The physiology of growth arrest: uniting molecular and environmental microbiology

Abstract

Most bacteria spend the majority of their time in prolonged states of very low metabolic activity and little or no growth, in which electron donors, electron acceptors and/or nutrients are limited, but cells are poised to undergo rapid division cycles when resources become available. These non-growing states are far less studied than other growth states, which leaves many questions regarding basic bacterial physiology unanswered. In this Review, we discuss findings from a small but diverse set of systems that have been used to investigate how growth-arrested bacteria adjust metabolism, regulate transcription and translation, and maintain their chromosomes. We highlight major questions that remain to be addressed, and suggest that progress in answering them will be aided by recent methodological advances and by dialectic between environmental and molecular microbiology perspectives.

Much of our knowledge of the molecular machinery that is responsible for energy generation, gene expression and DNA replication comes from studying fast-growing model organisms, such as Escherichia coli, during exponential phase. in nutritionally complete medium. Under these conditions, a single E. coli cell would grow to a population with the mass of the Earth within 2 days. Clearly, this does not occur, but the discrepancy between potential and actual growth underscores that the estimated 5 × 10³⁰ bacteria that are on our planet spend the vast majority of their time in energy-limited states and not dividing. Although environmental microbiologists have long appreciated the importance of extremely energy-limited states^1,2, their focus has primarily been on exploring theoretical and empirical energetic limits of diverse metabolisms^3,4. With a few notable exceptions^5–7, molecular microbiologists have largely avoided the study of growth-arrested cells. Consequently, relatively little is known about the mechanisms that underpin the dominant modes of bacterial existence; this gap in our knowledge hinders our understanding of the roles that bacterial communities have in driving global biogeochemical cycles and influencing plant and animal health.

The dearth of information is due, in part, to the challenges in defining, reproducing and measuring non-growing states of bacteria in the laboratory. Microbiologists have traditionally relied on population-level measurements of activity to draw conclusions about molecular mechanisms — an approach that benefits from high levels of biochemical activity and homogeneous populations. The few existing studies of non-growing states have focused on conditions that cause population growth arrest through the severe limitation of at least one basic resource (BOX 1). These studies have provided most of our insights into these growth states (referred to as ‘long-term stationary phase’ (REF 6), ‘continuous activity stationary phase’ (REF 8) or ‘starvation–survival’ (REF. 2)) and have shown that growth-arrested cells synthesize proteins at rates that can be orders of magnitude slower than in exponential phase but stay viable for several days to years, and are usually able to rapidly resume growth when nutrients become available. These populations are neither highly active nor likely to be homogeneous (BOX 2), potentially including subpopulations that undergo infrequent division cycles, but we assume, based on their static population numbers, that they represent the best available proxies for non-growing states. A major area for future research, as measurement sensitivity and selectivity continue to improve (BOX 3), will be to better understand how these populations are composed of distinct cellular states.

Box 1 |. Non-growing states.

Several approaches have been used to study bacterial cells in non-growing states, and for comparison to each other and to other familiar growth states, non-growing states of interest can be shown as portions of representative, idealized growth curves (see the figure, solid lines).

Stationary phase

Perhaps the simplest and most intuitive approach is to grow a batch culture until carbon and/or nitrogen sources are depleted. In 1971, this approach was used in Escherichia coli to show that, after an initial drop at the entry to stationary phase, the adenylate energy charge (AEC) decreased very slowly for days (see the figure, region a) before it began to decrease further and cells began to lose viability⁵³. These results are consistent with a much more recent study that used a microfluidic device to track protein synthesis by single cells during and after the transition into stationary phase, which found that after an initial large decrease, protein synthesis rates remained fairly stable during several days of starvation. This insight into the population heterogeneity of a growth-arrested state showed that, at least for the initial period of the stationary phase that lasted several days, non-growing status and protein synthesis rates are reasonably uniform across the majority of cells⁸.

Long-term stationary phase and the GASP phenotype

Following this initial period, many species exhibit a large decrease in viable cell counts. However, at least in batch cultures of E. coli, the cells that survive can use nutrients that are released from dead cells, such that the population enters into a state of balanced cell death and growth that can continue for years without any new input of nutrients (see the figure, region b). During this long-term stationary phase, mutants arise that take over the population, a phenomenon that is referred to as ‘growth advantage in stationary phase’ (GASP). The heterogeneity that underlies GASP poses some challenges for study (BOX 2). Studies of GASP have focused on genetic changes, which have shown that the functions that are under the strongest selective pressure include amino acid uptake and catabolism and modulation of RpoS activity22,89. In addition, the selective sweeps of beneficial mutations that are observed in GASP populations have confirmed that cell division is occurring, although this must be infrequent because the population numbers remain static. We consider the ‘long-term stationary phase’ condition in which GASP phenotypes are observed a reasonable proxy for non-growth, albeit a different one than the ‘continuous activity stationary phase’ (see the figure, regions b and a, respectively).

Oxygen limitation

In the nutrient-rich environments inhabited by Pseudomonas aeruginosa, Mycobacterium tuberculosis and other bacteria that are responsible for causing chronic infections, oxygen is much more likely than nutrients to become limiting, which can lead to growth arrest as oxygen is preferred for oxidative phosphorylation. For P. aeruginosa, the cell density of a culture that is incubated in an anaerobic chamber with either pyruvate or arginine as a carbon source remains relatively constant over a period of 1–2 weeks before cells begin to lose viability, and some cells survive anaerobic conditions for periods of at least several months^33,34 (see the figure, region c). For M. tuberculosis, the slow depletion of oxygen from a stirred, closed culture flask, which models the latent stage of infection, induces a long-term non-replicating survival state in which the cell density remains stable for at least several months³⁰ (see the figure, region d).

VBNC

It remains possible that some cells that seem to lose viability during growth-arrested states are actually entering a ‘viable-but-nonculturable’ (VBNC) state, which is characterized by an inability to form colonies on rich media but the continued maintenance of the proton motive force (PMF). First observed in Vibrio cholerae (see the figure, region e) and E. coli⁹², but subsequently in phylogenetically diverse bacteria⁹³, the VBNC state can often be induced by stresses that overlap with stationary phase, such as prolonged starvation or high osmolarity, which suggests that it may be part of the ‘growth-arrested state’ continuum⁹⁴. The widespread existence of VBNC states in the environment, and a lack of understanding of the triggers for emerging from them, has been proposed as one of the reasons why many bacteria remain uncultured in the laboratory⁹⁵.

Persisters

A different state in which cells remain growth-arrested despite an abundance of nutrients is the persister state that is associated with antibiotic tolerance (see the figure, region f). Persisters exist in a growth-arrested state in exponential phase batch cultures, in which most of the population is dividing at the maximum rate, which results in a substantial heterogeneity (BOX 2) in replication rates⁹⁶. Entry into the persister state seems to be stochastic and reversible, and the isolation of mutants that increase the rate of formation of persisters has suggested mechanisms that may be responsible, such as toxin–antitoxin systems that target the translational machinery and the stringent response^97,98. Thus, the mechanisms by which cells impose growth arrest in the context of high nutrient availability might share components with the mechanisms by which cells enter growth arrest in response to nutrient deprivation^97,98.

Box 2 |. Heterogeneity and growth arrest.

Population heterogeneity is clearly present in many of the contexts in which growth arrest has been investigated, including the ‘growth advantage in stationary phase’ (GASP) phenotype, persisters and biofilms. Phenotypic heterogeneity in cultures that exhibit a GASP phenotype⁶ probably helps determine which individuals survive the initial loss of viability; much later, genetic heterogeneity also develops in these populations⁸⁹. Perhaps the best-studied example of phenotypic heterogeneity in growth rates is that of persisters. Persisters are not mutants; indeed, isolated persisters give rise to a new population of exponentially growing cells that again has a tiny minority of persisters⁹⁹. Finally, heterogeneity in growth rates is clearly observable and important in biofilms. Laser-capture microdissection of cross-section samples of Pseudomonas aeruginosa biofilms^47,100 has shown that, although cells at the edges of biofilms, which have the best access to nutrients and/or oxygen, have gene expression signatures that are similar to cells that are entering stationary phase, cells at the interior of biofilms show much lower rates of metabolic activity and different gene expression profiles, which is consistent with growth arrest (see the figure, part a). Different regions of these biofilms also exhibit different antibiotic sensitivities and rates of respiration^47,100, which provides further evidence that cells in the different regions are in different growth states. Recently, it has been shown using time-lapse microscopy that Bacillussubtilis biofilms can also exhibit temporal heterogeneity in growth states, at least under some defined growth conditions, with cells at the periphery cycling through pulses of growth and non-growth; the authors propose that the non-growing periods provide the cells in the interior with better access to nutrients¹⁰¹.

Although population heterogeneity has not been addressed in most studies of gene expression during growth arrest, it seems likely that cells are heterogeneous to some extent in their molecular physiology, given that prolonged growth arrest often leads to a loss of viability in some, but not all, cells. Substantial heterogeneity can pose serious problems for the interpretation of measurements at the population level. Even the question of whether cells are really growth arrested becomes difficult to answer. In the context of emergence of the GASP phenotype, the number of viable cells in the culture remains stable, but cell division is clearly occurring, at a low rate and/or among a subset of cells, as shown by the observation of a selective sweep of beneficial mutations through the population. In this heterogeneous population, the population numbers are static because growth is balanced by death. The same might be true in other examples cited in this Review. In terms of understanding molecular mechanisms, this distinction is very important: in the absence of heterogeneity, the mechanisms proposed must account for very low levels of average activity, occurring in all cells of the population. If the populations that are under growth arrest are instead nearly always heterogeneous, then it is possible that the cells accounting for all of the observed activity are working at rates much closer to those observed for cells in exponential phase, but the population average is substantially lowered by the majority of nearly dormant cells. The increasing availability of methods for measuring single cells (BOX 3) should help to address this question and enable further exploration of the mechanisms that underlie heterogeneity. For example, nanoscale secondary ion mass spectrometry (NanoSIMS) was used to investigate heterogeneity of microbial metabolic activity, as measured by the incorporation of ¹⁵N, in sputum collected from a patient with cystic fibrosis and incubated in the presence of ¹⁵N-labelled ammonium, which showed that metabolic activity is heterogeneous and not well correlated with the abundance of 16S rRNA in matched cells. DAPI, 4’,6-diamidino-2-phenylindole; FISH, fluorescence in situ hybridization. Fluorescence microscopy and NanoSIMS images in part b courtesy of S. Kopf, University of Colorado Boulder, USA.

Box 3 |. New tools for studying non-growing states.

The key technical challenges that are associated with studying growth arrest are the low rates of metabolic activity; the low intensity of ‘signals’ of newly synthesized macromolecules compared with the ‘noise’ of the pre-existing background; the contribution of degradation to changes in steady state levels, which could be comparable to that of new synthesis; the probable high level of heterogeneity in populations; and the difficulty in maintaining constant or known conditions in laboratory cultures for long periods of time. Many advances in the development, modification and accessibility of methods for measuring the levels and activity rates of macromolecules, both in populations and at the single-cell level, are improving the feasibility of accurately observing non-growing states.

Sensitive and selective population measurements

Retentostat cultivation methods have been used by environmental microbiologists to overcome the challenge of maintaining constant conditions that arrest growth¹⁰². Similarly to chemostats, retentostats maintain a flow of growth medium at a defined rate, but unlike chemostats, they retain all biomass, thus enabling the study of very slow or non-growing states. Such devices could be used to culture organisms for the study of the molecular physiology of minimum energy states.

Bio-orthogonal non-canonical amino acid tagging (BONCAT) uses click chemistry to attach biotin or fluorescent tags to amino acid analogues that are incorporated into proteins¹⁰³. For example, pulse labelling of l-azidohomoalanine, which cells can naturally incorporate in place of methionine, was used to study anaerobic survival by Pseudomonas aeruginosa⁴⁸. Incorporation of l-azidonorleucine, which requires a mutated tRNA synthetase, can be used to measure the proteomes of a subpopulation of cells by placing the mutant synthetase under the control of a promoter that is specifically active in that subpopulation.

Next-generation sequencing has numerous applications (see the figure, part a) that, in many cases, could be straightforwardly applied to non-growing cells, owing to the generally low quantity of starting material that is required. Direct sequencing of genomes has been used to detect mutations that arise during chronic infection¹⁰⁴, whereas transposon sequencing (Tn-seq) determines mutations in a transposon mutagenesis screen that affect fitness during exposure to a stress¹⁰⁵, and RNA sequencing (RNA-seq) has been widely used to study transcriptomes under various conditions¹⁰⁶. Other applications include chromatin immunoprecipitation followed by sequencing (ChIP-seq), in which regions of the genome bound by a factor of interest are captured by immunoprecipitation¹⁰⁷; ribosome profiling, which uses nuclease degradation to remove sequences of mRNAs that are not physically protected by ribosomes¹⁰⁸; native elongating transcript sequencing (NET-seq), in which nascent elongating transcripts are captured by immunoprecipitation of transcribing RNA polymerases¹⁰⁹; and RNA-protein immunoprecipitation in tandem sequencing (RIPiT-seq), which reveals RNA-protein interactions through the sequencing of mRNAs that are captured by immunoprecipitation of RNA-binding proteins¹¹⁰.

Population isotope labelling can be used in microbiology to trace the metabolic incorporation of substrates that are composed of heavy isotopes with low natural abundance (such as ¹⁵N ammonium, ¹³C glucose or ²H water). Such an approach was recently combined with the extreme sensitivity of gas chromatography–pyrolysis–isotope ratio mass spectrometry (GC–pyrolysis-IRMS) to measure growth rates of Staphylococcus aureus in cystic fibrosis sputum¹¹¹ (see the figure, part Ba). Calculating isotope enrichment in labelled samples provides a measure of biosynthetic turnover that is independent of changes to total biomass, which makes this method applicable to a wide range of growth states. Stable isotope probing (SIP) can also be used to separate DNA of metabolically active cells from that of inactive cells in a mixed population, as metabolic incorporation of a stable heavy carbon isotope can lead to sufficient changes in DNA density for separation on a caesium chloride gradient¹¹² (see the figure, part Bb).

Single-cell measurements

Microfluidic devices enable the isolation and cultivation of single or small numbers of cells for study by microscopy, sequencing or other techniques¹¹³.

Microscopy is inherently well suited to observations of single cells, and can be combined with fluorescent markers that might be relevant in the study of growth-arrested states, such as engineered protein ‘biosensors’ of redox state or ATP levels^114,115 and fluorescent dyes that detect membrane permeability and defects in membrane polarization, or even distinguish between the two³².

Single-cell isotope methods that use radioactive isotopes have, for many years, been used by environmental microbiologists in conjunction with fluorescence in situ hybridization (FISH) to identify individual metabolically active cells in mixed populations¹¹⁶. New methods include nanoscale secondary ion mass spectrometry (NanoSIMS), which provides sufficiently high resolution to accurately investigate bacterial metabolism on a cellular, and even subcellular, scale. NanoSIMS uses stable, heavy isotope substrates, which, following metabolic incorporation and sample fixation, yield distinct secondary ions on bombardment with a caesium ion beam. A sensitive detector can determine the ratio of heavy to light isotopes present in the sample with high spatial resolution and this ratio gives insight into single-cell rates of metabolism¹¹¹ (see the figure, part Bc). Raman spectroscopy, in which the wavelengths of photons that travel through a sample are shifted by characteristic amounts as they interact with different chemical bonds, is also sensitive enough to distinguish between heavy and light stable isotopes and can be used to make rapid, high-throughput measurements of substrate uptake for individual cells¹¹⁷.

Our focus on non-growing states of relatively fast-growing bacteria purposefully excludes some related states and other survival strategies. We view the responses to nutritional downshift that are part of the transition to stationary phase, which have been well studied and reviewed elsewhere (see REFS 9,10), as distinct from the strategies that are used later in stationary phase, and we point out these differences throughout the review (see also BOX 1). Another state that we do not address is the endospore — a mode of survival that uses almost no energy for potentially thousands of years — which has been well reviewed elsewhere (see REF. 11). Finally, inherent slow growth in bacteria with low-energy core metabolisms is outside the scope of this review. These organisms, which have doubling times of months or much longer, might be viewed as ‘non-growing’ for substantial periods of time even when doubling at their maximum rates (reviewed in REF 4). How basic molecular mechanisms function in such contexts is fascinating, and may be related to the growth-arrested strategies of otherwise fast-growing bacteria, but the challenges of direct study remain extreme.

In this Review, we focus on general insights from current research and highlight remaining questions for three fundamental cellular challenges that are encountered during growth arrest. First, how are cellular energy stores maintained and managed during starvation? Second, how is gene expression regulated under extreme limitation for nucleotide and amino acid substrates? Third, how is chromosomal DNA protected in a way that allows occasional replication and low levels of transcription?

Metabolism: maintaining energy supply

The primary goal of growth-arrested organisms is to maintain the supply of the energy and biosynthetic precursors that are required to maintain essential macromolecular components of the cell, sustain active regulatory mechanisms for sensing and responding to the environment, and, perhaps most importantly, preserve the electrochemical gradient of the membrane^12,13. Preserving this gradient (commonly known as the proton motive force (PMF)) is, in turn, crucial for transporting energetic and biosynthetic substrates into the cell in all bacteria, for flagellar motility in flagellated bacteria and for ATP synthesis in bacteria that are unable to sustain substrate-level phosphorylation¹⁴. Thus, the continual supply of energy and building blocks is highly interdependent on the preservation of the PMF. Many bacteria can re-route metabolism to respond to specific limitations in this supply with impressive flexibility, enabling them to shift to alternative sources of energy and building blocks while balancing flux through central metabolic pathways.

Alternative sources of substrates for energy and biosynthesis.

Severe substrate limitation is often the underlying cause of growth arrest, as it leads to reduced rates of both anabolic and catabolic metabolism, and forces cells to rely on alternative sources of energy and biosynthetic substrates, including internal stores. Indeed, cellular components themselves can be catabolized, which solves two problems that arise from severe substrate limitation: it provides nutrients and it removes the burden of maintaining cellular machinery that has become dispensable as overall cellular activity decreases. The result of this catabolism is a large reduction in cell size and volume⁴, which increases the surface area-to-volume ratio of the cell and thus, in theory, could increase the efficiency of substrate transport (see below). Catabolism in growth-arrested bacteria has been shown most clearly to target ribosomes (see below) and membrane phospholipids. For example, genetic studies in E. coli found that derepression of genes that are involved in β-oxidation, and thus fatty acid degradation, is required for long-term survival in stationary phase¹⁵. Vibrio cholerae also undergoes a substantial decrease (>99%) in total lipids within 7 days of starvation, despite an increase in viable cell count due to reductive divisions (see below), which led to the proposal that membrane phospholipids are an endogenous energy source to maintain viability in these cells¹⁶. A recent microarray analysis of carbon-starved Vibrio harveyi cells supported this hypothesis, showing that genes that are involved in fatty acid β-oxidation were upregulated during starvation, concomitant with a decrease in cell size. Genes that have important roles in metabolic reactions that require acetyl-CoA, which is the major metabolite that is generated from the degradation of fatty acids, were also upregulated; many of these genes are part of the glyoxylate shunt, which is an anaplerotic pathway that bypasses most of the tricarboxylic acid (TCA) cycle, which is substantially downregulated during starvation¹⁷ (FIG. 1).

Figure 1 |. Metabolic rewiring during growth arrest.

Different organisms use distinct metabolic strategies to meet cellular requirements under growth-limiting conditions. Both Mycobacterium spp. and Pseudomonas aeruginosa must adjust their strategies for maintaining their membrane electrochemical gradient under oxygen-limited conditions. Mycobacterium spp. continue to use the electron transport chain, using either low levels of oxygen or fumarate as the terminal electron acceptor, and also contribute to the membrane electrochemical gradient by secreting succinate, which is generated through reversal of the tricarboxylic acid (TCA) cycle in Mycobacterium tuberculosis and by the glyoxylate shunt in Mycobacterium smegmatis. In P. aeruginosa, fluxes through the TCA cycle and electron transport chain drop close to zero under anoxic conditions, and substrate-level phosphorylation generates ATP to run the ATP synthase in reverse, thus pumping protons outward across the membrane. Under carbon-limiting conditions, fatty acid β-oxidation generates acetyl-CoA, which can be fed into biosynthetic pathways through the glyoxylate shunt; this generates the TCA cycle intermediates that are most useful as precursors without overproducing other intermediates. P_i, inorganic phosphate; Q, quinone.

In addition to catabolism of internal stores, increased active acquisition of exogenous nutrients is an important survival strategy during nutrient limitation. In heterotrophs, limitation of organic carbon causes shortages of both energy and biosynthetic precursors. Many heterotrophs respond to this shortage by upregulating high-affinity transporters to scavenge carbon from the environment. For example, Vibrio sp. strain Ant-300 effectively depletes nanomolar concentrations of arginine from its environment and, interestingly, only exhibits chemotaxis toward arginine after starvation¹⁸. Both high-affinity and low-affinity transport depend on respiration rather than ATP hydrolysis for energy, which suggests that the maintenance of the PMF is crucial for uptake by these transporters. Limitation of phosphate, nitrogen, sulfur and other nutrients also induces the upregulation of scavenging mechanisms in diverse organisms^19–21.

The evolutionary significance of competitive scavenging for resources is illustrated by the dynamics of populations of E. coli in batch culture. During long-term stationary phase in rich complex medium, in which the ‘growth advantage in stationary phase’ (GASP; BOXES 1,2) phenotype arises, E. coli mutants arise that take over the population. These mutants are able to outcompete ‘naive’ wild-type cells in stationary phase competition experiments²². Strikingly, the majority of the GASP mutations that have been studied result in improved amino acid uptake and catabolism, by affecting genes such as those that encode leucine-responsive regulatory protein (Lrp; a nucleoid-associated protein) or the sigma factor RpoS (see below), or genes that are located in the amino acid transport locus ybeJ-gltJKL^6,23,24. Increased amino acid uptake ability is particularly important in the long-term stationary phase that is associated with the GASP phenotype, as the breakdown products of dead and dying cells are the major sources of nutrients.

Re-routing metabolic pathways.

A final strategy by which cells respond to nutrient limitation is to re-route metabolic fluxes to maintain acceptable levels of affected intermediates. For example, the preferentially aerobic pathogens Mycobacterium tuberculosis and Pseudomonas aeruginosa have very different responses to the hypoxic conditions that they experience during human infections²⁵. As oxygen is preferred by these bacteria as a terminal electron acceptor, hypoxia causes decreased flux through the electron transport chain, which, in turn, causes a decrease in the PMF and an accumulation of reducing equivalents (primarily NAD(P)H). In M. tuberculosis, hypoxic conditions lead to a decreased but stable level of ATP, the maintenance of which requires the ATP synthase and some degree of PMF across the membrane²⁶. The PMF is maintained by the forward fluxes through the TCA cycle and the electron transport chain that can be sustained by the trace amounts of oxygen available to accept electrons, supplemented by electrogenic secretion of succinate. This succinate can be produced by one or both of two pathways: a reversal of the reductive branch of the TCA cycle through the upregulation of phosphoenolpyruvate carboxykinase, malic enzyme and fumarate reductase (which also re-oxidizes NADH); and the glyoxylate shunt, which bypasses steps of the TCA cycle that produce reducing equivalents^27–29 (FIG. 1). Interestingly, M. tuberculosis preferentially generates succinate by reversing the TCA cycle, whereas Mycobacterium smegmatis uses the glyoxylate shunt and Mycobacterium bovis uses both pathways²⁹, which shows that the re-routing of metabolic pathways in response to the limitation of terminal electron acceptors can vary, even between closely related bacterial species. M. tuberculosis can use metabolic reorganization as part of a strategy to survive hypoxic conditions for years without observable growth^30,31; indeed, even when oxygen is completely absent, M. tuberculosis has been suggested to survive by using endogenously generated fumarate as an alternative terminal electron acceptor for oxidative phosphorylation²⁶ (FIG. 1). In contrast to mycobacteria, P. aeruginosa dispenses with oxidative phosphorylation altogether when oxygen is severely limited, and can instead maintain the PMF anaerobically by reversing the reaction that is catalysed by the ATP synthase³², with ATP supplied by substrate-level phosphorylation using pyruvate or arginine as a substrate^33,34 (FIG. 1). Under these conditions, fluxes through the TCA cycle and the electron transport chain are presumably close to zero. Together, these examples illustrate the flexibility of metabolic networks in maintaining crucial metabolic parameters during severe nutrient and energy limitation.

Further questions.

Although recent work has advanced our understanding of the metabolic strategies that are used to survive growth arrest, much is still unknown. At the level of single cells, questions remain about the absolute minimal energy requirements for viability, and how these limits vary according to the organism and environment, despite progress in investigating these boundary conditions⁴. At an environmental level, complexities such as co-limitation for different substrates in a changing environment and interaction with other microorganisms are likely contributors to energy dynamics in the natural world and significant forces in the evolution of strategies to regulate metabolism^35,36. For example, in at least some cases, including low-energy environments, different species of microorganism can form cooperative metabolic interactions — an arrangement known as syntrophy³⁷. Indeed, mutually beneficial metabolic interactions may be more common than is currently appreciated; for example, redox-active phenazine pigments that are synthesized by Pseudomonas spp. were recently shown to support substrate-level phosphorylation by P. aeruginosa³². Furthermore, studies have suggested that other organisms may also benefit from the presence of phenazines produced by P aeruginosa in some contexts³⁸. Understanding metabolic network connectivity, even within one organism, still presents a research challenge, and we are just beginning to probe the metabolic interactions of microbial communities. Identifying new energy yielding pathways within community contexts is a priority for future research.

Regulation of gene expression

Limitation for energy, nucleotides and amino acids are common features of non-growing states that probably impose general constraints on gene expression, although the precise identities of upregulated and downregulated genes vary depending on the organism and the underlying causes of growth arrest. The constraints on gene expression may differ in non-growing states from those imposed by the better understood challenges of exponential growth and the initial transition to stationary phase, and recent technical advances have made feasible the study of how gene expression might adapt to the constraints that are imposed by growth arrest at a molecular level.

Regulatory paradigms of different growth states.

During exponential growth in nutritionally complete media, most gene expression is directed towards the biosynthesis of ribosomes, which are the principal mass constituent of each new cell being produced and are the drivers of the biosynthesis of all other proteins (reviewed in REF 39; FIG. 2a). Under these conditions, rates of translation elongation are limited by the intrinsic properties of the ribosome rather than the availability of amino acids⁴⁰. However, the unlimited availability of amino acids is short-lived even in rich medium batch culture. Thus, some resources will soon become limiting, at which time global regulatory systems will coordinate a transition away from maximum growth. Two global regulators of this transition that are phylogenetically widely distributed but best characterized in E. coli are the sigma factor RpoS, which is induced by stresses such as heat and osmotic shock as well as nutrient downshift, and the stringent response alarmone guanosine pentaphosphate ((p)ppGpp), which is synthesized in response to signals of nutrient limitation (reviewed in REF. 9 and REF 41, respectively). RpoS has relaxed sequence specificity compared with the housekeeping sigma factor RpoD, thus favouring the expression of a regulon of stress-adaptive genes with non-consensus promoters, but also drives lower levels of transcriptional activity⁴². (p)ppGpp has pleiotropic effects on many cellular processes, and, together with the co-regulator DksA, strongly represses rRNA and ribosomal protein gene expression^43,44. Together, RpoS and (p)ppGpp lead to a reduction in the total rates of gene expression, but also lead to a redirection of biosynthetic capacity away from ribosome biogenesis and towards more urgent concerns, such as preventing or repairing DNA damage (see next section), osmoprotection, refolding damaged proteins or increasing the synthesis of missing biosynthetic intermediates (FIG. 2b).

Figure 2 |. Transcription and translation during different growth phases.

a | During exponential phase, rRNA genes are among the most highly transcribed in the cell, as ribosome biogenesis is a top biosynthetic priority. In addition, genes with promoters that have strong consensus sequences for RpoD binding are highly expressed and efficiently co-transcriptionally translated, aided by the transcription elongation factor NusG, which helps physically associate a ribosome with the RNA polymerase. The stationary phase sigma factor RpoS is synthesized to some extent but fails to compete with RpoD for RNA polymerase; consequently, stress-responsive genes with promoters that do not match the RpoD consensus sequence are not efficiently expressed. b | In the transition to stationary phase, limitation for amino acids activates RelA, which senses uncharged tRNAs and synthesizes the alarmone guanosine pentaphosphate ((p)ppGpp). (p)ppGpp, in conjunction with DksA, represses the transcription of rRNA by destabilizing the rRNA open promoter complex. The decrease in abundance of the nucleoid-associated protein Fis and the increase in abundance of leucine-responsive regulatory protein (Lrp) also contribute to rRNA repression. Hibernation promoting factor (HPF) and ribosome modulation factor (Rmf) are upregulated and lead to the dimerization of ribosomes to 100S complexes that are inactive for translation. RNA polymerase complexes with RpoD are selectively sequestered through several mechanisms, including binding to a small RNA (6S RNA), and levels of RpoS are also increased, which leads to increased transcription of stress-responsive genes in the RpoS regulon. RpoS can also drive transcription of housekeeping genes that have RpoD-consensus promoters, but does so at much lower levels than transcription of these genes that is mediated by RpoD. c | During growth arrest, overall gene expression activity is much lower than in exponential phase or the transition to stationary phase. Although much remains to be elucidated about how these very low levels of activity are regulated, the observations that several global regulators change in abundance suggest some possible mechanisms. Regulators that are important during the transition to stationary phase, such as DksA, (p)ppGpp, RpoS and HPF, seem to be downregulated during growth arrest. Also, the complement of nucleoid-associated proteins (NAPs) changes substantially, which probably affects the expression of rRNA and other genes, although details remain to be explored. In Pseudomonas aeruginosa, some factors that are thought to contribute to transcription and translation elongation (GreA, S10 and elongation factor P (Efp)) were upregulated, possibly suggesting that they could help buffer against pausing and arrest in severely substrate-limited conditions. Although some ribosomes are catabolized, with the dual benefit of decreasing the number of ribosomes that are competing for amino acid substrates and liberating nutrients to be used for energy and maintenance, the newly identified transcriptional regulator SutA, which is upregulated during growth arrest in P aeruginosa, enhances rRNA and ribosomal protein gene expression, which suggests that some repair and replacement of ribosomes may also be important.

Decreased availability of nucleotides and amino acids probably begins to affect transcription and translation rates during this transition phase, but (p)ppGpp and DksA function, in part, by further sensitizing the initiation of transcription to the availability of nucleotides — a strategy that makes sense during a transition from high to low nucleotide availability, but perhaps not if limiting nucleotides become a long-term challenge. Indeed, several lines of evidence suggest that RpoS and (p)ppGpp govern a transient, active adjustment to dynamic conditions rather than survival of a long-term non-growing state. DksA and (p)ppGpp have been shown to decrease to low levels in late stationary phase in P. aeruginosa and E. coli, respectively^45,46, and two studies have suggested that levels of RpoS in P. aeruginosa may actually be lower in the non-growing state than in mildly nutrient-limited states^47,48. In addition, mutations in rpoS that decrease function cause a GASP phenotype and provide a selective advantage during continuous culture at the lowest possible dilution rates, which suggests that prolonged RpoS activity may be maladaptive^7,49. Although gene products of the RpoS regulon have important roles in growth-arrested states, the many layers of regulation that affect levels of RpoS suggest a delicate balance between preparing for growth arrest and actually surviving a prolonged non-growing state^50–52.

Nucleotide and energy levels continue to decrease after growth is arrested⁵³, potentially affecting both the priorities and mechanisms of gene expression regulation. As discussed, the priorities that are reflected by upregulated genes include using alternative energy and nutrient sources, and carrying out essential repair and replacement of cellular components. The mechanisms of expression regulation are less clear — the biochemical implications of severe nucleotide and amino acid shortages are of great interest. Studies of E. coli in exponential phase and early stationary phase suggest that even moderate nutrient downshifts lead to profound changes in the stability of open promoter complexes and the tendencies of both polymerases and ribosomes towards pausing, permanently arresting or terminating^40,3,44,54 (see REFS 55,56 for reviews). A successful regulatory strategy for non-growing states must favour the expression genes that are essential for survival, but must also ensure that the transcriptional and translational activities of the cell are matched to the available resources, and that relevant cellular machinery can handle slow elongation rates and long pauses (FIG. 2c).

Tuning the capacity and rate of gene expression to match limited substrates.

One strategy for tolerating shortages of nucleotides and amino acids is to limit the number of active polymerases and ribosomes so that fewer of these complexes are competing for the limited supply of substrates. Indeed, both RNA polymerases and ribosomes are sequestered at the entry to stationary phase, which reduces the number of active complexes. RNA polymerases bound to RpoD are sequestered by binding to a small RNA, at least in E. coli⁵⁷, and ribosomes are sequestered by hibernation promoting factor or ribosome modulation factor (or both, depending on the organism; reviewed in REF. 56), which mediate the formation of inactive ribosome dimers. In Listeria monocytogenes, the abundance of hibernation promoting factor peaks at the entry to stationary phase, and then decreases⁵⁸. Further into growth arrest, ribosome degradation begins to have an important role in the regulation of the number of ribosomes, and has the dual benefit of both limiting the number of active ribosomes and converting unused ribosomes to nutrients^58,59. In E. coli, studies have shown that individual 30S and 50S subunits are preferred substrates for degradation during starvation; therefore, ribosome dimerization at the entry to stationary phase may help delay degradation until it is essential⁶⁰. Many bacteria also encode additional ribosome-binding proteins that are upregulated in stationary phase and that may have distinct roles in modulating the activity, sequestration or degradation of ribosomes during growth arrest. For example, RsfA, which is a highly conserved factor that has been shown to prevent subunit joining in E. coli⁶¹, is upregulated during anaerobic survival in P. aeruginosa⁴⁸. Understanding how these factors coordinate to affect the fates of ribosomes during growth arrest will require further study.

Another strategy may be to tune the transcription and translation machineries to be less sensitive to pausing and substrate limitation, through changes in the levels of accessory factors. Detailed in vivo studies of such factors have not, for the most part, been conducted in non-growing bacteria, but some possible mechanisms are suggested by considering the proteins that are upregulated during anaerobic survival in P. aeruginosa in light of their functions in other growth states. One such factor is an RNA polymerase-binding protein, SutA, which enhances transcription of rRNA and ribosomal protein genes during slow growth. SutA may decrease the sensitivity of these genes to shortages of nucleotides and amino acids, and thus enable some repair and replacement of ribosomes during growth arrest⁴⁸. Although homologues of SutA are found only in a subset of Gammaproteobacteria (not including E. coli), another RNA polymerase accessory factor that may have a similar function is CarD, which is widely distributed outside the Gammaproteobacteria. CarD has been shown to be upregulated during nutrient limitation⁶² and to enhance the transcription of rRNA genes⁶³ in M. smegmatis, which suggests that this type of regulatory response may be of broad importance. Two additional factors that are known to have roles in transcription elongation, at least in growing cells, and that are upregulated in P. aeruginosa during anaerobic survival are the ribosomal protein S10, which can ‘moonlight’ as a transcription elongation factor through its interaction with NusG⁶⁴, and GreA, which can help to rescue RNA polymerases that are arrested in ‘backtracked’ states (known as backtracked RNA polymerases)⁶⁵. Finally, upregulation of the translation factor elongation factor P (Efp) may help buffer translation against stalling-induced arrest⁶⁶.

Further questions.

Although the observations described here imply that the non-growing state is actively regulated, the molecular details of the specific mechanisms that are responsible remain largely unknown, and there are hints that the solutions to fundamental problems of non-growing states may be quite diverse between phylogenetically divergent bacteria. This means that researchers must broaden the scope of inquiry to include more model organisms, including ones that are less adapted to laboratory growth conditions, and indicates that comparative and functional genomics techniques may become increasingly useful for identifying and understanding regulatory paradigms. Exploring molecular mechanisms in low-activity, heterogeneous states is challenging, but new applications for doing so have shown promise, notably those that are based on next-generation sequencing or proteomics techniques (BOX 3).

Maintenance and replication of DNA

A defining feature of non-growing states is the repression of DNA replication and cell division, and accompanying changes to the nucleoid have mostly been studied in E. coli and other model organisms as they transition towards stationary phase. The most observable feature of the stationary-phase nucleoid is a condensed, even crystalline, morphology when viewed by electron microscopy⁶⁷, which is thought to protect the DNA from damage and confer a survival advantage.

The generation times of exponentially growing E. coli cells are shorter than the time that is required to replicate the chromosome, which means that a new round of replication must start before the previous one completes. However, this process must be regulated to ensure a yield of exactly one replicated chromosome per cell division (FIG. 3a). One of the main factors in the regulation of the initiation of chromosome replication is DnaA, which is the ATPase that binds to the origin of replication. To prevent new rounds of initiation as the cell transitions to a non-growing state, the level, availability and activity of DnaA are markedly downregulated by several mechanisms (reviewed in REFS 68,69). During this transition to stationary phase, ongoing rounds of replication can still be completed, but the accompanying cell divisions result in small progeny because ribosome biosynthesis and other biosyntheses are suppressed. Recent work suggests that the cell division machinery in E. coli may be capable of directly sensing the nutritional status of the cell through interactions between the membrane glycosyltransferase osmoregulated periplasmic glucans H (OpgH; also known as MdoH) and the cytokinesis regulator FtsZ. Poor nutrient status leads to lower levels of OpgH, which removes a block to FtsZ assembly and cytokinesis, permitting reductive divisions during nutrient starvation⁷⁰ (FIG. 3b).

Figure 3 |. Overview of cellular morphology with emphasis on nucleoid.

Cells undergo gross morphological changes in transitions between different growth states. a | In exponential phase, the chromosome has a high degree of negative supercoiling, owing to large amounts of active transcription. RNA polymerase is mostly bound to DNA and is gathered in large clusters of highly transcriptionally active genes, which tend to migrate to the periphery of the nucleoid region. Ribosomes are observed in the central portion of the cell, sharing space with the nucleoid. New rounds of replication, initiated by DnaA, begin even before previous rounds have completed. Cytokinesis is inhibited by interaction of an abundant metabolic enzyme (for example, OpgH in Escherichia coli) with FtsZ, thus maintaining a larger cell size. b | In the transition to stationary phase, decreases in total transcription, and rRNA transcription in particular, lead to less supercoiling of the chromosome but a more condensed overall morphology, with fewer RNA polymerases and ribosomes associated with the nucleoid region in the centre of the cell. New rounds of replication are inhibited by the decreased abundance and activity of DnaA, but cell division to segregate chromosomes that have already been replicated can still take place (facilitated, in part, by a decrease in OpgH, which releases FtsZ), leading to progeny with a small cell size. The abundance of the nucleoid-associated DNA-binding protein from starved cells (Dps) begins to increase, owing to transcriptional upregulation by the stationary phase sigma factor RpoS. c | During growth arrest, an extremely high abundance of Dps leads to a highly condensed, crystalline appearance of the nucleoid region. Reductive divisions in stationary phase, combined with catabolism of ribosomes and membranes, leads to much smaller cell sizes. Transcription and translation still occur, despite reduced ribosome abundance, but at greatly decreased rates.

The nucleoid: condensation and protection from DNA damage.

As cells progress towards growth arrest, the nucleoid undergoes increasingly extreme morphological changes. In general, the nucleoid becomes more condensed, while clusters of bound, transcribing RNA polymerase dissipate and ribosomes move from the central region of the cell, which they had shared with the nucleoid, to the poles⁷¹. Some of these changes are driven by the global changes in gene expression that are discussed above; in particular, the substantial decrease in rRNA transcription can sufficiently alter the distribution of RNA polymerase to affect the gross morphology of the nucleoid⁷² (FIG. 3b,c). However, a crucial factor that drives changes in the nucleoid is the more specific transcriptional changes that occur in the expression of individual nucleoid-associated proteins (NAPs), which are proteins that bind to DNA with low or no sequence specificity and can give rise to higher-order structures. Most organisms encode several NAPs, which range in specificity from nearly universal to species specific (reviewed in REF 73). The most abundant NAP shifts from factor for inversion stimulation (FIS)⁷⁴ during nutrient upshift and early exponential phase⁷⁵ to DNA-binding protein from starved cells (Dps) in stationary phase — a change that is driven, in part, by the regulatory activity of RpoS^76,77. Dps, which has homologues in many bacterial species, can condense DNA through ordered self-aggregation, both in vivo and in vitro^67,78, and transmission electron micrographs and atomic force microscopy have shown that nucleoid condensation in stationary-phase E. coli is dependent on Dps^79–80 (FIG. 3c). Condensation of the nucleoid by DPS separates the DNA from potentially damaging reactants in the cytoplasm, holding it in a low-energy equilibrium state that is distinct from the dynamic disequilibrium of the nucleoid in exponential phase⁷⁹. In addition to protecting DNA by this spatial separation, Dps may protect DNA from oxidative damage⁸¹. Dps has structural homology to ferritin⁸² and can oxidize ferrous iron to its ferric, insoluble form⁸¹. This reaction is proposed to compete with the Fenton reaction, thus protecting cellular DNA from oxidative damage by decreasing the production of oxidizing free radicals⁷⁶.

Despite the potentially protective effects of Dps, studies suggest that repairing DNA damage during growth arrest is still a priority for bacterial cells. In a detailed study of their expression, all three DNA polymerases that are responsible for error-prone repair (Polll, PolIV and PolV) were upregulated at various times during long-term stationary phase, and combinatorial deletions of these genes conferred competitive fitness disadvantages⁸³. In addition, microarrays of V. cholerae entering a viable-but-nonculturable (VBNC) state showed that polB, which encodes PolII, was the most upregulated gene in the transcriptome⁸⁴, and proteomics of P. aeruginosa during anaerobic survival showed upregulation of PolA (the polymerase that fills in gaps in the lagging strand) and DNA ligase⁴⁸, both of which have functions that could contribute to DNA repair. The sources of DNA damage during growth arrest are probably diverse and could include increases in free radicals and oxidative damage that may arise from challenges in maintaining flux through the electron transport chain; the possible accumulation of ongoing abiotic damage over long periods of time; and the double-stranded breaks that could arise if the cell fails to efficiently complete replication.

Further questions.

Many questions regarding the maintenance of the nucleoid during growth arrest remain unaddressed. For example, how does the condensed nucleoid interact with enzymes that mediate transcription, replication and repair? Several possible scenarios seem plausible. The distribution of NAPs along the chromosome in non-growing states has not been investigated in detail, and some chromosomal regions may remain free of NAPs and thus be less condensed, even in growth states in which NAPs are some of the most abundant proteins in the cell. NAPs may also be capable of translocating along the chromosome or temporarily dissociating from the DNA, possibly through a mechanism that takes advantage of the high sensitivity of Dps–DNA interactions to concentrations of divalent cations and pH⁸⁵.

Another question is, how do other NAPs contribute to nucleoid organization and protection during growth arrest, both in E. coli and in other organisms? Work in E. coli has shown that the NAPs curled DNA-binding protein A (CbpA), integration host factor (IHF), histone-like nucleoid-structuring protein (H-NS), heat-unstable protein (HU) and Lrp all interact with the chromosome during stationary phase, despite being less abundant than Dps^73,86. These proteins can substantially affect gene expression, as exemplified by the role of Lrp in regulating the transport of amino acids (see above), and rates of mutagenic break repair⁸⁷; therefore, the mechanisms by which they interact with specific regions of the chromosome during growth arrest could be important. Furthermore, different organisms have different complements of NAPs encoded in their genomes⁷³, and even homologous NAPs seem to have different regulatory parameters in different organisms. For example, in Staphylococcus aureus, Dps is induced by oxidative stress and not by the stationary phase as observed in E. coli⁸⁸, and in P. aeruginosa, the only NAP that is upregulated during anaerobic survival is HU⁴⁸.

Finally, little is known about how and why individuals of a non-growing population might occasionally engage in cycles of replication and cell division. An important insight from studies of the GASP phenotype is that the ‘evolution’ evident in the selective sweeps is possible at all; this proves that a numerically stable population of viable cells does in fact experience replication and division⁸⁹. Can genetic changes in a seemingly non-growing population be explained by relatively normal division cycles that occur infrequently in a small number of cells? Or does DNA replication and repair instead occur in larger numbers of cells but at very slow rates? What are the mechanisms? As new techniques facilitate the investigation of the non-growing state in diverse organisms, such questions should become easier to address.

Concluding remarks

Although the characterization of molecular mechanisms in non-growing states of bacteria remains in its infancy, observations from diverse fields and organisms suggest that these states are actively controlled to ensure surprisingly robust survival in the face of deprivation. Gene expression programmes in these states reflect the priorities of catabolizing internal energy stores and scavenging trace nutrients from the environment; re-routing metabolism to integrate these alternative sources into central pathways; protecting and repairing DNA; and maintaining biosynthetic capacity for replenishing other essential molecular components of the cell. Furthermore, the gene expression machinery itself may adapt to the shortages of nucleotides and amino acids that are encountered. A better understanding of these priorities and mechanisms will open new doors for investigation of contexts in which non-growing states are probably required and regulated, such as symbiotic associations with plants and animals, and stable, structured microbial communities in both natural and applied settings^90,91.

Towards this understanding, we have highlighted specific unanswered questions regarding changes in gene expression and two other important aspects of the biology of non-growing states: the maintenance of metabolic energy and the replication and maintenance of DNA. In doing so, we emphasize that fundamental questions about growth-arrested states are still not fully understood, even in model organisms. The examples that are cited in this Review are predominantly from the Gammaproteobacteria, owing to the biases of the authors and the literature, but further exploration of the profound diversity across bacterial species should be highly encouraged. We have focused on the potentially unifying characteristics that are observed in various non-growing bacterial populations as a starting point and the sample of studies discussed indicate the presence of general principles that transcend individual organisms — but also of diversity in the mechanisms by which these principles are applied. We expect that broader investigations into non-growing states of diverse bacteria will uncover substantial deviations from the picture that has emerged based on the information that is currently available. Further complexity arises from the heterogeneity observed in growth states, even for clonal populations in the laboratory. Such heterogeneity is probably central to the function of microbial communities, and deciphering how it arises and what it contributes to population fitness will require attention to be paid to both the intracellular mechanisms and the environmental factors that constrain non-growing states. Growth arrest is becoming ever more amenable to laboratory or in situ study with the development of tools that enable sensitive and specific measurements of subsets of (or even individual) cells and their microenvironments, and the increasing ease of creating genetic systems in diverse organisms. We hope that a piqued interest in the cellular and molecular biology of non-growing bacteria will help bridge a gap between environmental and molecular microbiology perspectives, and ultimately reveal the molecular underpinnings of the most dominant mode of microbial existence on the planet.

Glossary

Exponential phase

Microbial population growth that can fit to the exponential equation N(t) = N₀e^kt, where N(t) is the population size at time t, N₀ is the starting population size, e is the base of the natural logarithm, and k is a constant. Exponential growth is generally assumed to occur when no resource is limiting, to be balanced and at steady state

Stationary phase

A growth phase of microbial populations that occurs after at least one resource becomes limiting for growth. At the transition to stationary phase, the population continues to increase in size, but the rate of increase decreases; in stationary phase, the population size stops increasing

Adenylate energy charge (AEC)

A value based on the ratio of high energy phosphate bonds in ATP and ADP molecules to the total amount of adenylate in the cell

Antibiotic tolerance

The survival of cells that are exposed to high doses of antibiotics for periods of time that would usually be lethal. Antibiotic tolerance extends the length of time that a cell survives exposure to the drug, whereas resistance enables a cell to survive an increased concentration of the drug

Persisters

A subpopulation of cells that exhibits antibiotic tolerance in a population in which other cells are killed by the same dose and length of exposure to a drug. Persisters were first noted in an exponential-phase culture that was treated with high doses of antibiotics for extended periods of time

Anabolic

Metabolic reactions that construct larger macromolecules from smaller substrates

Catabolic

Metabolic reactions that break down macromolecules into smaller components for the generation of energy or for recycling

Reductive divisions

Cell divisions that are uncoupled from biosynthesis and growth, leading to progeny that are smaller in size. These contribute to the decrease in cell size that is observed during stationary phase

Glyoxylate shunt

An alternative to the standard tricarboxylic acid (TCA) cycle in which steps that generate reduced NAD(P)H are bypassed to enable succinate, fumarate, malate and oxaloacetate to be produced for biosynthetic reactions without generating reducing equivalents. The glyoxylate shunt is useful in the context of limitation for terminal electron acceptors or the catabolism of lipids

Anaplerotic

Reactions that replenish key intermediates of central metabolic cycles to compensate for their use by other biosynthetic pathways

Nucleoid

The chromosome and associated proteins

Sigma factor

A protein that recruits RNA polymerase to a specific set of promoters on DNA. Some sigma factors have large regulons, whereas others drive expression from only a few loci

Electrogenic secretion

Symport of a substrate (with its chemical gradient) and a proton (against its chemical gradient) that results in a net increase in the proton motive force across a membrane

Syntrophy

A mutually beneficial metabolic interaction between two (or more) species of microorganism

Stringent response

A conserved regulatory mechanism that coordinates the responses of bacteria to nutrient downshift. The response is mediated by the small-molecule alarmone (p)ppGpp, the synthesis of which from ATP and GDP or GTP is stimulated by uncharged tRNAs or disrupted lipid biosynthesis

Regulon

The group of genes that is regulated by a specific regulatory factor

Open promoter complexes

The intermediate in transcription initiation in which RNA polymerase has bound to a promoter and unwound the double-stranded DNA, which allows the template strand of the DNA to pass through the active site of the polymerase

Backtracked RNA polymerases

Transcribing RNA polymerases that slip backwards along a template after pausing, which causes the RNA–DNA hybrid at the 3′ end of the nascent transcript to unwind

Origin of replication

The site on the bacterial chromosome, determined by its sequence, where the two strands of DNA are unwound to enable replication of the chromosome to begin

Fenton reaction

A metal-catalysed free radical chain reaction in which Fe²⁺ is oxidized by H₂O₂ to produce OH⁻ and OH^• which is a highly reactive radical species