Strong increase in the autofluorescence of cells signals struggle for survival

Prokaryotic and eukaryotic cells exhibit an intrinsic natural fluorescence due to the presence of fluorescent cellular structural components and metabolites. Therefore, cellular autofluorescence (AF) is expected to vary with the metabolic states of cells. We examined how exposure to the different stressors changes the AF of Escherichia coli cells. We observed that bactericidal treatments increased green cellular AF, and that de novo protein synthesis was required for the observed AF increase. Excitation and emission spectra and increased expression of the genes from the flavin biosynthesis pathway, strongly suggested that flavins are major contributors to the increased AF. An increased expression of genes encoding diverse flavoproteins which are involved in energy production and ROS detoxification, indicates a cellular strategy to cope with severe stresses. An observed increase in AF under stress is an evolutionary conserved phenomenon as it occurs not only in cells from different bacterial species, but also in yeast and human cells.

Among different endogenous fluorophores, flavins and NAD are extensively studied because they are responsible for most of the cytoplasmic AF and because of their prominent role in cell metabolism. Flavins, which comprise a category of molecules involving riboflavin and its derivatives flavin adenine dinucleotide (FAD) and flavin mononucleotide (FMN), and NAD are involved in various redox reactions. For example, FAD and NAD, which emit fluorescence when they are oxidized and reduced, respectively 2 play a key role in the conversion of energy from acetyl CoA to ATP. Therefore, AF derived from these molecules is expected to vary as a function of the ATP production in different cellular growth phases, and because of the variable nutrient availability and presence of stressors. FAD and FMN are also associated with proteins, some of which are included in the protection against reactive oxygen species (ROS) 14 .
In this study, we examined how AF of E. coli cells changes as a function of exposure to different stressors. As stressors, we used antibiotics that, besides their medical importance, are powerful tools for unraveling complexity of bacterial physiology 15 , and sodium hypochlorite (bleach) which is a widely used bactericidal agent. We used two classes of antibiotics: ß-lactams that are cell wall synthesis inhibitors, and the protein synthesis inhibitors tetracycline and gentamicin. We found that treatment of E. coli with the ß-lactam antibiotic ampicillin or with sodium hypochlorite, significantly increased cellular AF, while no significant AF increase was observed in protein synthesis inhibitor-treated cells. Our data suggest that flavins are major contributors to bactericidal treatment-induced AF and that AF increase reveals cellular adaptive response to cope with the life-threating stressors. Finally, we demonstrated that the increase in green cellular AF subjected to life-threating treatments is an evolutionary conserved phenomenon as it occurs not only in cells from different bacterial species but also in yeast and human cells.

Results
Ampicillin-induced increase of autofluorescence. We first tested whether treatment with the ß-lactam antibiotic ampicillin increases AF of E. coli cells. We used two E. coli strains which are susceptible (7705035) and resistant (8812112) to ampicillin according to the Clinical & Laboratory Standards Institute guidelines (www.clsi. org) (Table S1). Minimum inhibitory concentrations (MIC) of these strains are 4 mg/L and 64 mg/L respectively. Exponentially growing cultures of both strains were treated for 3 hours with a range of ampicillin concentrations. Cellular AF, as well as light scattering profiles, were analyzed by flow cytometry (Figs 1, S1 and S2). When compared to untreated conditions, all tested ampicillin concentrations induced an increase in green AF (excitation at 488 nm with bandpass filters of 530/30). Maximal AF was observed at ½ and 1× MIC. Of note, susceptible cells treated at 1× MIC revealed important heterogeneity in the autofluorescence distribution profile, which reflects cell-to-cell metabolic state heterogeneity. A similar trend was observed with the resistant strain. At concentrations above the MIC, cellular AF was higher than that of the untreated control, but lower than the AF measured at the MIC. Such an increase in the cellular AF, as a response to cytotoxic antibiotic treatment, was observed for other ampicillin-treated E. coli strains, but also for meropenem-treated Klebsiella pneumoniae and Serratia marcescens (Fig. S2).
The forward scattering (FSC) profiles of ampicillin-treated cells became dispersed, due to the appearance of cells of different shapes and sizes relative to the untreated control, and they changed over time (Figs 1 and S1). After one hour of incubation, cells treated with 1× and 16× MIC started increasing in size. After two hours of incubation, cells treated with 1× MIC continued to increase in size and smaller cells increased in frequency within populations of cells treated with 16× MIC. After three hours of incubation, smaller cells increased in frequency within populations of cells treated with both ampicillin concentrations. This could be explained by the mode of action of ß-lactam antibiotics that interfere with the synthesis of the peptidoglycan layer, which provides structural strength to the cell and balances osmotic pressure 16 . Consequently, treatment with the ß-lactams results in a loss of cell-wall integrity, elongation, bulge formation and ultimately cell lysis 17,18 . This raised the question of whether increased AF results from the increase in cell size upon antibiotic treatment, as was previously suggested 19 .
To address this question, we calculated the correlation coefficient r between the cell AF (FL1) and: (i) the FSC that is a proxy for cell size; (ii) the side scattering (SSC) that reflects cell internal complexity (Table S2). r values between FL1 and FSC increased over time for populations treated with the ampicillin concentrations below the MIC, while they decreased for the ampicillin concentration above the MIC. The same tendency was observed for the correlation between FL1 and the SSC.
However, when we normalized AF values of each individual cell with its FSC values, we observed that AF values increased significantly in E. coli cells treated with ≥MIC of ampicillin. The same phenomenon was observed for K. pneumoniae and S. marcescens treated with meropenem (Fig. S2). This result, combined with the correlation data between light scattering and cell fluorescence values (Table S2), suggests that the increase in AF is not due to the increase in bacterial cell size. This was observed for four different E. coli strains and two strains from other species, illustrating that the observed phenomena do not depend on a particular genetic background.
Because the light scattering parameters only give approximate values of cell size and shape, we also analyzed ampicillin-treated E. coli (MIC 4 mg/L) cells using the ImageStream imaging flow cytometer system (Fig. 2). We collected images of 480,000 cells (20,000 cells for each of the six concentrations and four repetitions). We quantified the AF of each cell. We also calculated the volume of each cell using the following formula: 4  width  length  width  3  ,   2 as previously described 20 . We observed that after 2 hours of incubation, both the AF intensity and the volume of cells treated with 0.25× and 0.5× MIC of ampicillin, increased in comparison to untreated controls. However, because the cell volume increased more than the AF, the AF normalized per cell volume decreased significantly (p value < 0.05 -One-way ANOVA Dunn's comparison test). In cells treated with 1×, 2× and 4× MIC, the normalized AF increased significantly (p value < 0.05 -One-way ANOVA Dunn's comparison test). The normalized AF increase was particularly strong at higher ampicillin concentrations because of the increased frequency of the small cells. Therefore, it can be concluded that an increase in the AF of ampicillin-treated bacterial cells is not an artifact due to increased cell size. Importantly, we also observed a significant increase in the AF of the yeast Saccharomyces cerevisiae and HeLa human cancer cells treated with a cytotoxic agent, sodium hypochlorite 21 (Fig. 3). Taken together, our data indicate that AF increases in stressed prokaryotic and eukaryotic cells.
Protein synthesis is a prerequisite for the increase in autofluorescence. Because treatment with bactericidal antibiotic ampicillin increases cellular AF, we tested the impact of a bacteriostatic antibiotic tetracycline on AF. Tetracycline inhibits bacterial protein synthesis by preventing the binding of aminoacyl-tRNA to ribosome-mRNA complex 22 . Therefore, tetracycline prevents cell growth but does not kill cells. Exponentially  Sodium hypochlorite 0.08% (green), 0.16% (violet), 0.32% (red). S. cerevisiae was treated with 0.08% sodium hypochlorite for 1, 2 and 3 hours, while HeLa cells were treated with sodium hypochlorite for 1 hour. The asterisk represents significant differences relative to the non-treated controls according to the unpaired t-test; p value < 0.05. Each value represents the mean (+/−standard error) of the median values of three independent experiments.
growing cultures of a sensitive E. coli strain were treated with a range of tetracycline concentrations for 3 hours. We did not observe increased AF in tetracycline-treated cells at any of the concentrations tested (Fig. 4).
To confirm that tetracycline did affect treated cells, we used two plasmid-borne reporters which consist of fusions of the promoters of the cspA and lacZ genes to a gene coding for green fluorescent protein (GFP) 23 . As expected, the expression of the cspA gene, which is known to be induced by treatment with antibiotics that inhibit protein synthesis 24 , increased upon tetracycline treatment (Fig. 4e). A peculiar structure of the cspA mRNA allows preferential translation, even in the absence of synthesis of other proteins due to antibiotic treatment or low temperature 25 . The lacZ promoter-driven GFP fluorescence decreased with increasing doses of tetracycline in spite of its induction with IPTG, which indicates that tetracycline blocks protein synthesis (Fig. 4). These experiments confirmed that tetracycline had affected treated cells but that this did not increase AF.
We also evaluated the viability of cells treated with tetracycline using the viability markers, Alexa Fluor ™ 633 Hydrazide (AFH) which targets carbonyl groups, and TO-PRO ™ -3 which binds to DNA in cells with damaged membranes 26,27 . Both viability markers confirmed that tetracycline did not kill the cells (Fig. 4) but that they both stained heat-killed cells.
This raised the question of whether AF did not increase in tetracycline-treated cells because they were alive or because tetracycline inhibited protein synthesis. To address this question, we treated cells with the bactericidal protein synthesis inhibitor gentamicin 28 and with the bactericidal agent sodium hypochlorite (Fig. 5). We observed that gentamicin treatment did not increase cellular AF. However, sodium hypochlorite strongly increased cellular AF. Taken together, our data indicate that protein synthesis is a prerequisite for the increase in AF.
Autofluorescence spectra correspond to flavin fluorescence. Different fluorescent cellular components and metabolites have different excitation and emission wavelengths 2,4 . We aimed to identify the endogenous fluorophores that are responsible for the observed increase in cell AF. Using several excitation and emission wavelength combinations, we investigated the fluorescence per FSC unit of ampicillin-treated E. coli by flow cytometry and fluorimetry. There were two peaks of excitation, one between 320 and 400 nm (maxima at 354 and 372 nm) and another between 408 and 488 nm (maximum at 458 nm, Fig. 6). The peak of emission was 525 nm (Fig. S3). These excitation and emission AF wavelengths correlated well with the flavin spectral properties 4 . The comparison of the excitation and emission AF spectra of cells treated with ampicillin in LB ( Fig. S4) or in M9 medium ( Fig. 6), with the spectra of purified flavins: riboflavin, FMN and FAD, further confirmed the correlation.
To further investigate possible involvement of flavins in the observed ampicillin treatment induced cellular AF, we used reporters to measure expression of ribA, ribB, ribC and ribE, which are essential genes required for FMN and FAD biosynthesis 29 (Fig. S5). Expression of ribA, ribC and ribE genes increased significantly in cells treated with ≥1× MIC of ampicillin relative to untreated controls (Fig. 7a). The expression of the ribB gene increased significantly in cells treated with ≥4× MIC of ampicillin relative to untreated controls.
Finally, we tested the impact of the deletion of the yeeO gene that codes for the MATE transporter protein, which exports both FMN and FAD 30 on ampicillin-induced AF (Fig. 7b). At 1× MIC of ampicillin, we found that the ΔyeeO strain had higher AF than the control strain, while there was no difference at 0.25× or 0.5× MIC. Taken together, our data suggest that flavins are major contributors to the observed increase in AF in ampicillin-treated cells.

Protective induction of genes coding for flavoproteins in ampicillin-treated cells.
It was estimated that 1 to 3% of all bacterial enzymes, most of which are involved in redox processes, depend on the FMN or FAD cofactors for their activity 14 . Many flavoproteins are involved in the response to various cellular damages, which may explain why flavin biosynthesis pathway is induced by cytotoxic treatments. Binding to flavin cofactors was experimentally confirmed for only a small number of proteins 31 . We examined the activity of promoters of 28 genes coding for the confirmed flavoproteins that we found in the Uri Alon library of transcriptional reporters 23 . In addition, we tested the promoter of the soxR gene which codes for the non-flavin-requiring protein that controls SoxRS regulon, including ribA gene 32 . This regulon is involved in the protection against superoxide and nitric oxide-induced stress. Finally, we used the lacZ gene promoter as a control. The advantage of this control reporter is that it can be fully induced with IPTG. All tested promoters were fused to the gene coding for GFP protein 23 . The cells were grown with and without ampicillin in 96-well microplates, and optical density and GFP fluorescence were measured for 3 hours (Table 1).
We broadly classified the gene promoters that showed significant induction when compared to the lacZ control, in three functional categories according to the gene functions described in the EcoCyc E. coli database (http://ecocyc.org/). (i) Genes coding for protection against ROS and reactive nitrogen species (RNS); soxR, sensor and transcriptional regulator of the SoxRS regulon; fpr, a member of the SoxRS regulon, which is involved in the protection against oxidative agents such as methyl viologen and paraquat; nfnB, reduction of nitroaromatic compounds including the antibiotics nitrofurazone and nitrofurantoin; norV/norW, nitric oxide reductase. (ii) Genes coding for the respiratory chain components: poxB, ndh, nuoA and nuoF. (iii) Genes coding for proteins involved in different metabolic pathways: Sulfur metabolism: cysJ, sulfite reductase involved in the assimilation of sulfate (A cysJ mutant is less sensitive to paraquat than wild type); ssuD/ssuE, alkanesulfonate monooxygenase allowing utilization of alkanesulfonates as a sulfur source (ssuD deletion mutant is sensitive to heat shock). Amino acid metabolism: aroC involved in biosynthesis of aromatic amino acids; ilvI, participates both in valine and isoleucine biosynthesis; metF, involved in biosynthesis of methionine. Energy metabolism: dfp; catalyzes two sequential reactions in the coenzyme A biosynthetic pathway; mqo, TCA cycle enzyme malate:quinone oxidoreductase; wrbA having an NAD(P)H:quinone oxidoreductase activity; frdA, catalytic subunits of the fumarate reductase complex. Increased transcription of genes coding for respiration, energy metabolism, different biosynthetic pathways and protection against ROS and RNS indicates that cells treated with ampicillin induce different cellular mechanisms in order to fix treatment-induced cellular damage and survive. If the increase in these flavoproteins is directly correlated with the increased AF, then the increase in AF should correlate with higher cellular robustness. In order to test this hypothesis, we incubated E. coli cells with the 1× MIC of ampicillin for 2 hours, after which 200 cells per condition, with low and high AF intensity were sorted using FACS, plated on LB agar plates without antibiotic, incubated overnight at 37 °C and compared with untreated controls (Fig. 8). Untreated bacteria did not show any significant difference in the plating efficiency between low AF and high AF populations. However, regarding ampicillin-treated bacteria, the plating efficiency of the high AF population was about 5-fold higher than that of the low AF population. This result supports our hypothesis that AF increase reveals the induction of protective responses to severe stresses.

Discussion
In this study, we examined how the AF of E. coli cells changes as a function of exposure to the different stressors. We found that bactericidal stressors, such as ß-lactam ampicillin, significantly increased cellular AF and that de novo protein synthesis is required for the AF increase. We also showed that an increased AF is not an artifactual consequence of an antibiotic-induced cellular morphology. Several lines of evidence suggest that most of the observed increase in the AF comes from flavins. Firstly, the excitation and emission spectra of the AF of ampicillin-treated cells correspond to the spectra of purified flavins: riboflavin, FMN and FAD. Similarly, Mihalcescu et al. 11 have studied AF of E. coli grown in comparable conditions and have shown that flavins FMN, FAD and riboflavin, were responsible for more than 80% of the fluorescence at wavelength 525 nm. They also reported that most of the remaining 20% of fluorescent molecules could be attributed to other flavin derivatives. Secondly, ribA, ribB, ribC, and ribE genes required for the FMN and FAD biosynthesis are induced with the ampicillin treatment. Thirdly, ampicillin-treated cells with deletion of the yeeO gene that codes for the MATE transporter protein exporting both FMN and FAD, have higher AF increases than ampicillin-treated cells with the functional yeeO. Fourthly, many of the tested genes coding for flavoproteins are overexpressed in ampicillin-treated cells.
We also demonstrated that increased AF is not just a collateral consequence of the cytotoxic treatment, but that it corresponds to the activation of a complex adaptive response that increases cell robustness to severe stresses. This is primarily illustrated by the fact that the plating efficiency of the high AF intensity ampicillin-treated cells was, on average, 5-fold higher than that of the low AF intensity ampicillin-treated cells. Cellular functions of  the observed ampicillin-induced genes indicate how increased flavin production can contribute to increased cell robustness. These genes code for flavoproteins that are involved in respiration, energy metabolism, different biosynthetic pathways and protection against ROS and RNS. The observed induction of the ribA, ribB, ribC and ribE genes by ampicillin treatment is important for increased supply of the FMN or FAD cofactors for flavoproteins. In addition, the ribA gene is a member of the SoxRS regulon and is also inducible by paraquat and superoxide-generating compounds 32 . This protein can also hydrolyze 8-oxo-dGTP which is a potent mutagenic substrate for DNA synthesis 33 . Importantly, AF increased significantly relative to the untreated control when the concentration of bactericidal stressor was close to or above the MIC. Therefore, AF can be used for the evaluation of susceptibility of bacterial pathogens to the bactericidal, non-protein synthesis inhibiting antibiotics.
A common reaction of the prokaryotic and eukaryotic cells to stress is the increase in energy production, which is required to deal with macromolecular damages. For this, cells increase respiratory activity, which increases ATP production, but also accidental ROS production that must be rapidly eliminated. The AF of the NAD and FAD cofactors is widely used to follow changes in energy production by the respiratory chain in eukaryotic cells, as well as to monitor the redox state of the stressed cells 34,35 . FAD fluorescence can be quenched when they are protein bound. However, it was also observed that FAD fluorescence increases when they are bound to mitochondrial lipoamide dehydrogenase (LipDH) and electron transfer flavoprotein (ETF), which are responsible for 75% of the flavin autofluorescence in rat liver mitochondria 36,37 . Importantly, it was reported that the amount of electron-transfer fluorescent flavoproteins correlates with respiration rates in eukaryotic cells 34,36,38 .
NAD and FAD fluorescence reflects the redox state of cells. For example, the reduced form of the NAD, NADH, is strongly fluorescent, while its oxidized form NAD + is poorly fluorescent 2,4 . Thus, it was observed that oxidative stress results in a decrease in fluorescence due to the oxidation of NADH in human cells treated with hydrogen peroxide or menadione 39,40 . The opposite is the case for FAD, which is fluorescent in the oxidized form and not in the reduced form. Our data show that the spectra of ampicillin-treated cells are very similar to the spectra of oxidized FAD, while they did not fit with the spectra of reduced FAD 41 .
An increase in the oxidized/reduced ratio of flavins may result in increased fluorescence without necessarily increasing flavin concentration. This could be the case in the cells treated with sodium hypochlorite, which is a potent oxidizing agent. However, the neosynthesis of flavoproteins also contributes to an increase in AF in stressed cells. The flavoproteins play a crucial role in energy metabolism, drug and xenobiotic metabolism, immune defense, chromatin remodeling, DNA repair, and apoptosis 14,42,43 (and references therein). Therefore, the increased AF of the eukaryotic cells corresponds to the induction of responses to cope with the deleterious effects of the life-threatening stress, like we observed in bacterial cells. For example, increased flavin concentrations were reported to be involved in the increased AF of the human and murine cells following ionizing radiation 44 .
In conclusion, this study shows that increased AF indicates that prokaryotic and eukaryotic cells are experiencing severe life-threatening stress. The observed increase in AF may come from both oxidation and synthesis of flavins. The relative contribution of oxidation and/or flavin neosynthesis to the AF increase is likely determined by the nature and strength of the stressor. Therefore, measurement of AF variations in general and that of flavins and OD 600 nm of treated cells and non-treated control were measured using a fluorimeter. The fluorescence increase was calculated using the non-treated control of each strain as reference. (b) Wild type BW25113 strain and its ∆yeeO derivative were treated with a range of ampicillin concentrations during exponential growth. Autofluorescence (λ ex 440/9/λ em 525/20 nm) and OD 600nm of treated cells and non-treated controls were measured using a fluorimeter. To calculate the increase in the autofluorescence due to ampicillin treatment, autofluorescence of treated cells was first normalized by OD 600nm and then by the normalized autofluorescence of the non-treated controls of each strain. Each value represents the mean (+/−standard error) of the fluorescence increase from six independent experiments. The asterisk represents significant differences with the lacZ control according to unpaired t-test p value: < 0.05.
Scientific REPORtS | (2018) 8:12088 | DOI:10.1038/s41598-018-30623-2 and flavoproteins in particular, can be used to study the cellular metabolic state and the molecular mechanisms involved in the response to stress, but also for various diagnostic applications in prokaryotic and eukaryotic cells.

Material and Methods
Strains, growth media and culture conditions. Bacterial and yeast strains, as well as human cell lines used in this study, are described in Table S1. Overnight liquid bacterial cultures were grown aerobically in 50 mL Falcon tubes in Lysogeny broth (LB) medium, Difco LB broth, Miller (Becton Dickinson, Franklin Lakes, USA) at 37 °C with shaking. For experiments, overnight cultures were diluted 10 4 -fold in appropriate media supplemented with antibiotics or chemicals when needed, and grown aerobically at 37 °C with shaking in 50 mL Falcon tubes.
Yeast strains were grown aerobically in 50 mL Falcon tubes at 30 °C for 48 hours in liquid Yeast Extract-Peptone-Dextrose (YPD) (Becton Dickinson) medium supplemented with 2% (final) glucose. For    Flow cytometry and cell sorting. We used a Gallios (Beckman Coulter, Brea, USA) analyzer 10 color, 4 laser system to analyze samples. FL1 channel (excitation at 488 nm with bandpass filters of 525/20 nm) was used for autofluorescence analysis. FL6 channel (excitation at 638 nm with bandpass filters of 660/20 nm) was used for AF633H and TOPRO ® -3 analysis. Samples were run at a rate between 1,000 and 2,000 events/s; 20,000 to 50,000 events were collected depending on sample type (prokaryotic or eukaryotic cells). Data acquisition was performed using the Kaluza software.
For ImageStream cytometry, samples were acquired on the ImageStreamX MarkII system (Amnis Corporation, Seattle, USA) using a 488 nm laser (Bandpass filter 505-560 nm). 20,000 events per condition were collected. The IDEAS software (Amnis) was used to analyze the measurements. Image acquisition and image analysis were performed using the IMAG'IC Facility (Institut Cochin, Paris, France).
For cell sorting, we used a S3e cell sorter (Bio-Rad, Hercules, USA). FL1 channel (excitation at 488 nm emission at 525/30 nm). Samples were run at an event rate of 1,200 events/s and 200 cells per condition were sorted in 200 µL 1x PBS with the "single cell" sort mode. Data acquisition was performed using the ProSort TM software (Bio-Rad).

Autofluorescence variation analysis.
Overnight bacterial and 48 hour yeast cultures were diluted in 20 mL LB or YPD medium respectively, to obtain a final 600 nm Optical Density (OD 600nm ) ≤ 0.05 with a Biophotometer plus (Eppendorf, Hamburg, Germany). Cultures were incubated aerobically at 37 °C for bacteria and 30 °C for yeast, in 50 mL Falcon tubes with orbital shaking until reaching an OD 600nm of 0.2 (AU). At this point, the bacterial cultures were subdivided into 50 mL Falcon tubes and sodium hypochlorite or different amounts of antibiotics were added to obtain a range of antibiotic concentrations. Incubation was performed at 37 °C for bacteria and 30 °C for yeast, with agitation. Samples of treated and untreated control cells were taken every hour, diluted in 500 µL 1x PBS to obtain around 1,000 to 2,000 events/s for bacteria and 500 to 1,000 events/s for yeast and immediately analyzed with the flow cytometer. For the Imagestream X analysis, aliquots of cells were fixed with 2% (final) methanol free paraformaldehyde for 10 minutes at room temperature, washed with PBS and kept on ice until analysis. HeLa cells were treated with sodium hypochlorite (0.08%; 0.16%, 0.32%). After 1 hour of incubation, cells were collected, washed with PBS, and analyzed by flow cytometry.
Cell viability assessment. Samples of bacterial cells prepared for autofluorescence variation analysis (see above) were killed by incubation at 65 °C for 30 min, and stained with 2 µg/mL AF633H or 0.5 µM TOPRO ® -3 iodide dyes. Cells labeled with AF633H were incubated for 15 min on ice and in the dark, and then washed with cold PBS before flow cytometry analysis. Cells labeled with TOPRO ® -3 were incubated at room temperature for 15 min in the dark before flow cytometry analysis.
Analysis of the fluorescence spectra. Bacterial overnight cultures were diluted in LB or M9 -glucose medium (Becton Dickinson) (supplemented with 0.003% B1 vitamin and 1 mM MgSO 4 ) to obtain an OD 600nm ≤ 0.05. M9 medium was used to avoid background fluorescence from proteins and other natural molecules usually found in complex media like LB. Diluted cultures were subsequently incubated aerobically at 37 °C in 50 mL Falcon tubes with orbital shaking until reaching an OD 600nm of 0.2 (AU). At this point, bacterial cultures or medium without bacteria as a control, were distributed in 5 Falcon tubes and different amounts of antibiotic were added to each tube to obtain a range of antibiotic concentrations. Tubes were subsequently incubated at 37 °C with agitation. Every hour, 100 µL of the bacterial cultures or the control medium were sampled and added to the wells of the 96-well black microplate containing 100 µL 1x PBS.
Purified flavonoid compounds were diluted in sterile water to obtain a stock concentration of 20 µM. 100 µL of the stock solution were added to 100 µL 1x PBS in the 96-well microplate. OD 600nm measurements and fluorescence scans were performed using an Infinite M200 PRO and an Infinite M1000 (Tecan, Männedorf, Switzerland) reader.

Gene expression measurements.
To evaluate how antibiotic treatments affect gene expression, we used strains from a library of E. coli strains carrying transcriptional fusions of gfp to different promoters 23 . Overnight cultures containing 12.5 µg/mL kanamycin were diluted in LB medium to obtain an OD 600nm of 0.1. 200 µL samples were then dispensed in the 96-well microplates. After 2 hours of incubation at 37 °C with orbital shaking allowing to obtain bacteria in an exponential growth phase, 40 µL of culture were added to a 37 °C warmed microplate containing a range of ampicillin concentrations (0 to 16 mg/L). 25 µL of mineral oil were added to the top of each well to prevent evaporation. Using a programmable robotic system "Freedom Evo" (Tecan), plates were incubated at 37 °C. OD 600nm and GFP fluorescence (λ ex 488/9 nm λ em 525/20 nm) were measured using an Infinite M200 PRO. lacZ promoter was used as a negative control. When the same promoter was used as a positive control, 0.2 mM IPTG were added to the growth medium. The medium background was subtracted and GFP fluorescence was divided by the OD 600nm . Gene expression was presented as the increase in fluorescence compared to the non-treated condition.
Autofluorescence and cell survival experiment. Overnight E. coli cultures were diluted in 20 mL LB in 50 mL Falcon tubes to a final OD 600nm ≤ 0.05, and incubated aerobically at 37 °C with orbital shaking until an OD 600nm of 0.2 (AU) was reached. Half of the obtained bacterial cultures were treated with 4 mg/L of ampicillin, which corresponds to the MIC, while the other half was used as an untreated control. After 2 hours of incubation, aliquots of the untreated and treated cultures were diluted in PBS and analyzed using the S3e cell sorter. Samples were run at 1,200 events/s, the FSC/FL1 dot plot was established and 2 gates were defined. The first gate corresponded to the least fluorescent cells and the second to the most fluorescent ones. 400 cells in each gate were sorted into 200 µL 1X PBS using the "single cell" sort mode. A 100 µL suspension of each sorted cell population was deposited on LB agar plates without ampicillin allowing cell recovery. Colony forming units were counted after overnight incubation at 37 °C. Data analysis. Flow cytometry data analysis and fluorescence normalization were performed using the Flowjo © V10.2 software (Flowjo, Ashland, USA). Data from the microplate reader and Imagestream X were processed using Magellan 7.2 (Tecan) and Excel © 2010 (Microsoft ® , Redmond, USA). Statistical analyses were performed using Prism V7.0 (GraphPad TM , La Jolla, USA). FCS Express V4 (De novo software, Los Angeles, USA) was used to export raw data from the flow cytometer for correlation analysis.