During site‐specific pseudouridylation of eukaryotic rRNAs, selection of correct substrate uridines for isomerization into pseudouridine is directed by small nucleolar RNAs (snoRNAs). The pseudouridylation guide snoRNAs share a common ‘hairpin–hinge– hairpin–tail’ secondary structure and two conserved sequence motifs, the H and ACA boxes, located in the single‐stranded hinge and tail regions, respectively. In the 5′‐ and/or 3′‐terminal hairpin, an internal loop structure, the pseudouridylation pocket, selects the target uridine through formation of base‐pairing interactions with rRNAs. Here, essential elements for accumulation and function of rRNA pseudouridylation guide snoRNAs have been analysed by expressing various mutant yeast snR5, snR36 and human U65 snoRNAs in yeast cells. We demonstrate that the H and ACA boxes that are required for formation of the correct 5′ and 3′ ends of the snoRNA, respectively, are also essential for the pseudouridylation reaction directed by both the 5′‐ and 3′‐terminal pseudouridylation pockets. Similarly, RNA helices flanking the two pseudouridylation pockets are equally essential for pseudouridylation reactions mediated by either the 5′ or 3′ hairpin structure, indicating that the two hairpin domains function in a highly co‐operative manner. Finally, we demonstrate that by manipulating the rRNA recognition motifs of pseudouridylation guide snoRNAs, novel pseudouridylation sites can be generated in yeast rRNAs.
The nucleolar biosynthesis of eukaryotic rRNAs consists of three major steps. First, RNA polymerase I synthesizes a large precursor RNA (pre‐rRNA), in which the 18S, 5.8S and 25/28S rRNAs are flanked and separated by external and internal spacer sequences (Hadjiolov, 1985). After transcription, the rRNA regions of the pre‐rRNA undergo extensive covalent modifications. Many precisely selected nucleotides are methylated at the 2′‐O‐hydroxyl position, and several uridine residues are converted into pseudouridine (Maden, 1990; Eichler and Craig, 1994). Finally, the mature‐sized rRNAs are nucleolytically processed from the modified pre‐rRNA (Eichler and Craig, 1994; Venema and Tollervey, 1995; Sollner‐Webb et al., 1996).
In vertebrates, the mature 18S, 5.8S and 28S rRNAs contain >100 2′‐O‐methyl groups and ∼95 pseudouridines (Maden, 1990; Ofengand et al., 1995; Ofengand and Bakin, 1997). Although substantial evidence points to a catalytic role for rRNAs in protein synthesis (Green and Noller, 1997; Nitta et al., 1998; Schimmel and Alexander, 1998), the function of modified nucleotides remains entirely speculative. Since pseudouridylation and ribose methylation sites cluster on the universally conserved functional centres of rRNAs and their positions show significant conservation during evolution (Maden, 1990; Ofengand and Bakin, 1997), we can anticipate that the modified nucleotides contribute to ribosome assembly or/and function (Lane et al., 1995; Ofengand et al., 1995).
In the nucleolus of eukaryotic cells, 2′‐O‐methylation and pseudouridylation of pre‐rRNA is accomplished by a large number of different small ribonucleoprotein particles (snoRNPs) (Smith and Steitz, 1997; Tollervey and Kiss, 1997). Each snoRNP contains a specific small nucleolar RNA (snoRNA) and a set of associated snoRNP proteins. The 2′‐O‐methylation and pseudouridylation guide RNAs possess distinctive structural elements and are associated with different sets of proteins (Maxwell and Fournier, 1995; Smith and Steitz, 1997; Tollervey and Kiss, 1997). The methylation guide snoRNAs carry the conserved C, C′, D and D′ box elements that are essential for both the nucleolar accumulation and function of snoRNAs (Caffarelli et al., 1996; Cavaillé and Bachellerie, 1996; Cavaillé et al., 1996; Kiss‐László et al., 1996, 1998; Watkins et al., 1996). The methylation guide snoRNAs select the target ribosomal nucleotides by forming a 10–21 bp Watson–Crick helix with rRNA sequences. This snoRNA–rRNA interaction, in conjunction with the D and C or D′ and C′ boxes of the snoRNA, provides the structural information for the methyltransferase activity to methylate the correct ribosomal nucleotide (Cavaillé et al., 1996; Kiss‐László et al., 1996, 1998; Tycowski et al., 1996).
Information on the structural requirements for accumulation and function of pseudouridylation guide snoRNAs is much more limited. This group of snoRNAs share a common secondary structure that consists of two major hairpins connected by a hinge and followed by a short tail (Ganot et al., 1997b). The single‐stranded hinge and tail regions contain the conserved H (consensus AnAnnA) and ACA box elements, respectively (Balakin et al., 1996; Ganot et al., 1997b). In vertebrates, the box H/ACA snoRNAs, similar to the box C/D methylation guide snoRNAs, are processed from introns of pre‐mRNAs, whereas in yeast, most H/ACA snoRNAs are transcribed from their independent genes by RNA polymerase II (pol II) or, less frequently, are processed from pre‐mRNA introns or polycistronic pre‐snoRNA transcripts (Maxwell and Fournier; 1995; Balakin et al., 1996; Tollervey and Kiss, 1997). Previous studies suggested that the human intron‐encoded U17 and U19 box H/ACA snoRNAs are processed from the removed and debranched host introns by 5′→3′ and 3′→5′ exonucleolytic activities (Cecconi et al., 1995; Kiss and Filipowicz, 1995; Kiss et al., 1996). Supporting this view, recent studies on the biogenesis of yeast box C/D snoRNAs demonstrated that debranching of host intron lariats by the Dbr1p RNA debranching enzyme (Ooi et al., 1998; Petfalski et al., 1998) or endonucleolytic cleavages of polycistronic pre‐snoRNAs by endonuclease III (Chanfreau et al., 1998) are essential for snoRNA production. These cleavage reactions provide the entry sites for exonucleolytic activities responsible for processing of mature snoRNAs (Petfalski et al., 1998; C.Allmang, P.Mitchell, E.Petfalski and D.Tollervey, personal communication; Qu et al., 1999).
The ACA box motif that is located three nucleotides from the 3′ end of box H/ACA snoRNAs, together with the adjacent 3′‐terminal stem, plays an essential role in the processing and/or stability of the RNA. This structural motif determines the correct 3′ terminus of the RNA, most likely by binding snoRNP proteins and thereby protecting the snoRNA sequences from the processing exonucleases (Balakin et al., 1996; Ganot et al., 1997b). The H box was proposed to contribute to the 5′ end formation of box H/ACA snoRNAs (Ganot et al., 1997b). Consistent with this, it is required for the accumulation of at least the intron‐encoded members of H/ACA snoRNAs.
The pseudouridylation guide snoRNAs select the substrate uridines by forming two short base‐pairing interactions with rRNA sequences that flank the target uridine (Ganot et al., 1997a). The two rRNA recognition motifs occupy the opposite strands of an internal loop, termed the pseudouridylation pocket, which is located in the 5′ and/or 3′ hairpin domain of the snoRNA. In this study, the essential elements for accumulation and function of pseudouridylation guide snoRNAs have been analysed by expressing various mutant yeast snR5, snR36 and human U65 snoRNAs in yeast cells. Our results demonstrate that the pseudouridylation guide snoRNAs, in marked contrast to the methylation guide snoRNAs, have to meet very strict and complex structural requirements to ensure efficient snoRNA accumulation and guide RNA function.
The 5′‐terminal cap structure contributes to the stability of box H/ACA snoRNAs
The H and ACA box‐containing snoRNAs are either synthesized from their own genes by pol II or are processed from longer RNA transcripts, such as pre‐mRNAs or polycistronic pre‐snoRNAs (see Introduction). The latter group of snoRNAs undergo exonucleolytic 5′ and 3′ end maturation, whereas the 5′ terminus of snoRNAs synthesized by pol II from independent genes is delineated by the polymerase itself and co‐transcriptionally acquires a 7‐methylguanosine cap structure that is modified later to 2,2,7‐trimethylguanosine (Maxwell and Fournier, 1995).
It has been demonstrated that elements directing the processing of human and yeast intron‐encoded H/ACA snoRNAs are located within the snoRNA sequence itself (Cecconi et al., 1995; Kiss and Filipowicz, 1995; Kiss et al., 1996; Ganot et al., 1997a,b). To assess whether box H/ACA snoRNAs that are synthesized by pol II from independent genes also possess the elements essential for processing from pre‐mRNA introns, a cDNA of the yeast snR5 snoRNA was inserted into the intron of the yeast actin gene that had been placed under the control of the promoter and terminator of the yeast alcohol dehydrogenase (ADH) gene (Figure 1A; Kiss‐László et al., 1996). The resulting ACT/SNR5 expression construct or the wild‐type SNR5 gene were inserted into the pFL45 yeast shuttle vector (Bonneaud et al., 1991) and transformed into the ΔsnR5 yeast strain that lacks a functional SNR5 locus (Ganot et al., 1997a). Northern analyses showed that, as expected, the ΔsnR5 cells transformed with the wild‐type SNR5 gene (Figure 1A, lane 3) accumulated the snR5 RNA at the levels of the control cells (Figure 1A, lane 1). However, in cells transformed with the ACT/SNR5 construct, no snR5 accumulation was observed (Figure 1A, lane 5). Since the actin mRNA was expressed correctly (data not shown), we concluded that the snR5 RNA lacks elements that are essential for RNA accumulation. In fact, the major difference between the wild‐type and the intronic snR5 RNA is that the latter lacks a 5′‐terminal cap, suggesting that the cap is crucial to the stability of snR5.
To investigate this assumption further, the snR5‐coding regions of the pFL45/SNR5 and the pFL45/ACT/SNR5 expression constructs were replaced with a cDNA of the yeast snR36 box H/ACA snoRNA that had been shown to be processed correctly from the yeast actin pre‐mRNA (Ganot et al., 1997a). The resulting pFL45/SNR36 and pFL45/ACT/SNR36 constructs were transformed into the ΔsnR36 yeast strain (Ganot et al., 1997a). Northern analysis showed that, either processed from the actin pre‐mRNA (Figure 1B, lane 1) or transcribed from the SNR5 promoter (lane 6), the snR36 was expressed efficiently. The differences in the levels of RNA accumulation probably reflect the fact that the SNR5 promoter supports transcription more efficiently than does the ADH promoter.
Integrity of the 3′‐terminal helical stem of box H/ACA snoRNAs is essential for RNA accumulation (Balakin et al., 1996). To test whether the 5′‐terminal stem is of similar importance, two mutant snR36 RNAs were constructed; snR36‐S1 and snR36‐S2, which contain two base substitutions in the 5′ stem (Figure 1B). When the mutant snR36 cDNAs were inserted into the pFL45/ACT intronic snoRNA expression construct and transformed into the ΔsnR36 strain, neither snR36‐S1 (Figure 1B, lane 3) nor snR36‐S2 (lane 4) was expressed. However, accumulation of the intron‐encoded snR36 was restored when the S1 and S2 base substitutions were combined to restore the base‐pairing of the 5′‐terminal stem (lane 5), indicating that integrity of this stem, rather than its nucleotide composition, is essential for snoRNA accumulation. Interestingly, both snR36‐S1 and snR36‐S2, as well as snR36‐S1+S2, were expressed under the control of the SNR5 promoter (Figure 1B, lanes 7–9). Immunoprecipitations with antibodies directed against the m3G cap demonstrated that all snR36 transcripts generated from the SNR5 promoter contained the 5′‐terminal m3G cap (data not shown). Collectively, our results demonstrate that the cap structure, most likely by protecting the 5′ end of the RNA, plays an important role in the accumulation of the wild‐type snR5 and the mutant snR36‐S1 and snR36‐S2 box H/ACA snoRNAs.
The H box was shown to be essential for the accumulation of intron‐encoded H/ACA snoRNAs (Ganot et al., 1997b). We tested whether this motif is also required for accumulation of box H/ACA snoRNAs which are transcribed from their independent genes. When the wild‐type H motif (AGACCA) of the yeast SNR5 was replaced with a mutant H box (gGgCCA) (Figure 1A), no snR5 accumulation was detected in the transformed ΔsnR5 strain (lane 4). This demonstrates that the H box is equally important for accumulation of both the intron‐borne and indepedently transcribed box H/ACA snoRNAs and that the 5′‐terminal cap structure cannot compensate the lack of an intact H box. The yeast snR36, like many other yeast H/ACA snoRNAs, carries a short stem–loop structure (IH1) inserted into the major 5′ hairpin domain (Figure 1B) (Ganot et al., 1997b). However, deletion of the IH1 element of the intron‐encoded snR36 (snR36‐ΔIH1) did not interfere with maturation of snR36 (Figure 1B, lane 2), indicating that the IH1 element does not contribute to the correct processing of the snoRNA.
Position of the H box determines the 5′ terminus of intron‐encoded box H/ACA snoRNAs
In H/ACA snoRNAs that are processed from pre‐mRNAs or polycistronic pre‐snoRNAs, the 5′‐terminal stem is either followed by the H box immediately or they are separated by one or two nucleotides (Figure 2A). SnoRNAs that lack spacer nucleotides always carry one unpaired 5′‐terminal nucleotide, while snoRNAs with one or two spacer nucleotides feature two free 5′‐terminal nucleotides, suggesting that the position of the H box relative to the 5′‐terminal stem determines the 5′ terminus of the snoRNA. The snR36 possesses two potential H box motifs that are separated by one or two nucleotides from the 5′ stem (Figure 2A). Nevertheless, utilization of either one of these motifs is expected to result in two unpaired 5′‐terminal nucleotides. Wild‐type and mutant snR36 RNAs carrying H box motifs with altered positions (Figure 2A) were expressed in the ΔsnR36 strain and their 5′ termini were mapped by primer extension (Figure 2B). As expected, for the wild‐type snR36 and the mutant snR36‐H1 and snR36‐H2 RNAs in which the H box was placed unambiguously one or two nucleotides far from the 5′ stem, respectively, the major stop was observed at position U +1 (Figure 2B, lanes 2–4). However, when the H box was placed in the immediate vicinity of the 5′ stem (snR36‐H0), the major transcription stop identified the U +2 residue as the 5′‐terminal nucleotide (Figure 2B, lane 5). Northern analyses showed that the 3′ terminus of the U65‐H0 snoRNA was processed correctly (data not shown). This demonstrates that the H box, most likely in concert with the 5′‐terminal helix, delineates the correct 5′ terminus of box H/ACA snoRNAs that are processed from precursor RNAs.
Human U65 snoRNA directs site‐specific pseudouridylation of yeast 25S rRNA
The 5′ and 3′ hairpins of yeast snR34 and human U65 snoRNAs carry two pseudouridylation pockets that are predicted to direct pseudouridylation of yeast 25S and human 28S rRNAs at two equivalent positions, at Ψ2822/Ψ2876 and Ψ4374/Ψ4428, respectively (Ganot et al., 1997a; Ni et al., 1997) (Figure 3A). To test whether the human U65 can support pseudouridylation of yeast 25S rRNA, the U65 RNA was placed under the control of the SNR5 promoter and transformed into the ΔsnR34 yeast strain (Ni et al., 1997). Northern analysis of cellular RNAs isolated from the resulting U65 (Figure 3B, lane 4) and the control CMY133 and ΔsnR34 yeast strains (lanes 2–3), as well as from human HeLa cells (lane 1), showed that the human U65 snoRNA was expressed efficiently in yeast. Note that the 5′‐terminal region of U65 transcribed from the SNR5 promoter carries 11 leader nucleotides.
The state of pseudouridylation of the 25S rRNA was monitored by primer extension performed on CMC‐treated cellular RNAs. CMC reacts irreversibly with pseudouridines, and the modified pseudouridine‐CMC residue stops reverse transcriptase one nucleotide before the pseudouridylation site (Bakin and Ofengand, 1993). Predictably enough (Ganot et al., 1997a; Ni et al., 1997), disruption of the SNR34 locus abolished pseudouridylation of the 25S rRNA at Ψ2822 (Figure 3B, compare lanes 1 and 2) and Ψ2876 (compare lanes 4 and 5). However, expression of the human U65 RNA in the ΔsnR34 cells restored pseudouridylation of 25S rRNA at Ψ2822 (Figure 3B, lane 3) and Ψ2876 (lane 6). These results demonstrate that the U65 snoRNA represents the human orthologue of yeast snR34 and, more importantly, structural elements required for the function of pseudouridylation guide snoRNAs are conserved between humans and yeast.
Neither the 5′ nor the 3′ hairpin domain of U65 can direct rRNA pseudouridylation alone
Since the H and ACA boxes and the 5′‐ and 3′‐terminal helical stems are essential for accumulation of H/ACA snoRNAs, the function of these elements in the rRNA pseudouridylation reaction cannot be investigated directly by mutational analyses. Recently, we have shown that the human intron‐encoded U24 box C/D snoRNA, when expressed in yeast cells, is processed correctly and packaged into a functional snoRNP particle (Kiss‐László et al., 1998). Therefore, a U65/U24 fusion snoRNA was constructed and placed under the control of the SNR5 promoter (Figure 4A), anticipating that the U65/U24 transcript, even if carrying mutant U65 sequences, would be stable in yeast cells due to the presence of the 5′‐terminal cap structure and the 3′‐terminal U24 snoRNP particle. Indeed, Northern blot analyses of cellular RNAs obtained from the ΔsnR34 strain transformed with the pFL45/SNR/65/U24 construct showed that the U65/U24 snoRNA accumulated efficiently (Figure 4B, lane 2). Moreover, mapping of rRNA pseudouridylation demonstrated that expression of the U65/U24 snoRNA in the ΔsnR34 strain restored the wild‐type levels of pseudouridylation of yeast 25S rRNA at Ψ2822 (Figure 4C, lanes 1–3) and Ψ2876 (lanes 4–6). This demonstrates that the U65/U24 fusion snoRNA is not only stable but also functional in yeast cells.
The 5′‐terminal hairpin (5′hp) followed by the H box motif and the 3′‐terminal hairpin (3′hp) with the ACA box seem to represent two elements that are structurally and perhaps functionally equivalent. To test whether either of these domains can direct rRNA pseudouridylation independently, the 5′‐ or 3′‐terminal domain of the human U65 RNA, U65‐5′hp and U65‐3′hp, respectively, were fused to the U24 RNA and expressed in the ΔsnR34 strain (Figure 5A). Although the two chimeric RNAs accumulated efficiently, in contrast to the U65/U24 RNA (Figure 5B, lanes 3 and 8), neither U65‐5′hp/U24 (lanes 4 and 9) nor U65‐3′hp/U24 (lanes 5 and 10) could restore pseudouridylation of 25S rRNA at Ψ2822 or Ψ2876. This demonstrates that the presence of both the 5′hp and 3′hp is essential not only for snoRNA accumulation, but also for the rRNA pseudouridylation reaction.
Intact H and ACA boxes are required for rRNA pseudouridylation
In the rRNA–guide RNA interaction, the distance between the pseudouridylation site and the H or ACA box of the snoRNA shows a remarkable conservation. Normally, they are separated by 14 nucleotides (Ganot et al., 1997a; Ni et al., 1997). Alteration of the distance between the ACA box of the yeast snR8 and the target uridine residue resulted in either reduced pseudouridine synthesis at the correct site and/or partial modification of an adjacent uridine (Ni et al., 1997). To test directly the importance of the H and ACA boxes in the pseudouridylation reaction, the wild‐type H (AUAGUA) or ACA box sequences of the U65/U24 fusion snoRNA were replaced with the gggGgg and cCc mutant sequences, respectively (Figure 6A). The resulting U65‐H/U24 and U65‐ACA/U24 RNAs accumulated efficiently when expressed in the ΔsnR34 strain (Figure 6A, lanes 3 and 4).
The state of pseudouridylation of 25S rRNA derived from the control CMY133, the ΔsnR34, U65/U24, U65‐H/U24 and U65‐ACA/U24 strains was monitored by primer extension (Figure 6B). The U65‐H/U24 RNA that carried a mutant H box motif failed to direct pseudouridylation of the 25S rRNA at Ψ2822 and, to our surprise, also at Ψ2876 (Figure 6B, lane 9). Likewise, lack of an intact ACA box in the U65‐ACA/U24 RNA inhibited the pseudouridine synthesis at both Ψ2876 (Figure 6B, lane 10) and Ψ2822 (lane 5), demonstrating that the H and ACA boxes play equally important roles in the pseudouridylation reactions directed by both the 5′‐ and 3′‐terminal pseudouridylation pockets.
Helical stems bracketing the pseudouridylation pockets are essential for rRNA pseudouridylation
In the two hairpins of box H/ACA snoRNAs, the pseudouridylation pockets are always flanked by two well‐defined stem structures (Ganot et al., 1997b) (Figure 7A). While the stems at the bases of the 5′ and 3′ hairpins (stems B) are essential for snoRNA accumulation (Balakin et al., 1996; this study), helices above the pseudouridylation pockets (stems U) do not contribute significantly to the stability of the snoRNA (M.‐L.Bortolin, P.Ganot and T.Kiss, unpublished results). Nonetheless, we investigated whether the lower (B) and upper (U) helical stems flanking the 5′ and 3′ pseudouridylation pockets of the U65 snoRNA are essential for rRNA pseudouridylation. To this end, these stems were destroyed in the U65/U24 fusion snoRNA by substitution of the descending (3′ side) strands of each stem with non‐complementary sequences (d). In the next step, the base‐pairing for each helix was restored by substitution of the 5′ side of the stems with appropriate complementary sequences (r). Northern analysis demonstrated that upon transformation of the resulting pFL45/SNR/U65/U24 constructs into the ΔsnR34 strain, all the mutant U65/U24 snoRNAs were expressed.
We tested whether the mutant U65 snoRNAs can restore the pseudouridylation of 25S rRNA at Ψ2822 and Ψ2876 in the ΔsnR34 strain (Figure 7B). When the upper or lower stem in the 5′ hairpin of U65 was destroyed (U65‐5′Ud and U65‐5′Bd), pseudouridylation of 25S rRNA was abolished at Ψ2822 (Figure 7B, lanes 4 and 6) and, to our surprise, also at Ψ2876 (lanes 11 and 13). Restoration of these stem structures in the U65‐5′Ur and U65‐5′Br RNAs re‐established the pseudouridylation of 25S rRNA both at Ψ2822 (Figure 7B, lanes 5 and 7) and at Ψ2876 (lanes 12 and 14). Similar results were obtained upon mutation of the stem structures flanking the 3′ pseudouridylation pocket of U65. Disruption of either the upper (U65‐3′Ud) or the lower stem (U65‐3′Bd) had detrimental effects on pseudouridine formation at both Ψ2822 (Figure 7B, lanes 18 and 20) and Ψ2876 (lanes 25 and 27). When base‐pairing interactions were re‐established, the resulting U65‐3′Ur and U65‐3′Br RNAs restored pseudouridine formation at Ψ2822 (Figure 7B, lanes 19 and 21) and Ψ2876 (lanes 26 and 28). These observations demonstrate that the helical stems bracketing the 5′ and 3′ pseudouridylation pockets of box H/ACA snoRNAs are essential for the rRNA pseudouridylation reaction.
Targeted pseudouridylation of yeast 25S rRNA directed by artificial guide snoRNAs
To select a ribosomal pseudouridylation site, two short sequence motifs in the pseudouridylation pocket of the guide RNA were proposed to form double helices with rRNA sequences that flank the substrate uridine (Ganot et al., 1997a). The importance of this interaction has been confirmed by showing that alteration of rRNA recognition motifs of the yeast snR36 and snR8 interferes with the rRNA pseudouridylation reactions directed by these snoRNAs (Ganot et al., 1997b; Ni et al., 1997). However, it remains unclear whether the two short rRNA recognition motifs provide all the information for selection of a unique pseudouridylation site in pre‐rRNA. To address this question, two slightly altered versions of the yeast snR36 pseudouridylation guide snoRNA, snR36‐m1 and snR36‐m2, were created (Figure 8A). The rRNA recognition motif of snR36 that directs the synthesis of Ψ1185 in the 18S rRNA (Ganot et al., 1997a) was substituted for sequences that, in principle, could select the U2871 residue in the 25S rRNA. The snR31‐m1 and snR36‐m2 RNAs carried identical recognition motifs except that the potential snoRNA–rRNA interaction region was extended by one base pair for the snR36‐m2 RNA. The wild‐type snR36 and the mutant snR36‐m1 and snR36‐m2 snoRNA were expressed from the SNR5 promoter in the ΔsnR36 strain (Figure 8A).
The state of pseudouridylation of 25S rRNA obtained from the control and ΔsnR36 strains as well as from the ΔsnR36 strain expressing either the wild‐type snR36 or the mutant snR36‐m1 or snR36‐m2 snoRNA was assayed (Figure 8B). Expression of the snR36‐m2 (Figure 8B, lane 1) and snR36‐m1 (lane 2) snoRNAs resulted in conversion of the U2871 residue into pseudouridine. Clearly, this pseudouridine was not detectable in the control (Figure 8B, lane 5), ΔsnR36 (lane 4) and snR36 (lane 3) strains. Contrary to expectations, the extended rRNA recognition motif of the snR36‐m2 snoRNA did not improve the efficiency of pseudouridine synthesis. However, as compared with a known pseudouridylation site in the yeast 25S rRNA at Ψ2861, it was apparent that the U2871 residue was converted into pseudouridine with only 40–50% efficacy. Similar results were obtained when another novel pseudouridylation site was introduced into the U2976 position of the yeast 25S rRNA (data not shown). Since snoRNA‐guided pseudouridylation sites are found in both helical and single‐stranded regions of rRNAs, it seems unlikely that the local structural environment of the substrate uridine would greatly alter the efficiency of the pseudouridine synthesis reaction (Ofengand et al., 1995; Ofengand and Bakin, 1997). We therefore propose that in the catalytic centre of these artificial snoRNPs, the substrate uridines occupy a suboptimal sterical position that impairs the isomerization reaction. It is also unlikely that additional, not yet identified snoRNA–rRNA interactions would contribute to the substrate recognition event. This conclusion is supported by the facts that, apart from the short rRNA recognition motifs in the pseudouridylation pockets, no obvious conservation can be found between the yeast snR34 and the human U65 snoRNAs and, more importantly, ribosomal target sequences as short as 12 nucleotides are pseudouridylated efficiently when expressed in the nucleolus (Ganot et al., 1997a). Hence, we conclude that the substrate specificity of the rRNA pseudouridylation reaction mediated by the box H/ACA snoRNAs is provided exclusively by the rRNA recognition motif of the snoRNA.
The nucleolar maturation of eukaryotic rRNAs is assisted by an unexpectedly complex population of snoRNAs (Smith and Steitz, 1997; Tollervey and Kiss, 1997). While a few snoRNAs are required for the nucleolytic formation of mature‐sized rRNAs (Maxwell and Fournier, 1995; Sollner‐Webb et al., 1996), most of them direct the site‐specific 2′‐O‐ribose methylation (reviewed in Maden, 1996; Peculis and Mount, 1996; Tollervey, 1996; Bachellerie and Cavaillé, 1997; Tollervey and Kiss, 1997) or pseudouridylation (reviewed by Maden, 1997, Peculis, 1997; Smith and Steitz, 1997) of rRNAs. We report here a comprehensive analysis of the structural elements essential for accumulation and function of rRNA pseudouridylation guide snoRNAs.
The pseudouridylation guide snoRNAs feature a highly conserved ‘hairpin–hinge–hairpin–tail’ secondary structure with two conserved sequence motifs, the H and ACA boxes. Previous works (Balakin et al., 1996; Ganot et al., 1997b) together with this study (Figure 1A) demonstrate that the box H and ACA motifs are absolutely required for accumulation of both mammalian and yeast H/ACA snoRNAs. These single‐stranded sequence motifs most likely represent protein‐binding signals that are recognized by snoRNP proteins common to this class of snoRNPs (Henras et al., 1998; Watkins et al., 1998). However, it seems very unlikely that either the H (consensus AnAnna) or the ACA (consensus AcA) motif alone could provide sufficient information for binding of snoRNP proteins. Supporting this assumption, the H and ACA boxes are always located in the close vicinity of the 5′‐ or 3′‐terminal helical stems, respectively, that are also required for snoRNA accumulation (Balakin et al., 1996; Figure 1B). Alteration of the distance between the ACA box and the 3′‐terminal stem interferes with the accumulation of yeast snR11 RNA (Balakin et al., 1996). In this study, we demonstrate that the position of the H box relative to the 5′‐terminal stem determines the 5′ end of the intron‐encoded yeast snR36 (Figure 2). These observations strongly support the notion that the H box together with the 5′‐terminal stem, and the ACA box in concert with the 3′‐terminal stem, constitute the recognition signals for snoRNP proteins. Most probably, snoRNP proteins associated with the 5′‐ and 3′‐terminal ‘stem–box’ structural motifs protect the snoRNA sequence from the processing exonucleases and, thereby, control the correct 5′ and 3′ end formation (Balakin et al., 1996, Ganot et al., 1997b).
In yeast, the majority of H/ACA snoRNAs are synthesized from independent transcription units by pol II. Selection of the transcription initiation site and the co‐transcriptionally added 5′ cap determines the 5′ terminus of these RNAs and, therefore, they undergo maturation only at their 3′ ends. The other group of H/ACA snoRNAs that are processed from intronic or polycistronic pre‐snoRNA transcripts undergo both 5′ and 3′ end maturation. Apparently, the basic structural requirements for accumulation, such as the presence of the 5′‐ and 3′‐terminal stems and the H/ACA boxes, are identical for both groups of snoRNAs. However, our results demonstrate that the steric structure of the 5′ and 3′ end‐processed snoRNAs has to conform to more rigorous requirements. When transcribed within the intron of the yeast actin pre‐mRNA, neither the yeast snR5 that is normally transcribed by pol II from its own gene (Figure 1A) nor the human intron‐encoded E3, U17, U19, U64 and U65 (our unpublished results) snoRNAs accumulated in yeast cells. However, when expressed under the control of the SNR5 promoter, all these snoRNAs accumulated efficiently (Figures 1 and 3; data not shown). It is notable that the human U65, when it was synthesized by pol II and carried a 5′‐terminal m3G cap, not only accumulated, but also directed the pseudouridylation of yeast rRNA (Figure 3). Moreover, accumulation of a mutant version of the intron‐processed snR36 RNA carrying two unpaired nucleotides in its 5′‐terminal stem was rescued when it was transcribed from the SNR5 promoter and possessed a m3G cap (Figure 1B). Collectively, our observations suggest that the 5′‐terminal cap structure, through stabilization of the snoRNA transcripts, contributes to the efficient accumulation of box H/ACA snoRNAs in yeast cells.
Thus far, four common snoRNP proteins, Gar1p, Nhp2p, Cbf5p and Nop10p, have been identified for yeast H/ACA snoRNPs (Balakin et al., 1996; Ganot et al., 1997b; Henras et al., 1998; Lafontaine et al., 1998; Watkins et al., 1998). Since the Cbf5 protein shows striking structural similarities to known pseudouridine synthases (Koonin, 1996; Watkins et al., 1998), it is most probably the enzyme that is responsible for the synthesis of ribosomal pseudouridines (Lafontaine et al., 1998). Therefore, each H/ACA snoRNP particle can be considered as a site‐specific pseudouridine synthase. While the Cbf5p provides the catalytic activity, the snoRNA component of the particle provides the specificity for the rRNA pseudouridylation reaction. Indeed, demonstration that novel pseudouridines can be introduced into the yeast 25S rRNA by manipulating the rRNA recognition motif of pseudouridylation guide snoRNAs (Figure 8) proves that all the information necessary to select the correct pseudouridylation sites is carried by the RNA component of the snoRNP particle. These experiments also provide direct evidence that pseudouridylation guide snoRNAs select the target uridines by forming direct Watson–Crick base‐pairing interactions with the target rRNA sequences.
The two major structural domains of box H/ACA snoRNAs, the 5′hp followed by the H box and the 3′hp together with the ACA box, share striking structural and functional similarities. Pseudouridylation pockets are found equally frequently in the 5′hp and 3′hp and, even more tellingly, many snoRNAs carry pseudouridylation pockets in both the 5′hp and 3′hp domains (Ganot et al., 1997a). It has been documented experimentally that the 5′ and 3′ pseudouridylation pockets of yeast snR5, snR34 and human U65 snoRNAs can direct pseudouridylation of rRNAs at two different positions (Ganot et al., 1997a; Figure 3). The H and the ACA boxes are located normally ∼14 nucleotides downstream of the catalytic centre of the corresponding pseudouridylation pocket in the 5′ or 3′ hairpin, respectively (Ganot et al., 1997a; Ni et al., 1997). Alteration of the wild‐type spacing between the ACA box and the 3′ catalytic centre of yeast snR8 impairs the efficiency and correctness of rRNA pseudouridylation directed by the 3′ pseudouridylation pocket of this snoRNA (Ni et al., 1997). This shows that for selection of the correct ribosomal uridine, in addition to the snoRNA–rRNA base‐pairing interaction, the pseudouridine synthase activity also relies on the position of the ACA box relative to the catalytic centre of the snoRNA. Although not yet experimentally supported, it is easy to imagine that the H box that is located 14 nucleotides downstream from the 5′ pseudouridylation centre possesses a function analogous to that of the ACA box.
Demonstration that the yeast snR5, snR34 and the human U65 snoRNPs, and probably many others, possess two independent catalytic centres for rRNA pseudouridylation implies that these snoRNPs carry two copies of the Cbf5p pseudouridine synthase (Ganot et al., 1997a; Figure 3). Moreover, the notion that the 5′‐ and 3′‐terminal domains of these snoRNAs are functionally equivalent presupposes that they bind the same set of snoRNP proteins. This view was strongly supported by recent purification of yeast snR42 and snR30 box H/ACA snoRNPs (Watkins et al., 1998). The isolated snoRNP core particles contained three common snoRNP proteins, the Gar1p, Nhp2p and Cbf5p. Electron microscopy revealed a highly symmetric bipartite structure for these complexes and, intriguingly, predicted a molecular mass that would be consistent with a particle consisting of a snoRNA and two copies of each of the Gar1, Nhp2 and Cbf5 proteins. The detailed architecture of box H/ACA snoRNPs remains to be understood. Since the Cbf5p lacks an apparent RNA‐binding motif, it seems unlikely that it would bind directly to the snoRNA. Another H/ACA snoRNP protein, the Nhp2p, would be a more likely candidate to bind to box H/ACA snoRNAs, since it contains an RNA‐binding motif also present in ribosomal proteins (Koonin et al., 1994; Watkins et al., 1998). The Gar1 snoRNP protein, although it has been reported to interact in vitro with snR10 and snR30 snoRNAs (Bagni and Lapeyre, 1998), seems to bind to the snoRNP particle through interaction with the Cbf5p (Henras et al., 1998).
The 5′hp and 3′hp domains of box H/ACA snoRNPs act apparently in a highly co‐operative manner. Destruction of any of the functionally essential elements—the H or the ACA box and the helical stems bracketing the pseudouridylation pockets either in the 5′ or 3′ hairpin—impeded rRNA pseudouridylation mediated by both the 5′‐ and 3′‐terminal pseudouridylation centres (Figures 6 and 7). We envisage that to construct a functional snoRNP complex, a direct or perhaps an adaptor protein‐mediated interaction is required between the two sets of snoRNP proteins which are bound to the 5′ or the 3′ hairpin domain of the snoRNA. Of course, this model would also explain why H/ACA snoRNAs that possess only one functional pseudouridylation pocket still contain two hairpin domains. The hairpin element that lacks a pseudouridylation pocket is required to provide scaffolding for snoRNP proteins to construct the functionally active bipartite structure of the snoRNP.
Similarly to pseudouridylaton guide snoRNAs, many 2′‐O‐methylation guide snoRNAs feature two rRNA methylation centres (Kiss‐László et al., 1998). The two rRNA methylation domains consist of an rRNA complementary sequence that is followed by either the C/D or C′/D′ box motifs. Interestingly, this bipartite structural organization is preserved even in those methylation guide snoRNAs which do not contain an RNA recognition motif next to the internal C′/D′ boxes (Kiss‐Lászó et al., 1998). At present, the significance of these intriguing structural and functional parallels drawn between the rRNA methylation and pseudouridylation guide snoRNAs is nebulous. However, it might underscore further the notion that the two major classes of eukaryotic snoRNAs evolved from a common ancestor molecule (Ganot et al., 1997b). In the future, an understanding of the molecular mechanisms of the snoRNA‐directed rRNA modification reactions will provide us with more insights into the complex world of small nuclear RNAs and may also facilitate the understanding of other RNA‐guided processes, such as certain RNA editing mechanisms.
Materials and methods
All procedures used for manipulating DNA, RNA and oligodeoxynucleotides were done according to standard laboratory techniques (Sambrook et al., 1989). The identity of all constructions was verified by sequence analyses. Growth and handling of Saccharomyces cerevisiae were done by standard techniques (Sherman, 1991). Construction of yeast strains ΔsnR5, ΔsnR34 and ΔsnR36 has been described (Ganot et al., 1997a; Ni et al., 1997). Expression plasmids were introduced into yeast cells using the lithium acetate transformation procedure (Ito et al., 1983). The following oligodeoxynucleotides were used in this study: (1) TATAAGCTTAATAGGAACTCATGGTG; (2) TATGGTACCTGATGGTTTTCTTATCCTGA; (3) TTTGGTACCTTTCTCGAGCTTCACTTCATTACTCTCTTGTTTAC; (4) TTTAGATCTATAATTGAAGTATATGTACG; (5) TTTGGTACCATCATTCAATAAACTGATC; (6) TTTCTCGAGATATGTACACCTAGAGCG; (7) TATGGTACCTTGCCCTGTGCCTCGCTCG; (8) TTACTCGAGTGATATGAGACGTTCTAATTA; (9) TTTGGTACCTTGAACTGTGCCTCGCTCG; (10) TATGTCGACGGGCTAAAACAATTAGACTTC; (11) AATTGTTTTAGAACGTTGATC; (12) CAACGGGCTAAAGCAATTAGACTTC; (13) CAACGG‐ GCTGAAACAATTAGACTTC; (14) CAACGGGCAAAGCAATTTAGACTTC; (15) GGCTGGAAGTGGGCCAATTTTTTTTGTTCC; (16) TTTGGTACCTCAGCCACCCGCCACTGC; (17) ATTCTCGAGCTGTTCCCATGCTTTCGG; (18) CCGCTCGAGATGCGGCTTACTGTGCAGATGATGTAAAAG; (19) TTCCGCGGCCGCTATGGCCGA‐ CGTCGACTGCAGTGCATCAGCGATCTTGG; (20) ATAGACATATGGAGGCGTG; (21) CCAGCTCAAGATCGTAATAT; (22) GTTATTACATCATTTGA; (23) CCGCTCGAGCGGGTCCCATGCTTTCG; (24) ATGCTTTCGGCACAGAGTCATCC.
Construction of plasmids for transformation of yeast cells
Construction of yeast expression vectors pFL45/SNR5 and pFL45/ACT has been described (Ganot et al., 1997a). Construction of many plasmids relied on PCR amplification reactions. For each amplification reaction, hereafter, the 5′ end‐specific primer is indicated first and it is followed by the 3′ end‐specific primer. The pFL45/SNR expression cassette was obtained as follows. The promoter (oligos 1 and 2) and terminator (oligos 3 and 4) regions of the yeast SNR5 gene were PCR amplified and, after HindIII–KpnI and KpnI–BglII digestions, respectively, were ligated and inserted into the HindIII and BamHI sites of pFL45. To obtain pFL45/ACT/SNR5, pFL45/ACT/SNR36 and pFL45/SNR36, the cDNAs of snR5 and snR36 were PCR amplified using oligos 5 and 6 or oligos 7 and 8, respectively. The amplified fragments were digested by KpnI–XhoI and inserted into the same sites of pFL/45/ACT and pFL45/SNR. The same approach was used to construct pFL45/SNR36‐S1, except that a mutagenic 5′ end‐specific primer (oligo 9) that contained appropriate base changes was used. To create pFL45/ACT/SNR36‐ΔIH1, the 5′ (oligos 7 and 9) and 3′ (oligos 10 and 8) halves of snR36 were amplified, religated using the PCR‐introduced SalI site and inserted into the KpnI and XhoI sites of pFL45/ACT.
Construction of cDNAs of snR36 or snR5 carrying altered H box motifs or 5′ stem sequences was achieved by the megaprimer amplification approach (Datta, 1995). To obtain snR36‐S2 and snR36‐S1+S2 cDNAs, the 5′‐terminal region of snR36 was amplified using primers 7 and 11 or 9 and 11, respectively. In the second amplification step, the resulting fragments were used as megaprimers in combination with a common 3′ end‐specific primer (oligo 8). The mutant snR36 cDNAs were inserted into the KpnI–XhoI sites of pFL45/ACT and pFL45/SNR, resulting in pFL45/ACT/SNR36‐S2, pFL45/ACT/SNR36‐S1+S2, pFL45/SNR36‐S2 and pFL45/SNR36‐S1+S2. To create snR36‐H1, snR36‐H2 and snR36‐H0 cDNAs, megaprimers encompassing the 3′ half of snR36 were generated by using oligos 12, 13 and 14 as 5′‐specific primers that carried altered H box motifs, respectively, and a common 3′ end‐specific primer (oligo 8). In the second amplification reaction, a common 5′‐specific primer (oligo 7) was used with the mutant megaprimers. The amplified cDNAs were inserted into the KpnI–XhoI sites of pFL45/ACT. To obtain pFL45/SNR5‐H, using oligos 15 and 4 and the pFL45/SNR5 plasmid as a template, the 3′ half of the SNR5 gene was amplified and used as a megaprimer with oligo 1 in the second amplification reaction. The resulting snR5‐H cDNA was digested by HindIII and BglII and inserted into the HindIII and BamHI sites of pFL45.
To generate pFL45/SNR/U65, the human U65 snoRNA was PCR amplified using oligos 16 and 17, cut with KpnI and XhoI and inserted into the same sites of pFL45/SNR. The cDNA of human U24 snoRNA was amplified (oligos 18 and 19), digested with XhoI and SalI and cloned into the XhoI site of pFL45/SNR/U65, resulting in pFL45/SNR/U65/U24.
A series of mutant U65 cDNAs, U65‐5′hp, U65‐3′hp, U65‐H, U65‐ACA, U65‐5′Ud, U65‐5′Ur, U65‐5′Bd, U65‐5′Br, U65‐3′Ud, U65‐3′Ur, U65‐3′Bd and U65‐3′Br (Figures 5, 6 and 7), were constructed by PCR amplification using appropriate mutagenic oligonucleotides in combination with either a common 5′ end‐ (oligo 16) or 3′ end‐specific primer (oligo 17). Structures of the mutagenic oligonucleotides are available upon request. The amplified cDNAs were digested with KpnI and XhoI and were used to replace the wild‐type U65 RNA sequences in pFL45/SNR/U65/U24.
Yeast cellular RNAs were isolated by the guanidine thiocyanate/phenol–chloroform extraction method (Tollervey and Mattaj, 1987). For Northern analyses, if not stated otherwise, 10 μg of cellular RNAs were separated on a 6% sequencing gel, electroblotted onto a Hybond‐N nylon membrane (Amersham) and probed with kinase‐labelled oligonucleotides complementary to yeast snR5 (oligo 20), snR36 (oligo 21), U24 (oligo 22), and to either the 5′‐ (oligo 23) or 3′‐terminal (oligo 24) sequences of human U65. Primer extension mapping of the 5′ terminus of snR36 RNA was performed using 5′ end‐labelled oligo 21 and 10 μg of yeast cellular RNAs. Mapping of pseudouridine residues at positions 2822 and 2876 in the yeast 25S rRNA was performed by primer extension analyses of CMC‐alkali‐treated yeast cellular RNAs as described earlier (Bakin and Ofengand, 1993; Ganot et al., 1997). Primer extension products were analysed on 6% sequencing gels.
We are grateful to L.Poljak for critical reading of the manuscript. We thank Y.de Préval for synthesis of oligonucleotides. This work was supported by the Centre National de la Recherche Scientifique and by grants from la Ligue Nationale Contre le Cancer, l'Association pour la Recherche sur le Cancer and la Fondation pour la Recherche Médicale.
- Copyright © 1999 European Molecular Biology Organization