Adaptive strategies under prolonged starvation and role of slow growth in bacterial fitness

Adaptive evolution has the power to illuminate genetic mechanisms under a pre-defined set of selection factors in a controlled environment. Laboratory evolution of bacteria under long-term starvation has gained importance in recent years because of its ability to uncover adaptive strategies to overcome prolonged nutrient limitation- a condition thought to be encountered often by natural microbial isolates. In this evolutionary paradigm, bacteria are maintained in an energy-restricted environment in the growth phase called as long-term stationary phase or LTSP. This phase is characterized by a stable viable population size and highly dynamic genetic changes. Multiple independent iterations of LTSP evolution experiments have given rise to mutants that are slow-growing compared to the ancestor. Although the antagonistic regulation between rapid growth and stress response is fairly well-known in bacteria (especially Escherichia coli), the reason behind the growth deficit of many LTSP-adapted mutants has not been explored in detail. In this review, I revisit the trade-off between growth and stress response and delve into the regulatory mechanisms currently known to control growth under nutrient deficiency. Focusing on the theme of “sigma-factor competition” I try to search for the evolutionary reasoning of slow growth amongst mutants adapted to prolonged starvation. Additionally, I present novel experimental data indicating the dynamics of four such slow-growing variants that evolved during a 30-day long LTSP evolution experiment with Escherichia coli.


Abstract
Adaptive evolution has the power to illuminate genetic mechanisms under a pre-defined set of selection factors in a controlled environment. Laboratory evolution of bacteria under long-term starvation has gained importance in recent years because of its ability to uncover adaptive strategies to overcome prolonged nutrient limitation-a condition thought to be encountered often by natural microbial isolates. In this evolutionary paradigm, bacteria are maintained in an energy-restricted environment in the growth phase called as long-term stationary phase or LTSP. This phase is characterized by a stable viable population size and highly dynamic genetic changes. Multiple independent iterations of LTSP evolution experiments have given rise to mutants that are slowgrowing compared to the ancestor. Although the antagonistic regulation between rapid growth and stress response is fairly well-known in bacteria (especially Escherichia coli), the reason behind the growth deficit of many LTSP-adapted mutants has not been explored in detail. In this review, I revisit the trade-off between growth and stress response and delve into the regulatory mechanisms currently known to control growth under nutrient deficiency. Focusing on the theme of "sigmafactor competition" I try to search for the evolutionary reasoning of slow growth amongst mutants adapted to prolonged starvation. Additionally, I present novel experimental data indicating the dynamics of four such slow-growing variants that evolved during a 30-day long LTSP evolution experiment with Escherichia coli.
preservation (also called SPANC) (20). Whenever the ambient nutrient availability becomes suboptimal, resources are allocated to express stress-response proteins as opposed to proteins assisting rapid growth (21), thereby reducing the growth rate.
A fraction of slow-growing bacteria are observed to form small-sized colonies on solid agar platesa phenotype ubiquitously observed in specific isolates of both natural and laboratory-maintained strains. The first documented report of small colony variants(SCV) dates back to 1910 (22). Over time, more studies reported the emergence of small colonies upon exposure to chemicals like copper sulphate (23), phenol (24)), and antibiotics like gentamycin (18). SCV also poses a threat in the public health sector, as multiple groups demonstrated key pathogens to form SCV after being isolated from the site of infection, establishing SCV as a dominant phenotype among pathogens (25). Staphylococcus aureus (26) remains the most well-characterized species to form SCV, although other pathogens like Pseudomonas aeruginosa (18,27), Staphylococcus epidermidis (28), Escherichia coli (17), Serratia marcescens (29), and Neisseria gonorrhoeae (30) are also known to form small colonies. A reduced rate of respiration and pigmentation are characteristic of pathogenic SCVs (25). They consistently demonstrate slow growth primarily through metabolic mutations, either being auxotrophic to thymidine or being deficient in the electron transport chain (ETC) pathway. Due to their near-dormant metabolic state, SCVs may avoid immune responses from the host and multiple externally administered antibiotics (25) and persist within host niches for prolonged periods, driving chronic diseases (17). Enhanced tolerance of SCV to multiple stressors, a feature identified consistently even in non-pathogens, makes it a public health challenge. In order to specifically target and kill infectious SCVs further investigations into the regulatory origin of this phenotype are required both in the laboratory and clinical settings. 4 2.2 Repeated emergence of slow-growing variants during adaptive lab evolution in longterm stationary phase Batch cultures of bacteria have been shown to remain viable for years without the addition of external nutrients in a paradigm called long-term stationary phase (LTSP) (31)(32)(33)(34). In contrast to evolution under one or a set of pre-defined stressors, the evolving population is exposed to a dynamic set of stress factors during prolonged starvation. Under such conditions, batch cultures are shown to experience cycles of "feast-and-famine" in their environment akin to bacteria in natural habitats like the host, river bed sediment, or soil (12,32,35). Multiple independent studies have shown that the bacterial population increases in both genotypic and phenotypic diversity during evolution in batch cultures for prolonged periods (13,34,36). Slow-growing variants have also been observed to emerge during LTSP across different studies (12,13,34).
To check the repeatability of emergence of small colony variants in the LTSP paradigm, the LTSP evolution experiment was re-iterated following a previously published study from our group (36).
As compared to the 5 strains that were evolved in the previous work (36), 11 evolving isolates were followed this time for 60 days (Fig 1). From these 11 isolates, SCV emerged in 4 lines, along with other interesting colony morphologies in the same sequence as compared to the previous runs ( Fig   2). Across both runs of the evolution experiment, SCV was observed to emerge around the third week (Day 19-22 post-inoculation) of the experiment (Fig 1.A, Additional files 1), and was observed for two weeks until the cultures were monitored (Fig 1. B,C, Additional file 2). This suggests that underlying changes in media conditions select small colony phenotype during this period in prolonged stationary phase. The genetic basis of the SC described in Nandy et al (13) was mapped to the rpo operon, secifically -a point mutation in rpoC gene. Coding for the β' subunit of RNA polymerase core, this gene is a fairly common target to be mutated across LTSP experiments (37). While analyzing the whole-genome sequences of the mutants from the second run of the evolution experiment, it was found that different SC isolates harbored unique mutations in different alleles and no alleles were featured in more than one strain (Table 1). No mutations were observed in the rpo operon as unique in any of the new SC strains, suggesting that slow growth is strongly selected in the LTSP but the genetic pathway to slow down growth is not fixed. In the second batch of SC mutants, mutations were also observed in several DNA-binding transcription factors like PaaX and Lon protease, some of which were previously implicated in survival under stress (38).
Apart from global regulators, unique deletions in acrB a multi-drug efflux pump, and cpdA, the regulator of intracellular cAMP levels were found, indicating novel strategies to achieve slow growth ( Table 1). regulation under nutrient deficiency and focus on sigma-factor competition as a genetic mechanism that orchestrate bacterial growth under dynamic stress environments akin to those encountered in the long-term stationary phase.

Bacterial growth under prolonged nutrient limitation
Natural habitats are characterized by long periods of nutrient scarcity interjected by short stints of resource abundance termed as the "feast-and-famine" lifestyle (2,3,39). This metabolic cycle is thought to be the primary selection pressure in bacterial evolution (39). Within their natural niche bacteria spend most of their lifetime dividing with extremely low division rates under limited availability of carbon and nitrogen source, as opposed to extreme responses like the complete lack of growth (dormancy) or exponential growth (12,40). conditions (20,40,42,44).
In this manner, the bacterial cell constantly monitors the nutrient status of its external environment and regulates its growth accordingly. In the following sections, I will discuss some of the mechanistic details that make the aforementioned regulation feasible.
3.1 rpoS-dependent strategies to control growth rate based on the ambient nutrient After bacterial cells exit the lag phase, gene expression is primarily modulated by σ D , promoting rapid growth because of the high ambient nutrient levels. As the population grows, the levels of carbon and nitrogen sources in the media steadily decline, compelling the bacterial cell to scavenge nutrients from the ambient media. This is achieved by activating the cellular "scavenging response" in two parts: (1) changing the permeability of the outer membrane through selective expression of porin channels ompF (higher permeability) or ompC (lower permeability) via the transcriptional regulator ompR, and (2) overexpression of ABC type transporters (mgl, mal, LamB) via an increase in the cAMP concentration (42). This "early stationary phase" response is prevalent at sugar concentrations ranging from 0.3 mM, when the growth rate is 70% of maximum growth rate, to 10 -6 8 156 mM when growth becomes negligible (20,42). When the concentration of the carbon source drops below a threshold of 10 -7 mM, the growth-stress response balance could not be maintained by regulating only the membrane porosity, as most of the porins and transporter proteins are saturated in very low sugar concentration in the media. At this point, the rpoS or σ S -mediated "starvation response" is triggered.
The σ S mediated starvation response induces an array of changes like the compaction of the genome (49), reduction in the permeability of the cell membrane by expressing the low porosity porin ompC (50), increase in the production of osmoprotectants like trehalose by upregulation of osmY and treA (40), and repression of all genes directly promoting growth which helps bacteria to adapt to different stress factors that it endures in the stationary phase (46).
The ability of a bacterial cell to mount the σ S -mediated response is not devoid of cost, especially under prolonged periods of slow growth. This is evident from evolutionary dynamics in two different paradigms-a rapid decline of rpoS WT carrying variants in populations evolving in chemostats under low growth rates (D < 0.3 hr -1 ) (44,51), and selection of rpoS partial mutants in bacteria maintained in long-term stationary phase (52).

Balance between growth and stress response through sigma factor competition
Because of the dynamic nature of nutrient availability and presence of stressors in its environment, a constant balance has to be maintained between increasing biomass for rapid growth and selfpreservation (46). Maintenance of this balance is crucial because over-optimizing rapid growth makes the population sensitive to even minute perturbations in the environment while putting excess weightage on self-preservation/stress-resistance results in the loss of fitness (46). It is known that in vivo, the concentration of the core RNA polymerase is limiting for transcription (46). Across different phases of bacterial growth, competition for the RNA polymerase core ensues between σ D and σ S to initiate transcription from specific promoters by integrating environmental cues (46). expression of genes under the control of the other one due to the competition as mentioned earlier (53). In E. coli, the intracellular levels of σ D have been shown to remain more-or-less constant across growth phases, whereas σ S concentration increases sharply during the onset of stationary phase (54).
It has been observed that strains carrying wild-type rpoS are able to utilize a lower number of carbon sources as compared to partial and null rpoS mutants (20,55). Hence, σ S acts as a 'necessary evil' to the cell presenting a particular "cost" (lower flexibility in terms of carbon utilization) under nutrient-rich conditions, but becomes essential under stress (schematically shown in Fig. 3). It is not surprising that bacteria have evolved multiple regulatory circuits to control the expression and activity of σ S at all levels (comprehensively reviewed by Susan Gottesman and colleagues here: (56)).

Regulatory programs to modulate sigma-factor competition
Multiple independent regulatory circuits have been demonstrated to influence the competition between σ D and σ S , thus regulating the balance between self-preservation and nutritional competence (SPANC) in response to the ambient nutrient and stress status (20,56). Some of these regulators increase the efficiency of σ S to bind to the RNA polymerase core, while others interfere with the binding of σ D with the core polymerase and hence repress its activity. In this review, I will focus on the role of a few global regulators essential for survival under prolonged nutrient limitation.

ppGpp: the master regulator of sigma factor competition
One of the most important reporters of ambient carbon and nitrogen levels that controls global gene expression in bacteria is the small molecule stress alarmone guanosine pentaphosphate or (p)ppGpp (53). The intracellular levels of this modified nucleotide govern the rapid growth versus self-preservation balance. During exponential growth, the concentration of ppGpp in the cell remains low due to the nutritional abundance. During starvation periods, free/stalled ribosomes resulting from a lack of amino-acyl t-RNAs induce the production of ppGpp via the expression of relA and spoT (53). ppGpp functions by destabilizing the DNA-protein interactions leading to the dissociation of σ D from the RNA polymerase holoenzyme. This event increases the availability of core polymerase for σ S to bind. σ S -bound holoenzyme can then transcribe from cognate promoters that encode genes involved in stress response (59). In fact, the expression and activity of σ S are themselves regulated by ppGpp, establishing the latter as the "master regulator" of sigma-factor competition (46,53,75).

Role of Crl and rssB in governing sigma factor competitions
Crl is a global regulator, active during the transition from exponential growth to stationary phase, that positively affects the activity of σ S by facilitating the binding of RpoS with the core RNA polymerase (76) (56,77). In effect, the sigma factor competition between σ S and σ D for the core RNA polymerase is shifted by Crl in favor of σ S -firstly by increasing the expression of a large subset of the σ S regulon under low RpoS concentrations, and secondly by inducing proteolysis of RpoS via increased expression of RssB (69,77). Hence regulation by Crl leads to a lesser but more active σ S protein which indicates that some σ S targets are probably expressed during exponential growth (77).
This regulation has implications in the log-to-stationary transit in batch cultures and during the long-term evolution of starved cultures (77).
6S RNA, a small, non-coding RNA generated from ssrB gene is also instrumental in controlling sigma factor activity (56,78). This RNA binds to the σ D -core polymerase complex by mimicking a promoter and inhibits transcription from several σ D promoters, and also allows σ S to occupy the free core polymerase and transcribe from relevant targets (78,79). The 6S RNA is shown to regulate different sets of genes across various growth phases, indicating that its activity is governed by the ambient nutrient status (79). Another protein Rsd upregulates a subset of σ S regulon by directly Recently, it was shown that in the initial stages of LTSP evolution, mutations in global regulators confer a higher fitness advantage-an effect that declines over time, leading to local regulators being mutated at a higher frequency later in the experiment (88). Evidence and role of GASP have been detected in pathogen populations as well-vector-borne pathogens like X. nematophila balance a trade-off between pathogenicity and the GASP phenotype, with the competitively advantageous lrp mutant subpopulations lacking transmissivity and virulence later in evolution (89).  (34). In another LTSP experiment with E. coli K-12 str ZK819 as the ancestor, rpoABC mutations were found to be widespread among the variants that were sequenced at different timepoints for a month (36). Most of these alleles are found to decrease growth rate in the exponential phase and hence present a cost to the bearing cell (13,91). However, the selection of these mutations across different ancestors and runs of LTSP evolution point to the obvious advantage of slow growth in nutrient-constricted spent media. Indeed, media age has emerged as a major determinant of the competitive fitness of LTSP-adapted mutants (13). Likely, the balance between rapid growth and stress response (SPANC) will be affected by the growth defect acquired by mutations in the rpoABC operon. Diverse strategies to gain fitness under prolonged starvation are developed by the bacterial cell by fine-tuning this SPANC balance (55) (Fig. 3).

Genetic determinants of GASP
Apart from the rpo operon, other genes that are frequently mutated in LTSP adapted variants include the cyclic-AMP associated global regulated CRP, cAMP phosphodiesterase cpdA, and the translational elongation factor fusA (EF-G). All of the above-mentioned genes hold the potential to modulate the SPANC balance via controlling the sigma factor competition.
Apart from the above-mentioned factors, mutations in other sigma factors like σ A in Bacillus subtilis and σ N in E. coli are also reported to alter the SPANC balance, enabling the bacterial cell to respond to environmental stressors (92,93). Victoria Shingler has proposed a theory where the binding between σ D and RNA polymerase core is affected by mutations in σ D , or concomitant players like alternative sigma factors and core RNA polymerase subunits. This increases the availability of free core polymerase to be occupied by other 'non-housekeeping' sigma factors, likely via the involvement of small-molecule regulators like ppGpp (75). Evidence of this theory has been provided by work done in Ruth Hershberg and Seshasayee groups.

The path ahead: Challenges in studying the long-term stationary phase
The long-term evolution of bacteria in chemostats (under constant nutritional status) has been studied for more than 20 years (94). The short generation time of a few species has enabled these seminal studies to uncover general principles about bacterial gene regulation under different growth rates and how it shapes the genome over time (95,96 Similarly, persistor subpopulations that can endure antibiotic regimes and drive chronic diseases are formed by pathogens in nutrient-limited host niches (10,100). Resistance to novel non-exposed stress factors in evolved mutants has been identified in LTSP evolution experiments (13,101). The reason for gaining these novel resistances is not clearly understood currently, but it might be caused by a global up-regulation of the general stress response (GSR) pathway. The translational utility of the LTSP paradigm can be harnessed by evolving pathogens in conditions resembling specific host niches and identifying strategies developed by the strain to gain fitness (89,102).
Bacterial gene regulation in long-term stationary phase is dynamic, complex, and spread across multiple subpopulations in an ever-changing environment. The role of external factors in the evolutionary dynamics in this phase is being increasingly recognized. Although the idea of sustained cultivation of microbes is not new, systematic observation of genomic and transcriptomic changes in this survival phase has only been started in the last 30 years (Fig 4). LTSP being a temporally open-ended paradigm, the amount of information that could be mined via novel, highthroughput multi-omics tools is endless. The natural evolutionary dynamics in biomes like host gut and soi are closely mimicked by the LTSP paradigm since it is free from any perturbation by the experimenter. Further developments against current medical challenges like the evolution of multidrug resistance (18), de-novo stress resistance (101), and phenotypic persistence (100) could be made by exploring the eco-evolutionary dynamics in LTSP.

Ethics approval and consent to participate
Not applicabe for this work

Consent for publication
Not applicable for this work

Availability of data and materials
The whole genome data analysed for the mutations in Table 1

Authors and Contributors
This manuscript was conceived, drafted, and written by PN. All the raw data described in the manuscript was acquired and analysed by PN.

Competing interests
The author reports no conflict of interest

Funding information
The author received a integrated M.Sc-PhD fellowship from NCBS, TIFR from 2014 to 2020-the duration that the evolution experiments were carried out.

Acknowledgements
The author wants to thank Dr. Aswin Sai Narain Seshasayee, Dr. Deepa Agashe, and Dr. Laasya Samhita for providing critical feedback and improvising the manuscript.