Nucleoid structure and dynamics influence natural chromosomal transformation in Bacillus subtilis: the role of EbfC

Abstract

Natural chromosomal transformation (NCT) in Bacillus subtilis requires RecA and its accessory proteins including RecX and RecD2. Inactivation of the nucleoid-associated protein (NAP) Hbsu, Rok, or LrpC results in a dual effect: it enhances NCT but exacerbates the NCT defect of ΔrecX cells. Purified EbfC exhibits characteristic features of NAPs: it binds both single- and double-stranded DNA, protects them from degradation, and forms higher-order protein–DNA complexes via its DNA bridging activity. NCT is reduced upon EbfC inactivation but enhanced by ebfC overexpression. hbs55, Δrok, or ΔlrpC mutations suppress the NCT defect of ΔebfC, yet synergistically increase NCT upon ebfC overexpression. The NCT defect in ΔrecD2 cells is worsened by ebfC overexpression or by hbs55, Δrok, or ΔlrpC mutations. The nucleoid was more compacted in ΔebfC cells, an effect counteracted by hbs55, Δrok, or ΔlrpC mutations. EbfC contributes to DNA repair, and ebfC is epistatic to hbs or lrpC in response to DNA damage. We propose that chromosome folding, modulated by NAPs, plays a critical role in NCT and DNA repair. In this context, EbfC, by regulating nucleoid dynamics as a NAP, opposes the functions of Hbsu, Rok, and LrpC in NCT, while their interconnected roles contribute to DNA repair.

Graphical Abstract

Graphical Abstract.

Introduction

Bacterial chromosomes are often mosaics, containing patches of horizontally acquired foreign DNA that contribute to genomic diversity [1]. Natural transformation—the non-sexual uptake and exchange of genetic material—is a key mechanism for acquiring genetic diversity, mitigating genetic interference, and enhancing adaptation to new ecological niches [2–4]. Indeed, natural transformation plays a crucial role in the dissemination of antibiotic resistance genes, the emergence of pathogenic strains, and vaccine escape [2–4]. However, the acquisition of novel and horizontally transferred chromosomal DNA among naturally competent cells incurs biological costs [2, 4]. The integration and/or maintenance of the genetic material may be constrained by biological barriers, reflecting a trade-off between the fitness advantage conferred by the new trait and the potential cost of its expression. Bacillus subtilis natural chromosomal transformation (NCT) presents a unique system for studying such antagonistic situations. Unless otherwise indicated, all genes and gene products mentioned are of B. subtilis origin.

Natural competence is developed by a stochastic subpopulation of B. subtilis cells [5] and is characterized by the following features: (i) DNA replication is halted; (ii) global transcript elongation is significantly reduced, although expression of <4% of total genes is selectively increased, including recA, ssbA, ssbB, dprA, and rok; (iii) nucleoid compaction is induced, and messenger RNA (mRNA) levels of DNA topoisomerases are moderately decreased; (iv) adaptive immune systems such as CRISPR–Cas are absent; and (v) since DNA is internalized as single-stranded (ss)DNA, restriction-modification systems are generally ineffective at fragmenting it [5–10].

The recombinase RecA assembles on the incoming ssDNA to form a nucleoprotein filament, the growth of which is negatively regulated by modulators such as RecX and RecD2 [11–13]. The resulting filament performs a homology search on the compacted nucleoid [7, 14]. When a region of homology with a minimal efficient processing segment (MEPS) [≥30-nucleotide (nt) long homologous region] is found between the internalized ssDNA and the duplex recipient genome [15–17], homology-directed, RecA-dependent NCT is initiated [5, 7]. In this context, the RecA nucleoprotein filament facilitates pairing of the donor ssDNA with its complementary recipient dsDNA, catalysing the integration of homologous (intraspecies NCT) or homeologous (interspecies NCT) donor DNA into the recipient chromosome through a single homologous recombination (HR) event [18–20]. Alternatively, a two-step HR event—contributing to pan-genome expansion—may mediate the integration of heterologous DNA flanked by two homologous regions (>400-nt each) with the recipient genome. This form of NCT, also processed by the HR machinery, occurs with ∼10-fold lower efficiency than a single homologous gene replacement event and will not be further discussed here [7].

If the internalized ssDNA is heterologous to the recipient chromosome but carries a functional replication origin, it can be reconstituted into a circular duplex molecule prior to its establishment via RecA-independent natural plasmid transformation (NPT) [21, 22]. However, in the absence of an active origin of replication or a site-specific integration system, the incoming ssDNA is degraded [5, 7].

The frequency of NCT involving homogeneously divergent DNA decreases exponentially with increasing local sequence divergence (SD) between donor and recipient DNA sequences [23, 24]. This frequency plateaus at ∼15% of homogeneously distributed SD, and no DNA strand exchange (DSE) is observed beyond 23% SD [19, 25]. A higher rate of gene transfer involving homogeneously divergent DNA—if left unregulated—could be detrimental. Accordingly, naturally competent cells are thought to have evolved barriers and counter-barriers to regulate the integration of incoming DNA into the host genome. At the local level, a discontinuous RecA filament interacts with and facilitates intersegmental homology recognition via an 8-bp foothold, with each RecA protomer stretching and transferring DNA in 3-nt steps [26, 27]. A single mismatch within one such 8-bp micro-homologous recognition unit only delays homology-directed RecA-mediated DSE, both in vivo and in vitro [19, 26–30]. These DNA mismatches have limited impact on homeologous recombination, as the mismatch repair system is either inactive in non-replicating competent B. subtilis cells or readily saturated during NCT [30, 31]. However, when ≥2 consecutive micro-homologous 8-bp steps each contain at least two mismatches, the associated energetic penalty cannot be overcome. This leads to the stripping of the dsDNA or RecA from the D-loop intermediate [26, 27, 29]. The integration of transforming ssDNA is likely halted when local SD level exceeds 20% across a 24-nt region (three consecutive footholds) [10, 30]. This explains the observed 30- to 40-fold reduction in interspecies NCT frequency and the shorter length of DNA integrated when using donor DNA with an overall SD of 8.35% and 18 MEPS ranging from 157 to 31 nt in length [19, 25, 30].

Other factors also contribute to the delicate cost-benefit balance of NCT. As demonstrated in our recent studies, inactivation of several unrelated nucleoid-associated proteins (NAPs)—which exhibit DNA bending, looping and bridging properties in vitro—significantly enhances both intra- and interspecies NCT [10]. The architecture and internal organization of the bacterial nucleoid, as well as its relationship with the cytoplasm, remain unclear. Numerous DNA-interacting proteins—including NAPs, as well as topoisomerases, RNA and DNA polymerases, etc.—likely form platforms on the nucleoid surface that accommodate various functional chromosomal regions [32–35]. These NAPs include Rok, LrpC (functional homolog of Lrp_Eco), and the essential Hbsu (gene name hbs, functional counterpart of HUαβ_Eco or HUβ_Eco) (see Supplementary Annex S1). Any combination of a leaky mutation in hbs (hbs55, Hbsu R55A) or inactivation of lrpC or rok—none of which affect the processing of incoming DNA—synergistically increases both intra- or interspecies NCT [10]. Conversely, simultaneous absence of Rok, LrpC, or Hbsu along with the RecA negative modulator RecX leads to a synergistic reduction in both intra- and interspecies NCT, with no significant impairment in ssDNA internalization, as evidenced by unchanged NPT frequencies [10]. This strongly suggests a functional interplay between NAPs and the HR machinery. We have proposed that Rok, Hbsu, and LrpC act sequentially or cooperatively to transiently condense the nucleoid, thereby altering the local chromosomal organization in competent cells through three principal mechanisms: DNA bridging (Rok) [36, 37], DNA bending (Hbsu) [38, 39], and DNA wrapping (LrpC) [40, 41]. In doing so, these NAPs may hinder the formation of active recombination structures required for chromosomal acquisition of donor ssDNA, acting as a failsafe to RecA-mediated DSE, rather than silencing the expression of the integrated genes [10]. If this hypothesis holds, a counteracting NAP—one that promotes nucleoid decondensation and facilitates DSE—should thereby enhance NCT, antagonizing the role of Rok, Hbsu, or LrpC. To test and extend this hypothesis, we searched for additional NAP(s) that could be involved in modulating NCT. Additionally, we revisited the role of negative RecA modulators that might influence the effects of NAPs. A potential role in NCT had previously been ruled out for three other NAPs: AbrB, Dps, and Smc-ScpAB [10]. Protein mining identified a candidate NAP in B. subtilis, YaaK, a protein of unknown function which shares sequence identity and predicted architecture (Supplementary Figs S1 and S2) with members of the widely conserved EbfC/YbaB protein family, which display NAP-like activities in various bacterial species due to its capacity to bind dsDNA as a multimer, bridge and condense DNA, thereby altering DNA conformation [42–46]. Since YaaK is annotated in several databases under the EbfC/YbaB family, and the Y-nomenclature often varies even among species of the same genus, we will henceforth refer to YaaK as EbfC for simplicity.

The widespread presence of EbfC orthologues across diverse bacterial taxa has led to the hypothesis of a conserved kingdom-wide role for this protein. However, its precise biological relevance remains largely undefined [45, 46]. EbfC is not likely to function as an enzyme, as its orthologues lack conserved residues necessary to form functional groups and assemble a catalytic machinery. Within the EbfC/YbaB protein family, Borrelia burgdorferi EbfC_Bbu seems to be essential, whereas its functional analogues in Haemophilus influenzae (YbaB_Hin), Escherichia coli (YbaB_Eco), Caulobacter crescentus (YbaB_Ccr), B. subtilis (EbfC), Deinococcus radiodurans (EbfC_Dra), and Mycobacterium tuberculosis [EbfC_Mtu (a.k.a. Rv3716c)] are dispensable [42–49].

Understanding the role of NAPs in various DNA processes—such as replication, transcription, chromosome segregation, and HR—is complicated by the simultaneous occurrence of these events. B. subtilis NCT provides a unique framework to study the contribution of NAPs to HR independently of their roles in other DNA transactions. In this study, we investigate whether B. subtilis EbfC is a NAP, whether it acts antagonistically or collaboratively with Hbsu, LrpC, and Rok in modulating DNA topology, and how NAPs contribute to NCT. We demonstrate that EbfC binds and bridges ssDNA and dsDNA in an apparent sequence-independent manner in vitro. In vivo, EbfC co-localizes with the nucleoid. Loss of EbfC reduces, while increased gene dosage (ebfC⁺⁺ strain) enhances both intra- and interspecies NCT relative to the wt control. Inactivation of the negative modulator RecD2 synergistically impairs both forms of NCT in ebfC⁺⁺, hbs55, ΔlrpC, or Δrok cells. While ebfC deletion leads to nucleoid compaction, cells harbouring ebfC⁺⁺, hbs55, ΔlrpC, or Δrok mutations exhibit nucleoid relaxation. When combined, the latter mutations additively increase nucleoid decondensation. hbs55, ΔlrpC, or Δrok mutations partially restore nucleoid decondensation in the ΔebfC context. These findings suggest that EbfC and Hbsu, LrpC, or Rok exert opposing effects on chromosomal DNA integration and nucleoid compaction in competent cells. Moreover, in exponentially growing cells—where ongoing replication and transcription influence global supercoiling—ebfC deletion renders cells sensitive to 4-nitroquinoline-1-oxide (4NQO)-induced DNA distortions, mitomycin C (MMC)-induced intra- and inter-strand cross-links, and H₂O₂-induced oxidative damage and DNA nicks, but not to methyl methanesulfonate (MMS), a DNA alkylating agent. Notably, ebfC is epistatic to hbs, rok, or lrpC in response to DNA damage. We propose that EbfC functions as a NAP that counterbalances the roles of Hbsu, Rok, and LrpC during genetic recombination in competent cells, while cooperating with other NAPs to facilitate DNA repair. Given that only Hbsu and EbfC are broadly conserved across bacteria, it is likely that these interdependent proteins are critical in maintaining nucleoid architecture.

Materials and methods

Bacterial strains and plasmids

The B. subtilis strains used in this study, which lack extrachromosomal elements and some prophages, are listed in Supplementary Table S1. Hbsu is an essential protein, but its Hbsu R55A mutant variant (encoded by the hbs55 gene) is viable [50]. A null ebfC mutation (ΔebfC) and the ebfC gene cloned under the control of the strong hybrid isopropyl β-d-1-thiogalactopyranoside (IPTG)-inducible promoter P_spac (hereafter referred to as P-ebfC for simplicity) were kindly provided by Prof. Kei Asai (Tokyo University of Agriculture, Tokyo, Japan) [51]. The ΔebfC mutation was introduced into the BG214 strain background by SPP1-mediated generalized transduction [52], resulting in strain BG1937. The P-ebfC cassette was ectopically integrated into the amyE locus of either BG1937 or BG214, generating strains BG1951 and BG683, respectively. BG1951 carries a single copy of the IPTG-inducible ebfC gene (ΔebfC P-ebfC). BG683 contains two copies of ebfC: one at its native locus, which is relatively highly expressed [8], and a second copy under the control of the IPTG-inducible promoter (ebfC⁺P-ebfC genotype). This strain and its derivatives are referred to here as ebfC⁺⁺ to emphasize the presence of two gene copies. The ΔebfC mutation or the P-ebfC cassette was subsequently transferred to ΔrecD2, Δrok, ΔlrpC, or hbs55 strains by SPP1-mediated generalized transduction. All genetic constructs were confirmed by sequencing the disrupted gene regions. Strains harbouring the P-ebfC cassette were cultivated and plated on medium containing 500 μM IPTG.

To determine whether EbfC co-localizes with the nucleoid, the gene encoding monomeric superfolder GFP was fused to the 3′-terminus of the ebfC gene, generating an ebfC–msfgfp gene fusion, which was used to replace the native ebfC gene at its original locus in an otherwise wt background. Notably, cells expressing this fusion protein exhibit wt behaviour in vivo, indicating that the fusion is functional (Supplementary Tables S3 and S5).

E. coli BL21(DE3) cells bearing pLysS and the pET3b-based plasmid expressing ebfC (pCB1237) were used to overproduce native EbfC. The plasmid pGEM3-Zf(+) was used as a source of circular ssDNA (cssDNA) and negatively supercoiled dsDNA (scDNA), as previously described [53]. EcoRI-linearized pGEM3-Zf(+) served as a source of linear dsDNA (ldsDNA). Plasmid pHP13 [54], digested with BsmI, MslI, StuI, and BbsI, was used as a source of ldsDNA fragments of various sizes and sequences for binding specificity assays.

Restriction enzymes, DNA ligases, DNA polymerases, and DNase I were purchased from New England Biolabs (Ipswich, MA, USA). Plasmids were purified using the Qiagen Miniprep kit (Hilden, Germany).

Donor DNAs and recipient strains

To minimize potential interference in gene transfer from defence systems targeting genetic parasites and from kin discrimination during interspecies NCT, purified donor DNAs encoding an essential, easily selectable gene were employed, along with isogenic recipient strains devoid of major mobile genetic elements, such as bacteriophages and conjugative transposons [25].

As donor DNA for intra- and interspecies NCT assays, the promoter-less 2997-base pair (bp) coding region of the essential housekeeping rpoB gene, which encodes the β-subunit of RNA polymerase (RNAP), was selected. A single C-to-T transition mutation was introduced at codon 482 (H482Y, rpoB482 gene), conferring rifampicin resistance (Rif^R), which serves as a convenient selectable marker [25]. This mutation is centrally located at position 1443 within the coding sequence of the rpoB482 gene. For intraspecies NCT, homologous rpoB482 DNA from B. subtilis 168 (Bsu 168 rpoB482, 0.03% SD, with a single mismatch relative to the recipient’s native rpoB gene) was used. For interspecies NCT, homeologous DNA from Bacillus atrophaeus 1942 (Bat 1942 rpoB482, 8.35% SD, containing 250 mismatches relative to the recipient’s native rpoB gene distributed nearly homogeneously across the gene) was used [25]. These plasmid-based rpoB482 DNAs were purified from E. coli cells. Plasmids were of equal length and base composition. As B. subtilis cells exhibit no codon usage bias, and the B. atrophaeus RNAP β subunit shares 99% amino acid sequence identity with its B. subtilis counterpart, no impairment in translation is anticipated [10, 25, 30].

For NPT assays, the 4548-bp Firmicutes (a.k.a. Bacillota) plasmid pUB110 DNA, which confers neomycin resistance (Nm^R), was used. pUB110 was purified from B. subtilis cells. Because monomeric plasmid DNA is inactive for transformation in B. subtilis [21], the plasmid was linearized using EcoRI and subsequently self-ligated with T4 DNA ligase under high DNA concentration conditions to artificially promote oligomerization. Approximately half of the resulting molecules are ligated in a head-to-tail configuration, rendering them active for transformation.

Transformation assays

Naturally competent B. subtilis cells were prepared as previously described [55]. Competent cell cultures [0.5–1 × 10⁹ colony-forming units (CFUs)·ml⁻¹] were incubated with 0.1 μg·ml⁻¹ of either the homologous Bsu 168 rpoB482, the homeologous Bat 1942 rpoB482, or the plasmidic and oligomeric pUB110 DNAs for 60 min at 37 °C. Following incubation, cells were plated on Rif-containing (8 μg·ml⁻¹) plates for NCT or Nm-containing (5 μg·ml⁻¹) plates for NPT, incubated overnight (o/n, 16–18 h) at 37 °C [55, 56], and CFUs were counted. Transformation frequency was calculated as the ratio of transformants to the total number of CFUs, and data are presented as mean ± SD.

The proportion of the cell population that transiently develops natural competence can range from 0.1% to 10% of the total cell population [55]. To normalize DNA uptake across strains, the fraction of competent cells was estimated by quantifying the uptake of radiolabelled donor DNA (measured as DNase I-resistant radiolabelled linear dsDNA internalized by competent cells), and total cell numbers were determined as previously described [55, 56]. However, this normalization introduces a certain degree of noise, and thus changes in transformant numbers of <3-fold are considered poorly significant. Statistical significance was assessed using t-tests.

Mapping of integration lengths

To map integration lengths and recombination endpoints, the rpoB gene from Rif^R transformants was amplified by polymerase chain reaction (PCR) and sequenced. The resulting nucleotide sequences were compared with those of the recipient and donor DNAs to determine, at each position, whether the sequence corresponded to the recipient or the donor. Recombination endpoints were defined as previously described, and the integration length was calculated as the distance between the two endpoints [57]. For each strain, a minimum of five independent Rif^R clones were sequenced. Statistical significance was assessed using t-tests.

Survival assays

To analyze bacterial growth, cultures were diluted to an OD₅₆₀ of 0.05 and incubated at 37°C for 300 min in both LB rich and GM defined medium [57]. OD₅₆₀ readings were taken every 30 min, and the resulting data were plotted against time to generate growth curves for each bacterial strain. When cultures reached an OD₅₆₀ of 0.4, appropriate dilutions were plated on LB agar, incubated o/n at 37°C, and CFUs were counted to assess plating efficiency. Statistical analysis was evaluated using a two-tailed Student’s t-test.

To determine cell sensitivity to chronic exposure to DNA-damaging agents (MMS, 4NQO, MMC, or H₂O₂, Merck, Darmstadt, Germany), cultures were grown at 37°C with agitation to an OD₅₆₀ of 0.4 and appropriate dilutions plated onto LB agar supplemented with increasing concentrations of MMS (1.5–2.0 mM), 4NQO (0.1–0.4 μM), MMC (50–150 nM), or H₂O₂ (0.2–0.6 mM), as described previously [58]. Plates were incubated o/n at 37°C and CFUs were counted.

For acute MMS or 4NQO sensitivity, cultures were grown in LB at 37°C with agitation to an OD₅₆₀ of 0.4 and exposed to MMS (20–60 mM) or 4NQO (100–300 μM) for 15 min. Cultures were then diluted and plated on LB agar. Plates were incubated o/n at 37°C and CFUs were counted.

All experiments were performed independently at least four times. The surviving fraction was calculated as the ratio of drug resistant CFUs to the total number of CFUs, and data are presented as mean ± SD. Statistical analyses were performed using a two-tailed Student’s t-test.

Fluorescence microscopy

Bacterial cultures were grown exponentially in LB medium at 37°C with shaking until reaching an OD₅₆₀ = 0.4. To analyze nucleoid area, cells were harvested, fixed with 2% formaldehyde, and stained with 4′,6′-diamino-2-phenylindole (DAPI) (Merck, Darmstadt, Germany) (1 μg·ml⁻¹) and FM4-64 (Molecular Probes, OR, USA) (1 μg·ml⁻¹). When indicated, cells were acutely exposed to MMS or 4NQO for 15 min at 37°C prior to fixation. To examine EbfC localization, cells were collected, washed with phosphate-buffered saline (PBS), and stained with DAPI.

Samples were mounted on poly-L-lysine-coated coverslips and visualized using a DMI6000B fluorescence microscope (Leica, Wetzlar, Germany) equipped with an ×100 immersion oil objective and appropriate filters (CNB Advanced Optical Microscopy Facility, Madrid, Spain). Images were captured with an OrcaR2 Digital Camera C7780 (Hamamatsu Photonics, Shizuoka Japan).

Image analysis was performed using ImageJ software (NIH, USA), which was employed to merge phase contrast, FM4-64, and DAPI-fluorescence images, and to quantify both the proportion of the cell area occupied by the nucleoid and the degree of co-localization between EbfC and the nucleoid. Statistical significance was assessed using t-tests.

EbfC purification and analysis of oligomerization

EbfC was expressed in E. coli BL21(DE3) [pLysS] cells harbouring plasmid pCB1237. Cultures were grown in LB at 37°C with agitation to an OD₆₀₀ of 0.5, then induced with 1 mM IPTG for 120 min at 37°C. To reduce host protein expression, Rif (200 μg·ml⁻¹) was added 30 min after induction. Cells were harvested by centrifugation, resuspended in buffer A [50 mM Tris–HCl (pH 7.5), 1 mM 1,4-dithiothreitol (DTT), and 10% glycerol] containing 500 mM NaCl, and lysed using a French press. Following centrifugation to remove cell debris, the traces of nucleic acids in the supernatant were precipitated by polyethyleneimine, while EbfC remained soluble. The supernatant was subjected to 50% saturation ammonium sulfate precipitation; EbfC remained in solution and was subsequently precipitated at 70% saturation ammonium sulfate. EbfC pellet was resuspended and loaded onto a Q Sepharose Fast Flow Column (Cytiva, MA, USA) equilibrated with buffer A containing 50 mM NaCl. The column was washed with buffer A containing increasing concentrations of NaCl (50, 100, and 200 mM), and pure EbfC was eluted using buffer A containing 250 mM NaCl. Fractions containing EbfC were pooled and dialyzed against buffer B [50 mM Tris–HCl (pH 7.5), 1 mM DTT, and 50% glycerol] containing 300 mM NaCl. The purified protein was aliquoted and stored at −20°C. EbfC was judged to be >99% pure by SDS–PAGE (Supplementary Fig. S4A) and quantitative analysis of polypeptide composition by LC-ESI-MS/MS (CNB Proteomics Facility, Madrid, Spain), and showed no detectable contamination with EbfC_Eco protein. The concentration of dimeric EbfC was determined using an extinction coefficient of 11780 M⁻¹ cm⁻¹ at 280 nm, as previously described [59].

EbfC oligomerization was assessed by suberic acid cross-linking. Briefly, 2 μg of EbfC (4 μM) was incubated in buffer C [50 mM Tris–HCl (pH 7.5), 25 mM NaCl, 1 mM DTT, and 5% glycerol] with 10 mM MnCl₂ and increasing concentrations of suberic acid (15–120 μM) for 15 min at 37°C. Cracking buffer [1% SDS, 1% 2-mercaptoethanol, 0.004% bromophenol blue, 10% glycerol, and 50 mM Tris–HCl (pH 6.8)] was added, and samples were resolved on 17.5% SDS–PAGE at 200 V for 60 min in running buffer (25 mM Tris, 200 mM glycine, and 0.1% SDS) at 22°C. Gels were stained with Coomassie Brilliant Blue.

In vivo EbfC protein quantification

Cells were grown in LB medium at 37°C with agitation to an OD₅₆₀ of 0.4, in the presence or absence of 500 μM IPTG. Aliquots of 3 ml were harvested by centrifugation, resuspended in 150 μl of PBS, and lysed by sonication. Cellular extracts from each experimental condition, normalized to equivalent total protein concentrations, were separated by 15% SDS–PAGE alongside purified EbfC protein standards (0.6–2.5 ng). Gels were either stained with Coomassie Brilliant Blue or transferred (300 mA, 100 min, 4°C) to Immobilon-P polyvinylidene fluoride (PVDF) membranes (Merck, Darmstadt, Germany) for subsequent western blot analyses. Western blots were probed with a rabbit polyclonal anti-EbfC primary antibody (120 min, 22°C) (Davids Biotechnologie, Regensburg, Germany), followed by a goat anti-rabbit IgG-horseradish peroxidase conjugate secondary antibody (60 min, 22°C) (Merck, Darmstadt, Germany). Signal detection was performed using the Clarity Western ECL Substrate (Bio-Rad, Hercules, CA, USA).

EbfC-specific bands were visualized and quantified using a ChemiDoc Touch Imaging System and ImageLab software (Bio-Rad, Hercules, CA, USA). No signal was detected in extracts lacking EbfC expression, indicating that the polyclonal anti-EbfC antibody did not exhibit cross-reactivity with unrelated proteins. A linear correlation was observed between signal intensity and EbfC concentration in the purified standards, enabling quantification of EbfC levels in cell extracts by interpolation from the standard curve. To estimate the intracellular concentration and the number of molecules/cell of EbfC, the total number of CFUs (1 × 10⁸ CFUs/ml) and the molecular weight of EbfC (11.64 kDa) were used. Since most CFUs consisted of single or non-separated paired cells, an average of ∼1.6 cells/CFU was applied as a correction factor [60].

Measurement of EbfC–DNA complexes, DNA bridging, and DNA ligation in bulk

EbfC binding to DNA was assessed by electrophoretic mobility shift assays (EMSAs). DNA substrates (3.1 nM in molecules) were incubated with increasing concentrations of EbfC in buffer C containing increasing concentrations of the indicated metal ion or EDTA for 15 min at 37 °C in a 20-μl reaction mixture. Reactions were halted by the addition of loading buffer [50 mM Tris–HCl (pH 7.8), 30% glycerol, 0.25% bromophenol blue, and 0.25% xylene cyanol].

DNA protection from DNase I digestion was analyzed using the pGEM3-Zf(+) plasmid (ldsDNA and scDNA). DNA (3.1 nM in molecules) was incubated with increasing concentrations of EbfC in buffer C containing 10 mM MnCl₂ for 15 min at 37 °C in a 20-μl reaction mixture. Subsequently, 2.5 mM MgCl₂, 0.5 mM CaCl₂, and 0.05 U DNase I were added, and samples were incubated for 5 min at 22°C.

DNA bridging activity was tested using pGEM-3Zf(+) plasmid (ssDNA and scDNA). DNA (3.1 nM in molecules) was incubated with increasing concentrations of EbfC in buffer C containing 10 mM MnCl₂ or EDTA for 15 min at 37 °C in a 20-μl reaction mixture. Reactions were centrifuged at 21 300 × g for 10 min to separate pellet and supernatant fractions.

DNA ligation was analyzed using EcoRI-linearized pGEM3-Zf(+) plasmid DNA (1.8 nM for intermolecular ligation or 0.3 nM for intramolecular ligation of DNA molecules) and was incubated with increasing concentrations of EbfC for 15 min in buffer C containing 1 mM MnCl₂ at 37 °C in a 20-μl reaction mixture. Thereafter, 10 mM MgCl₂, 1 mM ATP, and 3U T4 DNA ligase were added, and the mixture was incubated for 3 h at 22°C.

For DNA protection, bridging, and ligation assays, reactions were deproteinized with 1 mg/ml proteinase K, 20 mM EDTA, and 1% SDS for 5 min at 37°C prior to the addition of loading buffer. All DNA products were resolved by 0.8% agarose gel electrophoresis at 100 V for 60 min in TAE buffer at 22°C and visualized by ethidium bromide staining using a GelDoc Imaging System and ImageLab software (Bio-Rad, Hercules, CA, USA).

Atomic force microscopy analyses

The pGEM-3Zf(+) ssDNA and scDNA (0.7 nM in molecules) were mixed with increasing concentrations of EbfC (2.5–10 μM) in buffer C. Subsequently, 10 mM MnCl₂ was added, and the reaction was incubated for 15 min at 37°C in a 20-μl reaction mixture. Samples were then deposited onto freshly cleaved mica surfaces pre-coated with 1 mM spermidine, subsequently dried under a gentle stream of nitrogen gas, washed with water, and re-dried.

DNA–protein complexes were examined in air at 22°C using a Bruker Nanoscope VIII (Bruker ASX, Germany) atomic force microscopy (AFM) system. Imaging was performed in tapping mode using a 15 μm scanner (E scanner). Probes consisted of single silicon crystal with a cantilever length of 160 μm and a spring constant of 26 N/m (OTESPA-R4; Bruker). Data were acquired in height mode at a scan rate of 0.5–1.0 Hz. Images were captured in 1024 × 1024-pixel format and were flattened and plane-fitted before analysis [61].

To objectively assess the distribution of sheet-like structure widths, “grid-analysis” was employed. This involved measuring fibrous structures along 10 × 10 grid lines spaced at 200 nm regular intervals on a 2 μm × 2 μm scale image. Due to the edge curvature and point angle of the cantilever, the horizontal dimensions observed in AFM images typically overestimate actual sample size. To estimate the true dimensions of fibres, the “circular cone model”, proposed by [62] was used. Briefly, for samples >35 nm imaged with an OTESPA-R4 cantilever, the real dimension (S) can be calculated using the equation: S (nm) = 0.8588 × W (nm) – 23.435, where W is the apparent sample width, and the parameter values (0.8588 and 23.435) derived from calibrations using 5, 10, and 80 nm gold particles. For samples smaller than 35 nm, the real size was estimated with the equation: S (nm) = 0.0051 × W² (nm), where W is the apparent sample width, and the parameter value (0.0051) based on calibrations using dsDNA with a known radius of 1 nm. Fibre width distributions were evaluated by generating histograms followed by multiple Gaussian fitting using Origin 2025 software (Light Stone, Japan).

In silico analyses

Information regarding the gene composition and organization of the ebfC operon across diverse bacterial species was obtained from annotated reference genomes available in the Genome and Gene databases of the National Center for Biotechnology Information (NCBI, NIH, USA) and the Kyoto Encyclopedia of Genes and Genomes (KEGG, Kanehisa Laboratories, Tokyo, Japan). Reference protein sequences for EbfC, RecR, and Hbsu homologues were retrieved from the NCBI Protein database (NIH, USA). Sequence similarity searches were performed using the blastp algorithm (NCBI, NIH, USA), employing full-length amino acid sequences as queries.

The predicted structure of the EbfC dimer structure was generated using AlphaFold 3 [63]. Experimentally determined structures of EbfC homologues were retrieved from the RCSB Protein Data Bank [64]. Structural alignments between EbfC and its homologues were conducted using TM-align to evaluate structural conservation and similarity [65]. Molecular visualization and structural analyses were carried out using UCSF Chimera (Resource for Biocomputing, Visualization, and Informatics, University of California, USA) [66].

Results and discussion

Ubiquitous EbfC evolves more rapidly than other proteins involved in DNA transactions

To understand the role of EbfC in bacteria, we first examined its genomic synteny and compared it with orthologues in other bacterial species. For simplicity, the gene and protein are referred to as ebfC and EbfC, respectively, for each species in which no prior designation has been reported. In B. subtilis, the ebfC gene is located near the oriC region, within an operon that also includes the essential dnaX gene (encoding for the DnaX subunit of the clamp loader complex) upstream and the non-essential recR gene (encoding for the RecR mediator of RecA) downstream [67]. This genomic arrangement may facilitate coordination between DNA replication and HR. However, a pairwise comparative analysis of the ebfC locus revealed that its genomic organization and chromosomal location (within or distal to oriC) vary more than would be expected given the genetic distance separating the bacterial pairs under comparison (Supplementary Annex S2 and Supplementary Fig. S1A). These observations suggest that a ubiquitous ancestral ebfC gene was lost and later reacquired via horizontal gene transfer from a distantly related bacterium, as inferred from the gene rearrangements observed.

To investigate this further, we analyzed sequence identities among EbfC-like proteins from diverse bacterial species (Supplementary Fig. S1A). Within the same bacterial Class, EbfC sequence identity varied and was significantly lower than would be expected by neutral mutational models. Within the Bacillales Order, B. subtilis EbfC [107 amino acids (aa) long] shares >65% sequence identity with its counterparts from Listeria monocytogenes and Staphylococcus aureus. Within the Lactobacillales Order, it shares ∼57% and ∼40% sequence identity with EbfC from Enterococcus faecalis and Lactococcus lactis, respectively, but <32% with homologues from Lactobacillus helveticus, Streptococcus pyogenes, Streptococcus pneumoniae, or Streptococcus suis. Conversely, B. subtilis RecR shares >75% sequence identity with RecR proteins from L. monocytogenes, S. aureus, and S. pyogenes, 65%–62% with those of E. faecalis, S. suis, and S. pneumoniae, and sequence identity is only reduced to 60%–56% with its L. helveticus and L. lactis counterparts (Supplementary Fig. S1A).

A comparable pattern was observed when comparing EbfC with its homologues from other evolutionarily distal phyla (e.g. Pseudomonadota, Spirochaetota, etc.). It shares only ∼32% and ∼27% sequence identity with YbaB_Hin and EbfC_Bbu, respectively, across a 92-amino acid interval, with the common sequence identity among all three proteins reduced to ∼17% (Supplementary Fig. S1A and B). In contrast, RecR shows 39%–44% sequence identity with RecR_Hin and RecR from Borrelia turcica (RecR_Btu), with the common sequence identity among all three proteins of ∼30% (Supplementary Fig. S1A and B). These data suggest that EbfC is subject to different selective pressures than typical housekeeping proteins.

To further extend this hypothesis, we also made a pairwise comparison of the other ubiquitous NAP, Hbsu. Except for L. lactis (∼58% sequence identity), Hbsu shares 64%–73% sequence identity with homologues from other members of the Phylum (Supplementary Fig. S1A). Among more distantly related bacteria, it still retains 61% and 34% sequence identity with homologues from H. influenzae and B. burgdorferi, respectively, and the shared sequence identity among all three is reduced to 32%.

Altogether, these results suggest that (i) EbfC-like proteins may be adapted to specific environmental conditions and ecological niches; (ii) EbfC-like proteins have evolved to a faster rate than RecR or Hbsu; and (iii) EbfC may not require interaction with other proteins to perform its function, unlike RecR. However, whether these EbfC-like proteins perform analogous functions across those different bacterial species remains unclear.

EbfC is a manganese-dependent DNA-binding protein

To determine whether B. subtilis EbfC exhibits activities characteristic of NAPs, we developed a protocol to purify native, soluble EbfC. The purified 11.7 kDa protein formed dimers in solution (Supplementary Fig. S3A), consistent with behaviour observed for YbaB_Hin or YbaB_Eco [42].

We first examined the DNA-binding activity. EbfC did not apparently bind either circular 3199-nt ssDNA (cssDNA) or linear 3199-bp dsDNA (ldsDNA) in buffer C containing 10 mM MgCl₂, ZnCl₂, CaCl₂, or a chelating agent (EDTA) at 37 °C (Fig. 1A and B, lanes 2–13). However, in the presence of Mn²⁺, EbfC formed high molecular weight complexes with both cssDNA or ldsDNA, which failed to migrate into the gel (Fig. 1A and B, lanes 14–16).

Figure 1.

EbfC binds preferentially to ssDNA in a Mn²⁺-dependent manner. (Aand B) EMSAs showing EbfC (2.5–10 μM) binding to 3.1 nM (in molecules) of pGEM-3Zf(+) cssDNA (A) or EcoRI-linearized dsDNA (B). Reactions were incubated in buffer C containing 10 mM MgCl₂, ZnCl₂, CaCl₂, MnCl₂, or EDTA for 15 min at 37°C. Complex formation was assessed by agarose gel electrophoresis followed by ethidium bromide staining. Quantification of EbfC–DNA complex formation (E) from three independent experiments is shown below each gel as mean and SD. Representative gels are shown. (C) Quantification of EbfC–DNA complex formation with pGEM-3Zf(+) cssDNA, scDNA, and ldsDNA in buffer C containing 10 mM MnCl₂. Reactions were performed as in panels (A and B). Three independent experiments were performed, and the mean ± SD plotted here. (D) Binding of EbfC to ldsDNA fragments either containing or lacking the EbfC_Bbu consensus target sequence (5′-^A/_T^A/_TCTA-3′). pHP13 DNA was digested with BsmI, MslI, StuI, and BbsI, generating five discrete fragments: F-1, F-2, and F-4 (black) contain the target motif, whereas F3 and F5 (grey) do not. EbfC (0.15–10 μM) was incubated with the fragments as in panel (C). Quantification of EbfC-DNA complex formation (labelled with an E) from three independent experiments is shown below the gel as mean and SD. A representative gel is shown.

The formation of EbfC–cssDNA complexes increased exponentially before reaching a plateau, suggesting a cooperative binding (Fig. 1A, lanes 14–16, and 1C). Reducing MnCl₂ concentration from 10 to 1 mM neither affected the affinity nor the type of complex formed (Supplementary Fig. S3B). The apparent equilibrium constant (K_app) for the EbfC–cssDNA interaction was calculated as 0.75 μM—a protein concentration compatible with estimated intracellular EbfC levels in wt and ebfC⁺⁺ strains (ranging from 0.5 to 1.6 μM; see below, Supplementary Fig. S4B). Under this K_app, the stoichiometry corresponds to one EbfC dimer/13-nt, or ∼242 EbfC dimers/cssDNA molecule (Fig. 1C). EbfC exhibited ∼3- and ∼5-fold lower affinity for ldsDNA and scDNA, respectively, when compared to cssDNA (Fig. 1B, lanes 14–16, and 1C). Notably, no canonical DNA-binding domain is predicted within the EbfC structure (Supplementary Annex S3 and Supplementary Fig. S2).

Previous studies have reported varying DNA-binding characteristics among EbfC homologues. For instance, EbfC_Bbu was shown to bind dsDNA with certain sequence preference (5′-GTnAC-3′), whereas YbaB_Hin and YbaB_Eco displayed undefined specificity for a sequence, AT-content, or DNA topology. In contrast, EbfC_Dra and EbfC_Ccr bind dsDNA but show no preference for the palindromic 5′-GTnAC-3′ motif [42–48]. EbfC_Mtu also binds dsDNA, although its binding specificity has not been characterized [49]. Variability in DNA-binding profiles across EbfC orthologues has been attributed to amino acid sequence differences in their putative DNA-binding regions [46, 47]. Additionally, experimental differences may also contribute to these discrepancies; however, binding conditions of these assays are poorly defined since they were mainly performed using EMSA kits.

To test sequence specificity-particularly for the palindromic 5′-GTnAC-3′ motif recognized by EbfC_Bbu,we digested the 4748-bp pHP13 plasmid using BsmI, MslI, StuI, and BbsI, generating five discrete fragments. Three of these (F-1, F-2, and F-4) contained the motif, while the other two (F-3 and F-5) did not. All fragments were bound by increasing EbfC concentrations with comparable efficiency (Fig. 1D), indicating no significant preference for the 5′-GTnAC-3′ sequence.

EbfC bridges DNA

To explain the formation of high order EbfC–DNA complexes, we hypothesized that EbfC either binds and induces intramolecular looping of DNA or binds and bridges two or more cssDNA or ldsDNA molecules, forming sandwich-like or intermolecular complexes. Such DNA bridging activity would generate extensive protein–DNA networks too large to enter the gel matrix.

To test whether EbfC promotes DNA bridging, either the 3199-nt cssDNA or 3199-bp ldsDNA was incubated with increasing concentrations of EbfC for 15 min, and the reaction mixture centrifuged. In the absence of the transition metal Mn²⁺, both cssDNA or ldsDNA remained in the supernatant, even at EbfC concentrations up to 5-fold higher than the K_app (Fig. 2A–D, lanes 1–7). Only a small fraction, 6% to 10% of DNA, was recovered in the pellet when EbfC was present in large excess (>1500 EbfC/DNA molecule) (Fig. 2A–D, lane 8).

Figure 2.

EbfC promotes DNA bridging and protects DNA. (A–D) pGEM-3Zf(+) cssDNA (A and B) or scDNA (C and D) (3.1 nM in molecules) was incubated with increasing concentrations of EbfC (0.078–10 μM) in buffer C, either in the absence of metal or containing 10 mM MnCl₂ for 15 min at 37°C. Samples were then centrifuged at 21 300 × g for 10 min. DNA present in the supernatant (SN [A and C]) and in the pellet (B and D) was analyzed by agarose gel electrophoresis followed by ethidium bromide staining. The fraction of DNA recovered in the SN and pellet was quantified from three independent experiments, and the mean and SD are shown below each gel. Representative gels are presented. (E) pGEM-3Zf(+) ldsDNA [1.8 nM (lanes 2–9) or 0.3 nM (lanes 10–17) in molecules] was incubated with increasing concentrations of EbfC (2.5–10 μM) in buffer C containing 10 mM MnCl₂ for 15 min at 37°C. Subsequently, 3 U T4 DNA ligase were added, and the reaction continued for 3 h at 22°C. After deproteinization, samples were analyzed by agarose gel electrophoresis followed by ethidium bromide staining. (F) EbfC (0.15–10 μM) was incubated with ldsDNA or scDNA (3.1 nM in molecules) in buffer C containing 10 mM MnCl₂ for 15 min at 37°C. Then, 0.05 U DNase I was added, and the reaction continued for 5 min at 22°C. Following deproteinization, samples were analyzed by agarose gel electrophoresis followed by ethidium bromide staining. The extent of DNA digestion was quantified from three independent experiments, and the mean and SD is shown below the representative gel image.

By contrast, in the presence of the Mn²⁺ ion—which is required for EbfC–DNA binding (Fig. 1)—considerably lower concentrations of EbfC were sufficient to obtain similar results (Fig. 2A–D, lanes 10–18). Approximately 60% of the DNA was recovered in the pellet with ∼100 EbfC dimers/cssDNA or ldsDNA molecule (or 1 EbfC dimer/ 32-nt or 32-bp) (Fig. 2B and D, lane 13). At a ratio of 400 EbfC dimers per DNA molecule, >95% of the DNA was pelleted, with a corresponding near-complete depletion from the supernatant (Fig. 2A–D, lanes 15–18).

These results suggest that EbfC promotes the compaction or bridging of both cssDNA and ldsDNA. However, this assay does not allow us to definitely determine whether EbfC promotes DNA looping. DNA looping by architectural proteins can facilitate DNA ligase-mediated ring closure of restriction fragments (intramolecular ligation) or promote ligation between distinct DNA molecules (intermolecular ligation), as previously observed for LrpC [68]. To investigate whether EbfC promotes these activities, a DNA ligation assay was conducted.

Two concentrations of EcoRI-linearized dsDNA were tested: a low concentration to favour in cis reactions (intramolecular looping and self-ligation), and a high concentration to favour in trans reactions (intermolecular ligation between separate DNA fragments and bridging). In the presence of EbfC but absence of T4 DNA ligase, the migration pattern of deproteinized linear dsDNA remained unchanged (Fig. 2E, lanes 3–5 and 11–13), indicating that EbfC alone does not covalently alter the DNA. At the higher DNA concentration (1.8 nM of DNA molecules, in trans ligation), T4 DNA ligase converted a portion of the ldsDNA into higher-order circular oligomeric forms that migrate more slowly than nicked circular DNA (Form II) at high EbfC concentrations (2.5–5 μM) (Fig. 2E, lanes 8–9). At the lower DNA concentration (0.3 nM of DNA molecules, in cis ligation), T4 DNA ligase converted a portion of the ldsDNA into an oligomeric circular form that migrates slower than Form II DNA, without a corresponding increase in supercoiled circular DNA (Form I DNA) (Fig. 2E, lane 17).

The observation of DNA dimers and multimers at both low and high concentrations of DNA, without an increase in Form I DNA, suggests that EbfC primarily favours the formation of intermolecular DNA complexes and bridging, rather than promoting intramolecular looping that would result in an increase of supercoiled circular DNA.

EbfC protects DNA

Subsequently, we investigated whether EbfC protects ldsDNA or scDNA substrates from DNase I degradation, as previous studies have shown that EbfC_Dra and YbaB_Ccr perform this function by polymerizing along DNA [46, 48].

In the absence of EbfC, DNase I fully degraded both ldsDNA and scDNA substrates (Fig. 2F, lanes 2 and 11). At a low EbfC concentration (0.15 μM, 1 EbfC dimer/ 66-bp), ∼80% of the ldsDNA was degraded by DNase I. One EbfC dimer/8-bp (1.25 μM EbfC) was required to protect >95% of the ldsDNA from DNase I digestion (Fig. 2F, lane 3 versus lane 6), suggesting that EbfC acts as a physical barrier to enzymatic degradation. However, given typical in vivo cellular concentrations of EbfC, protection may be limited to small DNA regions. At present, we cannot exclude the possibility that ldsDNA becomes densely wrapped around a multimeric EbfC core under these conditions.

In contrast, scDNA was more susceptible to degradation by DNase I in the presence of low EbfC concentrations (0.15–1.25 μM) (Fig. 2F, lanes 12–15). At higher concentrations (2.5–10 μM), the formation of nicked circular and ldsDNA was observed, and a saturating concentration of EbfC was required to protect up to 30% of the scDNA substrate from degradation (Fig. 2F, lanes 16–18).

Nano-scale analysis of EbfC–DNA complexes

To gain detailed insight into the nature of the complexes formed by EbfC with DNA, we analyzed them at the nano-scale level using AFM. Briefly, increasing concentrations of EbfC (2.5–10 μM) were incubated with 3199-nt/bp cssDNA and scDNA, and >6 000 nucleoprotein complexes were imaged by AFM.

Theoretical estimates suggest diameters of 2.8 nm for an EbfC monomer and 5.6 nm for a dimer, in the globular form. During AFM sample preparation, EbfC appeared to multimerize in solution, forming particles with width ranging from 43.7 ± 2.3 to 63.4 ± 0.7 nm (Supplementary Fig. S3A), as observed with YbaB_Hin [42]. These likely correspond to aggregates containing 7–12 EbfC dimers, making it unfeasible to experimentally determine the volume of a single monomer. The mean width of the dsDNA substrate was 1.7 nm, both in the presence or absence of Mn²⁺ ions, which aligns well with the theoretical value of 2 nm (Fig. 3A and Supplementary Table S2). Upon EbfC binding, distinct higher-order structures emerged, likely resulting from multiple interaction modes between EbfC and DNA. These included coating, bridging, and bunching, although no stiffening of DNA was observed. To characterize these nucleoprotein complexes, fibre widths were measured and their distribution plotted for each protein concentration (Fig. 3A and Supplementary Table S2). At 2.5 μM EbfC, three representative classes of complexes were identified, with mean widths of 4.0, 15.2, and 20.2 nm. At 5 μM EbfC, the width of the largest complex increased to 39.1 nm, and at 10 μM, complexes as wide as 48.4 nm were observed (Fig. 3A and Supplementary Table S2). We estimate that the narrowest nucleoprotein complexes (4.1–4.6 nm) represent EbfC monomers bound to dsDNA, while the 10.9 nm complexes likely correspond to two DNA molecules bridged by two EbfC dimers. Wider complexes probably consist of large EbfC aggregates (>10 dimers) interacting with multiple dsDNA molecules.

Figure 3.

Analysis of EbfC-ssDNA/dsDNA complexes by AFM. pGEM3 Zf(+) dsDNA (A) or ssDNA (B) (0.7 nM in molecules) was incubated with increasing concentrations of EbfC (2.5–10 μM) in buffer C containing 10 mM MnCl₂ for 15 min at 37°C. Each sample was deposited onto freshly cleaved mica, processed, and visualized by AFM. Representative images are shown. Scale bars: 500 nm. The widths of nucleoprotein complexes were measured and displayed as histograms. The data were fitted to curves, with black arrowheads indicating the most frequent complexes (peaks). The measured width of the DNA-bound protein complexes displayed a linear relationship with the measured width of DNA. Peak widths are provided in Supplementary Table S2. Each experiment was carried out more than three times with similar results. Schematics illustrating the representative types of complexes observed are provided.

Similar complex formation was observed with cssDNA. The mean width of the ssDNA substrate was 1.2 nm, both in the presence or absence of Mn²⁺ ions (Fig. 3B and Supplementary Table S2). EbfC appeared to coat the ssDNA, with bush-like structures observed less frequently than with dsDNA. Fibre width analysis revealed broader complexes compared to those formed on dsDNA (Fig. 3A and B, and Supplementary Table S2). At 2.5 μM EbfC, two dominant classes of complexes were detected, with mean widths of 12.5, and 29.9 nm. At 5 μM EbfC, a third wider population appeared, with a mean width of 36.9 nm. At 10 μM EbfC, significantly larger complexes were observed, with mean widths >60 nm, indicative of extensive EbfC–ssDNA nucleoprotein complexes (Fig. 3B and Supplementary Table S2).

In conclusion, EbfC appears to sheath DNA and facilitates the formation of large, higher-order DNA complexes in a concentration-dependent manner, supporting its role in modulating nucleoid architecture. These findings reinforce the classification of EbfC as a NAP.

EbfC localizes to the nucleoid

To investigate the subcellular localization of EbfC in B. subtilis, a C-terminal msfGFP fusion of EbfC was expressed under the control of its native promoter. Nucleoids were simultaneously visualized using DAPI staining. As expected, wt cells lacking the fusion protein displayed no green fluorescence above background levels (Fig. 4). Expression of EbfC–msfGFP resulted in distinct clusters of faint green fluorescence. These clusters likely represent discrete EbfC complexes coating various regions throughout the nucleoid, as merged images revealed strong co-localization between EbfC–msfGFP and the nucleoid (>90%), with the fluorescence signal absent from DNA-free regions of the cell (Fig. 4). This observation further supports that EbfC behaves as a bona fide NAP. This is consistent with previous findings in other bacteria, where EbfC orthologues also co-localize with the nucleoid, including in B. burgdorferi, D. radiodurans, and C. crescentus [45, 46, 48].

Figure 4.

EbfC co-localizes with the nucleoid. Cells expressing an EbfC–msfGFP fusion protein were grown at 37°C in LB medium to exponential phase (OD₅₆₀ = 0.4), stained with DAPI, and visualized by fluorescence microscopy. The wt strain was included as a control. Co-localization was assessed in >500 cells, and representative images are shown. Scale bars: 2.5 μm.

ebfC inactivation reduces and ebfC overexpression increases natural chromosomal transformation

Several of the activities demonstrated here for EbfC, including sequence-independent binding to both ssDNA and dsDNA, enhancement of DNA ligation, DNA bridging, protection from DNase I degradation, and localization to the nucleoid, are hallmark features of bona fide NAPs that modulate chromosomal architecture and domain organization. Having characterized these molecular functions, we next examined the in vivo roles of EbfC. To investigate whether EbfC plays a role in genetic recombination, we analyzed homology-directed, RecA-dependent NCT. During the development of natural competence, cells enter a transient, persistent, non-growing, haploid state [69]. Thus, the contribution of replication or transcription to the formation of positive and negative supercoils—factors that influence chromosome dynamics and activity—can be neglected (see “Introduction” section).

To assess whether EbfC affects NCT frequencies, the wt strain and its isogenic ΔebfC derivative were induced to enter the competent state. First, we tested whether ebfC inactivation alters the rate of mutagenesis, which could indirectly confound our transformation frequency measurements. In the absence of externally added DNA, naturally competent cells were diluted and plated on LB agar with or without Rif (8 μg·ml⁻¹). The number of spontaneous Rif^R colonies was comparable in ΔebfC and wt cells (6–9 × 10⁻⁹), suggesting that EbfC does not influence spontaneous mutagenesis.

Second, competent cells were incubated with donor DNAs. The frequency of intra- and interspecies NCT in ΔebfC cells was reduced ∼11- and ∼8-fold, respectively, compared to the wt control (P < 0.01) (Fig. 5A and Supplementary Table S3). Since cells competent for NCT are also competent for NPT, we used NPT as a control to exclude the possibility that reduced transformation was due to defects in competence development and/or DNA uptake [21]. NPT frequencies were only marginally reduced 4- to 5-fold in ΔebfC cells compared to the wt strain (P > 0.1) (Supplementary Table S3). It is worth noting that only a subpopulation of cells becomes competent, and slight variations relative to the wt control are not considered significant [19].

Figure 5.

Inactivation of EbfC reduces, and ebfC overexpression increases, NCT. (A) The indicated strains were rendered naturally competent. Competent cells were transformed with 0.1 μg·ml⁻¹ of Bsu 168 rpoB482 (intraspecies NCT, filled bars) or Bat 1942 rpoB482 DNA (interspecies NCT, empty bars) with selection for Rif^R (8 μg·ml⁻¹). The yield of chromosomal transformants was normalized relative to that of the wt strain, set as 1. (B) As in panel (A), but data were normalized relative to the corresponding NAP single mutant strain background. Bars represent the mean ± SD of at least four independent experiments. Numerical data are provided in Supplementary Table S3.

To confirm that the NCT defect was due specifically to ebfC inactivation, and not to polar effects on the downstream recR gene (Supplementary Fig. S1A), we assessed NCT and NPT in ΔrecR cells. Inactivation of recR did not significantly affect intra- [55, 56] or interspecies NCT, nor NPT, compared to wt (P > 0.1) (Fig. 5A and Supplementary Table S3). To further support that the transformation defect is attributable solely to ebfC inactivation, a P-ebfC cassette was ectopically integrated at the amyE locus of the ΔebfC strain (BG1951) (Supplementary Table S1), and competence development and transformation assays were performed in the presence of IPTG (500 μM). Intra- and interspecies NCT, as well as NPT, were not significantly different from the wt control (P > 0.1) (Fig. 5A and Supplementary Table S3), demonstrating that the ectopically expressed P-ebfC cassette was both necessary and sufficient to restore the NCT defect in ΔebfC cells.

Third, to further characterize the role of EbfC on NCT, we increased ebfC gene dosage. The P-ebfC cassette was introduced into the wt background to generate the ebfC⁺P-ebfC strain (BG683), termed ebfC⁺⁺ for simplicity (Supplementary Table S1), which carries the native copy of ebfC and an ectopic second one at the amyE locus. The native ebfC gene is highly transcribed from its native locus and promoter, generating a long-lived (>30 min) mRNA [70, 71], while the ectopic copy is transcribed from a strong hybrid IPTG-inducible promoter induced with 500 μM IPTG. Unexpectedly, the ebfC⁺⁺ strain showed a ∼11- and ∼7-fold increase in intra- and interspecies NCT frequencies, respectively, compared to wt (P < 0.01) (Fig. 5A and Supplementary Table S3). This enhancement cannot be explained by an increased proportion of competent cells, as NPT frequency was only marginally (∼2-fold) increased relative to wt and ΔebfC P-ebfC strains (P > 0.1) (Supplementary Table S3).

To understand these effects, we quantified EbfC protein levels in each strain (Supplementary Fig. S4A and B). Given that EbfC is translated from a long-lived mRNA, protein levels could not be directly proportional to transcript levels, and thus RNA-seq or qPCR techniques are inappropriate for measuring expression [70, 71]. Instead, EbfC levels were determined via immunoblotting. As expected, EbfC was undetectable in ΔebfC cells. wt and IPTG-treated ΔebfC P-ebfC strains expressed similar levels of EbfC (717 ± 45 versus 643 ± 199 EbfC/monomers cell, P > 0.1). In contrast, IPTG-induced ebfC⁺⁺ cells exhibited a significant 3- to 4-fold increase in EbfC levels (2450 ± 190 EbfC/monomers cell, P < 0.01) compared to the wt strain. A low induction of EbfC expression was also detected in uninduced ΔebfC P-ebfC and ebfC⁺⁺ strains, likely due to a certain degree of promoter leakiness. Notably, neither deletion nor overexpression of ebfC affected growth in LB rich (doubling time of 35 ± 3 min) or GM defined (doubling time of 54 ± 8 min) medium or plating efficiency when compared to wt (Supplementary Fig. S4C–E).

Altogether, these results suggest that (i) the absence of EbfC impairs NCT without altering the fraction of competent cells in the population, or DNA uptake; (ii) the NCT defect in both intra- and interspecies is specifically attributable to the ebfC deletion and not a downstream polar effect on recR; (iii) NPT, which bypasses nucleoid-related topological constrains, remains largely unaffected; and (iv) ectopic expression of ebfC enhances intra- and interspecies NCT frequencies in wt cells and complements the ΔebfC NCT defect. We propose that EbfC, acting as a NAP, modulates the intrinsic inflexibility of naked DNA and/or bridges DNA molecules, and facilitates the formation of a nucleoprotein complex suitable for RecA to catalyze integration of the incoming linear ssDNA. In fact, the NAPs Rok, LrpC, and Hbsu, also influence intra- and interspecies NCT by modulating DNA architecture, although the DNA architecture they promote appear to hinder the process [10]. Alternatively, to explain these phenotypes, a possible role for NAPs in gene silencing of incoming DNA was considered. However, this is unlikely, as (i) mutants in rok, hbs, and lrpC exhibit a phenotype opposite to that of ΔebfC in NCT; (ii) members of the EbfC/YbaB, Hbsu/HU, and LrpC families have not been implicated in xenogeneic silencing; and (iii) in B. subtilis, Rok is a silencer of foreign DNA, and it has been proposed that typically only one NAP family acts as a xenogeneic silencer in a given bacterial species [72].

Inactivation of ebfC antagonizes and ebfC overexpression synergistically enhances NCT in hbs55, Δrok, or ΔlrpC cells

How NAPs link nucleoid architecture and the efficiency of genetic recombination in naturally competent cells remains poorly understood. In our recent studies, we showed that (i) the hbs55, ΔlrpC, and Δrok mutation significantly increased intra- (∼15–20-fold) and interspecies NCT (∼2–8-fold) (Supplementary Table S3), without altering the size of the subpopulation that becomes competent, DNA uptake, or the frequency of untargeted Rif^R mutagenesis [10]; and (ii) combinations of these mutations act synergistically to further enhance both intra- (∼105–140-fold) and interspecies NCT (∼20–30-fold) (Supplementary Table S3) [10]. Given the findings described in the previous section, we hypothesized that increasing ebfC gene dosage might further elevate NCT frequencies in hbs55, ΔlrpC, or Δrok cells. Additionally, we posited that changes in nucleoid structure caused by the inactivation of Hbsu, Rok, or LrpC might facilitate RecA-mediated DSE and thus indirectly compensate the NCT defect observed in ΔebfC cells. To test these hypotheses, the ΔebfC mutation or the P-ebfC cassette was introduced into the hbs55, Δrok, and ΔlrpC backgrounds (Supplementary Table S1).

As predicted, the hbs55, Δrok, and ΔlrpC mutations significantly suppressed the ΔebfC-associated defect in both intra- and interspecies NCT (P < 0.01). Moreover, when these mutations were combined with ebfC overexpression, a synergistic increase in both intra- and interspecies NCT was observed (P < 0.01) (Fig. 5A and Supplementary Table S3). These results support the following conclusions: (i) physiological levels of EbfC antagonize the barrier introduced by Hbsu, Rok, or LrpC on intra- and interspecies NCT; (ii) EbfC overexpression further enhances NCT in the hbs55, Δrok, or ΔlrpC contexts; and (iii) nucleoid organization should be considered an integral factor in horizontal gene transfer in naturally competent cells, potentially acting as a dynamic scaffold for homology-directed, RecA-mediated NCT. The precise role of NAPs in NCT remains to be elucidated. Whether they promote the rejection of dsDNA, facilitate strand separation and assimilation, or stabilize D-loop formation during recombination is not yet clear.

It is worth noting that RecA cannot tolerate ≥2 mismatches within each of two or more consecutive 8-bp micro-homologous steps during interspecies NCT [19, 25]. Interestingly, although NAPs influence the acquisition of donor DNA by modifying nucleoid architecture, they do not significantly alter the length of the divergent donor DNA fragments integrated into the host chromosome. Thus, NAPs do not impact the fidelity of NCT at the local level (see Supplementary Annex S4, Supplementary Fig. S5, and Supplementary Table S4).

Inactivation and overexpression of ebfC differentially affects NCT in ΔrecD2 cells

In our recent study, we demonstrated that the defect in NCT caused by the accumulation of saturated RecA·ATP filaments upon inactivation of the negative RecA modulator RecX is exacerbated in hbs55, ΔlrpC, or Δrok backgrounds [10]. To determine whether this effect extends to EbfC, and whether the aggravated defect is specific to the inactivation of RecX or instead reflects a more general impairment in RecA-mediated DSE caused by the inactivation of negative modulators, we investigated another RecA negative modulator, RecD2. Similar to RecX, RecD2 interacts with and promotes RecA filament disassembly, albeit through distinct molecular mechanisms [11, 13, 60]. Accordingly, the ΔrecD2 mutation—which also leads to saturated RecA·ATP filaments—was introduced into the ΔebfC, ebfC⁺⁺, hbs55, ΔlrpC, and Δrok backgrounds (Supplementary Table S1).

Firstly, the single and double mutant strains were rendered competent, and the frequency of spontaneous Rif^R mutants was assessed in the absence of exogenous DNA. As expected, the frequency of untargeted Rif^R mutants in the parental ΔrecD2 and corresponding double mutants increased ∼3-fold (1.3–2.6 × 10⁻⁸) compared to wt, ΔebfC, ebfC⁺⁺, hbs55, ΔlrpC, and Δrok single mutants. This confirms that spontaneous mutagenesis is moderately increased in competent ΔrecD2 cells [19, 22], and that this phenotype is unaffected by concurrent NAP inactivation.

Secondly, the same competent cells were incubated with donor DNA, and NCT and NPT frequencies were measured. The intra- and interspecies NCT frequencies in the ΔrecD2 ΔebfC mutant were similar to those of the ΔebfC strain (Fig. 5B and Supplementary Table S3).

Finally, increased ebfC gene dosage in the ΔrecD2 background significantly reduced both intra- and interspecies NCT compared to ebfC⁺⁺ cells (P < 0.01), while NPT remained unaffected (Fig. 5B and Supplementary Table S3). Similarly, the intra- and interspecies NCT frequencies were significantly decreased in hbs55 ΔrecD2, ΔlrpC ΔrecD2, and Δrok ΔrecD2 competent cells relative to their respective parental single mutants (P < 0.01), again without affecting NPT (Fig. 5B and Supplementary Table S3).

Together, these findings suggest that (i) the ΔrecD2 mutation does not modify the NCT defect seen in ΔebfC cells, but leads to a profound NCT defect upon EbfC overexpression or inactivation of hbs, lrpC, or rok; and (ii) this impairment in NCT is attributable to the lack of negative regulation of RecA nucleoprotein filament dynamics in the ΔrecX [10] or ΔrecD2 backgrounds, and indirectly a consequence of RecA·ATP filament saturation rather than a unique feature of RecX or RecD2 themselves.

Inactivation of ebfC antagonizes and ebfC overexpression additively increases nucleoid expansion in hbs55, ΔlrpC, and Δrok cells

Based on data presented in previous sections, EbfC may act in opposition to Rok, LrpC, and Hbsu with regard to chromosome structure. In B. subtilis, the chromosome is compacted into micro- and macrodomain-organized structures occupying a significant portion of the cellular volume (∼40%) [73–76]. In DAPI-stained, exponentially growing fixed cells, the nucleoid typically appears as one or two compact structures, spatially separated from the membrane by the cytosol and polyribosomes. Its compaction level is influenced by DNA-supercoiling, transcriptional activity, and the presence of NAPs [73, 75]. However, the molecular determinants governing nucleoid volume are still not well defined. For instance, Hbsu homologues in Firmicutes have been shown to enhance negative DNA supercoiling and promote compaction of previously relaxed DNA [77, 78]. Furthermore, Hbsu and LrpC are broadly distributed across the nucleoid, while Rok preferentially forms discrete nucleoid-condensing domains [36, 41, 74, 76, 78, 79]. To explore whether EbfC contributes to nucleoid structure regulation, cells were grown to OD₅₆₀ = 0.4 in LB at 37°C, their nucleoids stained with DAPI, and imaged via fluorescence microscopy to quantify nucleoid area relative to total cell area.

Firstly, we assessed whether NAP inactivation or ebfC overexpression altered cell morphology or nucleoid segregation. No significant differences in cell morphology, filamentation, or nucleoid abnormalities—such as anucleate cells or aberrant nucleoids—were observed across all strains examined (Fig. 6B–D). In wt cells, the nucleoid occupied ∼40% of the total cell area (n = 4121), closely aligning with previously reported scaling of nucleoid-to-cell size in B. subtilis cells grown to exponential phase, with a nucleoid/cytosolic ratio of ∼0.4 [75]. The area was not significantly altered in ΔebfC P-ebfC cells (∼43%, n = 1233, P > 0.1) (Fig. 6A and B). However, in ΔebfC cells, nucleoid occupancy dropped significantly to ∼31% (n = 962, P < 0.01), whereas nucleoid area in ΔrecR cells remained similar to that of the wt control (Fig. 6A and C), confirming nucleoid condensation stems from ebfC inactivation rather than a downstream polar effect on recR. In contrast, IPTG-induced ebfC overexpression (ebfC⁺⁺ cells) led to a substantial nucleoid expansion (∼58%, n = 1174, P < 0.01) (Fig. 6A and D), suggesting that elevated EbfC levels promote chromosomal relaxation. This is reminiscent of EbfC_Dra, which constrains DNA supercoils, without compacting the nucleoid [48].

Figure 6.

Inactivation of ebfC leads to nucleoid compaction, while its duplication results in nucleoid relaxation. (A) The indicated strains were grown to exponential phase, fixed, and stained with DAPI to visualize nucleoids by fluorescence microscopy. Nucleoid areas were quantified relative to total cell area. Data represent the mean ± SD from at least four independent experiments. (B–  D) Representative fluorescence microscopy images of the wt, ΔebfC, and ebfC⁺⁺ strains are shown. Nucleoids are stained with DAPI (blue), and cell membranes with FM4-64 (red). Scale bars: 5 μm.

We next examined whether inactivation of other NAPs also influenced nucleoid compaction. In hbs55, ΔlrpC, and Δrok cells, the nucleoid occupied a significantly greater fraction of the cell area (60%–63%, n > 1350, P < 0.01) than in wt cells (Supplementary Fig. S6). To rule out polar effects on the topB gene, which encodes Topo IIIβ (topB) and maps downstream of lrpC [80], we analyzed a ΔtopB mutant. This mutant showed only a modest increase (∼1.3-fold) in nucleoid area (51%, n = 999) (Supplementary Fig. S6), suggesting that the more pronounced relaxation observed in ΔlrpC cells is specific to the loss of lrpC.

To further explore the interplay between EbfC and other NAPs, double mutants were analyzed. In each case, ΔebfC reduced nucleoid area (44%–48%, n > 950), while IPTG-induced ebfC overexpression further expanded the nucleoid (69%–72%, n > 850, P < 0.01), in hbs55, ΔlrpC, or Δrok backgrounds (Fig. 6), indicating a dynamic antagonism between EbfC and Hbsu, Rok, or LrpC in maintaining nucleoid architecture. To confirm this hypothesis, we examined other double NAP mutants: Δrok hbs55, Δrok ΔlrpC, and ΔlrpC hbs55. These exhibited significantly greater nucleoid expansion than their respective single parental mutants (P < 0.01), suggesting additive effects on chromosomal decompaction (Supplementary Fig. S6).

In summary, these results support that (i) physiological levels of EbfC counteract the nucleoid architecture formed by Hbsu, Rok, and LrpC; and (ii) nucleoid relaxation correlates with increased NCT frequencies, whereas greater nucleoid condensation is associated with reduced NCT frequencies.

Inactivation of ebfC impairs DNA repair in response to 4NQO-, MMC-, or H₂O₂-induced DNA damage

How NAPs contribute to genome stability in exponentially growing B. subtilis cells remains poorly understood. Previous studies have shown that (i) the expression of ebfC, hbs, rok, or lrpC is not regulated by LexA, the central regulator of the SOS response [81]; (ii) the hbs55 mutation sensitizes cells to acute exposure to alkylating (10 mM MMS) or bulky (100 μM 4NQO) DNA lesions [50, 68]; (iii) deletion of ebfC results in nucleoid compaction, while its overexpression causes nucleoid expansion—similar to inactivation of rok, lprC, or hbs55 (Fig. 6); (iv) EbfC_Dra and YbaB_Eco contribute to maintain DNA integrity and cell survival following H₂O₂-, UV-, or γ-ray-induced DNA damage [48, 82]; (v) EbfC protects DNA from degradation (Fig. 2E), potentially contributing indirectly to genetic stability; and (vi) a DNA protecting role has been described for EbfC_Eco or Paenibacillus riograndensis EbfC [82, 83]. These observations support the hypothesis that NAPs contribute to genome organization and stability by modulating DNA compaction within the nucleoid and/or dynamically modulating domain boundaries across the chromosome. To test this, NAP-deficient strains were grown exponentially in LB medium at 37°C to OD₅₆₀ = 0.4 (with 500 μM IPTG for strains carrying the P-ebfC cassette), and then chronically exposed to increasing but sublethal concentrations of diverse DNA damaging agents (see Supplementary Annex S5). These included alkylating (MMS), distorting (4NQO), oxidizing and ssDNA nicking (H₂O₂), and cross-linking (MMC) agents. Survival was quantified with respect to their absence and compared to wt control.

To exclude polar effects of ebfC deletion on the downstream gene recR, we analyzed a ΔrecR mutant. This mutant served also to validate the selected doses as sufficient to reduce the survival rate of a genuine rec-deficient strain, as ΔrecR cells were previously shown to be hypersensitive to DNA damage [56, 67]. As expected, ΔrecR cells were highly sensitive to all four DNA damaging agents, with a significantly lower lethal dose to reduce survival by 90% (LD₉₀) than any of the NAP mutants (P < 0.01) (Fig. 7A, D, G, and J; Supplementary Table S5).

Figure 7.

Survival of ΔebfC and ebfC⁺⁺ cells, alone or in combination with hbs55, Δrok, or ΔlrpC mutations, following chronic exposure to DNA-damage. The indicated strains were exposed to MMS (A–  C), 4NQO (D–F), MMC (G–I), or H₂O₂ (J–L). The ΔrecR, hbs55, Δrok, ΔlrpC, and wt strain were used as controls. Cells were grown in LB medium to OD₅₆₀ = 0.4, serially diluted, and plated onto LB plates containing increasing concentrations of each DNA-damaging agent. Survival was calculated relative to untreated conditions, set to 1. Data represent mean ± SD from at least four independent experiments.

Upon chronic MMS exposure, the LD₉₀ for ΔebfC and ebfC⁺⁺ cells was >2 mM MMS, comparable to wt and ΔebfC P-ebfC complementing control strains (P > 0.1) (Fig. 7A and B, and Supplementary Table S5). Similarly, ΔlrpC and Δrok strains were MMS-resistant (P > 0.1), whereas the hbs55 mutant was sensitive (LD₉₀ ∼1.6 mM MMS, P < 0.01) (Supplementary Fig. S7A and Supplementary Table S5), consistent with previous reports [50, 68]. Double mutants ΔebfC Δrok, ΔebfC ΔlrpC, ebfC⁺⁺ Δrok, ebfC⁺⁺ ΔlrpC, and Δrok ΔlrpC were also as MMS-resistant as wt (P > 0.1). In contrast, combinations with hbs55 -ΔebfC hbs55, ebfC⁺⁺hbs55, Δrok hbs55 and ΔlrpC hbs55- retained MMS sensitivity similar to the hbs55 single mutant (P > 0.1) (Fig. 7A–C, Supplementary Fig. S7A, and Supplementary Table S5). These data suggest that Hbsu contributes to the repair of alkylating lesions, while EbfC, Rok, or LrpC do not. This is consistent with the previous observation that Hbsu, which enriches at abasic apurimidinic/apyrimidinic (AP) lesions [84], contributes to the repair of MMS-induced damage through a more specific mechanism: Hbsu stimulates the endonuclease activity of the Nth enzyme at AP sites [85].

Chronic exposure to 4NQO revealed that ΔebfC and hbs55 strains were both sensitive (LD₉₀ ∼0.17 and ∼0.2 μM 4NQO, respectively, P < 0.01), while Δrok and ΔlrpC strains remained proficient in DNA repair as the wt and ΔebfC P-ebfC controls (P > 0.1) (Fig. 7D and E, and Supplementary Table S5). Interestingly, the ebfC⁺⁺ strain exhibited slightly increased resistance (LD₉₀ ∼0.35 μM 4NQO, p > 0.1), with ebfC⁺⁺ Δrok and ebfC⁺⁺ ΔlrpC strains being even more proficient (P > 0.1) than wt (Fig. 7D–F and Supplementary Table S5), suggesting that elevated EbfC levels may facilitate the removal of bulky adducts or activate alternative repair pathways. The sensitivity of ΔebfC Δrok and ΔebfC ΔlrpC double mutants mirrored that of the ΔebfC parent strain (P > 0.1), and the ΔebfC hbs55 strain that of the hbs55 parent strain (P > 0.1), indicating no additive effect (Fig. 7D and E, and Supplementary Table S5). Similarly, the Δrok hbs55 or hbs55 ΔlrpC cells were as sensitive as the hbs55 parent strain (P > 0.1), while Δrok ΔlrpC cells showed slightly increased resistance compared to wt (P > 0.1) (Supplementary Fig. S7B and Supplementary Table S5).

Given that the ΔebfC mutation sensitizes cells to chronic exposure to 4NQO but not to MMS (Fig. 7A and D), we hypothesized that EbfC may contribute to overcoming replication stress, but the low dose of MMS and chronic exposure are insufficient to detect the phenotype. To evaluate this hypothesis, we acutely exposed exponentially growing ΔebfC and ebfC⁺⁺ cells to MMS (20–60 mM) or 4NQO (100–300 μM) for 15 min and plated them in their absence. The ΔebfC and ebfC⁺⁺ strains displayed a resistance to MMS comparable to wt and ΔebfC P-ebfC controls (P > 0.1), while the hbs55 positive control was significantly more sensitive (P < 0.01) (Supplementary Fig. S8A). By contrast, both ΔebfC and hbs55 mutations sensitized cells to acute exposure to 4-NQO (P < 0.01), while the ebfC⁺⁺ strain remained as resistant as wt and ΔebfC P-ebfC controls (P > 0.1) (Supplementary Fig. S8B). These results corroborate the findings of the chronic assay and suggest that EbfC does not function in the overall response to replication stress but instead contributes to the specific tolerance to or removal of 4NQO-induced lesions.

Since EbfC and Hbsu differentially modulate nucleoid compaction (Fig. 6 and Supplementary Fig. S6), we hypothesized that the distinct sensitivity to DNA damage observed in specific NAP mutants might reflect a failure to modulate global nucleoid compaction upon DNA damage. To investigate this, cells were acutely exposed to increasing concentrations of MMS or 4NQO, and nucleoid area was assessed by fluorescence microscopy. Nucleoids became significantly more compacted after DNA damage, with up to ∼10% compaction observed at the highest doses tested (Supplementary Fig. S8C and D). Importantly, the pattern of nucleoid compaction in ΔebfC, ebfC⁺⁺, ΔebfC P-ebfC, and hbs55 cells closely matched that of wt cells following DNA damage, irrespective of each strain’s sensitivity (or lack thereof) to the damaging agents. These findings suggest that the sensitivity of NAP mutants to DNA damage does not arise from failures in the ability of the cell to modulate global nucleoid compaction but rather reflects impaired local nucleoid architecture remodelling. Additionally, we cannot currently rule out the possibility that NAPs directly affect DNA repair mechanisms, as has been shown for Hbsu [85].

Following MMC exposure, ΔebfC cells were significantly sensitive (P < 0.01), while ebfC⁺⁺, Δrok, hbs55, and ΔlrpC mutants remained recombination proficient and as resistant as wt (LD₉₀ > 150 nM, P > 0.1) (Fig. 7G–I and Supplementary Table S5). Double mutants ΔebfC hbs55, ΔebfC Δrok, and ΔebfC ΔlrpC were as sensitive as the ΔebfC parent (P > 0.1) (Fig. 7G and H, and Supplementary Table S5). The remaining double mutant strains were as sensitive to MMC as wt (P > 0.1) (Fig. 7G–I, Supplementary Fig. S7C, and Supplementary Table S5).

H₂O₂ treatment revealed high sensitivity in ΔlrpC (LD₉₀ ∼0.13 mM H₂O₂, p < 0.01), sensitivity in ΔebfC (P < 0.05), and mild sensitivity in ebfC⁺⁺ (P < 0.01), while hbs55, Δrok, and Δrok hbs55 strains remained resistant as the wt control (LD₉₀ > 0.6 mM H₂O₂, P > 0.1) (Fig. 7J–L, Supplementary Fig. S7D, and Supplementary Table S5). Notably, ΔlrpC cells were as sensitive as the ΔrecR control (Fig. 7J and K, and Supplementary Table S5). Double mutants ΔebfC ΔlrpC, ebfC⁺⁺ ΔlrpC, Δrok ΔlrpC, and hbs55 ΔlrpC were as sensitive as the most sensitive parent strain (P > 0.1), indicating no additive effect of ΔebfC and ΔlrpC mutations (Fig. 7J–L, Supplementary Fig. S7D, and Supplementary Table S5). Given this non-additive effect, and the fact that LrpC neither regulates the DNA damage response nor the expression of function(s) involved in the removal of oxidative DNA damage [41], we must assume that the observed phenotype results either from LrpC-mediated DNA bridging [68, 86] or from a polar effect on the downstream topB gene. We discarded the latter hypothesis, as the ΔtopB mutant strain remained as capable of repairing H₂O₂-induced DNA damage as the wt control (P > 0.1) (Supplementary Fig. S7D and Supplementary Table S5). Characterizing the molecular mechanism by which LrpC influences the removal of oxidative lesions should be a focus of future studies.

Together, these findings suggest that (i) EbfC contributes to genome stability upon 4NQO, MMC, and H₂O₂ treatment by modulating local nucleoid compaction and global chromosomal loop anchoring; (ii) inactivation of ebfC does not induce dysregulation of the DNA damage response to replication stress as ΔebfC cells remain proficient in the repair of MMS-induced lesions that trigger a replication stress (reviewed in 87); iii) EbfC promotes cellular resilience to different types of DNA damage rather than merely physically shielding DNA or untwisting the DNA for the removal of alkylating lesions via base excision repair, as Hbsu does [85]; (iv) Hbsu is required for repair of both MMS- and 4NQO-induced lesions, LprC for oxidative H₂O₂-induced DNA lesion repair, whereas Rok is dispensable for DNA repair; (v) both hbs55 and ΔebfC cells compact the nucleoid in response to DNA damage, indicating that their sensitivity cannot be attributed to defective global nucleoid compaction modulation; and (vi) NAPs likely cooperate to maintain a dynamic nucleoid structure, as ebfC is epistatic to hbs and lrpC in response to DNA damage.

Conclusions

In this study, we integrate both in vivo and in vitro evidence to identify EbfC as a bona fide NAP, and propose the existence of two distinct classes of NAPs in B. subtilis: a nucleoid-condensing, including Hbsu, Rok, and LrpC, and a nucleoid-decondensing class, comprising EbfC. Specifically, we demonstrate that, in vitro, EbfC binds both ssDNA and dsDNA in a Mn²⁺-dependent manner, with a preference for ssDNA (Fig. 1). EbfC forms higher-order nucleoprotein structures, including DNA bridges, and protects DNA from DNase I digestion (Figs 2 and 3), supporting its function as a DNA architectural NAP. In vivo, EbfC co-localizes with the nucleoid (Fig. 4) and modulates nucleoid architecture: deletion of ebfC results in nucleoid condensation, while its overexpression promotes decondensation (Fig. 6). Conversely, inactivation of rok, lrpC, or hbs leads to nucleoid decondensation, and such effect is partially suppressed in the ΔebfC background (Fig. 6), suggesting mechanistic differences among these NAPs, though all four NAPs participate in DNA bridging and wrapping [36, 41, 50, 68, 74, 76–78, 86].

This work also provides a foundation for understanding the crucial role of NAPs in horizontal gene transfer in naturally competent B. subtilis cells, a state during which DNA transcription and replication cannot affect chromosome organization (see “Introduction” section). We propose that NAPs, by contributing to chromosome reorganization, indirectly regulate the acquisition of foreign DNA. Under these conditions, global topological homeostasis of the chromosome is dynamically maintained by a delicate balance between the antagonistic activities of (i) Hbsu, Rok, and LrpC, which promote nucleoid condensation and thereby restrict RecA-mediated DNA strand invasion; and (ii) EbfC, which promotes a more relaxed chromosomal topology, essential for homology-directed RecA-mediated DNA integration (Fig. 5). Importantly, these effects cannot be attributed to altered competence development or DNA uptake, as NPT remains unaffected.

To summarize our findings on the role of NAPs during NCT, we propose the following model (Fig. 8). In a decondensed chromosomal state (as upon inactivation of rok, hbs, or lrpC, or ebfC overexpression), the unconstrained base-stacking within dsDNA should overcome topological barriers to NCT, thereby enhancing RecA-mediated homology recognition and DSE, and increasing NCT frequency without compromising the fidelity of the reaction (Fig. 8a → b). EbfC overexpression synergistically boosts nucleoid relaxation, and thereby intra- and interspecies NCT, in the hbs55, ΔlrpC, or Δrok backgrounds (Fig. 8b → f). However, in the absence of RecD2—a negative modulator of RecA dynamics—RecA nucleoprotein filaments should be longer and long lived [13], impairing DSE and reducing NCT efficiency (Fig. 8a → c). Deletion of recD2 in ebfC⁺⁺, Δrok, hbs55, or ΔlrpC backgrounds leads to synergistic NCT reduction (Fig. 8b → d or 8c → d), with similar effects observed upon removal of the negative modulator RecX [10, 11, 13]. We propose that, in the absence of the topological constraints imposed by Hbsu, Rok, LrpC, or prevented by increased levels of EbfC, impaired RecA filament dynamics results in pronounced unproductive DSE and further impairment in NCT. In contrast, ebfC inactivation results in a more compacted nucleoid, likely hindering early homology search steps and reducing NCT frequency (Fig. 8a → e). Deletion of ebfC reverses the nucleoid relaxation seen in hbs55, Δrok, or ΔlrpC strains, restoring NCT frequencies to near wt levels (Fig. 8b → a).

Figure 8.

Model for the roles of NAPs and recombination functions during NCT. (a) The chromosome is maintained in a steady-state topology by NAPs. (b) Absence (-) of Hbsu, Rok, or LrpC, or overexpression of ebfC, transiently decondenses the nucleoid. This relaxed chromosomal state facilitates RecA-mediated DSE without compromising fidelity, by promoting more efficient access to homologous regions. (c) In the absence of RecD2 (or RecX), which assists RecA, RecA interactions with short homologous regions should be random, persistent and unproductive (dotted lines), reducing DNA strand invasion and three-strand exchange efficiency. (d) The accumulation of such unproductive intermediates in the absence of RecD2 should be favoured under relaxed nucleoid conditions (as in the absence of Hbsu, Rok, or LrpC, or upon ebfC overexpression), synergistically impairing NCT efficiency. (e) Deletion of EbfC results in nucleoid hyper-condensation, reducing NCT frequencies, though when combined with absence of Hbsu, Rok, or LrpC, the nucleoid becomes decondensed, restoring NCT frequencies. (f) Combined absence of Hbsu, Rok, or LrpC, or overexpression of ebfC, with deletion of a different NAP (Rok, LrpC, or Hbsu), leads to enhanced nucleoid decondensation, which synergistically improves RecA-mediated DSE efficiency and increases NCT frequency.

As deletion or overexpression of NAPs neither affects bacterial fitness nor NCT fidelity, manipulating their expression offers promising opportunities for optimizing DNA transformation processes in the biotechnological industry. This strategy could enhance the yield of bacterial transformants engineered to express biosynthetic or biodegradative pathways of industrial interest. Future studies should focus on key questions including whether these NAPs regulate the formation or residence time of protein–DNA complexes, and whether their expression levels impact transformation efficiency in other bacterial species. Understanding these mechanisms will also shed light on the broader roles of NAPs in key processes such as bacterial speciation and the dissemination of antibiotic resistance.

Beyond their roles in NCT, these NAPs—with the exception of Rok—also contribute to DNA repair and to mitigate replication stress, and facilitate repair-by-recombination following fork stalling and/or arresting, during exponential growth. Specifically, as shown in Fig. 7 and Supplementary Fig. S7: (i) inactivation of ebfC renders cells sensitive to distorting, oxidizing, nicking, and cross-linking DNA-damaging agents; of hbs to alkylating and distorting agents; and of lrpC to oxidizing and nicking agents; and (ii) the ebfC, hbs, and lrpC genes display epistatic interactions in DNA damage responses, and thus the division of labour among NAPs observed during NCT is here less evident.

Altogether, our data support the hypothesis that modulation of nucleoid structure and dynamics by NAPs plays a crucial role in NCT and DNA repair processes. Additional functions of EbfC in B. subtilis as a NAP cannot be ruled out and warrant further investigation. For instance, NAP binding near promoter regions may influence gene expression by modulating DNA topology or through other mechanisms, as observed for EbfC_Bbu, EbfC_Dra, and YbaB_Eco [43, 44, 88], thereby impacting diverse bacterial processes.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

Funding

This work was supported by the Ministerio de Ciencia e Innovación (MCIN)/Agencia Estatal de Investigación (AEI)/10.13039/501100011033/FEDER, EU, PID2021-122273NB-I00, as well as by the CSIC 202520E100 to J.C.A. Funding to pay the Open Access publication charges for this article was provided by CSIC.

Data availability

The data underlying this article are available in the article and in its online supplementary material. Materials used in this study are available from the authors upon request.