Both α and β tanycytes express GLUT2 in adult mice
Before inactivating Glut2, its protein localization was evaluated in hypothalamic tissue from bregma AP -1.82 mm of adult mice. GLUT2 was detected in the cell bodies that form the 3V at the height of VMN (Fig. 1A, asterisk) and the ARC (Fig. 1B). Colocalization with vimentin, a tanycyte marker, was detected in proximal a and b-tanycytes, and a strong reaction was detected in the processes of b-tanycytes that project to the ARC and ME (Fig. 1B, asterisk). Several studies have shown that GFAP-positive tanycytes are mainly located in the VMN and DMN [17, 23, 24]. At the bregma − 1.82, we observed GFAP-positive tanycytes facing the DMN and VMN that are also vimentin-positive (Fig. 2A). High magnification of the dorsal periventricular regions showed intense colocalization between both intermedia filaments (Fig. 2A,1–3). In contrast, tanycytes facing the ARC and ME are vimentin-positive and GFAP-negative (Fig. 2A). A previous study showed that these tanycytes are GLUT2-positive , and 3D reconstruction confirms GLUT2 immunoreaction and localization in dorsal tanycytes (Supplementary Fig. 1A, arrowheads), which is more evident at high magnification (Supplementary Fig. 1B, arrowheads). In addition, GLUT2 was not detected in parenchymal astrocytes (Supplementary Fig. 1C, arrowhead) as reported previously [10, 15].
Because GFAP-positive tanycytes have been poorly described , we analyzed their anteroposterior location based on 3D GFAP and vimentin immunoreactivity. We evaluated the proportion of GFAP-positive versus GFAP-negative tanycytes on the anteroposterior axis of the hypothalamic areas located between the bregma − 1.34 mm to the − 2.54 mm. The different bregma locations were identified as zones, as shown in Fig. 2B. Our results showed that GFAP-positive tanycytes are mostly located from zone 1 to 3 (Fig. 2C). In zone 4, we found a transition zone where the percentage of GFAP-positive and GFAP-negative tanycytes is similar. Interestingly, we detected a higher percentage of GFAP-negative tanycytes in zones 5 and 6 of the hypothalamus (Fig. 2C). These observations indicate that GFAP-expressing tanycytes have an asymmetric location along the anteroposterior axis with a marked localization in the more anterior zones. Supplementary Fig. 2 shows the projections of GFAP-positive tanycytes to the different hypothalamic nuclei in detail. Next, we analyzed the hypothalamic dorsoventral location of the GFAP-positive tanycytes from zone 1 to 6. In the anterior region (zones 1 and 2), GFAP-positive tanycytes are mainly located in the VMN and ARC (Fig. 2D-E and Supplementary Fig. 2). Interestingly, in the medial region (zone 3), GFAP-positive tanycytes project into the DMN, VMN, and ARC in similar proportion (Fig. 2F). In the posterior zones of the hypothalamus, GFAP-positive tanycytes contact the DMN, ARC, and the dorsal tuberomammillary nucleus (DTM) (Fig. 2G-I and Supplementary Fig. 2).
Glut2 genetic inactivation in GFAP-positive tanycytes
To delete the Glut2 gene in GFAP-positive tanycytes from zones 1 to 6, Slc2a2loxP/loxp, mice were stereotactically injected in the 3V with adeno-associated virus (AAV)-Gfap-Cre-GFP at bregma AP -1.82 mm, where the GFAP-expressing tanycytes are projecting to DMN, VMN, and ARC in similar proportions (Fig. 3A). Injection of the AAV-Gfap-CRE-GFP induced the expected recombination of the Glut2 allele, as shown by a 326 bp PCR product (Fig. 3B). The presence of the non-recombined floxed allele corresponding to the 281 bp product was due to the existence of non-transduced cells in this region. To confirm that recombination occurred exclusively in tanycytes, the GFP signal and the percentage of GFAP/GFP-positive tanycytes in the hypothalamus of mice transduced for 2-weeks were evaluated through confocal microscopy. As expected, GFP was detected in the apical region and processes of tanycytes (Fig. 3C, arrowheads). Moreover, GFP fluorescence (green) was observed in GFAP-positive tanycytes (yellow) that contain endfeet with button morphology, which closely contact other cells present in the hypothalamic parenchyma (Fig. 3D, arrowhead). To determine the percentage of transduction, the number of GFAP-expressing tanycytes positive for GFP fluorescence was quantified. As observed in Fig. 3E, the percentage of GFAP-expressing tanycytes positive for GFP fluorescence was close to 28.5%.
Since the processes of GFAP-expressing tanycytes contact different nuclei in the anteroposterior and dorsoventral axes, we next evaluated the proportion of transduced cells on both axes using GFP fluorescence. The transduced cells correspond to GFAP-positive tanycytes and their processes extending toward the ARC, ARC/VMN, and DMN, respectively (Fig. 3F, arrowheads). While GFP fluorescence was poorly present in the GFAP-expressing tanycytes in zone 1 (Fig. 3G), it was detected from zones 2 to 6 (Fig. 3H-L). Of the transduced GFAP-positive tanycytes located in zone 2, 25% and 75% contact the ARC and VMN, respectively (Fig. 3H). At the middle region of the hypothalamus, transduced GFAP-positive tanycytes located in zone 3 trace their processes to the ARC, VMN, and DMN in similar proportions (Fig. 3I), while 27.9% and 71.9% GFAP-positive tanycytes located in zone 4 contact the DMN and ARC, respectively (Fig. 3J). At the more posterior region of the hypothalamus, all GFAP-positive tanycytes contact the ARC (Fig. 3K-L). GFP-fluorescence was not detected in GFAP-positive cells located in other circumventricular organs, such as the area postrema (Supplementary Fig. 3A-E).
GFAP-expressing tanycytes regulate feeding behavior through Glut2
We next evaluated if GLUT2 expressed in GFAP-positive tanycytes is necessary to regulate feeding behavior and energy balance. Feeding behavior in a 24 h feeding cycle (05:00 p.m-05:00 p.m) was first evaluated. Body weight (control: 25.89 ± 0.75 g vs treated: 26.40 ± 0.78 g, p = 0.646, df = 16), cumulative meal events (control: 89.15 ± 4.98 events vs treated: 83.18 ± 6.86 events, p = 0.479, df = 22), and cumulative food intake (control: 5.01 ± 0.36 g vs treated: 4.19 ± 0.17 g, p = 0.067, df = 22) in 24 h were similar between the control and treatment groups (Fig. 4A-C). Moreover, no significant differences were observed during the 12 h of dark and 12 h of light phase of feeding (Fig. 4D). To more exhaustively analyze the feeding pattern in the 12 h dark cycle, the amount of food consumed every 1 h was evaluated. No significant differences were observed in the meal pattern between experimental groups (Fig. 4E). To further explore whether Glut2 inactivation in GFAP-positive tanycytes can impact satiety and satiation in basal conditions, we evaluated the mean meal events duration (min), mean events interval duration (min), meal size (g/event), latency of the first meal (min), first meal duration (min), and eating rate (mg/min) parameters. No statistically significant differences were observed in any of the parameters mentioned (Supplementary Fig. 4A-F). Taken together, the results suggest that GLUT2, expressed in GFAP-positive tanycytes, is not necessary to regulate the feeding behavior under basal conditions.
Subsequently, we evaluated the feeding behavior of Glut2-inactivated mice in response to fasting. The feeding behavior was evaluated through a fasting-refeeding protocol (24 h/24 h) (Fig. 4F). Interestingly, Glut2 inactivation in GFAP-positive tanycytes generated a significant decrease in cumulative meal events (events/24 h) (control: 76.27 ± 5.83 events vs treated; 59 ± 4.70 events, p = 0.027 events, df = 34) and cumulative food intake (g/24 h) (control: 5.65 ± 0.20 g vs treated: 5.00 ± 0.17 g, p = 0.019, df = 34) (Fig. 4G-H). To determine if Glut2 inactivation affects food consumption during the dark and light phases, we evaluated the food consumption during the 12 h of the night cycle and the 12 h of the day cycle. As expected, the amount of food consumed is significantly higher in the dark phase than the light phase in both groups (Fig. 4I), which suggests that while the Glut2 inactivation affects the total food consumption, it does not alter the circadian rhythm of feeding.
Our previous results show that GLUT2 inhibition, in all type of tanycytes, increases total food intake . We next performed a detailed evaluation in the dark phase. Data showed that GLUT2 inactivation led to minor food consumption (control: 0.695 g vs. treated: 0.45 g, p = 0.047, df = 40) during the first hour of feeding (Fig. 4J), suggesting that GLUT2 expressed in GFAP-positive tanycytes could regulate the beginning of feeding. To test this hypothesis, we compared the total food consumption during hours 1 and 6 in the dark phase. As observed in Fig. 4K, the control group had a normal response to fasting, presenting a peak in food intake at hour 1 of feeding and significantly decreasing the food consumption at hour 6 of the cycle (1h: 0.695 ± 0.06 g vs. 6h: 0.340 ± 0.05 g, p = 0.0001, df = 68), indicating that after 6 h feeding the control group is satiated. However, the treated group consumed an amount of food similar during the hour s1 and 6 of the dark phase of feeding (1h: 0.457 ± 0.06 g vs. 6h: 0.512 ± 0.04 g, p = 0.0001, df = 68), indicating that Glut2 gene inactivation affects the feeding initiation following a fasting period (Fig. 4K). To confirm that Glut2-inhibited mice did not show a response to negative energy balance, we compared the cumulative food intake, the eating rate (mg/min), and the duration of the first feeding (min) during the first hour of the dark phase (g/1 h) after 24 h fasting or in mice with ad libitum access to food. As seen in Fig. 4L, the control group with ad libitum feeding had a significant increase in the cumulative food intake in response to 24 h of fasting (ad-libitum: 0.244 ± 0.05 g vs. fasting: 0.695 ± 0.06 g, p = 0.0001, df = 56). Nevertheless, this response to fasting is lost in Glut2-inactivated mice (ad libitum: 0.259 ± 0.05 g vs fasting: 0.457 ± 0.06 g, p = 0.285, df = 56) (Fig. 4L). Moreover, Glut2 gene inactivation decreased the eating rate in response to fasting (ad libitum: 9.14 ± 1.23 mg/min vs fasting: 11.63 ± 1.00 mg/min, p = 0.999, df = 56) (Fig. 4M). However, we did not observe a significant difference in the first meal duration of the treated group between both conditions (Fig. 4N).
To further explore if these results were related to altered satiety, we analyzed the following parameters: mean meal events duration (min), mean events interval duration (min), meal size (g/events), the latency of the first meal (min), first meal duration (min), and eating rate (mg/min). No significant difference was observed between the control and treated groups in any of the parameters analyzed (Supplementary Fig. 4G-L), suggesting that the inactivation of Glut2 does not affect satiety. Altogether, the data strongly suggest that a GLUT2-dependent mechanism in GFAP-positive tanycytes is necessary to stimulate feeding and regulate the feeding initiation following a fasting period.
GLUT2 expression in GFAP-tanycytes is required to control the ghrelin secretion
In vivo studies suggest that GLUT2 expression in the CNS is necessary to control the secretion of hormones controlling glucose homeostasis, such as insulin and glucagon [5, 9]. With this in mind, we evaluated whether GLUT2 must be expressed in GFAP-positive tanycytes to maintain normal blood values of glucose and peripheral hormones in fasted and refed mice. Glycemia, insulin, and glucagon at 24 h fasting and 6 h refeeding were similar in control and Glut2-inactivated mice (Fig. 5A-C). Concurrently, glucose tolerance tests (GTT) did not show any significant differences between the control and treated groups, indicating that the Glut2 deletion in GFAP-positive tanycytes did not alter glucose homeostasis (Fig. 5D).
Studies conducted in ripglut1;glut2−/− mice showed that GLUT2 expression is required to control the peripheral ghrelin secretion during the fasting-to-refeeding transition . Therefore, the effect of Glut2-gene inactivation on ghrelin, leptin and GLP-1 secretion was evaluated. No significant difference was observed in the plasma levels of leptin and GLP-1 between the control and treated groups (Fig. 5E-F). Nevertheless, we detected abnormal plasma levels of the orexigenic hormone ghrelin in the treated group (Fig. 5G). As expected, after 24 h fasting, total ghrelin was high and decreased significantly after 6 h refeeding in the control (fasting: 2.44 ± 0.20 ng/ml vs refeeding: 0.94 ± 0.05 ng/ml, p = 0.0001, df = 66) and treated (fasting: 3.14 ± 0.24 ng/ml vs refeeding: 0.87 ± 0.03 ng/ml, p = 0.0001, df = 66) groups. However, Glut2-inactivated mice showed significantly higher total ghrelin values than control at 24 h of fasting (control: 2.44 ± 0.20 ng/ml vs treated: 3.014 ± 0.24 ng/ml, p = 0.04, df = 66) (Fig. 5G). When acylated ghrelin was i.c.v. injected, a similar increase in food intake was detected (control: 0.17 ± 0.03 g vs treated: 0.21 ± 0.02 g, p = 0.999, df = 93) (Fig. 5H). These results indicate that the secretion of total-ghrelin in response to fasting is partly regulated in the hypothalamus by a GLUT2-dependent mechanism involving GFAP-positive tanycytes.
Genetic inactivation of Glut2 increases the c-Fos expression in the VMN
We previously showed that Glut2 inactivation affects feeding exclusively in response to fasting. Therefore, to determine whether GFAP-expressing tanycytes regulate the activity of neighboring neurons through GLUT2, we evaluated c-Fos activation in response to a 24 h fasting. As seen in Fig. 6A-B, Glut2 inactivation generated a significant increase in c-Fos expression in the VMN in zone 1 (control: 4 ± 0.5 positive cells vs treated: 12 ± 1.5 positive cells, p = 0.02, df = 132) and zone 2 (control: 5 ± 0.6 positive cells vs treated: 14 ± 1.9 positive cells, p = 0.009, df = 156) of the hypothalamus, without impacting DMN and ARC activation. Moreover, in zone 3, Glut2 inactivation generated a large decrease in c-Fos-positive cells number in the DMN (control: 59 ± 2.7 positive cells vs treated: 37 ± 5.2 positive cells, p = 0.0001, df = 150) and a significant increase in the VMN (control: 5 ± 0.6 positive cells vs treated: 32 ± 6.1 positive cells, p = 0.0001, df = 150) without impacting c-Fos expression in the ARC (control: 29 ± 1.3 positive cells vs treated: 20 ± 1.7 positive cells, p = 0.418, df = 150) (Fig. 6C). Conversely, no significant changes in the expression of c-Fos were detected in zones 4–6 compared to the control group (Fig. 6D-E). Together, these findings suggest that GFAP-positive tanycytes regulate neuronal activation in response to fasting through a GLUT2-dependent mechanism mainly in the VMN.
Glut2 inactivation in GFAP-expressing tanycytes disturbs Pomc gene expression
To determine if c-Fos activation affects orexigenic or anorexigenic neurons, the genetic expression of neuropeptides, Npy, Cart, and Pomc, was evaluated in response to a 24-h fasting and 6-h refeeding period (Fig. 7A). In control mice, the orexigenic neuropeptide, Npy, was high after a 24 h fasting, decreasing significantly after the 6 h of refeeding (Fasting: 1.13 ± 0.35 vs. refeeding: 0.42 ± 0.08, p = 0.02, df = 11) (Fig. 6B). Interestingly, in Glut2-inactivated mice, Npy mRNA (Fasting: 1.00 ± 0.11 vs refeeding: 0.60 ± 0.12, p = 0.01, df = 14) (Fig. 7B) was not altered. Regarding anorexigenic neuropeptides, the levels of Cart mRNA in response to 6 h of feeding were increased as expected in both groups (fasting: 1.17 ± 0.21 vs. refeeding: 2.28 ± 0.29 in the control group, p = 0.005, df = 12 and fasting: 1.01 ± 0.12 vs. refeeding: 2.37 ± 0.16 in the treated group, p = 0.0001, df = 14) (Fig. 7C), whereas the loss in the POMC normal response to refeeding was observed in Glut2-inactivated mice (Fasting: 1.68 ± 0.31 vs. refeeding: 1.92 ± 0.19, p = 0.157, df = 13) (Fig. 7D). Altogether, the results indicate that GFAP-expressing tanycytes modulate the gene expression of POMC neurons during the fasted-to-refeeding transition.