Promoter-proximal transcription–translation coupling controls early transcription in Escherichia coli

Abstract

The functional coupling of transcription and translation contributes significantly to maintaining messenger RNA (mRNA) expression in bacterial cells. Premature transcription termination and fast mRNA decay are known to limit the expression of mRNAs when transcription is decoupled from translation. Here, we report that inhibiting the generation of untranslatable mRNAs from the promoter-proximal region is a newly identified but essential pathway of mRNA quality control by transcription–translation decoupling. The promoter-proximal region of mRNAs, the amount of which reflects early transcription in the 5′-untranslated region, is not generated without translation. The decoupling between transcription and translation results in RNA polymerase (RNAP) traffic within 250 bp from the transcription start site, hindering productive early transcription. The limited processivity of RNAP without a coupled ribosome in the promoter-proximal region is further supported by the observation that decoupled RNAP elongates mRNA by only 80–90 bp on average in vivo. Our results demonstrate that ribosome coupling near the promoter-proximal region is critical for the efficient synthesis of translatable mRNAs by RNAPs.

Graphical Abstract

Graphical Abstract.

Introduction

In prokaryotes, transcription and translation occur simultaneously on the same messenger RNA (mRNA) transcript due to the lack of spatial separation between transcriptional and translational machinery [1, 2]. A ribosome begins translation on a nascent mRNA shortly after an RNA polymerase (RNAP) synthesizes a ribosome binding site (RBS). The transcription elongation complex and the leading ribosome are in close contact [3], forming an RNAP–mRNA–ribosome complex accompanied by other transcription factors, such as NusG and NusA [4–6]. The physical interaction between RNAP and the pioneering ribosome may lead to the functional coupling of transcription and translation processes, such as tight coordination between the elongation rate of transcription and translation [7, 8]. Additionally, the following ribosome can physically push the arrested transcription elongation complex and maintain its transcriptional activity [9–11].

Arguably, the most crucial function of transcription–translation coupling (TTC) in cell maintenance is mRNA quality control [2]. Poor translation initiation efficiency reduces not only the expression level of proteins but also that of full-length mRNAs [12–15]. The correlation between mRNA expression level and translation initiation efficiency has generally been explained by premature transcription termination (PTT) and enhanced mRNA decay [12–14]. PTT occurs during the elongation of RNAPs when transcription and translation are decoupled [8]; transcription termination occurs by intrinsic terminators or a Rho factor [16–18]. On the other hand, mRNA degradation becomes faster when TTC is disrupted, as the steric protection of mRNA by ribosomes from RNases disappears [19, 20]. As a result, the amount of untranslatable mRNAs is reduced, which should be favorable for efficiently using resources in bacterial cells. Although PTT and fast mRNA degradation have been commonly believed to be mechanisms responsible for translation-regulated transcription, these regulatory mechanisms have a drawback, i.e. they prevent the synthesis of untranslatable mRNAs only after substantial transcription elongation has occurred. The inhibition of mRNA generation at the earlier stage of transcription should be a more efficient strategy for saving resources in bacterial cells.

Here, we report a new pathway of mRNA quality control by TTC, which inhibits the synthesis of untranslatable mRNAs from the promoter-proximal region when transcription is decoupled from translation in Escherichia coli. We investigated the effect of TTC at the early stage of transcription by quantitatively measuring the transcriptional kinetics and the abundance of the promoter-proximal region of mRNAs. We found that the synthesis rate of promoter-proximal mRNA before the generation of the RBS is significantly reduced when translation is absent, which implies that coupling with translation is required for efficient early transcription near the transcription start site (TSS). The length of RNAP elongation before coupling with the ribosome [5′-untranslated region (UTR) length] determined the mRNA level, i.e. the longer the 5′-UTR length was, the less mRNA was expressed. In line with this, without translation, we also observed that RNAPs stalled within 250 bp from the TSS. RNAPs decoupled from translation showed low processivity, elongating mRNA only 80–90 bp on average before transitioning into an inactive state, in sharp contrast with the high processivity observed in in vitro studies [21–24]. Based on these findings, we propose that inhibiting mRNA synthesis at the promoter-proximal region is the first checkpoint of mRNA quality control occurring ahead of PTT and fast mRNA decay by TTC in E. coli. Our results also provide a potential explanation for why genes with 5′-UTR lengths longer than 100 bp are rare in E. coli.

Materials and methods

Bacterial strains and mutant construction

We constructed strains in which the RBS and start codon (ATG) of each target gene (xylA,galP,galE,rhaB,and lacZ) are changed to block their translational initiation by homologous recombination. The pCas-CDF plasmid, single-guide RNA (sgRNA)-expressing plasmids, and respective DNA editing templates were used (Supplementary Table S1). The residual translation initiation rate was predicted by the RBS calculator [25] to confirm the loss of translational initiation (Supplementary Table S2). Homologous recombination and plasmid curing were performed by the previously described method [26, 27] with modifications. Editing DNA templates were prepared for each gene by primer extension polymerase chain reaction (PCR) to harbor homologous sequences using the designed dRBS and TAA codon as overlapping regions (Supplementary Table S3).

To construct strains with different 5′-UTR lengths for the xylA gene with a common binding region for the promoter-proximal probe, homologous recombination was used as described above except for the preparation of editing templates. First, the template DNA for UTR54 was designed to insert a single probe binding sequence at the promoter-proximal region (20 nt) between the 10 bp promoter-adjacent sequence and the 25 bp start codon-adjacent sequence to obtain a minimal 5′-UTR for minimal interference with translation initiation. This sequence was TA-cloned and used as a template to construct other variants, UTR100, UTR200, and UTR300; the number after the UTR denotes the 5′-UTR length. Random sequences to extend the 5′-UTR length were generated with a random sequence generator (46 bp to be inserted after promoter-proximal 29 bp and start codon adjacent 25 bp, GC%: 50, https://faculty.ucr.edu/∼mmaduro/random.htm), and designed to exclude any predicted RBS or start codon, as verified using the RBS calculator [25] and UTR designer [28]. The DNA carrying the extended 5′-UTR sequence was amplified using inverse PCR with 5′ phosphorylated oligo1/oligo2 and circularized by blunt-end ligation. The 5′-UTR regions were extended sequentially to make UTR200 and UTR300 strains by inverse PCR and blunt-end ligation (Supplementary Table S2). The lengthened 5′-UTR sequences were amplified with oligo1/oligo4-harboring homologous sequences and used as editing templates in homologous recombination. We also constructed a NusG-myc strain to perform NusG-targeting ChIP-exo experiments. A synthetic 10AAlinker-8myc sequence with homologous regions was introduced upstream of the stop codon of the nusG gene using the pCas-CDF plasmid. An editing DNA template was synthesized and directly used for recombination (Supplementary Table S3). To construct strains for constitutive lacZ expression, we replaced the lacI gene with a kanamycin resistance gene using lambda red recombination (Supplementary Table S3).

Reagents for cell cultivation were purchased from BD Bioscience, except for casamino acid, which was purchased from MyBiosource. Antibiotics were purchased from Sigma–Aldrich, except for bicyclomycin (BCM), which was purchased from Santa Cruz Biotechnology. Other chemicals were purchased from Sigma−Aldrich. Plasmid and amplified DNA fragments were purified using the GeneAll^® Exgene^TM Cell SV Kit (GeneAll Biotechnology) and GeneAll^® Expin^TM PCR SV Kit (GeneAll Biotechnology), respectively. Q5 DNA polymerase and Quick ligation^TM kit were purchased from NEB and used for inverse PCR and blunt-end ligation, respectively. TaKaRa Taq^TM and EmeraldAmp^® GT PCR Master Mix were purchased from Takara Bio Inc. for routine recombination and cloning procedures. Oligonucleotides listed in Supplementary Table S3 were synthesized by Bioneer.

Two-color fluorescence in situ hybridization

We followed a previously described protocol [29]. E. colicells were grown in 3 ml of LB medium at 37°C with shaking. Overnight cultures were diluted 1:200 into M9 medium supplemented with 0.4% glycerol and 1% casamino acids unless otherwise stated. For measuring mRNA quantity in the steady state, the cells were cultured for ∼4 h until they reached an OD₆₀₀ of ∼0.3 with the proper inducer to induce mRNA expression (0.2% L-rhamnose, 0.2% D-galactose, 0.4% D-xylose, or 1 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) for different target genes). Then, the cells were pelleted by centrifugation (4500 × g, 5 min) and fixed using 3.7% formaldehyde (Sigma–Aldrich) for 30 min. To measure the kinetics of mRNA expression, the cells were cultured without inducer and resuspended in prewarmed medium at 37°C. After adding the appropriate inducer, the cell culture was withdrawn at 15 s intervals and directly fixed with 3.7% formaldehyde. After fixation, the cells were pelleted by centrifugation (400 × g, 8 min) and washed twice using 1× phosphate buffered saline (PBS) (Ambion). The cells were resuspended in 70% ethanol (Sigma-Aldrich) for permeabilization. After 1 h of incubation, the cells were collected (600 × g, 7 min) and resuspended in wash buffer [25% formamide (Ambion) and 2 × SSC (Ambion)]. Then, the cells were pelleted again and resuspended in hybridization buffer [25% formamide, 2 × SSC, 10% dextran sulfate (Sigma–Aldrich), 0.1% tRNA (Sigma–Aldrich), 0.2 mg/ml bovine serum albumin (Ambion), and 2 mM ribonucleoside vanadyl complex (NEB)]. The DNA probes (Biosearch Technologies) labeled with fluorescent dyes (Alexa 514 and ATTO 594) shown in Supplementary Table S4 were added to the hybridization buffer before resuspension to a final concentration of 13.8 nM each. The hybridization mixture was incubated overnight at 30°C. Before imaging, 10 μl of the mixture was washed twice using wash buffer to remove free DNA probes and resuspended in 2 × SSC. Then, the cells were placed on a gel pad made of 3% low-melting-temperature agarose (Lonza) in 1× PBS.

Single-molecule mRNA fluorescent in situ hybridization (smFISH) samples were observed by an inverted microscope (Olympus, IX-71) with a 100× oil-immersed objective (Olympus) and an EMCCD camera (Andor iXon DU897). Phase contrast images and two fluorescence images for Alexa 514 and ATTO 594 were acquired at multiple fields of view. For imaging of mRNA labeled with Alexa 514, FF03-510/20 (Semrock) (excitation), HQ550/50m (Chroma) (emission), and FF520-Di01-25 × 36 (Semrock) (dichroic mirror) were used. Excitation was provided by a 514 nm Ar-ion laser (Melles Griot 43 Series Ion Laser). For imaging of mRNA labeled with ATTO 594, FF01-572/28–25 (Semrock) (excitation), FF01-641/75–25 (Semrock) (emission), and FF593-Di02-25 × 36 (Semrock) (dichroic mirror) were used. Excitation was provided by a 580 nm fiber laser (VFL-P-Series, MPB Communications Inc.). Metamorph software was used to control the automated measurements and maintain focus during data acquisition.

FISH data analysis

Home-built software (MATLAB) was used for image analysis. Phase contrast images were used to segment the cell area and shape. Then, we used Spatz cells [29] in the fluorescence images to quantify the target mRNA copy numbers in individual cells. To exclude the effect of autofluorescence and nonspecific binding of DNA probes, we used each target gene deletion strain (ΔxylAB, ΔlacOperon, ΔgalP, ΔgalETKM, ΔrhaB) as a negative control for each target mRNA measurement. For quantifying 5′- and 3′-end mRNA, which was detected by 10∼15 DNA probes each, false-positive spots were discarded according to the spot intensity statistics of negative samples. Promoter-proximal mRNA was detected using only one DNA probe, and it was difficult to distinguish false-positive and positive spots based on their intensities. Therefore, we subtracted the average quantity of promoter-proximal mRNA measured in negative samples (which is <0.1) from that of positive samples. The spot intensity of a single mRNA molecule was obtained from the spot intensity histogram measured from low-expression samples. To calculate the mRNA quantity of each cell, the total spot intensity in each cell was divided by the single-mRNA spot intensity. The average quantity of mRNA was calculated for the numbers of cells exceeding ∼300 in each independent experiment, and three independent experiments were performed (n = 3).

ChIP-exo

To determine the changes in RNAP density resulting from removing the RBS and start codon (dRBSdATG), we constructed sequencing libraries for ChIP-exo by the method described before with modifications [30]. Briefly, E. coli cells induced for 4 h with the appropriate inducer were cross-linked by 1% formaldehyde at room temperature for 30 min. The cross-linked cells were harvested and lysed, followed by sonication for DNA fragmentation. To immunoprecipitate the DNA–RNAP complexes, the sonicated cell extract was treated with 6 μl of purified anti-E. coli RNAP β Antibody (Biolegend) overnight at 4°C, followed by another overnight incubation at 4°C with 50 μl of Dynabeads™ Pan Mouse IgG magnetic beads (Invitrogen). ChIP-DNA was used for enzymatic modification and library construction as previously described procedure [27]. The ChIP-DNA was end-repaired, dA-tailed, and first adapter-ligated using the KAPA HyperPrep Kit (KAPA Biosystems) according to the manufacturer’s instructions. At this stage, nick repair, λ exonuclease and RecJ_f exonuclease (NEB) treatment were conducted in sequence according to the manufacturer’s instructions. The exonuclease-treated DNA samples were eluted from the beads by reversing protein−DNA cross-linking overnight at 65°C. After reversal of the cross-links, the second strand was synthesized, followed by dA-tailing and second adapter ligation. The DNA samples were enriched by PCR using KAPA HotStart ReadMix in HyperPrep Kit (KAPA Biosystems). Samples were quantified using the Qubit ds DNA HS Assay Kit (Invitrogen), and of which quality was determined using the HT DNA NGS 3K Reagent Kit (PerkinElmer). Samples were sequenced using NextSeq (Illumina) according to the manufacturer’s instructions. For the experiments involving NusG, synthetic 8myc tags were added at the C-terminus of the protein. DNA-NusG complexes were immunoprecipitated using 15 μl c-Myc antibody (Santa Cruz Biotech), and all other experimental steps were performed as described earlier. ChIP-exo measurements were conducted in biological duplicate (n = 2).

ChIP-exo data processing

Sequence reads were mapped onto the reference genome (GenBank accession number NC_000913) using Bowtie (version 1.2.2) [31] with the following options to generate SAM output files: -X 5000 -n 2 -p 3–3 9 -S. For Bowtie alignment, we prepared multiple variants of the reference genome in advance by updating some of the sequences on the genome, such as dRBSdATG and modified UTR, to properly map the sequence reads generated from different genotype samples. In the xylA samples, the aligned reads were manually normalized by dividing 10⁹ × (reads on the xylAB operon and its upstream region) into total reads aligned to the genome. The reads from the other gene samples, galE, galP, lacZ, and rhaB, were also normalized in accordance with their operon sequences. Genome-scale data during the data processing stage were visualized using either MetaScope (https://github.com/SBML-Kimlab/MetaScope) or NimbleGen’s SignalMap software.

mRNA lifetime measurements using quantitative real-time PCR

Cells were grown in the same medium as in smFISH experiments. When the OD₆₀₀ of the culture reached ∼0.3, the cells were concentrated and resuspended in medium with an appropriate inducer to induce target gene expression for 30 min at 37°C. Fifty microliters of the cells were withdrawn into 150 μl of ice-cooled RNAlater solution (Ambion) before shutting off lacZ expression for the time-zero sample. At time zero, rifampicin (Sigma–Aldrich) was added to a final concentration of 300 μg/ml. Fifty microliters of the cells were withdrawn into 150 μl of ice-cooled RNAlater solution at each sampling time point. After sampling, the cells were incubated for 20 min on ice, and 400 μl of ice-cooled M9 medium was added to the cells and gently mixed by pipetting. The cells were collected by centrifugation (10 000 rpm, 4°C, 2 min), and then the pellets were resuspended in 100 μl of ice-cooled lysozyme solution [100 mg/ml lysozyme (Sigma–Aldrich), 10 mM Tris–HCl, pH 8.0, 0.1 mM EDTA] and vigorously vortexed for 20 s. Then, 0.5 μl of 10% SDS solution was added to the cell. The cells were incubated at RT for 5 min. Then, total RNA was extracted using the PureLink RNA Mini Kit (Ambion) with optional DNase (Invitrogen) digestion.

The extracted RNA product was converted to cDNA using Superscript III Reverse Transcriptase (Invitrogen) and RNaseOut Recombinant Ribonuclease Inhibitor (Invitrogen). Quantitative real-time PCR (qRT−PCR) was performed to determine the target mRNA using SYBR Green premix (Enzynomics) and a StepOne Real-Time PCR system (Applied Biosystems). The sequences of the primers used are described in Supplementary Table S3. The results were plotted to determine the lifetime of each target mRNA (Supplementary Figs S5 and S7).

Correction for differences in mRNA lifetime and downstream transcription reduction

mRNA lifetime correction was performed by multiplying the ratio of mRNA lifetime between wild type (WT) and dRBSdATG (mRNA lifetime of dRBSdATG/mRNA lifetime of WT) for each target gene [32]. Downstream transcription reduction correction was performed by multiplying the expression ratio between 3′-end mRNA and 5′-end mRNA (<3′-end mRNA>/<5′-end mRNA>) in dRBSdATG for each target gene.

Results

Decoupling transcription from translation significantly reduces the mRNA level

To investigate the effect of translation on transcription, we selected five different inducible genes (rhaB, galP, galE, lacZ, and xylA) across the E. coli genome and constructed translation-blocked mutant strains (dRBSdATG) by removing the RBS and replacing the start codon (ATG) of each gene with a stop codon (TAA) (Fig. 1A and Supplementary Fig. S1). Deletion of both the RBS and ATG significantly reduced the LacZ levels by 320-fold [33]. To quantitatively measure the mRNA amounts of WT and dRBSdATG strains, we used two-color smFISH experiments at a steady state where different oligo-probe sets labeled with Alexa Fluor 514 and ATTO 594 were hybridized to the 5′- and 3′-end mRNA regions (∼500 bp) of target genes, respectively (Fig. 1B).

Figure 1.

Decoupling transcription from translation significantly reduces the mRNA level. (A) Target genes in the E. coli K12 MG1655 chromosome, rhaB, galP, galE, lacZ, and xylA. The mRNA copy numbers of these genes were measured by single-molecule mRNA FISH. (B) Schematic illustration of the two-color smFISH measurements and representative images of xylA mRNAs. Two different oligo probe sets labeled with Alexa Fluor 514 or ATTO 594 hybridized to the 5′- and 3′-end mRNA regions (∼500 bp) of target genes, respectively. Overlay of fluorescence and phase contrast images shows a significant reduction in xylA mRNA levels in dRBSdATG (Scale bar, 2 μm). The mean 5′- and 3′-end mRNA copy numbers per cell as measured by smFISH in inducible (C, D) and constitutive (E, F) expression conditions. The average mRNA quantities were determined from measurements of over ∼300 individual cells per condition in each independent experiment, consistent with all smFISH data presented in this study. The bars and error bars show the mean ± SD of three independent experiments. The dots represent the data from three experiments. (G,H) Expression ratio of the mean mRNA copy number between WT and dRBSdATG strains. Error bars were obtained from the SDs of replicates and error propagation. (I,J) The mean 5′- and 3′-end mRNA copy numbers in WT with and without kasugamycin (KSG). 1 or 10 mg/ml KSG was added to cells 15 min before the FISH sample preparation. The bars and error bars show the mean ± SD of three independent experiments. The dots represent the data from three experiments.

In all cases, the average copies of both 5′- and 3′-end mRNAs per cell decreased significantly without translation (Fig. 1C and D). For example, in the case of xylA, the mean 5′-end (3′-end) mRNA copy number per cell was 83.6 ± 17.3 (80.8 ± 11.4) in WT, which was reduced to 2.3 ± 0.4 (1.4 ± 0.5) in the dRBSdATG strain. Since these genes have different regulatory mechanisms that may affect the observed behavior, we constructed mutants that constitutively expressed each mRNA and measured their mRNA copy numbers (Fig. 1E and F). The mean 5′-end (3′-end) mRNA copy number per cell of the constitutively expressed xylA was 125.5 ± 18.2 (101.8 ± 12.0) in WT, while it was only 3.1 ± 0.8 (2.1 ± 0.9) in dRBSdATG. In the same manner, mRNA copy numbers decreased significantly without translation in all other genes we tested. These results indicate that mRNA levels are affected by translation regardless of the types of genes and regulation.

To more clearly compare mRNA levels with and without translation, we defined the expression ratio for each region of mRNA as the mean mRNA copy number in the presence of translation (WT) divided by that in the absence of translation (dRBSdATG). The expression ratio was in the range of 3.3–59.0 depending on the gene. The xylA gene showed the highest expression ratio, 35.9 ± 9.6 and 59.0 ± 22.3 for 5′- and 3′-end mRNA, respectively (Fig. 1G and H). The higher expression ratio of 3′-end mRNA than that of 5′-end mRNA probably resulted from downstream transcription reduction occurring without translation. The expression ratios of mRNA from inducible and constitutive expression were almost identical, demonstrating that the effect of translation on the mRNA level is independent of gene regulation at the molecular level.

As a control experiment to inhibit translation in another way, we treated the translation inhibitor KSG [34, 35]. A drastic reduction in WT mRNA levels to a similar level as that of dRBSdATG was observed for lacZ and xylA genes, demonstrating that global translation inhibition also leads to a reduction of mRNA levels (Fig. 1I and J).

Determining the effects of PTT and mRNA degradation on the mRNA level

The reduction in mRNA levels under transcription–translation decoupling has previously been explained by PTT and mRNA degradation [8, 12–14, 19, 20] (Fig. 2A). We determined how these mechanisms contribute to the reduced mRNA levels observed in dRBSdATG strains. First, we examined the possibility of Rho-dependent transcription termination [16–18]. We measured mRNA levels using qRT−PCR under treatment with Bicyclomycin (BCM), an antibiotic that inhibits the activity of Rho. Consistent with previous work [36], mRNA levels in WT were unaffected by BCM treatment (Fig. 2B and Supplementary Fig. S2A). Similarly, no restoration of mRNA levels was observed in any of the dRBSdATG mutants (Fig. 2C and Supplementary Fig. S2B). This lack of recovery was further confirmed by smFISH analysis (Supplementary Fig. S3). Together, these results indicate that the significant decrease in mRNA levels in dRBSdATG is unlikely to result from Rho-dependent PTT.

Figure 2.

Determining the effects of PTT and mRNA degradation on the mRNA level. (A) Schematic illustration of the process of mRNA synthesis. “Active translation” represents proper coupling between transcription and translation. “No translation” denotes the decoupling between transcription and translation, which has been known to induce PTT and fast mRNA decay. The relative 5′-end mRNA quantity in WT (B) and dRBSdATG (C) with and without BCM divided by that of WT, which was measured using qRT−PCR. 50 μg/ml BCM was added to the cells 15 min before the measurement. The bars and error bars show the mean ± SD of three independent experiments. The dots represent the data from three experiments. (D) mRNA lifetime measurements. The 3′-end mRNA lifetime of each gene was measured by rifampicin assay using qRT−PCR. Error bars were obtained from SDs of replicates in Supplementary Fig. S5. (E) The expression ratio of 3′-end mRNA after considering the measured effect of the difference in the mRNA lifetime and downstream transcription reduction. Error bars were obtained from the SDs of replicates and error propagation.

It has previously been reported that BCM treatment restores the reduced mRNA synthesis rate caused by transcription–translation decoupling within the coding region, such as by introducing a stop codon at the 500th nucleotide from the TSS, highlighting the role of Rho-dependent transcription termination [8]. To test whether BCM has a similar effect and to compare the result with dRBSdATG, we constructed a similar mutant strain by introducing a stop codon at the 261st nucleotide from the TSS of the xylA gene (xylA_TAA261). We then examined the induction kinetics of 5′- and 3′-end mRNAs in xylA_TAA261 with and without BCM (Supplementary Fig. S4A and B). Consistent with the previous study [8], BCM treatment recovered the reduced 3′-end mRNA synthesis rate by transcription–translation decoupling (Supplementary Fig. S4C), confirming that translation termination in the middle of the coding region induces Rho-dependent PTT. In contrast, in dRBSdATG, the significantly reduced synthesis rate of both 5′- and 3′-end mRNA was not restored by BCM treatment (Supplementary Fig. S4D and E). The difference between xylA_TAA261 and dRBSdATG lies in the region where the translation is decoupled from transcription: as for xylA_TAA261, translation is decoupled from transcription in the middle of the coding region, whereas in the case of dRBSdATG, translation is blocked at the promoter-proximal region. Thus, all our results indicate that Rho-mediated PTT is not a major contributor to the significant decrease in mRNA levels in the absence of promoter-proximal coupling of transcription and translation.

Next, we asked whether the reduced mRNA levels in dRBSdATG could be explained by faster mRNA decay resulting from the loss of ribosome-mediated protection against RNases [19, 20]. The mRNA lifetime of the 3′-end mRNA was measured using the assay with rifampicin [32] (Supplementary Fig. S5). The extent of the effect of translation on mRNA stability was different depending on the gene (Fig. 2D). The lifetime of xylA 3′-end mRNA was decreased by 2.3-fold without translation, while other genes showed a smaller decrease in the lifetime of mRNA. Nevertheless, faster mRNA decay alone is insufficient to account for the dramatic 9- to 59-fold decrease in 3′-end mRNA levels observed in dRBSdATG compared to the WT.

Then, we integrated the effects of both mRNA degradation and downstream transcription reduction to examine whether these effects together could explain the decrease in 3′-end mRNA levels observed in dRBSdATG. The ratio of 3′-end to 5′-end mRNA levels was calculated to represent the decrease in downstream transcription during RNAPs transcribe from the 5′-end to the 3′-end region (Supplementary Fig. S6). The ratio was close to 1 in WT genes (0.8–1.1), which was reduced to 0.3–0.6 in dRBSdATG genes, indicating that downstream transcription decreased when translation was blocked in dRBSdATG strains. We corrected the expression ratio of the 3′-end mRNA level between WT and dRBSdATG by accounting for both mRNA lifetimes and the ratio of 3′-end to 5′-end mRNA levels we obtained. The corrected expression ratio ranged from 4.6 to 15.0 across all genes (Fig. 2E), reinforcing that the significant decrease in mRNA levels in dRBSdATG cannot be fully explained by previously known effects of TTC.

Translation is required for efficient early transcription

Although PTT and faster mRNA degradation were previously reported to contribute to the decrease in mRNA level for the transcription–translation decoupled system [8, 12–14, 19, 20], these effects were not sufficient to explain the significant decrease in mRNA level that we observed (Fig. 2). This strongly implies that there is an additional pathway of transcription regulation by translation, beyond the classic effects of TTC. The roles of both PTT and mRNA degradation have mainly been investigated at the transcription elongation state after the generation of the RBS. This approach was highly rational, considering that ribosome mostly becomes involved in transcription after binding to the RBS. However, the effect of TTC on mRNA synthesis before the generation of the RBS, if it exists, is entirely unknown. Indeed, our data support this possibility, as the levels of 5′-end mRNA, representing the initial ∼500 bp of coding sequences close to the TSS (Fig. 1B), also significantly decreased, similar to the 3′-end mRNA levels, suggesting that translation may influence mRNA synthesis near the TSS.

Furthermore, our experiments involving BCM treatment did not reveal any significant effect on Rho-induced PTT in dRBSdATG strains (Fig. 2C and Supplementary Figs S2–S4). The PTT mechanism by Rho requires a minimum mRNA length of ∼90 nt for Rho binding to occur [37–40]. This finding led us to hypothesize that decoupling with translation may impact transcription at the early stage, even before sufficient mRNA length for Rho binding is generated. If translation regulates mRNA generation as early as possible, such as during early transcription elongation shortly after RNAP escapes the promoter, it could be the most efficient way to prevent the generation of untranslatable mRNAs and, as a consequence, to save vital resources in cells.

To investigate the possibility of the effect of translation at the early stage of transcription, we quantified the promoter-proximal region of mRNA located upstream of the RBS (Fig. 3A and B). xylA and lacZ were chosen because they have a sufficient length of the promoter-proximal region before the RBS for detection using a single 20 nucleotide (20 nt) DNA probe for smFISH (Fig. 3A). The 20 nt probe detects the mRNA region upstream of the RBS before the active engagement of ribosome in transcription (Fig. 3B). As a result, the amount of promoter-proximal mRNA reflects the efficiency of early transcription in the 5′-UTR and is expected to be independent of coupling with translation. Surprisingly, however, the level of promoter-proximal mRNA significantly decreased in dRBSdATG strains compared to that of WT (Fig. 3C). The expression ratios between WT and dRBSdATG in lacZ and xylA were 11.6 ± 7.3 and 12.4 ± 5.3, respectively, at the steady state (Fig. 3D). These results strongly suggest that translation is crucial for the expression of the promoter-proximal mRNA before the generation of RBS, i.e. considered the early transcription stage.

Figure 3.

Translation is required for efficient early transcription. (A) Schematic illustration of promoter-proximal mRNA detection by smFISH and 5′-UTR sequences of xylA and lacZ. A single oligo probe labeled with ATTO594 was hybridized to the promoter-proximal mRNA region (red color) upstream of the RBS. The TSSs (+1) and the start codons (ATG) of the genes are denoted. (B) The promoter-proximal region of mRNA is generated before translation begins. (C) Representative overlay of fluorescence (mRNA level) and phase contrast images of WT and dRBSdATG strains (Scale bar, 2 μm). (D) Expression ratio of the promoter-proximal mRNA between WT and dRBSdATG. Error bars were obtained from the SDs of replicates and error propagation. (E,F) Kinetics of promoter-proximal mRNA synthesis after transcription induction. xylA and lacZ mRNA quantities were measured after the addition of the inducer, 0.4% xylose and 1 mM IPTG, respectively, at time 0 sec. The data represent the means ± SDs of three independent experiments. The dots represent the data from three experiments. RNAP occupancy of WT and dRBSdATG near xylA (G) and lacZ (H) measured by ChIP-exo. The genes shown in bold are the target genes analyzed by smFISH. (+) and (−) in ChIP-exo data indicate the forward and reverse reads, respectively. TSS denotes a transcription start site. (I) RNAP occupancies near xylA were measured in dRBSdATG strains under two conditions: rifampicin treatment (150 μg/ml for 15 min) or BCM treatment (50 μg/ml for 30 min). For the purpose of the comparison, ChIP-exo data without antibiotics (Fig. 3G) are presented. After rifampicin treatment, the RNAP footprint in dRBSdATG shows a narrower peak exclusively close to the TSS, in contrast to the broad RNAP signals within 250 bp from the promoter (yellow box) observed without antibiotics. BCM treatment did not change the promoter-proximal RNAP occupancy.

The reduction in promoter-proximal mRNAs in dRBSdATG strains cannot be explained by the known mechanisms of PTT occurring without translation, as these mRNAs are located in the 5′-UTR (Fig. 3A and B). Then, to rule out the possibility of fast decay of the promoter-proximal mRNA in dRBSdATG, we measured the lifetime and the induction kinetics of the promoter-proximal mRNA. The lifetimes of promoter-proximal mRNAs (∼100 nt from the TSS) were 2.1 ± 0.6 min and 1.6 ± 0.6 min for lacZ and xylA mRNAs, respectively (Supplementary Fig. S7). Then, the induction kinetics of the promoter-proximal mRNA were observed within a time window (<1 min) where mRNA degradation was not significant (Supplementary Note S1). The level of the promoter-proximal mRNA increased after induction in WT, while no change was detected in dRBSdATG (Fig. 3E and F). Therefore, mRNA decay is not responsible for reducing the promoter-proximal mRNAs in dRBSdATG strains. Thus, the absence of translation may directly lead to reduced synthesis of promoter-proximal mRNA.

Our results indicate that the promoter-proximal mRNAs are not generated in dRBSdATG strains. Then, how could ribosomes regulate the amount of promoter-proximal mRNA even before the RBS is generated? To investigate possible clues for the mechanism, we measured RNAP density on the genes of WT and dRBSdATG strains using ChIP-exo [30]. By using an antibody targeting the RNAP beta subunit, we discriminated the footprint of RNAP in the presence and absence of translation (Fig. 3G and H, and Supplementary Fig. S8). In the WT strain, RNAPs were observed throughout the gene region (red lines), while in dRBSdATG, the RNAP density was observed primarily near the TSS region, with the signal significantly decreased beyond 250 bp from the TSS for all genes (gray lines). Since the promoter-proximal mRNA levels were found to be very low in dRBSdATG, the biased RNAP density observed in dRBSdATG strains may result from RNAPs that are stalled on DNA, incapable of mRNA synthesis.

To determine whether the promoter-proximal density observed in dRBSdATG represents RNAPs prior to promoter escape or not, we examined the RNAP footprint under treatment with rifampicin, an antibiotic that blocks elongation after RNAP transcribes 2–3 nt of transcript [41] (Supplementary Fig. S9 and Fig. 3I). Under rifampicin treatment, the RNAP occupancy within 250 bp from the TSS disappeared in dRBSdATG, indicating that the promoter-proximal RNAP density includes RNAPs that have escaped the promoter but become inactivated after escape (Fig. 3I). Next, to confirm that the promoter-proximal RNAP density observed in dRBSdATG is not due to the Rho-dependent transcription termination, we compared the RNAP densities with and without BCM treatment (Supplementary Fig. S9 and Fig. 3I). The RNAP occupancy in dRBSdATG was observed near the TSS even under Rho inhibition (Fig. 3I). If Rho-mediated RNAP dissociation were responsible for the RNAP density observed near the promoter, inhibiting Rho should have led to RNAP density extending further downstream. However, the inhibition of Rho by BCM did not increase the RNAP density, which supports again that the promoter-proximal RNAP density is not due to the Rho-mediated dissociation of RNAPs as they proceed downstream. This result is also consistent with a previous report demonstrating that BCM does not abolish promoter-proximal peaks of RNAP and Rho occupancy [42]. Together, these results suggest that, in the absence of translation, RNAPs that escape from the promoter become stalled and arrested in the promoter-proximal region, resulting in the stacking of inactivated RNAPs.

The length of RNAP elongation before coupling with the ribosome (5′-UTR length) determines the mRNA level

Since ribosomes can initiate translation only after the loaded RNAP reaches the end of the 5′-UTR of each gene, it is unlikely that translation directly affects the first promoter loading or the stability of the RNAP in the 5′-UTR. We hypothesize that the promoter-proximal occupancy of RNAPs may be related to inefficient early transcription (Fig. 4A). If an inactivated RNAP is generated near the TSS, it can inhibit the clearance of the promoter-proximal region, ultimately obstructing early transcription by subsequent RNAPs. In contrast, with proper function of translation, elongation complexes are efficiently released to move downstream, preventing the accumulation of inactivated RNAPs in the promoter-proximal region and facilitating productive transcription.

Figure 4.

The length of RNAP elongation before coupling with the ribosome (5′-UTR length) determines the mRNA level. (A) Schematic illustration showing the proposed model for early transcription at the promoter-proximal region depending on translation. (B) Mutant strains of xylA with various 5′-UTR lengths. Promoter-proximal mRNA of each strain was measured using the identical single probe labeled with ATTO594 (red boxes). (C) Promoter-proximal mRNA quantity per cell in each 5′-UTR mutant strain. The bars and error bars show the mean ± SD of three independent experiments. The dots represent the data from three experiments. (D) RNAP occupancy depending on the 5′-UTR lengths. (+) and (−) in ChIP-exo data indicate the forward and reverse reads, respectively. TSS denotes a transcription start site. The red star next to the promoter indicates the region detected by the promoter-proximal probe shown in Fig. 4B. (E) Expression ratio of the promoter-proximal mRNA between wtRBS and dRBSdATG for different 5′-UTR length mutant strains. Error bars were obtained from the SDs of replicates and error propagation.

If this is the case, translation initiation should occur before RNAP undergoes a transition from the elongation state into inactive RNAP in the promoter-proximal region. To test this hypothesis, we varied the length of the 5′-UTR, i.e. the distance RNAP must travel before coupling with the ribosome (Fig. 4B). If RNAP undergoes a transition into an inactive state before coupling with the ribosome, the probability of RNAP reaching the RBS should decrease as the length of the 5′-UTR increases. We analyzed the level of promoter-proximal mRNA and RNAP density on the gene using xylA mutants with different 5′-UTR lengths (54, 100, 200, and 300 bp) (Fig. 4B-D). These mutants have identical sequences for 29 bp from the TSS, and thus, the identical FISH probe was used to detect the promoter-proximal mRNA (Fig. 4B). The promoter-proximal mRNA was barely synthesized in dRBSdATG strains regardless of 5′-UTR length (Fig. 4C). Conversely, xylA mutant strains with the RBS and ATG (wtRBS) showed a decreasing trend in the amount of promoter-proximal mRNA as the length of the 5′-UTR was increased, even though this mRNA region is positioned identically relative to the TSS in all 5′-UTR mutants (Fig. 4B). These results indicate that the enhancement effect of translation on early transcription compared with dRBSdATG strains diminishes as the 5′-UTR length increases.

Next, we investigated the RNAP footprint depending on 5′-UTR length and verified that RNAP densities obtained from ChIP-exo data were well correlated with the mean level of promoter-proximal mRNA (Fig. 4D). For all dRBSdATG strains, RNAP density was detected only in the promoter-proximal region, regardless of 5′-UTR length. In the wtRBS strains, the RNAP signal in the xylA coding sequence gradually decreased as the 5′-UTR length increased and almost disappeared when the 5′-UTR length reached 300 bp (Fig. 4D). Note that the RNAP footprints of wtRBS and dRBSdATG are nearly identical to the 300 bp 5′ UTR, showing only promoter-proximal occupancy. The observed promoter-proximal RNAP density in both UTR300 wtRBS and dRBSdATG strains did not extend further upon BCM treatment (Supplementary Fig. S10), indicating that it is not a result of Rho-mediated RNAP dissociation, even in strains with long 5′-UTRs.

To compare the effect of translation on early transcription in different 5′-UTR lengths, we calculated the expression ratio of mean promoter-proximal mRNA of wtRBS and dRBSdATG (Fig. 4E). The expression ratio is a sequence-independent value because the contribution of sequence difference between the strains on the mRNA level is canceled out by dividing the two values. The expression ratio showed a clear dependence on the 5′-UTR length; it decreased as the 5′-UTR length increased (Fig. 4E). When the 5′-UTR length was 300 bp, the expression ratio of the promoter-proximal mRNA was only ∼1.7, indicating that translation almost loses its effect on early transcription when translation starts 300 bp downstream from the promoter. These results show that the effect of translation on early transcription is strongly dependent on 5′-UTR length, i.e. the distance RNAP elongates mRNA before the formation of the RBS. Thus, coupling transcription with translation near the promoter-proximal region is vital to prevent RNAP stacking near the promoter.

Model for the inhibition of early transcription by transcription–translation decoupling

To quantitatively describe the observed 5′-UTR length dependence on early transcription, we derived an analytical expression for the expression ratio as a function of 5′-UTR length using a simple kinetic model (Supplementary Note S2; Fig. 5A and B). We assumed that RNAPs in the 5′-UTR exist in two states, an active or an inactivated state (Fig. 5A). In other words, RNAP has a probability of having a transition from an active elongation state into an inactive state until it is coupled with the ribosome. If RNAP travels beyond the 5′-UTR length (i.e. reaching the RBS), translation initiation occurs, and the transcription activity of RNAP is persistent. If the transition from the active elongation state into the inactive state of RNAP occurs randomly, the expression ratio should have an exponential dependence on the 5′-UTR length [43]. The function we derived is a reasonably good fit to the data in Fig. 4E (R²= 0.84, Fig. 5C), which supports our proposition that RNAPs are randomly inactivated without translation in the 5′-UTR. Fitting the data provided the average elongation length of RNAP before it is inactivated (< l >), which was 81.8 bp. As an additional test, we fit the data together with those obtained by measuring the 5′-end mRNAs of xylA mutants having different 5′-UTR lengths and sequences; the result gave a similar value of < l >, 88.1 bp (Fig. 5D). The estimated value of < l> explains why the average promoter-proximal mRNA level per cell barely changed after transcription induction in dRBSdATG (Figs 3E and F); since a single RNAP occupies approximately 40 bp of DNA [44, 45], the average number of RNAPs that can pass the promoter-proximal region would be only 1 or 2.

Figure 5.

Model for the inhibition of early transcription by transcription–translation decoupling. (A) Illustration of the three states of RNAP in the model: active RNAP, inactivated RNAP, and translation-coupled RNAP. (B) Schematic showing the inhibition of early transcription by inactivated RNAP within the 5′-UTR. The inactivated RNAP in the 5′-UTR blocks or inactivates the following RNAPs, thereby preventing transcription in the 5′-UTR by subsequent RNAPs. The inactivation of RNAP occurs randomly before coupling with ribosomes. (C) Expression ratio of the promoter-proximal mRNA shows exponential dependence on the 5′-UTR length. Experimental data (red squares, the reproduction of Fig. 4E) and the fitting curve (black line) to a single exponential function (see Supplementary Note S2) are shown. <l> denotes the average elongation length of RNAP before it is inactivated. (D) Expression ratio of xylA mRNA obtained from eight different xylA mutant strains that have various 5′-UTR lengths and sequences. The data were fitted to the same function as in Supplementary Note S2. (E) A proposed model for mRNA quality control by translation. Previously known mechanisms of untranslatable mRNA quality control are (2) PTT and (3) fast mRNA degradation. We observed a reduction in early transcription near the TSS in this work and proposed a new mechanism of (1) blocking early transcription without translation. The random inactivation of RNAP prevents the clearance of the promoter-proximal region, obstructing efficient early transcription. As a result, translation regulates transcription from early to late elongation and mRNA degradation.

Our results imply that RNAP has limited processivity before coupling with ribosome in cells, and thus, as the 5′-UTR length increases, transcription is less effective. If an inactivated RNAP is generated within the 5′-UTR before reaching the RBS, it can block the trailing active RNAPs. The resulting accumulation of inactivated RNAPs near the promoter-proximal region may hinder productive early transcription (Fig. 5B). Consistent with our observation, genes with a 5′-UTR longer than 100 bp are rare in E. coli [46].

Discussion

TTC in bacteria has been extensively studied, revealing how ribosomes affect transcriptional elongation or prevent PTT and mRNA degradation. It has been a general belief that TTC occurs after the generation of the RBS, which may have left the effect of TTC on the earlier stage of transcription, such as early elongation near the TSS, largely unexplored. In this work, we provided the first insight into translation-regulated early transcription in bacteria. Previously known effects of translation on transcription, such as PTT and mRNA degradation [8, 12–14, 19, 20], cannot explain the significant reduction in full-length mRNA synthesis brought by transcription–translation decoupling. Additionally, the promoter-proximal mRNA was barely synthesized without translation, which was unexpected based on the current understanding of the role of TTC. Our results suggest that without translation, inactivated RNAPs become stacked at the promoter-proximal region and inhibit the further transcription of untranslatable mRNA from its very first stage. As a result, translation is the underlying mechanism that regulates transcription from early to late elongation and mRNA degradation (Fig. 5E). Because inhibiting the synthesis of mRNA at the promoter-proximal region occurs ahead of PTT and mRNA decay, the mechanism we proposed here should be the first checkpoint of mRNA quality control by TTC and the most efficient strategy for conserving the required components and recycling the molecular machinery.

Our results suggest that RNAP inactivation randomly occurs during early elongation in vivo, which can be prevented by coupling with the ribosome. It is well known that RNAP shows abortive initiation after promoter loading, and its transition from the initiation to elongation state is often difficult [47]. RNAP undergoes frequent backtracking and pauses until the transition occurs [48–50]. However, RNAP in the elongation state usually maintains high processivity without translation in vitro. Although transcribing RNAPs stochastically paused/backtracked at every nucleotide in vitro [21–23, 51–53], the average pause lifetime was <∼10 s [21–23]. In addition, the stalled transcription elongation complexes observed in in vitro experiments maintain their transcriptional activity and easily restart transcription [54]. In contrast, according to our in vivo results, RNAPs are inactivated during early elongation and remain inactive for a reasonably long time to affect the upstream event. The measured transcription rate [55] of xylA WT was 5.3 copies/min (Supplementary Fig. S11), which means that the inactivated RNAP should remain at least for ∼10 s to block RNAPs loaded after. The discrepancy in RNAP processivity between in vitro studies and our in vivo results suggests that inhibitory mechanisms or RNAP-binding factors present in the cellular environment may limit RNAP processivity.

Translation initiation may prevent RNAP inactivation, as the leading ribosome can physically push RNAP, reactivating it from an arrested state [7, 9, 10]. A recent in vitro study showed that the translating ribosome accelerates the transcription speed of the leading RNAP [51]. Moreover, another study demonstrated that the ribosome enhances RNAP activity by exerting mechanical force and allosterically inducing RNAP to adopt an active conformation [11]. Another possibility is that translation plays a key role in removing inactivated RNAPs, thereby preventing RNAP traffic in the promoter-proximal region. A very recent in vitro study demonstrated that coupled ribosomes can dislodge stalled RNAPs, a function that trailing RNAPs cannot perform [56]. This suggests that ribosomes may help prevent transcription traffic jams caused by stalled RNAPs. Based on this, it is possible that in the promoter-proximal region where Rho-mediated PTT is not feasible, ribosome-mediated removal of stalled RNAPs could serve as the primary mechanism for clearing inactivated RNAPs. In the absence of translation, the lack of a pathway to remove stalled RNAPs may lead to promoter-proximal RNAP traffic.

Transcription factors, such as NusG and Rho, which are closely associated with the formation of transcription–translation complex, may influence RNAP inactivation in the absence of promoter-proximal TTC. NusG, known to stabilize the transcription elongation complex and suppress RNAP backtracking and pausing [57], could help prevent promoter-proximal RNAP inactivation. NusG binds both RNAP and ribosome simultaneously, and the NusG bridging mainly determines the structure and function of the RNAP-ribosome complex [5, 6]. In eukaryotes, where the regulatory role of promoter-proximal RNAP pausing is well described, DSIF, the homolog of NusG, is one of the factors that determines the activity of RNAP II at the promoter-proximal region [58]. In a similar manner, NusG binding to RNAP through TTC may be a factor for maintaining RNAP in a productive state. Rho, in contrast, may contribute to promoter-proximal RNAP inactivation. Rho shows promoter-proximal occupancy similar to RNAP [42], and recent studies suggest it initially associates with RNAP independently of RNA synthesis. It then allosterically inactivates the elongation complex before relatively slow dissociation occurs [59–62]. RNAP dissociation requires the ring closure of Rho, which is triggered by its interaction with RNA of at least ∼90 nt in length [37–39]. Therefore, RNAPs in the promoter-proximal region, where transcripts are too short, may be inactivated by Rho but fail to dissociate. In this context, NusG also becomes relevant, as Rho and ribosome compete for binding to NusG, which stabilizes their complex with RNAP [4]. We examined the NusG footprint in both xylA WT and dRBSdATG (Supplementary Fig. S12). In WT, NusG was detected throughout the gene region, consistent with RNAP density (Supplementary Fig. S12A). In dRBSdATG, the NusG density within 250 bp of the TSS closely matches the RNAP density (Supplementary Fig. S12B). As previously reported [42], NusG exhibits a strong footprint downstream of the promoter-proximal RNAP peak. These results indicate that the promoter-proximal RNAP density in dRBSdATG corresponds to RNAPs that have transitioned into the elongation phase, where NusG can bind [42]. The presence of NusG binding to promoter-proximal RNAPs in dRBSdATG raises the possibility that NusG may stabilize the RNAP-Rho complex instead of transcription–translation complex. Further investigation will be required to clarify whether and how NusG and Rho regulate RNAP activity in the absence of promoter-proximal TTC.

We summarized the expression ratio depending on the 5′-UTR length in Supplementary Fig. S13A, which showed that the expression ratio has a maximal value at 40–50 bp and then gradually decreases. The expression ratio distribution is similar to the 5′-UTR length distribution found in E. coli [46, 63] (Supplementary Fig. S13B). Rare long 5′-UTRs, over 100 bp, form part of special regulatory mechanisms, such as riboswitches and sRNA binding sites, and are expressed by specific elongation complexes. In contrast to E. coli RNAP, Bacillus subtilis RNAP does not require TTC to maintain processivity and exhibits a faster elongation rate [64, 65]. These differences may explain the longer 5′-UTRs [66] and greater prevalence of riboswitches in B. subtilis [67]. Thus, the bacterial 5′-UTR seems to have evolved to the optimal distance from the promoter to the translation initiation region, depending on the necessity of TTC for maintaining RNAP processivity.

Then, what could be the origin of the differing necessity for TTC in maintaining RNAP processivity between species like E. coli (which require TTC) and B. subtilis (which do not)? Compare to E. coli RNAP, B. subtilis RNAP forms less stable complexes at many promoters [68, 69] and escapes more efficiently, likely due to weaker RNAP-DNA contacts [70]. Additionally, the two enzymes respond differently to pause and termination signals, indicating distinct RNAP-DNA interactions [70]. Weaker RNAP-DNA interactions may make RNAP in species lacking TTC, such as B. subtilis, less prone to inactivation, enabling faster elongation without the need for TTC. Alternatively, if, as we suggested, Rho-induced RNAP inactivation is a factor limiting RNAP processivity in the absence of promoter-proximal TTC, the functional divergence of Rho across species may contribute to the difference. In B. subtilis, NusG and Rho are dispensable and Rho-dependent termination is rare [71, 72]. Additionally, many species predicted to have uncoupled RNAP and ribosomes lack rho [64, 73]. In species lacking functional Rho would not require TTC to prevent Rho-induced RNAP inactivation.

Here, we report a new pathway of transcription regulation in bacteria: RNAP inactivation at the promoter-proximal region regulates transcription. This type of strategy seems to be widely conserved in biological organisms since it is the most efficient way to prevent the unnecessary usage of resources. In eukaryotes, promoter-proximal Pol II pausing within 100 bp from the TSS plays a fundamental role in gene regulation [58]. Recent studies have shown that controlling the escape of the transcription elongation complex at the promoter-proximal region regulates transcription initiation in both archaea and eukaryotes, including humans [74–76]. In bacteria, it has recently been reported that sigma factors can induce RNAP pausing within 10–20 bp from the TSS, potentially causing RNAP retention at the promoter and inhibiting transcription [77]. Our study focuses on a distinct mechanism in which ribosomes prevent RNAP inactivation further downstream from the TSS (80–90 bp), revealing a regulatory pathway mediated by TTC, a hallmark of bacterial gene expression. Taken together, the regulation of promoter-proximal RNAP traffic likely plays a crucial role in bacterial transcription regulation, similar to eukaryotic systems.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by National Research Foundation [RS-2023-NR077154, RS-2020-NR049542], and the Bio & Medical Technology Development Program [RS-2024-00352569, NRF-2021M3A9I4024737] and a grant [RS-2024-00345885] of the National Research Foundation funded by the Korean government (MSIT). Funding to pay the Open Access publication charges for this article was provided by National Research Foundation of Korea.

Data availability

All raw and processed data for ChIP-exo performed in this work have been deposited in Gene Expression Omnibus (GEO) under the accession number GSE228686.