Two birds with one stone: SGI1 can stabilize itself and expel the IncC helper by hijacking the plasmid parABS system

Abstract The SGI1 family integrative mobilizable elements, which are efficient agents in distribution of multidrug resistance in Gammaproteobacteria, have a complex, parasitic relationship with their IncC conjugative helper plasmids. Besides exploiting the transfer apparatus, SGI1 also hijacks IncC plasmid control mechanisms to time its own excision, replication and expression of self-encoded T4SS components, which provides advantages for SGI1 over its helpers in conjugal transfer and stable maintenance. Furthermore, SGI1 destabilizes its helpers in an unknown, replication-dependent way when they are concomitantly present in the same host. Here we report how SGI1 exploits the helper plasmid partitioning system to displace the plasmid and simultaneously increase its own stability. We show that SGI1 carries two copies of sequences mimicking the parS sites of IncC plasmids. These parS-like elements bind the ParB protein encoded by the plasmid and increase SGI1 stability by utilizing the parABS system of the plasmid for its own partitioning, through which SGI1 also destabilizes the helper plasmid. Furthermore, SGI1 expresses a small protein, Sci, which significantly strengthens this plasmid-destabilizing effect, as well as SGI1 maintenance. The plasmid-induced replication of SGI1 results in an increased copy-number of parS-like sequences and Sci expression leading to strong incompatibility with the helper plasmid.


Introduction
Multidrug-resistant (MDR) pathogenic bacteria represent a global threat to public health and animal husbandry.The current practice of preventive overuse, or even misuse, of antibiotics in livestock farming could lead to a post-antibiotic era when drugs may be ineffective for treatment of bacterial infections ( 1 ).The emergence of MDR bacteria is mainly facilitated by the activity of various mobile genetic elements such as transposons, genomic islands and conjugative plasmids, which often carry antibiotic resistance (AR) genes.Mobile genomic islands, also referred to as integrative elements (IE), have recently been recognized as key players in multidrug resistance propagation ( 2 ,3 ) as they generally contain arrays of resistance genes associated with, or independently of, integrons.Two major types of IE have been distinguished based on their autonomous functions.Integrative and conjugative elements (ICEs, formerly known as conjugative transposons ( 4 )) possess the entire genetic apparatus required for chromosomal integration, excision, and conjugative transfer.Unlike ICEs, integrative mobilizable elements (IMEs) can integrate autonomously, but lack a complete set of conjugation genes, and therefore require a transfer-competent helper element, such as a conjugative plasmid or ICE, to ensure their horizontal transfer ( 5 ).
SGI1-family elements are mobilized by the broad host range, low-copy IncA and IncC conjugative plasmids (25)(26)(27).These closely-related plasmids ( 27 ,28 ) have the same conjugative system as the MOB H12 group ( 29 ).The conjugation genes are regulated by the FlhDC-family master activator, AcaCD, whose expression is controlled by negative and positive feedback loops established by the Acr1 and Acr2 repressors and the transcription activator, AcaB (30)(31)(32).SGI1 does not simply utilize this conjugation apparatus, but applies several refined interventions at the expense of the plasmid, to increase the success of its own lateral transfer and ensure stable maintenance in donor and recipient cells ( 20 , 18 , 33 ).
After entry into the new host, SGI1 integrates into the chromosome at the 3 end of trmE (also known as thdF or mnmE ) via site-specific recombination carried out by the SGI1 λ-family integrase (Int) ( 25 ).This ensures highly stable vertical transfer of SGI1 ( 34 ,24 ) until an IncC plasmid appears.When the plasmid is present, five SGI1 operons, including xis , rep , tr aN S , and tr aH S G S (Figure 1 A), that are controlled by AcaCD-responsive promoters ( 35 ), are activated, allowing SGI1 to exploit several functions of the plasmid.AcaCDactivation induces production of the recombination directionality factor, Xis, which promotes SGI1 excision catalysed by Int ( 30 ,21 ).This recombination event generates the extrachromosomal circular form of SGI1, which can be mobilized by the helper plasmid, but is also liable to segregation ( 21 , 24 , 36 , 18 ).To prevent SGI1 loss from cells in which it coexists with the helper plasmid, the island both expresses a TA system ( 24 ) and begins to replicate like a plasmid.Replication initiated by an IncN2-type RepA protein expressed from the S004rep operon, which is also activated by AcaCD, maintains excised SGI1 at 6-8 copies ( 36 ,18 ), significantly increasing its stability ( 18 ); however, another FlhDC-family activator, the SGI1encoded SgaDC (also known as FlhDC SGI1 ) is also necessary to reach the optimal SGI1 copy number.In the absence of SgaDC, AcaCD-activation of S004rep is reported to maintain SGI1 in an approximately single-copy state ( 18 ); however, complete loss of the SGI1 Kn sgaDC replication has also been reported ( 37 ).
Many observations suggest that coexistence of SGI1 and IncC plasmids is not stable.Natural isolates harbouring both elements have never been reported.Under experimental circumstances, a moderate frequency of SGI1 loss was observed in Salmonella Typhimurium strains when antibiotic selection was applied only for the IncC plasmid, R55 ( 21 ).SGI1 loss was also observed in E. coli under nonselective conditions; however, SGI1 appears to be more stable than the IncC plasmid, which is lost from 45%-85% of the cell population after 20 generations ( 38 , 36 , 18 , 39 ).Since SGI1 behaves like a plasmid in the presence of an IncC plasmid, their mutual destabilization resembles the phenomenon of plasmid incompatibility, which is defined as the failure of stable inheritance of co-resident plasmids without external selection.This phenomenon occurs if a plasmid destabilizes another within the same host cell because their maintenance functions, such as elements involved in replication and copy number control, or partitioning, interfere with one another due to their relatedness, and leading to irregular plasmid inheritance.Incompatibility is symmetric when both plasmids are lost with equal probability, or vectorial when one plasmid is lost with higher probability ( 40 ).
Replication-mediated incompatibility is a well-established model that has led to the paradigm of classifying plasmids into widely used incompatibility groups (Inc).In this case, the sharing of related elements (iterons, antisense RNAs, RepA proteins) that control the replication of both plasmids can lead to a complete or partial blocking of replication of one plasmid, or prevent the replication of both plasmids at the frequency required for stability ( 40 ).For stable maintenance, low-copy number plasmids need active partitioning by Par systems, which ensure that every daughter cell receives at least one plasmid copy after segregation ( 41 ).Par systems are also important for ICE stability (42)(43)(44).The systems responsible for partitioning are also significant incompatibility factors.Similarly to replication-mediated incompatibility, competition arises between two plasmids or ICEs for shared components if they have closely related Par systems, which can lead to partition-mediated incompatibility ( 41 ,45 ).Generally, partition systems comprise three main components: one or more copies of a partition site ('centromere'), a centromere binding protein (CBP), and an ATPase or GTPase protein.Three general system types have been identified and classified based on their partition NTPases ( 46 ,45 ).The most prevalent Type I systems ( parABS ), which occur in many plasmids, such as F, P1 or RK2, and are the only Par systems of bacterial chromosomes, include a Walker-type ATPase (ParA), a centromere binding protein (CBP or ParB), and one or more centromere-like DNA sites, parS .Interestingly, a number of plasmids have two Par systems, generally of different types (e.g.pB171, R27).In these cases, Type I systems appear to be more important for plasmid stabilization ( 47 ,48 ).In addition, TA system-mediated incompatibility has also been reported among Ti plasmids ( 49 ).
Such systems may be associated with the incompatibility between SGI1 and the IncA and IncC plasmids, although no Par system has yet been identified in SGI1.Replication of IncC plasmids and SGI1 are driven by the unrelated IncC  S1 ).(C) Proportion of R55 Tn+ cells after the 5th passage (100 generations growth).Four replicates from the 5th passage were pooled, spread on LB plates selective for SGI1, and retention of R55 Tn was individually tested by replica plating.Number of total and R55 Tn+ colonies tested are listed and the plasmid loss is indicated as % of the R55 Tn − colonies.If no R55 Tn − colonies were found, the threshold limit was calculated as 1 / total colony count.If high-frequency loss of R55 Tn was observed, the percentage of R55 Tn+ cells was calculated as mean of the Cm R / total cell titers of the four parallels measured in the 5th passage.
The third major factor that often mediates incompatibility is partitioning.IncC plasmids belong to a class of plasmids that have two Par systems, as both parABS and parMRC loci have been identified in their backbone.The Type I parABS system is crucial for plasmid maintenance, as no insertions were obtained in the parA or parB genes in a Tn 5 mutagenesis screen, indicating that these genes are essential.A further indispensable factor, ORF053 , was also detected ( 51 ), however, its exact role and identification of the cognate parS site(s) of the system remain to be established.In contrast, the parMRC system, which appears to be part of the AcaCD-regulon ( 30 ), may be less important for IncC stability, as this locus was not identified as non-mutable ( 51 ).Interestingly, this Par system is related to srpMRC identified in SXT / R391 ICEs, which is expressed under the control of the FlhDC-family activator, SetCD, and involved in reducing R391 loss ( 44 ).Unlike SXT-family ICEs, SGI1 appears to lack a functional Par sys-tem.Thus, apparently neither replication-mediated, nor TAor partition-mediated incompatibility plays a role in the SGI1-IncC relationship.
Recent works have indicated that the SGI1-encoded regulator, SgaDC, which has a key role in crosstalk between SGI1 and IncC plasmids, may also be involved in their incompatibility.The SGI1-K variant, lacking the sgaDC operon due to a deletion did not destabilize the IncC plasmid pRMH760 ( 38 ,39 ).Further support for this hypothesis came from stability assays, where the IncC plasmid, R55, appeared to be much more stable under nonselective conditions in the presence of an SGI1 mutant lacking sgaDC than in the presence of a wt SGI1-C ( 18 ).Since SgaDC is proven to be a key factor in SGI1 replication, a consensus has emerged that it acts on SGI1-IncC incompatibility by influencing the control of SGI1 replication and copy number ( 18 ,37 ), however, the exact mechanism remains unclear.Alhough SgaDC appears to have a key role in SGI1-IncC incompatibility, it has been previously suggested that the ultimate explanation for this phenomenon may rather be the presence of an, as yet unidentified, SGI1-encoded factor, whose dosage is increased due to SGI1 replication in the presence of the helper ( 18 ).In this study, we report discorvery of such factors and describe how SGI1 destabilizes the IncC plasmid and increases its own stability by disturbing and hijacking the main partition system of the plasmid using these factors.

DNA and microbial techniques
Standard molecular biology procedures were carried out according to ( 53 ).Enzymes were purchased from Thermo Fisher Scientific, New England Biolabs and Merck; chemicals were from Merck, Roth and Reanal.For cloning purposes, E. coli TG1 was used as the host strain.Pwo (Roche) or Phusion (Thermo Fisher Scientific) high-fidelity DNA polymerases were applied for PCR and cloned PCR products were sequenced on an ABI 3500xL Genetic Analyzer (Life Technologies).Oligonucleotides used in this work are listed in Supplementary Table S1 .Primers were designed using the published sequence of SGI1 (GenBank: AF261825).Test / colony PCRs were performed using DreamTaq polymerase (Thermo Fisher Scientific), as described previously ( 34 ).Relevant features of plasmids are listed in Supplementary Table S2 .Detailed methodology of construction of plasmids is described in Text S1.
Total DNA for RT-qPCR assays was isolated using Ex-ctractMe DNA & RNA Extraction kit (Blirt S.A.), according to the manufacturer's protocol.
Mating was carried out as described previously ( 23 ).Briefly, 100 μl of overnight (ON) cultures (1-2 × 10 9 cells / ml) of donor and recipient strains grown in LB supplemented with appropriate antibiotics were mixed, centrifuged for 1 min, washed with 0.5 ml 0.9% NaCl, spread on LB agar plates, and incubated for 6 h at 37 • C.After suspension and dilution of the bacterial lawn in 0.5 ml 0.9% NaCl, transconjugants were selected on LB plates for the chromosomal and SGI1 markers of the recipient (NalSmSp) and the transferred marker of the test plasmid (Cm).
KO fragments were electroporated using a BTX Electro Cell Manipulator 600 and 2-mm gap electroporation cuvettes as described previously ( 56 ).To maintain and cure the plasmids with a temperature-sensitive pSC101 replication system, incubation at 30 • C and 42 • C was applied, respectively.
β-galactosidase assays ( 57 ) were performed in four biological replicates with some modifications ( 21 ).P parA promoter activity was measured using the β-galactosidase test plasmid pMSZ1260 (Sm R Ap R ) harbouring the 173-bp upstream region of the parAB053 operon of R55 in front of a promoterless lacZ gene.Measurements were conducted in the presence or absence of R55 Tn (Cm R ), R16a (Km R ), the single copy minimal IncC plasmid, pMSZ1248 (Cm R ), or the pBeloBac11-derived expression vector, pMSZ1239 (Km R ), with the p15A-based IGR-bearing plasmids, pJKI1149 (Km R ) or pMSZ1164 (Cm R ), or the negative control plasmids, pJKI88 (Km R ) or pJKI405 (Cm R ).The test plasmid and the p15A-based IGR-bearing or negative control plasmids were co-transformed into E. coli strain TG1Nal harbouring R55 Tn , R16a, pMSZ1248, or pMSZ1239.Starter cultures were grown in LB medium supplemented with antibiotics selecting for all plasmids at 37 • C ON.These cultures were diluted 100-fold in LB + antibiotics and grown at 37ºC to OD 600 ∼0.3.In the case of pMSZ1239, expression of ParAB053 proteins was induced with 0.1% l -arabinose during growth to OD 600 ∼0.3.

Plasmid stability assays
To monitoring IncC plasmid stability of the in the presence of wt SGI1-C or its mutant derivatives, R55 Tn (Cm R ) was conjugated from E. coli TG90 / R55 Tn (Tc R Cm R ) into an E. coli TG1Nal (Nal R ) recipient strain carrying Sm R Sp R SGI1-C wt or one of the 1-8 mutants.Transconjugants harbouring both SGI1 and the IncC plasmid were selected on LB plates supplemented with Nal,Sm,Sp,Cm.
When the stability of R55 Tn was assayed in the presence of SGI1-derived fragments cloned into different copynumber plasmid vectors, empty plasmids (used as negative control) and their SGI1-fragment-containing derivatives were transformed into E. coli strain TG1Nal / R55 Tn (Nal R Cm R ).Transformants harbouring both R55 Tn and one of the test plasmids were selected on LB plates supplemented with Cm,Km or Cm,Ap.For in trans systems, Cm,Km,Ap selection was applied.
The truncated SGI1 of pJKI670 (and its derivatives in pGMY14, pGMY20, and pGMY28) originated from the SGI1-C variant of the S. Typhimurium DT104 isolate, ST28S / 1, which emerged spontaneously from a wt SGI1 in the parental strain ST1773 ( 34 ).The d1 deletion between DR1 and DR2 repeats occurred during the entrapment of SGI1-C, resulting in the plasmid pJKI669, which carries the SGI1-Cd1 variant (formerly designated SGI28S / 1d1).Since the SGI1-Cd1 variant has a different origin from that of our wt SGI1-C, it was sequenced and several SNPs were identified as follows: T9C in DRL, A53G in ORF S017, an A deletion 150 bp upstream of S018, and T499A, T502G, A530C, G557A, C669A and C582A in ORF S044.All other cloned SGI1 fragments were amplified from the wt SGI1-C.
Assays were conducted with four biologically independent replicates, as described in ( 38 ), with some modifications.Single transconjugant or transformant colonies were grown ON at 37 • C in 2 ml LB broth under selection for R55 Tn and SGI1, or the test plasmids.Then, 100 μl of a 10 5 × dilution of the starter culture (several thousand cells) was transferred into 5 ml fresh medium and cultured at 37 • C to stationary phase, with selection for SGI1 or the test plasmids, but without selection for R55 Tn .Passaging (approx.20 generations / passage) was repeated daily five times.Cultures of each passage were serially diluted and the cell counts were determined on LB agar plates with selection for SGI1 (Sm,Sp) or the test plasmids (Km or Ap or Km,Ap in the cases of trans systems) to determine the total number of cells, or for R55 Tn (Cm) to determine the frequency of cells that retained the IncC plasmid.Rate of IncC plasmid loss was calculated as the ratio of R55 Tn+ to total cell titres, mean and standard deviation values of the replicates were plotted against the number of passages.In most cases, stability assays were repeated 2-4 times, but references (e.g.pJKI670, pGMY64, pJKI1153) ( Supplementary Dataset 1 ) and negative controls were tested several times (pJKI88: 4 replicates in 31 experiments, pJKI298: 4 replicates in 4 experiments).Data from all assays were used to calculate mean and standard deviation values and for statistical comparisons.When low rates of plasmid loss were observed by titration, R55 Tn retention in the 5th passage was also confirmed by replica plating.In these cases, 10 6 × dilutions of the four parallel cultures were pooled and spread onto LB agar plates supplemented with antibiotics selective only for SGI1 or the test plasmids.After ON incubation at 37 • C, colonies were replica plated onto LB + Cm to determine the proportion of R55 Tn+ colonies.
To monitor the stability of Cm R SGI1-based minimal replicons in the presence of the R55-derived parAB053 operon, pMSZ1175b and pMSZ1200 were transformed into TG1 cells containing the Km R plasmids, pJKI625 (negative control) or pJKI1128 (P ara :: parAB053 ).Four transformant colonies with every plasmid combination selected on LB + Km + Cm plates were grown ON in LB + Km + Cm broth at 37 • C, then 100 μl of 10 5 × dilutions of ON cultures was transferred into 5 ml LB broth supplemented with Km and 0.001% l -arabinose.Passaging was repeated 5 times, as described above.
For stability assays under nonselective conditions, four colonies containing R55 Tn and wt SGI1-C, or one of the single (KODR1 or KODR2), double (KODR1 + KODR2) or triple (KODR1 + KODR2+ sci ) KO mutants, were grown ON in LB + Nal + Cm + Sm + Sp broth (selecting for both elements).Then, 40 μl cultures from the 10 5 × dilution ( ∼400 cells) were transferred into 2 ml fresh LB broth supplemented with 0.1% glucose and grown to stationary phase without antibiotic selection, to obtain cell populations of ∼22 generations, respectively.Total cell counts of four replicates were determined on LB + Nal agar plates.Proportions of cells retaining R55 Tn and / or SGI1 were determined by replica plating onto LB + Nal + Cm, LB + Nal + Sm + Sp, and LB + Nal + Cm + Sm + Sp plates.The percentages of bacteria harbouring R55 Tn , SGI1 or both were calculated as the fractions of Cm R Sm S Sp S / Nal R , Cm S Sm R Sp R / Nal R and Cm R Sm R Sp R / Nal R colonies, respectively.Ratios of the subpopulations of cells that lost both SGI1 and R55 Tn were calculated as Cm S Sm S Sp S / Nal R , where Cm S Sm S Sp S = Nal R -(Cm R Sm S Sp S + Cm S Sm R Sp R + Cm R Sm R Sp R ).

Mobility shift assay
Purification of ParB protein.Overnight culture of E. coli strain Tuner (DE3) transformed with pMSZ1166 (pET16bbased ParB-producer) was diluted 100-fold in 25 ml LB + Ap broth and grown to an OD 600 of 0.5 at 37ºC.The culture was induced with 0.3 mM IPTG at 30ºC for 3 h under vigorous shaking, then bacteria were harvested by centrifugation and resuspended in 1 ml lysis buffer (50 mM Tris pH 8.1, 300 mM NaCl, 0.01% Triton X-100) supplemented with 60 μg / ml lysozyme and 30 μl Complete protease inhibitor cocktail (Roche) prepared according to the manufacturer's recommendations.Cells were frozen at -70ºC, then thawed and sonicated on ice at 50% activity for 4 × 10 s.Lysates were centrifuged at 16000 g for 30 min at 4 • C and supernatants (cleared lysates) used for purification of ParB with a Dynabeads ® His-Tag Isolation & Pulldown Kit (Novex Life Technologies), according to the manufacturer's protocol.Purified protein ( ∼250 ng / μl) was kept on ice until use.

Real-time quantitative PCR (RT-qPCR) for relative quantification of the copy number and excision of DR1 KO, DR2 KO and Δsci mutant SGI1-C
Copy numbers of free circular SGI1 per cell ( attP ), and the unoccupied integration site ( attB ) in TG1Nal strains carrying wt, or the DR1 K O , DR2 K O , and sci ( IGR_ORF2) mutant SGI1-C, in the presence of R55 Tn were determined by RT-qPCR.Amounts of attB and attP were normalized to those of the single-copy chromosomal gene, trmE .The TG1Nal:: attP SGI1 strain, containing single chromosomal copies of att B and attP ( 18 ), was used to calibrate the RT-qPCR assay.The LJ3-RJ5, attsgifor2-attsgirev2 and attsgifor2-attsgirev3 primer pairs were used to amplify specific 251, 257 and 207 bp attP, attB and trmE fragments, respectively.For RT-qPCRs, total DNA was extracted using the EXtractMe Kit (Blirt DNA, Gdansk), according to the manufacturer's instructions.RT-qPCRs were performed in a final volume of 10 μl using a LightCycler ® 96 detection system (Roche).Each reaction mixture contained 1 × qPCRBIO SyGreen Lo-ROX master mix (PCR Biosystems), 400 nM forward and reverse primers, and 10 ng total DNA as template.PCR conditions were: initial denaturation at 95 • C for 2 min; 40 cycles of 95 • C for 5 s and 60 • C for 30 s. Amplification specificity was confirmed by generating melting curves.Relative amounts of amplified attP and attB sequences were calculated based on the Ct deviation of samples compared with the single-copy-containing control sample, TG1Nal:: attP SGI1 , and are expressed relative to the calibrator ( trmE ) sequence.Consequently, the ratio of attP and trmE was used to measure excised SGI1 copy number, while the ratio of attB and trmE measured the excision of the island.Results are presented as mean and standard deviation values of data obtained from total DNA of four individual colonies.

Bioinformatics and statistical analyses
Multalin Interface ( 58 ) was used to generate alignment of parS sites.The sequence logo was created using the WebLogo server ( 59 ).Promoter motifs were predicted using BPROM ( 60 ) and BDGP ( 61 ), as well as by manual search.All homology searches were performed using the NCBI BLAST server.SGI1-related elements in GenBank were identified via nucleotide BLAST search using the previously defined SGI1 backbone ( 22 ) as the query sequence.Putative small RNA secondary structures were predicted using mFold server ( 62 ).Protein structures were predicted using Phyre 2 ( 63 ), PSIPRED ( 64 ) and AlphaFold ( 65 ,66 ).Pfam ( 67 ) was used for protein family or domain searches.
Statistical analyses for comparisons of R55 Tn stability in the presence of test plasmids containing different SGI1 fragments were carried out by fitting a monoexponential decay model to the plasmid presence ratio data with passage number as an independent variable.The decay model was expressed in its common version of μ = 2 (−pn/t 1 / 2 ) where μ is the mean predicted amount of plasmid expressed as a percentage of the initial 100%, pn is the passage number, and t 1 / 2 is the virtual passage number at which the plasmid amount drops to 50%.This model allows direct estimation of the half-life parameter ( t 1 / 2 ) for each sample, together with the corresponding 95% confidence limits by means of nonlinear regression analysis.Comparisons of the decay kinetics between the samples and references were then conducted by computing the ratios and 95% confidence limits of the corresponding half-life parameters, and significant influences were identified when the confidence interval of the ratio did not contain 1.0.Regression models fitted to the measured data points were plotted against passage number to visualize the decay kinetics of the samples and the references in the same figures ( Supplementary Dataset 1 ).Due to the limited number of experiments, no adjustment for multiple comparisons was applied to avoid the loss of statistical power in detecting practically important differences in decay kinetics.All statistical analyses presented in Supplementary Dataset 1 were carried out using the web-based academic version of SAS Studio 3.81.

Deletion mapping of SGI1 functions responsible for destabilization of IncC plasmids
As a first approach to identify SGI1 regions that contribute to expelling IncC plasmids, the stability of R55 Tn 6187 (abbreviated as R55 Tn ), a Flo R / Cm R Km S Gm S Ap S derivative of the IncC Type2 plasmid, R55, was examined in an E. coli TG1Nal host harbouring wt or different deletion mutants of SGI1-C (Figure 1 A).Since several previous observations published re-cently ( 21 , 24 , 36 , 18 , 37 ) indicate that wt SGI1, wt SGI1-C, and especially the replication deficient SGI1 derivatives are unstable in the presence of the IncC helper plasmids without selection, to identify and analyse SGI1-encoded factors responsible for IncC plasmid destabilization, antibiotic selection for SGI1 was applied to avoid SGI1 loss during passaging.Although there have been reports of wt SGI1 being completely stable besides IncC plasmid pRMH760 ( 38 ,39 ), based on our own and others' experience, and due to a consistent experimental setup, all strains were grown without antibiotic selection for R55 Tn but with selection for SGI1-C wt and mutants for 5 passages (approx.20 generations / passage), and the proportion of cells retaining R55 Tn was determined at each passage.As observed previously, the plasmid was rapidly lost from a host carrying wt SGI1-C.The proportion of R55 Tn+ cells dropped to approximately 10% at the first passage and to 0.3% at the second.After 100-generations of growth, no R55 Tn+ cells were detected ( < 0.073%).Similar dynamics of plasmid loss were observed if the S023-S027 ( 7) or S044 ( 8) regions were deleted from SGI1-C (Figure 1 B, C, Supplementary Figure S1 ), indicating that these deletions had no effect on SGI1-mediated incompatibility.The next class of deletions affected the excision, replication, or control of these functions ( 1: P xis-oriV , 2: repA-S004 , 3: repA-traN S , 4: sgaDC , 5 flhC SGI1 -traH S , Figure 1 A) and impeded SGI1 in reaching its normal 6-8 excised copies per cell ( 18 ,37 ).These deletions significantly reduced incompatibility with R55 Tn and caused moderate plasmid loss, as 40-70% of cells retained the plasmid after the 2nd passage and a considerable fraction of the population (10%-40%) were R55 Tn+ even after the 5 th passage.The only exception was 3, eliminating the repA-S004-traN S region, which caused somewhat slower plasmid loss, detectable only by replica plating after the 5th passage (approx.86% of cells were R55 Tn+ , Figure 1 B, C).The fact that these deletions reduced, but did not fully eliminate, the incompatibility suggested that the major factor responsible for plasmid loss was still present in these mutants.Only the 6 deletion, removing the S013-S018 region including six short ORFs with unknown functions, led to high plasmid stability as no R55 Tn -free cells were observed, even after 100 generations of growth in this case (R55 Tn − < 0.23%, Figure 1 C).
These results directed our attention towards the S013-S018 region, however, the high level of deviation observed among parallel samples in stability experiments (mainly in cases where the deletion interfered with excision or replication of SGI1, Supplementary Figure S1 ), and the fact that fine mapping of the functions of interest would require many directed mutagenesis changes to the chromosomally integrated SGI1-C, forced us to change our experimental set up to a plasmidbased system.
We have previously captured SGI1 variants in an F plasmidbased trap vector ( 34 ) and one of them, SGI1-C-d1 (formerly known as SGI28S / 1d1), carried a large deletion (designated d1) that was generated by recombination between two 31-bp imperfect directly repeated sequence motifs, DR1 and DR2, located upstream of the promoters P S004 and P traHs , respectively (Figure 1 A).This spontaneous deletion removed the entire tr aN S -tr aH S region, including sgaDC , without altering the S013-S018 region.The SGI1-C-d1 variant was cloned into a p15A-based vector (copy number, around 15 / cell) and the 3 -part of SGI1 from within mpsB was deleted.The resulting plasmid, pJKI670 (Figure 2 A), was used to test whether this truncated SGI1 still exerted incompatibility with R55 Tn .  1 A. ( B ) Dynamics of R55 Tn -loss from TG1Nal strain harbouring R55 Tn along with pJKI670, pGMY28, pGMY20, pGMY14 or pJKI88 (empty vector used as negative control).Graph shows the mean percentage of the R55 Tn+ cells in biological replicates grown under selection for the p15A plasmids (Km), but without selection for R55 Tn , through 5 passages as described previously.For individual segregation curves see Supplementary Figure S2 A. ( C ) Dynamics of R55 Tn loss from TG1Nal strains harbouring one of the plasmids expressing repA SGI1 (pGMY9), sgaDC (pGMY6) or containing the entire rep region of SGI1 (pGMY18).The empty e xpression v ector, pJKI391, w as used as a negativ e control.T he data presented on the graph w ere obt ained in similar st abilit y assa y s described in part B. For individual segregation curves see Supplementary Figure S2 B ( D ).Proportion of R55 Tn+ cells after the 5th passage (100 generations growth) in the cultures assa y ed in parts B and C. Data were obtained and the threshold limit was calculated as described in Figure 1 C. SGI1-IncC incompatibility was examined using assays similar to those described previously, except that the SGI1 regions were cloned into a p15A-based plasmid vector and introduced to the R55 Tn -bearing host by transformation.While the p15A-based empty vector, pJKI88 (used as negative control), proved fully compatible with R55 Tn , pJKI670 caused rapid R55 Tn -loss, confirming that the SGI1-derived incompatibility factor was still present in the test plasmid (Figure 2 B), however, loss of R55 Tn in this system appeared to be somewhat slower than that induced by wt SGI1-C (see Figure 1 B and Figure 2 B).To confirm our previous results, four deletions were generated in pJKI670 (Figure 2 A, Supplementary Figure S2 A) and the incompatibility of these constructs with R55 Tn was monitored by the method applied for SGI1-C deletion mutants.As expected, partial or complete removal of the S013-S014 region (pGMY14 and pGMY20) abolished R55 Tn loss, although the smaller deletion, which left the P S004 -DR1 region intact, did not entirely extinguish the incompatibility, as very low-frequency R55 Tn loss was detectable in the 5th passage (Figure 2 ABD).In contrast, a deletion inactivating the int and xis genes and removal of the entire mpsAB region from within S016 (pGMY28) did not significantly alter R55 Tn loss dynamics relative to unmodified pJKI670 (Figure 2 A, B, D, Supplementary Dataset 1 A).
The fact that deletions affecting SGI1 excision, replication, or their control reduced, but did not eliminate, the incompatibility (Figure 1 BC) suggested that repA , the entire rep region or sgaDC may contribute to this phenomenon.To examine the incompatibility with R55 Tn exerted directly by these genes, plasmids expressing repA and sgaDC from the P tac promoter (pGMY9 and pGMY6, respectively) or containing the entire rep region (pGMY18) were assayed as described previously.All three constructs caused low-frequency loss of R55 Tn , which was only detectable at the 5th passage by replica plating (Figure 2 C, D, Supplementary Figure S2 B) and could not account for the strong incompatibility observed with wt SGI1 or its truncated derivative cloned in pJKI670.These results (Figure 2 ) support that the major factor(s) mediating incompatibility of SGI1 with R55 Tn is encoded in the S013-S014 region, although roles of SGI1 replication and flhDC expression in this phenomenon cannot be completely excluded.

Fine mapping of the S0 13-S0 14 region for identification of incompatibility factors
Our further analysis focused on the region whose deletion caused complete loss of incompatibility with R55 Tn (pGMY14, Figure 2 A).This fragment of SGI1-C-d1 includes the 5 -part and promoter region of S004, the DR1 repeat, the ORF S013 and S014, the intergenic region (IGR) between S014 and S015 and the 3 -part of S015.A series of plasmid constructs were created to identify which part of this segment was responsible for the incompatibility.Shortened or site-directed mutated segments (Figure 3 A) were cloned into the p15A-based vector, pJKI88, and the incompatibility of the resulting plasmids with R55 Tn monitored as previously.The results obtained with plasmids, where the 5 -or 3 -part of the analysed region was deleted or exchanged, indicated that both ends carry elements involved in the incompatibility.Removal of the DR motif (pGMY11) or the entire IGR (pGMY34) almost completely ablated the incompatibility with R55 Tn , however, some plasmid loss was detectable at the fifth passage (see pGMY11).In contrast, restoration of the original constitution present in wt SGI1 by replacing the P S004 -DR1 segment with the P S012 -DR2 located upstream of S013 (pGMY33 versus pGMY38), or even removal of P S012 (pGMY43) did not lead to any sizeable change in incompatibility (only slightly alleviated it) (Figure 3 A, C, Supplementary Dataset 1 A).Based on these data, the shortest region exerting similarly strong destabilization of R55 Tn to that observed with pJKI670 (Figure 2 B) was the DR2-IGR segment of SGI1 (pGMY43, Figure 3 A, Supplementary Dataset 1 A).
Since the 31-bp DR motif proved to be an important factor in the incompatibility, the DR2 sequence was first modified by replacing 13 bp at its 3 end and by filling in the Pci I site located in the centre of a 22-bp inverted repeat (IR) embedded in the 5 -part of the DR.Both mutations drastically reduced the incompatibility with R55 Tn (pGMY39, pGMY49, Figure 3 B, C, Supplementary Figure S3 ).The next question was whether the ORFs S013 and S014 are also involved in this phenomenon.Both ORFs were individually knocked out by inserting an 8-bp Xho I-linker into a unique restriction site localized near the 5 end of the ORFs.Although the insertions caused frameshifts, no remarkable effects were detected, however, both resulted in a minor strengthening of the incompatibility (compare pGMY38, pGMY41 and pGMY42, Figure 3 B, C, Supplementary Figure S3 , Supplementary Dataset 1 B).Thus, the minimal sequence causing a similar (somewhat higher) rate of R55 Tn loss to that induced by the original SGI1-derivative clone, pJKI670, was achieved by abutting the DR2 repeat and the IGR (pGMY64, Figure 3 B, C, Supplementary Dataset 1 A, C).

The incompatibility exerted by DR2 and IGR is dose-dependent
The strength of incompatibility is often dose-dependent ( 40 ), therefore, the copy-number dependence of the incompatibility was examined by introducing the DR2 + IGR into the R55 Tn -bearing host on different copy-number plasmids: a one-copy pBeloBac11-based (pGMY72), an approx.15-copy p15A-based (pGMY64), and a > 300-copy pEMBL19-based (pGMY65) vectors.The kinetics of R55 Tn loss indicated ro-bust copy-number dependence of incompatibility (Figure 4 A).While plasmid loss was almost undetectable if DR2 + IGR was present at 1 copy / cell (about 0.1% of cells were R55 Tn − in the 5 th passage), > 300 copies per cell caused drastic plasmid loss (in the 1st passage < 0.02% of cells were R55 Tn+ , and were no longer detectable).In the presence of empty vectors used as negative controls, R55 Tn loss was undetectable (Figure 4 D, Supplementary Figure S4 A).
Next, we separated the two elements contributing to the incompatibility to identify which caused the copy number dependence.The DR2 and IGR sequences were cloned separately into a p15A-and a pEMBL19-based vector and their incompatibility with R55 Tn was assayed.The results showed that both elements caused very weak incompatibility alone at low copy number (approx.15 / cell) as R55 Tn − cells were only detectable by replica-plating (about 2.5% of cells in the 5th passage), but DR2 and IGR had different effects on R55 Tn loss when present at high copy number.While the IGR did not show a strikingly stronger effect at high copy number than at 15 copies (17.4% versus 2.6% R55 Tn − cells in the 5th passage), DR2 promoted robust plasmid loss at high copy number (R55 Tn+ cells were undetectable from the 3rd passage, Figure 4 BD, Supplementary Figure S4 B).Finally, to test whether they could cooperate in trans , DR2 and IGR were supplied on different plasmids with similar copy-number (DR2: p15A-based plasmid pGMY66, IGR: pBR322-based plasmid, pGMY67, ∼20 copies / cell).The results (Figure 4 C, D, Supplementary Figure S4 C) showed that together, DR2 and IGR promoted similarly strong incompatibility, whether they were present in cis or in trans (compare pGMY64 and pGMY66 + pGMY67, Figure 4 C, Supplementary Dataset 1 A, C).These data also indicate that the two elements together exert a much stronger incompatibility than either of them alone (Figure 4 ).

SGI1 DR motifs contain functional parS sites of IncC plasmids
A thorough examination of the 31-bp DR motifs identified a 22-bp IR in the 5 -part of the DRs (Figure 5 A).IR motifs often serve as binding sites, thus we supposed that the 22-bp IRs in DR1 and DR2 bind a protein involved in IncC plasmid stability.The 22-bp IRs resemble those present at one or more copies in centromere sites of many plasmids and bacterial chromosomes ( 45 ).In addition, a very similar IR motif was found in the promoter region of the parAB operon of R55 and virtually all other IncC plasmids.Using BProm promoter prediction software, a putative promoter motif was identified 64 bp upstream from the parA start codon.The largest part of the 22-bp IR motif resembling those found in SGI1 DRs lies downstream of the predicted promoter, but partially overlaps its -10 box (Figure 5 B).The length and position of the IR in P parA suggested that it may serve as a parS site for the plasmid by which the par operon is controlled, as a similar genetic constitution can be seen in both the parABS and parMRC systems of several other plasmids ( 45 ).A global homology search of the R55 sequence identified 13 additional IR-like motifs scattered across the plasmid backbone, supporting that these sequences may represent parS sites of IncC plasmids.Alignment of 14 plasmid-derived and two SGI1-derived IRs, along with their several flanking bases, indicated that the repeats are assembled from a fully conserved symmetrical 14-bp core sequence, which is flanked by less conserved AAAC and GTTT motifs (Figure 5 C).The sequence logo generated for the 16 IRs suggested that the conserved G (16 / 16) and T (15 / 16) in the GTTT sequence of the right flanking arm also have an important role, presumably in protein binding.
To examine whether these IRs serve as binding sites for ParB protein, the putative CBP of the IncC ParABS system, ParB was purified ( Supplementary Figure S5 ) and used in EM-SAs with three FAM-labelled probes: the entire 31-bp DR2 sequence and the 22-bp putative parS sites identified upstream of parA and ORF R55_174 ( eexC ), respectively.The latter was chosen because its perfect symmetrical sequence is identical to both the IR motif in the other DR of SGI1 (DR1) and the consensus of the 16 IR motifs.EMSAs showed that ParB protein bound to all three putative parS sites, while it did not bind to the empty vector sequence flanking the parS sites in the labelled probes (Figure 5 D-G).Binding specificity was confirmed by adding 25-fold unlabelled DNA fragments to the binding reactions (Figure 5 H).These results support that all 14 plasmid-derived IR motifs can bind ParB and act as actual parS sites, and demonstrate that both SGI1-derived IRs found in DR1 and DR2 are also functional ParB binding sites.
Next, binding of ParB to the insertion mutant parS sequence in DR2 (DR2**, Figure 3 B) and to the 14-bp core sequence of the parS sites was also examined by EMSA.The 4-bp insertion at the centre of the parS site did not abrogate its symmetry, but abolished ParB binding (Figure 5 I), suggesting that the previously observed loss of incompatibility with R55 Tn by the same mutation (pGMY49, Figure 3 B, C) was related to its inability to be bound by ParB.Surprisingly, ParB could not bind the fully conserved symmetrical 14-bp core sequence, affirming that the mostly symmetrical flanking bases, in particular the highly conserved GT bases in the right side flanking sequence, may have an important role in the recognition of

Identification of the incompatibility factor localized in the IGR between S014 and S015
Our data indicated that the IGR between S014 and S015 encodes an unknown factor, which drastically enhances the incompatibility with R55 Tn exerted by a single DR (Figure 4 ).The 415-bp IGR region contains two short divergent ORFs, along with putative SD-boxes and promoter-like motifs, which presumably express functional proteins (Figure 6 A).Furthermore, there are two imperfect IR motifs near the 3 end of IGR, which may act as protein binding sites or form stem-loop structures with a regulatory role in a hypothetical RNA.Therefore, we first conducted deletion mapping focused on the role of these elements in incompatibility.Since the IGR alone had a weak effect on R55 Tn stability (Figure 4 ), the deletion mapping was performed on plasmid pGMY64, which contained the DR2 joined to the entire IGR and exhibited strong incompatibility with R55 Tn (Figure 3 B).This setup ensured easier detection of any changes in the strength of incompatibility in stability assays.Thus, the test plasmids were pGMY64 analogues containing one of the differently truncated or modified IGR fragments joined to the DR2 sequence.
Three deletions were generated in the 3 end of the IGR.The shortest removed the downstream noncoding region of IGR_ORF2, along with the majority of the distal 18 / 19-bp IR motif (pGMY91).The next also removed the last 44 bp of IGR_ORF2, but left the proximal 8 / 8-bp IR motif (pJKI1137) unaffected, while the longest deletion (pJKI1131) eliminated both IRs, along with the last 68 bp of the ORF (Figure 6 A).Stability assays indicated that the two smaller deletions had no significant effect, as pGMY91 and pJKI1137 showed strong incompatibility, comparable to that of pGMY64 ( Supplementary Dataset 1 C).However, the longest deletion reduced the strength of incompatibility to a similar level to that observed with DR2 alone (Figure 6  negative controls (empty vectors: pGMY69 -pBeloBac11-derivative, approx. 1 / cell; pJKI88 -p15A-derivative, approx.15 / cell; pJKI298 -pEMBL19-deriv ativ e, > 300 / cell) see Supplementary Figure S4 A. ( B ) Analysis of copy number dependence of either DR2-or IGR-promoted incompatibility with R55 Tn .Effects of DR2 and IGR were monitored separately in approx.15 and > 300 copies / cell.The DR2-bearing plasmids were pGMY66 (p15A-derivative, approx.15 / cell) and pMSZ1174 (pEMBL19-deriv ativ e, > 300 / cell), while the IGR-bearing plasmids were pJKI11 49 (p1 5A-deriv ativ e) and pJKI11 50 (pEMBL1 9-deriv ativ e).In the case of > 300 copies of DR2 (pMSZ1174), no R55 Tn+ cells were detectable after the 3 rd passage (see panel D).For individual segregation curves see Supplementary Figure S4 B. ( C) Analysis of incompatibility induced by DR2 and IGR arranged in trans .DR2 was provided on pGMY66, while the IGR was placed on the pBR322-derivative pGMY67 (approx.20 / cell).Empty vectors, pJKI88 (c1-) and pBR322 (c2-), were used as negative controls.Bacteria carrying R55 Tn and one of the four combinations of the p15A and pBR322-deriv ativ es w ere passaged under Ap + Km selection.R55 Tn loss in the presence of DR2 and IGR in trans sho w ed similar dynamics as observed with DR2 + IGR in cis (pGMY64, approx.15 / cell, the curve from Figure 3 C is presented here for comparison).For individual segregation curves see Supplementary Figure S4 C. ( D ) Proportion of R55 Tn+ cells in the 5th passages assa y ed in parts A and B (determined as described in Figure 2 D).Data of pGMY64 from Figure 3 D is also presented here for comparison).n.d.-not detectable in the 5th passage.Supplementary Figure S6 , see also DR2-bearing pGMY66 in Figure 4 B, D).This result suggests that the integrity of the long IR and the 3 end of IGR_ORF2 is not important for R55 Tn destabilization, while the region containing the 8 / 8-bp IR appears to be necessary.
Next, the 5 -region of the IGR was gradually shortened in a series of plasmid constructs, where the 3 -part was terminated after the 8 / 8-bp IR, as in pJKI1137.Removal of the non-coding sequence downstream of IGR_ORF1 (pJKI1138) and complete deletion of IGR_ORF1 (pJKI1141) did not considerably change (slightly increased) the strong incompatibility observed with pJKI1137 or pGMY64.Surprisingly, the next deletion, additionally removing the two predicted promoters preceding IGR_ORF2 (pJKI1144) also had no effect (Figure 6 A, C, D, Supplementary Figure S6 , Supplementary Dataset 1 C).In contrast, removal of the entire upstream region of IGR_ORF2, including the putative SD-box (pJKI1154), caused a drastic reduction in incompatibility, however, this construct was still a more efficient destabilizer of R55 Tn , than the single DR2-bearing plasmid, pGMY66.Finally, deletion of the entire upstream region of IGR_ORF2, along the replacement of the original GTG (Val) START codon with a GTC (Val) and the second GCA (Ala) codon with GAC (Asp) (pJKI1155), or additional removal of the first seven codons of IGR_ORF2 (pJKI1142) reduced the incompatibility to the basal level caused by DR2 alone.
Since DR2 and the promoterless IGR_ORF2 together exerted strong incompatibility with R55 Tn , we supposed that an outer promoter controls the expression of the putative 'incompatibility gene' in pJKI1144 (e.g. a promoter-like element found in DR2).To confirm this hypothesis, the rrnB T1T2 terminator was inserted into pJKI1144 between DR2 and IGR_ORF2 (pJKI1152).This led to loss of incompatibility, which was entirely restored by insertion of the strong P cat promoter (pJKI1153) downstream of the terminator (Figure 6 DEH, Supplementary Figure S7 , Supplementary Dataset 1 C).Inserting the same rrnB T1T2-P cat cassette upstream of the intact but SD-box-deleted IGR_ORF2 (pJKI1156) also restored the strong incompatibility (pJKI1154 vs. pJKI1156, Figure 6 DFH, Supplementary Figures S6 and S7 , Supplementary Dataset 1 D), while this did not considerably increase the weak incompatibility of the START codondeleted construct (pJKI1155 vs. pJKI1157, Figure 6 D, G, H, Supplementary Figures S6 and S7 ).
We conclude that an unknown incompatibility factor is expressed from the predicted promoter region located upstream of IGR_ORF2 and can be inactivated by deletion of the START codon or the last 31 codons of IGR_ORF2, including the 8 / 8-bp IR motif.The importance of intactness of the N-terminal region, and particularly the START codon of the ORF suggest that the sought incompatibility factor is the 51 amino acids (aa) small protein encoded by IGR_ORF2.However, as removal of the last 14 codons (almost one third of the putative protein) had no negative effect on the strength of incompatibility, and given the apparent importance of the 8 / 8 bp IR in the 3 -part of IGR_ORF2, we considered the possi-bility that a small RNA is synthesized from this region, and can act as an incompatibility factor.
A small RNA or a small protein?
To answer this question, we generated several mutations in pJKI1153, which contained DR2 and a 3 truncated IGR_ORF2 with its 22-bp upstream region under the control of the P cat promoter (Figure 7 A) and exerted strong incompatibility with R55 Tn .First, the unique Bgl II site located near the 5 end of IGR_ORF2 was filled in (pJKI1158), which generated a 4-bp insertion and a frameshift in the ORF.This mutation abolished the incompatibility (Figure 7 B, D, Supplementary Figure S8 A), reducing it to the level exerted by a single DR-containing pGMY66 (Figure 4 B, D), however, the insertion also significantly changed the predicted stemloop structure in the 5 -part of the presumed small RNA ( Supplementary Figure S9 A, B), which could also be responsible for the loss of function.The next frameshift mutation was generated by introducing a single base insertion after the 34 th bp position of IGR_ORF2 (pJKI1159) in a region not involved in forming specific secondary structures in the RNA, according to secondary structure prediction ( Supplementary Figure S9 A).Similar to the previous frameshift mutation, this also abolished the incompatibility (Figure 7 BD), despite causing no significant change in the predicted RNA structure ( Supplementary Figure S9 C).Next, two mutations were designed to modify the putative RNA structure, but not the protein sequence.First, two neighbouring Leu codons (aa positions, [30][31] overlapping the left arm of the 8 / 8-bp IR were reversed (pJKI1160), which destroyed the IR motif and caused a radical change in the predicted RNA secondary structure ( Supplementary Figure S9 AD) without changing the protein sequence.Then, codons 2-10, overlapping the sites of the two frameshift mutations, were replaced with same-sense codons (pJKI1165), which modified the predicted 5 hair-pin structure in the RNA ( Supplementary Figure S9 A, E).Despite these alterations in the putative RNA structure, both constructs exerted similarly strong (or even somewhat stronger) incompatibility with R55 Tn than the unmutated plasmid, pJKI1153 (Figure 7 B, D).These data strongly suggest that the sought incompatibility factor is more likely the small protein encoded by the IGR_ORF2, rather than a small RNA.
For additional proof, a trans system was constructed, where the DR2 sequence was supplied by the p15A-based plasmid pGMY66, while the protein was expressed from a pKK223-based vector, where the complete IGR_ORF2 was placed under the control of P tac (pJKI1166).In stability assays, the empty parental vectors of both constructs were used as negative controls and loss of R55 Tn was examined as previously.Together, the DR2-bearing and IGR_ORF2expressing plasmids exerted strong incompatibility (Figure 7 CD, Supplementary Figure S8 B), similar to or even stronger than that obtained using constructs with these elements in cis (e.g.pGMY64 or pGMY91, Figure 6 B, Supplementary Dataset 1 D).As expected, the negative control combinations caused no (c1-/ c2-) or low (c1-/ IGR_ORF2 and DR2 / c2-) rates of R55 Tn loss.These results are also consistent with those obtained with the system testing DR2 and the whole IGR in trans (Figure 4 C).Considering all these findings, we conclude that the small protein (named Sci from S GI1-Inc C i ncompatibility) expressed from IGR_ORF2 (the sci gene) acts as another SGI1-borne incompatibility factor, in addition to the parS sites localized in DR1 and DR2.
Based on protein structure predictions, Sci is mainly helical protein with a 13 aa disordered region at its C-terminus ( Supplementary Figure S10 ), which appears to be unnecessary for the activity (Figure 6 A-C).Sci has no homologs in the protein databases among proteins with known functions and does not contain any identified domains or belong to established protein families included in the Pfam database.However, a Blastn search in GenBank using IGR_ORF2 sequence retrieved 225 hits, including the vast majority of identified SGI1 variants and more distant relatives, such as PGI1 variants, PGI2, AGI1, and an unnamed element found on Shewanella sp.W3-18-1 chromosome.The wide distribution and high conservation (79-100%) of this small gene in the SGI1-family indicates that the encoded protein has important roles in the stability or propagation of SGI1-like elements.
A minimal SGI1 replicon bearing a parS site is stabilized by the ParABS system of R55 Partitioning systems for plasmids or ICEs have crucial roles in their stability.Although this work initially focused on the incompatibility phenomenon observed between SGI1 and IncC plasmids, the fact that SGI1 harbours two copies of the centromere-like parS sequence, a cis -element of the ParABS system of IncC plasmids, raised the possibility that their role is not limited to plasmid expulsion, but that they may also contribute to SGI1 stabilization.To test this hypothesis, we first inserted DR2 and the DR2-IGR cassette into an SGI1derived Cm R minimal replicon assembled from oriV SGI1 and repA SGI1 placed under the control of promoter P tac ( 18 ).The stability of the resulting plasmids was examined in the absence or presence of the parAB053 operon of R55 expressed in trans from the P araB promoter .Stability assays carried out without selection for the SGI1-based replicons showed that these plasmids were rather unstable in the absence, but became much more stable in the presence, of the par operon (Figure 8 A, Supplementary Figure S11 A).The segregation dynamics also indicated that this effect is mainly due to the presence of DR2 harbouring a parS site, as the presence or absence of IGR had no significant effect on the stability of the minimal SGI1 replicon.This result suggests that the Sci protein synthesized from the IGR acts independently from the plasmid stabilization exerted by the ParAB053 proteins.
The stabilizing effect of ParAB053 proteins was further analysed using three KO mutant parAB053 -expression plasmids, in which each one of the three ORFs ( parA , parB or ORF053 ) was inactivated by a frameshift mutation.In a stability assay, retention of the DR2-bearing minimal SGI1 replicon was examined in the presence of each of these mutant parAB053 -expressing plasmids.Loss of the minimal SGI1 replicon was similarly low when all the three proteins, or both ParA and ParB, were expressed without the 053 protein (Figure 8 B, Supplementary Figure S11 B), whereas the SGI1 replicon was drastically destabilized in the absence of either ParA or ParB.Lack of ParA caused even more rapid plasmid loss than that detected in the negative control, expressing no Par proteins.These observations indicate that the IncC-derived ParA and ParB proteins together stabilize the parS -bearing minimal SGI1 replicon and the 053 protein is not involved in this process.

Sci protein affects regulation of the IncC parAB053 operon
Expression levels of Par systems are fine-tuned and even small changes can cause plasmid destabilization, thus, we hypothesized that the Sci protein exerts its IncC-destabilizing effect by influencing expression of the parAB053 operon.Therefore, we cloned the promoter region of parA (P parA ) into a β-galactosidase test plasmid and measured its expression level in the presence of Type 1 (R16a) and Type 2 (R55 Tn ) IncC plasmids, which provided the ParAB053 proteins in trans .Sci protein was expressed from the IGR placed in p15A-based vectors with appropriate resistance markers.P parA acted as a strong promoter in the absence of IncC plasmids, and its activity was not directly modified by Sci (Figure 8 C).In contrast, the presence of IncC plasmids (R55 Tn or R16a) caused approximately 11-or 8-fold reduction in the promoter activity , respectively .Surprisingly , the presence of Sci significantly reduced the level of repression brought about by IncC plas-mids, as the promoter activity decreased by only 1.5-1.7-foldcompared to the unrepressed state (no IncC plasmid), indicating that the Sci protein somehow alleviates repression of the parAB053 operon.
A similar assay was carried out using a minimal IncC plasmid assembled from the rep region and the entire parAB053 operon with its own promoter region, which provided Par proteins.Surprisingly, the minimal replicon did not reach the repression level achieved by entire IncC plasmids and the Sci protein did not modify the strength of repression.In the next assay, ParAB053 proteins were expressed from a single-copy vector under the control of the arabinose-inducible P araBAD promoter, depriving the par operon of its own regulation.After induction with 1% l -arabinose, the strength of repression was similar to that obtained with the minimal IncC replicon, and Sci had no detectable effect on the repression (Figure 8 D).Summarizing these observations we conclude that autoregulation of the parAB053 operon (presumably by ParB) is modulated by Sci, however, this intervention does not occur through a direct effect on ParAB053 proteins.The fact that maximal repression was only achieved when the full IncC plasmids were present suggests that the proteins expressed from the parAB053 operon are not sufficient to tightly regulate the P parA promoter and that an additional regulatory factor is required, which is missing from the minimal IncC plasmid and certainly encoded on the IncC backbone, but outside the par operon.Our results also indicate that this additional regulator may be the target of Sci, as similar levels of repression were observed in the presence of Sci and the entire IncC plasmids expressing the putative regulator (Figure 8 C, R55 Tn / R16a + Sci), and in the absence of the IncC backbone, independently of Sci (Figure 8 C, miniIncC and Figure 8 D).

SGI1 stability is enhanced by its incompatibility factors in the presence of R55
The data we gathered suggested that the parS sites of SGI1 have a dual role, contributing to both destabilization of IncC plasmids and to stabilization of SGI1 itself, by serving as ParB binding sites for partitioning, whereas Sci appears to be involved only in IncC destabilization via disrupting regulation of the parAB053 operon.To examine the effects of the parS sites and / or the Sci protein on SGI1 stability and its coexistence with an IncC plasmid, KO mutations were generated.We knocked out the parS sites by introducing the same 4-bp insertion into DRs of SGI1-C that was previously applied in the DR2** mutant plasmid pGMY49 (Figure 3 B) and EMSA probe (Figure 5 I), while the Sci protein was eliminated by deletion of IGR_ORF2 ( sci gene).Stability of cohabitation of SGI1 KO mutants and R55 Tn was assayed in approx.22-generations of growth without antibiotic selection, and the proportions of cells retaining one or both elements was determined.Although the output of such assays is significantly biased by the stochastic constitution of the cell populations, since the plasmid can destabilize SGI1 ( 21 ) and vice versa ( 38 , 36 , 18 ), and cells that retain or lose one or both components show different growth rates, our results clearly demonstrated that both DR1 KO and sci deletion caused severe destabilization of SGI1, while DR2 KO had a similar, but much weaker, effect on its stability (Figure 8 E).Compared to wt SGI1-C, DR2 KO slightly decreased the proportion of the otherwise predominant 'SGI1-only' cells and increased that The three KO mutant expression plasmids were analogous to pJKI1128 that expressed all the three proteins (used as positive control), while pJKI625 (no parAB053 ) was used as a negative control.Expression of the par operon was induced with 0.001% L -arabinose.For individual segregation curves, see Supplementary Figure S11 B. ( C ) β-galactosidase assay for measuring the effect of the Sci protein on the promoter activity of P parA in the presence or absence of an IncC plasmid or a minimal IncC replicon.The test plasmid pMSZ1260 carried the P parA -lac Z fusion and Sci was provided in trans from the p15A-based vectors pJKI1149 (Km R ) or pMSZ1164 (Cm R ) (both containing the IGR), which were applied depending on the selection marker of the IncC plasmids R55 Tn (Cm R ) or R16a (Km R ) or the minimal IncC replicon (Cm R ).The minimal IncC replicon (pMSZ1248) contained the rep region with oriV and the parAB053 operon with its own promoter region.Data were obtained from four independent colonies.( D ) β-galactosidase assay for measuring the effect of the Sci protein on the promoter activity of P parA in the presence of ParAB053 proteins expressed from the P araB :: parAB053 cassette placed on the single copy pBeloBac11 vector (Bac-P araB = pMSZ1239, Km R ).To measure P parA activity, the P parA -lac Z -bearing plasmid, pMSZ1260, was used and the Sci was provided in trans from the p15A-based vector pMSZ1264 (Cm R ) carrying the entire IGR.Expression of Par proteins was induced with 1% L -arabinose.Data were obtained from five independent colonies.In the negative control (no parAB053 , no IGR), empty vectors pMSZ1242 and pJKI405 were applied along with the test plasmid pMSZ1260.Paired t-test was used to calculate the statistical significance.***** P < 10 −9 ; **** P < 10 −6 ; *** P < 10 −3 ; n.s., not significant ( P > 0.05).( E ) Segregation of wt, DR1 KO, DR2 KO and sci mutant SGI1-C and R55 Tn plasmid under nonselective conditions.Diagrams show the distribution of R55 Tn and SGI1-C after approx.22 generations growth without selection.Data are means obtained from four independent replicates.For the raw data set, see Supplementary Table S4 .( F ) Excision and copy number of extrachromosomal wt SGI1-C and KO mutants in cells harbouring also R55 Tn .Data of RT-qPCR were normalized to the that of control strain, TG1Nal:: attP SGI1 (marked with 'c-' on the chart), containing a single copy of attB and attP .Paired t-test was used to calculate the statistical significance, n.s., not significant ( P > 0.05).
of 'R55-only' and 'R55 + SGI1' cells to approximately 4%.In contrast, DR1 KO and sci deletion radically reduced SGI1 stability, leading to populations predominated by 'R55-only' cells.In these cultures, the second largest subpopulation was cells with both elements retained (6-7%), while 'SGI1-only' cells were almost (DR1 KO) or entirely ( sci ) absent.Similar distributions were observed with the double and triple mutants, however, the 'SGI1 + R55 ' fraction remained considerable in both cases.These data indicate that lack of parS sites and the Sci protein not only facilitates the co-existence of SGI1 and R55 Tn , but also significantly destabilizes SGI1.
Nevertheless, the difference in the stability between DR1 and DR2 KO mutant SGI1-Cs was bewilderingly large.Since parS sites in DRs differed by only a single, non-conserved position (Figure 5 A), which had no effect on ParB binding (Figure 5 D, F), we presumed that the difference in stability was caused by the locations of the DRs, rather than by their sequence divergence.It has previously been observed that replicationdeficient SGI1 mutants show similarly high instability in the presence of R55 Tn ( 18 ).Since DR1 is located close to the promoter of the S004-repA operon, which is responsible for replication and copy-number control of excised SGI1, we supposed that the KO mutation negatively affects the regulation of SGI1 replication.Therefore, we conducted an RT-qPCR assay to assess copy numbers of wt SGI1-C, sci , DR1 KO and DR2 KO mutants in the presence of R55 Tn (Figure 8 F).The results showed that DR2 KO and sci did not significantly effect SGI1 copy number, while confirming that DR1 KO reduced it to approx.one copy per cell.These findings are consistent with our previous observation that this low copy number is accompanied by extreme instability of SGI1 ( 18 ), and also indicate that integrity of the parS site in DR1 is necessary for the normal replication, and consequently for the stability, of SGI1.

Discussion
Conjugative plasmids and IEs are key factors in the distribution of genes conferring antibiotic and heavy metal resistance, as well as numerous other adaptive traits, among diverse host bacteria.Plasmids and IEs apply similar mechanisms to guarantee their stable maintenance, such as partitioning and postsegregational killing.Unlike plasmids, IEs ensure their vertical transfer by integration into the host chromosome, however, for horizontal transfer they have to excise, relinquishing this quiescent state, which can lead to sibling cells that have lost the element.
SGI1-family IMEs that are mobilized by IncA and IncC plasmids apply several refined mechanisms to reduce the risk of being lost when they coexist with the helper plasmid, which triggers their excision leading to segregation ( 21 , 24 , 36 , 18 , 37 ).Besides the TA system ( 24 ), another important function involved in SGI1 stabilization is the transient plasmid-like replication of the excised element ( 18 ).Further, SGI1 destabilizes the helper plasmid, which also contributes to its own stable maintenance.There is strong evidence that SGI1 replication both enhances its persistence in cells where the island coexists with the IncC helper plasmid, and appears is a key factor in plasmid destabilization.Incompatibility between SGI1 and its mobilizing IncC and IncA plasmids has been reported ( 38 , 36 , 18 , 68 , 39 ), and it is proposed that SgaDC, the SGI1encoded master regulator may have a central role in this phenomenon.Although the exact mechanism of plasmid expul-sion is unknown, interference with plasmid maintenance functions (replication, partitioning) has been suggested.Several possible ways have been proposed, such as titration of the host replication protein, DnaA, or perturbation of plasmid partitioning or post-segregational killing mechanisms, mediated by an unknown factor, the amount of which is increased by the elevated copy number of replicating SGI1 ( 36 , 18 , 37 ).
In this work, we present evidence that SGI1 intervenes in plasmid maintenance via two distinct factors.Deletion mapping and stability assays (Figures 1 -3 ) revealed that the 31bp direct repeats, DR1 and DR2, located upstream of ORFs S004 and traH S , respectively, and the IGR between ORFs S014 and S015, comprise elements responsible for incompatibility with IncC plasmids.One copy of the 31-bp DRs (DR2) and the IGR fragment, without any other SGI1-derived regions, exerted strong incompatibility with R55 Tn when both were present in either cis or trans in ∼15 copies.The joined DR2-IGR fragment resulted in a much lower level of IncC plasmid loss when present on a single copy plasmid, while it led to extremely rapid plasmid segregation in a high-copy state ( > 300 per cell) compared with that mediated in the 15-copy state (Figure 4 A).Analysis of the incompatibility exerted by the two elements separately indicated that DR2-promoted plasmid loss was robustly copy-number dependent, while that of the IGR was not (Figure 4 B, D), suggesting that the two elements act via different mechanisms.Furthermore, DR2 and IGR amplified the effects of one another as they induced much weaker incompatibility alone than together (Figure 4 ).
In the 5 -part of the DR1 and DR2 repeats, a 22-bp IR motif was identified, with striking similarity to an imperfect IR downstream of the putative -10 promoter box preceding the parAB053 operon in the IncC plasmid, R55 (Figure 5 AB).ParABS systems are generally autoregulated by ParA, ParB or both via binding to a parS -like sequence close to the promoter of the parABS operon ( 45 ) and their parS sites are often consists of one or more IR motifs.No parS sites have yet been identified in the IncC parABS system, however, the third ORF, ORF053 , is necessary for stable plasmid maintenance, and it was previously hypothesized that the parS site may be within ORF053 ( 51 ).Our findings, demonstrating that SGI1 carries two copies of an IR motif, which are involved in incompatibility with R55 and resemble the IR present in the promoter region of the parAB053 operon, raise the possibility that these IRs act as parS sites for IncC plasmids, through which they can generate centromere-mediated incompatibility.A thorough sequence search revealed that R55 contains 13 additional IR motifs scattered through its conserved backbone that are similar to those found in the P parA promoter and in SGI1.Most of these IRs are located in intergenic regions near R55_13, R55_14, R55_24 / parB -like regulator, R55_29, parAB053 , R55_55, R55_74, R55_109, R55_170 / DNA helicase and R55_174 / eexC genes, while four copies are within the ORFs, R55_6, R55_08, R55_91 and R55_130 / xerC .Alignment of the two SGI1-and 14 IncCderived IRs showed a fully conserved symmetrical 14-bp core sequence flanked by less conserved AAAC and GTTT motifs (Figure 5 A-C).EMSA experiments with three selected putative parS sites (one representing the consensus of the 16 IR sequences and also identical to that present in DR1 of SGI1, the others from DR2 of SGI1 and the promoter region of R55 parAB053 operon) proved that these motifs are binding sites for the ParB protein expressed by the 2nd ORF of parAB053 operon (Figure 5 D-F).Interestingly, the fully conserved 14-bp palindromic core sequence, TTTC AC ATGTGAAA, was not bound by ParB (Figure 5 I) indicating that the flanking AAAC and GTTT motifs, and particularly the highly conserved GT bases in the right side flanking sequence, are important for ParB recognition.ParB of IncC plasmids contains a ParB and a KorB domain ( 27 ).The prototype KorB protein, KorB encoded by the IncP α plasmid RP4 acts as ParB in the partitioning system of this plasmid, and also as a global regulator of housekeeping genes.A helix-turn-helix (HTH) motif was predicted in KorB, responsible for binding to 12 palindromic 13-bp operator sequences ( 69 ,70 ), similarly to other KorB homologs and ParB-family members ( 71 ).KorB can form tetramers in solution ( 72 ) and a C-terminal dimerization domain has been identified ( 73 ).Although similar details are not available for IncC ParB, structural modelling using SwissModel suggested that it also has an HTH near its Nterminus.Further, EMSAs with ParB gave unusually highly shifted bands with the three parS probes (Figure 5 D-F), suggesting that IncC ParB behaves similarly to several other ParB proteins that also cause high shifts and have been shown to spread from the parS site along the DNA ( 74 ,75 ), potentially forming higher order complexes.A 4-bp insertion in the centre of the parS site in DR2 abolished ParB binding, indicating that ParB may bind to the IR of parS as a symmetric dimer or tetramer form.The same 4-bp insertion or removal of half of the parS site in DR2 eliminated incompatibility with R55 Tn (Figure 3 B, C), supporting this hypothesis and indicating that ParB-parS binding is required for SGI1-mediated plasmid destabilization.
A Genbank search with the 22-bp parS sequence present in the P parA promoter revealed that the majority of IncC plasmids carry 12-14 copies of the conserved 14-bp core sequence of the parS sites, several plasmids have 9 copies (i.e.CP050727, CP029123 and CP050727) and plasmid pV266a (LC056472) carries only 7 copies.IncA plasmids identified to date ( 76 ) contain 10-14 copies and no IncA or IncC plasmids were found to have more than 14 or less than 7 copies, except pRAx (5 copies), a deletion derivative of pRA1, which has 14 copies ( 77 ).Interestingly, several plasmids not belonging to the IncA or IncC groups but sharing some parts of their backbone (i.e.LC501639, CP040595 and KX832927) also contain 5-6 copies of IncC parS .Many strains of Proteus , Salmonella , Klebsiella , E. coli , Shewanella and Morganella carry two copies of IncC parS sequence in their chromosome, mostly within integrated SGI1-family elements, however, Providencia stuartii FDAARGOS_1040 and Vibrio cholerae SA3G contain 3 and 4 chromosomal copies that do not appear to be associated with SGI1-related elements.
Helper-induced transient replication of SGI1 is established as a major factor in IncC plasmid destabilization ( 36 , 18 , 37 , 68 ).Our data demonstrate that increased parS copy number accelerates plasmid loss (Figure 4 B, D), consistent with a model where helper-plasmid-induced replication leads to higher SGI1 copy number (with two parS sites per copy), thus providing 14-16 extra parS copies, comparable to the number of parS sites in the plasmid itself, and this increased number of parS sites contributes to plasmid destabilization.The dose-dependence of incompatibility can explain why SGI1 replication appears to be a key factor in plasmid elimination.As Par systems are sensitive to such perturbations, doubling of parS site numbers will impair normal plasmid partitioning ( 78 ), likely due to ParB protein titration, or competition for attachment sites or other partitioning proteins ( 41 ,45 ).Thus, the effects of SGI1-derived parS sites on IncC stability are similar to, and can be treated as, a type of centromere-mediated incompatibility ( 45 ).
Nevertheless, introduction of approx.15 extra parS copies led to much weaker incompatibility than that observed with the full wt SGI1 (compare Figure 4 B and Figure 1 B), indicating that parS sites are just one factor contributing to incompatibility.Deletion mapping revealed that a second factor is expressed from the IGR between ORFs S014 and S015.After ruling out the role of IGR_ORF1 and a putative small RNA overlapping IGR_ORF2, a 51 aa (5.9 kDa) protein named Sci was identified as the parS -mediated incompatibility-enhancing factor.Sci encoded by IGR_ORF2 appears to be a simple helical protein with a C-terminal 13 aa disordered tract ( Supplementary Figure S10 ), which seems unimportant for its function (pJKI1137, Figure 6 B).Sci alone exerted weak incompatibility, which was slightly increased when expressed from a high-copy plasmid (Figure 4 B, D), however, Sci significantly enhanced the incompatibility exerted by the DR2-derived parS site when expressed from its own promoter or the strong P cat promoter (Figures 4 C  and 7C ).
Removal of the mobilizing helper plasmid, which triggers excision from the chromosome and hereby causes SGI1 destabilization, appears to be an adaptive strategy for IMEs like SGI1, to ensure their stable maintenance.Further, IMEs require the helper elements for their horizontal distribution, hence, transient stabilizing mechanisms are needed during periods of coexistence with their helpers.ICEs face similar challenges when they excise before conjugal transfer and exist in an extrachromosomal state, despite not being destabilized by other elements in this context.Plasmid-like replication is an important stabilizing mechanism for IEs ( 44 ,18 ), but cannot provide sufficient stability, as demonstrated for locked-out int mutant pSGI1 ( 36 ) or SXT / R391 elements ( 44 ).Thus, further stabilization mechanisms are applied.Transient expression of a parMRC system, srpMRC , under the control of the FlhDC-family activator, SetDC, significantly contributes to stable maintenance of SXT-family elements.This means of expression control also ensures that the srpMRC system functions only when the ICE is excised and replicates as a plasmid.Although the occurrence of Par systems in ICEs does not appear to be unusual ( 44 ,43 ), similar systems in IMEs have not previously been identified.Similar to SXT, SGI1 also requires transient stabilization when in an extrachromosomal state due to helper-induced excision ( 21 , 24 , 18 ).Although SGI1 and its relatives do not include a complete Par system, they do contain functional parS sites, mimicking those of IncA and IncC plasmids, by which this issue can be solved by exploiting the Par system of the plasmids.Here we demonstrate that the otherwise unstable SGI1-derived minimal replicon carrying a parS site located in DR2 became significantly more stable if R55 parAB053 was expressed in trans (Figure 8 A).This stabilizing effect was manifested only if both ParA and ParB were expressed (Figure 8 B), while 053 protein did not contribute to the effect.Consistently, KO mutation of parS sites in SGI1-C significantly reduced SGI1 stability alongside coresident R55 Tn under nonselective conditions (Figure 8 E).Together, these data strongly suggest that SGI1 uses parS sites to both expel helper plasmids and hijack the plasmid-encoded partitioning system to increase its own stability through active partitioning.
While the function of SGI1-borne parS sites is relatively clear, the mode of action of Sci protein in incompatibility and SGI1 stability is rather puzzling.The fact that the incompatibility exerted by a parS site was significantly amplified by Sci expression (Figures 4 C and 7 C) suggests that Sci also influences plasmid partitioning, for example, via modulating regulation of the parAB053 operon or ParB-parS binding.Expression of the parAB053 operon from a minimal IncC replicon reduced expression from P parA (Figure 8 C).Since the -10 box of the P parA promoter overlaps with the parS site to which ParB binds, we conclude that ParB is a repressor of its own synthesis, as established in many ParABS systems ( 45 , 79 , 80 ), however, Sci had no effect on expression from P parA . in this minimal system A similar result was obtained when ParAB053 proteins were expressed from the inducible P araBAD promoter (Figure 8 D), which released the operon from its original regulation, ruling out feedback effects between the β-galactosidase measuring plasmid and the ParAB053-producer plasmid.In contrast, when an entire IncC plasmid was present, stronger repression was observed, which was clearly alleviated by Sci (Figure 8 C).These findings indicate that correct regulation of P parA involves an additional IncC-plasmid-derived factor, encoded outside the parAB053 operon, and Sci presumably exerts its effect on this putative regulator.This effect may occur, for example, via competition of Sci with the unknown regulator for its binding site.However, direct DNA binding by small helical protein such as Sci, which lacks any known domains, seems unlikely.A more realistic hypothesis is that Sci acts via protein-protein interactions by reducing the binding specificity or strength of the regulator to its target DNA, or through interfering with other activities of this factor involved in par -regulation.Anyway, if Sci releases strict repression of the otherwise strong P parA promoter, the increased expression of ParB can easily destabilize the single-copy helper plasmid.Plasmid destabilization by overexpression of the cognate ParB or ParR proteins has been described (81)(82)(83), and was also observed in our experimental system, where the minimal SGI1 replicon, rep SGI1 , could be stabilized only if ParAB053 expression was induced by very low (0.001%) larabinose concentration, while stronger induction always led to rapid loss of the SGI1 replicon.Nevertheless, under natural conditions, elevated ParB expression due to the effects of Sci intervention on a single-copy IncC plasmid, may be optimal to increase the partitioning efficiency of the 7-8 copies of free plasmid-like SGI1, further enhancing SGI1 stability.This model is consistent with the observation that sci mutant SGI1-C stability was clearly decreased under nonselective conditions compared to wt SGI1-C (Figure 8 E).In light of these results, we favour the hypothesis that Sci disrupts control of the parAB053 operon by influencing a yet unknown co-repressor of the P parA promoter, leading to elevated expression of Par proteins, which assists in SGI1 partitioning, while simultaneously perturbing that of the IncC plasmid.However, regulation of the IncC ParABS system requires further exploration to facilitate better understanding of these interactions.Therefore, investigations to examine parAB053 operon control mechanisms and search for the Sci protein target are in progress.
Inactivation of SGI1-C parS sites decreased its stability, however, the segregation rate of the two DR KO mutants and their effect on the stability of the coresident IncC plasmid under nonselective conditions proved unexpectedly different (Figure 8 E).The 4-bp insertion introduced into the parS site of DR2 moderately decreased SGI1 retention and increased the fraction of R55-only and SGI1 + R55 cells relative to wt SGI1-C.In contrast, the same mutation in DR1 caused a dramatic change in SGI1 stability, as the proportion of SGI1-only cells dropped to 1.7%, while R55-only cells dominated the population, and the ratio of R55 + SGI1 cells also exceeded 7%.The low stability of DR1 KO mutant SGI1-C can be explained by its reduced copy number (approx. 1 / cell, Figure 8 F), as observed in rep mutants ( 18 ).This change in copy number indicates that parS insertion mutation in DR1 impairs normal replication control of SGI1.The facts that parS in DR1 is 25 bp upstream of the core AcaCD-binding motif of the P S004 promoter and that ParB cannot bind KO mutant parS (Figure 5 I) strongly suggest that ParB somehow participates in regulation of the S004-repA operon.A similar genetic constitution is found in the promoter region of the traG S H S operon, where the parS site is only 10 bp upstream of the AcaCD binding site, however, expression of the traG S H S operon presumably does not influence replication.Thus, DR2 KO mutant stability may be changed solely because of the loss of one copy of parS , while the copy number in this mutant remained unaffected (Figure 8 E, F).Interestingly, an analogous constitution occurs in the promoter regions of four AcaCD-regulated operons in the conserved IncC backbone ( 30 ).In these promoters (the first ORFs of the operons are R55_9 / vcrx012, R55_55 / vcrx036, HS904_RS00445 / vcrx068 and traF ) similar crosstalk can occur as that observed for P S004 of SGI1.Since ParB often acts as global regulator of various genes involved in the maintenance functions of several plasmids ( 84 ), this may also be the case for ParB and IncC plasmids.Thus, investigations on the presumed multiple roles of ParB are ongoing.
SGI1 appears to efficiently exploit helper plasmid features via an intricate control circuit elucidated in recent studies ( 21 , 36 , 18 , 37 ).The first signal is likely the IncC-entry-induced SOS response, which triggers expression of the SGI1-encoded activator, SgaDC ( 68 ).Although SgaDC is constitutively expressed ( 35 ) from quiescent SGI1, the amount produced is insufficient to induce SGI1 excision and replication.When the plasmid enters a cell containing SGI1 in its chromosome, it encounters a basal level of SgaDC, which likely quickly increases, due to SOS induction.This may activate expression of AcaB, the activator of the plasmid acr-acaCD operon ( 32 ).The acaB gene is part of the AcaCD regulon and its promoter, like all other AcaCD-responsive promoters, is also activated by SgaDC ( 37 ).Induction of the acr-acaCD operon leads to elevated levels of the AcaCD activator, which induces expression of genes encoding the plasmid-borne conjugation apparatus, as well as contributing to triggering SGI1 excision ( 30 ,21 ) and replication ( 18 ), however, SgaDC appears to have major role in the latter processes ( 18 ,37 ).Due to the positive feedback loop between acaB and acr-acaCD and SOSmediated induction of sgaDC , the concentrations of both activators will be sufficient for SGI1 excision and generation of 7-8 copies / cell.Meanwhile, all AcaCD-controlled conjugation genes encoded by the helper plasmid and SGI1 ( traN S , and traG S H S ) are also induced.Hence, everything is in place for the lateral transfer of SGI1, however, retention of the island is riskier in this state.To prevent SGI1-free sibling cells arising, the higher copy number of SGI1 achieved by transient plasmid-like replication, together with expression of the TA system, increase SGI1 stability, however these phenomena appear insufficient over evolutionary time scales, as SGI1 applies additional protective measures.Hijacking the parABS system of IncC helper plasmids using 'stolen' parS sites and a protein factor modification of expression from the parAB053 operon is an elegant 'two birds with one stone' solution.Expelling the helper plasmid and increasing SGI1 stability by active partitioning is a logical way to solve the stability issue.SGI1, which bears two parS sites, may be more efficiently dispersed during cell division using parABS system proteins of the helper.Simultaneously, this upsets the partitioning of the helper plasmid by competing for ParB, ParA, attachment sites, or other partitioning factors, and through the effects of the Sci protein.In this way, SGI1 negates the need for a self-encoded Par system and schedules its partitioning and incompatibility as required.
SgaDC is considered a key element in IncC plasmid destabilization ( 36-38 , 68 , 39 ) and is indispensable for induction of SGI1 replication and maintenance of 7-8 copies per cell.
Here we demonstrate that the presence of increased amounts of identified incompatibility factors due to SGI1 replication strengthens its incompatibility with R55.Thus, one can conclude that SgaDC may act only indirectly on incompatibility via its effect on replication control.Consistent with this model, no destabilization of the IncC plasmid, pRMH760, by SGI1-K variant lacking the sgaDC operon, due to a deletion from within traN S (S005) to within S009, was observed ( 38 ,39 ), despite it possessing both parS sites and the sci gene.Further, sgaDC -deleted mutants cannot replicate normally ( 36 , 18 , 37 ), therefore, the weak or undetectable incompatibility of SGI1-K can be explained by its low copy number, and consequent lower amount of Sci protein and fewer additional parS copies (Figure 8 E, F), as was also observed with our replication deficient deletion mutants ( 1-5, Figure 1 ).Although we did not detect similarly strong incompatibility of an SgaDC-expression vector as recently reported ( 39 ), we also observed that expression of SgaDC in trans and the presence of the entire IncN2 / N3-related replicon (including one parS site) on an approximately 15-copy plasmid caused low-level but detectable R55 Tn loss (Figure 2 C, D), even though RepA expression itself exerted no detectable incompatibility.These observations require further investigation, but we believe that the incompatibility is not primarily caused by components of SGI1 replication machinery themselves (RepA, S004, SgaDC), but rather by reinforcement of the effects of incompatibility factors due to the effect of increased SGI1 copy number following its normal replication.
Hence, SGI1 takes advantage of IncC (and possibly IncA) plasmids in several ways, utilizing the plasmid AcaCDdependent regulatory systems and transfer apparatus.Here we report an additional form of parasitism on IncC plasmids by SGI1 via exploitation and disruption of their partitioning system.Although elements from several other IME-families are also mobilized by IncC plasmids ( 30 ,85 ), the fact that SGI1family elements are far more common suggests that this complex behaviour is a winning evolutionary strategy.

Figure 1 .
Figure 1.Incompatibility of wt SGI1-C and its deletion mutants with the IncC plasmid R55 Tn .( A ) Schematic map of the conserved SGI1 backbone.Annotated ORFs S00 1 -S044 are indicated by colour-coded arrows: green -recombinase; orange -replication; blue -transcription regulator; yellow -T4SS components; purple -conjugation initiation, grey -TA system; white -unknown function.Names of genes with known functions or identified homologs are indicated.Abbreviations: x -xis , C and D -flhC SGI1 and flhD SGI1 , sgiT and sgiA -toxin and antitoxin genes of the TA system, B -mpsB .Genes of unknown functions are numbered according to the original numbering (e.g.'8 refers to S008, etc.).Terminal direct repeats are shown as black bo x es, green and red vertical lines indicate oriV and oriT , respectively.Purple vertical lines indicate the 31-bp imperfect direct repeats (DR1 and DR2), and green broken arrows represent AcaCD / SgaDC-responsive promoters.Positions of In104 (carrying the AR genes aadA2 , floR , tetG , pse-1 , sul1 -in the case of protoype SGI1) and eight deletions ( 1-8) generated in SGI1-backbone are indicated (the map is drawn to scale).( B ) Dynamics of R55 Tn loss from TG1Nal strains harbouring wt or one of the 1-8 deletion mutant SGI1-C.Graph shows the percentage of R55 Tn+ cells grown under selection for SGI1 but not for R55 Tn through 5 passages.Each passage represents approx.20 generations.The % values are means of four biological replicates (for the individual segregation curves with standard deviations, see Supplementary FigureS1).(C) Proportion of R55 Tn+ cells after the 5th passage (100 generations growth).Four replicates from the 5th passage were pooled, spread on LB plates selective for SGI1, and retention of R55 Tn was individually tested by replica plating.Number of total and R55 Tn+ colonies tested are listed and the plasmid loss is indicated as % of the R55 Tn − colonies.If no R55 Tn − colonies were found, the threshold limit was calculated as 1 / total colony count.If high-frequency loss of R55 Tn was observed, the percentage of R55 Tn+ cells was calculated as mean of the Cm R / total cell titers of the four parallels measured in the 5th passage.

Figure 2 .
Figure 2. Destabilization of R55 Tn by the 5 -part of SGI1-C-d1 variant and its further truncated derivatives.( A ) Map of the p15A-based plasmids containing parts of SGI1-C-d1.The 5 -part of SGI1-C-d1 present in pJKI670 includes SGI1 from DRL to the Nco I site of the mpsB gene, but lacks the S005-S012 region due to the 'd1' deletion occurred between DR1 and DR2 repeats.Deletion derivatives of pJKI670 were generated by cloning steps using the restriction enzymes indicated below the plasmid map (Ac -Acc I, P -Pst I, Bs -Bsa BI, Pa -Pas I, Nc -Nco I, B -Bam HI).'d1' indicates the spontaneous deletion removing the 3474-13401 bp region of SGI1, other symbols are as in Figure1A.( B ) Dynamics of R55 Tn -loss from TG1Nal strain harbouring R55 Tn along with pJKI670, pGMY28, pGMY20, pGMY14 or pJKI88 (empty vector used as negative control).Graph shows the mean percentage of the R55 Tn+ cells in biological replicates grown under selection for the p15A plasmids (Km), but without selection for R55 Tn , through 5 passages as described previously.For individual segregation curves see Supplementary FigureS2A.( C ) Dynamics of R55 Tn loss from TG1Nal strains harbouring one of the plasmids expressing repA SGI1 (pGMY9), sgaDC (pGMY6) or containing the entire rep region of SGI1 (pGMY18).The empty e xpression v ector, pJKI391, w as used as a negativ e control.T he data presented on the graph w ere obt ained in similar st abilit y assa y s described in part B. For individual segregation curves see Supplementary FigureS2 B ( D). Proportion of R55 Tn+ cells after the 5th passage (100 generations growth) in the cultures assa y ed in parts B and C. Data were obtained and the threshold limit was calculated as described in Figure1C.

Figure 3 .
Figure 3. Fine mapping of incompatibility factors in the surroundings of S0 13-S0 14 ORFs.Maps of SGI1 fragments cloned in the p15A-based vector pJKI88 are shown in scale.Charts show the dynamics of R55 Tn -loss during 5 passages.For comparison, curves obtained with pJKI670 (positive control) and the empty vector pJKI88 (negative control) are also shown.(A ) Effects of replacement of P S004 -DR1 with the P S012 -DR2 and deletion of DR2 or IGR region on R55 Tn -loss.( B) Effects of replacement or insertion mutations in DR2, S013 and S014 on R55 Tn -loss.Positions of mutations are indicated by vertical red lines.Replaced or inserted bases are in red.DR2* -the last 16 bp of DR2 is replaced, DR2** -the Pci I site of DR2 was filled in leading to 4 bp insertion, 13* and 14* -frameshift mutations generated by insertion of an 8-mer Xho I linker into the Hinc II or the BsaA I site of S013 and S0 14, respectively .( C ) P roportion of R55 Tn+ cells in the 5th passages assa y ed in parts A and B (determined as described in Figure 2 D).For individual segregation curves see Supplementary FigureS3.

Figure 5 .
Figure 5. Analysis of putative parS sites identified in R55 backbone and the 31-bp DRs of SGI1.( A ) Comparison of DR1 and DR2 sequences.Symmetry of the 22-bp in v erted repeat motifs corresponding to the putative parS sites (red box) is shown by arrows.Asterisk indicates the mismatching base in the IR. ( B ) Upstream region of R55 parA gene.Predicted promoter boxes ( −35, −10), ribosome binding site (SD) and the putative parS site (red box) are indicated (symbols are as in panel A).Coordinates shown above the sequence are according to the published R55 sequence (Acc.no.: JQ0 1 0984.1).( C ) Alignment and sequence logo for 16 putative parS sites identified in R55 and SGI1.R55-derived sequences are named according to the nearest ORF.In the consensus, red and blue indicate > 90% and > 60% conserved bases, respectively.The fully conserved symmetrical core sequence is shown by brackets.(D-G) EMSAs for detection of DNA-ParB complex.FAM-labelled probes were: ( D ) the putative parS site located in the intergenic region near ORF R55_174, ( E ) the putative parS preceding the parAB operon, ( F ) the entire DR2 of SGI1 and ( G ) the empty multicloning site (negative control) flanking the parS sequences in the other three probes.The parS R55_174 sequence represents the IR motif present in DR1 of SGI1 and also the consensus of the 16 IR motifs.Amount of ParB protein was increased from 0-900 ng (lanes 1-7: 0, 30, 75, 150, 300, 600 and 900 ng).Brackets indicate shifted comple x es, and filled arro wheads point to the unbound probe.( H ) Binding specificity tests f or three parS -cont aining probes.Binding specificit y was confirmed in the presence of 25-fold unlabelled probes.Samples in all lanes contained 3 ng of labelled probes.Lanes '-', no ParB protein; lanes '+' and 'c', 300 ng parB protein; lanes 'c', 25-fold unlabelled DNA fragments were added to the respective binding reaction run in lanes '+'.( I ) EMSA for detection of ParB binding to wt and insertion mutant DR2, and to the conserved symmetrical 14-bp core sequence of the putative parS sites.Probe 'DR2**' contained 4-bp insertion at the centre of the 22-bp IR of DR2, and probe ' parS core' carried the 14-bp symmetrical sequence shown by bracket in panel C. Amount of ParB protein was 0, 150 or 450 ng.In panels H and I, open and filled arrowheads point to shifted comple x es and unbound probes, respectively.

Figure 6 .
Figure 6.Localization of the incompatibility factor in IGR between ORFs S014 and S015.( A ) Schematic map of the IGR sequence (drawn to scale) present in the test plasmid pGMY64.Coordinates are according to the published SGI1 sequence AF261825, with the modification that the numbering starts from the 5 end of DRL.Arrowheads indicate the predicted -35 (filled) and -10 (open) promoter boxes.Blue rectangles show the predicted Shine-Dalgarno (SD) bo x es.Short and long green arrowheads indicate the 8 / 8-bp and the 18 / 19-bp IR motifs, respectively.Thin lines show the missing regions compared to pGMY64.Double red bo x es represent rrnB T1 and T2 terminators, and the dark blue arrowhead indicates the P cat promoter (these are not to scale).( B and C ) Dynamics of R55 Tn loss from TG1Nal strains harbouring test plasmids with 3 -and 5 -truncated IGRs.The curve obtained with pGMY64 is also shown for comparison.For individual segregation curves, see Supplementary Figure S6 .( D ) Proportion of R55 Tn+ cells in the 5 th passages assa y ed in parts B and C (determined as described in Figure 2 D). ( E-G ) Dynamics of R55 Tn loss from TG1Nal strains harbouring test plasmids supplemented with rrnB T1T2 terminator or the rrnB T1T2-P cat cassette.Data of unmodified test plasmids pJKI1144, pJKI1154 and pJKI1155 presented on panel C are also shown for comparison.For individual segregation curves see Supplementary Figure S7 .( H ) Proportion of R55 Tn+ cells in the 5 th passages assa y ed in parts E-G (determined as described in Figure 2 D).

Figure 7 .
Figure 7. Identification of the incompatibility factor encoded in IGR. ( A ) Schematic map of the functional part of IGR_ORF2 in the test plasmid pJKI1153.The green arrowheads represent the 8 / 8 bp IR motif.Coordinates and other symbols are as in Figure 6 A. The horizontal thick lines represent the rele v ant part of the mutated test plasmids.The 4 or 1-bp insertions are indicated by arrowheads, and the samesense codon replacements are highlighted by orange.Note that the replacement in pJKI1160 eliminated the left arm of the 8 / 8-bp IR. ( B ) Dynamics of R55 Tn loss from TG1Nal strains harbouring the plasmids with modified IGR_ORF2 3' (panel A).The curve obtained with the unmodified construct pJKI1153 (Figure 6 E) is also shown for comparison.For individual segregation curves see, Supplementary Figure S8 A. ( C ) Dynamics of R55 Tn loss from TG1Nal strains containing the components of the trans system in different combinations.DR2 -the DR2-bearing p15A plasmid pGMY66, IGR_ORF2 -the pKK223-3-based e xpression v ector pJKI1161 producing Sci protein (during passages, protein expression occurred by leaking of the P tac promoter without IPTG induction).The empty vectors, pJKI88 (p15A, c1-), and the pKK223-3 deriv ed e xpression v ector pJKI1130 (pBR322, c2-), w ere used as negative controls.For individual segregation curves see Supplementary Figure S8 B. ( D ) Proportion of R55 Tn+ cells in the 5th passages assa y ed in parts B and C (determined as described in Figure 2 D).

Figure 8 .
Figure 8. Functional analysis of DRs and the Sci protein produced by SGI1.( A ) Stabilization of the SGI1-derived minimal replicon containing DR2 or DR2 + IGR in the presence of the parAB053 operon of R55.Graph shows the dynamics of the loss of the SGI1-derived replicons supplemented with DR2 (pMSZ1200) or DR2 + IGR (pMSZ1175b) from TG1Nal strain with or without the parAB053 operon (pJKI1128 or pJKI625, respectively) under the control of P araB promoter.Expression of the par operon was induced with 0.001% L-arabinose.The empty expression vector pJKI625 was used as a negative control.For individual segregation curves, see Supplementary Figure S11 A. ( B ) Stability of the SGI1-derived replicon carrying DR2 in the presence of the KO mutants of parAB053 operon expressing no ParA or ParB or 053 proteins.Graph shows the dynamics of the loss of SGI1-derived minimal replicon, containing an IncC parS site present in DR2 (pMSZ1200, from TG1Nal strain, where only two of the three proteins of the parAB053 operon were expressed.ParB + 053 without ParA were expressed from pMSZ1263, ParA + 053 without ParB were expressed from pMSZ1272, and ParA + ParB without 053 were expressed from pGMY86.The three KO mutant expression plasmids were analogous to pJKI1128 that expressed all the three proteins (used as positive control), while pJKI625 (no parAB053 ) was used as a negative control.Expression of the par operon was induced with