An Hfq-dependent post-transcriptional mechanism fine tunes RecB expression in Escherichia coli

Abstract

All living organisms have developed strategies to respond to chromosomal damage and preserve genome integrity. One such response is the repair of DNA double-strand breaks (DSBs), one of the most toxic forms of DNA lesions. In Escherichia coli, DSBs are repaired via RecBCD-dependent homologous recombination. RecBCD is essential for accurate chromosome maintenance, but its over-expression can lead to reduced DNA repair ability. This apparent paradox suggests that RecBCD copy numbers may need to be tightly controlled within an optimal range. Using single-molecule fluorescence microscopy, we have established that RecB is present in very low abundance at mRNA and protein levels. RecB transcription shows high fluctuations, yet cell-to-cell protein variability remains remarkably low. Here, we show that the post-transcriptional regulator Hfq binds to recB mRNA and downregulates RecB protein translation in vivo. Furthermore, specific disruption of the Hfq-binding site leads to more efficient translation of recB mRNAs. In addition, we observe a less effective reduction of RecB protein fluctuations in the absence of Hfq. This fine-tuning Hfq-mediated mechanism might have the underlying physiological function of maintaining RecB protein levels within an optimal range.

Keywords: DNA repair; E. coli; Hfq protein; RecBCD enzyme; chromosomes; gene expression; noise suppression; physics of living systems; post-transcriptional regulation.

Figure 1.. recB mRNAs are low abundant, short-lived and constitutively expressed.

(A) Schematic description of the recBCD locus, its location on the E. coli chromosome and the corresponding mRNAs. (B) Examples of fluorescence and bright-field images of recB mRNA FISH experiments in wild-type (WT) and ΔrecB strains. Scale bars represent 2 µm. (C) Total recB mRNA distribution quantified with single-molecule fluorescent in situ hybridization (smFISH) and presented in molecule numbers per cell. The histogram represents the average across three replicated experiments; error bars reflect the maximum difference between the repeats. Total number of cells, included in the distribution, is 15,638. (D) recB mRNA degradation rate measured in a time-course experiment where transcription initiation was inhibited with rifampicin. Mean mRNA counts, calculated from the total mRNA distributions for each time point, were normalized to the mean mRNA number at time t = 0 and represented in the natural logarithmic scale. Vertical error bars represent the standard error of the mean (s.e.m.); horizontal errors are given by experimental uncertainty on time measurements. The shaded area shows the time interval used for fitting. The red line is the fitted linear function,–γ_mt, where γ_m is the recB mRNA degradation rate. The final degradation rate was calculated as the average between two replicated time-course experiments (Supplementary file 1G). (E) recB mRNA molecule numbers per cell from the experiments in (C) shown as a function of cell area. The black circles represent the data binned by cell size and averaged in each group (mean ± s.e.m.). The solid line connects the averages across three neighbouring groups. Based on the mean mRNA numbers, all cells were separated into three sub-populations: newborns, cells in transition, and adults. (F) Experimental data from (C) conditioned on cell size (<3.0 µm², newborns; total cell number is 2180) fitted by a negative binomial distribution, NB(r,p). The Kullback–Leibler divergence between experimental and fitted distributions is D_KL = 0.002.

Figure 1—figure supplement 1.. Single-molecule fluorescent in situ hybridization (smFISH) image analysis performed by Spätzcell.

(a) Peak height intensity profiles for spots detected in the wild-type and ΔrecB samples. The dashed line indicates the intensity threshold (99.9 %) that separates a specific signal from the background. (b) Integrated spot intensity histogram for foci detected in the wild-type sample. The data, binned and averaged in groups, are shown in red circles. The data were fitted by the sum of two Gaussian distributions corresponding to one and two mRNA molecule(s) per focus. The grey lines correspond to single Gaussian distributions, while the black solid curve is the sum of these two Gaussians. The red dashed line indicates the intensity equivalent corresponding to the total integrated intensity of FISH probes (in average) bound to a single mRNA (one mRNA).

Figure 1—figure supplement 2.. Sensitivity of the single-molecule fluorescent in situ hybridization (smFISH) protocol allows for quantification of low-abundant recB mRNAs.

(a) Fluorescence signal, detected in the samples with recB over-expression from an arabinose-inducible plasmid (pIK02), is presented as a function of arabinose concentration (orange-dashed line). The intensity of the fluorescence signal in the wild-type and ΔrecB strains is indicated with grey- and green-dashed lines, respectively. More than 350 cells for each induction condition were taken for the analysis. (b) Representative fluorescence images of the samples where recB expression was induced with 10^-3%, 10^-4%, or 10^-5% of arabinose. All fluorescence images are shown in the same intensity range. Scale bars represent 1 µm.

Figure 1—figure supplement 3.. Independent transcription from recB gene copies.

recB mRNA distributions for newborns (left) and adults (right). The experimental data from Figure 1C was conditioned on cell size (<3.0 µm² for newborns) and (>3.5 µm² for adults). The histograms represent the average across three replicated experiments; error bars reflect the maximum difference between the repeats. Total number of cells: 2180 (newborns) and 6413 (adults). The recB mRNA statistics for newborns is well described by a negative binomial distribution, NB(r,p) (Figure 1F). The sum of two independent variables, each of which is distributed as NB(r,p), is then described by a negative binomial distribution NB(2r,p) Thus, the adults data was compared to the predicted distribution, NB(2r,p), based on (1) the assumption of independent transcription from recB gene copies and (2) the fitting parameters inferred on the newborns. The results were verified with Gillespie’s simulations (SSA). The Kullback–Leibler divergence between the experimental and simulated distributions for newborns and adults is D_KL = 0.003 and D_KL = 0.007, respectively.

Figure 2.. RecB proteins are low abundant, long-lived and show evidence of noise reduction.

(A) Schematic of RecB-HaloTag fusion at the endogenous E. coli chromosomal locus of the recB gene. RecB-HaloTag is conjugated to the Janelia Fluor 549 dye (JF549). (B) Examples of fluorescence and bright-field Halo-labelling images for the strain with the RecBHalo fusion and its parental no HaloTag strain. Both samples were labelled with JF549 dye and the images are shown in the same intensity range. Scale bars represent 2 µm. Zoom-in: An example of a cell with five RecBHalo molecules (several single RecBHalo molecules are shown with light-green arrows). (C) Total RecB protein distribution quantified with Halo-labelling and presented in molecules per cell. The histogram represents the average of three replicated experiments; error bars reflect the maximum difference between the repeats. Total number of analysed cells is 10,964. Estimation of false positives in no HaloTag control resulted in ∼0.3 molecules/cell. (D) RecB removal rate measured in a pulse-chase Halo-labelling experiment where the dye was removed at time t = 0. Mean protein counts, calculated from the total protein distributions for each time point, were normalized to the mean protein number at time t = 0 and represented in the natural logarithmic scale. The shaded area shows the time interval used for fitting. The red line is the fitted linear function, –γ_pt, where γ_p is the RecB removal rate. The final removal rate was calculated as the average between two replicated pulse-chase experiments Supplementary file 1G. (E) Comparison between the experimental RecB molecule distribution from (C) conditioned on cell size (in grey) and the results of Gillespie’s simulations (SSA) for a two-stage model of RecB expression (in red). Parameters used in simulations are listed in Supplementary file 1G. The Kullback–Leibler divergence between the distributions is D_KL = 0.15. (F) Comparison of the coefficients of variation, CV_p² = σ_p²/〈p〉² for experimental (Data), simulated (SSA) data and analytical prediction (Theory). The experimental and simulated data from (E) were used; error bars represent standard deviations across three replicates; the black circles show the mean CV_p² in each experiment. Theoretical prediction was calculated using Equation 1.

Figure 2—figure supplement 1.. Image analysis of Halo-labelling experiments performed with the modified version of Spätzcell.

Peak height intensity profiles for spots detected in the strain with RecB-HaloTag fusion (RecBHalo sample) and its parental strain (no HaloTag control). The dashed line indicates a chosen intensity threshold in order to separate a specific signal from the background.

Figure 3.. Translational efficiency of RecB is increased under DNA damage conditions.

(A) Schematic of the experimental workflow. DSBs were induced with 4 ng/ml of ciprofloxacin for 2 hr, and protein (mRNA) quantification was performed with Halo-labelling (single-molecule fluorescent in situ hybridization, smFISH). Mean protein and mRNA concentrations were calculated from the distributions, and the average protein-to-mRNA ratio (translational efficiency) was estimated. (B) Examples of bright-field images for unperturbed (cipro^–) and perturbed (cipro⁺) conditions. Scale bars represent 2 µm. (C) Box plot with cell area distributions for perturbed (blue) and unperturbed (grey) samples. The medians of cell area in each sample are 1.9 µm² (cipro^–) and 3.6 µm² (cipro⁺). (D) recB mRNA concentration distributions quantified with smFISH in intact (grey) and damaged (blue) samples. The histograms represent the average of three replicated experiments. The medians of the recB mRNA concentrations are shown by dashed lines: 0.28 mol/µm² (cipro^–) and 0.12 mol/µm² (cipro⁺). Total numbers of analysed cells are 11,700 (cipro^–) and 6510 (cipro⁺). Inset: Box plot shows significant difference between the average recB mRNA concentrations (P-value = 0.0023, two-sample t-test). (E) RecB concentration distributions quantified with Halo-labelling in unperturbed and perturbed conditions. The histograms represent the average of three replicated experiments. The medians are shown by dashed lines: 2.12 mol/µm² (cipro^–) and 2.28 mol/µm² (cipro⁺). Total number of analysed cells is 1534 (cipro^–) and 683 (cipro⁺). The difference between the average RecB concentrations was insignificant (P-value = 0.36, two-sample t-test). (F) The average number of proteins produced per one mRNA in intact and damaged conditions. Average translational efficiency for each condition was calculated as the ratio between the mean protein concentration c_p and the mean mRNA concentration c_m. The error bars indicate the standard deviation of the data; statistical significance between the conditions was calculated with two-sample t-test (P-value = 0.0001). In (D, F), significance marks stand for: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***).

Figure 3—figure supplement 1.. Analysis of recB mRNA and RecB protein concentrations and growth rate measurements under DNA damage.

(a) recB mRNA concentration for perturbed (blue) and unperturbed (grey) samples represented as a function of cell area. The data from Figure 3D were binned by cell size and averaged in each group (mean ± s.e.m.). The solid lines connect the averages while the dashed lines and shaded areas show mean ± s.e.m. calculated across all cells in each sample. (b) RecB protein concentration for perturbed (blue) and unperturbed (grey) samples shown as a function of cell area. The data from Figure 3E were binned by cell size and averaged in each group (mean ± s.e.m.). The solid lines connect the averages while the dashed lines and shaded areas show mean ± s.e.m. calculated across all cells in each sample. (c) Growth curves obtained with optical density measurements (OD₆₀₀). Cell cultures were grown with 4 ng/ml (blue) or without (grey) ciprofloxacin. The red arrow indicates the time when ciprofloxacin was added to the cipro⁺ sample. The exponential phase of growth (OD₆₀₀ [0.08–0.4]) was used for growth rate estimation. The growth rates, averaged across three replicated experiments, are shown in the legend. Error bars on the bar chart represent 95% confidence intervals.

Figure 4.. RecB expression is regulated by Hfq protein in vivo.

(A) Genome browser track showing Hfq binding to recC-ptrA-recB-recD mRNAs. The coverage is normalized to reads per million (RPM). The major peaks of interest are highlighted by red dashed boxes. (B) Examples of fluorescence and bright-field images of RecB quantification experiments in Δhfq and wild-type strains. Yellow outlines indicate rough positions of bacterial cells in the fluorescence channel. Scale bars represent 2 μm. Both fluorescence images are shown in the same intensity range while different background modulation was applied in the zoom-in figures (for better spots visualization). (C) RecB concentration distributions quantified with Halo-labelling in wild-type cells and the Δhfq mutant. The histograms represent the average of five replicated experiments. The medians are shown by dashed lines: 2.12 mol/µm² for WT and 2.68 mol/µm² for Δhfq. Total number of analysed cells is 3324 (WT) and 2185 (Δhfq). Inset: Box plot shows significant difference (P-value = 0.0007) between the average RecB concentrations in wild-type and Δhfq cells verified by two-sample t-test. (D) Examples of fluorescence and bright-field images of recB mRNA FISH experiments in Δhfq and wild-type cells. Yellow outlines indicate rough positions of bacterial cells in the fluorescence channel. Scale bars represent 2 µm. (E) recB mRNA concentration distributions quantified with single-molecule fluorescent in situ hybridization (smFISH) in the Δhfq mutant and wild-type cells. The histograms represent the average of three replicated experiments. The medians are shown by dashed lines: 0.28 mol/µm² for WT and 0.21 mol/µm² for Δhfq. The total number of analysed cells is 7270 (WT) and 5486 (Δhfq). The insignificant difference between the average recB mRNA concentrations in both strains was verified by two-sample t-test (P-value = 0.11). (F) RecB translational efficiency in wild-type and Δhfq cells. Average translational efficiency (for each strain) was calculated as the ratio between the mean protein concentration c_p and the mean mRNA concentration c_m across replicated experiments. The error bars indicate the standard deviation of the data; statistical significance between the strains was calculated with two-sample t-test (P-value = 0.014). (G) The difference between theoretically predicted (CV²_Theory) and experimentally measured (CV²_Exp) coefficient of variation (squared). The theoretical values were computed according to 1. Error bars represent s.e.m. calculated from the replicated experiments. Statistical significance between the samples was calculated with a two-sample t-test (P-value = 0.0088). In (C, F, G), significance marks stand for: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***).

Figure 4—figure supplement 1.. The main Hfq-binding site within the ptrA gene.

The Hfq-binding peak located within ptrA-recB-recD mRNA, identified in the Hfq-CRAC experiment (Iosub et al., 2020), presented in raw counts (Hits). The vertical dashed line and black box indicate a translation initiation site (ATG) and Shine–Dalgarno sequence (SD), respectively. The black vertical arrows show the nucleotides where direct Hfq cross-linking was detected in the CRAC data. Hfq-binding motifs, A-R(A/G)-N(any nucleotide) (Tree et al., 2014; Link et al., 2009), are highlighted in the red boxes while the cluster of Hfq-binding motifs is underlined.

Figure 4—figure supplement 2.. Cell size analysis, RT-qPCR quantification, and growth rate measurements in the Δhfq mutant.

(a) Examples of bright-field images for wild-type and Δhfq strains. Scale bars represent 2 μm. (b) Box plots represent cell area distributions for wild-type and Δhfq samples. (c) ptrA, recB, recC, and recD transcripts quantified by RT-qPCR in the Δhfq mutant and normalized to the corresponding expression levels in wild-type cells. The data represent averages and standard deviations across three replicates for each gene. rrfD was used as a reference gene. (d) recB mRNA lifetime measured in the wild-type and Δhfq mutant in a time-course experiment where transcription initiation was inhibited with rifampicin. mRNA abundance was quantified with RT-qPCR at t = 0, 1, 2, and 4 min after rifampicin treatment. recB mRNA lifetime was calculated from an exponential fit and normalized to the lifetime in the wild type. The bar chart represents results of three replicated experiments; the error bar shows standard deviation. Significance was calculated with two-sample t-test (P-value 0.35 (ns)). (e) Growth curves obtained by optical density measurements (OD₆₀₀) in Δhfq (orange) and wild type (grey). The strains were grown in the medium supplemented with glucose and amino acids. The exponential phase of growth (OD₆₀₀ ∼ [0.08–0.4]), shown in the logarithmic scale in the inset, was used for growth rate calculation. The growth rates, averaged across three replicated experiments, are shown in the legend. Error bars on the bar chart represent 95% confidence intervals.

Figure 5.. Specific alteration of RecB translation by Hfq in vivo.

(A) A model of Hfq downregulating RecB translation by blocking the ribosome-binding site of the ptrA-recB mRNA. (B) Partial complementation of RecB expression to the wild-type level demonstrated by expression of a functional Hfq protein from a multicopy plasmid (pQE-Hfq) in Δhfq (shown in red). The histogram represents the average of two replicated experiments. RecB concentration histograms in wild type and Δhfq mutant from Figure 4C are shown for relative comparison. Dashed lines represent the average RecB concentration in each condition. (C) An increase in RecB protein production caused by sequestering of Hfq proteins with highly abundant small RNA ChiX (shown in light blue). The histogram represents the average of two replicated experiments. RecB concentration histograms in wild type and Δhfq mutant from Figure 4C are shown for relative comparison. Dashed lines represent the average RecB concentration in each condition. (D) recB mRNA concentration distribution quantified with single-molecule fluorescent in situ hybridization (smFISH) in the strain with the deletion of the main Hfq-binding site recB-5′UTR* (shown in blue). The histogram represents the average of three replicated experiments. The recB mRNA distribution in the wild-type cells is shown for relative comparison. The medians are indicated by dashed lines. An approximate location of the removed sequence is schematically shown in the insert (red star). (E) RecB protein concentration distribution quantified with RecBHalo-labelling in recB-5’UTR* strain (shown in blue). The histogram represents the average of three replicated experiments. RecB concentration histograms in wild type and Δhfq mutant from Figure 4C are shown for relative comparison. Dashed lines represent the average RecB concentration in each condition. (F) Translational efficiency of RecB in wild type, Δhfq mutant and the strain with the deletion of the main Hfq-binding site, recB-5′UTR*. Average translational efficiency was calculated as the ratio between the mean protein concentration c_p and the mean mRNA concentration c_m across replicated experiments. The error bars indicate the standard deviation of the data; statistical significance between the samples was calculated with a two-sample t-test. The P-value for WT and recB-5′UTR* is 0.0007; while the difference between Δhfq and recB-5′UTR* is non-significant (P-value >0.05). Significance marks stand for: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***).

Figure 5—figure supplement 1.. Controls for Hfq complementation experiment.

(a) RecB protein concentration distributions quantified in the Δhfq mutant carrying pQE80L (blue) or pQE-Hfq (red) plasmid. The upper panel is the same as in Figure 5B and presented here to facilitate visual comparison with the lower panel. The histograms represent the average across two replicated experiments per each condition. The RecB concentration histograms for wild type and Δhfq mutant are plotted as references in grey and orange, respectively. The dashed lines represent the mean RecB concentration in each condition. Significance was evaluated with two-sample t-test (P-values: 0.045(*) for Δhfq and Δhfq+pQE-Hfq; 0.71(ns) for Δhfq and Δhfq+pQE80L). (b) hfq RNA levels quantified by RT-qPCR in Δhfq cells carrying pQE-Hfq plasmid and normalized to the hfq RNA level in wild-type cells. (c) ptrA, recB, recC, and recD transcripts quantified by RT-qPCR in Δhfq carrying pQE80L (left) or pQE-Hfq (right) plasmid and normalized to the corresponding expression levels in Δhfq and Δhfq+pQE80L, respectively. In (b, c), the data represent averages and standard deviations across three replicates. (d) A fivefold serial dilution assay of Δhfq strain carrying pQE80L or pQE-Hfq plasmid. Cells were plated onto LB plates without or with 6 ng/ml of ciprofloxacin.

Figure 5—figure supplement 2.. Controls for ChiX over-expression experiment.

(a) RecB protein concentration distributions quantified in wild-type cells carrying pZA21-ChiX (light blue), pZA21 (dark blue) or pZA21-CyaR (purple) plasmid. The upper panel is the same as in Figure 5C and presented here to facilitate visual comparison with the lower panels. The histograms represent the average across two replicated experiments per each condition. The RecB concentration histograms for wild type and the Δhfq mutant are plotted as references in grey and orange, respectively. The dashed lines represent the mean RecB concentration in each condition. Significance was evaluated with two-sample t-test (P-values: 0.021(*) for WT and WT+pZA21-ChiX; 0.42(ns) for WT and WT+pZA21; 0.37(ns) for WT and WT+pZA21-CyaR). (b) chiX and cyaR RNA levels quantified by RT-qPCR in wild-type cells carrying pZA21-ChiX (left) or pZA21-CyaR (right) over-expression plasmid. The results were normalized to the corresponding expression in the cells carrying the backbone plasmid (pZA21). (c) ptrA, recB, recC, and recD transcripts quantified by RT-qPCR in wild-type cells carrying pZA21-ChiX (left), pZA21 (middle), or pZA21-CyaR (right) plasmid. RNA levels were normalized to the corresponding expression levels in wild type. The data represent averages and standard deviations across three replicates.

Figure 5—figure supplement 3.. RecB protein and mRNA quantification in ΔchiX.

(a) RecB protein concentration distribution quantified in ΔchiX cells. The histogram represents the average across two replicated experiments. The RecB concentration histograms for wild type and the Δhfq mutant are plotted as references in grey and orange, respectively. The dashed lines represent the mean RecB concentration in each condition. Significance was evaluated with a two-sample t-test: P-value for WT and ΔchiX is 0.96 (ns). (b) ptrA, recB, recD, and recC transcripts quantified by RT-qPCR in chiX mutants and normalized to the corresponding expression levels in the wild-type sample. The data represent averages and standard deviations across three replicates.

Appendix 3—figure 1.. RecB quantification upon expression of Hfq with point mutations in proximal or distal binding faces.

(a) RecB protein concentration distributions quantified in hfq mutants carrying a pQE80L-derivative plasmid with Hfq protein mutated either in proximal (K56A) or distal (Y25D) binding faces (Zhang et al., 2013). The histograms represent the average across two replicated experiments for each mutation. The RecB concentration histograms for wild type and hfq mutants are plotted as references in grey and orange, respectively. The dashed lines represent the mean RecB concentration in each condition. Significance was evaluated with two-sample t-test: P-values: 0.30(ns) for Δhfq and Δhfq+pQE-Hfq(K56A); 0.07(ns) for Δhfq and Δhfq +pQE-Hfq(Y25D). (b) hfq and recB RNA levels quantified by RT-qPCR in Δhfq cells carrying a pQE80L-derivative plasmid with a mutated Hfq protein. The results were normalized to the corresponding expression in Δhfq cells carrying the backbone plasmid, pQE-Hfq.

Appendix 3—figure 2.. Toxicity of RecBCD over-expression upon DSB induction.

Viability assays were performed for the wild type, wild type carrying RecBCD over-expression plasmid, pDWS2, and ΔrecB. Cells were plated onto LB plates supplemented with ampicillin and either without or with 8/10/12/16 ng/ml of ciprofloxacin. The average survival factors were calculated for at least three replicated experiments while the error bars indicate standard estimation of the mean.

Author response image 1.