Internal Ribosome Entry Sites in Cellular mRNAs: Mystery of Their Existence*

Although studies on viral gene expression were essential for the discovery of internal ribosome entry sites (IRESs), it is becoming increasingly clear that IRES activities are present in a significant number of cellular mRNAs. Remarkably, many of these IRES elements initiate translation of mRNAs encoding proteins that protect cells from stress (when the translation of the vast majority of cellular mRNAs is significantly impaired). The purpose of this review is to summarize the progress on the discovery and function of cellular IRESs. Recent findings on the structures of these IRESs and specifically regulation of their activity during nutritional stress, differentiation, and mitosis will be discussed.

Although studies on viral gene expression were essential for the discovery of internal ribosome entry sites (IRESs), it is becoming increasingly clear that IRES activities are present in a significant number of cellular mRNAs. Remarkably, many of these IRES elements initiate translation of mRNAs encoding proteins that protect cells from stress (when the translation of the vast majority of cellular mRNAs is significantly impaired). The purpose of this review is to summarize the progress on the discovery and function of cellular IRESs. Recent findings on the structures of these IRESs and specifically regulation of their activity during nutritional stress, differentiation, and mitosis will be discussed.
Initiation of protein synthesis in eukaryotes is a complex process requiring numerous accessory proteins called initiation factors (also termed canonical initiation factors) (1). Assembly of the 80 S ribosome at a start codon within the majority of eukaryotic mRNAs involves binding of the mRNA 5Ј-m 7 G cap structure to a group of proteins referred to as the cap-binding complex or eIF4F (which consists of three proteins: eIF4E, 1 eIF4G, and eIF4A) (1)(2)(3). This is followed by recruitment of the 40 S ribosomal subunit and associated initiation factors (43 S initiation complex comprising a 40 S subunit, eIF2⅐GTP⅐Met-tRNA i , and eIF3) and movement of the 43 S initiation complex along the 5Ј-untranslated region (5Ј-UTR) in search of the initiation codon (1-3) (Fig. 1, left panel). This mechanism of translation initiation is known as "ribosome scanning" (1)(2)(3). Initiation factor eIF4G functions as a scaffolding protein. It binds eIF4E (a capbinding protein) and eIF4A (an ATP-dependent RNA helicase, which is thought to unwind the secondary structure in the mRNA 5Ј-UTR) and bridges the mRNA and the ribosome via its interaction with the 40 S bound initiation factor eIF3 (1)(2)(3). eIF3 is a multiprotein complex directly associated with the small ribosomal subunit and was shown to impede the association of the 40 S and 60 S ribosomal subunits in the absence of eIF2⅐GTP⅐Met-tRNA i ternary complex (1)(2)(3). eIF2 binds GTP and Met-tRNA i and transfers Met-tRNA i to the 40 S ribosomal subunit (1)(2)(3)(4). It should be noted that the availability of eIF4E for binding to eIF4G is regulated in eukaryotic cells by the phosphorylation of a small family of eIF4E-binding proteins (the 4E-BPs). Furthermore, proteins that bind the poly(A) tails of the mRNA (PABPs) were shown to facilitate initiation and recycling of ribosomes through interaction with eIF4G (1-3). However, it has become clear that some viral and eukaryotic cellular mRNAs can be translated via internal initiation, a process that involves direct binding of the ribosome to specific mRNA regions termed internal ribosome entry sites (IRESs). This translation initiation mechanism is generally independent of recognition of the 5Ј-mRNA end and involves direct recruitment of the 40 S ribosomes to the vicinity of the initiation codon (5-10) (Fig. 1, right panel). Although the debate continues over whether internal initiation of some cellular mRNAs can be explained by a number of alternative mechanisms (such as alternative mRNA splicing or the presence of internal cryptic promoters rather than internal ribosome binding) (13)(14)(15)(16), an overwhelming amount of data have appeared, arguing that internal initiation of translation of eukaryotic cellular mRNAs is an important cellular mechanism (9,10). In this review, we will discuss the cellular IRES elements and summarize our current understanding of their mechanism of action.

Viral Ribosome Landing Pads: How the Story Began
The poliovirus and encephalomyocarditis virus (EMCV) mRNAs were the first to be described to utilize IRES elements (17,18) originally termed ribosome landing pads by Nahum Sonenberg (18). Pelletier and Sonenberg (18) developed the dicistronic test, which became the standard for assaying IRES activity. Their dicistronic mRNAs had two open reading frames of reporter genes (also called cistrons): thymidine kinase and chloramphenicol acetyltransferase and the entire poliovirus leader sequence or parts of it inserted in between (18). The first cistron (thymidine kinase) measured cap-dependent initiation, whereas the second (chloramphenicol acetyltransferase) reflected the existence of internal initiation of translation (18). The dicistronic mRNA expression system was used to show that IRES elements allow initiation to bypass many regulatory mechanisms specific for cap-dependent translation involving the eIF4F complex. Hence, it appeared that IRES-driven translation initiation prevails when cap-dependent initiation is severely compromised (6,7,19). During viral infection, inhibition of cellular protein synthesis can be caused by the cleavage and partial loss of activity of eIF4G (19,20), 4E-BP dephosphorylation (21), or the cleavage of PABP (22).
The subset of canonical initiation factors required for viral internal initiation can vary depending on the IRES type and structure (e.g. recruitment of the 40 S ribosomes to hepatitis C virus and swine fever virus IRES containing mRNAs does not require any of the initiation factors of the eIF4 "family" (23)). Furthermore the cricket paralysis virus IRES containing mRNA directs initiation without a requirement for any of the canonical initiation factors at all (24,25). The same holds true for the IRES element found in an RNA virus that infects penaid shrimp (26).
Thus, three major conclusions can be drawn regarding viral IRES function. First, viral IRES-driven translation initiation can prevail when cap-dependent initiation is severely compromised. Second, viral IRES-driven translation has a reduced requirement for the canonical translation initiation factors. Third, viral IRES-driven translation can often be enhanced by a number of trans-acting factors (ITAFs).

Cellular IRESs
Very soon after the first IRES was identified in the picornavirus 5Ј-UTR (17, 18), it was found by Peter Sarnow and colleagues (33) that a cellular mRNA, encoding the immunoglobulin heavy chain binding protein (BiP), can be translated in poliovirus-infected cells at a time when cap-dependent translation is inhibited. It was concluded that the BiP mRNA can mediate internal entry of ribosomes in mammalian cells, indicating that translation initiation by an internal ribosome-binding mechanism can be used by cellular eukaryotic mRNAs (33). For a long time this example of a cellular mRNA containing an IRES element (along with the mRNA for the Drosophila melanogaster homeotic genes Antennapedia (34) and Ultrabithorax (35)) has been considered as an exception for the general cap-dependent mode of translation initiation in eukaryotes. However, it was later shown that 3-5% of the cellular mRNAs remain associated with polyribosomes in poliovirus-infected cells at a time when cap-dependent initiation is impaired (36). Indeed, in the past few years, IRES elements have been detected in an increasing number of cellular mRNAs from various species (9,10), and the list is growing (37). So far IRES elements were mainly found in mRNAs involved in regulating gene expression during development, differentiation, cell cycle progression, cell growth, apoptosis, and stress (Refs. 9, 10, and 37 and supplemental Table I).
Canonical Factors Requirement-Despite the identification of IRESs in an increasing number of cellular mRNAs, the mechanism of internal initiation is still poorly understood. It is clear that similarly to viral IRES elements, cellular IRESs enable translation of the mRNAs under conditions when cap-dependent protein synthesis is impaired (such as a variety of stress conditions, the G 2 /M phase of the cell cycle, and apoptosis). Under these conditions, the activity of the cap-binding factor eIF4E is down-regulated (3). This favors the expression from the IRES elements (9,10). Interestingly, the majority of the stress conditions (such as starvation for growth factors/nutrients, heat shock, UV light irradiation, hypoxia, endoplasmic reticulum stress, and virus infection) also lead to the down-regulation of activity of initiation factor eIF2 through phosphorylation of its ␣ subunit (38 -40). The reduction in eIF2-GTP levels leads to the inhibition of the overall rate of cap-dependent protein synthesis (38 -40). However, a number of cellular mRNAs containing IRES elements such as platelet-derived growth factor-2, vascular endothelial growth factor (41), oncogene c-myc (41,42), and PITSLRE kinase (43) continue to be efficiently translated under these conditions. Interestingly, the IRES element found in the cationic amino acid transporter (Cat-1) requires prior eIF2␣ phosphorylation for its following activation by stress (44). A number of reports have also indicated that some viral (EMCV and hepatitis C virus) IRES elements can function under conditions of eIF2␣ phosphorylation (41,45). The case of the EMCV IRES element continues to be controversial (41,43,45). Al- At a third step, the scanning complex encounters and recognizes the initiation AUG codon (usually this is the first AUG met by the scanning complex). This recognition is followed by the release of the initiation factors, and subsequently the 40 S subunit is joined by a 60 S subunit to form an 80 S ribosome (subunit joining is catalyzed by the initiation factor eIF5B). Note that this scheme is a simplified representation of the pathway and omits additional initiation factors participating in the process. For a detailed review see Ref. 1. The internal initiation pathway postulates a one-step (or in some cases two-step) mechanism by which the 40 S ribosome approaches the initiation codon. This translation initiation mechanism is generally independent of the recognition of the 5Ј-mRNA end and involves direct recruitment of the 40 S ribosome to the vicinity of the initiation codon (directed by an IRES element) (5-10). The 40 S recruitment is assumed to be accompanied by the simultaneous recognition of the initiation codon. In certain cases, however, the 40 S ribosome is also assumed to be able to scan downstream of the internal landing "place" to locate the initiation codon (see Ref. 10 for a review). Single asterisk, the fate of eIF4F bound to the cap structure after the formation of the 48 S complex is unclear. It is assumed that it can either fall apart with at least eIF4A continuing to be part of the scanning complex, or even all 3 proteins (eIF4E, eIF4G, and eIF4A) can stay with the scanning ribosome (11). Double asterisks, the exact sequence of events and timing of the initiation factors release are not known. For the most recent model see Ref. 12. It is also not clear whether the mechanism of ribosomal subunit joining in the case of cap-dependent and cellular mRNA internal initiation pathways is similar or not. Triple asterisks, the subset of canonical initiation factors as well as ITAFs required for internal initiation varies with different viral and cellular IRES elements (9, 10, 23-26). though two reports have indicated that EMCV IRES can function under the conditions of eIF2␣ phosphorylation (41,45), the other did not support this observation (43).
Information on the requirement of other canonical initiation factors for the activity of cellular IRESs is just beginning to emerge (46). Because many cellular IRESs are active in poliovirus-infected cells (36) they would be anticipated to have a reduced requirement for the integrity of initiation factor eIF4G. Indeed, it was recently shown that initiation driven by the c-myc and BiP IRESs is resistant to proteolytic cleavage of eIF4G and even stimulated under these conditions (46). It should also be noted that inhibition of protein synthesis in apoptosis is accompanied by a caspase-dependent cleavage of initiation factors eIF4G, eIF4B (eIF4B stimulates the helicase activity of eIF4A), eIF2␣, and the p35 subunit of eIF3 (47). Proteolytic cleavage of these proteins yields distinct, characteristic products (47,48). However, there is strong evidence that translation of c-myc, death-associated protein 5 (DAP5), X chromosome-linked inhibitor of apoptosis protein (XIAP), inhibitor of apoptosis protein 2 (HIAP2/c-IAP1), a pro-apoptotic protein Reaper, chaperone Hsp70, anti-apoptotic proteins Bcl-2 and Survivin, protein kinase C␦, and the apoptotic protease activating factor 1 (Apaf-1) mRNAs is maintained under these conditions and is driven by their IRES elements (supplemental Table I, and references therein). This indicates that these IRESs would probably have reduced requirements for the integrity of initiation factor eIF4G as well as eIF4B, eIF2␣, and the p35 subunit of eIF3.
It was also demonstrated that eIF4A activity is essential and limiting for the activity of c-myc and BiP IRES (46). It is unclear whether the activity of other cellular IRESs will also be dependent on eIF4A. Information on other canonical initiation factors that might influence the translation driven by cellular IRESs is limited.
Start Site Location and Selection-It is assumed that translation of most of the cellular mRNAs that contain IRES elements proceeds predominantly through the internal entry of ribosomes. Although all cellular mRNAs are capped and in principle should be able to bind the eIF4F complex, it is generally believed that the conventional scanning from the 5Ј-end is not possible for the majority of IREScontaining cellular mRNAs because their 5Ј-UTRs are long and structured (9,10). The vast majority of the cellular IRES elements are located within the 5Ј-UTRs in close proximity to the initiation codon, and thus even if cap-dependent translation from the 5Ј-end were possible, the translation products would be indistinguishable. The mRNA for neurogranin, a neuronal calmodulin-binding protein, is an example of an mRNA that is translated by both 5Ј-cap-dependent and internal initiation mechanisms (49). However, some cellular mRNAs contain IRES elements located in their coding region (50 -53). Translation of such mRNAs results in the production of two different protein products (one produced by 5Ј-end cap-dependent initiation and the other by internal initiation) (50 -53).
In some cases, however, the translation can be much more complex and the balance between the cap-dependent and IRES-mediated expression can result in the production of several protein "isoforms" (54 -58). For example fibroblast growth factor 2 (FGF-2) mRNA drives the synthesis of five products (54,55). Expression of the longest 34-kDa FGF-2 product was shown to be exclusively cap-dependent, whereas expression of the other four isoforms (24, 22.5, 22, and 18 kDa) is driven by an IRES element (54). The choice of internal codons seems to be influenced by a number of cisregulatory elements and also controlled by trans-acting factors bound to the FGF-2 mRNA (55). It was shown that the hnRNP A1 binds directly to the FGF-2 IRES and stimulates the expression from the internal downstream codons (55). There is also strong evidence that in the case of the c-myc oncogene mRNA, translation of the two c-myc isoforms (c-Myc1 and c-Myc2), which are generated by alternative translation initiation at an upstream CUG and a downstream AUG codon, can be triggered both by cap-dependent and cap-independent mechanisms (58). Protein isoforms whose production is triggered by the above mentioned IRES elements generally possess functionally distinct properties and might even have different cellular functions (50 -58).
Structure-Function Relationship-The major and yet unanswered question is how do cellular IRES elements recruit the 40 S ribosomes to initiate translation. It is widely assumed that IRES elements possess complex secondary and tertiary structure, which allow for multiple interactions with the components of the translational machinery (canonical initiation factors, ITAFs, and 40 S ribosomes) (9,10). It should be noted that no nucleotide sequence similarity among cellular IRES elements has been found so far. Although some reports have indicated that cellular IRESs may contain a common Y-type stem-loop structural motif (59), evidence is scarce demonstrating that these Y-shaped RNA structures actually are responsible for internal entry of ribosomes. Dissection of the cellular IRES element boundaries by deletion analysis showed that most of them are 150 -300 nt in length (37) although functional fragments as short as 22 nt have been also reported to have full IRES activity (60). It was also demonstrated in some cases that separate, non-overlapping sections of cellular IRESs are able to promote internal initiation although not as efficiently as the full-length IRES (60,61). Moreover, a very small 9-nt IRES from the UTR of the homeodomain protein Gtx was characterized (61). The nucleotide sequence of this short IRES was found to be 100% complementary to 18 S rRNA at nucleotides 1132-1124 (61). It was suggested that base pairing between segments of mRNAs and 18 S rRNA can lead to ribosome recruitment and translation initiation (61) similar to the way initiation proceeds in prokaryotes (62). This and other observations suggest that IRES elements are composed of distinct structural and functional elements, and it is the combined effect of these elements that promotes internal initiation. Chemical and enzymatic probing of the structure of a subset of cellular IRESs (including c-myc, FGF-2, L-myc, Apaf-1, Cat-1, FGF-1, and others (54,(63)(64)(65)(66)(67)) revealed complex structures that included stem loops and pseudoknots (54,(63)(64)(65)(66)(67). Yet, it is unclear how these structures can promote and facilitate efficient internal entry of ribosomes. tRNA-like elements (68), initially found within the structures of viral IRES elements (69), could facilitate docking of the cellular IRESs to the E or P sites on the 40 S ribosome. Recent cryo-electron micrograph visualization of two viral (hepatitis C virus and cricket paralysis virus intergenic region CrPV) internal ribosome entry sites bound to the 40 S ribosome seems to support this suggestion (70,71).
Interestingly, a number of cellular mRNAs containing IRES elements in their 5Ј-UTRs also contain small upstream open reading frames (uORFs) located within the IRES sequence (60,65,(72)(73)(74)(75)(76). In most cases, the significance of these uORFs for translation is not clear. Initially it was believed that these uORFs are present in the 5Ј-UTRs of cellular mRNAs to inhibit cap-dependent translation (Ref. 65, and see Ref. 77 for a review). However, recent studies on Cat-1 IRES-mediated translation showed that translation of the 48-amino acid uORF is required for increased Cat-1 IRES activity during stress (65,68). This study introduced a new concept of a "dynamic IRES" (65). Thus, we believe that cellular IRES elements are not rigid structures but can undergo transitions that can substantially influence their activity (64,65,68).
ITAFs-All known ITAFs are cellular RNA-binding proteins that play a variety of functions in cells (reviewed in Refs. 10 and 29 -32). For example, the levels of ITAF expression were shown to correlate with pathogenic properties and tissue specificity of picornaviruses (29 -32). The properties and tissue distribution of ITAFs were suggested to determine the biological properties of a variety of viruses that use the IRES-dependent translation initiation (29 -32).
RNA-protein complexes containing multiple protein components have been also reported for a number of cellular IRESs (Ref. 10 and references therein; supplemental Table I), and some of these proteins have been shown to modulate the efficiency of internal ribosome entry. A striking feature of many of these ITAFs is that they belong to the group of heterogeneous nuclear ribonucleoproteins (A1, C1/C2, I, E1/E2, K, and L) known to shuttle between the nucleus and the cytoplasm. It was suggested that the relative levels of ITAFs present in the cytoplasm vary under different stress conditions and their cell/tissue distribution could significantly modulate the level of IRESmediated translation (10) (supplemental Table I). One hypothesis is that ITAFs may help to recruit the 40 S ribosomal subunit to the mRNA through specific interactions with canonical translation initiation factors or ribosomal components. On the other hand ITAFs may promote or stabilize specific active conformations of the IRES. For example remodeling of the Apaf-1 structure upon interaction with UNR protein was shown to promote binding of polypyrimidine tractbinding protein, and these events led to stimulation of Apaf-1 internal initiation (64). Therefore, it is possible that ITAFs play an important role in regulating IRES activity by causing a conformational change of the mRNA structure. Further experiments are required to elucidate the functional role of ITAFs.
The Mystery of Their Existence-Why do cellular IRESs exist? Why has "Nature" evolved an alternative and not very efficient way to initiate translation? The answers to these questions can be found if one considers the physiological conditions under which cellular IRES elements are utilized. As was mentioned above, many of them are activated under conditions when cap-dependent protein synthesis is greatly reduced (such as starvation for growth factors/nutrients, heat shock, UV light irradiation, hypoxia, endoplasmic reticulum stress). The rapid inhibition of protein synthesis under these conditions is believed to function as a protective homeostatic mechanism. Thus, internal initiation represents a cellular "backup plan" for survival under the above mentioned conditions. However, it should be noted that only transient cellular stress favors the expression from IRES elements that help cells to cope with these conditions (e.g. XIAP, Cat-1, and many others), whereas severe stress conditions result in the activation of the "pro-apoptotic" IRES elements in e.g. Apaf-1 and DAP5 mRNAs (78). Thus, regulation of gene expression through internal initiation can impact on many processes and trigger mechanisms either leading to cell survival or cell death.
It should be also noted that high levels of expression of many proteins (such as c-Myc, Apaf-1, Bcl-2, XIAP, DAP5, and others) that are under IRES control would be detrimental under normal conditions of cell growth. Therefore, cellular IRESs may have evolved to support low levels of expression in normal conditions (10,16,78) and an inducible expression in response to different stimuli. It has been suggested that increased expression of several proteins that are under IRES control such as oncogenes, growth factors, and proteins involved in the regulation of programmed cell death (supplemental Table I) can contribute to the development of a number of pathological conditions in humans like diabetes (79), cardiovascular diseases (80), and the development and progression of cancer (81).
Uncovering the mechanism of IRES-mediated translation and its regulation will be a major challenge in the field.