Box C/D snoRNP catalysed methylation is aided by additional pre‐rRNA base‐pairing

It has been proposed that snoRNA C′/D′ and C/D motifs are based on the same core consensus sequences, even though C′/D′ elements are generally less well conserved (Kiss‐Laszlo et al, 1998). Surprisingly, analysis of the primary sequence failed to clearly identify C′ and/or D′ boxes in many S. cerevisiae box C/D snoRNAs. We, therefore, compared the sequence of each S. cerevisiae box C/D snoRNA across multiple yeast species (Figure 1A; Supplementary Figure S1). This enabled the identification of C′/D′ motifs in all snoRNAs (Figure 1B; Supplementary Figure S1). There was a surprising amount of variation in C′/D′ sequences between individual snoRNAs; ranging from poorly (e.g., snR51) to highly conserved (e.g., snR53) motifs. Some snoRNAs harbour sequences that are highly conserved in yeast evolution but quite distinct from the accepted consensus, for both D′ (e.g., snR73, snR70, snR51 and snR87) and C′ (e.g., snR68 and U24). Surprisingly, nine C′ boxes in S. cerevisiae aligned better to the consensus with between one or two nucleotide insertions (Figure 1B; e.g., snR50 and snR69). In some cases, insertions were present in a subset of the orthologues of a single snoRNA (e.g., snR71) and in some snoRNAs there could be up to nine nucleotides inserted in the split C′ boxes (Supplementary Figure S1; snR190 and snR76). One or two nucleotide insertions were apparent in C′ boxes from vertebrate snoRNAs (e.g., rodent U15B, HBII‐234, HBII‐82 and mgh28S‐2411; data not shown), indicating that this is not specific to yeast. C′/D′ motifs were also highly conserved in snoRNAs that do not appear to use this motif to direct methylation (Figure 1B); the ‘guide’ region adjacent to the D′ motif in each of these RNAs was not conserved and no target has been identified (Supplementary Figure S1). Indeed, some of these ‘inactive’ motifs show better sequence conservation than active motifs, suggesting that the C′/D′ motif has a key role in the overall architecture of the snoRNP. The identification of many unusual C′/D′ motif sequences raised questions about the validity of the original consensus sequence. Re‐evaluation of the C′ and D′ sequences confirmed that, while the original consensus sequence was correctly identified, significant divergence is tolerated in this motif (Figure 1C). The alignments also identified highly conserved regions present in several snoRNAs (e.g., snR75 and snR70; Figure 1A; Supplementary Figure S1), which do not correspond to either the box or guide regions. We speculated that these could assist in snoRNA function through providing a protein‐binding site or additional base‐pairing potential (see below).

We were next interested in defining how the core box C/D proteins contact the box C/D snoRNAs. In particular, we were interested in how the proteins interacted with some of the more divergent C′/D′ motifs and with the additional conserved regions. We therefore performed CRAC, an RNA–protein crosslinking approach, followed by Illumina/Solexa sequencing, with Nop56, Nop58 and Nop1 to identify sites within the snoRNAs that crosslink to the core box C/D proteins (Granneman et al, 2009). Briefly, snoRNP proteins were C‐terminal tagged with a His6‐TEV‐Protein A (HTP) tag (Granneman et al, 2009) and the fusion proteins were expressed from the genome under control of the endogenous promoter. Cells were UV irradiated at 254 nm to induce RNA–protein crosslinks. RNAs associated with HTP‐tagged proteins were partially fragmented, purified and identified by linker ligation, cDNA synthesis and Illumina/Solexa sequencing. Analysis of the derived data revealed that 70–90% of the reads for each protein corresponded to box C/D snoRNA sequences. Reads were identified for every snoRNA, but U3 (snR17), U14 (snR128) and snR4 sequences were particularly enriched (Figure 2A). The different proteins showed significant variation in the number of reads recovered for each individual box C/D snoRNA (Figure 2A). Low‐level hits were also recovered for the H/ACA snoRNAs. The highest number of hits for this class of snoRNA were recovered for snR37, which was used as a baseline for background hits (Figure 2A). Note, however, that these reads may reflect genuine interactions taking place within pre‐ribosomes.

For each of the proteins, many reads were mapped to 3′ regions of the snoRNAs, near box D, presumably reflecting protein organization on the snoRNA (see Discussion). The average sequence length of the reads in the Solexa data was between 22 and 33 nts, resulting in significant overlap of individual reads. To better localize the binding sites, we generated heatmaps (Figure 2B) and graphs (Figure 2C) of reads smaller than 19 nucleotides. Comparison of the read distribution revealed differential protein localization. Nop58 primarily crosslinked to the 5′ and 3′ ends of the snoRNA, which include boxes C and D (Figure 2B). Nop56 preferentially crosslinked to the C′/D′ motif and the guide regions, in addition to binding near the 3′ end of the snoRNA. Nop1 frequently crosslinked to the guide adjacent to the D box and, occasionally, to the C′/D′ motif and associated guide (Figure 2B). The proteins, in particular Nop56, also make contacts outside the C/D and C′/D′ motifs. The data also indicate an asymmetric distribution of the three proteins on the snoRNAs.

The snoRNAs use either the guide adjacent to the D or D′ box to direct methylation, with some using both guide regions (Lowe and Eddy, 1999). However, no clear differences were observed between snoRNAs with one active guide region or two (Figure 2C; data not shown). The snoRNA sequence alignments revealed some unusual C′/D′ motifs. Several contained C′ and/or D′ boxes that, while evolutionarily highly conserved, differed substantially from the consensus. CRAC data confirmed that these unusual C′/D′ motifs are contacted by core box C/D snoRNP proteins (Figure 2C; e.g., snR70, snR79 and snR51). The same is true for the snoRNAs in which the C′ motif contains an insertion or have poorly conserved C′/D′ motifs (Figure 2C; e.g., snR69). The CRAC data further showed that several snoRNAs with conserved C′/D′ motifs but inactive adjacent guides (e.g., snR39, snR57, snR72 and snR79) are also bound by core proteins. Several snoRNAs contain conserved regions that do not correspond to the box elements or the methylation guides (Figure 1A; Supplementary Figure S1). As CRAC data resolution is roughly 15–20 nucleotides, we could not unambiguously determine whether core proteins bound these additional conserved regions in most snoRNAs. The conserved regions within snR70 and snR190 are, however, well over 20 nts away from either the guide regions or the box motifs. Analysis of the crosslinking data revealed that Nop1 crosslinked to these conserved regions (Figure 2C), suggesting that these conserved elements are recognized by at least Nop1.

Since many C′/D′ motifs diverge from the consensus, we next compared the activity of several such motifs in directing rRNA methylation. To perform this, we developed an expression system for an artificial snoRNA designed to target methylation of nucleotide S1316 in the 18S rRNA (Figure 3A) and expressed under the control of a GAL promoter. This site is not naturally methylated but its modification does not affect growth (Decatur and Fournier, personal communication; data not shown). We used the human U24 C′/D′ sequence as a standard in the artificial snoRNA as this motif matches the consensus sequence and was previously used to characterize C′/D′ sequence function in yeast (Kiss‐Laszlo et al, 1998; Qu et al, 2011). Several C′/D′ motifs were cloned into this snoRNA construct, including the divergent motifs from snR47, snR51 and snR70, the ‘split’ motif from snR78 and ‘inactive’ C′/D′ motifs from snR39, snR50, snR55 and snR57, that are not naturally adjacent to a guide sequence. The resulting plasmids were transformed into S. cerevisiae grown on galactose medium to induce snoRNA expression, and RNA was extracted. Methylation status was determined by primer extension (Figure 3B; Figure 5B (snR70)) and snoRNA levels were monitored by northern blotting (Supplementary Figure S2).

Expression of the artificial snoRNA containing the consensus hU24 C′/D′ motif resulted in methylation of S1316 (Figure 3B and C). This was not observed in the absence of the artificial snoRNA. The artificial snoRNAs, containing the C′/D′ motifs from snR51, snR78, snR50 and snR55, all methylated the rRNA to approximately the same level as seen with the hU24 C′/D′ motif. Slightly lower (50%) and higher (180%) methylation was seen, relative to the consensus motif, with the C′/D′ motifs of snR57 and snR47. This indicated that divergent C′/D′ motifs, including those with insertions in the C′ box (snR78), are roughly as active as the consensus C′/D′ motif (hU24). In contrast, a snoRNA carrying the C′/D′ motif of snR39 did not direct detectable methylation activity. The C′/D′ motif of snR39, although very conserved within the gene family, does not appear to naturally direct methylation and may be inactive.

The data on the ‘split’ C′ motifs predict that between 1 and 3 nt insertions in the C′ box will not block the function of the C′/D′ motif. Insertion of a U or G into the hU24 C′/D′ motif (Figure 3C) directed methylation at site S1316 to 85 and 140% of the levels seen with the wild‐type motif. Insertion of 2 nts reduced methylation activity of the snoRNA to 60%. Therefore, inserting 1 or 2 nts into the C′ motif did not significantly affect the activity of the snoRNA. Several C′ boxes contain only half of the C′ consensus sequence, suggesting that only one half of the sequence element may be sufficient for methylation, potentially explaining why insertions are tolerated. However, mutation of either the first or second GA dinucleotide in the C′ box (Figure 3C) reduced methylation at site S1316 by 5‐ and 10‐fold, respectively. Thus, both halves of box C′ used in the artificial snoRNA are essential for modification.

We have found that many snoRNAs contain conserved regions that do not correspond to C/D or C′/D′ box elements, or to methylation guides (Figure 1A; Supplementary Figure S1). In some cases, part or all of the conserved sequence appears to support secondary structures in the snoRNA (Supplementary Figure S1; e.g., snR66, snR70 and snR74). However, highly conserved regions identified in 13 yeast snoRNAs (snR13, snR39, snR47, snR48, snR61, snR64, snR70, snR73, snR75, snR76, snR87, snR190 and U18) are complementary to sequences immediately upstream or downstream of the rRNA methylation site (Figure 4A and B; Supplementary Figures S3 and S4). A further 11 yeast snoRNAs (snR38, snR40, snR54, snR55, snR56, snR60, snR62, snR68, snR69, snR71 and snR79) contained complementarity to regions flanking the rRNA methylation site, which was present in most but not all yeasts. Interestingly, these regions were mainly located either between the D′ and C′ boxes (e.g., snR70), or, where there is only one guide, in the second guide region (e.g., snR75; Figure 4B; Supplementary Figures S3 and S4). With one exception (snR69), however, the conserved regions were not found adjacent to a D or D′ box and would not be predicted to guide methylation. Occasionally, the regions were found in the spacer region between the guide and C or C′ box (e.g., snR47 and snR60). The potential base‐pairing interactions ranged from 4 to 11 base pairs (Figure 4A). While some of these interactions are relatively short, these regions are likely functioning as an extension of the normal guide region. For snR87, this potential interaction is highly conserved in evolution and the human orthologue, U16 could form 11 consecutive base pairs with the rRNA (Figure 4C). Further analysis revealed an additional 17 potential snoRNA–rRNA base‐pairing interactions for human box C/D snoRNAs that were evolutionarily conserved in higher eukaryotes (Figure 4C; Supplementary Figures S5–S7; U15, U21, U46, U49, U56, U103, snR39b, HBII‐180, HBII‐202, HBII‐429, HBII‐142, HBII‐210, HBII‐99, HBII‐296, HBII‐316, HBII‐82 and mgh28S‐2411). Interestingly, in both yeast and human snoRNAs, some of the extra potential base‐pairing interactions overlap with the canonical target‐snoRNA interaction by one or two nucleotides. In the case of HBII‐316, however, the overlap is five nucleotides (Supplementary Figure S6). The data, therefore, suggest that the additional base‐pairing might either enhance or regulate rRNA methylation.

The proximity of the base‐pairing to the methylation target site suggests a role in rRNA methylation, either stabilizing or facilitating formation of snoRNA–rRNA interactions. To test this possibility, we generated a yeast strain in which the snR72–snR78 cluster was deleted. This strain was then transformed with a plasmid expressing the wild‐type snoRNA cluster, or a cluster in which the additional base‐pairing region of either snR75 or snR76 was mutated (Figure 5A). RNA was extracted from each strain and the methylation levels were monitored by site‐specific RNase H cleavage (Yu et al, 1997) directed by chimeric 2′‐O‐methyl RNA/DNA oligonucleotides to sites L2288 (snR75) and L2197 (snR76). This approach was used as primer extension analysis of methylation at site L2197 proved unreliable. The snoRNA expression levels were analysed by northern blotting (Supplementary Figure S2). In the absence of snR75 and snR76, both oligonucleotides directed RNase H‐mediated cleavage of >90% of the 25S rRNA (Figure 5B). Each oligonucleotide also resulted in the non‐specific cleavage of 18S rRNA (Figure 5B; asterisk). Importantly, this was not affected by the presence or absence of the snoRNAs. Expression of wild‐type snR75 and snR76 rendered the 25S rRNA resistant to RNase H cleavage at the two methylation sites (<5% cleaved). In contrast, significant levels of cleavage were seen at sites L2288 and L2197 in 25S rRNA derived from cells expressing mutant snR75 (snR75mut) and snR76 (snR76mut), respectively. Mutation of the conserved extra base‐pairing region of snR76 resulted in a 2–3‐fold reduction in methylation (35–50% 25S rRNA uncleaved). Similarly, mutation of the conserved region of snR75 resulted in a 4–5‐fold reduction in methylation (20–25% 25S rRNA uncleaved). A similar reduction in rRNA methylation was observed for the snR75mut, relative to the wild‐type snoRNA, when the RNA was analysed by primer extension (Supplementary Figure S8). The data, therefore, indicate that the putative extra base‐pairing regions in both snR75 and snR76 are important for efficient methylation activity.

If the conserved regions function through base‐pairing, then they should only enhance methylation activity at a natural target site in the rRNA. Therefore, if the snoRNA is mutated to modify a new target site, the conserved region should only enhance methylation when it is mutated to be complementary to the region adjacent to the new target site. We, therefore, cloned the C′/D′ region of snR70, including the intervening stem structure and extra base‐pairing region, into the artificial snoRNA system so that it would direct methylation at site S1315 in the 18S rRNA (Figure 5C). A construct was generated in which the conserved loop region, containing the extra base‐pairing sequence, was mutated to be complementary to the sequence immediately upstream of the new target site. The snR70 construct directed methylation at site S1315 in the 18S rRNA (Figure 5D) at a level comparable to that seen with the hU24 consensus C′/D′ motif (at site S1316). Mutation of the loop sequence to generate a sequence complementary to the flanking region immediately upstream of the target site resulted in a reproducible five‐fold increase in methylation at site S1315. This, therefore, provides strong evidence that the conserved regions can function to enhance methylation through base‐pairing adjacent to the target site.

Genomic DNA sequences of Ascomycota (Saccharomycotina, Schizosaccharomycetes and Pezizomycotina) were searched iteratively for homologues using blastn with setting ‘expect 100’ (http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi?organism=fungi or http://www.yeastgenome.org/cgi-bin/blast-fungal.pl). Pezizomycotina snoRNAs were also identified by initiating blast searches with sequences of snoRNAs identified in a Neurospora Crassa cDNA library (Liu et al, 2009). Searching EMBL EST Fungi cDNA libraries (http://www.ebi.ac.uk/Tools/sss/wublast/nucleotide.html) yielded further snoRNAs from Basidiomycota, which were not found in the systematic blast searches. Sequences were aligned using Clustal W and annotated on the basis of visually identified conserved motifs and phylogenetically supported secondary structure. A similar approach, using just the basic nr/nt nucleotide collection database, was used to identify and align sequences for vertebrate U15 and U16 snoRNAs. All other vertebrate snoRNA alignment data were derived from the snoRNABase database (http://www-snorna.biotoul.fr/).

Nop1, Nop56 and Nop58 CRAC experiments were described elsewhere (Granneman et al, 2009). Reads were aligned against the S. cerevisiae genome using novoalign 2.05 (http://www.novocraft.com; settings –r Random, ‐s –h190 –a) and processed using in‐house python scripts (paper in preparation). Heatmaps of log‐transformed data shown in Figure 2A were generated using Java TreeView and Cluster3.0 with default settings. To calculate the densities of reads on the snoRNAs, we analysed reads between 15 and 19 nucleotides in length (after trimming linkers). This allows precise identification of protein‐binding sites, but is long enough to map reads uniquely to the yeast genome. Each snoRNA was divided into 10 regions: ‘before C box’, ‘C box’, ‘between C and guide’, ‘guide 1’, ‘D′ box’, ‘between D′ and C′’, ‘C′ box’, ‘between C′ and guide 2’, ‘guide 2’ and ‘D box’. For snoRNAs that lacked certain features, relevant regions were merged. The numbers of hits per million mapped reads were calculated separately for all nucleotide positions, and then averaged to yield read densities for each region. The densities were converted to a heatmap using Java TreeView, with ‘Contrast’ set to 5.0, 0.7 and 1.5 for Nop1, Nop56 and Nop58, respectively.

Methylation activity of various C′/D′ motifs was analysed in vivo in W303 (MATa/MATá; leu2‐3112 trp1‐1 can1‐100 ura3‐1 ade2‐1 his3‐11,15; [phi+]) using an artificial snoRNA construct inserted in the intron of the actin gene. The actin/snoRNA cassette (a PCR amplified 1‐kb Bam HI–Xba I fragment derived from pFL45/ACT/XK; Kiss‐Laszlo et al, 1996) was placed under the control of the GAL1 promoter (amplified as a 1‐kb Eco R1–Bam HI fragment from pBL143; Liu and Fournier, 2004) and cloned into pRS416 in which the Acc65 I and Xho I sites of the multiple cloning site had been deleted. Artificial snoRNAs were assembled from oligonucleotides and cloned in between the unique Xho I and Acc65 I sites present within the actin intron (Figure 3A; Supplementary Figure S10; Kiss‐Laszlo et al, 1996). C′/D′ regions, and the target site guide, were subsequently assembled from oligonucleotides and cloned into the Nhe I and Mlu I sites in the snoRNA‐coding sequence (Figure 3B; Supplementary Figure S10). The target site (S1316 or S1315) was chosen as a site detectable by reverse transcription that is not naturally modified. The expression levels of the artificial snoRNA were confirmed by northern blotting (Supplementary Figure S2).

Wild‐type and mutant snR75 and snR76 were expressed in their natural polycistronic context of the snR72–snR78 gene cluster. The cluster was PCR amplified using primers snR72r (5′‐AAAAGGTACCGTTATCCGTACACTTGACCTC‐3′) and snR78f (5′‐AAAACTCGAGAAGCATGAGGTATTATAGCGAC‐3′) and was cloned into the Acc65 I/Xho I sites of pRS416 and transformed into YPH499 (MATa, ura3‐52, lys2‐801, ade2‐101, leu2Δ1, his3‐Δ200, trp1‐Δ63) in which the non‐essential snoRNA gene cluster (Qu et al, 1999) was replaced by a natNT2 cassette (Janke et al, 2004). The snR75 and snR76 mutants were generated by site‐directed mutagenesis.

Methylation activity was determined by reverse transcription under limited nucleotide and enzyme concentrations (Maden, 2001). In all, 8 μg total RNA was annealed to ³²P‐, 5′‐end labelled primer and then incubated with M‐MLV reverse transcriptase (40 u, Promega), 2 μl 5 × RT buffer, 0.25 μl superasin and either 12.5 or 1.25 mmol dNTP's. The reactions were separated on either a 6 or 8% polyacrylamide/7 M urea gel and then visualized using a phosphorimager. Primers used for mapping were Map1316 (5′‐TAGTCCCTCTAAGAAGTGGATAACC‐3′) and Map75 (5′‐CTAGATAGTAGATAGGGACAGTGG‐3′). In addition, methylation was also monitored by site‐specific RNase H cleavage (Yu et al, 1997) directed by a chimeric 2′‐O‐methyl RNA/DNA oligonucleotides to sites L2197 (5′‐mAmCmUGGGCmAmGmAmAmAmUmCmAmCmAmUmU‐3′) and L2288 (5′‐mUmGmACGAGmGmCmAmUmUmUmGmGmCmUmAmC‐3′). Cleaved RNA was separated on a 1.2% agarose/glyoxal gel, stained with ethidium bromide and visualized using a transilluminator.

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).