Wide mutational analysis to ascertain the functional roles of eL33 in ribosome biogenesis and translation initiation

An extensive mutational analysis of RPL33A, encoding the yeast ribosomal protein L33A (eL33) allowed us to identify several novel rpl33a mutants with different translational phenotypes. Most of the rpl33a mutants are defective in the processing of 35S and 27S pre-rRNA precursors and the production of mature rRNAs, exhibiting reductions in the amounts of ribosomal subunits and altered polysome profiles. Some of the rpl33a mutants exhibit a Gcd− phenotype of constitutive derepression of GCN4 translation and strong slow growth phenotypes at several temperatures. Interestingly, some of the later mutants also show a detectable increase in the UUG/AUG translation initiation ratio that can be suppressed by eIF1 overexpression, suggesting a requirement for eL33 and a correct 60S/40S subunit ratio for the proper recognition of the AUG start codon. In addition to producing differential reductions in the rates of pre-rRNA maturation and perhaps in r-protein assembly, most of the point rpl33a mutations alter specific molecular interactions of eL33 with the rRNAs and other r-proteins in the 60S structure. Thus, rpl33a mutations cause distinctive effects on the abundance and/or functionality of 60S subunits, leading to more or less pronounced defects in the rates and fidelity of mRNA translation. Supplementary Information The online version contains supplementary material available at 10.1007/s00294-022-01251-1.


Introduction
Ribosomes are highly conserved ribonucleoprotein machines that catalyze the translation of mRNA into proteins. Eukaryotic ribosomes consist of a large 60S and a small 40S ribosomal subunits, which in Saccharomyces cerevisiae contain the 25S, 5.8S and 5S ribosomal RNAs (rRNAs) and 46 different ribosomal proteins (r-proteins), and the 18S rRNA and 33 r-proteins, respectively. Ribosome biogenesis is a highly dynamic process that starts in the nucleolus with the synthesis of a 35S precursor (pre-rRNA) by RNA polymerase I, and a 5S pre-rRNA by RNA polymerase III. Early cleavages occur co-transcriptionally at the A0, A1 and A2 sites of the 35S pre-rRNA. Cleavage at A2 generates the 20S precursor of the 18S rRNA and the pre-27SA2, which will yield first the 27SA3 and then the 27SB precursors of the 25S and 5.8S rRNAs (Fig. 3A). Pre-rRNA processing and rRNA modification and folding are closely coordinated with the assembly of r-proteins and non-ribosomal trans-acting factors to form different pre-ribosomal particles, which move from the nucleolus to cytoplasm where they undergo final maturation and acquire translation competence (reviewed in Gerhardy et al. 2014;Henras et al. 2015;Klinge and Woolford 2019;Konikkat and Woolford 2017;Kressler et al. 2017;Woolford and Baserga 2013). Ribosomal proteins play an active role in ribosome biogenesis and function. Most r-proteins initially bind with low affinity to nascent pre-rRNA in the nucleolus but become more stably associated during maturation of preribosomal particles. Ribosomal proteins are required for the efficient processing and proper folding of pre-rRNAs and to stabilize and maintain the structure of the pre-ribosomal particles. They participate in the nucleo-cytoplasmic transport of the pre-ribosomal particles and serve as binding sites for specific ribosome assembly factors and components of the translation machinery (reviewed in de la Cruz et al. 2015;Graifer and Karpova 2015)).
Systematic analyses of yeast mutant strains allowing conditional depletion of one 40S or 60S r-protein revealed their roles in distinct steps of ribosome maturation (Ferreira-Cerca et al. 2005;Poll et al. 2009). However, the examination of the effects of discrete mutations in different r-proteins can reveal additional functions of these components of mature ribosomes in ribosome biogenesis and in translation. Atomic resolution structures of eukaryotic ribosomes allow the prediction of the molecular interactions of most r-proteins with other r-proteins and with the rRNAs (Ben-Shem et al. 2011;Greber 2016;Greber et al. 2012;Klinge et al. 2011Klinge et al. , 2012Yusupova and Yusupov 2014). Semi-quantitative proteomic analyses of pre-ribosomal particles from yeast r-protein mutants provided further information about the roles of the corresponding r-proteins in stabilization, pre-rRNA processing and ribosome assembly (Gamalinda et al. 2014;Jakovljevic et al. 2012;Linnemann et al. 2019;Ohmayer et al. 2013Ohmayer et al. , 2015. Eukaryotic translation is a central biological process and a key point in the regulation of gene expression. Translation initiation begins with the formation of a ternary complex (TC) of initiator tRNA (tRNA i ), the eukaryotic initiation factor 2 (eIF2) and a molecule of GTP. Additional eIFs (1, 1A, 3 and 5) promote the recruitment of the TC to the 40S ribosomal subunit to form the 43S pre-initiation complex (PIC), which binds first to the 5ʹ end of the mRNA in a manner facilitated by eIF4F and the poly (A)-binding protein. The 43S PIC scans the mRNA leader base-by-base looking for an AUG in a suitable sequence context. Once the start codon is recognized, eIF1 is ejected from near the ribosomal P-site to permit base pairing between tRNA i and the AUG codon with the subsequent release of P i produced by hydrolysis of the GTP molecule in TC. These events trigger the transition to a closed conformation of the PIC that is incompatible with scanning. The GTPase eIF5B collaborates with eIF1A to catalyze the joining of the 60S ribosomal subunit, producing the 80S initiation complex that is competent for elongation (reviewed in Hinnebusch et al. 2007;Hinnebusch and Lorsch 2012).
Translational machinery components that are critical for accurate AUG selection were identified genetically in S. cerevisiae by the isolation of mutations that decrease initiation accuracy and allow increased translation from an UUG start codon (Sui − phenotype). Such Sui − mutations have been described in different translation initiation factors (i.e. eIF2, eIF3, eIF5, eIF1 or in the C-terminal tail of eIF1A (for review see (Hinnebusch 2014)). Mutations that suppress initiation at UUG codons in Sui − mutants and confer enhanced fidelity of start codon recognition (Ssu − phenotype) have been identified in eIF1 and the NTT of eIF1A (Fekete et al. 2007;Martin-Marcos et al. 2011Saini et al. 2010), in the 40S subunit r-proteins Rps5/uS7 and Rps3/uS3 Visweswaraiah et al. 2015) and in the 25S rRNA of the yeast 60S ribosomal subunit (Hiraishi et al. 2013), indicating an important role of both 40S and 60S subunits in accurate AUG codon recognition.
The 60S subunit protein rpL33 (eL33 according to the universal nomenclature proposed by (Ban et al. 2014)), present in eukaryotes and archaea but not in bacteria, is encoded by two paralogous genes in budding yeast, RPL33A and RPL33B, that are differentially expressed (Tornow and Santangelo 1994). RPL33A produces mRNA at a level ~ sixfold higher than that of RPL33B when fused to a reporter gene (Tornow and Santangelo 1994); and, similarly, steadystate levels of native RPL33A mRNA are ~ sixfold higher than those of RPL33B as judged by Northern blot analyses (Martin-Marcos et al. 2007). Accordingly, the rpl33aΔ null mutant is viable but exhibits a severe slow growth phenotype, the rpl33b∆ null mutant shows normal growth, and the rpl33a∆ rpl33b∆ double mutant is inviable, indicating that eL33 is an essential protein (Martin-Marcos et al. 2007;Tornow and Santangelo 1994). We previously showed that the rpl33a-G76R mutation in RPL33A impairs efficient processing of 35S, 27SA 2 and 27S pre-rRNAs, causing a deficit of 60S subunits. In addition, rpl33a-G76R constitutively derepresses GCN4 mRNA translation (Gcd − phenotype) (Martin-Marcos et al. 2007). GCN4 is a transcriptional activator of many biosynthetic genes whose translation is repressed in yeast under conditions of amino acid sufficiency through a reinitiation mechanism involving four short upstream open reading frames (uORFs 1-4) in the GCN4 mRNA leader that is very sensitive to TC levels. After translating the uORF1, the 40S subunits remain bound to the leader and rebind the TC in time to reinitiate translation at the inhibitory uORFs 3 or 4. This prevents 40S subunits from reaching the GCN4 start codon, keeping GCN4 translation repressed. In starvation conditions, TC abundance in cells is reduced by phosphorylation of eIF2α by kinase GCN2. As a result, a fraction of the reinitiating 40 subunits bypass uORFs 2-4 before reacquiring TC and proceed to initiate at the GCN4 start codon instead (Hinnebusch 2005). Under non-starvation conditions, reductions in the amount of 60S subunits caused by RPL33A deletion or the -G76R mutation, or alterations in the functions of 60S subunits provoked by the -G76R mutation, allow reinitiating 40S subunits to abort uORF4 translation, resume scanning and initiate at the GCN4 start codon, evoking the Gcd − phenotype (Martin-Marcos et al. 2007).
Diamond-Blackfan anemia (DBA) is a bone marrow failure syndrome characterized by anemia, congenital abnormalities, and cancer predisposition linked to defects in ribosome synthesis and function. Sequence analysis of a cohort of DBA probands revealed three mutations in RPL35A, the human ortholog of RPL33A, which could be a potential target for diagnosis or treatment of the disease (Farrar et al. 2008). Knockdown of RPL35A in hematopoietic cell lines revealed that rpL35A/eL33 is essential for maturation of pre-ribosomal rRNAs, 60S subunit biogenesis and cellular proliferation (Farrar et al. 2008). A mutant strain lacking RPL33A was analyzed as a "yeast model" of DBA and showed defects in 27SA2/A3 processing, accompanied by a decrease in the levels of 27SB precursors with the consequent reductions in the amounts of 5.8S and 25S rRNAs, which resulted in a deficit of 60S ribosomal subunits (Moore et al. 2010). Similar results were obtained in a systematic study of the rRNA maturation process in yeast mutant strains conditional for expression of individual large subunit ribosomal protein genes. Depletion of eL33 showed an elevated 27SA2 to 27SB pre-rRNA ratio under non-permissive versus permissive conditions and primer extension analyses revealed that in the rpl33a mutant strain, different 27SB pre-rRNAs species and the 7S pre-rRNA were detected at lower levels than in a wild-type (WT) strain, suggesting that eL33 is required to generate the major 5.8S rRNA 5'end at site B1S (Poll et al. 2009). eL33 was identified as a member of the yeast 60S r-protein dII/dVI cluster that binds at a region of the 60S secondary structure domain II (25S rRNA nucleotides 652-1455), and establishes contacts with the expansion segment 7 (ES7, 25S rRNA nucleotides 436-624), the 60S rRNA domain VI (25S rRNA nucleotides 2996-3397) and the ES39 (25S rRNA nucleotides 3152-3295) (Ohmayer et al. 2015). The dII/dVI cluster is composed of the r-proteins rpL6/eL6, rpL14/eL14, rpL16/uL13, rpL20/ eL20 and rpL33/eL33. The formation of this cluster is a downstream event in the 60S rRNA domain II assembly pathway and is required for early steps of 60S pre-rRNA maturation (processing at site B1S to generate the 5.8S rRNA 5' end) and stabilization by anchoring the 60S rRNA domain VI including ES39, to domain II and ES7, which facilitates the association of different ribosome biogenesis factors and recruitment of other ribosomal proteins required in successive steps of the 60S maturation pathway (Ohmayer et al. 2015).
In this work, we performed an exhaustive mutational genetic analysis of the yeast RPL33A gene with the aim of investigating which specific elements, motifs or residues are important for functions of eL33 in ribosome biogenesis, translation, and recognition of the AUG start codon. Among others, we selected to alter amino acid residues that interact with other r-proteins and rRNA domains of the 60S ribosomal subunit, and residues found mutated in human r-protein rpL35A/eL33 related with DBA and different tumors, to try and determine in S. cerevisiae the functional consequences of substitutions associated with DBA and cancer. We identified several mutations in rpl33a that lead to different cellular phenotypes. A first set of rpl33a mutants showed strong slow growth (Slg − ) phenotypes, a reduced production of 25S, 18S and 5.8S mature rRNAs and a strong deficit in 60S ribosomal subunits that could also be not completely functional, smaller reductions in the 40S subunit amounts and attendant defects in polysome assembly and general translation.
Moreover, these mutants have a modest 3AT R /Gcd − phenotype which results from constitutive derepression of GCN4 translation, and a slight Sui − phenotype (increased UUG/AUG initiation ratio) suppressable by overexpression of eIF1. A second set of rpl33a mutants showed severe Slg − phenotypes only at 37 °C and similar defects in pre-rRNA processing as those displayed by the first set of rpl33a mutants, resulting in reduced amounts of 60S subunits and defects in polysome assembly and general translation at the restrictive temperature, but not Gcd − or Sui − phenotypes. A recent review analyzed the known phenotypes of each r-protein mutant in S. cerevisiae, suggesting that the pleiotropic phenotype of those mutants, and associated gene expression changes, mainly results from the canonical roles of r-proteins in the ribosomes (Polymenis 2020).

Yeast strains
The Saccharomyces cerevisiae strains employed in this study are listed in Table 1.
Derivatives of strain Hm700 were obtained by transformation with low copy (lc) or high copy number (hc) LEU2   This study H2994 MATa ura3-52 trp1∆63 leu2-3, leu2-112 his4-301(ACG) This study Hm700   plasmids harboring the appropriate RPL33A alleles and selection on SC medium lacking leucine (Leu). Derivatives of strain Hm701 were obtained by transformation to Leu + with lc LEU2 plasmids harboring the appropriate RPL33A alleles on SC-Leu medium and the resident RPL33A + URA3 plasmid (pPM2) was evicted by selecting for growth on 5-fluoorotic acid (5-FOA) medium.

Plasmids
Plasmids used in this study are listed in Table 2. pAGH1 was created by replacing the kanMX module of plasmid pUG6 (loxP-kanMX-loxP) (Guldener et al. 1996) by NatMX4 obtained by digestion of plasmid pAG25 with SacI and BglII (Goldstein and McCusker 1999).
pPM17 (RPL33A in pRS315, lc LEU2) was constructed by inserting a 1.9 kb XbaI-SalI fragment from plasmid pPM1 (Martin-Marcos et al. 2007) into the XbaI-SalI sites of pRS315. To generate plasmids from pPM40 to pPM55 and from pPM65 to pPM73 the QuikChange site-directed mutagenesis system (Stratagene) was used with the appropriate primers listed in Table 3 using pPM2 (Martin-Marcos et al. 2007) or pPM17 templates. To avoid the putative appearance of extra-mutations in the vector by the Quick-Change system, 1.9 kb XbaI-SalI fragments containing rpl33a mutant alleles were cloned into the corresponding sites of pRS316 or pRS315. To generate the high copy (hc) versions of the rpl33a mutants, 2.3 kb PvuII fragments from lc plasmids (pRS315) harboring rpl33a with the indicated mutations, were cloned into the PvuII sites of pRS425.
Analysis and visualization of the 60S ribosomal subunit and eL33 protein structure models S. cerevisiae 60S subunit structure models were downloaded from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB-PDB). Structure files used were 4V88 (Ben-Shem et al. 2011) and 4V8T (Greber et al. 2012). Structures were displayed and analyzed with the PyMOL Molecular Graphics System, version 1.5.0 and the Discovery Studio 3.5 software.

Biochemical techniques
(i) Assays of β-galactosidase activity in whole-cell extracts (WCEs) were performed as described previously (Moehle and Hinnebusch 1991) with cells grown in SD medium + histidine (His) and tryptophan (Trp) to an A 600 of ~ 1.0. At least four biological replicates (independent transformants) were employed for all β-galactosidase activity measurements. For derepressing conditions, cells were grown for 6 h in SD + His + Trp with sulfometuron-methyl (0.5 μg/ ml). β-Galactosidase activities are expressed as nanomoles of o-nitrophenyl-β-D-galactopyranoside cleaved per minute per mg. Unpaired t tests were performed to compare wild type and mutants mean values and, when indicated, the change was considered significant if the two-tailed P value was < 0.05.
(ii) To measure the steady-state levels of RNAs, Northern analysis were carried as follows. Cells were grown in liquid SC-Leu medium at 28 °C to and OD 600 of ~ 1.0 and, when indicated, transferred to 37 °C for 6 h. Total RNA was extracted from equivalent numbers of cells (5 × 10 8 ) by the phenol-acid method (Schmitt et al. 1990). Samples containing 10 μg of total RNA were electrophoresed in 1.2% agarose and 4% formaldehyde gels. The RNAs were blotted to positively charged nylon membranes (Roche) and immobilized by UV cross-linking with a UV Stratalinker 2400 (Stratagene). To detect the pre-rRNAs and mature rRNAs, the blots were hybridized sequentially with ~ 20 nts long oligonucleotides labeled at their 5′ ends with [γ-32 P] ATP (6,000 Ci/mmol), and direct quantification of the corresponding hybridization signals was performed using a PhosphorImager. The oligonucleotides used as probes are listed in Table 3.
(iii) Polysome analysis by sucrose gradient centrifugation was basically conducted as previously described (Foiani et al. 1991). Cells were grown in SC-Leu at 28 °C to mid-logarithmic phase, and when indicated, cultures were incubated at 37 °C for 6 h to an A 600 of ~ 1.0. Cycloheximide was added at a final concentration of 100 μg/ml. WCEs were extracted as described (Foiani et al. 1991) and loaded onto 7% to 50% sucrose density gradients, which were scanned at 254 nm.

Identification of different rpl33a mutations
We previously identified the rpl33a-G76R mutation as a spontaneous suppressor of the inability of the gcn2 gcn3 double mutant strain H117 to derepress GCN4 mRNA translation under amino acid starvation, producing a Gcd − phenotype (Martin-Marcos et al. 2007). The G76R substitution also impairs the efficient processing of pre-rRNAs and reduces the levels of mature rRNAs, indicating a role for r-protein eL33 in ribosome biogenesis. Gly-76 is localized between two β-sheets in a motif of 22 amino acids well conserved in all the members of the r-protein L35Ae family. To investigate which eL33A residues are important for proper ribosome biogenesis or/and efficient translation, and to explore in yeast the functional consequences of substitutions associated with DBA and cancer in human rpL35A/ eL33, we conducted site-directed mutagenesis of the paralogous gene RPL33A to alter amino acid residues that (i) interact with r-proteins of the dII/dVI cluster (rpL6/eL6 and rpL16/uL13), with L32/eL32 and with different nucleotides of the 60S domains I, II and VI and ES7 and ES39 Single amino acids that interact with eL6 (Ser-4, Arg-6, Leu-7, Tyr-8, Gly-34, Val-35, Asp-40 Phe-43, Tyr-44 and from Leu-102 to Ile-107), with uL13 (Lys-12, Glu-33 and Gly-95) and with eL32 (Val-22), and residues localized close to Gly-76 (Gly-69, Arg-73, Gly-79 and Arg-82) were substituted either by Ala, which replaces bulky or charged side chains with a methyl group, or by Arg or Glu that introduce a positive or a negative charge, respectively. We also generated in the yeast RPL33A paralogous gene the corresponding human substitutions and truncations associated with DBA (∆L29, V35I and ∆99-107) and with different tumor entities (S4P, L7Q, S28T, R48I, A50T, S56N, G79C, R82C and F94T), and the above indicated amino acid residues were also substituted by Ala, Arg and Glu. The Lys residues at positions 31, 47 and 92, which are susceptible to post-translational modifications, were substituted by Ala. In addition, multiple-Ala substitutions of consecutive residues as indicated in Fig. 1A were generated; although in mutants A40-44 and A92-99, not every amino acid residue was substituted (DAQFY 40-44 AAQAA and KTFGASVR 92-99 AAA GAA VA). As shown in Fig. 1A, C, the substitutions are localized in different structural elements of eL33 as that would help to elucidate which specific elements or residues are required for proper function of this essential r-protein.
The appropriate mutations were generated in RPL33A under its native promoter on low copy (lc) LEU2 or URA3 plasmids; the resulting URA3 plasmids were introduced into the gcn2 gcn3 rpl33aΔ RPL33B Hm538 strain to screen for mutations that confer 3AT resistance (3AT R ), indicating the Gcd − phenotype. The Hm538 strain is sensitive to 3AT, an inhibitor of histidine biosynthesis, because its gcn2 and gcn3 mutations impede derepression of GCN4 under histidine starvation conditions imposed by 3AT. We found that in a first set of mutants (from now on Set 1), the single amino acid substitutions rpl33a-Y44R, G69R, L102R and the previously described G76R, as well as the double mutant G76R-G79R, the nine amino acid carboxy-terminal deletion (∆99-107), and carboxy-terminal blocks of Ala substitutions A102-105 and A102-107, all conferred 3AT R / Gcd − phenotypes of different degrees ( Fig. 2A). This Set 1 group of rpl33a of substitution mutations affect residues that (i) interact with eL6 (Tyr-44 and from Leu-102 to Ile107)) (Figs. S2-B and S1-B) and the 25S rRNA ES7 (Pro-104 and Asn-106) (Fig. S1-C) (ii) belong to the 22-residue motif conserved in the L35Ae family (Gly-69, Gly-76 and Gly-76, Gly-79), or (iii) were found mutated in patients with DBA (∆99-107) or in tumor entities (Gly-79).
The rpl33a∆ null allele in strain Hm538 was constructed to preserve the integrity of YPL142C, a hypothetical ORF encoded on the DNA strand opposite to that encoding RPL33A. Since we frequently observed the appearance of Slg + revertants or spontaneous suppressors in the Hm538 strain, we deleted the complete RPL33A ORF to produce strain Hm700, of genotype GCN2 GCN3 rpl33a∆ RPL33B his4-301, which did not produce Slg + colonies. This new strain also allowed us to investigate both, whether any of the mutant rpl33a alleles produce 3AT-sensitivity indicative of a Gcn − phenotype, and/or reduce the stringency of AUG start codon selection by suppressing the His − phenotype conferred by an initiation codon mutation of his4-301.
Under amino acid starvation conditions, the GCN4 protein activates transcription of the HIS4 gene among many other amino acid biosynthetic genes (Hinnebusch 1988). Gcd − mutations increase expression of a HIS4-lacZ reporter under non-starvation conditions, because they constitutively derepress GCN4 expression (Harashima and Hinnebusch 1986). To provide a quantitative estimation of the strength of the mutant Gcd − phenotypes, an empty vector, WT RPL33A, or mutant rpl33a alleles were introduced on lc LEU2 plasmids into the rpl33a∆ RPL33B his4-301 strain Hm700. The transformants were cultured in non-starvation conditions (SD medium) and β-galactosidase activities synthesized from a HIS4-lacZ reporter with a WT AUG start codon loxP-kanMX-loxP (Guldener et al. 1996) pAGH1 loxP-natMX4-loxP This study  TTT TAA CAT ACT TCT ATG GGA TTC 3ʹ  FL7R.2  5ʹ CTT TTT TTT TTT TAG GGT ACG TCA AAG GTA AG 3ʹ  RR7R.2  5ʹ CTT ACC TTT GAC GTA CCC TAA AAA AAA AAA AG 3ʹ  F∆L29  5ʹ CAA CCC AAA TGT CTC TAT CAA GAT CGA AGG TG 3ʹ  R∆L29  5ʹ CAC CTT CGA TCT TGA TAG AGA CAT TTG GGT TG 3ʹ  FV35R  5ʹ GAT CAA GAT CGA AGG TCG CGC TAC TCC ACA AG 3ʹ  RV35R  5ʹ CTT GTG GAG TAG CGC GAC CTT CGA TCT TGA TC 3ʹ  FF43R 5ʹ were measured in whole-cell extracts (WCEs). Consistent with their Gcd − phenotypes, all the 3AT R mutants showed a two-to threefold increase in β-galactosidase expression under amino acid sufficiency conditions, which were similar to those of the rpl33a∆ transformants containing only empty vector, in comparison to the WT RPL33A strain (Fig. 2B).
In addition to the 3AT R /Gcd − phenotypes, the Set 1 of rpl33a mutations confer a slow growth (Slg − ) phenotype on synthetic complete (SC) medium at 28 °C, a severe temperature sensitivity at 37 °C (Ts − ) and cold sensitivity (Cs − ) at 18 °C (Fig. 2C, lc panels), suggesting that these mutations impair functions of eL33 that affect essential processes of Secondary structures are represented as a bar (α-helix), and arrows (β-sheets). Residues substituted by site-directed mutagenesis are highlighted in colors: green, red and pink for those amino acids that interact with ribosomal proteins L6 (eL6), L16 (uL13), and L32 (eL32), respectively; blue, for those that map to a conserved motif of 22 amino acids located between Gly-69 and Pro-90 in the carboxy-terminal region of eL33, or those that were found mutated (i) in patients with Diamond-blackfan anemia (DBA) or (ii) in different tumor entities (Catalogue o Somatic Mutations in Cancer, COSMIC); orange, for residues that are targets of post-translational modifications; the black triangle indicates the position of a C-terminal deletion of the L33A protein described in patients with DBA. Multiple-Ala substitutions of consecutive residues are represented with bars.
Red bars and red asterisks above indicate amino acid substitutions that produce Gcd − and/or slow growth phenotypes at different temperatures. B Amino acid sequence comparison between yeast and human eL33. Residues highlighted were found mutated in patients with DBA (red) or in several tumor entities (blue) and the correspondent mutation is marked above. The ∆ symbol indicates a leucine deletion C Structure of eL33 where mutated residues that confer different phenotypes are signaled in red. D Position of eL33 in the 60S ribosomal subunit. The yeast 60S ribosomal subunit is shown viewed from the solvent exposed side with the 25S rRNA in yellow and ribosomal proteins in grey. Ribosomal protein L33 (eL33) is shown in blue, L6 (eL6) in green, L16 (uL13) in red, and L32 (eL32) in pink the cell. The Slg − phenotypes at 28 °C are relatively more severe in the rpl33a-G69R, -G76R, -G76R-G79R mutants and the strong Slg − phenotype of rpl33a-∆99-107 is comparable to that of the rpl33a∆ mutant. At 37 °C, all the rpl33a mutants showed similar Ts − phenotypes; and at 18 °C, the Slg − phenotypes are similar or stronger than those observed at the permissive temperature in all the mutants, except for rpl33a-Y44R.
The parental rpl33a∆ RPL33B his4-301 Hm700 strain cannot grow on medium lacking histidine (His), because the mutant his4-301 mRNA lacks an AUG start codon (His − phenotype). Sui − mutations increase initiation at the third (UUG) codon at his4-301 and restore the ability to grow in a medium without His (His + phenotype). However, we found that none of the rpl33a mutations allowed detectable growth on medium without His (His − phenotype) suggesting the absence of marked Sui − phenotypes (data not shown).
The rpl33a mutant alleles were then introduced on high copy (hc) plasmids into the same rpl33a∆ RPL33B his4-301 Hm700 strain, to examine whether overexpression of the mutant proteins would compensate for possibly suboptimal levels of eL33 and attendant reductions of 60S subunits when the alleles are expressed from lc number plasmids instead. As shown in Fig. 2C, at 28 °C, hc rpl33a-Y44R suppressed the Slg − phenotype observed with lc rpl33a-Y44R. Surprisingly, although expression of all rpl33a mutant alleles from hc plasmids slightly suppressed the Ts − phenotypes of all Set 1 mutants at 37 °C, the same phenotypes were observed in the corresponding mutants bearing a high copy number empty vector, (Fig. 2C, central panels, bottom raw) indicating that the suppression, at least at 37ºC cannot be attributed to increased expression of the eL33 variants. At 18 °C, the Cs − phenotypes of rpl33a-Y44R, G69R, L102R, A102-105 and A102-107 mutants were slightly suppressed when mutant rpl33a alleles were expressed from hc plasmids. This suppression is most pronounced in the rpl33a-G69R mutant (Fig. 2C). In contrast, we did not observe any suppression of the Cs − phenotype by rpl33a-G76R, G76R-G79R or 99-107 mutant alleles expressed from hc plasmids (Fig. 2C).

Defects in pre-rRNA processing caused by mutations in rpl33a
To investigate defects in the pre-rRNA maturation pathway caused by mutations in different regions of RPL33A, we conducted Northern analysis of pre-rRNA and mature mRNA species in WT and the two sets of rpl33a mutant cells grown to mid-logarithmic phase in SC medium at 28ºC (Figs. 3B, C and 4A), and after shifting the Set 2 group of mutants to 37 ºC for 6 h (Fig. 4B). The pre-rRNA processing pathway in S. cerevisiae and the probes used for Northern analysis are shown in Fig. 3A. The RNAs were quantified relative to the level of SCR1, the RNA component of the Signal Recognition Particle (SRP) transcribed by RNA polymerase III, used as an internal control.
First, we analyzed the effects of Set 1 rpl33a mutations expressed from lc plasmids in rpl33a∆ RPL33B strain Hm700 on the pre-rRNA processing pathway and on the level of RPL33A mRNA (Fig. 3B). Appreciable reductions (~ 40-90%) in the steady-state levels of 25S rRNA were observed in all the Set 1 lc rpl33a mutants; and in lc rpl33a-Y44R, G76-G79R and ∆99-107 those reductions were comparable to that of the empty vector transformant of Hm700 (~ 90%) (Fig. 3B, Table 4). The abundance of 18S rRNA was also reduced in all the rpl33a mutants, but in general to a lesser extent than the reductions of Fig. 2 Set 1 rpl33a mutations confer both Gcd − and/ or Slg − phenotypes. A 10 5 cells of gcn2 gcn3 rpl33a∆ RPL33B strain Hm538 containing the indicated RPL33A alleles on low copy (lc) plasmids or an empty vector were spotted on synthetic complete (SC) medium lacking uracil (Ura) and SC medium containing 5 mM or 10 mM 3-Amino-1,2,4-triazole (3AT) and incubated for 5 days at 28 °C. B Derivatives of strain Hm700 his4-301 rpl33a∆ RPL33B containing the indicated RPL33A alleles on lc plasmids or an empty vector also harboring HIS4-lacZ reporter with an AUG start codon (plasmid p367) were cultured in synthetic dextrose minimal medium (SD) supplemented with histidine (His) and tryptophan (Trp) at 28 °C to A 600 of ~ 1.0, and β-galactosidase activities were measured in wholecell extracts (WCEs) in units of nanomoles of o-nitrophenyl-β-Dgalactopyranoside cleaved per min per mg. The mean and standard error (SE) from at least four independent transformants are reported. The values in the right column are the results expressed relative to the corresponding WT value (1.0). C Ten-fold serial dilutions of his4-301 rpl33a∆ RPL33B strain Hm700 containing the indicated RPL33A alleles on lc or high copy (hc) plasmids or empty vectors were grown at 28 °C, 37 °C and 18ºC for 3.5, 4.5 and 7 days, respectively, on SC medium lacking leucine (Leu). D Ten-fold serial dilutions of his4-301 rpl33a∆ rpl33b∆ strain Hm701 containing the indicated RPL33A alleles on lc plasmids were grown at 28 °C, 37 °C and 18ºC for 2.5, 3 and 6 days, respectively, on SC-Leu ◂ 1 3 25S rRNA, (~ 25-60% and ~ 70% in rpl33a∆) (Fig. 3B, Table 4), indicating a stronger effect of these rpl33a mutations in the maturation pathway of the 60S versus 40S subunit. Early cleavages at the A0-A1-A2 sites of the 35S pre-rRNA were reduced in all the rpl33a mutant strains, with attendant increases in 35S/32S, 35S/27SA2, 35S/27S, and 35S/20S ratios, with relatively greater increases observed in rpl33a-Y44R, G76R, G76-G79R, ∆99-107, A102-107 and rpl33a∆ mutants (Fig. 3B, Table 5). Almost no detectable amounts of 33-32S precursors, and substantial reductions in the levels of 27SA 2 , 27S, 7S and 20S pre-rRNAs were observed in all the rpl33a mutants, which is consistent with the lower amounts of 25S, 5.8S and 18S rRNAs detected in the mutants in comparison with the WT. However, the 27SA2/25S, 27S/25S and 27S/7S ratios were higher in the mutants than in WT (Table 5), likely reflecting defects in processing of the 27S pre-rRNA species at A3, B1S and C2 sites, as previously reported for the rpl33-G76R mutant and strains lacking RPL33A (Farrar et al. 2008;Martin-Marcos et al. 2007;Poll et al. Fig. 3 Defects in ribosomal RNA processing caused by Set 1 rpl33a mutations. A Scheme of the yeast rRNA processing pathway. The 35S pre-rRNA contains the sequences for mature 18S, 5.8S, and 25S rRNAs separated by two internal transcribed spacer sequences (ITS1 and ITS2) and flanked by two external transcribed spacer sequences (5' ETS and 3'ETS). The rRNAs are represented as bars and the transcribed spaces as lines. The processing sites and the annealing positions of oligonucleotides used as probes are indicated by the letters A to E above the diagram and by the numbers 1 to 6 beneath all rRNAs, respectively. B Derivatives of strain Hm700 (his4-301 rpl33a∆ RPL33B) containing the indicated RPL33A alleles on lc or hc plasmids or empty vectors were cultured in liquid SC-Leu at 28 °C to A 600 of ~ 1.0. Total RNA was extracted and samples containing 10 µg were resolved on 1.2% agarose 4% formaldehyde gels and subjected to Northern blot analysis.  Fig. 4 Defects in ribosomal RNA processing caused by Set 2 rpl33a mutations. Derivatives of strain Hm701 (his4-301 rpl33a∆ rpl33b∆ containing the indicated RPL33A alleles on lc plasmids were cultured in liquid SC-Leu at 28 °C to A 600 of ~ 1.0 (A) or grown to mid-logarithmic phase in SC-Leu at 28 °C and then shifted to 37 °C for 6 h (B). Total RNA was extracted and samples containing 10 µg were subjected to Northern analysis as indicated in Fig. 3B A W T Y 1 0 3 R V 3 5 R L 7 R F 4 3 R A 9 2 -9 9 A 4 0 -4 4 G 7 9 R ∆ L 2 9

33-32S
mRNA RPL33A 1 3 2009). Although mature 18S rRNA is less abundant in these mutants than in the WT (~ 25-70%), there was also a comparable reduction in the level of 20S pre-rRNA, indicating that these mutations do not affect processing of 20S pre-rRNA to 18S rRNA. Although the levels of 5.8S were diminished (~ 40-65%) in the rpl33a mutants and ~ 70% in rpl33a∆ (Fig. 3B, Table 4), the ratio 7S/5.8S was similar or slightly lower in the mutants than in the WT ( Table 5), suggesting that the processing of 7S to 5.8S is normal in rpl33a mutant cells. As expected, the RPL33A mRNA is absent in the strain transformed with an empty vector (rpl33a∆) (Fig. 3B). From lc plasmids, rpl33a-102R, A102-105 and A102-107 mutants showed RPL33A mRNA levels similar than those of WT, whereas rpl33a-Y44R, G69R, G76R, G76R-G79R, and -∆99-107 showed significant increases in RPL33A mRNA abundance (Fig. 3B, Table 4). When WT RPL33A is expressed from a hc plasmid, the levels of RPL33A mRNA increase ~ fivefold with respect to lc RPL33A ( Fig. 3B and C); and similar increases from ~ 3 to ~ ninefold (A102-107 mutant) in the amounts of RPL33A mRNA were detected when rpl33a mutant alleles were expressed from hc plasmids (Fig. 3C). Although stabilization of the specific RPL33A mRNAs by an unknown mechanism cannot be excluded, the amounts of 25S and 60S subunits in these mutants did not increase (Table 4), suggesting that there would not be an increase in the amount of the corresponding overexpressed eL33A mutant proteins assembled into ribosomes in these mutants.

SCR1
We found similar defects in pre-rRNA processing produced by Set 1 rpl33a mutations on hc plasmids (Fig. 3C). Amounts of 25S, 18S and 5.8S rRNAS were reduced to a similar extent as observed in cells harboring the corresponding lc rpl33a mutants, with a slight increase in the 25S rRNA level in the hc rpl33a-Y44R versus the lc version ( Fig. 3B and C; and Table 4), which is consistent with the less accentuated Slg − phenotype of the hc rpl33a-Y44R strain (Fig. 2C). In contrast, hc rpl33a-G69R leads to a greater reduction of 25S rRNA than does lc rpl33a-G69R. This might be explained by the observation that in the lc rpl33a-G69R mutant we frequently observe the appearance of revertants or suppressors that could be masking the defects produced by the G69R mutation, or by a possible dominant-negative effect of the overexpressed G69R mutant in the hc rpl33a-G69R strain. Ratios of pre-rRNAs to mature rRNAs for Set 1 rpl33a mutations on hc plasmids are shown in Table 5. Thus, with the notable exception of rpl33a-Y44R, the pre-rRNA processing defects are very similar in cells with rpl33a mutant alleles expressed from hc or lc plasmids, suggesting that those defects do not arise merely from reduced expression of the eL33A mutant proteins when rpl33a mutant alleles are present in lc plasmids. Rather, the mutations appear to affect the function of eL33A in contributing to 60S biogenesis.
In summary, the Northern analyses in Figs. 3, 4 revealed defects indicative of impaired processing of 35S pre-rRNA at sites A0-A1-A2 in all of the analyzed Set 1 rpl33a mutants, in Set 2 rpl33a mutants Y103R and A40-44 at both 28 °C and 37 °C, and the rest of the Set 2 rpl33a mutants at 37 °C, leading to strong reductions in the levels of 33-32S, 27SA 2 , 27S, 7S and 20S pre-rRNAs in all of the mutants. Defects in processing of 27S pre-rRNAs and probably a pronounced destabilization of early and intermediate 60S pre-RNAs, could contribute to the reduced production of 25S, and 5.8S mature rRNAs. Together, these data support the idea that a wide range of mutations affecting key eL33A residues substantially impair the efficiency of different pre-rRNA processing steps in the ribosomal maturation pathway.

Defects in polysome assembly caused by rpl33a mutations
To investigate whether the rpl33a mutations reduced steady-state levels of 60S subunits and also confer reductions in general translation of mRNAs, we analyzed profiles of free-ribosomal subunits, 80S monosomes and polysomes from WT and Set 1 and Set 2 rpl33a mutants by sucrose gradient-velocity sedimentation. The ratio of polysomes to monosomes (P/M) was calculated to determine the effects of the mutations on the rate of bulk translation initiation. As shown in Fig. 5A, all the Set 1 rpl33a mutants showed an appreciable decrease in the P/M ratio and reduced polysome abundance, in comparison with the WT at 28 °C, suggesting that these rpl33a mutations reduce the rate of bulk translation initiation. The pool of free 40S ribosomal subunits is elevated and the levels of free 60S subunits is severely reduced in all the mutants, concomitant with the presence of "halfmer" shoulders on the monosome (80S) and disome peaks of the polysome profiles. The appearance of halfmers, representing mRNAs with one or more 80S ribosomes plus a single 43S or 48S PIC is characteristic of a reduced level of 60S ribosomal subunits, resulting in a delay in the 60S subunit joining reaction at the AUG codon. The reduction in P/M ratio was strongest in the rpl33a-G69R, G76R-G79R and ∆99-107 mutants and comparable to that of the rpl33a∆ mutant. In contrast, rpl33a-102R and the two blocks of Ala substitutions in residues of the C-terminal region of eL33 (A102-105 and A102-107) produced only a ~ 20% reduction in the P/M ratios whereas G76R and Y44R decreased the P/M ratio by ~ 30-40%. The hc rpl33a-Y44R mutant showed an increase in both the P/M ratio (~ 40%) and the polysome content when compared with the lc rpl33a-Y44R mutant, which is in accordance with the partial suppression of the Slg − phenotype and the modest increase in the levels of 25S rRNA seen above on overexpressing this variant. The decrease in the P/M ratios of polysomes in the Set 1 rpl33a mutants correlated with the intensity of the Slg − phenotypes displayed by these mutants at 28 °C (Fig. 2C, left panels and Table 7), suggesting that this phenotype resulted, at least in part, from a reduction in the amounts of 60S subunits that generally reduces the initiation rate of protein synthesis.
We also analyzed total polysome profiles from WT and selected Set 2 rpl33a mutants. At 28 °C, we observed in the rpl33a-A40-44 mutant an increase in free 40S ribosomal subunits, a decrease in free 60S subunits, 80S monosomes and polysomes, and the presence of halfmers (Fig. 5B), all consistent with the ~ 25% reduction in the levels of 25S rRNA revealed by this mutant at the permissive temperature ( Fig. 4A and Table 4). The P/M ratio of A40-44 was very similar to that of the WT; however, the proportion of > 4-mer polysomes was somewhat reduced (~ 50% of WT), suggesting a modest reduction in the translation initiation rate at the permissive temperature. After 6 h of incubation at 37 °C, we observed accumulation of free 40S subunits accompanied by a marked reduction in the amount of free 60S subunits and the appearance of halfmers in all five mutants. Notably, the levels of free 40S subunits and those of 80S monosomes are quite similar in the rpl33a-V35R, Y103R, and A40-44 mutants (Fig. 5C). The P/M ratios are reduced in all the mutants compared with the WT, as is the abundance of 80S ribosomes and polysomes. The rpl33a-A40-44 mutant showed a P/M decrease of ~ 50%, the largest reduction in P/M ratio among the Set 2 mutants analyzed, whereas rpl33a-∆L29 displays only a slight reduction of ~ 15% in the P/M ratio. Interestingly, the P/M ratio is lower in the rpl33a-L7R versus V35R and Y103R mutants, although the defects in ribosome biogenesis were more severe in the latter mutants, thus indicating that L7R elicits a relatively stronger defect in translation initiation beyond its deleterious effect on ribosome assembly.

Set 1 rpl33a mutants exhibit Gcd − phenotypes
As shown in Fig. 2A, B, the Set 1 rpl33a mutations conferred 3AT R phenotypes in a gcn2 gcn3 rpl33aΔ strain, and increased HIS4-lacZ expression under amino acid sufficiency conditions (Gcd − phenotypes). To obtain direct evidence that the Set 1 rpl33a mutations cause derepression of GCN4 at the translational level, we measured β-galactosidase activities from the GCN4-lacZ reporter on plasmid p180, which contains the WT GCN4 mRNA leader with the four regulatory uORFs (Fig. 6A), in cells grown at 28ºC. To evaluate the effects of the mutations on translational control by the uORFs, the increases in expression from p180 were normalized with the values obtained from measuring the expression of the control GCN4-lacZ reporter on plasmid p227 lacking the four uORFs (Fig. 6B). Modest but significant increases of 1.5-2.4-fold in normalized GCN4-lacZ expression were observed for all the rpl33a mutants compared with the WT, indicating that all the mutants diminished translational repression by the uORFs. The rpl33a-L102R, A102-105 and A102-107 mutants exhibited greater normalized derepression of WT GCN4-lacZ expression than did the other Set 1 rpl33a mutants.
It has been previously described that "leaky scanning of uORF4" by fully assembled PICs elicits the Gcd − phenotypes of the rpl16b∆ (Foiani et al. 1991), rpl33a-G76R and rpl33a∆ mutations (Martin-Marcos et al. 2007). Thus, we investigated whether other rpl33a mutations elevate Fig. 5 Substitutions in the yeast ribosomal protein eL33A result in a deficit of 60S subunits and accumulation of halfmer polysomes. A Derivatives of strain Hm700 (his4-301 rpl33a∆ RPL33B) containing different RPL33A alleles on lc or hc plasmids or an empty vector as indicated (Set 1 mutants), were cultured in liquid SC-Leu at 28 °C to A 600 of ~ 1.0. Cycloheximide was added at 100 µg/ ml before harvesting the cells and WCE were prepared in the presence of 30 mM Mg 2+ . 10 A 260 of each extract were resolved on 7-50% sucrose gradients and analyzed by continuous monitoring of A 254 . In the left top panel, peaks representing free-ribosomal 40S and 60S subunits, 80S monosomes and polysomes are indicated. Mean Polysome/Monosome (P/M) ratios were calculated from two or three independent experiments. B-C WCEs of derivatives of strain Hm701(his4-301 rpl33a∆ rpl33b∆ containing the indicated RPL33A alleles (Set 2 mutants) were obtained and analyzed as in Fig. 5A from cells cultured in SC-Leu at 28 °C (B) or shifted to 37ºC for 6 h (C) ◂ GCN4-lacZ expression from the reporter in p226 containing uORF4 alone (Fig. 6A) (Mueller and Hinnebusch 1986). As shown in Fig. 6B, rpl33a mutants produced increases of twoto threefold in GCN4-lacZ expression from plasmid p226 compared to that seen in WT cells, after normalizing with the values of the expression of GCN4-lacZ from plasmid p227. As this effect could also arise from increased reinitiation after termination at uORF4, we measured GCN4-lacZ expression from the pM226 construct containing a single elongated version of uORF1, which overlaps the beginning of GCN4 and is incompatible with reinitiation downstream (Fig. 6A) (Grant et al. 1994). Thus, only mutations that cause leaky scanning of uORF1 would increase GCN4-lacZ expression from pM226. All Set 1 rpl33a mutants showed a 2-4-fold increase in normalized GCN4-lacZ expression compared to WT (Fig. 6B). These data indicate that the Gcd − phenotype conferred by some mutations in rpl33a could arise from leaky scanning of uORF4 by 48S PICs that continue scanning, and on reaching the GCN4 AUG codon, are able to join a 60S ribosomal subunit and initiate translation, thus elevating the expression of GCN4-lacZ from p180 (Gcd − phenotype). It is possible that the modest (< twofold) derepression of the WT GCN4-lacZ reporter by rpl33a-Y44R, G69R, G76R, and G79R-G76R reflects leaky scanning of the GCN4 AUG codon as well, offsetting increased leaky scanning of uORF4. In agreement with this idea, levels of GCN4-lacZ expression from the pM226 reporter were relatively higher in the rpl33a-Y44R, G69R, G76R, and G79R-G76R mutants, indicating that leaky scanning of AUGs by fully assembled PICs occurs more frequently in the latter mutants than in rpl33a-L102R, A102-105 and A102-107 (Fig. 6B).
Mutations that impair derepression of GCN4 translation under starvation conditions confer sensitivity to inhibitors of amino acid biosynthesis, including sulfometuron (SM), which inhibits an isoleucine/valine biosynthetic enzyme, leading to a Gcn − phenotype (non-derepression of genes induced by Gcn4 under starvation). It has been reported before that an rpl16b∆ mutant strain exhibits a Gcn − phenotype resulting from leaky scanning of the GCN4 AUG codon under starvation conditions (Foiani et al. 1991). To investigate whether rpl33a mutations provoked Gcn − phenotypes under starvation conditions, we measured β-galactosidase activities of GCN4-lacZ from p180 in cells grown in minimal medium containing 0.5 μg/ml of SM for 6 h at 28 °C. rpl33a Set 1 mutants showed modest reductions in GCN4-lacZ expression in the presence of SM compared to WT RPL33A cells (Fig. 6B, p180 SM), consistent with a moderate leaky scanning of the GCN4 AUG codon under starvation conditions. Leaky scanning of uORF1 could also contribute to the Gcn − phenotype of these mutants, given that translation of uORF1 is required for reinitiation at the GCN4 AUG codon under starvation conditions.
We observed a ~ twofold increase in GCN4-lacZ expression in the Hm700 strain carrying a LEU2 empty vector in comparison when assayed in the absence of an empty vector ( Fig. 6B (vector vs rpl33a∆*). This also occurs even in the WT parental strain H2994 (data not shown), suggesting that the presence of a LEU2 empty vector could affect the β-galactosidase assays or interferes with GCN4 translational regulation, at least in this genetic background.
β-galactosidase assays revealed that the expression levels of GCN4-lacZ reporters in hc rpl33a mutants were very similar to those of the corresponding lc rpl33a mutants (data not shown), except for rpl33a-Y44R that showed a marked reduction of normalized GCN4-lacZ expression from plasmids p180 (~ 50%), p226 (~ 30%) and pM226 (~ 40%) compared with sc rpl33a-Y44R (Fig. 6B). These data are consistent with the partial suppression of the Slg − phenotype, the increase in the amounts of 25S rRNA, 60S subunits and in polysome content observed in the hc rpl33a-Y44R mutant.
Although Set 2 rpl33a mutants did not exhibit 3AT R phenotypes in a gcn2 gcn3 rpl33a∆ strain, we wished to check if it was the presence of the RPL33B paralogous gene the reason of attenuating a possible effect of these mutations on GCN4 translational regulation. However, measuring expression of GCN4-lacZ reporters in rpl33a∆ rpl33b∆ his4-301 strains containing several Set 2 rpl33a mutants revealed a weak Gcd − phenotype only for the rpl33a-A40-44 mutant (Fig. 6B).

Several Set 1 rpl33a mutations increase the UUG/ AUG initiation ratio and confer Sui − phenotypes
As described above, none of the rpl33a mutations analyzed in this work restored growth in medium lacking His in the parental strain containing the his4-301 mutation, suggesting the absence of strong Sui − defects that elevate initiation at the UUG start codon of his4-301. In contrast, several Ssu − mutations that suppress initiation at UUG codons, and enhance the accuracy of start codon recognition, were identified in the 25S rRNA of the yeast 60S ribosomal subunit (Hiraishi et al. 2013). Similar to what occurs in rpl33a mutants, those 25S mutations increase the expression of a GCN4-lacZ fusion from plasmid pM226 by leaky scanning of the uORF1 start codon (Hiraishi et al. 2013). To determine whether rpl33a mutations behave in a similar manner and suppress initiation at UUG codons, evoking Ssu − phenotypes, we examined whether they could suppress the Sui − phenotype of the dominant eIF2β Sui − allele SUI3-2 (Huang et al. 1997). To that end, we measured β-galactosidase activities from matched HIS4-lacZ reporters containing an AUG or UUG as start codons in strains carrying SUI3-2 and either WT RPL33A, Set 1 rpl33a mutations or an empty vector. As expected (Huang et al. 1997), SUI3-2 increases the ratio of expression of the UUG to AUG HIS4-lacZ fusion ~ 5.2-fold compared to the otherwise WT strain (Fig. 7A). Surprisingly, rpl33a∆ and eight rpl33a mutations did not lower the high UUG/AUG ratio conferred by SUI3-2 in a WT strain but five increased it between ~ 1.25 and ~ 2.0-fold (Fig. 7A) Fig. 6 Comparison of the Gcd − and Gcn − phenotypes conferred by mutations in rpl33a. A Schematic depiction of GCN4-lacZ reporters containing the four uORFs in the leader of GCN4 (p180), only uORF4 (p226), an elongated version of uORF1 overlapping with the beginning of GCN4 (pM226) and GCN4 without uORFs (p227). B Strain Hm700 (his4-301 rpl33a∆ RPL33B), marked as rpl33a∆*, derivatives of strain Hm700 containing the indicated RPL33A alleles on lc plasmids (upper panel) or on hc plasmids, or an empty vector (middle panel), and derivatives of strain Hm701(his4-301 rpl33a∆ rpl33b∆ containing the indicated RPL33A alleles (lower panel), were transformed with GCN4-lacZ fusions on plasmids p180, p226, pM226 and p227 and cultured in liquid SD + His + Trp, or for 6 h in SD + His + Trp containing 0.5 μg/ml sulfometuron (SM) at 28 °C to A 600 of ~ 1.0 and assayed for β-galactosidase activities as in Fig. 2A. The means and SE from at least four independent transformants are reported. The values highlighted in bold in the right columns are the results normalized to correct for the different expression of p227 and relative to the corresponding WT value (1.0) at least five rpl33a mutants significantly enhance the Sui − phenotype conferred by SUI3-2. Although none of the Set 1 rpl33a mutations conferred His + phenotypes, it was previously reported that certain Sui − mutations also displayed no detectable growth in a medium without His, i.e. Arg-36 or Lys-37 substitutions in eIF1 (Martin-Marcos et al. 2013); therefore, we examined whether Set 1rpl33a mutations increased the UUG/AUG initiation ratio when they are examined in the absence of another Sui − mutation. Except for rpl33a-G69R and rpl33a-∆99-107, we observed small but significant increases in UUG/AUG initiation for all the analyzed rpl33a substitutions (Fig. 7B), indicating that they conferred weak Sui − phenotypes. Fig. 7 Set 1 rpl33a mutations increase initiation UUG/AUG initiation ratio. (A) Derivatives of strain H700 (his4-301 rpl33a∆ RPL33B) containing the indicated RPL33A alleles or an empty vector and episomal SUI3-2 (p4280/ YCpSUI3-S264Y-W) or empty vector (WT) also harboring HIS4-lacZ reporters with AUG or UUG start codons (plasmids p367 and p391, respectively) were cultured in SD + His at 28 °C to A 600 of ~ 1.0, and β-galactosidase activities were measured in WCEs as in Fig. 2B. B Derivatives of strain Hm700 containing the indicated RPL33A alleles or an empty vector also harboring HIS4-lacZ reporters on plasmids p367 and p391 were cultured in SD + His + Trp at 28 °C to A 600 of ~ 1.0, and β-galactosidase activities were measured in WCEs as in Fig. 2B. C Transformants of strain Hm700 containing the indicated RPL33A alleles or an empty vector and hc empty TRP1 vector (hc vector) or SUI1 (encoding eIF1) on a hc plasmid (hc eIF1) were grown in SD + His at 28 °C to A 600 of ~ 1.0, and β-galactosidase activities were measured in WCEs as in Fig. 2B  eIF1 blocks accommodation of Met-tRNA i in 48S PICs clashing with Met-tRNA i in the P IN conformation, such that dissociation of eIF1 from the 40S subunit is required for start codon recognition (Hussain et al. 2014;Rabl et al. 2011). This clash can be efficiently overcome with the perfect codon-anticodon duplex formed by Met-tRNAi at AUG codons, but not with the mismatches formed at near-cognate codons like UUG, thus discriminating against the latter. Consistent with this, it was previously shown that overexpression of eIF1 suppresses the increased UUG initiation produced by Sui − mutations in different initiation factors, including the SUI3-2 mutation in eIF2β (Valasek et al. 2004), conferring the suppressor Ssu − phenotype. Hence, we investigated whether the weak Sui − phenotype of rpl33a mutants could be overcome by overexpressing eIF1; and indeed, hc SUI1 (encoding eIF1) reduced the UUG/AUG initiation ratios of rpl33a Sui − mutations to those of WT levels (Fig. 7C). These data support the notion that the 60S subunit has a role in accurate recognition of the AUG start codon, and that eL33A alterations or the lack of eL33A increase aberrant utilization of near-cognate UUG start codons, most likely by allowing inappropriate dissociation of eIF1 at Met-tRNAi:UUG mismatches.

Discussion
In this study, we generated a large array of mutations altering different structural elements or residues of ribosomal protein eL33 with the aim of investigating which are required for the proper function of this protein in ribosome biogenesis, mRNA translation, and recognition of the AUG start codon. Mutations assigned to Set 1 conferred strong Slg − phenotypes at both permissive (28 °C) and restrictive temperatures (37 °C and 18 °C), and modest Gcd − and Sui − phenotypes, whereas Set 2 mutations conferred Slg − phenotypes alone, and only at the restrictive temperature of 37 °C.
Except for a few Set 2 mutants at 37ºC, all the remaining mutant alleles produce levels of rpl33a mRNA higher or similar those of the WT, suggesting that the mutant proteins are well expressed. Only in the case of rpl33a-Y44R, overexpression of the mutant allele in a hc plasmid conferred a partial suppression of the Slg − phenotype, reduced the P/M ratio and diminished amounts of 25S rRNA and 60S subunits observed in the corresponding lc mutant, suggesting that only this variant is under-expressed compared to WT eL33A. However, we cannot discard the possibility that some eL33A mutant proteins are expressed at lower than WT levels owing to reduced stability versus WT eL33A, and their recruitment or the stability of the corresponding assembled 60S subunits could also be diminished, all of which could contribute to both the general translation initiation defects and other phenotypes (i.e., rRNA processing and subunit biogenesis) observed in these mutants.
We showed previously that the rpl33a-G76R mutation impairs efficient processing of 35S, 27SA2 and 27S pre-rRNAs, leading to substantial reductions in the steady-state levels of 25S, 18S and 5.8S rRNAs (Martin-Marcos et al. 2007). It was also reported that in vivo depletion of eL33 inhibited formation of the major 5ʹ end of 5.8S rRNA (B1S site in yeast) and the endonucleolytic cleavage at site C2 (Poll et al. 2009). In accordance with previous results, our Northern blot analysis revealed that Set 1 rpl33a mutations impair early cleavages at A0, A1 and A2 in the 35S pre-rRNA with attendant reduced steady-state levels of the resulting subsequent precursors in the pathway. This phenotype has been described before for several mutants of genes encoding 60S r-proteins (i.e. Babiano and de la Cruz 2010;Espinar-Marchena et al. 2016, 2018Jakovljevic et al. 2012;Martin-Marcos et al. 2007;Poll et al. 2009;Rosado et al. 2007;van Beekvelt et al. 2001)). Substantial reductions in the amounts of 33-32S, 27SA 2 , 27S, 7S and 20S pre-rRNAs were found in all the rpl33a mutants, indicating a pronounced destabilization of early and intermediate precursors with the concomitant reduced production of 25S, 18S and 5.8S mature rRNAs. Additional defects in 27S pre-rRNA species processing were seen in all Set 1 rpl33a mutants which led to stronger reductions in the steady-state levels of 25S rRNA than in those of 18S rRNA. These results are in accordance with previous findings showing that eL33 is assembled very early into pre-60S ribosomal particles (Gamalinda et al. 2014;Ohmayer et al. 2013).
The rpl33a-G76R substitution could affect the interaction of eL33 with residues of 60S subunit rRNA domain II (Fig. S1-A), and Set 1 mutations A102-105 and A102-107 in the C-terminal domain of eL33 could alter contacts of the protein with r-protein eL6 and with expansion segment ES7 of 25S rRNA (Fig. S1 B-C). In general, mutant A102-107 displayed stronger defects than those of A102-105 which could be attributed to the fact that A102-107 affects more contacts with ES7 and eL6 in the 60S ribosomal structure than does A102-105 (Fig. S1 B-C). The strongest Slg − and defective pre-rRNA processing phenotypes are observed in the rpl33a-∆99-107 mutant, which lacks the last nine amino acids at the C-terminus of the protein and behaves almost identically to eliminating the entire RPL33A ORF by rpl33a∆, thus indicating that the C-terminal domain of eL33 is essential for proper pre-rRNA processing and maturation, and ribosome assembly. The ∆99-107 truncation would impede interactions of eL33 with both the ES7 and the ES39 (Figs. S1-C and S2-A) as well as with residues of eL6 (Fig. S1-B). The ∆99-107 truncation of human rpL35A, which removes Arg-102 to Ile-110, was identified by sequence analysis of a cohort of DBA probands (Farrar et al. 2008). The results obtained in this work suggest that a nearly complete loss of rpL35 function caused by the ∆99-107 truncation could contribute to the disease phenotypes. rpl33a-Y44R and L102R substitutions could also disturb contacts between r-proteins eL33 and eL6 (Fig. S2-B).
Five Set 2 mutations, rpl33a-L7R, V35R, F43R, Y103R and A40-44 could perturb different contacts of eL33 with r-protein eL6 (Fig. S3 A-C), whereas ∆L29, the equivalent in yeast of the ∆L27 mutation in human rpL35A/eL33 identified in DBA probands (Farrar et al. 2008), could impair interaction with 25S rRNA domain II (Fig. S4-A). The Set 2 rpl33a-A104-107 mutation could alter interactions with both the ES7 and r-protein eL6, as Set 1 rpl33a substitutions A102-105 and A102-107, altering the C-terminal domain of eL33 (Fig. S1 B-C). However, the effects on growth and pre-rRNA processing caused by A102-105 and A102-107 are relatively more pronounced. The A102-105 includes Leu-102 and Tyr-103, which establish several interactions with eL6 and were not substituted by Set 2 mutant A104-107. In fact, the rpl33a-L102R single point mutation alone caused Slg − and strong defects in pre-rRNA processing at 28 °C. In addition, the A102-107 mutation is larger and encompasses A104-107. In contrast, the Set 2 rpl33a-A92-99 mutation could disturb interactions of eL33 with several residues of the 25S rRNA domain I and the ES39 (Fig. S4-B), but only produced slight Ts − phenotype at 37 °C and a mild decrease in levels of mature rRNAs.
In summary, several Set 1 and Set 2 rpl33a mutations would alter contacts of eL33 with several residues of eL6, nucleotides of 25S rRNA domains I and II, and the ES7 and ES39. The loss of these connections could impair the formation of the dII/dVI ribosomal protein cluster, with attendant defects in pre-rRNA processing, destabilization of the pre-ribosomal 60S particles and inhibition of recruitment of ribosomal proteins and ribosome biogenesis factors required for further steps in ribosome assembly and pre-rRNAs maturation. The ultimate consequence of the loss of function of eL33 in pre-rRNA processing and ribosome maturation would be a reduction in the assembly of mature 60S subunits and/or the production of defective 60S subunits that are not fully functional in translation.
Total polysome profiles revealed that rpl33a mutations caused defects in the general translation of mRNAs.
Eight Set 1 rpl33a mutants ( Fig. 5A and Table 7) showed a decrease in both the P/M ratio and polysome content, suggesting that those mutations produce defects in translation initiation that are more rate-limiting than possible reductions in the rate of elongation.
The reduction in P/M ratio is strongest in rpl33a-G69R, G76R-G79R and ∆99-107 and is comparable to that of rpl33a∆. We observed a drastic shortage in the amount of free 60S ribosomal subunits and the appearance of halfmers, indicating a diminished rate of 60S subunit joining near the completion of initiation. These polysome assembly defects correlated well with the Slg − phenotypes displayed by the group of Set 1 rpl33a mutants at 28 °C.
One question to address is whether eL33A mutant proteins are efficiently assembled into ribosomes and contribute to the defects in 60S subunit joining. The rpl33a mutant alleles were expressed in strain Hm700 containing a chromosomal deletion of RPL33A but an intact RPL33B allele that is expressed to a much lower level than its paralogous gene RPL33A (Martin-Marcos et al. 2007;Tornow and Santangelo 1994). Thus, in Hm700 transformed with empty vector, only a small amount of WT 60S ribosomal subunits, all containing eL33B, would be synthesized. In fact, we have observed that the steady-state amounts of the RPL33B mRNA do not increase in rpl33a∆ cells (data not shown). The lack of eL33A protein produces severe defects in pre-rRNA processing and ribosome assembly, which leads to a pronounced deficit in 60S ribosomal subunits and consequent defects in general mRNA translation responsible of the strong Slg − phenotype of the rpl33a∆ mutant. As all Set 1 rpl33a mutants, except ∆99-107, grow better than rpl33a∆, it appears that at least a fraction of the eL33A mutant proteins is correctly assembled into mutant 60S ribosomal subunits that are partially functional in translation. It is difficult to say to what extent the reductions in translation and growth conferred by these mutations, as well as the more severe rpl33a-∆99-107 mutation, arise from reduced 60S function versus reduced 60S biogenesis/abundance.
At 28 °C, the Set 2 rpl33a mutant A40-44 revealed moderate defects in polysome abundance but the P/M ratio remained practically unaffected, suggesting that translation initiation is efficient at this temperature. However, after incubation at 37 °C, the P/M ratio showed a marked decrease of ~ 50% and the polysome defects were more accentuated, which is consistent with the strong Ts − phenotype showed by this mutant (the eL33a-A40-44 protein could be unstable at 37 °C). Polysome profiles of Set 2 rpl33a mutants at 37 °C revealed accumulation of free 40S subunits, reduction of free 60S subunits, the appearance of halfmers and a decreased P/M ratio which would account for their Ts − phenotypes.
Set 1 rpl33a mutants exhibit modest 3AT R /Gcd − phenotypes that could be explained by leaky scanning of uORF4 by PICs reaching it due to the lack of 60S ribosomal subunits. In one hand, the leaky scanning of uORF1 would have a negative effect in the derepression of GCN4 translation (which would produce a Gcn − phenotype), because the initiation by the ribosomes at the uORF1 (which has a positive effect on GCN4 synthesis) is a necessary requirement for that derepression, and this event can be reduced in mutants with a strong deficit of 60S subunits. However, leaky scanning would also occur at the uORF4 (which has a strong negative effect on GCN4 synthesis), favoring initiation events downstream, so that PICs would continue scanning and reach the AUG initiation codon of GCN4. Then, they could join a 60S ribosomal subunit and translate GCN4 (Gcd − phenotype), or bypass the AUG codon, depending on the amount of available 60S subunits. Thus, mutants with a lesser deficit in 60S subunits (rpl33a-L102R, A102-105 and A102-107, Fig. 3B, Table 4 and Fig. 5A) will produce relatively greater Gcd − phenotypes (Fig. 6B) than those mutants with a stronger deficit in 60S subunits. However, we cannot discard the possibility that some of the Set1 rpl33a mutations affect the functionality of the mature 60S subunits and alter initiation events at the four uORFs and/or at the AUG initiation codon of GCN4.
Assaying the expression of matched HIS4-lacZ reporters containing an AUG or UUG as start codon in rpl33a Set 1 mutants revealed small but significant increases in the UUG/AUG initiation ratio, which were generally of lesser magnitude than those given by other Sui − mutations in various initiation factors (i.e. Alone et al. 2008;Fekete et al. 2007;Huang et al. 1997;Martin-Marcos et al. 2011;Saini et al. 2010;Valasek et al. 2004). Nevertheless, the elevated UUG/AUG ratios were fully suppressed by overexpressing eIF1, indicating a bona fide decrease in the fidelity of start codon selection in the rpl33a mutants. It is possible that rpl33a mutations and the lack of RPL33A increase the UUG/ AUG initiation ratio, given that some PICs are paused in the first AUG of mRNAs (as indicated by the presence of halfmers in the polysome profiles) awaiting junction with a 60S ribosomal subunit. Thus, both 40S subunits stalled at AUG codons and the reduced amounts of 60S subunits showed in different rpl33a mutants would impair translation initiation and contribute to increase the proportion of initiation events at UUG codons relative to AUG canonical initiation. It can also occur that some PICs leaky scan the AUG codons-as seen in the mRNA of GCN4-which could also contribute to decrease general translation initiation at AUG codons. This would explain the slight Sui − phenotype conferred by certain rpl33a mutations compared with stronger phenotypes observed in Sui − mutants with alterations in different initiation factors. It was shown before that in cells carrying an rpl16b mutant allele ribosomes initiate translation at the non-canonical codon AUA, and such initiation may be enhanced because of the shortage of 60S subunits in this strain (Moritz et al. 1991). But it is also possible that the amounts of certain initiation factors could be reduced in rpl33a mutants that showed severe defects in general translation of mRNAs, and thus conferred the modest Sui − phenotype observable in these mutants.
We found that the increased UUG/AUG initiation ratios observed in some Set 1 rpl33a mutants were suppressed by eIF1 overexpression. Increasing the abundance of eIF1 that monitor codon-anticodon interactions during translation initiation would prevent its release from the PICs, which occurs at a higher frequency at UUG codons in Sui − mutants, and thereby the PICs continue scanning downstream without initiating at UUG codons. Thus, the overexpression of eIF1 increases the accuracy of start codon recognition and suppresses the Sui − phenotype caused by rpl33a mutations.
One hypothesis to explain tissue-specific effects observed in some human ribosomopathies caused by single-copy mutations in specific ribosomal proteins (i.e. DBA probands carrying rpL35A mutations, (Farrar et al. 2008)) is that ribosome deficiency or dysfunction can affect global and mRNAspecific translational control, and that certain cells-like bone marrow-derived cell lineages and skeletal tissues-are more vulnerable than others to those defects (reviewed by (Mills and Green 2017)). In this model, mRNAs are variably dependent on cellular ribosome concentration, with more poorly initiated mRNAs being relatively more sensitive to perturbations in ribosome concentrations or function. Ribosomal subunit deficiencies could explain differences in the intensities of the yeast eL33A mutant phenotypes analyzed in this work (growth rates, thermo-sensitiveness, global translation rates, GCN4-mRNA translation), depending on how much each type of mutation reduces the 60S amounts. Initiation events at the specific AUGs of the 4 uORFs and/or at the canonical AUG of GCN4 could be also differentially affected by each type of eL33A mutation in the rpl33a∆ RPL33B genetic background of Set 1 mutants. The same explanation can be invoked to explain the Sui − phenotype observed in some rpl33a mutants.
A second hypothesis proposes that ribosome heterogeneity could affect global mRNA translation and would be critical to the translation of specific mRNAs. Ribosome variants result from the incorporation of r-protein paralogs (Komili et al. 2007) and evidence supporting ribosome heterogeneity and its regulation is very abundant in bacteria and yeast (Gilbert 2011). Moreover, it has been reported that heterogeneous ribosomes preferentially translate distinct sub-pools of mRNA genome-wide (Shi et al. 2017). Some of the strongest rpl33a mutations generated in this work are only viable in presence of the eL33B -protein paralog producing Slg − , Gcd − and Sui − phenotypes observable at 28ºC in the rpl33a∆, RPL33B genetic background. The yeast A and B protein paralogs only differ in one amino acid at position (40-D in eL33A, E in eL33B) in the eL33 protein sequence, and the B protein is much less abundant than the A (Martin-Marcos et al. 2007;Tornow and Santangelo 1994). In these Set 1 mutant cells, 60S-B wild-type subunits would coexist in the cytoplasm with 60S-a mutant subunits, providing an example of "ribosome heterogeneity". Thus, the ribosomal subunit balance and "homogeneity" regarding the two paralogs of eL33 that would exist in WT cells will be differently altered in the eL33A mutants, likely depending on the a-mutant ribosomal protein assembly efficiency into 60Ss that would lead to distinct a/B 60S imbalances.
Mutated eL33A proteins expressed from lc or hc plasmids must be necessarily assembled into pre-60S subunits, leading to enough amounts of mature and at least partially functional 60S subunits to support viability of Set 2 mutants in the rpl33a∆ rpl33b∆ genetic background. These Set 2 mutants would carry less severe rpl33a mutations than those of Set 1 and accordingly they exhibit phenotypes mostly observable at 37ºC and not at 28ºC. Moreover, as for eL33A/ eL33B wild-type cells, ribosome heterogeneity cannot be invoked either in the Set 2 mutants.
It would be possible that disrupted molecular interactions of eL33A with domains and ESs of the 25S rRNA or with other RPLs in the 60S structure, may affect in different manners, or to several extents the fidelity of translation initiation, depending on the specific rpl33a mutation and the types and structural features of the mRNAs. Perturbations related to ribosome recycling and rescue could also affect the rates of global translation in a yeast cell, as well as the translational control of specific mRNAs and the expression of important proteins, like the transcription factor GCN4.
Our data indicate that mutant versions of the eL33A protein may have different defects on its own assembly into 60S-preribosomes, altering the pre-rRNA maturation rates and molecular interactions with rRNAs or RPLs in the 60S structure, and affecting at different extents the amounts and/or functionality of mature 60S subunits, leading to impaired mRNA translational efficiencies and/or in the fidelity of translation initiation.