A Novel Role of the Mammalian GSPT/eRF3 Associating with Poly(A)-binding Protein in Cap/Poly(A)-dependent Translation*

The mammalian GSPT, which consists of amino-termi-nal (N) and carboxyl-terminal (C) domains, functions as the eukaryotic releasing factor 3 (eRF3) by interacting with eRF1 in translation termination. This function requires only the C-domain that is homologous to the elongation factor (EF) 1 (cid:1) , while the N-domain interacts with polyadenylate-binding protein (PABP), which binds the poly(A) tail of mRNA and associates with the eukaryotic initiation factor (eIF) 4G. Here we describe a novel role of GSPT in translation. We first determined an amino acid sequence required for the PABP interaction in the N-domain. Inhibition of this interaction significantly attenuated translation of capped/poly(A)-tailed mRNA not only in an in vitro translation system but also in living cells. There was a PABP-dependent linkage between the termination factor complex eRF1-GSPT and the initiation factor eIF4G associating with 5 (cid:1) cap through eIF4E. Although the inhibition of the GSPT-PABP interaction did not affect the de novo formation of an 80 S ribosomal initiation complex, it appears to suppress the subsequent recycle of ribosome. These results indicate that GSPT/eRF3 plays an important role in translation cycle through the multicloning site of pUC18. synthetic oligonucleotides (cid:2) (cid:2) (cid:2) (cid:2) Production of Recombinant Proteins— Proteins were induced by the addition of 0.1 m M isopropyl-1-thio- (cid:2) - D -galactopyranoside at 37 °C for 3 h in Escherichia coli JM109. The cells were resuspended in buffer A consisting of 50 m M Tris-HCl (pH 8.0), 1 m M EDTA, 150 m M NaCl, 1% Nonidet P-40, 2 (cid:3) g/ml aprotinin, 100 (cid:3) M phenylmethylsulfonyl fluoride, and 2 (cid:3) g/ml of leupeptin. After incubation with 1 mg/ml lysozyme at 4 °C for 30 min, the cell lysate was sonicated for 3 min on ice. The supernatant after centrifugation at 100,000 (cid:3) g for 60 min was subjected to glutathione-Sepharose 4B (Amersham Biosciences) and/or Ni- NTA-agarose (Qiagen). If necessary, GST was removed using PreScis-sion TM Protease (Amersham Biosciences). The purified proteins were dialyzed against buffer B consisting of 20 m M Tris-HCl (pH 8.0), 50 m M NaCl, and 1% Nonidet P-40.

The process of eukaryotic protein biosynthesis is divided into three steps: initiation, elongation, and termination. Among them, termination had been the least investigated aspect. However, the identification of two releasing factors, eRF1 1 (1) and eRF3 (2), provided a breakthrough in understanding the termination process. eRF1 recognizes all stop codons to release the completed polypeptide chain from the ribosome (1), and eRF3 is essential for the GTP-dependent releasing activity (2).
Mammalian eRF3 gene, GSPT, was first isolated based on its ability to complement a temperature-sensitive gst1 mutant of Saccharomyces cerevisiae (3). At present, two distinct eRF3 genes termed GSPT1 and GSPT2 have been identified (4) and mapped to chromosomal band 16p13.1 (5) and Xp11.23 to p11.21 (6), respectively, in the human genome. The structural analysis revealed that both subtypes consist of an N-terminal region (ϳ200 amino acids) and a C-terminal EF1␣-like GTPbinding domain (428 amino acids). The C-domain, which interacts with eRF1, is sufficient not only for the termination reaction (2,4,7,8) but also for the compensation of yeast gst1growth arrest (3). In contrast, the N-domain is not required for the eRF1 binding and the termination reaction. We previously reported that the N-domain associates with PABP and inhibits its multimerization (9,10).
PABP has two major functions: mRNA stabilization (11)(12)(13)(14) and translation enhancement (15)(16)(17)(18)(19)(20). PABP prevents mRNA from deadenylation, a late-limiting step of mRNA decay, by binding to the poly(A) tail. On the other hand, the involvement of PABP in translation enhancement is based on the finding that efficient translation requires the synergistic interplay between the 5Ј cap and 3Ј end-poly(A) tail of mRNA. The 5Ј cap and 3Ј poly(A) tail are recognized by eIF4E and PABP, respectively, and eIF4G mediates the association between them. These interactions result in the formation of a circularized mRNA (21)(22)(23), and this suggests the hypothetical machinery of efficient translation; ribosome after translation termination is recruited to the next cycle of translation initiation. However, it is noteworthy that PABP was also reported to stimulate translation of capped, nonpolyadenylated mRNA (24).
In this study, we analyzed the biological significance of the interaction between GSPT and PABP in the several steps of translation reaction. Inhibition of this interaction significantly attenuated translation of capped/poly(A)-tailed mRNA. There was a PABP-dependent linkage between eRF1-GSPT and the 5Ј cap-initiation factor complex, and this linkage appeared to be responsible for the reentry of ribosome to the initiation factor complex. Thus, GSPT/eRF3 plays important roles not only in translation termination with eRF1 but also in the translation cycle through its interaction with PABP.

EXPERIMENTAL PROCEDURES
Plasmids-For production of N-terminally GST-fused GSPT2 mutants, PCR products were inserted to pGEX4T1 (Amersham Biosciences). To produce full-length GSPT2 and its deletion mutants fused with N-terminally GST and C-terminally His 6 epitope, respectively, the SalI-NotI fragment of pGEX6P1 vector (Amersham Biosciences) was ligated with the synthetic adaptor HIS (HIS5Ј plus HIS3Ј as described blow) to make pGPH6. PCR products encoding the corresponding sequences (amino acid sequence 1-632, 1-204, 58 -204, and 80 -204) of GSPT2 were inserted to pGPH6, resulting in pGP2-full, pGP2-1, pGP2-58, and pGP2-80. To produce the amino acid 45-204 of eIF4G I fused * This work was supported in part by research grants from the "Research for the Future" Program of the Japan Society for the Promotion of Science (JSPS-RFTF 96L00505) and the Scientific Research Funds of the Ministry of Education, Culture, Sports, Science, and Technology of the Japanese Government. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Production of Recombinant Proteins-Proteins were induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside at 37°C for 3 h in Escherichia coli JM109. The cells were resuspended in buffer A consisting of 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, 2 g/ml aprotinin, 100 M phenylmethylsulfonyl fluoride, and 2 g/ml of leupeptin. After incubation with 1 mg/ml lysozyme at 4°C for 30 min, the cell lysate was sonicated for 3 min on ice. The supernatant after centrifugation at 100,000 ϫ g for 60 min was subjected to glutathione-Sepharose 4B (Amersham Biosciences) and/or Ni-NTA-agarose (Qiagen). If necessary, GST was removed using PreScission TM Protease (Amersham Biosciences). The purified proteins were dialyzed against buffer B consisting of 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, and 1% Nonidet P-40. PABP I was purified as described previously (10).
Cell Culture, DNA Transfection, and in Vivo Translation Assay-COS-7 and HeLa cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum and maintained at 37°C in 5% CO 2 . Transfections were performed with Lipofectin (Invitrogen).
HeLa cells that had been transfected with pcDNA3/GSPT2 mutants and a reporter pUC18-T7-R-luc-HCV IRES-F-luc were incubated for 40 h and infected with vaccinia virus vTF-3 (25) for 4 h. Dual luciferase activities were measured using Stop & Glo luciferase assay system (Promega).
In Vitro Binding Assay-Recombinant GST-fused proteins were incubated with glutathione-Sepharose 4B for 30 min at 4°C. After removal of the unbound fraction, the resin was mixed with recombinant PABP in buffer B and further incubated at 4°C for 30 min. The resin was washed with buffer B and incubated with synthetic peptides or recombinant proteins at 4°C for 60 min. After washing with buffer B, proteins were eluted from the resin with SDS-polyacrylamide sample buffer and subjected to SDS-PAGE and immunoblot analysis.
Immunoprecipitation and Ni-NTA Pull-down Assay-The transfected cells were lysed in buffer C consisting of 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, 1% Nonidet P-40, 1 mM dithiothreitol, 10 g/ml boiled RNase A, 100 M phenylmethylsulfonyl fluoride, 2 g/ml of aprotinin, and 2 g/ml of leupeptin. After centrifugation at 15,000 ϫ g for 20 min, the lysate was incubated at 4°C for 30 min with anti-FLAG IgGagarose (Sigma) or Ni-NTA-agarose, and then the resin was washed with buffer C. As the need arose, recombinant proteins or synthetic peptides were added, and the resin was further incubated at 4°C for 60 min. After washing with buffer C, proteins retained in the resin were subjected to SDS-PAGE and immunoblot analysis. Immunoprecipitation from nuclease-treated rabbit reticulocyte lysate (RRL, Promega) was performed in the same manner as described above using an anti-GSPT polyclonal antibody and protein A-agarose 4B (Amersham Biosciences).
In Vitro Translation Assay-Luciferase mRNAs containing poly(A) tail or not were synthesized with T7 RNA polymerase after linearization of pGL3:T7-pA with BamHI or XbaI, respectively. When capped mRNAs were synthesized, m 7 GpppG (Stratagene) was used. In vitro translation reaction was performed as described below. Nuclease-treated RRL (10 l) was reconstituted with 10 l of a buffer consisting of 10 mM Hepes-KOH (pH 7.5), 142 mM KCl, 1.32 mM MgCl 2 , 0.1 mM EDTA, 7 mM ␤-mercaptoethanol, 20 M each of complete amino acid mixture (Promega), 1.6 units/l RNasin (Promega), 5 g/ml of luciferase mRNA, and the indicated amounts of synthetic peptides or recombinant proteins and further incubated at 30°C for 60 min or for the indicated times. Luciferase activity was measured using Bright-Glo luciferase assay regent (Promega).
Assay for the Formation of an 80 S Ribosomal Initiation Complex-Globin mRNA (Invitrogen) was 3Ј-32 P-labeled using T4 RNA ligase and 5Ј-32 P-labeled pCp. For the formation of a de novo 80 S ribosomal initiation complex, the nuclease-treated RRL (26 l) was reconstituted with 14 l of a buffer consisting of 6.5 mM Hepes-KOH (pH 7.5), 65 mM KCl, 0.65 mM MgCl 2 , 65 M EDTA, 4.5 mM ␤-mercaptoethanol, 28 M each complete amino acid mixture, 2.3 units/l RNasin, 25 ng of 3Ј-32 Plabeled globin mRNA, and 0.14 mM cycloheximide. The reaction mixture was incubated at 30°C for 15 min, and aliquots (20 l) were analyzed on 5 ml of 15-30% (w/v) linear sucrose gradient. After centrifugation at 160,000 ϫ g for 45 min, fractions were collected using a piston-gradient fractionator (Biocomp), and the radioactivity of each fraction was measured by a liquid scintillation counter.

Interaction between GSPT and PABP in Intact Cells-We
have previously shown that the N-domain of GSPT interacts with PABP in a yeast two-hybrid assay and in vitro binding experiments (10). To investigate the significance of the interaction in intact cells, N-terminally FLAG-tagged GSPT2 was expressed in COS-7 cells, and the cell extract was subjected to immunoprecipitation using an anti-FLAG antibody. Endogenous PABP co-precipitated with FLAG-GSPT2 (Fig. 1A, lane 2). Co-immunoprecipitation of PABP was also observed if GSPT1 was expressed in the cells (data not shown). To check if this interaction is dependent on the N-domain rather than the C-domain of GSPT, either of the two domains was produced in COS-7 cells. PABP co-immunoprecipitated with the N-domain (lane 3), whereas eRF1 co-precipitated with the C-domain (lane 4). To investigate whether these three factors are in the same complex, an immunoprecipitation experiment was performed using extracts from COS-7 cells expressing N-terminally FLAG-tagged eRF1. Endogenous GSPT and PABP co-immunoprecipitated with FLAG-eRF1 (Fig. 1B). These interactions appear to be independent of RNA tethering because cell extracts had been treated with RNase. Thus, these experiments show that GSPT associates with PABP and eRF1 via its N-domain and C-domain, respectively, in living cells, and consequently GSPT mediates the association between eRF1 and PABP.
Identification of the Site Critical for PABP Binding in the N-domain of GSPT-To identify a PABP-binding sequence in the N-domain of GSPT, a co-precipitation assay was performed using COS-7 cells expressing deletion mutants of the N-domain ( Fig. 2A). As shown in Fig. 2B, deletion mutants starting from amino acid positions 1, 19, 36, and 58 interacted with PABP (lanes 7-11), while a mutant starting from the 80th amino acid (aa 80 -204) failed to associate with PABP (lane 12). Since the expression of this mutant (aa 80 -204) was rather low in COS-7 cells, we performed a binding assay using recombinant proteins. GSPT2/aa 58 -204 associated with PABP, but aa 80 -204 did not (Fig. 2C). Furthermore, GSPT2/aa 1-79 but not aa 78 -141 was sufficient for the PABP binding (Fig. 2D). Thus, the amino acid sequence 58 -79 of GSPT2 (see Fig. 2A) was identified as a critical region for PABP binding. This sequence is conserved well between GSPT1 and GSPT2 in mice and humans (4).
The significance of the identified sequence was further in-vestigated using a synthetic peptide corresponding to amino acid 58 -75 (Fig. 3A). As shown in Fig. 3, B and C, the GSPT-PABP interaction was progressively inhibited with increasing amounts of the whole N-domain or the synthetic peptide but not with GST or a control peptide consisting of the same amino acid composition in a scrambled order (see Fig. 3A). The complete inhibition of the interaction by the synthetic peptide supports the notion that the sequence aa 58 -75 of GSPT2 constitutes a critical site for the PABP-binding. However, since the half-maximum inhibition by the synthetic peptide was observed at about 100 M, which is almost three orders of magnitude higher than that of the whole N-domain (Fig. 3C), we cannot exclude the possibility that other regions might also be involved in the interaction.

Involvement of the Interaction between the N-domain of GSPT and PABP in Cap/Poly(A)-dependent Translation-
PABP was reported to regulate translation in a cap/poly(A)-dependent manner by mediating the interaction between the cap-binding complex eIF4F and the poly(A) tail of mRNA (21)(22)(23). To investigate whether the interaction between GSPT and PABP is involved in cap/poly(A)-dependent translation, we utilized nuclease-treated RRL as a cell-free translation system. It was previously reported that the synergistic stimulation by cap and poly(A) was observed in the RRL system with partial removal of ribosome and the associated initiation factors (26,27). However, such a synergistic stimulation was observed only by changing the concentrations of MgCl 2 and KCl (Fig. 4A). In this system, cap-dependent translation was markedly stimulated by the simultaneous presence of the poly(A) tail, while mRNA containing only the poly(A) tail had little activity. The N-domain of GSPT2 fused to GST markedly inhibited the cap/ poly(A)-dependent translation (Fig. 4B, closed triangles). Interestingly, the N-domain was also capable of inhibiting translation of capped and non-poly(A)-tailed mRNA (closed circles). In contrast, GST alone had little effect on any of the mRNAs (open circles and triangles). The effect of the synthetic peptides was also investigated. The synthetic peptide aa 58 -75 inhibited translation not only of capped/poly(A)-tailed mRNA but also of capped mRNA (Fig. 4C). In accordance with the results in Fig.  3, the concentration of the synthetic peptide required for translation inhibition was much higher than that of the whole N-domain of GSPT2. Thus, both the GSPT-PABP interaction and the cap/poly(A)-dependent translation are inhibited by the synthetic peptide aa 58 -75 and the whole N-domain in a similar concentration-dependent manner. In addition, the results presented here suggest that the interaction between GSPT and PABP may also be involved in poly(A)-independent translation (Fig. 4, B and C). Although the exact mechanism is still unclear, it is noteworthy that PABP was reported to stimulate cap/poly(A)-dependent and poly(A)-independent translation by distinct mechanisms (24).
Previous studies reported that the C-domain of GSPT is sufficient for translation termination (2,4,7,8). However, it is possible that the translation inhibition observed here might be the result of inhibition of translation termination or elongation. Therefore, we investigated whether the inhibition of the GSPT-PABP interaction may affect translation termination or elongation by using an uncapped/non-poly(A)-tailed mRNA. If the termination step is inhibited, luminescence would diminish because luciferase has no activity when it is not released from ribosome (28), and if the elongation step is inhibited, luminescence would also diminish. As shown in Fig. 4D, the N-domain of GSPT2 had no inhibitory effect on the cap/poly(A)-independent translation, which is in sharp contrast to the results obtained with capped/poly(A)-tailed mRNA. These results suggest that the GSPT-PABP interaction is not involved in translation termination or elongation.

No Involvement of Paip1 in the Inhibition of Cap/Poly(A)-dependent Translation by the N-domain of GSPT-In addition to
GSPT, two other proteins that interact with PABP have been reported. One is Paip1 identified as a translation activator (29), and the other is Paip2 identified as a translation repressor (19,20). The PABP-binding sites, which were recently reported in Paip1 and Paip2 (19,20,30), are similar to that of GSPT2 identified in this study. Thus, it is possible that the translation inhibition by the N-domain of GSPT2 may have resulted from the inhibition of the Paip1-PABP interaction. To examine this possibility, we used RRL immunodepleted of Paip1 by anti-Paip1 antibodies. As shown in Fig. 5A, Paip1 was completely depleted, but PABP and GSPT were little affected. Under these conditions, significant effect was not observed in cap/poly(A) synergy (Fig. 5B), and the N-domain of GSPT2 had still inhibitory effect on translation (Fig. 5C) the N-domain of GSPT2 on translation by monitoring the synthesis of R-luc and F-luc from the bicistronic construct T7-Rluc-HCV IRES-F-luc (Fig. 6A). We used HCV IRES because it functions in eIF4G-(31) and a poly(A) tail-(20, 32) independent manners. By means of this bicistronic mRNA, efficiencies of cap/poly(A)-dependent translation (R-luc activity) and HCV IRES-dependent translation (F-luc activity as an internal control for both transfection efficiency and the amount of the reporter mRNA) can be measured at the same time. The overexpression of the N-domain of GSPT2 caused a marked decrease in the ratio of R-luc/F-luc (Fig. 6B), indicating that cap/poly(A)-dependent translation was inhibited by the N-domain. Moreover, the mutant lacking the amino acids 1-57 (⌬1-57), which can interact with PABP (Fig. 2), still has the inhibitory effect. Such inhibition was, however, reduced in the mutant lacking the amino acids 1-79 (⌬1-79), which cannot associate with PABP (Fig. 2B). To confirm this result, we constructed a mutant of aa 1-204 whose amino acids 65-71 are all converted to alanine (Ala-65-Ala71). This mutant could not interact with PABP any more (Fig. 6C) and had a lesser inhibitory effect on cap/poly(A)-dependent translation than the original N-domain (Fig. 6D). Taken together, these results further substantiate the idea that the GSPT-PABP interaction is involved in cap/poly(A)-dependent translation in living cells. However, the results that both of the mutants, ⌬1-79 and Ala-65-Ala71, still exhibited just a little inhibitory activity are consistent with the idea that besides the region 58 -79, the N-domain (aa 1-204) of GSPT2 contains an additional sequence(s) responsible for PABP binding as suggested by the results in Figs. 3 and 4.
PABP Mediates the Interaction between GSPT and eIF4G-Since the GSPT-PABP interaction is involved in cap/poly(A)dependent translation, we examined the possibility that GSPT could associate with the translation initiation factor. To this end, an immunoprecipitation assay was performed against nuclease-treated RRL. As shown in Fig. 7A, PABP and eIF4G were co-immunoprecipitated with GSPT by anti-GSPT antibodies. This complex was also detected when cell extracts from COS-7 cells were used instead of RRL (data not shown). To substantiate these findings, we examined the interaction using recombinant proteins. PABP and the N-terminally GST-fused eIF4G/aa 45-204, which is sufficient for the PABP binding (23), were mixed with GSPT2-(His) 6 and subjected to a glutathione-Sepharose pull-down assay. As shown in Fig. 7B, the interaction between GST-eIF4G/aa 45-204 and PABP was observed both in the presence and absence of GSPT2 (lanes 4 and 5). However, the association between eIF4G/aa 45-204 and GSPT2 was observed only in the presence of PABP (compare lane 5 with 3), indicating that PABP mediates the association. These results provide a possibility that GSPT may be involved in a translation initiation step, in addition to termination.
The Interaction between GSPT and PABP Is Involved in the Multiple Rounds of Translation but Not in the de Novo Formation of an 80 S Ribosomal Initiation Complex-We next examined the involvement of the GSPT-PABP interaction in translation initiation. The final output of the translation initiation process was examined by monitoring the formation of an 80 S ribosomal initiation complex. 3Ј-32 P-labeled globin mRNA was incubated with a nuclease-treated RRL in the presence of cycloheximide, and the complex formation was monitored by a sucrose-density gradient analysis. As shown in Fig. 8A, labeled mRNAs were shifted at a position corresponding to the 80 S ribosomal initiation complex. The complex formation was, however, little affected by the addition of the N-domain of GSPT2, suggesting that the GSPT-PABP interaction does not function in the de novo formation of an 80 S initiation complex.
To further elucidate the role of GSPT-PABP interaction in translation reaction, we next analyzed the effect of the N-domain of GSPT2 on the kinetics of luciferase production in RRL. Regardless of the presence of the N-domain, production of luciferase was observed after an absolute lag time of about 8 min (Fig. 8), which was also not affected by the increasing amount of mRNA or preincubation of the reaction mixture before the addition of mRNA (data not shown). Since luciferase becomes active after its release from ribosome (28), the lag  [3][4][5] or GST (lane 2) was immobilized to glutathione-Sepharose and incubated with the purified PABP and/or GSPT2-(His) 6 . Proteins that associated with the resin (lanes 2-5) were resolved by SDS-PAGE and immunoblotted with anti-His (upper), anti-PABP (middle), and anti-GST (lower) antibodies. Lane 1 shows the purified GSPT2-(His) 6 and PABP (marked by asterisks) used in the pull-down assay. means time required for completion of the first round of translation. Thus, consistent with the results in Fig. 8A, the first round indexed by the lag time was not affected by the N-domain of GSPT2. In contrast, the production of luciferase after the time lag, which is indicative of the subsequent rounds of translation, was markedly inhibited by the addition of the N-domain. These results indicate that the interaction between GSPT and PABP functions in the translation cycle, possibly the recycle of ribosome to the initiation factor complex rather than the initial formation of 80S complex. DISCUSSION GSPT Interacts with PABP through a Site in Its N-domain -We previously presented evidence that GSPT interacts with PABP in in vitro experiments (9). This conclusion was confirmed and further extended by our present experiments. First, the interaction between the N-domain of GSPT and PABP was observed with cell extracts (Figs. 1, 2B, and 7A) and with purified proteins (Figs. 2C and 7B). Moreover, we identified a possible PABP-binding sequence in the N-domain (Figs. 2 and  3). The GSPT2-PABP interaction is mediated at least through the amino acid sequence aa 58 -75 of GSPT2, since the synthetic peptide completely inhibited the association (Fig. 3).
In addition to GSPT, Paip1 and Paip2 have been reported to interact with PABP. PABP-binding sites in Paip1 and Paip2 are similar to the sequence aa 58 -75 of GSPT2, and this motif is important for their interactions with the C-terminal domain of PABP (19,20,30). Thus, GSPT may compete with Paips for PABP binding. However, the relationship between GSPT and Paips may not be so simple, since Paips interact with both the N-and C-terminal regions of PABP (19,20,29,30,33). In contrast, GSPT interacts only with the C-terminal site (10). A rabbit reticulocyte lysate, which we used in this study, has a much smaller amount of Paips than a rabbit liver lysate when compared with the amount of PABP (data not shown). Thus, it is possible that Paips may be the factors modifying the function of PABP on the requirement of each tissue.
A Novel Role of GSPT/eRF3 in the Eukaryotic Translation System-It is generally believed that the function of GSPT was solely to facilitate the release of completed peptide chains from ribosome as a GTP-dependent stimulator of eRF1. However, the present study reveals that GSPT associates with eIF4G through PABP (Fig. 7) and that the GSPT-PABP interaction is involved in the multiple rounds of translation (Figs. 4, 6, and 8). The synergistic enhancement of translation by cap and poly(A) is explained by the circularization of mRNA, which is mediated through a complex consisting of poly(A)-PABP-eIF4F-cap (21)(22)(23). This fact suggests the hypothetical model that a translation-terminating ribosome may be recruited to the next translation initiation. However, this idea is unsatisfactory since translation is terminated at stop codons that are not always close to the poly(A) tail of mRNA. Therefore, some factors are likely to mediate the physical coupling between the terminating ribosome on the stop codon and the poly(A) tail. The fact that GSPT interacts with eRF1 and PABP at the same time (Fig. 1B) suggests that GSPT may be the bridging protein to connect the stop codon with the poly(A) tail. In this hypothesis, a 3Ј-untranslated region, which locates between a stop codon and a poly(A) tail, could be looped out, and the terminating ribosome could be passed to the 5Ј cap structure through the novel protein bridge consisting of eRF1, GSPT, PABP, and eIF4F (Fig. 9).
In addition to the role of PABP in translation, several lines of evidence suggest that PABP might affect translation in a poly(A)-independent manner (18,20,34,35). Furthermore, this function appears to be independent of its binding to eIF4G (24). The results presented here suggest that the GSPT-PABP interaction may also be involved in poly(A)-independent translation (Fig. 4, B and C), though the exact mechanism is still unclear.
It is well established that PABP has another function; it prevents mRNA degradation by protecting the poly(A) tail. In general, mRNA degradation, an important aspect of gene expression, is a strictly regulated process that is often linked to translation (12,36,37), and translation-dependent deadenylation is an important step of this mechanism in which PABP is probably involved. Moreover, several reports show that GSPT is involved in nonsense-mediated decay, a mechanism by which mRNAs containing a premature termination codon are rapidly degraded (38,39). These mechanisms are not well understood, but it is conceivable that they are linked to the translation  termination. Further studies on GSPT/eRF3 should be important for the understandings of not only translation machinery but also mRNA-decay mechanism.