IRES-mediated cap-independent translation, a path leading to hidden proteome

1In eukaryotes, a m7G cap is added to the 5′ end of most precursor messenger RNAs (pre-mRNAs) during RNA synthesis. This cap structure can be recognized by the eukaryotic initiation factor 4E (eIF4E), a component of the eIF4F complex consisting of eIF4E, eIF4G, and eIF4A. The eIF4F complex further facilitates the recruitment of the pre-assembled 43S pre-initiation complex (PIC) that includes the 40S small ribosomal subunit, eIF1, eIF1A, eIF3, eIF5, and the eIF2/Met-tRNAi/GTP ternary complex. Following assembly, the 43S PIC complex scans the 5′ untranslated region (5′ UTR) of the mRNAs in a 5′ to 3′ direction to reach the start codon (typically AUG) and subsequently employs 60S large ribosomal subunit to form an 80S ribosome to initiate peptide synthesis (for a mechanistic review, see ). This process is known as the cap-dependent translation initiation, which is a primary mode of translation initiation in eukaryotic cells.

2Under certain conditions, mRNAs cannot be translated through the cap-dependent translation. For example, the cap-dependent translation is inhibited under cellular stress and viral infection, and some viral mRNAs that are efficiently translated in their host cells do not even contain a 5′ cap structure. In such cases, an alternative mechanism known as cap-independent translation is often used to initiate mRNA translation through the internal ribosome entry site (IRES). The endogenous genes capable of cap-independent translation are usually involved in biological pathways in response to cellular stress or viral infection, implying that IRES-mediated cap-independent translation plays an important role under such cellular conditions. Alternatively, the protein translation under stress conditions can also be regulated through upstream open reading frames (ORFs) within the 5′ UTRs. Typically, upstream ORFs suppress translation of their associated downstream coding regions under normal growth condition. However, under cellular stresses, the suppression by upstream ORFs is released and thus translation of the downstream ORFs can be activated/increased. In this review, we will focus on the mechanism of the IRES-driven cap-independent translation and the experimental systems that are used to study them. The related topics such as biological functions of cap-independent translation and regulation through upstream ORFs can be found in other excellent reviews.

3By definition, IRESs are the RNA elements that recruit ribosomes to the internal region of mRNAs to initiate translation through a cap-independent pathway. They were originally discovered in the viruses of the Picornaviridae family such as poliovirus (PV) and encephalomyocarditis virus. A large number of IRESs were later identified in pathogenic viruses, including human immunodeficiency virus, hepatitis C virus (HCV), and foot and mouth disease virus. Although these viral IRESs contain diverse sequences, many of them have similar secondary structures and initiate translation through similar mechanisms. In addition, the activities of IRESs often require assistance from other factors known as IRES-transacting factors (ITAFs). Based on the structures and the requirement of translation initiation factors (IFs) and ITAFs, the viral IRESs are classified into four groups (Figure 1; ). Generally, IRESs with more tightly folded RNA structures require fewer protein factors (including IFs and ITAFs) to promote translation initiation; however, the mechanism behind this seemingly compensatory relationship is unclear.

4The Group I viral IRESs generally have strong activities and can initiate translation from a non-AUG start codon without additional ITAFs or even eIF2/Met-tRNAi/GTP ternary complex. These IRESs are folded to a compact structure that includes three pseudoknots (PKI, PKII, and PKIII) and multiple stem-loops to directly interact with the 40S small ribosomal subunit. The structures of Group II IRESs in ribosome complex reveals that the PKI pseudoknot mimics a codon-anticodon interaction between an mRNA and tRNA and can be loaded into the A site of ribosome without initiator Met-tRNAi. Interestingly, the CrPV IRES can initiate translation by interacting with bacterial ribosomes using the tRNA mimicry, suggesting a similarity of translation initiation between eukaryotes and bacteria.

5The Group II IRESs can also directly interact with 40S small ribosomal subunit with specialized RNA structure, but their activities usually require assistance of several IFs including eIF2 and eIF3 and initiator Met-tRNAi (;;; Figure 1). For example, the HCV IRES contains three structural domains that interact with 40S small ribosomal subunit. The Domain II binds to the 40S small ribosomal subunit at the E site, and the domain IIId paired with 18S rRNA to stabilize the interaction between IRES and 40S small ribosomal subunit. More details on the structures and the interaction between IRES and ribosome can be found in the in-depth review by others.

6The other two groups of viral IRESs, Group III and Group IV, cannot bind to the 40S small ribosomal subunit directly. Instead, they recruit the 40S small ribosomal subunit through different ITAFs and require canonical IFs in the cap-dependent translation (i.e. eIF2, eIF3, eIF4A, eIF4B, and eIF4G). The major difference between Group III and Group IV IRESs is the requirement of 40S ribosome scanning. In the in vitro translation systems, Group III IRESs usually initiate translation at the 40S subunit recruitment site and thus 40S ribosome scanning is unnecessary, whereas the 40S ribosome scans the untranslated region in a 5′ to 3′ direction to reach the downstream AUG start site in the Group IV IRES-mediated translation initiation. However, such distinction may change when the IRES activities are measured using in vivo systems: some Group IV IRESs (e.g. EV7, PV1, and EV-A71) were recently reported to initiate translation from the upstream ORF's start codon to produce the UP protein without 40S ribosome scanning during viral infection. Since translation initiation by IRESs of these two groups shares many common features with the canonical translation, they are probably affected by the reagents that inhibit cap-dependent translation of mRNA.

7In addition to the viral mRNA, IRESs were also found in cellular mRNAs, many of which encode proteins required in stress response, e.g. in conditions of apoptosis, mitosis, hypoxia, and nutrient limitation (reviewed in ). Generally, cellular IRESs contain fewer RNA structures compared to the viral IRESs and share little sequence conservation among them, and thus are difficult to be classified into different groups. The lack of sequence/structure similarity in cellular IRESs also makes it difficult to predict novel endogenous IRESs in mRNAs. On the other hand, the large diversity of endogenous IRESs in eukaryotic mRNAs also suggests that more diverse and complicate mechanisms may be used by endogenous IRESs to drive cap-independent translation.

8The cellular IRESs can be roughly classified into two types based on the mechanisms of ribosome recruitment: type I IRESs interact with ribosomes through ITAFs that bound on the cis-elements, e.g. RNA binding motifs and N-6-methyladenosine (m6A) modification (;;; Figure 2A); whereas type II IRESs contain a short cis-element that pairs with 18S rRNA to recruit ribosomes (similar to Shine-Dalgarno sequences in bacteria) (; Figure 2B). Hundreds of studies have shown that different eukaryotic mRNAs contain certain regions with IRES activity, and a recent screen using an in vivo translation reporter has demonstrated that ~10% of mammalian mRNAs contain some elements to function as IRESs. However, the existence of cellular IRESs is still under debate, mainly because the systems to measure IRES activity are not reliable (discussed in detail later). In addition, the mechanisms of cellular IRESs are largely unclear compared to viral IRESs.

9With the exception of Group I and Group II viral IRESs, most IRESs require the assistance of several ITAFs to recruit ribosomes for translation initiation. Almost all ITAFs are RNA binding proteins recruited by IRES elements to facilitate the ribosome assembly onto pre-mRNA (Figure 2). Most ITAFs were identified as nuclear proteins or proteins shuttling between the nucleus and cytoplasm, implying a crosstalk between translation and RNA transcription/processing. The La protein (La autoantigen) is the first ITAF discovered to promote the translation driven by PV IRES. Many other proteins were later reported as ITAFs to promote IRES activity using in vitro or in vivo experiments (see the list of IRESs and bound ITAFs in IRESites: http://www.iresite.org;; ). A well-studied example is the polypyrimidine tract-binding protein (PTB) that can promote the activity for both viral and cellular IRESs. Interestingly, this protein is also involved in the regulation of mRNA splicing, stability, and transport, further suggesting the crosstalk between translation and RNA processing.

10The exact mechanisms of how ITAFs facilitate IRES-dependent translation are largely unclear. It has been shown that some ITAFs (e.g. PTB) can function as RNA chaperones to remodel RNA structures around IRES, allowing for ribosome binding to the IRES. Alternatively, ITAFs can function as adaptor proteins to interact with ribosome or other translation IFs. The activity of most IRESs is fairly weak as judged by in vitro assays without the addition of ITAFs, which can increase the translation efficiency of cognate IRESs. Therefore, the efficiency of IRES-mediated translation generally varies in different cell types or tissues due to the expression level of ITAFs. Following this idea, a synthetic IRES containing binding sites for tissue- or cell type-specific ITAFs may be used to achieve tissue-specific protein production through IRES-dependent translation.

11The existence of IRESs in cellular mRNAs has been under debate for many years due to the lack of reliable evidence that translation initiation is indeed driven by these IRESs. To measure the activity of IRESs, various in vitro and in vivo translation systems were developed. In an in vitro translation system, the rabbit reticulocyte lysate or the translationally active HeLa cell extract is incubated with synthesized RNAs to examine their IRES activities. Such cell-free translation system is a powerful tool to determine the minimally required elements of the IRES; however, several cellular and viral IRESs do not have activity without additional ITAFs supplemented in this system. The in vivo systems, on the other hand, use various translation reporters that are transfected into cells to assay for the products of cap-independent translation. In this case, the ITAFs are provided by the cell, and thus the activities of IRESs may vary in different cell types.

12The activities of most viral IRESs have been validated using both in vitro and in vivo systems, and in several cases there are structural evidences for the direct interaction between IRESs and ribosomes. However, the activity of many cellular IRESs was only supported by limited evidence in in vivo systems, and most of cellular IRESs do not have the IRES activity as judged by in vitro system. Several reports have shown that the cellular IRESs require a `nuclear experience’ for their function, i.e. they can only drive translation in mRNAs that have been transcribed and spliced in the nucleus before being transported into the cytoplasm. The mechanism behind the requirement of `nuclear experience’ is unclear. A possible reason is that the in vitro synthesized RNAs lack necessary modifications or nuclear RNA binding proteins, and thus the activity of most cellular IRESs can only be observed by using in vivo systems.

13The most commonly used system to measure the IRES activity is the bicistronic translation reporter that contains two complete ORFs separated by an IRES within a single transcription unit (Figure 3A). The ORFs coding for renilla and firefly luciferases are often used in such reporters because their activities can be accurately quantified using a simple system with high sensitivity. After transcribed as a single mRNA, the first ORF is translated through cap-dependent translation, while the translation of the second ORF is initiated from the IRES. Such bicistronic reporters can be transfected into cells using a plasmid vector, or occasionally using in vitro transcribed RNA. As mentioned earlier, the in vitro synthesized RNAs lack `nuclear experience’, and thus it may not be efficiently translated. Although using plasmid vectors coding for bicistronic mRNAs can avoid this concern of `nuclear experience’, there are other problems that often induce artifacts in IRES activity.

14Three common artifacts may contribute to the false discovery of IRESs by the bicistronic translation vectors: first, the DNA sequence for the `IRES’ may serve as a cryptic promoter to initiate alternative transcription, producing a new mRNA with only the second ORF; second, the transcribed mRNA may contain cryptic splicing sites that generate in-frame fusion ORF through alternative splicing; finally, the ribosomes may be able to restart the translation after finishing the first ORF, either by reading through the stop codon of the first ORF, shunting to the second ORF, or through re-initiation at the second ORF (Figure 3A).

15Several careful control experiments are usually required to eliminate these artifacts. For the artifact caused by the promoter activity of `IRES’, the deletion of the first promoter or the siRNA knockdown targeting the first ORF should be performed as control experiments. When the translation of the second ORF is not reduced in these conditions, it is likely that the observed `IRES’ activity is, in fact, a false positive caused by the internal promoter. For the artifacts caused by cryptic splicing sites, one should carry out the control experiments using RT-PCR to directly detect additional splicing isoforms produced from the bicistronic reporter. The `abnormal’ splicing isoform containing the second ORF should be further validated by sequencing. Finally, to examine the false positive IRESs due to translation re-initiation or ribosome readthrough, a hairpin structure is usually inserted before the start codon of the first ORF to inhibit cap-dependent translation. The IRES-driven translation of the second ORF should not be affected by such hairpin structure (i.e. independent of the first ORF), whereas the readthrough translation will be inhibited.

16A large amount of circular RNAs (circRNAs) have been identified in all known eukaryotes, most of which are generated from linear pre-mRNA through back-splicing using canonical splicing machinery. Increasing evidences have demonstrated that circRNAs can function as mRNA to direct protein translation. Since circRNAs lack 5′ or 3′ end, the translation of circRNAs can only be initiated through cap-independent fashion, providing a reliable system to measure IRES activity.

17The translatable circRNAs can be generated by the circularization of in vitro synthesized RNA using chemical or enzymatic ligation, or through a self-splicing Group I intron system. However, it is technically challenging to ensure that all in vitro synthesize RNAs in such preparation are actually circular. In addition, the in vitro synthesized RNAs will have the same concern of lacking `nuclear experience’. Alternatively, the circRNAs can be generated from plasmid vectors encoding pre-mRNAs that undergo back splicing, providing an in vivo system to measure IRES activity using circRNA translation. The typical circRNA reporter contains an IRES and split GFP sequences that can be joined together by precise back splicing (;;; Figure 3B). Since the circRNA can only be translated through cap-independent fashion, this system will be able to avoid the false discovery of IRESs that occurred in linear RNAs (e.g. cryptic promoters, cryptic splicing sites, and translation readthrough). The only false positive scenario could be introduced by transcription readthrough, which produces a very long linear RNA concatemer including two or more exons with the split GFP ORFs. This artifact can be easily eliminated by linearizing the plasmid before transfection.

18Although circRNA is naturally suitable for the IRES activity measurement, there are several general rules that should be applied to ensure the specificity of this system. First, the ORF of the reporter gene should be divided into two fragments in reverse order, and thus the intact ORF can only be formed in circRNA. The introns flanking the circular exon usually contain complementary elements that can form base pairs to facilitate back-splicing (for the detailed design, see ). This design can reduce the artifacts caused by potential cryptic promoters, cryptic splicing sites, and translation readthrough. In addition, the linearized plasmid should be used to eliminate the false discovery that induced by transcription readthrough. Finally, when the candidate IRES does not contain the stop codon, the stop codon depleted reporter gene can be used to produce large protein concatemers through rolling cycle translation, which is probably the strongest evidence of cap-independent translation.

19It is well accepted that only 1%–2% of the human genome codes for proteins, while the rest largely comprises regulatory elements (e.g. promoters, introns, and non-coding RNAs). However, the precise number of protein-coding genes in the human genome is still an unanswered fundamental question. Currently, 19986 genes are annotated as being protein-coding by the GENCODE project (GENCODE version 30), ~80% of which have been supported by mass-spectrometry evidence in ProteomicsDB. The rest 20% of the annotated protein-coding genes (also referred as `missing proteome’) are resulted from either technical limitations in proteomic survey (e.g. small secreted proteins are difficult to be detected) or the genome mis-annotation (e.g. some protein-coding genes are in fact mis-annotated pseudogenes).

20Conversely, there could also be unknown proteins translated from the regions outside of the annotated protein-coding genes, suggesting the existence of a ‘hidden proteome’. Increasing evidence demonstrated that various small peptides are translated from non-coding regions as judged by ribosome profiling or by improved analysis of mass-spectrometry data. Some of these newly identified peptides are found to play important roles in the regulation of transcription (e.g. Pgc in fruitfly) or enzymatic activity (e.g. the MLN and DWORF that regulate SECRA activity in human). More examples and information of these peptides can be found in recent in-depth reviews.

21The RNAs encoding hidden proteome were originally annotated as non-coding RNAs because they lack typical features of mRNA. For example, these RNAs usually have a long 5′ UTR and short ORF with low translation efficiency. The long 5′ UTRs usually contain multiple stop codons, indicating that the downstream ORFs are probably translated through cap-independent mechanism driven by IRESs. In this case, the frequency and location of the cellular IRESs in the transcriptome become a critical issue.

22Previous studies have shown that 10% of the mRNA may contain the IRESs; however, the number of IRESs in non-coding RNAs or regions is still unknown. Interestingly, using circRNA reporter genes, it was reported that many short RNA elements containing m6A have IRES-like activity to drive the cap-independent translation of circRNAs. In addition, a recent unbiased screen of short sequences using circRNA translation reporter has also identified many short elements (<10 nt) with IRES-like activity to drive cap-independent translation. In fact, any fragment >50 nt is expected to contain an IRES-like short element by chance. Such surprising finding suggests that the requirement of IRESs is easy to fulfill, and thus many sequences can drive cap-independent translation, especially when the cap-dependent translation is inhibited under stress conditions. Analogous to the ‘pervasive transcription’ of the genome, the abundance of short IRES-like elements may suggest a ‘pervasive translation’ of the transcriptome. This finding also implies that multiple ORFs can be coded within a single mRNA, with some ORFs being translated internally through cap-independent translation. Similar to alternative splicing, this `alternative translation’ may serve as a new mechanism to increase the proteome complexity encoded by the human genome.

23Cap-dependent translation becomes impaired under stress conditions, therefore it is tempting to speculate that many of the novel ORFs might encode peptides with functions in cellular stress response pathways. In addition, since circRNAs do not have 5′ and 3′ ends, the circRNAs containing an infinite ORF may also be translated in a rolling cycle fashion to produce a large protein concatemer. Such large protein concatemers have a high tendency to form aggregates due to a large amount of repetitive sequences, and thus may be toxic to the cells. Additional research is required in the future to illustrate the scope of cap-independent translation and the biological functions of these new proteins translated from circRNAs or from alternative ORFs of mRNAs.

24The mechanism by which viral IRESs drive cap-independent translation is well understood; however, cellular IRESs still need more reliable evidences to illustrate their scope and mechanisms. The major issue of cellular IRESs is that their IRES activity is relatively weak when tested with the in vitro translation system, possibly due to the lack of necessary RNA modification and/or the accessory nuclear RNA binding proteins. Therefore, a reliable in vivo system will be critical to examine the activity of cellular IRESs. Based on the previous studies and our knowledge, the circRNA reporter system will be a useful tool to examine the activity of cellular IRESs. Several screen experiments using different in vivo systems suggest that IRES-driven translation is more prevalent than previously expected.

25This work is supported by the National Natural Science Foundation of China (31570823, 31661143031, and 31730110 to Z.W.; 91753135 and 31870814 to Y.Y.). Z.W. is also supported by the CAS Pioneer 100-Talent Program (type A). Y.Y. is also sponsored by the Youth Innovation Promotion Association CAS and Shanghai Science and Technology Committee Rising-Star Program (19QA1410500).

26Conflict of interest: none declared.