Catalogues of mammalian long noncoding RNAs: modest conservation and incompleteness

Abstract

Background: Despite increasing interest in the noncoding fraction of transcriptomes, the number, species-conservation and functions, if any, of many non-protein-coding transcripts remain to be discovered. Two extensive long intergenic noncoding RNA (ncRNA) transcript catalogues are now available for mouse: over 3,000 macroRNAs identified by cDNA sequencing, and 1,600 long intergenic noncoding RNA (lincRNA) intervals that are predicted from chromatin-state maps. Previously we showed that macroRNAs tend to be more highly conserved than putatively neutral sequence, although only 5% of bases are predicted as constrained. By contrast, over a thousand lincRNAs were reported as being highly conserved. This apparent difference may account for the surprisingly small fraction (11%) of transcripts that are represented in both catalogues. Here we sought to resolve the reported discrepancy between the evolutionary rates for these two sets.

Results: Our analyses reveal lincRNA and macroRNA exon sequences to be subject to the same relatively low degree of sequence constraint. Nonetheless, our observations are consistent with the functionality of a fraction of ncRNA in these sets, with up to a quarter of ncRNA exons having evolved significantly slower than neighboring neutral sequence. The more tissue-specific macroRNAs are enriched in predicted RNA secondary structures and thus may often act in trans, whereas the more highly and broadly expressed lincRNAs appear more likely to act in the cis-regulation of adjacent transcription factor genes.

Conclusions: Taken together, our results indicate that each of the two ncRNA catalogues unevenly and lightly samples the true, much larger, ncRNA repertoire of the mouse.

Figure 1

G+C content of mouse ncRNA exons and ancestral repeats. The figure shows the cumulative distribution of G+C fraction as measured for macroRNA exons (red), lincRNA exons (black) and ancestral repeats (blue). LincRNAs tend to have higher G+C contents than macroRNAs. Ancestral repeats tend to possess a low G+C content.

Figure 2

Substitution rates of ncRNA and protein-coding genes. The cumulative distributions of substitution rate for (a) exons and (b) promoters as measured for macroRNAs (red), lincRNAs (black) and protein-coding genes (blue). MacroRNA and lincRNA exons exhibit similar degrees of constraint and appear to evolve faster than protein-coding exons. Protein-coding gene promoters evolve under stronger constraint than ncRNA exons. MacroRNA promoters have lower substitution rates than lincRNA promoters.

Figure 3

Distribution of highly conserved sequence across ncRNA exon sequences. Examples of phastCons elements (as in [38]) within (a) lincRNA (located on chromosome 10, 68730506-68731547) and (b) macroRNA (located on chromosome 1, 47378880-47380310) exons. Blue histograms represent the conservation in 17 vertebrates based on a phylogenetic hidden Markov model [13]. Green histograms represent pairwise conservation to other vertebrate species. Images have been taken from the UCSC genome browser.