It is now clearly demonstrated that fatty acids (FAs) may modulate neural control of energy homeostasis and specifically affect both insulin secretion and action. Indeed, pancreatic β-cells receive rich neural innervation and FAs induce important changes in autonomic nervous activity. We previously reported that chronic infusion of lipids decreased sympathetic nervous system activity and led to exaggerated glucose-induced insulin secretion (GIIS), as would be expected from the known inhibitory effect of sympathetic splanchnic nerve activity on insulin secretion. Intracarotid infusion of lipids that do not change plasma FA concentrations also lead to increased GIIS. This effect of FAs on GIIS was prevented by inhibition of β-oxidation. It is noteworthy that a single intracarotid injection of oleic acid also induced a transient increase in plasma insulin without any change in plasma glucose, suggesting that FAs per se can regulate neural control of insulin secretion. Finally, using whole cell current clamp recordings in hypothalamic slices and calcium imaging in dissociated hypothalamic neurons, we identified a hypothalamic subpopulation of neurons either excited (13%) or inhibited (6%) by FAs. Thus, FAs per se or their metabolites modulate neuronal activity, as a means of directly monitoring ongoing fuel availability by central nervous system nutrient-sensing neurons involved in the regulation of insulin secretion.
FATTY ACIDS AND NERVOUS CONTROL OF ENERGY HOMEOSTASIS
It is now well established that fatty acids (FAs) act on the central nervous system as important physiological regulators of glucose metabolism and overall energy homeostasis (1,2). For example, central administration of the long-chain FA oleate inhibits food intake and glucose production in rats (3,4). Such effects are partly related to a decreased expression of hypothalamic neuropeptide Y and of hepatic glucose-6-phosphatase after short-term overfeeding (4). This suggests that the slight increase in plasma FA concentrations in the postprandial state (5) might be detected by the central nervous system as a satiety signal. Beside these physiological effects, we previously showed that central triglyceride infusions into the brain through the carotid artery in rats over 24 h initiated hepatic insulin resistance independently of any change in plasma FA concentration (6). In this model, glucose-induced insulin secretion (GIIS) was markedly increased in lipid-infused rats compared with controls (6). Those toxic effects of lipids were prevented when β-oxidation was inhibited (6). Finally, central inhibition of lipid oxidation is sufficient to restore the hypothalamic levels of long-chain fatty acyl-CoA and to normalize food intake and glucose homeostasis in overfed rats (7). Such studies suggest an important link between the central actions of FA on both food intake and glucose metabolism under both physiological but also pathological conditions.
FAs AND NERVOUS CONTROL OF INSULIN SECRETION
FAs directly act to regulate insulin exocytosis from the pancreatic β-cell (8). In vitro, long-term exposure (48 h and more) to FAs results in a decreased insulin response to glucose in the pancreatic β-cells (rev. in 9). However, when studied in vivo in healthy subjects, the precise effects of FAs on pancreatic β-cells remain controversial. On the basis of experiments with lipid infusion, the insulin secretory response to glucose was reported to be increased (10), unchanged (11), or decreased (12). Different mechanisms might explain these discrepancies. The insulinotropic potency of FAs is influenced by chain length and degree of saturation (13). Moreover, the duration of lipid infusion also seems crucial, since opposite effects on insulin secretions are obtained by short-term (6 h) versus long-term (24 h) triglyceride infusions (12). It is also critical to carefully control the way in which glucose is administered (intravenously versus intraperitoneally) as a challenge of pancreatic β-cell function (10–12). Discrepancies between in vivo and in vitro studies concerning the effect of a long-term increase in FA concentrations on pancreatic β-cell function suggest the involvement of extrapancreatic factors. Among these factors, those of neural origin may be crucial. On the one hand, the endocrine pancreas is richly innervated by both parasympathetic and sympathetic inputs. Parasympathetic activation stimulates, whereas sympathetic activation inhibits, insulin secretion (14,15). On the other hand, several experiments clearly showed that lipids may alter autonomic nervous system activity in humans (16,17). More precisely, clinical and epidemiological studies showed that reduced autonomic activity, especially sympathetic, might be involved in the onset of obesity (18,19). Particularly in Pima Indians, a low sympathetic nervous system activity is also described as a predictor of weight gain (20). There is an important link between the development of obesity in rodents fed on diets of moderate to high FA content, pancreatic sympathetic function, and the development of abnormal insulin sensitivity. When outbred rats are fed such diets chronically, ∼50% become obese. Those rats do display reduced pancreatic sympathetic activity that precedes the development of obesity and that persists in association with hyperinsulinemia after obesity develops (21,22). These data suggest a critical link between impaired pancreatic sympathetic function, dietary FA intake, and the genetic propensity to develop diet-induced obesity.
To further study this potential relationship, we developed an in vivo model based on a 48-h intravenous infusion of triglyceride emulsion in rats (23). Briefly, a catheter was implanted under ketamine anesthesia (125 mg/kg, intraperitoneally; Imalgène, Mérieux, Lyon, France) in the right atrium of a 250-g male Wistar rat via the jugular vein. A technique previously described (24) for a 48-h infusion in unrestrained rats was used for triglyceride or saline infusion. The infusion period started on day 2 after surgery. Rats were randomly divided into two groups. In the first group, rats were infused with a mixture of a 20% triglyceride emulsion (2,000 kcal/l Intralipid KabiVitrum and 20 units/ml heparin; Intralipid KabiVitrum, Stockholm, Sweden) at a rate of 20 μl/min. Control rats were infused with saline and heparin. At the end of infusion, we measured in vivo GIIS in response to a single intravenous injection of glucose after 5 h of food deprivation. In parallel, sympathetic firing-rate activities were recorded at the level of the superior cervical ganglion (Fig. 1). Such infusions approximately double plasma FA concentration in association with decreased sympathetic nerve activity (Fig. 1) and a concomitant increase in GIIS compared with control rats. These findings are in accord with the known inhibitory effect of sympathetic activity on insulin secretion. To demonstrate the potential role of the sympathetic activity in the control of insulin secretion after lipid infusion, GIIS was also studied in the presence of oxymetazoline, an α2A-adrenoceptor (the main β-cell inhibitory adrenoceptor) agonist, to mimic an activated sympathetic nervous system. As expected, oxymetazoline produced a dose-related blunting of GIIS in lipid-infused rats, whereas controls were not sensitive to low concentrations of oxymetazoline. At a dose of 0.3 pmol/kg, GIIS became similar in both groups, suggesting that decreased sympathetic nervous system activity was partly responsible for pancreatic β-cell hyperresponsiveness to glucose (23). Interestingly, there was also a reduced Bmax and Kd for α2-adrenoceptor binding in the pancreatic β-cell of lipid-infused rats (25). Finally, neither β-adrenoceptor agonists nor antagonists affected GIIS in either lipid-infused or control rats. This result suggests that the insulin stimulating β-adrenergic pathway was not involved in the β-cell hyperresponsiveness to glucose in our model. This is in keeping with the 100-fold lower number of β-adrenoceptors compared with α-adrenoceptors on the β-cell surface (26,27) and is in agreement with the fact that α-adrenergic control of insulin secretion predominates over the β-adrenergic one in vivo (28). We have also demonstrated that these findings in rats are relevant to humans, where 48-h intravenous lipid infusion in healthy subjects also leads to glucose-induced insulin hypersecretion in association with a decrease in plasma norepinephrine concentration and urinary excretion, as an indirect measurement of sympathetic nervous system activity (29).
To assess whether these effects of FAs may be due to a direct action in the central nervous system, 250-g male Wistar rats were infused with lipids toward the brain through the carotid artery for 24 h at a rate of 2 μl/min (6). This approach allowed us to raise brain FA levels without bypassing the blood-brain barrier. Additionally, to assess the role of β-oxidation in the effects of central FAs, rats were infused intracerebroventricularly with etomoxir, an inhibitor of carnitine palmitoyl transferase 1, before and during intracarotid infusion of FAs. Importantly, this route of administration produced increased GIIS without any change in plasma FA concentrations. This FA effect on GIIS was abolished by co-infusion of etomoxir, suggesting that β-oxidation is required for the FA effects on insulin secretion (6). This is similar to the demonstration that β-oxidation in the hypothalamus also appears to be an important regulator of food intake and hepatic glucose output (30). Because FAs induced hepatic insulin resistance in intracarotid lipid-infused rats, the increased GIIS could partly be related to changes in insulin sensitivity. To specifically study the effect of FAs on nervous control of insulin secretion, we measured both plasma insulin and glucose in response to an acute intracarotid injection of oleic acid (OA), which did not affect circulating lipid levels. The effect of OA was compared with saline, glucose, and both OA/glucose injection. As depicted in Fig. 2, an acute OA overload induced a transient significant increased insulinemia at time 1 and 3 min after injection, but did not affect glycemia. Glucose induced a greater increase in plasma insulin. Both OA and glucose injected together had an additive effect, thus leading to a higher insulinemia. It may be a bit surprising that blood glucose is not at all altered, although there is a rise in insulin secretion. It could suggest that insulin resistance is induced, in agreement with decreased hepatic insulin sensitivity observed in 24-h intracarotid interleukin-infused rats (6). The additive effect of both nutrients may partially lead to further dysregulation of nervous control of insulin secretion and finally contribute to the etiology of type 2 diabetes in predisposed subjects.
FA-SENSITIVE NEURONS
Whereas the in vivo effects of FA on insulin and glucose metabolism are interesting, it is imperative to demonstrate the cellular and molecular mechanisms by which FA can act to alter neural activity. Thirty years ago, Oomura et al. (31) showed that FAs modulated central nervous system activity in vivo. Experiments were performed on the rat lateral hypothalamic neurons. Indeed, the hypothalamus contains large populations of nutrient-sensitive neurons (either glucose or FA sensitive). Thus, it is now assumed that, like glucose, FAs per se or their metabolites could modulate neuronal activity as a means of directly monitoring ongoing fuel availability to allow central nervous system nutrient-sensing neurons to regulate energy homeostasis (2,32). We previously showed that intracarotid lipid infusion decreased the number of cfos-like immunoreactive (FLI) neurons (a marker of neuronal activity) in four of the five hypothalamic nuclei studied: arcuate nucleus (ARC), dorsomedial hypothalamus, ventromedial hypothalamus, and paraventricular nucleus. In contrast, the number of FLI neurons in the lateral hypothalamus was increased in lipid-infused rats (6). The decreased number of FLI neurons in ventromedial hypothalamus indicated a lower activity in this nucleus, which could likely promote a decreased sympathetic tone, since both ventromedial hypothalamus and ARC are known to activate the sympathetic nervous system via polysynaptic efferent pathways (28). On the other hand, stimulation of the lateral hypothalamus, which contributes to the control of parasympathetic activity, promotes insulin secretion in the presence of a simultaneous rise in glucose (33). Thus, a decrease in number of FLI neurons in the ventromedial hypothalamus and an increase in number of these FLI neurons in lateral hypothalamus could reflect a concomitant decrease in sympathetic activities and an increase in parasympathetic activities, which inhibit and stimulate insulin secretion, respectively.
MOLECULAR MECHANISMS INVOLVED IN FA-SENSITIVE NEURONS
Molecular mechanisms involved in FA-sensitive neurons are still being debated. However, enzymes involved in their metabolism such as fatty acid synthase have been demonstrated to be present and active in hypothalamic neurons (34). Furthermore, carnitine palmitoyl transferase 1 is also expressed in some brain areas, including the hypothalamus, which supports our finding that β-oxidation seems crucial to mediate some the effects of FA on glucose metabolism (6,30). Thus, while FA signaling in the hypothalamus is important for peripheral glucose and energy homeostasis, the mechanisms by which FAs alter neuronal activity are not clear. One possibility is via modulation of ion channels. A body of literature indicates that FAs regulate the conductance of a wide variety of ion channels, which include Cl−, GABAA (35), CIC-2 (36), potassium, K+-Ca2+ (37), ATP-sensitive K+ channel (37), and calcium channels (38). Additionally, FAs inhibit the Na+-K+ ATPase pump (39). These effects of FAs may be directly on ion channels or through the actions of metabolic intermediates. Because there is a reciprocal relationship between glucose and FA metabolism, flux through these metabolic pathways may interact to regulate neuronal activity in the hypothalamus (40). Using patch clamp recordings, we recently investigated the effects of OA on ARC neurons in hypothalamic slices from 14- to 21-day-old Sprague-Dawley (SD) rats (41). We also recorded discharge rates in ARC neurons in 8-week-old fed and fasted SD rats in vivo in response to a single injection of OA. We found that OA regulated three distinct subpopulations of ARC neurons in a glucose-dependent fashion. Patch clamp studies showed that in 2.5 mmol/l glucose, 12 of 94 (13%) ARC neurons were excited by 2 mmol/l OA (OA-excited [OAE] neurons, Fig. 3), whereas 6 of 94 (6%) were inhibited (OA-inhibited2.5 [OAI2.5] neurons). In contrast, in 0.1 mmol/l glucose, OA inhibited 6 of 20 (30%) ARC neurons (OAI0.1 neurons); none were excited. None of the OAI0.1 neurons responded to OA in 2.5 mmol/l glucose. Thus, OAI2.5 and OAI0.1 neurons are distinct. Similarly, in fed rats, 7 of 20 (35%) ARC neurons were OAE, whereas only 3 of 20 (15%) were OAI. Note that in these rats, there was an increase in insulin secretion in response to a single injection of OA (Fig. 2). In contrast, in fasted rats only, OAI neurons were observed (3 of 15 [20%]) in ARC neurons. Thus, neural control of insulin secretion by FAs probably involved more than one neuronal population and also several different molecular mechanisms. Indeed, we identified at least two types of channels relaying the effects of OA (41). Excitatory effects of OA were probably involved in the closure of Cl− channels assessed by whole-cell current clamp recording (41), whereas the inhibitory effects of OA were reversed by the ATP-sensitive K+ channel blocker tolbutamide (in 2.5 mmol/l glucose; V.H.R., R.W., unpublished data). Additionally, activation of AMP-activated kinase with 5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR) reversed the excitatory effect of OA (Fig. 4). This suggests that AMP-activated kinase could be a key molecule-mediating effect of FAS in nutrient-sensing neurons.
We have further demonstrated the ability of FA to alter neuronal activity in dissociated neurons from the ventromedial hypothalamic nucleus of 3- to 4-week-old rats using the fura-2 Ca2+ imaging technique. These neurons either increase (glucose excited) or decrease (glucose inhibited) their intracellular Ca2+ flux when extracellular glucose levels are raised from 0.5 to 2.5 mmol/l (42). As mentioned above, in vivo OA infusion in ventromedial hypothalamic nucleus resulted in a decreased in FLI neuron number (6). In preliminary studies (Fig. 5), we found that OA induced a significant decrease in intracellular Ca2+ oscillations in both glucose-excited and glucose-inhibited neurons. These data independently reinforce the ability of FAs to alter activity in hypothalamic nutrient-sensitive neurons.
To summarize, neural control of insulin secretion by FAs is crucial to finely regulate pancreatic β-cell function. As a consequence, central dysregulation of FA metabolism could be an early event leading to inappropriate GIIS and that could, at least in part, contribute to development of metabolic diseases such as type 2 diabetes in predisposed subjects.
Illustrative recording segments of sympathetic nerve activity from control (A) and lipid-infused (B) rats. The sympathetic nerve, which is close to the carotid artery, was dissected free of underlying tissues over a distance of ∼1 cm to the superior cervical ganglion. The nerve was then covered with paraffin oil to prevent dehydration and carefully placed on a pair of recording silver wire electrodes (0.6-mm diameter). Electrodes were connected to a high-impedance probe, amplified by 104 with an alternative-current amplifier (time constant 0.2 s) and filtered at low- and high-frequency cutoffs (1–80 kHz). The filtered amplified signal was routed to an oscilloscope and a pen recorder and stored on a magnetic tape for further analysis.
Illustrative recording segments of sympathetic nerve activity from control (A) and lipid-infused (B) rats. The sympathetic nerve, which is close to the carotid artery, was dissected free of underlying tissues over a distance of ∼1 cm to the superior cervical ganglion. The nerve was then covered with paraffin oil to prevent dehydration and carefully placed on a pair of recording silver wire electrodes (0.6-mm diameter). Electrodes were connected to a high-impedance probe, amplified by 104 with an alternative-current amplifier (time constant 0.2 s) and filtered at low- and high-frequency cutoffs (1–80 kHz). The filtered amplified signal was routed to an oscilloscope and a pen recorder and stored on a magnetic tape for further analysis.
![FIG. 2. Time course of plasma glucose (A) and insulin (B) levels in response to intracarotid injection of saline (○), glucose (50 mmol/kg, 30-s injection, □), OA (50 nmol/kg, •), and OA and glucose (▪) in rats implanted with intracarotid catheters pointed toward the brain (under pentobarbital [50 mg/kg i.p.; Sanofi, Libourne, France] anesthesia 5 days before testing). ***P < 0.01, different from time 0.](/view-large/figure/2330252/zdb1120646270002.jpeg)
Time course of plasma glucose (A) and insulin (B) levels in response to intracarotid injection of saline (○), glucose (50 mmol/kg, 30-s injection, □), OA (50 nmol/kg, •), and OA and glucose (▪) in rats implanted with intracarotid catheters pointed toward the brain (under pentobarbital [50 mg/kg i.p.; Sanofi, Libourne, France] anesthesia 5 days before testing). ***P < 0.01, different from time 0.
Time course of plasma glucose (A) and insulin (B) levels in response to intracarotid injection of saline (○), glucose (50 mmol/kg, 30-s injection, □), OA (50 nmol/kg, •), and OA and glucose (▪) in rats implanted with intracarotid catheters pointed toward the brain (under pentobarbital [50 mg/kg i.p.; Sanofi, Libourne, France] anesthesia 5 days before testing). ***P < 0.01, different from time 0.
Whole cell current clamp recordings in an OAE neuron in 2.5 mmol/l glucose in a hypothalamic slice from a 2- to 3-week-old rat. Resting membrane potential is –45 mV. Similarly, downward deflections represent the membrane voltage response to a constant hyperpolarizing pulse.
Whole cell current clamp recordings in an OAE neuron in 2.5 mmol/l glucose in a hypothalamic slice from a 2- to 3-week-old rat. Resting membrane potential is –45 mV. Similarly, downward deflections represent the membrane voltage response to a constant hyperpolarizing pulse.
Whole-cell current clamp recordings in an OAE neuron in 2.5 mmol/l glucose in a hypothalamic slice from a 2- to 3-week-old rat. 5-Aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR) reversed the excitatory effect of OA.
Whole-cell current clamp recordings in an OAE neuron in 2.5 mmol/l glucose in a hypothalamic slice from a 2- to 3-week-old rat. 5-Aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR) reversed the excitatory effect of OA.
![FIG. 5. Effect of OA on intracellular Ca2+ ([Ca2+]i) oscillations in a representative glucose-excited neuron (A) and a representative glucose-inhibited neuron (B). Neurons were dissociated from the ventromedial hypothalamic nucleus of 2- to 3-week-old rats. The ventromedial hypothalamic nucleus was punched out of slices made through the ventrobasal hypothalamus. Single ventromedial hypothalamic nucleus neurons were dissociated by papain digestion and trituration and were then maintained in a HEPES-buffered balanced salt solution. Fluorescent imaging measurement of [Ca2+]i was carried out using fura-2. The values for Rmin, Rmax, and constant b (the 380-nm wavelength in the Ca2+-free and Ca2+-bond states) were calculated from measurements with fura-2–free acid in solution in the microscope chamber. The dissociation constant (Kd) for the fura-2 · Ca2+ complex was taken as 224 nmol/l. Regarding the data analysis, integrated area under the curve (AUC) for changes in [Ca2+]i oscillations was calculated for every experimental condition, and AUC values were compared between 2.5 mmol/l glucose and 2.5 mmol/l glucose + OA at different concentrations for GE neurons (A). For GI neurons (B), the OA + 0.5 mmol/l glucose conditions were compared with 0.5 mmol/l glucose alone. Data were analyzed using a statistical paired t test.](/view-large/figure/2330255/zdb1120646270005.jpeg)
Effect of OA on intracellular Ca2+ ([Ca2+]i) oscillations in a representative glucose-excited neuron (A) and a representative glucose-inhibited neuron (B). Neurons were dissociated from the ventromedial hypothalamic nucleus of 2- to 3-week-old rats. The ventromedial hypothalamic nucleus was punched out of slices made through the ventrobasal hypothalamus. Single ventromedial hypothalamic nucleus neurons were dissociated by papain digestion and trituration and were then maintained in a HEPES-buffered balanced salt solution. Fluorescent imaging measurement of [Ca2+]i was carried out using fura-2. The values for Rmin, Rmax, and constant b (the 380-nm wavelength in the Ca2+-free and Ca2+-bond states) were calculated from measurements with fura-2–free acid in solution in the microscope chamber. The dissociation constant (Kd) for the fura-2 · Ca2+ complex was taken as 224 nmol/l. Regarding the data analysis, integrated area under the curve (AUC) for changes in [Ca2+]i oscillations was calculated for every experimental condition, and AUC values were compared between 2.5 mmol/l glucose and 2.5 mmol/l glucose + OA at different concentrations for GE neurons (A). For GI neurons (B), the OA + 0.5 mmol/l glucose conditions were compared with 0.5 mmol/l glucose alone. Data were analyzed using a statistical paired t test.
Effect of OA on intracellular Ca2+ ([Ca2+]i) oscillations in a representative glucose-excited neuron (A) and a representative glucose-inhibited neuron (B). Neurons were dissociated from the ventromedial hypothalamic nucleus of 2- to 3-week-old rats. The ventromedial hypothalamic nucleus was punched out of slices made through the ventrobasal hypothalamus. Single ventromedial hypothalamic nucleus neurons were dissociated by papain digestion and trituration and were then maintained in a HEPES-buffered balanced salt solution. Fluorescent imaging measurement of [Ca2+]i was carried out using fura-2. The values for Rmin, Rmax, and constant b (the 380-nm wavelength in the Ca2+-free and Ca2+-bond states) were calculated from measurements with fura-2–free acid in solution in the microscope chamber. The dissociation constant (Kd) for the fura-2 · Ca2+ complex was taken as 224 nmol/l. Regarding the data analysis, integrated area under the curve (AUC) for changes in [Ca2+]i oscillations was calculated for every experimental condition, and AUC values were compared between 2.5 mmol/l glucose and 2.5 mmol/l glucose + OA at different concentrations for GE neurons (A). For GI neurons (B), the OA + 0.5 mmol/l glucose conditions were compared with 0.5 mmol/l glucose alone. Data were analyzed using a statistical paired t test.
This article is based on a presentation at a symposium. The symposium and the publication of this article were made possible by an unrestricted educational grant from Servier.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
