Translation Initiation on mRNAs Bound by Nuclear Cap-binding Protein Complex CBP80/20 Requires Interaction between CBP80/20-dependent Translation Initiation Factor and Eukaryotic Translation Initiation Factor 3g*

Background: How the eIF3 complex and ribosomes are recruited during translation on CBP80/20-bound mRNAs remains obscure. Results: CTIF interacts with eIF3g to recruit the eIF3 complex. Conclusion: Translation on CBP80/20-bound mRNAs requires CTIF-eIF3g interaction. Significance: The use of different eIF3 subunits for recruiting eIF3 complex implies that translation on CBP80/20-bound mRNAs differs mechanistically from translation on eIF4E-bound mRNAs. In the cytoplasm of mammalian cells, either cap-binding proteins 80 and 20 (CBP80/20) or eukaryotic translation initiation factor (eIF) 4E can direct the initiation of translation. Although the recruitment of ribosomes to mRNAs during eIF4E-dependent translation (ET) is well characterized, the molecular mechanism for CBP80/20-dependent translation (CT) remains obscure. Here, we show that CBP80/20-dependent translation initiation factor (CTIF), which has been shown to be preferentially involved in CT but not ET, specifically interacts with eIF3g, a component of the eIF3 complex involved in ribosome recruitment. By interacting with eIF3g, CTIF serves as an adaptor protein to bridge the CBP80/20 and the eIF3 complex, leading to efficient ribosome recruitment during CT. Accordingly, down-regulation of CTIF using a small interfering RNA causes a redistribution of CBP80 from polysome fractions to subpolysome fractions, without significant consequence to eIF4E distribution. In addition, down-regulation of eIF3g inhibits the efficiency of nonsense-mediated mRNA decay, which is tightly coupled to CT but not to ET. Moreover, the artificial tethering of CTIF to an intercistronic region of dicistronic mRNA results in translation of the downstream cistron in an eIF3-dependent manner. These findings support the idea that CT mechanistically differs from ET.

mRNAs undergo several modification processes in the nucleus as follows: 5Ј-capping, 3Ј-polyadenylation, and splicing (1)(2)(3). The 5Ј cap structure and 3Ј poly(A) tail are bound by the nuclear cap-binding protein complex, which is a heterodimer of CBP80 and CBP20 (CBP80/20), and poly(A)-binding proteins, respectively (1,3). Properly spliced mRNAs are then exported from the nucleus to the cytoplasm through the nuclear pore complex. During mRNA export, CBP80/20 recruits ribosomes to initiate the first round of translation in what is called the pioneer round of translation. To clearly specify the cap-binding protein used, we will hereafter refer to this process as CBP80/ 20-dependent translation (CT) 4 (4,5). After CT, the cytoplasmic cap-binding protein, eukaryotic translation initiation factor (eIF) 4E, takes over the function of binding the cap (4,6,7). The ribosome is recruited by eIF4E to carry out the on-going process of polypeptide synthesis in what is called steady-state translation, responsible for the bulk protein synthesis. We will hereafter refer to this process as eIF4E-dependent translation (ET) (8,9).
CT and ET are functionally and mechanistically different from each other (4,10). Although ET is the step where large amounts of proteins are produced, CT is generally considered a step where the quality of mRNA is monitored and surveyed. For instance, nonsense-mediated mRNA decay (NMD), in which faulty mRNAs harboring premature termination codons (PTCs) are recognized and down-regulated (11,12), is tightly coupled to CT but not ET (4). In addition, although CT and ET may involve common translation initiation factors (4,10,13,14), they may also use distinct initiation factors to recruit ribo-* This work was supported in part by the Korea Science and Engineering somes, as suggested by our recent identification of CBP80/20dependent translation initiation factor (CTIF) (5). CTIF, which contains a middle domain of eIF4GI (MIF4G), directly interacts with CBP80, is complexed with eIF3 and eIF4AIII, and associates with the active form of the CT initiation complex (5). In addition, depletion of endogenous CTIF from in vitro translation reactions or down-regulation of CTIF using small interfering RNA (siRNA) from cultured cells inhibits the efficiencies of CT and consequently NMD (5). Based on our previous findings, we proposed that CT requires a series of protein interactions of CBP80/20-CTIF-eIF3, analogous to eIF4E-eIF4GI/II-eIF3 interactions for ET (5).
Mammalian eIF3 consists of at least 13 different polypeptides that are designated eIF3a to eIF3m depending on their protein mass (8,9). The eIF3 complex plays multiple roles in translation by (i) stabilizing the 40S ribosomal subunit; (ii) promoting the formation of the 43S pre-initiation complex; (iii) enhancing ribosome recruitment to mRNA via interactions with other eIFs; and (iv) helping ribosome scanning for re-initiation (8,9,15). In particular, eIF3g, one of the core subunits of eIF3 in yeast (8,9) and part of a stable eIF3 subcomplex in mammals (16), interacts with poly(A)-binding protein-interacting protein 1 (Paip1), helping mRNAs circularize (17).
Whereas diverse roles of eIF3 in ET have been characterized in detail as described above, the underlying molecular mechanism by which eIF3 complex is recruited to mRNA during CT remains elusive. Here, we show that CTIF directly interacts with eIF3g and demonstrate that the CTIF-eIF3g interaction is important for the formation of the CT initiation complex and efficient CT.
Protein Expression, Purification, and GST Pulldown Assay-Plasmids expressing GST, GST-eIF3g, or GST-AMSH were transformed into Escherichia coli BL21(DE3)pLysS strain. Isopropyl ␤-D-thio-galactoside (1 mM) was added to the culture to induce GST-fused protein expression, when the absorbance at 600 nm (A 600 ) reached 0.5. The cultures were then incubated for an additional 3 h at 37°C.
Recombinant His-CTIF(365-598) was overexpressed in E. coli BL21(DE3)RIL by the addition of 1 mM isopropyl ␤-Dthio-galactoside when the A 600 value was 0.5. After additional cultivation at 18°C for 24 h, cells were harvested, resuspended in lysis buffer (50 mM Tris-HCl (pH 8.0), 100 mM NaCl, 5% (v/v) glycerol, and 1 mM phenylmethanesulfonyl fluoride (PMSF)), and sonicated. Total cell extracts were loaded onto a Hi-trap nickel nitrilotriacetic acid column (GE Healthcare), and then the column was washed with the same buffer. The bound His-CTIF(365-598) was eluted at ϳ250 mM imidazole with buffer A (50 mM Tris-HCl (pH 8.0) and 5% (v/v) glycerol). Subsequent purification was conducted on a Hi-trap Q FF anion exchange column (GE Healthcare), and the protein was eluted at ϳ350 mM NaCl with buffer A.
Far Western Blotting-Far Western blotting was performed using purified rabbit eIF3 (23,24), recombinant His-CTIF(365-598), and recombinant GST-eIF3g, GST-AMSH, and BSA (New England Biolabs). Briefly, the purified eIF3 was resolved by SDS-PAGE and transferred to a Hybond ECL nitrocellulose FIGURE 1. Human CTIF interacts with eIF3g. A, yeast two-hybrid analysis of human CTIF⌬N and eIF3 subunits. Yeast strain (PBN204) was co-transformed with plasmids expressing BD-CTIF⌬N and AD-eIF3 subunits ( a-k). Transformed yeast cells were spread on selective medium lacking leucine and tryptophan (SD-LW) to select for co-transformants (Master plate). Specific interactions between two proteins were tested as follows: (i) by the appearance of visible blue color by LacZ expression (Filter assay); (ii) growth on selective medium lacking leucine, tryptophan, and uracil (SD-LWU), and (iii) growth on selective medium lacking leucine, tryptophan, and adenine (SD-LWA). The dimerization of polypyrimidine tract-binding protein served as the positive control (ϩ). The empty vector pBCT and pACT2 served as the negative control (Ϫ). B, in vitro GST pulldown assays of CTIF(365-598). Extracts of E. coli expressing GST, GST-eIF3g, and GST-AMSH were mixed with purified recombinant His-CTIF(365-598). After pulldown using GST resin, the precipitated proteins were analyzed by Western blotting (WB) using ␣-GST antibody (upper panel) or ␣-His 6 antibody (lower panel). The locations of markers for molecular weight are indicated on the left. C, far Western blotting analysis (FW) of purified eIF3 complex. Left panel, Coomassie Blue staining results showing the integrities and relative abundances of input proteins. The degraded product of eIF3a is denoted as ⌬eIF3a. Right panel, purified rabbit eIF3 complex was resolved by SDS-PAGE. The purified His-CTIF(365-598) and ␣-His 6 antibody were used as a probe and a primary antibody, respectively. D, far Western blotting analysis using recombinant GST-eIF3g. As in C, except that BSA, recombinant GST-AMSH, and GST-eIF3g were resolved by SDS-PAGE. membrane (Amersham Biosciences). Each membrane was incubated for 24 h at 4°C in blocking buffer (100 mM Tris (pH 7.5), 100 mM potassium acetate, 2 mM magnesium acetate, 0.1 mM EDTA, 10% glycerol, 1 mM PMSF, 1 mM benzamidine, and 0.05% Tween 20) containing 5% skimmed milk. The membranes were then further incubated for 24 h at 4°C in blocking buffer containing 5 g of purified His-CTIF(365-598). The membrane was then analyzed by Western blotting using ␣-His 6 antibody.
Polysome Fractionation-HEK293FT cells (two 150-mm culture dishes) were transiently transfected with the indicated siRNA. Three days after transfection, cytoplasmic extracts of the cells were prepared and then subjected to polysome fractionation as described previously (5).

C-terminal Half of Human CTIF Directly Interacts with eIF3g-
Our previous findings showed that CTIF interacts with CBP80 via its N-terminal half and associates with eIF3, likely via the MIF4G domain in its C-terminal half (5). However, whereas the direct interaction between CBP80 and CTIF has been demonstrated by the GST pulldown assay, the CTIF-eIF3 association was only suggested by immunoprecipitations (IPs) (5). Therefore, we aimed to clearly determine whether CTIF interacts with eIF3 directly or indirectly, and, if directly, which subunit of eIF3 interacts with CTIF. To this end, we first conducted yeast FIGURE 2. CTIF is an essential protein for recruiting eIF3 into CT complex. A, IPs of FLAG-CTIF. HeLa cells were transiently transfected with plasmid, either pcDNA3-FLAG or pcDNA3-FLAG-CTIF. Cell extracts were then either untreated (Ϫ) or treated (ϩ) with RNase A and subjected to IPs using ␣-FLAG-conjugated resin. Western blotting was carried out to detect the indicated proteins (upper panel). Cellular protein ␤-actin served as a negative control. The complete digestions of endogenous RNAs by RNase A treatment were demonstrated by RT-PCR (lower panel). B, IP of FLAG-eIF3g. As in A, except that cells were transfected with either pcDNA3-FLAG or pcDNA3-eIF3g. C, IP of Myc-CBP80. The extracts of HeLa cells that were either undepleted (Control siRNA) or depleted of endogenous CTIF (CTIF siRNA) and were transiently transfected with either pCMV-Myc or pCMV-Myc-CBP80 were subjected to IP using ␣-Myc antibody. D, IP of Myc-eIF4E. As in C, except that cells were transfected with either pCMV-Myc or pCMV-Myc-eIF4E.
two-hybrid analysis using GAL4 DNA-binding domain (BD)fused human CTIF⌬N, which contains the C-terminal half of CTIF, and a GAL4 transcription AD-fused eIF3 subunit (Fig.  1A). Specific interactions between BD-CTIF⌬N and AD-eIF3 would direct the expression of LacZ, Ura3, and Ade2 in yeast cells. The results showed that yeast cells expressing both BD-CTIF⌬N and AD-eIF3g exhibited the blue color in the filter assay and growth on selective media (Fig. 1A). Although we could not observe the expression of each AD-fused eIF3 subunit in yeast by Western blotting because of weak promoter (data not shown), our results suggest that CTIF⌬N and eIF3g specifically interact in vivo.
The interaction between CTIF and eIF3g was further demonstrated by two independent approaches: in vitro GST pulldown assay (Fig. 1B) and far Western blotting (Fig. 1, C and D). The results of the GST pulldown assay showed that GST-eIF3g, but neither GST only nor GST-AMSH, which served as a negative control, interacted with purified recombinant His-CTIF(365-598) (Fig. 1B), thus indicating the direct interaction between CTIF(365-598) and eIF3g. In addition, the results of far Western blotting using purified rabbit eIF3 complex and, as a probe, purified recombinant His-CTIF(365-598) showed that His-CTIF(365-598) specifically reacted with a single protein that corresponded to eIF3g or eIF3h in size (Fig. 1C, right  panel). The amount and integrity of purified eIF3 and BSA, which served as a negative control, were determined by Coomassie Blue staining (Fig. 1C, left panel). As observed in other reports (23,24), the largest eIF3 subunit, eIF3a, was proteolyzed during the purification of the eIF3 complex from rabbit reticulocyte lysate. Furthermore, the results of far Western blotting using BSA, recombinant GST-AMSH, GST-eIF3g, and, as a probe, recombinant His-CTIF(365-598) showed that His-CTIF(365-598) specifically reacted with GST-eIF3g (Fig. 1D). Altogether, these results clearly indicate that CTIF(365-598) directly interacts with eIF3g.
CTIF Serves as an Adaptor Protein to Link CBP80 and eIF3-Based on the CTIF-eIF3g interaction, we next aimed at characterizing the CTIF-or eIF3g-containing protein complex. To this end, we performed IPs using extracts of cells expressing FLAG-CTIF ( Fig. 2A) or FLAG-eIF3g (Fig. 2B). Previous studies showed that although the CT complex contains CBP80/20, CTIF, and eIF3, the ET complex contains eIF4E, eIF4GI/II, and eIF3 (5). The IP results revealed that the CBP80 and the eIF3 subunits, eIF3a, eIF3b, eIF3c, and eIF3g, but not eIF4E, co-immunopurified with FLAG-CTIF in an RNase A-resistant manner ( Fig. 2A, upper panel), suggesting that CTIF and eIF3g coexist in the CT complex. The sufficient removal of endogenous mRNA by RNase A treatment was demonstrated by measuring the level of endogenous GAPDH mRNA ( Fig. 2A, lower panel). The reciprocal IP to immunopurify FLAG-eIF3g revealed that all tested translation factors co-immunopurified with FLAG-eIF3g in an RNase A-resistant manner (Fig. 2B). Because both CT and ET use eIF3 as a common factor to recruit ribosomes (1,4,5,14), it is reasonable that eIF3g associates with both CT and ET initiation complexes.
Based on a previous report that the N-terminal half of CTIF interacts with CBP80 (5) and our findings in this study that (i) the C-terminal half of CTIF interacts with eIF3g ( Fig. 1) and (ii) the CTIF-immunoprecipitated complex contains CBP80 and eIF3 ( Fig. 2A), we speculated that CTIF may function to link CBP80 and eIF3 during CT. Thus, the down-regulation of CTIF would abolish an interaction between CBP80 and eIF3, without affecting the association between eIF4E and eIF3. To test this hypothesis, we performed IPs using extracts of cells expressing either Myc-CBP80 or Myc-eIF4E and depleted of endogenous CTIF by siRNA. The results showed that CTIF, eIF3a, eIF3b, and eIF3c, but not eIF4GI, co-immunopurified with Myc-CBP80 upon treatment with a nonspecific "Control" siRNA (Fig. 2C). However, a reduced or undetectable amount of coimmunoprecipitated eIF3 subunits was observed upon CTIF down-regulation (Fig. 2C). On the contrary, eIF4GI, eIF3a, eIF3b, and eIF3c co-immunopurified with Myc-eIF4E in a CTIF-independent manner (Fig. 2D). All these results strongly support the idea that CTIF serves as an adaptor protein to link CBP80 and eIF3 within the CT complex.

CTIF-eIF3g Interaction Is Required for Ribosome Recruitment
tions was assessed using polysome fractionation (Fig. 3). Because CTIF down-regulation abolishes the association between CBP80 and eIF3 (Fig. 2C), CTIF down-regulation may lead to the inhibition of CT, consequently reducing the level of polysome-associated CBP80. To test this hypothesis, HEK293FT cells were either treated with CTIF siRNA to known down endogenous CTIF or treated with a nonspecific control siRNA. The cytoplasmic extracts of cells were separated by polysome fractionation, and the relative distribution of endogenous CBP80 and eIF4E was analyzed by Western blotting (Fig. 3).
The results showed that the level of endogenous CTIF was down-regulated to 15% of normal (Fig. 3, A and B, compare the intensity of CTIF protein). Intriguingly, although both CBP80 and eIF4E were detected in most fractions upon treatment with a nonspecific control siRNA, CBP80 but not eIF4E was redistributed from the polysome fractions to subpolysome fractions upon CTIF down-regulation (Fig. 3, A and B). The intensities of endogenous CBP80 and eIF4E in each fraction were quantitated, and the relative distributions of endogenous CBP80 (Fig.  3C) and endogenous eIF4E (Fig. 3D) in each fraction are shown graphically. Therefore, taking our findings in Fig. 2 into account, it is most likely that CTIF recruits ribosomes into the CT complex via its interaction with eIF3.
Down-regulation of Endogenous eIF3g Abolishes NMD of PTC-containing Globin mRNA and GPx1 mRNA-It is well established that NMD of PTC-containing mRNAs takes place during CT but not ET (4,10). Because recruitment of the eIF3 complex via the CTIF-eIF3g interaction is important for CT (Figs. 2 and 3), an efficient NMD might require CTIF-mediated eIF3 recruitment. We therefore analyzed the change of NMD efficiency upon down-regulation of endogenous eIF3 subunits. To this end, HeLa cells were transfected with siRNA against one of the endogenous eIF3 subunits, siRNA against Upf1, a key factor for NMD (11,12), or a nonspecific Control siRNA. Two days later, cells were transiently re-transfected with plasmids expressing the following: (i) ␤-globin (Gl) mRNA, either normal (Norm) or PTC-containing at the 39th amino acid position (Ter); (ii) glutathione peroxidase 1 (GPx1) either normal (Norm) or PTC-containing at the 46th amino acid position (Ter); (iii) mouse urinary protein (MUP) mRNA, which served to control for variations of transfection and RNA harvest; and
Under the same conditions, we also analyzed the overall translation efficiencies by measuring the ratio of RLuc activity to RLuc mRNA. It is generally considered that overall translation efficiency mirrors the ET efficiency, as CT efficiency is very weak by comparison (1,4,5). The results showed that, although down-regulation of eIF3b or Upf1 had no significant influence on the abundance of RLuc mRNA, down-regulation of eIF3c and eIF3g showed an increase of the level of RLuc mRNAs by ϳ3and 1.5-fold, respectively, for unknown reason(s). The ratio of RLuc activity to RLuc mRNA, overall translation efficiency, was not affected by the down-regulation of eIF3b, eIF3c, and eIF3g. However, down-regulation of Upf1 slightly increased the level of overall translation of RLuc mRNAs, consistent with the previous findings (27). All these results suggest that, under our conditions, transient and temporal down-regulation of eIF3 subunits by siRNA is not sufficient for inhibition of overall translation. On the contrary, the level of down-regulation of eIF3 subunits in this study was sufficient for inhibition of CT and consequently NMD, which suggests that CT is more sensitive to the level or activity of eIF3 than ET.
Artificially Tethered CTIF to an Intercistronic Region Directs the Translation of a Downstream Cistron in an eIF3-dependent Way-The results in Fig. 4 suggest that CT is inhibited by the down-regulation of eIF3 subunits under the conditions where ET was minimally affected. However, we cannot completely rule out the possibility of an indirect effect by eIF3 down-regulation. Therefore to clearly address the direct role of CTIF-eIF3 interaction in CT, we employed a tethering assay using the N/BoxB derived from bacteriophage (28). Indeed, it has previously been shown that when the middle region of eIF4GI is tethered to the intercistronic region of a dicistronic mRNA it can recruit ribosomes to activate the translation of the downstream cistron (29).
HEK293T cells depleted of eIF3b, eIF3c, or eIF3g were transiently co-transfected with tethering dicistronic reporter plasmid expressing dicistronic F/BoxB/R mRNAs (Fig. 5A) and effector plasmid expressing either N-EGFP or N-EGFP-CTIF. Western blotting results showed that the levels of endogenous eIF3b, eIF3c, and eIF3g were reduced to 39, 41, and 36% of normal, respectively, and that comparable levels of N-EGFP and N-EGFP-CTIF were expressed (Fig. 5B). RT-PCRs using specific oligonucleotides and [␣-32 P]dATP revealed that tethering of N-EGFP-CTIF increased the F/BoxB/R mRNAs in abundance by 1.7-o 2.4-fold (Fig. 5C) in an eIF3-independent manner, suggesting that CTIF may contribute to another step of post-transcriptional gene regulation. More intriguingly, tethering of N-EGFP-CTIF increased the relative RLuc activity (the ratio of RLuc activity to firefly luciferase activity) by 4-fold, compared with tethering of N-EGFP. The increase of the relative RLuc activity was abolished by down-regulation of eIF3 subunits (Fig. 5D). Considering that eIF3 down-regulation had a marginal influence on ET efficiency under our conditions (Fig. 4), all these results suggest that a tethered CTIF recruits ribosomes via its interaction with eIF3, consequently triggering translation of the downstream cistron.

DISCUSSION
Here, we propose that CTIF serves as an adaptor protein to bridge CBP80/20 at the 5Ј-end of mRNA and eIF3 during CT. More specifically, the N-terminal half and C-terminal half of CTIF directly interact with CBP80 (5) and eIF3g (in this study), respectively. The CTIF-eIF3g interaction mediates the entire (or functional) eIF3 complex recruitment, which in turn leads to the ribosome loading onto mRNAs for efficient CT (Fig. 6). To support this idea, recent studies on the assembly of the eIF3 complex (16,30) showed that eIF3g is not part of a reconstituted functional core of mammalian eIF3. Indeed, eIF3g is located on the periphery of the eIF3 complex and is not involved in interactions among three stable subcomplexes of eIF3. Therefore, during CT, eIF3g might provide a docking platform for the CTIF-eIF3 interaction.
Although we could not observe any significant interaction between CBP80 and eIF4GI/II in this study (Fig. 2C) and in a previous report (5), we cannot rule out the possibility that certain environmental changes or stresses may trigger the formation of an alternative complex, in which eIF4GI/II replaces CTIF during CT (Fig. 6), as proposed by others (4,13).
After CT, CBP80/20 is replaced by eIF4E (1, 4, 6, 7). Subsequently, eIF4E interacts with eIF4GI, which in turn recruits eIF3 via a direct interaction between eIF4GI and eIF3e, another eIF3 subunit (31). The use of a different eIF3 subunit for recruiting a functional eIF3 complex during CT and ET via CTIF-eIF3g interaction and eIF4GI/II-eIF3e interaction, respectively, may contribute to the differences in the regulation, efficiency, and biological roles of CT and ET. This view is supported by the fact that ET is able to respond quickly to changes in physiological conditions or to cellular stresses, whereas CT is relatively resistant to these changes (32)(33)(34)(35). Further studies will be required to fully dissect the mechanistic differences between CT and ET.