Interpreting ribosome dynamics during mRNA translation

Abstract

Translation takes a central position in gene expression, and its swift response to environmental stress is evolutionarily conserved. Upon chemical damage to the messenger RNA (mRNA) or the lack of building blocks, the ribosome stalls during elongation and halts the production line. Even under normal growth conditions, the translation machinery encounters constant hindrances such as varied codon composition or nascent chains with distinct features. However, it is challenging to define these kinetics experimentally, partly due to the inherent variations of ribosome behavior during mRNA translation. To ensure the flow of ribosomal traffic, cells employ several mechanisms to circumvent the traffic jam. When the roadblock is not resolved timely, trailing ribosomes can collide with stalled ribosomes. However, the boundary between physiological queuing and pathological collision is often blurred, representing a fundamental gap in our understanding of ribosome dynamics. To cope with translational barriers, several signaling pathways are activated to adjust the rate of global translation and rescue the local stalled ribosome. Deficiencies in cellular response to translational stress have been associated with a wide array of human diseases. In this review, we focus on fundamental aspects of the ribosome dynamics during mRNA translation. We provide an overview of causes, outcomes, and cellular responses to ribosome stalling and collision on mRNA. We highlight questions that may clarify the biological roles of distinct ribosome behavior during mRNA translation and emphasize the mechanistic connection between altered ribosome dynamics and human diseases.

Following transcription, a messenger RNA (mRNA) begins a journey of processing and transportation toward translating its nucleotide sequence into the amino acid residues of a protein. Translation is divided into four sequential phases: initiation, elongation, termination, and recycling. During initiation, the scanning 43S preinitiation complex identifies the start codon, typically the first AUG, followed by the assembly of the functional 80S ribosome at the start codon. The elongation is a complex polymerization process, where the ribosome moves along the mRNA while reading one codon (three nucleotides) at a time through the action of tRNAs. Aminoacyl tRNAs enter the ribosomal A site via codon: anticodon base pairing, which involves a sampling process to ensure accommodation of cognate tRNAs from the tRNA pool. Amino acids carried by tRNAs are polymerized within the ribosome until a stop codon is encountered. It has been estimated that, on average, an elongating ribosome adds 5 ∼ 10 amino acids per second in Escherichia coli and 3 ∼ 5 amino acids in mammalian cells (1, 2, 3). Along individual mRNAs, ribosomes do not travel at a constant speed but rather in a stop-and-go traffic manner. Besides tRNA availability, many other factors influence the elongation rate, which ultimately impacts the quality and quantity of translational products. In some circumstances, elongation slowdown leads to ribosomal pausing and prolonged stalling causes ribosome collision. Cells respond to ribosome stalling and collision by recruiting distinct effectors to resolve the traffic jam or trigger cell death when the ribosome stall cannot be timely resolved. While our knowledge of translational regulation is steadily increasing, how cells distinguish a physiological pause from a pathological stall remains poorly understood.

The importance of ribosome dynamics on mRNA is not limited to protein production. Altered ribosome dynamics have far-reaching impacts on mRNA stability as well as cellular stress responses. A halted cellular production line often leads to mRNA degradation and triggers stress signaling pathways at cellular and organismal levels. While mRNA decay occurs on individual problematic messengers, the systemic stress response affects a broad range of cellular processes. Very little is known about how cells gauge translational stress and decide between local resolution and global response to cope with the roadblock.

In this review, we discuss current knowledge of molecular mechanisms, cellular responses, and translational outcomes of altered ribosome dynamics along the translated open reading frame (ORF). We focus on both cis-sequence features and trans-factors contributing to distinct ribosome behaviors. We do not cover ribosome-associated mRNA decay pathways that have been extensively reviewed elsewhere (4, 5). We highlight fundamental questions underlying the decoding process, illustrate sources and consequences of altered ribosome dynamics, and suggest how ribosome dynamics could be harnessed to optimize protein production and facilitate the development of therapeutic strategies against human diseases.

Measurements of ribosome dynamics

Quantitative analysis of ribosome dynamics is crucial for elucidating the regulatory mechanisms and functional consequences of translation. Given the heterogeneous coding sequences across the transcriptome, it is challenging to calculate the elongation rate at codon positions inside cells. Even the best estimate does not indicate a specific ribosome dynamics status. Although multiple parameters need to be considered, we are still far from being clear in defining ribosome slowdown, pause, stall, and collision (Fig. 1).

Figure 1.

A wide range of ribosome dynamics during mRNA translation. An elongating ribosome could exhibit different kinetics ranging from slowdown, temporary pause, persistent stall, to collision (top panel). These distinct ribosome stages are context-dependent, and their definition is often ambiguous if judged by the A-site ribosome density (bottom panel). During ribosome pausing, the deacylated tRNA (light blue t) likely dissociates from the E-site due to prolonged dwell time. Upon persistent stalling, additional factors like eIF5A (blue oval) could bind to the empty E-site. When the trailing ribosome collides with the stalled ribosome, the collided ribosome takes on a rotated state and the disome interface can be occupied with collision factors (orange oval). Ribosome collision also leads to elevated A-site ribosome peaks (blue line) upstream of the stall site. The size of single ribosome and disome is shown with light blue lines as codons of ribosome footprints.

Ribosome profiling

By sequencing ribosome-protected fragments (RPFs), ribosome profiling (Ribo-seq) readily shows ribosome positions on individual mRNAs (6, 7). When viewed across the entire transcriptome, the ribosome density map displays remarkable 3-nt periodicity, an indicator of the decoding process. However, the resolution of translational snapshots is subject to variations of sample preparation, cDNA library construction, and sequencing data analysis. For instance, endonuclease incomplete digestion or overdigestion not only affects the RPF pool but also alters 5′ and 3′ end positions (8). During small RNA library construction, adapter ligation could introduce nucleotide bias that distorts the subsequent read distribution (9). Similarly, caution must be taken in applying calculations of offset to infer ribosome positions from sequencing reads with a length different from 28 to 30 nt.

When individual codons are considered, it is generally believed that the ribosome density at each position corresponds to the ribosome dwell time. While the overall ribosome occupancy on individual mRNAs is often used to infer translation efficiency, the ribosome density at individual positions is widely interpreted as decoding time for the A-site codon (10). This feature is best represented at start and stop codons with higher ribosome density relative to the coding sequences (CDS) due to slow initiation and termination, respectively. Among the 61 sense codons, the ribosome exhibits a wide range of density at the A-site with glycine and glutamate codons showing the highest occupancy (11). Notably, the same codon at different positions can display a range of ribosome occupancies, raising fundamental questions about how cis-sequence elements (such as codon contexts) and trans-acting factors (such as RNA binding proteins) coordinate the decoding rate (12).

Regular Ribo-seq can be modified to capture disome footprints, revealing ribosome queuing and collision events where the stalled leading ribosome is followed by a trailing ribosome (13, 14, 15, 16). However, disome formation does not distinguish between ribosome queuing and collision. A recent disome-seq study using mouse liver reported that ∼10% of all translating ribosomes are in a disome state without evident stress response (15). Similarly, queuing of >2 ribosomes leaves longer RPFs than the disome footprints. With this caveat in mind, disome-seq is informative in identifying collision sites across the transcriptome.

Reporter assays

While ribosome profiling is powerful in uncovering slow decoding regions at the global level, those bottlenecks require experimental validation. A commonly used approach is the dual reporter assay, in which the query sequence is flanked by two distinct reporters in the same reading frame (17). The upstream reporter serves as an internal control for mRNA abundance and translation initiation. To avoid the influence of inserts on the readout of fusion proteins, a self-cleaving sequence such as 2A can be inserted into the construct before and after the insert (18). The reporter assay can be further developed into massively parallel reporter assays (MPRAs), which simultaneously evaluate millions of sequence inserts by coupling cell sorting with deep sequencing. Although reporter assays are highly quantitative in assessing the translational output, they cannot reveal mechanisms underlying the altered ribosome dynamics.

Single-molecule imaging

Single-molecule fluorescence methods have provided rich detail on macromolecule interactions. A quantitative single-molecule microscope technology has been developed to visualize translation dynamics inside cells (2, 19, 20). To achieve this, mRNA is labeled with fluorogenic RNA aptamers or RNA-binding proteins such as bacteriophage-derived coat protein. Nascent peptides can be traced by adding epitopes recognizable by single-chain variable fragments (scFVs) such as frankenbodies (21). By monitoring translational events with high spatiotemporal resolution in live cells, single-molecule imaging helps dissect translational bursting (22). More recently, long-term imaging of individual ribosomes translating stopless circular RNAs revealed that translating ribosomes frequently undergo transient collisions (23). Unexpectedly, transient ribosome collisions lead to ribosome cooperation, thereby increasing the translational output. Despite the real-time feature of single-molecule imaging in live cells, it cannot achieve codon resolution. Advances in the development of fluorescence resonance energy transfer (FRET) on single ribosomes permitted examination of the ribosome decoding process in real-time through multiple cycles of elongation (24). The zero-mode waveguide approach, which enables single-molecule fluorescence imaging at high reactant concentrations by confining the illumination volume to the zeptoliter (10⁻²¹ L)scale, allows direct observation of the compositional dynamics of tRNA occupancy on the elongating ribosome, despite its limitation to an in vitro reconstituted system (25).

Ribosome dynamics by tRNA biology

Within the ribosome decoding center, the tRNA anticodon pairs with the codon triplet on mRNA, thereby defining the rate and fidelity of protein synthesis. Therefore, tRNA abundance and functional status play crucial roles in ribosome dynamics (Fig. 2). In the following sections, we discuss how various aspects of tRNA biology influence ribosome dynamics.

Figure 2.

The central role of tRNA in ribosome dynamics. The availability and functional integrity of tRNAs are key determinants of ribosome dynamics during translation. Besides differential gene expressions of tRNA isoacceptors and isodecoders, tRNA molecules are subject to cleavage by endonucleases. Aminoacyl synthetase (ARS)-mediated tRNA charging is central to the decoding process and tRNA modification also modulates decoding fidelity by influencing codon-anticodon base pairing. Amino acids are depicted as orange.

tRNA availability

The composition of intracellular tRNA abundance affects the decoding speed of elongating ribosomes, resulting in varied decoding efficiencies among different codons. The recent advent of high-throughput tRNA sequencing technologies has revealed tRNA abundance at both isoacceptor and isodecoder levels (26). A tRNA isoacceptor refers to a tRNA molecule that carries the same amino acid but has a different anticodon, while a tRNA isodecoder has the same anticodon but differs in its body sequence (27). Among different tissues or upon environmental perturbations, tRNA isodecoders show more differential expression than isoacceptors, suggesting the existence of isodecoder buffering to maintain the isoacceptor pool (28, 29). Not surprisingly, pathogenic mutations causing specific tRNA depletion led to ribosome stalling, as exemplified by tRNA-Arg^UCU mutation in central nervous system (30). Among the five tRNA-Arg^UCU isodecoders, the mutated gene exhibits predominant expression in mouse neurons relative to non-neuronal cells (31). Notably, ribosome stalling at the corresponding arginine AGA codon was exaggerated in the absence of guanosine triphosphate-binding protein 2 (GTPBP2), a ribosome rescue factor binding to the ribosome recycling protein PELO (30).

Endonucleolytic cleavage of some tRNAs also causes an imbalance of tRNA pools. Commonly found in bacteria as toxins (32), tRNA nucleases exist in eukaryotic cells as well. For instance, Schlafen family member 11 (SLFN11) specifically cleaves tRNA-Leu^UAA, resulting in translational inhibition (33). Recently, sterile alpha motif domain-containing 9 (SAMD9) and SAMD9-like (SAMD9L) have been identified as endonucleases targeting tRNA-Phe^GAA (34). As a result, prominent ribosome pausing was observed at both phenylalanine codons. Angiogenin, a member of the RNase A family, cleaves the anticodon loop of tRNAs within the ribosome A-site (35). It is unclear whether angiogenin-mediated tRNA cleavage causes codon-specific ribosome pausing.

tRNA charging

tRNA charging, or aminoacylation, is mediated by specific aminoacyl-tRNA synthetases (ARSs). In mammalian cells, nine different ARSs, including glutamine (QRS), proline and glutamate (EPRS), isoleucine (IRS), leucine (LRS), methionine (MRS), lysine (KRS), arginine (RRS), and aspartate (DRS), assemble with three accessory interacting multifunctional proteins (AIMPs) of p43, p38, and p18 to form a multi-tRNA synthetase complex (MSC) (36, 37). MSC formation has been suggested to channel charged tRNAs to ribosomes for efficient protein synthesis (38). While each ARS in MSC is responsible for the aminoacylation of its cognate tRNAs, depletion of RRS and QRS from MSC did not cause charging loss for tRNA^Arg and tRNA^Gln, and thus there was no ribosome pausing at arginine or glutamine codons (39). Depletion of valyl-tRNA synthetase (VRS) also did not impact the ribosome A-site codon occupancy and global protein synthesis despite its absence in the MSC complex (40). One possibility is tRNA misacylation, which allows ribosomes to move forward with mistranslation of the corresponding codons. For example, a missense mutation of alanyl-tRNA synthetase leads to misacylation from alanine to serine in mouse Purkinje cells (41).

The ribosome elongation rate at a codon depends on recruitment of the aminoacylated tRNA. The charging status can be evaluated by tRNA microarray, which utilizes a periodate treatment that selectively oxidizes uncharged tRNAs and blocks the subsequent fluorophore-labeled oligonucleotide ligation (42, 43). Alternatively, the relative charging ratio can be measured by quantifying the intensity of the bands representing deacylated and acylated tRNAs on northern blotting (44). Although these low-throughput methods remain the gold standard, short-read sequencing and, more recently, nanopore sequencing have been developed as complementary approaches to assess tRNA aminoacylation (45, 46, 47). Amino acid deprivation has been commonly used to reduce charged tRNA levels, but different tRNA isoacceptors exhibit different sensitivity to amino acid limitation. For instance, arginine starvation primarily reduces the charging status of tRNA-Arg^ACG, resulting in ribosomal pausing at CGC and CGU codons (48). Notably, the translational effect is intertwined with amino acid response. Unlike arginine starvation, leucine deprivation does not lead to discernable ribosome pausing at leucine codons partly because of reduced translation initiation by suppressed mTORC1 signaling pathways.

tRNA modification

More than 110 modifications have been identified on tRNAs, some of which can be dynamically reversed (49). N⁷-methylguanosine (m⁷G) modifications at position 46 in the variable loop are one of the most prevalent tRNA modifications found in several tRNAs (50). Depletion of the responsible methyltransferase complex composed of methyltransferase-like 1 (METTL1) and WD repeat domain 4 (WDR4) results in moderate ribosome pausing at some of the m⁷G-tRNA-decoded codons (51, 52, 53, 54). Methylation at the wobble base is thought to be important for translational fidelity via correct codon-anticodon pairing. DNA methyltransferase 2 (DNMT2), also known as tRNA-aspartic acid methyltransferase (TRDMT1), catalyzes methylation of cytosine at position 38 (m⁵C38) of tRNA-Asp^GUC, tRNA-Gly^GCC, and tRNA-Val^ACC. In mouse primary bone marrow cells lacking Dnmt2, mRNAs undergoing translational downregulation showed increased A-site occupancy at codons corresponding to Dnmt2 substrate tRNAs (55).

The impact of tRNA modifications on ribosome dynamics also holds true in the mitochondrial translation system. Silencing the mitochondrial folate enzyme serine hydroxymethyltransferase 2 (SHMT2) decreases the level of taurinomethyluridine modification (τm⁵U) at the wobble position of mt-tRNAs, resulting in mitochondrial ribosome stalling at lysine AAG and leucine UUG codons (56). Depletion of methyltransferase-like 8 (METTL8), a mitochondrial writer protein responsible for 3-methylcytidine (m³C) modification at position 32 of mt-tRNA-Ser^UCN and mt-tRNA-Thr, causes ribosome stalling at P-site serine codons and A-site threonine codons, respectively (57). These methyltransferases are crucial for the integrity of mitochondrial oxidative phosphorylation complexes.

Ribosome dynamics by mRNA features

As the ribosome moves along the mRNA template, the protein production factory must accommodate distinct sequence features to ensure the robustness of the production process. These sequence features are inseparable from tRNA “assembly workers” mentioned above and directly contribute to varied ribosome dynamics (Fig. 3).

Figure 3.

Intrinsic mRNA features influence ribosome dynamics. The rate of translation elongation is modulated by codon optimality. Optimal codons (blue) are typically translated more efficiently, accurately, and rapidly than the non-optimal ones (orange). mRNA modifications and secondary structures affect elongation rates directly or indirectly. Certain nascent peptides enriched in specific amino acids can significantly influence ribosome dynamics during translation.

Codon optimality

In many organisms, highly expressed genes tend to contain frequently used codons. A popular hypothesis is that codon usage controls the speed of translation elongation. Codon optimality is a concept that reflects a balance between the supply of charged tRNAs and the demand determined by codon usage. Although the codon frequency alone is not sufficient to define optimality, it correlates with codon optimality, particularly in highly expressed genes. It is generally believed that optimal codons are decoded by the abundant cognate tRNAs more efficiently, accurately, and faster than non-optimal codons (58). Consistent with this notion, live cell imaging of ribosome movements on single mRNA molecules revealed faster elongation rates on codon-optimized reporter mRNAs than non-optimal ones in mammalian cells (2). However, the presence of non-optimal codons could be beneficial in co-translational events via elongation slowdown. The codon optimality is also position-dependent, with sub-optimal codons tend to be enriched at the beginning of CDS, likely reducing the traffic jam of ribosomes after initiation (59). It should be noted that the effects of codon usage on ribosome behavior are organism- or tissue-specific. The CGG- and CGA-encoded polyarginine stretches induce strong ribosome stalling in yeast but not in mammalian cells (60, 61). Intriguingly, genes containing non-optimal codons are efficiently translated in mouse brain and testis, suggesting the existence of regulatory factors in codon usage (62).

The codon optimality on ribosome dynamics is not limited to the A-site. Glutamate and aspartate codons preceded by proline-proline or glycine-proline dicodons tend to show high A-site occupancy, suggesting the subtle coordination between the P-site and A-site tRNAs during elongation. P-site proline, glycine, and aspartate are among the most influential stalling sequences conserved across different organisms (63). Interestingly, codon pairs could impact the elongation rate in different ways. For example, the codon pair CUC-CCG, both of which are rare and non-optimal codons in yeast, strongly inhibits translation, but the same pair in the reverse order (i.e., CCG-CUC) is no longer inhibitory (64).

mRNA structure

With the improved resolution of mRNA structure measured inside cells, mRNA secondary structures within the ORF have been suggested to influence translation elongation (65). Although elongating ribosomes are capable of unwinding downstream structures, local mRNA structures are thought to slow down ribosomes and cause transient pausing (66, 67). The structural hindrance also comes from interactions between mRNA and the ribosome. While there are several mRNA-ribosomal protein interactions discovered in the context of initiation pausing, those involved in elongation pausing remain to be fully characterized. Cross-linking experiments revealed the proximity between the ribosomal protein eS26 and mRNA near the ribosome exit tunnel, which is likely to contribute to the Kozak sequence-dependent initiation (68). The ribosomal protein uS19 has been shown to interact with mRNAs at glutamate and lysine codons at high frequency when the ribosome pauses under cycloheximide treatment (69). Additionally, the highly conserved 3′ terminal sequence of 18S rRNA has the potential to interact with mRNA, thereby influencing the elongation rate. This possibility partly explains the wide range of ribosome occupancy at the same codon in different positions.

mRNA modification

mRNA is subject to modification with N⁶-methyladenosine (m⁶A) as the most abundant form. The asymmetric distribution of m⁶A across the mRNA complicates the translational effect. In vitro single molecule FRET study demonstrated that m⁶A slows down the ribosomal decoding process (70). Consistent with this notion, ribosome profiling studies revealed that m⁶A sites are associated with elevated ribosome density (71, 72). However, the presence of m⁶A tends to disrupt RNA secondary structures, thereby promoting ribosome movement. The m⁶A reader protein YTHDC2 contains RNA helicase activity, and its binding helps resolve the translational roadblock. Given the development of programmable m⁶A modification at specific sites, functional characterization of individual m⁶A sites is now possible to tease out their translational impacts (73).

Nascent peptide

The ribosome large subunit has a polypeptide exit tunnel accommodating 30 to 50 amino acid residues of the nascent polypeptide chain. In vitro translation experiments showed that positively charged amino acids, such as arginine or lysine residues, slow down elongation through the interaction with the negatively charged exit tunnel (74). However, this is not the case in mammalian cells. Translation of poly-lysine is linked to the decoding of poly(A) sequences on mRNA, which is one of the best-characterized sequence motifs that cause strong ribosome stalling. A recent study reported that both the poly(A) sequences and poly-lysine residues contribute to ribosome slowdown by structural rearrangement of ribosomes (75). The 4 A nucleotides on the mRNA rearrange the decoding center by flipping out the 18S rRNA base positions that impede the delivery of the tRNA-eEF1A-GTP ternary complex. In the meantime, the lysine residues in the exit tunnel mispoint the peptidyl-tRNA to alter the peptidyl-transferase center that impairs peptide bond formation. These mechanisms explain why short poly(A) sequences or AAG-encoded poly-lysine sequences do not cause ribosome stalling (76).

Among the 20 amino acids, proline is unique because of its reactive amine found within a five-membered ring, which makes tRNA-Pro a poor peptidyl acceptor in the A-site. Additionally, owing to entropic constraints, proline is not an ideal donor substrate in the P-site either. Not surprisingly, translation of poly-proline is kinetically slow. However, eukaryotic cells employ the translation factor eIF5A (EF-P in E. coli) to overcome this hurdle by promoting peptide formation via E-site binding (77). Recent studies uncovered a much broader role of eIF5A in promoting the synthesis of different peptide bonds, including the release of nascent chains during termination (78).

Translational outcomes of altered ribosome dynamics

Physiological pausing can be employed for regulatory purposes, and controlled translational delays are used to help co-translational folding of the nascent polypeptide on the ribosome. By contrast, non-productive translational obstacles can be temporarily solved by emergency “bailing out” procedures, including frameshifting, codon reassignment, codon bypassing, and translational abandonment (Fig. 4).

Figure 4.

Translational outcomes of altered ribosome dynamics. To ensure the flow of ribosomal traffic, cells employ several mechanisms to overcome the non-productive stalling, such as ribosomal frameshifting, codon reassignment, codon bypassing, or translation abortion. In many cases, these translational products are non-functional and subject to degradation. While mRNA is depicted in light blue, protein products are shown in dark green. A light green line indicates a different frame. An orange dot refers to an altered amino acid, whereas the sign of “v” indicates the missing amino acid. Dark triangle, start codon; dark square, stop codon.

Frameshifting

Under the limitation of cognate tRNAs, the paused ribosome with an empty A-site tends to undergo frameshifting to resolve the roadblock. This so-called hungry frameshifting often occurs in bacteria. For example, tRNA^Lys depletion caused −1 frameshifting at lysine(AAG) codons especially when the 3′ overlapping codon was AGC or AGU (79, 80). In human cells, cytochrome c oxidase 1 (MTCO1) and NADH-ubiquinone oxidoreductase core subunit 6 (MTND6) are mitochondria encoded proteins produced by translational termination at AGA and AGG codons by −1 frameshifting, respectively (81). In cancer cells, tryptophan (W) depletion induced by interferon-γ-dependent indoleamine 2,3-dioxygenase 1 (IDO1) inhibition caused W-bumps, an accumulation of ribosomes downstream of tryptophan codons. This is largely ascribed to ribosome pausing and frameshifting at tryptophan codons, leading to aberrant protein production (82).

A recent study has shown that a repeat of ∼5 codons could trigger frameshifting in a human cell line (83). The simultaneous frameshifting occurs on 46 out of 61 codons, with phenylalanine codons the most potent. Frameshifting has also been observed in pathological codon expansions such as CAG repeats in huntingtin and CGG repeats in Fragile X-associated tremor/ataxia syndrome (FXTAS) (84, 85, 86, 87). Although the underlying mechanism remains unclear, −1 frameshifting at CAG repeats in Huntington’s disease can be partially explained by the decrease in charged tRNA-Gln^CUG levels due to the excessive translation of expanded CAG codons (84).

Codon reassignment

The universal genetic code is not absolute, and some organisms have established unique decoding rules during evolution (88). The context-specific codon reassignment could be derived from misincorporation of tRNA into the A-site or tRNA misaminoacylation by the ARS. For instance, the alanyl-tRNA synthetase mutation impairs proofreading activity during aminoacylation and causes alanine-to-serine misincorporation in mouse neurons (41). A recent study reported tryptophan-to-phenylalanine substitution in human cancer cells, in which tryptophan depletion causes misacylation of tRNA-Trp^CCA (89). Despite sporadic cases reported in mammalian cells, the scope of codon reassignment upon ribosome stalling remains to be determined.

Codon bypassing

Another strategy for a ribosome to overcome the non-productive stalling is to skip the difficult-to-decode codon, resulting in codon bypassing. However, this non-canonical translational event is rare as it requires coordination between cis-regulatory elements and trans-acting effectors. The best-characterized example is bacteriophage T4 DNA topoisomerase gene product 60 (gp60), which undergoes ribosome “take-off” and “landing” (90). Similar events have been observed in yeast mitochondria, but not yet in cytosolic ribosomes (91).

Aborted translation

Pausing-induced translational abandonment often occurs in bacteria via the process of trans-translation mediated by transfer-messenger RNA (tmRNA) (92). Eukaryotic cells evolved a different mechanism by catalyzing the formation of C-terminal alanine and threonine (CAT) tails on stalled nascent chains (see below). Consecutive negatively charged residues in nascent chains can stochastically induce discontinuation of translation, in a phenomenon termed intrinsic ribosome destabilization (IRD) (93). Originally reported in bacteria, IRD also occurs in eukaryotic cells, translating acidic residues-enriched nascent peptide (such as glutamate or aspartate) (94). Interestingly, yeast ORFs largely avoid negatively charged amino acid clusters in their N-terminal regions to minimize the interrupted translation.

Cellular response to altered ribosome dynamics

A growing body of evidence suggests that ribosomes serve as molecular sensors of cellular stress. In response to translational obstacles, multiple signaling pathways are activated to resolve such situations, globally and locally (Fig. 5). It remains an open question how different ribosome surveillance mechanisms are coordinated to ensure cellular homeostasis.

Figure 5.

Cellular response to altered ribosome dynamics. In response to ribosome pausing and collision, cells activate several signaling pathways to cope with the translational hinderance. While single ribosome pausing or stalling triggers integrated stress response (ISR) and ribotoxic stress response (RSR), ribosome collision exacerbates such global responses. Ribosome collision also induces ribosome collision response (RCR) or ribosome-associated quality control (RQC) pathways to clear the roadblock by acting locally, resulting in ribosome splitting, nascent chain degradation and mRNA decay.

Integrated stress response (ISR)

ISR is an evolutionarily conserved signaling pathway for cellular adaptations to environmental changes (95, 96). The core of ISR is phosphorylation of eukaryotic initiation factor 2α (eIF2α), a regulatory subunit of the eIF2 complex. The phosphorylated eIF2α blocks GDP-to-GTP exchange mediated by guanosine nucleotide exchange factor (GEF) eIF2B, which is necessary for the ternary complex (TC) formation and subsequent translation initiation. At least four kinases mediate eIF2α phosphorylation in response to distinct cellular stresses. General amino acid control nonderepressible 2 (GCN2) serves as a nutrient sensor by binding to uncharged tRNA molecules upon amino acid deprivation. However, recent studies demonstrated that GCN2 could be directly activated by stalled ribosomes (30). By limiting ribosome loading, ISR appears to be the first line of global response to mitigate ribosome collisions.

Ribotoxic stress response (RSR)

In metazoans, ribosomal impairment also integrates into general stress response pathways by activating mitogen-activated protein (MAP) kinases such as p38 and JNK (97). The MAP kinase cascade is a large kinase network that regulates a plethora of biological processes in response to intra- and extracellular stimuli. Originally described in 1997, the RSR is a MAP kinase signaling cascade mounted in response to defective ribosomes caused by translation inhibitors, ribotoxins, chemotherapeutic reagents, and RNA damages (98, 99, 100). One of the upstream signaling kinases responsible for RSR activation is sterile alpha motif and leucine zipper containing kinase (ZAK) (101, 102). Among the two alternative splicing isoforms, the longer one, ZAKα, plays an important role in ribotoxic stress response (RSR) (103). Recent studies reported that stalling of ribosomes is sufficient to activate ZAKα, which senses structural conformations of stalled ribosomes via binding to the ribosome exit site of the mRNA channel (103, 104). Unlike ISR, ZAKα-mediated RSR activation triggers cell-cycle arrest and apoptosis via p38 and JNK, respectively.

Ribosome collision response

Upon ribosome collision, the interface created by collided “disomes” can be recognized by ribosome quality control factors to facilitate clearance of the roadblock and recycling of the stalled ribosome. The E3 ubiquitin ligase ZNF598 (Hel2 in yeast) is proposed to sense collided ribosomes and ubiquitinate ribosomal proteins such as RPS10 (eS10) (61, 105, 106). However, a recent work using in vitro assays demonstrated that ribosome collision is not a prerequisite for ZNF598-mediated ribosome ubiquitination (107). It is thus possible that single stalled ribosome could be sensed and cleared by surveillance pathways. Supporting this notion, the E3 ubiquitin ligase Mag2 and Fap1 target initiating ribosomes arrested at the start codon, although it is possible that a scanning ribosome could collide with the initiating ribosome (108). Whether ribosome stalling response and collision response use distinct pathways remains unclear.

Quantitative proteomics from emetine-induced ribosome collision have revealed another sensor of collided ribosomes: endothelial differentiation-related factor 1 (EDF1) (109, 110). Structural analysis showed that EDF1 binds to the mRNA entry channel and RPS3 (uS3) of the trailing ribosome, which is close to the collision interface (109). The abundant EDF1 recruits GIGYF2 to repress translation initiation via eIF4E-homologous protein 4EHP, thereby preventing continuous translation on problematic mRNAs. While EDF1 recruitment by collided ribosomes is independent of ZNF598, ZNF598-mediated ribosomal ubiquitination is facilitated by EDF1, suggesting that EDF1 serves the early responder to ribosome collision.

Ribosome-associated quality control (RQC) pathway

A dedicated surveillance mechanism evolved in eukaryotic cells to rescue the stalled ribosome by actively splitting ribosomal subunits and degrading incomplete protein products. Pelota and HBS1L (Dom34-Hbs1 in yeast) are responsible for ribosome dissociation during stalling using a mechanism analogous to termination, where stalled ribosome complexes are dissociated and peptidyl-tRNAs are released (111). Another ribosome splitting factor is activating signal cointegrator 1 complex (ASCC), whose action is ZNF598-dependent (112). Depletion of ZNF598 or ASCC enables ribosomes to slow down but read through the poly(A) sequences, which are known to cause ribosome stalling. Unlike normal termination, the dissociated 60S subunit remains attached to a nascent chain-tRNA conjugate, which is sensed by the RQC subunit NEMF (Rqc2 in yeast) and a ubiquitin ligase Ltn1 (113). While NEMF catalyzes the formation of C-terminal alanine and threonine (CAT) tail, Ltn1 ubiquitinates the emerging nascent chain whose extraction is facilitated by p97/VCP (114, 115). Prior studies identified ankyrin repeat and zinc finger domain-containing protein 1 (ANKZF1) (Vms1 in yeast) as endonucleases to cleave 3′-terminal CCA of the peptidyl-tRNA to promote the release of nascent peptides from the obstructed 60S subunit (116, 117).

Pathological ribosome dynamics in human diseases

Neurodegenerative disease

A growing body of evidence suggests that aberrant translation is a common feature across multiple neurogenerative diseases (118). ARS mutations have been reported in several peripheral nervous system disorders (119). At least six ARS mutations have been identified in Charcot-Marie-Tooth (CMT) disease, a diverse group of peripheral neuropathy disorders characterized by the degeneration of both motor and sensory nerves (120, 121, 122, 123, 124, 125). One potential mechanism involves slowed elongation rates, as one of the pathogenic GARS variants retains aminoacylation activity but shows ribosome stalling at glycine codons, triggering integrated stress response through eIF2α phosphorylation (126).

The genomic expansion of tandem repeats has been linked to at least 50 human diseases, especially those affecting central nervous system. Many triplet repeats encode glutamine and alanine, and translating such homopolymeric sequences is challenging for ribosomes and could result in pausing and/or frameshifting. For example, glutamine CAG expansions of huntingtin (HTT) and ataxin 3 (ATXN3) generate polyalanine peptides as well as in-frame polyglutamine (84, 127, 128, 129, 130). How exactly altered ribosome dynamics contributes to the disease onset and progression remains poorly understood. A recent study reported that mutant HTT inhibits global protein translation by promoting ribosome stalling on specific mRNAs (131). Lack of GTPBP2 in the background of tRNA-Arg^UCU mutations causes ribosome stalling at arginine AGA codons in a mouse model of neurodegenerative disease (30, 132). It is unclear why pathogenic repeats preferentially affect neurons despite their ubiquitous expression across tissues (133). Similarly, disruption of the downstream RQC pathway components LTN1 and NEMF is associated with severe neurodegeneration in mice and human patients. The underlying pathology appears to result from accumulation of 60S-nascent chain complexes and proteotoxicity.

ISRIB, a small-molecule ISR inhibitor, promotes the assembly of the active eIF2B complex, is under early-stage human trials for several neurological diseases that are characterized by chronic eIF2α phosphorylation (134, 135). Similarly, a selective inhibitor of a regulatory subunit of protein phosphatase 1 (Sephin1) has already been designated by the FDA as an orphan drug for CMT by blocking eIF2α dephosphorylation mediated by GADD34-PP1 (136, 137).

Cancer

In cancer, translation is commonly upregulated to satisfy the increased anabolic demands associated with malignant transformation and tumor growth. As a result, efficient translation with less ribosome pausing could be a driving force of tumor growth. Relaxed initiation pausing via reduced m⁶A modification on mRNAs facilitates oncogenic translation (138). Similarly, tRNA-Arg^UCU upregulation leads to efficient translation of specific transcripts involved in cell cycle regulation, which are enriched with AGA codons (139). Following a similar line, ribosome pausing is linked to the vulnerability of tumors to specific amino acids (140). Comparing codon occupancy between tumors and normal cells identified proline vulnerability in kidney cancer. Downregulation of pyrroline-5-carboxylate reductase 1 (PYCR1), the enzyme catalyzing the last step of proline synthesis from glutamate, showed therapeutic potential in treating kidney cancer. A similar approach, when applied to human breast epithelial cells, uncovered ribosome stalling at leucine codons in response to transforming growth factor-β1 (TGFβ1), which reflects the downregulation of a leucine transporter (141).

Ribosome stalling could also trigger apoptosis in response to cancer treatments. A recent study reported that DNA damage induces p53-independent apoptosis via SLFN11-mediated ribosome stalling at leucine UUA codons (99). This mechanism explains why cancers with p53 mutations could still undergo apoptosis by the therapeutic compounds damaging their DNAs. Although many of the existing therapeutic approaches for cancers target translation initiation, a few strategies were developed to modulate elongation. It is conceivable that inducible ribosome stalling and subsequent stress response could be exploited as therapeutic strategies against tumorigenesis.

Infectious disease

The host translation machinery is a common target of viral and bacterial pathogens to establish infection. How viruses hijack the host protein synthesis machinery has been extensively reviewed elsewhere (142, 143). Although well-known mechanisms are centered on the rate-limiting initiation step, recent studies showed that translation elongation can also be targeted under viral and bacterial infection. For instance, protein effectors secreted by an intracellular bacteria Legionella pneumophila serve as elongation inhibitors targeting eukaryotic elongation factor 1A (eEF1A) (144, 145). A recent study showed that SidI secreted from L. pneumophila functions as a tRNA mimic that glycosylates the host ribosomes, causing ribosome stalling, RSR activation, and host cell death (146). In addition to pathogenic effector proteins, the host defense system could also modulate translation to restrict viral infection. Sterile alpha motif domain-containing 9 (SAMD9) and SAMD9-like (SAMD9L) are paralogs of interferon-stimulated genes that ubiquitously express in human tissues. Upon activation by poxvirus infection, SAMD9/9L serve as antiviral factors by repressing global protein synthesis. Mechanistically, SAMD9/9L specifically cleaves tRNA-Phe^GAA, resulting in ribosome pausing at phenylalanine codons (34). The subsequent ribotoxic stress inhibits poxvirus replication and cellular proliferation.

Metabolic disease

Organisms experience considerable fluctuation in food availability, necessitating an ability to store energy when resources are abundant and consume it when resources are scarce. Ribosome impairment has recently been implicated as key signaling intermediates in the context of metabolic adaptation. Mice with GCN2 knockout (KO) decreased blood glucose levels and increased insulin sensitivity in response to high-fat diet (HFD) while exhibiting no overt phenotypes under normal chow (147, 148, 149). Similarly, ZAK^−/− mice exhibit an overall lean phenotype and protection against glucose intolerance and liver steatosis induced by a high-fat high-sucrose diet (104, 150). The phenotypes of Zak KO and Gcn2 KO mice highlight the crucial role of RSR and ISR in metabolic homeostasis.

Aging

The RQC pathway is tightly linked to cellular proteostasis, which is often declined during aging. A recent study reported that aging is associated with ribosome pausing at polybasic stretches in both Caenorhabditis elegans and Saccharomyces cerevisiae (151). Nascent polypeptides exhibiting age-dependent ribosome pausing in C. elegans were strongly enriched among age-dependent protein aggregates. During aging, male mice show metabolic decline, including insulin resistance (152). Similar to the protection against obesity-induced metabolic dysfunction, ZAK−/− mice were also protected from the aging-related metabolic decline (150). Although the altered ribosome dynamics are a crucial conductor of the ageing process, the original causes of age-dependent ribosome pausing remain to be elucidated.

Conclusion and perspectives

It has been more than 60 years since the genetic code was deciphered. Understanding the genetic design principles that determine protein production remains a major challenge. Different nucleotide sequences encoding an identical protein sequence can have tremendous variations in protein production levels. Apart from naturally occurring fluctuations in translation speed, ribosomes can encounter obstacles that lead to a permanent arrest, requiring disassembly by a dedicated machinery. Our current knowledge of ribosome surveillance pathways is largely based on artificial reporters bearing stalling sequences, chemically-induced global RNA damages, and non-specific elongation inhibitors (98, 109, 110). It is still an open question to what extent such translational abnormalities are representative of physiological translational challenges that occur naturally.

Perhaps the most fundamental question is how cells distinguish between physiological pausing and pathological stalling. An attractive model is that the abundance and duration of collisions functions as a molecular rheostat, although very little is known how the molecular timer is embedded in cellular surveillance pathways (98). Another model is based on ribosome conformations with the A-site occupancy dictating slow decoding or problematic decoding. The disome interface appears to be the most apparent feature for collision surveillance (109), which could serve as a signal of problematic decoding that needs to be resolved by quality control pathways. However, disome-seq revealed remarkable 10% of all translating ribosomes in a disome state under physiological conditions without triggering ISR or RSR (15). On the other hand, ribosome collisions are not strictly required for downstream responses like RSR. More strikingly, a recent study using single molecule imaging suggests that transient collision promotes elongation via ribosome cooperativity (23). Whether this finding from artificial reporters reflects physiological phenomena of native genes remains to be seen.

A general limitation of studying ribosome dynamics on endogenous mRNAs is the complexity in detecting “weak signals” from relevant factors within sequences that underwent optimization during billions of years of evolution. Among different genes, all 64 codons are distributed unevenly, forming the basis of selective translation in response to translational perturbation. GCN2-mediated ISR is a well-established cellular response pathway to amino acid starvation. While global translation is suppressed, many stress genes are upregulated at the translational level. One example is the selective translation of ATF4, whose non-canonical initiation mechanism has been extensively studied (153). Since selective translation of stress genes also involves elongation, very little is known about whether ATF4 translation overcomes the elongation hurdle formed by stalled ribosomes.

Our limited understanding of the fundamental rules in ribosome dynamics remains a significant challenge for its applications, especially for synthetic biologists trying to construct designer genes with optimized production. Where to slow down the ribosome along the mRNA and how much delay is desirable has become a challenging task in fine-tuning the quality and quantity of translational products. Given the stress signaling pathways initiated from the stalled ribosome, modulating ribosome dynamics has the potential to mitigate neurodegeneration and achieve metabolic homeostasis. On the other hand, ribosome stalling could be harnessed to treat cancers, although there is a long way to go to achieve codon-specific and gene-specific ribosome collision. Supported by high-throughput analysis of synthetic gene libraries and state-of-the-art sequencing techniques, machine-learning approaches may help to further elucidate unknown features and principles underlying ribosome dynamics.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Ronald Wek