Introduction
Two discoveries have softened the traditional differentiation between the classic model of nutrient sensing1 and the concept of endocrine signals controlling energy status2 by drawing attention to the regulation of energy homeostasis by circulating long-chain fatty acids. One group recently reported that one specific adipocyte-derived long-chain fatty acid (C16:1n7), the lipokine palmitoleate, functions as a hormone regulating systemic insulin sensitivity3. A recent study followed with the discovery that a gastrointestinal lipid metabolite, N-acylphosphatidylethanolamine (NAPE), can function as an endocrine signal that targets hypothalamic energy balance centers to control food intake, particularly when the acyl NAPE species is C16:0 (ref. 4). Now, ten years after ghrelin's discovery5, 6, it is revealing itself as yet another nutrient-hormone hybrid with the specific role of linking macronutrient composition with central nervous system energy balance regulation.
Ghrelin, the only circulating orexigenic peptide hormone and the only peptide known to require a fatty acid modification5, is also the only known afferent endocrine factor that depends on intraneuronal fatty acid metabolism6. Unique characteristics of the predominantly stomach-derived ghrelin include: appetite stimulation7, promotion of adiposity7, activation of hypothalamic neuropeptide Y–and agouti-related protein–expressing neurons8, meal-associated secretory peaks9 and its distinctive chemistry requiring acylation with medium-chain fatty acids (MCFAs) at its Ser3 residue in order to become a proper ligand for its only known receptor, the growth hormone secretagogue receptor-1a (GHS-R1a)5. The inactive form of ghrelin, des-acyl ghrelin, although more abundant in plasma than the active acyl ghrelin, lacks a fatty acid side chain and thus cannot activate GHS-R1a (ref. 5).
We10 and others11 recently identified the specific enzyme responsible for the acylation and activation of ghrelin and named it GOAT. Here we show that GOAT seems to function as a gastric lipid sensor linking selected ingested nutrients with hypothalamic energy balance regulation via the endocrine ghrelin system. Using murine genetic models with specific loss (GOAT-null) or gain (mice overexpressing both human GOAT and human ghrelin) of GOAT function, we show that GOAT is necessary and sufficient to mediate the impact of certain dietary lipids on body adiposity. Using mass spectrometry and gene expression profiling, we also find that activation of the GOAT-ghrelin system is triggered by a specific lipid-rich environment rather than by caloric deprivation. These observations suggest that the ghrelin-GOAT system informs the central nervous system about availability, rather than absence, of calories, thereby reversing the previously accepted model.
Because ghrelin plasma concentrations surge shortly before a meal, GOAT messenger RNA levels are expected to follow the ghrelin secretion pattern, increasing in the fasted state and decreasing postprandially. To test this hypothesis, we fasted four groups of C57BL/6 mice over a time course of 36 h and examined the expression patterns of both the GOAT gene (Mboat4) and the ghrelin gene (Ghrl). In addition, we measured the plasma concentrations of both acyl and des-acyl ghrelin in these mice.
To our surprise, gastric Mboat4 expression was highest under ad libitum conditions and decreased significantly with fasting for 12, 24 or 36 h (Fig. 1a). Ghrl mRNA expression also tended to be decreased after prolonged fasting and resembled the expression pattern of Mboat4; however, these changes did not reach statistical significance (Fig. 1a). Notably, blood acyl ghrelin concentrations were not changed over the time course of fasting (Fig. 1b). However, blood concentrations of inactive des-acyl ghrelin increased significantly with fasting (Fig. 1b).
Figure 1: Long-term fasting decreases Mboat4 expression and does not increase ghrelin acylation.
(a) Fold change in gastric Mboat4 and Ghrl mRNA levels of C57BL/6 mice, measured with quantitative PCR after fasting for 12 h (n = 6), 24 h (n = 5) or 36 h (n = 8) compared to mice fed chow ad libitum (ad lib) (n = 7). (b) Blood concentrations of acyl and des-acyl ghrelin in the same mice as in a after fasting for 12 h, 24 h or 36 h or feeding ad libitum. (c) Ghrelin immunoprecipitation matrix-assisted laser desorption-ionization–time-of-flight mass spectometry (IPMS) analyses of stomachs from mice exposed to normal (control) or glyceryl triheptanoate (C7)-containing diets. Downward arrow denotes peak corresponding to heptanoyl-modified ghrelin peptide (m/z 3,301). Peaks at m/z 3,189, 3,212, 3,315 and 3,339 correspond to mouse forms of des-acyl ghrelin, des-acyl ghrelin standard (STD), octanoylated ghrelin and octanoylated ghrelin standard, respectively. (d) Total concentration of ghrelin forms isolated from stomachs of C57BL/6 mice fed chow or triheptanoate enriched diet. (e) Total blood concentration of ghrelin forms in C57BL/6 mice fed chow or glyceryl triheptanoate-enriched diet. All data are represented as means 
s.e.m. (*P = 0.0205; #P < 0.0001; †P = 0.0066 versus C8-ghrelin control diet; ††P = 0.0126 versus C8-ghrelin control diet).
Owing to the fact that there are two substrates for GOAT, ghrelin and MCFA, and that Ghrl expression is relatively constant during fasting, these data suggest that Mboat4 is downregulated in the absence of available MCFAs. To further test this hypothesis, we fed C57BL/6 mice with a diet rich in glycerol triheptanoate, a medium-chain triglyceride (MCT) containing heptanoic acid (C7:0), which is not synthesized de novo in mice. It has previously been shown that dietary lipids can directly influence ghrelin acylation12. Similar to those results, we have found that heptanoic acid is used for ghrelin acylation. The C7 ghrelin can be found in stomach tissue (Fig. 1c,d) and blood samples (Fig. 1e). Acylation of ghrelin with heptanoic acid seems to be at the expense of acylation with octanoic acid, as octanoylated ghrelin abundance was significantly lower in the triheptanoate-fed mice (Fig. 1d,e). The fact that heptanoylated ghrelin was more abundant than octanoylated ghrelin in mice fed this special diet supports our theory that dietary lipids have a major role in ghrelin acylation and that regulation of acyl ghrelin production and secretion is dependent on the MCFA substrate. To investigate whether small and acute changes in the amount of incoming dietary lipids have an impact on Mboat4 regulation or GOAT activity, we measured Mboat4 and Ghrl expression as well as GOAT substrate (des-acyl ghrelin) and the GOAT enzyme product (acyl ghrelin) in wild-type (WT) mice at several time points during the light cycle and the dark cycle (Supplementary Fig. 1a). Gene expression and GOAT enzyme activity tended to be more associated with the feeding period in the dark phase (Supplementary Fig. 1a). Blood concentrations of acyl ghrelin were significantly increased two hours after beginning the dark phase (Supplementary Fig. 1b), just after the most intense feeding period in mice during the first hours of the dark phase13. To study the role of chronically increased food intake and obesity on the GOAT-ghrelin system, we measured gastric expression of Mboat4 and Ghrl in ob/ob mice fed ad libitum and compared it to age-matched WT mice. Neither Mboat4 nor Ghrl mRNA levels were changed by leptin deficiency–induced obesity (Supplementary Fig. 1c).
Next, we generated and examined GOAT-knockout mice (Mboat4-/-; Supplementary Fig. 2), which only lack acyl modified forms of ghrelin10 and show increased levels of des-acyl ghrelin in contrast to WT mice (Supplementary Fig. 2b), to study the physiological role of GOAT on diets of various macronutrient compositions in comparison with WT littermate mice. Mboat4-/- mice fed standard chow diet had normal body weight and fat mass (Supplementary Fig. 3a). However, when fed high-fat diet for at least 8 weeks, the body weight of Mboat4-/- mice decreased significantly, whereas we detected no significant body composition changes over that observation period and on that diet (Supplementary Fig. 3b).
In an attempt to enhance and further clarify the phenotype, we therefore studied a new cohort of GOAT-deficient mice and age-matched littermate controls under modified dietary conditions. Because we10 and others12 showed in previous experiments that MCT dietary lipids are a direct source for ghrelin acylation, and because MCFAs (C8, C10) are substrates for GOAT-mediated ghrelin acylation10, we wondered whether ghrelin activation can be enhanced by a diet rich in octanoic and decanoic acids. Therefore, we designed a diet that contains 10% of digestible calories from glyceryl trioctanoate and glyceryl tridecanoate, hypothesizing that this diet would specifically trigger ghrelin octanoylation and decanoylation in WT mice. Mboat4-/- mice should not be influenced by this diet, as they are not able to acylate ghrelin owing to the global GOAT deletion. Feeding an MCT-containing diet to Mboat4-/- and WT mice resulted in significantly lower body weights of Mboat4-/- mice (Fig. 2a). Analysis of body composition revealed that the difference in body weight can be explained by substantially lower fat mass (Fig. 2a). Such lower fat mass did not result from lower food intake, which was actually higher in Mboat4-/- mice, but was probably based on higher energy expenditure in the light phase (Fig. 2a). Furthermore, in spite of the decreased adiposity of the Mboat4-/- mice, glucose homeostasis in this strain was not statistically different from WT mice, as judged by intraperitoneal glucose tolerance tests (Supplementary Fig. 3c).
Figure 2: GOAT regulates energy homeostasis.
(a) Body weight (29.39 0.64 g versus 27.47 0.44 g), fat mass (24.14 2.45% versus 17.90 1.74%; measured by nuclear magnetic resonance), food intake (3.3 0.1 g d-1 versus 4.2 0.3 g d-1), energy expenditure (light phase: 157.3 14.74 kcal per d per kg0.75 versus 179.5 12.72 kcal per h per kg0.75) and respiratory quotient of WT and Mboat4-/- mice chronically fed MCT diet. (b) Body weight (31.88 1.32 g versus 34.93 0.48 g), fat mass (10.29 0.84 g versus 12.88 0.43 g; measured by nuclear magnetic resonance), food intake, energy expenditure and respiratory quotient of WT and GOAT and ghrelin–transgenic (Tg) mice chronically fed an MCT-enriched diet. (c) Ghrelin IPMS analyses of blood from either WT or GOAT and ghrelin–transgenic mice exposed long term to an MCT-enriched diet. Downward arrows denote endogenous mouse des-acyl (m/z 3,189) or octanoyl-modified (m/z 3,315) ghrelin forms or the standard peptides (m/z 3,212, 3,229) used in these studies. Peaks at m/z 3,245, 3,287 and 3,371 represent human des-acyl, acetyl-modified and octanoyl-modified forms of ghrelin 1–28. Peaks at m/z 3,429, 3,471 and 3,555 correspond to human des–acyl, acetyl-modified and octanoyl-modified forms of ghrelin 1–30. Bar graphs represent mean ghrelin concentrations of WT (n = 8) and transgenic (n = 8) mice fed an MCT-enriched diet. All data are represented as means s.e.m. (*P = 0.016; #P = 0.0403 ; **P = 0.0216; $P = 0.032; †P = 0.0478; ††P = 0.0068).
In a third approach to uncover the physiological function of GOAT, we generated transgenic mice designed to express the human GHRL and human MBOAT4 genes in the liver under control of the human APOE (encoding apolipoprotein E) promoter. We generated multiple genetic lines that stably carry and express both genes (data not shown) and selected one line for additional studies of energy homeostasis (Fig. 2b) on the basis of robust detection of human des-acyl and acyl-modified ghrelin forms in circulation. The selected mouse line segregated both GHRL and MBOAT4 genes in a mendelian fashion, transmitted them stably to progeny and had high circulating levels of the principal 1–28 ghrelin and a two–amino-acid ghrelin extension variant, 1–30 ghrelin, both of which were present as des-acyl and acetyl (C2) forms. However, these mice lack octanoyl-modified forms of human ghrelin in circulation, probably owing to the lack of MCFAs in liver under normal dietary conditions (Supplementary Fig. 4a). We reasoned that dietary supplementation with MCTs would allow for octanoylation of the hepatically expressed ghrelin forms, similar to our previous cell-based studies10. To test this notion, we fed the transgenic mice a diet containing triglyceryl octanoate. On this diet, transgenic mice showed high circulating concentrations of 1–28 and 1–30 forms of des-acyl, acetyl (C2) and octanoyl (C8) ghrelin (Supplementary Fig. 4b). Similarly, the transgenic mice on the MCT-containing diet produced large amounts of total ghrelin (98.71 28.55 ng ml-1) compared to WT controls (0.61 0.33 ng ml-1; Fig. 2c). Human octanoyl-modified ghrelin concentrations were approximately 32 ng ml-1 in transgenic mice and were undetectable in WT mice (Fig. 2c). Such ghrelin overproduction led to significantly higher body weight and fat mass in transgenic mice in contrast to WT controls on the same diet over 4 weeks (Fig. 2b). Furthermore, food intake did not differ between WT and transgenic mice, but energy expenditure was significantly lower in transgenic mice during both the light and the dark phases (Fig. 2b). In addition, there was a strong trend toward increased respiratory quotient during the light phase. Overall, these data indicate that transgenic mice oxidize less fat than do WT littermates.
To show the dependence of ghrelin acylation on dietary MCT lipid availability, we switched the transgenic mice back to regular chow diet. After 2 weeks, the body weight difference between transgenic and WT mice disappeared, and we did not observe differences in energy expenditure and respiratory quotient (Supplementary Fig. 5a). Therefore, our model of overexpression of human ghrelin with human GOAT and octanoylation is diet inducible and reversible. To confirm the observed phenotype of the GOAT and ghrelin loss- or gain-of-function models described here, we generated additional populations of transgenic GOAT and ghrelin–overexpressing and Mboat4-/- mice, further corroborating the body weight and fat mass phenotype differences in these larger groups of mutant and WT mice (Supplementary Fig. 5b). To exclude the possibility that the phenotype observed here results from downstream effect of MCT feeding, we compared body weight and composition of Mboat4-/- mice on regular chow with age-matched Mboat4-/- mice on an MCT-containing diet. Because Mboat4-/- mice are not able to acylate ghrelin, any potential outcome of this comparison could possibly be explained by an MCT-diet effect. However, we found no phenotypic indication for non–GOAT-specific effects of an MCT-containing diet that might have theoretically been mediated further downstream or through secondary effects of MCTs independently from GOAT (Supplementary Fig. 5c).
To identify mechanisms for the observed GOAT-mediated changes in energy expenditure and fat mass, we used the 'switch on' transgenic GOAT-ghrelin mouse model, where an MCT-enriched diet is essential for the phenotype. Specifically, we compared genetic profiles relevant for fat oxidation, as well as concentration of uncoupling protein-1 and thyroxin. Our results point toward decreased mitochondrial fat oxidation in skeletal muscle and potentially liver but not brown adipose tissue, where neither gene expression nor protein concentrations of uncoupling protein-1 were changed (Supplementary Fig. 6a–d). Plasma concentrations of thyroxin tended to be lower in transgenic mice; however, the changes were not significant, supporting a more muscle–specific mechanism of decreased energy expenditure (Supplementary Fig. 6e).
We show that Mboat4 mRNA levels decrease upon prolonged food deprivation and we also improve upon previous immunoassay-based models by using mass spectrometry–based analyses to show that, in keeping with our GOAT expression studies, blood concentrations of acyl ghrelin are not increased during long-term fasting. Mboat4 mRNA levels are negatively correlated with des-acyl ghrelin concentrations in blood, leading to our hypothesis that Mboat4 expression, or GOAT protein expression, directly or indirectly represses translational control of Ghrl. This conclusion is further supported by the increased plasma levels of desacyl ghrelin in the GOAT-knockout mice, in contrast to WT mice (Supplementary Fig. 2b).
Ghrl mRNA remains constant during prolonged fasting, but blood concentrations of des-acyl ghrelin double. As the latter seems to be a stable increase, the data suggest translational control of the message, with potential involvement of microRNA. The twofold decrease of Mboat4 mRNA after fasting does not result in lower acyl ghrelin concentrations in blood. Therefore, GOAT activity might not be a bottleneck for ghrelin acylation in ad libitum or fasting conditions. The measurement of GOAT activity and development of GOAT antibodies capable of determining amounts of the mostly intramembrane GOAT protein are necessary to confirm this theory.
Furthermore, we show that dietary lipids are directly used for ghrelin acylation, as heptanoyl ghrelin, a form of ghrelin not normally found, occurs when mice are fed triheptanoate and, likewise, more than double the concentration of octanoyl ghrelin occurs when mice are fed an MCT-enriched diet. Mice lacking GOAT have lower body weight and fat mass on an MCT-containing diet compared to WT mice, whereas mice transgenically overexpressing human ghrelin and GOAT show higher body weight and fat mass than WT littermates, proving a role for the endogenous GOAT-ghrelin system in the control of energy balance and adiposity. Further, we find that sufficient dietary supply of MCTs is crucial for ghrelin acylation, given that transgenic mice are not able to produce large amounts of octanoylated human ghrelin when fed regular chow. Notably, transgenic mice on regular chow show substantial amounts of inactive C2-acetyl–modified ghrelin in the absence of octanoylated ghrelin, suggesting that at least under these experimental conditions the GOAT fatty acid substrate for acylation, acetyl-CoA, is available for ghrelin acylation.
In humans, acylated ghrelin concentrations rise before a meal and drop afterward, thus suggesting a role for ghrelin as a hunger signal. However, a recent study showed that plasma concentrations of acyl ghrelin do not increase under conditions of prolonged fasting14. By establishing a new, superior sandwich ELISA with highly specific protection of acyl ghrelin against deacylation of the octanoyl side chain, the researchers corroborated the well-known ghrelin secretion pattern with surges before a meal and suppression after a meal for the inactive and the active form of ghrelin. Nevertheless, during prolonged fasting for 36 h or longer, plasma concentrations of acyl ghrelin were consistently at the basal level. The results of their study therefore suggested that ghrelin may be regulated by separate secretion and activation processes and probably acts more as meal preparation cue rather than as a meal initiation factor or a hunger signal that responds to persistent starvation. The interpretation of ghrelin as being a meal preparation factor preparing the organism to optimally metabolize and store energy could explain why we do not find increased food intake in the GOAT and ghrelin–transgenic mice. Additionally, gastric ghrelin might have more of a role in the regulation of energy metabolism in muscle and liver.
Along the line of thought that ghrelin is acting more as a meal preparation cue than as a hunger cue, prolonged food deprivation in our studies also did not lead to increased acyl ghrelin blood concentrations. In addition, neither gastric Mboat4 expression nor Ghrl mRNA levels were upregulated during fasting, as we would have expected on the basis of the traditional model of ghrelin function. Taken together, our results therefore do not point toward a prominent role for both Ghrl and Mboat4 mRNA expression as a key process for the generation of a hunger signal that indicates an empty stomach. However, ghrelin acylation and thereby activation seem to be influenced substantially by food intake and fatty acid composition of the ingested food. It remains unclear why endogenously derived fatty acids mobilized from adipose stores are apparently not used to massively acylate ghrelin during fasting. MCTs, the substrate for GOAT10, can principally be produced during metabolism of long-chain fatty acids by beta oxidation, where the long-chain fatty acid is shortened by two carbons with each cycle in an markedly rapid process15. Only recently, more data on modulation of fatty acid profiles have become available, but such profiles were usually measured after a substantial fast rather than between meals when acyl ghrelin concentrations would be affected in a considerable manner14. However, even after 12 h of fasting, MCFAs remain present in plasma15 and are not completely oxidized. We therefore conclude, on the basis of the existing body of published data in combination with our results, that ghrelin acylation and the secretion of acylated ghrelin probably represent two independent processes and that GOAT-ghrelin might act as a lipid sensor that is activated when certain fatty food is consumed.
Notably, ghrelin-secreting cells in the gastrointestinal tract are of the closed type and of the open type, with the open-type cells more abundant in the duodenum where fatty acids are absorbed. Thus, it is possible that dietary lipids are taken up by the open-type ghrelin cells and directly used for ghrelin acylation. Follow-up studies providing improved insight into such mechanistic details could help to understand and more efficiently explore the rapid metabolic benefits of some bariatric surgeries, which are becoming increasingly popular for the treatment of severe obesity with type 2 diabetes mellitus.
Two recently published studies also highlight the emerging importance of dietary fat as a substrate for endogenous modulators of energy balance4, 16. The lipid mediator oleoylethanolamide and its precursor NAPE both affect food intake and energy metabolism using dietary fat as a substrate, and an acyltransferase is involved in oleoylethanolamide activation4, 16. MCTs show a number of particular features: they are hydrolyzed much more rapidly than long-chain triglycerides in the intestinal lumen, the presence of pancreatic enzymes and bile salts are not required for their absorption and they are transported via the portal vein as free fatty acids bound to albumin, whereas long-chain triglycerides must undergo esterification and chylomicron formation and are transported in the lymph. In nature, MCTs are found in milk along with long-chain fatty acids, and because MCTs are more efficiently digested, some infant formulae are enriched with MCTs. For the same reason, MCT therapy is indicated for individuals with malabsorption17. Our findings indicate that the orexigenic, positive energy balance–promoting GOAT-ghrelin system is triggered by dietary MCTs and offer a new perspective for the potential role of MCTs in milk and infant formula that deserves further study.
Coconut is a natural source of large amounts of MCTs. Studies conducted in the Polynesian islands, where meals are composed of up to 60% from coconut, show high prevalence of obesity and dyslipidemia, which can be linked to the high intake of coconut products18. Although ghrelin abundance was not measured in this population, it can be assumed that the extreme levels of MCT intake triggers ghrelin acylation, possibly leading to increased peripheral lipid storage and hypercholesterolemia18, two phenomena that are associated with ghrelin signaling6, 19.
The new model we propose, with GOAT-ghrelin as a lipid sensor, seems logical when reviewing the main functions of ghrelin; namely, lipogenesis and the triggering of growth hormone release, two features that seem to be counterintuitive during hunger. It would make more sense for an organism to thrive and support fat storage while nutrients are coming in. Therefore, although more studies are needed to completely dissect acyl from des-acyl ghrelin functions or ghrelin activation from ghrelin secretion processes, we propose that ghrelin's predominant physiological function may not necessarily, or at least not exclusively, be that of a hunger signal that reflects an empty stomach as previously assumed. On the basis of the data presented here, it seems more likely that the GOAT-ghrelin system acts as a nutrient sensor by using readily absorbable MCFAs to signal to the brain that high caloric food is available, leading to optimization of nutrient partitioning and growth signals (Supplementary Fig. 7).
Our findings highlight the importance of identifying the physiological role of 'inactive' des-acyl ghrelin. The fact that des-acyl ghrelin seems to be secreted in highly regulated fashion in response to caloric deprivation (Fig. 1b and ref. 14) suggests the likely existence of a yet to be discovered physiological function, which, on the basis of indirect evidence, could be related to glucose metabolism20 and might involve an additional ghrelin receptor19.
Finally, gastrically expressed GOAT offers an accessible and elegant drug target for the treatment of metabolic diseases. As loss- and gain-of-function models prove its role in diet-induced adiposity regulation, GOAT modulators may be potential new antiobesity drugs or anticachexia therapeutics.
All studies were approved by and performed according to the guidelines of the Institutional Animal Care and Use Committee of the University of Cincinnati and the Eli Lilly Company. For the methods used in this study, please see Supplementary Methods.
Note: Supplementary information is available on the Nature Medicine website..
* In the version of this article initially published, the fourth condition from the top in the key to the bar graphs in Figure 2c was mislabeled as 'mC8'. The correct label is 'hC8'. The error has been corrected in the HTML and PDF versions of the article.
