A multi-level approach of gene expression data analysis to investigate translatome dynamics across multiple tissues, stages, and mouse models of SMA

Spinal Muscular Atrophy (SMA) is an autosomal recessive neurodegenerative disease, which, before the approval of therapies, was the leading genetic cause of infant mortality. The primary features of this pathology are progressive muscle weakness and atrophy, due to the degeneration of α-motor neurons in the anterior horn of the spinal cord. SMA is caused by deletions or mutations in the Survival Motor Neuron gene (Smn1), which induce reduced levels of the SMN protein. Since 1999, this disease has been primarily associated with splicing defects caused by loss of SMN protein due to its role in ribonucleoparticle biogenesis. However, further research revealed that this mechanism alone is not sufficient to explain the pathogenesis of the disease. More recent findings revealed that deficient SMN levels lead to defective translation in primary motor and cortical neurons, and in multiple tissues at the late stage of disease in the severe Taiwanese mouse model of SMA. Furthermore, SMN protein has been confirmed to be a ribosome-associated protein in vitro, in mouse cell lines and in vivo, and to physiologically regulate the translation of a particular subset of transcripts (defined as SMN-specific transcripts), which are characterized by specific sequence features. Upon SMN loss, the translation of this subset of transcripts is defective. SMN protein is ubiquitously expressed and its levels vary at different developmental stages and tissues in physiological conditions, leading to the hypothesis that translational defects may vary accordingly. However, the effect of SMN loss on translation across different tissue types, SMA mouse models, and disease stages is yet to be clarified. To investigate the link between SMN loss and translational defects in SMA, I took advantage of ribosome profiling to obtain the translatome from multiple tissues, stages and disease mouse models. Given that SMN is ubiquitously expressed, brain, spinal cord and liver were collected to investigate if common features underly translational defects upon its loss in these tissues. Since little is known about how translational impairments are modulated over time, tissues were collected from various developmental and disease stages, ranging from the embryo to the post-natal early-symptomatic stage of SMA. Furthermore, translation defects were studied in multiple models of SMA have ranging from severe to mild (i.e., Taiwanese, Delta7 and Smn2b/-), allowing for the exploration of the heterogeneity of the SMA clinical phenotype. Hence, the tissues were collected from three SMA mouse models (i.e., Taiwanese, Delta7, and Smn2b/-), allowing for the investigation of translational impairments in conditions that range from severe to mild SMA. A wide range of computational approaches was adopted to analyze ribosome profiling data from multiple perspectives, including Principal Component Analysis (PCA), pipelines for the analysis of RiboSeq positional information, differential and Gene Ontology enrichment analysis, and network methodologies. This set of tools applies to the study of ribosome profiling data and allows to investigate the translational mechanisms underlying SMA. This multilevel analysis holds difficulties in the representation and interpretation of the obtained results due to the number of variables (i.e., tissue, stage, model, and disease condition). I hence developed an R package to support the visualization of changes occurring in omics data from complex experimental designs. Next, I focused on the identification of translational defects in SMA through pairwise differential analyses performed on each set of experiments. This allowed me to identify significantly altered transcripts within each comparison. Despite poor overlaps between the sets of translationally dysregulated transcripts across the different stages, tissues, and models, commonly enriched biological processes were found. The analysis of sequence features on translationally dysregulated transcripts across all the stages, tissues, and models revealed the presence of features similar to those already found on the SMN-specific transcripts. In addition, based on network methodologies, I investigated the system-wide effects of SMN loss on connectivity patterns at the translational level, by taking advantage of network-based methodologies to integrate all sets of experiments and unravel any relationships between genes at the translatome level. Causal-inference networks, coupled with differential network analysis, complemented the standard differential analysis by modeling how the fluctuations in reciprocal transcript-specific ribosome occupancy might influence each other. This allowed to detect disrupted relationships in the disease condition across the multiple tissues, stages and models. In summary, this thesis provides, to my knowledge, the first multi-tissue, -stage, and -model translatome analysis to investigate the mechanisms underlying SMA. Furthermore, results provided within this work confirm that translation dysregulation is a common feature of SMA pathology across multiple tissues, stages, and SMA models. This highlights that the presence of specific sequence features of translationally dysregulated transcripts is a common link between defective translational regulation and SMN loss. Moreover, the detection of disrupted connectivity patterns at the translatome level underlies that a strong remodeling occurs upon SMN loss, and further emphasizes the pivotal role of this protein in translation. These outcomes highlight the importance of further investigating the mechanisms underlying defective translation in SMA from a system perspective to provide a comprehensive understanding of this pathology and promote the development of effective therapeutic strategies.