|Home | About | Journals | Submit | Contact Us | Français|
The 5′-exonuclease Rat1 degrades pre-rRNA spacer fragments and processes the 5′-ends of the 5.8S and 25S rRNAs. UV crosslinking revealed multiple Rat1-binding sites across the pre-rRNA, consistent with its known functions. The major 5.8S 5′-end is generated by Rat1 digestion of the internal transcribed spacer 1 (ITS1) spacer from cleavage site A3. Processing from A3 requires the ‘A3-cluster' proteins, including Cic1, Erb1, Nop7, Nop12 and Nop15, which show interdependent pre-rRNA binding. Surprisingly, A3-cluster factors were not crosslinked close to site A3, but bound sites around the 5.8S 3′- and 25S 5′-regions, which are base paired in mature ribosomes, and in the ITS2 spacer that separates these rRNAs. In contrast, Nop4, a protein required for endonucleolytic cleavage in ITS1, binds the pre-rRNA near the 5′-end of 5.8S. ITS2 was reported to undergo structural remodelling. In vivo chemical probing indicates that A3-cluster binding is required for this reorganization, potentially regulating the timing of processing. We predict that Nop4 and the A3 cluster establish long-range interactions between the 5.8S and 25S rRNAs, which are subsequently maintained by ribosomal protein binding.
During ribosome subunit biogenesis in eukaryotes, the nascent pre-rRNA can either undergo cotranscriptional cleavage by the small subunit processome or be transcribed into the 35S pre-rRNA, which is cleaved within the 90S pre-ribosome complex. In either pathway, the early cleavages separate the pre-40S and pre-60S complexes, which are then matured independently (Figure 1A). During 60S maturation, internal transcribed spacers 1 and 2 (ITS1 and ITS2) are removed from the pre-rRNA by endonuclease cleavage followed by exonuclease digestion to yield the mature 5.8S and 25S rRNA. The 60S ribosome maturation pathway involves several intermediate particles and a many different proteins. Several late steps in 60S maturation have recently been characterized (see Lo et al, 2010 and Panse and Johnson, 2010), but much less is known about early pre-60S particles or the actual function of most of the protein components. Understanding of yeast ribosome assembly will be greatly facilitated by the recent determination of the crystal structure of the mature 80S ribosomes (Ben-Shem et al, 2010).
Early in the eukaryotic lineage, the 5.8S rRNA was derived from the 5′-end of an ancestral 23S rRNA-like molecule by insertion of the ITS2 region. Reflecting this, the 5.8S rRNA associates with the 25S rRNA via two regions of extended base pairing between the 5′-region of 5.8S (nts 4–13) and 25S rRNA (nts 404–413), and between the 3′-region of 5.8S (nts 139–155) and 25S rRNA (nts 4–19). 5.8S–25S base pairing is established before or at C2 cleavage, since the 7S pre-rRNA remains associated with 25S rRNA following deproteinization (DT, unpublished). However, secondary structure models of the ITS1 region of the 27SA pre-rRNAs strongly predict that the 5′-region of 5.8S is base paired to the 3′-end of ITS1 in a stem-loop that is incompatible with base pairing to 25S rRNA, implying the presence of a conformational switch (Yeh et al, 1990). It is currently unclear whether the transfer of the 5′-region of 5.8S from ITS1 base pairing to 25S base pairing is provoked by exonuclease digestion of ITS1 from A3 to B1S, or is required for this processing to take place.
The 5′-end of the yeast 5.8S rRNA is heterogeneous due to the use of two alternative processing pathways (Henry et al, 1994). The major, short form of the rRNA (5.8SS) has a 5′-end at site B1S (Figure 1A). This is generated by exonuclease digestion from an upstream cleavage site (A3), which is the target for cleavage by RNase MRP. Processing from A3 to B1S involves two exonucleases: the Rat1–Rai1 heterodimer (Schmitt and Clayton, 1993; Henry et al, 1994; Lygerou et al, 1996; Xue et al, 2000) and Rrp17 (Oeffinger et al, 2009). In addition, the cytoplasmic exonuclease Xrn1 can degrade the A3 -B1S region when Rat1 is inactive (Henry et al, 1994), but probably does not contribute to normal processing (El Hage et al, 2008). Formation of the 5′-end of the less abundant 5.8SL rRNA does not require RNase MRP or the 5′-exonucleases and is believed to involve an unidentified endonuclease (Faber et al, 2006). Similarly, 5′ heterogeneity is observed for 5.8S rRNA in metazoans, plants and other eukaryotes (Henry et al, 1994), suggesting the use of conserved processing pathways.
In addition to the exonucleases, A3 processing requires components of the pre-60S particles (Dunbar et al, 2000; Pestov et al, 2001; Gadal et al, 2002; Oeffinger et al, 2002; Fatica et al, 2003; Oeffinger and Tollervey, 2003; Miles et al, 2005). Loss of these factors (including Rlp7, Cic1/Nsa3, Erb1, Nop7 and Nop15) leads to accumulation of the 27SA3 pre-rRNA and reduced synthesis of 27SBS and mature 5.8SS without concomitant loss of 27SBL or 5.8SL. A further factor, Nop12, is non-essential but required for efficient 25S rRNA synthesis (Wu et al, 2001) and, together with Brx1 and Ebp2 is predicted to be physically close to Rlp7, Cic1/Nsa3, Erb1, Nop7 and Nop15 within the pre-60S particle (Tarassov et al, 2008; Figure 1B). The role of Nop12 in 5.8S rRNA maturation was not assessed (Wu et al, 2001). A further protein, Nop4 (Nop77) was predicted to bind in this region (Tarassov et al, 2008), but its depletion confers a different phenotype, with specific inhibition of cleavage at both A3 and B1L (Bergès et al, 1994; Sun and Woolford, 1994, 1997).
To better understand the roles of the non-enzymatic ribosome synthesis factors implicated in 5.8S 5′-maturation, we identified their sites of interaction with the pre-rRNA using the crosslinking and analyses of cDNA (CRAC) crosslinking technique (Supplementary Figure S1; Granneman et al, 2009). Unexpectedly, binding sites for five factors required for exonucleolytic removal of the A3-B1S fragment clustered near the 3′-end of the 5.8S rRNA (E-site), not the 5′-end. In contrast, Nop4, required for endonucleolytic cleavages in ITS1, crosslinked near the 5′-end of 5.8S. This suggests that a ribonucleoprotein complex, hereafter referred to as ‘A3 cluster', assembled near the 3′-end of 5.8S rRNA is required for the activity of the 5′-exonucleases acting at the 5′-end of 5.8S, whereas endonucleolytic cleavage in ITS1 requires Nop4 binding to the 5′-end of 5.8S. High-throughput CRAC on the Rat1 5′-exonuclease indicated significant crosslinking at the 3′-end of 5.8S and in ITS2 and we furthermore demonstrate that Rat1 can bind the 5′-end of 25S independently of Nop4 and A3-cluster protein Nop15. We propose that A3-cluster proteins act to coordinate events at the 3′- and 5′-ends of 5.8S rRNA, ensuring that 5′-end maturation of 5.8S precedes ITS2 cleavage.
Rat1 has several distinct targets in the pre-rRNA; it accurately processes the 5′-ends of the mature 5.8S (A3-B1S) and 25S rRNAs (C2-C1) and completely degrades the excised A0-A1 and A2-A3 pre-rRNA fragments (Figures 1 and and2).2). In addition, Rat1 functions in the 5′-maturation of intronic snoRNAs, surveillance of mRNAs with defects in splicing and 5′-capping, and transcription termination on both Pol I and Pol II genes. Given this large number of substrates, we anticipated that Rat1 binding would be very transient. To maximize the recovery of RNA binding sites, we therefore developed a novel UV crosslinking system that allows UV irradiation of large volumes of cells actively growing in culture medium. This crosslinking device (the ‘Megatron') consists of a 1.2-m metal tube with a central 205 W, 254 nm UV lamp, and can irradiate 2.5 l of culture in ~1 min (Supplementary Figure S1A). Cells expressing HTP-tagged Rat1 and a non-tagged negative control strain (BY4741) were UV irradiated in the Megatron. A3-B1S processing also requires the 5′-exonuclease Rrp17 and the Rat1 cofactor Rai1 (Fang et al, 2005; Oeffinger et al, 2009). Both were tested as HTP-tagged constructs but neither gave usable crosslinking efficiencies. Illumina-Solexa deep sequencing of two independent Rat1 CRAC cDNA libraries revealed that around a quarter of the reads mapped to rDNA repeats. A multiple sequence alignment of rDNA-mapped reads is shown in Supplementary Table S2. As expected, many other putative Rat1 substrates were identified, which will be discussed elsewhere (manuscript in preparation).
Prominent peaks of Rat1 crosslinking were observed over the 5′-region of the 5′-ETS (Figure 2B), which is degraded by pathways including endonuclease cleavage (Lebreton et al, 2008; Schaeffer et al, 2009; Schneider et al, 2009). These data indicate that, like its human counterpart Xrn2 (Wang and Pestov, 2010), Rat1 is involved in degradation of the cleaved 5′-ETS fragments. A very prominent peak was present at 3′-end of the 5′-ETS, directly upstream of the A1 cleavage site, in agreement with the reported role of Rat1 in degradation of the excised A0-A1 fragments (Petfalski et al, 1998).
Inspection of the regions surrounding the known Rat1 pre-rRNA processing substrates, A3-B1 in ITS1 and C2-C1 in ITS2 (Figure 2) revealed apparent similarities. High levels of crosslinking were seen immediately 5′ to the A3 and C2 cleavage sites; over the A2-A3 region in ITS1 and over the 3′-region of 5.8S and between the 3′-end of 5.8S (site E) and cleavage site C2 in ITS2. Reads including the 3′-region of 5.8S predominately extended through site E, at least 2 nt into ITS2 (Supplementary Figure S2A), indicating that binding occurred on the pre-rRNA, rather than on mature 5.8S rRNA. Substantial Rat1 crosslinking was also observed 3′ to the Rat1 targets, over the 5′-regions of the mature 5.8S (helices H3 and H4) and 25S rRNAs (H11) (Figure 2). In contrast, crosslinking was much lower over the Rat1 processing substrates A3-B1 and C2-C1. The intermediates in A3-B1 and C2-C1 processing are almost undetectable in wild-type yeast indicating high processivity during Rat1 processing. Finally, hits located at H66, H79 and H99 were also frequently found in negative control experiments (marked with asterisks in Figures 2A, B and and3C)3C) and were therefore considered background.
Rat1 also participates in degradation of the excised A2-A3 spacer fragment (Petfalski et al, 1998). This may contribute to Rat1 crosslinking over this region, but mutational analyses (below) indicate that this is not the major source.
The location of site C2 was originally inferred from fingerprinting of in vivo labelled RNA and predicted to lie within the G133–G136 region (Veldman et al, 1980). However, Rat1-associated sequences frequently extended ~8 nt further 3′ to terminate at U140 and A141 in ITS2 (Supplementary Figure S2A and C). This end was previously detected by primer extension in a rat1 mutant (Geerlings et al, 2000) and the same end is observed in the wild-type (see below). 7S pre-rRNA was immunoprecipitated using HTP-tagged Nop7 and 3′-ends were mapped by RNA cloning (Supplementary Figure S2B). The results confirmed that the 3′-end of 7S predominately falls at U140 and A141. We conclude that the major endonuclease cleavage site lies between A141 and G142, as proposed previously (Geerlings et al, 2000), which we refer to herein as site C2.
We conclude that Rat1 is associated with sites flanking its target regions for pre-rRNA processing.
Based on published data (Miles et al, 2005; Tarassov et al, 2008), a group of synthesis factors were predicted to physically interact in early pre-60S particles: Brx1, Cic1/Nsa3, Erb1, Ebp2, Nop4, Nop7, Nop12, Nop15, Pwp1 and Ytm1 (see Figure 1B for interaction map).
Except for Rlp7 and Ebp2, all of these factors were tested in CRAC analyses, and six (Cic1, Erb1, Nop4, Nop7, Nop12 and Nop15) efficiently crosslinked to RNA. The locations of crosslinking sites along the rDNA are indicated in Figure 3A and C, and are displayed on the predicted secondary structures of 25S, 5.8S and ITS2 in Figure 3D. CRAC experiments were performed 2–5 times and to be considered a bona fide RNA binding site, a nucleotide sequence had to be significantly enriched in every experiment. As negative controls, four independent CRAC experiments were performed with the non-tagged parental strain (Figure 3B). Sanger sequencing of Nop4 CRAC cDNA libraries had indicated multiple pre-rRNA binding sites and therefore high-throughput Solexa sequencing was performed together with a negative control to obtain sufficient sequence depth for each binding site (Figure 3C). Recovered sequences that overlapped with control peaks in H66, H79 and H99 at the 3′-end of the 25S were considered background (asterisks in Figure 3). Crosslinking sites were precisely identified by the presence of multiple point deletions or substitutions at a specific position in sequence reads (see Granneman et al, 2009), or a minimal RNA binding site was determined from overlapping sequences.
In CRAC analyses, Erb1 was crosslinked to the 5′-region of the 25S rRNA over H21 and H22 (Figure 3A and D). In the yeast 60S structure, this region forms a compact structure, probably associated with Rpl42, Rpl8, Rpl28 and Rpl15 (Figure 4A and B; Ben-Shem et al, 2010). The average length of RNAs crosslinked to Erb1 was relatively long (59 nt), possibly because this structure confers protection during the RNase digestion used to generate the cloned fragments.
Nop7 predominately crosslinked to H54 in 25S rRNA (Figure 3A and D). This site is within Domain III, in close proximity to the binding sites for Rpl25 and Rpl35 (Figures 3D, ,4A4A and andB).B). Nop7 is genetically linked to Rpl25 (Oeffinger et al, 2002) and cells lacking Rpl35A accumulate 27SA3 pre-rRNA and are defective in ITS2 processing (Babiano and de la Cruz, 2010). In the yeast 60S structure, the N-terminal domain of Rpl25 makes extensive contacts with 5.8S, whereas the C-terminal region mainly contacts Domain III of 25S (Figure 3D). Both CRAC and genetic data, therefore, support direct interactions between Nop7, Rpl25 and Rpl35. Notably, Rpl35, together with Rpl17 and Rpl26, binds in the vicinity of the peptide exit tunnel in mature 60S subunits (Figure 4E and F).
Cic1 and Nop15 crosslinked preferentially to the 5′-end of the ITS2 spacer region (Figure 3A and D). Although many of the RNAs crosslinked to Nop15 and Cic1 originated from the same region, analysis of shorter fragments and specific mutations in the RNAs allowed us to distinguish the respective RNA binding sites (Supplementary Figure S3). Cic1 also consistently crosslinked to H10 in 25S (Figure 3A), albeit less frequently. High-throughput data suggested that Cic1 and Nop15 interact (Tarassov et al, 2008), and the locations of their RNA binding sites strongly support their direct association in pre-60S ribosomes.
The major Nop12 binding sites were located at H8 (Figure 3A), near the 3′-end of the 5.8S rRNA, and in H10, which is formed by base pairing between the 3′-end of 5.8S and the 5′-end of the 25S rRNA (Figure 3D, ITS2 proximal stem). Thus, Nop12 could have a role in formation or stabilization of the interaction between the 5.8S and 25S rRNAs. The proximity of Nop12 binding sites to A3-cluster protein crosslinking sites prompted us to re-analyze the role of Nop12 in 5.8S processing. Because cells lacking Nop12 are cold-sensitive (Wu et al, 2001) pre-rRNA processing was analysed in nop12Δ cells grown at 18 °C for 24 h. Similar to A3-cluster-depleted cells, nop12Δ cells accumulated 35S and 27SA3 pre-rRNAs (Supplementary Figure S4B and C, lanes 11 and 12), albeit modestly. We conclude that Nop12 is a novel A3-cluster protein.
Nop4 crosslinking sites were dispersed in the primary rRNA sequence and mapped to 25S rRNA domains II and III (Figure 3C and D). In the yeast 60S crystal structure, these sites are ~30–80 Å apart and cluster near the 5′-end of 5.8S (Figure 4C and D). Nop4 is an ~80-kDa protein and could conceivably contact many of these sites simultaneously, bringing domains II and III in proximity to each other. The Nop4 binding sites in H26, H32, H33 and H47 are close to 5.8S rRNA and to binding sites for Rpl17, Rpl35 and Rpl37. We predict Nop4 directly contacts these ribosomal proteins. The binding site in H60 is adjacent to Rpl19 binding sites (Figures 3D, ,4C4C and andD).D). Consistent with previous data (Bergès et al, 1994), Nop4 depletion leads to delayed cleavage at site A2, with more severe inhibition at A3, B1L and concomitant decrease in 27SB levels (Supplementary Figure S4B and C, lanes 7 and 8). The 27SB1L/27SB1S ratio was not significantly altered in these cells (Supplementary Figure S4D, compare lane 2 with lane 1), suggesting Rat1 can process residual 27SA3 generated in the absence of Nop4.
We conclude that cleavage at A3 and B1L requires Nop4 binding near the 5′-end of 5.8S, whereas Rat1-dependent processing from A3 requires A3-cluster proteins binding near the 3′-end of 5.8S.
To validate the CRAC data, we performed immunoprecipitation experiments with HTP-tagged proteins and the non-tagged parental strain and analysed co-precipitated RNAs by northern blot analysis, primer extension analysis and EtBr staining (Figure 5). The data were quantified by calculating fold enrichment over background levels observed in experiments with untagged strains (Supplementary Figure S5). Consistent with previous reports, tagged Nop7, Cic1 and Nop15 co-precipitated 27SA/B and 7S pre-rRNAs (Nissan et al, 2002; Fatica et al, 2003; Oeffinger & Tollervey, 2003; Figure 5B and C). HTP-tagged Nop7 also associated with 6S pre-rRNA, a 5.8S rRNA precursor that is 3′-extended by only 8–9 nt (Figure 5B, lane 12). These observations are consistent with the CRAC data and indicate that ITS2 may be a major determinant for pre-60S binding by Cic1 and Nop15, but not by Nop7.
The association of Nop4, Nop12 and Erb1 with pre-rRNA species has not previously been analysed in detail. Both Nop4 and Nop12 co-precipitated 27SA/B and 7S pre-rRNAs, albeit in modest amounts compared with other proteins tested (Figure 5B and C, lanes 11 and 13). Nop12 and Nop4 also reproducibly showed modest enrichment of mature 5.8S and 25S (Figure 5A and B, lanes 11 and 13; Supplementary Figure S5A), suggesting that these proteins remain associated with pre-ribosomes after removal of ITS2. HTP-tagged Erb1 efficiently precipitated 27S pre-rRNA, but precipitated low amounts of 7S and 5′-extended 25S pre-rRNAs (25S' and 26S) compared with other proteins tested (Figure 5B and C, lane 10; Supplementary Figure S5B and C), implying that Erb1 dissociates from pre-ribosomes following cleavage of 27SB pre-rRNA at site C2.
In the 25S rRNA secondary structure, the Erb1 and Nop7 binding sites appear distant from the 3′-end of 5.8S and ITS2 (Figure 3D). However, in the yeast 60S structure (Ben-Shem et al, 2010), these sites are in the vicinity of the Nop12 binding sites and ~30–50 Å (equivalent to 10–15 nt) from the 3′-end of 5.8S rRNA (Figure 4A and B). Nop7 and Erb1 directly interact in pre-ribosomes (Tang et al, 2008), suggesting their participation in long-range interactions bringing domains III and I of 25S close to the 3′-end of 5.8S (Figure 4B).
Collectively, UV crosslinking, protein–protein interaction and pre-rRNA processing data strongly indicate that A3-cluster proteins physically and functionally interact with the pre-60S region surrounding the 3′-end of 5.8S and with ITS2.
Sites of Rat1 crosslinking may reflect targets undergoing active degradation, pre-rRNA docking sites directed by protein association or both. To distinguish between these possibilities, Rat1 crosslinking was analysed in strains depleted of Nop4 or the A3 cluster protein Nop15.
Rat1-HTP strains carrying PGAL::3xHA-NOP4 or PGAL::3xHA-NOP15 were compared with the parental Rat1-HTP strain. Cells were harvested at an OD600 of ~0.5 after 12 h of depletion in glucose-containing minimal medium. At this depletion time point, only modest defects in growth and pre-rRNA processing were observed (Supplementary Figure S6A and B). Crosslinked RNAs were cloned using barcoded 5′ linkers (Supplementary Table S1) and pooled cDNA libraries were Illumina deep sequenced. Depletion of Nop4 or Nop15 caused a ~1.6-fold reduction in the total number of reads that mapped to the rDNA (Figure 6A). To normalize the data sets, 2000 reads mapping to the rDNA from two independent experiments were randomly selected and the averaged number of hits for each nucleotide in the rDNA was plotted. Only binding sites present in all four Rat1-HTP CRAC experiments were considered.
The results for the 5′ ETS and a region comprising ITS1, 5.8S and ITS2 are presented in Figure 6C and D. Average hit densities (sum of the hits for nucleotides in a region) for the 5′ ETS and 5.8S-ITS2 boundary region (nts 2970–3170, shaded region in Figure 6D) from the normalized data sets are plotted in Figure 6B. Numerous changes were observed, but these were strikingly similar following depletion of either Nop4 or Nop15.
Three classes of changes were apparent. Loss of Nop4 or Nop15 did not dramatically alter the distribution of Rat1 crosslinking to the 5′-ETS region including A0-A1 (Figure 6C), whereas crosslinking was increased 2–3-fold over H3/4 at the 5′-end of 5.8S and H11 near the 5′-end of 25S, and also over the 5′-region of ITS1. Binding of Rat1 at all of these sites, therefore, appears to be independent of Nop4 and the A3 cluster. In marked contrast, crosslinking was dramatically (~70%) reduced, across the A2-A3 and 5.8S 3′ to C2 (Figure 6B and D). Notably, association of 5′ with A3 was equally reduced in strains depleted of Nop4 (which inhibits processing before A3 cleavage) and Nop15 (which inhibits processing following A3 cleavage), indicating that loss of A2-A3 fragment synthesis is not solely responsible for the effect. Rat1 association over the 5.8S 3′- to ITS2 5′-region and the number of reads that terminated at C2 (the 3′-end of 7S pre-rRNA) and the level of 7S were similarly reduced (Supplementary Figure S6C). We could not detectably co-precipitate 7S with HTP-tagged Rat1 (unpublished observation), but binding to ITS2 may be transient in vivo and unstable during the extended incubations needed for precipitation.
We propose that Rat1 binding 5′ to A3 at least in part represents binding to 27SA2 and precedes cleavage at A3, while binding 5′ to C2 represents binding to 27SA2 and/or 27SB and precedes cleavage at C2. Pre-binding of Rat1 at the A3 and C2 cleavage sites may explain the strikingly low steady-state abundance of the 3′ products of these cleavages (27SA3 and 26S pre-rRNA, respectively) in wild-type cells. Rat1 association at these sites is apparently dependent on both Nop4 and the A3 cluster.
Based on phylogenetics and structure probing, two distinct structures were proposed for ITS2, termed the ‘ring' (Figure 7C) and ‘hairpin' (Figure 7D) structures (Yeh and Lee, 1990; Joseph et al, 1999). These share the same distal stems but differ in the 25S and 5.8S proximal regions. Notably, the 5′-end of 25S and 3′-end of 5.8S are base paired in the mature ribosome, and are brought into proximity by the hairpin but not by the ring structure. Subsequent genetic data indicated that ITS2 undergoes a conformational change from the ring to hairpin structure during maturation (Cote et al, 2002). Structurally conserved regions of ITS2 are important for pre-rRNA processing (shaded in blue in Figure 7C and D; van der Sande et al, 1992; Van Nues et al, 1995; Peculis and Greer, 1998; Joseph et al, 1999; Cote et al, 2002), including domains II, III.A and III.B. These were proposed to be binding sites for ribosome synthesis factors (Van Nues et al, 1995; Cote et al, 2002) and they are indeed the crosslinking sites for Cic1 and Nop15 (see Figure 7C and D). Deletions and mutations in these regions abrogated 25S rRNA synthesis and caused 27SA accumulation, with a strong block in conversion of 27SA to 27SB (Van Nues et al, 1995; Cote et al, 2002); processing defects similar to those observed A3-cluster protein-depleted cells (Supplementary Figure S4B and C; Van Nues et al, 1995). These findings strongly implicated Cic1 and Nop15 in the structural rearrangement of ITS2.
To assess the RNA secondary structure, we performed in vivo and in vitro chemical modification experiments using dimethyl-sulphate (DMS) (Figure 7), which preferentially methylates adenines and cytosines in single stranded or flexible RNA not bound by proteins. In vitro chemical probing reactions were performed on refolded total RNA, which helps to identify RNA regions potentially protected by protein binding in vivo. We analysed DMS modifications in conditional (PGAL) Nop15 and Cic1 strains, grown for 12 h in glucose medium to deplete the proteins. As controls, we analysed the parental strain (BY4741) and the strain depleted of Nop4, which did not significantly crosslink to ITS2 (Figure 3C). DMS modifications were mapped by primer extension analysis and autoradiography and quantified by normalizing signal intensities for each band to remove differences between lanes (Figure 7E). Note that primer extension was performed on total RNA, revealing the ‘average' secondary structure of all 27S pre-rRNA species.
The 3′ most region of ITS2 could not be analysed but the distal region (stems VI and V) adopts virtually the same secondary structure in vitro or in vivo and in the depleted and parental strains (Figure 7B). In contrast, ITS2 stem II has a more accessible structure in vivo than in refolded RNA, as judged by the intensity of the primer extension stops at nucleotides C2 and C5 (indicated by open circles in Figure 7A, compare lane 9 with lane 1, and see Supplementary Figure S8). This is consistent with C2 and C5 base pairing with G22 and G25 in vitro (dotted arrows in Figure 7), forming a longer stem II very similar to that proposed in the hairpin model (Figure 7D).
In Cic1- and Nop15-depleted cells, C2 and C5 in stem II were almost completely protected from DMS modification in vivo (Figure 7A, compare lanes 3, 4 with lane 1). We conclude that binding of Cic1 (and perhaps also Nop15) to this region blocks formation of these base-pairing interactions, maintaining the shorter stem II structure of the ring structure. Notably, these nucleotides are located in a highly conserved region required for A3-B1S processing (Van Nues et al, 1995).
Stem III (stem III.A in the ring model) was more protected from DMS modification in vivo compared with the refolded RNA (Figure 7A, compare lane 9 with lane 1), but this protection was lost in Nop15-depleted cells (Figure 7A, compare lane 4 with lane 1). The CRAC data indicate that Nop15 binds to this region, which would largely explain the observed differences in DMS modification in stem III.
Depletion of Nop4 did not dramatically affect the in vivo DMS modification pattern in ITS2 compared with the other strains tested (Figure 7A and B). Notably, we did observe a reduction in DMS reactivity in depleted cells at G319 in ITS1, between the A3 and B1L cleavage sites (Supplementary Figure S7, indicated with arrows); however, primer extension stops at this position were also found in unmodified RNA, indicating that it represents a natural primer extension stop (Supplementary Figure S7, lanes 5–8). The significance of this observation remains unclear.
We conclude that Cic1 and Nop15 binding is necessary to maintain or establish a more flexible and open structure in the 5′-end of ITS2, potentially preventing premature formation of the hairpin structure.
The eukaryotic pre-rRNA processing and ribosome assembly pathways are strikingly complex making coordination and timing of key importance. Following cleavage of site A3 by the endonuclease RNase MRP, the 5′-exonuclease Rat1 digests the pre-rRNA back to the major 5′-end of the 5.8S rRNA at site B1S (Schmitt and Clayton, 1993; Henry et al, 1994; Lygerou et al, 1996; Figure 1A). We initially assumed that A3 cleavage to generate a free 5′-end would be necessary and sufficient for recruitment of Rat1, which shows robust, processive activity in vitro on non-specific substrates (Kenna et al, 1993). However, subsequent data revealed that A3 processing requires a group of eight proteins (Nop12, Nop7, Erb1, Ytm1, Rlp7 Nop15, Nsa3/Cic1 and Rrp1), termed the A3 cluster (Tang et al, 2008) (Sahasranaman et al, submitted) (this work). Similarly, Nop4 and other factors are required in vivo for A3 cleavage (Bergès et al, 1994; Sun and Woolford, 1994), even though this site can be cleaved by RNase MRP alone in vitro (Lygerou et al, 1996).
Rat1 showed strong pre-rRNA binding over several different regions, reflecting its diverse targets for processing and degradation. Strong signals were seen over the A0-A1 region of the 5′-ETS, a known Rat1 target, and at sites further 5′, presumably reflecting a role in spacer degradation following endonuclease cleavage by the PIN domain of the Rrp44 component of the exosome complex (Lebreton et al, 2008; Schaeffer et al, 2009; Schneider et al, 2009; Wang and Pestov, 2010). In contrast, fewer hits were recovered over the A3-B1S and C2-C1 regions that are targets for Rat1 processing during pre-rRNA maturation. The steady-state levels of A3-cleaved and C2-cleaved pre-rRNAs (27SA3 and 26S, respectively, in Figure 1A) are very low, indicating that exonuclease digestion very rapidly follows endonuclease cleavage, and processing intermediates are essentially undetectable. These observations indicate that Rat1 is actively recruited to, and highly processive on, the 27SA3 and 26S pre-rRNAs.
Rat1 was strongly crosslinked to regions immediately 5′ to the A3 and C2 cleavage sites, suggesting that it is bound to these locations before cleavage. Rat1 binding 5′ to A3 presumably also partly reflects its role is A2-A3 fragment degradation (Petfalski et al, 1998). However, the similar effects of depletion of Nop4 and Nop15, which differ in their effects on A3 processing, strongly indicates that loss of A2-A3 is not solely responsible for the reduction. In contrast, crosslinking of Rat1 to the 5′-region of 5.8S (H3/4) and H11 in the 5′-region of 25S rRNA was independent of the A3-cluster protein Nop15 and of Nop4. This suggests that these helices provide distinct access points for Rat1 in pre-ribosomes, potentially representing initial binding sites.
CRAC analyses placed Nop7, Erb1, Cic1, Nop12 and Nop15 on the surface of the pre-60S ribosomes in the vicinity of the 5.8S 3′–25S 5′ stem. Using the CRAC data, we can extend the interaction map for the A3-cluster proteins (see Figure 1B). Some of these interactions are superimposed onto a schematic of the pre-rRNA in Figure 8, allowing potential mechanisms to be proposed.
In the mature 60S structure, both the 5′- and 3′-regions of 5.8S rRNA are base paired to the 25S rRNA. However, based on in vitro chemical modification experiments, it was proposed that in the 35S pre-rRNA the 5.8S 5′-region is initially base paired to ITS1 (Yeh et al, 1990), forming a stable stem-loop structure that will be strongly favoured over the long-range interaction with 25S during folding of the nascent transcript. Our in vivo chemical probing experiments are largely consistent with this model (Supplementary Figure S7), as are analyses of the intermediates detected in rat1 mutants in vivo (Henry et al, 1994) and phylogenetic analyses (Van Nues et al, 1994). Rat1 processing of ITS1 will free the 5′-end of 5.8S, which might be a prerequisite for the establishment of base pairing with 25S (Figure 8).
The 3′-end of 5.8S and 5′-end of 25S are separated by ITS2, which was previously reported to be refolded during pre-rRNA maturation from a more open ‘ring' conformation into a closed ‘hairpin' structure (Yeh and Lee, 1990; Peculis and Greer, 1998; Joseph et al, 1999). In the hairpin structure, a proximal stem brings the 5.8S and 25S rRNAs into close proximity, presumably promoting base pairing between them. The ring-to-hairpin transition was shown to be an active process involving unidentified ribosome synthesis factors (Peculis and Greer, 1998). The locations of the binding sites for Nop15 and Cic1 are in strikingly good agreement with the binding sites predicted for these factors. Moreover, the phenotypes of ITS2 mutations expected to block these interactions confer processing defects similar to those seen in A3-cluster mutants. In vivo structure probing indicated that depletion of Nop15 and Cic1 indeed shifted the conformation of ITS2 from the ring to the hairpin structure. The hairpin structure is predicted to be more thermodynamically stable and is very likely to fold spontaneously. Binding of Nop15 and Cic1 blocks folding of the hairpin, potentially establishing the timing of base pairing between 5.8S 3′- and 25S 5′-regions (Figure 8A).
A3-cluster proteins show interdependent pre-rRNA binding (i.e., depletion of a single protein causes loss of many or all A3-cluster proteins; Tang et al, 2008) (Sahasranaman et al, submitted). This implies that they interact—so their binding sites, which are dispersed in the pre-rRNA sequence, must all be brought into proximity. Notably, several 60S r-proteins make contact with multiple sites in the rRNA, which must also be brought together during subunit assembly. We speculate that this is a key function of the ribosome synthesis factors.
Nop7 directly binds Erb1, forming an inter-domain bridge in the 27S pre-rRNA. Cells lacking these A3-cluster proteins fail to assemble r-proteins Rpl35, Rpl37, Rpl26 and Rpl17 in pre-60S complexes (Sahasranaman et al, submitted). All of these r-proteins have binding sites in close proximity to Nop12, Erb1 and Nop7 crosslinking sites (and presumably also Cic1 and Nop15; Figure 3D) and contact both the 25S rRNA and the 5.8S rRNA (Ben-Shem et al, 2010). Nop4 crosslinking sites were located near the binding sites for Rpl17, Rpl35 and Rpl37 in each of two different domains. In addition, Nop7 was genetically linked to Rpl25 and crosslinked adjacent to an Rpl25 binding site in 25S domain III. The Nop4, Nop7 and Erb1 interactions with pre-rRNA bring together sites that are distant in the primary sequence. We speculate that this establishes an assembly platform for Rpl17, Rpl25, Rpl26, Rpl35 and Rpl37, allowing these proteins to simultaneously interact with multiple regions of the pre-rRNA. The H. marismortui homologue of Rpl26 (HmRpl24) is an initiator protein for 50S assembly (Nowotny and Nierhaus, 1982) and appears critical for the tertiary structure of 23S Domain I (Klein et al, 2004). HmRpl24 binds in the centre of 5.8S rRNA and to the region corresponding to H24 near the 5′-end of 5.8S (Klein et al, 2004). Notably, the binding sites for Rpl26 include at least four rRNA regions that are dispersed in the primary and secondary structures, including contacts at H19, in proximity to the Erb1 binding site, and the 5′-end of 5.8S (Figures 3D and 8). We propose that interactions between the A3-cluster proteins bring these regions together to promote Rpl26 binding. In turn, Rpl26 may be required for correct 5.8S rRNA folding, correct positioning of the 5′-region of 5.8S and binding of further r-proteins, including Rpl17, Rpl35 and Rpl37.
Erb1 appears to dissociate from pre-ribosomes after C2 is cleaved (Figure 5), suggesting that this bridging interaction is required only at early stages of 60S assembly. This is consistent with bridging by the A3-cluster proteins establishing long-range interactions that are subsequently maintained by the r-proteins. Future structural analyses and assays for r-protein association in strains depleted of individual ribosome synthesis factors will be required to test these models.
Previous analyses indicated that recruitment of the 18S rRNA 3′-endonuclease Nob1 to the pre-ribosomes is distinct from its activation for pre-rRNA cleavage (Lamanna and Karbstein, 2009; Pertschy et al, 2009; Granneman et al, 2010). This also appears to be the case for Rat1, suggesting that this might be a more general property of eukaryotic pre-rRNA processing enzymes. Pre-rRNA cleavage or digestion is an irreversible step and separation of binding and activation of the processing enzymes presumably permits better regulation and quality control during pre-ribosome assembly.
Rat1 can terminate Pol I transcription by degradation of the nascent transcript (El Hage et al, 2008). Rat1 also degrades the excised A2-A3 fragment (Petfalski et al, 1998), and is therefore capable of initiating degradation at site A2. This makes it important that exonuclease degradation from A2 and A3 be delayed until the transcript is released from the polymerase by cleavage at site B0, almost 4 kb away (Figure 1A). We propose that the structural reorganization of ITS2, induced by A3-cluster binding, is important for regulating the timing of 5′-processing of the 5.8S rRNA. ITS2 was introduced into an ancestral 23S-like rRNA early in eukaryotic evolution and this mechanism offers a potential rationale for the advantage of having, and retaining, an ITS2 region.
Saccharomyces cerevisiae strain BY4741 (MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0) was used as the parental strain (Brachmann et al, 1998). The HTP carboxyl-tagged strains were generated by PCR as described (Rigaut et al, 1999). Strains were grown in YPD (1% yeast extract, 2% peptone, 2% dextrose), YPG/R (1% yeast extract, 2% peptone, 2% galactose, 2% raffinose) or yeast minimal medium (Formedium) lacking uracil and tryptophan, at 30 °C.
In vivo CRAC, western and northern blot analyses were performed essentially as described previously (Granneman et al, 2009) with the following modifications: For the Rat1 CRAC experiments shown in Figure 6, cells were pre-grown in filter sterilized synthetic minimal medium containing galactose and raffinose, but lacking uracil and tryptophan (SG/R-URA-TRP), to an OD600 of 1.0, subsequently shifted to glucose containing medium (SD-URA-TRP) for 12 h to an OD600 of ~0.5. Cultures were UV irradiated in the Megatron (Supplementary Figure S1A) at room temperature for 100 s (equivalent to an average dose of ~1.6 mJ/cm2) and cells were harvested by centrifugation, washed with phosphate-buffered saline (PBS), and stored at −80 °C. Megatron parts were purchased from UVO3 (http://www.uvo3.co.uk; contact Peter Wadsworth ku.oc.3ovu@reteP//:ptth). Sanger sequencing of cDNAs was performed as described (Granneman et al, 2009). Negative control cDNAs were generated from RNA extracted from excised membrane fragments from mock experiments run in parallel. For the high-throughput sequencing analysis, crosslinked RNAs were sequentially ligated to L3 and barcoded L5 adaptors and amplified by RT/PCR (see Supplementary Table S1 for oligonucleotide sequences). Oligonucleotides were purchased from Integrated DNA technologies (IDT). Illumina sequencing (single end 50-bp reads) was performed according to the manufacturer's procedure and reads were aligned to the yeast genome using novoalign (http://www.novocraft.com). Data analyses were performed using pyCRAC, a set of Python tools for high throughput sequence analysis (Webb, Tollervey and Granneman, manuscript in preparation).
Parental (BY4741) and GAL depletion strains (GAL::3HA-nop4, GAL::3HA-cic1 and GAL::3HA-nop15) were grown in YPG/R to logarithmic phase, shifted to YPD and grown for 12 h to an OD600 of ~0.5. For the in vivo modification experiment, 200 μl of 33% DMS solution (diluted in ethanol) was added to 10 ml of cells and shaken for 3 min at room temperature. DMS was quenched using 10 ml of 0.7 M β-mercaptoethanol and 5 ml of water-saturated isopropanol. Cell pellets were subsequently washed once with 10 ml 0.7 M β-mercaptoethanol and once with ice-cold phosphate saline (PBS). RNA extractions were performed as previously described (Tollervey, 1987). For the in vitro DMS experiments, RNA extracted from 5 ODs of cells was dissolved in 400 μl of refolding buffer (50 mM HEPES-KOH pH 7.5, 150 mM NaAc, 1.5 mM MgCl2). RNA was refolded by slowly cooling heat denatured RNA (65 °C 10 min) to room temperature. Refolded RNA was incubated with 20 μl 33% DMS solution (diluted in ethanol) for 3 min at room temperature and reactions were quenched by adding 160 μl of 0.7 M β-mercaptoethanol. One microgram of total RNA was used for primer extension reactions. Primer extension products were resolved on 6% polyacrylamide/7 M urea gels and visualized by autoradiography. Chemical probing data were quantified as described in the main text using a Fuji FLA5000 phosphoimager and the Aida software package.
We thank Ernst Granneman, Peter Davis and the Darwin workshop for help with designing and modifying the Megatron; John Woolford and Aarti Sahasranaman for communicating unpublished results; Adam Ben-Shem, Lasse Jenner, Gulnara Yusupova and Marat Yusupov for making the yeast 80S ribosome crystal structure available; Tollervey laboratory members for helpful suggestions; the Edinburgh Gene Pool Sanger and Solexa Sequencing Facilities; and the Swann Building kitchen staff. This work was supported by the Welcome Trust, the BBSRC (BB/D019621/1), EMBO long-term fellowship (SG) and a Marie Curie EIF fellowship (SG).
Author contributions: SG and DT designed the research. SG, EP performed the research and analysed the data. SG and DT wrote the paper.
The authors declare that they have no conflict of interest.