Identification of the caspase-dependent cleavage site of 4E-BP1.
Our previous studies have shown that the treatment of cells with potent inducers of apoptosis, either DNA-damaging agents (38
) or staurosporine (39
), leads to the cleavage of 4E-BP1 to produce a product termed Δ4E-BP1 and that this is blocked by a broad-spectrum caspase inhibitor, Z-VAD.FMK. It has also been shown that eIF4G1
is cleaved in a caspase-dependent manner during apoptosis (4
). The time courses of both eIF4G1
and 4E-BP1 cleavage in Swiss 3T3 cells undergoing apoptosis induced by etoposide, a DNA-damaging agent, were analyzed (Fig. ). The amount of full-length eIF4G1
, measured in total-cell lysates, gradually decreased during apoptosis. However, eIF4G1
was still present in cells after 24 h of etoposide treatment, albeit at reduced levels compared to the level at 12 h. Complete loss of intact eIF4G1
was apparent by 36 h. 4E-BP1 underwent cleavage to yield a product termed Δ4E-BP1 after 18 h of etoposide treatment.
FIG. 1. Identification of the caspase-dependent cleavage of 4E-BP1. (A) Swiss 3T3 cells were treated with etoposide for the times indicated. The samples were analyzed for total eIF4G1 (top) and 4E-BP1 (bottom). The α-, β-, and γ-species (more ...)
Δ4E-BP1 has previously been shown to bind to eIF4E in apoptotic cells (6
). To study this further, eIF4E was purified by affinity chromatography on m7
GTP-Sepharose from extracts of Swiss 3T3 cells that had been treated with staurosporine (Fig. , top). The majority of the 4E-BP1 bound to eIF4E was in the cleaved state (63% of the total; Fig. , top). Analysis of 4E-BP1 within these cells revealed that Δ4E-BP1 was 28% of the total (Fig. , bottom). These data suggest that Δ4E-BP1 bound preferentially to eIF4E compared to the uncleaved 4E-BP1 within apoptotic cells.
To study further the properties of Δ4E-BP1, it was necessary to identify the caspase-dependent cleavage site. To do this, Swiss 3T3 cells were treated with staurosporine and Δ4E-BP1 was copurified with eIF4E on m7GTP-Sepharose and then subjected to N-terminal sequence analysis (Fig. ). This revealed that 4E-BP1 was cleaved after Asp24 at an atypical caspase cleavage motif, ALGD. The scissile bond (Asp-Gly) and adjacent residues are fully conserved in rat and mouse 4E-BP1 sequences and partially conserved in human 4E-BP1 (VLGD) but absent from either 4E-BP2 or 4E-BP3 (Fig. ). Given that Δ4E-BP1 is recognized by our antiserum raised against the peptide ESQFEMDI, which corresponds to the extreme C terminus of 4E-BP1, Δ4E-BP1 must contain an intact C terminus. Δ4E-BP1 therefore lacks the first 24 residues of the full-length protein. As a control to confirm that the interaction of Δ4E-BP1 with eIF4E (Fig. ) was specific and not due to its binding to the Sepharose resin, affinity chromatography was carried out using either m7GTP-Sepharose or unmodified Sepharose beads (Fig. ). Δ24/BP1 (a truncation mutant that mimics Δ4E-BP1; for more detail see Fig. ) was not retained on Sepharose itself, showing that the binding of Δ24/BP1 to m7GTP-Sepharose is specific and presumably facilitated through its interaction with eIF4E.
FIG. 2. The cleaved fragment of 4E-BP1 (Δ24/BP1) acts as a dominant inhibitor of eIF4F complex assembly and cap-dependent translation. (A) Diagrammatic representation of mutants 4E-BP1myc/his and Δ24/BP1. The amino acid sequence indicates the (more ...) Δ4E-BP1 potently inhibits eIF4F complex assembly.
A 4E-BP1 mutant lacking residues 1 to 24 and with C-terminal Myc and His tags (Δ24/BP1; see Materials and Methods) was generated to mimic Δ4E-BP1. Full-length 4E-BP1 with C-terminal Myc and His tags (4E-BP1myc/his) was also generated for use as a control. 4E-BP1myc/his and Δ24/BP1 were overexpressed in HEK293 cells, and these cells were treated as indicated in the legend for Fig. . As judged by immunoblotting, the level of expression of Δ24/BP1 was considerably lower than that of 4E-BP1myc/his and slightly lower than that of endogenous 4E-BP1 (Fig. ). Both endogenous 4E-BP1 and 4E-BP1myc/his became more highly phosphorylated after insulin treatment, as indicated by increased proportions of the upper hyperphosphorylated γ-isoform and the loss of the least-phosphorylated α-species. Δ24/BP1 underwent only partial phosphorylation, with a modest increase in the more slowly migrating form. The levels of 4E-BP1 and eIF4G1 bound to eIF4E were then compared (Fig. ). 4E-BP1myc/his competed with the endogenous 4E-BP1 for binding to eIF4E. Insulin treatment, as expected, caused a time-dependent dissociation of 4E-BP1myc/his from eIF4E, allowing consequent association of eIF4G1 with eIF4E. This indicates that the response of 4E-BP1myc/his to insulin is similar to that of endogenous 4E-BP1. 4E-BP1myc/his completely dissociated from eIF4E by 60 min. However, this was slower than the dissociation of endogenous 4E-BP1, probably reflecting the greater total amount of 4E-BP1 within these cells that must be phosphorylated for this to occur. In contrast, Δ24/BP1 was not released from eIF4E, i.e., the ratio of Δ24/BP1 to eIF4E did not decrease with insulin. As a consequence, Δ24/BP1 inhibited eIF4F complex assembly even though its expression levels are lower than those for 4E-BP1myc/his.
Since the endogenous 4E-BP1 still became phosphorylated in these cells (Fig. ), it is unlikely that Δ24/BP1 inhibits signaling through mTOR. To confirm this, the regulation of p70 S6 kinase (another target of mTOR signaling) was also examined. The activity of p70 S6 kinase is regulated by phosphorylation: more highly phosphorylated forms exhibit higher kinase activity and migrate more slowly during SDS-PAGE. Insulin treatment caused similar shifts to the more highly phosphorylated isoforms of p70 S6 kinase in cells expressing Δ24/BP1 and in cells transiently transfected with the empty vector (Fig. ), confirming that Δ24/BP1 does not inhibit mTOR signaling.
Δ24/BP1 inhibits cap-dependent translation.
Since Δ24/BP1 blocked insulin-stimulated eIF4F assembly, it was anticipated that, in insulin-stimulated cells, it would specifically inhibit cap-dependent translation, while 4E-BP1myc/his would not. To study this, HEK293 cells were cotransfected with a bicistronic vector and the 4E-BP1 constructs. The bicistronic vector contained both a GFP reporter gene downstream from the β-globin gene 5′ untranslated region, to monitor cap-dependent mRNA translation, and a CAT reporter gene downstream from a picornavirus (encephalomyocarditis virus) IRES. The latter monitors cap-independent mRNA translation (Fig. ). Overexpression of either 4E-BP1myc/his or Δ24/BP1 inhibited the basal level of cap-dependent mRNA translation, compared to that in cells transiently transfected with the empty vector (observed as a fivefold decrease in [35
S]methionine incorporation into GFP). Incorporation into GFP increased significantly and by similar amounts after insulin treatment in cells transfected with either the empty vector or 4E-BP1myc/his, demonstrating an increase in cap-dependent mRNA translation in both cases. In contrast, no such increase was observed in cells overexpressing Δ24/BP1 (ratio of 0.2, sixfold less than that observed with 4E-BP1myc/his). The levels of incorporation into CAT for cells expressing 4E-BP1myc/his and those expressing Δ24/BP1 did not differ significantly, suggesting that these two proteins do not interfere with IRES-driven (cap-independent) mRNA translation, as expected, and reported earlier for 4E-BP1 (1
Δ24/BP1 shows reduced phosphorylation upon insulin stimulation.
The data in Fig. suggested that all or part of the N-terminal region removed from Δ4E-BP1 is necessary for the full regulation of phosphorylation of 4E-BP1 in response to insulin. To verify this, we compared the levels of incorporation of radiolabel into 4E-BP1myc/his and Δ24/BP1 in insulin-treated cells. HEK293 cells expressing either 4E-BP1myc/his or Δ24/BP1 were metabolically labeled with [32P]orthophosphate and treated with insulin. Insulin caused a threefold increase in incorporation of the 32P radiolabel into 4E-BP1myc/his (Fig. ). In contrast, insulin failed to increase the extent of incorporation of the 32P radiolabel into Δ24/BP1.
FIG. 3. The cleaved form of 4E-BP1 does not undergo efficient insulin-stimulated phosphorylation. (A) HEK293 cells overexpressing either 4E-BP1myc/his or Δ24/BP1. In vivo radiolabeling was carried out as described in Materials and Methods and 32P label (more ...)
4E-BP1 undergoes phosphorylation on at least five sites (11
) in vivo. To investigate the phosphorylation of individual sites in 4E-BP1myc/his and Δ24/BP1, we employed the available phospho-specific antibodies that detect 4E-BP1 when phosphorylated at Thr36/45, Ser64, or Thr69 (29
). When the same cell extracts used in Fig. were analyzed, the phosphorylation of Ser64 coincided with the release of 4E-BP1myc/his from eIF4E, which became maximal at 1 h (data not shown). This indicates that phosphorylation at this site is important for the release of 4E-BP1 from eIF4E. It is interesting that Ser64 phosphorylation is only observed for the γ-species, which is not able to bind eIF4E. This is consistent with recent data indicating that phosphorylation of Ser64 weakens the binding of 4E-BP1 to eIF4E (20
). Because the expression levels of Δ24/BP1 were much lower than those of 4E-BP1myc/his, the experiment was repeated under altered conditions. A reduced amount of the 4E-BP1myc/his expression vector was used during the transient transfections so that both 4E-BP1myc/his and Δ24/BP1 were expressed in HEK293 cells to similar degrees for better comparison of their phosphorylation (Fig. ). In all cases, there was no detectable phosphorylation of Δ24/BP1 at Thr36 and 45, Ser64, or Thr69, while this was clearly observed for 4E-BP1myc/his. Longer exposure of the Western blot, when using the Thr69 phospho-specific antibody, revealed a trace of Thr69 phosphorylation (data not shown), indicating that the phosphorylation of Δ24/BP1 in response to insulin is substantially, but not completely, impaired.
The region between residues 9 and 16 in 4E-BP1 is required for efficient phosphorylation.
The N terminus of 4E-BP1 thus appears to be essential for the regulation of phosphorylation at Thr36/45, Ser64, and Thr69 in vivo. To identify the region necessary for these phosphorylation events, thus allowing release from eIF4E, two progressively shorter truncations were generated (Δ8/BP1, lacking residues 1 to 8, and Δ16/BP1, lacking residues 1 to 16; Fig. ). The total levels of these overexpressed 4E-BP1 peptides were similar (Fig. ). Δ16/BP1 was phosphorylated less well in response to insulin treatment (Fig. ) than 4E-BP1myc/his and Δ8/BP1, as manifested by the decreased proportion undergoing a mobility shift from the faster-migrating hypophosphorylated isoforms (α) to the slower-migrating hyperphosphorylated isoforms (β and γ). Association of the truncated 4E-BP1 polypeptides with eIF4E was analyzed (Fig. ). In all cases, the overexpressed 4E-BP1 proteins competed with endogenous 4E-BP1 for binding to eIF4E. 4E-BP1myc/his and Δ8/BP1 each dissociated from eIF4E after insulin treatment; in contrast Δ16/BP1 did not, again showing defective regulation.
FIG. 4. Residues between 8 and 16 are required for efficient phosphorylation of 4E-BP1. (A) Diagrammatic representation of mutants 4E-BP1myc/his, Δ8/BP1, and Δ16/BP1. The amino acid sequences indicate the N termini of these truncated proteins. (more ...)
The levels of phosphorylation at Thr36/45, Ser64, and Thr69 (Fig. ) were also compared (Fig. shows a loading control). Phosphorylation of Δ16/BP1 at all sites was markedly decreased compared with that of 4E-BP1myc/his and Δ8/BP1. At longer exposures, a faint band appeared when Δ16/BP1 was probed with the Thr36/45 phospho-specific antibody, suggesting that there was some phosphorylation at either Thr36 or Thr45 after insulin treatment, albeit much less than for 4E-BP1myc/his or Δ8/BP1. Δ16/BP1 was phosphorylated at Thr69 upon insulin stimulation, although to a significantly lesser degree than 4E-BP1myc/his and Δ8/BP1. The partial mobility shift of Δ16/BP1 from the α- to the β-species upon stimulation with insulin (Fig. ) is likely due to the phosphorylation of this protein at Thr69 (Fig. ), as this event causes a shift to the β-form, as previously reported by Mothe-Satney et al. (30
). These data therefore reveal that a region between residues 9 and 16 of 4E-BP1 is required for efficient phosphorylation of 4E-BP1 and therefore for its release from eIF4E (to allow eIF4F complex formation). Considering that the loss of this region of 4E-BP1 does not fully impair the phosphorylation of 4E-BP1 observed upon insulin treatment (especially at Thr69), it appears that other regions within 4E-BP1 still permit some insulin-responsive phosphorylation events.
It has been reported that mTOR phosphorylates 4E-BP1 (at least in vitro) at Thr36 and -45 (see, e.g., reference 29
). It was thus possible that residues 9 to 16 of 4E-BP1 played a role in the efficiency with which mTOR could phosphorylate 4E-BP1. To examine this, 4E-BP1 and Δ16/BP1 were tested as substrates for mTOR (Fig. ). Phosphorylation was inhibited by two agents which block mTOR kinase activity, FKBP12-rapamycin and wortmannin, confirming that the phosphorylation observed was mTOR dependent. mTOR phosphorylated both proteins with similar efficiencies in vitro. Given that phosphorylation of Δ16/BP1 in vivo was greatly impaired, compared to that of 4E-BP1, these data suggest that mTOR may not be the major kinase involved in vivo.
Identification of the RAIP sequence as an important regulatory element in 4E-BP1.
The analysis above strongly implies that residues 9 to 18 are critical for the regulation of 4E-BP1 phosphorylation in vivo. This region is highly conserved between 4E-BP1 and 4E-BP2 homologues from mammalian species (Fig. ). However, human 4E-BP3 has a shorter N-terminal region, which differs markedly from that of 4E-BP1 or 4E-BP2.
FIG. 5. 4E-BP1 R13A and 4E-BP1 P16A mutants show reduced basal and insulin-stimulated phosphorylation, preventing dissociation from eIF4E. (A) N-terminal sequence homology of 4E-BPs. The Thr36 phosphorylation site and truncations Δ8/BP1, Δ16/BP1, (more ...)
To investigate further the potential importance of this region, four point mutations were made within this section of full-length 4E-BP1 (Fig. ). When each of these polypeptides was expressed in HEK293 cells, the total levels of overexpressed 4E-BP1 were similar (Fig. ). The R13A and P16A mutations each caused a significant defect in the basal phosphorylation of 4E-BP1, resulting in a higher proportion of the protein migrating as the least-phosphorylated α-species. The wild-type, P11A, and R18A proteins each showed similar shifts in migration to more highly phosphorylated species upon insulin stimulation, where the β-species is the main form, with a smaller amount resolving as the γ-isoform. In contrast, insulin apparently elicited a smaller increase in phosphorylation of either R13A or P16A mutants than of the wild type, as manifested by an equal distribution between α- and β-isoforms and no appearance of the γ-isoform. To analyze the extent of phosphorylation more thoroughly, the degrees of phosphorylation at Thr36/45, Ser64, and Thr69 were also compared (Fig. ), with Fig. showing the loading control. For the R13A and P16A mutants, the basal phosphorylation of all sites was less than that for the wild type. For Thr69, basal phosphorylation was reduced for the R13A, P16A, and R18A mutants, although insulin increased this to a level comparable to that for the wild type in each case. For Ser64, insulin was only able to induce a small degree of phosphorylation of the R13A mutant and no signal at all was observed for the P16A mutant. The R18A mutant also showed a reduced degree of Ser64 phosphorylation upon insulin stimulation compared to the wild-type 4E-BP1 protein. The abilities of insulin to bring about the release of these 4E-BP1 mutants from eIF4E were also compared (Fig. ). Wild-type 4E-BP1 and the P11A and R18A mutants each substantially dissociated from eIF4E upon insulin stimulation. In contrast, the R13A and P16A point mutants did not.
The above data suggest that the conserved sequence that includes Arg13 and Pro16 is important for normal 4E-BP1 regulation. To explore the role of this RAIP sequence more thoroughly, additional point mutants were made (summarized in Fig. ). The phosphorylation of these mutants was analyzed (Fig. and ). In all cases, mutants containing the Ile15-to-Ala mutation showed the greatest defects in phosphorylation compared to the corresponding mutants where Ile15 was retained. For example, the basal phosphorylation of Thr36/45, Ser64, and Thr69 was almost undetectable in the AAAP, RAAA, and AAAA proteins. Insulin increased the signal with all the phospho-specific antisera for R13A, P16A, and AAIA mutants, while little or no phosphorylation of Thr36/45 or Ser64 was observed in the corresponding point mutants having Ala at position 15 (Fig. ). Insulin caused increased phosphorylation of Thr69 in these proteins, but this was again much weaker than in the mutants where Ile15 was retained (Fig. ). This suggests that Ile15 is one of the critical determinants for regulating 4E-BP1 phosphorylation in vivo.
FIG. 6. Residues within the RAIP motif (13 to 16) are necessary for efficient phosphorylation of 4E-BP1. (A) Diagrammatic representation of 4E-BP1myc/his, R13A, I15A, P16A, AAAP, AAIA, RAAA, and AAAA mutants. (B and C) HEK293 cells were transiently transfected (more ...)
To assess whether the phosphorylation of the N-terminal mutant 4E-BP1 polypeptides was dependent on mTOR, as for wild-type 4E-BP1, HEK293 cells expressing these proteins were pretreated with rapamycin for 30 min prior to being stimulated with insulin. The phosphorylation of truncation mutants Δ8/BP1 and Δ16/BP1, as well as that of the AAIA mutant, was compared to that of the wild-type 4E-BP1myc/his (Fig. ). The AAIA mutant was chosen for this study since it showed only an intermediate defect in insulin-induced phosphorylation, i.e., it still showed some regulation by insulin (Fig. ). The basal and insulin-induced phosphorylation of both 4E-BP1myc/his and Δ8/BP1 was reduced by rapamycin treatment but not completely blocked, i.e., some of the β-species and a signal with the Thr36/45 phospho-specific antibody were still observed. Rapamycin also blocked the phosphorylation of the AAIA mutant induced by insulin (Fig. ). The faint Thr69 signal observed with the Δ16/BP1 and AAIA proteins after insulin stimulation was abolished with rapamycin treatment, showing that this residual phosphorylation, like that of wild-type 4E-BP1, is dependent on mTOR signaling.
The N terminus of 4E-BP1 facilitates insulin-stimulated phosphorylation of 4E-BP3 and its release from eIF4E.
The above data suggest that the RAIP motif is necessary for the mTOR-dependent regulation of 4E-BP1 phosphorylation upon insulin stimulation. 4E-BP3 lacks this motif (Fig. ). We, therefore, asked whether insulin stimulation could also regulate the phosphorylation of 4E-BP3. Following insulin stimulation, 4E-BP3 did not dissociate from eIF4E and eIF4F complex formation did not occur, i.e., no eIF4G1 was copurified with eIF4E (Fig. ). This indicates that the binding of 4E-BP3 to eIF4E is not modulated in response to insulin and that its regulation thus differs from that of 4E-BP1. We tested whether the observed difference between 4E-BP1 and 4E-BP3 might be due to their differing N termini by analyzing two chimeric 4E-BPs. In one chimera (nBP1/cBP3), residues 1 to 19 of 4E-BP3 were replaced by residues 1 to 32 of 4E-BP1. The other chimera (nBP3/cBP1) had the converse replacement. The constructs are summarized in Fig. . Insulin caused a shift in the migration of 4E-BP1 toward the more slowly migrating species, while no significant shift was apparent with either 4E-BP3myc/his or nBP1/cBP3 (Fig. ). Samples of the same cell extracts used in Fig. were subjected to m7GTP-Sepharose chromatography (Fig. ). Insulin stimulation caused 50% of nBP1/cBP3 to dissociate from eIF4E (Fig. ) but did not elicit significant release of nBP3/cBP1. It thus appears that the N-terminal regions of these proteins are indeed important in the regulation of their functions by insulin. In particular, the N terminus of 4E-BP1 is required for insulin to bring about the phosphorylation events necessary for the release of 4E-BPs from eIF4E. In contrast, the N terminus of 4E-BP3 is unable to promote a sufficient degree of phosphorylation of 4E-BP3myc/his and nBP3/cBP1 during insulin stimulation to cause their release from eIF4E (Fig. ). It seemed probable that the anti-Thr36/45 antibody would cross-react with 4E-BP3 at residue Thr23 (which has the same flanking sequence as Thr36 in 4E-BP1; Fig. ). Insulin stimulation increased the signal observed with this antibody for 4E-BP3, showing that it can bring about some degree of phosphorylation of 4E-BP3, albeit not a sufficient degree to cause its release from eIF4E. The basal signal for the nBP1/cBP3 chimera was stronger than that for 4E-BP3myc/his, suggesting a higher basal level of phosphorylation, although this could reflect a higher affinity of this antibody for the former protein.
FIG. 7. Insulin-induced phosphorylation of 4E-BP3 is increased when 4E-BP3 contains the N terminus of 4E-BP1. (A) Diagrammatic representation of mutants used. White and grey boxes, 4E-BP1 and 4E-BP3 proteins, respectively. These mutants each possess C-terminal (more ...)
Due to sequence differences between 4E-BP1 and 4E-BP3, it was considered that the 4E-BP1 Ser64 and Thr69 phospho-specific antibodies would probably not cross-react with 4E-BP3. It is also interesting that neither the insulin-induced phosphorylation of 4E-BP3 at Thr23 nor the phosphorylation of nBP1/cBP3 was sufficient to alter significantly the mobility of these polypeptides during SDS-PAGE (Fig. ). We therefore employed an isoelectric focusing technique to resolve differentially phosphorylated species of 4E-BP3 and nBP1/cBP3 (Fig. ). One main species of 4E-BP3 was apparent when using the anti-Myc antiserum, and no signal was observed with the anti-phospho-Thr36/45 serum. Insulin caused a shift in the migration of 4E-BP3, yielding two more-acidic forms, which both reacted with the phospho-specific antiserum for Thr36/45. The picture for nBP1/cBP3 was quite different, as it showed a much higher basal level of phosphorylation (Fig. ) based on its migration on isoelectric focusing. Furthermore, these species also reacted strongly with the anti-phospho-Thr36/45 antibody. Insulin induced a further shift toward the anode, with the appearance of an additional acidic form probably containing five phosphate groups. This indicates that the N terminus of 4E-BP1 confers a much higher basal level of phosphorylation on 4E-BP3. To examine whether the RAIP sequence element was necessary for the high degree of phosphorylation observed in nBP1/cBP3, we generated an nBP1/cBP3 triple-alanine substitution mutant in which Arg13, Ile15, and Pro16 were replaced [designated nBP1/cBP3(AAAA)]. This nBP1/cBP3(AAAA) mutant showed only a limited degree of basal phosphorylation such that the pattern was similar to that for 4E-BP3myc/his. Its phosphorylation was not increased further upon insulin treatment (Fig. ). This experiment reveals that the RAIP sequence element is indeed required for insulin to elicit the extent of phosphorylation observed for nBP1/cBP3.