Inhibiting translation elongation can aid genome duplication in Escherichia coli

Abstract Conflicts between replication and transcription challenge chromosome duplication. Escherichia coli replisome movement along transcribed DNA is promoted by Rep and UvrD accessory helicases with Δrep ΔuvrD cells being inviable under rapid growth conditions. We have discovered that mutations in a tRNA gene, aspT, in an aminoacyl tRNA synthetase, AspRS, and in a translation factor needed for efficient proline–proline bond formation, EF-P, suppress Δrep ΔuvrD lethality. Thus replication-transcription conflicts can be alleviated by the partial sacrifice of a mechanism that reduces replicative barriers, namely translating ribosomes that reduce RNA polymerase backtracking. Suppression depends on RelA-directed synthesis of (p)ppGpp, a signalling molecule that reduces replication-transcription conflicts, with RelA activation requiring ribosomal pausing. Levels of (p)ppGpp in these suppressors also correlate inversely with the need for Rho activity, an RNA translocase that can bind to emerging transcripts and displace transcription complexes. These data illustrate the fine balance between different mechanisms in facilitating gene expression and genome duplication and demonstrate that accessory helicases are a major determinant of this balance. This balance is also critical for other aspects of bacterial survival: the mutations identified here increase persistence indicating that similar mutations could arise in naturally occurring bacterial populations facing antibiotic challenge.


INTRODUCTION
Cell survival from one generation to the next relies on efficient and faithful replication of the genome. However, the template for replication frequently harbours obstacles that have the potential to interfere with the progression of replisome complexes, the macromolecular machines responsible for driving genome duplication. Protein-DNA complexes are important sources of such obstacles and those associated with transcription are especially problematic, in part because of their abundance and in part because they present multiple different challenges to replisome movement (1,2). One major challenge is the very high free energy of binding of transcription complexes that creates a need to disrupt many RNA polymerase-nucleic acid interactions as replication proceeds. Transcribing RNA polymerases (RNAPs) also pause frequently either spontaneously or upon encountering DNA template damage (3). Paused RNAPs can also diffuse backwards along the template resulting in displacement of the transcript 3 end from the RNAP active site. This causes an inability to resume transcription, creating barriers to replication that threaten genome stability (4,5).
Mechanisms exist that reduce the probability of a replication fork encountering paused transcription complexes. Upstream RNAPs on the DNA and ribosomes on the emerging transcript may inhibit backtracking of a paused transcription complex in bacteria (5)(6)(7). RNA translocases such as the bacterial Rho hexamer can also translocate along transcripts not coated with ribosomes and displace paused and blocked RNAPs (5,8,9). Rho translocation also disrupts R-loops, structures in which an RNA transcript hybridizes to the DNA template strand (9). Nucleases can also aid genome duplication by promoting the resumption of transcription by backtracked RNAP (5,10). GreA and GreB bind to E. coli RNAP and stimulate cleavage of backtracked transcripts, creating a new RNA 3 end that can be used to restart transcription (11). Conversely, the 3 -5 helicase UvrD can increase rather than decrease backtracking of RNAP and this may increase access of nucleotide excision repair enzymes UvrABC to DNA damage thus helping to clear barriers to RNAP and aid genome duplication (12).
Signalling molecules such as guanosine tetraphosphate and pentaphosphate in bacteria (referred to collectively as (p)ppGpp) can also reprogramme transcription. (p)ppGpp is central to the stringent response in bacteria, identified originally as a response to amino acid starvation in which the presence of uncharged tRNA within the ribosomal A site stimulates ribosome-bound RelA to synthesize (p)ppGpp (13)(14)(15). Control of the stringent response also requires a (p)ppGpp-specific pyrophosphohydrolase, SpoT, that also has low (p)ppGpp synthase activity (15). However, the synthase activity of SpoT is induced by other stresses such as depletion of fatty acids (16). Central to the stringent response is binding of (p)ppGpp to RNAP which leads to downregulation of stable RNA production and upregulation of stress response genes (15,17,18). Expression of ribosomal rRNA operons accounts for half of all transcription during rapid growth when (p)ppGpp levels are low and thus provides the majority of transcriptional obstacles to replisomes in vivo when cells are dividing quickly (19)(20)(21)(22). Elevated (p)ppGpp therefore reduces conflicts between replication and transcription by decreasing the density of transcribing RNAPs on the genome. (p)ppGpp can also decrease the stability of transcription complexes blocked by template damage in vitro which may decrease the density of blocked RNAP on the genome (10). This destabilization has been questioned, though, and a role for (p)ppGpp in increasing the fidelity of transcription and hence reducing RNAP pausing and backtracking has been proposed (23). Other work implicates (p)ppGpp acting synergistically with UvrD in promoting backtracking of paused RNAP, one function of which could be to facilitate transcriptioncoupled repair of any pause-inducing DNA damage (24). Thus, how (p)ppGpp might act on transcription elongation complexes, as opposed to transcription initiation, is still far from clear. Transcription is not the only target of (p)ppGpp, though, since (p)ppGpp also binds multiple other targets. Inhibited targets include DnaG primase (25)(26)(27)(28) and the translation elongation factors EFG and EF-Tu (29), indicating the pleiotropic impact of elevated (p)ppGpp on replication, transcription and elongation. (p)ppGpp also affects other diverse aspects of metabolism such as phospholipid synthesis, oxidative metabolism and resistance to antibiotics (30). These signalling molecules also enhance the ability of E. coli to persist in the presence of antibiotics. Persistence is the non-heritable ability of a small fraction of a bacterial population to survive exposure to an otherwise lethal concentration of antibiotic (31). (p)ppGpp levels vary stochastically in a bacterial population and increased (p)ppGpp activates toxin-antitoxin systems that induce slow growth in a sub-population of cells, leading to antibiotic tolerance (32).
Mechanisms also exist to increase the probability of continued replisome movement in the event of a collision with an RNAP. Accessory replicative motors provide a supply of additional helicases at the fork to aid protein displacement ahead of the replisome (33)(34)(35)(36)(37). In E. coli Rep helicase promotes movement of replisomes along protein-bound DNA in vitro and in vivo (20,34). Rep appears to be the main replisome anti-pausing factor in E. coli (38) and absence of Rep results in at least a two-fold increase in genome duplication time (39,40) and higher dependence on recombinationdirected repair of damaged forks (41)(42)(43)(44). However, rep cells remain viable since the homologous helicase UvrD can substitute for the absence of Rep at the replication fork (20,34). This substitution is only partial, though, and correlates with a physical and functional interaction between DnaB and Rep but not between DnaB and UvrD (34,45). This partial functional overlap is sufficient for single deletion mutants to be viable during rapid growth whereas rep uvrD cells are not (46). rep uvrD inviability can be suppressed by growth on minimal medium, conditions under which (p)ppGpp levels are high, and also by elevation of (p)ppGpp levels on rich medium via the spoT1 mutation encoding a (p)ppGpp pyrophosphorylase-defective SpoT (20,34). Mutations within rpo genes can also suppress rep uvrD rich medium lethality (20,34,47). These mutant RNAPs display different phenotypes suggesting distinct mechanisms of suppression but some seem to phenocopy elevated (p)ppGpp and/or reduce RNAP backtracking (10,47,48).
The above mechanisms reduce conflicts between replication and transcription but under rapid growth conditions forks are still blocked sufficiently frequently to require replisome reloading enzymes to maintain viability (49,50). Replisome reloading also often requires remodeling of the DNA at the blocked fork by recombination enzymes in order to generate a DNA structure suitable for replisome reloading (51). However, recombinational processing can lead to genome instability and is thus tightly controlled. One control is exerted by UvrD as it can disrupt RecA-ssDNA filaments and this disruption prevents excessive RecFOR-dependent loading of RecA onto ssDNA at blocked forks (52). Such excessive RecA loading contributes to rep uvrD lethality, evinced by the weak suppression of rep uvrD rich medium lethality via mutations in recF, recO or recR (34,53,54).
We have searched for suppressors of rep uvrD rich medium lethality that are not within rpo, spoT, recF, O or R. We identified a spontaneous suppressor in a tRNA gene, aspT [t8c], that mutated a highly conserved residue that is structurally important in other tRNA species. Probing the basis of this suppression revealed that defective tRNA aminoacylation or inefficient peptide bond formation within the ribosome also suppressed rep uvrD lethality. In all cases, suppression required RelA-directed (p)ppGpp synthesis indicating that stalling of ribosome translocation by uncharged tRNA in the A site of ribosomes underpinned suppression. Thus, although efficient translation elongation aids genome duplication (5,7), the partial inhibition of ribosome translocation is more effective at mitigating replication-transcription conflicts. These data illustrate the fine balance between the multiple mechanisms that promote simultaneous gene expression and genome duplication and reveal the importance of accessory replicative helicase activity in determining this balance.

Genome sequencing and analysis
DNA was extracted from 1 culture each of strain N7153 and N7182 grown in LB broth to stationary phase using Qiagen 100/G genomic tips from 5 ml cultures following the manufacturer's protocol. Genome sequencing was performed using an Illumina GAIIx instrument with 100 bp paired end reads. Paired reads were trimmed to remove adapters and mapped against the E. coli K12 strain MG1655 NC 000913 genome using bwa (57), duplicates marked using Picard (http://broadinstitute.github. io/picard) and variant analysis performed with SAMtools (58), followed by merging of variant tables using perl. Identified high quality synonymous and non-synonymous single nucleotide polymorphisms were annotated manually using the Integrative Genomics Viewer (59). The Illumina data were submitted in the form of fastq files to the European Nucleotide Archive (ENA) and are available under accession number PRJEB14483 at http://www.ebi.ac.uk/ena/ data/view/PRJEB14483.

Growth assays
All steps in plasmid loss assays were carried out at 37 • C except those shown in Figures 2B and 3B in which all steps were conducted at 30 • C or 25 • C as indicated. Colonies were grown in LB broth and agar except the assays in Figure 2B and Supplementary Figures S3B and S4 which were performed with a defined rich medium broth and agar containing 0.2% glycerol ( Figure 2B) or 0.2% glucose (Supplementary Figures S3B and S4) as a carbon source (60). The plates used in Supplementary Figures S3B and S4 also had decreasing concentrations of aspartate or phenylalanine, as indicated. Strains carrying derivatives of pRC7 were grown in LB broth or defined rich medium with 100 g ml −1 of ampicillin overnight, diluted 100-fold into the same type of fresh liquid medium without ampicillin and grown to A 650 0.4. Dilutions were then plated onto LB or defined rich medium agar plates containing 120 g ml −1 X-gal and 1 mM IPTG and incubated for 48 h except those shown in Figure 2B. Plates were then photographed and scored for blue/white colony formation.
For assays to assess colony forming ability, strains were grown in LB broth overnight at 37 • C or, when temperature sensitive strains were involved, at 30 • C. Serial 10-fold dilutions were made with 56/2 salts (61) on ice and then 5 l of each dilution was spotted onto LB agar plates. Plates were then incubated at 37 • C for 16 h unless otherwise stated. Ampicillin and bicyclomycin were included in LB agar plates where indicated at 100 and 25 g ml −1 , respectively.
For the colony formation assays in Supplementary Figure  S3A, strains were grown in defined rich medium broth (60) containing 0.2% glucose, all amino acids and 100 g ml −1 ampicillin overnight at 37 • C. Tenfold serial dilutions were made in 56/2 salts and then 5 l of each dilution spotted onto defined rich medium plates containing 0.2% glucose, 100 g ml −1 ampicillin and with all amino acids or missing either aspartate or phenylalanine. These plates were incubated at 37 • C for 16 h. Minimal medium agar plates (61) were used in Supplementary Figure S3C without and with 100 g ml −1 ampicillin as indicated and incubated at 37 • C for 72 h.
Colony-forming ability at increasing doses of UV light was assessed as described (62). Mismatch repair capacity was measured as the fraction of cells in a culture that acquired spontaneous mutations leading to rifampicin resistance. Briefly, overnight LB cultures were washed once in 56/2 salts and serially diluted 10-fold. 100 l of the neat and the 10 −1 dilution were spread on LB agar containing 15 g ml −1 rifampicin. To estimate the total cfu ml −1 in the overnight culture, 10 l of the 10 −6 dilution was spotted in triplicate on LB agar without rifampicin. Plates were then incubated at 37 • C for 16 h and the colonies then counted.
The liquid growth assays in Figure 2A were performed using a Tecan Infinite M200 Pro plate reader. Overnight LB cultures were diluted to A 600 0.005 in LB and 150 l of each culture was transferred into 20 wells of a 96-well flat bottom plate (Corning). The microplate was incubated at 37 • C for 24 h and A 600 measurements were performed every 30 min preceded by plate shaking.

Persistence
Single colonies were inoculated into 5 ml LB broth and grown with shaking at 37 • C overnight. 50 l of overnight culture was then inoculated into 5 ml of LB broth and grown at 37 • C with shaking until 1-2 × 10 8 colony forming units ml −1 reached. 500 l was then transferred into two 1.5 ml microcentrifuge tubes. The first tube was centrifuged at 6000 rpm for 5 min at room temperature, resuspended in 500 l 56/2 salts, serially diluted tenfold in 56/2 salts and 10 l volumes of the 10 −4 and 10 −5 dilutions spotted onto LB agar plates containing 20 mM MgSO 4 in triplicate. To the second tube 5 l of 10 g ml −1 ciprofloxacin was added and the tube was then inserted into a 50 ml Falcon tube and placed in a shaking incubator at 37 • C for 5 h. After the 5 h incubation, this tube was centrifuged at 6000 rpm for 5 min at room temperature, resuspended in 500 l 56/2 salts, recentrifuged and resuspended in 500 l 56/2 salts. Colonyforming units were then assayed as for the first tube by serial dilution into 56/2 salts and plating in triplicate onto LB agar containing 20 mM MgSO 4 . MgSO 4 was included to inhibit the activity of any traces of ciprofloxacin not removed by washing of the cells (63). All plates were incubated at 37 • C overnight and then the fraction of colony forming units surviving ciprofloxacin exposure calculated.

Flow cytometry
Analysis of chromosome content after completion of all ongoing rounds of replication, so-called run-out conditions, was performed on mid-log phase cultures after treatment with rifampicin and cephalexin as described (64) using a Becton Coulter CyAn ADP cytometer with 488 nm excitation and a 530/540 nm bandpass filter. Analysis of (p)ppGpp-dependent formation of RpoS-mCherry by flow cytometry was performed by growing the indicated strains overnight in LB, transferring 100 l into 10 ml LB in a 125 ml flask followed by incubation at 37 • C for 2 h. Then 1 ml of the culture was centrifuged, resuspended in 2 ml of 10% LB in M9 medium and analysed by flow cytometry on a Becton Dickinson LSRFortessa X-20 flow cytometer using 561 nm laser excitation and dection using a 610/620 nm bandpass filter.

Translation assays
␤-Galactosidase activity assays monitoring the relative levels of translation of cadC-lacZ fusions were performed in LB broth as described (56).

A tRNA mutation suppresses the requirement for accessory replicative helicases
pRC7 is a very low copy plasmid which encodes ␤ lactamase and can therefore be maintained in E. coli cells by the inclusion of ampicillin in the medium (65). However, the inefficient origin of replication within pRC7 results in rapid loss of the plasmid in the absence of ampicillin. This rapid loss can be detected if the host strain has a chromosomal deletion of lacIZYA since pRC7 encodes the lac operon and hence cells with and without the plasmid appear blue and white, respectively, on plates containing IPTG and X-gal (65). This retention or loss of pRC7 can be used to assess whether a gene is essential for colony formation by cloning of the test gene into pRC7, transformation of a lacIZYA strain with the pRC7 derivative and subsequent deletion of the test gene from the chromosome. Plating of the strain onto medium containing IPTG and X-gal results in formation of only blue colonies if the test gene is essential or white and segregating colonies if the test gene is not essential (65). Rapid growth of E. coli requires accessory helicase activity and so pRC7 encoding uvrD can be lost rapidly from lacIZYA rep + uvrD + cells on rich medium but pRC7uvrD cannot be lost from lacIZYA rep uvrD cells as monitored by blue/white colony colour (34) (see also Figure 1A, compare i and ii). Spontaneous mutations that suppress this requirement for an accessory helicase can be isolated by exploiting the ability of rep uvrD cells to grow on minimal medium in the absence of a complementing pRC7 plasmid and subsequent plating of plasmid-less cells onto rich medium (34). Rare survivors on rich medium can then form colonies and the mutation(s) responsible for allowing rep uvrD cells to grow under rapid growth conditions can be analysed. We identified one such suppressor, the strain designation of which is N7182 (Supplementary Table S1). Potential linkage of the suppressor mutation to rep was tested by transducing rep::cat from the suppressor strain into pRC7uvrD/rep + ΔuvrD::dhfr (N6639). Eleven chloramphenicol-resistant transductants were tested for loss of pRC7uvrD on rich medium. Four transductants could not lose pRC7uvrD but seven could, indicating close linkage of the suppressor mutation with rep::cat (see also Figure 1A, compare ii and iii).
We sequenced the genome of this suppressor strain and compared it with the genome of N7153, a rep uvrD strain that contains a well-characterized suppressor mutation rpoB*35 (5,10,34,66). Use of the rep uvrD rpoB*35 strain as a reference genome avoided the need for the reference strain to retain a complementing plasmid for viability. Only one mutation in N7182 that was not present in N7153 was located sufficiently close to rep::cat to explain the above linkage (see Supplementary File 1). This mutation resulted in replacement of T with C at position 8 within the aspT gene, one of three identical tRNA Asp genes in E. coli. Sequencing of the aspT gene from the 11 transductants obtained in the above cross revealed that all seven strains able to lose pRC7uvrD contained aspT[t8c] whereas all four that could not lose pRC7uvrD retained a wild type copy of aspT. Suppression of the inviability of rep uvrD cells on rich medium was therefore associated with the aspT[t8c] allele, with suppression comparable to that of rpoB*35 in a plasmid loss assay ( Figure 1A, compare iii and iv).
rep uvrD + cells have a growth defect defect since UvrD can compensate only partially for the absence of Rep accessory helicase activity, resulting in slower movement of replication forks in rep cells (39,40). Thus, the median number of copies of oriC in rep uvrD + cells is twice that of rep + uvrD + cells in rich medium due to an extended cell cycle in rep cells and hence more replication initiation events per cell cycle (38). This doubling in oriC numbers results in a doubling of chromosome content when cells are treated with cephalexin and rifampicin to inhibit cell division and reinitiation of replication, so-called runout conditions (38) (see also Figure 1B, compare i and iii). aspT[t8c] suppressed the increased chromosome copy number in rep uvrD + cells, reducing the median number of chromosomes from eight to four ( Figure 1B, compare iii and iv). These data support the conclusion that this tRNA mutation reduces the need for accessory helicase activity. Furthermore, aspT[t8c] had no detectable impact on other UvrD-mediated processes. Defects in nucleotide excision repair, mismatch repair and control of recombination in rep + uvrD cells (52,67,68) were not suppressed by aspT[t8c] (Supplementary Figure S1A, B and C, respectively), providing further support for the specific suppression of the accessory helicase defect in rep uvrD cells by aspT [t8c].
The requirement for accessory helicase activity is driven primarily by transcriptional barriers to replication (20,34). Backtracking of paused RNA polymerases results in formation of particularly stable replicative barriers and multiple factors have evolved to reduce the numbers of backtracked complexes (5). Cells lacking two homologous anti-backtracking factors, GreA and GreB, display a temperature-sensitive growth defect that is a consequence of more frequent collisions between replisomes and backtracked transcription complexes (5,10,69). aspT [t8c] suppressed the temperature-sensitive phenotype of greA greB cells (Figure 1C), similar to the suppression seen with rpoB*35 (48). We conclude that aspT[t8c] reduces the need for GreA/GreB-dependent rescue of backtracked RNA polymerase.
Taken together, these data indicate that aspT[t8c] reduces the need for Rep and UvrD to underpin replication and for anti-backtracking factors to resuscitate transcription complexes.

Suppression by aspT[t8c] does not occur via growth rate restriction
The uridine encoded at position 8 within the wild type aspT gene is post-transcriptionally modified to 4-thiouridine and this modified nucleotide is conserved across all kingdoms of life (70). This residue is involved in a triple non-Watson-Crick pairing interaction and is important in coordination of magnesium within tRNA (71,72). This central structural role is reflected in the temperature-dependent destabilisation of human mitochondrial tRNA Met structure by the same T to C mutation as found in aspT[t8c]. This transition mutation in tRNA Met inhibits aminoacylation and any mutated tRNA Met that is aminoacylated fails to form a stable ternary complex with elongation factor EF-Tu (70).
Given the conservation of tRNA structure, aspT[t8c] may result in similar structural destabilisation of the encoded tRNA Asp and consequent inhibition of interactions with aspartyl tRNA synthetase and EF-Tu. However, aspT[t8c] did not have a major impact on growth of rep + uvrD + cells in liquid culture (Figure 2A, compare i and ii). This absence of a significant growth defect in aspT[t8c] strains may be due to the presence of two other identical tRNA Asp genes in E. coli, aspU and aspV.
Other means of growth restriction did not suppress rep uvrD inviability. Growth on defined rich medium containing all 20 amino acids but with a poor carbon source, glyc- erol, did not result in suppression of rep uvrD aspT + inviability at either 37 • C or 25 • C as determined using a plasmid loss assay ( Figure 2B). Reduced growth rates therefore do not provide a general means of reducing the need for accessory helicase activity.

Defective tRNA aminoacylation suppresses rep uvrD inviability
As reduced growth rate was not the cause of suppression by aspT[t8c], we probed whether defective translation was responsible. We could not detect significant defects in translation in vivo at tandem aspartate codons in an aspT[t8c] strain (Supplementary Figure S2). Either this allele does not result in translational pausing or such pausing is below the limits of detection using this assay. We therefore used alternative approaches to probe the aspT[t8c] suppression mechanism. The same t8c mutation in human mitochondrial tRNA Met inhibits aminoacylation (70). Inhibition of tRNA Asp aminoacylation was therefore tested for suppression of rep uvrD lethality. The E. coli tls-1 allele encodes a P555S mutation in aspartyl tRNA synthetase that reduces the thermal stability of the synthetase and causes a severe reduction in growth rate at 42 • C under low salt conditions (73,74) (see also Figure 3A). Given that AspRS P555S is less stable than wild type enzyme regardless of the tem-perature (74), we tested whether aspS P555S suppressed rep uvrD lethality at 30 • C on low salt medium, conditions under which viability is similar to wild type (73) (see also Figure 3A, compare i and ii). pRC7uvrD could be lost from rep uvrD aspS P555S at 30 • C on low salt medium but not on high salt medium (Figure 3Biii). Suppression of rep uvrD lethality correlates therefore with a defect in aspartyl tRNA synthetase.
We also tested whether an alternative means of inhibiting tRNA Asp aminoacylation could suppress rep uvrD inviability. A pRC7uvrD/ rep uvrD strain defective in aspartate biosynthesis could lose the complementing pRC7uvrD plasmid on defined rich medium upon restriction of aspartate availability (Supplementary Figure S3). Thus a defect in the enzyme needed to synthesize aspartyl tRNA, AspRS, or a limiting concentration of one of the substrates needed for formation of aspartyl tRNA, aspartate, can suppress the need for accessory replicative helicase activity. Moreover, suppression was not specific to aspartate starvation as restriction of availability of phenylalanine also allowed rep uvrD cells to survive in the absence of a complementing plasmid (Supplementary Figure S4). As expected, restricting amino acid availability also resulted in very poor growth (Supplementary Figures S3 and S4) but a restricted growth rate does not by itself provide suppression of rep uvrD lethality ( Figure 2B). These data support the conclusion that inhibition of aminoacylation of tRNA Asp , and of other tRNAs, reduces the need for accessory helicase activity.

Translational pausing suppresses rep uvrD inviability
Suppression of rep uvrD inviability via inhibition of tRNA Asp or tRNA Phe aminoacylation might occur directly via pausing of ribosomes but could also occur via decreased production of one or more specific proteins containing both aspartate and phenylalanine. We tested therefore whether increased translational pausing by a well-defined mechanism that does not rely on decreased tRNA aminoacylation also suppresses the need for accessory helicase activity. Peptide bond formation by ribosomes occurs with low efficiency within polyproline tracts as compared with other amino acids and this low efficiency is compensated for by interaction of elongation factor P (EF-P) with the ribosome (56,75). When EF-P is absent ( efp) ribosomes pause for extended periods at tandem proline codons (56,75) (see also Supplementary Figure S2) resulting in reduced growth rates (76) (see also Figure 2A). Despite its negative effect on growth rate, introduction of efp clearly suppressed rep uvrD lethality (Figure 4, compare A and B). yjeA and yjeK encode enzymes needed for post-translational modification of EF-P to form functional enzyme (56,(75)(76)(77)(78) and deletion of either yjeA or yjeK also resulted in suppression ( Figure  4C and D). Enhancement of ribosomal pausing at polyproline sequences can therefore reduce the need for accessory helicase activity.

Suppression by aspT[t8c] and efp requires (p)ppGpp synthesis
One consequence of ribosomal pausing is increased synthesis of (p)ppGpp by RelA upon binding of RelA to a ribo-Nucleic Acids Research, 2017, Vol. 45, No. 5 2577   some containing a non-acylated tRNA in the A site (13,14). Elevated (p)ppGpp is known to suppress rep uvrD inviability (34) and so aspT[t8c] and efp might both therefore suppress via elevation of (p)ppGpp concentration.
Direct measurement of (p)ppGpp using 32 P cannot be performed on cells grown in rich media (79) which prevents direct assessment of (p)ppGpp levels in aspT [t8c] and efp strains under conditions relevant to suppression of rep uvrD lethality. An indirect assay was therefore used that employs an RpoS-mCherry translational fusion (32). This reporter provides a fluorescence signal that correlates with intracellular (p)ppGpp concentration due to (p)ppGpp both stimulating rpoS transcription and inhibiting RpoS degradation (32). Fluorescence was assayed by flow cytometry of cells grown to mid-logarithmic phase in rich medium. relA + spoT + cells lacking the RpoS-mCherry fusion and rpoS-mCherry relA spoT cells which are unable to synthesize (p)ppGpp gave similar levels of background fluorescence (Figure 5Ai and ii). An increase in fluorescence was observed in wild type cells carrying the fusion but this increase was modest ( Figure 5A, compare iii with i and ii), consonant with low (p)ppGpp levels in wild type cells growing in nutrient-rich environments (32,80). mCherry fluorescence increased substantially in efp cells but not in aspT[t8c] cells with respect to wild type (Figure 5A, compare iv and v with iii). We conclude that absence of EF-P results in elevation of (p)ppGpp concentration. Higher (p)ppGpp concentration in efp cells as compared with wild type or aspT[t8c] cells is consistent with the significant retardation of growth of efp cells (Figure 2A). In contrast, aspT[t8c] either does not increase (p)ppGpp levels as compared with wild type cells or any increase is below the limits of detection using this assay. To gauge the sensitivity of this assay we monitored fluorescence in spoT1 cells. spoT1 encodes a SpoT enzyme that retains (p)ppGpp synthase activity but lacks (p)ppGpp pyrophosphorylase activity, resulting in elevated (p)ppGpp (81) and the ability to suppress rep uvrD lethality (34). spoT1 rpoS-mCherry cells did not result in increased fluorescence as compared with spoT + rpoS-mCherry cells indicating that (p)ppGpp levels sufficient to reduce the need for accessory replicative helicases could go undetected using this assay (Figure 5Aiii and vi).
To address this potential sensitivity problem we employed a second assay to determine whether either efp or aspT[t8c] results in elevated (p)ppGpp. Cells with elevated (p)ppGpp display elevated levels of persistence and thus spoT1 enhances whereas relA spoT alleles reduce persistence (32,82) (see also Figure 5Bi-iii). aspT[t8c] and efp both increased persistence with the increase being higher for efp (Figure 5Biv and v). The large increase in persistence in efp cells correlates with the enhancement of fluorescence in efp rpoS-mCherry cells. The level of persistence in aspT[t8c] cells is consistent with a smaller increase in (p)ppGpp levels as compared with efp that is below the limits of detection in the mCherry fluorescence assay.
We tested whether increased (p)ppGpp contributed to suppression of rep uvrD inviability by aspT[t8c] and efp by deleting the gene encoding the primary (p)ppGpp synthase, relA. Deletion of relA prevented loss of pRC7uvrD from rep uvrD aspT[t8c] and rep uvrD efp cells, indicating that RelA was required for suppression by both mutant alleles ( Figure 5C, compare iii with iv and v with vi). In contrast, pRC7uvrD could be lost from rep uvrD relA rpoB*35 cells indicating that RelA was not required for the viability of suppressed rep uvrD strains under all circumstances ( Figure 5C, compare vii and viii). This lack of dependence of rep uvrD rpoB*35 on RelA likely reflects the recapitulation by rpoB*35 of many phenotypes associated with elevated (p)ppGpp even in the absence of RelA (48,66).
Taken together, these data indicate that RelA-directed synthesis of (p)ppGpp is important for suppression of rep uvrD lethality by both aspT[t8c] and efp.

aspT[t8c] and Δefp confer differing requirements for Rho activity
Translocation 5 -3 by Rho along untranslated and unstructured nascent transcripts can displace transcription complexes thus reducing both RNA polymerase occupancy on the chromosome and R-loop formation (9,83,84). Rep and Rho therefore provide two different mechanisms that reduce the impact of transcription on replication. Consequently, cells need either Rep or wild type levels of Rho activity to maintain genome duplication in the face of transcriptional barriers (8,85). One manifestation of this requirement is the hypersensitivity of rep uvrD + cells to low concentrations of the Rho-specific inhibitor bicyclomycin (8) (see also Figure 6A and B, compare i and iii). We assessed whether efp or aspT[t8c] could suppress rep bicyclomycin hypersensitivity. Neither efp nor aspT[t8c] altered bicyclomycin sensitivity in a rep + background ( Figure 6A and B, compare i and ii). In rep cells hypersensitivity was suppressed by efp but not by aspT[t8c] ( Figure 6A and B, compare iii and iv). These data demonstrate that efp but not aspT[t8c] can reduce the requirement for Rho activity in the absence of Rep.
We also tested whether this differential need for Rho activity was reflected in cells lacking both rep and uvrD by exploiting rho[A243E] which encodes a partial loss of function enzyme (86). Rho dependency was assessed by the ability of rep uvrD strains to lose a complementing pRC7rep rather than pRC7uvrD plasmid, allowing construction of pRC7rep/ rep uvrD rho[A243E] strains regardless of rep rho double mutant lethality (85). The similar numbers and sizes of white plasmidless colonies formed by rep uvrD efp rho + and rep uvrD efp rho[A243E] cells indicate that wild type Rho function was not required for suppression by efp ( Figure 6C, compare v and vi). In contrast, rep uvrD aspT[t8c] rho[A243E] gave much smaller white plasmidless colonies as compared with the isogenic rho + strain indicating that wild type Rho function was important for growth of aspT[t8c] rep uvrD cells ( Figure 6C, compare iii and iv).
The data in Figures 5 and 6 indicate that the viability of aspT[t8c] rep uvrD cells requires both the major (p)ppGpp synthase in the cell, RelA, and wild type Rho activity. In contrast, efp rep uvrD cells require RelA but not wild type Rho activity. efp cells have higher levels of (p)ppGpp as compared with aspT[t8c] ( Figure 5A) and these elevated levels might explain the differential requirement for Rho, given the ability of (p)ppGpp to reduce replication/transcription conflicts (10,34,66). Such a model implies that (p)ppGpp synthesis is critical not only for the viability of efp rep uvrD cells ( Figure 5C) but also for suppression of bicyclomycin sensitivity of rep cells ( Figure 6A). Absence of the primary (p)ppGpp synthase RelA did not hypersensitize otherwise wild type cells to bicyclomycin (Figure 7Aiii). However, the suppression of rep bicyclomycin hypersensitivity by efp was abolished upon deletion of relA ( Figure 7A, compare viii with v). Thus RelA-dependent (p)ppGpp synthesis in rep efp cells is essential for survival with lowered Rho activity, supporting the hypothesis that elevated (p)ppGpp can reduce the need for Rho.
We tested this possible link between (p)ppGpp concentration and requirement for Rho by analysing the impact of other means of increasing (p)ppGpp concentration on rep bicyclomycin hypersensitivity. The spoT1 allele suppressed rep bicyclomycin hypersensitivity supporting the conclusion that elevated (p)ppGpp can reduce the need for Rho ( Figure 7B, compare iv with ii).
The data in Figures 5-7 indicate that both aspT[t8c] and efp require RelA-directed synthesis of (p)ppGpp to maintain the viability of rep uvrD cells. In contrast, efp has a much lower dependence on Rho activity as compared with aspT[t8c]. This differential requirement for Rho activity correlates with higher (p)ppGpp levels in efp cells as compared with aspT [t8c]. Resolving conflicts between replication and transcription involves therefore a fine balance between accessory replicative helicases, levels of (p)ppGpp and Rho activity.

DISCUSSION
We have discovered three new types of suppressors of rep uvrD rich medium lethality: mutations in tRNA genes, in aminoacyl-tRNA synthetases and in translation elongation factors. Whilst the mechanistic consequences of the aspT[t8c] mutation are unclear, the inhibition of tRNA aminoacylation by aspS P555S and of translation elongation by efp are well-characterized (56,74,75), indicating that partial inhibition of translation can suppress rep uvrD lethality. Previous studies have demonstrated that this lethality is caused primarily by the conflict between replication and transcription (20,34,54). Thus partial inhibition of translation can compensate for the impact of transcription on DNA replication in the absence of Rep and UvrD. Suppression requires RelA-directed synthesis of (p)ppGpp with (p)ppGpp concentration being higher with efp than with aspT[t8c] ( Figure 5). Wild type levels of Rho activity are also needed in aspT[t8c] rep uvrD but not efp rep uvrD cells indicating that the relative importance of (p)ppGpp and Rho function depends upon the nature of the translational mutation ( Figure 6). rep uvrD cells are inviable on rich medium primarily because of lack of accessory replicative helicase function (20,34,54). Suppression of this growth defect by the above translation mutations ( Figures 1A, 3B and 4) therefore suggest that it is the lack of accessory helicase activity in rep uvrD cells that is being suppressed. Suppression by aspT[t8c] of the chromosome content defect of rep cells, a direct consequence of slower fork movement in the absence of Rep accessory helicase activity (34,38,39), supports this view ( Figure 1B). UvrD also inhibits RecFORdependent loading of RecA at blocked replication forks (52,68) and absence of this function makes a minor contribution to rep uvrD lethality (20,34,53). However, the lack of suppression of this RecA displacement defect by aspT[t8c] (Supplementary Figure S1C) indicates that suppression of rep uvrD lethality by aspT[t8c] does not operate via an effect on RecA loading. UvrD might also aid replication of transcribed DNA by inducing backtracking of RNAP stalled at DNA lesions, one consequence of which may be to promote repair of the lesion and allow the RNAP to then continue transcription (12). However, rep uvrA cells, lacking nucleotide excision repair, are viable (54) indicating that lesion repair via UvrD-catalysed backtracking of RNAP cannot be a major contributor to rep uvrD lethality. aspT[t8c] also does not suppress the sensitivity of uvrD cells to UV light indicating that aspT[t8c] does not suppress nucleotide excision repair defects (Supplementary Figure S1A). Furthermore, aspT[t8c] suppression of the greA greB growth defect indicates that aspT[t8c] can suppress the absence of an anti-backtracking activity, making it difficult to explain how aspT[t8c] could also suppress the absence of a UvrD pro-backtracking function ( Figure  1C). RelA-directed synthesis of (p)ppGpp plays a central role in maintaining rep uvrD viability by aspT[t8c] and by efp ( Figure 5C). (p)ppGpp dramatically inhibits transcription of rrn operons (17) which are the primary transcriptional barriers to genome duplication (19,22), together with other highly expressed operons (18). (p)ppGpp may also destabilize stalled RNAP (10) or increase the fidelity of transcription (23), both of which could decrease the impact of transcription on fork movement. aspT[t8c]-and efpdirected suppression of rep uvrD lethality via (p)ppGpp is therefore likely to occur by impacting on transcription initiation, stalled RNAP stability and/or decreased pausing due to enhanced fidelity.
RelA is stimulated to synthesize (p)ppGpp when an uncharged cognate tRNA is bound to the ribosomal A site (13,14). aspT[t8c] and efp mutations may therefore increase the probability of A site-bound uncharged tRNA either directly or indirectly. The uridine at position 8 is highly conserved in tRNA species and the equivalent t8c mutation within human mitochondrial tRNA Met results in inhibition of tRNA aminoacylation (70). If aspT[t8c] also resulted in inhibition of aminoacylation then the resulting increase in non-acylated tRNA Asp could lead to an increased probability of uncharged tRNA Asp occupying the ribosomal A site. However, whether the mutant tRNA Asp can still bind to the A site is unknown. Alternatively tRNA asp[t8c] might poison the aspartyl tRNA synthetase by binding to it and forming a dead end complex with respect to aminoacylation. Formation of such a dead end complex might titrate out aspartyl tRNA synthetase and increase levels of uncharged wild type tRNA Asp leading to increased non-acylated tRNA Asp occupying the ribosomal A site. For cells lacking EF-P, YjeA or YjeK it is difficult to conceive how a reduced rate of prolineproline bond formation within the ribosome could lead directly to increased occupancy of the A site by uncharged tRNA. Absence of any one of these three factors, though, does impact on expression of many genes and so altered expression of one or more genes in efp, yjeA or yjeK cells could result in an increased probability of A site-bound uncharged tRNA. For example, efp cells have reduced expression of valyl tRNA synthetase (87) which could lead to accumulation of uncharged tRNA Val and triggering of the stringent response. Suppression of rep uvrD lethality by aspS P555S demonstrates that suppression via partial loss of tRNA synthetase function can occur (Figure 3).
In contrast to the requirement for RelA to sustain suppression of rep uvrD lethality by both aspT [t8c] and efp, the requirement for Rho activity is reduced in efp as compared with aspT[t8c] cells (Figures 6 and 7). This difference correlates with the higher concentration of (p)ppGpp in efp as compared with aspT[t8c] cells (Figure 5). A role for (p)ppGpp in reducing the need for Rho activity is supported by the ability of spoT1 to suppress rep bicyclomycin hypersensitivity ( Figure 7B). Suppression of rep bicyclomycin hypersensitivity by efp also depends on RelA which supports a critical balance between (p)ppGpp concentration and Rho activity in maintaining viability ( Figure 7A). Elevated (p)ppGpp can therefore reduce the need for Rho.
The greater dependence of aspT[t8c]-directed suppression on Rho might reflect not just lower (p)ppGpp levels in aspT[t8c] versus efp cells but possibly also enhancement of Rho binding on nascent transcripts by aspT [t8c]. In other words, suppression via aspT[t8c] might occur at least partly via the increased generation of Rho binding sites on emerging transcripts. Aspartate codons are more frequent than polyproline tracts and hence aspT[t8c] has the potential to impact on ribosome translocation more frequently than efp. Enhancement of Rho binding would also be dependent only on ribosomal pausing and not specifically require ribosomal A site occupancy by a non-aminoacylated tRNA (88), in contrast to stimulation of RelA activity (13,14). However, any aspT[t8c]-directed increase of transcription complex displacement by Rho cannot be sufficient by itself to suppress rep uvrD lethality since RelA is also needed ( Figure 5C).
Our data indicate that whilst translation helps prevent RNAP backtracking (5,7) the reduction of conflicts between replication and transcription can be achieved more effectively by partial inhibition of translation. Of course, the growth defect in efp cells (Figure 2A) indicates why the balance between replication, transcription, translation and (p)ppGpp synthesis is poised as it is in wild type cells. Thus for the maintenance of rapid growth there is a very fine balance to be struck between gene expression and accurate, rapid genome duplication. Furthermore, accessory replicative helicases play a major role in determining this balance. In the absence of both Rep and UvrD the wild type balance between various other mechanisms that reduce the impact of transcription on replication is unable to effectively counter the adverse effects of transcription on completion of genome duplication. There may also be circumstances under which mutations such as those identified here confer a selective advantage even in rep + uvrD + cells. Elevated (p)ppGpp is a key factor in determining bacterial persistence in the face of antibiotic challenge (32) and both aspT[t8c] and efp mutations enhance persistence (Figure 5B). It remains possible therefore that mutations such as aspT[t8c] that have only a modest inhibitory effect on growth could arise in bacterial populations continually exposed to antibiotics, especially given the many mutations known to affect translation (89). Increased persistence in strains lacking EF-P function also imply that EF-P and its unique post-translational modification pathway are poor potential targets for antibiotics.

ACCESSION NUMBERS
Sequencing data were submitted in form of fastq files to the European Nucleotide Archive (ENA) and are available under accession number PRJEB14483 at http://www.ebi.ac. uk/ena/data/view/PRJEB14483.