A Second Role for the Second Messenger Cyclic-di-GMP in E. coli: Arresting Cell Growth by Altering Metabolic Flow

ABSTRACT c-di-GMP primarily controls motile to sessile transitions in bacteria. Diguanylate cyclases (DGCs) catalyze the synthesis of c-di-GMP from two GTP molecules. Typically, bacteria encode multiple DGCs that are activated by specific environmental signals. Their catalytic activity is modulated by c-di-GMP binding to autoinhibitory sites (I-sites). YfiN is a conserved inner membrane DGC that lacks these sites. Instead, YfiN activity is directly repressed by periplasmic YfiR, which is inactivated by redox stress. In Escherichia coli, an additional envelope stress causes YfiN to relocate to the mid-cell to inhibit cell division by interacting with the division machinery. Here, we report a third activity for YfiN in E. coli, where cell growth is inhibited without YfiN relocating to the division site. This action of YfiN is only observed when the bacteria are cultured on gluconeogenic carbon sources, and is dependent on absence of the autoinhibitory sites. Restoration of I-site function relieves the growth-arrest phenotype, and disabling this function in a heterologous DGC causes acquisition of this phenotype. Arrested cells are tolerant to a wide range of antibiotics. We show that the likely cause of growth arrest is depletion of cellular GTP from run-away synthesis of c-di-GMP, explaining the dependence of growth arrest on gluconeogenic carbon sources that exhaust more GTP during production of glucose. This is the first report of c-di-GMP-mediated growth arrest by altering metabolic flow.

was important for this phenotype, as observed by lack of growth arrest in the active site mutant pYfiN(GGAAF), where the GGDEF signature motif is changed to GGAAF (Fig. 1C). This mutation did not compromise protein stability (Fig. S1B).
To test whether localization of YfiN to the mid-cell is required for the growth arrest observed in Fig. 1A and B, the location of both cYfiN and pYfiN GFP was monitored in M9M. cYfiN was found dispersed throughout the cell, while pYfiN GFP relocated to the mid-cell in 60% of the cells ( Fig. 2A). Thus, the observed growth arrest by cYfiN is not related to its ability to localize at the division site, while that of pYfiN GFP may include division arrest ( Fig. 1A and B). To investigate if the mid-cell localization of pYfiN GFP in 60% of the cells was related to the higher expression levels from a plasmid, we monitored the behavior of a plasmid encoded R260A mutant of YfiN (R260 is located in the cytoplasmic domain) that does not localize to the mid-cell ( Fig. 2B; see Fig. S1B for protein expression). Like pYfiN GFP , pYfiN(R260A) GFP induction resulted in severe motility inhibition on soft agar plates, indicating robust c-di-GMP production (Fig. 2C). Despite its failure to localize to the mid-cell, the R260A mutant arrested growth similar to WT YfiN (Fig. 2D). We conclude that the inability to grow in M9M is a new phenotype of YfiN, distinct from its ability to arrest cell division by interacting with the divisome at the mid-cell.
E. coli has 12 DGCs, of which five (including YfiN) harbor a transmembrane region. To test if this new phenotype of YfiN is related to its membrane location, and to also control for overexpression artifacts, we cloned the other four transmembrane DGCs (C, E, F, and J) (27) under PBAD control. In contrast to YfiN, none of these four DGCs arrested growth in M9M (Fig. 2E). Except for DgcF, the ectopic expression of DgcC, E, and J inhibited motility, suggesting that they were active as DGCs (Fig. 2F), which is supported by previous work (28)(29)(30). We conclude that early growth arrest in M9M is unique to YfiN.
A clear difference between the early growth arrest observed in this study ( Fig. 1A and B) and the late growth-arrest phenotype of YfiN reported earlier (24) was the growth medium: minimal M9M in the present study, and nutrient-rich LB in the earlier one. When glycerol was substituted with glucose, the most favored carbon source for E. coli, cYfiN no longer inhibited growth in M9M (Fig. 1D, third set of bars from the left). In E. coli growing aerobically, glucose is directly integrated into glycolysis as glucose 6-phosphate (G6P) and consumed through the tricarboxylic acid cycle (TCA), while glycerol, an energy-poor carbon source is incorporated into central metabolism as dihydroxyacetone phosphate (DHAP), a metabolite that can participate in both gluconeogenic and glycolytic processes (31) (Fig. 1E). To explore this finding further, we provided several gluconeogenic and non-gluconeogenic substrates as carbon sources in the M9 medium. For the former, we chose G6P, lactose, glucose, and arabinose, and for the latter, we chose the following: sugar alcohols in addition to glycerol (sorbitol, mannitol), various other sugars (fructose, mannose, maltose), acetate, and TCA cycle intermediates (pyruvate, succinate). There was a clear-cut difference in the ability of YfiN to arrest growth on the two sets of carbon sources, which was patterned after the glycerol and glucose examples, i.e., only gluconeogenic sugars arrested growth. If the differential growth-arrest phenotype of YfiN on these two types of substrates is because of the higher energy produced by one substrate type versus the other, the expectation is that providing both substrates would override the effect of the gluconeogenic substrate. This was found to be the case (Fig. S2A to D). To test whether YfiN-mediated growth arrest is reversible, either glucose or LB was added 2 h post-induction of YfiN. Bacteria resumed growth immediately, suggesting that YfiNmediated growth arrest is fully reversible and that cells are not dead (Fig. S2E).
Growth arrest is correlated with absence of autoinhibitory I sites in YfiN. Given that the DGC activity of YfiN is required to mediate growth arrest (Fig. 1C), we first tested whether the canonical function of c-di-GMP (turning on biofilm pathways/shutting down motility) played a role. This was done in two ways: (i) simultaneous expression of YfiN with YhjH (the most active PDE in E. coli) expected to degrade c-di-GMP, and (ii) disruption of six key genes that control the major biofilm pathways in E. coli (production of cellulose, PGA, colanic acid, Type 1 fimbriae, EPS, etc.) (32). In the first case, i.e., expression of YhjH, the inhibitory effect of YfiN on motility was substantially relieved (Fig. S3A); the reduction in c-di-GMP levels was confirmed using a riboswitch-based biosensor (see Materials and Methods) (33,34) (Fig. S3B). Yet, the growth-arrest function of YfiN was not relieved (Fig. S3C). In the second case, i.e., disruption of multiple biofilm pathways, biofilm levels decreased as expected (Fig. S3D). However, YfiN still arrested growth (Fig. S3E). In summary, the canonical functions of c-di-GMP do not contribute to the observed growth-arrest phenotype of YfiN.
In thinking about what distinguishes YfiN from all the other inner membrane E. coli DGCs previously tested (DgcC, E, F, J), we noted that YfiN lacks the consensus autoinhibitory I site residues (I p and I s ), which bind c-di-GMP to feedback regulate enzyme activity ( Fig. 3A and B) (12,35). We reasoned that lack of autoregulation would lead to continual c-di-GMP production; hence, depletion of cellular GTP. Depletion of GTP as a mechanism of growth arrest is well-established in the (p)pGpp-mediated pathway (36). Imbalance in nucleotide pools is also known to disrupt growth (37,38). To test the GTP depletion idea, we rebuilt the two YfiN I sites (I p and I s ) both sequentially and together. I p was restored by changing GLRH to the consensus RXXD, and I s by changing N280 to R. Only when both I sites were reconstituted, was growth arrest fully reversed (Fig. 3C). These changes did not perturb either the DGC activity of the I-site reconstituted strains as monitored by motility inhibition in soft agar plates (Fig. 3D) or by protein expression levels (I ps mutant; Fig. S1B). We also measured c-di-GMP levels in YfiN and its mutants using the riboswitch-based biosensor. As expected, introduction of I site mutations reduced enzymatic function (I p and I s ), with I ps showing the lowest c-di-GMP levels (Fig. S4). To test whether the I-site is integral to the growth arrest phenotype, we mutated the I p site (RESD ! GESD) in DgcA, a robust heterologous DGC from C. crescentus (12). Expression of WT DgcA in E. coli did not elicit growth arrest in M9M, but expression of the DgcA I p mutant did (Fig. 3E). We note that c-di-GMP levels in DgcA were similar to those when YhjH was co-expressed with YfiN ( Fig. S3B), yet growth arrest was seen in the latter, i.e., YfiN 1 YhjH (Fig. S3C), but not the former, i.e., DgcA (Fig. 3E). Thus, both loss of the growth-arrest phenotype in YfiN and gain of this phenotype by DgcA by merely restoring and inactivating I-site function, respectively, allows us to conclude that unregulated c-di-GMP production in the absence of the I site is responsible for the arrest not by increasing c-di-GMP levels per se, but likely by depleting cellular GTP levels through continuous c-di-GMP synthesis.
YfiN expression depletes intracellular GTP. To test our conjecture that GTP depletion due to unregulated c-di-GMP synthesis is the cause of growth arrest by YfiN, we first compared cellular nucleotide levels with (1) or without (2) induction of c YfiN by metabolomic analysis of a standard panel of 219 intracellular metabolites. Of these, the levels of five ribonucleotides relevant to this study are shown in Fig. 4A and B. c-di-GMP levels were high in the cYfiN1 strain as expected. A consistent decrease in the levels of GTP, ATP, UTP, and CTP was seen in experimental samples compared to the controls. To confirm the drop in GTP concentration upon cYfiN expression, we next used a luciferase-linked GTP assay (GTPase-Glo), which converts GTP into ATP, the latter detected by the light produced by the luciferase reaction (Fig. 4C). We observed a 50% decrease in GTP levels in cYfiN-expressing cells, supporting the GTP metabolomics data. To monitor GTP levels by a third method, as well as to test the proficiency of various YfiN I-site mutants in converting GTP to c-di-GMP, cell lysates expressing pYfiN and its I-site double mutant (I ps ) variant were incubated with [a-32 P] GTP at regular time intervals from 0 to 15 min and the products analyzed by thin-layer chromatography (TLC); only data for the 10-min time point are shown because in this experiment, WT pYfiN lysates depleted almost all of the input GTP at this time, while ;20% of input GTP still remained in the I ps mutant (Fig. 4D). We then compared the time course of GTP consumption by WT pYfiN with that of I-site regulated pDgcA (Fig. 4E). By16 min, YfiN had consumed most of the GTP, while half of the input GTP still remained during DgcA expression. This difference in enzymatic activity may be critical when intracellular GTP is low to begin with in glycerol-fed cells (;2.5-fold lower than glucose) (39). We finally used a fourth method to verify GTP depletion by quantifying the absolute levels of GTP using mass spec as described under Materials and Methods. GTP levels decreased significantly in cells expressing pYfiN (Fig. 4F). Although p YfiN(I ps ) and p DgcA cells showed higher levels of GTP compared to p EmptyVector, the difference was not statistically significant. Thus, four different kinds of measurements show that YfiN expression depletes cellular GTP as expected. The data in Fig. 4B show that YfiN expression impacts all ribonucleotide triphosphate levels as well. An alternative explanation for the depletion of cellular GTP observed in Fig. 4 could be activation of the (p)ppGpp synthesis pathway, during which a diphosphate from ATP is transferred to the 39-OH oxygen of GTP/GDP; (p)ppGpp is the primary regulator of GTP homeostasis in E. coli (39), and deployed in bacteria experiencing various stresses (40,41). In E. coli, RelA and SpoT are the only known enzymes that synthesize as well as degrade (p)ppGpp, which is not made in the absence of both proteins (ppGpp 0 ) (42). To test this alternate explanation, we induced YfiN in DrelA or DrelADspoT(ppGpp 0 ) background, expected to have minimal amounts of (p)ppGpp. Both mutant strains showed growth arrest as early as 2 h after induction of YfiN (Fig. S5), ruling out involvement of (p)ppGpp in YfiNinduced arrest.
To run an independent check of studies concluding that less GTP is available during growth on poor carbon sources such as glycerol and acetate (43), we considered tweaking enzymes contributing to gluconeogenesis. We chose phosphoenolpyruvate carboxy kinase (PCK) for our test, as it mediates an irreversible step in the pathway, converting oxaloacetate to phosphoenolpyruvate, and promotes gluconeogenesis (44) (Fig. S6A). Increasing PCK levels might, therefore, be expected to increase severity of the growth arrest. We monitored growth in M9 mannitol, where cYfiN-mediated growth arrest is not as severe as in M9M (Fig. 1E). In this growth medium, PCK induction resulted in a severity of growth arrest similar to that seen in M9M (Fig. S6B). These data, along with those demonstrating that YfiN-mediated growth arrest can be reversed by the addition of a non-gluconeogenic sugar (Fig. S2E), support our hypothesis that the availability of GTP pools underlies the mechanism of the YfiN-mediated growth arrest.
In summary, we infer from the data in this and prior sections that absence of feed-back control of c-di-GMP synthesis in YfiN, combined with the energy-expensive metabolism of gluconeogenic sugars, deplete cellular GTP to levels unsustainable for growth, which can be rescued by utilizing non-gluconeogenic sugars.
YfiN-arrested cells are tolerant to a broad class of antibiotics. To investigate whether all major cellular processes-cell wall, protein, and DNA synthesis-are arrested by YfiN during growth on M9M, we exposed cells to the antibiotics Ampicillin, Gentamicin, and Ciprofloxacin (45). At the bactericidal concentrations used, c YfiN-expressing cells survived better than control cells in the presence of all three antibiotics when tested as follows. Cells were first grown with or without cYfiN induction on M9M 1 antibiotic plates for 12 h (Fig. 5A, middle). They were then transferred with an inoculation loop to LB plates without added antibiotics for recovery (Fig. 5A, right; Gen/Amp/Cip stamps only point to the antibiotic condition from where cells were taken). The results are shown in Fig. 5B (note again that there are no antibiotics in these plates, as illustrated in Fig. 5A, right). Without cYfiN expression, cells succumbed to all three antibiotics they were initially plated on, as judged by loss of recovery of viable cells (Fig. 5B, left); this was not the case when cYfiN was expressed (Fig. 5B, right). To quantify the number of YfiN-expressing cells that survive antibiotic treatment, a similar assay was performed in M9M, where after 4 h of antibiotic exposure, survival was measured by CFU counts on LB plates ( Fig. 5C and D). The results were similar to those reported in Fig. 5B, i.e., YfiN-expression increases survival to multiple antibiotics. We conclude that YfiN effectively shuts down all major cellular processes in M9M.
Given that the enzymatic function of YfiN is closely related to the growth-arrest phenotype, we reasoned that this function should also be linked to its antibiotic tolerance property. This was tested in two ways using c YfiN: by concomitantly increasing the expression of its periplasmic inhibitor YfiR (pYfiR) (Fig. S7A), as well as by testing the GGDEF active site mutant (Fig. S7B). Both manipulations abrogated the antibiotic (Cip) tolerance phenotype. We conclude that the DGC activity of YfiN mediates both growth arrest and antibiotic tolerance.
Native YfiN delays exit from lag phase during exponential growth. The results presented thus far have relied on ectopic expression of YfiN from non-native promoters, either from the chromosome or from a plasmid. But are they relevant to the native situation? Because we do not know what environmental conditions induce native YfiN, we took advantage of a report that showed a brief burst of YfiN protein levels when overnight cultures are inoculated into fresh media (46). We, therefore, prepared overnight cultures of WT and DyfiN strains in LB, and inoculated them to either M9 glycerol or M9 glucose, recording their growth rates over three 2-h intervals are shown in Fig. 6A and B (DyfiN data are in orange). In M9 glycerol (Fig. 6A), both strains were still in the lag phase in the first growth interval (0 to 2 h). In the next interval (2 to 4 h), the growth rate of WT was unchanged, i.e., it was still in the lag phase, but that of the DyfiN strain increased. In the third interval or exponential phase of growth (4 to 6 h), both strains grew at the same rate. Thus, the WT strain showed a longer lag or delayed transition to exponential phase compared to DyfiN. In M9 glucose, however, both strains grew similarly, and the early lag seen in M9 glycerol for WT was not observed (Fig. 6B). To test whether this observation was strain specific, we repeated the same experiment using WT and isogenic DyfiN derivatives of the uropathogenic E. coli strain CFT073. The longer lag pattern seen for our WT strain (MG1655) in glycerol (Fig. 6A), was also seen for CFT073 compared to its DyfiN counterpart ( Fig. S8A and B).
To test if the antibiotic tolerance seen during growth delay/arrest by ectopic expression of YfiN (Fig. 5) would also be seen under native conditions where growth of the WT strain was delayed (Fig. 6A), we sampled aliquots of the YfN 1/2 strains at 2, 3, and 4 h (which fall within the time windows monitored in Fig. 6A and B), and measured their ability to survive a 4-h exposure to Ampicillin. The survival data for the WT strain inoculated in either glycerol for glucose could be superimposed on growth periods that coincided with the extended lag for the WT strain ( Fig. 6A and B), i.e., WT survived better than DyfiN after antibiotic treatment only during the lag phase and only in M9M (compare Fig. 6C and D). The larger killing effect of Ampicillin on the strains grown on glucose could be due to the reported enhanced killing on this metabolite (47,48).
If the lag was due to a burst of YfiN synthesis as reported (46), this should be reflected in c-di-GMP levels. These were monitored by the riboswitch sensor at the same time points where antibiotic survival was measured, i.e., 2, 3, and 4 h. The data are plotted as a ratio of c-di-GMP in WT versus DyfiN (Fig. S8C). Compared to DyfiN, the WT strain shows higher c-di-GMP levels at 2 and 3 h but not at 4 h, mirroring the antibiotic survival pattern in Fig. 6C. In summary, the growth delay/arrest property of YfiN is seen even when the protein is expressed from its native chromosomal location, and is not an artifact of ectopic expression.

DISCUSSION
c-di-GMP is the most ubiquitous signaling nucleotide in bacteria, with dozens of DGCs involved in its production. These enzymes enable a variety of downstream outputs. In this study, we have discovered a new output for the DGC YfiN, made possible by a particular structural feature of the enzyme that restricts growth in specific nutrient conditions, allowing E. coli to survive through stressors like antibiotics.
YfiN exploits the absence of autoinhibitory I sites to enable a novel mode of survival. I sites (I p to I s ) are common in DGCs, and serve an important role in binding c-di-GMP to feed-back regulate DGC activity (12,20)' thus, controlling the amount of c-di-GMP available to bind to downstream effector proteins. YfiN belongs to a small fraction of DGCs in E. coli (3/12, based on sequence gazing) that do not encode these sites. Absence of this regulatory structural feature appears to be the hallmark of the majority (.90%) of all bacterial YfiN homologs found in the UniProt database. YfiN activity is instead controlled by the periplasmic repressor YfiR, known to be inactivated by redox stress (16,24). Given that YfiN is the most robust DGC in E. coli (24), we suspect other periplasmic stresses that unfold proteins may also activate this enzyme (49). Signals that might activate transcription of yfiN are still unknown. Once activated, one would not expect YfiN to stop c-di-GMP synthesis until the inducing stressors are gone. We show in this study that unregulated c-di-GMP production comes with a metabolic cost. The nature of this cost came to attention when E. coli grown on gluconeogenic carbon sources such as glycerol, mannitol, or sorbitol were observed to arrest cell growth, a phenotype that required the DGC activity of YfiN A Second Role for the Second Messenger c-di-GMP mBio ( Fig. 1), but did not require its relocation to the cell division site at the mid-cell (Fig. 2). That it was the unregulated DGC activity of YfiN that was responsible for growth arrest was established by reconstituting the consensus sequence of both I p and I s sites, which restored growth (Fig. 3C). Conversely, inactivating the I p site of DgcA from C. cresentus, conferred on it the growth arrest phenotype (Fig. 3E). The studies described above were performed under ectopic expression of YfiN alone, in order to mimic conditions where YfiR is nonfunctional. Deleting yfiR was not sufficient to induce growth arrest from the native levels of YfiN. While we do not as yet know which environmental signals might activate yfiN expression, a brief spike in YfiN levels was reported when an overnight culture of E. coli was inoculated into fresh media (46). As discussed below, we have leveraged this finding to show that during this brief spike, native YfiN recapitulates the data from ectopic expression. We imagine that in natural habitats, a combination of periplasmic stress plus the stress of growing on energetically costly substrates, combined perhaps with environmental signals that activate yfiN transcription, might create fertile grounds for growth arrest. That this arrest is reversible, as seen by revival of growth upon adding non-gluconeogenic sugars (Fig. S2E), suggests that E. coli employs YfiN to weather inhospitable conditions. A reversible quiescent state is known to favor adaptive evolution from microbes to humans (50)(51)(52). Our findings likely extend to all YfiN-encoding bacteria given that the majority of the homologs lack the I p site. In those that do have this site (Pseudomonas and Yersinia species), the I s site is absent (Fig. S9). In light of our data showing that both I sites are required to relieve growth arrest (Fig. 3C), feed-back inhibition of YfiN activity is likely inefficient in the bacteria harboring only the I p site (21,22).
YfiN activity during growth on gluconeogenic carbon sources depletes cellular GTP. The clear distinction between gluconeogenic and glycolytic carbon sources in promoting growth arrest by YfiN is striking (Fig. 1D). Gluconeogenesis is the process by which cells synthesize glucose 6-phosphate from non-hexose sugars (53). While eukaryotic cells as well as other bacteria consume both ATP and GTP in this process, E. coli is known to only consume ATP (54). Each cycle of gluconeogenesis requires four ATPs plus two NADHs, resulting in net deficit of nine ATPs. Lower levels of ATP will consequently lower those of GTP, because the g phosphate of ATP is used by nucleoside diphosphate kinase (ndk) to synthesize GTP from GDP (55) (Fig. 7). In glycerol, there is less ATP produced during TCA cycle as well (20 ATPs compared to 36 to 38 ATPs in glucose). Indeed, E. coli grown in M9 glycerol and acetate has been demonstrated to have less intracellular ATP/GTP than in glucose (43). Unregulated consumption of GTP by YfiN will further increase cellular metabolic stress under these conditions. The lower cellular GTP levels we see upon induction of YfiN during growth on M9M are therefore expected (Fig. 4A). Lowering GTP levels even by 2fold (Fig. 4B, C, and F) should be sufficient to arrest growth given in vitro data showing that the transcription rate from a rRNA promoter (rrnB) is highly sensitive to GTP concentrations (56). It was reported earlier that depletion of intracellular GTP by 50% along with decrease in ATP levels by overexpression of RelA homologues, contributes to persister formation by arresting cell growth (57). In eukaryotes, cancer drugs that cause similar drop in GTP levels resulted in cell death in a human cancer cell line (58). In the ecological niches that heterotrophic bacteria inhabit, they obtain carbon from dissolved organic matter. Although E. coli is primarily a commensal of mammals, and to a lesser extent birds, it can be isolated from a variety of host species as well as soil, sediments, and water. We imagine that the particular YfiN function we have uncovered in this study may manifest in certain ecological niches rich in gluconeogenic sugars or in non-carbohydrate substrates not tested in this study. For example, each part of the human gut has different concentrations of carbon sources and types (59).
In summary, we conclude that uncontrolled synthesis of c-di-GMP by YfiN is the proximate cause of depletion of cellular GTP. These nucleotides are essential for the synthesis of the major cellular macromolecules DNA, RNA, and protein, explaining why cells enter growth arrest when using the energetically more costly gluconeogenic carbon sources for synthesis of glucose-6-P.
YfiN contributes to increased tolerance to antibiotics. YfiN is known to inhibit cell division in response to redox and envelope stresses, and to protect E. coli from envelope-disrupting environments when ectopically expressed (24). We show here that ectopic expression of YfiN in M9M also increases tolerance to a broad class of antibiotics such as Ampicillin, Gentamicin, and Ciprofloxacin, which target and disrupt a variety of cellular processes (Fig. 5). This is a result of YfiN's unregulated DGC activity ( Fig. 1C and 3C). Bacteria are continuously exposed to environmental stresses and knowing when to stop proliferating is key to their survival. Our finding that transferring E. coli cells from a stationary culture to M9 glycerol delays growth by extending the lag phase ( Fig. 6A and B), might suggest that transitioning from one nutritional environment to another is one such stress that activates YfiN expression (46). The growth delay upon transitioning from LB to M9M is accompanied by increased c-di-GMP levels (Fig. S8C), and increased tolerance to antibiotics compared to a DyfiN strain (Fig. 6C). This result is similar to antibiotic tolerance observed when cell growth is delayed/arrested by ectopic expression of YfIN. Although it may seem counterintuitive to harbor a gene preventing bacteria from proliferating, persisting in the lag phase may help bacteria in transitioning to the new environment, particularly if the environment is stressful or nutrients are scarce. An increase in the lag phase has been reported as an integral step for developing antibiotic resistance (60,61). It has been suggested that a prolonged lag could buy the bacteria time to diversify adaptive phenotypes (62,63). The importance of the lag phase is not limited to surviving antibiotic stress alone. A transcriptional profiling study of S. enterica showed that genes associated with DNA repair and protein degradation were induced during lag phase, implicating a crucial role for this phase in repairing damaged cellular components (61,64). Bacteria are also known to reorganize their metabolism during the lag phase to achieve optimal growth (64), and two distinct lag phases were observed in E. coli supplied with arabinose (65). Thus, by regulating the duration of the lag phase, we suspect that YfiN not only confers increased tolerance to antibiotics but also other advantages such as optimizing DNA repair and metabolic pathways for growth.
Coda. YfiN seems to have emerged as a DGC that responds to multiple metabolic stresses-redox, envelope, gluconeogenic substrates-and has thus far shown multiple output responses designed to hunker down, make biofilms, and persist.

MATERIALS AND METHODS
Strains, growth conditions, mutagenesis, and plasmid constructions. Strains and plasmids used in this study are listed in Table S1. The WT parent strain for E. coli was MG1655 for all experiments. All strains were grown in M9M (M9 minimal media 1 0.2% glycerol 1 0.2% casamino acid) or in LB broth (10 g/ L tryptone, 5 g/ L yeast extract, 5 g/ L NaCl) unless noted otherwise. When appropriate, the following antibiotics were used: Ampicillin (100 mg/mL), Chloramphenicol (20 mg/mL), Kanamycin (50 mg/mL), and Gentamicin (30 mg/mL). For inducible plasmids, 100 mM isopropyl-b-d-thiogalactopyranoside (IPTG) or 0.02%(wt/vol) L-arabinose were added as indicated in the figures or legends. To monitor the growth of cells, optical density was measured at the wavelength of 600 nm (OD 600 ) HK 533 strain was constructed similarly to HK532 strain (24). Using lambda red recombination (66), yfiR was replaced with PTrc promoter and a kanamycin cassette inserted in an orientation opposite to the direction ( ) of yfi operon transcription, in order to prevent polar effects on yfiN expression.
For cloning of expression plasmids, gene sequences were amplified from the genomic DNA of WT strains by using PCR and introduced into pBAD30 or pBAD33.
To restore or introduce I-sites, specific primers designed for single base pair substitutions were used for PCR amplification using pBAD_YfiN or _DgcA as templates. Following amplification, the original templates were digested with 1 unit of DpnI and PCR products were used for transformation and selection. All constructs were confirmed by DNA sequencing.
Recovery assay. For the experiment shown in Fig. S2E, cYfiN strain was propagated in M9M with IPTG for induction. After 2 h post-induction, 0.2% glucose was added for recovery, followed by measurement of OD 600 every 2 h.
Motility assay. LB soft agar or swim plates were made using 0.3% Fisher agar. A total of 5 mL of an overnight culture was inoculated in the center and plates incubated at 30°C for 8 to 12 h. Swim ring diameter was measured to compare motility across strains.
Biofilm assay. Cells were propagated in M9M in 96-well plates. After 20-h incubation at 37°C, plates were washed twice with water and dried for 2 h at RT (Room Temperature). Then, 125 mL of 0.01% Crystal violet solution (wt/vol) was added to each well and incubated at RT for 15 min. The plate was rewashed three times with water and then 125 mL 30% Acetic acid (vol/vol) were added to each well. After 15 min incubation at RT, solutions in each well were transferred to a new 96-well plate. OD 550 nm was used to measure biofilm formation, and OD 600 for data normalization.
Fluorescence microscopy. Overnight cultures of cells with plasmids encoding fluorescent fusion proteins were diluted 1:100 in fresh M9M or LB medium with antibiotics and grown at 30°C with 0.02% arabinose for 4 h (unless otherwise stated). For imaging cells, a cell suspension (5 mL) was applied to a slide and incubated for 5 min before imaging. Images were acquired using an Olympus BX53 microscope, appropriate filters, and cellSens standard software (version 1.6) from Olympus.
Western blot. Overnight cultures of cells with plasmids encoding YfiN and its mutants (I ps , R260A, GGAAF) were diluted 1:100 in M9M and propagated with 0.02% arabinose for 3 h. Cells were collected and then resuspended in a lysis buffer (12.5 mM Tris pH 6.8, 4% SDS) with a final concentration of 2.5 Â 10^9/mL cells. Upon cell lysis using a heat block (100°C), 1 Â 10^7 cells were loaded into each lane on a SDS-gel. Proteins were then transferred to a nitrocellulose membrane and blocked with 5% non-fat dry milk (NFDM) in Tris-buffered saline with 20% Tween (TBST) for 1 h. The membrane was incubated overnight with a 1:1000 dilution of either GFP antibody (Sigma) or FLAG antibody (Sigma Monoclonal FLAG antibody M2) in TBST with 5% NFDM (wt/vol), and then washed 3 times with TBST every 10 min intervals. This step was followed immediately by an hour incubation with 1:5,000 dilution of Goat a-mouse-HRP (Bio-Rad) in TBST. ECL Select (Amersham; chemiluminescence) was used for a development and the blot image was taken using ChemiDoc (G:BOX).
ImageJ software was used to measure the difference in pixel intensities to compare the difference in the levels of expression.
c-di-GMP biosensor assay. Overnight cultures of cells with plasmids encoding YfiN and riboswitchbased c-di-GMP biosensor were diluted (1:100) and then propagated in M9M media with an addition of 0.02% arabinose for induction. After a 3-h incubation, cells were transferred to a 96-well plate (100 mL/well). Riboswitches (bc3, 4, 5) that are situated upstream of turborfp (encoding a more photo-stable RFP variant) are activated upon binding of c-di-GMP and allow for downstream gene expression, while amcyan, an enhanced CFP variant, is expressed constitutively. Samples were diluted to OD 600 prior to measurement and the ratio of RFP/CFP was used to estimate c-di-GMP concentration/cell. Excitation/Emission wavelengths of 405/488 nm and 553/574 nm were used for amcyan and turborfp, respectively, using a FlexStation3 Plate Reader.
Gene alignment for identification of primary and secondary I-sites. Primary I-site designation for YfiN was based on the identified consensus RXXD, situated five amino acids upstream of the GGDEF catalytic site (12,19,67). Secondary I-site designation for YfiN was predicted by comparing sequences from other DGCs that harbor experimentally identified secondary sites. Clustal Omega, available online (https://www.ebi.ac.uk/), was used for the prediction and alignment; DgcA was excluded from the analysis, because it has two predicted secondary I-sites.
For analysis of YfiN homologues, we used YfiN protein sequence of E. coli MG1655 from the UnitProt database (http://www.uniprot.org; accession number: P46139). Using UniRef50, we acquired 1,974 orthologous sequences with minimum 50% sequence identity. Conservation of primary and secondary I-sites was calculated using the Clustal Omega tool.
Antibiotic survival assay in liquid. An overnight culture of cYfiN strain was inoculated at 1:100 dilution in M9M. Cell cultures were grown for 2 h to reach the OD 600 of 0.1, at which IPTG was added. Then, 2-h postinduction, either Gentamicin (20 mg/mL) or Ciprofloxacin (10 mg/mL) was added. After a 4-h treatment with antibiotics, cell cultures were washed with PBS (phosphate-buffered saline) buffer twice and plated on LB agar for CFU counts. The percent survival was calculated as follows: (final CFU/CFU before antibiotic treatment) Â 100. The results are presented as the average results from at least three biological replicates.
Thin layer chromatography assay. YfiN, YfiN I-site mutants (I p , I s , I ps ), and DgcA were induced from pBAD plasmids with arabinose in M9M for 3 h. Then, cells were washed twice with PBS buffer and lysed with a VCX-750 Vibra-Cell sonicator at 20 kHz (30-s pulse followed by 10-s cool-down for total of five cycles). Cell lysates were incubated with 1 mM GTP/[a-32 P]GTP (0.1 mCi/mL) (PerkinElmer) and incubated at 30°C for various time intervals for comparison pYfiN and pYfiN(I ps ). Reactions were stopped by the addition of 1 volume of 0.5 M EDTA. Radio-labeled products were analyzed by polyethyleneimine-cellulose thin-layer chromatography (TLC; Millipore), by spotting 2 mL samples onto TLC plates. Plates were dried at RT for 5 min, and developed in 1.3 M KH 2 PO 4 (pH 3.2) (12). TLC data were analyzed with GE Typhoon Phosphorimager and ImageJ software by comparing the intensities of each blots on the TLC plate. GTP consumption was estimated based on the input GTP concentration.
GTPase assay. The GTP levels were measured using a GTPase-Glo assay kit (Promega), which employs an ATP-linked luciferase reaction. Total concentration of GTP in cells was calculated based on the standard curve generated using a standard provided by manufacturer, and normalized to OD 600 (OD 600 1.0 = 5 Â 10^8 cells). The final concentration was represented in ng/cell.
Metabolomics. E. coli MG1655 yfiR::kan pTrc99a-cYfiN GFP (HK533) was grown at 30°C for 2 h in 10 mL M9M, with IPTG for induction. Cells were incubated for additional 3 h post-induction (OD 600 reached around 0.4). Pellets were collected and flash frozen. They were shipped on dry ice to the Metabolomics Core Facility at the Mayo Clinic at Rochester, which routinely analyzes a panel of 219 commonly investigated metabolites such as dNTPs, NTPs, carbohydrates, and acids. We provided them with c-di-GMP, and they estimated these concentrations (68). Samples were analyzed by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) at the facility, using a Thermo Fisher Q Exactive mass spectrometer. The analysis is qualitative, and data are provided as areas under elution peaks (A.U., arbitrary unit). Of the three biological replicates (each with two technical replicate) sent to the facility, one was unusable. COVID-19 work protocols during this time (August to November, 2020), prevented us from sending more samples.
Absolute quantification of the targeted nucleotides c-di-GMP and rNTPs was done by (HPLC-MS/MS) at the Metabolomics Core Facility at the University of Texas Medical Branch (UTMB). Cells were grown in M9M except with arabinose for induction. After 3.5-h post-induction, cells were collected and their wet weight was measured for data normalization. Pellets were then flash frozen and mailed to UTMB. Three biological replicates for each set consisting p Empty, p YfiN, p YfiN(I ps ), p DgcA were prepared. The facility used 13 C-labeled standards to obtain a standard curve against which experimental values were derived.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.