Glucose, the main energy source of the CNS, enters the brain from the bloodstream crossing the BBB through specific transporters. As normal brain function depends on its glucose supply, this step is highly regulated. That is, the transport of glucose across the BBB adapts in response to cerebral energy demand in order to maintain glucose homeostasis in the brain. The facilitative glucose transporter (GLUT)-1 is largely responsible for glucose transport across the BBB. This protein is expressed in non-neuronal cells throughout the CNS, especially in astrocytes and endothelial cells of the BBB (101), as well as in tanycytes along the BCSFB (50). However, GLUT-1 in endothelial cells is highly glycosylated, having a higher molecular weight than the isoform expressed in astrocytes and other glial cells (101, 102). As indicated in a recent review, some authors suggest different functional characteristics between the two forms of GLUT-1, although there is no consensus on this subject (103).

Changes in glucose concentration are rapidly detected in the hypothalamus, which adapts to such variations and emits a response to maintain glucose homeostasis not only in the brain, but also systemically as glucose-sensing neurons in the hypothalamus send signals to the autonomous nervous system, reaching peripheral organs such as the pancreas or the liver (104–107). There is more than one mechanism for central glucose sensing and different cell types are involved in this essential task (107–110). Two populations of glucose-sensing neurons have been identified: glucose-excited and glucose-inhibited neurons (GE and GI, respectively) (111) and glial cells also participate in these important glucose-sensing mechanisms. Astrocyte endfeet express GLUT-2 which, in addition to its transport functions, participates in glucose sensing (110, 112). This GLUT is highly expressed in tanycytes along the BCSFB (109), with these specialized glial cells also participating in glucose-sensing processes. In addition to expressing GLUT-2, astrocytes and tanycytes express sodium glucose transporter (SGLT)-1, glucokinase (GCK), and K_ATP channels (110), proteins that are all known to be involved in glucose-sensing mechanisms. Indeed, the classical mechanism for glucose sensing in pancreatic β-cells requires glucose uptake through GLUT-2 in rodents or GLUT-1 in humans, GCK, and activation of ATP-sensitive K⁺ channels (112, 113). This system shares some similarities with glucose-sensing pathways in astrocytes and tanycytes.

One proposed model for glucose sensing in tanycytes involves glucose entering the cell through GLUT-2 and phosphorylation by GCK. Subsequently, glucose-6-phosphate undergoes glycolysis, producing pyruvate and, through the action of lactate dehydrogenase, lactate. Lactate is transported to the extracellular space by monocarboxylate transporter (MCT)-4 or MCT-1, and then taken up by neurons through MCT-2 (109). Depending on the kind of neuron, GE or GI, an excitatory or inhibitory signal will be produced in the hypothalamus and sent to other brain areas and the autonomic nervous system (108). Tanycytes can also respond rapidly to glucose and other inputs by producing calcium waves, a process requiring ATP release and autocrine signaling through purinergic P2Y receptors (56, 59). The precise mechanisms involved in this tanycytic response are not yet fully elucidated, but it constitutes a possible model for tanycyte–neuron interaction.

Glucose sensing in astrocytes involves a similar process. Indeed, according to the “astrocyte-neuron lactate shuttle” hypothesis proposed by Pellerin and Magistretti over two decades ago (4), lactate from glucose or glycogen metabolism released by astrocytes is not only used by neurons as an energy source but can also signal energy availability to glucose-sensing neurons. Glucose transport into astrocytes is facilitated by GLUT-2 or occurs through gap junctions in a passive manner (112, 114–116). This glucose can be metabolized or stored as glycogen. However, it is still debated as to whether astrocytes secrete only lactate or also glucose to the extracellular fluid to act on glucose-sensing neurons and to be used as fuel (112). Moreover, astrocytes and tanycytes can respond to an increase in glucose or to other signals (i.e., some neurotransmitters) by secreting endozepines, anorexigenic peptides that act on hypothalamic neurons to maintain energy homeostasis (107, 117) and that also participate in unsaturated long-chain fatty acid metabolism in astrocytes (118).

The precise mechanisms of glucose transport and sensing in the hypothalamus are yet to be fully elucidated. For example, SGLT, an active sodium co-transporter, is reported to be involved in glucose sensing in the ventromedial nucleus of the rodent hypothalamus (119), although it is not clear whether this sensing occurs in glucose responsive neurons or in astrocytes. By using genetically engineered mouse models, García-Cáceres and collaborators recently demonstrated that insulin signaling in astrocytes plays a role in the regulation of systemic glucose homeostasis. Specific ablation of the insulin receptor (IR) in astrocytes was shown to impair their uptake of glucose and the ability to correctly respond to changes in glycemia (120). Other studies suggest a role of leptin in increasing (121) or ghrelin in reducing (122) glucose uptake by astrocytes, which might also affect glucose sensing. It thus appears that the transport of glucose by astrocytes is highly regulated by diverse nutrient and hormonal signals.

Monocarboxylates are molecules with one carboxylate group; some examples with metabolic functions include not only lactate, but also pyruvate and ketones, all of which can be used by neurons as an alternative energy source in addition to acting as metabolic signals (123–126). The brain expresses MCTs-1, -2, and -4, with MCT-1 being found in endothelial and ependymal cells, as well as in astrocyte endfeet at the BBB (127, 128). MCT-2 is expressed in endothelial cells, but not in astrocytes, whereas MCT-4 appears to be specific for astrocytes (58, 129–131). Ketone bodies and other monocarboxylates from the bloodstream cross the BBB through specific MCTs present in both the luminal and abluminal sides of the endothelial cells (132, 133). Astrocyte endfeet not only takes up monocarboxylates through MCT-1 (132–134), but these glial cells are also able to synthesize ketone bodies from fatty acid β-oxidation and secrete them as an energy source for neurons and other glial cells (Figure 1). Tanycytes have also been suggested to transport lactate through MCT-2 in a photoperiodic model of Siberian hamster (58). These authors found that MCT-2 and the glutamate transporter GLAST were decreased during a short photoperiod, which could indicate a change in seasonal neurotransporter supply. In the rat brain, tanycytes were shown to express functional MCT-1 and MCT-4 in an anatomically specific manner (135), suggesting that these glial cells may also participate in lactate transport to neurons.

Regulation of the transport and production of ketone bodies in the brain is important in metabolic control as hypothalamic sensing of these monocarboxylates also participates in the regulation of food intake (126, 136). Indeed, after the initial HFD-induced hyperphagia, there is a reduction in food intake that is reported to be mediated, at least in part, by ketone body signaling to hypothalamic neurons. These ketone bodies are synthesized by hypothalamic astrocytes as products of fatty acid metabolism (136, 137).

Lipid sensing in the hypothalamus is necessary for the correct regulation of energy balance (138). There are lipid sensing neurons that are excited or inhibited by fatty acids, depending on the type of neuron and also the metabolic state, i.e., fasting versus overfeeding (139). Although the role of glial cells in this process is not fully understood, astrocytes are the primary lipid metabolizers in the CNS. They also express proteins related to lipid sensing, such as transporter CD36 and peroxisome proliferator-activated receptor gamma, an important lipid-activated nuclear receptor that regulates transcription of numerous genes, including some involved in lipid metabolism (140, 141). In addition, astrocytic production of ketone bodies from fatty acids and their release to neurons could be one way by which an excess of fatty acids is signaled to metabolic neuronal circuits. Recent evidence suggests an increase in fatty acid β-oxidation in hypothalamic astrocytes from obese mice fed a HFD, as well as a role for tanycytes in restricting the passage of saturated fatty acids into the hypothalamus (142).

Although the brain produces lipids, it also has mechanisms to transport them from the bloodstream, but how they go through the BBB is not yet fully understood. Short and medium chain fatty acids appear to enter the CNS by simple diffusion through the plasma membrane (143). In contrast, long chain fatty acids (>12 carbons) need transporters to cross the BBB (144), with several fatty acid transport proteins (FATP) and fatty acid binding proteins (FABP) having been identified (145). In vitro studies indicate that FATP-1, FATP-4, and FABP-5 are the major isoforms expressed in microvessel endothelial cells and the gray matter of the human brain (145, 146). When the fatty acid translocase (FAT) CD36 is knocked-out in mice (CD36−/−), the uptake of monounsaturated fatty acids is significantly decreased, with no effect on polyunsaturated fatty acid uptake (147). In the CNS, CD36 is expressed in endothelial cells, microglial cells, astrocytes, and in ventromedial hypothalamic neurons (148–151). Although it is not the most highly expressed FATP, studies indicate that CD36 is responsible for fatty acid sensing in the hypothalamus and is thus important for the control of energy homeostasis (136, 137, 152).

In addition to the passage of free fatty acids through the BBB, lipids can also enter or exit the CNS as lipoproteins. This process is mediated by apolipoprotein E (ApoE) interacting with lipoprotein receptors (153). In the CNS, ApoE is expressed in astrocytes and tanycytes and its levels are upregulated by both leptin and TH (154, 155), with this process being involved in the regulation of food intake and energy balance (156).

Leptin is an anorexigenic hormone that exerts part of its effects by inhibiting orexigenic neurons and activating anorexigenic neurons in the hypothalamus (157–159). It also has a role in the regulation of systemic lipid and glucose metabolism (160, 161). The leptin (or obesity) receptor (ObR), which has six isoforms, is highly expressed in brain endothelial cells, astrocytes and tanycytes (162–165), and endothelial and astroglial cells have been studied in attempt to unravel the mechanisms of leptin transport into the brain (163, 166). However, González-Carter and colleagues have recently reported that, in a human in vitro BBB model, leptin–ObR interaction is not necessary for the transport of this hormone across the BBB. They propose that lipoprotein receptor-related protein-2, expressed in endothelial cells at the BBB, is responsible for the passage of leptin across the BBB (167). Increasing evidence points to the BCSFB as the main pathway for entry of leptin into, at least, the hypothalamus (165, 168) (Figure 2).

The median eminence, a circumventricular organ close to the third ventricle, is the first site in the brain reached by blood-borne leptin (165). After an intraperitoneal leptin injection, there is a 1–2 h lag between the activation of leptin signaling pathways in the ventral and dorsal nuclei of the hypothalamus. This time-lag disappears when leptin is administered intracerebroventricularly, instead of intraperitoneally (169), suggesting that leptin transport from blood to the CSF is an important step in the action of this hormone in the brain and that it involves a delay in circulating changes reaching central target sites. Moreover, this process appears to be a finely regulated step in the control of energy balance as tanycytes act as “gatekeepers” for the passage of leptin into the mediobasal hypothalamus. Evidence suggests that leptin is first taken up by tanycyte processes in contact with the fenestrated capillaries at the median eminence (165) and that this uptake requires the activation of ObRb and the internalization of leptin by clathrin-coated vesicles (165). According to research carried out by Vincent Prevot and his team, this process involves signal transducer and activator of transcription (STAT)-3, protein kinase B (PKB)/Akt, and extracellular signal regulated kinase (ERK) phosphorylation, but is janus kinase-2 independent (165). Leptin is then transported toward the tanycyte cell body and, finally, released to the CSF and hypothalamic parenchyma (Figure 2) employing an ERK-dependent pathway (165). By using STAT-3 phosphorylation as an indicator of leptin signaling (170–172), Balland and collaborators reported that neutralization of leptin in the CSF impairs leptin signaling in mediobasal hypothalamic neurons (165), supporting the idea of the blood–CSF–hypothalamic gateway for leptin entry into the brain.

Taking into account the above mentioned studies, it appears that both endothelial cells and tanycytes contribute to the transport of leptin through the BBB and between different brain regions (163, 166, 167). In contrast, there is no clear evidence of the involvement of astrocytes in leptin transport, but a number of studies demonstrate that leptin signaling in astrocytes is important for energy homeostasis (173, 174).

Leptin transport into the brain is modulated by conditions including obesity and fasting, as well as metabolic factors. Obesity associated to HFD intake is reported to induce central leptin resistance. There are two main mechanisms or levels of leptin resistance suggested to occur: impairment of leptin transport into the brain (165) and reduction in the central response to leptin (175). Mice exposed long term to a HFD develop leptin resistance only when high levels of plasma leptin are reached (176). This suggests that hyperleptinemia is at least one of the causes of diet-induced leptin resistance. In addition, hypothalamic inflammation associated with diet-induced obesity could contribute to leptin resistance by altering the cellular networks and molecular pathways that control energy homeostasis (177). Nevertheless, recent studies suggest that leptin resistance does not imply a loss of responsiveness to endogenous leptin, but rather that there is a threshold above which exogenous leptin barely increases the response to leptin (178, 179). Glucose and insulin are reported to increase the transport of leptin across the BBB (180), while an increase in circulating triglycerides could impair leptin transport across the BBB (181). The latter suggests a possible mechanism for the reported reduction in leptin transport into the brain during fasting (182).

Ghrelin is an orexigenic hormone produced and secreted in the stomach (183). It has similar targets as leptin in the CNS and also plays an important, but opposite, role in energy balance (184). There are two forms of ghrelin, acylated and unacylated, depending on the post-translational acylation with octanoic or decanoic acid (183, 185). This modification occurs mainly in the stomach, but there is evidence that it can also take place in target tissues (186). The acylated form of ghrelin exerts the majority of the metabolic effects of this hormone in the CNS and it binds more efficiently to the ghrelin receptor than the unacylated form (187). This receptor, also called the growth hormone secretagogue receptor 1a, is widely expressed in the hypothalamus (188). The mechanism underlying the passage of ghrelin across the BBB is not yet fully understood, but recent studies indicate that ghrelin possibly uses a similar route as leptin into the brain (189), i.e., through tanycytes in contact with the median eminence (Figure 2). Other studies indicate that this process is carried out by saturable transporters, at least for the acylated form, whereas transport of des-acyl ghrelin is not saturable (190). Entry of acylated ghrelin into the CNS is increased by serum triglycerides and fasting and is decreased in obese mice (191), in contrast with leptin transport. Diet-induced obesity is reported to impair the hypothalamic response to peripherally or centrally administered ghrelin (192). The role of unacylated ghrelin on metabolism is largely unknown, but an increasing number of studies reveal that des-acyl ghrelin has similar and opposite functions as the acylated form (193–196).

Insulin is a pancreatic hormone directly involved in glucose metabolism and homeostasis. Within the brain, it acts to increase energy expenditure and reduce food intake and energy storage (197). Insulin binds to its receptor in the plasma membranes of endothelial cells at the BBB and is internalized following a saturable pathway (198, 199). Recent studies have shown that IRs in astrocytes are involved in the entry of this hormone into the CNS (120). Also, as mentioned above, insulin signaling in astrocytes is necessary for the regulation of systemic glucose levels (120). Insulin transport into the brain is enhanced by satiation hormones like cholecystokinin (200). Although estradiol is known to impair insulin’s actions in the brain, its effects appear to be unrelated to insulin transport (201). Some studies show that leptin increases insulin sensitivity in the hypothalamus at the molecular level (202), while others have found that leptin impairs insulin signaling in the brain (203). This discrepancy could be a matter of the experimental model employed, but further research is needed to understand the relationship between the effects of leptin and insulin at the level of the CNS. Leptin shares some signaling pathways with insulin, but the effects of these two hormones are not entirely parallel, as they exert opposite effects in some hypothalamic neurons (204). Saturated fatty acids induce insulin resistance in the hypothalamus (205), as has been previously described in peripheral tissues (206).

As hydrophobic molecules, estrogens, androgens, and progesterone can enter the brain by simple diffusion (207). Moreover, steroids are synthesized in the brain (208). These neurosteroids are produced in the CNS either from brain-borne cholesterol or from peripherally synthesized steroid precursors, like pregnenolone, deoxycorticosterone, and testosterone (209). The enzymes necessary for this synthesis are found in non-neuronal cells, including astrocytes, tanycytes, ependymal cells, and oligodendrocytes (210, 211), and in some neurons (212). As steroid hormones are known to regulate neurosteroid metabolism in glial cells (213–217) and also the expression levels of steroid receptors in the hypothalamus (218, 219), neurosteroids could have paracrine/autocrine functions within the brain.

Steroids and neurosteroids exert neuroprotective effects in the brain following brain injury, neurological disease, or inflammation (220–227). The expression of aromatase, the enzyme that catalyzes the conversion of testosterone into estradiol, is stimulated in reactive astrocytes after brain injury as a neuroprotective measure (228–230). Both microglial cells and astrocytes play an important role in the neuroprotective functions of steroids (231), as sex steroids diminish microglia reactivity (232–234) and astrocyte production of proinflammatory molecules (235–238).

Sex steroids, but specially estrogens, modulate energy homeostasis at the hypothalamic level decreasing food intake (239–241), increasing energy expenditure (242), and modulating the sensitivity to other metabolic hormones (243, 244). Their effect differs depending on the neuronal population (245, 246), but with an overall anorectic effect (247–249), although the underlying mechanisms are not yet fully understood. While nuclear estrogen receptors (ERs) are involved, especially ER α (247, 250–252), evidence indicates that estrogen responsive G-coupled membrane receptors can also regulate these effects (253, 254). The apparently contradictory results in the literature regarding the mechanism of action of estrogens on metabolism indicate a complex system for estrogens’ function in energy homeostasis, where the different ERs could be acting in combination (255). Moreover, the mechanisms of action used by estrogens in metabolic control could be sexually dimorphic (256). The involvement of neurosteroids in energy homeostasis remains unknown.

The role of TH in increasing the metabolic rate has been known for more than a century (257). The involvement of these hormones in the control of energy homeostasis at the central level is a more recent discovery (258, 259). They promote lipogenesis at the level of the hypothalamus, which eventually leads to brown adipose tissue thermogenesis (259) and blockage of TH signaling in the hypothalamus reverts this process, leading to weight gain without an increase in feeding (259). Clinical studies and animal models with a pathological excess of TH synthesis and secretion (hyperthyroidism) have shed light on TH action in the hypothalamus and control of energy balance (260). Most hyperthyroid patients have an increased appetite and food intake and decreased body weight (261). Moreover, these same symptoms that are observed in animal models of hyperthyroidism are associated with the upregulation of orexigenic neuropeptides AgRP and NPY and downregulation of anorexigenic neuropeptides derived from POMC in the arcuate nucleus (259). There is evidence that TH are involved in brain inflammation, promoting survival, and processes growth in microglial cells and also in astrocytes (262–264). TH are also involved in systemic glucose homeostasis and insulin sensing (265, 266).

The thyroid gland produces and secretes mainly tetraiodo-l-thyronine or thyroxine (T4), which is generally transformed into triiodo-l-thyronine (T3) through deiodination at the level of peripheral tissues (267). Thus, deiodinase enzyme expression in peripheral tissues is important for the control of TH actions (268), as they catalyze the transformation of T4 into T3 and of both hormones into reverse T3 (rT3) and 3,5-diiodo-l-thyronine (T2), respectively (269). These two last forms are usually considered inactive, although in the last few years new roles have been proposed for them and other non-classical TH (270).

Thyroid hormones enter the hypothalamus mainly through MCT-8 (271) and organic anion transporting polypeptide-1C1 (272) in rodents (273). These transporters are expressed in endothelial cells of the BBB and epithelial cells of the choroid plexus (274), besides neurons and glial cells of the hypothalamus (275–277). Tanycytes act as gatekeepers for TH at the BBB (61) (Figure 2). These cells express the enzyme deiodinase II (DII) (278–280), catalyzing the formation of hormone T3 from the prohormone T4. Tanycytes uptake T4 from the capillaries and release T3 to the extracellular space in the hypothalamus, where this hormone can exert its central actions (258, 259, 281) (Figure 2). Modulation of deiodinase expression is a key point in TH homeostasis. For example, DII expression in tanycytes is promoted in fasting conditions (282). DII-expressing tanycytes are in direct contact with AgRP/NPY-expressing neurons of the arcuate nucleus. Upregulation of DII results in an increased production of T3, which activates AgRP/NPY neurons and, therefore, feeding behavior (283, 284). Tanycytes also express deiodinase III (DIII), which deiodinates T4 into reverse T3 which is biologically inactive, and T3 into T2. TH is important in the adaptation to different photoperiods in seasonal animals were, for example, there is a decrease in food intake and body weight during short photoperiods. The study of hypothalamic metabolism of TH during photoperiodic changes in seasonal mammals has shown that the there is an upregulation of DII during periods of long days, which would increase the levels of T3. In Siberian hamsters an upregulation of DIII in tanycytes has been shown to occur during short photoperiods, lowering active T3 levels (285). The retinoic acid pathway in tanycytes appears to be similarly regulated by photoperiodicity and also leads to modifications in energy expenditure (286, 287).

Thyroid hormone signaling usually occurs through nuclear thyroid receptors α and β (288) that function as transcription factors modulating gene expression (289). TH can also exert rapid non-genomic actions through membrane-associated receptors (290, 291). This signaling pathway could mediate TH effects on vasodilatation (292) and has been shown to be involved in neuronal excitability in the hippocampus (293, 294).

Centrally, THs control their own homeostasis in various ways, with non-neuronal cells having an important role, i.e., regulation of deiodinase expression (278) and inactivation of thyroid releasing hormone (295). Other hormones involved in metabolic control can enhance the secretion, synthesis, or sensing of TH, including leptin (296–298) and sex steroids (299–301).

Perivascular astrocytes take-up blood-borne glucose that then undergoes glycolysis or glycogenesis (112). Lactate produced from glucose or glycogen metabolism in these cells is released to the extracellular space and enters neurons to be used as energy, constituting their primary energy source as suggested by some studies (302, 303). However, the question about the identity of the main energy source for neurons—lactate or glucose—is still debated. Tanycytes can metabolize and sense glucose in a similar manner (109).

Glucose storage as glycogen in astrocytes provides a way to guarantee energy release to neurons when it is needed, i.e., when faced with a raise in neuronal activity (304), by production of lactate from glycogenolysis. Several factors can regulate glycogen production and utilization in astrocytes, with insulin, insulin-like growth factor (IGF)-1 (305, 306), and leptin (203, 307) increasing their production of glycogen. More recently, ghrelin has been reported to possibly promote glycogenolysis in hypothalamic neurons (122).

It has been suggested that some fatty acids, like erucic acid (308, 309), suffer metabolic changes as they cross the BBB, whereas others do not (310, 311). Studies indicate that lipoproteins are hydrolyzed as they cross the BBB by the enzyme lipoprotein lipase associated to the membrane of endothelial cells (312–316).

In the absence of glucose and when glycogen stores are exhausted, such as in fasting conditions, astrocytes increase their uptake and utilization of fatty acids (136, 317, 318), which enter the mitochondria through carnitine palmitoyltransferase-1 to undergo β-oxidation (319). In the mitochondria, the enzymes 3-hydroxy-3-methylglutaryl-CoA synthase and lyase (320–322) transform fatty acids into β-hydroxybutyrate, a ketone body (323). Ketone bodies produced from this metabolic pathway are used by astrocytes themselves for fuel or secreted to be used by neurons and other glial cells (318).

Microglia, the innate immune cells of the CNS, are the first line of defense in response to foreign substances (417, 418) and are activated in response to saturated fat consumption (36, 403, 408, 419). Indeed, these glial cells are suggested to dictate the inflammation that occurs in response to saturated fats (419). Microglia can also be activated when weight gain is due to excess intake of a normal diet and due to high fat intake, (402), indicating that not only dietary signals are involved. Leptin stimulates the release of inflammatory cytokines from microglia (420), suggesting that hyperleptinemia could be involved in microglial activation in obese subjects.

Astrocytes also respond to HFD intake (36, 384, 421) and can be directly activated in vitro by fatty acids (408, 409, 422). Hyperleptinemia associated with weight gain may also participate in the activation of glia in situation of obesity (121, 173, 388, 408, 423). Indeed, ob/ob mice, which are dramatically obese due to the genetic lack of leptin, do not exhibit astrogliosis and leptin-induced weight loss actually increases astrocytic profiles in the hypothalamus of these animals (408). However, we have found that in some situations of increased weight gain, such as increased carbohydrate intake in the form of sucrose, astrocytic markers may actually be decreased (407).

The astrogliosis response to HFD differs between males and females, as does the in vitro response to PA (409). The protective effects of estrogens are exerted through ERα in astrocytes (424), with estrogens protecting against PA activation of astrocytes in vitro (409). Morselli et al demonstrated that HFD-intake reduces hypothalamic ERα levels in males, but not in females, which may be involved in the decreased protection against diet-induced obesity in males.

Astrocytes have also been implicated in determining the preference for a HFD, with this mechanism involving cannabinoid receptor 1 (CB1) (425, 426). The intake of a HFD induces the preference for this type of diet and this appears to involve the production of endocannabinoids in the hypothalamus (426). Leptin signaling in astrocytes is regulated by CB1, with disruption of CB1 in these glial cells resulting in the inability of leptin to regulate glycogen storage (307) and thus possibly affecting central energy storage and glucose sensing/signaling.