. Therefore, here we used a highly-resolved and sensitive Förster Resonance Energy 97 Transfer (FRET) approach to measure ligand-induced modulation of basal cAMP/PKA levels 98 at the sub-neuronal scale in vitro, downstream of endogenous CB1Rs, in well-differentiated 99 hippocampal neurons. 100 101 Materials and methods 102 103 Animals 104 Animal procedures were conducted in strict compliance with approved institutional protocols 105 and in accordance with the provisions for animal care and use described in the European 106 Communities council directive of November 24, 1986 (86 � 609 � EEC). SpragueDawley rats 107 (Janvier) were used for dissociated cell culture experiments. 108 109 Chemicals, antibodies and DNA constructs 110 CB1R agonists WIN55,212,2 (WIN), CP55,940 (CP) and 2-arachydonoylglycerol (2-AG), 111 CB1R inverse agonist AM281 (AM) and DAGL inhibitor RHC80267 (RHC) were obtained 112 from R&D Systems Europe. Dimethyl Sulfoxide (DMSO), Tetrahydrolipstatin (THL), Δ9- 113 Tetrahydrocannabinol solution (THC), Pertussis Toxin (PTX), Forskoline (Fsk), monoclonal 114 mouse anti-Tau antibody, monoclonal mouse anti-microtubule-associated protein 2 (anti- 115 MAP2) antibody, Bovine Serum Albumin (BSA) and Poly-D-Lysine were obtained from 116 SIGMA-ALDRICH. Polyclonal anti-DAGLα antibody was obtained from Frontier Institute 117 co., ltd (JAPAN). B27, Lipofectamine 2000 and Neurobasal were obtained from Life 118 Technologies. 119 AKAR4, Lyn-AKAR4 and AKAR4-Kras probes were obtained from Dr Jin Zhangs 120 laboratory (Baltimore, USA). TEpacVV probe was obtained from Dr Kees Jalink laboratory 121 (Amsterdam, Netherlands).

122 123 Hippocampal neuronal cultures 124 Hippocampal neuronal cultures were performed essentially as described previously (Leterrier 125 et al., 2006). Briefly, hippocampi of rat embryos were dissected at embryonic day 18. After 126 trypsinization, dissociation was achieved with a fire-polished Pasteur pipette. Cells were 127 counted and plated on poly-D-lysine-coated 18-mm diameter glass coverslips, at a density of 128 300400 cells/mm2. The plating medium was Neurobasal supplemented with 2% B27 and 129 containing Stabilized Glutamine (0.5 mM) and penicillin G (10 U/ml)/streptomycin (10 g/ml). 130 Four hours after plating, the coverslips were transferred into Petri dishes containing 131 supplemented Neurobasal medium that had been conditioned for 24 h on a 80% confluent glia 132 layer. Neurons were transfected after 6 days in vitro (DIV6) using Lipofectamine 2000, 133 following the manufacturers instructions. 134

135 FRET imaging 136 Neurons transfected either with TEpacVV or AKAR4-KRAS were imaged by videomicroscopy 137 between DIV7 and DIV11 on a motorized Nikon Eclipse Ti-E/B inverted microscope with the 138 Perfect Focus System (PFS) in a 37°C thermostated chamber, using an oil immersion CFI

139 Plan APO VC 60X, NA 1.4 objective (Nikon). 140 Acquisitions were carried out at the excitation wavelength of the CFP (434nm +/- 15nm) 141 using an Intensilight (Nikon). Emitted light passed through an Optosplit II beam-splitter 142 (Cairn Research) equipped with a FF509-FDi01 dichroïc mirror, a FF01-483/32-25 CFP filter 143 and a FF01-542/27-25 YFP filter and was collected by an EM-CCD camera (Evolve 512, 144 Photometrics), mounted behind a 2x magnification lens. Acquisitions were performed by 145 piloting the set-up with Metamorph 7.7 (Molecular Devices). All filter sets were purchased 146 from Semrock. 147 Cultured neurons on 18-mm coverslips were placed in a closed imaging chamber containing

148 an imaging medium: 120 mM NaCl, 3 mM KCl, 10 mM HEPES, 2 mM CaCl2, 2 mM

149 MgCl2, 10 mM D-glucose, 2% B27, 0,001% BSA. 150 The acquisition lasted 90 minutes registering one image each 2 minutes, registering in parallel 151 10 to 15 neurons on the same coverslip. 30 minutes after the beginning of the acquisition, 152 pharmacological treatment was applied then 60 minutes after the beginning of the acquisition, 153 Forskoline 10µM was applied. 154 155 FRET data analysis

156 Images were divided in two parts on ImageJ to separate the CFP channel from the YFP 157 channel. Data were then analyzed on Matlab by calculating the FRET ratio at each time point 158 for one or several Regions Of Interest (ROIs). The user defined ROIs for each position. For

159 each image, the value of the FRET ratio corresponds to -IC BC for TEpacVV probe and to IY - BY 160 -IY BY for AKAR4-KRAS probe IC - BC 161 IY : Mean Intensity of ROI in YFP channel; 162 BY : Mean Intensity of the background in YFP channel; 163 IC : Mean Intensity of ROI in CFP channel; 164 BC : value of the background in CFP channel 165 For each ROI, the FRET ratio was then normalized by the baseline mean, defined as the 7 Rc  Ro 166 time points before first treatment injection. FRET Ratio  100 * Ro 167 Rc : Value of crude FRET ratio 168 Ro : Mean of the baseline 169 The quantitative results obtained for each neuronal compartment were grouped together and, 170 for each time point, the mean FRET ratio normalized to baseline and SEM were calculated. 171 Deviation was corrected for somata and dendrites on Matlab. Mean slope was calcultated for 172 all neurons in somata and dendrites, respectively, for the last 7 time points before addition of 173 treatment and substracted from all FRET ratio time points. 174 175 FRET statistical analysis 176 FRET Response was obtained by calculating the mean FRET ratio on Matlab for 6 time points 177 after injection, from +4 minutes to +14 minutes (Response). 178 Groups were compared using GraphPad Prism. Significance of differences between various 179 conditions was calculated using unpaired t-tests or one-way ANOVA with Newman-Keuls 180 post-tests for computing p estimates. NS p>0.05, * p<0.05, ** p<0.01 and *** p<0.001. 181 182 Immunocytochemistry 183 DIV9 hippocampal cultured neurons were briefly rinsed with Dulbecoos PBS (DPBS; PAA 184 laboratories) and fixed in DPBS containing 4% paraformaldehyde and 4% sucrose. After 185 permeabilization with a 5 minutes incubation in DPBS containing 0.1% Triton X-100 and 186 blocking for 30 minutes in antibody buffer (DPBS supplemented with 2% BSA and 3% 187 normal goat serum), neurons were incubated with primary antibodies diluted 1:200 (DAGLα)

188 or 1:250 (MAP2 and Tau) in antibody buffer for 1h at room temperature. After DPBS rinses, 189 neurons were labelled with secondary antibodies 1:400 in antibody buffer for 30 minutes at 190 room temperature. Coverslips were fixed with Mowiol containing Hoechst. Images were 191 obtained using a dry 40x objective lens on Zeiss Axio Imager M1. Excitation wavelengths of 192 488 nm (DAGLα) and 568 nm (MAP2 or Tau) were used. 193 194 Confocal microscopy 195 Hippocampal cultured neurons were cotransfected at DIV6 with DsRed2 and various FRET 196 probes (TEpacVV, AKAR4, AKAR4-Kras, Lyn-AKAR4) and fixed in DPBS containing 4% 197 paraformaldehyde and 4% sucrose at DIV7. Images were obtained using an oil immersion 198 objective lens (Plan-Apochromat 60X, NA 1.4) on a Nikon A1 confocal microscope. 199 Excitation wavelengths of 488 nm (FRET probes) and 568 nm (DsRed2) were used. Stacks 200 were obtained with one image per optical section and 300 nm between each section. 201 202 Results 203 204 Endogenous CB1Rs modulate basal PKA activation levels in neurons. The characteristic

205 inhibition of cyclic AMP production and PKA activity by Gi/o-protein coupled GPCRs is 206 usually detected in pharmacological assays after GPCR over-expression and forskolin- 207 induced artificial activation of adenylyl cyclases. Here we aimed to directly measure 208 cannabinoid-induced changes in basal levels of neuronal PKA signaling, downstream of 209 endogenous CB1Rs in cultured hippocampal neurons. Pilot experiments indicated that by 210 using a sensitive EM-CCD camera and hardware-based focus stabilization (see Materials and 211 Methods) we are able to measure cannabinoid-induced inhibition of cyclic AMP production 212 and PKA activity in relatively large cytoplasmic volumes such as neuronal somata by using 213 the soluble TEpacVV probe (Klarenbeek et al., 2011) (Fig. 1A) and AKAR4 (Depry et al., 214 2011) (Fig. 1B), respectively. However, smaller diameter neurites such as distal dendrites and 215 axons gave weak (low amplitude) and highly variable responses, leading to a low signal-to- 216 noise ratio, which impeded the reliable measure of the relatively small amplitude 217 cannabinoid-induced changes in the FRET ratio with the AKAR4 probe. To overcome this 218 experimental limitation, we hypothesized that, since the PKA activator cAMP is produced by 219 membrane-bound adenylyl cyclases at the plasma membrane and PKA deactivator 220 phosphodiesterases are cytosolic (Neves et al., 2008), targeting a PKA probe to the plasma 221 membrane may strongly increase experimental sensitivity. Indeed, results of a previous report

222 show both higher FRET responses and higher PKA-sensitive potassium current responses

223 downstream of Gs-protein activation in dendrites that have a high surface-to-volume ratio as 224 compared to the soma (Castro et al., 2010). Therefore, we expressed separately two 225 membrane-targeted PKA biosensors: AKAR4-Kras (Depry et al., 2011), which is targeted to 226 the non-raft domains of the plasma membrane (Figure 1C), and Lyn-AKAR4 (Depry et al., 227 2011), which is targeted to the raft regions of the plasma membrane (Figure 1D) in well- 228 differentiated hippocampal neurons. In our experimental conditions, AKAR4-Kras showed a 229 more homogenous distribution that segregated well with the plasma membrane at different 230 optical sections of the somatodendritic domain, while Lyn-AKA4 was more strongly localized 231 to relatively small membrane microdomains and intracellular structures (Figure 1C-D). In 232 order to focus on plasma-membrane localized endogenous CB1Rs, further experiments were 233 therefore performed using AKAR4-Kras. 234 Does the membrane-targeted AKAR4-Kras probe permit the measurement of cannabinoid- 235 induced modulation of basal PKA levels downstream of endogenous CB1Rs in all neuronal 236 sub-compartments? We tested the sensitivity of our experimental setup by determining the 237 minimal amount of cytoplasmic volume necessary to the detection of cannabinoid-induced 238 modulation of basal PKA levels, in individual thin (mean diameter = 0.7 µm, see Figure 3C) 239 and mature (day in vitro 9 - DIV9) axons of AKAR4-Kras expressing neurons. By using a 240 large region of interest (ROI) to measure the FRET ratio change, treatment with the synthetic 241 CB1R agonist, WIN55,212-2 (WIN) at 100 nM (Figure 1F), but not with vehicle (Figure 1E), 242 induced a strong change of the FRET ratio within 2 minutes, indicating that CB1R activation 243 induces a decrease of basal PKA activity downstream of endogenous CB1Rs. By gradually 244 decreasing the size of the ROI, we determined the minimum cytoplasmic volume necessary 245 for reliable measurement of the WIN-induced FRET signal change. To calculate the volume 246 (V) corresponding to a ROI, the axon was considered as a cylinder; the diameter (d) and 247 length (L) were measured and volume was calculated as: V=π x L x (d/2)2. We found that a 248 significant decrease of WIN-induced basal PKA activity downstream of endogenous CB1Rs 249 could be measured in volumes as small as 1µm3, which equates to 1 femtoliter of axonal 250 cytoplasm (Figure 1G). 251 In conclusion, this experimental approach enables the measurement of modulation of basal

252 neuronal PKA activation levels, downstream of an endogenous Gi/o protein coupled receptor, 253 in extremely small cellular volumes, such as the cytoplasm of mature axons. 254 255 Transient somatodendritic CB1Rs constitutively inhibit the cAMP/PKA pathway.

256 Previous ultrastructural analysis of hippocampal neurons has shown that in the 257 somatodendritic region, the steady-state presence of endogenous CB1Rs at the plasma 258 membrane is very low both in vitro (Leterrier et al., 2006) and in vivo (Katona et al., 1999; 259 Thibault et al., 2013). However, previous studies have also reported that most axonally 260 targeted CB1Rs accomplish a transient passage on the somatodendritic plasma membrane 261 (Leterrier et al., 2006; McDonald et al., 2007; Simon et al., 2013). Currently, it remains 262 unknown whether somatodendritic CB1Rs are able to inhibit cAMP/PKA signaling in this 263 neuronal compartment. Therefore, we measured modulation of basal somatodendritic PKA 264 activity downstream of endogenous CB1Rs and found that treatment with WIN at 100nM, but 265 not with vehicle, induced a moderate decrease of the FRET ratio in individual neurons within 266 a few minutes (Figure 2A-B). To precisely analyze this WIN-induced response, we compared 267 PKA activity in two groups of neurons treated either with vehicle or WIN (100nM) during 30 268 minutes, followed by Forskolin (Fsk, 10µM) treatment (Figure 2C, 2D), to induce saturating 269 activation of adenylyl cyclases. Fsk induced strong somatodendritic PKA activation with a 270 FRET-ratio increase between 20% and 30%. Conversely, activation of CB1Rs with WIN 271 induced a rapid decrease of basal PKA activity in somata (-2.54 +/- 0.38%) and dendrites (- 272 3.21 +/- 0.52%), which was significant as compared to vehicle (somata: 0.08 +/- 0.27%,

273 dendrites: -0.15 +/- 0.41%) (Figure 2C1, 2C2, 2D1, 2D2). This effect was blocked after

274 overnight treatment with 100 ng/mL of the Gi/o-protein specific inhibitor pertussis toxin 275 (PTX) (somata: -0.34 +/- 0.24%, dendrites: -1.5 +/- 0.29) as well as after 3 hour pre-treatment 276 with 1 µM of the CB1R-specific antagonist/inverse-agonist AM281 (somata: 0.26 +/- 0.26%, 277 dendrites: -1.12 +/- 0.49%). Therefore, endogenous CB1Rs, transiently present on the 278 somatodendritic plasma membrane, can be activated by exogenous cannabinoids and are able

279 to subsequently inhibit basal PKA signaling through their coupling to Gi/o proteins both in 280 somata and dendrites. 281 We have previously reported that somatodendritic CB1Rs are constitutively endocytosed 282 because of constitutive receptor activation, which can be inhibited by pharmacological or 283 genetic tools (Leterrier et al., 2006; Simon et al., 2013). To investigate whether CB1Rs also 284 constitutively inhibit cAMP/PKA signaling in the somatodendritic compartment, we applied 285 the CB1R inverse agonist, AM281 at 100 nM, to neurons expressing AKAR4-Kras. This 286 treatment led to a rapid and significant increase of the FRET ratio both in somata and

287 dendrites (somata: 1.34 +/- 0.20%, dendrites: 1.95 +/- 0.37%) (Figure 2C1, 2C2, 2D1, 2D2). 288 Therefore, somatodendritic CB1Rs exert a constitutive inhibition on PKA activity that is 289 removed by inverse agonist treatment.

290 Taken together these results indicate that somatodendritic CB1Rs constitutively inhibit PKA

291 activity through the mobilization of Gi/o proteins, which is likely due to the inhibition of 292 adenylate cyclase and subsequent decrease of cAMP production. To confirm this mechanism, 293 we directly measured the modulation of basal somatodendritic cAMP concentration, 294 downstream of CB1Rs, by using the TEpacVV probe (Klarenbeek et al., 2011). The activation 295 of endogenous CB1Rs with WIN (100nM) induced a rapid and significant decrease of cAMP 296 concentration in somata (-1.46 +/- 0.25%) and dendrites (-2.55 +/- 0.55%), while application 297 of the inverse agonist AM281 at 100 nM led to a rapid and significant increase of cAMP

298 concentration both in somata (1.66 +/- 0.26%) and dendrites (2.56 +/- 0.50%) (Figure 2E, 2E1,

299 2E2, 2F, 2F1, 2F2). 300 These results show that endogenous CB1Rs exert a constitutive inhibition on the cAMP/PKA 301 signaling pathway both in somata and dendrites. In addition, somatodendritic CB1Rs can be 302 further activated by exogenous cannabinoids leading to a rapid decrease of cAMP

303 concentration and PKA activity through activation of Gi/o proteins. 304 305 Axonal CB1R signaling is different from dendritic signaling. Previous studies have found 306 a polarized accumulation of transcytosed CB1Rs on the axonal plasma membrane due to 307 reduced internalization levels as compared to dendrites (Leterrier et al., 2006; McDonald et 308 al., 2007; Simon et al., 2013). We asked whether the recruitment of signaling pathways 309 downstream of CB1Rs in axons also differ from somata and dendrites. Application of 100nM 310 WIN led to a rapid and significant decrease of basal PKA activity in axons (-14.63 +/- 1.38%)

311 compared to vehicle (-1.81 +/- 0.60%) (Figure 3A, 3A1, 3A2). This effect was blocked by pre- 312 treatment with AM281 1µM (-2.61 +/- 1.00%) and PTX 100 ng/mL ( -5.50 +/- 1.06%),

313 showing that PKA inhibition is specifically mediated by CB1Rs acting through Gi/o proteins. 314 In addition, CB1R activation decreased PKA activity more strongly in the axon than in 315 dendrites (dendrite response normalized to vehicle: -3.07 +/- 0.52%, axonal response 316 normalized to vehicle: -12.83 +/- 1.38) (Figure 3B). Interestingly, in contrast to dendrites, 317 application of the inverse agonist AM281 at 100nM did not induce a detectable change of 318 PKA activity in the axon (0.08 +/- 0.84%), suggesting that axonal CB1Rs are not

319 constitutively activated (Figure 3A, 3A1, 3A2). 320 Why does axonal CB1R activation lead to a significantly higher amplitude of PKA inhibition 321 in axons than in dendrites and somata? First, similarly to their distribution in vivo (Katona et 322 al., 2001; Bodor et al., 2005; Thibault et al., 2013), CB1Rs display an axonally polarized 323 distribution in cultured neurons (Coutts et al., 2001; Leterrier et al., 2006; McDonald et al.,

324 2007; Simon et al., 2013). Second, theoretical models predict, and experiments show, that, for 325 signaling molecules produced at the plasma membrane and degraded in the cytoplasm, such 326 as cAMP, the ratio of the surface area of the plasma membrane to the cytoplasmic volume 327 (surface/volume ratio (S/V)) becomes important (Neves et al., 2008). As such, we asked 328 whether the strong decrease of PKA activity observed after CB1R activation in the axon is 329 related to the high S/V ratio of this compartment. However, for both axons and distal 330 dendrites, we found no correlation between neurite diameter and FRET response amplitude

331 after CB1R activation (Pearson's correlation coefficient: rdistal dendrites = -0.065 and raxons = 332 0.03649) (Figure 3C). Moreover, a sub-population of distal dendrites has the same diameter 333 range as axons. In these thin dendritic segments, the amplitude of the FRET response after 334 CB1R activation was again significantly different from the axonal response (dendrites 335 normalized to vehicle: -4.04 +/- 0.80%, axons normalized to vehicle: -12.83 +/- 1.38) (Figure 336 3D). Therefore, morphological differences between axons and dendrites do not explain the 337 observed signaling disparity among these two compartments, suggesting that the polarized 338 distribution of neuronal CB1Rs is the main reason for the enhanced agonist response in axons. 339 340 Constitutive activation of somatodendritic CB1Rs requires local synthesis of 341 endocannabinoids. Next we investigated why CB1Rs are constitutively activated in the 342 somatodendritic compartment but not in the axon, by focusing on the contribution of 343 endocannabinoids, which play an important role in basal CB1R activation in several 344 experimental systems (Turu et al., 2007; Howlett et al., 2011). The major endocannabinoid 2- 345 arachidonoylglycerol (2-AG) is a lipid molecule present in the cell plasma membrane and is 346 synthesized by DAG Lipases (DAGL). DAGLα, the major DAGL in the postnatal brain, is 347 segregated to axonal tracts during embryonic development but was shown to accumulate after 348 birth in the somatodendritic plasma membrane in several brain areas, such as the cerebellum 349 (Bisogno et al., 2003), striatum (Uchigashima et al., 2007), hippocampus (Katona et al., 2006; 350 Yoshida et al., 2006) and amygdala (Yoshida et al., 2011). Similarly, we found a 351 somatodendritic segregation of DAGLα in fully-polarized (DIV9) cultured hippocampal

352 neurons, while no labeling was found in the axon (Figure 4A, 4A1). This indicates local 353 production of 2-AG in the plasma membrane of the somatodendritic compartment but not in 354 the axonal counterpart. To investigate whether such polarized 2-AG production may explain 355 the aforementioned differences in constitutive CB1R activation between dendrites and axons, 356 we pre-treated neurons expressing the AKAR4-Kras probe with two DAGL inhibitors, 357 Tetrahydrolipstatin (THL) or RHC80267 (RHC), during 3 hours before treatment with the

358 inverse agonist AM281 100 nM (Figure 4B, 4B1, 4C, 4C1). The FRET ratio did not increase 359 in these neurons after adding AM281, neither in somata (AM281: 1.56 +/- 0.28%, vehicle: 360 0.21 +/- 0.29%, AM281 after THL 1µM: 0.43 +/- 0.34%, AM281 after RHC 25µM: 0.06 +/- 361 0.28%) nor in dendrites (AM281: 2.01 +/- 0.53%, vehicle: 0.40 +/- 0.44%, AM281 after THL 362 1µM: 0.40 +/- 0.29%, AM281 after RHC 25µM: -0.13 +/- 0.45%). Thus, the constitutive 363 inhibition on PKA activity was removed by DAGL blockade, demonstrating that constitutive 364 activation of somatodendritic CB1Rs requires locally produced 2-AG, which may activate the 365 receptors through a cell-autonomous mechanism. 366 367 Signaling responses to exogenous ligands WIN, CP55,940 and Δ9-THC are differentially 368 shaped by local production of 2-AG in the somatodendritic compartment. Previous 369 results show that, after DAGL inhibition, the amount of CB1Rs increase on the plasma 370 membrane, both in the somatodendritic compartments of neurons and in CHO cells (Turu et 371 al., 2007). In CHO cells, the elevated CB1R levels at the plasma membrane yield enhanced G- 372 protein activation following WIN administration (Turu et al., 2007). We asked whether the 373 THL-induced accumulation of CB1Rs on the somatodendritic membrane is able to produce 374 similar enhanced inhibition of PKA activity after WIN administration, as compared to basal 375 conditions. Therefore, we pre-treated neurons with THL (1 µM) during 3 hours before 376 acquisition and applied WIN during the FRET acquisition (Figure 5A). Surprisingly, DAGL 377 inhibition blocked the effect of 100 nM WIN in the somatodendritic compartment instead of 378 signaling enhancement (somatic response to WIN 100nM: -2.35 +/- 0.40%, P<0.01 compared 379 to vehicle (0.06 +/- 0.19%) and P<0.01 compared to response to WIN 100nM after 3h THL 380 1µM (-0.20 +/- 0.58%), one-way ANOVO followed by Newman-Keuls post-test; dendrite 381 response to WIN 100nM: -2.76 +/- 0.43%, P<0.01 compared to vehicle (-0.36 +/- 0.41%) and 382 P<0.01 compared to response to WIN 100nM after 3h THL 1µM (-0.56 +/- 0.58%), one-way

383 ANOVO followed by Newman-Keuls post-test) (Figure 5A, 5A1, 5C, 5C1), while it did not 384 change the FRET response in the axon (response to WIN 100nM: -14.61 +/- 1.14%, P<0.001 385 compared to vehicle (-1.88 +/- 0.89%) and P>0.05 compared to WIN 100nM after 3h THL 386 1µM (-12.44 +/- 1.21%), one-way ANOVO followed by Newman-Keuls post-test) (Figure

387 5A2, 5C2). This suggests that a local 2-AG production drop, caused by THL pre-treatment, 388 was responsible for the somatodendritic signaling decrease, which indeed could be rescued by 389 2-AG (100 nM), applied for 10 minutes before WIN treatment (somata: -3.30 +/- 1.04%, 390 dendrites: -3.68 +/- 1.15%, axons: -11.80 +/- 1.55%; WIN responses were compared to

391 vehicle) (Figure 5B, 5B1, 5C, 5C1, 5C2). By itself, 2-AG used at 1µM is able to decrease PKA

392 activity in all neuronal compartments, with a stronger effect in axons as compared to the 393 somatodendritic compartment (somata: -2.58 +/- 0.74%, dendrites: -5.53 +/- 0.54%, axons: -

394 17.68 +/- 1.49%) (Figure 5C, 5C1, 5C2). To verify if the presence of local 2-AG is a general 395 requirement for somatodendritic CB1R activation, we tested two other, structurally different, 396 CB1R agonists: CP55,940 (CP) and Δ9-THC (THC), the psychoactive compound of 397 marijuana. In control neurons, the effect of CB1R activation with 100 nM CP was comparable 398 to WIN, with a decrease of PKA activity in both dendrites and axons as well as a stronger 399 amplitude in the axonal response compared to dendrites (somata: -0.95 +/- 0.40%; dendrites: -

400 2.10 +/- 0.68%; axons: -18.43 +/- 1.61%) (Figure 5C, 5C1, 5C2). However, blockade of 401 DAGL by THL pretreatment did not decrease the effect of CP (100 nM) in the 402 somatodendritic compartment. On the contrary, and according to our previous expectations 403 for WIN, this response was enhanced as compared to control neurons (somata: -5.03 +/- 404 1.15%, dendrites: -5.68 +/- 1.26%, axons: -19.23 +/- 1.36%; responses to CP 100nM after 3h 405 THL 1µM were compared to CP 100nM alone). Finally, treatment with THC (1 µM) also 406 decreased PKA activity in all neuronal compartments, with a stronger effect in the axon 407 compared to the somatodendritic compartment (somata: -1.48 +/- 0.35%, dendrites: -2.76 +/-

408 0.53%, axons: -15.75 +/- 2.10%) (Figure 5C, 5C1, 5C2). However, the somatodendritic effect 409 of 1µM THC was suppressed by THL pretreatment while it did not affect the axonal response 410 (somata: -0.28 +/- 0.69%, dendrites: 0.14 +/- 0.98%, axons: -14.22 +/- 2.49%; responses to

411 THC 1 µM after 3h THL 1µM were compared THC 1µM alone) (Figure 5C, 5C1, 5C2). Thus, 412 THC and WIN require local presence of 2-AG to activate somatodendritic CB1Rs. 413 In conclusion, targeting of CB1Rs by exogenous cannabinoids can have contrasted effects on 414 the mobilization of somatodendritic signaling pathways: these effects are highly shaped by 415 local presence of 2-AG which is necessary for both WIN and THC, but not for CP action in 416 this neuronal compartment. 417 418 419 Discussion 420 421 We developed a highly sensitive quantitative in vitro method to evaluate, for the first time to 422 our knowledge, the modulation of the cAMP/PKA signaling pathway downstream of an

423 endogenous Gi/o protein coupled receptor with sub-neuronal resolution. We measured 424 modulation of basal cAMP/PKA signaling after activation or blockade of endogenous CB1Rs, 425 in somata, dendrites and axons of well-differentated cultured rat hippocampal neurons. Our

426 results show that polarized distribution of two neuronal proteins, the endocannabinoid 427 synthesizing DAGLenzyme and the CB1R receptor, leads to previously unappreciated 428 quantitative sub-domain dependent differences in intraneuronal GPCR signaling. In axons, the 429 combined effect of high CB1R density and absence of DAGL activity leads both to elevated 430 response amplitude following agonist stimulation, as well as to a lack of constitutive 431 activation. In the somatodendritic compartment, relatively low CB1R density and high 432 DAGL activity, locally producing the membrane component endocannabinoid 2-AG, results 433 in constitutive activation of CB1R-activated signaling which is accompanied by relatively low 434 amplitude agonist-induced signaling responses. 435 In addition, we show that the 2-AG content of the somatodendritic plasma membrane has 436 contrasted effects on CB1R activation by various exogenous cannabinoid ligands: at the 437 ligand concentrations used in the present study, CP acts as a classical agonist while both WIN 438 and THC require the presence of endogenous 2-AG to efficiently activate CB1Rs. 439 440 CB1Rs constitutively inhibit cAMP/PKA signaling in the somatodendritic compartment 441 but not in the axon. Several studies reported that CB1Rs display constitutive activity in 442 neurons (Pan et al., 1998; Hillard et al., 1999) and notably, these receptors are constitutively 443 endocytosed in the somatodendritic compartment, but not in axons, due to basal activation 444 (Leterrier et al., 2006; Simon et al., 2013). Here we show that application of the inverse 445 agonist AM281 leads to a rapid increase in both somatodendritic cAMP concentration and 446 PKA activity, suggesting cell-autonomous constitutive CB1R activation in the 447 somatodendritic compartment but not in the axon. In non-polarized cells, constitutive CB1R 448 activity is highly diminished in the absence of endocannabinoid 2-AG (Turu et al., 2007). We 449 found here that DAGL is segregated in the somatodendritic compartment and its inhibition 450 removes the effect of AM281. Therefore, somatodendritic CB1Rs are constitutively activated 451 by a high-tone of locally produced 2-AG, and the lack of constitutive activity in the axon is 452 due to the absence of 2-AG. 453 Our results show important somatodendritic effects on cAMP/PKA regulation for an axonal 454 (i.e. presynaptic) receptor. Previously, CB1R-mediated somatodendritic slow self-inhibition 455 (SSI) was reported in neocortical interneurons (Bacci et al., 2004) and pyramidal neurons 456 (Marinelli et al., 2009). During SSI, activation-induced post-synaptic increase of calcium 457 stimulates somatodendritic DAGL, leading to local 2-AG production and cell-autonomous 458 activation of somatodendritic CB1Rs and G protein inwardly rectifying K+ (GIRK) channels

459 (Marinelli et al., 2008). Our results are coherent with these observations and extend the

460 mechanical understanding of the phenomenon. βγ subunits of Gi/o proteins may directly

461 activate GIRK channels (Lujan et al., 2009). Here we directly demonstrate that such Gi/o 462 proteins can be activated by CB1Rs in the somatodendritic region and we show that this 463 activation impacts on local cAMP and PKA activation levels. Therefore, it is likely that SSI- 464 inducing activation leads to a parallel decrease of somatodendritic cAMP levels and to PKA 465 inhibition. We also report that basal cell-autonomous production of 2-AG is both necessary

466 and sufficient to activate Gi/o proteins through CB1Rs to achieve measurable constitutive 467 inhibition of somatodendritic cAMP/PKA signaling. GPCRs may also display constitutive 468 activity due to conformational instability (Kenakin, 2004) and several studies reported that 469 CB1Rs may display constitutive activity in systems apparently free of endocannabinoids 470 (review in (Pertwee, 2005)). However, it is difficult to formally exclude the presence of 471 endocannabinoids, since these lipid molecules may be present in cell plasma membrane at 472 high levels even in non-stimulated neurons (Alger and Kim, 2011). 473 Here, our results indicate the complete elimination of measurable constitutive somatodendritic 474 CB1R activation after pharmacological inhibition of DAGL and the lack of constitutive 475 activation in the mature axon, where the absence of DAGL suggests low levels of membrane- 476 borne cell-autonomous 2-AG. However, a certain level of conformational instability may be 477 necessary to enable constitutive activation of CB1R by 2-AG. Alanine substitution of the 478 T210 residue, which is located in the 3rd transmembrane helix and is well-conserved in the 479 cannabinoid receptor family but absent in other class A GPCRs (D'Antona et al., 2006), 480 results in change of the CB1R conformational state (Simon et al., 2013) and yields a 481 hypoactive receptor form, which displays significantly lower constitutive activity but 482 preserves responsiveness to agonists (D'Antona et al., 2006). Overexpressed T210A mutant 483 CB1Rs accumulate on the somatodendritic surface because of reduced steady-state 484 endocytosis and this accumulation leads to elevated somatodendritic responses to WIN 485 treatment (Simon et al., 2013). To further understand the effect of conformational instability 486 induced by T210 on CB1R signaling, it would be useful in the future to induce the T210A 487 mutation in the endogenous CB1R through a genetic editing approach, in order to avoid the 488 putative effects of receptor overexpression on the signaling response. 489 490 CB1R activation by exogenous cannabinoids in axons differs from that in dendrites, 491 where local 2-AG modulates the response to agonists. Activation of endogenous CB1Rs 492 leads to a stronger decrease of PKA activity in axons compared to dendrites. This difference

493 is not due to the shape of neurites. CB1Rs are enriched in the axonal plasma membrane, 494 leading to approximately 10-fold more endogenous CB1Rs receptors at the plasma membrane 495 in axons as compared to dendrites (McDonald et al., 2007). Here, we observed that the 496 decrease of PKA activity after CB1R activation is about 3-fold stronger in the axon than in 497 dendrites. Thus, differences in sub-neuronal signaling and receptor density are in the same 498 range, suggesting that the main cause of the polarized signaling response is polarized CB1R 499 distribution. This polarized distribution of CB1Rs is driven by steady-state somatodendritic 500 activation and endocytosis, which is a result of local 2-AG production by DAGL. Therefore 501 our results suggest that the principal cause of polarized CB1R signaling responses is the 502 polarized spatial distribution (i.e. somatodendritic segregation) of DAGL. 503 Our results also indicate that inhibition of 2-AG synthesis prevents WIN-induced activation of 504 CB1Rs in the somatodendritic compartment, whereas, in the axon, absence of 2-AG leads to a 505 lack of constitutive activity but does not prevent activation by WIN. Presently, possible 506 interactions between 2-AG and WIN on CB1R activation are not clearly understood. CB1R 507 intramembrane loops were proposed to shape a binding pocket that 2-AG could reach 508 through a gap allowing lipidic ligands to enter from membrane bilayer, without need of 509 extracellular access (Hurst et al., 2013). Aminoalkylindole cannabinoids such as WIN bind at 510 a different site (McAllister et al., 2003; Hurst et al., 2013), so WIN could act as a positive 511 allosteric modulator for 2-AG, by increasing 2-AG-induced constitutive CB1R activation, 512 leading to enhanced inhibition of cAMP/PKA signaling in the somatodendritic compartment. 513 Interestingly, the agonist CP55,940 binds at a different site than WIN (Kapur et al., 2007) and 514 dissimilarly to WIN, CP55,940-mediated inhibition of somatodendritic PKA activity is 515 significantly stronger after DAGL inhibition. After DAGL inhibition, CB1R levels increase at 516 the somatodendritic plasma membrane because of reduced endocytic elimination (Turu et al., 517 2007) possibly explaining the enhanced CP effect in the somatodendritic compartment. In the 518 axon, CP-induced PKA activity decrease is not modified by THL, as DAGL is absent in this 519 compartment. Finally, the phytocannabinoid THC induces a decrease of PKA activity in the 520 somatodendritic compartment that is removed after DAGL inhibition. Therefore, neuronal 521 pharmacology of THC is similar to WIN but not to CP, suggesting that THC may also act as 522 an exogenous positive allosteric modulator, that amplifies the CB1R-activating effect of 523 locally produced 2-AG in the somatodendritic compartment. These surprising interactions 524 between 2-AG and exogenous cannabinoid ligands may result from changes in CB1R levels 525 on the somatodendritic surface but also from different, potentially overlapping and to date not 526 completely understood mechanisms, such as conformation-induced changes in ligand affinity

527 and efficiency and competition for ligand binding sites. Full comprehension of these effects 528 requires further technical development that, through enhancing the sensitivity of the 529 experimental approach presented here, may allow detailed pharmacological characterization 530 in the future, such as precise measurement of ligand affinity and efficacy, of endogenous 531 GPCR signaling in neuronal sub-domains. 532 In conclusion, our results show that pharmacological responses to activation of a major 533 neuronal GPCR are different in axons and dendrites. In the somatodendritic compartment, 534 CB1Rs are constitutively activated by locally produced 2-AG, constitutively inhibit the 535 cAMP/PKA pathway and can be further activated, significantly, albeit moderately, by 536 exogenous cannabinoids. A similar activation profile was reported in non-polarized cells 537 (Turu et al., 2007). However, the pharmacological profile of axonal CB1Rs is different: their 538 activation leads to a strong decrease of PKA activity and no constitutive activity is observed. 539 This highly contrasted difference in sub-neuronal signaling responses warrants caution in 540 extrapolating pharmacological profiles, which are typically obtained in non-polarized cells, to 541 predict in vivo responses of axonal (i.e. presynaptic) GPCRs. Therefore, the in situ 542 pharmacological approach presented in our study may also be useful for a better 543 understanding of the physiology of other neuronal GPCRs. 544

545 References 546 Alger BE, Kim J (2011) Supply and demand for endocannabinoids. Trends Neurosci 34:304- 547 315. 548 Bacci A, Huguenard JR, Prince DA (2004) Long-lasting self-inhibition of neocortical 549 interneurons mediated by endocannabinoids. Nature 431:312-316. 550 Bisogno T, Howell F, Williams G, Minassi A, Cascio MG, Ligresti A, Matias I, Schiano- 551 Moriello A, Paul P, Williams EJ, Gangadharan U, Hobbs C, Di Marzo V, Doherty P 552 (2003) Cloning of the first sn1-DAG lipases points to the spatial and temporal 553 regulation of endocannabinoid signaling in the brain. J Cell Biol 163:463-468. 554 Bodor AL, Katona I, Nyiri G, Mackie K, Ledent C, Hajos N, Freund TF (2005) 555 Endocannabinoid signaling in rat somatosensory cortex: laminar differences and 556 involvement of specific interneuron types. J Neurosci 25:6845-6856. 557 Castro LR, Gervasi N, Guiot E, Cavellini L, Nikolaev VO, Paupardin-Tritsch D, Vincent P 558 (2010) Type 4 phosphodiesterase plays different integrating roles in different cellular 559 domains in pyramidal cortical neurons. J Neurosci 30:6143-6151. 560 Coutts AA, Anavi-Goffer S, Ross RA, MacEwan DJ, Mackie K, Pertwee RG, Irving AJ 561 (2001) Agonist-induced internalization and trafficking of cannabinoid CB1 receptors 562 in hippocampal neurons. J Neurosci 21:2425-2433. 563 D'Antona AM, Ahn KH, Kendall DA (2006) Mutations of CB1 T210 produce active and 564 inactive receptor forms: correlations with ligand affinity, receptor stability, and 565 cellular localization. Biochemistry 45:5606-5617. 566 Depry C, Allen MD, Zhang J (2011) Visualization of PKA activity in plasma membrane 567 microdomains. Mol Biosyst 7:52-58. 568 Freund TF, Katona I, Piomelli D (2003) Role of endogenous cannabinoids in synaptic 569 signaling. Physiol Rev 83:1017-1066. 570 Hillard CJ, Muthian S, Kearn CS (1999) Effects of CB(1) cannabinoid receptor activation on 571 cerebellar granule cell nitric oxide synthase activity. FEBS Lett 459:277-281. 572 Horton AC, Ehlers MD (2003) Neuronal polarity and trafficking. Neuron 40:277-295. 573 Howlett AC (2005) Cannabinoid receptor signaling. Handb Exp Pharmacol:53-79. 574 Howlett AC, Reggio PH, Childers SR, Hampson RE, Ulloa NM, Deutsch DG (2011) 575 Endocannabinoid tone versus constitutive activity of cannabinoid receptors. Br J 576 Pharmacol 163:1329-1343.

577 Hurst DP, Schmeisser M, Reggio PH (2013) Endogenous lipid activated G protein-coupled 578 receptors: emerging structural features from crystallography and molecular dynamics 579 simulations. Chem Phys Lipids 169:46-56. 580 Kano M, Ohno-Shosaku T, Hashimotodani Y, Uchigashima M, Watanabe M (2009) 581 Endocannabinoid-mediated control of synaptic transmission. Physiol Rev 89:309-380. 582 Kapur A, Hurst DP, Fleischer D, Whitnell R, Thakur GA, Makriyannis A, Reggio PH, Abood 583 ME (2007) Mutation studies of Ser7.39 and Ser2.60 in the human CB1 cannabinoid 584 receptor: evidence for a serine-induced bend in CB1 transmembrane helix 7. Mol 585 Pharmacol 71:1512-1524. 586 Katona I, Sperlagh B, Sik A, Kafalvi A, Vizi ES, Mackie K, Freund TF (1999) 587 Presynaptically located CB1 cannabinoid receptors regulate GABA release from axon 588 terminals of specific hippocampal interneurons. J Neurosci 19:4544-4558. 589 Katona I, Rancz EA, Acsady L, Ledent C, Mackie K, Hajos N, Freund TF (2001) Distribution 590 of CB1 cannabinoid receptors in the amygdala and their role in the control of 591 GABAergic transmission. J Neurosci 21:9506-9518. 592 Katona I, Urban GM, Wallace M, Ledent C, Jung KM, Piomelli D, Mackie K, Freund TF 593 (2006) Molecular composition of the endocannabinoid system at glutamatergic 594 synapses. J Neurosci 26:5628-5637. 595 Kenakin T (2004) Principles: receptor theory in pharmacology. Trends Pharmacol Sci 25:186- 596 192. 597 Klarenbeek JB, Goedhart J, Hink MA, Gadella TW, Jalink K (2011) A mTurquoise-based 598 cAMP sensor for both FLIM and ratiometric read-out has improved dynamic range. 599 PLoS One 6:e19170. 600 Leterrier C, Laine J, Darmon M, Boudin H, Rossier J, Lenkei Z (2006) Constitutive activation 601 drives compartment-selective endocytosis and axonal targeting of type 1 cannabinoid 602 receptors. J Neurosci 26:3141-3153. 603 Lujan R, Maylie J, Adelman JP (2009) New sites of action for GIRK and SK channels. Nat 604 Rev Neurosci 10:475-480. 605 Marinelli S, Pacioni S, Cannich A, Marsicano G, Bacci A (2009) Self-modulation of 606 neocortical pyramidal neurons by endocannabinoids. Nat Neurosci 12:1488-1490. 607 Marinelli S, Pacioni S, Bisogno T, Di Marzo V, Prince DA, Huguenard JR, Bacci A (2008) 608 The endocannabinoid 2-arachidonoylglycerol is responsible for the slow self- 609 inhibition in neocortical interneurons. J Neurosci 28:13532-13541.

610 McAllister SD, Rizvi G, Anavi-Goffer S, Hurst DP, Barnett-Norris J, Lynch DL, Reggio PH, 611 Abood ME (2003) An aromatic microdomain at the cannabinoid CB(1) receptor 612 constitutes an agonist/inverse agonist binding region. J Med Chem 46:5139-5152. 613 McDonald NA, Henstridge CM, Connolly CN, Irving AJ (2007) An essential role for 614 constitutive endocytosis, but not activity, in the axonal targeting of the CB1 615 cannabinoid receptor. Mol Pharmacol 71:976-984. 616 Neves SR, Tsokas P, Sarkar A, Grace EA, Rangamani P, Taubenfeld SM, Alberini CM, 617 Schaff JC, Blitzer RD, Moraru, II, Iyengar R (2008) Cell shape and negative links in 618 regulatory motifs together control spatial information flow in signaling networks. Cell 619 133:666-680. 620 Nomura DK, Morrison BE, Blankman JL, Long JZ, Kinsey SG, Marcondes MC, Ward AM, 621 Hahn YK, Lichtman AH, Conti B, Cravatt BF (2011) Endocannabinoid hydrolysis 622 generates brain prostaglandins that promote neuroinflammation. Science 334:809-813. 623 Pan X, Ikeda SR, Lewis DL (1998) SR 141716A acts as an inverse agonist to increase 624 neuronal voltage-dependent Ca2+ currents by reversal of tonic CB1 cannabinoid 625 receptor activity. Mol Pharmacol 54:1064-1072. 626 Pertwee RG (2005) Inverse agonism and neutral antagonism at cannabinoid CB1 receptors. 627 Life Sci 76:1307-1324. 628 Phillips R, Ursell T, Wiggins P, Sens P (2009) Emerging roles for lipids in shaping 629 membrane-protein function. Nature 459:379-385. 630 Simon AC, Loverdo C, Gaffuri AL, Urbanski M, Ladarre D, Carrel D, Rivals I, Leterrier C, 631 Benichou O, Dournaud P, Szabo B, Voituriez R, Lenkei Z (2013) Activation- 632 dependent plasticity of polarized GPCR distribution on the neuronal surface. J Mol 633 Cell Biol 5:250-265. 634 Thibault K, Carrel D, Bonnard D, Gallatz K, Simon A, Biard M, Pezet S, Palkovits M, Lenkei 635 Z (2013) Activation-dependent subcellular distribution patterns of CB1 cannabinoid 636 receptors in the rat forebrain. Cereb Cortex 23:2581-2591. 637 Turu G, Simon A, Gyombolai P, Szidonya L, Bagdy G, Lenkei Z, Hunyady L (2007) The role 638 of diacylglycerol lipase in constitutive and angiotensin AT1 receptor-stimulated 639 cannabinoid CB1 receptor activity. J Biol Chem 282:7753-7757. 640 Uchigashima M, Narushima M, Fukaya M, Katona I, Kano M, Watanabe M (2007) 641 Subcellular arrangement of molecules for 2-arachidonoyl-glycerol-mediated 642 retrograde signaling and its physiological contribution to synaptic modulation in the 643 striatum. J Neurosci 27:3663-3676.

644 Yoshida T, Fukaya M, Uchigashima M, Miura E, Kamiya H, Kano M, Watanabe M (2006) 645 Localization of diacylglycerol lipase-alpha around postsynaptic spine suggests close 646 proximity between production site of an endocannabinoid, 2-arachidonoyl-glycerol, 647 and presynaptic cannabinoid CB1 receptor. J Neurosci 26:4740-4751. 648 Yoshida T, Uchigashima M, Yamasaki M, Katona I, Yamazaki M, Sakimura K, Kano M, 649 Yoshioka M, Watanabe M (2011) Unique inhibitory synapse with particularly rich 650 endocannabinoid signaling machinery on pyramidal neurons in basal amygdaloid 651 nucleus. Proc Natl Acad Sci U S A 108:3059-3064. 652 653 654

655 Figure Legends 656 657 Figure 1: Quantitative measure of basal cAMP/PKA pathway modulation downstream 658 of endogenous neuronal CB1Rs in small cytoplasmic volumes. 659 A-D: Cultured hippocampal neurons expressing soluble (cytoplasmic) DsRed2 and various 660 FRET probes designed to measure cAMP concentration or PKA activity: TEpacVV (A), 661 AKAR4 (B), AKAR4-Kras (C), LYN-AKAR4 (D). After fixation, confocal imaging at two 662 different optical sections shows sub-cellular localization of the probes. AKAR4-Kras probes 663 are well-localized to the plasma membrane in the somatodendritic region. 664 E-F: Modulation of basal PKA activity downstream of endogenous CB1Rs in large axonal 665 Regions of Interest (ROI, orange) in AKAR4-Kras expressing neurons. The mean FRET ratio 666 is shown at 4 minutes (-t4) before (E1, F1) and at 6 minutes (t6) after (E2, F2) the addition of 667 treatment at t0. Incubation with vehicle does not change the FRET ratio (E2) compared to 668 baseline (E1) but addition of agonist WIN 55-212,2 (WIN) 100nM induces a rapid FRET- 669 ratio decrease (F2), as compared to baseline (F1). 670 G: Test of FRET imaging sensitivity by determining the smallest axonal cytoplasmic volume 671 allowing the measurement of significant PKA activity decrease after WIN-induced activation 672 of endogenous CB1Rs. We calculated the mean value of the FRET response amplitude 673 normalized to baseline (Amp) and its standard deviation (SD), in different axonal ROIs, 674 between t4 (4 minutes after drug treatment) and t14. The ratio of the FRET response 675 amplitude to its standard deviation (Amp/SD) is represented in function of the volume (see 676 text). The WIN effect is significantly different from control (modeled as an effect of Amp=0 677 with the same standard deviation than the corresponding WIN-stimulated response) at the 678 Amp/SD ratio equal to -0.91 (in grey, p<0.05, Student's t-test, N=10), which is reached 679 starting from ~1µm3 axonal volume. 680 Data information: Scale bar: 10µm (A, B, C, D, E, F) 681 682 Figure 2: Somatodendritic CB1Rs constitutively inhibit the cAMP/PKA pathway. 683 A-B: Two representative neurons expressing the membrane-targeted PKA sensor AKAR4- 684 Kras. The mean FRET ratio in somatic ROIs (orange) is shown at 4 minutes (-t4) before (A1, 685 B1) and at 6 minutes (t6) after the addition of treatment at t0: Vehicle (A2) or agonist WIN 686 55-212,2 (WIN) 100nM (B2). 687 C-F: Averaged responses of AKAR4-Kras (C, D) or the cAMP sensor TEpacVV expressing 688 neurons (E, F). The FRET ratio normalized to baseline was calculated for each neuron with a

689 time-resolution of 2 minutes, separately in somata and dendrites. The curves represent mean 690 +/- SEM of the FRET ratio for all imaged neurons at each time point. Addition of agonist 691 WIN 100 nM but not of vehicle at t0 results in rapid FRET ratio decrease while inverse- 692 agonist AM281 100 nM (AM) treatment results in elevated PKA-activation. At 30 minutes, 693 adenylyl-cyclase activator Forskolin (Fsk) was added at 10 µM, inducing a saturating increase 694 of the FRET ratio. C1, D1, E1, F1: Zoom between -t14 and t30 of C,D,E,F, respectively, 695 shows significant modulation of basal PKA activity after activation or blockade of CB1Rs. 696 C2, D2, E2, F2: FRET responses were calculated as the mean response between t4 and t14 697 minutes (shaded zone labeled "Response" on C1, D1, E1, F1), using data normalized to the 698 baseline (shaded zone between -t14 and -t2, labeled "Baseline" on C1, D1, E1, F1), as

699 described in the Materials and Methods section. Implication of Gi/o-proteins was shown by the 700 specific inhibitor pertussis toxin (PTX), applied overnight at 100ng/mL before the beginning 701 of the experiment. The WIN effect was CB1R-induced as shown by pre-treatment with the 702 CB1R-specific antagonist AM281 (1µM 3 hours before the beginning of the experiment). 703 Data information: Data are expressed as mean +/- SEM; Statistical analysis was realized with 704 one-way ANOVA followed by Newmann-Keuls post-test; NS p>0.05, * p<0.05, ** p<0.01, 705 *** p<0.001. Scale bar: 10µm (A, B). 706 707 Figure 3: Axonal CB1R signaling differs from dendritic signaling. 708 A: Averaged axonal responses of AKAR4-Kras expressing neurons, as shown on Fig.1E-F. 709 The FRET ratio normalized to baseline was calculated for each neuron with a time-resolution 710 of 2 minutes. The curves represent mean +/- SEM of the FRET ratio for all imaged neurons at 711 each time point. Addition of agonist WIN 100 nM but not of vehicle or inverse-agonist 712 AM281 100 nM (AM) at t0 results in rapid high-amplitude FRET ratio decrease. At 30 713 minutes, adenylyl-cyclase activator Forskolin (Fsk) was added at 10 µM, inducing a 714 saturating increase of the FRET ratio. A1: Zoom between -t14 and t30 of A shows significant 715 modulation of basal PKA activity after activation of CB1Rs. A2: FRET responses were 716 calculated as the mean response between t4 and t14 minutes (shaded zone labeled "Response" 717 on A1), using data normalized to the baseline (shaded zone between -t14 and -t2, labeled

718 "Baseline" on A1). Implication of Gi/o-proteins was shown by the specific inhibitor pertussis 719 toxin (PTX), applied overnight at 100ng/mL before the beginning of the experiment. The 720 WIN effect was CB1R-induced as shown by pre-treatment with the CB1R-specific antagonist 721 AM281 (1µM 3 hours before the beginning of the experiment). 722

723 B: Vehicle-normalized FRET response to WIN is significantly stronger in axons than in 724 dendrites. 725 C: Individual FRET responses in axons and distal dendrites are represented in function of 726 their respective diameter. For each group (distal dendrites and axons), a Pearson correlation

727 test was calculated showing no correlation between FRET response and diameter (rdistal

728 dendrites=-0.065 and raxons=0.03649). 729 (D) Distal dendrites having the similar diameter than axons still display significantly weaker 730 vehicle-normalized FRET responses to WIN compared to axons. 731 Data information: Data are expressed as means +/- SEM; Statistical analysis was realized with 732 one-way ANOVA followed by Newmann-Keuls post-test (A2) or unpaired t-test (B, D); NS 733 p>0.05, *** p<0.001. 734 735 Figure 4: Constitutive activation of somatodendritic CB1Rs requires locally synthesized 736 endocannabinoids. 737 A: Simultaneous immunolabeling of fully-polarized (DIV9) neurons with anti-DAGLα 738 antibody and either anti-MAP2 (A) or anti-Tau (A1) antibodies. Large arrows indicate 739 dendrites, arrow-heads indicate axon and the thin arrow shows an astrocyte. 740 B-C: Averaged somatic and dendritic responses of AKAR4-Kras expressing neurons to 741 inverse-agonist AM281 (AM), with our without inhibiting DAGL activity. The FRET ratio 742 normalized to baseline was calculated for each neuron with a time-resolution of 2 minutes. 743 The curves represent mean +/- SEM of the FRET ratio for all imaged neurons at each time 744 point. Addition of 100 nM AM but not of vehicle at t0 results in elevated PKA activity, 745 revealing constitutive CB1R activation, which is significantly decreased after DAGL 746 inhibition either by tetrahydrolipstatin (THL) 1µM or RHC80267 25µM. 747 Data information: Data are expressed as means +/- SEM; Statistical analysis was realized with 748 one-way ANOVA followed by Newmann-Keuls post-test; * p<0.05, ** p<0.01. Scale bar: 749 20µm (A, A1). 750 751 Figure 5: Endogenous 2-AG significantly modifies CB1R responses to exogenous 752 cannabinoids. 753 A-B: Averaged somatic, dendritic and axonal responses of AKAR4-Kras expressing neurons 754 to agonist WIN 55-212,2 (WIN). The FRET ratio normalized to baseline was calculated for 755 each neuron with a time-resolution of 2 minutes. The curves represent mean +/- SEM of the 756 FRET ratio for all imaged neurons at each time point. Addition of WIN 100 nM but not of

757 vehicle at t0 results in decreased PKA activity, which effect is significantly inhibited after

758 DAGL inhibition by tetrahydrolipstatin (THL) 1µM in somata (A) and dendrites (A1) but not

759 in axons (A2). The effect of THL pre-treatment on the WIN effect in somata (B) and dendrites 760 (B1) can be rescued by appling 2-AG at 100 nM 10 minutes before WIN. 761 C: Bidirectional variation of neuronal 2-AG levels (similarly to A-B) differently modifies the 762 FRET responses to exocannabinoids WIN 55-212,2 100 nM (WIN), CP55,940 100nM (CP) 763 and ∆9-THC 1µM (THC), shown as the mean response between t4 and t14 minutes (shaded 764 zone labeled "Response" on A-B), using data normalized to the baseline (shaded zone

765 between -t14 and -t2, labeled "Baseline" on A-B) in somata (C), dendrites (C1) or axons (C2). 766 2-AG levels were reduced by THL 1µM, applied 3 hours before the beginning of the 767 experiment and elevated by 2-AG 100nM at 10 minutes before agonist treatment. 768 Data information: Data are expressed as means +/- SEM; Statistical analysis was realized with 769 unpaired t-test (2-AG) or one-way ANOVA followed by Newmann-Keuls post-test (WIN, CP 770 and THC); NS p>0.05, * p<0.05, ** p<0.01 and *** p<0.001. 771 772 773 774 775 776 777 778

25 A FRET Probe Cytoplasmic DsRed Overlay B FRET Probe Cytoplasmic DsRed Overlay VV Opticalsection 1 (0.3 µm(0.3 from coverslip) Opticalsection 1 (0.9 µm(0.9 from coverslip) 10 µm Epac AKAR4 T Opticalsection 2 (3.9 µm(3.9 from coverslip) Opticalsection 2 (3.9 µm(3.9 from coverslip)

C D Opticalsection 1 Opticalsection 1 (0.9 µm(0.9 from coverslip) (1.2 µm(1.2 from coverslip) Lyn -AKAR4 AKAR4-Kras Opticalsection 2 Opticalsection 2 (3.0 µm(3.0 from coverslip) (4.2 µm(4.2 from coverslip)

EVehicle F CB1R agonist

1.8 E1 E2 F1 F2 1.6 G

1.4 RatioFRET (%)

1.2 AKAR4-Kras AKAR4-Kras 1 p=0.05 ROI ROI 0.8

0.6

PKA 1.5 1.5

1.4 1.4

1.3 1.3

1.2 1.2 FRET RatioFRET (%) FRET RatioFRET (%) 1.1 1.1 Vehicle WIN 1.0 1.0 -10 0 10 20 30 -10 0 10 20 30 time (min) time (min)

Figure 1 A Vehicle B CB1R agonist

1.8 A1 A2 B1 B2 1.6

1.4 FRETRatio (%) AKAR4-Kras

1.2 AKAR4-Kras

ROI 0.8

0.6 ROI

1.20 1.30

1.15 1.25

1.10 1.20 FRET RatioFRET (%) Vehicle RatioFRET (%) WIN 1.05 1.15 -10 0 10 20 30 -10 0 10 20 30 10 µm time (min) time (min)

WIN 100nM WIN 100nM CSomata AM 100nM DDendrites ESomata AM 100nM F Dendrites Vehicle Vehicle Treatment Treatment Treatment Treatment Fsk 10µM Fsk 10µM Fsk 10µM Fsk 10µM 35 35 35 35 30 30 30 30 25 25 25 25 20 20 20 20 15 15 15 15 10 10 10 10 5 5 5 5 FRET RatioFRET (%) FRET RatioFRET (%)

0 RatioFRET (%) 0 0 RatioFRET (%) 0 -5 -5 -5 -5 -10 -10 -10 -10 -10 0 10 20 30 40 50 60 -10 0 10 20 30 40 50 60 -10 0 10 20 30 40 50 60 -10 0 10 20 30 40 50 60 time (min) time (min) time (min) time (min) E1 F1 C1 Treatment D1 Treatment Treatment Treatment 4 4 4 4 Baseline Baseline Response Baseline Response Baseline Response 3 Response 3 3 3 2 2 2 2 1 1 1 1 0 0 0 0 -1 -1 -1 -1 -2 -2 -2 -2

PKA -3 -3 -3 -3 -4 -4 cAMP -4 -4 FRET RatioFRET (%) FRET RatioFRET (%) FRET RatioFRET (%) -5 -5 -5 RatioFRET (%) -5 -6 -6 -6 -6 -7 -7 -10 0 10 20 30 -7 -7 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 time (min) time (min) time (min) time (min)

C2 D2 NS NS E2 F2 3 3 NS NS *** 2 2 4 4 * ** *** (%) (%) 3 *** 3 *** 1 *** 1 *** 2 2 0 0 1 1 -1 -1 0 0 -1 -2 -1 -2 -2

FRET Response FRET -2 FRET Response FRET FRET Response FRET

-3 Response FRET -3 -3 -3 -4

normalized normalized to(%) baseline -4 -4 normalized normalized to(%) baseline -4 baseline to normalized normalized to baseline to normalized

WIN (N=59) AM (N=48) Vehicle (N=51) AM (N=43) Vehicle (N=48) WIN (N=56) Vehicle (N=29) Vehicle (N=29) WIN 100nMAM (N=40) 100nM (N=36)PTX+WIN (N=29)AM+WIN (N=20) PTX+WIN (N=27)AM+WIN (N=20) WIN 100nM (N=39)AM 100nM (N=36)

Figure 2

WIN 100nM Somata Vehicle 1 Dendrites Axons AATHL 1µM + WIN 100nM A2

Treatment Treatment Treatment 4 4 Baseline Response 10 3 3 Baseline Response Baseline Response 2 2 5 1 1 0 0 0 -1 -1 PKA -2 -2 -5 -3 -3 -10 -4 -4 FRET RatioFRET (%) FRET RatioFRET (%) -5 RatioFRET (%) -5 -15 -6 -6 -7 -7 -10 0 10 20 30 -20 -10 0 10 20 30 -10 0 10 20 30 time (min) time (min) time (min)

BBSomata 1 Dendrites WIN 100nM WIN 100nM 2-AG 100nM 2-AG 100nM THL 1µM (3h) THL 1µM (3h) 5 5 Baseline Response Baseline Response

0 0

-5 -5 PKA FRET RatioFRET (%) FRET RatioFRET (%) -10 -10

-10 0 10 20 30 40 -10 0 10 20 30 40 time (min) time (min)

Vehicle without pretreatment C Somata THL 1µM C1 Dendrites THL 1µM + 2-AG 100nM

2 2-AG 1µM WIN 100nM CP 100nM THC 1µM 2-AG 1µM WIN 100nM CP 100nM THC 1µM ** 1 *** ** *** ** NS * ** ** ** *** * *** ** NS ** 0 N=46 N=46 N=46 N=46 N=27

-1 N=42 N=46 N=46 N=46 N=46 N=29 N=42 N=32 -2 N=33

-3 N=53 N=31 N=26 N=52 N=31 FRET Response FRET -4 N=27 normalized normalized to(%) baseline -5 N=27

-6 N=30 N=26 -7 N=28 C2 Axons

2-AG 1µM WIN 100nM CP 100nM THC 1µM PKA *** NS NS NS *** *** *** *** 0

N=39 N=39 N=39 N=39 -5

-10

N=18 -15 N=36

FRET Response FRET N=46 N=19 N=22 normalized normalized to(%) baseline -20 N=23 N=27 N=25

-25

Figure 5 4. Main results and remaining questions Measuring PKA activity in cultured hippocampal neurons, we demonstrated that endogenous CB1Rs display differential pharmacology between the somatodendritic compartment and the axon. Notably, somatodendritic CB1Rs are moderately activable by exogenous agonists leading to an inhibition of the cAMP/PKA pathway and constitutively inhibit this pathway due to basal DAGLα activity that synthesis the endocannabinoid 2-AG. Axonal CB1R activation induces a decrease of PKA activity that is stronger than the induced-decrease in dendrites and do not constitutively inhibit PKA activity due to lack of DAGLα in this compartment. These results are in accordance with previous results that demonstrate a differential traffic of CB1Rs between the somatodendritic compartment and the axon (see Part 1). Moreover, our study is the first demonstrating a modulation of cAMP/PKA pathway in the somatodendritic compartment downstream of CB1Rs. In addition, we demonstrated, for the first time to our knowledge, that 2-AG modulates the efficiency of several CB1R agonists: synthetic WIN55-212,2, CP55,940 and phytocannabinoid Δ9-THC. Indeed, in the somatodendritic compartment, basal tone of endogenous 2-AG is required for WIN55-212,2 and Δ9-THC effect but, on the contrary, decreases the effect of CP55,940. Interestingly, in axons, lack of 2-AG does not prevent CB1R activation by these agonists. These data strongly suggest that somatodendritic CB1Rs do not adopt the same conformations as axonal CB1Rs. Moreover, our results are the first evidence for polarized distribution of a lipid that modulate GPCR activity. Finally, our data prevent for extrapolation of pharmacological data obtained in non-polarized cells to polarized neurons.

However, some questions remain: - Can CB1R polarized pharmacology be applied to other neuronal GPCRs? - To what extent is the CB1R T210 residue involved in constitutive activation of somatodendritic receptors?

In order to answer these questions, we used the same FRET imaging technique and measured the modulation of basal PKA activity downstream of axonal 5-HT1B receptors (Part 3) and hypoactive CB1R-T210A mutant (Part 4) in cultured hippocampal neurons.

151

Part 3: Influence of neuronal polarity on the pharmacology of

serotoninergic 5-HT1B receptors.

152 1. Introduction Serotonin (5-HT) is a neurotransmitter widely present in the CNS. Synthesized by serotoninergic neurons and stored in vesicles, it is exocytosed in serotoninergic synapses. 5- HT is involved in numerous physiological functions, such as sleep-wake cycle control (Ursin, 2002), food intake (Curzon, 1990), and nociception (Oliveras & Besson, 1992). It also participates to the regulation of mood and emotions. In particular, serotoninergic system dysfunctions are linked to depression (Coppen & Doogan, 1988) and general anxiety (Iversen, 1984; Thiebot, 1986). Selective serotonin reuptake inhibitors (SSRIs) have demonstrated high efficiency as antidepressants (Fuller, 1992). Moreover, typical and atypical antipsychotics, such as clozapine, used to treat schizophrenia, are potent antagonists of 5-HT receptors subtypes (Hery & Hamon, 1993; Meltzer, 1995), suggesting 5-HT dysfunction in this pathology. The high amounts of physiological functions regulated by 5-HT is notably due to the high number and variety of identified receptors to this molecule. To date, fifteen 5-HT receptors have been identified (Barnes & Sharp, 1999; Boess & Martin, 1994).

Among these receptors, 5-HT1B receptors show interesting analogies with CB1Rs. Predominantly localized on the plasma membrane of mature axons, our group previously reported that, after their synthesis, these GPCRs appear transiently on the somatodendritic plasma membrane from where they are rapidly removed by constitutive endocytosis and targeted to the axon through transcytosis (Carrel et al, 2011). Moreover, axonal 5-HT1B receptors are not only found in terminals but also in the extrasynaptic regions (Sari et al, 1999). Differences in traffic between somatodendritic compartment and axon involve a differential steady-state endocytosis: mostly internalized in somata and dendrites, 5-HT1B receptors accumulate on axonal plasma membrane where endocytosis is reduced.

Importantly, the genuine constitutive activity of 5-HT1B receptors involves structural determinants, notably the complete third intracellular loop. In addition, these receptors are predominantly coupled to Gi/o proteins and their activation inhibits cAMP production (Bouhelal et al, 1988; Schoeffter & Hoyer, 1989). Thus, using the method that I developed to investigate sub-neuronal CB1R pharmacology, I proposed to answer the following questions:

- Do somatodendritic 5-HT1B receptors constitutively recruit signaling pathway in this compartment? - Are differential steady-state endocytosis between somatodendritic compartment and axon linked to differential pharmacology?

153 In order to answer these questions, I used the FRET imaging in live cultured hippocampal neurons of rats. Experiments were performed on cultured neurons expressing AKAR4-Kras probe to detect PKA activity. However, the high number of 5-HT receptors could lead to problems of specificity and to the measure of responses from several 5-HT receptor sub-types, while we are interested only in 5-HT1B receptor responses. To overcome this problem, I overexpressed 5-HT1B receptors in neurons by transfection. Moreover, I used Sumatriptan as agonist that is specific to 5-HT1B and 5-HT1D receptors and SB224289 as inverse agonist that is specific for 5-HT1B receptors.

2. Materials and Methods Animals Animal procedures were conducted in strict compliance with approved institutional protocols and in accordance with the provisions for animal care and use described in the European Communities council directive of November 24, 1986 (86 � 609 � EEC). SpragueDawley rats (Janvier) were used for dissociated cell culture experiments.

Chemicals and DNA constructs

5-HT1B agonist Sumatriptan and inverse agonist SB224289 were obtained from R&D Systems Europe. Dimethyl Sulfoxide (DMSO), Forskoline (Fsk), Bovine Serum Albumin (BSA) and Poly-D-Lysine were obtained from SIGMA-ALDRICH. B27, Lipofectamine 2000 and Neurobasal were obtained from Life Technologies. AKAR4-Kras probes were obtained from Dr Jin Zhangs laboratory (Baltimore, USA).

Flag-5HT1B construct was previously described (Carrel et al, 2006).

Hippocampal neuronal cultures Hippocampal neuronal cultures were performed essentially as described previously (Leterrier et al, 2006). Briefly, hippocampi of rat embryos were dissected at embryonic day 18. After trypsinization, dissociation was achieved with a fire-polished Pasteur pipette. Cells were counted and plated on poly-D-lysine-coated 18-mm diameter glass coverslips, at a density of 300400 cells/mm2. The plating medium was Neurobasal supplemented with 2% B27 and containing Stabilized Glutamine (0.5 mM) and penicillin G (10 U/ml)/streptomycin (10 g/ml). Four hours after plating, the coverslips were transferred into Petri dishes containing supplemented Neurobasal medium that had been conditioned for 24 h on a 80% confluent glia

154 layer. Neurons were transfected after 6 days in vitro (DIV) using Lipofectamine 2000, following the manufacturers instructions.

FRET imaging

Neurons cotransfected with AKAR4-KRAS and Flag-5HT1B were imaged by videomicroscopy between DIV8 and DIV11 on a motorized Nikon Eclipse Ti-E/B inverted microscope with the Perfect Focus System (PFS) in a 37°C thermostated chamber, using an oil immersion CFI Plan APO VC 60X, NA 1.4 objective (Nikon). Acquisitions were carried out at the excitation wavelength of the CFP (434nm +/- 15nm) using an Intensilight (Nikon). Emitted light passed through an Optosplit II beam-splitter (Cairn Research) equipped with a FF509-FDi01 dichroïc mirror, a FF01-483/32-25 CFP filter and a FF01-542/27-25 YFP filter and was collected by an EM-CCD camera (Evolve 512, Photometrics), mounted behind a 2x magnification lens. Acquisitions were performed by piloting the set-up with Metamorph 7.7 (Molecular Devices). All filter sets were purchased from Semrock. Cultured neurons on 18-mm coverslips were placed in a closed imaging chamber containing an imaging medium: 120 mM NaCl, 3 mM KCl, 10 mM HEPES, 2 mM CaCl2, 2 mM

MgCl2, 10 mM D-glucose, 2% B27, 0,001% BSA. The acquisition lasted 90 minutes registering one image each 2 minutes, registering in parallel 10 to 15 neurons on the same coverslip. 30 minutes after the beginning of the acquisition, pharmacological treatment was applied then 60 minutes after the beginning of the acquisition, Forskoline 10µM was applied (positive control, data not shown).

FRET data analysis Images were divided in two parts on ImageJ to separate the CFP channel from the YFP channel. Data were then analyzed on Matlab by calculating the FRET ratio at each time point for one or several Regions Of Interest (ROIs). The user defined ROIs for each position. For each image, the value of the FRET ratio corresponds to -IY BY C -I BC IY : Mean Intensity of ROI in YFP channel; BY : Mean Intensity of the background in YFP channel; IC : Mean Intensity of ROI in CFP channel; BC : value of the background in CFP channel

155 For each ROI, the FRET ratio was then normalized by the baseline mean, defined as the 7 Rc  Ro time points before first treatment injection. FRET Ratio  100 * Ro Rc : Value of crude FRET ratio Ro : Mean of the baseline The quantitative results obtained for each neuronal compartment were grouped together and, for each time point, the mean FRET ratio normalized to baseline and SEM were calculated. Deviation was corrected for somata and dendrites on Matlab. Mean slope was calcultated for all neurons in somata and dendrites, respectively, for the last 7 time points before addition of treatment and substracted from all FRET ratio time points.

Statistical analysis FRET Response was obtained by calculating the mean FRET ratio on Matlab for 6 time points after injection, from +4 minutes to +14 minutes (Response). Groups were compared using GraphPad Prism. Significance of differences between various conditions was calculated using unpaired t-tests or one-way ANOVA with Newman-Keuls post-tests for computing p estimates. NS p>0.05, * p<0.05, ** p<0.01 and *** p<0.001.

3. Results

3.1. Overexpressed 5-HT1B receptors constitutively inhibit PKA activity in all neuronal compartments.

5-HT1B receptors display high constitutive activity. Predominantly localized on the axonal plasma membrane, we reported previously their presence in the somatodendritic compartment and that their constitutive endocytosis in this compartment is required for axonal targeting through transcytosis (Carrel et al, 2011). First, we asked if these somatodendritic constitutively activated receptors recruit signaling pathways in this compartment.

Predominantly coupled to Gi/o proteins, 5-HT1B receptors are known to decrease cAMP concentration (Bouhelal et al, 1988; Schoeffter & Hoyer, 1989). Thus, we measured PKA activity in cultured hippocampal neurons overexpressing 5-HT1B receptors, by using the genetically encoded FRET probe AKAR4-Kras (Depry et al, 2011). As described previously, we are able to detect basal PKA activity modulation in non-overactivated systems in all neuronal compartments, including thin mature axons (see Part2, Ladarre et al, unpublished

156 data). During the acquisition, we applied either vehicle solution (Figure 45A), agonist Sumatriptan 500 nM (Figure 45B) or inverse agonist SB224289 500 nM (Figure 45C). In individual somata, compared to baseline (Figure 45A1-45B1-45C1), vehicle solution did not change the FRET ratio (Figure 45A2), while Sumatriptan induced a decrease of the FRET ratio (Figure 45B2) and SB224289 induced an increase of the FRET ratio (Figure 45C2). Measured in a population of neurons, inverse agonist SB224289 500nM induced a significant increase of PKA activity in all neuronal compartments (somata, dendrites and axons) compared to vehicle (responses in somata: SB224189: 2.13 +/- 1.21%, vehicle: -0.81 +/- 0.40%; responses in dendrites: SB224189: 4.97 +/- 1.26%, vehicle: -0.10 +/- 0.60%; responses in axons: SB224189: 5.58 +/- 2.71%, vehicle: -5.63 +/- 1.85%) (Figure 45D-45D1-

45E-45E1-45F-45F1). Thus, application of 5-HT1B inverse agonist SB224289 removes a constitutive inhibition on PKA activity in all neuronal compartments, demonstrating that 5-

HT1B receptors constitutively recruit signaling pathways.

3.2. Activation of 5-HT1B receptors leads to a stronger decrease of PKA activity in axons as compared to dendrites.

Previous studies reported a differential distribution of 5-HT1B receptors in neurons: in the somatodendritic compartment, receptors are predominantly internalized and appear only transiently on the plasma membrane, while axonal receptors are accumulated on the plasma membrane with higher concentration than in somata and dendrites (Carrel et al, 2011). Thus, we asked if activation of membrane 5-HT1B receptors lead to the same pharmacological response in the various neuronal compartments. In a population of neurons, the application of agonist Sumatriptan 500 nM led to a significant decrease of PKA activity in dendrites and axons compared to vehicle (responses to Sumatriptan in dendrites: -3.46 +/- 0.97%; in axons:

-23.24 +/- 2.84%) (Figure 45E-45E1-45F-45F1). However, in somata, no significant difference was measured as compared to vehicle (responses to Sumatriptan in somata: -1.37

+/- 0.59%) (Figure 45D-45D1). Interestingly, normalized to their respective vehicle, decrease of PKA activity in axons is significantly stronger than in dendrites (responses normalized to vehicle in dendrites: -3.34 +/- 0.97%; in axons: -17.61 +/- 2.84%) (Figure 45G), suggesting that the higher concentration of membrane 5-HT1B receptors in axons leads to a stronger effect of agonist on signaling pathways compared to dendrites.

157 4. Figures and legends

Figure 45: Serotoninergic receptors 5-HT1B modulate PKA activity differently in axons and dendrites.

Cultured neurons were cotransfected with AKAR4-Kras and Flag-5HT1B.

158 (A-B-C) Individual neuronal responses to vehicle (A), Sumatriptan 500nM (B) and SB224289 (C) were measured by FRET and presented as color ratio images. In somata, compared to baseline (A1-B1-C1), vehicle did not change the FRET ratio (A2), agonist Sumatriptan induced a FRET ratio decrease (B2) and inverse agonist SB224289 induced an increase of the FRET ratio (C2). (D-E-F) FRET ratio was calculated in a neuronal population. For each neuron, responses to

Vehicle, Sumatriptan 500 nM or SB224289 was measured in soma (D-D1), dendrites (E-E1) and axon (F-F1). Curves represent mean +/- SEM of the FRET ratio normalized to the baseline for each time point (D-E-F). Baseline and response interval are indicated. Responses to treatment were calculated as mean FRET ratio in the response interval for each neuron.

Mean +/- SEM are represented for somata (D1), dendrites (E1) and axons (F1). (G) FRET responses to Sumatriptan 500nM in dendrites and axons were normalized to their respective vehicle and compared.

Data information: Data were expressed as means +/- SEM; Statistical analysis were realized with one-way ANOVA followed by Newmann-Keuls post-test (D1-E1-F1) or unpaired t-test (G); NS p>0.05, * p<0.05, ** p<0.01, *** p<0.001. Scale bar: 10µm (A-B-C).

5. Discussion

Our results demonstrate that 5-HT1B receptors display differential pharmacology due to neuronal polarization. Indeed, activation of these receptors leads to stronger decrease of PKA activity in the axon compared to dendrites. By analogy with results obtained with CB1Rs, one can assume that this difference is due to polarized distribution of 5-HT1B receptors. However, contrary to CB1Rs, no significant decrease of PKA activity was observed in somata after addition of Sumatriptan. Two hypotheses could explain this result. First, one can assume that internalization of 5-HT1B receptors in somata is stronger than CB1R internalization rate. Thus, there are maybe not enough 5-HT1B receptors on the somatic plasma membrane to detect a decrease of PKA activity in this compartment after their activation. Secondly, overexpression of receptors may lead to higher constitutive inhibition of PKA activity, resulting in a very low basal PKA activity in somata. Thus, it could become difficult to detect further decrease.

Moreover, we demonstrated here that somatodendritic 5-HT1B receptors constitutively inhibit

PKA activity in this compartment. However, contrary to endogenous CB1Rs, 5-HT1B

159 receptors also constitutively recruit signaling pathways in axons, even through previous studies reported accumulation of receptors on plasma membrane in this compartment (Carrel et al, 2011; Sari et al, 1999). Lack of CB1R internalization in axons have been reported to be due to slow endocytosis in this compartment, maybe because of reduced endocytic material

(Leterrier et al, 2006). Thus, 5-HT1B receptors could constitutively recruit signaling pathways in axons, while they do not internalize due to slow endocytosis in this compartment.

Taken together, these data support that polarized distribution of 5-HT1B receptors leads to differential pharmacology in axons and dendrites. Thus, results previously obtained on CB1Rs may be generalized to other axonal GPCRs targeted through transcytosis.

However, these results have to be completed. Indeed, this series of experiments has been perfomed only twice. Thus, a third series has to be added to confirm these results. Moreover, very important controls are lacking, such as the validation of Gi/o protein involvement (with a pertussis toxin treatment) and validation of Sumatryptan selectivity for 5-HT1B receptors (with SB224289 pretreatment).

160

Part 4: Structural determinants for the constitutive activity of type-1 cannabinoid receptors - Role of T210.

161 1. Introduction In the Part 2, we demonstrated that neuronal polarity influences CB1R pharmacology. Indeed, activation of endogenous CB1Rs induces a decrease of PKA activity in both somatodendritic compartment and axon with a significantly stronger effect in the axon that can be linked to higher receptor density at the plasma membrane. Moreover, somatodendritic CB1Rs constitutively inhibit PKA activity in somata and dendrites but not in axons that is due to somatodendritic presence of endocannabinoid 2-AG. Interestingly, the group of Kendall demonstrated that Threonine 210 (T210) plays a major role for the passage from inactive to active conformations of CB1R. Indeed, one-point mutation of T210 in alanine (T210A-CB1R) displays a reduced activability by agonists compared to wild-type receptors (WT-CB1Rs) (D'Antona et al, 2006). Moreover, expression of T210A-CB1Rs in HEK 293 cells increases cAMP concentration in these cells and potentiate forskolin stimulation as compared to WT-CB1Rs (D'Antona et al, 2006). These results suggest that T210A-CB1Rs are hypoactive receptors with reduced constitutive adenylyl cyclase inhibition. Moreover, our group also demonstrated that T210A-CB1Rs display an increased presence on somatodendritic plasma membrane compared to WT- CB1Rs, demonstrating that somatodendritic constitutive endocytosis is reduced with this mutant (see Part 1, (Simon et al, 2013)). Taken together, T210 appears as a structural determinant for constitutive CB1R activation. How this structural determinant affects recruitment of signaling pathways downstream of CB1Rs in various neuronal compartments? In order to answer this question, I used cultured hippocampal neurons of rats expressing AKAR4-Kras probe and performed FRET imaging to measure basal PKA activity downstream of T210A-CB1Rs. As a control, I used neurons that overexpressed WT-CB1Rs. Finally, I compared the responses obtained with T210A-CB1Rs, WT-CB1Rs and endogenous CB1Rs.

2. Materials and Methods See Part 2 (Ladarre et al., submitted manuscript). Flag-CB1R-WT and Flag-CB1R-T210A were obtained as previously described (Leterrier et al, 2004; Leterrier et al, 2006).

162 3. Results 3.1. Overexpressed wild-type CB1Rs constitutively inhibit PKA activity in all neuronal compartments independently of 2-AG presence. First, we overexpressed wild-type CB1Rs (WT-CB1Rs) in cultured hippocampal neurons and measured by FRET how they modulated PKA activity in all neuronal compartments (soma, dendrites and axon) (Figure 46A-46A1-46B-46B1-46C-46C1). Addition of WIN 100nM led to a rapid and significant decrease of FRET ratio as compared to vehicle in somata (response to WIN: -2.27 +/- 0.62%; response to vehicle: 0.45 +/- 0.41%), dendrites (response to WIN: - 2.17 +/- 0.62%; response to vehicle: 0.88 +/- 0.48%) and axons (response to WIN: -9.72 +/- 1.34%; response to vehicle: -1.92 +/- 0.80%). Thus, activation of WT-CB1Rs leads to a decrease of PKA activity in all neuronal compartments. Moreover, application of the inverse agonist AM281 100nM led to a rapid and significant increase of the FRET ratio as compared to vehicle in somata (3.82 +/- 0.69%), dendrites (6.36 +/- 0.73%) and strikingly also in axons (15.25 +/- 1.73%). Thus, inverse agonist removes a CB1Rs constitutive inhibition of PKA activity. We reported previously that endogenous CB1Rs constitutively inhibit PKA activity in somatodendritic compartment due to high basal tone of 2-AG, segregated in this compartment, and do not display constitutive activity in axons, where DAGL is absent. Thus, we tested whether the somatodendritic constitutive activity of overexpressed CB1Rs is also mediated by constitutive 2-AG production. Thus, we treated neurons with THL 1µM, a DAGL inhibitor, during 3 hours before acquisition and added AM281 100nM during FRET acquisition. Interestingly, we found that THL pretreated neurons display a significant increase of PKA activity in both somata (4.28 +/- 0.76%) and dendrites (4.62 +/- 0.67%) as compared to vehicle (Figure 46A-46A1-46B-46B1). Thus, overexpressed WT-CB1Rs constitutively inhibit PKA activity in the somatodendritic compartment, even after DAGL inhibition, and in axon, where DAGL is absent. These results show that overexpressed WT-CB1Rs constitutively inhibit PKA activity independently of the presence of 2-AG.

3.2. Overexpressed hypoactive CB1Rs do not display constitutive activity, even in the somatodendritic compartment where 2-AG is present. As WT-CB1Rs constitutively inhibit PKA activity even in absence of 2-AG, we asked if the structural determinant for CB1R activation T210 (D'Antona et al, 2006) is responsible for this constitutive activity. T210A mutant has been described to display reduced constitutive activity compared to WT-CB1Rs in neurons, leading to an increased presence at the

163 somatodendritic plasma membrane (Simon et al, 2013). Thus, we overexpressed T210A-

CB1Rs in cultured hippocampal neurons of rats (Figure 46D-46D1-46E-46E1-46F-46F1). In these neurons, addition of WIN 100 nM induced a rapid and significant decrease of PKA activity as compared to vehicle in somata (response to WIN: -2.12 +/- 0.34%; response to vehicle: -0.15 +/- 0.51%), dendrites (response to WIN: -3.53 +/- 0.47%; response to vehicle: 0.10 +/- 0.74%) and axons (response to WIN: -15.05 +/- 1.88%; response to vehicle: -2.79 +/- 3.17%). Moreover, addition of the inverse agonist AM281 100nM did not change PKA activity as compared to vehicle neither in somata (0.15 +/- 0.39%), nor in dendrites (0.03 +/- 0.68%), nor in axons (-1.57 +/- 1.44%). Thus, T210A-CB1Rs do not constitutively inhibit PKA activity, even in somatodendritic compartment, where DAGL and 2-AG are present.

3.4. Activity pattern of endogenous CB1Rs is similar to overexpressed WT-CB1Rs in dendrites and to overexpressed T210A-CB1Rs in axons. Finally, we compared the results obtained with WT-CB1R, T210A-CB1R and endogenous CB1Rs (Figure 46G-46H). In dendrites, activation of endogenous CB1Rs led to decrease of PKA activity, such as overexpressed WT-CB1Rs and T210A-CB1Rs (responses to WIN normalized by vehicle for endogenous CB1Rs: -3.07 +/- 0.52%; for WT-CB1Rs: -3.05 +/- 0.77%; for T210A-CB1Rs: -3.63 +/- 0.47%) (Figure 46G). Application of the inverse agonist AM281 on endogenous CB1Rs led to a significantly lower increase of PKA activity than overexpressed WT-CB1Rs but significantly higher than overexpressed T210A-CB1Rs (responses to AM normalized by vehicle for endogenous CB1Rs: 2.16 +/- 0.45%; for WT- CB1Rs: 5.48 +/- 0.73%; for T210A-CB1Rs: -0.08 +/- 0.68%). Thus, activation profile of endogenous CB1Rs is similar to WT-CB1Rs in dendrites. WT-CB1Rs display higher constitutive activity, probably due to their overexpression. In axons, endogenous CB1Rs activation led to a strong decrease of PKA activity, in the same range as T210A-CB1Rs and significantly stronger than WT-CB1Rs (responses to WIN normalized by vehicle for endogenous CB1Rs: -12.83 +/- 1.38%; for WT-CB1Rs: -7.80 +/- 1.34%; for T210A-CB1Rs: - 12.27 +/- 1.88%) (Figure 46H). Moreover, application of the inverse agonist AM281 on endogenous CB1Rs led to no increase of PKA activity, such as T210A-CB1Rs, while WT- CB1Rs display high constitutive activity in this compartment (responses to AM normalized by vehicle for endogenous CB1Rs: 1.89 +/- 0.83%; for WT-CB1Rs: 17.17 +/- 1.73%; for T210A-CB1Rs: 1.22 +/- 1.44%). Thus, activation profile of endogenous CB1Rs in the axon is similar to T210A-CB1Rs, demonstrating a reduced constitutive activity in this compartment.

164 4. Figures and legends

Figure 46: T210 is a structural determinant for CB1R constitutive activity. (legend: see next page)

165 (A-B-C) Cultured neurons were cotransfected with AKAR4-Kras and Flag-CB1-WT. FRET was measured in a population of neurons in somata (A), dendrites (B) and axons (C) simultaneously. During the acquisition, either Vehicle or WIN 100nM or AM 100nM was added. Curves represent mean +/- SEM of the FRET ratio normalized to baseline for each time point (A-B-C). Baseline and response intervals are indicated. Responses to treatment were calculated as mean FRET ratio in response interval for each neuron. Mean +/- SEM are represented for somata (A1), dendrites (B1) and axons (C1). THL treatment was applied 3 hours before acquisition and AM added during the acquisition; calculated response is the response to AM. (D-E-F) Cultured neurons were cotransfected with AKAR4-Kras and Flag-CB1-T210A. FRET was measured in a population of neurons in somata (D), dendrites (E) and axons (F) simultaneously. During the acquisition, either Vehicle or WIN 100nM or AM 100nM was added. Curves represent mean +/- SEM of the FRET ratio normalized to baseline for each time point (D-E-F). Baseline and response intervals are indicated. Responses to treatment were calculated as mean FRET ratio in response interval for each neuron. Mean +/- SEM are represented for somata (D1), dendrites (E1) and axons (F1). (G-H) Cultured neurons were transfected with AKAR4-Kras only (endogenous) or with AKAR4-Kras and Flag-CB1-WT or Flag-CB1-T210A. Comparison of responses in dendrites (G) and axons (H) between endogenous CB1Rs, overexpressed CB1R-WT and overexpressed CB1R-T210A was performed, normalizing all responses to WIN and AM to their respective vehicle.

Data information: Data were expressed as means +/- SEM; Statistical analysis were realized with one-way ANOVA followed by Newmann-Keuls post-test (A1-B1-C1-D1-E1-F1-G-H); NS p>0.05, * p<0.05, ** p<0.01, *** p<0.001.

5. Discussion 5.1. Receptor overexpression changes activation profile. Using overexpressed WT-CB1Rs as a control, we expected to obtain the same activation profile as endogenous receptors. However, contrary to endogenous CB1Rs, overexpressed WT-CB1Rs constitutively inhibit PKA activity in all neuronal compartments, even in axon, independently of the presence of 2-AG. This result could be due to the overexpression.

166 Indeed, overexpressing a receptor increases number of both inactive and active receptors, leading to a possible detection of constitutive activity that is too low to be detected with endogenous receptors. Moreover, we measured a strong constitutive inhibition of PKA activity in axons, while previous studies clearly demonstrated a very low proportion of internalized receptors in this compartment (Simon et al, 2013). This result is comparable with those obtained with serotoninergic 5-HT1B receptors (see Part 3). Thus, one can assume that the endocytotic machinery is reduced in axons, avoiding receptor internalization, but constitutive recruitment of signaling pathways can still occur. Interestingly, while the response to the inverse agonist AM281 in axons is stronger with overexpressed WT-CB1Rs compared to endogenous receptors, the response to the agonist WIN is diminished. One can assume that high constitutive recruitment of signaling pathways decreases the efficiency of the agonist for two reasons: 1) when agonist is added, high proportion of receptors are already signaling and thus can not be activated by the agonist; 2) high constitutive activity results in lower basal PKA activity that can not decrease as much as in a system where basal PKA activity is higher, e.g. with endogenous receptors that do not constitutively inhibit this pathway in axons.

5.2. T210 is required for CB1R activation by 2-AG. T210A-CB1Rs do not display constitutive activity in all neuronal compartments. However, we demonstrated previously that endogenous receptors are constitutively activated in somatodendritic compartment, due to the presence of DAGL that synthesizes the endocannabinoid 2-AG. Thus, even in the somatodendritic compartment, where 2-AG is present, T210A-CB1Rs do not display constitutive activity. These results demonstrate that T210 is necessary for CB1R activation by 2-AG. Previous studies showed that T210A-CB1Rs are activable by agonists (D'Antona et al, 2006; Simon et al, 2013) but with a reduced affinity. Here, we demonstrate that the endogenous concentration of 2-AG present in neuronal plasma membrane is not sufficient to activate T210A-CB1Rs. Thus, one can assume that in somatodendritic plasma membrane, the limiting factor for CB1R activation is not the receptor number but the 2-AG concentration. That could explain why it is possible to activate somatodendritic endogenous CB1Rs with exogenous cannabinoids (see Part 2). On the other hand, it is possible that 2-AG is not a genuine ligand for CB1Rs but a lipid present in plasma membrane that favours receptors activation. In that case, the constitutive activity observed with overexpressed WT-CB1Rs in absence of 2-AG and lack of constitutive

167 activity with T210A-CB1Rs, demonstrate that T210 is necessary and sufficient to provide constitutive CB1R activity.

5.3. Responses to WIN in neurons expressing T210A-CB1Rs are due to endogenous CB1R activation. Previously, we reported that T210A-CB1Rs display higher levels at the somatodendritic membrane as compared to WT-CB1Rs (Simon et al, 2013). Thus, we were expecting that receptor activation with WIN induces a stronger inhibition of PKA activity in somata compared to WT-CB1Rs or endogenous CB1Rs. However, the decrease is in the same range for all conditions. As mentioned above, previous studies demonstrated that T210A-CB1Rs display about 20-fold lower affinity for agonists compared to WT-CB1Rs (D'Antona et al,

2006). Indeed, DAntona and colleagues measured a Ki=1.44µM for WIN on T210A-CB1Rs. Moreover, cultured neurons express both endogenous CB1Rs and overexpressed T210A- CB1Rs. Thus, using WIN 100nM and supposing that endogenous CB1Rs display about the same affinity as WT-CB1Rs for WIN, one can assume that measured responses to WIN were due to endogenous CB1Rs activation. This hypothesis explains why WIN responses with T210A-CB1Rs fit perfectly those obtained with endogenous CB1Rs both in dendrites and axon. However, to date, it is not possible to provide proper results with neurons expressing only T210A-CB1Rs and none endogenous receptors. One possibility could be to use knock-in mice, where CB1R gene has been invalidated and replaced by the T210A-CB1R sequence.

168

Part 5: Role of type-1 cannabinoid receptors in neuronal development and the establishment of neuronal polarity.

169 1. Introduction Prenatal cannabis exposure, due to cannabis use by pregnant women, can lead to various serious disorders such as impulsive and hyperactive behaviours with increased inattention (Huizink & Mulder, 2006). Indeed, CB1R is present in embryonic brains, even before synapses formation. It is even expressed in both embryonic (Aguado et al, 2005; Rueda et al, 2002) and postnatal (Aguado et al, 2006; Jin et al, 2004) neuronal progenitors. In mice, it is highly expressed in glutamatergic neurons from embryonic day 12.5 and is downregulated at approximately 5 days after birth (Vitalis et al, 2008), leading to the adult CB1R distribution with predominant expression in GABAergic neurons. Moreover, in embryonic neurons, CB1Rs are mostly internalized in both somatodendritic compartment and axon (Vitalis et al, 2008). This sub-neuronal localization differs from its polarized distribution in mature neurons. CB1R role during brain development was widely studied and showed important implications in migration and differentiation of neuronal progenitor cells, neuritogenesis, neurite outgrowth, axonal tracts structure, and synapse establishment. Notably, CB1R activation was shown to have an overall inhibitory effect on the development of polarized neuronal morphology. In order to prepare the development of a novel research axis in the group that studies the molecular mechanisms and the pathophysiological relevance of these effects, we established a bibliographic review on this subject. The published literature data and current results of the team suggest that not only neuronal polarization influences both CB1R traffic and pharmacology but CB1Rs also contribute to the achievement of neuronal polarization.

170

2. Review (published): Type-1 cannabinoid receptor signaling in neuronal development. Pharmacology, 2012. Anne-Lise Gaffuri, Delphine Ladarre, and Zsolt Lenkei.

171 Contribution on Cannabinoids

Pharmacology 2012;90:19–39 Received: April 13, 2012 DOI: 10.1159/000339075 Accepted: April 13, 2012 P ublished online: July 3 , 2012

Type-1 Cannabinoid Receptor Signaling in Neuronal Development

Anne-Lise Gaffuri Delphine Ladarre Zsolt Lenkei

Neurobiology Laboratory, ESPCI-ParisTech, ESPCI-CNRS UMR 7637, Paris , France

Key Words of cytoskeletal effectors, which act in concert with positive- GPCR � Brain development � Neuritogenesis � Axonal tracts feedback local-excitation loops, to ultimately yield highly polarized neurons. Copyright © 2012 S. Karger AG, Basel

Abstract The type-1 cannabinoid receptor (CB1R) was initially identified as the neuronal target of �9-tetrahydrocannabinol Introduction (THC), the major psychoactive substance of marijuana. This receptor is one of the most abundant G-protein-coupled re- The main neuronal receptor for � 9 -tetrahydrocan- ceptors in the adult brain, the target of endocannabinoid li- nabinol (THC), the major psychoactive substance of mar- gands and a well-characterized retrograde synaptic regula- ijuana, is the type-1 cannabinoid receptor (CB1R), one tor. However, CB1Rs are also highly and often transiently ex- of the most abundant G-protein-coupled receptors pressed in neuronal populations in the embryonic and early (GPCRs) in the nervous system [1] . The endogenous li- postnatal brain, even before the formation of synapses. This gands of CB1R, i.e. endocannabinoids (eCBs), are cur- suggests important physiological roles for CB1Rs during rently recognized as retrograde messengers that are ca- neuronal development. Several recent reviews have summa- pable of modulating synaptic plasticity at different time rized our knowledge about the role of the endocannabinoid scales. First, short-term forms of eCB-mediated sup- (eCB) system in neurodevelopment and neurotransmission pression of synaptic transmission, called depolarization- by focusing on the metabolism of endocannabinoid mole- induced suppression of inhibition (DSI) or excitation cules. Here, we review current knowledge about the effects (DSE), are induced by a transient stimulation of CB1Rs of the modulation of CB1R signaling during the different which inhibits voltage-gated calcium channels leading to phases of brain development. More precisely, we focus on the inhibition of neurotransmitter release [2]. Second, the reports that directly implicate CB1Rs during progenitor cell autocrine activation of CB1R can lead to synaptic slow migration and differentiation, neurite outgrowth, axonal self-inhibition (SSI) that lasts around 20 min, where the pathfinding and synaptogenesis. Based on theoretical con- receptor modulates calcium-dependent potassium considerations and on the reviewed experimental data, we pro- ductance [3] . Finally, eCB signaling mediates long-term pose a new model to explain the diversity of experimental synaptic depression (LTD) in excitatory and inhibitory findings on eCB signaling on neurite growth and axonal pathfinding. In our model, cell-autonomus and paracrine eCBs acting on CB1Rs are part of a global inhibitory network A.-L. Gaffuri and D. Ladarre contributed equally to this work.

© 2012 S. Karger AG, Basel Zsolt Lenkei 0031–7012/12/0902–0019$38.00/0 Laboratoire de Neurobiologie-UMR 7637 Fax +41 61 306 12 34 ESPCI-ParisTech E-Mail [email protected] Accessible online at: 10 rue Vauquelin, FR–75005 Paris (France) www.karger.com www.karger.com/pha E-Mail zsolt.lenkei @ espci.fr afferents. Extended activation (several minutes) of pre- see Turu and Hunyady [21] ). In neurons, CB1Rs were gen- synaptic CB1Rs coupled to pre- and/or postsynaptic acti- erally reported to couple to similar signaling pathways vation leads to a decrease of presynaptic PKA (protein [22–24] as in nonpolarized cells, but we do not yet know kinase A) activity, via the activation of Gi/o proteins that how the specific neuronal structure, such as the unusu- inhibit adenylyl cyclase. This results in dephosphoryla- ally high surface-to-volume ratio of mature axons, mod- tion of target proteins and a long-lasting reduction of ifies CB1R signaling. Furthermore, direct information neurotransmitter release (recently reviewed in Heifets about the mobilization of CB1R effectors in developing and Castillo [4] ). neurons is still sparse. In addition to these now relatively well-characterized GPCRs are highly dynamic sensory molecules, dis- functions, in the last two decades it became widely recog- playing a flexible and dynamic three-dimensional struc- nized that eCB actions in the brain are not limited to the ture. Evidence from both functional and biophysical regulation of neurotransmission at established adult syn- studies suggests that GPCRs permanently sample multi- apses. Developmental studies have shown that, from ze- ple conformations but are held in a basal conformational brafish to mammals, CB1R is highly expressed in the de- equilibrium at steady state by intervening loops and non- veloping brain [5–11] . Remarkably, after birth, when syn- covalent intramolecular interactions [25] such as the aptic contacts of axons consolidate, CB1R expression is highly conserved ‘ionic lock’ between the 3rd and 6th greatly reduced in structures such as cortical projection transmembrane domains (reviewed in refs. [25–28] ). Ag- neurons [7, 8, 12] . This peculiar expression pattern opens onist binding modifies the energy landscape and leads to the possibility that neuronal eCBs may play important a prolonged adoption of an active receptor conformation, novel roles, different from their established roles in syn- resulting in the rearrangement of the cytoplasmic do- aptic regulation. Indeed, currently eCBs and CB1R are main of the receptor and in subsequent mobilization of known to be involved in brain development at the synap- intracellular signaling pathways mainly through a cog- tic [7, 12–15] , neuronal [7–12, 15–19] and network level [5, nate heterotrimeric G-protein. Agonist ligands shift the 10, 20] . equilibrium toward activated states, whereas inverse ago- The main purpose of this review is to summarize cur- nists shift the equilibrium toward inactive states. Signal- rent knowledge about the effects of CB1R signaling during by stabilized activated states is typically terminated ing different phases of brain development. More precise- on the timescale of seconds by phosphorylation of spe- ly, we will focus on studies reporting the direct effects of cific in tracellular serine and threonine residues, which CB1R during migration and differentiation of progenitor leads to decoupling from effectors (desensitization) and cells, neurite outgrowth, axonal pathfinding and synap- to the recruitment of scaffolding proteins such as togenesis. As the exact role of CB1Rs in several of these �-arrestins, ultimately resulting in GPCR endocytosis functions is still controversial, we have attempted to pre- through the classical clathrin-mediated endocytic path- cisely summarize experimental data and to consider way. After endosomal elimination of bound ligands and whether a consensus model could be proposed. dephosphorylation, GPCRs are either recycled back to the plasma membrane or degraded in lysosomes, depend- The Type-1 Cannabinoid Receptor CB1R ing on the receptor subtype, the cell type and the level of Cloned in 1990, the CB1R was established as the target activation. of � 9 -tetrahydrocanna binol, the major psychoactive sub- Interestingly, the CB1R, like numerous other GPCRs stance of marijuana [1] , and as the main neuronal recep- [29] , displays a high level of constitutive activity, i.e. con- tor of the eCB system. CB1R is a class A or rhodopsin-like, stitutive activation of intracellular signaling pathways in seven-transmembrane-domain GPCR. CB1Rs are pre- absence of exogenous ligands, either when heterologous- dominantly coupled to G proteins of the Gi/o family, con- ly expressed in non-neuronal cells [30] or in neurons sequently CB1R activation leads in most tissues to inhi- where CB1Rs are endogenous [31, 32] . The cannabinoid bition of adenylate cyclase, resulting in diminished receptor family shares the highly conserved DRY motif production of cAMP and protein kinase A (PKA) as well as well as several neighboring residues with the GPCR as in activation of G-protein-coupled inwardly rectifying consensus sequence [33] . However, Debra Ken dall’s group potassium channels (GIRKs), in the inhibition of several has observed that all cannabinoid receptors identified to types of voltage-gated calcium channels, and in the acti- date differ from the consensus GPCR sequence by a Thr vation of the ERK 1/2, p38 MAPK and JNK pathways (for substitution at position 3.46 as well as one helical turn a recent review on CB1R signaling in nonpolarized cells, amino-terminal to Arg 3.50, which is part of the basal-

20 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei state stabilizing, highly-conserved ionic lock. Substitu- nolamines (NAPEs), by a phospholipase D selective for tion of Thr 3.46 with Ala, a residue with a higher helical NAPEs (NAPE-PLD). AEA is inactivated by hydrolysis by packing moment, resulted in a receptor (T210A) with di- the fatty acid amine hydrolase (FAAH). The other major minished constitutive activity [34] . endocannabinoid, 2-AG, is synthesized principally by hy- Interestingly, while CB1R receptors are localized to the drolysis from diacylglycerols (DAGs) containing arachid- plasma membrane in axons of mature neurons [35–38] , a onate in position 2. This reaction is catalyzed by two iso- predominantly intracellular localization was reported in forms of DAG lipases (DAGL� and DAGL� ) selective for nonpolarized cells [34, 39–44] and in the somatodendrit- the sn-1 position. The degradation of 2-AG is mainly due ic domain of mature neurons [35–38, 45] as well as in the to monoacylglycerol lipases (MAGLs). Other members of somatodendritic and axonal domain of embryonic neu- the eicosanoid family are also synthesized in neurons and rons [8] . The exact origin and role of this intracellular are able to bind the CB1R: dihomo-� -linolenoyl ethanol- CB1R population is not settled yet [43, 46] . We have pre- amide and docosatetraenylethanolamide. Several other viously reported that an important proportion of intra- arachidonic-acid-derived molecules showing eCB activ- cellular CB1Rs are localized to endosomes, that a signifi- ity have been identified: the 2-arachidonoylglyceryl ether cant proportion of intracellular CB1Rs is of endocytic (or noladin ether), the O-arachidonoylethanolamine origin [38, 39] , and that the constitutive endocytosis of (virhodamine) and the N-arachidonoyldopamine CB1Rs is necessary for the correct axonal targeting of (NADA) (for review, see Bisogno et al. [62] ). CB1Rs [38, 45] , by a mechanism similar to that of the con- While synthesis and degradation of the major eCBs stitutively active 5-HT1B serotonin receptor [47] . Block- AEA and 2-AG are now relatively well understood, stor- ade of constitutive activation, either through the T210A age and release mechanisms are not well known (for a mutation [34] , inverse agonist treatment [38, 39] (but see recent review, see Alger and Kim [59] ). Notably, it appears McDonald et al. [45] ) or reduction of steady-state cell- that significant amounts of 2-AG are present in resting autonomous production of eCB 2-arachidonoylglycerol neurons, that only a fraction of stimulation-induced (2-AG) [48] reduced the endosomal CB1R population 2-AG is released from cells and that basal levels of AEA suggesting that constitutive endocytosis is a consequence are sufficient to constitutively induce CB1R signaling of constitutive CB1R activation at steady state. These re- [59] . We have also shown previously that basal levels of sults are in line with the view that an important intracel- 2-AG are sufficient to tonically activate CB1Rs in non- lular pool, reported also for several constitutively active stimulated neurons [48] . Given that eCB ligands are di- GPCRs [49–56] , is a dynamic phenomenon and is related rectly derived from membrane phospholipids, and that to constitutive activation of CB1Rs. However, as detailed no effective intramembrane sequestration of the impor- below, in addition to putative structural determinants, tant basal pool of eCBs has been identified yet, it is high- eCBs, which are ubiquitously expressed in neurons [57] , ly probable that eCBs may reach the membrane-bound may contribute to constitutive activation of CB1Rs. CB1Rs by two-dimensional diffusion, leading to cell-autonomous basal (i.e. constitutive) activation of CB1Rs. In- deed, the possibility of this direct entry of eCBs from the eCBs Are Ubiquitous Neuronal Membrane lipid bilayer to the CB1R-binding site was recently con- Components firmed by biochemical tools as well as molecular simula- tion (reviewed in Howlett et al. [57] ). As seen above, eCBs are lipophilic molecules, which are thought to be CB1Rs display structural characteristics [63] that may synthesized on demand from ubiquitous plasma mem- contribute, in addition to the ubiquitously present endo- brane components through multiple biosynthetic path- cannabinoid molecules, to the well-described constitu- ways, in response to elevations of intracellular calcium tive stimulation of downstream-signaling pathways in alone or combined with activation of several Gq/11 -pro- absence of exogenous cannabinoid ligands [57, 64] . The tein-coupled receptors (for recent detailed reviews, see idea that membrane bilayer composition may have sig- references [58, 59] ). The first of these eCBs to be identified nificant effect on the conformation of embedded GPCRs were N-arachidonoylethanolamine, also called anan- was first described by studying the effect of membrane damide (AEA) [60] and 2-arachidonoylglycerol (2-AG) cholesterol content on rhodopsin activation and has been [61] , both belonging to the eicosanoid family of poly-un- extended since to other GPCRs and lipids (recently re- saturated fatty acids. AEA is synthesized mainly by hy- viewed in Oates and Watts [65] ). In general, cell mem- drolysis of the corresponding N-acyl-phosphatidyletha- branes are increasingly considered as key dynamic com-

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 21 Development ponents in sensory and signaling pathways, where the tal stages (E21) and that its expression is downregulated highly regulated lipid environment significantly regu- after birth leading to a disappearance of significant CB1R lates the structure, conformation and function of embed- binding in these structures (corpus callosum, fornix, ded proteins [66] . Thus, a plausible scenario proposes that stria terminalis, stria medullaris and fasciculus retroflex- CB1Rs, due to structural determinants and the presence us) [70] . On the other hand, at fetal stages, CB1R is already of specific binding sites which are accessible for lipids expressed in structures such as hippocampus, cerebel- present in the lipid bilayer, are able to continuously trans- lum, caudate-putamen and cerebral cortex, and its ex- late the plasma membrane concentration of highly regu- pression increases to adult levels after birth. Moreover, lated lipidic eCBs into the activation of specific intracel- stimulation of [35 S]GTP� S binding by cannabinoid ago- lular signaling pathways, and be the subject of steady- nists suggests that embryonic CB1Rs are already func- state endocytosis. The source of eCBs acting on CB1Rs tional [71] . may be in many instances cell-autonomous. In the endo- Following the identification of the CB1R gene [1] , in cannabinoid literature, this cell-autonomous action is of- situ hybridization histochemistry was used to determine ten referred to as ‘autocrine’. However, as seen above, li- the localization of the corresponding mRNA transcript pophilic eCBs likely do not need to leave the neuronal in both fetal and adult rodent brains. CB1R mRNA was plasma membrane to gain access to the ligand binding identified in structures such as cerebral cortex, hippo- site of CB1Rs on the same neuron as proposed by the clas- campus, caudate-putamen and cerebellum at E16 [71] . In- sical autocrine model (from Greek kr īnō , ‘to separate’ or terestingly, CB1R binding and mRNA expression pat- ‘to secrete’) where hormones or chemical messengers are terns are mismatched in white matter areas of fetal brains, secreted to bind to receptors on the same cell. In addition as CB1R binding is elevated whereas mRNA expression is to the cell-autonomous mode of CB1R activation, a large low [72] . The limited spatial resolution of receptor auto- body of indirect experimental evidence, gathered using radiography and in situ hybridization initially precluded electrophysiological, anatomical and genetic tools sug- to determine whether the transiently elevated CB1R ex- gest that eCBs may be able to leave the parent cell mem- pression in white matter areas, which are composed brane and cross short distances in the aqueous extracel- mainly of axons of projection neurons, is due to axonally lular milieu, leading to paracrine activation of CB1Rs on transported CB1Rs or to local expression in astrocytes. neighboring neurons, leading to the well-described ret- This question was finally settled by using immunohisto- rograde modulation of synaptic transmission in the adult chemistry, which showed that most CB1Rs present in fe- brain [2] . tal white matter areas are located in axons of projection Taken together, the lack of clearly understood storage neurons in contrast with the adult brain, where CB1R is and release mechanisms for eCBs, as well as the struc- predominantly expressed in GABAergic interneurons. tural characteristics of CB1Rs, the high level of cellular Thus, several groups, including ours, showed that CB1R eCB content in membranes of non-stimulated neurons, is highly expressed in glutamatergic projection neurons the multiple reports on cell-autonomous CB1R activation from E12.5 and is downregulated in these neurons at ap- mechanisms and high level of CB1Rs outside of presyn- proximately 5 days after birth (P5) [7, 8, 12] . Moreover, aptic specializations [67], suggest that in addition to the immunochemistry allowed CB1R to be localized with classical modus operandi of presynaptic GPCR function high resolution by electron microscopy. In contrast to the [68] , novel functional paradigms should also be envi- adult brain, where CB1R is located on the axonal plasma sioned for CB1Rs, both in the adult and developing brain. membrane and in somatodendritic endosomes [36, 73] , in the fetal brain, CB1R is mostly localized to endosomes both in axons and in the somatodendritic region [8] . CB1Rs and eCBs Are Highly Expressed in the Developing Brain eCB Signaling Modulates Progenitor Cells Migration CB1R is one of the most abundant GPCRs expressed and Differentiation in the adult brain [69] . However, CB1Rs are also highly expressed from early fetal stages, starting from embry- One of the most remarkable natural phenomena is the onic day 12.5 (E12.5) in mice [8] . First, autoradiography emergence of the highly complex vertebrate brain, which using radiolabeled cannabinoid ligands showed that is accomplished by an extraordinary succession of self- CB1R is highly expressed in white matter areas at late fe- organized developmental events and finely tuned through

22 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei experience. Initial establishment of functional structure The effect of CB1Rs to regulate the ability of neuronal and connectivity in the developing cerebral cortex rely progenitors to differentiate and reach mature neuronal upon three major, self-organized early developmental phenotype has been evaluated in various studies. In 2002, events: (1) the proliferation and differentiation of neural Rueda et al. [75] showed that AEA inhibits differentiation progenitors, leading to the timely generation of appropri- of cortical neuron progenitor and NGF-induced PC12 ate neuronal subtypes, (2) the migration of neurons to cells via CB1R-dependent ERK activation. They also specific locations, and (3) the establishment of functional found that administration of methanandamide, an ana- synaptic connections between neurons after completing log of AEA, decreases the number of new mature neurons neuronal differentiation [74] . in the dentate gyrus of rats, whereas the antagonist Using rodent models, CB1Rs were detected in embry- SR141716 increases neurogenesis in the same zone. An- onic [75, 76] and postnatal neuronal progenitors [77–79] , other study has shown that CB1R activation promotes the suggesting a developmental role for the eCB system start- differentiation of neuronal progenitors into astroglial ing at the earliest stages of brain development. Indeed, cells, confirmed by the higher presence of these cells in studies in the last decade have established the implica- FAAH knock-out mice and the lower level of differenti- tions of CB1R in the survival, proliferation, migration ated cells in CB1R knock-out mice [79]. Chronic treat- and differentiation of neuronal progenitors ( table 1 ). ment with the cannabinoid agonist HU-210 did not lead Studies of the effects of cannabinoids on the survival to an increase in neuronal differentiation of newly-born of neuronal progenitors have been attempted with vari- progenitors [78] . Interestingly, a recent study indicates ous CB1R agonists and produced conflicting results: that, in vitro, a 4-day treatment of AEA promotes glial ACEA (arachidonyl-2-chloroethylamide, an AEA ana- differentiation, whereas a 7-day treatment promotes neu- log) and HU-210 ((–)-1,1-dimethylheptyl analog of 11-hy- ronal differentiation [82] . By studying adult spinal cord droxy-� 8-tetrahydrocannabinol) two agonists of CB1R, progenitors in primary cultures and in vitro slices, the enhanced progenitor survival [80, 81] , whereas URB297, knock-down of CB1Rs and treatment with the antagonist a FAAH inhibitor, AEA and THC seemed to have no ef- AM-251 resulted in enhanced neuronal differentiation fect [79, 81, 82] . On the other hand, WIN55,212-2, an oth- [83] . er agonist of CB1R, did not influence the survival of neu- Finally, pharmacological activation of CB1R increases ronal progenitors from cortices [79] but enhanced oligo- migration of GABAergic interneurons both in vitro and dendrocyte progenitor survival through activation of Gi/o in vivo through a mechanism involving the activation of proteins, enhanced PI3K (phosphoinositide 3-kinase) ac- TrkB receptor (BDNF/NT-3 growth factor receptor) [18] , tivity and Akt (protein kinase B) phosphorylation [80] . whereas inhibition of CB1Rs decreases neuroblast migra- In vitro, pharmacological CB1R activation enhances tion in rostral migratory stream explants [84] . progenitor proliferation and neurosphere generation via Overall, in spite of the rather moderate expression lev- activation of Gi/o proteins and ERK phosphorylation. els in neuronal progenitors as compared to embryonic This mechanism does not involve the PI3K/Akt pathway projection neurons ( fig. 1 ) [8] or adult GABAergic neu- [78] . On the contrary, application of the antagonist/in- rons, CB1Rs appear to regulate survival, proliferation, verse agonist SR141716 leads to a decrease of progenitor migration and differentiation of neuronal progenitors. proliferation [76, 78] . The presence of this proliferative The overall picture that seems to emerge from the above effect has not yet been unambiguously shown in vivo. studies is that CB1R activation promotes the prolifera- Studies on CB1R knock-out mice indicate indeed a de- tion, survival and migration of progenitor cells but delays crease of neuronal progenitor proliferation [77] and the transition from the multipotent, proliferating and FAAH knock-out mice show also an increase of progen- migration-competent progenitor phenotype towards a itor proliferation [76] . Chronic treatment with the can- settled, well-differentiated, postmitotic neuronal phe- nabinoid agonist HU-210 also led to an increase in pro- notype. However, the duration of the treatment and the genitor proliferation [78] . However, mice treated with presence of growth-promoting factors may modulate this THC or cannabidiol during 6 weeks had a lower number effect. Notably, long-term treatment was reported in sev- of proliferating cells in the dentate gyrus in comparison eral instances to yield the opposite effects of short-term with the controls [81] , which is reminiscent of previous treatment [81, 82] . Thus, further investigation is required findings of reduced progenitor cell proliferation in vitro to clarify the validity of the above results in both embry- following an elevated dose of AEA [82] . onic and post-natal physiological settings. Specific intracellular signaling mechanisms have also yet to be identi-

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 23 Development Table 1. E ffects of CB1R activation (a) or inactivation (b) on progenitor cell survival, proliferation, migration and differentiation a CB1R activation

Model CB1R activation Treatment Functional readout Ref. duration survival proliferation migration differentiation

Cultured oligodendrocyte WIN55,212-2 12 h + 80 progenitors (rat) 25 nmol/l HU-210 12 h + 500 nmol/l E17 cortical neuron progenitors AEA 24 h – 75 (rat) 5 �mol/l FAAH knock-out mice Genetic knock-down of FAAH + 76 Cultured neurospheres and neural WIN55,212-2 7 days + 76 progenitors (rat) 30 nmol/l URB597 7 days + 30 nmol/l AEA 7 days + 10 �mol/l 2-AG 7 days + 10 �mol/l Pregnant rats THC 1 injection/day + 18 0.15 mg/kg from E5 to P2 Cultured hippocampal neural HU-210 12 h + = 78 progenitors (rat) 10 nmol/l to 1 �mol/l AEA 12 h = 1 to 5 �mol/l Adult rats HU-210 chronic treatment + 78 100 �g/kg HU-210 acute treatment = 100 �g/kg Cultured neurospheres and neural WIN55,212-2 7 days + 78 progenitors (rat) 30 nmol/l URB597 7 days + 30 nmol/l AEA 7 days + 10 �mol/l 2-AG 7 days + 10 �mol/l FAAH knock-out mice genetic knock-down of FAAH + 79 into astroglial cells Cultured neuronal progenitors WIN55-212,2 16 h = + 79 (rat) 30 nmol/l into astroglial cells URB597 16 h = + 30 nmol/l into astroglial cells Cultured neurospheres (mouse) AEA 16 h + 110 1 �mol/l 2-AG 16 h + 1 �mol/l FAAH knock-out mice Genetic knock-down of FAAH + 7 Amygdala on coronal male WIN55,212-2 2 injections: = = 111 brain sections of P4 rats 1 mg/kg P0 and P1 female WIN55,212-2 2 injections: –– 1 mg/kg P0 and P1 into astroglial cells Coronal brain sections male WIN55,212-2 2 injections: = = 111 of amygdala of P14 rat 1 mg/kg P0 and P1 female WIN55,212-2 2 injections: –– 1 mg/kg P0 and P1 into astroglial cells male WIN55,212-2 2 injections: = 1 mg/kg P0 and P1 female WIN55,212-2 2 injections: = 1 mg/kg P0 and P1

24 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei Table 1 (continued)

Model CB1R activation Treatment Functional readout Ref. duration survival proliferation migration differentiation

Amygdala on coronal male WIN55,212-2 2 injections: = = 111 brain sections of P0 (8 h) 0.5, 1 and 10 mg/kg 0 and 4 h rats female WIN55,212-2 2 injections: – 0.5, 1 and 10 mg/kg 0 and 4 h male FAAH inhibitor 2 injections: = 20 mg/kg 0 and 4 h female FAAH inhibitor 2 injections: – 20 mg/kg 0 and 4 h male MGL inhibitor (NAM) 2 injections: = 1 mg/kg 0 and 4 h female MGL inhibitor (NAM) 2 injections: – 1 mg/kg 0 and 4 h Cultured neural progenitors AEA 1 or 2 days = – 82 (mouse) 10 �mol/l AEA 4 days + 5 or 10 �mol/l into astroglial cells AEA 7 days + 5 or 10 �mol/l Adult mice THC 6 weeks feeding = – 81 41.2% cannabidiol 6 weeks feeding + – 38.8% Cultured RMS neuroblasts ACEA during + 84 (mouse) 0.75 �mol/l the assay (0–24 h) b CB1R inactivation

Model CB1R inactivation Treatment Functional readout Ref. duration survival proliferation migration differentiation

CB1R knock-out mice genetic knock-down of CB1R – 77 Cultured neurospheres and SR141716 7 days – 76 neural progenitors (rat) 2 �mol/l CB1R knock-out mice genetic knock-down of CB1R – 79 into astroglial cells Hippocampal neurospheres genetic knock-down of CB1R 1 day – 110 (mouse) genetic knock-down of CB1R 7 days – rimonabant 30 days = CB1R knock-out mice genetic knock-down of CB1R – 7 Nestin GFP reporter mice AM-251 1 or 24 h + 81 0.25 g/kg 7 days CB1R knock-out mice genetic knock-down of CB1R 24 h + 81 7 days – Cultured RMS neuroblasts AM-251 during the assay – 84 (mouse) 1 �mol/l (0–24 h)

Th is table summarizes the effects of direct modulation of CB1R activity on progenitor cell survival, proliferation, migration and differentiation. The ‘+’ and ‘–’ signs represent, respectively, statistically significant increase or decrease of the measured parameter, whereas a ‘=’ sign means no modification as compared to control. The mentions ‘+ into astroglial cells’ and ‘– into astroglial cells’ in the column named ‘differentiation’ mean that the modulation of CB1R activity activates or inhibits, respectively, the differentiation of progenitor cells into astroglial cells.

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 25 Development Fig. 1. Overview of CB1R protein expression in the developing cortico-hippocampal formation of the mouse telencephalon III beta-tubulin (Tuj-1) + CB1R + nuclei at E13.5. Differentiating and migrating progenitors, located in the subventricular zone (svz) and in the ganglionic eminence

cp Color version available online (ge) do not express high levels of CB1Rs cp (red; colors refer to the online version only). However, CB1R is expressed by the majority of newly differentiated neurons svz of the early cortical plate (cp) as indicated by the co-expression of CB1R with the neuron-specific early-differentiation mar ker class III beta-tubulin (Tuj-1, ar- hc rows). As neurons mature, CB1R protein is svz gradually relocalized into growing axons of projection neurons (arrowheads). Thus, in the more mature lateral parts of the cerebral cortex (on the left-hand side), CB1R immunolabeling is mainly localized in the lv intermediate zone, where Tuj-1-positive corticofugal axons are located. At the same time, in the less mature medial parts of the cortical region (right-hand side), which ge will later give rise to the hippocampus (hc), punctuate CB1R immunolabeling is localized to somatic endosomes of newly differentiated neurons, which lack axonal processes at this stage. Sections are counter- stained for nuclei with DAPI (blue). Coronal section. lv = Lateral ventricle. Bar = 200 � m. For experimental details, refer to Vitalis et al. [8] .

fied. As not all proliferating cells in a neurogenic niche However, immortalized neuronal cell lines, which are de- become neurons [85] , particular care must be taken in rived from neuroblastoma cells, are capable of long-term order to distinguish effects of CB1Rs on progenitor pro- growth and generally represent a single cell type, provid- liferation from their effects on neurogenesis. ing a high degree of reproducibility. Therefore, they are widely used for studies that aim to better understand neuronal development and function. Consequently, the eCB Signaling as a Modulator of Neuritogenesis and bulk of our initial knowledge on cannabinoid effects on Neurite Growth neurite development was obtained by using immortalized neuronal cell lines, which in addition often express CB1R Effects on Neurite Outgrowth: CB1R receptors endogenously and may grow neurites in Neuroblastoma-Derived Cell Lines certain conditions ( table 2 ). However, these neurites will One of the first steps of neuronal differentiation is not polarize into axons and dendrites, thus do not express neuritogenesis, i.e. the development of extensions, called key aspects of neuronal differentiation [86] . neurites, which will differentiate into one axon and sev- The first study of cannabinoid effects on neuronal cell eral dendrites, ultimately leading to a highly polarized morphology used rat B103 neuroblastoma cells and re- mature neuron. Growing and manipulating well-polar- ported dose-dependent morphological changes following ized cultured primary neurons in a controlled environ- treatment with � 9 -THC at micromolar concentrations, ment is not an easily accessible experimental model. such as cell rounding and retraction of neurites [87] . A

26 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei Ref. 93 19 92 89 90 75 17 16 18 dendrite length number of dendrites axonal branching length number of axonal branching primary axon length median length mean length + + + average neurite length total neurite length number of neurites Functional readout all neurites% cells with neurites main neurite dendrites 16 h = 16 h = duration 16 h + 16 h + 18 h = 18 h + 16 h + 16 h + 15 min – 14 h – 48 h – 48 h – 18 h 18 h + 18 h 6 days – – = 48 h48 h – – 18 h ) on neurite) on outgrowth b mol/l mol/l mol/l mol/l mol/l mol/l mol/l mol/l mol/l mol/l mol/l � � � � � � � � � � � HU-210 100 nmol/l HU-210 1 HU-210 10 nmol/l CB1R activation Treatment HU-210 10 100 nmol/l HU-210 1 10 HU-210 1 methanandamine 1 AEA 5 HU-210 from 0.01 to 100 nmol/l methanandamine 5 1 2-AG 5 200 nmol/l 1 200 nmol/l 100 nmol/l HU-210 50 nmol/l Age before activation DIV0 ACEA DIV0 WIN55,212-2 DIV0 NA DIV0 WIN55,212-2 DIV0 AEA ) and inactivation ( a starvation neurite outgrowth 1 h serum starvation HU-210 24 h serum starvation HU-210 treatment ng/ml NGF during treatment starvation E ffects activation CB1R of ( E activation CB1R Table 2.Table Neuronal cell lines Neuro2A cells overnight serum a Model Activation of NG108-15 cells 1% serum during PC12 cells 2% serum and 100 N1E-115 cells 24–48 h serum Cerebellar neurons (rat) Primary cultures of neurons Cerebellar neurons (rat) CB1R+ interneurons (rat)

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 27 Development 7 19 11 15 84 90 Ref. Ref. 112 dendrite length dendrite length number of dendrites = = 8 = –– number number of dendrites axonal branching length axonal branching length number of axonal branching number of axonal branching – – 11 primary axon length primary axon length median length median length mean length – = = = + + = mean length average neurite length average neurite length total neurite length total neurite length number of neurites number of neurites Functional readout all neurites% cells with neurites main neurite dendrites all neurites% cells with neurites main neuriteendrites d Functional readout overnight 3 days = + = duration 7 days 24 h overnight – 15 h 24 h 16 h = + duration 14 h + 4 days 24 h – mol/l mol/l+ under � � 200 nmol/l 200 nmol/l CB1R activation Treatment 200 nmol/l 100 nmol/l 50 nmol/l 100 nmol/l 500 nmol/l 100 nmol/l 300 100 CB1R inactivation Treatment SR141716A 100 nmol/l (oricular injection) N-cadherin stimulation DIV3 WIN55,212-2 DIV0 AEA Age before activation DIV3 WIN55,212-2 DIV3DIV2 T210I-CB1R HU-210 24 h 5 h JZL184 DIV0 ACEA DIV0 ACEA 30 h JZL184 P5 ACEA DIV0 RHC-80267 Age before activation neurite outgrowth treatment neurite outgrowth (continued) inactivation CB1R Table Table 2 Hippocampal neurons (rat) Pyramidal cells (rat) Model Activation of Cortical neurons (rat) Cortical neurons (rat) Retinal explants (mouse) Explants from rostral migratory stream (mouse) In vivo Retinal projection (hamster) Primary cultures of neurons Retinal neurons (Xenopus) Neuronal cell lines NG108-15 cells 1% serum during b Model Activation of

28 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei 7 Ref. 16 11 dendrite length number of dendrites + + + + + 8 at,inthis context, activa- CB1R axonal branching length er, whereas a ‘=’ signmeans modificationwhereas no ‘=’ a er, number of axonal branching + + 11 primary axon length median length mean length = = + = – + 84 = average neurite length total neurite length number of neurites Functional readout all neurites% cells with neurites main neurite dendrites uriteoutgrowth. Experiments performed Neuro2Aon cells demonstrate th statisticallysignificant increase decreaseor the measured of paramet duration 18 h 18 h 4 days 18 h overnight + 3 days + = + 24 h 7 days overnight mol/l � mol/l mol/l mol/l mol/l � � � � CB1R activation Treatment 1 1 1 1 300 (ocular injection) 300 nmol/l 500 nmol/l 200 nmol/l 200 nmol/l Age before activation DIV0 AM-251 DIV0 AM-281 DIV0 AM-630 DIV0 AM-251 DIV0 O2050P5 AM-251 overnight + DIV0 AM-251 DIV3DIV0 T210A-CB1R ACEA 24 h DIV3 AM-251 DIV3 AM-251 neurite outgrowth (continued) Thi s tables summarizes Thithe effects directof modulation activity CB1R of ne on Cerebellar neurons (rat) Model Activation of as compared to control. tioninduce neuritic signs outgrowth.represent,and‘–’ respectively, The‘+’ Pyramidal cells (rat) In vivo Retinal projection (hamster) Retinal explant (mouse) Explants from rostral migratory stream (mouse) Hippocampal neurons (rat) Table Table 2

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 29 Development subsequent study also reported anandamide-induced cell CB1R Effects on Neurite Outgrowth: Primary rounding in CB1R-overexpressing B103 cells; this effect Neuronal Cultures was PTX-insensitive and Rho-kinase dependent [88] . As with neuroblastoma-derived cell lines, the effects Similar effects were later reported in other neuroblas- of CB1R activation or deactivation on neurite outgrowth toma cell lines such as N1E-115 [89] and NG108–15 [90] in primary neuronal cultures are also somewhat contro- and also in pheochromocytoma-derived, differentiating versial ( table 2). Indeed, several studies found that CB1R PC12 cells [75] , by using either micromolar concentra- activation promotes neurite outgrowth [16, 17, 19] , where- tions of the endocannabinoids AEA, methanandamine as others reported that similar treatments inhibit neurite (mAEA) and 2-AG or nanomolar concentrations of the development [8, 18] in cultured neurons. synthetic high-affinity agonist HU-210. Typically, treat- On the one hand, neuronal CB1R has been described ment durations were between 14 and 48 h, but the N1E- as an activator of a hierarchical signaling network that 115 cell study reported that cell-rounding and neurite re- promotes neurite growth. First, Patrick Doherty’s group, traction were already detectable after as little as 15 min using cultured cerebellar granule neurons, demonstrated of low-dose HU-210 treatment [89] . a positive effect of CB1R activation on neurite outgrowth In contrast to these studies, which reported neurite [16] . They showed that cannabinoid receptor antagonists retraction as a general effect following cannabinoid treat- inhibit the axonal growth response induced by N-cad- ment, the group of Ravi Iyengar reported induction of herin and FGF2 (fibroblast growth factor 2), whereas neurite outgrowth in neuroblastoma-derived Neuro2A CB1R agonist treatment mimics the N-cadherin/FGF2 cells [19, 91–93] . The underlying mechanisms reportedly responses. This modulation acts a step downstream of involved the activation of CREB (cAMP response ele- FGF receptor and upstream of calcium influx. The effect ment-binding), which is activated by MAPK (mitogen- of CB1R activity on neurite outgrowth was further con- activated protein kinase), PAX6 (paired box protein-6), firmed in another study from the same group [17] , which downstream of PI3K, and STAT3 (signal transducer and showed that inhibiting DAGL� and DAGL � by tetrahy- activator of transcription 3) which is regulated by Rap1 drolipstatin reduces the neurite outgrowth response in- (Ras-related protein 1), Src (proto-oncogene tyrosine- duced by FGF2. Finally, after reporting the coincidence protein kinase) and BRCA1 (breast cancer type 1 suscep- between eCBs and IL-6 signaling in Neuro2A cells (see tibility protein). Interestingly, IL-6 can also activate above), the team of Ravi Iyengar also demonstrated that STAT3 through Jak, and CB1R and IL-6 receptors are able co-incident stimulation of CB1R and the IL-6 receptor to act in synergy to induce neurite outgrowth in Neuro2A increases median length of neurites in rat primary corti- cells at sub-threshold concentration of each agonist (HU- cal neurons [19] . 210 and IL-6) [19] . These reports are intriguing for sev- On the other hand, when the group of Tibor Harkany eral reasons: (1) the main signaling pathway of CB1Rs is addressed whether eCBs affect the morphological speci- through G o/i proteins, which are usually negatively cou- fication of CB1R expressing interneurons, they found pled to CREB activation, and (2) reported effects were that CB1R activation inhibits neurite extension by modu- achieved after long-term incubation, i.e. 16–18 h, by using lating TrkB receptor-dependent signaling [18]. They also micromolar concentrations of the high-affinity ligand showed that this negative effect of CB1R activation on HU-210. Lower concentrations such as 10 and 100 nmol/l neurite outgrowth (1) requires Src and MAPK activity, were ineffective. These results are intriguing since the (2) is negatively regulated by PI3K (phosphatidylinositol dissociation constant (Kd ) of HU-210 is 45 pmol/l, so 3-kinase) and (3) is phospholipase C-� -kinase indepen- low-nanomolar concentrations are already saturating at dent. These results partially contradict those obtained by CB1Rs. This raises the possibility that the reported in- Williams et al. [16] , who could not find any evidence that duction of neurite growth, which is in variance with the CB1R is coupled to the TrkB receptor to modulate neurite above reports that used different neuroblastoma cell growth in cultured cerebellar granule neurons. lines, may also be due to indirect effects of long-term, Finally, our group’s observation of morphological high-dose HU-210 treatment. In this context, it is of inter- changes of CB1R-transfected neurons in vitro indicate est that the same group has recently demonstrated that that CB1R is a predominantly negative regulator of neu- neurite outgrowth induced by retinoic acid in Neuro2A rite growth [8] , since CB1R activation is inversely corre- cells is also achieved through an increased expression of lated with the total length of dendrites and axons. These both the DAGL� and DAGL � diacylglycerol lipases, results were obtained by two complementary techniques: which synthesize 2-AG [91] . pharmacological treatment of neurons expressing the

30 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei wild-type CB1R or by transfection of point-mutant CB1Rs targeting in the developing brain ( table 3 ). First, Watson that display different levels of spontaneous activation, et al. [6] have shown in chick and zebrafish embryos that originally published by Kendall’s group [34] . modifying CB1R activity disrupts axonal projections and Finally, two recent studies [7, 84] performed on cul- fasciculation. This effect was confirmed in the develop- tured mouse explants from rostral migratory stream and ing mammalian brain: genetic or pharmacological inac- on cultured pyramidal cells, respectively, yielded results tivation of embryonic CB1R leads to defasciculation and which allow consensus to be reached. On the one hand, mistargeting of axonal tracts in rodents [7, 9, 10] . In the Mulder et al. [7] showed that CB1R activation by AEA developing mouse brain, 2-AG appears to be the major positively regulates axonal length; on the other, CB1R in- eCB acting through CB1R to modulate reciprocal con- activation by AM-251 increases the number of axonal nections between the thalamus and the cortex [10] . branching points. These results were consistent with In order to precisely characterize how eCB signaling those obtained on cultured mouse explants from the ros- regulates axonal pathfinding, several studies have devel- tral migratory stream that demonstrated that inhibition oped in vitro and in vivo approaches based on the direct of CB1R activity via AM-251 decreases the length of the observation of growth cone morphology, motility and main neurites while increasing the number of branching directionality of growth. Indeed, in vitro observations points [84] . Thus, these data suggest that the eCB system clearly show that CB1R is present in growth cones and is helps in maintaining a polarized morphology in growing particularly accumulated in filopodia [11, 12, 15, 37]. In neurons by promoting the extension of a single major vitro growth cone turning assays have allowed further leading process. understanding of the effects of local cannabinoid exposure on axonal navigation. Thus, anisotropic applications of CB1R agonists induce growth cone repulsion in CB1R Activity Modulates Targeting and Structure of rat and mouse CB1R-expressing cultured neurons [11, Axonal Tracts 12] . Similarly, CB1R agonists increase the frequency of electric field-induced cathodal repulsion in Xenopus lae- During development, neurons extend axons in order vis spinal neurons [12] . Furthermore, in mouse retinal to innervate their targets. Developing axons navigate explants cultures, constitutive CB1R activation has been through highly complex environments for long distances. shown to diminish the surface expression of the deleted This process has to be precisely controlled in order to ob- in colorectal cancer receptor (DCC, a receptor for axonal tain correct network architecture. Located at the tip of guidance molecule netrin-1) in a PKA-dependent man- the growing axons, actin-rich growth cones are highly ner, thus opposing the growth-promoting effect of ne- motile structures that explore the extracellular environ- trin-1, ultimately leading to growth cone retraction [11] . ments, determine the direction of growth, and guide ax- Thus, an increasing body of evidence suggests that CB1R onal elongation. signaling controls growth cone navigation by opposing CB1R is strongly expressed in projection axons during attractive chemical and electrical directional cues, indi- neuronal development of embryonic chicken [5, 6] , ze- cating an important role of eCB signaling in axonal brafish [6] , mice [7, 8, 10] , rats [9] and in mouse retina [11] . pathfinding. Interestingly, this expression is dynamically regulated in Using in vitro chemotropic growth cone turning as- cortical projection neurons during brain development says, recent studies have started to characterize the mo- [6–8, 12] . Whereas projection neurons express only rela- lecular mechanisms underlying the effects of CB1R acti- tively low levels of CB1Rs in the adult hippocampus and vation on growth cone navigation. Selective inhibition of cerebral cortex, these neurons express high amounts of ROCK (Rho-associated protein kinase) switched CB1R functional CB1Rs during axonal elongation (between agonist induced neurite repulsion into chemoattraction E12 and birth). After birth, when synaptic contacts con- without affecting neurite extension, indicating that in solidate, CB1R expression is gradually reduced in corti- this process CB1R is coupled to the RhoA (Ras homolog cal projection neurons but remains highly expressed gene family, member A) signaling pathway [12] . More- throughout adulthood in GABAergic interneurons. over, the same group showed that agonist stimulation in- Thus, after the discovery that CB1R expression is dy- duces CB1R removal from the growth cone filopodia and namically regulated in projection axons, several recent Erk1/2 phosphorylation in the central growth cone do- studies were devoted to investigate whether eCB signal- main. However, Argaw et al. [11] did not succeed in rep- ing, through CB1R, exerts differential effects on axonal licating these results using primary cortical glutamat-

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 31 Development Table 3. E ffects of CB1R activation (a) and inactivation (b) on axonal growth cones a CB1R activation

Cellular model DIV before CB1R Treatment Functional readout Ref. treatment activation duration growth number of extension neurite growth cone filopodia at rate out- cone turn- surface area the growth cone growth ing angle

Bath application Spontaneous retinal ganglion DIV0 ACEA 24 h – – 11 modifications of the cells (mouse) 50 nmol/l growth cone cortical neurons DIV0 ACEA 48 h – – (mouse) 50 nmol/l Electric field-induced spinal neu-rons AEA 10 nmol/l 3 h – 12 modifications of the (rat) AEA 100 nmol/l 3 h – growth cone WIN55,212-2 3 h – 50 nmol/l WIN55,212-2 3 h – 100 nmol/l cortical neurons 30 h JZL184 16 h – 15 (rat) 100 nmol/l Microgradient Spontaneous spinal neu-rons DIV2–3 WIN55,212-2 1 h – – 12 modifications of the (rat) 200 nmol/l growth cone retinal ganglion DIV2 ACEA 1 h – – 11 cells (mouse) 50 nmol/l b CB1R inactivation

Cellular DIV before CB1R Treatment Functional readout Ref. model treatment inactivation duration growth number of extension neurite growth cone filopodia at rate out- cone turn- surface area the growth cone growth ing angle

Bath application Spontaneous retinal neurons RHC-80267 30 min – 112 modifications of the (Xenopus) 500 �mol/l (only during growth cone application) RHC-80267 9 h – 250 or 500 �mol/l retinal ganglion DIV0 AM-251 24 h + + 11 cells (mouse) 300 nmol/l DIV0 O2050 48 h + + 300 nmol/l cortical neurons DIV0 AM-251 48 h + + (mouse) 300 nmol/l DIV0 O2050 48 h + + 300 nmol/l

Thi s table summarizes the effects of direct modulation of CB1R activity on spontaneous or induced growth cone advance. The three articles that focus on the role of CB1R activity on growth cone navigation have shown that eCBs act as chemorepulsive cues for axonal growth cones. The ‘+’ and ‘–’ signs represent, respectively, statistically significant increase or decrease of the measured parameter as compared to control.

32 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei ergic neurons, so the implication of the Erk signaling data was obtained through indirect methods; molecular pathway remains unclear. mechanisms as well as the in vivo relevance of the inhib- Overall, in vivo and in vitro studies have now clearly itory CB1R effect on synaptogenesis remain to be eluci- demonstrated that eCB signaling plays a central role in dated. axonal pathfinding in the developing brain. Through the activation of CB1R, eCBs locally modulate growth cone morphology, motility and directionality in order to allow eCBs, Surmountable Negative Regulators of axons to reach their specific targets. However, the spe- Neurite Growth? cific downstream molecular mechanisms underlying this process are still poorly described, and further research The establishment of neuronal polarity depends on in- still needs to be performed to fully understand it. trinsic factors and on extrinsic cues from the environment [96, 97] , as proposed by Ramon Y Cajal more than a century ago, who observed that developing neurites re- CB1R Activity Modulates the Establishment of semble migrating cells, and thus neurite growth may be Neuronal Connectivity regulated by gradients of extracellular cues [98] . Recent models of gradient sensing and subsequent neuronal po- Once pre- and postsynaptic elements have come into larization propose a delicate balance between local stim- contact, both structures are modified in order to create a ulation, amplified by positive feed-back loops, and long- functional synapse. As CB1R is present both in growth range diffusible inhibitory factors [96] , whose integrated cones and dendrites of developing neurons, several studies output may drive an excitable downstream network of have investigated its role in the modulation of synaptogen- cytoskeletal effectors [99] . Recent reports from the group esis [7, 12, 13, 15] . cAMP/PKA activity, which has been of Mu-Ming Poo have started to formally identify key el- proven instrumental to the induction of new synapses be- ements of the local-excitation, global-inhibition mecha- tween hippocampal neurons in culture [94, 95] , is expect- nism implicated in neuronal polarization by showing the ed to be inhibited by the Gi/o -protein-coupled CB1R, sug- antagonistic effects of the effectors cAMP/cGMP [100] , gesting that CB1R activation will result in reduced synap- regulated by multiple factors that are yet to be fully char- togenesis. Indeed, cannabinoids were found to inhibit the acterized, such as BDNF, a self-amplifying autocrine pro- formation of new synapses in cultured hippocampal neu- moter of axon growth [101] . These reports depict a net- rons, induced by a forskolin-mediated increase in cAMP work of mutually inhibitory, competitive interactions [13] . Similarly, in cultured cortical neurons, inhibition of during outgrowth of neural processes, both during axon DAGL� induces an increase in SNAP25 expression, specification and later during axon branching [102] . Con- strongly suggesting that inhibition of CB1R signaling trig- sequently, an extrinsic factor may be a positive or a nega- gers synaptogenesis [7] . These results were indirectly cor- tive regulator of neurite growth, depending on the devel- roborated by in vivo findings, which showed that targeted opment stage, the identity of the neurite (axon or den- deletion of CB1Rs in GABAergic neurons results in a sig- drite), and the presence or absence of other extrinsic nificantly elevated density of perisomatic GABAergic ter- factors, as it was shown for Semaphorin3A, a secreted fac- minals in the neocortex and hippocampus [12] . Collec- tor which was reported to inhibit axonal growth and tively, these results depict CB1Rs as a negative regulator stimulate dendritic development [103] . Such dependence of synaptogenesis. This scenario was also indirectly con- from the cellular and tissue context should be kept in firmed by a recent study which showed that in growth mind by researchers who want to understand the action cones reaching their target, the 2-AG degrading enzyme of cannabinoids on neurite development. monoacylglycerol lipase (MGL), previously excluded from Based on these theoretical considerations and on the the advancing growth cone through proteosomal degra- experimental data summarized above, we propose a new dation, shows elevated concentrations [15] . As such down- model which may help to untangle the complexity of ex- regulation of cell-autonomous 2-AG levels would result in perimental findings on eCB signaling on axonal growth diminished constitutive activation of CB1Rs, MGL-medi- and pathfinding ( fig. 2 ). We propose that the primary ef- ated release of the inhibitory CB1R-tone could be impor- fect of CB1R activation is the mobilization of cytoskeletal tant for the initiation of synaptogenesis. effectors such as Rho-activated kinase (ROCK), which are In conclusion, CB1Rs are likely negative regulators of negatively coupled to cell spreading and neurite growth the formation of new synapses. However, the bulk of the [104, 105] . In consequence, the direct short-term effect of

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 33 Development Normal Growth

n Color version available online

Bath application of CB1R agonist

CB CB CB

Local application of CB1R agonist

Fig. 2. Proposed model for eCBs as sur- CB mountable negative regulators of neurite CB growth. We propose that the primary effect of CB1R activation (blue arrows; col- n ors refer to the online version only) is the mobilization of cytoskeletal effectors, which are negatively coupled to cell spreading and neurite growth. In most physiological contexts, cell-autonomous or paracrine activation of CB1Rs (CB) leads to a relatively moderate level of inhibition, Bath application of CB1R antagonist/inverse agonist which is surmountable by local growth- promoting factors (red arrows) at the growth cone, such as BDNF or netrin-1, re- IA sulting in more efficient polarized growth. IA However, when the growth cone reaches a region highly enriched with eCBs, the balance between eCB signaling and growth- IA promoting molecules would be modified, n and could result, depending on the new equilibrium, in growth cone arrest, repulsion or collapse. Overall inhibition of this negative regulatory effect, e.g. through an- IA IA tagonist/inverse agonist (IA) ligands, leads to enhanced neurite growth, increased axonal branching and dendrite development. Arrow size is proportional to signal strength. n = neuronal nucleus.

34 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei CB1R activation would be a principally negative effect on ject to significant basal activation, there is room for fur- neurite growth. However, basal cell-autonomous or para- ther axonal activation [8] . Taken together, in developing crine activation of CB1Rs would yield only relatively mod- axons, there are several indirect indications that cell-au- erate inhibition in most physiological contexts, which is tonomous and local production of eCBs leads to a basal surmountable by local growth-promoting effectors at the and moderate activation of the receptor, which would be growth cone, such as the self-amplifying autocrine pro- in line with our proposed model. moter BDNF [101] or netrin-1 [11] . In such case, the ‘chan- Overall, the above data argue in favor of controlled neling’ effect of overall CB1R-mediated inhibition would and mostly cell-autonomous basal activation of CB1Rs in help the neuron to focus its resources to a limited amount developing axonal tracts. Unfortunately, detailed onto- of growth sites, resulting in more efficient polarized genic profiling of DAGL � / � and MGL expression is cur- growth. However, when the growth cone reaches a region rently lacking. However, the colocalization of DAGLs highly enriched with eCBs, the balance between eCB sig- and CB1R diminishes in the postnatal cerebellum where naling and growth-promoting molecules would be modi- DAGLs have been shown to be enriched in dendritic trees fied, and could result, depending on the new equilibrium, of Purkinje cells and excluded from CB1R enriched axo- in growth cone arrest, repulsion or collapse. Notably, this nal tracts [17] . CB1R receptor localization also shows a system would be an efficient mechanism for coordinated developmentally regulated shift since in postnatal axons guidance of axonal fascicles in the brain, where produc- CB1Rs are mostly localized at the axonal plasma mem- tion of eCBs by nearby axons would be used as repulsion brane [35, 36, 73, 107] , whereas in the somatodendritic cues that allow axons to grow directly towards their target region they are mostly intracellular [36] because of con- without unnecessary branching. stitutive endocytosis [38] , partially due to somatoden- This model is based mainly on two principal cohorts dritic DAGL action [48] . This somatodendritic endocyto- of experimental findings which shed light on direct cell- sis is important for the correct transcytotic targeting of autonomous effects of CB1R activation. First, in the ma- CB1Rs to the axonal plasma membrane [38] , but the exact jority of neuroblastoma cell models, the most readily de- mechanism of somatodendritic CB1R activation is still tected effect of CB1R activation, detected even at low under debate [45] . These results collectively suggest that agonist concentrations and short timescales, is cell round- eCB-induced basal activation of axonal CB1Rs may be a ing and neurite retraction. Second, direct observation of developmentally regulated phenomenon. Direct mea- the effects of CB1R activation on growth cones consis- surements of CB1R activity levels at the sub-neuronal tently yielded cone arrest, repulsion or collapse at the scale, a technically challenging task, may shed further timescale of minutes ( table 2 ). light on this interesting question. An important corollary to this model is that the activation of CB1Rs would be subject to fine spatial regulation. In the developing axon, it would display a constant Conclusions and moderate level of activation, which would require the constant and local production of eCBs. This hypothesis There is growing evidence that eCB signaling, through is consistent with results from immunohistochemical the activation of the CB1R plays a central role in neuronal studies, performed on the developing mouse brain, which development. Studies from developmental biology to cell reveal that DAGLs and CB1R are expressed in the same biology, performed on several animal models, indicate axonal tracts [10, 11, 17] and in the same growth cones [7, that the fine regulation of this signaling pathway is im- 12] . That suggests that eCBs could be locally produced to portant for sculpting the temporal and spatial diversity of activate CB1R in developing axons. Moreover, using ul- neuronal networks during brain development. These de- trastructural analysis in E14–E16 mice, we have shown velopmental effects could, at least partially, explain the that while CB1Rs are present on the axonal plasma mem- neurobehavioral abnormalities and the alteration of cog- brane, an important pool of receptors is found in axonal nitive functions in marijuana-exposed offspring [108] . endosomes [8] . As the activation of the receptor has been However, none of the available CB1R null mutant mice clearly associated with its internalization [37–39, 106] , display a marked developmental CNS defect or clearly ap- this finding strongly suggests that axonal CB1R are con- parent behavioral phenotype. This robustness or resil- tinuously activated in the embryonic brain. Moreover, ience of CNS development, often described in rodent agonist treatment was shown to further increase the en- knock-out models, could be explained by redundancy at dosomal pool. Therefore, even if embryonic CB1R is sub- different levels: (1) at the cellular level, upstream signal-

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 35 Development ing pathways converging on the same effectors as CB1Rs microenvironmental requirements that drive the regula- may show compensatory regulation, and (2) at the net- tion of neuronal development by eCBs. For example, work level, the topology of CB1R-regulated neural net- identifying the precise interactions of this pathway with works may show elevated resilience towards the altered cytoskeletal actors will give important clues to better un- structural integrity of the network following deletion of derstand the mechanisms that drive neuronal growth, hubs or links [109] . migration or remodeling. This review also emphasized the current lack of knowledge concerning the precise downstream signaling pathways involved in the effect of eCB system on neuronal Acknowledgments development. Even if some key players have been identified, further identification of endocannabinoid ligands, We are grateful to Natalia Velez-Sanchez (MIT, Boston, Mass.) for correction of the English syntax. metabolic enzymes or second-messenger cascades will be necessary to better understand the cellular context and

References

1 Matsuda LA, Lolait SJ, Brownstein MJ, Young 9 Gomez M, Hernandez ML, Pazos MR, Tolon 16 Williams EJ, Walsh FS, Doherty P: The FGF AC, Bonner TI: Structure of a cannabinoid RM, Romero J, Fernandez-Ruiz J: Colocal- receptor uses the endocannabinoid signaling receptor and functional expression of the ization of CB1 receptors with L1 and Gap-43 system to couple to an axonal growth re-

cloned CDNA. Nature 1990; 346: 561–564. in forebrain white matter regions during fe- sponse. J Cell Biol 2003; 160: 481–486. 2 Freund TF, Hajos N: Excitement reduces tal rat brain development. Evidence for a role 17 Bisogno T, Howell F, Williams G, Minassi A, inhibition via endocannabinoids. Neuron of these receptors in axonal growth and Cascio MG, Ligresti A, Matias I, Schiano-

2003; 38: 362–365. guidance. Neuroscience 2008; 153: 687–699. Moriello A, Paul P, Williams EJ, Ganga- 3 Bacci A, Huguenard JR, Prince DA: Long- 10 Wu CS, Zhu J, Wager-Miller J, Wang S, dharan U, Hobbs C, Di Marzo V, Doherty P: lasting self-inhibition of neocortical inter- O’Leary D, Monory K, Lutz B, Mackie K, Lu Cloning of the first Sn1-Dag lipases points to neurons mediated by endocannabinoids. HC: Requirement of cannabinoid CB(1) re- the spatial and temporal regulation of endo-

Nature 2004; 431: 312–316. ceptors in cortical pyramidal neurons for ap- cannabinoid signaling in the brain. J Cell

4 Heifets BD, Castillo PE: Endocannabinoid propriate development of corticothalamic Biol 2003; 163: 463–468. signaling and long-term synaptic plasticity. and thalamocortical projections. Eur J Neu- 18 Berghuis P, Dobszay MB, Wang X, Spano S,

Ann Rev Physiol 2009; 71: 283–306. rosci 2010; 32: 693–706. Ledda F, Sousa KM, Schulte G, Ernfors P, 5 Begbie J, Doherty P, Graham A: Cannabi- 11 Argaw A, Duff G, Zabouri N, Cecyre B, Mackie K, Paratcha G, Hurd YL, Harkany T: noid receptor, CB1, expression follows neu- Chaine N, Cherif H, Tea N, Lutz B, Ptito M, Endocannabinoids regulate interneuron mi- ronal differentiation in the early chick em- Bouchard JF: Concerted action of CB1 can- gration and morphogenesis by transactivat-

bryo. J Anat 2004; 205: 213–218. nabinoid receptor and deleted in colorectal ing the TrkB receptor. Proc Natl Acad Sci

6 Watson S, Chambers D, Hobbs C, Doherty P, cancer in axon guidance. J Neurosci 2011; 31: USA 2005; 102: 19115–19120. Graham A: The endocannabinoid receptor, 1489–1499. 19 Zorina Y, Iyengar R, Bromberg KD: Canna- CB1, is required for normal axonal growth 12 Berghuis P, Rajnicek AM, Morozov YM, binoid 1 receptor and interleukin-6 receptor

and fasciculation. Mol Cell Neurosci 2008; Ross RA, Mulder J, Urban GM, Monory K, together induce integration of protein kinase

38: 89–97. Marsicano G, Matteoli M, Canty A, Irving and transcription factor signaling to trigger

7 Mulder J, Aguado T, Keimpema E, Barabas AJ, Katona I, Yanagawa Y, Rakic P, Lutz B, neurite outgrowth. J Biol Chem 2010; 285: K, Ballester Rosado CJ, Nguyen L, Monory Mackie K, Harkany T: Hardwiring the brain: 1358–1370. K, Marsicano G, Di Marzo V, Hurd YL, Guil- endocannabinoids shape neuronal connec- 20 Berrendero F, Sepe N, Ramos JA, Di Marzo

lemot F, Mackie K, Lutz B, Guzman M, Lu tivity. Science 2007; 316: 1212–1216. V, Fernandez-Ruiz JJ: Analysis of cannabi- HC, Galve-Roperh I, Harkany T: Endocan- 13 Kim D, Thayer SA: Cannabinoids inhibit the noid receptor binding and mRNA expres- nabinoid signaling controls pyramidal cell formation of new synapses between hippo- sion and endogenous cannabinoid contents

specification and long-range axon pattern- campal neurons in culture. J Neurosci 2001; in the developing rat brain during late gesta-

ing. Proc Natl Acad Sci USA 2008; 105: 8760– 21:RC146. tion and early postnatal period. Synapse

8765. 14 Bernard C, Milh M, Morozov YM, Ben-Ari 1999; 33: 181–191. 8 Vitalis T, Laine J, Simon A, Roland A, Leter- Y, Freund TF, Gozlan H: Altering cannabi- 21 Turu G, Hunyady L: Signal transduction of rier C, Lenkei Z: The type 1 cannabinoid re- noid signaling during development disrupts the CB1 cannabinoid receptor. J Mol Endo-

ceptor is highly expressed in embryonic cor- neuronal activity. Proc Natl Acad Sci USA crinol 2010; 44: 75–85.

tical projection neurons and negatively regu- 2005; 102: 9388–9393. 22 Glass M, Felder CC: Concurrent stimulation lates neurite growth in vitro. Eur J Neurosci 15 Keimpema E, Barabas K, Morozov YM, Tor- of cannabinoid CB1 and dopamine D2 recep-

2008; 28: 1705–1718. toriello G, Torii M, Cameron G, Yanagawa Y, tors augments cAMP accumulation in stria- Watanabe M, Mackie K, Harkany T: Differ- tal neurons: evidence for a Gs linkage to the

ential subcellular recruitment of monoacyl- CB1 receptor. J Neurosci 1997; 17: 5327–5333. glycerol lipase generates spatial specificity of 23 Childers SR, Deadwyler SA: Role of cyclic 2-arachidonoyl glycerol signaling during AMP in the actions of cannabinoid recep-

axonal pathfinding. J Neurosci 2010; 30: tors. Biochem Pharmacol 1996; 52: 819–827. 13992–14007.

36 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei 24 Pan X, Ikeda SR, Lewis DL: Rat brain can- 36 Bodor AL, Katona I, Nyiri G, Mackie K, Le- 47 Carrel D, Simon A, Emerit MB, Rivals I, Le- nabinoid receptor modulates N-type Ca2+ dent C, Hajos N, Freund TF: Endocannabi- terrier C, Biard M, Hamon M, Darmon M, channels in a neuronal expression system. noid signaling in rat somatosensory cortex: Lenkei Z: Axonal targeting of the 5-HT1b se-

Mol Pharmacol 1996; 49: 707–714. laminar differences and involvement of spe- rotonin receptor relies on structure-specific

25 Kobilka BK, Deupi X: Conformational com- cific interneuron types. J Neurosci 2005; 25: constitutive activation. Traffic 2011; 12: plexity of G-protein-coupled receptors. 6845–6856. 1501–1520.

Trends Pharmacol Sci 2007; 28: 397–406. 37 Coutts AA, Anavi-Goffer S, Ross RA, Mac- 48 Turu G, Simon A, Gyombolai P, Szidonya L, 26 Rosenbaum DM, Rasmussen SGF, Kobilka Ewan DJ, Mackie K, Pertwee RG, Irving AJ: Bagdy G, Lenkei Z, Hunyady L: The role of BK: The structure and function of G-pro- Agonist-induced internalization and traf- diacylglycerol lipase in constitutive and an-

tein-coupled receptors. Nature 2009; 459: ficking of cannabinoid CB(1) receptors in giotensin AT1 receptor-stimulated cannabi-

356–363. hippocampal neurons. J Neurosci 2001; 21: noid CB1 receptor activity. J Biol Chem 2007;

27 Ballesteros J, Palczewski K: G protein-cou- 2425–2433. 282: 7753–7757. pled receptor drug discovery. Implications 38 Leterrier C, Laine J, Darmon M, Boudin H, 49 Mohammad S, Baldini G, Granell S, Narduc- from the crystal structure of rhodopsin. Rossier J, Lenkei Z: Constitutive activation ci P, Martelli AM: Constitutive traffic of mel-

Curr Opin Drug Discov Devel 2001; 4: 561– drives compartment-selective endocytosis anocortin-4 receptor in neuro2a cells and 574. and axonal targeting of type 1 cannabinoid immortalized hypothalamic neurons. J Biol

28 Shapiro DA, Kristiansen K, Weiner DM, receptors. J Neurosci 2006; 26: 3141–3153. Chem 2007; 282: 4963–4974. Kroeze WK, Roth BL: Evidence for a model 39 Leterrier C, Bonnard D, Carrel D, Rossier J, 50 Marion S, Weiner DM, Caron MG: RNA ed- of agonist-induced activation of 5-hydroxy- Lenkei Z: Constitutive endocytic cycle of the iting induces variation in desensitization

tryptamine 2a serotonin receptors that in- CB1 cannabinoid receptor. J Biol Chem 2004; and trafficking of 5-hydroxytryptamine 2C

volves the disruption of a strong ionic inter- 279: 36013–36021. receptor isoforms. J Biol Chem 2004; 279: action between helices 3 and 6. J Biol Chem 40 Ellis J, Pediani JD, Canals M, Milasta S, Mil- 2945–2954.

2002; 277: 18244–18244. ligan G: Orexin-1 receptor-cannabinoid CB1 51 Chanrion B, Mannoury la Cour C, Gavarini 29 Seifert R, Wenzel-Seifert K: Constitutive ac- receptor heterodimerization results in both S, Seimandi M, Vincent L, Pujol JF, Bockaert tivity of G-protein-coupled receptors: cause ligand-dependent and -independent coordi- J, Marin P, Millan MJ: Inverse agonist and of disease and common property of wild- nated alterations of receptor localization and neutral antagonist actions of antidepres-

type receptors. Naunyn Schmiedebergs Arch function. J Biol Chem 2006; 281: 38812–38824. sants at recombinant and native 5-hydroxy-

Pharmacol 2002; 366: 381–416. 41 Tappe-Theodor A, Agarwal N, Katona I, tryptamine2C receptors. Differential mod- 30 Bouaboula M, Perrachon S, Milligan L, Rubino T, Martini L, Swiercz J, Mackie K, ulation of cell surface expression and signal

Canat X, Rinaldi-Carmona M, Portier M, Monyer H, Parolaro D, Whistler J, Kuner T, transduction. Mol Pharmacol 2008; 73: 748– Barth F, Calandra B, Pecceu F, Lupker J, Maf- Kuner R: A molecular basis of analgesic tol- 757.

frand JP, Le Fur G, Casellas P: A selective in- erance to cannabinoids. J Neurosci 2007; 27: 52 Whistler JL, Enquist J, Marley A, Fong J, verse agonist for central cannabinoid recep- 4165–4177. Gladher F, Tsuruda P, Murray SR, Von Zas- tor inhibits mitogen-activated protein ki- 42 Wu DF, Yang LQ, Goschke A, Stumm R, trow M: Modulation of postendocytic sort- nase activation stimulated by insulin or Brandenburg LO, Liang YJ, Hollt V, Koch T: ing of g protein-coupled receptors. Science

insulin-like growth factor 1. Evidence for a Role of receptor internalization in the ago- 2002; 297: 615–620. new model of receptor/ligand interactions. nist-induced desensitization of cannabinoid 53 Morris DP, Price RR, Smith MP, Lei B,

J Biol Chem 1997; 272: 22330–22339. type 1 receptors. J Neurochem 2008; 104: Schwinn DA: Cellular trafficking of human 31 Pan X, Ikeda SR, Lewis DL: Sr 141716a acts as 1132–1143. alpha1a-adrenergic receptors is continuous an inverse agonist to increase neuronal volt- 43 Rozenfeld R, Devi LA: Regulation of CB1 and primarily agonist-independent. Mol 2+ age-dependent Ca currents by reversal of cannabinoid receptor trafficking by the Pharmacol 2004; 66: 843–854.

tonic cb1 cannabinoid receptor activity. Mol adaptor protein AP-3. FASEB J 2008; 22: 54 Morisset S, Rouleau A, Ligneau X, Gbahou F,

Pharmacol 1998; 54: 1064–1072. 2311–2322. Tardivel-Lacombe J, Stark H, Schunack W, 32 Hillard CJ, Muthian S, Kearn CS: Effects of 44 Grimsey NL, Graham ES, Dragunow M, Ganellin CR, Schwartz JC, Arrang JM: High CB(1) cannabinoid receptor activation on Glass M: Cannabinoid receptor 1 trafficking constitutive activity of native H3 receptors cerebellar granule cell nitric oxide synthase and the role of the intracellular pool: impli- regulates histamine neurons in brain. Na-

activity. FEBS Lett 1999; 459: 277–281. cations for therapeutics. Biochem Pharma- ture 2000; 408: 860–864.

33 Ballesteros J, Kitanovic S, Guarnieri F, Da- col 2010; 80: 1050–1062. 55 Holliday ND, Holst B, Rodionova EA, vies P, Fromme BJ, Konvicka K, Chi L, Millar 45 McDonald NA, Henstridge CM, Connolly Schwartz TW, Cox HM: Importance of con- RP, Davidson JS, Weinstein H, Sealfon SC: CN, Irving AJ: An essential role for constitu- stitutive activity and arrestin-independent Functional microdomains in G-protein- tive endocytosis, but not activity, in the axo- mechanisms for intracellular trafficking of

coupled receptors – the conserved arginine- nal targeting of the CB1 cannabinoid recep- the ghrelin receptor. Mol Endocrinol 2007;

cage motif in the gonadotropin-releasing tor. Mol Pharmacol 2007; 71: 976–984. 21: 3100–3112.

hormone receptor. J Biol Chem 1998; 273: 46 Benard G, Massa F, Puente N, Lourenco J, 56 Jacquier V, Prummer M, Segura JM, Pick H, 10445–10453. Bellocchio L, Soria-Gomez E, Matias I, Dela- Vogel H: Visualizing odorant receptor traf- 34 D’antona AM, Ahn KH, Kendall DA: Muta- marre A, Metna-Laurent M, Cannich A, He- ficking in living cells down to the single-

tions of CB1 t210 produce active and inactive bert-Chatelain E, Mulle C, Ortega-Gutierrez molecule level. Proc Natl Acad Sci USA 2006;

receptor forms. Correlations with ligand af- S, Martin-Fontecha M, Klugmann M, Gug- 103: 14325–14330. finity, receptor stability, and cellular local- genhuber S, Lutz B, Gertsch J, Chaouloff F, 57 Howlett AC, Reggio PH, Childers SR, Hamp-

ization. Biochemistry 2006; 45: 5606–5617. Lopez-Rodriguez ML, Grandes P, Rossignol son RE, Ulloa NM, Deutsch DG: Endocan- 35 Katona I, Rancz EA, Acsady L, Ledent C, R, Marsicano G: Mitochondrial CB(1) recep- nabinoid tone versus constitutive activity of

Mackie K, Hajos N, Freund TF: Distribution tors regulate neuronal energy metabolism. cannabinoid receptors. Br J Pharmacol 2011;

of CB1 cannabinoid receptors in the amyg- Nat Neurosci 2012; 15: 558–564. 163: 1329–1343. dala and their role in the control of GABAer-

gic transmission. J Neurosci 2001; 21: 9506– 9518.

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 37 Development 58 Pertwee RG, Howlett AC, Abood ME, Alex- 71 Berrendero F, Garcia-Gil L, Hernandez ML, 83 Sideris A, Bekker T, Chan WS, Montoya- ander SPH, Di Marzo V, Elphick MR, Greas- Romero J, Cebeira M, de Miguel R, Ramos Gacharna JV, Blanck TJ, Recio-Pinto E: A ley PJ, Hansen HS, Kunos G, Mackie K, JA, Fernandez-Ruiz JJ: Localization of role for the cannabinoid 1 receptor in neuro- Mechoulam R, Ross RA: International union mRNA expression and activation of signal nal differentiation of adult spinal cord proof basic and clinical pharmacology. LXXIX. transduction mechanisms for cannabinoid genitors in vitro is revealed through pharma- Cannabinoid receptors and their ligands: receptor in rat brain during fetal develop- cological inhibition and genetic deletion.

beyond CB(1) and CB(2). Pharmacol Rev ment. Development 1998; 125: 3179–3188. Front Neurosci 2012; 6: 4.

2010; 62: 588–631. 72 Wang X, Dow-Edwards D, Keller E, Hurd 84 Oudin MJ, Gajendra S, Williams G, Hobbs 59 Alger BE, Kim J: Supply and demand for en- YL: Preferential limbic expression of the C, Lalli G, Doherty P: Endocannabinoids

docannabinoids. Trends Neurosci 2011; 34: cannabinoid receptor mrna in the human fe- regulate the migration of subventricular

304–315. tal brain. Neuroscience 2003; 118: 681–694. zone-derived neuroblasts in the postnatal

60 Devane WA, Hanus L, Breuer A, Pertwee RG, 73 Morozov YM, Freund TF: Post-natal devel- brain. J Neurosci 2011; 31: 4000–4011. Stevenson LA, Griffin G, Gibson D, Mandel- opment of type 1 cannabinoid receptor im- 85 Kriegstein A, Alvarez-Buylla A: The glial na- baum A, Etinger A, Mechoulam R: Isolation munoreactivity in the rat hippocampus. Eur ture of embryonic and adult neural stem

and structure of a brain constituent that J Neurosci 2003; 18: 1213–1222. cells. Ann Rev Neurosci 2009; 32: 149–184. binds to the cannabinoid receptor. Science 74 Marin O, Rubenstein JLR: Cell migration in 86 Banker G, Goslin K: Culturing Nerve Cells.

1992; 258: 1946–1949. the forebrain. Ann Rev Neurosci 2003; 26: Cambridge, MIT Press, 1991. 61 Sugiura T, Kondo S, Sukagawa A, Nakane S, 441–483. 87 Cabral GA, Mcnerney PJ, Mishkin EM: In- Shinoda A, Itoh K, Yamashita A, Waku K: 75 Rueda D, Navarro B, Martinez-Serrano A, teraction of delta-9-tetrahydrocannabinol 2-Arachidonoylgylcerol – a possible endog- Guzman M, Galve-Roperh I: The endocan- with rat B103 neuroblastoma-cells. Arch

enous cannabinoid receptor-ligand in brain. nabinoid anandamide inhibits neuronal Toxicol 1987; 60: 438–449.

Biochem Biophys Res Commun 1995; 215: progenitor cell differentiation through at- 88 Ishii I, Chun J: Anandamide-induced neuro- 89–97. tenuation of the Rap1/B-Raf/ERK pathway. J blastoma cell rounding via the CB1 cannabi-

62 Bisogno T, Ligresti A, Di Marzo V: The en- Biol Chem 2002; 277: 46645–46650. noid receptors. Neuroreport 2002; 13: 593– docannabinoid signalling system. Biochemi- 76 Aguado T, Monory K, Palazuelos J, Stella N, 596.

cal aspects. Pharmacol Biochem Behav 2005; Cravatt B, Lutz B, Marsicano G, Kokaia Z, 89 Zhou D, Song ZH: CB1 cannabinoid recep-

81: 224–238. Guzman M, Galve-Roperh I: The endocan- tor-mediated neurite remodeling in mouse 63 D’antona AM, Ahn KH, Wang L, Mierke DF, nabinoid system drives neural progenitor neuroblastoma N1e-115 cells. J Neurosci Res

Lucas-Lenard J, Kendall DA: A cannabinoid proliferation. FASEB J 2005; 19: 1704–1706. 2001; 65: 346–353. receptor 1 mutation proximal to the dry mo- 77 Jin K, Xie L, Kim SH, Parmentier-Batteur S, 90 Chemin J, Nargeot J, Lory P: Neuronal T- tif results in constitutive activity and reveals Sun Y, Mao XO, Childs J, Greenberg DA: De- type alpha 1H calcium channels induce neu- intramolecular interactions involved in re- fective adult neurogenesis in CB1 cannabi- ritogenesis and expression of high-voltage-

ceptor activation. Brain Res 2006; 1108: 1–11. noid receptor knockout mice. Mol Pharma- activated calcium channels in the NG108–15

64 Pertwee RG: Inverse agonism and neutral col 2004; 66: 204–208. cell line. Journal of Neuroscience 2002; 22: antagonism at cannabinoid CB1 receptors. 78 Jiang W, Zhang Y, Xiao L, Van Cleemput J, Ji 6856–6862.

Life Sci 2005; 76: 1307–1324. SP, Bai G, Zhang X: Cannabinoids promote 91 Jung KM, Astarita G, Thongkham D, Pio- 65 Oates J, Watts A: Uncovering the intimate embryonic and adult hippocampus neuro- melli D: Diacylglycerol lipase-alpha and relationship between lipids, cholesterol and genesis and produce anxiolytic- and antide- -beta control neurite outgrowth in neuro-2a

GPCR activation. Curr Opin Struct Biol pressant-like effects. J Clin Invest 2005; 115: cells through distinct molecular mecha-

2011; 21: 802–807. 3104–3116. nisms. Mol Pharmacol 2011; 80: 60–67. 66 Phillips R, Ursell T, Wiggins P, Sens P: 79 Aguado T, Palazuelos J, Monory K, Stella N, 92 He JC, Gomes I, Nguyen T, Jayaram G, Emerging roles for lipids in shaping mem- Cravatt B, Lutz B, Marsicano G, Kokaia Z, Ram PT, Devi LA, Iyengar R: The G alpha(o/

brane-protein function. Nature 2009; 459: Guzman M, Galve-Roperh I: The endocan- i)-coupled cannabinoid receptor-mediated 379–385. nabinoid system promotes astroglial differ- neurite outgrowth involves Rap regulation

67 Mátyás F, Yanovsky Y, Mackie K, Kelsch W, entiation by acting on neural progenitor of Src and Stat3. J Biol Chem 2005; 280:

Misgeld U, Freund TF: Subcellular localiza- cells. J Neurosci 2006; 26: 1551–1561. 33426–33434. tion of type 1 cannabinoid receptors in the 80 Molina-Holgado E, Vela JM, Arevalo-Mar- 93 Jordan JD, He JC, Eungdamrong NJ, Gomes rat basal ganglia. Neuroscience 2006;137(1): tin A, Almazan G, Molina-Holgado F, I, Ali W, Nguyen T, Bivona TG, Philips MR, 337-361. Borrell J, Guaza C: Cannabinoids promote Devi LA, Iyengar R: Cannabinoid receptor- 68 Starke K: Presynaptic receptors. Annu Rev oligodendrocyte progenitor survival: in- induced neurite outgrowth is mediated

Pharmacol Toxicol 1981; 21: 7–30. volvement of cannabinoid receptors and by Rap1 activation through G(alpha)o/i-trig- 69 Herkenham M, Lynn AB, Johnson MR, Mel- phosphatidylinositol-3 kinase/Akt signal- gered proteasomal degradation of Rap1Ga-

vin LS, de Costa BR, Rice KC: Characteriza- ing. J Neurosci 2002; 22: 9742–9753. pII. J Biol Chem 2005; 280: 11413–11421. tion and localization of cannabinoid recep- 81 Wolf SA, Bick-Sander A, Fabel K, Leal-Gali- 94 Kavalali ET, Klingauf J, Tsien RW: Activity- tors in rat brain. A quantitative in vitro au- cia P, Tauber S, Ramirez-Rodriguez G, Mul ler dependent regulation of synaptic clustering

toradiographic study. J Neurosci 1991; 11: A, Melnik A, Waltinger TP, Ullrich O, Kem- in a hippocampal culture system. Proc Natl

563–583. permann G: Cannabinoid receptor CB1 me- Acad Sci USA 1999; 96: 12893–12900. 70 Romero J, Garcia-Palomero E, Berrendero F, diates baseline and activity-induced survival 95 Ma L, Zablow L, Kandel ER, Siegelbaum SA: Garcia-Gil L, Hernandez ML, Ramos JA, of new neurons in adult hippocampal neuro- Cyclic AMP induces functional presynaptic

Fernandez-Ruiz JJ: Atypical location of can- genesis. Cell Commun Signal 2010; 8: 12. boutons in hippocampal CA3-CA1 neuronal

nabinoid receptors in white matter areas 82 Soltys J, Yushak M, Mao-Draayer Y: Regula- cultures. Nat Neurosci 1999; 2: 24–30.

during rat brain development. Synapse 1997; tion of neural progenitor cell fate by anan-

26: 317–323. damide. Biochem Biophys Res Commun

2010; 400: 21–26.

38 Pharmacology 2012;90:19–39 Gaffuri/Ladarre/Lenkei 96 Arimura N, Kaibuchi K: Neuronal polarity: 102 Hutchins BI: Competitive outgrowth of 108 Huizink AC, Mulder EJH: Maternal smok- from extracellular signals to intracellular neural processes arising from long-dis- ing, drinking or cannabis use during preg-

mechanisms. Nat Rev Neurosci 2007; 8: tance cAMP signaling. Sci Signal 2010; nancy and neurobehavioral and cognitive 194–205. 3:jc1. functioning in human offspring. Neurosci

97 Barnes AP, Solecki D, Polleux F: New sights 103 Shelly M, Cancedda L, Lim BK, Popescu Biobehav Rev 2006; 30: 24–41. into the molecular mechanisms specifying AT, Cheng PL, Gao HF, Poo MM: Semapho- 109 Bullmore E, Sporns O: Complex brain net- neuronal polarity in vivo. Curr Opin Neu- rin3a regulates neuronal polarization by works. Graph theoretical analysis of struc-

robiol 2008; 18: 44–52. suppressing axon formation and promoting tural and functional systems. Nat Rev Neu-

98 Cajal Y, Berg OG, Jain MK: Direct vesicle- dendrite growth. Neuron 2011; 71: 433–446. rosci 2009; 10: 186. vesicle exchange of phospholipids mediated 104 Dickson BJ: Rho GTPases in growth cone 110 Aguado T, Romero E, Monory K, Palazue-

by polymyxin B. Biochem Biophys Res guidance. Curr Opin Neurobiol 2001; 11: los J, Sendtner M, Marsicano G, Lutz B,

Commun 1995; 210: 746–752. 103–110. Guzman M, Galve-Roperh I: The CB1 can- 99 Xiong YA, Huang CH, Iglesias PA, 105 Luo L: Rho GTPases in neuronal morpho- nabinoid receptor mediates excitotoxicity-

Devreotes PN: Cells navigate with a local- genesis. Nat Rev Neurosci 2000; 1: 173–180. induced neural progenitor proliferation

excitation, global-inhibition-biased excit- 106 Hsieh C, Brown S, Derleth C, Mackie K: In- and neurogenesis. J Biol Chem 2007; 282:

able network. Proc Natl Acad Sci USA 2010; ternalization and recycling of the CB1 can- 23892–23898.

107: 17079–17086. nabinoid receptor. J Neurochem 1999; 73: 111 Krebs-Kraft DL, Hill MN, Hillard CJ, Mc- 100 Shelly M, Lim BK, Cancedda L, Heilshorn 493–501. Carthy MM: Sex difference in cell prolif- SC, Gao HF, Poo MM: Local and long- 107 Nyiri G, Cserep C, Szabadits E, Mackie K, eration in developing rat amygdala medi- range reciprocal regulation of cAMP and Freund TF: CB-1 cannabinoid receptors are ated by endocannabinoids has implications cGMP in axon/dendrite formation. Science enriched in the perisynaptic annulus and for social behavior. Proc Natl Acad Sci USA

2010; 327: 547–552. on preterminal segments of hippocampal 2010; 107: 20535–20540.

101 Cheng PL, Song AH, Wong YH, Wang S, GABAergic axons. Neuroscience 2005; 136: 112 Lom B, Hopker V, McFarlane S, Bixby JL, Zhang X, Poo MM: Self-amplifying auto- 811–822. Holt CE: Fibroblast growth factor receptor crine actions of BDNF in axon develop- signaling in Xenopus retinal axon exten-

ment. Proc Natl Acad Sci USA 2011; 108: sion. J Neurobiol 1998; 37: 633–641. 18430–18435.

CB1R Signaling in Neuronal Pharmacology 2012;90:19–39 39 Development

General discussion and Perspectives

195 Differential pharmacology and maintenance of neuronal polarity We demonstrated that both the traffic and the pharmacology of CB1Rs are different between the somatodendritic compartment and the axon. In the somatodendritic compartment, CB1Rs are constitutively activated by the endocannabinoid 2-AG, leading to their internalization and to constitutive inhibition of the cAMP/PKA pathway. In axons, where DAGL is lacking, CB1Rs are accumulated at the plasma membrane and do not display constitutive activation (see Part 1 and Part 2). Moreover, we also demonstrated that the T210 residue is required for somatodendritic activation of CB1Rs by endogenous 2-AG (see Part 4). This residue is well conserved among the CB1R family but is absent in other GPCRs (D'Antona et al, 2006). Thus, specific structural determinants enable CB1Rs to be activated by 2-AG. Interestingly, 2- AG is present in invertebrates, such as nematodes, that lack CB1Rs (Lehtonen et al, 2008). Given that, CB1R appeared in the evolution after 2-AG. Why did evolution create a neuronal receptor sensitive to 2-AG? A possible explanation is in order to finely modulate synaptic transmission. Indeed, CB1R is widely known for its role in DSI, DSE and LTD. However, it does not explain the predominantly extrasynaptic localization of CB1Rs. Three interesting facts remain: (i) CB1Rs are already expressed in neuronal progenitors and play a major role in neuronal development (see Part 5). (ii) The group of Mu-Ming Poo showed that, during neuronal polarity establishment, axon maintains a high cAMP concentration and send negative feedbacks that decrease cAMP concentration in other neurites. These low-cAMP concentration neurites will develop as dendrites (Shelly et al, 2010). Thus, it seems that neuronal polarity requires high cAMP concentration in axon and low cAMP concentration in dendrites. (iii) Measure of PKA activity showed that activation of endogenous SST2Rs leads to a decrease of about 5% of FRET ratio in somata (see Annexe 2). Thus, although SST2Rs are highly present on the somatodendritic plasma membrane, the decrease of PKA activity observed after their activation is not stronger than the one observed after CB1R activation (see Part 2). This observation suggests that basal cAMP concentration in somata is low compared to cAMP concentration in axons (where CB1R activation leads to a decreased FRET ratio of about 15%). Taken together, these data suggest that CB1Rs could be involved in the maintenance of neuronal polarity. Indeed, as DAGLα is segregated in the somatodendritic compartment of mature neurons, a receptor activable by 2-AG and coupled to Gi/o proteins could enable to maintain low basal cAMP concentration in dendrites. Moreover, the lack of 2-AG in axons

196 indirectly enables the maintenance of high basal cAMP concentrations in this compartment. Thus, it is possible that the observed differential pharmacology of CB1Rs is directly linked to the maintenance of neuronal polarity. In conclusion, there are reciprocal effects between CB1R and neuronal polarity.

DAGLα is an interesting target to modulate neuronal CB1R activity The results we obtained show that somatodendritic DAGLα plays a major role in CB1R pharmacology. Indeed, its blockade removes the constitutive somatodendritic CB1R activation by 2-AG but also prevents its activation by both WIN and Δ9-THC (see Part 2). These data suggest that WIN and Δ9-THC could act as allosteric modulators that enhance response to 2-AG in the somatodendritic compartment. It could be an analogous mechanism as the positive allosteric effects between membrane lipidic ligand S1P and extracellular Nogo-

A-Δ20 on S1PR2 (Kempf et al, 2014). Moreover, in axons, 2-AG is not required for CB1R activation by WIN and Δ9-THC. Thus, it is possible that CB1Rs do not display the same basal conformation in axon and dendrites. Indeed, GPCRs are known to adopt a multitude of conformations and membrane components can strongly influence their activity. Thus, it is possible that axonal CB1R conformation enables the activation of receptors after agonist binding on all binding sites, while somatodendritic CB1R conformation is locked for some binding sites in absence of 2-AG, preventing its activation. It could be interesting to test the validity of these experiments in vivo. For example, comparing Δ9-THC effects on rodents in presence or absence of a DAGL inhibitor could enable to determine the physiological effects mediated specifically by somatodendritic CB1Rs. It is already known that CB1Rs on GABAergic neurons do not display the same physiological functions as CB1Rs on glutamatergic neurons (Monory et al, 2006) and could interestingly display different affinity for agonists, such as HU-210 (Steindel et al, 2013). These data suggest that it could be possible to design specific drugs for either GABAergic or glutamatergic CB1Rs and perform a finer modulation of effects. Applied to sub-neuronal CB1Rs, acting on DAGL activity could also enable the modulation of effects due to somatodendritic CB1R activation only. Interestingly, the group of Patrick Doherty has developed DAGLα-/- mice (Gao et al, 2010). Studying the responses of these mice to cannabinoids could also show if somatodendritic CB1Rs have specific in vivo effects. Using also these mice to perform neuronal cultures of hippocampi and live FRET imaging, we could also measure if constitutive activation of CB1Rs remains in the somatodendritic compartment that would be due only to structural and ligand-independent constitutive activity. Moreover, inhibition of DAGL (Turu et al, 2007) but

197 also blockade of CB1Rs (Simon et al, 2013) increases the number of receptors at the somatodendritic plasma membrane of cultured neurons. Thus, it could be interesting to study CB1R traffic in neurons of DAGLα-/- mice, using quantitative methods shown in Part 1. Notably, the group of Patrick Doherty showed that DSI was removed in DAGLα-/- mice and interpreted this result as the lack of postsynaptic DAGLα to synthesize retrograde 2-AG. However, one can also assume that lack of DAGLα decreases strongly the quantity of presynaptic CB1Rs leading to the lack of DSI.

Possible further investigations using live FRET imaging on cultured neurons The technique we developed enabled the sub-neuronal pharmacology study of endogenous and overexpressed CB1Rs, overexpressed serotonin 5-HT1B receptors and endogenous SST2Rs. The results obtained with overexpressed WT-CB1Rs warn for using overexpressed receptors to study their neuronal pharmacology (see Part 4). Thus, the studied system (type of cultured neurons) should be adapted in order to measure endogenous GPCR pharmacology. Moreover, development of new FRET biosensors could enable the investigation of other signaling pathways, such as cGMP pathway for example (see Annexe 2). High adolescent cannabis use is known to be associated with the development of serious psychiatric diseases development, such as schizophrenia. Thus, establishment of molecular mechanisms involved in cannabis dependence would be of great interest in order to treat patients. Twin studies have established that cannabis dependence is highly heritable (Agrawal et al, 2007). Polymorphism of CNR1 gene that codes for CB1R is linked to the development of cannabis dependence (Hopfer et al, 2006). Indeed, single-nucleotide polymorphism rs806380 is associated with the development of one or more cannabis dependence symptoms. Moreover, other genes coding for proteins of the endocannabinoid system but also for other neurotransmitter systems have been identified as candidates for polymorphisms that enhance cannabis dependence (Agrawal & Lynskey, 2009). Interestingly, specific phenotypes of CNR1 (rs2023239) and FAAH (rs324420) genes are correlated with a greater activation of reward-related areas of the brain after exposure to marijuana (Filbey et al, 2010). Which molecular mechanisms could explain these differences? One can assume that there is a difference in CB1R pharmacology between patients suffering from cannabis dependence and control patients. Performing live FRET imaging on cultured neurons of these patients could answer this question. How can we obtain cultured neurons of patients? Induced pluripotent stem cells (iPSCs) are pluripotent stem cells obtained from de-differentiation of fibroblasts from adult patients (Takahashi & Yamanaka, 2006). Then, it is possible to differentiate iPSCs

198 into several cell types, including neurons (Pasca et al, 2014). Thus, using this technique, it could be possible to obtain cultured neurons with the same genetic background as patients suffering from cannabis dependence and control patients. Finally, using live FRET imaging, it will be possible to study CB1R pharmacology in these neurons. My first hypothesis is that polymorphisms of CB1R could lead to signaling bias downstream of the receptor, increasing activation of some pathways and decreasing others. Notably, a stronger activation of Gi/o proteins after Δ9-THC-induced CB1R activation would increase induced-disinhibition of dopaminergic neurons and is coherent with the increased activation of reward-related areas of the brain of cannabis-dependent patients after exposure to marijuana. My second hypothesis is that CB1R pharmacology modification could be due to a change of basal tone of endocannabinoids, such as 2-AG. For example, as suggested by me results (see Part 2) an increased level of 2-AG would increase the activability of somatodendritic CB1Rs by THC and could explain the increased response after marijuana exposure. This hypothesis is supported by the identification of FAAH gene polymorphisms correlated with cannabis dependence.

199

Annexe 1: Article 3 (under review): Fast retrograde signaling in Drosophila axons. Anne-Lise Gaffuri, Lu Li, Thomas Riemensperger, Alexandre Roland, Nicolas Gervasi,

Delphine Ladarre, Pierre-Yves Plaçais, Anne Simon, Hervé Willaime, Patrick Tabeling, Paul

Tchénio, Serge Birman, Thomas Preat, and Zsolt Lenkei

200 The Journal of Neuroscience

http://jneurosci.msubmit.net

JN-RM-3914-13

Fast retrograde cAMP signaling in Drosophila axons

Zsolt Lenkei, ESPCI-ParisTech Anne-Lise Gaffuri, ESPCI-ParisTech Lu Li, Institut du Fer à Moulin Thomas Riemensperger, Georg-August-Universität Göttingen Alexandre Roland, ESPCI-ParisTech Nicolas Gervasi, Institut du Fer à Moulin Delphine Ladarre, ESPCI-ParisTech Pierre-Yves Plaçais, ESPCI- ParisTech Anne Simon, ESPCI-CNRS UMR 7637 Hervé Willaime, ESPCI- ParisTech Patrick Tabeling, ESPCI-ParisTech Paul Tchénio, ESPCI-ParisTech Serge Birman, ESPCI ParisTech, CNRS Thomas Preat, ESPCI, CNRS

Commercial Interest: No

This is a confidential document and must not be discussed with others, forwarded in any form, or posted on websites without the express written consent of The Journal for Neuroscience. ��

�

��

�

��

�

��

�

��

�

��

� ��

��

� ��

��

�

��

�

� ��

��

�

��

�

��

� ��

��

�

��

�

��

� ��

��

�

��

�

��

�

��

�

��

� ��

��

�

��

�

��

� ��

��

�

��

�

��

� ��

��

��µ��

�

��

�

� ��

��

�

��

�

� ��

��

�

��

� ��

��

�

��

�

��

� ��

��

�� γ�� α�β�� α′�β′� ��

�� α� ��

��

��α��

�� α� ��

��

�

��

� ��

��

� ��

��

�

� ��

��

� ��

��

�

� ��

��

� ��

��

�

��

� ��

��

�

��

�

��

� ��

��

α��

��α��

��

�

��

� ��

��

�

� ��

��

�

� �

� �� Figure 1

A B DIV1 DIV5 b mCD8-SYB-GFP/+;MB247/+ Nucleus Kenyon cell Kenyon Ac-tubulin

C mCD8-SYB-GFP/+;DvGlut/+ DmCD8-SYB-GFP/+;TH/+ ETrH/mCD8-SYB-GFP;+/+ F *** 5 * ** 4 3 2

Number of Number 1

primary neurites primary 0 N=79 N=57 N=68 N=92

KCs Glut. Dopaminergic neuron Dopaminergic Glutamatergic neuron Glutamatergic Serotoninergic neuron Serotoninergic Dopa. Sero. G GFP Serotonin Nucleus Merge TrH/mCD8-SYB-GFP;+/+

H GFP TH Nucleus Merge mCD8-SYB-GFP/+;TH/+

I Elav;mCD8-SYB-GFP/+;+/+ Nucleus mCD8-RFP/+;Repo/+ mCD8-RFP/+;NP3233/+ NP2276/mCD8-RFP;+/+ NP2222/mCD8-RFP;+/+ mCD8-RFP/+;NP6520/+ Figure 2

A +/+;TrH/mCD8-SYB-GFP Serotoninergic neuron Serotoninergic Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 B 100 D Kenyon Cells: mCD8-SYB-GFP/+;MB247/+ Nucleus Axons Somata and dendrites Cultured neurons 80 Stage 1 60 Stage 2 Stage 3 40 Stage 4

% of neurons of % 20 Stage 5 0

4DIV 7DIV 0.5DIV1.5DIV 0.25DIV C 100 E Serotoninergic neurons: TrH/mCD8-SYB-GFP;+/+ Nucleus 80 Axons Somata and dendrites Cultured neuron Stage 1 60 Stage 2 Stage 3 40 Stage 4 % of neurons of % 20 Stage 5 0

4DIV 7DIV Neuron Serotoninergic vivo In 0.5DIV1.5DIV 0.25DIV F 103Y/+;MB247-DsRed/mCD8-SYB-GFP; G Brain Culture Stage 3 larvae Adult P6 DIV0 DIV4

50 GFP 40

10 DsRed 0

% double-labeled Kenyon Cells Kenyon double-labeled % 0 2 4 6 8 1012 ; MB247-DsRed/+ Day in vitro Overlay Nucleus Nucleus Figure 3

A +/+;MB247/Apc2-GFP F Elav;mCD8-SYB-GFP/+;+/+ Apc2-GFP Fas2 Overlay GFP NC82 Nucleus Nucleus Kenyon cell Kenyon

B Fas2; TrH/+;+/Apc2-GFP Apc2-GFP Fas2 Overlay Nucleus Serotninergic neuron Serotninergic G 1 2 C Fas2;+/+;TH/Apc2-GFP Apc2-GFP Fas2 Overlay r Nucleus Microchannel

1 : Somatodendritic 2 : Axonal compartment compartement H Dopaminergic neuron Dopaminergic

D Fas2;+/+;DvGlut/Apc2-GFP AKAR2;238Y Apc2-GFP Fas2 Overlay AKAR2;238Y

Dextran 10 kDa

Glutamatergic neuron Glutamatergic I E DenMark-mCherry/+;MB247/+ Elav; mCD8-GFP/+;DenMark-mCherry/+;MB247/+ DenMark Fas2 Overlay DenMark-mCherry mCD8-GFP Kenyon cell Kenyon Figure 4

A B C RAKAR2=0.85 t=-1min t=8min 16 17.5 14 15.0 FSK 15 μM - bath application 12 12.5 10 10.0 8 Cell body Cell 6 7.5 Axons 4 5.0 Dendrites 2 2.5 Cell bodies

0 (%) ratio FRET AKAR2 0.0 -2 -10 -9 -8 -7 -6 -5 -4 -10 5 0 5 10 15 20 25 30 log (FSK concentration)

Mean AKAR2 FRET ratio change (%) change ratio FRET AKAR2 Mean time (min) D Dendrite *** 14 *** 12 *** 10 ** 8 ** ** 6 4 ** 2 N=44 Axon ratio change (%) change ratio N=53 N=41 N=35 N=45 N=42

Mean AKAR2 FRET AKAR2 Mean 0 _ _ _ RP-cAMPs + + +

Cell bodies Dendrites Axons RAKAR2=0.65 Figure 5

A B FSK 15 μM - axonal stimulation 150 μm *** 17.5 14 ** 15.0 12 Axons 12.5 10 10.0 8 7.5 Dendrites 6 5.0 4 2.5 Cell bodies

ratio change (%) change ratio 2

0.0 FRET AKAR2 Mean 0 N=30 N=8 N=30 -10 AKAR2 FRET ratio (%) ratio FRET AKAR2 -2.5 -5 0 5 10 15 20 25 30

-5.0 time (min) Axons Dendrites Cell bodies

C D 8 *** *** 7 FSK 15 μM 4 6 5 3 Bath 4 *** ** 2 3 150 μm 2 1 ns 200 μm 1 N=40 N=38N=22 N=21 N=15 N=16 0 0 Relative delay to axon (min) axon to delay Relative

AKAR2 FRET ratio (%) ratio FRET AKAR2 -10 -5 0 5 10 15 20 25 30 -1 time (min) Axons Axons Axons

Cell Bodies Cell Bodies Cell Bodies

Bath application Axonal stimulation Axonal stimulation 150 μm 200 μm E F

α lobes 20 FSK 15 M Cell bodies μ calyx 16 Calyx 12 α 8 150 μm 4 Axons 0

γ (%) ratio FRET AKAR2 -8 -6 -4 -2 0 2 4 6 8 10 β time (min) G I α lobes 10 calyx Dopamine 100 μM * 8 3.0 α lobes 6 2.5 calyx 4 2 2.0 ns 0 1.5 -8 -6 -4 -2 0 2 4 6 8 10 -2 ns AKAR2 FRET ratio (%) ratio FRET AKAR2 time (min) 1.0 H N=5 10 α lobes 0.5 Octopamine 100 M 8 μ calyx N=6 N=6 N=3 N=5 N=6 0.0 6 (min) axon to delay Relative 4 -0.5 _ _ _ _ 2 FSK + + _ _ _ _ 0 Dopamine + + _ _ _ _ + + -2 -8 -6 -4 -2 0 2 4 6 8 10 Octopamine AKAR2 FRET ratio (%) ratio FRET AKAR2 time (min) Figure 6

A 150 μm Proximal axon Distal axon Stimulation region 10 μm 20 μm 20 μm 12 μm 25 μm

Microchannel

7 μm 200 μm

B C 2 2 DPKAc= 1.8 μm /s DPKAc= 100 μm /s M) M)

17.5 μ 17.5 μ 15.0 15.0 Distal axon (Theoretical) Distal axon (Theoretical) 12.5 12.5 Distal axon (Measured) 10.0 10.0 Distal axon (Measured) 7.5 7.5 Cell bodies, 5.0 5.0 150 mm (Theoretical) 2.5 Cell bodies, 2.5 Cell bodies, 0.0 150 mm (Theoretical) 0.0 200 mm (Theoretical) Phosphorylated AKAR2 ( AKAR2 Phosphorylated Phosphorylated AKAR2 ( AKAR2 Phosphorylated 0 5 10 15 20 25 30 0 5 10 15 20 25 30 time (min) time (min) D E F *** *** 9 *** * 2.5 *** 12 FSK 15 M - axonal stimulation μ 8 10 7 2.0 6 8 1.5 5 6 4 1.0 4 3 2 0.5 2 1

AKAR2 FRET ratio (%) ratio FRET AKAR2 N=78 N=55 N=63 N=40 N=36 N=50 0 0 0.0 Mean AKAR2 FRET ratio change (%) change ratio FRET AKAR2 Mean

-2 -10 -5 0 5 10 15 20 25 30 (min) axon distal to delay Relative -0.5 time (min) Distal Distal Proximal axons Proximal axons Cell bodies axons Cell bodies axons

G H I FSK 15 μM - axonal stimulation 9 *** * 12 8 3 10 7 Distal 6 8 2 axons 5 6 4 4 3 1 Proximal 2 2 axons 1 N=36 N=30 N=20 N=25 Epac1 FRET ratio (%) ratio FRET Epac1 0 0 0 -2 -10 -5 0 5 10 15 20 25 30 Distal time (min) (%) change ratio FRET Epac1 Mean axons Proximalaxons (min) axon distal to delay Relative -1 Distal axons Proximalaxons J K

12 M) 12 μ M)

μ 10 10 Distal axons (measured) 8 8 Distal axons (theoretical) 6 6 Proximal axons (measured) 4 4 2 2 Proximal axons (theoretical) 0 0 Cell bodies (measured)

Epac1 bound cAMP ( cAMP bound Epac1 -2 -2 Cell bodies (theoretical) 0 5 10 15 20 25 30 ( AKAR2 Phosphorylated 0 5 10 15 20 25 30 time (min) time (min)

Annexe 2: Somatostatin SST2 receptors modulate both cGMP pathway and PKA activity in hippocampal neurons.

232 1. Introduction Nitric oxide (NO) mediates several physiological processes. It is synthesized by Nitric Oxyde Synthase (NOS) that oxides L-arginine to liberate NO and citrulline (Bredt & Snyder, 1992). Three isoforms of NOS have been identified: the neuronal NOS (nNOS), the endothelial NOS (eNOS) and the inducible NOS (iNOS). NO mediates its effects by activating soluble guanylate cyclase (GC) (Koesling et al, 2004). GC converts guanosine triphosphate (GTP) to cyclic Guanosine Monophosphate (cGMP), a second messenger able to activate several cellular targets, such as cGMP-dependent protein kinases and cGMP-regulated ion channels. cAMP and cGMP are known to have opposite biological effects, notably on neuronal development. Indeed, neurites enriched in cAMP tend more to differentiate into axons while those enriched in cGMP tend to grow as dendrites. Using FRET imaging in cultured hippocampal neurons, the team of Mu-ming Poo demonstrated a reciprocal regulation of cAMP and cGMP in undifferentiated neurites of developing neurons (Shelly et al, 2010). Notably, an increase of cAMP concentration induced by forskolin decreases cGMP concentration while an increase of cGMP concentration induced by a NO donor induces a decrease of both cAMP concentration and PKA activity. This mechanism involves activation of phosphodiesterases and could enable the establishment of neuronal polarity. We asked if reciprocal regulation of cAMP and cGMP could also take place downstream of GPCRs. Thus, we investigated the recruitment of signaling pathways downstream of endogenous neuronal somatostatin SST2 receptors (SST2Rs). SST2Rs are predominantly localized at the somatodendritic plasma membrane of mature hippocampal neurons and display no constitutive activity (Csaba et al, 2007; Lelouvier et al, 2008). When activated by endogenous somatostatin, a polypeptide hormone, SST2Rs inhibit exocytosis through inhibition of adenylyl cylclase and cAMP concentration (Theodoropoulou & Stalla, 2013). Interestingly, some studies reported also that activation of SST2Rs increases the cGMP concentration in isolated rat retinas (Mastrodimou et al, 2006) and activates cGMP-dependent kinase in chick ciliary ganglion (Meriney et al, 1994). Using live FRET imaging in cultured hippocampal neurons of rats, I started to investigate the modulation of cGMP concentration downstream of endogenous SST2Rs. Here, I will present preliminary results I obtained on that project.

233 2. Materials and Methods Animals Animal procedures were conducted in strict compliance with approved institutional protocols and in accordance with the provisions for animal care and use described in the European Communities council directive of November 24, 1986 (86 � 609 � EEC). SpragueDawley rats (Janvier) were used for dissociated cell culture experiments.

Chemicals and DNA constructs Dimethyl Sulfoxide (DMSO), Forskoline (Fsk), DEANONOate (NO), Bovine Serum Albumin (BSA) and Poly-D-Lysine were obtained from SIGMA-ALDRICH. B27, Lipofectamine 2000 and Neurobasal were obtained from Life Technologies. 1H- [1,2,4]Oxadiazolo[4,3-a]quinoxalin-1-one (ODQ) was obtained from R&D Systems. Digestion enzymes (NheI, SmaI and EcoRV) were obtained from New England Biolabs. Octreotide was nicely lended by Dr Pascal Dournaud (INSERM U676, Paris, France). AKAR4-Kras probes were obtained from Dr Jin Zhangs laboratory (Baltimore, USA). TEpacVV probe was obtained Dr Kees Jalink laboratory (Amsterdam, Netherlands).

FRET imaging Neurons transfected with either AKAR4-KRAS or with newly developed THPDE5VV probe were imaged by videomicroscopy between DIV8 and DIV11 on a motorized Nikon Eclipse

234 Ti-E/B inverted microscope with the Perfect Focus System (PFS) in a 37°C thermostated chamber, using an oil immersion CFI Plan APO VC 60X, NA 1.4 objective (Nikon). Acquisitions were carried out at the excitation wavelength of the CFP (434nm +/- 15nm) using an Intensilight (Nikon). Emitted light passed through an Optosplit II beam-splitter (Cairn Research) equipped with a FF509-FDi01 dichroïc mirror, a FF01-483/32-25 CFP filter and a FF01-542/27-25 YFP filter and was collected by an EM-CCD camera (Evolve 512, Photometrics), mounted behind a 2x magnification lens. Acquisitions were performed by piloting the set-up with Metamorph 7.7 (Molecular Devices). All filter sets were purchased from Semrock. Cultured neurons on 18-mm coverslips were placed in a closed imaging chamber containing an imaging medium: 120 mM NaCl, 3 mM KCl, 10 mM HEPES, 2 mM CaCl2, 2 mM

MgCl2, 10 mM D-glucose, 2% B27, 0,001% BSA. The acquisition lasted 120 minutes registering one image each 2 minutes, registering in parallel 10 to 15 neurons on the same coverslip. 30 minutes after the beginning of the acquisition, vehicle solution was applied then 60 minutes after the beginning of the acquisition, Octreotide 100nM or 1µM was applied. Finally, 90 minutes after the beginning of the acquisition, Forskolin (Fsk) 10µM or DEANONOate (NO) 50µM was applied.

235 The quantitative results obtained for each neuronal compartment were grouped together and, for each time point, the mean FRET ratio normalized to baseline and SEM were calculated. Deviation was corrected for somata and dendrites on Matlab. Mean slope was calcultated for all neurons in somata and dendrites, respectively, for the last 7 time points before addition of treatment and substracted from all FRET ratio time points.

Statistical analysis FRET Response was obtained by calculating the mean FRET ratio on Matlab for 6 time points after injection, from +6 minutes to +16 minutes (Response). Groups were compared using GraphPad Prism. Significance of differences between various conditions was calculated using unpaired t-tests or one-way ANOVA with Newman-Keuls post-tests for computing p estimates. NS p>0.05, * p<0.05, ** p<0.01 and *** p<0.001.

3. Results 3.1. Activation of endogenous SST2Rs decreases PKA activity in somata but not in the axons. To begin with, we measured the modulation of PKA activity downstream of endogenous SST2Rs in mature cultured hippocampal neurons of rats. We expressed AKAR4-Kras FRET probe in neurons and performed live FRET imaging. We measured FRET ratio in somata and axons of a population of neurons treated consecutively with a vehicle solution (30 minutes), the SST2R agonist octreotide 100nM or 1µM (30 minutes) and forskolin 10µM (Fsk) (30 minutes). Both in somata and axons, the addition of Fsk induced a strong increase of FRET ratio, in accordance with the increase of PKA activity (Figure 47A-47B). In somata, while vehicle solution did not change the FRET ratio, octreotide both at 100nM and 1µM induced a significant decrease of FRET ratio as compared to vehicle (Figure 47A1-47A2). Thus, activation of SST2Rs induces a decrease of basal PKA activity in somata. However, in axons, compared with the vehicle, octreotide did not change the FRET ratio (Figure 47B1-47B2). These results are in accordance with the somatodendritic localization of SST2Rs.

236

Figure 47: Activation of endogenous SST2Rs decreases PKA activity in somata but not in the axons. Hippocampal cultured neurons were transfected with AKAR4-Kras and imaged by FRET videomicroscopy. After 30 minutes of baseline, Vehicle solution was added. 30 minutes later octreotide 100nM (red) or 1µM (black) was added. Finally, Fsk 10µM was added. Curves represent mean +/- SEM of the FRET ratio for each time point in a population of neurons either in somata (A-A1) or in axons (B-B1). (A2-B2) For each treatment, FRET responses were calculated between 6 and 16 minutes after treatment addition (Vehicle or octreotide) in somata (A2) and axons (B2). Histograms represent mean +/- SEM of FRET responses. Statistical analysis were realized with one-way ANOVA followed by Newmann-Keuls post-test; NS p>0.05, * p<0.05, ** p<0.01.

237 3.2. Construction and validation of THPDE5VV, a new cGMP FRET probe. In order to measure cGMP concentration in live cells, several FRET probes have been constructed and validated. Notably, Nikolaev and colleagues constructed a cGMP FRET probe called cGES-DE5 (Nikolaev et al, 2006). This fusion protein contains a CFP and a YFP surrounding the human PDE5A1 GAF-A domain that binds cGMP. cGES-DE5 displays a good selectivity for cGMP as compared to cAMP. Compared to a previous construction called Cygnet 2.1, the amplitude of cGES-DE5 response to an increase in cGMP concentration was found higher and more reversible. We sought to improve cGES-DE5 probe by using the very efficient FRET fluorophores from TEpacVV probe that contains one mTurquoise coupled to two mVenus (Klarenbeek et al, 2011). For this, we obtained the sequence of the truncated human PDE5A1 (HPDE5) by gene synthesis in pUC57 vector (Genscript) and digested it with SmaI and NheI enzymes (New England Biolabs). Then, we removed Epac1 sequence from TEpacVV vector by digestion with EcoRV and NheI (New England Biolabs) (Figure 48A) and replaced it with HPDE5 (Figure 48B). We transfected this new construct, called THPDE5VV, in cultured hippocampal neurons of rats and performed FRET imaging. First, we added 50µM of DEA NONOate, a compound that releases NO in the medium and activates guanylate cyclases. We observed an increase of the FRET ratio both in the somatodendritic compartment (Figure 49A) and in the axon (Figure 49B) of individual neurons. In a population of neurons, this increase is 10-fold stronger than the increase induced by addition of Forskolin (10µM) both in somata (Figure 49C) and axons (Figure 49D). Moreover, response to NO is blocked by a 10-minutes pretreatment with ODQ, an inhibitor of guanylate cyclases (Figure 49C). These results indicate that THPDE5VV probe displays a good specificity for cGMP as compared to cAMP. Thus, our construct specifically detects cGMP concentration variations with high FRET responses (around 20% in somata and 40% in axons after DEA NONOate 50µM).

238 Figure 48: cGMP biosensor construction.

(A) Sequence map of TEpacVV FRET probe expression vector. Digestion sites EcoRV and NheI are indicated (framed).

(B) Sequence map of THPDE5VV FRET probe expression vector.

239

Figure 49: Validation of THPDE5VV probe. (A-B) Hippocampal cultured neurons were transfected with THPDE5VV probe and imaged by FRET videomicroscopy. Addition of DEA NONOate 50µM induced a strong increase of the FRET ratio both in the soma (A2) and axon (B2) compared to their respective baselines (A1-B1). (C-D) FRET responses were measured between 6 and 16 minutes after treatment in a population of neurons and represented as mean +/- SEM for somata (C) and axons (D). DEA NONOate (NO) was used at 50µM and Forskoline (Fsk) at 10µM. 1H- [1,2,4]Oxadiazolo[4,3-a]quinoxalin-1-one (ODQ), an inhibitor of guanylate cyclases was added at 100µM 10 minutes before the addition of DEA NONOate. Statistical analysis were realized with one-way ANOVA followed by Newmann-Keuls post-test; NS p>0.05, * p<0.05, ** p<0.01.

240 3.3. Activation of endogenous SST2Rs potentiates guanylate cyclases activation. We measured the modulation of cGMP concentration downstream of endogenous SST2Rs in cultured hippocampal neurons of rats expressing THPDE5VV probe. Neurons were treated consecutively with a vehicle solution (30 minutes), the SST2R agonist octreotide 100nM or 1µM (30 minutes) and DEANONOate 50µM (NO) (30 minutes). As above results show that SST2Rs modulate PKA activity in somata and not in axons, we measured FRET ratio in somata of a population of neurons (Figure 50A-50A1). While addition of octreotide did not seem to change FRET ratio as compared to vehicle (Figure 50A1), NO response seems affected by the increase of octreotide concentration (Figure 50A). Indeed, NO response after treatment with octreotide 1µM is significantly stronger than NO response after treatment with octreotide 100nM (Figure 50A2). These results suggest that the activation of endogenous SST2Rs does not directly modulate cGMP concentration in somata but it is able to potentiate guanylate cyclase activity.

Figure 50: Activation of endogenous SST2Rs potentiates guanylate cyclases activation. T VV (A-A1)Hippocampal cultured neurons were transfected with HPDE5 and imaged by FRET videomicroscopy. After 30 minutes of baseline, Vehicle solution was added. 30 minutes later octreotide 100nM (red) or 1µM (black) was added. Finally, DEANONOate (NO) 50µM was added. Curves represent mean +/- SEM of FRET ratio for each time point in somata of a population of neurons. (A2) For each treatment, FRET responses were calculated between 6 and 16 minutes after treatment addition (Vehicle or NO) in somata. Histograms represent mean +/- SEM of FRET responses. Statistical analysis were realized with one-way ANOVA followed by Newmann-Keuls post-test; *** p<0.001. 241 4. Discussion These preliminary results show that we developed a new FRET probe, THPDE5VV, that detects cGMP concentration. Some experiments remain to be performed in order to validate properly this construction (increase the number of quantified neurons in Figure 49 and perform a dose-response of DEANONOate). Interestingly, endogenous SST2R activation inhibits PKA activity in somata but not in axons of cultured neurons between DIV9 and DIV10 indicating that neurons are mature at this developmental stage. The decrease of basal PKA activity after application of octreotide 100nM is not significantly different from the decrease measured after octreotide 1µM treatment. This result indicates that either octreotide 100nM corresponds to saturating concentration of SST2R agonist or that the PKA activity can not be more inhibited. Moreover, while no increase of the cGMP concentration is measurable in somata after SST2R activation, increasing the concentration of octreotide increases the response to NO 50µM. This indicates that octreotide 100nM is not a saturating concentration for SST2R activation. Thus, one can assume that the decrease of PKA activity observed after octreotide 100nM or 1µM in somata corresponds to the minimum PKA activity that the system can achieve. The stronger NO-induced increase of cGMP concentration in somata after octreotide 1µM compared to octreotide 100nM indicates that SST2R activation potentiates guanylate cyclase activity. Which mechanism could explain this potentiation? Some studies have already reported that SST2R activation leads to cGMP concentration increase through the activation of Src homology region 2 domain-containing phosphatase-1 (SHP-1) (Mastrodimou et al,

2006). Notably, SHP-1 associates with SST2Rs at basal state through Giα3 proteins (Lopez et al, 1997). Activation of SST2Rs leads to the dissociation and activation of SHP-1 that dephosphorylate nNOS and increases NO production (Lopez et al, 2001) (Figure 51, pathway 1). The release of NO increases cGMP concentration through guanylate cyclases activation. However, if this pathway is involved downstream of somatic SST2Rs in cultured hippocampal neurons, we should have measured an increase of cGMP concentration after octreotide treatment and it does not explain the observed potentiation of NO-induced cGMP increase. Another possible mechanism involves the reciprocal regulation of cAMP and cGMP observed in developing neurons (Shelly et al, 2010). This process implies that cAMP activates Phosphodiesterase 3 (PDE3) that hydrolysis cGMP. Thus, an increase of cAMP concentration leads to a decrease of cGMP. Thus, one can assume that a decrease of basal cAMP concentration could lead to a decrease of PDE3 activity and thus increase cGMP

242 concentration (Figure 51, pathway 2). However, with this mechanism, we should have measured an increase of basal cGMP concentration and it does not explain the guanylate cyclase potentiation. Finally, an interesting endogenous molecule, called isatin (indole-2,3- dione), is known to inhibit guanylate cyclase activity (Medvedev et al, 2002). Interestingly, in rat brain, the highest concentration of isatin was found in the hippocampus (Watkins et al, 1990). One can assume that activation of SST2Rs decreases isatin concentration, leading to a removal of inhibition on guanylate cyclases (Figure 51, pathway 3). However, the synthesis and degradation pathway of isatin in neurons are still unclear (Medvedev et al, 2007) and further study is required in order to determine if SST2Rs can influence isatin concentration in hippocampal neurons.

Figure 51: How do SST2Rs potentiate guanylate cyclase activity? (i) SST2R activation is known to increase cGMP concentration through SHP- 1/nNOS/GC activation (pathway 1, orange) (Mastrodimou et al, 2006). (ii) Reciprocal regulation of cAMP and cGMP could also explain the potentiation of GC after SST2Rs activation (pathway 2, blue) (Shelly et al, 2010). (iii) Isatin (indole-2,3-dione) is an endogenous antagonist of guanylate cyclases. Possibly, SST2R activation can modulate isatin concentration leading to a removal of inhibition on GC (pathway 3, green). AC: adenylyl cylclase; PDE3: phosphodiesterase 3; SHP-1: Src homology region 2 domain-containing phosphatase-1; nNOS: neuronal NO synthase; GC: guanylate cylclase.

243

References

244

Adams SR, Harootunian AT, Buechler YJ, Taylor SS, Tsien RY (1991) Fluorescence ratio imaging of cyclic AMP in single cells. Nature 349: 694-697

Adan RA, Kas MJ (2003) Inverse agonism gains weight. Trends Pharmacol Sci 24: 315-321

Agrawal A, Lynskey MT (2009) Candidate genes for cannabis use disorders: findings, challenges and directions. Addiction 104: 518-532

Agrawal A, Lynskey MT, Bucholz KK, Martin NG, Madden PA, Heath AC (2007) Contrasting models of genetic co-morbidity for cannabis and other illicit drugs in adult Australian twins. Psychol Med 37: 49-60

Aguado T, Monory K, Palazuelos J, Stella N, Cravatt B, Lutz B, Marsicano G, Kokaia Z, Guzman M, Galve-Roperh I (2005) The endocannabinoid system drives neural progenitor proliferation. FASEB J 19: 1704-1706

Aguado T, Palazuelos J, Monory K, Stella N, Cravatt B, Lutz B, Marsicano G, Kokaia Z, Guzman M, Galve-Roperh I (2006) The endocannabinoid system promotes astroglial differentiation by acting on neural progenitor cells. J Neurosci 26: 1551-1561

Alger BE, Kim J (2011) Supply and demand for endocannabinoids. Trends Neurosci 34: 304- 315

Allen JA, Halverson-Tamboli RA, Rasenick MM (2007) Lipid raft microdomains and neurotransmitter signalling. Nat Rev Neurosci 8: 128-140

Allen MD, Zhang J (2006) Subcellular dynamics of protein kinase A activity visualized by FRET-based reporters. Biochem Biophys Res Commun 348: 716-721

Altenbach C, Klein-Seetharaman J, Cai K, Khorana HG, Hubbell WL (2001) Structure and function in rhodopsin: mapping light-dependent changes in distance between residue 316 in helix 8 and residues in the sequence 60-75, covering the cytoplasmic end of helices TM1 and TM2 and their connection loop CL1. Biochemistry 40: 15493-15500

Andersen SS, Bi GQ (2000) Axon formation: a molecular model for the generation of neuronal polarity. Bioessays 22: 172-179

Arimura N, Kaibuchi K (2007) Neuronal polarity: from extracellular signals to intracellular mechanisms. Nat Rev Neurosci 8: 194-205

Ascano M, Richmond A, Borden P, Kuruvilla R (2009) Axonal targeting of Trk receptors via transcytosis regulates sensitivity to neurotrophin responses. J Neurosci 29: 11674-11685

Azzi M, Charest PG, Angers S, Rousseau G, Kohout T, Bouvier M, Pineyro G (2003) Beta- arrestin-mediated activation of MAPK by inverse agonists reveals distinct active conformations for G protein-coupled receptors. Proc Natl Acad Sci U S A 100: 11406-11411

Bacci A, Huguenard JR, Prince DA (2004) Long-lasting self-inhibition of neocortical interneurons mediated by endocannabinoids. Nature 431: 312-316

245

Bacskai BJ, Hochner B, Mahaut-Smith M, Adams SR, Kaang BK, Kandel ER, Tsien RY (1993) Spatially resolved dynamics of cAMP and protein kinase A subunits in Aplysia sensory neurons. Science 260: 222-226

Baillie GL, Horswill JG, Anavi-Goffer S, Reggio PH, Bolognini D, Abood ME, McAllister S, Strange PG, Stephens GJ, Pertwee RG, Ross RA (2013) CB(1) receptor allosteric modulators display both agonist and signaling pathway specificity. Mol Pharmacol 83: 322-338

Baker D, Pryce G, Davies WL, Hiley CR (2006) In silico patent searching reveals a new cannabinoid receptor. Trends Pharmacol Sci 27: 1-4

Ballesteros JA, Shi L, Javitch JA (2001) Structural mimicry in G protein-coupled receptors: implications of the high-resolution structure of rhodopsin for structure-function analysis of rhodopsin-like receptors. Mol Pharmacol 60: 1-19

Banker GA, Cowan WM (1977) Rat hippocampal neurons in dispersed cell culture. Brain Res 126: 397-342

Barnes NM, Sharp T (1999) A review of central 5-HT receptors and their function. Neuropharmacology 38: 1083-1152

Bartlett WP, Banker GA (1984) An electron microscopic study of the development of axons and dendrites by hippocampal neurons in culture. II. Synaptic relationships. J Neurosci 4: 1954-1965

Baur R, Kielar M, Richter L, Ernst M, Ecker GF, Sigel E (2013) Molecular analysis of the site for 2-arachidonylglycerol (2-AG) on the beta(2) subunit of GABA(A) receptors. J Neurochem 126: 29-36

Bel C, Oguievetskaia K, Pitaval C, Goutebroze L, Faivre-Sarrailh C (2009) Axonal targeting of Caspr2 in hippocampal neurons via selective somatodendritic endocytosis. J Cell Sci 122: 3403-3413

Berg KA, Harvey JA, Spampinato U, Clarke WP (2005) Physiological relevance of constitutive activity of 5-HT2A and 5-HT2C receptors. Trends Pharmacol Sci 26: 625-630

Berger C, Schmid PC, Schabitz WR, Wolf M, Schwab S, Schmid HH (2004) Massive accumulation of N-acylethanolamines after stroke. Cell signalling in acute cerebral ischemia? J Neurochem 88: 1159-1167

Bertalovitz AC, Ahn KH, Kendall DA (2010) Ligand Binding Sensitivity of the Extracellular Loop Two of the Cannabinoid Receptor 1. Drug Dev Res 71: 404-411

Birnbaumer L, Abramowitz J, Brown AM (1990) Receptor-effector coupling by G proteins. Biochim Biophys Acta 1031: 163-224

Bisogno T, Howell F, Williams G, Minassi A, Cascio MG, Ligresti A, Matias I, Schiano- Moriello A, Paul P, Williams EJ, Gangadharan U, Hobbs C, Di Marzo V, Doherty P (2003)

246 Cloning of the first sn1-DAG lipases points to the spatial and temporal regulation of endocannabinoid signaling in the brain. J Cell Biol 163: 463-468

Bisogno T, Ligresti A, Di Marzo V (2005) The endocannabinoid signalling system: biochemical aspects. Pharmacol Biochem Behav 81: 224-238

Boekholdt SM, Peters RJ (2010) Rimonabant: obituary for a wonder drug. Lancet 376: 489- 490

Boess FG, Martin IL (1994) Molecular biology of 5-HT receptors. Neuropharmacology 33: 275-317

Bond RA, Leff P, Johnson TD, Milano CA, Rockman HA, McMinn TR, Apparsundaram S, Hyek MF, Kenakin TP, Allen LF, et al. (1995) Physiological effects of inverse agonists in transgenic mice with myocardial overexpression of the beta 2-adrenoceptor. Nature 374: 272- 276

Bos JL (2003) Epac: a new cAMP target and new avenues in cAMP research. Nat Rev Mol Cell Biol 4: 733-738

Bouhelal R, Smounya L, Bockaert J (1988) 5-HT1B receptors are negatively coupled with adenylate cyclase in rat substantia nigra. Eur J Pharmacol 151: 189-196

Bradke F, Dotti CG (1997) Neuronal polarity: vectorial cytoplasmic flow precedes axon formation. Neuron 19: 1175-1186

Bradke F, Dotti CG (1999) The role of local actin instability in axon formation. Science 283: 1931-1934

Bredt DS, Snyder SH (1992) Nitric oxide, a novel neuronal messenger. Neuron 8: 3-11

Brinkmann V, Billich A, Baumruker T, Heining P, Schmouder R, Francis G, Aradhye S, Burtin P (2010) Fingolimod (FTY720): discovery and development of an oral drug to treat multiple sclerosis. Nat Rev Drug Discov 9: 883-897

Broussard JA, Rappaz B, Webb DJ, Brown CM (2013) Fluorescence resonance energy transfer microscopy as demonstrated by measuring the activation of the serine/threonine kinase Akt. Nat Protoc 8: 265-281

Bruggemann EP, Melchior DL (1983) Alterations in the organization of phosphatidylcholine/cholesterol bilayers by tetrahydrocannabinol. J Biol Chem 258: 8298- 8303

Buggia-Prevot V, Fernandez CG, Riordan S, Vetrivel KS, Roseman J, Waters J, Bindokas VP, Vassar R, Thinakaran G (2014) Axonal BACE1 dynamics and targeting in hippocampal neurons: a role for Rab11 GTPase. Mol Neurodegener 9: 1

Burack MA, Silverman MA, Banker G (2000) The role of selective transport in neuronal protein sorting. Neuron 26: 465-472

247 Burdyga A, Conant A, Haynes L, Zhang J, Jalink K, Sutton R, Neoptolemos J, Costello E, Tepikin A (2013) cAMP inhibits migration, ruffling and paxillin accumulation in focal adhesions of pancreatic ductal adenocarcinoma cells: effects of PKA and EPAC. Biochim Biophys Acta 1833: 2664-2672

Caille S, Alvarez-Jaimes L, Polis I, Stouffer DG, Parsons LH (2007) Specific alterations of extracellular endocannabinoid levels in the nucleus accumbens by ethanol, heroin, and cocaine self-administration. J Neurosci 27: 3695-3702

Campbell FA, Tramer MR, Carroll D, Reynolds DJ, Moore RA, McQuay HJ (2001) Are cannabinoids an effective and safe treatment option in the management of pain? A qualitative systematic review. BMJ 323: 13-16

Carrel D, Hamon M, Darmon M (2006) Role of the C-terminal di-leucine motif of 5-HT1A and 5-HT1B serotonin receptors in plasma membrane targeting. J Cell Sci 119: 4276-4284

Carrel D, Simon A, Emerit MB, Rivals I, Leterrier C, Biard M, Hamon M, Darmon M, Lenkei Z (2011) Axonal targeting of the 5-HT1B serotonin receptor relies on structure-specific constitutive activation. Traffic 12: 1501-1520

Chandrashekar J, Mueller KL, Hoon MA, Adler E, Feng L, Guo W, Zuker CS, Ryba NJ (2000) T2Rs function as bitter taste receptors. Cell 100: 703-711

Cheng PL, Poo MM (2012) Early events in axon/dendrite polarization. Annu Rev Neurosci 35: 181-201

Chevaleyre V, Castillo PE (2003) Heterosynaptic LTD of hippocampal GABAergic synapses: a novel role of endocannabinoids in regulating excitability. Neuron 38: 461-472

Chevaleyre V, Takahashi KA, Castillo PE (2006) Endocannabinoid-mediated synaptic plasticity in the CNS. Annu Rev Neurosci 29: 37-76

Chiu TT, Yung LY, Wong YH (1996) Inverse agonistic effect of ICI-174,864 on the cloned delta-opioid receptor: role of G protein and adenylyl cyclase activation. Mol Pharmacol 50: 1651-1657

Choi JW, Chun J (2013) Lysophospholipids and their receptors in the central nervous system. Biochim Biophys Acta 1831: 20-32

Chun J, Hartung HP (2010) Mechanism of action of oral fingolimod (FTY720) in multiple sclerosis. Clin Neuropharmacol 33: 91-101

Conner SD, Schmid SL (2003) Regulated portals of entry into the cell. Nature 422: 37-44

Coppen AJ, Doogan DP (1988) Serotonin and its place in the pathogenesis of depression. J Clin Psychiatry 49 Suppl: 4-11

Corbin JD, Keely SL, Park CR (1975) The distribution and dissociation of cyclic adenosine 3':5'-monophosphate-dependent protein kinases in adipose, cardiac, and other tissues. J Biol Chem 250: 218-225

248

Costa T, Herz A (1989) Antagonists with negative intrinsic activity at delta opioid receptors coupled to GTP-binding proteins. Proc Natl Acad Sci U S A 86: 7321-7325

Craig AM, Banker G (1994) Neuronal polarity. Annu Rev Neurosci 17: 267-310

Csaba Z, Lelouvier B, Viollet C, El Ghouzzi V, Toyama K, Videau C, Bernard V, Dournaud P (2007) Activated somatostatin type 2 receptors traffic in vivo in central neurons from dendrites to the trans Golgi before recycling. Traffic 8: 820-834

Curzon G (1990) Serotonin and appetite. Ann N Y Acad Sci 600: 521-530; discussion 530-521

D'Ambra TE, Estep KG, Bell MR, Eissenstat MA, Josef KA, Ward SJ, Haycock DA, Baizman ER, Casiano FM, Beglin NC, et al. (1992) Conformationally restrained analogues of pravadoline: nanomolar potent, enantioselective, (aminoalkyl)indole agonists of the cannabinoid receptor. J Med Chem 35: 124-135

D'Antona AM, Ahn KH, Kendall DA (2006) Mutations of CB1 T210 produce active and inactive receptor forms: correlations with ligand affinity, receptor stability, and cellular localization. Biochemistry 45: 5606-5617

Davis MI, Ronesi J, Lovinger DM (2003) A predominant role for inhibition of the adenylate cyclase/protein kinase A pathway in ERK activation by cannabinoid receptor 1 in N1E-115 neuroblastoma cells. J Biol Chem 278: 48973-48980

De Deurwaerdere P, Navailles S, Berg KA, Clarke WP, Spampinato U (2004) Constitutive activity of the serotonin2C receptor inhibits in vivo dopamine release in the rat striatum and nucleus accumbens. J Neurosci 24: 3235-3241

De Lean A, Stadel JM, Lefkowitz RJ (1980) A ternary complex model explains the agonist- specific binding properties of the adenylate cyclase-coupled beta-adrenergic receptor. J Biol Chem 255: 7108-7117

Depry C, Allen MD, Zhang J (2011) Visualization of PKA activity in plasma membrane microdomains. Mol Biosyst 7: 52-58

Derkinderen P, Ledent C, Parmentier M, Girault JA (2001) Cannabinoids activate p38 mitogen-activated protein kinases through CB1 receptors in hippocampus. J Neurochem 77: 957-960

Deupi X, Kobilka B (2007) Activation of G protein-coupled receptors. Adv Protein Chem 74: 137-166

Devane WA, Dysarz FA, 3rd, Johnson MR, Melvin LS, Howlett AC (1988) Determination and characterization of a cannabinoid receptor in rat brain. Mol Pharmacol 34: 605-613

Devane WA, Hanus L, Breuer A, Pertwee RG, Stevenson LA, Griffin G, Gibson D, Mandelbaum A, Etinger A, Mechoulam R (1992) Isolation and structure of a brain constituent that binds to the cannabinoid receptor. Science 258: 1946-1949

249 Di Marzo V, De Petrocellis L (2012) Why do cannabinoid receptors have more than one endogenous ligand? Philos Trans R Soc Lond B Biol Sci 367: 3216-3228

Diana MA, Marty A (2004) Endocannabinoid-mediated short-term synaptic plasticity: depolarization-induced suppression of inhibition (DSI) and depolarization-induced suppression of excitation (DSE). Br J Pharmacol 142: 9-19

Dotti CG, Sullivan CA, Banker GA (1988) The establishment of polarity by hippocampal neurons in culture. J Neurosci 8: 1454-1468

Drews J (2000) Drug discovery: a historical perspective. Science 287: 1960-1964

Egertova M, Elphick MR (2000) Localisation of cannabinoid receptors in the rat brain using antibodies to the intracellular C-terminal tail of CB. J Comp Neurol 422: 159-171

Engelman DM (2005) Membranes are more mosaic than fluid. Nature 438: 578-580

Farrens DL, Altenbach C, Yang K, Hubbell WL, Khorana HG (1996) Requirement of rigid- body motion of transmembrane helices for light activation of rhodopsin. Science 274: 768-770

Filbey FM, Schacht JP, Myers US, Chavez RS, Hutchison KE (2010) Individual and additive effects of the CNR1 and FAAH genes on brain response to marijuana cues. Neuropsychopharmacology 35: 967-975

Francischetti EA, de Abreu VG (2006) The endocannabinoid system: a new perspective for cardiometabolic risk control. Arq Bras Cardiol 87: 548-558

Fredriksson R, Lagerstrom MC, Lundin LG, Schioth HB (2003) The G-protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints. Mol Pharmacol 63: 1256-1272

Freund TF, Katona I, Piomelli D (2003) Role of endogenous cannabinoids in synaptic signaling. Physiol Rev 83: 1017-1066

Fuller, R. W. (1992). Clinical applications of 5-HT uptake inhibitors. In Serotonin, CNS receptors and brain function (eds P. B. Bradley, S. Handley, S. J. Cooper, B. J. Key, N. M. Barnes and J. H. Coote), pp. 255-70. Oxford: Pergamon Press

Futerman AH, Khanin R, Segel LA (1993) Lipid diffusion in neurons. Nature 362: 119

Gaffuri AL, Ladarre D, Lenkei Z (2012) Type-1 Cannabinoid Receptor Signaling in Neuronal Development. Pharmacology 90: 19-39

Gao Y, Vasilyev DV, Goncalves MB, Howell FV, Hobbs C, Reisenberg M, Shen R, Zhang MY, Strassle BW, Lu P, Mark L, Piesla MJ, Deng K, Kouranova EV, Ring RH, Whiteside GT, Bates B, Walsh FS, Williams G, Pangalos MN, Samad TA, Doherty P (2010) Loss of retrograde endocannabinoid signaling and reduced adult neurogenesis in diacylglycerol lipase knock-out mice. J Neurosci 30: 2017-2024

250 Garrido JJ, Fernandes F, Giraud P, Mouret I, Pasqualini E, Fache MP, Jullien F, Dargent B (2001) Identification of an axonal determinant in the C-terminus of the sodium channel Na(v)1.2. EMBO J 20: 5950-5961

Gatley SJ, Gifford AN, Volkow ND, Lan R, Makriyannis A (1996) 123I-labeled AM251: a radioiodinated ligand which binds in vivo to mouse brain cannabinoid CB1 receptors. Eur J Pharmacol 307: 331-338

Gatley SJ, Lan R, Volkow ND, Pappas N, King P, Wong CT, Gifford AN, Pyatt B, Dewey SL, Makriyannis A (1998) Imaging the brain marijuana receptor: development of a radioligand that binds to cannabinoid CB1 receptors in vivo. J Neurochem 70: 417-423

Gebremedhin D, Lange AR, Campbell WB, Hillard CJ, Harder DR (1999) Cannabinoid CB1 receptor of cat cerebral arterial muscle functions to inhibit L-type Ca2+ channel current. Am J Physiol 276: H2085-2093

Gerard C, Mollereau C, Vassart G, Parmentier M (1990) Nucleotide sequence of a human cannabinoid receptor cDNA. Nucleic Acids Res 18: 7142

Gerdeman GL, Ronesi J, Lovinger DM (2002) Postsynaptic endocannabinoid release is critical to long-term depression in the striatum. Nat Neurosci 5: 446-451

Gimpl G, Burger K, Fahrenholz F (1997) Cholesterol as modulator of receptor function. Biochemistry 36: 10959-10974

Glass M, Felder CC (1997) Concurrent stimulation of cannabinoid CB1 and dopamine D2 receptors augments cAMP accumulation in striatal neurons: evidence for a Gs linkage to the CB1 receptor. J Neurosci 17: 5327-5333

Gloriam DE, Fredriksson R, Schioth HB (2007) The G protein-coupled receptor subset of the rat genome. BMC Genomics 8: 338

Gomez Del Pulgar T, De Ceballos ML, Guzman M, Velasco G (2002) Cannabinoids protect astrocytes from ceramide-induced apoptosis through the phosphatidylinositol 3-kinase/protein kinase B pathway. J Biol Chem 277: 36527-36533

Goslin K, Banker G (1989) Experimental observations on the development of polarity by hippocampal neurons in culture. J Cell Biol 108: 1507-1516

Greene LA, Tischler AS (1976) Establishment of a noradrenergic clonal line of rat adrenal pheochromocytoma cells which respond to nerve growth factor. Proc Natl Acad Sci U S A 73: 2424-2428

Hanoune J, Defer N (2001) Regulation and role of adenylyl cyclase isoforms. Annu Rev Pharmacol Toxicol 41: 145-174

Hansen HH, Ikonomidou C, Bittigau P, Hansen SH, Hansen HS (2001) Accumulation of the anandamide precursor and other N-acylethanolamine phospholipids in infant rat models of in vivo necrotic and apoptotic neuronal death. J Neurochem 76: 39-46

251 Hanson MA, Cherezov V, Griffith MT, Roth CB, Jaakola VP, Chien EY, Velasquez J, Kuhn P, Stevens RC (2008) A specific cholesterol binding site is established by the 2.8 A structure of the human beta2-adrenergic receptor. Structure 16: 897-905

Hanson MA, Roth CB, Jo E, Griffith MT, Scott FL, Reinhart G, Desale H, Clemons B, Cahalan SM, Schuerer SC, Sanna MG, Han GW, Kuhn P, Rosen H, Stevens RC (2012) Crystal structure of a lipid G protein-coupled receptor. Science 335: 851-855

Hanus L, Abu-Lafi S, Fride E, Breuer A, Vogel Z, Shalev DE, Kustanovich I, Mechoulam R (2001) 2-arachidonyl glyceryl ether, an endogenous agonist of the cannabinoid CB1 receptor. Proc Natl Acad Sci U S A 98: 3662-3665

Hanus L, Gopher A, Almog S, Mechoulam R (1993) Two new unsaturated fatty acid ethanolamides in brain that bind to the cannabinoid receptor. J Med Chem 36: 3032-3034

Hanyaloglu AC, Vrecl M, Kroeger KM, Miles LE, Qian H, Thomas WG, Eidne KA (2001) Casein kinase II sites in the intracellular C-terminal domain of the thyrotropin-releasing hormone receptor and chimeric gonadotropin-releasing hormone receptors contribute to beta- arrestin-dependent internalization. J Biol Chem 276: 18066-18074

Hausdorff WP, Caron MG, Lefkowitz RJ (1990) Turning off the signal: desensitization of beta-adrenergic receptor function. FASEB J 4: 2881-2889

Hayflick JS (2000) A family of heptahelical receptors with adhesion-like domains: a marriage between two super families. J Recept Signal Transduct Res 20: 119-131

Heifets BD, Castillo PE (2009) Endocannabinoid signaling and long-term synaptic plasticity. Annu Rev Physiol 71: 283-306

Heijman J, Dewenter M, El-Armouche A, Dobrev D (2013) Function and regulation of serine/threonine phosphatases in the healthy and diseased heart. J Mol Cell Cardiol 64: 90-98

Heimann AS, Gomes I, Dale CS, Pagano RL, Gupta A, de Souza LL, Luchessi AD, Castro LM, Giorgi R, Rioli V, Ferro ES, Devi LA (2007) Hemopressin is an inverse agonist of CB1 cannabinoid receptors. Proc Natl Acad Sci U S A 104: 20588-20593

Hemar A, Olivo JC, Williamson E, Saffrich R, Dotti CG (1997) Dendroaxonal transcytosis of transferrin in cultured hippocampal and sympathetic neurons. J Neurosci 17: 9026-9034

Herkenham M, Lynn AB, Johnson MR, Melvin LS, de Costa BR, Rice KC (1991) Characterization and localization of cannabinoid receptors in rat brain: a quantitative in vitro autoradiographic study. J Neurosci 11: 563-583

Hermann A, Kaczocha M, Deutsch DG (2006) 2-Arachidonoylglycerol (2-AG) membrane transport: history and outlook. AAPS J 8: E409-412

Hery F, Hamon M (1993) [Neuroleptics and serotonin]. Encephale 19: 525-532

252 Hillard CJ, Harris RA, Bloom AS (1985) Effects of the cannabinoids on physical properties of brain membranes and phospholipid vesicles: fluorescence studies. J Pharmacol Exp Ther 232: 579-588

Hillard CJ, Muthian S, Kearn CS (1999) Effects of CB(1) cannabinoid receptor activation on cerebellar granule cell nitric oxide synthase activity. FEBS Lett 459: 277-281

Hillard CJ, Pounds JJ, Boyer DR, Bloom AS (1990) Studies of the role of membrane lipid order in the effects of delta 9-tetrahydrocannabinol on adenylate cyclase activation in heart. J Pharmacol Exp Ther 252: 1075-1082

Hoffman AF, Lupica CR (2000) Mechanisms of cannabinoid inhibition of GABA(A) synaptic transmission in the hippocampus. J Neurosci 20: 2470-2479

Hopfer CJ, Young SE, Purcell S, Crowley TJ, Stallings MC, Corley RP, Rhee SH, Smolen A, Krauter K, Hewitt JK, Ehringer MA (2006) Cannabis receptor haplotype associated with fewer cannabis dependence symptoms in adolescents. Am J Med Genet B Neuropsychiatr Genet 141B: 895-901

Horton AC, Ehlers MD (2003) Neuronal polarity and trafficking. Neuron 40: 277-295

Howlett AC, Fleming RM (1984) Cannabinoid inhibition of adenylate cyclase. Pharmacology of the response in neuroblastoma cell membranes. Mol Pharmacol 26: 532-538

Howlett AC, Qualy JM, Khachatrian LL (1986) Involvement of Gi in the inhibition of adenylate cyclase by cannabimimetic drugs. Mol Pharmacol 29: 307-313

Huizink AC, Mulder EJ (2006) Maternal smoking, drinking or cannabis use during pregnancy and neurobehavioral and cognitive functioning in human offspring. Neurosci Biobehav Rev 30: 24-41

Hurst DP, Schmeisser M, Reggio PH (2013) Endogenous lipid activated G protein-coupled receptors: emerging structural features from crystallography and molecular dynamics simulations. Chem Phys Lipids 169: 46-56

Ikeda SR (1996) Voltage-dependent modulation of N-type calcium channels by G-protein beta gamma subunits. Nature 380: 255-258

Ishikawa Y, Homcy CJ (1997) The adenylyl cyclases as integrators of transmembrane signal transduction. Circ Res 80: 297-304

Iversen SD (1984) 5-HT and anxiety. Neuropharmacology 23: 1553-1560

Jafurulla M, Tiwari S, Chattopadhyay A (2011) Identification of cholesterol recognition amino acid consensus (CRAC) motif in G-protein coupled receptors. Biochem Biophys Res Commun 404: 569-573

Jbilo O, Ravinet-Trillou C, Arnone M, Buisson I, Bribes E, Peleraux A, Penarier G, Soubrie P, Le Fur G, Galiegue S, Casellas P (2005) The CB1 receptor antagonist rimonabant reverses

253 the diet-induced obesity phenotype through the regulation of lipolysis and energy balance. FASEB J 19: 1567-1569

Ji TH, Grossmann M, Ji I (1998) G protein-coupled receptors. I. Diversity of receptor-ligand interactions. J Biol Chem 273: 17299-17302

Jin K, Xie L, Kim SH, Parmentier-Batteur S, Sun Y, Mao XO, Childs J, Greenberg DA (2004) Defective adult neurogenesis in CB1 cannabinoid receptor knockout mice. Mol Pharmacol 66: 204-208

Kaneko M, Takahashi T (2004) Presynaptic mechanism underlying cAMP-dependent synaptic potentiation. J Neurosci 24: 5202-5208

Kapur A, Hurst DP, Fleischer D, Whitnell R, Thakur GA, Makriyannis A, Reggio PH, Abood ME (2007) Mutation studies of Ser7.39 and Ser2.60 in the human CB1 cannabinoid receptor: evidence for a serine-induced bend in CB1 transmembrane helix 7. Mol Pharmacol 71: 1512- 1524

Katona I, Sperlagh B, Sik A, Kafalvi A, Vizi ES, Mackie K, Freund TF (1999) Presynaptically located CB1 cannabinoid receptors regulate GABA release from axon terminals of specific hippocampal interneurons. J Neurosci 19: 4544-4558

Katona I, Urban GM, Wallace M, Ledent C, Jung KM, Piomelli D, Mackie K, Freund TF (2006) Molecular composition of the endocannabinoid system at glutamatergic synapses. J Neurosci 26: 5628-5637

Kavalali ET, Klingauf J, Tsien RW (1999) Activity-dependent regulation of synaptic clustering in a hippocampal culture system. Proc Natl Acad Sci U S A 96: 12893-12900

Kawasaki H, Springett GM, Mochizuki N, Toki S, Nakaya M, Matsuda M, Housman DE, Graybiel AM (1998) A family of cAMP-binding proteins that directly activate Rap1. Science 282: 2275-2279

Kempf A, Tews B, Arzt ME, Weinmann O, Obermair FJ, Pernet V, Zagrebelsky M, Delekate A, Iobbi C, Zemmar A, Ristic Z, Gullo M, Spies P, Dodd D, Gygax D, Korte M, Schwab ME (2014) The sphingolipid receptor S1PR2 is a receptor for Nogo-a repressing synaptic plasticity. PLoS Biol 12: e1001763

Kenakin T (2004) Principles: receptor theory in pharmacology. Trends Pharmacol Sci 25: 186-192

Kenakin T (2009) Quantifying Biological Activity in Chemical Terms: A Pharmacology Primer to Describe Drug Effect. ACS Chem Biol

Kim SH, Won SJ, Mao XO, Jin K, Greenberg DA (2006) Molecular mechanisms of cannabinoid protection from neuronal excitotoxicity. Mol Pharmacol 69: 691-696

Klarenbeek JB, Goedhart J, Hink MA, Gadella TW, Jalink K (2011) A mTurquoise-based cAMP sensor for both FLIM and ratiometric read-out has improved dynamic range. PLoS One 6: e19170

254

Kobayashi T, Storrie B, Simons K, Dotti CG (1992) A functional barrier to movement of lipids in polarized neurons. Nature 359: 647-650

Kobilka BK, Deupi X (2007) Conformational complexity of G-protein-coupled receptors. Trends Pharmacol Sci 28: 397-406

Koesling D, Russwurm M, Mergia E, Mullershausen F, Friebe A (2004) Nitric oxide-sensitive guanylyl cyclase: structure and regulation. Neurochem Int 45: 813-819

Kreitzer AC, Regehr WG (2001) Retrograde inhibition of presynaptic calcium influx by endogenous cannabinoids at excitatory synapses onto Purkinje cells. Neuron 29: 717-727

Kunishima N, Shimada Y, Tsuji Y, Sato T, Yamamoto M, Kumasaka T, Nakanishi S, Jingami H, Morikawa K (2000) Structural basis of glutamate recognition by a dimeric metabotropic glutamate receptor. Nature 407: 971-977

Lagerstrom MC, Schioth HB (2008) Structural diversity of G protein-coupled receptors and significance for drug discovery. Nat Rev Drug Discov 7: 339-357

Lambert NA (2008) Dissociation of heterotrimeric g proteins in cells. Sci Signal 1: re5

Laprairie RB, Bagher AM, Kelly ME, Dupre DJ, Denovan-Wright EM (2014) Type 1 Cannabinoid Receptor Ligands Display Functional Selectivity in a Cell Culture Model of Striatal Medium Spiny Projection Neurons. J Biol Chem

Lasiecka ZM, Winckler B (2011) Mechanisms of polarized membrane trafficking in neurons - - focusing in on endosomes. Mol Cell Neurosci 48: 278-287

Lefkimmiatis K, Zaccolo M (2014) cAMP signaling in subcellular compartments. Pharmacol Ther

Lehtonen M, Reisner K, Auriola S, Wong G, Callaway JC (2008) Mass-spectrometric identification of anandamide and 2-arachidonoylglycerol in nematodes. Chem Biodivers 5: 2431-2441

Leker RR, Gai N, Mechoulam R, Ovadia H (2003) Drug-induced hypothermia reduces ischemic damage: effects of the cannabinoid HU-210. Stroke 34: 2000-2006

Lelouvier B, Tamagno G, Kaindl AM, Roland A, Lelievre V, Le Verche V, Loudes C, Gressens P, Faivre-Baumann A, Lenkei Z, Dournaud P (2008) Dynamics of somatostatin type 2A receptor cargoes in living hippocampal neurons. J Neurosci 28: 4336-4349

Leterrier C, Bonnard D, Carrel D, Rossier J, Lenkei Z (2004) Constitutive endocytic cycle of the CB1 cannabinoid receptor. J Biol Chem 279: 36013-36021

Leterrier C, Dargent B (2013) No Pasaran! Role of the axon initial segment in the regulation of protein transport and the maintenance of axonal identity. Semin Cell Dev Biol

255 Leterrier C, Laine J, Darmon M, Boudin H, Rossier J, Lenkei Z (2006) Constitutive activation drives compartment-selective endocytosis and axonal targeting of type 1 cannabinoid receptors. J Neurosci 26: 3141-3153

Levitt ES, Clark MJ, Jenkins PM, Martens JR, Traynor JR (2009) Differential effect of membrane cholesterol removal on mu- and delta-opioid receptors: a parallel comparison of acute and chronic signaling to adenylyl cyclase. J Biol Chem 284: 22108-22122

Lichtman AH, Hawkins EG, Griffin G, Cravatt BF (2002) Pharmacological activity of fatty acid amides is regulated, but not mediated, by fatty acid amide hydrolase in vivo. J Pharmacol Exp Ther 302: 73-79

Little PJ, Compton DR, Johnson MR, Melvin LS, Martin BR (1988) Pharmacology and stereoselectivity of structurally novel cannabinoids in mice. J Pharmacol Exp Ther 247: 1046- 1051

Liu J, Gao B, Mirshahi F, Sanyal AJ, Khanolkar AD, Makriyannis A, Kunos G (2000) Functional CB1 cannabinoid receptors in human vascular endothelial cells. Biochem J 346 Pt 3: 835-840

Lohse MJ, Benovic JL, Codina J, Caron MG, Lefkowitz RJ (1990) beta-Arrestin: a protein that regulates beta-adrenergic receptor function. Science 248: 1547-1550

Lopez F, Esteve JP, Buscail L, Delesque N, Saint-Laurent N, Theveniau M, Nahmias C, Vaysse N, Susini C (1997) The tyrosine phosphatase SHP-1 associates with the sst2 somatostatin receptor and is an essential component of sst2-mediated inhibitory growth signaling. J Biol Chem 272: 24448-24454

Lopez F, Ferjoux G, Cordelier P, Saint-Laurent N, Esteve JP, Vaysse N, Buscail L, Susini C (2001) Neuronal nitric oxide synthase: a substrate for SHP-1 involved in sst2 somatostatin receptor growth inhibitory signaling. FASEB J 15: 2300-2302

Lujan R, Maylie J, Adelman JP (2009) New sites of action for GIRK and SK channels. Nat Rev Neurosci 10: 475-480

Luttrell LM, Lefkowitz RJ (2002) The role of beta-arrestins in the termination and transduction of G-protein-coupled receptor signals. J Cell Sci 115: 455-465

Ma L, Zablow L, Kandel ER, Siegelbaum SA (1999) Cyclic AMP induces functional presynaptic boutons in hippocampal CA3-CA1 neuronal cultures. Nat Neurosci 2: 24-30

Mackie K (2006) Cannabinoid receptors as therapeutic targets. Annu Rev Pharmacol Toxicol 46: 101-122

Mackie K, Hille B (1992) Cannabinoids inhibit N-type calcium channels in neuroblastoma- glioma cells. Proc Natl Acad Sci U S A 89: 3825-3829

Mackie K, Lai Y, Westenbroek R, Mitchell R (1995) Cannabinoids activate an inwardly rectifying potassium conductance and inhibit Q-type calcium currents in AtT20 cells transfected with rat brain cannabinoid receptor. J Neurosci 15: 6552-6561

256

Malenka RC, Bear MF (2004) LTP and LTD: an embarrassment of riches. Neuron 44: 5-21

Manganiello VC, Degerman E (1999) Cyclic nucleotide phosphodiesterases (PDEs): diverse regulators of cyclic nucleotide signals and inviting molecular targets for novel therapeutic agents. Thromb Haemost 82: 407-411

Marinelli S, Pacioni S, Bisogno T, Di Marzo V, Prince DA, Huguenard JR, Bacci A (2008) The endocannabinoid 2-arachidonoylglycerol is responsible for the slow self-inhibition in neocortical interneurons. J Neurosci 28: 13532-13541

Marinissen MJ, Gutkind JS (2001) G-protein-coupled receptors and signaling networks: emerging paradigms. Trends Pharmacol Sci 22: 368-376

Marrs WR, Blankman JL, Horne EA, Thomazeau A, Lin YH, Coy J, Bodor AL, Muccioli GG, Hu SS, Woodruff G, Fung S, Lafourcade M, Alexander JP, Long JZ, Li W, Xu C, Moller T, Mackie K, Manzoni OJ, Cravatt BF, Stella N (2010) The serine hydrolase ABHD6 controls the accumulation and efficacy of 2-AG at cannabinoid receptors. Nat Neurosci 13: 951-957

Marsicano G, Goodenough S, Monory K, Hermann H, Eder M, Cannich A, Azad SC, Cascio MG, Gutierrez SO, van der Stelt M, Lopez-Rodriguez ML, Casanova E, Schutz G, Zieglgansberger W, Di Marzo V, Behl C, Lutz B (2003) CB1 cannabinoid receptors and on- demand defense against excitotoxicity. Science 302: 84-88

Marsicano G, Lutz B (1999) Expression of the cannabinoid receptor CB1 in distinct neuronal subpopulations in the adult mouse forebrain. Eur J Neurosci 11: 4213-4225

Marsicano G, Wotjak CT, Azad SC, Bisogno T, Rammes G, Cascio MG, Hermann H, Tang J, Hofmann C, Zieglgansberger W, Di Marzo V, Lutz B (2002) The endogenous cannabinoid system controls extinction of aversive memories. Nature 418: 530-534

Martin NP, Whalen EJ, Zamah MA, Pierce KL, Lefkowitz RJ (2004) PKA-mediated phosphorylation of the beta1-adrenergic receptor promotes Gs/Gi switching. Cell Signal 16: 1397-1403

Martini L, Waldhoer M, Pusch M, Kharazia V, Fong J, Lee JH, Freissmuth C, Whistler JL (2007) Ligand-induced down-regulation of the cannabinoid 1 receptor is mediated by the G- protein-coupled receptor-associated sorting protein GASP1. FASEB J 21: 802-811

Mastrodimou N, Kiagiadaki F, Hodjarova M, Karagianni E, Thermos K (2006) Somatostatin receptors (sst2) regulate cGMP production in rat retina. Regul Pept 133: 41-46

Matsuda LA, Lolait SJ, Brownstein MJ, Young AC, Bonner TI (1990) Structure of a cannabinoid receptor and functional expression of the cloned cDNA. Nature 346: 561-564

Matulef K, Zagotta WN (2003) Cyclic nucleotide-gated ion channels. Annu Rev Cell Dev Biol 19: 23-44

257 Matyas F, Yanovsky Y, Mackie K, Kelsch W, Misgeld U, Freund TF (2006) Subcellular localization of type 1 cannabinoid receptors in the rat basal ganglia. Neuroscience 137: 337- 361

McDonald NA, Henstridge CM, Connolly CN, Irving AJ (2007) An essential role for constitutive endocytosis, but not activity, in the axonal targeting of the CB1 cannabinoid receptor. Mol Pharmacol 71: 976-984

Mechoulam R, Ben-Shabat S, Hanus L, Ligumsky M, Kaminski NE, Schatz AR, Gopher A, Almog S, Martin BR, Compton DR, et al. (1995) Identification of an endogenous 2- monoglyceride, present in canine gut, that binds to cannabinoid receptors. Biochem Pharmacol 50: 83-90

Mechoulam R, Braun P, Gaoni Y (1967) A stereospecific synthesis of (-)-delta 1- and (-)- delta 1(6)-tetrahydrocannabinols. J Am Chem Soc 89: 4552-4554

Mechoulam R, Gaoni Y (1967) The absolute configuration of delta-1-tetrahydrocannabinol, the major active constituent of hashish. Tetrahedron Lett 12: 1109-1111

Medvedev A, Buneeva O, Glover V (2007) Biological targets for isatin and its analogues: Implications for therapy. Biologics 1: 151-162

Medvedev A, Bussygyna O, Pyatakova N, Glover V, Severina I (2002) Effect of isatin on nitric oxide-stimulated soluble guanylate cyclase from human platelets. Biochem Pharmacol 63: 763-766

Meltzer HY (1995) Role of serotonin in the action of atypical antipsychotic drugs. Clin Neurosci 3: 64-75

Meriney SD, Gray DB, Pilar GR (1994) Somatostatin-induced inhibition of neuronal Ca2+ current modulated by cGMP-dependent protein kinase. Nature 369: 336-339

Meye FJ, Ramakers GM, Adan RA (2014) The vital role of constitutive GPCR activity in the mesolimbic dopamine system. Transl Psychiatry 4: e361

Monory K, Massa F, Egertova M, Eder M, Blaudzun H, Westenbroek R, Kelsch W, Jacob W, Marsch R, Ekker M, Long J, Rubenstein JL, Goebbels S, Nave KA, During M, Klugmann M, Wolfel B, Dodt HU, Zieglgansberger W, Wotjak CT, Mackie K, Elphick MR, Marsicano G, Lutz B (2006) The endocannabinoid system controls key epileptogenic circuits in the hippocampus. Neuron 51: 455-466

Morisset S, Rouleau A, Ligneau X, Gbahou F, Tardivel-Lacombe J, Stark H, Schunack W, Ganellin CR, Schwartz JC, Arrang JM (2000) High constitutive activity of native H3 receptors regulates histamine neurons in brain. Nature 408: 860-864

Munro S, Thomas KL, Abu-Shaar M (1993) Molecular characterization of a peripheral receptor for cannabinoids. Nature 365: 61-65

Namkung Y, Sibley DR (2004) Protein kinase C mediates phosphorylation, desensitization, and trafficking of the D2 dopamine receptor. J Biol Chem 279: 49533-49541

258

Neves SR, Tsokas P, Sarkar A, Grace EA, Rangamani P, Taubenfeld SM, Alberini CM, Schaff JC, Blitzer RD, Moraru, II, Iyengar R (2008) Cell shape and negative links in regulatory motifs together control spatial information flow in signaling networks. Cell 133: 666-680

Nicol X, Gaspar P (2014) Routes to cAMP: shaping neuronal connectivity with distinct adenylate cyclases. Eur J Neurosci

Nie J, Lewis DL (2001) Structural domains of the CB1 cannabinoid receptor that contribute to constitutive activity and G-protein sequestration. J Neurosci 21: 8758-8764

Nikolaev VO, Bunemann M, Hein L, Hannawacker A, Lohse MJ (2004) Novel single chain cAMP sensors for receptor-induced signal propagation. J Biol Chem 279: 37215-37218

Nikolaev VO, Gambaryan S, Lohse MJ (2006) Fluorescent sensors for rapid monitoring of intracellular cGMP. Nat Methods 3: 23-25

Oh DY, Kim K, Kwon HB, Seong JY (2006) Cellular and molecular biology of orphan G protein-coupled receptors. Int Rev Cytol 252: 163-218

Ohki EC, Tilkins ML, Ciccarone VC, Price PJ (2001) Improving the transfection efficiency of post-mitotic neurons. J Neurosci Methods 112: 95-99

Ohno-Shosaku T, Maejima T, Kano M (2001) Endogenous cannabinoids mediate retrograde signals from depolarized postsynaptic neurons to presynaptic terminals. Neuron 29: 729-738

Oliveras, J. L. and Besson, J. M. (1992). Serotonin, stimulation-produced analgesia and spinal mechanisms of antinociception. In Serotonin, CNS receptors and brain function (eds P. B. Bradley, S. Handley, S. J. Cooper, B. J. Key, N. M. Barnes and J. H. Coote), pp. 383-406. Oxford: Pergamon Press

Ollmann MM, Wilson BD, Yang YK, Kerns JA, Chen Y, Gantz I, Barsh GS (1997) Antagonism of central melanocortin receptors in vitro and in vivo by agouti-related protein. Science 278: 135-138

Pacher P, Batkai S, Kunos G (2006) The endocannabinoid system as an emerging target of pharmacotherapy. Pharmacol Rev 58: 389-462

Palmer SL, Thakur GA, Makriyannis A (2002) Cannabinergic ligands. Chem Phys Lipids 121: 3-19

Pan X, Ikeda SR, Lewis DL (1996) Rat brain cannabinoid receptor modulates N-type Ca2+ channels in a neuronal expression system. Mol Pharmacol 49: 707-714

Pan X, Ikeda SR, Lewis DL (1998) SR 141716A acts as an inverse agonist to increase neuronal voltage-dependent Ca2+ currents by reversal of tonic CB1 cannabinoid receptor activity. Mol Pharmacol 54: 1064-1072

259 Pao CS, Benovic JL (2005) Structure/function analysis of alpha2A-adrenergic receptor interaction with G protein-coupled receptor kinase 2. J Biol Chem 280: 11052-11058

Park PS (2012) Ensemble of G protein-coupled receptor active states. Curr Med Chem 19: 1146-1154

Park SH, Das BB, Casagrande F, Tian Y, Nothnagel HJ, Chu M, Kiefer H, Maier K, De Angelis AA, Marassi FM, Opella SJ (2012) Structure of the chemokine receptor CXCR1 in phospholipid bilayers. Nature 491: 779-783

Pasca SP, Panagiotakos G, Dolmetsch RE (2014) Generating Human Neurons In Vitro and Using Them to Understand Neuropsychiatric Disease. Annu Rev Neurosci

Pettit DA, Harrison MP, Olson JM, Spencer RF, Cabral GA (1998) Immunohistochemical localization of the neural cannabinoid receptor in rat brain. J Neurosci Res 51: 391-402

Phillips R, Ursell T, Wiggins P, Sens P (2009) Emerging roles for lipids in shaping membrane-protein function. Nature 459: 379-385

Pierre S, Eschenhagen T, Geisslinger G, Scholich K (2009) Capturing adenylyl cyclases as potential drug targets. Nat Rev Drug Discov 8: 321-335

Piomelli D (2003) The molecular logic of endocannabinoid signalling. Nat Rev Neurosci 4: 873-884

Pippig S, Andexinger S, Lohse MJ (1995) Sequestration and recycling of beta 2-adrenergic receptors permit receptor resensitization. Mol Pharmacol 47: 666-676

Ponsioen B, Zhao J, Riedl J, Zwartkruis F, van der Krogt G, Zaccolo M, Moolenaar WH, Bos JL, Jalink K (2004) Detecting cAMP-induced Epac activation by fluorescence resonance energy transfer: Epac as a novel cAMP indicator. EMBO Rep 5: 1176-1180

Pucadyil TJ, Chattopadhyay A (2004) Cholesterol modulates ligand binding and G-protein coupling to serotonin(1A) receptors from bovine hippocampus. Biochim Biophys Acta 1663: 188-200

Rall TW, Sutherland EW (1958) Formation of a cyclic adenine ribonucleotide by tissue particles. J Biol Chem 232: 1065-1076

Reimann EM, Walsh DA, Krebs EG (1971) Purification and properties of rabbit skeletal muscle adenosine 3',5'-monophosphate-dependent protein kinases. J Biol Chem 246: 1986- 1995

Ribas C, Penela P, Murga C, Salcedo A, Garcia-Hoz C, Jurado-Pueyo M, Aymerich I, Mayor F, Jr. (2007) The G protein-coupled receptor kinase (GRK) interactome: role of GRKs in GPCR regulation and signaling. Biochim Biophys Acta 1768: 913-922

Rinaldi-Carmona M, Barth F, Heaulme M, Shire D, Calandra B, Congy C, Martinez S, Maruani J, Neliat G, Caput D, et al. (1994) SR141716A, a potent and selective antagonist of the brain cannabinoid receptor. FEBS Lett 350: 240-244

260

Rivera JF, Ahmad S, Quick MW, Liman ER, Arnold DB (2003) An evolutionarily conserved dileucine motif in Shal K+ channels mediates dendritic targeting. Nat Neurosci 6: 243-250

Robbe D, Kopf M, Remaury A, Bockaert J, Manzoni OJ (2002) Endogenous cannabinoids mediate long-term synaptic depression in the nucleus accumbens. Proc Natl Acad Sci U S A 99: 8384-8388

Ross RA (2007) Allosterism and cannabinoid CB(1) receptors: the shape of things to come. Trends Pharmacol Sci 28: 567-572

Rueda D, Galve-Roperh I, Haro A, Guzman M (2000) The CB(1) cannabinoid receptor is coupled to the activation of c-Jun N-terminal kinase. Mol Pharmacol 58: 814-820

Rueda D, Navarro B, Martinez-Serrano A, Guzman M, Galve-Roperh I (2002) The endocannabinoid anandamide inhibits neuronal progenitor cell differentiation through attenuation of the Rap1/B-Raf/ERK pathway. J Biol Chem 277: 46645-46650

Samama P, Cotecchia S, Costa T, Lefkowitz RJ (1993) A mutation-induced activated state of the beta 2-adrenergic receptor. Extending the ternary complex model. J Biol Chem 268: 4625- 4636

Sampo B, Kaech S, Kunz S, Banker G (2003) Two distinct mechanisms target membrane proteins to the axonal surface. Neuron 37: 611-624

Santucci V, Storme JJ, Soubrie P, Le Fur G (1996) Arousal-enhancing properties of the CB1 cannabinoid receptor antagonist SR 141716A in rats as assessed by electroencephalographic spectral and sleep-waking cycle analysis. Life Sci 58: PL103-110

Sari Y, Miquel MC, Brisorgueil MJ, Ruiz G, Doucet E, Hamon M, Verge D (1999) Cellular and subcellular localization of 5-hydroxytryptamine1B receptors in the rat central nervous system: immunocytochemical, autoradiographic and lesion studies. Neuroscience 88: 899-915

Saucerman JJ, Zhang J, Martin JC, Peng LX, Stenbit AE, Tsien RY, McCulloch AD (2006) Systems analysis of PKA-mediated phosphorylation gradients in live cardiac myocytes. Proc Natl Acad Sci U S A 103: 12923-12928

Sawzdargo M, Nguyen T, Lee DK, Lynch KR, Cheng R, Heng HH, George SR, O'Dowd BF (1999) Identification and cloning of three novel human G protein-coupled receptor genes GPR52, PsiGPR53 and GPR55: GPR55 is extensively expressed in human brain. Brain Res Mol Brain Res 64: 193-198

Schmid PC, Krebsbach RJ, Perry SR, Dettmer TM, Maasson JL, Schmid HH (1995) Occurrence and postmortem generation of anandamide and other long-chain N- acylethanolamines in mammalian brain. FEBS Lett 375: 117-120

Schoeffter P, Hoyer D (1989) 5-Hydroxytryptamine 5-HT1B and 5-HT1D receptors mediating inhibition of adenylate cyclase activity. Pharmacological comparison with special reference to the effects of yohimbine, rauwolscine and some beta-adrenoceptor antagonists. Naunyn Schmiedebergs Arch Pharmacol 340: 285-292

261

Scott JD, Dessauer CW, Tasken K (2013) Creating order from chaos: cellular regulation by kinase anchoring. Annu Rev Pharmacol Toxicol 53: 187-210

Scott JD, Fischer EH, Takio K, Demaille JG, Krebs EG (1985) Amino acid sequence of the heat-stable inhibitor of the cAMP-dependent protein kinase from rabbit skeletal muscle. Proc Natl Acad Sci U S A 82: 5732-5736

Seifert R, Wenzel-Seifert K (2002) Constitutive activity of G-protein-coupled receptors: cause of disease and common property of wild-type receptors. Naunyn Schmiedebergs Arch Pharmacol 366: 381-416

Shelly M, Lim BK, Cancedda L, Heilshorn SC, Gao H, Poo MM (2010) Local and long-range reciprocal regulation of cAMP and cGMP in axon/dendrite formation. Science 327: 547-552

Shi L, Liapakis G, Xu R, Guarnieri F, Ballesteros JA, Javitch JA (2002) Beta2 adrenergic receptor activation. Modulation of the proline kink in transmembrane 6 by a rotamer toggle switch. J Biol Chem 277: 40989-40996

Sidell N (1982) Retinoic acid-induced growth inhibition and morphologic differentiation of human neuroblastoma cells in vitro. J Natl Cancer Inst 68: 589-596

Sigel E, Baur R, Racz I, Marazzi J, Smart TG, Zimmer A, Gertsch J (2011) The major central endocannabinoid directly acts at GABA(A) receptors. Proc Natl Acad Sci U S A 108: 18150- 18155

Simon AC, Loverdo C, Gaffuri AL, Urbanski M, Ladarre D, Carrel D, Rivals I, Leterrier C, Benichou O, Dournaud P, Szabo B, Voituriez R, Lenkei Z (2013) Activation-dependent plasticity of polarized GPCR distribution on the neuronal surface. J Mol Cell Biol 5: 250-265

Simons K, Ikonen E (1997) Functional rafts in cell membranes. Nature 387: 569-572

Singer SJ, Nicolson GL (1972) The fluid mosaic model of the structure of cell membranes. Science 175: 720-731

Skalhegg BS, Tasken K (2000) Specificity in the cAMP/PKA signaling pathway. Differential expression,regulation, and subcellular localization of subunits of PKA. Front Biosci 5: D678- 693

Song AH, Wang D, Chen G, Li Y, Luo J, Duan S, Poo MM (2009) A selective filter for cytoplasmic transport at the axon initial segment. Cell 136: 1148-1160

Song ZH, Bonner TI (1996) A lysine residue of the cannabinoid receptor is critical for receptor recognition by several agonists but not WIN55212-2. Mol Pharmacol 49: 891-896

Soubias O, Teague WE, Jr., Hines KG, Mitchell DC, Gawrisch K (2010) Contribution of membrane elastic energy to rhodopsin function. Biophys J 99: 817-824

262 Spivak CE, Kim W, Liu QR, Lupica CR, Doyle ME (2012) Blockade of beta-cell K(ATP) channels by the endocannabinoid, 2-arachidonoylglycerol. Biochem Biophys Res Commun 423: 13-18

Steindel F, Lerner R, Haring M, Ruehle S, Marsicano G, Lutz B, Monory K (2013) Neuron- type specific cannabinoid-mediated G protein signalling in mouse hippocampus. J Neurochem 124: 795-807

Stella N, Schweitzer P, Piomelli D (1997) A second endogenous cannabinoid that modulates long-term potentiation. Nature 388: 773-778

Stenmark H (2009) Rab GTPases as coordinators of vesicle traffic. Nat Rev Mol Cell Biol 10: 513-525

Stephenson RP (1956) A modification of receptor theory. Br J Pharmacol Chemother 11: 379-393

Stork PJ, Schmitt JM (2002) Crosstalk between cAMP and MAP kinase signaling in the regulation of cell proliferation. Trends Cell Biol 12: 258-266

Stowell JN, Craig AM (1999) Axon/dendrite targeting of metabotropic glutamate receptors by their cytoplasmic carboxy-terminal domains. Neuron 22: 525-536

Subtil A, Gaidarov I, Kobylarz K, Lampson MA, Keen JH, McGraw TE (1999) Acute cholesterol depletion inhibits clathrin-coated pit budding. Proc Natl Acad Sci U S A 96: 6775- 6780

Sugiura T, Kodaka T, Nakane S, Miyashita T, Kondo S, Suhara Y, Takayama H, Waku K, Seki C, Baba N, Ishima Y (1999) Evidence that the cannabinoid CB1 receptor is a 2- arachidonoylglycerol receptor. Structure-activity relationship of 2-arachidonoylglycerol, ether-linked analogues, and related compounds. J Biol Chem 274: 2794-2801

Sugiura T, Kondo S, Sukagawa A, Nakane S, Shinoda A, Itoh K, Yamashita A, Waku K (1995) 2-Arachidonoylglycerol: a possible endogenous cannabinoid receptor ligand in brain. Biochem Biophys Res Commun 215: 89-97

Sugiura T, Yoshinaga N, Kondo S, Waku K, Ishima Y (2000) Generation of 2- arachidonoylglycerol, an endogenous cannabinoid receptor ligand, in picrotoxinin- administered rat brain. Biochem Biophys Res Commun 271: 654-658

Takahashi K, Yamanaka S (2006) Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126: 663-676

Tanimura A, Yamazaki M, Hashimotodani Y, Uchigashima M, Kawata S, Abe M, Kita Y, Hashimoto K, Shimizu T, Watanabe M, Sakimura K, Kano M (2010) The endocannabinoid 2- arachidonoylglycerol produced by diacylglycerol lipase alpha mediates retrograde suppression of synaptic transmission. Neuron 65: 320-327

263 Tappe-Theodor A, Agarwal N, Katona I, Rubino T, Martini L, Swiercz J, Mackie K, Monyer H, Parolaro D, Whistler J, Kuner T, Kuner R (2007) A molecular basis of analgesic tolerance to cannabinoids. J Neurosci 27: 4165-4177

Taylor SS, Buechler JA, Yonemoto W (1990) cAMP-dependent protein kinase: framework for a diverse family of regulatory enzymes. Annu Rev Biochem 59: 971-1005

Taylor SS, Yang J, Wu J, Haste NM, Radzio-Andzelm E, Anand G (2004) PKA: a portrait of protein kinase dynamics. Biochim Biophys Acta 1697: 259-269

Theodoropoulou M, Stalla GK (2013) Somatostatin receptors: from signaling to clinical practice. Front Neuroendocrinol 34: 228-252

Thiebot MH (1986) Are serotonergic neurons involved in the control of anxiety and in the anxiolytic activity of benzodiazepines? Pharmacol Biochem Behav 24: 1471-1477

Thomas A, Baillie GL, Phillips AM, Razdan RK, Ross RA, Pertwee RG (2007) Cannabidiol displays unexpectedly high potency as an antagonist of CB1 and CB2 receptor agonists in vitro. Br J Pharmacol 150: 613-623

Tian X, Guo J, Yao F, Yang DP, Makriyannis A (2005) The conformation, location, and dynamic properties of the endocannabinoid ligand anandamide in a membrane bilayer. J Biol Chem 280: 29788-29795

Tsao PI, von Zastrow M (2000) Type-specific sorting of G protein-coupled receptors after endocytosis. J Biol Chem 275: 11130-11140

Turu G, Simon A, Gyombolai P, Szidonya L, Bagdy G, Lenkei Z, Hunyady L (2007) The role of diacylglycerol lipase in constitutive and angiotensin AT1 receptor-stimulated cannabinoid CB1 receptor activity. J Biol Chem 282: 7753-7757

Uchigashima M, Narushima M, Fukaya M, Katona I, Kano M, Watanabe M (2007) Subcellular arrangement of molecules for 2-arachidonoyl-glycerol-mediated retrograde signaling and its physiological contribution to synaptic modulation in the striatum. J Neurosci 27: 3663-3676

Ursin R (2002) Serotonin and sleep. Sleep Med Rev 6: 55-69

Van Gaal LF, Rissanen AM, Scheen AJ, Ziegler O, Rossner S (2005) Effects of the cannabinoid-1 receptor blocker rimonabant on weight reduction and cardiovascular risk factors in overweight patients: 1-year experience from the RIO-Europe study. Lancet 365: 1389-1397 van Meer G, de Kroon AI (2011) Lipid map of the mammalian cell. J Cell Sci 124: 5-8

Van Sickle MD, Duncan M, Kingsley PJ, Mouihate A, Urbani P, Mackie K, Stella N, Makriyannis A, Piomelli D, Davison JS, Marnett LJ, Di Marzo V, Pittman QJ, Patel KD, Sharkey KA (2005) Identification and functional characterization of brainstem cannabinoid CB2 receptors. Science 310: 329-332

264 Vasquez C, Lewis DL (1999) The CB1 cannabinoid receptor can sequester G-proteins, making them unavailable to couple to other receptors. J Neurosci 19: 9271-9280

Visel A, Alvarez-Bolado G, Thaller C, Eichele G (2006) Comprehensive analysis of the expression patterns of the adenylate cyclase gene family in the developing and adult mouse brain. J Comp Neurol 496: 684-697

Vitalis T, Laine J, Simon A, Roland A, Leterrier C, Lenkei Z (2008) The type 1 cannabinoid receptor is highly expressed in embryonic cortical projection neurons and negatively regulates neurite growth in vitro. Eur J Neurosci 28: 1705-1718

Vrecl M, Anderson L, Hanyaloglu A, McGregor AM, Groarke AD, Milligan G, Taylor PL, Eidne KA (1998) Agonist-induced endocytosis and recycling of the gonadotropin-releasing hormone receptor: effect of beta-arrestin on internalization kinetics. Mol Endocrinol 12: 1818-1829

Wade DT, Robson P, House H, Makela P, Aram J (2003) A preliminary controlled study to determine whether whole-plant cannabis extracts can improve intractable neurogenic symptoms. Clin Rehabil 17: 21-29

Waksman Y, Olson JM, Carlisle SJ, Cabral GA (1999) The central cannabinoid receptor (CB1) mediates inhibition of nitric oxide production by rat microglial cells. J Pharmacol Exp Ther 288: 1357-1366

Walker JM, Huang SM (2002) Cannabinoid analgesia. Pharmacol Ther 95: 127-135

Walker JM, Huang SM, Strangman NM, Tsou K, Sanudo-Pena MC (1999) Pain modulation by release of the endogenous cannabinoid anandamide. Proc Natl Acad Sci U S A 96: 12198- 12203

Walsh DA, Ashby CD, Gonzalez C, Calkins D, Fischer EH (1971) Krebs EG: Purification and characterization of a protein inhibitor of adenosine 3',5'-monophosphate-dependent protein kinases. J Biol Chem 246: 1977-1985

Walsh DA, Perkins JP, Krebs EG (1968) An adenosine 3',5'-monophosphate-dependant protein kinase from rabbit skeletal muscle. J Biol Chem 243: 3763-3765

Watkins P, Clow A, Glover V, Halket J, Przyborowska A, Sandler M (1990) Isatin, regional distribution in rat brain and tissues. Neurochem Int 17: 321-323

Wei H, Ahn S, Shenoy SK, Karnik SS, Hunyady L, Luttrell LM, Lefkowitz RJ (2003) Independent beta-arrestin 2 and G protein-mediated pathways for angiotensin II activation of extracellular signal-regulated kinases 1 and 2. Proc Natl Acad Sci U S A 100: 10782-10787

Weiss JM, Morgan PH, Lutz MW, Kenakin TP (1996) The cubic ternary complex receptor- occupancy model. III. resurrecting efficacy. J Theor Biol 181: 381-397

Wen W, Harootunian AT, Adams SR, Feramisco J, Tsien RY, Meinkoth JL, Taylor SS (1994) Heat-stable inhibitors of cAMP-dependent protein kinase carry a nuclear export signal. J Biol Chem 269: 32214-32220

265

Wen W, Taylor SS (1994) High affinity binding of the heat-stable protein kinase inhibitor to the catalytic subunit of cAMP-dependent protein kinase is selectively abolished by mutation of Arg133. J Biol Chem 269: 8423-8430

Wettschureck N, Offermanns S (2005) Mammalian G proteins and their cell type specific functions. Physiol Rev 85: 1159-1204

Williams CM, Rogers PJ, Kirkham TC (1998) Hyperphagia in pre-fed rats following oral delta9-THC. Physiol Behav 65: 343-346

Wilson RI, Nicoll RA (2001) Endogenous cannabinoids mediate retrograde signalling at hippocampal synapses. Nature 410: 588-592

Wisco D, Anderson ED, Chang MC, Norden C, Boiko T, Folsch H, Winckler B (2003) Uncovering multiple axonal targeting pathways in hippocampal neurons. J Cell Biol 162: 1317-1328

Wootten D, Christopoulos A, Sexton PM (2013) Emerging paradigms in GPCR allostery: implications for drug discovery. Nat Rev Drug Discov 12: 630-644

Xu W, Yoon SI, Huang P, Wang Y, Chen C, Chong PL, Liu-Chen LY (2006) Localization of the kappa opioid receptor in lipid rafts. J Pharmacol Exp Ther 317: 1295-1306

Yamamoto KK, Gonzalez GA, Biggs WH, 3rd, Montminy MR (1988) Phosphorylation- induced binding and transcriptional efficacy of nuclear factor CREB. Nature 334: 494-498

Yona S, Lin HH, Siu WO, Gordon S, Stacey M (2008) Adhesion-GPCRs: emerging roles for novel receptors. Trends Biochem Sci 33: 491-500

Yoshida T, Fukaya M, Uchigashima M, Miura E, Kamiya H, Kano M, Watanabe M (2006) Localization of diacylglycerol lipase-alpha around postsynaptic spine suggests close proximity between production site of an endocannabinoid, 2-arachidonoyl-glycerol, and presynaptic cannabinoid CB1 receptor. J Neurosci 26: 4740-4751

Yoshida T, Uchigashima M, Yamasaki M, Katona I, Yamazaki M, Sakimura K, Kano M, Yoshioka M, Watanabe M (2011) Unique inhibitory synapse with particularly rich endocannabinoid signaling machinery on pyramidal neurons in basal amygdaloid nucleus. Proc Natl Acad Sci U S A 108: 3059-3064

Zaccolo M, De Giorgi F, Cho CY, Feng L, Knapp T, Negulescu PA, Taylor SS, Tsien RY, Pozzan T (2000) A genetically encoded, fluorescent indicator for cyclic AMP in living cells. Nat Cell Biol 2: 25-29

Zhang J, Hupfeld CJ, Taylor SS, Olefsky JM, Tsien RY (2005) Insulin disrupts beta- adrenergic signalling to protein kinase A in adipocytes. Nature 437: 569-573

Zhang J, Ma Y, Taylor SS, Tsien RY (2001) Genetically encoded reporters of protein kinase A activity reveal impact of substrate tethering. Proc Natl Acad Sci U S A 98: 14997-15002

266

Abstract Polarized neuronal architecture is achieved and maintained mainly through highly controlled targeting of proteins to axons versus to the somatodendritic compartment. Among these proteins, neuronal G protein coupled receptors (GPCRs) are key therapeutic targets. However, their pharmacology is generally studied in non-polarized cell lines, and results obtained in such systems likely do not fully characterize the physiological effects of brain GPCR activation. Therefore, a main research subject of our group is to understand how neuronal polarity influences GPCR pharmacology, by studying one of the most abundant GPCR in the brain: the type-1 cannabinoid receptor (CB1R).

Previous studies of the group suggested that CB1Rs achieve axonal polarization through transcytotic targeting: after their synthesis, these receptors appear on the somatodendritic plasma membrane from where they are removed rapidly by constitutive endocytosis and then targeted to the axonal plasma membrane where they accumulate due to relatively reduced endocytosis rate. At the beginning of my PhD project we directly demonstrated this differential endocytosis and transcytotic transport of CB1Rs by using cultured neurons in microfluidic devices. Moreover, we showed that chronic pharmacological treatments may strongly change neuronal GPCR distribution on the neuronal surface. These results demonstrate that subdomain-dependent steady-state endocytosis, which is pharmacologically controllable, is important for GPCR distribution in neurons.

In a second part, we asked if differential traffic of CB1Rs between axons and dendrites is correlated with differential pharmacology. CB1R is predominantly coupled to Gi/o proteins and is known to inhibit cAMP production. Thus, we developed live Föster Resonance Energy Transfer (FRET) imaging in cultured hippocampal neurons in order to measure basal cAMP/PKA pathway modulation downstream of endogenous CB1Rs in all neuronal compartments: in somata, in dendrites but also in the very thin mature axons. Our results show that CB1R displays differential pharmacology between axon and dendrites. Notably, its activation leads to a stronger decrease of PKA activity in axons compared to dendrites, due to increased number of membrane receptors in this compartment. Moreover, we demonstrate that somatodendritic CB1Rs constitutively inhibit cAMP/PKA pathway, while axonal receptors do not. This difference is due to polarized distribution of DAGLipase, the enzyme that synthesizes the major endocannabinoid 2-arachidonoylglycerol (2-AG). Moreover, blocking DAGL by pharmacological treatment modifies somatodendritic, but not axonal effects of several CB1R agonists, possibly through allosteric action.

In a third part, we asked if the above results may be generalized to other GPCRs. Because the axonal targeting and in vitro pharmacology of 5-HT1B serotonin receptors demonstrate strong similarities with CB1Rs, we studied their neuronal pharmacology by using the previously developed FRET technique. We found similar differential responses to pharmacological treatments between axon and dendrites.

In a fourth part, we investigated the role of the threonine 210 (T210) residue in the constitutive activity of neuronal CB1R. We showed that the hypoactive mutant T210A-CB1R do not constitutively recruit signaling pathways even in somatodendritic compartment, where 2-AG is present. This result demonstrates that T210 is necessary for constitutive CB1R activation by 2-AG.

Finally, previous results of our group demonstrated the involvement of CB1R in neuronal development. Notably, CB1R activation was shown to have an overall inhibitory effect on the development of polarized neuronal morphology. We established a bibliographic review on this subject. The published literature data suggest that not only neuronal polarization influences both CB1R traffic and pharmacology but CB1Rs also contribute to the achievement of neuronal polarization.

To conclude, the studies completed during my PhD thesis suggest that neuronal polarization has a strong effect on the traffic and pharmacology of neuronal GPCRs. The highly contrasted difference in sub-neuronal signaling responses warrants caution in extrapolating pharmacological profiles, which are typically obtained in non-polarized cells, to predict in vivo responses of axonal (i.e. presynaptic) GPCRs. Therefore, the in situ pharmacological approach presented here may be also useful for a better understanding of the physiology of other neuronal GPCRs.