CapSeq and CIP-TAP identify Pol II start sites and reveal capped small RNAs as C. elegans piRNA precursors

Abstract

Piwi-interacting (pi) RNAs are germline-expressed small RNAs linked to epigenetic programming. C. elegans piRNAs are thought to be transcribed as independent gene-like loci. To test this idea and to identify potential transcription start (TS) sites for piRNA precursors, we developed CapSeq, an efficient enzymatic method for 5' anchored RNA profiling. Using CapSeq, we identify candidate TS sites, defined by 70-90 nt sequence tags, for >50% of annotated Pol II loci. Surprisingly, however, these CapSeq tags failed to identify the overwhelming majority of piRNA loci. Instead, we show that the likely piRNA precursors are ∼26 nt capped small (cs) RNAs that initiate precisely 2 nt upstream of mature piRNAs and that piRNA processing or stability requires a U at the csRNA +3 position. Finally, we identify a heretofore unrecognized class of piRNAs processed from csRNAs that are expressed at promoters genome wide, nearly doubling the number of piRNAs available for genome surveillance.

Figure 1

Flowcharts illustrating the CapSeq (left) and CIP-TAP (right) protocols.

Figure 2. Comparative analysis of RNA-seq protocols that enrich long- and short-capped RNAs (CapSeq and CIP-TAP respectively) and a protocol (CIP-PNK) that clones uncapped short RNAs such as siRNA, piRNA and miRNA species

(A) Histograms representing the 5′ ends of mapped reads, as indicated, at a typical protein-coding locus, rps-4. The height of each histogram bar is proportional to the number of reads sharing the same 5′ nt and the scale (log2) is shown. Candidate pre-mRNA 5′ ends and csRNAs are indicated. Trans-splicing at some genes including rps-4 results in removal of the 5′ UTR of the pre-mRNA, called an “outron”, and the addition of a “Spliced leader”. The major trans-splice site for rps-4 is off the scale as indicated by a break in the bar, and the total number of SL-containing reads is indicated. Two minor trans-splice sites flank the major trans-splice site as indicated by triangles below the CapSeq reads. The outron is indicated by a line below the CapSeq reads; dashes indicate the variable 5′ end of the outron. The blue bars beneath the rps-4 coding sequences in the CIP-TAP and CIP-PNK samples correspond to antisense 22G-RNAs. (B) Schematic representation of the nucleotide composition around candidate TS sites (the +1 position) identified by CapSeq and CIP-TAP reads (here only sense csRNAs). The nucleotide height (in bits) represents the log2 ratio of the frequency observed relative to the expected frequency based on genomic nt composition. The enriched YR motif is indicated. (C) Pie charts indicating the relative composition of small RNAs recovered in the CIP-TAP and CIP-PNK samples. (D) Histograms representing the 5′ ends of mapped reads at the mir-54 –56 miRNA cluster. Candidate pri-miRNA 5′ ends and csRNAs are indicated. The asterisks indicate reads corresponding to miRNA star-strands. See also Figure S1 and Table S1.

Figure 3. CapSeq analysis of mouse testes RNA

(A and B) Browser views of representative protein-coding and piRNA cluster regions are shown. The histograms (log2 scale) represent the frequency of reads sharing the same 5′ ends from CapSeq or MILI IP (Robine et al., 2009) as indicated. Bidirectional reads were observed around the TS sites of Gpatch4. (C) Schematic representation of the nucleotide composition around candidate TS sites (YR). The nucleotide height (in bits) represents the log2 ratio of the frequency observed relative to the expected frequency based on genomic nt composition. See also Table S2.

Figure 4. Analysis of annotated 21U-RNA loci

(A) Cumulative analysis of the 5′ ends of unique CIP-TAP (orange) and CIP-PNK (green) sequences with respect to the YRNU motif of a consensus 21U-RNA locus. The scale (linear) is shown. The red segment preceded by U indicates the mature 21U-RNA. (B) Graph showing the length distribution of CIP-TAP/csRNA reads (orange) and CIP-PNK/21U-RNA reads (green) mapping to 21U-RNA loci. (C) Graph of csRNA levels plotted against corresponding 21U-RNA levels for annotated 21U locus. The points in red indicate previously annotated 21A/G/C piRNAs. Points near the X-axis (under the bracket) include 22G-RNAs previously mis-annotated as piRNAs. See also Figure S2 and Table S3.

Figure 5. Analysis of 21U-like loci

(A) An example of a 21U-like locus with a piRNA cluster on chromosome IV. Both 21ur-747 (blue triangle at left) and a nearby 21U-like or csRNA locus (open triangle) are enlarged to single nucleotide resolution. The bars indicate the number of reads (linear scale provided) sharing the corresponding 5′ nt from CIP-TAP (orange) and CIP-PNK (green), relative to the YRNU motif (yellow) of 21ur-747 and YRNA motif (gray) of the 21U-like locus (open triangle). (B) Schematic representation of the nucleotide composition at canonical 21U-RNA loci (top) and 21U-like loci (bottom). The nucleotide height (in bits) represents the log2 ratio of the frequency observed relative to the expected frequency based on genomic nt composition. The upstream and TS-site (YR) motifs are indicated. The observed 5′ end of mature 21nt-RNAs is indicated by the arrow at the +3 position. See also Figure S2 and Table S3.

Figure 6. Analysis of Type-2 21U-RNA loci

(A and B) Comparative analysis of reads from RNA-seq protocols (as indicated). The histograms represent the frequency of reads sharing the same 5′ end at the rps-4 locus (A), and at a non-annotated locus on chromosome X (B). The genome coordinates are indicated. A log2 scale is provided for each set of histograms. Mature 21U-RNA reads are enriched in the “oxidized” vs “control” (A) and “PRG-1 IP” vs “Input” (A and B) samples as indicated. Closed triangles indicate the positions of mature 21U-RNAs enriched in the PRG-1 IP. In (B), the red bars correspond to WAGO-dependent 22G-RNAs likely targeting non-annotated transcripts (dashed arrows) of unknown length. (C) Enlarged regions indicating the precise positions of corresponding CIP-TAP (orange) and PRG-1 IP (green) reads relative to the YRNU motif indicated below the sequence. Note that one of the TS sites for rps-4 is “RR” rather than “YR”. The open triangles above the CIP-TAP bars point to the likely precursor of the mature 21U-RNAs, which are indicated by the closed triangles below the PRG-1 IP bars. See also Figure S3 and Table S3.

Figure 7. C. elegans piRNAs are processed from capped-small RNAs

(A and B) Correlation analyses between 21U-RNAs and csRNAs or long-capped RNAs (as indicated) at (A) the 44 Type-1 loci where long-capped RNA reads were obtained by CapSeq, and at (B) 982 Type-2 loci where csRNA and CapSeq reads were both found at +1, relative to the downstream (+3) U of a 21U-RNA. For a perfect correlation, the Spearman rank correlation coefficient (r) = 1 or −1, and for no correlation, r = 0. P-values (p) were calculated using non-parametric correlation model. (C) Model for the biogenesis of 21U-RNA. Arrows indicate TS sites of Type-1 and Type-2 piRNA loci.