Small RNAs derived from snoRNAs

Abstract

Small nucleolar RNAs (snoRNAs) guide RNA modification and are localized in nucleoli and Cajal bodies in eukaryotic cells. Components of the RNA silencing pathway associate with these structures, and two recent reports have revealed that a human and a protozoan snoRNA can be processed into miRNA-like RNAs. Here we show that small RNAs with evolutionary conservation of size and position are derived from the vast majority of snoRNA loci in animals (human, mouse, chicken, fruit fly), Arabidopsis, and fission yeast. In animals, sno-derived RNAs (sdRNAs) from H/ACA snoRNAs are predominantly 20-24 nucleotides (nt) in length and originate from the 3' end. Those derived from C/D snoRNAs show a bimodal size distribution at approximately 17-19 nt and >27 nt and predominantly originate from the 5' end. SdRNAs are associated with AGO7 in Arabidopsis and Ago1 in fission yeast with characteristic 5' nucleotide biases and show altered expression patterns in fly loquacious and Dicer-2 and mouse Dicer1 and Dgcr8 mutants. These findings indicate that there is interplay between the RNA silencing and snoRNA-mediated RNA processing systems, and that sdRNAs comprise a novel and ancient class of small RNAs in eukaryotes.

FIGURE 1.

sdRNAs size distributions in metazoa. (Black) C/D and (yellow) H/ACA sdRNA sizes are plotted for (A) human, (B) mouse, (C) chicken, and (D) Drosophila. In all species, sdRNAs segregate by size dependent on the snoRNA type they originate from. In animals, C/D sdRNAs are more highly expressed and smaller than H/ACA sdRNAs. Vertical axes are in units of counts per million (cpm)—a ratio of the abundance of sdRNAs per million mapped tags in the total library. In A–C, C/D sdRNA abundance is denoted on the left of each panel, H/ACA on the right.

FIGURE 2.

sdRNA position-of-origin within parent snoRNAs. The position-of-origin of (black) C/D and (yellow) H/ACA is shown for (A) human, (B) mouse, (C) chicken, and (D) Drosophila. To normalize for varying host snoRNA lengths, we divided snoRNAs into 10% blocks (deciles) from the 5′ to the 3′ end, and plotted the positions of the 5′ end of each sdRNA. C/D sdRNAs are primarily derived from the 5′ end of snoRNAs. H/ACA sdRNAs are predominantly derived from the 3′ end, a position that correlates to one of the two H/ACA snoRNA hairpins. C/D sdRNA abundance is denoted on the left of each panel, H/ACA on the right. (E) The structure of mouse ACA58b. SdRNAs are predominantly localized to the 3′ end of the RNA secondary structure. Sequences shaded in red and blue indicate miRNA-like mature and star sequences, respectively.

FIGURE 3.

Mouse sdRNAs are affected by the loss of Dicer1 and Dgcr8. Wild-type and mutant ES cell libraries were normalized to t/sn/srpRNA abundance (Supplemental Table 4; see the text). (A, black) C/D and (yellow) H/ACA sdRNA sizes in (solid lines) wild-type (WT), (large dashed lines) dicer1Δ/Δ, and (small dashed lines) dgcr8Δ/Δ embryonic stem cells. H/ACA sdRNAs also exhibit an increase of 21–24-nt species in dgcr8Δ/Δ ES cells. (B) The relative enrichment of sdRNA abundance compared to WT of 39 H/ACA snoRNAs in (blue) dicer1Δ/Δ and (black) dgcr8Δ/Δ embryonic stem cells. SnoRNAs were only included in the analysis if their sdRNAs had an abundance greater than 5 in each library and greater than 10 in at least one library.

FIGURE 4.

sdRNAs are associated with Arabidopsis AGO7. (A) The relative enrichment of (black) C/D and (yellow) H/ACA sdRNA abundance from individual snoRNAs in Arabidopsis AGO7 immunoprecipitations. Enrichments are calculated in comparison to IP input deep sequencing libraries. SnoRNAs were only included in the analysis if their sdRNAs had an abundance greater than 5 in each library and greater than 10 in at least one library. H/ACA and C/D sdRNAs are strongly enriched in the AGO7 library. (B,C) The proportions of C/D and H/ACA sdRNAs 5′ end nucleotides associated with Arabidopsis AGO7. C/D sdRNAs are 5′U biased and H/ACA sdRNAs are 5′A biased.