A Quantitative Molecular Model for Modulation of Mammalian Translation by the eIF4E-binding Protein 1*

Translation initiation is a key point of regulation in eukaryotic gene expression. 4E-binding proteins (4E-BPs) inhibit initiation by blocking the association of eIF4E with eIF4G, two integral components of the mRNA cap-binding complex. Phosphorylation of 4E-BP1 reduces its ability to bind to eIF4E and thereby to compete with eIF4G. A novel combination of biophysical and biochemical tools was used to measure the impact of phosphorylation and acidic side chain substitution at each potentially modulatory site in 4E-BP1. For each individual site, we have analyzed the effects of modification on eIF4E binding using affinity chromatography and surface plasmon resonance analysis, and on the regulatory function of the 4E-BP1 protein using a yeast in vivo model system and a mammalian in vitro translation assay. We find that modifications at the two sites immediately flanking the eIF4E-binding domain, Thr46 and Ser65, consistently have the most significant effects, and that phosphorylation of Ser65 causes the greatest reduction in binding affinity. These results establish a quantitative framework that should contribute to understanding of the molecular interactions underlying 4E-BP1-mediated translational regulation.

Regulation of protein interactions and protein activity via changes in phosphorylation status is a common theme in biology (see, for example, Refs. 1 and 2). Eukaryotic translation is an example of a major cellular process that may be regulated in this way at a number of steps in the overall pathway (3). A key point of translational regulation is at the step of ribosome binding to the 5Ј end of cellular mRNA. Ribosome-mRNA binding involves interactions mediated by the cap-binding complex eIF4F 1 (Fig. 1A). Mammalian eIF4F comprises the eukaryotic initiation factors eIF4E, eIF4G, and eIF4A (4). The largest eIF4F component (5), eIF4G, has binding sites for eIF4E, eIF3, eIF4A, the poly(A)-binding protein and the mitogen-activated protein kinase-activated protein kinase Mnk1 (5, 6 -10). More recent work indicates that yeast eIF4G also binds to the mRNA decapping protein (Dcp1; Ref. 11). The eIF4F complex is teth-ered to the mRNA cap by the 25-kDa cap-binding protein, eIF4E. eIF3 is believed to form a bridge between the 40 S ribosomal subunit and eIF4F, whereas the binding of eIF4G to poly(A)-binding protein may play a role in promoting interaction between the 3Ј and 5Ј ends of mRNA. The significance of the interaction between eIF4G and eIF4A is not clear, but the latter factor (together with eIF4B) catalyzes ATP-dependent RNA helicase activity that may promote ribosomal scanning along structured mRNA (12). Mnk1 phosphorylates eIF4E, which may in turn increase this factor's affinity for the mRNA cap (13).
The cell regulates eIF4F activity in response to environmental changes and various stimuli, including temperature stress and stimulation by hormones (3,4,14,15). Deviations from normal levels of eIF4F activity can have drastic effects on cellular functions. For example, overexpression of eIF4E leads to cell transformation and changes in cell morphology (16,17). The eIF4E-eIF4G interaction is of central importance for capdependent initiation, and can be blocked by small regulatory proteins that bind to eIF4E, the 4E-binding proteins (4E-BPs; Refs. 18 and 19). Inhibition by 4E-BP1 is apparently inversely dependent on its level of phosphorylation, whereby the hypophosphorylated form binds most tightly to eIF4E (19 -21). Cell transformation caused by overexpression of eIF4E is reversed by simultaneous overexpression of the 4E-BP1 gene (22).
Recent work (23)(24)(25) has shown that both the yeast and the mammalian eIF4E proteins each interact with eIF4G via a site that maps to the face opposite to the ventral cap-binding slot (26). The binding sites for eIF4G of Saccharomyces cerevisiae and the yeast eIF4E-binding protein p20 (which is also a phosphoprotein (27)) overlap within this dorsal region on yeast eIF4E (23). There is also another yeast 4E-binding protein, called Eap1, that is suspected to act partially as a 4E-BP1 functional analogue (28), and possesses an eIF4E-binding motif similar to that of p20. 4E-BP1 binds to the equivalent region on the dorsal face of human eIF4E (24,25), but this regulatory protein uses a slightly different set of molecular contacts to those involved in eIF4G binding (24,29). The binding of protein ligands at the dorsal site of yeast or human eIF4E is capable of stabilizing cap binding at the ventral cap-binding slot (23,24,30,31), although the physiological significance of this effect has yet to be determined.
The mammalian 4E-BPs seem to mediate strong regulation of translation in vivo (3,12). This is in contrast to yeast p20 (30,32,33), which may possibly be involved in the fine tuning of translation rates (23,24). Given the key role of the 4E-BPs in the regulation of translation, it is important to elucidate the mechanism underlying this modulation. The binding affinities of the nonphosphorylated 4E-BPs for human eIF4E are high (estimated K D values of 10 Ϫ8 -10 Ϫ9 M) and comparable with recent estimates of the binding affinity between eIF4E and the eIF4E-binding domain of eIF4G of yeast (23,24,34). Thus, hypophosphorylated 4E-BPs have the potential to compete effectively with eIF4G for binding to eIF4E and therefore to act as potent translational inhibitors. In the present study, we extend this comparison by estimating the binding affinities of phosphorylated forms of a 4E-BP.
Six sites of phosphorylation on rat 4E-BP1 (also known as PHAS-I) have been mapped (Fig. 1B), but there is disagreement about the roles of the respective sites in regulation, while the kinases involved in vivo remain poorly characterized. These correspond in the human 4E-BP1 sequence (which is almost identical to PHAS-I) to Thr 37 , Thr 46 , Ser 65 , Thr 70 , Ser 83 , and Ser 112 (35, 36; Fig. 1iB). In rat adipocytes, phosphorylation at the first four of these sites is increased in response to insulin and partially decreased in the presence of rapamycin (35,37). The effect of rapamycin is currently explained by proposing that at least two of the sites (Thr 37 and Thr 46 ) are phosphorylated by a FRAP/mTOR-associated pathway (38). The phosphoinositide 3-kinase and the Akt/PKB protein kinase also seem to act via a further pathway that does not involve FRAP/mTOR (39). However, the kinases directly responsible for phosphorylating 4E-BP1 in vivo have not been identified, and much uncertainty remains about the signaling pathways leading to phosphorylation of the 4E-BPs (see Ref. 3 for review).
The other key area yet to be resolved concerns the influence of phosphorylation on the molecular function of 4E-BP1. Some reports have concluded that phosphorylation of Thr 37 and Thr 46 (40), or of Thr 46 alone (35,41), is capable of inhibiting the eIF4E-4E-BP1 interaction. In contrast, another group observed no loss of binding associated with phosphorylation at these two sites, and proposed that their primary function is to trigger phosphorylation at the other sites on 4E-BP1 (38). This interdependence of phosphorylation events has been questioned by others (37,42). Heesom and Denton (42) have concluded that the phosphorylation of Ser 65 leads to dissociation of 4E-BP1-eIF4E, whereas phosphorylation at Thr 37 and Thr 46 does not disrupt this complex. In this paper we focus on the molecular consequences of phosphorylation at the sites on 4E-BP1 that are most relevant to eIF4E binding, showing how these can contribute to the regulatory function of this protein in vivo. The result is a quantitative biophysical model of molecular regulation that helps explain how different degrees of inhibition can be achieved by phosphorylation at the respective phosphorylatable sites.

MATERIALS AND METHODS
Mutagenesis and Plasmid Construction-Mutagenesis via a polymerase chain reaction-based technique (43) yielded products which could be inserted between either the NdeI and EcoRI sites of the yeast expression vector YCpSUPEX2 (44) or the NdeI and BamHI sites of the Escherichia coli expression vector pCYTEXP3 (45). The genes for expression in yeast all encode an N-terminal "Flag," MDYKDDDDKT, which appears to enhance the stability of the protein in yeast (24,29). The genes for expression in E. coli all contain 12 histidine codons between the first and second 4E-BP1 codons, thus facilitating (using nickel ion affinity) both purification of the protein and binding to the sensor chip surface used for SPR. The nucleotide sequence of each of the variants in both vectors was confirmed to be free of any spurious mutations.
Yeast Growth Assay for 4E-BP Function-4E-BP1 function was assayed in yeast strain Jo56 (cdc33::LEU2 his3 ura3 ade3 (YCpTrp-hu4E TRP1), Ref. 29) after transformation to uracil prototrophy in glucosecontaining selective medium with the YCpSUPEX2 plasmid derivatives. The cdc33 disruption in Jo56 was complemented by the human eIF4E gene in the plasmid YCpTrp-hu4E. 4E-BP1 expression was induced by transferring the cells to galactose-containing selective medium. Expression of each of the 4E-BP1 derivatives was also tested in a strain with intact CDC33 as control. The phenotypes of two independently isolated derivatives of each 4E-BP1 allele were tested, and each allele was re-tested after re-cloning into a fresh YCpSUPEX2 vector background.
Synthesis and Purification of Recombinant Proteins in E. coli-Expression of the 4E-BP1 genes contained in pCYTEXP3 was achieved by a temperature shift from 30 to 42°C in the protease-deficient E. coli strain CAG629 (46). The proteins were then purified denatured on nickel-nitrilotriacetic acid-agarose (Qiagen).
Phosphorylation of 4E-BP1 in Vitro-For analysis of the sites of phosphorylation, 2 g of 4E-BP1 were incubated in 10 mM HEPES, pH 7.4, 1 mM dithiothreitol, 2 mM magnesium chloride, 1% (v/v) glycerol, and 5 mM ATP with 5 Ci of [␥-32 P]ATP and 0.45 units of activated recombinant ERK2 in a total volume of 20 l for 2 h at 23°C. The reference activity of ERK2 was estimated using myelin basic protein as substrate. ERK2 was generously given by Dr. C. Armstrong, Division of Signal Transduction Therapy, University of Dundee. For SPR analysis, 20 g of recombinant 4E-BP1 was phosphorylated with 2.25 units of ERK2 under the same conditions, but in a total reaction volume of 100 l and excluding radiolabeled ATP.
Phosphopeptide Mapping-The phosphorylation reaction was stopped by adding SDS and ␤-mercaptoethanol, each to 1%, and heating at 95°C for 5 min. The proteins were separated by SDS-polyacrylamide gel electrophoresis and radiolabeled 4E-BP1 was eluted from the gel by shaking overnight in 50 mM Tris-HCl, pH 7.5, 0.1% SDS, and 5% ␤-mercaptoethanol. Gel pieces were removed with Spin-X centrifuge tube filters (Costar) and the labeled protein was precipitated in 10% trichloroacetic acid. After removal of the trichloroacetic acid by extensive washing, the protein was digested overnight at 30°C with trypsin (sequencing grade, Roche Molecular Biochemicals) in 20 mM ammonium bicarbonate, 0.1% Zwittergens, or with trypsin and chymotrypsin (Roche Molecular Biochemicals) in 20 mM ammonium bicarbonate and 0.2% n-octylglucopyranoside (Calbiochem). The peptide mixtures were acidified by adding 3 volumes of 0.1% (v/v) trifluoroacetic acid in water and analyzed by high pressure liquid chromatography (HPLC) (Gilson) using a Vydac C 18 (250 mm long ϫ 4.6-mm internal diameter) reversedphase column developed with 0.1% trifluoroacetic acid in water/acetonitrile. The main radiolabeled fractions were analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (47). Two-dimensional mapping was performed as previously described (27,48). Tryptic peptide mixtures were dried, then redissolved in buffer for the first dimension separation. Electrophoresis in the first dimension was performed at pH 1.9, and chromatography in the second dimension with buffer containing isobutyric acid.
Reticulocyte Lysate Translation Assays-Capped and polyadenylated luciferase mRNA was synthesized in vitro from a luciferase gene inserted into plasmid pHST7 (11). 16.5 l of a rabbit reticulocyte lysate (micrococcal nuclease-treated, Promega) was mixed with 125 ng of 4E-BP1 in 0.5 l of buffer A (20 mM HEPES, pH 7.4, 100 mM KCl, 2 mM MgCl 2 ), plus RNasin (final concentration 1 unit/l) and amino acids (final concentration of each, 40 M) in a total volume of 24 l. The mixture was preincubated at 30°C for 10 min, programmed by the addition of 1 l containing 0.1 g of mRNA and incubated a further 90 min. Luciferase activity was then measured using a luminometer. m 7 GTP-Sepharose Chromatography-2 g of 4E-BP1 were labeled with 32 P, as described above, and separated from [␥-32 P]ATP by gelfiltration through Sephadex G-25 (Amersham Pharmacia Biotech). The sample was mixed with a 5-fold molar excess of eIF4E, to give a total volume of 370 l (in buffer A, see above), then incubated at 15°C for 15 min before mixing with 30 l of m 7 GTP-Sepharose resin (Amersham Pharmacia Biotech) and incubating for a further 90 min with gentle agitation. The resin was subsequently washed three times with 400 l of the same buffer, and bound proteins were eluted with 100 l of 0.1 mM m 7 GTP in the same buffer. In controls, further elution steps were performed to demonstrate that all of the eIF4E (and, if present, associated proteins) had been eluted. The protein fractions were precipitated with ice-cold acetone and resolved by denaturing 15% polyacrylamide gel electrophoresis. The quantities of radiolabeled proteins in the gel were determined using a Molecular Dynamics Typhoon 8600 image analyzer (Amersham Pharmacia Biotech) with Image Quant software.
Purification of Phosphorylated 4E-BP1-Phosphorylated 4E-BP1 was separated from nonphosphorylated 4E-BP1 by native gel electrophoresis after loading the phosphorylation reaction mixture directly onto a 12.5% polyacrylamide gel (49), with an aliquot of 32 P-labeled 4E-BP1 run in parallel as a marker. Phosphoprotein was eluted from macerated gel slices directly into buffer B (10 mM HEPES pH 7.4, 150 mM NaCl, 0.005% surfactant P20 (Biacore), and 50 M EDTA). Gel pieces were removed using a Spin-X centrifuge tube filter (Coster) and the solution was concentrated using a Vivaspin polyethersulfone membrane with a 5-kDa exclusion limit (Vivascience).
Surface Plasmon Resonance Analysis-All SPR assays were performed using a Biacore 3000 instrument essentially as described previously (24). The 4E-BP1 proteins in Buffer B (described above) were immobilized on the sensorchip of type NTA (Biacore). For estimation of the dissociation constant (K D ), six different eIF4E concentrations, ranging from 10 nM to 1 M, were allowed to interact with the immobilized 4E-BP1 protein. The resulting sensorgrams were evaluated using the BIA Evaluation software package. The curves were analyzed using local fittings for Langmuir binding and averaged for each protein.  Fig. 3). Ser 112 was not included in this study because its phosphorylation is neither rapamycin-sensitive nor is it thought to affect 4E-BP1 binding to eIF4E (36).

Modification of 4E-BP1 Phosphorylation Sites to Glu and
We assessed the effects of these changes initially using a yeast growth assay. Previous work established a system whereby the synthesis of human 4E-BPs in yeast causes inhibition of translation and consequent reduction in growth (24,29). This allows us to study translational regulation by 4E-BP1 in vivo. The system is dependent on replacement of the endogenous yeast eIF4E gene (CDC33) by the human eIF4E gene, which complements cdc33. Human, but not yeast, eIF4E is able to bind 4E-BP1 tightly, therefore only the strain with the eIF4E substitution is sensitive to 4E-BP1 expression; expression of the human 4E-BP1 gene in an unmodified yeast strain has no inhibitory effect on growth.
The 4E-BP1 gene sequences were introduced into the test strain on a plasmid that allows conditional induction. We used a colony growth assay to provide semiquantitative estimates of growth inhibition and thus first indications as to the impact of the amino acid substitutions on 4E-BP1 repressive activity. When synthesis of "wild-type" 4E-BP1 was induced, colony growth was inhibited, the cell doubling time being increased to over 15 h at 30°C (Fig. 2, WT), compared with the control, in which no 4E-BP1 was synthesized and the doubling time was about 3 h (Fig. 2, CON). When the modified 4E-BP1 genes were induced, the phenotypes of most of the substituted forms were indistinguishable from that of the wild type and showed strong growth inhibition. Two of the forms, however, involving substitution of Asp for Thr at position 46 and of Glu for Ser at position 65, showed partial relief of the growth inhibition (Fig. 2, S65E and T46D; Fig. 3). All of the yeast strains grew normally when the 4E-BP1 genes were repressed (Fig. 2, Repressed). Moreover, in an unmodified control strain expressing the endogenous yeast eIF4E, none of the modified 4E-BP1 genes caused any inhibitory effects (data not shown). The fact that acidic side chain substitution has effects at positions 46 and 65 suggests that these two sites are of primary importance in terms of their potential influence on the affinity of 4E-BP1 for eIF4E.
Effects on Translational Repression of Glu and Asp Substitutions-If the effects on yeast growth caused by the T46D and S65E substitutions are due to reduced binding to eIF4E, then it should be possible to observe these effects on translation more directly. The effects of the acidic residue substitutions in 4E-BP1 on translation were therefore investigated using a reticulocyte lysate translation assay. The Asp-or Glu-substituted 4E-BP1 proteins were synthesized in, and purified from, E. coli (an N-terminal poly-His "tag" was used to facilitate purification). We then investigated their effects on the translation of a capped, firefly luciferase mRNA in the assay system. The optimal ratio of 4E-BP1 to capped mRNA to give maximal inhibition of cap-dependent translation was established prior to the experiment by determining dose-response curves, which were found to follow a saturation profile (data not shown). The effects, expressed as the proportion of luciferase activity com-FIG. 1. A, schematic representation of how a 4E-BP is thought to block eIF4F formation and association with the cap structure at the 5Ј end. The 4E-BP competes with eIF4G for binding to the dorsal site on eIF4E, forming a stable heterodimeric complex that, although still able to bind to the 5Ј cap (not shown here; see Ref. 24), is no longer able to interact with eIF4G. Phosphorylation of the 4E-BP reduces the affinity between the 4E-BP and eIF4E, thus allowing eIF4G to compete more effectively for binding to eIF4E. This means that the active eIF4F complex can assemble on the mRNA, promoting cap-dependent translation initiation. The relationship between eIF4E-4E-BP formation and interactions between eIF4G and eIF4A (and thus eIF4B) are not understood. B, the human 4E-BP1 sequence and the seven known phosphorylation sites of 4E-BP1. The sites functionally studied in this work are marked with p. The phosphorylation of Ser 101 , described in this work, is not known to occur in vivo, while the phosphorylation of Ser 112 is not thought to be relevant to modulation of 4E-BP1 activity in vivo (36). pared with the reaction in the absence of added 4E-BP1, are illustrated in Fig. 3. As observed previously (19), wild-type 4E-BP1 imposes partial, rather than complete, inhibition on translation in a reticulocyte lysate system. This inhibition is specifically attributable to 4E-BP1 binding to eIF4E since, in control experiments, the addition of an equimolar amount of recombinant eIF4E together with 4E-BP1 was found to reverse the effect. In the case of the Glu substitutions, just as is observed in the yeast growth assay, only substitution of Glu for Ser at position 65 has any significant effect in alleviating the inhibition of translation (Fig. 3, BP1-S65E). In the case of the Asp substitutions, T46D and S65D have the most significant effects, and of these, as observed in the yeast growth assay, the effect of T46D appears to be the strongest (Fig. 3, BP1-T46D and -S65D). These results are therefore broadly consistent with those of the yeast growth assay in showing that the presence of an acidic side chain at either positions 46 or 65 causes the greatest reduction in the repressive potential of 4E-BP1.
Asp and Glu Substitutions Reduce 4E-BP1:eIF4E Binding Affinity-We then asked whether these effects of acidic side chain substitutions on translation and cell growth correlate with changes in the affinity of eIF4E binding. The affinities of the substituted forms of 4E-BP1 for eIF4E were estimated by means of surface plasmon resonance (SPR) analysis. This technique was performed using 4E-BP1 as the ligand (immobilized on the sensor surface) and eIF4E as the analyte (passed in solution over the sensor surface). The substituted forms of 4E-BP1 were polyhistidine tagged at the N terminus to facilitate binding to the sensor chip. These were the same forms as those analyzed in the reticulocyte lysate assay and were therefore known to inhibit translation. Changes in the resonance response observed upon the introduction and then removal of eIF4E were measured, and estimates of the kinetic constants based on the respective response curves allowed us to derive apparent dissociation constants for each of the substituted 4E-BP1⅐eIF4E complexes (Fig. 3). Typical response curves for the Glu-substituted and Asp-substituted 4E-BPs at a single eIF4E concentration are shown in Fig. 4, A and B, respectively. The pattern of variation in the K D values of the respective complexes broadly reflects the differences observed in the effects on yeast growth and on translation in vitro (Figs. 2 and 3). Thus, of the Glu substitutions, only the BP1-S65E⅐eIF4E complex manifests a significant elevation in K D (2-3-fold compared with wild-type), and S65E is the only Glu substitution that allows increased yeast growth and an increase in translation in the reticulocyte lysate system, compared with wild-type 4E-BP1. Of the Asp substitutions, both the BP1-T46D⅐eIF4E and BP1-S65D⅐eIF4E complexes have elevated K D values (again 2-3-fold compared with wild-type 4E-BP1). BP1-T46D and BP1-S65D also allow the highest levels of translation in reticulocyte lysate and, of these two, T46D causes the most prominent increase in K D and translation in vitro, and also allows an increased rate of yeast growth.
We conclude that substitution of an acidic side chain at either of positions 46 or 65 within the 4E-BP1 molecule can cause a 2-3-fold reduction in the binding affinity for eIF4E, and that this reduction in affinity is sufficient to cause detectable changes in cap-dependent translational activity and in cell growth rate. Because acidic side chain substitutions are potentially capable of partially mimicking the effects of phosphorylation, these findings provided a good starting point for a comparative analysis of the effects of phosphorylation at the respective (T/S)P sites.
Generating Single-site Phosphorylated Forms of 4E-BP1-In order to investigate the effects of phosphorylation at individual sites, a series of 4E-BP1 proteins were synthesized in which alanine had been substituted at four of the five (T/S)P sites, leaving in each case one of the sites unchanged and susceptible FIG. 4. SPR analysis reveals changes in the binding characteristics of 4E-BP1-eIF4E as a function of Asp and Glu substitutions at known phosphorylation sites. The interaction data for wild-type 4E-BP1 (WT) and BP1-S65E (A) and BP1-T46D (B) are compared. The plots depict the resonance unit levels detected by a BIAcore 3000 instrument as a function of time. In the first phase of each experiment, eIF4E was present in the buffer (eIF4E present). eIF4E was omitted in the second phase.

FIG. 3. Asp and Glu substitutions at 4E-BP1 phosphorylation sites change the affinity of 4E-BP1 for eIF4E and the ability of 4E-BP1 to act as a translational inhibitor.
The approximate relative rate of yeast growth in the plate assay (Fig. 2) during 4E-BP1 synthesis (from multiple trials) is represented on a two-point scale as ϩ (slight) or ϩϩ (moderate). The relative rates of synthesis of luciferase in the reticulocyte lysate assay are given as percentages of the control experiment value obtained in the absence of 4E-BP1. The K D estimates obtained via SPR are represented on a bar chart and given as numbers on the right-hand side. All experiments were performed at least three times and the white portions of the bars represent the standard deviations. to phosphorylation ("4A"-substitutions). An additional protein (BP1-5A), in which all five sites were replaced by Ala, was also synthesized as a control. Each protein was phosphorylated in vitro using recombinant activated ERK2 mitogen-activated protein kinase. ERK2 is a useful tool in these studies because it is known to phosphorylate the (T/S)P sites of PHAS-I in vitro (35) The 32 P-labeled proteins, resolved electrophoretically in a denaturing polyacrylamide gel, are shown in Fig. 5A. Wild type 4E-BP1 was phosphorylated at multiple positions, but primarily at sites Thr 46 and Ser 65 (Fig. 5B). It is apparent that BP1-5A underwent a low level of phosphorylation, indicating that in this protein at least some phosphorylation of an additional site or sites occurs. Isolation from phosphorylated BP1-5A of the phosphopeptide after tryptic digestion, followed by analysis using mass spectrometry and solid-phase sequencing, revealed that the phosphorylated residue was Ser 101 (data not shown), which is not a site where phosphorylation has been previously reported. Like the five known phosphorylation sites upstream of it, Ser 101 is also followed by proline. To examine the possibility that the other Ala-substituted forms were significantly phosphorylated at Ser 101 , we employed two different techniques: separation of the phosphopeptides by HPLC and two-dimensional mapping. The phosphorylated proteins were analyzed after cleavage either with trypsin alone or in combination with chymotrypsin. The HPLC analyses yielded single phosphopeptide peaks (see, for example, Fig. 5, C and D), indicating that the proteins are indeed phosphorylated almost exclusively at the expected sites. A few cases where the HPLC resolution was not clear were investigated further, as outlined below.
The HPLC profile for BP1-T37(4A) indicated incomplete digestion by one of the two enzymes, while mass spectrometric analysis of the major tryptic phosphopeptides derived from BP1-T46(4A) resolved two products (not resolved by HPLC) resulting from alternative cleavages at adjacent arginines (Arg 19 and Arg 20 ). Therefore, to resolve the phosphorylation patterns of BP1-T37(4A) and BP1-T46(4A) in more detail, the tryptic digests were analyzed by two-dimensional mapping (Fig. 5E). The two different phosphopeptides, one containing and one lacking an N-terminal arginine, were resolved by this method (marked by arrows in Fig. 5E). Amino acids 37 and 46 are contained in the same tryptic peptide, thus yielding almost identical peptide maps. In conclusion, these data confirm that BP1-T37(4A) and BP1-T46(4A) are phosphorylated by ERK2 primarily at Thr 37 and Thr 46 .
Due to the failure of cleavage at Arg 73 , as revealed by analysis of the peptides by mass spectrometry (data not shown), Ser 83 occurs within a large peptide spanning residues 70 -99. The HPLC profile of the BP1-5A protein digested with trypsin and chymotrypsin showed phosphopeptides containing phosphorylated Ser 101 with retention times of 11.9 and 14.5 min (data not shown); these can also be seen as minor peaks in the profile of BP1-S83(4A) (Fig. 5D). We were unable to incorporate phosphate at position Thr 70 in BP1-T70(4A) using ERK2. This suggests that prior phosphorylation at one or more other sites might be necessary. We note that others have encountered the same difficulty using a similar protein as substrate (37,41).
Taken together, these data show that the 4A-substituted proteins were indeed phosphorylated predominantly at the expected single sites and that the relative degree of phosphorylation at Ser 101 was very low. The two major peaks seen in the HPLC profile of the wild-type protein (Fig. 5B) have retention times corresponding to those of the Ser 65 and Thr 46 containing phosphopeptides (13.1 and 13.7 min, respectively). In conclusion, Ser 65 and Thr 46 are the major ERK2 phosphorylation sites in the wild-type protein in vitro.
Site-dependent Differential Effects of Phosphorylation on 4E-BP1-eIF4E Binding Affinity-How do individual phosphorylations affect the affinity of 4E-BP1 for eIF4E? We analyzed the effects of single-site phosphorylations on eIF4E binding using two different techniques. First, we tested binding in a semiquantitative manner by determining the proportion of 32 Plabeled 4E-BP1 eluting with eIF4E from m 7 GTP-Sepharose (to which eIF4E binds directly). This was achieved by preincubating 32 P-labeled 4E-BP1 preparations with a molar excess of eIF4E, allowing the eIF4E to bind to m 7 GTP-Sepharose, then resolving the unbound and elutable fractions by SDS-polyacrylamide gel electrophoresis and measuring the proportion of labeled protein in each fraction (Fig. 6A). The distribution of each phosphoprotein between the bound and unbound fractions provides an indication of their relative affinities for eIF4E. Thus, the wild-type phosphoprotein remains predominantly (more than 80%) unbound (Fig. 6A, WT ϩ P), and therefore has a relatively low affinity for eIF4E, as expected. Of the four singly phosphorylated proteins, BP1-S65(4A) ϩ P has the lowest affinity for eIF4E (over 75% remains in the unbound fraction), BP1-T46(4A) ϩ P and BP1-T37(4A) ϩ P each have moderate affinity (50 and 46% unbound, respectively), and BP1-S83(4A) ϩ P retains relatively high affinity (35% unbound; Fig. 6A). These results therefore indicate that phosphorylation at different sites affects 4E-BP1-eIF4E binding affinity to varying degrees, and that phosphorylation at Ser 65 has the single most pronounced effect on binding affinity.
We tested the effects of phosphorylation of Ser 65 on translation directly using the rabbit reticulocyte lysate assay. First it was necessary to purify the phosphoprotein from any nonphosphorylated protein, and we achieved this by resolving the two forms by native polyacrylamide gel electrophoresis and by eluting the phosphoprotein from the gel (see "Materials and Methods"). In a comparative investigation, we observed that both unphosphorylated wild-type and BP1-S65(4A) inhibited translation to about the same degree, whereas phosphorylation of each of the proteins caused pronounced alleviation of the inhibition; phosphorylation of the wild-type somewhat more so than of BP1-S65(4A) (Fig. 6B). These results correlate well with the effects of phosphorylation on eIF4E binding (Fig. 6A), indicating that the effects on translation are almost certainly the direct result of changes in 4E-BP1-eIF4E affinity. They also confirm that phosphorylation of Ser 65 strongly reduces the ability of 4E-BP1 to inhibit translation.
We measured quantitatively the effects of individual phosphorylations on 4E-BP1-eIF4E binding affinity by means of SPR analysis. The purified 4E-BP1 phosphoproteins were used as chip ligands and eIF4E as analyte, as before. Sample sets of response curves for the phosphorylated and nonphosphorylated BP1-WT and BP1-S65(4A) are illustrated in Fig. 7, A and B, respectively, while the apparent dissociation constants for the 4E-BP1-eIF4E complexes derived from the complete analysis are listed in Fig. 8.
Phosphorylation of BP1-WT caused the greatest change in affinity: the apparent K D increased by more than 3 orders of magnitude. Of the single site phosphorylations, phosphorylation of BP1-S65(4A) had the greatest effect, causing a change in affinity of nearly 2 orders of magnitude. Phosphorylation of either BP1-T46(4A) or BP1-T37(4A) caused only a 3-5-fold change in affinity, and phosphorylation of BP1-S83(4A) caused a less than 2-fold change in affinity. These differences are represented graphically in Fig. 9A, in which the area of each circle is proportional to the relative magnitude of the apparent dissociation constant (K D ) for each of the phosphoproteins. DISCUSSION We have characterized the modulatory 4E-BP1-eIF4E interactions in order to be able to ascribe relative levels of functional significance to phosphorylation events at the respective (T/S)P sites on the regulator protein (Fig. 9). Our SPR analyses yield an estimated change in affinity of wild-type 4E-BP1 for eIF4E of approximately 3 orders of magnitude as a result of phosphorylation. Since ERK2 phosphorylates primarily Thr 46 and Ser 65 in the wild-type protein (Fig. 5), it is likely that even greater reductions in the affinity of 4E-BP1 for eIF4E would be observed if full phosphorylation of all sites were achieved. However, the large reduction in affinity seen here clearly demonstrates the wide range over which affinity can be varied in response to phosphorylation. It follows that phosphorylation of all sites on each 4E-BP1 molecule is not a requirement for strong regulation to be achievable.
The results obtained with the Asp and Glu substitutions have provided useful comparative information about the roles of the respective phosphorylation sites in vitro and in vivo. The changes in affinity associated with these substitutions were found to be comparatively small (Fig. 3), but gave a first indication that Ser 65 is of particular significance in terms of modulating 4E-BP1 affinity for eIF4E and thus its ability to inhibit translation. Thr 46 was also identified as potentially playing an important role. Yet, it is interesting to note that the Asp and Glu substitutions fall short of being good mimics of phospho- rylated Ser or Thr, at least in a quantitative sense (compare Figs. 3 and 8). The resolution of differences in binding affinity associated with the respective Asp and Glu mutations is limited, indicating overall a clear distinction between the role of Ser 65 and the other phosphorylation sites, but not differentiating as convincingly between the effects of mutations at positions 37, 46, 70, and 83.
The distinction between the roles of the respective sites on 4E-BP1 is made fully apparent by the results obtained using the single site phosphorylated forms of this protein. Phosphorylation of Ser 65 reduces the estimated affinity of 4E-BP1 for eIF4E ϳ100-fold (Fig. 8). A second group of sites (Thr 37 and Thr 46 ) shows a 3-5-fold decrease in apparent binding affinity in response to single phosphorylation events. Finally, phosphorylation of Ser 83 has little effect on binding affinity. For technical reasons, we were not able to study 4E-BP1 with a single phosphorylation at position 70. However, the results of the Asp and Glu substitution experiments suggest that the effect of phosphorylation of Thr 70 will fall somewhere in the range between Ser 83 and Thr 37 . Overall, the relationship between phosphorylation at each site and its influence on binding affinity is consistent with a model in which the proximity of the added phosphate to the eIF4E-binding motif plays a major role in determining the impact of the modification on the interaction between the two proteins (Fig. 9A). It should, however, be emphasized that while the above data are useful indicators of how the competitiveness of 4E-BP1 as an eIF4E-binding partner is likely to change in response to phosphorylation at each of the major sites, they are not equivalent to absolute estimates of the real affinity values in the cell.
How can we envisage the molecular basis of the effects of 4E-BP1 phosphorylation? The dorsal face binding site on eIF4E and the eIF4E-binding motif on 4E-BP1 are the key sites of interaction between these two proteins, although it seems likely that further intermolecular contacts play a role (23,24). The respective motifs bind via a combination of hydrophobic, van der Waals, hydrogen bonding, and salt bridge interactions (25). Presumably, differences in the three-dimensional structure of the dorsal face motif, and most likely of other residues on S. cerevisiae eIF4E, prevent a sufficiently tight association with the human 4E-BP1 (24), which is why 4E-BP1 does not function as a translational regulator in wild-type yeast cells (29).
It has been proposed that electrostatic repulsion between like-charged groups on the surfaces of 4E-BP1 and eIF4E plays a dominant role in the regulatory interplay between 4E-BP1 phosphorylation and inhibition of eIF4E function. More specifically, the acidic side chains of Glu 70 and Asp 71 in eIF4E have been identified as potential sources of electrostatic forces repelling a phosphate group on Ser 65 (25). The relative position of the Ser 65 phosphate in the eIF4E⅐4E-BP1 complex has not been determined (25), but it might be sufficiently close to Glu 70 and Asp 71 for electrostatic effects to be significant (Fig. 9C). The observation of a very large change in binding affinity upon  9. Effects of phosphorylation on the molecular environment at the binding surfaces. A, the relative impact of phosphorylation at each 4E-BP1 site on the eIF4E binding affinity and inhibitory function of this translational repressor. The relative areas of the circles are proportional to the changes in estimated affinity between eIF4E and 4E-BP1 observed with phosphorylation at the respective sites (indicated by P followed by the amino acid position). The largest circle represents the impact of phosphorylation of wild-type 4E-BP1 (P-WT). Thr 46 and Ser 65 are closest to the N-terminal limit and the C-terminal limit, respectively, of the known eIF4E-binding domain. B, a threedimensional ribbon diagram of the structure of m 7 GDP:eIF4E[28 -217]-4E-BP1 [51][52][53][54][55][56][57][58][59][60][61][62][63][64][65][66][67] reported by Marcotrigiano et al. (25). The peptide backbone structure for 4E-BP1 is not resolved beyond position 64 because no electron density data were obtainable. C, the surface structure of eIF4E in the region of interactions with the 4E-BP1 peptide shown in B. A potential position of the phosphate group of Ser 65 (P(S65)) in the 4E-BP1 binding motif peptide is shown as a circle defined by a broken line. This could potentially interact with Glu 70 , Asp 71 , and other amino acids on the dorsal surface of eIF4E. The eIF4E residue Trp 73 interacts with the helical region of the 4E-BP1 peptide used in the crystal structure (25), and is essential for eIF4E binding to eIF4G and 4E-BP1 (10,23,24). phosphorylation of Ser 65 is consistent with this model.
However, there are also other potential explanations of the large quantitative impact of Ser 65 phosphorylation reported here. First of all, it should be noted that Glu 70 and Asp 71 are in salt-bridge interactions with His 37 and Lys 36 , respectively (Fig.  9C), which may reduce the significance of the repulsive charge effects between Glu 70 , Asp 71 , and the phosphate group. Second, the presence of a phosphate group in the near vicinity could affect the pK a of Glu 70 , thus influencing the stability of the Glu 70 to His 37 side chain and main chain hydrogen-bond network. This could mean that at least the local structure of eIF4E could be altered in the presence of 4E-BP1 bearing a phosphorylated Ser 65 , thus destabilizing the eIF4E⅐4E-BP1 complex. Third, while ␣-helix N-terminal NH groups can provide effective solvation for a negatively charged phosphate group (50), the CO groups available at the C terminus of an ␣-helical region are expected to be incompatible with phosphate solvation. This means that the helical structure of the binding motif region of 4E-BP1 may be destabilized by phosphorylation of Ser 65 , which in turn would be expected to destabilize the set of interactions seen with the nonphosphorylated peptide (Ref. 25; Fig. 9C). In summary, analysis of the available structural data reveals that the marked destabilization of the eIF4E⅐4E-BP1 complex by Ser 65 phosphorylation may be attributable to electrostatic repulsion, conformational effects, or steric effects, or a combination of these. The greater distance of the other phosphorylation sites on 4E-BP1 from this interaction region may mean that they are unable to influence this core set of interactions as significantly as Ser 65 . Further work is needed to elucidate the consequences of phosphorylation at this molecular interface.
In this study, we have characterized the consequences of phosphorylation at the respective sites on 4E-BP1. The resulting data provide a first approximation quantitative framework that should contribute to more precise interpretations of the significance of the action of specific kinases on this important regulatory protein, as well as to the design of new experiments that should yield insight into the molecular mechanism underlying regulation of eIF4E-4E-BP1 binding. It should be pointed out that the presence of phosphorylation sites which can impose different degrees of influence on eIF4E binding raises the intriguing possibility that the cell can impose fine or coarse control by targeting different sites on 4E-BP1. This might be achieved via kinases of distinct specificities, providing considerable flexibility in terms of control, perhaps at the individual tissue level. Finally, to what extent the principles of quantitative control characterized here for 4E-BP1 apply to the other human 4E-BPs remains to be determined. This is a particularly interesting question because of the reported evidence that 4E-BP1 and 4E-BP2 are subject to distinct modulatory pathways in human myeloid cell differentiation (51).