Sorana Ciura. Department of Neurology and Neurosurgery. McGill University, Montreal. August A thesis submitted to

Transcription

1 MECHANISMS OF OSMOTIC AND MECHANICAL TRANSDUCTION IN THE ORGANUM VASCULOSUM LAMINA TERMINALIS By Sorana Ciura Department of Neurology and Neurosurgery McGill University, Montreal August 2010 A thesis submitted to The Faculty of Graduate Studies and Research, McGill University In partial fulfillment of the requirements for the degree of Doctor of Philosophy Sorana Ciura, 2010

2 TABLE OF CONTENTS Abstract/Résumé... X Acknowledgements... XIV Publications arising from this work... XV Contibutions and originality... XVII Abbreviations and symbols... XX CHAPTER ONE: INTRODUCTION Osmoregulation The osmotic homeostatic machinery Behavioral homeostatic response: water intake Endocrine homeostatic reponse: the neurohypophysis The supraoptic nucleus The paraventricular nucleus Neurohypophysial hormone function Neurohypophysis hormone synthesis and release Osmosensors Peripheral osmosensors Central osmosensors The OVLT Morphology and function II

3 The role of the OVLT in osmosensing: lines of evidence Lesion studies Stimulation studies OVLT neurons are activated by hypertonicity and inhibited by hypotonicity Neural pathways involved in the osmotic control of the effector mechanisms mediated by the OVLT The control of thirst The control of MNC activity Factors mediating osmosensory transduction in osmosensitive neurons Osmometry and mechanosensory transduction TRPV channels as candidate molecular osmosensors a. TRPV b. TRPV Gliotransmission Integration of different sensorial modalities Hypotheses addressed by this thesis 41 III

4 CHAPTER TWO: TRPV1 IS REQUIRED FOR INTRINSIC OSMORECEPTION IN OVLT NEURONS AND FOR NORMAL THIRST RESPONSES TO SYSTEMIC HYPEROSMOLALITY 2.1. Overview Materials and Methods Extracellular recordings from hypothalamic explants Whole cell recordings from acutely Isolated OVLT neurons Volume measurements Immunohistochemistry Water intake measurements in mice Statistics Results TRPV1 is expressed in OVLT neurons Osmosensitivity of OVLT neurons in hypothalamic.53 explants OVLT neurons are intrinsically osmosensitive Osmosensory transduction involves hypertonicity-activated cation channels OVLT neurons from Trpv1 / mice fail to transduce osmotic stimuli Osmosensory transduction is blocked by ruthenium red TRPV1 contributes to osmotic thirst in vivo.58 IV

5 2.4. Discussion The OVLT is a primary homeostatic osmosensory nucleus OVLT neurons are intrinsically osmosensitive Ionic basis for osmoreception in OVLT neurons Expression of the Trpv1 gene is required for osmoreception in OVLT neurons Impaired osmotic thirst in Trpv1 / mice...64 CHAPTER THREE: OSOMOSENSORY TRANSDUCTION IN OVLT NEURONS INVOLVES A TRPV1-, BUT NOT TRPV4-, DEPENDENT MECHANOSENSORY EVENT 3.1. Overview Materials and Methods Explant recordings Isolated cell recordings Cell-attached single channel recordings Volume measurements Results OVLT neurons are specialized osmosensors Hyperosmolality activates calcium-permeable cation channels Hyperosmotic stimuli activate a 32 ps cation channel V

6 Hypertonicity causes osmometric volume changes in OVLT neurons Hypertonicity-sensing by OVLT neurons is a mechanical process Osmotic and mechanical transduction are quantitatively equivalent Spike encoding is equivalent during osmotic and mechanical transduction Osmo-mechanical transduction requires expression of Trpv1 and not Trpv Hypertonicity sensing in hypothalamic explants requires Trpv1 not Trpv Discussion OVLT neurons display osmometry Hypertonicity-sensing involves mechanotransduction Osmosensory transduction involves a 32 ps calciumpermeable cation channel Osmosensory channels require expression of Trpv1 but not Trpv Mechanism of osmotic gating of the transduction channel Trpv4 does not contribute to hypertonicity sensing by OVLT neurons...95 VI

7 CHAPTER FOUR: TRPV1 GENE REQUIRED FOR THERMOSENSORY TRANSDUCTION AND ANTICIPATORY SECRETION FROM VASOPRESSIN NEURONS DURING HYPERTHERMIA 4.1.Overview Materials and Methods Preparation and perfusion of isolated neurons Whole-cell recording Immunocytochemistry Induction of hyperthermia without osmotic stress in vivo Serum VP measurements Statistical analysis Results VP neurons display a heat-activated current VP neurons are thermosensitive VP neurons express heat-activated calcium-permeable cation channels Trpv1 gene expression is required for thermosensory transduction in VP neurons Contribution of other RuR-sensitive channels to thermosensitivity Trpv1 gene contributes to anticipatory VP release during hyperthermia Discussion VII

8 CHAPTER FIVE: GLIOTRANSMISSION RESCUES HYPOTONICITY SENSING IN MICE LACKING Trpv Overview Results and Discussion Materials and Methods Preparation of acute hypothalamic explants Extracellular recording and analysis Whole-cell recordings from acutely isolated neurons Sodium Dodecyl Sulfate (SDS) Polyacrylamide Electrophoresis (SDS/PAGE) and Western Blotting Whole cell recordings from mouse astrocytes Statistics CHAPTER SIX: GENERAL DISCUSSION OVLT neurons are intrinsic osmosensors An osmosensory TRPV1 variant: a multimodal channel Glia participate in sensing hypotonic stimulation in the absence of TRPV TRPV4 does not mediate the intrinsic osmosensitivity of OVLT neurons. Is there a role for TRPV4 in osmoregulation? Osmosensory neurons can integrate osmotic and thermic sensing due to the presence of the TRPV1 channel VIII

9 6.6. Remaining Questions/ Future directions Projections to the thalamus/prefrontal cortex Characterization of markers expressed by osmosensory OVLT neurons Integration of other modalities: vasopressin as a thirst stimulus Concluding remarks CHAPTER SEVEN: REFERENCES APPENDIX IX

10 ABSTRACT Mammals maintain a constant internal osmotic environment due to the existence of specialized osmosensory neurons which detect fluctuations in body fluid osmolality and can trigger homeostatic regulatory mechanisms to correct such deviations. The organum vasculosum of the lamina terminalis (OVLT) is the main osmosensory area of the brain. Despite its importance in whole body homeostasis, little is known about the electrophysiological properties of the OVLT. In this thesis we use murine OVLT neurons in isolation or in brain explants to study their osmosensory transduction properties. Hypertonic stimulation resulted in an increase in the cation conductance of the cells, the generation of an inward current and depolarization which triggered increased firing of action potentials. Two members of the TRPV family of channels, TRPV1 and TRPV4, were tested for their contribution to osmosensory transduction. We found that TRPV1, but not TRPV4, forms an osmotically gated channel in OVLT neurons. Absence of TRPV1 function, either through genetic engineering or pharmacological blockade, prevented the response of OVLT neurons to osmotic stimulation. In contrast, OVLT neurons obtained from Trpv4 -/- mice showed normal osmosensitive responses. The transduction of hypertonic stimuli was found to be entirely mediated by the mechanical process associated with cellular shrinkage during the presentation of the stimulus. Previous studies have shown that osmosensory neurons are also able to detect other modes of stimulation, such as temperature. Here we show that TRPV1 enables osmosensory neurons in the supraoptic nucleus to X

11 respond to increases in external temperature. Another contributing factor to osmosensory detection are the glia surrounding osmosensory neurons which can release neuromodulatory factors during hypotonic stimulation. In this thesis we show that astrocytes in the OVLT can release glycine receptor agonists which cause hyperpolarization during a hypotonic stimulus, but this effect is secondary to the intrinsic ability of these neurons to detect such stimuli. In conclusion, OVLT neurons are intrinsic osmosensors equipped with stretch receptors encoded by Trpv1 functioning in the central control of body fluid homeostasis. XI

12 RÉSUMÉ Les mammifères maintiennent l'osmolalité plasmatique près d'une valeur constante grâce à l`existence de neurones osmosensoriels capables de détecter les fluctuations d`osmolalité fluidique et d'initier les mécanismes homéostatiques nécessaires pour corriger ces déviations. L`Organum vasculosum lamina terminalis (OVLT) représente un site primaire pour l'osmoréception dans le système nerveux central. Même si le rôle physiologique de cette structure est bien établi, presque rien n'est connu au sujet des propriétés membranaires des neurones de l'ovlt qui soustendent la détection et le codage électrique des perturbations osmotiques. Nous avons utilisé une approche électrophysiologique pour étudier les neurones de l`ovlt de souris suite à l'isolation in vitro ainsi que dans des tranches d'hypothalamus. La stimulation hypertonique induit une élévation de la conductance aux cations, causant ainsi la génération d`un courant entrant, la dépolarisation du potentiel de repos et une augmentation de la fréquence de décharge des potentiels d`action. Nos expériences démontrent que ces réponses reflètent un processus entièrement mécanique qui est initié par la dépression du volume cellulaire. Deux membres de la famille des canaux ioniques Transient Receptor Potential Vanilloid (TRPV) pourraient être impliqués: TRPV1 et TRPV4. Nos données montrent que seul le canal TRPV1 est requis pour supporter l'osmosensibilité des neurones de l`ovlt. En effet, des études pharmacologiques avec des inhibiteurs du TRPV1 ou des mesures sur des souris «knock-out» (Trpv1 -/ - ) démontrent que le blocage ou l'absence XII

13 d'expression du canal TRPV1 élimine l'osmosensibilité des neurones, tandis que ceux des souris Trpv4 -/- détectent les changements d`osmolalité de façon normale. Des études sur les souris Trpv1 -/ - ont aussi révélé un rôle secondaire pour les astrocytes de l'ovlt dans la détection des conditions hypotoniques. Nos résultats montrent que ces cellules libèrent un agoniste des récepteurs à la glycine qui contribue à la réduction de l'activité électrique en condition hypotonique. Finalement, des études antécédentes avaient indiqué que les neurones osmorécepteurs peuvent aussi détecter l'augmentation de la température corporelle. Nos résultats démontrent que le canal TRPV1 confère aux neurones hypothalamiques la capacité intrinsèque de répondre au réchauffement du milieu extérieur. En conclusion, les neurones de l`ovlt sont des osmorécepteurs intrinsèques dont l'activité électrique est modulée par un canal cationique mécanosensible dont l'expression est gérée par le gène Trpv1. XIII

14 ACKNOWLEDGEMENTS I would like to thank Dr. Charles Bourque for being such a thoroughly wonderful supervisor. When I arrived in the lab I had little experience with electrophysiology. Through these years in his lab, Charles has always been available to help me understand this black magic science, by patching, clamping, recording, bourquing, analyzing data, as well as taking down and putting together several rigs when necessary. I also want to thank members of my committee for their scientific input, Drs. David Ragsdale, Anne McKinney and Don Van Meyel. I am also grateful for having such wonderful lab mates, all the past and present members of the Bourque lab: Zizheng, Eric, Tev, Reza, Pierce, Katrina, Jes, Masha, Cristian and Ara. Their help and friendship was invaluable to my work. I want to thank my parents, my brother Mihai, and his wife Ji-Eun, as well as Claudia and Alex, for their love and for being very understanding with my role as an eternal student throughout these years. Also, I am grateful to Edor for more than I can express in words, and his wonderful family (Kujtim, Irma and Sokol). My friends Madalina, Carmen, Andrei, Marius, Claudiu, and all my family in Romania have been incredibly supportive throughout all my studies. Also, want to acknowledge the friends who have made my life in Montreal so memorable: Ernesto, Jodat, Emily, Ara, Sorin, Madi, Ilia, Amelia, Klodi, Adrian, Max, Sara, Jarrod, Emma, Andi, etc. Finally, I want to acknowledge the support of the Canda Graduate Scholarhips program and CIHR for project funding. XIV

15 PUBLICATIONS ARISING FROM THIS WORK Peer-reviewed articles: Ciura S, Bourque CW. (2010) Gliotransmission rescues hypotonicity sensing in mice lacking TRPV1. Manuscript in preparation Ciura S, Bourque CW. (2010) Physiological steps involved in osmodetection in the OVLT. Manuscript in preparation Ciura S, Bourque CW. (2010) Vasopressin response in osmosensitive neurons in the OVLT. Manuscript in preparation Stachniak, T.J. Sudbury, J.R., Trudel, E., Choe, K.Y., Ciura, S. and Bourque, C.W. (2010). Osmoregulatory Circuits in Slices and En-Bloc Preparations of Rodent Hypothalamus. In: Isolated brain circuits. Ballanyi K.(Ed.). Humana-Springer. (In press). Sharif-Naeini R, Ciura S, Stachniak TJ, Trudel E, Bourque CW. (2008) Neurophysiology of supraoptic neurons in C57/BL mice studied in three acute in vitro preparations. Prog Brain Res.170: XV

16 Sharif-Naeini R, Ciura S, Bourque CW. (2008) TRPV1 gene required for thermosensory transduction and anticipatory secretion from vasopressin neurons during hyperthermia. Neuron. 58(2): Ciura S, Bourque CW. (2006) Transient receptor potential vanilloid 1 is required for intrinsic osmoreception in organum vasculosum lamina terminalis and for normal thirst responses to systemic hyperosmolality. J Neurosci. 26(35): Reviews: Sharif-Naeini R, Ciura S, Zhang Z, Bourque CW. (2008) Contribution of TRPV channels to osmosensory transduction, thirst, and vasopressin release. Kidney Int. 73(7): Review. Bourque CW, Ciura S, Trudel E, Stachniak TJ, Sharif-Naeini R. (2007) Neurophysiological characterization of mammalian osmosensitive neurones. Exp Physiol. 92(3): Review. Sharif Naeini R, Ciura S, Bourque CW. (2006) TRPVs: ion channels that make you thirsty! Med Sci (Paris). 22(12): XVI

17 CONTRIBUTION TO ORIGINAL KNOWLEDGE This thesis was written in the manuscript based format according to guidelines by McGill University. An overview of the original contributions in each chapter is given below. I have performed all the experiments reported in these chapters with the exception of Chapter 4, as detailed below. Chapter 1: INTRODUCTION, provides a review of the scientific literature on the osmoregulatory homeostatic mechanisms. The role of the OVLT in controlling these mechanisms is highlighted. Also reviews the evidence implicating members of the Transient receptor potential vanilloid (TRPV) family of channels in osmosensory transduction. Chapter 2: TRPV1 IS REQUIRED FOR INTRINSIC OSMORECEPTION IN OVLT NEURONS AND FOR NORMAL THIRST RESPONSES TO SYSTEMIC HYPEROSMOLALITY. Original observations in this chapter include the demonstration that OVLT neurons are intrinsically osmosensitive and this property requires a function Trpv1 gene. It also shows the characteristic responses of murine OVLT neurons when challenged with hypertonic stimulation and establishes that 60% of these neurons are osmosensitive, both in situ and in isolation. Chapter 3: OSOMOSENSORY TRANSDUCTION IN OVLT NEURONS INVOLVES A TRPV1-, BUT NOT TRPV4-, DEPENDENT XVII

18 MECHANOSENSORY EVENT. In this chapter we describe in detail the electrophysiological steps required for osmosensory transduction in OVLT neurons, showing that osmosensory transduction involves the activation of a non-selective cation conductance. Single channel recordings reported in this channel show the presence of an osmotically-modulated channel in OVLT neurons. Furthermore, we demonstrate that osmosensing is a mechanosensory process. Also, this chapter investigates for the first time the role of TRPV4 in the transduction of osmotic stimulation in the OVLT. The manuscript containing these results is under preparation for submission: Ciura S, Bourque CW. (2010) Physiological steps involved in osmodetection in the OVLT. Manuscript in preparation Chapter 4: Trpv1 GENE REQUIRED FOR THERMOSENSORY TRANSDUCTION AND ANTICIPATORY SECRETION FROM VASOPRESSIN NEURONS DURING HYPERTHERMIA. The data presented in this chapter have been obtained by Reza Sharif-Naeini (Fig. 4.1, 4.4 and partially Fig 4.2 and Supp Fig. 4.1) and me (Fig 4.3 and partially Fig. 4.2 and Supp Fig. 4.1). The original contribution of this chapter consists of the demonstration that osmosensory vasopressinreleasing neurons of the supraoptic nucleus are intrinsically thermosensitive, and heat induces a ruthenium red-sensitive current in these neurons. Furthermore, Trpv1 contributes to this intrinsic thermosensitivity. These data have been published in the following article: XVIII

19 Sharif-Naeini R, Ciura S, Bourque CW. (2008) TRPV1 gene required for thermosensory transduction and anticipatory secretion from vasopressin neurons during hyperthermia. Neuron. 58(2): Chapter 5: GLIOTRANSMISSION RESCUES HYPOTONICITY SENSING IN MICE LACKING Trpv1. The original contributions of this chapter include the demonstration that isolated OVLT neurons can respond to hypotonic stimulation, and Trpv1 is required for this intrinsic response; in situ, OVLT neurons from Trpv1 -/- mice respond equally well to hypotonicity as their WT controls; astrocytes in the OVLT have the capacity to contribute to osmosensory transduction in hypotonic conditions. This chapter was written as a manuscript for a short communication: Ciura S, Bourque CW. (2010) Gliotransmission rescues hypotonicity sensing in mice lacking TRPV1. Chapter 6: GENERAL DISCUSSION represents a general discussion of this research in light of current literature, open questions as well as presents several avenues for future directions. XIX

20 LIST OF ABBREVIATIONS -/-: Knock-out Ω: Ohm μm: μm: μl: Micrometer (10-6 meter) Micromolar (10-6 molar) Microliter (10-6 liter) A: Ampere ACSF: ACC: AMPA: AMPAR: BBB: CGRP: CNS: CSA: CVO: DCPIB: Artificial cerebro-spinal fluid anterior cingulated cortex α-amino-3-hydroxy-5-methyl-4-isoxalone propionic acid AMPA receptor Blood brain barrier Calcitonin gene related peptide Central nervous system Cross-sectional area Circumventricular organ 4-[(2-Butyl-6,7-dichloro-2-cyclopentyl-2,3-dihydro-1-ox o-1hinden-5-yl)oxy]butanoic acid Erev: FC: Equilibrium potential Fluorocitrate G: Conductance GABA: γ-amino butyric acid XX

21 Gd 3+ : GlyR: H 2 O: H 2 Od: HNA: Hz: Gadolinium Glycine receptor Water Distilled water Hypothalamo-neurohypophyseal axis Hertz I: Current INS: Insular cortex K: Potassium Kyn: LHRH: Kynurenic acid Luteinizing hormone releasing hormone M: Molar ma: Mg: min: ml: mm: mm: MNC: mosm: MRI: ms: mv: MNC: Milliampere Magnesium Minute Millilitre Millimolar Millimetre Magnocellular Neurosecretory Cells Milliosmol/kg Magnetic resona Millisecond Millivolt Magnocellular neurosecretory cell XXI

22 MnPO: NMDA: NMDAR: Na: NaCl: NTS OVLT: ON: OT: pa: PFA: Median Preoptic Nucleus of the Hypothalamus N-methyl-D-aspartate NMDA receptor Sodium Sodium chloride Nucleus of the Tractus Solitaris Organum vasculosum laminae terminalis Osmosensory neurons Oxytocin Picoampere Paraformaldehyde ph: -log 10 [H + ] PBS: PNZ: PVN: RVD: RR: Phosphate buffered saline Perinuclear zone Paraventricular nucleus Regulator volume decrease Ruthenium red s: Second s.e.m.: SFO: SIC: SON: TRP: Standard error of mean Subfornical Organ Stretch inactivated channel Supraoptic nucleus Transient receptor potential XXII

23 TRP: TRPC: TRPM: TRPL: TRPP: TRPV: Transient receptor potential ankyrin Transient receptor potential canonical Transient receptor potential melastatin Transient receptor potential mucolipin Transient receptor potential polycystin Transient receptor potential vanilloid TRPV1: Transient receptor potential vanilloid type 1 TTX: Tetrodotoxin V: Volt VP: VRAC: WT: Vasopressin Volume regulated anion channels Wild type XXIII

24 CHAPTER ONE: INTRODUCTION 1.1 Osmoregulation Historically, a crucial question in animal physiology has been to understand how organisms maintain an optimal water and solute concentration of the internal environment. In fact, one of the basic physical parameters affecting the biological processes of a multicellular organism is osmolality, or the concentration of solutes in the extracellular and intracellular fluid (Strange, 2004). Cellular membranes are semi-permeable, that is, they allow water and certain osmolytes (e.g. urea) to cross freely between the internal and the external compartments, while being impermeable to most ions and macromolecules (Brahm and Wieth, 1977, Yancey et al., 1982). This means that external osmotic pressure can force water flux in or out of the cell, resulting in cellular shrinking or swelling as well as in a change of the internal ionic concentration (Steenbergen et al., 1985). These perturbations affect the biological processes of the cell, causing in extreme cases physical trauma to organs (Ayus et al., 2000, Verbalis, 2003, Machino and Yoshizawa, 2006). It is therefore imperative for the survival of the organism that the internal osmotic environment should be strictly maintained around a constant value that is optimal for the proper cellular functioning (Verney, 1947, Stricker et al., 1987, Bourque et al., 1994, Denton et al., 1996, Bourque, 1998, Voisin and Bourque, 2002, Bourque, 2008). We can speculate that this constraint 1

25 represented one of the primary homeostatic requirements during the evolution of multicellular organisms, since most vertebrates tested to date have a very well conserved internal osmolality set-point of ~300 mosm (Denton, 1948, Bourque, 2008) (Figure 1.1). As reviewed by Denton, it was previously postulated that this narrow osmotic range of internal conditions reflects the physiochemical composition of the ancient ocean where cellular organisms have evolved (Denton, 1948). Controlling this osmotic parameter represents a significant challenge, given the large diversity of environmental conditions where organisms dwell today (dry land, fresh water, ocean). Few organisms have developed strategies to escape this requirement and adopt the osmolality of their environment, therefore being termed osmoconformers (Dietz et al., 1995, McAllen et al., 2002). One strategy employed by these animals (e.g. sharks, Australian desert frog) is to match the external osmolality by allowing higher concentrations of urea into the bloodstream (Schmidt-Nielsen, 1960, Withers and Guppy, 1996). In biophysical terms, osmolality translates to the osmotic pressure across the cellular membrane. Therefore, solutes to which the membrane is permeable, such as urea, do not contribute to the effective osmolality impinging on the cell (Pierce et al., 1984, Pasantes-Morales et al., 2000, Yancey et al., 2002, Lang, 2007). Other osmoconformers (mussels, clams) have evolved mechanisms of maintaining a constant cellular volume by actively secreting osmolytes to regulate intracellular osmolality (Hosoi et al., 2005). 2

26 Mammals as a group are osmoregulators, that is, they actively oppose external changes in osmolality to maintain a constant internal environment as proposed by Claude Bernard in 1878 and defined by Cannon (Cannon and Querido, 1924, Bourque, 2008). Negotiating water and solute intake, evaporative water loss, excretion of waste and a variable external osmotic pressure requires a finely tuned homeostatic apparatus of sensorial and operational functions linked together by specialized neural networks. I will briefly introduce the main homeostatic apparatus responsible for the maintenance of osmotic set-point and then focus on the role that the primary CNS osmosensor plays in regulating this apparatus The osmotic homeostatic machinery Osmoregulation is mediated by a complex network of behavioral and endocrine processes [reviewed in (Andersson, 1971, McKinley et al., 1974, McKinley et al., 1978, Thrasher et al., 1980a, Thrasher et al., 1981, Thrasher et al., 1982a, McKinley et al., 1992c, Denton et al., 1996, Verbalis, 2003, Bourque, 2008) and summarized in Figure 1.2]. For example, increases in body fluid osmolality as small as 1-3% or ~3-9 mosm, lead to robust stimulation of water ingestion (Robertson, 1983, Thompson et al., 1986) and release of the anti-diuretic hormone, vasopressin (VP) which acts on the kidneys collecting duct to increase their permeability to water and thus concentrate urine (Dunn et al., 1973, Robertson and Athar, 1976, Thrasher et 3

27 al., 1980a, de Wardener and Clarkson, 1985). Sodium excretion, or natriuresis, is stimulated through the release of oxytocin (OT) (Dorn et al., 1969, Chiu and Sawyer, 1974, Bealer et al., 1983a, Bealer et al., 1983b, McKinley et al., 1992c, Huang et al., 1995) (this effect has been established in rodents, but not in humans, (Rasmussen et al., 2003)). This is often accompanied by a reduction in sodium preference/appetite (Weisinger et al., 1979, Stricker et al., 1987, Weisinger et al., 1987a, Weisinger et al., 1987b, Blackburn et al., 1993, 1995, Johnson and Thunhorst, 1997). Conversely, low solute concentration in the blood stream triggers sodium appetite and suppression of thirst and VP release (Dunn et al., 1973, Claybaugh et al., 2000, Maresh et al., 2001). For the scope of this thesis I will address the pathways of water homeostasis, which can largely be summarized by the induction or suppression of water intake (thirst) and the mechanisms of water conservation through the release of neurohypophysial hormones Behavioral homeostatic response: water intake Water is lost from the body through a multitude of ways. There is an obligate amount required for the elimination of waste, a constant evaporative loss through breathing, sweating, as well as panting and grooming in certain animals, and in medical conditions leading to hypovolemia water can be lost in large quantities in a short amount of time (e.g. bleeding) (Verbalis, 2003). Therefore, a number of conditions regulate thirst in addition to increases in 4

28 blood osmolality: high temperature, low blood pressure, activation of peripheral sensors on the GI tract (e.g. in response to the sensation of dry mouth or ingestion of dry food) (Stricker and Hoffmann, 2007), as well as subjective aspects such as fluid palatability and social context (Verbalis, 2003). The contribution of osmotically regulated thirst to everyday water intake is therefore difficult to assess. Under experimental conditions however, osmotic thirst can be triggered by injecting hypertonic solutions into the blood stream (McKinley et al., 1978, Thrasher et al., 1980a, Zerbe and Robertson, 1983, Egan et al., 2003). Functional imaging in healthy human subjects has revealed the brain areas where activation is correlated with the sensation of thirst under this experimental paradigm (Egan et al., 2003, Farrell et al., 2008). Parts of the anterior cingulate cortex (ACC) and the insular cortex (INS) showed correlates of activity which were proportional to the strength of the induced thirst sensation (Egan et al., 2003, Sewards and Sewards, 2003, Farrell et al., 2008). Satiation of thirst caused by allowing the subjects to drink resulted in a rapid deactivation of these areas. Studies have suggested that the INS is required for the conscious sensations correlated to specific physiological needs (e.g. hunger, thirst) and ACC activation provides the motivational drive for the appropriate behavioral response (Craig, 2002, 2003). In accordance with this hypothesis, electrical stimulation of specific parts of the ACC induced drinking in water satiated monkeys (Robinson and Mishkin, 1968) and conditions that increase thirst were found to induce expression of the early response gene c-fos in the INS (Pastuskovas et al., 5

29 2003). However, there is evidence that the subcortical regions of the periaqueductal gray can sustain some level of water intake since decorticate animals containing this area have a residual thirst response (Sorenson and Ellison, 1970). While thirst is a homeostatic response to increased osmolality, numerous physiological conditions that can cause an osmotic shift can be powerful thirst triggers. Most notably, increases in body temperature can initiate thirst before any increase in body fluid osmolality due to evaporative water loss takes place (Forsling et al., 1976, Takamata et al., 1995, Niimi et al., 1997, Harikai et al., 2003, Saat et al., 2005). While these responses could reflect a conditioned response learned during development, or the convergence of osmotic and thirst sensors on the same cortical pathway, there remains the interesting possibility that such an integration of multiple stimuli occurs on the same sensory neuron. This possibility will be the subject of Chapter 4 of this thesis Endocrine homeostatic response: the neurohypophysis The hypothalamo-neurohypophysial (HNA) hormone release system was one of the first peptidergic systems to be characterized in mammals and it is essential for solute homeostasis. It involves the production and release of VP and OT by the HNA axis. The HNA consists of the magnocellular neurosecretory cells (MNC) and the axons of these neurons that coalesce to form the neural stalk and project into the posterior pituitary, or 6

30 neurohypophysis, where they branch out to release the hormones directly into the neurohypophysial capillaries (Bargmann, 1951, Sherlock et al., 1975). MNCs are mainly located in the supraoptic (SON) and paraventricular nuclei (PVN) of the hypothalamus (Poulain and Wakerley, 1982), in addition to several scattered small groups of MNCs located in between these nuclei (Riva et al., 1999, Simmons and Swanson, 2009) The supraoptic nucleus In rats and mice, the SON is located on the ventral part of the brain, adjacent to the dorso-lateral side of each optic tract. It is composed of a homogeneous population of MNCs characterized, as the name implies, by their large cell bodies (~25µm in diameter in rats) (Randle et al., 1986). These MNCs secrete either VP or OT, and in rare instances both (in less than 5% of the cells) (Swaab et al., 1975, Vandesande and Dierickx, 1975). The populations of VP and OT secreting neurons are segregated across the nucleus. In rats, the VP cells occupy the ventral aspect of the nucleus, with a higher density in the posterior half of the nucleus, while OT cells on the other hand are more prominent in the antero-dorsal part of the nucleus (Swaab et al., 1975, Vandesande and Dierickx, 1975). In mice, the distribution is largely similar, with a higher proportion of VP cells than in rats (75% vs 60%) (Sharif- Naeini et al., 2008b). Due to its accessible location and its homogeneous 7

31 population of neurons, the SON has been used extensively in the study of osmoregulation (Bourque, 2008, Sharif-Naeini et al., 2008b) The paraventricular nucleus The PVN neurons are found in the dorsal part of the hypothalamus, flanking the top of the third ventricle. Unlike the SON, the PVN is more heterogeneous in its neuronal composition (Armstrong and Hatton, 1980, Armstrong et al., 1980, Swanson and Sawchenko, 1983). In addition to magnocellular VP and OT secreting cells, which project to the posterior pituitary (Alonso and Assenmacher, 1981), the PVN contains at least two other morphologically and functionally distinct types of cells. The parvocellular neurosecretory neurons project to the median eminence portal system and synthesize hypophysiotropic hormones which regulate the release of adenohypophysial hormones (Swanson et al., 1980, Pelletier, 1991). Another group of PVN neurons are pre-autonomic neurons that project to the brain stem and spinal cord and modulate the activity of preganglionic sympathetic and parasympathetic neurons (Kiss, 1988) Neurohypophysial hormone function Early on it was recognized that extracts from the pituitary have very pertinent physiological effects, including vasoconstriction (Oliver and Schafer, 8

32 1895), milk ejection (Ott and Scott, 1911) and antidiuresis (Von den Velden, 1913). The bioactive components were purified as two different humoral products: OT (Taylor et al., 1953) and VP (Taylor et al., 1953, Katsoyannis and Du Vigneaud, 1959). Vasopressin is a nine amino-acid peptide which was shown to be present in all mammals as either arginine-vasopressin or lysine-vasopressin and it functions as the main determinant of free water excretion and urine flow (Frank and Landgraf, 2008). Circulating vasopressin promotes antidiuresis by binding to V2 receptors in the collecting ducts of the kidneys, resulting in the insertion of aquaporin-2 water channels into the luminal membrane (Nielsen et al., 1995) and the subsequent increase in the permeability and the reabsorption of water out of the ducts [reviewed in (Jard, 1983)]. Oxytocin is a similar peptide, with only two amino acids difference from VP in most mammals, including humans and rats (Russell et al., 1980, de Bree, 2000). The most conclusively described function of oxytocin is its role in inducing uterine contractions during birth and milk ejection (Poulain and Wakerley, 1982). The role of oxytocin is not altogether clarified in relation to fluid homeostasis. In rodents, oxytocin can stimulate sodium excretion (Verbalis et al., 1991, Huang et al., 1995) and inhibit sodium appetite (Stricker et al., 1987), but this role has not been confirmed in humans (Rasmussen et al., 2003). The role of oxytocin in other biological functions, such as social bonding and pain perception, form currently a dynamic area of research (Minakata, 2010). 9

33 Neurohypophysial hormone synthesis and release VP and OT are produced in MNCs from high molecular-weight preprohormones which are cleaved to the final peptide prior to release (Russell et al., 1980, Alstein et al., 1988). The maturation of the hormones takes place in Golgi-derived neurosecretory vesicles which are used for packaging, transporting and finally releasing VP and OT into the vasculature of the posterior pituitary (Russell et al., 1980). The release of neurohypophysial hormones from the axonal terminals is controlled by the frequency of action potentials generated in the MNC soma (Poulain and Wakerley, 1982). Electrical activity in the neurosecretory terminals causes depolarization and influx of Ca2+ through voltage-gated Ca2+ channels resulting in the fusion of vesicles to the plasma membrane and release of the neurohypophysial hormones (Douglas and Poisner, 1964, Brethes et al., 1987). The relationship between action potential generation and neurohypophysis hormone release has been well established with the help of experiments in isolated neurohypophysis. Increasing the firing in the MNC axons results in frequencydependent facilitation of hormone release, i.e. a progressive increase in the amount of hormone released with each spike (Dreifuss et al., 1971, Dutton and Dyball, 1979, Bicknell and Leng, 1981, Bicknell, 1988). A number of different conditions can stimulate hormone release from the SON, including temperature (Forsling et al., 1976, Takamata et al., 1995, Niimi et al., 1997, 10

34 Harikai et al., 2003, Saat et al., 2005), pain (Day and Sibbald, 1990, Suzuki et al., 2009) and hypotension (Howe et al., 2004) [reviewed in (Doris, 1984)], but the best studied and arguably the most finely tuned homeostatic response is the osmotic regulation of VP and OT release (Verbalis, 2003, Bourque, 2008). Three major factors control the frequency and pattern of firing in the MNCs in response to an osmotic challenge: the synaptic inputs from osmosensitive areas (Miselis et al., 1979, Richard and Bourque, 1995), their own intrinsic osmosensitivity (Mason, 1980, Oliet and Bourque, 1992, 1993a) and the surrounding glia which release neuromodulatory substances (Hussy et al., 1997) from processes that closely surround neuronal cell bodies (Bourque, 1998). All the aforementioned factors function in synchrony to fine-tune the release of VP and OT from MNCs. A number of lesion studies have established that the synaptic input from primary osmosensory areas is required for the osmosensory activation of the effector areas, such as the SON [reviewed in (Johnson and Gross, 1993, McKinley et al., 2004b)]. Indeed, both behavioral and endocrine responses to changes in body fluid osmolality depend on the existence of specialized cells which constantly monitor extracellular osmolality and relay the information to effector neurons. In the next sections I will discuss which are these specialized osmosensors as well as the underlying CNS pathways controlling osmoregulation. 11

35 1.3. Osmosensors As it has already been introduced, osmoregulatory responses are finely tuned and depend on early detection of very minute changes in ECF osmolality. This function is attributed to specialized neurons both in the PNS and CNS, which have the intrinsic ability to detect changes in osmolality. Ernest Verney coined the term osmoreceptors and hypothesized that these sensory components are able to transform changes in external osmolality into electrical signals, with their resting potential representing the osmotic setpoint (Verney, 1947, Jewell and Verney, 1957). Electrical signals are transmitted to downstream effector neurons to control body fluid homeostasis (Gilman, 1937, Wolf, 1950, McKinley et al., 1974). Thus, osmosensors are specialized cells with an intrinsic ability to detect changes in ECF osmolality and to trigger appropriate responses (Bourque, 2008) Peripheral osmosensors A number of studies have reported that perfusion of water or hypertonic saline at several sites throughout the alimentary tract could modulate the osmotic homeostatic responses of thirst, VP release and MNC firing (Baertschi et al., 1980, Vallet and Baertschi, 1980, Baertschi and Vallet, 1981, Vallet and Baertschi, 1982, Arsenijevic and Baertschi, 1985, Choi- Kwon et al., 1990). Such peripheral osmosensory areas have been identified specifically in the oropharyngeal cavity (Kuramochi and Kobayashi, 2000), in 12

36 the gastrointestinal tract (Dooley and Valenzuela, 1984, Carlson et al., 1997, Andersen et al., 2000), in the mesenteric and portal veins subserving the upper part of the small intestine (Baertschi and Vallet, 1981, Choi-Kwon and Baertschi, 1991), and in the liver (Adachi et al., 1976, Adachi, 1984). Signals from these osmoreceptors monitoring the ingested food and water can precede a change in blood osmolality and therefore induce an anticipatory homeostatic response that will prevent large swings in body fluid solute concentrations (Haberich, 1968). Indeed, gastric water loading was shown to inhibit VP release before any significant change in blood osmolality takes place (Geelen et al., 1984, Blair-West et al., 1985, Baertschi and Pence, 1995, Huang et al., 2000b). Similarly, the sensation of thirst in humans and animals can be terminated by the ingestion of water before its absorption into the circulation (Egan et al., 2003, Stricker and Hoffmann, 2007). Conversely, salt ingestion or infusions of hypertonic saline into the hepatic portal and the mesentery veins trigger VP release and thirst before a measurable change of blood osmolality takes place (Choi-Kwon and Baertschi, 1991, Carlson et al., 1997, Huang et al., 2000a, Bykowski et al., 2007, Stricker and Hoffmann, 2007). The structure of the peripheral receptors is unknown, but there is evidence that they signal to the brain via the vagus and possibly the splachnic nerve which relays information to the nucleus of the tractus solitaris (NTS) (Kobashi and Adachi, 1985, 1993) and the ventrolateral medulla (King and Baertschi, 1991). These CNS areas feature direct connections to the median preoptic nucleus of the hypothalamus (MnPO) (Saper and Levisohn, 1983, 13

37 Edwards et al., 1989), believed to be an important integration site for central and peripheral osmotic signals (Thrasher, 1989, McKinley et al., 1992b) and to effector neurons of the SON (Kannan and Koizumi, 1981, Tribollet et al., 1985). While peripheral sensors are undoubtedly important in signaling to the CNS the state and content of the GI tract, denervation of vagal or splachnic nerves does not significantly reduce VP release in response to ingestion of hypertonic saline (Carlson and Osborn, 1998) or to systemic infusion of hypertonic saline (Raff et al., 1996). A large body of literature spanning more than 60 years of research has established that the osmosensors responsible for constantly monitoring, integrating peripheral signals and initiating homeostatic functions to correct any deviation from the osmotic set-point are found in the CNS Central osmosensors The concept of osmoreceptors was one that was recognized early as a property of specialized neurons within the CNS. It was Alfred Gilman who postulated in 1937 the existence of specialized neurons able to respond to cellular dehydration as opposed to increases in osmolality per se (Gilman, 1937). Gilman performed experiments in dogs where he injected either a membrane permeable osmolyte such as urea, or the membrane impermeable NaCl, which causes water to exit the cell. Even though both stimuli caused 14

38 equivalent osmotic increases in the blood serum, only the NaCl treatment induced significant thirst in the animals (Gilman, 1937). Verney (1947) took this line of experimenting further by investigating the effect of infusing hypertonic solutions in the blood circulating to the brain (Verney, 1947). He was able to show that not only NaCl, but any membrane impermeable osmolyte such as fructose, sucrose or sodium sulfate were able to induce antidiuresis in water replete animals, but membrane permeable substances such as urea and glucose failed to do so. Injection of moderately hypertonic solutions of membrane impermeable osmolytes into the systemic circulation also failed to elicit an equivalent antidiuretic response. This led him to hypothesize the existence of specialized neurons in the brain able to respond to the effective osmotic pressure, rather than to the absolute osmotic pressure, impinging on the cell. As explained in section 1.1., this type of stimulation forces water flux across the plasma membrane and cellular volume changes, which led Verney to propose that the specialized osmosensors are equipped with stretch receptors which monitor this osmotically-induced mechanical disturbance. In a subsequent study (Jewell and Verney, 1957), ligations of different intradural branches of the internal carotid artery, coupled with infusions of hypertonic saline into this artery, allowed the localization of the central osmosensors to a site encompassing the preoptic area and the anterior hypothalamus. This area was validated as the essential osmosensory site by a study where decerebrate dogs were shown to still be able to osmoregulate if the preoptic/anterior hypothalamus 15

39 was not damaged (Woods et al., 1966). Verney s prediction that the vasopressin releasing neurons are the primary central osmosensors were partially validated by further research showing that neurons in the supraoptic nucleus of the hypothalamus are intrinsically osmosensitive (Oliet and Bourque, 1992, 1993b). Furthermore, as will be detailed in section , these specialized neurons can detect changes in external osmolality through the activation of a mechanosensitive channel which is modulated by membrane stretch (Oliet and Bourque, 1993a). However, it became soon clear that the original conclusions of Verney and Jewell had overlooked the effect of the blood-brain barrier. Indeed, even membrane permeable solutes, such as urea, cannot pass the BBB, with the result that injection of hypertonic urea into the circulation has the same dehydrating effect on the CSF as hypertonic NaCl (Oldendorf, 1971). With this insight in mind, in a new series of experiments the authors monitored simultaneously antidiuresis, thirst and the osmolality of the CSF while hyperosmolar solutions of NaCl, sucrose or urea were injected into the carotid artery of conscious sheep (McKinley et al., 1978, Thrasher et al., 1980a). Since neither of these substances readily passes the BBB, all three treatments resulted in equivalent increases of the effective osmolality of the CSF. However, only hypertonic NaCl and sucrose, and not hypertonic urea, were found to induce antidiuresis and thirst (McKinley et al., 1978, Thrasher et al., 1980a). For an osmosensor within the brain to make this distinction, it needs to be bathed in the circulating blood and not in the CSF. The authors 16

40 therefore concluded that osmoreceptors must be located in brain regions which are not protected by the blood-brain barrier (BBB). Such areas are the circumventricular organs (CVOs), which are nuclei found in close proximity to the ventricles at various sites in the mammalian brain (Kuenzel and van Tienhoven, 1982, Ganong, 2000, Oldfield B. J. and McKinley, 2004, Oldfield and McKinley, 2004, Duvernoy and Risold, 2007). CVOs all have in common extensive vascularisation and an altered BBB due to specialized ependymal cells (Wislocki and Leduc, 1952, Bouchaud et al., 1989, Petrov et al., 1994). Since the original observation that osmosensing takes place in a BBB-free area of the brain, two of these sensory CVOs have been heavily implicated in hydromineral homeostasis: the subfornical organ (SFO) and the organum vasculosum lamina terminalis (OVLT) (McKinley et al., 1982, Thrasher et al., 1982a, Thrasher et al., 1982b, McKinley et al., 2003b, Noda, 2006). A body of multi-disciplinary research, which includes lesion studies in animal models (Simpson and Routtenberg, 1973, Thrasher et al., 1982b, Weisinger et al., 1990) and recordings of neuronal activity (Renaud et al., 1985, Ferguson and Renaud, 1986, Jhamandas et al., 1989), has shown that the SFO plays an important role in solute homeostasis. Interestingly, it appears that the SFO is mainly specialized in the detection of two signals essential for hydromineral homeostasis: AngII, via the angiotensin receptors (Joy and Lowe, 1970, Simpson and Routtenberg, 1973, van Houten and Posner, 1981, Ferguson, 2009), and sodium via the sodium channel NaX (Hiyama et al., 2002, Hiyama et al., 2004, Noda, 2006). Electrophysiological recordings both in 17

41 vitro and in vivo show that neurons in the SFO respond with increased firing rates when exposed to increases in extracellular tonicity (Gutman et al., 1988, Sibbald et al., 1988). However, in all these studies increases in osmolality were accompanied by increases in sodium concentration, or synaptic inputs were present in the preparation, leaving the question of whether the SFO can sense changes in osmolality per se unanswered. This question was addressed by the study of Anderson and colleagues who recorded isolated SFO neurons and showed that they can respond to increases in external osmolality (Anderson et al., 2000). It is still not clear if the SFO participates substantially in the regulation of osmotically induced VP/OT release and thirst, as studies of discrete lesions of this nucleus yielded conflicting results (Weisinger et al., 1990, Starbuck and Fitts, 1998). For this thesis we considered the physiology of the OVLT, which plays a major role in the regulation of osmotic control of thirst and VP/OT release, as will be reviewed below The OVLT The OVLT was described anatomically early in the 20 th century by Gerhard Behnsen (Behnsen, 1927). Through a series of experiments involving the administration of dyes into the CNS of mice, Behnsen was able to observe that the midline anterior wall of the third ventricle has special 18

42 permeability properties which allowed a systemically injected dye to accumulate within the surrounding tissue. Further work has confirmed that the OVLT belongs to a group of specialized structures within the brain that lack a proper BBB (Wislocki and King, 1936, Wislocki and Leduc, 1952, Kuhlenbeck, 1954) Morphology and function The OVLT is situated on the ventral part of the lamina terminalis, which is the tissue forming the anterior border of the third cerebral ventricle (Weindl, 1969). The OVLT lines the anterior border of the third ventricle and it extends midline from the optic chiasm to the dorsally-laying MnPO, connecting the two brain hemispheres (Landas and Philips, 1987). One of its most prominent features is the extensive intrapial vasculature which subserves the OVLT and which originates from the preoptic arteries (McKinley et al., 2003a). The terminating capillaries in the OVLT are wrapped in fenestrated endothelial cells in many species (e.g. rodents) (McKinley et al., 1983, Szabo, 1983) or endothelial cells lacking tight junctions (in humans) (Wenger and Toro, 1971), therefore not conferring a functional BBB (Landas and Philips, 1987). Interestingly, on the ventricular side, the OVLT is lined with a layer of specialized ependymal cells termed tanycytes (fusiform ependymal cells) which extend long processes into the nucleus. These tanycytes, which are joined together through tight junctions, are also found at the separation 19

43 between the OVLT and the adjacent pre-optic region, forming an effective barrier between these areas (Krisch et al., 1978). Therefore, dyes such as HRP which are used to delineate the BBB can penetrate the OVLT tissue, but are excluded from the surrounding areas due to the presence of these specialized glial cells (Broadwell and Brightman, 1976, Krisch et al., 1978). The same penetration characteristics have been shown for blood borne hormones, such as LHRH (Krisch et al., 1978). In situ hybridization, immunohistochemistry, autoradiography binding studies and neuropharmacological studies in the OVLT have showed the binding, or the expression of their respective receptors, of an impressively large number of circulating hormones, cytokines and neurotransmitters [reviewed in (McKinley et al., 2003a)]. These include amylin (Sexton et al., 1994b, van Rossum et al., 1994), calcitonin and CGRP (Henke et al., 1985, Sexton et al., 1986, Sexton et al., 1994a), angiotensin II (Van Houten et al., 1983, Mendelsohn et al., 1984), bradykinin (Murone et al., 1997), endothelin (Yamamoto et al., 1997), imidazoline (Ruggiero et al., 1998), lipopolysaccharide (Lacroix et al., 1998), natriuretic peptide (Quirion et al., 1986), the prostaglandin E2 receptor (Matsumura et al., 1990), relaxin (Osheroff and Phillips, 1991), and steroids: estrogen, glucocorticoid, progesterone and testosterone (Stumpf et al., 1992). Recently, we have also observed the expression of the VP receptor V1A in about 50% of the OVLT neurons (unpublished results). In addition to unveiling the great diversity of sensorial functions in the OVLT, what emerges from these expression studies is a clear picture of the heterogeneity of this 20

44 nucleus. For example, while angiotensin receptors are expressed mostly in the lateral and caudal parts of the OVLT (Giles et al., 1999, McKinley et al., 2004b), acetylcholinergic innervation is confined to a dense group of neurons closely bordering the dorsally lying MnPO, termed the dorsal cap (Meszaros et al., 1969). Where the experimental procedure allowed for detailed observation, each of the markers mentioned above were found to bind to, or to activate only a distinct population of OVLT neurons [e.g. (Quirion et al., 1986, Murone et al., 1997, Riediger et al., 1999)]. Little effort has been made to characterize from a molecular or morphological standpoint the various cell types (neurons, glia or vasculature) expressing the above mentioned markers and the functional interaction of the different receptors in the OVLT The role of the OVLT in osmosensing: lines of evidence Lesion studies A number of lesion studies where the lamina terminalis has been targeted have highlighted the role of the OVLT as the primary osmosensory area of the brain. While carotid artery ligation studies had narrowed the sites that trigger AVP release and thirst in response to osmolality in a range from the preoptic area to the lateral/anterior hypothalamus [section , (Jewell and Verney, 1957)], it was the seminal work of Bengt Andersson and colleagues that has unveiled the dramatic effect on osmoregulation of lamina terminalis lesions (Andersson et al., 1975). In experiments done in goats 21

45 where the anterior wall of the third ventricle was destroyed from the OVLT to the dorsally lying SFO and encompassing the MnPO in the middle, animals became adipsic, hypernatrimic due to water diuresis, and lacked an antidiuretic response to a systemic hypertonic challenge (Andersson et al., 1975). Furthermore, these effects were confirmed by similar lesions experiments in rats showing that ablation of large parts of the lamina terminalis caused severe adipsia (Johnson and Buggy, 1978, Rundgren and Fyhrquist, 1978). The star role of the OVLT in sensing osmolality and orchestrating thirst and VP release was definitely clinched by the elegant experiments performed in dogs (Thrasher et al., 1982a, Thrasher and Keil, 1987) and in sheep (McKinley et al., 1982, McKinley et al., 1984). Being able to induce precise and localized lesions, both groups have shown that ablating the OVLT alone caused a significant attenuation of the drinking response to intracarotid infusions of hypertonic saline (McKinley et al., 1982, Thrasher et al., 1982a, McKinley et al., 1984, Thrasher and Keil, 1987). Consistently, plasma osmolality was significantly increased following OVLT ablations. Despite this increase, vasopressin levels remained low in lesioned animals. Other stimuli, such as hypovolemia, did however stimulate normal VP release, demonstrating that OVLT ablations specifically affected the osmosensory regulation of the water homeostatic apparatus. Lesions that extended to the dorsally lying MnPO further reduced the ability of experimental animals to osmoregulate (McKinley et al., 2003b). This is consistent with the hypothesis that the MnPO, which is protected by the BBB, 22

46 plays an important role in integrating osmosensory signals [see section ]. The contribution of the OVLT to VP release has also been demonstrated in vitro, where lesioning the OVLT in hypothalamic explants prevents hypertonicity-induced VP release from magnocellular neurons (Sladek and Johnson, 1983). Further studies have shown that ablation of the area encompassing the OVLT in vivo disrupts both vasopressin and oxytocin release (Buggy and Jonhson, 1977, Blackburn et al., 1987, Thrasher and Keil, 1987, Allen et al., 1988, Negoro et al., 1988, Leng, 1989) Stimulation studies Consistent with lesion studies, in vivo stimulation of the region encompassing the OVLT showed that activation of this area was sufficient to trigger homeostatic osmoregulatory responses (Andersson et al., 1975, Peck and Blass, 1975, Buggy et al., 1979, Honda et al., 1987). Experiments by Andersson et al. showed that microinjections of hypertonic saline into the anterior third ventricle were an effective VP release stimulus (Andersson et al., 1974). Indeed, electrical stimulation of the anterior wall of the third ventricle, where the OVLT is located, induced VP release in vivo (Andersson et al., 1974). Local application of hypertonic saline into the OVLT area in vivo was also shown to enhance oxytocin secretion (Russell et al., 1992). Buggy and Johnson have also established that hypertonic saline targeted to the 23

47 AV3V area by microinjection is a potent stimulus for eliciting drinking in rats (Johnson and Buggy, 1978) OVLT neurons are activated by hypertonicity and inhibited by hypotonicity In vivo recordings of the OVLT are particularly difficult to obtain, owing to the technical challenges given the extremely small size of the nucleus and its location in the brain. However, the group of Honda and colleagues was able to obtain recordings of OVLT neurons in vivo while simultaneously delivering hypertonic stimulation by applying positive pressure to the recording pipette which was filled with hypertonic saline. Thus they were able to show that OVLT neurons were excited in vivo by local application of hypertonic saline, or by systemic injection of a hypertonic solution (Honda et al., 1990). Numerous publications took advantage of various in vitro preparations to obtain electrical recordings from the OVLT. Extracellular recordings in brain explants containing the OVLT showed increase in the firing rate of neurons in this area when the explant was superfused with a hypertonic solution of NaCl or mannitol and the activity of these neurons was decreased in hypotonic conditions (Sayer et al., 1984, Nissen et al., 1993, Renaud et al., 1993, Bourque et al., 1994). Similar results have been obtained with brain slices taken through the preoptic area, firmly establishing the ability of OVLT neurons to act as osmosensors in situ (Vivas et al., 1990). 24

48 Numerous studies have used the expression of the early response gene c-fos as an indication of OVLT activation in response to diverse stimuli. While not as precise as electrophysiological recordings, these studies allow observation of correlates of in vivo neural activity (Hunt et al., 1987, Morgan et al., 1987). One of the first such studies tested the effects of intravenous injections of hypertonic solutions of different osmolytes on c-fos expression in rats (Oldfield et al., 1991a, Oldfield et al., 1991b). As predicted earlier by Verney s (Verney, 1947) pioneering experiments, only hypertonic NaCl and sucrose were effective stimuli in promoting c-fos expression throughout the OVLT (Oldfield et al., 1991a, Oldfield et al., 1991b), while urea, which readily crosses the cellular membrane, did not cause significant increases in c-fos immunoreactivity (McKinley et al., 2003b). In addition to the OVLT, other structures from the lamina terminalis, the MnPO and the SFO, as well as the VP-releasing nuclei of the SON and PVN became activated. The most prominent activation however, was reported in the dorsal cap of the OVLT (McKinley et al., 2003b). Other investigators have reported a similar pattern of induced c-fos expression following a systemic hypertonic challenge (Sharp et al., 1991, Guldenaar et al., 1992, Hamamura et al., 1992, Xu and Herbert, 1996, Luckman, 1997). The activation of the OVLT in response to hypertonic saline was also tracked by the expression of other inducible transcription factors such as Jun, egr1, and Krox24 (Kovacs and Sawchenko, 1993, Luckman, 1997, Moellenhoff et al., 1998). Taken together, these studies showed that correlates of neuronal activation are induced by hypertonic 25

49 stimuli in the OVLT, corroborating the hypothesis that increases in osmolality are translated by increases in action potential firing in osmosensory neurons. The role of the OVLT in osmosensation has been established in all the experimental animal models to date: rats, dogs, sheep, goats, etc. The question of whether it participates in osmoregulation in humans was tackled in a study by Egan et al. using MRI to track the activity of this area during infusion of hypertonic saline in healthy subjects (Egan et al., 2003). Increase in blood osmolality and thirst in the subjects was positively correlated to blood flow, and implicitly activity, in the OVLT region and in regions in the prefrontal cortex related to thirst, the ACC and the INS (Egan et al., 2003). When subjects were allowed to drink prefrontal activity was rapidly quenched before the internal osmolality was changed in any significant measure by water absorption. While the activity of prefrontal cortex structures was at this point independent of the osmolality of the blood, the MRI signal in the preoptic area maintained precise correlation with this parameter throughout the study (Egan et al., 2003). Therefore, the activity of the OVLT appears to accurately monitor the osmolality of the blood, which is the expected behavior for a neurosensitive area. However, no study to date had clearly established that OVLT neurons are intrinsically osmosensitive. Vivas and colleagues used brain slices containing the OVLT to show that in the absence of synaptic input (low Ca 2+ and Mg 2+ ) OVLT neurons showed increases in firing rate to application of 26

50 hypertonic NaCl solutions (Vivas et al., 1990). This does not, however, exclude the possibility of non-synaptic modulation of neuronal activity in the OVLT. For example, as will be elaborated in section , gliotransmitters released by astrocytes under osmotic stress have been shown to participate in osmosensory transduction in the SON (Hussy et al., 2000). A more direct and definitive evidence for the intrinsic osmosensitivity of OVLT neurons, which can only be obtained by studying these neurons in isolation, will be the presented in Chapter 2 of this thesis Neural pathways involved in the osmotic control of the effector mechanisms mediated by the OVLT The control of thirst While the effect of OVLT lesions on inhibiting water intake, and conversely, the stimulation of OVLT inducing drinking in water loaded animals are well established [sections and ], so far it has proven difficult to characterize the neural networks underlying this drive. A recent study from Hollis and colleagues (Hollis et al., 2008) has recently thrown some light on the putative connections between the OVLT and the prefrontal cortex areas responsible for initiating water intake: the INS and the ACC [section 1.2.1]. Using a combination of retrograde tracers and c-fos immunostaining, the authors were able to show that osmosensory units in the OVLT send axons to the dorsal midline region of the thalamus (Hollis et al., 2008). These thalamic 27

51 neurons which show c-fos activation in response to hypertonicity, project in turn to the INS and the ACC (Hollis et al., 2008). Of interest are also the connections from the MnPO to the same region of osmotic information integration of the thalamus (Hollis et al., 2008), which could imply that some efferent signals from the OVLT could also be processed in the MnPO before being relayed to the thalamic nuclei. Although a previous study had identified direct connections from the OVLT to the ACC (Camacho and Phillips, 1981), Hollis and colleagues did not confirm the existence of direct connections from the OVLT to the prefrontal cortex in this study (Hollis et al., 2008). Thus, it remains unclear if OVLT neurons could also control the generation of thirst via direct projections to the prefrontal cortex The control of MNC activity Lesions of the OVLT result in a failure of VP release in response to osmolality, as reviewed in section Electrophysiological recordings have established that stimulation of the OVLT in vivo (Honda et al., 1987, Honda et al., 1990) and in vitro (Richard and Bourque, 1995, Trudel and Bourque, 2010) results in increased firing rate of the MNCs. This control is accomplished through direct projections from the OVLT to the SON and to the PVN, as well as indirect afferents which are relayed through the MnPO. Direct projections from the OVLT to the SON and PVN have been observed by injection of anterogradely transported tracers which resulted in the labeling of terminals in the SON (Camacho and Phillips, 1981, Uschakov et al., 2009) 28

52 and PVN (Gu and Simerly, 1997, Uschakov et al., 2007). Reversely, retrograde tracers injected into the SON (Tribollet et al., 1985, Wilkin et al., 1989, Anderson et al., 1990, McKinley et al., 1992a, Oldfield et al., 1994, Sunn et al., 2001), or into the PVN (Tribollet and Dreifuss, 1981, Sunn et al., 2001) results in labeling of OVLT neurons, with a higher density in the dorsal cap. Direct projections from the OVLT to the SON have also been mapped by electron microscopy on sections of the rat hypothalamus (Armstrong et al., 1996). Moreover, a large proportion of the OVLT neurons which project to the SON (approximately 70%) (Oldfield et al., 1994) and to the PVN (Larsen and Mikkelsen, 1995) showed induction of c-fos expression when an intravenous or intraperitoneal injection of hypertonic saline was administered, indicating a prevailing osmosensory input from the OVLT to these neurosecretory nuclei. Tracing studies have also revealed that neurons in the OVLT send efferent projections to the MnPO (Saper and Levisohn, 1983, Gu and Simerly, 1997). These connections are considered important for the control of SON and PVN since ibotenic acid injections into the MnPO, which kill cell bodies without affecting neural tracts in the area, severely decrease the VP release capabilities of the SON (Cunningham et al., 1992). Taken together these studies clearly establish that OVLT neurons regulate the activity of SON and PVN neurosecretory cells during systemic osmotic challenges. 29

53 Factors mediating osmosensory transduction in osmosensitive neurons Since the initial descriptions of osmotic homeostatic processes, studies have been mainly focused on determining the location of the osmosensory areas in the brain [section 1.3.2] and mapping the neural pathways controlling body fluid homeostasis [sections and ]. Relatively recently, however, research has started to investigate how these specialized osmosensory neurons are able to detect and translate osmotic information into electrical signals. Most of what we know about the mechanisms of osmosensation comes from studies of MNCs in the SON. Being located on the ventral surface of the brain, the SON is relatively easily accessible for in vivo [via trans-pharyngeal approach, e.g. (Day et al., 1985, Ludwig and Leng, 1997, Brown et al., 2004)], or brain explant recordings (Armstrong et al., 1982, Bourque and Renaud, 1983). Moreover, in stark contrast to the OVLT [section 1.4.1], the SON is distinctively homogeneous, being mainly composed of VP and OT releasing neurons which so far have exhibited similar osmosensing properties, presenting a clear advantage for the study of this specialized sensorial modality Osmometry and mechanosensory transduction As mentioned in section 1.3.2, one of the first and most pertinent observations with respect to osmosensation was the finding that solutes 30

54 which can permeate the plasma membrane, such as urea, are not effective in stimulating thirst and VP release. It became clear that the actual stimulus sensed by the specialized osmosensory neurons is cellular dehydration caused by the water movement out of the cells when an effective hyperosmotic gradient is established across the plasma membrane. From the correlation between the plasma osmolality and thirst initiation, Wolf was able to calculate that a 1-3% cell shrinkage is required for the triggering of osmoregulatory responses (Wolf, 1950). Verney had proposed, as early as 1947, that osmosensory neurons are equipped with stretch receptors that are sensitive to the tension in the plasma membrane (Verney, 1947). However, in addition to the mechanical changes occurring with cell shrinkage, there is a simultaneous increase in the relative concentration of ions in the intracellular milieu which could theoretically represent the actual trigger of neuronal responses. This possibility was addressed in the SON using isolated MNCs. Electrophysiological studies have shown that isolated SON neurons are intrinsically osmosensitive (Oliet and Bourque, 1992, 1994). Under whole cell patch configuration, hypertonicity was found to induce an increase in membrane conductance which was correlated to the decrease in cell volume (Oliet and Bourque, 1994). MNCs also behave as osmometers, i.e. their volume change in response to osmotic challenges is accurately proportional to the strength of the stimulus (Zhang and Bourque, 2003). This property represents a specialized function of these osmosensory neurons, as most other neuronal cells show some degree of resistance to volume changes 31

55 when placed in solutions of different osmolality (Leng, 1989). To determine if volume changes are important in osmosensing, Zhang and Bourque measured the electrophysiological responses of MNCs in a paradigm where changes in cell volume were attained by mechanical, as opposed to osmotic, means (Zhang and Bourque, 2003). Applying suction to the recording pipette resulted in equivalent conductance changes and increased firing rate in MNCs as when similar volume changes were obtained by osmotic stimulation. Moreover, neuronal responses to osmotic stimulation could be blocked when the shrinkage of the cell was prevented with positive pressure in the pipette (Zhang and Bourque, 2003). Taken together, these results indicate that osmotic stimulation in MNCs is transduced by the mechanical changes to the cellular structures. Whether the same mechanism is at play in OVLT neurons remains to be determined, and will be the subject of our investigations in Chapter TRPV channels as candidate molecular osmosensors Efforts have also been made to identify the molecular transducer of osmotic information. In MNCs, both osmotic and mechanical stimulation evoke an increase in a non-selective cation conductance (Erev of -42mV) which is sensitive to application of RR or Gd 3+ (Oliet and Bourque, 1992, 1994, Zhang and Bourque, 2003, Sharif Naeini et al., 2006). Single channel recordings in the MNCs have demonstrated the existence of a channel of a 32

56 similar reversal potential which can be modulated by changes in the stretch of the membrane (Oliet and Bourque, 1994). The identity of this molecular transducer of osmotic information remains unknown and is currently the subject of many lines of research in the field of mechano- and osmotransduction. The most prominent candidates are at the moment members of the transient receptor potential-vanilloid family of channels as will be discussed in the following sections. TRPV channels are one of the six related subfamilies of the transient receptor potential (TRP) family of cation channels, which also include the canonical (TRPC), melastatin (TRPM), mucolipin (TRPL), polycystin (TRPP) and ankyrin (TRPA) channels (Clapham, 2007, Damann et al., 2008, Patapoutian et al., 2009). All these related channels are putative sixtransmembrane domain proteins which form cation permeable pores in a tetrameric formation (Clapham, 2003). They also have in common a well conserved TRP domain which refers to a stretch of 25 residues in the proximity of the pore region (Clapham, 2003, Montell, 2005). TRP channel full length proteins and numerous reported splice-variants are ubiquitously expressed and appear to function mainly in sensory systems: light (Cosens and Manning, 1969, Montell et al., 1985, Hardie and Minke, 1992), pheromones (Stowers et al., 2002), taste (Zhang et al., 2003), olfaction (Colbert et al., 1997), mechanical touch (Corey, 2003), temperature (Caterina et al., 1997a, Sharif-Naeini et al., 2008a), ph (Caterina et al., 1997a), etc. TRPV1 is the founding member of the vanilloid family of TRP channels, 33

57 TRPV, and was cloned from a rat sensory neuron cdna library as the receptor for capsaicin, the pungent ingredient in hot chili peppers (Caterina et al., 1997b). Five more TRPV channels, TRPV2-6 were later identified by homology. However, based on functional and structural similarities, it is currently considered that TRPV5 and 6 constitute a different group from the rest of the TRPVs (Clapham, 2003, Montell, 2005). In general, TRPV1-4 can be characterized by a specific temperature activation profile and by their permeability to cations, with a moderate preference for Ca (pca/pna=1-10) (Clapham et al., 2003). TRPV channels came into the spotlight of the mammalian osmotransduction field in 1999, when Suzuki et al. reported the cloning of the first stretch inhibited channel (SIC) to function in a heterologous system (Suzuki et al., 1999). This channel was identified by screening a rat kidney cdna library using a probe for the TRP domain of rat TRPV1. The reported channel was a chimera formed out of exons 6-14 of TRPV1 channel in the N-terminus and the pore region, and the C-terminus of TRPV4. When expressed in CHO cells, this channel displayed calcium permeability, was activated by hypertonicity and cellular shrinkage and blocked by Gadolinium (Gd + ), a well-known blocker of mechanosensitive channels (Yang and Sachs, 1989, Suzuki et al., 1999). This chimeric construct turned out to be a cloning artifact which is not expressed in vivo (Xue et al., 2001). The in vitro experiments however did raise the possibility that members of the TRPV family of channels could act as molecular mechanosensors transducing 34

58 osmotic stimulation. Indeed, further work has highlighted the possible contribution of TRPV1 and TRPV4 to central osmosensation a. TRPV1 When expressed in cell lines, TRPV1 forms homomeric ion channels with a single channel conductance of 35 ps at hyperpolarized voltages and showing a slight preference for fluxing calcium over other cations (PCa 2+ /PNa + = 9.6) (Caterina et al., 1997a). Beside the capsaicin activation used for cloning this channel, the most extensively studied mode of activation for TRPV1 is temperature. In heterologous expression systems TRPV1 is activated by temperatures above 42 o C, whether tested under whole cell patch recordings or in excised membrane patches (Caterina et al., 1997a, Tominaga et al., 1998) The threshold of activation can be reduced to physiological temperatures (33-37 o C) by intracellular modulators such as PIP2, PKA or PKC (Premkumar and Ahern, 2000, Vellani et al., 2001, Prescott and Julius, 2003). Recently, TRPV1 was also shown to function in osmoreception in the SON (Sharif Naeini et al., 2006). TRPV1 immunoreactivity was detected in MNCs when probed for with an antibody directed against the C-terminus of the protein (Sharif Naeini et al., 2006). Similarly, tissue RT-PCR shows expression of the 3 end of the mrna encoding TRPV1, but not the first 4 exons, suggesting a gene splicing event 35

59 which removes part of the N-terminus of the protein (Sharif Naeini et al., 2006). Mice lacking Trpv1 (Trpv1 -/- ) have a chronically high internal osmolality and blunted VP release in response to a hypertonic challenge. Moreover, the gene is required for transduction of osmosensory stimulation in SON neurons, as isolated MNC failed to increase conductance or firing rate in response to hypertonicity-induced shrinking (Sharif Naeini et al., 2006). The possibility that TRPV1 also functions as a molecular osmosensor in the OVLT will be the subject of Chapter 2 of this thesis b. TRPV4 The possibility that TRPV channels can act as molecular osmosensors was reinforced when the C.elegans orthologue of TRP channels, Osm-9 (Colbert et al., 1997) was shown to be the osmosensor responsible for a wellcharacterized avoidance behavior to hypertonicity in these organisms (Liedtke et al., 2000). Interestingly, mammalian TRPV4 was shown to rescue the Osm-9 knock out phenotype (Liedtke et al., 2000). In cellular expression systems, TRPV4 forms a homomeric channel with a conductance of 90pS at hyperpolarized voltages which can be activated by temperatures in the physiological range [>30 o C, (Guler et al., 2002)] and by hypotonicity (Strotmann et al., 2000, Wissenbach et al., 2000, Liedtke et al., 2003). Both these modalities are hypothesized to activate TRPV4 indirectly through a secondary messenger, since this channel loses its heat sensitivity in excised 36

60 patches of membrane (Watanabe et al., 2002) and pressure applied to the excised patch does not modulate TRPV4 activity (Strotmann et al., 2000). Furthermore, activation of TRPV4 by hypotonicity is dependent on PLA2 activation and arachidonic acid release (Vriens et al., 2004). Studies of two different strains of genetically engineered mice lacking TRPV4 investigated the role of this channel in osmoregulation. Liedtke and Freedman created a Trpv4 -/- line by eliminating the entire pore loop domain and the adjacent transmembrane domains 5 and 6 (Liedtke and Friedman, 2003). These mice were hyperosmolar under normal conditions and released significantly lower amounts of vasopressin in response to systemic injections of hypertonic saline. These mice also drank less water than WT controls when challenged with a hypertonic saline injection and displayed reduced c-fos induction in the OVLT to this stimulus. The results led the authors to propose that TRPV4 forms the osmosensory channel in the central osmosensory area of the OVLT (Liedtke and Friedman, 2003). However, in a different strain of Trpv4 -/- mice where the authors inserted a stop signal in exon 4, the phenotype with respect to osmoregulation was considerably different (Mizuno et al., 2003). The authors observed no effect on the intake of water or on the serum osmolality. Hypertonic stimulation of hypothalamic slices from Trpv4 -/- mice resulted in significantly higher VP release than when the slices were obtained from WT mice (Mizuno et al., 2003). The difference in these two studies is not easily explained. With respect to VP release, it is essential to keep in mind that in the former study the hypertonic stimulus is systemic, which can 37

61 engage peripheral osmosensors that impinge upon the OVLT and the SON. Indeed, TRPV4 was reported to play a role in the function of the osmopressor response in the portal vein (McHugh et al., 2010). In the second study, the IP injections of propylene glycol as a stimulus caused hypovolemia in addition to systemic hypertonicity, therefore implicating baroreceptor and angiotensinergic stimulation of VP release (Mizuno et al., 2003). Whether TRPV4 is involved in osmosensory transduction in the primary osmosensor of the brain, the OVLT, remains to be determined and will be investigated in Chapter 3 of this thesis Gliotransmission In addition to the intrinsic ability of MNCs to sense osmolality, and the inputs that modulate its firing rate, another emerging contributor to osmotic sensing is represented by the glial component of osmosensory structures. Astrocytes in the SON, and presumably all over the brain, respond to hypotonic stimulation by actively opposing extreme cellular swelling through regulatory volume decrease (RVD) (Law, 1994, Pasantes-Morales et al., 2000). Volume increase due to water entering the cell in hypotonic surroundings triggers a series of poorly characterized steps which result in the excretion of intracellular osmolytes, therefore reducing the osmotic pressure on the cell (Pasantes-Morales et al., 2000). The excreted osmolytes consist of inorganic ions, mainly potassium and chloride, and organic 38

62 macromolecules, mainly amino acids, amines and polyalcohols (Sanchez- Olea et al., 1992, Pasantes-Morales et al., 1994a, Pasantes-Morales et al., 1994b). Interestingly, TRPV4 appears to play an important role in the initial step triggering RVD in various tissues in the body: lung airways, vasculature, salivary glands, etc (Becker et al., 2005, Becker et al., 2009, Aure et al., 2010). In the SON, glial release of taurine during RVD has been suggested to play a neuromodulatory role during hypotonic conditions (Pasantes-Morales et al., 1990, Law, 1994, Hussy et al., 1997, Hussy et al., 2000). Taurine is an agonist of strychnine-sensitive glycine receptors (Hussy et al., 1997, Hussy et al., 2000) which are chlorine permeable channels expressed at high levels on SON neurons (Randle and Renaud, 1987) at extra-synaptic sites (Deleuze et al., 2005). Application of taurine in vitro causes hyperpolarization in MNCs (Randle and Renaud, 1987) and in vivo it inhibits the firing rate of SON neurons (Hussy et al., 1997). Immunohistochmistry has allowed the localization of taurine in the SON predominantly in the cell bodies of astrocytes which line the ventral side of the nucleus and in their processes found in close proximity to neuronal cell bodies (Decavel and Hatton, 1995). In cultured astrocytes taurine is released mainly from volume regulated anion channels (VRAC), as DCPIB, a blocker of these channels prevents taurine release (Bres et al., 2000). In the SON, taurine was shown to be released by decreases of osmolality as small as 5% (Deleuze et al., 1998, Hussy et al., 2000). There appears to be a basal level of taurine release which can be inhibited by exposure to hypertonicity or by application of the specific glial 39

63 toxin fluorocitrate (Deleuze et al., 1998). Whether this iso-osmotic level of taurine release is sufficient to have an impact on the electrical activity of SON neurons has not been reported to date. With respect to the contribution of glia in the osmosensory process in the OVLT there is to date no report. Our preliminary studies on rat OVLT explants have allowed us to detect a significant amount of taurine released in response to hypotonic stimulation (unpublished observation). Therefore the possibility that OVLT glia contribute to osmotic sensing will be investigated in chapter 5 in the context of this thesis Integration of different sensorial modalities As mentioned before, the OVLT contains neurons that can respond to a variety of stimuli. As reviewed in section 1.4.1, in addition to osmolality, OVLT neurons respond to circulating hormones, such as angiotensin and LHRH, to blood borne cytokines, such as the fever and inflammation related factors, and to other environmental factors, most remarkably, to temperature. This effect is highly relevant in terms of homeostatic physiology. Hyperosmotic conditions often coincide and interfere with other physiological states, such as hypovolemia or hyperthermia. For example, both dehydration and hypovolemia actively inhibit panting and evaporative water loss during high body temperature in animals (Hainsworth et al., 1968, Baker et al., 1983, Barney et al., 1991). Reversibly, high body temperature induces water 40

64 conservation mechanisms, such as vasopressin release and thirst, before any significant amount of evaporative water loss occurs (Forsling et al., 1976, Horowitz and Meiri, 1985). This preemptive effect helps prevent large loss of fluid and fluctuations in body fluid osmolality. How such integration occurs remains unclear. In the OVLT, studies have shown c-fos induction in response to both thermic and osmotic stimulation (Scammell et al., 1993, Patronas et al., 1998). There is also electrophysiological evidence that there is a population of neurons in the OVLT area that responds to local application of both osmotic and heat stimuli (Nakashima et al., 1985, Boulant and Silva, 1987, Hori et al., 1988). A mechanism that could explain the integration of these two sensorial modalities is at present unknown Hypotheses addressed by this thesis The purpose of the research presented in this thesis is to address some of the unanswered questions about OVLT physiology and the role it plays in osmosensing. Specifically, we have focused on showing that OVLT neurons are intrinsically osmosensitive, identifying a molecular mechanism of osmosensation, on characterizing the steps involved in transducing osmotic stimulation into action potentials, as well as describing a mechanism for multimodal sensing in osmosensory neurons and unveiling the glial contribution to osmotic sensing. Here is a brief summary of the hypotheses that were tested and the results presented in the subsequent chapters: 41

65 Chapter 2: OVLT neurons are intrinsically osmosensitive and they rely on a functional TRPV1 product for transducing osmosensory information. As summarized above, an extensive body of evidence implicates OVLT neurons as the brain s primary osmosensors but it is not known if neurons in this nucleus are intrinsically osmosensitive. Here we use acutely isolated OVLT neurons to show that they have an intrinsic ability to sense osmotic changes. A major candidate for the role of the osmosensory transduction channel is TRPV1. We take advantage of the mouse Trpv1-/- line available in our laboratory to investigate the possibility that a product of the Trpv1 gene is required for osmosensory transduction in the OVLT. Chapter 3: Osmosensing in the OVLT involves mechanosensory transduction. Previous studies have linked the change in cellular volume to the transduction of osmotic stimuli. Here we test and characterize every electrophysiological step involved in the translation of osmotic information, or mechanical disturbance, into action potential generation, and we show that these two processes are equivalent and relying on TRPV1 for transduction. Furthermore, we test the hypothesis that TRPV4 participates in the intrinsic osmosensitivity of OVLT neurons. Chapter 4: Gliotransmission can rescue hypotonicity sensing in Trpv1-/- mice. Little is known about hypotonicity sensing and particularly if the same 42

66 mechanisms participate in this sensory modality. Astrocytes in particular have been reported to modulate neuronal activity during hypotonic stress. Here we show that hypotonicity sensing in OVLT neurons relies on the intrinsic mechanosensory abilities of OVLT cells, which are mediated by a channel encoded by TRPV1. The absence of intrinsic osmosensitivity in Trpv1-/- mice allows us to unveil the contribution of gliotransmission to hypotonicity sensing in situ. Chapter 5: Temperature stimulation is transduced through the osmosensory channel in osmosensory neurons. Thermic and osmotic stimulation can activate the same osmosensory neurons in both OVLT and SON. Here we show a mechanism by which SON neurons can sense both temperature and osmolality and a way by which these two physiological parameters can trigger the same homeostatic pathways. 43

67 Figure 1.1. Extracellular fluid (ECF) osmolality in a variety of organisms, with black circles representing mammalians and white circles other animals. This illustration is adapted from (Bourque, 2008). 44

68 Figure 1.2. Feedback regulation of plasma osmolality. A series of fine-tuned homeostatic mechanisms detect changes in ECF osmolality to maintain a defined set-point in ECF osmolality as shown in Fig Changes in body fluid osmolality (center) evoke behavioral (left) and endocrine (right) responses. Dashed lines represent potential homeostatic responses that have not been, as of yet, demonstrated. This figure is modified from (Voisin and Bourque, 2002, Bourque, 2008). 45

69 Figure 1.3. Anatomy and connectivity of the OVLT to the main osmoregulatory centres in the CNS and PNS. Osmotic information is directly sensed in the OVLT (red) and transmitted directly to effector neurons (green) of the SON and PVN, and possibly the ACC (white), or to osmotic integration sites in the MnPO and the thalamus, where it is further distributed to the INS and the ACC (white). Peripheral information is believed to reach the MnPO via the NTS and VLM (Bourque, 2008). 46

70 CHAPTER TWO: TRPV1 IS REQUIRED FOR INTRINSIC OSMORECEPTION IN OVLT NEURONS AND FOR NORMAL THIRST RESPONSES TO SYSTEMIC HYPEROSMOLALITY 2.1. OVERVIEW As reviewed in chapter 1, the brain's primary homeostatic osmosensor is located in the OVLT (Johnson and Gross, 1993, McKinley et al., 1999), a circumventricular organ located at the rostral ventral border of the third ventricle. Although electrophysiological recordings in vivo (Honda et al., 1990) and in vitro (Sayer et al., 1984, Vivas et al., 1990, Nissen et al., 1993) have shown that OVLT neurons are excited by hypertonic stimuli, the functional basis for this osmoresponsiveness remains unknown. Specifically, it is not known if these neurons are intrinsically osmosensitive, or if the effects of osmotic perturbations are mediated indirectly. Indeed, hyperosmolality could provoke excitation through the depolarizing effect of hypertonicityactivated cation channels expressed in the OVLT neurons themselves, as found in vasopressin releasing neurons of the supraoptic nucleus (Oliet and Bourque, 1993a, Bourque et al., 1994). Alternatively, increases in osmolality might turn off basally active hypotonicity-activated channels expressed in a neighboring cell, thereby suppressing the release of an inhibitory substance onto OVLT neurons (Mizuno et al., 2003). In this study, we used electrophysiological approaches to determine if OVLT neurons are intrinsically sensitive to hypertonic stimuli. Since recent experiments have 47

71 suggested that TRPV1 may contribute to the formation of a cation channel activated by hypertonicity in osmosensory neurons of the SON (Sharif Naeini et al., 2006) [see section ] we compared the effects of osmotic stimuli on OVLT neurons obtained from wild type (WT) and Trpv1 / mice MATERIALS AND METHODS Extracellular recordings from hypothalamic explants C57/Bl and Trpv1 -/- mice from the same background, 6-8 wks of age, were anaesthetized using halothane, decapitated and the brains removed rapidly according to a procedure approved by the McGill University Animal Care Committee. A block of tissue mm containing the optic chiasm and the OVLT was excised using razor blades and pinned, ventral surface up, to the Sylgard base of a superfusion chamber. Within 3 5 min from decapitation, the excised OVLT explant was superfused (at 1.5 ml.min -1 at 32 to 33 C) with carbogenated (95% oxygen; 5% CO 2 ) artificial cerebrospinal fluid (ACSF) delivered via a Tygon tube placed over the rostral pole of the preparation. The optic nerves were removed to provide direct access to the OVLT, and a cotton wick was placed at the caudal tip of the explant to minimize the level of fluid over the preparation and to facilitate the removal of ACSF. The ACSF (ph 7.4) comprised (in mm): NaCl (125), NaHCO 3 (26), KCl (5), MgCl 2 (1.5), CaCl 2 (1), D-glucose (10). Mannitol was added to adjust the basal osmolality of the solution to 312 ± 2 mosmol.kg -1, or to make hypertonic solutions. Extracellular recordings were made from OVLT neurons using micro-electrodes filled with 1M potassium acetate (resistance 5 to 10 48

72 MΩ) and hyperosmotic stimuli were bath-applied by switching the solution being delivered through the Tygon tube. For the analysis of changes in firing rate, the average rate of spike discharge of the neuron measured during the final minute of an osmotic stimulus was subtracted from the average firing rate observed prior to the onset of the stimulus Whole cell recordings from acutely Isolated OVLT neurons To obtain acutely isolated OVLT neurons, blocks of tissue containing the OVLT (<1 mm 3 ) were dissected from the brains of anaesthetised C57/BL or Trpv1 -/- mice (6-8 weeks old) and incubated in oxygenated (100% O 2 ) PIPES solution comprising (in mm): NaCl (120), KCl (5), MgCl 2 (1), PIPES (20), CaCl 2 (1), D-Glucose (10); ph 7.3, as well as 0.5 mg/ml Protease-X and 0.5 mg/ml Protease-XIV at room temperature for 30 min. Individual tissue blocks in enzyme-free PIPES solution were triturated, and the resulting suspension was plated onto petri dishes. Cells were perfused (1.2 ml.min -1 ) with HEPES solution comprising (in mm): NaCl (150), KCl (3), MgCl 2 (1), HEPES (10), CaCl 2 (1) and D-Glucose (10), ph 7.3, through one of the adjacent barrels of a three-channel glass capillary tube assembly. Mannitol was added to adjust the basal osmolality of the solution to 312 ± 2 mosmol.kg -1, or to make hypertonic solutions. Glass pipettes (1 mm O.D.) were pulled on a Flaming-Brown puller (Sutter Instruments) and filled with an internal solution (ph 7.25) containing (in mm): K-Gluconate (120), MgCl 2 (1), EGTA (1), Na 2 ATP (4), NaGTP (1), Phosphocreatine (14), HEPES (10). Cells 49

73 were patch-clamped in the whole cell current-clamp and voltage-clamp modes using an Axopatch-1D amplifier (Molecular Devices Corp., Sunnyvale, CA). In the current-clamp experiments cells were deemed responsive to hypertonic stimuli if they displayed a membrane depolarization >2 mv and increases in firing rate >15 % compared to control. In the voltage-clamp experiments 0.5 μm tetrodotoxin (TTX) was added to the HEPES solution to block voltage-gated sodium channels. In both conditions, hyperosmotic stimuli were applied by switching the channel of the delivery tube assembly to one containing hyperosmotic HEPES using a piezoelectric stepper device (SF-77B, Warner Instruments Co., Hamden, CT). The TRPV channel blocker Ruthenium Red (RR) was used at a concentration of 10 µm in the perfusion solution. Gadolinium (Gd + ) was diluted from a stock concentration of 30 mm in water to the final concentration of 300 µm in the perfusing solution. In all experiments the overflow solution was sucked away via a vacuum system Volume measurements We quantified percent changes in cell volume from measurements of maximal cross-sectional area (CSA, from digitized video images) of the cell using Scion Image for Windows 4.02 (Scion Corp., Frederick MA) as performed elsewhere (Oliet and Bourque, 1993b). All values of CSA measured in the control period were averaged (CSA o ) and values of normalized volume at individual time points (nv t ) were calculated from the CSA value at that time point (CSA t ) using the equation; nv t = 50

74 [(CSA t ) 1.5 /(CSA o ) 1.5 ]. Percent changes in volume (ΔV%) were calculated as ΔV%=(1-nV t )* Immunohistochemistry Coronal sections through the OVLT (15 μm thick) were obtained from C57/BL and Trpv1 -/- mice brains perfused with 4% paraformaldehyde in phosphate buffered saline (PFA/PBS). Sections were blocked for 1 hour at room temperature in 4 % normal goat serum, 0.5 % bovine serum albumin and 0.1 % Triton-X, followed by incubation with primary antibody overnight at 4 o C. For the detection of TRPV1 we used a rabbit anti-trpv1 C-terminus antibody (1:400, Neuromics, Inc., Northfield, MN) which recognizes the peptide sequence EDAEVFKDSMAPGEK (Price et al., 2005). In order to colabel neurons we used mouse anti-neun antibody (1:50, Chemicon Inc., Temecula, CA). Secondary antibodies (Alexa Fluor 488 goat anti-mouse and Alexa Fluor 568 goat anti-rabbit, both diluted 1:500 in PBS; Invitrogen Canada Inc., Burlington, ON) were applied for 30 min. Fluorescence images were acquired using a spinning disk confocal microscope equipped with an ORCA-ER camera and MetaMorph Imaging software (PerkinElmer Biosignal Inc. Montreal, QC). 51

75 Water intake measurements in mice Age-matched WT and Trpv1 / mice housed in individual cages were injected intraperitoneally with hypertonic (1 M NaCl) or isotonic (150 mm NaCl) solution at a dose of 10 μl per gram of body weight. Water and food were removed at this point. After 30 minutes, water was reintroduced in the cages and the amount of water consumed (per gram body weight) during a 30 minute period was determined Statistics All values are reported as mean ± s.e. Comparisons of the means between groups were made using a paired t-test, or a two-way analysis of variance (ANOVA, Sigmastat 2.1, Jandel), as appropriate. Differences between the means were considered significant when p<0.05. Where differences were found, the Student-Newman-Keuls test for multiple comparisons was performed post hoc to identify specific distinctions (p<0.05). Slopes of linear regressions were compared using Prism 4.03 (GraphPad Software, Inc.) RESULTS TRPV1 is expressed in OVLT neurons To determine if the Trpv1 gene is expressed in murine OVLT, we performed immunohistochemical staining of coronal tissue sections using an 52

76 antibody directed against the C-terminus of TRPV1. As shown in Figure 2.1A, immunolabeling was detected throughout the OVLT of WT mice, including in the dorsal cap region that has been associated previously with osmodetection (Oldfield et al., 1994). Notably, intense TRPV1 immunoreactivity was present in the somata of most, but not all of the neurons contained within this region (Figure 2.1B), whereas no signal could be detected in the OVLT of Trpv1 / mice Osmosensitivity of OVLT neurons in hypothalamic explants We first investigated the possible involvement of the Trpv1 gene in the osmotic control of OVLT neurons by comparing the effects of hypertonic stimulation on the electrical activity of these cells in superfused hypothalamic explants obtained from adult WT and Trpv1 / mice. Increasing the osmolality of the external solution by the addition of mannitol caused reversible increases in the firing rate of many WT OVLT neurons (e.g. Figure 2.2A). As illustrated in Figure 2.2B, this increase in firing rate was proportional to the magnitude of the osmotic stimulus, and linear regression analysis of the data revealed that firing rate in the population as a whole increased at a rate of Hz/mosmol.kg -1 (r = 0.56, p < ; n = 128). In contrast, OVLT neurons recorded from hypothalamic explants prepared from age-matched Trpv1 / mice consistently failed to respond to osmotic stimulation (e.g. Figure 2.2C). Indeed, regression analysis (Figure 2.2D) revealed that the slope of the relation between changes in firing rate and osmolality observed 53

77 in these cells ( Hz/mosmol.kg -1 (r = 0.12, n = 62) was not significantly different from zero (p = 0.35). To estimate the true sensitivity of the subset of WT OVLT neurons that are osmoresponsive, we next separated WT cells according to the magnitude of their response to osmotic stimulation. Since the overall standard deviation of changes in firing rate measured in (nonosmoresponsive) OVLT neurons from Trpv1 / mice was ±0.48 Hz, WT OVLT neurons showing responses greater than this value were considered to be osmoresponsive, whereas cells showing changes in firing rate smaller than this value were considered to be non-osmosensitive. Based on this analysis, 63% (40/64) of the WT OVLT neurons we sampled were deemed to be osmoresponsive, and the sensitivity of these particular cells was Hz/mosmol.kg -1 (r = 0.72, n = 80). As expected, there was no relation between changes in firing rate and osmolality in the group of OVLT cells identified as non-osmosensitive (slope = Hz/mosmol.kg -1 ; r = 0.29, n = 48). These findings suggest that a majority of OVLT neurons are osmoresponsive and that a product of the Trpv1 gene is required for expression of this phenotype in situ OVLT neurons are intrinsically osmosensitive The excitatory effects of hypertonicity observed in OVLT neurons in explants obtained from WT mice could reflect an intrinsic osmosensitivity of these neurons, or an indirect osmoresponsiveness mediated by substances released by adjacent non-neuronal cells or by synaptic afferents originating 54

78 from neurons in other nuclei. To determine if OVLT neurons are intrinsically osmosensitive, we therefore examined the effects of hyperosmotic stimulation on neurons acutely isolated from the OVLT of adult WT mice. Under whole cell current clamp conditions, a hypertonic stimulus of +20 mosmol.kg -1 caused a mean depolarization of the WT OVLT neurons from ± 1.3 mv to ± 1.8 mv (n = 36, p = 1.01 x 10-5 ; Figure 2.3A,B). In addition, this effect was associated with a significant increase in the firing rate of the cells from 0.11 ± 0.05 Hz to 1.43 ± 0.42 Hz (n = 36, p = 0.002; Figure 2.3C). Indeed, when considering the entire population of responsive neurons (>61%), the increase in firing rate observed during osmotic stimulation was positively correlated with the amplitude of the depolarization (slope = 0.26 Hz/mV; r = 0.69; data not shown; see also chapter 3 and Figure 3.8), indicating that the magnitude of the osmoreceptor potential determines the degree of excitation under these conditions. These findings reveal that a population of OVLT neurons has the intrinsic ability to transduce hyperosmotic stimuli into depolarising receptor potentials that generate proportional increases in the rate of action potential discharge. To determine if the Trpv1 gene contributes to the intrinsic osmosensitivity of OVLT neurons, we repeated the same test on neurons isolated from the OVLT of agematched Trpv1 / mice. As illustrated in Figure 2.3A, hyperosmotic stimuli failed to affect the membrane potential (from ± 1.6 mv to ± 1.4 mv, n = 20, p = 0.4; Figure 2.3B) or the firing rate of Trpv1 / neurons (from 0.07 ± 0.04 Hz to 0.15 ± 0.10 Hz, n = 20, p = 0.21; Figure 2.3C). Taken 55

79 together these data indicate that OVLT neurons are intrinsically osmosensitive, and suggest that the presence of a product of the Trpv1 gene is necessary for osmosensory transduction Osmosensory transduction involves hypertonicity-activated cation channels If a channel comprising TRPV1 protein operates as the osmosensory transduction channel, then the depolarizing and excitatory responses to hypertonicity should be caused by an inward current associated with an increase in membrane cation conductance. We therefore investigated this issue by performing steady-state current-voltage (I-V) analysis on isolated WT OVLT neurons under whole cell voltage clamp in the presence of TTX. In cells clamped at a holding voltage near resting potential (-60 mv), bath application of a hypertonic stimulus caused a reversible inward current (Figure 2.4A). Moreover, steady-state I-V relations measured every 15 seconds revealed that the amplitude of this inward current was associated with a proportional increase in membrane conductance, confirming that ion channels are activated under these conditions. The mean reversal potential of the osmotically evoked current was ± 4.5mV, consistent with the involvement of non-selective cation channels under our experimental conditions (see section 2.4). Additionally, increases in membrane conductance induced by hyperosmolality could be reversed by bath 56

80 application of 300 μm Gd 3+ (ΔG = 0.51 ± 0.15 ns before and ΔG = ± 0.06 ns after Gd 3+ ; n = 8, p = 0.003; data not shown), a blocker of nonselective cation channels (Hase et al., 1995, Cho et al., 2002) OVLT neurons from Trpv1 / mice fail to transduce osmotic stimuli In order to determine if the presence of TRPV1 is required for the conductance increase and activation of an inward current, we examined the effects of hypertonic stimuli on voltage-clamped OVLT neurons from Trpv1 / mice. As illustrated in Figure 2.4B, OVLT neurons from Trpv1 / animals failed to show detectable changes in holding current or membrane conductance upon exposure to hypertonicity. Whereas the membrane conductance of WT OVLT neurons increased significantly during the hypertonic stimulation (from 0.91 ± 0.08 ns to 1.24 ± 0.12 ns, n = 38, p = 1.56 x 10-5 ), OVLT neurons isolated from Trpv1 / mice showed no significant changes in conductance in response to the same stimulus (control 0.94 ± 0.09 ns vs ± 0.09 ns during hypertonicity, n = 23; p = 0.7; Figure 2.4C). Since changes in cell volume are hypothesized to participate in osmosensory transduction (Oliet and Bourque, 1993a, Hussy et al., 2000), we also compared the degree of volume change induced by hypertonic stimulation in OVLT neurons isolated from WT and Trpv1 / mice. As shown in Figure 2.4D, no significant difference was observed between the two strains, indicating that the absence of osmotically induced responses in 57

81 OVLT neurons of Trpv1 / mice is not due to an alteration of the effect of osmotic stimulation on cell volume Osmosensory transduction is blocked by ruthenium red The data presented so far show that cation channels transduce hyperosmotic stimuli into conductance increases and depolarising receptor potentials to excite WT OVLT neurons, and such effects are lost in Trpv1 / mice. If a product of the Trpv1 gene is a functional component of the osmosensory transduction channel, then responses to hypertonicity should be inhibited by Ruthenium Red (RR), a generic blocker of TRPV channels (Watanabe et al., 2003). As illustrated in Figure 2.5A, the presence of 10 μm RR abolished the conductance increase induced by hypertonic stimulation in isolated WT OVLT neurons. Under control conditions, the membrane conductance of OVLT neurons increased from 0.87 ± 0.12 ns to 1.14 ± 0.16 ns, n = 19; p = 0.005), whereas RR-treated cells failed to show a significant change (from 0.67 ± 0.07 ns to 0.70 ± 0.14 ns, n = 8; p = 0.8; Figure 2.5B) TRPV1 contributes to osmotic thirst in vivo Since the osmotic activation of OVLT neurons is believed to contribute to the generation of thirst (Thrasher, 1989, McKinley et al., 2004a, McKinley et al., 2004b), we hypothesized that Trpv1 / mice would show deficits in drinking behavior following a systemic hypertonic challenge. To examine this 58

82 hypothesis, access to water was removed for a period of 30 minutes following a single intraperitoneal injection (10 ml/kg of body weight) of either saline (0.15 M NaCl) or hypertonic solution (1 M NaCl) (Figure 2.6A). Analysis performed on separate animals revealed that injection of the hypertonic solution caused an equivalent increase in serum osmolality in both WT (+15.8 ± 1.7 mosmol.kg -1 ; n = 8) and Trpv1 / mice (+15.2 ± 1.9 mosmol.kg -1 ; n = 10) compared to genetically matched saline injected controls (p = 0.827), a stimulus sufficient to motivate drinking in most mammals (Thrasher et al., 1982a, Egan et al., 2003). Indeed, during the first 30 min following the reintroduction of water, osmotically stimulated WT mice drank significantly more per unit body weight (30.7 ± 2.05 μl/g) than saline injected controls (5.3 ± 1.2μl/g; p = 3.3 x 10-8 ). The same test performed in Trpv1 / mice also evoked drinking in these animals (23.4 ± 2.1 μl/g in hypertonic vs. 7.9 ± 1.9 μl/g in control; p = 5.5 x 10-5 ). However, osmotically-challenged Trpv1 / mice ingested significantly less water than WT animals (p = 0.02) (Figure 2.6B). Thus, a product of the Trpv1 gene contributes to the generation of thirst in response to systemic hypertonicity. 59

83 2.4. DISCUSSION The OVLT is a primary homeostatic osmosensory nucleus Pioneering experiments by Andersson and colleagues (Andersson et al., 1975) first demonstrated the importance of the anterior wall of the third ventricle in the generation of thirst over thirty years ago. Studies combining systemic infusions, intracerebroventricular injection and fluid sampling, along with neuroendocrine and behavioral analysis, subsequently established that an increase in the osmotic pressure of the cerebrospinal fluid is sufficient to induce thirst and vasopressin release in unanaesthetised animals (Andersson, 1971, McKinley et al., 1978, Buggy et al., 1979, Thrasher et al., 1980b). Lesion studies performed in vivo and in vitro ultimately localized the primary homeostatic osmosensory area to the OVLT, a small circumventricular organ that borders the dorsal aspect of the preoptic recess of the third ventricle (McKinley et al., 1982, Thrasher et al., 1982a, Sladek and Johnson, 1983). Indeed, damage to brain areas encompassing the OVLT in humans causes diabetes insipidus, due to an absence of osmotically regulated vasopressin release and chronic hypernatremia due to adipsia (Rudelli and Deck, 1979, Kubota et al., 1991) OVLT neurons are intrinsically osmosensitive Previous studies have shown that infusion of hypertonic solutions into the OVLT area triggers drinking and antidiuresis in awake, water satiated animals (Andersson, 1971, Buggy et al., 1979). Moreover, hypertonic stimuli 60

84 have been shown to activate a proportion of OVLT neurons as revealed by increased c-fos immunolabeling (Oldfield et al., 1991a, McKinley et al., 1994, Xu and Herbert, 1996) and by electrophysiological recordings in in vitro brain slices (Vivas et al., 1990) or explants (Nissen et al., 1993). Although these observations suggest that hypertonicity-induced excitation of OVLT neurons is essential for systemic osmoregulation, definitive evidence indicating that these cells are intrinsically osmosensitive has been lacking. Indeed, although hypertonic NaCl can elicit excitatory responses from OVLT neurons in brain slices even when they are perfused with low Ca 2+ solutions to block synaptic transmission (Vivas et al., 1990), such experiments do not exclude a possible role for substances released from non-neuronal osmosensory cells, or the involvement of Na + receptors rather than osmoreceptors. Indeed, recent studies have demonstrated that osmotically-regulated taurine release by glial cells can contribute to neuronal osmoresponsiveness through the activation of glycine receptors (Hussy, 2002), and subpopulations of neurons in the lamina terminalis have been shown to be specifically sensitive to changes in sodium concentration rather than osmolality (Grob et al., 2004). Our finding that isolated OVLT neurons respond to solutions made hypertonic via the addition of mannitol therefore provide the first direct evidence that a large proportion of these cells are intrinsically osmosensitive. 61

85 Ionic basis for osmoreception in OVLT neurons Several ionic mechanisms could hypothetically mediate an inward current and neuronal excitation under hypertonic conditions. Specifically, hypertonicity could induce membrane depolarization either by inhibiting ion channels that have an equilibrium potential (E rev ) more negative than the resting membrane potential (Abe and Ogata, 1982), or by opening channels that have an E rev more positive than the resting potential (Oliet and Bourque, 1993a). The first mechanism can be excluded here since depolarizing osmoreceptor potentials were associated with channel opening, as indicated by the increased membrane conductance observed under hypertonic conditions. Moreover, under our recording conditions the Nernst potentials for K + (-88 mv) and Cl - (-125 mv) ions were both more negative than the resting potential. Thus, channels selectively permeable to either of these ions could not have mediated osmosensory transduction. Since the E rev of the osmotically evoked current lies between the equilibrium potentials for K + and Na + (+84 mv) ions, it is likely that nonselective cation channels mediate osmoreception in OVLT neurons. In support of this hypothesis, bath application of Gd 3+, a blocker of nonselective cation channels (Hase et al., 1995, Cho et al., 2002), suppressed the effects of hypertonic stimulation on OVLT neurons. 62

86 Expression of the Trpv1 gene is required for osmoreception in OVLT neurons Recently, the search for molecular components responsible for osmotic detection has brought attention to the TRPV family of cationic channels (Liedtke and Kim, 2005, O'Neil and Heller, 2005). For example, mice lacking Trpv1 (Naeini et al., 2006) or Trpv4 (Liedtke and Friedman, 2003, Mizuno et al., 2003) have been found to display impaired osmoregulatory responses to systemic hypertonicity. Moreover, hypoosmotic solutions can activate heterologously expressed channels comprised of homomultimers of TRPV4 (Liedtke et al., 2000, Strotmann et al., 2000, Wissenbach et al., 2000). Although heterologously expressed homomultimers comprising the native capsaicin receptor (i.e. TRPV1) do not respond to hypotonicity (Liedtke et al., 2000), a recent study has suggested that a capsaicin-insensitive amino-terminal splice variant of TRPV1 may contribute to the formation of a hypertonicity-activated cation channel in vasopressin releasing neurons of the supraoptic nucleus (Sharif Naeini et al., 2006). Since the cation channel mediating osmosensory transduction in OVLT neurons is also activated by hypertonicity, we examined if responses to hypertonicity were affected in Trpv1 / mice. Our findings show unequivocally that a product of the Trpv1 gene is required for osmosensitivity in the primary osmosensory neurons of the OVLT. 63

87 Hypothetically, the osmosensory deficit of OVLT neurons in Trpv1 / mice could be due to a developmental defect. However, Trpv1 / mice display no gross neuroanatomical abnormalities and the density of cells expressing the neuronal marker NeuN was equivalent in the OVLT of WT and Trpv1 / animals. A more likely possibility is that a product of the Trpv1 gene plays a functional role as a pore-forming channel in the osmosensory transduction complex of OVLT neurons. In support of this hypothesis we found that the osmosensory transduction current can be blocked by RR, a generic blocker of TRPV channels (Watanabe et al., 2003). Although additional work will be required to determine the precise molecular nature of the osmosensory transduction channel in OVLT neurons, preliminary analysis indicates that these cells are insensitive to capsaicin (Ciura & Bourque, unpublished), suggesting that a splice variant of TRPV1 may contribute to the osmosensitivity of these cells, as recently shown for supraoptic nucleus neurons (Sharif Naeini et al., 2006) Impaired osmotic thirst in Trpv1 / mice The involvement of the trpv1 gene in systemic osmoregulation was first indicated in a report showing that Trpv1 / mice display a significantly elevated basal plasma osmolality, a lack of intrinsic osmosensitivity in supraoptic nucleus neurons, and impaired vasopressin release in response to a systemic hypertonic challenge (Sharif Naeini et al., 2006). Although the 64

88 intrinsic osmosensitivity of supraoptic nucleus neurons is important for the osmotic control of vasopressin release, these cells do not share osmosensory information with other parts of the brain and are unlikely to modulate other CNS-dependent osmoregulatory responses because they project exclusively to the posterior pituitary (Bourque et al., 1994). In contrast, OVLT neurons are extensively interconnected with other nuclei implicated in the control of hydromineral balance and their activation is believed to be a key step in the activation and coordination of most osmoregulatory responses, including thirst (Thrasher, 1989, McKinley et al., 2004b). Since a product of the Trpv1 gene is required for the intrinsic osmosensitivity of OVLT neurons, we compared the effects of a systemic hypertonic challenge on the drinking responses of WT and Trpv1 / mice. Mutant mice showed a significantly attenuated response when compared to WT counterparts, confirming a role for the Trpv1 gene in osmotically driven thirst. Interestingly, intraperitoneal injections of hypertonic saline still evoked a significant drinking response in the Trpv1 / animals. This residual response could have been mediated by a variety of factors. First, although hypertonic stimuli failed to excite isolated OVLT neurons, and OVLT neurons in hypothalamic explants, it is possible that additional mechanisms contribute to the osmotic activation of these cells in vivo, and thus mediate partial drinking responses in Trpv1 / mice. A second possibility is that CNS areas other than the OVLT can contribute to the osmotic modulation of thirst. Indeed, osmosensitive and Na + -sensitive neurons have been found in the subfornical 65

89 organ (Anderson et al., 2000) and in the median preoptic nucleus (McAllen et al., 1990, Travis and Johnson, 1993, Grob et al., 2004), areas known to be part of the integrated neural circuit that underlies osmoregulation (Johnson and Gross, 1993). Finally, it is possible that our protocol did not lead to a pure osmotic stimulus and that additional (i.e. non-osmotic) factors contributed to the thirst response. For example, intraperitoneal injection of hypertonic saline causes water to shift from the vascular compartment to the peritoneal cavity, it is therefore possible that the hypovolemia caused by such a stimulus could have induced part of the thirst response (Johnson and Thunhorst, 1997). Although further studies will be required to define the mechanisms responsible for the residual drinking response in Trpv1 / mice (and possibly for the non-lethality of this mutation), our data establish that the Trpv1 gene is required for osmosensory transduction in OVLT neurons, and provide support for the role of this process in the CNS control of systemic osmoregulation. 66

90 Figure 2.1. Immunolocalization of TRPV1 in OVLT neurons. A, Low magnification micrographs taken through the area of the OVLT (Bar, 50 μm) showing the expression of the neuronal marker NeuN (left panels), TRPV1 (middle panels) and colocalization (right panels), in WT (upper) and Trpv1 / mice (lower). Note the lack of specific TRPV1 signal in Trpv1 / mice. Also, note that neuronal density is not decreased in the OVLT of Trpv1 / mice. B, Higher magnification micrographs (Bar, 50 μm) showing expression of TRPV1 in a majority of WT OVLT neurons. 67

91 Figure 2.2. Extracellular recordings from OVLT neurons in hypothalamic explants. A, Representative examples of extracellular recordings obtained from different osmoresponsive OVLT neurons in acute hypothalamic explants prepared from WT mice. Note the reversible increases in firing rate observed in response to hypertonic stimulation (bar). B, Scatter plot analysis of the changes in firing rate evoked by hypertonic stimuli in the whole population of WT OVLT neurons sampled. For each trial two points are plotted; one corresponding to the basal state (i.e. ΔHz = 0 at 312 ± 2 mosmol.kg -1, and 68

92 one plotting the change in rate observed in response to the particular osmotic stimulus). The dashed line is a linear regression fit to the data points. Note that many but not all cells are osmoresponsive. C, Representative examples of extracellular recordings obtained from different OVLT neurons during osmotic stimulation (bars) in explants prepared from Trpv1 / mice. D, Scatter plot analysis of the changes in firing rate evoked by hypertonic stimuli in all of the Trpv1 / OVLT neurons sampled. In this case the slope of the linear regression fit (dashed line) is not significantly different from zero, indicating a lack of osmoresponsiveness. 69

93 Figure 2.3. Current-clamp analysis of osmotically stimulated OVLT neurons. A, Representative examples of current clamp recordings from osmotically stimulated (bar) OVLT neurons acutely isolated from WT (left panels) and Trpv1 / mice (right panels). Note that hypertonic stimuli consistently depolarize and excite WT but not Trpv1 / neurons. B, Bar graphs show the mean±s.e. membrane potential (V m ) observed under control and osmotically stimulated conditions in OVLT neurons from WT and Trpv1 / mice. Note that no significant depolarization was observed in neurons from Trpv1 / mice. C, Bar graphs show the mean±s.e. changes in firing rate observed under control and osmotically stimulated conditions in OVLT neurons from WT and Trpv1 / mice. Note the absence of osmotically-induced excitation in neurons from Trpv1 / mice. * indicates p<

94 Figure 2.4. Voltage clamp analysis of osmotically stimulated OVLT neurons. A, Representative example of a whole cell voltage clamp recording obtained from a WT OVLT neuron (left). Lower trace shows current responses to voltage ramp commands (27mV/s; upper) from -100 to -10mV applied every 15 sec. before, during (bar) and after the delivery of an osmotic stimulus. Note the appearance of a reversible inward current (compared to dotted line) in response to the hypertonic stimulus. Right hand panel shows steady-state current-voltage (I-V) relations obtained by averaging current responses to 71

95 three consecutive voltage ramps applied under control conditions (1) and during osmotic stimulation (2), as indicated on the left-hand trace. Note the increase in slope of the I-V relation under hypertonic conditions. B, A representative example of a similar experiment performed on an OVLT neuron from a Trpv1 / mouse. Note the absence of a change in holding current (left) or slope of the I-V relation (right) during hypertonic stimulation. C, Bar graphs show the mean±s.e. changes in membrane conductance (slope of I-V relations) observed under control and osmotically stimulated conditions in OVLT neurons from WT and Trpv1 / mice. Note the absence of osmotically-induced conductance increase in neurons from Trpv1 / mice. * indicates p< D, Bar graphs showing the mean±s.e. maximum percent changes in cell volume observed during hypertonic stimulation in isolated OVLT neurons from WT and Trpv1 / mice. 72

96 Figure 2.5. Ruthenium Red blocks osmosensory transduction in WT OVLT neurons. A, Top panel shows steady-state I-V relations recorded from an isolated WT OVLT neuron under control and hypertonic (+20 mosmol.kg -1 ) conditions. Lower panel shows I-V relations recorded under the same conditions, from the same cell, but in the presence of 10μM Ruthenium Red (RR). Note that RR blocks the conductance increase induced by hypertonicity. B, Bar graphs show mean±s.e. values of membrane conductance observed before and during hypertonic stimulation in absence (HEPES, left) and presence of RR (HEPES+RR, right). * indicates p<

97 A B Figure 2.6. Water intake test in WT and Trpv1 -/- mice. A. Schematic representation of the timeline of the experiment. Age-matched WT and Trpv1 - /- were injected intraperitoneally with hypertonic saline (1M NaCl, 10μl/g body weight). The mice were allowed to recover with no access to food or water for 30 min. Water was reintroduced at this point and the amount drank in the following 30 minutes was measured. B. As a control, WT and Trpv1 -/- were injected with the same volume of isotonic saline to show that both strains had equivalent water intake under normal conditions. When challenged with systemic hypertonicity, Trpv1 -/- mice drank significantly less than their WT counterparts in the 30 min following the injection. 74

98 CHAPTER THREE: OSOMOSENSORY TRANSDUCTION IN OVLT NEURONS INVOLVES A TRPV1-, BUT NOT TRPV4-, DEPENDENT MECHANOSENSORY EVENT 3.1. Overview In chapter 3 we have shown that OVLT neurons are intrinsically osmosensitive. More specifically, hypertonic stimulation of isolated OVLT neurons activates a Trpv1-dependent, RR-sensitive cation conductance. It remains unclear however if the osmosensory transduction channel is mechanically gated and if additional channels, namely TRPV4, participates in this transduction. As described in section , from the earliest studies on central osmoreception it became clear that eliciting a response to osmotic pressure requires cellular dehydration (i.e. establishing an osmotic pressure across the membrane resulting in water exiting the cell). Such an observation led to the hypothesis that the mechanical shrinkage of the cell in a hypertonic solution triggers the electrical response recorded in osmosensory neurons, as is the case with supraoptic neurons [section ]. However, in addition to the mechanical changes taking place during exposure to a hyperosmotic environment, cellular dehydration also results in an obligatory increase in the ionic concentration inside the cell as a result of water efflux. This increase in ionic strength could have an impact on the activity of a number of cellular processes and could thus be the actual stimulus gating the osmosensory transduction channel. 75

99 Two distinct channel proteins have been reported as candidate molecular sensors of osmotic stimuli in the OVLT, TRPV1 and TRPV4. As reported in the previous chapter, TRPV1 is required for the transduction of osmotic stimuli in osmosensory OVLT neurons (Ciura and Bourque, 2006), as mice lacking TRPV1 showed impaired responses to hypotonicity. Previous studies have also shown that TRPV4 KO (Trpv4 -/- ) mice have impaired systemic osmoregulation which has been associated with decreased neuronal activation in the OVLT in conditions of systemic hypertonicity. These results have prompted the hypothesis that TRPV4 acts as the osmosensor in OVLT neurons. Thus, it is conceivable that TRPV4 participates in osmosensory transduction either as a heteromeric partner of TRPV1 (Hellwig et al., 2005) or independently. In this chapter we describe the electrophysiological steps involved in hypertonicity sensing in the OVLT and we show that osmotransduction involves a mechanosensory process. Furthermore, we investigate the role of TRPV1 and TRPV4 in mediating this transduction MATERIALS AND METHODS Explant recordings All animal procedures were approved by the McGill University Animal Care Committee. Mice, 6-8 wks of age, were anaesthetized, decapitated and the brains removed. The explants were prepared as described elsewhere (Ciura and Bourque, 2006). Briefly, a block of tissue mm containing 76

100 the optic chiasm and the OVLT was cut out of the brain using razor blades and pinned, ventral surface up, to the Sylgard base of a superfusion chamber. Within 3 5 min from decapitation, the excised OVLT explant was superfused (at 1.5 ml/min at 32 to 33 C) with carbogenated (95% oxygen; 5% CO 2 ) artificial cerebrospinal fluid (ACSF) delivered via a Tygon tube placed over the rostral pole of the preparation. The optic nerves were removed to provide direct access to the OVLT, and a cotton wick was placed at the caudal tip of the explant to minimize the level of fluid over the preparation and to facilitate the removal of ACSF. The ACSF (ph 7.4) comprised (in mm): NaCl (125), NaHCO 3 (26), KCl (5), MgCl 2 (1.5), CaCl 2 (1), D-glucose (10). Mannitol was added to adjust the basal osmolality of the solution to 312 ± 2 mosmol/kg, or to make hypertonic solutions. Extracellular recordings were made from OVLT neurons using micro-electrodes filled with 1M potassium acetate (resistance 5 to 10 MΩ) and hyperosmotic stimuli were bath-applied by switching the solution being delivered through the Tygon tube. For the analysis of changes in firing rate, the average rate of spike discharge of the neuron measured during the final minute of an osmotic stimulus was subtracted from the average firing rate observed prior to the onset of the stimulus Isolated cell recordings To obtain acutely dissociated OVLT cells, blocks of brain tissue containing the OVLT were removed from C57/BL or Trpv1 / adult mice from 77

101 the same background and were incubated in oxygenated PIPES solution (120 mm NaCl, 5 mm KCl, 1 mm MgCl 2, 20 mm PIPES, 1 mm CaCl 2, 10 mm D-Glucose; ph 7.3) containing 5mg/ml Protease X and XIV at room temperature for 30 min after which they were washed in PIPES solution without protease. Individual tissue blocks in PIPES solution were triturated and the resulting suspension was plated onto 30mm cell culture dishes (Falcon). Cells were perfused with HEPES solution (150 mm NaCl, 3 mm KCl, 1 mm MgCl 2, 10 mm HEPES) during the experiments. The internal solution of the recording electrode contained 120 mm K-Gluconate, 1 mm MgCl 2, 1 mm EGTA, 4 mm ATP, 1 mm GTP, 14 mm Phosphocreatine, 10 mm HEPES (ph 7.25). Cells were patch-clamped in the whole cell mode and imaged every 15 second for offline analysis. In the voltage-clamp experiments 0.5 μm tetrodotoxin (TTX) was added to the HEPES solution. SB was added where indicated as a 1:5000 dilution from a stock of 15 mm (in DMSO) to a final concentration of 3 μm. Where indicated, RR was added at a final concentration of 10 μm. Gd + was added at a final concentration of 300 μm where described. Hyperosmotic stimulation was delivered by switching the perfusion solution to hyperosmotic HEPES (adjusted with mannitol everywhere) using a piezoelectric stepper device. Osmolalities of all solutions were measured using a freezing-point micro-osmometer. For the ion substitution experiments the solutions were as follows. For low Na +, the external solution contained 50 mm NaCl, 3 mm KCl, 1 mm MgCl 2, 10 mm HEPES. For the High K +, the external solutions 150 mm 78

102 NaCl, 6 mm KCl, 1 mm MgCl 2, 10 mm HEPES. For Ca 2+ only, the internal solution contained: 130 mm CsCl, 10 mm NaCl, 1 mm EGTA, 0.1 mm CaCl 2, 10 mm HEPES (ph 7.25) and the external: 70 mm CaCl 2, 6 mm Ca(OH) 2, 10 mm Glucose, 20 mm HEPES (ph 7.25). The low Cl - experiment was performed with the external solution 100 mm Na-Gluconate, 50 mm NaCl, 1 mm MgCl 2, 3 mm KCl, 10 mm HEPES Cell-attached single channel recordings Isolated cells obtained as described above were patched with polished glass electrodes having a resistance of 9-10 MΩ. The internal solution of the patch electrode consisted of (in mm): NaCl 140, KCl 3, HEPES 10, EDTA 0.5, TEA 6, 4-AP 4, CsCl 5, TTX 0.5x10-6. Currents were captured and analyzed using pclamp-9 (Axon Corp.) Volume measurements We quantified percent changes in cell volume from measurements of maximal cross-sectional area (CSA, from digitized video images) of the cell using Scion Image for Windows 4.02 (Scion Corp., Frederick MA) as performed elsewhere (Oliet and Bourque, 1993a). All values of CSA measured in the control period were averaged (CSA o ) and values of normalized volume at individual time points (nv t ) were calculated from the CSA value at that time point (CSA t ) using the equation; nv t = [(CSA t ) 1.5 /(CSA o ) 1.5 ]. Percent changes in volume (ΔV%) were calculated as ΔV%=(1-nV t )*

103 3.3. RESULTS OVLT neurons are specialized osmosensors To establish if OVLT neurons are specialized for the intrinsic detection of hyperosmotic conditions we first used whole cell current clamp recordings to examine the effects of raising the osmolality of the perfusing solution on neurons acutely isolated from the OVLT or adjacent piriform cortex (PFC) of adult C57/BL mice (Figure 3.1A). As illustrated in Figure 3.1B, hyperosmotic conditions (+25 mosmol/kg applied for seconds) caused a progressive and sustained depolarization and excitation in most of the OVLT neurons tested (22 out of 36), whereas PFC neurons were unaffected by equivalent stimuli. Indeed, the average membrane potential and firing rate of OVLT neurons (>10% change) exposed to a +25 mosmol/kg stimulus were significantly increased (+7.30 ± 1.71 mv and ± 0.40 Hz; n =36 ; P = 0.002), whereas those of PFC neurons were not (+0.46 ± 0.07 mv and ± 0.51 Hz; n =11 ; P = 0.97; Figure 3.1C). A majority of OVLT neurons are therefore specifically and intrinsically sensitive to hyperosmotic conditions. channels Hyperosmolality activates calcium-permeable cation Previous work has shown that hyperosmotic conditions activate a ruthenium red (RR)-sensitive membrane conductance in OVLT neurons and 80

104 the equilibrium potential of the current induced was consistent with the involvement of a putative non selective cation current [see chapter 2; (Ciura and Bourque, 2006)]. However it remains unknown if these channels can also flux either calcium or chloride ions. We therefore measured the reversal potential of the hypertonicity-induced current (+ 25 mm mannitol) by recording current-voltage (I-V) relations during whole cell voltage clamp recordings from isolated OVLT neurons bathed in solutions of different ionic composition (Figure 3.2A). When tested under normal conditions (with 500 nm TTX to block voltage gated Na + channels) the mean reversal potential (E REV ) of the osmosensory current was ± 1.9 mv (n =20; Figure 2B,C). As shown in Figure 3.2(B,C), this value was significantly more negative in low [Na + ] solution ( ± 5.2 mv; n = 5 ; P <0.05) and significantly more positive in high [K + ] solutions ( ± 2.2 mv; n = 7 ; P <0.05, one-way Anova). In contrast, E REV was not significantly different in low [Cl - ] solution ( ± 2.9 mv; n = 7; P >0.05). These results indicate that the channels are permeable to cations, but not anions. To determine if the channels are calcium permeable, we examined the effects of hypertonicity in cells bathed in an external solution comprising only Ca 2+ as a cation charge carrier. As shown in Figure 3.2(B,C), inward currents and increases in membrane conductance were readily detected in these conditions, indicating that Ca 2+ permeates osmosensory transduction channels in OVLT neurons. Finally, we examined the effects of Gd 3+, a non-specific blocker of many subtypes of non-selective cation channels (Yang and Sachs, 1989) and ruthenium red (RR), a broad 81

105 spectrum inhibitor of TRPV channels (Watanabe et al., 2002). As shown in Figure 3.2(B-D), OVLT neurons showed no significant increase in membrane conductance when tested in the presence of either Gd 3+ (300 μm) or RR (10 μm). These data are therefore consistent with the possible involvement of TRPV channels in osmosensory transduction by OVLT neurons Hyperosmotic stimuli activate a 32 ps cation channel While our whole cell recording experiments indicate that osmosensory transduction involves the activation of a non-selective cation conductance, it is not clear if this macroscopic response reflects the involvement of a single type of ion channel with mixed cation permeability, or the simultaneous activation of distinct populations of channels permeable to different ion species. To address this question, we performed cell-attached single channel recordings in isolated OVLT neurons. The solution filling the recording pipette was equivalent to the normal extracellular solution used for whole cell recording, but also included blockers of other types of ion channels that are not involved in generating the osmosensory response. The solution included 0.5 μm TTX to block voltage-gated Na + channels, 1 mm 4-aminopyridine and 10 mm tetraethylammonium to block voltage-gated K + channels, 5 mm CsCl to block hyperpolarization-activated cation channels and 0.1 μm apamin to block Ca 2+ -activated K + channels. Under these conditions, three types of spontaneous unitary inward currents were observed at a trans-patch potential 82

106 of -60 mv. I-V analysis of open channel currents confirmed the existence of three subtypes of channels displaying unitary slope conductances of 18, 32 and 55 ps. Interestingly, hyperosmotic stimulation (25 mm mannitol added to the bath) had no effect on the probability of opening (P O ) of the 18 ps (n = 5 ; P= 0.80) and 55 ps (n = 33 ; P = 0.55) channels. However the P 0 of 32 ps channels was found to increase by more than 15% in 17 out of 24 patches (Figure 3.3 A,B). The P 0 of the responsive channels was found to increase from 3.2x10-4 ± 1.3 x10-4 to 6.7 x10-4 ± 1.9 x10-4 (P=0.004). As shown in Figure 3.3C, the average E REV for this channel was -30.6±0.07 mv, a value corresponding to the E REV of the macroscopic current observed during whole cell recording (e.g. Figure 3.2A). The non-selective cation current induced by hyperosmotic stimulation in OVLT neurons is therefore generated by a population of 32 ps channels with mixed permeability to Na + and K + ions Hypertonicity causes osmometric volume changes in OVLT neurons Although previous work suggests that osmosensory transduction requires "cellular dehydration" (sections and ), the activation of cation channels in OVLT neurons could be due either to a mechanical effect associated with hypertonicity-induced shrinking, or to the change in ionic strength or solute concentration that occurs in concert with osmotic water extraction from the intracellular compartment. In principle the involvement of a 83

107 mechanical effect would seem unlikely since most cells reportedly undergo regulatory volume decrease (RVD) or increase (RVI) within 2-5 minutes of swelling or shrinking (Lang et al., 1998). We therefore examined the effects of osmotic stimulation on cell volume in isolated OVLT neurons. Surprisingly, OVLT neurons were found to display sustained and fully reversible decreases in nv when exposed to hyperosmotic stimuli lasting as long as 60 minutes (Figure 3.4A). Even though the changes in osmolality were quasiinstantaneous (i.e. <20 ms when delivered via a piezoelectric stepper), the observed changes in nv occurred gradually (τ = ~100 s), indicating that the rate of water efflux is relatively slow in these cells, or that the cytoskeleton can temporarily oppose osmotic forces during the early stages of a volume decrease. Finally as shown in Figure 3.4B, steady-state nv was found to vary as an inverse function of test osmolality (π) and could be accurately described by the Boyle Van't Hoff equation: nv = [(1/π)(π 0 (1 b)]+b where b is the osmotically inactive fraction of the cell and π 0 is the osmotic set point (i.e. 312 mosmol/kg). These findings indicate that changes in cell volume provide an accurate physical representation of the osmotic stimulus and could therefore be involved in the osmometric activation of cation channels in OVLT neurons. 84

108 Hypertonicity-sensing by OVLT neurons is a mechanical process To determine if the osmotic activation of cation channels is due to a concentration of intracellular solutes or to a mechanical effect we examined the impact of reducing cell volume by applying negative pressure to the inside of the recording pipette during whole cell patch clamp recording. For this purpose I-V relations were generated from a holding voltage of -60 mv by applying slow voltage ramps (from -100 to 0 mv; 20 mv/s) at regular intervals (10-30 s) during the experiment and the slope of the I-V plot between -80 and -60 mv was taken as a measure of membrane conductance (G). Values of G measured from I-Vs taken before the stimulus were averaged and taken as control conductance (G 0 ) and changes in conductance (ΔG) during subsequent trials were calculated by subtracting G 0 from G measured at various time points. We found that values of G increased in proportion with the degree of cellular shrinkage (Figure 3.5A,B) whether the volume change was induced by osmotic or mechanical means. As quantified in Figure 3.5C,D, osmotically- and mechanically-induced decreases in cell volume caused equivalent and proportional increases in G. We hypothesized that if the cation channel is gated by a mechanical effect, this response to hypertonicity should be reversed when cell volume was restored by applying positive pressure to the recording electrode (Figure 3.6A). As illustrated in Figure 3.6B and quantified in Figure 3.6C, increases in G associated with hypertonicity-induced decreases in nv could be reversed 85

109 by mechanical inflation of the cell in the continued presence of the osmotic stimulus Osmotic and mechanical transduction are quantitatively equivalent In order to firmly establish the equivalence of the neuronal responses to osmotic and mechanical stimulation, we performed a quantitative comparison of the transduction events coupling volume changes to the generation of a transduction current (I T ). Values of ΔG were computed as described above and the net shift in holding current measured at -60 mv (relative to control) was taken as a measure of I T. As shown in Figure 3.7 (A,B), changes in G observed during cell shrinking were proportionally related to the magnitude of the normalized volume decrease (nδv = 1 - nv) observed at corresponding time points whether they were induced by a hyperosmotic stimulus or by applying negative pressure to the recording pipette. Statistical analysis revealed that the slopes of linear regressions fitted through both data sets were not significantly different (osmotic 5±0.01 ns/nδv vs mechanical 4±0.01 ns/nδv ; P=0.54). Similarly, the magnitude of I T observed at various time points was linearly related to the underlying conductance change (ΔG) and the slopes of linear relations were not significantly different in cells 86

110 challenged osmotically (20.8±2.9 pa/ns) and mechanically (25.5±1.3 pa/ns; P = 0.16) Spike encoding is equivalent during osmotic and mechanical transduction Changes in membrane potential and firing caused by changes in I T could be strongly influenced by the presence of voltage-gated conductances at near-threshold potentials, and the impact of these could potentially differ when nv is modified by osmotic or mechanical stimuli. We therefore examined the magnitude of the depolarizing osmoreceptor potential (ΔV T ) associated with increases in I Τ by tracking the voltage at which zero current was observed in I-V relations collected at various time points. As shown in Figure 3.8A, ΔV T increased in proportion with I T, and rose asymptotically toward the maximal voltage change predicted by the driving force of the cation current. Values of I T and corresponding ΔV T derived from osmotically stimulated and mechanically stimulated cells were equally well described by the equation ΔV T = DF.ΔG/(G 0 +ΔG), where DF is the driving force (holding potential - E REV ) and G 0 is the conductance of the cell under resting conditions. We next examined if the relation between ΔV T and action potential firing rate (i.e. spike encoding) was different during osmotic and mechanical stimulation of OVLT neurons recorded under whole cell current-clamp. As illustrated in Figure 3.8B, an equivalent and proportional relation was 87

111 observed between firing rate and ΔV T when the latter were provoked by either type of stimulus. Indeed, the slopes of linear relations were not significantly different in cells challenged osmotically (0.31±0.03 pa/ns) and mechanically (0.34±0.05 pa/ns; P = 0.66) Osmo-mechanical transduction requires expression of Trpv1 and not Trpv4 We previously reported that osmosensory transduction in OVLT neurons requires expression of the Trpv1 gene (Ciura and Bourque, 2006). However other studies have suggested that the Trpv4 gene may also play a role in this process (section b). To determine if Trpv1 also supports mechanical transduction and to evaluate the possible importance of Trpv4, we examined the responses of OVLT neurons isolated from wild-type (WT) mice, and animals lacking expression of Trpv1 (Trpv1 / ) or Trpv4 (Trpv4 / ). As described in Chapter 2, we found that OVLT neurons isolated from Trpv1 / mice were unable to generate increases in G in response to a +25 mosm stimulation (+0.013±0.04 ns; n = 23; P = 0.73; Figure 3.9A,B). In contrast, OVLT neurons obtained from animals generated significant increases in G (+0.50 ± 0.16 ns; n = 17; P = 0.008) and these changes were equivalent to those observed in WT OVLT neurons (+0.33 ± 0.07 ns; n = 38; comparison with Trpv4 / P = 0.27). Similarly, we found that OVLT neurons isolated from Trpv1 / animals lacked responses to suction-evoked shrinking 88

112 (+5.21x10-3 ± 0.04 ns; n = 12; P = 0.9; Figure 3.9C,D), whereas significant responses were observed in WT (+0.72 ± 0.20 ns; n = 32; P = 0.001) and Trpv4 / animals (+0.43 ± 0.11 ns; n = 13; P = 0.008). To confirm that the loss of osmoresponsiveness in Trpv1 / OVLT neurons is due to the absence of a channel encoded by Trpv1 rather than a secondary effect of gene deletion (e.g. an indirect developmental abnormality), we examined the effect of SB366791, a selective inhibitor of TRPV1 channels on responses recorded in WT OVLT Neurons. As shown in Figure 3.9, addition of 3 μm SB prevented the increase in conductance in WT OVLT neurons evoked by hyperosmolality (+0.12 ± 0.06 ns; n = 19; P = 0.06) or by suction (+0.07 ± 0.12 ns; n = 13; P = 0.55). Osmotically and mechanically-evoked decreases in cell volume were equivalent in all groups (data not shown) Hypertonicity sensing in hypothalamic explants requires Trpv1 not Trpv4 The data reported above indicate that expression of Trpv1 but not Trpv4 is required for cell-autonomous hypertonicity sensing in OVLT neurons. However detection of osmotic stimuli in situ can potentially involve a number of extrinsic mechanisms (see section 1.4.5). We therefore examined the effects of hyperosmotic stimuli on the action potential firing rate of OVLT neurons detected by non-invasive single unit extracellular recordings in superfused explants of mouse hypothalamus. As previously reported in 89

113 chapter 2, we found that OVLT neurons recorded in explants from Trpv1 / animals were unaffected by hyperosmotic conditions (0.003 ± Hz/mOsm; n = 32, P = 0.31). However as shown in Figure 3.10, OVLT neurons recorded from explants obtained from Trpv4 / mice were strongly affected by hyperosmotic conditions and the sensitivity of these cells measured by linear regression (0.027 ± Hz/mOsm; n = 61; P < ) was not significantly different than that of WT neurons (0.038 ± Hz/mOsm; n = 99; P < ; statistical comparison with Trpv4 / ; P = 0.22) DISCUSSION Previous work has established that neurons in the OVLT encode hyperosmotic stimuli as proportional increases in action potential firing rate [chapter 2; (Ciura and Bourque, 2006)] and that activation of these neurons is required to elicit appropriate osmoregulatory responses such as thirst, natriuresis and antidiuretic hormone release ( section 1.4.2). In this study we reveal that the intrinsic osmotic excitation of OVLT neurons is a specific property of these cells that reflects the cell-autonomous and mechanical modulation of a non-selective cation channels encoded by the Trpv1 gene (Ciura and Bourque, 2006). 90

114 OVLT neurons display osmometry Early physiological studies indicated that systemic hyperosmotic conditions are detected within the brain (Verney, 1947, Jewell and Verney, 1957, Andersson, 1971) through a process that involves cellular dehydration. Since osmoregulatory responses do not adapt as a function of time (Verney, 1947, Verbalis and Dohanics, 1991) and rise in proportion with extracellular fluid osmolality (Walters and Hatton, 1974) it is presumed that the cells mediating osmosensory transduction lack the mechanisms that normally support volume regulation (Strange, 2004, Lang, 2007). Here we show that neurons isolated from the mouse OVLT show sustained and graded decreases in volume in response to hypertonic stimuli lasting up to one hour. The basis for the apparent lack of volume regulation in response to mild osmotic stimulation (<+40 mosmol/kg) remains to be determined. Interestingly, the steady state amplitude of cell volume in solutions of varying tonicity could be described by the Boyle van't Hoff equation, indicating that OVLT neurons perform as near-perfect osmometers Hypertonicity-sensing involves mechanotransduction The observation that OVLT neurons display osmometry is consistent with the possibility that these cells use a mechanical process to transduce osmotic perturbations. However, it does not exclude the possibility that the obligatory changes in ionic strength that accompany osmotically-induced 91

115 volume changes participate or even mediate this process. Our results provide two lines of evidence indicating that osmosensory responses are strictly mediated by the mechanical aspect of volume change. First, when suction was used to decrease cell volume in the absence of a change in ionic strength the neuronal responses obtained were equivalent to those caused by hypertonicity-induced shrinking. Indeed, the analysis shown in Figures 3.7 and 3.8 revealed that the individual steps coupling volume decreases to channel activation (increased G), transduction current, membrane depolarization, and increased firing rate were all quantitatively equivalent. Second, neuronal responses induced by hypertonicity-evoked shrinking could be reversed by restoring cell volume via application of positive pressure to the recording pipette. These results provide strong evidence indicating that the mechanical effect of the osmotically-induced volume change is both necessary and sufficient to induce the excitatory effect of hypertonicity in OVLT neurons Osmosensory transduction involves a 32 ps calciumpermeable cation channel Previous studies indicated that hypertonicity-sensing by OVLT neurons involves the activation of a non-selective cation conductance that generates a net inward current and membrane depolarization at resting potential (Ciura and Bourque, 2006). However it remained unclear if this osmotically-induced 92

116 non-selective conductance was due to the simultaneous activation of various subtypes of ion channels, or that of a single population of channels with mixed permeability to various cations. Our experiments provide clear evidence that the macroscopic osmosensory current is mediated by the flux of Na +, K + and Ca 2+ ions, but not Cl - ions. Moreover, single channel recordings established that hypertonicity increases the P O of a single population of ion channels characterized by a unitary conductance (γ) of 32 ps. Unfortunately, the extremely low probability of opening of the channels observed under our conditions (<0.001) precluded a more extensive analysis of the gating properties of these channels. Nonetheless, the reversal potential of the open channel current of the channel was in good agreement with that of the macroscopic current observed during whole cell recording. Thus osmosensory responses are mediated by channels that can flux Na +, K + and Ca 2+ ions, rather than by distinct subsets of Na +, K + and Ca 2+ selective channels Osmosensory channels require expression of Trpv1 but not Trpv4 We previously reported that osmosensory transduction is lost in OVLT neurons isolated from the OVLT of Trpv1 / mice and that OVLT neurons from WT animals are insensitive to capsaicin (Ciura and Bourque, 2006), a highly potent agonist of the full length TRPV1 channel (Caterina et al., 1997a). One 93

117 possibility is that OVLT neurons, like osmosensitive neurons in the supraoptic nucleus (Sharif Naeini et al., 2006), express a variant of the TRPV1 channel that enables osmosensing but not capsaicin responsiveness. Direct evidence that a channel encoded by Trpv1 mediates osmosensing in either type of neuron has yet to be obtained. Interestingly, previous work has shown that Trpv4 is also expressed in the OVLT (Liedtke et al., 2000), where it contributes to hypertonicity-induced increases in expression of the activitydependent immediate early gene c-fos (Liedtke and Friedman, 2003). It therefore remained unclear if osmosensory channels in mouse OVLT are encoded by either Trpv1 or Trpv4. To address this question, we examined the impact of deleting expression of these genes on osmotic and mechanical transduction in isolated OVLT neurons. Our results show that wild type-like responses to cell shrinking induced by hypertonicity or pipette suction are abrogated in OVLT neurons obtained from Trpv1 / animals, but are unaffected in those obtained from Trpv4 / mice. Moreover, osmosensory responses recorded from WT OVLT neurons are eliminated by acute exposure to SB366791, a selective inhibitor of current flux through TRPV1 channels (Gunthorpe et al., 2004). Taken together, these results indicate Trpv1 encodes a mechanosensitive channel which confers osmotic sensitivity to OVLT neurons. 94

118 Mechanism of osmotic gating of the transduction channel Exactly how a Trpv1-encoded channel could be gated by mechanical disturbances remains to be determined. In essence, mechanosensory transduction in mammalian cells could involve the gating of a channel either indirectly, for example through an enzymatic process linked to a mechanosensory unit (which is not the channel itself), or through forces applied directly to the channel protein (e.g. via changes in the local lipidchannel interface or links to the cytoskeleton) (Sukharev and Corey, 2004). Recent work has shown that the actin cytoskeleton is required for the Trpv1- dependent osmosensory response of supraoptic nucleus neurons (Zhang et al., 2007, Zhang and Bourque, 2008). Although the wild type TRPV1 channel features domains that could mediate interactions with microtubules (Goswami et al., 2004) or actin (Caterina and Julius, 2001), it remains unclear if supraoptic neurons use actin as a force-mediating channel tether or simply to support the plasma membrane. The potential involvement of cytoskeletal elements in the osmotic and mechanical gating of OVLT neurons remains to be established Trpv4 does not contribute to hypertonicity sensing by OVLT neurons Although loss of Trpv4 does not prevent cell-autonomous hypertonicity sensing by isolated OVLT neurons, there is compelling evidence suggesting 95

119 that this gene participates in systemic osmoregulation (Mizuno et al., 2003) and in the expression of c-fos by OVLT neurons following hyperosmotic stimulation in vivo (Liedtke and Friedman, 2003). One possibility is that the TRPV4 protein, either independently or in a heteromeric complex with TRPV1 protein (Hellwig et al., 2005), contributes to the osmosensory conductance in OVLT neurons and that loss of the former subunit somehow causes a deficit that can be detected in situ, but not in isolated neurons studied in vitro. Another possibility is that TRPV4 mediates a non-cell autonomous form of osmosensing that cannot be observed in isolated neurons. For example, recombinant homomeric TRPV4 channels are activated by hypotonicity (Liedtke et al., 2000, Strotmann et al., 2000) and it has been proposed that these channels could mediate the hyperosmotic activation of OVLT neurons via network-dependent disinhibition (Mizuno et al., 2003). We therefore further investigated the possible contribution of Trpv4 to hypertonicity sensing by OVLT neurons in superfused hypothalamic explants. Although hypertonicity-sensing was abolished in neurons from Trpv1 / animals, excitatory responses to hypertonicity were similar in explants from wild type and Trpv4 / animals. Our results suggest that Trpv4 does not contribute to cell autonomous or non-cell autonomous hypertonicity sensing by OVLT neurons. The basis for the decrease in hypertonicity-mediated c-fos expression observed in Trpv4 / mice in vivo therefore remains to be explained. One possibility is that channels encoded by Trpv4 can mediate a long-range influence that depends on circuitry that is not retained in our 96

120 explant preparations. Indeed, recent work suggests that TRPV4 may be used by peripheral osmoreceptors (McHugh et al., 2010) that could potentially signal to the OVLT via ascending pathways (Bourque, 2008). 97

121 Figure 3.1. Osmosensitivity of acutely isolated OVLT neurons. A, Representative examples of current-clamp responses of isolated OVLT (upper) or piriform cortex (lower) neurons to application of +25 mosm hypertonic stimulus (bar). B,Quantification of firing rate responses during hypertonic stimulation in OVLT and PFC neurons. *<

122 Figure 3.2. Ionic substitution experiments in OVLT neurons. A, Representative traces showing current-voltage relationships of acutely isolated OVLT neurons during isotonic (black trace) and hypertonic stimulation (gray, +25 mosm).b, Mean reversal potentials for the osmotically induced current in experiments where concentration of ions in the solutions have been manipulated as described in Materials and Methods. C, Quantification of conductance changes induced by hypertonic stimulation in control experiments and in the presence of RR or Gd 3+. *shows significant difference from control (p<0.05) 99

123 Figure 3.3. Single channel recording in OVLT neurons. A.Raster plot showing distribution of individual channel (32pS) openings (circles) in responding (filled circles) and non-responding (clear) patches before and after the application of a hypertonic stimulus (+25 mosm). The gray vertical bar marks the start of the stimulus, which lasts for the remainder of the recording. B. Histogram quantifying the frequency of channel openings shown in A. C. Bar graph quantifying the probability of opening of the channel during control and hypertonic conditions. D. I-V plot showing the conductance and the reversal potential of the osmotically activated channel in OVLT neurons. 100

124 Figure 3.4. Osmometry in OVLT neurons. A, Normalized volume decrease caused by the application of a hyperosmotic (+40 mosm) stimulus in acutely isolated OVLT neurons. Each point is the average of volume measurements obtained from 5 cells. B, Normalized steady-state volume responses to various osmotic stimuli of isolated OVLT neurons. 101

125 Figure 3.5. Correlation of conductance and volume responses to hypertonic and mechanical stimuli in OVLT neurons. A, Example of the conductance and volume responses of an OVLT neuron to the application of a hypertonic stimulus (+25 mosm, bar) recorded under V-clamp. B, Examples of the conductance and volume responses to mechanical stimulation (negative pressure). C and D, Scatter plot of all the cells analyzed under these conditions. Note the similar sensitivity to volume change under hypertonic and mechanical stimulations. 102

126 Figure 3.6. Responses to hypertonic stimulation can be reversed with positive pressure. A. Examples of conductance (upper) and volume changes (lower) during stimulation with hypertonic (+25 mosm, red bar) and positive pressure in the recording pipette (blue bar) in an OVLT neuron. B. Examples of I-Vs from the same cell taken during control (black trace), hypertonicallystimulated (red trace) and mechanically-induced volume recovery (blue) parts of the recording. C. Quantification of conductance during these three conditions for all the recordings. 103

127 Figure 3.7. Correlation of volume changes with conductance and current responses. A. Scatter plot showing the correlation of normalized volume changes with the change in conductance (upper plot) and the relation between conductance and current responses (lower plot) during a hypertonic stimulus (+25 mosm). B. Similar relationships between volume and conductance responses (upper plot) or between conductance and current responses (lower plot) are observed during mechanically-induced volume decreases. 104

128 Figure 3.8. Correlation between electrophysiological steps leading to increase in firing rate in OVLT neurons under hypertonic and mechanical stimulation. A. Scatter plot showing the relation between the induced current and membrane depolarization (upper plot) or between depolarization and the rate of action potential generation (lower plot) during stimulation with hypertonic solution (+25 mosm). B. Similar correlations observed during stimulation with negative pressure in the recording pipette. 105

129 Figure 3.9. Voltage-clamp recordings in isolated OVLT neurons. A. representative current-voltage relationships (I-Vs) during control (black) and hypertonic stimulation (gray, +25 mosm) showing increase in the conductance (slope) in OVLT neurons taken from WT (panel 1) and Trpv4 -/- (panel 3). OVLT neurons from Trpv1 -/- (panel 1) and WT neurons treated with SB (panel 4) do not show any significant increase in conductance with hypertonic stimulation. B. Quantification of the conductance change in all four conditions. C. representative I-Vs of OVLT neurons from WT, Trpv1 -/-, Trpv4 -/- and SB treated WT showing conductance responses to mechanical stimulation in these cells. D. Quantification of the responses in C. OVLT neurons from WT and Trpv4 -/- genotypes show significant responses to mechanically-induced shrinking, while the absence of Trpv1 function prevents such responses.*p<

130 Figure Extracellular recordings in OVLT explants. A, Firing rate histograms showing representative responses to application of a hypertonic solution (+35 mosm, bar) recorded in OVLT explants obtained from WT, Trpv1 -/- and Trpv1 -/- mice. B. Scatter plot showing firing rate changes associated with increases in the extracellular osmolality in OVLT explants obtained from WT, Trpv4 -/- and Trpv1 -/- genotypes. Note the absence of osmosensory neurons in the Trpv1 -/- background. 107

131 CHAPTER FOUR: Trpv1 GENE REQUIRED FOR THERMOSENSORY TRANSDUCTION AND ANTICIPATORY SECRETION FROM VASOPRESSIN NEURONS DURING HYPERTHERMIA 4.1. OVERVIEW In the previous chapters we revealed that OVLT neurons are intrinsically osmosensitive and explored the cellular and molecular basis for this property. Our results indicated that hypertonicity-sensing by these cells relies on a mechanism that is equivalent to that used by osmosensory neurons in the SON. However as discussed in section 1.4.6, osmosensory neurons are able to respond to more than one physiological modality and this response is important for triggering anticipatory homeostatic mechanisms of water conservation. The best studied example of this interaction is the ability of both OVLT and SON osmosensory neurons to respond to increases in temperature (Forsling et al., 1976, Silva and Boulant, 1984). Water homeostasis can be dramatically affected by a rise in core body temperature (hyperthermia) as this condition promotes evaporative heat loss through panting, grooming, or sweating in mammals (Shibasaki et al., 2006). Remarkably, the potential impact of this response on body fluid balance is blunted due to a simultaneous increase in the release of vasopressin (VP, antidiuretic hormone) from the neurohypophysis (Segar and Moore, 1968). How this important anticipatory homeostatic response is achieved is not known, but previous work has indicated that a central thermosensory 108

132 mechanism may be required for this effect to occur (Szczepanska-Sadowska, 1974, Forsling et al., 1976). Indeed, VP release can be stimulated by thermode-heating of the hypothalamus (Szczepanska-Sadowska, 1974), and the anticipatory release of this hormone in hyperthermic animals can be blunted by thermode-cooling of this part of the brain (Forsling et al., 1976). Interestingly, the osmosensitivity of OVLT and SON neurons relies on an N- terminal variant of TRPV1 (Chapters 2 & 3) (Ciura and Bourque, 2006, Sharif Naeini et al., 2006). As discussed in section , TRPV1 is a member of tremo-trpv family of heat-gated channels. We therefore examined if expression of the trpv1 gene might render SON neurons intrinsically thermosensitive and be required for VP release during hyperthermia in vivo MATERIALS AND METHODS Preparation and perfusion of isolated neurons For rats, tissue blocks (~1 mm 3 ) comprising the SON and PNZ were isolated from the brains of adult ( g) male Long Evans rats according to a procedure reported previously (Zhang and Bourque, 2006). Blocks were incubated for min in oxygenated PIPES solution (33.5 o C; ph 7.0) containing (in mm): NaCl 120, KCl 3, MgCl 2 1, CaCl 2 1, glucose 20, PIPES 10, and 0.7 mg/ml trypsin, then washed in trypsin-free PIPES (22 o C) solution, triturated and plated on glass coverslips. Cells were recorded within minutes of trituration. For mice [10- to 12-week-old male C57/BL mice and Trpv1 -/- mice (BL126 S4)], blocks containing SON were incubated for 30 min in an oxygenated PIPES solution (23 o C, ph 7.34) containing (mm): NaCl 130, 109

133 KCl 5, MgCl 2 1, CaCl 2 1, glucose 10, PIPES 20, and 0.5 mg/ml 1 protease X and protease XIV. Blocks were then washed in protease-free PIPES solution, triturated and plated on glass coverslips. Coverslips were placed in a Leiden chamber (internal volume 500 μl) perfused with HEPES-buffered saline containing (in mm): NaCl (140) [150 for mice HEPES], KCl 3, MgCl 2 1, CaCl 2 1, glucose 10, HEPES 10 (298 mosmol.kg -1 in rat experiments and 312 mosmol.kg -1 in mouse experiments). Where required, RR or SB (both from Sigma Chemical Co.) were dissolved in HEPES before the experiment Whole-cell recording Cells were patch-clamped with glass pipettes containing (in mm): K- gluconate 120, MgCl 2 1, EGTA 1, HEPES 10, Na 2 -ATP 4, Na-GTP 1, adjusted (280mosmol.kg -1 for mouse experiments and 265mosmol.kg -1 for rat experiments) with mannitol (ph 7.34). All recordings were made using an Axopatch-200B amplifier (Molecular Devices Inc.). In current-clamp experiments, all cells were initially held at a membrane potential of approximately ~50 mv and firing frequency was measured at each 0.5 C. The perfusate was heated using a TC-324B temperature controller and SH-27B solution heater (both from Warner Instruments Inc.). To examine the temperature responses of VP and PNZ neurons in the physiological range, perfusate temperature was initially increased to approximately 32 C over 20 s (0.35 C/s), then slowly ramped to 39 C over 100 seconds (0.07 C/s). The 110

134 temperature of the solution was monitored with a miniature thermocouple placed within 1 mm of the patch-clamped cell Immunocytochemistry Recorded neurons were identified post-hoc as VP or PNZ by immunocytochemistry. Following whole-cell recording, isolated SON neurons were incubated in 4% paraformaldehyde (1 h, 4 C), then in blocking solution (2% normal goat serum, 0.2% Triton-X, 1 h, 4 C), and finally in a mixture of polyclonal rabbit anti-vp (1/1000) and monoclonal anti-oxytocin-associated neurophysin (1/150) overnight at 4 C. Following three washes, cells were incubated (2 h, 23 C) with a mixture of secondary antibodies: Alexa-Fluor 568 goat anti-rabbit IgG and Alexa-Fluor 488 goat anti-mouse IgG (both at 1/200 dilution). Following three washes, cells were visualized under a fluorescence microscope Induction of hyperthermia without osmotic stress in vivo Because hyperosmolality stimulates VP release (Bourque, 1998), it was important to induce hyperthermia without a rise in serum osmolality. Cages housing single weeks-old WT or Trpv1 -/- mice were placed inside a warm incubation chamber with an ambient temperature between 40 and 41 C (measured inside the cage). Preliminary experiments with WT mice revealed that a 20 min heat exposure significantly increased average serum osmolality (from 314 ± 4 mosmol.kg -1 in controls, n=9, to 323 ± 2 mosmol.kg -1 after 20 minutes, n = 5 ; p < 0.05), whereas a 15 minute exposure did not 111

135 (see Figure 4.4A). A 15 min stimulus was therefore used in these experiments. Following the incubation period, animals were quickly anaesthetized with halothane and killed by decapitation. Trunk blood was immediately collected and core body temperature was determined by inserting a digital thermometer (L-0509, Conglom Inc., Canada) inside the body cavity Serum VP measurements Serum VP concentration was determined using an enzyme-linked immunosorbant assay (Assay Design, U.S.A) based on a colorimetric reaction read at 405 nm using a 96-well microplate reader (EAR400AT, SLT- Instruments, Austria). For each mouse, 0.5 ml of blood was collected in a tube containing 500 kiu aprotinin (Sigma-Aldrich, Canada), placed on ice for 30 minutes, and then centrifuged at 1,600g. Serum samples were assayed for VP concentration and osmolality (both in duplicates). To minimize the impact of individual differences in serum osmolality, we rejected data obtained from mice displaying serum osmolality values more than 6 mosmol.kg -1 higher than the average osmolality of control mice in each genotype (i.e. above 320 and 326 mosmol.kg-1 in WT and Trpv1 -/- mice, respectively) Statistical analysis All values are reported as mean ± s.e.m. Comparisons of linear regressions and fits through the data were performed using Prism (v5.01, GraphPad Software Inc., San Diego CA). Comparisons of the means 112

136 observed in different groups were performed using a Student s t-test or an analysis of variance (ANOVA), as appropriate (SigmaStat 2.03; SPSS Inc., Chicago IL) RESULTS VP neurons display a heat-activated current To determine if osmosensory neurons can be specifically activated by increases in temperature, whole-cell voltage clamp recordings were obtained from VP neurons acutely isolated from the rat supraoptic nucleus (SON) and from non-neurosecretory neurons isolated from the adjacent perinuclear zone (PNZ), which do not express TRPV1 (Sharif Naeini et al., 2006). Each cell was thermally challenged by gradually increasing the temperature of the perfusate from 25 to 39 C over a period of ~2 min. In contrast to PNZ neurons, which showed a small and quasi-linear increase in holding current over the entire range of temperatures tested (n = 13), VP neurons displayed a prominent increase in temperature-sensitive inward current at temperatures above 35 C (n = 15; Figure 4.1A,B). Although similar increases in holding current could be evoked reproducibly in single VP neurons (not shown), the increase in holding current induced by consecutive heat stimuli rarely showed a complete recovery (average recovery was 63.7 ± 6.5%). Because the effect of increasing heat was repeatable, we focused our analysis on this part of the response, rather than on the protracted and incomplete recovery phase. For quantification purposes, heat-evoked changes in holding current were 113

137 expressed in Arrhenius plots (e.g. Figure 4.1B) and the absolute slope of a linear regression through the data points was used to quantify the thermal current coefficient (i.e. Q 10 ) of the response above and below 35 C. Whereas both types of cells showed similar slopes in the C range (Figure 4.1B), the mean values of Q 10 observed in the C range were significantly greater in VP neurons (13.1 ± 1.0; n = 15) than in PNZ neurons (2.9 ± 0.4; n = 13; p = 3 x 10-8 ) VP neurons are thermosensitive To determine if the heat-activated current generated in VP neurons can mediate thermosensitive changes in electrical activity, we examined the effects of thermal stimuli applied under current-clamp. Indeed, increasing the temperature of the perfusate from 36.0 C to 38.0 C over a period of ~2 min caused a significant increase in the firing rate of VP neurons (from 2.91 ± 0.77 Hz to ± 2.93 Hz; p = 0.001; n = 10), but had no effect on the firing rate of PNZ neurons (from 1.87 ± 1.17 Hz to 1.12 ± 0.41 Hz; p = 0.65; n = 5; Figure 4.1C,D). Linear regression analysis indicated that the thermal activity coefficient of VP neurons (+5.6 ± 1.1 Hz/ C; n = 10) was well above Hz/ C, a criterion commonly used to define a neuron as being warm sensitive (Boulant and Dean, 1986), whereas PNZ neurons were not thermosensitive (thermal activity coefficient -0.3 ± 0.5 Hz/ C; n = 5; data not shown). Previous studies have shown that heating does not cause a significant depolarization of warm-sensitive neurons in the preoptic area (Zhao and Boulant, 2005). 114

138 Rather, the thermosensitivity of these neurons appears to be caused by a heat-induced inhibition of the transient potassium current (Lee and Deutsch, 1990; Pahapill and Schlichter, 1990) and a corresponding reduction of the interspike interval (Zhao and Boulant, 2005). However, in contrast to warmsensitive preoptic neurons (Zhao and Boulant, 2005) and temperatureinsensitive PNZ neurons (1.17 ± 0.40 mv/ C ; n = 5; data not shown), thermally-stimulated VP neurons displayed a significant membrane depolarization in response to heat (+5.59 ± 1.31 mv/ C; p = 0.03; n = 10). Together with the positive thermal current coefficient reported above, these observations suggest that an active depolarizing receptor current underlies the thermosensitivity of VP neurons in the SON VP neurons express heat-activated calcium-permeable cation channels To determine the nature of the ionic current responsible for the thermosensitivity of VP neurons we performed steady-state current-voltage (I- V) analysis and ion substitution experiments under whole cell voltage clamp. Thermal stimulation of rat VP neurons was associated with an increase in membrane conductance (Supp. Figure 4.1), indicating that ion channels were activated by heat. When compared to that observed in control solution ( 24.2 ± 2.9 mv; n = 6), the reversal potential of the thermosensitive current was not different when measured in a low chloride solution ( 23.7 ± 2.8 mv; n = 5; p = 0.75), or in a solution comprised mainly of CaCl 2 ( 29.3 ± 8.2 mv; n = 115

139 5; p = 0.60). In contrast, the reversal potential of the heat-evoked current was significantly more negative when measured in the absence of external sodium ( 60.4 ± 9.3 mv, n = 3, p = 2 x 10 4 ). Thus the thermosensitive current of VP neurons is mediated by calcium-permeable non-selective cation channels, in agreement with the permeability features of TRPV channels (Venkatachalam and Montell, 2007). If such channels are involved, then the thermosensitivity of these neurons should be inhibited by ruthenium red (RR), a generic blocker of TRPV channels (Ramsey et al., 2006). Indeed, the heat-induced increase in membrane conductance recorded in VP neurons was significantly smaller in the presence of 10 μm RuR (18 ± 5 ps; n = 6) than in control conditions (183 ± 30 ps; n = 7; p = 4 x 10 4 ; see Supp. Figure 4.1) Trpv1 gene expression is required for thermosensory transduction in VP neurons To determine if expression of the Trpv1 gene is specifically required for thermosensitivity, we first compared the thermal responses of VP neurons obtained from wild-type (WT) and Trpv1 -/- mice. As illustrated by the currenttemperature (I/T ) relations shown in Figure 4.2A,B, heat stimuli were less effective at evoking responses in VP neurons from Trpv1 -/- animals than from WT animals. Indeed, the mean thermosensitivity of VP neurons between 35 and 39 C was significantly greater in WT mice (Q 10 of 19.3 ± 2.8 ; n = 19) than in Trpv1 -/- mice (Q 10 of 8.5 ± 1.2 ; n = 13; p = ; Figure 4.2C). The impaired thermosensitivity of VP neurons in Trpv1 -/- animals is consistent with 116

140 a Trpv1 gene product acting as an integral part of the heat-sensing channel. However, an alternate possibility is that the heat-sensing channels do not comprise a product of the Trpv1 gene, but that expression of the latter is required for specification of a thermosensory phenotype during development, or for correct production or targeting of the channel in mature neurons. Therefore, as a second approach, we performed an acute loss-of-function analysis using the selective TRPV1 channel antagonist SB (Gunthorpe et al., 2004). As illustrated in Figure 4.2, (I/T ) relations and average Q 10 values were significantly smaller in WT VP neurons tested in the presence of 1.5 μm SB than under control conditions. Moreover, addition of SB could reverse the inward current induced by heating WT VP neurons, but not that of VP neurons from Trpv1 -/- mice (Figure 4.3A), confirming that the effects of this compound are mediated specifically via blockade of channels encoded by the Trpv1 gene. Interestingly, mean values of Q 10 (Figure 4.2C), as well as the mean amplitudes of heat-evoked current (Figure 4.3B) were respectively equivalent in VP neurons taken from Trpv1 -/- mice and in WT VP neurons tested in the presence of SB These observations indicate that the reduced thermosensitivity of VP neurons in Trpv1 -/- mice reflects an absence of heat-sensitive channels comprising poreforming protein subunits encoded by the Trpv1 gene. 117

141 Contribution of other RR-sensitive channels to thermosensitivity As illustrated in Figure 4.3A, addition of RR abolished the residual heat-activated current recorded from VP neurons treated with SB regardless of genotype, indicating that other subtypes of RR-sensitive channels mediate the residual thermosensitivity of VP neurons. Although the nature of the channels involved remains to be determined, the average magnitude of the SB insensitive current blocked by RR was not significantly different in VP neurons obtained from WT or Trpv1 -/- mice (Figure 4.3C), suggesting that the density or properties of these other channels are not altered as an attempt to compensate for the thermosensory deficit of VP neurons in adult Trpv1 -/- animals Trpv1 gene contributes to anticipatory VP release during hyperthermia To determine if expression of the Trpv1 gene contributes to anticipatory VP release during hyperthermia in vivo, we measured the relationship between serum VP concentration ([VP]) and core body temperature in groups of control and thermally challenged WT and Trpv1 -/- mice. Because increases in serum osmolality associated with dehydration are a potent stimulus for vasopressin release (Bourque, 1998), it was important to induce hyperthermia in the absence of a significant change in serum osmolality. As illustrated in Figure 4.4A, exposure to a warm (40-41 C) 118

142 environment for 15 minutes significantly increased core body temperature in both WT (from 37.5 ± 0.2 C, n = 4 to 38.7 ± 0.1 C, n = 4; P=0.001) and Trpv1 -/- mice (from 37.8 ± 0.1 C, n = 8 to 39.1 ± 0.1 C, n = 8; P<0.001), without provoking concurrent changes in serum osmolality (p=0.66 for WT and P=0.931 in Trpv1 -/-. In the WT animals, hyperthermia caused a significant increase in average serum [VP] (from 163 ± 16 to 311 ± 50 pg.ml -1 ; p<0.05), whereas no significant change could be detected in Trpv1 -/- mice (from 78 ± 11 to 86 ± 17 pg.ml -1 ; P>0.05). Indeed, linear regression analysis of the combined data (i.e. data from control and heated animals) revealed that the slope of the relation between serum [VP] and core body temperature was significantly lower in Trpv1 -/- mice (6 ± 13 pg. C.ml -1, n= 16) than in WT mice (113 ± 41 pg. C.ml -1 ; n = 8; p < 0.005; Figure 4.4B). Thus Trpv1 -/- mice display a significant deficit in anticipatory VP secretion during hyperthermia DISCUSSION The primary function of circulating VP is to promote water reabsorption from the kidney when body fluid volume is reduced and plasma osmolality is elevated, as occurs during dehydration (Bourque, 1998, Verbalis, 2006). Interestingly, increases in core body temperature caused by exposure to heat can provoke VP secretion well before changes in either of these parameters can be measured (Forsling et al., 1976). This release, therefore, confers a preemptive osmoregulatory benefit by reducing the ultimate impact of 119

143 thermoregulatory evaporative water loss on body fluid balance. How the brain performs this anticipatory thermal modulation of VP release is poorly understood. Our results demonstrate for the first time that hypothalamic VP neurons are intrinsically thermosensitive, and that expression of the Trpv1 gene is essential both for thermosensory transduction in these cells and for heat-induced VP secretion in vivo. Moreover, we found that the thermosensitive current could be inhibited by acute exposure to SB366791, a compound that selectively inhibits heat activated currents in cells expressing recombinant TRPV1, but not TRPV4 (Gunthorpe et al., 2004). This finding suggests that the Trpv1 gene encodes a protein that contributes to the pore of the thermosensory channel, rather than a protein that is indirectly required for thermosensitivity. Thermal activation of VP neurons was observed at temperatures above 35 C. Although the apparent temperature threshold for TRPV1 activation in recombinant systems is commonly reported to be near 42 C (Caterina et al., 1997a), the temperature dependence of this channel s activity is known to extend well below this value (Voets et al., 2004, Caterina, 2007). Indeed, the thermal threshold of TRPV1 can be lowered to physiological temperatures through sensitization of the channel by various intracellular mediators such as PKC (Premkumar and Ahern, 2000, Vellani et al., 2001, Numazaki et al., 2002) or PIP 2 (Chuang et al., 2001, Prescott and Julius, 2003). It is also possible that alternative splicing of TRPV1 might lower the thermal activation threshold of TRPV1 (Caterina, 2007). Indeed, VP neurons 120

144 in the SON do not express the full-length capsaicin receptor, but a variant of TRPV1 that lacks part of the N-terminus (Sharif Naeini et al., 2006). Finally the TRPV1 variant expressed in VP neurons may co-assemble with other TRPV subunits, as does TRPV1 (Hellwig et al., 2005, Liapi and Wood, 2005, Rutter et al., 2005, Cheng et al., 2007), and might thus form a channel with significant thermosensitivity in the C range. Indeed, other TRPV channels are known to be expressed in VP neurons (e.g. TRPV2 (Wainwright et al., 2004)). Identifying the molecular structure of the TRPV1 variant expressed in VP neurons is now required to determine if its expression alone, or in combination with other Trpv subunits, can encode a channel with the appropriate thermosensitivity. Our experiments on VP neurons isolated from Trpv1 -/- mice indicate that RR-sensitive channels other than TRPV1 contribute to the residual thermosensitivity of these neurons (Figure 4.3). Although RR can affect mitochondrial calcium homeostasis, this compound does not easily cross the cell membrane and commonly has to be dialysed into cells to affect mitochondrial calcium transport (e.g. (Tang and Zucker, 1997)). Since the blocking effect of RR could be observed within seconds of exposure to extracellular RR (e.g. Figure 4.3), we believe that the effects of RR on heatinduced currents are more likely to be due to channel blockade than to an effect on mitochondrial transport. The heat-activated and RR-sensitive channels TRPV2, TRPV3 and TRPV4 (Dhaka et al., 2006) are thus possible candidates for the residual thermosensitivity observed in Trpv1 -/- mice. 121

145 Additionally, three other TRP channels, namely TRPM2, TRPM4 and TRPM5 have been shown to be thermosensitive (Talavera et al., 2005, Togashi et al., 2006) and blocked by RR (Ramsey et al., 2006). However, TRPM2 is poorly expressed in the hypothalamus (Nagamine et al., 1998), making it an unlikely thermosensor in VP neurons. TRPM4 and TRPM5, for their part, are essentially impermeable to calcium (Launay et al., 2002, Hofmann et al., 2003, Liu and Liman, 2003, Nilius et al., 2003). In VP neurons, not only were heat-evoked currents still observed in external solution containing Ca 2+ as the only inorganic cation (Supp. Figure 4.1), but this current could be blocked (by 66 ± 10 %) by SB Since this fraction is equivalent to the effect of the drug in normal solution (e.g. Figure 4.3B), we conclude that the majority of the residual heat-activated current is Ca 2+ -permeable and, thus, unlikely to be carried by TRPM4 or TRPM5. Additional experiments will be required to identify the nature of the SB resistant thermosensitive channels in VP neurons. We have shown previously that the intrinsic sensitivity of VP neurons to changes in external osmolality also requires expression of the Trpv1 gene (Sharif Naeini et al., 2006). Together with the present results, these observations indicate that the central integration of thermal and osmotic signals can take place within single VP neurons, and might be mediated by a single molecular complex incorporating TRPV1 protein. Previous studies have shown that thermal (Tominaga et al., 1998) and osmotic (Liu et al., 2007) stimuli can sensitize the responsiveness of TRPV1 channels to other 122

146 sensory modalities. Thus when both hyperthermia and hyperosmolality are present, these stimuli may enhance each other s effect on the activity of VP neurons. Indeed, studies in humans have shown that the slope of the relation between plasma VP and osmolality is significantly enhanced by small increases in core body temperature (Takamata et al., 1995). Although the mechanism underlying this synergism is unknown, it is tempting to speculate that this effect may involve a temperature-dependent modulation of the osmosensory channel. Evidence for a similar interaction has been provided by the demonstration that TRPV4-expressing cells show a noticeably greater response to hypotonicity at 37 C than at room temperature (Liedtke et al., 2000). Further studies will be required to investigate interactions between thermosensitivity and osmosensitivity in central osmosensory neurons. 123

147 Figure 4.1. Vasopressin neurons are thermosensitive. A, Sample traces showing the effects of raising temperature (T) from 25 to 39 C (heat; gray ramps) on holding current (V H -60 mv) in single VP and PNZ neurons. B, Arrhenius plots of the data in (A). The Log of the average absolute current at each temperature point was plotted against 1000/T (in Kelvin). Separate linear regressions were fit through data taken between C and between C. The absolute value of the slope is the thermal coefficient, Q 10. Note 124

148 the greater thermosensitivity of the VP neuron above 35 C. C, Sample traces demonstrate the effects of temperature on electrical activity in single VP and PNZ neurons recorded in current-clamp. D, Mean (±s.e.m.) effect of temperature on action potential firing rate in VP (filled circles) and PNZ (empty circles) neurons (* p < 0.05). 125

149 Figure 4.2. The Trpv1 gene contributes to thermosensitivity in isolated mouse VP neurons. A, Sample I-T relations measured in VP neurons obtained from WT mice in the absence (WT) or presence of SB (1.5 µm), and from Trpv1 -/- mice. B, Plots show mean (±s.e.m.) I-T relations measured in VP neurons from WT mice in the absence (WT) or presence of 1.5 μm SB366791, or from Trpv1 -/- mice (* p < 0.05; ** p < 0.01 compared to Trpv1 -/- group). C, Mean (±s.e.m.) Q 10 values observed in the three different groups between 35 and 39 C (WT: 19.3 ± 2.8, n=19; WT + SB366791: 5.8 ± 1.5, n=10; Trpv1 -/- 8.5 ± 1.2, n=13; ** p < 0.01). 126

150 Figure 4.3. SB inhibits the heat-activated current in VP neurons from WT but not Trpv1 -/- mice. A, Whole cell current recordings (V hold -60 mv) show the effects of heat on VP neurons isolated from a WT (upper) and Trpv1 -/- mice (lower). Arrows indicate where SB (1.5 μm) or RR (10 μm) were added to the bath (small gaps are segments removed to eliminate artifacts caused by the addition of drug). Note that SB inhibited the heat-evoked current in the WT neuron but not the Trpv1 -/- neuron, and that RR blocked all of the residual current in both genotypes (dashed lines show baseline holding current). B, Bar graphs show mean (±s.e.m.) current amplitudes measured at 39 o C in the absence (control) or presence of SB SB significantly reduced the heat-evoked current in WT (control ± pa vs ± 31.3pA in SB366791, n=8, p=0.03) but had no effect in Trpv1 -/- neurons (control ± 11.0 pa vs ± 13.4 pa in SB366791, p=0.4, n=7). C, Mean (±s.e.m.) amplitude the RR-sensitive current measured in the presence of SB in VP neurons obtained from 127

151 WT (55.4 ± 28.7pA; n=8) and Trpv1 -/- mice (70.1±12.0 pa, n=7; P=0.66). n.s. = not significant. 128

152 Figure 4.4. Trpv1 / mice show impaired heat-evoked VP release in vivo. A, Bar histograms compare mean (±s.e.m.) values of body temperature (left panel) and serum osmolality (right panel) observed in groups of unheated (control, open bars) and heated (filled bars) WT (n = 8) and Trpv1 -/- mice (n = 16). The heat treatment significantly increased body temperature in both genotypes (** indicates P<0.005), but had no effect on serum osmolality (n.s., not significant; P>0.05). Note that basal serum osmolality is higher in Trpv1 -/- mice, as previously reported (Sharif Naeini et al., 2006). B, Linear regression analysis of the relation between serum [VP] and core body temperature in the WT (open circles) and Trpv1 / mice (filled circles) reported in (A). The solid line is the best fit through the WT data obtained by linear regression (slope ± 40.5 pg. C.ml -1 ; n = 8 ; r 2 = 0.567) whereas the dashed line is the best fit through data obtained from Trpv1 / mice (slope 6.2 ± 13.3 pg. C.ml -1 ; n = 16 ; r 2 = 0.015). The slopes of the lines are statistically different (p = ). 129

153 Supplementary Figure 1. An RR-sensitive and calcium permeable cation conductance underlies thermosensitivity in VP neurons. (A) Sample I-V relations from isolated rat VP neurons show that heating (35 to 39 C) under increases membrane conductance and evokes an inward current with an E rev of 26 mv. The E rev of the heat-evoked current is not significantly different when recorded in (B) a low (5 mm) Cl solution, or (C) a CaCl 2 solution (Calcium), but is shifted to more negative potentials in Na + -free solution (D). (E) Sample I-V relations showing the lack of effect of heating a VP neuron (35 to 39 C) on membrane conductance in the presence of RR (10 μm). (F) Mean (±s.e.m.) values of E rev in different external solutions (* P < 0.001; n.s. not significant). 130

154 CHAPTER FIVE: GLIOTRANSMISSION RESCUES HYPOTONICITY SENSING IN MICE LACKING Trpv OVERVIEW As discussed in section 1.2, the mammalian brain orchestrates its homeostatic defense of the osmotic "set-point" via bi-directional control of behavioral (thirst & salt appetite) (Denton et al., 1996, McKinley et al., 2006) and physiological mechanisms (natriuresis, diuresis & sympathetic tone) (Weisinger et al., 1979, Shi et al., 2007). Central to this are osmosensory neurons (ONs) in the OVLT (Buggy et al., 1979, Denton et al., 1996, McKinley et al., 2006, Bourque, 2008). In chapters 2 and 3 we have shown that hypertonicity increases action potential firing in ONs via depolarization mediated by channels encoded by Trpv1 (Ciura and Bourque, 2006, Sharif Naeini et al., 2006), and previous work has demonstrated that this hypertonically-induced activation modulate effector responses via distinct pathways (Bourque, 2008, Hollis et al., 2008). Surprisingly, nothing is known about the process of systemic hypotonicity sensing. Hypothetically, this function could be mediated by the intrinsic osmosensory mechanism of OVLT neurons. Specifically, cellular swelling could inhibit a basal mechanosensory conductance, possibly mediated by the same channel responsible for hypertonicity sensing, Trpv1. As outlined in section , previous work has also implicated glial neurotransmission in inhibition of neuronal firing rate during a hypotonic stimulus in SON neurons (Hussy et al., 1997). In this study we investigate whether intrinsic osmosensitivity of OVLT neurons, or the glial- 131

155 release of inhibitory neurotransmitters, or both, are responsible for hypotonicity sensing in the OVLT RESULTS AND DISCUSSION To address this issue, we obtained extracellular single unit recordings from OVLT neurons in superfused explants of mouse hypothalamus. Bath application of hypotonic solutions reversibly inhibited firing in 34 of 49 neurons (Figure 5.1A,B). The remainder of the cells showed no significant response (Figure 5.1B), suggesting that the OVLT lacks hypotonicityactivated neurons. To determine if ONs detect hypotonicity via suppression of a basal Trpv1-mediated depolarizing current, we first compared the effects of various hypertonic and hypotonic stimuli in explants prepared from wild type (WT) and Trpv1-knockout (Trpv1 / ) mice. In contrast to hypertonicity-induced excitation, which was eliminated in Trpv1 / mice, hypotonicity inhibited ONs in both genotypes (Figure 5.1). As reported previously [chapter 2; (Ciura and Bourque, 2006)], regression analysis showed a proportional relation between firing rate and hypertonicity in ONs from WT (slope = ± Hz/mOsm; r 2 = 0.412; n = 57; p < 0.05) but not Trpv1 / mice (slope = ± Hz/mOsm; r 2 = 0.057; n = 33; P > 0.05) (Figure. 5.1B,D). However, neurons in both genotypes displayed an equal sensitivity (p > 0.05) in their 132

156 responsiveness to hypotonic stimuli (WT slope = ± Hz/mOsm, r 2 = 0.579, n=43; Trpv1 / slope=0.049 ± Hz/mOsm; r 2 = 0.767; n = 19) (Figure 5.1C,D). This absence of phenotype could indicate that Trpv1 does not mediate hypotonicity sensing, or that extrinsic factors mediate this response in Trpv1 / animals. To determine if Trpv1 enables cell-autonomous hypotonicity sensing in WT mice, we examined the effects of hypoosmolality (-25 mosm) on acutely isolated OVLT neurons. Voltage recordings (not shown) and current-voltage analysis (Figure 5.2A) revealed that hypotonicity inhibits ONs via hyperpolarization caused by suppression of a cation current reversing at ± 2.1 mv (n = 21). Indeed, hypotonicity significantly reduced average input conductance in WT OVLT neurons (-0.08 ± 0.01 ns; n = 32; p < 0.001; Figure 5.2B). In contrast, this treatment did not affect conductance in WT OVLT neurons recorded in the presence of the Trpv1 blocker SB ( ± ns; n = 7; p = 0.095; Figure 5.2A,B) (Gunthorpe et al., 2004), or in OVLT neurons isolated from Trpv1 / mice (-0.01 ± ns; n = 32; p = 0.16; Figure 5.2). These findings reveal that Trpv1 imparts cell-autonomous hypotonicity sensing in WT ONs and indicates that the inhibitory effect of hypotonicity on ONs in explants from Trpv1 / mice is caused by extrinsic factors. Two possibilities were considered. Inhibition might result from; (i) changes in synaptic drive from neurons outside the OVLT or (ii) non-synaptic factors 133

157 released by neighboring cells. However, synaptic blockade had no effect on the hypotonicity induced responses in OVLT neurons from Trpv1 / explants (Supp. Figure 5.1). Studies in the supraoptic nucleus have shown that glia can release the glycine receptor (GlyR) agonist taurine during hypotonicity via volume regulated anion channels (VRACs) [see section ; (Hussy et al., 1997)]. GlyRs are chloride permeable channels and previous work has shown that E Cl in OVLT neurons lies below the resting membrane potential (Nissen et al., 1993). Thus activation of GlyRs in these cells has an inhibitory effect (see Supp. Figure 5.2). We therefore examined if hypotonicity sensing by ONs in Trpv1 / mice involved GlyR-dependent inhibitory gliotransmission (Figure 5.3). In agreement with this hypothesis, the inhibitory effects of hypotonicity were suppressed by the GlyR antagonist strychnine. Moreover, these responses were abolished by the glia selective metabolic toxin (Imamura et al., 1993) fluorocitrate (FC) and by the selective VRAC inhibitor DCPIB (4-[(2-Butyl-6,7-dichloro-2-cyclopentyl-2,3-dihydro-1-ox o-1h-inden-5- yl)oxy]butanoic acid; (Decher et al., 2001)) (see control experiments in Supp. Figure 5.2). Additional experiments confirmed that hypotonicity activates a DCPIB-sensitive anion current in astrocytes but not ONs (Supp. Figure 5.3). Hypotonicity sensing by OVLT neurons in Trpv1 / is therefore mediated by GlyR-dependent gliotransmission. Surprisingly, hypotonicity sensing was blocked by SB in explants from WT mice, but not by strychnine, FC or DCPIB (Figure 5.3). 134

158 These observations indicate that a product of the Trpv1 gene enables cell-autonomous hypotonicity sensing in WT ONs, and that inhibitory gliotransmission does not contribute to this process under normal conditions in situ. Our results also show that OVLT neurons lacking Trpv1 retain the capacity to detect hypotonic stimuli in situ due to an upregulation of GlyRdependent gliotransmission. The basis for this switch is unknown, but does not appear to involve changes in functional GlyR density (Supp. Figures 5.2, 5.4). Whether changes in astrocyte morphology, spatial distribution of GlyRs, or in the accumulation, synthesis and release of GlyR agonists contribute to this compensatory mechanism remains to be determined MATERIALS AND METHODS Preparation of acute hypothalamic explants Adult WT and Trpv1-KO mice (2-6 months) from the C57Bl background were anaesthetized with halothane and decapitated according to a procedure approved by the McGill University Animal Care Committee. The brain was rapidly removed and an explant of basal hypothalamus (5 x 5 x 5 mm) containing the OVLT was excised using razor blades. The optic nerves were removed and the preparation was placed in a superfusion chamber. The OVLT region was superfused (1.5 ml/min) with warm (32 C) carbogenated (95% CO 2, 5% O 2 ) artificial cerebrospinal fluid (ACSF, containing (in mm): 105 NaCl, 26 NaHCO 3, 5 KCl, 1.5 MgCl 2, 1 CaCl 2, 10 D-glucose, all from 135

159 Sigma Chemical Corp., St-Louis MO; ph 7.4) via a tygon tube placed over the rostral side of the explant. The osmolality of the ACSF was adjusted with mannitol as required. Where indicated, kynurenic acid (1 mm, diluted in ACSF from a 1M stock solution prepared in 1N NaOH; Sigma), bicuculline (10 µm, diluted from a 50 mm stock solution in water; Tocris Bioscience, Ellisville MO), SB (3 μm, diluted from a 15 mm stock solution prepared in DMSO; Sigma), strychnine hydrochloride (1 µm, diluted from a stock solution of 10 mm in water; Sigma), fluorocitrate (50 µm, prepared by the barium precipitation method as described below from DL-fluorocitric acid barium salt; Sigma) or DCPIB (20 μm, diluted in ACSF from a stock solution of 100 mm in DMSO) were added to the ACSF. Where specified, glycine (Tocris Bioscience) was dissolved directly in ACSF to the required concentration. To obtain a 50 μm FC solution, 41.5 mg of DL-Fluorocitric acid barium salt (Sigma) was dissolved in 500 μl of 1M HCl to which 500 μl of 1M NaSO 4 was subsequently added. To precipitate the barium, 1ml of 1M Na2HPO 4 was added and the solution was centrifuged at r.p.m. for 10 min. The supernatant was carefully collected and added to 1L ACSF. In the experiments involving fluorocitrate, the drug was superfused for at 90 minutes prior to the start of recording, and cells were recorded for up to 60 minutes. In experiments involving DCPIB, ACSF containing the drug was superfused on each cell at the beginning of the recording, at least 5 minutes prior to the onset of the hypotonic stimulation. 136

160 Extracellular recording and analysis Extracellular recordings of single unit action potential firing were made via borosilicate glass microelectrodes filled with 0.5M potassium acetate (resistance MΩ) using an Axoclamp-2A amplifier (Molecular Devices Corp., Sunnyvale, CA). The voltage signal was amplified 100 times, bandpass filtered ( khz) and acquired to a computer using a Digidata 1440A interface and pclamp10 software (both from Molecular Devices Corp.). Action potentials were detected using the software threshold detection feature of Clampfit 10. Baseline firing rate was computed by averaging 1-2 minutes of activity immediately before the onset of the stimulus. Hypotonic stimuli were typically applied for 2-3 minutes and the firing rate in hypotonic solution was determined from the average firing rate observed during the last 30 seconds of activity recorded before washout. The standard error of the mean (s.e.m.) of the basal firing rates measured in 149 WT OVLT neurons (± 0.41 Hz) was arbitrarily chosen as a minimal threshold to determine if changes in firing rate provoked by hypertonic or hypotonic stimuli were significant (gray areas in Figure 5.1b,d). Scatter plots showing the effects of hypertonicity (right side of panels b and d in Figure 5.1) are expanded data sets comprising data originally reported in reference #7, supplemented with additional data derived from new recordings. 137

161 Whole-cell recordings from acutely isolated neurons The brain of adult WT and Trpv1 -/- mice was obtained as described above. A block of tissue containing the OVLT (< 1 mm 3 ) was excised from the ventral midline tissue lying immediately dorsal to the point where the optic nerves fuse into the optic chiasma. Tissue blocks were incubated for 30 minutes at room temperature in oxygenated (100 % O 2 ) PIPES buffer containing (in mm): 120 NaCl, 5 KCl, 1 MgCl 2, 20 PIPES, 1 CaCl 2, 10 D- glucose, ph 7.4, to which 0.5 mg/ml Protease-X and 0.5 mg/ml Protease-XIV (Sigma) were added. The blocks were then incubated for 30 minutes in protease-free PIPES buffer, then triturated using polished Pasteur pipettes and plated onto 35 mm Petri dishes. The cells were superfused with HEPES buffer containing (in mm): 120 NaCl, 3 KCl, 1 MgCl 2, 10 HEPES, 1 CaCl 2, and 10 D-glucose, ph 7.4 (Sigma). In voltage-clamp experiments tetrodotoxin (0.5 μm; Sigma) was added to the HEPES solution to block voltage-gated sodium channels. Where indicated, DCPIB (20 μm) or SB (3 μm) were added to HEPES. The osmolality of the HEPES buffer was adjusted using mannitol as specified. Borosilicate glass microelectrodes (3-5 MΩ) were filled with an internal solution containing (in mm): 120 K-gluconate, 1 MgCl 2, 1 EGTA, 4 Na 2 ATP, 1 NaGTP and 10 HEPES (Sigma), ph 7.4, 280 mosm (adjusted with mannitol). In the experiments shown in Figure 5.2 (Effects of hypotonicity on isolated OVLT neurons) the reversal potential of the cation current being suppressed is near 30 mv, a value distinct from that observed 138

162 in the capsaicin receptor (~0 mv). The reason for this discrepancy is presently unknown but suggests that the apparent permeability of the osmosensory channel in postnatal rodent neurons is slightly different than that of the homomeric full-length Trpv1 channel expressed in recombinant systems. Like Trpv1, the native osmosensory channel is blocked by Ruthenium red (RR) (Ciura and Bourque, 2006), Gadolinium (Oliet and Bourque, 1996) and SB (Figure 5.2), and it is permeable to calcium (PCa/PNa ~ 5) (Zhang and Bourque, 2006). However the native channel appears to be somewhat more potassium selective (PK/PNa >3) (Voisin et al., 1999, Zhang and Bourque, 2006), causing the reversal potential to be slightly below zero mv. This effect is slightly exaggerated here because a solution containing a relatively low amount of sodium was used (120 mm Na) to allow the preparation of control and hypoosmotic solutions (mannitol removed) comprising the same ion concentrations. The basis for this effect is not known, but could be due to a difference in the characteristics of the TRPV1-variant expressed in the osmosensory neurons, or to the impact of accessory proteins linked to the native channel. In experiments where the density of the glycine currents was investigated, the electrode solution consisted of (in mm): 120 KCl, 5 NaCl, 10 EGTA, 4 Na 2 ATP, ph 7.35, 280 mosm adjusted with mannitol. Cells were patched in the whole-cell mode using an Axopatch-1D amplifier (Molecular Devices Corp.) Sodium Dodecyl Sulfate (SDS) Polyacrylamide Gel Electrophoresis (SDS/PAGE) and Western Blotting 139

163 Equal amounts of OVLT tissue obtained from each of 3 WT and 3 Trpv1-/- mice (as described above), was triturated and homogenized in a buffer for protein denaturation containing 1% SDS, 8M urea and 0.7M 2- mercaptoethanol and flash frozen on dry ice. Following centrifugation at r.p.m. for 10 min, one tenth of the supernatant was loaded onto a 12% SDS Polyacrylamide gel and proteins were separated by electrophoresis at 100V for 3 h. The proteins were transferred to a nitrocellulose membrane in transfer buffer (192 mm glycine, 25 mm Tris-HCl, 20 % methanol, ph 8.3) overnight at 40 V. Membranes were incubated with mouse monoclonal antibody against glycine receptors (mab4a; Synaptic Systems, Germany, 1:1000) for 2 hours and with HRP goat anti-mouse secondary antibody (Jackson ImmunoResearch, PA, 1:5000) for 30 min. Detection was performed using an enhanced chemiluminescence kit (ECL; Perkin Elmer). Membranes were stripped using the Restore Western Blot Stripping Buffer (PierceNet, IL) and probed with mouse anti α-actin antibody (MP Biomedicals, CA, 1:5000) for normalization. The intensity of the bands was quantified using ImageJ analysis software (NIH Scion Image) Whole cell recordings from mouse astrocytes Mouse astrocyte cultures were generously provided by Ms. Denise Cook and Dr. Keith Murai. Briefly, the hippocampi of P0 - P2 mice were removed in glia growth medium (MEM supplemented with 1 % penicillin, 0.6 % glucose and 10 % horse serum; Invitrogen), triturated with a flamed 140

164 Pasteur pipette, passed through a cell strainer and plated in tissue culture flasks coated with poly-d-lysine. Prior to recording, cells were rinsed with phosphate-buffered saline and incubated with 1 ml Trypsin (Invitrogen), after which they were removed from the flask and plated into 30 mm Petri dishes for at least 30 min before the start of the recording. For whole cell patch clamp recordings, borosilicate glass microelectrodes (3-5 MΩ) were filled with the internal recording solution consisting of (in mm): 110 CsCl, 1 MgSO 4, 10 HEPES, 1 EGTA. The ph of the internal solution was adjusted to 7.35 with NaOH and the osmolality to 280 mosm with mannitol. Cells were superfused with a solution consisting of (in mm): 110 CsCl, 2 CaCl 2, 1 MgSO 4, 10 HEPES, ph 7.4, 315 mosm. The hypotonic solution of 250 mosm was obtained by adding less mannitol. DCPIB was used at a final concentration of 20 μm Statistics All values in this paper are reported as mean plus or minus the standard error of the mean (± s. e. m.). Comparisons between groups were made using Student's paired t-test or one-way ANOVA, as appropriate. Differences between means were considered significant when p<0.05. Slopes of linear regressions were compared using Prism 4.03 (GraphPad Software, CA). 141

165 Figure 5.1. Effects of osmotic stimulation in superfused explants. a, upper plots show effects of hypotonicity (-10 to -30 mosm by removal of mannitol) and hypertonicity (lower, +9 to +30 mosm by addition of mannitol) on firing rate in ONs from WT explants. Traces below plots are 10s segments of unit activity during corresponding periods. b, scatter plot showing changes in firing induced by hypotonic (blue) and hypertonic (white) conditions. Lines are linear regressions through each data set. c, effects of osmotic stimulation in Trpv1 / explants (layout as in a). d, scatter analysis of osmoresponsiveness in Trpv1 / explants. Layout as in (b). 142

166 Figure 5.2. Trpv1 is required for hypotonicity sensing in WT ONs. a, examples of steady-state current-voltage plots (upper) and difference currents (lower) recorded from isolated ONs during control and hypotonicity ( 25 mosm). The cation current suppressed by hypotonicity in WT is blocked by SB and is absent in Trpv1 /. b, Bar graphs show mean (± s.e.m.) changes in conductance induced by hypotonicity in each condition (WT, n = 32, *p = ; WT + SB366791, n = 7, p = ; Trpv1-KO, n = 32, p = ). 143

167 Figure 5.3. Different mechanisms mediate hypotonicity sensing in WT and Trpv1 /. a, ratemeter records show the effects of hypotonicity ( 25 mosm, horizontal bar) on the firing rate of ONs in explants from WT and Trpv1 / mice tested under different conditions. Controls show responses recorded in the absence of any drug. Other panels are obtained in the presence of kynurenic acid and bicuculline to block synaptic transmission (see section 144

168 5.3), with either SB (SB), strychnine (Strych), fluorocitrate (FC) and DCPIB. b, Bar graphs show mean (± s.e.m.) changes in firing rate induced by hypotonicity in each condition. Numbers above each bar indicate the corresponding number of cells tested. All changes are significant (p < 0.05) except where n.s. (non significant) is indicated. 145

169 Supplementary Figure 5.1. Synaptic blockade does not prevent hypotonicity sensing in OVLT neurons in explants prepared from trpv1-ko mice. a, sample ratemeter recordings showing the effects of hypoosmotic stimuli (removal of mannitol) on single unit action potential firing by OVLT neurons recorded in superfused hypothalamic explants prepared from trpv1-ko mice. The upper trace shows the response of a neuron under control conditions whereas the lower trace is from another neuron recorded from an explant superfused with artificial cerebrospinal fluid containing 10 μm bicuculline and 1 mm kynurenic acid. These conditions block all spontaneous excitatory and inhibitory postsynaptic currents under our conditions (not shown). b, bar histograms show the mean (± s.e.m.) changes in firing rate (inhibition) observed in groups of OVLT neurons exposed to hypoosmotic stimuli in the absence (control; n = 12) or presence of synaptic block (n = 12). In both groups the inhibitory effect of the stimulus compared to basal firing rate was statistically significant (p < 0.05). However the amplitude of the inhibitory effect observed in the two groups was not significantly different (n.s.; p = ). 146

170 Supplementary Figure 5.2. Single unit recordings and effects of exogenous glycine in various conditions. a, five second excerpts of extracellularly recorded single unit action potential firing by representative OVLT neurons sampled in explants from wild type (WT) and trpv1-ko (KO) mice, as well as KO explants treated with fluorocitrate (KO+FC) or DCPIB (KO+DCPIB). Kynurenic acid (1 mm) and bicuculline (10 μm) were present in all recordings to block synaptic transmission. The ranges of action potential amplitude and firing rate (see c below) were similar in all conditions. b, rate meter plots showing the effects of exogenously applied glycine (bar; 200 μm) on spontaneous action potential firing rate in the conditions indicated above each trace. The inhibitory effect of glycine was potent, reversible and reproducible. c, bar histograms plot mean (± s.e.m.) action potential firing rate in the absence (control, open bars) and presence of glycine (filled bars) by OVLT neurons in WT (n = 6), KO (n = 9), KO+FC (n = 8) and KO+DCPIB explants (n = 8). * indicates p < 0.05 vs. corresponding control (paired t-test). In each case, the mean firing rate recorded in the presence of glycine was not significantly different from zero (p > 0.05). In addition to showing that GlyRs are equally functional in all of these conditions, these data suggest that OVLT neurons were not metabolically impaired in the presence of FC or DCPIB under our recording conditions. 147

171 Supplementary Figure 5.3. DCPIB-sensitive volume regulated anion channels (VRACs) are expressed in glia but not OVLT neurons. Whole cell voltage clamp recordings were obtained from cultured mouse glial cells (a, b) and OVLT neurons isolated from trpv1-ko mice (c, d). Cells were subjected to voltage (ΔV) commands applied as steps (e.g. b) or ramps (e.g. d) and corresponding current responses (ΔI) were used to measure input conductance (G = ΔI/ΔV) under control conditions, following application of a large hypoosmotic stimulus (Hypo ; -60 mosm ; mannitol removed), and after adding the VRAC inhibitor DCPIB (20 μm) in the continued presence of the hypoosmotic medium. As shown in e, hypoosmotic stimuli applied to glial cells caused a significant increase in G that could be significantly reduced by DCPIB (n = 7 cells). In contrast, G was not significantly affected (n.s.) by either hypoosmolality or DCPIB in OVLT neurons isolated from trpv1-ko mice (n = 7 cells). 148

172 Supplementary Figure 5.4. Glycine receptor (GlyR) density is equivalent in WT and trpv1-ko mice. Detection of GlyR protein and α-actin in OVLTs obtained from 3 WT and 3 trpv1-ko mice by Western blot (a). Bar graphs in b quantify the mean (± s.e.m.) density of GlyR relative to actin from the bands shown in a (p = 0.919). Examples of current responses (c) to application of 100 μm glycine (Gly) in isolated OVLT neurons from WT and trpv1-ko mice under voltage clamp (V hold = 60 mv). Note that currents are inward because E Cl ~ 0 mv in our conditions. d shows mean (± s.e.m.) peak (left panel) current and steady state (right panel) current densities measured in WT (n = 10) and trpv1-ko (n = 15) isolated OVLT neurons. No significant differences were observed in either peak (p = 0.738) or steady state (p = 0.861) current density. 149