MazF activation promotes translational heterogeneity of the grcA mRNA in Escherichia coli populations

Bacterial strains and plasmids

We used derivatives of E. coli strain K-12 MG1655 (Blattner et al., 1997; Müller et al., 2016) when addressing effects of antibiotic treatments and amino acid starvation, and derivatives of E. coli strain BW27784 containing a native copy of the mazEF operon (Khlebnikov et al., 2001) for experiments with arabinose-inducible systems. The open reading frame of the employed reporter gene EmgfpΔACA is devoid of ACA sites (Sauert, 2015; Oron-Gottesman et al., 2016), whilst the amino acid sequence corresponds to the wild-type fast-folding Emerald GFP (Tsien, 1998). The chromosomally integrated mCherry reporter (Cox III, Dunlop & Elowitz, 2010) under control of the phage λ promoter P_R is used to infer constitutive gene expression. All strains and plasmids are listed in Table S1.

All primers used in this study are listed in File S1, section ‘Sequences’. Briefly, the translational grcA_wt reporter was constructed by fusing the fragment of the grcA gene comprising nucleotides −278 to +45 in frame to the gfp coding sequence via a linker encoding three glycine residues. The grcA_ATA reporter was constructed by fusion-PCR, modifying C to T in the sequence of the grcA_wt reporter at positions −1 and −32 upstream of the start codon. Both grcA reporters were cloned into a modified CRIM plasmid pAH68-frt-chlor (kindly provided by T. Bergmiller; Haldimann & Wanner, 2001) via EcoRI/PstI restriction sites. The grcA_wt reporter was further cloned into a low copy plasmid pZS*12-GFP (kindly provided by C. Guet) via XhoI/HindIII restriction sites. The transcriptional reporter P_maz-gfp was cloned into a modified CRIM plasmid pAH120-frt-chlor (kindly provided by T. Bergmiller; Haldimann & Wanner, 2001), and the promoter sequence, i.e., intergenic region between relA and mazE, was flanked with XhoI/BamHI restriction sites. All constructs were inserted into neutral phage attachment sites in the E. coli genome using the CRIM system: the plasmids pAH68-frt-chlor and pAH120-frt-chlor were integrated into attHK022 and attλ, respectively (Haldimann & Wanner, 2001). After integration, the chloramphenicol resistance marker was removed by using the site-specific Flp-recombinase (Cherepanov & Wackernagel, 1995).

Growth conditions

Rich, defined media was composed of 1× M9 salts, 1 mM MgSO₄, 0.1 mM CaCl₂, 0.5% casamino acids (Fluka), and 10 mM maltose. Clones were first streaked from frozen glycerol stocks on LB (Lennox modification) agar to obtain single colonies. A single colony was inoculated overnight for 15–16 h, in 4 ml of media, at 37 °C with shaking at 165 rpm. Bacterial growth was monitored by measuring the optical density at 600 nm (OD₆₀₀) with a spectrophotometer. The overnight cultures were diluted 1 to 1,000 into fresh media, OD₆₀₀ ∼  0.007, and grown for 2 h 15 min. The exponential cultures were then split in two or more flasks; with one flask serving as a control, while different stressors (or arabinose to induce mazF expression) were added to other flasks. After 3–4.5 h of stress, 4 ml of stressed cultures were washed with 1×  PBS, resuspended in 4 ml of pre-warmed fresh media and allowed to regrowth. In general, the cultures were analyzed by flow cytometry at three time points: (t1) 2–3 h after stress induction; (t2) 5.5–7 h after stress induction, or recovered for 1.5–2.5 h after initial 3–4.5 h under stress; (t3) 21 h after stress induction, or recovered for 15.5–18 h after initial 3–4.5 h under stress.

Flow cytometry

Data analysis and visualization

Raw fcs files were exported from FlowJo (Tree Star, Inc.) into a custom R script, similar to the script published as Dataset S2 in Silander et al. (2012), with R commands for gating indicated in File S1, section ‘R commands’. Flow cytometry plots were made in FlowJo. All events were analyzed for the experiments, unless indicated otherwise. For the experiment presented in Fig. 1, in total 10,000 cells were analyzed with R script after filtering (gating) events on the basis of FSC and SSC measurements (size parameters), and visualized by using auto-gating tool on the size parameters in FlowJo.

Cytometry setup and analysis for Fig. 2 was as follows. The cultures were analyzed after approximately 3 h (time point t1), 7 h (time point t2) and 23 h (time point t3). We used concentrations of stressors above their minimum inhibitory concentration (MIC), referred to as ‘harsh’ conditions (Table 1), and below MIC referred to as ‘mild’ conditions (Table 2) (Tosa & Pizer, 1971; Sulavik et al., 2001; Chaudhuri et al., 2010): ampicillin (Amp) 100 µg/ml and 1 µg/ml, chloramphenicol (Cam) 15 µg/ml and 1 µg/ml, nalidixic acid (NA) 50 µg/ml and 1 µg/ml, spectinomycin (Spec) 90 µg/ml and 1 µg/ml, and serine hydroxamate (SHX, Sigma) 100 µg/ml and 1 µg/ml, as ‘harsh’ and ‘mild’ conditions, respectively.

Table 1:

Expression of the P_maz-gfp reporter in the presence of stressors at concentrations above the MIC levels (‘harsh’ stress conditions).

Percentage of the cells that express the transcriptional reporter P_maz-gfp above background level, measured before, during and after experiencing stress.

Time point			Percentage of the cells with GFP signala (gated populations)	SCV of log mCherry fluorescence (a.u.)a (entire populations)	∼OD
t0	Exponential		∼0	0.0140 ± 0.001	0.23
t1	5 h of growth	Control	∼0	0.0156 ± 0.002	2.10
	3 h under stress	cAmp, 100 µg/ml	∼0	0.0327 ± 0.002	0.07
		Cam, 15 µg/ml	∼0	0.0176 ± 0.002	0.35
		NA, 50 µg/ml	∼0	0.0273 ± 0.005	0.68
		SHX, 100 µg/ml	0.78 ± 0.56	0.0157 ± 0.001	0.56
		Spec, 90 µg/ml	1.05 ± 0.35	0.0142 ± 0.002	0.42
t2	9 h of growth	Control	0.39 ± 0.07	0.0210 ± 0.002	7.16
	7 h under stress	Amp, 100 µg/ml	∼0	0.0265 ± 0.001	0.07
		Cam, 15 µg/ml	0.18 ± 0.11	0.0227 ± 0.004	0.56
		NA, 50 µg/ml	∼0	0.0315 ± 0.008	0.77
		SHX, 100 µg/ml	b1.01 ± 0.19	0.0211 ± 0.004	3.00
		Spec, 90 µg/ml	2.35 ± 0.46	0.0110 ± 0.002	0.62
	4.5 h under stress,	Regrowth, Amp-treated	0.02 ± 0.01	0.0300 ± 0.001	0.23
	then 2.5 h of regrowth	Regrowth, Cam-treated	5.41 ± 4.32	0.0148 ± 0.001	1.40
		Regrowth, NA-treated	7.01 ± 2.94	0.0360 ± 0.000	0.77
		Regrowth, SHX-treated	0.61 ± 0.05	0.0165 ± 0.001	4.21
		Regrowth, Spec-treated	1.31 ± 0.94	0.0171 ± 0.002	1.34
t3	25 h of growth	Control	1.13 ± 0.02	0.0075 ± 0.001	7.27
	4.5 h under stress,	Regrowth, Amp-treated	1.99 ± 0.38	0.0123 ± 0.001	3.85
	then 15.5 h of regrowth	Regrowth, Cam-treated	4.56 ± 0.38	0.0113 ± 0.001	7.81
		Regrowth, NA-treated	2.58 ± 0.75	0.0167 ± 0.001	6.71
		Regrowth, SHX-treated	4.04 ± 0.32	0.0147 ± 0.004	6.68
		Regrowth, Spec-treated	5.65 ± 0.92	0.0091 ± 0.001	8.62

DOI: 10.7717/peerj.3830/table-1

Notes:

a(mean ±  standard error of the mean).

bValues in bold font are the examples depicted in Fig. 2.

cAmp, ampicillin; Cam, chloramphenicol; NA, nalidixic acid; SHX, serine hydroxamate; Spec, spectinomycin.

Table 2:

Expression of the P_maz-gfp reporter in the presence of stressors at concentrations below MIC levels (‘mild’ stress conditions).

Percentage of the cells that express the transcriptional reporter P_maz-gfp above background level, measured before, during and after experiencing mild stress conditions.

Time point			Percentage of the cells with GFP signala (gated populations)	SCV of log mCherry fluorescence (a.u.)a (entire populations)	∼OD
t0	Exponential		b∼0	0.0167 ± 0.002	0.22
t1	5 h of growth	Control	∼0	0.0157 ± 0.000	1.34
	3 h under stress	cAmp, 1 µg/ml	3.13 ± 2.21	0.0159 ± 0.000	1.25
		Cam, 1 µg/ml	0.95 ± 0.34	0.0163 ± 0.003	0.64
		NA, 1 µg/ml	2.96 ± 1.26	0.0113 ± 0.002	1.15
		SHX, 5 µg/ml	2.36 ± 1.58	0.0173 ± 0.000	1.17
		Spec, 1 µg/ml	2.01 ± 1.14	0.0125 ± 0.002	1.08
t2	9 h of growth	Control	1.02 ± 0.46	0.0211 ± 0.003	5.93
	7 h under stress	Amp, 1 µg/ml	1.98 ± 0.51	0.0162 ± 0.000	5.67
		Cam, 1 µg/ml	0.29 ± 0.13	0.0188 ± 0.000	3.00
		NA, 1 µg/ml	1.22 ± 0.18	0.0185 ± 0.001	4.72
		SHX, 5 µg/ml	1.14 ± 0.29	0.0173 ± 0.000	5.13
		Spec, 1 µg/ml	1.65 ± 0.57	0.0173 ± 0.001	4.17
	4 h under stress,	Regrowth, Amp-treated	1.92 ± 0.08	0.0117 ± 0.001	5.88
	then 2.5 h of regrowth	Regrowth, Cam-treated	0.16 ± 0.06	0.0182 ± 0.000	3.48
		Regrowth, NA-treated	0.90 ± 0.04	0.0155 ± 0.000	5.34
		Regrowth, SHX-treated	1.84 ± 0.37	0.0111 ± 0.001	7.33
		Regrowth, Spec-treated	0.96 ± 0.18	0.0160 ± 0.000	5.37
t3	25 h of growth	Control	4.94 ± 1.77	0.0088 ± 0.001	7.83
	4 h under stress,	Regrowth, Amp-treated	3.90 ± 0.12	0.0072 ± 0.000	9.45
	then 18 h of regrowth	Regrowth, Cam-treated	2.18 ± 0.12	0.0118 ± 0.002	7.99
		Regrowth, NA-treated	11.69 ± 0.80	0.0094 ± 0.001	9.24
		Regrowth, SHX-treated	2.95 ± 0.24	0.0163 ± 0.006	9.99
		Regrowth, Spec-treated	12.04 ± 1.96	0.0081 ± 0.001	9.51

DOI: 10.7717/peerj.3830/table-2

Notes:

a(mean ±  standard error of the mean).

bValues in bold font are the examples depicted in Fig. 2.

cAmp, ampicillin; Cam, chloramphenicol; NA, nalidixic acid; SHX, serine hydroxamate; Spec, spectinomycin.

In total, 50,000 events were acquired by the flow cytometer Fortessa for strain NN230. Minimally 20,000 events were processed per each sample, by filtering on the basis of FSC and SSC measurements (size parameters). The gating captured cells in the densest part of the FSC-SSC plot and therefore the gates consisted of cells with similar size and physiology due to the narrow range of FSC and SSC, respectively. Such analysis allows one to minimize the variation in fluorescence due to different cell size or cell damage, differences in stage of cell cycle, etc. Of those gated events, we only considered events that contained a mCherry signal above background level. The mCherry background level is defined as the 99th percentile of the measurements for the strain that does not contain the mCherry fluorescent reporter gene, which corresponds to autofluorescence of the MG1655 strain. This was one way how we could determine that the cells were alive or at least their membranes were not damaged during stress. For instance, cells damaged or lysed due to Amp-treatment could still be detected by flow cytometry within defined FSC and SSC parameters, but a constitutive mCherry signal would not be detected, and therefore we do not include those dead cells into further analysis.

The GFP background level is defined as the 99th percentile of measurements for the strain that does not contain the gfp fluorescent reporter gene (which corresponds to autofluorescence of the TB205 strain; Bergmiller et al., 2017), averaged over all conditions. These subsets were used for the subsequent analysis to calculate the number of cells that expressed gfp above background level. Finally, the number of GFP-positive cells was corrected with the number of cells with GFP fluorescence above background level measured in reporterless strains in the same condition, to eliminate false-positive GFP signals.

Microscopy and image analysis

An exponentially growing culture of strain NN207 was divided into two flasks, and 0.2% Ara was added to one flask to induce mazF expression. Samples were taken from untreated cultures and cultures treated for 3 h and 22 h with Ara. The cells were fixed with 1% formaldehyde in 1× PBS, incubated for 2 h in the dark at room temperature using a nutator, followed by two washing steps with 1× PBS. The samples were kept in dark at 4 °C until microscopy the following day. 1.2 µl of a fixed bacterial sample was applied on a 1.5% agarose pad using a cavity slide as previously described (Bergmiller et al., 2011). Images were acquired with Olympus IX-81 inverse widefield epifluorescence microscope using objective UPlanSApo 100×/1.4 Oil, and Hamamatsu ORCA-ER detection system. The lamp intensity was set on 100%; exposure time of 130 ms for phase contrast images; and exposure time of 200 ms (filter TexasRed U-MWG, Ex. 510–550 nm, Em. LP 590 nm) for mCherry images. Cells were analyzed with the Matlab-based package Schnitzcells (Young et al., 2011), and segmented using mCherry fluorescence images.

Statistical analysis

Statistics was done in SPSS, R and Microsoft Excel. Error bars in all graphs depict standard deviation from N replicates (independent cultures); N is indicated in every Figure Legend. We log₁₀-transformed fluorescence data prior to analysis. The squared coefficient of variation (SCV) was used as a measure of variation in the reporter fluorescence. SCV is the squared value of coefficient of variation (CV), and CV is defined as standard deviation divided by the mean. We computed Pearson’s second skewness coefficient (median skewness) as skewness = 3*(mean − median)/standard deviation; and kurtosis type 2 in R. We used 2-tailed, heteroscedastic Student’s t-test to infer differences between two datasets for each measurement: between the fluorescence values of the grcA_wt and grcA_ATA reporters, and between the fluorescence values of the strain harboring P_maz-gfp reporter (NN230) and the strain without gfp reporter gene (TB205). We used One-way ANOVA with Post Hoc Test Bonferroni to infer differences in four datasets concomitantly (fluorescence measurements of strains NN200, NN211, NN208, NN221). We used non-parametric Spearman’s two-tailed test to assess correlations between mCherry and GFP signals.

Monitoring repression and de-repression of the mazEF transcription using the transcriptional reporter P_maz-gfp

Expression of the mazEF module is negatively autoregulated, and the mazEF transcription from promoters P_maz1 and P_maz2 is strongly repressed under non-stressful conditions by the MazE protein or the MazE-MazF complex (Marianovsky et al., 2001; Keseler et al., 2013). However, repression by the MazE antitoxin alone is reduced by approximately 40% when compared to the MazE-MazF complex, as determined in the E. coli strain MC4100 (Marianovsky et al., 2001). In contrast, stressful conditions promote degradation of the MazE antitoxin (Aizenman, Engelberg-Kulka & Glaser, 1996; Maisonneuve et al., 2011), which subsequently results in de-repression of the mazEF transcription. Thus, the state of mazEF transcription depends on the ratio between toxin T and antitoxin A within a cell: weak repression (when T ≪ A), repression (when T ≈ A), or de-repression (when T ≫ A). This complex mechanism of transcriptional autoregulation of TA systems is generally termed ‘conditional cooperativity’ (Cataudella et al., 2013; Gelens et al., 2013).

Here, we used the transcriptional reporter P_maz-gfp to indirectly monitor the regulation of the mazEF transcription in the E. coli strain MG1655 via GFP fluorescence (Fig.  1A). As a proof-of-principle for using this fluorescent reporter to infer the conditional cooperativity mechanism, we first quantified the activity of the P_maz promoters in the presence and absence of the MazF toxin. To this end, we calculated the difference of the mean levels of the GFP signal between the strain MG1655 ΔmazF and the isogenic wild-type strain, divided by the signal obtained with the wild-type strain (see File S1 for Source Data). Our data show that in the early exponential phase the mean GFP fluorescence level is increased by 60% in the ΔmazF strain relative to the wild-type strain (Fig. 1B), what can be attributed to the weak repression of transcription in the absence of MazF, i.e., when [MazF] ≪ [MazE].

During and after exposure to antibiotics or amino acid starvation mazEF is transcribed at low levels

Next, we aimed to quantify mazF expression levels during adverse conditions that were reported to activate MazF using the transcriptional reporter P_maz-gfp. Concomitantly, we determined the degree of cell-to-cell heterogeneity under these conditions using the constitutive mCherry reporter (Fig. 2). We measured the GFP fluorescence by flow cytometry upon treatment with ‘mild’ (below MIC values) and ‘harsh’ (above MIC values) concentrations of ampicillin (Amp), chloramphenicol (Cam), nalidixic acid (NA), or spectinomycin (Spec), as well as upon mimicking amino acid starvation by adding serine hydroxamate (SHX) (Table 1 for ‘harsh’ and Table 2 for ‘mild’ stress). To calculate the percentage of cells that display GFP signals above background level, we first gated the populations based on the size parameters, and second, based on a constitutively expressed mCherry. Such gating allowed us to analyze live cells with similar size, as described in details in ‘Material and Methods’.

Our analysis using the transcriptional reporter P_maz-gfp revealed that GFP fluorescence was not detectable in the exponential phase (Fig. 2A). However, during SHX-treatment (Fig. 2B) and antibiotic treatments (Tables 1 and 2) the obtained GFP signals exceeded background levels. GFP fluorescence was generally lower during ‘harsh’ than during ‘mild’ stress because the high concentrations of the stressors either killed a fraction of the stressed populations or severely inhibited growth (OD₆₀₀ values are given in Tables 1 and 2). Severe growth retardation implies reduction in overall translational capacity of cells, including decreased GFP synthesis. However, increased GFP signals were measured in all recovered cultures after removal of the stressor (Fig. 2C) when translation of the reporter gfp mRNA was no longer inhibited by the stressor. The maximal GFP signal was observed during recovery after treatment with 1 µg/ml Spec (Fig. 2D), i.e., about 12% of the cells expressed gfp above background level. We also included an additional statistical test indicating significantly higher mean levels of GFP fluorescence for the strain harboring P_maz-gfp reporter compared to the strain without gfp reporter gene (t-test, p < 0.05), during stress and in the recovery periods (all p-values in File S1). These results suggest that during antibiotic treatment and amino acid starvation, and following stress removal, de-repression of the mazEF transcription occurs at low levels. Taken together, transcription and therefore de novo production of MazF is maintained at low levels during stress.

Besides determining the GFP fluorescence, we concomitantly monitored the expression of the reporter gene λP_R-mCherry to infer constitutive expression (Fig. 2E), as the λ promoter P_R is specifically regulated only in the presence of the phage λ proteins (Berthoumieux et al., 2013). Given that expression of bacterial genes is regulated both through specific factors, such as transcription factors and post-transcriptional regulators, and global factors, such as the number of available RNA polymerases and ribosomes (Berthoumieux et al., 2013; Gerosa et al., 2013), the constitutive mCherry expression, which is solely affected by global factors, provides an estimate for the influence of the growth rate on gene expression in single cells (Klumpp, Zhang & Hwa, 2009; Scott et al., 2010; Klumpp & Hwa, 2014) (N Nikolic et al., 2017, unpublished data). Moreover, mCherry is a highly stable protein, thus the amount of mCherry fluorescence reflects mCherry production and dilution through cell division. As growth is impeded during adverse conditions (and mCherry is not diluted through cell division), and some stressful conditions additionally impair translation, these global factors would collectively affect the mCherry fluorescence signal.

Measurements of the entire bacterial populations, without gating, showed a high degree of variation in the constitutive mCherry expression during stressful conditions (Figs.  2F and 2G), which can signify heterogeneity in cell size, viability, and growth rate. The variation in mCherry fluorescence was highest during and after treatments with high concentrations of nalidixic acid (Fig. 2G), or ampicillin (Table 1), suggesting that cell lysis and severe growth retardation increase population heterogeneity. In contrast, the variation in mCherry fluorescence was lowest in populations experiencing longer recovery periods, i.e., in cultures initially stressed and analyzed after overnight recovery (Fig. 2H, Tables 1 and 2).

Interestingly, we found no correlation between the GFP and mCherry signals in the analyzed populations after mild Spec-treatment (gated only based on size parameters, Fig.  2I), and the same trend is also found in other populations initially stressed and analyzed after overnight recovery (see File S1 for Source Data). If we assume that constitutive mCherry fluorescence is an indicator of growth (Leveau & Lindow, 2001; Klumpp, Zhang & Hwa, 2009; Scott et al., 2010; Klumpp & Hwa, 2014) (N Nikolic et al., 2017, unpublished data), and GFP fluorescence is an indicator of the mazEF transcription, this lack of correlation could imply that de-repression and activation of the mazEF transcription do not depend on the cellular growth rate. This result is in line with previous population-based studies that have shown that the expression of an unregulated gene is growth-rate dependent, whereas the expression of a negatively autoregulated gene is a growth-rate independent process (Klumpp, Zhang & Hwa, 2009; Scott et al., 2010).

MazF-mediated mRNA processing monitored by gfp reporter gene fusions

The increasing evidence that TA systems might introduce phenotypic heterogeneity in genetically identical populations (Gerdes & Maisonneuve, 2012; Klumpp & Hwa, 2014) prompted us to study the impact of the MazF-mediated processing on the translation of the respective mRNAs (Vesper et al., 2011; Sauert et al., 2016). To this end, we selected the grcA transcript, which has been shown to be cleaved by MazF at two ACA sites closely upstream of the start codon (Vesper et al., 2011; Sauert et al., 2016). The toxin MazF can thus modify the grcA mRNA by keeping a short leader sequence or making the mRNA leaderless, which allows its selective translation by stress-ribosomes (Vesper et al., 2011). Hence, the grcA mRNA represents a bona fide model to investigate the alteration of gene expression by MazF-mediated mRNA processing.

To follow the grcA expression pattern at the single-cell level, we constructed a fluorescent reporter gene comprising the 5′-UTR and the first 15 codons of grcA fused in frame to the gfp coding sequence, under control of the native P_grcA promoters (Fig. 3A). This wild-type reporter grcA_wt was used as a proxy for the impact of MazF-mediated mRNA processing on gene expression. As a control, we generated the grcA_ATA fluorescent reporter by substituting the ACA sequence motifs by ATA, to prevent MazF-mediated processing. In both reporters, we used an Emerald gfp version that does not contain ACA sites in the open reading frame (EmgfpΔACA) (Sauert, 2015; Oron-Gottesman et al., 2016), while the amino acid sequence remains the same as in the wild-type Emerald GFP (Tsien, 1998).

First, we confirmed by flow cytometry that both, the grcA_wt and the grcA_ATA reporters display similar fluorescence signals during exponential growth without stressors, which would indicate that there are no significant differences in translation of the reporter mRNAs. Indeed, we observed no significant difference in the mean GFP fluorescence level between the grcA_wt and the grcA_ATA reporter (Fig. 3B, Fig. S1). Likewise, there was no difference in the variation in GFP fluorescence, measured as squared coefficient of variation (SCV, Fig. 3C). Alongside, we measured constitutive mCherry expression from the reporter λP_R-mCherry. Correspondingly, we observed no significant differences in the mean level and variation in the mCherry fluorescence in strains harboring the different grcA reporters (Figs. 3B and 3C).

Heterogeneity in the grcA_wt reporter fluorescence during recovery from mazF expression

Next, we aimed to investigate differences in fluorescence between the grcA_wt and grcA_ATA reporters in conditions that promote MazF activation. Because our fluorescent system is based on the GFP translation and subsequent fluorescence, and most of the known inducers of MazF activation are translational inhibitors (Sat et al., 2001; Christensen et al., 2003; Hazan, Sat & Engelberg-Kulka, 2004), we opted for an experimental setup in which mazF expression is tightly controlled by a synthetic promoter. Thus, we employed a chromosomally integrated system to express mazF under transcriptional control of the arabinose-inducible promoter P_BAD in order to investigate direct effects of MazF activation on translation of reporter mRNAs (sequence P_BAD-mazF taken from the plasmid-based system pBAD-mazF (Amitai, Yassin & Engelberg-Kulka, 2004) (N Nikolic et al., 2017, unpublished data)). In this setup, mazF expression was ectopically induced at low levels (Fig. S2), as our results suggested that MazF is transcribed at low levels during antibiotic treatments or amino acid starvation (Tables 1 and 2).

The analysis was performed using derivatives of the E. coli strain BW27784 (Khlebnikov et al., 2001). Strain BW27784 is suitable for arabinose-inducible expression systems such as P_BAD-mazF because it constitutively takes up L-arabinose but is devoid of L-arabinose metabolism (Fig. S2). mazF expression was ectopically induced by addition of 0.1% arabinose (Ara) to exponentially growing cultures. It is important to note that bacterial growth did not cease under these conditions. This low level of ectopic mazF expression led to a 20% reduction in the bacterial growth rate and a decrease in the final optical density (OD₆₀₀) by 36% (Fig. 4A). In contrast to previous observations that ectopic toxin overexpression causes cell elongation (Kasari et al., 2010), the mild ectopic mazF expression used here did not result in the formation of filamentous cells (Fig. 4B; all values in File S1).

After arabinose induction of mazF expression and during the recovery period after the removal of the inducer, we analyzed the increase in the mean level of GFP fluorescence encoded by the grcA_wt and the grcA_ATA reporters. We calculated normalized mean GFP fluorescence as the mean GFP level of each sample divided by the mean GFP level of the respective culture measured in the exponential phase. Despite the normalized mean levels of GFP fluorescence were similar between the grcA reporters (Fig. 5A), there were significant differences in variation in the GFP fluorescence (Fig. 5B). Interestingly, variation in the grcA_wt reporter fluorescence was higher than the grcA_ATA reporter fluorescence 17 h after recovery from 3 h of ectopic mazF expression. Likewise, we observed increased variation when cultures experienced nutrient deprivation during the overnight growth (stationary phase), without addition of the inducer (‘control’, time point t3 in Fig. 5B). Moreover, variation in the grcA_wt reporter fluorescence after overnight growth was significantly higher in the wild-type strain MG1655 containing native mazEF operon compared to the ΔmazF background (Fig. S3). These results suggest that conditions that promote MazF activation increase heterogeneity in the translation of the reporter mRNA that is processed by MazF, compared to the reporter mRNA that is not affected by MazF cleavage.

To further understand variable gfp expression within clonal populations during ectopic mazF expression, we employed two additional measures of variation, namely skewness and kurtosis, which signify the degree of asymmetry and the “tailedness” of GFP fluorescence distributions. Quantitative analyses of skewness and kurtosis showed statistically significant differences (p < 0.01) between the grcA reporters only for the fluorescence measurements 17 h after recovery from 3 h of ectopic mazF expression (see File S1 for Source Data). Interestingly, the corresponding histograms of GFP fluorescence revealed that the grcA_wt reporter GFP distributions had a right tail (Fig. 5C), i.e., the distributions were skewed toward higher values (for comparison see Fig. 5D). Specifically, GFP distributions of the grcA_wt reporter contained more highly fluorescent cells as compared to the grcA_ATA GFP distributions. This result is indicative of the generation of a small subpopulation of cells with an increased level of reporter gene expression as a consequence of the MazF-mediated processing of the grcA_wt mRNA. It has been previously established that stress response promotes functional ribosome heterogeneity (Byrgazov, Vesper & Moll, 2013; Starosta et al., 2014), and that during MazF-induced stress the total ribosome pool is comprised of canonical ribosomes as well as stress-ribosomes that are able to translate MazF-processed mRNAs (Vesper et al., 2011; Sauert et al., 2016). Thus, one plausible explanation is that during ectopic mazF expression both, the un-processed and processed forms of the grcA_wt reporter mRNA can be translated by canonical and stress-ribosomes, respectively, whereas the grcA_ATA reporter mRNA can solely be translated by canonical ribosomes that comprise one part of the total ribosome pool. Given that MazF processing affects more than 300 mRNA species in E. coli by trimming their 5′-UTRs various lengths (Sauert et al., 2016), it is feasible that processed mRNAs maintain different translational regulatory elements that influence their translation in a specific manner. To generally investigate translational heterogeneity of other MazF-processed mRNAs, a reporter system similar to the grcA reporter could be used. However, the reporter sequence has to be carefully optimized and adapted to each candidate mRNA, since we already observed differences in the reporter fluorescence between the wild-type and the ΔACA reporter prior to MazF stress for another MazF-processed mRNA, namely the rpsU mRNA (Fig. S4).