Polyamines are Required for tRNA Anticodon Modification in Escherichia coli

Biogenic polyamines are natural aliphatic polycations formed from amino acids by biochemical pathways that are highly conserved from bacteria to humans. Their cellular concentrations are carefully regulated and dysregulation causes severe cell growth defects. Polyamines have high affinity for nucleic acids and are known to interact with mRNA, tRNA and rRNA to stimulate the translational machinery, but the exact molecular mechanism(s) for this stimulus is still unknown. Here we exploit that Escherichia coli is viable in the absence of polyamines, including the universally conserved putrescine and spermidine. Using global macromolecule labelling approaches we find that ribosome efficiency is reduced by 50– 70% in the absence of polyamines and this reduction is caused by slow translation elongation speed. The low efficiency causes rRNA and multiple tRNA species to be overproduced in the absence of polyamines, suggesting an impact on the feedback regulation of stable RNA transcription. Importantly, we find that polyamine deficiency affects both tRNA levels and tRNA modification patterns. Specifically, a large fraction of tRNA, tRNA and tRNA lack the queuosine modification in the anticodon “wobble” base, which can be reversed by addition of polyamines to the growth medium. In conclusion, we demonstrate that polyamines are needed for modification of specific tRNA, possibly by facilitating the interaction with modification enzymes. 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CCBY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Polyamines are organic polycations that are essential in many organisms. 1 Imbalances in polyamine levels cause severe cell growth defects. They are synthesized by fundamental and conserved pathways and found in all domains of life from bacteria to human. 2,3 Intracellular polyamine concentrations are in the millimolar range and carefully regulated at the levels of biosynthesis, uptake, secretion and degradation. The gram-negative model bacterium Escherichia coli primarily synthesizes three types of polyamines: the diamine putrescine (1,4-butanediamine) is synthesized by decarboxylation of arginine and ornithine (by enzymes encoded by speA, speB, speC and speF) and the diamine cadaverine (1,5-pentanediamine) is synthesized from lysine (by cadA and ldcC). 4 The triamine spermidine (N-(3-aminopropyl)-1,4-b utandiamine) is synthesized by amino propylation of putrescine (by speE and speD). Putrescine and spermidine are common and present in all domains of life, whereas cadaverine belongs to a more diverse repertoire found only in bacteria and archaea. 5 Despite their chemical simplicity, polyamines have been implicated directly or indirectly in the acid stress response 6 biofilm formation 7 outer membrane porin formation 8 oxidative stress response, antibiotic tolerance [9][10][11] and in hostbacterium interactions and virulence. 12,13 Because of their cationic properties and their flexible hydro-carbon chain, polyamines are thought to bind and form bridges between negatively charged sites in nucleic acids inside the cell (reviewed in 14). Affinity studies have estimated that 47.9% of putrescine and 89.7% of spermidine is associated with RNA and polyamines therefore exist predominantly as a polyamine-RNA complex. 15 Hints to polyamines' role in RNA binding and structure stabilization arose from early structural studies on ribosomes and tRNA. 16,17 Interactions of polyamines with tRNA were observed to occur with the T-loop and the D-stem close to the anticodon stem. 17 These binding sites were later confirmed using NMR spectroscopy and photoactivatable crosslinking studies. 18,19 Binding of polyamines to RNA have been suggested to stabilize and increase melting temperatures of RNA duplexes. 20 Consistently, it has recently been observed that rRNA, tRNA his and tRNA tyr are highly unstable in Thermus Thermophilus during heat stress (at 80°C) in the absence of long-chain and branched polyamines. 21 Early studies have reported that polyamines also directly stimulate in vitro translation. 22 Here, polyamines can accelerate codon recognition on the ribosome without loss of translational fidelity. 23 Consistently, the lack of polyamines has been reported to decrease translation fidelity and increase mistranslation. 24,25 On the other hand, polyamines were reported to be required for amber stop codon read-through, a specific case of mistranslation. 26 Polyamines have been reported to directly or indirectly regulate a set of genes referred to as the "polyamine regulon". [27][28][29] In E. coli, polyamines have been shown to increase translation of at least 20 proteins, which could be involved in transcriptional regulation of 300 genes (recently reviewed in 27). The majority of the affected mRNAs contain non-optimal Shine-Dalgarno sequences, weak start codons or internal frameshifts or nonsense codons, and are dependent on polyamines for their efficient translation. Interestingly, in some of the protein-encoding mRNA it seems to be bulged-out regions in doublestranded RNA that are necessary for the stimulatory effect, but it is still not known how polyamines regulate these regions. 30,31 Curiously, many of the polyamine-dependent genes encode proteins central for cell growth during logarithmic phase, while others are involved in stationary phase or stress survival. 27 This could indicate a very general and possibly growth-state-independent role of polyamines in translation, which is still unknown.
Here, we applied whole-cell labelling approaches in E. coli to initially identify global effects on translation. While macromolecule synthesis including translation was severely inhibited in the absence of polyamines, we observed an overproduction of RNA during steady-state growth to levels usually obtained at much higher growth rates. This imbalance could not be explained by increased mistranslation, but could be caused by a general effect of polyamines on the translation machinery, which was observed to decrease the translation elongation speed. Using tRNA northern blot analysis we identified a severe defect in queuosine modification of tRNA his , tRNA tyr and tRNA asn . We also observed changes in other tRNAs species, which suggested a general effect of polyamines on tRNA modification. Consistent with the lack of tRNA modification, we observed elevated levels of tRNA and rRNA in the absence of polyamines, in accordance with reduced feedback from inefficient ribosomes to the stringent control of transcription. Finally, by structural probing we found that polyamines did not change the overall secondary structure of tRNA his , but affected the accessibility of the anticodon loop. In conclusion, we present here a new general mechanism by which polyamines stimulate global protein translation through tRNA modification.

Results
Polyamines stimulate growth rate and macromolecule synthesis Polyamines are not required for viability in E. coli, which makes this organism a suitable model for studying the effects of polyamines. 6,32 We generated a polyamine auxotrophic strain (here referred to as D8) in which the eight genes needed for polyamine biosynthesis were deleted (see Supplementary Information) and verified its genome sequence to ensure the absence of undesired genetic changes. When grown in defined liquid medium (MOPS MM) containing all 20 amino acids at 37°C, MG1655 D8 had a doubling time (T d ) of about 88 min, which is a growth rate approximately 70% slower than the MG1655 wildtype (WT) strain ( Figure 1(A)). This is slightly slower, but in line, with previous reports. 32 Consistent with the polyamine auxotrophy, the D8 mutant reached the same growth rate as the WT (T d~2 6 min) when 100 lM spermidine ( Figure 1(A), D8 + spermidine) was added to the growth medium. This showed that D8 is auxotrophic for polyamines and that addition of polyamines (spermidine) to the growth medium neutralizes the defect that is responsible for growth retardation.
Quantitative measurements of macromolecule synthesis were expected to give useful insights into the molecular mechanism underlying the reduced growth rate. We therefore applied a method to measure and calculate synthesis rates of DNA, RNA and protein from incorporated ionization counts with radiolabeled uracil (for DNA and RNA) and leucine (for protein) (see Supplementary Figure S1(A-B)). To prevent biosynthesis of uracil and leucine, we deleted the leuA and pyrE genes of the WT and D8 strain set. These auxotrophic strains grew at rates comparable to WT and D8 in the presence of uracil and leucine (Supplementary Figure S1(A)). Consistent with the slow growth rate of D8 we observed 66.5-67.1% reduction in the steadystate synthesis rates (lmol min À1 mL À1 ) of protein, RNA and DNA (Figure 1(B), see Supplementary Figure S1(C-R) for calculations). 33 When cultures of WT E. coli are made to grow at a range of steady-state growth rates by culturing them in media of different nutritional qualities, a linear correlation is observed between the growth rate and the cellular ratio of RNA to protein or RNA to DNA. 33,34 The relative reduction in RNA at lower growth rates reflect that the vast majority of cellular RNA is components of the protein synthesis apparatus (rRNA and tRNA), and that fewer resources are invested in protein synthesis capacity in the nutrient-limited slow growing cells. In agreement with this interpretation, treatments or mutations that reduce the growth rate by directly interfering with the functionality of the protein synthesis apparatus, rather than by nutrient limitation, do not conform to the same linear relationship. Such cells have a higher relative RNA content than nutrient-limited WT cells growing at the same rate. 35,36 By comparing the relative macromolecular contents of D8 to the data compiled by Bremer and Dennis 33 we observed that the ratio of RNA to protein was higher in D8 than expected from a WT strain growing at a similar rate (Figure 1(C) compare red squares with gray triangles). Strikingly, the RNA to DNA ratio was~3-fold higher ( Figure 1 (D)), and a similar~3-fold difference was observed in the protein to DNA ratio (Figure 1(E)). The ratios of protein, RNA and DNA we measured in the WT were slightly higher but comparable to the ratios reported by Bremer and Dennis 33 (Figure 1(C-E), compare blue diamonds with gray triangles). This suggested an increased production of ribosomes and MG1655 D8 (D8, red squares) were grown exponentially in MOPS MM with 1.32 mM K 2 HPO 4 , 10 lg/mL Uracil, 0.2% glucose and FN20 amino acid mix at 37°C. At time zero, spermidine was added to a culture containing MG1655 D8 at a final concentration of 100 lM (green triangles). Optical density (OD 436 ) was measured at times indicated. Doubling times (T d in min) are indicated above the growth curves. The dashed line indicates the growth rate achieved by the MG1655 D8 strain after growth for more than one hour in the presence of 100 mM spermidine. (B) Macromolecule (Protein, RNA and DNA) synthesis (lmol mL À1 ) in MG1655 DleuADpyrE (WT, blue diamonds) and MG1655 D8 DleuADpyrE (D8, red squares) over time (min) were calculated based on steady state continuous incorporation of 3 H-leucine and 14 C-uracil as described in Figure S1. (B) shows a representative plot of a single experiment. (C-E) Ratios between calculated absolute amounts of macromolecules (mg/mL of Protein, RNA and DNA) in 1 mL of cells (OD 436 = 1). The data of WT (blue diamonds) and D8 (red squares) are based on biological independent experiments (n = 3) summarized in supplementary Figure S1. Gray triangles indicate ratios obtained from wild type E. coli growing at different growth rates in doublings per hour (m), as presented in the study by Bremer and Dennis. 33 and tRNA during steady-state growth in D8, possibly caused by an underlying effect of polyamine deficiency on the translation machinery. The growth rate of E. coli is tightly coupled with the steady-state RNA/protein ratio. 37,38 The complete restoration of wild type growth rate in the D8 mutant supplemented with spermidine therefore indicates that the imbalanced macromolecular ratios of the D8 mutant are directly caused by the lack of polyamines, rather than an unknown indirect effect of the eight gene deletions.

Polyamines affect translation of specific proteins and global translation elongation speed
To investigate any global effects on protein synthesis and translation machinery we performed proteome analysis by 2D-PAGE using 35 Smethionine pulse-labeled cell samples. Total protein was initially separated by isoelectric focusing, and then by size, which can reveal changes in translation fidelity and mistranslation due to "stuttering" of protein spots. 39 When comparing labeled total protein from MG1655 (WT) with MG1655 D8 (D8) we did not observe any obvious differences in the proteome, which could account for the severe growth phenotype of MG1655 D8 (Figure 2(A), biological repeat see Supplementary Figure S2(A)). Some protein levels were distinctly increased or decreased in the D8 strain relative to the WT, but we did not observe any apparent effect of polyamines on mistranslation, which was confirmed by a dual luciferase assay for mistranslation (Supplementary Figure S2(B)). The overall decline in signal in D8 samples suggested that polyamines not only regulate specific proteins, but also the global translation rate. This global effect could be caused by a direct effect of polyamines on the translation elongation speed. To test this we measured the translation elongation speed of bgalactosidase production by sub-minute sampling and activity assays in WT and D8 (Figure 2 ). We observed a clear effect of polyamines on b-galactosidase synthesis times, which corresponded to approximately 80 (~13 amino acid per second) and 180 seconds (~6 amino acid per second) in WT and D8, respectively. In conclusion, while the polyamine-deficient strain did not show significant mistranslation, it showed differential expression of specific proteins and more importantly showed severely decreased translation elongation speed.

Polyamines are needed for anticodon loop modifications of tRNA
The strong effect on the translation machinery and the recently described effects of long-chain and branched polyamines on tRNA his and tRNA tyr stability in thermophile T. thermophilus 21 prompted us first to look at tRNA with our polyamineauxotrophic strain (Figure 3(A)). We analysed the tRNA charging levels in WT and D8 in the absence and presence of 100 lM spermidine, for tRNA hisR and tRNA tyrTV (Figure 3(A), lanes 1-6). tRNA nomenclature indicate corresponding tRNA gene names in E. coli K-12. In the WT strain, we observed two bands corresponding to charged and uncharged species of tRNA (compare lane 1 and 2). However, when analysing these tRNA species in the D8 mutant, we observed four bands, two corresponding to the bands observed in WT and two that migrated further in the gel (indicated with arrows, compare lanes 1 and 3). This suggested that large fractions of tRNA hisR and tRNA tyrTV were not fully modified in the absence of polyamines. Furthermore, the fully charged and modified tRNA species was less abundant in the D8 mutant than in WT. Consistent with this interpretation, growth in the presence of 100 lM spermidine counteracted this effect in D8 and resulted in a band pattern identical to WT (compare lanes 1 and 5). The queuosine or epoxyqueuosine modification is a large tRNA modification that is conserved on tRNA his and tRNA tyr from bacteria to eukaryotes. This modification is synthetized from GTP and transferred to the first anticodon base ("wobble base") as prequeuosine (preQ1) by tRNA-guanine transglycosylase (Tgt) and then matured to form epoxyqueuosine. This is converted to queuosine by a corbamidedependent reaction that is generated under anaerobic conditions 40,41 (Supplementary Figure S3(A)) and could account for the faster migrating tRNA species observed in D8. To test this hypothesis we removed the tgt gene (Dtgt), which would block transfer of prequeuosine and performed northern blot analysis ( Figure 3(A), lane 7 and 8). Strikingly, the tRNA of the Dtgt mutant migrated at the same size as observed for the furthest migrating tRNA in D8 (compare lanes 3-4 with 7-8). This clearly indicated that queuosine/epoxyqueuosine modification was inhibited in the absence of polyamines. tRNA asn and tRNA asp are also modified with queuosine 42 and we also performed northern blot analysis of tRNA asnTUVW and tRNA aspTUV (Figure 3(A)). As observed, tRNA asnTUVW was also severely affected, whereas tRNA aspTUV was less affected (compare lanes 3-4 and 6-7). This suggested that polyamines are needed for queuosine/epoxyqueuosine modification of tRNA hisR , tRNA tyrTV and tRNA asnTUVW , but less so for tRNA aspTUV . Furthermore, this observation suggested that the polyamine requirement could be selective for specific tRNA. We therefore randomly selected different tRNA species and found that tRNA gltTUVW , tRNA leuPVQT and tRNA ileTUV were also slightly affected in migration in D8, although they are not modified by Tgt (Supplementary Figure S3(B)). Other tRNA species like tRNA valTUVWX , tRNA trpT and tRNA argVYZQ were not significantly affected by polyamines in our assay ( Supplementary Fig-ure S3(C)). Interestingly, tRNA levels were in many instances observed to be elevated in D8 as compared to WT (compare lane 1 with 3, Figure 3(A) and Supplementary Figures S3(B and C)). In order to normalize for variation in purification yields and gel loading, a tRNA selC spike-in internal control was included (Figure 3(B), see Material and Methods for description). Since tRNA selC levels are constant in our blots, this suggested that tRNA levels for some tRNA species were elevated in D8 (Figure 3(B) and Supplementary Figure S3(B and C)).
Queuosine modifications have recently been reported to increase translation elongation speed and the lack of queuosine modifications could contribute to the decreased elongation speed observed in D8. 43 We therefore assayed translation elongation speed in MG1655 Dtgt (Supplementary Figure S2(E)). We found that MG1655 Dtgt on average synthesized b-galactosidase in 94 seconds (~11 amino acids per second), which was slightly slower than WT (~13 amino acids per second), but not as slow as D8 (~6 amino acids per second). In conclusion, the results suggest that polyamines are not only required for queuosine modification, but could also affect other RNA modifications, some of which were not necessarily resolvable in our setup. The cellular levels of tRNA were observed to be elevated for some tRNA species in D8, which suggested increased tRNA transcription or increased tRNA stability.
Polyamine deficiency increases the levels of certain tRNA and 16S rRNA As mentioned, tRNA his and tRNA tyr have reduced stability in the absence of long-chain and branched polyamines at high temperatures (80°C) in T. thermophiles. 21 We therefore analysed the stability of several tRNA species including tRNA hisR and tRNA tyrTUV upon treatment with the transcriptional inhibitor rifampicin, and using tRNA selC spike-in  Figure 4(B)). tRNA leuPQVT levels were 1.5-fold higher in D8 as compared to WT and also observed to be stable after 180 min of rifampicin treatment (Figures 4(A  and B)). tRNA asnTUVW and tRNA ileTUV levels were  inhibiting transcription with 100 lg/mL rifampicin. The RNA was separated by polyacrylamide gel electrophoresis, transferred to a membrane and the tRNA (indicated) was detected by hybridization with a tRNA-specific radiolabeled oligonucleotide probe. Lanes are indicated by numbers. See Figure S4 for a biological repeat. (B) Quantification of the tRNA bands in (A). Wildtype: blue diamonds, D8: red squares, D8 + spe: green triangles. The tRNA band signal for each sample was first normalized to the signal of spike-in tRNA selC internal control and then normalized to the signal of WT before rifampicin treatment. The values from independent biological replicates (n = 2) were plotted and the average values indicated with a line (Wildtype: blue, D8: red and D8 + spe: green). also approximately 1.5-fold higher in D8 compared to WT (Figure 4(A and B)). These tRNA species underwent notable degradation during the experiment, but at comparable rates in D8 and WT. Interestingly, no significant difference in tRNA levels or stability was observed for tRNA alaTUV and tRNA gltTUVW between the two strains ( Figure 4(A  and B)). The increased tRNA levels observed in D8 can be attributed to the lack of polyamines as growth with 100 lM spermidine reversed this effect (Figure 4(A and B), compare lanes 1-4 with 9-12).
Ribosomal RNA in T. thermophilus was also observed to be more unstable in the absence of polyamines in response to elevated temperature. 21 We therefore included 16S rRNA in our analysis (Supplementary Figure S4(B)). Both premature 16S rRNA (pre16S) and total 16S rRNA (16S) levels were observed to be elevated in D8 as compared to WT, 2 and 1.3-fold respectively (Supplementary Figure S4(C)). pre16S rRNA was observed to be highly unstable as compared to 16S rRNA in the D8 mutant and WT after 180 min of treatment with rifampicin. In conclusion, among the RNA levels we have measured, the polyamine-deficient D8 strain had increased levels of tRNA hisR , tRNA tyrTUV , tRNA asnTUVW , tRNA leuPQVT , tRNA ileTUV and 16S rRNA, but not tRNA alaTUV and tRNA gltTUVW . Interestingly, while some tRNA species were degraded faster than others upon rifampicin treatment, we did not observe increased degradation in D8 as compared to WT, suggesting the different steady-state levels arose due to a higher transcription rate in D8 relative to WT. The observation that polyamine deficiency affects tRNA differently, could suggest that polyamines selectively bind to specific RNA species.

Polyamines increase the accessibility of the anticodon loop of tRNA hisR
Here, we observed that polyamines can affect tRNA species differently. Furthermore, the specific queuosine/epoxyqueuosine modification was affected to different degrees in tRNA hisR and tRNA aspTUV . This suggested that polyamines were needed for anticodon modification in tRNA hisR , but less so in tRNA aspTUV . A simple, feasible explanation is that polyamines directly make the anticodon loop of tRNA hisR more accessible for modification enzymes. To investigate this hypothesis, we performed in vivo structural probing of tRNA hisR by treating WT and D8 cells with dimethyl sulfate (DMS) followed by primer extension. DMS methylates single-stranded RNA, which leads to inhibition of reverse transcriptase and can be observed as stops in primer extension (RT-stops) ( Figure 5(A) and Supplementary Figure S5(A) for a biological replicate). Consistent with the data presented above, we observed increased intensities of the primer extension stop signals in D8 as compared to WT in general, which is consistent with slightly elevated levels of tRNA hisR (compare lane 1 and 4). Strikingly, with and without DMS treatment we observed increased RT-stops in the anticodon loop in D8 samples, indicating increased structure in the anticodon loop as compared to WT (Figure 5(B) and Supplementary Figure S5(B)). These stops occurred at G35 and U36, of which the first is modified with queuosine/epoxyqueuosine in the WT, and suggested that the anticodon loop could be more structured in the absence of polyamines ( Figure 5(C)). Apart from RT-stops at G35 and U36, treatment with DMS produced an overall similar pattern, consistent with a similar tRNA secondary structure in the two strains ( Figure 5(A  and B), compare untreated samples in lanes 1 with 4 and treated samples in lanes 2-3 with 5-6).
In conclusion, the structure probing suggested that polyamines did not affect the overall structure of the tRNA, but locally affected the structure of the anticodon loop of tRNA hisR .

Discussion
Here we present data that supports a new fundamental role of the universally conserved polyamines in protein synthesis. By steady-state labelling of macromolecules in exponentially growing cells lacking eight polyamine biosynthesis genes, we find that the synthesis rates are down by 66.5-67.1% in line with the observed effect on growth rate (Figure 1). Strikingly, we calculate an imbalance in the RNA/protein, RNA/DNA and protein/DNA ratios, which could be caused by an underlying effect of polyamines on the translational machinery that leads to increased synthesis of ribosomes and tRNA in the polyamine auxotroph mutant (Figure 1(C-E)). Ribosomal RNA and tRNA synthesis is regulated by stringent control when (p) ppGpp binds to RNA polymerase and decreases rRNA and tRNA transcription. [44][45][46] The starvation alarmone (p)ppGpp is synthesized by stringent factor RelA, which senses decreased tRNA charging within the cell. We did not observe significantly decreased tRNA charging levels in cells lacking polyamines (Figure 3), but instead increased synthesis of ribosomes and tRNA (Figure 4 and Supplementary Figure S4). Taken together with the observed slow translation rate in cells lacking polyamines (Figure 2(B)) we suggest that the hypomodification of the tRNA causes ribosomes to reject otherwise cognate ternary complexes. An increased rejection rate would cause the ribosomes to move slower, but since there is no shortage of ternary complexes with charged tRNA arriving at the ribosomes, RelA is not induced to make (p)ppGpp, resulting in an overproduction of ribosomes and tRNA relative to the growth rate. Consistently, polyamine auxotrophs have previously been reported to have impaired stringent control. 47 A comparable phenotype has been observed for streptomycindependent and pseudo-dependent ribosome mutants that translate at reduced rates due tohyperaccurate proofreading. 35,36 Furthermore, polyamines are required for the activity of streptomycin and for the growth of cells with streptomycin-dependent hyper-accurate ribosomes. 24,48,49 The impaired recognition of ternary complexes was not observed to cause a significant increase in mistranslation (Figure 2). We observed the expression of some proteins to be differentially regulated in D8 (Figure 2(A)). These proteins could be part of the "polyamine modulon" that are translationally regulated in the presence of polyamines. 27,50,51 The main effect of polyamine deficiency was on translation elongation speed (of b-galactosidase), which was~54% slower in D8 than in WT (Figure 2(B)). A reduced elongation speed in D8 could partly be explained by the impaired queuosine/epoxyqueuosine modification in tRNA hisR , tRNA tyrTV and tRNA asnTUVW (Figure 3). Queuosine modifications have previously been reported to increase the translation elongation rate 43,52 and we observed a decreased translation elongation speed in Dtgt (~15%, Supplementary  Figure S2(E)), although it was not as slow as D8. This suggests that decreased queuosine/ epoxyqueuosine modification levels is not the only effect responsible for the slow elongation speed of D8. This is not surprising considering that polyamines can stimulate translation of specific mRNAs, and also stimulate protein synthesis in in vitro translation systems. 27 Furthermore, tRNA hisR , tRNA tyrTV and tRNA asnTUVW are not the only tRNA, which were observed to be affected by polyamines. Indeed tRNA gltTUVW , tRNA leuPVQT and tRNA ileTUV also showed altered migration patterns (Supplementary Figure S3(B)). In line with this interpretation, queuosine/epoxyqueuosine modification of tRNA aspTUV was not affected, which suggests that polyamine effects could be specific to the tRNA and not necessarily the type of modification. This observation also shows that the lack of modification is not due to an indirect effect of disrupting the polyamine biosynthesis genes in D8 i.e. S-Adenosyl methionine is both a substrate of SpeC (S-adenosyl methionine decarboxylase, needed for spermidine synthesis) and QueA (tRNA preQ 1 34 S-adenosylmethionine ribosyltransferase-isomerase, needed for epoxyqueuosine synthesis). More importantly, the modification defect in D8 is reversed by addition of 100 lM spermidine (Figure 3), and thus independent of S-adenosyl methionine decarboxylase. Consistent with the proteome analysis we also observed that the lack of queuosine/epoxyqueuosine on tRNA asnTUVW did not increase mistranslation of asparagine-coding codons in the more sensitive dual-luciferase assay (Supplementary Figure S2(B)).
In eukaryotes, queuosine modifications protect tRNA against ribonuclease cleavage 53 and similarly tRNA his and tRNA tyr were observed to be unstable in T. thermophilus in the absence of polyamines at extreme temperatures (80°C). 21 We did not observe a significant difference in the stability of tRNA in D8 and WT at 37°C (Figure 4). We suggest that while polyamines can stabilize tRNA and rRNA at extreme temperatures, they could also be important for stabilizing structural transitions in tRNA at 37°C that are needed for efficient activity of tRNA-modification enzymes. It should be mentioned here that Nakashima and co-workers did not observe any difference in modified nucleosides in T. thermophilus polyamine auxotrophs by High Performance Liquid Chromatography. 21 Some modifications, including queuosine modifications, were however not observed in the HPLC analysis and potential differences may therefore have missed detection. We want to note a discrepancy between the tRNA stability observed here after rifampicin-treatment of WT cells, and that previously reported by two of the co-authors. 54 The difference is caused, at least in part, by a difference in RNA extraction methods (manuscript in preparation).
Polyamines are known to bind to the anticodon stem of the tRNA, which could have structural effects on the anticodon loop. 17,19 Similarly, modifications in the anticodon loop can make the anticodon more accessible without affecting the secondary structure of the tRNA. 55,56 Indeed, by structural probing of tRNA hisR using DMS we did not observe any significant changes in tRNA secondary structure between WT and D8 ( Figure 5).
We did observe increased RT-stops in the anticodon loop of tRNA hisR , which is consistent with increased structure in the loop in the absence of the queuosine/epoxyqueuosine modification. This is also consistent with the need of polyamines to transiently stabilize the anticodon loop structure to make it accessible for modification enzymes. Intriguingly, it has recently been observed that polyamine auxotrophy is synthetically lethal with tRNA modification enzymes encoded by mnmE and mnmG. 57 The mnmE and mnmG genes are responsible for generating 5-methylaminomethyluridine and 5-carboxymethylaminemethyluridine modifications at U34 of tRNAs that decode NNG codons. The polyamine-dependent queuosine modification is located in the same 3 0 "wobble" position, which could indicate that in the absence of polyaminedependent anticodon modification other modification enzymes become essential, which could explain the observed lethality.
In conclusion, we uncover a new fundamental role of polyamines in tRNA modification, which provides a molecular handle on the complex role that this highly conserved class of molecules play in cellular growth physiology.
Strains and plasmids are described in detail in Supplementary Information. Oligonucleotides used in this study are listed in Supplementary Table S1.

Absolute macromolecule synthesis rates
Synthesis rates were determined by following steady-state labelling of macromolecules by incorporation of 3 H-leucine and 14 C-uracil with known specific activities. To prevent biosynthesis of leucine and uracil affecting incorporation of radiolabels, DleuA and DpyrE were deleted in MG1655 and MG1655 D8. MG1655 DleuADpyrE and MG1655 D8 DleuADpyrE were grown exponentially in 50 mL MOPS MM containing 1.32 mM K 2 HPO 4 , 0.2% glucose, FN19 amino acid mix without leucine, 20 lg/mL leucine and 10 lg/mL uracil at 37°C. At Optical Density, OD 436 = 0.1, 25 mL of each culture was transferred into separate flasks and 8.25 lL 3 H-leucine (42,5 lCi/mmol, 1 mCi/mL) and 3.75 lL 14 Curacil (60 mCi/mmol, 0.1 mCi/mL) was added. The culture without radiolabel was used for OD 436 measurements, which were measured when samples were collected from radiolabeled cultures. 0.5 mL samples were collected at time points indicated and transferred to two tubes, one containing 0.5 mL 5% TCA (trichloroacetic acid) and one containing 0.5 mL 0.5 M NaOH, in an ice bath. NaOH hydrolyses RNA and is used to measure the synthesis of DNA. NaOH-treated samples were incubated at 37°C for two hours before precipitation in 1 mL 10% TCA in an ice bath for 30 min. All TCA-precipitates were transferred to separate Whatman glass microfiber filters GF/C (GE healthcare) by vacuum filtration, which were subsequently washed four times using 5 mL 5% ice-cold TCA. Dried filters were transferred to scintillation vials containing 5 mL scintillation liquid and disintegrations per minute (DPM) were measured by scintillation counting.
Two "specific activity" cultures were used to calculate the specific activity of 3 H-leucine and 14 C-uracil inside the labeled cells. They were prepared as follows: Ten min after addition of radioactivity to the experimental cultures, 100 lL culture was transferred to each of 5 mL medium containing either 20 lg/mL leucine or 10 lg/mL uracil. The medium cultures were left to incubate overnight and 0.5 mL samples were then transferred to 0.5 mL 5% TCA and to 0.5 mL 0.5 M NaOH in an ice bath, and treated as described for the samples above.

Translation elongation speed by bgalactosidase activity measurements
Translation elongation speed was measured by bgalactosidase induction lag, essentially as described in 60. MG1655 and MG1655 D8 were grown exponentially in MOPS MM containing 1.32 mM K 2 HPO 4 , 10 lg/mL uracil, 0.2% glucose and FN20 amino acid mix at 37°C. At To determine the translation elongation speed, bgalactosidase activity was plotted as ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi E t ð Þ À E ð0Þ p against time (sec), where E(0) is the background activity and E(t) is the total activity. The translation time of b-galactosidase is the intersection with the X-axis.

tRNA charging by northern blot analysis
The acylation state and migration of different tRNAs in exponentially growing MG1655, MG1655 D8, MG1655 D8 with 100 lM spermidine and MG1655 Dtgt were determined by the northern blotting method described in 54. Bacterial cultures were grown exponentially in MOPS MM supplemented with 0.2% glucose, 10 mg ml À1 uracil and FN20 amino acids at 37°C. At OD 436 = 0.5, a 1.5 mL sample was mixed with 0.3 mL stop solution (Ethanol + 5% phenol pH 4.3). In parallel, MAS1074 (BL21(DH3)/pSelC) spike-in cells 61 were grown exponentially at 37°C in LB medium containing 100 lg/mL ampicillin and selC transcription was induced with 1 mM IPTG at OD 436 = 0.1. Induced MAS1074 cells were collected after 2-3 hours of expression and were used as whole-cell spike-ins for the samples in stop solution. MAS1074 cells were added to the samples in the amount corresponding to 5% of the sample optical density (OD 436 ). The spiked cell aliquots were collected by centrifugation and the RNA purified by acidic phenol extraction and ethanol precipitation. The tRNA was resuspended in 20 lL tRNA buffer (10 mM Na-Acetate pH 4.5 and 1 mM EDTA). To deacylate the tRNA, 4 lL of resuspended tRNA was treated with 6 lL 1 M Tris-HCl pH 9 and 50 lL dH 2 O for 2 hrs at 37°C, ethanol precipitated and resuspended in tRNA buffer. 4 lL aliquots of acylated and deacylated tRNA samples were mixed with 6 lL loading buffer (0.1 M succinate pH 5, 8 M Urea and 0.05% bromophenol blue and xylene cyanol) and loaded onto a 6% polyacrylamide sequencing gel with 8 M Urea buffered in 0.1 M succinate pH 5. The RNA was separated overnight at 10 W at 4°C. The tRNA was then elektrotransferred to a Hybond N + membrane in 40 mM Tris-Acetate pH 8.1 and 2 mM EDTA. After transfer, the membrane was UV crosslinked and pre-hybridized in 6 mL hybridization solution (0.9 M NaCl, 0.05 M NaH 2 PO 4 (pH 7.7), 5 mM EDTA, 5 Â Denhardt's solution, 0.5% (w/v) SDS and 100 mg/mL salmon sperm DNA) at 42°C. After 1 hr, 30 pmol of 5 0 radiolabeled DNA oligonucleotide probe (see Supplementary  Table S1) (1Â PNK buffer, 1 lL Polynucleotide Kinase, 40 lCi [c 32 P]-ATP, 3 lL 10 lM oligonucleotide in a total volume of 20 lL) was added to the hybridization solution and left to incubate overnight. After hybridization, the blot was washed in 2xSSC + 0.1% SDS and the signal detected by phosphorimaging.
tRNA quantification and stability by northern blot analysis MG1655 and MG1655 D8 cells were cultured, harvested, supplemented with MAS1074 spike-in cells and total RNA extracted as in the northern blot analysis of tRNA charging (see above). At time zero (OD 436 = 0.5) the first sample was collected from cultures in balanced growth, and rifampicin was added to a final concentration of 100 lg/mL. Cell samples were subsequently collected 60, 120 and 180 min after rifampicin addition. 4 lL of total RNA from each sample was denatured in 10 lL formamide loading buffer and briefly boiled before loading onto an 8% (for tRNA) or 4.5% (for rRNA) denaturing polyacrylamide gel containing 8 M urea buffered in Tris-Borate-EDTA (TBE).
The RNA was separated by electrophoresis and transferred to a zetaprobe membrane (Bio-Rad) by electroblotting in TBE buffer. The membrane was crosslinked and subsequently pre-hybridized and hybridized with probe and washed as described for northern blotting of tRNA charging. The radioactive signal was visualized by phosphorimaging. The bands were quantified using imageJ and normalized to the signal measured for tRNA selC .
DMS structural probing by primer extension analysis.
Cell cultures of MG1655 and MG1655 D8 were treated with dimethylsulfate (DMS) essentially as described in 62. The cells were grown exponentially in 15 mL MOPS MM containing 1.32 mM K 2 HPO 4 , 10 lg/mL uracil, 0.2% glucose and FN20 amino acid mix at 37°C. At time zero (OD 436 = 0.4), a 4.5 mL sample was collected and 525 lL DMS was added to the remaining culture which was left to incubate and samples were collected after 2 and 10 min of incubation. DMS reactivity was immediately quenched by transfer of the collected sample to 1.5 mL DMS stop solution (50% b-Mercaptoethanol, 45% Ethanol and 5% phenol) on ice. The sample was washed in 5 mL 30% b-Mercaptoethanol and the cell pellet stored at À80°C . Total RNA was purified by acidic phenol extraction, chloroform extraction and ethanol precipitation. 1 lg of total RNA from each sample was mixed with 0.2 pmol of radiolabeled oligonucleotide. The PE_tRNAHisR-rv oligonucleotide (4 pmol) was phosphorylated using 30 lCi [c 32 P]-ATP and T4 Polynucleotide kinase (Fermentas) and subsequently desalted using a G-25 desalting column (GE healthcare). To facilitate hybridization, the RNA/oligonucleotide mixtures were incubated at 80°C for 5 min, transferred to an ice bath and left to incubate for 5 min. 1 Â FS buffer (Invitrogen), 10 mM DTT and 1 mM dNTP was added to the chilled tube, which was transferred to 54°C and the temperature stabilized by incubation for 2 min. Then 20 U of Superscript III reverse transcriptase (Invitrogen) was added, followed by incubation for 1 hr. Finally, the reaction was terminated by addition of an equal volume of formamide loading buffer. To separate the cDNA, the reactions were briefly boiled and loaded onto a 6% polyacrylamide gel containing 8 M urea and 1 Â TBE. A dideoxy sequencing ladder made from a PCR template generated using oligonucleotides PE_tRNAHisR-f and PE_tRNAHisR-rv was also loaded on the gel. After separation by gel electrophoresis, the gel was fixed (fixing solution: 50% Ethanol and 20% acetic acid), dried and the bands visualised by phosphorimaging.