Phosphorylation of eIF4GII and 4E-BP1 in response to nocodazole treatment

Abstract

Translation mechanisms at different stages of the cell cycle have been studied for many years, resulting in the dogma that translation rates are slowed during mitosis, with cap-independent translation mechanisms favored to give expression of key regulatory proteins. However, such cell culture studies involve synchronization using harsh methods, which may in themselves stress cells and affect protein synthesis rates. One such commonly used chemical is the microtubule de-polymerization agent, nocodazole, which arrests cells in mitosis and has been used to demonstrate that translation rates are strongly reduced (down to 30% of that of asynchronous cells). Using synchronized HeLa cells released from a double thymidine block (G₁/S boundary) or the Cdk1 inhibitor, RO3306 (G₂/M boundary), we have systematically re-addressed this dogma. Using FACS analysis and pulse labeling of proteins with labeled methionine, we now show that translation rates do not slow as cells enter mitosis. This study is complemented by studies employing confocal microscopy, which show enrichment of translation initiation factors at the microtubule organizing centers, mitotic spindle, and midbody structure during the final steps of cytokinesis, suggesting that translation is maintained during mitosis. Furthermore, we show that inhibition of translation in response to extended times of exposure to nocodazole reflects increased eIF2α phosphorylation, disaggregation of polysomes, and hyperphosphorylation of selected initiation factors, including novel Cdk1-dependent N-terminal phosphorylation of eIF4GII. Our work suggests that effects on translation in nocodazole-arrested cells might be related to those of the treatment used to synchronize cells rather than cell cycle status.

Introduction

Protein synthesis is performed in 3 stages (initiation, elongation, and termination), with the initiation stage of translation generally accepted as a major site of regulation of gene expression.^1-6 This pivotal role reflects the regulated binding of mRNA to the ribosome, facilitated by the assembly of eIFs (eukaryotic initiation factors) into a multi-protein complex known as eIF4F (composed of eIF4E, eIF4A, and eIF4G), which associates with the 7-methylguanylate (m⁷G) cap structure at the 5′ end of the mRNA. eIF4G is expressed as 2 paralogs in mammalian cells, eIF4GI and eIF4GII. Although initial reports suggested eIF4GI and eIF4GII were interchangeable,⁷ overexpression of the eIF4GII paralog in an eIF4GI-knockdown background was unable to restore translation to the same extent as eIF4GI.⁸

In turn, the activity of the eIF4F complex is regulated by both phosphorylation and the inherent structural properties of the recruited mRNA.^1,4,6 The formation of the eIF4F complex reflects the regulated availability of eIF4E to participate in initiation, a process controlled by a number of general and mRNA-specific regulatory proteins. Using a conserved motif, 4E-BPs (eIF4E-binding proteins) compete with the eIF4Gs for a common surface on eIF4E and inhibit eIF4F assembly. The regulated association of 4E-BPs with eIF4E is acutely modulated by multi-site phosphorylation dependent upon mammalian target of rapamycin complex 1 (mTORC1) signaling,^2,5 effectively integrating signals from mitogens and nutrients with the translational apparatus.^2,3 In addition to modification of the eIF4 family of proteins, phosphorylation of the α-subunit of the canonical initiation factor eIF-2 has been shown to be increased at G₂/M.⁹ This phosphorylation has been shown to have a permissive effect on the efficiency of cap-independent initiation of translation at 2 cell cycle-dependent internal ribosome entry sites (IRESs), thereby mediating specific mRNA translation at this phase of the cell cycle.⁹

Work over a number of years has suggested that the rate of protein synthesis varies throughout the cell cycle.^10-12 In HeLa cells arrested in mitosis using colcemid, cap-dependent, but not IRES-mediated, translation initiation was impaired along with a dephosphorylation of eIF4E.¹³ In nocodazole-arrested HeLa cells, 4E-BP1 was shown to be in a more hypophosphorylated state, interacting strongly with eIF4E, decreasing eIF4F complex levels.¹¹ The same study showed that eIF4GII was hyperphosphorylated during nocodazole treatment of cells, which correlated with a decrease in its binding to eIF4E. In contrast, in HeLa cells synchronized and released from a sequential thymidine and aphidicolin block, phosphorylation of 4E-BP1 was increased at Thr 70 by the cyclin-dependent kinase Cdk1, with phosphorylation of this site being permissive for Ser 65 phosphorylation.¹⁴

Using synchronized HeLa cells, incubated with nocodazole, the Cdk1 inhibitor RO3306, or released from a double thymidine block, we have systematically re-addressed whether rates of protein synthesis are changed during mitosis, and how this correlates with initiation factor phosphorylation. The work described here indicates that exposure of cells to nocodazole promotes increased eIF2α phosphorylation and disaggregation of polysomes, characteristics of an inhibition of initiation. However, this was not observed with cells released from a G₁/S block using the double-thymidine approach. Furthermore, we have teased apart some of the phosphorylation events that occur in response to nocodazole, and present here a reappraisal of how initiation factor complexes are affected in response to this treatment.

Results

Translation rates are maintained throughout the cell cycle, and initiation factors are associated with components of the tubulin cytoskeleton during mitosis

Double thymidine block was used to synchronize exponentially growing HeLa cells at the G₁/S phase border. Cells were then released from this block and the nuclear content analyzed by propidium staining and flow cytometry analysis, showing that a cycle was completed within 20 h (Fig. 1A). The majority of mitosis was observed to occur between 8 and 11 h post-release. These results are in agreement with previous work.¹⁵ In parallel experiments, the incorporation of [³⁵S] methionine into precipitable protein was measured for the 30 min prior to harvest in order to gauge whether there was a decrease in protein synthesis rate in cells going through M phase. While translation rates did fluctuate during the cell cycle, there was no decrease observed at any of the time points where the majority of cells were undergoing mitosis (Fig. 1B). In comparison, in cells that were synchronized with thymidine, released, then incubated with nocodazole, there was a robust decrease in translation to around ~30% of that of asynchronous cells, in agreement with previous work.¹⁶

Figure 1. Protein synthesis rates are not decreased during M phase when observing cycling synchronized cells. (A) Exponentially growing HeLa cells were maintained as an asynchronous population or synchronized at the G₁/S boundary by double thymidine block and released for the number of hours indicated. Nuclei content of cells at each time point was determined by staining with propidium iodide and FACS analysis. (B) In parallel experiments, translation rates were examined by pulsing cells with [³⁵S] methionine for 30 min prior to harvest. Extracts were prepared, and the incorporation of radioactive methionine into protein (cpm/μg total protein) determined as described in the text. For the T/N sample, HeLa cells were blocked once in thymidine before being released then synchronized in M phase by incubation with the microtubule depolymerizing agent nocodazole for 19 h. (C) Equal amounts of extract from (B) were subjected to SDS-PAGE and proteins transferred to PVDF. The membrane was then probed with the antibodies shown and visualized with secondary antibodies conjugated to HRP.

Immunoblotting of cell extracts showed that the levels of the cell cycle markers Cyclin B1 and Survivin increased and decreased as predicted throughout the time course (Fig. 1C). When determining the influence of the cell cycle stages on members of the translation initiation factor machinery, there were no changes in levels of components of the eIF4F complex (eIF4GI, eIF4A, and eIF4E), and no increased phosphorylation of eIF2α was observed. There was an upward shift in mobility of 4E-BP1 in cells in early G₁ and immediately post-M phase (12–14 h post release), suggesting an increase in phosphorylation. According to models of translational control, these results all point to a generally favorable environment for translation to proceed.

The results from Figure 1B and C indicate that, rather than there being a reduction in protein synthesis associated with passage into mitosis, translation is maintained at a relatively consistent level throughout the cell cycle. We and others have previously shown that a subpopulation of components of the initiation factor machinery are localized to the nucleus.¹⁷ We therefore embarked on confocal microscopy experiments to examine how these proteins localize in cells in which the nuclear membrane undergoes breakdown,¹⁸ thus removing the physical barrier between the cytoplasm and nucleoplasm. We examined synchronized cells that were proceeding through mitosis (Fig. 2), staining cells for a component of the translation machinery (eIF4GI, PABP, eIF4E, 4E-BP1, or ribosomal protein S6) alongside α-tubulin and DNA, the definitive morphologies of which could be used to define the various characteristic stages. At each stage co-localization of each component was observed with the tubulin cytoskeleton, implying that there may be some ongoing translation localized to this important component of the cell. This perhaps aids in the synthesis of further tubulin or other components (e.g., chromosomal passenger proteins) that are required for efficient mitosis.

Figure 2. The translation machinery co-stains with the microtubule network during different stages of mitosis and cytokinesis. Cells released from a double thymidine block for 9 h were fixed with 4% (w/v) paraformaldehyde and permeabilized with 0.1% (v/v) Triton X-100. Cells in different stages of mitosis and cytokinesis were examined by confocal microscopy, as determined by the positioning of microtubules or DNA (A–E). Upper panels show components of the translation machinery which were detected with rabbit polyclonal antibodies, as indicated, followed by an anti-rabbit secondary antibody conjugated to Alexa Fluor 555 (red). Lower panels show this signal merged with α-tubulin detected with a mouse monoclonal antibody conjugated to FITC (green) and DNA detected with DAPI (blue).

Synchronization in M phase with nocodazole inhibits translation rates and leads to robust phosphorylation of key translation factors

Given that our work showed that there was unlikely to be any inhibition of translation in cells transiting through mitosis, we returned to those which had undergone treatment with nocodazole, to determine the underlying mechanisms of the observed reduction in protein synthesis (Fig. 1B). FACS analysis showed the expected peak of cells arrested in G₂/M when compared with cells maintained in an asynchronous state (Fig. 3A), and sucrose density gradients show that nocodazole treatment leads to a reduction in polysomes, a further confirmation of the inhibition of protein synthesis (Fig. 3B). Extracts from asynchronous cells or those having undergone thymidine/nocodazole synchronization showed that there was a strong phosphorylation of the α subunit of eIF2 on Ser 51 (Fig. 3C), confirming the work of others.¹⁶ This post-translational modification inhibits the eIF2B guanine nucleotide exchange factor, which prevents the recycling of eIF2 from a GDP to a GTP bound state, thus preventing ternary complex formation (eIF2·Met-tRNAi·GTP), thus inhibiting further rounds of translation initiation. Other modifications of translation initiation factors were observed, with a retarded migration observed of eIF4GI, eIF4GII and of some of the 4E-BP1 population (Fig. 3C). The maintenance of a proportion of 4E-BP1 in a hypophosphorylated (termed “α”) state shows another situation which would be likely to inhibit translation. This phospho form of the protein is able to bind to and sequester eIF4E and inhibit the interaction of this cap-binding protein with the eIF4G proteins, which act as molecular “scaffolds” onto which a number of other factors bind. However, we also observed that around 70% of the 4E-BP1 migrated higher than the standard phospho forms of the proteins, reflecting modification over and above those already established. This modification was not prevented by preincubation of cells with the mTOR inhibitor RAD001 (Fig. S1). We used our low-percentage SDS-PAGE method¹⁹ to enable delineation of eIF4G isoforms that arise through alternative translation initiation from either different AUGs (in the case of eIF4GI^20,21) or a CUG and AUG (eIF4GII⁸) and therefore differ at their N termini. The shift in migration was observed in all the isoforms, (Fig. 3C) suggesting that no one eIF4G isoform is subjected to differential regulation. Both these factors, and 4E-BP1 are known phosphoproteins, so we next analyzed whether the migratory changes were due to phosphorylation.

Figure 3. Synchronization with nocodazole inhibits translation and leads to substantial phosphorylation of components of the translation initiation machinery. (A) HeLa cells were either maintained asynchronously, or blocked once in thymidine before being released, then synchronized in M phase by incubation with nocodazole for 19 h. As before, FACS analysis was used to determine nuclear DNA content. (B) Samples were subjected to sucrose density gradient centrifugation to determine the effect of incubation with nocodazole on polysome formation. Translation rates from these cells are shown as part of Figure 1 (B), which confirm the large inhibition of translation, as determined by the reduction in [³⁵S] methionine incorporation. (C) Equal amounts of cell extracts from parallel samples were analyzed by immunoblotting for the proteins indicated. To separate isoforms of eIF4G proteins that arise from alternative translation initiation, low-percentage SDS-PAGE was used. (D) To determine whether shifts in migration of eIF4GI/II and 4E-BP1 were due to phosphorylation, extracts from asynchronous cells or cells which had been incubated with nocodazole for 19 h were prepared. These were then incubated in the presence or absence of lambda protein phosphatase for 30 min before the treated extracts were separated by SDS-PAGE and immunoblotted for the indicated proteins.

Extracts from asynchronous or nocodazole-treated cells were prepared and then treated with bacteriophage lambda protein phosphatase (Fig. 3D). In all cases, the nocodazole-dependent changes in migration or phosphorylation were either greatly reduced or completely abrogated, suggesting that phosphorylation is the post-translational modification underlying these observations. In the case of eIF4GII, it was also observed that there is an underlying phosphorylation in asynchronously maintained cells, which is likely to be due to phosphorylation at alternative sites to those modified in this treatment.²²

Robust phosphorylation of eIF2α and 4E-BP1 does not occur following release from G₂/M block or with reduced incubation with nocodazole

To investigate the nocodazole-dependent phosphorylation of initiation factors in more detail, we employed an alternative mechanism to arrest cells at the G₂/M border. Phosphorylation of key substrates by the Cdk1/cyclin B complex acts as a trigger for the commitment to mitosis, and so the reversible cell-permeable small-molecule Cdk1 inhibitor RO3306²³ was employed, as it acts in an independent manner to those like nocodazole, which act on microtubule dynamics. In our experience, we have struggled to release cells from the nocodazole-induced arrest, but this issue is not the case when using RO3306. Therefore, following removal of this inhibitor from the cell medium, the cell cycle status was assayed by FACS in comparison to asynchronous cells, or those arrested by thymidine nocodazole block (Fig. 4A), and translation rates measured as before (Fig. 4B), albeit with a shorter 15 min pulse of [³⁵S] methionine prior to harvest. While we did observe some inhibition of protein synthesis in the cells treated with RO3306, this was not as strong as that observed with nocodazole, and we measured no further reduction in translation rates as cells progressed through mitosis. Analysis of extracts prepared from these cells (Fig. 4C) showed that there was no shift in migration of eIF4GI, only a slight phosphorylation of eIF2α and a transient limited migration of 4E-BP1 to the upper forms. This suggests that any of the changes of signaling events to these proteins are only limited or transitory, but exposure to nocodazole itself may be creating a situation where the modifications are maintained for too long, thus allowing a state of reduced translation to develop.

Figure 4. Robust phosphorylation of eIF2α does not occur following release from G₂/M block, and arresting synchronous cells with nocodazole has negligible effects on translation rates. (A) HeLa cells were maintained asynchronously, subjected to a single thymidine nocodazole block, or incubated with the Cdk1 inhibitor RO3306 for 20 h at a concentration of 9 μM to arrest cells at the end of G₂. As this latter inhibitor is reversible, the drug was washed off, and the cells progressed through mitosis, as observed by FACS analysis every 30 min. (B) Translation rates (obtained by pulsing for 15 min prior to harvest) were obtained from parallel experiments to (A). (C) Mobility shifts of eIF4GI and 4E-BP1 and the phosphorylation status of eIF2α were determined from cell extracts prepared in (B) by SDS-PAGE and immunoblotting with antibodies raised against standard or phospho-epitopes, as indicated. (D) Cells were synchronized with a double thymidine block as before, then released into medium which was either normal, supplemented with nocodazole at release, or supplemented with nocodazole 6 h after release. FACS analysis of cells at the time points indicated was used to show the arrest in G₂/M in the cells incubated with nocodazole. (E) The amount of incorporation of [³⁵S] methionine into protein for the 15 min prior to harvest was determined from parallel experiments to (D) as previously described. (F) As before, cell extracts from parallel experiments to (D) were made and subjected to SDS-PAGE and immunoblotting to show the phosphorylation status of eIF2α and 4E-BP1 in the 3 different treatments. (G) In an attempt to determine the novel phosphorylation site of 4E-BP1 that is detected following nocodazole treatment, the open reading frame of the human protein was inserted into a vector in frame with an N-terminal myc-tag and C-terminal 3× FLAG tag. Further variants were constructed where putative phosphorylation sites (S83 and S112) were mutated to alanine either singly or in combination. The 4 variants were then transfected into HeLa cells and either maintained asynchronously or treated with nocodazole for 19 h. Cell extracts were prepared and the migration of the exogenous proteins determined by SDS-PAGE, detected with anti-FLAG antibody. (H) As for (G) mutants of 4E-BP1 were prepared where putative phosphorylation sites (T82 and S83) were mutated to alanine. After transfection into HeLa cells and nocodazole treatment as before, migration of FLAG-tagged proteins was examined by immunoblotting.

To further determine the effects of sustained incubation with nocodazole, we again synchronized cells by double thymidine block and then, post-release, incubated them without nocodazole or by adding nocodazole at different times. This would induce the arrest as usual, but the nocodazole would have been present for less time than in the standard protocol. FACS analysis showed that the nocodazole induced the expected arrest, whether added at an initial or later time after release of the thymidine block (Fig. 4D). Pulse labeling of cells with [³⁵S] methionine prior to harvest showed that there was only a limited inhibition of protein synthesis at the last time point (Fig. 4E), by which point some of the cells which would have otherwise completed mitosis would have been arrested for around 4 h, rather than the 20 h utilized in the standard protocol. It is therefore not surprising that the modifications of eIF2α and 4E-BP1 were also only observed at later time points in these cells, and at a limited level as determined by immunoblotting (Fig. 4F).

In addition to the 4 mTORC1-dependent phosphorylation sites of 4E-BP1 identified at Thr 37, Thr 46, Ser 65, and Thr 70,^24,25 other phosphorylation sites were postulated at Ser 83 and Ser 112.^26,27 Therefore, we attempted to identify whether these were the sites modified following nocodazole treatment. A cDNA corresponding to the open reading frame of 4E-BP1 was obtained by RT-PCR of HEK293 cell RNA and subcloned into a vector which fused a myc-tag to the N terminus and 3 copies of the FLAG tag to the C-terminus. This vector was then subjected to site-directed mutagenesis to convert the serines at 83 and 112 to alanines, either singly or together. The wild-type vector and the 3 mutated variants were then transfected into HeLa cells which were incubated with or without nocodazole (Fig. 4G). Although it is not as obvious as observed with the endogenous protein (possibly due to the overexpression of the exogenous tagged 4E-BP1), there is an upward shift in response to nocodazole with the WT and S112A variants. However, the movement of S83A is less evident, perhaps indicating that this residue is a further site of phosphorylation in cells treated with nocodazole. Previous work has shown that 4E-BP1 can be phosphorylated by Cdk1 in vitro,¹⁴ and it is possible that a further threonine residue could be a subsequent target following phosphorylation at Thr 70 by this kinase. Given these results, Thr 82 could perhaps be such a site. To this end, we created further mutants (T82A and T82A/S83A) to examine if the nocodazole-dependent modifications were abrogated. To aid in the identification of the phosphoforms of both endogenous and exogenous proteins, the membrane was probed with both murine anti-FLAG and leporine anti-4E-BP1, which enabled us to visualize the migration using appropriate species-specific fluorescently tagged secondary antibodies. Our results (Fig. 4H, upper panel) show that Thr 82 is not likely to be a phosphosite, as the upward shift is still observed in T82A (compare lanes 4 and 6). However, the relative lack of a shift in the T82A/S83A mutant does again confirm that while Ser 83 may be the residue modified in the presence of nocodazole, or be modified in tandem with Thr 82, neither of these is the major site of phosphorylation. The large shift observed in the endogenous 4E-BP1 could indicate that nocodazole treatment affects a pair of phosphosites, and it is a consequence of overexpressing exogenous 4E-BP1 that prevents us observing these modifications to their full extent. Indeed, probing the membrane with an anti-4E-BP1 antibody shows the substantial expression of exogenous protein vs. the endogenous (Fig. 4H, lower panel).

The phosphorylation of eIF4GII maps to a novel N-terminal site

Previous work has established the presence of 3 serum-sensitive serine phosphorylation sites on eIF4GI (Fig. 5A) at 1148, 1188, and 1232 (using numbering for the longest eIF4GI isoform^20,21) with PKCα responsible for a further phosphorylation at Ser 1186.²⁸ In addition, CAMK1 has been described as phosphorylating eIF4GII (Fig. 5D) in a similar C-terminal region of the molecule (Ser 1308²²), using the numbering from our work which extended the N-terminus from a CUG initiation codon.⁸

Figure 5. Deletion analysis of eIF4G proteins reveals two novel phosphoserine sites in eIF4GII. (A) A schematic representation of eIF4GI, showing the alternative AUG translation initiation sites (f through a) and the caspase-3 cleavage sites used to generate deletion fragments of this protein. Numbering conforms to the longest isoform as independently identified by others.^20,21 Each cDNA was inserted into a vector containing an N-terminal myc-tag and C-terminal 3xFLAG tag to enable detection. (B) Plasmids containing cDNAs representing the 3 fragments of apoptotic cleavage of eIF4GI (FAGs) were transfected into HeLa cells either untreated or incubated with nocodazole for 19 h. Any changes in migration of the fragments due to the presence of nocodazole was determined by SDS-PAGE and immunoblotting using anti-FLAG antibody to detect fragment expression. (C) Further deletion fragments of N-FAGf were constructed as before to delineate regions likely to be phosphorylated, these were transfected into HeLa cells and migration in the presence or absence of nocodazole was examined. (D) Deletion fragments of the N-terminus of eIF4GII were constructed as before, in order to map the phosphorylation site(s) of this protein. The amino acid numbering conforms to the CUGb initiated isoform previously identified by our group,⁸ which extends the open reading frame N-terminally from the initially published AUG. (E) Migration of the N-terminal fragments was examined following treatment with nocodazole for 19 h, with proteins detected with anti-FLAG antibody. (F) Further subfragments were made of the eIF4GII open reading frame, as indicated in (D), and their migration in cells treated with nocodazole determined as before.

To delineate the sites of phosphorylation observed in nocodazole treatment, we created cDNAs corresponding to regions of both proteins. First, eIF4GI was separated into fragments corresponding to those that arise from caspase-directed cleavage of the protein during apoptosis. This separates the molecule into 3 “FAGs,”^29-31 with each corresponding cDNA inserted into the same N-myc, C-3xFLAG vector as before. Each vector was then transfected into HeLa cells, and 48 h after transfection, the cells were incubated with or without nocodazole. Figure 5B shows that there was a small, reproducible shift in only the N-terminal fragment (N-FAGf, lanes 1 and 2) and also more of the upper band in the doublet from C-FAG expressing cells. As 2 of the 3 known phosphoserines on eIF4GI are in this latter fragment, we concentrated our efforts on N-FAGf, splitting it into further fragments, and carrying out experiments in a similar manner (Fig. 5C). There was a slight shift in the f-QIAP and e-QIAP fragments, suggesting there may be a novel phosphorylation site between amino acids 41 and 147 (compare lane 1 with 2, and 3 with 4). However, this phosphorylation does not appear to be as robust as that observed in the full-length protein and implies either that the protein must be intact or, more likely, that there are multiple sites present on eIF4GI and only when phosphorylated together is migration sufficiently retarded.

Our focus then turned to eIF4GII, with the N terminus of this protein studied by again making cDNAs for insertion in the epitope-tagged vector (Fig. 5D). In this case, there was a much clearer response to nocodazole treatment (Fig. 5E), with a clear shift in migration of the MSGG-EELS (311–688) fragment (compare lanes 7 and 8). To map the site of modification, further fragments were constructed and examined in Figure 5F. While there was a perceptible shift in STVP-EELS, there was a much more obvious retardation of MSGG-AAPT (311–435) in nocodazole-treated cells (compare lanes 4 and 5). This subfragment was then subjected to immunoprecipitation by virtue of the C-terminal 3xFLAG tag, allowing us to examine the uppermost migrating species by mass spectroscopy and identify the site of modification.

Phosphorylation of eIF4GII occurs at 2 sites, with Cdk1 likely to be the kinase responsible for their modification

Mass spectrometric analysis of peptides generated from the MSGG-AAPT fragment identified 2 novel phosphorylation sites at Ser 384 and Ser 392 (Fig. S2). These sites were aligned to the eIF4GI sequence (Fig. 6A), which shows that they lie in a region of poor conservation and are hence unique to this paralog. However, it is interesting to note that Ser 384 is a glutamic acid in eIF4GI, which suggests that the phosphorylation of this residue shifts the charge in this region of eIF4GII closer to that of eIF4GI. Our first task after making this identification was investigating whether alanine mutants of either of these sites would abrogate the nocodazole-dependent shift. Therefore site-directed mutagenesis was carried out on one, the other, or both sites and the migration of these mutants and the wild-type MSGG-APPT fragment were assayed as before (Fig. 6B). The shift is no longer observed in the double mutant (S384/392A), implying that both sites need to be phosphorylated to cause the shift, but there is no preference as to the order of modification.

Figure 6. Identification of phosphorylation of eIF4GII at S385 and S389 by Cdk1, and consequences for initiation factor complex formation. (A) Mass spectroscopy was used to identify phosphorylation of two serine residues (384 and 392) within the MSGG-APPT fragment of eIF4GII following immunoprecipitation of the fragment from cells treated with nocodazole. The sites are shown in bold and underlined, and the surrounding sequences were aligned with eIF4GI using CLUSTALW, showing the poor conservation of this particular region of the protein. The site of PABP binding is shown in gray, with a dashed underline. The symbols *: and . respectively denote identical residues, or conserved and semi-conserved substitutions in the alignment. (B) Ser–Ala mutants, either single or in combination, of the MSGG-APPT fragment of eIF4GII were made and inserted into the myc-3× FLAG vector as before. Cells transfected with these vectors were incubated with or without nocodazole and migration of the proteins examined as before. (C) Prediction software (GPS 2.1.2) indicated that likely kinases responsible for this phosphorylation were p38 MAPK or Cdk1. Therefore cell extracts were made from cells transfected with MSGG-APPT that had been incubated with or without nocodazole and cell-permeable kinase inhibitors. These would either block p38 MAPK signaling (SB202190) or Cdk1 signaling (RO3306). Migration of the eIF4GII fragment was determined as previously described. (D) To confirm the results from panel (C), immunoprecipitated MSGG-AAPT protein was incubated in the presence of recombinant Cdk1/cyclin B and ATP. This was run alongside a positive control of extract from cells treated with or without nocodazole. (E) To determine whether the novel phosphorylation sites have any influence on the interactions of eIF4GII with other components of the translation initiation factor machinery, the S384/392A double mutant was introduced into the full-length eIF4GII CUGb ORF. This vector contains an N-terminal myc tag, and therefore exogenous eIF4GII could be immunoprecipitated from transfected cells which had been incubated with nocodazole. The left hand panel shows the levels of endogenous proteins as determined by SDS-PAGE and immunoblotting, and the right hand panel shows the eluate from a co-immunoprecipitation where myc-9E10 antibody had been captured on agarose resin.

Bioinformatic analysis using GPS 2.1.2³² of the novel eIF4GII phosphorylation sites indicated that the most likely kinases responsible were Cdk1 and p38 MAPK. An experiment was therefore set up to investigate whether inhibitors of either kinase would abolish the nocodazole-dependent shift (Fig. 6C). The modification was still evident in cells treated with the p38 MAPK inhibitor SB202190, but was lost in those treated with nocodazole and RO3306, the inhibitor of Cdk1. As the latter compound causes a synchronization of the cells prior to the arrest induced by nocodazole treatment, this result is perhaps not surprising. It therefore indicates that Cdk1 may be influencing the phosphorylation of eIF4GII either directly or indirectly; however, this experiment does not offer definitive proof.

In vitro kinase assays were therefore undertaken to allow the direct phosphorylation of the MSGG-APPT fragment, which had been immunoprecipitated from asynchronous cells. In a “cold” kinase assay (Fig. 6D), there is a slight mass increase in the upper band of the MSGG-AAPT fragment (compare lanes 1 and 2), giving direct evidence that Cdk1 is indeed the kinase responsible for phosphorylation of eIF4GII. However, in the context of this experiment, the phosphorylation is nowhere near as robust as that observed with the same fragment expressed in HeLa cells exposed to nocodazole (lane 4).

Having confirmed the presence of novel phosphorylation sites in the N terminus of eIF4GII, we finally wished to determine how they influenced the binding capacity of this scaffold protein in the context of the full-length protein. The double S384/392A mutation was introduced into a vector which contained the full eIF4GII open reading frame arising from our recently identified CUG initiation codon, fused with a myc-tag at the N terminus. This tag was then used to co-immunoprecipitate wild-type or mutant eIF4GII from extracts from cells treated with or without nocodazole (Fig. 6E). Extract from untransfected cells was also subjected to immunoprecipitation to ensure that any eluted proteins were present due to interaction with eIF4GII, not the agarose resin (lane 8). Due to the conditions of the assay, it was not possible to identify any eIF4A in the immunoprecipitation eluate, but PABP and eIF4E both co-eluted. Quantification of the ratios of eIF4GII or myc to either protein shows that there is a reduction in PABP binding on nocodazole treatment, which is much more severe in the S384/392A mutant. Given the proximity of the PABP binding site to the phosphorylated region, it is perhaps unsurprising that there is some influence on the interaction.

Subjecting the eluates to low-percentage SDS-PAGE indicates that there is still some possible residual modification of the eIF4GII protein which contains the S384/392A mutant, indicating that there are other nocodazole-dependent phosphorylations elsewhere in the protein, as discussed below.

Discussion

Our work addresses some of the discrepancies that have been reported when observing cycling cells vs. those that are arrested at various points in the cell cycle using pharmacological, nutritional, or environmental treatments. The general consensus that there is a lack of protein synthesis during mitosis has come from the observation that treatment with colcemid, nocodazole, or other microtubule disrupters (always for an extended period of time) results in a reduction in the incorporation of radiolabeled amino acids into precipitable proteins.^10,16

However, our work following cycling cells after release from a double thymidine block shows that there are no substantial changes in overall protein synthesis rates as measured by incorporation of radiolabeled amino acids (Fig. 1). This is coupled with our observations that initiation factors and ribosomal proteins co-localize with tubulin at various points during mitosis (Fig. 2), suggesting that ongoing translation is indeed important for the process. Previous work examining the protein content of isolated CHO cell midbodies also revealed the presence of other ribosomal proteins and translation elongation factors,³³ which supports our hypothesis that new proteins must be synthesized in order for proper abscission to proceed. Others have observed ribosomal protein S3 at the mitotic spindle³⁴ and suggested that this is due to the protein having a function in a process other than translation. However, given our observation of other ribosomal proteins and initiation factors at this location, we suggest that involvement in protein synthesis is more likely. The observation that rpS3 knockdown causes arrest in metaphase,³⁴ could also be interpreted as there being a key requirement for translation of mRNA during mitosis. Furthermore, our previous work examining the effects on cell morphology following knockdown of eIF4GI by shRNA¹⁹ showed an uncoupling of karyokinesis from cytokinesis, resulting in production of large cells with multiple nuclei, with similar results observed when eIF3e/Int-6 was depleted from cells.³⁵ This phenotype is typically observed when similar experiments knockout or overexpress dominant-negative forms of chromosomal passenger proteins (e.g., Aurora B, survivin³⁶), suggesting these may be at least some of the particular components whose translation must be maintained during mitosis.

As there are well-characterized interactions between cytoskeletal components and ribosomal proteins, translation initiation factors, and various mRNPs,³⁷ it is therefore not surprising that using an agent such as nocodazole, which stops assembly of microtubules and also disassembles those previously made, should have such a drastic effect on protein synthesis. The link between cellular stress, the phosphorylation of eIF2 on Ser 51 of the α-subunit by one of four kinases, and the consequential downregulation of translation rates has been comprehensively studied.³⁸ As others have previously observed, there is robust phosphorylation of eIF2α following incubation with nocodazole, and we also show that other components of the translation initiation machinery are phosphorylated. As phosphorylation of eIF2α still occurs in PKR^−/− MEFs (Fig. S1), these data indicate a role for PERK in modification of eIF2. Unfortunately, PERK^−/− MEFs did not arrest with nocodazole and consequently showed no change in eIF2α phosphorylation.

The eIF4E-binding protein 4E-BP1 was observed to migrate as 2 forms, one corresponding to a completely unphosphorylated form (commonly referred to as the α form), while the other was retarded further in its migration compared with the gamma form which is phosphorylated at 4 sites (Thr 37, Thr 46, Ser 65, and Thr 70).²⁴ This led us to attempt to pinpoint the further sites of modification but our results mutating Thr 82, Ser 83, and Ser 112 proved inconclusive. Advances in the sensitivity of proteomic instrumentation has led several groups to undertake large-scale phosphoproteomic analysis of cells (including HeLa cells) treated with nocodazole. For example, a recent publication involving HEK293 cells found a further 11 sites of phosphorylation over the standard 4, which included Ser 83 and Ser 112.³⁹ Given that the shift in migration we observe does not correspond to such a drastic change in phosphorylation, which would result in multiple species on an immunoblot, it is possible that a single 4E-BP1 protein only undergoes some but not all possible modifications, in a particularly disorganized manner. In addition, how some 4E-BP1 protein remains in a completely unphosphorylated state is an as yet unanswered question. This could reflect some sort of protection in a compartment or complex which is inaccessible to the kinase(s) activated upon sustained incubation with nocodazole. It is also feasible that only pre-phosphorylated forms of 4E-BP1 can then undergo further modifications, analogous to the observation that only phosphorylated 4E-BP1 is subjected to ubiquitination.^40,41

Our investigations of the eIF4G paralogs were successful in identifying 2 sites of phosphorylation on eIF4GII at Ser 384 and Ser 392, through a combination of deletion analysis and unphosphorylatable serine–alanine mutations. While we have not conclusively identified the kinase responsible for this modification, it does appear likely to be Cdk1. The larger-scale study of HEK293 cells treated with nocodazole independently confirmed these modifications and also indicates several others, including one at Ser 382, one at Ser 644 (so may be evident when comparing Fig. 5E, lanes 8 and 9) and another at Ser 1560. A caveat of this previously published work is that, due to the experimental design, no distinction can be made between those residues which were phosphorylated in response to nocodazole or already modified. However, of particular interest was a phosphothreonine residue identified in the extended N terminus, initiated from a CUG which is completely absent in eIF4GI proteins; the analysis could not determine whether Thr 64 or Thr 66 was the amino acid in question. This particular modification could account for the altered banding pattern observed above the main band in Figure 5E, lane 4.

This comprehensive study did not identify any phosphopeptides in eIF4GI, although an earlier independent study⁴² did find phosphorylation of this paralog at both known sites (Ser 1186 and Ser 1188), together with another at Ser 1232. All these sites reside in the C-FAG fragment of eIF4GI and again could reflect the alteration in intensity of the 2 bands observed in Figure 5B, comparing lanes 5 and 6.

The sites that we and others have identified in eIF4GII are between the regions known to bind PABP and eIF4E, which is notable as being the most poorly conserved region and lacks any currently unassigned binding partners in either paralog (Fig. 5A and D). Our previous work identified a nuclear localization signal in eIF4GI further downstream of this region, which is not present in eIF4GII.¹⁷ Taken together these findings confirm that, despite initial experiments showing that the paralogs were functionally interchangeable, they are subject to several alternative regulatory controls. We have not been able to identify any clear changes in the composition of the eIF4F complex in response to these modifications, but they may be acting functionally somewhere else in the initiation pathway other than during the assembly of the mRNA-binding complex of eIF4E:eIF4G:PABP.

It is clear from our work and the larger-scale phosphoproteomic surveys that a number of initiation factors are modified by kinases following sustained nocodazole treatment, with the general consequence that translation rates are significantly reduced. However, as others have previously shown,¹⁶ a subset of messages is maintained in actively translating pools, with IRES-dependent translation postulated to be responsible for maintaining their translation, while general cellular messages are inhibited. Different IRESs have alternative requirements for the canonical initiation factor machinery, and it would be interesting to determine the eIF4G requirement and which particular modifications exert effects on IRES-driven translation.

For a cell to progress through mitotis, a plethora of regulated phosphorylation and dephosphorylation events must occur.⁴³ Our work has demonstrated increases in phosphorylation of key translation initiation factors following nocodazole treatment, which arrests cells at the G₂/M boundary. However, when adding nocodazole for shorter times, these modifications were not observed. This suggests either that phosphorylation is a consequence of prolonged incubation, which may activate further stress pathways, or that phosphorylation is extremely transient and only perceivable when given the opportunity to accumulate in stalled cells. While it is still not clear why these modifications are specifically required for mitotic progression, they do show that there is still much to learn about phosphorylation of initiation factors in different cellular conditions. Given that there is a considerable amount of research into using microtubule and other cytoskeletal disruptors as chemotherapeutic agents,^44-47 it is important to consider how these treatments may also be impinging upon gene expression at the level of translation initiation.

Materials and Methods

Cell culture and transient transfection

Materials for tissue culture were from Invitrogen (UK), and fetal bovine serum (FBS) was from Labtech International or Invitrogen. HeLa (cervical cancer) cells were obtained from the ECACC and maintained in DMEM supplemented with 10% FBS at 37 °C in a humidified atmosphere containing 5% CO₂.

Cell cycle synchronization of HeLa cells was performed following conditions as described by others.¹⁵ Briefly, for the double thymidine block, cells were incubated for 16 h with 2 mM thymidine, released for 8 h, then synchronized for a further 16 h with 2 mM thymidine. For the thymidine/nocodazole treatments, the cells were incubated for 16 h with 2 mM thymidine, released for 6 h, then incubated in 0.6 µg/ml nocodazole for 20 h.

For transient transfection cells were seeded on 6-cm plates at a density of 1 × 10⁵ cells per plate, and 24 h later, cells were transfected with 1 µg of pcDNA plasmid using GeneJuice (http://www.merckmillipore.com/united-kingdom/chemicals/genejuice-transfection-reagent/EMD_BIO-70967/p_qI6b.s1OxjIAAAEjOhx9.zLX?PortalCatalogID=merck4biosciences), according to the manufacturer’s protocol. The transfection mixture was removed 24 h later, and the cells were washed twice with PBS and further incubated in fresh media for 24 h until the addition of nocodazole. Preparation of cell lysates, and measurement of protein synthesis rates by incorporation of [³⁵S] methionine into protein were performed as previously described.¹⁹ Cell extracts were also prepared using M-PER reagent (Pierce, http://www.piercenet.com/browse.cfm?fldID=06010420), supplemented with Halt phosphatase and protease inhibitor cocktail (Pierce, http://www.piercenet.com/browse.cfm?fldID=E780097B-5056-8A76-4EFC-572566DA6E04), as per the manufacturer’s instructions.

Plasmids

pcDNA plasmids containing N-terminal myc tags and C-terminal 3× FLAG tags have been described previously.^8,17,19 For this study, cDNAs corresponding to portions of eIF4GI/II or 4E-BP1 were subcloned from existing plasmids or HeLa cell cDNA using appropriate primers to maintain the open reading frame with the epitope tags, removing any endogenous stop codons where necessary. Site-directed mutagenesis to generate alanine mutations of phosphorylation sites was performed using the Agilent Quikchange Lightning kit (http://www.genomics.agilent.com/en/product.jsp?cid=AG-PT-175&tabId=AG-PR-1162&_requestid=640783) in accordance with the manufacturer’s instructions, using primers synthesized as directed by the Agilent Quikchange Primer design program. All sequences were confirmed by automated sequencing (Eurofins MWG or Beckman Coulter Genomics).

FACS analysis

At each time point, to ensure all cells both adherent and in suspension were harvested, all growth medium was removed and combined with a single PBS wash, and the remaining cells were subsequently harvested from the plate by dissociation with trypsin. Following centrifugation (1000 × g for 5 min) and a single wash with ice-cold PBS, cells were fixed in 3 ml of freshly made ice-cold 70% EtOH, 30% PBS, and left at 4 °C overnight. Each sample was washed twice in PBS before being resuspended in 500 µl PBS, containing 0.1 mg/ml RNase A to digest RNA and 0.03 mg/ml propidium iodide solution to stain DNA, prior to incubation at RT for 30 min. Samples were analyzed with a BD FACScan^™ using appropriate filter sets.

Immunoblotting

Antibodies to eIF4GI, eIF4GII, eIF4A, eIF4E and PABP were prepared in house and have been previously described.^17,19 Antibodies to total 4E-BP1 http://www.cellsignal.com/products/9452.html, phospho-4E-BP1 (Ser65) http://www.cellsignal.com/products/9451.html, phospho-4E-BP1 (Thr70) http://www.cellsignal.com/products/9455.html, total eIF2α http://www.cellsignal.com/products/9722.html and phospho-eIF2α http://www.cellsignal.com/products/9721.html were from Cell Signaling Technology. Antibodies to myc (9E10) http://www.sigmaaldrich.com/catalog/product/sigma/m4439?lang=en&region=GB FLAG (M2) http://www.sigmaaldrich.com/catalog/product/sigma/f1804?lang=en&region=GB and actin http://www.sigmaaldrich.com/catalog/product/sigma/a2668?lang=en&region=GB were from Sigma-Aldrich.

Polysome gradients

Sucrose density gradients to separate ribosomal subunits and those ribosomes associated with mRNAs were performed as previously described.⁴⁸

Isolation of protein complexes from cell extracts

m⁷GTP-Sepharose affinity isolation of eIF4E and associated factors, were performed as previously described.¹⁹ Immunoprecipitation and co-immunoprecipitation experiments were performed using the Pierce Co-Immunoprecipitation kit http://www.piercenet.com/browse.cfm?fldID=AB59D109-9690-4EAA-8521-E6C407156407, as per the manufacturer’s instructions, with anti-myc 9E10 or anti-FLAG M2 immobilised to the resin, depending on the experiment.

Mass spectrometric analysis

FLAG-tagged immunoprecipitated protein was first separated by 1D-SDS-PAGE using a precast Novex Bis-Tris 4–12% gel http://www.invitrogen.com/site/us/en/home/Products-and-Services/Applications/Protein-Expression-and-Analysis/Protein-Gel-Electrophoresis/Protein-Gels/Novex-NuPAGE-SDS-PAGE-Gels/NuPAGE-SDS-PAGE-Bis-Tris-Gel.html. Following staining with Colloidal Coomassie blue, the gel band of interest was subjected to in situ trypsin digestion using the method of Schevchenko et al.⁴⁹ Peptide extracts were enriched for phosphopeptides using TiO₂ ProteaTips (Protea Biosciences, https://proteabio.com/products/SP-125) and used in accordance with the manufacturer’s instructions. Lyophilized peptide samples were re-suspended in 50 mM sodium citrate pH 6.0 and loaded onto a reverse phase trap column (Symmetry C18, 5 µm, 180 µm × 20 mm, Waters Corporation http://www.waters.com/waters/en_GB/Symmetry-Columns/nav.htm?cid=513219&locale=en_GB), at a flow rate of 5 µl/min and washed for 10 min with 3% acetonitrile containing 0.1% formic acid (buffer A) prior to the analytical nano-LC separation using a C18 reverse-phase column (HSS T3, 1.8 µm, 200 mm × 75 µm, Waters http://www.waters.com/waters/en_GB/HSS-%28High-Strength-Silica%29Technology/nav.htm?cid=134618105). Separation of peptides was achieved using a gradient of 0–50% buffer B (95% acetonitrile containing 0.1% formic acid [v/v]) at a flow rate of 300 nl/min. Eluted samples were sprayed directly into a Synapt G₂-S mass spectrometer (Waters) operating in MSe mode. Data was acquired from 50 to 2000 m/z using alternate low and high collision energy (CE) scans. Low CE was 5V and elevated collision energy ramp from 15 to 40V. Ion mobility was implemented prior to fragmentation using a wave velocity of 650 m/s and wave height of 40V. The lock mass Glu-fibrinopeptide, ([M+2H]+2, m/z = 785.8426) was infused at a concentration of 100 fmol/µl with a flow rate of 250 nl/min and acquired every 60 s.

The raw mass spectra were submitted to ProteinLynx Global Server version 2.5.2 (Waters) and the data processed to generate reduced charge state and deisotoped precursor and associated product ion mass lists. These mass lists were searched against the human UniProt protein sequence database (August, 2012) using the MASCOT search algorithm. A maximum of one missed cleavage was allowed for tryptic digestion with a fixed modification of carboxyamidomethylation of cysteines and variable modifications set to allow for oxidation of methionine and the phosphorylation of serine, threonine, and tyrosine.

Precursor ion and sequence ion mass tolerances were set at 20 ppm and 0.25 Da, respectively. The significance threshold for search results was set at P < 0.05.

Kinase assays

Recombinant Cdk1/cyclinB1 was purchased from New England Biolabs https://www.neb.com/products/p6020-cdk1-cyclin-b, and the fragment of eIF4GII was obtained from an IP eluate and phosphorylated with 10 units of kinase for 30 min at 30 °C in the presence of 200 µM ATP, in accordance with the manufacturer’s instructions. Enzyme activity was confirmed in parallel experiments by phosphorylation of eEF2K⁵⁰ in the presence of 200 µM unlabeled ATP and 1 µCi of [γ-³²P] ATP.

Glossary

Abbreviations:

4E-BP

eIF4E-binding protein

CAMK1

calcium/calmodulin-dependent protein kinase type 1

Cdk1

cyclin-dependent kinase 1

eIF

eukaryotic initiation factor

FACS

fluorescent-activated cell sorting

IRES

internal ribosome entry site

mTORC1

mammalian target of rapamycin complex 1

PABP

poly(A)-binding protein

PKC

protein kinase C

10.4161/cc.26588

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Materials

Supplemental materials may be found here:

www.landesbioscience.com/journals/cc/article/26588