|Home | About | Journals | Submit | Contact Us | Français|
In eukaryotes, the 40 S ribosomal subunit serves as the platform of initiation factor assembly, to place itself precisely on the AUG start codon. Structural arrangement of the 18 S rRNA determines the overall shape of the 40 S subunit. Here, we present genetic evaluation of yeast 18 S rRNA function using 10 point mutations altering the polysome profile. All the mutants reduce the abundance of the mutant 40 S, making it limiting for translation initiation. Two of the isolated mutations, G875A, altering the core of the platform domain that binds eIF1 and eIF2, and A1193U, changing the h31 loop located below the P-site tRNAiMet, show phenotypes indicating defective regulation of AUG selection. Evidence is provided that these mutations reduce the interaction with the components of the preinitiation complex, thereby inhibiting its function at different steps. These results indicate that the 18 S rRNA mutations impair the integrity of scanning-competent preinitiation complex, thereby altering the 40 S subunit response to stringent AUG selection. Interestingly, nine of the mutations alter the body/platform domains of 18 S rRNA, potentially affecting the bridges to the 60 S subunit, but they do not change the level of 18 S rRNA intermediates. Based on these results, we also discuss the mechanism of the selective degradation of the mutant 40 S subunits.
The translation reaction, or mRNA-dependent protein synthesis, is catalyzed by the ribosome, a large ribonucleoprotein complex, with assistance from translation factor proteins. During the initiation phase, the ribosome dissociates into the large (50 S in Bacteria and 60 S in Eukarya) and small (30 S in Bacteria and 40 S in Eukarya) subunits, and the latter binds the methionyl initiator tRNA (Met-tRNAiMet; its amino group is formylated in Bacteria) and mRNA with the help of initiation factors (initiation factors for Bacteria and eIF for eukaryotic initiation factors) (reviewed in Refs. 1,–3). As the consequence, the anticodon of the tRNAiMet is base-paired to the start codon of the mRNA at the peptidyl-tRNA-binding site (P-site) of the small ribosomal subunit. The universally conserved factors, eIF1A/IF1 and eIF5B/IF2, are proposed to bind the small subunit at conserved locations and play distinct yet similar roles between Bacteria and Eukarya (4, 5).
The fidelity of start codon recognition is regulated by eIF1A/IF1 that binds the small subunit A-site (6, 7) as well as by other initiation factor(s) specific to each domain of life, IF3 in Bacteria (2) and eIF1, eIF2, eIF3, and eIF5, in Eukarya (eukaryotes). eIF2 is a heterotrimeric factor that binds Met-tRNAiMet depending on a bound GTP. The resulting ternary complex (TC)3 binds the 40 S subunit in the context of multifactor complex (MFC) with eIF1, -3, and -5, forming a 43 S preinitiation complex (PIC) (8). eIF1A also promotes TC loading onto the 40 S subunit (7). Start codon selection by PIC requires the eIF5-catalyzed GTP hydrolysis for eIF2 and is executed by a mechanism that involves the release of eIF1 and the resulting Pi, in response to AUG recognition (9). Subsequently, conformational changes in eIF2 allow it to leave the PIC together with eIF5. The clearance of at least eIF2 and eIF5 appears to be the requisite for subsequent 60 S subunit joining mediated by eIF5B (1, 3). In Archaea, aIF1 is a fidelity factor similar to eIF1 (10), in support of the phylogenetic relationship between Archaea and Eukarya, as proposed by Woese et al. (11).
X-ray crystallography of bacterial and archaeal ribosomes indicates that three-dimensional arrangement of rRNA determines the ribosome structure (reviewed in Ref. 12). Bacterial 16 S rRNA, as well as its eukaryotic counterpart, 18 S rRNA, contains 5′, central, 3′ major, and 3′ minor domains, which make up the body, platform, head, and a major projection from the body structure, respectively, of the small ribosomal subunit (see Fig. 1A for the secondary structure of yeast 18 S rRNA) (13). The larger mass of the eukaryotic 40 S subunit compared with the bacterial and archaeal 30 S subunit is partly due to several major expansion segments of 18 S rRNA inserted into the core rRNA sequence conserved between 16 S and 18 S rRNA (boxes in Fig. 1A) (14). Three-dimensional structure of the yeast and mammalian ribosomes have been predicted by fitting these additional segments to the portions of 40 S subunit cryo-EM structure, which could not be filled with the structure predicted with the bacterial 30 S subunit structure (15,–18). Hydroxyl radical cleavage, cryo-EM, and other studies elucidated approximate locations of eIF1, eIF1A, eIF2α/γ, eIF3, eIF4G (the major adaptor subunit of eIF4F), and eIF5 on the 40 S subunit (reviewed in Ref. 19). In particular, eIF1 was linked to helix (h)23 and h24 of the central domain and h44 of the 3′ minor domain by hydroxyl radical cleavage (20). Cryo-EM studies support this location close to the E-site (6).
Within the yeast ribosomal DNA (RDN) repeat, the RNA polymerase I drives transcription of 35 S rRNA, the precursor for 18 S, 5.8 S, and 25 S rRNA, whereas RNA polymerase III drives 5 S rRNA transcription (supplemental Fig. S1) (reviewed in Ref. 21). In the nucleolus, 35 S rRNA is processed, modified, and assembled with ribosomal proteins, producing the 90 S preribosome, which is cleaved into large and small presubunits. They are then separately transported into the cytoplasm while maturing into the final products, the 40 S and 60 S subunits, containing 18 S rRNA and 25 S, 5.8 S, and 5 S rRNAs, respectively (21). The ribosomes misprocessed or those containing mutations are degraded (21). Different “nonfunctional” rRNA decay pathways that degrade mutant rRNA altering bases important in bacterial rRNAs have been identified (22,–24).
The role of yeast 18 S rRNA in elongation and termination functions, as well as antibiotic resistance, has been extensively studied by site-directed mutagenesis of universally conserved residues important in bacterial 16 S rRNA (25,–28), taking advantage of strains deleted for all ~100 RDN repeats but carrying a high copy (hc) RDN plasmid (29, 30). Recently, Dong et al. (31) identified by random mutagenesis that 18 S rRNA nucleotides in the h28 region are important for initiator tRNA binding and AUG selection. Here, we report the characterization of 10 18 S rRNA mutations obtained by screening for slow growth (Slg−) or temperature-sensitive (Ts−) mutants that alter the polyribosome (polysome) profile, indicative of changes in translation status in the cell. Our results extend the earlier report (31) and provide further evidence that the 18 S rRNA domains making up the decoding site and the eIF platform are directly involved in translation initiation, regulating the stringent selection of start codons.
Plasmids and yeast strains carrying yeast rdn mutants are listed in Table 1. Other plasmids used include YCpU-SUI3 (SUI3 URA3) and YCpU-SUI3-2 (SUI3-2 URA3) (32) and modified GCN4-lacZ reporter plasmids pM199, pM226, and p227 (33, 34). pKA624 (his4-306Δ URA3) was generated by subcloning into pRS306 a DNA segment derived from PCR using a pair of HIS4 oligos and p391 (34) as template, as described for a similar his4Δ URA3 integration plasmid (32).
To prepare pNOY373 derivatives carrying rdn-41a and rdn-59a (Table 1, 6th column), we subcloned 2940-bp PstI-DraIII fragment of pNOY373-41 and -59, respectively, into the same sites of pNOY373. We confirmed by sequencing that the rdn-41b and rdn59b mutations found in 25 S rRNA-coding region of the originally isolated plasmids were eliminated in the resulting plasmids.
KAY488 is a remake of NOY890 (30), derived from NOY908 (30) transformant carrying pRDN-hyg (RDNhyg URA3). Because NOY890 was not available, we predominantly used KAY488 for plasmid shuffling in this study. Thus, the strain is referred to as NOY890 (KAY488) throughout the text. KAY165 (Table 1) was constructed by introducing his4-306(TTG) allele by transformation of NOY908 with an XhoI-linearized pKA624, followed by homologous recombination-mediated removal of URA3 by selecting against Ura+. Details in our method of introduction of HIS4 initiation codon mutations was described elsewhere (32). KAY171 was generated from a KAY165 transformant carrying pRDN-hyg. A Leu− Ura+ progeny derived from the transformant was selected from its culture grown in the presence of leucine, designated KAY171, and used for plasmid shuffling.
We mutagenized a pool of the hc LEU2 plasmid pNOY373 encoding a wild-type rDNA repeat (supplemental Fig. S1) using a mutator Escherichia coli strain XL1-Red, and we used it to transform NOY890 (KAY488). We used the drug 5-fluoroorotic acid (FOA) as the reagent to select against the URA3 plasmid (35); thus, the growth of the double plasmid transformants will produce only yeast derivatives bearing hc LEU2 rDNA plasmid, unless the latter carries a lethal rDNA allele. The frequency of formation of FOA resistant cells from NOY890 (KAY488) transformants with a hcLEU2 rDNA plasmid was lower at the regular temperature of 30 °C than in a typical plasmid shuffling done with commonly used yeast strains deleted for a protein factor, and even lower at a higher temperature of 36 °C. This prevented us from screening for Ts− mutants simply by comparing the cell growth on FOA medium at 30 and 36 °C, as typically done to obtain protein factor mutations (36). Instead, we re-streaked FOA-resistant cells that had been grown at 30 °C onto two rich YPD plates and compared the growth between the two temperatures.
pNOY373 derivatives isolated from mutant candidates, as described in the text and Table 1, were sequenced using oligodeoxyribonucleotides listed in supplemental Table S1 and supplemental Fig. S1. Prior to determining the mutant sequences, we determined the sequence of the 9150-bp PstI-BamHI RDN insert of wild-type pNOY373 (GenBankTM accession number DQ888227), using the same set of oligos (see the structure of the RDN locus in supplemental Fig. S1). The chromatogram files of the sequences generated from the mutant plasmids were then analyzed using Mutation Surveyor® version 3.0 (Softgenetics LLC). This software allowed us to compare up to 24 different mutant chromatogram patterns at once, thereby unambiguously identifying mutation sites, including base deletion or insertion within a long stretch of the same nucleotide. The supplemental Table S2 indicates that all the 18 S rRNA mutants listed contained other mutations besides base changes within 18 S rRNA. To evaluate these non-18 S mutations, we performed systematic BLAST searches using different RDN segment sequences as query, as described below.
To examine evolutionary conservation of each of the Saccharomyces cerevisiae rRNA residues, we utilized a small or large subunit RNA variability map available on line (37, 38). These maps categorize every residue of SSU and large subunit rRNA according to the level of conservation throughout eukaryotes, and the map for S. cerevisiae SSU (18 S) rRNA was used in supplemental Fig. S2.
To examine polymorphisms within the S. cerevisiae RDN locus, we undertook systematic BLAST searches with each of the nine segments derived from the 9.15-kb pNOY373 insert sequence as listed across the top of supplemental Table S2. We wrote Perl scripts to parse BLAST hits that exceed a predefined cutoff score as follows: (i) a specified length in bp (longer than 300 bp), (ii) a specified probability score (less than 1 e−1), and (iii) a specified identity (larger than 95%). rDNA sequences in the resulting hits were processed for multiple sequence alignment to identify polymorphisms among different S. cerevisiae strains/isolates or closely related Saccharomyces species. As a result, we identified 17 and 57 naturally occurring base changes (polymorphisms) within the RDN segment upstream of 18 S rRNA (supplemental Table S3) and the 18 S-coding region (supplemental Table S4), respectively (Polymorphisms found in the remaining RDN regions will be reported elsewhere.) The supplemental Table S5 summarizes variations at major T or A stretches found throughout the entire RDN locus during the searches. These results indicated that the remaining mutations found outside of 18 S rRNA (supplemental Table S2) are either a common polymorphism found at the site of consecutive A or T residues (supplemental Table S5) or a base change least likely to affect the production of the ribosome (see supplemental text for details).
Polysome profiling was performed by sucrose gradient velocity sedimentation of extracts of yeast treated with cycloheximide, as described previously (34). The mass ratio of polysome to monosome (P/M ratio) was measured from the A260 profile of the gradient samples. However, due to different abundance of the 60 S subunit, the P/M ratio here was calculated as the ratio of polysome abundance to the sum of the abundances of free 40 S, free 60 S, and the monosome. The P/M ratio was converted to the fraction of ribosomes found in polysome fractions by the formula 1/(1 + (1/a)), where a is the P/M ratio.
To determine the mass ratio of 60 S to 40 S subunit (60 S/40 S ratio) independent of polysome profiling, yeast cell extracts were prepared in the absence of free Mg2+ and resolved by sucrose gradient velocity sedimentation followed by monitoring subunit abundance at A260, as described previously (39). The 60 S/40 S ratio was converted to the fraction of the excessive 60 S subunit abundance compared with the mass of 60 S and 40 S subunits, which can form a stoichiometric 80 S couple (used in Fig. 3A). Assuming that the wild-type cells contain 1:1 stoichiometry of these subunits, this value was calculated as (b – a)/(1 + b), where a and b are the 60 S/40 S mass ratios for the wild-type (= 2.16) and mutant strains, respectively. The 60 S/40 S ratio was also converted to the percentage of 40 S subunit reduced by each mutation compared with the 40 S subunit abundance in wild-type cells (used in Fig. 3C). This was calculated as (1 – a/b) × 100, where a and b are the 60 S/40 S ratio for wild-type and mutant, respectively, assuming that the 60 S subunit abundance was unaltered.
To determine the RNA sequence of the ribosomes, rRNA was isolated by a hot phenol method (40) and used to perform reverse-transcriptase (RT)-directed PCR after DNase I treatment (DNase I removes any contaminating plasmid DNA). For RT-PCR, we used Access QuickTM RT-PCR system (Promega) and a pair of oligos, as listed in supplemental Table S1 and supplemental Fig. S1. The resulting cDNA product was used for sequencing.
To study the abundance of different PIC intermediates, cycloheximide-treated yeast culture was fixed with formaldehyde for preparing whole cell extracts, which are then resolved by sucrose gradient velocity sedimentation, as described previously (41). One-third of the gradient samples were precipitated with ethanol and analyzed with SDS-PAGE followed by immunoblotting with rabbit polyclonal anti-SUI2 (eIF2α) (42), -TIF5 (eIF5) (43), -TIF35 (eIF3g) (44), -SUI1 (eIF1) (45), -TIF11 (eIF1A) (46), -Rps0,4 and mouse monoclonal anti-TIF32 (eIF3a) (47) antibodies.
[3H]Met-tRNAiMet binding assay was performed in the presence of a nonhydrolyzable GTP analogue, GMPPNP, as described previously (48). 40 S subunit purification was performed also as described previously (48). To perform GST pulldown assays with the purified 40 S subunit, we purified GST-eIF1 and -eIF1A fusion proteins from E. coli BL21 derivatives bearing pGEX-SUI1 (36) and p3415 (46), respectively, and performed GST pulldown assays, as described with minor modifications (49). Briefly, the glutathione resin adsorbed with 5 μg of GST fusion proteins was blocked by preincubating with 100 μg/ml preheated total yeast tRNA (Sigma), and after washing the resin once with the binding buffer containing 100 mm KCl, the blocked resin-protein complex was incubated with two A260 units of purified 40 S subunit in the same binding buffer but with dry milk at 20 °C. Then the resin was washed three times with the binding buffer without dry milk and analyzed by SDS-PAGE, followed by immunoblotting with anti-RpsO antibodies.
Standard molecular biology methods, including the β-galactosidase assay, were used throughout (34). To determine yeast titer on different agar plates, overnight culture was diluted to A600 = 0.15, and 5 μl of this and 10-fold serial dilutions were spotted, incubated under specified conditions, and photographed (34). The method of Northern blotting using PCR-generated 32P-probe was also described (40). The 32P-probe for 20 S, 23 S, and 35 S rRNA was generated by PCR with oligos 20 S-F and 20 S-R (supplemental Fig. S1 and supplemental Table S1), which were designed based on oligo probes previously used to detect 20 S rRNA with modifications (50, 51).
To understand the structure-function relationship of the eukaryotic ribosome in translation initiation, we attempted to isolate temperature-sensitive (Ts−) or slow growth (Slg−) mutations in yeast rRNA transcribed from an hc LEU2 plasmid by random mutagenesis. We used the strain NOY890 (KAY488) deleted for all the chromosomal rDNA repeats (rdnΔΔ) and instead carrying an hc URA3 plasmid with a copy of RDN as the sole source of rRNA (30). Of >4000 clones, we obtained ~80 mutant candidates that showed Ts− or Slg− phenotypes after evicting the URA3 RDN helper plasmid using 5-fluoroorotic acid (FOA) as the drug to select against URA3 (see “Materials and Methods”). We purified rDNA mutant plasmids from these candidates and reintroduced them to fresh NOY890 (KAY488), and the resulting transformants were subjected for growth on FOA-containing plates. ~30 FOA-resistant clones reproduced the growth phenotypes, suggesting that their phenotypes are due to mutations in the hc rDNA LEU2 plasmid.
To find out if the isolated mutations compromise protein synthesis, we examined the polysome profile of all the mutants, grown at permissive (30 °C) and restrictive (36 °C) temperatures. Active translation of mRNA by multiple ribosomes results in formation of polysomes detected by sucrose gradient velocity sedimentation (34). 16 of the tested mutants substantially altered polysome profiles in two different ways. Type I, including nine mutants, initially designated rdn-41 to rdn-76 based on clone numbers, showed a strong accumulation of free 60 S subunit, with reduced polysome contents (see Fig. 2). Type II, including seven mutants, displayed a distinct polysome profile, with the decrease in free 60 S subunits and the appearance of half-mers (polysomes containing 40 S subunits awaiting 60 S subunit joining), which will be reported elsewhere.5 Sequencing of the entire rDNA repeat from the 16 mutants (see “Materials and Methods”) indicated that the type I mutations altered an 18 S rRNA base (supplemental Table S2), whereas the type II mutations changed a 25 S rRNA base. Because rdn-41 and rdn-59 also contained a mutation in 25 S rRNA (supplemental Table S2), we created yeast mutants just containing the 18 S rRNA mutations (called rdn-41a and rdn-59a, respectively) (Table 1) and confirmed that their polysome profile and growth defects are almost identical to those of the parental strains (supplemental Fig. S3, A and B, also see below). We also found that rdn-63, showing a weak Slg− phenotype at 30 °C, had an insertion mutation in 18 S rRNA (Tables 1 and supplemental Table S1). Here, we report the characterization of the 10 mutants with altered 18 S rRNA (Table 1).
Most of the 18 S rRNA mutations (rdn-41, -42, -44, -61, -71, and -76) altered an evolutionarily conserved 18 S rRNA base (category 0 or 1, the highest or second highest degree of conservation, respectively, according to the 7-scale SSU RNA variability map), although rdn-59, -60, and -72 altered a less conserved base (categories 3, 2, and 5, respectively) (red arrows in supplemental Fig. S2). In addition, we confirmed that the identified 18 S rdn mutations are different from naturally occurring base changes (polymorphisms) found in different strains or isolates of S. cerevisiae or its closely related species (black arrows in supplemental Fig. S2 and supplemental Table S4).
To verify that the 18 S rRNA mutant yeast express the ribosomes with the corresponding mutations, the RNA samples isolated from the mutant yeast were treated with DNase I, to eliminate any contaminating plasmid DNAs, and then used for RT-PCR (see ”Materials and Methods”). After confirming that the reaction generates cDNA in the presence of RT (supplemental Fig. S4A, lanes labeled +) but not in its absence (lanes labeled −), we determined the cDNA sequence around the mutation site. As shown in supplemental Fig. S4B, all of the cDNAs generated from the mutant 18 S rRNA contained the mutated nucleotide, confirming that the detectable majority of the ribosomes found in the rdn mutants is the expected mutant ribosome.
Because the analyses as described above indicated that the single 18 S rRNA base change identified was the cause of the phenotype related to protein synthesis, hereafter, we designate the individual 18 S rdn alleles with their base changes in 18 S rRNA. To relate this study to Escherichia coli ribosome, we also add the corresponding E. coli 16 S rRNA base number in parentheses. Thus, rdn-42 is designated A1193U (A968), etc. For rdn-41 and rdn-59 with additional mutations in 25 S rRNA, we designate the originally isolated alleles as A952C* (G742) and A333G* (A262), respectively, and we distinguish them from A952C (G742) and A333G (A262), which were created as rdn-41a and rdn-59a, respectively, as mentioned above.
Comparative genomics predicts that yeast 18 S rRNA consists of four domains as follows: 5′, central, 3′ major, and 3′ minor domains (Fig. 1A). Examination of the sites of the mutations within the proposed secondary structure of yeast 18 S rRNA indicated that all the mutations except A1193U (A968) changed a residue in the 5′ or central domain of 18 S rRNA. Specifically, A138G (A143), G322A (G251), A333G (A262), and G362A (C290) altered the 5′ domain, whereas G875A (A663), A952C (G742), G922A (G711), G1108A (G885), and ΩG1112 (G888) mapped in the central domain (Fig. 1, A and B, shows bases corresponding to those in E. coli sequence). Cryo-EM structure of eukaryotic ribosomes (15,–17) suggests that five of them directly disrupt a bridge to the 60 S subunit (Table 1, 5th column) and that other rdn mutations, as well as A333G (A262) (Fig. 1A), affect the 60 S subunit interface indirectly in part by disrupting A-minor interactions (blue and red boxes and lines in Fig. 1A) predicted from bacterial ribosome structure at atomic resolutions (see “Discussion”).6 These speculations provide a potential basis for phenotypes observed with the isolated mutations, as described below, but need to be justified in the future by x-ray structure of eukaryotic ribosomes at an atomic resolution.
A1193U (A968) is the only mutation found here to alter the 3′ major domain. The altered base in h31 loop locates directly below the codon-anticodon helix at the P-site (Fig. 1C). Mutations directly affecting yeast tRNAiMet binding to the 40 S subunit have been mapped in a different helix, h28, of the 3′ major domain (31). Therefore, in addition to h28, the h31 loop appears to constitute tRNAiMet-binding site. This idea is further tested, as shown below.
As shown in Fig. 2, the 18 S rdn mutants show strong 60 S subunit accumulation in their polysome profiles. Agarose gel electrophoresis and ethidium bromide staining confirmed that the accumulated mass at the 60 S position of every mutant profile indeed contains 25 S rRNA with the correct size (data not shown). Thus, it was conceivable that a significant proportion of the mutant 40 S subunit was depleted by defective biogenesis or stability in each of the 18 S rRNA mutant strains. To confirm the depletion of 40 S subunits relative to 60 S subunits, we prepared the ribosome fractions in the absence of free Mg2+ and measured the relative abundance of 40 S and 60 S subunits by sucrose gradient analysis, independent of polysome profiling (see “Materials and Methods”). As shown in the insets of Fig. 2, panel 1, the 60 S/40 S ratio for wild-type, which would correspond to 1:1 stoichiometry of the subunits, was 2.16, comparable with previously reported values (52). The insets of other panels in Fig. 2 show that all the 18 S rRNA mutants contained a higher ratio of the 60 S subunit mass against the 40 S mass than the ratio for wild type (see supplemental Fig. S3A insets for reconstructed alleles, A952C and A333G). rdn-63 or ΩG1112 (G888) barely increased the 60 S/40 S ratio at 36 °C (Fig. 2, panel 8), qualifying this mutation as the weakest of “type I” polysome defect. These results confirm the depletion of the 40 S subunit in our Ts− or Slg− mutants.
To examine whether the observed reduction in 40 S subunit abundance can quantitatively account for the strong 60 S subunit accumulation in the 18 S rRNA mutant polysome profiles, we calculated the percentage of free 60 S subunit found in the polysome profiles (examples shown in Fig. 2) compared with the total ribosome mass in the profiles (gray bars in Fig. 3A, panel 1). As shown in Fig. 3A, these values correlate significantly (r = 0.941, p = 2.16 × e−10, see panel 2) with the percentage of excessive 60 S subunit relative to the mass of the 40 S and 60 S subunits that can form a single ribosome (filled bars in panel 1), values obtained from the independent measurement of 60 S/40 S ratio. Thus, the accumulation of free 60 S subunit observed in the mutant polysome profiles is largely due to the depletion of mutant 40 S subunits.
The polysome profiles in Fig. 2 also indicated that the polysome contents of most of the 18 S rRNA mutants, as measured by the ratio of polysome-to-monosome abundance (P/M), were lower than that of the wild-type RDN strain at permissive (30 °C) and restrictive (36 °C) temperatures, suggesting defects in translation initiation. The plot of the fraction of ribosomes found in polysome fractions (calculated from the P/M ratio) against the growth rate (times doubling/h) of each mutant indicates a linear correlation for the dataset measured at 30 °C (Fig. 3B, panel 1; r = 0.906, p = 0.0003), confirming that the decrease in cellular translation initiation resulted in a corresponding decrease in growth rate. We noted, however, that the growth at 36 °C disrupted this correlation (Fig. 3B, panel 2; r = 0.53, p = 0.11). We do not know the reason for this, but among many other possibilities, this may be because ribosomes defective in the elongation phase slow down the migration of translating ribosomes, thereby increasing the polysome abundance (34, 53).
We also found that the extent of 40 S subunit depletion, as calculated from 60 S/40 S abundance ratio, significantly correlated with decrease in the growth rate of each mutant at 30 °C (Fig. 3C; r = 0.876, p = 0.0019) if we eliminate the data from the A1193U (A968) mutant showing the most severe functional defect (see below). Thus, the 40 S subunit depletion by 18 S rRNA mutations directly resulted in a decrease in translation initiation rate. Our results also indicate that the 40 S subunit is a limiting component of translation initiation, as observed for essential yeast initiation factors (eIF1A, eIF4G1, and eIF4E) by their quantitative depletion studies (54).
In an effort to understand the mechanism whereby the 18 S rdn mutations reduce 40 S subunit abundance and translation initiation rate, we first examined the processing of 18 S rRNA in yeast strains bearing these mutations. During 18 S rRNA biogenesis, RNA polymerase I transcribes 35 S rRNA, whose 5′ external transcribed sequence is removed to generate the 32 S species by cleavages A0 and A1. 32 S rRNA is then cleaved at A2 to generate 20 S rRNA. Finally, 20 S rRNA is transported to the cytoplasm, and cleavage D eliminates its excess 3′ terminal sequence (supplemental Fig. S5A) (21). Cleavage A3 not only is the first cleavage leading to the maturation of 5.8 S and 25 S rRNAs, but it also generates an aberrant 23 S species from the 5′-third of 35 S rRNA, which escaped A0, A1, and A2 cleavages (50). Northern blot analyses using the probe specific to the region 3′ of 18 S (generated by PCR using oligos 20 S-F and 20 S-R in supplemental Table S1) indicated that no detectable 18 S biogenesis intermediates, such as 35 S, 20 S, or 23 S species, abnormally accumulated in the 18 S rRNA mutants (supplemental Fig. S5B, panel 2) except for A1193U (A968) with a buildup of 20 S pre-rRNA (supplemental Fig. S5B, lanes 3 and 7; see panel 3 for quantification). We loaded RNA samples isolated from the same A600 units of the exponentially growing yeast culture. Therefore, the increase in 20 S rRNA observed by A1193U (A968) is most likely due to the increase in its steady-state abundance, although detailed analyses on its production rate are necessary to conclude this. The 20 S species increased by this mutation was not detected by EtBr staining (supplemental Fig. S5B, panel 1, lanes 3 and 7), and therefore its effect on ribosome function would be minimal. These results suggest that the processing of 18 S rRNA is normal in the rdn mutants except A1193U (A968) and that the reduced 40 S subunit abundance in these mutants are due to degradation of the final product, the 40 S subunit (also see “Discussion”).
To study the steady-state abundance of different initiation complex intermediates in rdn mutants, we fractionated formaldehyde-treated yeast extracts by sucrose gradient velocity sedimentation, and analyzed the gradient samples by immunoblotting for the PIC components, eIF1A, -1, -3, and -5, and the Rps0 subunit of the 40 S subunit (see “Materials and Methods”). We examined four strongest Ts−/Slg− mutants, A333G (A262), G875A (A663), A952C (G742), and A1193U (A968), together with the RDN+ control strain. As shown in Fig. 4A, free 40 S subunits found in the RDN+ cells are loaded with up to ~5% of eIF2α, -3a, -3g, and -5 (see Fig. 4F for % of eIF2α found in free 40 S fraction). Given that these factors occur substoichiometrically to the 40 S subunit (molar ratio of 1 (eIF2):1 (eIF5):0.5 (eIF3):3 (40 S)) and that ~10% of 40 S subunit occur free of the 80 S ribosome (55), the majority (>75%) of the 40 S mass detected by A254 and anti-RpsO antibodies (Fig. 4A, top and bottom) is considered as 40 S subunits free of these eIFs, in agreement with previous observations (55). Fig. 4A also identified eIF complexes free of the 40 S subunits, including MFC (eIF1/2/3/5/tRNAiMet; lanes 3–5) (8, 56) and the eIF2/eIF5 complex free of Met-tRNAiMet (lanes 1–3) (57).
Strikingly, A1193U (A968) impaired the MFC interaction with the 40 S subunit (Fig. 4C, lanes 7 and 8), allowing the former to accumulate at a significant level (lanes 5 and 6).7 This agrees with the idea that A1193 is directly involved in tRNAiMet binding. In contrast, other mutations increased the fraction of eIF2α bound to the 40 S subunit, even though they decreased the abundance of free 40 S subunit detected by A254 (Fig. 4, B, D, and E; see F for quantification of bound eIF2α). In the case of A952C (G742) altering h22, the majority of the 40 S subunit mass detected by A254 appears to be loaded with full PIC components (Fig. 4B, lanes 8–10), with ~5 times more eIF2α loaded onto the 40 S subunit than in wild type (Fig. 4F, column 2). The observed accumulation of the 43 S/48 S PIC suggests a defect in step(s) after 48 S complex assembly, including scanning or 60 S subunit joining (see below for its phenotype indicating defective initiation). In theory, the accumulation can be explained by the accelerated PIC assembly, but the Slg− phenotype of the mutant may exclude this possibility.
In the case of G875A (A663) also altering h22, most of the free 40 S subunit seems to be bound with eIF2 only (partial 43 S complex) (Fig. 4D, lanes 7 and 8). The free 40 S subunit found in A333G (A262) appears to be the mixture of the partial 43 S complex and the fully loaded PIC, because the relative abundance of eIF2α on the 40 S subunit seems to be greater than that of other PIC components (Fig. 4E, lanes 8 and 9).7 Although the mechanism whereby the partial 43 S complex is formed requires further examination (see “Discussion”), the reduced amount of eIF1, -3, and -5 in the 40 S fractions suggests that G875A (A663) and A333G (A262) impair the interaction of MFC components (eIF1, -3, and -5) with the 40 S subunit. These results indicate that the amount of free 40 S subunit devoid of eIFs is greatly reduced in all the four rdn mutants examined, consistent with the limitation of free 40 S subunit available for initiation. Moreover, they suggest that the individual rdn mutations impair different steps of PIC formation or function.
To identify defects in specific steps of translation initiation in vivo, we took advantage of yeast general amino acid control pathway, which depends on the Gcn4p transcription factor (58). Translational control of GCN4 in response to amino acid availability depends on upstream ORFs (uORFs) in the leader region (schematics at top of Fig. 5B). When amino acids are plentiful, special cis-elements around uORF1 allow the terminating ribosome to re-initiate translation of one of the downstream uORFs, thereby repressing translation of GCN4. When amino acids are limited, Gcn2 eIF2 kinase is activated to phosphorylate eIF2, thereby inhibiting eIF2B-catalyzed formation of active eIF2-GTP complex. This prevents a timely formation of the PIC prior to translation of the inhibitory uORFs, allowing translation of GCN4. Because of this mechanism of GCN4 translational control, general amino acid control expression is, in some cases, a sensitive indicator of defects in translation initiation activities (59). For example, if a translation initiation mutation causes faulty recognition of the uORF1 start codon, this leads to failure to express GCN4 under the amino acid starvation conditions. As a consequence, the mutant yeast cannot overcome growth inhibition caused by the starvation (general control nonderepressible or Gcn− phenotype).
The drug 3-aminotriazole (3AT) is an inhibitor of His3p enzyme, causing histidine starvation. The growth of yeast in the presence of 3AT requires general amino acid control expression, and therefore, 3AT sensitivity conferred by a genetic mutation signifies a Gcn− phenotype. As shown in Fig. 5A, panel 3, all the 18 S rRNA mutants are more or less sensitive to 50 mm 3AT at a semi-restrictive temperature of 34 °C, suggesting that they are Gcn− (Table 1, column 9). Four of the strongest 18 S rRNA mutants, A333G* (A262), G875A (A663), A952C* (G742), and A1193U (A968) were even sensitive to a lower concentration of 3AT (30 mm) (Fig. 5A, panel 2, rows 2–5), suggesting that their Gcn− phenotypes are stronger (also see Fig. 5B, below).8
To investigate the mechanism underlying the observed Gcn− phenotype, we measured β-galactosidase activities from GCN4-lacZ plasmid, pM226, in a select set of the 18 S rRNA mutants. The GCN4-lacZ allele in this plasmid is preceded by a modified uORF1 whose stop codon locates within GCN4 reading frame (schematics in Fig. 5B, panel 1, top), which in turn strongly prevents translation of GCN4-lacZ. Thus, any increase in GCN4-lacZ expression compared with the value obtained with wild-type strain reflects the increased frequency of the bypass or “leaky scanning” of the uORF1 start codon. We found that two of the Gcn− mutations, A1193U (A968) and G875A (A663), strongly, and A952C* (G742) and A333G* (A262) weakly but significantly, increased GCN4-lacZ expression from pM226 (p < 0.01, n = 8) at the semi-restrictive temperature of 34 °C (Fig. 5B, panel 1). As a control, GCN4-lacZ expression with no uORF preceding GCN4 (p227) was increased only modestly (<1.5-fold) by the rdn mutations tested (Fig. 5B, panel 2), confirming that changes observed with pM226 were at the translational level. Thus, the Gcn− phenotypes of the rdn mutants likely occurred at least partially due to bypassing uORF1.
To investigate more specific defects in stringent AUG selection in vivo, we used the his4-306 allele with the AUG-to-UUG initiation codon mutation as a phenotypic reporter. SUI3-2 mutation altering eIF2β reduces eIF2 interaction with tRNAiMet, thereby increasing the spontaneous release of eIF2 from the scanning ribosome (43). This increases the chance of initiating translation from the noncanonical UUG start codon, to suppress His− phenotype of his4-306UUG allele (suppressor of initiation codon mutation or Sui− phenotype) (Fig. 5C, rows 1 and 2). As shown in Fig. 5C, rows 5–8, A1193U (A968) and G875A (A663) suppressed the ability of SUI3-2 to express His4p from his4-306 at 30 °C (Ssu− or suppressor of Sui phenotype). A1193U (A968) alters the h31 loop, located below the P-site Met-tRNAiMet, whereas G875A (A663) changes h22, the core helix of the central domain, which is suggested to serve as the eIF platform. Thus, the mutant ribosome with an altered decoding site appears to be unable to mistakenly form a productive PIC on a UUG codon, even though SUI3-2 mutation promotes eIF2 release from the PIC positioned at the UUG codon.
Finally, we investigated whether A1193U (A968) and G875A (A663) directly impair integrity of the PIC by in vitro interaction assays. First, we purified mutant ribosomes and assayed their ability to promote [3H]Met-tRNAiMet binding to the 40 S subunit in the wild-type yeast extracts. As shown in Fig. 6A, solid line, the addition of purified wild-type 40 S subunit increased [3H]Met-tRNAiMet binding by 2.6 (+1.1, n = 4)-fold compared with the amount bound by the endogenous 40 S subunit. Importantly, 40 S subunits purified from A1193U (A968) and G875A (A663) strains were unable to promote the assembly, indicating defects in the interaction with Met-tRNAiMet. This defect observed with A1193U (A968) was as expected, because it alters a base potentially proximal to the P-site tRNA. This is also consistent with the analysis of PIC intermediates in vivo (Fig. 4C). In contrast, G875A (A663) does not directly alter the predicted Met-tRNAiMet-binding site. However, Met-tRNAiMet is delivered to the 40 S subunit in the TC with the GTP-bound eIF2, which in turn binds to the 40 S subunit in the context of MFC made with eIF1, -3, and -5 (8, 60). Thus, we hypothesize that G875A (A663) disrupts a critical interface to the MFC components, eIF1 or eIF3, in particular, as they are suggested to directly bind the 40 S subunit. Again, this is consistent with the observation of partial 43 S complex (the 43 S PIC lacking eIF1, -3, and -5) in the G875A (A663) mutant (Fig. 4D; also see under “Discussion”).
To directly test if A1193U (A968) and G875A (A663) are defective in the interaction with eIF1 and eIF1A, we performed GST pulldown assays. The 40 S subunits purified from RDN+ and G875A (A663) bound efficiently to GST-eIF1 or -eIF1A (Fig. 6B). However, A1193U (A968) altering the h31 loop reduced the interaction with both GST-eIF1 and -eIF1A (Fig. 6B, panel 3), suggesting that the conformation of the P-site is important for the small factors to bind near the decoding site. In agreement with this finding, A1193U (A968) severely impaired eIF1 and eIF1A loading to the 40 S subunit (Fig. 4C). In conclusion, the data presented here together suggest strongly that 18 S rRNA structure around the decoding site and the eIF platform is crucial for the integrity of the scanning PIC and the control of stringent AUG selection.
Here we identified, using yeast genetic screening, 10 new residues of yeast 18 S rRNA important for its function and/or maintenance. Not surprisingly, from reduced 40 S subunit abundance observed with the 18 S rRNA mutants (Fig. 2), the primary consequence of these mutations was to compromise the initiation, rather than elongation, phase of translation. Accordingly, these mutations led to strong accumulation of free 60 S subunit in the polysome profile characterized as type I (Fig. 2), and this effect directly led to slower growth rate (Fig. 3C). 60 S subunit accumulation has been observed in previous site-directed mutagenesis studies on 18 S rRNA (26), but here it became clear that the depletion of 40 S subunit is a general consequence of mutations altering 18 S rRNA (also see below).
We found that at least two of the strongest rdn mutations directly impair the 40 S subunit function in regulating start codon selection. A1193U (A968) and G875A (A663) increased the frequency of leaky scanning an AUG codon (Fig. 5B). We also provided genetic evidence suggesting that A1193U (A968) and G875A (A663) prevented the PIC on a UUG codon from maturing into the 40 S IC due to weak interaction between eIF2 and tRNAiMet (Fig. 5C). Biochemical evidence is presented that A1193U (A968) impairs the interaction with TC, eIF1A, and eIF1 (Fig. 6), thereby disrupting the integrity of the PIC in vivo (Fig. 4C). Therefore, it is conceivable that the h31 loop close to the P-site, altered by A1193U (A968), indeed interacts with Met-tRNAiMet during the scanning process and that the defective interaction leads to bypassing a start codon, regardless of whether it was correct or incorrect.
As for the h22 altered by G875A (A663), this is not only a part of the core helix structure (h21/22/23) of the 40 S subunit platform domain but is also predicted to constitute a universally conserved bridge to the 60 S subunit termed Bridge 2e (B2e) (also see Table 1, column 5) (15,–17). In vivo, G875A (A663) impaired the PIC formation, allowing accumulation of the partial 43 S complex containing only eIF2 (Fig. 4D). Although eIF3 binds the solvent side of the 40 S subunit, the eIF1/3/5 complex together can cover a significant portion of the interface side, including the intersubunit bridges (19). Disruption of B2e or the platform/body structure by G875A (A663) can affect the interaction with the eIF1/3/5 complex, such that this interaction cannot tolerate fractionation by sucrose gradient velocity sedimentation. Alternatively, the h22 mutation might have prevented the 43 S PIC formation via MFC, thereby allowing TC alone to bind the 40 S subunit. This model is consistent with the slower TC binding to the 40 S subunit bearing the same mutation observed in yeast extracts (Fig. 6A), because TC binding in the absence of MFC is expected to be slow. The h22 is also altered by A952C (G742), another Gcn− mutation (Fig. 5, A and B; supplemental Fig. S3), which introduces an additional base pair to the same bulge of h22 altered by G875A (A663) (Fig. 1A). Together with the Ssu− phenotype observed with G875A (A663) (Fig. 5C), this study highlights the importance of the h22 during the initiation function and, perhaps, the conformational change of the 40 S subunit in response to AUG selection.
We showed that mutations in 18 S rRNA can impair general control response (Gcn− phenotype) in part due to leaky scanning of uORF1 in the GCN4 mRNA leader (Fig. 5). This is not the consequence of depleting 40 S subunits observed with these mutations (Fig. 2), because a strong depletion of the 40 S subunits (2-fold increase in 60 S/40 S ratio compared with wild-type) caused by the deletion of RP51 encoding a small ribosomal protein does not cause Gcn− phenotype (61). In agreement with strong leaky scanning observed with A1193U (A968) and G875A (A663) (Fig. 5B), Dong et al. (31) showed that point mutations disrupting h28 increased leaky scanning of uORF1 and uORF4, the positive and negative regulatory elements for GCN4 translation, respectively.
Remarkably, however, some of the Dong et al. (31) mutations, e.g. A1152U (A928), were isolated as mutations that increase GCN4 expression independent of Gcn2 eIF2 kinase (General control derepressed or Gcd−). Because the leaky scanning of uORF1 impairs GCN4 translation and attendant general control response (3AT sensitivity or Gcn− phenotype), as observed here in Fig. 5A, the constitutive GCN4 translation by these mutations must be due to an additional mechanism. Indeed, Dong et al. (31) showed that the Gcd− mutations were suppressible by overproduction of eIF2 TC in vivo and impair TC loading to the 40 S subunit in vitro (31). Together with our findings (Fig. 5A), these results indicate that 18 S rRNA mutations altering P-site (h28 or h31) impair TC loading as well as post-assembly function in scanning or AUG selection. It is also noteworthy that the deletion of RPL16b encoding a large ribosomal protein has long been known to display a weak Gcd− phenotype and a modest leaky scanning of uORF4 (61). In contrast to our 18 S rRNA mutations, this deletion mutation leads to the depletion of the 60 S subunit. Thus, the phenotypes observed with rpl16b deletion would be solely due to the defect in 60 S subunit joining, which apparently allowed even the committed (wild type) 40 S subunit to bypass uORFs 2–4, the negative regulatory elements. This suggests that the 40 S subunit commitment to form 80 S initiation complex in response to AUG selection is canceled by the absence of the 60 S subunit after a certain time. More work involving thorough mutagenesis of 18 S and 25 S rRNAs is needed to fully understand the role of the ribosome structure and function in gene regulation, including uORF-mediated translational control of GCN4.
A1193U (A968) altering the h31 loop would likely affect the modification of Ψ 1191, located two bases away from A1193. The lack of Ψ 1191 modification due to the small nucleolar RNA mutation causes a cold-sensitive Slg− phenotype and accumulation of free 60 S subunit in the polysome profile and of 20 S species (62), all as observed with A1193U (A968) (this mutant grows slowly at 30 °C but not at 36 °C, consistent with a cold-sensitive growth, as shown in Fig. 2, panel 3). However, A1193U (A968) apparently confers a stronger Slg− phenotype than a strain defective in Ψ 1191 modification, so A1193U (A968) would directly impair the 40 S subunit function, rather than doing so via defective modification of the neighboring base.
Although the possible defect in Ψ 1191 modification may explain 40 S subunit depletion by A1193U (A968), the mechanism leading to reduced 40 S subunit abundance in other 18 S rdn mutants remains unclear. However, because we did not observe any change in 18 S rRNA intermediates except for A1193U (A968) (supplemental Fig. S5) and because all of the 10 18 S rRNA mutations conferred sensitivity to 3AT by a mechanism likely involving defective GCN4 translation (Fig. 5), it could be hypothesized that the functional defect of each mutant 40 S subunit led to a corresponding degree of its cytoplasmic degradation. That all the mutations except A1193U (A968) alter the body (5′) and platform (central) domains, which together include the major 60 S interface, is consistent with the idea that defective 60 S subunit joining can trigger such degradation (23, 24). On the other hand, we observed the accumulation of vacant monosomes in the polysome profiles of all the mutants (Fig. 2); thus, the rdn mutant 40 S subunits appear to able to form vacant ribosomes efficiently, and therefore, their selective degradation may not be explained by the failure to form vacant, inactive ribosomes.
Based on the data presented here, we suggest among other possibilities that most of the individual 18 S mutations disrupt intersubunit bridges directly or indirectly (see Table 1, column 5).6 This not only leads to defective 60 S subunit joining at the last step of initiation, potentially leading to selective degradation of the mutant 40 S subunit, but also results in defective formation of scanning-competent PIC. This is consistent with the idea that eIF1, -1A, -2, and -5, which govern the fidelity of start codon recognition, bind the 60 S subunit interface of the 40 S subunit (19). There is currently little information on the role of individual bridges in stabilizing subunit association, nor is it known which bridges are sequestered by eIFs for anti-association. Mutational studies of 18 S rRNA bases around some of the strong rdn mutations identified here would provide a profitable direction to be explored in the future.
We are indebted to Masayasu Nomura and Alan Hinnebusch for discussion and generous gifts of rdnΔΔ strains and [3H]Met-tRNAiMet, anti-TIF11, and p3415, respectively. We also thank Tatsuya Ii for technical assistance and Tsutomu Suzuki, Kotaro Fujii, and Makoto Kitabatake for discussion.
*This work was supported, in whole or in part, by National Institutes of Health Grant R01 Grant GM64781. This work was also supported by Innovative Award from Kansas State University Terry Johnson Cancer Center (to K. A.) and Wellcome Trust Grants 076456/Z/05/Z and Institutional Research Concept AVOZ50200510 (to L. V.).
The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1–S5, Tables S1–S5, and additional references.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EBI Data Bank with accession number(s) DQ888227.
4L. Valášek, unpublished material.
5T. Udagawa, N. Nemoto, C. R. Singh, S. Wang, S. Brown, and K. Asano, unpublished observations.
6G362A (C290) might disrupt the potential A-minor interaction with a helix in the central domain (denoted as “?” in Fig. 1A, as h21 does not possess an adenosine donor for this interaction). G1108A (G885) and ΩG1112 (G888) alter residues at the foot of h27, connecting a structure tightly packed by at least three A-minor interactions (blue boxes and lines in Fig. 1A), which may bring the h27 loop and adjacent areas closer to the coaxially stacked 21/22/23 helix, forming B2c, B2d, B2e, and eB9. Finally, A138G (A143) changes a base connecting two coaxially stacked helices made of h7/10 and h8/9, potentially disrupting the structure of 5′ domain that constitutes the major interface to the 60 S subunit.
7The RpsO-containing product smaller than 40 S was found in A1193U (A968) and A333G (A262) mutants (Fig. 4, C and E). This intermediate corresponds to only <1–2% of the total 40 S subunit and was observed by a longer exposure to show the diminished abundance of 40 S subunit. The abundance of the Rps0 product observed in A1193U likely reflects the destruction of the 40 S subunit with the most severe functional defect. Thus, we do not believe that the observation of these levels of the intermediate indicates a direct defect in the structure and composition of the pool of 40 S subunits; this would be rather due to partial destruction in response to the 18 S rRNA functional defect.
8The Gcn− phenotypes observed in Fig. 5A with the original constructs, A333G* (A262) and A952C* (G742) (each with a different 25 S mutation, see supplemental Table S2), were also observed with A333G (A262) and A952C (G742) carrying only the 18 S rRNA mutations (supplemental Fig. S3C).
3The abbreviations used are: