Tanycytic VEGF-A Boosts Blood-Hypothalamus Barrier Plasticity and Access of Metabolic Signals to the Arcuate Nucleus in Response to Fasting
Introduction
Appetite, energy balance and metabolism are all controlled by select neurons of the hypothalamic arcuate nucleus (ARH) (see for reviews (Elmquist et al., 2005; Gao and Horvath, 2007; Levin et al., 2011; Sawchenko, 1998)). These interoceptive sensory neurons participate in neural networks that sense circulating factors such as glucose and adiposity hormones that signal changes in metabolic state (Cowley et al., 2001; Cowley et al., 2003; Dunn-Meynell et al., 2002; Elias et al., 1999; Hill et al., 2010; Liu et al., 2012). However, the physiological mechanisms that control the access of these factors to ARH circuits and their regulation in response to changes in feeding status remain largely unexplored.
Molecular traffic between the periphery and the central nervous system (CNS), including the hypothalamus, is restricted by regulated interfaces, such as the blood-brain barrier (BBB), composed of tight junctions between endothelial cells lining brain microvessels (Neuwelt et al., 2011). The blood-cerebrospinal-fluid (CSF) barrier, a lesser known interface, is composed of tanycytes, specialized hypothalamic glia that line the floor of the 3rd ventricle, and microvessels of the median eminence (ME), a circumventricular organ adjacent to the ARH (Mullier et al., 2010). While the endothelial cells of the ME are unique in being fenestrated, and thus highly permeable to blood-borne molecules, tight junction complexes between adjacent tanycytes act as a physical barrier preventing their diffusion to the rest of the brain via the CSF. Transcellular transport across the blood-brain and blood-CSF barriers is usually mediated by saturable carriers (Banks, 2006; Hawkins and Davis, 2005). However, whether specific hypothalamic areas that regulate energy balance, such as the ARH, directly access peripheral homeostatic signals through the fenestrated microvessel plexus in the adjacent ME is a matter of debate (Ciofi et al., 2009; Flier, 2004; Mullier et al., 2010).
Here we show that food deprivation, by inducing both tight junction complex reorganization in tanycytes and the increased fenestration and permeability of ME microvessel loops that reach the ventromedial ARH (vmARH) (Ambach and Palkovits, 1979), leads to such direct access. Refeeding, glucose infusion, the selective inhibition of vascular endothelial growth factor (VEGF)-A expression in tanycytes and the manipulation of VEGF signaling all reverse these structural changes and the resulting ability of molecules to enter the ARH.
Results
Fasting-induced plasticity of the blood-hypothalamus barrier (BHB)
To investigate the role of BHB plasticity in the adaptive response to fasting, we compared the hypothalami of mice deprived of food for 24h and those fed ad libitum using immunofluorescence for two constitutive tight junction proteins, zonula occludens-1 (ZO-1) and occludin, expressed in BBB endothelial cells and tanycytes, and claudin-1, expressed in tanycytes at the blood-CSF barrier (Mullier et al., 2010). In parallel, antibodies to MECA-32 were used to selectively label the fenestral diaphragms of ME endothelial cells (Ciofi et al., 2009; Mullier et al., 2010).
Fasting increased the organization of tanycytic tight junction complexes in both the ME and ARH (Fig. 1A,B; Fig. S1A), but not in standard non-ARH hypothalamic vessels composing the BBB (Fig. S1B). Fasting intensified the honeycomb-like pattern of ZO-1 and occludin around the apical pole of tanycytes overlying the ME (Fig. 1A, insets 2,5; Fig. S1A, insets 3,6), and switched ZO-1 and occludin distribution from a diffuse apical pattern to a honeycomb pattern in tanycytes lining the third ventricle (3V) next to the ARH (Fig. 1A, insets 1,4; Fig. S1A, insets 2,5). Claudin-1 expression in ARH tanycytes also became organized in food-deprived mice (Fig. 1C,D). This reorganization was associated with increased protein levels of ZO-1, but not of occludin or claudin-1, in fasting animals (Fig. S1D). Finally, food-deprived mice refed for an additional 24h displayed a reappearance of the hypothalamic barrier phenotype of fed mice (Fig. 1B-D).
In fasting mice, ME microvessel loops, some of which extend up to the ARH, also demonstrated a marked increase in fenestration associated with changes in MECA-32 distribution (Fig. 1A, insets 1,4; Fig. 1B) and the appearance of fenestrated diaphragms at the ultrastructural level (Fig. 1E, inset 2; Fig. S1C). Western blotting revealed that this increased fenestration was accompanied by a significant increase in MECA-32 protein levels (p<0.05; Fig. 1F). Together, the fasting-induced structural reorganization of the tight junction complexes of tanycytes lining the ventricular wall adjoining the ARH potentially limits paracellular diffusion between the tissue and the CSF, while ME microvessels simultaneously become more leaky. These reversible morphological alterations at the BHB suggest that nutritional state modulates the access of metabolic signals from the periphery to ARH neurons critical for energy homeostasis.
Glucose deprivation mediates fasting-induced plasticity at the BHB
What nutritional factors underlie these profound morphological changes in fasting mice? Blood glucose acts as a metabolic signal that can alter the activity of hypothalamic neurons (Levin et al., 2011; Thorens and Larsen, 2004) and evokes robust signaling in tanycytes, which act as glucosensors (Frayling et al., 2011), suggesting that reduced glucose levels could underlie the morphological changes seen during fasting. As expected, blood glucose levels were significantly lower after 24h of food deprivation (Fig. 2A, p<0.001). Levels were normalized and fasting-induced BHB reorganization prevented (Fig. 2B) by intravenous (i.v.) glucose (30%, 1–3 µl/min; n = 6) but not saline infusion (n = 4). On the other hand, central neuroglucopenia induced by the intraperitoneal (i.p.) or intracerebroventricular (i.c.v.) injection of 2-deoxy-D-glucose (2-DG, 300 mg/kg i.p. or 1 mg/mouse i.c.v.; n = 4 each, Fig. 2B; Fig. S2A), a glucose analog that inhibits glucose metabolism but can act on tanycytes (Frayling et al., 2011), caused the reorganization of tanycytic tight junction complexes and associated microvessel loops to resemble those in fasting mice. The central detection of glucose deprivation thus appears to play a key role in BHB reorganization after fasting.
VEGF promotes changes in ME microvessel loops and tanycytic tight junction complexes during fasting
Quantitative RT-PCR analyses showed that fasting induced a constellation of hypothalamic transcripts involved in controlling the structural plasticity of the brain (Fig. 2C). However, only VEGF-A expression was upregulated in both fasting and 2-DG-treated animals (Fig. 2C). VEGF protein levels increased in the hypothalamus of 24h-fasting mice when compared to fed mice (p<0.05; Fig. 2D), and 2DG-treatment for 12h directly triggered VEGF secretion from ME-ARH explants in vitro (p<0.05; Fig. 2E). We therefore treated mice for 24h with Axitinib (25 mg/kg/12h, i.p. or 70 µg/mouse, i.c.v.), a tyrosine kinase inhibitor that selectively inhibits VEGF receptors (VEGFR) 1, 2 and 3 (Mancuso et al., 2006). Axitinib reversed both fasting- and 2DG-evoked changes in tight junction complex organization in tanycytes and the associated microvessel loops in the ME/ARH region (Fig. 2B,F; Fig. S2A,B). Conversely, VEGF infusion (60 µg/kg/12h, i.p. or 100 ng/mouse, i.c.v.) for 24h mimicked the effects of fasting on BHB plasticity in fed mice (Fig. 2F,G; Fig. S2B) without affecting tight junction organization in BBB capillaries (Fig. S1B).
Next, we examined the contribution of VEGFR1, 2 and 3, which are all expressed in the mediobasal hypothalamus (MBH; Fig. S2C), to fasting-induced BHB plasticity, by administering selective neutralizing monoclonal antibodies to each receptor (40 mg/kg, i.p.) (Pytowski et al., 2005; Wang et al., 2004). While antibodies to VEGFR1 and 3 bound to both hypothalamic BBB microvessels and the ME microvessel plexus, antibodies to VEGFR2 bound only to the latter (Fig. S2D). Importantly, the blockade of VEGFR2 and, to a lesser extent, VEGFR1, inhibited the BHB rearrangement observed after fasting (Fig. S2E,F), while antibodies to VEGFR3 had no effect on this plasticity (Fig. S2F). Together, these data suggest that increased hypothalamic VEGF levels during food deprivation target VEGFR1 and 2 in ME endothelial cells to promote microvessel permeability and tight junction complex reorganization in the ME and ARH.
Fasting-induced BHB plasticity requires tanycytic VEGF-A expression
VEGF-A mRNA expression in the hypothalamic tuberal region is restricted to tanycytes (Allen Brain atlas, http://mouse.brain-map.org/experiment/show/74988747, Fig. S3A), suggesting that these cells play a role in the control of VEGF-mediated BHB plasticity. Cell sorting experiments using tdTomato reporter mice in which the tat-cre fusion protein, whose cellular uptake is enhanced compared to cre recombinase (Peitz et al., 2002), was stereotaxically infused into the 3 ventricle, where it selectively targeted tanycytes (Fig. S2G), to study fasting-dependent changes in tanycytic gene expression in vivo (Fig. 2H). Sorted Tomato-positive cells abundantly expressed the tanycytic marker DARPP-32 (Hokfelt et al., 1988), which was barely detectable in Tomato-negative cells (Fig. S2I). Purified tanycytes also expressed GLUT1, GLUT2 and glucokinase transcripts, as suggested by others (Rodriguez et al., 2005), and low levels of VEGFR1 and 2, although VEGFR3 was undetectable (Fig. S2I). Intriguingly, fasting upregulated VEGF-A but not VEGF-B or VEGF-C mRNA expression in tanycytes (Fig. 2H; Fig. S2J), with a concomitant increase in the transcript for hypoxia-inducible factor 1 alpha (HIF-1α) (Fig. S2K), recently shown to be involved in hypothalamic glucosensing (Zhang et al., 2011) and known to promote VEGF expression (Carmeliet et al., 1998). Finally, Vegfa deletion in tanycytes by tat-cre infusion into the third ventricle of Vegfa mice (Fig. S3A) abolished fasting-induced BHB reorganization (n = 4; Fig. 2F) but did not affect BHB properties in animals fed ad libitum (n = 4; not shown). Together these data demonstrate that tanycytic VEGF-A expression plays a key role in regulating fasting-induced BHB plasticity.
BHB plasticity modulates the access of blood-borne metabolic factors to the ARH
To establish whether the morphological changes described in the MBH above were associated with altered permeability and access of blood-borne molecules to the ARH, the extravasation of intravenously-injected Evans Blue dye from brain microvessels into the hypothalamus was compared between 24h-fasting and fed mice. In fed mice, Evans Blue diffusion was restricted to the vascular bed of hypothalamic BBB microvessels and the ME (Fig. 3A, top panels), and did not spread to neighboring structures such as the ARH. In striking contrast, in 24h-fasting mice, the dye was observed in the ventromedial ARH (vmARH), where plastic BHB changes were observed in previous experiments (Fig. 3A,B). As with the morphological changes, refeeding reversed dye diffusion into the vmARH (Fig. 3B). Importantly, fasting did not promote dye extravasation from the intrinsic hypothalamic microvessels that compose the BBB (Fig. S3B). The inhibition of BHB reorganization in food-deprived mice by Axitinib, VEGFR2-neutralizing antibodies, Vegfa gene targeting or glucose infusion prevented dye diffusion into the ARH, whereas the induction of barrier plasticity with VEGF in fed mice clearly elicited such diffusion (Fig. 3B). These findings strongly suggest that the anatomical changes at the BHB in fasting animals facilitate the access of blood-borne signals to the vmARH due to its increased permeability.
To directly test whether fasting-induced structural rearrangements at the BHB do indeed increase the access of critical metabolic substrates to the ARH, we simultaneously assessed glucose levels in the ARH and the adjacent ventromedial nucleus of the hypothalamus (VMH) by microdialysis in fed or 24h-fasting rats (Fig. 3C). First, rats exhibited similar fasting-induced morphological and functional changes to the BHB as did mice (Fig. S3C-E). While ARH glucose levels in fed rats were comparable to those in the VMH (Fig. 3C), in 24h-fasting rats, ARH glucose levels were 300% higher than VMH levels (Fig. 3C). In keeping with the restricted occurrence of microvessels whose permeability changes in response to feeding status in the vmARH, and unlike what would have been expected if the BBB were completely permeable throughout the ARH, ARH glucose levels during fasting (≈ 2mM) never reached blood levels (≈ 5mM). However, these data do suggest that fasting-induced structural changes at the BHB create a privileged route for the access of circulating glucose to glucosensing ARH neurons, bypassing both the BBB and the blood-CSF barrier (Fig. 3F).
We further explored this hypothesis by examining the ability of exogenous leptin, a 16-kDa peptide hormone, to access ARH neurons by quantifying the leptin-stimulated phosphorylation of STAT3 in fasting and fed mice. Leptin (3 mg/kg, i.p.) induced a 30% increase in immunoreactivity for phosphorylated STAT3 (pSTAT3) in food-deprived mice when compared to fed mice (Fig. 3D). This increase was restricted to the vmARH (Fig. S3F). Axitinib (which inhibits BHB rearrangement; Fig. 2F) prevented the leptin-induced increase in pSTAT3 in fasting animals (Fig. 3D,E). Conversely, the treatment of fed mice (in which endogenous leptin levels are significantly higher than in fasting mice; Fig. S3G), with VEGF, which promotes BHB reorganization (Fig. 2F,G), markedly increased STAT3 activation in the ARH (Fig. 3D,E). This effect was blunted by the i.p. injection of a mutated recombinant leptin antagonist (LAN; 3 mg/kg) that is devoid of biological activity but effectively binds to the leptin receptor (Niv-Spector et al., 2005) 45 min before death (Fig. 3D,E), suggesting that leptin access to the ARH is facilitated by VEGF treatment in fed mice. Together, these data suggest an important role for BHB plasticity in modulating the access of metabolic factors to the ARH.
BHB plasticity modulates feeding
To evaluate the functional consequences of BHB plasticity to feeding behavior, we measured refeeding after fasting in control and Axitinib-treated mice. Compared to vehicle-infused mice refed after a 24h fast, food intake was significantly lower in Axitinib-treated refed mice (Fig. 4A). This difference, which occurred primarily during the first 30 min of refeeding (Fig. 4A), was associated with decreased body weight gain 24h and 48h after Axitinib administration (Fig. 4B). Importantly, Axitinib alone did not inhibit food intake in animals fed ad libitum (Fig. 4A), suggesting that reduced feeding in fasting mice was not due to food aversion. Conversely, food intake significantly increased in the afternoon and at lights-off (when endogenous anorectic hormones are thought to stimulate food intake) in fed mice 24h after the initiation of VEGF treatment, when compared to vehicle-treated mice (Fig. 4C). Subsequently, we found that the anorectic and weight-loss-inducing effects of exogenous leptin were greater in fed mice treated with VEGF than in vehicle-treated controls (Fig. 4D). Together, these findings suggest that VEGF-mediated structural changes at the BHB, by modulating the access of blood-borne metabolic substrates to the ARH, play an important role in the adaptive response to fasting.
Fasting-induced plasticity of the blood-hypothalamus barrier (BHB)
To investigate the role of BHB plasticity in the adaptive response to fasting, we compared the hypothalami of mice deprived of food for 24h and those fed ad libitum using immunofluorescence for two constitutive tight junction proteins, zonula occludens-1 (ZO-1) and occludin, expressed in BBB endothelial cells and tanycytes, and claudin-1, expressed in tanycytes at the blood-CSF barrier (Mullier et al., 2010). In parallel, antibodies to MECA-32 were used to selectively label the fenestral diaphragms of ME endothelial cells (Ciofi et al., 2009; Mullier et al., 2010).
Fasting increased the organization of tanycytic tight junction complexes in both the ME and ARH (Fig. 1A,B; Fig. S1A), but not in standard non-ARH hypothalamic vessels composing the BBB (Fig. S1B). Fasting intensified the honeycomb-like pattern of ZO-1 and occludin around the apical pole of tanycytes overlying the ME (Fig. 1A, insets 2,5; Fig. S1A, insets 3,6), and switched ZO-1 and occludin distribution from a diffuse apical pattern to a honeycomb pattern in tanycytes lining the third ventricle (3V) next to the ARH (Fig. 1A, insets 1,4; Fig. S1A, insets 2,5). Claudin-1 expression in ARH tanycytes also became organized in food-deprived mice (Fig. 1C,D). This reorganization was associated with increased protein levels of ZO-1, but not of occludin or claudin-1, in fasting animals (Fig. S1D). Finally, food-deprived mice refed for an additional 24h displayed a reappearance of the hypothalamic barrier phenotype of fed mice (Fig. 1B-D).
In fasting mice, ME microvessel loops, some of which extend up to the ARH, also demonstrated a marked increase in fenestration associated with changes in MECA-32 distribution (Fig. 1A, insets 1,4; Fig. 1B) and the appearance of fenestrated diaphragms at the ultrastructural level (Fig. 1E, inset 2; Fig. S1C). Western blotting revealed that this increased fenestration was accompanied by a significant increase in MECA-32 protein levels (p<0.05; Fig. 1F). Together, the fasting-induced structural reorganization of the tight junction complexes of tanycytes lining the ventricular wall adjoining the ARH potentially limits paracellular diffusion between the tissue and the CSF, while ME microvessels simultaneously become more leaky. These reversible morphological alterations at the BHB suggest that nutritional state modulates the access of metabolic signals from the periphery to ARH neurons critical for energy homeostasis.
Glucose deprivation mediates fasting-induced plasticity at the BHB
What nutritional factors underlie these profound morphological changes in fasting mice? Blood glucose acts as a metabolic signal that can alter the activity of hypothalamic neurons (Levin et al., 2011; Thorens and Larsen, 2004) and evokes robust signaling in tanycytes, which act as glucosensors (Frayling et al., 2011), suggesting that reduced glucose levels could underlie the morphological changes seen during fasting. As expected, blood glucose levels were significantly lower after 24h of food deprivation (Fig. 2A, p<0.001). Levels were normalized and fasting-induced BHB reorganization prevented (Fig. 2B) by intravenous (i.v.) glucose (30%, 1–3 µl/min; n = 6) but not saline infusion (n = 4). On the other hand, central neuroglucopenia induced by the intraperitoneal (i.p.) or intracerebroventricular (i.c.v.) injection of 2-deoxy-D-glucose (2-DG, 300 mg/kg i.p. or 1 mg/mouse i.c.v.; n = 4 each, Fig. 2B; Fig. S2A), a glucose analog that inhibits glucose metabolism but can act on tanycytes (Frayling et al., 2011), caused the reorganization of tanycytic tight junction complexes and associated microvessel loops to resemble those in fasting mice. The central detection of glucose deprivation thus appears to play a key role in BHB reorganization after fasting.
VEGF promotes changes in ME microvessel loops and tanycytic tight junction complexes during fasting
Quantitative RT-PCR analyses showed that fasting induced a constellation of hypothalamic transcripts involved in controlling the structural plasticity of the brain (Fig. 2C). However, only VEGF-A expression was upregulated in both fasting and 2-DG-treated animals (Fig. 2C). VEGF protein levels increased in the hypothalamus of 24h-fasting mice when compared to fed mice (p<0.05; Fig. 2D), and 2DG-treatment for 12h directly triggered VEGF secretion from ME-ARH explants in vitro (p<0.05; Fig. 2E). We therefore treated mice for 24h with Axitinib (25 mg/kg/12h, i.p. or 70 µg/mouse, i.c.v.), a tyrosine kinase inhibitor that selectively inhibits VEGF receptors (VEGFR) 1, 2 and 3 (Mancuso et al., 2006). Axitinib reversed both fasting- and 2DG-evoked changes in tight junction complex organization in tanycytes and the associated microvessel loops in the ME/ARH region (Fig. 2B,F; Fig. S2A,B). Conversely, VEGF infusion (60 µg/kg/12h, i.p. or 100 ng/mouse, i.c.v.) for 24h mimicked the effects of fasting on BHB plasticity in fed mice (Fig. 2F,G; Fig. S2B) without affecting tight junction organization in BBB capillaries (Fig. S1B).
Next, we examined the contribution of VEGFR1, 2 and 3, which are all expressed in the mediobasal hypothalamus (MBH; Fig. S2C), to fasting-induced BHB plasticity, by administering selective neutralizing monoclonal antibodies to each receptor (40 mg/kg, i.p.) (Pytowski et al., 2005; Wang et al., 2004). While antibodies to VEGFR1 and 3 bound to both hypothalamic BBB microvessels and the ME microvessel plexus, antibodies to VEGFR2 bound only to the latter (Fig. S2D). Importantly, the blockade of VEGFR2 and, to a lesser extent, VEGFR1, inhibited the BHB rearrangement observed after fasting (Fig. S2E,F), while antibodies to VEGFR3 had no effect on this plasticity (Fig. S2F). Together, these data suggest that increased hypothalamic VEGF levels during food deprivation target VEGFR1 and 2 in ME endothelial cells to promote microvessel permeability and tight junction complex reorganization in the ME and ARH.
Fasting-induced BHB plasticity requires tanycytic VEGF-A expression
VEGF-A mRNA expression in the hypothalamic tuberal region is restricted to tanycytes (Allen Brain atlas, http://mouse.brain-map.org/experiment/show/74988747, Fig. S3A), suggesting that these cells play a role in the control of VEGF-mediated BHB plasticity. Cell sorting experiments using tdTomato reporter mice in which the tat-cre fusion protein, whose cellular uptake is enhanced compared to cre recombinase (Peitz et al., 2002), was stereotaxically infused into the 3 ventricle, where it selectively targeted tanycytes (Fig. S2G), to study fasting-dependent changes in tanycytic gene expression in vivo (Fig. 2H). Sorted Tomato-positive cells abundantly expressed the tanycytic marker DARPP-32 (Hokfelt et al., 1988), which was barely detectable in Tomato-negative cells (Fig. S2I). Purified tanycytes also expressed GLUT1, GLUT2 and glucokinase transcripts, as suggested by others (Rodriguez et al., 2005), and low levels of VEGFR1 and 2, although VEGFR3 was undetectable (Fig. S2I). Intriguingly, fasting upregulated VEGF-A but not VEGF-B or VEGF-C mRNA expression in tanycytes (Fig. 2H; Fig. S2J), with a concomitant increase in the transcript for hypoxia-inducible factor 1 alpha (HIF-1α) (Fig. S2K), recently shown to be involved in hypothalamic glucosensing (Zhang et al., 2011) and known to promote VEGF expression (Carmeliet et al., 1998). Finally, Vegfa deletion in tanycytes by tat-cre infusion into the third ventricle of Vegfa mice (Fig. S3A) abolished fasting-induced BHB reorganization (n = 4; Fig. 2F) but did not affect BHB properties in animals fed ad libitum (n = 4; not shown). Together these data demonstrate that tanycytic VEGF-A expression plays a key role in regulating fasting-induced BHB plasticity.
BHB plasticity modulates the access of blood-borne metabolic factors to the ARH
To establish whether the morphological changes described in the MBH above were associated with altered permeability and access of blood-borne molecules to the ARH, the extravasation of intravenously-injected Evans Blue dye from brain microvessels into the hypothalamus was compared between 24h-fasting and fed mice. In fed mice, Evans Blue diffusion was restricted to the vascular bed of hypothalamic BBB microvessels and the ME (Fig. 3A, top panels), and did not spread to neighboring structures such as the ARH. In striking contrast, in 24h-fasting mice, the dye was observed in the ventromedial ARH (vmARH), where plastic BHB changes were observed in previous experiments (Fig. 3A,B). As with the morphological changes, refeeding reversed dye diffusion into the vmARH (Fig. 3B). Importantly, fasting did not promote dye extravasation from the intrinsic hypothalamic microvessels that compose the BBB (Fig. S3B). The inhibition of BHB reorganization in food-deprived mice by Axitinib, VEGFR2-neutralizing antibodies, Vegfa gene targeting or glucose infusion prevented dye diffusion into the ARH, whereas the induction of barrier plasticity with VEGF in fed mice clearly elicited such diffusion (Fig. 3B). These findings strongly suggest that the anatomical changes at the BHB in fasting animals facilitate the access of blood-borne signals to the vmARH due to its increased permeability.
To directly test whether fasting-induced structural rearrangements at the BHB do indeed increase the access of critical metabolic substrates to the ARH, we simultaneously assessed glucose levels in the ARH and the adjacent ventromedial nucleus of the hypothalamus (VMH) by microdialysis in fed or 24h-fasting rats (Fig. 3C). First, rats exhibited similar fasting-induced morphological and functional changes to the BHB as did mice (Fig. S3C-E). While ARH glucose levels in fed rats were comparable to those in the VMH (Fig. 3C), in 24h-fasting rats, ARH glucose levels were 300% higher than VMH levels (Fig. 3C). In keeping with the restricted occurrence of microvessels whose permeability changes in response to feeding status in the vmARH, and unlike what would have been expected if the BBB were completely permeable throughout the ARH, ARH glucose levels during fasting (≈ 2mM) never reached blood levels (≈ 5mM). However, these data do suggest that fasting-induced structural changes at the BHB create a privileged route for the access of circulating glucose to glucosensing ARH neurons, bypassing both the BBB and the blood-CSF barrier (Fig. 3F).
We further explored this hypothesis by examining the ability of exogenous leptin, a 16-kDa peptide hormone, to access ARH neurons by quantifying the leptin-stimulated phosphorylation of STAT3 in fasting and fed mice. Leptin (3 mg/kg, i.p.) induced a 30% increase in immunoreactivity for phosphorylated STAT3 (pSTAT3) in food-deprived mice when compared to fed mice (Fig. 3D). This increase was restricted to the vmARH (Fig. S3F). Axitinib (which inhibits BHB rearrangement; Fig. 2F) prevented the leptin-induced increase in pSTAT3 in fasting animals (Fig. 3D,E). Conversely, the treatment of fed mice (in which endogenous leptin levels are significantly higher than in fasting mice; Fig. S3G), with VEGF, which promotes BHB reorganization (Fig. 2F,G), markedly increased STAT3 activation in the ARH (Fig. 3D,E). This effect was blunted by the i.p. injection of a mutated recombinant leptin antagonist (LAN; 3 mg/kg) that is devoid of biological activity but effectively binds to the leptin receptor (Niv-Spector et al., 2005) 45 min before death (Fig. 3D,E), suggesting that leptin access to the ARH is facilitated by VEGF treatment in fed mice. Together, these data suggest an important role for BHB plasticity in modulating the access of metabolic factors to the ARH.
BHB plasticity modulates feeding
To evaluate the functional consequences of BHB plasticity to feeding behavior, we measured refeeding after fasting in control and Axitinib-treated mice. Compared to vehicle-infused mice refed after a 24h fast, food intake was significantly lower in Axitinib-treated refed mice (Fig. 4A). This difference, which occurred primarily during the first 30 min of refeeding (Fig. 4A), was associated with decreased body weight gain 24h and 48h after Axitinib administration (Fig. 4B). Importantly, Axitinib alone did not inhibit food intake in animals fed ad libitum (Fig. 4A), suggesting that reduced feeding in fasting mice was not due to food aversion. Conversely, food intake significantly increased in the afternoon and at lights-off (when endogenous anorectic hormones are thought to stimulate food intake) in fed mice 24h after the initiation of VEGF treatment, when compared to vehicle-treated mice (Fig. 4C). Subsequently, we found that the anorectic and weight-loss-inducing effects of exogenous leptin were greater in fed mice treated with VEGF than in vehicle-treated controls (Fig. 4D). Together, these findings suggest that VEGF-mediated structural changes at the BHB, by modulating the access of blood-borne metabolic substrates to the ARH, play an important role in the adaptive response to fasting.
Discussion
Energy homeostasis requires increased food intake when energy stores are depleted, and a means of signaling this depletion to central neurons that control feeding behavior (Cowley et al., 2001; Cowley et al., 2003; Dunn-Meynell et al., 2002; Elias et al., 1999; Liu et al., 2012), such as ARH neurons (Hill et al., 2010). Here, we provide evidence that the BHB (Mullier et al., 2010) undergoes dynamic and reversible structural changes that modulate its permeability in response to glucose and tanycytic VEGF levels, thereby acting as a checkpoint in the access of peripheral metabolic signals to ARH neurons.
Our data show that fasting-evoked dips in blood glucose levels trigger VEGF-A expression in tanycytes and VEGF accumulation in the hypothalamic ME, which acts via VEGFR to promote endothelial cell fenestration. Tanycytes contacting these newly permeable microvessel loops then reorganize their tight junction complexes to seal the paracellular space between the parenchyma and the CSF. In consequence, some target neurons in the vmARH are no longer insulated by the blood-brain and blood-CSF barriers but become directly exposed to peripheral metabolic signals (Fig. 3F), a situation reversed upon refeeding. This increased accessibility is confirmed both by the leakage of intravenously injected dye into the vmARH and by higher physiological glucose levels in the ARH than in the adjacent VMH in the fasting state. Fasting-induced reorganization is blocked by the VEGFR inhibitor Axitinib, which, consistent with the importance of these changes in the adaptive response to fasting, reduces food intake and weight gain when food-deprived mice are refed. Even though Axitinib alone did not affect food intake in control animals fed ad libitum, chronic Axitinib treatment, which causes a marked reduction in endothelial cell fenestration (Kamba et al., 2006), is known to affect glucose homeostasis (Kamba et al., 2006) and to lead to decreased appetite and weight loss in patients (Fruehauf et al., 2011; Rini et al., 2009).
Conversely, triggering BHB permeability in mice fed ad libitum with exogenous VEGF, which signals food deprivation, significantly increases their food intake and sensitivity to the anorectic effects of leptin. We used leptin as a surrogate for other large metabolic peptides such as ghrelin because we could assess its access to the vmARH by monitoring the activation of STAT3. In fact, leptin levels are decreased during fasting, leading to a marked anabolic state within the hypothalamus, with increased neuropeptide Y and agouti-related peptide (AgRP) and decreased proopiomelanocortin levels. This combination acts as a potent stimulus for the animal to seek and ingest food. Contrarily, ghrelin levels increase during fasting, and the increased permeability of ARH microvessels to this peptide could facilitate its selective activation of anabolic AgRP neurons (Elmquist et al., 2005; Gao and Horvath, 2007; Levin et al., 2011; Sawchenko, 1998).
Falling blood glucose levels and decreased glucose metabolism appear to be critical signals for the initiation of the BHB response to starvation. Glucose replacement prevents this rearrangement while 2-DG-induced central glucopenia reproduces it. The molecular pathways that underlie the subsequent accumulation of VEGF in the ME are unknown. However, our results showing that VEGF-A inactivation in these cells blunts fasting-induced BHB plasticity strongly suggest that they are intrinsic to tanycytes, which directly and rapidly respond to changes in glucose levels (Frayling et al., 2011). This response could involve HIF-1α, which is upregulated in tanycytes by fasting (Fig. S2J), and is known to be activated by glucoprivation and to promote VEGF expression (Carmeliet et al., 1998; Zhang et al., 2011).
VEGF has long been associated with increased vascular permeability (Esser et al., 1998; Ioannidou et al., 2006) and is required for ependymal cell function and the maintenance of key brain-periphery interfaces such as the choroid plexus (Maharaj et al., 2008). VEGF might also contribute to increased BBB permeability in diseased (Argaw et al., 2009) but not healthy brains (Hawkins et al., 2010) (Fig. S1B; Fig. S3A). Our data suggest that VEGF and its signaling receptor, VEGFR2, are key determinants of fasting-induced structural rearrangements at the BHB. Although VEGF involvement in endothelial cell fenestration and the expression of diaphragm proteins (named PV-1 in rats and MECA-32 in mice) is clearly documented (Esser et al., 1998; Kamba et al., 2006), the mechanisms underlying tight-junction-complex remodeling in tanycytes, which express low levels of VEGFR2, are unknown. These could also involve VEGF signaling, which induces posttranslational modifications in tight junction proteins under some pathological conditions (Murakami et al., 2009), or other as-yet undiscovered signals released by endothelial cells upon fenestration.
Tight junction proteins such as occludin, whose organization is modified in tanycytes under fasting conditions, could also be involved in brain metabolic sensing and body-weight regulation. Occludin-null mice are leaner than their wild-type littermates (Saitou et al., 2000), and the i.c.v. infusion of antisense oligodeoxynucleotides to occludin restores leptin sensitivity in an animal model of diet-induced leptin resistance and hyperglycemia (Oh et al., 2005) in which fasting appears not to promote changes in BHB permeability (unpublished observations). This, together with recent findings showing that a high-fat diet triggers neurogenic activity in ME tanycytes (Lee et al., 2012), shows that these cells play a dynamic role in metabolic sensing and hold therapeutic potential in metabolic disorders.
Overall our data unveil a new physiological concept in the maintenance of energy homeostasis, in which blood glucose levels, by regulating tanycytic VEGF-A expression, modulate the organization of their tight junctions as well as the permeability of ME capillary loops in the vmARH, and thereby control the access of circulating homeostatic signals to brain circuits that regulate metabolism.
Experimental procedures
Animals
Male C57Bl/6 mice, 3–4 months old (Charles River, France) and male Sprague Dawley rats (Charles River, USA) were given ad libitum access to water and standard laboratory chow. tdTomato reporter mice were purchased from the Jackson laboratories (Maine, USA) and Vegfa mice (Gerber et al., 1999) were a gift from N. Ferrara (Novartis, USA). Animal studies were approved by the Institutional Animal Care and Use Committee of Lille and the East Orange Veterans Affairs Medical Center.
Treatment protocols
Glucose infusion and 2-deoxy-D-glucose (2-DG) injection
Mice fed ad libitum and fasting mice were anesthetized with isoflurane and the jugular vein catheterized. After a 7-day recovery period, fed mice were infused with saline solution (0.9%) for 24h while fasting mice were infused with glucose (30% in saline) or saline. Moreover, mice fed ad libitum were given an i.p. or i.c.v. injection of 2-DG (RDS, France) in saline or an equal volume of saline alone, following procedures described previously (Mullier et al., 2010).
Anti-VEGF and VEGF treatments
Mice were subjected to an i.p. or i.c.v. infusion of Axitinib (in DMSO, LC Laboratories, France) or an equal volume of DMSO during the 24h fasting period. Finally, mice fed ad libitum were given an i.p. or i.c.v. infusion of recombinant mouse VEGF 164 (RDS, France) in PBS for 24h.
Tat-cre delivery
A tat-cre fusion protein produced as detailed previously (Peitz et al., 2002) was stereotaxically infused into the third ventricle (1.5 μl over 5 min at 2.1 mg/ml; AP: −1.7 mm, ML: 0 mm DV: - 5.6 mm) of isoflurane-anesthetized floxed mice 24h before experiments.
Fluorescence-activated cell-sorting and real-time PCR analyses
Tomato-positive cells were sorted and collected from ME explants microdissected from fasting and fed mice and processed for quantitative RT-PCR, as described in the Supplemental Experimental Procedures.
Physiological measurements
Food intake
Mice were housed 3 per cage with preweighed amounts of food dispensed through the wire cage tops, and food intake was measured every 30 min for the first 3 hours and every hour for 24 hours. The average and cumulative food intake of 3 mice was used for statistical comparisons (n = 4 cages per group).
In vivo leptin sensitivity test
Mice were housed in individual cages two days before the beginning of the experiment. Mice fed ad libitum and fed mice infused with VEGF were injected i.p. at 18:00 with vehicle (5 mM sodium citrate buffer, pH 4.0) or leptin (3 mg/kg, PeproTech, France). Body weight and food consumption were measured at 08:00 the next day.
Permeability assays, immunohistochemistry and image analysis
Mice were given i.v. injections of sterile 1% Evans Blue dye (Sigma, France) in 0.9% saline (50 μl) into the tail vein and killed by decapitation 20 min later. Brains were processed for immunofluorescence as described previously (Mullier et al., 2010). The primary antibodies used were: polyclonal rabbit anti-zonula occludens-1 (ZO-1, 1:500, Zymed, USA), rabbit anticlaudin-1 (1:200, Zymed, USA), chicken anti-vimentin (1:2000, Chemicon, France), and rat anti-MECA-32 (1:200, gift from Pr Britta Engelhardt, Switzerland). Additional details appear in the Supplemental Experimental Procedures and Fig. S4.
For pSTAT3 immunolabeling and analysis, mice were injected i.p. with vehicle (5 mM sodium citrate buffer, n = 4 per group), leptin (PeproTech, France) or LAN (Protein Laboratories Rehovot Ltd, Israel) and perfused 45 min later with a 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were processed for pSTAT3 immunolabeling and quantification as described previously (Bouret et al., 2012). Additional details appear in the Supplemental Experimental Procedures.
Microdialysis of the hypothalamus
Placement of hypothalamic cannulae and the assessment of ARH and VMH glucose levels in male rats (n = 9) was performed as described previously (Dunn-Meynell et al., 2009) and detailed in the Supplemental Experimental Procedures.
Immunoblotting
Frozen microdissected ME and MBH of mice fed ad libitum (n = 3) and those fasting for 24h (n = 3) were immunoblotted as described in the Supplemental Experimental Procedures. Rabbit anti-claudin-1 (1:1000, Zymed, USA), mouse anti-VEGF (1:500, SantaCruz), rat anti-MECA-32 (1:500, Santa Cruz) and goat anti-actin antibodies (1:500, Santa Cruz, France) were used in these experiments.
Statistical analysis
All values are expressed as means ± SEM. Data were analyzed for statistical significance with SigmaPlot software (Version 11.0), using one-way or two-way ANOVA followed by a Tukey post hoc test when appropriate. P-values of less than 0.05 were considered to be statistically significant.
Animals
Male C57Bl/6 mice, 3–4 months old (Charles River, France) and male Sprague Dawley rats (Charles River, USA) were given ad libitum access to water and standard laboratory chow. tdTomato reporter mice were purchased from the Jackson laboratories (Maine, USA) and Vegfa mice (Gerber et al., 1999) were a gift from N. Ferrara (Novartis, USA). Animal studies were approved by the Institutional Animal Care and Use Committee of Lille and the East Orange Veterans Affairs Medical Center.
Treatment protocols
Glucose infusion and 2-deoxy-D-glucose (2-DG) injection
Mice fed ad libitum and fasting mice were anesthetized with isoflurane and the jugular vein catheterized. After a 7-day recovery period, fed mice were infused with saline solution (0.9%) for 24h while fasting mice were infused with glucose (30% in saline) or saline. Moreover, mice fed ad libitum were given an i.p. or i.c.v. injection of 2-DG (RDS, France) in saline or an equal volume of saline alone, following procedures described previously (Mullier et al., 2010).
Anti-VEGF and VEGF treatments
Mice were subjected to an i.p. or i.c.v. infusion of Axitinib (in DMSO, LC Laboratories, France) or an equal volume of DMSO during the 24h fasting period. Finally, mice fed ad libitum were given an i.p. or i.c.v. infusion of recombinant mouse VEGF 164 (RDS, France) in PBS for 24h.
Tat-cre delivery
A tat-cre fusion protein produced as detailed previously (Peitz et al., 2002) was stereotaxically infused into the third ventricle (1.5 μl over 5 min at 2.1 mg/ml; AP: −1.7 mm, ML: 0 mm DV: - 5.6 mm) of isoflurane-anesthetized floxed mice 24h before experiments.
Glucose infusion and 2-deoxy-D-glucose (2-DG) injection
Mice fed ad libitum and fasting mice were anesthetized with isoflurane and the jugular vein catheterized. After a 7-day recovery period, fed mice were infused with saline solution (0.9%) for 24h while fasting mice were infused with glucose (30% in saline) or saline. Moreover, mice fed ad libitum were given an i.p. or i.c.v. injection of 2-DG (RDS, France) in saline or an equal volume of saline alone, following procedures described previously (Mullier et al., 2010).
Anti-VEGF and VEGF treatments
Mice were subjected to an i.p. or i.c.v. infusion of Axitinib (in DMSO, LC Laboratories, France) or an equal volume of DMSO during the 24h fasting period. Finally, mice fed ad libitum were given an i.p. or i.c.v. infusion of recombinant mouse VEGF 164 (RDS, France) in PBS for 24h.
Tat-cre delivery
A tat-cre fusion protein produced as detailed previously (Peitz et al., 2002) was stereotaxically infused into the third ventricle (1.5 μl over 5 min at 2.1 mg/ml; AP: −1.7 mm, ML: 0 mm DV: - 5.6 mm) of isoflurane-anesthetized floxed mice 24h before experiments.
Fluorescence-activated cell-sorting and real-time PCR analyses
Tomato-positive cells were sorted and collected from ME explants microdissected from fasting and fed mice and processed for quantitative RT-PCR, as described in the Supplemental Experimental Procedures.
Physiological measurements
Food intake
Mice were housed 3 per cage with preweighed amounts of food dispensed through the wire cage tops, and food intake was measured every 30 min for the first 3 hours and every hour for 24 hours. The average and cumulative food intake of 3 mice was used for statistical comparisons (n = 4 cages per group).
In vivo leptin sensitivity test
Mice were housed in individual cages two days before the beginning of the experiment. Mice fed ad libitum and fed mice infused with VEGF were injected i.p. at 18:00 with vehicle (5 mM sodium citrate buffer, pH 4.0) or leptin (3 mg/kg, PeproTech, France). Body weight and food consumption were measured at 08:00 the next day.
Food intake
Mice were housed 3 per cage with preweighed amounts of food dispensed through the wire cage tops, and food intake was measured every 30 min for the first 3 hours and every hour for 24 hours. The average and cumulative food intake of 3 mice was used for statistical comparisons (n = 4 cages per group).
In vivo leptin sensitivity test
Mice were housed in individual cages two days before the beginning of the experiment. Mice fed ad libitum and fed mice infused with VEGF were injected i.p. at 18:00 with vehicle (5 mM sodium citrate buffer, pH 4.0) or leptin (3 mg/kg, PeproTech, France). Body weight and food consumption were measured at 08:00 the next day.
Permeability assays, immunohistochemistry and image analysis
Mice were given i.v. injections of sterile 1% Evans Blue dye (Sigma, France) in 0.9% saline (50 μl) into the tail vein and killed by decapitation 20 min later. Brains were processed for immunofluorescence as described previously (Mullier et al., 2010). The primary antibodies used were: polyclonal rabbit anti-zonula occludens-1 (ZO-1, 1:500, Zymed, USA), rabbit anticlaudin-1 (1:200, Zymed, USA), chicken anti-vimentin (1:2000, Chemicon, France), and rat anti-MECA-32 (1:200, gift from Pr Britta Engelhardt, Switzerland). Additional details appear in the Supplemental Experimental Procedures and Fig. S4.
For pSTAT3 immunolabeling and analysis, mice were injected i.p. with vehicle (5 mM sodium citrate buffer, n = 4 per group), leptin (PeproTech, France) or LAN (Protein Laboratories Rehovot Ltd, Israel) and perfused 45 min later with a 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were processed for pSTAT3 immunolabeling and quantification as described previously (Bouret et al., 2012). Additional details appear in the Supplemental Experimental Procedures.
Microdialysis of the hypothalamus
Placement of hypothalamic cannulae and the assessment of ARH and VMH glucose levels in male rats (n = 9) was performed as described previously (Dunn-Meynell et al., 2009) and detailed in the Supplemental Experimental Procedures.
Immunoblotting
Frozen microdissected ME and MBH of mice fed ad libitum (n = 3) and those fasting for 24h (n = 3) were immunoblotted as described in the Supplemental Experimental Procedures. Rabbit anti-claudin-1 (1:1000, Zymed, USA), mouse anti-VEGF (1:500, SantaCruz), rat anti-MECA-32 (1:500, Santa Cruz) and goat anti-actin antibodies (1:500, Santa Cruz, France) were used in these experiments.
Statistical analysis
All values are expressed as means ± SEM. Data were analyzed for statistical significance with SigmaPlot software (Version 11.0), using one-way or two-way ANOVA followed by a Tukey post hoc test when appropriate. P-values of less than 0.05 were considered to be statistically significant.
Supplementary Material
01
01
Acknowledgments
This research was supported by the NEUROBESE International Associated Laboratory (Inserm, SABAN, University of Lille 2; to V.P. and S.G.B.), the Agence National pour la Recherche (ANR, France) grants ANR-05-JCJC (NT_NV_18 to V.P.), ANR-09-BLAN-0267 (to V.P. and S.L.), and ANR 11 BSV1 02102 (to S.G.B. and S.L.) and the Fondation pour la Recherche Médicale (Equipe FRM 2005, France to V.P.; Régulation Métabolique to S.G.B.; postdoctoral fellowship to A.M.), the Institut Fédératif de Recherche 114 (IFR114, France; imaging, electron microscopy cores), the National Institute of Diabetes, Digestive and Kidney Diseases and the Veterans Administration (BEL, AAD-M), the National Institute of Health (Grant DK84142, to S.G.B.), the EUFP7 Integrated Project (Grant agreement n°266408, Full4Health, to S.G.B.). F.L. was supported by a doctoral fellowship from the Ministère délégué à la Recherche et aux Nouvelles Technologies. We thank Drs. Britta Engelhardt and Philippe Ciofi for their generous gift of antibodies to MECA-32 and PV-1, respectively, Dr. S. Rasika for the editing of our manuscript, and Delphine Taillieu, Julien Devassine, Delphine Cappe (animal facility, IFR 114), Nathalie Jouy (cell sorting facility, IFR114) and Dr. Emilie Caron (metabolomic facility, IFR114) for expert technical assistance.
Summary
The delivery of blood-borne molecules conveying metabolic information to neural networks that regulate energy homeostasis is restricted by brain barriers. The fenestrated endothelium of median eminence microvessels and tight junctions between tanycytes together compose one of these. Here, we show that the decrease in blood glucose levels during fasting alters the structural organization of this blood-hypothalamus barrier, resulting in the improved access of metabolic substrates to the arcuate nucleus. These changes are mimicked by 2-deoxyglucose-induced glucoprivation and reversed by raising blood glucose levels after fasting. Furthermore, we show that VEGF-A expression in tanycytes modulates these barrier properties. The neutralization of VEGF signaling blocks fasting-induced barrier remodeling and significantly impairs the physiological response to refeeding. These results implicate glucose in the control of blood-hypothalamus exchanges through a VEGF-dependent mechanism, and demonstrate a hitherto unappreciated role for tanycytes and the permeable microvessels associated with them in the adaptive metabolic response to fasting.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Supplemental information
The supplemental information includes four figures, Supplemental Experimental Procedures, and Supplemental References.