Long noncoding RNA repertoire and targeting by nuclear exosome, cytoplasmic exonuclease, and RNAi in fission yeast

Abstract

Long noncoding RNAs (lncRNAs), which are longer than 200 nucleotides but often unstable, contribute a substantial and diverse portion to pervasive noncoding transcriptomes. Most lncRNAs are poorly annotated and understood, although several play important roles in gene regulation and diseases. Here we systematically uncover and analyze lncRNAs in Schizosaccharomyces pombe. Based on RNA-seq data from twelve RNA-processing mutants and nine physiological conditions, we identify 5775 novel lncRNAs, nearly 4× the previously annotated lncRNAs. The expression of most lncRNAs becomes strongly induced under the genetic and physiological perturbations, most notably during late meiosis. Most lncRNAs are cryptic and suppressed by three RNA-processing pathways: the nuclear exosome, cytoplasmic exonuclease, and RNAi. Double-mutant analyses reveal substantial coordination and redundancy among these pathways. We classify lncRNAs by their dominant pathway into cryptic unstable transcripts (CUTs), Xrn1-sensitive unstable transcripts (XUTs), and Dicer-sensitive unstable transcripts (DUTs). XUTs and DUTs are enriched for antisense lncRNAs, while CUTs are often bidirectional and actively translated. The cytoplasmic exonuclease, along with RNAi, dampens the expression of thousands of lncRNAs and mRNAs that become induced during meiosis. Antisense lncRNA expression mostly negatively correlates with sense mRNA expression in the physiological, but not the genetic conditions. Intergenic and bidirectional lncRNAs emerge from nucleosome-depleted regions, upstream of positioned nucleosomes. Our results highlight both similarities and differences to lncRNA regulation in budding yeast. This broad survey of the lncRNA repertoire and characteristics in S. pombe, and the interwoven regulatory pathways that target lncRNAs, provides a rich framework for their further functional analyses.

Keywords: NMD pathway; RNA degradation; Schizosaccharomyces pombe; antisense RNA; cytoplasmic exonuclease; pervasive transcription.

FIGURE 1.

Hierarchical clustering of gene expression in different RNA-processing mutants and physiological conditions. Expression profiles are shown for (A) all 5177 mRNAs, (B) 1573 annotated lncRNAs, and (C) 5775 novel, unannotated lncRNAs. Changes in RNA levels in response to the different genetic and physiological conditions (indicated at bottom) relative to control cells grown in minimal medium are color-coded as shown in the color legend at bottom right (log₂ expression ratios). The rrp6 and exo2 samples indicated by asterisks have been depleted for rRNA instead of poly(A) purification used for all other samples. Data obtained from these two types of RNA-seq libraries are compared in Supplemental Figure S2b. Details on strains and conditions are provided in Supplemental Tables S1 and S2, and all expression data are provided in Supplemental Table S3.

FIGURE 2.

Histograms showing numbers and proportions of induced (orange), repressed (blue), and all other (gray) transcripts for the different RNA classes as indicated. Differentially expressed genes were defined as those being ≥2-fold induced (mean of two biological repeats) or repressed and showing significant changes (P < 0.05) compared to reference as determined by DESeq2.

FIGURE 3.

Gene expression in major groups of genetic and environmental conditions. (A) (Left graph) Box plot of expression ratios (condition relative to control) of all mRNAs, annotated and novel lncRNAs in nuclear exosome (rrp6Δ, rrp6-ts), RNAi (ago1Δ, dcr1Δ, rdp1Δ), and cytoplasmic exonuclease (exo2Δ) mutants. (Right graph) As left but for quiescence, stationary phase, and meiosis conditions. The horizontal dashed lines indicate twofold induction and repression. (B) As in panel A, but for expression levels (RPKM scores). All expression data are provided in Supplemental Table S3.

FIGURE 4.

Classification of lncRNAs into CUTs, DUTs, and XUTs. (A) The lncRNAs significantly induced (DESeq2; adjusted P ≤ 0.05) in rrp6Δ, dcr1Δ, or exo2Δ mutants were clustered into CUTs, DUTs, and XUTs, respectively, using the Mfuzz R package (default parameters, three clusters specified). The clustering shows the 5586 uniquely classified lncRNAs after filtering those with a membership score <0.7. The red/blue colors indicate the mean RPKM values in the three mutants as indicated, scaled by subtracting the mean of the row and division by the standard deviation of the row (z-score). The assigned clusters are indicated at left. (B) Box plots of RPKM values (log₂) of all CUTs (left), DUTs (middle), and XUTs (right) in control (MM) and mutant cells as indicated. The data for the biological repeats 1 and 2 are plotted separately. (C) Hierarchical clustering of genetic conditions for all CUTs, DUTs, and XUTs as indicated. Changes in RNA levels in response to the different genetic conditions (indicated at bottom) relative to control cells grown in minimal medium are color-coded as shown in the legend at bottom right (log₂ expression ratios). All expression and classification data are provided in Supplemental Table S3.

FIGURE 5.

Analyses of lncRNAs by positions relative to mRNAs. (A) We grouped the annotated and novel lncRNAs into three main positional types as represented schematically: 1577 bidirectional, 4474 antisense, and 1189 intergenic RNAs, leaving only 108 lncRNAs (105 annotated, three novel) that overlapped mRNAs in sense direction (Materials and Methods). Pie charts of the corresponding proportions of CUTs, DUTs, XUTs, and other lncRNAs are provided beneath each positional type, and also for all annotated and known lncRNAs. Significantly enriched slices are indicated with asterisks (R prop.test function, P < 10⁻⁶). (B) Pie charts of the proportions of bidirectional, antisense, intergenic, and sense lncRNAs for all (annotated and known) lncRNAs, and among the CUTs, DUTs, and XUTs. Significantly enriched slices are indicated with asterisks (R prop.test function, P < 10⁻⁴). (C) Pearson correlation coefficients for RPKM expression data of each bidirectional lncRNA-mRNA pair (left) and each antisense lncRNA-mRNA pair (right). The correlation data are shown separately for all genetic and physiological conditions as indicated on x-axes. For antisense lncRNAs, the difference between the distributions in genetic versus physiological conditions is highly significant (P_Wilcoxon = 4.6 × 10⁻¹⁷⁰). All classification data are provided in Supplemental Table S5.

FIGURE 6.

Expression patterns of bidirectional, antisense, and intergenic lncRNAs. (A) Histograms showing the numbers and proportions of the induced (orange), repressed (blue), and all other (gray) transcripts for the different lncRNA classes as indicated. Differentially expressed genes were defined as those being >2-fold induced (average of two biological repeats) or repressed and showing significant changes (P < 0.05) compared to reference as determined by DESeq2. (B) Hierarchical clustering of physiological conditions for different lncRNA classes as indicated. Changes in RNA levels in response to the different physiological conditions (indicated at bottom) relative to control cells grown in minimal medium are color-coded as shown in the legend at bottom (log₂ expression ratios). Hierarchical clustering of physiological conditions for CUTs, DUTs, and XUTs is shown in Supplemental Figure S4. All expression and classification data are provided in Supplemental Tables S3 and S5.

FIGURE 7.

Nucleosome patterns for coding and noncoding transcribed regions. Nucleosome profiling data for all mRNA loci (left), 509 bidirectional lncRNA loci, and 1119 intergenic lncRNA loci (right) in proliferating wild-type cells. Bidirectional lncRNA loci that overlap mRNAs in antisense direction were not included. We omitted 70 intergenic lncRNA loci that showed unusually high histone occupancies (Supplemental Table S8); these loci are mostly located in centromeric or subtelomeric regions. The top graphs show average nucleosome profiles for the different types of transcribed regions, aligned to the transcription start sites (TSS). The lower graphs show heatmaps for the first two kilobases of all transcribed regions analyzed, ordered by transcript length from top (longest RNAs) to bottom (shortest RNAs). Sequencing scores are color-coded as shown in legend at bottom right.