Hepatocyte expression of the micropeptide adropin regulates the liver fasting response and is enhanced by caloric restriction

The micropeptide adropin encoded by the clock-controlled Energy homeostasis-associated gene is implicated in the regulation of glucose metabolism. However, its links to rhythms of nutrient intake, energy balance and metabolic control remains poorly defined. Using surveys of Gene Expression Omnibus datasets, we confirm that fasting suppresses liver adropin expression in lean mice. However, circadian rhythm data are inconsistent. In lean mice, CR induces bouts of compulsive binge feeding separated by prolonged fasting intervals, increasing NAD-dependent deacetylase sirtuin-1 signaling important for glucose and lipid metabolism regulation. CR upregulates adropin expression and induces rhythms correlating with cellular stress-response pathways. Furthermore, adropin expression correlates positively wth phosphoenolpyruvate carboxokinase-1 ( Pck1 ) expression, suggesting a link with gluconeogenesis. Our previous data suggest that adropin suppresses gluconeogenesis in hepatocytes.

pathways. Furthermore, adropin expression correlates positively wth phosphoenolpyruvate carboxokinase-1 (Pck1) expression, suggesting a link with gluconeogenesis. Our previous data suggest that adropin suppresses gluconeogenesis in hepatocytes. Liver-specific adropin knockout (LAdrKO) mice exhibit increased glucose excursions following pyruvate injections, indicating increased gluconeogenesis. Gluconeogenesis is also increased in primary cultured hepatocytes derived from LAdrKO mice. Analysis of circulating insulin levels and liver expression of fasting-responsive PKA signaling pathways also suggests enhanced responses in LAdrKO mice during a glucagon tolerance test (250 µg/kg ip.). Fasting-associated changes in PKA signaling are attenuated in transgenic mice constitutively expressing adropin, and in fasting mice treated acutely with adropin peptide. In summary, hepatic adropin expression is regulated by nutrient-and clock-dependent extrahepatic signals. CR induces pronounced post-prandial peaks in hepatic adropin expression. Rhythms of hepatic adropin expression appear to link rhythms in energy balance and cellular stress to the intracellular signal transduction pathways that drive the liver fasting response.
The human genome contains thousands of short open reading frames (sORFs) predicted to encode 'micropeptides' of <100 amino acids (1)(2)(3). The translation and functions of most have not been studied. Early studies indicate that micropeptides can function as 'fine-tuners' of homeostatic processes, and are potential leads for developing treatments against the metabolic diseases of obesity (2)(3)(4).
Adropin is a micropeptide translated from a unique highly conserved sORF in mammalian genomes (5). The 76 amino acid sORF (adropin  is situated in exon 2 of the Energy homeostasis associated (ENHO) gene (5,6). Bioinformatic analysis suggest adropin 1-33 is a signal peptide targeting the secretory pathway (5,7). Adropin  is predicted to be released by proteolysis and is biologically active in rodents and cultured cells (5,8). Experiments using RNAi knockdown suggest the orphan G protein-coupled receptor GPR19 is required for biological responses to adropin   (9)(10)(11). However, the coupling of adropin  with GPR19 is controversial (12). Yeast 2 hybrid protein screens also identified an interaction between adropin  and NB3/Contactin6, a non-canonical membrane-tethered Notch1 ligand, that may regulate neural development (13). The identity, binding and signal transduction characteristics of cell-surface receptor(s) that mediate adropin's physiological actions remain under investigation.
The discovery of adropin was driven by liver transcriptomic data from C57BL/6J (B6) mice (5). Adropin expression is suppressed by diet-or genetically induced obesity at least in part by an increase in miR29b-directed degradation of Enho mRNA (5,14,15) . The link to miR29b is interesting as the miR29 family also regulate liver lipid metabolism (16) and liver insulin sensitivity (17). Early studies of liver adropin expression identified rapid changes due to energy deficit and food intake (5). Enho transcription also appears to be controlled by regulatory components of the circadian clock driving rhythms in glucose and lipid metabolism (18). However, the impact of energy balance on liver adropin expression and its significance is controversial. For example, liver adropin expression in teleosts (bony fish) increases with fasting (19). Moreover, whether the sORF encoding adropin is translated in liver is unclear (13). And while caloric-restriction (CR) is an energy deficit, it increases plasma adropin concentrations in mice (20).
Several studies have implicated adropin in the control of glucose metabolism. In mouse and cellbased models, adropin regulates glucose production by the liver and fuel selection (glucose vs. fat oxidation) in cardiac and skeletal muscles (5,6,11,(21)(22)(23)(24)(25)(26)(27). These results linking adropin to metabolic processes appear clinically relevant. Nonhuman primate (NHP) models are more closely related to humans compared to rodents (28). In rhesus macaques fasting plasma adropin concentrations correlate with indices of liver glucose and lipid metabolism (29)(30)(31). The expression of ENHO in NHP liver suggests daytime peaks that anticipate nutrient intake and coregulation with enzymes involved in carbohydrate and lipid metabolism (29). Plasma adropin concentrations in NHP also appear to peak during the daytime (18). Plasma adropin concentrations in NHP correlate with indices of metabolic dysregulation that indicate increased risk for type 2 diabetes and cardiovascular disease in humans (29). The relationship between plasma adropin concentrations and metabolism in humans is less clear. However, several groups have observed correlations with various indices of glucose and hepatic lipid metabolism (18,(32)(33)(34)(35)(36)(37)(38).
The current investigation had two objectives. First, liver transcriptome (RNA-seq) data were used to further define the relationship between energy balance, circadian rhythm and fed-fasting cycles on hepatic Enho expression. Whether the suppression of adropin expression in liver has a causal relationship with the dysregulation of carbohydrate and lipid metabolism in obesity is not known. Thus, the second objective was to investigate the phenotype of liver-specific adropin knockout (LAdrKO) mice developed using B6 mice with a floxed adropin coding sequence (6).
There were significant differences in Enho expression between ZTs (p<0.05) in this study, but no interaction between ZT and genotype (Suppl. Fig S3B). Caloric restriction (CR) upregulates and enhances rhythms in liver adropin expression. Mice will exhibit feeding behavior throughout the light and dark cycle when given free access to food. Subjecting mice to CR induces compulsive 'binge' feeding 90% of the calories provided are consumed within 1-2 hr leading to prolonged intervals of fasting (44)(45)(46). The current study used open access transcriptome data from mice exhibiting profound changes in liver gene expresion, nicotinamide adenine dinucleotide (NAD) metabolism, and protein acetylation (GSE93903) (47). Liver samples were collected at 4h intervals from 'young' (19-29 weeks of age) and 'middle-aged' (55-69 weeks of age) B6 mice of mixed sex (n=3-6/group). Control mice were provided 'normal diet' (ND) ad libitum, other mice were adapted to a CR protocol providing 70% of normal calories at ZT12 for 25 weeks. Enho expression data were extracted and analyzed using a 3-way ANOVA (fixed variables: feeding condition, age, ZT There was a significant interaction between feeding condition (ND vs. CR) and ZT (p=0.001). CR induced a peak in Enho expression 8-12h after food presentation (ZT20 and ZT24) (Fig. 1A, B). Post hoc analysis using ND or CR data alone indicated significant differences of expression between ZT in the CR group only (p<0.05) (Fig.  1B). Liver adropin expression correlates with cellular stress responses. Two complementary studies were performed to identify metabolic pathways correlating with low or high Enho expression. In the first study, a correlation matrix was computed using a correlation co-efficient of 0.6 as a cut-off. This approach identified 64 genes positively correlated with that of Enho, and 253 genes whose expression correlated negatively. The biological functions of these genes were queried against the ToppGene server (https://toppgene.cchmc.org/), which returned the lists of enriched genes and pathways under a low false discovery rate (FDR) <= 0.05 cutoff level (suppl. Fig. 4A, B).
Genes positively associated with Enho regulate enzymatic protein cleavage/degradation and amino acid biosynthesis. Taking a candidate gene approach, a strong positive correlation was observed between Enho and phosphoenolpyruvate carboxykinase 1 (Pck1), the main control point in the regulation of gluconeogenesis induced during energy deficit (Fig. 1C, D). Other genes correlating positively associate with insulin sensitivity and immune responses (Fig. 1D). Genes negatively associated with adropin are involved role in triglyceride metabolism (Fig. 1C, E ). Increases in Enho expression with CR thus correlate with the suppression of lipogenic pathways that occurs with CR (47). Other genes of interest that exhibit strong negative correlations are involved in glycogen synthesis, glycolysis and appetite control (Fig. 1C).
A limitation of this approach is the arbitrary selection of cut-off values, and that it only identifies "enriched" but not "depleted" genes. To further validate the results and allow unbiased discovery of enriched and depleted genes, we performed a genome-wide gene set enrichment analysis using the "fgsea" package in R as described above. Four independent studies were performed using (i) all 5529 biological gene sets, and its subsets belonging to (ii) REACTOME, (iii) KEGG or (iv) the remaining lists of the genes.
The top 20 enriched (NES > 0) and or depleted gene lists/pathways are shown in suppl. Fig. 4C-E. This analysis also found that liver Enho expression is incompatible with lipid and triglyceride metabolism. It also correlates positively with amino acid metabolism, translation and ribosomal functions. Interestingly, Enho expression also correlates with genes that respond to viral infection, supporting a role as part of the cellular stress response.
To further investigate the relationship between liver Enho expression and stress responses, we analyzed data from B6 mice treated with acetaminophen, isoniazid, or paraquat (GSE51969) (48). Acetaminophen and paraquat trigger production of reactive oxygen species, oxidative stress and mitochondrial dysfunction (49,50). Isoniazid-induced liver toxicity is still not well understood, but also result from mitochondrial dysfunction and inflammatory responses (51). Treatment with acetaminophen or paraquat resulted in dose-dependent increases in liver Enho expression (Fig. 1F).

Deletion of liver Enho expression.
To investigate the physiological functions of liver adropin, the adropin 1-76 sORF in exon 2 of the Enho gene was targeted using the Cre/LoxP system ( Fig.  2A) (6). Expression in hepatocytes was suppressed using B6.Cg-Speer6-ps1 Tg(Alb-cre)21Mgn /J transgenic mice (52). Liver Enho gene expression was reduced ~80% in LAdrKO mice relative to Enho fl/fl 'wild type' (WT) controls; expression in other tissues was normal (Fig.2B). Adropin immunoreactivity is detected in crude brain lysate by western blot (Fig. 2C) (53); adropin immunoreactivity could be detected using immunoprecipitation in WT but not LAdrKO liver samples (Fig. 2C). LAdrKO mice have normal body mass and composition determined by NMR (Fig. 3A,B). Liver weight and histology were also normal in LAdrKO (data not shown), indicating adropin is not essential for normal liver development. Adropin expression is highest in the nervous system (5,13,(53)(54)(55)(56), with low expression in liver. Fasting plasma adropin concentrations were similar in LAdrKO and control mice, suggesting compensation from adropin secreted by other tissues (data not shown).
While gene expression data suggest an association with stress responses and lipid metabolism, there was no evidence of dyslipidemia or liver injury in LAdrKO mice fed low or high-fat diets (data not shown). Expression of genes involved in de novo lipogenesis and triglyceride production, circulating lipids and plasma markers of liver injury (ALT, AST) were normal in LAdrKO mice (data not shown).

Increased glucose production in LAdrKO mice.
Based on data showing that adropin  suppresses gluconeogenesis in primary cultured hepatocytes (27,57,58), we hypothesized that suppressing liver adropin expression would dysregulate glucose metabolism. As predicted , glucose excursions following ip. injections of pyruvate in male and female LAdrKO mice were more pronounced ( Fig.  3C-G). A modest but significant impairment of glucose tolerance was also observed in females ( Fig. 3H-L). Expression of Pck1 was increased in LAdrKO mice of both sexes compared to agematched controls (Suppl. Fig S5A). The effect of genotype on Pck1 expression cannot be attributed to differences in fasting-responsive pancreatic hormones. Fasting plasma glucagon and insulin levels were normal in LAdrKO mice (Suppl. Fig  S5B and data not shown). Endogenous glucose production was also increased by approx. 1/3 in hepatocytes cultured from LAdrKO mice (Fig. 4A).
The liver gluconeogenic response to fasting is regulated by several signal transduction pathways affecting gene transcription. The NAD-dependent deacetylase sirtuin-1 (SIRT1) has been implicated in the control of gluconeogenesis (59,60). The transcriptional co-activator Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1A) has a critical role in the liver response to fasting, promoting gluconeogenesis and lipid catabolism (61,62). While both SIRT1 and PGC1A expression increased with glucagon treatment, this response was not affected by genotype ( Fig. 4D and data not shown).

Adropin-deficiency enhances hepatic responses to glucagon in vivo.
The primary cultured hepatocyte data suggest an enhanced response of gluconeogenic enzymes to glucagon. Whether liver adropin deficiency alters glucagon signaling in vivo was investigated using a modified glucagon tolerance test (63). Food was removed for 1h, and mice then administered an ip. injection of glucagon (250 µg/kg). Tissue and blood samples were collected at baseline (t=0 min), and then after 60 or 180 min. Insulin and glucose data were assessed using a 2-way ANCOVA (fixed variables: genotype, treatment; covariates: age, body weight for glucose, age, body weight and glucose for insulin). Age and body weight data for the mice used in the experiment are provided as supplemental information (supplemental information, table 1). The groups were evenly matched for age and body weight.
Glucagon treatment significantly increased blood glucose concentrations (p<0.001) and plasma insulin concentrations (p<0.05) ( Fig. 5A-D). While there was no effect of genotype, there was a trend (p=0.09) for an interaction between treatment and genotype for insulin values. Depletion of liver glycogen due to increased glycogenolysis was not significantly different between genotype (Supplemental Fig. S6).
Glucagon receptors (GCGR) couple to G protein complexes containing stimulatory (Gs) subunits (64). GCGR activation increases adenylate cyclase activity, accumulation of cAMP and activation of cAMP-dependent protein kinase A (PKA) signal transduction (64). While the specific mechanism remain unclear, adropin 34-76 appears to suppress cAMP-dependent PKA signal transduction in primary cultured hepatocytes (57). Analysis of PKA signaling using a PKA substratespecific antibody indicated an enhanced response in livers of LAdrKO mice at t=60 and t=180 (Fig.  5E,F). Phosphorylation of CREB on Ser 133 , a PKA substrate, also appeared to be higher at t=60 and t=180 min (Fig. 5E,F). Phosphorylation of PKA on Thr 197 in the activation loop of the catalytic subunit is required for normal function (65). Glucagon treatment increased Thr 197 phosphorylation in this study, and this effect was prolonged in LAdrKO mice (Fig. 5E,F).
SIRT1 protein expression was increased in LAdrKO mice compared to controls at baseline and t=60 min (Fig. 5E, F). HNF4A and FOXO1 protein levels were also increased in LAdrKO, particularly at baseline and t=60 min (Fig. 5E, F and data not shown). However, while glucagon treatment appears to increase PGC1A protein irrespective of genotype, overall expression levels were lower in LAdrKO mice (Fig 5E, F).
At the level of gene expression, there was an effect of genotype on Pgc1a mRNA expression (25% increase), but no effect of glucagon treatment (Fig. 6A). While glucagon treatment appeared to increase the expression of SIRT1 (Fig. 6B) and HNF4A (Fig. 6C), the differences were not statistically significant. Most of the genes involved in gluconeogenesis and glycolysis were expressed at similar levels (Suppl. Fig S7).
Activation of Protein kinase B (PKB)/AKT is essential for insulin to reduce glucose production by the liver (66). Adropin 34-76 treatment improves hepatic insulin sensitivity in diet induced obese B6 mice, activating AKT as indicated by phosphorylation on Ser 473 (8,57). Ser 473 phosphorylated AKT expression was lower in livers of LAdrKO mice, irrespective of glucagon challenge (Fig. 5E). Constitutive adropin expression dysregulates the liver fasting response. We next compared PKA signaling in transgenic mice constitutively expressing adropin (AdrTG) and WT littermates in fed and fasted conditions. There was a trend (p<0.1) for highr Enho expression in AdrTG, however this relatively shortterm fast did not significantly affect liver Enho expression (Fig. 7A). It did however increase Pck1 expression nearly two-fold, irrespective of genotype (Fig. 7A). There was an increase in Thr 197 phosphorylation of PKA, as well as in the phosphorylation of several PKA substrates including CREB (Fig. 7B, C). This response was not observed in AdrTG mice. This phenotype was not limited to PKA signaling, as fasting-induced increases in HNF4A and FOXO1 were also prevented in AdrTG mice (Fig. 7B, C). However, this deficit does not translate to major changes in circulating insulin or glucagon concentrations in fed or fasted conditions (Fig. 7D, E). Fasting had no significant effect on circulating glucagon concentrations. This finding is consistent with a previous study reporting that fasting of 24 hr is required to increase glucagon concentrations in lean B6 mice (67). However, there was a trend (p<0.1) for lower plasma glucose levels in AdrTG mice (Fig. 7F). Administration of adropin  to fasted mice resulted in a similar outcome, disrupting the normal responses of PKA signaling to fasting (Fig.  8A, B).
The AdrTG transgene uses a ubiquitously active promoter that will induce expression in the pancreas. While circulating insulin levels appearred to be normal, we investigated whether adropin regulates glucose-stimulated insulin secretion. However, no significant effect was observed in response to 10 ng/ml adropin 34-76 (data not shown).

DISCUSSION
These results support the hypothesis that adropin expressed by hepatocytes is a critical modulator of the liver fasting response. Open access liver transcriptome data indicate that liver adropin expression is labile in response to transitions between different states of energy balance. They also indicate that nutrient intake and cellular stress are positive signals. Data from the phenotyping of LAdrKO mice indicate that adropin may function to regulate liver responses to fasting, acting to moderate glucagon action in the liver. The absence of a response of cAMP-PKA pathways in liver to systemic signals of energy deficit is also remarkable. These observation suggest a heretofore unrecognised critical regulator of signaling pathways mediating adaptive changes in liver metabolism during transitions between energy deficit and nutrient sufficiency.
Our recent investigations of adropin function suggest that it acts as a co-regulator of liver glucose production (57). This previous study primarily examined the effects of pharmacological doses on liver metabolism in the context of diet-induced obesity and insulin resistance. However, the role of adropin in the liver in the homeostatic control of glucose production in physiological conditions has remained poorly defined. The current study therefore had two primary objectives. The first was to verify that liver adropin expression is indeed responsive to changes in energy balance. This goal was achieved by surveying open access transcriptome databases. One strength of this approach is that combining data from independent laboratories controls for stochastic variables such as diet, climate, housing and microbiota that can influence the metabolic phenotype of inbred mouse strains. The second goal was to determine whether loss of adropin expression in hepatocytes affects glucose homeostasis. This goal was achieved using the Cre/LoxP system to suppress adropin synthesis in the liver.

Liver adropin expression is sensitive to energy deficit and cellular stress responses.
Results from our analysis of transcriptome data verified that the adropin sORF is translated in liver. Moreover, liver adropin expression rapidly responds to changes in the balance of energy intake and expenditure. The expression profile of the Enho transcript in liver is however complex. Previous investigations suggested effects of energy balance, dietary macronutrients and regulatory elements of circadian clocks (5,18). The current study confirms that energy balance has a significant role in determining adropin expression in liver.
Our assessment of liver adropin expression in animals experiencing a range of conditions suggests increases in expression correlate with cellular stress-responses. This finding is consistent with other data suggesting a role for adropin in situations of oxidative stress. Adropin has been reported to protect against diet-induced nonalcoholic steatohepatitis by activating Nrf2 (68). Adropin also protects the vascular endothelium during ischemia/reperfusion, activating the reperfusion injury salvage kinase pathway. (69) While adropin expression appears to be incompatible with situations where liver de novo lipogenesis is stimulated, loss of adropin in liver does not appear to regulate in marked changes of lipid metabolism as assessed by measurement of gene expression or plasma lipid load.
Most investigations of adropin expression in liver have used mouse models. It is important to distinguish between experiments that investigated acute responses with those that examine the consequences of long-term exposure to diets differing in macronutrient composition. Studies investigating responses to fasting and re-feeding suggest that stimulation is most pronounced with high energy (high fat/high sucrose) diets, and that this difference can persist for up to 4 weeks (5). In rodent models, the consumption of high energy refined diets rich in dietary fats and refined sugars result in dysregulation of metabolism and oxidative stress (70). In the current study, the analysis of adropin expression in mice fed ad libitum or CR, and responses to drug-induced liver injury, suggests a positive correlation with cellular stress responses. At one level, increases in liver adropin synthesis may therefore be a response to nutrient load and oxidative stress.
On the other hand, chronic consumption of high fat/high sucrose diets resulting in hepatic insulin resistance suppresses liver adropin expression (5,14). Adropin expression is also suppressed in genetic models of obesity, a response that is prevented by reducing calories to delay obesity onset (5). These results indicate that chronic excess consumption and caloric overload lead to a dysregelation of adropin expression. A recent study provided a mechanism for the suppression of liver adropin expression in obesity. A micro-RNA (miR-29) upregulated in multiple rodent models targets the Enho transcript for degradation (15). Collectively, these suggest a decline in expression as obesity develops. Moreover, Enho expression is regulated at the level of transcription and mRNA stability. The phenotype of LAdrKO mice suggests that the suppression of adropin synthesis in liver is consequential and could contribute to the dysregulation of glucose metabolism.
The expression profile of adropin during CR suggests dynamic responses to shifts in energy balance. Expression clearly peaks after food presentation and late in the dark phase, and then declines during the period of fasting. Overall, the 24h average of adropin expression appears to increase during CR. Indeed, CR has also been reported to increase circulating adropin levels (20). These observations suggest that factors other then signals of energy balance drive adropin expression, and could indicate a role for adropin in signaling metabolic adaptations to CR. Indeed, consumption of a single large meal corresponding to 90% of daily caloric intake could place the liver under duress requiring adaptation, when compared to processing several small meals throughout the day.
The parent study used for comparing Enho expression in AL and CR conditions compared B6 mice from different age groups (47). At termination, mice in the 'young' age group were 10-12 months old, while mice in the 'old' age group were 18-22 months old. In human terms, 10-12 months is considered middle aged, while 18-22 months is considered old. Surprisingly, there was a modest (15%) but significant increase in liver Enho expression in the old group. The significance of this finding is not clear. There was however no significant impact of age group in the response of liver Enho expression to CR.
Two families of nuclear receptors may act to regulate Enho transcription (18). The retinoid acid receptor-related orphan receptors (ROR) are transcriptional activators that appear to activate Enho transcription. The transcriptional repressor Rev-erb was proposed to inhibit Enho transcript expression by competing for binding to ROR response elements. ROR and Rev-erb function as regulatory elements of the circadian clock, and also work to regulate rhythms in the expression of genes encoding enzymes involved in liver carbohydrate and lipid metabolism (71). We reported a modest rhythm in liver adropin expression with a peak late in the dark period in male AKR/J mice, (18). However, the evidence for a circadian profile in the GSE databases explored herein is weak. This discrepancy could be due to differences between mouse strains (AKR/J vs. B6), and in methods of measuring gene expression (transcriptomic data versus normalizing using a single housekeeping gene).
Adropin expression in the liver is markedly reduced in mice lacking functioning clocks. However, a functioning liver clock is not sufficient to restore normal expression. The liver clock is also not sufficient to restore expression of a large number of liver genes normally exhibiting circadian profiles (43). Rhythmicity in the expression of these genes has been suggested to be dependent on cues emanating from other tissues. It is therefore possible that suppression of adropin expression in clock deficient mice does not reflect regulation by the circadian clock expressed in hepatocytes.
B6 mice lacking the CLOCK transcription factor develop a metabolic disorder that includes obesity and hepatic steatosis (72). Suppression of Bmal1/Arnlt results in fasting hyperglycemia and insulin resistance (73)(74)(75), and an accelerated aging phenotype that includes increases in oxidative stress and inflammation (76). Suppression of liver adropin expression in clock mutant mice may therefore be a secondary consequence of metabolic dysregulation.

Adropin expressed in hepatocytes regulates the fasting response.
Targeted deletion of adropin from the mouse liver results in increased gluconeogenesis. Data supporting this conclusion came from live mice and from primary cultured hepatocytes. Surprisingly however, this phenotype appears to occur in the context of a paradoxical reduction of PGC1A expression. PGC1A protein levels were markedly reduced in liver of LAdrKO (Fig. 5E). PGC1A is a transcriptional co-activator that regulates the expression of genes involved in glucose and lipid metabolism during fasting. PGC1A expression in liver is increased in mouse models of type 2 diabetes (62). Conversely, suppression of PGC1A in liver using RNAi results in fasting hypoglycemia (77). The increases in glucose production thus cannot be explained by increased expression of PGC1A. How reduced PGC1A expression would impact other aspect of liver metabolism was also not determined.
In our laboratory conditions glucagon treatment increased PGC1A protein levels in vivo and PGC1A mRNA expression in primary cultured hepatocytes, irrespective of genotype. The lower levels of PGC1A protein observed in liver of LAdrKO mice are thus not explained by reduced mRNA levels. The mechanisms explaining lower PGC1A protein expression in the liver of LAdrKO mice were not determined in this study. However, PGC1A protein levels are regulated posttranslationally. For example, nuclear PGC1A is targeted for degradation in proteasomes after activation by acute oxidative stress (78).
It is also important to acknowledge that other interpretations of the pyruvate tolerance test results are possible (79). Pyruvate is a substrate in many tissues and could alter glucose concentrations by competing for oxidative processes. However, the PTT data are supported by other observations that are consistent with our interpretation of increased glucose production. LAdrKO mice exhibit changes in their responses of plasma insulin, blood glucose and expressions of liver signaling enzymes suggesting enhanced glucagon responses. In addition, primary cultured hepatocytes from LAdrKO mice also exhibit increased glucose production.
Another intriguing observation is that the responses of canonical liver signal transduction pathways regulating the fasting response are essentially absent in mice constitutively expressing adropin. The fasting response of the liver involving increases in HN4A and FOXO1 protein expression and activation of PKA were clearly suppressed in LAdrKO mice. A decline in AKT activity indicated by Ser 473 phosphorylation was also attenuated. Indeed, AKT activity appears to be lower in LAdrKO mice. This suggests that adropin expressed in hepatocytes is required to maintain normal AKT signaling. Moreover, this could also indicate a state of insulin resistance given AKTs role in mediating insulin action (66).
The changes observed in signal transduction pathways implicated in glycemic control did not affect circulating levels of pancreatic hormones or glucose. This finding has precedent, as several studies have suggested that mice can compensate for profound changes in liver signal transduction pathways. The control of hepatic glucose metabolism by insulin involves direct and indirect pathways (80). Constitutive activation of PKA in the liver results in fasting hyperglycemia and evidence of increased glycogenolysis and gluconeogenesis (81). However, suppression of PKA activity by over-expression of inhibitory subunits has no effect on fasting induced changes in Pck1, Pgc1A, G6pase or Gck expression; circulating levels of insulin, NEFA and TG were also normal in the fasted condition and after refeeding (82). Inhibition of liver PKA can result in lower glucose levels (80,82). Compensatory mechanisms thus allow for normal fasting and fed levels of glucose, insulin and lipids. A trend for lower fasting glucose was observed in AdrTG mice. However, whether this results from reduced glucose production or increased clearace was not determined.
The suppression of liver adropin expression with fasting suggests an association with catabolic processes. Declining adropin expression during fasting associates with suppression of lipogenesis and activation of glucose production. However, when data from mice fed AL and CR are viewed collectively, adropin expression correlates negatively with lipogenesis, and positively with glucose production. This comparison draws from data in mice responding to a range of conditions (fasting, normal voluntary feeding behavior, binge feeding leading to overconsumption). Adopin expression thus appears to associate with conditions where the liver could be producing glucose. Elevated adropin expression thus appears to correlate with situations in which glucose production is increased. In this situation, adropin may function to 'fine tune' the response of liver to signals regulating gluconeogenesis. It is however important to note that this correlation is limited to gene expression.
Dysregulation of liver gluconeogenesis is a major contributor to fasting and postprandial hyperglycemia observed in type 2 diabetes (83)(84)(85). This facet of type 2 diabetes has been attributed to insulin resistance in hepatocytes (84,85) and hyperglucagonemia resulting from dysregulation of pancreatic a-cells (86). The current results are therefore important for advancing understanding of a potential 'local modulator' of the livers responses to fasting signals. Moreover, they suggest suppression of liver adropin expression in obesity may contribute to dysregulation of signal transduction pathways that control glucose metabolism.
In summary, the current findings suggest that adropin expressed in the liver functions as a nutrient-sensing modulator of the fasting response (Fig. 9). The findings indicate a model where the response of the liver to "master regulators" of anabolic (insulin) and catabolic (glucagon) responses to changes of energy balance can be finetuned by a local micropeptide.

EXPERIMENTAL PROCEDURES
For animal studies and experiments using primary cultured hepatocytes, adropin 34-76 was purchased from ChinaPeptides (Shanghai, China) in batches of 100 µg; the lyophilized peptide was resuspended on the day of use in sterile 0.9% saline. For islet studies, adropin 34-76 was purchased from Phoenix Pharmaceuticals and was reconstituted in water. Animals. Experiments using mice were reviewed and approved by the Institutional Animal Care and Use Committees of Saint Louis University.
The mouse Enho gene was altered by introducing loxP sites to flank the ORF (Open Reading Frame) in exon 2. The neomycin cassette flanked by Frt sites was removed using ACTB-Flpe transgenic mice; the resulting strain (Enho fl/fl ) was crossed with mice expressing Cre under the control of an albumin promoter/enhancer [B6.Cg-Tg(Alb-cre)21Mgn/J; also known as Albumin-Cre]. The Albumin-Cre transgene was crossed onto the Enho fl/fl background; offspring used for these studies were produced by mating Enho fl/fl mice with Albumin-Cre; Enho fl/fl mice resulting in 50% Enho fl/fl mice (WT) and 50% Albumin-Cre; Enho fl/fl mice (LAdrKO). Body composition was analyzed by NMR using BioSpin LF50 Body Composition Analyzer (Bruker, Germany). Mice were housed under 12 hr L:D conditions with access to standard rodent chow and water. Tolerance tests using glucose, pyruvate or glucagon. Mice were fasted for 6 hr for glucose tolerance tests (GTT, glucose, 2 mg/g fat-free mass); 16 hr for pyruvate tolerance tests (PTT: sodium pyruvate from Sigma, 2mg/g fat-free mass), and 1hr for glucagon tolerance tests. Blood was collected by tail nicks at different time points post intraperitoneal injection for GTT and PTT. Blood glucose levels were monitored using OneTouch Blood Glucose Meters (LifeScan Europe, Switzerland) at the times indicated. Plasma insulin levels were measured using an Ultrasensitive Mouse Insulin ELISA kit (Crystal Chem, Downers Grove, IL). Glucagon was measured using an ELISA (Mercodia Developing Diagnostics). Primary hepatocyte culture. Primary hepatocytes were isolated using the protocol as previously described (57,87). Experiments were performed 12-16h after plating cells. Hepatocytes were washed twice with PBS and incubated in Hank's balanced salt solution (HBSS, pH 7.4) for 2 hr and then washed with fresh HBSS. Finally, cells were incubated for 3 hr in HBSS with 100ng/ml glucagon and 5mM pyruvate in presence or absence of 100nM freshly prepared adropin  . At the end of 3hr incubation, media was collected and assayed for glucose concentrations with a glucose-oxidase based glucose assay kit (Sigma GAGO kit). Rates of glucose production were expressed as µg glucose/mg cell protein/hr.

RNA extraction, cDNA synthesis and real-time quantitative RT-PCR.
The extraction of the total RNA from was performed by using kits from Applied Biosystems. cDNA was synthesized with cDNA reverse transcription kit from Quanta. PCR was conducted using a QuantStudio Realtime PCR machine, Applied Systems using primers from Integrated DNA Technology. Data were normalized using three reference genes (HPRT1, 36B4, PPIB) (88) Immunoblotting.
Standard immunoblotting procedures were performed as previously described (25,26,57). Lysate was quantitated using BCA and equal amount of protein was loaded and blotted using antibodies listed in table S2 in supplemental information. Mouse islet isolation. C57BL/6J male mice at 8 weeks of age were anesthetized with ketamine:xylazine (0.18 mg:0.012 mg per gram of animal) by intraperitoneal injection. Pancreas was perfused with collagenase (0.21 mg/mL of Liberase TL [Roche]) through the common bile duct prior to euthanasia (by incision of the chest cavity to produce a bilateral pneumothorax). After verification of death, pancreas was removed out of animal. Islets from three mice per experiment (6 independent experiments were conducted) were isolated and pooled after pancreas digestion (37ºC for 14 min), followed by Histopaque® 1077 (Sigma) density gradient separation and handpicked purification. Islets were incubated for 24 h in RPMI 1640 medium containing 11.2 mmol/L glucose, 10% FBS, 110 µg/mL sodium pyruvate, 110 U/mL penicillin and 110 µg/ml streptomycin.

Glucose-stimulated insulin secretion (GSIS).
Five islets per condition (in triplicate) were incubated for 24 h at 37°C in RPMI 1640 medium (containing 11.2 mmol/L glucose, 10% FBS, 110 µg/mL sodium pyruvate, 110 U/mL penicillin and 110 µg/ml streptomycin) under following conditions: Control (no addition of adropin 34-76 ); 10 ng/ml adropin  and 100 ng/ml adropin 34-76 R After 24 h of treatment, islets were washed for one hr by incubating with Krebs-Ringer-Hepes buffer (KRH in mmol/L: 137 NaCl, 4.8 KCl, 1.2 KH2PO4, 1.2 MgSO4, 2.5 CaCl2, 5 NaHCO3, 16 HEPES and 0.1% BSA) at 2.8 mmol/L glucose, and supernatant was eliminated. Islets were incubated for one hr with KRH 2.8 mmol/L glucose, followed for one-hr incubation at 16.7 mmol/L glucose. Supernatants were collected after each one-hr period. All incubations were performed at 37°C and 5% CO2. At the end of the experiment, islets were lysed in HCl-Ethanol. Supernatants and lysates were stored at -20°C for later insulin determination by a commercial ELISA (Merck-Millipore). Insulin secretion is expressed as a percentage of total insulin content (i.e. intracellular insulin + extracellular insulin after basal and GSIS incubation). After 24h incubation with or without adropin 34-76 , cell death was assessed by chemiluminescent quantification of adenylate kinase activity (Toxilight, Lonza Group) in media. Data were normalized to intracellular insulin content. A muscle lysate was used as a positive death control. All experiments were run in triplicate.

Analysis of open access datasets.
Enho expression data were extracted from GSE107787, GSE73554 and GSE51969 data sets and managed in .xl files. For pathway analysis, the public liver gene expression dataset, GSE93903, which contains 21261 probe ids corresponding to 20832 unique genes, was downloaded and analyzed using the Bioconductor in R. We first performed quality control analyses using the genefilter package and the results confirmed that the dataset has been log2 transformed and properly normalized.
Enho, measured with the 1428739_PM_at probe, showed an interesting, cyclic pattern of expression across the 103 samples, under different time points, ZT4, ZT8, ZT12, ZT16, ZT0, ZT20, and feeding condition. Enho thus showed a strong variation in the expression across different samples, providing a good opportunity to analyze Adropin-regulated genes and related pathways, both positively and negatively correlated with Enho expression. To best capture these genes and pathways,we used all samples to perform the pathway analysis.

Genome-wide gene set enrichment analysis based on a scaled correlation matrix.
Gene set enrichment analysis (GSEA) is an algorithm to evaluate the degree of enrichment or depletion of a given gene set relative to one of the two functional states that are distantly connected by a gradient of pre-ranked genes based a measurable activity. Typically, these states are defined by deferentially expressed genes. In our analysis, we adopted this algorithm to reveal the function of Enho. We defined the two states as the one associated with Enho expression and the other antagonistic/incompatible with Enho expression. We ranked the 20832 genes based on their Pearson Correlation coefficients as defined by Enho. This was further scaled to enhance the detection sensitivity. The pre-ranked matrix was then used to search for enriched or depleted gene sets of experimentally verified biological pathways from a total of 5529 downloaded Statistical analysis. All mouse data were analyzed using Microsoft Excel or Graph Pad Prism software. (SPSS to be included)All experiments were repeated thrice with similar outcome; one representative experiment is presented. Data are represented as Mean ± SEM. *P<0.05; **P<0.01; ***P<0.001.

DATA AVAILABILITY.
The transcriptome data used for these studies are open access datasets managed by the NCBI that were deposited by the authors of the published studies. All data are contained within this article and supplemental information. The data shown in panels A-C were from young (Y, 19-29 weeks of age) or older (O, 55-69 weeks of age) mice subjected to ND or CR conditions. Liver samples were collected at the zeitgeber times (ZT) indicated on the x-axis (n=3-6/group). Enho expression correlates positively with phosphoenolpyruvate carboxykinase (Pck1) which is rate limiting for gluconeogenesis (D) but is incompatible with expression of genes involved de novo lipogenesis such as fatty acid synthase (Fasn) (E). Note that the units of expression in panels D and E are z-scores used for the heatmaps shown in panel C, while in panels A and B Enho expression is shown as probe intensity. (F) Enho expression relative to vehicle control in B6 mice 24 hr after treatment with low, medium or high doses of acetaminophen (APAP at 169, 225 or 300 mg/kg), isoniazid (INZ at 22, 44 or 88 g/kg), or paraquat (PQ at 12.5,25 or 50 mg/kg) (n=5/group). Data in panels A-F were drawn from GSE93903, data in panel G are from GSE51969. Correlation coefficients are shown for all data, and for ND or CR separately. a p<0.05 vs. ZT12, ZT16; b p<=0.05 vs. ZT12, ZT16. In C, the heatmap was generated using z scores. In D and E, the values shown z scores. * p<0.05 vs. vehicle, ** p<0.01 vs. vehicle.    were collected at the times indicated. Blood glucose levels (A) and % change in blood glucose levels relative to baseline within genotype (B); plasma insulin concentration (C) and % change in plasma insulin concentration within genotype (D). (E) Western blot data showing changes in activity of fasting-responsive signaling pathways in liver at baseline (t=0), and 60 or or 180 min after glucagon injections. HSP90 was used as a loading control. Note the two bands recognized by the pCREB antibody; the upper band is pCREB-Ser 133 , the lower band is the phosphorylated form of a CREB-related protein (ATF-1). Panels to the right show scatterplots of quantified data expressed as a ratio; pCREB Ser 133 and pATK Ser 473 are normalized against total protein, other protein measured are normalized using HSP90. Phosphorylation of proteins with PKA consensus sequences following glucagon treatment was more robust in liver lysates from LAdrKO mice (bands used for quantitation are indicated by asterisks in panel E). Phosphorylation of Thr 197 of PKA required for activation remained elevated at the 180min timepoint. AKT activity was markedly reduced in LAdrKO, irrespective of glucagon treatment, indicated by low phosphorylation of Ser 473 . glucagon treatment increased expression of PGC1A and HNF4A. These effects on PGC1A and HNF4A protein levels were preserved in LAdrKO mice. However, PGC1A protein levels were on lower at all time points while HNF4A and SIRT1 expression were initially higher in LAdrKO mice.   The box and whisker graph below the plot titled "pPKA substrates" shows data for all bands grouped by genotype and treatment (fed or fasted ± adropin treatment). This graph shows the clear effect of fasting to increase the phosphorylation of PKA substrates, and that this effect is attenuated with adropin 34-76 treatment. For the experiment , 6 wild type littermates and 6 LAdrKO mice (males) were fed ad libitum (Fed, n=2) or fasted for 6h (n=4). Mice indicated by the "+" received a single ip. injection of 450 nmol/kg adropin  or vehicle (0.9% saline plus 0.1% BSA) 1h before tissue collection. Figure 9. Model integrating adropin expressed by hepatocytes into the control of glucose metabolism by pancreatic hormones. In this model, adropin expression in the liver is affected by dietary nutrients and cellular stress responses, and serves to moderate the response of the liver to regulatory inputs from the pancreas. Euglycemia requires the coordinated release of insulin (INS) and glucagon (GCG) from b-cells and a-cells located in pancreatic islets. INS and GCG released in a pulsatile secretory pattern (89) travel to liver sinusoids via the hepatic portal vein. In the liver, INS and GCG have opposing actions on glucose metabolism in hepatocytes. INS suppresses glucose production, while GCG enhances glucose production. The results of the current study and our previous publication (57) suggest adropin 34-76 expressed by hepatocytes suppresses glucose production in part by cross-talk with GCGR. However, a recent study suggests that adropin  suppresses hepatic glucose production independently of GCG (27). The current study indicates that adropin expressed by hepatocytes has a tonic inhibitory effect, and is required for normal regulation. Adropin may also act to enhance or complement INS signaling, particulary in situations of metabolic stress that can distrupt INS signaling (27,57). Suppression of adropin expression with prolonged fasting or metabolic dysregulation in situations of obesity and/or disruption of circadian rhythms may enhance GCG signaling while limiting INS action. However, it should be noted that both GCG and INS action in liver are disrupted in the context of diet-induced obesity (90). Both INS and GCG regulate glycogen and lipid metabolism via pathways that are not shown here; the role of adropin in modulating these actions require further investigation. It should also be noted that the model incorporates data from the current work and other sources (6,22,27,57).