The relaxase of the plasmid R1162 is a large protein essential for conjugative transfer and containing two different and physically separate catalytic activities. The N-terminal half cleaves one of the DNA strands at the origin of transfer (oriT) and becomes covalently linked to the 5′ terminal phosphate; the C-terminal half is a primase essential for initiation of plasmid vegetative replication. We show here that the two parts of the protein are independently transported by the type IV pathway. Part of the domain containing the catalytic activity, as well as an adjacent region, is required in each case, but the required regions do not physically overlap. Both transport systems contribute to the overall frequency of conjugative transfer. MobB is a small protein, encoded within mobA but in a different reading frame, that stabilizes the relaxase at oriT. MobB is required for efficient type IV transport of both the complete relaxase and its two, separate functional halves. MobB inserts into the membrane and could thus stabilize the association between the relaxase and the type IV transfer apparatus.
The type IV secretion systems (T4SS) in Gram-negative bacteria are complex mechanisms for the transport of macromolecules out of the bacterial cell (Cascales and Christie, 2003). While often involved in the export of effector molecules and toxins during pathogenesis, T4SS is also responsible for the conjugative transfer of plasmid DNA between bacterial cells. The close relationship between conjugation and protein secretion by the T4SS pathway is based on three observations: the homology between protein components of the transporting machines in each case (Christie, 2001), the ability of some pathogenic secretory systems also to transport plasmid DNA (Vogel et al., 1998), and the secretion of components of conjugative transfer, independently of DNA transfer (Luo and Isberg, 2004).
During conjugation, a protein called the relaxase, along with one or more accessory proteins, assembles on the DNA at the origin of transfer (oriT), cleaves one of the DNA strands and becomes covalently linked to the 5′ end (Guiney and Helinski, 1975; Pansegrau et al., 1990; Scherzinger et al., 1992; Matson et al., 1993; Grandoso et al., 2000). The relaxase is then secreted by the T4SS transport system after first interacting with a component called the coupling protein (Christie, 2001); the linked DNA is cotransported by a mechanism not yet understood. It is the coupling protein of a T4SS that largely determines the set of molecules identified for transport, and thus as well those plasmids efficiently transferred by conjugation (Cabezon et al., 1997). There is very little information on how a coupling protein identifies substrates for type IV secretion. In the case of the Vir system of Agrobacterium tumefaciens, transported Vir proteins have C-terminal ends that are relatively unstructured and rich in basic amino acids. These ends, possibly with the basic amino acids distributed in some way, might be all that is required on a substrate for selection by the Vir transporter (Vergunst et al., 2005). The signal is more complex for the Vir system of Bartonella henselae. Here, a positively charged, C-terminal end, as well as a conserved, 142-amino-acid domain, is required (Schulein et al., 2005). In addition, the CagA protein, which is secreted by Helicobacter, requires an as-yet-uncharacterized N-terminal domain in addition to a basic C-terminal region (Hohlfeld et al., 2006).
MobA, the large relaxase of the plasmid R1162 (and the virtually identical RSF1010), has two different and independent catalytic domains (Fig. 1). An N-terminal fragment (amino acids 1–186) cleaves the plasmid DNA strand and is strictly required for conjugative transfer (Brasch and Meyer, 1986). The C-terminal part of MobA, amino acids 388–709, which is also translated separately (Haring and Scherzinger, 1989; Scholz et al., 1989), is a primase required for vegetative replication of the plasmid. The linkage of two, distinct functional activities, possibly by the ancestral fusion of separate proteins, is not an invariable feature of members of the MobA family of proteins (the closely related MobA of the plasmid pSC101, for example, is not fused with another, catalytically active polypeptide) (Meyer, 2000). For the R1162 primase, the fusion might have been selected to increase the efficiency of priming complementary strand synthesis following transfer of the plasmid DNA strand (Henderson and Meyer, 1996; Parker and Meyer, 2005).
The two active regions of MobA are linked by translation through the small gene mobB, which is in a different reading frame. MobB is part of the relaxosome, and the purified protein has been shown to enhance strand cleavage at oriT by relaxosomes assembled in vitro (Scherzinger et al., 1992; Perwez and Meyer, 1996). It also appears to stabilize relaxosomes insolated from cleared cell lysates (Perwez and Meyer, 1996).
A MobA polypeptide fragment consisting of the first 284 amino acids, and therefore lacking the primase domain, is functional in conjugative transfer with the T4SS machines provided by IncP1 plasmids such as RK2 and R751 (Brasch and Meyer, 1986). Thus, this fragment must be recognized, directly or indirectly, by the coupling protein. However, a signal for type IV secretion by the Vir system has been identified within the last 48 amino acids of the RSF1010 MobA, and is therefore within the primase domain (Vergunst et al., 2005). We therefore decided to look more closely at the regions of MobA required for type IV transport. Our results show that there are two signals in MobA for R751-mediated transport. Each signal is associated with one of the functional domains of the molecule, and neither is identical to the Vir signal. Moreover, both signals unexpectedly require MobB.
MobA contains two, MobB-dependent type IV transport signals
We used the vector shown in Fig. 2A to fuse Cre protein to full-length MobA (Vergunst et al., 2000). The plasmid is a derivative of pBR322; details of its construction are given in Supplemental methods. MobA was cloned between the KpnI and SalI sites, so that the hybrid protein consisted of the last codon of Cre joined to the second codon of MobA. The hybrid protein was expressed in donor cells also containing the plasmid R751. After mating into our reporter strain, the transport frequency (the percentage of R751-containing transconjugants expressing β-galactosidase due to Cre-directed site-specific recombination) was 0.097 (Fig. 2B). Because mobB is embedded within mobA, the product of this gene is also expressed in the donor. In agreement with earlier results (Perwez and Meyer, 1999), deletion of 54 amino acids within MobB (MobBΔ1, Fig. 1) decreased the frequency of conjugative transfer of R1162 1000-fold, but the mutation can be complemented by MobB in trans (Fig. 2B). This deletion also reduced type IV transport of the corresponding CreMob fusion to an undetectable level, but transport was restored by providing MobB in the cell (Fig. 2B). Thus, MobB is important not only for stabilization of the relaxosome at oriT, but also for the secretion of the relaxase in the absence of the R1162 oriT. For clarity, we will refer to the mobilization of CreMob fusions as transport, and use the term transfer to refer to the conjugative mobilization of DNA. A second deletion, MobBΔ2 (Fig. 1), lowers the transfer frequency of R1162 less than 10-fold, but again results in undetectable transport of the corresponding CreMob derivative (Fig. 2B). Normal levels of both transport and transfer are again restored by MobB in trans.
We next tested for transport N-terminal and C-terminal MobA fragments, each containing one of the catalytic activities of MobA. A fusion containing amino acids 2–284 [CreMob(2–284)], which is active for conjugative transfer (Brasch and Meyer, 1986), was transported (Fig. 3), and again MobB was required. CreMobA(324–709), a fusion containing the C-terminal amino acids 324–709, which include the primase, was also transported (Fig. 3), and at a very high frequency relative to the N-terminal fragment. Transport of this fragment was also undetectable in the absence of MobB. Because the N-terminal and C-terminal fragments contain no overlapping amino acids, MobA must contain two separate signals, which we designate sig1 and sig2, for transport by the R751 type IV machinery.
The structural integrity of each of the functional domains of MobA is important for transport
We generated additional mutations to localize regions of MobA required for transport. An N-terminal fragment containing amino acids 2–204 (Fig. 1) was shown previously to be inactive in conjugative transfer although it still cleaves oriT DNA (Meyer, 1989; Becker and Meyer, 2002). This fragment is not transported [CreMob(2–224), Fig. 3]. Because the additional amino acids 225–284 are included in the transported fusion CreMob(2–284), these could contain all or part of the signal. However, fusion of these amino acids to Cre did not result in a transportable fusion (Fig. 3). On the other hand, changing amino acids in this region did affect transfer. We selected for mutagenesis the arginines at positions 281 and 282 (Fig. 1), because C-terminal positive charges were shown to be important for transport by the Vir system (Vergunst et al., 2005). Changing these residues to alanines, to form the fusion protein CreMob(2–284)AA, resulted in loss of transport (Fig. 3). This is not due to decreased stability of the protein, because CreMob(2–284), and the CreMob(2–284)AA derivative, are recovered from the cell in similar amounts (Fig. 4). However, more than positive charges are probably required, because changing the terminal arginines residues to lysines did not restore transport (frequency < 0.0037%).
Insertion of three amino acids between residues 113 and 114, in the N-terminal catalytic domain (Fig. 1), also eliminated transport [CreMob(2–284) + lDR (Fig. 3)]. Thus, perturbations in the structure of the Mob(2–284) fragment in a large region, and in both the catalytic and non-catalytic regions, can affect transport.
We drew a similar conclusion from the effects of alterations in the C-terminal fragment (amino acids 324–709). We first tested whether the terminal 48 amino acids of MobA, active in transport by the Vir system (Vergunst et al., 2005), are sufficient for transport by Escherichia coli containing R751. We constructed a plasmid encoding a fusion protein, CreMob(662–709), identical to that reported by Vergunst et al. (2005), but were unable to detect transport of the protein from our donors (Fig. 3). Next, we generated truncations from the C-terminal end of MobA. Fragments containing amino acids 324–599 and 324–588 were still transported (Fig. 3), although at a lower rate than the parent fusion. Smaller, C-terminal truncations active for transport were not obtained. There are two arginines at the C-terminal end of the 324–588 truncation (Fig. 1), and because active proteins deleted for these were not found, they might have been required for transport. However, changing these to alanines in the CreMobA(324–588) and CreMobA(324–709) proteins did not eliminate transport (frequencies 0.34 and 9.8 respectively). Finally, insertion of six amino acids between residues 447 and 448 in the MobA(324–709) hybrid sharply reduced transport [CreMob(324–709) + LEASSV (Fig. 3)]. The effect of this large insertion appeared to be on transport itself, because, as with the other mutations tested, it caused no large decrease in the amount of fusion protein in the cell (Fig. 4).
The 324–709 fragment highly active for transport contained the complete primase region (amino acids 387–709) as well as a ‘leader’ region encoded by DNA overlapping mobB (Fig. 1). A Cre fusion containing the leader region alone was not transported (Fig. 3), but the region was required as part of the signal, because a protein consisting of Cre fused directly to the full-length primase was also inactive (Fig. 3). This failure was not due to Cre-induced, extensive misfolding of the primase segment, because the hybrid protein could complement for replication an R1162 derivative containing a deletion in the primase gene (results not shown). In addition, insertion of a 45-amino-acid ‘spacer’ polypeptide at the Cre–primase junction, to form the fusion CreMob(387–709)K, did not restore transport (frequency < 0.0014%).
Overall, our results indicate that the transport of the primase region of MobA requires both a large, N-terminal portion of the primase domain and an adjacent, leader region. This is consistent with the idea that transport again requires considerable secondary structure, rather than a defined small and unstructured region.
Both transport signals are active in full-length MobA
To determine which transport signals are normally active in the full-length relaxase, we constructed several internal, in-frame deletions in mobA. The first of these (CreMobΔsig1) results in protein lacking MobA amino acids 205–323, and is thus defective for the signal located in the N-terminal half of the molecule and associated with the nicking domain. A second deletion (CreMobΔsig2) eliminates the leader region required for the C-terminal transport signal (above). When the deletions were introduced into R1162, both of the resulting plasmids were conjugatively transferred, but at frequencies below that of the intact plasmid (Fig. 2B). These results indicate that both signals can independently support transfer of plasmid DNA. A plasmid deleted for both sig1 and sig2 was not transferred (Fig. 2B). The effect of each deletion on transfer differed in magnitude, with Δsig1 and Δsig2 reducing the frequency 154- and 24-fold respectively. The effect of the deletions on transport was different. Here, Δsig1 had no large effect on transport (the frequency was fourfold higher than with the intact molecule, Fig. 1), but transport of the fusion containing Δsig2 was undetectable. Western gel analysis showed that failure of CreMobΔsig2 to be transported was not due to instability of the protein (Fig. 4).
MobB is a membrane protein
MobB contains the putative transmembrane domain, strongly predicted by the program PhDhtm (Rost et al., 1996), underlined in Fig. 1. We asked whether a C-terminal, his-tagged MobB derivative, which is active in transfer (unpublished), is associated with the membrane. A strain was constructed that contains a plasmid with MobBhis under control of the lac repressor. After induction for 1 h, a cell lysate was first cleared by low-speed centrifugation, and then fractionated by high-speed centrifugation to pellet the membrane fragments and associated proteins (Churchward and Holland, 1972). We then used Western blotting, with antibody directed against the his tag, to examine the partitioning of MobBhis into the supernatant and pellet (Fig. 5A). As a control, his-tagged minMobA was also induced in the cells. This protein, which contains the first 186 amino acids of MobA and is highly soluble after induction, is active for nicking but not for transport or transfer (Becker and Meyer, 2002). The results show that MobB remains associated with the membranous pellet, whereas minMobAhis is mostly in the soluble fraction. The smaller amount of minMobAhis in the MobBhis-containing cells is probably due to the toxicity of MobB after induction.
We had observed that MobB tends to aggregate in solution, and were concerned that the presence of the protein in the membranous pellet reflected microaggregates rather than a true association with the membrane. We therefore further purified the membrane fragments on a sucrose step gradient (Roy and Isberg, 1997). The preparation was then incubated with either buffer or Triton X-100 and recentrifuged. The distribution of MobB in the pellet and supernatant was again assayed by Western blotting (Fig. 5B). The results show that MobB remains associated with the membrane and is released by incubation with Triton X-100, suggesting it is an integral membrane protein. However, MobB did aggregate as well, and could be recovered from the bottom of the sucrose step gradient.
Finally, we determined whether the deletion derivatives of MobB (Fig. 1) were also associated with the high-speed membrane fraction. Only the his-tagged MobBΔ2, which retains the transmembrane domain, remained visibly associated with the membrane fraction (Fig. 5C) when probed with anti-his antibody. Thus, deletion of part or all of the putative transmembranous domain disturbed association of MobB with the membrane fraction.
The large, R1162 MobA contains two separate catalytic centres, an N-terminal nickase, which forms a covalent complex with one of the plasmid DNA strands at the origin of transfer (Scherzinger et al., 1993), and a C-terminal primase, which is highly specific for the two initiation sites within the R1162 origin of replication (Lin and Meyer, 1987; Honda et al., 1991). We have associated with each of these a type IV transport signal active with the R751-encoded transfer apparatus. Neither the N-terminal nor C-terminal halves of MobA contain an obviously similar amino acid sequence that might reflect a simple recognition signal for type IV transfer. In addition, part of the active domain as well as an adjacent region is required for transport in each case. This suggests that the requirements for transport are complex, either for recognition of the coupling protein or for subsequent passage through the transport pore.
Although the two secretion signals are also involved in conjugative transfer (Fig. 2), different mutations do not have co-ordinate effects on the two processes. There could be many possible reasons for this; at the very least, formation of an excisant requires at least four molecules of Cre (Hoess and Abremski, 1984), and thus four active and transported hybrid proteins. In contrast, a single round of conjugative DNA transfer can potentially form a transconjugant, and there is generally more than one round of transfer in our mating pairs (Rao and Meyer, 1994). Therefore, everything else being equal, partially active proteins would have a lesser effect on the frequency of conjugative transfer than on the frequency of transport.
The crystal structure of the nickase domain of the R1162 MobA has recently been solved (Monzingo et al., 2007). The protein is structurally very similar to the relaxases of R388 and the F factor (Datta et al., 2003; Guasch et al., 2003), despite little similarity in primary amino acid sequence. Threading analyses (McGuffin et al., 2000; Torda et al., 2004) to identify possible structural similarities between the R1162 primase and other members of the RCSB protein database (Berman et al., 2000) did not bring up any of the relaxases as proteins likely to have similar structures, but this does not rule out more localized regions of structural similarity.
Secretion of a plasmid-encoded primase is not unprecedented and has been shown for ColIb (Wilkins and Thomas, 2000). In the case of R1162, the primase can initiate synthesis of the complement to the entering strand (Parker and Meyer, 2005), and its early presence in the recipient cell would therefore be an advantage. However, it is remarkable that transport of either the N-terminal or C-terminal halves of MobA require MobB; we have never observed transport from a cell in which MobB was not provided. The bifunctional R1162 MobA seemed most likely to have arisen as the fusion of two separate proteins, a relaxase and primase, due to translational readthrough of the small MobB gene (Fig. 1). It is hard to understand in that case how the primase would come to depend on MobB, particularly because it can be synthesized separately from the relaxase (Scholz et al., 1989) and is not then part of the relaxosome. Further, part of sig2 is located within the presumed linker region joining the two functional halves (Fig. 3). The fusion might in fact be closer to the ancestral condition, with unlinked MobA proteins, such as that encoded by pSC101, secondarily generated deletion derivatives. Consistent with this, the sequence QRQQEKAR, amino acids 580–587 in R1162 MobA and located in sig2, is also found (with one change, A to E) at the C-terminal end of the pSC101 MobA. The pSC101 signal might then have been built at least in part from sig2, following deletion of other regions of the protein.
Under our mating conditions, R751 is transferred very efficiently (about 0.15 transconjugants per donor). Because about 30% of the transconjugants are activated for lacZ when the Cre Mob(324–709) fusion is transported, this means that protein secretion by the R751 T4SS can also be very efficient, particularly because four molecules are required for each activation of the lacZ reporter. The lower activation frequencies for the other fusions could be due to weaker transport of the protein or lower specific activity of Cre in the fusion. In any case, the high transport frequency of Cre Mob(324–709) indicates that there can be a large flux of molecules into the recipient during plasmid transfer.
How is MobB facilitating type IV transport and conjugative transfer? Transfer can still occur, at a frequency well above background, even when MobB contains a large deletion (Fig. 2B), arguing against only MobB having the correct conformation for docking to the coupling protein and thus acting as an adaptor for MobA. This also means that sig1 and sig2 on MobA probably have elements for binding to the coupling protein, MobB, or both; our data do not allow us to distinguish between these possibilities. MobB purifies with the membrane, and this targeting appears important for activity: MobBΔ2his, which is still inserted in the membrane (Fig. 5C), also sustains a high level of transfer (Fig. 3). MobB might then act to stabilize an inherently weak interaction between the R1162 relaxase and different coupling proteins, by acting as a membrane anchor at the site of transport (Fig. 6). It is noteworthy that RSF1010, which is essentially the same plasmid as R1162, has been successfully transferred by a number of different type IV transport machines; A. tumefaciens and Legionella pneumophila alone each encode at least two active on this plasmid (Beijersbergen et al., 1992; Segal et al., 1998; 1999; Chen et al., 2002). R1162, a broad host-range plasmid, might be especially versatile in this regard, because weak interactions with the coupling protein could be amplified both by anchoring to the membrane and by the multiple signalling sites on the MobA protein.
Plasmids and strains
Construction of the reporter strain to detect Cre transport and the plasmids used are described in detail in Supplemental methods. The reporter strain is derived from the E. coli strain TOP10 (Invitrogen). Strains for conjugation and type IV transfer were all derived from MV10 (C600 ΔtrpE5) (Hershfield et al., 1974). For conjugation, the recipient was DF1019, a derivative of C600 resistant to nalidixic acid (Figurski et al., 1976).
Assays for type IV transport of MobA and conjugative transfer of plasmid DNA
All donor strains contained the IncP1 plasmid R751 (Thorsted et al., 1998), which encodes resistance to trimethoprim, to provide the T4SS. To detect protein transfer, we adapted the Cre fusion reporter assay developed by Vergunst et al. (2000). We constructed a streptomycin-resistant recipient strain (Supplemental methods) where Cre-mediated excision resulted in activation of lacZ, detected by plating on medium containing Xgal. Prior to mating experiments, all protein fusions were tested for Cre activity by transforming the reporter strain with plasmid DNA and verifying that the transformants were blue on Xgal.
Cultures of Cre donor strains and the recipient were grown separately overnight in broth medium containing antibiotics to maintain the plasmids, then diluted 10-fold into drug-free medium and incubated further for 90 min. Donor (0.5 ml) and recipient (1.0 ml) were then drawn by gentle vacuum onto a 25 mm diameter, 0.48 μM pore-size nitrocellulose filter membrane, and incubated cell-side up on a broth plate for 2 h at 37°C. The cells were then washed into 1 ml broth and plated at various dilutions onto medium containing trimethoprim (200 μg ml−1), streptomycin (25 μg ml−1) and Xgal (60 μg ml−1). Cre transport was measured as the percentage of (trimethoprim-resistant) transconjugant colonies that were blue in the presence of Xgal. The procedure allowed us to estimate easily the frequency of transfer of the Cre fusion, relative to the conjugative activity of the R751 T4SS in each mating. However, it should be noted that the frequency of blue colonies depends not only on the transport frequency of the fusion proteins, but on their stability and the Cre specific activity for each.
The assay for conjugative transfer was similar, except that cells were mated on filters for 90 min before resuspension and plating. Transconjugants were selected on medium containing nalidixic acid (25 μg ml−1) and streptomycin (25 μg ml−1) or trimethoprim (200 μg ml−1).
CreMob proteins (Fig. 4). A total of 5 ml of mid-log phase cultures was pelleted, washed with 2 ml phosphate buffer containing 137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4, 1.4 mM KH2PO4, pH 7.3, and resuspended in 1 ml of the same buffer containing 0.1% SDS and 1 mM phenylmethylsulphonyl fluoride. The cells were heated to 65°C and cleared by centrifugation at 20 400 g for 15 min. Soluble protein in the supernatant was determined by the Bio-Rad Protein Assay. In total, 20 μg of protein was precipitated by adjusting the volume to 200 μl and adding 800 μl acetone. The samples were kept at −20°C for at least 1 h, and the precipitated protein collected by centrifugation at 15 800 g for 20 min and at 4°C. The pellets were air-dried, then dissolved in 10 μl of the phosphate buffer. A total of 5 μl of sample buffer (150 mM Tris HCl, 6% SDS, 0.5% bromphenol blue, 30% glycerol) was added, and the samples were boiled for 5 min. The proteins were resolved on a 4–15% linear gradient polyacrylamide gel.
The separated proteins were transferred to a nitrocellulose filter in a Bio-Rad Mini Transfer Cell at 30 V overnight at 4°C. The filter was then processed as described in the Promega Technical Manual for the Protoblot II System. Primary, rabbit anti-Cre antibody and secondary goat anti-rabbit antibody conjugated to alkaline phosphatase were obtained from Novagen and used at 1:10 000 dilution, according to the manufacturer's instructions. Western transfers were developed using the Alk Phos detection kit (Novagen).
MobBhis and minMobAhis proteins (Fig. 5). E. coli strains M15 expressing MobB or deletion derivatives in the his-tag vector pQE60 (Qiagen) were used for detection of MobB in membrane. The activity of the MobBhis derivative was first confirmed by complementation of R1162 mobBΔ1 (Fig. 2B) for transfer. In total, 250 ml cultures was grown to OD540 = 0.8, induced for 1 h with 0.5 mM IPTG, and the cells then collected in 8–10 ml CBB (20 mM Tris, pH 8, 25 mM NaCl, 5 mM Na2EDTA and 3.6 mM 2-mercaptoethanol) (Hale and de Boer, 1997). The cells were disrupted by sonication and the lysates cleared by centrifugation at 12 000 g for 5 min. The supernatants were then centrifuged at 48 000 g for 30 min, and the pelleted membrane fractions resuspended in 200 μl CBB by brief sonication. Both supernatants and resuspended pellets were stored at −20°C. Approximately equal amounts of protein, judged from a Coomassie-stained gel, were applied to a 10% Tricine gel and transferred to nitrocellulose as described above. Proteins were detected with AP-conjugated monoclonal anti-his antibody (US Biochemical).
Membranes were further purified by centrifugation through 25% sucrose and collection on a 60% sucrose shelf according to the procedure of Roy and Isberg (1997). A total of 2 ml of CBB was then added, and the membranes pelleted by centrifugation at 48 000 g for 30 min. Membrane fractions were incubated in an equal volume of either CBB or CBB + 1% Triton X-100 for 1 h at room temperature, then recentrifuged at 48 000 g for 30 min.
We thank P.J.J. Hooykaas for providing pSDM3204 used in the construction of the CreMob derivatives, and Dorian Henderson and Tariq Perwez for constructing several of the plasmids.