Exclusion systems preserve host cell homeostasis and fitness, ensuring successful dissemination of conjugative plasmids and associated resistance genes

Abstract

Plasmid conjugation is a major driver of antibiotic resistance dissemination in bacteria. In addition to genes required for transfer and maintenance, conjugative plasmids encode exclusion systems that prevent host cells from acquiring identical or redundant plasmids. Despite their ubiquity, the biological impact of these systems remains poorly understood. Here, we investigate the importance of the exclusion mechanism for plasmid dynamics and bacterial physiology at the single-cell level. Using real-time microscopy, we directly visualize how the absence of exclusion results in plasmid unregulated self-transfer, causing continuous and repeated plasmid exchange among host cells. This runaway conjugation severely compromises cell integrity, viability, and fitness, a largely undescribed phenomenon termed lethal zygosis. We demonstrate that lethal zygosis is associated with membrane stress, activation of the SOS response, and potential reactivation of SOS-inducible prophages, as well as chromosome replication and segregation defects. This study highlights how exclusion systems maintain host cell homeostasis by limiting plasmid transfer. Paradoxically, this restriction is critical to the successful dissemination of conjugative plasmids by conferring a selective advantage, which explains their evolutionary conservation and underscores their role in the spread of antibiotic resistance among pathogenic bacteria.

Graphical Abstract

Graphical Abstract.

Introduction

Conjugation is a horizontal gene transfer mechanism through which recipient bacteria acquire genetic information by direct contact from donor bacteria. It was first identified with the discovery of genetic exchange in bacteria involving the F (fertility) factor, an extrachromosomal DNA element capable of autonomous replication and transfer between Escherichia coli cells [1]. Since then, plasmid conjugation has been revealed as a widespread mechanism of horizontal gene transfer, playing a central role in promoting genetic diversity and adaptation within microbial populations [2, 3]. Conjugative plasmids are recognized as the main contributors to the global spread of antibiotic resistance among bacterial pathogens, raising significant concerns about their impact on public health [4–7]. In particular, conjugative plasmids belonging to the IncF incompatibility group are responsible for the worldwide dissemination of resistance and virulence traits among clinical isolates of E. coli and other enterobacterial species [8].

Conjugation in Gram-negative bacteria primarily relies on plasmid-encoded factors involved in the key steps of DNA transfer, including the relaxosome protein complex that prepares the single-stranded DNA (ssDNA) plasmid before transfer, the type IV secretion system (T4SS) that transports the ssDNA plasmid through the bacterial membrane, and the sex pili that mediate mating pair formation (Mpf) [9–11]. Another conserved feature of conjugative plasmids is exclusion systems, encoded by diverse conjugative elements, including IncF, IncP, IncN, IncW, IncI, and IncH plasmids, mobilizable plasmids, integrative and conjugative elements, and conjugative elements found in Gram-positive bacteria [12–15]. Exclusion systems prevent the redundant acquisition of conjugative plasmids into a host cell that already contains identical or a closely related element. They achieve this by making the plasmid-carrying cell a poor recipient for conjugation, thereby reducing its efficiency of plasmid uptake. Although the precise molecular mechanisms underlying exclusion remain elusive, exclusion is known to be mediated by plasmid-encoded membrane proteins, which are classified into two groups based on their mode of interference in the conjugation process. In the F-like plasmids, surface exclusion factors (the outer-membrane protein TraT) inhibit Mpf, while entry exclusion factors (the inner-membrane protein TraS) block DNA translocation following Mpf [16–20].

Early observations underscore the importance of exclusion for plasmid stability, reporting that F plasmid mutants harboring traS and traT mutations can never be isolated by chemical mutagenesis, suggesting that self-transfer-proficient mutants become highly unstable [12, 21]. Furthermore, early genetic investigations have revealed the role of exclusion in protecting cells from lethal zygosis, a phenomenon initially described during mating with a large excess of Hfr (high frequency of recombination) donor cells capable of transferring the entire 4.6 Mb chromosome, which severely compromises the viability of the recipient cells [22, 23]. These findings prompted the proposal that exclusion systems serve to shield host cells from excessive conjugation, which can disrupt essential cellular processes. Two main mechanisms have been proposed to explain lethal zygosis. First, the physical attachment of the conjugative pilus to the cell surface, along with the establishment of a complex protein channel through the cellular envelopes during mating, may affect the integrity of the cell membranes. Second, superinfection by multiple copies of the ssDNA plasmid or increased copy number of double-stranded DNA (dsDNA) plasmids could deplete the recipient cell’s metabolic or DNA-processing resources and overwhelm the host cell resources. Both mechanisms have received some support from early studies. The addition of the outer membrane-disrupting antibiotic polymyxin B to the mating mix reduced the viability of transconjugant cells [24], while the specific inhibition of DNA transfer using nalidixic acid was found to partially suppress lethal zygosis [25]. Despite these early proposals, the precise mechanisms underlying the phenomenon of lethal zygosis are still uncharacterized, and the specific physiological significance of exclusion systems remains largely unknown. To address this gap in understanding, we investigate the role of exclusion systems in plasmid dynamics and host cell physiology at the single-cell level.

Materials and methods

Bacterial strains, plasmids, and growth

Bacterial strains are listed in Supplementary Table S1, plasmids in Supplementary Table S2, and oligonucleotides in Supplementary Table S3. Fusion of genes with fluorescent tags and gene deletion on the F plasmid used λRed recombination [26, 27]. Modified F plasmids were transferred to the background strain K12 MG1655 by conjugation. Where multiple genetic modifications on the F plasmid were required, the kan and cat genes were removed using site-specific recombination induced by expression of the Flp recombinase from plasmid pCP20 [27]. Plasmid cloning was done by Gibson assembly and verified by Sanger sequencing (Eurofins Genomics biotech). Strains and plasmids were verified by Sanger sequencing (Eurofins Genomics). Cells were grown at 37°C in M9 medium supplemented with glucose (0.2%) and casamino acid (0.4%) (M9-CASA) before imaging, and in Luria–Bertani (LB) broth for conjugation efficiency assays. When appropriate, supplements were used in the following concentrations: ampicillin (Ap) 100 μg/ml, chloramphenicol (Cm) 20 μg/ml, kanamycin (Kn) 50 μg/ml, streptomycin (St) 20 μg/ml, and tetracycline (Tc) 10 μg/ml.

Conjugation assays

Overnight cultures in LB of recipient and donor cells were diluted to an OD₆₀₀ of 0.05 and grown until an OD₆₀₀ between 0.7 and 0.9 was reached. Twenty-five microliters of donor and 75 μl of recipient cultures were mixed into an Eppendorf tube and incubated for 90 min at 37°C. One milliliter of LB was added gently, and the tubes were incubated again for 90 min at 37°C. The conjugation mix was vortexed, serially diluted, and plated on LB agar supplemented with X-gal (40 μg/mL), IPTG (20 μM), and the appropriate antibiotics to select for recipient or donor populations. Recipient (R) colonies were then streaked onto LB agar plates containing the appropriate antibiotic to select for transconjugants (T), and the frequency of transconjugants is calculated using the following formula: T/R + T.

Live-cell microscopy experiments

Overnight cultures in M9-CASA were diluted to an A₆₀₀ of 0.05 and grown until A₆₀₀ = 0.8 was reached. Conjugation samples were obtained by mixing 25 μl of donor and 75 μl of recipient into an Eppendorf tube. For time-lapse experiments, 50 μl of the pure culture or conjugation mix was loaded into a B04A microfluidic chamber (ONIX, CellASIC^®) [28]. Nutrient supply was maintained at 1 psi and the temperature maintained at 37°C throughout the imaging process. Cells were imaged every 1 or 5 min for 90–120 min. For snapshot imaging, 10 μl samples of clonal culture or conjugation mix were spotted onto an M9-CASA 1% agarose pad on a slide [29] and imaged directly.

Image acquisition

Conventional wide-field fluorescence microscopy imaging was carried out on an Eclipse Ti2-E microscope (Nikon), equipped with ×100/1.45 oil Plan Apo Lambda phase objective, ORCA-Fusion digital CMOS camera (Hamamatsu), and using NIS software for image acquisition. Acquisitions were performed using 50% power of a Fluo LED Spectra X light source at 488 and 560 nm excitation wavelengths. Exposure settings were 100 ms for Ypet, sfGFP, and mCherry, and 50 ms for phase contrast.

Image analysis

Quantitative image analysis was done using Fiji software with the MicrobeJ plugin [30]. For snapshot analysis, cells’ outline detection was performed automatically using MicrobeJ and verified using the manual-editing interface. For time-lapse experiments, detection of cells was done semiautomatedly using the manual-editing interface, which allows to select for the cells to be monitored and automatically detects the cell outlines. Within conjugation populations, donor (no mCh-ParB signal), recipient (diffuse mCh-ParB signal), or transconjugant (mCh-ParB foci) category was assigned using the “Type” option of MicrobeJ. Recipient cells were detected on the basis of the presence of red fluorescence above the cell’s autofluorescence background level detected in the donors. Among these recipient cells, transconjugants were identified by running MicrobeJ automated detection of the ParB fluorescence foci (Maxima detection). This approach was used independently of the presence or the absence of the Ssb-Ypet or sfGFP fusions within donor and recipient cells. Within the different cell types, mean intensity fluorescence (a.u.), skewness, signal/noise ratio (SNR), or cell length (μm) parameters were automatically extracted and plotted using MicrobeJ. SNR corresponds to the mean intracellular signal/mean noise signal ratio, where the mean intracellular signal is the fluorescence signal per cell area and the noise is the signal measured outside the cells (due to the fluorescence emitted by the surrounding medium). In contrast with the total amount of fluorescence per cell, which depends on the cell size/age and accounts for the background, SNR quantitative estimate is more appropriate for unbiased quantification of intracellular fluorescence over time. Ssb-Ypet, SsbF-mCh, and mCh-ParB foci were detected using MicrobeJ’s Maxima detection function, and foci localization and fluorescence intensity were extracted and plotted automatically. Plots presenting time-lapse data were aligned to the first frame where the transconjugant cell exhibits either a conjugative Ssb-Ypet focus (ssDNA acquisition) or an mCh-ParB focus (ss-to-dsDNA conversion) as indicated in the corresponding figure legend.

Statistical analysis

P-value significance was analyzed by running specific statistical tests on the GraphPad Prism software. Single-cell data from quantitative microscopy analysis were extracted from the MicrobeJ interface and transferred to GraphPad. P-value significance of single-cell quantitative data was performed using unpaired nonparametric Mann–Whitney statistical test, which allows to compare differences between independent data groups without normal distribution assumption. P-value significance for the frequency of transconjugants obtained by plating assays was evaluated using one-way analysis of variance (ANOVA) with Dunnett’s multiple comparison test, which allows to determine the statistical significance of differences observed between the means of three or more independent experimental groups against a control group mean (corresponding to the Fwt). When required, P-value and significance are indicated on the figure panels and within the corresponding legend.

Results

Inactivation of exclusion triggers unrepressed plasmid self-transfer between host cells

The efficiency of exclusion is quantitatively assessed by estimating the exclusion index (EI), calculated by comparing the frequency of plasmid acquisition by a recipient strain harboring the plasmid to the frequency observed in the isogenic plasmid-free strain [12]. The wild-type F (Fwt) plasmid, which carries the exclusion genes traS and traT, exhibits an EI of ∼474. This means that E. coli cells containing the Fwt plasmid are 474 times less receptive to acquiring an additional plasmid than plasmid-free cells. Consistent with previous estimates, we calculate that single-deletion mutants lacking either traT (FΔtraT) or traS (FΔtraS) exhibit reduced EIs of 28 and 3, respectively. Most importantly, we report for the first time the double deletion mutant (FΔtraST), which completely abolishes exclusion, as reflected by an EI of ∼1 (Supplementary Fig. S1A). Exclusion is restored by complementation plasmids, both in the presence (Supplementary Fig. S1B) and in the absence (Supplementary Fig. S1C) of F plasmids, with TraS playing a more significant role than TraT. However, full exclusion requires the synergistic action of both proteins.

To investigate the impact of exclusion loss on plasmid dynamics and host cell physiology, we examined cells harboring the FΔtraST plasmid. In this clonal population, each cell is expected to function as both a plasmid donor and recipient. However, because all cells are genetically identical, traditional approaches based on resistance marker selection to measure plasmid transfer are not applicable. Therefore, we used live-cell microscopy coupled with established fluorescent reporters to visualize ssDNA plasmid transfer at the single-cell level [31, 32]. Specifically, we used a fluorescent fusion of the chromosomally encoded single-strand binding protein Ssb-Ypet and observed the frequent formation of bright membrane-associated foci on the ssDNA plasmid on both sides of the conjugation pore, reflecting transfer between FΔtraST-carrying cells (Fig. 1A and Supplementary Movie S1). We estimate a dramatic increase in ssDNA plasmid frequency from 0.0063 transfers per cell per hour for Fwt compared to 1.31 for FΔtraST (Fig. 1B). This visualization approach provides the first direct observation of unrepressed ssDNA plasmid self-transfer among host cells triggered by the absence of exclusion system.

Figure 1.

Loss of exclusion leads to deregulated plasmid transfer and impaired host cell viability. (A) Time-lapse microscopy images of a clonal population of FΔtraST cells carrying the ssb-ypet fusion gene, illustrating the transfer of ssDNA. The intensity of Ssb-Ypet foci is represented by the color scale bar on the right. Scale bar: 1 μm. (B) Histogram showing the number of plasmid transfer events per cell per hour for cells carrying either Fwt or FΔtraST. Data represent mean and standard deviation (SD) from (n) individual transfer events across at least three independent biological replicates (black dots). Statistical significance was assessed using an unpaired t-test (**P = .0072). (C) Time-lapse phase-contrast images showing the filamentation phenotype in clonal populations of FΔtraST, FΔtraST cells treated with sodium dodecyl sulfate (SDS), FΔtraIST, FΔtraAST, and lexA3^ind−FΔtraST. Scale bar: 10 μm. (D) Histograms showing the proportion of live and dead in Fwt, FΔtraS, and FΔtraT strains and their FΔtraST plasmid derivatives, as determined by the live/dead staining assay. Bars represent mean and SD from at least three independent experiments. (E) Growth curves of Fwt and FΔtraST mutant cells, measured by OD₆₀₀ every 30 min over 240 min. Cells were grown in LB medium in triplicate; only the average curve is shown. (F) Histograms showing the percentage of viable cells, defined by their ability to divide during time-lapse microscopy. Bars represent the mean ± SD from at least three independent biological replicates. (G) Schematic representation of the Fwt and FΔtraST transfer regions and the corresponding genetic maps of the three suppressor clones (A, B, and C).

Unrepressed plasmid self-transfer induces host cell viability defects

Monitoring bacterial growth within a microfluidic chamber over 180 min using phase-contrast imaging shows that FΔtraST-carrying cells exhibit severe morphology defects, including the formation of filamentous and ghost cells (Fig. 1C and Supplementary Movie S2). Using live/dead staining and microscopy snapshot analysis, we quantified ∼29% dead cells in the population (Fig. 1D). These phenotypes result in growth defects and decreased viability, as demonstrated by OD monitoring (Fig. 1E) and microscopy-based microcolony assays (Fig. 1F), respectively. These deficiencies are observed in the complete absence of exclusion genes, while only minor or no effects are observed in the traS or traT single mutants (Fig. 1D and Supplementary Fig. S1D). We further show that FΔtraST-carrying cells exhibit increased uptake of SYTOX staining dye (Supplementary Fig. S1E), which penetrates cells with compromised plasma membranes, and DiBAC₄[3] staining dye, which penetrates depolarized cells with impaired transmembrane potential (Supplementary Fig. S1F). Such alterations of membrane permeability are likely related to the formation of ghost cells (Fig. 1C and D).

Crucially, cell filamentation and death, viability, and growth defects induced by the absence of exclusion systems are suppressed when conjugation is abolished, either chemically by adding SDS to the microfluidic chamber or genetically by deleting the relaxase gene traI or the pilin gene traA, which are essential for plasmid processing and transfer and pilus formation, respectively (Fig. 1C–F and Supplementary Movie S2). This indicates that derepressed plasmid self-transfer is the sole cause of the observed proliferation defects. Consistent with this interpretation, suppressor mutants of the FΔtraST-carrying strain that regained normal growth in liquid conditions (Supplementary Fig. S1G) have acquired mutations that drastically reduced or eliminated their capacity to transfer the FΔtraST plasmid (Supplementary Fig. S1H). These mutations are found in the tra region, including a Tn10 insertion interrupting the traA pilin gene essential for pilus formation, an IS2 insertion disrupting the traQ gene involved in pilin maturation, and a 22.85-kb deletion encompassing the origin of transfer (oriT) and part of the tra operon (Fig. 1G). Altogether, these results demonstrate that unrepressed plasmid self-transfer results in severe impairment of cell integrity and viability, which imposes an important fitness burden to host cells.

Unrepressed plasmid self-transfer induces membrane stress

In search for the mechanism by which unrepressed self-transfer impacts the physiology of the cells, we addressed its effect on the induction of five envelope stress response (ESR) pathways, namely Cpx, Rcs, Psp, SigmaE, and Bae. These ESR systems are known to be activated by specific perturbations of the cell envelope [33, 34], i.e. the Cpx system by the presence of misfolded periplasmic proteins and by surface adhesion [35, 36], Rcs by disturbances in the outer membrane [37], Bae by toxic compounds [38, 39], and the phage-shock response Psp system by alterations in proton motive force [40–42], while the envelope heat-shock sigma factor σ^E responds to the accumulation of unfolded outer membrane proteins or lipopolysaccharide alterations [43]. We used transcriptional fusions coupling the promoters of genes specifically regulated by these ESRs (cpxP, rcsA, pspA, micA, and mdtA) to the gene encoding the fluorescent protein mNeonGreen (mNG) [34]. The activation of these reporters was monitored in FΔtraST-carrying cell populations during exponential growth using single-cell microscopy analysis. No induction of Psp, Bae, or SigmaE pathways is observed in cells carrying FΔtraST cells compared to cells carrying an Fwt plasmid (Supplementary Fig. S2A–C). However, we measured a significant induction of the Rcs (Fig. 2A) and the Cpx (Fig. 2B) pathways in FΔtraST-carrying cells, indicative of a response to outer membrane perturbations and surface adhesion during unregulated self-transfer. Rcs and Cpx activations are mainly due to the absence of the surface exclusion protein traT, while weaker induction was observed in cells lacking entry exclusion (FΔtraS) (Fig. 2A). This suggests that stress pathways activation is related to processes occurring during the formation of the mating pair between cells. Consistently, Rcs and Cpx induction is fully abolished in FΔtraAST-carrying cells, which are unable to form mating pairs due to the lack of pili. However, Rcs and Cpx induction is also abolished by the deletion of the relaxase (FΔtraIST) that retains pilus formation while impeding the processing of the plasmid required for transfer (Fig. 2A and B). This suggests the activation of membrane stress pathways is due to interactions occurring between mating cells during repeated self-transfer of ssDNA plasmid through the conjugation pore.

Figure 2.

Exclusion systems limit ESRs during plasmid transfer. (A) Quantification of P_rcsA-mNG induction at the single-cell level in Fwt, FΔtraS, FΔtraT(in grey), FΔtraST(in blue), and its derivative mutants during exponential growth. Each dot represents an individual cell; black dots indicate the mean fluorescence value for each independent biological replicate. Statistical significance was assessed using an unpaired t-test on replicate means (*P< .05;**P< .005;****P< .0001). (B) Quantification of P_cpxP-mNG induction at the single-cell level in Fwt, FΔtraS, FΔtraT (in grey), FΔtraST (in blue), and its derivative mutants during exponential growth. Each dot represents an individual cell; black dots indicate the mean fluorescence value for each independent biological replicate. Statistical significance was assessed using an unpaired t-test on replicate means (*P< .05; ***P< .001). (C) Histograms showing the proportion of live and dead cells in strains carrying deletions in the Rcs (ΔrcsDB) and/or Cpx (ΔcpxQPRA) pathways, either without a plasmid (F−) or harboring FΔtraST. Cell viability was assessed using the live/dead staining assay. Bars represent the mean ± SD from at least three independent biological replicates.

We measured a lesser but detectable induction of the Rcs pathway when conjugation was performed between Fwt donor cells and wt recipients carrying the Rcs reporter (Supplementary Fig. S2D). This indicates that outer membrane stress also occurs during normal conjugation, but is dramatically amplified by unrepressed self-transfer. Finally, we questioned whether Rcs and Cpx response pathways play any role in the viability of host cells acquiring plasmid DNA. However, we observed no effect of the deletion of these pathways on the proportion of dead cells in the presence or absence of exclusion (Fig. 2C). Altogether, these results show that, while membrane stress is induced during conjugation, their downstream regulatory effects and biological functions are not critical to the survival of the host cells.

Unrepressed plasmid self-transfer induces activation of the SOS response

Next, we addressed whether the excessive entry of ssDNA during unregulated self-transfer induces the SOS response, possibly accounting for the observed viability and cell morphology defects. Using a transcriptional sulA fluorescent reporter (P_sulAGFP), we measure a significant induction of the SOS response in FΔtraST-carrying cells compared to cells containing Fwt or no plasmid (Fig. 3A). SOS induction is strictly attributable to unrepressed self-transfer, as it is fully suppressed when conjugation is impeded by the addition of SDS (Fig. 3A) or by deletion of traI or traA (Supplementary Fig. S3A). Consistent with this interpretation, we confirmed that the activation of the SOS is not due to the presence of DNA damage in FΔtraST-carrying cells. To show this, we visualized a fluorescent fusion of the recombination protein RecA (RecA-YFP), which is diffuse in normally growing cells and forms intracellular structures in response to DNA damage [44, 45]. The formation of intracellular structures by fluorescent proteins can be reported by an increase in fluorescence skewness [31, 46]. Analysis showed that while RecA-YFP skewness is significantly increased in response to UV-induced DNA damage (0.025 J/m²), it remained unchanged in FΔtraST cells compared to F− and Fwt cells (Supplementary Fig. S3B).

Figure 3.

Induction of the SOS response and phage production upon exclusion system loss. (A) Quantification of P_sulA-GFP reporter induction at the single-cell level in F−, Fwt(in grey), FΔtraST(in blue), or its derivatives. Each dot represents an individual cell; black dots indicate the mean fluorescence value for each independent biological replicate. Statistical significance was assessed using an unpaired t-test on replicate means (**P< .005,***P< .001). (B) Analysis of P_sulA-GFP induction as a function of cell length in wt (grey circles) and lexA3^ind− (green circles) strains carrying the Fwt plasmid. The number of cells analyzed (n) is indicated, based on at least three independent experiments. (C) Analysis of P_sulA-GFP induction as a function of cell length in wt (blue circles) and lexA3^ind− (pink circles) strains carrying the FΔtraST plasmid. The number of cells analyzed (n) is indicated, based on at least three independent experiments. (D) Quantification of Φ80 prophage production in Fwt, FΔtraST, and plasmid-free cells (F−). Data represent the mean ± SD from three independent experiments (black dots). Statistical significance was determined using a two-sided Mann–Whitney test (*P=.0443; ns, not significant). (E) P_sulA-GFP intensity measured at the single-cell level before mating and after 2 and 4 h of conjugation between a donor strain lacking ompA and carrying FΔtraST plasmid and a plasmid-free recipient cell. Each dot represents an individual cell; black dots indicate the mean fluorescence value for each independent biological replicate.

To test the possibility that SOS activation could account for viability defects observed in the absence of exclusion, we examined the effect of the SOS-defective lexA3^ind−mutation. While the lexA3^ind mutation abolishes the activation of the SOS response in the FΔtraST cells (Fig. 3A), it does not suppress filamentation (Figs 1C and 3B and C) or cell death (Fig. 1D), and does not improve viability or cell growth (Fig. 1E and F). We therefore conclude that the onset of the SOS response is not responsible for the viability defects triggered by the absence of exclusion. However, we further hypothesized that the induction of strong levels of SOS response due to the loss of exclusion could result in the awakening of SOS-inducible prophage potentially located into the host chromosome. To test this possibility, we measured the concentration of phage in the supernatant of cultures of wt E. coli strains carrying the Φ80 prophage, with or without F derivatives. We observed a 2-log increase in phage plaque-forming units (PFU/ml) in cultures with the FΔtraST plasmid compared to those carrying Fwt or F− cells (Fig. 3D).

Our experimental setup gave us the unique opportunity to address whether, in the FΔtraST population where each cell can act as both a donor and a recipient, the induction of SOS occurs within the donor due to plasmid donation or within the recipient due to plasmid acquisition, or both. To address that specific question, we used a ΔompA donor strain that has reduced capacity to receive the plasmid, as OmpA protein is critical to the mating pair formation [47, 48]. The ΔompA/FΔtraST donor retains plasmid donation proficiency (Supplementary Fig. S3C), but has ∼200-fold reduced plasmid acquisition capability (Supplementary Fig. S3D). During conjugation, SOS induction is not observed in the ΔompA/FΔtraST donor, but only in the subpopulation of transconjugant cells that have received the FΔtraST plasmid (Fig. 3E).

All together, these findings demonstrate that the activation of the SOS response in the absence of exclusion systems is due to repeated plasmid acquisition rather than repeated plasmid donation. However, SOS induction is not responsible for the viability defects induced by unrepressed plasmid self-transfer.

Impact of the loss of exclusion on replicon maintenance

Next, we questioned whether the loss of exclusion could result in a deregulation of plasmid copies per cell or maintenance, which could in turn have deleterious impact on the physiology of the host cell. Indeed, an excess in plasmid copies per cell might overwhelm the cell’s resources and induce a dramatic fitness cost [49]. Hence, we addressed whether the ssDNA plasmids acquired through unrepressed self-transfer can be successfully converted into dsDNA plasmids. To this end, we examined mating between two parental strains, one harboring the FΔtraST plasmid labeled with a green fluorescent sfGFP-ParB_P1/parS_P1 system (F^GFP+ cells) and the other carrying a red fluorescent mCh-ParB_PMT1/parS_PMT1 system (F^mCh+ cells). Upon mating between FΔtraST^GFP+ and FΔtraST^mCh+ cells, we observed the rapid emergence of dual-labeled cells containing both plasmids (Fig. 4A and Supplementary Movie S3), constituting ∼70% of the population after 4 h of mating (Fig. 4B). Notably, dual-labeled cells were rarely detected when the conjugation mix was treated with SDS, which impedes conjugation by depolymerizing the conjugative pilus (Fig. 4B), or when the experiment was performed with Fwt plasmids (Fig. 4C). These observations indicate that ssDNA FΔtraST plasmids that are transmitted among and between FΔtraST-bearing cells can, at least partially, be successfully converted into dsDNA.

Figure 4.

Impact of the loss of exclusion on replicon maintenance. (A) Time-lapse microscopy images showing the transfer of dsDNA in FΔtraST-carrying cells, indicated by the appearance of a GFP focus in an mCH⁺ cell or vice versa. Scale bar: 1 μm. (B) Quantification of plasmid transfer over time in mating assays between FΔtraST^GFP+and FΔtraST^mCH+ strains. Bars represent the proportion of GFP⁺, mCH⁺, and double-positive GFP⁺mCH⁺ cells at the indicated time points, with or without SDS treatment. Data represent the mean and SD from three independent biological replicates (n = 3). (C) Quantification of plasmid transfer over time in mating assays between Fwt^GFP+and Fwt^mCH+ strains. Bars represent the proportion of GFP⁺, mCH⁺, and double-positive GFP⁺mCH⁺ cells at the indicated time points. Data represent the mean and SD from three independent biological replicates (n = 3). Histograms showing the distribution of oriC (D) and oriT (E) foci number per cell in populations carrying either the Fwt or FΔtraST plasmid, as determined by fluorescence microscopy. (F) Histogram showing the oriT/oriC ratio in cells carrying the Fwt, FΔtraS, FΔtraT, or FΔtraST plasmid. The experiment was performed from at least three independent biological replicates. P-value significance and ns were obtained from Mann–Whitney two-sided statistical test. (G) Stability assay of Fwt, FΔtraS, FΔtraT, or FΔtraST plasmid in the absence of antibiotic selection showing plasmid retention rate after 20, 40, 60, and 80 generations of growth.

Because newly acquired ssDNA plasmids can be converted into dsDNA plasmids, we asked whether repeated self-transfer alters the intracellular copy number of the F plasmid. To test this, we quantified the oriT/oriC ratio in a strain carrying a parS_P1 insertion in the ilvA chromosome locus located near the origin of replication (oriC) and harboring the F plasmid with a parS_PMT1 insertion adjacent to the origin of transfer (oriT) (Supplementary Fig. S4A and B). Quantitative image analysis reveals that the FΔtraST population exhibits ∼10% of cells without oriC and oriT foci, a phenotype absent from the Fwt control (Fig. 4D and E). Size distribution analysis showed that these focus-negative cells are predominantly small, consistent with recently divided newborn cells, which are likely nonviable and leading to the eventual formation of ghost cells (Supplementary Fig. S4C). Nonetheless, most FΔtraST cells are normal-length or filamentous and exhibit an oriT/oriC ratio of ∼2.4, similar to Fwt cells, indicating that the chromosome-to-plasmid ratio is maintained in absence of exclusion (Fig. 4F). Consistent with plasmid maintenance, stability assay shows that, similar to Fwt, all viable cells retain FΔtraST plasmid when grown over 80 generations without selection pressure (Fig. 4G).

Unrepressed plasmid self-transfer induces cell cycle disruption

Next, we characterized the impact of superinfection with ssDNA plasmid on the host’s cell chromosome dynamics. In particular, we characterized chromosome intracellular positioning using DAPI staining and replication intracellular organization using a fluorescent fusion of the replisome’s B2 subunit (mCh-DnaN). In Fwt-containing cells, DAPI-fluorescence demographs of cells sorted by length and corresponding localization heatmaps reflect the progressive segregation of well-separated nucleoid DNA in the course of cell cycle (Fig. 5A). In contrast, FΔtraST-containing cells exhibit impaired nucleoid separation, especially marked in small cells (<4 μm) (Fig. 5A) as well as in filamentous (>8 μm) cells (Supplementary Fig. S5A). This phenotype is associated with mislocalization of the replication machinery. In Fwt-containing cells, mCh-DnaN forms discrete foci regularly positioned at the cell quarter positions, reflecting replisome localization during DNA replication progression (Fig. 5B). In contrast, localization analysis of FΔtraST cells revealed a dramatic impairment of mCherry-DnaN localization, which appears largely diffuse in smaller cells (<4 μm) and strongly mispositioned foci in larger cells (>4 μm) (Fig. 5B), as well as in filamentous cells (Supplementary Fig. S5B). In parallel, we quantified the cell DNA content using Syto9 staining and flow cytometry (Fig. 5C). The results show that most FΔtraST-carrying cells have significantly lower DNA content compared to those carrying the Fwt plasmid, reflecting a significant disruption of chromosome replication.

Figure 5.

Effect of transferred DNA on chromosome organization and replication dynamics. (A) Nucleoid positioning visualized by DAPI staining in strains carrying Fwt or FΔtraST. (B) Replisome localization using an mCh-DnaN fusion in the same strains. For both panels, demographs show signal distribution in cells sorted by length (4–8 μm), and heatmaps display average fluorescence intensity along the normalized cell axis for short (<4 μm) and elongated (4–8 μm) cells. The number of analyzed cells (n) is indicated, based on data from three independent experiments. (C) Quantification of DNA content using Syto9 staining in cells carrying Fwt and FΔtraST. (D) Heatmaps showing the spatial distribution of DAPI-stained nucleoids and replisome localization using mCh-DnaN fusion (E) in exponentially growing cells. Cells are grouped by size: small (<4 μm) and normal-sized (4–8 μm). Data are shown for clonal populations of lexA3^ind−/FΔtraST, wt/FΔoriTST, and wt/FΔoriTST/pmob. The number of analyzed cells (n) is indicated, based on data from three independent experiments.

Nucleoid and replisome mislocalizations persist in lexA3^ind− cells (Fig. 5D and E, and Supplementary Fig. S5C and D), further supporting our previous interpretation that the SOS response does not account for the defects associated with the absence of exclusion. We then wanted to know whether the amplitude of these defects depends on the amount of DNA transferred. To test this, we first abolished FΔtraST plasmid transfer by deleting its origin of transfer (FΔoriTtraST), which restores normal chromosome and replisome dynamics (Fig. 5D and E), suppresses SOS induction as measured using the sulA transcriptional fluorescent reporter (Supplementary Fig. S5E), and prevents the formation of dead cells as assessed by live and dead staining (Supplementary Fig. S5F). We then introduced a 6-kb plasmid (pmob) that carries the oriT from the F plasmid, rendering it mobilizable by the conjugation machinery produced from the coresident FΔoriTtraST plasmid. We observe that the reduction of the size of the transferred DNA from 108 kb of the F plasmid to the 6 kb of the pmob plasmid significantly reduces SOS induction and suppresses the formation of dead cells (Supplementary Fig. S5E and F). Most importantly, it significantly improves nucleoid separation and restores normal replisome localization (Fig. 5C and D). Conversely, we attempted to increase the amount of transferred DNA by removing exclusion systems from an Hfr strain that is able to transfer its entire 4.6 Mb chromosome. However, despite multiple attempts, we were unable to construct a viable deletion strain, further suggesting that unregulated self-transfer with extreme DNA amounts imposes a lethal burden on host cells. These findings are consistent with the interpretation that the amount of transferred DNA is a key determinant of the cellular stress induced by conjugation.

Exclusion safeguards successful plasmid dissemination

The absence of exclusion systems and the consequent unrepressed self-transfer negatively impacts the fitness of the host cell. We wanted to evaluate how these fitness defects influence the dissemination capability of a plasmid devoid of exclusion system within a bacterial population. To address this question, we performed competition assays by mixing equal proportions of cells carrying FΔtraST^RFP+ (RFP fusion) and cells carrying Fwt^GFP+ (GFP fusion). We then monitored their relative ratio of over several timescales. Using fluorescence microscopy, we observe that the proportion of cells carrying FΔtraST relative to Fwt decreases from 1:1 to 1:13 over the first 20 h (Fig. 6A). In comparison, competition between two Fwt populations showed no significant change in their relative proportions (Fig. 6A). This trend was further confirmed after 24 h by fluorescence imaging of bacterial colonies derived from the initial competition mix, revealing a clear reduction in the FΔtraST^RFP+ signal relative to the Fwt^GFP+ (Fig. 6B). To investigate this effect on longer period, we monitored the competition mix over a week using plating assays. Fwt rapidly outcompetes FΔtraST, reaching up to a 5-log advantage after 7 days (Fig. 6C). These experiments demonstrate that exclusion systems are critical to the successful dissemination and maintenance of the plasmid within a bacterial population.

Figure 6.

Exclusion safeguards successful plasmid dissemination. (A) Quantification of plasmid ratios within cell population during mating between Fwt^GFP+ and either FΔtraST^RFP+ or Fwt^RFP+. Ratios were estimated by fluorescence microscopy at 0, 2, 4, and 20 h post-mating, across three independent experiments. (B) Quantification of the GFP/RFP fluorescence signal ratio after 24 h of mating between Fwt^GFP+ and either FΔtraST^RFP+ or Fwt^RFP+, with and without SDS treatment. Representative merged fluorescence images are shown for each mating. Data represent the mean ± SD from three independent experiments (n = 3). (C) Quantification of plasmid ratios over time between Fwt and FΔtraST strains during co-culture. The ratio of colony forming units (CFU) was measured daily over a 7-day mating experiment.

Discussion

This study demonstrates that exclusion systems play a crucial role in maintaining a healthy host population capable of effectively propagating plasmids by mitigating the harmful effects associated with unregulated plasmid transfer and resulting in lethal zygosis. While this conclusion aligns with previous proposals [13, 25, 50], our findings provide a detailed description of the events occurring during lethal zygosis. In the absence of functional exclusion mechanisms, host cells simultaneously act as efficient donors and recipients, initiating a runaway conjugation process characterized by frequent and continuous ssDNA plasmid exchanges. This unrestricted self-transfer profoundly impacts host physiology, causing filamentation, cell death, reduced viability, and significant fitness defects.

A direct consequence of excessive conjugation is the induction of membrane stress and activation of the Rcs and Cpx phosphorelay systems. Several steps within the conjugation process can compromise membrane integrity, including physical interactions mediated by the conjugative pilus, stabilization of mating pairs via OmpA–TraN interactions [51–53], channel formation through the cell envelope by the T4SS, and the passage of ssDNA through the conjugation pore. However, our observations that neither Rcs nor Cpx inactivation affected transconjugant cell viability during either regulated or runaway conjugation strongly suggest that these pathways do not significantly contribute to maintaining host cell homeostasis during plasmid acquisition. Furthermore, membrane stress responses depend primarily on the absence of the surface exclusion protein TraT, which interferes with OmpA–TraN interaction [54], and were entirely suppressed by traA deletion. Thus, membrane stress predominantly arises from events occurring during or after mating pair formation. Inactivation of the traI gene, which impedes plasmid transfer but does not affect pilus production, suppresses membrane stress activation and fully prevents filamentation, cell death, and growth defects. This finding supports the conclusion that the harmful effects on host physiology primarily result from the repeated passage of the ssDNA plasmids through the membrane-spanning conjugation pore rather than by pilus-mediated interactions per se.

We then investigated several potential consequences of superinfection by excessive ssDNA.

We report that unrepressed ssDNA entry triggers the activation of the LexA-dependent SOS response. Our data reveal that unrepressed ssDNA transfer strongly induces the SOS response, specifically due to repeated plasmid acquisition rather than donation. However, SOS induction itself does not explain the proliferation defects observed during unregulated self-transfer. Nonetheless, SOS activation led to secondary effects, notably the reactivation of SOS-inducible prophages. This can lead to further implications, as reactivated prophages can trigger host cell lysis and alter competition among mobile genetic elements, potentially favoring certain plasmids or mobile elements over others.

The absence of exclusion does not result in an increase in FΔtraST plasmid copy number per cell, eliminating resource overload as a cause of fitness defects. The maintenance of FΔtraST copy number supports the model that, despite repeated ssDNA acquisition, the regulation of F plasmid by host-chromosome replication preserves a constant plasmid-to-chromosome ratio [55], further suggesting that not all incoming ssDNA molecules are converted into dsDNA plasmids.

The absence of exclusion also results in the formation of a significant proportion of cells that have lost at least part of their chromosome or plasmid DNA content. While these cells are expected to degenerate into ghost cells, it is more difficult to determine their mechanism of formation. Two mechanisms could theoretically generate these DNA-deficient cells, i.e. the asymmetric polar division of filaments that results in cells with aberrant DNA content [56, 57], or post-segregational killing (PSK) triggered by the activation of toxin–antitoxin modules upon plasmid loss [58]. However, multiple lines of evidence demonstrate that PSK does not account for the growth defects triggered by plasmid self-transfer. First, both normal and filamentous FΔtraST cells retain their plasmids, ruling outkilling through toxin–antitoxin activation after plasmid loss. Second, chemically blocking conjugation with SDS or genetically deleting traA or traI genes fully restores normal growth. Taken together, these observations implicate runaway conjugation itself as the cause of proliferation deficiencies.

We report that a major consequence of superinfection with ssDNA plasmids is a pronounced disruption of the intracellular organization of host chromosome replication and segregation. Based on these findings, we propose a model in which the effects of superinfection on chromosome organization result directly from replication stress, independent of SOS-mediated responses. This replication stress likely stems from the persistent recruitment of essential host factors, such as ssDNA-binding protein (Ssb) and DNA polymerase III that converts the ssDNA plasmids into dsDNA [59, 60], and potentially other replication machinery components onto incoming ssDNA plasmids. The hijacking of these essential factors would consequently limit their availability for chromosome maintenance, explaining the observed chromosome replication defects, reduced DNA content, and impaired cell proliferation. Similar stresses may also arise transiently during regulated conjugation, when functional exclusion systems limit ssDNA entry. Although brief and moderate, these effects likely contribute to the immediate plasmid-acquisition cost that reduces host fitness upon plasmid entry [61, 62]. Under uncontrolled self-transfer, however, the same stresses accumulate continuously, eventually overwhelming cellular functions and imposing an insurmountable fitness burden.

Overall, this work sheds light on the pivotal role of exclusion systems in preserving host cell homeostasis and fitness. Although counterintuitive, restricting plasmid transfer actually enhances the plasmid’s dissemination potential by balancing the need for horizontal gene transfer against maintaining host viability. In evolutionary terms, plasmids that carry well-functioning exclusion systems gain a selective advantage, thereby promoting a more stable coexistence with their bacterial hosts and facilitating the dissemination of conjugative elements and associated antibiotic resistance genes among bacteria. These findings also emphasize the importance of further investigating the largely unknown molecular mechanisms underlying surface and entry exclusion systems. Targeting these systems could induce lethal zygosis in plasmid-bearing cells, opening a novel strategy to limit the spread of antibiotic resistance by conjugation.

Notes

Present address: de Duve Institute, UCLouvain, 75 avenue Hippocrate, 1200 Brussels, Belgium

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by funding from the Foundation for Medical Research (grant number FRM-EQU202103012587 to C.L. and A.C.) and from the French National Research Agency (grant numbers ANR-22-CE12-0032 and ANR-23-CE12-0037). Funding to pay the Open Access publication charges for this article was provided by the French National Research Agency (ANR-23-CE12-0037).

Data availability

All data to understand and assess the conclusions of this research are available in the main text and supplementary material.