Preservation of liver protein synthesis during dietary leucine deprivation occurs at the expense of skeletal muscle mass in mice deleted for eIF2 kinase GCN2.

In eukaryotic cells, amino acid depletion reduces translation by a mechanism involving phosphorylation of eukaryotic initiation factor-2 (eIF2). Herein we describe that mice lacking the eIF2 kinase, general control nonderepressible 2 (GCN2) fail to alter the phosphorylation of this initiation factor in liver, and are moribund in response to dietary leucine restriction. Wild-type (GCN2(+/+)) and two strains of GCN2 null (GCN2(-/-)) mice were provided a nutritionally complete diet or a diet devoid of leucine or glycine for 1 h or 6 days. In wild-type mice, dietary leucine restriction resulted in loss of body weight and liver mass, yet mice remained healthy. In contrast, a significant proportion of GCN2(-/-) mice died within 6 days of the leucine-deficient diet. Protein synthesis in wild-type livers was decreased concomitant with increased phosphorylation of eIF2 and decreased phosphorylation of 4E-BP1 and S6K1, translation regulators controlled nutritionally by mammalian target of rapamycin. Whereas translation in the liver was decreased independent of GCN2 activity in mice fed a leucine-free diet for 1 h, protein synthesis in GCN2(-/-) mice at day 6 was enhanced to levels measured in mice fed the complete diet. Interestingly, in addition to a block in eIF2 phosphorylation, phosphorylation of 4E-BP1 and S6K1 was not decreased in GCN2(-/-) mice deprived of leucine for 6 days. This suggests that GCN2 activity can also contribute to nutritional regulation of the mammalian target of rapamycin pathway. As a result of the absence of these translation inhibitory signals, liver weights were preserved and instead, skeletal muscle mass was reduced in GCN2(-/-) mice fed a leucine-free diet. This study indicates that loss of GCN2 eIF2 kinase activity shifts the normal maintenance of protein mass away from skeletal muscle to provide substrate for continued hepatic translation.

Mammals require an adequate supply of dietary essential amino acids (i.e. those that cannot be synthesized de novo) to grow and thrive. Deficiency in one or more essential amino acids in the diet of rodents results in anorexia and loss of body protein (1,2). Loss of body proteins in growing animals is due in large part to a depression in protein synthesis at the initiation stage of mRNA translation (3,4).
Translation initiation is an intricate process coordinated by a family of protein factors called eukaryotic initiation factors (eIFs) 1 (for review, see Ref. 5). eIF2 plays a central role in regulating global translation rates via phosphorylation of serine 51 on its ␣ subunit (6). Under diverse conditions of cell stress, phosphorylation of eIF2␣ inhibits guanine nucleotide exchange on eIF2 and restrains mRNA translation. At the same time, the phosphorylation of eIF2␣ serves to enhance gene-specific translation important for regulating the expression of genes that manage cellular insults (7)(8)(9). A family of four known protein kinases have been described to phosphorylate eIF2␣ in response to different cell stressors (6, 7, 10 -12). Coined the "integrated stress response" (13), this process places eIF2␣ phosphorylation as the decisive event in regulating the ability of the cell to cope with environmental stress.
One of the eIF2␣ kinases, termed general control nonderepressible 2 (GCN2), is activated under conditions of nutrient deprivation or UV irradiation (14 -16). The role of GCN2 in response to amino acid starvation has been characterized extensively in yeast (10,17,18). In yeast, starvation for any single amino acid results in intracellular accumulation of uncharged tRNA, which binds to GCN2 on a domain homologous to histidyl-tRNA synthetases. This binding event triggers a conformational change in GCN2, enhancing phosphorylation of eIF2␣. Whereas eIF2␣ phosphorylation serves to slow global rates of translation, at the same time it enhances translation of the transcriptional activator, GCN4. Increased levels of GCN4 leads to induction of genes that encode amino acid biosynthetic enzymes and other related metabolic proteins in an attempt to correct for the nutritional deficiency. Whereas GCN4 is not present in mammalian systems, several transcription factors have been shown to be induced in response to phosphorylation of eIF2␣ by amino acid starvation in mammalian cells, including members of the activating transcription factor and CCAAT/ enhancer-binding protein (C/EBP) family of transcriptional transactivators (8,9).
Another signaling event altered in response to nutrients involves the mammalian target of rapamycin (mTOR) protein kinase, a downstream effector of the phosphatidylinositol 3-kinase/Akt (protein kinase B) signaling pathway. Signaling downstream of mTOR is implicated in many aspects of cell growth including cell cycle control and ribosome biogenesis. mTOR activates both the ribosomal p70 S6 kinase (S6K1) and the mRNA cap-binding protein inhibitory protein, 4E-BP1, and its pharmacological inhibition causes G 1 phase cell cycle arrest (19,20). Recently, there have been reports of potential crosstalk between GCN2 and the TOR signaling pathway in yeast (21,22). Specifically, rapamycin releases TOR-directed phosphorylation of yeast GCN2, contributing to enhanced eIF2␣ kinase activity (21,22). Although this finding has not yet been extended to mammals, it suggests that events downstream of mTOR, namely the phosphorylation of 4E-BP1 and S6K1, may be coordinated with eIF2␣ phosphorylation via GCN2. Previous work by one of the authors (4) demonstrates that dietary leucine deprivation results in both the increased phosphorylation of eIF2␣ and the concomitant reduced phosphorylation of 4E-BP1 and S6K1 in rat liver.
The mammalian form of GCN2 was identified several years ago (23,24), and soon thereafter a GCN2 knockout mouse was developed and partially described (25). Its phenotype was initially unremarkable, reportedly growing and reproducing normally under freely fed conditions. However, pregnant GCN2 Ϫ/Ϫ dams fed a diet devoid of leucine bore fewer viable pups as compared with pregnant GCN2 ϩ/ϩ dams, suggesting that GCN2 is important for managing nutritional stress during embryogenesis. Currently, there are no studies addressing the role of GCN2 in managing postnatal stress in these mice. Thus, the focus of this investigation was to characterize the growth response of GCN2 Ϫ/Ϫ mice to dietary amino acid restriction, and to determine whether GCN2 is important for coping with nutritional stress postnatally. We were also interested in understanding if GCN2 is the primary kinase involved in catalyzing the phosphorylation of eIF2␣ in response to amino acid deprivation in vivo, or if other eIF2␣ kinase family members also serve a role in sensing amino acid deprivation. Finally, we wished to examine whether dysregulation of eIF2␣ phosphorylation would impact signaling downstream of mTOR.

EXPERIMENTAL PROCEDURES
Animals and Diets-The following study protocol was approved by the Institutional Care and Use Committee at the Indiana University School of Medicine, Evansville Center for Medical Education. Mice were obtained by in-house breeding or from Jackson Laboratories (Bar Harbor, ME). Four to 6-week-old male C57BL/6J (GCN2 ϩ/ϩ ) and GCN2 Ϫ/Ϫ mice (either on a heterozygous hybrid background of C57BL/6J and TL1 129SvEvTac, as reported in (25) and designated GCN2 Ϫ/ϪHH ) or backcrossed onto the C57BL/6J background 8 generations (GCN2 Ϫ/ϪBC ) were maintained on a 12-h light:dark cycle and provided free access to commercial rodent chow (PMI International, Brentwood, MO) and tap water prior to the experiment. At the start of the feeding experiment, mice were acclimated to a nutritionally complete, control diet for a minimum of 3 days, and then randomly assigned to one of four dietary treatments as follows: AA, continued free access to the nutritionally complete, control diet; ϪLEU, free access to a diet that was devoid of the essential amino acid leucine; ϪGLY, free access to a diet that was devoid of the non-essential amino acid glycine; PF, restricted access to the AA diet that equaled the intake of mice freely consuming the ϪLEU diet as previously detailed (4). Rodents, when fed a diet deficient in an essential amino acid, will reduce their intake (26,27). Consequently, the PF group was used to compare the effects of amino acid versus calorie deprivation on signaling events regulating protein synthesis. PF mice received their daily food ration as one meal at the beginning of the light cycle. All diets were isocaloric and compositionally the same in terms of carbohydrate and lipid components as described previously (4). The ϪLEU and ϪGLY diets were isonitrogenous with the AA diet, with alanine, glutamate, and aspartate compensating for the missing amino acid. Animals in each group were offered their diets at the beginning of the light cycle following an overnight fast and euthanized at 1 h and 6 days (144 h) following commencement of feeding. Food intake and body weight was recorded daily and body and tissue wet weights (liver and skeletal muscle) were recorded at each point of euthanasia. Animals were killed by decapitation and trunk blood was collected for the determination of serum amino acids and insulin concentrations.
Serum Measurements-Serum was obtained by centrifugation of clotted blood and then snap-frozen and stored at Ϫ20°C. Serum samples were sent to the Indiana University School of Medicine Quantitative Amino Acid Core Facility (under the direction of Edward Liechty) for the determination of amino acid profiles by the ninhydrin method, using standard ion exchange chromatography with a Beckman 6300 automated amino acid analyzer. Serum insulin was measured using a commercial radioimmunoassay kit (Linco, Inc., St. Louis, MO).
Protein Synthesis-Ten minutes before euthanasia, all mice were injected intraperitoneally with a bolus solution of 250 mg of DL-[ 2 H 5 ]phenylalanine per kg of body weight for the measurement of tissue protein synthesis. Each tissue sample was processed to determine the enrichment of labeled phenylalanine in liver protein as previously described (28). Tissue L-[ 2 H 5 ]phenylalanine enrichments were measured in Stony Brook by monitoring the ions at m/z 336 and 341 of the tertiary butyldimethylsilyl derivative on a model MD800 GC-MS (Fisons Instruments) operated under electron impact (29). The intraperitonal route of injection to ensure constant precursor enrichment has been previously validated (30). Additionally, time course studies performed using the intraperitonal route of injection demonstrated constant enrichment of the free phenylalanine within the liver and linear incorporation of injected tracer into tissue protein for at least 20 min (data not shown).
Phosphorylation of eIF2␣-Phosphorylation of eIF2␣ was measured as described (4) using an antibody that recognizes the protein only when it is phosphorylated at serine 51 (Cell Signaling Technology, Inc., Beverly, MA). Results were normalized for total eIF2␣ with an antibody that recognizes the protein irrespective of phosphorylation state (Santa Cruz Biotechnology, Santa Cruz, CA).
Phosphorylation of 4E-BP1-Phosphorylation of 4E-BP1 was measured as a change in migration during SDS-polyacrylamide gel electrophoresis as detected by immunoblot analysis as described previously (4). Briefly, an aliquot of the 10,000 ϫ g supernatant was boiled for 10 min and centrifuged at 10,000 ϫ g for 30 min at 4°C. The resultant supernatant was added to 1 volume of SDS sample buffer and then subjected to protein immunoblot analysis using a polyclonal 4E-BP1 antibody (Bethyl Laboratories, Montgomery, TX).
Phosphorylation of S6K1-Phosphorylation of S6K1 was measured as a decrease in mobility during SDS-polyacrylamide gel electrophoresis as described previously (4). Briefly, an aliquot of the 10,000 ϫ g supernatant was added to 1 volume of SDS sample buffer. Immunoblot analysis was then performed using a polyclonal S6K1 antibody (Santa Cruz Biotechnoloogy, Santa Cruz, CA).
Statistics-All data were analyzed by the STATISTICA statistical software package for the Macintosh, volume II (StatSoft, Tulsa, OK). Data were analyzed using two-way ANOVA to assess main effects, with mouse strain and diet treatment as the independent variables. When a significant overall effect was detected, differences among treatment groups were assessed with Duncan's Multiple Range post-hoc test. The level of significance was set at p Ͻ 0.05 for all statistical tests. Data are reported as mean Ϯ S.E.

GCN2
Ϫ/ϪBC Mice Are More Fragile Than GCN2 Ϫ/ϪHH or GCN2 ϩ/ϩ Mice-A prior characterization of the GCN2 null mouse indicated that these mice grow normally with complete nutrition (25). GCN2 Ϫ/Ϫ mice used in this earlier study were of a heterozygous hybrid background, a genetic mixture of the C57BL/6J and TL1 129 SvEvTac strains (designated in this study as GCN2 Ϫ/Ϫ HH ). It is recognized that the genetic background of engineered mice can influence physiological response to challenge (31); therefore, GCN2 Ϫ/ϪHH mice were backcrossed into the C57BL/6J lineage for 8 generations (designated in this study as GCN2 Ϫ/ϪBC ) to produce a mouse that is 99.6% genetically identical to a C57BL/6J mouse (designated in this study as GCN2 ϩ/ϩ ). Average litter size of the GCN2 Ϫ/ϪHH strain was greater than GCN2 Ϫ/ϪBC or GCN2 ϩ/ϩ (6.73 Ϯ 0.26 versus 5.65 Ϯ 0.29 versus 5.46 Ϯ 0.45 pups per litter, respectively, p Ͻ 0.05). Although similar in initial litter size, GCN2 Ϫ/ϪBC mice were decidedly more fragile than GCN2 ϩ/ϩ mice, as a subset of most litters died shortly (1-2 days) after birth. GCN2 Ϫ/ϪHH mice could be weaned at day 21, but weaning had to be delayed until day 28 with GCN2 Ϫ/ϪBC mice to prevent additional mortality, particularly when housed singly or on wire-bottom cages. By contrast, GCN2 ϩ/ϩ and GCN2 Ϫ/ϪHH mice adapted to these environmental conditions upon weaning without incident. Growth curves from birth until 28 days of age demonstrated that GCN2 Ϫ/ϪHH mice grew larger than GCN2 Ϫ/ϪBC or GCN2 ϩ/ϩ mice (p Ͻ 0.05 at days 12-28; Fig. 1). GCN2 ϩ/ϩ and GCN2 Ϫ/ϪBC mice grew similarly during the suckling period until day 27 when the body weights of GCN2 Ϫ/ϪBC mice began to plateau as compared with GCN2 ϩ/ϩ mice, ending up statistically the smallest in size at day 28.
GCN2 Ϫ/Ϫ Mice Consume Less Food, Lose More Body Weight, and Have Increased Mortality When Fed a Leucine-devoid Diet-Intake of a leucine-free diet caused the expected reduction in food intake in all strains (two-way ANOVA p Ͻ 0.05 main effect of ϪLEU on reducing intake) (Fig. 3B). However, GCN2 Ϫ/ϪBC mice consumed less food than GCN2 ϩ/ϩ or GCN2 Ϫ/ϪHH mice independent of dietary treatment (two-way ANOVA p Ͻ 0.05 main effect of GCN2 Ϫ/ϪBC strain on food intake). Intake of the leucine-devoid diet in both GCN2 Ϫ/Ϫ strains was less than that in GCN2 ϩ/ϩ mice (Ϫ30% in GCN2 Ϫ/ϪBC versus Ϫ34% in GCN2 Ϫ/ϪHH versus Ϫ22% in GCN2 ϩ/ϩ compared with respective AA intake). Consumption of the ϪGLY diet was similar to intake of the AA diet in each respective strain, whereas the PF intake successfully matched the intake of the ϪLEU mice.
Although GCN2 Ϫ/ϪHH mice grew better than GCN2 ϩ/ϩ mice, they also lost more body weight on a leucine-devoid diet. GCN2 Ϫ/ϪHH mice fed ϪLEU lost 23% of their starting body weight by day 6, whereas GCN2 ϩ/ϩ mice lost 14.3% (Fig. 3A). GCN2 Ϫ/ϪHH mice further became scruffy and lethargic in appearance by day 6 (Fig. 2), yet none died. In contrast, GCN2 Ϫ/ϪBC mice fed ϪLEU experienced a substantial rate of death before the end of the experiment (Table I). Forty percent (8 of 20) of the initial starting number of GCN2 Ϫ/ϪBC mice assigned to a leucine-devoid diet would die within 3-5 days. GCN2 Ϫ/ϪBC mice that remained viable throughout the 6-day experimental period also appeared sickly and experienced loss of body weight (Ϫ19.4%) that was slightly greater than GCN2 ϩ/ϩ mice fed the ϪLEU diet, and statistically intermediate between the GCN2 Ϫ/ϪHH and GCN2 ϩ/ϩ strains. Mice fed a glycine-devoid diet generally gained weight throughout the study period, although GCN2 Ϫ/ϪBC mice gained less weight than GCN2 ϩ/ϩ or GCN2 Ϫ/ϪHH . This is likely because of reduced intake of the ϪGLY diet by GCN2 Ϫ/ϪBC mice (Fig. 3B). All PF mice lost weight because of reduced caloric intake equal to GCN2 ϩ/ϩ mice fed a leucine-devoid diet. Thus, leucine deficiency in GCN2 Ϫ/Ϫ mice resulted in greater weight loss than that because of caloric restriction alone.
A small number of GCN2 Ϫ/ϪBC mice fed the other diets also died before the end of the study period (Table I). Closer examination of the expired GCN2 Ϫ/ϪBC AA, ϪGLY, and PF mice revealed their starting body weights to be smaller than the average group weights (below 12 versus 15 g average). Therefore, similar to younger GCN2 Ϫ/ϪBC mice, their demise may be related to an inability to adapt to being housed singly in a wire-bottom cage, as noted above.
GCN2 Ϫ/Ϫ Mice Experience Sparing of Liver Mass during Dietary Leucine Starvation-Leucine deprivation reduced wet weight of liver in all strains. However, when expressed relative to body weight, only GCN2 ϩ/ϩ mice fed on the leucine-devoid diet experienced significant loss of liver relative to body mass (Ϫ20% in ϪLEU, p Ͻ 0.05 as compared with AA; Fig. 3C). In contrast, both GCN2 Ϫ/Ϫ strains were resistant to losing a significant percentage of their liver mass. Interestingly, food restriction (PF group) significantly reduced liver mass in both GCN2 ϩ/ϩ (Ϫ15%) and GCN2 Ϫ/ϪBC (Ϫ22%) mice. A glycine-free diet did not alter liver mass in any strain compared with the AA diet.
Dietary Leucine Starvation Reduces Skeletal Muscle Mass in GCN2 Ϫ/Ϫ Mice-The wet weights of hind limb muscle (gastrocnemius plus plantaris) were reduced in all mice in response to caloric restriction (PF) or leucine deprivation. However, when expressed as a ratio to body weight, skeletal muscle from GCN2 ϩ/ϩ mice generally remained in proportion to body weight irrespective of dietary treatment (Fig. 3D). By contrast, GCN2 Ϫ/ϪHH and GCN2 Ϫ/ϪBC mice demonstrated proportionately greater loss of muscle mass by feeding ϪLEU (Ϫ20.5% and Ϫ17.5%, respectively, p Ͻ 0.05) and GCN2 Ϫ/ϪHH strain  a Experimental diets: AA, free access to a nutritionally complete diet; ϪLEU, free access to a diet that was devoid of the essential amino acid leucine; ϪGLY, free access to a diet that was devoid of the non-essential amino acid glycine; PF, restricted access to the AA diet that equaled the intake of mice freely consuming the ϪLEU diet. demonstrated marked loss by PF (Ϫ16.5%, p Ͻ 0.05). Thus, loss of functional GCN2 resulted in greater loss of skeletal muscle in response to nutrient deprivation.
Amino Acid Profiles Suggest GCN2 Ϫ/Ϫ Mice Fed ϪLEU or PF for 6 Days Are More Catabolic Than GCN2 ϩ/ϩ Mice-Serum concentrations of the 20 essential and non-essential amino acids were measured at 1 h and 6 days following consumption of experimental diets (Tables II and III). Data in the two GCN2 null strains were pooled, because values were similar. At 1 h, serum leucine concentrations were lower and serum isoleucine and valine concentrations were higher, in all mice fed ϪLEU (main effect of leucine-free diet by two-way ANOVA, p Ͻ 0.05) ( Table I). In all mice fed ϪGLY, serum concentrations of glycine and serine were lower, irrespective of mouse strain (main effect of glycine-free diet by two-way ANOVA, p Ͻ 0.05). There were no changes in the concentrations of any other amino acids.
Six days later, serum concentrations of the same amino acids were measured again (Table III). Mice fed the ϪLEU diet continued to have reduced leucine concentrations as compared with mice fed the AA diet (two-way ANOVA main effect of ϪLEU diet, p Ͻ 0.05), but circulating levels of leucine in GCN2 Ϫ/Ϫ mice fed ϪLEU were 42% lower than GCN2 ϩ/ϩ mice fed ϪLEU. In GCN2 ϩ/ϩ mice fed the ϪLEU diet, serum concentrations of isoleucine and valine were equal to GCN2 ϩ/ϩ mice fed AA. Conversely, isoleucine and valine concentrations in GCN2 Ϫ/Ϫ mice fed the ϪLEU were 2.7-and 2.1-fold higher, respectively, than GCN2 Ϫ/Ϫ mice fed the AA diet. In addition, concentrations of isoleucine, valine, alanine, and serine in GCN2 Ϫ/Ϫ mice fed ϪLEU were equal to PF mice. The PF mice, which were food-restricted ϳ30% in daily energy, displayed elevated concentrations of leucine, isoleucine, valine, alanine, and asparagine at day 6 as compared with same strain using the AA diet. The PF amino acid profiles suggest increased catabolism of body proteins in response to acute food deprivation, consistent with an adopted "meal feeding" pattern of in-take where most of the daily rations were consumed in the early hours following presentation of food. Furthermore, GCN2 Ϫ/Ϫ PF mice displayed additional elevations in circulating concentrations of serine (ϩ186%), lysine (ϩ156%), threonine (ϩ221%), and arginine (ϩ173%) as compared with GCN2 Ϫ/Ϫ mice fed AA, signifying further catabolism of body proteins. Mice fed the ϪGLY diet maintained marked to significant reductions in serum glycine and serine, but did not demonstrate changes in the concentration of any other amino acid.
Serum insulin was measured at both 1 h and 6 days following experimental diet intake, and again, data from the two GCN2 null strains were combined (Tables II and III). At both time points, mice fed ϪLEU demonstrated reduced circulating concentrations of insulin (two-way ANOVA main effect of a leucine-devoid diet, p Ͻ 0.05) likely due in part to reduced intake of food during these periods. There was no effect of mouse strain or other dietary treatments on serum insulin values.
Reductions in Relative Protein Synthesis by Dietary Leucine Deprivation Are Maintained in GCN2 ϩ/ϩ but Not GCN2 Ϫ/Ϫ Mice-Protein synthesis was measured as the enrichment of phenylalanine labeled with deuterium into liver protein, using the "flooding dose" method to saturate and equilibrate all tissue precursor pools of tracer amino acid. Data in the two GCN2 null strains were pooled, as values were similar. One hour following commencement of feeding, mice fed ϪLEU demonstrated significant reductions in labeling of liver protein that were similar between strains (GCN2 ϩ/ϩ : Ϫ29% versus GCN2 Ϫ/Ϫ Ϫ33%; two-way ANOVA main effect of leucine-devoid diet, p Ͻ 0.05) (Fig. 4A). There was no effect of feeding ϪGLY to the labeling of liver protein in either strain. On day 6, there was a differential effect of dietary leucine deprivation between wild-type and null strains (Fig. 4B). Hepatic tracer enrichment remained suppressed in GCN2 ϩ/ϩ mice fed ϪLEU as compared with GCN2 ϩ/ϩ AA (Ϫ29%, p Ͻ 0.05). In contrast, enrichment of FIG. 3. GCN2 ؊/؊ mice consume less food, lose more body weight, and experience sparing of liver mass at the expense of muscle mass during dietary leucine starvation. Wild-type (GCN2 ϩ/ϩ ), GCN2 null heterozygous hybrids (GCN2 Ϫ/ϪHH ), and GCN2 null backcrossed (GCN2 Ϫ/ϪBC ) mice were fed experimental diets for 6 days. Experimental diets: AA, free access to a nutritionally complete diet; ϪLEU, free access to a diet that was devoid of the essential amino acid leucine; ϪGLY, free access to a diet that was devoid of the non-essential amino acid glycine; PF, restricted access to the AA diet that equaled the intake of mice freely consuming the ϪLEU diet. A, percent body weight change between days 0 and 6. Values are mean Ϯ S.E., n ϭ 8 -14 per group. Means not sharing the same letter are different, p Ͻ 0.05. B, total grams of food consumed over 6 days. Values are mean Ϯ S.E., n ϭ 8 -14 per group. *, main effect of leucine to reduce intake, p Ͻ 0.05; ϩ, main effect of PF to reduce intake, p Ͻ 0.05; #, main effect of mouse strain to reduce intake, p Ͻ 0.05. C, liver weight expressed relative to body weight. Values are mean Ϯ S.E., n ϭ 8 -14 per group. *, different from GCN2 ϩ/ϩ AA, p Ͻ 0.05; ϩ , different from GCN2 Ϫ/ϪBC AA, p Ͻ 0.05. D, skeletal muscle (gastrocnemius plus plantaris) weight expressed relative to body weight. Values are mean Ϯ S.E., n ϭ 8 -14 per group. *, different from same-strain AA control, p Ͻ 0.05. phenylalanine into GCN2 Ϫ/Ϫ ϪLEU liver protein was reduced by only Ϫ17% compared with GCN2 Ϫ/Ϫ AA, failing to reach statistical significance. There was again no effect of ϪGLY feeding on enrichment of labeled phenylalanine into liver protein. Pair-feeding caused slight, but not significant, reductions in incorporation of label into protein in both strains, with PF values intermediary between AA and ϪLEU diets (Fig. 4B).

Signaling Events in Translation Initiation in Response to Dietary Leucine Deprivation Are Disrupted in GCN2 Ϫ/Ϫ
Mice-As expected, GCN2 ϩ/ϩ mice demonstrated robust phosphorylation of eIF2␣ at 1 h and 6 days (Fig. 5, A and B). By contrast, GCN2 Ϫ/Ϫ mice (both null strains) fed ϪLEU demonstrated no increase in the phosphorylation of eIF2␣ at either 1 h or 6 days. Results were similar at intermediate time points examined (3, 6, 24, and 72 h) (data not shown). Pair-feeding also increased eIF2␣ phosphorylation in GCN2 ϩ/ϩ mice but not GCN2 Ϫ/Ϫ mice. Therefore, GCN2 appears to be the primary or dominant eIF2␣ kinase that responds to nutrient deprivation.
To examine the possibility that abrogation of eIF2␣ phosphorylation might alter other events in translation initiation, the phosphorylation states of 4E-BP1 and S6K1, two downstream effectors of mTOR that are involved in regulating the translation of growth-related mRNAs were also examined. Phosphorylation of both proteins was examined by migration following SDS-PAGE and immunoblot analysis. During electrophoresis, 4E-BP1 and S6K1 are resolved into multiple phosphorylated species, labeled as ␣, ␤, and ␥ (Figs. 6 and 7).
For 4E-BP1, the ␥-form is the most highly phosphorylated species and is the only species that is reported to not bind eIF4E; therefore, the results are expressed as a ratio representing the amount of protein present in the ␥-form relative to the total. At 1 h and 6 days, GCN2 ϩ/ϩ mice fed ϪLEU demonstrated significant reductions in the amount of protein present in the ␥-form (Fig. 6). By comparison, neither strain of GCN2 Ϫ/Ϫ mice fed ϪLEU demonstrated significant reductions in 4E-BP1 hyperphosphorylation at either 1 h or 6 days. It should be noted that while no change in the ␥-form was seen, GCN2 Ϫ/Ϫ mice fed ϪLEU did show a slight increase in the ␣-form at day 6 ( Fig. 6B); indicating that some dephosphorylation was occurring. Interestingly, calorie restriction resulted in decreased 4E-BP1 phosphorylation in all mice (p Ͻ 0.05).
Phosphorylation of S6K1 was calculated as a ratio of the density of the upper bands (␤ ϩ ␥ forms) over the total density of all bands present. Similar to 4E-BP1, only GCN2 ϩ/ϩ mice demonstrated significant reductions in the phosphorylation state of S6K1 by feeding a leucine-devoid diet for 1 h (Fig. 7A) or 6 days (Fig. 7B). Again, similar to 4E-BP1, food restriction by pairfeeding resulted in a significant reduction in S6K1 phosphoryla-  tion in all mice. These results indicate that loss of functional GCN2 precludes maximal inhibition of mTOR signaling by amino acid deprivation.

DISCUSSION
Herein we report the following novel observations: 1) GCN2 is the dominant kinase involved in the phosphorylation of eIF2␣ in the liver in response to amino acid deprivation, because GCN2 Ϫ/Ϫ mice do not display increased eIF2␣ phosphorylation in the liver following 1 h and 6 days of feeding a leucine-devoid diet; 2) GCN2 is important for viability under nutritional stress. GCN2 Ϫ/Ϫ mice that consumed a leucinedevoid diet became scruffy in appearance and either lost significantly more weight than GCN2 ϩ/ϩ mice, or died before the end of the study period; 3) mice lacking functional GCN2 had reduced abilities to chronically down-regulate hepatic protein synthesis in response to leucine starvation over time, resulting in preservation of liver mass relative to body size; 4) the inability to adequately restrain protein synthesis in the liver results in increased loss of skeletal muscle to supply needed substrate (leucine); and 5) loss of GCN2 reduces dephosphorylation of 4E-BP1 and S6K1 in response to leucine deprivation, suggesting a role for this eIF2␣ kinase in the regulation of the mTORdirected pathway. Together, these results demonstrate that GCN2 contributes to the regulation of protein balance and the coordination of translation initiation during dietary amino acid deprivation.

GCN2 Is the Primary eIF2␣ Kinase Activated in Liver during
Amino Acid Limitation-GCN2 is one member in a family of eIF2␣ kinases. Previously, it was reported that while eIF2␣ phosphorylation was completely dependent on GCN2 in mouse embryonic fibroblasts subjected to leucine starvation for up to 6 h, extended periods of starvation induced one or more alternative eIF2␣ kinases (25). In the current study, compensatory phosphorylation of eIF2␣ did not occur (Fig. 5). The difference between our findings in vivo and those in cultured cells are most likely because of the fact that complete starvation of a single amino acid cannot be produced by dietary manipulation alone, because breakdown of body proteins can produce a continued supply of the deficient amino acid. Thus, it can be argued that amino acid starvation in cultured cells represents a more extreme situation that may not be encountered in animal tissues. A possible exception to this idea involves treatment of amino acid-depleting enzymes to treat cancer. During treatments using anti-cancer drugs such as asparaginase, the circulating concentration of the targeted amino acid is depleted to such a low level that the resulting response may be comparable with amino acid-free culture conditions (32). 2 This concept remains to be fully investigated.
Whereas genetic diversity impacted the gross phenotype in this study (Fig. 2 and Table I), common themes remained in both GCN2 null strains at the tissue, cellular, and subcellular levels. Directional changes in serum and tissues were generally consistent in both GCN2 Ϫ/Ϫ strains, even if the response was more marked in one null strain versus the other. And although higher rates of mortality only occurred in the GCN2 Ϫ/ϪBC strain, both null strains appeared generally unwell as compared with GCN2 ϩ/ϩ mice fed the leucine-devoid diet. Nevertheless, this study substantiates previous reports indicating that the genetic background of a rodent can influence its phenotypic response to challenge (31).
Loss of GCN2 Triggers Death in Nutritionally Stressed Animals-The precise reason for the premature death of the GCN2 Ϫ/ϪBC mice remains ambiguous. It is noted that some GCN2 Ϫ/ϪBC mice completely stop eating within 1-3 days after presentation of the ϪLEU diet; thus, their declining health may be related to starvation. It is known that rodents develop a taste aversion to amino acid imbalanced diets, leading to a reduction in intake (1,2). Work by Gietzen's (26) group demonstrated that a specific area in the brain, the anterior pyriform cortex, is reactive to amino acid imbalance, demonstrating increased phosphorylation of eIF2␣ and expression of c-Jun. Interestingly, GCN2 is abundantly expressed in brain (24). However, no study to date has identified GCN2 activation as being central to the regulation of food intake.
Alternatively, the failing health of the GCN2 Ϫ/ϪBC mice fed ϪLEU may relate to altered glucoregulation in response to self-starvation. Kaufman and co-workers (33) created a mouse in which serine 51 on eIF2␣ was mutated to an alanine, precluding all phosphorylation at this site. This S51A eIF2␣ "knock-in" mouse was born viable but died within 18 h because of hypoglycemia associated with defective gluconeogenesis (33). Thus, the lack of eIF2␣ phosphorylation in the liver of nutritionally stressed GCN2 null mice may result in the altered ability to regulate blood glucose during these conditions. Preliminary studies of GCN2 ϩ/ϩ and GCN2 Ϫ/ϪBC lactating dams fed a leucine-free diet for 6 days demonstrate hypoglycemia in GCN2 Ϫ/ϪBC pups (65 Ϯ 8 mg/dL) but not GCN2 ϩ/ϩ pups (101 Ϯ 3 mg/dL). 3 Further study addressing the role of GCN2 in glu- FIG. 5. Absence of GCN2 preludes phosphorylation of eIF2␣ in the liver of mice fed a leucine-deprived diet for 1 h or 6 days. Wild-type (GCN2 ϩ/ϩ ) and GCN2 null (combination of GCN2 Ϫ/ϪHH and GCN2 Ϫ/ϪBC ) mice were fed experimental diets for 1 h (A) or 6 days (B). Values are mean Ϯ S.E., n ϭ 6 -8 per group. *, different from all other treatment groups, p Ͻ 0.05; ϩ, different from GCN2 ϩ/ϩ mice fed AA, p Ͻ 0.05.
An important finding in this study is that liver mass is spared in GCN2 Ϫ/Ϫ mice fed ϪLEU, and instead, skeletal muscle is lost (Fig. 3). In normal liver, amino acid starvation stimulates macroautophagy, resulting in marked reduction in liver mass (34). Skeletal muscle is normally maintained in the shortterm, as breakdown rates do not initially increase when amino acid supply is reduced (35). During extended periods of protein or calorie deprivation, rates of proteolysis in the liver gradually diminish, whereas corresponding responses in skeletal and cardiac muscle move in the opposite direction (36). In the present study, mice lacking functional GCN2 were unable to suppress hepatic translation in response to leucine starvation, resulting in moderate derepression of protein synthesis in the liver over time. While not measured, loss of functional GCN2 may also have affected the formation of hepatic autophagic vacuoles during leucine deprivation, because both the eIF2␣ kinase and mTOR signaling cascades are implicated in the regulation of autophagy (37,38). Future studies addressing the role of GCN2 in regulating proteolysis during amino acid deprivation are planned.
Regulation of Translation Initiation by GCN2-Reductions in growth by amino acid deficiency are due in part to an inhibition of protein synthesis at the level of translation initiation. Indeed, rats fed diets lacking one or more essential amino acids demonstrate hepatic protein synthesis to be lowered in association with reduced eIF2B activity (4), supporting the general idea that phosphorylation of eIF2␣ regulates global protein synthesis (39). Whereas our study shows that GCN2 is the major kinase that phosphorylates eIF2␣ during dietary leucine restriction, it also demonstrates that increased eIF2␣ phosphorylation is not required for immediate reduction of general translation (Figs. 4A and 5A). These results, which may appear contrary to contemporary descriptions in the literature, are in fact consistent with a previous report characterizing GCN2 Ϫ/Ϫ mice (25). In the earlier study, in situ perfusion of histidinol induced eIF2␣ phosphorylation and restrained eIF2B activity in livers from GCN2 ϩ/ϩ but not GCN2 Ϫ/ϪHH mice (25). Nevertheless, liver protein synthesis was equally suppressed by histidinol in both strains of mice. Collectively, these data suggest that phosphorylation of eIF2␣ may function in conjunction with other regulators of translation, and emphasizes a significant role for eIF2␣ kinases in regulation of genespecific translation.
To determine whether regulatory mechanisms other than eIF2␣ phosphorylation are altered by loss of GCN2 function, we addressed whether signaling events downstream of mTOR were reduced in GCN2 Ϫ/Ϫ mice fed ϪLEU. Hyperphosphorylation of both 4E-BP1 and S6K1 are associated with the increased translation of specific classes of mRNAs important for growth (4,40). In the current study, phosphorylation of both 4E-BP1 and S6K1 was reduced in GCN2 ϩ/ϩ mice fed ϪLEU at both 1 h and 6 days, supporting an inhibitory role for these translation regulators in global protein synthesis (Figs. 6 and 7). However, in GCN2 Ϫ/Ϫ mice deprived of leucine, the phosphorylation of 4E-BP1 and S6K1 in liver was not decreased. These observations support the idea that GCN2 contributes to mTOR regulation of 4E-BP1 and S6K1 in response to amino acid depletion, but not to general caloric reduction because there was no difference in the phosphorylation levels of these translation regulators between PF fed GCN2 ϩ/ϩ and GCN2 Ϫ/Ϫ mice. Furthermore, increased phosphorylation of 4E-BP1 and S6K1 in the GCN2 Ϫ/Ϫ mice deprived of leucine would not be conductive to dampened protein synthesis. Thus, it appears that signaling via mTOR does not exhibit major control of hepatic protein synthesis in the absence of the primary eIF2␣ kinase GCN2.
In the combined absence of changes in the phosphorylation of eIF2␣, S6K1, and 4E-BP1 at 1 h, alternative or novel mechanisms may serve to initially repress hepatic protein synthesis in GCN2 Ϫ/Ϫ mice during leucine deprivation. One possibility involves the phosphorylation of eIF2B on its ⑀ subunit, associated with reductions in eIF2B guanine nucleotide exchange activity in muscle (41). However, decreased eIF2B activity does not appear to coincide with general reductions in protein synthesis in the liver of GCN2 Ϫ/Ϫ mice (25). Another possibility is the suppression of elongation by means of eEF2 phosphorylation. Whereas the kinase that phosphorylates and inhibits eEF2 is regulated in part by a rapamycin-sensitive manner (19), the activity of eEF2 is also regulated by cellular energy levels via an mTOR-independent mechanism (42). It is reported that pharmacological activation of the AMP-activated protein kinase reduces general protein synthesis rates in the skeletal muscle of rodents (43), but it is unknown what effect AMPK activation has on liver protein synthesis. Further exploration in this area is warranted. the scientific counsel of Carla Aldrich, Guiseppe Caso, Robert Harris, and Ed Leichty.