Towards an integrated model of bacterial conjugation

Bacterial conjugation is one of the main mechanisms for horizontal gene transfer. It constitutes a key element in the dissemination of antibiotic resistance and virulence genes to human pathogenic bacteria. DNA transfer is mediated by a membrane-associated macromolecular machinery called Type IV secretion system (T4SS). T4SSs are involved not only in bacterial conjugation but also in the transport of virulence factors by pathogenic bacteria. Thus, the search for specific inhibitors of different T4SS components opens a novel approach to restrict plasmid dissemination. This review highlights recent biochemical and structural findings that shed new light on the molecular mechanisms of DNA and protein transport by T4SS. Based on these data, a model for pilus biogenesis and substrate transfer in conjugative systems is proposed. This model provides a renewed view of the mechanism that might help to envisage new strategies to curb the threating expansion of antibiotic resistance.

INTRODUCTION

Over the last half a century, human health has profited from the discovery and use of antibiotics to fight against infectious diseases. However, after many years of extensive use, bacteria fight back and our arsenal of antibiotics seem not to be enough to cope with the avalanche of multi-resistant new pathogens (McGowan and Tenover, 2004; Payne et al., 2007; Hawkey, 2008; Boucher et al., 2009). Although new antibiotics are emerging in the pipeline (Fischbach and Walsh, 2009; Gould and Bal, 2013), very few have reached the market in the last four decades. One of the main mechanisms whereby bugs become resistant to antibiotics is the acquisition of antibiotic resistance genes by bacterial conjugation (Mazel and Davies, 1999; Waters, 1999; de la Cruz and Davies, 2000). Therefore, the search for specific conjugation inhibitors is of special interest in the fight against these superbugs. In this context, the recent development of new Eco-Evo drugs has opened a promising strategy to control plasmid dissemination using conjugation inhibitors (Baquero et al., 2011).

Conjugation is a process by which genetic material is transferred from a donor cell to a recipient cell. The transfer of these conjugative genes requires a sophisticated machinery that ensures DNA mobilization [brought about by MOB genes (Garcillan-Barcia et al., 2009; de la Cruz et al., 2010)] and mating pair formation (organized by the MPF genes). These genes can be encoded by an autonomous replicating plasmid (Smillie et al., 2010) or by integrative conjugative elements (ICE) inserted in the chromosome (Wozniak and Waldor, 2010; Guglielmini et al., 2011). Conjugation in Gram-negative bacteria is mediated by the Type IV secretion system (T4SS), a large macromolecular complex involved in substrate transport and pilus biogenesis. T4SSs are implicated not only in bacterial conjugation, but also in the secretion of virulence factors to eukaryotic cells. Many effectors secreted by T4SS are virulence factors involved in pathogenic diseases, such as brucellosis, whopping cough, cat scratch disease, pneumonia or gastric ulcer, caused by bacterial infection with Brucella suis, Bordetella pertussis, Bartonella henselae, Legionella pneumophila or Helicobacter pylori, respectively (Vogel et al., 1998; Burns, 2003; Seubert et al., 2003; Schulein et al., 2005; Nystedt et al., 2008; Engel et al., 2012; Myeni et al., 2013). Excellent reviews on T4SSs have appeared recently (Waksman and Fronzes, 2010; Christie et al., 2014; Waksman and Orlova, 2014), and we are not going to dwell on the general architecture of these systems. Instead, we will focus on the issues that pertain specifically to conjugative systems, analysing recent advances that have increased our understanding on the structure and function of conjugative T4SS.

Conjugative systems can be considered as a particular subfamily of T4SS. This subfamily is unique among bacterial secretion systems as it is not only able to transport protein effectors but also DNA (covalently bound to a pilot protein) [for recent reviews see (de la Cruz et al., 2010; Zechner et al., 2012)]. DNA transfer during bacterial conjugation provides plasticity to bacterial genomes, constituting one of the main mechanisms of horizontal gene transfer (Frank et al., 2005). Conjugative T4SSs are essential for mating pair formation (MPF) and, therefore, genes encoding T4SS proteins are called MPF genes. Conjugative plasmids encode their own set of MPF genes and are, hence, self-transmissible. Other plasmids lack their own set of MPF genes and depend on the T4SS of a conjugative plasmid to be mobilized. Accordingly, four distinct MPF families could be described in Proteobacteria: MPF_F (characterized by the conjugative plasmid F), MPF_I (referring to IncI plasmids like RP4), MPF_T, (based on Agrobacterium tumefaciens T_i plasmid) and MPF_G (related to ICEs) (Smillie et al., 2010). Signatures of some MPF genes have also been found in Cyanobacteria, Firmicutes, Bacteroides, Actinobacteria and Archaea (Guglielmini et al., 2011), suggesting that conjugation is an ancient process that arose very early among ancient Proteobacteria and then it spread out to all clades of prokaryotes (Guglielmini et al., 2012).

MPF_T systems encode the simplest T4SS, consisting of 11 proteins, named VirB1 to 11 after A. tumefaciens T4SS (Christie et al., 2005). These 11 proteins, together with the coupling protein VirD4, assemble into a macromolecular complex that spans the inner and outer membrane and the periplasm in between. The genes encoding these proteins are organized in operons under the control of different promoters, with a clear distinction between those involved in mating pair formation (Mpf) and DNA transfer replication (Dtr) (Fernández-López et al., 2006; de la Cruz et al., 2010). Figure 1 shows the organization of these genes in R388 plasmid, which together with F and A. tumefaciens Ti plasmids are the best characterized conjugative systems. This gene architecture is relatively well conserved in most conjugative systems (Fernández-López et al., 2006; Guglielmini et al., 2012), which is a striking feature considering that gene synteny is an evolutionary trait that is quickly lost in evolution (Mushegian and Koonin, 1996; Itoh et al., 1999). Synteny conservation is also observed in other systems involved in the assembly of large multi-subunit complexes, such as the bacteriophage capsid (Brussow and Hendrix, 2002) or the archaeal (Patenge et al., 2001) and eubacterial flagella (Canals et al., 2006). Therefore, synteny conservation seems to have an adaptive value, perhaps harnessing information related to the temporal programme of expression of the elements of the complex.

T4SS architecture is relatively well preserved regardless of its biological function (bacterial conjugation, secretion of virulence factors into eukaryotic cells or DNA uptake) (Waksman and Fronzes, 2010). Four protein domains can be distinguished in T4SS: the pilus, the core channel complex, the inner membrane platform and the hexameric ATPases at the base of the channel that supply the energy for pilus biogenesis and substrate transport (see Fig. 2 for subunit composition). A near complete structure of T4SS from R388 plasmid has been recently published (Low et al., 2014), revealing the general architecture of the secretion system (Fig. 2). This new structure has provided new insights into T4SS assembly. However, some of the components of the T4SS, such as the pilus or the VirB11 traffic ATPase, are missing in the structure. Here, in the light of this new structure and the genetic, biochemical and structural information provided by numerous laboratories along the last years, a description of the architecture, biogenesis and function of the four different T4SS domains is provided.

PILUS BIOGENESIS

A prevalent way for bacteria to accomplish adhesion to recipient cells is by the use of external appendages of different sizes but similar helical structure known as pili (Bradley and Cohen, 1976; Achtman et al., 1978; Lai and Kado, 2000; Schroder and Lanka, 2005; Thanassi et al., 2012). At the distal end of these filamentous structures, there is a molecule called adhesin (Aly and Baron, 2007), which plays an essential role in cell-to-cell contact (Anthony et al., 1994; Dehio, 2004). The major component of the T4SS pilus is a hydrophobic protein named pilin or VirB2 (TrwL in plasmid R388) (Fullner et al., 1996; Lai and Kado, 1998), being the adhesin VirB5 (TrwJ in R388) a minor component (Schmidt-Eisenlohr et al., 1999). Although pilus biogenesis in T4SS is an intricate process that is not well understood, a picture is emerging in which different steps leading to pilus formation can be distinguished (Fig. 3).

Step1. Prepilin processing

Pilin proteins (VirB2) are synthesized as prepilin precursor molecules, which contain an N-terminal leader signal sequence. Once in the inner membrane, the N-terminal leader sequence is processed by a signal peptidase I, presumably LepB (Majdalani et al., 1996; Eisenbrandt et al., 2000; Lai et al., 2002). In some VirB2 homologues, the N-terminal residue of the cleaved protein binds covalently to the C-terminus, resulting in a mature, cyclic pilin (Eisenbrandt et al., 1999; Lai and Kado, 2002). However, not all VirB2 homologues adopt this cyclic conformation. Some VirB2-homologues, such as TraA of plasmid F, retain their linear sequence, probably by modification of the N-terminus by N-acetylation (Moore et al., 1993). Pilin acetylation appears to ensure the correct assembly of the F-pilus filament (Anthony et al., 1999).

Step 2. Pilin extraction from the inner membrane

Because of their hydrophobic nature, pilin molecules can partition into the lipid bilayer without the additional help of any other component of the T4SS (Paiva et al., 1992). Once the protein is accumulated in the inner membrane and, upon reception of an unknown signal that triggers conjugation (Silverman, 1997; Hwang and Gelvin, 2004), pilin molecules are dislocated from the membrane towards the periplasmic space with the assistance of VirB4/TrwK and VirB11/TrwD ATPases (Kerr and Christie, 2010). VirB4 seems to act as the primary dislocation motor, whereas VirB11 would act as a modulator of the VirB4 dislocase activity, inducing conformational changes in the pilin.

VirB5 is essential for the incorporation of VirB2 to the pilus (Schmidt-Eisenlohr et al., 1999; Lai et al., 2000). VirB5 stable expression is dependent on the presence of VirB6 (Hapfelmeier et al., 2000), a polytopic inner membrane component of T4SS. VirB5 localizes at the pilus distal end (Aly and Baron, 2007) although it has also been found in the cytoplasm and the inner membrane fractions of pili-producing bacteria (Thorstenson et al., 1993). The crystal structure of TraC, the VirB5 homologue in the conjugative plasmid pKM101, revealed an elongated α-helical structure which, presumably, is involved in cell-to-cell contact (Yeo et al., 2003; Hwang and Gelvin, 2004). Interestingly, in contact independent T4SS, such as that of Bordetella pertussis, VirB5 is absent, but not a VirB2-like protein (PtlA) (Weiss et al., 1993; Craig-Mylius and Weiss, 1999). As VirB2, VirB5 also possesses an N-terminal leader signal, likely to be processed by a signal peptidase I. However, in contrast to VirB2, the mature VirB5 is not hydrophobic and, therefore, it is not expected to partition into the inner membrane. Upon cleavage of VirB5 leader sequence, the mature protein would be exported to the periplasm. This step is likely to trigger pilus assembly.

Steps 3 and 4. Pilus elongation

Packing geometry of VirB2 into the assembled pilus would be in such a way that hydrophobic patches of the protein will be buried at the interface between adjacent pilin monomers (Marvin and Folkhard, 1986). These hydrophobic-driven interactions would promote the oligomerization of pilin molecules into bundles which, in turn, would result in the formation of the pilus, as observed in the structure of F-pili (Wang et al., 2009). The mechanism driving conjugative pili assembly remains unclear, and it is unknown whether pili assembly initiates from the inner membrane platform at the base of the T4SS or from the cap complex at the outer membrane. Pilus polymerization from the inner membrane might be driven by a mechanism similar to that of T2SS and Type IV pilus assembly (reviewed in Filloux, 2004; Craig and Li, 2008; Thanassi et al., 2012). This model of pilus growth from an inner membrane base platform is supported by the finding of VirB2 in both the inner and outer membranes in cell fractionation experiments (Shirasu and Kado, 1993). Interestingly, in Type II secretion systems, the minor pseudopilins nucleate pilus assembly (Cisneros et al., 2012), reinforcing the idea that VirB5 might also act by initiating filament assembly in T4SS. However, this model of pilus elongation from the inner membrane would be incompatible with the recently solved structure of T4SS from conjugative plasmid R388 (Low et al., 2014) in which a central stalk formed by VirB6/VirB8 subunits rises from the inner membrane occupying the putative position of a pilus at this location (Fig. 2). On the basis of this structure, pilus biogenesis might only occur in the periplasm, with the VirB8/VirB6 central stalk acting as a nucleation platform.

Little is known about the size and morphology of VirB-like conjugative pili. In contrast, the morphology of F-pili is well characterized (Wang et al., 2009). They are flexible, long appendages (up to 20 μm), with an outside diameter of 8.5–9.5 nm and a central lumen of 3 nm. The outer diameters of Type IV (T4P), Type I and P pili are also around 6 to 9 nm wide (Hahn et al., 2002; Mu and Bullitt, 2006; Craig and Li, 2008). Therefore, VirB-like conjugative pili might have similar outer dimensions. Considering the inner diameter of the T4SS outer membrane ring, which is 2–3 nm, (Chandran et al., 2009; Fronzes et al., 2009; Low et al., 2014), it is difficult to envisage pilus polymerization within the secretion channel without a big conformational change in the distal end of the core complex. Therefore, an alternative scenario in which pilin polymerization occurs at the outer membrane is also plausible. That is the case, for instance, of the Type I pilus in the chaperone usher (CU) pathway (Allen et al., 2012), where the folding energy released by each pilin subunit drives polymerization (Fig. 3).

In the Type IV pilus (T4P) apparatus, pilus extension and retraction are mediated by PilB and PilT, two hexameric ATPases located in the inner membrane (Merz et al., 2000; Mattick, 2002; Craig and Li, 2008). In contrast to the T4P systems, conjugative pilus retraction has been a puzzling issue for a long time. Demonstration that conjugative pilus can retract came from fluorescent experiments in which F-pilus retraction could be visualized in real time (Clarke et al., 2008). However, evidence for retraction of short-rigid pili like those of R388-like conjugative systems is still missing. It has been suggested that F-pilus retraction does not require energy (Frost et al., 1985). Instead, F-pilus assembly and retraction would be similar to the filamentous phage ‘telescopic inovirus’ model (Marvin, 1989), whereby pilus rotation along its longitudinal axis drives assembly and retraction (Clarke et al., 2008).

ASSEMBLY OF THE SECRETION CHANNEL (CORE COMPLEX)

In recent years, several advances have allowed to gather a glimpse on the structural architecture of T4SS (Chandran et al., 2009; Fronzes et al., 2009; Low et al., 2014), providing information on the organization of the secretion channel. This channel is formed by a large structure (1.05 MDa), known as the core complex, which spans across the inner and outer membranes and the periplasm in between. The structure comprises 14 monomers of three distinct proteins (VirB7, VirB9 and VirB10). Below, some of their most salient properties are described.

VirB7 is a small lipoprotein inserted in the outer membrane (Fernandez et al., 1996a) firmly associated by disulphide bridges with VirB9 (Anderson et al., 1996; Bayliss et al., 2007), which is mainly involved in maintaining the integrity of the core complex (Fernandez et al., 1996b). VirB7 contains a globular domain that shares structural similarities with the N0 domains of Type IV pilus PilQ, T2SS GspD, T3SS EscC and filamentous phage pIV secretins (Souza et al., 2011).

VirB9 is a hydrophilic protein also located at the outer membrane. The C-terminal domain of VirB9 contains a β-sandwich fold around which VirB7 binds, forming a stable outer membrane channel complex (Bayliss et al., 2007). VirB9 interacts with VirB10 and, hence, contributes to the stabilization of the secretion channel (Beaupre et al., 1997). The crystal structure of this outer membrane complex formed by VirB7, VirB9 and the C-terminal end of VirB10 has recently been solved (Chandran et al., 2009), revealing a cap structure with an inner pore of 3.2 nm.

VirB8 is another component traditionally associated with the channel complex (Das and Xie, 2000; Kumar et al., 2000). VirB8 is anchored to the inner membrane by its N-terminal region (Thorstenson and Zambryski, 1994; Buhrdorf et al., 2003). The three-dimensional structure of a periplasmic fragment (residues 77-239) of the VirB8 protein of B. suis consists of a large extended β-sheet and five α-helices, with an overall globular fold (Terradot et al., 2005). VirB8 interacts with VirB9 and VirB10 (Das and Xie, 2000) which suggests that it plays an essential role in the assembly of the secretion channel (Kumar et al., 2000; Judd et al., 2005). In the recent EM structure of the T4SS from conjugative plasmid R388 (Low et al., 2014), VirB8 appears to be associated with VirB6 forming a central stalk structure that occludes the pore of the secretion channel (Fig. 2). This disposition of the central stalk formed by VirB8/VirB6 does not seem to be compatible with a channel able to secrete any type of protein or DNA substrate. It is possible that this structure represents an intermediate step during T4SS and pilus assembly in which the central stalk plays an essential role. Then, upon completion of T4SS biogenesis, this central stalk could be unplugged from the channel pore, allowing the passage of substrates. This notion is supported by the fact that VirB8 seems to be loosely attached to VirB9 and VirB10, as reflected by its absence from the EM structure of the core complex, despite of being initially co-expressed (Fronzes et al., 2009). In summary, VirB8 could be an assembly factor, needed for complex assembly but not a component of the secretion channel.

VirB10 is the largest protein in the core complex, spanning the inner and outer membranes and the periplasmic space (Jakubowski et al., 2009). VirB10 shares structural similarities with TonB-like proteins (Cascales and Christie, 2004a), such as a common bitopic membrane topology and a prolin-rich extended region in the periplasm (Evans et al., 1986). These structural features were later confirmed by the crystal structure of the periplasmic domain (residues 146-376) of ComB, the VirB10 homologue in H. pylori (Terradot et al., 2005). In this crystal structure, the VirB10 periplasmic domain was shown to form dimers, confirming data obtained by two-hybrid experiments (Ding et al., 2002). Therefore, given that there are 14 copies of VirB10 in the core structure (Fronzes et al., 2009), it could be assumed that there is a heptameric repetition of the dimeric unit observed in the crystal structure of ComB periplasmic domain. However, the subunit packing observed in this structure, with the N-termini of each monomer at opposite ends, is not compatible with that observed in the core complex structure.

One of the most interesting features of the EM structures of the T4SS (Fronzes et al., 2009; Low et al., 2014) is the size of the inner channel cap constriction at the outer membrane domain of the core complex (10–20 Å). Such a small size implies that any protein substrate crossing the channel should be in an unfolded state, unless a large aperture of the channel takes place. Interestingly, the crystal structure of the outer membrane part of the core complex, formed by 14 copies of VirB7, VirB9, and the C-terminal domain of VirB10 (Chandran et al., 2009) revealed a much larger dimension of the inner ring (32 Å). Therefore, the EM and the crystal structures significantly differ in the inner diameter dimensions of the pore. Comparison of the outer membrane cap crystal structure (Chandran et al., 2009) with that of the EM core complex (Fronzes et al., 2009) revealed that the outer transmembrane helices of VirB10 in the crystal structure were in a ‘relaxed’ conformation, forming a 45° angle relative to the vertical disposition observed in the EM structure. This conformational change could be due to the fact that the cap crystal structure was obtained after removal by proteolysis of the N-terminal half of VirB10. Interestingly, comparison between the EM structures of the pKM101 core complex (Fronzes et al., 2009) and R388 (Low et al., 2014) revealed a flexibility of the VirB10 N-terminus, which fluctuates from a compacted conformation under the I layer of the core complex and an extended conformation in the R388 structure. These conformational differences could reflect natural occurring transitions but more effort is needed to understand the arrangement of the VirB10 N-terminus within the core complex and its interactions with the stalk domain formed by VirB8/VirB6 and, probably, VirB3. In any case, it seems that VirB10 N-terminus plays an essential role in the transmission of the signal from the inner membrane to the rest of the core complex, which could result in the opening of the secretion channel.

ATP-dependent conformational transitions have been reported for VirB10 (Cascales and Christie, 2004). Protease susceptibility experiments showed that VirB10 can adopt two different conformations: ATP-independent and ATP-dependent ‘energized’ forms, although only the energized conformation is able to form a complex with VirB9. Such structural transitions occur in response to ATP utilization by the VirD4 and VirB11 ATPases (Cascales and Christie, 2004). Interactions of VirB10 with the coupling protein VirD4 have been reported (Gilmour et al., 2003; Llosa et al., 2003; Cascales et al., 2013). These interactions take place in the inner membrane (de Paz et al., 2010), which suggests that conformational changes at the base of the channel are transmitted across the periplasmic and OM domains of VirB10, leading to the aperture of the external cap. Interestingly, heteromeric transmembrane interactions in VirB10 have been found to be essential for T-pilus biogenesis but not for substrate secretion (Garza and Christie, 2013).

VirB10 is also able to interact with VirB4. A low-resolution EM structure of the core complex bound to the VirB4-homologue of the conjugative plasmid pKM101 revealed an intimate relation between VirB4 and the channel complex (Wallden et al., 2012). However, the recently solved structure of the T4SS from R388 shows that the VirB4 N-terminus is not embedded in the I layer of the core complex. Instead, an association to the complex through the integral membrane protein VirB3 seems a more plausible explanation. This idea is supported by the finding of VirB3–VirB4 fusion proteins in the IncX branch (Batchelor et al., 2004). It is enticing to speculate that the ATPase activity by VirB4 induces conformational changes in the core complex during pilus biogenesis whereas VirD4 does it during substrate transport.

MOLECULAR MOTORS THAT DRIVE PILUS SYNTHESIS AND SUBSTRATE TRANSFER

Type IV secretion assembly and substrate transport are processes that require energy, usually coming from ATP hydrolysis catalysed by specific ATPases (Peña and Arechaga, 2013). VirB4 and VirB11, two ATPases encoded by T4SS (Rivas et al., 1997; Arechaga et al., 2008), have been suggested to be involved in pilus biogenesis and substrate transport (Rabel et al., 2003). In addition, conjugative systems have a third ATPase, VirD4 (Tato et al., 2005), termed coupling protein (T4CP) as it couples relaxosome DNA processing to the transport of the nucleoprotein complex across the T4SS (Cabezón et al., 1997; Llosa et al., 2002).

VirB4

VirB4 is the only ATPase that is present in all T4SS and, therefore, the presence of a VirB4-like protein constitutes a signature of a T4SS (Alvarez-Martinez and Christie, 2009). For instance, in F-plasmids, VirB4-like TraC is the only ATPase required for pilus production (Lawley et al., 2003). VirB4 proteins are also the largest and most conserved constituents of T4SS (Fernández-López et al., 2006; Guglielmini et al., 2012). The protein consists of two distinct domains: an N-ter (NTD) and a C-terminal domain (CTD). The CTD of VirB4 proteins is ancestrally related to the Type IV coupling proteins T4CP (Guglielmini et al., 2012) and contains the Walker A and Walker B motifs, responsible for the ATPase activity (Arechaga et al., 2008). The crystallographic structure of the CTD of a VirB4 homologue in Thermoanaerobacter pseudethanolicus has recently been reported (Wallden et al., 2012). The structure turned out to be strikingly similar to the computer generated models of the CTD of A. tumefaciens (Middleton et al., 2005) and conjugative R388 VirB4 homologues (Peña et al., 2011), which were built using the structure of TrwB (Gomis-Ruth et al., 2001), the VirD4 homologue in plasmid R388 (see Fig. 1), as a template. VirB4_CTD also shares a high degree of structural similarity with the CTD of FtsK and the soluble domain of TrwB (Cabezón et al., 2012). FtsK and TrwB are DNA pumps involved in bacterial chromosome segregation and bacterial conjugation, respectively. These three motor proteins seem to have evolved from a common ancestor (Iyer et al., 2004; Guglielmini et al., 2012) and, interestingly, a DNA-binding activity has been recently demonstrated in different VirB4 homologues (Li et al., 2012; Peña et al., 2012). At the C-terminal end of VirB4 proteins, there is a conserved α-helical region, which is absent in TrwB. This α-helical region is a regulatory domain that down-regulates the ATPase activity of the protein, probably to prevent a futile waste of energy in the absence of a biological activity, with a mechanism that resembles the regulation of other ATPases (Peña et al., 2011).

In contrast to the CTD, there is little information about the NTD of VirB4 proteins. The IncX subfamily of VirB4 proteins is characterized by the presence at the N-terminus of an extra region corresponding to a fusion with a VirB3 protein (Batchelor et al., 2004). VirB3 proteins are polytopic inner membrane proteins that require VirB4, VirB7 and VirB8 for stabilization (Mossey et al., 2010). VirB3 has also been found to co-elute with VirB2 and VirB5 in membrane fractionation experiments (Yuan et al., 2005), suggesting that VirB3 is essential in pilus biogenesis. VirB4 has been reported to be an integral cytoplasmic membrane protein (Dang and Christie, 1997). However, computer predictions of VirB4 topology are negative for transmembrane spans in most of the VirB4 members, excluding those of the IncX branch, which contain a VirB3-like sequence at the N-termini (Arechaga et al., 2008). Moreover, TraB, the VirB4 homologue in the conjugative plasmid pKM101, has been isolated both, as a soluble and as a membrane-associated form (Durand et al., 2010). It is altogether very likely that VirB3 acts as an anchor of VirB4 to the membrane, assisting it in its VirB2 dislocase function (Kerr and Christie, 2010).

Although there is not any high-resolution structure of the VirB4_NTD available, a low-resolution structure of the full-length VirB4 homologue of the conjugative plasmid R388 has been reported (Peña et al., 2012). This structure was obtained in its hexameric form, which is likely to be the catalytic active conformation of the enzyme (Arechaga et al., 2008). The oligomeric state of VirB4 proteins has been largely under dispute. Initial reports on the oligomeric state of A. tumefaciens VirB4 stated that it assembled minimally as a dimer, although higher-order structures were not discarded (Dang et al., 1999). On the other hand, the VirB4 homologue in the conjugative plasmid pKM101 has been reported to be present both as dimer and hexamer, being the hexameric form soluble and catalytically active and the dimeric form inactive and membrane-associated (Durand et al., 2010). This debate on the oligomeric state of VirB4 seems to have been resolved by the recent structure of the T4SS from R388 (Low et al., 2014), which clearly shows that VirB4 forms hexamers at the base of the secretion channel, as previously suggested (Middleton et al., 2005; Arechaga et al., 2008; Peña et al., 2012). Moreover, this structure shows not only one but two hexamers attached to the inner membrane domain of the T4SS. The implications of this arrangement in the molecular mechanism are however still unclear.

VirB11

VirB11 proteins belong to a large family of hexameric AAA⁺ traffic ATPases, which includes proteins involved in Type II secretion and in Type IV pilus and flagellar biogenesis (Planet et al., 2001). VirB11 proteins consist of two domains, an N-ter (NTD) and a C-ter (CTD), connected by a flexible linker of variable length. The VirB11_NTD has been suggested to interact with the membrane (Yeo et al., 2000), whereas the VirB11_CTD contains the Walker A and Walker B motifs and is thus involved in ATP hydrolysis (Rivas et al., 1997; Ripoll-Rozada et al., 2012). The linker region between both domains has been proposed to play a key role in enzyme catalysis (Hare et al., 2006). In fact, although the crystallographic structures of two non-conjugative homologues, HP0525 from Helicobacter pylori (Yeo et al., 2000; Savvides et al., 2003) and VirB11 from Brucella suis (Hare et al., 2006), revealed similar hexameric rings, the dynamics of hexamer formation and the monomer-monomer interactions are completely different. These differences are produced by a large domain swap of the central linker, which is much larger in the Brucella homologue than in the Helicobacter counterpart (Hare et al., 2006). Based on biochemical and structural evidence, a common model for the mechanism of action of these secretion ATPases has been proposed (Yamagata and Tainer, 2007; Ripoll-Rozada et al., 2012). According to this model, the VirB11_NTD of the nucleotide-free form of the enzyme would be pivoting over the flexible linker. Binding of magnesium and ATP would lock the enzyme in a closed conformation. Upon ATP hydrolysis, a Mg-ADP state would be stabilized and, consequently, at physiological Mg²⁺ concentrations this Mg-ADP state would remain in an inhibited, closed conformation. A specific signal (for instance, substrate binding or release) would unlock this state, resuming the catalytical cycle by releasing the ADP for the next turnover (Ripoll-Rozada et al., 2012).

The biological function of VirB11 is still unknown. VirB11 seems to participate in pilus biogenesis by assisting VirB4 dislocase activity on VirB2 (Kerr and Christie, 2010). Actually, interactions between VirB11 and VirB4 have been observed (Draper et al., 2006; Ripoll-Rozada et al., 2013). These interactions depend on the length of the linker region, as inferred from experiments in which VirB4 was tested for interactions with VirB11 homologues of different linker size (Ripoll-Rozada et al., 2013). These interactions seem to be unstable as judged by the fruitless attempts to purify stable VirB4–VirB11 complexes (J. Ripoll-Rozada, unpublished) and the absence of VirB11 in the T4SS structure (Low et al., 2014).

VirB11 has also been suggested to play an essential role in the first steps of the DNA translocation pathway (Atmakuri et al., 2004b), as shown by transfer DNA immunoprecipitation assays (Cascales and Christie, 2004b). Supporting this dual functionality, mutations in VirB11 uncouple T-pilus production from substrate translocation (Sagulenko et al., 2001). It has been suggested that VirB11 acts as a molecular switch between pilus biogenesis and substrate transport by replacing VirB4 and VirD4 ATPases at the entrance of the secretion channel (Ripoll-Rozada et al., 2013). It is worth noting that VirB11, although widely distributed among conjugative systems, is not found in all T4SS. MPF_F plasmids, for instance, do not code for a cognate VirB11 homologue (Frost et al., 1994; Lawley et al., 2003), suggesting that there are important differences between MPF_T and MPF_F conjugative systems, mainly in the morphology, length and flexibility of the pilus. In DNA uptake T4SS, like H. pylori ComB system, a cognate VirB11 orthologue is also missing (Karnholz et al., 2006). Therefore, in these systems, the function of VirB11 seems to be dispensable or replaceable by other proteins with similar structure.

VirD4

The T4SS coupling protein (T4CP) or VirD4 in A. tumefaciens is essential for substrate translocation but, in contrast to VirB11 and VirB4 ATPases, is dispensable for pilus biogenesis (Lai et al., 2000). The main function of conjugative coupling proteins is to drive the nucleoprotein susbtrate to the secretion channel (Cabezón et al., 1997). The coupling function of T4CP is mediated, on one side, by interactions with the relaxosome auxiliary proteins (Disque-Kochem and Dreiseikelmann, 1997; Tato et al., 2007) and, on the other side, by interactions with the inner membrane domain of the core complex VirB10 protein (Gilmour et al., 2003; Llosa et al., 2003; de Paz et al., 2010).

TrwB is a DNA-dependent ATPase (Tato et al., 2005). The structure of the cytoplasmic, soluble domain of TrwB revealed a hexamer with a six-fold symmetry and a central channel of about 20 Å in diameter (Gomis-Ruth et al., 2001). The protein presents structural similarity to hexameric molecular motors such as F₁-ATPase, FtsK or ring helicases (Iyer et al., 2004), suggesting that TrwB also operates as a motor, using energy released from ATP hydrolysis to pump ssDNA through its central channel (Cabezon and de la Cruz, 2006; Gomis-Ruth and Coll, 2006). At the N-terminal domain of TrwB, there are two transmembrane α-helices which anchor the protein to the inner membrane. This transmembrane domain also regulates nucleotide binding affinity (Vecino et al., 2010), and it is important in stabilizing the structure of the protein (Vecino et al., 2011).

TrwB ATPase activity is enhanced by TrwA (Tato et al., 2007), an accessory protein, which together with the DNA and TrwC forms the relaxosome of R388 plasmid. TrwA binds specifically at the origin of transfer of the plasmid (oriT) as a tetramer and facilitates TrwB assembly on DNA (Moncalian and de la Cruz, 2004). The C-terminal domain of TrwA interacts with TrwB, whereas its N-terminal domain is involved in DNA binding. In IncF plasmids, interactions between the coupling protein TraD and the accessory protein TraM have also been reported (Lu and Frost, 2005) and the structure of the C-terminal tail of TraD bound to the TraM tetramerization domain has been solved (Lu et al., 2008). Intriguingly, the TrwA–TrwB system resembles the mechanism used by other DNA-dependent molecular motors such as the RuvA-RuvB recombination system (Yamada et al., 2002; Tato et al., 2007). Recently, it has been shown that TrwB is a structure-specific DNA-binding protein, with very high affinity for G4 DNA structures (Matilla et al., 2010). The protein might be involved in resolving G4 secondary structures that arise during conjugative DNA processing. Alternatively, this higher affinity for G-quadruplex might reflect the need of a secondary DNA structure as a loading site for the motor. Further experiments are needed to clarify the in vivo function of G-quadruplex structures in the conjugative process.

SUBSTRATE TRANSPORT

Biochemical and genetic analysis of T4SS components provide evidence supporting a model in which DNA/protein transfer and pilus production are uncoupled and independent processes. A pathway of interactions between the T-DNA substrate and the different T4SS subunits during DNA transfer has been proposed (Cascales and Christie, 2004b). However, despite the mounting evidence describing the interactions between T4SS components and the interactions between them and the translocated substrate, there are still a number of questions that remain open. For instance, the involvement of the core complex subunits during substrate transport, the specific role of the three ATPases or the fate of the pilus after the formation of the mating pair are still not fully understood processes. Last but not least, an important question that remains unanswered is the conformational state, native or unfolded, of the protein substrate during translocation. Here, based on the information available up to date, a detailed description of the events that take place during DNA processing and transport in bacterial conjugation is provided (Fig. 4).

Step 1. DNA recognition and relaxase cleavage reaction

Conjugative and mobilizable plasmids contain a cognate DNA sequence, oriT, which is recognized by specific auxiliary factors within the relaxosome. The relaxosome, called like this by analogy with the nucleosome, is formed by the relaxase and auxiliary transfer proteins bound to a fragment of DNA several hundred base pairs in length, which contains the origin of transfer (oriT) (Fürste et al., 1989).

Relaxases are specific phosphodiesterases which cleave DNA at a nic site within the oriT (Byrd and Matson, 1997; Chandler et al., 2013). Relaxases are usually, but not always, large multi-domain proteins (for a recent review see Garcillan-Barcia et al., 2009). In all cases, the relaxase domain, which is approximately 300 aa long, is located at the N-terminus of the protein. At the C-terminus, a DNA helicase or a DNA primase can be found. Occasionally, an extra domain of unknown function is also observed. For instance, TraI, the relaxase of plasmid F, is a 182 kDa protein (Abdel-Monem et al., 1983) with three distinct domains (Cheng et al., 2011): the relaxase (1–306 aa), the helicase (309–1504) (Byrd et al., 2002) and a third domain, comprising the last 252 aa, which is involved in DNA binding (Guogas et al., 2009). In most cases, relaxases contain a conspicuous signature consisting of a histidine triad (3H motif) (Garcillan-Barcia et al., 2009), which is used by the protein to bind divalent cations. This metal ion is essential for the relaxase cleavage reaction (Boer et al., 2006) but not for DNA binding (Lucas et al., 2010). An analogous histidine motif is also found in nucleases involved in rolling circle DNA replication (RCR) (Koonin and Ilyina, 1993; Chandler et al., 2013), suggesting that the generation of a single-stranded DNA copy of the plasmid occurs by a similar RCR mechanism.

Crystal structures of the relaxase domains of TrwC of plasmid R388 (Guasch et al., 2003; Boer et al., 2006), TraI of plasmid F (Datta et al., 2003; Larkin et al., 2005), MobA (Monzingo et al., 2007) and the relaxase of Salmonella typhimurium pCU1 plasmid (Nash et al., 2011) have been determined, providing details of the cleavage reaction and ssDNA-protein interactions. The cleavage reaction occurs via a nucleophilic attack by the hydroxyl group of the relaxase catalytic tyrosine residue on the 5′-side of the DNA phosphate. This transesterification reaction results in a covalent linkage between protein and DNA (Guasch et al., 2003; Gonzalez-Perez et al., 2007). After the cleavage reaction, donor DNA synthesis begins from the 3′-end of the cleaved strand. Then, the helicase domain displaces the old DNA strand, which is replaced by the new synthesized DNA strand, and the covalently bound nucleoprotein is transferred to the recipient cell.

The auxiliary proteins of the relaxosome are involved in the regulation of gene expression. For instance, TraY and TraM in IncF plasmids bind to the F tra operon promoters, which results either in a positive or negative regulation of gene expression (Schwab et al., 1993; Penfold et al., 1996; Silverman and Sholl, 1996; Taki et al., 1998). TraY and its orthologues, such as TrwA in R388, are structurally related to DNA-binding proteins with ribbon-helix-helix (RHH) motifs. Proteins with this RHH motif interact in antiparallel fashion with the major groove of the DNA (Bowie and Sauer, 1990). An additional auxiliary factor IHF (integration host factor) binds to the DNA inducing a bending in the oriT region (Rice et al., 1996; Seong et al., 2002) that helps to bring TraY to the nic site. The auxiliary proteins are not absolutely required for linear DNA cleavage by MOB_F relaxases, but they seem to be essential for nicking activity on supercoiled DNA (Garcillan-Barcia et al., 2009). The role of the different auxiliary proteins and the mechanism of cleavage and processing of conjugative DNA is described in further detail in (de la Cruz et al., 2010).

Step 2. DNA driving to the membrane

Once the cleavage reaction has ended, the nucleoprotein complex must be recruited to the membrane channel to initiate transfer (Llosa et al., 2002). This step is mediated by the coupling protein (Cabezón et al., 1997), which binds to the relaxosome with the assistance of the auxiliary proteins. Binding of the coupling protein to the DNA is sequence unspecific but, thanks to the auxiliary proteins that recognize a specific sequence in the oriT (Moncalian and de la Cruz, 2004), the coupling protein is able to bind to an specific region within the relaxosome. As mentioned above, TrwB has been found to bind G4-quadruplex DNA with much higher affinity than any other type of DNA (Matilla et al., 2010). Therefore, it is possible that DNA G-quadruplex structures can also act as loading sites for TrwB.

Step 3a. Relaxase unfolding

T4SS-dependent transfer of relaxases to human (Schulein et al., 2005), plant (Vergunst et al., 2000) and bacterial (Draper et al., 2005) recipient cells has been documented. TrwC has been shown to display site-specific and recombinase activities in the recipient cell (Draper et al., 2005). However, the mechanism of transport across the T4SS channel is unknown. Given the large size of most relaxases (over 100 kDa), it is unlikely that they can go through the T4SS channel in a native state conformation and an unfolding mechanism should be envisaged. Nothing is known about this unfolding mechanism but it would involve the assistance of some kind of unfoldase. In this regard, it is worth noting the proposed role of the secretion ATPase superfamily proteins as traffic ATPases (Planet et al., 2001). The overall structure of the traffic ATPase VirB11 (Yeo et al., 2000; Hare et al., 2006) is highly conserved among distantly related ATPases associated to other transport systems (Peña and Arechaga, 2013). T3SS also have an AAA+ hexameric ATPase, InvC, which is a chaperone involved in the release of the unfolded substrate to the secretion channel in a process coupled to ATP hydrolysis (Akeda and Galan, 2005). Interestingly, and similarly to TrwD (Rivas et al., 1997), InvC fractionates both as cytoplasmic and as a peripheral inner membrane protein. Therefore, a tempting hypothesis is that TrwD/VirB11 binds and unfolds the relaxase in a similar manner to that of InvC and then it delivers the unfolded substrate to the inner membrane platform at the base of the T4SS secretion channel. However, experimental evidence of interactions between VirB11 and the relaxase is still missing. The putative role of VirB11 in substrate transport as an acceptor of the nucleoprotein complex from the coupling protein is supported by several lines of evidence, such as the T-DNA immunoprecipitation (TrIP) assays carried out in A. tumefaciens (Cascales and Christie, 2004).

Step 3b. Relaxase shooting

Once the nucleoprotein complex has been recruited to the T4SS by the coupling protein and transferred to VirB11, the substrate is delivered to VirB6 and VirB8 proteins in a process mediated by VirB4 (Cascales and Christie, 2004). However, the mechanism by which the relaxase is transported across the secretion channel is still unclear. In the VirB system of Brucella, this type of signal sequence has been identified in several effectors (de Jong et al., 2008; de Barsy et al., 2011; Marchesini et al., 2011). This signal has been localized in some cases at the N-termini, whereas in other cases this signal was located at the C-termini of the translocated protein (de Jong et al., 2008). In contrast, relaxase translocation signals (TSs) in conjugative T4SS have been localized to internal positions. The study on the relaxase TraI of R1 and F plasmids (Lang et al., 2010), identified two internal TSs (TSA and TSB), with conserved residues that defined a consensus sequence (G[E/D]R[L/M]R[V/F]T). The three-dimensional structure of TSA has recently been reported (Redzej et al., 2013). Based on conservation of the consensus sequence within a RecD2-like domain of the proteins, an extension of these results to other relaxases was proposed (Lang et al., 2010). Accordingly, two TSs (TS1 and TS2) in the relaxase TrwC of plasmid R388 were also identified (Alperi et al., 2013). Mutations within TS1 region almost abolished conjugation by the T4SS of plasmid R388, whereas mutations in both TSs barely altered DNA transfer by the T4SS of B. henselae. These results seem to indicate that a single substrate can be recruited by two different T4SSs through different signals (Alperi et al., 2013).

Step 4. DNA passage across the secretion channel

According to a shoot and pump model (Llosa et al., 2002), once the relaxase, covalently bound to the DNA, is shot through the channel, TrwB wraps on the DNA as a hexamer and uses energy derived from ATP hydrolysis to pump the DNA. The proposed mechanism (Cabezon and de la Cruz, 2006) is a modified version of the so-called binding change mechanism proposed for the F₁-ATPase, which involves a sequential binding and hydrolysis of ATP (Boyer, 1997). After the transfer of a complete copy of a single DNA strand into the recipient cell, the relaxase would recognize the nic site and it would carry out a second transesterification reaction, resulting in re-circularization of the transferred strand in the recipient (Draper et al., 2005).

One important question that remains unanswered is the extent of the conformational changes that take place in the core complex channel. It is evident that these structural conformational changes should be large enough to permit the nucleoprotein passage across the T4SS. These dramatic conformational transitions are likely to require energy, which might be provided by one of the two ATPases at the entrance of the secretion channel. Moreover, by protease susceptibility assays, it has been shown that VirB10 undergoes a structural transition in response to ATP utilization by the VirD4 and VirB11 ATPases. Such an energy-induced conformational change in VirB10 is required to form the complex with the outer membrane-associated proteins VirB9 and VirB7 (Cascales and Christie, 2004a). The mechanism resembles that of the TonB-dependent transporters, where TonB also acts as an energy sensor (Postle and Kadner, 2003). VirB10 shares structural similarities with TonB-like proteins (Cascales and Christie, 2004a). TonB is also sequence and structurally related to the T2SS protein GspB. TonB and GspB act as energy sensors of the proton motive force (pmf), and in some cases, like in Klebsiella oxytoca, this is the only energy source required (Possot et al., 1997). However, in other T2SS, ATP hydrolysis is required in addition to the pmf (Letellier et al., 1997). The pmf has also been postulated to be the only energy source required for substrate secretion by T3SS (Wilharm et al., 2004; Paul et al., 2008), although ATP hydrolysis is essential for substrate unfolding (Akeda and Galan, 2005). Hence, the translocation of substrates across the T4SS might also require both the pmf and ATP hydrolysis as energy sources.

Regardless of the source of energy, it has become evident that VirD4 interacts with VirB10 during substrate transport (Cascales et al., 2013). The interactions between these two proteins would only be possible if the resident VirB4 protein, involved in pilus biogenesis, is displaced from the base of the core channel. It has been suggested that VirB4 and VirD4 could be interchanged on this location in the conjugative systems (Peña et al., 2012) in a process mediated by VirB11 (Ripoll-Rozada et al., 2013). Interactions between VirB4 and VirD4 have been reported (Li et al., 2012; Peña et al., 2012). Moreover, it has been reported that VirB4/TrwK is able to inhibit the DNA-dependent ATPase activity by TrwB (Peña et al., 2012), further supporting the hypothesis of the formation of transient TrwB–TrwK heterocomplexes at the base of the secretion channel.

FUTURE WORK

Although much progress has been made in describing the molecular architecture of the T4SS, these efforts have not been accompanied by a better understanding of the dynamics of the secretion mechanism. In DNA transfer systems, many essential questions remain unsolved, such as the mechanism by which the nucleoprotein complex enters in the recipient cell, overcoming the membrane barrier, or the mechanism by which entry exclusion proteins prevents the entry of a second identical plasmid in the cell (Garcillan-Barcia and de la Cruz, 2008). Little is known about specific receptors in the recipient cell that allow pilus attachment. Interactions between the pilus tip and lipopolysaccharides (LPS) in the recipient cells have been reported (Anthony et al., 1994; Gyohda et al., 2004), and modifications of the cell wall LPS were found to affect efficient mating ability (Watanabe et al., 1970). However, recent data have shown that mutations in the LPS pathway had little effect on conjugation and that no nonessential recipient E. coli genes played an essential role in conjugation. These results suggest that conjugation would take place with little regard to the recipient cell constitution (Perez-Mendoza and de la Cruz, 2009).

Another important question that remains unsolved is the exact role of pili in DNA transfer. Although the most universally accepted idea is that pili only provide cell-to-cell contacts, the possibility that ssDNA travels through the lumen of the pili is still debatable. F-pilus, for instance, presents a 20 Å central lumen (Marvin and Folkhard, 1986), which might well accommodate ssDNA. However, it is difficult to envisage the passage of the protein substrate through the pilus. It could be also possible, especially in the short-rigid pili encoded by IncP and IncW plasmids, that after engagement of the recipient cell, pili depolymerize facilitating the membrane contacts between donor and recipient cells, leading to membrane fusion. At that stage, T4SS could easily deliver its substrates in the interior of the recipient cell. This membrane fusion hypothesis would also explain the mechanism of interaction between the entry exclusion factor in the recipient cell and T4SS components in the donor cell (Garcillan-Barcia and de la Cruz, 2008). Alternatively, a mechanism similar to the T3SS injectisome, in which the pilus works as a needle, shooting up the virulence factors directly into the recipient cell, is also feasible (Galan and Wolf-Watz, 2006).

In summary, there are still challenging questions to be solved to understand the intimate mechanisms of DNA and protein transport in bacterial conjugation. Understanding these mechanisms is essential to develop strategies in the battle against the dissemination of antibiotic resistance genes. Additionally, a better understanding of the conjugative machinery could be helpful in the development of farfetched biotechnological applications, such as gene therapy (Llosa and de la Cruz, 2005; Schroder et al., 2011; Llosa et al., 2012) or plant transformation (Tzfira and Citovsky, 2006).

This work was supported by the Spanish Ministerio de Economía y Competitividad (MINECO) grants BFU2011-22874 (to E. C. and I. A.) and BFU2011-26608, EU VII Framework Program projects 282004/FP7-HEALTH-2011-2.3.1-2 and 612146/ICT-2013-10 (to FdlC).

Conflict of interest statement. None declared.

REFERENCES