|Home | About | Journals | Submit | Contact Us | Français|
Translational control in eukaryotic cells is critical for gene regulation during nutrient deprivation and stress, development and differentiation, nervous system function, aging, and disease. We describe recent advances in our understanding of the molecular structures and biochemical functions of the translation initiation machinery and summarize key strategies that mediate general or gene-specific translational control, particularly in mammalian systems.
Gene expression is regulated at multiple levels, including the translation of mRNAs into proteins. Compared to transcriptional regulation, translational control of existing mRNAs allows for more rapid changes in cellular concentrations of the encoded proteins and, thus, can be used for maintaining homeostasis in addition to modulating more permanent changes in cell physiology or fate. The process of translation can be divided into initiation, elongation, termination, and ribosome recycling. Most regulation is exerted at the first stage, where the AUG start codon is identified and decoded by the methionyl tRNA specialized for initiation (Met-tRNAi). The mechanism of start codon selection differs fundamentally between bacteria and eukaryotes and, accordingly, so do strategies for regulating initiation. In bacteria, base-pairing of the Shine-Dalgarno (SD) sequence located just upstream of the start codon with the complementary anti-SD sequence in 16S rRNA recruits the 30S ribosomal subunit directly to the initiation region of the mRNA. Hence, most translational control in bacteria involves modulating accessibility of the SD sequence. In eukaryotes, by contrast, the start codon is generally identified by a scanning mechanism, where the small (40S) ribosomal subunit loaded with Met-tRNAi in a preinitiation complex (PIC) binds to the mRNA near the 5′ end and scans the 5′ untranslated region (5′UTR) for an AUG (or rarely a near-cognate AUG) codon. Consequently, RNA structures that impede the ability of ribosomes to interact with the 5′UTR in single-stranded form, or subsequently to scan the 5′UTR, reduce the efficiency of initiation. Decoy AUG codons in the 5′UTR can also waylay scanning ribosomes as a means to impede recognition of the correct start codon.
The mRNA is activated for PIC binding by eukaryotic initiation factors (eIFs) that recognize the mRNA’s m7G cap structure at the 5′ end or the poly(A) tail at the 3′ end. This activation process can be downregulated by inactivating these eIFs to reduce translation for most mRNAs under starvation or stress conditions. The same strategy is used for mRNA-specific controls by sequence-specific RNA-binding proteins, or microRNA (miRNA) RNPs, which generally bind to the 3′UTR and recruit eIF-inhibitory factors to the mRNA. A subset of eukaryotic mRNAs can circumvent the scanning process by way of specialized sequences, called internal ribosome entry sites (IRESs), that recruit the PIC to the start codon in a manner analogous to, but generally more complicated than, the SD/anti-SD interaction in bacteria. The use of IRES elements is common in viral mRNAs and allows their translation to continue when eIFs necessary for mRNA recruitment or scanning are inhibited in infected cells. Another prominent mechanism for blocking translation initiation is to reduce the activities of the eIFs that stimulate Met-tRNAi recruitment to the 40S subunit. Although this might be expected to exert only a general inhibition of initiation, specialized mechanisms have evolved that allow certain mRNAs encoding transcription factors to be upregulated translationally under stress conditions where most translation is repressed.
There has been enormous progress over the last decade in dissecting the molecular mechanisms of eukaryotic translation initiation, fueled by advances in several areas. Structural biologists are providing high-resolution structures of ribosomal complexes and eIFs, and both classical and reverse genetics are being harnessed to identify the key domains and residues required for the biochemical reactions of the initiation machinery and the regulatory mechanisms available to control these activities. Reconstituted systems and sophisticated new assays for partial reactions in the pathway are providing an independent approach to deciphering mechanisms and are especially powerful when combined with genetics. In parallel with these mechanistic studies, biologists in multiple disciplines are producing a rapidly expanding list of translationally regulated genes and processes of central importance in signal transduction, development, neuroscience, aging, and disease. A case in point is the recent emergence of the vast potential for translational control of the mammalian genome by miRNAs (see Review by R.W. Carthew and E.J. Sontheimer on page 642 of this issue). Here we seek to highlight some of the exciting recent developments involving both the mechanisms of translation initiation and the diverse strategies for targeting this process to regulate gene expression in vertebrates.
Because most regulation occurs at the initiation stage of translation, the molecular basis of this process is being studied intensively to elucidate the molecular details of every potential control point. The initiation pathway is comprised of a set of reactions that place the AUG start codon of mRNA in the P (peptidyl) decoding site of the ribosome, base-paired with the anticodon of Met-tRNAi. As indicated above, the start codon is generally identified by the scanning mechanism (Figure 1). A preassem-bled 43S PIC containing Met-tRNAi and eIFs 1, 1A, 2, 3, and 5 is recruited to the capped 5′ end of mRNA, which is facilitated by the cap-binding factor eIF4E and its partners, eIF4G and eIF4A, in the eIF4F complex. The PIC then scans downstream, inspecting successive triplets as they enter the P-site for complementarity to the anticodon of Met-tRNAi. The Met-tRNAi is anchored to the PIC by the GTP-bound form of eIF2, and a perfect match with an AUG start codon triggers the arrest of scanning and irreversible hydrolysis of GTP in the eIF2-GTP-Met-tRNAi ternary complex (TC). With the release of eIF2-GDP and other eIFs, the large (60S) subunit joins to form an 80S initiation complex ready to accept the appropriate aminoacyl-tRNA into the A (aminoacyl) site and synthesize the first peptide bond (Pestova et al., 2007).
There has been exciting progress in elucidating the molecular basis of AUG codon selection during scanning. Work in a reconstituted system showed that eIF1 collaborates with eIF1A to promote scanning from the 5′ end through near-cognate start codons in the 5′UTR, which occurs on an unstructured mRNA without ATP hydrolysis by eIF4A, the helicase subunit of eIF4F (Pestova and Kolupaeva, 2002). It appears that eIF1 promotes scanning when non-AUG codons occupy the P-site by stabilizing an open conformation of the mRNA binding cleft of the 40S subunit, possibly by modulating the conformation of the “latch” to the mRNA entry channel (Passmore et al., 2007) (Figures 2A and 2B). eIF1 also rejects non-AUG codons during scanning by blocking the release of Pi from the partially hydrolyzed eIF2-GDP-Pi in the scanning PIC. These “gate-keeper” functions are neutralized at the AUG codon when eIF1 dissociates from its location near the P-site (Algire et al., 2005; Maag et al., 2005). Consistently, eIF1 mutations increase initiation at near-cognate (e.g., UUG) start codons by allowing premature eIF1 dissociation (Cheung et al., 2007), whereas overexpressing eIF1 suppresses UUG initiation in relaxed-fidelity (Sui−) yeast mutants. The eIF1A, thought to occupy the A-site, also regulates start codon selection. Its C-terminal region promotes continued scanning at non-AUG codons, whereas its N-terminal extension acts in the opposite way to arrest scanning and promote eIF1 release at AUG codons (Fekete et al., 2007) (Figure 2A).
The assembly and function of eukaryotic PICs appear to be enhanced by formation of a higher-order complex, dubbed the multifactor complex (MFC) in yeast, comprised of TC, eIF1, eIF5, and the multisubunit factor eIF3. The fact that eIF3 interacts with all other MFC components (Hinnebusch et al., 2007), and with multiple sites on the solvent-exposed “backside” of the 40S subunit (Pisarev et al., 2008; Siridechadilok et al., 2005; Valášek et al., 2003) (Figure 2C), probably underlies its ability to promote PIC assembly. eIF3 also plays a key role in recruiting the 43S PIC to the mRNA (Hinnebusch et al., 2007), possibly interacting with mRNA as it emerges from the exit channel of the 40S subunit (Pisarev et al., 2008). At least in mammals, eIF3 also forms a protein bridge to the mRNA by interacting with eIF4G (Pestova et al., 2007) (Figure 1). Interestingly, the j-subunit of eIF3 can bind near the A-site and mRNA entry channel of the 40S subunit and impede mRNA binding in the absence of TC, presumably to ensure that TC loading precedes 40S binding to mRNA (Fraser et al., 2007).
mRNAs that contain secondary structure in the 5′UTR require ATP and helicase activity to enhance binding of the 43S PIC at the cap and for subsequent scanning. The ability of eIF4F to promote scanning through the structured β-globin mRNA leader has been reconstituted in vitro (Pestova and Kolupaeva, 2002). High-resolution structures were solved for eIF4E bound to the cap (Pestova et al., 2007) and for eIF4E bound to a segment of eIF4G, which increases eIF4E’s affinity for the cap (Gross et al., 2003). Binding to eIF4G helps to hold the two lobes of eIF4A together and align the DEAD-box motifs in the orientation required for ATP-dependent helicase activity (Schutz et al., 2008). A cryo-EM model of the eIF3-eIF4G-40S complex places eIF4G near the mRNA exit channel on the solvent side of the 40S subunit (Siridechadilok et al., 2005). An intriguing new structural model envisions that eIF4G could interact with mRNA sequences both upstream and downstream of the nucleotides located in the decoding sites and could position eIF4A at the mRNA entry channel for its presumed role in unwinding secondary structure in advance of the scanning PIC (Marintchev et al., 2009) (Figure 2D).
Recently, it was shown that the mammalian DExH-Box protein DHX29 is needed for scanning through highly structured 5′UTRs in a reconstituted system in vitro, and that DHX29 occupies 40S ribosomes and might influence the conformation of the mRNA entry channel to increase the processivity of scanning (Pisareva et al., 2008). Knockdown of DHX29 in cells inhibits translation initiation as it causes polysome dissociation (A. Parsyan, N.S., and T.V. Pestova, unpublished data). It will be important to determine whether DHX29 knockdown specifically impairs translation of mRNAs with highly structured 5′UTRs.
The ability of the poly(A) binding protein, PABP, to interact with eIF4G can mediate circularization of the mRNA by linking the cap and poly(A) tail in a “closed loop” (Figure 1). This property is thought to underlie PABP’s ability to stimulate mRNA binding to the 43S PIC (Pestova et al., 2007), at least partly by enhancing eIF4F binding to the capped 5′ end of mRNA (Kahvejian et al., 2005). Forming the closed loop could facilitate reinitiation by post-termination ribosomes, and there is evidence that PABP interacts with polypeptide release factors eRF1 and eRF3. Surprisingly, this last interaction also stimulates initiation complex (IC) formation in vitro independently of a prior termination event (Amrani et al., 2008). PABP also stimulates 60S subunit joining (Kahvejian et al., 2005), and findings from yeast suggest that the poly(A) tail is required to block the inhibitory effects of two RNA helicases on subunit joining factor eIF5B (Searfoss et al., 2001). All of this helps to explain why increasing poly(A) tail length by the cytoplasmic poly(A) polymerase GLD2 activates translation of dormant maternal mRNAs in Xenopus oocytes. In fact, the poly(A) tail length is dynamically regulated by the opposing activities of GLD2 and poly(A) ribonuclease (PARN), both of which are recruited to the cis-acting polyadenylation element (CPE) in the 3′UTR by CPE-binding protein (CPEB) (Kim and Richter, 2006).
Discovery of the role of eIF5B in catalyzing 60S subunit joining, and the second GTP hydrolysis reaction, at the end of the eukaryotic initiation pathway has been a major recent advance (Pestova et al., 2007) (Figure 1). Interaction of eIF5B with the extreme C terminus of eIF1A stimulates both subunit joining and the GTP hydrolysis that triggers eIF5B release from the initiation complex (Acker et al., 2006; Shin et al., 2002). Its interaction with eIF1A helps to recruit eIF5B to the 48S PIC for subunit joining while eIF5B, in turn, promotes release of eIF1A from the 80S initiation complex to open up the A-site for the first aminoacyl-tRNA (Fringer et al., 2007). This means that eIF1A is present throughout the entire initiation pathway, participating in TC recruitment, scanning, AUG codon selection, and subunit joining.
Even though the mechanisms of AUG selection differ fundamentally between bacteria and eukaryotes, there are important structural and mechanistic features of initiation conserved between these two kingdoms. To begin with, we know that eukaryotic ribosomes have 3D structures similar to those in bacteria. The stunning progress on crystal structures of bacterial 70S ribosomes bound to mRNAs and tRNAs (Selmer et al., 2006; Yusupova et al., 2006) has provided atomic details of the path of mRNA and contacts made with tRNAs in the decoding sites. Although no crystal structures exist for eukaryotic ribosomes, a detailed molecular model was produced by docking homologous regions of bacterial rRNA and ribosomal proteins into a cryo-EM density map of an 80S ribosome. This model reveals strong similarities to 70S ribosomes, including the inter-subunit space containing the mRNA binding cleft and decoding sites (Spahn et al., 2004). Indeed, UV crosslinking of substituted mRNAs in reconstituted mammalian 48S PICs revealed many similarities with the mRNA path in bacterial 70S complexes (Pisarev et al., 2008). Moreover, there is genetic evidence that certain rRNA contacts with the P-site tRNA in 70S complexes are functionally conserved in eukaryotic PICs (Dong et al., 2008).
In bacteria, only three single-polypeptide initiation factors, IF1, IF2, and IF3, are required to stimulate assembly of the 30S IC, with the formylated Met-tRNAi bound to an AUG codon in the P-site. IF1 and IF3 promote initiation accuracy by destabilizing the binding of near-cognate elongator tRNAs at the expense of Met-tRNAi (Antoun et al., 2006). A recent cryo-EM model of the 30S IC provides the most detailed structural view obtained thus far of a translation initiation complex. It shows Met-tRNAi bound to the P-site, IF2 bound in a manner similar to that of other ribosome-dependent GTPases, and IF1 bound to the A-site (Figure 3A). Interestingly, it is seen that domain IV of IF2 interacts with the accepter end of Met-tRNAi, explaining how IF2 stabilizes fMet-tRNAi binding until the 50S subunit joins (Simonetti et al., 2008).
Although initiation is more complicated in eukaryotes than in bacteria, requiring >30 polypeptides that comprise the different eIFs, a simplifying fact has emerged, namely that the three bacterial IFs have structural or functional counterparts in eukaryotes and archaea. Thus, eIF1A shares with IF1 a conserved globular domain, which likely occupies the A-site in the manner described for IF1, but eIF1A contains additional N- and C-terminal segments required for its eukaryotic-specific functions of recruiting TC and scanning (Fekete et al., 2007; Pestova and Kolupaeva, 2002). IF2 and eIF5B are structurally similar GTPases that catalyze joining of the large subunit to the PIC, and both depend on GTP hydrolysis for final release from the initiation complex (Pestova et al., 2007; Shin et al., 2002). Although eIF1 and IF3 are not structurally related, they both bind the small subunit near the P-site and function analogously in rejecting non-initiator tRNAs (IF3) or non-AUGs (eIF1). IF3 and eIF1 can even substitute for one another in vitro, suggesting that they elicit similar conformational changes in the PIC according to codon-anticodon pairing in the P-site (Lomakin et al., 2006). Analogous to the findings described for eIF1, dissociation of IF3 from reconstituted bacterial 30S initiation complexes is accelerated by AUG and a favorable SD/anti-SD interaction. Moreover, IF1 inhibits subunit joining to the 30S IC when these mRNA signals are absent (Milon et al., 2008), reminiscent of the role played by the eIF1A in blocking non-AUG selection during scanning.
Some exceptional mRNAs, particularly in viruses, bypass the conventional scanning mechanism and at least a subset of eIFs by using IRESs to recruit the 40S subunit more directly to the initiation region (see Essay by B.R. Cullen in this issue of Cell). IRESs in picornaviruses, the first to be discovered, generally do not require eIF4E but need all other eIFs to recruit the 40S subunit (Doudna and Sarnow, 2007). These IRESs are long, highly structured sequence elements (Figure 3B) and are stimulated by noncanonical IRES trans-activating factors (ITAFs), including polypyrimidine tract binding protein PTB, ITAF45, or La autoantigen, most likely to stabilize their active conformations. The IRESs of hepatitis C virus (HCV) dispense with eIF4F entirely, binding directly to the 40S, and require only eIF3 and either eIF2/eIF5 or eIF5B to pair tRNAi with the IRES start codon and produce a 48S PIC competent for subunit joining (Pestova et al., 2008; Terenin et al., 2008). The ability to use eIF5B, the IF2 homolog, instead of eIF2 for tRNAi recruitment may allow initiation to proceed in virus-infected cells when eIF2 is inactivated by phosphorylation.
Cryo-EM analysis has revealed that the HCV IRES binds between eIF3 and the 40S on the backside of the subunit (Figure 2C), with domain II making contacts near the E (exit) site and inducing rotation of the head and the opening of the mRNA entry channel latch similar to that evoked by eIF1 and eIF1A (Doudna and Sarnow, 2007; Siridechadilok et al., 2005). Domain II is also crucial for eIF5-dependent GTP hydrolysis by TC (in the eIF2-dependent mode of initiation) and subunit joining (Doudna and Sarnow, 2007; Locker et al., 2007; Pestova et al., 2008), suggesting that it also facilitates the conformational change(s) elicited by perfect codon-anticodon pairing in the canonical initiation pathway.
The notion that IRES elements perform the functions of eIFs in placing a charged tRNA in the P-site and manipulating ribosome conformation is taken to the extreme by the IRES of cricket paralysis virus (CrPV), which dispenses with all eIFs and even Met-tRNAi (Doudna and Sarnow, 2007). This IRES uses different pseudoknot domains to contact 40S and 60S components and occupy the decoding center, placing the GCU triplet in the A-site (Schuler et al., 2006) where translation begins after a “pseudotranslocation” event moves the alanyl-tRNA decoding the GCU triplet into the P-site (Doudna and Sarnow, 2007; Jan et al., 2003). Remarkably, the IRES domain that occupies the P-site mimics precisely the anticodon stem loop of tRNAi and an AUG start codon (Costantino et al., 2008) (Figure 3B).
Clearly, HCV and CrPV are exemplars of viral IRESs that can highjack host ribosomes without competing for limiting eIFs in infected cells. In the same vein, several IRESs were discovered in cellular mRNAs that are active during mitosis or apoptosis when cap-dependent translation is impaired, presumably to allow efficient expression of key regulators in these special states. Research on cellular IRESs, and the involvement of non-canonical ITAFs in IRES function, has so far indicated a surprising diversity of structure and mechanism (Elroy-Stein and Merrick, 2007). Thus, IRES activity has been described for polypyrimidine sequences that bind the ITAF PTB (Mitchell et al., 2005), poly(A) tracts that bind PABP (Gilbert et al., 2007), and sequences capable of Shine-Delgarno-like pairing with rRNA (Dresios et al., 2006). IRES function has not yet been reconstituted in vitro for any cellular IRES, however, precluding a detailed understanding of their molecular mechanisms.
One of the key mechanisms of translational control during stress is the phosphorylation of eIF2 on Ser51 of its α subunit, converting eIF2-GDP into a competitive inhibitor of the 5-subunit GEF (guanine nucleotide exchange factor), eIF2B, and decreasing TC assembly. Remarkably, only a small portion of one subunit (ε) of eIF2B is sufficient for GEF function (Gomez et al., 2002), and three of the remaining subunits provide a binding site for phosphorylated eIF2α that inhibits GEF function (Hinnebusch et al., 2007). In addition to reducing general initiation, eIF2(αP) paradoxically induces translation of yeast transcriptional activator GCN4 by overcoming the inhibitory effects of four uORFs on reinitiation at the GCN4 ORF. After translating the 5′-most uORF (uORF1), post-termination 40S subunits can resume scanning and reinitiate downstream at uORFs 2, 3, or 4 after rebinding the TC. This is a dead-end, however, as scanning does not resume after termination at these uORFs. When TC levels are reduced by eIF2α phosphorylation by GCN2 (activated by amino acid starvation), a proportion of post-termination 40S subunits rebind TC only after bypassing uORFs 2–4 and reinitiate at the GCN4 start codon instead, and the Gcn4 thus produced transcriptionally activates amino acid biosynthesis (Hinnebusch et al., 2007).
Translation of ATF4 mRNA is upregulated by eIF2(αP) in mammalian cells by essentially the same reinitiation mechanism, leading to transcriptional activation of stress response genes, including the regulatory subunit of an eIF2(αP) phosphatase (GADD34) to provide negative feedback (Ron and Harding, 2007; Vattem and Wek, 2004). There are four different eIF2α kinases in mammals activated by different stresses, PKR (double-stranded RNA in virus infection), PERK (unfolded proteins in the ER), HRI (heme deprivation), and GCN2 (amino acid starvation), that phosphorylate the same residue in eIF2α and, hence, elicit the same “integrated stress response” involving downregulation of general translation and translational induction of specific transcription factors (Ron and Harding, 2007). The crystal structure of human kinase PKR bound to eIF2α reveals a novel interaction of the kinase domain (KD) G-helix with a segment of eIF2α remote from Ser51, explaining the exquisite substrate specificity of eIF2α kinases (Dar et al., 2005).
A key feature of GCN4’s uORF1 is the ability to allow a high frequency of reinitiation by post-termination 40S subunits, which depends on its short length (3 codons), and enhancer sequences both 3′ and 5′ of the uORF. There is now evidence that the 5′ enhancer sequences interact with the eIF3a-NTD (N-terminal domain), presumably as they emerge from the mRNA exit pore, to promote retention of post-termination 40S subunits on the mRNA (Szamecz et al., 2008). The eIF3, and also eIF4G, are likely retained on the ribosome during elongation of small uORFs to make them available for renewed scanning following termination. eIF3 also stimulates reinitiation after translation of a long uORF in a polycistronic mRNA of feline calicivirus (FCV) by binding to a sequence at the 3′ end of the uORF (Poyry et al., 2007). eIF3 has also been identified as a reinitiation factor in plants (Kim et al., 2004) and is targeted by the protein TAV (transactivator/viroplasmin) that enables efficient reinitiation on polycistronic mRNAs of cauliflower mosaic virus (Park et al., 2001). Interestingly, eIF3 was implicated in dissociation of 60S subunits from post-termination ribosomes following polypeptide termination, when ribosome recycling was reconstituted in vitro (Pisarev et al., 2007). Thus, the eIF3-40S complexes released at the stop codon could rebind to the stimulatory element just upstream in the FCV uORF to facilitate resumed scanning and reinitiation.
Another prominent, but completely distinct, mechanism of translational control by uORFs involves a “roadblock” to scanning PICs produced by an 80S ribosome that stalls while translating the uORF in a manner dictated by the amino acid sequence of the attenuator peptide encoded by the uORF. For the uORF that inhibits translation of the cytomegalovirus UL4 gene, the uORF-encoded peptidyl-tRNA interacts with release factor eRF1 to block polypeptide hydrolysis and stall the ribosome at the stop codon (Janzen et al., 2002). Stalling at the stop codon of the uORF controlling translation of yeast CPA1 (encoding an arginine biosynthetic enzyme) is dependent on arginine and, interestingly, the stalled ribosome activates nonsense-mediated decay to also reduce CPA1 mRNA levels (Gaba et al., 2005).
A second extensively used mechanism in eukaryotes to control the rate of translation initiation involves the mRNA 5′-cap recognition process by eIF4F. Binding of eIF4F to the cap structure can be hindered by the eIF4E homolog, 4E-HP (see below). The interaction between eIF4G and eIF4E in the eIF4F complex is inhibited by members of a family of related proteins, termed eIF4E-binding proteins (4E-BPs) (Figure 1) (Raught and Gingras, 2007). The 4E-BPs compete with eIF4G for a shared binding site on eIF4E (Marcotrigiano et al., 1999). Consequently, 4E-BPs inhibit cap-dependent, but not IRES-dependent, translation. 4E-BP binding to eIF4E is controlled by phosphorylation. Hypo-phosphorylated 4E-BPs bind strongly to eIF4E, whereas phosphorylation of 4E-BPs weakens their interaction with eIF4E (Raught and Gingras, 2007).
A critical kinase, which phosphorylates 4E-BPs, is mTOR (mammalian target of rapamycin). mTOR is a downstream Ser/Thr kinase in the PI3K/Akt signaling pathway and senses and integrates signals from extracellular stimuli, amino acid availability, and oxygen and energy status of the cells (Figure 1). mTOR is responsible directly or indirectly for the phosphorylation of several substrates, which are relevant to translation, including eIF4G, the S6 kinases (S6Ks), which phosphorylate eIF4B on Ser422, to enhance the interaction with eIF3 (Holz et al., 2005), and eukaryotic elongation factor-2 kinase (eEF2K). S6Ks also phosphorylate Pdcd4, which is a tumor suppressor that binds and suppresses eIF4A activity (Yang et al., 2003). Phosphorylation of Pdcd4 leads to its ubiquitination and degradation by the proteasome (Dorrello et al., 2006). Another major cellular signaling pathway that strongly impacts translation is the Ras-MAPK pathway. It is responsible for the phosphorylation of eIF4E and eIF4B. eIF4B phosphorylation occurs at Ser422, the site which is phosphorylated by S6K (Holz et al., 2005).
MicroRNAs (miRNAs) are short (~22 nt) oligonucleotides, which are major regulators of gene expression and function at the post-transcriptional level (see Review by R.W. Carthew and E.J. Sontheimer on page 642 of this issue). It is estimated that approximately half of the human genome is controlled by miRNAs, as there are ~1000 miRNAs and each could control ~10 mRNAs. Once processed from its primary transcript precursor, an miRNA is then loaded into a protein complex, referred to as an RNA-induced silencing complex (RISC), which targets and inhibits protein expression from specific mRNAs (Hammond et al., 2001; Hutvagner and Zamore, 2002; Martinez et al., 2002). Specificity of miRNA function is determined through direct base pairing of an miRNA-loaded RISC to miRNA-complementary target sites located within the 3′UTRs of specific mRNAs (Doench and Sharp, 2004). A large number of studies both in vivo and in vitro demonstrated that miRNAs either inhibit translation or destabilize the mRNA or both, depending on many factors. Studies in cells led to different conclusions concerning the mechanisms of translational repression at the level of initiation or elongation, mRNA degradation, or proteolysis of the nascent protein. Even for those studies concluding that inhibition occurs at the level of translation initiation, the data are conflicting as in some studies translation inhibition was cap dependent and mediated by eIF4F, whereas in others it was cap independent and mediated via eIF6 (Filipowicz et al., 2008). The factor eIF6 binds to 60S subunits and could thereby regulate ribosomal subunit joining, but evidence for a role in general translation initiation in vivo was obtained only recently (Gandin et al., 2008). In contrast to the disparate results obtained from studies in cells, in vitro experiments in cell-free translation extracts prepared from mammalian or Drosophila embryos all indicated that miRNAs suppress cap-dependent but not IRES-mediated translation (Mathonnet et al., 2007; Thermann and Hentze, 2007; Wakiyama et al., 2007). The recent report that the Argonaute (Ago) protein, a core component of RISC, binds directly to the cap structure and competes with eIF4E for binding could potentially provide a mechanistic explanation for the cap-dependent inhibition of translation (Kiriakidou et al., 2007). This would resemble the repression of the caudal gene by Bicoid and d4E-HP (Figure 4A). However, more recent results (Eulalio et al., 2008) challenge this conclusion.
Surprisingly, it was reported recently that miRNAs can also stimulate rather than inhibit translation when cells enter quiescence (Vasudevan et al., 2007). Under these conditions, two RISC-associated proteins, FXR2 and Argonaute 2, bind to the mRNA 3′UTR and stimulate translation. Thus, it is possible that miRNAs function as inhibitors or activators of translation of specific mRNAs in a cell-cycle-dependent manner.
It is striking that all of the components that are involved in the miRNA-mediated repression of mRNA expression concentrate in P-bodies, which were first characterized in Saccharomyces cerevisiae and are thought to be sites where mRNAs are sequestered from the translation machinery and sometimes degraded (Parker and Sheth, 2007). P-bodies contain components of the mRNA degrading machinery such as decapping enzymes and 5′ to 3′ exonucleases. The demonstration that Ago proteins, the associated proteins GW182 and Rck/p54, miRNAs, and mRNAs repressed by miRNAs are present in P-bodies suggests that P-bodies play a role in miRNA-mediated repression. Another type of cytoplasmic body that contains repressed mRNAs is stress granules, which accumulate in response to various stress conditions or inhibition of translation initiation (Anderson and Kedersha, 2008). Leung et al. (2006) showed that Ago proteins, miRNAs, and their target mRNAs accumulate in stress granules. In contrast to P-bodies, localization of Ago proteins to stress granules is miRNA dependent. Thus, stress granules might function together with P-bodies in the miRNA-mediated regulation of translation. It is noteworthy that some studies demonstrated the association of P-bodies and stress granules and suggested the possibility of exchange of material between P-bodies and stress granules (Anderson and Kedersha, 2008).
Learning and memory can be explained by alterations in synaptic strength of neurons, which describe the changes to synaptic strength in response to experience. Long-lasting forms of synaptic plasticity and memory are dependent on new protein synthesis. There is a large body of evidence to demonstrate that translational control plays a key role in regulating long-term changes in synaptic plasticity and behavior. Translational control is important for regulating both general protein synthesis and synthesis of specific proteins in response to neuronal activity.
Long-term potentiation (LTP) is a cellular model for the changes in synaptic strength that occur during learning and memory. Given that one neuron can have up to ~1000 synapses, different groups of synapses are strengthened in response to different stimuli. The PI3K/Akt/mTOR pathway and the MAPK/Erk signaling pathway are important for the control of protein synthesis-dependent LTP and learning and memory (Costa-Mattioli et al., 2009; Kelleher et al., 2004). These two signaling pathways also stimulate local translation at synapses (Sutton and Schuman, 2006). Numerous studies have provided evidence that dendrites and synapses contain ribosomes, mRNAs, and the components of the translation machinery (see Review by K.C. Martin and A. Ephrussi on page 719 of this issue).
Because synaptic activity and the learning and memory processes are dependent on translation, it has been anticipated that translational control would play an important role in regulating these processes. In the past decade results obtained from genetic manipulations demonstrated that this is the case. Mice lacking 4E-BP2 (the major 4E-BP form in the brain) require a lower threshold of stimulation to achieve lasting LTP but exhibit impaired spatial learning and long-term contextual fear memory. These results were interpreted to mean that the excessive translation could be deleterious to the synaptic plasticity required for learning and memory (Banko et al., 2005). Mice lacking the eIF2α kinase GCN2 show altered synaptic plasticity and memory formation. These mutant mice exhibit enhanced memory with a weak training protocol. This could be explained by augmented synaptic plasticity, as evidenced by a lower threshold of stimulation to achieve lasting LTP (Costa-Mattioli et al., 2005). This conclusion was bolstered by the finding that eIF2αser51ala/+ heterozygote “knockin” mice also performed better in several behavioral tasks and exhibited a facilitated long-lasting LTP (Costa-Mattioli et al., 2007). The molecular basis for the enhanced memory in mice with reduced eIF2α phosphorylation is explained by the fact that GCN2-mediated eIF2α phosphorylation causes an increase in translation of ATF4 mRNA in the brain. ATF4 suppresses memory because it inhibits the transcription factor CREB (cyclic AMP response element binding protein)-mediated gene expression, which is critical for long-term synaptic plasticity and memory (Barco et al., 2002).
The translation of specific mRNAs is frequently repressed by sequence-specific RNA-binding proteins, which bind to response elements in the 5′ or 3′UTR. The first such mechanism elucidated at the molecular level involves translational repression of ferritin mRNAs, encoding an iron-storage protein, under conditions of low iron. The iron regulatory proteins (IRP1 or IRP2) bind to a stem loop near the cap in the 5′UTR and impede binding of the 43S PIC to the 5′UTR of the activated eIF4F-mRNA complex (Hentze et al., 2007).
Message-specific translational control by RNA binding also plays a critical role in the early stages of embryogenesis, where it is the major determinant of gene expression, because transcriptional activity is low. This was best studied in Drosophila where most of the translational control is exerted on the expression of morphogens, which are targeted to distinct regions of the cytoplasm to define the embryonic axes and, thus, body pattern. Translational regulation plays a central role in the localization of morphogens, including Bicoid and Nanos (Thompson et al., 2007). The mRNAs encoding these morphogens are localized to the opposite poles of the embryo, establishing concentration gradients of the encoded proteins across the embryo, posterior-to-anterior or anterior-to-posterior. Bicoid and Nanos proteins function in the cytoplasm to suppress the translation of mRNAs encoding other morphogens (caudal and maternal hunchback), which are uniformly distributed in the embryo. Bicoid represses the translation of caudal mRNA at the anterior of the embryo, where it is present, by binding simultaneously to the 3′UTR of caudal mRNA and to an eIF4E-related protein (d4E-HP) at the 5′ end of the mRNA (Figure 4A) (Cho et al., 2005). In contrast to eIF4E, d4E-HP does not interact with eIF4G and therefore blocks the assembly of eIF4F and attendant recruitment of the 43S PIC. Similarly, d4E-HP binds to a protein complex containing Nanos at the 3′UTR of hunchback mRNA to inhibit its translation at the posterior (Cho et al., 2006). The foregoing mechanism is similar to that suggested originally from studies in Xenopus oocytes for the inhibition of translation by CPEB through Maskin, which binds to eIF4E (Stebbins-Boaz et al., 1999) (Figure 4A). In early oogenesis when Maskin is absent, a different mechanism operates to silence the translation of maternal mRNAs in which CPEB binds to eIF4E1b (a homolog of eIF4E) through 4E-T (4E-transporter) (Figure 4A) (Minshall et al., 2007).
Cup, an eIF4E-binding protein, is recruited to nanos mRNA by Smaug (Nelson et al., 2004) and to oskar mRNA by Bruno (Chekulaeva et al., 2006; Nakamura et al., 2004). In both cases, Cup inhibits recruitment of the mRNA to the ribosome by competing with eIF4G for eIF4E binding. Translational repression of oskar and nanos mRNA outside the posterior pole of the embryo is essential to buttress the targeting of the morphogens they encode, as the localization mechanisms for these mRNAs are inefficient (Bergsten and Gavis, 1999; Zimyanin et al., 2008).
Posttranscriptional repression mechanisms that involve sequestering target mRNAs into large silencing mRNPs (Chekulaeva et al., 2006; Nakamura et al., 2004), or recruiting the CCR4 deadenylase complex to specific target mRNAs (Chicoine et al., 2007; Kadyrova et al., 2007; Semotok et al., 2005; Zaessinger et al., 2006), also contribute importantly to embryonic patterning and development. These mechanisms might be similar to those by which Dhh1/Pat1 repress general translation in yeast during carbon starvation (Parker and Sheth, 2007).
Morphogen targeting requires not only translational repression, but also mechanisms to alleviate repression and activate translation in the appropriate spatial domain. Oskar contributes to derepressing nanos mRNAs by preventing its association with Smaug (Zaessinger et al., 2006). Activation of grk mRNA translation at the anterodorsal cortex of the developing oocyte is mediated by poly(A) binding protein 55D (PABP55D) in association with the large cytoplasmic protein Encore (Clouse et al., 2008). Vasa (Vas) is a DEAD-box helicase that localizes during oogenesis to the posterior and that binds to eIF5B (Carrera et al., 2000). In vas null oocytes, grk translation is greatly reduced, an effect also observed in oocytes that express only a form of Vas (VasΔ617) that is specifically compromised for eIF5B binding (Johnstone and Lasko, 2004). These results have led to the hypothesis that Vas positively regulates specific mRNAs such as grk by recruiting eIF5B, presumably to stimulate 60S subunit joining to the 48S PIC.
There are other instances of message-specific translational repression targeting the step of 60S subunit joining. ZBP1 (zip-code-binding protein 1), which binds to the 3′UTR of β-actin mRNA and is important for localizing the mRNA to the leading edge of fibroblasts or neurite growth cones, inhibits translation of the mRNA during its transport. ZBP1 blocks 60S joining to the 43S PIC (Huttelmaier et al., 2005). Inhibition of 60S joining has also been documented in the control of 15-Lipoxygenase (LOX) mRNA translation in erythroid precursor cells by binding of hnRNPs K and E1 to the 3′UTR of the mRNA (Ostareck et al., 2001).
In Drosophila, dosage compensation equalizes the expression of X-linked genes in males and females and involves the male-specific-lethal-2 (MSL-2) protein. Dosage compensation is eliminated in females by translational repression of MSL-2 mRNA by Sex-lethal (SXL) protein, expressed only in females, by a combination of two mechanisms: SXL binds to the 3′UTR of the msl-2 mRNA together with the RNA-binding protein, UNR, and inhibits recruitment of the 43S PIC to the mRNA, while SXL binding to the 5′UTR blocks scanning (Beckmann et al., 2005) (Figure 4C).
The mRNA transcripts for inflammatory genes—such as those encoding vascular endothelial growth factor (VEGF)-A or ceruloplasmin—are translationally repressed by interferon-γ (IFNγ) through 3′UTR elements that bind to a heteromeric complex termed IFNγ-activated inhibitor of translation (GAIT) (Fox et al., 2007). GAIT consists of the 60S subunit protein L13a, glutamyl-, prolyl-tRNA synthetase, NS-1 associated protein-1 (NSAP1), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). IFNγ induces phosphorylation of L13a, which leads to its release from the 60S ribosomal subunit, and assembly of the GAIT complex. Phosphorylated L13a in the complex bound to the target mRNA binds to eIF4G and blocks the eIF3-eIF4G interaction to inhibit 43S PIC recruitment (Figure 4B). Recently, Fox and colleagues (Mukhopadhyay et al., 2008) showed that the zipper-interacting protein kinase (ZIPK) phosphorylates L13a on a single site and, in turn, is activated by phosphorylation of a second IFNγ-induced kinase, death-associated protein kinase-1.
Translational control also plays an important role in the differentiation of embryonic stem cells (ESCs) into adult cell lineages. Sampath et al. (2008) demonstrated a striking increase in ribosome recruitment to mRNA during the differentiation of ESCs into embryoid bodies. This is explained by increased phosphorylation of 4E-BP1 during ESC differentiation via the mTOR signaling pathway, as the mTOR inhibitor rapamycin blocked 4E-BP1 phosphorylation and embryoid body growth.
Aberrant function of components of the translation machinery underlies a variety of human diseases including certain cancers and metabolic disorders. Most cancers are caused by dysregulation of signaling pathways that control cell growth and proliferation. These pathways also affect translation. Cancer is associated with aberrant changes in the amounts and activities of initiation factors, translation regulatory factors, and tRNAs. In addition, in many types of cancers there is an increase in the amounts of ribosomal subunits and accelerated ribosome biogenesis (Schneider and Sonenberg, 2007). A strong causal relationship between levels of expression and cancer was established for eIF4E, whose overexpression causes malignant transformation of human and mouse cells in tissue culture and tumors in mice (reviewed in Schneider and Sonenberg, 2007). In many cancers, the amounts of eIF4E are increased, and when eIF4E abundance is reduced by siRNA in cells or its activity repressed by 4E-BPs, the Ras- or Src-oncogene transformed phenotype reverts to normal. Likewise, antisense oligonucleotides to eIF4E inhibit tumor growth in mice while not causing any deleterious effects (Graff et al., 2007). Moreover, the amounts and phosphorylation status of 4E-BP1 were found to predict survival outcome for ovarian and breast cancers and childhood rhabdomyosarcoma. Thus, when 4E-BP1 levels were high or the protein was hypophosphorylated, the prognosis for survival was good (Armengol et al., 2007). The transforming activity of eIF4E can be explained by its ability to promote translation of a subset of mRNAs encoding proteins involved in regulating growth, proliferation, and apoptosis, most likely in a combinatorial fashion (Larsson et al., 2007; Mamane et al., 2007).
The phosphorylation of eIF4E by the Ser/Thr kinase Mnk is important for transformation and oncogenicity. A Ser209Ala mutant of eIF4E is unable to transform NIH 3T3 cultured cells (Topisirovic et al., 2004). Furthermore, Wendel et al. (2007) demonstrated that Ser209 is required for eIF4E’s oncogenic activity in a mouse model of B cell lymphoma. This study also concluded that an important mechanism by which phosphorylated eIF4E causes tumorigenesis is by suppressing apoptosis. Interestingly, one of the targets of phosphorylated eIF4E is the antiapoptotic protein Mcl-1, which promotes tumorigenesis.
Recent studies showed that the balance between cap-dependent and IRES-dependent translation in the cell plays an important role in tumorigenesis. Inhibition of cap-dependent translation in vitro causes an increase in IRES-dependent translation (Svitkin et al., 2005). The tumor suppressor 14-3-3σ inhibits cap-dependent translation during mitosis through its binding to several initiation factors, which is expected to result in elevated IRES-dependent translation (Wilker et al., 2007). In cells lacking 14-3-3σ, cap-dependent translation is not suppressed and consequently the IRES-dependent translation of the mRNA for cyclin-dependent kinase Cdk11 (p58 PITSLRE) is reduced, leading to impaired cytokinesis, aneuploidy, and tumorigenesis (Wilker et al., 2007). In the same vein, Barna et al. (2008) showed that activation of the Myc oncogene causes an increase in cap-dependent translation at the expense of IRES-dependent translation, and thus a reduction in Cdk11 levels, resulting in tumorigenesis. A switch from cap-dependent to cap-independent mRNA translation is mediated by hypoxia in large advanced breast cancers and is required for promoting angiogenesis and tumor survival and progression (Braunstein et al., 2007). The switch is caused by overexpression of 4E-BP1, resulting in the inhibition of cap-dependent translation, but enhanced translation of mRNAs containing IRESs that encode proteins, such as HIF1α, VEGF, and Bcl2, which are required for growth under hypoxic conditions.
Translational regulatory proteins control metabolism, and impairment in their function results in metabolic diseases, such as obesity. eIF2 phosphorylation plays a critical role in homeostasis of the endoplasmic reticulum (Ron and Harding, 2007). Knockin mice for the Ser51Ala (S51A) mutation of eIF2 suffer from serious hypoglycemia and die shortly after birth. Mice heterozygous for the S51A mutation are sensitive to a high-fat diet and display glucose intolerance, increased body weight, and hyperlipidemia. This type 2 diabetic-like phenotype is most probably caused by defects in pancreatic β cell function and lipid metabolism. Similarly, deletion of PERK in mice and humans results in β cell destruction after birth and diabetes mellitus (Ron and Harding, 2007).
4E-BPs and S6Ks, which are downstream targets of the PI3K/Akt/mTOR signaling pathway, also control metabolism. Activation of mTOR, and consequent phosphorylation of 4E-BPs and S6Ks, promote anabolic processes partly through enhanced translation. Accordingly, mice lacking 4E-BPs or S6Ks demonstrate altered metabolism. Mice lacking both 4E-BP1 and 4E-BP2 are hypersensitive to a high-fat diet and become obese. In addition, they display an insulin-insensitive and glucose-intolerant phenotype (Le Bacquer et al., 2007). Mice lacking S6K1 are resistant to age- and diet-induced obesity and exhibit an increase in insulin sensitivity(Um etal.,2004). The molecular basis underlying the metabolic phenotypes is not known.
Because translational control plays a role in learning and memory, it has been implicated in certain neuropsychiatric disorders, such as fragile X mental retardation (FMR) syndrome. This disorder is caused by changes in the FMR protein (FMRP), which is mutated or produced in reduced amounts in this disease. FMRP is an RNA-binding protein, which normally inhibits the translation of mRNAs whose products have critical roles in synaptic plasticity. Therefore, mutation or reduced expression of FMRP results in excessive synaptic plasticity. Although several conflicting models have been proposed to explain the mechanism of translational repression by FMRP (reviewed in Klann and Richter, 2007), it was recently reported that FMRP binds to eIF4E via CYFIP1 (cytoplasmic FMRP interacting protein 1), thus displacing eIF4G and consequently inhibiting translation (Napoli et al., 2008). This molecular mechanism conforms to the established model of specific translational suppression through tethering of the mRNA 3′ and 5′ ends (Figure 4A).
Viruses are dependent on the host translational apparatus to synthesize their proteins. They have evolved intricate strategies to gain access to the cellular translation machinery and to counteract host defense mechanisms that act at the level of translation. Viruses also manipulate cellular signal transduction pathways, which control the phosphorylation and activity of translation initiation factors. Many viruses, but not all, shut off host protein synthesis to promote their own replication and possibly prevent host innate immune defenses from mounting an antiviral response. The mechanism of host protein synthesis shut off is particularly clear for poliovirus and its relatives. These viruses selectively inhibit cap-dependent translation of cellular mRNAs by cleaving eIF4G, while viral mRNA translation continues unabated through the use of IRESs. Surprisingly, large DNA viruses that shut off host protein synthesis (Herpes simplex virus, Vaccinia virus) stimulate eIF4F assembly and consequently enhance translation. This is important for acute virus replication in quiescent cells and is also observed upon reactivation of latent infections (Arias et al., 2009). Stimulation of eIF4F assembly is caused in part by the activation of the PI3K/Akt/mTOR pathway, which results in phosphorylation of 4E-BPs through different mechanisms (Buchkovich et al., 2008). In addition, viral gene products, such as the ICP6 of Herpes simplex virus, which binds to eIF4G, enhance eIF4F assembly (Walsh and Mohr, 2006). Finally, Vaccinia virus promotes eIF4F assembly and also recruits initiation factors to cytoplasmic replication compartments, effectively increasing their local concentration (Katsafanas and Moss, 2007; Walsh et al., 2008). Interestingly, for DNA viruses the host protein synthesis shut off occurs primarily via changes in mRNA metabolism, including enhanced global mRNA turnover and inhibition of host cell splicing (reviewed in Mohr et al., 2007). DNA viruses, such as SV40, Herpes simplex virus, and cytomegalovirus can also impair translation of select host mRNAs by using viral encoded miRNAs, which inhibit the translation of cellular mRNAs that are required for apoptosis (reviewed in Sullivan and Ganem, 2005).
Aging is associated with decreased protein synthesis (Norsgaard et al., 1996). Studies in the nematode Caenorhabditis elegans demonstrated that inhibition of several components of the eIF4F complex including eIF4E and eIF4G or lowering their levels extends life span (Pan et al., 2007; Syntichaki et al., 2007; Curran and Ruvkun, 2007; Hansen et al., 2007; Henderson et al., 2006).
As we gain better insight into the mechanisms of translation it is clear that the use of emerging technologies will lead to a more complete understanding of this paramount cellular process. In particular single-molecule imaging and analysis is absolutely required to bolster the models generated from genetic and biochemical studies. Recently, single-molecule experiments were used to study the kinetics of ribosome movement on the mRNA during translation (Wen et al., 2008). The authors followed individual ribosomes translating single mRNA molecules tethered by the ends to optical tweezers. These kinds of experiments to study the movement of ribosomes on the mRNA 5′UTR or their direct interaction with IRESs will be of immense importance. High-resolution structures of eukaryotic PICs in different functional states, and new biochemical and biophysical assays for functions of eIFs and their regulators in vitro, also will be important for further dissecting initiation reactions. Reconstituting cellular IRES activity in vitro is crucial to validating the proposed functions of these elements and elucidating their molecular mechanisms.
Although the mechanisms by which miRNAs inhibit translation are enigmatic, it is highly likely that they will be elucidated in the near future, mainly because of the development of cell-free translation systems that recapitulate key features of miRNA-mediated mRNA repression in cells. It is noteworthy that recent large-scale studies in cells demonstrated that the magnitude of translation inhibition of most mRNAs is moderate (~2-fold) (Baek et al., 2008; Selbach et al., 2008), and that similar results were obtained in cell-free translation extracts. This is consistent with the requirement of several miRNA sites on an mRNA to achieve inhibition and with the presence of multiple miRNA-binding sites on most target mRNAs. It is also consistent with the proposed molecular mechanism by which miRNAs inhibit cap-dependent translation via competition with cap-binding proteins for binding to the cap structure. Because one miRNA can inhibit as many as 10 different mRNAs, it is expected that miRNAs would act in a combinatorial manner and that the target mRNAs would function in a common biological pathway or process. Enhanced knowledge of this complex biological system will derive from systems biology experiments using microarray platforms. The precedents for coordinate regulation of classes of mRNAs at the translational level have already been documented in yeast via Puf RNA-binding proteins (Gerber et al., 2004) and in mammalian systems where the coordinately expressed mRNAs were called “RNA operons” (reviewed in Keene, 2007).
Natural and synthetic antibiotics that interdict bacterial protein synthesis machinery especially targeting the ribosome have been used with great success for decades to treat bacterial infections. The availability of cocrystal structures of ribosomes complexed with antibiotics explains the structural basis of antibiotic action, development of resistance to antibiotics, and the promise of developing superior antibiotics. In contrast, efforts to discover small-molecule compounds that inhibit eukaryotic translation have begun only in the past 5 years. Screening natural compound libraries yielded potent inhibitors of eIF4A (Bordeleau et al., 2006; Pelletier and Peltz, 2007). Another study generated compounds that inhibit the eIF4E-eIF4G interaction and retard the growth of tumor cells (Moerke et al., 2007). It is hoped that these and related discoveries will lead to the development of drugs to combat cancer and other diseases and perhaps to slow the process of aging.
This work was supported in part by the Intramural Research Program of the National Institutes of Health (A.G.H.) and the Canadian Institutes of Health Research, The National Cancer Institute of Canada, the Howard Hughes Medical Institute, and the National Institutes of Health (N.S.).