Ceramide accumulation in ventromedial hypothalamus during high fat feeding impacts body weight control and impairs glucose homeostasis

Sphingolipids are a broad class of lipids derived from a decarboxylation reaction of serine and palmitoyl-CoA and the addition of a second variable-length fatty acid chain. These bioactive lipids have a role in a multitude of biological processes, and in the past few decades have become recognized to have a vital role in body weight control and insulin sensitivity. To date, a thorough examination of sphingolipid metabolism in response to high fat feeding has not been reported in the literature. Although it is well-characterized that ceramides and other sphingolipids play a role in body weight control and insulin sensitivity, little is known of the kinetics and tissue distribution of these effects in the evolution of obesity. Indeed, numerous reports have measured ceramide content in a tissue of interest after sundry manipulations such as inflammatory insults, high fat diet, and genetic obesity. However, none have systematically interrogated changes in the expression of enzymes of sphingolipid metabolism in response to metabolic stress. Importantly, ceramides accumulate in the hypothalamus during obesity and following intravenous saturated fat infusions. The ventromedial hypothalamus (VMH), specifically, has been recognized for its role in these actions for multiple decades. Indeed, perturbations of the VMH have been shown to alter the response to hypoglycemia, muscle glucose homeostasis, and body weight regulation. As hypothalamic lipid metabolism has become recognized as an important player in whole-body metabolism, the logical hypothesis emerges that ceramides play a role in hypothalamic control of fat and glucose metabolism during the evolution of obesity.

ABSTRACT Sphingolipids are a broad class of lipids derived from a decarboxylation reaction of serine and palmitoyl-CoA and the addition of a second variable-length fatty acid chain. These bioactive lipids have a role in a multitude of biological processes, and in the past few decades have become recognized to have a vital role in body weight control and insulin sensitivity. To date, a thorough examination of sphingolipid metabolism in response to high fat feeding has not been reported in the literature. Although it is well-characterized that ceramides and other sphingolipids play a role in body weight control and insulin sensitivity, little is known of the kinetics and tissue distribution of these effects in the evolution of obesity. Indeed, numerous reports have measured ceramide content in a tissue of interest after sundry manipulations such as inflammatory insults, high fat diet, and genetic obesity. However, none have systematically interrogated changes in the expression of enzymes of sphingolipid metabolism in response to metabolic stress. Importantly, ceramides accumulate in the hypothalamus during obesity and following intravenous saturated fat infusions. The ventromedial hypothalamus (VMH), specifically, has been recognized for its role in these actions for multiple decades. Indeed, perturbations of the VMH have been shown to alter the response to hypoglycemia, muscle glucose homeostasis, and body weight regulation. As hypothalamic lipid metabolism has become recognized as an important player in whole-body metabolism, the logical hypothesis emerges that ceramides play a role in hypothalamic control of fat and glucose metabolism during the evolution of obesity. ii TABLE OF CONTENTS ABSTRACT ii INTRODUCTION 1 METHODS 5 RESULTS 9 DISCUSSION 25 CONCLUSION 29 REFERENCES 31 iii 1 INTRODUCTION Despite its broad recognition as a public health menace, obesity rates in the USA remain at record high levels (Flegal, 2010), carrying with it an increased risk for an array of metabolic diseases (Adams 2006). This public health crisis points to the need for novel treatments of both obesity and obesity-related illnesses, such as Type 2 diabetes; unfortunately, success in developing such treatments has lacked (Rodgers 2012). According to research published in Current Opinion in Pharmacology, one attractive target of recent interest has been the hypothalamus, which is vital to the control of both body weight and glucose homeostasis. Multiple populations of neurons reside in the hypothalamus, with diverse roles in regulating body weight control and glucose metabolism. The ventromedial hypothalamus (VMH), specifically host to SF1 neurons, has been recognized for its role in these actions in recent times (Choi, 2013). Indeed, manipulations of the VMH have been shown to alter the response to hypoglycemia and body weight regulation (Klöckener, 2011). Sphingolipids are created by a decarboxylation reaction of palmitoyl-CoA with serine, and the subsequent addition of a second, variable-length fatty acid chain, ultimately resulting in the production of ceramide. As outlined in his paper, Ceramides in Insulin Resistance and Lipotoxicity, Dr. Scott Summers contends that “a great deal of work has been done in recent years to characterize the effects of sphingolipid accumulation on insulin sensitivity and lipotoxic organ failure in any number of scenarios, with the general finding that accumulation of sphingolipids such as ceramide and glucosylceramide promote insulin resistance.” Sphingolipids are a ubiquitous lipid species found in cell and organelle membranes. The Ceramide is then able to be 2 derivatized in a number of enzymatically-driven reactions, yielding ceramide-1phosphate, sphingosine, glucosylceramide, or sphingomyelin. Although sphingolipids have been thoroughly studied in the context of type 2 diabetes mellitus and lipotoxic organ damage, comparatively less focus has been devoted to the study of sphingolipids and body weight regulation and overall glucose storage and gluconeogenesis. These bioactive lipids have a role in a multitude of biological processes, and in the past few years have become recognized to have a vital role in body weight control and insulin sensitivity. Notably, work published in the Journal of Clinical Investigation by Dr. William L. Holland proved that ceramides accumulate in the hypothalamus during obesity and following intravenous fat infusions. As hypothalamic lipid metabolism has become recognized as an important player in whole-body metabolism, the logical hypothesis emerges that ceramides play a role in hypothalamic control of metabolism during the evolution of obesity. As the hypothalamus is another vital player in the control of body weight, the literature suggests that paradigms which modulate ceramide metabolism exert their effects on body weight through altering brain ceramide metabolism. Generally, hypothalamic lipid metabolism is becoming recognized as a participant in body weight regulation, and recent evidence suggests that sphingolipids specifically play a key role in altering hypothalamic control of body weight. Importantly, multiple reports demonstrate that ceramide content is increased in the hypothalamus of mice in obesity or in response to lipid infusions, suggesting that hypothalamic weight control may be impaired by ceramide accumulation during the 3 evolution of obesity. To understand the importance of ceramides in central body weight regulation, investigators have successfully modulated hypothalamic ceramide levels by applying pharmacologic agents, hormones, or short-chain ceramides to the brain through intracerebroventricular cannulae. These studies have revealed a few interesting insights into sphingolipid influences on central control of body weight. First, hypothalamic ceramide synthesis and degradation appear to be mediated in part by hormonal cues; acute administration of leptin or ghrelin rapidly alters ceramide content in mouse hypothalamus. These changes in ceramide content (decreased by leptin, increased by ghrelin) are congruent with the hormones’ effects on food intake (decreased by leptin, increased by ghrelin), supporting a role for dynamic alterations in ceramide content in controlling food intake. In agreement, arcuate nucleus infusions of myriocin (which blocks ceramide synthesis) were able to decrease food intake, consistent with the effects of leptin and ghrelin on both food intake and ceramide levels. Interestingly, infusion of ceramide itself does not acutely alter food intake, despite being able to block leptininduced decreases in food intake. This raises the question: why does altering ceramide content indirectly (by modulating synthesis or disposal via hormonal or pharmacological methods) alter food intake, but directly increasing ceramide content (with exogenous supply) does not? Many plausible explanations exist, including differences in the sites of endogenous ceramide synthesis compared with the effects of a broad ceramide delivery, an insufficient mass of C2 ceramide being provided to recapitulate an opposing effect to myriocin, or myriocin-mediated changes in the content of other sphingolipids. Indeed, a few recent reports point to a role for ceramides in altering hypothalamic function in its regulation of metabolism and body weight control. Two reports in 4 Neuroendocrinology by Dr. Ismael González-Garcia, have indicated that “altering hypothalamic ceramide content via external ceramide administration or inhibitors of ceramide synthesis can change food intake”. Importantly, these effects were in part mediated by the VMH. As several reports are beginning to identify hypothalamic (and specifically SF1 neurons of the VMH) ceramides in the control of whole-body metabolism, we decided to investigate the effect of reducing endogenous ceramides in the VMH during high fat diet-induced obesity. Regulated and decreased ceramide levels in the VMH will serve as a profound effort in beginning to developing targets for treatment against obesity-induced metabolic diseases as a result of improved glucose storage, glucose metabolism, and altered satiety. 5 METHODS Animals and Diets SF1-AC (short for Acid Ceramidase) animals were generated by interbreeding of previously generated SF1-cre, rosa26-lox-stop-lox-rtTA, and tet-response-element driven AC—tre-AC (Belteki, 2005). Animals were previously bred to a C57Bl6 background. rtTA+/tre-AC+ animals were bred to rtTA+/tre-AC+/SF1-cre+ animals to generate experimental cohorts; wild-type (WT) animals lacked SF1-cre. All animals used in these studies were littermate/cage-mate males, and all metabolic phenotyping was done on animals of identical body weight. For the high fat diet (HFD) studies, high fat-high sucrose diet containing 600 mg/kg doxycycline was utilized. The chow diet also contained 600 mg/kg doxycycline. For the pre-fattening, doxycycline-reversal experiments, animals were provided high fat diet without dox for 6 weeks before being provided HFD-dox. All animals were kept on a standard 12-hr light-dark cycle, and all manipulations were done between 0700-1700. Body composition was measured at the indicated time point by NMR. mRNA Measurement Tissue mRNA was extracted using Trizol and converted into cDNA with a commercially-available kit according to the manufacturer’s instructions. qPCR was carried out in a Roche qPCR machine using specific primer sequences. Gapdh or Rps18 was used as housekeeping genes. 6 Oral Glucose Tolerance Tests For oral glucose tolerance tests (OGTT), animals were placed in a procedure room to acclimate and fasted for 6 hours. Glucose (2.5 g/kg) was injected to the stomach via oral gavage at the end of the fast, and blood glucose was measured at the indicated points—15, 30, 60, and 120 minutes. To measure glucose- stimulated insulin secretion, serum was drawn before and 15 minutes after the glucose bolus. Insulin Tolerance Tests Unless otherwise noted, animals were placed in a procedure room to acclimate and fasted for 3 hours. Insulin was prepared in PBS with varying milligrams of BSA as a carrier, and injected at 0.75 U/kg intraperitoneally. Blood glucose was measured at the indicated points. To experimentally induce hypoglycemia, animals were fasted overnight (20 hours) before being injected with a double dose (1.5 U/kg) of insulin, and serum was drawn immediately before the injection and 60 minutes after the injection to monitor glucagon secretion. Serum Insulin and Glucagon Measurement. For all serum insulin measurements, a commercially available insulin ELISA (Crystal Chem, Elk Grove Village, IL, USA) was used according to manufacturer instructions. For serum glucagon measurements, a commercially available glucagon ELISA (Mercodia, Uppsala, Sweden) was used according to manufacturer instructions. 7 Gene Expression Tissues were harvested, dissected as needed, and snap frozen in liquid nitrogen. To extract mRNA, tissues were homogenized in Trizol and extracted according to manufacturer specifications (Thermo Fisher, Waltham, MA, USA). mRNA concentration was measured with a spectrophotometer, and cDNA was created using a kit according to manufacturer instructions (Thermo Fisher). Gene expression was assayed using the primers as will be described. Hyperinsulinemic-Euglycemic Clamps Hyperinsulinemic-euglycemic clamps were done as previously described. Briefly, animals were fasted for five hours in the morning. Next, 3H-labeled glucose was infused for 90 minutes to assay basal hepatic glucose output. During the clamp, animals were clamped at 4U/kg/min insulin, and glucose infusion rate was varied until animals achieved a steady-state blood concentration of 150 mg/dL glucose. Glucagon Tolerance Tests To assay glucagon tolerance, animals were fasted for 1 hour to empty the stomach before being injected with glucagon at the indicated dose. Blood glucose was monitored at the indicated time points. To understand the effect of insulin sensitivity on the glucagon tolerance assay, myriocin (Sigma-Aldrich, 0.3 mg/kg) was injected for 3.5 weeks on every other day schedule. Arginine Tolerance Test 8 To assay the blood glucose response to endogenous glucagon and insulin secretion, arginine (2 g/kg) was injected into 2-hr fasted animals. Blood glucose was measured at the indicated time points, and serum was drawn at t = 0 and t = 10 minutes. Western Blotting For immunoblotting analysis, total protein extracts were prepared from liver tissues of mice and transferred to a nitrocellulose membrane (Bio-Rad Laboratories). The blotted membrane was blocked in 1× Tris-buffered saline (TBS) containing 0.1% Tween and 5% (wt/vol) nonfat dry milk (TBST- MLK) for 1 h at room temperature with gentle, constant agitation. After incubation with primary antibodies anti–phospho-CREB, antiCREB, anti-PEPCK (Cell Signaling Technologies), anti-glucagon receptor (Alomone Labs catalog number AGR-021), or anti–γ-tubulin (Sigma) in freshly prepared TBSTMLK at 4 °C overnight with agitation, the membrane was washed two times with TBST buffer. This was followed by incubating with secondary anti-rabbit or anti-mouse horseradish peroxidase-conjugated Ig antibodies in TBST-MLK for 1 h at room temperature with agitation. The membrane was then washed three times with TBST buffer, and the proteins of interest on immunoblots were detected by an ECL plus Western blotting detection system (GE Healthcare Life Sciences). 9 RESULTS To study the effects of ceramide accumulation in the VMH during obesity in the context of metabolism, we used previously generated animals to overexpress acid ceramidase (AC) in SF1 neurons of the VMH. These animals (termed SF1-AC) are able to inducibly deplete ceramides in SF1 neurons when provided doxycycline in the diet. To generate SF1-AC mice, previously generated and validated SF1-cre mice were bred to rosa26-loxP-stop-loxP-rtTA and tet-ON-AC mice (Xia, 2015). Cre-mediated excision of a stop cassette allows for expression of rtTA in SF1-expressing cells. When animals are provided doxycycline in the diet, the triple transgenic SF1-AC animals will overexpress AC in cells expressing SF1, and deplete ceramide content within these cells (Figure 1A). Animals were generally obtained in the expected Mendellian ratio (data not shown). To validate overexpression of AC in SF1-expressing tissues, WT and SF1-AC animals were provided doxycycline (600 mg/kg) in high fat diet (HFD). As expected, in tissues which are known to express SF1-cre (VMH, adrenal, and testes), an increase in AC mRNA was detected in SF1-AC animals, in a dox-dependent manner(Figure 1B-D). Tissues without expression of SF1-cre did not demonstrate increased AC expression (Figure 2A). As adrenals and testes can alter systemic metabolism through their canonical roles in the secretion of corticosterone and testosterone, we measured these hormones to ensure that changes in corticosterone and testosterone were not the reason for any ensuing phenotypic changes. Importantly, serum corticosterone (in chow and HFD-fed animals) and testosterone (in HFD-fed animals) levels were unchanged in SF1-AC animals (Figure 2B-D), and adrenal tissue histology was grossly normal in chow-fed SF1-AC animals 10 (Figure 2E), arguing against any non-neuronal effects in mediating the phenotypes noted below. Figure 1: Validation of acid ceramidase overexpression in SF1-expressing cells. (A) Schematic of mouse genetics allowing for inducible overexpression of AC in SF1 neurons. Animals expressing Cre recombinase in SF1-expressing cells were bred to animals expressing rtTA under the control of loxP- flanked stop cassettes and tet response element-driven acid ceramidase. When animals are provided doxycycline in the diet, acid ceramidase in overexpressed in SF1-expressing tissues. Expression of acid ceramidase in SF1-expressing tissues: (B) VMH, (C) adrenal, (D) testes. Figure 2: SF1-AC animals do not have altered adrenal or testicular hormone secretion. Expression of AC is not changed in SF1-AC animals in liver (A). Despite increases in AC expression in testes and adrenals, testosterone (B) and corticosterone (C-D) are unchanged in SF1-AC animals. Representative adrenal H&E staining is shown in (E). 11 To understand the effect of acid ceramidase overexpression on neuronal integrity and function, we bred SF1-AC mice to a fluorescent SF1-eGFP reporter mouse, allowing for visualization and electrophysiological measurement of SF1-neurons in SF1-AC animals. WT:SF1-GFP and SF1-AC:SF1-GFP animals were provided high fat diet containing doxycycline for 6 weeks. GFP-labeled neurons in the VMH appeared grossly normal in SF1-AC mice (Figure 4). Although not significant, a trend of depolarized resting membrane potential (RMP) was noted in SF1-AC animals, possibly indicating reduced threshold to excitability (Figure 3A). However, this increase in resting membrane potential did not alter spontaneous action potential frequency (Figure 3B). Importantly, SF1-AC mice display a significant increase in the frequency of excitatory post synaptic currents (EPSC) and a trend toward an increased frequency of inhibitory post synaptic currents (IPSC) (Figure 3C-D). Similarly, a trend toward increased input resistance, and therefor increased excitability of SF1 neurons was noted in SF1-AC mice (Figure 3H). However, the evoked response to EPSCs and IPSCs remained unchanged between groups (Figure 3E-F). 12 Figure 3: Acid ceramidase overexpression alters excitability of SF1 neurons WT and SF1-AC mice expressing an SF1-eGFP reporter fluorescent were fed high fat diet + dox for 6 weeks before electrophysiology recordings. (A) Resting membrane potential of SF1-GFP neurons in WT and SF1-AC mice. (B) Frequency of spontaneous action potentials in SF1-GFP neurons. Spontaneous excitatory post- synaptic current frequency (C) and amplitude (E) and inhibitory post-synaptic current frequency (D) and amplitude (F). Figure 4: Demonstration of electrophysiology technique. (A) Brightfield, (B) GFP, and (C) Alexa 594 illumination of the same neurons in WT (above) and SF1-AC animals (below). Scale bar is 50 μm. While VMH ceramide accumulation does not affect body weight or composition control on chow or high fat diet, as others have previously implicated hypothalamic ceramide accumulation (and specifically VMH ceramide accumulation) in altering body weight control through food intake and brown adipose tissue function, we studied the effects of VMH ceramide depletion on body weight gain and body composition. To do this, animals were provided dox-chow (Figure 5A-B) or dox-HFD (Figure 5C-D). Body weight (Figure 5A & 5C) was measured at the indicated time points and was unchanged through the entire experiment, while body composition (Figure 5B &5D) was also unchanged when measured at three months (chow) or two months (HFD) of the 13 experiment, demonstrating that depletion of VMH ceramides does alter body weight and composition in SF1-AC mice in response to either chow or HFD. To assay whether tissue size was altered, tissue mass was measured for a cohort of 4-month HFD-fed animals, demonstrating no changes in the weight of any of the harvested tissues themselves aside from fat (Figure 5E). Figure 5: SF1-AC mice do not have altered weight gain or body composition on chow or HFD. WT and SF1-AC were provided chow (A-B) or HFD (C-D) containing dox. Body weights (A&C) were measured at indicated time points. Body composition (B&D) was measured by NMR prior to sacrifice. As SF1-neurons are known to regulate insulin sensitivity, we hypothesized that altering ceramide content in the VMH may alter glucose homeostasis and insulin sensitivity. To test this, we analyzed WT and SF1-AC mice fed dox-chow for 3 months. Random-fed blood glucose and insulin (Figure 7A-B) were unchanged. Similarly, oral glucose tolerance (Figure 7C) and insulin tolerance (Figure 7D) were unchanged, demonstrating that VMH ceramide depletion does not alter glucose homeostasis in basic 14 regular chow fed animals. As it has been noted that manipulations of SF1-neurons typically require a metabolic challenge to unveil any phenotypic changes, we also assayed measures of glucose metabolism in WT and SF1-AC mice fed dox-HFD. On HFD, oral glucose tolerance was improved in SF1-AC mice without changes in insulin secretion (Figure 6A-C). Interestingly, AC overexpression in the VMH was able to strikingly improve high fat-mediated glucose intolerance in animals pre-fattened with 8 weeks of HFD before the addition of dox to the diet (Figure 6E-F). However, despite these improvements in glucose homeostasis, insulin tolerance did not have as much of a significant change as glucose tolerance (Figure 6D). To understand the mechanism driving this improvement in glucose homeostasis, we performed hyperinsulinemic-euglycemic clamps on WT and SF1-AC animals fed HFD-dox. During the clamp, there was no difference in the glucose infusion rate required to clamp the animals at steady state, and the radiolabeled-glucose disposal remained unchanged between groups (Figure 6G-H), validating the lack of change in insulin tolerance noted above. We were surprised to find, however, that hepatic glucose output was decreased in SF1-AC animals only in the basal (unstimulated) state (Figure 6I). Typically, when changes in hepatic glucose output are discovered during a hyperinsulinemic-euglycemic clamp, they are found either during both the basal and clamp state, or only during the insulin-stimulated clamp state; we are unaware of any other report of a mouse model where hepatic glucose output is lower during the basal state, and not the clamp state. During both the basal and clamped state, no significant differences were detected in plasma glucose, insulin, or free fatty acids (Table 1). 15 Table 1: Serum chemistry during hyperinsulinemic-euglycemic clamp. Depletion of VMH ceramides alters glucagon sensitivity. As SF1-neurons have been well-characterized to control insulin sensitivity, we were surprised to see an improvement in glucose homeostasis but no change in insulin sensitivity in SF1-AC mice. We were particularly confounded by the insights revealed in the clamp studies, where hepatic glucose output was lowered only in the basal state. However, previous reports have demonstrated alterations in hepatic glucose output in manipulations of SF1neurons. In VMH-lesioned rats and mice with blunted SF1-neuron signaling capability, animals are unable to defend against hypoglycemia due to a defect in secretion of the counterregulatory hormones glucagon, epinephrine, and norephinephrine, which serve to raise blood glucose (Tong 2007). As the basal (unstimulated) state of the clamp is a fasted state, we hypothesized that glucagon secretion or sensitivity was impaired in SF1AC animals, leading to decreased hepatic glucose output and accounting for the improvements in glucose tolerance that had been noted earlier. In support of this, we showed lowered blood glucose and increased serum triglyceride in HFD-fed, overnight fasted SF1-AC mice (Figure 6J-K). The decrease in blood glucose was not a result of impaired glucagon secretion itself, as circulating glucagon in SF1-AC mice was not altered in either the fed or fasted condition (Figure 6L).This phenotype is reminiscent of 16 that seen in glucagon receptor knockout animals, which have lowered blood glucose and increased fasting triglycerides (Longuet, 2008). Figure 6: SF1-AC mice have improved glucose tolerance without altered insulin resistance. In HFD-fed WT and SF1-AC mice, glucose tolerance (2.5g/kg, (A-C)) and insulin tolerance (0.75u/kg,(D)) were analyzed. A separate cohort of animals were provided high fat diet without dox for 8 weeks, to examine whether HFD-induced glucose intolerance could be reversed. Glucose tolerance tests (2.5 g/kg) were administered before and 5 days after animals were switched to HFD-dox. Paired responses of WT (E) and SF1-AC (F) are shown. (G-I) Hyperinsulinemic-euglycemic clamps were used to assay changes in insulin sensitivity. Glucose uptake (G), glucose infusion rate (H), and hepatic glucose output (I), are unchanged during the clamp. However, basal hepatic glucose output (I) is altered. Serum triglyceride (J), glucose (K), and glucagon (L) are shown in 18-hr fasted and refed conditions. 17 Figure 7: Glucose tolerance is unchanged in chow-fed SF1-AC animals. Random-fed glucose (A) and insulin (B) were taken from chow-fed WT and SF1-AC animals. Glucose tolerance tests (2.5 g/kg, (C)) and insulin tolerance tests (0.75 u/kg, (D)) were performed. Three paradigms were thus developed by which to evaluate in vivo glucagon secretion and sensitivity. First, we used a previously validated pharmacologic intraperitoneal glucagon injection, and analyzed blood glucose through the duration of the study. In chow-fed, 5-day-HFD-fed, and 4-month HFD-fed SF1-AC animals, glucose excursion in response to glucagon was reduced (Figure 8A-F). As a comparison, we also carried out these experiments in a well-validated model of improved insulin sensitivity: HFD-fed myriocin-injected animals (Figure 9), where reductions in blood glucose are only apparent after the time of glucose excursion in response to glucagon (the first ~ 30 minutes of the injection). Thus, these acute glucagon injections suggest that AC overexpression in VMH can blunt glucagon sensitivity. Although these experiments were intriguing, they are also a response to an exogenous, nonphysiologic bolus of glucagon. Is endogenous glucagon action similarly 18 blunted? To answer this, we fasted HFD-fed WT and SF1-AC animals for 20 hours, and then performed an insulin tolerance test at double our standard dosage (1.5 U/kg). This experimental setup will rapidly drive animals into hypoglycemia, where both endogenous glucagon secretion and action can be assessed. Although HFD- fed SF1-AC animals were able to defend from hypoglycemia during fasting and had similar fasting glucagon levels, euglycemia was not sufficiently maintained during the insulin challenge; in response to a strong insulin bolus, hypoglycemia was reached more rapidly in SF1-AC animals (Figure 8G). Importantly, glucagon secretion in the hypoglycemic state was unchanged (Figure 8H), suggesting that the difference noted in blood glucose was due to changes in glucagon sensitivity rather than changes in glucagon secretion. To see whether these effects could be magnified, we used a stronger stimulant of glucagon secretion: an intraperitoneal arginine injection. Arginine is well-known for its ability to depolarize pancreatic alpha and beta cells, causing massive degranulation and release of insulin and glucagon. As we have thoroughly demonstrated above, insulin sensitivity is unchanged in SF1-AC animals, so any alterations in blood glucose during the experiment must result from changes in glucagon secretion or glucagon sensitivity. Indeed, SF1-AC animals had strikingly lower blood glucose following an arginine injection, while both insulin and glucagon secretion were unchanged (Figure 8I-L). This again suggests that glucagon sensitivity is markedly reduced, as the lower blood glucose levels during the assay must result from changes in glucagon (and not insulin) sensitivity. Again, glucagon or insulin secretion was identical between groups after the arginine injection, ruling out a possibility of a defect or improvement of pancreatic hormone 19 secretion. These experiments demonstrate that SF1-AC animals have reduced glucagon sensitivity independent of any changes in glucagon secretion. Figure 8: Glucagon sensitivity is reduced in SF1-AC mice. Glucagon tolerance was assayed in chow-fed (120 ug/kg, (A&B)), 5-days HFD-fed (60 ug/kg, (C&D)), and 4month HFD-fed (120 ug/kg, (E&F)) WT and SF1-AC mice. To assay endogenous glucagon action, HFD-fed WT and SF1-AC animals were fasted overnight (20 hrs) before an intraperitoneal insulin injection (1.5 u/kg); blood glucose (G) and glucagon secretion (H) were monitored. To further demonstrate that endogenous glucagon action is impaired, HFD-fed WT and SF1-AC mice were fasted for two hours prior to an intraperitoneal arginine injection (2g/kg), and blood glucose (I-J), insulin secretion (K), and glucagon secretion (L) were monitored during the assay. Figure 9: Demonstration of glucagon tolerance test kinetics in a model of improved insulin sensitivity. As a demonstration of the kinetics of a glucagon tolerance test in a 20 model of improved insulin sensitivity, mice were fed high fat diet for 5 months, before animals were injected with myriocin (0.3 mg/kg) or vehicle (PBS and MeOH) every other day for 25 days. As expected, myriocin reduced body weight gain from HFD feeding (A). On the 25th day, animals were fasted for 1 hour before glucagon (120 ug/kg) was injected and blood glucose (B) was monitored. Altered glucagon sensitivity does not arise from altered hepatic lipid uptake and oxidation or glucagon receptor signaling. To explain the alterations seen in glucagon sensitivity, we first hypothesized that changes in hepatic glucagon receptor protein or signaling had occurred. However, glucagon receptor protein levels were unaltered in the liver of SF1-AC animals, and SF1-AC mice had a surprising increase in PEPCK, the transcription of which is driven by glucagon receptor signaling (Figure 10A). Activated glucagon receptor is known to increase intracellular cAMP, which then activates protein kinase A to phosphorylate cAMP response element-binding protein (CREB), leading to its downstream effects on glucose output and gluconeogenesis. We theorized that the changes in glucagon sensitivity in SF1-AC mice may be due to changes in glucagonstimulated phospho-CREB. However, glucagon-stimulated CREB was also unaltered in the liver of SF1-AC mice (Figure 10A). Thus, glucagon receptor signaling is intact, and expression of both the receptor itself and its downstream effectors is unchanged, leading us to conclude that hepatic glucagon receptor signaling is grossly intact and does not cause the reduced glucagon sensitivity phenotype. We then hypothesized that hepatic energy supply was disturbed in SF1- AC mice, as this may impair the energy supply needed to drive gluconeogenesis in the liver. We thus carried out 3H-labeled triolein lipid tracing experiments, where radiolabeled lipid is packaged into intralipid micelles and injected intravenously into fasted animals. This allows for measurement of both tissue 21 lipid uptake and oxidation. In liver (and all other measured tissues), no changes in either tissue lipid uptake or oxidation were found (Figure 10B-C). To further assay adipose tissue function, we conducted β3-adrengergic receptor (β3-AR) stimulation, but did not note any changes in serum free fatty acids, glycerol, glucose or insulin, suggesting that there are no changes in β3-AR sensitivity (Figure 11). Thus, it is unlikely that altered lipolysis, lipid uptake, or lipid oxidation impairs hepatic gluconeogenesis in SF1-AC mice. Figure 10: Hepatic glucagon signaling and lipid metabolism are unaltered in SF1-AC animals. Glucagon receptor protein and PEPCK protein were analyzed by Western Blot (A). Quantifications are shown to the right. To analyze whole-body lipid metabolism, HFD-fed WT and SF1-AC animals were injected with 3H-labeled triolein, and tissues were analyzed for lipid oxidation (B) and lipid uptake (C). 22 Figure 11: Adipose tissue β3-adrengergic receptor signaling is unaltered. HFD-fed WT and SF1-AC animals were injected with CL-316,243 (1.5 mg/kg) immediately following removal of food. Blood glucose (A), serum glycerol (B), serum free fatty acids (C), and serum insulin (D) were assayed during the experiment. With all of this being said, enzymes of sphingolipid synthesis are not increased in the brains of HFD-fed mice. Following the insulin tolerance test, the hypothalamus and hippocampus were dissected from rest of the brain to compare, and the mRNA of sphingolipid synthesis enzymes was measured. Transcriptional control of sphingolipid metabolism enzymes was remarkably tight in the brain regions interrogated in response to HFD. In hypothalamus (Fig 12A), no significant changes in any enzymes of sphingolipid metabolism was noted, although a trend to decreases in some of the ceramide synthesis enzymes was observed. In the hippocampus (Fig 12B), some enzymes of ceramide synthesis were significantly decreased in HFD-fed animals, while ceramide degradation and modification enzymes were mostly unchanged. Finally, in the remaining brain (Fig 12C) no consistent changes in response to HFD were noted, although expression of two 23 enzymes of ceramide synthesis increased and expression of three enzymes of ceramide degradation decreased. As we were particularly surprised by the lack of any noticeable changes in the hypothalamus, we wished to validate our dissection and cDNA preparation. To do this, we also measured neuropeptides specific to the hypothalamus from these animals, and noted that (as expected from properly prepared tissues) there was a strong increase in the mRNA of Pomc and Agrp detected, two proteins enriched in the hypothalamus (Figure 13A). Similarly, we measured Dlg3 as a marker of the hippocampus, and noted a strong increase in Dlg3 in the hippocampus when compared to both hypothalamus and brain (Figure 13B). Figure 12: Measurement of sphingolipid metabolism enzymes in brain during high-fat feeding. Brains were dissected at the end of the insulin tolerance test. mRNA from the enzymes of sphingolipid metabolism was then measured in A) hypothalamus, B) hippocampus, and C) remaining brain. Data are shown as fold induction over WT as mean ± SEM. * p < 0.05 vs. chow. n = 4-6 per group. 24 Figure 13: Expression of hypothalamic and hippocampal markers in dissected brain regions. A) To validate proper dissection and cDNA synthesis, two well-characterized hypothalamic markers (POMC and AgRP) were measured in brain, hippocampus, and hypothalamus from chow and 1- d HFD fed animals. B) To validate proper dissection of hippocampus, Dlg3 (a hippocampus marker) was measured. ** p <0.01, *** p < 0.001. n = 7-8 per group. 25 DISCUSSION Aside from its effects on food intake, one report has also demonstrated that infusion of ceramide into the hypothalamus reduces energy expenditure (without altering food intake) by reducing sympathetic outflow to brown adipose tissue, resulting in increased body weight. This effect was attributed to ER stress-mediated alterations in ventromedial hypothalamus (VMH) function; when chaperone protein was exogenously supplied via adenovirus, ceramide-induced ER stress in the VMH was reversed and brown adipose tissue function was restored. Interestingly, the report also demonstrated that central ceramide administration can induce insulin resistance, which was similarly reversed by reducing ER stress. Further recent work has demonstrated that ceramide levels are dynamically altered in the hypothalamus by thyroid hormone, and that this action is important in the control of brown adipose tissue (and subsequently body weight) by the VMH. Of greater importance, however, would be the expanded use of genetic mouse models that alter ceramide synthesis/disposal in the CNS. These models allow much greater temporal, spatial, and pathway-specific control over the manipulations of ceramide concentrations. Importantly, many of the requisite mouse lines have already been generated for study in other contexts: many neuron-specific Cre recombinase lines have been generated by neuroscientists, and a number of loxP-flanked ceramide synthesis enzymes, derivation enzymes, and sphingolipid receptors have been generated by lipid biologists. It is vital that studies which aim to alter sphingolipid synthesis and degradation in neurons be done in an inducible manner. Aside from the standard concerns 26 of developmental and synaptic programming of hypothalamic neurons, sphingolipids (and glycosphingolipids in particular) are known to be vital for neuronal development; constitutive whole-body knockouts of glucosylceramide synthase, ceramide galactosyltransferase, ceramide kinase, and acid sphingomyelinase in humans and mouse have led to severe neurological and behavioral defects, sometimes to the point of lethality. Thus, to ensure that proper developmental programming of neurons is retained, inducible models of sphingolipid perturbation should be utilized, such as tamoxifeninducible cre and doxycycline- inducible Tet-on/Tet-off systems should be used. The above experiments represent the first demonstration of altering internal ceramide metabolism within targeted neuronal populations. Overexpression of acid ceramidase did not grossly impact neuronal function, as confirmed by both the electrophysiology data and the normal body weight regulation in SF1-AC mice; disruption of VMH neurons is a classical obesogenic stimulus, and thus the unchanged body weight in SF1-AC mice suggests intact VMH function (Duggan and Booth, 1986). The ceramidase overexpression did not change rates of spontaneous firing or evoked responses in SF1 neurons, but did increase excitability in recorded SF1 neurons from SF1-AC animals. We speculate that these effects may be mediated by first order SF1 neurons synapsing onto our recorded SF1 neurons in the case of EPSCs, or in the case of IPSCs by SF1 neurons increasing activity of inhibitory neurons which then synapse onto recorded SF1 neurons. This model implies that ceramidase overexpression serves to increase spontaneous synaptic vesicle release, a hypothesis which requires further study. Separately, it has been demonstrated that ceramides play a role in lipid raft function and 27 synaptic plasticity; altering ceramide content may alter synapse numbers, which may explain the effects seen in increasing both EPSCs and IPSCs (Colombaioni 2004). We demonstrate here that ceramide accumulation in SF1 neurons of the hypothalamus during high fat feeding worsens glucose tolerance in a reversible manner. Further, these changes in glucose tolerance are not related to insulin sensitivity, but rather reflect changes in glucagon sensitivity through an as-yet undescribed mechanism. During the analysis of these experiments, we were continuously surprised about the differences in our findings compared to what others have reported when C6- ceramide is exogenously supplied to the hypothalamus (Contreras 2014); we had expected our results to be directly opposite in nature to the C6-ceramide challenges. As these reports showed that C6-ceramide blunts sympathetic nerve activation of brown adipose tissue and subsequently can increase body weight in rats, we had expected an opposite finding when overexpressing AC in the VMH of mice. However, our data clearly and consistently show over multiple cohorts of animals that there is no effect of AC overexpression in SF1 neurons on body weight regulation in both chow and high fat fed animals. Similarly, investigators noted that C6-ceramide application to the hypothalamus blunts insulin sensitivity, whereas we found no effect of acid ceramidase overexpression on insulin sensitivity during insulin tolerance tests and hyperinsulinemic-euglycemic clamps, again over multiple cohorts of animals. Although there was no effect of VMH ceramide depletion on body weight or insulin sensitivity in our studies, there was a non-significant trend toward smaller brown adipose tissue fat pads, and a similar trend toward increased lipid uptake and oxidation in those fat pads (Figure 3E & 6B-C). Thus, although our data would rule out any effect of ceramide accumulation during diet-induced obesity in SF1 28 neurons on body weight or insulin sensitivity, these data may agree with the effects of modulating VMH ceramides on BAT function that others have noted. Regardless, this work reveals a heretofore unknown biology of SF1 neurons in the VMH: the regulation of systemic glucagon sensitivity. It will be of interest to validate or disprove these findings in other models of SF1 neuron manipulation where changes in glucose tolerance are seen without changes in body weight. 29 CONCLUSION The lack of any noticeable phenotype in chow-fed animals argues against any effect of AC overexpression on normal neuron function; unlike some other models of perturbed SF1 neurons, chow-fed SF1-AC animals had normal body weight, body composition, and insulin sensitivity. The vast majority of phenotypic changes were only noted with high fat feeding, in agreeance with previous findings that SF1 neurons generally require a metabolic challenge (fasting, exercise, or diet-induced obesity) before seeing the effects of SF1 neuron manipulation. Although others have demonstrated that altering hypothalamic ceramide metabolism alters insulin sensitivity, weight gain, and brown adipose tissue activation, we see none of these effects. Why are these results so divergent from other studies of ceramide in the hypothalamus? First, this work only modulates ceramide content in SF-1 neurons of the hypothalamus, whereas others have changed ceramide content or synthesis in the entire hypothalamus, where many other neuronal populations vital to the regulation of body weight and insulin sensitivity also reside. Further, these manipulations are acute, exogenous treatments rather than alterations of endogenous sphingolipids during chronic high fat feeding. Similarly, others have modulated ER stress in response to ceramide in the VMH with adenoviral-mediated gene expression; the VMH is comprised of neurons other than just SF1-neurons, and it is thus possible that those neurons mediate ceramide’s effects on insulin sensitivity and body weight control. Finally, the exogenous administration of ceramide will also affect glial populations, which are rapidly being recognized as key players in the central control of metabolism. 30 Although it is known that ceramides accumulate in the hypothalamus during high fat feeding, an open question is the origin of these hypothalamic ceramides. A few possibilities exist: first, hypothalamic ceramides may be synthesized from de novo precursors within neurons. Second, ceramides may be delivered to and across the blood brain barrier either on lipoproteins in circulation or as constituents of freely floating lipid rafts. Finally, hypothalamic astrocytes may produce ceramide and traffic this ceramide on lipoprotein to neurons. We prefer the theory of astrocyte effects which may also play a role in central ceramide metabolism (unpublished observations). This work provides further evidence of the importance of hypothalamic ceramide accumulation in driving aspects of metabolic syndrome. It will be of interest to continue this work in other neuronal populations, and separately to focus on multiple aspects of ceramide metabolism not covered here such as alterations in endogenous synthesis and disposal pathways and perturbations of ceramide derivation (i.e. the production of glucosylceramide, ceramide-1-phosphate, etc.). 31 REFERENCES 1. Flegal, K.M., et al., Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999-2010. Jama, 2012. 307(5): p. 491-497. 2. Adams, K.F., et al., Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. New England Journal of Medicine, 2006. 355(8): p. 763-778. 3. Rodgers, R.J., M.H. Tschöp, and J.P. Wilding, Anti-obesity drugs: past, present and future. Disease models & mechanisms, 2012. 5(5): p. 621-626. 4. Elmquist, J.K., C.F. Elias, and C.B. Saper, From lesions to leptin: hypothalamic control of food intake and body weight. Neuron, 1999. 22(2): p. 221-232. 5. Coll, A.P. and G.S. Yeo, The hypothalamus and metabolism: integrating signals to control energy and glucose homeostasis. Current opinion in pharmacology, 2013. 13(6): p. 970-976. 6. Coppari, R., Hypothalamic Neurones Governing Glucose Homeostasis. Journal of neuroendocrinology, 2015. 27(6): p. 399-405. 7. Choi, Y.-H., et al., Revisiting the ventral medial nucleus of the hypothalamus: the roles of SF-1 neurons in energy homeostasis. Frontiers in neuroscience, 2013. 7. 8. Borg, M.A., et al., Local ventromedial hypothalamus glucose perfusion blocks counterregulation during systemic hypoglycemia in awake rats. Journal of Clinical Investigation, 1997. 99(2): p. 361. 32 9. Borg, W.P., et al., Ventromedial hypothalamic lesions in rats suppress counterregulatory responses to hypoglycemia. Journal of Clinical Investigation, 1994. 93(4): p. 1677. 10. Tong, Q., et al., Synaptic glutamate release by ventromedial hypothalamic neurons is part of the neurocircuitry that prevents hypoglycemia. Cell metabolism, 2007. 5(5): p. 383-393. 11. Kamitakahara, A., B. Xu, and R. Simerly, Ventromedial hypothalamic expression of Bdnf is required to establish normal patterns of afferent GABAergic connectivity and responses to hypoglycemia. Molecular metabolism, 2016. 5(2): p. 91-101. 12. Ramadori, G., et al., SIRT1 deacetylase in SF1 neurons protects against metabolic imbalance. Cell metabolism, 2011. 14(3): p. 301-312. 13. Fujikawa, T., et al., SF-1 expression in the hypothalamus is required for beneficial metabolic effects of exercise. Elife, 2016. 5: p. e18206. 14. Kim, K.W., et al., FOXO1 in the ventromedial hypothalamus regulates energy balance. The Journal of clinical investigation, 2012. 122(7): p. 2578. 15. Zhang, R., et al., Selective inactivation of Socs3 in SF1 neurons improves glucose homeostasis without affecting body weight. Endocrinology, 2008. 149(11): p. 5654-5661. 16. Dhillon, H., et al., Leptin directly activates SF1 neurons in the VMH, and this action by leptin is required for normal body-weight homeostasis. Neuron, 2006. 49(2): p. 191-203. 33 17. Klöckener, T., et al., High-fat feeding promotes obesity via insulin receptor/PI3Kdependent inhibition of SF-1 VMH neurons. Nature neuroscience, 2011. 14(7): p. 911-918. 18. Gault, C.R., L.M. Obeid, and Y.A. Hannun, An overview of sphingolipid metabolism: from synthesis to breakdown, in Sphingolipids as Signaling and Regulatory Molecules. 2010, Springer. p. 1-23. 19. Chavez, J.A. and S.A. Summers, A ceramide-centric view of insulin resistance. Cell metabolism, 2012. 15(5): p. 585-594. 20. Borg, M.L., et al., Consumption of a high-fat diet, but not regular endurance exercise training, regulates hypothalamic lipid accumulation in mice. The Journal of physiology, 2012. 590(17): p. 4377-4389. 21. Holland, W.L., et al., Lipid-induced insulin resistance mediated by the proinflammatory receptor TLR4 requires saturated fatty acid–induced ceramide biosynthesis in mice. The Journal of clinical investigation, 2011. 121(5): p. 1858. 22. González-García, I., et al., Hypothalamic lipids: key regulators of whole body energy balance. Neuroendocrinology, 2017. 104(4): p. 398-411. 23. Gao, S., et al., Important roles of brain-specific carnitine palmitoyltransferase and ceramide metabolism in leptin hypothalamic control of feeding. Proc Natl Acad Sci U S A, 2011. 108(23): p. 9691-6. 24. Ramirez, S., et al., Hypothalamic ceramide levels regulated by CPT1C mediate the orexigenic effect of ghrelin. Diabetes, 2013. 62(7): p. 2329-37. 34 25. Contreras, C., et al., Central Ceramide-Induced Hypothalamic Lipotoxicity and ER Stress Regulate Energy Balance. Cell Rep, 2014. 9(1): p. 366-77. 26. Martínez-Sánchez, N., et al., Hypothalamic AMPK-ER Stress-JNK1 Axis Mediates the Central Actions of Thyroid Hormones on Energy Balance. Cell Metabolism, 2017. 26(1): p. 212-229. e12. 27. Xia, J.Y., et al., Targeted induction of ceramide degradation leads to improved systemic metabolism and reduced hepatic steatosis. Cell metabolism, 2015. 22(2): p. 266-278. 28. Belteki, G., et al., Conditional and inducible transgene expression in mice through the combinatorial use of Cre-mediated recombination and tetracycline induction. Nucleic acids research, 2005. 33(5): p. e51-e51. 29. Stallings, N.R., et al., Development of a transgenic green fluorescent protein lineage marker for steroidogenic factor 1. Molecular Endocrinology, 2002. 16(10): p. 2360-2370. 30. Sun, J., et al., Adiponectin potentiates the acute effects of leptin in arcuate Pomc neurons. Molecular metabolism, 2016. 5(10): p. 882-891. 31. Kusminski, C.M., et al., MitoNEET-driven alterations in adipocyte mitochondrial activity reveal a crucial adaptive process that preserves insulin sensitivity in obesity. Nature medicine, 2012. 18(10): p. 1539-1549. 32. Longuet, C., et al., The glucagon receptor is required for the adaptive metabolic response to fasting. Cell metabolism, 2008. 8(5): p. 359-371. 33. Duggan, J. and D. Booth, Obesity, overeating, and rapid gastric emptying in rats with ventromedial hypothalamic lesions. Science, 1986. 231: p. 609-612. 35 34. Colombaioni, L. and M. Garcia-Gil, Sphingolipid metabolites in neural signalling and function. Brain Research Reviews, 2004. 46(3): p. 328-355. 35. Douglass, J.D., M.D. Dorfman, and J.P. Thaler, Glia: silent partners in energy homeostasis and obesity pathogenesis. Diabetologia, 2017: p. 1-11. 36 Name of Candidate: James Banks Date of Submission: May 3, 2022