Disruption of the Interaction of Mammalian Protein Synthesis Eukaryotic Initiation Factor 4B with the Poly(A)-binding Protein by Caspase- and Viral Protease-mediated Cleavages*

Eukaryotic initiation factor (eIF) 4B interacts with several components of the initiation pathway and is targeted for cleavage during apoptosis. In a cell-free system, cleavage of eIF4B by caspase-3 coincides with a general inhibition of protein synthetic activity. Affinity chromatography demonstrates that mammalian eIF4B interacts with the poly(A)-binding protein and that a region consisting of the N-terminal 80 amino acids of eIF4B is both necessary and sufficient for such binding. This interaction is lost when eIF4B is cleaved by caspase-3, which removes the N-terminal 45 amino acids. Similarly, the association of eIF4B with the poly(A)-binding proteinin vivo is reduced when cells are induced to undergo apoptosis. Cleavage of the poly(A)-binding protein itself, using human rhinovirus 3C protease, also eliminates the interaction with eIF4B. Thus, disruption of the association between mammalian eIF4B and the poly(A)-binding protein can occur during both apoptosis and picornaviral infection and is likely to contribute to the inhibition of translation observed under these conditions.

In plants, eIF4B can interact with PABP, and this leads to enhancement both of the RNA binding activity of PABP (42) and of the RNA helicase activity of the [eIF4A⅐eIF4B] complex (43). It has remained unclear whether such an interaction occurs in mammalian cells and, if so, whether it might be affected by the cleavage of eIF4B that occurs in apoptotic cells. PABP itself can also be a target for proteolytic cleavage, notably in picornavirus-infected cells, in which inhibition of host protein synthesis and the selective translation of viral mRNAs correlate with the cleavage of both eIF4GII and PABP (30, 44 -48). The cleavage of PABP by the picornavirus-encoded 2A or 3C proteases separates a large N-terminal fragment from a C-terminal homodimerization domain (47,48), and this may contribute to the inhibition of protein synthesis.
Here, we provide evidence for a direct interaction between mammalian eIF4B and PABP. We demonstrate that recombinant eIF4B can bind to PABP via the N terminus of eIF4B and that this interaction is ablated when eIF4B is cleaved by caspase-3 in vitro or when cells are induced to undergo apoptosis. Furthermore, the cleavage of PABP with human rhinovirus (HRV) 3C protease eliminates the ability of PABP to bind to eIF4B. These data suggest that in addition to the cleavage of eIF4G that occurs during apoptosis and picornavirus infection, the association between mammalian eIF4B and PABP is targeted under these conditions by the cleavage of eIF4B and PABP, respectively. forms of full-length and truncated eIF4B, 4E-BP1, and the C-terminal region of eIF4GI were expressed in BL21 cells and purified on nickel-NTA-agarose (Qiagen) (49). GST-PABP was expressed in baculovirusinfected Sf9 insect cells and purified on glutathione-Sepharose (Amersham Pharmacia Biotech), as per the manufacturer's instructions. Purified, recombinant proteins were dialyzed against Buffer A (20 mM MOPS-K ϩ , pH 7.2, 25 mM NaCl, 10 mM KCl, 7 mM 2-mercaptoethanol, 2 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride) and stored in aliquots at Ϫ70°C. The baculovirus vector encoding untagged Xenopus PABP was provided by Dr. D. Schoenberg (Ohio State University). Seventy-two hours after baculovirus infection, insect cell extracts were prepared by cell lysis with 1% (v/v) Igepal and stored in aliquots at Ϫ70°C (34).
Immunoblotting-Antisera to eIF4B and PABP were gifts from Drs. J. W. B. Hershey and D. Schoenberg, respectively. Antiserum specific for the C-terminal domain of eIF4G was as described previously (36). Proteins were resolved by SDS-PAGE (50,51), transferred to polyvinylidene difluoride membranes, and visualized by immunoblotting with the antisera described, using alkaline-phosphatase-coupled secondary antibodies.
Protein Cleavage in Vitro-Reticulocyte lysate was incubated for the indicated times at 30°C with purified caspase-3 (TCS Biologicals, Buckingham, UK) (5 or 10 g/ml, as indicated), in the absence or presence of 50 M zVAD⅐FMK (Alexis, UK). Lysates were also incubated with 200 units/ml recombinant HRV protease 3C at 5°C for 2 h prior to translation assays. Purified His-eIF4B was incubated in phosphate-buffered saline containing 2 mM benzamidine and 0.5% (v/v) Triton X-100 (PBS-TB) with 0.5 g/ml purified caspase-3 for 2 h at 37°C. Untagged PABP or GST-PABP was incubated with 60 units/ml HRV protease 3C in PBS-TB for 30 min at 37°C.
In Vitro Translation Assays-Protein synthesis assays used nonnuclease-treated reticulocyte lysates essentially as described in Ref. 52. Incubations were carried out in the absence or presence of 25 g/ml poly(A) ϩ or poly(A) Ϫ mRNA encoding luciferase, which was generated using the Ambion Message Machine system as per the manufacturer's instructions. Samples were subsequently resolved by SDS-PAGE. Translation of the luciferase reporter mRNA was quantified by analysis on a PhosphorImager (Molecular Dynamics Storm 860) using Image-Quant software.
Co-isolation of Recombinant Proteins-Following incubation together in PBS-TB, His-eIF4B or GST-PABP and their associated proteins were recovered on nickel-NTA-agarose or glutathione-Sepharose, respectively. All resins were washed five times each (with 1 ml of PBS-TB), and the recovered proteins were eluted in SDS-PAGE sample buffer prior to analysis by SDS-PAGE and immunoblotting.
Induction of Apoptosis and Analysis of Cell Extracts-Jurkat cells were incubated in the absence or presence of 250 ng/ml anti-Fas antiserum for 2 h at 37°C. Extracts from control and apoptotic cells were prepared as described previously (33).

RESULTS
eIF4B Cleavage Coincides with Inhibition of Protein Synthesis by Caspase-3-Initiation factor eIF4B is cleaved in apoptotic cells to yield a smaller fragment (⌬eIF4B). In addition, other factors, including eIF4GI, eIF4GII, eIF3p35, and eIF2␣ are cleaved under the same conditions (31)(32)(33)(34)(35)(36)(37)(38)(39)(40). However, the relative contribution of each of these events to the overall down-regulation of protein synthesis during apoptosis is not yet known. Using a cell-free system, we have investigated the time at which cleavage of eIF4B occurs relative to that of other caspase targets in the protein synthetic machinery upon addition of purified caspase-3. Fig. 1A (right panel) shows that in the reticulocyte lysate system, eIF4B is cleaved within 10 min of caspase-3 treatment, whereas eIF4G remains intact (as judged by the appearance of the specific cleavage products p120 and M-FAG (35)) until at least 20 min of incubation. At similarly early times following caspase-3-treatment, eIF3p35 and eIF2␣ are also not significantly cleaved (data not shown). The early cleavage of eIF4B coincides with the start of caspaseinduced loss of protein synthetic activity, which affects both endogenous globin translation (  (Fig. 1B, right panel), as is the inhibition of protein synthesis (Fig. 1, A and B).
eIF4B Interacts with PABP in a Caspase-sensitive Manner-Cleavage of eIF4B by caspase-3 removes an N-terminal frag-ment of 45 amino acids (36). As eIF4B interacts with PABP in plants to modulate the activity of the cap-binding complex (42,43), we have examined, using affinity chromatography, whether eIF4B and PABP interact directly in the mammalian system and whether caspase-3 cleavage of eIF4B affects this binding. Fig. 2A shows that whereas mammalian His-eIF4B did not itself bind to glutathione-Sepharose (lane 3), there was substantial recovery of the factor on these beads following incubation with GST-tagged PABP (lane 5). Conversely, PABP could be isolated from a reticulocyte lysate by chromatography on nickel-agarose in the presence but not in the absence of His-eIF4B that was bound to the beads (Fig. 2B). An irrelevant His-tagged protein (the eIF4E-binding protein 4E-BP1) did not bind PABP in this system, ruling out the possibility of nonspecific interactions between PABP and nickel-agarose or the His tag. Similarly, purified His-eIF4B associated with GST-PABP could be recovered with the latter protein on glutathione-Sepharose, whereas another His-tagged protein (the C-terminal part of eIF4GI (49)) did not bind under the same conditions (Fig. 2C). This eliminates the possibility of nonspecific interactions between the His tag and PABP, GST, or glutathione-Sepharose itself. The interaction between eIF4B and PABP was not prevented by incubation with RNase A (Fig. 2A, lane 6), suggesting that it is not mediated by RNA bridging between the two proteins. However, we cannot entirely discount a role for a fragment of RNA that is protected from RNase action by one or both proteins.
To delineate the region of eIF4B that interacts with PABP, we incubated truncated versions of His-eIF4B with untagged PABP. Fig. 3 shows that whereas PABP alone was not recovered on nickel-agarose (lane 1), the N-terminal region of eIF4B (amino acids 1-150) permitted the co-isolation of PABP (lane 5). Further delineation of this region showed that amino acids 81-180 of eIF4B did not interact with PABP (lane 3), but amino acids 1-80 (lane 4) were sufficient to allow the interaction.
To determine whether the caspase-mediated truncation of eIF4B at the N terminus disrupts the interaction of the factor with PABP, we incubated full-length and caspase-cleaved forms of eIF4B with GST-PABP. Protein complexes were subsequently isolated on glutathione-Sepharose. caspase-3, is sufficient to ablate the interaction with PABP.
PABP Interacts with eIF4B in a Protease-sensitive Manner-Mammalian PABP is cleaved during picornavirus infection by the virally encoded proteases 2A and 3C, and the sites of cleavage have been mapped to near the C terminus (47,48). We have used HRV protease 3C to determine whether such cleav-age interferes with the ability to interact with eIF4B. Protease 3C was able to utilize both GST-PABP (Fig. 4B, lane 3) and untagged PABP (Fig. 4C) as substrates but did not degrade eIF4B (Fig. 4C). Following cleavage of GST-PABP, there was a reduction in the co-isolation of full-length eIF4B with PABP (Fig. 4B, lane 3 versus lane 2). Protease 3C cleavage also prevented the association of untagged PABP with the Histagged N terminus of eIF4B (Fig. 4C, lane 5 versus lane 4). These data suggest that the highly conserved C-terminal domain of PABP (53) is required for the interaction with the N-terminal region of eIF4B.
We have investigated the consequences for protein synthesis of the cleavage of PABP in the reticulocyte lysate. In contrast to the reported effect of foot-and-mouth disease virus protease 3C in vivo (54), HRV protease 3C, at the concentration and times used here, did not degrade eIF4G or eIF4A in vitro but caused cleavage of ϳ50% of the PABP in the lysate (Fig. 5A). Consequently, it was possible to examine the effect of this enzyme specifically on PABP cleavage and protein synthesis in parallel. In a non-nuclease-treated lysate, to which was added either poly(A) ϩ or poly(A) Ϫ luciferase mRNA, HRV protease 3C had a small inhibitory effect on the translation of both the endogenous globin mRNA (data not shown) and the exogenous poly(A) ϩ RNA (Fig. 5B, top panel). In contrast, the protease treatment resulted in a substantial increase in the ability to translate the poly(A) Ϫ luciferase mRNA (Fig. 5B, bottom panel). These data suggest that either the PABP cleavage product shows a gain of function toward translation of poly(A) Ϫ mRNA or, more likely, the latter acquires a translational advantage in competition with the endogenous globin mRNA. These results also indicate that exposure to HRV protease 3C does not impair the function of other initiation factors required for the translation of both poly(A) ϩ and poly(A) Ϫ mRNAs.
The Interaction between PABP and eIF4B Is Diminished in Apoptotic Cells-A prediction that can be made on the basis of our results is that in cells induced to undergo apoptosis, where eIF4B is cleaved (36), the association of PABP with eIF4B eIF4B Interaction with PABP should be disrupted. To address this, extracts were prepared from control and apoptotic Jurkat cells. The integrity of eIF4G, PABP, eIF4B, and eIF4E and the association of these proteins with each other were monitored by immunoblotting. In agreement with published data (31,36), Fig. 6A shows that eIF4G  (lane 2 versus lane 1) and eIF4B (lane 4 versus lane 3), but not PABP or eIF4E, were cleaved during apoptosis. PABP and associated proteins were isolated by affinity chromatography on poly(A)-Sepharose (Fig. 6B). Although similar levels of PABP were recovered from apoptotic and control cell extracts, the amount of eIF4B associated with PABP was substantially reduced in apoptotic extracts (lane 2 versus lane 1). In addition, the amount of eIF4E recovered was greatly diminished (lane 2 versus lane 1), as would be predicted from the cleavage of eIF4G, which results in the separation of the eIF4E binding site from the PABP binding site (35). We have also used FIG. 6. Induction of apoptosis reduces the association of eIF4B with poly(A)-interacting proteins in vivo. A, Jurkat cells were incubated for 2 h in the absence (control (C)) (lanes 1 and 3) or presence (Fas) (lanes 2 and 4) of an agonistic antibody to the Fas receptor, and extracts were prepared. Equal amounts of protein were analyzed by immunoblotting for eIF4G (and its cleavage product M-FAG), PABP, eIF4B/⌬eIF4B, and eIF4E, as indicated.  (22,29). For PABP, the four RNA recognition motifs and the proline-rich domain are shown (22,24,47,48). The region of eIF4B required for PABP binding contains a sequence (residues 6 -20) with similarity to amino acid sequences 133-145 and 135-147 within the PABP-binding domains of human eIF4GI and eIF4GII, respectively (17,60). Homologous regions in the three sequences are boxed, with identical residues indicated in italics. B, disruption of the initiation complex by proteolytic cleavages of initiation factors. The known interactions between eIF3, eIF4A, eIF4B, eIF4E, eIF4G, and PABP are shown on the left, and the sites at which these factors are cleaved by caspase-3 and by the picornavirus-encoded 2A, 3C, and L proteases are indicated. In apoptotic cells, caspase-mediated cleavages result in the appearance of three fragments of eIF4G and a truncated form of eIF4B (top right). The cleavages separate the N-terminal region of eIF4G that binds PABP from the central region that binds eIF4E, eIF4A, and the [eIF3⅐eIF4B] complex (4,35) and also disrupt the interaction of eIF4B with PABP. In cells infected with poliovirus, HRV, or foot-and-mouth disease virus, viral protease-mediated cleavage results in the appearance of two fragments of eIF4G and a truncated form of PABP (bottom right). The fragmentation of eIF4G again separates the regions that bind PABP and the [eIF3⅐eIF4B] complex (9,63,64). In both apoptotic and infected cells the links between the 5Ј end of an mRNA (via recognition of the cap by eIF4E) and the 3Ј end (via recognition of the poly(A) tail by PABP) or between the mRNA and the ribosome (via binding of eIF3 to the latter) are broken, potentially reducing the efficiency with which reinitiation of protein synthesis on capped poly(A) ϩ mRNA can take place.
eIF4B Interaction with PABP m 7 GTP-Sepharose chromatography to investigate the association of PABP with eIF4F (Fig. 6C). Following induction of apoptosis, there was a decrease in the level of PABP recovered with eIF4E (lane 2 versus lane 1). However, in this instance, it is not possible to discern whether this effect is attributable to the loss of integrity of eIF4B or to the caspase-mediated cleavage of eIF4G (31,35,37), because both these proteins bind PABP independently. DISCUSSION The direct interaction between mammalian eIF4B and PABP, in conjunction with the binding of eIF4G to PABP, may facilitate the functional association of the 5Ј and 3Ј ends of mRNA (1)(2)(3)(55)(56)(57). In wheat germ, the association of eIF4B with PABP increases the efficiency of reinitiation of protein synthesis (58), a process that is also enhanced by the interaction between eIF4G and PABP (17,43). As depicted in Fig. 7A, the loss of the N terminus of eIF4B (by caspase-mediated cleavage) or of the C terminus of PABP (by picornavirus protease-mediated cleavage) disrupts complex formation between the two proteins. Such an effect would be predicted to inhibit reinitiation. Indeed, cleavages of eIF4B ( Fig. 1 and Ref. 36) and PABP ( Fig. 5 and Refs. 47 and 48) are associated with inhibition of protein synthesis in vivo and in vitro.
We have previously pointed out (36) that ⌬eIF4B still contains the DRYG domain required for self-association and binding to eIF3p170 (29) and also retains the RNA recognition motif domain required for binding RNA (22,59). As the region of eIF4B required for the stimulation of the RNA helicase activity of eIF4A is located in the C-terminal half of the protein (24,29), this activity also may not be directly affected. In contrast, interaction of the extreme N terminus of eIF4B with PABP provides a potential mechanism by which cleavage of eIF4B during apoptosis may contribute to the down-regulation of protein synthesis (4). Because PABP also enhances the RNA helicase activity of the [eIF4A⅐eIF4B⅐eIFiso4F] complex in wheat germ (43), such a regulatory mechanism would also be impaired when the eIF4B-PABP interaction is abolished, resulting in less efficient unwinding of mRNA secondary structure.
The site of binding of PABP on eIF4GI has been assigned to amino acids 132-160 (17), and there is strong conservation of this sequence in eIF4GII (60). Interestingly, a region with significant similarity to parts of these sequences is present near the N terminus of eIF4B (Fig. 7A), and our data are consistent with the possibility that this site is critical for eIF4B-PABP interaction. Further deletion mapping and sitedirected mutagenesis studies will be required to test this.
Inhibition of the eIF4B-PABP interaction during infection, as a consequence of the cleavage of PABP by viral proteases (Fig. 4), may be additive or synergistic with that of the cleavages of eIF4GI and eIF4GII in contributing to the shut-off of translation in infected cells (44 -48). Only a partial cleavage of PABP in vitro is sufficient to enhance the translation of nonpolyadenylated mRNA relative to that of mRNA with a poly(A) tail, perhaps because the latter is translated less efficiently under these conditions and competes less well.
As illustrated in Fig. 7B, disruption of several protein-protein interactions, as a result of cleavages of initiation factors, is a feature of both the early stages of apoptosis and of infection with picornaviruses. The patterns of the cleavages are different in the two situations, but a degree of evolutionary convergence has apparently occurred between the mechanisms by which protein synthesis is targeted for down-regulation. Interestingly, both poliovirus proteases 2A and 3C can themselves induce apoptosis (61,62), adding weight to the view that initiation factor modifications may have important functional significance for the development of an apoptotic response.