Sodium Fluoride Exposure Leads to ATP Depletion and Altered RNA Decay in Escherichia coli under Anaerobic Conditions

ABSTRACT Although fluoride-containing compounds are widely used to inhibit bacterial growth, the reprogramming of gene expression underlying cellular responses to fluoride, especially under anaerobic conditions, is still poorly understood. Here, we compare the genome-wide transcriptomic profiles of E. coli grown in the absence (control) or presence (20 and 70 mM) of sodium fluoride (NaF) under anaerobic conditions and assess the impact of fluoride-dependent ATP depletion on RNA turnover. Tiling array analysis revealed transcripts displaying altered abundance in response to NaF treatments. Quantile-based K-means clustering uncovered a subset of genes that were highly upregulated and then downregulated in response to increased and subsequently decreased fluoride concentrations, many of which (~40%) contained repetitive extragenic palindromic (REP) sequences. Northern blot analysis of some of these highly upregulated REP-containing transcripts (i.e., osmC, proP, efeO and yghA) confirmed their considerably enhanced abundance in response to NaF treatment. An mRNA stability analysis of osmC and yghA transcripts demonstrated that fluoride treatment slows down RNA degradation, thereby enhancing RNA stability and steady-state mRNA levels. Moreover, we demonstrate that turnover of these transcripts depends on RNase E activity and RNA degradosome. Thus, we show that NaF exerts significant effects at the whole-transcriptome level under hypoxic growth (i.e., mimicking the host environment), and fluoride can impact gene expression posttranscriptionally by slowing down ATP-dependent degradation of structured RNAs. IMPORTANCE Gram-negative Escherichia coli is a rod-shaped facultative anaerobic bacterium commonly found in microaerobic/anaerobic environments, including the dental plaques of warm-blooded organisms. These latter can be treated efficiently with fluoride-rich compounds that act as anticaries agents to prevent tooth decay. Although fluoride inhibits microbial growth by affecting metabolic pathways, the molecular mechanisms underlying its activity under anaerobic conditions remain poorly defined. Here, using genome-wide transcriptomics, we explore the impact of fluoride treatments on E. coli gene expression under anaerobic conditions. We reveal key gene clusters associated with cellular responses to fluoride and define its ATP-dependent stabilizing effects on transcripts containing repetitive extragenic palindromic sequences. We demonstrate the mechanisms controlling the RNA stability of these REP-containing mRNAs. Thus, fluoride can affect gene expression posttranscriptionally by stabilizing structured RNAs.

inhibition of these enzymes reduces ATP levels (8). Here, we deployed increasing NaF concentrations (0, 20,40,50,60,70,80, and 160 mM) to study ATP depletion in E. coli under anaerobic conditions. We cultured the E. coli MG1655 strain on minimal medium (M9) containing 0.4% glucose in a 1 L Winpact Bench-Top fermentor, as described previously (19), and then measured ATP levels in the cultures. We found that ATP levels gradually declined in response to increasing NaF concentration and prolonged incubation (Fig. 1A). In particular, we found that treatment with 80 or 160 mM NaF led to a sharp decrease (;5-or 10-fold, respectively) in ATP level even after only 2 min, which remained low for up to 32 min of treatment. In contrast, at NaF concentrations in the range of 20 to 70 mM, ATP levels declined more gradually, reaching a minimal level after ;8 min of treatment and remaining low thereafter. To study the impact of NaF treatment, we chose to compare the effects of 20 mM NaF and 70 mM NaF after 8-min treatments (causing moderate and profound ATP depletion, respectively) versus control (0 mM NaF) in further experiments. Moreover, to determine if NaF treatment is accompanied by any phenotypic change, we also examined cell morphology, which revealed that untreated (control) and NaF-treated cells displayed a similar filamentous morphology (i.e., cell length .5 mm; Fig. 1B), mimicking the cell morphology of anaerobically grown E. coli MG1655 (19). Under the aforementioned experimental conditions, there were no obvious differences in cell size, appearance or integrity, indicating that our NaF treatments did not elicit any morphological changes in cells intended for further study of the impact of NaF on gene expression in E. coli grown anaerobically.
Preparation of anaerobic cultures and genome-wide gene expression patterns under different NaF treatments. To analyze the impact of NaF by tiling array, we prepared two sets of cultures in M9/glucose medium. The first one (Cd cultures) included the untreated cell culture (0 mM NaF, control), as well as two cultures treated with either 20 or 70 mM NaF (Fig. 1C). The second set of cultures (Ce cultures) focused on partly reversing the effects of NaF treatment by diluting the Cd cultures with fresh M9/glucose medium and further incubating them to reach the cell density of the original (undiluted) cultures (Fig. 1C). To obtain both sets of cultures, cells were grown anaerobically, as described previously (19), with details of our culture conditions and reagents provided in Materials and Methods. Total RNA was extracted using a hot acidic phenol method (21). After checking total RNA quality, it was converted to labeled cDNA and hybridized with a whole-genome tiling array (NimbleGen) to determine gene expression. These experiments were performed on three biological replicates of each culture. Tiling array signals were mapped to the E. coli K-12 substrain MG1655 reference genome (NC_000913.3) (22), allowing us to detect the expression of 4,417 genes in NaF-treated and control E. coli cells under anaerobic conditions (NCBI GEO accession no. GSE211579).
Differential expression of genes in NaF-treated E. coli cells under anaerobic conditions. (i) Clustering of differentially expressed genes. We used a quantilebased K-means clustering approach to identify gene clusters with similar expression profiles arising from NaF treatment (Fig. 2). That analysis of the expression data revealed six clusters displaying distinct profiles, of which cluster 1 (n = 1,293 genes), cluster 2 (n = 975), cluster 3 (n = 974), and cluster 4 (n = 714) did not show any differences (log 2 fold changes) in gene expression potentially correlated with the level of NaF in the examined cultures. In contrast, gene expression notably decreased in cluster 5 (n = 341) and increased in cluster 6 (n = 100) in a dose-dependent manner in response to NaF treatment (Fig. 2). We used the functional gene categories assigned according to the STRING platform (https://string-db.org/) to assess the significance of enriched expression of the functionally related genes in clusters 5 and 6 ( Fig. 3A and B). Intriguingly, we found that twice as many transcripts in cluster 6 carried repetitive extragenic palindromic (REP) sequences than those in cluster 5 (Fig. 3D). Indeed, ;50% of the top 50 abundant transcripts associated with cluster 6 carry REPs (Table 1).
Northern blot validation of altered transcript expression levels in response to NaF treatments. Our tiling array data revealed that the abundance of numerous RNAs increased in a dose-dependent manner in response to NaF treatment (0 ! 20 ! 70 mM NaF), and subsequently decreased upon dilution of the NaF-treated cultures. To validate this           expression pattern, we assessed the levels of several transcripts-namely, osmC, proP, efeO, and yghA-by means of Northern blot analysis. Consistently, we found that levels of all four transcripts increased upon NaF treatment ( Fig. 3Ca-d; compare lines 1, 2, and 3), and their abundance subsequently decreased following dilution of the NaF-treated cultures (Fig. 3Ca-d; compare lines 4, 5, and 6). We selected osmC and yghA as representatives for further analysis as model mRNA species, because of following features: (i) both transcripts are in the top list of upregulated transcripts upon NaF treatment (Table 1); (ii) these transcripts carry different numbers of REPs (osmC and yghA carry 1 and 5 REPs, respectively), thus allowing us to investigate if less or more REPs can alter RNA decay; and (iii) their abundance was sufficiently high for REP detection by Northern blotting using specific probes. As the impacts of increased salinity (NaCl) on gene expression have been studied previously (23,24), we have referred to these studies to address whether NaCl (70 mM) has impact on gene expression of osmC, proP, efeO, and yghA. Our experimental data ( Fig. S1; compare lines 1 and 2) support the conclusion that the increased levels of the above transcripts are mainly due to exposure to fluoride rather than due to a general increase in osmolarity.
Degradation of REP-containing osmC and yghA mRNAs depends on RNase E and is impaired by NaF treatment. To determine if osmC and yghA transcript accumulation is posttranscriptionally regulated, we employed Northern blot analyses to measure the stability of osmC and yghA mRNAs in the presence or absence of NaF. As shown in ) in untreated (0 mM NaF) E. coli were considerably shorter than those following exposure to 70 mM NaF (i.e., .16 min). These results demonstrate that upregulation of osmC and yghA upon NaF treatment is exerted at the posttranscriptional level. To investigate the mechanisms potentially involved in degrading these RNAs, we explored if RNase E is required for osmC and yghA RNA degradation. To do so, we compared their transcript stability in wild-type E. coli N3433 and in RNase E temperature-sensitive strain N3431 (25), in which RNase E activity is inhibited under nonpermissive temperatures, thereby leading to RNase E-dependent transcript accumulation at nonpermissive temperatures solely in N3431. Thus, RNase E-dependent transcripts could be identified by comparing the corresponding Northern blot signals obtained at a permissive (30°C) versus nonpermissive (44°C) temperature. We observed that osmC transcripts accumulated in the N3431 strain ( Fig. 4Cb; compare left and right), and their half-lives increased at 44°C (i.e., .8 min for both FL and IM RNAs) relative to at 30°C (. ). Likewise, we detected no increase in stability of yghA transcripts in N3433 (Fig. 4Ga), in which RNA half-lives were 2.2 6 0.12 min (FL) or .8 min (IM) at 30°C and 1.6 6 0.12 min (FL) or 1.7 6 0.06 min (IM) at 44°C, respectively. In contrast, stabilization of full-length and intermediate yghA mRNAs was observed in the temperature-sensitive N3431 strain (Fig. 4Gb), in which half-lives were 3.3 6 0.08 min (FL) or .8 min (IM) at 30°C and .8 min for both FL and IM RNAs at 44°C. Thus, the higher stabilities of the osmC and yghA transcripts upon inactivation of RNase E indicate that their degradation is RNase Edependent.
In addition, experiments were conducted on Rned500-1061 mutant (19) coding for C-terminal-truncated RNase E (1 to 499 aa; see Fig. 4D and H). This polypeptide contains only the catalytic domain and lacks the scaffolding region involved in interacting with major degradosome components, including PNPase, RhlB helicase and enolase (26,27).
Our data show that osmC and yghA mRNAs are stabilized (half-lives .16 min; Fig. 4Da  and 4Ha) even in the absence of NaF, supporting that degradation of the FL and IM mRNAs is degradosome-dependent. We also demonstrate that maximal stabilization of these transcripts is achieved in the absence of fluoride ions, so addition of NaF does not elicit further increases in transcript levels ( Fig. 4; compare 4Da and Db, and 4Ha and Hb). Collectively, our data indicate that the levels of the osmC and yghA transcripts are greatly impacted by RNase E/degradosome-dependent degradation (Fig. 4C and G). Namely, FL and IM transcripts of osmC and yghA mRNAs are stabilized in an RNase E-temperature sensitive strain and RNase E mutant coding for C-terminal-truncated (500 to 1,061 aa deleted) enzyme (half-lives .8 to 16 min; Fig. 4C and G). In other words, in RNase E thermosensitive and mutant strains, we observe accumulation/stabilization of these transcripts.

DISCUSSION
Fluorine is ubiquitous on Earth. Its ion (fluoride) plays a vital role in many organisms, including humans (4). Clinical studies have revealed the physiological range of fluoride concentrations and demonstrated that insufficient or excessive fluoride intake causes various diseases (28). At high concentrations, fluoride acts as an antimicrobial agent that can inhibit bacterial growth and metabolic activities (20). Therefore, using fluoride-containing toothpaste or other anticavity drugs should impact the composition and function of the oral and gut microbiomes. Recent studies have shown that fluoride exerts its inhibitory effects on bacterial metabolism in many ways (4,29). For instance, it can act as an enzyme inhibitor by interacting with the various metal ions present in many phosphatases, kinases, hydrolases, and other metalloproteins, or as an enhancer of membrane permeability by increasing proton uptake, thereby inducing cytoplasmic acidification. Most studies on fluoride-mediated enzymatic inhibition have focused on eukaryotic enzymes, especially those of yeast. Very similar inhibitory effects have since been observed for Streptococcus sp. (30) and other bacteria. However, the antibacterial actions of fluoride appear to be complex and remain incompletely understood.
One of the best-known protein targets of fluoride is enolase (6,31). This enzyme catalyzes the penultimate step of glycolysis by converting 3-phosphoglycerate to phosphoenolpyruvate (5). Fluoride-mediated inhibition of enolase and phosphoenolpyruvate synthetase blocks glycolysis, thereby potentially depleting ATP. Indeed, NaF treatment of eukaryotic cells has been found to reduce ATP levels (8). Although diminished ATP levels can have profound effects on gene expression at the transcriptional and posttranscriptional levels, especially in animal microbiota under microaerobic conditions, the actual consequences of fluoride on gene expression in bacteria had been unclear.
Here, we used E. coli MG1655 grown under anaerobic conditions to mimic the environment of the dental plaque gastrointestinal tract, allowing us to study the impact of NaF-mediated ATP depletion on gene expression at the whole-transcriptome level as well as RNA decay. We found that ATP levels gradually declined in response to E. coli being exposed to elevated concentrations of NaF under anaerobic conditions (Fig. 1A). Despite inhibited growth, the E. coli cells retained their filamentous morphology (i.e., the typical appearance of anaerobically grown E. coli cells [19]) upon treatment with moderate (20 mM) or high (70 mM) NaF concentrations, and the integrity of control (untreated) and treated cells was almost indistinguishable (Fig. 1B).
As E. coli adapts to altered oxygen availability and/or upon exposure to molecules/ chemicals such as NaF, it typically adjusts its metabolism by reprogramming gene expression (32,33), resulting in changes in steady-state levels of mRNAs that are determined by rates of transcription and decay. Tiling microarrays (34) represent a valuable tool for establishing the whole-genome transcriptomes of microbial organisms exposed to various stress conditions. Here, we used such an array built at 25 nt resolution to observe changes in E. coli MG1655 gene expression upon NaF treatment. We anticipated that NaF-mediated ATP depletion would result in downregulated expression of genes involved in ATP-dependent processes. Nevertheless, we also expected that a subset of genes whose transcripts are stabilized upon NaF-associated inhibition of their ATP-dependent RNA turnover would be upregulated. NaF-dependent cell growth inhibition has previously been reported as reversible, both in vitro and in vivo (9, 10), following dilution of the NaF-treated cultures with fresh media lacking this reagent. To explore that scenario further, first we inhibited enolase activity by means of dose-dependent NaF treatments (i.e., at 0, 20 and 70 mM NaF; Fig. 1C). Then, we diluted the NaF-treated cultures (Fig. 1C) with the intention to reverse the changes in gene expression elicited by NaF. Next, we isolated RNA from each of the cultures and hybridized it to tiling array. Using a quantile-based K-means clustering approach on the tiling array data set, we identified two clusters (5 and 6) comprising genes whose expression changed in a dose-dependent manner upon NaF treatment (Fig. 2). Our data revealed that the 341 genes in cluster 5 were first downregulated and then upregulated following NaF treatment and subsequent dilution of the NaF-treated cultures, respectively. In contrast, the 100 genes in cluster 6 were initially upregulated and then downregulated upon NaF treatment and subsequent dilution.
An increase in expression of the talA-tktB operon controlling the pentose phosphate pathway indicates that the fluoride-dependent inhibition of glycolysis due to the negative effect of fluoride on glycolytic enzymes induces E. coli to increase glucose catabolism via alternative routes such as the pentose phosphate pathway.
Moreover, another group of upregulated genes involves those controlling metal ion transport. As fluoride can associate and form stable complexes/precipitates with a number of metals (e.g., calcium, magnesium, and iron) in bacteria and other organisms (4), exposure to fluoride can potentially elicit deficiencies in the biologically active forms of these ions, thus necessitating cells to increase the concentrations of available free metal ions. Consistently, we found that NaF treatment upregulated the expression of a large number of genes controlling the uptake of iron (e.g., entCEBA nfeF, exbBD, feoABC, etc.) and magnesium (i.e., mgtA and mgtS).
Furthermore, many upregulated genes are involved in cell envelope stress (e.g., pspA otsBA, treF, and osmY) and lipid biosynthesis (pgpC, ybhP, cslB, and ybhN). Many of them are positively regulated by the nucleotide alarmone ppGpp. It seems likely that their upregulation is a result of the ppGpp-mediated stress response caused by defects in ATP production upon NaF exposure. This notion is well supported by a recent study reporting novel roles for ppGpp in bacterial physiology (35). In contrast, the putative increase in ppGpp level in response to NaF treatment might account for the downregulation of numerous genes involved in translation (e.g., infA, rrf, prfC, etc.).
Further examination of our data revealed that the mentioned transcriptional/posttranscriptional regulators likely contribute to NaF-dependent E. coli gene regulation under anaerobic conditions. In particular, we found that downregulation of the dual regulator H-NS (encoded by hns) appears to relieve (activate) expression of several genes/operons (e.g., proVWX, osmC, and osmY) involved in the osmotic stress response, and concomitantly prevent its role as a transcriptional activator, thus reducing transcription (downregulating) of cydAB (energy production) and degP (periplasmic endoprotease), among others. Similarly, downregulation of Fis, a transcriptional activator of many operons, including those coding for tRNA precursors, is consistent with an observed decrease in level of numerous tRNAs ( Table 2; cluster 5).
In addition to transcription factors, a number of global regulators (Rnc, Poly[A] polymerase I and RNA pyrophosphohydrolase) act at the posttranscriptional level. As the level of the corresponding genes (rnc, pcnB and rppH, respectively) is downregulated in the presence of NaF, it can be envisaged that NaF treatment slows down RNA processing and turnover.
Notably, we observed that the abundance of numerous transcripts increased upon NaF treatment. Although transcript abundances can generally increase or decrease in response to various environmental signals, the NaF-mediated reduction in ATP level we observed for E. coli under anaerobic conditions-and the potentially concomitant decrease in ATP-dependent metabolic activities such as translation, transcription, and degradation of biomolecules (e.g., DNA and RNA)-suggest that profoundly increased abundance of certain transcripts might not be due to a considerably higher rate of transcription (which, in fact, should be less efficient upon ATP depletion). Instead, it could be a result of RNA stabilization as part of the posttranscriptional control playing essential role under various stress conditions (36,37). Indeed, our analysis of the top upregulated transcripts indicated that a large proportion of them contain stable secondary structures (such as REPs elements) that impede 39 to 59 end RNA decay by exonucleases. In fact, we found that a large percentage (;40%) of highly stabilized transcripts (70 versus 0 mM NaF) carried REPs (Fig. 3D).
Although RNA degradation by 39 exonucleases (e.g., PNPase or RNase II) can be promoted by RhlB-mediated RNA unwinding or 39 end polyadenylation, both those reactions are ATP-dependent and, consequently, they should be less efficient in circumstances where fluoride depletes ATP. Consistently, our Northern blot hybridization using a probe targeting 59 regions of several upregulated transcripts (i.e., osmC, proP, efeO, and yghA) revealed stabilization of these REP-containing transcripts in a dose-dependent manner upon NaF treatment, and their abundances subsequently decreased after dilution of the cultures with fresh medium lacking NaF (Fig. 3Ca-d). Previous studies have revealed that a number of osmotic stress-related genes (i.e., proP, proV, otsBA, osmC, osmY; Table 2) upregulated by NaF treatment were also upregulated by 0.4 M NaCl at both transcript levels (24,38). However, in contrast to 0.4 M NaCl, we only used 70 mM NaF, i.e., at a considerably lower concentration (.5-fold). Therefore, it seems more likely that transcript stabilization per se (rather than osmotic stress) is responsible for the observed regulation. Moreover, our additional experimental data (Fig. S1) support the conclusion that the increased levels of osmC, proP, efeO, and yghA transcripts are mainly due to exposure to fluoride rather than due to increase in osmolarity.
Further analysis on two selected transcripts, i.e., osmC (carrying 1 REP) and yghA (carrying 5 REPs), revealed that the observed changes in their abundance were likely caused by enhanced RNA stability, as manifested by the prolonged half-lives of the transcripts in the presence of 70 mM NaF (Fig. 4Ba and Bb and 4Fa and Fb). Moreover, the initial increase in transcript level of full-length osmC we detected (Fig. 4Bb) is likely attributable to its efficient stabilization in the presence of 70 mM NaF before complete rifampicin-mediated transcriptional inhibition is attained. To date, studies on the structure of RNA polymerase and inhibition by rifampicin have predominantly focused on the holoenzyme containing the most abundant sigma factor 70, which is responsible for recognizing most E. coli genes. However, in E. coli cells, there are at least seven alternative sigma factors responsible for recognizing different types of promoters (39), and it has been reported that transcription from the sigma70-and sigma32-dependent promoters is differentially inhibited by rifampicin (40). Furthermore, it remains unclear if binding of different s subunits to the core enzyme represents a simple case of factor substitution that has no effect on core structure or if their binding significantly alters the conformation of the core under different growth conditions. We speculate that transcription of osmC under anaerobic conditions is dependent on an alternative sigma factor and so it is less sensitive to rifampicin, particularly upon NaF treatment. As a result, the remaining percentage of FL osmC mRNA increases within a short time after rifampicin treatment, as shown in Fig. 4Bb.
Our experimental data support the notion that regulation of osmC transcript levels is exerted at the posttranscriptional level by reducing the degradation rate of fulllength transcript and its decay intermediate (Fig. 4Ba and Bb and 4Fa and Fb).
The main pathway of RNA turnover in E. coli is controlled by the endoribonuclease RNase E, a primary component of the multienzyme ribonucleolytic complex termed the RNA degradosome (16,17). Other major components of degradosomes include enolase, RhlB RNA helicase and the 39-to-59 exonuclease polynucleotide phosphorylase (PNPase) (15)(16)(17). RNA degradosome assembly is required for efficient RNA turnover under both aerobic and anaerobic conditions. We have recently elucidated a mechanism by which E. coli uses enolase-bound degradosomes and the small RNA DicF to alternate from rodshaped to filamentous form in response to anaerobiosis (19). However, despite some advances, our understanding of how RNA degradation is regulated in bacteria under anaerobic conditions remains limited. To explore anaerobic RNA decay further, we examined if RNase E is required to degrade E. coli osmC and yghA transcripts using an RNase E temperature-sensitive N3431 strain (25) (Fig. 4Ca and Cb and 4Ga and Gb).
As described above, the osmC and yghA genes carry REPs at the 39 end. Although REPs possess stable structures that inhibit 39-to-59 exonucleolytic degradation of the cognate transcripts, and/or restrict cleavage at potentially susceptible RNase E sites (41), E. coli deploys mechanisms to overcome the stabilizing effect of REPs. For instance, it was reported previously that degradosomes exert ATP-dependent activity that aids in the unwinding of structured RNAs by RhlB helicase, facilitating their subsequent degradation by PNPase (17) in the 39 to 59 direction (42). Alternatively, REP-containing RNAs can be destabilized by polyadenylation, which is normally catalyzed by poly(A) polymerase I (PAPI) that adds poly(A) tails to the 39 end of E. coli transcripts, thereby expediting exonuclease binding and subsequent digestion of the RNAs by PNPase, RNase II, and RNase R (43).
Given that osmC and yghA transcript abundance increased considerably upon NaF fluoride treatment (Fig. 4Ba and Bb and 4Fa and Fb), it seems likely that this effect could be caused, at least in part, by inhibition of their RhlB-dependent and/or PAPI-dependent decay, presumably due to NaF-dependent ATP depletion. Moreover, degradation of both transcript types may be dependent on RNase E, the key player in E. coli mRNA decay, whose quaternary structure regulates RNA turnover in a substrate length-dependent manner (44). Indeed, our assessment of RNA half-lives in wild-type (N3433) and temperature-sensitive RNase E mutant (N3431) strains revealed enhanced stability of osmC and yghA transcripts upon inactivation of RNase E (i.e., in N3431 at the nonpermissive temperature). This finding indicates that RNase E-dependent mechanisms play a critical role in degrading structured RNAs under anaerobic conditions. It will be interested to address in future, whether NaF treatment affects these mechanisms by altering RNase E localization or its tertiary structure under anaerobic conditions.
Collectively, our study reveals that ATP depletion under anaerobic conditions upon exposing E. coli to 20 mM or 70 mM NaF leads to dramatic changes in gene expression. Moreover, NaF-induced upregulation of certain genes in E. coli is likely exerted at the posttranscriptional level, apparently by impeding ATP-dependent unwinding (or polyadenylation) of the respective transcripts. Inhibition of these pathways prevents efficient 39-to-59 degradation by exonucleases of the structured RNAs. This scenario is supported by the increased half-lives (chemical stabilization) of structured RNAs upon NaF treatment, as demonstrated by Northern blotting analysis of osmC and yghA transcripts, which are protected from exonucleolytic decay at their 39 ends by REP sequences. Furthermore, we report that degradation of these transcripts is RNase E-dependent, implying that anaerobic turnover of structured RNAs likely involves the combined action of both exo-and endoribonucleases.

MATERIALS AND METHODS
Bacterial strains, growth conditions and sodium fluoride treatment. E. coli MG1655 K-12 strain was grown anaerobically in a 1 L Winpact bench-top fermentor (Major Science Inc., USA) with M9 medium (45) supplemented with 0.4% glucose, as described previously (19,46). In brief, the overnight culture was subcultured in a 1 L fermentor vessel containing M9/glucose medium that had initially been incubated for ;12 h with a constant flow rate of 0.5 L/min of sparged pure nitrogen gas to maintain anaerobic conditions. Then, overnight cultures that had been diluted to an optical density at 460 nm (OD 460 ) of ;0.05 were added. The pH, temperature, and agitation were maintained at 7.0, 37°C, and 200 rpm, respectively.
For our ATP depletion experiments, 500 mL of each culture was grown anaerobically to OD 460 ;0.4, and then NaF was added to obtain a final concentration of 20, 40, 50, 60, 70, 80, or 160 mM, whereas the control culture (0 mM NaF) remained untreated. After incubating the cultures for 8 min, aliquots were withdrawn for ATP measurement and microscopy imaging.
It was reported previously that fluoride ions present at millimolar concentrations in bacterial culture media inhibit cell growth in E. coli under aerobic conditions (47). We tested the effect of fluoride on growth of E. coli under anaerobic conditions (Fig. S2; compare 0 mM versus 20 to 160 mM). We found that sodium fluoride likewise inhibits cell growth, leading to its arrest at a concentration of 70 mM NaF. In our experiments, we treated cells for 8 min to ensure that cells are alive and physiologically active.
Two sets of cultures were prepared in M9/glucose medium for tiling microarray (Fig. 1C). The first one (Cd cultures) included untreated cells (0 mM NaF, control) and two cultures treated with 20 or 70 mM NaF, respectively. The second set of cultures underwent partial reversal of the effects of NaF treatment by diluting the Cd cultures with fresh M9/glucose medium and incubating them to reach the cell density of the original (undiluted) Ce cultures (Fig. 1C). To obtain both sets of cultures, 2 mL M9/glucose medium inoculated with a single E. coli colony was incubated overnight at 37°C to obtain the starting culture (Ca) for inoculation into 50 mL M9/glucose medium placed in a 250 mL Erlenmeyer flask and grown overnight under the same growth conditions, resulting in Cb culture. This later was transferred to a 1 L fermentor vessel containing 300 mL of M9/glucose (maintained under anaerobic conditions at an OD 460 of ;0.05), and the cells were grown anaerobically until the OD 460 reached ;0.4 (culture Cc), before transferring the fermentor vessel to an Anaerobic System Glove Box to maintain the anaerobic conditions. Aliquots (100 mL each) of culture Cc were transferred to three 500 mL Erlenmeyer flasks (representing the Cd cultures), and NaF was added to two of them to obtain a final concentration of 20 mM or 70 mM, with the third culture remaining untreated (control). After incubating all Cd cultures for 8 min, aliquots were withdrawn for RNA isolation, and the remaining NaF-treated cultures (20 mL) were diluted (1:5) with fresh M9/glucose medium to reduce the NaF concentration to 4 or 14 mM, respectively. These diluted cultures (Ce) were further incubated under the same conditions and, when they reached an OD 460 of ;0.4, aliquots were withdrawn for RNA isolation. These experiments were performed on three biological replicates of each culture.
To measure RNA half-lives, MG1655 cultures were grown anaerobically as described above. Then each 0 mM and 70 mM NaF (8 min) culture was treated with rifampicin to inhibit new RNA synthesis. In the experiments carried out on the rne temperature-sensitive N3431 and isogenic wild-type N3433 strains (25), the cultures were grown anaerobically at 30°C. Then, each culture was treated with rifampicin or shifted to 44°C for 1 h prior to addition of rifampicin (0.5 mg/mL).
ATP measurement. ATP was extracted from cells using perchloric acid according to a previously described procedure (48). In brief, before withdrawing an aliquot of culture, 200 mL of ice-cold 3.0 N HClO 4 was placed into an Eppendorf tube and kept on ice, to which the culture aliquot (800 mL) was added. The tube was vortexed and kept on ice for 10 min, before adding 200 mL of 3.0 N KOH in 0.3 M HEPES (pH 7.8) to neutralize HClO 4 , to stabilize the ATP against acid-catalyzed hydrolysis, and to precipitate KClO 4 . After the precipitate had formed and settled, the supernatant was removed and stored frozen at 220°C for later assays. ATP level was measured by luciferase assay in a luminometer (EnSpire multilabel plate reader, PerkinElmer) using an ATP detection kit (Invitrogen) according to the manufacturer's instructions. In brief, the assay was carried out on black flat-bottomed 96-well microtiter plates. Each sample (10 mL) and 90 mL of a standard reaction solution were mixed in the individual wells and luminescence (a.u.) was recorded at 570 nm. Changes in ATP level were calculated and the percentage of ATP (normalized to untreated control) was determined. All measurements of ATP level were performed in triplicate.
Microscopy imaging. For cell imaging, 2 mL of cell culture was placed on the middle of a glass plate, then covered with a cover slide and sealed. Differential interference contrast (DIC) images were captured using an Axio Imager Z1 (Carl Zeiss) microscope at 100Â magnification. The images were analyzed in ZEN Blue Edition software (Carl Zeiss).
RNA isolation and Northern blot analysis. Total RNA was extracted by means of the hot acidic phenol method (21). In brief, aliquots (43 mL) of cell cultures grown anaerobically in M9/glucose medium as described above were mixed with cold stop solution (5% phenol in ethanol) at an 8:1 ratio. The cells were collected by centrifugation (4,000 g, 4°C, 15 min), suspended in 2 mL KJ medium (50 mM glucose, 25 mM Tris-HCl pH 8.0, 10 mM EDTA pH 8.0, 100 mM NaCl), then added into 2 mL boiling lysis buffer (0.2 M NaCl, 20 mM Tris-HCl pH 7.5, 40 mM EDTA, 0.5% SDS), and incubated in boiling water for 30 sec before adding 2 mL acidic phenol (pH 4.5). The contents of each tube were gently mixed by inverting the tube ;20 times. Total RNA was extracted into aqueous phase by centrifugation (4,000 g, 4°C for 1 h). Then, the RNA was precipitated by adding 1 volume of isopropanol and 1/10 volume of 3 M sodium acetate (pH 7.8). The resulting RNA suspensions were stored at 220°C. Prior to use, the RNA was precipitated by centrifugation (30,000 g, 4°C for 15 min), washed with 70% ethanol, centrifuged again (30,000 g, 4°C for 15 min), and suspended in 20 mL water.
The purified RNA was first analyzed by Bioanalyzer (Agilent 2100). A ratio of rRNAs [23S/16S] . 1.4 was considered to reflect RNA of high quality, so it was used for subsequent tiling array analysis.
For Northern blot analysis, aliquots of total RNA (10 mg) were individually mixed with equal volumes of 2Â RNA loading dye (0.03% bromophenol blue, 0.03% xylene cyanol FF, 0.5 mM EDTA in formamide), incubated at 65°C for 10 min, chilled on ice, and separated on 3.5% or 6% polyacrylamide-urea gels. The fractionated RNAs were transferred to ZetaProbe blotting membranes (Bio-Rad) at 400 mA (60 min at 4°C) in 0.5ÂTBE buffer and UV cross-linked (1,200 mJ) using a Stratalinker UV Crosslinker 2400 system (Stratagene). RNAs were stained with 0.3% methylene blue/0.3 M sodium acetate solution (pH 5.2) for 1 min and further washed with miliQ water for 5 min on a shaker, and then scanned using an Epson Perfection 4990 photo scanner. The membranes were cut into smaller pieces and, after prehybridization with ULTRAhyb hybridization buffer (Ambion) at 65°C for 12 h, they were further hybridized with the specific oligonucleotide probes internally labeled with [a-32 P] ATP oligodeoxynucleotide or 59-labeled with [g -32 P] ATP (Table S1). The internally labeled RNA probes were synthesized using an in vitro transcription kit (MAXIscript, Ambion) and 59-labeled DNA probes were labeled using T4 polynucleotide kinase (T4 PNK, Thermo Scientific) according to the manufacturer's instructions. DNA templates for transcription of individual RNA probes were generated by PCR using gene-specific primers (Table S1). Radioactive probes were purified using a MicroSpin G-25 column (GE Healthcare). We used RiboRuler high and low range RNA ladders (Thermo Scientific) to estimate the approximate size of RNAs. After hybridization at 65°C (internally labeled probes) or 42°C (oligonucleotide probes) overnight, the membranes were washed twice with preheated wash buffer (5Â SSC with 0.5% [wt/vol] SDS) at 65°C (internally labeled probes) or 42°C (oligonucleotide probes) and exposed to Phoshor imaging plates (FujiFilm) at 280°C. The RNA bands were detected using a Typhoon FLA 9000 biomolecular imager (GE Healthcare), and the relative amount of individual RNA species in each band was calculated by normalizing its signal to 5S or 16S rRNA signals using as controls for our Northern blots. Ribosomal RNAs are very stable and therefore their steady-state intracellular levels are almost the same, which was supported by their expression profiles in our tiling array data. In other words, all three ribosomal RNAs (i.e., 23S, 16S, and 5S) belong to clusters 1 or 2 (Fig. 2), which indicates that their levels were not significantly affected by NaF treatment, so they can be used as loading controls for Northerns. For transcript level comparisons we used the same cell mass.
Tiling array. The purified RNA was first analyzed by Bioanalyzer (Agilent 2100), with a ratio of rRNAs (23S/16S) .1.4 being considered of sufficiently high quality for tiling array analysis. The RNA was then converted to labeled cDNA, which was hybridized to a whole-genome tiling array (NimbleGen) using standard NimbleGen operating protocols. In brief, double-stranded cDNA was synthesized using a Nimblegen cDNA synthesis kit from 10 mg total RNA and up to 1 mg of the resulting cDNA was plabeled with Cy3-9mer primers by using a Nimblegen labeling kit. Following labeling, 5 mg of Cy3-cDNA was hybridized for 18 h at 42°C in a MAUIHybridizer unit with a Nimblegen 385K E. coli tiling array by means of the Nimblegen hybridization kit. Afterwards, the array was washed using Nimblegen washing buffers, followed by drying on a microarray dryer. The microarray slides were scanned using the Axon GenePix 4200A microarray scanner at 5 mm resolution and photomultiplier tube (PMT) at 480 V. The scanned images were processed using NimbleScan software. The array images were further processed using Nimblegen's standard protocol for Nimblescan ChIP data extraction.
Bioinformatics. The expressions for each gene were quantified by summarizing the tiling-array probe signals (subColSummarize) and normalizing across all the samples ("normalize.quantiles") using the R package "preprocessCore" (version 1.58.0). For each sample, relative fold changes of each transcript were calculated by comparing it to the gene expression value of the control sample. The k-means clustering analysis (k = 6) was performed by running the R function "kmeans" on the log 2 -transformed median fold changes in expression across all samples. We applied STRING analysis (49) to obtain the interaction network of differentially expressed genes. The Biocyc Database Collection (https://biocyc .org) was used to identify REPs in the E. coli MG1655 K-12 strain.
Quantification and statistical analysis. Integrated band intensity was quantitated using Fiji (50) software based on the intensity of signals obtained by scanning the phosphor image plates. Optical density was obtained using a Prema PRO-739 visible spectrophotometer. All statistical tests were performed using GraphPad Prism version 9.0.
Data availability. The tiling array data and related information have been deposited to NCBI GEO (accession no. GSE211579).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.5 MB.