Bacterial strains used in this study are listed in Supplementary Table S1. E. coli DH5α strain was the general cloning host. All E. coli strains were derived from MG1655 strain and grown at 37°C in LB (Lysogeny Broth) liquid medium from Oxoid (LP0021B and LP0042B; Clark and Maaløe, 1967) or NA (Nutrient Agar) solid medium from Oxoid (CM0003B). M9 minimal liquid medium was composed of M9 salt (60mM Na₂HPO₄, 22mM KH₂PO₄, 8mM NaCl, and 20mM NH₄Cl), 1mM MgSO₄, 100μM CaCl_2, 1μg/ml thiamine, and 0.025% or 0.2% glucose. MOPS minimal liquid medium was prepared as previously described (Neidhardt et al., 1974) and was free of nucleobases and amino acids. When necessary, media were supplemented with 80μg/ml X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) and antibiotics used at the following concentrations: 50μg/ml ampicillin, 50μg/ml chloramphenicol, and 25μg/ml kanamycin. P1 transductions were performed as previously described (Thomason et al., 2007).

Expression of lambda recombinase from pKD46 (Datsenko and Wanner, 2000) was induced by adding 0.2% of arabinose and by growing cells at 30°C during 1.5h, pKD46 plasmid was then eliminated by streaking colonies on NA solid medium and by incubating plates overnight at 37°C. Kanamycin resistance cassette was flipped out as previously described (Cherepanov and Wackernagel, 1995).

Plasmids used in this work are listed in Supplementary Table S1 and were extracted using Monarch plasmid miniprep kit (Biolabs). PCRs were carried out from colonies with Phusion DNA polymerase (Thermo Scientific) to amplify DNA fragments used for cloning or strain constructions and Gotaq flexi DNA polymerase (Promega) for diagnostic PCR. PCR products were purified using the Monarch DNA gel extraction kit (Biolabs).

The plasmid derivatives used in this study (Supplementary Table S1) were constructed by amplifying genes by PCR from template chromosomal DNA using primers listed in Supplementary Table S2 and by digesting DNA with restriction enzymes indicated in Supplementary Table S2.

Transcriptional and translational fusions reporter strains were constructed by two-step λ red-mediated recombination, adapted from Blank et al. (2011). A chloramphenicol resistance cassette was amplified by PCR together with an I-SceI recognition site using pWRG100 plasmid as template with primers (357/358) for translational and (353/354 or 577/354) for transcriptional fusions (Supplementary Table S2) containing a 50bp homologous sequence with upstream and downstream of the ytfK locus. MG1655 strain harboring pKD46 plasmid was electroporated with the resulting fragment and the insertion of the I-SceI:cat fragment into the target locus was verified by PCR and then P1 transduced into the TB28 strain. The resulting strain was then transformed with pWRG99 harboring the I-SceI endonuclease under the control of an anhydrotetracycline inducible promoter. The chloramphenicol cassette was then removed by counter selection using lambda red recombination to insert a PCR product complementary to the flanking regions of the I-SceI:cat cassette on the chromosome. The PCR product of the different transcriptional fusions was generated with primers (355/356) and pGH254:P_{ytfK P1+P2}:lacZ or pGH254:P_{ytfK P2}:lacZ as templates or primers (708/356) and pGH254:P_{ytfK P1}:lacZ as template (Supplementary Tables S1, S2). The translational fusion PCR product was obtained by PCR amplification with primers (360/361) and pGH254 as template (Supplementary Tables S1, S2). Selection of successful recombinants was mediated by spreading cells on NA plates containing ampicillin, X-gal, and 1μg/ml anhydrotetracycline. The proper integration of transcriptional or translational fusions was confirmed by diagnostic PCR and then sequenced.

Deletion of the crp gene was achieved by replacement of the crp locus with a kanamycin resistance cassette using λ red-mediated recombination as previously described (Datsenko and Wanner, 2000). The kanamycin resistance cassette was amplified from pKD4 plasmid template with primers (362/363; Supplementary Tables S1, S2) containing a 50bp homologous extension with upstream and downstream of the coding sequence of crp. The resulting PCR product was then used to transform, by electroporation, MG1655 strain harboring pKD46 plasmid. Deletion of crp was confirmed by diagnostic PCR.

A collection of plasmids containing 6His-tagged genes (minus GFP) from the ASKA library (Kitagawa et al., 2005) was used to transform the TB28 translational fusion (ytfK TL P1+P2) reporter strain by electroporation. Cells were diluted and spread on NA plates supplemented with chloramphenicol, 50 or 200μM IPTG and 80μg/ml X-gal. Petri plates were incubated overnight at 37°C and colonies were screened for their dysregulated expression of ytfK (Supplementary Figure S1), then streaked and their plasmids were sequenced.

The pEG25:6His-crp plasmid expressing crp with an N-terminal 6His tag, under the control of the T5 lac promoter inducible by IPTG was used to transform BL21 (DE3) cells. Several transformants were grown at 37°C overnight in LB medium containing 100μg/ml of ampicillin. Culture was then diluted 50-fold in 2l of the same medium and incubated at 37°C with shaking until OD_600nm 0.6. The expression of the 6His-CRP protein was induced by adding IPTG at a final concentration of 0.5mM for 2h. Bacteria were harvested (9,000 × g, 20min at 4°C) and the pellet was stored at – 80°C. Cells were resuspended and incubated in lysis buffer (50mM Tris-HCl pH 8, 300mM NaCl, 1mM EDTA, 10mM imidazole, 0.5mg/ml lysozyme, 1mM phenylmethylsulfonyl fluoride (PMSF), 20μg/ml DNase and 15mM MgCl₂) for 1.5h at 4°C with gentle shacking and were disrupted using three cycles of French press lysis steps. The cleared lysate was recovered by centrifugation (6,080 × g, 25min) and 6His-CRP protein was purified by ion metal affinity chromatography using a 5ml Nickel (HiTrapHP) column on an AKTA pure 25 (GE healthcare) and desalted using Hiprep 26/10 Desalting column, as previously described (Germain et al., 2019; Supplementary Figure S2A). The last step in 6His-CRP purification was achieved by size-exclusion chromatography (SEC) using a HiLoad 26/600 Superdex 200pg. column pre-equilibrated with 50mM Tris-HCl pH 8, 500mM NaCl, 500mM KCl, 2mM β-mercaptoethanol, and 2% glycerol. The SEC chromatogram is visualized in Supplementary Figure S2B. The purity of the 6His-CRP was verified by SDS-gel electrophoresis (Supplementary Figure S2C). The 6His-CRP protein was stored at −80°C in storage buffer (40mM Tris-HCl pH 8, 238mM NaCl, 22% glycerol, and 1.6mM β-mercaptoethanol).

5' or 3'-Cy5-labeled DNA fragments of the ytfK promoter region were obtained by PCR amplification with appropriate oligonucleotides (Supplementary Table S2). DNA fragments (5nM) were incubated for 15min at 37°C with 6His-CRP (12.5, 25, 65, 130, 195, 300, or 500μM) or not, in 14μl of the binding buffer [10mM Tris-HCl pH 7.5, 50mM NaCl, 5mM MgSO4, 1mM DTT (dithiothreitol), 1mg/ml BSA (bovine serum albumin), 200μM cAMP, 10% glycerol and 12μg/ml poly (dI-dC)]. Ten microliters of the sample were then loaded into a 5% polyacrylamide (37.5/1 [wt/wt] acrylamide-bisacrylamide) gel containing 0.25X TBE (tris-borate EDTA) pH 7.5, 10% glycerol and 200μM cAMP. After migration (10 volts/cm, 1h, 4°C) in migration buffer (TBE 0.25X, 20μM cAMP), the bands were visualized using phosphoImaging (GE Healthcare). The DNA fragment of ytfK promoter region deleted for the putative CRP-binding site was obtained by PCR extension of overlapping DNA fragments using appropriate oligonucleotides (Supplementary Table S2).

Cells expressing transcriptional fusion (ytfK P1+P2) were grown at 37°C in MOPS minimal medium containing 0.025% of glucose and 0.4mM KH₂PO₄. At the indicated time point, OD_600nm was measured and 200μl of the culture was taken at indicated times and incubated with 800μl of Z-Buffer (0.06M Na₂HPO₄ 7H₂O, 0.04M NaH₂PO₄ H₂O, 0.01M KCl, 0.001M MgSO₄ 7H₂O and 0.05M β-mercaptoethanol). Twenty microliters of chloroform was added and cells were vortexed 3×10s followed by incubation at room temperature for 2min. Fifty microliters of the sample was incubated with 150μl of Z-Buffer pre-heated to 28°C. ONPG was added at a final concentration of 0.67mg/ml and the β-galactosidase activity was measured according to the Miller method (Miller, 1992) by following the OD_420nm with TECAN microplate reader.

Bacteria were grown at 37°C overnight in MOPS medium containing 2mM KH₂PO₄ and 0.2% glucose. Cells were then diluted 100-fold in 500μl of fresh MOPS medium containing 0.4mM phosphate (KH₂PO₄) and 0.025% glucose. Cells were continuously and uniformly labeled with 5μl of ³²P (0.37MBq/μl, PerkinElmer) and grown at 37°C with shaking (440rpm). Fifty microliters of samples were taken at the indicated times and 20μl of 21M ice-cold formic acid was added to stop the reaction. Samples were kept on ice for 20min and then stored at −20°C. Cell extracts were recovered by centrifugation (14,000 × g for 60min at 4°C). Five microliters of each sample were spotted into PEI Cellulose TLC (thin layer chromatography) plates (purchased from MercK-Millipore) and resolved with 1.5M KH₂PO₄ pH 3.4. TLC plates were then revealed by PhosphoImaging (GE Healthcare) and analyzed using ImageQuant software (GE Healthcare). The amount of ppGpp was normalized by the amount of pppGpp, ppGpp and GTP for the wild-type strain and the ΔytfK mutant and by the amount of ppGpp and GTP for the ΔrelA and ΔrelAΔytfK mutants.

Overnight cultures of M9 minimum medium containing 2% of glucose were diluted 100-fold in fresh M9 medium containing 0.025% of glucose. Bacteria were cultivated during 120h with shaking at 37°C. Evaporated water was measured and compensated by regularly adding the missing volume of water throughout the culture. Aliquots were taken at indicated hours, cells were serially diluted and plated on NA medium. Plates were then incubated at 37°C overnight and the number of CFU/ml was determined.

ytfK is transcribed as a monocistronic unit (Figure 1A). One promoter, referenced in this study as P1 promoter, has been shown to be recognized by the sigma factor σ^S (Lacour and Landini, 2004), which is proposed to be involved in the induction of ytfK expression at the onset of stationary phase (Salgado et al., 2013). In addition, two pho boxes recognized by PhoB are located upstream the P1 transcriptional start site and are required for induction of ytfK expression in response to phosphate starvation (Baek and Lee, 2006; Yoshida et al., 2011; Figure 1A). Moreover, a second putative P2 promoter is predicted (Lacour and Landini, 2004) upstream the P1 promoter but has not yet been experimentally validated.

In order to dissect the promoter region of ytfK and to analyze the genetic regulation of ytfK expression, we first generated (Figure 1A; see Materials and Methods) chromosomal translational (ytfK TL P1+P2) and several truncated transcriptional lacZ fusions (ytfK P1, ytfK P2, or ytfK P1+P2) to follow promoter activity on X-gal plates. As shown in Figure 1A, ytfK expression is driven from two-independent promoters when cells were spotted on nutrient rich agar, thus confirming the existence of a distal P2 promoter.

To gain further insight on how ytfK is regulated, we used cells harboring the translational fusion (ytfK TL P1+P2) and searched for activators and inhibitors of ytfK expression by overexpressing E. coli genes from the ASKA library pool (Kitagawa et al., 2005) and by using screening assay on X-gal plates for selection of clones with dysregulated lacZ activity (Figure 1B). The ASKA library encompasses almost all E. coli genes cloned into the high-copy-number vector pCA24N, under the control of a P_T5-lac promoter inducible by IPTG (Kitagawa et al., 2005). Out of the approximately 40,000 clones screened, 29 candidates were selected and their plasmids were sequenced (Table 1; Supplementary Figure S1).

We confirmed, after re-cloning into a more suitable physiological plasmid harboring a tight IPTG-inducible P_T5-lac promoter (pEG25), that the ectopic overexpression of cpdA, encoding a cAMP phosphodiesterase, strongly decreases ytfK expression level (Figure 1B). Therefore, this result suggests that the intracellular level of cAMP plays an important role in adjusting ytfK expression level.

In E. coli, cAMP is degraded by CpdA (Imamura et al., 1996) and is synthesized by the adenylate cyclase CyaA (Notley-McRobb et al., 1997). To understand how cAMP plays an essential role in controlling the YtfK level, we first followed lacZ activity of wild-type (TB28 strain) and ΔcyaA cells harboring translational or transcriptional lacZ fusions. We first observed that deletion of the adenylate cyclase cyaA dramatically reduces ytfK expression and this regulation occurs at the transcriptional level (Figure 2A). Moreover, adding 1mM of cAMP in the medium fully restored ytfK expression (Figure 2A). These results confirm that cAMP is necessary to induce ytfK transcription under standard rich conditions.

The CRP is the best-known cAMP target in E. coli (Postma et al., 1993). Once activated by cAMP, the cAMP-CRP complex has a central role in the integration of external signals, such as carbon starvation by regulating, in absence of glucose, the expression of several hundred genes involved in the uptake and catabolism of other carbon sources (Zheng et al., 2004; Shimada et al., 2011). We naturally investigated the involvement of CRP in regulation of ytfK. Interestingly and similarly to what was observed in a ΔcyaA strain, we found that transcriptional expression of ytfK is also strongly impaired in the Δcrp mutant (Figure 2A). Importantly, this phenotype is fully trans-complemented by pEG25:6His-crp (Figure 2B). However, addition of 1mM of cAMP does not restore expression of ytfK in the Δcrp mutant (Figure 2A). Therefore, these results show that the cAMP-CRP complex positively regulates transcription of ytfK.

Glucose is transported into the cell and phosphorylated to glucose-6-phosphate, by the phosphotransferase system (PTS), which is composed of several proteins (i.e., EI, HPr, and EIIA^Glc; Postma et al., 1993; Bettenbrock et al., 2007). The phosphorylation state of the PTS is lower when glucose is available in the medium, whereas once glucose is consumed, phosphorylated PTS proteins accumulate. The phosphorylated EIIA^Glc protein interacts with CyaA and stimulates its activity, thus increasing the intracellular cAMP concentration (Notley-McRobb et al., 1997). In agreement with these data, we observed that the ytfK expression is also highly reduced, when cells are grown in a NA rich medium supplemented with 0.2% of glucose (Figure 2C). In addition, we found that cells expressing transcriptional (ytfK P1) or (ytfK P2) fusions are both submitted to carbon catabolite repression (Figure 2C), suggesting that in the absence of glucose, the cAMP-CRP complex promotes ytfK transcription from both promoters.

Taken together, our results support that the cAMP-CRP complex is required to regulate ytfK transcription in response to glucose availability.

To address whether the cAMP-CRP complex directly regulates ytfK expression, we first produced and purified the CRP protein by two consecutive chromatography steps (affinity and size exclusion; Supplementary Figure S2) and electrophoretic mobility shift assay (EMSA) were performed with 5' or 3'-Cy5-labeled DNA fragments selected from the ytfK promoter region (Figure 3A).

Using DNA covering the entire promoter sequence, a band shift is observed even in the presence of a large amount of poly (dI-dC) competitor (Figures 3A,B). Moreover, this interaction occurs only in the presence of cAMP (Figures 3A,B), as is commonly observed for E. coli CRP homologs (Green et al., 2014). Importantly, no band shift is detected when the intragenic sequence of hofB is used as internal negative control as previously shown (Cameron and Redfield, 2006; Supplementary Figure S3). In addition, the CRP binding to the ytfK promoter sequence appears upon addition of a 2.5-fold molar excess CRP/DNA and the estimated dissociation constant (K_D) is around 25nM range suggesting that CRP has a strong affinity for the ytfK promoter region (Figures 3A,B). Moreover, no additional shifted band of lower mobility was observed even when the CRP/DNA mixing ratio was raised up to 100 (Supplementary Figure S4) suggesting existence of a single-binding site. Finally, and consistent with this observation, we show that CRP specifically interacts with the P2 promoter region and that no binding was observed with the P1 promoter region (Figures 3A,B).

To further determine the accurate localization of the CRP-binding site, we made several truncated deletions in the ytfK P2 promoter region (Figure 3A). Using this approach, we observed that cAMP-CRP complex binds a region located between 248 and 221 base pairs upstream the ytfK start codon (Figures 3A,B). We searched in this short region of interest for DNA sequence similar to the known consensus sequence (5'-TGTGAT-N6-TCACA-3') recognized by CRP (Shimada et al., 2011) and we found as a potential CRP-binding site, the 5'-TGTGATGCCAGTTTGC-3' sequence located at 229bp upstream the start codon (Figure 3A). To validate this prediction, the putative CRP-binding site has been deleted and we observed that CRP no longer binds to the ytfK promoter region (Figures 3A,B). These results suggest that CRP induces ytfK transcription by its binding to the 5'-TGTGATGCCAGTTTGC-3' sequence located in the P2 promoter and that induction of transcription from P1 promoter is likely to be indirect (Figures 3A,B).

Under carbon starvation, SpoT promotes the (p)ppGpp accumulation (Xiao et al., 1991; Gentry and Cashel, 1996; Murray and Bremer, 1996), but little is known about how this environmental change is sensed by bacteria and linked to SpoT-dependent (p)ppGpp accumulation. As mentioned before, YtfK protein level plays an important role in adjusting intracellular (p)ppGpp level in E. coli (Germain et al., 2019). Given that cAMP-CRP complex plays an important role in carbon sources catabolism and positively regulates the ytfK expression, we naturally investigated the role of YtfK in SpoT-dependent (p)ppGpp accumulation under glucose starvation. For that purpose, we first followed expression of ytfK during glucose exhaustion, a condition known to trigger SpoT-dependent (p)ppGpp accumulation (Xiao et al., 1991; Gentry and Cashel, 1996). As shown in Figure 4A and consistent with the observed positive regulatory role of cAMP-CRP complex, ytfK expression gradually increased 30min after growth arrest due to glucose exhaustion. We then compared (p)ppGpp accumulation in the wild-type strain and the ΔytfK mutant in response to glucose exhaustion. However, and as shown in Supplementary Figure S5, both strains seem to have similar kinetics of (p)ppGpp accumulation in response to glucose exhaustion. Importantly, in the absence of amino acids, both RelA and SpoT contribute to (p)ppGpp accumulation during carbon source or diauxic growth transition (Gentry and Cashel, 1996; Fernández-Coll and Cashel, 2018). Therefore, we further decided to address the role of YtfK in response to glucose starvation in cells devoid of RelA. We observed that compared to the wild-type strain, deletion of relA causes earlier growth arrest and a 30min delayed kinetic of (p)ppGpp accumulation (Figures 4B,C and Supplementary Figures S5, S6). Interestingly, we observed that compared to the ΔrelA mutant, the ΔrelAΔytfK mutant exhibits a strong decrease in ppGpp accumulation in response to glucose exhaustion (Figures 4B,C). Hence, YtfK contributes to the full SpoT-dependent (p)ppGpp accumulation in response to carbon starvation.

SpoT-dependent (p)ppGpp accumulation is also known to be essential for survival during prolonged glucose starvation (Nystöm, 1994). Therefore, we addressed the role of YtfK in cell survival during carbon starvation. While survival of the wild-type, ΔrelA and ΔytfK strains are not significantly affected after 5days of carbon starvation (Figure 4D), the ΔrelAΔytfK double mutant rapidly loses its viability after 24h and retains less than 30% of viability after 5days of prolonged carbon starvation (Figure 4D). Taken together, our results show that regulation of ytfK is a fine-tuned regulated process that allows cells to rapidly adapt and survive during long-term carbon starvation by triggering SpoT-dependent (p)ppGpp accumulation.

When E. coli is cultivated in a medium containing various carbon sources, the glucose is preferentially consumed until its exhaustion. (p)ppGpp accumulation rapidly occurs resulting in a transitional growth arrest allowing the establishment of regulatory networks that coordinate the resumption of growth on another carbon source (Harshman and Yamazaki, 1971; Traxler et al., 2006). This phenomenon causes biphasic growth, well known as diauxie (Jacob and Monod, 1961). Importantly a tight control of (p)ppGpp level governs the length of the diauxic lag (Fernández-Coll and Cashel, 2018). We therefore addressed the role of YtfK in diauxic shift. For this purpose, we cultivated the wild-type strain, the simple mutants ΔrelA, ΔytfK and the double-mutant ΔrelAΔytfK in MOPS minimal medium containing a limiting concentration of glucose (0.025%) and an excess of lactose (0.4%). The diauxic lag times were calculated and normalized by generation times on glucose, as previously described (Fernández-Coll and Cashel (2018). We showed that the wild-type strain and the ΔytfK mutant display a similar diauxic lag times of 53min (Figure 5A) and a diauxic lag time/generation time on glucose close to 0.7 (Figure 5B). Moreover and as previously reported (Fernández-Coll and Cashel, 2018), we observed that compared to the wild-type strain, deletion of relA, significantly increases the diauxic lag time (64min; Figure 5A) and the ratio diauxic lag time/generation time on glucose (0.9; Figure 5B). Interestingly, the ΔrelAΔytfK double mutant displays an important extended diauxic lag time (92min; Figure 5A) and the ratio diauxic lag time/generation time on glucose reaches 1.2 (Figure 5B), showing that YtfK also plays an important role in SpoT-dependent (p)ppGpp accumulation during diauxie.

Taken together, these results showed that YtfK plays a key role in sensing carbon starvation though the cAMP-CRP complex and transducing the signal to SpoT to orchestrate (p)ppGpp accumulation and thus coordinate cellular metabolism.