LincRNA sequences are biased to counteract their translation
Long intergenic non-coding RNAs (lincRNAs) represent a large fraction of transcribed loci in eukaryotic genomes. Although classified as non-coding, most lincRNAs contain open reading frames (ORFs), and it remains unclear why cytoplasmic lincRNAs are not or very inefficiently translated. Here, we analysed signatures of repressed translation in lincRNA sequences from six eukaryotes. In species under stronger selection, i.e. fission yeast and worm, we detected significantly shorter ORFs than in intronic and non-transcribed control regions, a suboptimal sequence context around start codons for translation initiation, and trinucleotides ("codons") corresponding to less abundant tRNAs than codons in control regions, potentially impeding translation elongation. We verified that varying tRNA expression levels affect ribosome-binding to lincRNAs by analyzing data from five human cell lines. Notably, for three cell lines, codons in abundant cytoplasmic lincRNAs corresponded to lower expressed tRNAs than control codons, substantiating cell type-specific repression of lincRNA translation in human. Finally, comparing non-coding with peptide-encoding ORFs suggested that codon usage at the start of ORFs to be of particular relevance for ribosome-binding. The identified sequence signatures may assist distinguishing peptide- from real non-coding lincRNAs in a cell.