Regulation of eukaryotic translation initiation factor 6 dynamics through multisite phosphorylation by GSK3

Eukaryotic translation initiation factor 6 (eIF6) is essential for the synthesis of 60S ribosomal subunits and for regulating the association of 60S and 40S subunits. A mechanistic under-standing of how eIF6 modulates translation in response to stress, specifically starvation-induced stress, is lacking. We here show a novel mode of eIF6 regulation by glycogen synthase kinase 3 (GSK3) that is predominantly active in response to serum starvation. Both GSK3 a and GSK3 b phosphorylate human eIF6. Multiple residues in the C terminus of eIF6 are phosphorylated by GSK3 in a sequential manner. In response to serum starvation, eIF6 accumulates in the cytoplasm, and this altered localization depends on phosphorylation by GSK3. Disruption of eIF6 phosphorylation exacerbates the translation inhibitory response to serum starvation and stalls cell growth. These results suggest that eIF6 regulation by GSK3 contributes to the attenuation of global protein synthesis that is critical for adaptation to starvation-induced stress.

Eukaryotic translation initiation factor 6 (eIF6) is essential for the synthesis of 60S ribosomal subunits and for regulating the association of 60S and 40S subunits. A mechanistic understanding of how eIF6 modulates translation in response to stress, specifically starvation-induced stress, is lacking. We here show a novel mode of eIF6 regulation by glycogen synthase kinase 3 (GSK3) that is predominantly active in response to serum starvation. Both GSK3a and GSK3b phosphorylate human eIF6. Multiple residues in the C terminus of eIF6 are phosphorylated by GSK3 in a sequential manner. In response to serum starvation, eIF6 accumulates in the cytoplasm, and this altered localization depends on phosphorylation by GSK3. Disruption of eIF6 phosphorylation exacerbates the translation inhibitory response to serum starvation and stalls cell growth. These results suggest that eIF6 regulation by GSK3 contributes to the attenuation of global protein synthesis that is critical for adaptation to starvation-induced stress.
Eukaryotic initiation factor 6 (eIF6) is a key modulator of translation initiation that regulates the biogenesis and availability of the 60S ribosomal subunits (1)(2)(3). eIF6 directly associates with pre-60S complexes in the nucleolus and is exported into the cytoplasm in complex with the 60S where it aids in 60S maturation (4)(5)(6)(7)(8). The well-characterized role of eIF6 is its antiassociation activity that prevents interactions between the 60S and 40S ribosomal subunits (2,(9)(10)(11)(12). Structural and biochemical studies indicate that eIF6 binds to 60S at the 40S-binding interface and sterically hinders association of the 40S subunits (10)(11)(12)(13)(14)(15)(16). Release of eIF6 from the mature 60S allows association of the 40S-mRNA complex, which leads to active 80S formation and translation initiation (10)(11)(12)(13)(14)(15)). An inhibition of eIF6 release (or premature release) from the 60S subunits can greatly influence intersubunit association and translation initiation. A block in eIF6 release leads to an increase in the fraction of eIF6bound 60S subunits that are unable to join 40S, which hinders the assembly of active 80S monosomes (10-13, 15, 16). Alternatively, insufficient levels of eIF6 seem to lead to spurious association of 60S with the 40S subunits that are devoid of mRNA. This causes an increase in the assembly of inactive 80S monosomes, which impairs translational response to growth stimuli as seen in eIF6 1/2 mice (17)(18)(19). Thus, an impairment of eIF6 function can substantially affect ribosome levels, limiting protein synthesis specifically in response to growth stimuli, and this deregulation has been shown to contribute to the underlying pathologies of diseases such as Shwachman-Bodian-Diamond syndrome, cancer, and certain metabolic disorders (13,(17)(18)(19)(20).
In response to stimuli such as insulin or phorbol esters, primary hepatocytes and mouse embryonic fibroblasts (MEFs) derived from the eIF6 1/2 mice do not up-regulate protein synthesis, unlike the WT cells (17)(18)(19). Similar defects in translation are also observed in vivo, where livers of eIF6 1/2 mice are smaller and exhibit an accumulation of inactive/empty (mRNA-free) 80S ribosomes compared with WT (17)(18)(19). The lack of insulin-dependent stimulation of protein synthesis in the eIF6 1/2 mice has been associated with a reprogramming of fatty acid synthesis and glycolytic pathways with implications for muscle, liver, and fat metabolism (19,21).
In terms of mechanism, the increased formation of inactive 80S complexes in the eIF6 1/2 cells is attributed to an impairment of its anti-association function in the cytoplasm (17)(18)(19). Interestingly, such an accumulation of inactive 80S monosomes is commonly observed in cells subjected to stress, especially stress induced by nutrient deprivation or limitation (22)(23)(24). Starvation or nutrient limitation in yeast and mammalian cells invokes an adaptive metabolic response that conserves energy by restricting global protein synthesis and leads to an accumulation of inactive (empty) 80S ribosomes along with an increase in the pool of free 60S subunits (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33). The role of eIF6 as a ribosome anti-association factor and 60S biogenesis factor in modulating this starvation response has not been thoroughly explored. Given the key role of eIF6 in regulating translational response to insulin and growth factors in an mTOR-independent manner (19,34), it is important to understand the mechanisms that control eIF6 in response to growth inhibitory stress responses. To address this, here we report a novel regulation of eIF6 by glycogen synthase kinase 3 (GSK3) that is active under conditions of serum starvation-induced stress.
Global proteomic and biochemical studies indicate that human and murine eIF6 is phosphorylated at multiple sites, and the majority of these sites are conserved and cluster around the C-terminal tail (35)(36)(37)(38)(39)(40)(41)(42). However, most of these phosphorylation sites have not been validated in vivo, and the identities of the associated kinases have not been clarified. It is also unclear whether these uncharacterized phosphorylation sites carry any functional relevance. Here we report a novel This article contains supporting information. * For correspondence: Sofia S. Origanti, sofia.origanti@slu.edu. regulation of eIF6 by GSK3 that is active under conditions of serum starvation-induced stress. Based on sequence-motif prediction analysis, we identified the presence of a GSK3-specific motif within the C-terminal tail of eIF6. Previous studies have shown that GSK3 is a unique kinase that is activated in response to growth inhibitory conditions such as starvation or under resting states and is inhibited in response to insulin and other growth stimulatory conditions by AKT and mTORC1-p70S6K1-dependent phosphorylation (43)(44)(45)(46)(47)(48)(49). GSK3 plays a prominent role in translational control by inhibiting the nucleotide exchange function of the translation initiation factor: eIF2B, which is reversed in response to insulin (43,(50)(51)(52). GSK3 phosphorylates the Ser-540 site in the e subunit of eIF2B to inhibit its activity (43,(50)(51)(52). Here, we show that GSK3 also influences translation by regulating human eIF6 through multisite phosphorylation at Ser-243, Ser-239, Ser-235, and Thr-231 sites. GSK3 interacts with endogenous eIF6, and phosphorylation by GSK3 alters the subcellular localization of eIF6 in response to serum starvation. Interestingly, the phosphodead mutant and the C-terminal deletion mutant displayed increased levels of free 60S subunits in response to serum starvation. Altering the steady-state levels of ribosomal subunits even by a small percentage can have detrimental effects as seen in several ribosomopathies (53,54). Expression of the phosphodead mutant decreased translation rates and stalled cell growth in response to serum starvation. Based on these findings that eIF6 is a novel substrate of GSK3, we propose a model wherein GSK3-dependent phosphorylation of eIF6 contributes to translation inhibition in response to serum starvation by regulating 60S availability.

GSK3 phosphorylates the C-terminal tail of eIF6
Sequence analysis of the C terminus of human eIF6 revealed the presence of sequential (S/T)XXX(S/T) motifs as potential candidates for phosphorylation by GSK3. We identified two such multisite motifs (motif 1 and 2) in the C terminus that are highly conserved in higher eukaryotes from Xenopus to mammals ( Fig. 1A) but show variability in lower eukaryotes such as yeast. Canonically, GSK3 phosphorylates multiple residues on its substrates in a sequential manner starting with the C terminus. GSK3 often works synergistically with a priming kinase that phosphorylates a priming site located 4 residues away from the GSK3 recognition site (Fig. 1A); however, GSK3 can also work independently of a priming kinase (43,45,46,(55)(56)(57). The potential priming sites on eIF6 are indicated in Fig. 1A. To determine whether eIF6 is a substrate of GSK3, we purified full-length human eIF6 from Escherichia coli and incubated recombinant eIF6 with GSK3. GSK3a and GSK3b are two isoforms of GSK3 that are largely redundant in their activities. Human eIF6 was phosphorylated by both the isoforms of GSK3 (Fig. 1, B and C). We also observed autophosphorylation of GSK3 as previously reported (45). The kinases were used at low concentrations that would not be detected by Coomassie staining, but the autophosphorylation showing similar kinase activities could be detected. Phosphorylation of eIF6 was detected by the incorporation of radiolabeled phosphate in eIF6 in the presence of GSK3 only. Control reactions lacking eIF6 or GSK3 did not exhibit a band at these molecular weights highlighting specificity.
To validate these results and to determine whether eIF6 obtained under a physiological context was also a substrate for GSK3, eIF6 was immunoprecipitated from HCT116 cells that were briefly serum-starved, to capture potentially preprimed eIF6. We observed that the immunoprecipitated eIF6 is also phosphorylated by GSK3b (Fig. 1D), thus confirming eIF6 as a bona fide target of GSK3. Also, GSK3b displayed a time-dependent activation in response to serum starvation, which was verified by probing for the loss of inhibitory phosphorylation of the serine 9 residue that is phosphorylated by kinases such as AKT (Fig. S1, A and B) (44, 46-49, 55, 56).
We next tested for specificity of phosphorylation by eliminating the two multisite motifs in the C terminus by creating a deletion mutant that lacks the last 36 amino acid residues (eIF6-DC36). GSK3b did not phosphorylate eIF6-DC36, whereas it phosphorylated the full-length (FL) eIF6, suggesting that the last 36 residues are critical for phosphorylation (Fig.  1D). We did observe a faint background signal for eIF6-DC36 incubated with kinase (Fig. 1D, lane 3). Prolonged exposures of the blots have captured a similar nonspecific band for the Mycempty vector incubated with the kinase (Fig. S1C), which could be associated with immunoprecipitation.
We also observed a doublet of eIF6 that was more obvious for the immunoprecipitated substrate, and such a doublet has been observed before (14). This gel effect likely arises from the high cysteine content of human eIF6 (9 Cys residues for a 26-kDa protein); when denatured, the exposed Cys residues likely form nonspecific disulfide bonds that are resistant to reduction. Also, the addition of multiple negatively charged phosphates could further alter its gel migration patterns. The intensity of doublet varied based on the concentration of recombinant eIF6 (Fig. S1D) and also based on resolving gel percentage and concentration of DTT in sample buffer (data not shown). Both the bands of the doublet were confirmed to be full-length recombinant human eIF6 by MS (data not shown). Such a denaturing gel effect has been observed for other cysteine-rich proteins (58).

GSK3 phosphorylates multiple sites within the last 20 amino acid residues of eIF6
To identify the specific motif that is phosphorylated by GSK3, we analyzed a C-terminal tail deletion mutant lacking the last 20 residues (motif 2) (eIF6-CD20). Indeed, deletion of just motif 2 greatly reduced eIF6 phosphorylation ( Fig. 2A) and confirmed that motif 2 is critical for recognition by GSK3b. In this case, motif 2 is in the disordered region of the C-terminal tail of eIF6 that is predicted to protrude outside the core structure, which would be favorable for regulatory interactions (10,16). Also, all the predicted sites indicated in motif 2 were previously identified to be modified by phosphorylation in global proteomic studies, including those performed under cellular conditions of stress (35)(36)(37)(38)(39)(40)(41).
Computational analyses indicated that the C-terminal tail carrying motif 2 is highly conserved in vertebrates (especially birds and mammals), and analysis of 50 species of vertebrates indicated 100% identity in the GSK3-specific motif 2 in the Ctail (Fig. 2B). In addition, the serine/threonine residues at 239 and 243 are highly conserved in lower eukaryotes including budding and fission yeast.
To identify the specific residues in motif 2 that are phosphorylated by GSK3b, we carried out nanoscale liquid chromatography with tandem MS analysis (nano-LC-MS/MS) on the immunoprecipitated eIF6 sample incubated with and without GSK3b. MS analysis revealed that the Thr231 residue was phosphorylated by GSK3b ( Fig. 2C and Fig. S2), and its detection by MS indicated a better preservation of the Thr-231 phospho site. The relative abundances of the peptides in samples incubated with and without the kinase were found to be identi-cal (Fig. S2, A-D). The technical limitations of MS to capture and accurately identify two or more sites of phosphorylation within a single peptide, especially when sample levels are limiting, made it challenging to capture other C-terminal phospho sites (59,60). However, these results validated that motif 2 in the C-terminal tail is phosphorylated by GSK3b.

GSK3 phosphorylates multiple sites on eIF6 in a sequence
To directly test whether other GSK3-specific sites in motif 2 are phosphorylated, we used the site-directed mutagenesis approach in which the indicated serine or threonine sites of phosphorylation within motif 2 were substituted with alanines ( Fig. 2D). In the presence of multiple and adjacent recognition sites, GSK3 phosphorylates residues in a sequence starting from the C terminus. Examples of sequential phosphorylation Figure 1. GSK3 phosphorylates a multisite motif in the C-terminal tail of eIF6. A, sequence alignment of the amino acid residues 2205 to 245 in the C-terminal tail of human eIF6 (NP_001254739) with Xenopus (NP_001083080.1), mouse (NP_034709.1), and rat eIF6 (NP_001032429). The conserved residues are marked with asterisks. Putative GSK3-specific phospho sites and priming site in multisite motifs 1 and 2 are indicated in pink and red letters, respectively. B and C, in vitro kinase reactions were carried out in the presence and absence of recombinant full-length human eIF6 and GSK3b (B) and GSK3a (C). The reactions were analyzed by autoradiography (autorad.) and Coomassie staining. Each experiment was repeated three independent times. D, the deletion mutant lacking both motifs 1 and 2 (eIF6-DC36) was generated by substitution of Thr-210 with a stop codon. Myc-eIF6-FL or Myc-eIF6-DC36 were transfected into HCT116 cells, and 24 h later, the cells were briefly serum-starved in 0.1% FBS for 4 h. For the in vitro kinase assay, Myc-eIF6-FL or Myc-eIF6-DC36 were immunoprecipitated and incubated in the presence or absence of recombinant GSK3. In vitro kinase reactions were then subjected to autoradiography and Western blotting using anti-Myc antibody. Each experiment was repeated three independent times. An asterisk indicates the presence of a slower-migrating species.
include substrates such as b-catenin and glycogen synthase that carry three and four GSK3-specific sites, respectively (61)(62)(63)(64)(65). Substitution of the potential priming site serine 243 with alanine resulted in a significant 40% reduction in phosphorylation compared with eIF6-FL (full-length and WT) (Fig. 2, E and F). This suggested that in the absence of the Ser-243 site, phosphorylation of the Ser-239, Ser-235, and Thr-231 sites are reduced. However, it indicated that the initiating phosphorylation at Ser-243 is required for robust phosphorylation of eIF6. It also suggested that the Ser-239 site could serve as a potential initiation site in the absence of the Ser-243 site. The significance of the other three sites was confirmed by substitution of the Thr-231, Ser-235, and Ser-239 sites with alanines (Fig. 2E). Mutation of these three residues greatly reduced phosphorylation of eIF6, suggesting that they are also the key sites of phosphorylation (Fig. 2E).
These results were further validated by analysis of the individual site mutants (Fig. S3A). The T231A mutation reduced phosphorylation by 40%, indicating that it is one of the sites of phosphorylation (Fig. S3, A and B), which was consistent with MS analysis. It also indicated that in the absence of the Thr-231 site, the other sites in the sequence are phosphorylated. Both the S235A and S239A mutations markedly reduced phosphorylation of eIF6. Altogether, the studies with the single-site, multisite, and deletion mutants and MS analysis indicate that the Ser-243, Ser-239, Ser-235, and Thr-231 sites are key sites of phosphorylation.
Further to determine whether phosphorylation of eIF6 is sequential, we used a classical approach of incubating the substrate peptide (C-terminal peptide of eIF6) with the kinase, followed by MS analysis (Fig. 3A). MS analysis revealed that the Ser-243 site is phosphorylated by GSK3. Doubly phosphorylated peptides were also detected with phosphorylation at both the Ser-243 and Ser-239 sites (Fig. 3A). These results strongly indicated that GSK3 sequentially phosphorylates eIF6 starting with the Ser-243 site followed by the Ser-239 site. Because of the technical limitations associated with MS, detection of peptides with triple or quadruple phosphorylation were found to be at or below baseline (Fig. 3A). There are several other serine and threonine residues in the C-tail; however, peptides with phosphorylation at other sites could not be detected. MS analysis only detected phosphorylation in the GSK3-specific motif, which highlighted the specificity of phosphorylation.
GSK3 often works in concert with priming kinases that phosphorylate the initiating/priming site in the sequence. It is currently unclear whether the initiating Ser-243 site is also recog-nized by another priming kinase in vivo. It is also unclear as to the likely effect of enhanced priming phosphorylation of the Ser-243 site on the subsequent sites in the sequence. We therefore synthesized a phosphopeptide with phosphorylation at the Ser-243 site (Fig. 3B) and subjected it to kinase assay. MS analysis revealed that GSK3b phosphorylated the phosphopeptide at subsequent Ser-239 and Ser-235 sites ( Fig. 3B and Fig. S3C) sequentially. Detection of quadruple phosphorylation was at or below baseline (Fig. 3B). However, the Thr-231 site was identified earlier in our MS analysis of immunoprecipitated eIF6 (Fig.  2C). Interestingly, for both the primed and nonprimed peptide analysis, we were unable to detect any doubly phosphorylated Ser-235/Ser-243 or Thr-231/Ser-243 peptides or singly phosphorylated peptides with just the Ser-239, Ser-235, or Thr-231 phospho sites or peptides with phosphorylation at other serines and threonines in the C terminus. This clearly indicated that phosphorylation is specific and only occurs in a sequence. These MS results in combination with the analyses of site-specific mutants strongly indicate that the phosphorylation of eIF6 by GSK3b occurs in a sequence starting with the Ser-243 site, followed by the Ser-239, Ser-235, and Thr-231 sites.

Cellular phosphorylation of eIF6
Because GSK3 is activated under serum-starved conditions, we also determined whether eIF6 is phosphorylated in serumstarved 293T cells by MS. Nano-LC-MS/MS analysis revealed that the C-terminal tail of eIF6 is phosphorylated and that the sites of phosphorylation mapped to the Ser-243 site (Fig. S4A) or the Ser-239 site (Fig. S4B). As mentioned before, these phospho sites have been detected on endogenous eIF6 by several other phosphoproteomic studies.
We next tested whether GSK3b interacts with eIF6 under serum-starved conditions. GST-GSK3b interacted with endogenous eIF6 in serum-starved cells (Fig. 4A). The interaction was found to be specific to WT-GST-GSK3b because no such interaction was observed with either the empty vector or with a priming-site recognition mutant of GSK3 (GST-GSK3b-R96A) (Fig. 4, A and B). These results clearly indicate that there is a cellular interaction between GSK3b and eIF6.

Altered subcellular localization of eIF6 in response to starvation is regulated by GSK3
We then determined the effect of GSK3-dependent phosphorylation on eIF6 levels and/or localization. We observed Figure 2. Sequential and multisite phosphorylation of eIF6 by GSK3. A, the deletion mutant lacking just motif 2 (eIF6-DC20) was generated by substitution of Glu-226 with a stop codon. Myc-eIF6-FL or Myc-eIF6-DC20 was immunoprecipitated from serum-starved HCT116 cells, and the in vitro kinase reactions were assayed by autoradiography (autorad.) and Western blotting. Each experiment was repeated three independent times. B, sequence logo of the 23 residues in the C-terminal tail of human eIF6 generated by BLASTp (Geneious) alignment of human eIF6 C-tail with 50 species of vertebrates. Letter height corresponds to the degree of sequence conservation, and the percentage of identity is color-coded, with green representing 100% identity and brown representing 30-99% identity. C, MS analysis was carried out on eIF6 that was immunoprecipitated from 293T cells serum-starved for 4 h and incubated with or without recombinant GSK3. The samples were excised from Coomassie-stained polyacrylamide gels and analyzed by nano-LC-MS/MS analysis. The plot shows the MS/MS spectrum along with matched fragmentation mass table showing specific Thr-231 phosphorylation in the region of interest. D, the GSK3-specific sequential sites of phosphorylation are indicated in the C-terminal tail of eIF6. An arrow indicates the putative priming site that is primed by GSK3 itself or by another unidentified priming kinase. E, Myc-eIF6-FL or Myc-tagged phospho-site mutants were immunoprecipitated from HCT116 cells that were serumstarved for 4 h. In vitro kinase reactions were carried out for a shorter duration (25 min) in the presence and absence of recombinant GSK3b and were analyzed by autoradiography and by Western blotting using anti-Myc antibody (n = 3). F, graph represents the percentage of change in the phosphorylated levels of the eIF6-S243A phospho-site mutant relative to eIF6-FL as determined from the autoradiographs indicated in E. The values represent the S.E. of three independent experiments. Asterisks indicate significantly different as determined by an unpaired two-tailed t test: p , 0.0001. that the total levels of eIF6 were not altered by overexpression of GST-GSK3b, suggesting that GSK3 may not regulate the total eIF6 protein levels (Fig. 4, A and B). This was also confirmed by probing for eIF6 protein levels in response to long-term serum starvation when GSK3 is fully active. We did not detect significant changes in total eIF6 protein levels even after 18 h of starvation, suggesting that GSK3 does not regulate the stability or synthesis of eIF6 protein (Fig. 4, C and D). . Sequential phosphorylation of specific residues in the C terminus of eIF6. A, for nano-LC-MS/MS analysis, a C-terminal eIF6 peptide (23 amino acids) was synthesized (Aapptec) and incubated with recombinant GSK3b for 20 min. Extracted ion chromatograms show relative abundance levels of singly (precharged) and multiple residues phosphorylated by GSK3b in the region of interest. B, for nano-LC-MS/MS analysis, an eIF6 phosphopeptide (23 amino acids) carrying a phosphate at Ser-243, the priming site, was synthesized (Aapptec) and incubated with recombinant GSK3b for 20 min. Extracted ion chromatograms show relative abundance levels of singly (precharged) and multiple residues phosphorylated by GSK3b in the region of interest.
Because the total levels of eIF6 were unaltered, we next tested whether GSK3 regulates eIF6 function by altering its subcellular localization in response to serum starvation. Interestingly, immunofluorescence staining of endogenous eIF6 showed that eIF6 was predominantly localized to the cytoplasm in response to serum starvation in HeLa and HCT116 cells (Fig. 5, A-C, and Fig. S5, A and B). Serum starvation caused cytoplasmic (punctate) accumulation of eIF6, which was reversed by the readdition of serum, indicating that the response is prompt (Fig. 5A). These results were also confirmed by using a different mAb specific for eIF6, which ruled out nonspecific staining associated with the antibody (Fig. 5B). To further determine whether this altered subcellular localization depended on GSK3 activity, HeLa cells were treated with the GSK3 inhibitor LiCl. Treatment with the GSK3 inhibitor rescued eIF6 localization to the nucleus (Fig. 5C). Because LiCl could have other cellular targets, we also used CHIR99021, a highly selective GSK3specific inhibitor, which also rescued eIF6 localization to the nucleus (Fig. S5B) (66,67). The potency of the GSK3-specific inhibitor was confirmed by probing for a loss of phosphorylation of b-catenin at GSK3-specific sites (Ser-33/Ser-37/Thr-41) (61)(62)(63)68), which was also validated with SB415286 treatmentanother GSK3-specific inhibitor (Fig. S5, C and D). Inhibitor treatment resulted in greater than 50% loss in b-catenin phosphorylation, suggesting that GSK3 is inhibited under these conditions (Fig. S5, C and D). Similar effects on localization were observed for a normal rat intestinal epithelial cell line (RIE-1) treated with the GSK3-specific inhibitor. In RIE-1 cells, eIF6 shows predominant cytoplasmic accumulation in response to serum starvation as shown by fewer nuclei (pink) co-stained with eIF6 antibody, which is reversed by inhibition of GSK3 (Fig. S6A).
We further quantitated eIF6 levels in the nuclear and cytoplasmic fractions by subcellular fractionation. The purity of different fractions was confirmed by probing for topoisomerase IIb, a nuclear marker, and a-tubulin, a cytoplasmic marker. eIF6 was distributed almost equally between the nuclear and cytoplasmic fractions in serum-fed controls (53% nucleus and 47% cytoplasm), and no significant difference was found between the two fractions (Fig. 6, A and B). Similar to the imaging results, we found that in the serum-starved cells, there was a significant decrease in eIF6 levels in the nucleus with a corresponding increase in the cytoplasm (;32% in nucleus and ;68% in cytoplasm) (Fig. 6, A, C, D, and E). Inhibition of GSK3 completely restored the nuclear-cytoplasmic distribution as seen in the CHIR990021-treated cells (54% nucleus, 46% cytoplasm) (Fig. 6, D and F).
To determine the significance of phosphorylation in regulating the subcellular localization of eIF6 in response to starvation, we generated Myc-tagged phosphodead mutant of eIF6 (eIF6-4A) in which all four sites of phosphorylation (Ser-243, Ser-239, Ser-239, and Thr-231) were substituted with alanines. We also generated the Myc-tagged eIF6-DC mutant that lacked the last 20 amino acid residues in the C terminus, including the  4). b-Actin and a-tubulin were used as the loading control. C, lysates of HCT116 cells were collected just prior to serum starvation at 0 h and again at 4, 8, and 18 h of starvation. The samples were analyzed by Western blotting, and the blots were probed with anti-eIF6 and anti-a-tubulin antibodies (loading control) (n = 3). D, total eIF6 protein levels were quantitated using the blots represented in C. The values represent the S.E. of three independent experiments. All time points were compared with the 0-h time point, and no significant differences were found, as determined by one-way analysis of variance (Dunnett's multiple comparison test).
GSK3-specific sites of phosphorylation. To analyze the effect of Myc-tagged WT and mutants in the absence of endogenous eIF6, they were either expressed transiently (Fig. 6, G and H) or stably (Fig. S6, B-D) in a cell line that was stably knocked down for eIF6 (eIF6-KD) (Fig. S6, B and C). Endogenous eIF6 levels were reduced to greater than 85% by using shRNA targeted against eIF6 (Fig. S6, B and C). Analysis of the subcellular localization of the phosphodead mutants revealed that eIF6-4A and eIF6-DC do not accumulate in the cytoplasm in response to starvation unlike the WT cells (Fig. 6, G-I, and Fig. S7, A-C). The total input levels of eIF6 under all conditions tested are shown in Fig. S7 (D-F). These results strongly suggest that eIF6 exhibits predominant cytoplasmic localization in response to serum starvation, and this altered localization is regulated by GSK3-dependent phosphorylation of the C-terminal tail. Figure 5. Altered subcellular localization of eIF6 in response to serum starvation is regulated by GSK3. A, HeLa cells were serum-fed or serum-starved for 24 h. For refeeding, the cells were fed with 10% FBS for 2 h. The cells were fixed and stained for endogenous eIF6 (red) using anti-eIF6 mAb (Cell Signaling Technology), and the nuclei were stained with DAPI (blue) and analyzed by immunofluorescence microscopy. Each experiment was repeated three independent times. Scale bar, 50 mm. B, HCT116 cells were fed with 10% FBS or starved in 0.1% FBS for 24 h. The cells were fixed and stained for endogenous eIF6 (red) using anti-eIF6 mAb (Santa Cruz Biotechnology), and the nuclei were stained with DAPI (blue) and analyzed by immunofluorescence microscopy (n = 3). Scale bar, 50 mm. C, HeLa cells were serum-starved in 0.1% FBS for 24 h and treated with either vehicle (ultrapure water) or 20 mM LiCl for 3 h and fixed and stained for endogenous eIF6 (red) using anti-eIF6 mAb (Cell Signaling Technology), and the nuclei were stained with DAPI (blue) and analyzed by immunofluorescence microscopy (n = 3). Scale bar, 50 mm.

Expression of the phosphodead mutant decreases translation rates and stalls cell growth
We performed the polysome profile assay to determine the physiological effect of disrupting the GSK3-specific sites of phosphorylation on translation rates. Polysome profiles of starved (WT control) cells compared with the serum-fed cells showed that serum starvation causes a profound inhibition of translation, as shown by an increase in the levels of free 60S, 40S, and 80S subunits with a concomitant decrease in the heavier polysome levels (Fig. S8), and this was consistent with previ-ous reports (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33). Interestingly, in response to starvation, cells stably expressing the phosphodead eIF6-4A mutant and eIF6-DC mutant in the eIF6-KD cell line, displayed a significant 20-25% increase in the steady-state levels of free 60S subunits compared with eIF6-WT (Fig. 7, A-D). A consistent increase in the levels of 80S monosomes and a further reduction in heavier polysome levels that were already quite reduced because of serum starvation was also observed for both mutants (Fig. 7, A  and C). These results suggest that translation is further inhibited in the mutant cells under starved conditions. This was also Figure 6. The phosphodead mutant does not exhibit predominant cytoplasmic accumulation in response to serum starvation. A-C, for nuclear (Nuc) and cytoplasmic (Cyto) extraction, HCT116 cells were serum-fed (lanes 1 and 2) or serum-starved for 24 h (lanes 3 and 4) followed by subcellular fractionation. The samples were analyzed by Western blotting. The blots were probed with anti-eIF6, anti-a-tubulin (cytoplasmic marker), and anti-topoisomerase IIb antibodies (nuclear marker) (A). The blots in A were quantitated, and the values represent the S.E. of four independent experiments (B and C). The percentages of nuclear and cytoplasmic fractions relative to the total (sum of nuclear and cytoplasmic eIF6 levels) are plotted. The differences in eIF6 levels between the nuclear and cytoplasmic fractions for nonstarved controls (B) were not significant as determined by an unpaired two-tailed t test but were found to be significant for the starved sample (C), and asterisks indicate p = 0.0015. D-F, HCT116 cells were serum-starved for 24 h and treated with vehicle (lanes 1 and 2) or 10 mM CHIR99021 (lanes 3 and 4) (5 h) followed by subcellular fractionation. The samples were analyzed by Western blotting, and the blots were probed with the indicated antibodies (D). The blots in D were quantitated, and the values represent the S.E. of three independent experiments (E and F). The percentages of nuclear and cytoplasmic fractions relative to the total (sum of nuclear and cytoplasmic eIF6 levels) are plotted. The difference in eIF6 levels between the nuclear and cytoplasmic fractions for the starved-vehicle control (E) was found to be significant as determined by an unpaired two-tailed t test, and the asterisk indicates p = 0.015 and not significant for the CHIR99021-treated fractions (F). G-I, Myc-tagged eIF6-WT or eIF6-4A mutant were transiently expressed in HCT116 cells stably knocked down for eIF6 (eIF6-KD), and the cells were serum-starved for 24 h followed by subcellular fractionation. The samples were analyzed by Western blotting, and the blots were probed with the indicated antibodies (G). The blots represented in G were quantitated, and the values represent the S.E. of three independent experiments (H and I). The percentage of nuclear and cytoplasmic fractions relative to the total are plotted. The difference in Myc-eIF6 levels between the nuclear and cytoplasmic fractions for eIF6-WT (H) was found to be significant as determined by an unpaired two-tailed t test, and the asterisks indicate p = 0.0052 and not significant for eIF6-4A (I).
confirmed by measuring the incorporation rate of L-azidohomoalanine (AHA), an analog of methionine. Compared with eIF6-WT, a significant reduction in AHA incorporation (;25%) was observed for the eIF6-4A mutant relative to eIF6-WT under starvation conditions, suggesting that the translation rate is much more impaired in these cells (Fig. 7E).
We then determined whether the altered translation profile in the eIF6-4A mutant affected cell growth in response to Figure 7. Cells expressing the phosphodead mutant display reduced rate of protein synthesis and stalled cell growth. A and B, representative polysome profile of serum-starved HCT116 (eIF6-KD) cells stably expressing Myc-eIF6-4A mutant was overlaid with the profile of cells stably expressing Myc-eIF6-WT. The cells were serum-starved for 24 h prior to analysis (A). AUC of 60S peak for eIF6-4A mutant normalized to AUC of the 60S peak of eIF6-WT is indicated in the bar graph (B). The graph represents the S.E. of three independent experiments. Asterisks indicate significant differences between eIF6-WT and eIF6-4A with p = 0.0035 as determined by an unpaired two-tailed t test. Arrows indicate differences in 60S, 80S, and polysome peaks. C and D, representative polysome profile of HCT116 (eIF6-KD) cells stably expressing Myc-eIF6-DC mutant was overlaid with the profile of cells stably expressing Myc-eIF6-WT. The cells were serum-starved for 24 h prior to analysis. Profiles are representative of three independent experiments (C). The AUC of 60S peak for eIF6-DC mutant normalized to AUC of the 60S peak of eIF6-WT is indicated in the bar graph (D). The graph represents the S.E. of three independent experiments. Asterisks indicate significant differences between eIF6-WT and eIF6-DC with p = 0.0055 as determined by an unpaired two-tailed t test. E, AHA incorporation normalized to Hoechst 33342 (nuclear stain) fluorescence was measured for HCT116 (eIF6-KD) cells stably expressing either eIF6-WT or eIF6-4A. The bar graph represents the fold change in AHA incorporation relative to eIF6-WT. The data represent S.E. of three independent experiments, and asterisks indicate a significant difference with p = 0.003. F, bar graph represents the percentage of viable HCT116 (eIF6-KD) cells stably expressing eIF6-4A mutant relative to eIF6-WT. The cells were serumstarved for 24 h and subjected to trypan blue dye exclusion assay. The graph represents the S.E. of three independent experiments. Asterisks indicate significant differences between eIF6-4A and eIF6-WT with p , 0.0001 as determined by an unpaired two-tailed t test. G, plot represents the fold change in viable HCT116 (eIF6-KD) cells stably expressing eIF6-4A mutant relative to eIF6-WT. The cells were serum-starved for 24 h and subjected to MTS assay. The plot represents the S.E. of three independent experiments. Asterisk indicates significant differences between eIF6-4A and eIF6-WT with p = 0.0131 as determined by an unpaired two-tailed t test.
starvation. As shown in Fig. 7 (F and G), there was a significant 20-25% decrease in the number of viable eIF6-4A mutant cells compared with eIF6-WT. Analysis of the MTS assay further indicated that even in the absence of serum, cell proliferation was not fully inhibited for eIF6-WT but was completely stalled for the eIF6-4A mutant (Fig. 7G). These results indicate that the eIF6-4A mutant significantly impacts cell growth in response to starvation.

Discussion
The metabolic response to starvation is to conserve energy and limit processes with high-energy requirements such as translation and ribosome biogenesis (22)(23)(24)(25)(26)(27)(28)(29)(30)(31). Here, we show that the regulation of translation initiation factor eIF6 by GSK3 affects translation response to starvation. Despite its extensive role in metabolism, eIF6 function in starvation response is poorly understood. We show that eIF6 is regulated in response to starvation-induced stress and that the regulation hinges on the GSK3 signaling pathway that is prolifically active under such nutrient-deprived conditions. This study also identifies GSK3 as one of the kinases that phosphorylates the multiple sites in the C-terminal tail of eIF6. Although several global proteomic and biochemical studies have shown that the C-terminal tail of eIF6 is heavily phosphorylated, the identity of the kinases involved, other than PKCbII, were largely unknown. Our results show that GSK3 sequentially phosphorylates the C-terminal tail of eIF6 at Ser-243, Ser-239, Ser-235, and Thr-231 sites, with Ser-243 being the initiation site in the sequence. It is currently unclear whether GSK3 works in concert with a priming kinase in cells to enhance phosphorylation at the initiation site.
Interestingly, one of the sites of phosphorylation, the Ser-235 site, is also phosphorylated by the RACK1-PKCbII complex (12). Substitution of the serine 235 site with alanine blocks eIF6 release from 60S and inhibits active 80S formation (12). Apart from in vitro studies, the Ser-235 site has also been shown to be important for normal growth and translation in vivo (17,18). Studies in eIF6 1/2 mouse embryonic fibroblasts reconstituted with the S235A mutant show that the Ser-235 site is critical for eIF6 function in stimulating translation in response to insulin and phorbol esters (17)(18)(19). It is not uncommon for a single site to be regulated by multiple kinases in a context-dependent manner. Because PKCbII is active under conditions of growth and proliferation, whereas GSK3 is active under growthdeprived conditions, the two kinases could phosphorylate the Ser-235 site based on the polarizing cellular cues of growth and starvation or resting. However, in the case of eIF6 phosphorylation by GSK3, additional sites are also phosphorylated, which could further perturb the local protein structure and alter eIF6 localization.
We observed predominant cytoplasmic accumulation of eIF6 in response to serum starvation and inhibiting the GSK3dependent phosphorylation of eIF6 rescues the subcellular distribution. Because GSK3 is predominantly cytoplasmic with a small nuclear fraction, it is likely that phosphorylation by GSK3 leads to increased cytoplasmic retention of eIF6. In our study, the predominant cytoplasmic accumulation of eIF6 in response to starvation was consistently observed among different cell lines including the normal RIE-1, HCT116, and HeLa cells. A previous study showed that the nuclear import of eIF6 was mediated by calcium-activated calcineurin phosphatase, whereas the nuclear export was mediated by phosphorylation of eIF6 at Ser-174/175 residues by nuclear casein kinase 1 in COS-7 cells (69). Thus, our results and previous data suggest that altering the subcellular localization of eIF6 may be a predominant mode of regulating eIF6 function, and the underlying signaling mechanisms may vary based on the growth stimulus and stressed states. However, it is yet to be understood whether this change in localization is due to altered association with 60S subunit or due to altered association with a nuclear-export factor.
We and others (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33) have shown that in response to nutrient deprivation, protein synthesis is inhibited with a decrease in polysome levels along with a corresponding increase in free 60S subunits and inactive 80S monosomes. Even mild changes to the 60S or 40S levels can drastically impact cell growth as seen in ribosomopathies. Our results indicate that phosphorylation of the C-tail of eIF6 and its cytoplasmic accumulation is important to maintain a basal level of protein synthesis under starved conditions. It is unclear as to how eIF6 function in 60S maturation is altered in response to starvation and whether the effects on 60S availability are direct or indirect. eIF6 is known to work with upstream regulators of 60S maturation, such as Shwachman-Bodian-Diamond Syndrome factor and the GTPase EFL1 (elongation factor-like 1) complex, and RPL10 that facilitate eIF6 release, and it remains to be determined how such release is affected in response to stress (4,11,13,20,70). However, these results indicate that any disruption of GSK3dependent phosphorylation of eIF6 can shift the balance from adaptation to enhanced sensitivity to starvation. Analysis of eIF6 phosphorylation in the free eIF6 pool that is unbound to 60S relative to those that are bound to 60S will help to understand the potential effects of phosphorylation on eIF6 release.
Based on these results, we propose the following model, in which in response to nutrient or growth-limiting conditions, AKT signaling is inactivated, which in turn activates GSK3, and GSK3 phosphorylates eIF6 at multiple sequential sites. Phosphorylation by GSK3 enhances cytoplasmic accumulation of eIF6. The timely release of eIF6 from the 60S in the late stages of maturation is critical for the 60S and 40S subunits to associate and initiate translation. Phosphorylation at the Ser-235 site has been shown to facilitate eIF6 release under fed conditions. Because this site and other sequential sites are phosphorylated under starvation conditions by GSK3, it is likely that eIF6 release from the 60S permits 60S association with the limited pool of mRNA-bound 40S to engage in translation even under starvation conditions. This regulation ensures that a basal rate of translation continues during starvation to maintain slow growth, which is essential to adapt to starvation and to permit recovery upon nutrient addition. This model is based on the results that in the absence of eIF6 phosphorylation, the rates of translation are further decreased under serum starvation conditions, and there is an increase in free 60S pool that is not engaged in translation. Future studies will assess the role of phosphorylation in altering the dynamics of eIF6 and 60S interaction and determine the mechanism of altered subcellular localization.

Experimental procedures
Cell culture and transfections

Plasmid constructions
To clone human eIF6, mRNA was extracted from MCF7 (ATCC) cells using RNAaqueous total RNA isolation kit (Invitrogen), and cDNA was synthesized using the Superscript firststrand cDNA synthesis kit (Invitrogen). Using PCR, eIF6 ORF was cloned into the pCMV-Myc plasmid (Clontech) using the respective forward and reverse primers carrying EcoRI and XhoI restriction sites: 59-TATAAGAATTCTAATGGCG-GTCCGAGCTTCGTTCGAGAAC-39 and 59-TATAACTC-GAGTCAGGTGAGGCTGTCAATGAGGGAATC-3. GSTtagged GSK3b constructs were cloned into the pDEST TM 27 vector (Thermo Fisher Scientific) as described previously (71). All of the pCMV and pcDNA3.1-Myc-eIF6 phospho-site mutants were generated using the Q5 ® site-directed mutagenesis kit (New England Biolabs) using the forward and reverse primers Table S1. All plasmid constructs used in this study were verified by Sanger sequencing (Genewiz).

Generation of stable cell lines expressing eIF6-WT and mutants
HCT116 cells were transfected with pLKO.1 vector expressing either an eIF6 targeted shRNA (CCGGGTGCATCCCAA-GACTTCAATTCTCGAGAATTGAAGTCTTGGGATG-CACTTTTTG, Sigma, TRC327700) or a control nontargeting shRNA (Sigma) as described above. Stably transfected clones were selected in 2 mg/ml puromycin and maintained in 0.5 mg/ ml puromycin. ;30 clones were screened for eIF6 knockdown (eIF6-KD) by Western blotting. To express eIF6 WT and mutants in the eIF6-KD cell line, the cells were stably trans-fected with pcDNA3.1 plasmid expressing either the Myctagged eIF6-WT, eIF6-4A, eIF6-DC, or the empty vector. To evade shRNA targeting, mismatches were generated in the Myc-eIF6-WT, Myc-eIF6-4A, or Myc-eIF6-DC plasmids using site-directed mutagenesis. All stably transfected clones were selected in 500 mg/ml Geneticin and 0.5 mg/ml puromycin. ;25 clones were screened for expression of Myc-eIF6-WT or mutant by Western blotting. All stable lines were maintained in 100 mg/ml Geneticin and 0.5 mg/ml puromycin.
Cell viability assays 7,000 HCT116 cells expressing eIF6-WT or eIF6-4A were cultured per well of a 96-well plate in McCoy's culture medium without phenol red. After overnight incubation, the cells were serum-starved for 24 h in McCoy's medium supplemented with 0.1% FBS. 20 ml of MTS solution (CellTiter 96 ® Aqueous One solution reagent, Promega) was added to each well, and the plates were incubated for 1 h. Absorbances were read at 490 nm (Spectra Max i3x, Molecular Devices). To obtain background-corrected absorbances, average absorbance values of the control wells containing media and MTS only were subtracted from all other absorbances. For trypan blue assay, HCT116 cells expressing eIF6-WT or eIF6-4A were plated as indicated for polysome profile assays. The cells suspended in culture media were mixed with trypan blue dye at 1:1 ratio, and the numbers of blue-stained (dead) cells and unstained (viable) cells were counted using a hemocytometer.

Immunoprecipitation
For immunoprecipitation of Myc-eIF6, or Myc-empty vector, lysates with 1.2 mg of total protein were suspended in a final volume of 500 ml of MCLB buffer supplemented with inhibitors. The lysates were precleared with 20 ml of protein A/ G plus-conjugated agarose beads (Santa Cruz Biotechnology) and 1 ml of normal mouse IgG (Santa Cruz Biotechnology). Protein A/G plus-agarose beads were washed in MCLB buffer before use. For preclearing, the lysates were rotated for 20 min at 4°C. Precleared lysates were collected by centrifugation at 3500 rpm for 3 min at 4°C, and the beads were discarded. Precleared lysates were incubated with 20 ml of protein A/G plusagarose beads and 11 ml of Myc-tag antibody (9E10) (Santa Cruz Biotechnology) and rotated overnight at 4°C. The samples were washed three times in MCLB buffer, and the beads were collected by centrifugation at 3000 rpm for 3 min and washed for a fourth time in incomplete kinase buffer (50 mM Tris, pH 7.4, 10 mM MgCl 2 , 2 mM DTT), and after the final wash, the beads were suspended in 45 ml of incomplete kinase buffer. Immunoprecipitation of the Myc-tagged phospho-site mutants of eIF6, eIF6-DC20, and Myc-eIF6-WT was carried out as described above except that 1.5 mg of total protein was used for the assay. For the Myc-eIF6-DC36 mutant, because of a small reduction in its expression, 2 mg of total protein was used for the assay. This ensured that the total eIF6 protein input for the kinase assays were comparable between eIF6-WT and the eIF6-DC36 mutant.

Regulation of eIF6 by GSK3
In vitro kinase assay For the in vitro kinase assays, Myc-tagged versions of eIF6 were immunoprecipitated as indicated above, and the immunoprecipitated beads were incubated with or without 350 units of recombinant GSK3b (rabbit skeletal muscle) (New England BioLabs) or 0.7 unit/mg of recombinant human GSK3a (Abcam) and 13 hot kinase buffer (50 mM Tris, pH 7.4, 50 mM MgCl 2 , 10 mM DTT, 50 mM cold ATP, [g-32 P]ATP) and incubated at 30°C for 30 min. For kinase reactions with pure eIF6 protein, codon-optimized recombinant human His 6 -eIF6 was cloned into RSF-Duet vector and purified from E. coli as described previously, except purification using a nickel-nitrilotriacetic acid column was followed by negative selection using Heparin column (72). 2.5 mM recombinant-eIF6 protein was incubated with the kinase as described above. Kinase reactions with the phospho-site mutants of eIF6 and eIF6-WT control were set up as detailed above, except that the reactions were incubated at 30°C for shorter duration of 25 min to capture time-dependent kinetic differences. Kinase reactions were analyzed by SDS-PAGE and exposed to X-ray film for autoradiography. For Western blotting of the kinase assay, duplicate kinase reactions were simultaneously set up as described above except that the reactions were set-up using the cold kinase buffer (50 mM Tris, pH 7.4, 50 mM MgCl 2 , 10 mM DTT, 250 mM cold ATP). The blots were probed with anti-eIF6 mAb (1:1000) or anti-Myc-tag mAb (1:750) and anti-GSK3b mAb (1:1000) (Cell Signaling Technology) diluted in TBS with 0.1% Tween 20 (TBS-T).

Pulldown assay
To capture interactions with the endogenous eIF6, HCT116 cells were transfected with GST-GSK3b or empty vector as described above. Cell lysis was carried out in MCLB buffer, and a small fraction was saved for analysis of input levels. For the pulldown assays, 1.1 mg of total protein suspended in 500 ml of MCLB buffer and supplemented with inhibitors was incubated with 50 ml of GSH-agarose resin beads (Gold Biotechnology) and rotated overnight at 4°C. The next day, the samples were washed three times in MCLB buffer supplemented with inhibitors and resuspended in (30 ml) of MCLB buffer and 23 Laemmli buffer. The samples were then analyzed by Western blotting. For the pulldown of GST-GSK3b-WT and GST-GSK3b-R96A, 293T cells were transfected with the respective vectors, and 1.9 mg of total protein was used for the pulldown analysis.

Mass spectrometry
For MS, 7 3 10 6 293T cells were plated per 10-cm plate and transfected as indicated before. 24 h later, the cells were serumstarved in Dulbecco's modified Eagle's medium containing 0.1% FBS for 3 h. The cells were lysed in MCLB buffer supplemented with inhibitors. Immunoprecipitations were carried out as previously described using 3 mg of total protein. For the kinase reactions, 15 ml of Myc-eIF6-bound beads were incubated with/without recombinant GSK3b (NEB) and suspended in cold kinase buffer and incubated at 30°C for 30 min. The reactions were washed twice in incomplete kinase buffer and resuspended in incomplete kinase buffer and 1.53 Laemmli buffer. The samples were boiled and analyzed by SDS-PAGE followed by Coomassie staining. Stained gels were sent to the University of Wisconsin-Madison mass spectrometry facility.

Enzymatic "in gel" digestion
In-gel digestion and mass spectrometric analysis was done at the mass spectrometry facility of the Biotechnology Center at the University of Wisconsin-Madison. In short, Coomassiestained gel pieces were destained twice for 5 min in MeOH: H 2 O:NH 4 HCO 3 (50%:50%:100 mM), dehydrated for 5 min in ACN:H 2 O:NH 4 HCO 3 (50%:50%:25 mM) and then once more for 30 s in 100% ACN, dried in a SpeedVac for 2 min, rehydrated completely and reduced in 25 mM DTT (in 25 mM NH 4 HCO 3 ) for 30 min at 56°C, alkylated by solution exchange with 55 mM iodoacetamide (in 25 mM NH 4 HCO 3) in the dark at room temperature for 30 min, washed once in 25 mM NH 4 HCO 3 , dehydrated twice for 5 min in ACN:H 2 O: NH 4 HCO 3 (50%:50%:25 mM) and then once more for 30 s in 100% ACN, dried in a SpeedVac again, and finally rehydrated with 20 ml of trypsin digestion solution (10 ng/ml trypsin (Promega) in 25 mM NH 4 HCO 3 and 0.01% ProteaseMAX w/v (Promega)). Additional 30 ml of rehydration solution (25 mM NH 4 HCO 3 and 0.01% ProteaseMAX w/v (Promega)) was added to facilitate complete rehydration with excess overlay needed for peptide extraction. The digestion was conducted for 3 h at 42°C. Peptides generated from digestion were transferred to a new tube and acidified with 2.5% TFA to 0.3% final concentration. Gel pieces were extracted further with ACN:H 2 O:TFA (70%:29.25%:0.75%) for 10 min, and the solutions were combined and then dried completely in a SpeedVac (;15 min). Extracted peptides were solubilized in 30 ml of 0.1% formic acid, degraded in ProteaseMAX, and removed via centrifugation (maximum speed, 10 min). Supernatant was taken for immediate solid phase extraction (ZipTip ® C18 pipette tips, Millipore) according to manufacturer's protocol. Peptides eluted off ZipTip C18 tips with ACN:H 2 O:TFA (70%:29.9%:0.1%) were dried and finally solubilized in 8 ml of 0.1% formic acid.

Nano-LC-MS/MS
Peptides were analyzed by nano-LC-MS/MS using the Agilent 1100 nanoflow system (Agilent) connected to a hybrid linear ion trap-Orbitrap TM mass spectrometer (LTQ-Orbitrap Elite TM , Thermo Fisher Scientific) equipped with an EASY-Spray TM electrospray source. Chromatography of peptides prior to mass spectral analysis was accomplished using a capillary emitter column (PepMap ® C18, 3 mM, 100 Å, 150 3 0.075 mm, Thermo Fisher Scientific) onto which 3 ml of extracted peptides was automatically loaded. The nano-HPLC system delivered solvents A: 0.1% (v/v) formic acid, and B: 99.9% (v/v) acetonitrile, 0.1% (v/v) formic acid at 0.50 ml/min to load the peptides (over a 30-min period) and 0.3 ml/min to elute peptides directly into the nano-electrospray with gradual gradient from 3% (v/v) B to 20% (v/v) B over 17 min and concluded with 5 min fast gradient from 20% (v/v) B to 50% (v/v) B, at which time a 5 min flush-out from 50-95% (v/v) B took place. As peptides eluted from the HPLC-column/electrospray source survey, MS scans were acquired in the Orbitrap with a resolution of 120,000 followed by MS2 fragmentation of 20 most intense peptides detected in the MS1 scan from 350 to 1800 m/z; redundancy was limited by dynamic exclusion.

Data analysis
Raw files were searched against Uniprot human amino acid sequence database (downloaded on October 31, 2016) containing a list of common contaminants (67,042 total entries) using Sequest HT search engine within Proteome Discoverer suite (version 2.2.0.388). Full trypsin specificity, two missed cleavages, 15 ppm for precursor, and 0.6 Da for fragment mass tolerance plus fixed carbamidomethylation (Cys) and dynamic oxidation (Met), deamidation (Asn/Gln), and phosphorylation (Ser/ Thr) were selected in the search parameters. A strict confidence threshold of 0.01 false discovery rate in decoy database searching on protein and peptide levels were applied. In addition, at least two peptides per protein plus 0.01 q value high confidence threshold was selected to affirm proper identification.
MS of synthetic eIF6 phosphopeptide and nonphosphorylated peptide 100 mM of the synthesized phosphopeptides (Aapptec) or nonphosphorylated peptides were incubated with 350 units of recombinant GSK3b (New England Biolabs) and cold kinase buffer. The reactions were incubated for 20 min at 30°C. The reactions were then quenched by 40 mM EDTA and by freezing.

Nano-LC-MS/MS
100 mM of synthesized version of human eIF6 phosphopeptide was acidified with 2.5% TFA to 0.4% final and solid phase extracted (ZipTip C18 pipette tips, Millipore) according to the manufacturer's protocol. Peptide was eluted off ZipTip C18 tip with 2 ml of ACN:H 2 O:TFA (70%:29.9%:0.1%) and diluted to 20 ml of final volume with 0.1% formic acid. Mass spectrometric analysis followed using the Agilent 1100 nanoflow system (Agilent) connected to a hybrid linear ion trap-Orbitrap TM mass spectrometer (LTQ-Orbitrap Elite TM , Thermo Fisher Scientific) equipped with an EASY-Spray TM electrospray source. Chromatography of peptides prior to mass spectral analysis was accomplished using capillary emitter column (PepMap C18, 3 mM, 100 Å, 150 3 0.075 mm, Thermo Fisher Scientific) onto which 2 ml of extracted peptides was automatically loaded. Nano-HPLC system delivered solvents A: 0.1% (v/v) formic acid, and B: 99.9% (v/v) acetonitrile, 0.1% (v/v) formic acid at 0.50 ml/min to load the peptides (over a 30-min period) and 0.3 ml/min to elute peptides directly into the nano-electrospray with gradual gradient from 3% (v/v) B to 20% (v/v) B over 17 min and concluded with 5 min of fast gradient from 20% (v/v) B to 50% (v/v) B, at which time a 4-min flush-out from 50-95% (v/v) B took place. As peptides eluted from the HPLC-column/ electrospray source survey, MS scans were acquired in the Orbitrap with a resolution of 120,000 followed by MS2 fragmentation of 20 most intense peptides detected in the MS1 scan from 350 to 1800 m/z; redundancy was limited by dynamic exclusion. Raw files were searched against targeted amino acid sequence database containing a list of common laboratory contaminants plus a sequence of synthetic version of human eIF6 (202 total entries) using Sequest HT search engine within Proteome Discoverer suite (version 2.2.0.388). No-enzyme specificity, 0 missed cleavages, 15 ppm for precursor, and 0.6 Da for fragment mass tolerance plus dynamic oxidation (Met), deamidation (Asn/Gln), and phosphorylation (Ser/Thr) were selected in the search parameters. Confidence threshold of 0.02 false discovery rate in decoy database searching on protein and peptide levels were applied. Dynamic phosphorylation cascade distributions were manually interrogated for precursor ion abundance and quality of MS/MS fragmentation.

MS of serum-starved samples
For cellular MS analysis, Myc-eIF6 was immunoprecipitated from 293T cells that were serum-starved for 24 h. Immunoprecipitations were carried out as previously described except that 6 mg of total protein was used for analysis, and both a faster and slower migrating species of eIF6 was excised from Coomassie-stained gel and subjected to MS analysis as described above. Phosphopeptides were only detected in the slower migrating/ super-shifted band. Table S2 summarizes the list of phosphopeptides indicated in this study.
Immunofluorescence staining and microscopy 50,000 HCT116 cells were plated in each well of a 12-well plate carrying poly-D-lysine-coated glass coverslips (Neuvitro). After overnight incubation, the cells were either serum-fed with McCoy's 5A medium containing 10% FBS or serum-starved in medium containing 0.1% FBS, and 24 h later, the cells were treated with either 0.1% DMSO (vehicle) or 10 mM CHIR99021 (Tocris). The cells were fixed overnight in 2% paraformaldehyde/PBS. For immunostaining, fixed cells were permeabilized with 2% Triton X-100/PBS for 20 min and incubated in blocking buffer (2% BSA, 0.1% Igepal, PBS) for 30 min. The following primary antibodies diluted in the blocking buffer were used for staining: anti-eIF6 mAb (D16E9) (1:300) (Cell Signaling Technology) or anti-eIF6 mAb (1:350) (Santa Cruz Biotechnology), and incubated in a humidified chamber at room temperature for 1 h. Coverslips were washed with 0.1% Igepal/PBS and incubated with Cy3-conjugated Affinipure donkey anti-rabbit IgG secondary antibody (Jackson ImmunoResearch) diluted in blocking buffer (1:300) at room temperature for 1 h in the dark. Coverslips were washed in 0.1% Igepal/PBS and mounted using ProLong Gold antifade reagent with 49,6-diamidino-2-phenylindole (DAPI) (Invitrogen). Immunofluorescence studies in HeLa cells were carried out as described above. Staining of normal rat intestinal epithelial cells (RIE-1) was carried out as described above except that 30,000 cells were plated per well, and the cells were probed with anti-eIF6 mAb (D16E9) (1:200) (Cell Signaling Technology). RIE-1 cells were cultured and maintained as described previously (73,74). For imaging, the slides were analyzed using the Leica DM6 B upright fluorescent microscope. The images were acquired using Leica DFC 3000G (Bin 1 3 1, Gamma1) camera and processed by Leica LAS X imaging software.

Subcellular fractionation
700,000 HCT116 cells were plated per 100-mm dish, and 24 h later, the cells were serum-starved in medium containing 0.1% FBS or fed with medium containing 10% FBS for an additional 24 h. To assay the effect of inhibitors, serum-starved cells were treated with either 0.1% DMSO (vehicle) or 10 mM CHIR99021 (Tocris). For subcellular fractionation of WT and mutants, 1.2 million cells were plated per 60-mm dish. To assay for total protein input, the cells were lysed in MCLB buffer supplemented with inhibitors as previously described. To extract cytoplasmic fractions, the cells were collected in cell lysis buffer (10 mM HEPES, pH 7.5, 10 mM KCl, 0.1 mM EDTA, 0.5% Igepal) supplemented with the following inhibitors: 10 mM b-glycerophosphate, 1 mM sodium vanadate, 2 mM DTT, 13 protease inhibitor mixture (Sigma-Aldrich), 13 phosphatase inhibitor mixture (Santa Cruz Biotechnology), and 0.5 mM PMSF. Nuclear-cytoplasmic fractionation was carried out as described before (75). For cytoplasmic fractionation, the lysates were incubated on ice for 15 min with intermittent gentle mixing and vortexed on high for 10 s at 14000 rpm. The cells were centrifuged at 12,000 3 g for 10 min at 4°C, and cytoplasmic lysates were collected. For nuclear extraction, nuclear pellets were washed four times in cell lysis buffer and centrifuged at 3000 rpm for 5 min at 4°C. Nuclear pellets were suspended in 75 ml of nuclear extract buffer (20 mM HEPES, pH 7.5, 400 mM sodium chloride, 1 mM EDTA) supplemented with the following inhibitors: 10 mM b-glycerophosphate, 1 mM sodium vanadate, 2 mM DTT, 1 3 protease inhibitor mixture (Sigma-Aldrich), 13 phosphatase inhibitor mixture (Santa Cruz Biotechnology), and 0.5 mM PMSF. Nuclear pellets were solubilized on ice for 30 min followed by centrifugation at 12,000 3 g for 15 min at 4°C. Both nuclear and cytoplasmic fractions were analyzed by Western blotting.

AHA translation assay
To measure the rate of AHA incorporation, the Click-iT AHA kit was utilized, and the assay was performed as per the manufacturer's instructions (Life Technologies) as reported before (76). 15,000 cells were plated per well of a 96-well plate, and after overnight incubation they were washed twice in 13 PBS. The cells were serum-starved in medium containing 0.1% FBS. 24 h later, the cells were incubated in methionine-free medium containing 0.1% FBS and Click-iT AHA reagent (50 mM) for 30 min at 37°C in a CO 2 incubator. The cells were fixed, and the azide-alkyne cycloaddition reaction was carried out using azide conjugated to Alexa Fluor 488, and nuclei were stained with Hoechst 33342. Fluorescence intensity was measured using a microplate reader.

Data availability
The MS data have been uploaded to MassIVE under accession number MSV000085596. All other data have been included in the article. intellectual input and for providing initial resources for the work and Nathaniel Letcher and Jenna Fee for technical help. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.