Cerebrospinal Fluid-Contacting Neurons Sense Ph Changes and Motion in the Hypothalamus

This Accepted Manuscript has not been copyedited and formatted. The final version may differ from this version.

Research Articles: Systems/Circuits

Cerebrospinal fluid-contacting neurons sense pH changes and motion in the hypothalamus

Elham Jalalvand1, Brita Robertson1, Hervé Tostivint2, Peter Löw1, Peter Wallén1 and Sten Grillner1 1The Nobel Institute for Neurophysiology, Department of Neuroscience, Karolinska Institutet, SE-171 77 Stockholm, Sweden 2Evolution des Régulations Endocriniennes, UMR 7221 CNRS, and Muséum National d'Histoire Naturelle, Paris, France.

DOI: 10.1523/JNEUROSCI.3359-17.2018

Received: 24 November 2017

Revised: 4 July 2018

Accepted: 15 July 2018

Published: 23 July 2018

Author contributions: E.J., B.R., P.W., and S.G. designed research; E.J., H.T., and P.L. performed research; E.J., B.R., P.L., and P.W. analyzed data; E.J., B.R., P.W., and S.G. wrote the paper.

Conflict of Interest: The authors declare no competing financial interests.

This work was supported by the Vetenskapsrådet VR-M-K2013-62X-03026 and VR-M-2015-02816, VR- NT-621-2013-4613, the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no 604102 (HBP), EU/Horizon 2020 under grant agreement no 720270 (HBP SGA1), StratNeuro Karolinska Institutet, the Karolinska Institutet's Research Funds and the Centre National de la Recherche Scientifique and the Muséum National d'Histoire Naturelle. We thank Dr. Charlotta Borgius for assistance with the in situ hybridization, and Dr. Liang Wang for conducting experiments on dopamine CSF-c neurons in the spinal cord, which did not respond to deviations in pH.

Correspondence to: Prof. Sten Grillner, The Nobel Institute for Neurophysiology, Department of Neuroscience, Karolinska Institutet, SE-171 77 Stockholm, Sweden., E-mail: [email protected]

Cite as: J. Neurosci ; 10.1523/JNEUROSCI.3359-17.2018

Alerts: Sign up at www.jneurosci.org/cgi/alerts to receive customized email alerts when the fully formatted version of this article is published.

Accepted manuscripts are peer-reviewed but have not been through the copyediting, formatting, or proofreading process.

1 Cerebrospinal fluid-contacting neurons sense pH changes and motion in the

2 hypothalamus

3 Abbreviated title: Hypothalamic CSF-c neurons sense pH and motion

4 Elham Jalalvand1, Brita Robertson1, Hervé Tostivint2, Peter Löw1, Peter Wallén1 and Sten 5 Grillner1*

6 1The Nobel Institute for Neurophysiology, Department of Neuroscience, Karolinska Institutet, 7 SE-171 77 Stockholm, Sweden and 2Evolution des Régulations Endocriniennes, UMR 7221 8 CNRS, and Muséum National d'Histoire Naturelle, Paris, France.

9 *Correspondence to: Prof. Sten Grillner, The Nobel Institute for Neurophysiology, 10 Department of Neuroscience, Karolinska Institutet, SE-171 77 Stockholm, Sweden.

11 E-mail: [email protected] 12 Number of pages: 33

13 Number of figures: 6

14 Number of words in Abstract: 237

15 Number of words in Introduction: 595

16 Number of words in Discussion: 1766

17 We declare no conflict of interest.

18 Authors contributions: S.G., P.W., B.R. and E.J. designed research; E.J., P.L. and H.T. 19 performed experiments; E.J., P.W. P.L. and B.R. analyzed data; E.J., B.R., P.W. and S.G. 20 wrote the paper. 21 22 Acknowledgements: This work was supported by the Vetenskapsrådet VR-M-K2013-62X- 23 03026 and VR-M-2015-02816, VR-NT-621-2013-4613, the European Union Seventh 24 Framework Programme (FP7/2007-2013) under grant agreement no 604102 (HBP), 25 EU/Horizon 2020 under grant agreement no 720270 (HBP SGA1), StratNeuro Karolinska 26 Institutet, the Karolinska Institutet’s Research Funds and the Centre National de la Recherche 27 Scientifique and the Muséum National d’Histoire Naturelle. We thank Dr. Charlotta Borgius 28 for assistance with the in situ hybridization, and Dr. Liang Wang for conducting experiments 29 on dopamine CSF-c neurons in the spinal cord, which did not respond to deviations in pH.

31 ABSTRACT

32 Cerebrospinal fluid-contacting (CSF-c) cells are present in the walls of the brain ventricles

33 and the central canal of the spinal cord and found throughout the vertebrate phylum. We

34 recently identified ciliated somatostatin/GABA-expressing CSF-c neurons in the lamprey

35 spinal cord that act as pH sensors as well as mechanoreceptors. In the same neuron, acidic and

36 alkaline responses are mediated through ASIC3-like and PKD2L1 channels, respectively.

37 Here, we investigate the functional properties of the ciliated somatostatin/GABA-positive

38 CSF-c neurons in the hypothalamus, by performing whole-cell recordings in hypothalamic

39 slices. Depolarizing current pulses readily evoked action potentials, but hypothalamic CSF-c

40 neurons had no or a very low level of spontaneous activity at pH 7.4. They responded,

41 however, with membrane potential depolarization and trains of action potentials to small

42 deviations in pH, in both the acidic and alkaline direction. Like in spinal CSF-c neurons, the

43 acidic response in hypothalamic cells is mediated via ASIC3-like channels. In contrast, the

44 alkaline response appears to depend on connexin hemichannels, and not PKD2L1 channels.

45 We also show that hypothalamic CSF-c neurons respond to mechanical stimulation induced

46 by fluid movements along the wall of the third ventricle, a response mediated via ASIC3-like

47 channels. The hypothalamic CSF-c neurons extend their processes dorsally, ventrally and

48 laterally, but as yet the effects exerted on hypothalamic circuits are unknown. With similar

49 neurons being present in rodents, the pH- and mechano-sensing ability of hypothalamic CSF-c

50 neurons is most likely conserved throughout vertebrate phylogeny.

51 Keywords: CSF-c neurons; hypothalamus; somatostatin; pH sensor; ASIC3; mechanosensor.

54 Significance statement

55 CSF-contacting neurons are present in all vertebrates, located mainly in the hypothalamic area

56 and the spinal cord. Here, we report that the somatostatin/GABA-expressing CSF-c neurons

57 in the lamprey hypothalamus sense bidirectional deviations in the extracellular pH, and do so

58 via different molecular mechanisms. They also serve as mechanoreceptors. The hypothalamic

59 CSF-c neurons have extensive axonal ramifications and may decrease the level of motor

60 activity via release of somatostatin. In conclusion, hypothalamic somatostatin/GABA-

61 expressing CSF-c neurons, as well as their spinal counterpart, represent a novel homeostatic

62 mechanism, designed to sense any deviation from physiological pH and thus constitute a

63 feedback regulatory system, intrinsic to the CNS, possibly serving a protective role from

64 damage caused by changes in pH.

65 INTRODUCTION

66 All organisms are sensitive to changes in the extracellular pH and consequently, for their

67 survival, it is necessary to maintain the pH stable within the physiological range. Variations in

68 extracellular or intracellular pH in brain tissue modulates neuronal excitability and function

69 (Ruusuvuori and Kaila, 2014). Increased CO2 levels, as in ischemia, will result in acidosis,

70 while conversely hyperventilation due to low O2 will result in alkalosis, and similarly

71 metabolic events can influence the pH in both directions (Levin and Buck, 2015). Moreover,

72 during high levels of neuronal activity, pH within the CNS itself is lowered through enhanced

73 levels of lactate that is produced by astrocytes (Magistretti and Allaman, 2015, 2018).

74 We recently showed that the ciliated somatostatin/GABA-expressing CSF-c neurons

75 located at the lateral aspect of the central canal in the lamprey spinal cord act as pH sensors.

76 The same cell detects both acidic and alkaline deviations from around pH 7.4, through the

77 acid-sensing ion channel 3 (ASIC3)-like and the polycystic kidney disease (PKD)-protein-2-

78 like 1 (PKD2L1) channel, respectively, resulting in an increased depolarization and firing of

79 action potentials (Jalalvand et al., 2016a, b). Similarly, CSF-c neurons in the mouse dorsal

80 vagal complex respond to alkaline pH via the PKD2L1 channel (Orts-Del’Immagine et al.,

81 2012, 2016).

82 Even moderate changes in the extracellular pH, whether in the alkaline or acidic

83 direction (0.3 pH units or less) will enhance CSF-c neuronal spike firing, which in turn will

84 suppress the locomotor activity by negative feedback to the spinal cord circuits. The response

85 to any pH change in the spinal cord will thus be a reduced motor activity, which should help

86 the organism to recover its normal pH (Jalalvand et al., 2016a, b).

87 CSF-c neurons come in many different forms that express different transmitters and

88 peptides. They are present in all major groups of vertebrates, from cyclostomes to mammals,

89 and line the wall of the brain ventricles and the central canal (Vigh-Teichmann et al., 1983a;

90 Brodin et al., 1990; Vigh et al., 1977, 2004; Vigh and Vigh-Teichmann, 1998; Russo et al.,

91 2008; Marichal et al., 2009; Orts-Del’Immagine et al., 2012, 2016; Jalalvand et al., 2014).

92 The largest number of CSF-c neurons in the brain is found in the third ventricle in the

93 diencephalon, mainly in the hypothalamus (Vigh and Vigh-Teichmann, 1998). Hypothalamus

94 is an evolutionarily conserved area of the brain that regulates a variety of homeostatic

95 mechanisms such as osmoregulation, food intake and thermoregulation (Boulant and Dean,

96 1986; Hofman and Swaab, 1993; Broberger, 2005; Saper at al., 2002, 2005; Ball, 2007;

97 DiMicco and Zaretsky, 2007). Hypothalamic cultured neurons (mouse) have been reported to

98 be sensitive to acidic pH changes through ASICs (Wang et al., 2007).

99 The hypothalamic CSF-c neurons have a ciliated bulb-like ending that protrudes into

100 the third ventricle (Vigh et al., 1980) and their axons branch and extend laterally, dorsally and

101 ventrally. Our goal here is to investigate the functional role of the somatostatin/GABA-

102 expressing subpopulation of CSF-c neurons in the lamprey hypothalamus, and to determine if

103 they respond to the composition of the cerebrospinal fluid. Patch clamp recordings were

104 performed while the cells were subjected to moderate alterations in the extracellular pH, or to

105 mechanical stimulation by applied fluid movement. In contrast to non-somatostatin-

106 expressing CSF-c neurons in the third ventricle, the somatostatin/GABA-expressing

107 hypothalamic CSF-c neurons responded to both acidic and alkaline pH and to mechanical

108 fluid motion. The acidic and mechanical responses are mediated through ASIC3-like

109 channels, whereas the alkaline response appears to be mediated through the connexin

110 hemichannel. These results show that also the hypothalamic somatostatin/GABA CSF-c

111 neurons play a role as pH sensors, as well as mechanosensors.

112 MATERIAL AND METHODS

113 Animals. Experiments were performed on a total of 44 adult river lampreys (Lampetra

114 fluviatilis) of both sexes that were collected from the Ljusnan River, Hälsingland, Sweden.

115 The experimental procedures were approved by the local ethical committee (Stockholm’s

116 Norra Djurförsöksetiska Nämnd) and were in accordance with the Policy on the Use of

117 Animals in Neuroscience Research (Society for Neuroscience). During the investigation,

118 every effort was made to minimize animal suffering and to reduce the number of animals

119 used.

120 Electrophysiology. Animals (n=32) were deeply anesthetized through immersion in 0.01M

121 phosphate buffered saline (PBS) containing tricane methanesulfonate (MS-222; 100mg L-1;

122 Sigma, St. Louis, MO, USA). Following decapitation, the exposed brain was removed and

123 placed in ice-cold artificial cerebrospinal fluid (aCSF; extracellular solution) of the following

124 composition (in mM): 125 NaCl, 2.5 KCl, 1 MgCl2, 1.25 NaH2PO4, 2 CaCl2, 25 NaHCO3, 10

125 glucose, pH 7.4. The aCSF was oxygenated continuously with 95% O2 and 5% CO2 and

126 osmolarity was udjusted to 290 mOsm L-1 with glucose. Transverse slices (300 μm) of the

127 hypothalamic area were cut using a vibrating microtome (Microm HM 650V; Microm

128 International GmbH, Walldorf, Germany), and mounted in a cooled (5-8˚C) recording

129 chamber. The chamber was was continuously perfused with cooled aCSF. Responses to

130 deviations in extracellular pH were recorded by adjusting the pH of the perfusate to various

131 pH values (6.5, 6.8, 7.1, 7.4, 7.7, 8.0, and 8.3) with HCl or NaOH. Although it takes two

132 minutes to exchange the fluid in the chamber, it has the advantage of knowing the exact pH

133 applied. Previous authors have ejected solutions from a micropipette (Marichal et al., 2009;

134 Orts-Del´Immagine et al., 2016), which has the advantage of short latencies, but the exact pH

135 levels cannot be determined. For mechanical fluid motion stimulation, glass pipettes (2-4

136 MΩ) were filled with aCSF. Fluid pulse stimuli were given by applying 5-20 psi pressure

137 pulses of 10-80 ms duration by a PicoSpritzer II unit (Parker Hannifin Corporation, NJ, USA).

138 Patch electrodes (8-12 MΩ) were prepared from borosilicate glass microcapillaries

139 (Hilgenberg GmbH, Malsfeld, Germany) using a two-stage puller (PP-830, Narishige, Japan)

140 and filled with an intracellular solution of the following composition (in mM): 130 K-

141 gluconate, 5 KCl, 10 HEPES, 4 Mg-ATP, 0.3 Na-GTP and 10 phosphocreatine sodium salt.

142 The pH of the solution was adjusted to 7.4 with KOH and osmolarity to 270 mOsm L-1 with

143 water.

144 CSF-c cells in the hypothalamic area were recorded in whole-cell configuration, and

145 in current or voltage clamp mode using a Multiclamp 700B amplifier (Molecular Devices

146 Corp., CA, USA). The gigaseal resistance during recording was higher than 10 G:, i.e. at

147 least five times higher than the input resistance of the cells. Access to the cell in voltage

148 clamp mode was repeatedly confirmed during the course of the experiment. Cells were

149 visualized with differential interference contrast/infrared optics (DIC/IR). The following

150 drugs were added to the extracellular solution and applied by bath perfusion: the specific

151 ASCI3 blocker APETx2 (1 μM; Cat # 500527; Calbiochem, Merck Chemicals Ltd.

152 Nottingham UK), the connexin hemichannel blocker lanthanum chloride (100 μM; Cat #

153 298182; Sigma-Aldrich), the GABAA receptor antagonist gabazine (20 μM; Cat # 1262;

154 Tocris, Ellisville, MO, USA), the NMDA receptor antagonist AP5 (50 μM; Cat # 0105;

155 Tocris), the AMPA receptor antagonist CNQX (40 μM; Cat # 1045; Tocris) and TTX (1.5

156 μM; Cat # T8024; Sigma-Aldrich).

157 Immunohistochemistry. The brains (n=8) were fixed by immersion in 4% formalin and 14%

158 saturated picric acid solution in 0.1M phosphate buffer (PB; pH 7.4; 1% glutaraldehyde was

159 added for GABA immunohistochemistry) for 12-24 hours at 4˚C, and subsequently

160 cryoprotected in 20% sucrose in PB for 3-12 hours. 20 μm transverse sections were cut on a

161 cryostat (Microm HM 560), collected on gelatin-coated slides and kept at -20˚C until

162 processing. For detection of somatostatin (n=2), co-expression of somatostatin/GABA (n=2),

163 co-expression of somatostatin/α-tubulin (n=2) and somatostatin/connexin 35/36 (n=2),

164 hypothalamic sections were incubated overnight with a rat monoclonal anti-somatostatin

165 antibody (1:200; MAB354; RRID: AB_2255365; Millipore, MA, USA), a mouse monoclonal

166 anti-GABA antibody (0.1 μg/mL; MAB3A12; RRID: AB_2314450; a generous gift from

167 Prof. Peter Streit, Zürich, Switzerland), a mouse monoclonal anti-acetylated α-tubulin

168 antibody (1:500; 6-11B-1; RRID: AB_477585; Sigma-Aldrich) or a mouse monoclonal anti-

169 connexin 35/36 antibody (1:200; MAB3045; RRID:AB_94632; Millipore). The sections were

170 subsequently rinsed thoroughly in 0.01M PBS. To visualize co-expression of somatostatin

171 and GABA, the sections were incubated with a mixture of Cy3-conjugated donkey anti-rat

172 IgG (1:500; Cat # 712-165-150; RRID: AB_2340667; Jackson ImmunoResearch, West

173 Grove, PA, USA) and Alexa Fluor 488-conjugated donkey anti-mouse IgG (1:200; Cat # 715-

174 545-150; RRID: AB_2340846; Jackson ImmunoResearch) for 2 hours. For co-expression of

175 somatostatin and α-tubulin or connexin 35/36, the sections were incubated with Alexa Fluor

176 488-conjugated donkey anti-rat IgG (1:200; Cat # 712-545-153; RRID: AB_2340684;

177 Jackson ImmunoResearch) and Cy3-conjugated donkey anti-mouse IgG (1:500; Cat # 715-

178 165-150; RRID: AB_2315777; Jackson ImmunoResearch). The sections were then rinsed in

179 0.01 M PBS and mounted with glycerol containing 2.5% diazabicyclooctane (DABCO; Cat #

180 D27802; Sigma-Aldrich).

181 To examine the morphology of hypothalamic CSF-c neurons, cells were

182 intracellularly labeled with 0.3% Neurobiotin (Cat # SP-1120; Vector Laboratories,

183 Burlingame, CA, USA) during the whole-cell recording. The hypothalamic slices were fixed

184 overnight in 4% formalin and 14% picric acid in 0.1 M PB. They were subsequently rinsed

185 thoroughly in 0.01M PBS and incubated with Alexa Fluor 488-conjugated streptavidin

186 (1:1000; Cat # 016-640-084; Jackson ImmunoResearch) for 3 hours. The sections were then

187 rinsed in 0.01 M PBS and mounted with glycerol containing 2.5% DABCO. Alternatively,

188 after fixation and rinsing in PBS, slices were incubated in Vectastain (Cat # PK6100; Vector

189 Laboratories) followed by diaminobenzidine (DAB; Cat # SK-4100; ImmPACT, Vector

190 Laboratories) for 5 min, then rinsed and dehydrated in alcohol prior to mounting in Entellan

191 (Merck).

192 To investigate if the intracellularly Neurobiotin labeled CSF-c cells express

193 somatostatin, the slices were incubated overnight with a rat monoclonal anti-somatostatin

194 antibody (1:200; MAB354; RRID: AB_2255365; Millipore) then rinsed thoroughly in 0.01M

195 PBS and incubated with a mixture of Alexa Fluor 488-conjugated streptavidin (1:1000; Cat #

196 016-640-084; Jackson ImmunoResearch) and Cy3-conjugated donkey anti-rat IgG (1:500; Cat

197 # 712-165-150; RRID: AB_2340667; Jackson ImmunoResearch). All primary and secondary

198 antibodies were diluted in 1% bovine serum albumin (BSA), 0.3% Triton X-100 in 0.1 M PB.

199 In situ hybridization. Animals (n=2) were deeply anesthetized as described above and the

200 brain and rostral spinal cord (the spinal cord was used as a positive control) were removed,

201 fixed in 4% formalin in 0.1M phosphate buffer (PB) over night at 4°C and then cryoprotected

202 in 20% sucrose in 0.1M PB. 10 and 20 μm thick cryostat sections were made and stored in -

203 80°C until processed. Single-stranded digoxigenin-labeled sense and antisense pkd2l1

204 riboprobes were generated by in vitro transcription of the previously cloned pkd2l1 cDNA by

205 using the Digoxigenin RNA Labeling kit (see Jalalvand et al., 2016b; Cat # 11 277 073 910;

206 Roche Diagnostics, Mannheim, Germany). Briefly, sections were incubated for 1h in

207 prehybridization mix (50% formamide, 5X SSC, 1% Denhardts, 50 μg/mL, salmon sperm

208 DNA, 250 μg/mL yeast RNA) at 60°C. Sections incubated with the heat-denaturated

209 digoxigenin-labeled riboprobe were hybridized overnight at 60°C. Following the

210 hybridization, the sections were rinsed twice in 1X SSC, then washed twice in 1XSSC (30

211 min each) at 60°C and two times in 0.2XSSC at room temperature. After blocking in 0.5%

212 blocking reagent (PerkinElmer), the sections were incubated overnight in anti-DIG antibody

213 coupled to HRP (1:2000; Cat # RRID: AB_514497; Roche Diagnostics) at 4°C. The probe

214 was then visualized by TSA Cy3 Plus Evaluation Kit (Cat # NEL763E001; PerkinElmer). The

215 specificity of the hybridization procedure was verified by incubating sections with the sense

216 riboprobe (data not shown). The sections were rinsed thoroughly in 0.01M PBS and then

217 incubated with a rat monoclonal anti-somatostatin antibody (1:200; MAB354; Millipore)

218 overnight at 4°C, rinsed in PBS and incubated with Alexa Fluor 488-conjugated donkey anti-

219 rat IgG (1:200; Cat # 712-545-153; RRID: AB_2340684; Jackson ImmunoResearch) for 2

220 hours and mounted with glycerol containing 2.5% DABCO. All primary and secondary

221 antibodies were diluted in 1% bovine serum albumin (BSA), 0.3% Triton X-100 in 0.1 M PB.

222 Western blot. Lamprey brains (n=2) were solubilized by sonication in 1% Na-SDS, 0.32M

223 sucrose, 20 mM Hepes pH 7.6 and a cocktail of protease inhibitors. Samples was then

224 subjected to SDS Page electrophoreises on Bio-Rad gradient gels (4-15%) and subsequently

225 transfered to PVDF membrane according to the manufacturer. The membrane was incubated

226 with mouse anti-connexin 35/36 (1:1000; MAB3045; RRID: AB_94632; Millipore) overnight

227 at 4°C and then with goat anti-mouse IgG-HRP (1:5000; P0447; DAKO) for 2 hours at room

228 temperature. It was developed with SuperSignal© West Dura (Thermo Scientific).

229 Morphological analysis. Photomicrographs were taken with an Olympus XM10 digital

230 camera, mounted on an Olympus BX51 fluorescence/bright field microscope (Olympus

231 Sverige AB, Stockholm, Sweden). Intracellularly labeled cells were drawn using a camera

232 lucida tube attached to a bright field microscope (Leitz, Germany). Confocal images were

233 obtained using a Zeiss confocal laser scanning microscope (LSM 510 NLO, Carl Zeiss AB,

234 Stockholm, Sweden) and processed using Zeiss LSM software. Illustrations were prepared

235 using Adobe Illustrator and Photoshop CS6. Images were adjusted only for brightness and

236 contrast.

237 Experimental design and statistical analysis. Hypothalamic CSF-c neurons were recorded

238 in whole-cell configuration, and in current or voltage clamp mode. Resting membrane

239 potentials were determined in current clamp mode. Bridge balance and pipette capacitance

240 compensation were adjusted and signals were digitized and recorded using Clampex software

241 and analyzed in Clampfit (pCLAMP 10, Molecular Devices, CA, USA). Membrane- and

242 firing properties were analyzed by injection of short (20 ms) and long (500 or 1000 ms)

243 depolarizing and hyperpolarizing current pulses. Responses to deviations in extracellular pH

244 were recorded as changes in membrane potential and firing frequency (current clamp), and in

245 the frequency of inward current events (voltage clamp). Effects were analyzed by comparing

246 the responses in the same neuron at the control condition at physiological pH (7.4) and at

247 acidic or alkaline pH, respectively, or in the absence or presence of an ion channel blocker. In

248 all cases, paired comparisons were made between two conditions (e.g. control vs acidic pH, or

249 absence vs presence of blocker). Data are presented as means ± standard deviation (SD) or

250 standard error of the mean (SEM), and statistical comparisons were made using Student’s

251 paired two-tailed t-test. : p<0.001, **: p<0.01 significant difference compared to control;

252 n.s.: no significant difference (for details, see figure legends).

253

254 RESULTS

255 Somatostatin/GABA-expressing CSF-c neurons in the hypothalamus

256 An area in the rostral hypothalamus contains numerous somatostatin-positive CSF-c neurons

257 (Fig. 1A, B). These neurons have a short, thick apical process with a bulb-like ending (Fig.

258 1C) from which an α-tubulin-immunoreactive cilium protrudes into the CSF (Fig. 1D, arrow).

259 To further identify the phenotype of the hypothalamic CSF-c neurons we examined if these

260 cells co-express somatostatin and GABA, similar to spinal CSF-c neurons. Most of the

261 somatostatin-positive neurons within the hypothalamus were CSF-c cells (Fig. 1B, C) that co-

262 expressed GABA (Fig. 1E, arrows). However, some CSF-c neurons in this region were only

263 immunoreactive to GABA (Fig. 1E, arrowhead). In the lamprey, the periventricular area of

264 hypothalamus is the main source of somatostatin. Intracellular Neurobiotin labeling of

265 recorded hypothalamic CSF-c neurons showed that their axonal branches extend laterally,

266 dorsally and ventrally (Fig. 1F). Somatostatin-positive fibers and terminals were found

267 throughout the brain including a dense innervation of motor-related areas, such as the deep

268 layer (DL) of the optic tectum and the reticulospinal nuclei (Fig. 1G, H). Sensory areas

269 throughout the brain were, on the other hand, devoid of somatostatin innervation, as shown in

270 the retinorecipient superficial layer (SL) of the optic tectum and the brainstem alar plate (Fig.

271 1G, H).

272 Presence of gap junctions in somatostatin-expressing CSF-c neurons

273 Intracellular injection of Neurobiotin into a single somatostatin-expressing CSF-c neuron in

274 hypothalamus sometimes resulted in staining of one or two additional CSF-c neurons in close

275 vicinity to the injected cell, suggesting dye coupling via gap junctions (n=4; Fig. 2A). To

276 examine this possibility further, we applied the neuronal connexin-35/36 antibody, a gap

277 junction marker (O’Brien et al., 1998; Srinivas et al., 1999). Somatostatin-positive CSF-c

278 neurons co-expressed the connexin protein (Fig. 2B-D), suggesting that gap junctions are

279 present between at least some CSF-c neurons in the hypothalamus. Fig. 2E is a western blot of

280 the anti-connxin 35/36 antibody (immunogen recombinant perch connexin 35), showing that

281 this antibody recognizes a band at around 35kD and is thus specific for the lamprey brain and

282 a band around 75kD probably representing a dimer (O’Brian et al., 2004). Searching the

283 genome build for Petromyzon_marinus-7.0 (petMar2) using mouse connexin we also fond a

284 potential ortholog for lamprey connexin (PMZ_00001234-RA) with more then 80% identity

285 to both mouse and perch connexin 35/36 (results not shown).

286 Neuronal properties of hypothalamic CSF-c neurons

287 To examine the membrane properties of somatostatin-expressing CSF-c neurons in the

288 hypothalamus we performed whole-cell recordings in voltage- or current-clamp mode (n=45

289 cells). The hypothalamic CSF-c neurons had a mean resting membrane potential of -65 ± 4

290 mV, a mean input resistance of 1.9 ± 0.4 GΩ (n=15) and a mean spike threshold of -45 ± 1.8

291 mV (n=15). In contrast to spinal CSF-c neurons, hypothalamic CSF-c neurons did not fire

292 action potentials spontaneously under control conditions. However, spiking was readily

293 evoked in response to depolarizing current injection. For a current pulse of short duration (20

294 pA, 20 ms), the neuron responded with a single action potential followed by an

295 afterhyperpolarization (Fig. 3A, left trace). A train of action potentials was evoked with longer

296 current injections (20 pA, 500 ms; Fig. 3A, right trace). Hypothalamic CSF-c neurons

297 generally showed spontaneous GABAergic and glutamatergic postsynaptic potentials, the

298 latter mediated by both α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and

299 N-Methyl-D-aspartic acid (NMDA) receptors (Fig. 3B). The voltage responses recorded

300 during depolarizing current steps in current clamp mode showed reliable spiking with spike

301 frequency adaptation (Fig. 3C). The I-V curve revealed a linear current-voltage relationship

302 (Fig. 3D).

303 Hypothalamic CSF-c neurons expressing somatostatin respond to acidic and alkaline

304 pH.

305 To examine whether hypothalamic somatostatin-expressing CSF-c neurons respond to

306 changes of the extracellular pH, acidic (pH 6.5; red trace in Fig. 4A) or alkaline (pH 8.0; blue

307 trace in Fig. 4A) solutions were bath-applied to hypothalamic slices during whole-cell

308 recording.. To exclude the possibility of indirect effects through synaptic inputs, blockers of

309 GABA (gabazine) and glutamate (AP5, CNQX) receptors were applied throughout these

310 experiments. In 30 out of 45 tested CSF-c neurons, in the same cell a decrease as well as an

311 increase of pH depolarized the membrane potential by 10-12 mV, eliciting a sustained

312 discharge of action potentials (Fig. 4A). After application of tetrodotoxin (TTX), the

313 membrane potential depolarization induced by acidic or alkaline pH remained (Fig. 4B;

314 n=12).

315 These results thus suggest that the response to both acidic and alkaline extracellular

316 pH is due to a direct effect on the CSF-c neuron. The recorded cells were intracellularly

317 labeled with Neurobiotin and co-stained for somatostatin, verifying that the pH-sensing CSF-c

318 cells express somatostatin (Fig. 4C). All recorded CSF-c neurons that were non-responsive to

319 acidic or alkaline extracellular pH, did not express somatostatin (n=15). Figure 4D illustrates

320 the mean membrane potential changes of CSF-c neurons recorded in whole-cell mode,

321 showing that individual CSF-c neurons in the hypothalamus respond with membrane potential

322 depolarization (and firing) to step changes in pH (6.5, 6.8, 7.1, 7.7, 8.0, 8.3) from the normal

323 value of 7.4, yielding a U-shaped response curve (p< 0.001, paired t test; n=15 cells).

324 The response to acidic and alkaline pH was also analyzed in voltage clamp mode

325 (Fig. 4E-G). Only a few low amplitude current events were seen at pH 7.4 (AP5, CNQX,

326 gabazine and TTX present; Fig. 4E). Upon decreasing or increasing the extracellular pH to 6.5

327 or 8.0, respectively, the frequency of inward current events increased markedly (Fig. 4E, F; p

328 < 0.001, paired t test; n=7 cells). We also recorded unitary inward current events, presumably

329 corresponding to single channel openings, at acidic and alkaline pH (Fig. 4G).

330 At the transition from the brainstem to the spinal cord, where the central canal starts,

331 there is a small region in which CSF-c neurons co-express somatostatin and dopamine. These

332 cells also responded to acidic and alkaline pH deviations (n=3; not illustrated). However,

333 CSF-c neurons, both at the obex level and in the spinal cord, that only express dopamine did

334 not respond to pH changes (n=4). It thus appears that the common factor is an expression of

335 somatostatin in CSF-c neurons responding to deviations of the extracellular pH.

336 Specific blockade of ASIC3 eliminates the response of hypothalamic CSF-c neurons to

337 acidic pH

338 To investigate if the sensitivity of hypothalamic CSF-c neurons to decreases in pH depends on

339 the ASIC3 subtype, as is the case with their spinal counterparts (Jalalvand et al., 2016a), the

340 specific ASIC3 blocker APETx2 was applied. In current clamp mode, changing the pH of the

341 extracellular solution to acidic or alkaline resulted in firing of action potentials, as well as a

342 net depolarization of the membrane potential (Fig. 5A). Following application of APETx2,

343 both these aspects of the response were eliminated at acidic pH, while the alkaline response

344 was not affected (Fig. 5A; n=5). Changing the pH of the extracellular solution to 6.5 or 8.0 in

345 voltage-clamp mode (gabazine, CNQX, AP5 and TTX present), caused inward current

346 deflections which were abolished by APETx2 at acidic pH, whereas the current events

347 induced by alkaline pH remained (Fig. 5B, C pH 6.5 vs pH 7.4 non-significant; pH 8 vs pH

348 7.4 p < 0.01, paired t test; n=3 cells). Thus, as in the spinal CSF-c neurons, ASIC3-like

349 channels represent the acid sensor in somatostatin-expressing hypothalamic CSF-neurons.

350 Specific blockade of connexin hemichannels eliminates the response of hypothalamic

351 CSF-c neurons to alkaline pH

352 To investigate which type of channel that might be sensing alkaline pH in hypothalamic CSF-

353 c neurons, we performed in situ hybridization for PKD2L1 (no specific PKD2L1 antagonist

354 exists). This channel was a main candidate since it is considered to mediate the alkaline

355 response in spinal somatostatin-positive CSF-c neurons (Jalalvand et al., 2016b). However, no

356 hypothalamic somatostatin-expressing nor any other hypothalamic CSF-c neurons expressed

357 the PKD2L1 channel (Fig. 5D). As a positive control, we performed in situ hybridization on

358 spinal cord sections that were mounted on the same slide as the hypothalamic sections. Strong

359 PKD2L1 expression was observed in spinal CSF-c neurons (Fig. 5E), in correspondence with

360 our previous findings (Jalalvand et al., 2016b). Thus, PKD2L1 is not the alkaline sensor in

361 hypothalamic somatostatin-expressing CSF-c neurons.

362 The connexin hemichannels, which can be present in the cell membrane although

363 there is no coupling to an adjacent cell, corresponds structurally to one half of the gap

364 junction channel, and represents another candidate that has been suggested to sense alkaline

365 pH (Schalper et al., 2010). The somatostatin-positive hypothalamic CSF-c neurons are

366 immunoreactive to connexin 35/36 (see Fig. 2), and we therefore tested whether the connexin

367 hemichannel blocker lanthanum (La3+) could block the alkaline response. Lanthanum blocks

368 connexin hemichannels but has no effect on the gap junction channels themselves at the

369 concentrations we have used (100 μM and 70 μM; Contreras et al., 2002; Spray et al. 2006).

370 In a CSF-c neuron that responded to both acidic and alkaline pH, application of LaCl3 (100

371 μM) completely blocked responses to alkaline pH including the net depolarization of the

372 membrane potential, as well as the firing of action potentials (Fig. 5F, G; control: pH 6.5 and

373 pH 8 vs pH 7.4, significant difference p < 0.001; lanthanum, 100 μM: pH 8 vs control,

374 significant difference p < 0.001, paired t test; n=5). LaCl3 at 100 μM, however, also tended to

375 reduce the acidic response (non-significant, p = 0.06, paired t test; n=5). At 70 μM, there was

376 no effect on the acidic response, but a clear and significant reduction of the alkaline response

377 (Fig. 5G; p < 0.001, paired t test; n=5). The recorded CSF-c neurons were intracellularly

378 labeled with Neurobiotin and co-stained for somatostatin, verifying that the hypothalamic

379 CSF-c neurons responding to the changes of extracellular pH, express somatostatin (Fig. 5H).

380 Some hypothalamic CSF-c neurons did not respond to either alkaline or acidic pH

381 (Fig. 5I), and none of these non-pH sensitive CSF-c neurons expressed somatostatin (Fig. 5J).

382 These somatostatin-negative CSF-c neurons exhibited spontaneous postsynaptic potentials

383 (Fig. 5I, top trace), as well as action potential firing upon current injection.

384 Taken together, these results suggest that the alkaline response in hypothalamic,

385 somatostatin-expressing CSF-c neurons is mediated via connexin hemichannels.

386 Hypothalamic CSF-c neurons sense fluid movements via ASIC3-like channels

387 To investigate whether hypothalamic CSF-c neurons are mechanosensitive to fluid

388 movements, brief pressure pulses were applied to an aCSF-filled micropipette placed close to

389 the bulb-like protrusion of the CSF-c neuron with its cilium (Fig. 1E), while performing patch

390 recordings (Fig. 6A). To verify that the responses were not indirect or synaptically evoked,

391 GABAergic and glutamatergic synaptic transmission was blocked by bath-application of

392 gabazine, AP5 and CNQX. Graded receptor potential responses were reliably elicited by fluid

393 pulses (duration 40-100 ms) in all CSF-c neurons tested (n=8; Fig. 6B, C). The receptor

394 potential amplitude increased with increasing pulse magnitudes (from 10 to 20 psi in Fig. 6B),

395 and action potentials were evoked when the membrane potential was held at -55 mV (top

396 trace in Fig. 6B).

397 Since ASIC3-like channels mediate the mechanosensitivity underlying the fluid pulse

398 response in spinal CSF-c neurons (Jalalvand et al., 2016a), we investigated whether this

399 channel may also act as a mechanosensor in hypothalamic CSF-c neurons that express

400 somatostatin. The ASIC3 blocker APETx2 was bath-applied while recording the mechanical

401 response to fluid pulses. In all cells tested (n=4), all of which were somatostatin-positive (data

402 not shown), the response was eliminated in the presence of APETx2 (Fig. 6C, D). Thus, also

403 the mechanosensitivity of the hypothalamic CSF-c neurons appears to be mediated by ASIC3-

404 like channels. The same CSF-c neuron responded to decreased (pH 6.5) and increased

405 extracellular pH (pH 8; Fig. 6E).

406 DISCUSSION

407 An accurate control of the acid-base balance is one of the main homeostatic tasks of an

408 organism. With regard to the brain it would be important to sense the pH of the cerebrospinal

409 fluid, since it integrates the overall impact of pH changes within the brain. To have pH

410 sensors in the wall of the third ventricle at the hypothalamic level would seem to be an

411 optimal location, given the role of hypothalamus in a variety of other homeostatic

412 mechanisms. A high level of neuronal activity can by itself lead to an extracellular acid shift,

413 through the glucose-lactate shuttle in the astrocytes to furnish neurons with lactate as a source

414 of energy (Chelser and Kaila, 1992; Magistretti and Allaman, 2015, 2018). Moreover,

415 increased CO2 levels (Voipio and Kaila 1993), as in ischemia, will lead to acidosis, while

416 conversely hyperventilation due to low O2 levels will result in alkalosis. Similarly, metabolic

417 events can influence the pH in both directions (Levin and Buck, 2015). Somatostatin/GABA

418 CSF-c cells lining an area within the 3rd ventricle, as shown here, sense both acidic and

419 alkaline deviations from the control level. This also applies to the same type of CSF-c neurons

420 located around the central canal (Jalalvand et al., 2016a, b) and mammalian CSF-c neurons at

421 the level of obex (Orts-Del´Immagine et al., 2016).

422 The pH-sensing capacity in both the hypothalamus and the spinal cord appears to be

423 exclusive to the somatostatin CSF-c subpopulation of neurons, since somatostatin-negative

424 neurons do not respond to pH changes. In the spinal cord, acidic and alkaline deviations of the

425 extracellular pH lead to an activation of the CSF-c neurons and a release of somatostatin in

426 the spinal cord, which in turn exerts a depressing effect on the network generating locomotor

427 activity (Jalalvand et al., 2016a, b). The response to any pH change in the spinal cord will thus

428 be a reduced motor activity, which should help the organism to recover its normal pH.

429 Do the hypothalamic somatostatin-positive CSF-c neurons have a similar role as their

430 spinal counterparts? There is a very rich somatostatin innervation in the motor areas of the

431 brain extending from the forebrain to the brainstem, whereas sensory areas have little or no

432 innervation (see also below). Although not addressed here, it raises the possibility that the

433 hypothalamic somatostatin/GABA-expressing CSF-c-neurons could act at the supraspinal

434 level to reduce motor activity and thereby help restore the pH level.

435 pH sensitivity of the CSF-c neurons in relation to physiological and pathophysiological

436 conditions

437 The U-shaped curves of pH sensitivity of both hypothalamic and spinal CSF-c neurons show

438 a significant depolarization at pH 7.1 or 7.7, from a neutral value of 7.4, but a much larger

439 response occurs outside this area, at pH 6.5 and 8.3 (cf. Figure 4D). Whereas pH values

440 between 7.1 and 7.7 can be regarded as within the physiological range, values outside this

441 range would imply pathophysiology. The somatostatin CFS-c neurons will thus respond

442 steeply if the pH shifts to pathophysiological levels, but will also detect smaller changes

443 within the physiological range. In lamprey, the pH of the blood can change with as much as

444 0.5 units during excercise (Tufts et al., 1992), and the pH-values in the interstitial fluid is

445 somewhat lower than in the blood around pH 7.3 to 7.4 (Chesler, 1986). During epilepsy or

446 ischemia, pH values of 6.5 have been reported in mammalian brain tissue (Siesjö et al., 1985;

447 Rehncrona 1985). In the blood of patients with severe acidosis, pH just below 7 and for

448 alkalosis above 7.65 have been reported (Thorén, 1960). Also, under physiological conditions,

449 high levels of neuronal activity can lead to deviations in the acidic direction due to the

450 glucose-lactate shuttle in the astrocytes, which provides energy to neurons through lactate

451 (Chelser and Kaila, 1992; Magistretti and Allaman, 2015).

452 The fact that the somatostatin CSF-c neurons respond to both acidic and alkaline

453 deviations with increased activity can be regarded as an elegant evolutionary solution to

454 respond to, and master, a life-threatening condition. The response in both cases can result in

455 reduction of the motor activity. In the case of metabolic demands during ischemia or

456 hyperactivity this will lead to a restoration of the pH. Similarly, during conditions with low

457 pO2, when hyperventilation will cause a lower pCO2 and thereby an alkaline pH, a reduction

458 of the motor activity will also help restore the pH. This applies to spinal cord somatostatin

459 CSF-c neurons and may also apply to the hypothalamic somatostatin CSF-c cell

460 subpopulation.

461 The response to alkaline pH is mediated by different molecular mechanisms in spinal

462 and hypothalamic CSF-c neurons

463 PKD2L1 is the alkaline sensor in brainstem and spinal CSF-c neurons (Jalalvand et al 2016,

464 a,b; Orts-Del´Immagine et al., 2016), and was therefore considered a main candidate for the

465 hypothalamic CSF-c neurons. Unexpectedly, we could not detect any expression of PKD2L1

466 in hypothalamic CSF-c neurons (Fig. 5D), suggesting that another unidentified channel is

467 responsible for the alkaline response in hypothalamus. This is in line with studies in the

468 mouse, in which PKD2L1 channels are not expressed at a level just rostral to obex (Orts-

469 Del´Immagine et al., 2014). A very small group of mouse hypothalamic cells in cell culture

470 were shown to be PKD2L1-positive, but were not identified as CSF-c cells (Huang et al.

471 (2006).

472 Different extracellular alkaline pH sensors have been described in the vertebrate

473 brain, including connexin hemichannels, insulin receptor-related receptors (IRR), and two-

474 pore-domain K+ (K2P) channels (for review, see Murayam and Maruyama, 2015). Since the

475 connexin 35/36 protein is expressed in hypothalamic CSF-c neurons (Fig. 2B-D), in contrast

476 to the cells around the central canal, we considered connexin hemichannels as a possibility.

477 The connexin hemichannels are sensitive to pH changes (Francis et al., 1999; Contreras et al.,

478 2002; Sáez et al., 2005; Spray et al., 2006; Yu et al., 2007), and extracellular alkalization

479 increases intracellular Ca2+ levels in cells expressing connexin hemichannels (Schalper et al.,

480 2010).

481 The somatostatin CSF-c neurons in hypothalamus express connexin 35/36, and the

482 alkaline response could be totally blocked with lanthanum, suggesting that connexin

483 hemichannels are sensors for alkaline pH. We used two different concentrations of lanthanum,

484 70 and 100 PM. At 70 PM, lanthanum did not affect the acidic response but very significantly

485 reduced the alkaline response, whereas at 100 PM, a complete blockade of the alkaline

486 response was observed, together with a non-significant reduction of the acidic response. At

487 very high concentrations (≥100 μM) lanthanum has an inhibitory effect on many channels

488 including kainate and AMPA receptors (1-30 mM; Reichling and MacDermott 1991; Hong et

489 al., 2004), as well as TRP channels (at concentrations higher than 1mM; Zhao et al., 2015),

490 and might thus at such high levels also affect ASIC3 in hypothalamic CSF-c neurons.

491 Another difference between lamprey hypothalamic and spinal CSF-c neurons is that

492 the former have a lower mean resting potential (-65 mV compared to -52 mV for spinal CSF-c

493 neurons) at pH 7.4 and therefore rarely are spontaneously active at this pH. Presumably as a

494 consequence of this difference, hypothalamic CSF-c neurons showed a more prominent

495 membrane potential depolarization in response to a change in pH than spinal CSF-c cells (cf.

496 Jalalvand et al., 2016a).

497 ASIC3-like channels mediate the response to acidic pH and to fluid motion

498 The acidic response in hypothalamic CSF-c neurons, as in spinal CSF-c neurons (Jalalvand et

499 al. 2016a), is mediated by ASIC3-like channels, since APETx2, a specific antagonist of

500 ASIC3 (Diochot et al., 2004), blocks the acidic sensing. In rodents, ASIC3 has been reported

501 to be expressed throughout the hypothalamus, including the suprachiasmatic nucleus (Wang

502 et al., 2007; Chen et al., 2008; Meng et al., 2009). Lamprey hypothalamic CSF-c neurons are

503 also mechanosensitive and activated by fluid movements, like motion of the CSF. This

504 response is also mediated by ASIC3-like channels, as in the spinal cord (Jalalvand et al.,

505 2016b). In CSF-c neurons of the most caudal part of the mouse brainstem, the acidic response

506 was shown to be mediated through the general class of ASICs (Orts-Del´Immagine et al.,

507 2016). It would therefore seem that vertebrate CSF-c neurons responsive to acidic pH have

508 developed this capacity through the expression of ASICs, in contrast to the alkaline sensors

509 that can take different forms.

510 The role of somatostatin CSF-c neurons in the brain and spinal cord

511 In the lamprey brain, the main source of somatostatin is the hypothalamic CSF-c neurons,

512 although small populations exist in the ventral thalamus and the isthmic region, and of course

513 in the cells around the central canal (Buchanan et al., 1987; Wright, 1986; Yanez et al., 1992,

514 Christenson et al., 1991; unpubl. observations). Periventricular somatostatin-expressing

515 neurons are highly concentrated along the third ventricle also in other vertebrates, including

516 rat (Kronheim et al., 1976; Finley et al., 1981; Johansson et al., 1984), reptiles (Fasolo and

517 Gaudino, 1982; Wang et al., 2016), amphibians (Fasolo and Gaudino, 1981) and fish (Vigh-

518 Teichmann et al., 1983b; Sas and Maler, 1991). The lamprey pallium and striatum have a rich

519 somatostatin innervation originating from CSF-c neurons in hypothalamus (Suryanarayana et

520 al., 2017), but no intrinsic somatostatin interneurons, in contrast to the mammalian cortex and

521 striatum (see Liguz-Lecznar et al., 2016).

522 The hypothalamic somatostatin CSF-c neurons innervate most parts of the brain,

523 however, with a higher density of fibers and/or terminals in motor areas, including the output

524 layer of the optic tectum and the reticulospinal nuclei, but not in sensory areas. Interestingly,

525 somatostatin-expressing CSF-c neurons at the obex level, which co-express dopamine also

526 respond to pH changes, whereas dopaminergic CSF-c neurons that do not express

527 somatostatin are unresponsive. This further strengthens the notion of a functional link

528 between somatostatin expression in CSF-c neurons and pH sensitivity. In the lamprey spinal

529 cord, the suppression of motor activity induced by activation of the somatostatin/GABA-

530 expressing CSF-c cells following a change of extracellular pH, is mediated by a release of

531 somatostatin (Jalalvand et al., 2016a, b). The hypothalamic CSF-c neurons may well have a

532 similar role.

533 In conclusion

534 Taken together, somatostatin-expressing CSF-c neurons in the hypothalamus, like in the

535 spinal cord, may serve as pH sensors as well as mechanosensors. They will sense any pH-

536 deviation and motion in the cerebrospinal fluid of the third ventricle and respond by

537 membrane potential depolarization and action potential firing. The acidic response and the

538 mechanosensitivity is mediated by ASIC3-like channels, while the alkaline response appears

539 to be mediated through connexin hemichannels, in contrast to their spinal counterparts in

540 which PKD2L1 is responsible. Activation of hypothalamic CSF-c neurons will cause a release

541 of somatostatin in the different target areas, many of which are motor related. As in the spinal

542 cord, this may contribute to restore the deviation in pH and thus to homeostasis.

543

544 REFERENCES

545 Ball SG (2007) Vasopressin and disorders of water balance: the physiology and

546 pathophysiology of vasopressin. Ann Clin Biochem 44:417-431.

547 Boulant JA, Dean JB (1986) Temperature receptors in the central nervous system. Annu

548 Rev Physiol 48:639-654.

549 Brazeau P, Vale W, Burgus R, Ling N, Butcher M, Rivier J, Guillemin R (1973)

550 Hypothalamic polypeptide that inhibits the secretion of immunoreactive pituitary

551 growth hormone. Science 179:77-79.

552 Broberger C (2005) Brain regulation of food intake and appetite: molecules and

553 networks. J Intern Med 258:301-327.

554 Brodin L, Hökfelt T, Grillner S, Panula P (1990) Distribution of histaminergic neurons in the

555 brain of the lamprey Lampetra fluviatilis as revealed by histamine-

556 immunohistochemistry. J Comp Neurol 292:435-442.

557 Buchanan JT, Brodin L, Hökfelt T, Van Dongen PA, Grillner S (1987) Survey of

558 neuropeptide-like immunoreactivity in the lamprey spinal cord. Brain Res 408:299-

559 302.

560 Chen, C.H., Hsu, Y.T., Chen, C.C., Huang, R.C. (2009) Acid-sensing ion channels in

561 neurones of the rat suprachiasmatic nucleus. J Physiol 587, 1727–173.

562 Chesler M (1986) Regulation of intracellular pH in reticulospinal neurones of the lamprey,

563 Petromyzon marinus. J Physiol 381:241-261.

564 Chesler M, Kaila K (1992). Modulation of pH by neuronal activity. Trends Neurosci 15:396-

565 402.

566 Christenson J, Alford S, Grillner S, Hökfelt T (1991) Co-localized GABA and somatostatin

567 use different ionic mechanisms to hyperpolarize target neurons in the lamprey spinal

568 cord. Neurosci Lett 134:93-97.

569 Contreras JE, Sanchez HA, Eugenin EA, Speidel D, Theis M, Willecke K, Bukauskas FF,

570 Bennett MV, Saez JC (2002) Metabolic inhibition induces opening of unapposed

571 connexin 43 gap junction hemichannels and reduces gap junctional

572 communication in cortical astrocytes in culture. Proc Natl Acad Sci USA

573 99:495-500.

574 Diochot S, Baron A, Rash LD, Deval E, Escoubas P, Scarzello S, Salinas M, Lazdunski M

575 (2004) A new sea anemone peptide, APETx2, inhibits ASIC3, a major acid-sensitive

576 channel in sensory neurons. EMBO J 23:1516-1525.

577 DiMicco JA, Zaretsky DV (2007) The dorsomedial hypothalamus: a new player in

578 thermoregulation. Am J Physiol Regul Integr Comp Physiol 292:R47-63.

579 Fasolo A, Gaudino G (1981) Somatostatin immunoreactive neurons and fibers in the

580 hypothalamus of the newt. Gen Comp Endocrinol 44:256-263.

581 Fasolo A, Gaudino G (1982) Immunohistochemical localization of somatostatin-like

582 immunoreactivity in the hypothalamus of the lizard, Lacerta sicula. Gen Comp

583 Endocrinol 48:205-212.

584 Finley JC, Maderdrut JL, Roger LJ, Petrusz P (1981) The immunocytochemical

585 localization of somatostatin-containing neurons in the rat central nervous system.

586 Neuroscience 6:2173-2192.

587 Francis D, Stergiopoulos K, Ek-Vitorin JF, Cao FL, Taffet SM, Delmar M (1999)

588 Connexin diversity and gap junction regulation by pHi. Dev Genet 24:123-136.

589 Hofman MA, Swaab DF (1993) Diurnal and seasonal rhythms of neuronal activity in the

590 suprachiasmatic nucleus of humans. J Biol Rhythms 8:283-295.

591 Hong Z, Wang DS, Lu SY, Li JS (2004) Suppression of kainate-evoked AMPA receptor

592 mediated responses by lanthanum in rat sacral dorsal commissural neurons.

593 Neurosignals 13:258-264.

594 Huang AL, Chen X, Hoon MA, Chandrashekar J, Guo W, Trankner D, Ryba NJ, Zuker CS

595 (2006) The cells and logic for mammalian sour taste detection. Nature 442:934-938.

596 Jalalvand E, Robertson B, Wallén P, Grillner S (2016a) Ciliated neurons lining the central

597 canal sense both fluid movement and pH through ASIC3. Nat Commun 7:10002.

598 Jalalvand E, Robertson B, Tostivint H, Wallén P, Grillner S (2016b) The Spinal Cord Has

599 an Intrinsic System for the Control of pH. Curr Biol 26:1346-1351.

600 Jalalvand E, Robertson B, Wallén P, Hill RH, Grillner S (2014) Laterally projecting

601 cerebrospinal fluid-contacting cells in the lamprey spinal cord are of two distinct

602 types. J Comp Neurol 522:1753-1768.

603 Johansson O, Hökfelt T, Elde RP (1984) Immunohistochemical distribution of somatostatin-

604 like immunoreactivity in the central nervous system of the adult rat. Neuroscience

605 13:265-339.

606 Kronheim S, Berelowitz M, Pimstone BL (1976) A radioimmunoassay for growth

607 hormone release-inhibiting hormone: method and quantitative tissue distribution.

608 Clin Endocrinol (Oxf) 5:619-630.

609 Levin LR, Buck J (2015) Physiological roles of acid-base sensors. Annu Rev Physiol

610 77:347-362.

611 Liguz-Lecznar M, Urban-Ciecko J, Kossut M (2016) Somatostatin and somatostatin-

612 containing neurons in shaping neuronal activity and plasticity. Front Neural

613 10:48.

614 Magistretti PJ, Allaman I (2015) A cellular perspective on brain energy metabolism and

615 functional imaging. Neuron 86:883-901.

616 Magistretti PJ, Allaman I (2018) Lactate in the brain: from metabolic end-product to signaling

617 molecule. Nature Rev Neurosci 19:235-249.

618 Marichal N, Garcia G, Radmilovich M, Trujillo-Cenoz O, Russo RE (2009) Enigmatic central

619 canal contacting cells: immature neurons in "standby mode"? J Neurosci 29:10010-

620 10024.

621 Meng QY, Wang W, Chen XN, Xu TL, Zhou JN (2009) Distribution of acid-sensing ion

622 channel 3 in the rat hypothalamus. Neuroscience 159:1126-1134.

623 Murayama T, Maruyama IN (2015) Alkaline pH sensor molecules. J Neurosci Res

624 93:1623-1630.

625 O'Brien J, Bruzzone R, White TW, Al-Ubaidi MR, Ripps H (1998) Cloning and

626 expression of two related connexins from the perch retina define a distinct

627 subgroup of the connexin family. J Neurosci 18:7625-7637.

628 O’Brian J, Nguyen HB, Mills SL (2004) Cone photoreceptors in bass retina use two

629 connexins to mediate electrical coupling. J Neurosci 24:5632-5642.

630 Orts-Del'immagine A, Wanaverbecq N, Tardivel C, Tillement V, Dallaporta M, Trouslard J

631 (2012) Properties of subependymal cerebrospinal fluid contacting neurones in the

632 dorsal vagal complex of the mouse brainstem. J Physiol 590:3719-3741.

633 Orts-Del'Immagine A, Kastner A, Tillement V, Tardivel C, Trouslard J, Wanaverbecq N

634 (2014) Morphology, distribution and phenotype of polycystin kidney disease 2-like

635 1-positive cerebrospinal fluid contacting neurons in the brainstem of adult mice.

636 PLoS One 9:e87748.

637 Orts-Del'Immagine A, Seddik R, Tell F, Airault C, Er-Raoui G, Najimi M, Trouslard J,

638 Wanaverbecq N (2016) A single polycystic kidney disease 2-like 1 channel opening

639 acts as a spike generator in cerebrospinal fluid-contacting neurons of adult mouse

640 brainstem. Neuropharmacology 101:549-565.

641 Rehncrona S (1985) Brain acidosis. Ann Emergen Medicine 14: 770-93. 642 Reichling DB, MacDermott AB (1991) Lanthanum actions on excitatory amino acid-gated

643 currents and voltage-gated calcium currents in rat dorsal horn neurons. J

644 Physiol 441:199-218.

645 Russo RE, Reali C, Radmilovich M, Fernandez A, Trujillo-Cenoz O (2008) Connexin 43

646 delimits functional domains of neurogenic precursors in the spinal cord. J Neurosci

647 28:3298-3309.

648 Ruusuvuori E, Kaila K (2014) Carbonic anhydrases and brain pH in the control of

649 neuronal excitability. Subcell Biochem 75:271-290.

650 Saez JC, Retamal MA, Basilio D, Bukauskas FF, Bennett MV (2005) Connexin-based gap

651 junction hemichannels: gating mechanisms. Biochim Biophys Acta 1711:215-224.

652 Saper CB, Chou TC, Elmquist JK (2002) The need to feed: homeostatic and hedonic

653 control of eating. Neuron 36:199-211.

654 Saper CB, Scammell TE, Lu J (2005) Hypothalamic regulation of sleep and circadian

655 rhythms. Nature 437:1257-1263.

656 Sas E, Maler L (1991) Somatostatin-like immunoreactivity in the brain of an electric fish

657 (Apteronotus leptorhynchus) identified with monoclonal antibodies. J Chem

658 Neuroanat 4:155-186.

659 Schalper KA, Sanchez HA, Lee SC, Altenberg GA, Nathanson MH, Saez JC (2010)

660 Connexin 43 hemichannels mediate the Ca2+ influx induced by extracellular

661 alkalinization. Am J Physiol Cell Physiol 299:C1504-1515.

662 Siesjö BK, von Hanwehr R, Nergelius G, Nevander G, Ingvar M (1985) Extra- and

663 Intracellular pH in the brain during seizures and in the recovery period following the

664 arrest of seizure activity. J Cerebral Blood Flow and Metabolism 5: 47-57.

665 Spray DC, Ye ZC, Ransom BR (2006) Functional connexin "hemichannels": a critical

666 appraisal. Glia 54:758-773.

667 Srinivas M, Rozental R, Kojima T, Dermietzel R, Mehler M, Condorelli DF, Kessler JA,

668 Spray DC (1999) Functional properties of channels formed by the neuronal gap

669 junction protein connexin36. J Neurosci 19:9848-9855.

670 Suryanarayana SM, Robertson B, Wallén P, Grillner S (2017) The Lamprey Pallium Provides

671 a Blueprint of the Mammalian Layered Cortex. Curr Biol 27:3264-3277 e3265.

672 Thorén L (1960) Vätskebalans. Almqvist & Wiksells, Boktryckeri Aktiebolag. Uppsala,

673 Sweden.

674 Tufts BL, Bagatto B, Cameron B (1992) In vivo analysis of gas transport in arterial and

675 venous blood of the sea lamprey Petromyzon marinus. J exp Biol 169:105-119.

676 Vigh B, Vigh-Teichnann I, Sikora K, Jennes L (1980) Scanning electron microscopy of the

677 cerebrospinal fluid-contacting neurons of the central canal in various vertebrate.

678 Verh Anat Ges 74:707-714.

679 Vigh B, Manzano e Silva MJ, Frank CL, Vincze C, Czirok SJ, Szabo A, Lukats A, Szel A

680 (2004) The system of cerebrospinal fluid-contacting neurons. Its supposed role in

681 the nonsynaptic signal transmission of the brain. Histol Histopathol 19:607-628.

682 Vigh B, Vigh-Teichmann I (1998) Actual problems of the cerebrospinal fluid-contacting

683 neurons. Microsc Res Tech 41:57-83.

684 Vigh B, Vigh-Teichmann I, Aros B (1977) Special dendritic and axonal endings formed

685 by the cerebrospinal fluid contacting neurons of the spinal cord. Cell Tissue Res

686 183:541-552.

687 Vigh-Teichmann I, Vigh B (1983a) The system of cerebrospinal fluid-contacting neurons.

688 Arch Histol Jpn 46:427-468.

689 Vigh-Teichmann I, Vigh B, Korf HW, Oksche A (1983b) CSF-contacting and other

690 somatostatin-immunoreactive neurons in the brains of Anguilla anguilla,

691 Phoxinus phoxinus, and Salmo gairdneri (Teleostei). Cell Tissue Res 233:319- 334.

692 Voipio J, Kaila K (1993) Interstitial PCO2 and pH in rat hippocampal slices measured by

693 means of a novel fast CO2/H(+)-sensitive microelectrode based on a PVC-gelled

694 membrane. Pflugers Arch 423:193-201.

695 Wang W, Yu Y, Xu TL (2007) Modulation of acid-sensing ion channels by Cu2+ in

696 cultured hypothalamic neurons of the rat. Neuroscience 145:631-641.

697 Wang H, Zhang R, Zhang S, Zhou Y, Wu X (2016) Immunohistochemical localization of

698 somatostatin in the brain of chinese alligator Alligator sinensis. Anat Rec

699 (Hoboken) 300:507-519.

700 Wright GM (1986) Immunocytochemical demonstration of growth hormone, prolactin and

701 somatostatin-like immunoreactivities in the brain of larval, young adult and

702 upstream migrant adult sea lamprey, Petromyzon marinus. Cell Tissue Res

703 246:23-31.

704 Yanez J, Rodriguez-Moldes I, Anadon R (1992) Distribution of somatostatin-

705 immunoreactivity in the brain of the larval lamprey (Petromyzon marinus). J

706 Chem Neuroanat 5:511-520.

707 Yu J, Bippes CA, Hand GM, Muller DJ, Sosinsky GE (2007) Aminosulfonate modulated

708 pH-induced conformational changes in connexin26 hemichannels. J Biol Chem

709 282:8895-8904.

710 Zhao PY, Gan G, Peng S, Wang SB, Chen B, Adelman RA, Rizzolo LJ (2015) TRP

711 channels localize to subdomains of the apical plasma membrane in human fetal

712 retinal pigment epithelium. Invest Ophthalmol Vis Sci 56:1916-1923.

713

714 FIGURE LEGENDS

715 Figure 1. Ciliated CSF-c neurons in the hypothalamus express GABA and somatostatin. A,

716 The location of somatostatin-expressing CSF-c neurons (red dots) indicated in schematic

717 drawings of transverse sections through the hypothalamic area. B, Photomicrograph of

718 somatostatin-immunopositive CSF-c neurons in the hypothalamus (arrows). Scale bar, 200

719 μm. C, High-magnification photomicrograph of the boxed area in C. CSF-c neurons

720 expressing somatostatin have short, thick apical processes with a bulb-like ending protruding

721 into the ventricle (arrows). Scale bar, 20 μm. D, Confocal image of a somatostatin-expressing

722 hypothalamic CSF-c neuron (green) with an α-tubulin-immunoreactive cilium (magenta,

723 arrow). Scale bar, 5 μm. E, Somatostatin (magenta) and GABA (green) are co-localized in

724 hypothalamic CSF-c neurons (arrows). Some of the CSF-c neurons only expressed GABA

725 (arrowhead). Scale bar, 20 μm. F, Illustration of a reconstructed intracellularly labeled

726 hypothalamic CSF-c neuron with a bulb-like ending protruding into the ventricle and axonal

727 branches extending dorsally, laterally and ventrally. G, Somatostatin-immunoreactive

728 fibers/terminals at the level of the optic tectum. Note the dense innervation of the deep layer

729 (DL) and absence of labeling in the retinorecipient superficial layer (SL). Scale bar, 400 μm.

730 H, Rich somatostatin labeling in the ventral brainstem at the level of MRRN and the

731 trigeminal nucleus (nV). Scale bar, 200 μm. Th, thalamus; DL, deep layer; cpo, postoptic

732 commissure; Hb, habenula; Hyp, hypothalamus; LPal, lateral pallium; MPal, medial pallium;

733 MRRN, middle rhombencephalic reticular nucleus; ncpo, nucleus of the postoptic

734 commissure; nh, neurohypophysis; ot, optic tract; SL, superficial layer.

735 Figure 2. Hypothalamic CSF-c neurons express connexin. A, Intracellular injection of

736 Neurobiotin in a hypothalamic CSF-c neuron (patched cell) resulted in two filled neurons

737 revealed by DAB-staining. Scale bar, 20 μm. B, Somatostatin-expressing hypothalamic CSF-

738 c neurons. Scale bar, 20 μm. C, Connexin 35/36-immunoreactivity in the same hypothalamic

739 area as (B). Scale bar, 20 μm. D, Merged image showing connexin 35/36-immunoreactivity

740 (arrows) co-expressed with somatostatin in hypothalamic CSF-c neurons. Scale bar, 20 μm.

741 E, Westen blot of a lamprey brain extract detected with a monoclonal mouse anti-connexin

742 35/36 antibody, showing a distinct band around 35kD.

743 Figure 3. Electrophysiological properties of hypothalamic CSF-c neurons. A, Whole-cell

744 current clamp recording of a hypothalamic CSF-c neuron showing its firing pattern. A brief

745 current injection (20 pA, 20 ms; left) elicited a single action potential while longer current

746 injections (20 pA, 500 ms; right) generated repetitive firing. B, A CSF-c neuron showing

747 spontaneous GABA- and glutamate-mediated postsynaptic potentials (upper trace) that were

748 successively blocked by gabazine (20 μM), CNQX (40 μM) and AP5 (50 μM; lower traces),

749 respectively. C, Voltage responses to 12 consecutive hyperpolarizing and depolarizing current

750 injections. The red and green traces illustrate a single action potential and spike frequency

751 adaptation evoked by depolarizing steps, respectively. D, I-V curve showing a linear current-

752 voltage relationship.

753 Figure 4. Hypothalamic CSF-c neurons are activated by both acidic and alkaline pH. A, A

754 change to acidic (pH 6.5) as well as to alkaline (pH 8.0) extracellular pH (gray area)

755 depolarized the membrane potential (10-12 mV) and triggered action potentials. Upon return

756 to pH 7.4, firing ceased and the membrane potential repolarized back to the control value (-63

757 mV). In this and all subsequent recordings, gabazine (20 μM), AP5 (50 μM) and CNQX (40

758 μM) were applied to exclude any indirect, synaptically mediated effects. B, In the presence of

759 TTX (1.5 μM), both acidic and alkaline pH resulted in depolarization of the membrane

760 potential that recovered after return to pH 7.4. C, Photomicrographs of the CSF-c neuron

761 recorded in A, after intracellular labeling with Neurobiotin (green, arrow), showing that the

762 cell expressed somatostatin (magenta). Scale bars, 20 μm. D, Mean membrane potential

763 changes in hypothalamic CSF-c neurons for each pH condition (Mean ± SD; n=15; Student’s

-19 764 paired t test: *** p < 0.001 significant difference vs pH 7.4 at 6.5 (p = 1.43 x 10 , t14 = 22.7),

-18 -14 765 at 6.8 (p = 5.56 x 10 , t14 = 19.76), at 7.1 (p = 1.97 × 10 , t14 = 14.35), at 7.7 (p = 2.20 ×

-13 -20 -21 766 10 , t14 = 13.0), at 8.0 (p = 2.24 x 10 , t14 = 24.33), and at 8.3 (p = 1.38 x 10 , t14 =

767 26.99)). E, In voltage-clamp mode, frequent inward current deflections appeared at pH 6.5

768 and 8.0 with a maximal amplitude of about 10 pA. F, Mean frequency increase of events (5-

769 15 pA) in response to acidic and alkaline pH (Mean ±SEM; n=7; Student’s paired t test: *** p

-7 770 < 0.001 significant difference vs pH 7.4 at pH 6.5 (p = 2.34 × 10 ; t6 = 25.59) and at 8 (p =

-8 771 4.48 × 10 ; t6 = 33.79)). G, Unitary current deflections, recorded at acidic and alkaline pH.

772 TTX (1.5 μM) present in B, E, F and G.

773 Figure 5. The response to acidic and alkaline pH is mediated by different channels. A,

774 Whole-cell current clamp in the presence of gabazine (20 μM), AP5 (50 μM), and CNQX (40

775 μM). Application of APETx2 (1μM) abolished the response to acidic pH but not to alkaline

776 pH in the same cell. B, Recordings from a hypothalamic CSF-c neuron in voltage-clamp

777 mode. No current events were seen at pH 7.4 (black traces) in the presence of gabazine (20

778 μM), AP5 (50 μM), CNQX (40 μM) and TTX (1.5 μM). Inward current deflections appeared

779 after a decrease or increase in the extracellular pH. The current events recorded in acidic pH

780 (6.5) were completely blocked in the presence of APETx2 (1μM; red trace), whereas those

781 recorded in alkaline pH (8.0; blue trace) remained. C, Mean frequency increase of events (5-

782 15 pA) in response to acidic and alkaline pH in the presence of APETx2. (Mean ± SEM; n=3;

783 Student’s paired t test: no significant difference vs pH 7.4 at 6.5 (p = 1, t2 = 0), but a

784 significant difference vs pH 7.4 at 8 (** p < 0.01, p = 0.007, t2 = 11.55)). D, Hypothalamic

785 somatostatin-immunopositive CSF-c neurons (green) do not express the PKD2L1 channel

786 (magenta). Scale bar, 20 μm. E, Spinal somatostatin-CSF-c neurons (green) co-express

787 PKD2L1 (arrows; magenta). Scale bar, 20 μm. cc, central canal. F, Whole-cell current clamp

788 in the presence of gabazine (20 μM), AP5 (50 μM), and CNQX (40 μM), showing the

789 response to both acidic (pH 6.5; red trace) and alkaline (pH 8.0; blue trace) pH. The connexin

790 hemichannel blocker lanthanum (100 μM) abolished the alkaline response but not the acidic

791 response. G, Mean frequency increase of action potential firing in response to acidic and

792 alkaline pH before and after application of lanthanum chloride (100 and 70 μM). Before

793 application (control): Means ± SEM; n=5; Student’s paired t test: *** p < 0.001 significant

-4 -5 794 difference vs pH 7.4 at 6.5 (p = 1.31 × 10 , t4 = 14.51), and at 8 (p = 9.43 × 10 , t4 = 15.77).

795 In the presence of lanthanum, a complete (100 μM) or partial (70 μM) blockade of the spiking

796 response to alkaline pH was seen. Means ± SEM; n=5; Student’s paired t test: *** p < 0.001

797 significant difference at pH 8 vs control in the presence of lanthanum at 100 μM (p = 9.4 ×

-5 -4 798 10 , t4 = 15.77; as well as at 70 μM (p = 9.3 × 10 , t4 = 8.76). A tendency for a small, non-

799 significant (n.s.) frequency reduction was also observed at acidic pH (6.5) vs control in the

800 presence of lanthanum at 100 μM (p = 0.06, t4 = 2.58). H, The recorded CSF-c neuron in F,

801 intracellularly labeled with Neurobiotin, expressed somatostatin. Scale bars, 10 μm. I, Whole-

802 cell current clamp of a hypothalamic CSF-c neuron in control condition and in the presence of

803 gabazine (20 μM), AP5 (50 μM), and CNQX (40 μM), showing that this cell did not respond

804 to either acidic (pH 6.5; red trace) or alkaline (pH 8.0; blue trace) pH. J, The recorded CSF-c

805 neuron in I, intracellularly labeled with Neurobiotin, did not express somatostatin. Scale bars,

806 10 μm.

807 Figure 6. The same hypothalamic CSF-c neuron is sensitive to both fluid movement and pH

808 changes. A, In vitro preparation of the hypothalamus with CSF-c neurons protruding into the

809 third ventricle. An aCSF-filled pressure pipette was placed close to a bulb-like ending of a

810 recorded hypothalamic CSF-c neuron. Scale bar, 20 μm. B, A short fluid-pulse (80 ms)

811 elicited receptor potential responses and action potentials while holding the membrane

812 potential at -65 mV and -55 mV, respectively, in the presence of gabazine (20 μM), CNQX

813 (40 μM) and AP5 (50 μM). C, The receptor potential elicited by fluid pulse stimulation (20

814 p.s.i., 80 ms) was blocked by application of the ASIC3 blocker APETx2 (1 μM). D, Complete

815 blockade of responses after application of APETx2 (n=4). E, In the same hypothalamic CSF-c

816 neuron as in C, exposure to acidic (pH 6.5) as well as alkaline (pH 8.0) pH depolarized the

817 membrane potential (10-12 mV) and triggered action potentials (GABA and glutamate

818 receptor antagonists present).

819