Translon Symposium

Ribosome profiling has revealed the extensive utilization of multiple start codons in eukaryotic mRNAs, including in humans. Translation initiation at these start codons results in the synthesis of proteoforms with alternative N-termini and also in translated regions (translons) not annotated as protein coding. Some of such enigmatic translons encode microproteins and immunopeptides while many function as critical elements of translation control that sense cellular environment. This regulatory function of translons is often independent of the polypeptide products encoded by them.

The emerging picture of RNA translation is in striking contrast with the current data structures for annotating protein coding genes in which a sequence of an RNA molecule could have only a single protein-coding sequence (CDS). To address this limitation, we developed a new abstract representation of mRNA translation termed Ribosome Decision Graphs (RDGs). RDGs allow for a biologically realistic representation of eukaryotic translation complexity, focusing on locations critical for translation regulation. RDGs can be used to depict the mutual organization of translons reflecting the interplay between their translation. RDGs can be used in the analysis of ribosome profiling data to identify critical regulatory events. They also can be used for the interpretation of genetic variants, enabling a mechanistic understanding of pathogenic variants occurring outside of CDS regions.

The cellular response to stress often includes translational downregulation, which is widely attributed to phosphorylation of eIF2α. Studies of UV-induced stress responses in fission yeast prompted a re-examination of the relationship between eIF2α phosphorylation and global translational repression. In this talk, I will discuss observations from fission yeast and mammalian systems that challenge this simple model of translation control, briefly touch upon emerging non-canonical links between stress signalling and translation, and present evidence linking translational responses to stress with actin cytoskeleton dynamics.

In BAG1 (BCL2-associated athanogene) mRNA, at least three start codons initiate translation within the main protein coding frame, generating protein isoforms with alternative N-termini. The efficiency of start site selection determines the relative abundance of these isoforms, which differ in their subcellular localisation and binding partners. Regulation of start site selection in BAG1 mRNA has been extensively studied under normal and stress conditions, revealing a complex interplay between RNA secondary structure and translon architecture.

Leveraging the recent wealth of aggregated ribosome profiling data, we identify a role for a 16-codon translon (nested within the 5′ region of the sequence encoding the longest BAG1 protein isoform, and upstream of the sequence encoding the shorter isoforms) in selection of translation initiation sites within the BAG1 protein coding frame. Aggregated ribosome profiling data reveal extensive translation of this short translon from the AUG most proximal to the 5′ end of BAG1 mRNA. Notably, the sequence conservation of this translon exceeds that of the BAG1 coding sequence across the first part of the coding region.

We show that translation of the conserved 16-codon translon exerts complex effects on start site selection in BAG1 mRNA. While its translation inhibits initiation at AUG1 in the BAG1 encoding frame, it also facilitates translation from AUG2 in the BAG1 encoding frame through reinitiation. In addition, the short translon appears to have a stimulatory effect on upstream CUG-initiated translation, as evidenced by TCP-seq data.

Taken together, we have characterised a highly conserved, efficient translon and we present evidence for its role in coordinating start site selection in BAG1 mRNA both upstream and downstream of itself.

Programmed ribosomal frameshifting expands coding capacity by allowing ribosomes to switch reading frame during translation. However, the sequence features that determine whether a potential frameshift site is utilized remain poorly understood. Here, we characterize two novel +1 frameshift signals in yeasts. First, we investigate a frameshift site identified at the KLMX_30357 locus in Kluyveromyces marxianus, which promotes efficient +1 frameshifting when tested in Saccharomyces cerevisiae. Mutational analyses revealed that efficient frameshifting depends not only on the heptameric shift site but also on a downstream enhancer element comprising a conserved di-proline motif and a predicted RNA secondary structure. Second, we describe a hybrid frameshift heptamer, GCG_A.GGC, combining features of the canonical TY1 and TY3 frameshift signals. In addition, genome-wide analysis of known frameshift heptamers and ribosome profiling data revealed numerous instances in which predicted frameshift signals do not exhibit detectable frame switching. Together, these findings demonstrate that the presence of a frameshift heptamer alone is insufficient to drive efficient recoding and highlight the critical role of local sequence context in determining frameshift utilization.

In addition to standard mechanisms, translation regulation also involves non-canonical events, including non-AUG initiation, stop-codon readthrough or programmed ribosomal frameshifting. These events result in the the production of protein variants with different characteristics, such as subcellular localization or function, thus expanding the translatome and the proteome diversity and regulating protein synthesis in different cellular or environmental conditions. The identification of such rare non-canonical events, which are more common than previously predicted, has been enabled by the availability of a growing number of Ribo-seq datasets.

Analysis of large Ribo-seq dataset collections for yeast Saccharomyces cerevisiae using the RiboCrypt browser revealed several unexpected findings. Notably, we observed a characteristic signature of translation termination on standard stop codons within the ORFs. The presence of specific termination signature enabled the detection of potential translation termination events in many protein-coding genes. Using standard biochemical assays, we have observed premature ribosome dissociation occurring in some cases with an efficiency as high as 40% and most often at positions near-cognate to stop codons. Our preliminary results suggest that the mechanism responsible for this phenomenon does not depend on ribosome release factors but to some extent is linked to Ribosome-associated protein Quality Control (RQC) mechanism.

The human transcriptome contains thousands of previously unannotated translons, yet functional validation lags behind the Ribo-Seq evidence for their translation. Standard CRISPR-Cas9 screens are poorly suited to these compact ORFs, where few targetable sites exist and frameshifting indels disrupt them stochastically and unpredictably. Base editing overcomes this by introducing precise point mutations - disrupting start codons or creating premature stop codons - without double-stranded DNA breaks. We first leveraged almost 4,000 Ribo-Seq libraries (Ribocrypt, protected repositories, and in-house data) to map human translation with tissue specificity and isoform resolution. Translons were called robustly using three independent callers with quality-feature-aware filtering and independent scoring, enabling an in-depth survey of contextual expression, evolutionary conservation, and co-expression via network analysis. We designed and optimised a versatile base-editing library targeting non-canonical start codons, benchmarking base editors and sgRNA design strategies for editing efficiency in a pilot screen, and scaled this to a targeted survey of all start codons across human tissues. Together, this pairs a transcriptome-wide translation map with a scalable platform for systematic functional dissection of translons.

In the cell, proteins begin to fold co-translationally as they emerge from the ribosome's exit tunnel during biosynthesis. This process is fundamental to cellular homeostasis; disruptions can drive protein misfolding, which is implicated in severe pathologies including cystic fibrosis, alpha-1 antitrypsin deficiency, and a wide spectrum of protein aggregation diseases. Despite its critical importance, co-translational folding (co-TF) remains poorly understood, as capturing high-resolution structural data using experimental methods alone is a significant challenge. Consequently, integrating accurate computational techniques with experimental data is crucial. Furthermore, because most of our current understanding stems from E. coli, there is a significant knowledge gap regarding evolutionary variations in ribosome structure and their impact on co-TF. In my talk, I will describe an integrative structural biology approach to study co-TF, combining molecular dynamics (MD) simulations with structural restraints from NMR data and cryo-EM maps. I will also demonstrate how we can use bioinformatics and MD simulations to analyse ribosome heterogeneity and its influence on nascent chain pathways inside the ribosomal tunnel. Ultimately, this research sheds light on the intricate dynamics of co-TF and paves the way for a deeper understanding of how evolutionary variations in the ribosome shape protein folding across different organisms.

Circular RNAs (circRNAs) are endogenous RNA transcripts characterized by intrinsic resistance to exonucleolytic degradation, resulting in prolonged intracellular stability compared to linear mRNA. Although endogenously expressed circRNAs are typically non-coding, flaws in statistical analyses may have in some cases falsely classified a subset of circRNAs as protein-coding, underscoring the importance of rigorous evaluation of the coding potential of non-canonical RNA species.

However, recent advances in circRNA engineering have enabled effective cap-independent translation of circRNAs and positioned the circRNA technology as a promising next-generation therapeutic modality capable of durable protein expression at low doses.

At Circio, we are developing circVec, a modular gene expression platform engineered to enhance payload expression through intracellular generation of protein-coding circRNA. circVec leverages spliceosome-mediated circRNA biogenesis from viral and non-viral DNA vectors, enabling sustained gene expression through the intrinsic stability and exonuclease resistance of circRNA transcripts. Systematic optimization of circVec architecture identified key regulatory elements that markedly improve expression efficiency. The latest-generation construct, circVec4.0, achieves up to 37x higher protein expression compared with early constructs and outperforms conventional linear mRNA expression vectors (mVec) in both expression magnitude and durability.

In vivo, circVec enables robust, dose-dependent enhancement of AAV transgene expression across tissues, serotypes, and promoters. In both heart and eye, circVec-AAV vectors yield markedly higher luciferase expression than mVec-AAV across all dose levels with a significant increase of RNA steady state levels. These findings indicate that the enhanced expression is driven primarily by increased RNA stability and transcript accumulation rather than improved transduction efficiency. Moreover, circVec shows higher on-tissue expression, likely due to reduced circRNA stability in the liver, and despite the enhanced expression profile, cellular stress pathways, such as the unfolded protein response (UPR), are unaffected or less activated compared to the benchmark mVec-AAV.

The therapeutic efficacy of adeno-associated virus (AAV) gene therapy is often constrained by inefficient and variable transgene expression, necessitating high clinical doses which can cause severe toxicities and narrow the therapeutic window. However, by using circVec-based AAV gene expression as a next-generation gene therapy platform, may enable substantial transgene expression enhancements with reduced cellular stress and dose reduction potential, supporting the development of safer and more effective therapies for rare and chronic diseases, particularly within the heart and eye.

TENT5 proteins function as non-canonical poly(A) polymerases that selectively re-adenylate mRNAs encoding secreted proteins, enhancing their stability and translation. Following our previous work on TENT5 function in immunity, we investigated how TENT5A influences the efficacy of mRNA therapeutics. Using primary macrophages and vaccines encoding viral and parasitic antigens, we assessed mRNA poly(A) tail dynamics, antigen output, and immune response. We also evaluated the spatial dependency of re-adenylation in the endoplasmic reticulum (ER) by disrupting the interaction with FNDC3, a docking protein critical for TENT5A localisation. Upon delivery, TENT5A-dependent re-adenylation enhanced mRNA stability, resulting in increased antigen production and elevated IgG levels. The ER localisation of TENT5A via FNDC3 proved essential for this process, highlighting a previously unrecognised spatial dimension of post-transcriptional regulation. To further explore translational outcomes, we applied the ribosome profiling (Ribo-seq) method to quantify ribosome occupancy and distribution across different mRNA constructs. Interestingly, our Ribo-seq data revealed distinct translational dynamics between two clinically approved SARS-CoV-2 vaccines, suggesting that subtle differences in mRNA design affect translational efficiency in vivo.

Eukaryotic cells adjust biochemical pathways in response to environmental stress, stalling energetically costly processes such as protein synthesis and ribosome biogenesis. Ribosome biogenesis occurs within the multiphase nucleolus, which contains distinct layers corresponding to rRNA transcription, processing, and assembly. While nucleolar perturbations induced by drugs and optogenetic gelation impair ribosome biogenesis, it remains unclear whether nucleolar properties are actively tuned in response to metabolic cues. Here, we show that cells adjust ribosome biogenesis to physiological needs by remodeling nucleolar material properties. Live-cell imaging reveals a tight relationship between the dynamics of the major nucleolar scaffolding protein NPM1 and nutrient availability. Under nutrient deprivation, NPM1 becomes enriched within the nucleolus, indicating enhanced phase separation and a more gel-like state. This change is accompanied by retention of ribosomal intermediates, reduced rRNA diffusivity, and contraction of nucleolar meshwork. Mechanistically, these changes are linked to cellular ATP levels, which broadly affect the energetically demanding process of ribosome biogenesis. Upon nutrient reintroduction, nucleolar composition recovers within minutes, indicating active regulation. Together, our findings reveal that nucleolar material properties are dynamically remodeled by cellular energy levels, linking phase separation to metabolic control of ribosome biogenesis.

Cellular translation undergoes dynamic changes in response to environmental conditions, leading to the production of diverse proteoforms, often synthesized by non-canonical translation. Identification of such rare translation events has recently become possible thanks to the availability of a growing number of ribosome profiling datasets. Using Ribocrypt.org we identified genes that, through translation initiation at non-AUG codons, generate N-terminally extended (NTE) protein variants containing mitochondrial targeting signal (MTS) in their NTE. This analysis identified components of the LSM complexes that normally localize to the nucleus (Lsm2-Lsm8) or cytoplasm (Lsm1-Lsm7). We confirmed that all Lsm proteins, except Lsm6, are indeed targeted to mitochondria. We focused on Lsm1 because its alternative NTE isoform contains a strong MTS, and several uORFs in the LSM1 5’UTR are translated, suggesting uORF-mediated translation regulation. In addition, deletion of LSM1 impairs yeast respiratory growth.

To assess the biological relevance of these observations, we generated a set of constructs expressing different LSM1 variants carrying specific changes in the 5’UTR, start codons and NTE-MTS. We checked mitochondrial localization of the resulting proteins and their impact on respiratory capacity and thermosensitivity. These tests confirmed that the Lsm1 NTE is responsible for its mitochondrial targeting, and showed that elements within the LSM1 5’UTR contribute to the LSM1-dependent thermotolerance. This demonstrates that translatome analyses enable systematic exploration of an alternative cellular proteome that shapes adaptation to changing environmental conditions.