|Home | About | Journals | Submit | Contact Us | Français|
In eukaryotes, ribosomes are made from precursor rRNA (pre-rRNA) and ribosomal proteins in a maturation process that requires a large number of snoRNPs and processing factors. A fundamental problem is how the coordinated and productive folding of the pre-rRNA and assembly of successive pre-rRNA-protein complexes is achieved cotranscriptionally. The conserved protein Mrd1p, which contains five RNA binding domains (RBDs), is essential for processing events leading to small ribosomal subunit synthesis. We show that full function of Mrd1p requires all five RBDs and that the RBDs are functionally distinct and needed during different steps in processing. Mrd1p mutations trap U3 snoRNA in pre-rRNP complexes both in base-paired and non-base-paired interactions. A single essential RBD, RBD5, is involved in both types of interactions, but its conserved RNP1 motif is not needed for releasing the base-paired interactions. RBD5 is also required for the late pre-rRNP compaction preceding A2 cleavage. Our results suggest that Mrd1p modulates successive conformational rearrangements within the pre-rRNP that influence snoRNA-pre-rRNA contacts and couple U3 snoRNA-pre-rRNA remodeling and late steps in pre-rRNP compaction that are essential for cleavage at A0 to A2. Mrd1p therefore coordinates key events in biosynthesis of small ribosome subunits.
Synthesis of ribosomes is an evolutionarily conserved process in eukaryotic cells. In Saccharomyces cerevisiae each rRNA gene is transcribed into a precursor rRNA, the 35S pre-rRNA, which is processed into 18S rRNA, 5.8S rRNA, and 25S rRNA. Processing includes chemical modifications and endonucleolytic cleavages and requires many small nucleolar RNAs (snoRNAs) (28) and at least 180 different proteins (52). In addition to and prior to the endonucleolytic processing steps, the pre-rRNA has to be both correctly folded and assembled into a 90S preribosomal complex. Three cleavages, A0, A1, and A2, result in a pre-40S subunit that is processed relatively independently of the pre-60S subunit (18). Furthermore, subcomplexes of the pre-rRNP have been identified (31, 43), which apparently associate with the pre-rRNA in a defined temporal order (43). Visualization of nascent pre-rRNP has revealed that sequential pre-rRNA-protein structures form cotranscriptionally, suggesting that assembly is coordinated with the gradual synthesis of the nascent transcript (40). Furthermore, specific proteins remodel pre-40S (46) and pre-60S (30, 39) particles, constituting important steps during maturation. Conformational changes in the pre-rRNP thus most likely occur during both transcription and subsequent processing steps.
Folding of the 35S pre-rRNA is closely connected to pre-rRNP assembly. However, our knowledge of how the 35S pre-rRNA undergoes conformational changes and adopts intermediate structures that allow processing and that precede the final compact fold of the 18S, 25S, and 5.8S rRNAs is very limited. 35S pre-rRNA faces the general problems of avoiding RNA misfolding (7, 22). Folding of RNA can be assisted through interactions with specific proteins and/or RNA molecules. U3 snoRNA base pairs with 35S pre-rRNA and is believed to prevent premature formation of the central pseudoknot in the 18S rRNA and assist cleavage at A0 to A2 (8, 25, 37, 49). These cleavages also require the more transiently interacting U14, snR30, and snR10 snoRNAs (for references, see reference 52). Proteins proposed to affect RNA and RNP folding and assembly can in general be separated into RNA chaperones, RNA annealers, and RNA helicases (44). In addition, specific RNA binding proteins exist, which stably bind to the RNA and contribute to the native RNA structure.
Several trans-acting proteins involved in 35S pre-rRNA processing are likely to assist folding of the 35S pre-rRNA. Among these are RNA helicases, mainly belonging to the DEAD box family (reviewed in references 4 and 34). In vivo, RNA helicases are involved in specific pre-rRNA processing steps, but, with few exceptions, we lack information about their precise function. Dbp4p (29), Has1p (33), and Rok1p (5) are necessary for releasing snoRNAs that base pair with pre-rRNA. Ribosomal proteins stabilize rRNA structures in the mature ribosomal subunits and may also play an early role in processing of 35S pre-rRNA, for example, by stabilizing assembly intermediates (14, 15).
Several eukaryotic ribosomal biogenesis factors have RNA binding capability (16), and binding specificity has been shown in vitro for some of these proteins (see, for example, references 1, 17, 32, and 53). In general, it is not known if and to what extent these or other ribosome biogenesis factors operate as chaperones, annealers, or stabilizers of certain RNA structures in vivo.
Mrd1p is likely to play an important role in pre-rRNA folding and pre-rRNP assembly. Mrd1p is essential for cleavages at A0 to A2 and hence for 18S rRNA biogenesis (26). Mrd1p contains five RNA binding domains (RBDs) that are conserved in sequence and position. Mrd1p is one of three RBD-containing proteins that are conserved throughout eukaryotic evolution and the only one involved in ribosomal maturation (10). Based on previous data (47), it is possible that Mrd1p plays a role in guiding and/or stabilizing productive folding of 35S pre-rRNA. Here, we have asked what is the functional importance of the individual RBDs in Mrd1p.
We conducted mutagenesis on the RBDs in MRD1 in vivo, creating full RBD deletions. We also substituted aromatic amino acid residues at position 3 and 5 in the RNP1 motif, which is important for RNA binding (for a review, see reference 36). The RNP1 substitutions caused a phenotype only in RBD5, where they were lethal, whereas the full deletions showed a graded effect on growth and ribosome biogenesis. An interallelic complementation assay showed that different RBD deletions exhibited allele-specific complementation and noncomplementation, and Miller spreads showed that the different mutants at least partially affected different processing stages and/or functions. Several RBD deletion mutants displayed snoRNA release defects. In the absence of Mrd1p, snoRNA-pre-rRNA base pairing was unresolved. In the case of U3 snoRNA, this effect was mainly coupled to RBD5. These results suggest that the different RBDs of Mrd1p are needed during different steps in preribosome processing, possibly facilitating the creation of physically distinct preribosomal intermediary particles, including the release of U3 snoRNA via the fifth RBD of Mrd1p, thereby enabling events leading to processome compaction.
The strains used are described in Table S1 in the supplemental material. Genomic tagging was performed by one-step PCR-based gene integration (35). The mutRBD5 allele was made in a diploid cell by integration of a fusion PCR fragment between the 3′ part of the MRD1 gene containing the RNP1 mutations in RBD5 and the 13Myc and KanMX6 from pFA6a-13Myc-KanMX6. Haploid mutants were obtained by tetrad analysis in the presence of plasmid pS001 (see Table S1 in the supplemental material). The RNP1 mutations in RBD1 to -4 and all the RBD deletions were constructed essentially as described previously (50). Briefly, the klURA3 marker from pBS1539 (a gift from B. Séraphin), was inserted in each RBD of a MRD1-13Myc allele in the presence of pRS424/MRD1 (see Table S1 in the supplemental material). Next, the cells were transformed with a PCR fragment containing the appropriate RNP1 substitutions or RBD deletions. After the mutations were verified by sequencing and Western analyses, growth was tested on medium containing 5-fluoroanthranilic acid (5-FAA) (Fluka).
Strains containing a mutant gal2 allele were transformed with plasmid pCA077 (see Table S1 in the supplemental material) (a gift from C. Andreasson and P. Ljungdahl) that was linearized with BglII in the GAL2 open reading frame. Uracil prototrophs were tested on antimycin A (Sigma). Resistant cells were selected for loop-out of the URA3 marker on 5-fluoroorotic acid (5-FOA) (Duchefa Biochemie) and retested on antimycin A. Viable cells were considered GAL2.
The plasmid pRS406-PGAL-HA-MRD1 was cut with StuI in the URA3 open reading frame. Transformation and selection for uracil prototrophy produced strains that contained an additional genomic MRD1 wild-type (wt) allele under control of the GAL1 promoter. Functionality, galactose-dependent expression of HA-MRD1, and unaltered expression of Myc-tagged mutants were verified (data not shown).
The ΔRBD mutants obtained by tetrad analyses were crossed and selected for diploids using standard techniques. Diploids were pregrown in nonselective medium to allow for plasmid loss, and cells in log phase were spotted in 10-fold dilutions on nonpermissive medium (5-FOA-, 5-FAA-, 5-fluorouracil (5-FU)-, or glucose-based SC medium). Spotting of equal amounts of cells was at all times verified on permissive medium (SC with glucose or galactose) (data not shown). For plates used at 37°C, 6% glucose was used mostly, as this gave more robust growth. Growth that was reproducible (same result in separate crosses and experiments) and representative for the population was scored.
Yeast cells were grown to an optical density at 600 nm (OD600) of ~0.5. Cycloheximide (Sigma) was added to a final concentration of 100 μg/ml. Extracts were made and analyzed by centrifugation in 10 to 50% sucrose gradients as described previously (47). The positions of 25S and 18S rRNAs, determined by Northern analyses, served as markers for 40S and 60S subunits and 80S ribosomes.
Deproteinization analyses were performed essentially as described previously (29). The treated extracts were analyzed by centrifugation in 10 to 30% sucrose gradients. RNA was analyzed as described above. Proteins were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.
RNA was separated on 1.5% agarose-formaldehyde gels as described previously (45) and blotted onto a Zeta-probe membrane (Bio-Rad). Labeling of probes and hybridization were performed as described previously (47). The oligonucleotide probes used are listed in Table S2 in the supplemental material, and their positions are shown in Fig. Fig.2A.2A. Hybridization signals were visualized with phosphorimager (Fuji, FLA3000) and quantified with Fuji Multi Gauge V3.0 software.
The primary antibody was monoclonal anti-c-Myc (diluted 1:1,000, 9E10; Santa Cruz Biotechnology). Polyclonal goat anti-mouse horseradish peroxidase-coupled antibodies (diluted 1:1,000 to 1:3,000; DakoCytomation), were used for immunodetection with ECL solution (Amersham Biosciences).
Miller chromatin spreads were made as described previously (40). The control strain (PLY094) and viable mutant strains (PLY290, PLY291, PLY292, and PLY350 × PLY306), were grown in YPD. For the lethal Mrd1p mutants (PLY178, PLY361, PLY362, and PLY363), depletion of the gal-controlled wt MRD1 gene was accomplished by switching from YPG to YPD for 17 to 20 h prior to spreading.
For semiquantitative analysis of rRNA transcription and processing (see Fig. Fig.3A),3A), electron micrographs from two to six different experiments for each strain were analyzed. In general, each micrograph displayed multiple genes from a single nucleolus. Multiple micrographs (range, 20 to 86; average, 48) were analyzed for each strain. For each micrograph, an assessment of the general state of the following normal features was made: transcription level, 5′ external transcribed spacer (5′ETS) particle formation, small-subunit (SSU) processome formation, and normal cotranscriptional cleavage at the A2 site. Each of these features was given a score of 0, +1, +2, or +3 for each micrograph. (For these normal features, a control gene would typically score +3) The scores were averaged and normalized to a scale of 0 to 1 for display as bar graphs.
PLY374, PLY375, and PLY388 were grown in YPG to an OD600 of 0.5, shifted to YPD for 6 h, and harvested. PLY094 and PLY023 were grown in YPD to mid-log phase before harvest. Extracts from 20 OD600 units of cells in immunoprecipitation buffer consisting of 10 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 150 mM NaCl, and 0.1% sodium dodecyl sulfate and containing 1× Complete protease inhibitor (Roche Diagnostics GmBH), 20 mM ribonucleoside vandyl complexes (Sigma), and 100U RNaseOUT (Ambion) were immunoprecipitated with anti-c-Myc antibodies as described previously (47). RNA was analyzed by Northern blotting and hybridized as described above.
Immunofluorescence analyses were essentially as described previously (6). Myc-tagged proteins were detected with a monoclonal anti-c-Myc antibody and fluorescein isothiocyanate-labeled secondary anti-mouse antibody (Dako). DNA was stained with DAPI (4′,6′-diamidino-2-phenylindole) in Vectashield mounting medium (Vector Laboratories, Inc.). Cells were viewed in a Zeiss Axioplan 2 fluorescence microscope (Zeiss).
To analyze the functional properties of each individual RBD in Mrd1p in vivo, two sets of mutations were introduced in the MRD1 gene (Fig. (Fig.1A).1A). One set of mutations resulted in the precise deletion of one RBD at a time, called ΔRBD1 to ΔRBD5 or simply Δ1 to Δ5. The other set of mutations introduced amino acid residue substitutions with leucines at positions 3 and 5 in the conserved RNP1 motif in each RBD, called mutRBD1 to mutRBD5. Replacement of the aromatic residues at positions 3 and 5 in RNP1 has been shown to drastically reduce RNA binding (17).
Each ΔRBD MRD1 allele lacked a single RBD and had a 3′-terminal in-frame 13Myc sequence. In all other aspects the alleles were wt. DNA and Western analyses confirmed that the correct RBD mutations were introduced and that the mutant Mrd1 proteins were expressed (Fig. (Fig.1B1B).
Deletion of individual RBDs in all cases had consequences for growth. The ΔRBD1, ΔRBD2, and ΔRBD4 strains were viable (Fig. (Fig.1C),1C), but their growth properties differed considerably (Fig. (Fig.1D).1D). ΔRBD4 exhibited a slight reduction in growth, whereas the effect of ΔRBD2 was more pronounced and ΔRBD1 had a severe slow growth phenotype. In addition, the ΔRBD1 mutant did not support growth at 37°C (Fig. (Fig.1D,1D, compare 30°C and 37°C). At 16°C, none of the ΔRBD1, ΔRBD2, or ΔRBD4 mutations had any significant additional effect on growth compared to the wt (data not shown).
Two of the RBD deletions, ΔRBD3 and ΔRBD5, were lethal (Fig. (Fig.1C).1C). After integration of a genomic wt MRD1 allele, expressed from the GAL1 promoter, a shift to YPD medium resulted in arrested growth for both ΔRBD3 and ΔRBD5, comparable to complete depletion of Mrd1p (Fig. (Fig.1E1E).
In the case of mutRBD1 to -4, no growth impairment was observed (data not shown). In contrast, the mutRBD5 cells were nonviable (see Fig. S1A in the supplemental material) and were dependent on wt MRD1 on a plasmid for growth (see Fig. S1B in the supplemental material). To facilitate analyses of the mutRBD5 mutant, the wt MRD1 gene under the control of the GAL1 promoter was integrated in the genome. In YPD medium, the effect on growth for the mutRBD5 strain was comparable to depletion of Mrd1p (Fig. (Fig.1E1E).
18S rRNA was reduced in all ΔRBD mutants, albeit to different extents (Fig. (Fig.2B).2B). ΔRBD4, the mutant least impaired in growth, had approximately half as much 18S rRNA as wt cells. The ΔRBD1 and ΔRBD2 mutants had a more dramatic drop in 18S rRNA levels, i.e., approximately 5.8 (ΔRBD1) and 4.5 (ΔRBD2) times less 18S rRNA than wt cells. Finally, the lethal ΔRBD3 and ΔRBD5 mutants both had about 4.5 times less 18S rRNA than wt cells after 6 h in glucose medium (Fig. (Fig.2B)2B) and an almost total loss of 18S rRNA after 16 h in glucose medium (data not shown).
The ΔRBD mutants accumulated different amounts of pre-rRNAs (Fig. (Fig.2B).2B). The ΔRBD1 and ΔRBD2 mutants produced little 20S pre-rRNA but accumulated modest amounts of 35S and 23S pre-rRNAs. The ΔRBD4 mutant accumulated relatively large amounts of 35S and 23S pre-rRNAs in spite of producing substantial amounts of 20S pre-rRNA. The ΔRBD3 and ΔRBD5 mutants accumulated considerable amounts of 35S and 23S pre-rRNAs but very little 20S pre-rRNA. Finally, none of the ΔRBD mutations significantly affected the levels of 27S and 7S pre-rRNAs and 25S rRNA (Fig. (Fig.2B2B).
We conclude that all ΔRBD mutations had an effect on 20S pre-rRNA and 18S rRNA steady-state levels and that this effect was proportional to their effect on growth. Since production of 20S pre-rRNA requires A0 to A2 cleavages, we also conclude that these cleavages were inhibited and/or delayed by all the ΔRBD mutations. Despite this common negative effect on the A0 to A2 cleavages, we detected different levels of pre-rRNAs. These differences did not appear to be explained by differences in transcription levels (Fig. (Fig.3).3). Instead, this indicates that there were differences in turnover of the pre-rRNAs in the different ΔRBD mutants.
In the mutRBD1 to -4 strains the levels of 25S, 18S rRNA, and the pre-rRNA processing intermediates were not influenced (see Fig. S2 in the supplemental material). In contrast, the mutRBD5 strain had a reduced level of 18S rRNA after 6 h in YPD medium (about 3.7 times less than in wt cells) and essentially wt amounts of 25S rRNA (Fig. (Fig.2B).2B). The reduction of 18S rRNA was comparable to the effect of total Mrd1p depletion. Likewise, the processing pattern for the mutRBD5 strain was very similar to the one seen after total depletion of Mrd1p (Fig. (Fig.2B).2B). The A0 to A2 cleavages were inhibited, with a drastic reduction of 20S pre-rRNA and substantial accumulation of 35S pre-rRNA and 23S rRNA. There was no or little effect on 27S and 7S pre-rRNAs (Fig. (Fig.2B).2B). The drastic effect of deletion of the entire RBD5 and depletion of the complete Mrd1p was therefore phenocopied by two amino acid residue substitutions in the RNP1 motif in RBD5, demonstrating the functional importance of RBD5.
All of the mutant strains showed defects in cotranscriptional processing compared to the control wt strain (Fig. (Fig.3).3). As shown in Fig. Fig.3A,3A, a semiquantitative analysis of multiple genes from each strain showed that the mutant strains tended to have fewer transcripts per gene, as well as fewer genes displaying transcripts with small 5′ETS particles (which are U3 snoRNP dependent) or SSU processomes (which represent a compaction of the 5′ half of the pre-rRNP into a larger particle) and normal levels of cotranscriptional cleavage. This cleavage, which corresponds to cleavage at A2 (Fig. (Fig.2A),2A), is shown near the 3′ end of the wt gene in Fig. Fig.3B.3B. Examples of rRNA genes from the mutant strains are shown in Fig. 3C to H. Cotranscriptional cleavage was absent in all the mutant strains as seen by the presence of mostly full-length transcripts at the 3′ ends of the genes (Fig. 3C to H). This can reflect either a slowing or an absence of the A0 to A2 cleavages (40). The decrease in cotranscriptional cleavage was accompanied by a decrease in 20S pre-rRNA for all of these strains (Fig. (Fig.2).2). Some of the mutant strains, especially ΔRBD1, displayed transcripts that appeared to be extensively degraded (P. Lundkvist et al., unpublished data).
Defects in particle assembly (both 5′ETS particles and SSU processomes) were observed to various degrees in all of the mutant strains compared to the wt, as summarized in Fig. Fig.3A.3A. Common configurations in the ΔRBD strains were genes with very few particles (Fig. (Fig.3E),3E), genes with 5′ETS particles but no SSU processomes (Fig. 3C and D), and genes with a few 5′ETS particles and SSU processomes on the more mature transcripts (Fig. 3F, G, and H). The genes from the ΔRBD4 strain were clearly the most normal on average, though delayed in cotranscriptional cleavage. The ΔRBD4 strain displayed the highest level of SSU processomes, correlating with the mild growth phenotype of this strain. Formation of 5′ETS particles, though less efficient than in control cells, was not predictive of the severity of the mutation and was reasonably similar in all mutant strains and similar to the level seen when MRD1 was depleted (Fig. (Fig.3A)3A) (47).
There is also a suggestion from the electron microscopy (EM) data that lethalities from the ΔRBD3 and ΔRBD5 mutations are due to different mechanisms, in that ΔRBD3 had the most severe defect in transcription level, while ΔRBD5 and mutRBD5 were the only strains other than ΔRBD4 that displayed SSU processomes, albeit at a greatly reduced level. In addition, these SSU processomes were in a loose configuration and had not progressed to the compact state seen in wt cells. We conclude that deletion of RBD1 to -3 interfered with normal cotranscriptional processome formation, whereas mutations in RBD5 allowed SSU processomes to form but inhibited subsequent processome compaction.
Immunofluorescence staining showed that all of the partially functional ΔRBD mutant Mrd1p's (ΔRBD1, ΔRBD2, and ΔRBD4) and the nonfunctional ΔRBD5 mutant Mrd1p (Fig. (Fig.4A),4A), as well as the nonfunctional mutRBD5 Mrd1p (Fig. (Fig.4B),4B), were present mainly in the nucleolus. The distribution in the nucleus was similar to that of wt Mrd1p (26). The ΔRBD3 mutant Mrd1p had a more uniform nuclear staining, although the most intense staining was still nucleolar (Fig. (Fig.4A).4A). The redistribution of the ΔRBD3 mutant Mrd1p was not due to competition by the wt Mrd1p, because the distribution of the ΔRBD3 mutant Mrd1p was the same in cells depleted for wt Mrd1p (data not shown).
We conclude that RBD1, -2, -4, and -5 in Mrd1p are redundant for achieving a predominantly nucleolar localization and that lack of RBD3 still permits the Mrd1p protein to enter the nucleus and the nucleolus and to be concentrated in the nucleolus, albeit with somewhat reduced efficiency.
wt Mrd1p is assembled into 5′ETS RNP complexes early during transcription and is required for compaction of the 18S part of the 35S pre-rRNA into SSU processomes (47). In sucrose gradients (Fig. (Fig.4C),4C), we found a broad distribution of wt Mrd1p, presumably reflecting association with various forms of pre-rRNPs during assembly and processing. In addition, we routinely detected wt Mrd1p in fractions corresponding to very large complexes, cosedimenting with polyribosomes.
All ΔRBD Mrd1ps and the mutRBD5 Mrd1p cosedimented with 35S pre-rRNA in 80S to 90S complexes (Fig. (Fig.4C).4C). In fact, some of the mutants, especially ΔRBD3 and ΔRBD5, accumulated in 80S to 90S complexes, possibly reflecting inhibited and/or delayed processing of 35S pre-rRNA or 23S rRNA. In agreement with these results, the mutRBD5, ΔRBD3, and ΔRBD5 Mrd1ps all associated with 35S pre-rRNA and 23S rRNA as shown by immunoprecipitation (Fig. (Fig.4D).4D). In combination with the sucrose gradient results, this shows that the lethal mutations in Mrd1p do not prevent association with 35S pre-rRNA in 90S pre-rRNP complexes. Instead, increased amounts of 35S pre-rRNA and 23S rRNA were associated with these mutated Mrd1ps.
Mrd1p is not required for U3 snoRNA association with pre-rRNA, but the addition of a green fluorescent protein domain to Mrd1p interferes with productive U3 snoRNA association within the preribosomal complex (47). Furthermore, the ΔRBD3, ΔRBD5, and mutRBD5 Mrd1ps all showed an increased association with U3 snoRNA compared to wt Mrd1p (Fig. (Fig.4D).4D). These results, in combination with the association of the ΔRBD3, ΔRBD5, and mutRBD5 proteins with pre-rRNP (Fig. (Fig.4C)4C) and 35S pre-rRNA (Fig. (Fig.4D),4D), indicated that the mutant proteins confer an increased association of U3 snoRNA with the pre-rRNP. Mrd1p and some of its RBDs could therefore have a role in release of U3 snoRNA and other snoRNAs from preribosomal complexes.
We examined the distribution in sucrose gradients of the three snoRNPs that are essential for A0 to A2 cleavages (U3, U14, and snR30) in the nonviable ΔRBD and mutRBD5 mutants. The mutants were assayed under conditions where snoRNPs in wt cells were clearly present both as free monoparticles and in preribosomal complexes (29). The cells were grown in synthetic medium with galactose to log phase before they were shifted to the same medium but with glucose for 6 h. In wt cells, a large fraction of U3 and U14 snoRNPs was present in free particles (fractions 1 to 6 in Fig. 5A and C). Compared to the wt, all the nonviable mutants, including that with depletion of Mrd1p, displayed a shift of U3 snoRNP to preribosomal complexes, cosedimenting with 35S pre-rRNA (Fig. (Fig.5A).5A). This shift was least pronounced in the ΔRBD3 mutant and most striking in the ΔRBD5 mutant, in which the shift was as extensive as after depletion of Mrd1p. An accumulation of U3 snoRNP in preribosomal complexes was also seen in the ΔRBD1 mutant after a 3-h shift to 37°C, a temperature at which this mutant rapidly loses viability (Fig. (Fig.5A5A).
For U14 snoRNP, a small but significant shift from the wt distribution was evident for the ΔRBD1, ΔRBD5, and mutRBD5 mutants (Fig. (Fig.5C).5C). The most pronounced effect was seen after depletion of Mrd1p. For snR30, depletion of Mrd1p resulted in a clear shift to large pre-rRNPs (Fig. (Fig.5E5E).
We investigated if the release defect in the Mrd1p mutants was due to unresolved snoRNA-pre-rRNA contacts or if it was entirely protein dependent. Cells were grown as described above, and the extracts were treated with proteinase K prior to analyses in 10 to 30% sucrose gradients. Proteinase K effectively digested the proteins in the extract under these conditions (see Fig. S3 in the supplemental material), and hybridization to 5S and 5.8S rRNAs supported the conclusion that base-paired interactions were preserved (Fig. 5G and H). Strikingly, the ΔRBD5 mutant still showed a strong presence of U3 snoRNA in preribosome complexes (overlapping with the position of 35S pre-rRNA); this was equally strong as after depletion of Mrd1p (Fig. (Fig.5B).5B). In contrast, U3 snoRNA in the mutRBD5 mutant was released from the larger complex by digestion of proteins. The ΔRBD1 and ΔRBD3 mutants also retained some U3 snoRNA in preribosomal complexes, but to a considerably lesser extent than ΔRBD5.
The effect of deproteinization on the U14 snoRNP distribution was less pronounced (Fig. (Fig.5D).5D). Although full Mrd1p depletion resulted in cosedimentation of U14 snoRNA and 35S pre-rRNA in preribosome complexes, all the RBD mutants showed a U14 snoRNP gradient distribution that was similar to that of the wt. After proteinase K treatment, a small amount of snR30 cosedimented with 35S pre-rRNA in wt cells, and no clear difference was observed upon depletion of Mrd1p (Fig. (Fig.5F5F).
We conclude that Mrd1p is needed for release of U3, U14, and snR30 snoRNPs from preribosomal complexes. In the absence of Mrd1p, a majority of the U3 and U14 snoRNPs are retained in RNA-RNA base-paired structures. The RBD mutants exhibited a clear difference in relation to the mechanism of retention of U3 snoRNP in preribosomal complexes. RBD5 was required for resolving base pairing to a greater extent than RBD1 and RBD3. Strikingly, the essential aromatic residues in RNP1 of RBD5 were not required for releasing U3 snoRNA from base-paired structures, indicating a functional separation between different parts of RBD5. For U14 snoRNP, no single RBD could be linked to release from snoRNA-pre-rRNA interactions.
To learn more about the function of the individual RBDs in Mrd1p, we asked whether different ΔRBD alleles could complement each other in vivo. We crossed the different ΔRBD alleles in all pairwise combinations. The cells contained a wt MRD1/URA3 plasmid, and we scored for growth complementation in the diploid strains selected on 5-FOA (Table (Table1).1). Suppression of growth defects was scored by comparing growth of the heterozygous strains to growth of homozygous diploids. The viable ΔRBD1, ΔRBD2, and ΔRBD4 alleles crossed with ΔMRD1 grew at a rate similar to that of the homozygous ΔRBD1, -2, and -4 diploid strains (Table (Table1),1), showing that none of these alleles were haploinsufficient.
We were further concerned that recombination or gene conversion between the ΔRBD alleles could result in recreation of a wt MRD1 allele in the diploid cells. We analyzed the ΔRBD1 × ΔRBD2 and ΔRBD1 × ΔRBD3 diploid cells in some detail, since these crosses showed the strongest suppression of growth defects (Table (Table1).1). In Western blots, we found that in both crosses, only the mutated versions of Mrd1p were present (see Fig. S4A in the supplemental material). Tetrad analysis of the ΔRBD1 × ΔRBD3 diploid further demonstrated that each haploid segregant showed reappearance of the original phenotypes and that the segregants contained the expected RBD deletion versions of Mrd1p (see Fig. S4B and C in the supplemental material). Thus, there was no sign of DNA rearrangement for the ΔRBD1 × ΔRBD2 and the ΔRBD1 × ΔRBD3 strains. All other crosses studied grew in a reproducible way. No or very few revertant cells with wt growth characteristics, indicative of the presence of the wt MRD1 gene, were seen. The growth properties described for all crosses were therefore most likely due to interallelic complementation/noncomplementation and not to DNA rearrangements in the mrd1 alleles.
In Table Table1,1, it is evident that interallelic complementation was most pronounced for the ΔRBD1 × ΔRBD2 and ΔRBD1 × ΔRBD3 crosses. It is also evident that all crosses involving ΔRBD4 grew roughly at the same rate as the ΔRBD4 × ΔRBD4 cells, with the exception of ΔRBD4 × ΔRBD5. The effect of possible complementation of RBD4 function in trans therefore was at most marginal. This was in agreement with the finding that the ΔRBD4 cells grew almost as wt and exhibited minor defects in pre-rRNA processing (Fig. (Fig.1B1B and Fig. Fig.2B),2B), suggesting that RBD4 is not contributing any essential property.
It was also significant that the ΔRBD2 × ΔRBD3 cross did not show interallelic complementation, suggesting that RBD2 and RBD3 have to be present in the same Mrd1p molecule. In fact, the ΔRBD2 × ΔRBD3 cross grew slightly worse than the ΔRBD2 homozygous diploid, suggesting a slight dominant-negative effect. In addition, the ΔRBD3 × ΔRBD5 cross showed very weak interallelic complementation and only at 30°C.
To verify the result from the 5-FOA selection, we repeated all crosses in the presence of a MRD1/TRP1 plasmid. We also performed crosses between the viable alleles (ΔRBD1, -2, and -4) and the lethal alleles (ΔRBD3 and -5) containing a genomic wt MRD1 gene expressed from the GAL1 promoter. The results so far described were essentially the same for 5-FAA selection (used for strains with the MRD1/TRP1 plasmid), glucose shutoff, and 5-FOA selection (used for strains with the MRD1/URA3 plasmid) (see Fig. S5A and B in the supplemental material, and data not shown).
There was a subtle but significant growth defect at 22°C when ΔRBD5 was crossed with wt MRD1 (Table (Table1;1; see Fig. S5C in the supplemental material). Also, ΔRBD1, ΔRBD2, and ΔRBD4 crossed to ΔRBD5 all exhibited a weaker growth than the respective homozygous diploids (Table (Table1).1). Since the ΔRBD1, -2, and -4 alleles were not haploinsufficient, the ΔRBD5 allele had a dominant-negative effect on the ΔRBD1, ΔRBD2, and ΔRBD4 alleles and on wt MRD1.
However, we observed significant differences in growth for crosses between the ΔRBD5 allele and the ΔRBD1, -2, and -3 alleles depending on the selection method (5-FOA versus 5-FAA and glucose shutoff). In the case of 5-FAA and glucose shutoff, no dominant-negative effect was seen for ΔRBD5 crossed with ΔRBD1 and -2 as had been seen with 5-FOA, and instead there was growth enhancement, especially at 37°C, for these crosses (see Fig. S5A and B in the supplemental material). Furthermore, the ΔRBD3 × ΔRBD5 cross, which did not grow on 5-FOA (Table (Table1),1), was able to grow on 5-FAA medium at 30°C at the same rate as ΔRBD1 × ΔMRD1 (see Fig. S5A in the supplemental material). We conclude that the ΔRBD5 allele can weakly complement slow growth and temperature-sensitive lethality and that this property is lost in the presence of 5-FOA. ΔRBD4 × ΔRBD5 was an exception, with a dominant-negative growth effect independent of the selection medium and most pronounced at 22°C (Table (Table1;1; see Fig. S5D in the supplemental material). In theory, 5-FOA is not in itself detrimental to cells, but in the presence of the URA3 gene, 5-FOA is converted to toxic 5-FU. The positive enhancement contributed by ΔRBD5 at 37°C for both ΔRBD1 × ΔRBD5 and ΔRBD2 × ΔRBD5 was not seen in the presence of 20 mM 5-FU (see Fig. S5E in the supplemental material). In contrast, the ΔRBD1 × ΔRBD3 cross still showed enhancement of growth on 5-FU medium. These results demonstrated a linkage between the negative effects of ΔRBD5 and 5-FOA/5-FU. 5-FU has previously been linked to ribosome biogenesis, as discussed below.
We conclude from all crosses that there is a ΔRBD allele-specific pattern of complementation and noncomplementation. This suggests that the RBDs in Mrd1p function in a specific relationship to each other. To assess if the interallelic complementation was related to processing of pre-rRNA, we analyzed pre-rRNA and rRNA in ΔRBD1 × ΔRBD3 cells by Northern analyses. The steady-state level of 18S rRNA was higher in the ΔRBD1 × ΔRBD3 cells than in the ΔRBD1 or ΔRBD3 strain (Fig. (Fig.6A).6A). Also, there was an increase in 20S pre-rRNA (Fig. (Fig.6B).6B). Furthermore, EM analyses showed signs of improved cotranscriptional processing in ΔRBD1 × ΔRBD3 cells, most notably increased numbers of SSU processomes (Fig. 3A and I). These data suggest that the observed suppression of the growth defects was at least partly connected to an increase in synthesis of 18S rRNA.
We show that the five RBDs of Mrd1p are all functionally important and contribute to the cleavages at A0 to A2 and that they at least to some extent perform different functions. This is in agreement with the position-specific evolutionary conservation of all five RBDs in Mrd1p (26). The linker regions connecting the RBDs in Mrd1p are long (257, 114, 64, and 23 residues in linkers 1 to 4, respectively), which suggests that several of the RBDs may interact with separate targets (48). The RNP1 mutations showed that the aromatic residues are essential only in RBD5. It is therefore possible that RNA contact is essential for RBD5, while RNA contacts for RBD1 to -4 are not individually functionally important. In the EM analyses we detected differences in RBD function. The ΔRBD5 and mutRBD5 mutants support formation of visible SSU processomes, while the ΔRBD1, -2, and -3 mutants do not (Fig. (Fig.3).3). Defects in transcription and 5′ETS particle formation are also more pronounced for ΔRBD3 than for any other mutant and even more so than after full Mrd1p depletion (Fig. (Fig.33).
The allele-specific complementation and noncomplementation results furthermore indicate that the RBDs in Mrd1p operate as separate functional units (Table (Table1).1). RBD1 seems to function relatively independently of RBD2 to -5, because its function could be provided in trans with all the other ΔRBD mutants. The considerable length of linker 1, which is conserved in spite of the presence of an extra RBD within this linker in most species, strengthens this conclusion. According to predictions using PONDR (data not shown), linker 1 is intrinsically disordered. This property may be functionally relevant, since it is conserved among Mrd1p orthologs (data not shown) and since a high level of structural disorder is common in known RNA chaperones (44).
RBD2 and -3 must be present in the same Mrd1p molecule, and apparently they are needed at the same step during assembly and processing, although RBD3 is of greater importance. The long distance between RBD2 and -3 (114 residues) and the fact that specific defects are seen in the ΔRBD3 mutant compared to the ΔRBD2 mutant (Fig. (Fig.33 and and4A),4A), argue that RBD2 and -3 may interact with separate targets. RBD3 has some characteristics in common with the U2AF homology unit type of RBDs (27), making it possible that RBD3 interacts with proteins.
In proteins with multiple RBDs, the binding affinity and specificity are often increased when two adjacent RBDs and a short linker cooperatively bind RNA (for reviews, see references 36 and 42). Linker 4 is the shortest one in Mrd1p, and RBD4 and -5 could function together, although apparently with unequal contribution since the lack of RBD4 does not affect Mrd1p function to any great extent, while RBD5 is essential. We also observed a dominant-negative effect of ΔRBD5 with ΔRBD4 that was FOA independent (see Fig. S5D in the supplemental material). This could possibly be explained by a steric effect, where binding of RBD4 present in an Mrd1p lacking the adjacent RBD5 could interfere with binding of a second Mrd1p with RBD5 but lacking RBD4, thereby delaying the essential action of RBD5.
Crosses involving ΔRBD5 displayed sensitivity to FOA and 5-FU but not to 5-FAA. 5-FAA toxicity is caused by its conversion to 5-fluorotryptophan (51), which has not been coupled to synthesis of ribosomes. 5-FU sensitivity has been linked to ribosome biogenesis and exosome-dependent rRNA quality control (13). The mechanism has been proposed to involve the covalent binding of the H/ACA-specific pseudouridinylase Cbf5p to 5-FU-containing RNA (24). In normal cells, these adducts would presumably be removed by exosome degradation.
It is possible that the effects of the ΔRBD5 mutation and incorporation of 5-FU in pre-rRNA converge. Disturbance of pre-rRNP remodeling induced by ΔRBD5 could, for example, impair removal of one or several Cbf5p-containing H/ACA snoRNAs.
We found full complementation only for the ΔRBD1 × ΔRBD2 combination and only at 37°C. This shows that optimal Mrd1p function requires that all RBDs be present in the same molecule, especially at lower temperatures. It is therefore likely that the different functions of the RBDs are coordinated and influence each other. It also suggests that the requirement for Mrd1p may change slightly from low to high temperature, possibly reflecting that low temperature requires activities that increase RNA/RNP fluidity and that higher temperature requires stabilization of structures.
Our interallelic complementation results show that more than one Mrd1p molecule can be involved in the processing of an individual pre-rRNP. This implies either that two or more Mrd1p molecules are simultaneously present in each pre-rRNP or that Mrd1p is in a dynamic equilibrium with each pre-rRNP during processing. The latter could possibly reflect that Mrd1p has not been identified as a stable component of biochemically isolated pre-rRNPs. However, at present, we cannot rule out either of the two possibilities.
A subset of ribosomal biogenesis factors is necessary for the association of U3 snoRNA with pre-rRNP complexes and subsequent stabilization (12, 43). Mrd1p is not required for U3 snoRNA association with pre-rRNP (47). The opposite function, of facilitating the release of snoRNAs from preribosomal complexes, has been found for three helicases, Dbp4p (29), Has1p (33), and Rok1p (5), and mainly involve the essential snoRNAs (5). In addition, two proteins that lack helicase motifs, Esf1p (41) and Esf2p (23), also cause a snoRNA release defect when depleted, but it is not known if the release defect is due to unresolved base pairing.
Mrd1p is required for release of snoRNAs (U3, U14, and snR30 snoRNAs) from pre-rRNP complexes, and furthermore, it is the first nonhelicase protein shown to be essential for resolving base pairing between the U3 and U14 snoRNAs and pre-rRNAs. Mrd1p lacks any recognizable helicase motif, suggesting that the mechanism of release is indirect. Mrd1p has not been specifically copurified with any of the RNA helicases directly needed for snoRNA release and is therefore probably not a helicase cofactor. The fact that Mrd1p affects release of snR30 from non-base-paired (protease-sensitive) structures and release of U3 and U14 snoRNAs from RNA-RNA (protease-resistant) interactions shows that Mrd1p influences snoRNA-pre-rRNP interactions in different ways. We therefore propose that Mrd1p is involved in structural rearrangements of the pre-rRNP that affect resolution of snoRNA-pre-rRNA structures.
Both deletion of RBD5 and replacement of the two aromatic residues in RNP1 of RBD5 inactivated the function of Mrd1p. Most strikingly, deletion of RBD5, but not the mutRBD5 mutation, impaired U3 snoRNA release from base pairing with pre-rRNA. This implies that RBD5 is involved in two essential functions in the pre-rRNP. One contributes to release of U3 snoRNA, and the other requires the aromatic residues in the RNP1 motif, suggesting involvement of an RNA contact.
It is unlikely that the mutRBD5 mutation prevents the initial formation of U3 snoRNA-pre-rRNA duplexes, since U3 snoRNA associates with pre-rRNA (47) and base pairs with pre-rRNA (Fig. (Fig.5B)5B) independently of the presence of Mrd1p. Both mutRBD5 and ΔRBD5 also supported association of U3 snoRNA with preribosomal complexes containing 35S pre-rRNA (Fig. (Fig.4D4D and and5A).5A). Furthermore, both mutations in Mrd1p allowed formation of SSU processomes (Fig. (Fig.3),3), which implies that association of U3 snoRNA and initial pre-rRNP assembly were not severely affected in these mutants.
U3 snoRNA base pairs with 35S pre-rRNA in the 5′ETS and at the 5′ end of 18S rRNA (2, 3, 49). The latter will form the 5′ side of the central pseudoknot that is conserved in SSU RNA (21). Potentially, U3 snoRNA can also base pair with the 3′ side of the pseudoknot located approximately 1,140 bases into the 18S rRNA in S. cerevisiae, but this interaction has not been verified (49). Although it remains unclear how the 3′ site is moved into position for pseudoknot formation, U3 snoRNA is believed to be important for directing and timing the formation of the pseudoknot (25).
Cotranscriptional assembly of the pre-rRNA containing the 5′ ETS and the 18S rRNA into pre-rRNP entails several compactions as visualized by EM (40). First the nascent pre-RNP folds into a small terminal knob, believed to include U3 snoRNA (38), followed by a compaction into a loose SSU processome. Before cotranscriptional cleavage at A2, the SSU processome tightens into a more compact structure. This compaction of the SSU processome may represent structural rearrangements that are connected to formation of the central pseudoknot (40). If so, pseudoknot formation may not take place in the presence of the ΔRBD5 and mutRBD5 Mrd1ps, since these mutations were not compatible with the compaction of the SSU processome. Depletion of Esf2p also reduces SSU compaction (23). Esf2p is a cofactor of Dbp8p (20), indicating that this RNA helicase could have a function at this stage.
The general outline of formation of the SSU RNA pseudoknot during pre-rRNA processing seems to be highly conserved. In archaea and eubacteria, the initial base pairing to the 5′ part of 16S rRNA is believed to be performed in cis by U3 box A-like sequences located in the 5′ETS of pre-rRNA (9, 11, 54). Since the U3-like sequences are also involved in forming the 16S processing stem, central pseudoknot formation is likely cotranscriptional and happens before RNase III cleavage that produces the 16S rRNA (11).
If central pseudoknot formation precedes the U3 snoRNA-dependent cleavages at A2 also in eukaryotes, this would imply that the base pairing of U3 snoRNA to the 5′ part of the 18S rRNA could be resolved well before U3 snoRNP dissociates from the SSU processome. This is compatible with the observation that association of U3 snoRNA with pre-rRNP is not dependent on base-pairing sequences (19). Such a scenario would also fit with our finding that the mutRBD5 mutant still retained almost all U3 snoRNA in pre-rRNP although the U3 snoRNA was released from base pairing. This interaction between U3 snoRNP and the pre-rRNP would then depend on other contacts, such as protein-protein interactions and/or weak RNA-RNA interactions.
The RNP1 motif of RBD5 is thus not needed to release U3 snoRNA from base pairing. It is therefore possibly involved in establishing contacts necessary for structural rearrangements leading to pseudoknot formation and/or A2 cleavage. Mrd1p could thus be involved both in U3 snoRNA release from the 5′ site of the pseudoknot and in forming structures that facilitate pseudoknot formation. An attractive consequence of such a model is that two steps in pseudoknot formation would rely on a single module (RBD5 of Mrd1p). This would ensure a highly efficient coupling. However, interference with the sequences involved in pseudoknot formation does not necessarily inhibit pre-18S cleavages (49). This makes it both possible and likely that the function of Mrd1p extends beyond this point and includes accommodating an SSU processome conformation compatible with coupling the A0 to A2 cleavages.
We thank Kerstin Bernholm and Magnus Wieslander for excellent technical assistance.
This work was supported by grants from Carl Tryggers Stiftelse and the Swedish Research Foundation and by NSF grants MCB-0448171 and MCB-0818818 to A.L.B.
Published ahead of print on 24 August 2009.
†Supplemental material for this article may be found at http://mcb.asm.org/.