The Regulation of Hepatic Protein Synthesis during Fasting in the Rat*

We have studied translational control in the model of 48 h of fasting in the rat. Our initial observations showed a paradoxical increase in ribosomal protein S6 (rpS6) phosphorylation and a decrease in eukaryotic initiation factor 2α (eIF2α) phosphorylation. These effects, which would favor an increase in protein synthesis, could be attributed to increased circulating concentrations of branched-chain amino acids in fasting. To determine what mechanisms might account for decreased hepatic translation in fasting, we examined the cap binding complex. eIF4E-bound 4E-BP1 did not increase. However, eIF4E-bound eIF4G and total cellular eIF4G were profoundly decreased in fasted liver. eIF4G mRNA levels were not lower after fasting. Based on the hypothesis that decreased eIF4G translation might account for the reduced eIF4G content, we fractionated ribosomes by sucrose density centrifugation. Immunoblotting for rpS6 showed modest polysomal disaggregation upon fasting. PCR analysis of polysome profiles revealed that a spectrum of mRNAs undergo different translational regulation in the fasted state. In particular, eIF4G was minimally affected by fasting. This indicated that reduced eIF4G abundance in fasting may be a function of its stability, whereas its recovery upon refeeding is necessarily independent of its own involvement in the cap binding complex. Western immunoblotting of polysome fractions showed that phosphorylated rpS6 was disproportionately present in translating polysomes in fed and fasted animals, consistent with a role in translational control. However, the translation of rpS8, an mRNA with a 5′-oligopyrimidine tract, did not coincide with rpS6 phosphorylation, thus dissociating rpS6 phosphorylation from the translational control of this subset of mRNAs.

We have studied translational control in the model of 48 h of fasting in the rat. Our initial observations showed a paradoxical increase in ribosomal protein S6 (rpS6) phosphorylation and a decrease in eukaryotic initiation factor 2␣ (eIF2␣) phosphorylation. These effects, which would favor an increase in protein synthesis, could be attributed to increased circulating concentrations of branched-chain amino acids in fasting. To determine what mechanisms might account for decreased hepatic translation in fasting, we examined the cap binding complex. eIF4E-bound 4E-BP1 did not increase. However, eIF4E-bound eIF4G and total cellular eIF4G were profoundly decreased in fasted liver. eIF4G mRNA levels were not lower after fasting. Based on the hypothesis that decreased eIF4G translation might account for the reduced eIF4G content, we fractionated ribosomes by sucrose density centrifugation. Immunoblotting for rpS6 showed modest polysomal disaggregation upon fasting. PCR analysis of polysome profiles revealed that a spectrum of mRNAs undergo different translational regulation in the fasted state. In particular, eIF4G was minimally affected by fasting. This indicated that reduced eIF4G abundance in fasting may be a function of its stability, whereas its recovery upon refeeding is necessarily independent of its own involvement in the cap binding complex. Western immunoblotting of polysome fractions showed that phosphorylated rpS6 was disproportionately present in translating polysomes in fed and fasted animals, consistent with a role in translational control. However, the translation of rpS8, an mRNA with a 5-oligopyrimidine tract, did not coincide with rpS6 phosphorylation, thus dissociating rpS6 phosphorylation from the translational control of this subset of mRNAs.
Prolonged fasting in rodents is accompanied by a decrease in liver size (1)(2)(3). This can be accounted for by a decrease in liver protein content while liver cellularity is maintained at prefasting levels (1,4). A number of studies have focused on the cellular mechanisms that could account for the decrease in hepatocyte and liver size that occurs in the fasted state. It appears that both autophagy and inhibition of protein synthesis play a role (2,(5)(6)(7). The role of autophagy has been conclu-sively established, whereas the mechanisms accounting for translational inhibition remain unexplained. Recent advances in the area of translational control (8 -10) led us to undertake a detailed analysis of the hepatic translational apparatus in fasted rats.
Cap binding complex formation, eIF2␣ activity state, and rpS6 phosphorylation are all regulated by the mammalian target of rapamycin (mTOR). mTOR is a nutrient-sensing kinase (18). It is well established that its activity is under the influence of branched-chain amino acids (BCAA). However, mTOR is also regulated by receptor tyrosine kinases, including those for insulin, insulin-like growth factor I, and a panoply of peptide growth factors (19). It is generally accepted that the mechanism by which mTOR regulates mRNA translation involves activation of the translation initiation factor eIF4E via phosphorylation of the eIF4E inhibitory binding protein, 4E-BP1 (9). Phosphorylation of 4E-BP1 results in its dissociation from eIF4E, which permits eIF4E to interact with the scaffold protein eIF4G. eIF4G, in turn, recruits the RNA helicase eIF4A, which unwinds the 5Ј-untranslated region of the mRNA, the Poly(A)-binding protein, which circularizes the mRNA for translation, and eIF3, which links the mRNA to the 40 S subunit of the ribosome (20). This multiprotein complex forms at the 5Ј-cap of eukaryotic mRNAs and results in translational initiation. Activation of mTOR also results in the activation of the rpS6 kinases S6K1 and S6K2 (21,22). More recently, mTOR signaling has been shown to be involved in controlling the phosphorylation of eIF2␣ by its cognate kinase, GCN2. GCN2-dependent phosphorylation of eIF2␣, which inhibits the ability of this initiation factor to participate in the activation of translation (9), is rapamycin-sensitive in yeast, indicating a role for mTOR in this phosphorylation event (23,24).
For the present studies, we have focused on the model of fasting in the adult rat to examine several aspects of translation control in the absence of cellular proliferation. Our rationale for doing so was based on long-standing observations that circulating BCAA concentrations increase in the fasted state (25,26). This leads to the conundrum that fasting might be associated with mTOR activation rather than inhibition. This, in turn, led to the following questions. Can a decrease in rpS6 phosphorylation in the fasted state account for a reduction in 5Ј-TOP mRNA translation? Is phosphorylated rpS6 preferentially found in ribosomes that are translating (i.e. in polysomes), thereby supporting a role for rpS6 phosphorylation in regulation of translation? Are 4E-BP1-dependent control of eIF4E and regulation of eIF2␣ the predominant mechanisms for control of translational initiation in this model? Liver homogenates separated by SDSpolyacrylamide gel electrophoresis were analyzed by Western immunoblotting with antibodies directed toward rpS6 and phosphorylated rpS6 (P-S6; Ser-235/236 and Ser-240/244). The immunoblots shown in A were analyzed by densitometry, the results of which are shown as the mean Ϯ 1 S.D. for the ratio of P-S6 to total S6 (B). *, p Ͻ 0.01 versus corresponding control group. These results were replicated in two additional experiments. Panel C shows the results of Western immunoblot analysis for phosphorylated (active) Akt and total Akt in duplicate liver homogenates prepared from control, fasted, and refed animals. The number of cycles is that figure for which semi-quantitation was optimal for analysis. c Semi-quantitation was optimal for liver RNA analysis. d Semi-quantitation was optimal for polysome analysis.

Materials-Primaria
eIF4G, eIF2␣, S6K1, and S6K2, as well as protein kinase A and protein kinase C inhibitors were obtained from Santa Cruz Biotechnology, Inc.
(Santa Cruz, CA). Protein A-Sepharose CL-4B, protein G-Sepharose 4FF and 7-methyl (7m) GTP-Sepharose 4B were purchased from Amersham Biosciences. [␥-32 P]ATP (3000 Ci/mmol) was purchased from PerkinElmer Life Sciences. TRI reagent was purchased from Molecular Research Center, Inc. (Cincinnati, OH). All primers for RT-PCR were purchased from Invitrogen except for ␤-actin primers, which were purchased from Clontech (Palo Alto, CA). Animals-Male Sprague-Dawley rats age 4 -5 weeks (Charles River Laboratories, Wilmington, MA) were used for all studies. Animals were fed standard laboratory chow ad libitum (control). Fasted animals had food withdrawn for 48 h. Where noted, animals were refed by replacing the laboratory chow in their cage cover and allowing them to eat ad libitum for 1 h. Rats were sacrificed under pentobarbital sodium anesthesia (50 mg/kg by intraperitoneal injection). Liver tissue was flash frozen in liquid nitrogen before storage at Ϫ70°C. All animals were treated humanely as per guidelines adhered to by the Rhode Island Hospital Institutional Animal Care and Use Committee.
Hepatocyte Isolation and Primary Culture-In vitro studies were done using fetal rat hepatocytes because of their high basal level of translation in vitro (as opposed to adult hepatocytes, which would require acute stimulation with insulin plus epidermal growth factor to activate translation). Fetal hepatocytes were isolated from liver obtained on embryonic day 19 by collagenase digestion (27). Immunocytochemical analyses have shown that these preparations consist of ϳ90% hepatocytes, with the remainder consisting of a mixture of nonparenchymal cell types. In vitro amino acid restriction was accomplished as described previously (28).
Preparation of Liver Homogenates and Cell Extracts-Rat liver homogenates were prepared for S6K assays using a 15 ml/g volume of S6K homogenization buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 10 mM sodium pyrophosphate, 100 mM NaF, 1.5 mM MgCl 2 , 1 mM EGTA, 200 M NaVO 3 , 1 M microcystin, and 10% glycerol) containing protease inhibitors (10 g/ml leupeptin, 10 g/ml aprotinin, and 34.4 g/ml 4-(2-aminoethyl)benzenesulfonyl fluoride). For the 7mGTP cap-binding studies, samples were prepared in 15 ml/g liver volume of a buffer containing 50 mM Tris (HCl), pH 7.4, 50 mM KCl, 1 mM EDTA, and 1 M microcystin with protease inhibitors. Following homogenization, Triton X-100 was added to S6K lysis buffer and Nonidet P-40 to the 7mGTP buffer to final concentrations of 1% and 0.5%, respectively. The samples were then incubated on ice for 30 min. The detergent-extracted homogenates were centrifuged at 1,000 ϫ g for 15 min. The resulting supernatant was centrifuged at 10,000 ϫ g for 20 min. The final supernatant was stored at Ϫ70°C pending analyses.
Cell extracts were prepared by lysing 2 ϫ 10 6 fetal hepatocytes/ 100-mm plate with 400 l of an ice-cold buffer containing 50 mM Tris, pH 7.4, 50 mM KCl, 1 mM EDTA, 1 M microcystin, and 0.5% Nonidet P-40 with protease inhibitors (as above). The mixture was sonicated then incubated on ice for 30 min. The resulting lysate was centrifuged for 20 min at 10,000 ϫ g, and the supernatant was frozen at Ϫ70°C until use.
Protein determinations were made using the bicinchoninic acid method (BCA, Pierce). Protein assays used bovine serum albumin as the standard.
RNA Isolation-RNA was extracted from liver using TRI reagent according to the manufacturer's instructions. To obtain total RNA from polysome pools, fractions were extracted twice using acid phenol:chloroform, pH 4.5 (Ambion, Austin, TX). The RNA was precipitated with ethanol overnight at Ϫ20°C and recovered by centrifugation at 4°C. The resulting RNA pellet was air-dried, dissolved in RNA storage solution (Ambion), and quantified by spectral absorption.
Protein Analyses-S6K1 and S6K2 activities were measured by immunoprecipitation kinase assay (14). Western immunoblotting was accomplished using standard electrophoresis and transfer methods with chemiluminescent detection (29). The association of 4E-BP1 and eIF4G with eIF4E was evaluated as their ability to bind to 7mGTP-Sepharose beads via the interaction with eIF4E. Sample preparation, 7mGTP-Sepharose affinity purification, and Western immunoblotting were carried out as described previously (29). PESTfind analysis of eIF4G used the European Molecular Biology Network-Austria site (emb1.bcc. univie.ac.at/embnet/tools/bio/PESTfind/).
Polysome Analyses-For the preparation of polysomes, a small piece of liver (50 -100 mg) was thawed on ice in 5 volumes of polysome buffer

FIG. 3. eIF2␣ phosphorylation in control and 48-h fasted rats.
Liver homogenates were separated by SDS-polyacrylamide gel electrophoresis and analyzed by Western immunoblotting with antibodies directed toward eIF2␣ and phosphorylated eIF2␣ (Ser-51; P-eIF2␣). The immunoblots shown in Panel A were analyzed by densitometry, the results of which are shown as the mean Ϯ 1 S.D. for the ratio of P-eIF2␣ to total eIF2␣ (B). *, p Ͻ 0.05 versus the control group.
X-100, 0.14 M sucrose) containing 100 g/ml heparin. Initial attempts to exclude the heparin resulted in polysomal disaggregation. The sample was first homogenized using a 1-ml Dounce homogenizer (Wheaton, IL) with a loose-fitting pestle. After the addition of 0.9 volume of buffer containing 10% Triton X-100 and 10% deoxycholate, the sample was re-homogenized on ice with a tight-fitting pestle. Nuclei and mitochondria were pelleted by microcentrifugation at 4°C for 2.5 min. The resulting cytosol preparation was diluted with an equal volume of polysome buffer containing 500 g/ml heparin. A 0.3-ml aliquot of this mixture was layered over a 0 -45% sucrose gradient made in 25 mM Tris (HCl), pH 7.4, 25 mM NaCl, 5 mM MgCl 2 . Ultracentrifugation was carried out at 4°C and 40,000 rpm in a SW40 Ti Beckman rotor. Tube piercing was accomplished using a Brandel apparatus that allowed collection starting with the densest portion of the gradient at the bottom of the tube. Absorbance at 254 nm was monitored directly using a flow-cell attached to a spectrophotometer. Twelve 1-ml fractions were collected, and pooled as follows: Fractions 1-4, long polysomes; fractions 5-8, intermediate polysomes; fractions 9 -12, short polysome/ monosome pool.
A small aliquot from each pool was prepared for Western immunoblotting. The remainder of the polysome pools was used for RNA extraction. Because heparin was present in the gradient, and because it potently inhibits PCR, the contaminating heparin was digested with heparinase. This was accomplished by incubating 1 g of RNA from pooled gradient fractions with 3 units of heparinase in a buffer containing 5 mM Tris (HCl), pH 8.0, 1 mM CaCl 2 , and 40 units of RNasin (Promega, Madison, WI) per sample. This mixture was incubated at 25°C for 2 h prior to analysis by RT-PCR.
Reverse Transcription and Polymerase Chain Reaction-cDNA was generated using 3 g of total RNA from liver or 1 g from polysome pools using SuperScript TM First-Strand Synthesis System (Invitrogen). Rat-derived sequence data were used for primer design except when not available (eIF4G), in which case the human sequence was used. PCR primers were designed using the MIT Primer3 design program (www.genome.wi.mit.edu//cgi-bin/primer/primer3.cgi). Ribosomal protein S8 (rpS8) primer sequences were provided by Dr. Scot Kimball (Pennsylvania State College of Medicine, Hershey, PA). RT-PCR was performed in a semi-quantitative manner, as described previously (30 -32). The primer sequences, PCR conditions, and expected product sizes are shown in Table I. The 5Ј-untranslated regions of ornithine decarboxylase (ODC) and signal transducer and activator of transcription 1 (STAT1) were folded using mfold (version 3.1) by Zuker and Turner (33,34).

RESULTS
Liver Mass in the Model of Starvation-Rats were fasted for 48 h by withdrawing food and providing water ad libitum. At the end of this period of starvation, liver mass was decreased as indicated by a 23% fall in liver:body mass ratio (control, 0.047 Ϯ 0.005 (mean Ϯ 1 S.D.); fasted, 0.036 Ϯ 0.002; p Ͻ 0.002). Liver protein content declined by 39% (control, 841 Ϯ 7.61 mg; fasted, 512 Ϯ 73 mg; p Ͻ 0.002). To determine if the decrease in mass and protein could be accounted for by a reduction in cell size, cryosections were stained with 4Ј,6-diamidino-2-phenylindole to detect nuclei and subjected to morphometric analysis. Results demonstrated a 49% increase in nuclei per unit surface area (control, 109 Ϯ 17 nuclei per high powered field; fasted, 162 Ϯ 10 nuclei per high powered field; p Ͻ 0.001), consistent with a marked decrease in hepatocyte size in the fasted animals.
Hepatic rpS6 and eIF2␣ Phosphorylation in 48-h Fasted Rats-Our initial studies were aimed at examining the phosphorylation state of rpS6, an indicator of the activation state of the mTOR pathway. The results we obtained (Fig. 1) were unexpected and counterintuitive. There was an increase in the amount of phosphorylated rpS6 in liver from fasted rats. In cultured cells, nutrient deprivation causes a precipitous drop in rpS6 phosphorylation (35,36). Therefore, we concluded that nutrient deprivation in vivo differs from starvation of cells in culture. To confirm that this increase could not be attributed to hormone-mediated signaling downstream from phosphatidylinositol 3-kinase, we examined the activation state of Akt (Fig.  1). As would be expected, hepatic Akt phosphorylation was decreased in the fasted animals and restored with refeeding.
Analysis for the activities of the rpS6 kinases S6K1 and S6K2 were consistent with the rpS6 phosphorylation results. S6K activities are high in cells involved in growth and proliferation and low in cells that are quiescent (22). It was therefore unexpected to find that S6K1 and S6K2 activities (Fig. 2) were 30 -40% higher in fasted compared with control rat livers. FIG. 4. Effect of altering leucine concentration on phosphorylation of rpS6 and eIF2␣. E19 hepatocytes in primary culture were exposed to either 0.4 mM (High) leucine or 0.04 mM (Low) leucine. Cell lysates were prepared and analyzed by Western immunoblotting for P-S6 (Ser-235/236) and total S6 (A) and phosphorylated/total eIF2␣ (B). In addition to the immunoblots, the densitometric analyses for the ratio of phosphorylated to total S6 or eIF2␣ are shown graphically in each panel. The numbers in parentheses above the bars show the raw densitometric data for the duplicate determinations.
These results were interpreted as indicating that under conditions of fasting hepatic mTOR signaling may increase rather than decrease.
Loss of eIF2 nucleotide exchange activity due to increased eIF2␣ phosphorylation is a mechanism that cells employ to shut down translational initiation (9). Based on prior observations that hepatocyte protein synthesis is inhibited during starvation, we predicted higher levels of eIF2␣ phosphorylation in fasted animals. We performed Western immunoblots for phosphorylated and total eIF2␣ in liver homogenates prepared from fed and fasted rats. Again, results we obtained (Fig. 3) were unexpected. eIF2␣ was relatively hypophosphorylated in livers of fasted animals.
As noted above, the regulation of translation is responsive not only to hormone and growth factor signals, but also to nutrient availability. Furthermore, fasting is associated with increased circulating concentrations of some amino acids, most notably BCAAs (25,26). To test the hypothesis that the changes in hepatic rpS6 and eIF2␣ phosphorylation could be accounted for by a response to elevated BCAA availability, we studied rat hepatocytes in primary culture. Cells were cultured overnight in the presence of 0.4 mM (high) or 0.04 mM (low) leucine. Results (Fig. 4) showed a direct association between leucine concentration and rpS6 phosphorylation, whereas the relationship between leucine concentration and eIF2␣ phosphorylation was an inverse association. These data are consistent with the hypothesis that changes in leucine availability could significantly contribute to the changes in these phosphorylation events that we observed in liver from the fasting rats.
eIF4E⅐eIF4G Complex Formation in Livers of Animals Fasted for 48 h-Having not identified the mechanism of hepatic translational inhibition in fasting, we used 7mGTP-Sepharose beads to affinity purify eIF4E and its binding partners 4E-BP1 and eIF4G. eIF4E binds to eIF4G and 4E-BP1 in a mutually exclusive manner; eIF4E has a single binding site that is shared by eIF4G and 4E-BP1 (37). Based on published data, we FIG. 5. Regulation of the hepatic eIF4E/eIF4G cap-binding complex in fasting. A, eIF4E-containing complexes, which include 4E-BP1 and eIF4G, were affinity-purified using 7mGTP-Sepharose beads from liver homogenates prepared from control and 48-h fasted rats. The affinity-purified proteins were analyzed by Western immunoblotting with antibodies directed toward eIF4E, 4E-BP1, and eIF4G. Densitometry was performed. The graphs show the ratio of 4E-BP1:eIF4E and eIF4G content. The numbers in parentheses above the bars show the raw densitometric data for the duplicate determinations. B, liver homogenates prepared from control, 48-h fasted, and 1-h refed rats were analyzed by Western immunoblotting for total eIF4E, 4E-BP1 and eIF4G. Similar results were obtained in a replicate experiment.
predicted that fasting would result in a marked decrease in 4E-BP1 phosphorylation with a resultant increase in 4E-BP1-eIF4E binding and a reciprocal decrease in eIF4G-eIF4E binding. Contrary to this expectation, fasting was associated with a decrease in the amount of 4E-BP1 that co-purified with eIF4E (Fig. 5A). However, we also saw a profound decrease in the amount of eIF4G bound to the 7mGTP beads (Fig. 5A). The explanation for this observation was apparent when we performed Western immunoblots on total liver homogenates for eIF4E, eIF4G, and 4E-BP1 (Fig. 5B). Although we found that the abundance of all the three proteins decreased in the fasted state, the decrease in eIF4G was disproportionately high. This raised the possibility that regulation of eIF4G abundance, not 4E-BP1 phosphorylation, may be responsible for translational inhibition of liver in vivo.
To determine if there is down-regulation of eIF4G secondary to a decrease in steady-state mRNA levels in fasted liver, we performed semi-quantitative RT-PCR (Fig. 6). eIF4G expression relative to ␤-actin was unaffected by fasting, indicating that translational and/or post-translational events account for the change in eIF4G abundance between livers of control and fasted rats.
Polysome Fractionation and RT-PCR of Control, Fasted, and Refed Rats-Post-nuclear supernatants of liver homogenates from control, 48-h fasted, and 1-h refed animals were fractionated on a sucrose gradient. The absorbance profile was used to confirm the fractions containing long polysomes, intermediate polysomes, and short polysomes/monosomes. We found that a duration of centrifugation of 105 min maximized the resolution of the fractionation. In an initial experiment (Fig. 7A), we analyzed 12 fractions across the entire polysome profile for rpS6 content. The results indicated that fasting was associated with a subtle shift from the fractions containing long polysomes to those containing intermediate polysomes. Under all conditions, a significant proportion of the rpS6 was present in fractions representing short polysomes and monosomes.
We then performed triplicate analyses comparing polysomes from control, fasted, and refed livers. For this experiment, shown in Fig. 7B, the polysome fractions were divided into three pools. Based on the absorbance profile for the experiment shown in Fig. 7A, we considered these pools as representing long polysomes (L), intermediate polysomes (I), and short polysomes/monosomes (S/M). These analyses confirmed that fasting was associated with a shift from long to intermediate polysomes. Samples obtained 1 h into refeeding showed a minimal effect on polysome distribution.
The polysome pools were analyzed using RT-PCR for the expression of eIF4G, rpS8 (a 5Ј-TOP mRNA), ␤-actin, and ODC and STAT1 (both with highly structured 5Ј-untranslated regions). The experiment was performed in triplicate. A representative experiment is shown in Fig. 7C. Our results showed a minimal effect of fasting and refeeding on the distribution of eIF4G, rpS8, and ␤-actin. Refeeding promoted an apparent shift in the distribution of the mRNA for ODC toward the longer polysomes. The most marked effect was seen with STAT1, which showed a clear shift toward the short polysomal/ monosomal fraction with fasting and back toward the long and intermediate polysome fractions with refeeding.
The Relationship between rpS6 Phosphorylation, Ribosome Distribution, and Translation-Liver polysome preparations from control, fasted, and refed rats were fractionated, pooled as above, and analyzed by Western immunoblotting for rpS6 and phosphorylated rpS6 (Fig. 8). To assess the stoichiometry of phosphorylation, results were expressed as the ratio of phosphorylated rpS6 to total rpS6. Triplicate analyses showed that rpS6 in the long and intermediate polysome fractions was relatively hyperphosphorylated under all conditions.
Translation of Specific mRNA in the Fasting Rodent Model-We reasoned that the data shown in Fig. 7C do not take into account the effect of the redistribution of ribosomes. We combined RNA abundance data with a measure of ribosome content based on rpS6 abundance by expressing our results as the product of the two. Thus, we were able to obtain an index of in vivo translation for individual RNA species. The results of this analysis (Fig. 9) showed varying patterns of re-distribution of transcripts for the different mRNAs that were studied. eIF4G showed no change in its distribution with fasting. Refeeding for 1 h was associated with a shift from the short polysome/monosome fraction to the long polysome fraction. The absence of an effect of fasting was noteworthy given the marked decrease in eIF4G abundance that was observed after 48 h of fasting. The shift that occurred with refeeding was consistent with the increase in the abundance of eIF4G that was observed.
rpS8 showed no redistribution under any of the three conditions. As described above, we had observed marked increase in the rpS6 phosphorylation in fasted compared with control livers. A correlation between S6 phosphorylation and translation of 5Ј-TOP mRNA would predict a large increase in translation of rpS8 in the fasted state. This was not the case.
The distribution of ␤-actin mRNA was not significantly affected by fasting. Like eIF4G, there was a shift in the distribution of ␤-actin mRNA to the long polysome fraction in the refed samples.
ODC mRNA distribution showed only a modest effect of fasting. In the control samples, the ODC mRNA content in the intermediate polysome pool was lower than in the other two pools. With fasting, there was a shift from the long to intermediate pools, consistent with a slowing of translation. There was no evidence for an activation of ODC translation upon refeeding.
As noted above, we chose to study STAT1 based on the structure of its 5Ј-untranslated region, which would predict a high degree of dependence on the presence of an active cap binding complex. Examination of the distribution of STAT1 mRNA was consistent with this prediction. It was significantly altered by fasting, with a movement toward shorter polysomes, while refeeding induced a marked shift toward long polysomes. DISCUSSION The mechanism by which hepatic protein synthesis is inhibited on prolonged fasting in the rodent remains undefined. A number of studies have shown that fasting in rodents is accom- panied by a decrease in liver size, liver protein content, and hepatocyte protein synthetic rate (1,(3)(4)(5). In the present study, we investigated translational control exerted by rpS6, eIF2␣, and the eIF4E-mRNA cap-binding complex in livers from fasted rats. eIF2␣ phosphorylation inhibits nucleotide exchange in the eIF2 complex, an event important for the delivery of methionyl-tRNA for translational initiation, thereby reducing protein synthesis (9). Global translational initiation is also regulated by mTOR-mediated phosphorylation of the eIF4E-binding protein, 4E-BP1 (38). Early studies on the phosphorylation of rpS6, a component of the 40 S subunit, indicated that it may be required for the translation of 5Ј-TOP mRNAs (12,13). 5Ј-TOP mRNAs may number as few as 100 to 200, but they can account for 20 -30% of total cellular mRNA. They encode many of the components of the translational apparatus, including ribosomal proteins and elongation factors, which are necessary for cell growth.
Our laboratory has focused on late gestation liver development in rodents as a model of liver growth and hepatocyte proliferation. We have shown that in fetal liver cellular proliferation can take place in the absence rpS6 phosphorylation (14). A number of recent reports have dissociated rpS6 phosphorylation from 5Ј-TOP mRNA translation (15)(16)(17). To better characterize the correlation between rpS6 phosphorylation and 5Ј-TOP mRNA translation, attempts have been made to inves-tigate the localization of phosphorylated rpS6 in translating polysomal versus non-translating subpolysomal fractions (39,40). However, these studies were done before the advent of phospho-specific anti-rpS6 antibodies, and they yielded conflicting results.
Our initial experiments led to the unexpected finding that phosphorylation of rpS6 is high in livers of fasted rats compared with control animals. This was surprising given the widely accepted correlation between rpS6 phosphorylation and cell growth and proliferation (22). We found a moderate but significant increase in the activities of S6K1 and S6K2. This increase did not appear to be sufficient to account for the profound increase in rpS6 phosphorylation in livers from fasted animals, raising the possibility that a decrease in phosphatase activity toward rpS6 could be involved. Studies by Olivier et al. (41) indicate that the catalytic subunit of protein phosphatase-1 mediates the dephosphorylation of ribosomal protein S6. Low protein phosphatase-1 expression, decreased localization to ribosomes through decreased expression of associated subunits, or inhibition of ribosomal protein phosphatase-1 activity could serve to maintain S6 in a hyperphosphorylated state, even in the absence of a marked increase in kinase activity.
We next examined phosphorylation of eIF2␣, an event that inhibits its ability to deliver the initiator methionyl-tRNA for translational initiation. We again obtained a counterintuitive result: phosphorylation of eIF2␣ was lower in fasted than in control livers.
Hyperbranched-chain aminoacidemia is known to occur in the fasting state in rodents and humans, primarily as a result of skeletal muscle catabolism. This provides substrate for the increase in hepatic gluconeogenesis and branched-chain amino acid oxidation in muscle that occur during fasting (25,26). We hypothesized that this increase in BCAA levels during fasting might contribute to increased signaling to rpS6 and eIF2␣ via mTOR. To examine this in an easily controlled environment, we studied two different concentrations of leucine using hepatocyte primary cultures. Consistent with our hypothesis, we were able to show a direct association between leucine concentration and rpS6 phosphorylation and an inverse relationship between leucine concentration and eIF2␣ phosphorylation. In other studies, 2 we have found that, unlike in yeast, rapamycin does not increase eIF2␣ phosphorylation in vivo in liver. Taken together, these results indicate that this event may be mTORdependent but rapamycin-resistant in hepatocytes in vivo.
Given that our rpS6 and eIF2␣ studies did not provide for a mechanism for hepatocyte translational inhibition during fasting, we turned our attention to the study of the mRNA capbinding complex. Based on the voluminous literature on phosphorylation of 4E-BP1 as a mechanism of global translational inhibition (38), we expected to find 4E-BP1 in a hypophosphorylated state and bound to eIF4E during fasting. The results we obtained were unexpected in that the ratio of 7mGTP-bound 4E-BP1⅐eIF4E was lower in fasted compared with control livers. Despite the apparent increased activation state of eIF4E in fasting, the amount of eIF4G co-purifying with 7mGTP-Sepharose beads was profoundly diminished in livers of fasted animals. This could be accounted for by a marked decrease in eIF4G abundance.
Based on the decreased eIF4G abundance in fasting, we made several predictions. Given the central role of eIF4G in linking mRNA to the ribosome, we expected that many mRNAs (including eIF4G mRNA) would not be associated with polysomes. By extension, we anticipated that there would be a reduced abundance of long polysomes in fasted livers with a shift toward the short polysomal/monosomal fraction. In addition, specific mRNAs, such as ODC and STAT1 that have highly structured 5Ј-untranslated regions would be predicted to have an increased dependence on the RNA helicase eIF4A. The secondary structure energy of the 5Ј-untranslated regions of rat ODC and STAT1 are extremely high: Ϫ238.9 and Ϫ91.5 kcal/mol, respectively. Therefore, one would expect that ODC and STAT1 translation would be more profoundly inhibited in the fasted animal than many other mRNA species. A caveat is that the mRNA for ODC contains an internal ribosome entry site (IRES), which under certain conditions precludes the requirement for the canonical mRNA cap-binding proteins for translational initiation. This is reflected in the previously published observation that ODC translation during G 2 /M is not sensitive to rapamycin despite inhibition of the eIF4E⅐eIF4G cap-binding complex, a presumed reflection of translational activation via an IRES-dependent mechanism (42).
Polysomal fractionation and Western immunoblotting of the long, intermediate, and short polysome/monosome fractions for rpS6 confirmed that some polysomal disaggregation occurs during fasting. In addition we examined the polysomal local-2 P. Anand and P. A. Gruppuso, manuscript in preparation. To combine data from three separate experiments, all results were first normalized by expressing the data for an individual fraction as the proportion of the sum of the three fractions for that analysis. Data for phosphorylated rpS6 were then expressed as a ratio to the total rpS6 in that fraction. The asterisks indicate a significant difference versus the other two fractions for the same condition, as determined by one-way analysis of variance with a Tukey post-hoc test.
ization of phosphorylated rpS6, the role of which is a topic of intense debate in the translational control arena. rpS6 is phosphorylated in a sequence starting at amino acid residue Ser-236 and then proceeding to Ser-235, Ser-240, Ser-244, and Ser-247 in that order (43,44). We examined phosphorylation at Ser-235/236 and Ser-240/244. There was a distinct and marked preferential localization of both phosphorylated species to the polysomal fractions. Based on this result, it appears that rpS6 phosphorylation may play a role in promoting translation.
We reasoned that uncorrected results of PCR analyses would be an inadequate indicator of the distribution of specific RNAs between long, intermediate, and short polysomes/monosomes. Given that rRNA constitutes 80 -85% of the RNA of a eukaryotic cell (45,46), we knew that RT-PCR analysis standardized per unit RNA would yield results that were independent of ribosome distribution. We took advantage of the fact that ribosomal proteins, including rpS6, and rRNA exist in a strict stoichiometry (47). This allowed us to correct the mRNA distribution for ribosome content by using rpS6 content as a surrogate for rRNA. We were thus able to examine the effect of fasting and refeeding on the distribution of specific RNAs between the three pools.
An examination of these analyses reveals one finding that was relatively unexpected. Our data did not show a generalized decrease in translation rates with fasting. In fact, the only mRNA that behaved as expected was STAT1, and even the STAT1 results showed only a modest effect of fasting. These results may reflect a relatively great contribution of autophagy in mediating the marked decrease in hepatic protein content that is seen after 48 h of fasting. Given our observation that eIF4G, an essential element of the cap binding complex, is markedly reduced in content by fasting, our data may also indicate that a significant proportion of hepatic protein synthesis is independent of the action of the cap binding complex. Our data do not provide any insight into alternative mechanisms that might reconcile these findings.
rpS8 translation was minimally affected by fasting despite higher levels of rpS6 phosphorylation in fasted compared with control livers. These data add to the body of evidence that indicates that rpS6 phosphorylation does not have a direct regulatory role for the control of 5Ј-TOP mRNA translation (15)(16)(17).
Like rpS8, ODC was minimally affected by fasting even though its 5Ј-untranslated region predicts regulation by the cap binding complex. This may relate to activity of the ODC IRES allowing its translation in the fasted animal.
Similar to ODC, eIF4G mRNA was not disproportionately redistributed to the short polysome/monosome fraction under conditions of fasting. However, it did show redistribution to long polysomes in response to refeeding. The marked reduction of eIF4G abundance in fasted liver would appear to require a form of autoregulation for eIF4G levels to rapidly recover with refeeding. Given the profound loss of eIF4G during fasting, its synthesis during the earliest stages of refeeding (see Fig. 5) could not result from a marked increase in the activity of the cap binding complex. We hypothesize that the presence of an IRES in eIF4G and short polysome/monosome (solid bars) fractions were prepared by sucrose density centrifugation. mRNA abundance in each pool was corrected for ribosome content in that pool. The latter was determined by Western immunoblotting for rpS6 content. To normalize the data between the triplicate experiments that were performed, the signal for each pool was first converted to a proportion of the total of the three pools combined. Results, expressed as the mean plus the standard error of the mean, are shown for the following mRNAs (top to bottom): eIF4G, rpS8, ␤-actin, ODC, and STAT1. Asterisks indicate results that are significantly different from the other two fractions for that condition, as determined by one-way analysis of variance with a Tukey post-hoc test. mRNA (48) mediates a rapid increase in eIF4G during refeeding, thus replenishing eIF4G protein that can then function to mediate cap binding complex-initiated translation.
Our data showing the absence of an effect of fasting on the polysomal distribution of eIF4G mRNA indicates that the loss of eIF4G during fasting is not necessarily a function of a cap binding complex-mediated decrease in eIF4G translation. Rather, the loss of eIF4G may be a result of its proteolytic degradation. This could be part of the autophagy response to fasting (2,7), which is rapidly and potently reduced by refeeding. The eIF4G protein sequence supports the hypothesis that its degradation may be important to the regulation of its abundance. eIF4G contains a number of PEST sequences with high PESTFIND scores (residues 252-335 (ϩ10.65); 361-430 (ϩ12.51); 431-476 (ϩ16.85); and 532-574 (ϩ18.66)). PEST sequences are hydrophilic stretches of amino acids enriched in proline (P), glutamic acid (E), serine (S), and threonine (T) that target proteins for rapid destruction. During the past several years, evidence has accumulated that PEST regions serve as proteolytic signals in certain conditions. PESTFIND scores above zero denote a possible PEST region, whereas values Ͼϩ5, as in this case, strongly suggest regulated proteolysis (49,50). We did not detect proteolytic degradation products of eIF4G in our analyses, but this does not rule out the possibility of regulated proteolysis as the primary mode of control of eIF4G abundance in our model.
In summary, our studies suggest that previously accepted mechanisms for global translational regulation cannot account for changes in hepatic translation during fasting. Problems arise in the assignment of a key role for mTOR signaling given the increase in circulating BCAAs that occurs in the fasted state. Our data indicate that translation can be down-regulated in the face of increased mTOR signaling and that this downregulation is associated with a decrease in eIF4G abundance. Thus, hepatic translation during periods of prolonged fasting may shift to mechanisms that are independent of the cap binding complex. Alternatively, there may be functional surrogates for eIF4G that were not identified in our studies. Our data also provide further evidence that 5Ј-TOP mRNA translation can be dissociated from rpS6 phosphorylation but that phosphorylated rpS6 localizes to translating polysomes, suggesting an as of yet uncharacterized role for rpS6 in translation control.