Global Gene Expression Analysis of Cross-Protected Phenotype of Pectobacterium atrosepticum

The ability to adapt to adverse conditions permits many bacterial species to be virtually ubiquitous and survive in a variety of ecological niches. This ability is of particular importance for many plant pathogenic bacteria that should be able to exist, except for their host plants, in different environments e.g. soil, water, insect-vectors etc. Under some of these conditions, bacteria encounter absence of nutrients and persist, acquiring new properties related to resistance to a variety of stress factors (cross-protection). Although many studies describe the phenomenon of cross-protection and several regulatory components that induce the formation of resistant cells were elucidated, the global comparison of the physiology of cross-protected phenotype and growing cells has not been performed. In our study, we took advantage of RNA-Seq technology to gain better insights into the physiology of cross-protected cells on the example of a harmful phytopathogen, Pectobacterium atrosepticum (Pba) that causes crop losses all over the world. The success of this bacterium in plant colonization is related to both its virulence potential and ability to persist effectively under various stress conditions (including nutrient deprivation) retaining the ability to infect plants afterwards. In our previous studies, we showed Pba to be advanced in applying different adaptive strategies that led to manifestation of cell resistance to multiple stress factors. In the present study, we determined the period necessary for the formation of cross-protected Pba phenotype under starvation conditions, and compare the transcriptome profiles of non-adapted growing cells and of adapted cells after the cross-protective effect has reached the maximal level. The obtained data were verified using qRT-PCR. Genes that were expressed differentially (DEGs) in two cell types were classified into functional groups and categories using different approaches. As a result, we portrayed physiological features that distinguish cross-protected phenotype from the growing cells.


Introduction
Adaptation is an important property of living organisms that enables them to survive under stress. In higher organisms, the stress-response is divided into several stages: alarm, resistance and exhaustion. The alarm stage is characterized by the violation of homeostasis (shock) and subsequent induction of specific regulatory systems (e.g. mediated by stress hormones-corticosteroids) provoking mobilization of the endogenous resources to restore homeostasis (contra-shock). During the resistance stage, the cell (organism) restores homeostasis and acquires resistance that lasts until resources are exhausted [1,2].
Bacteria also have the ability to resist environmental challenges and stress effects despite their relatively simple organization. This ability is related to functioning of specific regulatory networks that change the metabolism of the cells and make them resistant [3]. Well-known components of these networks are bacterial alarmone (p)ppGpp synthesized by RelA or SpoT and stress-induced sigma factor RpoS [4,5]. These components largely reprogram gene expression that is crucial for survival and formation of so-called cross-protection when one stress factor forms the general resistance to multiple ones. Activation of these stress-related regulatory networks proceeds at the beginning stage of stress response (within a few hours of exposure to stress factors) [6,7]. Although some of the changes in gene expression that occur during the initial stages of bacterial stressresponse were described [8][9][10], the physiological portrait of the cells that acquired resistance and cross-protective phenotype during the resistance stage of stress-response remains largely unknown.
Cross-protection to multiple stressors in various bacterial species is formed in response to some but not all stress factors [11]. Starvation and acid shock evoke the formation of crossprotection to many if not all different environmental and anthropogenic stressors. Only starvation and acid stresses, but not oxidative or chemical challenge, resulted in increased resistance of Escherichia coli and Salmonella typhimurium to heat [12,13]. Starving E. coli cells were shown to acquire multi-resistance against ionizing radiation, heat and acid shocks, osmotic challenge, and antibiotic treatment [13][14][15][16][17].
Persistence under unfavorable conditions is common for many plant pathogenic bacteria since the life cycle of their host plants usually depends on seasonal changes [18]. Therefore, phytopathogens have adapted to survival in various ecological niches, including water, soil, insects, etc [19][20][21]. Representatives of harmful soft-rot Enterobacteriaceace (Pectobacterium and Dickeya species) that cause serious plant damages all over the world are known to be able to survive, besides plants, in soil, clouds, sea water, fresh surface water, ground water, insects, and mollusks [22]. Moreover, pectobacteria may apply alternative types of adaptive reactions depending on various exogenous and endogenous factors [23][24][25][26]. Some of these reactions were shown to result in the formation of resistance to multiple stress factors: oxidative stress, heat shock, and antibiotic treatment [26]; however, the time necessary for the formation of the cross-protected phenotype was not determined. Altogether, pectobacteria represent suitable model microorganisms for studying microbial adaptation.
The aim of the present study was the elucidation of temporal parameters of cross-protective phenotype formation in Pba cultures under stress (starvation) conditions and determination of physiological characteristics that distinguish unstressed growing cells from those that acquired resistance to multiple unfavorable factors as a result of stress-response, in terms of transcriptome profiling using RNA-Seq approach.

Cross-protection effect in starving Pba cultures
Prior to starvation, Pba cells were susceptible to heat shock, oxidative and osmotic stresses: only 0.0005% of the cells were recovered as CFU after exposure to NaCl and no recovery was observed when H 2 O 2 or heat shock were applied (Fig 1). After four hours of starvation, 0.01, 0.09 and 0.02% of initial cells survived after the exposure to heat shock, H 2 O 2 and NaCl, respectively, pointing to the formation of cross-protection phenotype in a minor portion of the population. A sharp increase of cross-protective effect was observed after 24 h of starvation: 1.0 ± 0.4%, 85 ± 12.5% and 86 ± 2.7% of cells were recovered by plating after heat shock, oxidative and osmotic stresses, respectively (Fig 1). The cross-protective effect did not increase further after 24 h of starvation. Therefore, the maximal cross-protection effect that is close to 100% (for oxidative and osmotic stresses) is achieved after 24 h of starvation of Pba cells.

RNA-Seq Data Analysis
To characterize whole transcriptome of Pba cell after the induction of cross-protective effect, cDNA libraries of growing and 24 h starving cells were prepared. A total of 27.4 and 26.2 million reads were obtained from nutrient rich and starvation conditions, respectively (Table 1). About 98% of total reads were successfully mapped to the reference Pba genome sequence using the Bowtie2. In total, approximately 26 million reads were uniquely mapped, with quality scores above Q30.  In total, out of 4626 loci predicted in the genome of Pba (NCBI Reference Sequence: NC_004547.2), 4613 transcripts were detected across analyzed samples. The results of the edgeR analysis indicated that 1648 genes had statistically significant differences in expression levels (FDR <0.05) (S1 Table). Among them, 610 were up-regulated and 1038 were down-regulated after the formation of cross-protected phenotype under starvation conditions (Fig 2 and S1 Table). Besides, 265 additional transcripts were revealed using old genome assembly (NC_004547.1); among them additional 29 down-regulated and 55 upregulated DEGs were found.

Verification of RNA-seq data
To validate RNA-Seq data, the expression levels of 15 up-and 15 down-regulated genes (according to RNA-Seq) under starvation belonging to different pathways were assessed by qPCR analysis. The results showed that qPCR data were in good agreement with those of RNA-Seq (Spearman's correlation coefficient 0.78, P-value 1.76 × 10 −6 ) (Fig 3).

Classification of DEGs into functional gene modules and pathways
To gain better insight into physiological processes that distinguish non-adapted growing Pba cells from those that acquired cross-protected phenotype as a result of starvation induced stress response, the classification of genes into functional gene modules and pathways was performed. Gene ontology (GO) analysis of differentially expressed genes (DEGs) revealed 76 gene categories: 51 up-and 35 down-regulated (ten categories were present among both upand down-regulated) (Fig 4, S2 Table and S3 Table). 330 DEGs were not classified into GO categories.
The summation of the results of automatic analysis using GO and KEGG and manual gene classification using various databases revealed functional gene modules and pathways induced in the cross-protected phenotype (Table 4 and S4 Table). Several defense-related genes, including those for universal stress protein UspB, cell division inhibitor SulA, DNA protection during starvation protein Dps, heat and acid shock proteins DjlA and Ars were up-regulated. Besides, 25 genes participating in antioxidant defense, including putative catalase ECA1216, hydrogenase-4 complex, thioredoxins ECA1267 and trxC were induced in starving Pba cells. A group of the genes related to biofilm formation was up-regulated in cross-protected phenotype (Table 4 and S4 Table). This group includes genes whose products are engaged in the synthesis or transport of capsular polysaccharides and several genes encoding regulators involved in biofilm formation: YbaJ and BssS [28,29], and diguanylate phosphodiesterases (ECA1841, ECA2008, ECA3271, ECA3549) and diguanylate cyclases (ECA3199, ECA3270, ECA3374, ECA3886) engaged in the synthesis of cyclo-di-GMP that was shown to induce biofilm formation in Pba [30].
Many conjugation-related genes were up-regulated in Pba after the formation of cross-protected phenotype under starvation conditions. Among these genes are those located within HAI-7 that was assumed to encode a full conjugative transfer system [27,31], HAI-2 and HAI13 that encode integrative and conjugative elements and putative integrated plasmids participated in conjugation, respectively [32] (S4 Table). Several genes belonging to a group of prokaryotic defense system, including restriction-modification, DND and toxin-antitoxin system were up-regulated under starvation (Table 4 and S4 Table).
Activation of several pathways related to amino acid metabolism occurred in adapted cells. Operon argADECBH for arginine and citrulline biosynthesis was up-regulated during the starvation period. The expression levels of the genes for methionine synthesis from homocysteine (ECA0181 and ECA3126), methionine synthase (ECA1113) as well as metF, which product catalyses the conversion of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate, a cosubstrate for homocysteine remethylation to methionine, were also increased in adapted cells. Almost all genes for methionine salvage cycle were up-regulated under starvation: methylthioribose kinase (ECA3476), methylthioribose-1-phosphate isomerase (ECA3477), acireductone dioxygenase 1 (ECA2986). Transport of methionine was also activated partially in starving bacterial cells: methionine import ATP-binding protein MetN1 (ECA2063), phosphate transport system permease protein (ECA4475), lipoprotein (D-methionine transport system substrate-binding protein, ECA1498) ( Table 4 and S4 Table). Adapted cells had an increased level of transcripts related to chemotaxis. Twenty one of 36 genes for methyl-accepting chemotaxis proteins along with coupling protein CheW, two component kinase CheA, methyltransferase CheR, methylesterase CheB and response regulator CheY, which phosphorylated form is able to interact with flagellar motor were up-regulated under starvation. Genes for stator component of motor (motA and motB) and flagellar hookassociated proteins (flgK, flgL, fliR and fliD) were also up-regulated. Herewith, genes encoding flagellar assembly proteins and genes assigned to bacterial motility (pathways eca02035 and eca02040) were down-regulated. Additionally, gene encoding negative regulator of flagellin synthesis FlgM (ECA1700) was up-regulated (Table 4 and S4 Table).
Many genes related to signaling and signal transduction were differentially expressed in growing cells cross-protected Pba phenotype. Out of 292 genes encoding one-component regulatory systems (according to database MiST 2.2.) 56 and 17 were up-and down-regulated under starvation, respectively. The level of transcripts for genes encoding proteins for biosynthesis of secondary messenger cyclo-di-GMF (diguanylate cyclases and diguanylate phosphodiesterases) was increased in the starving cells. Many genes for transcriptional regulators of different families (GntR, AraC, LysR, TetR) were induced under starvation. Forty and 24 genes related to two-component regulatory systems were up-and down-regulated under starvation, respectively. Some of extracytoplasmic function (ECF) sigma-factors (ECA0813, ECA1540) or their cognate anti-sigma-factors (ECA0233, ECA2086) were induced in adapted cells (Table 4 and S4 Table). Table 4. Modules and pathways of genes expressed differentially in P. atrosepticum during growth in nutrient rich medium and after the formation of cross-protected phenotype under starvation conditions. Manually assigned modules and pathways are marked with stars.    Table). Sixteen KEGG pathways were significantly down-regulated in the cross-protected phenotype (30-80% of KEGG pathway genes): Metabolic pathways (eca01100), Oxidative phosphorylation (eca00190), Microbial metabolism in diverse environments (eca01120), Ribosome (eca03010), Carbon metabolism (eca01200), Fructose and mannose metabolism (eca00051), Bacterial secretion system (eca03070), Inositol phosphate metabolism (eca00562), Glycolysis/ Gluconeogenesis (eca00010), Amino sugar and nucleotide sugar metabolism (eca00520), Citrate cycle (TCA cycle) (eca00020), Biosynthesis of secondary metabolites (eca01110), Sulfur metabolism (eca00920), Methane metabolism (eca00680), Starch and sucrose metabolism (eca00500), Glyoxylate and dicarboxylate metabolism (eca00630) ( Table 3).
Thus, by functional classification of down-regulated genes, the following functional gene modules and pathways were found to be repressed in the cross-protected Pba phenotype: genes related to oxidative phosphorylation, metabolism of carbon and carbohydrates (including TCA cycle), sulfur, nucleotide, lipid, biosynthesis of secondary metabolites, processes of translation and protein modification. Herewith, RNA-Seq did not indicate on the decrease in general level of transcription; however, many genes for transcriptional regulators were among DEGs. Some genes related to amino acid metabolism were down-regulated, however pathways for methionine (S-adenosyl methionine) and arginine were up-regulated (Table 4 and S5  Table).
Most of the genes encoding virulence factors (plant cell wall degrading enzymes as well as other proteins transported to the periplasm via the Sec pathway and secreted proteases) were down-regulated in adapted cells. Virulence genes of cfa-like cluster (encoding synthases of coronafacic acid) located within HAI-2 were also down-regulated. Genes for motility and flagellar assembly proteins were down-regulated in cross-protective phenotype. However, many genes assigned to chemotaxis and flagellar motor were up-regulated (Table 4, S4 Table and S5  Table).
Many genes related to DNA repair (eca03430 -mismatch repair, eca03400 -DNA repair and recombination, eca03440 -homologous recombination) were repressed in adapted cells. The expression level of genes of SOS-regulon did not vary significantly in growing and adapted cells (Table 4 and S5 Table).
Many ABC transport uptake systems were down-regulated under starvation. However, several genes for transporters of metal cation (zinc, iron) and phosphorus and proteins related to amino acid transport were up-regulated under starvation (Table 4, S4 Table and S5 Table). Genes encoding well-known regulators of stress response (RelA, SpoT and RpoS) that are induced in Pba at early stages (4-8 h) of starvation and later their expression decreases to initial level (Petrova et al., 2014) were not among differentially expressed genes (S1 Table).

Discussion
In the present study, we determined the period necessary for the formation of cross-protected phenotype in Pba under starvation conditions and portrayed physiological features of cells that acquired resistance to various stress factors compared to unstressed growing cells applying transcriptome analysis using RNA-Seq approach. Under starvation, Pba, as with many other bacterial species [33][34][35], becomes resistant to a number of stressors. In the present study, we additionally show that in Pba, maximal level of cross-protection is reached within 24 hour of starvation and does not increase further. Within this period, a transition from non-adaptive to adaptive physiological status is achieved and bacteria may persist under stress effect. Transcriptome analysis revealed 610 up-regulated and 1038 down-regulated genes in the cross-protected phenotype versus growing cells. These DEGs were grouped into functional modules and categories by KEGG database, GO enrichment analysis and manual classification using various databases. RNA-Seq results were verified by means of qRT-PCR and the data obtained by two methods showed good correlation.
The formation of cross-protected phenotype is known to require the action of global stressinduced regulators, particularly alarmone (p)ppGpp necessary for the induction of the stringent response and synthesized by RelA or SpoT and stress-induced sigma-factor RpoS [36]. The induction of expression of relA, spoT and rpoS genes under stress conditions in Pba and many other bacteria is known to occur during first few hours after stress exposure; thereafter their expression reaches the initial levels [7,25,37]. Here we show that maximal cross-protective effect does not coincide in time with the highest expression level of relA and rpoS and occurs when transcript level of these genes reaches the degree of unstressed cells [25 and this study]. Thus, the induction of relA, spoT and rpoS may be attributed to the alarm stage of stress response when cells are preparing to become resistant, and this stage precedes and triggers the resistance phase when the cross-protection is achieved. It is reasonable to note that the effects of ppGpp and RpoS (increase membrane stability, synthesis of glucose/trehalose and glycogen, activation of lipolytic activity and oxidation of fatty acids, etc.) resemble the reactions that are induced by animal stress hormones during the alarm stage of the stress response [38][39][40][41][42][43][44][45].
Our work allowed determination of several physiological features reflected in the transcriptome profile inherent to Pba cross-protective phenotype. General metabolic activity was evident to be lower in adapted cells than in growing ones since genes for oxidative phosphorylation, carbon and carbohydrate metabolism (including TCA cycle) and processes related to translation and protein modification were significantly repressed. Some genes related to amino acid metabolism were also down-regulated. However, several amino acid pathways were induced under starvation. Arginine biosyntesis pathway that includes genes for arginine and citruline metabolism was up-regulated in cross-protected phenotype. Additionally, the activation of formation of methionine and its derivatives from homocystein as well as methionine salvage cycle that is a universal pathway recycling sulfur-containing metabolites to methionine, occurred under starvation. It was previously shown, that arginine increased the level of c-di-GMP in Salmonella by modulating the expression of diguanilate cyclase gene that is necessary for the formation of biofilms under starvation condition [46][47]. The increase in citrulline content is known to result in the formation cross-protection in starving bacteria [48]. The role of methionine in bacterial adaptation is unknown. However, it is reasonable to hypothesize that the induction of methionone metabolism, including the methionine cycle, where this amino acid is converted to S-adenosyl methionine (SAM)-a principal biological methyl donor for numerous prokaryotic methyltranaferases, is related to the activation of the methylation processes that are evident from the RNA-Seq analysis as up-regulation of various methyltransferase genes in adapted Pba cells.
The maintenance of the cross-protective phenotype requires the presence of specific defensive proteins that support cell homeostasis under adverse conditions. The RNA-Seq analysis revealed the induction of several protective genes, including those for antioxidant defense. This is in accordance with the up-regulation of oxidative stress regulon, including the gene encoding catalase, in cross-protected E. coli phenotype [49]. Additionally, many genes related to biofilm formation, including those that encode regulators YbaJ and BssS, and diguanylate phosphodiesterases and diguanylate cyclases engaged in the synthesis of cyclo-di-GMP that was shown to induce biofilm formation [30], were activated under starvation, as well as genes related to the synthesis of capsule polysaccharides.
Although virulence genes such as those encoding the cfa-cluster, plant cell wall degrading enzymes and type III secretion system, were repressed in adapted cells, their transcripts were observed under starvation pointing to the "readiness" of bacteria to switch on virulence program fast when it is required. This is in accordance with the preservation of virulence of starving Pba cells shown in our previous study [25].
Cross-protected phenotype was characterized by the up-regulation of chemotaxis-related genes and genes for the flagellar stator motor. However, genes for motility and flagellar assembly proteins were repressed. This likely means that cross-protected cells do not spend exogenous resources for the synthesis of flagella de novo, however, they maintain activity of the flagellar motor and increase the sensitivity to chemoattractants to be able to "feel" the environment. This fact is in accordance with the ultrastructural changes of starving Pba cell: under starvation the volume of the cytoplasm decreases significantly, however, cells seek to maintain sufficient cell volumes and considerable outer surfaces that presumably serve as peculiar cell antennas catching various external signals [25]. This probably may enable bacteria to "escape" the stress-factor. The activation of chemotaxis under starvation was also noted previously [50].
Most of the genes related to DNA repair were either non-differentially expressed or downregulated in adapted cells pointing to the possibility of the increase in mutation process in the starving bacterial population. Genes that are classified into a group of prokaryotic defense systems protecting cell from foreign DNA were differentially expressed in growing and starving cells. These defense systems are divided into two broad groups that differ in their modes of action [51]. The defense systems that function on the self-non-self discrimination principle constitute the first group that includes restriction-modification (RM) system, DND system (which labels DNA by phosphothiolation and destroys unmodified) and CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats)-Cas (CRISPR-associated genes) system-a prokaryotic adaptive immunity system able to memorize encounters with foreign DNA and attack it specifically. Several RM-and DND-related genes were up-regulated in the starving cells. However, the expression of CRISPR-Cas locus was down-regulated. Toxin-antitoxin (TA) systems acting as inducers of programmed cell death or dormancy represent the second group of prokaryotic defense systems. In starving cells, 18 genes related to TA-systems were up-regulated.
Additionally, significant portion of genes within HAI7 where the Type IV secretion system and the rest of the conjugative functions annotated as an "integrated plasmid" are located [27,31], as well as genes of HAI2 and HAI13 encoding conjugation-related genes were up-regulated in adapted cells. Thus, the obtained results show that although there were several up-regulated genes which products may protect cells from the foreign DNA, the whole transcriptome portray shows that in adapted cells, the background for mutational process, including DNA exchange through conjugation and existence of some foreign DNA, is formed. This likely enables bacteria to acquire new traits that may increase their survival under stress conditions. It is obvious that the acquirement of cross-protection and the ability to maintain homeostasis at "novel level" requires switching on/off various regulatory systems. The RNA-Seq analysis revealed the regulatory systems that may coordinate physiological processes in adapted cells. These systems include multiple one-and two-component regulatory systems, transcriptional factors, sigma and anti-sigma factors and genes which products are involved in the synthesis of cyclo-di-GMP-a regulatory signal that is involved in particular in stress reactions of microorganisms. Additionally, in our recent study various small RNAs that are induced under starvation and potentially have regulatory effect on the formation of cross-protective phenotype were unraveled [52].

Summary and outlook
Cross-protected cell phenotype of Pba formed under stress (starvation) conditions at the resistance stage of stress response is characterized by multiple physiological features reflected in the trascriptome profile that distinguish it from unstressed cells. The described alterations in the expression of gene modules and pathways likely enable cross-protected cells to maintain their resistance under prolonged stress effect. Stress response is known to be expressed in two types of adaptive strategies: "fight"-to resist stress factor, or "flight" to escape the stress factor. The RNA-Seq analysis of growing and cross-protected Pba cells indicated on several physiological parameters that may contribute to maintain the ability of adapted Pba cells to both "fight" the stress factor (e.g. expression of defense genes, decrease of general metabolism, increased mutational background, biofilm formation) and to "flight" to better life (e.g. increase in chemotaxis, the presence of transcripts of virulence genes). Herewith, proteins (or other metabolites) that might have been synthesized during the alarm stage of the stress response and preserved in the cells may also play roles in determination of the physiology of cross-protected Pba phenotype, though the encoding genes were not described as up-regulated in the cross-protected Pba phenotype in our study. Additionally, transcriptome analysis enabled determination of potential regulators involved in the promotion and/or maintenance of cross-protective features under adverse conditions.
For future research work on the described phenomenon, it would be informative to elucidate general and species-specific aspects of stress-induced resistance by comparing transcriptomes of cross-protected cells of bacteria belonging to different ecological groups. Additionally, since bacteria are able to resist stressors applying alternative adaptive reactions, many of which lead to the formation of cross-protection [25,26], it would be interesting to compare transcriptome profiles of cross-protected phenotypes formed in the course of realization of different adaptive strategies. Valuable information may be also obtained by analyzing stress-induced transcriptomes of mutant bacteria that are unable to form cross-protection phenotypes under stress conditions (e.g. rpoS-, relA-deficient mutants). In addition, the analysis of mutants deficient in regulatory genes that were shown to be induced in the cross-protected Pba phenotype in our study, would allow verification of the roles played by these regulators in cell resistance to multiple stressors.

Materials and Methods
Bacterial strains, media and culture conditions A strain of P. atrosepticum SCRI1043 (Bell et al., 2004), was used in this study. The cultures with inoculation titer of 2-3 × 10 6 CFU (colony forming units) per ml were grown in Luria-Bertani medium [53], with aeration (200 r.p.m.) at 28˚C for 16 h (growth-promoting conditions). Aliquots of these cultures were used for total RNA extraction. The remaining cells were transferred (after double wash) to carbon and phosphorus deficient AB medium containing 1 g/L NH 4 Cl; 0.62 g/L MgSO 4 × 7H 2 O; 0.15 g/L KCl; 0.013 g/L CaCl 2 × 2H 2 O, pH 7.5 and incubated under starvation conditions with initial population density of 5.4 × 10 8 ± 6.1 × 10 7 CFU per ml in glass vials without aeration at 28˚C.

Cross-protection assay
To assess the dynamics of the cross-protective effect of starvation stress (primary stress), the tolerance of starving (or non-starving) Pba cells toward heat shock, hydrogen peroxide and NaCl (secondary stress) was examined. For secondary stress challenge, 1 ml aliquots of cell suspensions incubated under starvation for 0, 4, 8, 24, and 48 h were subjected to 50˚C for 5 min or supplemented with 2.5 mM H 2 O 2 or 20% NaCl for 1 hour. Before and right after exposure to one of the secondary stress factors (heat shock, or H 2 O 2 , or NaCl), suspensions were plated onto 1.5% LB agar as serial 10-fold dilutions. Plates were incubated at 28˚C for two days and the CFUs were counted. The survival of starving cell was assessed by the comparison of cell titer prior and after secondary stress factor exposure.

Total RNA preparation
Total RNA was isolated from bacterial cells using the RNeasy Protect Bacteria Mini Kit (Qiagen, USA), according to the manufacturer's instructions. Contaminating DNA was removed from the samples by DNAse (Qiagen) treatment. RNA was quantified using a Qubit fluorometer (Invitrogen, USA).
cDNA library construction and bacteria strand-specific RNA sequencing Library construction and strand-specific sequencing were carried out at the Beijing Genomics Institute (BGI-Shenzhen, China; http://www.genomics.cn/en/index), following the manufacturer's protocols. Briefly, the rRNA was depleted from 1 microgram of total RNA using the Ribo-Zero Magnetic Gold Kit (Epicenter). TruSeq RNA Sample Prep Kit v2 (Illumina) was used for library construction. RNA was fragmented into small pieces using Elute Prime Fragment Mix. First-strand cDNA was synthesized with First Strand Master Mix and Super Script II (Invitrogen) reverse transcription (25˚C for 10 min; 42˚C for 50 min; 70˚C for 15 min). After product purification (Agencourt RNAClean XP Beads, AGENCOURT) the secondstrand cDNA library was synthesized using Second Strand Master Mix and dATP, dGTP, dCTP, dUTP mix (1 hour at 16˚C). Purified fragmented cDNA was end repaired (30 min at 30˚C) and purified with AMPureXP Beads (AGENCOURT). Addition of the poly (A) tail was done with A-tailing Mix (30 min at 37˚C) prior to ligating sequencing adapters (10 min at 30˚C). The second-strand cDNA was degraded using the Uracil-N-Glycosylase (UNG) enzyme (10 min at 37˚C) and the product purified by AMPureXP Beads (AGENCOURT). Several rounds of PCR amplification with PCR Primer Cocktail were performed to enrich the cDNA fragments and the PCR products were purified with AMPureXP Beads (AGENCOURT). Sequencing was performed using the Illumina HiSeq TM 2000 platform with pair-end 90 base reads. The sequencing data used in this study can be accessed from NCBI's Gene Expression Omnibus, under the accession number GSE68547.

RNA-seq data analysis
Strand-specific RNA-seq was used to compare whole transcriptomes of Pba cells under nutrient-rich and starvation conditions. Paired-end read quality control, trimming, read mapping and indexing of resulting BAM files, were performed as reported in Kwenda et al [52]. Gene quantification was conducted using bedtools multicov [54], to make counts of reads mapped to Pba protein-coding regions. In order to determine differentially expressed genes, edgeR package was used [55]. A false discovery rate (FDR) of 5% was used as cut-off for significantly differentially expressed genes.  (Bell et al., 2004). Gene ontology categories were visualized using REVIGO http://revigo.irb.hr/.
Total RNA was extracted from Pba cells and digested with DNAze as described above and used for cDNA synthesis. The reaction mixture for reverse transcription contained 1 mM random hexamers, 300 mM dNTPs, 1x reverse transcriptase buffer, 100 U RevertAid reverse transcriptase (Thermo Scientific, USA). First, water and random hexamers were mixed with RNA and incubated for 5 min at 70˚C and cooled immediately on ice. The other components were then added. Incubation was performed using a DNA Engine thermocycler (Bio-Rad, Hercules, CA, USA). Reverse transcription was performed as follows: 10 min at 25˚C, followed by 1 h at 37˚C and 10 min at 70˚C. Two μl of a dilution (1/5) of cDNA were used as the template for qPCR.
qRT-PCR was performed using the 2.5x kit containing Taq-polymerase and corresponding buffer, dNTPs and Eva Green dye (Synthol, Russia). Primers (S6 Table) designed according to the genome sequence of P. atrosepticum SCRI1043 (NCBI Gen-Bank accession number NC_004547) using Vector-NTI Version 9 software (Invitrogen, Carlsbad, CA, USA) and synthesized by Evrogen (Moscow, Russia) were added to 0.4 μM. PCR was performed under the following conditions: 94˚C for 2 min, followed by 45 cycles at 94˚C for 10 s, 60˚C for 15 s and 72˚C for 30 s. After PCR, the melt curve analysis (60-95˚C) was performed. The reactions were run and changes in fluorescence emission were detected using a CFX96 quantitative PCR system (Bio-Rad, USA). The amount of fluorescence was plotted as a function of the PCR cycle using CFX Manager Software (Bio-Rad, USA). The amplification efficiency for all primers was determined using a dilution series of a pool of cDNAs. Additional controls included the omission of reverse transcriptase to measure the extent of residual genomic DNA contamination and omission of template.
The expression level of target genes was calculated relative to reference genes. The geNorm software was used to choose genes that displayed stability of expression under the experimental conditions. Among the candidate reference genes (groES, recA, rpoD, ffh and gyrB), the latter three had the most stable expression levels (M value-0.862, 0.757 and 0.652, respectively; Pairwise Variations-0.256). These genes were also among stably expressed according to RNA-Sed data. Therefore, rpoD, ffh and gyrB genes were used as reference ones. Relative expression levels were determined as the ratio between the quantity of cDNA corresponding to the target and reference gene transcripts. The data are presented as the average ± SD of five biological replicates. The correlation between the data obtained by RNA-Seq and qPCR was assessed by Spearman test.
Supporting Information S1 Table. Relative expression levels of genes in adapted P. atrosepticum cells vs. growing ones.