Phosphorylation screening identifies translational initiation factor 4GII as an intracellular target of Ca(2+)/calmodulin-dependent protein kinase I.

CaMKI is a Ca2+/calmodulin-dependent protein kinase that is widely expressed in eukaryotic cells and tissues but for which few, if any, physiological substrates are known. We screened a human lung cDNA expression library for potential CaMKI substrates by solid phase in situ phosphorylation ("phosphorylation screening"). Multiple overlapping partial length cDNAs encoding three proteins were detected. Two of these proteins are known: 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatase and eukaryotic translation initiation factor (eIF) 4GII. To determine whether CaMKI substrates identified by phosphorylation screening represent authentic physiological targets, we examined the potential for [Ca2+]i- and CaMKI-dependent phosphorylation of eIF4GII in vitro and in vivo. Endogenous eIF4GII immunoprecipitated from HEK293T cells was phosphorylated by CaMKI, in vitro as was a recombinant fragment of eIF4GII encompassing the central and C-terminal regions. The latter phosphorylation occurred with favorable kinetics (Km = 1 microm; kcat = 1.8 s-1) at a single site, Ser1156, located in a segment of eIF4GII aligning with the phosphoregion of eIF4GI. Phosphopeptide mapping and back phosphorylation experiments revealed [Ca2+]i-dependent, CaMKI site-specific, eIF4GII phosphorylation in vivo. This phosphorylation was blocked by kinase-negative CaMKI consistent with a requirement for endogenous CaMKI for in vivo eIF4GII phosphorylation. We conclude that phosphorylation screening is an effective method for searching for intracellular targets of CaMKI and may have identified a new role of Ca2+ signaling to the translation apparatus.


eukaryotic cells (reviewed in
). Because such effects regulate cellular events critical to cell survival, differentiation, and growth, identification of the corresponding intracellular targets of CaM kinases is of paramount importance. In those cases in which the respective CaM kinase is dedicated to a single physiological substrate, such identification has been relatively straightforward and has revealed mechanistic details of the Ca 2ϩ control of intermediary metabolism (phosphorylase kinase), cytoskeletal dynamics (myosin kinases), and peptide chain elongation (elongation factor-2 kinase). There has also been considerable progress recently in identification of targets of the broad specificity CaM kinases, CaMKII and CaMKIV, involved in the molecular events underlying synaptic plasticity and transcriptional activation (reviewed in Refs. [3][4][5][6]. In counterpoint has been the notable lack of progress in identifying the intracellular targets of CaMKI, despite its identification as a distinct CaMK contemporaneously with CaMKII (7). CaMKI has a substrate recognition motif defined using synthetic peptides that is highly similar (although not identical) to those of CaMKs II and IV (8 -11) and has been demonstrated to be capable of phosphorylation of a relatively small number of protein substrates in in vitro assays (12)(13)(14)(15)(16)(17)(18)(19). It is unclear, however, whether these in vitro substrates represent the authentic physiological targets of CaMKI.
Nonetheless, despite the lack of information regarding its physiological role(s), an important, perhaps essential, role for CaMKI in the regulation by [Ca 2ϩ ] i of eukaryotic cell physiology may be postulated based on circumstantial observations. At both the mRNA and protein levels CaMKI is expressed in all mammalian cell and tissue types examined to date (19 -21). CaMKI also exhibits a widespread phylogenetic distribution, with either obvious orthologous cDNAs or cross-immunoreactive bands detectable in species as diverse as fission yeast, Caenorhabditis elegans, Drosophila, and vertebrates (21)(22)(23). Finally, CaMKI is a downstream target of a signaling cascade initiated by elevation of [Ca 2ϩ ] i (24). This cascade is tightly regulated by [Ca 2ϩ ] i via three distinct mechanisms: (i) activating phosphorylation by two upstream CaMKI kinases, CaMKK␣ and ␤, which themselves are Ca 2ϩ /CaM-activated; (ii) binding of Ca 2ϩ /CaM to CaMKI promoting efficient phosphorylation of CaMKI by the CaMKKs; and (iii) direct allosteric activation of CaMKI by Ca 2ϩ /CaM (Refs. 25-27; additional references in Ref. 2). Whereas the rationale for such an elaborate scheme for regulation by [Ca 2ϩ ] i is not immediately apparent, it may be speculated that inappropriate (i.e. Ca 2ϩ /CaMindependent) expression of CaMKI activity may be deleterious to the cell, an idea consistent with the observation that expression of a hyperactive mutant of CaMKI in Schizosaccharomyces pombe resulted in cell cycle arrest (Ref. 23; similarly observed with the CaMKI/IV homologue from Aspergillus nidulans) (28).
To detect potential intracellular targets of CaMKI and thereby gain insight into its cellular role(s), we screened a cDNA expression library using solid phase in situ phosphorylation (the "phosphorylation screening" method developed by Fukunaga and Hunter (29)). Using this method, we identify as potential CaMKI targets, 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatase and eukaryotic translation initiation factor (eIF) 4GII (30). We further show that eIF4GII is an effective in vitro CaMKI substrate and also an in vivo target of [Ca 2ϩ ] i -dependent and CaMKI-catalyzed phosphorylation. These results demonstrate phosphorylation screening to be an effective method for identifying intracellular targets of protein kinases, generalizable to the CaMK family, and have potentially revealed a new role for [Ca 2ϩ ] i in the control of the initiation of mRNA translation.
Phosphorylation Screening of a Human Lung cDNA Expression Library-E. coli XL1-Blue cells were infected with a human lung cDNA library directionally constructed in TriplEx TM (Clontech) and plated at 1.2 ϫ 10 4 plaque-forming units/150-mm plate. After a 3-5-h growth period at 42°C, the plates were overlaid with isopropyl-␤-D-thiogalactopyranoside-saturated nitrocellulose filters and incubated at 37°C for an additional 6 -10 h. The filters were removed and immersed in 200 ml of blocking buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 3% BSA) and gently agitated for 60 min at room temperature (or overnight at 4°C). The filters were washed 4 ϫ 15 min in a Triton X-100 containing wash buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, 1 mM DTT, and 0.2 mM phenylmethylsulfonyl fluoride) followed by rinsing for 10 min in CaMKI reaction buffer (50 mM Tris-HCl, pH 7.5, 0.5 mM DTT, 10 mM MgCl 2 , 0.5 mg/ml BSA, 1 mM CaCl 2 ). The filters were then incubated for 60 min at room temperature in CaMKI reaction buffer containing 25 M unlabeled ATP and then rinsed for 10 min in this buffer without ATP. 2 GST-CaMKI 3 (2.0 M) was preincubated with 1.3 M MBP-CaMKK␤ for 20 min at room temperature in a buffer containing 50 mM Tris-HCl, pH 7.6, 0.5 mM DTT, 10 mM MgCl 2 , 0.5 mg/ml BSA, 1 mM CaCl 2 , 10 M CaM, and 200 M ATP. The CaMKI/ CaMKK ratio used was established in preliminary experiments to yield maximal CaMKI activation, determined as previously described (35) (data not shown). The activated CaMKI was diluted in CaMKI reaction buffer and [␥-32 P]ATP to achieve final concentrations in the complete CaMKI reaction mixture of 0.2 M GST-CaMKI, 0.13 M MBP-CaMKK␤, 50 mM Tris-HCl, pH 7.6, 0.5 mM DTT, 10 mM MgCl 2 , 0.5 mg/ml BSA, 1 mM CaCl 2 , 4 M CaM, and 25 M [␥-32 P]ATP (6 Ci/ml), and the expressed proteins were phosphorylated by incubation of filters in this mixture for 30 min at room temperature. The filters were then washed extensively in CaMKI wash buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 1 mM EGTA, 20 mM NaF, 10 mM NaPPi, and 0.1% Triton X-100). The filters were air-dried for 45 min and exposed to x-ray film for 12-48 h at Ϫ70°C. Secondary and tertiary screens were performed as above to obtain well isolated positive plaques. The purified TriplEx clones were converted to pTriplEx clones by Cre-Loxmediated subcloning following the manufacturer's instructions. The inserts were sized by restriction mapping and sequenced in either the 5Ј or 5Ј and 3Ј directions using an ABI automated DNA sequencer.
Transfection of HEK 293T Cells-The cells were plated at 5.6 ϫ 10 6 cells/100-mm dish and grown overnight to 80 -90% confluency in a 37°C and 5% CO 2 incubator in DMEM with 10% fetal bovine serum without antibiotics. The cells were transfected with plasmids using LipofectAMINE 2000 TM according to the standard protocol provided by the manufacturer. Transfection efficiency, determined using a LacZ reporter, was 70.0 Ϯ 7.1% (S.E., n ϭ 4). Following transfection, the cells were returned to the incubator for 24 -48 h, at which point replicate dishes were either subjected to manipulation of [Ca 2ϩ ] i followed by immunoprecipitation and phosphorylation or to immunoblotting to determine the relative expression levels.
Manipulation of [Ca 2ϩ ] i -After the period of expression, the transfected cells were washed twice with 10 ml of prewarmed serum-free DMEM and placed in a 37°C and 5% CO 2 incubator for 15 min in 5 ml of serum-free DMEM containing 3 mM EGTA and 0.1-1 M ionomycin. Either replicate dishes were maintained under these conditions for another 30 min (minus Ca 2ϩ ), or the media were changed to serum-free DMEM containing 2 mM Ca 2ϩ and 0.1-1 M ionomycin (plus Ca 2ϩ ); then the dishes incubated for 30 min as above.
Antibodies and Related Reagents-Horseradish peroxidase-conjugated goat anti-rabbit IgG and horseradish peroxidase-conjugated donkey anti-goat IgG were purchased from Jackson ImmunoResearch Laboratories Inc. Anti-FLAG M2-Biotin was obtained from Sigma. Anti-HA (Y-11) was from Santa Cruz Biotechnology. Streptavidin-horseradish peroxidase was from Invitrogen. Protein A-Sepharose was purchased from Amersham Biosciences. A polyclonal antibody to eIF4GII was described previously (32,34).
Immunoprecipitation-The cells were rinsed with PBS twice at room temperature, and the plates were placed on ice. All of the subsequent steps were performed at 4°C. The cells were lysed for 15-30 min in lysis buffer (immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EGTA, and 5 mM EDTA), 1% Triton X-100, 1% protease inhibitor mixture (Sigma), 2 mM DTT, 1 mM sodium orthovanadate, 200 nM okadaic acid, 25 mM NaF, and 10 mM NaPP i ) and then microcentrifuged at top speed for 10 min. Supernatants, precleared with protein A-Sepharose, containing 1 mg total protein were incubated with primary antibodies (1:100 anti-HA or 1:250 anti-eIF4GII) in lysis buffer overnight. Immunocomplexes were pelleted by incubation with 3 mg of protein A-Sepharose for 60 min followed by microcentrifugation for 5 s. The immunoprecipitation pellets were washed with 2 ϫ 0.5 ml of lysis buffer, 2 ϫ 0.5 ml of immunoprecipitation buffer, and 1 ϫ 0.5 ml of 50 mM Tris-HCl, pH 7.5.
Phosphorylation of Immunoprecipitated Proteins-To prepare 32 Plabeled eIF4GII for peptide mapping experiments (described below) or assay eIF4GII phosphorylation at CaMKI sites in vivo by "back phosphorylation" (described under "Results"), the following standard procedure was employed. Immunoprecipitation pellets (prepared as described above) were incubated for 30 min at room temperature in reactions (30 l of total volume) containing (in final concentrations) 50 mM Tris-HCl, pH 7.6, 0.5 mM DTT, 10 mM MgCl 2 , 25 M [␥-32 P]ATP (5 Ci/ml), 0.5 mg/ml BSA, 1 mM CaCl 2 , 4 M CaM, 0.2 M GST-CaMKI (activated by preincubation for 20 min at room temperature using the CaMKI/CaMKK ratio established to yield maximal activation; see above) and MBP-CaMKK␤ (0.13 M, residual from the preincubation). The reactions were stopped by the addition of 30 l of 2ϫ SDS dissociation buffer, heating at 95°C for 5 min, and brief centrifugation. The supernatants were subjected to SDS-PAGE, and phosphorylation was analyzed by autoradiography or phosphorimaging as indicated in figure legends. The concentration of CaMKI utilized was determined in preliminary experiments to maximally phosphorylate immunoprecipitated eIF4GII under these conditions (data not shown) so that in back phosphorylation experiments (see Figs. 5 and 6), signal intensity is inversely proportional to the degree of in vivo eIF4GII phosphorylation.
Western Blotting and Tryptic Phosphopeptide Mapping-The cellular lysates were subjected to SDS-PAGE and electrophoretically transferred at 100 V for 1.5 h to a polyvinylidene difluoride membrane. The membrane was blocked overnight at 4°C in 5% nonfat dry milk with PBST (PBS with 0.05% Tween 20), washed 2ϫ with PBST, and incubated for 1 h at room temperature in 5% nonfat dry milk/PBST containing primary antibodies (1:1000 anti-HA or 1:1500 anti-FLAG). The membrane was washed 5 ϫ 5 min in PBST and then incubated in the appropriate horseradish peroxidase-linked secondary antibody for 1 h at room temperature. Following washing 5 ϫ 5 min with PBST, immunoreactive bands were detected by ECL. ATP pools of nontransfected 293T cells were metabolically labeled using 32 P i , and endogenous eIF4GII was immunoprecipitated and subjected to two-dimensional tryptic phosphopeptide mapping as previously described (34).
Kinetic and Phosphorylation Site Analysis-To determine kinetic parameters (K m , k cat ), GST-CaMKI (activated by preincubation with MBP-CaMKK␤) was incubated with varying concentrations of bacterially expressed, affinity-purified HIS-eIF4GII(590 -1451). Given its low K m (1 M; see under "Results"), the reaction conditions were modified from those described under "Phosphorylation of Immunoprecipitated Proteins" as follows. The reaction times were reduced to 3-6 min, and the final concentrations of CaMKI and ATP were 1 nM and 200 M, respectively. Under these conditions the initial rates were linear even at the lowest substrate concentration (25 nM). Following the reactions, eIF4GII was subjected to SDS-PAGE and excised from the gel, and its 32 P incorporation was quantified by liquid scintillation spectrometry. Kinetic parameters (K m , k cat ) were determined by nonlinear fit to the Michaelis-Menten equation. Comparison of the phosphorylation of HIS-eIF4GII(590 -1451) WT and (S1156A) mutant proteins was as above at fixed substrate concentrations of 50 g/ml of both, followed by SDS-PAGE and autoradiography. Stoichiometry of phosphorylation was calculated from the time course of 32 P incorporation into HIS-eIF4GII(590 -1451) by fit to the first order rate equation. Nonlinear curve fitting was performed using the program Enzfitter (Elsevier, BioSoft).

RESULTS
As defined through the use of synthetic peptides, CaMKI has an optimal consensus sequence for substrate recognition of Hyd-Xaa-Arg-Xaa-Xaa-Ser*/Thr*-Xaa-Xaa-Xaa-Hyd (8). The highly degenerate nature of this motif, typical of other CaMKs and of protein kinases in general (Ref. 10 and references therein) precludes its use as an efficient mechanism for identifying substrates by direct search of protein data bases. The excessive numbers of hits generated in such a search (Ͼ500), 4 many of which are undoubtedly false positives, may in part be due to instances of identification of the site in a deduced sequence but located in a region of the protein not normally accessible to the kinase because of higher order protein structural determinants. The phosphorylation screening method (29) can potentially circumvent this obstacle in that, in principle, it should allow the probing kinase to select its most kinetically favorable substrates from all proteins expressed in the cell or tissue type from which the library is constructed.
Prior to its application for the purpose of identifying CaMKI substrates, we modified the original phosphorylation screening method (29) in several ways. First, the method, to our knowledge has only been applied twice (29,36), and both times it was with the custom library/vector combination, HeLa cell/GEX5 (29). However, in principle it should be applicable to any cDNA library prepared from a tissue in which substrates for a particular protein kinase are expected to be expressed, and because CaMKI is abundantly expressed in lung, we used instead of the HeLa cell/GEX5 library a human lung cDNA library constructed in TriplEx. A potential advantage of the latter vector is an apparent ability to express inserts in all reading frames. 5 Second, in a preliminary screening we were unable to obtain positive clones using a mutationally activated (hyperactive) form of the enzyme, CaMKI(T177D) (26). The reason for this may be that phosphorylation of CaMKI at Thr 177 by CaMKK provokes a dramatic drop in K m for substrates (11), 6 and although CaMKI(T177D) is 2-3-fold activated relative to wild type, it has only 10% or less the activity of phosphorylated CaMKI (26) and thus presumably has K m values too high to efficiently detect the small quantity of substrate protein expressed in the recombinant phage plaques. We therefore used CaMKI maximally activated by CaMKK␤ as probe in these screens.
Of 1.0 ϫ 10 6 primary recombinant phage screened, eight positive clones were detected. 7 They were judged to be authentic positive clones based on two criteria. First, all were positive through the secondary and tertiary plaque purification screens. Second, all were positive by the following control experiment. Plaques purified through the tertiary screen were replated at low density and grown on fresh isopropyl-␤-D-thiogalactopyranoside-impregnated filters. Following induction, the filters were divided into quadrants and separately incubated under standard reaction conditions with either the complete CaMKI reaction mixture, CaMKK omitted, CaMKI omitted, or both CaMKI and CaMKK omitted. For the eight purified clones, positive plaques were observed only with the complete mixture, indicating that detection of positive plaques requires activated CaMKI and is not due either to phosphorylation by CaMKK or to autophosphorylating protein kinases. Three representative experiments of this type are shown in Fig. 1. Because the inserts were directionally cloned, the three forward frames were conceptually translated. In all cases, two of the three were short open reading frames (average length, 29 amino acids) and are likely to be out of frame. The remaining long open reading frames represented partial length cDNAs encoding proteins identifiable in GenBank TM by BLAST search. The characteristics of these cDNAs and the proteins they encode are listed in Table I. It is of note that although eight independent cDNA clones were isolated, they represent partially overlapping coding sequences for only three proteins. The independent isolation of multiple cDNAs for each of the proteins is consistent with the idea that they may be favorable substrates for CaMKI.  Table I. Two of the sequences represent known proteins. Clones 1 and 2 encode 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatase (PFK-2/Fru-2,6-P(2)ase), a bifunctional enzyme involved in the regulation of gluconeogenesis/glycolysis. Clones 3-5 code for eIF4GII, a key eukaryotic translation initiation factor. The protein encoded by the final three clones (6 -8) (DNA Data Bank of Japan accession number AB022657/8) has not been characterized to date. To evaluate the significance of the identification of these proteins as potential substrates of CaMKI, we have focused, in the present investigation, on eIF4GII. Studies of a similar nature may in the future be directed toward PFK-2/Fru-2,6-P(2)ase.
Although it has been established that protein kinases can phosphorylate protein substrates in solid phase (29), the efficiency of such phosphorylation and hence their ability to be recognized in a screen could be aided if the substrate were bound by the kinase with moderate-to-high affinity. Such a situation is apparently the case for Mnk1, a substrate of mitogen-activated protein kinase detected in both phosphorylation and two-hybrid screens (29,40). This does not, however, appear to be the case for CaMKI detection of eIF4GII because we were not able to detect their association by co-immunoprecipitation from transfected cell lysates, suggesting that in the cell their association is of a transient nature (data not shown). In the absence of stable binding to eIF4GII, we considered the alternate possibility that eIF4GII represents a kinetically favorable substrate of CaMKI. Because full-length eIF4GII is rapidly degraded during bacterial expression and because this region of eIF4GII was detected in the initial screening as the region phosphorylated, we expressed in E. coli and affinity-purified a large (104 kDa) fragment of eIF4GII (amino acids 590 -1451) for kinetic characterization. As shown in Fig. 2, HIS-eIF4GII(590 -1451) is an excellent substrate for CaMKI having kinetic constants of K m ϭ 1 M and k cat ϭ 1.8 s Ϫ1 . Its specificity constant (k cat /K m ratio) of 1.8 s Ϫ1 M Ϫ1 is considerably higher (ϳ25-170-fold) than that of any previously described CaMKI protein substrate for which kinetic parameters are available, for example, myosin light chain (17) or Hsp25 (18). The rapid phosphorylation of eIF4GII by CaMKI is consistent with the "kinetic selection" assumption of phosphorylation screening proposed above and with the possibility that eIF4GII may represent an in vivo target of CaMKI.
To explore the latter hypothesis we first verified that endogenous eIF4GII, after isolation from its native mammalian cell environment, is a CaMKI substrate. As shown in Fig. 3, CaMKK␤-activated CaMKI phosphorylated endogenous eIF4GII immunoprecipitated from HEK293T cells. The phosphorylation of eIF4GII was evidently not because of co-immunoprecipitating kinase(s) (lanes 1 and 2 as compared with lane  4) nor to phosphorylation by CaMKK␤ (lane 3 as compared with lane 4). The lack of phosphorylation upon the addition of Ca 2ϩ /CaM in the absence of added CaMKI suggests that the latter does not stably interact with eIF4GII, a conclusion consistent with their lack of co-immunoprecipitation from lysates of transfected 293T cells as noted above.
We then investigated whether the phosphorylation observed in vitro was indicative of [Ca 2ϩ ] i -dependent, CaMKI-catalyzed phosphorylation of eIF4GII in intact cells. Endogenous 293T cell eIF4GII phosphorylated in vitro using CaMKI in the presence of [␥-32 P]ATP or in vivo in response to the Ca 2ϩ -ionophore, ionomycin in cells metabolically labeled with 32 P i , revealed a similar pattern of 32 P-labeled peptides by two-dimensional tryptic phosphopeptide map analysis (Fig. 4). As shown in Fig.  4A, immunoprecipitated 293T cell eIF4GII phosphorylated in vitro, generated three prominent (peptides 1, 2, and 4) and three faintly labeled (peptides 3, 5, and 6) tryptic phosphopeptides. Treatment of 293T cells with ionomycin markedly stimulated phosphorylation of peptides 1 and 2 and to a lesser extent peptides 3 and 4 (Fig. 4B). A faintly labeled peptide (peptide 6) could not be detected in the in vivo maps, presumably because of the low level of incorporation. The final peptide (peptide 5), which was only faintly labeled in vitro was detectable in vivo, but the level of incorporation appeared to be unaffected by ionomycin treatment, suggesting that it is not a primary site of phosphorylation by CaMKI and may be targeted by other kinases in vivo. In addition, 5-10 other phosphopeptides were detected that were neither phosphorylated by CaMKI in vitro nor responsive to ionomycin in vivo. The presence of background (ionomycin nonresponsive) phosphorylation suggests that eIF4GII may be the target of a signal transduction pathway remaining to be identified, and its level is consistent with the little apparent change in the total extent of phosphorylation of eIF4GII in vivo after ionomycin treatment as assessed by one-dimensional SDS-PAGE (data not shown). This is reminiscent of the pattern of eIF4GI phosphorylation in 293T cells, in which there is a redistribution of levels of phosphate incorporation into different sites stimulated by serum or mitogen treatment without apparent change in the total extent of phosphorylation (34). Unlike eIF4GII however, the pattern of eIF4GI phosphorylation was unaffected by ionomycin treatment (data not shown).
To confirm and extend these results by a complementary method, we employed a technique referred to here as back phosphorylation (24). In this procedure, cells cultured without metabolic 32 P i labeling of intracellular ATP pools are lysed, and a putatively in vivo phosphorylated protein is immunoprecipitated from the lysate and subsequently incubated in vitro with a kinase of interest in the presence of [␥-32 P]ATP. Because of the occupation of phosphorylation sites with nonradioactive phosphate prior to cell lysis, the extent of 32 P incorporation in vitro is inversely proportional to the degree of in vivo phosphorylation. A notable advantage of this technique is that it detects phosphorylation in vivo only at those sites capable of being phosphorylated by the specific kinase in vitro. It was therefore employed here to avoid the extensive basal phosphorylation of eIF4GII seen with phosphopeptide mapping. Using back phosphorylation, we then tested whether in vivo phosphorylation of eIF4GII at CaMKI site(s) was dependent upon the elevation of [Ca 2ϩ ] i as suggested by the phosphopeptide map analysis shown in Fig. 4. In the experiments shown in Figs. 5 and 6, [Ca 2ϩ ] i was directly manipulated by preincubation of 293T cells with ionomycin and 3 mM EGTA, followed either by incubation with ionomycin/2 mM Ca 2ϩ or by continued incubation with ionomycin/EGTA, essentially according to the protocol of Reilly et al. (41). As shown in Fig. 5, elevation of [Ca 2ϩ ] i led to pronounced CaMKI site-specific in vivo phosphorylation of endogenous eIF4GII (Fig. 5A) or transiently expressed HA epitope-tagged eIF4GII (Fig. 5B) without alteration of eIF4GII expression levels. An ionomycin concentration of 100 nM was suboptimal in promoting Ca 2ϩ -dependent phosphorylation of HA-eIF4GII but was sufficient to promote endogenous eIF4GII phosphorylation (Fig. 5, compare B with A). This presumably reflects a kinetic requirement for a higher [Ca 2ϩ ] i (promoted by higher ionomycin concentrations) to drive CaMKI-catalyzed phosphorylation of a greater total eIF4GII concentration achieved by the latter's transient overexpression.
The phosphorylation of eIF4GII in 293T cells in response to the elevation of [Ca 2ϩ ] i and in vitro by CaMKI at an equivalent site(s) (Figs. 4 and 5) is consistent with the possibility that CaMKI mediates [Ca 2ϩ ] i -dependent phosphorylation of eIF4GII in vivo. We evaluated this hypothesis further using the back phosphorylation technique. Confirming the data shown in Fig. 5B, elevation of [Ca 2ϩ ] i induced CaMKI site-specific in vivo phosphorylation of HA-eIF4GII (Fig. 6, compare lanes 1 and 4). CaMKI site-specific phosphorylation was noticeably augmented by overexpression of recombinant FLAG-tagged CaMKI in the presence of elevated [Ca 2ϩ ] i (lane 5). In addition, although considerably less than that observed in the presence of increased [Ca 2ϩ ] i , a low level of eIF4GII phosphorylation was detected in the absence of Ca 2ϩ upon CaMKI overexpression (lane 2 compared with lane 1). This is hypothetically due to the high concentration of CaMKI as a result of its overexpression resulting in partial activity even at low [Ca 2ϩ ] i . In vivo eIF4GII phosphorylation (in either the presence or absence of Ca 2ϩ ) was not observed, however, when cells were transfected with a form of CaMKI (CaMKI(K49A), Kin Ϫ CaMKI) made catalytically inactive by disruption of its ability to bind ATP (26) through substitution of Lys 49 with an Ala residue (lanes 3 and 6). Significantly, the ability of Kin Ϫ CaMKI to block the effect of elevated [Ca 2ϩ ] i on eIF4GII phosphorylation (compare lanes 4 and 6) is consistent with it acting as a dominant negative inhibitor of endogenous CaMKI. These data therefore support the hypothesis that endogenous CaMKI mediates eIF4GII phosphorylation in vivo. However, because we cannot eliminate the possibility that Kin Ϫ CaMKI could inhibit eIF4GII phosphorylation by an additional mechanism(s), this hypothesis will require additional validation in future studies as for example through the use of genetic silencing techniques.

FIG. 3. Phosphorylation of endogenous 293T cell eIF4GII by CaMKI in vitro.
Endogenous eIF4GII was immunoprecipitated from 293T cells, phosphorylated in vitro under the indicated conditions, and subjected to SDS-PAGE and phosphorimaging. Note that 32 P incorporation into CaMKK␤ is due to its autophosphorylation and that the phosphorylation of CaMKI by CaMKK␤ is not shown. 926 -1386, respectively (Table I). The overlap of these sequences suggested that there is a kinetically favorable site of phosphorylation by CaMKI between eIF4GII residues 994 and 1386. That this site may be Ser 1156 was further suggested by examination of the complete sequence of eIF4GII identifying it as the only site incorporating all of the determinants of CaMKI substrate recognition (8). To test this hypothesis, bacterially expressed HIS-eIF4GII(590 -1451) was phosphorylated to completion by CaMKI in vitro. The calculated stoichiometry of 0.6 mol 32 P/mol is consistent with a single phosphorylation site within this fragment. Finally, as shown in Fig. 7A, abolition of CaMKI phosphorylation by substitution with a nonphosphorylatable Ala residue confirmed its identity as Ser 1156 .
In other studies, an equivalent S1156A mutant of nearly full-length eIF4GII (140 -1585) was expressed in 293T cells and tested for its ability to serve as a substrate for CaMKI. In this case the S1156A mutant was phosphorylated although at a reduced rate (43% that of wild type under equivalent conditions). Thus, additional site(s) of phosphorylation exist (presumably N-or C-terminal to residues 590 -1451), but it is probable that these site(s) are phosphorylated relatively slowly in comparison to Ser 1156 in vivo. These sites might therefore be represented by the faintly labeled phosphopeptides seen in peptide mapping experiments (Fig. 4) (although it is likely that multiple phosphopeptides are also generated by incomplete tryptic hydrolysis, a common occurrence in two-dimensional peptide map analysis). Ser 1156 therefore appears to be the major site of phosphorylation of eIF4GII by CaMKI.
In mammals, eIF4G is represented by two homologues, eIF4GI and eIF4GII, with 46% amino acid sequence identity (39). Ser 1156 of eIF4GII lies within a region that aligns closely with an equivalent region of eIF4GI containing a cluster of serum-responsive phosphorylation sites termed the "phosphoregion" (34) (Fig. 7B). Although we did not detect eIF4GI as a potential CaMKI substrate by phosphorylation screening, the mapping of phosphorylation sites to the equivalent restricted region of the two proteins prompted us to examine whether eIF4GI can also be phosphorylated in vivo in a [Ca 2ϩ ] i -dependent fashion. Accordingly, HA-eIF4GI and HA-eIF4GII were expressed in 293T cells and immunoprecipitated using anti-HA antibodies after manipulation of [Ca 2ϩ ] i in the presence of 0.1 M ionomycin. Their respective levels of in vivo CaMKI sitespecific phosphorylation were then quantified by back phosphorylation and scanning densitometry (with correction for minor differences in expression levels by Western blotting using anti-HA antibodies). The plus Ca 2ϩ /minus Ca 2ϩ ratio of eIF4GII back phosphorylation was 0.53 Ϯ 0.10, n ϭ 2, indicating 47% in vivo phosphorylation in response to the elevation of [Ca 2ϩ ] i . In contrast, under the same conditions, the plus Ca 2ϩ /minus Ca 2ϩ ratio of eIF4GI back phosphorylation was 1.14 Ϯ 0.06, n ϭ 2, indicating that eIF4GI was not phosphorylated at CaMKI site(s) in response to the elevation of [Ca 2ϩ ] i . In addition, we have been unable to observe an effect of ionomycin on eIF4GI phosphorylation in 293T cells by phosphopeptide map analysis (data not shown). Consistent with these results, immunoprecipitated eIF4GI was phosphorylated by CaMKI at a rate Ͻ10% that of eIF4GII in vitro. As discussed below, the presence of phosphorylation sites in eIF4GI and eIF4GII apparently targeted by different signaling pathways but mapping to the same region (the phosphoregion) raises the possibility that the two functional homologues of eIF4G may be required to provide responsiveness to different extracellular stimuli in the control by phosphorylation of a similar regulatory function(s) in translation initiation. DISCUSSION Identification by phosphorylation screening of two protein substrates of CaMKI, eukaryotic translation initiation factor 4GII and 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatase, may be helpful in removing what to date has been the major impediment in understanding the physiological role of the CaMKI cascade, namely a lack of information regarding its downstream targets. Because the elements of this cascade, CaM, CaMKK, and CaMKI, are evolutionarily conserved from lower eukaryotes to mammals and in the latter demonstrate a broad cellular and tissue distribution (1)(2)(3)(4)(5)(6), it may be hypothesized that the cascade functions as part of a network for the regulation by [Ca 2ϩ ] i , of an essential cellular function(s). Based on their involvement in translation initiation and carbohydrate metabolism, respectively, the possibility that eIF4GII and PFK-2/Fru-2,6-P(2)ase may be physiological targets of this cascade is consistent with this hypothesis.
The evidence that at least one of these substrates, eIF4GII represents an authentic in vivo substrate of CaMKI may be summarized as follows: (i) eIF4GII was multiply selected in a proteome-wide screen based on its potential to be phosphorylated by CaMKI. (ii) Both mammalian, endogenously expressed eIF4GII, and recombinant, bacterially expressed eIF4GII(590 -1451) were phosphorylated by CaMKI in vitro with the latter found to have rapid kinetics of phosphorylation. (iii) eIF4GII phosphopeptides generated by CaMKI phosphorylation in vitro demonstrated similar electrophoretic and chromatographic mobilities as those observed in vivo after treatment of cells with the calcium ionophore, ionomycin. (iv) eIF4GII was phosphorylated in vivo at CaMKI-specific site(s), and this phosphorylation was blocked by a kinase negative form of CaMKI. Together, these data are consistent with eIF4GII being the downstream target of a Ca 2ϩ /calmodulin-dependent signaling cascade and implicate CaMKI in this pathway, although we cannot rule out the possibility that a distinct Ca 2ϩ -dependent kinase, of similar substrate specificity as CaMKI, and inhibitable by Kin Ϫ CaMKI is responsible for this phosphorylation in vivo.
The significance of [Ca 2ϩ ] i -dependent eIF4GII phosphorylation for mechanisms of translational control remains to be explored. eIF4G functions as a molecular scaffold, engaging in the multiple protein-protein and protein-RNA interactions involved in cap-dependent and -independent forms of translation initiation (39). Because mRNA recruitment to the ribosome is considered to be the rate-limiting step in initiation and because initiation is, under the majority of circumstances, rate-limiting for overall rates of translation (reviewed in Refs. 46 and 47), eIF4G represents a potential target for the physiological regulation of protein synthesis in response to intracellular signaling events. In a variety of cell systems, mitogens such as serum, insulin, phorbol 12-myristate 13-acetate, or epidermal growth factor induce eIF4G phosphorylation (34, 48 -51). In several instances, eIF4G phosphorylation in vitro (52) or in vivo (53,54) was found to correlate with enhanced eIF4F formation and The cellular lysates were prepared, and eIF4GII was immunoprecipitated using either anti-eIF4GII (A) or anti-HA antibodies (B) and phosphorylated in vitro with CaMKI, and the level of back phosphorylation was analyzed by SDS-PAGE and autoradiography. The cellular lysates were electrophoresed on separate gels, and the protein expression levels were analyzed by Western blotting using either anti-eIF4GII (A) or anti-HA antibodies (B) as shown.
initiation. However, causally relating eIF4G phosphorylation with eIF4F complex formation has proven difficult because of the oftentimes simultaneous phosphorylation of eIF4E and the existence of the 4E-binding proteins (4E-BPs), which when phosphorylated, fail to compete with eIF4G for binding to eIF4E, resulting in greater eIF4F formation (55,56).
Because eIF4G function in mammalian cells may be served by eIF4GI and/or eIF4GII, it has been an unresolved question whether the homologues are entirely redundant in translation initiation. Both homologues perform the same adaptor function in mRNA-ribosome recruitment and undergo viral-dependent proteolysis as part of the mechanism underlying the shut-off of host cap-dependent protein synthesis by viruses (30,57). One possibility is that while providing the same functional end points, eIF4GI and eIF4GII have evolved to respond differen-tially to cellular regulatory signals. Along these lines, sitespecific alterations of eIF4GI phosphorylation occur in response to treatment with serum and mitogens (34) but not in response to ionomycin treatment (this report). Conversely, we report here that eIF4GII is responsive to alterations in [Ca 2ϩ ] i in vivo. Mapping of serum-and Ca 2ϩ /CaMKI-responsive phosphorylation sites in eIF4GI and II, respectively, to an equivalent ϳ170-amino acid segment (the phosphoregion) raises the possibility that specific protein-protein or protein-RNA interactions may be similarly modulated by phosphorylation in response to diverse extracellular signals in the two eIF4G homologues.
It has been known for some time that Ca 2ϩ is required for optimal rates of protein synthesis (58,59). Two mechanisms have been proposed to account for this relationship. At the level FIG. 6. [Ca 2؉ ] i -dependent, CaMKIsite specific eIF4GII phosphorylation is blocked by kinase negative CaMKI in 293T cells. [Ca 2ϩ ] i was manipulated by ionomycin treatment followed by either 3 mM EGTA (Ϫ) or 2 mM Ca 2ϩ (ϩ) as indicated and described under "Experimental Procedures" in 293T cells expressing HA-eIF4GII. The cells were co-transfected with either empty vector, FLAG-CaMKI(WT), or FLAG-CaMKI(K49A) (Kin Ϫ ). The cellular lysates were prepared, immunoprecipitated using anti-HA antibodies, and phosphorylated in vitro by CaMKI, and the level of back phosphorylation was analyzed by SDS-PAGE and autoradiography. The cellular lysates were electrophoresed on separate gels, and the protein expression levels were analyzed by Western blotting using either anti-FLAG or anti-HA antibodies as indicated. of initiation, depletion of endoplasmic reticulum Ca 2ϩ stores results in eIF2␣ phosphorylation, inhibition of ternary complex formation, and consequent translational suppression (60,61). In addition, [Ca 2ϩ ] i -dependent phosphorylation of an elongation factor, eEF2 by a Ca 2ϩ /CaM-dependent protein kinase, elongation factor-2 kinase, is thought to reduce the rate of peptide chain elongation to the point where it becomes ratelimiting for overall rates of translation (62)(63)(64). The stimuli capable of mobilizing [Ca 2ϩ ] i and inhibiting translation, such as ionomycin, thapsigargin, or the excitotoxic glutamate receptor agonist N-methyl-D-aspartate (41,(63)(64)(65), evoke cellular stress responses that, if prolonged, lead to cell death, a process that may underly pathological conditions such as transient ischemia or neurodegenerative disease (reviewed in Ref. 66). Accompanying this stress response is a shift from the translation of cap-dependent to internal ribosome entry site (IRES)dependent mRNAs (67). Because eIF4G (or a fragment of eIF4G) may mediate translation of some IRES-containing cellular mRNAs, it would be of interest to determine whether [Ca 2ϩ ] i -dependent eIF4GII phosphorylation might serve to promote switching between these two modes of initiation.
Another situation where such a mechanism might play a role is during progression of the cell cycle (68,69). It has been observed that eIF4GII is more extensively phosphorylated during mitosis than in interphase and that this phosphorylation is correlated with reduced eIF4E association (70). The consequent reduction in eIF4F complex formation might in turn allow greater utilization of specific cellular IRESes. It is also of interest in this context that genetic deletion of either the CaMKI/IV or CaMKK homologues in A. nidulans results in disruptions in the nuclear division cycles of this organism (28) and that, in addition, inappropriate increases in the activity of CaMKI in S. pombe or of the CaMKI/IV homologue in A. nidulans lead to cell cycle arrest (24,28).
CaMKI has a substrate recognition motif, defined using synthetic peptides, very similar to those of the broad specificity kinases, CaMKII and CaMKIV, suggesting that CaMKI may also recognize multiple targets in vivo (8 -11). Based on the phosphorylation screening reported here, the key glycolytic/ gluconeogenic regulator PFK-2/Fru-2,6-P(2)ase could represent another, out of potentially multiple, intracellular targets. Although this possibility has not yet been addressed through in vivo studies, it is of interest that evidence for the coordination of carbohydrate metabolism and protein synthetic rates has been reported in both yeast and mammals (71)(72)(73). Whether additional substrates of CaMKI exist in mammalian cells and can be identified, perhaps productively through the use of phosphorylation screening, should be addressed in subsequent investigations.