Lethal perturbation of an Escherichia coli regulatory network is triggered by a restriction-modification system's regulator and can be mitigated by excision of the cryptic prophage Rac

Abstract Bacterial gene regulatory networks orchestrate responses to environmental challenges. Horizontal gene transfer can bring in genes with regulatory potential, such as new transcription factors (TFs), and this can disrupt existing networks. Serious regulatory perturbations may even result in cell death. Here, we show the impact on Escherichia coli of importing a promiscuous TF that has adventitious transcriptional effects within the cryptic Rac prophage. A cascade of regulatory network perturbations occurred on a global level. The TF, a C regulatory protein, normally controls a Type II restriction-modification system, but in E. coli K-12 interferes with expression of the RacR repressor gene, resulting in de-repression of the normally-silent Rac ydaT gene. YdaT is a prophage-encoded TF with pleiotropic effects on E. coli physiology. In turn, YdaT alters expression of a variety of bacterial regulons normally controlled by the RcsA TF, resulting in deficient lipopolysaccharide biosynthesis and cell division. At the same time, insufficient RacR repressor results in Rac DNA excision, halting Rac gene expression due to loss of the replication-defective Rac prophage. Overall, Rac induction appears to counteract the lethal toxicity of YdaT. We show here that E. coli rewires its regulatory network, so as to minimize the adverse regulatory effects of the imported C TF. This complex set of interactions may reflect the ability of bacteria to protect themselves by having robust mechanisms to maintain their regulatory networks, and/or suggest that regulatory C proteins from mobile operons are under selection to manipulate their host's regulatory networks for their own benefit.


Introduction
Prokaryotic genomes are highly plastic and can exploit horizontal gene transfer, point mutations, and gene rearrange-ments to adapt to environmental changes and other challenges.Being tolerant of constant gene influx is selected for, in order to preserve fitness in the absence of obligate sexual recombination ( 1 ).However, acquiring genetic modules that often include regulators, such as transcription factors (TFs), can pose challenges, as a new TF must not only exert its specific function, but also avoid disrupting the host's existing regulatory networks upon modification.Bacterial TFs collectively build up and coordinate flexible regulatory circuits ( 2 ,3 ).However, significant perturbations by a new TF of the existing regulatory orchestration and interconnectivity may result in reduced fitness and even cell death.Perhaps this is why regulatory networks are subject to continuous rewiring and fine-tuning to optimize functionality ( 4 ).Still, we poorly understand the impact of exogenous TFs on horizontal gene transfer and gene regulatory network robustness, though mechanisms for temporary silencing of imported genes appears to be part of the general solution (5)(6)(7).
Another aspect of gene flux dynamics is associated with integration of bacteriophages into bacterial genomes, which then co-evolve with their hosts as prophages (8)(9)(10)(11).Often prophage DNA remains as 'foreign islands' with other horizontally accepted gene fragments, resulting in host genomic mosaicism.Prophages often seem to be beneficial for the host, improving its survival, persistence, metabolic strategies and prevalence in microbial communities, yet still more needs to be explored ( 12 ,13 ).Genomic analysis of diverse bacterial species has revealed multiple prophages present within each genome in the form of 'cryptic' prophages, unable to proceed through a lytic cycle due to mutations accumulated during phage-host co-existence.The vast majority of such defects is related to process of phage DNA excision, virion assembly or host cell lysis ( 9 ).Still, even cryptic phage DNA undergoes positive genetic selection to maintain those of its genes that drive bacterial adaptation and fitness, whereas other genes are lost ( 10 , 14 , 15 ).Among characterized E. coli strains, E. coli O157:H7 str Sakai has the largest known number of prophages, harboring 18 that constitute almost 16% of the total genome ( 16 ).Well-studied E. coli K-12 has nine cryptic prophages, with DLP12, e14, Rac, CPZ-55, and Qin prophages as the best characterized ( 9 ,10 ).
In recent reports, we monitored the transfer of an operon into a new host, in real time to understand the fate of horizontally transferred DNA fragments.As a model, we used the Csp231I Type II restriction-modification (R-M) system ( 17 ).In this particular operon, the two enzymes comprising R-Msrestriction endonuclease (REase) and DNA methyltransferase (MTase)-are precisely controlled at the transcriptional level by a dedicated TF, called a C protein ( 18 ), part of a large family of such proteins ( 19 ,20 ).The C protein not only activates and represses (depending on its concentration) the REase gene and its own as an autoregulator, but also acts as a temporal regulator during RM mobility .Specifically , C protein delays REase expression so the MTase has time to completely modify the genome, to prevent cell death due to REase cutting ( 17 , 21 , 22 ).
During our studies, we noticed that this R-M system's C protein (C.Csp231I) also has some unpredicted side-effects, which initially were observed as inducing a toxic effect reflected in a cell morphology phenotype.We studied this phenomenon in detail using a combination of genetics, biochemistry and transcriptomics, and found that the C regulator C.Csp231I changed global E. coli K-12 genetic networks.In particular, we established a link between C protein binding within the Rac prophage, and cell toxicity manifested by profound filamentation (Figure 1 A, B) ( 23 ,24 ).Our results indicated that the C protein was engaged in transcriptional cross-talk, via off-target DNA binding, rather than by action at its native site (target C-box) within its own promoter.By cross-talk, we mean here the adventitious binding of a TF with an unrelated (off-target, non-cognate) DNA site, sometimes called TF promiscuity ( 25 ).
This observation brought our attention to the racR-ydaS-ydaT locus, coding for three TFs, and its intergenic region, where the RacR master repressor competes with YdaS to bind their operators -if RacR wins this competition ydaS and ydaT expression is silenced and lysogeny is maintained ( 24 , 26 , 27 ).However, when C protein appears, RacR gene expression is derepressed, YdaS and YdaT are produced, and cell filamentation is triggered leading to cell death if these conditions are prolonged (Figure 1 A) ( 23 ,24 ).Still, the basis for this unleashed toxicity was not known, so we designed experimental approaches to find the E. coli genomic target site(s) for YdaS and YdaT.As we reported before, no toxicity is detectable in an E. coli rac variant ( 23 ).We hypothesized that the coexistence of C protein gene and Rac locus with RacR repressor gene creates a genetic conflict.Hence, selective pressure may lead to emergence of inactivating mutations in one component out of these two ( 23 ), which we confirm in this study.
Our main objective here was to generate E. coli mutant libraries, in the presence and absence of C.Csp231I-RacR conflict, to find genes involved in the toxic pathway initiated by C protein.Our results support an analogy of RacR repressor to λCI, where both proteins act as lysogeny maintenance sentinels.We also found that reduction of RacR expression is critical for Rac excision from genome and eventual loss of the prophage DNA.Rac induction appears to counteract the lethal toxicity of YdaT.Cells can apparently survive the relatively short 'toxic' period, until they are cured of Rac prophage.Overall, our studies on Rac prophage biology show an interesting example of a flexible strategy, in which bacteria use potent TFs to maintain beneficial prophage genes, while silencing those likely to be deleterious.

Bacterial strains, plasmids, bacteriophages and oligonucleotides
The bacterial strains, phages and plasmids used in this study are listed in Supplementary Table S1 ; the oligonucleotides are in Supplementary Table S2 of the Supplementary Materials.

Generation of Tn 5 mutant libraries
The plasmid pRL27 ( 28 ), which carries the transposon Tn 5 with a kanamycin (Kan) resistance gene, was used to transfer the transposon into MG1655 rac + cells carrying the plasmid pBAD-CWT, with the C.Csp231I regulatory protein expressed under arabinose induction.A second library was prepared in MG1655 rac , with pBAD-ydaT expressing the ydaT gene under arabinose induction.Creation of the mutant libraries was done by conjugating the strains BW20767 and MG1655 which served as a donor and a recipient of plasmid pRL27, respectively .Briefly , 50 μl of log phase cell samples were washed away from antibiotics and inoculated simultaneously in one spot on an LB agar plate supplemented with 0.1% arabinose.After 16 h of incubation at 37 • C, the whole spot was transferred to 500 μl LB, serially diluted and spread on LB agar plates supplemented with Kan, chloramphenicol (Cam) and 0.1% arabinose.Following overnight incubation formed colonies were visually inspected for the desired phenotype with a stereo microscope (Olympus).

Transposon insertion localization
The selected Tn 5 integrants were grown in LB medium supplemented with Kan, Cam and 0.1% arabinose.Chromosomal DNA was extracted according to the manufacturer's recommendations (Roche) and digested with a restriction enzyme which does not cut within the transposon sequence of pRL27 ( Supplementary Table S1 ).In about a quarter to onethird of the cases digestion with one enzyme (EcoRV) was sufficient to reveal products in subsequent PCR reactions, while in others a combination of two enzymes out of EcoRV, BamHI, NdeI, NheI, NcoI, ScaI was required.The resulting fragments were cleaned (A&A Biotechnology, Poland), and self-ligated using the T4 ligase (Eurx, Gdansk, Poland) at 16 • C overnight.The obtained DNA fragments were again cleaned and inverse-PCR amplified by using tpnRL17-1 and tpnRL13-2 ( Supplementary Table S2 ), which anneal to positions within the transposon sequence in pRL27 and read outwards into flanking DNA regions.To avoid amplification of non-specific fragments initial annealing temperatures were set to 74 • C and were gradually reduced as specified in Supplementary Table S3 and Supplementary Table S4 .Obtained DNA fragments were analyzed on 1% agarose gels to confirm random insertion.The bands that differed from the rest were cut from the gel and cleaned up (bands appearing at the same height in every path were omitted).The site of insertion of the transposon was determined by sequencing (Genomed, Poland) with the tpnRL17-1 primer.The BLAST (NCBI) and EcoCyc was used for genes identification ( 29 ).All library integrants are presented in Table 1 , and Supplementary Table S5 of Supplementary Materials.Strains, for which the above-described procedure was not sufficient to identify transposon insertion site, were whole genome sequenced (DNA Sequencing and Oligonucleotide Synthesis Laboratory IBB PAS, Warsaw, Poland).

Complementation of the Tn 5 mutation
To confirm that the obtained defect in cell morphology is due to C protein / YdaT expression and it is blocked by inactivation of certain genes selected from Tn 5 library, these selected genes were cloned as the intact genes under an inducible promoter.Cloning vector pHM1786 contains the lacI q repressor gene, and the IPTG (isopropyl β-D-1-thiogalactopyranoside) inducible P tac promoter ( Supplementary Table S1 ).Both the vector and the PCR-amplified insert were digested with HindIII and EcoRI, ligated and transferred to the cloning strain.Positively verified recombinant plasmids were used in the complementation assay.The assay included two compatible plasmids: one with the intact selected gene (pHM1786 derivative), and the second with an inducible form of the gene for either C.Csp231I (pBAD-CWT) or YdaT (pBAD-ydaT).These two plasmid pairs (and their empty vectors as control) were introduced into the Tn 5 insertional mutant strains under antibiotic selection and gene induction (IPTG, arabinose).The fresh transformant cells were examined under a microscope for the filamentous phenotype.

Chromosomal gene knock-outs
The knockout strains were constructed using the lambda-red recombination method with a pSIM5 plasmid carrying the recombineering proteins, Gam, Exo and Beta ( 30 ,31 ), and using pKD46 as a template plasmid for ampicillin resistance cassette amplification.The constructed strains and primers used are listed in Supplementary Table S1 and Supplementary Table S2 (Supplementary Materials).

Assay for excision of the Rac prophage
Three types of MG1655 rac + strain were tested with bla cassette (ampicillin (Amp) resistance gene) inserted into the rac locus: within ydaS , ydaT or ralRA .Each strain was transformed with pBAD-CWT or pBAD-Cmut (non-DNA-binding C protein variant, described in ( 18 )).The ydaT expression (pBAD-ydaT plasmid) or ydaS expression (pBAD-ydaS) was also tested for effects on Rac excision.Transformants were subcultured every 10 generations in the presence of 0.1% arabinose over a period of five consecutive days, and samples were spread onto Cam LB-agar plates (without Amp or arabinose).A hundred colonies per strain were streaked in parallel on both LB-agar and LB-agar Amp plates.After overnight incubation colonies were inspected for the selective lack of growth in the presence of Amp.Colonies were counted in duplicate and data presented in percentage of total colonies tested.In addition, colonies were also examined for the filamentous phenotype.

Quantification of prophage excision
E. coli MG1655 rac + cells, carrying the plasmid with C protein WT gene (or Cmut, unable to bind DNA ( 18 )) under control of the inducible P BAD promoter, were grown overnight in LB medium with glucose and appropriate antibiotics.After dilution, the two cultures were grown in LB-glucose to an OD 600nm of 0.4, then cells were gently pelleted, washed, and split into three replicate cultures.Pre-induction samples were taken (generation 1), then the LB was supplemented with 0.1% arabinose and cultures were left for continuous subculturing under the induced conditions (samples were taken up to generation 50).
The prophage excision triggered by C protein expression was quantified by PCR (qPCR).Two sets of primers were selected.One pair (ttcAqPCR, intRqPCR; Supplementary Table S2 ) flanked the region between the Rac left attachment site ( ttcA gene in chromosome) and the first gene within the Rac prophage ( intR ) (amplicon 216 bp) (schematic location of primers; Supplementary Figure S1 ).The second set of primers amplified part of the ydaS and ydaT genes (primers ydaS-pLEX3Bfor, ydaTpLEX3Brev, 183 bp amplicon).For normalization, two alternative housekeeping genes were used: idnT (gluconate transporter; idnTfor, idnTrev, 200 bp amplicon; Supplementary Table S2 ), and zntB (Zn 2+ / H + symporter, 2865 bp from the ttcA start; primers zntBqPCRfor, zntBqPCRrev, 129 bp amplicon).Total DNA was isolated using a High Pure DNA isolation kit (Roche) and was used as the template for the qPCR reaction using the SG qPCR Master Mix (Eurx, Poland).The qPCR conditions were as follows: predenaturation step 95 • C, 5 min and 30 cycles of 95 • C for 10 s, annealing 56 • C for 10 s and elongation at 72 • C for 10 s.The reaction and analysis was performed using the Roche Light Cycler.Each reaction was performed in biological triplicates and repeated at least twice independently.Data were averaged and standard deviation was calculated.Melting curve analysis was used to confirm the formation of the specific products.The calculations also included the PCR efficiences and its C t was plotted against DNA input to calculate the slope corresponding to PCR efficiency obtained with high linearity ( R 2 > 0.97) (32)(33)(34).
Determination of transcript levels by q-RT-PCR One ml of exponentially grown E. coli MG1655 rac + cells on LB-arabinose medium, harboring plasmids carrying: the inducible C gene (pBAD-CWT), the inducible RacR gene (pBAD-RacR) or empty vector (pBAD33), were harvested.The bacterial pellet was resuspended in 1 ml of StayRNA reagent to prevent RNA degradation (A&A Biotechnology).The total cellular RNA was then extracted using the Total RNA Mini Plus kit (A&A Biotechnology) according to the manufacturer's instructions.After elution, the RNA was treated with DNaseI (A&A Biotechnology) for 60 min at 37 • C following enzyme inactivation for 15 min at 65 • C. The RNA was then used to synthesize first-strand cDNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific).Several sets of primers ( Supplementary Table S2 ) were used to estimate transcript levels in qPCR reactions (Roche Light Cycler 480), specific to Rac genes ( intR , racR , ydaQ ) with idnT being used as the reference gene.Each qPCR reaction (25 μl) contained 12.5 μl SG qPCR Master Mix (2 ×) with SYBR Green I fluorescent dye, Perpetual Taq DNA polymerase and dNTPs (Eurx Poland), 7.75 μl H 2 O, 0.25 μl of UNG (Uracil-N-glycosylase), 1 μl of 10 μM forward and reverse primers mix and 2.5 μl of diluted cDNA as a template.The qPCR conditions were as follows: UNG pre-treatment 50 • C for 2 min, pre-denaturation step in 95 • C for 10 min and 35 cycles of 94 • C for 15 s, annealing 55 • C for 30 s and polymerization at 72 • C for 30 s and final acquisition 80 • C for 15 s.The expected sizes of products were between 171 and 200 bp.Each reaction was performed in biological triplicates and repeated at least twice independently.Data were averaged ( ±SD) and Student's t -test was calculated.Melting curve analysis was used to confirm the formation of the specific products.The calculations also included the PCR efficiencies, where each cDNA was serially diluted and its Ct was plotted against cDNA input to calculate the slope corresponding to PCR efficiency obtained with high linearity ( R 2 > 0.98) (32)(33)(34).

Sequencing and genome analysis
Genomic DNA was extracted from cells using Genomic Midi AX (A&A Biotechnology) according to the manufacturer's instructions.DNA was ethanol-precipitated and resuspended in TE buffer.DNA was separated in agarose gel to assess its high quality (no observed degradation).Samples were diluted to achieve A 260 / A 280 ratio in the range 1.8-2.0 with concentrations above 10 ng / μl.
For whole-genome sequencing, paired-end whole-genome next-generation sequencing (NGS) was performed on Illumina Sequencing PE150 (Novagene, UK), with read lengths of 150 bp.The raw reads were subjected to quality control with FastQC and due to overall good quality were not processed further.Next, reads were aligned to the reference genome E. coli K-12 MG1655 (GenBank assembly accession GCA_000005845.2) using BWA alignment software, version 0.7.17-r1188 ( 35 ).Mapping quality was assessed using QualiMap version 2.2.1 ( 36 ).The latter allowed identification of regions with larger genomic interruptions, that were further verified manually and visualized using JBrowse2 ( 37 ).Data were deposited in the NCBI (bioproject accession number PRJNA993445).
An RNA-seq library was prepared according to the TruSeq RNA Sample Preparation, version 2 Guide (Illumina, San Diego, C A, US A) and sequencing of the libraries was performed using an Illumina HiSeq2500 platform at Macrogen.The results have been published ( 23 ) and raw data deposited in the NCBI GEO (accession number GSE126248).

The ydaT and ydaS toxicity assay
An overnight culture of E. coli BL21(DE3) carrying two compatible plasmids, p51ydaT (IPTG-inducible ydaT expression) and pBAD-ydaS (arabinose-inducible ydaS expression), was inoculated into LB containing 100 μg / ml Amp and 34 μg / ml Cam, grown to early log phase (OD 600nm ∼0.2), followed by induction of gene expression with either 1 mM IPTG, 0.1% l -arabinose, or both 1 mM IPTG and 0.1% L-arabinose.Control cultures remained uninduced, treated with glucose.Bacteria were further grown with shaking for additional 3 h, and the samples were taken at indicated time points and after serial dilution spotted on LA containing antibiotics and appropriate inducer.Other E. coli strains were tested at identical way.

Overproduction, purification and size-exclusion chromatography of the RacR repressor
For the C-terminally-6His-tagged RacR overproduction and purification were performed as reported in Supplementary Materials in ( 24 ).Oligomerization status was tested by size exclusion chromatography on a Superdex 75 10 / 300 GL column using the ÄKTA Pure 25 system (GE Healthcare).The predicted size for monomer of 6His-tagged RacR repressor is 18.5 kDa.The 0.5 mg protein sample was loaded onto the column equilibrated with 50 mM potassium phosphate buffer (pH 7.0), 150 mM NaCl, and eluted at a flow rate of 0.5 ml / min in the same buffer.The column was calibrated with proteins of known molecular masses: alcohol dehydrogenase (tetramer), 146.8 kDa; bovine serum albumin, 66 kDa; ovalbumin, 43 kDa; trypsin inhibitor, 22 kDa; and cytochrome C, 12.4 kDa (Sigma-Aldrich, St. Louis, MO, USA).

Bacteriophage spot assay
To determine the efficiency of phage lytic productivity, virulent bacteriophage samples (P1 vir and λ vir ) were used.Log phase BW25113 rac + bacteria cells carrying the pBAD-YdaT or pBAD33 were 0.1% arabinose induced for 2 h, and 30 μl samples were mixed with 10 μl phage samples, serially diluted in TM+ buffer (10 mM MgSO 4 , 10 mM TRIS and 0.01 mM CaCl 2 ).After 10 min incubation time at 37 • C, successive dilutions were spot plated (5 μl) using LB-agar with Cam and incubated overnight at 37 • C. Bacterial colony forming units (CFU) without phage addition were checked as control.
The files related to the bioinformatics protein modelling are accessible at RePOD repository: https:// doi.org/ 10.18150/ FV9R0P

Statistical analysis
Unless otherwise indicated, we performed statistical analysis and presented plot data as the averages and standard deviations.For column plots, data were analyzed using GraphPad Prism 8. Student's t test was performed to calculate a p value for the difference between related pairings (*** P < 0.001; * P < 0.05 and ns: P > 0.05).

Screening and analysis of the Tn 5 insertion mutant library
In order to identify the E. coli genetic loci participating in the C.Csp231I protein-dependent cell filamentous phenotype (presumably mediated by the YdaS and YdaT gene products) (Figure 1 ), we used an unbiased, functional approach.We generated a Tn 5 transposon random insertion library of E. coli MG1655 cells, employing a vector carrying a mini-Tn 5 transposon (Kan R ) and a modified hyperactive version of the gene coding for the Tn 5 transposase (pRL27) ( 28 ).In this library, after induction of C protein expression all colonies display a distinct flat, abnormal curly morphology (Figure 1 ).In theory, if a gene responsible for the C-dependent phenotype was interrupted by Tn 5 -leading to loss of the filamentation then supplying a WT copy of that gene should restore the filamentous phenotype.This morphological difference enables us to easily distinguish standard colonies among the majority of morphologically-abnormal colonies under a magnifying glass, or even by eye alone.
The representative E. coli MG1655 Tn 5 insertional mutant library was created on agar plates with Kan (selecting for Tn 5 ), Cam (selecting for the csp231IC plasmid), and arabinose (to induce C protein production).Inspection of the plated library revealed the expected defective colony morphology in the great majority of colonies (Figure 2 A).The prepared library, as two technical replicates, each consisted of about 8 × 10 5 integrants, which were screened thoroughly to find 'regular', opaque colonies, as shown by red arrow in Figure 2 A. Such Tn 5 mutants were isolated, restreaked in sectors, and examined under a microscope to ensure that they were filament-free.Forty-one mutants were isolated and frozen for further studies.In total, we found 108 colonies with normal morphology, among 1.6 × 10 6 colonies screened, for a hit rate of 6.75 × 10 −5 (the total ratio of the number of normal to filamentous colonies).Filamentous cells were on average between 15 to 25 μm in length and consisted of several segments with visible DNA inside ( 23 ).Such cells formed the majority of defective colonies, which seemed to have flawed septum formation during cell division, as indicated by scattered distribution of FtsW protein, which assists in Z-ring assembly ( 49 ,50 ) (white arrows on Figure 2 B using FtsW::GFP, pDSW360).
Next, the locations of Tn 5 insertions within the selected group of mutants was determined by Tn 5 subcloning and inverse-PCR ( 28 ), as described in Methods, Supplementary Figure S2 , and Supplementary Tables S3 , S4 .This approach worked well for most mutants, however for a few of them we performed whole genome sequencing using E. coli MG1655 as the reference strain control.The site of Tn 5 integration was unique for each mutant isolated (Table 1 , Supplementary Table S5 and Figure 3 ).DNA sequence analysis showed that Tn insertion had no particular hot-spots.Nevertheless, some genes were disrupted in several distinct positions (Figure 3 ), such as wcaD (colonic acid polymerase), c y aA (adenylate cyclase) and arpA (a regulator of acetyl CoA synthetase).Five insertions were within the waa operon, involved in assembly of the core region of the lipopolysaccharide, which stabilize the cell's outer membrane (Table 1 , Figure 3 ).There were also some insertions within TF genes, such as rcsA , which controls a phosphorelay system; aaeR, which is a LysR-type regulator; and yjhL, which regulates phage-like genes.A few genes within the cryptic CP4-5 prophage region were also disrupted: fadE , phoE , yagE , yagK , rclB and yahN (Figure 3 ).Many identified interrupted genes are involved in sugar metabolism or transport.For some strains with interrupted genes such as w aaG , w aaQ and rirA , the cells revealed a mucoid, shiny colony phenotype.
To confirm that the filamentous and colony phenotypes of cells depends on C protein expression in our system, and to rule out polar effects, we performed complementation assays.Introducing the respective intact gene in trans on a plasmid should revert the phenotype from normal to 'sick' colony morphology, in the strain where the tested gene is interrupted by Tn 5 .Thus, we cloned selected genes (from Table 1 ) under an inducible promoter.We used the following genes: arnB, fadE, nlpA, rcsA, waaQ, wcaD, yagE and yagK.For all of them, co-expression with C protein gene in the genetic context of Tn 5 -inactivated genes did not change the cell morphology significantly.So, the complementation assay was not successful under our conditions.The reasons for this are described in the remainder of this section and the following one.
We next tested whether the revealed mutants are C protein responsive.Hence, we made a directed gene knock-out for two genes: waaQ (involved in lipopolysaccharide synthesis ( 51 ,52 ), which appeared twice in the library) and rcsA (a multifunctional transcriptional activator involved in biofilm production, cell division, motility, swarming and other func-tions ( 53 ,54 )).Next, we introduced the plasmid specifying C.Csp231I and found out that these cells remained normal and failed to filament despite C protein overexpression ( Supplementary Figures S3 , S4 ).This result indicated that the library had indeed yielded mutants (at least waaQ and rcsA ) associated with resistance to C protein overexpression, even though the complementation results are unclear.
In addition, we knew that induction of C protein directly affects the Rac genes based on our transcriptomic data analyzed previously ( 23 ,24 ).Accordingly, we expected some Tn 5 insertions in the Rac region itself, but surprisingly there were none among the library mutants we characterized (Figures 1 B  and 3 , Table 1 ).This fact made us address the issue of Rac presence in the library genomes.Thus, we first decided to test some Tn 5 mutants for the presence of the Rac racR-ydaS-ydaT region, using PCR amplification with lysed frozen cells samples as a reaction template.In most cases, we obtained very weak or undetectable amplification products (Table 1 and Supplementary Figure S5 ).We concluded that deletion within the Rac region might have occurred, or possibly prophage excision.In addition, we analyzed the whole-genome sequence of selected mutant DNAs, and observed the larger deletion within Rac locus.

Insufficient RacR repressor level is critical for rac excision
To further verify the role of Rac prophage genes, and especially of the racR-ydaS-ydaT operon (Figure 1 B), in the Cprotein-responsive phenotype, we tested whether cells lose the filamentous phenotype spontaneously after long exposure to C protein expression.We subcultured MG1655 rac + cells carrying the p24 plasmid (with csp231IC under its natural promoter) for many generations, by daily passaging, as we had done earlier ( 23 ).About 75 cell generations produced almost 90% colonies with round and regular shapes, losing the initial phenotype (Figures 1 , 2 B and 4 A).We took nine such colonies and one control colony (prior to p24 plasmid introduction), isolated their genomes and subjected all 10 to whole genome sequencing.We expected to find some suppressing mutations, which neutralized the YdaS / YdaT toxicity.Indeed, we found one common mutation among all nine tested genomes, that was not present in the control genome (Figure 4 B).We de- tected three base changes within the 5 end of ttcA (involved in post-transcriptional tRNA thiolation).Significantly, this locus is the specific integration / excision site for the Rac prophage ( 55 ).Close inspection of this region revealed identical 23060 bp deletions of the Rac prophage in all nine tested genomes (Figure 4 B).The analysis revealed the same specific point mutations as had been reported previously, during studies on Rac excision in E. coli K-12 BW25113 cells after the overproduction of the Rac excisionase ( ydaQ = xisR ) and integrase ( intR ) proteins ( 55 ).Thus, both assays, utilizing our Tn 5 library, as well as the spontaneous mutants, suggest a common selective pressure, favoring Rac prophage excision to block prophage gene expression, apparently including the ydaS / ydaT genes.This finding explains why our complementation assays did not succeed -after prolonged cell growth with C overexpression, Rac prophage undergoes induction, and is subsequently lost, along with expression of the potentially toxic genes.This cell propagation might have happened as recently as during preparation of competent cells from Tn 5 mutants for the complementation assays.Our observation would also explain the accumulation of morphologically normal colonies, as these strains had lost the Rac prophage (including ydaS / ydaT ).In addition, it is possible that some of the Tn 5 integrants had lost Rac before Tn 5 integrated, and were actually false mutants, that played no role in the filament forming phenotype, though this is clearly not the case for waaQ or rcsA (see above, and Supplementary Figures S3 and S4 ).

Kinetics of rac excision in the presence of C protein over-expression
Our previous whole-transcriptome study ( 23 ) showed that several genes in the Rac prophage are significantly activated in C protein-expressing cells, as compared to C-absent cells.The Rac genes showing increased expression included: ydaS , ydaC , kilR, ralA, ydaF, ydaG, ydaE and racC (ranging from 40-to 70-fold change) (Figure 1 B).Hence we did not expect to see Rac prophage excision and circularization.
In light of our studies here with the Tn libraries, it is significant that certain stresses result in Rac DNA excision ( 55 ).We sought to find out how quickly this process occurs.To monitor the Rac excision over time, we first used a PCR approach, growing cells with the plasmid carrying the C protein gene under the inducible P BAD promoter (pBAD-CWT).We observed that Rac induction did not take place within 15 h after adding arabinose.Thus we decided to monitor this process over many cell generations.In the first approach, we used the two genomic MG1655 mutants: ydaS::bla and ydaT::bla, where the respective Rac genes ydaS and ydaT were interrupted with a bla cassette (Amp R ).Specifically, these strains grow on Amp plates only when Rac is stably maintained in the genome, while Rac excision would eventually lead to loss of resistance, as the cells would not express Rac genes, including the bla cassette.Next, we introduced plasmid pBAD-CWT with the C protein gene.The resulting transformants were induced with arabinose for continuous C protein expression, and sub-cultured every ∼10 generations for up to 5 days in liquid culture.Single colonies were isolated and screened on Amp plates.Overall, the results showed that phage excision is not a rapid process, as during the first 10 generations the number of Amp-resistant colonies hardly changed (Figure 5 A).However, after that a sharp decline in number of such cells is observed, reaching nearly no Amp-resistant colonies at generation 30.These results are consistent with earlier data showing lower viability of cells with C protein expressed only in rac + strain, but not in rac ( 23 ).In addition, this observation explains also why our whole-transcriptome analysis have shown Rac genes' expression ( 23 ), as cell samples for RNA-seq were taken from fresh transformants, within the generations, where Rac excision was not detected.
The implied excision and loss of Rac DNA during C protein expression prompted us to investigate this process in a more sensitive manner, using qPCR analysis.We used cells with C protein (CWT), or as a negative control its mutated inactive variant (Cmut = A33G; R34E; Q37A, which is unable to bind its specific DNA; pBADCmut ( 18 )) (Figure 5 B-D).We employed two pairs of primers: (i) flanking the Rac attachment site, within genomic ttcA and Rac intR genes (Figure 5 B, C, Supplementary Figure S1 ); and (ii) in the center of Rac, within ydaS -ydaT (Figure 5 D) using a primer pair within the distant host gene idnT for normalization (Figure 1 B; Figure 3 ).Both assays revealed a similar time course for the relative Rac DNA levels, with a surprising initial increase up to 9-fold within the first 10 cell generations, which was not seen in the context of the mutated C protein (Figure 5 BD).After 20 generations passed, the Rac excision was highly visible, as the DNA amplification product was weak (reduced by 8-fold).The control with Cmut remained stable throughout the entire course of the experiment.Next, we changed the normalization gene for the same DNA samples, from idnT to zntB , which is located much closer to the attachment site, though still on the genomic flank ( ∼2000 bp from ttcA start).We saw complete loss of the apparent increase of Rac copies (Figure 5 C).The initial difference of Rac copies compared to genome copies is not clear, as the attachment site and Rac DNA circle have been characterized before.Still, this observation was consistent and repeatable, using different pairs of primers within Rac.We tried also to monitor the Rac induced DNA circles by PCR, but with no success ( Supplementary Figure S6 ).It is not known how deficient the Rac replication process is, but other lambdoid phages use rolling circle replication to produce one long concatameric molecule with many copies of the phage genome ( 56 ), and this may explain our observed peak of Rac copies (Figure 5 B, D).

RacR acts as a sentinel for Rac prophage maintenance
To study further the direct effect of RacR as a master regulator of Rac induction, we analyzed the mRNA levels of two crucial proteins responsible for prophage excision: Rac excisionase (YdaQ = XisR) and integrase (IntR) ( 55 ).These genes are adjacent, and lie down next to the rac attachment site, but far away from RacR repressor (Figure 1 B).We aimed to measure their transcript levels in response to a gradient of RacR repressor concentrations: from high expression ( in trans from a pBAD-racR plasmid), through the medium natural genomic level (chromosomal Rac plus empty vector pBAD33), to the low level where RacR expression is repressed by C protein (from pBAD-CWT).Control experiments have shown that C gene overexpression inhibited genomic racR expression by over 100-fold, while arabinose induction of racR expression from a plasmid was over 1000-fold (Figure 6 A).In the presence of these three RacR levels, the relative level of xisR and intR increased about 100-fold only when the RacR level was reduced relative to its natural level (Figure 6 B).In contrast, the natural RacR level was sufficient to maintain optimal xisR and intR levels, and increased racR expression did not change their expression.
Previously, we hypothesized that RacR repressor may function in analogous manner as CI repressor of λ phage.We had determined RacR recognition sites within the racR upstream region, and found four inverted repeats as potential operators ( 24 ).Now, we tested whether purified RacR protein can form oligomers as λCI does.Using the size-exclusion chromatography approach, with co-chromatographed molecular size markers, we confirmed that RacR forms octamers, which is consistent with number of recognition sites and bioinformatics predictions (Figure 7 A, B).It seems likely that each of the four inverted repeats is bound by a RacR dimer.The 3D structure prediction reveals that the RacR protein consists two distinct domains (Figure 7 C and Supplementary Figure S7 ).The N-terminal domain (NTD) is responsible for DNA binding.Despite limited sequence identity with the CI repressor (NP_415874.1 versus P03034.2), the NTD demonstrates a similar 3D fold, composed of five helices, with the α3-helix (residues 38-47) interacting with the major groove of DNA (Figure 7 C and Supplementary Figure S8 ).Notably, while the C-terminal domain (CTD) of RacR and the CI repressor exhibit entirely different 3D folds (Figure 7 D), they serve a common purpose: facilitating oligomerization.This suggests that the RacR repressor behaves analogously to the CI repressor of the phage.

YdaT, but not YdaS, activity reduces cell viability, and YdaT has no effect on Rac excision
The Rac excision results obtained using E. coli strains MG1655 ydaS::bla or ydaT::bla (Figure 5 A; with inactive ydaS or ydaT genes respectively) made us re-consider the ydaST effects on cell viability.We cloned ydaS under the inducible P BAD promoter, and ydaT under the P T7 promoter, in separate, compatible plasmids.Initially, we used E. coli strain BL21(DE3) and found that only ydaT expression caused a severe detrimental growth effect, and not expression of ydaS, as observed via cell samples spotted onto agar plates supplemented with appropriate inducers (Figure 8 A).Further, inducing YdaS together with YdaT did not show any additive effect, so they are unlikely to be co-regulators.We also tested ydaT induction in strains MG1655 rac + and Δrac (Figure 8 B), and it was clear that the toxicity comes solely from YdaT activity, participating in an unknown pathway outside of the rac locus.Interestingly, E. coli B strain, ER2566, a derivative of BL21(DE3), showed complete loss of YdaT toxicity, indicating a possible lack of certain genetic mediators of filamentation (Figure 8 B).
We also wanted to confirm that induction of ydaT or ydaS will not trigger Rac excision, so we repeated the Rac excision assay.Whether measuring cell growth or relevant qPCR amplification products (Figure 8 C, D), both approaches showed that increased expression of ydaS induces Rac prophage excision (Figure 8 D), but not ydaT (Figure 8 C).

Generation of Tn 5 library in rac genetic context under ydaT overexpression
To find an effector gene for YdaT outside the rac locus, we again created a Tn 5 library, but this time in a rac context.We used a plasmid carrying ydaT linked the P BAD promoter, and then followed exactly the same protocol as before.We obtained a representative library in technical duplicates.This time, the screen of colonies on arabinose-LB agar plates did not reveal as many mutants with changed colony morphology as before.In fact, majority of regular colonies could not be passaged, having obvious viability problems.We managed to isolate just six Tn 5 -integration mutants.We identified three disrupted genes: three occurrences in rcsA , two in c y aA , and one in pitA .The rcsA and c y aA genes were among Tn 5 integrants determined before (Table 1 , Figure 3 ).The pitA gene specifies low-affinity inorganic phosphate transport (Pit), and is the major uptake system for phosphate under conditions of P i abundance ( 57 ).Of these three, we chose to examine the involvement of rcsA more closely.

Toxic YdaT effect is alleviated when the rcsA gene is inactive
We repeated the complementation assay for rcsA gene, which is a transcription factor affecting various bacterial pathways including cell division.We generated an E. coli rac + rcsA::bla strain and transformed with two compatible plasmids, respectively carrying inducible rcsA and ydaT genes ( Supplementary Table S1 ), with empty vectors as controls.The assay showed that MG1655 rac + has 69-fold reduced viability when ydaT is overexpressed, but only 12-fold lower if the strain was rcsAdeficient (Figure 9 A).We also confirmed that only combined overexpression of rcsA and ydaT genes in the rcsA::bla background could restore the altered phenotype from normal bacterial cells back to filamentous ones (Figure 9 B).This is strong evidence that the RcsA TF plays a significant (though partial) role in the YdaT-initiated pathway leading to a cell division defect.In addition, we noticed that E. coli cells overproducing RcsA have a distinct colony morphology, with shiny, mucoid characteristics (Figure 9 C).
To determine whether YdaT affects rcsA expression directly, we fused the rcsA gene (under its native promoter) to the lacZ reporter gene, and induced ydaT expression in trans (Figure 9 D).Clearly, higher ydaT levels led to increased rscA expression, by almost five fold, indicating that YdaT is a gene regulator that activates rcsA , whether directly or indirectly.In addition, some reports have shown RcsA affects lipopolysacharide biosynthesis via interaction with RcsB ( 53 ).We wondered whether increased levels of RcsA / RcsB changes the LPS, using protection against phages that recognize LPS receptors as a biomarker for LPS alteration.P1 vir bacteriophage recognizes the LPS core oligosaccharide of E. coli K-12 as a receptor ( 58 ).λ vir phage served as a negative control, as its LamB receptor protein should not be affected by LPS changes ( 59 ), though LPS alteration can reduce somewhat amounts of LamB on the cell surface ( 60 ).λ vir is insensitive to repression by CI, so λ-lysogenic hosts are still susceptible.The assay, in which serially diluted phages and a constant amount of bacterial cells were spotted onto LB-agar plates, showed protection against P1 vir only in cells with elevated levels of YdaT (Figure 9 E).The same cells were still comparable in plaquing efficiency when λ vir phage was tested, irrespective of YdaT levels.Thus, it appears that the receptors recognized by phage are ( E ) Bacteriophage test on cells under ydaT overexpression: P1 vir (left) and λ vir (right).Serially diluted phages and the same amount of cells were spotted onto LB-agar plates.For λ vir the efficiency of lysis is unchanged upon ydaT induction (in comparison to the empty vector (V) control).For P1 vir recognition receptor is a terminal glucose moiety of the lipopolysaccharide (LPS) for which synthesis is disturbed during ydaT induction.
lost or changed when YdaT levels are elevated, as reflected by the loss of P1 phage efficiency of plaquing.More experiments are needed to verify whether P1 resistance is related to the change in the cell envelope or phage inability to reach its receptor.

Discussion
Cryptic Rac prophage uses a regulatory scheme similar to that of the Lambda phage immunity region Our initial objective in this report was to find a function for YdaT protein, and assign the possible mechanism of its toxicity.However, our experimental approach led us to consider the broader phenomenon of Rac prophage maintenance versus excision.The Rac immunity region is located between racR and ydaST , and was previously investigated and noted to be similar to the immunity regions of other lambdoid prophages in terms of gene organization, operator locations, and gene regulation ( 10 , 16 , 24 , 26 , 61 , 62 ).Such regions typically impose the lysis vs .lysogeny decision.Our bioinformatic analyses have uncovered an intriguing discovery: while RacR and CI repressor exhibit low sequence similarity, they share a structurally similar N-terminal domain responsible for DNA binding.Remarkably, although the C-terminal domains of these proteins differ significantly in structure, they both serve as the foundation for oligomerization.Bioinformatics predictions, along with size-exclusion chromatography supports octamers in active form.Both repressors ensure that prophage DNA is maintained in the genome, by acting within the lysis / lysogeny decision region (between the CI and Cro genes in the case of bacteriophage λ) ( 63 ).RacR has multiple binding sites, comprising four inverted repeats between racR and ydaS , which cover the racR promoter ( 24 ).In this context, in Rac, the functional analog of the λCro protein (an antagonist of λCI) seems to be played by YdaS, as its increased expression leads to Rac excision, comparable to λCro protein, which switches λ phage from lysogeny into the lytic phase ( 64 ).However, unlike in active Lambda phage, Rac induction does not cause the cell lysis and plaques are not formed.
In a similar way, we also tested the proposal that YdaT acts as a λCII functional analog ( 61 ).In bacteriophage λ, CII activates several promoters during the switch to lytic growth ( 65 ,66 ).Under our conditions in E. coli MG1655, YdaT has no effect on Rac excision.Perhaps in this defective prophage excision is solely dependent on RacR concentration without any other TF involvement needed.Rac excision is not as rapid a process as λ phage excision (switch of cycle from lysogeny to lysis), which in λ occurs within a 2-3 h time window after Cro overexpression or CI proteolysis in rich media ( 67 ).In our case, Rac excision (under RacR deficit), takes generations, not hours, and never ends with cell lysis.In addition, interestingly, the YdaT-like protein of the lambdoid prophage CP-933P does not have any toxic effects ( 61 ), whereas Rac YdaT yields serious viability problems, depending on the host genetic context (Figure 8 ).Thus again, Rac prophage regulation and the gene functions it encodes are distinct from other Rac-like regions described so far, and require further study to reveal its peculiarities.

Rac excision and increased YdaT levels are coupled
The RacR regulon has not yet been characterized in detail, and there are several genes without determined functions, such as the racR -proximal operons ydaFG and ydaUVW (Figure 1 B).The RacR regulon was previously reported as being essential to E. coli ( 26 , 27 , 68 ).Due to variation in Rac genes content across E. coli strains ( 26 ), the RacR regulon might also vary.Quite often, gene manipulation within the RacR regulon triggers a lethal effect manifested by growth defects ( 26 , 55 , 69 , 70 ).The Keio collection of E. coli single gene deletion mutants does not contain a racR mutant ( 71 ) , consistent with our observation that such mutation is possible only in ydaST background.The gene for YdaT is completely silenced under normal conditions by RacR repressor ( 24 ,26 ).Previously, YdaS / YdaT action was linked indirectly to inhibition of cell division by acting on DNA replication or chromosome segregation ( 72 ).YdaT is not only a TF acting in cis within Rac (work in progress), but also has target sites outside of the Rac region.Thus, YdaT-mediated toxicity may be a diverse process, affecting several target genes.Recently, a DNA target site for the YdaT ortholog from the E. coli CP-933P lambdoid prophage was determined, which may help to identify other putative genomic sites of YdaT interaction ( 73 ).No toxic effect has been shown for this ortholog, but there is only ∼30% amino acid identity between the YdaT orthologs from Rac and CP-933P, thus their functions might differ to some extent.
Our transcriptomic data revealed that, under decreased RacR expression, expression of certain Rac genes increases substantially ( 23 ).This should help to identify members of the RacR regulon, as well as other gene associations ( 24 ).The global regulator H-NS can also trigger Rac prophage excision ( 74 ).Nevertheless, under conditions of RacR deficit, caused in this study by C protein action, two processes are initiated: (i) RacR regulon upregulation, including YdaT and (ii) Rac excision due to rise of xisR and intR expression.We hypothesize that these two pathways, although functionally independent, are coupled.As YdaT toxic activity accumulates in cells (possible routes described below), it seems that Rac excision occurs in parallel.The Rac prophage is defective for replication, so its circular DNA might be maintained in cells for several generations and subsequently be lost, which also possibly is what halts Rac gene expression.
Overall, we propose that the Rac excision acts as a countermeasure to cell lethality due to YdaT action.Cells can apparently survive the relatively short 'toxic' period, until they are cured of Rac prophage.We previously observed that the filamentation process is reversible with long-term culturing ( 23 ), but now we believe that the cells either die or lose the Rac DNA to survive.This may represent an excellent example of how bacteria use different regulatory strategies involving master regulators (here RacR) to maintain beneficial prophage genes, while silencing those likely to be deleterious.Non-cryptic (intact) prophages are likely to kill the cell upon induction of the lytic cycle, so there should be a strong evolutionary selection for mutations leading to inactivation of prophage lytic capabilities ( 14 ,75 ).

YdaT-dependent deficiencies in lipopolysaccharide biosynthesis and cell division
We confirmed that foreign TF entry into a cell (C.Csp231I in this case) may perturb existing regulatory networks in a chain reaction.We showed previously that C protein directly inhibits RacR repression of its regulon ( 23 ,24 ), which boosts expression of, in particular, two more TFs: YdaS and YdaT.Here we tried to define the further steps in the molecular mechanisms of the toxic effect initiated by YdaT.We used Tn 5 library analysis, transcriptomic data, and other reports (Figure 10 , Supplementary Figures S9 , S10 ).Some missing links were proposed and will be tested in the future.Unexpectedly, it seems YdaT may not only act in cis , in its gene's upstream region, but also at other effector sites in the E. coli genome.The transposon mutagenesis screening identified at least one effector gene, again a TF, RcsA, of the two-component regulatory Rcs system.The Rcs system can sense cell envelope damage or defects, and regulate gene expression to counteract this stress ( 53 ).In particular, RcsA alone is an autoactivator ( 76), but forms a heterodimer with another partner TF, RcsB, that can activate not only the rcsA gene, but other genes as well, including the wza gene of the exopolysaccharide and colanic acid biosynthesis ( cps ) operon in enteric bacteria ( 77 ,78 ) (Figure 10 ).This is in accord with our observation that some mutants have mucoid colonies, as is seen in rcsA overexpression (Figure 9 C) or in lon − cells (RcsA proteolysis absent) ( 54 ,79 ).Expression of cps operon genes is also upregulated in our C.Csp231I protein-producing cells ( Supplementary Figure S10 ), supporting our hypothesis that YdaT might mediate a regulatory cascade leading from C protein to the cps operon (Figure 10 ).A gene involved in colanic acid synthesis ( wcaD ) was also detected in our library (Table 1 , Supplementary Figure S9 ).The genetic complexity gets higher if we add global stress regulators to the model, such as the sigma factor RpoE, which initiates regulon expression in response to extracytoplasmic stress ( 80 ).The rpoE promoter is positively regulated by RcsAB in response to defects in LPS core biosynthesis ( 81 ).E. coli in conditions of cell envelope stress or morphology defects cannot regenerate its normal rod-shape unless RcsB activity is present ( 82 ).In addition, RcsAB represses motility via negative control of flhDC expression ( 83 ,84 ), which is also observed in our transcriptomic data (strongly decreased expression of flhD and flhC genes, Supplementary Figure S10 ).We clearly have shown a positive effect of YdaT on the promoter activity of rcsA , whether direct or indirect, and more detailed work is in progress (Figure 9 D).
We noticed in our C protein -responsive Tn 5 library a large representation of genes involved in LPS biosynthesis (six insertion within the waa operon and RirA regulatory RNA, Table 1 ).In our transcriptomic analysis with C-expressed cells, at least 16 genes involved in LPS biosynthesis exhibited reduced expression, including wbbI , wbbJ and wbbK ( Supplementary Figure S10 ).Hence we infer that YdaT expression significantly changes a lipopolysaccharide composition.This is consistent with our observation that YdaTexpressing cells were resistant to P1 vir phage, for which the terminal glucose moiety of the lipopolysaccharide is the recognition receptor (Figure 9 E).In addition, production of mucoid layer by stimulation of cps operon might mask some phage receptors, thus preventing the cell against phage lysis.Cell conversion to becoming phage resistant is usually associated with losing a part of the outer core sugars, having certain modifications in the LPS or masking a surface phage receptors ( 59 ,85 ), usually as an effective tool in the bacteria arms race (86)(87)(88).Although, Rac prophage being cryptic cannot fully compete with other active phages, still some remaining genes, like ydaT might be also considered as a putative functional superinfection exclusion system helping Rac to avoid its host lysis by other active phages.More addressed approaches are needed to verify this highly speculative hypothesis.
We also hypothesize that YdaT may indirectly inhibit the expression of transcription elongation factor rfaH (Figure 10 ).RfaH is a specialized antiterminator, required for expression of genes that encode LPS core, capsule biosynthesis enzymes, toxin-antitoxin systems and some dedicated secretion systems in Enterobacteriaceae ( 89 ).The small regulatory RNA, RirA binds to RfaH to inhibit its effect so that LPS assembly (by waa operon) is expressed in a balanced manner ( 81 ,90 ).The regulatory links between RfaH and RcsA may involve the global regulator H-NS (Figure 10 ).Indeed, in our transcriptome data, expression of hns was decreased almost five-fold in C proteinexpressing cells, and a Tn 5 insertion (isolated in the presence of Rac) was found within the rirA regulatory RNA gene (Table 1 ).H-NS functions almost exclusively as a transcriptional repressor, and its activity may be opposed by RfaH, so they may function as silencer / countersilencer pair ( 91 ).In cells lacking hns , pathways leading to LPS and cell envelope synthesis were overexpressed ( 92 ).In turn, H-NS and Lon together have the highest negative effect on RcsA, H-NS by inhibiting the rcsA promoter, and Lon by proteolysis of RcsA itself ( 54 ).Thus, it seems the effect of YdaT on rcsA expression might also be mediated by H-NS, RfaH and RirA (Figure 10 ).This needs to be verified experimentally.
In parallel, RcsAB may modulate cell division by directly activating the ftsZ promoter.FtsZ gene expression depends on several upstream promoters ( 93 ), and FtsZ is an essential cell division protein responsible for septum formation ( 94 ).FtsZ polymerizes into a dynamic ring (Z-ring) that defines the division spot and recruits other proteins to drive localized peptidoglycan synthesis ( 95 ).We observed in C protein / YdaTexpressing cells, a defect in Z-ring formation within cell filaments (Figure 2 A).The Rac protein, KilR is a potent inhibitor of FtsZ ( 96 ), but in our case the same filaments are formed in the absence of kilR ( 23 ).
Our schematic model of the cross-talking TF cascade in E. coli (Figure 10 ) shows some triggered pathways that might be verified experimentally.It is also challenging, amid this complexity, to separate the primary genetic signal from the secondary outputs or feedbacks.Of note, YdaT might stand as an interesting prophage-encoded example of a TF with pleiotropic effects on E. coli physiology.Another prophageorigin TF , AppY , was identified that triggered regulatory pathways increasing bacterial survival under low pH conditions, and also causing biofilm formation and decreased motility ( 97 ).Both studies provide molecular insights into prophageencoded TF integration into the E. coli regulatory network.The comprehensive picture of bacterial physiology is inextricably linked with prophages, with benefits for both parties ( 98 ).Phage integration into a bacterial chromosome is highlyeffective way to be replicated along with the rest of the bacterial genome ( 99 ).Bacteria can benefit in many ways including: phage immunity, antibiotic resistance, virulence factors, motility, new nutrient metabolism, biofilm formation, and overall a high level of heterogeneity ( 13 , 55 , 74 , 100-104 ).These collectively contribute to more efficient and competitive behavior of bacteria to thrive in various microhabitats and ecological niches.However, as shown here, prophages can also complicate horizontal gene transfer by interfering with existing regulatory networks.

Figure 1 .
Figure 1.A dv entitious transcriptional cross-t alk occurs bet ween t w o transcription f actors: C protein regulating the Csp231I restriction-modification system that might be acquired by horizontal gene transfer, and RacR repressor controlling Rac prophage genes.Under ph y siological conditions, the RacR repressor binds within the Rac intergenic region, blocking the possible common promoter / operator region for ydaS and ydaT .As a result, ydaS and ydaT are completely silenced.When a DNA fragment carrying the C protein gene enters the cell, both transcription factors compete for binding to the same region.C protein interferes with RacR repressor binding, so y daS T gene expression is not sufficiently blocked.When YdaS and YdaT are produced, the likely toxicity of YdaT triggers the cell division defect (or others) and cell filamentation occurs, along with an aberrant colony morphology ( A ). Part of genetic map of rac locus is presented with at tac hment site (attL) within genomic ttcA gene ( B ).

Figure 2 .
Figure 2. Screening the Tn 5 insertional mutant library.The vast majority of colonies showed an abnormal morphology, with a flat center and irregular, sw arming-lik e boundary, once C protein expression was induced.The objective was to find normal colonies (shown by red arrow), where the 'sick' colony phenotype was suppressed.( A ) A representative image is shown, acquired by Olympus stereo microscope.( B ) To visualize the Z-ring prior cell division from 'sick' colony, cells with two plasmids carrying inducible C protein and fusion protein FtsW::GFP were grown under arabinose and IPTG induction.A single filamentous cell containing se v eral randomly distributed spots of FtsW::GFP (white arrows).

Figure 3 .
Figure 3. Genomic map of E. coli MG1655 showing identified genes interrupted with Tn 5 transposon insertions.Rac prophage and CP4-6 prophage regions indicated by red sectors.The positions are taken from the reference genome E. coli MG1655 (GenBank U0 0 096), and are also shown in Table1.Underlined are genes used in the complementation assay.For two genes: waaQ and rcsA , a directed knock-out was made (indicated by **).Genes used for normalization of DNA level: zntB and idnT are marked in red.Origin of replication (oriC) is also indicated in box.

Figure 4 .
Figure 4. Rac prophage e x cision from the E. coli K-12 MG1655 genome is triggered by C protein e xpression.( A ) T he nucleotide sequence of the left and right at tac hment sites of Rac within the host t tcA gene before and af ter R ac induction.Excision of R ac into a DNA circle also results in changes of amino-acid residues in the TtcA protein (Q6E, T8S), as shown.Prolonged rac + cells exposure for C protein expression induced a selective pressure for rac + cells to lose the filamentation phenotype.Initially all cells were filamentous and formed uniformly small colonies, but after passaging for ∼75 generations only ∼10% of cells had this phenotype.The remaining 90% of cells spontaneously changed to producing large colonies containing normal, rod-shaped cells.( B ) Nine such colonies were grown, their chromosomal DNA was isolated, sequenced, and compared to the genome sequence from generation 1.All nine genomic sequences under comparison re v ealed identical deletion of the Rac prophage from the at tac hment site, as presented in the top sequence for rac .Control genomes contained the entire WT rac locus.

Figure 5 .Figure 6 .Figure 7 .
Figure 5. Kinetics of Rac prophage e x cision from E. coli K-12 genome.(A ) Phage e x cision w as measured as cell gro wth on Amp-supplemented LB-agar plates.Two derivatives of E. coli MG1655 rac + were tested, where an Amp-resistance cassette ( bla ) was inserted into one of two Rac genes, yielding ydaS::bla or ydaT::bla .These cells, carrying a plasmid with the C protein gene csp231IC controlled by the inducible promoter P BAD , were incubated in glucose-LB (generation 1).A small fraction was gently pelleted, washed and supplemented with 0.1% arabinose-LB for continuous sub-culturing (up to generation 50).At each indicated time, isolated colonies were picked and streaked onto LB-agar plates with and without Amp in duplicate, and the % of resistant colonies was determined.(B-D) Relative levels of Rac DNA after induction of C protein were also determined using qPCR (see Methods) on isolated total DNA from the same assay described in (A).However, E. coli MG1655 rac + cells were used with induced C protein WT or its non-DNA binding mutant, Cmut, as a negative control.( B ) First we used the DNA segment for amplification that spans the Rac at tac hment site, using primers matching genes from both sides: ttcA on genomic side and intR on the Rac side.For normalization of DNA amounts, the housekeeping idnT gene was used.( C ) The same samples were analyzed using different normalization primers, within the zntB gene located much closer to the Rac attachment site than idnT .( D ) Additional measurements were also taken for a DNA segment amplified from the middle of Rac, within ydaS / ydaT , again using idnT for normalization.Statistical analysis were performed as indicated in Methods, each reaction was performed in biological triplicates.Averages and standard deviations are shown.Comparisons between column bars for CWT for 1st versus 50th generation number indicate significant difference, with a P value = 0.0025 (panel B); P < 0.0001 (panel C); P = 0.0042 (panel D).

Figure 8 .
Figure 8. YdaT, but not YdaS e x erts cell toxicity.Only YdaS affects Rac excision.( A ) Effect of overexpressing ydaT, ydaS or both genes on growth of E. coli cells.Culture of E. coli BL21(DE3) carrying two compatible plasmids: p51ydaT (IPTG-inducible ydaT expression) and pBAD-ydaS (arabinose-inducible y daS e xpression).T hese w ere induced with 1 mM IPTG and / or 0.1% L-arabinose, and control cultures remained uninduced, treated with glucose only.Samples were collected at indicated times after induction, and spotted onto LB-agar supplemented with Amp, Cam and the appropriate inducer.( B ) The same approach was used to compare the growth reduction in selected E. coli strains.CFU's (colony-forming units) ratio (CFU ydaT / CFU vec ) of arabinose-induced cells carrying plasmid with ydaT gene (pBAD-ydaT) versus cells with empty vector (pBAD33) was measured in the indicated strains: E. coli K-12: MG1655 rac + and MG1655 Δrac ; E. coli B: BL21(DE3) and ER2556.Averages and standard deviations are shown.( C ) YdaT does not affect Rac e x cision.YdaT effect w as measured as a cell surviv al ratio in an e xperiment identical to that described in Figure 5 A, where strain MG1655 y daT::bla w as e xposed on Amp o v er a range of YdaT concentrations.( D ) T he same e xperiment w as conducted f or YdaS.Rac e x cision w as measured in E. coli MG1655 rac + under YdaT induction, using qPCR under identical conditions, as noted in Figure 5 B with housekeeping idnT gene as a reference gene.Statistical analysis for q-RT-PCR in panel C, D were performed as indicated in Methods, each reaction was performed in biological triplicates.Averages and standard deviations are shown.For all panels Student's t -test was performed with values: *** P < 0.0 0 01; * P < 0.05 and ns: P > 0.05.

Figure 9 .
Figure 9.The RcsA transcription factor complements the toxic YdaT phenotype, suggesting their possible common interaction within the pathway leading to lipopolysaccharide biosynthesis.( A ) Growth of MG1655 rac + cells when ydaT expression is low.The effect is alleviated when rscA is inactive (MG1655 rac + rcsA :: bla ).The averages ( ± SD) are shown with Student's t-test: P *** P < 0.0 0 01; * P < 0.05.( B ) In contrast, only overexpression of two genes at the same time: ydaT and rcsA in rcsA::bla background restores the initial filamentous cell phenotype.( C ) Overexpression of rcsA gene alone yields a mucoid, shiny colony morphology.( D ) Miller assay showing increased rcsA promoter activity under ydaT induction.The error bars indicate the a v erages ( ± SD) of at least three independent measurements.Student's t-test was performed with values: *** P < 0.0 0 01; * P < 0.05 and ns: P > 0.05.

Figure 10 .
Figure 10.Hypothetical model for a regulatory TF's cascade triggered by C protein and resulting in YdaT-dependent deficient lipopolysaccharide biosynthesis and cell division.C protein (y ello w he xagon) is accepted as a f oreign TF, whic h affects a Rac locus TFs (red bloc ks): RacR, YdaS, YdaT.In turn, YdaT affects a genomic TF, an RcsA TF, which together with RcsB (green blocs) modulate e x opoly saccharide biosynthesis.Thick lines indicate the e xperimentally v erified pathw a y e xplained in this w ork, normal lines pathw a y s kno wn from the literature data and dashed lines pathw a y s to be clarified in the future.Description explained in Discussion.

Table 1 .
Integration sites for Tn 5 mini-transposon insertions