Multiple Pathways of Plasmid DNA Transfer in Helicobacter pylori

Many Helicobacter pylori (Hp) strains carry cryptic plasmids of different size and gene content, the function of which is not well understood. A subgroup of these plasmids (e.g. pHel4, pHel12), contain a mobilisation region, but no cognate type IV secretion system (T4SS) for conjugative transfer. Instead, certain H. pylori strains (e.g. strain P12 carrying plasmid pHel12) can harbour up to four T4SSs in their genome (cag-T4SS, comB, tfs3, tfs4). Here, we show that such indigenous plasmids can be efficiently transferred between H. pylori strains, even in the presence of extracellular DNaseI eliminating natural transformation. Knockout of a plasmid-encoded mobA relaxase gene significantly reduced plasmid DNA transfer in the presence of DNaseI, suggesting a DNA conjugation or mobilisation process. To identify the T4SS involved in this conjugative DNA transfer, each individual T4SS was consecutively deleted from the bacterial chromosome. Using a marker-free counterselectable gene deletion procedure (rpsL counterselection method), a P12 mutant strain was finally obtained with no single T4SS (P12ΔT4SS). Mating experiments using these mutants identified the comB T4SS in the recipient strain as the major mediator of plasmid DNA transfer between H. pylori strains, both in a DNaseI-sensitive (natural transformation) as well as a DNaseI-resistant manner (conjugative transfer). However, transfer of a pHel12::cat plasmid from a P12ΔT4SS donor strain into a P12ΔT4SS recipient strain provided evidence for the existence of a third, T4SS-independent mechanism of DNA transfer. This novel type of plasmid DNA transfer, designated as alternate DNaseI-Resistant (ADR) mechanism, is observed at a rather low frequency under in vitro conditions. Taken together, our study describes for the first time the existence of three distinct pathways of plasmid DNA transfer between H. pylori underscoring the importance of horizontal gene transfer for this species.


Introduction
Helicobacter pylori is a highly motile, microaerophilic, Gramnegative bacterium, resident in the gastric mucus layer of about 50% of the human population. Infection with H. pylori is a major cause of gastroduodenal disease, including chronic active gastritis, peptic ulcer disease, mucosa-associated lymphoid tissue (MALT) lymphoma and gastric carcinoma [1,2]. A remarkable feature of H. pylori is its panmictic population structure, reflected by an extreme genetic heterogeneity, possibly resulting from frequent recombination events after import of small pieces of foreign DNA from other H. pylori strains during persistent or transient mixed infections [3][4][5][6]. Such an efficient DNA exchange has been attributed to the natural transformation competence of H. pylori [7], mediated by the comB type IV transport system [8][9][10], which is actually stimulated by DNA damage to trigger genetic exchange [11].
Besides natural transformation, bacterial conjugation is a further possible mechanism of lateral DNA transfer. Interestingly, about 50% of H. pylori isolates carry cryptic plasmids ranging between 2 and 100 kb in size [12]. Some isolates even carry multiple plasmids of different size; however, the role of these plasmids for H. pylori is not well understood. Chromosomal integration and excision of plasmid DNA after transfer from a donor into a recipient strain might be an alternative way to generate genome rearrangements [13].
Most H. pylori plasmids either replicate via the ''rolling circle'' mechanism [14,15], or carry direct sequence repeats, so-called ''iterons'', and replicate via the theta mechanism [16] (for review see [17]). A DNaseI-resistant, conjugation-like bidirectional chromosomal DNA transfer between H. pylori has been reported, but the mechanism has not been explored [18]. Conjugative transfer of indigenous plasmids between H. pylori strains, or even more unrelated bacterial species, has not yet been demonstrated.
Plasmids are generally classified as conjugative (autotransmissible) or mobilisable (transmissible only in the presence of a helper conjugative plasmid). In contrast to conjugative plasmids, which contain all the necessary genetic information to catalyse conjugative DNA processing and DNA transport, mobilisable plasmids lack part of this machinery. Mobilisable plasmids typically have an origin of conjugative transfer (oriT) and code for proteins involved in conjugative processing of DNA, such as oriT-specific relaxases and nicking accessory proteins [19].
We have initially characterized two cryptic H. pylori plasmids of 10.9 and 18.2 kb in size named pHel4 and pHel5, respectively [13], and recently pHel12 [20], a 10.2 kb plasmid. They were assigned to the group of theta plasmids. In pHel4 and pHel12, a putative mob region was identified, showing best homologies to proteins MbeA, MbeB, MbeC and MbeD of colicinogenic plasmids, such as pColE1 [13,21]. The MbeA protein of the ColE1 plasmid is an atypical relaxase, because it lacks two conserved histidine residues in the third conserved amino acid sequence motif (motif III) [22], present in most relaxases. MbeA, which reveals best sequence homology in the three conserved amino acid motifs (motif I, motif II and motif III) to the pHel4 relaxase [13], has been verified as a relaxase necessary for plasmid ColE1 mobilisation [23]. H. pylori plasmid pHel4 is 10.9 kb in size. In addition to its encoded replicase RepA, the plasmid contains a putative microcin operon, a putative mobilisation region (mobA) and a number of cryptic open reading frames (ORFs) with low homology to H. pylori chromosomal genes (Fig. 1A) [13]. Plasmid pHel12 is rather homologous to pHel4 in its gene content and primary sequence [20]. It carries a microcin region as well as a putative mobilisation region, but contains only a homologue of orf4M (HPP12_12), but no homologues of orf4K, L, N and O. We show here that cryptic plasmids of H. pylori can be transferred by at least three different mechanisms: natural transformation, mobilisation and a DNaseI-resistant alternate pathway.

Generation of a Model System to Monitor Directional H. pylori Plasmid Transfer
Several plasmids carrying putative mob regions have been described for H. pylori [17]. They are candidates for a conjugative transfer between H. pylori strains. In this study, plasmids pHel4, present in H. pylori strain P8, and pHel12, present in strain P12 (Fig. 1A), were analysed for their transfer potential. The comB T4SS (tfs2) is involved in plasmid and chromosomal DNA uptake by natural transformation and is present in all H. pylori strains tested so far [9,20]. Other T4SSs, such as the cag-T4SS (tfs1), tfs3 or tfs4, are variably present in individual H. pylori strains [20]. Plasmid transfer from a donor into a recipient strain is supposed to occur via plasmid DNA release by the donor and subsequent transformation of the recipient by the comB system [9]. Thus, any one of the known T4SSs could be involved in conjugative plasmid transfer (Fig. 1B). In strain P12, all four T4SS are present, whereas P8 carries comB and cag-PAI only [20].
To verify conjugative transfer of plasmids pHel4 or pHel12 between H. pylori strains, a chloramphenicol acetyltransferase gene (cat GC ) was inserted either upstream or downstream of the mobA gene (pHel4[mccA::cat]; pHel12::cat), which leaves the mob region intact (Fig. 1A). The donor strain (P8, P12) carried a recA deletion, whereas the recipient strain (P12) was streptomycin-(str) or kanamycin resistant (moeB::aphA-3) (Fig. 1B). Transfer of the plasmid was monitored by screening transconjugants containing the cat GC gene on the plasmid and the streptomycin or kanamycin (Str/Kan) resistance in the recipients chromosome (Fig. 1B). The recA mutation in the donor avoided a possible transfer of the chromosomal resistance marker from the recipient back into the donor strain by transformation. This guaranteed a transfer of plasmid pHel4 or pHel12 into the recipient (Fig. 1B). The unidirectional plasmid transfer from donor to recipient was verified by RAPD PCR (data not shown).

Verification of Plasmid Transfer between Different H. pylori Strains
Co-cultivation of H. pylori P8 or P12 as donor strains and P12 as recipient strain resulted in transfer of plasmids pHel4 or pHel12 at a frequency between 10 25 and 10 26 , respectively (agar plate mating, see Materials and Methods) (Fig. 1C columns 1 and 5). The inactivation of the relaxase gene (pHel4[mobA::cat, pHel12[D-mobA::cat], reduced the plasmid transfer rate slightly but not significantly (Fig. 1C, columns 3 and 7). The transferred plasmid pHel12::cat was successfully reisolated from the recipient strain after mating. The transferred as well as the indigenous plasmids co-existed in the recipient strain, although the transferred plasmid carrying the cat resistance gene was found in higher quantities, as demonstrated by gel electrophoresis and PCR analysis ( Fig. 2A-C).
To further verify and expand these data to other H. pylori strains, co-incubation experiments were performed using H. pylori strain P8 as a donor with a set of different recipient strains, such as H. pylori, P1, P8, P12, or B128. Plasmid transfer was observed with these independent recipient strains at slightly different rates (data not shown). This suggested that independent H. pylori strains can act as recipients for plasmid DNA transfer.

Plasmid Transfer Takes Place by Both, DNaseI-Sensitive Natural Transformation and a DNaseI-Resistant Transfer Mechanism
To study the mechanism of plasmid transfer, for all subsequent experiments the same strain, H. pylori P12, was used as donor (recA::erm) and as recipient (str or moeB::aphA-3) strain. This strategy ruled out potential plasmid incompatibilities or DNA restriction mechanisms [24] that might otherwise affect plasmid transfer efficiency. Furthermore, for strain P12 the complete genome sequence is available [20], so that all T4SS and relaxase genes present in the genome are known. To further corroborate the plasmid transfer as a plasmid mobilisation event, plasmid transfer was analysed with the addition of DNaseI, to remove extracellular DNA and thus to abolish natural transformation. Co-cultivation of H. pylori P8 or P12 carrying pHel4[mobA::cat] or pHel12[DmobA::cat] (donor) and P12DmoeB::aphA-3 (recipient) with DNaseI treatment, revealed a significant reduction (p,0.05) in transfer rates between intact and mobA-defective plasmids for both P8 and P12 donor strains (Fig. 1C, compare columns 2 versus 4 and 6 versus 8). These data are best explained by natural transformation of extracellular released plasmids pHel4 or pHel12 (e.g. by lysis of bacteria), as long as no DNaseI was present (Fig. 1B, blue, dashed arrows). The addition of DNaseI destroyed extracellular plasmid DNA and abolished transformation-mediated plasmid transfer, whereas the knockout of the mobA relaxase gene is supposed to obstruct conjugative plasmid DNA transfer only. Thus, our data suggest that the relaxase MobA has a small but significant effect on DNaseI-resistant plasmid transfer of pHel4 and pHel12, which verifies MobA as a functional relaxase for H. pylori plasmids. However, since inactivation of mobA in the presence of DNaseI did not completely abolish plasmid DNA transfer, a further DNaseIresistant transfer of pHel plasmids seems to occur, which is independent of the plasmid-encoded relaxase MobA.
Conjugative DNaseI-Resistant Plasmid Transfer is Mediated by comB, but no Other T4SS of H. pylori P12 Certain cryptic plasmids of H. pylori carry the relaxase gene mobA, but none of them contain a complete T4SS mediating its conjugative transfer. We therefore asked which T4SS is responsible for the DNaseI-resistant plasmid mobilisation between H. pylori strains. Since up to four unique T4SSs have been described in different H. pylori isolates, one (or several) of the chromosomally encoded T4SSs (tfs1-tfs4) might mediate plasmid mobilisation.
To study the role of these T4SSs for plasmid transfer, all T4SSs were successively removed from the genome by a marker-free genetic deletion method, the rpsL counterselection procedure [25]. First, the complete tfs3 and tfs4 systems were deleted in the P12 donor strain and the correct deletions were verified by PCR analysis (Fig. 3) (see Materials and Methods and Tables 1, 2 and  Table S1 for details of plasmid construction). Removal of TFS3 or TFS4 in the donor (Fig. 4A) or in the recipient strain (Fig. 4B) did not significantly change the plasmid transfer frequency, neither without nor with DNaseI. Also a double mutant (P12Dtfs3Dtfs4) did not show a defect in plasmid transfer. In the next step the cag-PAI was removed by the same procedure (Fig. 3C), resulting in a triple mutant (P12Dtfs3Dtfs4Dcag-PAI). Again, plasmid transfer rates were not reduced, as compared to the P12 wt situation (Fig. 4A, B). An additional deletion of the comB system in the donor strain (P12Dtfs3Dtfs4Dcag-PAIDcomB, also termed P12DT4SS), or deletion of the complete comB system only, did not result in a significant change of plasmid transfer rate (Fig. 4A, columns 11 and 12). However, the same deletions in the recipient strain significantly reduced plasmid transfer rates, especially in experiments without DNaseI (transformation) but also to a lower extent with DNaseI (p,0.05) (  Schematic map showing the gene content and basic functional regions of two cryptic plasmids pHel4 and pHel12. Both plasmids contain a microcin gene cluster with homology to E. coli mccC7 (orf4A, B, X, or orf01, 02, 03). Genes orf4C-F and orf04-orf07, respectively, show homology to the mob region of colicinogenic plasmids. Genes orf4G/H and orf08/09, respectively, are related in sequence to the toxin-antitoxin system RelE-RelB (TA). The plasmid replicase is encoded by orf4I (pHel4) or orf10 (pHel12). Both plasmids also carry a number of hypothetical genes (orf4J-O, orf11). The insertion of the cat GC antibiotic resistance gene cassette into various orfs is shown. (B) Procedure of H. pylori co-cultivation to determine plasmid DNA transfer via natural transformation or conjugative processes. All T4SS (cag-PAI, comB, tfs3, tfs4) potentially involved in plasmid transfer between an H. pylori donor (P8, 12) and a recipient (P12) strain are indicated. Generally, natural transformation of the recipient strain by released plasmid DNA after lysis of the donor strain and conjugation processes are superimposed and are discriminated by adding of DNaseI. Plasmid transfer of pHel4[mccA::cat] or pHel12::cat from strain P8 or P12 into the P12 recipient strain was monitored by selection of the recipient strain (P12) via streptomycin (P12 str ) or kanamycin (P12moeB::aphA-3). To avoid the transfer of the chromosomal marker of the recipient (str, aphA-3) into the donor strain by natural transformation, the recA gene in the donor strain was deleted (P8DrecA::erm, P12DrecA::erm). (C) Inter-or intrastrain transfer of the cryptic plasmids pHel4[mccA::cat] or pHel12::cat from donor strain P8 or P12 into the recipient strain P12 str in the absence (2) or presence (+) of DNaseI. Transfer rates were determined as the number of transconjugants/cfu/ml. Data shown are mean values of at least three independent experiments including standard deviations. * p,0.05, ns, not significant. doi:10.1371/journal.pone.0045623.g001 simultaneous deletion of the comB T4SS in both, the donor as well as the recipient strain, maximally reduced plasmid transfer rates, supporting the role of the donor strain comB T4SS for DNaseIresistant plasmid transfer (Fig. 4B, columns 15 and 16). To exclude any insufficient DNA degradation by DNaseI and therefore residual transformation, the extracellular DNA of the co-cultivation assay with and without DNaseI was extracted, sterile-filtrated and used for transformation of the H. pylori P12 wild type (wt) strain. The non-treated, but not the DNaseI-treated extract resulted in transformation of P12, indicating that the DNase used was efficient during the co-incubation conditions (data not shown). Plating of the donor and recipient strains separately (without coincubation) on double selective media did not result in any resistant bacteria, which excluded the generation of any spontaneous mutants (data not shown).
In conclusion, these data suggest that none except the comB T4SS contributes to plasmid transfer by transformation as well as a DNaseI-resistant mobilisation.

Plasmid-Encoded Relaxase MobA, but not Chromosomally Encoded Relaxases Contribute to Plasmid Transfer
The plasmid-encoded relaxase MobA contributed to DNaseIresistant plasmid transfer of pHel4 as well as pHel12 (Fig. 1C). H. pylori P12 harbours two chromosomally-encoded VirD2-homologous relaxases, HPP12_1353 and HPP12_0451, located next to tfs3 and tfs4, respectively (Fig. 3A, B). Therefore, it could not be excluded that these putative chromosomal relaxases partially compensate for the deletion of the plasmid-encoded relaxase in strain P12. We therefore sought to analyse the effect of the plasmid-as well as chromosomally encoded relaxases in strain P12 on plasmid transfer. The original deletions of TFS3 or TFS4 described above left the VirD2-homologous relaxases in the chromosome. Therefore, the complete tfs3 and tfs4 systems, including the adjacent relaxase genes were deleted using the rpsL counter-selection procedure. Precise deletions were verified by PCR as described for the tfs3/4 deletions (data not shown, see Material and Methods section and Table S1 for primers) Removal of HPP12_1353 together with tfs3 (Dtfs3DvirD2), as well as HPP12_0451 with tfs4 (Dtfs4DvirD2) did not significantly change the plasmid transfer rates (Fig. 5A). Also a double mutant (Dtfs3DvirD2, Dtfs4DvirD2) did not have a significant effect on plasmid transfer, as compared to the transfer rates between P12 strains (Fig. 5B). In conclusion, the chromosomally-encoded relaxases HPP12_1353 or HPP12_0451, located adjacent to TFS3 or TFS4, respectively, were not involved in plasmid DNA transfer, neither in a DNaseI-sensitive nor a DNaseI-resistant manner.

A T4SS-Independent Alternate DNaseI-Resistant (ADR) Mechanism of Plasmid Transfer in H. pylori
Our data provided evidence that plasmids can be transferred between H. pylori strains by natural transformation, as well as by conjugation. Both routes of DNA transfer are strictly dependent on the comB T4SS. For conjugation or mobilisation of plasmids between H. pylori, the mobA relaxase is essential. However, in our experiments DNaseI-resistant plasmid transfer was clearly seen in strains without any T4SS (P12DT4SS) used either as donor or as recipient (Fig. 4). To unequivocally prove the existence of such an alternate DNA transfer pathway in H. pylori operating independently of any T4SS, an experiment was designed using a P12DT4SS strain as a donor as well as a recipient. Such a strain cannot be transformed by plasmid DNA [9], nor can it act as a donor or recipient for conventional conjugative DNA transfer, since it is does not contain any T4SS (Fig. 6A).
Co-incubation experiments using such a pair of strains showed that plasmid pHel12::cat can be transferred between H. pylori, albeit at a lower efficiency as that seen for natural transformation or mobilization (,1610 7 ), but still at a reasonable rate (Fig. 6B). Furthermore, the additional deletion of the mobA gene of pHel12 (pHel12DmobA::cat) did not abrogate the transfer between the P12DT4SS strains. However, control plating experiments with the donor or the recipient strain alone on double selective media did not result in any resistant bacteria, excluding spontaneous mutations of the donor or the recipient strain as a source for the double resistant bacteria. Without any T4SS extracellular plasmid DNA should not contribute to this type of plasmid transfer via transformation, which is supported by the fact that a similar low rates of DNA transfer were seen with and without addition of DNaseI (Fig. 6B, C). Finally, the plasmids from the double resistant bacteria were isolated and confirmed to be pHel12::cat. The recipient strain carrying the transferred plasmid could also be verified as P12DT4SS_moeB::aphA-3.
In conclusion, these experiments clearly show that an alternative pathway for plasmid exchange is existent in H. pylori, which is Plasmid Transfer in Helicobacter pylori PLOS ONE | www.plosone.org distinct from natural transformation and conventional conjugation. We suggest naming this pathway as ADR pathway.

Discussion
The existence of plasmids in H. pylori has been known for more than twenty years [12]. More recently, a subgroup of H. pylori plasmids has been described that carries putative mobilisation genes [13,17,20,26], which might be involved in conjugative transfer between different H. pylori strains. This is of particular interest, since such plasmids might be involved in the generation of genetic variation in H. pylori [13]. Here we studied the transfer of such plasmids between identical and different, unrelated H. pylori strains. Since H. pylori is naturally competent for DNA transformation, the analysis of conjugative transfer of plasmid DNA between H. pylori is more complicated than in bacteria without natural transformation competence, since both processes are superimposed (Fig. 1B). A simple transfer of the chromosomal resistance gene marker (str, aphA-3) from the recipient bacteria into the donor strain by natural transformation, which would result in double resistant pseudo-transconjugants should be avoided. We  Table S1 for primers) and cloned into the pBluescript vector either with or without the rpsL-erm cassette in-between. Verification of the correct deletion was obtained by DNA amplification of a genomic fragment using oligos SR24/SR25 resulting in a PCR fragment of 0.77 kb, as expected (gel I). The PCR fragment was further verified by DNA sequencing. Furthermore, amplification of an internal region of Tfs3 (primers WS542/355) resulted in the expected PCR product from wt, but not P12DT4SS chromosomal DNA (gel II) (B) For deletion of the complete Tfs4 (14 kb region) the flanking sequences were amplified using oligos SR13/SR14 and SR32/SR34 (see Table S1 for primers) and cloned into the pBluescript vector either with or without the rpsL-erm cassette enclosed. Verification of the correct deletion was obtained by DNA amplification of a genomic fragment using oligos SR22/SR34, resulting in a 2.6 kb fragment, as expected (gel I). The PCR fragment was further verified by DNA sequencing. Amplification of an internal region of Tfs4 (primers SR48/49) resulted in the expected PCR product from wt, but not P12DT4SS chromosomal DNA (gel II) (C) The complete cag-PAI was deleted by the rpsL-counterselection procedure. Flanking fragments were cloned using oligos JP22/23 and JP24/25 for PCR (see Table S1 for primers). The correct deletion was verified by the generation of a PCR fragment of 300 bp spanning the deletion, as generated by primers hp519 and hp549. The PCR fragment was further verified by DNA sequencing. Amplification of an internal region of the cag-PAI (primers WS418/HP542f) resulted in the expected PCR product from wt, but not P12DT4SS chromosomal DNA (gel II). (D) Deletion of the comB6-comB10 genes of the comB T4SS by the rpsL-counterselection procedure (4.3 kb). Flanking regions were cloned using oligos AK59/AK65 and DHO10/DHO11 for PCR (see Table S1 for primers). therefore generated a recA mutation in the donor strains making them unable to integrate chromosomal markers (Fig. 1B).
Generally, plasmid mobilisation is defined by the DNaseIresistant transfer of DNA and its dependence on a functional T4SS, mediating the conjugative transfer. Classical conjugative transfer is also dependent on a functional relaxase, which nicks one strand of the plasmid DNA at a specific sequence in the mob site. Rolling circle replication elongates the DNA strand and the relaxase bound to the 59-end of the single-stranded DNA then mediates the specific transfer of the nucleoprotein complex into the T4SS and the recipient cell [27].
We show here for the first time that the comB T4SS, which up to now has been associated with natural transformation competence only, can accomplish DNaseI-resistant plasmid DNA transfer as well. We identified two distinct mechanisms of DNaseI-resistant plasmid transfer, one mechanism being dependent on a functional plasmid-encoded relaxase (plasmid mobilisation) and one transfer mechanism completely independent of any relaxase, and even any T4SS, but nevertheless resistant to DNaseI (ADR). Conjugative transfer of H. pylori shuttle vectors carrying an RP4 oriT sequence has been reported to occur between different H. pylori isolates [28].
These plasmids did not carry a relaxase gene on the plasmid, but contained an origin of transfer (oriT) from the broad host range plasmid RP4. Conjugative transfer of these plasmids was reported to be dependent on a chromosomal relaxase gene (rlx1, hp0996) and the coupling protein HP1006 [28]. Plasmid transfer with the endogenous plasmid pHel12, which carries a H. pylori-specific relaxase gene and no RP4 oriT, has been shown here to be independent of chromosomal relaxases.
The comB T4SS is well established as being absolutely necessary for transformation-mediated DNA uptake into H. pylori [9]. Our data show that in addition to transformation, comB is also important for DNaseI-resistant plasmid DNA transfer between H. pylori strains, generally designated as conjugative transfer or mobilisation. First, a comB deletion in the recipient strain significantly reduced plasmid DNA transfer rates, both for transformation and DNaseI-resistant transfer (Fig. 4B, columns [11][12][13][14][15][16]. Second, a mobA deletion in pHel12 resulted in a significant reduction of DNA transfer (Fig. 1C), whereas other T4SS did not have any effect. Thus, we showed here for the first time that a comB-dependent mobilisation of pHel12 plasmid is an obvious pathway of plasmid transfer between H. pylori. Surprisingly, a P12 recipient strain without any T4SS (P12DT4SS), for which DNA transformation was completely abolished, did not exhibit a significantly different efficiency of DNaseI-sensitive or resistant pHel12 plasmid transfer rate under contact-dependent co-incubation conditions (Fig. 4B, columns 11/ 12) than a P12DcomB strain (Fig. 4B, columns 13/14). Definite proof for pHel12 plasmid transfer in the absence of any T4SS allowing natural transformation or conjugation was coming from a co-incubation experiment with donor and recipient strains devoid of any T4SS (P12DT4SS). Since plasmid transfer could be demonstrated under these conditions, our data prove that plasmid transfer between H. pylori can occur by a novel, hitherto not characterized T4SS-independent, but DNaseI-resistant pathway. We propose to designate this pathway as alternate DNaseI-Resistant pathway (ADR pathway). The mechanism for plasmid DNA transfer via the ADR pathway is unclear and currently investigated in detail in our lab. From the current literature we would envisage two possible mechanisms for plasmid DNA transfer in H. pylori via the ADR pathway, (i) either outer membrane vesicles (OMVs) [29] or (ii) nanotubes [30]. Nanotubes have been recently described as variously sized tubular extensions connecting Gram-positive or Gram-negative bacteria, allowing non-conjugative DNaseI-resistant plasmid transfer between the same or even different bacterial species. The production of OMVs has been described for H. pylori, especially for a potential delivery of proteins (e.g. VacA) or peptidoglycan into host cells [31,32], but not for transfer of plasmid DNA.
Taken together, these data demonstrate that three different mechanisms of plasmid DNA exchange are operating in H. pylori. Whether these different mechanisms are active at the same time, or whether there is a spatial or temporal control for one or the other mechanism, is currently not known. However, it is possible that by knockout of the comB system, e.g. in the donor strain, the ADR pathway may compensate for the defect in plasmid transfer, which might explain the small effect of a P12DcomB donor strain on plasmid DNA transfer efficiency (Fig. 4A). Furthermore, our data were exclusively generated under in vitro conditions. The in vitro plasmid transfer rates cannot easily be transferred to the situation of the bacteria under in vivo conditions in the stomach mucosa. It is well possible that one or the other pathway might be turned on or off under in vivo conditions. In a recombination-based in vivo expression technology (RIVET) screen, a promoter in the mG27 H. pylori strain was identified, which turned on the expression of the plasmid-encoded mobA gene under in vivo conditions in the mouse stomach [33], indicating that plasmid transfer might be enhanced when the bacteria are in their natural environment, the stomach mucosa. In the in vitro situation, natural transformation seems to represent the most efficient way of plasmid transfer, followed by conjugative transfer, whereas the ADR pathway contributes only minimally. The relative contribu- tion of one or the other pathway under in vivo conditions has to be determined in future.

Electroporation, Natural Transformation and Conjugation of H. pylori
Transformation of H. pylori strains was performed with plasmid or chromosomal DNA as described earlier [34]. Bacteria were harvested from serum plates and suspended to an optical density at 550 nm (OD 550 ) of 0.2 in Brucella broth (BB) containing 10% fetal calf serum. DNA was added (1 mg), and incubation was extended for 4 h under microaerophilic conditions before the suspension was plated on selective serum plates. For electroporation of H. pylori, bacterial cells were harvested from serum plates and suspended in 1 ml of phosphate-buffered saline (PBS) solution. For each electroporation, bacteria were diluted to an OD 550 of 1 and 1 ml bacterial suspension was washed twice with PBS and suspended first in 500 ml and then in 40 ml electroporation buffer [36]. Forty microliters of H. pylori competent cells was mixed with 1 to 2 ml DNA in pre-chilled 0.2 cm electroporation cuvettes. Electroporation was performed at 2.5 kV, 200 V, and 25 mF by a Gene Pulser (Bio-Rad, Munich, Germany). After electroporation, 1 ml of BB containing 10% fetal calf serum was added immediately to each sample. The aliquots were incubated for 4 h in a CO 2 incubator before being plated on selective agar plates.

DNA Manipulations and Plasmid and Strain Constructions
Cloning and DNA analysis procedures were performed according to Sambrook et al. [35]. Chromosomal DNA from H. pylori was isolated with the QIAamp tissue kit (QIAGEN, Hilden, Germany). Plasmid DNA was purified from E. coli using the QIAprep Spin Miniprep kit (QIAGEN, Hilden, Germany).
Deletion and replacement of genes or genomic regions were achieved using the streptomycin susceptibility counterselection strategy [25]. A rpsL-erm cassette cloned into the BamHI restriction site of pBluescript IISK+ was used for all cloning procedures. For the deletion of the TFS3 system, the corresponding upstream and downstream regions were amplified by PCR using primers SR17/ SR18, and SR30/SR35, respectively (Table S1). Likewise, tfs4 upstream and downstream regions were amplified using primers SR13/SR14 and SR32/SR34, respectively. Upstream and downstream regions were cleaved with XhoI/ClaI and NotI/SacII, respectively, and cloned with or without the rpsL-erm cassette into the corresponding sites of pBluescript IISK+. The plasmids obtained were used in sequential transformations to generate marker-free deletions of the corresponding T4SS.
To inactivate the ComB system, comB6-comB10 were deleted using the rpsL-erm system and plasmid pAK23 [10]. For deletion of the cag-PAI, a plasmid containing the left and right flanking gene regions of the cag-PAI [37], was modified by insertion the rpsL-erm cassette. To monitor plasmid transfer between H. pylori strains, a chloramphenicol resistance cassette (cat GC ) was inserted between orf4M and orf4A in pHel12 using an inverse PCR reaction with primers SR53 and SR54. The cat GC cassette (pWS48) was cloned via a BamHI restriction site. Homologous recombination in plasmid pHel12 resulted in pHel12::cat GC .
For deletion of the plasmid-encoded relaxase of pHel12, a plasmid from a shotgun library was used as template and after inverse PCR using primers SR41/SR52, the cat GC cassette was inserted via a BamHI restriction site. In pHel4 the cat GC cassette was inserted within orf4C to disrupt the corresponding gene. To delete the chromosomal relaxases within tfs3 and tfs4, plasmids pSR29 and pSR31 were generated using primer pairs SR17/ SR18 and SR69/SR70 (tfs3/Rel deletion), or SR71/SR72 and SR28/SR34 (tfs4/Rel deletion). As already described above, the rpsL-erm counterselection strategy was applied to generate markerfree H. pylori mutant strains.

Mating Experiments
To monitor transfer of the donor plasmid into the recipient strain, donor strains were provided with a cat GC cassette in the plasmid. For selection of the recipient strain, either an aphA-3 cassette was inserted into the recipients chromosomal moeB locus, using plasmid pSP76-aphA-3, or a streptomycin resistant P12 mutant (pEG21) was used. To prevent transformation of donor strain with chromosomal DNA from recipient strain, the recA gene was deleted in the H. pylori donor strains by transformation with the plasmid pDH29 (recA::erm). Donor and recipient strains were harvested after 24 h of growth on selective GC agar plates and suspended in 1 ml of BB. For mating experiments between H. pylori 1.5?10 7 cells were used. 25 ml donor and 25 ml recipient cell suspension were either mixed with 25 ml DNaseI (Roche; 1 mg/ml in 20 mM Tris-Cl (pH 7.5), 1 mM MgCl 2 ) each in separate tubes and pre-incubated for 30 min at 37uC and 10% CO 2 , or left untreated. Subsequently donor and recipient cells were mixed, collected at 1000 g for 5 min and resuspended either in 50 ml DNaseI or 50 ml BB. The cells were spotted onto non selective GC agar plates directly and incubated over night at 37uC under microaerobic conditions. After incubation the cells were harvested and resuspended in 1 ml BB. The cell suspension was serially diluted and 100 ml were plated on GC agar plates containing chloramphenicol to determine the CFU/ml of the donor strain and 100 ml and harvested ,800 ml cell suspension were plated on appropriate double selective GC agar plates (chloramphenicol and kanamycin or streptomycin) to determine the number of transconjugants. The plasmid transfer rates were determined by colony counting and are presented as numbers of transconjugants per donor cfu. Plasmid sequence accession numbers: pHel4: NC_004950; pHel12: CP001218.

Statistical Analysis
The values shown are means 6 SD from at least three independent experiments. Students t-test was used to analyze the data. P values are indicated in the figures and were considered significant if they were ,0.05.

Supporting Information
Table S1 Oligonucleotide Primers used in this study. Different types of oligonucleotide primers are listed which were used for amplification of genes for cloning purposes, for verification of chromosomal deletions (e.g. complete T4S systems) or for sequencing. The cloning procedures are described in the methods section. (DOCX)