Discussion

In the work presented in this article, the pAD1 relaxase structure was modeled, showing to contain two structural domains. The C-terminal domain belongs to the PD-(D/E)XK family of REs. The acidic residues D152, E170, and E172 occur in a similar position as the DEE active site residues of BamHI, suggesting a role for them in the coordination of two Mg⁺² cations that activate the nucleophilic attack of the phosphodiester bond at the nic site (18). Actually, site-directed mutagenesis of these residues in TraX_pAD1 abolished DNA transfer as well as in vivo nic cleavage. The basic residues K176 and R180 are also conserved in the BamHI-like REs. In fact, site-directed mutagenesis of both residues showed undetectable nicking activity and a drastic reduction in conjugative transfer. At the C-terminal end of β4, we find Y203, a conserved residue in the MOB_C family of proteins (Figs. 1 and 2). This polar amino acid is near the active site. It could be involved in the conformation of the active site or the interaction with the DNA. TraX Y203A was not able to catalyze in vivo cleavage, although TraX Y203A mutant was able to catalyze DNA transfer in vivo. As the approach used to check cleavage in vivo relies on primer extension on relaxosome complexes isolated from cells, it might not be sensitive enough to detect a potential residual cleavage activity of TraX Y203A. A reduced cleavage efficiency showed by this mutant would contribute to its lower conjugation frequency (a 100-fold lower transfer frequency compared with WT TraX).

An interesting residue is Y181, in which a mutation to Ala abolished pAD1 conjugal transfer. By comparison with other known conjugative relaxases of the MOB_F or MOB_P systems, we predict that Y181 could be the catalytic tyrosine that remains covalently attached to the 5′ end of the nic site after cleavage. In fact, Y181 is essential for pAD1 conjugation (Table 1). Moreover, the relative locations of the acidic residues and the catalytic tyrosine are similar to the organization found in typical relaxases (i.e., HUH relaxases). HUH relaxases are related to rolling-circle replication proteins and a specific type of transposases. RE homologues have also been found to be involved in rolling-circle replication [Rep of plasmid pUA140 (25)] and transposition [TnsA transposase (26)]. Therefore, conjugative relaxases come from HUH replication proteins or REs. This structural similarity suggests an adaptive functional convergence of these two completely separate protein folds. However, no covalent protein–DNA intermediate could be found in CloDF13 (14), and no catalytic Tyr is required in the restriction enzymes in which the phosphodiester bond is cleaved by hydrolysis. The presence of a stable covalent intermediate has not been observed for other replication initiator proteins such as RepB (27) and gpII (28). However, later observations concluded that indeed there is a covalent complex in filamentous phage that dissociated rapidly unless gpII was inactivated (29). Thus, gpII possesses activities to form the covalent bond and to break it. RepB was also predicted to form an unstable covalent complex by the study on the chirality of phosphorothioate in the strand transfer reaction (30). Therefore, so far, all the characterized RCR replicases or conjugative relaxases require a covalent intermediate. This phosphotyrosine covalent complex is necessary to have the 5′ phosphate activated at the recipient cell so that the second reaction could be performed. We cannot envision how conjugative transfer could terminate without the relaxase being covalently bound to the DNA after the first cleavage reaction.

PD-(D/E)XK restriction enzymes have been found to have two metal ions at the active sites. BamHI active site contains two divalent cations at the active site that are separated by approximately 4 Å and are coordinated by Glu-77, Asp-94, and Glu-111 (31). PD-(D/E)XK nucleases hydrolyze the phosphodiester bond by direct inline nucleophile attack, resulting in the inversion of configuration at the phosphorous atom (32). When the location of the TraX conserved catalytic residues was compared with BamHI, we observed that they lay in the same spatial position. Thus, we propose the catalytic reaction to proceed in TraX as depicted in Fig. 5 by comparison with the BamHI catalytic model proposed in a previous work (31). According to this model, two divalent cations are also involved in the TraX reaction. The metal ion A is coordinated by E170, D152, and the carbonyl oxygen of I171, whereas the metal B could be bound by D152 and D138. Strikingly, when the position of the TraX Tyr181 was compared, we observed that the hydroxyl group is located where the attacking water is found in BamHI. We propose that no water is involved in the nucleophilic attack by TraX. Instead, the hydroxyl group of Y181 is perfectly positioned for the catalysis, resulting in a covalent bond between the phosphate group of the cleaved DNA bond and the Y181. In our model, E172 could be the general base that activates the hydroxyl group of Y181 by proton abstraction (Fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Catalytic mechanism of BamHI, TraX, and TrwC. Schematic diagram illustrating the proposed catalytic mechanism for the DNA cleavage reaction performed by BamHI (Left), TraX (Center), or TrwC (Right) (Discussion).

When the purified mutant proteins were checked for dsDNA binding activity, we observed (Fig. 4B) that all mutant proteins were able to bind dsDNA, except mutant E28A. According to the prediction, all the conserved residues that were mutated are located in the ssDNA binding and catalytic domain of TraX, whereas the E28 is located in the N-terminal domain, explaining this difference. The N-terminal domain of TraX is predicted to have approximately 110 aa with a topology related to the DNA binding domain of MarR (16, 17). E28A occurs at the C-terminal end of α2 helix. The main interactions with the DNA are predicted to be in the N-terminal end of α4 and even α2 and the loop connecting β1 and β2 . However, the conserved Glu28 could be involved in the stabilization of the winged-helix DNA binding domain, and this is why this mutant is not able to bind the DR sequences. Consequently, TraX E28A was not able to promote conjugation in the complementation experiment. Interestingly we have found that the three systems most closely related to pAD1, namely pAM373, TX1341, and VRS11, are conserved in the nic region but not in the direct repeat region (Fig. S1). This conservation in the binding sites perfectly correlates with conservation in the protein domains. C-terminal domains are well conserved in the four systems, whereas the N-terminal domains and, more specifically, the DNA interacting α-helices, are quite different.

After nic-cleavage in the absence of recipient cells, typical HUH relaxases remain bound to both sides of the nic site and the 3′OH is not free. In the case of pAD1, TraX catalyses cleavage of the nic site in a process strictly controlled by the presence of pheromone (cAD1). In a previous work, we demonstrated that nic cleavage was only visible when cells were exposed to cAD1 (9). Also in this regard, preliminary experiments show that relaxase expression is detected only in the presence of cAD1. Therefore, there is evidence that pAD1 DNA processing for transfer is triggered when in close proximity to potential recipient cells (Fig. S2). Only then is enough pheromone sensed by the donor cells, which, in response, activate the plasmid mating specific functions, including the relaxase.

We have reported here that TraX is able to bind dsDNA fragments containing the full oriT2 region or just the region containing the five DRs, but not the nic region (Fig. 4A). Moreover, TraX-specific ssDNA binding was not observed. Thus, we think MOB_C relaxases are not able to bind the double-stranded nic region unless the binding of the N-terminal winged-helix DNA binding domain to the DRs at oriT2 allows the creation of a partially melted region around the nic site. This domain could be necessary in these relaxases to recognize and cleave only one of the DNA strands, whereas dimeric REs without this second domain recognize and cleave both strands of a palindromic sequence. In fact, the RCR initiation protein RepB is also formed by a helix-turn-helix (HTH) DNA binding domain that recognizes three direct repeats and a HUH endonuclease domain to exert the specific ssDNA cleavage. Interestingly, although it has the catalytic Tyr and HUH motif, no covalent intermediate has been directly found in RepB, as previously discussed. After the binding of the N-terminal domain to the DR region, the RE-like binding domain would recognize and bind the ssDNA nic region catalyzing the hydrolysis of the nic site, which would initiate DNA transfer from donor to recipient (Fig. S2). Although there is no experimental evidence of TraX covalently bound to the 5′ end of the DNA to be transferred, we propose that, like in the other characterized conjugative systems, TraX will be transferred to the recipient cell by the T4SS piloting the DNA. Then, after one cycle of RCR, a TraX-catalyzed second cleavage reaction in the donor cell will create the free 3′OH to recircularize the plasmid in the recipient cell.

According to our results, not all the relaxases involved in plasmid conjugation converged into the same structural fold. MOB_F, MOB_P, MOB_Q, and MOB_V family showed the HUH fold (3, 33), but MOB_C present the RE fold. We think the different fold mainly results from the different triggering mechanism. In HUH relaxases, conjugative transfer is triggered by the binding of the recipient cell to the donor cell. Until this cell-to-cell interaction occurs, the HUH relaxase is bound to the oriT, but the nic-cleavage and joining activities are in a tight equilibrium in such a way that the plasmid is not relaxed unless the conjugation signal is received. In the case of RE relaxases like pAD1_TraX, conjugation is triggered by the pheromones produced by the recipient cell and detected by the donor cell. When the pheromone has reached the donor cell, relaxase expression is induced along with its ability to introduce the nic-specific cleavage generating the plasmid relaxation and releasing a free 3′OH that is used by the cellular polymerase to produce by RCR the ssDNA strand that is transferred. Any reaction performed by the RE relaxase will produce a relaxed plasmid with the relaxase covalently bound. If this covalent product is not transferred to the recipient cell, the plasmid could be unstable. The quick hydrolysis of the covalent complex could be the mechanism to protect the plasmid DNA unless the transfer of the relaxases produces a conformational change that avoids the hydrolysis. The conjugation of some HUH relaxase-containing plasmids, such as pCF10 (34), is also induced by pheromones, but the need for accessory proteins could prevent the relaxation of the plasmid in the absence of pheromone.

In summary, the results shown in this article define a unique type of conjugative relaxases. DNA processing by RE relaxases is, in general terms, similar to processing by HUH relaxases. Thus, both relaxases should covalently bind the plasmid DNA to be transferred to the recipient cell to perform the termination reaction. The catalytic mechanism is also similar, although the structure of the relaxase is different. This is a beautiful case of adaptive convergence, whereby two radically different proteins (HUH and RE) functionally converge to create similar constellation of residues within the catalytic center and thus become able to catalyze the same biological reaction.