Rapid turnover of long noncoding RNAs and the evolution of gene expression

Abstract

A large proportion of functional sequence within mammalian genomes falls outside protein-coding exons and can be transcribed into long RNAs. However, the roles in mammalian biology of long noncoding RNA (lncRNA) are not well understood. Few lncRNAs have experimentally determined roles, with some of these being lineage-specific. Determining the extent by which transcription of lncRNA loci is retained or lost across multiple evolutionary lineages is essential if we are to understand their contribution to mammalian biology and to lineage-specific traits. Here, we experimentally investigated the conservation of lncRNA expression among closely related rodent species, allowing the evolution of DNA sequence to be uncoupled from evolution of transcript expression. We generated total RNA (RNAseq) and H3K4me3-bound (ChIPseq) DNA data, and combined both to construct catalogues of transcripts expressed in the adult liver of Mus musculus domesticus (C57BL/6J), Mus musculus castaneus, and Rattus norvegicus. We estimated the rate of transcriptional turnover of lncRNAs and investigated the effects of their lineage-specific birth or death. LncRNA transcription showed considerably greater gain and loss during rodent evolution, compared with protein-coding genes. Nucleotide substitution rates were found to mirror the in vivo transcriptional conservation of intergenic lncRNAs between rodents: only the sequences of noncoding loci with conserved transcription were constrained. Finally, we found that lineage-specific intergenic lncRNAs appear to be associated with modestly elevated expression of genomically neighbouring protein-coding genes. Our findings show that nearly half of intergenic lncRNA loci have been gained or lost since the last common ancestor of mouse and rat, and they predict that such rapid transcriptional turnover contributes to the evolution of tissue- and lineage-specific gene expression.

Figure 1. Identification and characterization of ncRNAs in Mmus.

(A) Primary tissues were isolated and separate portions either flash frozen to permit RNA sequencing or treated with formaldehyde to crosslink protein-DNA contacts, which allows the chromatin immunoprecipitation reaction. Flow diagram illustrates assembly of liver expressed transcripts (RNAseq) marked by H3K4me3 at their transcriptional start sites (TSSs). Classification of long noncoding RNA (lncRNA, red) based on genomic location relative to annotated protein-coding genes (black) and directionality of transcription (arrows). Relative expression of 15 randomly selected intergenic lncRNA transcripts in (B) seven different adult Mmus tissues and (C) at five different developmental stages of Mmus liver was validated by RT-qPCR. Each heatmap row represents a single intergenic lncRNA. Areas are shaded according to the relative level of transcription in different tissues and developmental stages (in percent white: 0 to black: 100%).

Figure 2. Transcriptional turnover of liver expressed transcripts in rodents.

(A) Primary liver tissue was isolated from Mmus and two other rodents whose lineage split from Mmus one million years (Mcas) or 13 to 19 million years (Rnor). Examples of a Mmus-specific (locus7150, left), Mus genus-conserved (locus1400, middle), and rodent conserved (locus4179, right) lncRNA locus and their corresponding neighbouring protein-coding genes are illustrated. H3K4me3 enrichment is shown against a green background track and RNAseq signature against a yellow background track. The height (y-axis) of each track corresponds to the read depth. Beneath the enrichment tracks is the Refseq genome annotation for this region (UCSC genome browser). The mammalian conservation track (UCSC genome browser) shows degree of placental mammal base pair conservation (20 species) and sequence conservation. The syntenic positions of the predicted TSS of ncRNAs and neighbouring protein-coding loci are traced between species with dashed red lines. (B) H3K4me3 enriched regions (black: H3K4me3 bound DNA reads and white: no ChIPseq reads) within 5 kb of the peak summit for all identified intergenic lncRNA (one per line, categories I to V) and 136 randomly sampled protein-coding loci (category VI). Categories represent intergenic lncRNA loci that are transcribed and marked by H3K4me3 in all three rodents (I), in Mmus and Mcas but not in Rnor (II), in Mmus only (III), in Mcas only (IV) and Rnor only (V). Peaks were sorted according to their width. (C) Heatmap similar to (D) representing intergenic lncRNA transcripts anchored on predicted TSS (black: more than one RNAseq reads and white: less than one RNAseq reads). (D) Transcriptional turnover of liver-expressed protein coding (black) or intergenic lncRNA loci (red) in rodents.

Figure 3. Rodent conserved intergenic lncRNA loci and promoter sequences exhibit constraint.

Normalised nucleotide substitution rates for (A) 160 intergenic lncRNA loci conserved in rodents (expressed in Mmus, Mcas and Rnor) and 108 Mus genus specific intergenic lncRNA loci; and (B) 159 putative intergenic lncRNA promoters conserved in rodents and 104 Mus genus specific intergenic lncRNA putative promoters. Putative proximal intergenic lncRNA promoters were defined as the 400 bp upstream region of the predicted TSS. Yellow dashed line represents the expected neutral substitution rate. Compared to neutral sequence (ancestral repeats, AR) in the vicinity, nucleotide substitution rates differ significantly for loci and promoter of intergenic lncRNA transcripts conserved in rodents (as indicated by asterisks ***, p<0.001) and the promoters of Mus genus specific intergenic lncRNAs (***, p<0.001).

Figure 4. Lineage-specific intergenic lncRNAs are associated with increased expression of genomically adjacent protein-coding genes.

(A) Effect of intergenic lncRNA (red) transcription on their closest protein-coding genes A (black) and their respective closest protein-coding genes B (grey) was determined. (B) Fold-difference in expression for one-to-one orthologous protein-coding gene pairs (A and B) where gene A is adjacent to lineage-specific (Mus genus- or Rnor-specific) intergenic lncRNA loci. The fold-difference in expression between Mmus and Rnor for protein-coding gene A is significantly (represented by asterisks [**], p<0.005) higher than the expected variation in expression based on 230 housekeeping genes (white). Yellow dashed line represents median fold-difference in expression between Mmus and Rnor housekeeping genes. Lineage-specific intergenic lncRNA transcription has no significant effect on the expression levels of protein-coding genes B. (C) Fold-difference in expression for one-to-one orthologous protein-coding gene pairs (A and B) where genes A are adjacent to rodent conserved (conserved in Mmus, Mcas and Rnor) intergenic lncRNAs loci. Rodent conserved intergenic lncRNA gene expression has no significant effect on the transcription of neighbouring protein-coding gene A or B between mouse and rat. In parentheses are the numbers of protein-coding genes studied.