Correspondence: Günther Muth, Mikrobiologie/Biotechnologie, Interfakultäres Institut für Mikrobiologie und Infektionsmedizin Tübingen IMIT, Eberhard Karls Universität Tübingen, Auf der Morgenstelle 28, 72076 Tübingen, Germany. Tel.: +4970712974637; fax: +497071295979; e-mail: email@example.com
Antibiotic-producing soil bacteria of the genus Streptomyces form a huge natural reservoir of antibiotic resistance genes for the dissemination within the soil community. Streptomyces plasmids encode a unique conjugative DNA transfer system clearly distinguished from classical conjugation involving a single-stranded DNA molecule and a type IV protein secretion system. Only a single plasmid–encoded protein, TraB, is sufficient to translocate a double-stranded DNA molecule into the recipient in Streptomyces matings. TraB is a hexameric pore-forming ATPase that resembles the chromosome segregator protein FtsK and translocates DNA by recognizing specific 8-bp repeats present in the plasmid clt locus. Mobilization of chromosomal genes does not require integration of the plasmid, because TraB also recognizes clt-like sequences distributed all over the chromosome.
Mycelium-forming actinomycetes do not divide by binary fission but grow by apical tip extension and undergo a complex life cycle ending in sporulation (Flardh & Buttner, 2009). They are well known for the production of antibiotics, a feature probably developed to inhibit competitors in the soil community (Allen et al., 2010). During evolution of the antibiotic biosynthetic gene clusters, they also evolved specific resistance genes as a part of the cluster to protect themselves from their own compounds. Because a typical Streptomyces strain contains 10–20 different gene clusters for the production of antibiotics and other bioactive secondary metabolites (Bentley et al., 2002; Medema et al., 2011), streptomycetes form a huge reservoir of antibiotic resistance genes in the soil, which can be passed to other bacteria by horizontal gene transfer (D'Costa et al., 2006; Allen et al., 2010). Therefore, the antibiotic producers not only compete with other organisms by the production of antimicrobial compounds but they also provide resistance genes that can help others to survive.
In Streptomyces and related actinomycetes, even small multi-copy plasmids of < 10 kb in size are normally self-transmissible and able to mobilize chromosomal resistance genes and auxotrophic markers (Kieser et al., 1982; Kataoka et al., 1991; Servin-Gonzalez et al., 1995). These plasmids are normally cryptic and do not confer phenotypic traits (Hopwood & Kieser, 1993; Vogelmann et al., 2011a). Efficiency of transfer reaches nearly 100% and between 0.1% and 1% of the transconjugants obtain chromosomal fragments during mating (Kieser et al., 1982). DNA transfer takes place only on solid surfaces in the early growth phase of the life cycle, when Streptomyces grows as substrate mycelium (Pettis & Cohen, 1996; Possoz et al., 2001). The transfer determinants of many Streptomyces plasmids were initially identified as killing functions (kilA, traB), which could only be subcloned in the presence of the corresponding killing override (korA, traR) region (Kendall & Cohen, 1987; Hagege et al., 1993; Reuther et al., 2006a). Probably due to the toxic effects of the transfer determinants, plasmid transfer is associated with the formation of so-called pock structures having a diameter of 1–3 mm. Pocks are formed when donor spores germinate on a lawn of a plasmid-free recipient. Pocks represent temporally retarded growth inhibition zones and indicate the area, where the recipient mycelium has obtained a plasmid by conjugation (Fig. 1). Formation of pock structures has been interpreted as the result of intramycelial plasmid spreading via the septal cross-walls of the recipient mycelium (Hopwood & Kieser, 1993; Grohmann et al., 2003).
Conjugative transfer of a double-stranded plasmid molecule requires TraB and the noncoding clt region
The small size of the plasmid region determining conjugative transfer already indicated that the Streptomyces DNA transfer mechanism must differ considerably from the known conjugation systems of other bacteria, involving a conjugative relaxase and a complex type IV protein secretion system (Chen et al., 2005; de la Cruz et al., 2010). Characterization of several Streptomyces plasmids by subcloning and linker insertions revealed a plasmid region of about 3 kb being essential for transfer, while the adjacent region affected only the size of the pock structures (Kieser et al., 1982; Kataoka et al., 1991; Servin-Gonzalez et al., 1995; Reuther et al., 2006a). When the nucleotide sequence of the Streptomyces lividans plasmid pIJ101 was available (Kendall & Cohen, 1988) as the first complete sequence of a conjugative plasmid from a Gram-positive bacterium, it was realized that korA (traR) encoded a transcriptional regulator of the GntR family, while a small region of the KilA (TraB) protein showed some similarity to the FtsK protein involved in cell division and chromosome segregation (Begg et al., 1995; Wu et al., 1995; Sherratt et al., 2010).
Pettis & Cohen (1994) demonstrated that beside the TraB protein, a small non-coding plasmid region of about 50 bp was required for the transfer of plasmid pIJ101, the cis-acting-locus of transfer (clt). When clt was inserted into a nontransferable plasmid, this plasmid could be mobilized, if TraB was provided in trans. Interestingly, clt was only required for plasmid transfer but was dispensable for the mobilization of chromosomal markers (Pettis & Cohen, 1994), indicating that clt does not represent a classical origin of transfer (oriT). The clt regions of different Streptomyces plasmids do not show any sequence similarity, but often contain repetitive sequences that have the ability to form secondary structures (Franco et al., 2003; Vogelmann et al., 2011a).
The first experimental evidence on the novel mechanism of the Streptomyces conjugative DNA transfer system came from the work of Possoz et al. (2001) by demonstrating that conjugative transfer of the Streptomyces ambofaciens plasmid pSAM2 was sensitive to the presence of the SalI restriction/modification system in the recipient. In this study, a pSAM2 derivative could not be transferred into S. lividans TK23 expressing SalI, whereas pSAM2 was efficiently transferred to TK23 lacking the SalI restriction system. Because the transferred DNA was obviously degraded by SalI and because SalI recognizes only double-stranded DNA as substrate but not single-stranded DNA, the incoming DNA must be double-stranded.
TraB is a pore-forming hexameric ring ATPase homologous to FtsK
Multiple alignment of TraB homologues from various Streptomyces plasmids revealed a highly diverse family of proteins showing only very limited sequence similarity, in part < 20%. However, secondary structure predictions revealed an identical domain architecture for all TraB homologues, which resembles that of FtsK: a N-terminal membrane association domain that is followed by a DNA-translocase/ATPase domain with Walker A and B boxes and a C-terminal winged helix-turn-helix fold (wHTH) (Vogelmann et al., 2011a). ATPase activity and membrane association have been experimentally confirmed for TraB proteins of various plasmids (Kosono et al., 1996; Pettis & Cohen, 1996; Reuther et al., 2006a). Inactivation of the ATP binding site of TraB from the Streptomyces nigrifaciens plasmid pSN22 demonstrated that the ATPase activity is essential for conjugative transfer (Kosono et al., 1996).
The similarity of TraB to the septal DNA translocator proteins FtsK and SpoIIIE that direct chromosome segregation during cell division and sporulation (Bath et al., 2000; Massey et al., 2006; Bigot et al., 2007) suggests a similar function for TraB during conjugation. However, whereas FtsK translocates the DNA through a closing septum to the daughter cell/spore, TraB has to translocate the DNA through intact cell envelopes of the donor and the recipient. Because a TraB–eGFP fusion protein localized to the hyphal tips of substrate mycelium, it was suggested that Streptomyces conjugation involves the tips (Reuther et al., 2006a). Up to now it is still unclear, whether TraB contains a membrane-targeting sequence and is directed to the tip by the membrane composition or curvature or whether TraB is recruited to the tips by its interaction with other proteins, for example, DivIVA (Hempel et al., 2008; Lenarcic et al., 2009; Jyothikumar et al., 2012).
Despite the toxic effects of TraB, this protein of the Streptomyces venezuelae plasmid pSVH1 could be expressed in S. lividans with an N-terminal Strep tagII sequence (Voss & Skerra, 1997) and purified (Reuther et al., 2006a). Chemical crosslinking showed higher oligomeric structures that were also observed when the membrane association domain of TraB was eliminated. After separation of TraB oligomers from the monomer fraction by gel filtration chromatography, ring-shaped TraB particles could be detected by electron microscopy. 2D averaging of the images revealed symmetric hexamers of about 12 nm in diameter, which contained a central pore. This structure was in full agreement with a predicted TraB-DNA-translocase structure obtained by homology modelling with the Pseudomonas aeruginosa FtsK translocase domain crystal structure as a template (Vogelmann et al., 2011a). Both structures had a central pore of 3.0 and 3.1 nm, respectively, which is of sufficient size to accommodate a double-stranded DNA molecule.
Conjugative transfer of DNA by direct cell to cell contact implies that the DNA has to pass the cell envelopes of donor and recipient. For Streptomyces, this means: two cytoplasmic membranes and two peptidoglycan (PG) layers. PG-binding assays of purified TraB showed that TraB interacts with PG (Vogelmann et al., 2011a). For the membrane passage, one has to postulate a pore structure for TraB. This is in contrast to Escherichia coli FtsK that probably translocates the chromosome before closure of the septum and therefore does not rely on a pore-forming ability (Dubarry & Barre, 2010). The ability of TraB to form pore structures was analysed by single channel recordings using planar lipid bilayers. These studies demonstrated that TraB spontaneously inserted into the membrane at various voltages and formed pores with an opening time of about 47–81 ms (positive voltage applied) and 105–200 ms, respectively, when a negative voltage was applied (Vogelmann et al., 2011a).
TraB specifically recognizes 8-bp repeats in the plasmid clt region
Because only TraB and the non-coding clt region are required for plasmid transfer, it was studied whether clt represents the binding site of TraB. This hypothesis turned out to be correct, because gel retardation assays showed a specific interaction of TraB with a plasmid region at the 3′ end of traB, which represents the clt region of pSVH1 (Reuther et al., 2006a). The pSVH1 clt region contained nine imperfectly conserved copies of the GACCCGGA motif. Subcloning experiments revealed a minimal fragment containing only four copies, which still supported TraB binding. A more careful analysis detected even binding of TraB to a synthetic 20-bp fragment containing only two copies (Vogelmann et al., 2011a). This study confirmed the GACCCGGA motif as the TraB Recognition Sequence (TRS). Although two copies of TRS were sufficient for TraB binding in vitro, binding of TraB to a larger clt fragment containing additional TRS copies was more efficient and required lower protein concentrations for retardation (Reuther et al., 2006a) indicating that in vivo only the complete clt might be effective.
Analysing other Streptomyces plasmids for the presence of 8-bp repeats also detected specific 8-bp repeats in the (predicted) clt regions (Franco et al., 2003; Vogelmann et al., 2011a). With the notable exceptions of pIJ101 (Kieser et al., 1982) and the highly related plasmid p1424 (G. Muth, unpublished), the clt localizes in all Streptomyces plasmids to the 3′ end of traB, forming a transfer module of only 2.5 kb in size consisting of the DNA-translocase-encoding traB gene and its binding site clt next to it.
To characterize the TraB–clt interaction in more detail, TraB was incubated with covalently closed circular (ccc) DNA of the pSVH1 derivative pEB211 in the presence of ATP and divalent cations. An aliquot was directly loaded to the gel, while others were heat treated or phenol extracted to denature TraB previous to gel loading. These analyses revealed ccc-DNA that had not changed its conformation demonstrating that TraB binds noncovalently to plasmid DNA and that the plasmid molecule was not processed by TraB binding (Reuther et al., 2006a).
Because the clt loci of different plasmids contain different 8-bp TRS, it was possible to map the TraB region determining specific TRS recognition by constructing a series of chimeric proteins and analysing their DNA binding activities in gel retardation experiments with different clt loci. These studies identified the very C-terminal end of TraB forming a wHTH fold as being responsible for clt recognition. Further studies even narrowed down the TRS recognition region to helix α3 of the wHTH fold. Exchange of only 13 aa of TraBpSVH1 against the 13 aa corresponding to helix α3 of TraBpIJ101 switched clt recognition. The chimeric protein was no longer able to bind to the clt of pSVH1 but shifted the clt fragment of pIJ101 (Vogelmann et al., 2011a).
Colonization of the recipient by intramycelial plasmid spreading requires additional plasmid-encoded proteins
Generation of pock structures during Streptomyces conjugation has been interpreted as the result of intramycelial plasmid spreading following the primary DNA transfer from a donor into the recipient (Hopwood & Kieser, 1993; Grohmann et al., 2003). Whereas plasmid transfer from a donor into the recipient requires only TraB, plasmid spreading involves five to seven plasmid-encoded proteins (Spd) in addition to TraB. This probably reflects the challenge to cross the septal cross-walls. The Spd proteins have no significant similarity to any functionally characterized protein complicating prediction of their putative function. Inactivation of a single spd gene reduces the size of the pock structures (Kieser et al., 1982; Kataoka et al., 1994; Servin-Gonzalez et al., 1995; Reuther et al., 2006a). Only few reports address the biochemical characterization of the Spd proteins and their molecular function is more or less unknown. Genetic organization of the spd genes with overlapping stop and start codons, analysis of protein–protein interaction by chemical crosslinking, bacterial two-hybrid analysis or copurification experiments indicated that the Spd proteins form a multiprotein complex with TraB (Tiffert et al., 2007) (Thoma, Guezguez and Muth, unpublished).
Intramycelial plasmid spreading might also contribute to the stable maintenance of Streptomyces plasmids, because hyphal compartments that have lost a plasmid can recover a plasmid from the neighbouring compartment. In agreement with this hypothesis, a clear effect of spd1 inactivation on stable maintenance of the linear plasmid SLP2 was reported (Hsu & Chen, 2010).
Conjugative Streptomyces plasmids contribute in different ways to the evolution of the genomes
Streptomyces plasmids contribute to the evolution and shaping of the chromosome in different ways (Medema et al., 2010). Linear plasmids can recombine with the chromosome. Because the Streptomyces chromosome is normally linear (Lin et al., 1993), this results in the exchange of the ends, creating plasmids that carry chromosomal DNA. These plasmids can be transferred by conjugation to new Streptomyces species, where they can replicate either autonomously or recombine again with the chromosome.
But also circular plasmids have been reported to mobilize chromosomal fragments with high efficiency (Kieser et al., 1982; Hopwood & Kieser, 1993). However, the mechanism of chromosome mobilization differs fundamentally from the one reported for other bacteria. Classical High Frequency of Recombination strains (HFR) carry the conjugative plasmid at a specific location in the chromosome (Thomas & Nielsen, 2005). Plasmid integration normally occurred via homologous recombination between IS elements. Initiation of rolling-circle replication at the plasmid oriT by the conjugative relaxase creates a linear single-stranded DNA molecule that contains plasmid sequences followed by the chromosomal loci next to the integration site. This strand is guided by the covalently bound relaxase to the recipient, where it can recombine with the chromosome (de la Cruz et al., 2010).
Because the Streptomyces DNA-translocase TraB does not have a relaxase activity and most probably does not process the DNA (Reuther et al., 2006a) and because clt is dispensable for the transfer of chromosomal markers (Pettis & Cohen, 1994), the chromosome mobilization mechanism in Streptomyces must be different (Fig. 2). An explanation provides the finding that TraB recognizes 8-bp TRS motifs and that clt-like sequences containing repeated TRS are frequently found in Streptomyces chromosomes (Vogelmann et al., 2011a). Analysis of the Streptomyces coelicolor genomic sequence for pSVH1 clt-like sequences (four copies of the TRS GACCCGGA with a spacing of up to 13 bp, allowing one mismatch) identified 25 hits. These sequences are not part of integrated plasmids or represent remnants of plasmids, but are often located within genes without disrupting their coding region. These insertions are only found in the respective S. coelicolor genes but not in the corresponding homologues of Streptomyces avermitilis or those of other Streptomyces species, which carry clt-like sequences on other locations (Sepulveda et al., 2011). This demonstrates that these insertions have been acquired later and are probably not involved in the respective enzymatic activities. It is unclear how these insertions have been generated. But with respect to the prevalence of plasmids in Streptomyces, one can speculate that there is an adaptive selection for clt-like sequences in Streptomyces genomes to benefit from the presence of conjugative plasmids.
Concluding remarks/open questions
Pettis & Cohen (1994) clearly demonstrated that TraB is the only plasmid-encoded protein required for conjugative transfer of pIJ101. Similarity of TraB to the chromosome segregator proteins FtsK or SpoIIIE suggests a conjugative DNA translocation mechanism for the transfer between a donor and a recipient mycelium that resembles the intracellular segregation of chromosomal DNA during cell division and sporulation. TraB hexamers probably assemble at the plasmid localized clt or, with lower efficiency, at chromosomal clt-like sequences. These hexamers form pore structures in the membrane, which act as molecular motors, energized by ATP hydrolysis and translocate double-stranded DNA to the recipient (Fig. 3).
However, this simplified model has drawbacks and leaves several open questions. To fully understand TraB-mediated conjugative DNA transfer, one has to propose additional enzymatic activities for TraB waiting to be uncovered. Alternatively, TraB might recruit other chromosomally encoded proteins for the transfer process.
1. How to cross the PG barrier?
A TraB–eGFP fusion was localized at the hyphal tip, suggesting that the tips of the mycelium are involved in conjugation (Reuther et al., 2006a). Also, TraB was shown to bind isolated PG (Vogelmann et al., 2011a). Because TraB itself does not have a PG-lysing activity (Finger and Muth, unpublished), it is possible that TraB interacts with chromosomally encoded PG hydrolases at the tip to direct fusion of the PG layers of donor and recipient.
2. How to cross membranes of donor and recipient?
In contrast to FtsK that is found in both compartments during cell division, TraB is present only in the donor mycelium. Therefore, the TraB pore has to traverse two membranes (one from the donor, one from the recipient) or the two membranes have to fuse. For SpoIIIE that mediates translocation of the chromosome into the forespore during Bacillus sporulation, a membrane fusing activity has been reported (Sharp & Pogliano, 2003). Therefore, it is tempting to speculate that also TraB might have a membrane fusing activity allowing formation of a pore structure to the recipient.
3. How to translocate a circular covalently closed plasmid molecule?
During cell division or sporulation, the septum closes, while chromosomal DNA is already present, allowing FtsK to assemble at both chromosomal arms to translocate the DNA. DNA translocation causes topological stress to the DNA, which has to be relieved by topoisomerases. The interaction of E. coli FtsK with topoisomerase IV has been reported (Espeli et al., 2003). However, it is still unclear, how the remaining end of the circular chromosome becomes translocated through the membrane and fusion of the two FtsK hexamer structures has been postulated (Burton et al., 2007).
During Streptomyces conjugation, the situation is even more complex. The translocase TraB is definitely present only on the donor site of the mating hyphae, and a mechanism translocating a circular double-stranded DNA molecule is not very plausible. Because the plasmid DNA is not processed during TraB binding at clt, one has to propose involvement of an additional enzymatic activity, for example, a topoisomerase, which might produce a linear molecule that can be transported through the TraB pore.
4. How to pass the septal cross-walls in the recipient mycelium?
Crossing the septal cross-walls during intramycelial plasmid spreading seems to be an even more challenging task compared to the primary DNA transfer at the hyphal tip. It involves, in addition to TraB, several Spd proteins. The structure of the Streptomyces septal cross-walls has not been elucidated, and it is not clear whether preexisting channel structures in the cross-walls connect the compartments of the substrate mycelium (Jakimowicz & van Wezel, 2012). The Spd proteins might interact to build a cross-wall traversing complex, allowing translocation of the plasmid by the motor protein TraB.
5. Is TraB able to promote intergeneric DNA transfer?
The capability of the T4SS conjugation system to transfer plasmids between distantly related bacteria, even across kingdoms, is well documented (Bates et al., 1998; Thomas & Nielsen, 2005). Although conjugative transfer of Streptomyces plasmids between different Streptomyces species has been observed (Hopwood & Kieser, 1993), conjugative transfer to other bacteria has not been reported. Therefore, the relevance of the Streptomyces conjugative DNA transfer system in the dissemination of the Streptomyces reservoir of resistance genes is still concealed.