Xenogeneic nucleoid-associated EnrR thwarts H-NS silencing of bacterial virulence with unique DNA binding.

Type III and type VI secretion systems (T3/T6SS) are encoded in horizontally acquired genomic islands (GIs) that play crucial roles in evolution and virulence in bacterial pathogens. T3/T6SS expression is subjected to tight control by the host xenogeneic silencer H-NS, but how this mechanism is counteracted remains to be illuminated. Here, we report that xenogeneic nucleoid-associated protein EnrR encoded in a GI is essential for virulence in pathogenic bacteria Edwardsiella and Salmonella. We showed that EnrR plays critical roles in T3/T6SS expression in these bacteria. Various biochemical and genetic analyses demonstrated that EnrR binds and derepresses the promoter of esrB, the critical regulator of T3/T6SS, to promote their expression by competing with H-NS. Additionally, EnrR targets AT-rich regions, globally modulates the expression of ∼363 genes and is involved in various cellular processes. Crystal structures of EnrR in complex with a specific AT-rich palindromic DNA revealed a new DNA-binding mode that involves conserved HTH-mediated interactions with the major groove and contacts of its N-terminal extension to the minor groove in the symmetry-related duplex. Collectively, these data demonstrate that EnrR is a virulence activator that can antagonize H-NS, highlighting a unique mechanism by which bacterial xenogeneic regulators recognize and regulate foreign DNA.

INTRODUCTION

Horizontal gene transfer of mobile gene elements (MGEs) is a major driving force shaping genome evolution in bacteria (1). Regarding Gram-negative bacterial pathogens, type III and VI secretion systems (T3/T6SS) and other various virulence determinants are encoded in horizontally acquired genomic islands (GIs) that have evolved for bacterial virulence, fitness, and adaptation to hosts and various environmental niches (2). However, existing biases in the G + C composition and codon usage of these GI genes impose transcription and translation limitations that may affect the physiology and fitness of bacterial hosts. Thus, the expression of GIs is tightly controlled by host regulatory networks (3–4). How bacteria discriminate themselves from foreign MGE DNA and offset the fitness costs of acquisition, amelioration, maintenance, and expression of these genes remains to be determined (3).

Nucleoid-associated proteins (NAPs) are small, basic, and highly abundant proteins that bind DNA, act as pivotal regulators of chromosome organization and play critical roles in facilitating or limiting the acquisition and expression of MGE DNA (5). Among a dozen NAPs encoded in Escherichia coli, H-NS is highly conserved and well characterized as an essential xenogeneic silencer (XS) that targets various AT-rich GIs by recognizing structural features unique to the minor groove of AT-rich DNA (6–8). H-NS utilizes a “prokaryotic AT-hook motif” (Q/RGR) that inserts into the minor groove and forms extensive interactions along the groove floor (9). Global H-NS–DNA interactions alter nucleoid topology following folding into higher-order structures—i.e. superhelical nucleoprotein filaments—by wrapping, bending or bridging nucleic acids (5). Additionally, the cooperation between H-NS and other NAPs mediated by the spatial organization of their corresponding protein binding sites can further govern the higher-order architecture of the nucleoprotein complexes (10). Recent investigations have established that both chromosomally (endogenous) and MGE gene-encoded (xenogeneic) NAPs coexist in host bacteria (6,11–13). Although endogenous NAPs are well characterized in the xenogeneic silencing of MGE genes, including GIs and prophages, less is known concerning the roles of xenogeneic NAPs in their functional interaction with endogenous NAPs, in foreign element biology and in host cell physiology.

Edwardsiella bacteria are Gram-negative and zoonotic enteric bacterial pathogens that cause systemic infection in animals and humans (14). Among the several established species in the Edwardsiella genus, E. piscicida and E. anguilarum are the causative agents of edwardsiellosis for >20 species of piscine hosts, leading to huge economic losses in aquaculture worldwide (14,15). The E. piscicida genome harbours 24 horizontally acquired GIs (16,17), of which T3SS (GI7) and T6SS (GI17) have been established as pivotal for the bacterium to grow intracellularly and occupy the niche by translocating a repertoire of ∼20 putative and established effectors against host defences (14,18–20). T3SS in E. piscicida is analogous to the Salmonella pathogenicity island (SPI-2) in S. Typhimurium (15,21–22). In addition to the T3/T6SS gene clusters as in E. piscicida, E. anguillarum has acquired more GIs (n = 33), including the newly acquired locus of enterocyte effacement (LEE) genes (15).

GIs are tightly controlled by regulatory networks, as exemplified by the T3/T6SS in E. piscicida (14). The expression of T3SS and T6SS is controlled by several regulators, including the two-component system (TCS) EsrA-EsrB, which is homologous to SsrA-SsrB in S. Typhimurium (23) and essential for bacterial pathogenesis. Additionally, many other regulators, such as PhoP (24), RpoS (25) and EvrA (26), control T3/T6SS expression. Furthermore, H-NS has been established to control T6SS gene expression in E. piscicida (27–28). Using transposon insertion sequencing (TIS)-based technologies, a large compendium of genes essential for in vivo survival was revealed (19,26), warranting extensive exploration of their roles in modulating virulence programming in vivo and in vitro in the bacterium.

Here, we identified and characterized a novel regulator, EnrR, that resides in a GI harbouring MGE-associated genes and is closely related to the in vivo virulence of E. piscicida. EnrR controls T3/T6SS by competing with H-NS in the esrB promoter and derepressing its expression. EnrR colocalizes with bacterial nucleoid as an NAP. Structural analysis revealed a unique DNA binding mode of EnrR. Finally, this work revealed that EnrR can globally bind to and regulate GIs in Edwardsiella bacteria and Salmonella, controlling the expression of MGEs. Collectively, we show that EnrR is a unique NAP that plays a role as an H-NS-antagonizing virulence activator, providing insights into the mechanisms by which EnrR recognizes AT-rich DNA.

MATERIALS AND METHODS

Bacterial and cell strains, media and culture conditions

The bacterial strains and plasmids used in this study are listed in Supplementary Table S1. The Escherichia coli, S. Typhimurium, E. piscicida and E. anguillarum strains were cultured in Luria Bertani (LB) broth at 37 or 30°C. Notably, E. piscicida was grown statically at 30°C in DMEM for T3/T6SS production. Escherichia coli CC118 λpir was used for plasmid preparation. Unless otherwise indicated, the plasmids were introduced into E. coli strains by transformation and into E. piscicida, E. anguillarum or S. Typhimurium by electroporation at 2000 V for 3 ms. When needed, ampicillin (Amp; 50 μg/ml), colistin (Col; 12.5 μg/ml), chloramphenicol (Cm; 34 μg/ml), streptomycin (Str; 100 μg/ml) and kanamycin (Kan; 50 μg/ml) were added.

Bacterial genetic engineering

The in-frame deletion mutants were generated by sacB-based allelic exchange as previously described (25). Briefly, upstream and downstream fragments were amplified by polymerase chain reaction (PCR) and then ligated into the suicide vector pDM4. The plasmid was transformed into SM10 λpir and subsequently transferred into E. piscicida, E. anguillarum or S. Typhimurium by conjugation. The insertion mutants with single crossover recombination events were selected on LB agar medium containing Cm and Col, while double crossover recombination events were counterselected on LB agar medium containing 12% sucrose. An isogenic mutant with the wild-type esrB promoter (P) sequence replaced with mutant Placking EnrR binding sites (Pmut1+2+3) was similarly constructed. The respective complementation and overexpression plasmids were constructed with pUT as previously described (25).

Biofilm formation and autoaggregation assays

The biofilm formation ability of E. piscicida and Salmonella on polystyrene was quantified as described elsewhere (29–30). Briefly, bacteria were grown overnight in LB at 30°C. The cultures were diluted 1:100 in fresh LB, and 200 μl of the cell suspensions were statically subcultured into sterile 96-well polystyrene microtiter plates at 30°C for 24 h. Each well was washed with phosphate-buffered saline (PBS) three times. Next, total biofilm formation was measured using 2% crystal violet staining. Bacterial autoaggregation was observed after the cells were statically grown in DMEM at 30°C for 24 h (29).

Extracellular protein (ECP) assay

Overnight cultured bacteria were incubated in 5 ml of DMEM at 30°C for 24 h. Next, protease inhibitor (Roche, Switzerland) was added and incubated for 10 min at 4°C. After pelleting the cells at 5000 g for 10 min, the supernatants were filtered through a 0.22 μm low-protein-binding Millex filter (Millipore, USA) and concentrated to 250 μl using a 10-kDa-cut-off Amicon Ultra15 centrifugal filter (Millipore, USA). The concentrated ECP profile was detected by SDS-PAGE as previously described (25). The supernatants of the same amounts of cells were also resolved by SDS-PAGE following silver staining.

Quantitative RT-PCR (qRT-PCR) and RNA-seq analyses

qRT–PCR and RNA-seq detection of gene transcripts were performed as described previously (23,25). Edwardsiella piscicida and S. Typhimurium were cultured in DMEM at 30 or 37°C for 12 h. Total RNA was extracted using an RNA isolation kit (Tiangen, China). For qRT-PCR, 1 μg of RNA from each sample was used to remove genomic DNA with DNase I (Promega, Madison, WI, USA), and reverse transcription was performed using the FastKing One Step RT-PCR Kit (Tiangen, China). Three independent qRT-PCR experiments were performed in triplicate using specific primer pairs (Supplementary Table S2) (Applied Biosystems, USA). RNA-seq was performed using a service provided by Personalbio China, and the sequencing data were analysed using Rockhopper (Version 2) as previously described (23,31).

Chromatin immunoprecipitation sequencing (ChIP-seq) and ChIP–qPCR

ΔenrR strains expressing EnrR-FLAG or FLAG alone driven by P in a plasmid were used for ChIP assays as previously described (25). Notably, bacteria were incubated in DMEM at 30°C for 12 h. The cultures were treated with 1% formaldehyde at room temperature for 10 min, and the reaction was stopped by adding 125 mM glycine. Next, the bacteria were washed twice with cold PBS and resuspended in 15 ml of IP buffer (50 mM HEPES-KOH pH 7.5, 0.1% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA pH 8 and moderate protease inhibitor). High pressure and ultrasound were used to disrupt bacteria and for subsequent DNA fragmentation, respectively. The fragmented DNA was purified by phenol/chloroform and precipitated using a Dr GenTLE precipitation carrier (Takara, Japan). Next, the sequencing library was constructed using a VAHTS Turbo DNA library prep kit (Vazyme, China) and sequenced using the MiSeq platform (Illumina, USA). The sequencing data were analyzed using the MACS algorithm (32) and MEME analysis to generate the EnrR-binding motif (https://meme-suite.org/). KEGG pathway analysis was performed using Kobas 2.0 (33).

ChIP–qPCR analyses were performed as previously described to validate the binding of EnrR to the specific DNA motifs identified in the above-mentioned ChIP-seq assays (26). Briefly, E. piscicida strains chromosomally expressing EnrR-FLAG or FLAG alone in the enrR locus driven by native Pwere treated with ChIP followed by qPCR with various specific primer pairs (Supplementary Table S2) targeting the ∼100 bp central region in ChIP-seq peaks. For each DNA target, the ΔCT values of the input and IP fractions were calculated in both samples for the EnrR-FLAG- and FLAG-expressing strains. Each value was then divided by the corresponding ΔCT of the nonspecific gyrB promoter region in the strains. Next, the enrichment ratio was calculated using the ΔΔCT value in the EnrR-FLAG strain divided by that of the FLAG strain. The ChIP–qPCR assays of H-NS binding to specific gene promoters were performed similarly using the anti-GST magnetic beads (Beyotime, Shanghai).

Lysogeny phage-related experiments

Bacteria were fixed in 2.5% glutaraldehyde in PBS for 12 h at 4°C, dehydrated with a series of gradient ethanol solutions (30–100%) and freeze-dried. Subsequently, the samples were coated with gold film by sputter coating and then observed under a field emission scanning electron microscope (S3400-N; Hitachi, Japan). Phages released from the bacteria were extracted using the Universal Phage Genomic DNA Extraction Kit (Knogen Biotech, China). Phage particles were resuspended in SM buffer (100 mM NaCl, 8 mM MgSO4·7H2O, 50 mM Tris-Cl pH 7.5; 0.002% w/v gelatine) and negatively stained with 2% phosphotungstic acid for 10 min followed by observation under a transmission electron microscope (JEM-1400; JEOL, Japan).

Purification of protein

The expression of C-terminal-tagged EnrR or N-terminal-tagged H-NS was induced by 0.2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) in E. coli BL21 (DE3) at 20°C at 120 rpm. At 18 h postinduction, the bacteria were harvested by centrifugation and purified as previously described (25).

Electrophoretic mobility shift assay (EMSA), bacterial one-hybrid (B1H) assay and microscale thermophoresis (MST) analyses

EMSA and B1H assays were conducted as previously described (30). To measure the protein affinity to a specific DNA probe, Cy5-labelled DNA was used, and the purified protein was diluted with TE buffer (10 mM NaCl, 0.5 mM EDTA, 10 mM Tris-HCl pH 7.5) followed by mixing protein and DNA at a volume ratio of 1:1. For MST assays of EnrR or H-NS affinities to specific DNA probes, EnrR or H-NS was labelled with RED-tris-NTA dye (MO-L018; Nano Temper, Germany) and then was mixed with an equal volume of the diluted DNA probes. After incubation for 20 min at room temperature, a Monolith NT™ capillary (Nano Temper, Germany) was horizontally placed into the reaction tube to aspirate the sample, and thermophoresis was detected using a Monolith NT.115 system (Nano Temper, Germany).

DNA-protection assay

DNA-protection assays were performed as previously described (34). Notably, a series of concentrations of EnrR (0–16 μM) were incubated with 750 bp P DNA (200 ng of the promoter region of ETAE_2212) for 30 min at room temperature in TE buffer. Next, 20 μl of the mixture was treated with 0.5 U of DNase I (NEB, UK) for 5 min at 37°C. The reactions were terminated by incubation at 70°C for 10 min, followed by treatment with proteinase K (20 μg), 5 mM MgCl2, 2% SDS and 0.3 M sodium acetate for 1 h at 37°C. The samples were extracted using phenol/chloroform and precipitated using a Dr. GenTLE precipitation carrier (Takara, Japan). The pellets were dissolved in 10 μl of ddH2O and run on a 1% agarose gel in TAE at 100 V for 30 min.

Plasmid supercoiling conformation assay

The supercoil assay was performed according to a previously published procedure using freshly prepared pUT plasmid (35). The plasmid was incubated with various amounts of EnrR at 37°C for 30 min. Next, topoisomerase I (TopA) (Takara, Japan) was added to the mixtures and incubated at 30°C for 30 min. The reaction was subsequently terminated by proteinase K (20 μg). The samples were analysed by 1% agarose gel electrophoresis.

Fluorescence colocalization assays

An E. piscicida strain expressing the EnrR-GFP fusion protein driven by the P promoter at the enrR locus was constructed. After 6 h of growth, the bacteria were washed twice with cold PBS and stained with DAPI for 30 min on LBA. The samples were visualized by differential interference contrast (DIC) and fluorescence microscopy using a confocal microscope (A1R; Nikon, Japan).

Atomic force microscopy (AFM)

The circular pUT vector or linear P DNA was incubated with different concentrations of EnrR protein in TE buffer at room temperature for 30 min. A 10 μl droplet of the mixtures was placed onto fresh mica for 2 min. Next, the mica was rinsed with Milli-Q water and dried under a stream of nitrogen. The samples were visualized by scanning probe microscopy (Veeco, USA).

Crystallization and data collection

Both native and Se-substituted EnrR proteins in gel filtration buffer were concentrated to 10–12 mg/ml. To prepare the DNA complex, EnrR and DNA1 were mixed at a molar ratio of 1:1.2 and incubated at room temperature for 30 min. The crystallization conditions were initially identified at 16°C using crystallization kits from Hampton Research Company and a crystallization robot system using the sitting-drop vapour diffusion method at 0.4 μl per drop (1:1 ratio of protein–DNA mixture or protein alone to well solution). Crystallization was optimized using the hanging-drop vapour diffusion method. Apo-EnrR crystals were grown in 20% PEG 3350 and 0.2 M sodium formate. EnrR-DNA1 complex crystals were grown in 20% PEG 3350 and 0.2 M K2SO4 buffer supplemented with 0.2 M D-sorbitol. Crystals of apo-EnrR were cryoprotected using their mother liquor supplemented with 25% glycerol, whereas EnrR-DNA1 complex crystals were cryoprotected using 30% PEG 400. All the crystals were snap-frozen in liquid nitrogen. X-ray diffraction data (Table 1) were collected on beamlines BL17U and BL19U at the Shanghai Synchrotron Radiation Facility (SSRF). Data processing was performed using the HKL2000 or HKL3000 programs (36).

Table 1.

Data collection and refinement statistics

Structure	Apo-EnrR	EnrR/DNA1
(PDB ID)	7F9I	7F9H
Data collection a
Space group	P212121	C2
Cell parameter
a, b, c (Å)	34.1, 56.3, 72.4	104.3, 58.5, 67.6
α, β, γ (°)	90.0, 90, 90.0	90.0, 98.4, 90.0
Wavelength (Å)	0.9793	0.9793
Resolution (Å)	30.0–2.5	30.0–1.75
Highest resolution shell (Å)	2.59–2.50	1.84–1.75
Completeness (%)	97.2(93.2)	98.1(67.0)
Redundancy	6.1(4.2)	6.7(6.0)
I/σ(I)	10.1(2.7)	26.2(4.8)
R merge (%)	14.3(38.2)	5.4(38.3)
Refinement
Resolution (Å)	28.1–2.5	29.6–1.78
R work (%)/Rfree (%)	24.5/28.2	20.1/22.2
No. of atoms
Protein	1061	1164
DNA	0	898
Water	2	228
Wilson B factors (Å2)	19.3	17.6
R.m.s. deviations
Bond length (Å)	0.012	0.007
Bond angles (°)	1.492	1.309
Ramachandran plot (%)
Most favoured	96.2	97.1
Additional allowed	3.8	2.9

Structure determination and refinement

The phases of the Se-substituted EnrR-DNA1 complex structure were determined using the single-wavelength anomalous diffraction (SAD) method (37) and the anomalous signal of Se- atoms with the AutoSol program (38) embedded in the Phenix suite (39). The initial model was refined using the Refmac5 program (40) of the CPP4i suite (41). The apo-EnrR structure was solved using the molecular replacement method; the EnrR molecule of the complex was used as a search model. The model was manually built using COOT (42) and refined using either Refmac5 or phenix.refine programs (43). During refinement, 5% of randomly selected data were set aside for free R-factor cross-validation calculations. The 2Fo–Fc and Fo–Fc electron density maps were regularly calculated and used as guides to build the missing amino acid residues and solvent molecules with COOT. The structural refinement statistics are summarized in Table 1.

Animal survival, in vivo competition and luminescence assays

Fish-related experiments were performed as previously described (25). Briefly, 3-month-old healthy turbots weighing approx. 25 ± 3 g were used for the experiments. Fish were anaesthetized (10 min) in sea water supplemented with MS-222 (0.02% v/v) (Sigma-Aldrich) and aseptically dissected to harvest their tissues. For survival assays, different concentrations of bacteria (2 × 102–103 CFU/g) were inoculated into fish at 100 μl intraperitoneally (i.p.), and fish survival was monitored over 30 days.

For in vivo competition assays, the paired bacterial strains were resuspended and mixed 1:1 in PBS. Next, the mixtures were diluted and i.p. injected into fish at a dose of 3 × 105 CFU/fish, and fish tissues were harvested at 5 days post-infection (DPI) when E. piscicida established colonization and systemic infection (19,25). The cell-forming units (CFU) per gram of tissue were enumerated by plating homogenized tissue on DHL agar containing appropriate antibiotics.

For in vivo luminescence detection, the subcultures of reporter strains were diluted to 106 CFU/ml in PBS and i.p. injected into turbot at 100 μl. At 5 DPI, the fish were i.p. inoculated with beetle luciferin substrate (Promega). After 10 min, the fluorescence was detected using the Kodak In-Vivo Multispectral System FX (Carestream Health) as previously described (25). Next, the livers of the fish were sampled, and bacterial colonization was enumerated by CFU plating.

Ethics statement

All the animal protocols used in this study were approved by the Animal Care Committee of the East China University of Science and Technology (2006272). The Experimental Animal Care and Use Guidelines from the Ministry of Science and Technology of China (MOST-2011-02) were strictly followed.

RESULTS

EnrR is a new virulence regulator crucial for E. piscicida infection in vivo

To identify genes temporally required for E. piscicida in vivo survival in hosts, we devised and applied the algorithm termed pattern analysis of conditional essentiality (PACE) to time series transposon insertion sequencing (TIS) data collected during a 14-day in vivo infection experiment in a turbot infection model (19). Because T3SS and T6SS have been established as key determinants during the progression of E. piscicida in vivo infection (19,25–26), we compared genes with similar in vivo essentiality patterns to those of the T3SS or T6SS loci—e.g. esrB, eseB and evpP—from the PACE analysis of the TIS dataset (19). Thus, we identified that the mutant of ETAE_0051 (hereafter referred to as  nucleoid-associated regulator R, enrR) showed a normalized median fitness decrease of approximately -0.25 log2-fold change (FC) per day, comparable to that of T3/T6SS insertion mutants, indicating its attenuation in its colonization capacities in vivo (Figure 1A).

Figure 1.

Characterization of the enrR mutant. (A) Relative in vivo fitness profiles of transposon (Tn) insertion mutants of ETAE_0051 (enrR), esrB, eseB and evpP. The dynamic log2-fold change (FC) in abundance for each mutant strain recovered from the liver relative to the wt strain was plotted against the days postinjection (DPI) by intraperitoneal (i.p.) injection of a highly saturated mutant library in turbot. (B) The gene locus of ETAE_0051 (enrR) neighbours the MGEs. (C) Structural analysis of EnrR. Red arrows represent the residues that recognize specific nucleobases. Larger letters represent more conservative residues. (D) Survival curve of turbot fish challenged with wt and ΔenrR strains, as well as the ΔenrR strain harbouring the complement plasmid pUT-enrR (enrR) by i.p. injection (n = 30 per group). **, P < 0.01 by comparing the ΔenrR strain with the wt and enrRstrains using Kaplan–Meier survival analysis using the log rank test (Mantel–Cox).

The enrR gene is localized in E. piscicida GI2, flanked by genes encoding transposases, integrases and DNA repair proteins as well as clustered genes encoding uncharacterized phage-like proteins (Figure 1B) that appear to be horizontally acquired from decayed prophages and are present in various phylogenetically distant bacteria (Supplementary Figure S1). Bioinformatics analysis showed that the enrR gene encodes a hypothetical small regulatory protein of 90 amino acid (aa) residues with a winged helix-turn-helix (w-HTH) DNA-binding domain (PF13693) that belongs to the XRE superfamily (Figure 1C). To verify the in vivo TIS results, the enrR in-frame deletion mutant strain ΔenrR and its complement strain ΔenrR+ were constructed. The ΔenrR strain showed no growth difference from the wt strain when grown in either DMEM or LB medium (Supplementary Figure S2A,B). Compared with the wt strain, the ΔenrR strain showed drastically impaired virulence and caused significantly lower mortalities in turbot, a natural host (16), over 24 days post-infection (DPI) (Figure 1D). Introduction of the enrR gene into the ΔenrR mutant strain restored its infectivity in a fish model. These results indicated that enrR is essential for bacterial virulence towards hosts.

EnrR controls T3/T6SS expression in vitro and in vivo

To clarify the roles of EnrR in bacterial virulence, we examined whether the control of virulence by EnrR is related to T3/T6SS expression. First, we tested wt-, ΔesrB- and enrR-related strains using an assay of autoaggregation, a phenotype attributable to the production of the T3SS extracellular apparatus protein EseB (29), and extracellular protein (ECP) profiling by SDS–PAGE analysis (Supplementary Figure S2C,D). As expected, the ΔenrR strain and complement strain enrR+ showed patterns of autoagglutination and EseB production similar to ΔesrB and its complementation strains, respectively (Figure 2B). Western blot analysis of T3/T6SS secretion using the antiserum against the T3SS protein EseB and T6SS protein EvpP further confirmed that EnrR positively regulates T3/T6SS production (Figure 2B, bottom). Similar to the above analyses, qRT-PCR also showed that T3/T6SS gene transcripts were upregulated by EnrR (Figure 2C).

Figure 2.

EnrR is a positive regulator of virulence in vitro and in vivo and functions by regulating esrB. (A and B) Autoaggregation (A) and extracellular protein (ECP) profiles (B) of the indicated strains statically cultured in DMEM. The supernatants from the same amounts of cells were resolved by SDS–PAGE and silver stained (upper panel), and specific bands corresponding to T3SS and T6SS proteins are shown. The SDS–PAGE-resolved proteins were then blotted with anti-EseB- or anti-EvpP-specific antiserum (lower panels). RNAP was used as the loading control for the blots. (C) qRT–PCR assays for the indicated transcripts in the wt, ΔenrR and enrR strains. The results shown are the means ± S.D. (n = 3) relative to the wt results. gyrB was used as the internal control. ***, P < 0.001 compared with wt based on Student’s t-test. (D) In vivo competition assays of the equally mixed indicated strains (Cm resistance) and WTΔp, the wt strain with pEIB202 cured (Cm sensitive) inoculated into fish and recovered from the livers of turbot fish at 8 DPI (n = 5 per group). ** P < 0.01, * P < 0.05 and N.S. (not significant) for P > 0.05, based on one-way ANOVA and Fisher’s LSD multiple comparison post-test. (E) Activities of the promoters of evpA and eseB from T3/T6SS gene loci in wt or ΔenrR strains inoculated in vivo in fish. The indicated strains harbouring P-luc or P-luc reporter plasmids were inoculated into turbot, and bioluminescence was measured at 8 DPI and normalized by CFU. ***, P < 0.001 based on ANOVA of the relative fluorescence units (RFU). (F and G) EnrR regulates T3/T6SS gene expression through esrB in both E. piscicida EIB202 and E. anguillarum ET080813 strains. The wt, ΔesrB or ΔenrR strains of EIB202 or ET080813 harbouring the empty vector pUT or the plasmid expressing esrB or enrR driven by their respective promoters—i.e. P or P or by the constitutive promoter for the 30S ribosomal protein (P0456)—were investigated for their production of T3/T6SS proteins in the ECPs by SDS–PAGE.

Subsequently, we investigated the in vivo involvement of EnrR in T3/T6SS expression during infection in a fish model. Competitive index (CI) experiments were performed in which ΔenrR, ΔeseBCD, ΔevpAB and the ΔenrRΔeseBCD and ΔenrRΔevpAB mutant strains were mixed equally with WTΔp, the wt strain cured with pEIB202. The CI results at 5 DPI showed that the ΔenrR strain was significantly (P < 0.01) impaired in fitness in fish (Figure 2D) and comparable to that of T3SS- or T6SS-related mutants, further corroborating the idea that EnrR controls T3/T6SS expression. Next, in vivo bioluminescence imaging was used to investigate EnrR regulation of T3/T6SS during E. piscicida infection of turbot. Luciferase reporters of P and P expression were introduced into a neutral position on the chromosomes of the wt and ΔenrR strains, respectively, and these strains were intraperitoneally (i.p.) inoculated into turbot fish at the same dose. At 5 DPI, significantly decreased P and P activity was detected in the ΔenrR strain background compared with that in the wt strain (Figure 2E). Taken together, these results revealed that EnrR controlled T3/T6SS expression in vitro and in vivo during infection.

EnrR regulates E. piscicida virulence via EsrB

The response regulator EsrB is pivotal for the expression of T3/T6SS in E. piscicida (25). We reasoned that EnrR controls E. piscicida T3/T6SS expression by modulating esrB expression because esrB overexpression driven by the promoter for the 30S ribosomal protein ETAE_0456 (P) in the ΔenrR strain could restore the autoaggregation phenotype and T3/T6SS production (Figure 2A,B), and the transcript level of esrB was downregulated in the ΔenrR strain (Figure 2C). Next, we performed genetic analysis of esrB in mediating the EnrR control of T3/T6SS expression in various genetic backgrounds of E. piscicida EIB202 and E. anguillarum ET080813, a strain that encodes T3/T6SS GIs and the EsrA-EsrB two-component system but does not encode enrR homologue (15). In EIB202, the introduction of P-driven enrR and esrB or esrB encoding the D60E variant (esrBD60E) mimicking constitutive phosphorylation in EsrB fully restored the T3/T6SS production defects in the ΔenrR strain (Figure 2F, lanes 5–7), although esrB or esrBD60E showed similar capacities in activating T3/T6SS production, likely because of sufficient phosphorylation for EsrB. Additionally, the expression construct P-esrB, but not P-enrR, activated T3/T6SS expression in the ΔenrRΔesrB double mutant (Figure 2F, lanes 8 and 9), indicating that EnrR is epistatic to EsrB with respect to the T3/T6SS regulatory pathway. This notion was also supported by the result observed in ET080813, in which T3/T6SS production could be triggered by introducing an enrR overexpression plasmid (Figure 2F, lanes 10 and 11). The process was EnrR- and EsrB dependent because neither the expression of a DNA-binding mutant of EnrR, EnrRR47G in the wt strain, nor the ΔesrB mutant strain containing the enrR overexpression plasmid showed wt levels of T3/T6SS (Figure 2F, lanes 12 and 13). Notably, the crystal structure (to be described later) revealed that the R47 residue was essential for the DNA binding of EnrR.

Next, we asked whether the control of the esrB promoter activity by EnrR was associated with T3/T6SS production. In EIB202, the expression of esrB driven by both the native esrB promoter (P) and P fully restored T3/T6SS expression in the ΔesrB strain (Figure 2G, lanes 3 and 4). However, in the ET080813 background, these two expression constructs could not generate T3/T6SS production although P-esrB caused minimal T3/T6SS proteins in the ΔesrB strain (Figure 2G, lanes 7 and 8). We used the expression of esrBD60E driven by P or P in ΔesrB to minimize the effect of upstream phosphorylation by the cognate histidine kinase EsrA, and only the latter triggered the wt EIB202 level of T3/T6SS expression (Figure 2G, lanes 12 and 13). These results suggested that the activity of P could be regulated by a specific repressor in ET080813. Furthermore, compared with the ΔesrB strain chromosomally expressing EnrR—i.e. the ΔesrB::enrR strain, which showed deficiency in T3/T6SS production—the introduction of the plasmids expressing EsrB driven by P or P fully augmented T3/T6SS expression (Figure 2G, lanes 9–11). The restoration of T6SS protein EvpC secretion in ΔesrB::enrR harbouring plasmid P-esrB compared with the strain ΔesrB/P-esrB suggested that EnrR might derepress esrB and another T6SS-related promoter (lanes 8 and 11). Collectively, these data demonstrated that EnrR controls T3/T6SS expression in Edwardsiella bacterium primarily by modulating esrB promoter activation.

Edwardsiella piscicida T3SS and its regulatory mechanisms are analogous to those of S. Typhimurium SPI-2. Expression of EnrR in S. Typhimurium SL1344, a strain that does not encode an EnrR homologue, resulted in higher levels of the SPI-2 transcript (Supplementary Figure S3A,B) and biofilm formation (Supplementary Figure S3C) than that of the control strain harbouring the empty plasmid. EnrR regulated SsaJ expression dependent on SsrB, the known master regulator of SPI-2 (Supplementary Figure S3B) (44–45). Compared with the ΔssrB::pUt and ΔssrB::pUt-enrR strains, all of the mice challenged with the EnrR-producing wt strain lost body weight and died in 7 days, while most (75%) of the mice injected with the same dose of the wt strain were still alive after 12 days (Supplementary Figure S3D,E). Taken together, these results showed that EnrR activated the expression of SPI-2 and enhanced Salmonella virulence through SsrB.

EnrR competes with H-NS from the esrB promoter to activate virulence gene expression

H-NS represses the transcription of T6SS genes in Edwardsiella bacteria (27–28). Here, we confirmed that H-NS overexpression inhibited the production of T3/T6SS proteins and EsrB in the wt and ΔenrR strains (Figure 3A, lanes 2 and 4, and Supplementary Figure S2E). Notably, H-NS is an essential gene in E. piscicida (19,26); its deletion mutant could only be generated with another repressor mutation, including rpoS null mutation, as described in other bacteria (3,8). Deletion of hns in the ΔrpoS context resulted in no obvious changes in the level of T3/T6SS proteins (Figure 3A, lanes 5–6), and the additional disruption of enrR in the Δhns mutant led to marginally reduced T3/T6SS production (lanes 6 and 8). By contrast, enrR overexpression augmented the expression of these virulence genes in the presence or absence of hns (lanes 9–10). Thus, these data suggested a role of EnrR in overcoming H-NS inhibition of esrB expression, although we could not exclude the activity of EnrR as a transcriptional activator.

Figure 3.

EnrR binds to the esrB promoter region and antagonizes H-NS binding. (A) SDS–PAGE analysis of the ECPs from the indicated wt or ΔenrR strains overexpressing H-NS or with hns deletion in the ΔrpoS context. Note that hns is an essential gene that can only be disrupted in the presence of additional rpoS suppressor mutations. (B) MST analysis of EnrR or H-NS binding to esrB promoter DNA (P) probes. Purified EnrR and/or H-NS protein was labelled with RED-tris-NTA dye and then incubated with wt or mutant P probes of increased concentrations in NT standard capillaries in MST assays. The y-axis represents the fractions of proteins being bound. The x-axis is the mol concentration per litre of DNA. The curve is the fit of data points to the standard Kd-fit function. The respective Kd (dissociation constant) values from triplicate assays are shown as means ± S.D. in brackets. Note that the data points of the P mut4 + 5 probe could not fit into a curve. N.D., not detectable. (C and D) DNase I footprinting analysis of EnrR or H-NS binding to P DNA. Electropherograms of a DNase I digest of the P promoter probe after incubation with or without EnrR (C) and H-NS (D). The respective nucleotide sequences protected by the proteins are boxed in colours. (E) Diagram of the P region. The EnrR- or H-NS-protected regions, -10 and -35 boxes for RNAP binding and the transcriptional and translational start sites are highlighted. The probes with mutated regions for binding assays are shown as mut1-mut5, mut1 + 2 + 3, or mut4 + 5 (B). (F) H-NS competes with EnrR for binding to P DNA in vitro. Top panel, EMSAs containing Cy5-labelled P DNA and either purified GST-tagged H-NS or His6-tagged EnrR or EnrRR47G. Competitive EMSA reactions comprising 3 μM purified H-NS, and either 0.5, 2, 4 or 6 μM purified EnrR are shown in the top right. The reactions were run on polyacrylamide gels and laser scanned at 650 nm. B: bound DNA; F: free DNA. Bottom panel, the gels were then transferred to a PVDF membrane and probed for H-NS using anti-GST antibodies. (G) In vivo binding of H-NS to the promoter regions of the esrB, hns and gyrB genes was determined in GST-H-NS-expressing E. piscicida cells harbouring wt or mutant P or Pmut1+2+3 grown in DMEM for 12 h using ChIP with anti-GST magnetic beads. The relative enrichment of GST-H-NS on P, P and P DNA was subsequently assayed by qPCR experiments with primer pairs targeting these regions. ***, P < 0.001, N.S., P > 0.05 based on Student’s t-test of the relative H-NS binding enrichments normalized to that of rpoD (n= 3).

We then investigated whether EnrR and H-NS could directly bind to the promoter region of esrB using MST analyses. EnrR and H-NS could bind to P with dissociation constants (Kd) of 11.5 ± 0.4 μM and 18.1 ± 2.8 μM, respectively (Figure 3B). EMSA experiments and B1H also demonstrated the binding ability of EnrR with P (Supplementary Figure S4A,C) (30). Furthermore, dye primer-based DNase I footprinting assays mapped at least three AT-rich binding sites of EnrR (Figure 3C) and two binding regions of H-NS (Figure 3D) just upstream of the esrB promoter region (Figure 3E). Distinct mutations in these regions (mut1-mut3 and mut1 + 2 + 3 for EnrR, mut4-mut5 and mut4 + 5 for H-NS) resulted in various extents of decrease in the affinities of these two proteins to the target probes (Figure 3B). Additionally, EnrR did not appear to bind with H-NS under in vitro conditions (Supplementary Figure S4F). Collectively, these analyses demonstrated that both EnrR and H-NS could bind to the esrB promoter region.

MST assays intriguingly revealed that the DNA binding of H-NS to P was significantly reduced in the presence of EnrR (Figure 3B). To further determine whether EnrR competed with H-NS from the esrB promoter in vitro, we performed EMSA using both GST-H-NS and His6-EnrR (or otherwise the null mutant His6-EnrRR47G) added to esrB promoter DNA (Figure 3F). The GST- or His6-tagged H-NS or EnrR behaved similarly to their wt proteins in the control of T3/T6SS production (Supplementary Figure S2D). EMSA confirmed that their binding to Pincreased with increasing protein concentrations (0–3 μM), and many more oligomerized protein–DNA complexes for H-NS were found than for EnrR (Figure 3F, left column). When a constant amount (3 μM) of H-NS was added, increasing concentrations (0.5–6 μM) of EnrR resulted in enhanced shifts for the migration of the nucleoprotein complex, similar to that of the EnrR-P complex (Figure 3F, right column). This analysis suggested that EnrR competed with H-NS from this promoter DNA. The proteins in EMSA gels were then transferred to a PVDF membrane to determine whether any H-NS protein remained bound to the shifted DNA substrate via Western blotting. As expected, as the EnrR concentration increased from 0.5 to 6 μM across reactions and the signal corresponding to unbound H-NS accumulated (Figure 3F, right column, lanes 4–7). Additionally, the parallel increase in EnrRR47G from 2 to 6 μM could not prevent the DNA binding of H-NS and remove it from P (Figure 3F, right column, lanes 8–10). These results confirmed that EnrR antagonization of H-NS relied on EnrR-DNA binding activity.

We further investigated the effects of H-NS occupancy on the esrB promoter in cells grown under T3/T6SS induction conditions. The hns gene is highly conserved and self-controlled in E. coli and other bacteria (7). Using ChIP–qPCR assays, we determined that H-NS bound to the esrB promoter region more strongly in the strain with the mutant P lacking the identified EnrR binding sites (Figure 3C,E) than in the isogenic wild-type Pstrain (Figure 3G). Control experiments showed that the strain with the Pmut1+2+3 mutation retained normal H-NS binding to the hns DNA (Figure 3G) and extremely low binding to the DNA region upstream of gyrB (Figure 3G). In summary, these findings demonstrated that EnrR antagonized H-NS binding to the esrB promoter region and alleviated H-NS repression of the master regulator EsrB to promote virulence gene expression.

EnrR features NAP protein binding to DNA in vivo and in vitro

EnrR is a small, basic (PI = 11.4), and highly expressed DNA-binding protein with an estimated ∼29 000 copies per cell (Supplementary Figure S2H) comparable to the H-NS abundance and is hypothesized to be an NAP. To test this hypothesis, we first characterized the DNA-binding properties of EnrR. EnrR could specifically and widely protect against DNase I digestion treatment of the 750 bp DNA probe of the promoter region of ETAE_2122 (Figure 4A) that was enriched by ChIP-seq analysis (see below). Further evaluation of the in vitro effect of EnrR on the activity of bacterial topoisomerase TopA showed that the plasmid DNA relaxed by TopA gradually transformed to a supercoiled configuration with increasing concentrations of EnrR (Figure 4B), indicating EnrR’s inhibitory function against the DNA relaxation activity of TopA. By contrast, BSA protein showed no resistance to the DNA relaxation activity of TopA. In the DNA-binding competition experiments, the amount of EnrR-DNA complex decreased with increasing concentrations of actinomycin D or methyl green, the specific minor- and major-groove binding drugs, respectively (Figure 4C) (34), strongly suggesting that EnrR could bind to the DNA minor and major grooves.

Figure 4.

Identification of EnrR as a novel NAP. (A) DNase I protection ability assay of EnrR. DNA was incubated with different concentrations of EnrR and then digested with 0 or 0.5 U of DNase I for 5 min. The recovered DNA was run on an agarose gel. BSA was used as a control. (B) EnrR altered the plasmid supercoiling conformation. Plasmid DNA was incubated with different concentrations of EnrR or EnrRΔN17, an EnrR variant with truncation of the N-terminal 17 residues (to be described later), in the absence or presence of topoisomerase A (TopA). BSA was used as a control. The recovered plasmid was analysed on an agarose gel. (C) EnrR binds to the DNA minor and major grooves. Cy5-labelled P fragments (10 and 40 ng for actinomycin D and methyl green competition assays, respectively) were incubated with EnrR protein in the presence of methyl green and actinomycin D. (D) AFM imaging assay for the protein–DNA complexes formed by EnrR and DNA. Images of circular plasmid DNA (5 ng/μl pUT) with or without 1 μM EnrR were acquired. (E) Colocalization assays for GFP-tagged EnrR with the bacterial nucleoid. The E. piscicida strain chromosomally expressing GFP-tagged EnrR in the native enrR locus was incubated on LBA with DAPI for 30 min. Images were acquired under 288 and 405 nm excitation. GFP (green) and DAPI fluorescence (blue) profiles are shown with the relative fluorescence intensity (y-axis) and cell positions, with -1 and + 1 indicating the fluorescence area (x-axis). Pearson’s correlation coefficient (Rr) between DAPI and GFP localization is also shown.

Using an atomic force microscopy (AFM) assay, we further characterized EnrR-DNA complexes and the structural effects of EnrR on both 750 bp DNA fragments (Supplementary Figure S6) and circular plasmid DNA (Figure 4D). The addition of increasing concentrations of EnrR protein led to DNA oligomerization or aggregation into larger globular particle complexes on the mica surface (Supplementary Figure S5A-D), a finding that was consistent with the observation that the EnrR-DNA complex was blocked in the loading well during EMSA experiments (Figure 4C and Supplementary Figure S4A,B). As a control, a high concentration (16 μM) of EnrRR47G showed defects in forming globular particles with the DNA (Supplementary Figure S5E). Compared with the uniform protein-free structure of DNA, the mixture of EnrR and plasmid DNA resulted in the condensation of the proteins along the DNA contour. In particular, some EnrR-rich cores were observed around the DNA loops/bridges (Figure 4D).

A subsequent colocalization experiment was performed to determine whether EnrR localizes to bacterial genomic DNA. A functional green fluorescence protein (GFP)-fused recombinant EnrR protein (Supplementary Figure S2D) was produced in E. coli with a plasmid or chromosomally expressed in the native enrR locus in E. piscicida, and the bacterial nucleoid was stained with DAPI (Figure 4E and Supplementary Figure S5G). The GFP fluorescence signals closely coincided (Pearson's rho [Rr] > 0.9) with the DAPI-stained regions, demonstrating that EnrR colocalizes with the bacterial nucleoid (Figure 4E and Supplementary Figure S5G). Collectively, these results revealed that EnrR can broadly bind to different regions of DNA, plasmids and the nucleoid. Thus, we propose that EnrR represents a newly described NAP.

EnrR globally binds to and regulates genes involved in various processes

Next, we investigated whether other genes and processes are regulated by EnrR, which would be conducive to virulence and adaptation in host niches. We used ChIP-seq to determine the EnrR regulon (Supplementary Table S3). We used the ΔenrR strain harbouring a plasmid expressing FLAG-tagged EnrR in ChIP assays with a strain expressing FLAG alone as a negative control. EnrR-FLAG behaved similarly to wt EnrR because T3/T6SS production in the ΔenrR strain was fully restored by introducing the vector expressing EnrR-FLAG (Supplementary Figure S2D and S2F), thus validating the subsequent ChIP-associated analysis. Two independent ChIP-seq analyses with a FLAG-specific monoclonal antibody enabled exploration of the EnrR-binding landscape in the E. piscicida genome. In total, we identified 160 enriched loci harbouring EnrR-binding peaks (enriched by ≥2.0-fold compared with the control sample) (Supplementary Table S3) located across the genome, including the intergenic (58%) and coding regions (42%) of the genes (Figure 5A). These peaks were not present in ChIP-seq datasets from experiments in the control strain expressing FLAG alone (Figure 5A). Subsequent ChIP–qPCR analyses with E. piscicida strains chromomosally expressing FLAG-tagged EnrR (WT::P-EnrR-FLAG) or FLAG alone (ΔenrR::P-FLAG) at the native enrR locus validated that all 160 ChIP-seq-identified loci were bona fide EnrR-binding targets (Supplementary Table S3). In particular, an EnrR-binding peak could be found flanking the esrB gene (fold change = 2.0). Additionally, other obvious peaks were observed to localize to the promoter regions as well as intragenic regions of T3/T6SS gene clusters. These included peaks upstream of the major promoters of evpP (enriched by 2.8-fold), esaR (2.0) and esrC (1.5) (Figure 5A) (22). The recruitment of EnrR to these promoters also facilitates T3/T6SS gene expression, as combined deletion of esrB or esrC with enrR led to a much more severe decrease in T3/T6SS-related transcripts (Supplementary Figure S2G).

Figure 5.

Genome-wide mapping of EnrR binding and verification of the binding site. (A) MACS view of EnrR binding across the E. piscicida genome as determined by ChIP-seq. The peak height (y-axis) indicates the sequencing read depth at each genomic position (x-axis). The inset boxes show significant EnrR binding sites identified over the indicated genes and T3/T6SS loci. The fold enrichment of binding sites around the promoter region of the indicated genes is shown in brackets by comparison of the strain expressing FLAG-tagged EnrR and with that of the FLAG tag alone. The genomic G + C plot is shown in red (window size of 1000 bp for overview or 200 bp for indicated genes). (B and C) MA-plot illustration (B) of the differentially expressed genes in wt and ΔenrR cells grown in DMEM, and the Venn diagram (C) showing genes directly bound and regulated by EnrR as revealed by RNA-seq and ChIP-seq analysis. The log2 of the ratio of the abundances of each transcript between the two conditions (M) is plotted against the average log2 of the abundance of that transcript in both conditions (A). T3/T6SS genes and ChIP-seq-enriched genes are highlighted (B). (D) Functional categories of the EnrR-regulated genes identified by RNA-seq and ChIP-seq. (E) EnrR-binding motif (Logo 1) derived from MEME analysis of the ChIP-seq peaks as well as the respective mutants are shown. (F) MST analysis of EnrR binding to the ChIP-seq consensus probe (Logo 1) and derived mutants (Logo 2–5) (E). Purified EnrR protein labelled with RED-tris-NTA dye was incubated with diluted DNA probes in MST assays (n= 3).

A survey of the ten top scoring peaks (Supplementary Table S3) highlighted the major EnrR-binding gene targets (Figure 5A). The top two binding peaks occurred upstream of its own gene (enriched by 4.2-fold) and the gene ETAE_2212 (ybjD) (4.2), encoding ATP-dependent endonuclease of the OLD family, implying autogenous control and possible roles in the prophage lysogenizing system (46). The third highest scoring site overlaps the promoter region of torT (3.9), encoding the periplasmic sensory protein associated with the TorRS two-component system. Of the top hits, the fourth and eight encode galF (3.9) and agaW (3.8) in the phosphotransferase system (PTS), respectively. The fifth peak is located between the intergenic region of a putative gene involved in a chromosome partitioning protein (ETAE_3427) and a putative integrase encoding gene (ETAE_3428). The sixth top peak is located at genomic islands (GI20) encoding a putative invasion containing featured LysM and Big domains (47). The ninth and tenth highest scoring sites are attributed to mcp involved in chemotaxis and a gene putatively encoding a multidrug efflux pump, respectively. Collectively, these data suggest that EnrR may be involved in various processes. The direct binding of EnrR to these loci—e.g. P, P and P—was further validated using EMSA (Supplementary Figure S4A). Additionally, EnrR seems to bind to fragments of Pwith various sizes and/or G + C contents (Supplementary Figure S4B). Furthermore, the B1H assays, DNase I footprinting and MST further proved the interaction of EnrR with these DNA molecules in vivo and in vitro (Supplementary Figure S4C-E) (47).

Transcriptomics analysis of the wt and ΔenrR strains grown in DMEM indicated that 363 genes were regulated (|log2FC|≥1-fold, Padj< 0.05) (Figure 5B) (Supplementary Table S4), of which 232 genes were upregulated and 131 genes were downregulated in ΔenrR compared with wt (Figure 5B and C). Although the expression of most (103/160) genes highlighted by ChIP-seq was not affected in the ΔenrR strain, T3/T6SS-associated genes were included in the list of genes most prominently regulated by EnrR (Figure 5B) (Supplementary Table S4). COG analysis showed that 31% of the genes controlled by EnrR are hypothetical or have unknown functions (Figure 5D). These results suggest that EnrR is a global regulator that binds to various genes and modulates their expression.

EnrR targets DNA regions harbouring AT-rich consensus sequences

Analysis using the MEME (multiple EM for motif elicitation) algorithm revealed a palindromic consensus (5′-AATAAATNATTTATT-3′) (E < 0.05) in the EnrR-binding sequences revealed by ChIP-seq (Figure 5E). This consensus DNA (Logo 1) showed high affinity for EnrR, and the Kd value for their interaction was assessed as 2.5 ± 0.5 nM by MST analysis (Figure 5F). The EnrR-DNA binding affinity decreased with increasing spacer N length (N = C/G; Logo 2 - Logo 5 designed based on Logo 1), and when its length reached the 4-mer, the binding affinity was barely detected (Figure 5F). These analyses suggested that EnrR targeted DNA fragments with AT-rich consensus.

The overlapping ChIP-seq- and RNA-seq-enriched regions/genes were broadly distributed on the low G + C content regions corresponding to all GIs. Although GI7 (T3SS) and GI17 (T6SS) displayed overall higher G + C contents (62.5% and 64.6%, respectively) than the genome average of 59.7% (16), there was also consistency in EnrR recruitment and a lower G + C content in the ChIP-seq-identified regions, including the evpP promoter region (Figures 5A and 6A). qRT-PCR analysis validated that the transcription levels of selected genes at GI6, GI11 and GI16 were all significantly higher in the ΔenrR strain than in the wt strain and were reduced to that of the wt level in the ΔenrR/P0456-hns strain overexpressing H-NS (Figure 6B). Consistent with this result, we commonly observed that the cultures of E. coli BL21 harbouring the EnrR expression plasmid pET-enrR became transparent and that bacterial growth could be largely restored by coexpressing H-NS (Supplementary Figure S6A). Bacterial lysis and cell wall disruption were observed for the cells expressing EnrR (Supplementary Figure S6B). Following extrachromosomal DNA extraction and phage enrichment from the supernatants of EnrR-expressing E. coli, phage Mu D108-like DNA and the corresponding phage particles were isolated (Supplementary Figure S6C,D) followed by DNA sequencing confirmation.

Figure 6.

EnrR targets AT-rich GIs and gene regions across the chromosome. (A) Circular genomic map of the G + C content (circle 1 from outer circle; above or equal to the median G + C content (blue), less than the median (yellow)), RNA-seq (circle 2; orange indicating EnrR-activated genes and green depicting EnrR-repressed loci) and ChIP-seq (circle 3; the enriched binding sites). The annotated genomic islands (GIs) are shown. (B) Transcript levels of the annotated GI genes in wt-, ΔenrR-, and ΔenrR-overexpressing H-NS (ΔenrR/P0456-enrR) cells grown in LB medium for 12 h. gyrB was used as the internal control. The results shown are means ± S.D. (n = 3). ***, P < 0.001; **, P < 0.01 and *, P < 0.05 compared with wt based on Student’s t-test. (C) AT% content of ChIP-seq peak-associated genes (peaks) compared with the averages of GIs and the chromosome (Chrome). (D) Relationship between averaged AT% values (y-axis) of pooled ChIP-seq peaks (with 100 bp central region; blue square) or the corresponding ORFs (red circle) and mean EnrR enrichment folds (x-axis) determined in the ChIP-seq analysis. The dataset of the EnrR enrichment fold was scaled by 4 and divided into 20 ranges to pool the corresponding ChIP-seq peaks and ORF sequences in each of the regions. (E) The height of the red/blue bars represents the positive/negative coefficient (Spearman’s rho) between the oligonucleotide usage deviations and EnrR recruitment levels in ChIP-seq peaks. Bar scales: −0.171∼0.137 (dinucleotide), −0.167∼0.160 (trinucleotide), −0.299∼0.218 (tetranucleotide). Oligonucleotide compositions of the 100 bp central regions of ChIP-seq peaks were analyzed.

Additionally, statistical analysis of the ChIP-enriched DNA sequence verified that EnrR was more intensively recruited to higher A + T content regions, including the GI regions, than to other regions (Figure 6C). Similar to other NAPs for Gram-negative bacteria (3–5), a positive correlation between AT% and the protein recruitment level was also found for EnrR (Figure 6D; Pearson’s rho = 0.93; P < 0.001 compared with ORFs). We further characterized the potential relationship between the oligonucleotide composition of the ChIP-seq peaks (tetranucleotide, trinucleotide and dinucleotide) and EnrR recruitment levels. TpA dinucleotides (TpA steps), among all dinucleotide steps in AT-rich DNA, confer the most flexibility on DNA by facilitating the accommodation of DNA-interacting residues of DNA binding proteins—e.g. xenogeneic silencers—in the minor groove (3,48). EnrR-enriched sequences harbouring the TpA step and trinucleotides containing TpA had the highest (Spearman’s rho = 0.137; P = 0.00042) positive correlation with the EnrR recruitment level (Figure 6E). Taken together, these results suggest that EnrR is a global regulator prone to target and inhibit the expression of horizontally acquired foreign AT-rich genes.

Structural basis for target DNA binding by EnrR

EnrR shares approximately 45% sequence identity with the NER protein of phage Mu, whereas the sequence similarities between EnrR and other DNA-binding phage proteins, including cro and λ repressor proteins, are low (Figure 7A). To unravel the underlying mechanism of DNA binding by EnrR, we performed crystallographic studies and solved the EnrR-DNA complex structure at high resolution (Table 1). Initial trials of crystallization with EnrR and the 14-mer Logo 1 derived from the MEME analysis of the ChIP-seq data (Figure 5E), and many other putative oligos failed because of low diffraction qualities. We used one self-complementary DNA (DNA1: 5′-CGAAATATCTATAGATATTTCG-3′, the underlined base identical or close to Logo 1) in the crystallization trials. This AT-rich DNA oligo contains the modified Logo 1 DNA sequence and extra CGA in the 5′ upstream region and appears not to be involved in EnrR-DNA binding but may facilitate the stability of the DNA–protein complex with CG pairs. In the complex structure, each DNA1 duplex interacts with two EnrR molecules (Figure 7B). The overall folding of the two EnrR molecules is almost identical, as indicated by the very low root mean square deviation value (rmsd, 0.16 Å). EnrR contains four α-helices (α1- α4, Supplementary Figure S7A), and its surface is highly positive in charge (Supplementary Figure S7B).

Figure 7.

Crystal structure of the EnrR-DNA1 complex. (A) Sequence alignment of EnrR from Edwardsiella piscicida (Edw) using HTH motif-containing proteins from various phages. (B) Overall structure of the EnrR-DNA1 complex. The two EnrR molecules are shown as cartoons in brown and magenta and yellow, respectively. The DNA1 duplex is shown as sticks outlined with 2Fo-Fc electron density maps (contoured at the 1.5 sigma level). (C) Detailed interactions between DNA1 and the HTH motifs of EnrR. H-bond interactions mediated by the EnrR main chain and side chain atoms are indicated by dashed and solid lines, respectively. Nucleobase-specific recognitions are indicated by triangles. (D–I) Detailed interactions between the EnrR HTH motif and DNA.

EnrR binds at the major groove (Figure 7B and Supplementary Figure S7C), forming extensive interactions with the phosphate backbone of the DNA1 duplex (Figure 7C). A5 of DNA1 forms two hydrogen-bond (H-bond) interactions with the side chain of Arg40 (Figure 7D). T6 of DNA1 forms three H-bonds, including one with the side chain of His29 and two with the main chains of Leu35 and Ala36, which are located at the N-terminus of α2. A7 of DNA1 forms one H-bond with the side chain of Arg21 (Figure 7E). Arg21 and His29 reside at the N- and C- termini of α1, respectively. T10* and A11* of the complementary strand interact with the side chains of Tyr57, Arg59 and Arg62 of α4 (Figures 7F,G). Similar to other nucleotides, T12* also forms H-bond interactions with EnrR. However, instead of positively charged residues, T12* interacts with neutral residues Ala45 and Thr48 of α3 (Figure 7G).

In addition to the phosphate backbone, EnrR also recognizes the nucleobases T10* and A11*. The nucleobases of both T10* and A11* form H-bonds that interact with Arg55 (Figure 7H). The distance between the O4 atom of T10* and side chain NH1 atom of Arg55 is 2.8 Å, and the distance is 3.1 Å between the NH2 atom of Arg55 and the N7 atom of A11*. The conformation of Arg55 is further stabilized by its H-bond (2.6 Å) interaction with Asn51. Similar to T10* and A11*, the nucleobase of G14* also interacts with one Arg residue, Arg47, which is inserted into the DNA1 major groove. G14* and Arg47 form two H-bonds, one (2.7 Å) between the G14* O6 atom and Arg47 NH2 atom and the other (2.9 Å) between the G14* N7 atom and Arg55 NH1 atom (Figure 7I).

EnrR mediates conformational changes during DNA interaction

The crystal structures of the repressor of phage λ and cro and the repressor of phage 434 were reported previously (49). EnrR and these phage proteins all contain one HTH motif. However, possibly because of their low sequence similarities (Figure 7A), the overall folding of EnrR is significantly different from that of these proteins (Supplementary Figure S7D), a result supported by the high rmsd values (∼4.0 Å). EnrR is a monomer (Supplementary Figure S2I,J), and cross contacts between other monomers are mediated by Glu54, Arg55, and His56 (Supplementary Figure S8A). Through their side chains, Glu54 of one EnrR and Arg55 of the partner protein form direct H-bond interactions. Mediated by water molecules, Glu54 also forms extensive H-bond interactions with His56 of the partner protein. His56 is the only residue connecting the α3 and α4 helices. The lengths of the α3-α4 connecting loops are similar for EnrR and phage Mu NER protein, whereas they are significantly shorter than other HTH motif-containing proteins (Figure 7A). Perhaps because of the smaller size and tight contact of the α3-α4 loops, the two EnrR molecules pack closely to each other and recognize one potential consensus sequence of 5′-CTATAG-3′ (Figure 7C). Unlike the consensus sequences of the phage λ repressor and phage 434 cro and repressor, which are all very long and contain stretches of nonspecific nucleotides in the middle, the consensus sequence of EnrR is short and continuous.

We found that the DNA1 duplex was severely distorted following EnrR binding. The width of the DNA1 major groove is 22.9 Å at the central cross contact region of EnrR, which is wider than that of the regular B-form DNA duplex (17.6 Å) (Supplementary Figure S8B). Away from the central region, the DNA1 major groove was narrowed and only measured 15.3 Å at the boundary region of EnrR, which is 2.3 Å narrower than that of the regular B-form duplex (Supplementary Figure S8C). In addition to the DNA complex, we also solved one apo-structure of EnrR (Table 1). Although the overall folding of EnrR is similar in the apo- and complex structures, structural superposition could reveal some subtle conformational changes in EnrR (Supplementary Figure S8D). Following DNA binding, the N-terminus of EnrR α3 tilts slightly towards α2, perhaps because of the interaction between Arg47 and the nucleobase of G14* (Figure 7I). One apo NMR structure of phage Mu NER has also been reported (1NER). As revealed by structural superposition (Supplementary Figure S8E), the overall folding of NER is very similar to that of EnrR. Many DNA-interacting residues are conserved in EnrR and NER, indicating that the latter may follow a similar manner in target DNA binding.

The EnrR N-terminus promotes DNA binding and H-NS competition

EnrR is longer than phage Mu NER at the N-terminus for 10 aa residues. The N-terminus of EnrR is also longer than that of cro and the repressor of phage 434 by 16 and 18 aa, respectively (Figure 7A). In the apo-structure, the N-termini of all EnrR molecules are disordered. In the complex structure, the N-terminus of one EnrR molecule is disordered, but it folds into one short helix, α0 (aa 7–16), at the second molecule (Supplementary Figure S7A). α0 does not interact with the DNA1 duplex, but it inserts into the minor groove of the symmetry-related duplex, forming two direct H-bond interactions (Figures 8A–C). As depicted in Figure 8D, one H-bond is formed between the Arg11 NH1 atom and backbone OP1 atom of T8; the other H-bond is formed between the Ser12 OG atom and OP1 atom of C9. Gln17 resides at the α0–α1 linker region; its side chain forms one H-bond interaction with the nucleobase of C9 (Figure 8E). His20, the first residue of α1, also interacts with the backbone phosphate of the symmetry-related duplex (Figure 8F).

Figure 8.

N-terminal α0 facilitates the EnrR interaction with the DNA duplex and competition against H-NS. (A and B) Interactions between EnrR α0 and the symmetry-related DNA duplex. (C) Packing and interactions of EnrR and the DNA1 duplex in the crystal lattice of the complex structure. (D–F) Detailed interactions between EnrR α0 and the symmetry-related DNA1 duplex. (G) Production of T3/T6SS proteins in the wt, ΔenrR or ΔenrR strain expressing FLAG-tagged wt or variant EnrR (Q17G, R47G, R55G, H56G and ΔN17 (deletion of N-terminal 17 residues)) based on the pUT plasmid. The SDS–PAGE resolved proteins were blotted with anti-FLAG specific antiserum (lower panels). RNAP was used as the loading control for the blots. (H) Cy5-P DNA binding capacities of the wt and EnrR variants analysed by MST. The assays were performed in triplicate and are shown as means ± S.D. (I) The N-terminal α0 domain is essential for EnrR competition with H-NS in P DNA binding. The top panel shows EMSA containing Cy5-labelled P DNA and either purified GST-tagged H-NS or EnrRΔN17. In the bottom panel, the EMSA gels were transferred to a PVDF membrane and probed for H-NS using anti-GST antibodies.

Similar to many NAPs (11), EnrR prefers to bind AT-rich DNAs, as confirmed by our abovementioned analyses (Figures 5E,F and 6) and structural data (Figure 7C). Altering the conformation of the target DNA is one common mechanism in regulating gene transcription by NAPs. However, different NAPs may use different strategies to change gene conformation (50–51). In the crystal lattice of the EnrR complex structure, many DNA1 molecules stack on each other, forming a continuous pseudoduplex (Figure 8C). Via their HTH motifs, many cross-contacted EnrR bind the pseudoduplex at the major groove. Next, the N-terminus of EnrR molecules binds to the minor groove of the nearby DNA duplex, leading to the aggregation and/or putative bending of the DNA. Such DNA binding and bending mechanisms are unique for EnrR and have not been documented for other NAPs.

To verify the DNA binding mechanisms observed in the EnrR-DNA1 complex, we characterized several EnrR variants with single-point mutations or deletion of N-terminal residues 1–17 (ΔN17) corresponding to the α0 domain. As illustrated in Figure 8G, the ΔenrR mutant strains expressing wt or mutant forms of EnrR, EnrRQ17G, EnrRR47G, EnrRQ55G, EnrRH17G and EnrRΔN17 showed various levels of T3/T6SS production. The significantly decreased EseB/C/D and EvpC yields in the strains expressing R47G and R55G variants of EnrR and their abolished or markedly decreased DNA-binding capacities further demonstrated their essential roles in the HTH domain-mediated EnrR–DNA interaction (Figures 7H,I and 8G,H). Additionally, the mutants producing EnrRQ17G and EnrRΔN17 generated markedly lower levels of T3/T6SS proteins (Figure 8G), a finding that agrees with the drastically impaired DNA binding in EnrRQ17G (Figure 8H) and EnrRΔN17 (Supplementary Figures S5F, S9A and S9C). Similarly, EnrRΔN17 showed no resistance to the DNA relaxation activity of TopA (Figure 4B) and exhibited a deficiency in DNA oligomerization (Supplementary Figure S5F) compared with wt EnrR. Importantly, EnrRΔN17 could no longer compete with H-NS in P DNA binding (Figure 8I). Collectively, these analyses demonstrated that the unique DNA binding mediated by the N-terminal α0 domain is essential for EnrR to antagonize H-NS and facilitate bacterial virulence gene expression.

DISCUSSION

In the present study, analysis of an in vivo TIS dataset with the PACE algorithm (19) allowed identification of a new virulence regulator, EnrR, in E. piscicida. EnrR is encoded in GI2 and is possibly horizontally acquired (16). We revealed that EnrR derepresses esrB expression by competing with H-NS binding to facilitate T3/T6SS production and in vivo virulence in E. piscicida. The introduction of the enrR gene in E. anguillarum or S. tyhpimurium strains strongly augments their virulence gene expression. EnrR functions as a novel NAP that colocalizes with the bacterial nucleoid. Global profiling of EnrR-DNA binding and gene regulation of various processes indicated that the protein targets AT-rich genomic regions or GIs and represses their expression. Structural analysis confirmed EnrR binding to DNA containing the TA-rich palindrome and illuminated its unique DNA binding ability to both major and minor grooves. Collectively, we show that EnrR is a unique NAP that plays roles as an H-NS-antagonizing virulence activator targeting AT-rich DNA. Thus, our findings suggest that bacterial pathogens have evolved exquisite mechanisms to acquire and manage xenogeneic NAPs to interact with endogenous NAPs and MGEs to balance virulence adaptation and regulatory integrity.

Similar to analogues in Salmonella and other bacterial pathogens, Edwardsiella T3/T6SS gene clusters are crucial GIs horizontally acquired as MGEs and subjected to sophisticated regulation during various growth conditions. Usually, their expression is repressed by factors—e.g. RpoS—under in vitro conditions and during the initial stages of the in vivo infection process (25,52). Previous investigation and evidence from this study also showed that the major xenogeneic silencer H-NS inhibited T3/T6SS production because of its binding to the esrB promoter region and putative binding to other loci on these two GIs (Figure 3) (27–28). The local- and long-distance cis-effects of these H-NS binding or nucleation sites on the promoter activities of esrB and other T3/T6SS promoters—e.g. esrA, esrC, esaR, esaM, evpA and evpP—warrant further characterization. However, the bacterium must overcome these repression mechanisms to activate T3/T6SS expression for in vivo systemic infections or under other conditions requiring these secretion systems for survival. Here, we revealed that E. pisicicida has evolved a mechanism to leverage putatively horizontally acquired EnrR to abrogate H-NS-mediated silencing and activate the esrB promoter and T3/T6SS production.

Currently, Edwardsiella bacteria comprises five species, the T3/T6SS-containing species E. ictaluri and E. piscicida, E. anguillarum and the non-T3/T6SS lineages of E. hoshinae and E. tarda (15). Thus, Edwardsiella is analogous to Salmonella (14,53), the model genus to study GI evolution and regulation. E. ictaluri and E. piscicida encode only one set of T3SS and T6SS genes, while E. anguillrum harbours additional sets of T3/T6SS clusters. Additionally, E. anguillrum showed lower DDH/ANI similarities to E. tarda than those of E. ictalurid/E. piscicida to E. tarda, suggesting that E. anguillarum appeared later than E. piscicida (15). Because E. anguillarum strains do not encode EnrR, we showed that ectopic expression of EnrR in E. anguillarum could augment T3/T6SS yields (Figure 2F,G), further explaining their differences in pathogenesis and virulence gene evolution. Evolutionary scenarios could be proposed that after acquiring T3/T6SS genes, the ancestral Edwardsiella bacteria (representing the EdwGI lineage) (17) could evolve into EdwGII, including E. anguillarum, by the accompanying acquisition of EsrAB, or into E. piscicida by acquiring additional EnrR to allow a higher capacity for T3/T6SS expression by antagonizing H-NS repression (Figure 9A–D). Some strains of notorious enteric pathogens—e.g. S. enterica, E. coli and Yersinia enterocolitica—also encode EnrR homologues (Supplementary Figure S1); similar virulence evolution events involved in these proteins might exist.

Figure 9.

Model of horizontally acquired NAP EnrR interacting with endogenous H-NS to modulate the expression of T3/T6SS and other genes. (A–D) Along with Edwardsiella evolution, the ancestral strain expressing housekeeping H-NS (A) and then acquired T3/T6SS GIs (B), following obtaining of the genes for the two-component system EsrA-EsrB in E. anguillarum (C) and EsrA-EsrB and EnrR in E. piscicida (D). In E. piscicida, the acquisition of EnrR antagonizes H-NS repression, leading to high T3/T6SS production and pathogenicity toward hosts. (E) Summary of EnrR target genes and related expression profiles. Green and purple solid lines indicate activation and repression, respectively.

Here, using RNA-seq and ChIP-seq, we defined the EnrR control of gene expression in E. piscicida (Figure 9E). Among the 131 EnrR-upregulated genes, T3/T6SS genes ranked at the top of the list. EnrR also positively controls the PhoP/PhoQ two-component system, which is involved in global gene expression (54). Lipid transport and metabolism, F0F1-type ATPase, and lysogeny/lytic switch systems are also under direct or indirect control by EnrR (Supplementary Tables S3 and S4). However, EnrR negatively regulates ∼232 genes classified into several key processes, including flagellar biosynthesis, pilus assembly, biofilm formation, chemotaxis, antibiotic resistance and DNA recombinase, which have been established as MGE genes (Supplementary Tables S3 and S4). Many of these genes are located in GIs and may be essentially involved in the pathogenesis of the bacterium. The promoters of genes associated with all 24 GIs recruit EnrR and are under direct control by the protein (Figure 6A). Additionally, the core genes related to metabolic processes, including the sugar phosphotransferase system (PTS) and acetyl-CoA acetyltransferase, and those linked to cell division and cell wall biosynthesis, were also subject to EnrR regulation (Figure 5D). Furthermore, EnrR can regulate the lysogeny/lytic lifestyle of the prophages in E. coli (Supplementary Figure S6) and enhance virulence in S. Typhimurium SL1344 by activating SPI-2 expression (Supplementary Figure S3). These analyses feature EnrR as a master regulator and an NAP globally controlling gene expression (Figures 6A and 9E).

It is intriguing to examine the underlying mechanisms by which EnrR counteracts H-NS silencing on the esrB promoter. Bacterial pathogens deploy many counter silencers, the factors that counteract H-NS or other xenogeneic silencers on foreign gene expression for dynamic regulation of virulence gene expression during in vivo infection or in vitro growth. Salmonella employs countersilencers, including SsrB (32,45), PhoP (53), LeuO (55), CsgD (56) and SlyA (57), to counteract H-NS-mediated repression of virulence gene expression in response to various signals. Currently, we cannot exclude EsrB and other countersilencers that may be used to antagonize the H-NS repression of esrB and other promoters associated with T3/T6SS in an EnrR-dependent manner in Edwardsiella bacteria. The intrinsic promoter architecture of esrB is unique because it contains several binding sites for EnrR and H-NS and accommodates multiple EnrR/H-NS molecules (Figure 3C). Additionally, the binding sites of EnrR and H-NS appear not to overlap (Figure 3C–E) and may require EnrR’s N-terminal extension mediated minor groove contacts and cooperative DNA binding in the far apart DNA duplex to form DNA loop and compaction for its occlusion of H-NS binding to DNA minor grooves (Figure 3F–G). Thus, esrB could be derepressed under physiologically permissive conditions (14). Notably, the ChIP-seq enriched peak mapped to the esrA-esrB intergenic region did not exactly match the neighbouring locus of the lowest G + C content in P (Figure 5A). We hypothesized that EnrR might mediate long-range interactions between these cis-elements, a claim that warrants future investigation. Alternatively, because EnrR appeared to bind the promoter region of esrB and augment its transcription in an H-NS-independent manner (Figure 3A, lanes 9–10), it could also play a role in transcriptional activation by direct recruitment of RNAP to P. How these NAPs or transcriptional regulators intertwine and coordinate to tightly control esrB transcription in a direct activation mode or silencing-anti-silencing manner as well as other genomic transactions, such as chromosome supercoiling (58–59), remains to be investigated.

Usually, H-NS silences gene expression by occluding and blocking RNAP transcription initiation of specific promoter regions from the ends of a stiffened nucleoprotein filament (60). Counter silencers such as PhoP (53), VirB (61) and SsrB (45) use disruptive countersilencing mechanisms to alter the DNA–protein complex structure to disrupt H-NS-mediated silencing, which directs the binding or processing of RNAP. However, H-NS remains bound or condensed to a promoter region even under countersilencing conditions by these proteins (53,60). This finding is different from EnrR, which largely evicts H-NS from the promoter (Figure 3F). Given that EnrR does not interact with H-NS directly (Supplementary Figure S4F), we also hypothesized that EnrR might not prevent H-NS oligomerization to derepress gene expression similar to H-NST-like proteins (62–63), Ler (64) and LeuO (65). AFM also demonstrated that the DNA-binding and oligomerization modes of EnrR are different from those of Ler. Although EnrR forms large particles with DNA at high concentrations, it appears to wrap, stiffen and bridge DNA at lower concentrations (Figure 4D and Supplementary Figure S5A–E). Ler protein forms compact particles wrapped by contour-length shortened DNA (66). Additionally, the dynamic derepression mechanisms by which EnrR induces T3/T6SS in response to various cues during the infection state warrant future investigation. Collectively, these analyses suggested that EnrR might adopt a distinct mechanism to compete for H-NS binding to the esrB promoter and activate its transcription.

Xenogeneic silencers are a group of specific NAPs (67). To date, five families of xenogeneic silencers have been identified, such as H-NS, in numerous bacteria that selectively repress the expression of foreign sequences and contribute to bacterial genome evolution (48,68). Xenogeneic silencers share a domain architecture common with an N-terminal oligomerization domain and C-terminal DNA binding domain that facilitate cooperative binding to high- and low-affinity DNA sites and promote the formation of nucleation filaments, chromosome compaction and gene silencing (48,69). Although EnrR targets AT-rich DNA fragments, we hypothesized that EnrR, as a novel NAP, might not be classified as a typical xenogeneic silencer because of its apparent lack of an oligomerization domain (Figure 1C and 7A). Additionally, EnrR binds to both DNA major and minor grooves with its HTH domain and the N-terminal extension (α0) residues, respectively (Figure 7), in contrast to documented xenogeneic silencers that specifically bind to the minor grooves of DNA targets (48). The observed TpA steps in the EnrR target DNA (Figure 6E) explain its favoured binding to the minor grooves (48) by the N-terminal extension (α0) residues. Given the essential roles of the N-terminal helix (α0) in EnrR DNA binding (Figure 8I; Supplementary Figures S5F and S9C) and gene regulation and the observed tandem binding sites upstream of the esrB promoter region, this unique DNA binding mode might facilitate DNA looping, bridging or even compaction to organize GI regions or chromosomes in a higher-order hierarchy. Whether EnrR can cooperatively interact with or facilitate the cooperativity of xenogeneic silencers—e.g. H-NS—to bind and bridge DNA remains to be determined.

DATA AVAILABILITY

Accession numbers are available for the SRA data, PRJNA734738 and PRJNA767449 for RNA-seq and ChIP-seq, respectively, and the PDB structural data, 7F9I and 7F9H for apo-EnrR and EnrR-DNA complex, respectively.

Click here for additional data file.