Abstract

Background and rationale Translation is a fundamental biological process occurring in cells, carried out by ribosomes simultaneously bound to an mRNA molecule (polyribosomes). It has been exhaustively demonstrated that dysregulation of translation is implicated in a wide collection of pathologies including tumours and neurological disorders. Latest findings reveal the existence of translational regulatory mechanisms acting in cis or trans with respect to the mRNAs and governing the movement and the position of ribosomes along transcripts or directly impacting on the ribosome catalogue of its constituent proteins. For this reason, translational controls also account for widespread uncoupling between transcript and protein abundances in cells. To explain the poor correlation between transcripts and protein levels, many computational models of translation have been developed. Usually, these approaches aim at predicting protein abundances in cells starting from the mRNA abundance. Despite the efforts of these modelling studies, a consensus model remains elusive, drawing to contradictory conclusions concerning the role of mRNA regulatory elements such as the usage of codons (codon usage bias) and slowdown mechanism at the beginning of the coding sequence (ramp). More recently, following the rapid and widespread diffusion of ribosome footprinting assays (RiboSeq), which enables the dissection of translation at single nucleotide resolution, a number of computational pipelines dedicated to the analysis of RiboSeq data have been proposed. These tools are typically designed for extracting gene expression alterations at the translational level, while the positional information describing fluxes and positions of ribosomes along the transcript is still underutilized. Therefore, the polysome organization, in term of number and position of ribosomes along the transcript and the translational controls directed in shaping cellular phenotypes is still open to breakthrough discoveries. Broad objectives The aim of my thesis is the development of mathematical and computational tools integrated with experimental data for a comprehensive understanding of translation regulation and polysome organization rules governing the number of ribosomes per polysome and the ribosome position along transcripts. Project design and methods With this purpose, I developed riboWaves, an integrated bioinformatics suite divided in two branches. riboWaves includes in the first branch two modeling modules: riboAbacus, predicting the number of ribosomes per transcript, and riboSim, predicting ribosome localization along mRNAs. In the second branch, riboWaves provides two pipelines, riboWaltz and riboScan, for detailed analyses of ribosome profiling data aimed at providing meaningful and yet unexplored ribosome positional information. The models and the pipelines are implemented in C and R, respectively. riboAbacus and riboWaltz are available on GitHub. Results To predict the number of ribosomes per transcript and the position of ribosomes on mRNAs, I applied riboAbacus and riboSim, respectively, to transcriptomes of different organisms (yeast, mouse, human) for understanding the role of translational regulatory elements in tuning polysome in different organisms. First, I trained and validated performances of riboAbacus taking advantage of Atomic Force Microscopy images of polysomes, while performances of riboSim were assessed employing ribosome profiling data. Predictions provided by riboAbacus and riboSim were evaluated in parallel. I showed that the average number of ribosomes translating a molecule of mRNA can be well explained by the deterministic model, riboAbacus, that includes as features the mRNA levels, the mRNA sequences, the codon usage bias and a slowdown mechanism at the beginning of the CDS (ramp hypothesis). The predictions of ribosome localization by riboSim that used as features the mRNA sequence, the codon usage and the ramp, were run for yeast, mouse and human. I observed a good similarity between the predicted and experimental positions of ribosomes along transcripts in yeast, while poor similarity was obtained between predicted and experimental ribosome positions in the two mammals, suggesting the presence of more elaborate controls that tune ribosomes movement in higher eukaryotes than in simple species. After having developed two tools for the analyses of RiboSeq data and extraction of positional information on ribosome localization along transcripts, I applied both riboWaltz and riboScan in a case study. The aim was to dissect possible defects in ribosome localization in tissues of a mouse model of Spinal Muscular Atrophy (SMA). SMA is a neurodegenerative disorder caused by low levels of the Survival of Motor Neuron protein (SMN) in which translational impairments are recently emerging as possible cause of the disease. I analysed ribosome profiling data obtained from three different types of RiboSeq variants in healthy and SMA-affected mouse brains at the early-symptomatic stage of the disease. I observed i) a significant drop-off of translating ribosomes along the coding sequence in the SMA condition (using riboWaltz); ii) in SMA-affected mice, the possible accumulation of ribosomes along the 3' UTR in neuro-related mRNAs (using riboScan); iii) the involvement of SMN-specialized ribosomes in playing a very intimate role with the elongation stage of translation of the first codons of transcripts (riboWaltz), iv) the loss of ribosomes at the 3rd codon in SMA in transcripts bound by SMN-specialized ribosomes and v) a remarkable connection between SMN and the down-regulation of genes in SMA-affected mice. Overall, these findings confirmed previous observation about possible SMN-related dysregulations of local protein synthesis in neurons. More importantly, they unravel a completely new role of SMN in tuning translation at multiple levels (initiation, elongation and the recycling of terminating ribosomes), opening new hypotheses and scenarios for explaining the most devastating genetic disease, leading cause worldwide of infant mortality. Conclusions The present work provides a new comprehensive and integrated scenario for better understanding translation and demonstrates that this approach is a very powerful strategy to pave the way for new understanding of fine alteration in polysome organization and functional control in both physiological and pathological conditions.