The Adaptive Response to Long-Term Nitrogen Starvation in Escherichia coli Requires the Breakdown of Allantoin

Bacteria in their natural environments seldom encounter conditions that support continuous growth. Hence, many bacteria spend the majority of their time in states of little or no growth due to starvation of essential nutrients. To cope with prolonged periods of nutrient starvation, bacteria have evolved several strategies, primarily manifesting themselves through changes in how the information in their genes is accessed. How these coping strategies change over time under nutrient starvation is not well understood, and this knowledge is important not only to broaden our understanding of bacterial cell function but also to potentially find ways to manage harmful bacteria. This study provides insights into how nitrogen-starved Escherichia coli bacteria rely on different genes during long-term nitrogen starvation.

T he process of RNA synthesis, transcription, is often the first decision-making and regulatory point in the flow of genetic information and is catalyzed by the multisubunit enzyme, RNA polymerase (RNAP) (1). The transcriptional responses that establish and maintain specific cellular states are dynamic and are often controlled by many transcription regulatory factors that influence the activity of the RNAP (1). Conditions that sustain constant bacterial growth are seldom found in nature. Whether in a host or the environment, bacteria frequently encounter nutrient deprivation, which can attenuate bacterial growth (2)(3)(4). Hence, the growth-attenuated state is now widely acknowledged as an important physiological state in bacterial cell function. However, although the transcriptional basis of bacterial growth is widely studied, our understanding of the transcriptome of growth-attenuated bacteria remains in its infancy. In fact, several recent studies on different species of bacteria have revealed that growthattenuated bacteria are transcriptionally active. For example, Gefen et al. showed that a batch culture of E. coli cells retains transcriptional activity in the stationary phase of growth over a period of nearly 2 days (5); Stapels et al. showed that nongrowing intracellular Salmonella organisms in host macrophages are transcriptionally active, and this activity is important to shape the host environment to benefit bacterial survival (6); similarly, nonreplicating Mycobacterium tuberculosis cells retained transcriptional activity in the lungs of mice (7). More recently, Gray et al. reported that Bacillus subtilis cells under oligotrophic conditions are transcriptionally active and replicate at an extremely low rate (8). Therefore, understanding the transcriptional landscape in growth-attenuated bacteria is important, as this knowledge could inform and inspire new antibacterial strategies, such as disrupting genetic networks and products that are activated during growth attenuation, which subsequently become essential for survival and/or optimal growth recovery.
In this study, we used nitrogen (N) starvation as a model condition to understand how the transcriptional landscape in nutrient starvation-induced growth-attenuated E. coli affects bacterial survival and the ability to recover growth when nutrients (i.e., N) become replenished. N is an integral nutrient to bacterial cell function, since N is an essential element of most macromolecules in the bacterial cell (proteins, metabolites, nucleic acids, and cell wall components). Therefore, an exponentially growing batch culture of E. coli rapidly attenuates growth upon N run-out (9)(10)(11). The initial adaptive response to N run-out, the N regulation (Ntr) stress response, results in the large-scale and multilevel reprogramming of the transcriptome, involving ϳ40% of all E. coli genes, to allow E. coli to adjust to N starvation (12)(13)(14). Since E. coli perceives N run-out as a decrease in the internal concentration of glutamine, the genes expressed as part of the Ntr response are predominantly involved in the transport and assimilation of nitrogenous compounds into glutamine or glutamate, yielding these catabolically or sparing the requirement for them in cellular processes (12,13). Although our understanding of the initial adaptive transcriptional response to N starvation is well established, how the transcriptome changes during long-term N starvation is far from clear.

RESULTS AND DISCUSSION
Inhibition of transcription in N-starved E. coli compromises survival and growth recovery. To set up an experimental system to study how the transcriptome changes during long-term N starvation, we grew a batch culture of E. coli in a defined minimal growth medium in the presence of a limiting amount (3 mM) of ammonium chloride as the sole N source and excess (22 mM) glucose as the carbon (C) source. As shown in Fig. 1A (left and middle), under these conditions, when ammonium chloride becomes depleted (but sufficient glucose is still available to support growth), bacteria attenuate growth due to N starvation (12). From the onset of N starvation (referred to as NϪ), the number of cells in the population of growth-attenuated E. coli, as measured by the number of CFU in the population, increases at a low rate over the initial 24 h under N starvation (doubling time of ϳ39 h, which is ϳ80-fold less than that during exponential growth when N is abundant; referred to as Nϩ) (Fig. 1A, right, orange area). This growth seems to occur despite the absence of any detectable ammonium chloride in the growth medium (Fig. 1A, left and middle). However, as N starvation ensues beyond 24 h (referred to as N-24), the population ceases growth before the number of viable cells gradually begins to decline after 48 h under N starvation (Fig. 1A, red area). Therefore, the first 24-h period following NϪ, which we refer to as the N-starved adaptive phase (NSAP), seems to be an important adaptive period during long-term N starvation. Transcriptional changes likely occur during NSAP, and we sought to determine how the inhibition of transcription during NSAP affects cell survival and growth recovery when N starvation stress is alleviated. To do this, we treated the bacteria with 150 g/ml of the bacteriostatic antibiotic rifamycin (Rif) at ϳ1 h following NϪ for ϳ23 h. As previously shown by us, under these conditions, 150 g/ml Rif prevents transcription elongation by the RNAP by trapping it at promoters (12). We initially compared the proportion of viable cells in the Rif-treated and untreated population of bacteria at N-24 (i.e., ϳ23 h following Rif treatment at N-1). As shown in Fig. 1B, at N-24, only ϳ50% of cells in the Rif-treated culture were viable compared to the level of the untreated culture. Rif treatment also had a profound effect on the ability of bacteria from N-24 to resume growth compared to that of untreated bacteria. As shown in Fig.  1C and D, the lag time to growth recovery of the Rif-treated bacteria (following CFU correction to ensure an equal number of viable treated and untreated bacteria were used; based on data in Fig. 1B) from N-24 was delayed by ϳ4.8 h compared to that of untreated bacteria. We considered whether this difference in lag time between Riftreated and untreated bacteria was due to the presence of a slower-growing subpopulation(s) of bacteria in the Rif-treated population that resumed growth slowly. To investigate this, we used a method called ScanLag (15), which allows the measurement of the appearance time of individual colonies on N-replete solid growth medium as a function of time following inoculation. We expected that if the Rif-treated population consisted of one or more subpopulations compared to the untreated population, then the appearance of individual colonies of bacteria treated with Rif would be more heterogenous than that of untreated bacteria. As shown in Fig. 1E, no heterogeneity in growth in the populations of Rif-treated and untreated bacteria from N-24 was detected. However, the ScanLag data substantiated the data shown in Fig. 1C and D and demonstrated that Rif-treated cells have an increased lag time to growth recovery compared to that of untreated cells. Once growth resumed, the doubling time of Rif-treated and untreated bacteria did not markedly differ (Fig. 1D). Overall, we conclude that the inhibition of transcription at the onset of NSAP has a detrimental effect on the ability of E. coli to adapt to long-term N starvation in terms of maximizing chances of viability and optimal growth recovery when N becomes available.
The transcriptome E. coli undergoes changes during the NSAP. To study the transcriptional changes that occur during the NSAP, we compared, as indicated in Fig.  2A, the transcriptomes of E. coli during exponential growth (Nϩ), following initial N run-out (NϪ), and at N-24. We defined differentially expressed genes as those with expression levels changed Ն2-fold with a false discovery rate-adjusted P value of Ͻ0.05. As previously reported by us and others, the initial transcriptional response to N starvation involves differential expression of 1,787 genes primarily required for scavenging and catabolism of alternative N sources (13,14,16); thus, we did not investigate these genes further (phase A in Fig. 2A). However, during the NSAP, upon transition from NϪ to N-24 (phase B in Fig. 2A), 451 genes (ϳ10% of E. coli genes) are differentially expressed, of which 105 genes (ϳ2% of E. coli genes) that were not differentially expressed in phase A become differentially expressed in phase B (Fig. 2B). Functional categorization of the genes uniquely differentially expressed in phase B, here called NSAP genes, revealed that they are involved in diverse processes, including transcription, amino acid transport, and metabolism, as well as several genes of unknown function (Fig. 2C). Eleven (abgR, bfd, ileY, leuO, ompN, sibE, tauA, ycaD, yedL, yhdV, and yihO) of the upregulated and nine (evgA, glnA, gltI, hemH, ivy, pfkB, ybjX, yieP, and yjeJ) of the downregulated NSAP genes were known to be regulated by either NtrC (glnA and gltI) or Nac (all other genes listed above) (16,17), the master transcription regulators that activate the initial adaptive response to N starvation. However, the absence of either glnG (which encodes NtrC) or nac did not detectably affect the viability of bacteria following 24 h under N starvation or the ability of 24-h N-starved bacteria to recover growth when inoculated into fresh N-replete minimal growth media (Fig. 2D). Overall, we conclude that the transcriptome of E. coli undergoes changes during the NSAP, and this is consistent with the finding above that 150 g/ml Rif treatment during the NSAP has a detrimental effect on the viability of the E. coli population following 24 h under N starvation and its ability to recover growth when N becomes replenished (Fig. 1B to D). Therefore, it seems that the adaptive transcriptional changes that occur during the NSAP happen subsequently to the initial adaptive changes to N starvation and do not depend on NtrC (and Nac), which coordinates the initial adaptive transcriptional response to N run-out.
Four upregulated NSAP genes are important for optimal growth recovery from long-term N starvation. We next investigated how the upregulated NSAP genes affected the viability and growth recovery of long-term N-starved E. coli. Of the 51 upregulated NSAP genes, 44 are available in the Keio single-gene knockout collection (18). Thus, we determined how the absence of any of the upregulated NSAP genes affected growth recovery following 24 under N starvation. To do this, we grew the library of upregulated NSAP gene mutant bacteria in minimal growth media containing excess (10 mM) ammonium chloride as the sole N source and measured cell density after 16 h of growth. The growth of the library of upregulated NSAP gene mutant strains was done in a plate reader incubator, which allowed continuous monitoring of the growth of each mutant strain simultaneously to enable better side-by-side comparison. The results shown in Fig. 3A indicate that the absence of any of the NSAP genes did not markedly affect growth under N-replete conditions. However, we cannot exclude the possibility that the relative dynamics of growth of the individual mutant strains differed from that of the wild-type (WT) strain over the 16-h period. We note that bacteria in which the tryptophan biosynthesis genes are deleted, that lack the ␥ subunit of formate dehydrogenase-N (fdnI), or that lack the ferric rhodotorulic acid outer membrane transporter (fhuE) are unable to grow in our minimal growth media, most likely due to auxotrophy (19). We next reinoculated bacteria that grew under N-replete conditions into fresh minimal growth medium containing limiting, i.e., 3 mM, ammonium chloride and determined how the absence of the upregulated NSAP genes affected viability at N-24, i.e., following 24 h under N starvation. The results depicted in Fig. 3B show that the viabilities of the upregulated NSAP gene mutants were similar to that of the WT strain. When the 24-h N-starved mutant bacteria were inoculated into fresh N-replete minimal growth medium, the dynamics of growth recovery for the majority of the mutant strains were similar to those of the WT strain (Fig. 3C). However, ΔallB, ΔleuO, ΔynfP, and ΔybcJ mutant strains displayed noticeably compromised growth recovery properties compared to that of the WT strain (Fig. 3C). These genes encode an allantoinase (allB), a DNA-binding transcription dual regulator (leuO), a small, possibly fragmented, protein of unknown function (ynfP), and a putative RNA-binding protein (ybcJ) (19). As shown in Fig. 3D, although the doubling times of ΔallB, ΔleuO, ΔynfP, and ΔybcJ mutant strains were similar to that of the WT strain, their lag time to growth recovery was increased by ϳ1.5 to 4 h compared to that of the WT strain. Thus, it seems that the transcription of allB, leuO, ynfP, and ybcJ during the NSAP is required for optimal growth recovery from long-term N starvation. As the ΔallB mutant strain displayed the most pronounced compromised ability to recover growth compared to that of the WT strain (Fig. 3C), we decided to focus on the role of allB during the NSAP and recovery from long-term N starvation.
Degradation of allantoin during the NSAP is required for optimal growth recovery from long-term N starvation. To discount the possibility that the difference in lag time during growth recovery between the WT and ΔallB bacteria was not due to the presence of a subpopulation of slow-growing ΔallB bacteria, we measured the lag time distribution of the appearance of individual colonies during growth recovery using ScanLag (15), as described for Fig. 1E. As shown in Fig. 4A, we observed the expected lag time in growth recovery for the ΔallB bacteria, but no evidence of a slow-growing subpopulation was seen. Further, we did not detect a lag in growth recovery of WT and ΔallB bacteria sampled from NϪ when reinoculated into N-replete fresh growth medium (Fig. 4B). This result underscores the view that allB has a specific role in E. coli during the NSAP. To establish that the increased lag time in growth recovery of ΔallB bacteria was directly due to the absence of allB, we complemented the ΔallB bacteria with plasmid-borne allB. As shown in Fig. 4C and D, the lag time in growth recovery of ΔallB bacteria reverted to that of WT bacteria when allB was supplied via a plasmid. The product of allB is allantoinase: a tetrameric enzyme that catalyzes the hydrolytic conversion of (S)-(ϩ)-allantoin to allantoate. A previous study (20) reported that substitutions at invariant amino acid positions N94 and S317 (in both cases to D and V), which are located proximal to the active center, rendered allantoinase inactive (Fig. 4E). As shown in Fig. 4C and D, in the context of the complementation experiments,  plasmid-borne allB-N94V and allB-S317V failed to fully decrease the lag time in growth recovery of ΔallB bacteria, as seen with the WT version of allB. In contrast, plasmidborne allB-N94D and, to a greater extent, allB-S317D reverted the lag time in growth recovery of ΔallB bacteria, as seen with the WT version of allB. Although we are unable to explain why specifically substitution by D but not V at N94 or S317 renders allantoinase catalytically inactive under our experimental conditions, it seems that the degradation of allantoin occurs at some point during the NSAP in E. coli. This is conceivable because, as shown in Fig. 4F, the conversion of allantoate (the product of allantoinase) to (S)-ureidoglycine by allantoate amidohydrolase (AllC) and the subsequent conversion of (S)-ureidoglycine to (S)-ureidoglycolate by (S)-ureidoglycine aminohydrolase (AllE) produces ammonium, which may be needed for the maintenance of cellular functions required for optimal growth recovery from long-term N starvation. Intriguingly, compared to allB, transcripts corresponding to allC and allE were only moderately upregulated (ϳ1.2-to 1.5-fold) in phases A and B in Fig. 2A and, thus, did not fall into the category of NSAP genes. However, we observed that 24-h N-starved ΔallC and ΔallE bacteria displayed an increased lag time to growth recovery similar to that of the ΔallB bacteria ( Fig. 4G and H) when grown in fresh N-replete growth media. This suggests that the activity of the genes of the allantoin degradation pathway has a role in the adaptive response to long-term N starvation in E. coli. As allantoin is a product of the purine degradation pathway in N-starved E. coli (21), we considered whether the requirement for allB is a consequence of purine degradation in E. coli during the NSAP to sustain cellular functions via the allantoin degradation pathway (as shown in Fig. 4F). To indirectly determine whether purines are degraded in E. coli during the NSAP, we measured the integrity of total RNA as a surrogate for purine degradation in E. coli as a function of time in the NSAP by chip-based capillary electrophoresis (see Materials and Methods). As shown in Fig. 4I, strikingly, we observed a gradual degradation of 16S and 23S rRNA as the NSAP ensued. This observation further underscores the view that the RNA landscape and turnover during the NSAP is plastic and consistent with the view that the transcriptome of E. coli undergoes changes during long-term N starvation. Collectively, the results support the view that allB contributes to optimal growth recovery from long-term N starvation by the degradation of allantoin during the NSAP in E. coli, which likely accumulates as a result of purine degradation.
Conclusions. N is central to all metabolic processes in the bacterial cell. As such, bacteria can utilize a wide range of N-containing compounds as sole sources of cellular N. Upon N run-out, E. coli cells activate the transcription of genes associated with N assimilatory metabolic pathways necessary for the utilization of N from the extracellular environment and catabolic biosynthetic pathways for intracellular assimilation of N. We have presented evidence that the growth of a population of N-starved E. coli does not completely cease upon N run-out and that E. coli cells continue to divide over a period of up to 24 h (the NSAP) following entry into N starvation. It seems that the NSAP represents an important adaptive phase in E. coli experiencing long-term N starvation to maximize chances of viability and optimize growth recovery should N become replenished. We have also presented evidence that the adaptive transcriptional response to N starvation in E. coli occurs in at least two distinct temporally coordinated phases, where the NtrC-dependent initial response to N starvation is followed by, to a large extent, an NtrC-independent transcriptional response. We have uncovered that an aspect of the adaptive transcriptional response, which occurs in the NSAP subsequent to the initial adaptive transcriptional response in E. coli under long-term N starvation, involves the degradation of allantoin. As ammonium is a by-product of the degradation of allantoin, we propose that allantoin degradation allows the assimilation of ammonium into glutamine and glutamate to maintain cellular processes during the NSAP to maximize chances of viability and optimize growth recovery when N becomes replenished. Previous studies have shown that allantoin can only be used as the sole N source by E. coli grown under anaerobic conditions (22,23). Consistent with these results, we were unable to grow ΔallB and WT bacteria under our conditions when ammonium chloride (the sole N source used in our experiments) was replaced by allantoin in both the initial growth and recovery growth experiments (data not shown). Therefore, as the end product of allantoin degradation under aerobic conditions is 3-phosphoglycerate (in addition to the by-product, ammonium), which is used for energy production, we propose that the breakdown of allantoin could serve as some sort of metabolic threshold to allow optimal growth recovery of E. coli from long-term N starvation under aerobic conditions when N becomes replenished. Previous work by us (12,14) and others (13) has also shown that genes associated with the degradation of pyrimidine (e.g., the rut operon) become upregulated at the onset of N starvation and remain upregulated throughout the NSAP (14). As allantoin is a product of the purine degradation pathway in N-starved E. coli (21), collectively, these observations suggest that the degradation of nucleotides is an essential component of the adaptive response to N starvation in E. coli. Although future work will focus on deciphering the metabolic and nucleotide landscape to unravel in depth the adaptive responses that define the NSAP in E. coli, this work has underscored the importance of studying the transcriptional changes that occur in long-term nutrient-starved bacteria to further our understanding of bacterial stress physiology.

MATERIALS AND METHODS
Bacterial strains and plasmids, growth conditions, and viability measurements. WT and mutant E. coli BW25113 strains were used in this study (18). Bacteria were grown as described by Figueira (16). Where relevant, rifampin was added at a final concentration of 150 g/ml at N-1 or Nϩ as stated, and L-arabinose was added to complementation strains containing the pBAD18 plasmid at 0.2% (wt/vol) for induction of gene expression in growth cultures. Ampicillin was added at a final concentration of 100 g/ml to all cultures containing derivatives of the pBAD18 plasmid. Measurement of growth was taken by measuring the optical density at 600 nm (OD 600 ) of bacterial cultures over time to generate growth curves. Ammonium and glucose concentrations in the media were determined using the Aquaquant ammonium quantification kit and glucose assay kit, respectively (both by Merck Millipore, UK), according to the manufacturer's instructions. To determine population viability, the number of viable cells was calculated by conducting serial dilutions of cultures at designated time points, as shown in relevant figures, grown overnight on Luria-Bertani (LB) agar plates and measuring the number of CFU per milliliter. The pBAD18-allB plasmid was made by traditional restriction enzyme cloning to incorporate the ribosomal binding site of pET33b(ϩ) with the allB gene to allow arabinose-inducible expression of the AllB protein. The N94D, N94V, S317D, and S317V mutations in allB were done by standard site-directed mutagenesis using pBAD18-allB as the template.
ScanLag. Bacteria were grown in 3 mM Gutnick minimal medium to N-24, washed twice in 1 ml PBS, and diluted in PBS. One hundred microliters of diluted culture was spread on an LB agar plate and incubated at 33°C for 48 h with pictures of plates taken every 20 min. Analysis of colony growth rate and appearance time was modified from Levin-Reisman et al. (25) by Miles Priestman and is available at https://github.com/mountainpenguin/NQBMatlab. RNA-seq. The RNA sequencing (RNA-seq) data set was previously generated and analyzed by Switzer et al. (14). The analysis of differential gene expression during NSAP as well as the statistical analyses were also conducted in the same ways as those described previously. The sequencing data can be accessed under accession number E-MTAB-6491 in the ArrayExpress database at EMBL-EBI (www.ebi.ack.uk/arrayexpress).
Total RNA analysis. Cultures were grown as described above and sampled at the following time points: at the onset of N starvation (NϪ) and 3 h (N-3), 6 h (N-6), 9 h (N-9), 12 h (N-12), 18 h (N-18), and 24 h (N-24) under N starvation. At least three biological replicates were taken at each time point. RNA was acquired and stabilized at specified time points using Qiagen RNA Protect reagent (76526; Qiagen). RNA extraction was carried out using the PureLink RNA minikit (12183025; Invitrogen) and DNase set (12185010; Invitrogen) per the manufacturer's protocol. Electropherograms of total RNA integrity were generated using an Agilent Bioanalyzer 2100 instrument (G2939BA; Agilent) and RNA 6000 Nano kit (5067-1511; Agilent) per the manufacturer's protocol.
Statistical analysis and growth calculations. Unless otherwise stated above, all growth curves and CFU data show the mean average from three independent experiments, and variation was calculated by the standard errors of the means (SEM). The Student's t test or one-way analysis of variance was used to determine statistical significance if the probability (P) value was Ͻ0.05. Lag and doubling times were calculated by first generating a straight-line graph of the exponential growth section of each growth curve. Lag time was defined as the intersection of the straight-line graph with the x axis, while the doubling time was determined during exponential growth (26).