Deletion of tusB results in a complete loss of s2U and changes in other tRNA modifications.

Log phase bacterial cultures were grown for 2h at 37°C (untreated or treated with 1µg/mL of doxycycline). Small RNA was purified and digested into nucleosides. Changes in relative quantities of modified nucleosides were measured by LC-MS/MS. Each sample is normalized to the average peak area of four biological replicates for WT untreated (WT1-4), see Supplemental Data 1.

Deletion of tusB induces slowed growth and antibiotic tolerance to ribosome and RNA-polymerase targeting antibiotics due to loss of s2U tRNA modification.

Log phase bacterial cultures were grown at 37C and sampled every 2h to enumerate CFUs (colony forming units). Complemented strains carry a plasmid (pDS145)30 with the yrvO-mnmA open reading frames under the control of the Para promoter, and were treated with 0.2% arabinose throughout the experiment to induce expression of the construct. Cultures were treated with either 1µg/mL doxycycline, 25µg/mL chloramphenicol, 2µg/mL gentamicin, 2µg/mL rifampicin, 0.1µg/mL ciprofloxacin, or left untreated. Fold change in CFUs are shown by normalizing each timepoint to the 0h timepoint. Two-way ANOVA with Tukey’s multiple comparisons. Asterisks denote significant p-values as follows: **** p <0.0001, *** p <0.001, ** p < 0.01, and * p < 0.050. Data points represent the mean of three biological replicates, error bars represent standard deviation (SD). See Supplemental Data 2.

ΔtusB has a starvation like metabolic profile and reduced ribosomal protein levels.

A. Log phase bacterial cultures were grown for 2h at 37°C before isolating RNA. Significant changes in transcript levels between WT and ΔtusB were evaluated by DESeq2. Enriched KEGG Pathways (p-value < 0.05) are shown for downregulated and upregulated genes. The number of genes associated with each pathway is represented by circle size. Three biological replicates. B. Protein was isolated from bacterial cultures after 2h of growth at 37°C. Significant changes in protein abundance between WT and ΔtusB were evaluated by One-way ANOVA. The number of significantly regulated genes (p-value < 0.05) associated with KEGG pathways are shown. Three biological replicates. See Supplemental Data 3.

Ribosomes pause more frequently at Glutamate, Glutamine, and Lysine codons.

Cultures of bacteria were grown for 5h at 37°C and then flash frozen in liquid nitrogen to stop translation. Ribosome protected fragments were isolated and sequenced to measure the abundance of ribosomes at each codon. Pause scores are calculated by taking the average ribosome density at a codon normalized to the average ribosome density across a given gene for all genes in the genome. Relative ribosome occupancy (pause score of ΔtusB normalized to the pause score of WT) is shown for each codon. Data points represent the mean of two biological replicates, error bars represent standard deviation (SD). See Supplemental Data 4.

Switching Glutamine codons from CAA to CAG slightly improves GFP translation.

A. Graphical depiction of model. B, C, D. Log phase bacterial cultures were grown for 8h at 37°C. At 0h cultures were treated with 1mM IPTG to induce expression of gfp or left untreated as a control. B. Fold change in OD600 is shown by normalizing each timepoint to the 0h timepoint. C. GFP fluorescence is normalized to optical density (OD600) to account for differences in bacterial density between samples. Values are normalized to the 0h timepoint to show fold change in GFP fluorescence overtime from the point of IPTG induction. D. RNA was isolated at 4h during the experiment and gfp transcript levels were quantified by qPCR. GFP fluorescence is normalized to gfp transcript level as a proxy for GFP protein translation. GFP fluorescence is normalized to the untreated (uninduced) control to account for background autofluorescence of the bacteria. gfp transcript levels were also normalized to the untreated (uninduced) control to account for any leaky expression of the construct. See Supplemental Data 5.

A combination of Glu, Gln, and Lys codons are highly associated with ribosomal proteins.

Codon Z-scores were calculated for each gene by taking the frequency of a given codon in that gene and normalizing it the global frequency of that codon and the standard deviation of that codon. A. A data matrix of codon Z-scores for all genes that were significantly regulated at the protein level in ΔtusB vs. WT (2-fold change, p<0.05) were analyzed by principal component analysis. PC scores are plotted for each gene based on their codon usage. PC1 explains 11.3% of the variance and PC2 explains 6.4% of the variance. B. Biplot shows the loadings plot, linear association of each codon, with the PC scores of each gene. C. Biplot of genes associated with Lys AAG, Lys AAA, and Glu GAA (PC score >2 and <-0.5) are graphed separately for clarity. Boxplots of codon Z-scores amongst upregulated (red) and downregulated (blue) genes in this subset are shown for Glu GAA, Lys AAA and Lys AAG. One-way ANOVA with Dunnett’s T3 multiple comparisons (* denotes p < 0.05). D. Cumulative pause predictor scores were calculated by taking the sum of Glu (GAA), Glu (GAG), Gln (CAA), Lys (AAA), Lys (AAG), codon Z-scores weighted by their ribosome pause score. Cumulative pause predictor scores are higher on average amongst downregulated proteins (blue) versus upregulated proteins (red). Mann Whitney test, ** p < 0.01. E. Ribosomal proteins that are significantly downregulated at the protein level (purple) have on average higher cumulative pause predictor scores compared to all other downregulated proteins (blue). Mann Whitney test, *** p < 0.0001. F. The log2fold change in protein abundance in ΔtusB vs. WT (x-axis) is plotted against the cumulative pause predictor score for all 54 ribosomal proteins. Simple linear regression shows a negative slope (-3.215) and significant deviation from zero (p-value = 0.0008). See Supplemental Data 6.

tusB is downregulated in response to doxycycline while multiple other genes in the 2-thiolation pathway are downregulated in response to gentamicin.

Log phase bacterial cultures were grown at 37°C and treated with either 1µg/mL doxycycline, 15µg/mL gentamicin, or left untreated. After 2h RNA was isolated and relative transcript abundance of tusA, tusB, tusC, tusD, tusE, and mnmA was quantified by qPCR using 16S transcript levels as an endogenous control. Three untreated biological replicates were done in parallel with doxycycline experiments and three untreated biological replicates were done in parallel with gentamicin experiments. All six biological replicates for untreated samples and the three biological replicates for the treated samples are shown (dots). Values are normalized to the average transcript expression of the three untreated controls that were done in parallel for each experiment. Lines indicate mean relative transcript levels. Kruskal-Wallis test with Dunn’s multiple comparisons test, * p< 0.05, * p<0.01.

Expression of yrvO-mnmA restores growth and antibiotic susceptibility of ΔtusB.

Log phase bacterial cultures were grown at 37°C with arabinose induction (0.2% arabinose) and sampled every 2h to enumerate CFUs (colony forming units). At the 0h timepoint, cultures were treated with either 1µg/mL doxycycline, 25µg/mL chloramphenicol, 2µg/mL gentamicin, 2µg/mL rifampicin, 0.1µg/mL ciprofloxacin, or left untreated. Fold change in CFUs are shown by normalizing each timepoint to the 0h timepoint. Two-way ANOVA with Tukey’s multiple comparisons. Asterisks denote significant p-values as follows: **** p <0.0001 and * p < 0.050. A. ΔtusB + pDS145 (pBAD plasmid containing yrvO-mnmA under the regulation of Para promoter), three biological replicates same as shown in Figure 2. B. ΔtusB + pDS220 (pBAD plasmid yrvO(C325A)-mnmA under the regulation of the Para promoter), three biological replicates. C. ΔtusB + pARA13 (pBAD empty vector control), two biological replicates.

Deletion of relA does not restore growth or doxycycline susceptibility of ΔtusB.

Log phase bacterial cultures were grown at 37°C and sampled every 2h to enumerate CFUs (colony forming units). At the 0h timepoint, cultures were treated with either 1µg/mL doxycycline or left untreated. Fold change in CFUs are shown by normalizing each timepoint to the 0h timepoint. Two-way ANOVA with Tukey’s multiple comparisons. Asterisks denote significant p-values as follows: **** p <0.0001. WT and ΔtusB doxycycline replicates are the same as shown in Figure 2.

Loss of s2U may increase frequency of amino acid misincorporation at CAA codons.

A. Table shows representative peptides and the modification (Glutamine Q -> Histidine H) detected in red. B. The abundance of each peptide was quantified by LC-MS/MS. The abundance ratio of modified to unmodified peptides in ΔtusB is normalized to the modified/unmodified abundance ratio in WT.

GFP protein decay is slower in ΔtusB.

Overnight cultures were treated with 1mM IPTG to induce expression of gfp. After 16-18h of growth at 26°C, samples were resuspended in PBS and treated with 50 µg/mL of kanamycin (MIC) to stop translation. Fluorescence (480/520) was measured every 5 minutes for 120 minutes in a plate reader. Percent signal of the starting GFP signal (0h) is shown overtime. Half-life is calculated using the one phase decay equation using 25% GFP signal as a constraint for the plateau (limit of detection).

Predictor pause scores are not higher among downregulated proteins or ribosomal proteins for random codons where ribosomes do not pause at.

Five codons that the ribosome does not pause more frequently at in ΔtusB were selected to calculate the cumulative pause predictor score as a control. Cumulative pause predictor scores were calculated by taking the sum of Ala (GCA), Arg (GAG), Asn (AAC), Cys (TGC), Gly (GGA), codon Z-scores weighted by their ribosome pause score. A. Cumulative pause predictor scores of upregulated proteins and downregulated proteins. Mann Whitney test, * p < 0.05. B. Cumulative pause predictor scores of ribosomal proteins and all other downregulated proteins *** p = 0.001

Sequences were ribosomes pause frequently at are conserved among bacterial species.

A. Representative image of ribosome density across a sequence in the rpmG (ribosomal subunit L33) gene. B. Amino acid sequences of rpmG were pulled from UniProt for multiple different bacterial species. Highlighted are the residues ribosomes are predicted to pause more frequently at.

Ribosomal protein genes are downregulated at the protein level in ΔtusB (vs WT), but few are differentially regulated at the transcript level.

All 54 ribosomal protein genes and their transcript abundance in ΔtusB compared to WT (as measured by RNA-seq) and their protein abundance in ΔtusB compared to WT (as measured by mass spectrometry). NA denotes when there was no significant difference (by p-value) detected by RNA-seq.