From the Laboratory of Eukaryotic Gene Regulation, NICHD, National Institutes of Health, Bethesda, Maryland 20892
When subjected to starvation, stress, or viral
infections, mammalian cells down-regulate general protein synthesis by
phosphorylating the The unique induction of GCN4 translation in response to eIF2
phosphorylation is mediated by four short open reading frames (uORFs)
in the leader of GCN4 mRNA located 150-360 nucleotides upstream of the authentic initiation codon. Eliminating the start codons of all four uORFs results in high level GCN4
expression under both starvation and non-starvation conditions without
altering the mRNA (4). Thus, the uORFs inhibit GCN4
translation in non-starved cells by restricting the progression of
scanning ribosomes through the leader to the GCN4 start
codon. The first and fourth uORFs (from the 5
We proposed that under conditions where GCN4 is repressed,
the ribosomes that resume scanning following uORF1 translation are
forced to reinitiate at uORF4 because they rebind the
eIF2·GTP·Met-tRNAiMet ternary
complex before reaching the uORF4 start codon. Under starvation
conditions, phosphorylation of eIF2 Genetic Evidence for the Scanning-Reinitiation Mechanism of
Translating GCN4 mRNA Supporting the idea that any ribosomes which translate
GCN4 must have scanned past uORF4 without initiating
translation (Fig. 1), it was shown that translation of a
uORF4-lacZ fusion decreases under conditions where
GCN4 translation is stimulated (5). Another strong
indication that uORF4 must be skipped en route to GCN4 is
that mutations in the uORF4 stop codon that make uORF4 overlap the
GCN4 ORF have almost no effect on GCN4
expression, indicating that the location of the uORF4 stop codon is of
little consequence (5). In contrast, elongating uORF1 abolishes
GCN4 translation (8), supporting the idea that essentially
all ribosomes reach the GCN4 start site by reinitiation
following translation of uORF1. In addition, there is a critical
requirement for nucleotides flanking the uORF1 stop codon for efficient
reinitiation at GCN4. Replacing the last codon and 10 nucleotides 3 It was conceivable that following termination at uORF1, ribosomes would
be shunted directly to the GCN4 start site rather than
scanning the entire uORF1-GCN4 interval. This possibility is
inconsistent with the fact that insertions of stem-loop structures in
the vicinity of uORF4 abolish GCN4 translation. A shunting model also cannot explain the critical finding that GCN4
translation gradually decreased as the uORF1-uORF4 spacing was
progressively increased (5). According to the model in Fig. 1, under
derepressing conditions, 50% of the 40 S subunits scanning from uORF1
have not re-bound the ternary complex upon reaching uORF4 and continue scanning to GCN4. When the uORF1-uORF4 interval is
expanded, however, most of the 40 S subunits have bound the ternary
complex and are now competent to reinitiate when they reach uORF4.
Consequently, they cannot bypass uORF4 and reinitiate downstream at
GCN4. A final piece of evidence inconsistent with a shunting
mechanism is that the authentic uORFs were replaced with heterologous
small uORFs without destroying GCN4 translational control
(11-13). In our model, the only critical requirements for the first
uORF are to be recognized efficiently by ribosomes and then to allow
ribosomes to resume scanning. Presumably, both requirements can be met
by heterologous uORFs, albeit with lower efficiency than occurs with authentic uORF1. Interestingly, efficient reinitiation at uORF1 depends
not only on the sequence context of its stop codon but on sequences
upstream of the uORF (14).
A key feature of the model in Fig. 1 is that the reduction in ternary
complex levels in starved cells is large enough to allow a fraction of
reinitiating ribosomes on GCN4 mRNA to ignore uORF4 and
continue scanning to GCN4 but not extensive enough to allow ribosomes to skip uORF4 if they have not translated uORF1. This distinction may exist because in conventional initiation events ribosomes bind the ternary complex before interacting with the 5 The Protein Kinase GCN2 Stimulates GCN4 Translation by Inhibiting
Recycling of eIF2 by eIF2B GCN2 is a 180-kDa protein kinase that phosphorylates eIF2 and
thereby induces GCN4 translation in cells starved for
histidine (6) or several other amino acids (15). Substitution of Ser-51 in eIF2 There is strong evidence that phosphorylation of eIF2 Mutations were obtained in the GCD2, GCD7, and
GCN3 subunits of eIF2B that reverse the derepression of
GCN4 and general inhibition of translation in the presence
of high level eIF2 phosphorylation. These mutations appear to make
eIF2B insensitive to eIF2( The majority of the regulatory mutations isolated in GCD2,
GCD7, and GCN3, which decrease or abolish inhibition of
eIF2B by eIF2( Mutations in the GCD10 gene cause derepression of
GCN4 translation in the absence of eIF2 phosphorylation by
GCN2 and confer temperature-sensitive growth on rich medium (24, 25).
These are the same phenotypes associated with gcd mutations
affecting subunits of eIF2 or eIF2B. A reduction in polysome content
observed in gcd10 mutants at the non-permissive temperature
is indicative of a general defect in translation initiation, and a
small fraction of GCD10 protein was found associated with polysomes and
ribosomal subunits. GCD10 co-purified with eIF3 biochemical activity
and was shown to be identical to the RNA-binding 62-kDa subunit of eIF3
(46) described previously (47). These findings suggested that some
aspect of translational reinitiation on GCN4 mRNA
involves the GCD10-encoded subunit of eIF3. Based on the
known biochemical activities of mammalian eIF3 (2), one possibility is
that gcd10 mutations decrease the ability of eIF3 to
stimulate formation of
eIF2·GTP·Met-tRNAiMet ternary
complexes or promote their binding to ribosomal subunits.
Regulation of GCN2 Kinase Activity by Uncharged tRNA When GCN2 was replaced by the mammalian eIF2
It has been proposed that GCN2 binds specifically to GCN4
mRNA, and this interaction is required for efficient derepression of GCN4 translation. Presumably, binding to GCN4
mRNA would cause GCN2 to phosphorylate eIF2 in a localized manner
and stimulate GCN4 translation without affecting other
mRNAs (55). While this mechanism may increase the sensitivity of
GCN4 mRNA to eIF2 phosphorylation by GCN2, it cannot
account for the fact that mutations in eIF2 or eIF2B subunits lead to
high level GCN4 translation with little effect on other
mRNAs or on GCN4 derivatives lacking uORFs (24, 25). It
still seems necessary to postulate that reinitiation on GCN4
mRNA is much more sensitive than are conventional initiation events
to reductions in ternary complex levels.
The GCN1- and GCN20-encoded proteins are also
required for activation of GCN2 in starved cells. Mutations that
inactivate these proteins either reduce (GCN20) or abolish (GCN1)
phosphorylation of eIF2
subunit of eukaryotic translation initiation
factor 2 (eIF2)1 (1). eIF2
functions in translation initiation by delivering charged initiator
tRNAMet (Met-tRNAiMet) in a
ternary complex with GTP to the 40 S ribosomal subunit, forming a 43 S
preinitiation complex. In the translation of most mRNAs, the 43 S
complex binds near the capped 5
end, migrates downstream, and upon
reaching the first AUG codon, joins with the 60 S subunit to form an 80 S initiation complex (the scanning mechanism) (2). Following AUG
recognition, the GTP bound to eIF2 is hydrolyzed and eIF2 is released
as an inactive eIF2·GDP binary complex. Exchange of the GDP bound to
eIF2 with GTP is catalyzed by eIF2B. Phosphorylation of the
subunit
of eIF2 (eIF2
) on Ser-51 prevents the recycling of eIF2 by eIF2B; in
addition, the phosphorylated complex eIF2(
P)·GDP has a higher
affinity than non-phosphorylated eIF2·GDP for eIF2B, such that
GDP-GTP exchange on non-phosphorylated eIF2 is also impaired and
ternary complex formation is blocked (2). In Saccharomyces
cerevisiae, eIF2
is phosphorylated when cells are deprived of
an amino acid or purine, and interestingly, this leads to increased
translation of a specific mRNA encoding GCN4, a transcriptional
activator of at least 40 genes encoding amino acid biosynthetic enzymes (3).
end), which are
sufficient for nearly wild-type regulation, have different effects on
GCN4 translation. When present alone, uORF4 reduces
GCN4 translation to only 1% of the level seen in the
absence of all four uORFs, under both starvation and non-starvation conditions (4). In this situation, it appears that all ribosomes translate uORF4 and then dissociate from the mRNA. In contrast, uORF1 alone reduces GCN4 translation by only 50%,
presumably because half of the ribosomes that translate uORF1 resume
scanning and reinitiate at GCN4. When uORF1 is present
upstream of uORF4 and amino acids are abundant, all the ribosomes that
resume scanning following uORF1 translation will reinitiate at uORF4
and dissociate from the mRNA, preventing GCN4
translation. Under starvation conditions, however, half of the
ribosomes that resume scanning after uORF1 translation will bypass the
uORF4 start site (and also those at uORFs 2-3 when present) and
reinitiate at GCN4 instead (Fig.
1) (5). Thus, prior translation of uORF1
allows ribosomes to overcome the strong translational barrier at uORF4
by a reinitiation mechanism.
Fig. 1.
A model for translational control of yeast
GCN4 by phosphorylation of eIF2 by the protein kinase
GCN2. GCN4 mRNA is shown with uORFs 1 and 4 and the
GCN4 coding sequences indicated as boxes. 40 S
ribosomal subunits are shaded when they are associated with
the ternary complex composed of eIF2, GTP, and
Met-tRNAiMet and are thus competent to
reinitiate translation; unshaded 40 S subunits lack the ternary complex
and, therefore, cannot reinitiate. 80 S ribosomes are shown translating
uORF1, uORF4, or GCN4, with the synthesized peptides depicted by coils
attached to the 60 S subunits. Free 40 S and 60 S subunits are shown
dissociating from the mRNA following termination at uORF4
(left panel). The
,
, and
subunits of eIF2 in
yeast are encoded by SUI2, SUI3, and GCD11,
respectively. The subunits of eIF2B are encoded by GCD6, GCD2,
GCD1, GCD7, and GCN3, as shown. GCN1 and GCN20 are positive regulators of the eIF2
kinase GCN2 that mediate its activation by uncharged tRNA. See text for further details.
[View Larger Version of this Image (20K GIF file)]
reduces the concentration of
ternary complexes, such that many ribosomes scan the distance between
uORF1 and uORF4 without rebinding the ternary complex (6). Lacking
initiator tRNAMet, they cannot recognize the AUG codons at
uORFs 2, 3, and 4 (7) and continue scanning downstream. Most of these
ribosomes will bind the ternary complex while scanning between uORF4
and GCN4, enabling them to reinitiate at the GCN4
start codon (6). Thus, reducing the level of ternary complexes allows
ribosomes to bypass the inhibitory uORFs 2-4 and reinitiate at
GCN4 instead (6) (Fig. 1).
to the uORF1 stop codon with the corresponding
nucleotides from uORF4 converts uORF1 into a strong translational
barrier and destroys its ability to stimulate GCN4
translation when situated upstream from uORF4 (9). Mutational analysis
revealed that diverse AU-rich sequences at the third codon and
immediately 3
of uORF1 would promote high level reinitiation at
GCN4 (10). This led to the idea that base pairing between
the mRNA surrounding the uORF4 stop codon and the rRNA could
lengthen the time spent by the ribosome in the termination region and
increase the probability of ribosome release from the mRNA.
end
of the mRNA (2); therefore, reducing ternary complex levels may
decrease the frequency that ribosomes load at the 5
end, but once
bound to mRNA, their ability to recognize AUG codons while scanning
downstream should be independent of the concentration of ternary
complexes. It is also possible that reinitiating 40 S subunits are less
efficient than free 40 S subunits in binding the ternary complex
because they lack an initiation factor like eIF3 or eIF1A that promotes
this reaction (2). In any case, the reinitiation mechanism for
GCN4 translation is an extremely sensitive indicator of the
activity of eIF2 and associated factors that function in the formation
of ternary complex or promote its binding to ribosomes.
with Ala (the SUI2-S51A allele) completely
eliminates the increased phosphorylation in amino acid-starved cells
and impairs derepression of GCN4 to the same extent as a
deletion of GCN2. Immunopurified GCN2 specifically
phosphorylated the
subunit of eIF2 purified from rabbit or yeast
but not yeast eIF2 containing the Ala-51 substitution (6). These
results established that GCN2 stimulates GCN4 translation by
phosphorylating eIF2
on Ser-51. Low level expression of the
mammalian eIF2
kinases HRI and PKR in gcn2 mutants
induces GCN4 translation in a manner completely dependent on
Ser-51 in eIF2
(16). When expressed at high levels, PKR and HRI
produce a much higher level of eIF2 phosphorylation than occurs when
GCN2 is activated in amino acid-starved cells, and this severely
inhibits cell growth (16, 17). Mutationally activated forms of GCN2
(GCN2c kinases) also cause hyperphosphorylation of eIF2
and a general reduction in translation initiation (6, 18-20). These
last findings confirm that GCN4 translation is induced at a
lower level of eIF2 phosphorylation than is required for general
inhibition of protein synthesis.
by GCN2
down-regulates the formation of
eIF2·GTP·Met-tRNAiMet ternary
complexes. Mutations in the genes encoding the
,
, and
subunits of eIF2 (encoded by SUI2 (21), SUI3
(22), and GCD11 (23), respectively) mimic the effect of eIF2
phosphorylation in derepressing GCN4 translation
(Gcd
phenotype) independently of GCN2 and amino acid
starvation (24-27). These mutations produce a slow growth phenotype
(Slg
) on nutrient-rich medium and thus appear to decrease
eIF2 function in translation initiation. Deleting two of the four
IMT genes encoding tRNAiMet
elicits the same Gcd
and Slg
phenotypes
(28). Overexpression of the eIF2 complex in wild-type cells interferes
with derepression of GCN4 translation and suppresses the
growth inhibitory effects of GCN2c alleles.
Overexpressing eIF2 and tRNAiMet
together has a synergistic effect in suppressing the toxicity of eIF2
hyperphosphorylation (28). The
,
,
, and
subunits of yeast
eIF2B (encoded by GCD7 (29), GCD1 (30),
GCD2 (31), and GCD6 (29), respectively) were
first identified by point mutations with the same Gcd
and
Slg
phenotypes observed for mutations in eIF2 subunits
(24, 32-38). The fact that non-lethal mutations in the essential
subunits of eIF2B mimic eIF2 phosphorylation in derepressing
GCN4 supports the notion that eIF2 phosphorylation in yeast
leads to a reduction in eIF2B function. Even stronger evidence for this
conclusion comes from the fact that overexpressing the eIF2B complex in
yeast overcomes the growth inhibitory effects and the derepression of GCN4 associated with eIF2 phosphorylation (28).
P) without decreasing the ability to
catalyze nucleotide exchange on non-phosphorylated eIF2 (16, 39-41).
Interestingly, this is the only effect observed when GCN3 is
deleted (16, 39), indicating that GCN3 is required primarily for
inhibition of eIF2B by eIF2(
P). The same conclusion was recently
obtained for rat eIF2B by showing that a four-subunit complex lacking
the
-subunit (homologous to GCN3), reconstituted in
baculovirus-infected insect cells, was insensitive to inhibition by
phosphorylated eIF2 in vitro (42). The GCD2 and
GCD7 subunits are essential (29, 36) and thus make important
contributions to catalysis as well as regulation. GCN3, GCD7, and GCD2
show strong sequence similarities to one another throughout most of GCN3 and GCD7 and the C-terminal half of GCD2 (29, 31). Overexpression of GCD2, GCD7, and GCN3 reduces the inhibitory effect of eIF2(
P) on
general translation in vivo, and the excess amounts of these proteins form a stable subcomplex that can be co-immunoprecipitated from cell extracts. Formation of this subcomplex does not compensate for a loss of eIF2B function by mutation; thus the GCN3·GCD7·GCD2 subcomplex does not possess guanine nucleotide exchange activity but
instead appears to prevent eIF2(
P) from inhibiting native eIF2B
(43). Together these results provide strong evidence that GCN3, GCD7,
and the C-terminal half of GCD2 comprise a regulatory domain in eIF2B
that mediates the inhibitory effects of eIF2(
P).
P), fall into two clusters of 70 amino acids in
regions of strong similarity among all three proteins, and several
mutations alter equivalent positions in two or all three subunits.
These results suggest that structurally related segments in GCD2, GCD7, and GCN3 carry out similar roles in the regulation of eIF2B by eIF2(
P). One possible function would be to interact with residues in
eIF2
surrounding Ser-51 to increase the affinity of eIF2B for
phosphorylated eIF2; alternatively they could elicit a conformational change in the eIF2·eIF2B complex that prevents nucleotide exchange on
eIF2(
P) (41). In certain GCD2 regulatory mutants, nearly all of the eIF2 was phosphorylated with no effects on cell growth. These mutations probably allow eIF2B to accept phosphorylated eIF2 as a
substrate rather than simply reducing the affinity for eIF2(
P) (41).
Given the role of GCD2, GCD7, and GCN3 in the regulation of eIF2B, we
proposed that GCD1 and GCD6 interact with one another and form the
catalytic center for nucleotide exchange (44). This proposal is
supported by recent findings that the rat homologue of GCD6 expressed
in baculovirus-infected insect cells possessed low level exchange
activity that was greatly stimulated by coexpressing the other four
subunits in the same cells (42). The
subunit of eIF2 (encoded by
GCD11) is homologous to EF-Tu and contains sequences
conserved in all known GTP-binding proteins (23). Moreover, several
GCD11 mutations that reduce eIF2 function in vivo
map within the putative GTP-binding domain (45). Thus, it seems likely
that GCD11 contains the GTP-binding site on eIF2 and consequently would
be expected to interact directly with the catalytic center in
eIF2B.
kinases PKR or
HRI, GCN4 translation was stimulated independently of amino
acid levels (16), indicating that increased phosphorylation of eIF2
in starved cells reflects increased GCN2 activity, not inhibition of an
eIF2
phosphatase. There is little or no increase in GCN2 protein levels in response to amino acid starvation (19), showing that
GCN2 function, not its expression, is stimulated in starved cells.
Uncharged tRNA appears to be the activating ligand for GCN2 because
mutations in aminoacyl-tRNA synthetases lead to increased eIF2
phosphorylation by GCN2 (15) with attendant derepression of
GCN4 (48) and genes under its control (15, 49, 50) without
limitation for the cognate amino acids. GCN2 contains 530 residues
C-terminal to the kinase domain similar to the sequence of
histidyl-tRNA synthetase (HisRS) (51), including the conserved "motif
2" sequence that interacts with the acceptor stem of tRNA in class II
synthetases (52). Accordingly, it was proposed that binding of
uncharged tRNA to the HisRS-like domain stimulates the ability of the
kinase domain to phosphorylate eIF2
(51). Supporting this model, it
was shown that motif 2 in GCN2 is required for kinase function in
vivo and in vitro and for tRNA binding by the
HisRS-like domain in vitro (15, 61). In addition, numerous activating GCN2c mutations alter residues in the
HisRS-like region (18-20). GCN2 can interact with translating
ribosomes and free ribosomal subunits, and this association is
dependent on its C-terminal 120 amino acids (53). Interestingly,
GCN2c alleles that activate GCN2 most
effectively alter residues in this C-terminal domain (18-20), leading
to the suggestion that activation of GCN2 occurs on the ribosome by
uncharged tRNA bound to the ribosomal acceptor (A) site (18). This
would be akin to the activation of the RelA protein of
Escherichia coli by uncharged tRNA in the stringent response
(54) (Fig. 2).
Fig. 2.
A model describing an EF3-related function of
GCN1 and GCN20 on translating ribosomes in activation of protein kinase
GCN2 by uncharged tRNA. A, right arrow, EF3
stimulates binding of cognate aminoacyl-tRNA·GTP·EF1 complexes
to the ribosomal A site at the expense of non-cognate tRNA binding,
consistent with its proposed role in translational fidelity. Left
arrow, EF3 stimulates release of deacylated tRNA from the
ribosomal E site. B, in E. coli, limitation for
an amino acid (aa) results in accumulation of uncharged tRNA
in the cytosol. RelA interacts with ribosomes in the vicinity of the A
site, and binding of uncharged tRNA to the A site stimulates the
production of ppGpp by RelA. Increased levels of ppGpp trigger the
stringent response. C, GCN2 is a ribosome-associated eIF2
kinase with a HisRS domain that binds tRNA in vitro and is
required for kinase activation by uncharged tRNA in amino acid-starved
cells. By analogy with RelA in bacteria and considering the importance
of the EF3-like region in GCN1, we propose that the GCN1-GCN20 complex
binds near the A site and functions in an EF3-like manner to stimulate
the binding of uncharged tRNA to the A site or its delivery to the
HisRS-like domain in GCN2. PK, protein kinase;
ABCs, ATP-binding cassettes. E, P, and A, E, P, and A sites of the ribosome.
[View Larger Version of this Image (22K GIF file)]
by GCN2 but have no such effect in strains
expressing PKR in place of GCN2 (56, 57). The latter indicates that
GCN1 and GCN20 are required to increase GCN2 kinase function rather than inhibit an eIF2
phosphatase. Neither gcn1 nor
gcn20 mutations reduce GCN2 expression or
decrease GCN2 kinase activity in immune complex assays (56, 57),
suggesting that GCN1 and GCN20 are required to mediate activation of
GCN2 in vivo by uncharged tRNA. GCN1 and GCN20 are
physically associated with one another in cell extracts (57), and
substantial amounts of both proteins cosediment with polysomes and 80 S
ribosomes (58). GCN1 and GCN20 share sequence similarity with
translation elongation factor EF3 (56, 57), an essential protein that
stimulates binding of the EF1
·GTP·aminoacyl-tRNA ternary complex
to the A site and release of deacylated tRNA from the ribosomal E
(exit) site (59). GCN20 contains two domains highly related to the ATP
binding cassettes (ABCs) in EF3 but shows little sequence similarity to
EF3 outside of these regions (57), whereas GCN1 is most similar to the
region in EF3 N-terminal to the ABCs (56). The N-terminal 15-25% of
GCN20, which is critically required for its regulatory function,
interacts with the internal segment of GCN1 that is similar to EF3, and
the EF3-like domain was found to be highly conserved in a human
homologue of GCN1. Based on these findings, it was suggested that the
GCN1·GCN20 complex functions on the ribosome in an EF3-like fashion
to mediate activation of GCN2 by uncharged tRNA (Fig. 2) (58). A
Drosophila homologue of GCN2 was recently discovered whose
expression is developmentally regulated (60). This and the fact that
proteins highly related to GCN1 (58) and GCN20 (57) appear to be
present in human cells suggest that the GCN2·GCN1·GCN20 apparatus
for modulating eIF2 function in translation initiation may be conserved in all eukaryotes and play important roles in regulating cell growth
and differentiation.