A computational screen for mammalian pseudouridylation guide H/ACA RNAs

Abstract

The box H/ACA RNA gene family is one of the largest non-protein-coding gene families in eukaryotes and archaea. Recently, we developed snoGPS, a computational screening program for H/ACA snoRNAs, and applied it to Saccharomyces cerevisiae. We report here results of extending our method to screen for H/ACA RNAs in multiple large genomes of related species, and apply it to the human, mouse, and rat genomes. Because of the 250-fold larger search space compared to S. cerevisiae, significant enhancements to our algorithms were required. Complementing extensive cloning experiments performed by others, our findings include the detection and experimental verification of seven new mammalian H/ACA RNAs and the prediction of 23 new H/ACA RNA pseudouridine guide assignments. These assignments include four for H/ACA RNAs previously classified as orphan H/ACA RNAs with no known targets. We also determined systematic syntenic conservation among human and mouse H/ACA RNAs. With this work, 82 of 97 ribosomal RNA pseudouridines and 18 of 32 spliceosomal RNA pseudouridines in mammals have been linked to H/ACA guide RNAs.

FIGURE 1.

Flowchart of the snoGPS-C algorithm. The algorithm is divided into three main stages. The initial stage is the generation of the most highly conserved non-protein-coding human and mouse sequences. The second stage involves running snoGPS on one of these two data sets and extracting the top 200 candidate sequences for each Ψ target. The final stage consists of extracting multiZ homologs of each of these sequences, testing each of these sequences with snoGPS, and averaging the human, mouse, and rat snoGPS scores.

FIGURE 2.

Fraction of previously known H/ACA RNAs detected by snoGPS and snoGPS-C. An H/ACA RNA was considered “detected” if it was among the five highest scoring hmr20 sequences for at least one rRNA or snRNA Ψ. The “No target” sequences included H/ACA RNAs with no predicted target as well as those whose associated Ψ was only experimentally located subsequent to these studies. The “#1–5 snoGPS” sequences include those that ranked among the five snoGPS-highest-scoring human sequences for at least one target, but that were not detected using snoGPS-C composite scores. “Missed” sequences did not rank among the top five for any target either for snoGPS or snoGPS-C.

FIGURE 3.

Experimental verification of predicted guide RNAs by primer extension. (Lanes 1–7) ACA63, ACA65, ACA67, ACA64, ACA68, ACA66, and ACA62, respectively. Experimentally determined 5′-ends were always within 7 nt of predicted 5′-ends.

FIGURE 4.

Putative base-pairings between Ψ target regions and H/ACA RNAs. Pairings are shown for (A) the newly identified Ψ guide H/ACA RNAs; (B) new assignments of previously known H/ACA RNAs.

FIGURE 5.

Annotated sequence alignments. Dots (…) in the alignments indicate bases that are identical to those in the reference sequence. The letters in the annotation string indicate H/ACA features identified in the sequence: (H and A) the H and ACA motifs, respectively. (L and R) The two components of the rRNA guide sequence. Bases annotated with X and I refer to the lower (or “external”) and upper (or “internal”) stems of either the 5′- or 3′-hairpins. (A) Annotated alignment of an unverified candidate sequence (S35) showing the 3′-hairpin of the human, mouse, rat, and chicken homologs. The alignment shows four substitutions (in green) that conserve the proposed secondary structure, seven others do not affect any of the proposed H/ACA motifs (in blue), and a single substitution (in red) that conflicts with the proposed H/ACA structure. (B) Alignment of ACA50 showing substitutions in the multiZ-identified rat homolog that violate the snoGPS-predicted RNA structure. Note the single nucleotide substitution in the ACA motif and the deletion in the 3′-hairpin in the rat sequence.

FIGURE 6.

ACA66 in human and mouse have nonhomologous host genes. Screen shot of the UCSC Genome Browser showing ACA66 with its host gene USP32 on Homo sapiens Chromosome 17 (NCBI Build hg17, May 2004). The yellow color along most of the mouse net alignment track indicates that the homologous region of USP32 in mouse is on Chromosome 11. The red color at the location of ACA66 in the mouse net alignment track indicates that the multiZ homolog of ACA66 itself is on Chromosome 5 (where it is located in an intron of Wbscr22).

FIGURE 7.

Positions of H/ACA RNAs within host introns and host genes. (A) Histogram of distances of H/ACA RNAs to their upstream and downstream host-intron splice sites. (B) Histogram of relative positions of introns containing H/ACA RNAs within their host genes. Relative positions near zero (one) indicate introns near the 5′ (3′) ends of the host gene, respectively.