Transcriptional Organization of the Stability Module of Broad-Host-Range Plasmid RA3, from the IncU Group

The efficiently disseminating conjugative or mobilizable BHR plasmids play key roles in the horizontal spread of genetic information between closely related and phylogenetically distant species, which can be harmful from the medical, veterinary, or industrial point of view. Understanding the mechanisms determining the plasmid’s ability to function in diverse hosts is essential to help limit the spread of undesirable plasmid-encoded traits, e.g., antibiotic resistance. The range of a plasmid’s promiscuity depends on the adaptations of its transfer, replication, and stability functions to the various hosts. IncU plasmids, with the archetype plasmid RA3, are considered to constitute a reservoir of antibiotic resistance genes in aquatic environments; however, the molecular mechanisms determining their adaptability to a broad range of hosts are rather poorly characterized. Here, we present the transcriptional organization of the stability module and show that the gene transcript dosage effect is an important determinant of the stable maintenance of RA3 in different hosts.

tionally transcribed, highly compacted module might influence the expression of the genes in various hosts.
Here, we compared the gene expression patterns of the intact RA3 stability module and its deletion derivatives. Using this approach for the whole RA3 plasmid and also its stability module cloned into a heterologous replicon, we found that this module is organized as a long multicistronic operon with numerous internal promoters and a few terminators/attenuators modulating the expression of downstream genes. The RNA polymerase (RNAP) read-through and an impact of the multiple upstream promoters on downstream gene expression were also detected in hosts other than E. coli, although they varied in different species. The transcriptional studies on RA3 stability module deletion derivatives combined with stability assays revealed the important role of encoded proteins in plasmid maintenance in various hosts.

RESULTS
Transcriptional read-through from orf02p to orf11. The unidirectional transcription and highly compacted arrangement of the stability module suggested the possibility of reading through from the upstream transcriptional signals into the down- The replication module (brown) is comprised of the repA-repB operon surrounded by long direct repeats DR1 and DR2. Genes of the stability module (gray) show homology to IncP-1 plasmids. The conjugative transfer module (green) resembles one found in PromA plasmids. Most accessory genes (blue), so called plasmid "genetic" load, belong to the class I integron (24). Genes res (invertase/ recombinase) and mpR (putative zinc metallopeptidase) of unassigned plasmid functions are labeled in white (24). Arrows indicate the direction of transcription. (B) Close-up of the stability module. Thin black arrows indicate previously identified promoters (24,28,29). Regulatory circuits are shown as lines connecting the regulatory genes with their target sequences. KorB binds also to the parS centromere-like sequence, a cis-acting site of the partition complex. stream units. To verify this assumption, total RNA was isolated from the E. coli DH5␣(RA3) strain and subjected to reverse transcription followed by PCR (RT-PCR). Three different primers were used in the RT reaction; these were complementary to the 3= end of the klcA gene (primer 8), the 3= end of the kfrC gene (primer 14), and the orf11 gene (primer 29). The cDNAs obtained in each reaction were then amplified with pairs of primers corresponding to the coding and intergenic parts of the analyzed region ( Fig. 2A). PCR products synthesized on a cDNA template started from primer 8 encompassed orf02 and klcA (Fig. 2B), suggesting possibility of RNAP initiating transcription at orf02p and reading through the intergenic region orf02-klcA. PCR products obtained on cDNAs started from primer 14 indicated the reading through from orf02p to kfrC (Fig.  2C), whereas the results of PCRs on the cDNA template obtained from primer 29 confirmed that mRNA initiated at orf02p might extend through the whole stability region up to orf11 (Fig. 2D). Notably, no PCR products were obtained in any set of reactions with primers encompassing orf02p (primers 1/2), which excludes the transcription from the replication module ( Fig. 1 and 2A) into the stability region. For each set of PCRs, an adequate pair of primers downstream of the start site for the given cDNA was used: 13/14 to amplify kfrC for the shortest cDNA, 27/28 to amplify korB, and 30/31 to amplify the parS region for two longer cDNAs ( Fig. 2A). For all three cDNAs, the products of such control PCRs were not detected ( Fig. 2B, C, and D). Two additional control sets of PCRs were conducted, with RNA as a template (negative control) to verify lack of DNA contamination in RNA samples (Fig. 2E) and with RA3 DNA as a template to demonstrate the efficiency of reaction with all the primers pairs (Fig. 2F).
The RT-PCR results indicated the presence of long transcripts initiated at orf02p and continuing toward orf11, hence indicating cotranscription of the whole maintenance module. The fact that orf02, klcA, korC, kfrA, and korA are preceded by functional promoters (24,28,29) suggested the possibility of a polarized transcript dosage, with progressively more multiple transcript variants for the genes further downstream in the module. To check if indeed transcripts of various lengths exist for a given ORF, the 5= rapid amplification of cDNA ends (5=RACE) procedure (35) was applied. Total RNA isolated from E. coli DH5␣(RA3) was reverse transcribed with primers 8, 14, and 29 as described above. The cDNAs were 3= tailed with a stretch of dCTPs and used as templates for PCRs with an abridged anchor primer (AAP) complementary to the dC tail paired with appropriate nested primers 32, 33, and 26, as shown in Fig. 2G. Parts of these RT-PCR mixtures were separated on agarose gels to visualize products corresponding to the cDNAs synthesized on the transcripts that were presumably of various lengths (Fig. 2H, I, and J). The remaining parts were diluted and used as templates for PCRs with pairs of primers specific to the ORFs located upstream of the sequences complementary to primers 32, 33, and 26 ( Fig. 2H.1, I.1, and J.1). The best results were obtained for the potential cDNA mixture synthesized with primer 8 complementary to the 3= end of klcA. Two products were visualized in the first set of PCRs with primers AAP and 32, confirming the presence of two transcripts for klcA starting at orf02p and klcAp (Fig. 2H), which was further verified by the subsequent round of PCRs with primers specific to orf02 and klcA, respectively ( Fig. 2H.1). When the cDNAs obtained with the use of primer 14, complementary to the 3= end of the kfrC gene, were used as a template, two products were detected after the first round of PCR with AAP and the nested primer 33 (Fig. 2I). The intense band corresponded to the product synthesized on cDNA with the 3= end determined by the transcription start site (TSS) of korC. Another, much weaker band (inset in Fig. 2I) presumably corresponded to cDNA with the 3= end determined by the TSS of klcA. Despite the fact that no product was visible for the cDNA extended up to orf02p, the next round of PCRs led to amplification of all four ORFs from kfrC up to orf02, although with various efficiencies (inset in Fig. 2I.1). The most ambiguous were the results for cDNA(s) obtained with primer 29, which annealed in the region of orf11. The PCRs with AAP and the "nested" primer 26 unexpectedly showed the main three products with sizes between 800 and 1,100 bp (Fig. 2J). Whereas a fragment of ca. 1,100 bp might correspond to cDNA ending at the TSS of korA, the two smaller products suggested an additional TSS(s) upstream of incC. The presence of these putative promoter sequences has been further analyzed (see below). A product of ca. 2,200 bp, corresponding to cDNA ending at the TSS of kfrA, was also visible after intensity enhancement (inset in Fig. 2J). The next round of PCRs using the specific pairs of primers for ORFs upstream of korB led to the amplification of fragments corresponding to ORFs from incC to orf02 (except for klcA), although with the efficiency of the reactions clearly inverse to the distance from the 3= end of the stability module ( Fig. 2J.1). Although the 5=RACE experiments did not directly and clearly show the whole spectrum of transcripts for particular ORFs, they supported the hypothesis of a gradient transcript dosage along the stability module (Fig. 2K). The numerous small PCR products obtained with AAP and "nested primers" likely resulted from nonspecific annealing of the AAP to the C stretches in the GC-rich sequences of the kfrC, kfrA, and incC genes (24).
Analysis of transcription termination signals. The expression of multigenic modules is regulated not only at their promoters but also by transcription terminators. In silico inspection of the RA3 replication and stabilization regions uncovered two putative unidirectional Rho-independent transcription terminators (GC-rich palindromic sequence followed by a run of Ts) after the kfrC and orf11 genes in the stabilization module and a third one after the repB gene separating the replication and stabilization modules ( Fig. 3A and B). Also, the intergenic regions klcA-orf04 and kfrA-korA contain GC-rich palindromic sequences (however, without a long run of Ts) which could pause/modulate the progress of RNAP ( Fig. 3A and B). The five relevant intergenic regions were cloned individually into pGBT70 (36) between the strong trfAp-1 RK2 promoter and the xylE reporter gene to verify their putative terminator/modulator action. This plasmid had been successfully used to characterize transcription terminator sequences before (37). The plasmid constructs were introduced into E. coli C600K, and the activity of catechol 2,3-dioxygenase, encoded by xylE, was assayed. The regions downstream of repB, kfrC, and orf11 noticeably hampered the transcription initiated at trfAp-1, while the sequence downstream of klcA did not (Fig. 3C). The result for the region downstream of the kfrA gene was ambiguous, showing a 30% decrease in the XylE activity compared to that for the unmodified pGBT70 plasmid. This decrease may suggest E. coli RNAP pausing at this sequence, but that needs an independent confirmation. Thus, we experimentally demonstrated the presence of two efficient Rhoindependent transcription terminators in the RA3 stability module: one in the middle of the module downstream of the kfrC gene and the other at the very end of the module, following orf11.
Transcriptional analysis of the synthetic RA3 stability module and its mutated variants: effect of mutations on plasmid stability in various hosts. To assess the roles of the individual promoters in the expression of downstream genes/operons in the RA3 stability module, a set of deletion variants, deprived of one or a combination of several promoters and/or particular genes, was constructed (see the supplemental material). The obtained library of the module variants, cloned into an unstable BHR plasmid, allowed study of the roles of genes of interest in the maintenance of the plasmid in various hosts.
For these studies, the wild-type (WT) module and its Δ(orf02p-orf02), Δ(klcAp-klcA), Δ(kfrAp-kfrA), ΔkorAp, and Δ(orf02p-kfrA) (the partition operon on its own) variants were chosen (Fig. 4A). The rationale behind the deleting of the promoter regions with the adjacent ORFs in the case of the monocistronic and strongly regulated klcA and kfrA operons was to avoid altered expression of these presumably important ORFs from upstream promoters. As a precaution, no genetic manipulations were undertaken in the korC operon, since KorC is a potent transcriptional regulator of orf02p and klcAp (29). The constructed modules were cloned into a single-copy, extremely unstable pESB36 vector, a derivative of pABB32 (38). pESB36 is based on the RK2 minireplicon (pRK415), is mobilizable by the RK2 conjugative system owing to the presence of oriT RK2 , and carries a convenient system for detection of the plasmid presence in colonies (repAp RA3 -lacZ transcriptional fusion). pESB36, pESB36.35 (WT stability module inserted), and its deletion derivatives (Fig. 4A) were introduced into two representative Gammaproteobacteria species (E. coli and Pseudomonas putida), two Alphaproteobacteria species (Agrobacterium tumefaciens and Paracoccus aminovorans), and one Betaproteobacteria species (Cupriavidus necator). Transformants/transconjugants were analyzed with respect to stability module gene expression and stable maintenance of the pESB36.35 derivatives during growth without selection. The plasmid segregation assay demonstrated that the pESB36 loss rate (LR) was variable in the hosts tested and varied from 2% per generation in P. putida to 18% per generation in E. coli and to more than 25% per generation in P. aminovorans (Table 1). Such differences in the vector stability could be due to the variations in the plasmid copy number (PCN) caused by differential gene expression of the miniRK2 replicon in the analyzed species. Estimation of the pESB36 copy number per chromosome in cells from stationary-phase cultures grown under selection confirmed this conjecture. The PCN for two hosts, P. aminovorans and E. coli, was only 0.12 to 0.15, whereas that for P. putida was 2.9 ( Table 2).
The presence of the RA3 stability module significantly increased the persistence of the plasmid (pESB36. 35) in all the species tested, again to different extents. The plasmid was very stably maintained in P. aminovorans, P. putida, and A. tumefaciens, with a loss rate per generation of 0.4 to 0.5% (Table 1), and slightly less so in C. necator (LR, 1.1%). The highest plasmid loss rate, 3.5%, was found for E. coli. However, the ratio of the loss rates for the empty vector and pESB36.35 in a given strain, the so-called stability index (SI), was quite similar in all the hosts, with the exception of P. aminovorans, where it exceeded 60 (Table 1).
(i) Expression of the stability module and its deletion variants in E. coli. The E. coli EC1250 (Δlac) strain was transformed with pESB36, pESB36.35, and its derivatives, transformants were grown on rich medium with selection, and total RNA was isolated. cDNA was synthesized on the RNA with the use of a mix of random hexamer primers, and then quantitative real-time PCR (qPCR) was performed with pairs of primers specific for each ORF of the stability module, with the exception of the short orf04 overlapping korCp. The results were normalized to the amount of mRNA for the chromosomal marker cysG (39).
The expression of blocks of genes in the WT stability module varied significantly, confirming the functionality of the five previously identified promoters in the context of the whole module (Fig. 5A). The lowest levels of transcripts were found for orf02, the first ORF in the module, and for kfrA, both presumably encoded by the tightly regulated monocistronic operons (24). In turn, klcA expression was 10-fold higher than that of orf02, despite the expected strong repression of klcAp by KorC (29). The korC, kfrC, and korA genes showed an intermediate level of expression, and unexpectedly, the highest levels of transcripts were found for incC, korB, and orf11.The organization of the korA-incC-korB-orf11 region suggested a single operon structure (with closely packed or overlapped ORFs) whose expression was dependent on the strong but autorepressed korA promoter (24,28). The observation here of decoupling of korA expression and three downstream genes raised the possibility of the existence of an internal promoter(s) in this operon. The presence of the promoters driving a korAp-independent expression of incC-korB-orf11 (at least in E. coli) was verified by cloning of a 586-bp DNA fragment (Fig. 4B) encompassing the korA coding sequence and the 5= end of incC into the promoter-probe vector pPT01 (40) upstream of the xylE cassette (pESB13.76). Determination of the XylE activity in the extracts from exponentially growing E. coli C600K(pESB13.76) cells revealed a weak transcriptional activity of the cloned fragment [7.2 mU, versus 0.5 mU in control E. coli C600K(pPT01) cells]. A similar XylE activity (6.5 mU) was detected in extracts from E. coli C600K(pESB13.77) and C600K(pESB13.79) when the shortened fragments were cloned upstream of xylE, pointing out the localization of the incC promoter(s) in the 3= end of korA (Fig. 4B). Importantly, the 5=RACE experiments also suggested the possibility of additional TSSs within korA (Fig. 2J). The products detected in the 5=RACE experiments were isolated and sequenced. The 5=  ends of these fragments localized two putative TSSs in korA, one corresponding to the 3= end of korA and another 169 bp upstream, as marked in Fig. 4B together with corresponding promoter motifs. Importantly, these weak constitutive internal transcriptional signals seem to significantly increase the amounts of mRNAs for partitioning genes in comparison to korA mRNA (Fig. 5A). Such accumulation of shorter transcripts may result from the higher stability of these mRNAs than those initiated at korAp, as differential mRNA decay has been observed in other prokaryotic multicistronic operons (41,42). The transcription profiling of the stability module variants lacking single or several promoters confirmed the participation of RNAP read-through in the establishing of this region's expression pattern (Fig. 5B to F). It was most clearly seen when the expression of the korA-incC-korB-orf11 genes was compared between the intact WT stability module and the partition operon on its own (Δorf02p-kfrA variant, pESB36.29) (Fig. 5B). The transcripts for the partition operon genes were almost four times less abundant in the construct lacking all the DNA sequence upstream of korAp than in the WT stability module. Similarly, deletion of orf02p-orf02 at least halved the expression levels of the majority of downstream genes compared to the WT module (Fig. 5D). Deletion of the klcAp-klcA region decreased the expression of downstream genes 2-to 3-fold but, in parallel, increased the expression of orf02 3-fold (Fig. 5E). The apparent derepression of orf02p might be due to the lower level of expression of korC and korB, encoding orf02p repressors (29). Another plausible explanation of the observed "induction" might be a relief from a negative interference between the orf02p and klcAp transcriptional and regulatory signals, since these two promoter regions arose by duplication (24). The unexpected lack of transmission of the enhanced orf02p activity to the genes downstream of orf02 could be a result of the genetic manipulations leading to the deletion of the klcAp-klcA region.
In the absence of korAp, no change in expression of korA was detected and there was only a slight decrease in the expression of the partitioning genes (Fig. 5G), confirming significant participation of RNAP read-through from the upstream promoters into korA and a minor role of korAp in driving expression of downstream loci. The most unexpected result was the strong effect of the kfrAp-kfrA deletion on the expression of the whole stability module (Fig. 5F). Genes downstream of kfrA were even expressed 12-fold less abundantly despite the low level of transcriptional activity of kfrAp in the intact module. Intriguingly, the upstream genes also were strongly downregulated. Further studies should elucidate the consequences imposed on plasmid functions by the lack of kfrAp-kfrA.
E. coli EC1250 (Δlac) transformants carrying pESB36.35 variants were also tested for plasmid retention over 60 generations of growth in rich medium without selection. Analysis of plasmid stability showed that all deletion variants except pESB36.38 ΔkorAp were unstable in E. coli ( Fig. 5C and H to K). Since only the ΔkorAp derivative demonstrated hardly any effect on expression of the partition operon, it may be concluded that any manipulation leading to the decreased transcription of incC-korB-orf11 led to an increase in the rate of loss of these plasmids ( Fig. 5 and Table 1).
(ii) Expression of the stability module in hosts other than E. coli. The RA3 plasmid has a very wide range of hosts belonging to Alpha-, Beta-, and Gammaproteobacteria (24), and hence it was important to establish whether the transcriptional signals characterized in E. coli function in a similar way in the other hosts and whether the RNA polymerases of other bacteria were also capable of reading through the termination/attenuation signals in the stability module. The mRNA levels of the RA3 genes of interest were measured with respect to those of the recommended reference gene for each transconjugant, and the results are presented in Fig. 6, normalized to orf02 taken as 1. The analysis of the WT stability module expression (pESB36.35) by RT-qPCR demonstrated that in all transconjugants the levels of incC and korB transcription were the highest (or among the highest, as in C. necator), similarly to the case in E. coli, but the expression of the genes preceding the partition operon varied significantly depending on the host (Fig. 6A). The expression of orf02 and klcA seemed to be Applied and Environmental Microbiology correlated and severalfold lower than that of korC and kfrC in P. aminovorans, A. tumefaciens, and P. putida, whereas in C. necator the first four genes were expressed at comparable levels. Expression of kfrA was like the expression of orf02 and klcA in A. tumefaciens and P. putida, lower in C. necator, and slightly higher in P. aminovorans. These results indicated substantial modulations of the abilities of the host transcriptional machinery to recognize different initiation signals (e.g., more efficient expression from korCp in all species other than E. coli) and presumably to respond to regulators such as KorC (orf02p and klcAp) or KfrA (kfrAp).
To check whether read-through takes place in hosts other than E. coli, the expression of genes in the Δ(orf02p-orf02), Δ(klcAp-klcA), and Δ(kfrAp-kfrA) variants of the stability module was analyzed by RT-qPCR in the transconjugants of four hosts carrying pESB36.40 (Fig. 6B), pESB36.41 (Fig. 6C), and pESB36.44 (Fig. 6D), respectively. Plasmid retention in transconjugants was also monitored for 60 generations of growth without selection, and the results of the stability assays are shown in insets in Fig. 6. Additionally, the calculated plasmid loss rates and stability indexes (if feasible to calculate) are included in Table 1.
Expression of the truncated stability modules varied between the strains. Notably, the deletion of orf02p-orf02 did not have as strong a negative effect on the level of transcripts of the downstream genes as in E. coli ( Fig. 5D and 6B). A decrease in transcription, but only of korC and kfrC, was observed in P. aminovorans (with a simultaneous increase in the transcription of the partition operon), and no change in expression of the stability module was observed in A. tumefaciens (Fig. 6B, two upper panels). In contrast, in C. necator the orf02p-orf02 deletion had a positive effect on the expression of almost all remaining genes (with the highest effect on the partition operon), whereas an increase in korC, kfrC, kfrA, and korA expression (but not that of incC-korB-orf11) was detected in P. putida (Fig. 6B, two bottom panels). This may suggest that RNAP read-through from orf02p encompasses only klcAp and korCp in P. aminovorans, plays no role in A. tumefaciens, and negatively interferes with the expression of downstream operons in C. necator and P. putida. Expression of the partition operon seems to be uncoupled to some extent from other parts of the module in these four species. Notably, deletion of the orf02p-orf02 fragment had a large impact on pESB36.40 stability in the analyzed hosts ( Fig. 6B and Table 1). Similar to the case for E. coli (Fig. 5D), the construct was unstable in A. tumefaciens, C. necator, and P. putida (loss rate close to that of pESB36 vector). In E. coli the pESB36.40 loss might have resulted from the lower expression of the partition operon (or of other genes in the module) ( Fig. 5D and H). In species in which the remaining genes in the stability module were expressed at the same level as in the wild-type fragment (A. tumefaciens) or even higher (C. necator and P. putida), the instability was likely caused by lack of the Orf02 product itself. The reverse effect of the orf02p-orf02 deletion on plasmid stability was observed in P. aminovorans, in which the pESB36.40 loss rate dropped 8-fold. This coincides with the increase of the partition operon expression but also with the decrease in the level of korC and kfrC transcripts. Further studies should define the role of Orf02 in regulation of gene expression and stability of the plasmid in these hosts.
The klcAp-klcA deletion led to a severalfold increase of the transcription from orf02p in E. coli (Fig. 5E). A similar effect was observed in C. necator and P. putida (Fig. 6C); however, in these hosts, in contrast to the case for E. coli, it was conveyed into higher expression of korC-kfrC operon, as expected for the RNAP reading through from orf02p into downstream genes. In two other species, P. aminovorans and A. tumefaciens, the lack of the klcAp-klcA DNA fragment lowered the transcription of the korC-kfrC operon, indicating significant participation of klcAp in the expression of this part of the stability module. Remarkably, the deletion of klcAp-klcA fragment either had no impact on the expression of the partition operon (A. tumefaciens and P. putida) or led to an increase of the expression of this operon (P. aminovorans and C. necator). In these four strains, the effect of klcAp-klcA deletion on the expression of the adjacent operon, korC-kfrC, varied from the effect on expression of the partition operon, confirming the uncoupling of two parts of the stability module (Fig. 6C).
The maintenance of the construct deprived of klcAp-klcA (pESB36.41) varied from being stable in P. aminovorans and A. tumefaciens to being unstable in C. necator and P. putida (insets in Fig. 6C and Table 1) and E. coli (Fig. 5I). Instability of the construct in E. coli might be explained by lowered expression of the partition operon; however, in C. necator the partition operon transcription was elevated, and in P. putida it was unchanged, in comparison with the WT module. Hence, either increased production of Orf02 or lack of KlcA may account for the plasmid instability in these hosts. Further studies are needed to discriminate between these options. It is worthwhile to notice that deletions of either orf02p-orf02 or klcAp-klcA had an identical impact on the stability module expression in P. aminovorans ( Fig. 6B and C, upper panels); however, only the first deletion increased plasmid stability in this host. This favors the hypothesis of Orf02 playing the negative role in plasmid maintenance in this species.
Deletion of kfrAp-kfrA in E. coli had a significant influence on expression of the upstream and downstream genes in the module (Fig. 5F). A similar effect of Δ(kfrAp-kfrA) was detected in C. necator and A. tumefaciens (Fig. 6D). In P. aminovorans only expression of the partition operon was significantly lowered, but it did not affect the stability of pESB36.44 in this host. Modest instability of the construct was observed in A. tumefaciens, whereas the highest plasmid LR was observed in C. necator and P. putida ( Fig. 6D and Table 1). The decrease in pESB36.44 stability in A. tumefaciens, C. necator, and E. coli (Fig. 5J) may result from the lower expression of the partition operon or other genes; however, in the case of P. putida, the lack of KfrA itself seems to be deleterious for the plasmid maintenance (Fig. 6D, insets, and Table 1).

DISCUSSION
Plasmids usually provide advantageous traits to their hosts but also impose a fitness cost on the cells. In the absence of selection for the plasmid-borne beneficial traits, the cells that have lost the plasmid easily outcompete the plasmid-bearing rivals (43), which should lead to plasmid extinction. However, laboratory and environmental studies have shown long-term plasmid persistence under nonselective conditions, a phenomenon dubbed the "plasmid paradox" (44). To ameliorate the fitness cost, the plasmid and host genomes undergo compensatory coevolution, which has been of great research interest in recent years (17,43,(45)(46)(47)(48). To increase the probability of the compensatory evolution events taking place, plasmids have to ensure lowering the rate of their loss (49,50).
To this end, plasmids have evolved efficient copy number control mechanisms and tight regulation of gene expression to diminish the metabolic burden imposed on the host as well as evolved/acquired specific stability mechanisms to support their maintenance in the population, which is especially important for low-copy-number replicons. Besides the prevalent and well-known multimer resolution systems (51), active partition mechanisms (52), and postsegregational killing (toxin-antitoxin) systems (53,54), additional factors may play significant roles during the establishment of broadhost-range plasmids in phylogenetically diverse hosts (7,45,55).
Above all, the promiscuous plasmids have to secure an adequate production of vital plasmid proteins regardless of the highly diversified efficiency of the transcriptional and translational machinery in these hosts (56). The question of how the broad-host-range plasmids cope with various host factors to ensure adequate levels of plasmid gene expression is far from being fully understood.
In general, gene expression depends on the RNAP recognition of promoters (57), the level of transcriptional regulators, transcriptional organization, global and local DNA topology controlled by host-or plasmid-encoded nucleoid-associated proteins (NAPs) (58,59), and the position of the gene in the genome with respect to the origin of replication (60,61). The initial concept of the bacterial operon defined it as a transcriptional unit (TU) that provided simultaneous expression of genes within the operon through production of a single polycistronic mRNA initiated at the promoter upstream of the first gene (62). Control of this unique promoter allowed a coordinated regulation of all genes. However, recent studies have revealed that bacterial transcriptomes are far more complex than previously thought (63)(64)(65). The architecture of bacterial operons may involve multiple internal differently regulated promoters and various terminators generating multiple TUs and leading to differential gene expression within these operons (66). It has been shown that various blocks of genes in an operon are often alternatively transcribed under diverse conditions; e.g., the E. coli polycistronic flagella operon fliF to fliR of thirteen genes may be split in up to seven various suboperons after heat shock or phosphorus starvation (67).
Genome-wide transcript analysis not only demonstrated variability of transcripts corresponding to the operon-clustered genes but also stressed the important regulatory roles that the stability of the mRNAs and their translation efficiency play in gene expression (42,66,(68)(69)(70)(71). First, it was assumed that the prokaryotic transcripts are processed only by 3=-to-5= exonucleases, leading to an abundance of transcripts for the 5= part of the multicistronic operons (polarity expression). Discovery of the endoribonucleases and their role in mRNA decay in combination with regulatory functions of RNA binding proteins and noncoding RNAs (ncRNAs) revealed a complex regulatory network responsible for differential transcript abundance even in the operons (72)(73)(74)(75)(76).
Similarly to the organization of their hosts' genomes, the mosaic plasmids cluster the genes engaged in the same processes into functional modules, i.e., the replication, conjugation, and stability modules. This evolutionary trend has been usually explained by gene products forming complexes, e.g., replisome, relaxosome, or transferosome, and temporal/spatial requirements for the synthesis of such complexes (77). Even if various stability functions of the plasmids are encoded in several operons, they are often located next to each other, seemingly representing separate TUs integrated only by common regulators, such as in RK2 (GenBank accession no. BN000925) of IncP-1 (78), R388 (GenBank accession no. BR000038) of IncW (79), or R46 (GenBank accession no. AY046276) of IncN (80).
Our studies on plasmid RA3, the archetype of the IncU group, have revealed that this conjugative, low-copy-number, broad-host-range plasmid has a mosaic modular structure with functional modules intertwined by the global regulatory network (24,(28)(29)(30)(31)81). The present project focused on the expression of the RA3 stability module, comprising a unidirectionally transcribed set of ten ORFs split apparently into five differently regulated operons (24). In the course of these studies, additional, previously unanticipated TSSs as well as transcription terminator and attenuator signals were found in the RA3 stability module.
The active partition operon korA-incC-korB-orf11 has previously been identified as the main part of the stability module (24,28), but the roles of the remaining six ORFs have not been known. Before attempting to analyze them, it seemed important to establish the transcriptional organization of the region and elucidate if and how the individual regions of the module influence the expression of other genes. The possibility of RNAP read-through and its putative effect on the relative abundance of individual transcripts along the module (gene transcript dosage) were investigated.
Thus, it was shown here that the whole RA3 stability module was built as a polycistronic operon, orf02 to orf11, but transcribed into alternative mRNAs due to at least five internal promoters and various terminators. Such a transcriptional organization may have evolved to ensure the relative expression levels of the individual genes that would meet the requirements of various hosts. The transcription analysis of the stability module in representative strains of Alpha-, Beta-, and Gammaproteobacteria in which RA3 replicated and seemed to be stably maintained (24) indicated differential expression patterns.
The transcriptional profiles of the RA3 stability module in E. coli and the four other hosts, P. putida, P. aminovorans, A. tumefaciens, and C. necator, showed similar abundances of transcripts for the partition genes ( Fig. 5A and 6A) but a substantial variability for the upstream region of the module. The relatively high expression from korCp in P. putida, P. aminovorans, and A. tumefaciens suggests more efficient RNAP recognition of this transcription initiation signal, leading to a stronger repression of the first two promoters by KorC than in E. coli (Fig. 5D and 6B). In turn, in C. necator the expression of the first three operons was at a similar level, indicating yet another specificity of its transcriptional machinery.
Transcriptional analysis of pESB36.35 deletion derivatives in E. coli clearly demonstrated the impact of the upstream transcriptional signals on the expression of the downstream genes due to RNAP read-through. Some observations indicated that multiple transcripts originating in the stability module may be more prone to nucleolytic degradation than others, as was observed, e.g., for the active partition operon. The much higher level of incC mRNAs than of korA mRNAs not only seems to result from the presence of weak internal promoters within korA but indicates the distinct stability of transcripts starting at korAp versus incCp.
The transcriptional studies conducted in different hosts demonstrated the speciesspecific dependence of expression of particular cistrons on RNAP read-through, TSSs, and intrinsic and Rho-dependent terminators and also suggested differential mRNA decay. In various combinations of host and plasmid deletion variant, a high plasmid loss rate was observed (Table 1), but further studies are needed to define whether the plasmid instability is due to transcriptional dysfunction or lack or abundance of specific products.
Analysis of the stable maintenance of pESB36.35 deletion derivatives in various hosts combined with the transcriptional analysis by qRT-PCR allowed us to determine the important, though converse, roles of the Orf02 product in A. tumefaciens, C. necator, P. putida, and P. aminovorans. At this point, no conclusion about Orf02 function in E. coli may be drawn, since deletion of orf02p-orf02 diminished expression of all downstream genes and the effect on segregation might have been caused by lower expression of, e.g., the partition operon. As mentioned above, the destabilizing effect of KlcA deficiency in E. coli, C. necator, and P. putida is far more complicated to interpret due to the simultaneous increase in orf02 expression. Finally, KfrA is definitely important for plasmid establishment/maintenance in P. putida, E. coli, A. tumefaciens, and C. necator but not in P. aminovorans, and its role is under investigation.
Studies on functions of particular plasmid-encoded proteins in plasmid biology are usually carried out on deletion/substitution mutants. However, as shown here, such genetic modifications may introduce changes in the transcriptional organization (initiation and termination signals) and regulation of gene/operon expression locally but may also affect gene expression in other regions of the plasmid. To avoid erroneous conclusions, the analysis of plasmid functions should be complemented by a thorough transcriptional analysis. As demonstrated here, in the case of broad-host-range plasmids, such analysis should be conducted in diverse hosts.
Plasmid DNA isolation and analysis and DNA amplification and manipulation. Plasmids used and constructed in this study are listed in Table 3 (those referred to in the main text) and in Table S1 in the supplemental material (intermediate constructs). Plasmid DNA manipulations were carried out by standard procedures (87) and in accordance with the manufacturers' instructions. Standard PCRs (88) were performed with appropriate pairs of primers listed in Table 4 and the RA3 DNA template.
Complementary oligonucleotides corresponding to the putative transcriptional terminator sequences or introducing restriction sites were annealed by heating to 95°C, slowly cooled, and cloned into appropriately digested plasmids. All new plasmid inserts were verified using dye terminator sequencing at the Laboratory of DNA Sequencing and Oligonucleotide Synthesis, Institute of Biochemistry and Biophysics, Polish Academy of Science.
Plasmid construction. The medium-copy-number promoter-probe vector pPT01, based on the pSC101 replicon (40), was used to monitor the promoter activity of the cloned DNA fragments. PCR products were cut with BamHI and PaeI and ligated into pPT01 upstream of the promoterless xylE cassette. To test the activity of putative transcriptional terminators, pGBT70 trfAp-1-xylE (36), a pPT01 derivative, was used. The point mutation in the Ϫ10 box of a very strong trfAp of RK2 (trfAp-1) lowers its transcriptional activity more than 10-fold, making it suitable for assays of weak transcription termination signals (37). PCR-amplified fragments or double-stranded oligonucleotides corresponding to the putative transcriptional terminators/attenuators were inserted between KpnI and NcoI restriction sites in the test vector to separate trfAp-1 from the xylE cassette. Catechol-2,3-dioxygenase (XylE) activity assays were indicative of functionality of the inserts in modulation of xylE transcription.
The plasmid loss rate (LR) per generation (%) was calculated using the formula ( n is the number of generations, F i is the fraction of cells containing plasmid at the initial time point, and F f is the fraction of cells containing plasmid at the final time point. The stability index (SI) for each construct was calculated as the ratio of the rate of loss of pESB36 to the rate of loss of the vector with the RA3 stability module variant in each host (94). Plasmid copy number. The copy number of pESB36 and pESB36.35 in various hosts was estimated by real-time qPCR. E. coli EC1250, A. tumefaciens LBA1010R Rif r , P. aminovorans JCM7685 Rif r , P. putida KT2442 Rif r , and C. necator 7MP228r Rif r carrying pESB36 or pESB36.35 were grown on L broth with a selective antibiotic to the stationary phase. Total bacterial DNA was extracted using a modification of the method of Chen and Kuo (95,96). Briefly, cells were collected from 5 ml of the cultures by centrifugation and suspended in 400 l of TESO lysis buffer (40 mM Tris-HCl [pH 8.0], 1 mM EDTA [pH 8.0], 1% [wt/vol] SDS, 20 mM sodium acetate) and 132 l of 5 M NaCl. The samples were vortexed for 1 min and then centrifuged at 4°C for 10 min at 16,000 ϫ g. One volume of phenol-chloroform (1:1 [vol/vol]) was added to the supernatants, and samples were vortexed for approximately 1 min and centrifuged as before. The extraction procedure was repeated for the upper phase, with an equal volume of chloroform. Subsequently, the aqueous fraction was collected, and DNA was precipitated with 2 volumes of 98% ethanol (87). Fifty and 100 nanograms of each DNA template were used in qPCRs with 5ϫ Hot FIREPol EvaGreen qPCR Mix Plus (Solis Biodyne), and reactions were carried out according to the manufacturer's instructions in the LightCycler 480 Instrument II (Roche). Single-copy genes on the chromosomes were chosen as the reference genes: cysG for E. coli, rpoD for P. putida, P. aminovorans, and A. tumefaciens, and gyrB for C. necator. trfA was used as a plasmid target for pESB36 and pESB36.35 (pairs of primers are listed in Table 4). The primers were checked for specificity and efficiency (only primers with an amplification factor between 1.95 and 2 were used). All qPCRs were done in triplicates. The PCN, defined as the number of plasmid amplicons relatively to the number of chromosome amplicons, was calculated considering the amplification efficiencies of the primers used (95,96). The average results of at least four biological replicates with standard deviation were reported.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.4 MB. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.