Fig. 4

Imbalances between translation initiation and reinitiation reinforce a pre-existing length bias in TE. A Arrow-based scatter plot show the transitions in TE per gene, between SHR.BN-(3S) and SHR.BN-(3L) rats. The length of the arrow is representative of the absolute change in TE between both congenic lines, with the position of the arrow tail reflecting the SHR.BN-(3L) TE and the position of the arrowhead indicating the TE in SHR.BN-(3S) rats. Blue arrows indicate a decrease in TE in SHR.BN-(3S) rats, whereas red arrows indicate an increase in TE in SHR.BN-(3S). Two zoomed-in regions show arrow behavior in the top and bottom of the graph, respectively highlighting genes with very long and short CDSs. B Schematic of how ribosome biogenesis defects can lead to a change in translation initiation and reinitiation rates, driving a global shift in TE that correlates with CDS length. C Scatter plots showing expression levels of all cardiac-expressed snoRNAs as measured by totRNA-seq and Ribo-seq data, with SNORA48 highlighted in pink. For both Ribo-seq datasets, p value volcano plots show the significance of the differential regulation of SNORA48 (highlighted in pink). D Representation of the genomic location of SNORA48. This snoRNA is contained within intron 4 of its host gene Eif4a1. Dot plots with expression levels as measured by totRNA-seq and Ribo-seq for SNORA48 and its host gene Eif4a1, in both the HXB/BXH RI panel and the congenic rat lines. Error bars indicate mean values with standard deviation (SD). See also Additional file 1: Figure S5. E Heatmap with CDS length versus fold change in TE (FC Mutant vs Wild Type) calculated from public Ribo-seq and RNA-seq data from various translational machinery mutants. Scaled fold changes are given. Within each group, genes are divided into 20 equally sized bins by increasing CDS length (left to right). Samples are sorted by Pearson’s correlation coefficient (r, top to bottom). Datasets are grouped as “negative correlation” or “positive correlation” depending on whether Pearson’s correlation coefficient value is lower than − 0.1 or higher than 0.1. The remaining datasets that showed no global shift in TE are grouped as “no correlation.” Scatter plots and Pearson’s square correlation coefficients (r²) between total CDS length and the FC in TE are displayed for three selected samples with one of the strongest negative Pearson’s r correlations (our model, chr3p teQTL), no correlation (Rps28b yeast knockout), and the strongest positive Pearson’s r correlation (Rpl26b yeast knockout)