Selective elimination of donor bacteria enables global profiling of plasmid gene expression at early stages of conjugation

Abstract

Conjugative plasmids are a major driving force for the dissemination of antimicrobial resistance. During conjugation, plasmid DNA is transferred from the donor cell as a single-stranded (ss) linear molecule. Recent research has highlighted intriguing DNA reactions at the early stages of conjugation that are important for plasmid establishment in the recipient cell, including prompt expression of antidefense genes. However, genomics-based approaches to investigate plasmid establishment have been challenging because the identical fully established plasmids in donor cells mask transconjugant-specific signals. To overcome this limitation, we developed a new method, ED-TA, which exploits a donor mutant hypersensitive to hypoosmotic shock. ED-TA allows unprecedently quick and efficient Elimination of Donor population for Transconjugant Analysis. RNA-seq analyses revealed highly selective and robust induction of plasmid genes during the early stages of conjugation. We experimentally identified transcription start sites of six possible operons that are matched with ssDNA promoter predictions. We also showed that transconjugant gene expression is altered in recipients with perturbed plasmid establishment. As the ED-TA method is straightforward and broadly applicable, it will further our understanding of plasmid establishment processes in the new host cell, including not only gene expression but also ss to double-stranded DNA conversion and plasmid circularization.

Graphical Abstract

Graphical Abstract.

Introduction

Bacterial conjugation has been subjected to a wide range of scientific research for its profound implications in microbial evolution and bioengineering. Particularly, conjugative plasmids serve as the primary vehicles for the dissemination of antimicrobial resistance. Conjugative plasmids encode genes essential for their transfer (transfer genes) that include DNA processing mobility (MOB) and mating pair formation (MPF) genes. During conjugation, the donor cell produces a pilus that facilitates contact with the recipient cell. Subsequently, single-stranded (ss) DNA is stripped from origin of transfer (oriT) by the relaxosome and transferred to the recipient cell through a mating channel. In the recipient cell, ss linear plasmid DNA is converted to a double-stranded (ds) circular form so that it can be stably maintained in the new host cell [1, 2]. Despite the diversity of conjugative plasmids found in natural or clinical environments, their fundamental processes are conserved and their transfer machinery can be classified into only a few MOB/MPF systems [3]. There are several model conjugative plasmids that belong to different plasmid incompatibility classes such as F (IncF), ColIb-P9 (IncI1), R64 (IncI1), pKM101 (IncN), RK2 (IncP1), and R388 (IncW). The transfer genes of these model conjugative plasmids were identified through forward and reverse genetic approaches [4–10]. Tnseq, a high-resolution functional genomics approach [11], was recently used to identify transfer genes across multiple conjugative plasmids in different classes, such as pESBL-EA11 (hereafter referred to as pESBL) from IncI1, pEC958 from IncF, pMS6198A from IncA/C, and pOXA-48 from IncL/M [12–15].

Many conjugative plasmids exhibit lower transfer frequency due to repression of their transfer genes under normal conditions. Mutants defective in such repression, called derepressed or drd, show overexpression of transfer genes including conjugative pili and increased transfer frequency [4, 16, 17]. Early studies also showed that freshly conjugated cells exhibit higher transfer capacity than cells in which the plasmid has been established for a long time [18–20]. This phenomenon is linked to zygotic induction, corresponding to a burst of expression of newly acquired plasmid genes in transconjugant cells. Zygotic induction was initially proposed to result from the unrepressed transcription of transfer genes until a plasmid-encoded repressor protein is produced. However, transcriptional activation of some nontransfer genes shortly after conjugation was also suggested. They include psiB of the F plasmid [21] as well as ssb, psiB, and ardA genes of ColIb-P9 [22, 23]. These genes are not essential for transfer, but some of their functions, such as SOS inhibition and antirestriction, seem linked to plasmid establishment. Interestingly, they are located in the leading region (the first segment that enters the recipient cell during conjugation), and lines of evidence suggest that the underlaying molecular mechanism is different to that of unrepressed transcription. In 1997, Masai and Arai showed in vitro that the leading region of the F plasmid contains ssDNA sequences that can form secondary structures recognizable by RNA polymerase to initiate transcription, named Frpo [24]. Similar secondary structures for possible ssDNA promoters are also predicted in ColIb-P9 [23]. Furthermore, expression of selected proteins encoded in the leading region of the F plasmid was measured by time-lapse microscopy as GFP fusions [25]. Results showed that induction of fluorescence was observed before ss to dsDNA conversion, but declined after dsDNA complementary strand synthesis. In vitro experiments suggested that these ssDNA promoters are not active in dsDNA form [24, 26], allowing prompt but transient expression of leading region genes. Recently, genomics studies of many different conjugative plasmids have shown that leading regions are enriched with antidefense genes and some of them are linked to predicted ssDNA promoters [27], further adding to the notion that timely expression through zygotic induction is crucial for plasmid establishment.

Besides transcription, ssDNA promoters are suggested to be involved in DNA replication (dsDNA complementation from ssDNA template) through primer RNA synthesis [24]. Recent results also showed that the UvrD helicase is involved in the progress of ss to dsDNA conversion [28]. Although it is appreciated that there will be intriguing DNA reactions at the early stages of plasmid establishment during conjugation [1, 2], their systematic genome-wide evaluation is still missing. The major hurdle is that, while donor and recipient cells may be easily distinguished, plasmid DNA molecules in donor and transconjugant cells contain the same raw sequence information. In microscopy approaches, the separation of donors and transconjugants can be done at the image analysis level but for genomics approaches, the separation of cell populations has to be done before DNA/RNA extraction. Fluorescence-activated cell sorting (FACS) is widely regarded as the primary tool to separate specific cell type(s) from mixtures [29, 30]. However, FACS for conjugation studies would rely upon the introduction of plasmid-encoded fluorescent proteins within each putative transcriptional unit or as protein fusions to those early coding sequences, either of which may interfere with the plasmid establishment reactions of interest.

Here, we developed the new method that achieves the selective Elimination of Donor cells for Transconjugant Analysis (ED-TA) from a conjugation mixture. In the proof-of-concept experiments, we used a clinically relevant multidrug resistant plasmid (pESBL) that encodes two extended spectrum beta-lactamase (ESBL) genes. Although wild-type (WT) pESBL is highly transmissible, Tnseq identified a superspreader mutant that shows elevated conjugation frequency [12]. Furthermore, pESBL encodes the M.EcoGIX DNA methyltansferase in the leading region that has been shown to counteract host restriction systems [12, 31]. Our analyses of pESBL gene expression in transconjugant cells suggests the specific induction of a subset of genes, divided into six operons. We demonstrate that these operons are driven by ssDNA promoters, and that the induction from ssDNA promoters is extended in ∆uvrD cells. Finally, we present how pESBL plasmid gene expression is limited in out-of-host range bacteria.

Materials and methods

Strains and plasmids

Escherichia coli outer membrane–peptidoglycan disconnected mutant (hereafter referred to as the DISCO mutant) EP272 (MG1655 ∆lpp ∆omp::kan) and Sinorhizobium meliloti reference strain 1021 were kindly provided by Drs Seung Hyun Cho and Jean-François Collet at UCLouvain, and Dr Peter Mergaert at I2BC, respectively.

To prepare DISCO donors, WT or superspreader (pESBL::Tn) pESBL [32] and F-Tn10 [33] were introduced by conjugation. For recipient E. coli, MKW278 (MG1655 ∆lacZ::cat; [12]) and its isogenic ∆uvrD mutant bEYY2756 were used. To construct bEYY2756, ∆uvrD::kan was first introduced by P1 transduction followed by removal of the FRT-flanked kan cassette by Flp recombinase expressed from pCP20 [34]. Then, a ∆lacZ::cat allele was introduced by P1 transduction.

For quantitative PCR (qPCR), we constructed plasmids to act as DNA standards that encode each intended qPCR target sequence (kan, E. coli ompA, oriT-flanking region of the plasmid, and S. meliloti rpoE). First, an E. coli ompA fragment (primers oYo1618 and oYo1624) and a pESBL fragment encompassing Hp23 and nikA (primers oYo107 and oYo1625) were amplified. These two fragments were cloned together into the KpnI-SacI site of pYB199 [35] by Gibson Assembly, resulting in pEYY515. Similarly, standard for F plasmid (pEYY549) was constructed by cloning the oriT-flanking region amplified with oYo1749 and oYo1750 along with ompA fragment. For S. meliloti, rpoE was amplified with oYo1722 and oYo1723 and cloned into the AflII site of pEYY515, resulting in pEYY542. Nucleotide sequences of oligo DNAs are shown in Supplementary Table S1.

Growth conditions

Otherwise specified, cells were grown in lysogeny broth (LB) with agitation at 180 rpm or on solid LB agar (1.5% w/v) at 37°C for E. coli and at 30°C for S. meliloti. Supplements were used in following concentrations when appropriate; ampicillin (amp) at 100 μg/ml, chloramphenicol at 25 μg/ml, kanamycin at 25 μg/ml, streptomycin at 100 μg/ml, tetracycline at 15 μg/ml, sodium citrate at 5 mm, and magnesium chloride (MgCl₂) at 1 mM.

Conjugation frequency test

Conjugation frequency was measured by conventional surface mating. In brief, overnight cultures of recipient (100 μl) and donor (10 μl) cells were mixed, washed once with LB to remove antibiotics, and resuspended in 50 μl LB. Cells were spotted on a 0.45-µm HAWP filter membrane (Merck) that was placed on an LB agar plate. After incubation for 2 h at 37°C, cells were resuspended in 1 ml LB in a 50 ml centrifuge tube, and then dilutions were spread on LB plates with appropriate antibiotics to count the colony forming units (CFUs) of donor, recipient, and transconjugant cells. For comparison, log₁₀ values were used to measure means and standard deviations. Data analysis was performed in Microsoft Excel.

Elimination of donor

For the ED-TA method, cultures (OD_600nm ~1.2) of donor (16 ml) and recipient (10 ml) cells were centrifuged and each pellet was resuspended in 200 μl LB containing 1 mM MgCl₂ and 100 µL LB, respectively. Then, cells were mixed and placed on filter membranes placed on LB agar plate containing 1 mM MgCl₂. After incubation at 37°C for appropriate duration, cells were recovered from membranes with 20 ml of 2 mM ethylenediaminetetraacetic acid (EDTA) solution with or without 0.1% sodium dodecyl sulfate (SDS). Cell suspensions were subsequently incubated at room temperature for 2 min with occasional vortexing, then sedimented at 3000 × g for 3 min at 16°C. Cells were washed once with 20 ml of 2 mM EDTA solution with or without 0.1% SDS and centrifuged at 3000 × g for 3 min at 16°C. Pellets were resuspended again in 1 ml of 2 mM EDTA solution for CFU counting and downstream preparation of RNA or DNA. In initial experiments (60 min and 30 min conjugation of superspreader pESBL to E. coli), killing and washing was performed with simple EDTA solution but else, 0.1% SDS was supplemented as it effectively prevented contamination of nucleic acids from the donor (discuss later). For control experiments, only donor cells were placed on filter membranes. After mock conjugation, cells were recovered and washed with LB instead of being processed with hypoosmotic shock.

RNA experiments

Total RNA was extracted using the Monarch Total RNA Miniprep Kit (New England Biolabs, NEB), following the manufacturer’s protocol. Quality of the purified RNA was verified by spectrophotometry (A260/A280 ratio of 1.8~2.2 and A260/230 ratio >2.0). Absence of DNA contamination was verified by RT-qPCR with the Luna Universal One-Step RT-qPCR Kit (NEB). Oligo DNAs used are shown in Supplementary Table S1. RNA samples were stored at −70°C. For RNA-seq, the integrity of purified RNA was checked on a TapeStation 4150 (Agilent Tech). Qualified samples with lowest donor contamination rate were sent to Novogene for Bacteria RNA sequencing (Illumina PE150, rRNA removal, standard service) with 8 Gb of output (or 2 Gb for control experiments).

Transcription start sites for the six operons were experimentally identified using the 5′/3′ RACE Kit (Roche Applied Science). One microgram of total RNA starting material was purified from an ED-TA experiment using a 30 min conjugation time. Operon-specific primers for complemantary DNA synthesis, amplification, and sequencing (SP1-SP3) are shown in Supplementary Table S1. Sanger sequencing of purified PCR products was performed by Eurofins Genomics and the chromatogram traces were mapped to the pESBL genome (NC_018659.1 with revised annotation shown in [12]) with SnapGene.

RNA-seq data analysis

Bowtie2 [36] was used for paired-end (PE) mapping with following reference genomes: E. coli chromosome (NC_000913.2), S. meliloti genome (NC_003047.1; NC_003037.1; NC_003078.1), F (AP001918), and pESBL. For the presentation purpose, the start site of the circular plasmids was altered so that origin of replication is situated at the end of linearized file. The resulting .bam files were used to measure the read counts of each gene with featureCounts [37], followed by differential gene expression (DGE) analysis done by DESeq2 package in SARTools [38, 39], to calculate adjusted P-value (Padj) and log₂ fold change (l2FC) of fragments per kilobase of gene per million mapped reads (FPKM). For S. meliloti samples, mapping to S. meliloti genome + pESBL was used solely for quality control of two replicates, and DGE analysis was done with mapping only to pESBL. To analyze the rearrangement of shufflons, single-end mapping with read 2 was carried out against seven shufflon fragments without pilV.

Sequence analyses

To identify ss promoters, a global prediction of secondary structures along the whole leading region was performed with a 300-nt sliding window and a 10-nt step size, using the DNA folding model from RNAfold [40]. Segments of the leading region with a folding free energy beyond a given threshold (<–75 kcal/mol) were further analyzed. Refined ssDNA secondary structure predictions of each individual candidate were obtained using the DNA folding package from the UNAfold suite [41]. Identification of −35 and −10 boxes in these hairpins was performed separately on both strands of the stem loop using PromoterHunter [42]. Comparison of F and pESBL genomes were performed with Mauve v2.4 [43].

Quantitative PCR

For DNA qPCR, gDNA was prepared with a DNA miniprep Kit (Sigma–Aldrich) following the manufacturer’s protocol. qPCR and RT-qPCR reactions were done with Luna Universal qPCR master mix and One-Step RT-qPCR Kits, respectively (NEB), following the manufacturer’s protocols. Oligo DNAs used are shown in Supplementary Table S1. All qPCR reactions were done in three technical replicates on a Bio-Rad CFX384 Touch amplifier (Bio-Rad) and results were exported to Bio-Rad CFX Manager (Bio-Rad) to determine the C_q values. For DNA qPCR, we included serial dilutions of control plasmid DNA (3 × 10⁸ to 3 × 10⁴ molecules per reaction). Generation of standard curves and data quantification were performed with Microsoft Excel.

Presentation

Graphics and figures were prepared with MatLab, SnapGene, Inkscape, and Adobe Illustrator.

Results

Elimination of donor cells for transconjugant analysis

For separating cell populations, an alternative method to FACS is the elimination of specific cell type(s). However, this requires not only the specific killing of donor cells but also a rapid processing time to capture transient cellular processes. Deghelt et al. recently reported that E. coli DISCO mutants exhibit a severe survival defect against hypoosmotic shocks [44]. Importantly, release of periplasmic and cytoplasmic materials and irreversible cell death happen <2 min after the imposition of hypoosmotic conditions [44]. Divalent cations such as Mg²⁺ in solution protected the viability of the DISCO strains against hypoosmotic shock, while incubation with a cation chelator such as EDTA further increased their osmosensitivity [44]. For these reasons, the DISCO mutants were appealing as donor bacteria for trials of transconjugant gene expression. While the DISCO mutant showed a slight growth defect, it exhibited donor capacity for plasmid transfer comparable to the WT E. coli (Supplementary Fig. S1).

In conventional conjugation assays on solid media, a donor-recipient mixture is often deposited onto a filter membrane placed on agar media, rather than directly spotted on to the plate. This allows easy and total recovery of cells after incubation by transferring the membrane into a collection tube then flushing the cells off with media such as LB. In the ED-TA method, after conjugation, specific killing was executed by performing the membrane washing step with 2 mM EDTA solution. During experimental manipulation, such as incubation and two washing/sedimentation steps, conjugation mixtures were exposed to EDTA solution for <10 min. This resulted in >99.999% killing of the DISCO donor cells, while CFU counts of recipient and transconjugant cells were not affected (Fig. 1A). However, the fraction of WT pESBL transconjugant cells was only ~1% after 60 min of conjugation. Even though viability of donor bacteria was reduced by several orders of magnitude, gDNA from the initial experiments displayed high levels of donor gDNA contaminations, presumably the donor gDNA was not well released during the washing step and cosedimented with cell debris. The degree of contamination varied from experiment to experiment but was not lower than 1% (Supplementary Table S2). With low transfer rate, it was not feasible to carry out downstream genomics experiments. We switched to using the superspreader mutant of pESBL [12, 32], which resulted in transconjugant cells cell percentages in the population of ~25% and ~20% for 60 min and 30 min conjugation, respectively (Fig. 1B and C). In the course of the study, we found that addition of 0.1% SDS in EDTA solution efficiently reduced donor DNA contamination (Supplementary Table S2), allowing genomics investigation of plasmid reactions at an earlier time point of 15 min with an ~3% transfer rate (Fig. 1D).

Figure 1.

Elimination of donor cells by the ED-TA method. CFUs of donor (D), recipient (R) and transconjugant (T) cells before (input) and after conjugation either recovered by LB (−) or 2 mM EDTA solution (+). R* was measured by Cm-R colonies so that it does not consider whether it received the plasmid or not. Average and standard deviation of at least three independent experiments were shown, along with individual results as closed dots. When no viable cells were recovered, limit of detection was shown as an open dot. (A) Transfer of WT pESBL in 60 min conjugation. Transfer of superspreader pESBL in 60 min (B), 30 min (C), or 15 min (D) of conjugation.

Transcriptome of plasmid genes during early stages of conjugation

We carried out RNA-seq to investigate transcriptional profiles of plasmid genes following the conjugation of pESBL. To compare, RNA from donor cells was purified from mock conjugations, and two biological replicates were prepared for each condition (Supplementary Fig. S2). DGE analysis showed that, after 15 min of conjugation, a set of the genes located in the leading region exhibited very strong (max. ~500 fold) induction while many other genes including transfer genes were significantly repressed (Fig. 2A and B). Indeed, induction of some of the leading region genes in pESBL and another IncI1 plasmid, such as ssb, psiB, ardA, and M.EcoGIX, was previously reported [22, 31, 45]. However, our results clearly suggested that not all leading region genes were induced. Upregulated genes were present in six operons (Fig. 2). IncI1 and IncF plasmids have similar gene organization in a part of the leading region ([46], Fig. 2C), and in the F plasmid it was shown that ssDNA promoters (Frpo) are involved in the early expression of the leading region genes [24, 25]. Indeed, ED-TA RNA-seq experiments performed on 60 min conjugation with the F plasmid enabled detection of transcript enrichment in recipient cells for genes regulated by Frpo1 and Frpo2 (Fig. 2C and Supplementary Fig. S4), among which transconjugant-specific expression of Ssb, YfjA, YfjB, PsiB, and YgeA was previously shown by microscopic approach [25]. Similarly to pESBL, transconjugants showed increased transcript levels for two operons located upstream of Frpo2, which encode homologs of KlcA and M.EcoGIX (Fig. 2C). However, promoters controlling the expression of these two operons in pESBL and F are not yet known, prompting us to look for novel classes of ssDNA promoters located upstream of these operons.

Figure 2.

Plasmid gene expression during conjugation. (A) Linear map of pESBL. CDS are indicated in red or blue depending on their coding strand. Several important loci are labeled above. Direction and order of DNA transfer is also indicated (color gradient arrow). ssDNA promoters identified are indicated below (red icons). (B) L2FC of gene expression in E. coli (uvrD⁺) recipient cells with indicated conjugation time, compared to the plasmid gene expression in the donor. Color of bars corresponds to the padj value. (C) Comparison of leading regions between F and pESBL. Red circles on the right indicates oriT. Homologous proteins are connected by dotted line. For F, ssDNA promoters and genes induced at an early stage of conjugation reported in [25] are indicated in green. ssDNA promoters homologous to those we have newly identified in pESBL are shown in magenta. For pESBL, induced genes (>4 fold) are labeled with color corresponding to panel (B) (15 min). ssDNA promoters are also indicated. (D) Predicted secondary structure of the ssDNA (above) and trace file of 5′ RACE sequencing (bottom) for M.EcoGIX promoter. ∆G (kcal/mol) value obtained by UNAfold are shown. Nucleotides are color-coded as follows: A, green; C, blue; G, black; T, red. Results for other ssDNA promoter loci are shown in Supplementary Fig. S3.

To identify putative ssDNA promoters located in the leading region of pESBL, we carried out homology-independent mining of hairpins in this region. Interestingly, all six intergenic segments upstream to the possible operons showed a high tendency to form secondary structures. They were all predicted to fold as large stem–loop structures, with possible reconstitution of −35 and −10 boxes. Three of these hairpins correspond to previously identified single-stranded initiation (ssi) sites from ColIB-P9 (ssi1-3), which display high sequence similarity with Frpo1-2 from plasmid F [23] (Fig. 2C and Supplementary Fig. S3). However, the remaining three hairpins displayed no sequence homology, suggesting that they function as novel classes of ssDNA promoters. The two hairpins located in 5′ of klcA and ardA display high pairwise sequence and structure similarity (Supplementary Fig. S3B), while the hairpin located in 5′ of M.EcoGIX is different from all other identified hairpins. We therefore propose to classify these putative ssDNA promoters in three different classes with ssi1-3 and Frpos as class 1, promoters driving ardA and klcA transcription as class 2, and the promoter driving M.EcoGIX as class 3. To ascertain the transcription start sites of these ssDNA promoters, we carried out 5′ RACE using RNA samples obtained by the ED-TA method. Transcription start sites (+1) were identified for all six promoters and matched their in silico predictions (Fig. 2D and Supplementary Fig. S3).

Extensive induction of leading region genes was still evident in both 30 and 60 min conjugation but the induction level was gradually decreased with longer conjugation time (Fig. 2B). Conversely, suppression of other genes was generally lifted at 30 min of conjugation. These results are consistent with the idea that ssDNA promoters are not active in dsDNA form, while conventional promoters require dsDNA form.

In addition to transcription, ssDNA promoters may also contribute to the ssDNA-to-dsDNA transition by providing RNA primers for DNA synthesis [24]. Previously, we showed that conversion of ss- to ds-plasmid DNA is perturbed in recipient cells deficient for the UvrD helicase [28]. We tested our RNA-seq experiments with a DISCO donor and a ∆uvrD recipient (Supplementary Fig. S5C and D) followed by ED-TA treatment. Similar to the WT (uvrD⁺) recipient, strong and specific induction of leading region genes was detected (Fig. 3B). While the expression level of almost all genes was comparable in a 30-min conjugation experiment, the induction of leading region genes was much more profound after a 60-min conjugation to a ∆uvrD recipient (Fig. 3B and C). These results are consistent with the prior observations that ssDNA promoters are active in ss but not in ds form, and such ssDNA persists for a longer period in ∆uvrD cells.

Figure 3.

Plasmid gene expression during perturbed conjugation. Comparison of plasmid gene expression similarly presented in Fig. 2B but with (B) ∆uvrD E. coli as recipient and (D) S. meliloti as recipient. (C) Gene expression compared between E. coli ∆uvrD and uvrD⁺ recipients.

Plasmid gene expression during abortive conjugation

As pESBL belongs to the IncI1 incompatibility group, it has a plasmid host range limited to Enterobacteria [47]. Indeed, no amp-resistant transconjugant colonies of pESBL were obtained after mating with an alpha proteobacterium S. meliloti (Supplementary Fig. S5). However, it is well known that conjugation machineries can transfer DNA into remotely related organisms, even across the biological kingdoms [48]. Consistent with this, pESBL DNA was readily detected by qPCR from S. meliloti transconjugant gDNA following conjugation between a DISCO donor and S. meliloti with ED-TA treatment (Supplementary Table S2). This result suggests that pESBL ssDNA can be transferred into S. meliloti cells but the plasmid cannot be established.

We also purified total RNA from S. meliloti cells and carried out RNA-seq to investigate plasmid gene expression during these abortive conjugation reactions. DGE analysis showed that genes that are repressed in an E. coli transconjugant are also repressed at similar levels in S. meliloti. However, only modest induction of leading region genes was observed in the out-of-host range bacterium during both 30 and 60 min conjugation experiments (Fig. 3D).

Discussion

In this study, we invented a novel method for the expeditious elimination of donor cells from bacterial conjugation mixtures. In general, the selective killing of a bacterial subpopulation can be performed in different ways: through death by induction of toxic genes or lytic phages, or by environmental chemicals such as antibiotics or detergents. However, each of these methods requires a certain time to induce cell death, for example limited by protein production. In contrast, DISCO mutant bacteria are killed by hypoosmotic shock in only a few minutes at extremely high efficiency [44]. The immediate execution of our ED-TA methodology provides an unprecedented opportunity to investigate the rapid DNA transactions occurring at early stages of plasmid establishment in transconjugant cells, such as transcription from ssDNA promoters. Moreover, the ED-TA method was successfully applied for RNA-seq of two conjugative plasmids, pESBL and F, suggesting the robust applicability to different plasmids.

Our RNA-seq studies of pESBL strongly suggested the presence of six operons that are highly induced at the early stages of conjugation. ssDNA promoters were predicted for those operons and 5′ RACE confirmed their transcription start sites in vivo. Among the promoters, previously suggested class 1 promoters exhibited the highest transcription rates, yet newly identified class 2 and 3 promoters can also promote significant gene expression. Class 2 and 3 ssDNA promoters were found to be conserved in the E. coli F plasmid (Fig. 2C), suggesting that these novel promoters support transcription from ssDNA in both IncI1 and IncF plasmids. Furthermore, in ∆uvrD transconjugants in which conversion of ss- to ds-plasmid DNA is perturbed, the level of induction was comparable to uvrD⁺ after 30 min conjugation, but at least 4 times more after 60 min conjugation. Altogether, we provided strong evidence that ssDNA promoters are active in the transconjugant cells at early stages of plasmid establishment. Some of the leading region genes have been previously shown to function against host defense mechanisms [27, 31]. However, there are still many genes whose biological functions remain to be elucidated.

An additional advance is that we successfully applied our ED-TA-based investigations to conjugation reactions of pESBL into an out-of-host range bacterium. Although other factors such as plasmid stability and fitness cost have been implicated, plasmid host range is primary defined by the capability for replication of which the functionality of plasmid-encoded replication initiator proteins and their interaction with host factors (e.g. replicative helicases) play the most important role [49]. In contrast, conjugation machineries have a significantly wider range of activity for the DNA transfer itself. However, the fate and reactions of a plasmid DNA transferred to an out-of-host range bacteria have been under studied due to a lack of appropriate methods. In this work, we found that the induction of leading region genes was very much diminished in S. meliloti (Fig. 3D), suggesting that gene expression was perturbed. It is likely that the ssDNA promoters have not evolved for wide recognition by the sigma factors of phylogenetically distant bacteria. Nonetheless, expression of other genes, presumably thorough conventional dsDNA promoters, were readily detected (no genes were greater than 10-fold decreased). This result suggests that plasmid ss to dsDNA conversion had likely occurred within in the out-of-host range S. meliloti, but downstream establishment and/or maintenance properties of pESBL are compromised in this host.

IncI plasmids encode diversity-generating DNA recombination mechanism called shufflon. It consists of the pilV gene, shufflon cassettes (seven in case of pESBL; A, A’, B, B’, C, C’, D’), and the Rci recombinase ([50–52]; Supplementary Fig. S6). Recombination between cassettes creates different C-terminal sequences of PilV, a minor pilin protein, which exhibit different affinities to different recipient bacteria and are important for successful conjugation in liquid. In growing cells (including the donor), active shuffling generates a random distribution of fragments so that the transcription level among shufflon fragments is expected to be even. In the RNA-seq analysis, skewed expression of shufflon fragments was observed in many experiments, especially in 15 min conjugation or S. meliloti transconjugant cells (Fig. 2B and D). However, it turned out to be an artifact of the analysis. Essentially, the reference sequence does not reflect the outcome of recombination events so that PE mapping is not adequate, particularly for the shufflon fragments annotated on the complementary strand. Indeed, reanalysis using single-end mapping against shufflon fragments suggested rather even distribution of 7 segments in donor cell, and no specific enrichment of particular fragment was observed in any of the transconjugant samples (Supplementary Fig. S6). Since we carried out conjugation on surface, it is possible that enrichment of fragment(s) can be observed in liquid conjugation conditions.

Expression of the repZ gene, encoding the replication initiator of pESBL, was found to be downregulated in all transconjugant samples compared to donor bacteria (Fig. 2). This result was unexpected because fluorescence microscopy examinations using the ParB/parS system exhibited multiple ParB foci in transconjugant cells, indicating multiple ds DNA copies of the plasmid [28, 53]. Furthermore, detailed time-lapse analyses of the F plasmid suggested its accelerated replication in transconjugant cells [25]. Possible scenarios include following: (i) the minimal expression level of initiator is still sufficient for pESBL replication and other origin of replication mechanisms, such as initiator-iteron binding, plays a more important role, and (ii) plasmids use different strategies for their initial amplification, which can be linked to rolling circle-type replication and plasmid circularization.

Notably, the other genes repressed at the early stages of conjugation were the antibiotic ESBL genes (bla-TEM and bla-CTX-M-3). This result suggests that transfer (or abortive transfer) of pESBL does not grant immediate resistance to the beta-lactams. To survive in a conjugation event in an environment where the antibiotics are present, transconjugant cells have to sustain a period without their protection presumably via AcrAB-TolC multidrug efflux pump [33].

In conclusion, our ED-TA method described here has allowed the first-time investigation of a plasmid transcriptome following conjugation. However, the genome-wide population level approach is disadvantaged in temporal resolution compared to the microscopy-based studies. Particularly, from classic gene mapping with Hfr strains, DNA transfer speed was estimated at ~46 kb/min, such that transfer of ~100 kb plasmid will complete in under a few minutes. Therefore, it would be difficult to monitor ssDNA entry by our ED-TA method in a time-resolved manner. Yet, our methodology should be applicable for investigations of plasmid-wide post-transcriptional regulation, translation, and other DNA actions such as ss- to ds-DNA conversion and circularization. Furthermore, to understand how host defense and plasmid antidefense mechanisms compete, it is important to study abortive transfer events. Therefore, further research taking advantage of the ED-TA method will shed light on important molecular mechanisms of plasmid establishment after conjugation.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by French National Research Agency (grant number ANR-22-CE12-0032 to Y.Y. and C.L.) and by the Foundation for Medical Research (grant number EQU202103012587 to C.L.). M.W. was supported by the China Scholarship Council. Funding to pay the Open Access publication charges for this article was provided by French National Research Agency ANR-22-CE12-0032.

Data availability

RNA-seq data are available at GEO repository with accession number GSE298248 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE298248).