Rqc1 and other yeast proteins containing highly positively charged sequences are not targets of the RQC complex

Our first step was to find endogenous yeast proteins containing sequences with different arrangements of polybasic sequences. We selected four distinct but possibly overlapping architectures, covering most of polybasic sequence variations analyzed in previous studies: (a) sequences of eight or more lysines/arginines residues (K/R) in a window of ten amino acids, (b) six or more consecutive K/R, (c) adenine repeats longer than ten nucleotide residues, and (d) net charge ≥+12 in a window of 30 residues. The architectures a) and b) were selected because they are often present in RQC activation reporters and lead to the accumulation of reads in ribosome profiling data (, , , , , , , , , , , , , , , , ). The architecture c) was selected because poly (A) mRNA composition can lead to ribosome sliding, worsening the translation efficiency of these sequences (, ). Architecture d) was selected because proteins containing a stretch of 30 residues (approximately the polypeptide length enclosed by the ribosome exit tunnel) with net charge ≥ +12 are extremely rare in most proteomes (, ). We also compared these architectures with a group of 1800 genes containing ICPs that are known to impair translation with substantial downregulation in protein output (e.g., CGA-CGA and CGA-CCG) (). It is important to note that the ICP group, used herein as an endogenous positive control for ribosome stalling, has some potential pitfalls. For example, the sequence CGA-CCG-A, captured in our list of ICP as CGA-CCG (Table S1), likely induces frameshifting (). Even though the pair CGA-CCG is present in just 1% of the genes in our list, we cannot rule out that other ICP arrangements might cause frameshifting and lead to poor translation scores by mechanisms unrelated to ribosome stalling.

We found that 354 proteins (approximately 6% of the yeast proteome) contained at least one of the characteristics described above (Fig. 1A). The features most commonly found were high net charges and poly (A) sequences, with 170 and 121 sequences, respectively. Moreover, the Venn diagram (Fig. 1A) shows that just only 91 proteins have more than one feature at the same time, and only one protein, the uncharacterized YNL143C, has all the characteristics. Localization analysis revealed that no bias was found for the different categories analyzed (Fig. 1B). We selected YNL143C and other proteins with prominent polybasic sequences (Fig. 1, C–E) for further experimental analysis in the next sections.

If the presence of polybasic sequences represents an important obstacle to translation, we expect to find signatures of poor translatability in proteins that contain them. For example, endogenous genes containing stall sequences have a slower or less efficient translation initiation; this is thought to be an adaptation to avoid ribosome collision during elongation (). Therefore, we asked whether the polybasic and the ICP data sets would be significantly different from the rest of the yeast genes in previously published experimental measurements that reflect translation activity. We chose six parameters that indicate how well a transcript is translated: translation efficiency (TE), codon stabilization coefficient (CSC), time for translation initiation, Kozak score, Kozak score 50 nt, and 5’ UTR footprint density. The TE is a metric derived from ribosome profiling data, calculated from the ratio of ribosome-protected footprint read counts to the total mRNA read counts for each gene (). The CSC is a metric derived from mRNA half-life that positively correlates with TE and protein abundance (). The time for translation initiation is calculated from ribosome profiling density, relative mRNA concentration, and time of translation for individual codons (). The Kozak score is determined by reporter expression quantification from a library containing synthetic 5 nt sequences cloned upstream the reporter’s start codon, while the Kozak score 50 nt is determined by reporter expression quantification from a library containing 5’ 50 nt UTR sequences from yeast genes cloned upstream the reporter’s start codon (). Finally, the 5’ UTR footprint density is determined by ribosome profiling reads detected at 40S ribosomal subunits (). To make this analysis, we used publicly available data sets generated by previous studies (, , , , ). The Table S3 shows how these metrics correlate with each other for the entire genome data set. TE had a positive correlation with CSC and both Kozak scores and negative correlation with time for translation initiation and 40S 5’ UTR densities. Moreover, even though we are using different ribosome profiling experiments to obtain the TE and time for initiation values, we still observed the expected positive correlation between these parameters (Table S3).

In Figure 2A, we compare the distribution obtained with the entire genome with the distribution of values obtained with the ICP and polybasic gene groups. Consistent with the pattern expected for hard-to-translate sequences, the ICP group showed higher initiation times, higher 40S 5’ UTR densities, and lower TE, CSC, and Kozak scores. All these differences were supported by a t-test analysis comparing the individual data sets with the rest of the yeast genes (p-values are shown in Fig. 2B). For the polybasic group, the only differences with a significant p-value were lower TE and CSC values than the genome group. Separating the polybasic sequences into subgroups did not reveal a particularly deleterious architecture of polybasic sequence arrangement (Fig. 2B). The cumulative distributions for each of the measurements described above are shown in Fig. S1.

To evaluate the characteristics of each gene from the polybasic group, we created a unified rank for all the measurements above. Using the values obtained with the entire yeast data set, each score system was organized in ten bins; next, the maximum and minimum bin values were used to normalize the scale. We created a heat map (Fig. 2C, upper panel) where 1 refers to the maximum translation activity (high TE, high CSC) or optimal initiation (short translation initiation time, high Kozak's efficiency, and low 5’UTR footprint density of the 40S ribosomal small subunits). A cluster analysis of the 354 genes with polybasic sequences revealed that the group is rather heterogeneous and displays all spectra of characteristics, with some sequences having excellent translatability indicators with poor initiation scores and vice versa (Fig. 2C). The lower panel indicates the polybasic architecture present in each gene. Notably, the sequences with the highest positively charged sequences or longest poly (A) tracts also showed diverse features (Fig. 2C, lower panel). Therefore, we could not find a clear signature of poor translatability for genes containing polybasic sequences using the analyzed parameters (Fig. 2, C and D).

To test whether the selected polybasic sequences have any effect on translation, we again took advantage of previously published ribosome footprint profiling data (, , , , ), which were used on all subsequent ribosome profiling analyses. (please see Experimental procedures section Ribosome profiling data for detailed information). Analyses of the ribosome profiling (RP) data of the polybasic and ICP sequences indicated an enrichment of 27 to 29 nucleotide long footprints around the stalling sequence (Fig. 3A). We also plotted the number of reads at an approximate position of the ribosomal A site and observed an accumulation of reads downstream the first codon of the polybasic sequence (Fig. S2), which suggests a reduction in translational speed.

Additionally, we looked directly at the presence of ribosome collisions through the analyses of disome profiling. The disome profiling allows sequencing of mRNA fragments protected by two stacked ribosomes and shows stalling patterns undetected by traditional RP (, , ). Disome footprints could be detected as a sharp peak immediately after the ICP sequences (Fig. S3, A–C). The analysis of polybasic sequences also showed an increase in the disome footprints. However, the peak was broader than the one observed with ICPs and occurred 9 to 10 codons after the polybasic-coding nucleotides (Fig. S3, D–F). This observation supports the hypothesis that the arrest is caused by electrostatic interactions of the nascent positively charged polypeptide with the ribosome exit tunnel. Since disomes were detected in most transcripts, it is unlikely that the formation of these disomes alone can trigger the RQC activation (, ).

Recently, it was shown that slow decoding of the second codon of the ICPs results in an empty ribosomal A site, which leads to shorter 20 to 22 nucleotide long footprints in RP experiments and represents slow decoding rates. These shorter footprints correspond to ribosomes in a classical state (or post-state), which are waiting for the arrival of the next tRNA, different from ribosome in a rotated state (or pre-state), which results in longer 27 to 29 footprints (, ). Structural and biochemical evidences demonstrated that the electrostatic repulsion caused by the presence of a polybasic sequence inside the exit tunnel can decrease the binding efficiency of an incoming tRNA carrying an additional positively charged residue, which could also lead to empty A site ribosomes (, , ). To evaluate if the polybasic sequences of endogenous proteins can cause the accumulation of ribosomes in the classical state (open A sites), we looked for short footprints in RP data of these proteins and found no evidence. The ICP was the only group that showed an accumulation of short reads (Fig. 3B). The same result was observed with another RP data set obtained with a different combination of translation inhibitors (cycloheximide (CHX) and anisomycin) (Fig. S4). These results suggest that polybasic sequences can reduce elongation rates (Fig. 3A), but they are not enough to cause severe ribosome stalling leading to the accumulation of open A site ribosomes (Fig. 3B).

Another indicative of severe ribosome stalling is the occurrence of ribosome collisions upstream an empty A site ribosome (, ). If the endogenous polybasic sequences are strong stalling sequences, we should observe not only an increase in short nucleotide footprints corresponding to an empty A site, but also an increase in long nucleotide footprints upstream of the stalling site with a periodicity of roughly ten codons, which corresponds to the collided ribosomes (). Among the 354 genes with one or more polybasic sequences identified in our bioinformatics screening, we selected the four top genes (Fig. 1, C–E) for subsequent analyses: YBR054W/YRO2 (net charge = +17), YGL078C/DBP3 (net charge = +16), YHR131C (14 consecutive arginines), and YNL143C (32 consecutive adenines coding for ten lysines) (Figs. 2 and 3). To expand our analyses, we also included genes with the two longest polybasic sequences, stretches with six or more K/R in ten amino acids of the yeast proteome: YLR197W/NOP56 and YOR310C/NOP58 (42 and 44 residues long polybasic sequences, respectively). As a positive control, we included the gene SDD1, which has been recently described as an endogenous target of the RQC complex and shows the ribosome collision footprint pattern described above (). Our analyses revealed that, apart from SDD1, none of the selected genes had the periodic footprint pattern that would indicate ribosome collisions (indicated by dashed lines) (Fig. 3, C–G). A downside of this analysis was the low ribosome profiling coverage of the polybasic genes, and this is the reason why YHR131C and YNL143C were not included in Figure 3, C–G. In an attempt to enhance the analysis sensitivity, we carefully examined all the biological replicates looking for consistent read patterns that could suggest the presence of queued ribosomes. We also checked the ribosome profiling of a hel2Δ strain, where the stalled ribosomes are more stable and produce stronger signals corresponding to the periodic pattern of collided ribosomes than in the wt strain (Figs. S5–S9). SDD1 presented a reproducible pattern across all the replicates and a stronger queued ribosomes signal in the hel2Δ strain (Fig. S5). However, no consistent periodicity or queued ribosomes signal increase in the hel2Δ strain could be observed for the polybasic genes analyzed (Figs. S6–S9). The ribosome collision footprint pattern was also carefully investigated for the remaining 348 proteins with polybasic sequences but, due to the low ribosome profiling coverage, no conclusion could be reached (data not shown).

In conclusion, we observed that the ICP group of genes produced more features associated to ribosome stalling than the polybasic group (Fig. 3 and Fig. S3). Therefore, polybasic sequences may represent a superable obstacle to translation under normal growth conditions, and their presence is not enough to induce RQC activation.

To confirm that polybasic sequences are not sufficient to cause ribosome stalling and translation termination, we examined whether the deletion of RQC components had any effect on the final protein output of our selected proteins, which would imply a regulatory role for the RQC complex on their expression. TAP-tagged versions of the six genes selected from the previous section were obtained from a yeast library containing a chromosomally integrated TAP-TAG at the 3’ region of each gene, thus preserving the natural promoter region and the 5’ UTR of the transcripts (). We chose to delete LTN1, the E3 ubiquitin ligase of the RQC, and ASC1, involved in the modulation of RQC activity because they represent different steps in the RQC pathway and because they have been consistently used in the investigation of RQC mechanism of action (, , , , , , ). Each TAP-TAG strain was used to generate the ltn1Δ or asc1Δ variants, which were confirmed by PCR and western blot (Fig. S10). In Figure 4, we show the protein expression of each gene as probed by anti-TAP western blot in the three different backgrounds and the relative quantification of at least three independent experiments. For each gene, we also show their ribosome profiling (complete sequence, with the polybasic sequence marked in blue) and their TE (depicted as a blue bar relative to total distribution of TE values of the yeast genes). We observed that deletion of LTN1 caused no change in the protein levels of the selected polybasic proteins. Since the TAP-TAG is downstream the putative stalling sequence, the lack of Ltn1 should not stabilize the full-length construct. Although the results from the ltn1Δ strains may have a limited contribution for the understanding of the role of RQC in the regulation of the analyzed genes, it showed that these proteins are not regulated by Ltn1 in an RQC independent pathway, as we observed for Rqc1 (see next section).

The results with asc1Δ revealed no changes in YGL078C/Dbp3, YNL197W/Nop56, and YOR310C/Nop58 protein levels and a downregulation in YBR054W/Yro2, YHR131C, and YNL143C (Fig. 4). This pattern could not be explained by features such as ribosome density on the polybasic site, TE, or initiation time. Asc1 has been linked to the assembly of the translation preinitiation complex, and its deletion impairs translation of a subset of genes, particularly small ORFs and genes linked to mitochondria function (, ). These roles of Asc1 unrelated to RQC could be responsible for the decreased levels since YBR054W/Yro2 protein is localized to the mitochondria and YNL143C protein is encoded by a small ORF.

Nevertheless, all tested proteins differed from what was observed with reporters containing stalling sequences, which had their expression levels upregulated in asc1Δ strains (, , , , , ). Therefore, we conclude that, even though Asc1 positively regulates the expression of some of the proteins, the complete RQC pathway is not actively contributing to the regulation of the endogenous levels of any of the proteins tested.

As described above, the proteins Rqc1 and Sdd1 have been previously reported as targets of the RQC complex. The regulation of Sdd1 showed the canonical pattern expected for a protein regulated by the RQC activity: deletion of LTN1 increased the levels of the truncated N-terminal of Sdd1, while deletions of ASC1 or HEL2 increased the levels of its full-length protein (). On the other hand, the experimental data available on Rqc1 regulation showed increased levels of the full-length construct in an ltn1Δ background, when the expected result would be the stabilization of the N-terminal fragment up to the stalling polybasic region (). This result is not consistent with a model of cotranslational regulation of Rqc1 by the RQC.

To investigate the mechanism of Rqc1 regulation by the RQC, we analyzed Rqc1 expression in the backgrounds ltn1Δ, asc1Δ, rqc2Δ, and hel2Δ. We used a C-terminal TAP-tagged version of Rqc1, meaning that we can only observe an increase in the levels of the full-length peptide. As previously described, the deletion of LTN1 increased the expression levels of Rqc1 but, different from the results reported in Brandman et al., 2012, we observed that the deletion of ASC1, HEL2, or RQC2 had no impact on the full-length Rqc1 expression levels (Fig. 5A). Rqc1 regulation was different from that of GFP-R12-RFP, an artificial reporter with a polybasic sequence known to be a target of the RQC (, , , ). As expected, the reporter presented increased levels of GFP (truncated peptide) when LTN1 was deleted and increased levels of both GFP and RFP (full-length peptide) when ASC1 or HEL2 was deleted (Fig. S11). To test if the TAP-TAG had any influence in our results, we added it to the C-terminal of the construct GFP-R12-RFP (stalling) and GFP-ST6-RFP (nonstalling) (Fig. S12) and measured its expression in wt, ltn1Δ, asc1Δ, and hel2Δ cells (Fig. S13). Our results showed that the presence of the TAP-TAG did not alter the expression of the neither constructs (Fig. S13). The amount of TAP-TAG nonstalling reporter was similar for all strains, while higher levels of TAP-TAG stalling reporter were observed for asc1Δ and hel2Δ strains (Fig. S13).

It is noteworthy that the Rqc1 polybasic sequence (blue arrow) is located at the protein N-terminal, meaning that only the upstream polypeptide fragment should be stabilized in an ltn1Δ strain. Therefore, as previously mentioned, the overexpression of the full-length Rqc1 in the ltn1Δ strain (Fig. 5A and Brandman et al., 2012) is inconsistent with the current model for RQC action. The RQC1 mRNA levels were the same in all tested backgrounds (Fig. 5B). To test if the presence of the C-terminal TAP-TAG influenced RQC1 mRNA levels, we repeated the qPCR analysis with the nontagged wt RQC1, and the same results were observed (Fig. S14). Moreover, ribosome profiling of RQC1 did not show the characteristic footprint pattern of ribosome collisions and accumulation of short fragments (Fig. 5C—blue dashed lines, see also Fig. S15), further indicating that the Rqc1 polybasic sequence does not act as a strong stalling sequence.

Another way to verify whether a transcript is a target of the RQC complex is through the saturation of the RQC system; if the protein of interest is a target, the saturation of the RQC will increase its expression levels. The saturation can be accomplished by treatment with low concentrations of an elongation inhibitor such as CHX, which leads to multiple events of ribosome stalling and RQC activation (, ). Accordingly, Figure 5D shows that the GFP signal from the GFP-R12-RFP reporter increased at subsaturation CHX concentrations and that this effect was Ltn1-dependent (Fig. 5D). However, when we tested the Rqc1 TAP-TAG strain, we could not observe the same result, further indicating that the RQC complex does not control the final Rqc1 protein output during translation (Fig. 5E). In another experiment, yeast cells were treated with 50 μg/ml CHX, a concentration high enough to completely halt translational activity. We then followed the levels of the Rqc1-TAP protein over time. The half-life of Rqc1-TAP protein was much shorter in wt cells when compared with lnt1Δ cells (Fig. S16, A–D). This result suggests that Rqc1 decay rate depends on Ltn1, but is uncoupled from its translation process. This experiment provides further support to the model of a posttranslational regulation of Rqc1 by Ltn1 (Fig. S16E).

The raw data used to create Figures 1 and 2, and the statistical analysis t-test calculations are presented in Tables S2 and S4. The Spearman correlation among the parameters used in Figure 2 is presented in Table S3. All statistical analyses were performed with GraphPad Prism 7.

Coding sequences and the annotation of S. cerevisiae were obtained from the Saccharomyces Genome Database (SGD; https://www.yeastgenome.org) and Ensembl Genomes (http://ensemblgenomes.org/). We excluded dubious ORFs as defined in the Saccharomyces Genome Database from our analysis. The list of inhibitory codon pairs (ICP) and genes with ICP is presented in Table S1.

We developed a program that screens the primary sequence of a given protein and calculates the net charge in consecutive frames of a 30-amino-acid window, the approximate number of amino acids that fill the ribosome exit tunnel (, ). For the net charge determination, K and R were considered +1; D and E were considered −1; and every other residue was considered 0. The additional charges of the free amino and carboxyl groups of the first and last amino acids, respectively, were disregarded. In a previous publication, we have shown that these simplified parameters are equivalent to a calculation using partial charges of individual amino acids at pH 7.4, according to their pKa values and the Henderson–Hasselbalch equation (). The algorithm initially establishes a stretch containing a predetermined number of amino acids (#1 to #30, for example). The stretch charge is calculated, and the charge value and the position of the first amino acid are temporarily saved to the memory. Then, our algorithm advances to the next stretch, from amino acid #2 to amino acid #31, and performs the same analysis until the end of the protein.

Based on Pearson’s correlation between the frequency of occurrence of each codon in each mRNA and the mRNA half-lives, Coller and colleagues created the codon occurrence to mRNA stability coefficient (CSC) (). We designed and implemented an algorithm that calculated the mean value of the CSC for each gene (CSCg) ().

An algorithm was developed to locate the position of stretches containing only adenine (A) in an ORF. The algorithm finds adenine stretches of a chosen size for each ORF and generates as output a file containing the gene name, the stretch size, and its position in the ORF. For our analysis, we considered stretches of size equal to or larger than ten adenines.

We developed an algorithm that searches for stretches containing a certain number of arginines and/or lysines in a stretch of ten amino acids in a protein. The chosen number of K and/or R in the stretches was equal to or greater than 8. The input file of the algorithm was a proteome fasta file, and the output file returned the protein name and the position of the stretch found for each protein.

An algorithm was developed to find the location of stretches containing consecutive arginines and/or lysines in a protein. The stretches analyzed had a size equal to or larger than 6. The input file of the algorithm was a proteome fasta file, and the output file returned the protein name, the stretch size, and the position of the stretch in the protein.

Since no commercial antibodies are available for the proteins used herein, we used a TAP-TAG collection of yeast (). In this collection, each ORF was tagged with a C-terminal TAG, and the same antibody could be used to detect the full collection. For each potential RQC complex target analyzed, a knockout strain for the LTN1 or ASC1 gene was created by homologous recombination. For the Rqc1 TAP-TAG strain, knockout strains for RQC2 and HEL2 genes were also created. After the creation of the knockout strains, the level of the TAP-tagged proteins was compared with that of the wild-type strain by western blot analyses. The TAP-TAG collection strains were cultivated in YPD medium (1% yeast extract, 2% peptone, 2% glucose). The S. cerevisiae BY4741 strain transformed with GFP-R12-RFP vector (R12: CGG CGA CGA CGG CGA CGC CGA CGA CGA CGG CGC CGC) () was cultivated in minimum synthetic medium, uracil dropout (0.67% Difco Yeast Nitrogen Base without amino acids, 2% glucose, and 0.003% methionine, leucine, and histidine).

For knockout creation, S. cerevisiae BY4741 yeast (TAP-TAG collection) was transformed using the PEG/Lithium Acetate method (). The PCR products used in the transformation were obtained by the amplification of the KanMX6 gene contained in the pFA6a-KanMx6 vector. All genes were knocked out by the insertion of the geneticin resistance gene into their original ORFs. The knockouts were confirmed using four sets of PCR. Two PCR sets combined primers for the target gene flank region (5’UTR or 3’UTR) with resistance gene primers. The other two PCR sets combined primers for the target gene flank region and ORF gene target primers. The primers used for deletions, strains construction, and confirmation are depicted in the topic “List of primers and PCR condition.” The S. cerevisiae BY4741 strains transformed with GFP-R12-RFP-TAP-TAG or GFP-ST6-RFP-TAP-TAG vectors (ST6: UCU ACU UCA ACA UCU ACA UCA ACU UCU ACC UCA ACG) were cultivated in minimum synthetic medium, uracil, and histidine dropouts.

The protein extraction of the strains was conducted in the exponential phase of growth (O.D ∼ 0.6–0.8), treating equal quantities of cells using the NaOH/TCA method (). For the SDS-PAGE, equal sample volumes were applied on the acrylamide gels and submitted to an electrophoretic run. For the YHR131C TAP-TAG strain, protein extraction was conducted using a glass bead protein extraction protocol (). Total protein quantification of the glass bead protein extracts was conducted using the BCA Protein Assay kit (Thermo Fisher), and samples with equal amount of proteins were applied on SDS-PAGE gels. The acrylamide gel concentration changed between 10 and 12.5% according to the size of the analyzed protein.

Proteins separated by SDS-PAGE were transferred to PVDF membranes using the TRANS-BLOT semidry system from BIO-RAD or the wet blotting system in transfer buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, 20% methanol, pH 8.3). The membranes were blocked using a Li-Cor blocking buffer with an overnight incubation at room temperature. Incubation with primary antibodies was performed at 4 °C for at least 1 h. The membranes were incubated in Li-Cor Odyssey secondary antibodies at room temperature for at least 1 h and then revealed in an Odyssey scanner. The primary antibodies used were mouse anti-PGK1 (Invitrogen; dilution 1:10,000), rabbit anti-TAP-TAG (Thermo Scientific; dilution 1:10,000), mouse anti-GFP (Sigma; dilution 1:2000), and rabbit anti-Asc1 (dilution 1:5000) (). The secondary antibodies that were used were anti-mouse 680 LT Li-Cor (dilution 1:10,000) and anti-rabbit 800 CW Li-Cor (dilution 1:10,000).

For qRT-PCR analysis, the Rqc1 TAP-TAG knockout strains constructed in this work were cultivated until the exponential phase (O.D ∼ 0.6–0.8) and RNA from these cells were extracted following the hot phenol method. The same protocol was used for the corresponding knockouts from a commercial collection as a control. After extraction, the RNAs were submitted to DNase I (Thermo Scientific) treatment for DNA contamination elimination. The reverse transcription was performed using random primers with the High-Capacity cDNA Reverse Transcription Kit (Thermo Scientific). For the cDNA amplification, we used SYBR Green PCR Master Mix on the StepOne Plus System (Applied Biosystems). The following protocol was used on the thermocycler: 95 °C for 10 min and 40 cycles of 95 °C for 15 s and 60 °C for 1 min, melting curve analysis: 95 °C for 15 s, 60 °C for 1 min and 95 °C for 15 s. The primers used were: 5’CGGAATGCACCCGCAACATT and 5’CGGCCTTTCCCATAATTGCCA for RQC1, and 5’ TTCCCAGGTATTGCCGAAA and 5’ TTGTGGTGAACGATAGATGGA for ACT1. The analysis of differential expression was made by relative quantification using 2^−ΔΔCt method, and the statistical analysis was conducted on Prism.

First, the TAP-TAG sequence was amplified from the genome of a TAP-TAG collection strain using primers with restriction sites inserted on the extremities of the sequence. The primers were designed to amplify the TAP-TAG sequence and the His3Mx6 marker. The PCR product was cleaned using the Wizard SV Gel and PCR Clean-Up System (Promega) and the purified PCR was used in a double digestion reaction with Box I (Thermo Scientific) and Not I (Fermentas) restriction enzymes. The double digestion reaction was conducted in buffer Tango 1X in an overnight incubation at 37 °C. The same double digestion reaction was conducted with the plasmids (GFP-R12-RFP and GFP-ST6-RFP) (). Both PCR and plasmids digested were purified with Wizard SV Gel and PCR Clean-Up System (Promega). The ligation reaction was done using the 3:1 proportion of vector to insert, where 150 ng of vector and 126 ng of insert were used. The ligation reaction with the T4 DNA ligase (Invitrogen) was conducted following the manufacturer’s instructions. Escherichia coli DH5α competent cells were used to receive the plasmids after the ligation reaction was completed. To confirm the TAP-TAG insertion in the colonies two sets of colonies PCR were done. In the first one, we observed TAP-TAG presence in the 3’ extremity of RFP sequence. In the second set we observed the sequence of TAP plus His3Mx6 marker in the 3’ extremity of RFP sequence. The cloning strategy removes 63 nt of the RFP sequence for TAP-TAG sequence insertion. Another confirmation was done by western blot, where we observed an increase in the full-length protein by approximately 28 kDa, the size of the TAP-TAG insertion. Moreover, just strains transformed with the TAP-TAG plasmids presented immunoreactivity against the TAP-TAG antibody (Fig. S13). The primers used for both to amplify the TAP-TAG sequence from the genome and for the confirmation steps are depicted in the topic “List of primers and PCR condition.”

For the experiments using serial dilutions of CHX, concentrations ranging from 10⁻⁶ to 10⁻¹ mg/ml were used, with steps of ten times the dilution factor. Strains grew until the exponential phase and were treated with CHX for 10 h under shaking at 28 °C. For the pulse-chase experiment a concentration of 50 μg/ml of CHX was used. Strains also grew until the exponential phase and were treated with CHX under shaking at 28 °C. The points were removed in intervals of 20 min of incubation with the drug. Alternatively, points were also removed in intervals of 30 min after 10 h of incubation with CHX. The points collected were submitted to the protein extraction protocol.

S. cerevisiae ribosome profiling data were treated as described previously (, ). The data were analyzed as described by Ingolia and collaborators (), except that the program used here was Geneious R11 (Biomatter Ltd) instead of the CASAVA 1.8 pipeline. The data were downloaded from GEO (, ), and the adaptor sequence was trimmed. The trimmed FASTA sequences were aligned to S. cerevisiae ribosomal and noncoding RNA sequences to remove rRNA reads. The unaligned reads were aligned to the S. cerevisiae S288C genome deposited in the Saccharomyces Genome Database. First, we removed any reads that mapped to multiple locations. Then, the reads were aligned to the S. cerevisiae coding sequence database, allowing two mismatches per read. For 27 to 29 nt footprint analyses, the ribosome profiling data were obtained without any drug, and just fragments with 27 to 29 nt were used (). For 20 to 22 nt footprint analyses, the ribosome profiling data were obtained with cells lysed with buffer containing 0.1 mg/ml of CHX and tigecycline (TIG) or 0.1 mg/ml of CHX and anisomycin (ANS) (). We normalized the coverage within the same transcript by dividing each nucleotide read by the sum of the number of reads for each gene in a given window. A window of 150 nucleotides before and after the beginning of a stretch of interest was used to calculate the number of reads of each gene. For Figures 3, C–G, 5C, and Fig. S2, the ribosome profiling analysis was done following as described by Ingolia and collaborators (). The data sets used in our analyses were from Pop et al., 2014 () (GEO accession GSE63789) and from Matsuo et al., 2020 () (GEO accession GSE131214). The data were mapped to the R64.2.1 S288C Coding Sequence (CDS) from the Saccharomyces Genome Database Project. The alignment was done using Bowtie accepting two mismatches at most. The analysis was restricted to 28 to 32 nt reads for Fig. S2. The A-site was estimated based on the read length: it was defined as the codon containing the 16th nt for 28 nt reads; the 17th nt for 29, 30, and 31 nt reads; and the 18th nt for 32 nt reads, considering the first nt as 1. The A-site estimation and other Ribosome Profiling Analysis were performed by custom software written in Python 3.8.2. For Figs. S2 and S3, we compared the ribosome profiles normalized for genes of the five distinct groups (Fig. S2) or all polybasic group (Fig. S3) with 2000 aleatory genes. Statistical analyses were performed with multiple t-tests using the Holm–Sidak method, and the adjusted p-value was calculated for each point (GraphPad Prism 7). For Figure 4, the ribosome profiling of each gene was downloaded from Trips-Viz by aggregating all experiments in the database (). Disome Profiling Data: The data sets we used for disome profiling were from D’Orazio et al., 2019 () (GEO accession GSE129128), and Meydan and Guydosh, 2020 () (GEO accession GSE139036). For the Meydan and Guydosh data, we further trimmed two nucleotides from the 5’ end of each read (the ones with the lowest quality in every read) and also five nucleotides from the 3’ end (adapting for the postmapping trimmings done to the nonmapped reads, as described in Meydan and Guydosh, 2020). Noncoding RNA sequences were removed from both data sets using Geneious. Then they were also mapped to the R64.2.1 S288C Coding Sequence (CDS), using Bowtie, accepting two mismatches at most. For D’Orazio data, we accepted any read containing between 50 nt and 80 nt. We estimated the A site from the first ribosome (the one at the 3’ portion of the read) as the codon containing the 46th nucleotide. For Meydan and Guydosh data, we accepted reads containing between 57 nt and 63 nt. Also, we estimated the A site from the ribosome at the 3’ portion of the read as the codon containing the 46th nucleotide.