Amino Acid Sufficiency and mTOR Regulate p70 S6 Kinase and eIF-4E BP1 through a Common Effector Mechanism*

The present study identifies the operation of a signal tranduction pathway in mammalian cells that provides a checkpoint control, linking amino acid sufficiency to the control of peptide chain initiation. Withdrawal of amino acids from the nutrient medium of CHO-IR cells results in a rapid deactivation of p70 S6 kinase and dephosphorylation of eIF-4E BP1, which become unresponsive to all agonists. Readdition of the amino acid mixture quickly restores the phosphorylation and responsiveness of p70 and eIF-4E BP1 to insulin. Increasing the ambient amino acids to twice that usually employed increases basal p70 activity to the maximal level otherwise attained in the presence of insulin and abrogates further stimulation by insulin. Withdrawal of most individual amino acids also inhibits p70, although with differing potency. Amino acid withdrawal from CHO-IR cells does not significantly alter insulin stimulation of tyrosine phosphorylation, phosphotyrosine-associated phosphatidylinositol 3-kinase activity, c-Akt/protein kinase B activity, or mitogen-activated protein kinase activity. The selective inhibition of p70 and eIF-4E BP1 phosphorylation by amino acid withdrawal resembles the response to rapamycin, which prevents p70 reactivation by amino acids, indicating that mTOR is required for the response to amino acids. A p70 deletion mutant, p70Δ2–46/ΔCT104, that is resistant to inhibition by rapamycin (but sensitive to wortmannin) is also resistant to inhibition by amino acid withdrawal, indicating that amino acid sufficiency and mTOR signal to p70 through a common effector, which could be mTOR itself, or an mTOR-controlled downstream element, such as a protein phosphatase.

Brief starvation engenders a decrease in protein synthesis, particularly in skeletal muscle, which is rapidly reversed on refeeding (1). The contribution of the nutrients themselves to the regulation of protein synthesis as compared with the concomitant hormonal responses has been the object of considerable study. Insulin, e.g., is well known to stimulate protein synthesis in skeletal muscle; however, much evidence indicates that a major portion of the increase in the skeletal muscle protein synthesis in vivo seen on refeeding is independent of changes in insulin and perhaps attributable to the nutrients themselves (2). Nevertheless, in vivo it is difficult to isolate the contributions of amino acids and other nutrients from those of insulin, insulin-like growth factors, and other regulators whose availability in vivo is altered by nutrients.
Mammalian cells in culture exhibit an inhibition of overall protein synthesis with depletion of medium amino acids, which is quantitatively substantial, but rapidly reversible (3). As with short term fasting in vivo (4), amino acid withdrawal in vitro is characterized by a loss of polysomes and an increase in monomeric ribosomes pointing to a block in peptide chain initiation (5,6). As regards the mechanism of translational inhibition, attention focused initially on the role of eIF-2␣ phosphorylation. Phosphorylation of eIF-2␣ at Ser-51 enhances its affinity for eIF-2B, resulting in sequestration of this critical, low abundance guanyl nucleotide exchange factor, and thereby inhibiting regeneration of the eIF-2/GTP necessary to bind Met-tRNA i (7,8). In Saccharomyces cerevisiae, the depletion of amino acids leads to the activation of the eIF-2␣ kinase known as GCN2, presumably through the accumulation of uncharged tRNA (9). The occurrence of increased eIF-2␣ phosphorylation has been well documented in a variety of mammalian cells deprived of amino acids, and in CHO 1 lines that contain mutant aminoacyl tRNA synthetases (reviewed in Ref. 3).
More recently, the eIF-4F complex has emerged as an important site of translational control (reviewed in Ref. 10). This complex contains the scaffold protein eIF-4G, which binds eIF-4E, the m 7 GTP-cap binding protein, eIF-4A, an RNA helicase and the RNA-binding protein eIF-4B. The eIF-4E component binds the 5Ј mRNA cap; 4A/4B then unwind the secondary structure in the 5Ј-untranslated segment, thereby enabling the 40 S-Met-tRNA i complex, aided by eIF-3, to efficiently scan the mRNA 5Ј-untranslated segment to the AUG start site. Both the assembly and activity of the 4F complex are regulated by phosphorylation. In mammalian cells, eIF-4E, -4B, and -4G are each phosphorylated in vivo in a manner that correlates positively with overall translational rate. The regulation of eIF-4E has been characterized most extensively (10). Phosphorylation of 4E, predominantly at Ser-209 (11,12), occurs in response to insulin, mitogens (6,10), or stress (13) and is accompanied by a 3-fold increase in 4E affinity for mRNA (14). In addition, the availability of 4E for incorporation into the 4F complex is independently regulated through the PHAS/4E-BP polypeptides. PHAS1 was purified as a heat-and acid-soluble polypeptide that exhibited rapid and multiple phosphorylation in vivo in response to insulin and mitogens (15), and was independently isolated as an eIF-4E-binding protein by interaction cloning (16). The 4E-BP1/PHAS1 polypeptide binds to 4E in a manner that is competitive with 4G (17); thus, the insulin/ mitogen-stimulated phosphorylation of 4E-BP1, which inhibits 4E-BP binding to 4E (16,18), makes 4E available for incorporation into the 4F complex. Although 4E and 4E-BPs are coordinately regulated by insulin and mitogens, examples of discordant regulation are common (13,19), indicating that the phosphorylation of the two polypeptides is independently mediated.
Evidence from in vivo models points to an important role for the 4E-BPs in translational regulation by nutrients. Thus, the phosphorylation of 4E-BPs in mouse skeletal muscle is diminished by starvation and restored by refeeding. Concomitantly, the association of 4E with 4E-BP1 increases with starvation and diminishes with refeeding whereas the association of 4E with 4G, as expected responds in the reciprocal manner (20). As regards the relative importance of insulin versus nutrients in the control of 4E-BP1 phosphorylation, it is notable that the changes in phosphorylation of 4E-BP in response to fasting and feeding remain intact in NOD and ob/ob mice (20), diabetic models who fail to show any change in insulin concentration in response to these dietary maneuvers (2).
The phosphorylation of eIF-4E is not altered by fasting and refeeding; however, the phosphorylation of the p70 S6 kinase responds to fasting and refeeding in parallel to 4E-BP1 (20). The p70 S6 kinase participates in translational control in a manner quite distinct from eIF-4F. This insulin/mitogen-activated protein kinase, whose major known substrate is the 40 S ribosomal subunit protein S6 (21), is critical for the translation of a subclass of mRNAs whose 5Ј-untranslated region contains a short oligopyrimidine sequence immediately after the 5Ј cap (22)(23)(24). Nevertheless, despite their quite different roles in translational control, it appears that the phosphorylation of both 4E-BP and p70 S6 kinase, but not eIF-4E, is coordinately controlled not only by insulin and mitogens, but independently by nutrients, as reflected in vivo by the responses to fasting and feeding (20).
Another response shared by 4E-BPs (25)(26)(27) and the p70 S6 kinase (28,29), but not by eIF-4E, is a susceptibility to dephosphorylation in vivo in response to the macrolide immunosuppressant, rapamycin. This effect of rapamycin is mediated indirectly, through the ability of rapamycin, in complex with FKBP12 to bind to the mTOR kinase and inhibit its activity (30). Mutant mTORs that are unable to bind rapamycin/ FKBP12 can protect p70 S6 kinase and 4E-BP1 from rapamycin-induced dephosphorylation (30,31). Recently, mTOR has been shown to phosphorylate 4E-BP1 directly in vitro, at sites corresponding to those phosphorylated in vivo during insulin stimulation (58,59). Nevertheless, the kinase activity of mTOR is not altered by insulin treatment prior to extraction (at least as measured after mTOR immunoprecipitation, Refs. 31 and 32) and the physiologic regulators of the TOR kinase are not known. TORs are 280-kDa polypeptides whose kinase domain is most closely related to the checkpoint control kinases, ATM, TEL1, MEC1, and the DNA protein kinase (33). In S. cerevisiae, loss-of-function mutations in both TOR genes, or treatment with rapamycin, each cause a profound inhibition of overall translational initiation and elicit a phenotype characteristic of starved cells entering G 0 , the stationary phase (34).
The biochemical steps through which the TORs regulate translation in yeast are not known (35,36); however, they appear to participate in a nutrient-sensing pathway that controls protein phosphatase activity (37). The possibility that mTOR might participate in nutrient-sensing in mammalian cells was raised by the observation that hepatocytes incubated in the absence of amino acids exhibit a selective decrease in the phosphorylation of S6, that is rapidly reversed by addition of amino acids but blocked by rapamycin (38). These findings imply that amino acids, independent of insulin or mitogens, can regulate the activity of the p70 S6 kinase through an mTOR-dependent mechanism. We therefore set out to characterize the pathways that regulate the phosphorylation of p70 S6 kinase and eIF-4E BP1 in response to amino acid sufficiency, mTOR and receptor tyrosine kinases. Our results provide evidence that amino acids and mTOR signal to the translational apparatus through convergent pathways distinct from those controlled by the insulin receptor.
Antibodies-A monoclonal antibody against influenza virus hemagglutinin (12CA5) was purchased from Boehringer Mannheim; a monoclonal anti-FLAG antibody (M2) was from Eastman Kodak Corp. The monoclonal antibody 12CA5 was used to recover HA-tagged polypeptides. Phosphospecific antibody against p70 S6 kinase were a gift from New England Biolabs Inc. Phosphospecific and nonspecific antibody against ERK1/2 were purchased from New England Biolabs Inc. A polyclonal antiserum raised against a synthetic peptide corresponding to amino acids of 337-352 of p70␣1 S6 kinase was used for immunoblot of p70 (39) and a antiserum against GST fusion protein of C-terminal 104 amino acids of p70 was used for immunoprecipitation of endogenous p70. A polyclonal antiserum against mTOR was raised against GST fusion protein of amino-terminal 100 amino acids of mTOR expressed in Escherichia coli (31). A polyclonal antibody against eIF-4E BP1 was raised against a GST-eIF-4E BP1 fusion protein expressed in E. coli and purified by chromatography using GST-eIF-4E BP1.
Cell Culture-HEK293 cells (40) and CHO-IR (41) cells were cultured as described earlier in DMEM or Ham's F-12 medium, respectively. For the starvation of amino acids, cells were first incubated in DMEM (293) or Ham's F-12 (CHO-IR) medium without FCS for 16 h, washed once with Dulbecco's phosphate-buffered saline (D-PBS, containing 0.1 g/liter CaCl 2 ) and incubated in the same buffer for the times indicated for up to 2 h. Readdition of amino acids involved changing the medium to D-PBS containing individual amino acids or a mixture of amino acids as indicated. The concentration of each amino acids designated as 1ϫ is as follows (in mg/liter): L-Arg, 84; L-Cys, 48; L-Glu, 584; L-His, 42; L-Ile, 105; L-Leu, 105; L-Lys, 146; L-Met, 30; L-Phe, 66; L-Thr, 95; L-Trp, 16; L-Tyr,72; L-Val, 94. A mixture of all these amino acids, each at this concentration is designated as the "1ϫ amino acid mixture". Transient transfection was performed by lipofection method using LipofectAMINE (Life Technologies, Inc.) as described previously (31).
Supernatants immunoblotted for endogenous ERKs were separated on SDS-PAGE, transferred to PVDF membrane, and blotted with phosphospecific or nonspecific antibody against ERKs.
To detect eIF-4E BP1, the supernatants were heated at 100°C for 5 min and precipitated proteins were removed by centrifugation at 10,000 g for 30 min (31). The heat-stable proteins were separated on SDS-PAGE and immunoblotted with anti-FLAG antibody for detection of recombinant FLAG-eIF-4E BP1 or with anti-eIF-4E BP1 antiserum for detection of endogenous eIF-4E BP1.
Kinase Assay-p70 S6 kinase activity was determined by immunocomplex assay by using 40 S ribosomal subunits as substrate (42). PI 3-kinase activity coimmunoprecipitated with anti-phosphotyrosine an-tibody (PY-20) was determined as described previously (43). c-Akt/PKB kinase activity was measured by immunocomplex assay using myelin basic protein as substrate. CHO-IR cells transfected with HA-c-Akt/ PKB were lysed in ice-cold buffer A and immunoprecipitated with anti-HA antibody. The immunoprecipitates were washed twice with buffer A containing 0.5 M NaCl, and twice with the buffer composed of 20 mM Tris (pH ϭ 7.4) and 1 mM DTT. The kinase reaction was started by adding the reaction mixture (50 mM Tris (pH ϭ 7.4), 10 mM MgCl 2 , 1 mM DTT, 50 M ATP, 1 M protein kinase inhibitor (PKI) peptide, 1 mM EGTA, 1 mg/ml myelin basic protein, and 2 Ci of [␥-32 P]ATP), incubated at 30°C for 30 min, and stopped by adding SDS sample buffer. The mixture was separated by SDS-PAGE, transferred onto PVDF membrane, and analyzed by autoradiography.

RESULTS
Amino Acids Regulate the Phosphorylation and Activation of p70 S6 Kinase-Withdrawal of amino acids from the medium leads to a rapid, reversible deactivation of the p70 S6 kinase and partial dephosphorylation of eIF-4E BP1. This response was observed in the three cell types examined: CHO-IR, HEK293, and Swiss 3T3 cells. As regards p70, the already low activity of endogenous p70 in serum-deprived CHO-IR cells (henceforth called the "basal" activity) declines further during the 2 h following amino acid withdrawal, and the ability of insulin to stimulate activity is virtually abolished (Fig. 1, A and  B). This inhibition of p70 activity is fully reversible; reintroduction of the mixture of amino acids at the concentrations used in DMEM (a concentration henceforth called 1ϫ (see "Experimental Procedures"), and used as a reference level for all responses examined), restores both basal and insulin/serum-stimulated p70 kinase activity within 30 min (Figs. 1 (A and B) and 2). A similar response is seen in 293 cells, examining either endogenous p70 (Fig. 1D) or transiently expressed recombinant p70␣1 ( Fig. 1C and 6A). The decline in p70 activity that occurs after withdrawal of amino acids is not due solely to the loss of insulin/serum stimulation, because the p70 activ- After extraction, the endogenous p70 S6 kinase was immunoprecipitated with polyclonal anti-p70 antibody and subjected to immunocomplex assay using 40 S ribosomal subunits as substrate. B, amino acid withdrawal causes a reversible inhibition of insulin-stimulated p70 S6 kinase activity in CHO-IR cells. CHO-IR cells grown to confluence in Ham's F-12 medium were deprived of serum for 16 h. At time 0Ј cells were extracted directly (OE), or 10 min after the addition of insulin (100 nM) (f). The remaining cells were transferred to D-PBS lacking amino acids. Insulin (100 nM) was added into the D-PBS 10 min prior to extraction, and these incubations were terminated at the times indicated (f). After 120 min, two plates were transferred to D-PBS containing 1ϫ amino acids for an additional 30 min prior to extraction (f, OE); insulin was added to one of these (f) for the last 10 min. Endogenous p70 was immunoprecipitated with anti-p70 antibody and the kinase activity were determined by immunocomplex assay. C, the effect of amino acid withdrawal and rapamycin on the activity of recombinant p70 S6 kinase. HEK293 were transfected with Flag-p70. Thirty two hrs. later, the cells were transferred to DMEM lacking serum; after 16 h, the cells were stimulated by addition of 10% FCS for 10 min. Some cells were then harvested (lane 1), or the medium was changed to D-PBS, lacking (lanes 2-8) or containing (lanes 9 -13) 1ϫ amino acids, for the times indicated prior to harvest. In lane 7, cells deprived of amino acids for 120 min were incubated another 30 min in D-PBS containing 1ϫ amino acids. In lane 13 and 14, rapamycin (200 ng/ml) was added to cells in DMEM, 10 min after the addition of 10% FCS, and harvested thereafter at the times indicated. Flag-p70 was immunoprecipitated with anti-Flag antibody and subjected to p70 S6 kinase assay using 40 S subunits. The reaction mixture was separated by SDS-PAGE, transferred onto PVDF membrane and subjected to autoradiography (upper panel) or Western blot with polyclonal anti-p70 antibody (lower panel). D, recovery of basal and serum-stimulated p70 S6 kinase activity in 293 cells after readdition of amino acids. Confluent 293 cells were incubated in DMEM lacking FCS for 16 h, followed by incubation in D-PBS for 2 h. At t ϭ 0, the cells were transferred to D-PBS containing 1ϫ amino acids mixture, with (dashed line) or without (solid line) 10% FCS, for times indicated prior extraction. Endogenous p70 S6 kinase activity was determined by immunocomplex assay. ity in insulin/serum-deprived cells also diminishes greatly. This is best appreciated in 293 cells, where the basal activity of p70 is 30 -60% of the maximally stimulated activity (Fig. 1D); amino acid withdrawal, after a slight lag, results in a marked decrease in the already high basal p70 activity that is reversible on readdition of amino acids (Figs. 1D and 6A). The changes in p70 activity are specific for amino acid withdrawal and addition; omission and reintroduction of glucose or the vitamins found in DMEM do not alter p70 activity. The ability of p70 to be activated by serum or insulin is fully restored by concentrations of amino acids far below those that maximally increase basal activity (Fig. 2). In CHO-IR cells, full restoration of the response to insulin and serum is evident after readdition of the complete amino acid mixture at 0.25ϫ concentration, and the absolute magnitude of p70 activity after addition of insulin/ serum is not altered as the concentration of the amino acid mixture is raised to 2ϫ. In contrast, the basal activity of p70 rises progressively as the concentration of the amino acid mixture is raised, so that at the 2ϫ concentration, p70 kinase activity is near maximal and shows little or no further increase on addition of insulin. The response of p70 to 10% undialyzed serum parallels the response to insulin, allowing for the amino acid content of the serum itself (Fig. 2). The effects of medium amino acids on the activity of p70 are due to overall amino acid sufficiency, although the effects of individual amino acids are not equal (Table I). Thus, little or no restoration of p70 activity occurs with readdition of any single amino acid. Readdition of a complete mixture of amino acids at 1ϫ, but lacking single amino acids provides the most informative data (Table I). This removal of Arg or Leu is most inhibitory (70 -90%), Lys or Tyr omission next in potency (approximately 30 -50% inhibition), whereas Cys or Glu omission are well tolerated, and His, Iso Met, Phe, Thr, Trp, and Val omission give moderate, variable inhibition under the conditions examined. Despite the potent inhibition caused by Arg or Leu omission, readdition of either individually does not stimulate p70 activity at all, and the two together enable only slight reactivation.
The kinase activity of p70 is dependent on the multisite phosphorylation of the p70 polypeptide (40, 44 -46). At least three sets of independently regulated site-specific p70 phosphorylations have been identified; one is a set of Ser/Thr-Pro motifs clustered in a psuedosubstrate autoinhibitory domain in the noncatalytic carboxyl-terminal tail (Ser-434, -441, and -447; Thr-444 in p70␣1) 2 (45,48). A second site is Thr-412, 2 located in a unique hydrophobic segment; homologous sites of regulatory phosphorylation have been identified in PKCs and c-Akt/ PKB (45). A third site is Thr-252, 2 located in the catalytic loop, a site of activating phosphorylation in many protein kinases (40,(45)(46)(47). Phosphorylation at each of these sites is stimulated by insulin or serum, and the phosphorylation of Thr-412 (45) and Thr-252 (40) are each indispensable for p70 activity. The protein kinase 3-phosphoinositide-dependent protein kinase-1 (PDK1) phosphorylates p70 Thr-252 in vitro and activates the S6 kinase activity (46,47); the extent of activation is controlled by a powerful positive site-site interaction between Thr-252-P and Thr-412-P (46). Anti-phosphopeptide antibodies toward several p70 phosphorylation sites have been developed, which are reactive with p70 only when the Ser/Thr residue is phosphorylated (48). These have been employed to examine the effects of insulin and amino acid withdrawal on p70 phosphorylation. Thus, insulin increases the phosphorylation of the proline-directed sites Ser-434, Thr-444, and Ser-447, as well as Thr-412 (Fig. 3); these insulin-stimulated p70 phosphorylations are strongly inhibited by amino acid withdrawal, concomitant with the inhibition of p70 S6 kinase activity. As observed during inhibition by rapamycin and wortmannin (48), the decline in 40 S kinase activity tracks most closely with the phosphorylation of Thr-412 (Fig. 3). Moreover, the activation of p70 S6 kinase in CHO IR cells induced by increasing ambient amino acid concentration to 2ϫ is accompanied by increased p70 phosphorylation at (at least) some of the relevant sites (data not shown). Thus, the proximate cause of p70 inhibition by amino acid withdrawal is dephosphorylation of p70 at multiple sites, and the stimulation of p70 activity by increasing extracellular amino acid concentrations is probably caused by increased p70 phosphorylation.
Amino Acids Regulate the Phosphorylation of eIF-4E-BPs-Withdrawal of medium amino acids leads to a reversible dephosphorylation of eIF-4E-BPs. Endogenous eIF-4E-BP polypeptides can be detected in extracts of CHO IR cells by immunoblot, using a polyclonal anti-eIF-4E BP1 antiserum, and are visualized after SDS-PAGE as a ladder of bands. Exposure of the cells to insulin diminishes the abundance of the more rapidly migrating bands and increases the abundance of the 2 Subtract 23 for the amino acid number in p70␣2.  slowest migrating bands, a response shown previously to be due to increased eIF-4E-BP phosphorylation (25-27, 31, 32). Omission of medium amino acids is accompanied by the disappearance of the most slowly migrating eIF-4E-BP polypeptide band, and a downshift of eIF-4E-BP polypeptides into two rapidly migrating bands; a similar downshift in eIF-4E-BP mobility occurs in response to rapamycin and wortmannin (Fig.  4A). Addition of insulin in the absence of amino acids fails to cause an upshift in the mobility of eIF-4E-BPs; readdition of amino acids restores the most slowly migrating eIF-4E-BP polypeptide, as well as the response to insulin (Fig. 4A). A similar pattern is observed for recombinant eIF-4E-BP1 expressed transiently in 293 cells. Removal of medium amino acids after stimulation by 10% FCS, leads, after a slight lag, to the disappearance of the most slowly migrating eIF-4E-BP1 polypeptides (Fig. 4B). Readdition of amino acids results in a time-dependent reappearance of the hyperphosphorylated eIF-4E-BP1 polypeptides. As with p70, the response to serum is restored by amino acids before the basal phosphorylation of eIF-4E-BP1 is fully recovered (Fig. 4C).
Amino Acids Do Not Regulate Insulin Receptor Kinase, PI 3-Kinase, c-Akt/PKB, or MAPK Activities-Inasmuch as the withdrawal of amino acids from CHO-IR cells reversibly blocks the ability of insulin to stimulate the phosphorylation and activation of p70 S6K and the eIF-4E-BPs, we examined the effects of amino acid withdrawal on several steps in insulin signal transduction upstream of, or parallel with the activation of the p70 S6 kinase (Fig. 5, A-D). Amino acid withdrawal has no effect on insulin-stimulated receptor tyrosine autophosphorylation, or on the tyrosine-specific phosphorylation of endogenous IRS proteins (Fig. 5A). The insulin-stimulated increase in PI 3-kinase immunoprecipitated by anti-phosphotyrosine antibodies, which is mostly bound to the IRS proteins (41), is slightly (ϳ20%) decreased after 2 h of amino acid withdrawal, a decrease that is not corrected by readdition of amino acids (Fig. 5B). Nevertheless, the very potent insulin activation of recombinant c-Akt/PKB, expressed transiently in CHO-IR cells, is unaffected by amino acid withdrawal (Fig. 5C). The  1-11); some cells were harvested directly thereafter (lane 1), whereas the remainder were transferred into D-PBS lacking (lanes 2-6) or containing a 1ϫ amino acid mixture (lanes 7-11) for the indicated times prior to extraction. Cell extracts were heated at 100°C for 5 min, and the heat-stable proteins were separated by SDS-PAGE and immunoblotted with anti-Flag antibody. C, amino acids restore basal and serum-stimulated eIF-4E BP1 phosphorylation. HEK293 cells transfected with Flag-tagged eIF-4E BP1 were incubated in DMEM lacking FCS for 16 h (lanes 1-11). In lane 1, the cells were harvested directly thereafter, whereas the rest were transferred to D-PBS for 2 h. Some cells were harvested thereafter (lane 2), whereas the rest were transferred to D-PBS containing 1ϫ amino acids lacking (lanes 3-5) or containing (lanes 6 -8) 10% FCS, for the times indicated. In lanes 9 -11, the cells in serum-free DMEM were treated with 200 ng/ml rapamycin for the times indicated. The supernatants from heat-treated cell extracts were prepared as described in B, subjected to SDS-PAGE, and immunoblotted with anti-flag antibody.
insulin-stimulated phosphorylation at the activating sites on the endogenous p42/p44 MAPKs, monitored by anti-P-peptide immunoblot, is slightly diminished after amino acid withdrawal, but is not restored by amino acid readdition (Fig. 5D). In addition, neither amino acid withdrawal nor readdition altered the activation loop phosphorylation of stress-activated protein kinase (SAPK) or p38 endogenous to the CHO-IR cells (data not shown). Thus, the effects of amino acid withdrawal/ readdition on basal and insulin-activated Ser/Thr phosphorylation of p70 S6 kinase and eIF-4E BP1 reflects the operation of a signal transduction response clearly distinguished from those recruited by insulin or generalized stress.
A p70 Mutant Resistant to Rapamycin Is Also Resistant to Amino Acid Withdrawal-The selective inhibition of p70 S6 kinase and eIF-4E BP1 phosphorylation that occurs consequent to amino acid withdrawal corresponds closely to the cellular response seen on addition of rapamycin. Moreover, like rapamycin, amino acid withdrawal inhibits p70 activity in response to essentially all cell growth factors, phorbol esters, heat shock, vanadate and calyculin, as well as to cotransfected modulators, including V12 Ras and a constitutively active PI 3-kinase. In addition, as with rapamycin, the substantial activation of p70 S6 kinase activity caused by cycloheximide, an inhibitor of translational elongation, and anisomycin, an inhibitor of translational initiation, are both inhibited by amino acid withdrawal (Table II).
The similar selectivity of the cellular response to amino acid withdrawal and to rapamycin led us to examine the effect of amino acid withdrawal on the activity of a rapamycin-resistant p70 variant, p70 ⌬2-46/⌬CT104 (40). This p70 variant is activated by insulin and/or serum through site-specific phosphorylation by PtdIns-3P-dependent protein kinases, and like wild type p70, can be markedly inhibited by treatment of cells with wortmannin (Fig. 6C), but is quite resistant to inhibition by rapamycin (Fig. 6B) (40,45). As with the response to rapamycin, the withdrawal of medium amino acids inhibits the activity of p70 ⌬2-46/⌬CT104 minimally in marked contrast to profound inhibition of recombinant wild type p70 engendered by amino acid withdrawal (Fig. 6A). Inasmuch as activation of the rapamycin-resistant p70⌬2-46/⌬CT104 variant, like wild type p70, depends on site-specific p70 phosphorylation, we compared the effects of amino acid withdrawal and rapamycin on 3 and 4) or transferred to D-PBS containing 1ϫ amino acid mixture for 30 min (lanes 5 and 6). The cells were treated with 100 nM insulin for 1 min prior to extraction, as indicated. Cell extracts were immunoprecipitated and immunoblotted with anti-phosphotyrosine antibody.

B, effect of amino acid withdrawal on phosphotyrosine-associated PI 3-kinase activity in CHO-IR cells. Confluent CHO-IR cells were incubated in
Ham's F-12 medium lacking serum for 16 h and then either left in that medium (lanes 1 and 2) or transferred to D-PBS for the times indicated (lanes 3-7). In lane 8, cells incubated in D-PBS for 2 h were transferred to D-PBS containing 1ϫ amino acids for 30 min prior to extraction. The cells were stimulated with 100 nM insulin for 2 min prior to extraction, where indicated (ϩ). Cell extracts were subjected to immunoprecipitation with anti-phosphotyrosine antibody and the co-immunoprecipitated PI 3-kinase activity was determined. C, effect of amino acid withdrawal on the activity of recombinant c-Akt/PKB in CHO-IR cells. Thirty-two hours after transfection, CHO-IR cells transfected with HA-tagged c-Akt/PKB were transferred to Ham's F-12 lacking serum. Sixteen hours later, some cells were left in this medium (lanes 1 and 2), whereas others (lanes 3-6) were transferred to D-PBS. After 2 h, the cells in D-PBS were left in D-PBS (lanes 3 and 4) or transferred to D-PBS containing 1ϫ amino acids (lanes 5 and 6) for 30 min. The cells were treated with 100 nM insulin (ϩ) or carrier (Ϫ) for the last 10 min of each incubation, as indicated. Cell extracts were immunoprecipitated with anti-HA antibody, and the immunoprecipitates were subjected to immunocomplex assay, using myelin basic protein as substrate. The reaction mixture were separated on SDS-PAGE and transferred to a PVDF membrane, and analyzed by autoradiography (upper panel) or Western blot with anti-HA antibody (lower panel). Mock refers to an anti-HA immunoprecipitate prepared from mock-transfected cells. D, effect of amino acid withdrawal on site-specific (-TEY-) phosphorylation of endogenous MAPKs in CHO-IR cells. CHO-IR cells were incubated as described in Fig. 3. Cell extracts were separated by SDS-PAGE and immunoblotted with an pp44/42 MAPK antibody (upper panel) or an antibody specific for the phosphorylated activation loop of erk1/erk2 (lower panel).
p70 phosphorylation (Fig. 6D). In contrast to wild type p70, where amino acid withdrawal provokes marked p70 dephosphorylation, the insulin-stimulated phosphorylation at Thr-412, known to be critical for the activity of p70⌬2-46/⌬CT104, like the 40 S kinase activity, is relatively resistant both to rapamycin or amino acid withdrawal, and much more susceptible to inhibition by wortmannin. Thus, the activation of the insulin-stimulated, PtdIns-3P-dependent protein kinase that acts on Thr-412 is not inhibited by rapamycin or by amino acid withdrawal. The similar response of p70⌬2-46/⌬CT104 to rapamycin and amino acid withdrawal, in contrast to its susceptibility to wortmannin suggests that rapamycin and amino acid withdrawal may lead to inhibition of p70 through a common effector, i.e. amino acid/rapamycin-sensitive p70 kinase kinases or p70 kinase phosphatases, which are distinct from the mTOR-independent, PI 3-kinase-dependent effectors inhibited by the relatively low concentrations of wortmannin.
Rapamycin Inhibits p70 S6 Kinase Activation by Amino Acids-Rapamycin promotes the dephosphorylation of p70 and eIF-4E-BPs through an inhibitory action on the mTOR kinase (30,31). Rapamycin blocks the ability of amino acids to (re)activate p70 S6 kinase activity (Fig. 7) and to restore the phosphorylation of eIF-4E BP1 (data not shown), indicating that the mTOR-directed input is required for the response to amino acids. Consequently, if amino acids and mTOR control a common regulator of p70, this amino acid/mTOR-sensitive p70 regulator either has an absolute requirement for dual, independent inputs from both amino acids and mTOR or, alternatively, it receives a single input, with mTOR situated downstream of the amino acid signal.

Amino Acid Sufficiency Signals to the Translational Apparatus-
The results demonstrate that the availability of amino acids controls the phosphorylation and activity state of two proteins that participate in the regulation of translation, i.e. FIG. 6. A, effect of amino acid withdrawal on the S6 kinase activity of p70 wild type and p70 ⌬2-46/⌬CT104. HEK293 cells, transfected with cDNA encoding either with Flag-p70␣1 wild type (dashed line) or Flag-p70 (⌬2-46/⌬CT104) (solid line) were incubated in DMEM lacking serum for 16 h, followed by addition of 10% FCS. Twenty minutes thereafter, the cells were transferred to D-PBS for the times indicated, prior to extraction. After incubation in D-PBS for 60 min, one set of plates were transferred to D-PBS containing 1ϫ amino acids for 30 min prior to extraction. Flag-tagged polypeptides were immunoprecipitated with anti-Flag antibody. The kinase activity of recombinant Flag p70 wild type or Flag p70 ⌬2-46/⌬CT104 was measured by immune complex assay. B, effect of rapamycin on the S6 kinase activity of p70 wild type and p70⌬2-46/⌬CT104. HEK293 cells, transfected with cDNA encoding either HA-tagged p70 wild type or HA-tagged p70⌬2-46/⌬CT104 were deprived of FCS, and 16 h later, reexposed to 10% FCS. Ten minutes later, rapamycin (200 ng/ml) was added, and the cells were extracted thereafter at the times indicated. Anti-HA immunoprecipitates were prepared from the extracts and assayed for p70 S6 kinase activity. C, effect of wortmannin on the S6 kinase activity of p70 wild type and p70 ⌬2-46/⌬CT104. The procedures were those described in B, except that wortmannin (100 nM) was added instead of rapamycin. D, regulation of p70 site-specific phosphorylation. CHO-IR cells were transfected with cDNA encoding HA-tagged p70␣1 wild type or HA p70⌬2-46/⌬CT104. After 32 h, the cells were deprived of FCS for 16 h. The cells were extracted directly (B) or 10 min after treatment with 100 nM insulin (I, No AA, W, R). Some cells were treated with 200 mg/nl rapamycin (R) or 100 nM wortmannin (W), 20 min prior to insulin addition, or transferred into D-PBS lacking amino acids 50 min prior to insulin addition. Cell extracts were subjected to anti-HA immunoprecipitation, and the immunoprecipitates were assayed for S6 kinase activity (top panel) or subjected to SDS-PAGE, transfer to PVDF membranes followed by immunoblot with anti-p70 peptide antibody (second panel from top), anti-p70 T444P/S447P phosphopeptidespecific antibody (third panel from top), or anti-p70 T412P phosphopeptide-specific antibody (bottom panel). the p70 S6 kinase and eIF-4E BP1, an inhibitor of eIF-4E function. This response is specific for the amino acids and is not reproduced by changes in the availability of glucose or in overall energy sufficiency. The changes in p70 and eIF-4E BP1 phosphorylation appear to be part of a "general" control response, inasmuch as they are elicited by depletion of virtually any single amino acid, although the responses to omission of some amino acids, e.g. arginine and leucine, are exceptionally strong. The inability of arginine alone to restore activation of p70 even partially argues strongly against a contribution from NO generation. The responses to amino acid omission described here are probably part of a physiologic program directed at the down-regulation of protein synthesis. The extent of translational inhibition has not been measured directly, but might be estimated by considering the effects of rapamycin on overall protein synthesis, inasmuch as rapamycin leads to a similar inhibition of p70 S6 kinase and eIF-4E BP1 phosphorylation. In several mammalian cell types, rapamycin inhibits overall protein synthesis by 10 -15%, although the translation of mRNAs containing 5Ј-terminal oligopyrimidine tracts such as those encoding ribosomal proteins and EF-2, is inhibited much more strongly (22)(23)(24). Thus, it is likely that the deactivation of p70 and dephosphorylation of eIF-4E BP1 per se do not contribute in a quantitatively substantial way to the marked inhibition of protein synthesis known to be engendered by amino acid depletion. It is probable that other changes such as in eIF-2␣ or eIF-4E phosphorylation, or in the rate of polypeptide elongation, are of greater quantitative importance (3,6,10). Nevertheless, the dephosphorylation of p70 and eIF-4E BP1 reflect the operation of a signaling apparatus that provides for the orderly shutdown of translation in the face of substrate insufficiency. Conceivably, this amino acid-regulated signal transduction pathway might also control the increase in endogenous protein breakdown that occurs in many cells in response to amino acid depletion. Thus, in isolated rat hepatocytes, ambient amino acids control the extent of S6 phosphorylation in situ, and negatively regulate autophagic proteolysis (38). Moreover, the ability of increased medium amino acids to suppress proteolysis is partially antagonized by rapamycin, leading to the suggestion that S6 phosphorylation and autophagic proteolysis are regulated, at least in part, by common, amino acid-sensitive signal transduction pathways. The nature of the intracellular signal generated in response to amino acid sufficiency that initiates the changes in regulatory protein phosphorylation shown here is not known; however, several precedents in eucaryotic cells are available. Among the best characterized "general control" responses to amino acid depletion is that in S. cerevisiae mediated by GCN2, an eIF-2␣ kinase, which contains a domain with sequence similarity to histidyl tRNA synthetase. This noncatalytic segment of the kinase is thought to act as a receptor for deacylated tRNAs, and mediate activation of GCN-2 (9). Studies in mammalian cells depleted of amino acids, or treated with histidinol (50), a competitive inhibitor of tRNA His synthetase, or that contain a temperature-sensitive mutant of leucyl tRNA synthetase (51,52), have each also demonstrated increased eIF-2␣ phosphorylation. Although these changes have been attributed to accumulation of uncharged tRNA or altered tRNA aminoacyl synthetase activity, the direct demonstration of an eIF-2␣ kinase activated by amino acid withdrawal is lacking; some data point to a decrease in eIF-2␣ phosphatase induced by amino acid depletion (50). Nevertheless, a contribution by deacylated tRNAs to the responses shown here to be engendered by amino acid depletion remains attractive, in view of the regulatory role proposed for deacylated tRNAs bound to GCN2 in yeast or to the ribosomal A site in initiating the bacterial stringent response (9). The exceptional potency of leucine and arginine depletion (Table I) could relate to the frequency of their utilization in protein synthesis, and/or to the existence of multiple FIG. 7. Rapamycin inhibits the activation of p70 S6 kinase by amino acids. Confluent CHO-IR cells were incubated in Ham's F-12 medium lacking FCS for 16 h and then transferred to D-PBS. After 2 h, the cells were transferred to D-PBS with or without 2ϫ amino acids, with or without rapamycin, at the concentrations indicated. The cells were extracted 30 min later, and the S6 kinase activity was determined after immunoprecipitation of endogenous p70 S6 kinase. 32 P incorporation into S6 was quantified by PhosphorImager, and is expressed in arbitrary units (PI units). tRNAs for these amino acids arising from the 6-fold codon degeneracy (53). Alternatively, the ability of nontransportable or nonmetabolizable analogs of leucine or phenylalanine to inhibit hepatic macroautophagy (54 -56) has been offered in support of the existence of a plasma membrane binding element as the initiator of this cellular response to amino acids.
Amino Acid Withdrawal and Rapamycin Inhibit a Common Set of Signaling Elements-The striking similarity of the cellular response to amino acid depletion to that caused by rapamycin led us to inquire as to whether amino acid sufficiency and mTOR regulated a common signal transduction pathway. Both treatments induce selective dephosphorylation of p70 and eIF-4E BP1 without altering the insulin activation of PI 3-kinase, c-Akt/PKB, or MAPK. Moreover, the p70⌬2-46/⌬CT104 variant, which is highly resistant to the dephosphorylation engendered by amino acid withdrawal, is comparably resistant to rapamycin-induced dephosphorylation. The possibility that amino acid sufficiency signals through a pathway that includes TOR is further reinforced by the similarity in the response of S. cerevisiae to nitrogen depletion as compared with its response to rapamycin or TOR 1/2 deletion (34). Both rapamycin and TOR1/2 deletion cause S. cerevisiae to arrest in early G1 in an unbudded state which resembles closely the phenotype of cells arrested by nutrient deprivation (i.e., G 0 or stationary phase). In particular, cells accumulate glycogen and become thermotolerant; overall translation is inhibited by 90%, but certain mRNAs, e.g. UB14 (ubiquitin), continue to be vigorously translated.
In CHO-IR cells, rapamycin prevents the restoration of p70 and eIF-4E BP1 phosphorylation that occurs in response to reintroduction of amino acids. Thus, if amino acids and mTOR control these responses through a common pathway, amino acid regulation occurs upstream of mTOR. The in vitro autokinase activity of endogenous mTOR, as well as the mTORassociated eIF-4E BP1 kinase activity, responses that require a functionally intact mTOR catalytic domain, and the mTORassociated PtdIns kinase activity, were readily detected in mTOR immunoprecipitates prepared from CHO-IR or HEK293 cells; nevertheless, no significant alterations in these activities were evident in response to alterations in amino acid sufficiency (data not shown) or insulin (31). The failure to detect an amino acid-induced alteration in mTOR autokinase and eIF-4E BP1 kinase activity does not constitute strong evidence against such regulation occurring in situ. Thus, it has been shown that the rapamycin/FKBP12 complex can bind directly to mTOR in vitro with high affinity (30,31), and inhibit the mTOR autokinase activity (30,32), as well as the putative eIF-4E BP1 kinase activity (32). Nevertheless, mTOR extracted from rapamycin-treated cells, which had been fully inhibited in situ as judged by the inhibition of p70 kinase, also showed no difference in autokinase and 4E BP1 kinase activity in vitro as compared with mTOR from untreated cells (data not shown). Thus, mTOR inhibition by rapamycin/FKBP12 is apparently readily reversed during cell extraction, and a similar reversal may occur for mTOR inhibited by amino acid depletion. Efforts to identify physiologic regulators of mTOR kinase activity are ongoing.
The Proximate Effectors Mediating Amino Acid and Insulin Regulation of the p70 S6 Kinase and eIF-4E BP1-Amino acid readdition, like insulin, activates p70 in situ through a multisite phosphorylation, and also promotes the multiple phosphorylation of eIF-4E BP1. What are the proximate events accounting for these changes in phosphorylation in response to amino acid sufficiency? Although the p70 kinase kinases have not yet been fully defined, it is clear that the sites phosphorylated during activation are situated in a variety of amino acid sequence contexts, and reflect the independent operation of several different protein kinases (40, 42, 44 -49, 57). The initial modification involves the multiple phosphorylation of a cluster of (Ser/Thr) Pro sites, situated in the psuedosubstrate inhibitory domain of the p70 carboxyl-terminal tail (Ser-434, -441, -447, and -452; Thr-444) (44,47) and probably Ser-394 (58), situated in a 65-amino acid segment immediately carboxylterminal to the catalytic domain, that is conserved in sequence and location among p70, c-Akt/PKBs, PKCs, and Rsks (49). These sites can be phosphorylated in vitro by an array of proline-directed kinases including erk1, erk2, cdc2, and other as yet unidentified proline-directed protein kinases (57). Notably, the five sites of insulin-stimulated phosphorylation on eIF-4E BP1 identified thus far are all (Ser/Thr) Pro sites (59), and mTOR has been shown recently to phosphorylate these sites directly in vitro (60). The phosphorylation of p70 at these (Ser/Thr) Pro sites, although insufficient to activate the kinase (57) facilitates the further PI 3-kinase-regulated phosphorylation of p70 Thr-412 in the catalytic domain carboxyl-terminal extension (45,48), and p70 Thr-252 in activation loop of catalytic subdomain VIII (40,(45)(46)(47)(48). Phosphorylation of the latter two sites, acting together in a strongly cooperative fashion, serves to activate the p70 catalytic function; modification of either Thr-252 or Thr-412 singly, at least in vitro, is insufficient to generate more than 2-5% of the maximal p70 activity (46). Recent evidence indicates that phosphorylation of p70 Thr-252 is catalyzed 3-phosphoinositide-dependent protein kinase 1 (PDK1) (46,47) a kinase that also activates c-Akt/PKB by phosphorylation of c-Akt/PKB Ser-308, a site homologous with p70 Thr-252 (61); p70 Thr-412 is phosphorylated by a different, as yet unidentified, 3-phosphoinositide-regulated protein kinase.
The evidence presented in this study indicates that amino acid sufficiency does not alter significantly insulin activation of PI 3-kinase or the (Ser/Thr) kinases that act on p70 kinase. Thus, erk1 and erk2 are not significantly affected by amino acid depletion. In addition, the lack of inhibition of c-Akt/PKB by amino acid withdrawal is especially significant, inasmuch as c-Akt/PKB activation requires the PtdIns (3,4,5)P 3 -regulated phosphorylation at two sites entirely homologous to p70 Thr-252 and Thr-412, namely c-Akt/PKB Ser-308, a phosphorylation catalyzed by PDK1 (60), and c-Akt/PKB Thr-473.
A strong indication that amino acids do not influence the p70 Thr-412 kinase is provided by consideration of regulation of the p70⌬2-46/⌬CT104 mutant (Fig. 6D). This variant, like the parent wild type p70 kinase, is activated in situ through phosphorylation at Thr-252 and Thr-412. As seen in Fig. 6D, insulin-stimulated phosphorylation of Thr-412 is readily evident in p70⌬2-46/⌬CT104 and persists in the face of amino acid withdrawal. By contrast, insulin-stimulated Thr-412 phosphorylation of p70⌬2-46/CT⌬104 is largely inhibited by wortmannin, an inhibitor of PI 3-kinase, presumably because of inhibition of PtdIns-3P-dependent p70 (Thr-412) kinases. Conceivably, amino acid depletion could also be inhibiting these same PtdIns-3P-dependent kinases, but the phosphorylation of p70⌬2-46/⌬CT104 is maintained because this variant is more readily phosphorylated by these kinases. This possibility can be rejected because in the absence of amino acid depletion, insulin promotes a much more robust phosphorylation in situ of Thr-412 on p70 wild type than on p70⌬2-46/⌬CT104. Taken together, these observations suggest that amino acid withdrawal does not induce dephosphorylation of p70 primarily by inhibiting p70 kinase kinases, but by accelerating p70 dephosphorylation, i.e., by increasing the p70 phosphatase activity. Such putative, amino acid-sensitive phosphatases must be relatively insensitive to calyculin (Table II), and okadaic acid, inasmuch as these inhibitors of phosphatase 1 and 2A do not prevent the dephosphorylation of p70 and eIF-4E BP1 that occurs following amino acid withdrawal. Significantly, all the arguments presented pointing to the likelihood that amino acid withdrawal activates a p70 phosphatase apply with equal force to the action of rapamycin.
Conclusions-We offer the following model for the amino acid regulation of p70 S6 kinase and eIF-4E BP1 phosphorylation, and its relation to insulin/growth factor-regulated inputs (Fig. 8). We propose that amino acid sufficiency acts through a signal arising from the translational apparatus, which reflects the availability of substrate for protein synthesis. This signal is generated downstream of amino acid transport, but upstream of the steps inhibited by anisomycin and cycloheximide, translational inhibitors known to activate the p70 S6 kinase. The chemical identity of the signal is unknown; however, examples of candidate signaling molecules whose abundance is altered by amino acid insufficiency include uncharged tRNAs, adenosine tetraphosphate, and in bacteria, guanosine 5Ј,3Ј-bispyrophosphate. We propose that this signal regulates in situ the activity of a protein phosphatase that selectively dephosphorylates p70 S6 kinase, eIF-4E BP1, and perhaps other elements that control translation. Moreover, based on the close similarity in the pattern of responses to rapamycin and amino acid withdrawal, and especially the parallel insensitivity of p70⌬2-46/⌬CT104 to amino acid depletion and rapamycin, we propose that the putative amino acid-responsive phosphatase is also regulated by mTOR. Rapamycin, by inhibiting mTOR in situ, activates the phosphatase. The p70⌬2-46/⌬CT104 variant is resistant to amino acid depletion and rapamycin because of a diminished susceptibility to the putative phosphatase; however, because it requires an insulin-stimulated, PI 3-kinase-dependent phosphorylation of Thr-252 and Thr-412, it remains susceptible to inhibition by wortmannin. We emphasize that the existence of such a phosphatase is entirely speculative; an enzyme with the expected regulatory behavior (i.e. responsive to amino acid sufficiency and mTOR) and specificity has not been identified.
The recent demonstration that mTOR directly phosphorylates eIF-4E BP1 in vitro (32) indicates that the ability of mTOR to phosphorylate p70 directly at regulatory sites merits examination, but does not undermine the suggestion that mTOR acts through regulation of an eIF-4E BP1/p70 phosphatase; several examples are known wherein a protein (Ser/Thr) kinase regulates a phosphatase that acts on other substrates of the kinase. Protein kinase A, which phosphorylates directly phosphorylase b kinase and glycogen synthase, also regulates their dephosphorylation by its ability to phosphorylate protein phosphatase inhibitor-1, thereby conferring on PPI-1 the ability to inhibit protein phosphatase 1 (62). The Rho kinase ␣, which phosphorylates myosin light chain directly (63), also inhibits myosin light chain dephosphorylation by phosphorylating and inhibiting an MLC phosphatase (64).
If amino acid sufficiency and mTOR control p70 S6 kinase and eIF-4E BP1 phosphorylation through a common phosphatase, then this could occur through independent pathways that converge at this phosphatase, or through amino acid regulation of the mTOR kinase. Although we have not detected changes in mTOR kinase activity in vitro in response to alterations in amino acid sufficiency (or rapamycin addition) in situ, the idea that amino acid sufficiency could be a regulator of mTOR signaling remains attractive because it suggests that mTOR, like its most closely related homologs ATM, MEC1, etc., functions as a checkpoint control element (33) responsive to signals arising from the translational apparatus, rather than from receptor tyrosine kinases and PI 3-kinase. Consistent with this view, withdrawal of the amino acid and mTOR signal acts as an override switch, that inhibits p70 S6 kinase activation and eIF-4E BP1 phosphorylation irrespective of the RTK-PI 3-kinase signals. Although the model suggested for amino acid regulation of p70 and eIF-4E BP1 provides a rational framework for the available data, many aspects remain entirely speculative, and much additional work will be necessary to identify the components and evaluate the relationships proposed.