Regulation of Eukaryotic Protein Synthesis by Initiation Factors

Protein synthesis is, in one sense, a typical metabolic process which must be regulated to remain within the bounds of available cellular energy resources but also respond to the growth require- ments of the cell. Yet unlike other cellular processes, the products of protein synthesis, enzymes, receptors, structural proteins, etc., directly affect every other cellular process. Hence, the number of input signals controlling the rate of protein synthesis would be expected to be much greater, and the number of regulatory mechanisms employed much more diverse, than for other bio-chemical pathways. Recent studies indicate that the first of these predictions is fulfilled, but surprisingly, the second is not. Protein synthesis is regulated in most cases by changes in either the cellular concentration or phosphorylation state of only a few initiation factors. The purpose of the present article is to review recent results on the mechanisms involved in this regulation. Looking at the past several years in overview, several themes emerge. The primary structures of many initiation factors have now been deter- mined, which has facilitated the determination of phosphorylation sites, their site-directed mutagenesis, and the over- and underexpression of wild type and variant forms in vivo. Such studies have revealed the profound effect of protein synthesis rate on cell growth and phenotype and have produced the unex- pected finding that some initiation factors act as proto-oncogenes or anti-oncogenes. Another theme is that an increasing number of viruses have been shown to cripple the host protein synthesis machinery and antagonize host defense mechanisms as well, but they do so by attacking only a limited number of sites. Finally, genetic approaches in yeast have revealed a regulatory mechanism which is remarkably similar to those of mammalian cells.

Protein synthesis is, in one sense, a typical metabolic process which must be regulated to remain within the bounds of available cellular energy resources but also respond to the growth requirements of the cell. Yet unlike other cellular processes, the products of protein synthesis, enzymes, receptors, structural proteins, etc., directly affect every other cellular process. Hence, the number of input signals controlling the rate of protein synthesis would be expected to be much greater, and the number of regulatory mechanisms employed much more diverse, than for other biochemical pathways. Recent studies indicate that the first of these predictions is fulfilled, but surprisingly, the second is not. Protein synthesis is regulated in most cases by changes in either the cellular concentration or phosphorylation state of only a few initiation factors.
The purpose of the present article is to review recent results on the mechanisms involved in this regulation. Looking at the past several years in overview, several themes emerge. The primary structures of many initiation factors have now been determined, which has facilitated the determination of phosphorylation sites, their site-directed mutagenesis, and the over-and underexpression of wild type and variant forms in vivo. Such studies have revealed the profound effect of protein synthesis rate on cell growth and phenotype and have produced the unexpected finding that some initiation factors act as proto-oncogenes or anti-oncogenes. Another theme is that an increasing number of viruses have been shown to cripple the host protein synthesis machinery and antagonize host defense mechanisms as well, but they do so by attacking only a limited number of sites. Finally, genetic approaches in yeast have revealed a regulatory mechanism which is remarkably similar to those of mammalian cells.

Initiation of Protein Synthesis
Initiation factors are grouped into classes on the basis of the step in initiation at which they act (for more detail, see Refs. [1][2][3] . Initiation factors of the eIF-1' class are considered to be involved with pleiotropic stimulation of initiation complex assembly. The eIF-2 factors are involved in binding Met-tRNAi to the native 40 S ribosome (40 SN), forming the 43 S initiation complex ( Fig. 1). This group includes the factor eIF-2B, which catalyzes the exchange of GDP for GTP in eIF-2 after it has been released from the ribosome. The eIF-3 polypeptides have as their primary function the formation of the 40 SN ribosome. The eIF-4 factors are involved in binding of mRNA to the 43 S initiation complex t o form the 48 S initiation complex. Finally, eIF-5 factors have as their primary function the release of eIF-2. GDP from the 48 S complex and joining of the 60 S ribosomal subunit.
Although it is conceivable that regulation by numerous initiation factor polypeptides could occur, the most compelling evidence supports regulation by modification of only two: eIF-20 and eIF-4E. eIF-Sa is a subunit of the heterotrimeric eIF-2  Calculated from the sequences deposited in DNA data bases. *Refers to GenBank, EMBL, or DDBJ DNA data bases except entries marked with *, which are from the protein data bases.
complex which forms a ternary complex with Met-tRNAi and GTP. eIF-4E binds mRNA caps and is one of four interacting polypeptides (the others being eIF-4A, eIF-4B, and eIF-4y), which collectively serve as a cap-directed mRNA helicase. At least 13 of the initiation factor polypeptides have been shown to be phosphoproteins (4), and correlations between in vivo protein synthesis rate and phosphorylation have been noted for many of these (3). However, in the case of eIF-2a and eIF-4E, the evidence extends to molecular genetic studies. Yet cDNA clones are being isolated for an increasing number of initiation factors (Table I), and further molecular genetic studies may reveal key regulatory roles for initiation factor polypeptides in addition to eIF-20 and eIF-4E.

Regulation Involving the eIF-2 Polypeptides
In principle, eIF-2 could be regulated by changes in intracellular protein level or activity. Recently, examples of the former have appeared aging ( 5 ) and mitogen activation of T-lymphocytes ( 6 ) . However, the overwhelming majority of physiological events affecting eIF-2 cause a change in the state of phosphorylation of the eIF-20 subunit (Table 11). These physiological conditions signify unfavorable conditions for the cell in which a continuance of protein synthesis would aggravate the situation. The response of the cell is to shut down protein synthesis by phosphorylation of eIF-2a at Ser-51 (7). eIF-2a-P binds tightly to the recycling factor eIF-2B and prevents its catalytic action.
Recent in vivo experiments have added further support for this mechanism. Choi et al. (8) showed that expression of e1F-2aA'a-s1 prevented inactivation of eIF-2B and allowed synthesis of a reporter protein, dihydrofolate reductase, to proceed. Furthermore, expression of eIF-2aAsP-51, which mimicked the phosphorylated state, caused inhibition of protein synthesis without kinase activation.
eZF-2a Kinases-One of the kinases responsible for eIF-2a phosphorylation, the double-stranded RNA-activated inhibitor of protein synthesis (DAI), is induced by interferon and, in the presence of dsRNA, becomes autophosphorylated and active in  phosphorylating eIF-2a (7). The recent cloning of DAI cDNA (9) has confirmed these features and indicated that the 62-kDa polypeptide contains all conserved domains for Ser/Thr kinases. One of the most interesting chapters in the DAI story is the manner by which viruses circumvent this antiviral defense mechanism (10). Adenovirus produces an RNA termed VAI that specifically antagonizes the activation of DAI by dsRNA. Recent molecular genetic studies support the idea that this process is mediated through eIF-2a phosphorylation; expression of the eIFvariant allows adenovirus with a deletion in the VAI gene to grow (11). Other viruses which antagonize the antiviral response through an effect on eIF-2 include Epstein-Barr, influenza, polio, HIV-1, reo, and vaccinia (10). There is evidence that, in addition to its antiviral role, DAI plays a role in normal growth and differentiation. The expression of an inactive DAI kinase was shown to result in malignant transformation of NIH 3T3 cells, suggesting that DAI normally functions as a suppressor of cell proliferation (12).
The second well studied eIF-2a kinase is the heme-controlled repressor (HCR), which is also activated by autophosphorylation, but in response to a wide range of physiological changes ( Table  11). The recent cloning of HCR cDNA reveals a 90-kDa protein containing all conserved domains of Ser/Thr kinases and homologies to both DAI and GCN2 (13). Several studies have appeared that give insight into the role of heavy metals (14), heat shock proteins (15,16), and dimerization (17,18) in the activation of HCR.
A third eIF-2a kinase has been discovered in the course of studies on general amino acid control in yeast (19,20). GCN4 is a transcriptional activator of amino acid biosynthetic genes. When amino acids become limiting, GCN4 synthesis is up-regulated at the translational level in a mechanism which involves reinitiation (21). The mRNA contains four upstream open reading frames (uORF). After translation of uORFl under non-starvation conditions, the scanning 40 S ribosome reinitiates at one of the downstream uORFs, and this prevents it from initiating at the GCN4 ORF. However, under amino acid starvation conditions, the scanning 40 S ribosome fails to be "rearmed" with eIF-2 until it scans past the uORFs and reaches the initiation codon for GCN4, whereupon productive initiation takes place. One of the genes involved, GCNZ, codes for a protein that is homologous to eukaryotic protein kinases (22) and associated with ribosomes during translation initiation (23).  in vitro, and the phosphorylation of eIF-Sa increases in vivo in response to amino acid starvation (24). eIF-2aA1a-51 is not phos-phorylated in vivo or in vitro, and its expression abolishes GCN4 control. eIF-2aAsp-51 derepresses GCN4 even in the absence of GCN2. Furthermore, another gene product involved in general amino acid control, GLC7, is a phosphatase that acts in opposition to GCN2 in the regulation of eIF-2a phosphorylation (25). Mutations in genes encoding eIF-2B subunits mimic the effect of eIF-Sa phosphorylation by GCNZ (26,27). Regulation of GCN2 activity may involve a region homologous to histidyl tRNA synthetase, suggesting that amino acid starvation may lead to accumulation of uncharged tRNAHi" and activate the kinase (22,28). Also, another gene involved in general amino acid control, GCD5, was shown to be lysyl tRNA synthetase (29). Mutations in GCD5 reduce charging of tRNALYe and increase expression of GCN4, which in turn increases CCD5 transcription in an autoregulatory loop. This implies that uncharged tRNALy", and perhaps other uncharged tRNAs, will activate GCN2.
Regulation of eIF-2B Activity-There is a coordinate increase in eIF-2B activity and protein synthesis in sea urchin embryos, which is believed to be caused by an increase in NADPH concentration following fertilization (Ref. 30, and references therein). Conversely, eIF-2B activity is reduced in mammalian cells in the absence of amino acids, glucose 6-phosphate, or polyamines (31, 32) and increased by insulin, phorbol esters, serum, and epidermal growth factor (33). The means by which these various conditions alter eIF-2B activity are not presently clear. Phosphorylation of the t subunit of eIF-2B by casein kinase I and I1 has been reported, but the effect on eIF-2B activity is in dispute (32,34).

Regulation Involving the eIF-4 Polypeptides
The other major site of regulation of protein synthesis is at the mRNA joining step (Fig. 1, Step 3), catalyzed by the eIF-4 factors. Studies to date have centered on two eIF-4 polypeptides: eIF-4E and eIF-47.
Changes in Protein Concentration-There are no situations yet reported in which protein synthesis is physiologically regulated by changes in intracellular eIF-4E concentrations, but molecular genetic manipulations demonstrating the consequences of such changes have produced surprising results. Expression from transfected vectors or microinjection of eIF-4E causes a variety of mammalian cells to grow rapidly, lose contact inhibition, grow in soft agar, and form tumors in nude mice (reviewed in Ref. 35). Similar effects are not seen with EF-la, EF-lH, eIF-2, eIF-4A, and the phosphorylation site variant eIF-4EA1"-53. This has led to the definition of eIF-4E as a member of a new class of translational proto-oncogene products, distinct from cell membranelocalized and nuclear protooncogene products.
The relationship of eIF-4E to other oncoproteins has recently been explored. Lazaris-Karatzas et al. (36) showed that overexpression of eIF-4E caused activation of ras and that overexpression of GAP, the negative effector of ras, caused reversion of the transformed phenotype. Lazaris-Karatzas and Sonenberg (37) showed that eIF-4E cannot transform primary rat embryo fibroblasts unless special selective conditions of cell culture are used but that eIF-4E will cooperate with nuclear oncoproteins v-Myc or E1A to transform these cells. The mechanism by which eIF-4E transforms cells is not clear, but the answer may lie in the preferential translation of mRNAs that are normally translationally repressed (reviewed in Ref. 35). In fact, eIF-4E overexpres-

Minireview: Regulation of Initiation Factor
Activity 3019 sion causes selective enhancement of ornithine aminotransferase mRNA translation (38). The opposite phenotype is seen when eIF-4E levels in HeLa cells are selectively decreased by expression of antisense RNA complementary to eIF-4E mRNA (39). Expression of antisense RNA causes a loss of both eIF-4E mRNA and eIF-4E protein, decreased cell growth, disaggregation of polysomes, and decreased protein synthesis. Surprisingly, eIF-4y decreases in parallel with eIF-4E. A similar reduction in protein synthesis is observed when a strain of yeast expressing a temperature-sensitive eIF-4E variant is grown at the restrictive temperature (40).
Turning to eIF-47, there is a clear physiological event in which protein levels are decreased, uiz. infection with certain classes of picornaviruses (2). Proteolytic cleavage of eIF-4y occurs during picornaviral infection, coincident with the loss of the cell's ability t o translate capped mRNAs, and extracts from poliovirus-infected cells are severely compromised in their ability to translate capped mRNAs. Nonetheless, recent results have called into question whether eIF-4y cleavage per se is responsible for the shut-off in host mRNA translation (41).
The mechanism by which translation occurs in the absence of eIF-4y is not clear. In addition to the picornaviral and comoviral RNAs, certain other mRNAs appear to be translated in the absence of eIF-47. GRP78 is a cellular protein whose mRNA is translated in poliovirus-infected cells (42). Also, mRNAs containing the tripartite leader of adenovirus continue to be translated after superinfection with poliovirus (43).
Under conditions in which no detectable eIF-4y remains (after expression of antisense eIF-4E RNA), a subset of cellular mRNAs continues to be translated, representing approximately 8% of control protein synthesis (39). The major polypeptides in this class were shown to be HSP90,. New initiation factors may be involved in the translation of these unusual mRNAs (45,46). Unexpectedly, although eIF-4 enhances 5"terminal AUG selection (47), it also stimulates internal initiation (48,49).
Changes in Actiuity-Various types of experimental evidence support the idea that eIF-4E is physiologically regulated by phosphorylation at Ser-53 and that this controls the rate-limiting formation of 48 S complexes ( Fig. 1, Step 3, reviewed in Ref. 35).
There is an increasing number of physiological changes in which eIF-4E phosphorylation and protein synthesis have been observed to be directly correlated (Table 111). With a few exceptions, these represent normal events in cell growth and development as opposed to the unfavorable conditions listed in Table 11. I n vitro evidence that Ser-53 phosphorylation is essential for eIF-4E activity comes from three sources. eIF-4E is normally present in the 48 S but not 43 S and 80 S initiation complexes, suggesting that it participates somehow in the binding of mRNA to the 43 S complex. However, the eIF-4EA'a-53 (50) and eIF-4EG'"-53 (51) variants fail to appear in the 48 S complex, and the eIF-4ESer-53 present in 48 S complexes is nearly all phosphorylated in contrast to unbound eIF-4E. Second, there is less eIF-4 activity in heatshocked Ehrlich cells (in which protein synthesis is reduced) and less phosphorylation of that portion of eIF-4E that is in the eIF-4 complex (52). Third, i n vitro treatment of eIF-4 preparations with protein kinase C results in phosphorylation of eIF-4E and eIF-4y, and such preparations are enhanced 3-5-fold in their stimulation of an eIF-4-dependent cell-free system and formation of 48 S initiation complexes (53). I n vivo evidence for the essentiality of Ser-53 phosphorylation comes from cell transfection and microinjection studies (see above). In all experimental systems tested, expression of eIF-4EA'"-53 failed to cause the phenotypic changes observed with eIF-4ES""53.
In contrast to eIF-2, little is known concerning the eIF-4E kinase(s). McMullin et al. (54) partially purified an eIF-4E kinase activity from reticulocytes. Tuazon et al. (55) showed that protein kinase C phosphorylated eIF-4E in vitro at Ser-53 to approximately 10% of the theoretical maximum, but this was enhanced to 100% if the eIF-4E was present in the eIF-4 complex. Maximal phosphorylation of eIF-4E in uiuo, on the other hand, occurred when protein kinase C was completely down-regulated in B lymphocytes by phorbol ester (56). Further studies showed that protein kinase C and casein kinase I (57) phosphorylated eIF-4E, but at both Thr and Ser sites, unlike the in uiuo situation. Very recently, an insulin-stimulated protamine kinase from bovine kidney was found to quantitatively phosphorylate eIF-4E (58).
There is also evidence that eIF-4y is activated by phosphorylation. eIF-4y is a substrate for protein kinase A (59), protein kinase C (59,60), and S6 kinase (60), and the amino acid sequence of eIF-4y has consensus recognition sites for all of these kinases and others (61). The fact that eIF-4 preparations treated with S6 kinase, which phosphorylates eIF-4y but not eIF-4E, stimulate cell-free translation (53) suggests that there is an effect of eIF-47 phosphorylation that is independent of eIF-4E phosphorylation.
Inhibition of eIF-4 activity has also been observed in a number of physiological situations, but it is not clear to what degree this may be due to eIF-4E phosphorylation. Heat shock causes a loss in eIF-4 activity (52, 62). In unfertilized sea urchin eggs, protein synthesis is blocked by inhibition of both eIF-2B and eIF-4 (63). Following fertilization, inhibition of first eIF-2B and then eIF-4 is relieved.

Conclusions and Perspectives
This article has focused on the two regulatory events for protein synthesis that are the best characterized at the present time, phosphorylation of eIF-Pa, which down-regulates protein synthesis, and phosphorylation of eIF-4E, which up-regulates it. The evidence that these two phosphorylations represent true physiological regulatory mechanisms includes the delineation of biochemical activities, protein sequence, and phosphorylation sites, the construction of site-directed variants, and the determination of their behavior in vitro and in uiuo. The genetic approach in yeast adds a new dimension to our understanding of one of these mechanisms and indicates that the response to amino acid starvation in yeast is very similar to the response to stress conditions in mammalian cells. This may in fact explain the effect of uncharged tRNA on eIF-Pa phosphorylation in mammalian cells (64).
Although the topic of regulation at eIF-2 and eIF-4 has been treated separately in this review, it is unlikely that their regulatory mechanisms operate independently in the cell. Several phenomena have been discussed which affect both factors (heat shock, sea urchin fertilization, poliovirus infection, adenovirus infection, and stimulation with serum, phorbol esters, epidermal growth factor, and insulin). These may represent separate actions of the various effectors on each factor, but it is also possible that the eIF-2 and eIF-4 signaling pathways are interlinked.