Growth hormone acts on liver to stimulate autophagy, support glucose production, and preserve blood glucose in chronically starved mice

Significance

Nine million people starve to death annually. To understand starvation, we created a mouse model in which the mice receive a daily meal that contains 40% of normal calories. They consume each meal within 1 hour and fast for 23 hours. Normal mice maintain viable levels of blood glucose by increasing hepatic glucose production. Gluconeogenesis requires elevated plasma growth hormone (GH) and is accompanied by hepatic autophagy. GH-deficient mice become hypoglycemic, owing to decreased autophagy and glucose production. Here we show that mice lacking the liver GH receptor have reduced autophagy and become hypoglycemic. These data demonstrate that GH acts through its receptor in the liver to induce autophagy and enhance glucose production, maintaining blood glucose during famine.

Abstract

When mice are subjected to 60% calorie restriction for several days, they lose nearly all of their body fat. Although the animals lack energy stores, their livers produce enough glucose to maintain blood glucose at viable levels even after a 23-hour fast. This adaptation is mediated by a marked increase in plasma growth hormone (GH), which is elicited by an increase in plasma ghrelin, a GH secretagogue. In the absence of ghrelin, calorie-restricted mice develop hypoglycemia, owing to diminished glucose production. To determine the site of GH action, in the current study we used CRISPR/Cas9 and Cre recombinase technology to produce mice that lack GH receptors selectively in liver (L-Ghr^−/− mice) or in adipose tissue (Fat-Ghr^−/− mice). When subjected to calorie restriction and then fasted for 23 hours, the L-Ghr^−/− mice, but not the Fat-Ghr^−/− mice, developed hypoglycemia. The fall in blood glucose in L-Ghr^−/− mice was correlated with a profound drop in hepatic triglycerides. Hypoglycemia was prevented by injection of lactate or octanoate, two sources of energy to support gluconeogenesis. Electron microscopy revealed extensive autophagy in livers of calorie-restricted control mice but not in L-Ghr^−/− mice. We conclude that GH acts through its receptor in the liver to activate autophagy, preserve triglycerides, enhance gluconeogenesis, and prevent hypoglycemia in calorie-restricted mice, a model of famine.

Famine is a challenge faced periodically by all animal species. Survival requires metabolic adaptation to maintain blood sugar levels high enough to support brain function. Recent experiments in mice have demonstrated an essential role of growth hormone (GH) in this adaptation (1, 2). In these experiments, laboratory mice are fed a diet that contains only 40% of the calories that they normally consume (60% calorie restriction). Over several days the mice lose weight, and nearly all of their white adipose tissue is consumed, depriving them of energy stores. The mice are fed at 6 PM every day, and they consume all of their food within 1 h. They then fast for 23 h until the next feeding. During this fast, blood glucose levels fall from 110–130 to 50–80 mg/dL, a range that preserves normal brain function.

The requirement for GH in blood sugar maintenance was revealed in studies of another hormone, ghrelin, whose secretion from the stomach is stimulated by fasting. Ghrelin stimulates secretion of GH by the pituitary (3) via an initial interaction with the GH secretagogue receptor (also called the ghrelin receptor) in the hypothalamus (4). Ghrelin is unique because it contains a covalently attached octanoate that is essential for function (5). During successive days of 60% calorie restriction, the blood levels of ghrelin and GH rise progressively in the late afternoon, which is 20–24 h after the last meal (1, 2). The rise in GH is caused by the rise in ghrelin as indicated by the observation that the rise in GH is diminished in gene knockout mice that lack ghrelin or the enzyme ghrelin-O-acyltransferase (GOAT) that attaches the required octanoate (1, 2). After several days of calorie restriction in ghrelin-deficient mice, the reduction in GH leads to a failure to maintain blood glucose. By day 7, after the 23-h fast, blood glucose falls below 20 mg/mL, and the animals become moribund. Hypoglycemia in GOAT-deficient mice is prevented when GH is infused continuously during the 7-d calorie restriction (2). However, blood glucose is not raised in these animals when GH is injected acutely on day 7.

Glucose turnover studies indicate that hypoglycemia in the calorie-restricted, GH-deficient mice is caused by a reduction in glucose production rather than an increase in glucose clearance (2). Hypoglycemia on day 7 is prevented by repeated i.p. injections of lactate or alanine, which are gluconeogenic substrates, or octanoic acid, which is a source of energy (2).

On day 7, after the 23-h fast, livers of calorie-restricted, WT mice show prominent autophagic vacuoles as visualized by electron microscopy, and there is an accompanying increase in the LC3-II isoform, an indication of increased autophagy (6). In marked contrast, livers of GH-deficient mice show little evidence of autophagy. Acute injection of GH on day 7 of calorie restriction restores autophagy in the GH-deficient mice, but there is no increase in blood glucose, indicating that an acute increase in autophagy is not sufficient to maintain blood glucose under these conditions.

GH is known to act on many tissues. Among the most prominent are adipose tissue, where it increases lipolysis, and liver, where it increases gluconeogenesis (7, 8). To determine the relevant site of GH action during calorie restriction, in the current studies we have used gene knockout technology to produce mice with a selective elimination of the GH receptor (GHR) in liver or adipose tissue. Deficiency of the GHR in the liver reproduced the hypoglycemia seen in GH-deficient, calorie-restricted mice. In contrast, animals lacking the GHR in adipose tissue maintained blood glucose levels equivalent to WT mice. These data indicate that GH action in the liver is necessary to prevent hypoglycemia in fasted, calorie-restricted mice.

Results

To inactivate the Ghr gene selectively in liver, we used CRISPR-Cas9 technology to insert loxP sites into the introns flanking exons 4a and 4b of the Ghr gene in fertilized mouse eggs (Fig. 1A and SI Appendix, Fig. S1). Mice that were homozygous for the loxP sites were bred with mice that express Cre recombinase under control of the liver-specific albumin promoter to generate L-Ghr^−/− mice. Deletion of exons 4a and 4b produces a frameshift, eliminating 93% of the amino acids in the GHR, and is known to destroy its function (9, 10). To confirm the deletion of exons 4a and 4b in the liver, we performed PCR reactions on hepatic mRNA using primers P1 and P2 as indicated in Fig. 1A. mRNA from WT mouse liver yielded a PCR product of 1,054 bp (Fig. 1B). mRNA from L-Ghr^−/− livers gave a single band of 900 bp, as predicted from loss of the two exons. We subjected the PCR products to sequencing reactions that yielded a single sequence consistent with the intended deletion (Fig. 1C). Fig. 1D shows the complete amino acid sequence of WT GHR with the portion encoded by exons 4a and 4b shown in red and the cleaved signal sequence shown in purple. Fig. 1E shows the amino acid sequence encoded by the deletion-bearing mRNA. The WT sequence stops at amino acid 21 of the mature receptor. Thereafter, there is a frameshift that encodes 40 novel amino acids before termination (shown in green). The bulk of the extracellular domain has been eliminated along with the membrane spanning sequence (shown in yellow in Fig. 1D). RT-PCR measurements demonstrated that the loss of GHR mRNA containing exons 4a and 4b was confined to the liver in the L-Ghr^−/− mice (Table 1). Fig. 1F shows an immunoblot performed on membranes from mouse liver. The GHR band at 84 kDa is absent from the knockout animals. The other bands represent cross-reacting proteins.

Fig. 1.

Liver-specific deletion of mouse Ghr gene. (A) Amino acid sequence of mouse Ghr is encoded by exons 2–10 of the Ghr gene. The signal sequence (SS) and the transmembrane domain (TM) are denoted by purple and yellow boxes, respectively. CRISPR-Cas9 technology was used to insert one loxP site upstream of exon 4a and another loxP site downstream of exon 4b as described in SI Appendix, Fig. S1. (B) Detection of exon 4a and 4b deletion using PCR amplification of liver mRNA. Total RNA was isolated from livers of control (WT) and liver-specific Ghr knockout (L-Ghr^−/−) mice and subjected to RT-PCR with primers P1 and P2 (A). (C) DNA sequencing chromatograms of PCR products from B confirm that exons 4a and 4b of GHR transcripts are deleted in L-Ghr^−/− livers. (D) Amino acid sequence of full-length WT GHR protein. The signal sequence (aa 1–24) and transmembrane region (aa 274–297) are highlighted in purple and yellow, respectively. Amino acids encoded by exons 4a and 4b are highlighted in red. (E) Amino acid sequence of the predicted GHR protein after deletion of exons 4a and 4b. This deletion resulted in the splicing of exon 3 to exon 5, causing a frameshift and a premature stop codon. The predicted truncated GHR protein contains only the first 45 amino acids of the WT GHR (encoding exons 2 and 3) plus 40 unrelated residues (denoted in green italics). (F) Immunoblot analysis of whole-cell lysates prepared from WT and L-Ghr^−/− livers. Aliquots of liver whole-cell lysates (50 μg protein) were subjected to SDS electrophoresis and immunoblot analysis with 1 μg/mL goat anti-mouse GHR antibody as described in Materials and Methods. Arrow denotes GHR protein. The other bands represent cross-reacting proteins that are not affected by the deletion.

Table 1.

Relative amount of GHR mRNA containing exons 4a and 4b in tissues from control and tissue-specific Ghr knockout mice

Tissue	GHR mRNA (relative to control)
	L-Ghr−/− mice	Fat-Ghr−/− mice
Liver	0.003 ± 0.001****	0.91 ± 0.05
White adipose tissue	1.08 ± 0.07	0.005 ± 0.001****
Brown adipose tissue	1.12 ± 0.07	0.012 ± 0.002****
Skeletal muscle	0.88 ± 0.05	1.2 ± 0.19
Kidney	1.16 ± 0.04	0.93 ± 0.04
Heart	1.07 ± 0.08	0.68 ± 0.08*

Ad libitum-fed male control (Ghr^f/f) and the indicated male liver- or fat-specific Ghr knockout mice (8 wk old) were killed by cervical dislocation at 6 PM. The indicated tissues were removed, and total RNA was isolated and subjected to real-time quantitative PCR as described in Materials and Methods. The values for GHR mRNA containing exons 4a and 4b in the control mice were set to 1.0, and the values in the knockout mice are expressed relative to the control number. Each value represents mean ± SEM of five mice. The average Ct values for the control mice in the two experiments were as follows: liver, 24.5; white adipose tissue (epididymal fat pad), tissue 22.5; brown adipose tissue, 24.4; skeletal muscle, 27.1; kidney, 22.5; and heart, 24.9. These experiments were repeated two times with similar results. Asterisks denote level of statistical significance (Student’s t test) between control and knockout mice. *P = 0.029; ****P < 0.0001. Bold values highlight the marked reduction in GHR mRNA in livers of L-Ghr^−/− mice or in adipose tissue of Fat-Ghr^−/− mice compared with control mice.

The L-Ghr^−/− mice grew normally. At 8 wk of age, their body weights did not differ significantly from those of littermate controls. When subjected to 60% calorie restriction and studied 23 h after the last meal, blood glucose levels declined similarly in control and L-Ghr^−/− mice for the first 5 d (Fig. 2A). Thereafter, in the control mice, blood glucose stabilized in the range of 80 mg/dL. In the L-Ghr^−/− mice, blood glucose continued to decline each day, reaching a nadir in the range of 40 mg/dL on days 9–11. The body weights of the control and L-Ghr^−/− mice were initially similar, and they declined similarly over the 11 d (Fig. 2B). In both strains of mice, the fat mass declined to the range of 2% of body mass as determined by NMR (Fig. 2C). Plasma GH was elevated in the L-Ghr^−/− mice even before calorie restriction (day 0, Fig. 2E). On day 11, plasma ghrelin and GH levels rose markedly in control mice and were even higher in L-Ghr^−/− mice (Fig. 2 D and E).

Fig. 2.

Levels of blood glucose, ghrelin, and GH in control mice and liver-specific Ghr knockout mice (L-Ghr^−/−) subjected to 60% calorie restriction for 11 d. Littermate control mice and L-Ghr^−/− mice (male, 8 wk old) were housed individually and subjected to 60% calorie restriction as described in Materials and Methods. Each mouse was fed at 6 PM every day. (A) Blood glucose levels were measured within a Bayer glucometer each day at 5:30 PM before feeding. (B) Total fat mass at days 0 and 11 of calorie restriction. (C) Body weight at days 0 and 11 of calorie restriction. Plasma levels of (D) ghrelin and (E) GH. Blood samples were collected before the start of calorie restriction (day 0) or at 5:30 PM 11 d after the 60% calorie restriction. Plasma hormone levels of the indicated hormone were measured by ELISA. In A–E, each value represents mean ± range of data from six mice. Asterisks denote level of statistical significance (Student’s t test) between control and knockout mice: **P < 0.01; ***P < 0.001. Experiments in A–C and in D and E were done four and three times, respectively, with similar results.

GH is known to affect white adipose tissue where it stimulates lipolysis (7, 8). To determine whether selective loss of GH activity in adipose tissue would cause hypoglycemia in calorie-restricted mice, we bred our floxed Ghr mice with mice that express Cre recombinase under control of the fat-specific AdipoQ (adiponectin) promoter (Materials and Methods). RT-PCR measurements revealed a greater than 99% loss of GHR mRNA containing exons 4a and 4b in white and brown adipose tissue of the Fat-Ghr^−/− mice (Table 1). We also observed a small but reproducible decrease in GHR mRNA containing exons 4a and 4b in the heart, which likely reflects the mRNA in epicardial and pericardial fat. No changes in GHR mRNA were detected in liver, skeletal muscle, and kidney.

Fig. 3 shows an experiment in which we subjected L-Ghr^−/− mice and Fat-Ghr^−/− mice to calorie restriction along with littermate controls for each strain. Fat mass in the L-Ghr^−/− and Fat-Ghr^−/− mice was the same as that in littermate controls, and by 9 d it had declined to the range of 2% in all of the mice (Fig. 3 A and B). By day 11 the L-Ghr^−/− mice were hypoglycemic (Fig. 3C), whereas the Fat-Ghr^−/− mice had blood glucose levels that were the same as the controls (Fig. 3D). On day 11, plasma ghrelin and GH levels in the L-Ghr^−/− mice were elevated even more than in the controls (Fig. 3 E and G). In contrast, these hormones were elevated equally in the Fat-Ghr^−/− mice and the controls (Fig. 3 F and H).

Fig. 3.

Comparison of (A and B) fat mass and (C and D) blood glucose levels in liver-specific and adipose tissue-specific Ghr knockout mice subjected to calorie restriction and studied in the same experiment. Fat mass and blood glucose levels in L-Ghr^−/− and Fat-Ghr^−/− mice and their littermate controls (male, 7 wk old) were subjected to 60% calorie restriction as described in the legend of Fig. 1. For control and L-Ghr^−/− mice, body weights (mean ± SEM) at day 0 were 21.5 ± 1.2 and 21.4 ± 1.3 g, respectively. For control and Fat-Ghr^−/− mice, body weights (mean ± SEM) on day 0 were 22.3 ± 1.1 and 21.0 ± 1.6 g, respectively. Fat mass (A and B) was determined as described in Materials and Methods. Blood glucose levels (C and D) in the same mice were measured each day at 5:30 PM before feeding. Plasma levels of (E and F) ghrelin and (G and H) GH were measured at the start of calorie restriction (day 0) or 11 d after the 60% calorie restriction as described in Fig. 2. In A–H, each value represents mean ± range of data from five mice. Asterisks denote the level of statistical significance (Student’s t test) between control and knockout mice: ***P < 0.001. Experiments in A and C and in E and G were done four and three times, respectively, with similar results. Experiments in B, D, F, and H were done two times with similar results.

Hepatic triglycerides represent a store of energy that the liver might use to support gluconeogenesis during calorie restriction (11, 12). To determine the relation between hepatic triglycerides and blood glucose, we subjected mice to calorie restriction, killed cohorts after their 23-h fast on various days, and measured hepatic triglycerides as well as blood glucose. As expected, blood glucose declined more in calorie-restricted L-Ghr^−/− mice than in littermate controls over the 11-d period (Fig. 4A). Hepatic triglycerides fell in both groups, but the decline was greater in the L-Ghr^−/− mice (Fig. 4B). Fig. 4C plots the relation between blood glucose and hepatic triglycerides in each animal in this study. The data show a close correlation between blood glucose and hepatic triglycerides in both groups (correlation coefficient of 0.84).

Fig. 4.

Changes in blood glucose and hepatic TGs over 11 d in control and liver-specific Ghr knockout mice subjected to calorie restriction. Control and L-Ghr^−/− mice (male, 9 wk old) were subjected to 60% calorie restriction as described in Materials and Methods. (A) Blood glucose was measured at 5:30 PM (30 min before feeding) on the indicated day. (B) Hepatic TG content was measured in the same mice as in A. The mice were killed by cervical dislocation at 6 PM before feeding on the indicated day. The livers were excised and freeze-clamped in situ and stored at −80 °C. In A and B, each value represents mean ± range of data from six mice. Asterisks denote the level of statistical significance (Student’s t test) between the control and L-Ghr^−/− mice. **P < 0.01; ***P < 0.001. (C) Plot showing the relation between blood glucose levels and hepatic TG content in the 48 mice subjected to calorie restriction. R² value indicates goodness of fit for a positive linear relationship (Graphpad Prism). Experiments in A were done four times with similar results. Experiments in B and C were done two times with similar results.

Fig. 5 presents blood glucose and hepatic triglyceride measurements at different times during day 11 of calorie restriction. At 2 PM, which is 20 h after the last meal, blood glucose averaged 80 mg/dL in the L-Ghr^−/− mice as well as in the control littermates (Fig. 5A). Over the next 4 h, the control mice maintained a stable level of blood glucose, whereas the L-Ghr^−/− mice exhibited a precipitous fall. Similarly, hepatic triglycerides were the same in control and L-Ghr^−/− mice at 2 PM, but they subsequently fell to a greater degree in the L-Ghr^−/− mice (Fig. 5B). Fig. 5C demonstrates the strong correlation between blood glucose and hepatic triglycerides over this single day in control as well as in L-Ghr^−/− mice (correlation coefficient of 0.81).

Fig. 5.

Decline in blood glucose and hepatic TGs in control and liver-specific Ghr knockout mice on day 11 after calorie restriction. Control and L-Ghr^−/− mice (male, 9 wk old) were subjected to a 60% calorie restriction for 11 d as described in Materials and Methods. (A) Blood glucose and (B) hepatic TGs were measured in the same mice that were killed at the indicated time. In A and B, each value represents mean ± range of data from six mice. Asterisks indicate statistical significance at P < 0.001 (Student’s t test). (C) Relation between blood glucose and hepatic TGs in the 36 mice in this study. R² value indicates goodness of fit for a positive linear correlation (Graphpad Prism). These experiments were done two times with similar results.

We previously showed that repeated injections of lactate or octanoate can prevent hypoglycemia in calorie-restricted, ghrelin-deficient mice (2). To determine whether these injections restored hepatic triglycerides, we subjected L-Ghr^−/− mice to calorie restriction. On day 11, beginning at 2 PM, we injected the mice intraperitoneally at hourly intervals with saline, lactate, or octanoate. Blood glucose was measured before each injection. The mice were killed, and hepatic triglycerides were measured at 6 PM, 4 h after the first injection. As shown in Fig. 6A, blood glucose declined progressively in the saline-injected mice, and this fall was prevented by injections of lactate or octanoate. Both compounds raised hepatic triglycerides (Fig. 6B). Once again we observed a strong correlation between hepatic triglycerides and blood glucose (correlation coefficient of 0.825 in Fig. 6C). Although the data in Figs. 4–6 reflect a correlation between hepatic triglycerides and blood glucose, they do not indicate that the fall in hepatic triglycerides is the cause of the hypoglycemia in calorie-restricted mice (Discussion).

Fig. 6.

Prevention of hypoglycemia in calorie-restricted, liver-specific Ghr knockout mice by injections of lactate or octanoate. Control and L-Ghr^−/− mice (male, 9 wk old) were subjected to 60% calorie restriction for 11 d as described in Materials and Methods. (A) Beginning at 2 PM on day 11, each mouse was injected intraperitoneally at hourly intervals as indicated with either saline or one of the following: sodium lactate (2 mg/g body weight per injection) or sodium octanoate (0.5 mg/g body weight per injection). Each injection of saline, lactate, or octanoate (denoted by arrows) was delivered in a volume of 160–190 μL. Stock solutions of lactate (1.8 M) and octanoate (0.3 M) were dissolved in water and adjusted to pH 7.2 and pH 7.6, respectively. Blood glucose was measured immediately before each injection at the indicated time. (B) Hepatic TGs at 6 PM in the same mice as in A. The mice were killed at 6 PM on day 11, and the livers were freeze-clamped in situ, after which TGs were measured. In A and B, each value represents mean ± range of values from six mice. Asterisks denote statistical significance at P < 0.001 (Student’s t test) between lactate or octanoate and saline. (C) Relation between blood glucose and hepatic TGs in the 18 mice in this study. R² value denotes goodness of fit for a positive linear relationship (Graphpad Prism). These experiments were done three times with similar results.

To determine whether the hepatic GHR is required for the enhanced autophagy that was previously observed in calorie-restricted mice (6), we subjected L-Ghr^−/− mice and littermate controls to calorie restriction for 11 d and killed the mice at 5:30 PM after the 23-h fast. Livers were fixed, and random electron micrographs were printed. Autophagic vacuoles were counted by three independent observers who were blinded to the source of the specimens. Fig. 7 A and B show abundant autophagic vacuoles in representative micrographs of livers from two control animals. Fig. 7 C and D show markedly reduced vacuoles in micrographs from L-Ghr^−/− livers. Fig. 7E shows the results of the quantification by the three observers.

Fig. 7.

Electron micrographs of livers from calorie-restricted control and L-Ghr^−/− mice. Two littermate control and two L-Ghr^−/− mice (8 wk old) were subjected to 60% calorie restriction. On day 11 of calorie restriction, mice were perfused at 5:30 PM, after which the liver was fixed and processed as described in Materials and Methods. Fifty micrographs were obtained from random areas of (A and B) each control mouse and (C and D) each L-Ghr^−/− mouse. Representative pictures are shown. (Scale bar: 5 μm; magnification 440×.) (E) The number of autolysosomes per unit area was determined by three independent examiners who were blinded to the experimental groups. Each value represents mean ± SEM of data from 100 images. Asterisks denote level of statistical significance (****P < 0.0001) between control and L-Ghr^−/− mice (Student’s t test).

Discussion

The current paper establishes that the liver is the major site of GH action in preventing hypoglycemia in mice subjected to chronic calorie restriction. Previous data demonstrated a progressive rise in plasma GH in the late afternoon when mice have been subjected to 60% calorie restriction for several days (1, 2, 13). Gene knockout experiments revealed that the late afternoon increase in GH is caused by an increase in ghrelin. The increase in GH is essential to prevent severe hypoglycemia when the calorie-restricted animals have been without food for nearly 24 h. Severe hypoglycemia is caused by decreased gluconeogenesis, and it occurs only after body fat content has declined below 2% of body weight (2). Indeed, severe hypoglycemia was delayed by several days when the animals were previously fed a high-fat diet that expanded body fat stores and prolonged the time required for calorie restriction to lower these stores below 2% of body weight (2).

To determine whether GH exerts its antihypoglycemic effect by acting on the liver, in the current studies we used CRISPR/Cas9 and Cre recombinase technology to eliminate two crucial exons in the gene for the GHR in the liver. In these L-Ghr^−/− mice, plasma GH rose even more than in control mice during calorie restriction (Figs. 2 and 3). Nevertheless, the calorie-restricted L-Ghr^−/− mice developed late afternoon hypoglycemia (Figs. 2–6). In contrast, blood glucose was maintained at control levels in mice that lacked functional GHR in adipose tissue (Fig. 3).

In the current studies of L-Ghr^−/− mice, hypoglycemia took longer to develop (11 vs. 7 d) and was less severe on average (blood glucose about 40 vs. 20 mg/dL) than we observed in earlier studies of calorie-restricted ghrelin and GOAT-deficient mice. One possible explanation for this difference is that our previous calorie restriction studies used an F1 hybrid 129/C57Bl6 strain of mice (1, 2, 6, 13), whereas in the current study, the advent of CRISPR/Cas9 technology allowed us to create the L-Ghr^−/− genotype in pure C57Bl6 mice. However, we cannot exclude the possibility that the decreased hypoglycemia in L-Ghr^−/− mice compared with Goat^−/− and Ghrelin^−/− mice indicates that GH action in the liver accounts for most, but not all, of the protection from hypoglycemia during calorie restriction. It is possible that hypoglycemia would be more severe in L-Ghr^−/− mice if we blocked GH action in another tissue at the same time as the block in the liver. Blocking GH action in adipose tissue did not cause hypoglycemia itself, but it might have enhanced hypoglycemia if combined with the block in the liver.

In the current study, as well as in our previous experiments, we use young mice (9–11 wk old at time of killing) because the starting body fat levels are more uniform than those in older mice. Inasmuch as the time to hypoglycemia depends strongly on previous body fat stores, we routinely prescreen our mice before calorie restriction and discard animals with excess body fat >11%. At 9–11 wk, the L-Ghr^−/− mice have the same body weight and fat content as their control littermates. This observation is consistent with the observations of List et al. (10), who found that body weights in L-Ghr^−/− mice were normal until 6 mo of age. After this period the L-Ghr^−/− mice failed to grow as much as the controls.

We previously reported that livers of calorie-restricted control mice show profound autophagy as monitored by electron microscopy and by the level of LC3-II, an isoform that increases during autophagy (6). Fig. 7 shows that livers of L-Ghr^−/− mice failed to show such autophagy, indicating that autophagy is a direct result of GH action in the liver. Although hepatic autophagy may be necessary to maintain blood glucose in calorie-restricted mice, it is not sufficient, at least acutely. Thus, injection of GH into Goat^−/− mice in the afternoon of day 7 of calorie restriction restored hepatic autophagy but did not prevent hypoglycemia at 5:30 PM (6).

Although the correlation between the fall in hepatic TGs and the fall in blood glucose was striking (Figs. 5 and 6), it does not necessarily indicate causation. It is possible that the failure to maintain hepatic TGs and hepatic gluconeogenesis both result from another deficiency. In this regard, we previously found a decrease in hepatic ATP levels in calorie-restricted, ghrelin-deficient mice (6). The ATP deficit may well be responsible for the fall in hepatic TGs as well as the fall in gluconeogenesis. Although we did not measure hepatic ATP in the current study, the finding that hypoglycemia was prevented and hepatic TGs were restored by infusion of lactate and octanoate (Fig. 6), two potential sources of energy, supports the notion that a lack of ATP may underlie both abnormalities.

Considered together with our previous findings, the current data suggest that GH maintains blood glucose during famine through a direct action to maintain energy stores in the liver. Future experiments will be directed toward finding the mechanism for this hepatic energy conservation. All animal experiments described in this paper were approved and conducted under oversight of the University of Texas Southwestern Institutional Animal Care and Use Committee.

Materials and Methods

SI Materials and Methods include description of the following items: reagents; tissue-specific Ghr mice; measurements of body composition, blood glucose, plasma hormones, and liver triglycerides; electron microscopy; quantitative real-time PCR; and immunoblot analysis.