Hypothalamic ependymal-glial cells express the glucose transporter GLUT2, a protein involved in glucose sensing


Address correspondence and reprint requests to Dr Francisco Nualart, Laboratorio de Neurobiología Celular, Departamento de Histología y Embriología, Facultad de Ciencias Biológicas, Universidad de Concepción, casilla 160C, Chile. E-mail: fnualart@udec.cl


The GLUT2 glucose transporter and the K-ATP-sensitive potassium channels have been implicated as an integral part of the glucose-sensing mechanism in the pancreatic islet β cells. The expression of GLUT2 and K-ATP channels in the hypothalamic region suggest that they are also involved in a sensing mechanism in this area. The hypothalamic glial cells, known as tanycytes α and β, are specialized ependymal cells that bridge the cerebrospinal fluid and the portal blood of the median eminence. We used immunocytochemistry, in situ hybridization and transport analyses to demonstrate the glucose transporters expressed in tanycytes. Confocal microscopy using specific antibodies against GLUT1 and GLUT2 indicated that both transporters are expressed in α and β tanycytes. In addition, primary cultures of mouse hypothalamic tanycytes were found to express both GLUT1 and GLUT2 transporters. Transport studies, including 2-deoxy-glucose and fructose uptake in the presence or absence of inhibitors, indicated that these transporters are functional in cultured tanycytes. Finally, our analyses indicated that tanycytes express the K-ATP channel subunit Kir6.1 in vitro. As the expression of GLUT2 and K-ATP channel is linked to glucose-sensing mechanisms in pancreatic β cells, we postulate that tanycytes may be responsible, at least in part, for a mechanism that allows the hypothalamus to detect changes in glucose concentrations.

Abbreviations used



nitric oxide synthase-brain


cytokeratin of high molecular weight


glial-fibrillary acidic protein


glucose transporter


von Willebrand factor


luteinizing hormone-releasing hormone

K-ATP, channels

K-ATP-sensitive potassium channels


myelin basic protein


neurotrophic growth factor receptor



The hypothalamus is thought to be involved in modulating feeding behavior and corporal growth through its ability to detect changes in circulating glucose (Schwartz et al. 2000). The hypothesis that the hypothalamus is able to detect changes in glucose requires the identification of cells involved in this process, as well as the expression and secretion of the key molecules that participate in the hypothalamic glucose-sensing mechanism (Oomura et al. 1969; Levin et al. 2001). The glucose-sensing mechanism by pancreatic β cells involves a number of molecules, including GLUT2, glucokinase, glucagon-like peptide-1 receptors and the ATP-sensitive K+ channels (Guillam et al. 1997; Schuit et al. 2001). The finding that ventricular hypothalamic glial cells (tanycytes) express glucose sensor molecules suggests that they may be responsible, at least in part, for glucose sensing by the hypothalamus (Alvarez et al. 1996; Navarro et al. 1996; Thomzig et al. 2001).

The hypothalamic peri-ventricular neurons are grouped in nuclei (arcuate nucleus) that are in close contact with highly elongated ependymal cells, namely tanycytes (Flament-Durand and Brion 1985; Chauvet et al. 1995) (Fig. 1). Different studies using the Golgi impregnation method, or analysis by electron microscopy (Akmayev and Fidelina 1974), identified two main types of tanycytes, β1 and β2. The β1 tanycytes are present in the lateral lower part of the third ventricle and they are capable of developing cell processes that contact neurons of the arcuate nucleus as well as the blood capillary vessels in the hypothalamus (Chauvet et al. 1995; Peruzzo et al. 2000; García et al. 2001). The end feet of the cells processes reach the lateral sulcus of the infundibular region contacting luteinizing hormone-releasing hormone (LH-RH) terminals, which are involved in hormone release to the hypophyseal portal vessels.

Figure 1.

Schematic representation of mouse hypothalamus. The cells that are directly identified are: 1, ciliated ependymocytes lining the rostral wall of the third ventricle, 2, α tanycytes located in the dorsal lateral wall of the third ventricle, 3, β1 tanycytes located in the lower lateral wall of the third ventricle, 4, β2 tanycytes located in the median eminence. The cells form the median eminence–cerebrospinal fluid barrier (thick line). The projections of β1 and β2 tanycytes contact the portal blood vessels of the median eminence and pars tuberalis that are characterized by the absence of a blood–brain barrier. There is experimental evidence that tanycytes and neurons of the lower lateral wall of the third ventricle express glucokinase, glucagon-like peptide-1 receptor (GLP-1), and ATP-sensitive K+ channels; molecules involved in the glucose sensing mechanism. V-, portal blood vessels without blood–brain barrier.

The β2 tanycytes are located in the floor of third ventricle lining the median eminence (Fig. 1). The proximal part of the cells is in contact with the cerebrospinal fluid of the third ventricle, while the dorsal part of the cells forms processes in which the end feet reaches the pial surface of the brain or the local capillary plexus in the median eminence (Chauvet et al. 1995; Peruzzo et al. 2000; García et al. 2001). These cells develop tight-junctions that form the cerebrospinal fluid–median eminence barrier. The existence of two additional types of tanycytes, α1 and α2, has also been proposed (Akmayev and Fidelina 1974) (Fig. 1). These cells line the dorsal and lateral walls of the third ventricle mainly facing the ventromedial nucleus of the hypothalamus. The functions of tanycytes remain a matter of controversy and speculation. Originally, it was suggested that they might represent a link between the cerebrospinal fluid and the portal vessels, and that they could be involved in regulating neuroendocrine function (Rodríguez et al. 1985). It has also been suggested that tanycytes are involved in uptake and transport but the nature of the material that might be transported remains to be discovered (Flament-Durand and Brion 1985; García et al. 2001).

Differential gene expression of facilitative glucose transporters (GLUT1–13) mediates the uptake of hexoses in mammalian cells (Joost et al. 2002). Experiments designed to identify the precise cellular localization of these transporters indicate that GLUT1 and GLUT3 are the main isoforms expressed in brain (Kalaria et al. 1988; Vannucci 1994; Gerhart et al. 1995; Nualart et al. 1999). GLUT2 is a low-affinity transporter for glucose and fructose expressed at low levels in different regions of the brain (Brant et al. 1993). Because of this low affinity for glucose, but high transport capacity, GLUT2 is believed to play a major role in glucose-sensing mechanisms (Guillam et al. 1997, 2000). In situ hybridization data suggest that GLUT2 is present in hypothalamus (Navarro et al. 1996). Immunohistochamical analysis indicates GLUT2 expression in astrocytes-like cells (Leloup et al. 1994), however, GLUT2 has also been detected in ependymal cells (Ngarmukos et al. 2001). Maekawa et al. (2000) confirmed the expression of GLUT2 in ependymal cells of the dorsal third ventricle and cerebral aqueduct, but the hypothalamic ependymal cells (tanycytes) were negative. Thus, the exact localization of the specific glial cell types that express GLUT2 in the hypothalamus remains to be determined.

Here, we report that tanycytes are the main glial cells in the periventricular zone of the hypothalamus. Expression analysis of the glucose transporters in mouse hypothalamic tanycytes revealed that GLUT1 and GLUT2 are expressed in both α and β tanycytes. The functional properties of GLUT1 and GLUT2 transporters were demonstrated in tanycytes isolated from mouse hypothalamus. In addition to GLUT1 and GLUT2, we observed that tanycytes in culture express the pore-forming subunit Kir6.1, an essential molecule in the formation of K-ATP-sensitive potassium (K-ATP) channels.

Materials and methods

Immunocytochemistry and confocal microscopy

Mice (C57BL/J6) brains were dissected and fixed immediately by immersion in Bouin's solution. Fixation in situ was performed by vascular perfusion (Nualart et al. 1991). Samples were dehydrated in graded alcohol solutions and embedded in paraffin. Frontal sections (5 μm) of the hypothalamic area were mounted on poly l-lysine-coated glass slides.

For immunohistochemical analyses, we used a panel of affinity-purified antibodies raised against synthetic peptides encompassing the last 10–13 carboxy terminal amino acid of each isoform of the human facilitative hexose transporters (GLUT1 to GLUT5, Alpha Diagnostic, San Antonio, TX, USA). Sections were incubated, overnight at room temperature (22°C) in a humid chamber, with anti-GLUT polyclonal antibodies (1 : 200–1 : 1000) diluted in a Tris-HCl buffer (pH 7.8) containing 8.4 mm sodium phosphate, 3.5 mm potassium phosphate, 120 mm NaCl and 1% bovine serum albumin. After washing extensively, the sections were incubated for 2 h with Cy2-conjugated affinity-purified donkey anti-rabbit IgG (1 : 200; Jackson Immuno-Research, West Grove, PA, USA) at room temperature. Alternatively, anti-mouse IgG (1 : 50; Dako, Carpinteria, CA, USA) labeled with fluorescein isothiocyanate (1 : 30; Dako) was used as a secondary antibody. For confocal laser microscopy analysis, the tissue sections were incubated with propidium iodine in the absence of RNAase for cellular staining. As negative controls for GLUT1 and GLUT2, we utilized both primary antibodies pre-absorbed with the respective peptides used to elicit them, and pre-immune serum. To characterize glial cell distribution in the hypothalamic area, serial tissue sections were immunostained using an anti-glial fibrillary acidic protein (GFAP) polyclonal antibody (1 : 100; Dako).

In situ hybridization

A cDNA of approximately 2.2 kb subcloned in pGEM-4Z (Clontech, Palo Alto, CA, USA) and encoding the human GLUT2 was used to generate sense and anti-sense digoxigenin-labeled riboprobes. RNA probes were labeled with digoxigenin-UTP by in vitro transcription with SP6 or T7 RNA polymerase following the manufacturer's procedure (Boehringer Mannheim, Mannheim, Germany). In situ hybridization was performed on hypothalamic frontal sections mounted on poly l-lysine-coated glass slides. The sections were baked at 60°C for 1 h, deparaffinized in xylene, and rehydrated in graded ethanol. Following proteinase K treatment (5 min at 37°C in PBS, 1 μg/mL proteinase K), the tissue sections were fixed with 4% p-formaldehyde for 5 min at 4°C, washed in cold PBS and then acetylated in 0.1 m triethanol amine-HCl (pH 8.0) at room temperature for 10 min. After a brief wash, the sections were incubated in pre-hybridization solution for 15 min at 37°C, and then 25 μL of hybridization mix [50% formamide, 0.6 m NaCl, 10 mm Tris-HCl (pH 7.5), 1 mm EDTA, 1 × Denhart's solution, 10% PEG 8000, 10 mm DTT, 500 μg yeast tRNA/mL, 50 μg/mL heparin, 500 μg/mL DNA carrier, and 1 : 20–1 : 100 diluted riboprobe] were added to each slide. The slides were covered with glass coverslips and placed in a humidified chamber at 42°C overnight. After removal of the coverslip, the slides were rinsed in 4 × SSC and washed twice for 30 min at 42°C. The slides were washed at 37°C for 30 min each in 2 × SSC, 1 × SSC and 0.3 × SSC. Visualization of digoxigenin was performed by incubation with a monoclonal antibody coupled to alkaline phosphatase (anti-digoxigenin-alkaline phosphatase Fab fragments diluted 1 : 500; Boehringer Mannheim) for 2 h at room temperature. Nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl-phosphate (Boehringer Mannheim) were used as substrates for the alkaline phosphatase. Controls included use of the sense riboprobe and omission of the probe.

Tissue culture

Ependymal cells

Tissue culture experiments used cells obtained from mice at 19 days of gestation (C57BL/J6) (Gabrion et al. 1988; Chauvet et al. 1996). The brain was removed, the hypothalamic area taken out, and further dissected until obtaining a region close to the ependymal layer. The dissection was done with the samples submerged in 10 mm Hepes (pH 7.3) containing 10 mm glucose, 44 mm sacarose, 135 mm NaCl, 5 mm KCl, 0.15 mm Na2HPO4 (340–350 mOsm/L). Ventricular walls were incubated with 0.1% (w/v) trypsin for 15 min at 37°C. Trypsinized tissue was transferred to a 15-mL culture tube containing 10 mL of minimal essential medium (MEM, Gibco Co., Rockville, NY, USA) with 10% (v/v) of fetal bovine serum (FBS) to stop trypsin action. The tissue was dissociated by trituration through a siliconized Pasteur pipette, until a single cell suspension was obtained. Cells were centrifuged for 5 min at 200 g, the supernatant was aspirated off and the cells were resuspended in MEM and seeded in culture dishes at a concentration of 0.5–1.0 × 106 cells by dish. Culture medium was supplemented with 10% FBS, 4 mm l-glutamine, 100 U/mL penicillin and 100 μg/mL streptomycin (Nalgene, Rochester, NY, USA). Cells were cultured in the same dish for 5 weeks and fed every 3 days. The dishes with the highest concentration of epithelial confluent cells were expanded and used for the uptake experiments.


Cerebral hemispheres were removed and the meninges were excised carefully and discarded. Cells were incubated in 0.1% trypsin for 15 min at 37°C and mechanically dissociated. The trypsinized tissue was transferred to a 15-mL culture tube containing 10 mL of MEM with 10% of FBS to stop trypsin action. Cells were centrifuged for 5 min at 200 g, the supernatant aspirated off and the cells resuspended in MEM and seeded in culture dishes at a concentration of 0.5–1.0 × 106 cells/dish. The culture medium was supplemented with 10% FBS, 2 mm l-glutamine, 100 U/mL penicillin and 100 μg/mL streptomycin.

For immunocytochemistry, cells were grown on 8-well Laboratory-Tek chamber microscopy slides (Nunc, Neperville, IL, USA), fixed with 4% p-formaldehyde in PBS for 30 min at 4°C, washed with PBS and incubated in PBS containing 1% bovine serum albumin (BSA) and 0.2% Triton X-100 for 5 min at room temperature. Cells were incubated with the different antibodies overnight at room temperature: Anti-GLUT1-5 (1 : 250, Alpha Diagnostic), anti-Kir6.1 (1 : 100, Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-GFAP (1 : 200, Dako), anti-vimentin monoclonal antibody (1 : 10, Boehringer Mannheim), anti-cytokeratin of high molecular weight (CK-HMW) monoclonal antibody (1 : 500, Dako), anti-brain-S100a polyclonal antibody (1 : 400, Dako), anti-myelin basic protein (MBP) monoclonal antibody (1 : 300, Boehringer Mannheim), anti-neurotrophic growth factor receptor (p75NGFr) polyclonal antibody (1 : 5000; Chemicon, Temecula, CA, USA), anti-transthyretin (TTR) polyclonal antibody (1 : 300, Dako), anti-Tau monoclonal antibody (5 mg/mL, Boehringer Mannheim), anti-MAP-2 monoclonal antibody (5 mg/mL, Boehringer Mannheim), anti-von Willebrand factor (HVWF) polyclonal antibody (1 : 300, Sigma, St Louis, MO, USA), anti-endothelial cell antigen (CD31) monoclonal antibody (1 : 50, Dako), anti–blood brain barrier monoclonal antibody (HT-7, 1 : 200, Sigma), and anti-nitric oxide synthase-brain (bNOS) monoclonal antibody (1 : 250, Sigma). Cells were then incubated with fluorescein isothiocyanate conjugated with goat anti-rabbit IgG or rabbit anti-mouse IgG (1 : 25, Boehringer Mannheim) for 2 h, mounted, and analyzed by fluorescence microscopy. Similar analyses were done with cultured astrocytes.


For immunoblot analysis, mouse tanycytes, liver and whole brain cell membrane proteins were obtained by homogenizing the cells in 0.3 mm sucrose, 3 mm DTT, 1 mm EDTA, 100 μg/mL PMSF, 1 μg/mL Peptatin A and 2 μg/mL Aprotinin. Total membranes were collected by high-speed centrifugation. Membrane protein (30 μg) was loaded in each lane, separated by polyacrylamide gel electrophoresis in the presence of sodium dodecylsulfate, transferred to nitrocellulose membranes, and probed with anti-GLUT, anti-Kir6.1 or pre-absorbed antibodies (1 : 500–1 : 1500) (Zamora-León et al. 1996). The secondary antibodies were goat anti-rabbit IgG coupled to peroxidase (1 : 5000) or rabbit anti-goat IgG coupled to peroxidase (1 : 5000). The reaction was developed with enhanced chemiluminiscence using the ECL western blotting analysis system (Amersham Corporation, Arlington Heights, IL, USA).

Uptake analysis

For uptake assays, cells were grown in 6-well plates to a density of 2 × 105 cells per well. Cultures were carefully selected under the microscope to ensure that only plates showing uniformly growing cells were used. In each experiment, cells from six wells incubated with buffer were removed and used to quantify the number of cells present in each well. We did not observe a significant variation in cell numbers between the wells after buffer incubation (Vera et al. 1993; Spielholz et al. 1997; Nualart et al. 2003). Cells were washed with incubation buffer (15 mm HEPES, 135 mm NaCl, 5 mm KCl, 1.8 mm CaCl2, 0.8 mm MgCl2) and incubated in the same medium for 30 min at room temperature. Uptake assays were performed in 1 mL of incubation buffer containing 0.2 mm deoxy-glucose and 3 μCi of 2-deoxy-d-[1,2-(N)3H]glucose (30.6 Ci/mmol; DuPont–NEN, Boston, MA, USA). Uptake was stopped by washing the cells with ice-cold PBS. Cells were lyzed in 0.5 mL of lysis buffer (10 mm Tris-HCl, pH 8.0, 0.2% SDS), and the incorporated radioactivity was assayed by liquid scintillation counting. Fructose uptake assays were performed in incubation buffer containing 1 mm fructose and 0.8 μCi of D-[U-14C]fructose/mL (285 mCi/mmol; Amersham) (Zamora-León et al. 1996). Samples were processed as indicated for deoxyglucose uptake. Where appropriate, competitors and inhibitors were added to the uptake assays or pre-incubated with the cells. Data represent means ± SD of three experiments done in duplicate.


Astrocytes and tanycytes localization in the periventricular zone of the mouse hypothalamus

Astrocytes are stellated cells with multiple fine processes, some of which are in close contact to capillary walls. Other astrocyte processes are wrapped around synaptic contacts which possess receptors for a variety of neurotransmitters. In the periventricular area of the brain, the astrocytes are located in the subventricular zone close to the ependymal cells. Consistent with this characteristic, we found a normal stellated astrocytes distribution in the dorsal ventricular wall of the third ventricle using the classical astrocyte and propidium iodine staining (Figs 2a and b, arrows). The ependymal cells and neurons were negative for anti-GFAP marker, indicating very-low-to-absent expression of GFAP (Figs 2a and b). However, in the lower lateral walls of the hypothalamus, stellated and GFAP positive cells were not detected in the subventricular zone (SVZ) and the reaction was observed in α tanycytes which presented long immunoreactive processes (Figs 2c and d, arrows). These processes may come into contact, en passant, with the ventromedial and arcuate nucleus neurons and also the blood vessels (Fig. 2d, insert and data not shown). In the infundibular walls of the hypothalamus, where β1 tanycytes contact the neurons of the arcuate nucleus, we did not find positive reaction with anti-GFAP (Figs 2e and f), suggesting that astrocytes were totally replaced by tanycytes in this hypothalamic region. Our results confirm that tanycytes are the main glial cells in contact with the endocrine neurons localized in the arcuate nucleus. This is one of the most important areas responsible for the generations of signals involved in glucose sensing.

Figure 2.

Double-labeling studies to observe the astrocytes and tanycytes distribution in the hypothalamic area. Frontal sections of mouse hypothalamus were stained with anti-GFAP (green) (a, c, e) and propidium iodine (red) (b, d, f). High expression of GFAP was detected in the dorsal subventricular zone of the third ventricle, where the astrocytes are localized under the classical ependymal cells (a, b, arrows). In the subventricular area of the hypothalamus, the processes of the α tanycytes showed positive reaction with anti-GFAP (c, d). The immunoreactive cell processes (c, d, arrows) contact ‘en passant’ the neurons of the arcuate and ventromedial nucleus (d, insert). No immunoreaction is detected in the lower lateral wall of the third ventricle where the arcuate nucleus neurons and β1 tanycytes are located (e, f). III-V, third ventricle; AN, arcuate nucleus; SVZ, subventricular zone. Scale bar, 100 μm.

GLUT1 and GLUT2 are expressed in the hypothalamic ependymal cells

Several studies suggest the expression of the facilitative glucose transporters, GLUT1 and GLUT2, in ependymal cells and neurons of the hypothalamic region (Leloup et al. 1994; Maekawa et al. 2000; Ngarmukos et al. 2001). To confirm the expression of GLUT1 and to establish the precise localization of GLUT2 in the hypothalamus of adult mice, we used immunocytochemical detection with anti-GLUTs polyclonal antibodies and confocal microscopy analysis. Intense anti-GLUT1 immunoreactivity was observed in the hypothalamic area, and the immunoreactive material was associated with endothelial cells, neurophils and tanycytes (Figs 3a–c). High-magnification analysis showed that GLUT1 was localized mainly in the processes of the tanycytes (Fig. 3d, arrows). The apical cytoplasm and the blebs of the cells were also positives (Fig. 3e, arrow and asterisks). The reaction was negative when the antibody was pre-absorbed with the blocking peptide (Fig. 3f). As positive controls, we observed the intense immunoreaction detected in the ependymal cells of the third and lateral ventricle (Figs 3g–i) and in the basolateral membrane of the choroids plexus cells (Fig. 3h, arrows).

Figure 3.

Double-labeling studies in the mouse hypothalamus to detect GLUT1 expression. Frontal sections of mouse hypothalamus were stained with propidium iodine (red) and anti-GLUT1 (green) (a–f). In the arcuate nucleus zone of the hypothalamus, the β tanycytes show a marked immunostaining with anti-GLUT1 (b, c). The reaction is positive in the proximal parts of β tanycytes which make contact with the cerebrospinal fluid of the third ventricle and in the processes of the cells (d, e, arrows). The cellular blebs of the tanycytes are also immunopositives (e, asterisks). The ciliated ependymal cells of the third and lateral ventricle, the choroids plexus cells and the endothelial cells of the blood–brain barrier present an intense immunoreaction (g–i). To control for the specificity of anti-GLUT1, the antibody was pre-absorbed with the peptides used to elicit them (f). III-V, third ventricle; AN, arcuate nucleus. Ep, ependymal cells; LV, lateral ventricle; N, neurons. Scale bars in a–f, 10 μm; in g–i, 30 μm.

Additionally, we used immunofluorescent and confocal microscopy to analyze, with higher sensitivity, the expression of GLUT2 in the hypothalamic area. Our results showed GLUT2 staining in α and β tanycytes (Figs 4a–c, arrows) mainly localized in the apical cellular membranes and the ventricular cytoplasmic regions of the cells (Figs 4d and e, large arrows). The cell processes of both α and β tanycytes presented a weak immunoreactivity (Figs 4d and e, short arrows). In the lower infundibular region of the third ventricle the β tanycytes presented a low positive reaction in the blebs of the cells, similar to the reaction observed with anti-GLUT1 (Fig. 4f, asterisks and short arrows). No immunostaining was observed in the ependymal cells (Fig. 4g). An intense immunoreaction was observed in the cellular membrane of the islet β cells (Fig. 4i), demonstrating the high specificity of the anti-GLUT2 antibody. The reaction was always negative when we used the blocking peptide to absorb the antibody (Fig. 4h).

Figure 4.

Double-labeling studies in the mouse hypothalamus to detect GLUT2 expression. Frontal sections of mouse hypothalamus were stained with propidium iodine (red) (a–h) and anti-GLUT2 (green) (b–h). In the hypothalamus, α and β tanycytes show a positive immunostaining with anti-GLUT2 (b, c). The reaction is positive in the proximal part of α and β tanycytes, where the cells contact the cerebrospinal fluid of the third ventricle (b–e, large arrows). The processes of the cells presented lower immunoreaction (d–e, short arrows). The cellular blebs of the tanycytes present low positive immunoreaction (f, asterisks and short arrows). The ciliated ependymal cells of the third ventricle were negative (g). To control for the specificity of anti-GLUT2, the antibody was pre-absorbed with the peptides used to elicit them (h). The specificity of the anti-GLUT2 antibody was observed in pancreatic islet β cells, an intense immunoreaction was detected in the cellular membranes of the cells. III-V, third ventricle; AN, arcuate nucleus. Scale bars in a–h, 10 μm; in i, 15 μm.

Isotopic in situ hybridization analyses have suggested the expression of GLUT2 mRNA in hypothalamic neurons and ependymal cells, but the low resolution of this technique has prevented a clear identification of the specific cell types expressing GLUT2. We analyzed the gene expression of GLUT2 at the mRNA level by in situ hybridization using digoxigenin-labeled cRNA probes specific for GLUT2. Both hypothalamic tanycytes, α and β, showed a positive hybridization signal (Figs 5a and b, arrows) concentrated in the region facing the third ventricle (Fig. 5b, insert). The neurons showed a very low reaction similar to the staining observed with the sense riboprobe (Figs 5b and c), however, the reaction detected in tanycytes was completely abolished when the sense probe was used (Fig. 5c, arrows). In conclusion, these experiments confirmed GLUT1 expression in hypothalamic tanycytes and demonstrated that these cells also express the low affinity transporter GLUT2.

Figure 5.

GLUT2 mRNA expression in hypothalamic cells. (a–c) Frontal sections of the hypothalamus hybridized using digoxigenin-labeled riboprobes specific for GLUT2. α and β tanycytes showed positive hybridization signal (a, arrows). An intense hybridization was observed at the proximal cytoplasm of β tanycytes (b, arrows and insert). The arcuate nucleus neurons presented low or absent reaction similar to the control (b). (c) Frontal sections of the hypothalamus hybridized using digoxigenin-labeled sense probes, as a control. The reaction observed in the tanycytes was completely abolished (arrows). III V, third ventricle; AN, arcuate nucleus. Scale bar in a, 10 μm; in b, c, 30 μm.

GLUT2 is highly expressed in primary tanycyte cultures

We seeded our cultures with cells obtained by thoroughly dissecting the pre-natal mouse hypothalamic area. After 5 weeks in culture without passage, we selected flasks with confluent cell growth having an elongated, epithelial aspect (Fig. 6a). Most cells showed a polarized morphology that consisted of a wide proximal cytoplasmic region containing the nucleus and a long basal process (Fig. 6b). Immunohistochemical analysis revealed an intense positive reaction with anti-vimentin and anti-p75 NGFr (Table 1, Figs 6c and f, and 7a), and a low anti-S100a immunoreaction (Table 1, Fig. 6e) in cultured tanycytes. Anti-GFAP, anti-CK-HMW and anti-TTR produced a negative immunoreaction (Table 1, Fig. 6d). The cells showed negative immunoreactivity for antibodies against neurons (anti-Tau and anti-MAP2), oligodendroglia (anti-MBP) and endothelial cell markers (anti-HVWF, anti-BBB, anti-bNOS and anti-endothelial cells) (Table 1). Astrocytes showed an intense immunostaining with anti-GFAP (Table 1, Fig. 6g), however, they were negative for anti-vimentin (Table 1 and Fig. 6h), anti-S100a, anti-MBP and antip75 NGFr (Table 1).

Figure 6.

Immunocytochemistry analyses of cultured tanycytes and astrocytes. The cells were obtained from mouse hypothalamus at 19 days of gestation. (a) Tanycytes after 5 weeks in culture. The cells are organized in monolayers and show an elongated form. (b) High magnification of a single cell using Nomarsky optics. The cell shows a polarized aspect with an apical expanded area and a long single process (labeled P). The position of the nucleus is also indicated (labeled N). (c–f) Tanycytes immunostained with anti-vimentin (c), anti-GFAP (d), anti-brain S100a (e) and anti-p75 NGFr (f). The tanycytes were anti-GFAP negative, however, the cells presented a strong immunoreaction with anti-vimentin and anti-p75 NGFr, both markers for tanycytes. (g, h) Astrocytes at 2 weeks in culture immunostained with anti-GFAP (g) and anti-vimentin (h). Insert in h, Höechst staining indicating the cells present in the observed field. Scale bars in a, 100 μm; in b, 5 μm. Scale bars in c–h, 30 μm.

Table 1.  Comparative immunohistochemical analysis of cultured tanycytes and astrocytes
  1. nd, not determined; –, negative; +, weak reaction; +++, strong reaction.

Glial and ependymal cell markers
 Anti-p75 NGFr+++
Neuronal markers
Endothelial cell markers
 Anti-endothelial cellsnd
Glucose transporters markers
K-ATP channels subunit
Figure 7.

Cultured tanycytes express GLUT1 and GLUT2. Immunocytochemical and immunoblot analysis. (a–f) Tanycyctes at 5 weeks in culture analyzed with different antibodies. (a) Anti-vimentin. (b) Anti-GLUT1. (c) and (e, f) anti-GLUT2. GLUT2 is detected in the proximal part of the tanycytes (c) and in the long processes of the cells (e). (d) Höechst staining indicating the total cells present in the observed field. The numbers 1–4 represent the same cells shown in c with anti-GLUT2. (f) Control of the immunostaining using anti-GLUT2 pre-absorbed with the peptide. (g, h) Cultured astrocytes analysed with anti-GLUT1 (g) and anti-GLUT2 (h). The astrocytes were positives to anti-GFAP (h, insert). (i) Immunoblotting of total cellular membranes of mouse brain (lane 1) and tanycytes (lane 2) incubated with anti-GLUT1. The tanycytes showed a 55-kDa band (lane 2) similar to the band located in cellular membranes isolated from whole mouse brains (lane 1). (j) Immunoblotting of total cellular membranes of mouse liver (1) and tanycytes (2) incubated with anti-GLUT2. A 60-kDa band that corresponds to GLUT2 was detected in both samples. A total of 30 μg of membrane protein was loaded in each lane. Controls were performed incubating the membranes with antibodies pre-absorbed with the peptides (i, j, lanes 3–4). Scale bars in a–h, 30 μm.

Antibodies specific for facilitative glucose transporters revealed expression of GLUT1 and GLUT2 in cultured tanycytes (Table 1, Figs 7b–f). The anti-GLUT1 immunoreactivity showed some heterogeneity (Fig. 7b); in comparison, anti-GLUT2 staining was consistently more intense and was evenly distributed throughout the cell population (Figs 7c and d). Moreover, the cell processes showed a homogeneous and particularly intense reaction with anti-GLUT2 (Fig. 7e, arrows). To control for the specificity of the anti-GLUT1 and anti-GLUT2 immunoreactivity, we used primary antibodies pre-absorbed with the blocking peptides (Fig. 7f). To control for the expression of GLUT1 and GLUT2 in other glial cells in culture, we analyzed the expression of these transporters in astrocytes. We detected a clear immunoreaction for GLUT1 (Fig. 7g), however, anti-GLUT2 was always negative (Fig. 7h). The astrocytes used for these analyses presented intense staining for GFAP (Fig. 7h, insert).

GLUT1 and GLUT2 expression was also evaluated by western blotting using protein extracts isolated from cultured tanycytes. Our analyses indicated that the tanycytes expressed a protein band of approximately 55 kDa reactive to anti-GLUT1 (Fig. 7i, lane 2) that co-migrated with a similar protein band present in membranes prepared from whole brain (Fig. 7i, lane 1). Parallel experiments in tanycytes demonstrated the presence of a 60-kDa protein band immunoreactive with anti-GLUT2 (Fig. 7j, lane 2) which co-migrated with a similar protein band present in membranes prepared from mouse liver cells (Fig. 7j, lane 1). No immunoreactive proteins were detected when anti-GLUT1 and anti-GLUT2 antibodies were pre-absorbed with the blocking peptides (Figs 7i and j, lanes 3 and 4).

Functional characterization of GLUT1 and GLUT2 in primary cultures of tanycytes

Hexose uptake studies revealed that the cultured tanycytes incorporate 2-deoxy-glucose (2-DOG) (Fig. 8a). DOG is transported by both GLUT1 and GLUT2. Short-term uptake assays revealed that uptake proceeded in a linear fashion at a rate of 3.3 nmol per millon cells per min for the first 90 s of incubation (Fig. 8a). To demonstrate the presence of a functional GLUT2, we determined the capacity of tanycytes to transport fructose, a substrate that is transported by GLUT2, but not GLUT1. As shown in Fig. 8(b), tanycytes transport fructose in a time-dependent manner and competition studies indicated that high concentrations of fructose decrease the cellular uptake of 2-DOG (Fig. 8c), which is consistent with the concept that tanycytes express a transporter with GLUT2-like properties. Competition analysis demonstrates that l-glucose is not transported by GLUT1 or GLUT2 (Fig. 8c).

Figure 8.

Kinetic analysis of 2-deoxy-d-glucose (2-DOG) and fructose uptake in cultured tanycytes. (a) Time-course of 0.2 mm 2-DOG uptake. Experiments were performed at room temperature. (b) Time-course of 1 mm fructose uptake. Hexose uptake analyses demonstrated that cultured tanycytes were able to transport 2-DOG and fructose. (c) Semi-log plot of the concentration dependence for inhibition of deoxyglucose transport by fructose and l-glucose. Measurements were performed at 0.2 mm deoxyglucose using 60-s uptake assays. (d) Substrate dose-dependence for the uptake of 2-DOG using 1 min assays. (e) Eadie-Hofstee analysis of the data presented in panel d. Two components with different affinities for the transport of 2-DOG are observed. (f) Substrate dose-dependency for each of the high- and low-affinity hexose transporters present in cultured tanycytes. The data indicate that mouse hypothalamic tanycytes express functional GLUT1 and GLUT2. Data represent the mean ± SD of three samples.

To determine the kinetic properties of transporters involved in 2-DOG uptake we used a short uptake period of 1 min. As shown in Fig. 8(d), dose-dependence experiments revealed that uptake of 2-DOG by tanycytes approached saturation at substrate concentrations near 100 mm. Eadie-Hofstee analysis of the uptake rate data revealed the existence of two functional components with different affinities for the transport of 2-DOG in tanycytes (Fig. 8e). The high-affinity component for glucose transporter reveals an apparent KM of 2.8 mm, with a Vmax of 19.8 nmol/min per million cells (Fig. 8e). A second component with a lower affinity reveals an apparent KM of 41.6 mm and a Vmax of 115 nmol/min per million cells for the transport of 2-DOG (Fig. 8e). Assuming that both transporters transport glucose with an affinity similar to that for 2-DOG, the apparent KM and Vmax values for the transport of this sugar were used to estimate the relative contribution of each transporter to the acquisition of glucose under normal conditions. As shown in Fig. 8(f), both transporters would make a similar contribution to the uptake of glucose at a sugar concentration of 5 mm.

The transport data, combined with the results of the immunohistochemistry, indicate that the two transporters expressed in tanycytes are functionally active and correspond to GLUT1 (the higher affinity component) and GLUT2 (the lower affinity component). To confirm our finding, we examined the effect of cytochalasin B, a classical glucose transporter inhibitor, on the transport of 2-DOG. At 0.2 mm 2-DOG, more than 70% of transport is carried out by the higher affinity transporter, GLUT1 (Fig. 8f). In contrast, at 10 mm 2-DOG, more than 60% of transport is carried out by the lower affinity component, GLUT2 (Fig. 8f). Cytochalasin B, a potent non-competitive inhibitor of glucose transporters (Deves and Krupka 1978), shows a greater affinity for GLUT1 than GLUT2, this characteristic can be used to distinguish between these two isoforms. The transport of 0.2 mm 2-DOG in the presence of cytochalasin B was inhibited in a dose-dependent manner, with an IC50 of 0.1 μm. This is close to the inhibition constant of GLUT1 by cytochalasin B and more than 80% inhibition was observed at 1 μm cytochalasin B (Fig. 9a). Interestingly, inhibition of the uptake of 10 mm 2-DOG by increasing concentrations of cytochalasin B showed a biphasic behavior suggesting the presence of two independent transporter activities (Fig. 9b). The first component, with an IC50 for cytochalasin B of 0.1 μm accounted for 40% of the total transport activity, and is consistent with the expected relative contribution and the IC50 of GLUT1. The second component, with an IC50 of 2 μm accounted for the remaining 60% of transport, which is consistent with the expected properties of GLUT2. Consistently, cytochalasin E failed to affect the uptake of 0.2 and 10 mm 2-DOG by hypothalamic tanycytes, confirming the specificity of the inhibition effect of cytochalasin B.

Figure 9.

Effect of cytochalasin B on 2-deoxy-d-glucose uptake by cultured tanycytes. (a) Dose-dependent effect of cytochalasin B and cytochalasin E using a concentration of 0.2 mm 2-DOG to test the higher affinity transporter. (b) Dose-dependent effect of cytochalasin B and cytochalasin E using a concentration of 10 mm 2-DOG to test the lower affinity transporter. Data represent the mean ± SD of three samples. DOG, deoxyglucose; Cyt, cytochalasin.

Tanycytes in culture express the K-ATP channels pore-forming subunit Kir6.1

Another component of the glucose-sensing mechanism is the K-ATP channel. There is immunohistochemical data showing the expression of the K-ATP channel pore-forming subunit Kir6.1 in hypothalamic tanycytes in situ (Thomzig et al. 2001). As the expression of this channel is crucial to argument that tanycytes are potentially involved in glucose sensing, we studied the expression of Kir6.1 subunit in cultured tanycytes using immunofluoresce and immunoblot analysis.

Immunocytochemical analysis using specific antibodies for Kir6.1 subunit indicates that Kir6.1 subunit is expressed in tanycyte primary cultures (Table 1, Fig. 10a). Positive immunoreactivity is detected in the cytoplasm and processes of the tanycytes. Propidium iodine staining after RNAase treatment confirms that most of the cells in culture expressed Kir6.1 immunoreactivity (Fig. 10b). Anti-Kir6.1 primary antibodies pre-absorbed with the blocking peptide indicate that the immunoreaction was specific (data not shown). Western immunoblotting analysis using anti-Kir6.1 antibodies revealed a 47-kDa protein band (Fig. 10d, lane 3). A similar protein band was found in extract prepared from hypothalamus and liver (Fig. 10, lanes 1 and 2).

Figure 10.

Tanycytes in vitro express the K-ATP channel subunit Kir6.1. The cells were obtained from mouse hypothalamus at 19 days of gestation and used to perform immunocytochemical (a, b) or immunoblot (c) analysis. (a, b) Tanycytes after 5 weeks in culture stained with anti-Kir6.1 subunit antibody (a) and propidium iodine (b). The merge image shows that most of the cells in culture present Kir6.1 immunoreaction (b). (c) Immunoblotting of total cellular membranes of liver (lane 1), hypothalamus (lane 2) and cultured tanycytes (lane 3) incubated with anti-Kir6.1. The tanycytes show a 47-kDa band similar to the band located in cellular membranes isolated from liver and hypothalamus. Scale bar in c, 15 μm.


Initial biochemical studies indicate that the relative expression of GLUT2 in the brain is low (Brant et al. 1993; Leloup et al. 1994). However, low relative levels of GLUT2 may be due to restricted expression in a few cell types. Our detailed immunohistochemical and functional analysis of the low affinity glucose transporter, both in situ and in vitro, consistently indicates that GLUT2 is primarily expressed in α and β tanycytes of the mouse hypothalamus. These experiments were performed using an anti-GLUT2 antibody that gave a positive immunoblot reaction in membranes isolated from mouse hepatocytes expressing high levels of GLUT2 and in pancreatic β cells in situ. Interestingly, the preferential GLUT2 expression observed in hypothalamic α and β tanycytes was localized in the proximal part of the cell body, which corresponds to the region of the cell that is in contact with the cerebrospinal fluid. Other cells of the hypothalamus, such as neurons and endothelial cells, were consistently negative for GLUT2 in our immunohistochemical analysis.

Immunohistochemical data have indicated GLUT2 localization in numerous punctate structures localized in the hypothalamic area between the arcuate nucleus and ventromedial hypothalamus (Leloup et al. 1994). Further, immunodetection using electron-microscopy suggests GLUT2 localization in astrocytes processes in close relationship with nerve terminals or neuronal cell bodies (Leloup et al. 1994). However, the antibodies used in this study were unable to detect GLUT2 proteins by western blot analysis of samples isolated from the hypothalamic arcuate nucleus. Ngarmukos et al. (2001) were unable to detect GLUT2 expression in ventromedial hypothalamic cells; however, GLUT2 immunoreactivity was detected in the ependymal cells of the dorsal third ventricle and in scattered cells in the arcuate and periventricular nuclei. Further, Maekawa et al. (2000) detected a clear expression of GLUT2 in ependymal cells of the dorsal third ventricle and cerebral aqueduct, but the hypothalamic glial cells were negative. In detail, GLUT2 was localized in ciliated ependymal cells of the cerebral aqueduct, specifically in the cell membrane of the cilia (Maekawa et al. 2000). Thus, the localized expression of GLUT2 in a few ventricular ependymal cells may explain the low relative expression of this transporter when the analysis includes the whole brain. In situ hybridization analyses have revealed GLUT2 expression in a region of the hypothalamus containing neurons and ependymal cells (Navarro et al. 1996). However, the lower resolution capacity of autoradiographic detection of the GLUT2 probe prevented the identification of the specific cell types expressing GLUT2. Our in situ non-isotopic hybridization clearly identifies the α- and β-tanycytes as the hypothalamic cells expressing GLUT2. We observed a reduced hybridization in neuronal soma, however, the negative controls using the sense probe showed a similar reaction, indicating low-to-absent expression of GLUT2 mRNA in arcuate nucleus neurons and astrocytes-like cells.

The expression of GLUT2 in α- and β-tanycytes suggest that tanycytes contacting ventromedial hypothalamic neurons (α) and arcuate nucleus neurons (β) are involved in glucose uptake using the same low-affinity transporter. In both types of cells, the localization of GLUT2 is observed in the proximal pole of the tanycyte which contains the cerebrospinal fluid, and thus tanycytes may be primarily involved in detecting glucose concentration in the cerebrospinal fluid of the ventricular system. Similar expression of GLUT2 transporter has been detected in ependymal cells of the cerebral aqueduct and dorsal third ventricle (Maekawa et al. 2000), indicating that different cells facing the ventricular system may be involved in glucose-sensing mechanisms.

GLUT2 expression was also observed in primary cultures of tanycytes dissected from the mouse hypothalamic area. In vitro, cells presented a highly elongated and polarized form reminiscent of tanycytes observed in vivo. Immunostaining with antibodies specific for markers found in different brain cells strongly supports the concept that the cells in culture corresponded to differentiated tanycytes. Thus, the cells were immuno-positive for two tanycytes markers, vimentin and p75 NGFr (Gudiño-Cabrera and Nieto-Sampedro 2000), and were immunonegative for antibodies against markers specific for astrocytes (anti-GFAP), neurons (anti-Tau), endothelial cells (anti-HVWF, anti-BBB), and oligodendroglia (anti-MBP). Immunohistochemical analysis of glucose transporter expression showed the presence of GLUT1 and GLUT2 in the cultured cells, this is analogous to the in vivo situation. Immunoblot analyses confirmed the expression of GLUT1 and GLUT2 transporters in cultured tanycytes, and molecular weight analyses revealed that the tanycytes expressed a 55-kDa GLUT1 isoform. A similar analysis combining molecular weight determination with anti-GLUT2 immunoblotting revealed that the size of GLUT2 (60 kDa) in tanycytes was identical to that in hepatocyte GLUT2. Overall, the immunolocalization and immunoblotting data strongly support the concept that the cells grown in vitro, after dissection from the hypothalamic area, corresponded to hypothalamic tanycytes.

A combination of transport and competition assays revealed that the tanycytes expressed two functionally active transporters directly involved in the uptake of hexose by these cells. The kinetic data confirmed that these transporters correspond to the isoforms GLUT1 and GLUT2 as detected by immunocytochemical techniques. The higher affinity transport component had the expected properties (kinetic and sensitivity to inhibitors) of GLUT1, with a transport Km of 2.8 mm. In contrast, kinetic analysis of the lower affinity transporter revealed an apparent Km of 41 mm, which fits the description of GLUT2. The capacity of the cultured tanycytes to take up fructose, a substrate specific for GLUT2 (Gould and Holman 1993), confirmed that GLUT2 is fully functional.

The high-affinity glucose transporter GLUT3 plays a central role in the metabolism of glucose in the brain because it is the main glucose transporter expressed by neurons (Vannucci 1994). The results of our functional studies indicated absence of expression of a GLUT3-like high affinity glucose transporter (Km < 1 mm) in the cultured tanycytes. We confirmed the absence of expression of GLUT3 in the tanycytes in vitro and in situ by immunocytochemistry and immunobloting with anti-GLUT3 antibodies (Table 1 and data not shown). Thus, the data indicate that GLUT3 plays no role in glucose uptake by the tanycytes.

The expression of GLUT2 in hypothalamic tanycytes may have important physiological consequences. An elevated Km for glucose transport implies that tanycytes expressing GLUT2 will increase their glucose uptake rate in direct proportion to extracellular changes in glucose concentration. This property of GLUT2 determines its participation in the glucose-sensing mechanism of the pancreatic β cell (Guillam et al. 1997; Yang et al. 1999; Guillam et al. 2000; Schuit et al. 2001). Therefore, it is possible that GLUT2 might play a similar role in the brain (Wan et al. 1998), which opens up the possibility for the involvement of tanycytes in a glucose-sensing mechanism in the hypothalamus. In this context, there is data indicating that proteins of the type involved in the β cell glucose-sensing mechanism are expressed in the hypothalamus (Alvarez et al. 1996; Navarro et al. 1996; Miki et al. 2001) (Fig. 1).

In order to obtain more evidence to reinforce our theory that tanycytes are involved in glucose sensing, we analyzed the expression of ATP-sensitive potassium channels in cultured tanycytes. These channels are essential to couple the energy state of a cell to its excitability, participating in glucose-sensing mechanism in pancreatic islet β cells (Schuit et al. 2001). Recently, the presence of functional K-ATP channels in glial cells has been suggested. This study detected the presence of the Kir6.1 subunit, the main pore-forming protein detected in hippocampal, cortical and cerebellar astrocytes, tanycytes and Bergmann glial cells (Thomzig et al. 2001). We confirmed the expression of the K-ATP channel subunit Kir6.1 in cultured tanycytes, indicating that these cells maintain in vitro some of the molecular and functional properties observed in situ.

In addition to GLUT2 and K-ATP channels, cells involved in glucose sensing express a high Km glucokinase, an enzyme that shows low affinity for glucose and is not inhibited by glucose-6-phosphate. Glucokinase is expressed in the hypothalamus, specifically in ependymal cells (tanycytes) and some neurons (Roncero et al. 2000). The high Km glucokinase has also been detected in the lower brain stem (Lynch et al. 2000; Maekawa et al. 2000). Other areas within the brain, specifically the area post rema, the medulla oblongata and the tractus solitarius nucleus have also been postulated to have glucose sensor mechanisms that modulate feeding and reproduction (Leloup et al. 1994; Schwartz et al. 2000). Additionally, a sensor mechanism triggered by low glucose concentrations in the hind brain controls the secretion of GnRH that inhibits the pulsatile LH secretion in rats (Murahashi et al. 1996). It might be possible that the effects associated with a glucose-sensing mechanism in the lower brain stem may be functionally co-ordinated with a hypothalamic glucose-sensing mechanism (Grill and Kaplan 2002). GnRH is secreted in the basal and lateral part of the median eminence, an area in which the axonal terminals of the neurons that release GnRH are in contact with the end feet of the tanycyte processes (Meister et al. 1988). The end feet of the tanycyte processes wrap around and support the portal blood vessels of the median eminence, thereby blocking the nerve endings, containing hypothalamic factors, from reaching the peri-capillary spaces in the median eminence (Kozlowski and Coates 1985; Hökfelt et al. 1988). It has been proposed that dopamine regulates the contact between the tanycyte ending and the blood vessels (Bergland and Page 1979; Hökfelt et al. 1988). When dopamine is released, it induces the retraction of the tanycyte endings, allowing the access of the neuropeptide-containing nerve terminal to the portal blood vessels. This property of tanycytes is regulated by dopamine, but its function may also be induced by intracellular signals generated directly in the tanycytes. Tanycytes are in simultaneous contact with the cerebrospinal fluid and the blood vessels, and it is reasonable to think that they may be able to detect changes in glucose concentration occurring in both fluids. Thus, variations in glucose concentration may induce the first changes in tanycytes, and dopamine (whose release may be regulated by axons afferent from the lower brain stem) could modulate the structural change. These various systems provide a selection of routes by which glucose may reach the brain and affect phenomena as diverse as sleep, orgasm, reproduction cycle and feeding behavior (Bergland and Page 1979).

The level of GLUT2 expression in tanycytes may be similar to the expression of GLUT3 and GLUT1 in neurons and astrocytes. Thus, these transporters need high-sensitivity immunohistochemical methods (Leloup et al. 1994). Additionally, an up-regulation in the expression of GLUT1 and GLUT3 has been observed in cultured cells (Gould and Holdman 1993). A similar up-regulation of GLUT2 occurred in cultured tanycytes, allowing us to detect the transporter with conventional immunofluorescence. We did not detect changes in GLUT2 expression in cells after prolonged culture, suggesting that the regulation of GLUT2 expression in tanycytes is different to that in hepatocytes, where cells lost the expression of GLUT2 in short time cultures with a simultaneous induction of GLUT1 expression (Zheng et al. 1995).


The authors thank Ximena Campos, Universidad de Concepción for her technical support and Dr Simon Watkins, Department of Cell Biology and Physiology, University of Pittsburgh, for his support in confocal microscopy. We also thank Dr Rodolfo Medina for the critical reading of the manuscript. This work was in part supported by grants 1010843 from FONDECYT, Chile, DIUC-GIA 201.034.006–1.4, DIUC 202.031.089–1 and DIUC 201.035.002–1.0, Universidad de Concepción, Chile.