Identification of Phosphorylation Sites in the Translational Regulator, PHAS-I, That Are Controlled by Insulin and Rapamycin in Rat Adipocytes*

Phosphorylation of PHAS-I by mitogen-activated pro- tein (MAP) kinase in vitro decreased PHAS-I binding to eukaryotic initiation factor (eIF)-4E. The decrease in binding lagged behind the phosphorylation of PHAS-I in Ser 64 , the preferred site of MAP kinase. Binding of the Ala 64 mutant of PHAS-I to eIF-4E was abolished by MAP kinase, indicating that phosphorylation of sites other than Ser 64 control binding. To identify such sites, PHAS-I was phosphorylated with MAP kinase and [ (cid:103) - 32 P]ATP and then cleaved proteolytically before the resulting phosphopeptides were isolated by reverse phase chromatography and directly identified by amino acid sequencing. Phosphorylated residues were located by determining the cycles in which 32 P was released when phosphopeptides were subjected to sequential Ed- man degradation. With an extended incubation in vitro , MAP kinase phosphorylated Thr 36 , Thr 45 , Ser 64 , Thr 69 , and Ser 82 . In rat adipocytes, the phosphorylation of all five sites was increased by insulin and decreased by rapamycin although there were differences in the mag-nitude of the effects. A form of PHAS-I phosphorylated exclusively in Thr 36 remained bound to eIF-4E, indicating that phosphorylation of Thr 36 is insufficient for dissociation of the PHAS-I (cid:122) eIF-4E complex. In summary, our results indicate that multiple phosphorylation sites

Insulin and growth factors act within minutes to stimulate protein synthesis (1,2). This rapid response is due to activation of mRNA translation and involves phosphorylation of multiple translation factors. One of these factors is PHAS-I (also known as 4EBP-1) (3,4), a protein of M r Ϸ 12,500 that was first identified in 32 P-labeled adipocytes as a heat-and acid-stable species that was markedly phosphorylated in response to insulin (5). PHAS-I is now known to be expressed in a wide variety of cell types and to act as a regulator of eIF-4E, the mRNA cap-binding protein (6,7). eIF-4E is one of the least abundant of the known translation factors (8,9), and the amount of eIF-4E is believed to be limiting for initiation, which is generally the rate-limiting phase of protein synthesis (1,2). Thus, increasing eIF-4E in cells increases mRNA translation, particularly of those messages which possess a high degree of secondary structure in their 5Ј-untranslated regions (10). eIF-4E is a key component of the eIF-4F complex, which catalyzes melting of secondary structure in the 5Ј-untranslated region of the mRNA and allows efficient binding and/or scanning by the 40 S ribosomal subunit (1,2,8,9). eIF-4F contains two other subunits, eIF-4A, an ATP-dependent helicase, and eIF-4G, a relatively large subunit that binds to both eIF-4A and eIF-4E (1,2,8,9). Nonphosphorylated PHAS-I binds tightly to eIF-4E (4, 11) and blocks eIF-4E binding to eIF-4G (12,13). Consequently, increasing PHAS-I inhibits cap-dependent mRNA translation both in vitro and in intact cells (4). However, when PHAS-I is phosphorylated in response to insulin or certain growth factors, the PHAS-I⅐eIF-4E complex dissociates (6,7), thereby increasing eIF-4E available to bind to eIF-4G. Increasing the eIF-4F complex by this mechanism provides an explanation of the preferential stimulation by insulin of the translation of messages, such as ornithine decarboxylase, which have a high degree of secondary structure in their 5Ј-untranslated region (14).
The immunosuppressive drug, rapamycin, promotes dephosphorylation of PHAS-I in adipocytes (15)(16)(17) and a variety of other cell types (18 -21). Rapamycin is a potent inhibitor of the activation of p70 S6K (22,23), but this enzyme is almost certainly not a PHAS-I kinase in cells since it does not phosphorylate PHAS-I in vitro (6,16,24). PHAS-I can be phosphorylated in vitro by protein kinase C and casein kinase II (6,24,25), but whether either of these kinases phosphorylate PHAS-I in cells is not known. PHAS-I is an excellent substrate for MAP kinase 1 in vitro (11,24); and, Ser 64 , the site in PHAS-I preferred by MAP kinase, is phosphorylated in response to insulin in adipocytes (24). However, several lines of evidence indicate that MAP kinase is not the major mediator of insulin action on PHAS-I (6). For example, when bound to eIF-4E, the phosphorylation of PHAS-I by MAP kinase in vitro is relatively slow (6,15,16), and inhibiting activation of MAP kinase with the inhibitor of MEK activation, PD 098059, did not block the effects of insulin on PHAS-I in 3T3-L1 adipocytes (15).
The amino acid sequence surrounding a particular Ser or Thr residue contributes to the specificity of many protein kinases (26). Thus, identification of phosphorylation sites can provide clues as to which kinases phosphorylate a protein. PHAS-I isolated from adipocytes contains both phosphothreonine and phosphoserine (27)(28)(29), and results from peptide mapping studies (24) and two-dimensional electrophoretic analyses (15,30) indicate that PHAS-I is phosphorylated in multiple sites in cells. Only one phosphorylation site (Ser 64 ) has been identified directly (24), and the possibility exists that the sites most important in regulating the association of PHAS-I and eIF-4E has not been determined. The objectives of our experiments were to identify the sites of phosphorylation in PHAS-I in adipocytes, to investigate the role of these sites in modulating binding of PHAS-I to eIF-4E, and to determine the changes in phosphorylation produced by insulin and rapamycin.

EXPERIMENTAL PROCEDURES
Incubation of Adipocytes and Immunoprecipitation of PHAS-I-Fat cells were isolated by collagenase digestion of adipose tissue from male rats (Wistar, 120 -140 g, 20 per preparation) (31) and then washed four times and suspended in low P i medium (0.2 mM sodium phosphate, 125 mM NaCl, 1.2 mM MgCl 2 , 4 mM KCl, 2.6 mM CaCl 2 , 0.5 mM glucose, 10 mg/ml bovine serum albumin, and 24 mM sodium HEPES, pH 7.4). Na 32 P i (5 mCi/10-ml aliquot of cells) was added, and the cells were incubated at 37°C for 90 min, a time sufficient to achieve steady-state labeling of intracellular ATP (32). After treatments with insulin and rapamycin, the incubations were terminated by homogenizing the cells as described previously (24) in Homogenization Buffer (2 ml/ml packed cells), which contained 100 mM NaF, 10 mM EDTA, 2 mM EGTA, 1 mM benzamidine, 0.2 mM phenylmethylsulfonyl flouride, and 50 mM Tris, pH 7.8. The homogenates were centrifuged at 28,000 ϫ g for 30 min. PHAS-I was immunoprecipitated from the supernatants by using an affinity-purified antibody (3) coupled to protein A-Sepharose (Pharmacia Biotech Inc.) as described previously (11). To elute PHAS-I from the immune complexes, the beads were suspended in 300 l of 1% ␤-mercaptoethanol, 1 mM EDTA, and 10 mM Tris-HCl, pH 7.5, and incubated at 95°C for 15 min. PHAS-I, which is relatively stable to heat (3), was recovered in the supernatant after centrifuging samples at 10,000 ϫ g for 10 min.
Phosphorylation of PHAS-I by MAP Kinase in Vitro-Wild-type PHAS-I (17), PHAS-I tagged at the NH 2 terminus with a hexahistidine sequence ([H 6 ]PHAS-I) (24), [H 6 ]PHAS-I having a Ser to Ala mutation at position 64 ([H 6 ]PHAS-I Ala64 ) (11), the ERK2 isoform of MAP kinase (33), and a constitutively active form of MEK1 (34) were expressed in bacteria and purified as described in the references indicated. MAP kinase was activated by incubation with MEK1 as described by Scott et al. (34) P]PHAS-I samples (300 l) from either in vitro phosphorylation reactions or immunoprecipitated from 32 P-labeled adipocytes were incubated at 37°C for 15 h with lysyl endopeptidase (17 g/ml). The digests were acidified by adding 30 l of 1% trifluoroacetic acid solution and applied at 30°C to a reverse phase column (Waters Nova-Pak C 18 , 3.9 ϫ 150 mm) that had been equilibrated in 0.1% trifluoroacetic acid (Buffer A). The flow rate was maintained at 1 ml/min. The column was washed for 5 min with Buffer A before peptides were eluted with a linear gradient of acetonitrile (0 -60% in 120 min) produced by increasing the proportion of Buffer B (0.1% trifluoroacetic acid in acetonitrile). Fractions (1 ml) were collected, and peptides containing 32 P were identified by measuring Cerenkov emissions. A second digestion was required to resolve phosphorylation sites present in phosphopeptides in fractions 62 through 72. These fractions were pooled and evaporated to dryness before the peptides were dissolved in 250 l of solution containing 1 mM dithiothreitol, 1% hydrogenated Triton X-100 (Calbiochem), and 50 mM Tris-HCl, pH 8.0. The samples were incubated for 15 h at 37°C with chymotrypsin (5 g) and then acidified by adding 25 l of 1% trifluoroacetic acid before peptides were resolved by reverse phase HPLC as described above.
The 32 P-labeled peptides generated by phosphorylating recombinant PHAS-I with [␥-32 P]ATP and MAP kinase were identified by using a vapor phase amino acid sequencer (Applied Biosystems Procise 494). Phosphorylated residues within phosphopeptides were located by determining the cycles in which 32 P was released when samples were subjected to sequential Edman degradation under conditions that optimize recovery of 32 P (35).
Electrophoretic Analyses-Protein samples were subjected to electrophoresis in 20% polyacrylamide gels in the presence of SDS by using the method of Laemmli (36). Dried gels were exposed to film to enable detection of 32 P-labeled PHAS-I, and bands containing the protein were excised. The amounts of 32 P in the gel slices were determined by measuring Cerenkov emissions. Binding of PHAS-I proteins to eIF-4E was assessed by far-Western blotting performed using a 32 P-labeled recombinant eIF-4E fusion protein (FLAG-4E) as described previously (17). Phosphoamino acid analyses were performed by subjecting samples of phosphopeptides that had been purified by reverse phase HPLC to limited hydrolysis (5.7 N HCl for 2 h at 110°C) (37). After removing the acid under vacuum, samples were subjected to high voltage electrophoresis at pH 1.9, which provides superior separation of phosphoserine and phosphothreonine. The fact that phosphothreonine and phosphotyrosine are not resolved at this pH was not a problem as PHAS-I from adipocytes is not phosphorylated on tyrosyl residues (27)(28)(29). 2 Other Materials-[␥-32 P]ATP and Na 32 P i were purchased from Du-Pont NEN. Chymotrypsin and lysyl endopeptidase were obtained from Boehringer Mannheim and WAKO Pure Chemical Industries, respectively. HPLC grade solvents were from Baker Chemical Co. Amino acid sequencing supplies were from Applied Biosystems. Most commonly used chemicals were from Sigma. Rapamycin was obtained from Calbiochem-Novabiochem International, and human insulin was from Eli Lilly.

Phosphorylation of Sites Other Than Ser 64 Controls PHAS-I
Binding to eIF-4E-PHAS-I contains seven Ser/Thr-Pro motifs that represent potential sites of phosphorylation by MAP kinase ( Fig. 1), the most effective of the protein kinases that have been found to phosphorylate PHAS-I in vitro (6,16,24). In previous experiments Ser 64 was identified as the preferred site of phosphorylation by MAP kinase (24). However, with extended incubation with MAP kinase, the stoichiometry of PHAS-I phosphorylation exceeded 1 mol/mol (24), and some phosphorylation of an Ala 64 mutant PHAS-I was observed (11), indicating that MAP kinase phosphorylated more than one site in PHAS-I. Results from peptide mapping experiments provide further evidence of multisite phosphorylation (Fig. 2). Three peaks of 32 P-labeled peptides, designated LE-P2, LE-P3, and LE-P4 in order of elution, were resolved when a sample of recombinant [H 6 ]PHAS-I that had been phosphorylated in a 10-min incubation with [␥-32 P]ATP and MAP kinase was digested with lysine endopeptidase and subjected to reverse phase HPLC ( Fig. 2A, inset). LE-P2 was completely absent in digests of [H 6 ]PHAS-I Ala64 that had been phosphorylated by MAP kinase, indicating that LE-P2 represents the 32 P-labeled Ser 64 phosphopeptide. After 2 h with MAP kinase, relatively more 32 P was found in sites in LE-P3 plus LE-P4 than in Ser 64 ( Fig. 2A). 3 In addition, a relatively small peak, LE-P1, eluting very early in the acetonitrile gradient was observed. The phosphorylation of other sites by MAP kinase was more pronounced in Fig. 2 than in our previous studies (11,24) because approximately 10-fold higher concentrations of MAP kinase were used in the present experiments. However, consistent with previous findings (11,24), the initial rate of phosphorylation of Ser 64 (LE-P2) occurred more rapidly than the initial rate of phosphorylation of the other sites (Fig. 3A), which were phosphorylated at least as rapidly in the Ala 64 mutant as in the wild-type protein (Fig. 3B).
The influence of Ser 64 and other sites on eIF-4E binding was assessed by far-Western blotting using a 32 P-labeled FLAG-4E. MAP kinase markedly decreased binding of [H 6 ]PHAS-I to FLAG-4E (Fig. 3C, inset). However, the initial rate of decline in eIF-4E binding activity (Fig. 3C) produced by MAP kinase did not correlate well with the initial rate of Ser 64 phosphorylation (Fig. 3A). For instance, almost no decrease in eIF-4E binding was observed after 5 min of incubation with MAP kinase, even though over half of the Ser 64 in PHAS-I had been phosphorylated. The decrease in binding activity correlated much better with the phosphorylation of sites recovered in LE-P2 and LE-P3, indicating that phosphorylation of sites other than Ser 64 control binding of PHAS-I to eIF-4E. Findings with [H 6 ]PHAS-I Ala64 provide additional support for this interpretation. The initial decrease in eIF-4E binding activity produced by phosphorylation of the mutant protein occurred with a time course that was similar to that obtained with [H 6 ]PHAS-I (Fig. 3C), and extended incubation with MAP kinase abolished binding of the mutant PHAS-I to FLAG-4E (Fig. 3C, inset).
Identification of Sites in PHAS-I Phosphorylated by MAP Kinase in Vitro-The first approach to identify the sites other than Ser 64 was to attempt to identify the phosphopeptides derived from lysyl endopeptidase digests of PHAS-I. Except for the 4-amino acid peptide that would result from cleavage at Lys 68 and Lys 72 , all of the predicted peptides contain at least 12 amino acids (Fig. 1) and would be expected to elute relatively late in the acetonitrile gradient. For this reason, the 4-amino acid peptide containing Thr 69 was considered the most likely candidate for LE-P1. Phosphoamino acid analysis supported this assignment, as LE-P1 was found to contain phosphothreonine but little if any phosphoserine. 2 Amino acid sequencing indicated that the peptide contained Pro residues in the second and third positions (Fig. 4). Other than the Thr 69 peptide, the only two other occurrences of adjacent prolines in PHAS-I are Pro 29 -Pro 30 and Pro 88 -Pro 89 (Fig. 1). Determining the position of the phosphorylated residue within the LE-P1 phosphopeptide solidified the assigment of Thr 69 as the phosphorylation site. This was accomplished in a separate sequencing run, where the 32 P in LE-P1 was found to be released in cycle 1 (Fig.  4). Even without the sequence data, finding release in the first cycle with a peptide generated by lysyl endopeptidase would be indicative of Thr 69 as this residue is the only Ser or Thr adjacent to a Lys in PHAS-I (Fig. 1).
The absence of LE-P2 in peptides derived from [H 6 ]PHAS-I Ala64 provided strong evidence that Ser 64 was the site of phosphorylation in the LE-P2 phosphopeptide ( Fig. 2A, inset). Proof that Ser 64 was the phosphorylated residue was provided by amino acid sequencing which yielded a single sequence corresponding to the predicted Ser 64 peptide (Fig. 4). As Ser 64 is the only Ser/Thr in this peptide, this residue had to be the phosphorylated site. As expected, release of 32 P from the phosphopeptide in LE-P2 occurred in cycle 8 (Fig. 4).
Amino acid sequencing indicated that LE-P3 contained the peptide generated by cleaving [H 6 ]PHAS-I at Lys 72 and Lys 104 . As this peptide contained multiple Ser and Thr, it was necessary to measure 32 P release during sequential Edman degradation to determine the location of the phosphorylated residue. The surge in 32 P release from LE-P3 occurred in cycle 10, identifying Ser 82 as a phosphorylation site. The percentage of 32 P released from LE-P3 samples was relatively low. This was probably due in part to the NH 2 -terminal Asp residue, as Asp may form a cyclic imide structure after reaction with the 1-ethyl-3-dimethylaminopropyl carbodiimide used to couple the peptides to the solid support (Sequalon-AA), thereby blocking Ed- man degradation (38). However, LE-P3 was also found to contain some of a relatively large peptide predicted to extend from the His-tag to Lys 56 (Fig. 1). This peptide, which was the predominant species in LE-P4, 2 contained multiple Ser/Thr residues that were too far from the NH 2 terminus to be identified as phosphorylation sites by determining 32 P release during Edman degradation. Therefore, an additional digestion was needed to resolve phosphorylation sites. Fractions containing LE-P3 and LE-P4 were pooled, and concentrated under vacuum to remove the acetonitrile and trifluoroacetic acid. After the peptides were digested with chymotrypsin, three peaks of radioactivity, designated CT-P1, CT-P2, and CT-P3, were resolved by reverse phase HPLC (Fig. 2B).
Amino acid sequencing indicated that CT-P1 contained a phosphopeptide derived from the more COOH-terminal portion of a region in PHAS-I that contains a repeat of the sequence, STTPGGT (Fig. 4). When a sample of CT-P1 was subjected to sequential Edman degradation, the surge in 32 P release occurred in cycle 3, identifying Thr 45 as a site of phosphorylation (Fig. 4). The more NH 2 -terminal portion of the region containing the repeat was found in CT-P2, and with this phosphopeptide, the surge of 32 P also occurred in cycle 3, identifying Thr 36 as a site of phosphoryation (Fig. 4). [ 32 P]Phosphoserine was the only phosphoamino acid detected in CT-P3, 2 consistent with the interpretation that Ser 82 was present in this peak.
Identification of Sites in PHAS-I Controlled by Insulin and Rapamycin in Vitro-The control of PHAS-I phosphorylation in cells was investigated by incubating rat adipocytes in medium containing 32 P i . After treatments with insulin and rapamycin, the cells were homogenized, and 32 P-labeled PHAS-I was immunoprecipitated from extracts. Samples were subjected to SDS-PAGE, and autoradiograms were prepared (Fig. 5A, inset). Essentially all of the 32 P-labeled protein migrated in bands corresponding to the PHAS-I protein.
Insulin not only increased the amount of 32 P-labeled PHAS-I, but also increased the proportion of the 32 P-labeled protein that was found in the most slowly migrating form, a finding that is consistent with the previous demonstration that phosphorylation of the appropriate sites in PHAS-I decreases its electrophoretic mobility (11). Rapamycin treatment caused a net decrease in the 32 Pcontent of PHAS-I and increased the proportion of 32 P found in the forms of higher electrophoretic mobility (Fig. 5A, inset). Incubating cells with rapamycin prior to insulin attenuated the effects of the hormone on increasing phosphorylation of PHAS-I.
To investigate the distribution of 32 P among different sites, the immunoprecipitated PHAS-I was cleaved with lysyl endopeptidase, and samples of the digest were analyzed by HPLC under conditions used to resolve the MAP kinase phosphorylation sites (Fig. 5A). The pattern of 32 P-labeled peptides from cellular PHAS-I was similar to that obtained with the recombinant protein, although there were differences. Notably, the relative size of LE-P1 was much larger with the adipocyte PHAS-I, indicating that the site in this peak was highly phosphorylated in cells. The elution positions of LE-P1 and LE-P2 from cellular PHAS-I (Fig. 5A) were identical to those of the corresponding peaks derived from recombinant PHAS-I ( Fig.  2A). To verify that the peaks contained the same sites of phosphorylation, 32 P release was measured following Edman deg- radation of the peptides (Fig. 6). Almost all of the 32 P in the LE-P1 peptide was released in cycle 1, and phosphothreonine was the only phosphoamino acid detected in this peak (Fig. 7). These findings indicate that LE-P1 contains the Thr 69 peptide. With LE-P2 a single surge of 32 P occurred in cycle 8 (Fig. 6), and phophoserine was the only phosphoamino acid detected in this peak (Fig. 7). Thus, LE-P2 appears to contain the Ser 64 peptide.
LE-P3 and LE-P4 were not as well resolved using PHAS-I from cells as when the phosphorylated recombinant protein was used. Likewise, LE-P3 and LE-P4 from non-tagged recombinant PHAS-I lacking the His-tag were not resolved. 2 The better resolution obtained with [H 6 ]PHAS-I is presumably because the additional amino acids in the His-tag region cause the large peptide in LE-P4 to elute at higher concentrations of acetonitrile. Subjecting peptides in LE-P3 to sequential Edman degradation resulted in release of 32 P in cycle 10, indicative of Ser 82 phosphorylation. LE-P3 also contains other sites found in the large NH 2 -terminal peptide, but residues from this peptide are not released during Edman degradation because the NH 2 terminus of PHAS-I from adipocytes is blocked (3). To resolve sites in the large peptide, LE-P3 fractions were pooled and incubated with chymotrypsin. This treatment resulted in generation of peptides that eluted in sharp peaks having the same retention times as CT-P1, CT-P2, and CT-P3 (Fig. 5B). Thus, the elution pattern suggested that the peptides that were phosphorylated by MAP kinase in vitro were also phosphorylated in cells. The peptides in these peaks were found to contain phosphothreonine, phosphothreonine, and phosphoserine (Fig. 7), respectively, as would be expected if Thr 45 , Thr 36 , and Ser 82 were the phosphorylated residues. Surges of 32 P release occurred in cycle 3 when the phosphopeptides in either CT-P1 or CT-P2 were subjected to sequential Edman degradation (Fig.  6). Thus, determinations of the positions of the phosphorylated residues in CT-P1 and CT-P2 supported the assignments of Thr 45 and Thr 36 as phosphorylation sites.
The effects of insulin and rapamycin on the 32 P contents of the five sites are summarized in Fig. 8. Insulin increased the amount of 32 P in all five sites, but the effects of insulin on the different sites differed both in the extent of 32 P introduced into the sites and in the percentage change in 32 P. Relatively little 32 P was recovered in the Ser 82 peptide, and Ser 82 was least affected by insulin, which produced only a 2-fold increase in 32 P. The effects of insulin on the other four sites ranged from approximately 2.5-fold with Thr 69 to more than 4-fold with Thr 45 . However, the different fold increases in 32 P in Thr 36 , Thr 45 , Ser 64 , and Thr 69 produced by the hormone were a function of the basal state of phosphorylation, as the four sites contained approximately the same amount of 32 P after insulin treatment.
Incubating cells with rapamycin alone decreased the amount of 32 P in all of the sites except Ser 82 (Fig. 8). Differences among the sites in sensitivity to rapamycin emerged in the presence of insulin. Rapamycin markedly inhibited the insulin-stimulated phosphorylation of Thr 45 and Thr 69 (Fig. 8). In contrast, the increases in the 32 P content of Ser 64 produced by insulin in the presence and absence of rapamycin were approximately equal. Similarly, rapamycin had little if any effect on the increment in the 32 P content of Thr 36 produced by insulin.
Identification of a Phosphorylated Form of PHAS-I That Binds eIF-4E-A partially phosphorylated form of PHAS-I has been previously shown to bind labeled eIF-4E in far-Western analyses (17). Likewise, some phosphorylated PHAS-I was recovered with eIF-4E when PHAS-I⅐eIF-4E complexes were purified from adipocyte extracts by using m 7 GTP-Sepharose (15,16). To identify the phosphorylated sites in the eIF-4E-bound form of PHAS-I, PHAS-I⅐eIF-4E complexes were purified from extracts of 32 P-labeled adipocytes that had been incubated with rapamycin plus insulin. After digesting the PHAS-I with lysyl endopeptidase, essentially all of the 32 P was recovered in LE-P3. 2 When the peptides in this peak were digested with chymotrypsin and subjected to HPLC, the 32 P was found in CT-P2 (Fig. 9, lower panel), indicating that the eIF-4E bound form was phosphorylated exclusively in Thr 36 . For comparison, note that the PHAS-I protein that did not bind to the cap affinity resin contained 32 P in both Thr 36 and Thr 45 (Fig. 9, upper panel). DISCUSSION Our results provide definitive evidence that in rat adipocytes insulin stimulates the phosphorylation of PHAS-I in five sites, all of which fit a Ser/Thr-Pro motif (Fig. 1). The existence of multiple sites raises the possibility that more than one site is involved in the control of PHAS-I binding to eIF-4E. Moreover, identification of the sequence of amino acids surrounding the different phosphorylated residues has provided information needed to identify the protein kinases responsible for phosphorylating PHAS-I in cells. Thus, the results have important implications with respect to not only the control of PHAS-I binding to eIF-4E but also to the mechanisms of action of insulin and rapamycin.
Multisite Phosphorylation and the Control of PHAS-I Binding to eIF-4E-All five of the phosphorylation sites identified in rat PHAS-I are conserved in not only the mouse (15) and human proteins (4), but also in PHAS-II, another eIF-4E binding protein that is approximately 60% identical to PHAS-I (4,17). The conservation of sites suggests that all may be important for the function of the protein. However, our results indicate that phosphorylation of certain sites can have a much greater impact on the binding of PHAS-I to eIF-4E than phosphorylation of others. For example, phosphorylation of Ser 64 , the only site that had been previously identified in PHAS-I FIG. 9. A phosphorylated form of PHAS-I that remains bound to eIF-4E is phosphorylated exclusively in Thr 36 . Rat adipocytes were incubated with 32 P i and then incubated with rapamycin plus insulin as described in the legend to Fig. 5. eIF-4E was partially purified by using m 7 GTP-Sepharose as described under "Experimental Procedures." Samples of [ 32 P]PHAS-I that copurified with eIF-4E (upper panel) and of the [ 32 P]PHAS-I that did not bind to m 7 GTP-Sepharose (lower panel) were then digested with lysyl endopeptidase C. The digests were subjected to HPLC. Peptides recovered in LE-P3, which was the only peak detected with the [ 32 P]PHAS-I that copurified with eIF-4E, were incubated with chymotrypsin, and the resulting [ 32 P]phosphopetides were resolved by HPLC.

FIG. 8. Effects of insulin and rapamycin on the phosphorylation of sites in PHAS-I.
Rat adipocytes were incubated with 32 P i and then incubated without additions or with rapamycin and/or insulin before PHAS-I was immunoprecipitated and [ 32 P]phosphopeptides were isolated as described in the legend to (24), did not attenuate binding of PHAS-I to eIF-4E as assessed by far-Western blotting (Fig. 3). Indeed, in contrast to our previous hypothesis (4,11) phosphorylation of this site was neither necessary nor sufficient for inhibiting binding of PHAS-I to eIF-4E, at least in far-Western analyses. Examining the phosphorylation patterns of PHAS-I in cells provided clues concerning the relative importance of the other sites. For example, the finding that PHAS-I phosphorylated in Thr 36 remained bound to eIF-4E in extracts of rat adipocytes (Fig. 9) indicates that phosphorylation of Thr 36 is not sufficient for dissociation of the PHAS-I⅐eIF-4E complex. It would also seem unlikely that Ser 82 phosphorylation could explain the near complete inhibition of PHAS-I binding to eIF-4E produced by insulin (4,11) since Ser 82 appeared to contain much less 32 P than any of the other four sites (Fig. 8). However, the apparently low stoichiometry of phosphorylation of Ser 82 may have been due to poor recovery of CT-P3. Therefore, a role for Ser 82 phosphorylation in regulating PHAS-I is still possible. Moreover, it should be stressed that with five sites of phosphorylation, complex interactions are possible, and phosphorylation of sites such as Thr 36 and Ser 64 could increase or decrease the rate of phosphorylation and/or the influence of other sites on eIF-4E binding. Nevertheless, through the process of elimination, Thr 45 and Thr 69 have emerged as candidates for important regulatory sites.
Implications of Multisite Phosphorylation With Respect to Signaling Pathways-The protein kinases responsible for phosphorylating PHAS-I in adipocytes have not been determined. However, it is now possible to exclude certain protein kinases that are capable of phosphorylating PHAS-I in vitro. Diggle et al. (25) first demonstrated that casein kinase II phosphorylated PHAS-I purified from adipocyte extracts. Based on evidence that insulin activated casein kinase II in adipocytes, it was proposed that casein kinase II mediated the phosphorylation of PHAS-I by insulin (25). None of the five sites phosphorylated in rat adipocytes meet the minimum consensus requirement for phosphorylation by casein kinase II (26). We have recently found that casein kinase II preferentially phosphorylates Ser 111 , which is not phosphorylated in cells. 2 Thus, it is clear that casein kinase II cannot be the major mediator of the action of insulin on PHAS-I. Similarly, none of the sites phosphorylated in vitro meet the requirements for phosphorylation by protein kinase C (26), which phosphorylates recombinant PHAS-I (6,24).
MAP kinase is by far the most effective kinase that has been described for phosphorylating PHAS-I in vitro (6,16,24). MAP kinase is activated by insulin in adipocytes (39), and all five sites phosphorylated by MAP kinase in vitro were phosphorylated in response to the hormone (Fig. 6). Despite this finding, neither the ERK1 nor ERK2 isoform is the sole mediator of the phosphorylation of PHAS-I in response to insulin in adipocytes as blocking MAP kinase activation with the MEK inhibitor, PD 098059, did not attenuate the effect of insulin in these cells (15). As discussed previously (15), this result does not eliminate the possibility that MAP kinase contributes to the control of PHAS-I phosphorylation as more than one pathway may be involved. Interestingly, PD 098059 was recently found to promote dephosphorylation of PHAS-I in CHO cells (40). Thus, there is reason to suspect that MAP kinase, or another kinase regulated by MEK, is involved in the control of PHAS-I in certain cell types.
PHAS-I phosphorylation is controlled by a rapamycin-sensitive pathway that is distinct from the MAP kinase signaling pathway (15,16,19,21). The most extensively characterized enzyme regulated by rapamycin is p70 S6K , a protein kinase that is activated by insulin and a variety of mitogenic stimuli (23). Full activation of the kinase appears to require phosphorylation of two classes of sites. One class consists of three Ser/Thr residues (Thr 229 , Thr 389 , and Ser 404 ) that are flanked by aromatic residues (41). The other consists of four Ser/Thr-Pro residues found in a 14-amino acid stretch located near the COOH terminus of the kinase in a region referred to as the autoinhibitory domain (42,43). Rapamycin is a potent inhibitor of the activation of p70 S6K (44,45), and an obvious possibility was that p70 S6K was responsible for phosphorylating PHAS-I in cells. However, this hypothesis was eliminated by the finding that PHAS-I was not a substrate for p70 S6K (6,16,24).
The five phosphorylation sites in PHAS-I resemble the four sites in the autoinhibitory domain of p70 S6K (42). Both sets of sites are markedly increased in response to insulin or mitogenic stimulation (6,23), and it is an intriguing possibility that the Ser/Thr-Pro sites in the two proteins are phosphorylated by the same protein kinase. Phosphorylation of PHAS-I and p70 S6K by a common kinase would provide a mechanism linking capped mRNA translation, which is dependent on eIF-4E (9), and polypyrimidine tract mRNA translation, which appears to be regulated by phosphorylation of ribosomal protein S6 (46). However, circumstances might exist in which it would not be advantageous for a cell to up-regulate translation of both classes of mRNA, and the potential exists for selective regulation of the phosphorylation of PHAS-I and p70 S6K . PHAS-I lacks the hydrophobic class of sites found in p70 S6K as none of the Ser/Thr residues in PHAS-I are flanked by aromatic residues. The closest resemblance to the hydrophobic sites are two Ser that are adjacent to Phe and Tyr in a tandem repeat of the sequence, (Tyr/Phe)-Ser-Thr-Thr-Pro-Gly-Gly (Fig. 1). Neither of these Ser residues were phosphorylated to a significant extent in adipocytes as relatively little 32 P released in cycle 1 when the CT-P1 and CT-P2 peptides were subjected to sequential Edman degradation (Fig. 6). In view of the differences in sequence surrounding the two classes of sites, it seems clear that a single protein kinase is not responsible for phosphorylating PHAS-I and the hydrophobic sites in p70 S6K . This is an important point in considering the action of rapamycin, which decreases four of the Ser/Thr-Pro sites in PHAS-I (Fig. 8) and all three hydrophobic sites in p70 S6K (41). However, it should be noted that rapamycin does promote dephosphorylation of p70 S6K in Ser 411 , a site having Pro in the ϩ1 position (41).
It is well established that rapamycin inhibits the mTOR signaling pathway, but how mTOR signals is still a mystery (22). A region of mTOR has homology with the catalytic domain of phosphatidyl inositol 3Ј-OH kinase, but it is not clear that mTOR functions as a lipid kinase. Indeed, there is good reason to suspect that mTOR signals as a protein kinase as it is homologous to the new class of protein kinases, which include the catalytic subunit of DNA-dependent protein kinase and the ATM gene product that is mutated in ataxia telangiectasia (47). In addition, the protein undergoes autophosphorylation in a reaction that is subject to inhibition by rapamycin (48). The effect of rapamycin on autophosphorylation, as well as the inhibitory effect of rapamycin on mTOR in cells, requires an additional receptor protein, FKBP12, as it is the rapamycin⅐FKBP12 complex that actually binds to mTOR (22). Rapamycin binding to FKBP12 is competitively inhibited by FK506, and because the FK506/rapamycin complex does not bind mTOR, FK506 competitively inhibits the effects of rapamycin on mTOR-dependent pathways (22).
The findings that FK506 blocks the effects of rapamycin on PHAS-I (15,19) and p70 S6K (44,45) provide strong evidence that the mTOR signaling pathway regulates both proteins. Results of recent experiments in T lymphoma cells (YAC-1) provide additional evidence implicating mTOR in the control of PHAS-I (20). In wild-type lymphoma cells rapamycin potently decreased PHAS-I phosphorylation, but rapamycin had little, if any, effect on PHAS-I in mutant lymphoma cells that had been selected on the basis of resistance to the antiproliferative effects of rapamycin (20). There is strong evidence that the rapamycin-resistant phenotype is due to a mutation which decreases the affinity of mTOR for rapamycin⅐FKBP-12 (49). Thus, there is both pharmacologic and genetic evidence supporting the conclusion that the mTOR signaling pathway is involved in the control of PHAS-I. The challenge now is to identify the PHAS-I kinases and/or phosphatases involved in the mTOR-dependent control of PHAS-I.